The Developmental Basis for the Evolution of Muroid Dentition:

Analysis of gene expression patterns and tooth morphogenesisin the mouse and sibling vole

Soile V. E. Keränen

Research Program in Developmental Biology,Institute of Biotechnology,

University of Helsinki

andHelsinki Graduate School in Biotechnology and Molecular Biology

andDepartment of Biosciences, Division of Genetics

Faculty of ScienceUniversity of Helsinki, Finland

Academic Dissertation

To be presented for public criticism, with the permission of theFaculty of Science, University of Helsinki, in the auditorium 1041

at Viikki Biocenter, Viikinkaari 5, Helsinki, on May 31st 2000at 12 o’clock noon

Supervised by:

Docent Jukka Jernvall,University of Helsinki, Finland

andProfessor Irma Thesleff

University of Helsinki

Reviewed by:

Dr. Lars Werdelin, Senior Curator,Swedish Museum of Natural History, Sweden

andDocent Kirsi Sainio,

University of Helsinki, Finland

Opponent:

Professor Kenneth M. Weiss,The Pennsylvania State University, U.S.A.

Helsinki 2000

ISBN-951-45-9410-X (PDF version)

3

Vanhemmilleni,Ilman heidän kannustustaan ja kiinnostustaan

tuskin oltaisiin tässä

“Nothing in biology makes sense except in light of evolution.”T. Dobzhansky

5

Contents

Abbreviations 7Abstract 9Original publications 10Introduction 11Literature review 12

The problem of morphological evolution 12Developmental biology in evolution research 13

Morphogenesis as a target for evolutionary processes 14Morphogenesis is based on local differences in intercellular signaling 15Lateral inhibition and dissociation in pattern formation 17Serial homology and diversification 21The tooth as a model system 22

The evolution of the mammalian dentition 25The origin of teeth 25The integration of occlusion in mammals 25The evolution of molar cusp patterns 27The evolution of specialized rodent dentition 28The evolution of muroid teeth 29The lower first molars of the mouse and sibling vole 31

The developmental biology of the teeth 32Morphogenesis 32Determination of positional identity and tooth morphogenesis 34Molecular interactions in tooth morphogenesis 36Signaling centers in tooth development 40

Aims 42Results and discussion 43

The morphogenesis of mouse and vole teeth 43The dentitions of the mouse and vole 43Rudimentary tooth germs in mouse and sibling vole upper diastema regions 44Morphogenesis of mouse and sibling vole first lower molars (M1s) 45

The basic developmental pathways are conserved in the teeth of the mouse and sibling vole 50The gene sequences are conserved between the mouse and vole 50The genes are expressed similarly in mouse and vole molars 52Cross-species recombinations of mouse and vole tissues can produce a tooth 53

Epithelial signaling centers 54Enamel knots as epithelial signaling centers 55Early epithelial signaling centers 55Conservation of signaling center programs 56

Early epithelium and dental formulas 57Individual tooth germs develop from a continuous dental lamina 58Early epithelium controls the positional identities of the tooth germs 58Positional identity and tooth morphogenesis 60Evolution of muroid dental formulas 60

Enamel knots and crown morphogenesis 62The development of individual enamel knots 63Molecular and mechanical models for enamel knot induction and function 65Enamel knots and cusp patterns 68

6

The evolution of mouse and sibling vole molar cusp patterns 69Crown size and cusp numbers 70Orthogonal and diagonal cusp positions 71The evolution of prismatism and hypselodonty in Microtinae 71

Concluding remarks 72Acknowledgements 75Materials and methods 76

Tissues of the mouse and sibling vole 76Probes for in situ hybridization 76In situ hybridization of sections 77Three-dimensional analysis of epithelial expression patterns 77Wholemount in situ hybridization 78Apoptosis detection by TUNEL-staining 78In vitro tissue culture 78PCR-genotyping 79

Appendix 1 80References 81

7

Abbreviations

AER apical ectodermal ridgeAP Alkaline phosphataseAS AntisenseBarx BarH-like homeobox (homeobox-transcription factor)BBR Blocking reagentBCIP x-phosphate / 5-Bromo-4-chloro-3-indolyl-phosphateBmp Bone morphogenetic protein (peptide growth factor)BSA bovine serum albuminC CaninecDNA Complemetrary DNACHAPS 3-[(3-Cholamidopropyl)-dimethylammonium]-propane-sulfonatec-met Cellular MNNG-HOS transforming gene (receptor tyrosine kinase)c-src Cellular sarcoma gene (protein tyrosine kinase)dATP 2’-deoxyadenosine 5’-trisphosphateDEM Digital elevation modelDEPC Diethyl pyrocarbonateDlx Distal-less like homeobox (homeobox-transcription factor)DNA Deoxyribonucleic acidDNAse Deoxyribonuclease enzymedNTP 2’-deoxyribonucleoside 5’-trisphosphate (mixture)Dpp Decapentaplegic (peptide growth factor)dUTP 2’-deoxyuridine 5’-trisphosphateE10 embryonic day 10ECM extracellular matrixEDA Ectodermal dysplasia A –protein (tumor necrosis factor –family member)EDTA Ethylenedinitrotrilotetraacetic acidEgf Epidermal growth factor (peptide growth factor)Eve Even-skipped (homeobox-transcription factor)FAB Fragment antibody bindingFgf Fibroblast growth factor (peptide growth factor)GIS Geographic information systemGli Glioblastoma gene (Zn2+-finger-transcription factor)Hgf Hepatocyte growth factor (peptide growth factor)I IncisorI1/1 one upper and one lower incisorLef Lymphoid enhancer factor (high mobility group-transcription factor)M. MicrotusM1 first molarM1 first lower molarM2 second molarM3 third molarM3 third upper molarM3/3 three upper and three lower molarsMABT maleic acid buffer with Tween 20M-CSF Macrophage colony stimulating factor (cytokine)MQ Milli Q –filtered (water)

8

mRNA messenger RNAMsx Muscle segment homeobox gene (homeobox-transcription factor)NMT NaCl-MgCl2-Tris-bufferNTB 4-nitroblue tetrazolium chlorideNTE NaCl-Tris-EDTA-bufferP PremolarP2/1 two upper and one lower premolarp21CIP1/WAF1 21 kD cyclin dependent kinase interacting protein, wild-type p53 activating factor 1

(cyclin dependent kinase inhibitor)p53 53 kD DNA-binding tumor suppressor proteinPax Paired-like homeobox (homeobox-transcription factor)PBS phosphate buffered salinePCR polymerase chain reactionPdgf Platelet derived growth factor (peptide growth factor)Pdgfr-α Platelet derived growth factor receptor α (receptor tyrosine kinase)PFA ParaformaldehydePtc Patched (twelve transmembrane repressor subunit in Hh-receptor complex)Pitx2 Pituitary homeobox 2 (homeobox transcription factor)RIEG Rieger syndrome gene (=Pitx2)RNA Ribonucleic acidRNAse Ribonuclease enzymeRT Room temperaturerUTP Uridine 5’-trisphosphate35S-rUTP Uridine 5’-[α-35S]thiotrisphospateSDS Sodium dodecyl sulfateShh Sonic hedgehog (signaling molecule)Smo Smoothened (seven transmembrane signaling subunit in Hh-receptor complex)SP substance PSSC NaCl-sodium citrate bufferRT-PCR reverse transcription polymerase chain reactionTBT Tris-buffer with TritonX-100TCF T-cell specific transcription factorTdT terminal deoxytransferaseTGFβ Transforming growth factor β (peptide growth factor)Tris-HCl Tris(hydroxymethyl)aminomethane buffer with hydrochloric acidtRNA transfer RNAWnt Wnt-family member (signaling molecule)ZPA zone of polarizing activity

9

Abstract

One of the main problems in morphological evolution is how the conserved developmental processesproduce disparate morphologies. The highly disparate dentitions of the mouse (Mus musculus) andsibling vole (Microtus rossiaemeridionalis) were chosen as a suitable model for analyzing thedevelopmental basis of morphological evolution.

Developmentally, teeth are epithelial appendages and their development is regulated by epithelial-mesenchymal interactions. The numbers, positions and shapes of individual teeth are strictly andheritably controlled. The molecular basis of tooth development has been extensively studied in themouse, which is the most important mammalian model in genetics and developmental biology. Thesibling vole is so closely related to the mouse that the number of changes in developmental processesnot directly related to morphological differences is minimal.

Histological analysis of serial sections revealed that morphogenesis of mouse and sibling voleembryonic tooth germs progressed via similar developmental stages. However, the shapes of theirlower first molars developed in different directions from the beginning. Moreover, the developmentof three rudimentary tooth germs found in the mouse upper diastema region was arrested at an earlybud stage, whereas the single rudimentary tooth germ found in the sibling vole upper diastema regionwas arrested at a late bud stage.

In situ hybridization comparison of expression patterns of developmental regulatory genes Bmp2,Bmp4, Fgf4, Fgf8, Lef1, Msx1, Msx2, p21CIP1/WAF1, Pax9, Pitx2, Shh and Wnt10a, and recombinationof mouse and sibling vole embryonic dental tissues revealed that tooth developmental processes areconserved between the species. The expression patterns and recombinations also strongly suggestedthat the dental formulas, positional identities and morphogenesis of individual teeth are controlled byearly epithelial signals. However, the recombinations and the morphometric analysis of digitalelevation models (DEMs) of developing mouse and sibling vole M1 crowns revealed some of theroles for the epithelial-mesenchymal interactions during later morphogenesis.

Based on in situ hybridization analysis, three consecutive epithelial signaling centers that expresssimilar signaling molecules were identified. These signaling centers appear to be involved in thedevelopment of dental formulas, the bud to cap stage transition and the development of cusp patterns.All tooth germs, including the rudimentary ones, have at least one signaling center. The earlyepithelial signaling center was previously unknown, but the later signaling centers have been knownas the primary and the secondary enamel knots.

By mapping with a GIS-program the expression patterns of selected enamel knot marker genes intoDEMs of developing crowns, the development of the enamel knots was shown to predict the crownmorphogenesis. In particular, the development of the primary enamel knot could be used forpredicting species-specific cusp patterns. Each individual cusp has its own enamel knot. Small timeinterval comparison of gene expression in isolated molar epithelia showed that these enamel knotsdevelop separately. The numbers, shapes and sizes of cusps appear to depend on the length of thetime period when new secondary enamel knots can be initiated and on the crown growth rates. Thesemechanisms can also be used to explain the tooth morphologies of extinct muroid species. Hence, itis suggested that the molecular interactions controlling the initiation of the signaling centers and thecrown growth are important targets for morphological evolution.

10

Original publications

The thesis is based on the following original articles, which are referred to in the text by their Romannumerals, and on some unpublished results.

I Vaahtokari A. Åberg T. Jernvall J. Keränen S. Thesleff I. (1996): The enamel knot as asignaling center in the developing mouse tooth. Mechanisms of Development. 54(1):39-43.

II Keränen S.V.E. Åberg T. Kettunen P. Thesleff I. Jernvall J. (1998): Association ofdevelopmental regulatory genes with the development of different molar tooth shapes in twospecies of rodents. Development Genes & Evolution. 208(9):477-86.

III Keränen S.V.E. Kettunen P. Åberg T. Thesleff I. Jernvall J. (1999): Gene expression patternsassociated with suppression of odontogenesis in mouse and vole diastema regions. DevelopmentGenes & Evolution. 209(8):495-506.

IV Jernvall J., Keränen S.V.E., Thesleff I. (2000): Developmental basis of evolutionary change inmammalian molar topography. (Submitted)

11

Introduction

Morphological evolution results from heritablechanges in morphogenesis. The use of variouswell known organisms, so called model organ-isms, such as the fruit fly and mouse, hasallowed us to dissect the molecular basis forsignaling interactions and cellular differentiationin animal development, i.e., to link moleculeswith morphogenesis. During the last decade,developmental genetics has shown that differentanimal phyla not only share a common cellularmetabolism, but that also their intercellularcommunication networks are conserved provid-ing a common language for morphological evo-lution. Even the developmental functions of thegenes can be conserved between animals as dif-ferent as flies, nematodes and birds (Gerhard andKirschner 1997). However, the fast rate of mor-phological evolution between closely relatedspecies suggests that even small genetic changescan have large morphological effects, whichseemingly contradicts the deep conservation ofdevelopmental programs. Thus, the extensiveconservation of developmental processes hasmade the origin of species specific morphologi-cal differences perhaps the most bafflingquestion in evolutionary developmental biology.

To understand the developmental basis of mor-phological evolution, we must understand thedevelopmental basis for small, species specificmorphological differences. During morphogene-sis, undifferentiated clusters of cells divide anddifferentiate into different, strictly arranged celltypes. The increasing complexity and the localgrowth differences of the tissues, which are nec-essary for the development of specific mor-phologies, is widely considered to arise fromlocal differences in the intercellular signaling(Wolpert et al. 1998). The development of com-plex morphologies requires local differences ingrowth and differentiation. These depend onlocal inductive interactions or on intercellularsignaling controlled by the organism’s genome.Hence, changes in pattern formation are an im-portant mechanism behind morphological evo-lution.

A mammalian tooth is a histologically simpleorgan, which develops from oral epithelium andunderlying mesenchyme into a strictly deter-mined shape. Each species has its dental for-mula, in which the numbers, locations andshapes of individual teeth are heritably deter-mined (e.g., Owen 1840-45, Grüneberg 1965,Berry 1978, Hillson 1986). Early tissue recom-bination experiments between mouse molars andincisors have shown that until the early budstage, tooth development is controlled by theepithelium, and thereafter the control is trans-ferred to the dental mesenchyme (Kollar andMina 1991).

As teeth are mineralized organs, they have agood fossil record showing their morphologicalevolution. Vertebrate teeth are serially homolo-gous. However, in mammals, tooth developmentand initiation have became less dependent onenvironmental factors, and teeth in different po-sitions have differentiated from each other(Butler 1995, Huysseune and Sire 1998).Moreover, also teeth that occupy homologouspositions in the jaw can be morphologically dif-ferent in different mammalian species. Differenttooth shapes can be described using their cusppatterns (Jernvall 1995), a cusp being a convexarea of the tooth crown, separated from possibleother adjacent cusps by valleys with epithelialand mesenchymal contribution (Butler 1956).Both the determination of the dentition and thedetermination of the cusp patterns of individualteeth are examples of pattern formation at differ-ent hierarchical levels.

The morphogenesis and molecular biology oftooth development, especially that of the lowerfirst molars (M1), have been extensively studiedin mouse (Mus musculus), which is the mostimportant mammalian model organism in mo-lecular, cell and developmental biology. A num-ber of cloned mouse genes combined with theadvanced molecular tools makes the functionalanalysis of individual signaling pathways andgenes possible, and together with routine tissue

12

culture techniques make the deterministicallydeveloping mouse tooth an ideal model forstudying the genetic basis for morphogeneticprocesses.

However, the mouse is only one species, and itsdentition, composed of one open-rooted incisor,a diastema lacking teeth and three molars in eachjaw quadrant, is very specialized when comparedto the basic placental mammalian dental for-mula. Hence, it is unclear if the data from mousestudies is directly applicable to other mammals.Moreover, without comparisons to other speciesit is impossible to study the evolution of species-specific morphologies. Therefore, in this study,mouse teeth were compared to teeth of anothermuroid rodent, sibling vole (Microtus rossiae-meridionalis), which is of similar size and ges-

tation period as the mouse, but has a very dispa-rate M1 morphology. The excellent fossil recordof muroids makes it possible to compare thetooth morphogenesis in these two species to theevolution of tooth morphology, in order to ana-lyze the general evolvability of tooth develop-mental processes. The mouse and vole lineagesdiverged from each other about 20 million yearsago (Nikoletopoulos et al. 1992, Robinson et al.1997) and because important developmentalprocesses and genes tend to be conserved, thenull-hypothesis was that the mouse and volegenes, genomes and tooth developmental pro-grams would be quite close to each other. There-fore, the techniques and principles of mousedevelopmental genetics could be easily appliedto studying sibling vole teeth.

Literature review

The problem of morphological evolution

How do species specific morphologies evolve?Today, there exist approximately between 1 and30 million metazoan species with differingmorphologies (mostly insects), of which approx-imately 600 000 are listed and described (Wilson1992). Because morphological evolutiondepends on heritable changes in ontogeny, it isobvious that morphological evolution dependson changes in genes affecting the developmentalprocesses.

The fundamental problem is the apparentdisparity between the rates of DNA sequencedivergence and the rates of the morphologicaldivergence. It is known, that the rate of morpho-logical evolution can change from virtual stasislasting for millions of years (a rate slower than0.1 darwins) to significant morphological changein a few decades (faster than 50 000 darwins)(Martin 1993). This indicates selection drivenevolution until a new local optimal morphologyis reached (Simpson 1944, Futuyma 1986).However, e.g., the number of segments isconstant in different Drosophila species, thoughthe promoter sequences of the genes involved insegmentation genes may have diverged consider-

Tim

e

Neutralevolution

Evolutiondriven by selection

Figure 1. The rate and direction of the neutral evolution andevolution under selection pressures. After a lineage split,neutral changes tend to (on the average) accumulate at aconstant rate determined by mutation frequency, leading toincreasing divergence between lineages, whereas thefeatures under selection pressures may evolve at variablerates, and the direction of evolution may change. Sequenceand morphological evolution are both driven by selectionpressures and chance accumulation of neutral changes, butthe relative importance of neutral change and selection mayvary between morphological and sequence evolution. Thismay, e.g., occasionally explain the disparity betweendivergence times indicated by the fossil record andmolecular clocks. The dashed line indicates the time whenthe lineages split.

ably (Ludwig et al. 1998). Thus, the rate ofmorphological change does not apparentlycorrelate with the rate of sequence divergence.Moreover, most sequence changes are likely to

13

be effectively neutral (see below), suggestingthat on the average, the rate of sequence diver-gence is more likely to resemble the neutral thanthe selection driven model (Figure 1). Althougha part of this seeming disparity between morpho-logical and sequence evolution doubtlesslyresults from biases caused by difficulties inobjective definition and weighing of morpholo-gical features (see, e.g., Raff 1996), the function-al relationship between sequence changes andmorphological changes in general is poorlyunderstood (e.g., Leroi 2000). Hence, to resolvethis fundamental problem, one must understandhow genes are involved in morphogenesis.

Morphogenesis depends on signals to individualcells, which interpret them according to theirgenome and earlier differentiation status. If theonly relevant signals come from the cells’ neigh-bors, morphogenesis is controlled only by theorganisms own genome. If the signals interpret-ed by the cells also include factors from outsidethe organism, such as temperature, small mole-cules, gravity or light, the outcome of morpho-genesis can be variable (Gerhard and Kirschner1997). Even then, the final morphology, orphenotype, of the organism depends on deter-ministic processes controlled by the genome.Thus, morphological evolution depends onchanges in genes involved in intrinsic develop-mental processes.

Ecophysiological demands act on variation pre-sent in the population. Each population containsan enormous amount of morphological variation,which is caused by a combination of heritableand environmental factors. This means that thedevelopmental programs are not direct blueprintsfor morphology, but a set of instructions andconditions coded by genetic information, whichis read in various environmental contexts (e.g.,Dawkins 1986). Thus, the main problem in evo-lutionary developmental biology is to discoverthe basic morphogenetic rules according towhich the variation can exist and evolve, leadingto species specific and even macroevolutionarydifferences (e.g., Arthur 2000, Erwin 2000). Oneway to analyse the morphogenetic rules is tosearch for forbidden variation by mapping the

existing morphologies. Another is to experi-mentally analyse the molecular and physiologic-al mechanisms of morphogenesis. Because expe-rimental data must be repeatable, developmentalvariation must be minimized by studying gene-tically homogeneous model organisms instandardised conditions. For similar reasons, it iseasier to compare the morphogenesis of relatedspecies rather than that within a population.Nevertheless, a combination of experimentalresearch on different model organisms and com-parative analyses of their morphogenesis is apowerful method for analysing the genetic basisof morphological evolution and probably thebest way to bridge the gap between paleontologyand genetics in evolutionary research.

Developmental biology in evolutionary research

The importance of developmental biology inexplaining morphological evolution was alreadyrecognized by Darwin, who discussed it in theOrigin of Species (Darwin 1856). Comparativestudies of organogenesis by 19th century mor-phologists, such as Geoffroy Saint-Hilaire andRichard Owen, became an important part ofevolutionary research (for reviews see Gilbert etal. 1996, Raff 1996). At the turn of the century,organogenesis became more important subject initself, as Roux began experimental embryolog-ical studies and induction was discovered bySpemann and Mangold (for reviews, see Gilbertet al. 1996, Raff 1996). The induction studiesfaltered, because the molecular tools for study-ing its biochemical basis did not then exist. Sub-sequently, developmental biology becamemainly a study of developmental physiology ofindividual organs in different model animals.Because the different model organisms in devel-opmental biology were perceived to be too dif-ferent from each other, the evolutionary aspectsof developmental biology were therefore ignored(Raff 1996). Meanwhile, the geneticists tookover evolutionary biology, as the role of muta-tions and population genetics was integratedwith the morphological and ecological data intothe neo-darwinian modern synthesis (for reviewssee, e.g., Futuyma 1986, Hartl and Clark 1989,Li and Graur 1991). However, with the advances

14

in molecular biological techniques, researchersbegan to learn more about molecular evolution,and the gap between the sequence data and fossildata became more and more pronounced requir-ing new approaches to evolutionary genetics(Raff 1996, Gilbert et al. 1996, Gerhard andKirschner 1997, Akam 1998, Arthur 2000).

That developmental biology could be an import-ant part of evolutionary biology was slowly re-alized when it was discovered that the anterior-posterior body axis identities in animals as dif-ferent as fruitfly and mouse were controlled bythe same genes (for reviews, see, e.g., Manakand Scott 1994, Holland and Garcia-Fernàndez1996, Averof 1997). Lewis, Nüsslein-Volhardtand Wieschaus received the 1995 Nobel Prize inMedicine or Physiology for their study of home-otic mutations in the fruitfly, Drosophila mela-nogaster (e.g., Roos 1995). After the late 1980’sand early 1990’s, the combination of the newestmolecular biological methods and a long tradi-tion in morphological and genetic studies in tra-ditional model organisms of developmental biol-ogy, such as the mouse, the fruitfly, the nema-tode, the African clawed toad, the sea urchin andthe chicken, has finally produced results, andtoday the amount of data about the develop-mental basis of evolution is rapidly increasing.We now know that practically all families andpathways of the signaling molecules found in themouse are found also in fly or the nematode(Gerhard and Kirschner 1997), and we also

know that even the developmental functions ofthe molecules may be conserved (Gerhard andKirschner 1997), which provides a common lan-guage for morphological evolution in all Meta-zoan animals.

The concept of similar developmental mechan-isms in all eumetazoans (and maybe even inplants) is of utmost importance, because itallows a common frame of reference in studiesof the genetic basis of morphological evolution.The renaissance of evolutionary developmentalbiology has also seen the introduction of newmodel systems, because they are either superiorfor genetic studies, as compared to the tradition-al systems (e.g., the mustard plant, zebrafish,Japanese pufferfish), or because their phylogen-etic positions may provide crucial informationon long unresolved evolutionary questions (e.g.,ascidians, amphioxus, cnidarians). Whilst theincreasing numbers of well known model organ-isms greatly facilitate evolutionary research,their distant relationships still limit the kinds ofquestions we can ask. Because the most import-ant model species in developmental biology aredistantly related, studies of developmental evol-ution have been mainly carried out on the simil-arities between the different orders or phyla. Theopposite approach, comparison between a wellknown model organism and its morphologicallydisparate relative, allows us to assess the evolu-tion of morphological divergence.

Morphogenesis as a target for evolutionary processes

Morphological evolution depends on changes inthe strict spatial and temporal control of morpho-genesis. Because processes such as growth andmigration are quantitative, these must be trans-formed into meristic properties like numbers ofvertebrae, and during deterministic developmentthe spatial arrangement of the individual unitsmust also be correct. The spatial and temporalcontrol of metric processes can be accomplishedwith changes in pattern formation , which de-pend on the dynamics of intercellular signaling.During pattern formation spatially simple sig-naling leads hierarchically into an increasingly

complex series of inductions in strictly localizedparts of the differentiating tissue (Figure 2A).

The evolution of morphogenesis does not seemto result from changes in the actual proteinsequences encoded by the genes involved inmorphogenesis. Even only ca. 50% identicalmembers of the same gene family may replaceeach other functionally (e.g., Huang et al. 2000),although occasionally also biochemical innova-tions may occur (e.g., Hanks et al. 1998). Rather,the morphological evolution seems to resultfrom changes in the spatiotemporal control of

15

the gene expression (see below). For example, alocalized misexpression of molecules involvedin patterning of Drosophila limb discs lead toabnormally shaped limbs (e.g., Basler and Struhl1994, Jiang and Struhl 1996, Penton and Hoff-mann 1996).

Because the dynamics of pattern formation thusdepend on exact control of the expression pat-terns of the genes determining it, even smallchanges in this genetic prepattern can lead tochanges in morphogenesis (e.g., Stern 1998).Unfortunately, the interrelationships of the indi-vidual signaling pathways and their effects onmorphogenesis are currently largely unknown.Hence, the question remains what kinds of pat-terning changes lead to morphological evolu-tion? Comparison between the expression pat-terns of the developmental genes in related ani-mals with different morphologies allows therejection of unlikely candidate genes and sig-naling pathways, and makes it possible to under-stand the nature of the actual mutation eventsleading to the disparate morphologies.

Morphogenesis is based on local differences inintercellular signaling

The basic unit of morphogenesis is a cell. Duringorganogenesis, simple groups of undifferentiatedcells co-operate to form complex structures,which consist of several kinds of differentiatedcells in specific positions. Co-operation betweencells is controlled by extracellular signals. Eachcell communicates with its neighbors by secret-ing signaling molecules or extracellular matrix,and by producing cell surface ligands and re-ceptors. All cells receive multiple signals at anyone time, and different combinations of thesesignals lead to various holistic responses, whichdepend on the differentiation state of the cells.The cellular responses to these signals includenot only differentiation or cell cycle changes, butalso new signals to the surrounding cells. Intime, there will be cells of several differentia-tional statuses, which communicate with eachother. Their various life cycle responses, mitosis,quiescence, migration and death, then lead tolocal changes in growth and differentiation,

A B

Figure 2. The increase in complexity during patternformation depends on combinatorial signaling andautoregulation. Combinatorial effects of partiallyoverlapping fields (e.g., areas of local signals) producecomplex patterns of differentiation, which can create newsignal sources (dark gray) in spatially restricted locations,further increasing complexity (A). Autoregulatoryinteractions during branching morphogenesis of the lung.The tip of the lung bud grows towards the Fgf10 expressingmesenchyme, but Shh expressed by the tip of the bud locallydownregulates Fgf10 expression, dividing the mesenchymalarea, and the epithelium begins to grow in two directions.This process can be iterated numerous times to producemillions of alveoli in mammalian lung (B).

which in turn leads to the species-specific mor-phogenesis of an organ. Both the signaling envi-ronment and the previous differentiation deci-sions, according to which the cells respond, de-pend on the reciprocal signaling between cells.Hence, development is, in a sense, canalized.

Secreted signaling molecules are a commonmeans for intercellular communication, becausethey can diffuse beyond the immediate vicinityof the signaling cell, forming gradients that

16

PP

PP

P

ECM

Growth factors

CytoskeletonSecondmessenger

Transcriptionfactors

New signalingmolecules

AA

BC

1

2

B

Figure 3. Intercellular signaling with peptide growth factors. Growth factor dimer binds to a receptor, which forms a dimerand autophosphorylates itself. Activated receptor activates a second messenger, often a kinase, which phosphorylates othermolecules, such as transcription factors. The ECM may affect the availability of signals by, e.g., providing substrate for growthfactor dimerization. The activated transcription factor binds to the promoter region of the target gene and activates or inhibitstranscription. Aside from transcriptional activation, the signaling affects the organization of the cytoskeleton, and the cells’interactions with the ECM or other cells (indicated by continuous vs. dashed lines) (A). Several pathways may sharecomponents or may be required for activating a certain gene. Their effects may also be mutually antagonistic. Thus, becausethe cell recieves signals B and C, gene 1 is upregulated, whereas gene 2 is downregulated (B).

depend on the rate of signal diffusion opposedby the rate of signal degradation (e.g., Collier etal. 1996, Asai et al. 1999). Because gradients ofdifferent signaling strengths or times (created bythe diffusion rate from the source) can createlocal differences between initially similar targetcells (Ericson et al. 1996, Yang et al. 1997,Drossopoulou et al. 2000), simple diffusible orrelayed signals from local sources can be usedfor creating complex patterns (Hammer 1998).In developmental context such pattern formingsignals are called morphogens (Wolpert et al.1998)

Peptide growth factors are typical secreted sig-naling molecules. Growth factor signaling ismediated through receptors, their second mes-sengers and transcription factors (Figure 3A).The receptor is oligomerized and transphospho-rylated when growth factor binds to it, and afterthis there follows a complex signaling cascade,which may involve phosphorylations, dephos-phorylations, proteolytic cleavages, Ca2+ con-centration changes and other processes toonumerous to be discussed here (for referencessee Gerhart and Kirschner 1997, Hunter 2000).

The final target of the signal can be a part of thecell cycle machinery or cytoskeleton or it can bea transcription factor (Gerhart and Kirschner1997), which goes into the nucleus activating orinhibiting gene transcription, depending on thecurrent transcriptional machinery of the cell(e.g., De Sousa et al. 1999). The effects of anextracellular signal can be modulated by otherextracellular signals, which activate antagonistic,synergistic or parallel signaling pathways (Fig-ure 3B), as well as by the set of available recep-tors and intracellular target molecules. The sig-nal transduction machinery of the cell dependson the differentiation status of the target cell.This in turn depends on the previous signals thecell has received. Because the potential respon-ses to intercellular signals include production ofnew signals, the result is reciprocal com-munication between cells, leading to mutualchange in differentiation. Since the signals oftenreach only some of the receptive cells, the recip-rocal signaling interaction can lead to increasingspatial complexity.

Signaling responses are also linked to changes incellular adhesion and architecture, which not

17

only affect the shape or movement of the cells,but also modulate the processing of variousextracellular signals (e.g., Orsulic et al. 1999,Zeller et al. 1999). The transmission of signals tothe nucleus, the cytoplasmic localization of theintracellular signal transduction molecules, theplasmamembrane localization of the receptorsand the intake of the signaling molecules havebeen shown to depend on the cytoskeleton andits movements (e.g., Gumbiner 1998, Strumpfand Volk 1998, Ayscough and Drubin 1998,Oren et al. 1999, Orsulic et al. 1999, Zeller et al.1999; for reviews see Gilbert 1997, Gerhard andKirschner 1997). Since some components of thecytoskeleton, like β-catenin, also function asparts of the signaling cascade itself, the cellularsignaling and morphological responses of thecell seem to be different aspects of the sameprocesses during differentiation (e.g., Gumbiner1998, Zeller et al. 1999).

Cells sharing the same developmental historyform a target field, on which the signals can act.The local sources of signaling molecules can becalled signaling centers, and they often consistof cells which are specialized for a signalingfunction, such as the notochord cells in tetrapodembryos or AER cells in developing vertebratelimbs (Wolpert et al. 1998). The target fields canbe identified by differences in the expressioncombinations of the transcription factors, recep-tors and second messengers involved in signaltransduction, whereas the signaling centersexpress high amounts of signaling molecules.Because cells both transmit and receive signalsat the same time, the difference between asignaling center and a target field depends on thesignaling interaction studied.

As the target fields are often wider than the ac-tual organs formed from them, the final place-ment of these structures depends on the localizedsignal, which only reaches part of a potentiallyreceptive target tissue. For example, only part ofthe competent lateral plate mesoderm normallydevelops into limbs (Vogel et al. 1996, Ohuchiet al. 1997). Likewise, just three of the six com-petent vulval precursor cells in C. elegans formthe vulva (Félix and Sternberg 1997) and only

part of the vertebrate lens placode actuallymakes the lens (Saha et al. 1989). The targetfield may also be smaller than the range of theinducing signal, as is the mammary gland epi-thelium during hormone induced changes (e.g.,Brisken et al. 1998). In such a case, the spatialdifferences are likely to be caused by patterningbased on, e.g., earlier localized signaling orunequal localization of cytoplasmic determinantsduring earlier cell divisions (Gerhart andKirschner 1997, Wolpert et al. 1998, Gilbert1997, Gumbiner 1998). Because previous signal-ing events determine the locations and sizes ofthe receptive target fields, as well as the exactlocations of the signals inducing them, thedevelopment of complex morphologies frominitially simple patterns, i.e., deterministic deve-lopment, is possible. Hence, morphogenesisdepends on a hierarchical cascade of reciprocalinductions. The spatial and temporal correlationsbetween the activities of different signalingcascades and morphogenetic events can there-fore be used to analyze potentially importantsignaling interactions regulating morphogenesis.Comparisons between such correlations in twomorphologically disparate species can, in turn,be used for analyzing the differences in morpho-genesis which produce morphological evolution.

Lateral inhibition and dissociation in patternformation

In intercellular communication, responses to thesignal can either amplify or inhibit it. If onesignaling pathway is combined with another,these can affect each other in an inhibitory orsynergistic manner, or create a totally new res-ponse. Thus, overlapping target fields of differ-ent signals can be used for increasing spatialrefinement of the pattern, e.g., by determiningthe locations of new signaling centers (Figure2A). If the response leads to an amplification ofa local signal, the size of the signaling areaand/or the area of response increases (e.g.,Haramis et al. 1995, Wasserman and Freeman1998). This does not inherently lead to localdifferences between initially similar cells, butcombined with a temporal element can causelocal growth disequalities and spatial variation in

18

the differentiation status of these cells. Hence,time is also an important element in morpho-genesis. It is also common that a signal inducesreceptors and transcription factors, which areessential for transduction of another signal,whereas the target cell's receptivity to the origi-nal signal is downregulated by induction ofinhibitory molecules (e.g., Wasserman andFreeman 1998). Therefore, the individual signal-ing cascades can be joined into one signalingnetwork, which regulates itself.

When acting within a planar surface, autoregul-atory induction-inhibition loops create lateralinhibition , which is an important molecularmechanism for creating spatial patterns withininitially uniform fields (Collier et al. 1996).Lateral inhibition models have been proposed tobe important in the formation of two-dimen-sional pattern like stripes in zebrafish (Asai et al.1999), but they have also been shown to beimportant in the development of three-dimen-sional structures like Drosophila bristles(Simpson 1996, Fisher and Caudy 1998) andfeather patterns (Jung et al. 1998). When com-bined with a temporal element, autoregulatoryinduction-inhibition loops can create complexiterative patterns, as during the branching mor-phogenesis of the lung (Figure 2B) (Bellusci etal. 1997b). Because growth, differentiation andsignal diffusion rates are important for pattern-ing processes, both spatial and temporal aspectsof morphogenesis must be analyzed whenstudying the development and evolution ofspecies-specific morphologies.

As seen above, autoregulatory cascades activ-ated in a suitable target field can initiate the de-velopment of a discrete morphological unit, evena whole organ like a limb or an eye (Halder et al.1995, Vogel et al. 1996, Ohuchi et al. 1997). Byactivating these cascades in an ectopic location,either by making an ectopic signaling center ormisplacing the target field, existing structurescan be relocated, remodeled or duplicated. Thisis called heterotopy (Raff 1996). If one devel-opmental process occurs later or earlier as com-pared to other processes in the ancestral species,or the duration of the process is changed, these

differences are called heterochronies (Futuyma1986, Raff 1996). The dichotomy betweenheterochrony and heterotopy is actually artifi-cial, because heterochronic processes can lead toheterotopy and vice versa. For instance, aheterotopy creating excess units in an iterativeprocess like vertebrate somitogenesis (Palmeirimet al. 1997, Richardson et al. 1998) may becaused by heterochronically extending the periodduring which new units are added or byincreasing the rate of addition, whereas the rateor duration of somitogenesis may be changed by,for example, heterotopic expression of somemolecule in the node.

Heterochrony and heterotopy are only possible ifthe signaling pathways controlling different partsof morphogenesis are still at least partially inde-pendently controlled, despite the autoregulatoryand synergistic interactions. Independent path-ways can be modified separately without affect-ing other parts of the signaling system, whichreduces the probability of lethal mutations.Moreover, the pathways can be recombinedindependently in different organs, which meansthat even mutations in the control mechanisms ofthe most essential developmental regulatorygenes, like Pax6 may affect only one organ,leaving other parts of ontogenesis untouched.This phenomenon is called dissociation, andtogether with the development of integrated sig-naling cascades, it is one of the most importantdeterminants of evolvability (Gerhart andKirschner 1997, Kirschner and Gerhard 1998).

Ectopic activation of organogenetic programsrequires a high degree of autoregulatory integrat-ion between the various signaling cascadesneeded for organ development, so that ectopicexpression of one or few master regulatory genescan activate the program in a new location. Italso requires a degree of dissociation in order toallow the sizes and/or the locations of the targetfield and the topological associations of the sig-naling centers to change independently fromeach other. When only parts of a program orsingle genes are co-opted into a new context,even novel types of organs or tissues can becreated. For example, conodont elements and

19

exoskeletal scales of early jawless fish consistedof similar mineralized tissues, although theseorgans probably were not homologous bydescent (Smith and Coates 1998). Morecommonly, the dissociation of the subprogramswithin the organogenic program allows forflexible evolution of both quantitative andmeristic features in species-specific morpholog-ies of orthologous organs. Thus, e.g., the wingdiscs of the insect third thoracic segment giverise to hind wings in butterflies but halteres inflies, although the segment identities are con-trolled by the same genes in both taxa (Warren etal. 1994, Weatherbee et al. 1999). Moreover, thecomplex species specific spot patterns, which areso prominent especially in many butterflies, arecontrolled by the same signaling cascades as thebasic wing axes and identities, but the axisdetermining programs are not activated ectopi-cally, as the cascades involved have been disso-ciably co-opted for different developmentalfunctions (e.g., Carroll et al. 1994, French 1997,Galant et al. 1998, Keys et al. 1998, Weatherbeeet al. 1999).

Since the developmental programs controllingorganogenesis are often combinations of severalparallel signaling cascades that form a loosesignaling network, and since the genetic controlof morphogenesis can be altered at any level ofthe hierarchical information structure (Figure 4),any basic controlling units of information can becalled modules, because they can be recombinedto produce changes in the cellular contextcontrolling morphogenesis. Originally, a deve-lopmental module meant a developmental pro-gram activated by a (master) regulatory gene in aspecific location, but later it began to mean aconserved signaling cascade used in variousdevelopmental contexts (Raff 1996, Fisher andCaudy 1998). Since the effects of all regulatorygenes depend on the cellular differentiation stateand the biomechanical constraints of the deve-loping tissues, the term module might be betterused in a more evolutionarily meaningful senseas a basic target for morphological evolution,equivalent to Riedl’s standard parts (Raff 1996).

Organism

Organ

(Morphological repeat unit)

Signaling pathway

Gene

Promoter element

{

{}

Serialhomology

Geneticpiratism

Mutations

Selection

Lev

el o

f co

mp

lex

ity

Ev

olu

tio

nar

y p

roce

ss

Figure 4. The hierarchy of developmental information ascompared to the hierarchy of evolutionary change. Disso-ciable processes can be reorganized at any developmentallevel by mutations in control elements of genes involved incellular signaling and differentiation (below), depending onthe level of the patterning process the changes would affect(above). During embryogenesis the patterning progressesfrom determination of the body axis via determination oforgan fields to morphogenesis to individual organs (Gerhardand Kirschner 1997). Different levels of patterning can,however, evolve quite freely without affecting the otherlevels (Raff 1996). Hence, the hierarchy of geneticcomplexity is dissociable from the hierarchy of patterningcomplexity, although the evolutionary processes actsimultaneously on the DNA and organismal levels. Units ofreorganization can be called developmental modules.

Such repeated usage of the same genes andpathways in different organs and developmentalstages is one of the most important features inmetazoan evolution, and it is based on modular-ity of control elements in eukaryote genes. Atypical developmental gene has several differentcontrol modules, various promoter, enhancer andinhibitor elements, which each consist of a com-bination of several transcription factor responseelements. Thus, if one of the modules changes,the function of the gene in other developmentalcontexts does not change. These elements can beorgan or stage specific and they can mutateindependently from each other. For example, thetranscription factor Pax6 is involved in thedevelopment of the eye, central nervous system,pancreas and nasal placode (Quinn et al. 1996,Xu et al. 1999). If the eye specific promoter ismutated, only the eye expression is affected,whilst the other organs are normal. The modularcontrol of individual developmental genes thusincreases the dissociability between variousdevelopmental and cellular processes.

20

The evolution of increasingly diversifiedmetazoan organisms can be accomplished, notonly by increasing the complexity of the controlregions of the developmental genes (Fraiden-raich et al. 1998, Gerhart and Kirschner 1997),but also by increasing the number of develop-mental genes with concomitant divergence offunction (Holland and Garcia-Fernàndez 1996,Stock et al. 1996). Although gene duplicationdoes occur, usually leading to developmentaland biochemical redundancies (e.g., Stock et al.1996, Thomas et al. 1997, Coulier et al. 1997),the former option seems to be favored, becausemodifying only the promoter requires fewermutations than duplicating the gene and modi-fying its promoter (and maybe the codingsequence as well). This explains the many stud-ies which have shown that morphological evolu-tion in metazoans has largely resulted fromreapplication of the same genes and programs innew contexts (for a review see, e.g., Gerhart andKirshner 1997).

Mutagenesis studies and the fossil recordindicate that differences between phyla evolveby accumulation of small changes or micro-mutations, whereas evolution through saltationor macromutations is unlikely (Gould 1983). Thesequence changes are most likely to affect theindividual elements of the promoters of develop-mental genes, which may then alter the express-ion patterns of such a gene, occasionally evenactivating it in a new organ or developmentalstage. However, for instance, comparisonsbetween the controlling regions of Drosophilasegmentation gene Even-skipped of differentDrosophila species have shown that stabilizingselection pressures have ensured extremeconservation of Eve expression areas and theirtranscriptional control despite considerablepromoter sequence divergence, i.e., the effects ofthe many micromutations counter each other(Ludwig et al. 1998). Hence, comparisonsbetween the promoter sequences of thedevelopmental genes of related species are notnecessarily informative about the evolution ofthe gene’s usage during morphogenesis. Toobtain this information, actual expressionpatterns must be analysed.

Whenever the same developmental process iscontrolled by several redundant pathways, theuse of pathways can evolve, although theprocess itself remains conserved. The processitself would be analogous to the kernel processproposed by Brenner (1997), whereas thehypothetical individual refining processes mayvary during evolution. Thus, for example, theroles of Fgf8 and Shh in the development of left-right asymmetry are different in the chicken andmouse, although the program is homologous inboth species (Meyers and Martin 1999,Rodríguez-Esteban et al. 1999). Aside from thepossibility of related genes replacing each otherin the same pathway, the potential functionalredundancies between different, even unrelated,signaling pathways (e.g., different mitogens),together with fast evolution of control elements,means that most of the potential variation isprobably hidden. Moreover, redundancy com-bined to dissociation between the signalingpathways means that disturbances in one cascademay either affect all of them, or that the othercascades can replace the failed one, or that onlyone part of the differentiation process is affected.Therefore, the more closely related the speciesare, the more likely it is that gene expressioncomparisons are will give reliable developmentalinformation, because the morphogenetic funct-ions of the common genes and signalingpathways have had less time to diverge.

Although the repeated use of the genes meansthat the signaling interactions in various organsand developmental stages are likely to beconserved, their dissociability means that theseprograms can be flexibly modified. Becausestabilizing selection can keep morphogenesisunchanged, it is not known which gene express-ion changes are important for morphologicalevolution and which are simple evolutionarynoise, or unimportant for morphogenesis. Hence,it is important to compare the expressionpatterns of selected key genes in several signal-ing cascades known to be important for organo-genesis. The correlation between developingmorphologies and gene expression patternsduring morphogenesis can then be used toidentify the potential morphogenetic signaling

21

modules for later experimental verification oftheir role. These correlations are comparedbetween species with disparate morphologies,where the similarities between the pattern form-ation processes indicate the degree and mode ofconservation between the developmental pro-grams of the species and, inversely, the potentialmodes of morphological evolution.

Serial homology and diversification

Because the same genes can be borrowed forvarious purposes in new organs, genetic andhistorical homology are separate concepts.Historically homologous organs, e.g., forelimbsof various vertebrate species, are homologous bydescent. When the same programs are applied ina new place within an individual to generate anew structure, the processes controlling theirdevelopment are homologous, although thestructures themselves lack historical homology(Gilbert et al. 1996; however, see Bolker andRaff 1996). In nature, the evolutionary reductionof individual structures may be associated withloss of some necessary inductive signal, whereasnew structures may result from ectopic activat-ion of signaling. Serial homology, in which newduplicates of pre-existing structures, like addit-ional body segments or extra hairs, are added toor removed from a set of earlier, similar structur-es, is a special case of genetic homology, leadingto a special case of numerical heterotopy. Forexample, the increase of snake vertebral numberand homogeneity has been proposed to havebeen caused by an increase in somite numberscoupled to relaxed requirements for theirpositional identities, (Cohn and Tickle 1999),whereas the changes in the bristle patterns in theDrosophilids seem to have depended on changesin the increasingly complex spatial control ofcommon bristle forming program (Simpson1996, Simpson et al. 1999)

Serial homology has its analogy on a molecularlevel. When a single gene is duplicated within alineage, the result is two paralogous genes, ordifferent members of a gene family. Forexample, vertebrate HoxA1 and HoxA2 ormouse Hoxa1 and Hoxc1 are paralogous genes.

X A B

A' B'

A'' B''

1

2

3

Figure 5. Genes A, A’ and A’’ are paralogous to genes B, B’and B’’, because they have arisen by duplication of theancestral gene X into A and B within the genome of species1. A’ is an orthologue of A’’ and B’ is an orthologue of B’’,because A’ and A’’ have both evolved from A and B’ andB’’ have both evolved from B when species 1 evolved intospecies 2 and species 3.

Copies of the same gene in different species arecalled orthologous genes (Figure 5).Orthologous genes are homologous by descent,whereas paralogous genes, which can even existwithin the same genome are serially homo-logous. Mouse Hoxa1 and human HOXA1 areorthologous genes. The Hox-genes belong to thehomeobox-transcription factor superfamily,which contains also numerous other genes,which have arisen by gene duplication andsubsequent divergence (e.g., Manak and Scott1994, Holland and Garcia-Fernàndez 1996, Bha-rathan et al. 1997, Meyer and Málaga-Trillo1999). Aside from the evolution of genefamilies, the terms paralogous and orthologouscan also be applied to the evolution of seriallyhomologous organs, like vertebrae. In such acase, the term paralogous would mean differentcopies of the similar organs in the same animal,whereas the orthologous organs would be thecopies homologous by descent in differentanimals. Thus, the atlas vertebrae of the mouseand human would be paralogous whereascervical and thoracic vertebrae of the mousewould be paralogous.

Because random mutations will soon eliminatefully redundant copies of the same gene givingrise to pseudogenes (e.g., Li and Graur 1991),gene paralogues usually survive by specializingfor different functions. Because gene duplicat-ions are rarer events than lineage splits, ortho-logous genes are usually more similar to eachother than to their paralogues (e.g., Coulier et al.1997). Nevertheless, since the sequence of theprotein is often less important for its function

22

than the control of its expression pattern, geneduplication increases the evolutionary dissociab-ility of the patterning events controlled bysimilar inductive cascades, whilst also increasingthe potential for redundancies, thus stabilizingthe developmental processes.

The divergence of paralogous organs occupyingdifferent locations depends on the evolution oflocal inductive conditions. In mammals, theincreasing number of paralogous genes allowsthe evolution of different local combinations ofsignals and signaling responses. This, in turn, isprerequisite for the existing morphologicaldiversity of both paralogous and orthologousorgans, as the tissues interpret the coded spatialinformation. Hence, the local combinations ofsignals and responses result in positionalidentity of the organ paralogue. For example,the shapes of the cervical and thoracic vertebrae,which both are serially homologous structures,vary according to the combinations of expressedHox-genes (Gaunt 1994, Horan et al. 1995).Because the paralogous positional identities areusually evolutionarily much older than the splitbetween the compared species, the development-al differences between paralogous organs withinone species may be greater than the differencesin the genetic programs controlling the disparatemorphologies of orthologous organs from relatedspecies.

Nevertheless, even the orthologous organs ofclosely related sibling species can look radicallydifferent. An extreme example of this are thetwo Drosophila species D. silvestris and D.heteroneura, which look quite different, but areso closely related that crosses in both directionsproduce viable and fertile offspring (Bock 1984).This proves that two intercompatible genomescan produce very different outcomes, suggestingthat minor differences in almost completely con-served developmental programs can have vastmorphological (and behavioral) effects. Hence,to analyze the genetic basis for morphologicalevolution, one must compare expression patternsof potential key signaling genes and their targets(as well as their known paralogues) in determi-nistically developing orthologous organs of

closely related, but morphologically differentspecies.

The tooth as a model system

To study the genetic basis for development, it isessential to have a simple, deterministicallydeveloping and easily manipulatable model sys-tem with a well known genome. For the analysisof genetic mechanisms of morphological evolu-tion, this system must exhibit species-specificmorphological differences in orthologous organsand these must be comparable at the molecularlevel between related organisms. A good fossilrecord is also essential for determining theintermediate morphological stages throughwhich evolution has progressed into the com-pared extant forms, so that false hypothesesregarding the mechanisms of morphologicalchange based on morphogenesis of the modelsystems alone can be eliminated.

A mammalian tooth has all the required charac-teristics. A tooth is histologically a simple organ,and its morphogenesis is very deterministic.Moreover, different mammalian species differ inthe numbers, locations and shapes of individualteeth. A mature tooth consists of three mineral-ized tissues, dentine, cementum and enamel,which also makes it the most easily fossilizedvertebrate organ. In fact, many extinct verte-brates (including many muroid rodents) areknown only by their individual teeth or toothfragments, which often have even retained theircorrect individual three dimensional shapes (e.g.,Clemens and Kielan-Jaworowska 1979, Jacobset al. 1989).

Each tooth develops via the same developmentalstages and consists of the same tissues (Figure6), and they are believed to be serially homo-logous. However, as tooth development and ini-tiation became more deterministic during evolu-tion, teeth in different positions differentiatedfrom each other. This enabled morphologicalvariation between the different teeth in the samejaw, but does not explain why the teeth thatoccupy homologous positions in the mammalianjaw are morphologically different in different

23

A B C D E

F G H

Figure 6. A schematic drawing of the development of a mammalian tooth. Initiation stage (A), early bud stage (B), middle budstage (C), late bud stage (D), cap stage (E), early bell stage (F), late bell stage (G), eruption (H). The epithelium is light gray,the mesenchyme medium gray, the signaling centers dark gray and dentine, cementum and enamel white. Gray lines indicatethe dental mesenchyme, the borders between the epithelial layers and the dentine-enamel or dentine-cementum junction. Thedrawings in the lower row are not to scale, as the tooth crown grows extensively during the bell stage.

species. Therefore, teeth can be used for study-ing both the evolutionary divergence of seriallyhomologous structures with different positionalidentities within one species and the perhaps themore interesting question of the evolution ofdisparate morphologies in the homologousstructures in different species.

Because metazoan genomes are very complex,(Drosophila has approximately 13600 genes andmammals appr. 70000 to 100000 genes) (Adamset al. 2000, O’Brien et al. 1999), tooth develop-ment is most easily studied in model organismswith many known genes. Of the five geneticallybest known experimental model organisms,Caenorhabditis elegans, zebrafish, Drosophilamelanogaster, the mouse and Xenopus laevis,only the mouse is a mammal and has a usefulfossil record. Mice are also easy to keep, repro-duces fast and has been studied for a long time.The mouse genome is being mapped and on 24th

February 00, GenBank contained 995142 mousesequences and sequence fragments(http://www.ncbi.nlm.nih.gov/Entrez/). Also, the

morphology, development and physiology of allmouse organs, including teeth, are well known.Together with modern molecular biologicaltechnologies, especially transgenesis, this allowsdetailed analysis of the biological and biochemi-cal functions of a wide variety of genes in devel-oping mouse tooth. Hence, a mouse tooth is anideal baseline model organ for comparing thegenetic basis of morphological evolution ofmuroid teeth.

The multicusped lower first molars (M1) of themouse have been extensively studied (see, e.g.,Thesleff and Sharpe 1997, Maas and Bei 1997,Weiss et al. 1998a, Peters and Balling 1999).They are small organs, and isolated mouse M1

can be cultivated in vitro from the early budstage onwards; even mouse mandibles cultivatedfrom tooth initiation stage onwards can form M1

tooth germs. Although the three-dimensionalmorphogenesis is distorted when teeth are cul-tured, the germs undergo the normal develop-mental stages, including histodifferentiation andmineralization, if suitable conditions are pro-

24

vided (Schmitt et al. 1999, Thesleff and Sahlberg1999), and the basic features of the individualteeth, e.g., the number of cusps, are usuallyrecognizable. Dissected tooth germs can also beimplanted into the anterior chamber of the eye(Mina and Kollar 1987, Lumsden 1988) or thekidney capsule, for analysis of the correct three-dimensional morphogenesis (Kratochvil et al.1996, Tucker et al. 1998b).

The molecular basis of the reciprocal inductiveinteractions between oral epithelium and under-lying mesenchyme during tooth development hasbeen studied, but the mechanisms of patternformation during the morphogenetic processes,such as when the two-dimensional epitheliumfolds to form a bud and later to surround themesenchyme, thus forming a three-dimensionalcrown, are still unknown. Epithelial-mesen-chymal interactions during pattern formation canbe analyzed by separating and recombining thetissues at various developmental stages or byadding ectopic inductory or inhibitory moleculesto the cultured tooth germs or dental tissues(Thesleff and Sahlberg 1999). Combined withmolecular biological tools, such as targeteddisruption of genes and other transgene techno-logies or to various methods for gene expressionstudies, the traditional tissue manipulation tech-niques allow us now to ask questions that werepreviously unanswerable.

However, the mouse is only one species, with avery specialized dentition. To find out how toothmorphogenesis can evolve, we must compare thedevelopment of mouse teeth to the developmentof teeth of other mammals. If the evolution oftooth developmental mechanisms is fast, inform-ation from the mouse is not directly applicable toother species. Though closely related species aremore comparable to each other than distantlyrelated ones, the problem of general applicabilitywould remain if they are compared. Therefore,in comparisons between closely related species

with highly disparate morphologies, the geneticbasis for morphological differences between theorthologous teeth is not lost in the evolutionarynoise.

Because tooth morphogenesis can essentially beunderstood as a continuing transformation ofmetric processes into metric and meristic infor-mation, i.e., the development of the numbers andshapes of teeth and individual cusps, carefulmorphometry, combined with small time intervalgene expression correlation is important.Although the rough expression correlations tothe individual differentiating tissues and meristicfeatures like cusps can be found simply bylooking at expression on the tissue level, toanalyze pattern formation processes during theconversion from metric to meristic traits, morecomplex statistical analysis of the emergingspatial patterns of gene expression and morpho-genesis is necessary.

At another level of pattern formation, recombin-ation of tissues from two different species allowsfor analysis of the evolutionary changes in in-ductive interactions. Because the tissues containundisturbed genomes, effects are seen in how thetissues of one species interpret the signalsproduced by the other species. The in vitrorecombinations also allow for dissection of therelatively few interactions during specific deve-opmental stages, which makes them valuable fordiscrimination between the evolutionary noiseand morphogenetic information contained in thegene expression data. By combining the induct-ion data with morphometry and correlating theresults with the information about pattern form-ation, it is possible to construct models of theroles of the basic molecular processes and sign-aling interactions in the spatiotemporal organis-ation of cellular differentiation and morphogen-esis, and hence the potential role of the genes inmorphological evolution.

25

The evolution of the mammalian dentition

The origin of teeth

The evolutionary origins of the teeth as organsare unclear. The oldest known mineralized verte-brate masticatory organs are the conodontelements, which apparently are not homologouswith modern teeth, although they consisted ofenameloid as well as cartilage and cellular bone(Sansom et al. 1992, Smith and Coates 1998).However, in the lineage leading to jawed verte-brates, enamel, dentine and bone were onlyfound in the exoskeleton (Kardong 1995, Smithand Coates 1998), and it is unknown when andhow the teeth evolved. The basic unit of exo-skeletal element is called an odontode. A com-plete odontode consists of an outer enamel layer,dentine surrounding a nerve canal, and a mesen-chymal bone of attachment, and this organiza-tion is still seen in mammalian teeth (Reif 1982).The enamel is secreted by ameloblasts derivedfrom epithelium and the dentin is secreted byodontoblast derived from mesenchyme. Oropha-ryngeal mineralized elements with similarhistology as exoskeletal elements have beenfound in agnathan thelodonts and several earlygnathostome fish groups (Smith and Coates1998), but until the affinities of the early fishlineages have been resolved, the origin of teethwill remain a mystery.

As an ossified endoskeleton evolved after theexoskeleton, the developmental programs formaking teeth are among the oldest for makingmineralized tissues in vertebrate bodies, and asall teeth within the same species develop viasimilar epithelial-mesenchymal interactions,they are serially homologous organs. Initially,the odontogenic potential has been more widelydistributed than it is in mammals, as still can beseen in modern amphibians and fish (Hankenand Hall 1993, Huysseune and Sire 1998). Teethare present not only in the premaxilla, maxillaand mandible, but also in the vomer and palatinebones, and in some cases even the neo-palatines,parasphenoid, pterycoid and ectopterycoidbones, as well as other pharyngeal arches, and

there often is more than one row of teeth. As theodontogenic potential was limited to the edges ofthe jaws, the number of tooth rows became lim-ited in synapsids, or mammal-like reptiles, andin archosaurs (the lineage leading to crocodiles,dinosaurs and birds) into one row or dentallamina along the oral sides of the first branchialarch, although in some later lineages replace-ment teeth could erupt lateral to their predeces-sors (e.g., Carroll 1988, Westergaard andFerguson 1990, Dingus et al. 1995).

The sizes, shapes and numbers of the teethwithin this single tooth row have divergedgreatly (for reviews, see, e.g., Romer 1966, Sav-age and Long 1986). In many modern actinop-terygian fish, the divergence depends on externalsignals, and is possibly mediated by the innerva-tion of the jaw (Tuisku and Hildebrand 1994,Huysseune 1995). During the evolution of thetetrapods the development of the dentition andindividual teeth became even more determinis-tic, reaching its extreme in modern mammals(Butler 1995, Huysseune and Sire 1998). Grad-ual differences between neighboring teeth in thetooth row seem to have evolved before thedivergence of the tetrapod lineage, because theshapes and sizes of the proximal teeth are differ-ent from the distal ones, e.g., in certain sharksand the zebrafish (Huysseune and Sire 1998),indicating that the proximal-distal axis of thejaws affected tooth morphogenesis already inearly jawed vertebrates. Genetically determinedpositional information is required for evolution-ary divergence of serially homologous structureslike teeth, and it is apparent that already in earlyamphibians the anterior teeth were distinctlydifferent from the posterior teeth, suggesting thatthe divergence of incisor and molar identitiesmay date back to the Carboniferous or earlier(Romer 1966, Huysseune and Sire 1998).

The integration of occlusion in mammals

In mammals, the dentition functions as a whole.Hence, a major problem in the evolution of the

26

mammalian dentition is the integration of thelocations and the shapes of opposing teeth indifferent replacement generations despite thegrowth of the jaws. During the evolution of themammalian dentition, tooth numbers have be-come reduced and individual teeth have special-ized for different functions within the dentitionto improve feeding efficiency. The anterior teethspecialized into instruments for capturing andgrasping the food items, whereas the posteriorteeth specialized into instruments for chewing,cutting and mincing the food for digestion. Theincrease in chewing efficiency was accom-plished by increasing the occlusal surfaces ofopposing teeth. This, in turn, caused the evolu-tion of complex cusp patterns in molars andpremolars, which are the teeth predominantlyused for mastication.

The anterior teeth were already different fromthe posterior ones in early tetrapods, and insynapsids, the differences between incisors andcanines grew pronounced, and the number ofcanines stabilized into one in each jaw quadrant(Romer 1966, Savage and Long 1986, Carroll1988). The cheek teeth became markedly differ-ent from the incisors, becoming multicusped(Romer 1966, Carroll 1988). Nowadays, thecanines are defined to be the most anterior max-illary teeth and the mandibular teeth are thosethat occlude in front of them (e.g., Osborn 1978,Schwartz 1982). However, the number of canini-form, incisiform and molariform teeth can varybetween species, even though the number oftooth germs may be the same, creating confusionin classifying homologous teeth (Osborn 1978,Butler 1978, Schwartz 1982). The differencesbetween the premolars and molars of modernmammals have been proposed to have evolvedfrom different replacement generations of synap-sid cheek teeth, because in fossil synapsids theearlier generation cheek teeth look differentfrom later generation cheek teeth, and even inmodern mammals the deciduous premolars re-semble morphologically molars more than per-manent premolars (Butler 1978, Butler 1995),although the resemblance between deciduouspremolars and molars is not necessarily causedby their ontogenetic evolutionary history, but by

ecophysiological constraints, as young mammalseat the same food as adults and thus also need asefficient mastication.

In the lineage leading to mammals, the numberof teeth became reduced. In early synapsids thepalate, pterygoid, vomer and even ectopterycoidwere tooth bearing bones, unlike in late synap-sids and mammals (Romer 1966). Moreover, thenumber of tooth rows per dental lamina and thenumber of dental laminae per jaw quadrant be-came reduced to one, and the possible initiationof teeth in the competent epithelium outside orbefore the morphologically distinguishable den-tal laminae ended (Westergaard and Ferguson1990). The current basic marsupial dental for-mula I4/5, C1/1, P3/3, M4/4 had already been de-termined during the Cretaceous period (Clemens1979), as probably was the basic placental dentalformula I3/3, C1/1, P4/4, M3/3, and their replace-ment patterns (Bown and Kraus 1979, Kielan-Jaworowska et al. 1979, Rougier et al. 1998,O’Brien et al. 1999; however, see also Schwartz1982), whilst the monotremes lost their teethduring the Cenozoic period, although the duck-billed platypus (Ornithorrhychus anatinus) hasrudimentary tooth germs as a juvenile (Luckettand Zeller 1989). To increase the occlusal fit, thenumber of tooth replacements was also reducedfrom at least five in some synapsids to one orless in modern mammals (Butler 1995). In thebasic placental dentition, all antemolar teeth arereplaced once, whereas in the basic marsupialdentition only P3/3 is replaced. The basic dentalformulas have been further modified in differentlineages, either by reduction of tooth numbersand replacements, or by increasing the numbersof teeth (the latter being very rare, and usuallyassociated with homodonty) (Kowalski 1976,Fortelius 1985, Janis and Fortelius 1988). Thereduction in tooth number and replacementsusually occurs gradually. Non-essential teethfirst become smaller and then either cease todevelop or are shed before they are used (Kurtén1953, Moss-Salentijn 1978, Schwartz 1982). Thedevelopment of unnecessary tooth germs can bearrested at any stage before tooth eruption(Kurtén 1953, Luckett 1985, Schwartz 1982,Moss-Salentijn 1978), although population

27

studies suggest that at certain developmentalstages there seems to be size thresholds thatmust be passed before the transition to the nextdevelopmental stage (Kurtén 1953, Grewal1962). Despite this, rudimentary tooth germs canpersist in the dentition for surprisingly long peri-ods (Luckett 1985, Kozawa et al. 1998), and theevolution of tooth loss does not necessarily re-move the later forming replacement teeth. Forexample, although marsupial dental formulasbecame fixed already during Cretaceous, opos-sum Monodelphis domestica has some rudimen-tary primary incisors, whilst the teeth erupting inthat location are secondary incisors (Kozawa etal. 1998).

The problem in reducing tooth positions andreplacements is that when the final dentition isworn down or lost, the animal will starve.Hence, adaptations that counteract the effects ofthe wear have been repeatedly favored (Janis andFortelius 1988). On the other hand, exact occlu-sion enables efficient mastication. In earlymammals, the development of occlusal surfaceswas not exact, but instead the functional shear-ing blades developed from wear facets of theteeth (Jenkins and Crompton 1979, Crompton1995). Increasing the accuracy of the initialocclusion by changing the tooth developmentalmechanisms increased the masticatory effi-ciency. This required that the programs control-ling the relative tooth locations in the opposingjaws and their cusp patterns became more inte-grated. Both the differentiation of the individualteeth and the reduction of tooth locations dependon changes in the early development of the teeth,in the determination of the odontogenic potentialand in the increasingly complex interpretation ofthe positional identity of individual teeth. Hence,the evolution of genetic mechanisms for deter-mining the numbers, locations, identities andmorphologies of each individual tooth must haveoccurred before or during the appearance of theabove mentioned trends in the fossil record.These innovations in the basic tooth develop-mental program were prerequisites for othermammalian adaptations to new diets, the in-creases in tooth size, lophedness and cusp num-bers and the durability increasing changes in

cusp shapes (Chaline 1989, Chaline and Sevilla1989, Hunter and Jernvall 1995, Jernvall 1995,Jernvall et al. 1996).

The evolution of molar cusp patterns

Both the incisors and the canines of early mam-mals were probably unicuspid and conical orspatulate, but the premolars and the molars arebelieved to have been primitively trituberculate,giving rise to tribosphenic teeth. Although somehave proposed that multicusped molars arose byfusion of several unicuspid teeth (for a classicreview, see Osborn 1907), this view has beenabandoned, and it is recognized that new cuspsevolve by increasing the local growth withincingulae or lophs or by increasing the size of denovo created cuspules (for reviews, see Butler1956, Jernvall 1995). In developmental termsthis means either that the mechanisms for mak-ing the primary cusp in unicuspid teeth, likemouse incisors, are applied in new places, or thatnew programs for making new cusps evolve.

Synapsids had multicusped molars (Romer 1966,Carroll 1988), and even the earliest mammalscould have complex molar cusp patterns, and themolars and premolars of, e.g., the Triassicharamiyids and Cretaceous multituberculateswere as complex as the teeth found in any mod-ern mammals (Clemens and Kielan-Jaworowska1979, Jenkins et al. 1997). Obviously, themechanisms for making multiple cusps hadevolved early in the mammalian lineage, and theevolution of the cusp pattern seems to be morerelated to ecological, rather than developmentalconstraints (Hunter and Jernvall 1995, Jernvall etal. 1996). However, the lineages with the mostcomplex molar types went extinct long beforethe present, and even the dominant Cretaceousorder of small mammals, the multituberculates,disappeared during the Paleocene epoch(Clemens and Kielan-Jaworowska 1979). Thus,all the current types of premolars and molars inmarsupial and placental mammals evolved fromtribosphenic molars of small omnivorousmammals, like those of the modern opossums(e.g., Osborn 1907, Bown and Kraus 1979).

28

The upper tribosphenic molar consists of threemain cusps; the paracone, the metacone and theprotocone, and two cuspules, protoconule andmetaconule, as shown in Figure 7. These threecusps form a trigon basin and are the oldest inevolution (e.g., Osborn 1907, Clemens and Lille-graven 1986). In addition, a new cusp, the hypo-cone, has evolved independently at least 22times, indicating that the cuspal homologies ingeneral may be a rather tenuous long-termmarker in evolution (Hunter and Jernvall 1995).

The lower tribosphenic molar, which is also theancestral form to the muroid M1, consists of fivemain cusps; protoconid, metaconid, paraconid,entoconid and hypoconid, and accessory cuspul-ids such as the hypoconulid, shown in Figure 7.The evolutionarily oldest cusp in the lower mo-lars seems to be the protoconid, which togetherwith two other cusps, the metaconid and paraco-nid, form the trigonid at the anterior end of thetooth (Osborn 1907, Clemens and Lillegraven1986). The posterior cusps, the hypoconid, ento-conid and hypoconulid, which form the talonidbasin, evolved later to increase the occlusal sur-face between opposing tooth rows. Again, theocclusal patterns can evolve quite flexibly, as therelative heights of the main cusps change andaccessory cuspules and cusps are added or lost.Hence, cuspal homologies are reliable only dur-ing short intervals in the fossil record, such asthe record of muroid rodents.

The evolution of specialized rodent dentition

The order Rodentia, with 29 families, 429 generaand more than 1800 species, is the most speciousextant mammalian order (Nowak 1991). Rodentsmake up about 40% of all extant mammalianspecies. Rodents have a very specialized dent-ition adapted for gnawing. Each jaw quadrantconsists of a single, ever growing or rootlessincisor used for gnawing through tough mater-ials. This incisor probably is the homologue ofthe second incisor (I2) of other mammals, andhas enamel only on its buccal surface, whereasthe lingual surface is covered by cementum (e.g.,Luckett 1985). In place of lateral incisors, ca-nines and anterior premolars, there is a toothless

end

hyd

hy

me

med

pa

padpr

prd

hld

1

2/3

2/3

4

5

6

pl ml

Figure 7. The cusp pattern of upper and lower tribosphenicmolars during occlusion. The numbers beside the cusps ofthe lower molar indicate the evolutionary order of appear-ance according to Osborn (1907) and Clemens andLillegraven (1986). Rodents have lost the paraconid and thehypoconulid, but in muroids the posterolophid or postero-conid have later evolved in place of the hypoconulid. Thelower molars are in gray and the upper molar, including thehypocone (whose evolution correlates with the concurrentloss of the paraconid) is colourless. end entoconid, hyd hypo-conid, hld hypoconulid, hy hypocone, me metacone, medmetaconid, ml metaconule, pa paracone, pad paraconid, plprotoconule, pr protocone, prd protoconid.

area, a diastema, which has evolved for wastedisposal during gnawing. The diastema and largeincisors are also seen in other gnawing animals,like lagomorphs and aye-ayes. The size, com-plexity and durability of the individual teeth hasincreased in many rodent lineages, but the num-ber of cheek teeth has usually become more re-duced, so that the dental formulas in differentlineages can vary from the primitive I1/1, - , P

2/1,M3/3 to the highly reduced I1/1, - , - , M

1/1 or tothe rare increase in tooth numbers in Heliocto-nius mole rats (Nowak 1991).

Molecular phylogenies suggest that the ancestorsof rodents diverged from lineages leading toother mammalian families some time during thelatter part of the Cretaceous (O’Brien et al.1999). The morphological divergence and adapt-ation for a gnawing lifestyle must have occurredbefore or during the Paleocene, when the oldestknown rodents, the Paramyids, appeared in thefossil record (Carroll 1988). The cheek teeth ofanother ancient fossil rodent, Tribosphenomysminutus, were quite small and simple in shape,and its upper jaw had two premolars and threemolars, whereas its lower jaw had one premolarand three molars like paramyids (Carroll 1988,Meng et al. 1994). The same dental formula isfound in extant squirrels (Luckett 1985, Nowak1991).

29

Remarkably, many rodent species, includingsquirrels, have rudimentary tooth germs in loca-tions where no teeth have existed for tens ofmillions of years (Moss-Salentijn 1978, Luckett1985, Luckett et al. 1989, Peterková et al. 1993).These rudiments are known to be arrested at anydevelopmental stage between small epithelialswellings and dentinal bell stages (Moss-Salen-tijn 1978, Luckett 1985, Peterková et al. 1998).Unless mineralized, rudimentary tooth germs donot fossilize. Hence, rudimentary teeth as foundin rodents can only be studied in extant species.Even so, they can provide important insights intothe evolution of the rodent dentition and the roleof positional identities in tooth development.

Different rodent species have different diets, andthere is a huge variety of rodent molar crowntypes. Although the tribosphenic molar is thebasal molar type for extant mammals, none ofthe modern rodents have a paraconid, and evenin the fossil record, it is found only in the cheekteeth of Tribosphenomys. Later evolution hasproduced other main cusps in many lineages,such as the anteroconid and instead of the hypo-conulid a posteroconid has evolved in muroidrodents (Figure 8). Aside from these main cusps,there can be any number of additional cusps andcuspules, and the cusp pattern may be obscuredeither by extensive lophedness of the crown orby the prismatism of the cusps. Because rodentsare herbivores or herbivore/insectivores, hypso-donty is common. In rodents adapted for afibrous diet, crown wear is heavy, and conicalcusps are less efficient than prismatic ones ascutting blades. Therefore, prismatism is associat-ed with rodent hypsodonty or hypselodonty.

The evolution of muroid teeth

The suborder Muroidea includes about 65% ofall extant rodent species, and the adaptive radiat-ion of Muroidea has been fast – a new mousespecies may have evolved even in historicaltimes (see Stanley 1979) and the lifespan ofsome vole species is about 300 000 to 400 000,which, together with their fossil record suggeststhat a new species of vole can evolve in a shorttime, maybe even within 10 000 years (Chaline

1989, Brunet-Lecomte and Chaline 1991). How-ever, these estimates are hard to substantiate,because lineage divergence does not necessarilyequate with morphological divergence (Martin1993, Rekovets 1994). Nevertheless, new molartooth shapes can evolve rapidly, and there isconsiderable variation within and between thepopulations of extant species, proving that thedevelopmental mechanisms behind the forma-tion of the species specific cusp patterns must beflexible (Martin 1993).

The ancestors of Eurasian muroid rodents, thecricetids, appear in the fossil record during theMiocene (Kälin 1999). Their functional dentalformula was I1/1, - , - , M3/3. The same dentalformula is still the most typical one for muroids,although in some species the reduction has gonefurther (Nowak 1991). The rudimentary toothgerms do not fossilize easily, hence their num-bers are unknown. The molars of early muroidsresembled those of modern hamsters.

The primitive cricetid M1 consisted of the fiveconical main cusps in two rows, together withpossible additional lophs and cuspules (Figure8). The metaconid has shifted anteriorly relativeto the protoconid into the same relative positionas the lost paraconid. All cusps point directly up-wards. The posterior end of M1 is defined by theposterolophid, which extends lingually from thehypoconid. As in tribosphenic molars, the cuspsare arranged into a diagonal pattern, but themetaconid occupied the relative position of themissing paraconid being anterior, not posteriorto the protoconid. The relative positions of theentoconid and hypoconid mimic the relativepositions of the metaconid and protoconid, andwhenever the anteroconid is split into two cusps,the main cusps can be arranged into a buccal anda lingual row. The splitting of the anteroconid isa common occurrence in muroid evolution, as isthe increase or reduction of new cusps and ac-cessory cuspules, and in some lineages the pos-terolophid evolved into a posteroconid (e.g.,Guthrie 1965, Chaline 1989, Kälin 1999,Freudenthal and Suárez 1999). The cusps ofupper molars are also in two rows, and pointdirectly downwards. Similar cusp patterns can

30

Allophaiomys pliocenicus

Antemus primitivus

Democricetodon sp

Promimomys cor n. sp.

1,0 mm

Mus musculus

Microtus rossiaemeridionalis

l b

a

p

p

p

p p

p

m

m

m

m

e

e

ee

e

h

h

h h

hpd

pd

pd

pd

pd

a

a

a

a

a

m

mp

eh

pd

a

12

43

5

6

6

5/7

78

9

5/7

Figure 8. The evolution of muroid M1. The earliest muroids (represented by Democricetodon sp.) had five or six conical maincusps and additional cuspules or lophids. In the mouse lineage the posteroconid evolved and the relative positions of the cuspsshifted to a more orthogonal orientation. In the vole lineage, the cusps became more prismatic, as the crown height increasedand the increase in tooth length preceeded the appearance of new cusps from the anteroconid. The anteroconid area is indicatedby the vertical bar. a anteroconid, e entoconid, h hypoconid, m metaconid, pa paraconid, po posteroconid or posterolophid prprotoconid. All teeth are to the same scale, the horizontal bar is 1 mm. Democricetodon was drawn after Kälin (1999) and M.pliocenicus after Kurtén (1968), whereas A. primitivus was drawn from a photograph in Wessels et al. (1982), Promimomyscor from a photograph in Kretzoi (1954), and the mouse and sibling vole molars from photographs taken from cleaned skulls.

B

A

Figure 9. The M1 cusp patterns of mouse (A) and siblingvole (B), shown in occlusal view. These teeth do notrepresent the extensive strain and population specificmorphological variation. The scale is 0.5 mm.

still be seen in extant cricetids (Gaunt 1961).

The two main adaptive radiations within theMuroidea are the evolution of Muridae, whichincludes mice and rats, starting during the LateMiocene, and the evolution of Microtidae, which

includes the voles and lemmings, starting in theEarly Pliocene about 5 million years ago. Themorphological diversification leading to the 121extant species of modern Microtidae (Nowak1991) began in the late-middle Pliocene with theappearance of the genus Promimomys, althoughits relationship to some lineages, like Lemmini(lemmings), is unclear (Chaline 1989, Chalineand Graf 1988, Gromov and Polyakov 1992).However, based on the fossil record, the genusMicrotus has been proposed to have descendedfrom the genus Promimomys (Fejfar and Hein-rich 1989, Gromov and Polyakov 1992). Thediversification of Murinae at Late Miocene some11 million years ago from a Megacricetodon-like ancestor (Freudenthal and Suárez 1999), hasproduced about 460 extant species (Nowak1991), making the Murinae the largest extant orextinct mammalian subfamily. However, accord-ing to the molecular data, the subfamily Murinaediverged from the stem cricetids about 20 mil-lion years ago, and the Microtinae a couple ofmillion years later (Nikoletopoulos et al. 1992,Robinson et al. 1997). Hence, the morphological

31

evolution seen in the fossil record does not coin-cide with the lineage splitting events indicatedby the molecular data.

In Murinae, or the lineage leading to the mouse,the posterior cingulum extending from the hypo-conid in the cricetid M1 evolved into a hypo-conulid. The cusp pattern can be either diagonalor orthogonal. The cusps and possible cuspulesof the lower molars are still in two rows, but inthe upper molars they are arranged in three rows.The cusps are conical, but in the upper molarsthey point in a posteroventral direction, whereasin the lower molars they point in an anterodorsaldirection. The lingualmost row of upper molarsapparently originates from accessory cuspules.

In Microtinae, the molars have evolved to copewith a coarser diet containing more cellulose andsilicates. Crown size has increased in both lengthand height. After the cusps had become prism-atic, the increasing hypsodonty became in somegenera hypselodonty, i.e., the molars becameever growing like the incisors (e.g., Fejfar andHeinrich 1989). As the molar height increased,cementum appeared into the re-entrant anglesbetween the consecutive buccal or lingual prisms(e.g., Chaline 1989, Fejfar and Heinrich 1989).The cusp pattern remained more or less diago-nal, but the number of cusps increased. The fos-sil record indicates that in the lower molars, theadditional cusps budded off the anterior loop,which is homologous with the anteroconids ofthe other muroids, whereas in the upper molarsthe new cusps were added to the posterior end ofthe molars (Guthrie 1965). Because of this, theincrease in molar length of voles was greatest inM1 and M3. The posteriormost prism of the voleM1 has evolved from the posterolophid.

The lower first molars of the mouse and siblingvole

The M1 of the house mouse and the sibling volerepresent the morphological extremes of theirevolutionary lineages, and are therefore goodmodels for a rough comparison of the morpho-genetic processes. The house mouse appeared inthe fossil record as a species about 0.9 millionyears ago (Auffray 1988). Its molars are brachy-dont and semilophodont, with conical cusps with

a typical Murinae slant. The accessory cuspulesare usually missing (Grüneberg 1965). Theenamel is also thicker on the anterior sides of theupper molar cusps and on the posterior sides ofthe lower molar (Lyngstadaas et al. 1998). Thecusp tips develop to be enamel-free. The molarsare rooted and the roots are anchored to thealveolar bone by cementum. In mouse molars,cementum only exists in the roots.

The crown of the mouse M1 is 1.4 mm long and0.8 mm wide, and consists of seven cusps, whichare arranged into pairs connected by lophs (Fig-ure 9). The protoconid – metaconid and hypo-conid – entoconid pairs are arranged orthogo-nally to the anterior – posterior axis, but the twoanteroconid cups are asymmetrically arranged,so that the lingual cusp is anterior to the buccalone, and the posteroconid is central to the poste-rior end of the crown, as in other Muridae. Un-like in other Muridae, there are no clear acces-sory cuspules. Mouse M1 has two roots.

The common vole appeared in the fossil recordin the mid-Pleistocene, but because the siblingvole dentition and skeleton are morphologicallyidentical to the dentition and skeleton of its sib-ling species, the common vole (Microtus ar-valis), the fossils of this species pair are calledthe common vole (MacDonald and Barrett 1993,Rekovets 1994). Because the sibling vole hasmore chromosomes (2n = 56 as compared to 2n= 46 in the common vole), it is possible that thesibling vole represents the ancestral species(Rekovets 1994). The sibling vole molars arehypselodont and prismatic. The re-entrant anglesbetween the consecutive buccal or lingual prismsare filled with cementum. The enamel is strongeron the posterior sides of the upper molar cuspsand on the anterior sides of the lower molarcusps, thus forming the main cutting blades (vonKoenigswald 1982). The central anterior andposterior angles of the cusps are fused together,forming a central ridge in the molars. The ves-tigial cusp tips are worn away within one day oferuption. In the lower molars the anterior edgesof the prisms have thicker enamel that the poste-rior edges, the opposite to the upper molars.

The sibling vole M1 is 2.5 mm long and 1.0 mmwide, and consists of nine diagonally arranged

32

(Note: Possible rudimentary incisors are not shown in this figure)

d2

d3

d

pr

d1pc

Figure 10. Mouse (A) and vole (B) upper jaw dental formulas. The insets show the rudimentary tooth germs at their maximaldevelopment before their apoptotic removal. Molars and incisors are shown in light gray and the rudimentary diastema toothgerms in dark gray. Potential rudimentary incisors are not shown. d1 mouse first rudimentary tooth germ, d2 mouse secondrudimentary tooth germ, d3 mouse third rudimentary tooth germ, d vole rudimentary tooth germ, pc primary choana, pr pala-tal ruga. The scale bar in the insets is 150 µm.

prismatic cusps, that lie at about 30° to eachother, and an anterior loop (Figure 9). The fivelingual prisms are wider than the four buccalones, and the anteriormost prisms are the small-est. This means that the buccal re-entrant anglesare wider than the lingual ones.

Both the mouse and sibling vole have the typicalmuroid adult dental formula I1/1, - , - , M

3/3, but

the mouse has in its upper diastema region threerudimentary tooth gems (see also Lesot et al.1998). These tooth germs obviously antedate theorigin of Muroidea. Because even Tribospheno-mys had only two premolars in each upper quad-rant (Meng et al. 1994), some of the mouserudimentary tooth germs may antedate the adapt-ive radiation of the order Rodentia.

The developmental biology of the teeth

Morphogenesis

Tooth morphogenesis has been extensivelystudied since the last century (Owen 1840-45),and the morphogenesis of mouse dentition, espe-cially M1, is well known. The morphogenesis ofthe upper and the lower first molars has beenmapped in detail, because they are the first andlargest multicusped teeth to develop in themouse. As its basic morphogenesis was welldescribed before most molecular biology tech-niques had been invented, mouse M1 became thebasic model for studying the function of genesduring tooth morphogenesis.

All teeth develop via epithelial-mesenchymalinteractions. In mammals in general, the oralepithelium thickens to form the dental lamina,from which the individual tooth germs bud (Fig-ure 6). The bud stage is followed by the capstage, when crown morphogenesis begins. Dur-ing the bell stage, the cusp patterns of the teethare formed and mineralization begins. In mice,tooth morphogenesis is initiated at E11, whenthe oral epithelium begins to thicken in the inci-sor and molar areas. The identities of the toothgerms are apparently already determined by un-known mechanisms (Tucker et al. 1998b). Thedental lamina in mice is very thin, unlike inother mammals (e.g., opossums). Except for M2

33

and M3, which are formed from the posteriorends of molars anterior to them, the individualtooth germs, including the rudimentary ones, areformed as epithelial invaginations, which budfrom the dental lamina. Earlier recombinationexperiments have suggested, that prior to theearly bud stage the control of dental identity is inthe epithelium, but during the early bud stage thecontrol is switched to the newly induced dentalmesenchyme (Mina and Kollar 1987, Kollar andMina 1991). The mesenchyme condenses aroundthe developing molar buds, and later forms thedental papilla, dental follicle and alveolar bone.

As the teeth grow from the tip down, the differ-entiation of the cells also progresses in the samedirection (Figure 11). The cervical loops are stillgrowing when mineralization begins, and inopen rooted teeth the growth never ends. Odon-toblast differentiation induces ameloblast differ-entiation, which is followed by the secretion ofdentine and enamel prisms (Smith 1995,Thesleff 1995). Since the odontoblasts andameloblasts are exactly aligned, the resultingdentine and enamel prisms are also aligned.Normally, as in mouse molars, the epitheliumcloses around the crown, leaving openings forroots (Butler 1956). The roots of mouse molarsare formed when the crown has mineralized, andthe tooth erupts. In open rooted teeth eruptionbegins as the crown height increases, without thecrown ever becoming fully mineralized. Sincethe tooth is surrounded by alveolar bone, erup-tion requires at least bone resorption by osteo-clast activity (Tiffee et al. 1999).

At late bud stage, the primary enamel knot be-gins to develop (Jernvall et al. 1998). Theenamel knots are areas of condensed, quiescentepithelium, and been found in many mammalsand in crocodiles (Butler 1956, Jernvall 1995,Westergaard and Ferguson 1987), and their rolein tooth development has been debated (see, e.g.,Butler 1956, Jernvall 1995, Lesot et al. 1996,Coin et al. 1999, Lesot et al. 1999, Jernvall andThesleff 2000). At the onset of the cap stage, thecervical loops begin to grow around the primaryenamel knot, surrounding the dental papilla,which forms the mesenchymal part of the crown.

x

y

z

x

y

z

x

y

z

Figure 11. Teeth grow and differentiate from the tip downand the absolute differences in cusp heights on the y-axiscorrelate with the time of initiation (adapted from Butler1956). However, on the x- and z-axes the cusp tips becomemore separated as the cusps grow in width as well as inheight. The increasing differentiation is shown by thedarkening color.

During crown morphogenesis, the epitheliumdifferentiates into inner enamel epithelium andouter enamel epithelium, and the stellate reti-culum and stratum intermedium begin to appear(e.g., Thesleff 1995, Lesot et al. 1999). Theenamel knots consist of inner enamel epitheliumand stratum intermedium cells. Some reportsalso distinguish an enamel septum, a chord ofdenser stellate reticulum leading from theenamel knot in the inner enamel epithelium tothe enamel navel in the outer enamel epithelium(Butler 1956, MacKenzie et al. 1992). After thecap stage, most of the mouse M1 primaryenamel knot is lost apoptotically (Lesot et al.1996, Jernvall et al. 1998, Lesot et al. 1999), butnew enamel knots are initiated (Jernvall 1995,Jernvall and Thesleff 2000). The initiation ofsecondary enamel knots marks the beginning ofthe bell stage. The locations of these secondaryenamel knots correlate with the locations of theinitiated cusps in both mice and opossums (Jern-vall 1995), and have been proposed to beinvolved in crown morphogenesis (for reviews,see Butler 1956, Jernvall 1995).

Because the epithelium is essentially two-dimen-sional tissue, whereas the mesenchyme is essent-ially three-dimensional, tooth crown morphogen-esis can be described as remodeling of an epi-thelial sheet. It has been proposed that unequalcell division in the different parts of the dentalepithelium and dental papilla, combined withmechanical pressures inflicted by the swellingstellate reticulum and condensing dental follicle

34

and septa, shape the tooth germs, but exactlyhow this happens is unknown (Butler 1956).

In some cases, tooth germ development fromone stage to another is disturbed and the toothgerm is later resorbed. This evolutionary loss ofteeth often begins with vestigial teeth, which areeither too small to have a function and often donot even erupt or are shed before they can beused (e.g., Kurtén 1953, Schwartz 1982, Moss-Salentijn 1978). The development of rudiment-ary tooth germs can be arrested at any stage, andthe arrest is often associated with abnormal his-todifferentiation (Moss-Salentijn 1978, Luckett1985, Luckett et al. 1989). This suggests aqualitative inductive failure in during odonto-genesis. However, some reports indicate, thatbetween different developmental stages, thereexists size thresholds which must be passed be-fore transition to next stage (Grewal 1962,Kurtén 1953). This implies a quantitative geneticfailure. Since tooth shapes and numbers in manymammals, including humans, vary within thepopulation (e.g., Grüneberg 1951, Grüneberg1965, Kurtén 1953, Garn et al. 1964, Berry1978), and since the last developing teeth areaffected both in many hypodontia mutations inhumans (e.g., Garn et al. 1964, Vastardis et al.1996) and in some cases of evolutionary toothreduction (Kurtén 1953), it seems that quantita-tive changes in tooth development are importantmeans for evolutionary tooth loss. However, it isunknown whether the disturbances occur at themolecular level, e.g., as an insufficiency of asignal, or if they have a mechanistic basis, e.g.,an insufficient amount of tissue for morphogene-sis, or a combination of both.

Determination of positional identity and toothmorphogenesis

Although the morphogenesis of mouse teeth iswell known, the mechanisms regulating it arelargely unknown. All mouse molars have shapesthat differ from those of incisors and recombin-ation experiments have suggested that the earlyepithelium controls the shapes of the individualtooth germs. If an initiation stage incisor epithel-ium is recombined with induced molar mesen-

chyme, an incisiform tooth will develop indicat-ing a reprogramming of the molar mesenchymeby an early incisor epithelium (Kollar and Mina1991). In fact, at that stage the oral epitheliumcan, when recombined with non-oral cranialneural crest derived 2nd branchial arch mesen-chyme or even trunk neural crest mesenchyme,direct it to a dental fate (Mina and Kollar 1987,Lumsden 1988). After the initiation stage, thedental mesenchyme can determine tooth shape,and even direct odontogenesis in non-oral epi-thelium, like footpad epidermis (Kollar andBaird 1970). Thus, the early epithelium containsthe patterning information necessary for induc-tion of tooth germs, determining their identityand controlling their morphogenesis, but thisinformation is somehow transferred into dentalmesenchyme. The exact nature and mechanismsof this information transfer between epitheliumand mesenchyme is unknown.

The earliest induction in odontogenesis mayactually come from the cephalic neural crestderived ectomesenchyme, which, according tothe clonal theory, is programmed into three typesof dental fates either prior to or after its migra-tion to the first branchial arch, forming both theupper and the lower jaws of vertebrates (Osborn1978, Osborn 1993, see also Smith 1995). Insuch case, the incisors, the canines and themolars and premolars develop from one of theseclones by growth and division of the toothforming tissues. However, as the fate of theneural crest derived mesenchyme can be repro-grammed by recombining it at early stages withheterologous epithelium (Lumsden 1988, Kollarand Mina 1991), it seems unlikely that the clonaltheory as such is correct. The early neural crestderived ectomesenchyme may, however, providepermissive signals to the overlying ectoderm,inducing it to become odontogenic epithelium.

The field theory maintains that the first branchialarch is patterned by morphogenetic fields (Butler1978), and that the relative intensities of the sig-naling molecules determine the positional iden-tity of the forming tooth germ. In the positionalhomeobox code model the positional identitiesof the tooth germs depend on the local combina-

35

tions of Msx, Dlx and other homeobox tran-scription factors (Sharpe 1995, Stock et al. 1997,Thomas and Sharpe 1998, Thomas et al. 1998,Weiss et al. 1998b). A combination of these hy-potheses is that signals from early branchial archectoderm pattern the facial prominences by in-ducing different local combinations of homeo-box genes, and that this patterning then deter-mines the dental positional identities (Weiss etal. 1998b). This seems quite likely, since it ispossible to change incisors into molariform teethby placing a bead soaked in Noggin protein atthe anterior end of an early mandible. Noggin isan antagonist to the signaling molecule Bmp4,expressed in early anterior epithelium (Tucker etal. 1998b), and the morphological change is ac-companied by an ectopic anterior expression ofBarx1, which is normally associated with theposterior mandible and molar region mesen-chyme (Tissier-Seta et al. 1995, Mucchielli et al.1997, Tucker et al. 1998b). However, early in-duction of different Dlx-genes in mouse maxillaand premaxilla by the same signals (Fgf8 ordental epithelium) indicates that in the ectomes-enchymes of upper and lower jaws some intrin-sic differences may exist prior to the inductionof teeth (Ferguson et al. 2000).

Aside from the problem of dental identity, thereis the problem of determination of dental laminaand the numbers and locations of the individualtooth germs. Although it has been shown that thetooth developmental programs have diverged somuch according to the positional identity of theteeth that removing the function of genes be-longing to the odontogenic homeobox code(Sharpe 1995, Thomas and Sharpe 1998, Tho-mas et al. 1998) may affect some teeth differ-ently from others. All mammalian teeth, regard-less of their identity, bud from the dental laminaor its derivatives (e.g., Luckett 1993, Weiss et al.1998a; see below). Hence, the processes con-trolling the number and locations of individualtooth germs must exert their effect via the dentallamina. However, the molecular basis of toothsite determination has been connected with clo-nal and field theories controlling positionalidentity, although it is so far is unknown if theinitiation of tooth germ and the determination of

its positional identity are even controlled by thesame inductive pathways (Weiss et al. 1998a).One reason for this ambiguity may lie in theextreme specialization of the mouse dentition.Although the mouse is the main model in thesestudies, at early developmental stages there ex-ists only one incisor and one molar tooth germseparated by a diastema in each jaw quadrant.Hence, it has been difficult to separate the proc-esses controlling the initiation of the tooth germsfrom, e.g., the processes controlling toothidentity.

More is known about development after theearliest stages. Targeted gene disruption experi-ments have shown that the transition from thebud to the cap stage depends on mesenchymalsignal(s), suggesting mesenchymal control ofcrown morphogenesis. Loss of exclusively mes-enchymally expressed transcription factors Msx1or Pax9 arrest tooth development at the late budstage (Satokata and Maas 1994, Peters et al.1998). However, epithelial Lef1 must be presentin the tip of the late bud, presumably to initiateenamel knot development (Kratochwil et al.1996, Peters and Balling 1999). The genes in-volved in the development of positional identity,such as Barx1, Dlx1/2 or ActivinβA, also seemto act at the initiation stage (Table 1, Tucker etal. 1998b)

The epithelial-mesenchymal interactions involv-ed in cusp pattern formation during crown mor-phogenesis are practically unknown. It is known,however, that the differentiation of the variousdental cell types requires interaction between themesenchyme and the epithelium. For instance,the "tip down" ameloblast differentiation de-pends on signals from the odontoblasts, butodontoblast-like differentiation can be inducedunder cell culture conditions by applying ectopicfollistatin, which is normally produced by differ-entiating ameloblasts (Heikinheimo et al. 1997,Thesleff et al. 1995a). On the other hand, themesenchyme is known to produce many mito-gens and other signaling molecules, which maylocally affect the adjacent mesenchyme (see, forexamplehttp://honeybee.helsinki.fi/toothexp/index.htm).

36

The mesenchyme also determines the rate ofstem cell division in the cervical loops of inci-sors (Harada et al. 1999). Since both growth anddifferentiation are affected by these signals,epithelial-mesenchymal interactions are nodoubt important for morphogenesis and evolu-tion of species specific dental morphologies. Toanalyze how the mesenchyme patterns the epi-thelium or the epithelium the mesenchyme,cross-species recombinations are needed be-tween the orthologous teeth of different species.

Molecular interactions in tooth morphogenesis

The role of genes in the inductive interactionsduring tooth development has been studied ex-tensively in mice (see, e.g., Thesleff and Sharpe1997, Maas and Bei 1997, Weiss et al. 1998a).At least 170 genes are known to be expressed inthe teeth, and more are being found at a rapidrate, but the only tooth specific genes so farfound are some of the genes coding the matrixmolecules secreted by ameloblasts and odonto-blasts(http://honeybee.helsinki.fi/toothexp/index.htm).Therefore, the tooth development is controlledby the same developmental genes as many otherorgans (Thesleff et al. 1995b, Weiss et al.1998a).

Targeted disruptions of mouse genes and in vitroinduction experiments have shown that bothmesenchymal and epithelial genes are important(Table 1). In many cases, the loss of a gene leadsto the developmental arrest of the tooth germs ata late bud stage, suggesting that the inductiveinteractions required for initiation of the capstage are missing. For example, loss of Lef1,Msx1 and Pax9 leads to this phenotype. In othercases, only some of the later developing teethmay be missing, whereas other teeth are smallerand abnormal in shape, suggesting a quantitativefailure in growth and/or patterning (Grüneberg1965). Tabby/EDA is an example of such agene, as is its putative receptor, Downless(Headon and Overbeek 1999, Pispa et al. 1999).In quite a few cases, only one or more toothtypes are affected, suggesting that the gene isessential only for their development (Table 1).

For example, ActivinβA-/- targeted mutant micelack incisors and mandibular molars, whereasDlx1-/-/Dlx2-/- double targeted mutants lack up-per molars (Thomas et al. 1997, Ferguson et al.1998). Because teeth are believed to be seriallyhomologous, this indicates that the basic toothdevelopmental program has become modified inteeth with different positional identities, whichin turn may be essential for the evolutionarydivergence of dental morphologies in the samedentition. Many such genes are only expressedin parts of the jaw, whereas the others are ex-pressed in all teeth, but are absolutely essentialfor only a few of them.

The effects of individual genes depend on thegenetic background of the population, as shownby Pax6-/- mutant mice – in certain strains thereare ectopic incisors, whereas in others these donot occur (Quinn et al. 1997). The severity of theTabby mutation also depends on the background(Grüneberg 1965). This is probably due to thefunctional redundancy between different genesand pathways present in the same tissues. Al-though exact tooth shapes and the prevalence ofmissing teeth varies between different strains ofmice (Grüneberg 1951, Grüneberg 1965), thedevelopment of teeth is strictly regulated and notvery easily disturbed. Redundancy has beenshown, e.g., for Msx1 and Msx2 and for Dlx1and Dlx2. In the former case, the loss of Msx1alone causes the developmental arrest of all teethat the late bud stage, whereas the loss of Msx2alone causes abnormal apoptosis in the bell stagedental epithelium (Satokata and Maas 1994,Maas and Bei 1997). When both genes are lost,tooth development is arrested shortly after initat-ion, suggesting that Msx2 and Msx1 are inter-changeable during early development (Maas andBei 1997). Likewise, though the loss of bothDlx1 and Dlx2 genes arrests the upper molars atthe lamina stage, the loss of either gene alonedoes not cause dental defects (Thomas et al.1997). Such redundancies are probably commonbetween all related genes, including growthfactors.

Many growth factors and transcription factors, aswell as molecules involved in processes like cell

37

cycle control are developmentally important.These genes belong to conserved families, andare used for the development of several organs.Although their functions in other organs can giveclues to their effects in tooth morphogenesis,their role in teeth must be studied separately,because the combinations or exact morphoge-netic functions of pathways are not always con-served (Meyers and Martin 1999). Moreover,many genes are used repeatedly for differentprocesses during organogenesis (e.g., Chen et al.1996, Jernvall et al. 1998, Tucker et al. 1998b).Also, many genes are expressed in locationswhere they do not seem to have any function(e.g., Kratochwil et al. 1996), either because offunctional redundancy or because correct down-stream signaling components are missing. Sincegratuituous expression patterns are less likely tobe conserved in different species, the compara-tive analysis of the expression patterns of thecandidate developmental genes to the developingmorphologies in closely related species, e.g.,different muroids, would be informative. If,despite the fast morphological evolution ofmuroids, the correlation between expression andmorphogenesis was conserved, this would beevidence that the developmental pathway wasprobably involved in morphogenesis.

As the number of potential morphogenetic con-trol genes is so high, this analysis was limited toa few key signaling molecules (Bmp2, Bmp4,Fgf4, Fgf8, Shh and Wnt10a), transcription fac-tors (Lef1, Msx1, Msx2, Pax9 and Pitx2) and acyclin dependent kinase inhibitor p21CIP1/WAF1.These genes belong to various pathways knownor believed to function during tooth initiationand morphogenesis (Table 2), and they havebeen shown to be expressed in developing teeth.Although many of the pathways seem to act inparallel (Dassule and McMahon 1998), they canalso be interconnected at some level. For exam-ple, Bmp4 induces the expression of the homeo-box transcription factor Msx1, which is requiredboth for induction and maintainance of mesen-chymal Bmp4 expression and for Shh-inducibletranscription of Shh-signaling receptor/inhibitor-subunit Ptc (Chen et al. 1996, Zhang et al.1999). On the other hand, whilst Bmp4 is essen-

tial for Shh expression, excess Bmp4 down-regulates it (Zhang et al. 2000). Such autoregu-latory loops are probably involved in regulationof the epithelial-mesenchymal interactions intooth development.

Growth factors Bmp2 and Bmp4 are the mam-malian orthologues of Drosophila Dpp (Wozney1998). All three genes belong to Bmp-family ofTGFβ-superfamily of signaling molecules. Inci-dentally, Bmps are actually a better example ofparalogous genes than Hox-genes, in which thenomenclature has been complicated by the un-usually high clustering of the individual genesand consequent cluster duplication events. BothBmp2 and Bmp4 are associated with differentia-tion, but Bmp4 may also be a natural inducer ofapoptosis (Graham et al. 1994, Winnier et al.1995, Lough et al. 1996, Marazzi et al. 1997,Chen and Zhao 1998). The mammalian geneswere first associated with the differentiation ofosteoblasts, but in Drosophila, Dpp has tradi-tionally been considered to be a pattering mole-cule (e.g., Sampath et al. 1993, Jiang and Struhl1996). Because the Bmp2 and –4 sequences areclosely related, they are functionally inter-changeable (Padgett et al. 1993), and both pro-teins can induce expression of the homeoboxtranscription factors Msx1 and Msx2 and TCF-transcription factor Lef1 (Vainio et al. 1993,Chen et al. 1996, Dassule and McMahon 1998,Tucker et al. 1998a). Bmp4, Lef1, Msx1 andMsx2 have been experimentally shown to beessential for normal tooth development (Satokataand Maas 1994, van Genderen et al. 1994, Chenet al. 1996, Maas and Bei 1997). The role ofBmp2 is unknown, because mice lacking Bmp2die before tooth morphogenesis is initiated(Zhang and Bradley 1996), and conditional genedisruption for assessing its role in odontogenesishas not been done.

Growth factors Fgf4 and Fgf8 are potent mito-gens for both the epithelium and mesenchyme inseveral systems, including teeth (Mahmood et al.1995, Lee et al. 1997, Kettunen et al. 1998). Inlimb buds, they are expressed in the apicalectodermal ridge (AER), and are most probablyinvolved in limb growth and/or patterning

38

Table 1. Some genes with known function during odontogenesis and their known roles in developing teeth.

Effects ReferenceActivinβA In ActivinβA-/- mutant mice incisors and mandibular molars arrested

at bud stage, ActivinβA induces FollistatinMatzuk et al. 1995a,Matzuk et al. 1995b,Ferguson et al. 1998

ActivinRcII Some ActRcII-/- mutant mice lack (at least) mandibular incisors Matzuk et al. 1995aBmp2 Can mimic the effects of Bmp4 in vitro Vainio et al. 1993,

Jernvall et al. 1998,Neubüser et al. 1997

Bmp4 Involved in determination of incisor identity, prevents induction ofdental mesenchyme by inhibiting Pax9 expression prior to E11,essential for cap stage transition, possibly involved in apoptoticremoval of epithelial cells

Chen et al. 1996, Jernvallet al. 1998, Neubüser etal. 1997, Tucker et al.1998b

c-met Hgf-receptor Tabata et al. 1996c-src c-src-/- mutant mice have malformed incisors because teeth do not

eruptTiffee et al. 1999

Dlx1 Dlx1-/- mutant mouse dentition is normal. In Dlx1-/-/Dlx2-/- mutantmouse maxillary molars are arrested at epithelial thickening stage,other teeth are normal

Thomas et al. 1997

Dlx2 Dlx2-/- mutant mouse teeth are normal Thomas et al. 1997Dlx3 Dlx3+/- mutation causes taurodontism and enamel hypoplasia in

humansPrice et al. 1998

Downless dl-/- mice have small and malformed teeth, lack of third molarsdepending on the strain. Downless is probably an EDA receptor

Headon and Overbeek,1999

EDA EDA-/- mice and humans have small and malformed teeth andhypodontia

e.g., Srivastava et al. 1997

Egf AS Egf oligonucleotides prevent tooth development in vitro Kronmiller et al. 1991Fgf3 possible role in maintaining epithelial growth Harada et al. 1999Fgf4 possible role in pattern formation during crown morphogenesis Jernvall et al. 1994Fgf8 Induces mesenchymal ActivinβA and Pax9, induces epithelial and

mesenchymal proliferationFerguson et al. 1998,Kettunen and Thesleff1998, Neubüser et al.1997

Fgf9 possibly similar in role to Fgf4 and Fgf8 Kettunen et al. 1998Fgf10 maintains and induces cervical loop growth in incisor explants Harada et al. 1999Follistatin Missing or delayed (mandibular) incisor development in mutant

mice, Follistatin can induce odontoblast-like differentiation in vitroMatzuk et al. 1995c,Heikinheimo et al. 1997.

Gli2 Fused premaxillary incisors in Gli2-/- mutant mice, Gli2-/-/Gli3+/-

mutant mice had smaller than normal mandibular incisors and molarsand maxillary incisors arrest at epithelial thickening stage

Hardcastle et al. 1998

Gli3 Gli3-/- mutant mice have normal teeth, Gli2-/-/Gli3-/- mutant mice hadrudimentary incisor buds and no sign of molars,

Hardcastle et al. 1998

Hgf Hgf AS oligonucleotides cause abnormal “inverted” crownmorphogenesis in vitro.

Tabata et al. 1996

Lef1 Development of all teeth arrested at bud stage, essential in epithelium van Genderen et al. 1994M-CSF M-CSF-/- mutant mice have malformed incisors because teeth do not

eruptTiffee et al. 1999

Midkine Midkine is essential for differentiation and morphogenesis Mitsiadis et al. 1995Msx1 In Msx1-/- mutant mice development of molars arrested at bud stage

and incisors not found in newborns, essential for maintenance ofmesenchymal Bmp4 expression, Msx1 haploinsufficiency causesselective tooth agenesis in humans

Satokata and Maas 1994,Chen et al. 1996,Vastardis et al. 1996, Huet al. 1998

39

Table 1. (Continued)

Msx2 Msx2-/- mutant mice have brittle and malformed teeth, abnormalitiesin stellate reticulum and delayed M3 development. In Msx1-/-/Msx2-/-

mutant mice all teeth are arrested at dental lamina stage

Maas and Bei 1997, Beiand Maas 1998

Notch1 Notch1 has a possible role in epithelial stem cell maintenance anddifferentiation in incisor cervical loops

Harada et al. 1999

Pax6 Pax6-/- mutant mice have supernumerary upper incisors at straindependent penetrance

Quinn et al. 1997

Pax9 In Pax9-/- mutants all teeth arrest at late bud stage. Peters et al. 1998Pdgf-A Ectopic PDGF-A increases tooth germ size and supports its

development in vitroChai et al. 1998; Hu et al.1995

Pdgfrα tooth crown development does not occur in Pdgfrα-/- mutant mice Morrison-Graham et al.1992

Pitx2/RIEG Haploinsufficiency causes anodontia, microdontia and abnormallyshaped or implanted teeth in humans, maxillary teeth in Pitx2-/-

mutant mice are arrested at placodal and mandibular at bud stage.

Semina et al. 1996, Lin etal. 1999, Lu et al. 1999.

Shh Ectopic Shh causes ectopic invaginations and malformations in toothbuds

Hardcastle et al. 1998

Tachykinins(SP)

AS-oligonucleotides against Tachykinins arrests tooth developmentin vitro

Weil et al. 1995

SP-receptor SP-receptor block greatly retarded tooth development in vitro Weil et al. 1995

Table 2. The analyzed molecules (bold) and their known upstream regulators and some of their targets in the tooth.

Inducer Inhibitor Gene Upregulates Downregulates

Noggin Bmp2 Bmp4, Lef1, Msx1, Msx2,p21CIP1/WAF1

Pax9

Noggin Bmp4 Egr1, Msx1, Ptc, Msx2, Dlx2,p21CIP1/WAF1, apoptosis

Barx1, Pax9

Fgf4 proliferation, Syndecan-1 apoptosisPitx2 Fgf8 ActivinEA, Barx1, Dlx1, Dlx2,

Fgf3, Msx1, proliferationBmp2, Bmp4, Wnt10b Lef1Bmp2, Bmp4, Fgf8 Msx1 Bmp4, Ptc, Fgf3, Syndecan-1Bmp2, Bmp4 Msx2Bmp2, Bmp4 p21CIP1/WAF1

Fgf8, Fgf9 Bmp2,Bmp4

Pax9 Bmp4

Pitx2 Fgf8Shh proliferation, Gli1, Ptc Wnt10b, Pax6

Shh Wnt10b Lef1

(Niswander and Martin 1993, Mahmood et al.1995, Vogel et al. 1996, Ohuchi et al. 1997,Zúñiga et al. 1999; however, see also Moon etal. 2000). In the midbrain-hindbrain junction oristhmus, Fgf8 has a role in the patterning andgrowth of the mesencephalon (Crossley et al.1996, Lee et al. 1997). Both AER and isthmusare signaling centers that are important for pat-

tern formation and local differentiation. In teeth,both genes are epithelial (Kettunen and Thesleff1998). Growth factor Fgf8 has been proposed tobe involved in the early determination of thedental formula, because it can induce mesen-chymal expression of Pax9 and Msx1, both ofwhich are necessary for tooth development(Satokata and Maas 1994, Neubüser et al. 1997,

40

Bei and Maas 1998, Kettunen and Thesleff 1998,Peters et al. 1998), and Fgf4 has been proposedto be involved in crown morphogenesis, becauseit is a known mitogen and survival factor formouse dental tissues (Vaahtokari et al. 1996,Kettunen and Thesleff 1998).

The signaling molecule Shh has been shown tobe both a patterning factor and a mitogen presentin many organs (e.g., Bitgood and MacMahon1995, Bueno et al. 1996, Bellusci et al. 1997a,Oro et al. 1997, St-Jacques et al. 1998, Fan andKhavari 1999, Zúñiga et al. 1999). Its Droso-phila homologue Hh is essential for, e.g., seg-ment polarity and imaginal disc development(e.g., Fietz et al. 1994). In mammals it was firstdiscovered in the central nervous system (CNS),where it induces floor plate and motor neurondifferentiation (Echelard et al. 1993). DuringCNS development it is expressed both in themesenchymal notochord and in the epithelialfloor plate. In limb buds it is exclusively mesen-chymal in the zone of polarizing activity (ZPA).In teeth it has been shown to be exclusivelyepithelial, and involved in the proliferation ofepithelial cells (Iseki et al. 1996, Hardcastle etal. 1998, see below). It has also been implicatedin the budding of the individual tooth germs(Hardcastle et al. 1998). Epithelial Shh inducesits own receptor/inhibitor-subunit Ptc, in dentalmesenchyme, but this depends on the presenceof mesenchymal Msx1 (Murone et al. 1999,Zhang et al. 1999). This is an example of aninteraction between Bmp-, Fgf- and Shh-signal-ing pathways. Similar interactions probablyoccur between these pathways and the Wnt-path-way, because signal transduction in the Wnt-pathway requires TCF transcription factor (Kühland Wedlich 1997, Orsulic et al. 1999), and Shhhas been shown to inhibit the expression ofWnt10b (Dassule and McMahon 1998).

Transcription factor Pitx2 is also essential fortooth development. It is first expressed in thestomodeal endoderm and then in the future den-tal lamina and becomes limited to the epitheliaof the developing tooth germs (Mucchielli et al.1997). In humans, even Pitx2 haploinsufficiencycauses oligodontia, and in Pitx2-/- targeted mu-tant mice all the teeth are missing (Semina et al.

1996, Flomen et al. 1998, Lin et al. 1999, Lu etal. 1999).

Cyclin dependent kinase inhibitor p21CIP1/WAF1 isnot essential for tooth development (Deng et al.1995), but high levels of p21CIP1/WAF1 arrest thecell cycle (Harper et al. 1995, Harper and El-ledge 1996), and its expression in teeth has beenassociated with quiescent or slowly dividingcells (Bloch-Zupan et al. 1998). It has been pro-posed to be involved in the initiation of apopto-sis (Deng et al. 1995, Harper and Elledge 1996),but it is possible that it is also involved in theterminal differentiation of the dental epithelialcells, where it is expressed, possibly even pro-tecting them from apoptosis (Parker et al. 1995,Wang and Walsh 1996, Fan and Khavari 1999).Expression of p21CIP1/WAF1 can be induced inteeth by Bmp2 and –4 (Jernvall et al. 1998), andits expression patterns are regulated spatially andtemporally (Bloch-Zupan et al. 1998, Jernvall etal. 1998).

Signaling centers in tooth development

Since the shape of the tooth depends on its posi-tional identity, and since the positional identityof a tooth germ seems to be determined by mor-phogenetic fields before tooth development hasbeen initiated, one way to approach the problemof patterning is to analyze the role of morphoge-netic fields during the development of signalingcenters. The dental lamina can be considered asa tooth field, analogous to the presumptive limb-forming region in the lateral mesoderm of thevertebrate body (Vogel et al. 1996). The budsarise from a continuous lateral plate mesoderm,but the development of each bud is controlled byseparate signaling centers, which develop fromthe border between the dorsal and ventral epi-thelium and the posterior mesenchyme of eachbud (Tabin 1992, Altabef et al. 1997, Pearse andTabin 1998). Signaling centers, like AER oristhmus, control the local differences in differ-entiation and growth of tissues by providingspatially and temporally limited signals. There-fore, the first signaling centers in teeth ought tobe formed from the dental lamina during theearly bud, or dental lamina stages.

41

During tooth crown morphogenesis, the enamelknots have been proposed to be signaling cen-ters, because they are associated with the crownbase development and with the tips of each cuspin many mammalian species (Butler 1956, Jern-vall 1995). Since the teeth grow from the tipdown, the relative heights of the cusps dependon the initiation times of each cusp. Likewise,the relative locations and distances depend bothon the spatial patterning required for cusp devel-opment and on the growth rates of the intercuspepithelium. Hence, if enamel knots are essentialfor initiation of individual cusps, changes in thespatiotemporal control of enamel knot initiationand removal cause evolution of cusp patterns(Jernvall 1995, Jernvall 2000, Jernvall andThesleff 2000). The enamel knots had previouslybeen shown to express mitogen Fgf4 (Jernvall etal. 1994.). Hence, the possibility that the enamelknots could be signaling centers controlling thespecies specific crown morphogenesis wasanalyzed.

The epithelial growth downwards determines thedepths of the intercusp valleys and the totalheight of the crown (Butler 1956). The sharpnessand the distances of the cusps (and the crown)are determined by the relative growth rates ofthe epithelium and the mesenchyme. For exam-ple, for conical cusps to grow in height, the two-dimensional epithelium must grow slower rela-tive to the three-dimensional mesenchyme thanwhen growing prismatic cusps, which remainequally wide during growth (Appendix 1). How-ever, because mouse cusps are more complexthan simple cones (e.g., they are connected bylophs) and the prisms of sibling vole molarsmust also grow laterally, the complex mathe-matical solutions to cusp morphogenesis, letalone cusp placement, require spatial mappingand analysis of morphogenesis as a whole.

The growth rates of the dental tissues dependboth on induction-inhibition cascades and on theintrinsic rates of cellular maturation. Experi-ments with primary cell cultures have suggested

that each cell has an internal clock that tells itwhen to stop dividing (Gao et al. 1997). Afterthe last division, the cell withdraws permanentlyfrom the cell cycle for terminal differentiation orapoptosis, depending on its environment (e.g.,hematopoiesis in Gilbert 1997). The local induc-tion-inhibition cascades can modulate the fatesof the cells, and the cells can also retreat intotemporary quiescence, which can be reversed bylater signaling (e.g., Cornelison and Wold 1997,Seale and Rudnicki 2000). The AER is a signal-ing center consisting of terminally differentiatedcells, which are removed apoptotically afterAER is no longer necessary, and it has been pro-posed that apoptosis has an important role inregulating signaling center function (Vaahtokariet al. 1996). Primary enamel knots consist ofquiescent cells and are downregulated apoptoti-cally (Vaahtokari et al. 1996, Jernvall et al.1998), but some reports claim that they may splitto produce the secondary enamel knots for eachcusp (Coin et al. 1999). This would mean that atleast some of the enamel knot cells are not ter-minally differentiated but can change their fatesand begin to replicate again, forming theintercusp valleys.

The exact relationships between the primary andsecondary enamel knots or the enamel knots andcrown morphogenesis are unknown. The enamelknots’ role as inducers of cusp growth has beendebated (Butler 1956, Jernvall 1995, Lesot et al.1996, Lesot et al. 1999), and it has also beenproposed that they provide a fixed point forepithelial folding, although the theory of enamelknots as a source for new cells for cusp growthhas probably been abandoned (Butler 1956). Theenamel knots seem to appear sequentially, andthe exact rate and order of appearance correlateswith the relative heights of the cusps (Jernvall1995). Hence, understanding the relationshipbetween individual enamel knots as well asbetween the growth and the enamel knot patternformation is crucial for understanding theevolution of mouse and sibling vole molarmorphogenesis.

42

Aims

The ultimate aims of this study are to create a deeper understanding of some of the developmentaland molecular principles behind morphological evolution. The model system, the muroid tooth, waschosen firstly; because the mouse is the best known mammalian model in experimental develop-mental biology, secondly; because the tooth is morphologically a complex organ, which developsdeterministically and has great evolutionary diversity, and thirdly; because the molecular biology oftooth development has been extensively studied. The sibling vole was chosen as a comparisonbecause its teeth are morphologically very different from mouse teeth, although both animals aremuroid rodents and have similar sizes and gestation times. Since there are many unansweredquestions regarding morphogenesis and the molecular basis of tooth development, even in the mouse,and because practically nothing was known about the development of the sibling vole dentition, theproximate aims in this study were:

1# to map the basic tooth morphogenesis and the expression patterns of several candidatedevelopmental regulatory genes and pathways and to compare them between mouse and siblingvole teeth

2# to analyze the role of early epithelial patterning in tooth morphogenesis and specifically in thedevelopment of the toothless diastema region and disparate molar shapes in the mouse and siblingvole

3# to study the development and role of epithelial signaling centers in tooth development and in theevolution of species specific molar morphologies

4# to analyze epithelial-mesenchymal interactions in the formation of species specific cusp patternsby growth comparisons and in vitro recombination experiments

5# to provide a morphological and methodological basis for further comparative work betweenmouse and sibling vole teeth and the spatiotemporal analysis of pattern formation and toothmorphogenesis.

43

Results and discussion

The morphogenesis of mouse and vole teeth

The morphogenesis of the mouse and siblingvole embryonal dentitions, and especially themorphogenesis of M1 and the rudimentary dias-tema tooth germs in mouse and sibling vole wasanalyzed in this study from histological serialsections. The differences in morphogenesis werecompared in order to identify the processes onwhich evolution may have acted to produce spe-cies and identity specific morphologies in thesetwo species. After this, expression patterns ofvarious candidate developmental regulators canbe correlated with morphogenetic events frombefore the initiation stage to the end of differen-tiation in order to identify the potential signalingpathways that the heritable changes may haveaffected.

The dentitions of the mouse and vole(Articles II and III, and unpublishedobservations)

All mouse and vole tooth germs arise originallyfrom the temporary dental lamina or from theepithelium of the previous tooth germs. Thedental lamina begins to form from the thickenedoral epithelium first in molar and then in incisorregions during the initiation stage. Nevertheless,the dental laminae of each jaw quadrant are con-tinuous epithelial structures. The dental lamina ismorphologically visible throughout the upperdiastema region. In the lower diastema regionthe dental lamina is not morphologically visible,but can be visualized with genetic markers.

The M1 is the first tooth germ to become visiblein both species (E11,5), the next being theincisor (E12). The existence of three small rudi-mentary upper jaw diastema tooth germs (D1,D2 and D3) was already known in the mouse(Turecková et al. 1995), but the sibling vole sur-prisingly had only one rudimentary tooth germ(D) in its upper diastema region (Figure 10).These seem to be remnants of the primitiveplacental mammalian dental formula I3/3, C1/1,P4/4, M3/3. The rudimentary diastema tooth

germs form from anterior to posterior (E12 –E12,5). The M2 was formed in both species atE15. The development of M3 or potentialrudimentary incisors was not studied (Moss-Salentijn 1978, Witter et al. 1996), but in thecommon vole (M. arvalis) the M3 has beenreported to develop postnatally, and the same islikely in the sibling vole (Štorba 1981).

The sibling vole D was in the same relativelocation as the anteriormost mouse rudimentarytooth germ D1. The single upper diastema toothgerms of E13 bank voles (Clethrionomys rufo-canus), E13 common voles (Microtus arvalis)and E13 root voles (Microtus oeconomus) werein the same relative location and developed intolarge buds (not shown). The only diastema toothgerm that could be certainly identified in E15 rat(Rattus norvegicus) was much smaller, andlocated in the anterior diastema region (notshown).

Although the other tooth germs are initiatedfrom separate swellings, the M2s and M3s origi-nate from the budding epithelium of the poste-rior end of the molar anterior to them. Moreover,the initial swellings of the incisors and the rudi-mentary tooth germs are as wide along the dentallamina as their final buds, but the M1s are ini-tially much shorter and the epithelium continuesto invaginate into the mesenchyme posteriorlyfor several days. Since M2 and M3 arise fromthis posterior budding and since the initial M1swelling has been reported to disappear apop-totically, it has been suggested that (at least) theupper M1 is a composite structure and that theswelling actually belongs to some rudimentaryposterior premolar (Peterková et al. 1996).

All the tooth germs found are of the primarydentition, which is not deciduous but permanent.Although muroid molars would not have anyreplacement teeth even according to the basicplacental dental formula during their early bellstage (E16 in M1), the molars have a swelling in

44

A

B

C

D

E

F

G

H

I

J

K

L

M

N

Figure 12. Frontal serial sections of mouse and vole M1 development (A) E11 mouse (B) E12 mouse (C) E13 mouse (D) E14mouse (E) E15 mouse (F) E16 mouse (G) E17 mouse, (H) E11 vole, (I) E12 vole, (J) E13 vole, (K) E14 vole, (L) E15 vole,(M) E16 vole, (N) E17 vole. The arrow indicates the possible rudimentary secondary lamina in a mouse molar. Lingual is tothe left and buccal to the right. The scale bar is 150 µm.

their lingual outer enamel epithelium, which hasbeen proposed to correspond to a rudimentarysecondary lamina (Figure 12F; Gaunt 1966).

Rudimentary tooth germs in mouse and siblingvole upper diastema regions (Article II)

The mouse D1 and the vole D become visible atE12, whereas mouse D2 and D3 become distin-guishable from each other and the palatal rugaeat E12,5 – E13. Morphologically, the mouse D1and vole D are associated with the primarychoanae, but on the maxillary side, which makesthem the anteriormost teeth in the maxilla.Because the canines are the anteriormost teeth in

the maxilla in the full placental dental formula,these rudiments could be the rodent canines. Onthe other hand, some reports suggest that thelateral incisors (I3) may migrate from the max-illary side of the primary choana (for a review,see Schwartz 1982), which means that D and D1could also represent I3, as the single still existingrodent incisors have been identified as I2 (Luck-ett 1985). The mouse D2 and D3 develop inconnection with the palatal rugae, at the poste-rior end of the diastema lamina near the formingmolar buds, and are most likely to be premolars.The exact identities of D2 and D3 are stillunclear, because some reports state that M1 isactually a composite structure consisting of M1

45

and two premolars (Peterková et al. 1996, Lesotet al. 1998). The diastema buds do not elongatealong the dental lamina, but this may be normal,because the incisor bud is approximately as wideas the original swelling as shown bywholemount in situ hybridization of the geneexpression patterns in mouse and sibling voleupper jaws (Figures 2 and 4 in Article III).

The mouse diastema tooth germs are removedapoptotically one day after their appearance(E13 – E14). The sibling vole D disappearsapoptotically by E17. The apoptosis is seen onlyin the epithelium. In mice the apoptosis is seensimultaneously through the whole anlagen,whereas in voles apoptosis begins from the neckof the bud. The degeneration of the rudimentscan be predicted by their abnormal histodiffer-entiation. Unlike around the molars or incisors,the mesenchyme around the tooth germs doesnot condense more than under the oralepithelium (Figure 10).

Morphogenesis of mouse and sibling vole firstlower molars (M1s) (Articles II and IV,and unpublished observations)

Figure 12 shows the histological frontal sectionsof the embryonic lower first molars of the mouseand sibling vole from the initiation to the latebell stage. The lower first molars of the mouseand sibling vole develop at the same rate throughthe same developmental stages. The tooth germsare histologically similar, but their shapesbecome increasingly different from the early budstage onwards. This can be seen more clearlyfrom the separated dental epithelia of E12- toE13,5 mouse and sibling vole molars (Figure 13)and in the three-dimensional reconstructions ofthe basement membrane drawn from mouse andsibling vole E14, E14,5, E15, E16 and E17histological serial sections (Figure 14). As thebasement membrane separates the mesenchymeand the epithelium, its final shape before thedeposition of the predentin and the enameldefines the basic shape of the adult tooth (Butler1956). Hence the folds of the basement mem-brane are excellent landmarks for studying toothmorphogenesis.

The development of M1 becomes visible whenthe posterior mandibular oral epithelium thick-ens around E11. Some cells become pseudo-stratified, whilst cells next to the basementmembrane remain columnar. The maxillary endof the dental lamina will develop from thisthickening. The small, roundish epithelial swel-ling invaginates from the dental lamina into theunderlying mesenchyme in both species at E12-,marking the initiation of molar development (notshown).

During the early bud stage, the molar buds arewide and the initial swelling remains in theanterior lingual end of the E12 M1 tooth germs(Figures 12 and 13). The posterior elongation ofthe buds coincides with their narrowing. Duringthe bud stage, the sibling vole M1 buds elongatefaster than mouse M1 buds (Figure 13). Theswelling in the anterior lingual epithelium disap-pears in both species, and the mesenchyme con-denses around the developing molar bud. In bothspecies, the buccal mesenchyme condenses morethan the lingual mesenchyme.

The bud to cap stage transition begins in bothspecies with the formation of the primary enamelknot. The full length of the primary enamel knotat E14 is 160 µm in voles and 200 µm in mice.The initiation of cervical loops around the pri-mary enamel knot marks the beginning of thecap stage. The cervical loops define the futurecrown base by growing around the condensedmesenchyme under the tip of the bud (Figure12). From widening of the separated E13,5 epi-thelia (Figure 13) it can be seen that the cervicalloops first begin to grow around the anterior endof the primary enamel knot, which is also thefirst part of it to differentiate (see below). Thelingual cervical loop is initiated first. Since thelingual cervical loop also grows faster than thebuccal cervical loop, the tooth crown base has byE15 or early bell stage a distinct lingual bias(Figure 12E, 12L). By E15, most of the primaryenamel knot has in both species been removedapoptotically (not shown) or downregulated (seebelow). The remnant of the primary enamelknot, which was originally in the middle of thebud tip, is at E15 on the buccal side of the crown

46

D Shh mouseC p21 mouseA Shh vole

B Lef1 mouse

E13-

E13

E13

E13,5

E12-

E13,5

E12-

E13

E13

E13,5

Figure 13. Separated epithelia from mouse and vole M1s in temporal order from E12- to E13,5, showing the expression of Shhin the vole (A), Lef1 in the mouse (B), p21CIP1/WAF1 in the mouse (C) and Shh in the mouse (D). The tissues are arranged fromyoungest to oldest according to their morphologies, young up and old down. Though the vole molars are at E12- appr. as longas mouse molars (not shown), at E13 the vole molars (A) are longer than mouse molars (D). The mouse Shh expression isdivided into two columns. The epithelia are shown from antiocclusal direction. Anterior is to the left and buccal up.Arrowheads point at the initiating cervical loops during the bud to cap stage transition. The lingual cervical loop begin to growfirst. Arrows indicate the early epithelial signaling center. The scale bar is 200 µm.

47

(Figure 12L). These remnants in both speciesare associated with the protoconid, which beginsto grow after E15 (Figure 14).

After the bud to cap stage transition, the crownbase mesenchyme differentiates into the dentalpapilla and the columnar epithelium next to itbecomes the inner enamel epithelium (Figure12). The stratum intermedium and outer enamelepithelium, as well as the stellate reticulumbegin to differentiate. Enamel chords have beenseen both in the mouse and sibling vole(MacKenzie et al. 1992, not shown), extendingbuccally from the primary enamel knot to theouter enamel epithelium. The condensation ofthe dental papilla mesenchyme becomes slightlyunequal at the late cap stage. This inequalitybecomes more distinct during the bell stage.Mesenchymal differentiation, however, followsthe formation of epithelial enamel knots.

Geographic Information System (GIS) -analysisof three-dimensional reconstructions shows thatat E14 mouse and vole crown bases are equallylarge (Figure 14, Table 3). However, betweenE14 and E15, the vole molars elongate overthree times as fast as the mouse molars and theirsurface area, occlusal area and volume increaseover twice as fast as in mouse molars (Tables 3and 4). The growth rates then drop to about thesame as in mouse molars (Table 4), but thelength difference seen at E15 (60%) is not thesame as the final difference (92%) in M1 lengthsbetween the species. This means that the voletooth germs grow faster (or longer) than mousetooth germs before mineralization.

The ratios of epithelial surface area/occlusalarea, measured from three-dimensional recon-structions of mouse and sibling vole M1 crownsat one day intervals, remained constant in bothspecies (Table 3). However, the species-specificratios were different (average 1.27 for mouseand 1.52 for sibling vole M1). The relationshipbetween the mouse/vole crown volumes andocclusal areas are about 0.63±0.03 of therelationship between mouse/vole surface areasand occlusal areas at E14, E15 and E16. Thelatter values also define the occlusal complexity,

E14

E14,5

E15

E16

E17

E14

E14,5

E15

E16-

E16

E18

Mouse Sibling vole

Figure 14. Three-dimensional reconstructions of mouse andvole molar crown development at E14, E14,5, E15, E15,1,E16, E17 and E18 seen from the shape of the basementmembrane. Except for vole E14,25, vole E15,1, mouseE17,01 values and adult distances, the values in Tables 3, 4and 5 are from these reconstructions. The arrows point to thecenter of the primary enamel knot. Anterior is to the left andbuccal up. The scale bar is 200 µm.

which is a measure of the relative height differ-ences of the cusp and intercusp areas in theocclusal area. Because occlusal complexity isconstant, the increase in crown folding can bepredicted by calculating the occlusal complexityat the cap stage (Table 3). Moreover, it can beinferred that the shapes of the cusps depend onthe relationship between epithelial and mesen-chymal growth rates. The degree of foldingcannot, however, be used for predicting thepattern of folding (Appendix 1). It is notable thatthe shapes of the second and third molars of themouse or vole more resemble the first molars ofthe same species than the second or third molarsof the other species (Figure 1 in Article II). Thismight be explained if the constancy of speciesspecific relationships between epithelial andmesenhymal growth rates in M1 could also beapplied to the development of other molars.

48

Table 3. The surface areas (S) and occlusal areas (Occl.) and occlusal complexity (S/Occl.) in mouse (m) and sibling vole (v)M1 at different stages from E14 to E18. The average S/Occl. for mouse is 1.27 and the average S/occl. for sibling vole is 1.52.

Species Age Surface area (µm) Occl area (µm) S/Occl. Volume

m 14.00 33480.14 27344.16 1.22 985336

m 14.50 29704.60 25320.48 1.17 574774

m 15.00 66325.93 53982.48 1.23 2856752

m 16.00 117260.65 89817.12 1.31 6382404

m 17.00 188802.01 146867.76 1.29 11029768

m 17.01 208021.92 152184.00 1.37 18342737

v 14.00 40936.61 27776.64 1.47 1821869

v 14.25 54938.29 37099.44 1.48 2727179

v 15.00 138964.62 83431.92 1.67 10890368

v 15.10 134075.51 89237.76 1.50 9244139

v 15.75 169676.55 107357.04 1.58 15522754

v 16.00 225788.28 161878.08 1.39 19514402

v 18.00 416370.97 272282.88 1.53 30966853

_________________________________________________________________________________

Table 4. Differences between the stages in mouse and sibling vole teeth. The size of the later stage from Tables 3 and 5 isdivided by the size of the earlier stage. For surface area, occlusal area and volume, both E15 and E15,1 vole values were used,thus giving a range between which the actual values may vary.

length growth surface growth occlusal growth volume growthm14-m15 1.38 1.98 1.97 2.90m15-m16 1.23 1.76 1.66 2.23m16-m17 1.07 1.61 1.64 1.73v14-v15 2.17 3.28 – 3.39 3.00 – 3.21 5.07 – 5.98v15-v16 1.20 1.62 – 1.68 1.81 – 1.94 1.79 – 2.11

_________________________________________________________________________________

Table 5. The lengths of the crown base (L) and the distances between the approximate center of the protoconid enamel knotand the anterior edge of the crown (AntP) in E14 – adult mouse (M) and sibling vole (V), and the protoconid – anterior edgedistance divided by the crown length. Note that the tip of the mouse cusp is always a bit posterior to the enamel knot, and thiscan distort the adult measurement. The distances are measured in micrometers and the relationships are given as percentages.

L (µm) AntP (µm) AntP / L (%)

M14 265 - -M15 365 135 37.0M16 450 150 33.3M17 485 170 35.0M(adult) 1480 560 37.8V14 270 - -V15 585 400 68.4V16 700 430 61.4V18 905 575 64.5V(adult) 2840 1860 – 1900 65.5 – 66.9

49

p

m e

a

pd

h1

2 4

3

5

6

7

8

9

10

1

2

3

4

5

6

7

8 9

10

p

m e

a

pd

h1

2 4

3

5/6

6/5

7

1

2 3

4

5/6

6/5

7

a pb

l

sibling vole

mouse

Figure 15. The evolutionary (in italics) and developmental(in bold) orders of cusp appearance in mouse and vole mo-lars. a anteroconid, a anterior, b buccal, e entoconid, hhypoconid, l lingual, m metaconid, p posterior, pa para-conid, po posteroconid or posterolophid pr protoconid. Thescale bar is 1 mm.

In adult molars, the cusps form a buccal and alingual row, which extend posteriorly and ante-riorly from the protoconid and the metaconid. Inmouse and sibling vole molars, new cusps arealways initiated at the ends of the rows (Figures15 and 16). Although the mouse anteroconidseems to consist of two cusps, its mesenchymalcomponent actually consists of three cuspules,which partially fuse during mineralization (notshown). Since the sizes of and distances betweenthe cusps depend on their growth rates and thetimes of cusp development, the last formingcusps at the ends of the rows are smaller andcloser together than the earlier forming cusps.Hence, the cusp rows converge at their ends.

The number of cusps in a row depends on therate of cusp initiation and on the time windowfor initiating new cusps. The rate of cusp initia-tion is faster in voles than in mice. As seen fromthe three-dimensional reconstructions of devel-oping mouse and vole M1s, all the cusps can beseen in E17 vole M1 (Figure 16), whereas inmouse M1 the last cusp begins to grow at E18(not shown). Cusp sizes seem to be more equal

in vole M1s, where the cusps are initiated closerto each other (Figures 9 and 15).

Each cusp is associated with its own, histologi-cally distinct, enamel knot (Figures 12 and 16).In the forward slanting separate mouse cuspsthey are roundish, separate structures on theanterior side of the cusp tip. In vole cusps, theinitially roundish enamel knots are remodeled.They become histologically less distinct andwiden laterally with the anterior edges of theprisms, fusing centrally with each other. Theremodeling of the vole enamel knots correlatesboth with the lateral growth of the prisms (seebelow), and the development of the central ridge(Figure 16).

Mouse Sibling voleA

B

C

D

E

F

Figure 16. Three-dimensional reconstruction of Fgf4 exp-ression in the enamel knots of the bell stage mouse and voleM1s. (A) E15 mouse, (B) E16 mouse, (C) E17 mouse, (D)E15 vole, (E) E16 vole, (F) E17 vole. The scale bar is 100µm.

50

The basic developmental pathways are conserved in the teeth of mouse and sibling vole

The differences between mouse and vole denti-tions arise either from qualitative differences inthe tooth developmental program, or because theprogram is controlled differently in mouse andsibling vole. To confirm or reject these hypothe-ses, the expression patterns of several genesbelonging to different signaling pathways werecompared in the embryonic tooth germs of bothspecies. The selection included the secreted sig-naling molecules, Bmp2, Bmp4, Fgf4, Fgf8 andShh, as well as transcription factors Lef1, Msx1,Msx2, Pax9 and Pitx2, and a cyclin dependentkinase inhibitor, p21CIP1/WAF1. Most of thesegenes are known to be important for toothdevelopment in mice (see Tables 1 and 2). How-ever, it was not certain if these genes have simi-lar effects in other species. Haploinsufficiency ofMsx1, e.g., leads to missing teeth in humans, butit has no detectable effect on tooth developmentin mice (Satokata and Maas 1994, Vastardis etal. 1996, Hu et al. 1998). Hence, to understandthe developmental basis for morphologicalchange it was necessary to assess the degree ofconservation of mouse and vole tooth develop-mental programs. After this, the evolutionaryimportance of the individual genes and pathwayscan be further evaluated.

The gene sequences are conserved betweenmouse and vole (Articles II and III,unpublished observations)

All antisense-RNA probes made from mousecDNA sequences hybridized specifically to voletissues. The genes were expressed at similarstages in the same cell types in the mouse andvole, although the exact morphologies of thetissues were different. Paralogous genes be-longing to the same gene family were expresseddifferently when compared to each other, but thedifferences were similar in both species. Becausethe sequences of the orthologous genes in thetwo species are closer to each other than thesequences of the paralogous genes in the samespecies, this proves that the probes hybridizedspecifically. However, some probes hybridizedless efficiently in vole than in mouse tissues,

probably because of the sequence differences,whereas Msx2 and Wnt6 (not shown) actuallyseemed to hybridize more strongly in vole thanin mouse tissues. It is not known whether this isan artifact or represents genuinely strongerexpression in voles than in mouse. This questioncan be solved by cloning the sibling vole Msx2and Wnt6 sequences from corresponding areasand by doing a comparative in situ hybridizationwith both mouse and vole probes and tissues. Incontrast, of the probes used only Shh, which atthe protein level is 96% identical between ratand mouse and 84% between rat and chicken(Roelink et al. 1994), worked in opossumtissues. Lowering the hybridization stringencydid not improve the results. One reason for thiswas probably that whilst it increased thehybridization between less identical RNAsequences, it also increases the unspecificbackgound. Also, it is possible that the treatmentfor post-natal opossum tissues might be differentfrom the treatment of embryonic tissues. Thisindicates that the optimization of conditions maytake time, and that although cross-species in situhybridization can be a useful method forcomparing closely related species, it does notnecessarily work for more distantly relatedspecies. Hence, it is advisable to test each newprobe even in closely related species.

The hybridization efficiency of the probes alsovaried according to their length and generalspecificity. Because the digoxygenin moleculesare very large, whereas 35S-UTP molecules aresmall, fewer digoxygenin molecules than 35S-UTP molecules could be incorporated in thesame length of RNA. Since each antisense-RNAbinds into one mRNA molecule, the stainingefficiency is lower with digoxygenin labeledprobes than with the radioactive probes. Hence,as a rule, shorter probes could only be used forradioactive in situ hybridization, whereas longerprobes could also be used for wholemount in situhybridization. Some probes were associated withmore background signal than others, and thismay reflect either low levels of backgroundexpression or unspecific hybridization into a

51

Lef1 p21 Shh Fgf4

E16

E14-

E14

E14,5

E15

E15,5

E16,5

Figure 17. Small time interval comparisons of the expression patterns of Fgf4, Lef1, p21CIP1/WAF1 and Shh during mouse molarcrown development. The growth factor Fgf4 is the last gene to appear and it is only expressed in the enamel knots, which de-velop on the common overlap areas of the other genes The tissues are arranged from youngest to oldest by shape, the youngestbeing above. Tissues of equal developmental stage are at the same level. Anterior is to the left and buccal is up. The scale baris 200 µm.

52

distantly similar, ubiquitously expressed RNA.The signal strengths varied from one reaction toanother, which means that in situ hybridizationcan only be used as a semiquantitative methodfor detecting local patterning and for assessingthe relative strength differences in gene expres-sion within a sample.

The genes are expressed similarly in mouse andvole molars (Articles I, II, III and IV,unpublished observations)

To see if the usage of the signaling pathwaysduring tooth development is conserved, theexpression patterns of the studied genes wasmatched roughly to the tissues and the develop-mental stages in the M1s of the two species.Since the studied genes represent several par-tially dissociable pathways, it was thus also pos-sible to see which pathways might be affectedduring the morphological divergence of mouseand sibling vole. The spatiotemporal correlationsbetween the gene expressions and variousstructures in the developing molars of mouse andsibling vole are shown in Table 1 in Article II.

Although the radioactive in situ hybridizationwas more sensitive for detecting weak signals,the wholemount in situ hybridization was moreuseful for visualizing the three-dimensionalrelationships of the expression patterns and mor-phogenesis. Wholemount in situ hybridizationwas especially useful for the analysis of earlyexpression of developmental genes during theinitiation stage. By separating dental epithelia,wholemount in situ hybridization could be usedfor small time interval analysis of the formationof the enamel knots during the cap and bellstages. Figures 2 and 4 in Article III show theexpression patterns of Bmp2, Pitx2 and Shhduring the formation of the individual toothgerms in mouse and vole upper and lower jawsand Figure 17 shows how the expression patternsof Lef1, Shh, Fgf4 and p21CIP1/WAF1 changeduring the formation of the enamel knots inmouse molars.

Figure 4 in Article II and Figures 5 and 6 in Ar-ticle III show the expression patterns of tran-

A B

Figure 18. The frontal section of opossum (Monodelphisdomestica) tooth germ (A) and palatal ruga (B) showing theexpression of Shh. The scale bar is 100 µm.

scription factors Lef1, Msx1, Msx2 and Pax9 andcyclin dependent kinase inhibitor p21CIP1/WAF1 inthe frontal sections of mouse and vole molars.All these genes are expressed in similar tissuesand at similar stages in both species. The tran-scription factors Msx1 and Pax9 are seen only inthe mesenchyme, but Lef1 and Msx2 areexpressed also in the epithelium, whereasp21CIP1/WAF1 is seen clearly only in the epithe-lium. Figure 3 in Article II and Figures 4 and 6in Article III show the expression patterns of thesecreted signaling molecules Bmp2, Bmp4, Fgf4,Fgf8 and Shh in the frontal sections of mouseand vole molars. After the initiation stage, Bmp4is primarily expressed in the dental mesen-chyme, but the rest of the signaling moleculesare expressed in the epithelium. Since Shh isexpressed in similar locations (enamel knots andpalatal rugae) as in the mouse and vole even inthe opossum (Monodelphis domestica) (Figure18), the (tooth) developmental programs seem tobe very conserved. On the other hand, the voleM1 anterior and lingual swellings expressed Fgf4and Fgf8, whereas mouse swellings did not,suggesting that the developmental programs canalso change. This was the only qualitative geneexpression difference detected between mouseand vole molars, and its morphogenetic signifi-cance remains to be studied.

At all stages, the mesenchymal gene expressionpatterns were quite uniform through the dentalmesenchyme in both species. The main mesen-chymal patterning seen was buccal lingualasymmetry or dental identity specific differences(Table 1 in Article II). However, the epithelialexpressions were often localized to distinct partsof the epithelium (Table 1 in Article II). The

53

limitation of gene expression in the epitheliumwas often correlated with the differentiation ofameloblasts and stellate reticulum (Figures 3 and4 and Table 1 in Article II), but at various devel-opmental stages there were areas of epithelium,which co-expressed signaling molecules (e.g.,Bmp2, Bmp4, Bmp7, Fgf8, Fgf4, Shh andWnt10a). This co-expression was associatedwith early epithelial swellings of the tooth germsor in the primary and the secondary enamelknots (Table 1 in Article II, Figures 13, 17 and19). The correlation between differentiation ormorphology and gene expression was similar inboth species, but the shapes of the expressionpatterns within the epithelium were different.Thus, the potential patterning controlling themorphological differences was associated withthe epithelium.

The gene expression patterns are also visible inthe tooth gene expression database(http://honeybee.helsinki.fi/toothexp/).

E12 E14,5 E16

Figure 19. The expression of Wnt10a in the placodes andcrown areas of the mouse molars. During crown morpho-genesis Wnt10a is expressed in a wide area overlapping theenamel knots and in part also the other genes. The scale baris 200 µm.

Cross-species recombinations of mouse and voletissues can produce a tooth (Unpublishedresults)

Although the studied genes represent severalsignaling pathways, many other essential path-ways may be omitted. However, during recom-binations each tissue reacts according to its spe-cies specific genome to the signals expressed bythe other tissue with a different genome, whichmeans that the analysis of the results is based onall signals during morphogenesis. Hence, themost impressive proof of the conservation oftooth developmental pathways was the teethproduced by cross-species tissue recombinations.

Initiation stage (E11) or cap stage (E14) mouseand sibling vole M1 dental epithelia wererecombined with dental mesenchymes of thesame age from the other species. In this waymorphogenesis could be separated from thequestion of identity (see below). Four kinds ofrecombinations were done in both age groups,mouse epithelium – mouse mesenchyme, mouseepithelium – vole mesenchyme, vole epithelium– mouse mesenchyme and vole epithelium –vole mesenchyme.

Teeth with conical cusps and roots were identi-fied as mouse molars, whereas rootless teethwith prismatic cusps were identified as volemolars. The exact arrangement of the cusps inrelation to each other was usually abnormal.However, within these parameters, controlmouse – mouse recombinations always producedmouse molars and vole – vole recombinationsvole molars. Some of the cross-species recombi-nations, especially E11 vole epithelium – E11mouse mesenchyme were unsuccessful, produc-ing hairs instead of teeth or intermediatemorphologies (not shown). Hairs, like teeth, areepithelial mesenchymal appendages, and manyof the same genes are involved in the develop-ment of both teeth and hairs (e.g., Botchkarev etal. 1999, St-Jacques et al. 1998, Chiang et al.1999). Hence, it is possible that the early pertur-bations in the tooth developmental program leadto hair development instead. It is also possiblethat in addition to the dental epithelium, theearly mandibular epithelium included future hairproducing area, and that although teeth failed todevelop, the hairs did not. However, most of therecombinations produced histologically normalteeth (Figure 20) with recognizable speciesspecific cusp morphologies.

The successful recombinations between E11mouse epithelium and E11 vole mesenchymeproduced mouse molars, and E11 vole epithe-lium recombined with E11 mouse mesenchymeproduced vole molars (not shown). On the otherhand, the recombinations between E14 mouseepithelium and E14 vole mesenchyme producedvole molars (Figure 20), whereas recombinationsbetween E14 vole epithelium and E14 mouse

54

mesenchyme produced mouse molars. Even thesecond molar, which was often formed, showedthe mesenchymal identity of shape (not shown).Because in early recombinations the epitheliumdetermined the tooth shapes, whereas in laterecombinations the shape was controlled by themesenchyme, morphogenesis did not depend onthe species specific sequences of the extracellu-lar matrix or signaling molecules produced bythe epithelium or mesenchyme. Moreover, sincein late recombinations the control of morpho-genesis was mesenchymal and the mesenchymalgene expression is not spatially patterned untillater during morphogenesis, the crown growthand enamel knot pattern formation is likely to becontrolled by general mesenchymal induction,

which acts on all parts of the epithelium.

Hence, both mouse epithelium recombined withvole mesenchyme and vice versa produced teeth,which often have cusp shapes and roots thatrecognizably belonged either to a vole or to amouse. Even when the teeth were malformed,the tissues were arranged as in a normal tooth.This means that, even though the individualprotein sequences have probably divergedslightly, the genes and signaling pathways usedfor tooth development are the same in both spe-cies. Moreover, because teeth with discerniblemorphologies are formed, not only the pathwaysessential for histodifferentiation, but also thosefor pattern formation are conserved.

B

C

D

E

F

G

H

A

Figure 20. Recombinations between E14 mouse and E14 vole molar tissues cultured in nude mouse kidney capsules. WhenE14 mouse epithelium was recombined with E14 vole mesenchyme, vole-like teeth developed (A, B), whereas when E14mouse epithelium was recombined with E14 vole mesenchyme, mouse-like teeth developed (C, D). Control recombinationsE14 mouse epithelium and mouse mesenchyme produced mouse-like (E, F) and E14 vole epithelium and vole mesenchymeproduced vole like teeth (G, H). The scale bar is 1.5 mm for B and G, and 1.0 mm for A, C – F and H.

Epithelial signaling centers

In both mouse and sibling vole, the areas ofBmp2, Bmp4, Fgf4, Fgf8, Lef1, Msx2,p21CIP1/WAF1, Shh or Wnt10a co-expression cor-relate with the development of dental formulas,the bud to cap stage transition and the initiationof individual cusps. Cap and bell stage co-ex-pression correlates with primary and secondaryenamel knots (Table 1 in Article II), whereas theearly co-expression was seen in the initialepithelial invaginations of the individual tooth

germs. Because pattern formation depends onlocal gene expression differences, and these ar-eas co-expressed many secreted signaling mole-cules, it is likely that these areas are signalingcenters controlling tooth morphogenesis. Theenamel knots are histologically visible evenwithout gene expression data, and hence theirrole in tooth morphogenesis has been debated(Butler 1956). The existence of the early epithe-lial signaling center, however, had not previ-

55

ously been recognized because it is not his-tologically detectable without gene expressiondata (Figures 10, 12 and 13).

Enamel knots as epithelial signaling centers (Articles I and II, unpublished observations)

The in situ hybridization analysis of mouse mo-lars showed that the enamel knots express mostof the studied secreted growth and differentia-tion factors. These molecules are also expressedin other signaling centers, like ZPA and AER inlimbs, suggesting that enamel knots are signalingcenters. The primary enamel knot began todevelop before the bud to cap stage transition, asthe tip of the molar bud epithelium began toexpress p21CIP1/WAF1, Lef1 and Shh (Figure 13).The p21CIP1/WAF1 expression probably correlateswith epithelial quiescence, because in mousemolars, the p21CIP1/WAF1 expressing area does notshow BrdU labeling, which correlates withmitotic activity (Figures 1 and 2 in Article I).The development of the primary enamel knotbegan from the anterior end of the bud. Thisprobably correlates with the later initiationpatterns of cervical loops (see above).

The in situ hybridization comparison of the pri-mary and secondary enamel knots of the twospecies showed that they expressed Bmp2,Bmp4, Fgf4, Lef1, Msx2, p21CIP1/WAF1, Shh andWnt10a (Figures 3 and 4 in Article II, Table 1 inArticle II, Figures 13, 17, 19 and 21). Of thesegenes, only Fgf4 was limited to the enamelknots, whereas the others had wider and partiallyoverlapping expression areas, which changedconstantly. The expression of Bmp7 in mouseprimary enamel knot was noted, but not furtherinvestigated (Figures 1 and 2, Article I).

Epithelial Lef1 has been shown to be essentialfor the cap stage transition, but dispensable forlater morphogenesis (Kratochwil et al. 1996).Because Lef1 is strongly upregulated in allenamel knots of both species (Figure 4 in ArticleII and Table 1 in Article II, Figure 17), it may beessential for the differentiation and function ofthe primary enamel knot, but not the secondaryenamel knots. Moreover, the primary enamel

knot and the first secondary enamel knots ofboth mouse and sibling vole express Bmp2, butthis expression is lost at E15, after the bell stagetransition is over (Table 1 in Article II). Thus, itis associated mainly with the primary enamelknot. Hence, the primary and secondary enamelknots may be quantitatively different, althoughthe gradual loss of Bmp2 expression from theprimary enamel knot and the presence of Bmp2in the first secondary enamel knots of bothspecies suggest that the differences are stagerather than signaling center specific.

The appearances of Bmp4 and Msx2 have beenassociated with the apoptotic removal of enamelknots (Article I, Jernvall et al. 1998). Theapoptotic remodeling of the primary enamel knotoccurs in both species (not shown). BecauseAER is also removed apoptotically (Vaahtokariet al. 1996), apoptosis may be an importantmechanism for controlling the spatiotemporalpattern of signaling center function.

Early epithelial signaling centers (ArticlesII and III, unpublished observations)

Whilst the evolution of species specific cusppatterns is realized by changes in the latemolecular pattern formation, the recombinationexperiments (see above) indicate, that also theearly epithelial patterning events are important.More detailed expression comparison betweenthe species showed that the initial swellings ofthe budding tooth germs co-expressed Bmp2,Bmp4, Lef1, p21CIP1/WAF1, Shh and Wnt10a (Fig-ure 13 and 19, Figures 3 and 4 in Article II, Fig-ures 4, 5 and 8 in Article III). This suggests thatthey also consist of quiescent or slowly dividingcells. It further indicates that it is a local sourceof various secreted signals. The expressions ofBmp4 and Bmp2 have been proposed to beinvolved in the apoptotic removal of these struc-tures (Turecková et al. 1995, Peterková et al.1998). Although our TUNEL stainings could notdetect any apoptosis in the initial swellings ofmolars of either species (not shown), as reportedin mouse molars (Peterková et al. 1998, Viriot etal. 1998), the signaling area became smaller and

56

disappeared in both species, suggesting at leastdownregulation of gene expression.

Because all the initial swellings of the toothgerms budding from the dental lamina expressedat least a subset of these molecules, these areasprobably are early epithelial signaling centers,analogous to enamel knots. As the differentiationof the early epithelial signaling center coincideswith the local invagination of the epithelium, itis likely that the signaling from the initialepithelial swellings are involved in the determi-nation of tooth locations and the budding of thetooth germs. The same has been proposed onpurely morphological grounds by Westergaardand Ferguson (1986, 1990) regarding earlyalligator tooth germs.

It is interesting to note that in the sibling vole theM1 (but not the diastema bud) early epithelialsignaling centers expressed Fgf8 and Fgf4,whereas in the mouse none of the studied earlyepithelial signaling centers expressed thesegenes (Figure 3 in Article II, Figure 6 in ArticleIII). This expression continued in the vole (al-though in a diminishing area) until the primaryenamel knot was induced. Therefore, the earlysignaling center of vole M1 seems to be qualita-tively different from that of mouse M1. In addi-tion, the early epithelial signaling centers of therudimentary diastema tooth germs appear to bequalitatively different from the molar signalingcenters (Table 6 in Article III). These differencesbetween the rudimentary tooth germs andmolars, however, may be related to the develop-mental arrest of the rudimentary tooth germs.

Conservation of signaling center programs(unpublished observations)

Since the same genes and pathways are usedrepeatedly in the signaling centers of differentorgans and even in the different signaling centersduring tooth development (Figure 21), it seemsthat whenever a few robust modules for a “sig-naling center” are induced, a new signaling cen-ter is generated and it can organize development.The same genes are also involved in the pat-terning of other epithelial-mesenchymal append-

ages like hair and feather placodes (St. Jacqueset al. 1998, Jung et al. 1998) and the branchingmorphogenesis of the lung (Figure 2B; Bellusciet al. 1997a, b). It has been hypothesized that asthe co-option of conserved autoregulatory cas-cades is facilitated by their dissociability, thelikelihood that they become significant molecu-lar mechanisms for controlling organogenesisincreases, whereas the molecular interactionswithin the signaling cascades become more con-served (Gerhard and Kirschner 1997, Kirschnerand Gerhard 1999). Thus, hierarchical modular-ity (Figure 4) simultaneously increases theevolvability of morphogenetic mechanisms andthe conservation of developmental signalingpathways.

Bmp2Bmp4(Fgf4)Fgf8Lef1p21Shh

Bmp2Bmp4Fgf4Lef1p21Shh

Bmp4Fgf4Lef1p21Shh

early epithelialsignaling center

primaryenamel knot

secondary enamel knot(s)

Figure 21. Comparison between the expression profiles ofthe analyzed genes in placodes, primary enamel knot and thesecondary enamel knots. The Bmp2 expression is down-regulated in the secondary enamel knots after E15, whereasthe mouse placodes do not express Fgf4 or Fgf8. The gradualshift in expression profiles between vole placode andprimary enamel knot and the mouse and vole primary andsecondary enamel knots (not shown) suggests that thecombinations of placodal and enamel knot signals depend onthe current differentiation status of the tissues.

Because multiple signaling pathways often formautoregulatory loops (Table 2, see Literaturereview), the induction of an autoregulatory sig-naling cascade in a new developmental contextis possible. Thus, interconnectivity has probablyfacilitated the developmental co-option of thesame signaling center programs for differentpurposes during tooth development. The ob-served rapid change and variability in cusp pat-terns (Guthrie 1965, Martin 1993, Jernvall 2000)suggest that the cusp determination program canbe activated or deactivated easily, which impliesthat only a few controls are required at any par-

57

ticular location to switch the program on. As anexample of this, because the cusps can evolvefrom different morphological features, the localconditions for secondary enamel knot different-iation are likely to have arisen from differentinitial combinations of signaling pathwayactivators. Hence, although the developmentalorigin of the cusps is variable, it seems that asimilar signaling center can be easily applied indifferent developmental contexts by chance co-expression of necessary signaling pathways.Thus, new morphological features can evolve byan initially random overlap pattern of a fewsignaling modules, which create a heterotopicsignaling center. This produces new local signalswhich can be interpreted by the target tissuesaccording to their developmental context.

The autoregulatory interactions mean that once asignaling center is activated, it is likely to beself-maintaining. Because the pathways areinterconnected according to similar rules in allsignaling centers, “gratuitous” gene expressionis also possible. The expression of Lef1 is e.g.,only essential for the transition from the bud tocap stage, although it is expressed also in theearly epithelia signaling centers and the secon-dary enamel knots.

Different signaling centers can have similar pat-terning functions. In both the mouse and thesibling vole, the M1 primary enamel knot isinduced into the tip of the bud after the earlyepithelial signaling center has become com-pletely downregulated (Figure 13). The correla-tion between gene expression patterns and mor-phology in mouse molars (Figure 17) indicatethat the first signaling center found in M2 can beclassified as its primary enamel knot, as it isassociated with the beginning of the cap stage.

The same is also likely in the vole. As M2 doesnot have a separate early epithelial signalingcenter, both early epithelial signaling centers andprimary enamel knots are hence involved indetermining the dental formulas along the dentallamina and structures derived from it. Thoughnot verified with small time interval analysis,E13 Bmp2 and Shh expression in incisors, butnot in molars, suggests that (at least) in muroidrodents, the early epithelial signaling centers andthe primary enamel knots form one continuoussignaling center in incisors, unlike the situationin molars (see, however, Kieffer et al. 1999).

If the first signaling center, which defines thelocation of the tooth germ, also induces itscrown base, it is possible that the primaryenamel knots of the mouse and sibling vole M1sare their first actual signaling centers. If this isso, the early epithelial signaling centers in theanterior ends of the molar buds would actuallybe signaling centers for rudimentary premolars,which never develop beyond the early bud stage(Peterková et al. 1996, Lesot et al. 1998). Like-wise, since the muroid molar primary enamelknots develop into the secondary enamel knotsof the protoconid, it is possible that, primitively,the dental signaling centers of unicuspid teethwere early epithelial signaling centers, whichlater differentiated into enamel knots. Hence, theearly epithelial signaling centers, the primaryenamel knots and the secondary enamel knotsare all likely to be copies of the same primitivedental signaling center, their morphogeneticfunctions only depending on context. The laterevolution of multicusped teeth would then havebeen accomplished by evolutionary innovation,which would have enabled the induction ofadditional dental signaling centers in the sametooth germ.

Early epithelium and dental formulas

Before the actual tooth morphogenesis begins,the future tooth forming areas can be visualizedwith expression of several epithelial genes, firstBmp4, Fgf8, Lef1, Msx2, Pitx2 and Shh, thenBmp2 and p21CIP1/WAF1. Of these genes, at leastBmp4, Fgf8, Msx2, Pitx2 and Shh have been

shown to be involved in initiation of tooth germs(Table 1). The expression patterns of the restalso suggest a function in the early developmentof teeth (see above). Many of these genes(Bmp2, Lef1, p21CIP1/WAF1, Pitx2 and Shh) areinitially expressed in the dental lamina or its

58

derivatives, whereas others (Bmp4, Fgf8 andMsx2) may have wider expression areas. Bycomparing their expression patterns with toothmorphogenesis in the molar and diastemaregions, it is possible to assess their possibleroles in development and the evolution of dentalformulas.

Individual tooth germs develop from acontinuous dental lamina (Article III)

The sites of budding tooth germs are marked bythe early epithelial signaling centers, whichdifferentiate from the dental lamina after E11.Although the dental lamina may be morphologi-cally indistinct (e.g., in the mandibular diastemaregion of mouse), it can be visualized withmarker genes like Pitx2 or Shh (Figures 2 and 4Article III), see also Dassule and McMahon1998) The first gene to be expressed in the fu-ture dental lamina is Pitx2, which is seen in thestomodeal epithelium as early as E8,5 (Muc-chielli et al. 1997). The expression of Pitx2 iscontinuous and soon followed by dynamicBmp4, Fgf8, Lef1, Msx2 and Shh expression inpartially overlapping areas. The expressions ofBmp2 and p21CIP1/WAF1 follow later, when theindividual tooth germs are being formed.

During initiation of individual tooth germs, thepreviously continuous dental lamina expressionof genes like Shh (Figure 4 in Article III) andLef1 (not shown) becomes limited to small areasof lamina and is upregulated there. However,Pitx2, is at least in molars and incisors expressedin the dental epithelium outside the actualsignaling center area (Figure 2 in Article III).The gene expression patterns indicate that therudimentary diastema tooth germs also havesignaling centers (Figures 4, 5, 8 and 9 in ArticleIII). The formation of the dental lamina and theupregulation and complementary downregula-tion of genes into the early signaling centers ofeven rudimentary tooth germs suggest thatinduction-inhibition loops define the locations ofindividual tooth germs (see below). Because thelimitation of expression occurs within theepithelium, the determination of dental formulasis likely to depend on early epithelial signaling.

However, the phenotypes of ActivinβA-/- orMsx1-/-/Msx2-/- targeted mutant mice suggestthat early tooth development may also involvepermissive signals from (epithelially induced)dental mesenchyme.

Early epithelium controls the positionalidentities of the tooth germs (Article III,unpublished results)

To see if the epithelium controls the identity ofthe tooth germs, the E11 incisor-molar recom-bination experiments of Kollar and Mina (1991)were repeated. The recombinations betweenincisor epithelium and molar mesenchymeproduced incisiform teeth, whereas reciprocalrecombinations produced molariform teeth (notshown). These results support reports, whichsuggest that the early oral epithelium determinedthe tooth type (Lumsden 1988, Kollar and Mina1991). Therefore, it can be concluded that thepositional identities of the tooth germs aredetermined before their initiation by epithelialsignals.

Pax9

Bmp4

Fgf8

Msx1

Bmp4

Msx2

Lef1

Msx1

Bmp4

epithelium

mesenchyme

Anteriormaxilla Posterior

maxilla

Figure 22. The likely disturbance mechanism of mesen-chymal Bmp4-signaling pathway in the vole diastemaregion. The Bmp4 expression is relatively stronger in theanterior maxilla than in the posterior maxilla or premaxillain both species. The inhibition of Pax9 expression by Bmp4leads to the downregulation of mesenchymal Bmp4 sig-naling and subsequent loss of Bmp4-dependent transcriptionfactors The schematic diagram has been adapted from Chenet al. (1996), Neubüser et al. (1997) and Peters and Balling(1999).

59

Table 6 Comparison between expression of the analyzed developmental genes in epithelia and the mesenchymes in develop-ing diastema tooth germs (upper rows) and molars (lower rows) of sibling vole. ++ indicates strong expression, + moderate,(+) weak but detectable, – indistinguishable from the basal expression or background, NA not analyzed. Note that the correla-tions with the morphological features were not analyzed, because the diastema tooth germs are so simple.

Epithelial expressions:E11 E12 E13 E14 E15

Bmp-2 db + -(+) + - NAmolar + ++ (+) +(+) +

Bmp-4 db + - (+) (+) NAmolar + + - ++ ++

Fgf-4 db NA - - - -molar NA (+) + ++ ++

Fgf-8 db +(+) + - NA NAmolar ++ ++ (+) NA NA

Shh db +(+) ++ ++ (+) -molar ++ ++ (+) ++ ++

Lef-1 db ++ ++ ++ (+) (+)molar ++ ++ ++ ++ ++

Msx-2 db NA + + + NAmolar NA + ++ ++ NA

p21CIP1/WAF1 db NA (+) + (+) (+)molar NA + (+) ++ ++

Mesenchymal expressions:E11 E12 E13 E14 E15

Bmp-4 db + + (+) (+) -molar + ++ ++ ++ ++

Lef-1 db + + (+) - -molar + ++ ++ ++ ++

Msx-1 db ++ + (+) - (+)molar ++ ++ ++ ++ ++

Msx-2 db NA (+) - - NAmolar NA + ++ ++ NA

Pax-9 db + (-) (-) NA NAmolar ++ +(+) (+) NA NA

The first gene expression difference that wasnoted between the molars and the diastemaregion was the stronger expression of Bmp4 andweaker Fgf8 in the future diastema area than inthe future molar or incisor areas at E10 in bothspecies (Figure 7 in Article III). This was in thevole diastema region followed by downregula-tion of the mesenchymal Bmp4, Lef1, Msx1,Msx2 and Pax9 gene expressions (Table 6, Fig-ure 22). In the mouse diastema region, Msx1expression remained relatively strong (Figure 5in Article III), as reported by Turecková et al.(1995), but Bmp4, (not shown), Lef1 and Pax9were downregulated (Figures 5 and 6 in ArticleIII). The downregulation of mesenchymal genesappears to be associated with the inhibition of

mesenchymal condensation around the diastematooth germs, since only the rudimentary con-densed diastema mesenchyme expresses Bmp4in either species (Figure 8 in Article III, notshown). This may be connected with the down-regulation of Pax9, because Pax9 is essential forBmp4 expression whilst Bmp4 and Msx1 form apositive feedback loop (Chen et al. 1996, Peterset al. 1998). Because the vole diastema toothgerms advance into middle (late) bud stage likethe molars of Pax9-/- or Msx1-/- mice, thissuggests a failure of mesenchymal specification(Satokata and Maas 1994, Peters et al. 1998).However, Pax9 is not essential for initiation oftooth germs, because the diastema tooth germsexist, and because in Pax9-/- targeted mutant

60

mice the tooth development is arrested only atthe middle bud stage (Peters et al. 1998).

Regional specification of the mesenchymedepends on positional identity, but positionalidentity is not only a mesenchymal property. Thetranscription factor, Msx1, is expressed epithel-ially in incisors and anterior diastema toothgerms but not in molars. Moreover, although thevole diastema buds arrest at the late bud stage,the mouse diastema tooth germs degenerate atthe early bud stage, suggesting that some addi-tional early epithelial failure occurs in the devel-opment of the mouse diastema buds as comparedto the vole diastema buds. In addition, the devel-opment of the vole diastema buds was not“rescued” with BMP4, even when combined toFGF4 (data not shown), which suggests that inthe sibling vole, diastema region more signalingpathway(s) than the Bmp4-pathway are alsoaffected.

Positional identity and tooth morphogenesis(Article III, unpublished results)

The morphological differences between mouseand vole molar shapes prove that positionalidentity does not equate with the morphogeneticprogram. The E11 mouse-vole and vole-mouserecombinations, however, showed that the earlyepithelium controls the specific morphogenesisof the individual tooth germs. Hence, teeth withthe same positional identity, like mouse and voleM1, can have different morphologies dependingon the early epithelial signals.

The mouse incisor – mouse molar recombin-ations always produced always mouse, not voleteeth (not shown), however. This means thatpositional morphological differences determinedby the positional identities are not the same asthe morphological differences between species.To create the species-specific morphologies, thepositional identities are hence either translatedby species-specific morphogenetic programsdifferently or the positional identity-determiningprogram as a part of the tooth developmentalprogram has changed between the species. In theformer option, the tooth morphogenetic pro-

grams are epithelial and only controlled by thepositional identity programs, whereas in the lat-ter option the signals determining the positionalidentity of the tooth germ participate directly inmorphogenesis, e.g., by inducing local differ-ences in the mesenchymal tooth developmentalpotential. The known effects (Neubüser et al.1997) and the expression patterns of Bmp4 andFgf8 (see above) support the latter possibility.Moreover, the existence of the diastema toothgerms indicates that not all tooth developmentalpathways are affected similarly by the positionalidentities. In addition, the vole, but not themouse, molar early epithelial signaling centersexpressed Fgf4 and Fgf8 (Figure 3 in Article II,Figure 6 in Article III), showing that regardlessof dental identity, the Fgf-pathway can becontrolled separately from the Shh- or Bmp-pathways. Hence, the morphological divergenceof both paralogous and orthologous teeth islikely to result from changes in the patterning ofone or more dissociable early epithelial signalsrelative to others.

Evolution of muroid dental formulas (ArticleIII, unpublished results)

The comparison between mouse and vole molarand diastema tooth germ gene expression pat-terns indicated that mesenchymal specificationand other processes involving epithelial Fgf8and Bmp4 are parallel and separate from theinitiation of the early epithelial signaling centersand individual tooth germs. Therefore, theseprocesses are probably controlled by differentpatterning events. The positional identitiescontrol the initiation of the tooth germs, as seenfrom the numerical and developmental differ-ences between mouse and vole diastema toothgerms and between the diastema regions of theupper and the lower jaws. Hence, aside from thepositional identities, the numbers and locationsof the tooth germs also depend on the earlyepithelial signaling. In contrast, the formation ofthe dental lamina appears not to depend onpositional identity, because it occurs in the dias-tema region as well, and some genes (e.g. Bmp4,Fgf8 (Figure 7 in Article III) and Shh (notshown) are even expressed in the nasal pits.

61

Co-expression of dissociable Bmp-, Fgf- andShh-signaling pathways correlates with the loca-tions of the tooth germs and their developmentalfates. This may explain the recombination re-sults, which indicated that the epithelium definesthe areas of mesenchymal odontogenic potentialand determines the identity of the paralogoustooth germs (Lumsden 1988, Kollar and Mina1991, see above). However, the expression pat-terns and results from cross-species recombina-tions suggest that the epithelium also controlsthe placing, numbers and early budding of theindividual tooth germs, and that it determines thespecies specific differences in the morphogene-sis of orthologous tooth germs. Because theepithelial patterning controlling the identitiesand morphologies of the tooth germs are at leastpartially dissociable from patterning controllingthe tooth numbers and locations, heterotopicshifts in different signaling patterns relative toeach other may also provide a molecular expla-nation for the observed homeotic shifts in themorphologies of paralogous tooth germs invarious mammalian lineages (Butler 1978,Osborn 1978, Schwartz 1982).

The conserved expression of Bmp4, Fgf8 andPax9 (Figures 6 and 7 in Article III) in the dias-tema region suggests that the early determinationof the positional identities is conserved betweenthe mouse and vole. In situ hybridization showedthat both mouse and vole molar and diastemaregions, but not the incisor regions, also expressBarx1 (not shown). Barx1 is a mesenchymallyexpressed transcription factor associated with thedevelopment of molars (Tissier-Seta et al. 1995,Mucchielli et al. 1997, Tucker et al. 1998b).Signaling molecules Bmp4 and Fgf8 have beenassociated with the development of facial proc-esses in mouse and chicken (Heikinheimo et al.1994, Wall and Hogan 1995, Barlow andFrancis-West 1997, Tucker et al. 1999), and theycontrol the expression of mesenchymal Barx1similarly in mouse and chicken (Barlow et al.1999). The early epithelial Fgf- and Bmp-signaling has also been shown to affect the shapeof the skull bones (Barlow and Francis-West1997, Richman et al. 1997). Hence, it is likelythat the evolutionarily conserved expression

patterns of Bmp4 and Fgf8 has in mammals beenco-opted to control the positional identities andmorphogenesis of the teeth, e.g., by controllingthe mesenchymal Bmp4-pathway.

The determination of tooth locations is probablyconnected with the evolution of other cranialfeatures. The morphological connection ofmouse D1 and vole D with the maxillary edge ofthe primary choana is, e.g., repeated in theircommon expression areas with genes Shh andLef1 (Figures 4 and 5 in Article III). The mouseD2 and D3 Shh expression continues similarly inthe palatal rugae they are connected with (Figure4 in Article III). If signaling molecules involvedin early development of the primary choanae andthe palatal rugae can initiate tooth development,co-evolution with orofacial features mightexplain why the dental formulas tend to be lessreduced in the upper than in the lower jaws ofthe studied rodents (see above, Moss-Salentijn1978, Peterková 1985, Peterková et al. 1993,Luckett 1985, however, see also Luckett et al.1989). Since palatal rugae and primary choanaedo not exist in the lower jaw, the retention of thetooth germ initiation program would not be asfavored in the lower jaw. It is noteworthy thatthe rudimentary tooth germs of other studiedvoles also associate with the primary choanae. Ifnot for the report that field vole (M. agrestis) hasthree small epithelial swellings in its upper di-astema region similar to the mouse (Witter et al.1996), this could have been interpreted as a signof the dental formulas having been determined atthe subfamily level. Nevertheless, our resultssuggest that the connection to primary choanaefavors the initiation of tooth germs.

In the mouse, epithelial Bmp4 inhibits mesen-chymal Pax9 expression at E10,5 but not atE11,5 (Neubüser et al. 1997), and it also acti-vates mesenchymal Msx1 expression, thus beingessential for molar development (Vainio et al.1993, Chen et al. 1996, Tucker et al. 1998a).Even earlier, Bmp4 is essential for the determi-nation of incisor identity (Tucker et al. 1998b).Hence, the toothlessness in the rodent diastemaregion may have originally been caused bychanging the time window when Pax9 needs to

62

be initiated in the anterior dental mesenchyme ascompared to the inhibitory expression of Bmp4.The roles of Fgf8 and Bmp4 in inhibiting toothdevelopment in the early anterior maxillae havenot been fully elucidated, however, because inhuman Pax9 haploinsufficiency, the molars andcentral incisors seem to be affected, but not thecanines and anterior premolars, which are miss-ing in the muroid diastema areas (Stockton et al.2000). Moreover, the differences in tooth initia-

tion and degeneration suggest that the loss ofdiastema teeth may also have been caused bypositional control of other patterning processes.Hence, instead of heterochronic changes inBmp4, Fgf8 and Pax9 expression, there mayhave been qualitative changes, which have madethe tooth development in the anterior maxillavulnerable to local positional signals and/orinhibition of Pax9 expression.

Enamel knots and crown morphogenesis (Articles I, III and IV)

Because the enamel knots express several secret-ed signaling molecules (Figure 21), includingknown mitogens like Fgf4 and Shh (Jernvall etal. 1994, Kettunen and Thesleff 1998, Hardcastleet al. 1998) and because the development of theprimary enamel knot precedes the initiation ofcrown base development (see above, Jernvall etal. 1998) whilst the tips of individual cusps inthe mouse and opossum (Monodelphis domes-tica) are associated with a secondary enamelknot (Jernvall 1995), the enamel knots have beenproposed to be signaling centers which controlcrown morphogenesis, e.g., by inducing localgrowth (Jernvall et al. 1994, Jernvall 1995,Thesleff and Jernvall 1997).

The size of the primary enamel knot has beenproposed to control the size of the crown baseduring cusp initiation (Jernvall 1995), whereasthe timing and spacing of the initiation of thesecondary enamel knots, together with cuspgrowth, has been proposed to control the specificcusp patterns (Jernvall 1995, Jernvall 2000). Theenamel knots have been hypothesized to controlthe development of subsequent enamel knots,thus controlling not only the physical morpho-genetic processes, like growth and differentia-tion, but also the actual spatiotemporal pattern-ing governing morphogenesis (Jernvall 1995,Jernvall 2000, Jernvall and Thesleff 2000). Inthe molars and premolars of the Ladoga ringedseal (Phoca hispida ladogensis), the relativeheight difference between any three neighboringcusps is constant and correlates with the angleformed by the main cusps. This suggests that thesharpness of the angle determines the number of

cusps/tooth, which can vary between individualteeth. If the timing of enamel knot initiationdetermines the heights of the cusps, then theangle between individual cusps depends on theconstancy of the rate of enamel knot initiationand of growth rate. Regular initiation can beaccomplished with iteration of the sameinduction-inhibition program for each enamelknot (Figure 23). In this model, the previousenamel knot produces an inhibition field, whichprevents the formation of another enamel knotnear it, whereas new enamel knots are induced inthe growing epithelium beyond the inhibitionfield(s) of the previous enamel knot(s) (Jernvall2000). However, all these hypotheses, as well asthe molecular and developmental mechanisms ofenamel knot function, have been speculative,because the initiation dynamics and morphologi-cal associations of the individual enamel knotsare so far largely unknown.

To analyze the development and relationshipsbetween the individual enamel knots, the expres-sion patterns of four marker genes, Fgf4, Lef1,p21CIP1/WAF1 and Shh were mapped in developingmouse and sibling M1s. These genes are ex-pressed in all enamel knots in both species(Table 1 in Article II), and in situ hybridizationresults suggested that they appear at differenttimes during enamel knot development. Theirinductive relationships are unknown, however.Because Fgf4 is expressed in mouse and siblingvole teeth exclusively in the differentiatedepithelial signaling centers (Table 1 in Article II,see below), it was used as a molecular markerfor mature enamel knots.

63

By comparing the expression patterns of Fgf4,Lef1, p21CIP1/WAF1 and Shh to each other and todigital elevation models (DEMs) made fromthree-dimensional GIS-reconstructions of base-ment membranes each day during crownmorphogenesis, it is possible to correlate theirsignaling pathways with morphogenesis. On theother hand, small time interval analysis of thespatiotemporal expression changes in theseparated mouse and sibling vole molar epitheliawas used to analyze the development of theindividual enamel knots. By combining these ap-proaches with the comparative morphometry ofmouse and sibling vole crown growth patterns, itwas possible to generate hypotheses using boththe molecular basis of the enamel knot function,and its role in the development of specific crownmorphologies.

The development of individual enamel knots(Articles I and IV, unpublished observations)

Earlier sectional in situ hybridization in bothspecies (data not shown) suggested that Fgf4,Lef1, p21CIP1/WAF1 and Shh expressions appear atdifferent stages during enamel knot develop-ment. Small time interval analysis of Fgf4, Lef1,p21CIP1/WAF1 and Shh expression in separatedmouse molar epithelia confirmed this hypothe-sis; Lef1 was the first gene to appear in theenamel knot area, p21CIP1/WAF1 the next, whileFgf4 was the last following Shh expression(Figures 13 and 17). Lef1, p21CIP1/WAF1 and Shhwere expressed in wide and partially overlappingpatterns which changed rapidly between timepoints in both mouse (Figures 17, 24, 25, 26 and27) and vole molars (Figures 24, 25, 26 and 27).The differences in the expression patternssuggest that Lef1, p21CIP1/WAF1 and Shh arecontrolled by separate signaling combinations.Hence, the enamel knot/epithelial signalingcenter program actually consists of several,potentially dissociable signaling pathways.

The partial overlap of the expression patternswas not readily apparent during the developmentof the primary enamel knot (Figure 13). Whenthe cervical loops appear, however, the growinginner enamel epithelium around the enamel knot

A B

Figure 23. A comparison between the iterative induction-inhibition cycle (A) and lateral inhibition (B) as amechanism for enamel knot development. In the iterativemodel, the enamel knots develop into maturing epithelium(medium gray), which therefore induces enamel knotdevelopment (arrow). In the next stage, the mature enamelknot (dark gray) produces an inhibitory signal (blunt head),which prevents enamel knot development in its radius (lightgray). The mature enamel knot itself does not respond tothis inhibitory signal, either because it is incompetent orbecause it produces a local activating signal, which does nothave as wide a range as the inhibitory signal. In the case oflateral inhibition, the cells begin to simultaneously produceinducing and inhibitory signal(s), which increase the enamelknot maturation within the cells themselves and inhibit thematuration of nearby cells. The enamel knots are thus selfmaintaining and their sizes and distances are determined bythe diffusion/relay distances of the inductory and inhibitorysignal(s). In both cases, the end result looks similar, but theintermediate stages are different. The size of the matureepithelium is assumed to increase regardless of the enamelknot signals.

begins to express Lef1, p21CIP1/WAF1 and/or Shhin clearly distinct patterns both in the mouse(Figures 17, 24 and 25, Figures 3 and 4 inArticle II) and in the vole (Figures 24 and 25,Figures 3 and 4 in Article II). The expressionareas of Lef1 (Figures 17), p21CIP1/WAF1 (Figures17 and 24) and Shh (Figures 17, 25 and 26) oftenappear to spread from the enamel knots of bothspecies. However, as the size of the inner enamelepithelium increased, Lef1 and p21CIP1/WAF1

64

expression areas unconnected to the enamelknots can appear, as seen, e.g., in the small timeinterval analysis of mouse M1 talonid region(Figure 17). Therefore, the enamel knots are notthe local inducers of epithelial maturation.

Mesenchymal signals have been shown to beable to induce epithelial p21CIP1/WAF1 expression(Jernvall et al. 1998). Because the expressionareas seem to associate with older epithelium,the patterns of epithelial maturation may bedepend the age of the epithelium, which in turndepends on its growth patterns. Thus, the mesen-chyme may provide permissive signals for matu-ration and initiation of new enamel knots. Be-cause all mouse and sibling vole enamel knotsare induced within the common overlap areas oflarger Lef1, p21CIP1/WAF1 and Shh expressionareas (Figures 17 and 27), and because they areinduced consecutively on the later developingends of the cusp rows in both species (Figures 26and 27), it appears that the pattern of enamelknot initiation depends on the rates and patternsof epithelial maturation. By inference, the actualgrowth rates may determine the enamel knotpatterns, which then determine the folding pat-terns of the crown. These, in turn, are dependenton the specific relationship between epithelialand mesenchymal growth rates (see above).Hence, at the level of morphogenetic processes,the patterns of enamel knot initiation are insepa-rable from the epithelial and mesenchymalgrowth rates.

Small time interval analysis (Figures 13, 17, 25and 26) showed that expression of each genewas stabilized and upregulated where newenamel knots were forming, whereas it could belost from the epithelium between enamel knots.All new enamel knots are formed in areasexpressing p21CIP1/WAF1 and Lef1 (Figures 17, 27and 28), and in mouse M1, these seem to corre-spond to the mapped quiescent areas (see Coinet al. 1999). The initiation of isolated Lef1 andp21CIP1/WAF1 expression in the mouse talonidregion (Figure 17) suggests that quiescent areascan also be induced de novo, without being indirect contact with a previous enamel knot. Be-cause in mouse molars the quiescence appears

E14

E14,5

MouseVole p21

Figure 24. The p21CIP1/WAF1 expression during bell stagetransition in mouse and vole molar epithelium. The expres-sion is lost around the primary enamel knot and the E14expression is split into several compartments. The first sec-ondary enamel knot is later induced in the lingual compart-ment. Anterior is to the left and buccal up. The scale bar is200 µm.

before Fgf4 expression (Figures 1 and 2 in Arti-cle I, Coin et al. 1999), the quiescence is proba-bly essential for enamel knot differentiation.Although the expression areas of Lef1 (Figure17, not shown), p21CIP1/WAF1 (Figures 17 and 24)or Shh (Figures 17, not shown) may have beencontinuous before the appearance of the neigh-boring enamel knot, in M1s, the common localstabilization points of these genes were separate(Figures 17 and 27), and Fgf4 was always initi-ated within these areas (Figures 16, 17 and 27).Hence, the M1 mature enamel knots, as visual-ized with small time interval Fgf4 expressionanalysis, e.g., in mouse, did not split into multi-ple enamel knots (Figure 17). Therefore, despitean earlier report (Coin et al. 1999), the M1 sec-ondary enamel knots do not develop by directlydividing from the primary enamel knot. Thedivision hypothesis was based on cell labelingstudies, which clearly showed the secondaryenamel knots splitting away from the quiescentarea corresponding roughly to the primaryenamel knot. The early expansion of thep21CIP1/WAF1 expression area around the primaryenamel knot may explain the results, becausep21CIP1/WAF1 arrests cells in the G1-phase, and

65

Vol

eM

ouse

E14 E15Shh

Figure 25. Anterior downregulation of the primary enamel knot and induction of the first secondary enamel knot in vole andmouse molars, as shown with small time interval changes Shh expression. In both species, the secondary enamel knot begins tobe induced directly lingual to the primary enamel knot, but in the vole the anterior downregulation continues further. Thesimultaneous folding of the crown base is also visible. Anterior is to the left and buccal up. The scale bar is 200 µm.

the anterior secondary enamel knots in mousemolars form in the p21CIP1/WAF1 expressing areaaround the primary enamel knot when it splits(Figures 17 and 24). Unlike in mouse and voleM1, however, according to preliminary data, themature primary enamel knot in mouse M2

appears after widening to split to form the firstsecondary enamel knot (Figure 17). It is notknown, how the other M2 enamel knots form, asM2 development was not further analyzed.

In the mouse talonid region, the de novo expres-sion area of Lef1 and p21CIP1/WAF1 splits to formtwo enamel knots corresponding to the entoco-nid and hypoconid. The splitting occurs beforedetectable Fgf4 expression in either of the knots.Because the separation of these enamel knotsoccurs before their differentiation, the locationsof the individual enamel knots and the intercuspepithelia between them are determined beforethe maturation of the previous enamel knots.Together with the apparent splitting of the M2

primary enamel knot, these results suggest thatthe locations of the individual enamel knotswithin the epithelium are determined by lateralinhibition within mature epithelium rather thanby inhibitory signals from prior neighboringmature enamel knots.

Molecular and mechanical models for enamelknot induction and function

Although epithelial maturation appears to becontrolled by signals from the mesenchyme(Kratochwil et al. 1996, Jernvall et al. 1998),the expression patterns of the marker geneswere used for visualizing how epithelial genesmay control the enamel knot patterns by lateral

Mouse E16 Shh Mouse E17 Shh

Vole E16 Shh

Figure 26. Late Shh expression in E16 mouse, E16 siblingvole and E17 mouse molars. The expression spreads awayfrom the enamel knots. Anterior is to the left and buccal up.The scale bar is 200 µm.

inhibition. Very little is known about theinductive interactions between the analyzedmarker genes in teeth (Table 2), but they areunlikely to belong to a common signalingpathway (see above, Dassule and McMahon1998). However, the common associationbetween Fgf4, Lef1, p21CIP1/WAF1 and Shhexpression and predictable morphogeneticevents (Figure 28) suggests that the signalingpathways, in which they belong to, are involvedin the same morphogenetic processes. Themodel in Figure 29 was created by combiningthe expression data with what was previouslyknown about the inductive interactions.

Lateral inhibition can be created by differentialdiffusion or cell-to-cell relay and decay of in-ducing and inhibitory signals (reaction-diffusionmodel) (Collier et al. 1996, Hammer 1998, Asaiet al. 1999). However, because the distancesbetween the activation peaks and their sizes andshapes depend on the rates of signal production,diffusion/relay and decay (Collier et al. 1996,Hammer 1998, Asai et al. 1999), there may beminimum and maximum sizes and distances foreach enamel knot. This means that the molecular

66

mechanisms behind lateral inhibition may beimportant both for the initiation of new cuspsand for the remodeling of the enamel knotshapes. Since epithelial-mesenchymal interac-tions are important for determining the cuspshapes, it is possible that lateral inhibition in thecase of enamel knot development requires thepresence of permissive mesenchymal signals orcompetent mesenchyme to relay the epithelialinhibitory signals.

The enamel knot does not respond to the mito-genic signals it produces, suggesting that lateralinhibition includes local downregulation ofreceptivity to the mitogenic signals within thedeveloping enamel knot, whilst the production ofthese signals is upregulated. The stabilization ofquiescence occurs only within regions that co-express Lef1, p21CIP1/WAF1 and Shh. Because Shhis known to increase epithelial growth and toinhibit differentiation despite p21CIP1/WAF1

expression (Oro et al. 1997, Hardcastle et al.1998, Fan and Khavari 1999), the stabilizationrequires local prevention of mitogenic effects ofShh-signaling, whereas upregulation of Shhexpression apparently requires epithelial differ-entiation and/or quiescence. Because Fgf4expression does not necessarily precede theseparation of neighboring secondary enamelknots, Fgf4 is not likely to be involved in theinitial lateral inhibition process.

It is notable that the enamel knots do not expressany of the known Fgfrs, which are receptor tyro-sine kinases required for transduction of Fgf-signal, or Smo, Gli1, Gli2, Gli3 or Ptc, which areinvolved in Shh-signaling (Hardcastle et al.1998, Kettunen et al. 1998, Murone et al. 1999).Whilst the primary enamel knot expresses amitogen and survival factor Fgf4 (Jernvall et al.1994, Vaahtokari et al. 1996, Kettunen andThesleff 1998), it cannot respond to FGF4,probably because it lacks Fgfrs (Kettunen et al.1998). However, both the mesenchymally ex-pressed Bmp4 and epithelially expressed Bmp2proteins can promote epithelial Msx2 expressionand thus apoptosis in teeth (Jernvall et al. 1998,Marazzi et al. 1997), as well as the expression ofp21CIP1/WAF1 and Lef1 (Chen et al. 1996, Dassule

mouse E14 vole E14

mouse E15

vole E15mouse E16

vole E14,5

vole E16mouse E17

Figure 27. The spatial relationships of averaged expressionof Fgf4, Lef, p21CIP1/WAF1 and Shh projected on digitalelevation models of E14, E15, E16 and E17 mouse molarsand E14, E14,5, E15 and E16 vole molars. The degree ofco-expression is shown as the intensity of the shade in theexpression areas. The areas with only one expressed geneare shown in darkest gray, the areas with two genes inmedium gray, the areas with three genes in light gray andthe areas with all four genes expressed are in white. Thenon-expressing areas are black. Fgf4 was not analysed inE14,5 vole molars, which develop faster than mouse molars.Although the degree of overlap increased centripetallytowards the enamel knots, the development of which pre-ceded cusp development, the exact combinations of geneswere not the same (not shown). Anterior is to the left andbuccal up. The teeth are not to scale.

and MacMahon 1998, Jernvall et al. 1998).Hence, the primary enamel knot cannot grow bycell division, but it can promote the growth anddifferentiation of the surrounding epithelium andits own apoptotic removal.

Both the apparent splitting of the M2 primaryenamel knot (Figure 17) and the initial spread-ing patterns of Lef1 (Figures 13 and 17, notshown) and p21CIP1/WAF1 (Figures 13, 17 and 24)expression areas around the mature M1 primaryenamel knot suggest that the lateral inhibition of

67

p21CIP1/WAF1 and Lef1 does not occur simultane-ously with primary enamel knot differentiation,but begins during the bell stage transition. Be-cause in secondary enamel knots the inductionof intercusp epithelium and lateral inhibition issimultaneous with their differentiation, the mo-lecular patterning and morphogenetic functionsof the enamel knots may change over time (seealso Jernvall 1995). However, the confinementof the Fgf4 expression to the M1 primaryenamel knot suggests some level of lateral inhi-bition in them already at the cap stage, althoughthe growth of the intercusp epithelium is notinduced at that stage.

Later, the spreading of Lef1, p21CIP1/WAF1 andShh from the secondary enamel knots to thesurrounding epithelium (Figures 17, 24 and 26)suggests that the lateral inhibition of these genesis lost. The fusion of vole secondary enamelknots (Figure 16) implies that even lateral inhi-bition of Fgf4 can disappear. It is possible thatthe number of cusps depends on the time win-dow when the lateral inhibition is finally lost.Although the vole cusp pattern is finished byE17, however, three-dimensional reconstructionsshow that the vole molar germ continues to growfrom 1030 µm (measured from reconstruction inFigure 16) to about 2840 µm (measured from theadult molar). Hence, the ability to initiate newenamel knots does not equate with an ability togrow. Nevertheless, the cusp shapes are alsosimilar to the shapes of the secondary enamelknots associated with them (Figure 16). It there-fore seems that whilst the initiation of theenamel knots may depend on lateral inhibition,the development of epithelial patterning cannotbe dissociated from the epithelial and mesen-chymal growth rates.

Because the early epithelial signaling centers canbe first visualised by the local upregulation andcomplementary downregulation of Shh and Lef1expressing areas within Pitx2 expressing dentallamina (Figures 2 and 4 in Article III, notshown), it appears that, like enamel knots, earlyepithelial signaling centers are also formed bylateral inhibition process. Therefore the induc-

Figure 28. The expression patterns of Fgf4, Lef1,p21CIP1/WAF1 and Shh at each time point (arrow) correlatebest with the morphology of the next time point in both themouse and vole until E17. The lower correlation in the voleresults from a faster rate of pattern formation in volemolars, which means that the sampling intervals were notshort enough for higher resolution.

tion and lateral inhibition models (without speci-fying the actual signaling molecules) may func-tion in differentiation of all epithelial signalingcenters of teeth, although different mouse muta-tions show clear stage and dental identity spe-cific differences in the developmental arrest ofteeth (Table 1). Permissive signaling from theunderlying neural crest derived ectomesenchymehas, however, been shown to be involved inearly budding and in the transition from bud tocap stage, as well as being implicated in enamelknot differentiation (Chen et al. 1996, Jernvall etal. 1998, Ferguson et al. 1998, Thomas et al.1998, Peters and Balling 1999). Hence, the roleof epithelial-mesenchymal interactions in epi-thelial pattern formation still remains to beanalyzed.

68

Enamel knots and cusp patterns (Articles IIand IV, unpublished observations)

As seen from three-dimensional reconstructionsof Fgf4 expression in developing mouse andsibling vole molar crowns (Figure 16), the firsttwo enamel knots show the orthogonal or diago-nal arrangement of cusps as early as E15. Thenext enamel knots are always initiated at theends of two rows extending anteriorly and poste-riorly from these enamel knots. The buccalenamel knot corresponds to the future protoco-nid and the lingual enamel knot to the futuremetaconid. Because the distances between theindividual cusps are determined by their growthrates and periods (see above), the first two cuspsand the rate and period of cusp initiation appar-ently determine the relative locations of thebuccal and lingual cusps.

Small time interval analysis showed that therelationship between the protoconid and metaco-nid, and the relative location of the protoconidon the crown, base depend on the patterns ofprimary enamel knot downregulation and firstsecondary enamel knot initiation. As in mouseM1 (Jernvall et al. 1998), in sibling vole M1 mostof the primary enamel knot is removed apoptoti-cally (not shown) or its gene expression is down-regulated (Figures 25 and 27). In both species,however, the enamel knot associated with theprotoconid is actually a remnant of the primaryenamel knot, not a newly induced secondaryenamel knot (Figures 12, 25 and 27).

Moreover, in both species at E15, the relativedistance from the remnants of the primaryenamel knot to the anterior edge of the crownbase was approximately the same as the relativedistance from the protoconid tip to the anterioredge of the adult molar (Table 5). This relation-ship did not vary much, although the species andstage specific absolute growth rates could differgreatly (Table 4). This indicates that theposterior and anterior parts of the crown grow atthe same rate. Hence, the differences in therelative lengths of the anteroconid regions ofmouse and sibling vole M1s are not caused byallometric growth in different parts of the crown.

differentiation

proliferation

loss of receptorsLef1

p21

Fgf4

Shh

other signals?

Figure 29. The potential developmental roles of the studiedmarker genes and signaling pathways in the enamel knots.Since only a few of the gene functions and interactions areknown, the marker genes merely represent examples of thekinds of genetic interactions known to be present andrequired for the function of the enamel knots. An enamelknot may either induce proliferation around itself or at leastpattern it by remaining quiescent. The enamel knots do notexpress Fgfrs or Gli1, Gli2, Gli3, Ptc or Smo, whichsuggests that an inability to respond to their own mitogenicsignals is essential for their differentiation. However, byupregulating cell cycle elsewhere, the enamel knots preventdifferentiation. This suggests that the development of theenamel knots depends on lateral inhibition, but thepatterning signals, which promote the loss of receptors atthe sites of the future enamel knots, are unknown. Becauseplacodes express many of the same genes, the induction-inhibition loops may be similar. Certain aspects of lateralinhibition are repressed around the primary enamel knotduring the cap stage, but the expression analysis of thesecondary enamel knots shows that full differentiation ofthe enamel knots as manifested by Fgf4 expression is notessential for lateral inhibition.

Correlating the enamel knot marker gene expres-sion patterns with a GIS-program onto DEMs ofthe developing mouse and sibling vole crowns atdifferent developmental times showed that theexpression patterns correlated best with the mor-phologies of the following developmental stages(Figures 27 and 28). In other words, the enamelknots appeared before the actual growth of thecusps. Because all genes “predict” the morpho-genesis similarly, even the expression of theearliest enamel knot markers, p21CIP1/WAF1 andLef1, can be used to predict the locations of new

69

enamel knots, or cusps. In vole molars, the cor-relation was lower than in mouse molars (Figure28). Because the tempo of morphogenesis asmeasured in growth and initiation of new cuspswas greater in vole than in mouse molars, thelower correlation is likely to result from too longsampling intervals in vole molars.

The correlation between the genes and the de-veloping morphology (Figure 28) also illustratedhow the species-specific cusp angles appeared.In both species, the location of the first secon-dary enamel knot had by E14,5 been determinedto be directly lingual to the primary enamel knot(Figures 24, 25 and 27). In vole M1, the anteriordownregulation of the primary enamel knotexpression transformed the initial orthogonalenamel knot arrangement into a diagonal one(Figures 25 and 27). Hence, the species specificarrangements of the buccal and lingual cuspsdepend on the remodeling/downregulation of theprimary enamel knot as compared to the timingof the initiation of the first secondary enamelknot.

Because the primary enamel knot is associatedwith the protoconid and the development of eachcusp can be predicted from the expression pat-terns of the enamel knot marker genes, the roleof the enamel knots is not to shape the intercusp

valleys by creating enamel grooves, as previ-ously proposed (Lesot et al. 1996, Lesot et al.1999). Since enamel knots are areas of quiescentepithelium, all epithelial growth occurs outsidethem. The enamel knots have been proposed togenerate cusps by locally increasing the growthrates around themselves through the secretion ofmitogens like Fgf4 (Jernvall et al. 1994). As thecrown grows older, however, the growth occursmore distantly from the enamel knots. Moreover,BrdU-labeling studies of Coin et al. (1999) indi-cated a local lack of cell division within the epi-thelia of enamel knot and cusp tip areas, whichcombined to the p21CIP1/WAF1 and Lef1 expressionpatterns (Figures 17, 24 and 27) suggests that theenamel knots are actually centers for spreadingquiescence. This supports the view that “cuspsare centers of precocious maturation in innerenamel epithelium” (Butler 1956). Finally, thegrowth rates determining the future shape arealready visible at the cap stage, when the enamelknots are not yet associated with individualcusps. Hence, it is more likely that the enamelknots define the spatiotemporal relationships ofthe tips of the individual cusps by acting as localfoci of differentiation. Remodeling of enamelknot shapes would thus affect the cusp shapes bychanging the crown differentiation and growthpatterns.

The evolution of mouse and sibling vole molar cusp patterns

To be useful for evolutionary research, devel-opmental models must be able to explain evolu-tionary changes. The cusp pattern in mouse M1

consists of seven forward slanting conical cuspsin two rows, forming pairs orthogonal to theanterior-posterior axis of the tooth germ (Figures9 and 15). The sibling vole M1 crown, on theother hand, consists of an anterior loop and ninediagonally alternating prisms in two rows, whichare fused in the middle, forming a central ridge(Figures 9 and 15). The primitive muroid M1

cusp pattern consisted of five or six conical,nonslanting cusps paired in two rows, probablywith a diagonal arrangement, and a posterolo-phid (Figure 8). In the mouse lineage, the mainchanges in M1 have been the evolution of the

posteroconid from the posterolophid and theorthogonal pairing of the cusps. In the lineageleading to the sibling vole, the cusps becameincreasingly prismatic and crown heightincreased, until the crowns became rootless, andnew cusps have budded from the anteroconid, ascrown length increased.

As revealed by the recombinations, all aspects ofmolar morphogenesis are controlled by the earlyepithelium. The specific growth rates and patternformation are controlled by subsequent epithe-lial-mesenchymal interactions. The comparisonsbetween mouse and sibling vole molar crowndevelopment revealed that the apparently mosaicevolution of different parts of the crown and

70

features such as cusp shapes and crown heightscould theoretically be derived from a few metricgrowth processes and inductive and inhibitoryinteractions controlling cusp pattern andmorphology. If the developmental rules inferredfrom mouse and sibling vole are valid, theyshould explain also the morphologies of the teethof extinct muroids, represented by Democriceto-don sp., Antemus chinjiensis, Promimomys corsp. and Microtus (Allophaiomys) pliocenicus(Figure 8).

Crown size and cusp numbers

In mouse and sibling vole, the relative lengths ofthe anterior and posterior parts of the crown aredetermined by the location of the primaryenamel knot during the bell stage transition be-tween E14 and E15. In Democricetodon, whichrepresents an ancestral muroid, the relativedistance from the protoconid tip to the anterioredge of the crown was 0.45, in Antemus 0.44 andin mouse 0.46, whereas the relative distances inPromimomys, M. pliocenicus and in sibling volewere 0.63. Hence, though the location of theprotoconid does not directly correlate with theevolutionary increase in the number of anteroco-nid cusps, it can be quite conserved within thesubfamily level.

Because the growth rates are constant throughthe whole crown, the increase of specificallyanteroconid cusps in the genus Microtus has notbeen caused by allometric growth of differentparts of the crown during the initiation timewindow, but rather has been a result of the spe-cific patterns of elongation and subsequentdownregulation of the primary enamel knot. Thedownregulation patterns seen in sibling voleapparently were present already in the genusPromimomys. As the number of cusps increased,the crown length remained the same, whilst thesizes of the individual cusps decreased. Theanteroconid region, which makes up about 63%of the crown length in the analyzed M1s inMicrotine-lineage. Hence, relatively more cuspscould be initiated into the anteroconid than in thetalonid region.

The elongation and downregulation of the pri-mary enamel knot of mouse M1 have been pro-posed to be an adaptation for early initiation ofnew cusps. The early growth burst during thedevelopment of the sibling vole M1 crown indi-cates that other kinds of adaptations are possible.The relative sizes of cusps depend on the timedifference between their initiation and the crowngrowth rates and periods, whereas the numbersof the cusps depend on the length of the cuspinitiation period and the growth and maturationrates of the crown base. The cusp sizes are quiteunequal in Democricetodon, Antemus andmouse, and accessory cuspules or lophids can beseen in Democricetodon and Antemus. In themouse, the accessory cuspules have been lost,but the number of main cusps is the same as inAntemus chinjiensis (~ 11 million years ago).Hence, it appears that primitively the time win-dow for initiating new cusps was quite long ascompared to the total crown growth time, evenallowing for the initiation of accessory cuspulesor lophids between or beside the main cusps.

The number of cusps has increased in the lineageleading to the sibling vole, although the length ofthe M1 crown is practically the same in Promi-momys, M. pliocenicus and the sibling vole. Therelative sizes of the cusps are almost equal inPromimomys molars, whereas in M. pliocenicus,and to a lesser extent, sibling vole molars, theanteriormost cusps are smaller than the rest. Thissuggests that in the Microtine lineage, the totalgrowth has remained the same, but the period ofcusp initiation has changed. In Promimomys, thetime window for initiating new cusps probablywas quite short and early relative to the totalgrowth time of the crown base, because thenumber of main cusps is small, while the cuspsare large. In Microtus pliocenicus, there weremore and smaller cusps, but they were of une-qual size. Hence, it seems that the time windowfor cusp initiation was apparently lengthenedand postponed relative to the total crown growthperiod, or the early crown growth rates wereincreased in M. pliocenicus. In sibling vole, thecusps are smaller and more equal in size. Thiscorrelates with the early growth burst, whichsuggests that more cusps could be initiated dur-

71

ing the initiation window than in Promimomys.Because the individual cusps are smaller than inPromimomys, the actual crown growth time aftercusp initiation is likely to be shorter or thegrowth rates slower. Hence, it appears that in thelineage leading to sibling vole, the number ofcusps was increased by increasing the earlycrown growth rates during the period of cuspinitiation.

In adult molars of both mouse and vole, thedistance between consecutive large main cuspsin the anterior-posterior direction is about 0.5mm (Figure 9). The rate of cusp initiation is,however, faster in vole than in mouse molars,and the later growth rates are slower in mousemolars (Table 4). This suggests that the earlycusp pattern formation for several (equally) largecusps has been favored in the lineage leading tothe sibling vole. The greater inequality in prismsizes in M. pliocenicus indicates that there are nodevelopmental constraints on making unequalcusps in sibling vole molars, either. Thus, themore than doubled growth rate of the vole M1 ascompared to the mouse M1 between E14 andE15 (Table 4), when most of the vole secondaryenamel knots are initiated, may have been a so-lution to ecological pressures requiring manyequally (large) cusps from about 0.5 mm apart.

Orthogonal and diagonal cusp positions

In mouse and sibling vole molars, the relativepositions of the protoconid and the metaconidtogether with the crown growth rates and theenamel knot initiation periods effectively deter-mine the relative positions of buccal and lingualcusps. The enormous variation in the relativearrangement of buccal and lingual cusps in ex-tant and extinct muroid species suggests that therelative positions of the protoconid and metaco-nid have changed numerous times during evolu-tion (Gromov and Polyakov 1992, Freudenthaland Suárez 1999, Kälin 1999). Anteriordisplacement seems to be the ancestral conditionin muroid lineage (Figure 8), but it is derivedrelative to mammals in general (Figure 7). Thissuggests that the heterochronic (and possiblyheterotopic) shifts between the initiation of the

first secondary enamel knot and the down-regulation of the primary enamel knot arecommon, i.e., developmentally easy to generate.

The gene expression patterns tentatively suggestthat the initiation of the first secondary enamelknot is prevented by the early repression of thelateral inhibition around the primary enamelknot (see above). If this is so, the relative loca-tions of the protoconid and metaconid depend onthe time when the lateral inhibition is derepress-ed as compared to the primary enamel knotdownregulation. Hence, the shifts in the relativepositions of the buccal and lingual cusps maybest be explained by heterochronic changes inthe timing of the repressive signal and the down-regulation of the primary enamel knot. If theearly elongation of the primary enamel knot hasinitially evolved to facilitate more rapid earlydevelopment of the tooth crown in primitivemuroids, its side effects, the anterior and poste-rior elongation and downregulation of theprimary enamel knot, hence also seem to haveincreased the evolvability of the cusp arrange-ment in the muroid molar crown.

The evolution of prismatism and hypselodonty inMicrotinae

The constant relationship between the epithelialsurface area and the occlusal area in the mouseand sibling vole indicates that the relationshipcan be used for predicting the degree of foldingduring crown morphogenesis (Figure 14; Table3). However, the pattern of folding apparentlydepends on the signaling, which controls thedevelopment of the enamel knots. The lateralgrowth of the prisms, which creates the re-entrant angles, contributes to the folding of theocclusal surface (Figure 14). It also correlateswith the remodeling of the enamel knots shownin Figure 16. Because the signaling pathwaysinvolved in enamel knot development andspecific growth rates are likely to be dissociable,the evolution of hypsodonty is not necessarilycontrolled by the same molecular mechanismsthat control the shapes and locations of thecusps. Moreover, because the numbers andlocations of the prisms can vary between

72

hypselodont or rootless vole species (Figure 8),it is unlikely that the growth mechanisms behindhypselodonty directly affect the patterning of theenamel knots. Thus, although the evolution ofthe molars in the Microtine lineage has beencharacterized by an increase in prismatism andby an increase in hypsodonty (e.g., Chaline1989, Fejfar and Heinrich 1989), it is likely thatthese properties have evolved by concertedchanges in separate processes.

The most likely reason for this “co-evolution” isecological. The hypselodont molars of siblingvole can sustain the increased wear from acoarse diet longer that the brachydont mousemolars. The transition to hypselodonty requiresthat the growth of the cervical loops does notterminate. Because this growth is maintained bylocal epithelial-mesenchymal interactions in therodent incisor cervical loops (Harada et al.1999), it is possible that the evolution of openrooted molars in Microtinae has been caused byheterotopic co-option of incisor specific signalsunder molar positional information. However,the fossil record, as well as, extant variability inthe degree of hypsodonty and hypselodonty inMicrotinae suggests that the evolution of hyp-selodonty has been caused by gradual quantita-

tive and temporal shifts in the epithelial-mesen-chymal ability of the molar region to maintaincervical loop growth.

Whilst hypsodonty is an evolutionary responseto crown wear, conical cusps cannot be hypselo-dont (Appendix 1). Moreover, prismatic cusps’alternating with re-entrant angles has increasedthe number and size of the shearing blades. Theremodeling of the vole enamel knots suggeststhat prismatism depends on enamel knot shapes,not on crown growth. In the sibling vole, thecusps grow initially in the lateral direction,creating re-entrant angles between consecutivecusps. After mineralization, the cusps grow onlyin height. In the primitive microtoid cricetidBjornkurtenia, the crown was mesodont andnarrower near the top than near the root,suggesting similar growth rates as in conicalcusps, but even then, re-entrant angles weregenerated by folds between the consecutivecusps (Kowalski 1992). Hence, whilst therelationships between epithelial and mesenhymalgrowth rates apparently determines the degree offolding within a tooth crown, the creation of there-entrant angles is probably connected with theevolution of folding patterns, which is likely todepend on the evolution of enamel knots.

Concluding remarks As the deep conservation of molecular functionand patterning principles in all metazoans hasbecome increasingly apparent, a major problemin morphological evolution turns out to be howthe conserved signals and patterning mecha-nisms create disparate morphologies even inclosely related species.

Thus far, the role of molecules during morpho-genesis has been largely studied with rough loss-of-function or gain-of-function mutants orexperimental induction-inhibition analysis.However, since the development of specificmorphologies depends on local spatiotemporaldifferences in gene expression, future develop-mental research will require more spatial andtemporal expression data and their correlation

with morphogenesis. This problem wasapproached by using GIS programs for studyingthe spatial correlation between gene expressionand morphological features. Unlike previous insitu hybridization methods, this has enabledquantitative correlations between gene expres-sion and future morphologies to be determined.Such correlations can reveal signaling pathways,which are potential targets for morphologicalevolution.

Mammalian teeth are excellent models for ana-lyzing the relationship between growth and pat-tern formation, because they are histologicallysimple and their morphogenesis is deterministic.The morphogenesis and expression patterns ofseveral candidate developmental regulatory

73

genes were mapped in both mouse and volemolars from initiation to the late bell stage andthe upper diastema rudimentary tooth germsfrom initiation to degeneration. The first findingwas that mouse molar morphogenesis may beregulated by epithelial signaling centers, theenamel knots. Rough qualitative gene expressioncomparisons, together with cross-species tissuerecombinations, showed that the tooth develop-ment programs are conserved between the spe-cies. In both species, the development of dentalformulas and cusp patterns was similarly associ-ated with epithelial signaling centers. Althoughthe early development of mouse and vole molarsoccurs at the same rate via same developmentalstages, the actual growth patterns and spatiotem-poral arrangement of the signaling centers dur-ing the morphogenesis are nonetheless different.Sibling vole and mouse molars do not recapitu-late common morphogenetic patterns.

As revealed by molar-incisor and mouse-volerecombinations and early epithelial expressionpatterns, all tooth morphogenetic processes arecontrolled by the early epithelium. The numeri-cal and developmental differences betweenmouse and sibling vole rudimentary tooth germsand the different morphogenetic associationsbetween the studied developmental genes sug-gest that the hierarchy of pattern formation be-gins with several parallel and dissociable induc-tive events (Figure 30). The early patterningdefines the local odontogenic potential andhence the development of dental formulas. Atthe following levels, the shape of each individualtooth germ is determined by subsequent epithe-lial-mesenchymal interactions, which control thespecific growth rates and pattern formation.

Unlike the entire tooth, individual cusps do nothave positional identity. Instead, the evolution ofcusp patterns apparently depends on hetero-chrony between dissociable pathways controll-ing growth and lateral inhibition. The apparentlymosaic evolution of different parts of the crownas well as features like cusp shapes and crownheights could, actually, be derived from a fewmetric processes controlling the meristicpatterning.

Anterior-posterioraxis

Buccal-lingualaxis

Othermodifications

Positionalinformation

Site of thedental lamina

Identity of the tooth germ

Dental mesenchyme Location of tooth germ

Intrinsic rates of signaling

Lateral inhibition

Placodes

Budding

Epithelialgrowth rate

Mesenchymal growth rate

Cervical loops

Primary enamel knot

Crown width

Number ofcusp rows

Crown length

Number of cusps / row

Timing of epithelial / mesenchymal

differentiation

Angle betweenthe cusp rows

(Lateralinhibition?)

Figure 30. The hierarchy of minimum developmentalprocesses in teeth. The number of signaling pathwaysinvolved in each of these processes, as well as the extent oftheir co-option in parallel or consecutive events, isunknown. The initial patterning processes establish the jawaxis, and the positional information is then interpretedaccording to the species specific rates of signaling intodental formulas and epithelial and mesenchymal growthrates. The latter, together with the intrinsic pattern form-ation dynamics, will determine the specific cusp patterns ofeach individual tooth germ, unless development is arrestedbecause one or more of the underlined early patterningevents fails.

The budding of the individual tooth germs beganwith the differentiation of the early epithelialsignaling center from the dental lamina. Theprimary enamel knots appear to control crowndevelopment in general, whereas the initiation ofsecondary enamel knots predicts the develop-ment of individual cusps. The signaling centershad similar but not identical gene expressionprofiles. Because the expression profiles couldchange as the signaling centers matured, thedifferences between the signaling centers maydepend on the developmental stage of the tooth

74

rather than intrinsic differences between thesignaling centers.

The development and morphogenetic roles ofboth primary and secondary enamel knots wereanalyzed using molecular markers. The resultsabove indicate that the timing of enamel knotseparation and initiation and the growth rates ofthe intercusp areas determine the cusp patterns.In mouse and sibling vole molars, the actualcusp patterns appear at the molecular levelearlier than at the morphological level, and thispattern apparently depends on the molecularsignaling interactions, as well as the crowngrowth.

The likely involvement of lateral inhibition inthe development of signaling centers suggeststhat there may be a minimum size for an enamelknot, and a minimum distance between twoenamel knots. Their calculation, however,requires more knowledge about the molecularbasis for enamel knot formation and interactionsbetween growth and patterning. The constantrelationship between the epithelial and mesen-chymal growth rates appears to determine thedegree of crown folding (Figure 14, Table 3), asalso indicated by the cross-species recombin-ation experiments (Figure 20). However, thecusp shapes are similar to the shapes of thesecondary enamel knots associated with them(Figure 16). Hence, it seems that the develop-ment of enamel knot patterns cannot be dissoci-ated from the crown growth rates at the morpho-logical level. At the molecular level, however,the individual signaling pathways appear to bedissociable, which probably has been the basisboth for extensive heterochronic and heterotopicvariation in cusp pattern development, and (on ahigher level of the information hierarchy) for theevolution of positional control of tooth morpho-genesis.

To fully understand the molecular basis under-lying the evolution of tooth morphologies, sev-eral levels of information must be integrated. To

identify the potential pathways actually involvedin morphogenesis, in situ hybridization and spa-tial autocorrelation analysis between the markersfor candidate pathways and morphogenesis areneeded, whereas to understand the patternformation at the molecular level, the signalinginteractions between individual proteins must beunderstood. Hence, experimental data from, e.g.,induction or gene transfection experiments, areessential. The number of potential candidategenes and pathways to be analyzed can belimited by pinpointing the exact developmentalevents and processes, which define the futuremorphogenesis, e.g., by recombination experi-ments. In the future, as the genomic data be-comes available, the analysis of temporal (if notspatial) co-expression profiles of the thousandsof genes potentially involved in morphogenesiswill provide information on interconnectivitybetween signaling networks. Conversely, insteadof comparing individual genes and pathways, theeffects of the whole genome (or a proteome)during a specific time window can be analyzedby heterologous tissue recombinations. The taskahead is enormous, but once the molecular rulesfor morphogenesis are elucidated, these can beused for predicting the favored directions ofmorphological change and, if necessary, for gen-erating transgenic phenocopies of extant mor-phologies or even novel morphologies from theunexplored parts of the possible morphospace.

In conclusion: by elucidating the developmentalprocesses behind the pattern formation in mouseand vole teeth, the processes can be tentativelyapplied to extrapolate the potential mechanismsfor the evolution of the mammalian dentition,and even to predict the potential directions ofmorphological evolution. If enough comparativeresearch is carried out on a wide range of modelsystems, including Drosophilidae and plants,ultimately even the basic mathematical princi-ples behind the genomic basis for the evolutionof patterning processes may be identified andanalyzed

75

Acknowledgements

This work has been done in the Institute of Dentistry(University of Helsinki) and in the Institute of Biotechnology(University of Helsinki) under Professor Irma Thesleff, whois the director of the Research Program for DevelopmentalBiology, and who has provided me outstanding possibilitiesto learn and study among the best. For the same reason, mysincere thanks also to Professor Mart Saarma, the director ofthe Institute of Biotechnology.

Many heartfelt thanks to my thesis supervisors DocentJukka Jernvall and Professor Irma Thesleff, whose help wasessential for making this work possible. Without Jukka’skind and patient teaching and unusual genius this workwould not be the same. I hope to be able to keep up the samestandard (if not to improve it) also in future.

The reviewers Docent Kirsi Sainio and Dr. Lars Werdelinprovided constructive criticism and real help in writing thisbook.

Also all those who provided me financial and materialsupport are gratefully acknowledged, especially my parentsand grant parents, Institute of Dentistry (University ofHelsinki), Helsinki Graduate School in Biotechnology andMolecular Biology directed by Professor Heikki Rauvala,Sigrid Juselius Foundation, Finnish Academy and theInstitute of Biotechnology (University of Helsinki), as wellas all the people mentioned elsewhere, who allowed the useof their plasmid constructs.

Because science depends on co-operation, the many peopleworking in the Research Program for DevelopmentalBiology, especially in the “Tooth” and the “Kidney” groupsis gratefully acknowledged. Especially the superb technicalsupport by our expert technicians and computer personnelhas been absolutely essential for successful completion ofthis work; without their help and patient teaching this workwould not have been possible. There also are countlessothers, who at one point or another have given me goodadvice and help, but who are too numerous to mention here.Hence, my thanks to anybody feeling involved in my work inthe Institute of Biotechnology, Division of Genetics,Division of Biochemistry, Department of AnimalPhysiology, Department of Geology, Institute of Dentistryand elsewhere. This also involves the often-undervaluedadministrative efforts of the people under Mart and Heikki,and the people in the DNA-sequencing laboratory under LarsPaulin and “Elatus”. Thank you.

Also my co-authors, and to a degree, mentors, PäiviKettunen, Thomas Åberg and Anne Vaahtokari deserve theirthanks. I liked sharing the same room with them.

Han-Sung Jung is gratefully acknowledged, for hisinvaluable help in the recombination experiments. AnniHienola shared her expertise on small PCR-samples.

Docent Christophe Roos’s views and computer expertise,and all the advice and help with references by ProfessorMikael Fortelius deserve thanks, as does Professor HannuSariola’s advices in writing technique.

Elina Waris made it possible to work without realizing morethan a couple of times a year how much bureaucracy thescientific work actually contains.

The moral and intellectual support provided by my mentorsand friends, (in an alphabetical order, not in the order ofimportance) Anu and Ari Alho, Petri Auvinen, Simone deLourenco, Alla Hanninen, Tapio Heino, Maire Holopainen,Marjo Hytönen, Mervi Hyvönen, Ritva Härönen, TiinaImmonen, Risto Jaatinen, Heidi and Kaija Kettunen, PäiviKettunen, Hyun-Jung Kim, Petra Koppinen, Hanna Kurppa,Leevi Kääriäinen, Tuire and Sami Lahtinen, Anja Lampio,Johanna Laurikkala, Ritva Leponiemi, Keijo Luukko, EevaMatinolli (and all my excellent teachers wherever they noware), Xiaojuan Meng, Marja Mikkola, Maxim Moshnyakov,Tuija Mustonen, Merja Mäkinen, Pekka Nieminen, Anna-Maija Partanen, Juha Partanen, Maija Pekkanen, PirjoPekkarinen, Reijo Peltomaa, Marja-Leena Peltonen, UllaPirvola, Johanna Pispa, Anne Raatikainen-Ahokas, LeenaRantakokko, Marjatta Raudaskoski, David Rice, Eija andOlli Ruokonen, Susanna Saarinen, Carin Sahlberg, KirsiSainio, Mervi Salo, Riikka Santalahti, Hannu Sariola, LenaSelänne, Päivi Setälä, Virpi Syvälahti, Teemu Teeri, MikaTirronen, Nina and Ras Trokovic, Mark Tummers, AnjaTuomi, Arja Tuomi, Anne Vaahtokari, Maria von Numers,Tarja Välimäki, Gudrun Wahlström, Janna Waltimo,XiuPing Wang, Elina Waris, Kirmo and Jorma Wartiovaara,Jukka Ylikoski, Thomas Åberg and Satu Åkerberg, deservesmore than thanks. These people were always willing to listenand help, and I am more indebted to them than can be said inhere. Also many thanks to Saku.

Moreover, I would like to thank for interesting collaboration,Professor Kathleen K. Smith, who provided the opossummaterial and Dr. Alex van Nieveltd, whose hospitality andhelp in obtaining some obscure but essential articles must notbe forgotten, and Professor Mark Ferguson, who was alsowilling to help, (though this was one of the dead endprojects).

For interesting discussions, thanks to Drs. John Hunter, AnnHuysseune and especially Dr. Jonathan Bard (who probablyprovided me the greatest inspiration of all).

Finally, Extra Special Thanks to:My parents, without whom (on more than one sense) Iwould never have become an evolutionary biologist,and Leena Rantakokko and Arja Tuomi for theirnurturing support.

76

Materials and methods

Tissues of the mouse and sibling vole

Mouse (Mus musculus) teeth and jaws (from E10 toE17) were obtained from crosses between inbred CBAmale and outbred NMRI female or NMRI male andNMRI female. The morning of the plug day wasconsidered to be E0. Sibling vole (Microtus rossiae-meridionalis), bank vole (Clethrionomys rufocanus),common vole (Microtus arvalis) and root vole(Microtus oeconomus) tissues of similar age wereobtained from an inbred colony kept in the Dept. ofAnimal Physiology (University of Helsinki, Finland).The animals were allowed to get used to each other inseparate cages overnight and then mated overnight.The following morning was considered to be E0. Therat (Rattus norvegicus) tissues were obtained fromSprague-Dawley crosses. The animals were matedovernight and the following morning was consideredas E0. The postnatal opossum heads were a kind giftfrom Dr. Kathleen Smith. The morning when theopossum pups were first seen was considered as PN0.

The tissues were dissected with scissors and needlesin Dulbecco’s PBS. The tissues for sectional in situhybridization were fixed overnight in 4% PFA, and, ifnecessary, decalcified for two weeks in 2.5%PFA/12.5% EDTA, after which they were dehydrated inascending ethanol xylene series before paraffinembedding. The lengths of the washes depended onthe age and size of the tissues and could vary between5’ and 60’ or longer. The opossum heads wereskinned before embedding. The tissues forwholemount in situ hybridization were fixed similarlyand dehydrated either in ascending ethanol series or inascending methanol series and stored in 100%methanol in –20°C or in 70% ethanol in +4°C untilused. To analyse the development of epithelialpatterning, epithelia separated with 0.75% pancreatinand 2.25% trypsin in Tyrode’s solution were fixed forwholemount in situ hybridization.

Probes for in situ hybridization

Probe Length Reference Article

Barx1 (murine cDNA fragment) 0.98 kb in pKSII(+) from Dr. Mitsiadis UnpublishedBmp2 (murine cDNA fragment) 240 bp in pGEM3 Vainio et al. (1993) I, II, IIIBmp2 (murine cDNA fragment) 1.2 kb in pBS(II)SK Dickinson et al. (1990) IIIBmp4 (murine cDNA fragment) 285 bp in pGEM3 Vainio et al. (1993) I, II, IIIBmp4 (murine cDNA fragment) 1 kb in pSP72 Jones et al. (1991) IIIBmp7 (murine cDNA fragment) 220 bp in pGEM3 IFgf4 (full length murine cDNA) 620 bp in pBS(II)KS+/- Hébert et al. (1990) I, II, III, IV,

unpublishedFgf8 (full length murine cDNA) 997 bp in pBKCMV Heikinheimo et al. (1994) IIILef1 (murine cDNA(?) fragment) 660 bp in pBS Travis et al. (1991) II, III, IV,

unpublishedMsx1 (murine cDNA fragment) 600 bp in pSP72 MacKenzie et al. (1991) I, II, IIIMsx2 (murine cDNA fragments in

tandem) 850+850 bp in pSP72 Monaghan et al. (1991) II, III

p21CIP1/WAF1CIP1/WAF1

(murine cDNA fragment)740 bp in pBS SK(+/-) Jernvall et al. (1998) II, III, IV,

unpublishedPax9 (murine cDNA fragment) IIIPitx2 (murine cDNA fragment) 1.8 kb in pKSII(+) Mucchielli et al. (1997) IIIShh (rat cDNA fragment) 2.6 kb in pBS SK(+) from Dr T. Edlund I, II, III, IV,

unpublishedWnt10a (murine cDNA fragment) 1.93 kb in pJ32 from Dr. McMahon unpublished

77

In situ hybridization of sections (Articles I, II andIII, unpublished results)

Radioactive in situ hybridization is a more sensitivemethod for expression analysis than wholemount insitu hybridization, and therefore suitable for testingunknown probes in new kinds of tissues. Tissuesstored in paraffin blocks were cut into 7µm or 10 µmsections and transferred onto triethoxysilane andacetone treated slides, dried overnight and stored at+4°C until used. Plasmids containing cDNA werelinearised and in vitro RNA transcription was carriedout in presence of 35S-rUTP (Sigma-Aldrich) andRNAse inhibitor. The labeled riboprobes were ethanolprecipitated, air dried and dissolved into Wilkinson’shybridization buffer containing 0.1M dithiothreitol(Sigma-Aldrich) as in Wilkinson and Green (1990).At the beginning, the slides were deparaffinated andrehydrated in descending xylene-ethanol-PBS series,washed twice in TE pH8.0 (50mM Tris-HCl, 5mMEDTA) treated with proteinase K (7 µg/ml TE pH8.0,RT, 15’), washed with PBST, fixed with 4% PFA (20’at RT), washed with PBST and treated with aceticanhydride to improve the penetration and specificityof binding of the probes. The sections weredehydrated and the probes (40 000 – 60 000 cpm/µl)were hybridized overnight in +52°C in 15 – 90 µl ofhybridization buffer. All pre hybridization treatmentswere done with RNAse free solutions and instruments.After hybridization, the sections were washed, first inlow stringency conditions (5x SSC pH4.5 [0.75MNaCl, 75mM Na-citrate], 10mM dithiothreitol at+50°C), then in high stringency conditions (20 mMdithithreitol in 50% formamide and 2x SSC at +65°C),washed three times in NTE (500mM NaCl, 10 mMTris-HCl pH8.0, 5 mM EDTA pH8.0 at +37°C) andtreated with RNAse A in NTE to remove unboundRNA probe, after which the high stringency wash wasrepeated. Finally, the sections were dehydrated andcoated with NTB2 autoradiographic emulsion(Kodak), dried and exposed for 10 to 16 days. Afterthe exposure, the sections were developed with KodakD19, fixed with Unifix (Kodak), and counterstainedwith hematoxylin (Shandon, Pittsburgh, PA). Finallythe sections were mounted with DePex (BDH) andwashed clean of excess emulsion. The results weredigitized using an Olympus BX-50 microscope(Olympus, Tokyo, Japan), a black and white CCDvideo camera (Cohu, San Diego, CA) and NIH-image1.61 public domain program (U.S. National Institutesof Health), and the bright and dark field images wereprocessed with Adobe Photoshop 4.0 (AdobeSystems, San Jose, CA).

Three-dimensional analysis of epithelialexpression patterns (Article IV, unpublishedresults)

Many of the Fgf4 expression patterns were obtainedwith three-dimensional reconstructions of sectionshybridized with more sensitive radioactive probes,because we could not make the short probe work wellin vole tissues for wholemount in situ hybridization.The teeth were sectioned serially and the bright anddark field images of the sections were digitized,aligned and stacked with the NIH image 1.61 program(US National Institutes of Health, public domainprogram available via the Internet by anonymous FTPzippy.nimh.nih.gov). Basement membranes weremarked with a few selected points for rendering in theExtreme 3D (Macromedia) program to produce thethree-dimensional morphology of the tooth germs,whereas the silver grains indicating the expression indark field images were inverted and stacked in NIH-image 1.61. The rendered basement membrane andthe expression grains were oriented in the sameposition, and the expression was projected on themorphology according to the digital co-ordinates inPhotoshop 4.0 (Adobe).

When the gene expression was strong, the patterninganalysis could be done with epithelial wholemount insitu hybridization, which can be used for rapidanalysis of several samples for statistical or short timeinterval developmental studies. The separated epi-thelia were digitized with a Kodak Digital ScienceDcm120 Zoom digital camera (Kodak) mounted on aNikon SMZ-U stereomicroscope (Nikon) in antiocclu-sal direction after the color reaction. The colorproduced by the reaction was filtered apart from theshadows caused by epithelial folding in AdobePhotoshop 4.0 (Adobe Systems, Inc.), and theexpression patterns were aligned to each other and theDEMs of type serial sections of correspondingdevelopmental stages according to theirmorphological features. DEMs were generated fromhorizontal sections of tooth germs using the 3Dviewversion (public domain by Iain Huxley) of NationalInstitute of Health (NIH) Image software(http://www.physics.usyd.edu.au/physopt/3dview/)as described in Jernvall and Selänne (1999). The tipsof the cusps, the dental lamina and the length of thetooth germ, which were clearly visible in the separat-ed epithelia were used as alingment points. Becausethe epithelia were fixed and stored similarly to thesectioned tissues, shrinkage was equal, and did notaffect the measurements, but to diminish the effects ofepithelial distortion and variable hybridization

78

reactions, the expression was averaged both betweendifferent epithelia and within 30 µm x 30 µm gridsquares. The surface areas, volume areas and occlusalareas were obtained using the 3Dview version of NIHImage and MFWorks GIS package (Thinkspace).Averages of gene expression patterns were super-imposed using the combine-operation (MFWorks,Thinkspace, Tomlin 1992) which allow for theseparation of all the possible combinations of geneexpressions. The expression areas were correlatedusing Spearman rank correlation to the morphologiesof the occlusal areas of different time points (dividedinto 30 µm x 30 µm grid squares) and by adjusting theexpression area to the size of the occlusal area at eachdevelopmental time point. To avoid circular logic, thecusps were defined in the GIS program as isolatedconvex areas.

Wholemount in situ hybridization

The riboprobes were transcribed in the presence ofdigoxygenin labeled rUTP (Boehringer-Mannheim)and 25U RNAse inhibitor/10µl reaction (Promega),ethanol precipitated, air dried and dissolved in DEPC-H2O. The prehybridization washes began with dehyd-ration of tissues in either descending methanol series(if stored in methanol) or in descending ethanol series(if stored in ethanol), washed 3x in DEPC-PBST (1%Tween 20 in Dulbecco’s PBS), treated with proteinaseK (10 µg/ml in +37°C), washed and refixed with 4%PFA for 20’ in RT, washed and transferred intohybridization buffer (50% formamide, 1.3x SSCpH4.5, 5mM EDTA pH8.0, 0.5% CHAPS, 0.1%Tween 20, 1% BBR, 100 µg/ml yeast tRNA and 50µg/ml heparin, ad DEPC-H2O) for prehybridization in+55°C for 1 to 2 hours. The probes were hybridized inthe same buffer (concentration 0.2 – 1.5 µg/ml) in+55°C overnight. All the solutions and instrumentsduring prehybridization were RNAse free. The excessprobes were washed away in high-stringency post-hybridization washes (50% formamide, 2x SSC, 1%Tween 20, 0.5% CHAPS; 25% formamide, 1x SSC,0.5% Tween 20, 0.25% CHAPS, 0.5x MABT pH7.5[0.1M maleic acid, 0.15M NaCl, 0.1% Tween 20] in+55°C). The tissues were then gradually taken intoantibody blocking solution [20% normal goat serum(Gibco BRL), 2% BBR (Boehringer-Mannheim) inMABT at RT. After 1 – 5 hour preblocking withoutantibody at RT, the tissues were hybridized with1/2000 anti-digoxygenin-AP-FAB-fragments(Boehringer-Mannheim) in +4°C overnight. Excessantibody was washed away the next day with longMABT washes in RT and these washes were usuallyextented overnight at +4°C. The color reaction with

0.168 mg/ml NTB (Boehringer-Mannheim) and 0.087mg/ml BCIP (Boehringer-Mannheim) was performedin NMT (0.1M Tris-HCl pH9.5, 1% Tween 20, 50mMMgCl2, 0.1M NaCl, 2mM levamisole), into which thetissues were taken via three washes. The reaction wasstopped with PBST and the tissues were fixed with4% PFA overnight in +4°C, washed three times andstored in 50% glycerol in PBST.

Apoptosis detection by TUNEL-staining (ArticlesII and III)

Paraffin stored tissues were sectioned, deparaffinatedin xylene and ethanol, and treated with 66% ethanol +33% acetic acid and 100% methanol + 0.5% H2O2.Then they were rehydrated in descending methanolseries, and washed with PBST before proteinase Ktreatment as in sectional in situ hybridization. Thetissues were washed again with PBST and fixed with4% PFA for 20’ at RT. The positive controls werealso treated DNAse (0.2 U/ml for 15’ at +37°C) andwashed with PBST before fixation. After the PFA hadbeen washed away with PBST, the sections werelabeled with 20 U TdT (Promega) in 75 µl of labelingmix (1 pmol/µl Digoxygenin-11-dUTP, 1 pmol/µldATP and 0.5% CHAPS in 1x TdT buffer) for 1 h at+37°C. The negative controls were treated similarlybut without TdT. When the slides were slightlyoverstained the reaction was stopped in 300 mMNaCl, 30 mM Na-citrate, 0.1% CHAPS. The sectionswere washed in TBT (50 mM Tris-HCl pH7.5, 150mM NaCl, 0.1% Triton X-100) before preblocking in2% BSA (Sigma-Aldrich) and 10% normal goatserum in TBT for 2-3 h. The sections were hybridizedwith anti-digoxygenin-AP-FAB-fragments in 1:2000in TBT overnight in +4°C. After that, the sectionswere washed in TBT and NMT. The colour reactionwas done as in wholemount in situ hybridization.After stopping the reaction with PBST, the sectionswere washed with MQ-H2O and mounted withAquamount (Danbrit).

In vitro tissue culture (unpublished results)

The NMRI mouse and sibling vole tissues (E14 toothgerms and E11 jaws) were microdissected in sterileconditions under a stereomicroscope. The tissues wereseparated from each other with 3’ – 6’ incubation in0.75 % pancreatin (Gibco BRL) and 2.25% trypsin(Difco laboratories) in Tyrode´s solution. The tissueswere allowed to recover for 30’ in culture medium(Dulbecco’s modified Eagle’s medium, supplementedwith 0.2% glutamax (GibcoBRL), 20% fetal bovineserum (GibcoBRL) and 10 IU/ml penicillin/10 µg/ml

79

streptomycin (GibcoBRL)) before their final separat-ion with fine needles. Epithelia and mesenchymeswere recombined in the combinations indicated in theresults and discussion and transferred onto Nucleporefilters, poresize 0.1 µm (Corning). The filters wereplaced on Trowell-type grids in humidified incubatorsat +37°C, 95% humidity, 5% CO2 and cultured withthe same medium overnight to attach the tissues toeach other. Next morning, the recombinations werecarefully detached from the filter with needles, andtransplanted into the kidney capsules of male NMRInude mice (Jackson Laboratories) anesthesised withfreshly made Avertin. Only one kidney was treated inany animal, but several recombinant tooth germs wereplaced into each treated kidney as in Kratochwil et al.(1996). The kidneys were harvested after two or threeweeks culture and the tooth germs were dissectedaway. The soft tissues of the tooth germs were eitherdigested away with 2.25% trypsin to reveal the shapeof the mineralized crown or dissected carefully forPCR-genotyping

PCR-genotyping (Unpublished results)

To ensure that the mouse and vole tissues remainedreasonably uncontaminated in the kidney capsules andthat the recombinations were correctly made, DNAwas isolated from the dissected epithelia and mesen-chymes of the cultured tooth germs with 10 hourincubation in 17.5 µg proteinase K in 20 µl of PCRdigestion buffer (100 mM Tris-HCl pH 8.3, 2.5 mM

EDTA pH 8.0, 100 mM NaCl, 0.2% SDS in H2O) at50°C. The proteinase was inactivated at 94°C for 20’.The PCR reactions were done in 50 µl volume, with0.36 mM dNTP (Promega) and 2U Dynazyme(Promega). For each PCR reaction, 1 µl of templatewas used. The cycles used were 20’ in 94°C, 36 times1’ +94°C, 1’ +59°C, 2’ +72°C, and final elongation10’ +72°C. The primers were

5’GGCCATCTACAAGAAGTCACAG for the S

and

5’CCATGCAGGAGCTATTACACA for the AS

direction for partial genomic p53 sequences of mouseand sibling vole. Since the primer areas are 100%identical and the PCR products are approximatelyequally long in the two species, there is no inherentbias for faster amplification of either mouse or voleproduct, which makes the PCR product reliable forsemiquantitative analysis of relative amounts ofmouse and vole cells in the analysed tissue. The PCRproducts were purified using the Qiaquick PCR-purification kit (Qiagen) and dissolved in 60 µl ofMQ-H2O, before their restriction fragment analysis bydigesting part of the product with SphI and anotherpart with NdeI. The mouse sequence contains a NdeIsite, which the vole sequence does not, whereas thevole sequence contains a SphI site, which the mousedoes not, enabling easy restriction fragment analysis.

80

Appendix 1

The relative growth rates of the mesenchymeand epithelium determine the sharpness of thecusps. In prismatic growth the change between t2

and t1 is the same as the change between t3 andt2, whereas in conical cusps the change betweent3 and t2 is greater than the change between t2

and t1. The difference depends on the slope ofthe cusp. The prismatic cusp is represented by aprism with a triangular base in Figure 31A andthe conical cusp by a straight cone with acircular base in Figure 31B.

If the downwards growth rate ∆t = x, then theincrease in the surface area of a prismatic cusp is∆A0-1 between t0 and t1 and ∆A1-2 between t1 andt2 and the increase in volume is ∆V0-1 between t0and t1 and ∆V1-2 between t1 and t2.

Because A0-1 = xa + xb + xc = A1-2,

and because V0-1 = (½ab/cosα)x = V1-2,

in prismatic cusps the ratio of A0-1 and A1-2 is thesame as the ratio of V0-1 and V1-2. Hence, thegrowth rates of the two-dimensional epitheliumand three-dimensional mesenchyme are thesame.

However, in conical cusps A0-1 = πs(r1 + r0),whereas A1-2 = πs(r1 + r2),where s = xcosαsinα.

Since r1 = r0 + xcosα and r2 = r0 + 2xcosα,

the difference is A0-1 = 2ro + xcosα A1-2 2ro + 2xcosα

The increase in volume between t0 and t1 isV0-1 = x/3(πr0

2 + πr0πr1 + πr12),

whereas the increase between t1 and t2 isV1-2 = x/3(πr1

2 + πr1πr2 + πr22).

Hence, the difference is

V0-1 = 3ro2 + 3roxcosα +x2cos2αV1-2 3ro2 + 3roxcosα +x2cos2α

A1-2 is greater than A0-1, and V1-2 is greater thanV0-1. The differences in volume and area growthdepend on the sizes of α and x, i.e., on thesharpness of the cusp and the downward growthrates. The differences between the surface areasat consecutive stages are always the same(2xcosα). The differences between the volumesat consecutive stages increase rapidly, however,with the increase depending on x and α. Hence,in conical cusps the volume (mesenchyme)grows faster than the surface area (epithelium).Thus, by controlling the mesenchymal andepithelial growth rates, it is possible to adjust theshape of the cusps.

Figure 31.A simplisticgeometric model ofdownwards growthin prismatic (A) andconical cusp (B).

81

References:

Adams M.D., Celniker S.E., Holt R.A., Evans C.A., GocayneJ.D., Amanatides P.G., Scherer S.E., Li P.V., HoskinsR.A., Galle R.F., George R.A., Lewis S.E., Richards S.,Ashburner M., Henderson S.N., Sutton G.C., WortmanJ.R., Yandell M.D., Zhang Q., Chen L.X., et al. (2000)The Genome Sequence of Drosophila melanogaster.Science 287, 2185-2195.

Akam M. (1998) The Yin and Yang of Evo/Devo. Cell 92,153-155.

Altabef M., Clarke J.D.W., Tickle C. (1997) Dorso-ventralectodermal compartments and origin of apicalectodermal ridge in developing chick limb.Development 124, 4547-4556.

Arthur W. (2000) The concept of developmentalreprogramming and the quest for an inclusive theory ofevolutionary mechanisms. Evol. & Devel. 2, 49-57.

Asai R., Taguchi E., Kume Y., Saito M., Kondo S. (1999)Zebrafish Leopard gene as a component of the putativereaction-diffusion system. Mech. Dev. 89:87-92.

Auffray J.-C., Tchernow E., Nevo E. (1988) Origine ducommensalisme de la souris domestique (Mus musculusdomesticus) vis-à-vis de l’homme. C. R. Acad. Sci. Parist. 307 Série III, pp. 517-522.

Averof M. (1997) Arthropod evolution: same Hox genes,different body plans. Curr. Biol. 7, R634-R636.

Ayscough K.R., Drubin D.G. (1998) A role for the yeastactin cytoskeleton in pheromone receptor clustering andsignalling. Curr. Biol. 8, 927-930.

Barlow A.J., Francis-West P.H. (1997) Ectopic applicationof recombinant BMP-2 and BMP-4 can changepatterning of developing chick facial primordia.Development 124, 391-398.

Barlow A.J., Bogardi J.P. Ladher R. Francis-West P.H.(1999) Expression of Chick Barx-1 and Its DifferentialRegulation by FGF-8 and BMP Signaling in theMaxillary Primordia. Dev. Dyn. 214, 291-302.

Bei M. Maas R. (1998) FGFs and BMP4 induce both Msx1-independent and Msx1-dependent signaling pathways inearly tooth development. Development 125, 4325-4333.

Bellusci S., Furuta Y., Rush M.G., Henderson R., WinnierG., Hogan B.L.M. (1997a) Involvement of Sonichedgehog (Shh) in mouse embryonic lung growth andmorphogenesis. Development 124, 53-63.

Bellusci S., Grindley J., Emoto H., Itoh N., Hogan B.L.M.(1997b) Fibroblast Growth Factor 10 (FGF10) andbranching morphogenesis in the embryonic mouse lung.Development 124, 4867-4878.

Berry A.C. (1978) Anthropological and Family Studies onMinor Variants of the Dental Crown. In: Development,Function and Evolution of Teeth (eds. Butler PM.,Joysey KA.) Academic Press, London pp. 81-98.

Bharathan G., Janssen B.J., Kellogg E.A., Sinha N. (1997)Did homeodomain proteins duplicate before the originof angiosperms, fungi, and metazoa?. PNAS 94, 13749-13753.

Bitgood M.J., McMahon A.P. (1995) Hedgehog and BmpGenes Are Coexpressed at Many Diverse Sites of Cell-

Cell Interaction in the Mouse Embryo. Dev. Biol. 172,126-138.

Bloch-Zupan A., Leveillard T., Gorry P., Fausser J.L., RuchJ.V. (1998) Expression of p21WAF1/CIP1 during mouseodontogenesis. Eur. J. Oral Sci. 106 Suppl 1:104-111.

Bock I.R. (1984) Interspecific hybridization in the GenusDrosophila. Evol. Biol. 18:41-70.

Bolker J,A., Raff R.A. (1996) Developmental genetics andtraditional homology. Bioessays. 18, 489-494.

Bown T.M., Kraus M.J. (1979) Origin of the TribosphenicMolar and Metatherian and Eutherian Dental Formulae.In: Mesozoic Mammals The First Two-Thirds ofMammalian History. eds. Lillegraven J.A., Kielan-Jaworowska Z., Clemens W.A. University of CaliforniaPress, Berkeley, USA. pp. 172-191.

Botchkarev V.A., Botchkareva N.V., Roth W., NakamuraM., Chen L.-H., Herzog W., Lindner G., McMahon J.A.,Peters C., Lauster R., McMahon A.P., Paus R. (1999)Noggin is a mesenchymally derived stimulator of hair-follicle induction. Nature Cell Biol. 1, 158-164.

Brenner S. (1997) Technical knockouts. In: Loose Ends fromCurrent Biology. Current Biology Ltd, London, UK. pp.80-82.

Brisken C., Park S., Vass T., Lydon J.P., O'Malley B.W.,Weinberg R.A. (1998) A paracrine role for the epithelialprogesterone receptor in mammary gland development.PNAS 95, 5076-5081.

Brunet-Lecomte P., Chaline J. (1991) Morphologicalevolution and phylogenetic relationships of theEuropean ground voles (Arvicolidae, Rodentia). Lethaia24, 45-53.

Bueno D., Skinner J., Abud H., Heath J.K. (1996) Spatialand Temporal Relationships Between Shh, Fgf4 andFgf8 Gene Expression at Diverse Signalling CentersDuring Mouse Development. Dev. Dyn. 207, 291-299.79S-80S.

Butler P.M. (1956): The ontogeny of molar pattern. Biol.Rev. 31, 30-70.

Butler P.M. (1978) The ontogeny of mammalianheterodonty. J. Biol. Buccale. 6, 217-228.

Butler P.M. (1995) Ontogenetic aspects of dental evolution.Int. J. Dev. Biol. 39, 25-34.

Carroll R.L. (1988) Vertebrate Paleontology and Evolution.W. H. Freeman and Company, New York, USA.

Carroll S.B., Gates J., Keys D.N., Paddock S.W., PanganibanG.E.F., Selegue J.E., Williams J.A. (1994) PatternFormation and Eyespot Determination in ButterflyWings. Science 265, 109-114.

Chai Y., Bringas P,Jr., Mogharei A., Shuler C.F., SlavkinH.C. (1998) PDGF-A and PDGFR-α Regulate ToothFormation Via Autocrine Mechanism during MandibularMorphogenesis In vitro. Dev. Dyn. 213, 500-511.

Chaline J. (1989) An approach to studies of fossil arvicolids.In: International Symposium Evolution, Phylogeny andBiostratigraphy of Arvicolids (Rodentia, Mammalia)Rohanov (Czechoslovakia) May 1987. eds. Fejfar O. andHeinrich W.-D. pp. 45-82.Verlag Dr. Friedrich Pfeil,Munich, FRG.

82

Chaline J., Graf J.D. (1988) Phylogeny of the Arvicolidae(Rodentia): Biochemical and paleontological evidence.J. Mamm. 69, 22-33.

Chaline J., Sevilla P. (1989) Phyletic grdualism anddevelopmental heterochronies in a EuropeanPlio/Pleistocene Mimomys lineage (Arvicolidae,Rodentia). In: International Symposium Evolution,Phylogeny and Biostratigraphy of Arvicolids (Rodentia,Mammalia) Rohanov (Czechoslovakia) May 1987. eds.Fejfar O. and Heinrich W.-D. pp. 85-98.Verlag Dr.Friedrich Pfeil, Munich, FRG.

Chen Y., Bei M., Woo I., Satokata I., Maas R. (1996) Msx1controls inductive signaling in mammalian toothmorphogenesis. Development 122, 3035-3044.

Chen Y., Zhao X. (1998) Shaping Limbs by Apoptosis. J.Exp. Zool. 282, 691-702.

Chiang C., Swan R.Z., Gratchchouk M., Bolinger M.,Litingtung Y., Robertson E.K., Cooper M.K., GaffieldW., Westphal H., Beachy P.A., Dlugosz A.A. (1999)Essential Role for Sonic Hedgehog during Hair FollicleMorphogenesis. Dev. Biol. 205, 1-9.

Clemens W.A. (1979) Marsupialia. In: Mesozoic MammalsThe First Two-Thirds of Mammalian History. eds.Lillegraven J.A., Kielan-Jaworowska Z. Clemens W.A.University of California Press, Berkeley, USA. pp. 192-220.

Clemens W.A., Kielan-Jaworowska Z. (1979)Multibuberculata. In: Mesozoic mammals The first two-thirds of mammalian evolution. eds. Lillegraven J.A.,Kielan-Jaworowska Z., Clemens W. University ofCalifornia Press, Berkeley, USA. pp 99-149.

Clemens W.A., Lillegraven J.A. (1986) New LateCretaceous, North American advanced therianmammals that fit neither the marsupial nor eutherianmolds. Contributions to Geology, University ofWyoming, Special Paper 3, p55-85.

Cohn M.J., Tickle C. (1999) Developmental basis oflimblessness and axial patterning in snakes. Nature 399,474-479.

Coin R., Lesot H., Vonesch J.-L., Haikel Y., Ruch, J.-V.(1999) Aspects of cell proliferation kinetics of the innerdental epithelium during mouse molar and incisormorphogenesis: a reappraisal of the role of the enamelknot area. Int. J. Dev. Biol. 43, 261-269.

Collier J.R., Monk N.A.M., Maini P.K., Lewis J.H. (1996)Pattern Formation by Lateral Inhibition with Feedback:a Mathematical model of Delta-Notch IntercellularSignalling. J. Theor. Biol. 183, 429-446.

Cornelison D.D., Wold B.J. (1997) Single-Cell Analysis ofRegulatory Gene Expression in Quiescent and ActivatedMouse Skeletal Muscle Satellite Cells. Dev. Biol. 191,270-283.

Coulier F., Pontarotti P., Roubin R., Hartung H., GoldfarbM., Birnbaum D. (1997) Of Worms and Men: AnEvolutionary Perspective on the Fibroblast GrowthFactor (FGF) and FGF Receptor Families. J. Mol. Evol.44, 43-56.

Crompton A.W. (1995) Masticatory function innonmammalian cynodonts and early mammals. In:Functional morphology in vertebrate paleontology. eds.

Thomason J.A. Cambridge University Press, Cambridge,GB. pp. 55-75.

Crossley P.H., Martinez S., Martin G.R. (1996) Midbraindevelopment induced by FGF8 in the chick embryo.Nature 380, 66-68.

Darwin C. (1859) On the Origin of Species by Means ofNatural Selection, or the Preservation of Favoured Racesin the Struggle for Life (6th Edition, 1906), Murray,London.

Dawkins R. (1986) The Blind Watchmaker, Penguin BooksLtd, London, UK

Dassule H.R., McMahon A.P. (1998) Analysis of Epithelial-Mesenchymal Interactions in the Initial Morphogenesisof the Mammalian Tooth. Dev. Biol. 202, 215-227.

Deng C., Zhang P., Harper J.W., Elledge S.J., Leder P.(1995) Mice lacking p21CIP1/WAF1 undergo normaldevelopment, but are defective in G1 checkpoint control.Cell 82, 675-684.

De Sousa P.A., Watson A.J., Schultz R.M. (1998) TransientExpression of a Translation Initiation Factor IsConservatively Associated with Embryonic GeneActivation in Murine and Bovine Embryos. Biol. Repr.59, 969-977.

Dickinson M.E., Kobrin M.S., Silan C.M., Kingsley D.M.,Justice M.J., Miller D.A., Ceci J.D., Lock L.F., Lee A.,Buchberg A.M., Siracusa L.D., Lyons K.M., DerynckR., Hogan B.L.M., Copeland N.G., Jenkins N.A. (1990)Chromosomal Localization of Seven Members of theMurine TGF-β Superfamily Suggests Close Linkage toSeveral Morphogenetic Mutant Loci. Genomics 6, 505-520.

Dingus L., Gaffney E.S., Norell M.A., Sampson S.D. (1995)The Halls Of Dinosaurs A Guide to Saurischians andOrnitschians. American Museum of Natural History,New York, USA.

Drossopoulou G., Lewis K.E., Sanz-Ezquerro J.J., NikbakhtN., McMahon A.P., Hoffmann C., Tickle C. (2000) Amodel for anteroirporterior patterning of the vertebratelimb based on sequential long- and short-range Shhsignalling and Bmp signalling. Development 127, 1337-1348.

Echelard Y., Epstein D.J., St-Jacques B., Shen L., Mohler J.,McMahon J.A., McMahon A.P. (1993) Sonic Hedgehog,a Member of a Family of Putative Signaling Molecules,Is Implicated in the Regulation of CNS Polarity. Cell 75,1417-1430.

Ericson J., Morton S., Kawakami A., Roelink H. Jessell TM.(1996) Two Critical Periods of Sonic HedgehogSignaling Required for the Specification of MotorNeuron Identity. Cell 87, 661-673.

Erwin D.H. (2000) Macroevolution is more than repeatedrounds of microevolution. Evol. & Devel. 2, 78-84.

Fan H., Khavari P.A. (1999) Sonic Hedgehog OpposesEpithelial Cell Cycle Arrest. J. Cell Biol. 147, 71-76.

Fejfar O., Heinrich W.-D. (1989) Muroid rodentbiochronology of the Neogene and Quaternary inEurope. In: European Neogene Mammal Chronology.NATO ASI Series, Ser. A: Life Sciences. eds. LindsayE.H., Fahlbusch V., Mein P. Plenum Press, New York,USA. pp. 91-117.

83

Félix M.-A., Sternberg P.W. (1997) Two nested gonadalinductions of the vulva in nematodes. Development 124,253-259.

Ferguson C.A., Tucker A.S., Christensen L., Lau A.L.,Matzuk M.M., Sharpe P.T. (1998) Activin is an essentialearly mesenchymal signal in tooth development that isrequired for patterning of the murine dentition. GenesDev. 12, 2636-2649.

Ferguson C.A., Tucker A.S., Sharpe P.T. (2000)Temporospatial cell interactions regulating mandibularand maxillary arch patterning. Development 127, 403-412.

Fietz M.J., Concordet J.-P., Barbosa R., Johnson R., KraussS., McMahon A.P., Tabin C., Ingham P.W. (1994) Thehedgehog gene family in Drosophila and vertebratedevelopment. Development Suppl. 43-51.

Fisher A., Caudy M. (1998) The function of hairy-relatedbHLH repressor proteins in cell fate decisions.Bioessays. 20, 298-306.

Flomen R.H., Vatcheva R., Gorman P.A., Baptista P.R.,Groet J., Bariµie I., Ligutic I., NiÓetie D. (1998)Construction and Analysis of a Sequence-Ready Map in4q25: Rieger Syndrome Can Be Caused byHaploinsufficiency of RIEG, but Also by ChromosomeBreaks ≈ 90 kb Upstream of This Gene. Genomics 47,409-413.

Fortelius M. (1985) Ungulate cheek teeth: developmental,functional and evolutionary interrelations. Acta Zool.Fennica 180, 1-76.

Fraidenraich D., Lang R., Basilico C. (1998) DistinctRegulatory Elements Govern Fgf4 Gene Expression inthe Mouse Blastocyst, Myotomes, and DevelopingLimb. Dev. Biol. 204, 197-209.

French V. (1997) Pattern formation in colour on butterflywings. Curr. Op. Genet. & Dev. 7, 524-529.

Freudenthal M., Suárez E.M. (1999) Family Muridae. In:Land Mammals of Europe, The Miocene (eds. RössnerG.E. and Heissig K.), pp. 401-409. Verlag Dr. FriedrichPfeil, Munich, Germany.

Futuyma D.J. (1986) Evolutionary Biology, 2nd edition.Sinauer Associates, Sunderland, USA

Galant R., Skeath J.B., Paddock S., Lewis D.L. Carroll S.B.(1998) Expression pattern of a butterfly achaete-scutehomolog reveals the homology of butterfly wing scalesand insect sensory bristles. Curr. Biol. 8, 807-813.

Gao F.-B., Durand B., Raff M. (1997) Oligodendrocyteprecursor cells count time but not cell divisions beforedifferentiation. Curr. Biol. 7, 152-155.

Garn S.M., Lewis A.D., Kerewsky R.S (1964) Third MolarAgenesis and Variation in Size of the Remaining Teeth.Nature 201, 839.

Gaunt S..J. (1994) Conservation in the Hox code duringmorphological evolution. Int. J. Dev. Biol. 38, 549-552.

Gaunt W.A. (1961) The development of the molar pattern ofthe golden hamster (Mesocricetus auratus w.), togetherwith a re-assesment of the molar pattern of the mouse(Mus musculus). Acta Anat. 45, 219-251.

Gaunt W.A. (1966) The disposition of the developing cheekteeth in the albino mouse. Acta Anat. 64, 572-585.

Gerhart J., Kirschner M. (1997) Cells, Embryos, andEvolution. pp 1-642. Blackwell Science, USA.

Gilbert S.F. (1997) Developmental Biology, Fifth edition.Sinauer Associates, Inc., Sunderland, Massachusetts,USA.

Gilbert S.F., Opitz J.M., Raff R.A. (1996) ResynthesizingEvolutionary and Developmental Biology. Dev. Biol.173, 357-372.

Graham A., Francis-West P., Brickell P., Lumsden A. (1994)The signalling molecule BMP4 mediates apoptosis inthe rhombencephalic neural crest. Nature 372, 684-686.

Grewal M.S. (1962) The Development of an Inherited ToothDefect in the Mouse. J. Embryol. exp. Morph. 10, 202-211.

Gould S.J. (1983) Hens Teeth and Horses Toes. W.W.Norton and Company, New York, USA

Gromov I.M., Polyakov I.Ya. (1992) Voles (Microtinae).Smithsonian Institution Libraries and National ScienceFoundation, Washington, DC, USA.

Grüneberg H. (1951) The genetics of a tooth defect in themouse. Proc. Royal Soc. B 138, 437-451.

Grüneberg H. (1965) Genes and genotypes affecting theteeth of the mouse. J. Embryol. exp. Morph. 14:137-159.

Gumbiner B.M. (1998) Propagation and localization of Wntsignaling. Curr. Op. Genet. Dev. 8, 430-435.

Guthrie R.D. (1965) Variability in characters undergoingrapid evolution, an analysis of Microtus molars.Evolution 19, 214-233.

Halder G., Callaerts P., Gehring W.J. (1995) Induction ofEctopic Eyes by Targeted expression of the eyeless Genein Drosophila. Science 267, 1788-1792.

Hammer T. (1998) Diffusion and Direct Signaling Modelsare Numerically Equivalent. J. Theor. Biol. 192, 120-130.

Hanken J., Hall B.K. (1993) The Skull. The University ofChicago Press, Chicago, USA

Hanks M.C., Loomis C.A., Harris E., Tong C.-X., Anson-Cartwright L., Auerbach A., Joyner A. (1998)Drosophila engrailed can substitute for mouseEngrailed1 function in mid-hindbrain, but not limbdevelopment. Development 125, 4521-4530.

Harada, H., Kettunen, P., Jung, H.-S., Mustonen, T., Wang,Y.A., Thesleff, I. (1999) Localization of putative StemCells in Dental Epithelium and Their Associations withNotch and FGF Signaling. J. Cell. Biol. 147, 105-120.

Haramis A.G., Brown J.M., Zeller R. (1995) The limbdeformity mutation disrupts the SHH/FGF-4 feedbackloop and regulation of 5' HoxD genes during limbpattern formation. Development 121, 4237-4245.

Hardcastle Z., Mo R., Hui C.-c., Sharpe P.T. (1998) The Shhsignalling pathway in tooth development: defects in Gli2and Gli3 mutants. Development 125, 2803-2811.

Harper J.W., Elledge S.J., Keyomarsi K., Dynlacht B., TsaiL.-H., Zhang P., Dobrowolski S., Bai C., Connell-Crowley L., Swindell E., Fox M.P., Wei N. (1995)Inhibition of Cyclin-Dependent Kinases by p21CIP1/WAF1.Mol. Biol. Cell 6, 387-400.

Harper J.W., Elledge S.J. (1996) Cdk inhibitors indevelopment and cancer. Curr. Op. Genet. Dev. 6, 56-64.

Hartl D.L., Clark A.G. (1989) Principles of PopulationGenetics, 2nd edition. Sinauer Associates, Sunderland,USA.

84

Headon D.J., Overbeek P.A. (1999) Involvement of a novelTnf receptor homologue in hair follicle induction.Nature Genet. 22, 370-374

Hébert J.M., Basilico C., Goldfarb M., Haub O., Martin G.R.(1990) Isolation of cDNAs Encoding Four Mouse FGFFamily Members and Characterization of TheirExpression Patterns during Embryogenesis. Dev. Biol.138, 454-463.

Heikinheimo M., Lawshé A., Shackleford G.M., WilsonD.B., MacArthur C.A. (1994) Fgf-8 expression in thepost-gastrulation mouse suggests roles in thedevelopment of the face, limbs and central nervoussystem. Mech. Dev. 48, 129-138.

Heikinheimo K., Bègue-Kirn C., Ritvos O., Tuuri T., RuchJ.V. (1997) The Activin-binding Protein Follistatin isExpressed in Developing Murine Molar and InducesOdontoblast-like Cell Differentiation in vitro. J. DentalRes. 76, 1625-1636.

Hillson S. (1986) Teeth Cambridge manuals in archaeology.Cambridge University Press, GB

Holland P.W.H., Garcia-Fernàndez J. (1996) Hox Genes andChordate Evolution. Dev. Biol. 173, 382-395.

Horan G.S.B., Ramírez-Solis R., Featherstone M.S.,Wolgemuth D.J., Bradley A., Behringer R.R. (1995)Compound mutants for the paralogous hoxa-4, hoxb-4,and hoxd-4 genes show more complete homeotictransformations and a dose-dependent increase in thenumber of vertebrae transformed. Genes Dev. 9, 1667-1677.

Hu G., Vastardis H., Bendall A.J., Wang Z., Logan M.,Zhang H., Nelson C., Stein S., Greenfield N., SeidmanC.E., Seidman J.G., Abate-Shen C. (1998)Haploinsufficiency of MSX1: a Mechanism for SelectiveTooth Agenesis. Mol. Cell. Biol. 18, 6044-6051.

Hu J.C.-C., Zhang C., Slavkin H.C. (1995) The role ofplatelet-derived growth factor in the development ofmouse molars. Int. J. Dev. Biol. 39, 939-945.

Huang Z., Pokrywka N.J., Yoder J.H., Stephenson E.C.(2000) Analysis of swallow homologue from Drosophilapseudoobscura. Dev. Genes Evol. 210, 157-161.

Hunter J.P., Jernvall J. (1995) The hypocone as a keyinnovation in mammalian evolution. PNAS 92, 10718-10722.

Hunter T. (2000) Signaling – 2000 and Beyond. Cell 100,113-127.

Huysseune A. (1995) Phenotypic plasticity in the lowerpharyngeal jaw dentition of Astatoreochromis alluaudi(Teleostei: Cichlidae). Arch. Oral Biol. 40, 1005-1014.

Huysseune A., Sire J.-Y. (1998) Evolution of patterns andprocesses in teeth and tooth-related tissues in non-mammalian vertebrates. Eur. J. Oral Sci. 106 Suppl1:437-481.

Iseki S., Araga A., Ohuchi H., Nohno T., Yoshioka H.,Hayashi F., Noji S. (1996) Sonic Hedgehog is Expressedin Epithelial Cells during Development of Whisker,Hair, and Tooth. Biochem. Biophys. Res. Comm. 218,688-693.

Jacobs L.L., Winkler D.A. Murry P.A. (1989) Modernmammal origins: Evolutionary grades in the EarlyCretaceous of North America. PNAS 86, 4992-4995.

Janis C.M., Fortelius M. (1988) On the means wherebymammals achieve increased functional durability of theirdentitions, with special reference to limiting factors.Biol. Rev. 63: 197-230.

Jenkins F.A.Jr., Crompton A.W. (1979) Triconodonta. In:Mesozoic Mammals The First Two-Thirds ofMammalian History. eds. Lillegraven J.A., Kielan-Jaworowska Z. Clemens W.A. University of CaliforniaPress, Berkeley, USA. pp. 74-90.

Jenkins F.A Jr., Gatesy S.M., Shubin N.H., Amaral W.W.(1997) Haramiyids and Triassic mammalian evolution.Nature 385, 715-718.

Jernvall J. (1995) Mammalian molar cusp patterns:Developmental mechanisms of diversity. Acta Zool.Fennica 198, 1-61.

Jernvall J. (2000) A link between develoment andevolutionary transformations. PNAS (in press).

Jernvall J., Kettunen P., Karavanova I., Martin L.B., ThesleffI. (1994) Evidence for the role of the enamel knot as acontrol center in mammalian tooth cusp formation: non-dividing cells express growth stimulating Fgf-4 gene.Int. J. Dev. Biol. 38, 463-469.

Jernvall J., Hunter JP., Fortelius M. (1996) Molar ToothDiversity, Disparity, and Ecology in Cenozoic UngulateRadiations. Science 274, 1489-1492.

Jernvall J., Åberg T., Kettunen P., Keränen S., Thesleff I.(1998) The life history of an embryonic signaling center:BMP-4 induces p21 and is associated with apoptosis inthe mouse tooth enamel knot. Development 125, 161-169.

Jernvall J., Selänne L. (1999) Laser confocal microscopy andgeographic information systems in the study of dentalmorphology. Palaeont. Electr. 2,http://www.odp.tamu.edu/paleo/1999_1/confocal/issue1_99.htm

Jernvall J., Thesleff I. (2000) Iterative signaling andpatterning during mammalian tooth morphogenesis.Mech. Dev. (in press)

Jiang J., Struhl G. (1996) Complementary and MutuallyExclusive Activities of Decapentaplegic and WinglessOrganize Axial Patterning during Drosophila LegDevelopment. Cell 86, 401-409.

Jones C.M., Lyons K.M., Hogan B.L.M. (1991) Involvementof Bone Morphogenetic Protein-4 (BMP-4) and Vgr-1 inmorphogenesis and neurogenesis in the mouse.Development 111, 531-542.

Jung H.-S., Francis-West P.H., Widelitz R.B., Jiang T.-X.,Ting-Berreth S., Tickle C., Wolpert L., Chuong C.-M.(1998) Local Inhibitory Action of BMPs and TheirRelationships with Activators in Feather Formation:Implications for Periodic Patterning. Dev. Biol. 196, 11-23.

Kardong K.V. (1995) Vertebrates Comparative AnatomyFunction Evolution. Wm. C. Brown Publishers,Dubuque, USA.

Kettunen P., Karavanova I., Thesleff, I. (1998)Responsiveness of Developing Dental Tissues toFibroblast Growth Factors: Expression of SplicingAlternatives of FGFR1, -2, -3, and of FGFR4; andStimulation of Cell Proliferation by FGF-2, -4, -8, and -9. Dev. Genet. 22, 374-385.

85

Kettunen P., Thesleff I. (1998) Expression and Function ofFGFs-4, -8, and -9 Suggest Functional Redundancy andRepetitive Use as Epithelial Signals during ToothMorphogenesis. Dev. Dyn. 211, 256-268.

Keys D.N., Lewis D.L., Selegue J.E., Pearson B.J., GoodrichL.V., Johnson R.L., Gates J., Scott M.P., Carroll S.B.(1999) Recruitment of a hedgehog Regulatory Circuit inButterfly Eyespot Evolution. Science 283, 532-534.

Kielan-Jaworowska Z., Bown T.M., Lillegraven J.A. (1979)Eutheria. In: Mesozoic Mammals The First Two-Thirdsof Mammalian History. eds. Lillegraven

J.A., Kielan-Jaworowska Z. Clemens W.A. University ofCalifornia Press, Berkeley, USA. pp. 221-258.

Kirschner M., Gerhart J. (1998) Evolvability. PNAS 95,8420-8427.

Kollar E.J., Baird G.R. (1970) Tissue interactions inembryonic mouse tooth germs. II. The inductive role ofthe dental papilla. J. Embryol. exp. Morph. 24, 173-186.

Kollar E.J., Mina M. (1991) Role of the early epithelium inthe patterning of the teeth and Meckel's cartilage. JCraniofac. Genet. Dev. Biol. 11, 223-228.

Kowalski K. (1976) Mammals An Outline of Theriology.(Translated from Polish Ssaki. Zarys Theriologii andpublished for the Smithsonian Institution and theNational Science Foundation, Washington D.C.),available from U.S. Department of Commerce, NationalTechnical Information Service, Springfield, USA.

Kowalski K. (1992) Bjornkurtenia, a new genus of primitivevoles of Europe (Rodentia, Mammalia). Ann. Zool.Fennici 28, 321-327.

Kozawa Y., Iwasa Y., Mishima H. (1998) Degeneration oftooth germ in the developing dentition of the gray short-tailed opossum (Monodelphis domestica). Eur. J. OralSci. 106 Suppl 1:509-512.

Kratochwil K., Dull M., Farinas I., Galceran J., GrosschedlR. (1996) Lef1 expression is activated by BMP-4 andregulates inductive tissue interactions in tooth and hairdevelopment. Genes Dev. 10, 1382-1394.

Kretzoi M. (1954) Promimomys cor n. g. n. sp., einAlterümlicher Arvicolide aus dem UngarischenUnterpleistozän. Acta Geol. Acad. Sci. Hungaricae 3,89-94.

Kronmiller J.E., Upholt W.B., Kollar E.J. (1991) EGFAntisense Oligodeoxynucleotides Block MurineOdontogenesis in vitro. Dev. Biol. 147, 485-488.

Kühl M., Wedlich D. (1997) Wnt signalling goes nuclear.Bioessays 19, 101-104.

Kurtén B. (1953) On the variation and population dynamicsof fossil and recent mammal populations. Acta Zool.Fennica 76, 1-122.

Kurtén B. (1968) Pleistocene Mammals of Europe. pp.215.Weiderfeld and Nicolson, London, GB.

Kälin D. (1999) Tribe Cricetini. In: Land Mammals ofEurope, The Miocene (eds. Rössner G.E. and HeissigK.), pp. 373-385. Verlag Dr. Friedrich Pfeil, Munich,Germany.

Lee S.M.K., Danielian P.S., Fritzsch B., McMahon A.P.(1997) Evidence that FGF8 signalling from themidbrain-hindbrain junction regulates growth andpolarity in the developing midbrain. Development 124,959-969.

Leroi A.M. (2000) The scale independence of evolution.Evol. & Devel. 2, 66-77.

Lesot H., Vonesch J.-L., Peterka M., Turecková J., PeterkováR., Ruch J.V. (1996) Mouse molar morphogenesisrevisited by three-dimensional reconstruction. II. Spatialdistribution of mitoses and apoptosis in cap to bellstaged first and second upper molar teeth. Int. J. Dev.Biol. 40, 1017-1031.

Lesot H., Peterková R., Viriot L., Vonesch J.L., TureckováJ., Peterka M., Ruch J.V. (1998) Early stages of toothmorphogenesis in mouse analyzed by 3Dreconstructions. Eur. J. Oral Sci. 106 Suppl. 1, 64-70.

Lesot H., Peterková R., Schmitt R., Meyer J.-M., Viriot L.,Vonesch J.-L., Senger B., Peterka M., Ruch J.-V. (1999)Initial features of the inner dental epithelium histo-morphogenesis in the first lower molar in mouse. Int. J.Dev. Biol. 43, 245-254.

Li W.-H., Graur D. (1991) Fundamentals of MolecularEvolution. Sinauer Associates, Sunderland, USA.

Lin C.R., Kioussi C., O'Connell S., Briata P., Szeto D., LiuF., Izpisúa-Belmonte J.C., Rosenfeld M.G. (1999) Pitx2regulates lung asymmetry, cardiac positioning andpituitary and tooth morphogenesis. Nature 401, 279-282.

Lough J., Barron M., Brogley M., Sugi Y., Bolender D.L.,Zhu X. (1996) Combined BMP-2 and FGF-4, butNeither Factor Alone, Induces Cardiogenesis in Non-Precardiac Embryonic Mesoderm. Dev. Biol. 178, 198-202.

Lu M.-F., Pressman C., Dyer R., Johnson R.L., Martin J.F.(1999) Function of Rieger syndrome gene in left-rightasymmetry and craniofacial development. Nature 401,276-278.

Luckett W.P. (1985) Superordinal and intraordinal affinitiesof rodents: developmental evidence from the dentitionand placentation. In: Evolutionary Relationships AmongRodents, NATO ASI Series, Ser. A, vol. 92. eds. LuckettW.P., Hartenberger J.-L., Plenum, New York, USA.

Luckett W.P., Zeller U. (1989) Developmental evidence fordental homologies in the monotreme Ornithorrhynchusand its systematic implications Z. Säugetierkunde 54,193-204.

Luckett W.P., Schrenk F., Maier W. (1989) On occurrence ofabnormal deciduous incisors during prenatal life inAfrican “hystricomorphous” rodents. Z. Säugetierkunde54, 296-303.

Luckett W.P. (1993) An Ontogenetic Assessment of DentalHomologies in Therian Mammals. In: MammalPhylogeny: Mesozoic Differentiation, Multituberculates,Monotremes, Early Eutherians, and Marsupials, eds.Szalay F.S., Novacek M.J., McKenna M.C., SpringerVerlag, New York, USA. pp. 182-204.

Ludwig M.Z., Patel N.H., Kreitman M. (1998) Functionalanalysis of eve stripe 2 enhancer evolution inDrosophila: rules governing conservation and change.Development 125, 949-958.

Lumsden A.G.S. (1988) spatial organization of theepithelium and the role of neural crest cells in theinitiation of the mammalian tooth germ. Development103(suppl.), 155-169.

86

Lyngstadaas S.P., MUinichen C.B., Risnes S. (1998) CrownMorphology, Enamel Distribution, and Enamel Structurein Mouse Molars. Anat. Rec. 250, 268-280.

Maas R., Bei M. (1997) The Genetic Control of Early ToothDevelopment. Crit. Rev. Oral Biol. Med. 8, 4-39.

MacDonald D.W., Barrett P. (1993) Collins Field GuideMammals of Britain & Europe. HarperCollinsPublishers, London, GB.

MacKenzie A. Ferguson M.W.J., Sharpe P.T. (1991) Hox-7expression during murine craniofacial development.Development 113, 601-611.

MacKenzie A., Ferguson M.W.J., Sharpe P.T. (1992)Expression patterns of the homeobox gene, Hox-8, in themouse embryo suggest a role in specifying toothinitiation and shape. Development 115, 403-420.

Mahmood R., Bresnick J., Hornbruch A., Mahony C.,Morton N., Colquhoun K., Martin P., Lumsden A.,Dickson C., Mason I. (1995) A role for FGF-8 in theinitiation and maintenance of vertebrate limb budoutgrowth. Curr. Biol. 5, 797-806.

Manak J.R., Scott M.P. (1994) A class act: conservation ofhomeodomain protein functions. Development Suppl.61-77.

Marazzi G., Wang Y., Sassoon D. (1997) Msx2 is aTranscriptional Regulator in the BMP4-MediatedProgrammed Cell Death Pathway. Dev. Biol. 186, 127-138.

Martin R.A. (1993) Patterns in variation and speciation inQuaternary rodents. In: Morphological change inQuaternary mammals of North America, eds. MartinR.A., Barnosky A.D. Cambridge Unicversity Press, NewYork, USA. pp. 226-280.

Matzuk M.M., Kumar T.R., Bradley A. (1995a) Differentphenotypes for mice deficient in either activins oractivin receptor type II Nature 374, 356-360.

Matzuk M.M., Kumar T.R., Vassalli A., Bickenbach J.R.,Roop D.R., Jaenisch R., Bradley A. (1995b) Functionalanalysis of activins during mammalian development.Nature 374, 354-356.

Matzuk M.M., Lu N., Vogel H., Sellheyer K., Roop D.R.,Bradley A. (1995c) Multiple defects and perinatal deathin mice deficient in follistatin. Nature 374, 360-363

Meng J., Wyss A.R., Dawson M.R., Zhai R. (1994) Primitivefossil rodent from Inner Mongolia and its implicationsfor mammalian phylogeny. Nature 370 134-136.

Meyer A., Málaga-Trillo E. (1999) Vertebrate genomics:More fishy tales about Hox genes. Curr. Biol. 9, R210-R213.

Meyers E.N., Martin G.R. (1999) Differences in Left-RightAxis Pathways in Mouse and Chick: Functions of FGF8and SHH. Science 285, 403-406.

Mina M., Kollar E.J. (1987) The induction of odontogenesisin non-dental mesenchyme combined with early murinemandibular arch epithelium. Arch. Oral Biol. 32, 123-127.

Mitsiadis T.A., Muramatsu T., Muramatsu H., Thesleff I.(1995) Midkine (MK), a Heparin-bindingGrowth/Differentiation Factor, Is Regulated by RetinoicAcid and Epithelial-Mesenchymal Interactions in theDeveloping Mouse Tooth, and Affects Cell Proliferationand Morphogenesis. J. Cell Biol. 129, 267-281.

Monaghan A.P., Davidson D.R., Sime C., Graham E.,Baldock R., Bhattacharya S.S., Hill R.E. (1991) TheMsh-like homeobox genes define domains in thedeveloping vertebrate eye. Development 112, 1053-1061.

Moon A.M., Boulet A.M., Capecchi M.R. (2000) Normallimb development in conditional mutants of Fgf4.Development 127, 989-996.

Morrison-Graham K., Schatteman G.C., Bork T., Bowen-Pope D.F., Weston J.A. (1992) A PDGF receptormutation in the mouse (Patch) perturbs the developmentof a non-neuronal subset of neural crest-derived cells.Development 115, 133-142.

Moss-Salentijn L. (1978) Vestigial Teeth in Rabbit, Rat andMouse; Their Relationship to the Problem in LactealDentitions. In: Development, Function and Evolution ofTeeth (eds. Butler P.M., Joysey K.A.) Academic Press,London pp. 13-29.

Mucchielli M.-L., Mitsiadis T.A., Raffo S., Brunet J.-F.,Proust J.P., Goridis C. (1997) Mouse Otlx2/RIEGExpression in the Odontogenic Epithelium PrecedesTooth Initiation and Requires Mesenchyme-DerivedSignals for Its Maintenance. Dev. Biol. 189, 275-284.

Murone M., Rosenthal A., de Sauvage F.J. (1999) Sonichedgehog signaling by the Patched-Smoothened receptorcomplex. Curr. Biol. 9, 76-84.

Neubüser A., Peters H., Balling R., Martin G.R. (1997)Antagonistic Interactions between FGF and BMPSignaling Pathways: A Mechanism for Positioning theSites of Tooth Formation. Cell 90, 247-255.

Nikoletopoulos N.P., Chondropoulos B.P., Fraguedakis-Tsolis S.E. (1992) Albumin evolution and phylogeneticrelationships among Greek rodents of the familiesArvicolidae and Muridae. J. Zool. 228, 445-453.

Niswander L., Martin G.R. (1993) FGF-4 and BMP-2 haveopposite effects on limb growth. Nature 361, 68-71.

Nowak R.M. (1991) Walker’s Mammals of the World, 5th

edition. The John Hopkins University Press, Baltimore,USA.

O'Brien S.J., Menotti-Raymond M., Murphy W.J., NashW.G., Wienberg J., Stanyon R., Copeland N.G., JenkinsN.A., Womack J.E., Marshall Graves J.A. (1999) ThePromise of Comparative Genomics in Mammals.Science 286, 458-462; 479-481.

Ohuchi H., Nakagawa T., Yamamoto A., Araga A., Ohata T.,Ishimaru Y., Yoshioka H., Kuwana T., Nohno T.,Yamasaki M., Itoh N., Noji S. (1997) The mesenchymalfactor, FGF10, initiates and maintains the outgrowth ofthe chick limb bud through interaction with FGF8, anapical ectodermal factor. Development 124, 2235-2244.

Oren A., Herschkovitz A., Ben-Dror I., Holdengreber V.,Ben-Shaul Y., Seger R., Vardimon L. (1999) TheCytoskeletal Network Controls c-Jun Expression andGlucocorticoid Receptor Transcriptional Activity in anAntagonistic and Cell-Type-Specific Manner. Mol. Cell.Biol. 19, 1742-1750.

Oro A.E., Higgins K.M., Hu Z., Bonifas J.M., Epstein E.H.Jr., Scott M.P. (1997) Basal Cell Carcinomas in MiceOverexpressing Sonic Hedgehog. Science 276, 817-821.

Orsulic S., Huber O., Aberle H., Arnold S., Kemler R.(1999) E-cadherin binding prevents β-catenin nuclear

87

localization and β-catenin/LEF-1-mediatedtransactivation. J. Cell Sci. 112, 1237-1245.

Osborn H.F. (1907) Evolution of Mammalian Molar Teeth.The Macmillan Company, New York, USA.

Osborn J.W. (1978) Morphogenetic Gradients: Fields versusClones. In: Development, Function and Evolution ofTeeth (eds. Butler P.M., Joysey K.A.) Academic Press,London pp. 171-201.

Osborn J.W. (1993) A Model Simulating ToothMorphogenesis without Morphogens. J. Theor. Biol.165:429-455.

Owen R. (1840-45) Odontography or a Treatise on theComparative Anatomy of the Teeth. Hippolyte Baillier,London.

Padgett R.W., Wozney J.M., Gelbart W.M. (1993) HumanBMP sequences can confer normal dorsal-ventralpatterning in the Drosophila embryo. PNAS 90, 2905-2909.

Palmeirim I., Henrique D., Ish-Horowicz D., Pourquié O.(1997) Avian hairy Gene Expression Identifies aMolecular Clock Linked to Vertebrate Segmentation andSomitogenesis. Cell 91, 639-648.

Parker S.B., Eichele G., Zhang P., Rawls A., Sands A.T.,Bradley A., Olson E.N., Harper J.W., Elledge S.J.(1995) p53-Independent Expression of p21Cip1 in Muscleand Other Terminally Differentiating Cells. Science 267,1024-1027.

Pearse R.V. II., Tabin C.J. (1998) The Molecular ZPA. J.Exp. Zool. 282, 677-690.

Penton A., Hoffmann F.M. (1996) Decapentaplegic restrictsthe domain of wingless during Drosophila limbpatterning. Nature 382, 162-164.

Peterková R. (1985) The Common Developmental Originand Phylogenetic Aspects of Teeth, Rugae Palatinae,and Fornix Vestibuli Oris in the Mouse. J. Craniofac.Genet. Dev. Biol. 5, 89-104.

Peterková R., Peterka M., Ruch J.V. (1993) Morphometricanalysis of potential maxillary diastemal dental anlagenin three strains of mice. J. Craniofac. Genet. Dev. Biol.13, 213-222.

Peterková R., Lesot H., Vonesch J.-L., Peterka M., Ruch J.-V. (1996) Mouse molar morphogenesis revisited bythree dimensional reconstruction. I. Analysis of initialstages of the first upper molar development revealed twotransient buds. Int. J. Dev. Biol. 40, 1009-1016.

Peterková R., Peterka M., Vonesch J.L., Turecková J., ViriotL., Ruch J.V., Lesot H. (1998) Correlation betweenapoptosis distribution and BMP-2 and BMP-4expression in vestigial tooth primordia in mice. Eur. J.Oral Sci. 106, 667-670.

Peters H., Balling R. (1999) Teeth where and how to makethem. Trends Genet. 15, 59-65.

Peters H., Neubüser A., Kratochwil K., Balling R. (1998)Pax9-deficient mice lack pharyngeal pouch derivativesand teeth and exhibit craniofacial and limbabnormalities. Genes Dev. 12, 2735-2747.

Pispa J., Jung H.-S., Jernvall J., Kettunen P., Mustonen T.,Tabata M.J., Kere J., Thesleff I. (1999) Cusp patterningdefect in Tabby mouse teeth and its partial rescue byFGF. Dev Biol. 216, 521-534.

Price J.A., Bowden D.W., Wright J.T., Pettenati M.J., HartT.C. (1998) Identification of a mutation in DLX3associated with tricho-dento-osseous (TDO) syndrome.Human Mol. Genet. 7, 563-569.

Quinn J.C., West J.D., Hill R.E. (1996) Multiple functionsfor Pax6 in mouse eye and nasal development. GenesDev. 10, 435-446.

Quinn J.C., West J.D., Kaufman M.H. (1997) Geneticbackground effects on dental and other craniofacialabnormalities in homozygous small eye(Pax6Sey/Pax6Sey) mice. Anat. Embryol. 196, 311-321.

Raff R.A. (1996) The Shape of Life Genes, Development,and the Evolution of Animal Form. The University ofChicago Press, Chicago, USA.

Reif W.-E. (1982) Evolution of Dermal Skeleton andDentition in Vertebrates: The Odontode RegulationTheory. Evol. Biol. 15, 286-368.

Rekovets L.I. (1994) History and phylogeny of Arvicolidae(Rodentia) in Pleistocene of Eastern Europe. PolishEcol. Studies 20, 553-558.

Richardson M.K., Allen S.P., Wright G.M., Raynaud A.,Hanken J. (1998) Somite number and vertebrateevolution. Development 125, 151-160.

Richman J.M., Herbert M., Matovinovic E., Walin J. (1997)Effect of Fibroblast Growth Factors on Outgrowth ofFacial Mesenchyme. Dev. Biol. 189, 135-147.

Robinson M., Catzeflis F., Briolay J., Mouchiroud D. (1997)Molecular Phylogeny of Rodents, with SpecialEmphasis on Murids: Evidence from Nuclear GeneLCAT. Mol. Phylog. Evol. 8, 423-434.

Rodríguez Esteban C., Capdevila J., Economides A.N.,Pascual J., Ortiz A., Izpisúa Belmonte J.C. (1999) Thenovel Cer-like protein Caronte mediates theestablishment of embryonic left-right asymmetry. Nature401, 243-251.

Roelink H., Augsburger A., Heemskerk J., Korzh V., NorlinS., Ruiz i Altaba A., Tanabe Y., Placzek M., Edlund T.,Jessell T.M., Dodd J. (1994) Floor Plate and MotorNeuron Induction by vhh-1, a Vertebrate Homolog ofhedgehog Expressed by the Notochord. Cell 76, 761-775.

Romer A.S. (1966) Vertebrate Paleontology, 3rd edition. TheUniversity of Chicago Press, Chicago, USA.

Roos C. (1995) Lääketieteen Nobelin palkinnotbanaanikärpäsen tutkijoille (editorial). Duodecim 111,2357, 2359, 2361, passim.

Rougier G.W., Wible J.R., Novacek M.J. (1998)Implications of Deltatheridium specimens for earlymarsupial history. Nature 396, 459-463.

Saha M.S., Spann C.L., Grainger R.M. (1989) Embryoniclens induction: more than meets the optic vesicle. CellDiff. Dev. 28, 153-171.

Sampath T.K., Rashka K.E., Doctor J.S., Tucker R.F.,Hoffmann F.M. (1993) Drosophila transforming growthfactor β superfamily proteins induce endochondral boneformation in mammals. PNAS 90, 6004-6008.

Sansom I.J., Smith M.P., Armstrong H.A., Smith M.M.(1992) Presence of the Earliest Vertebrate Hard Tissuein Conodonts. Science 256, 1308-1311.

88

Satokata I., Maas R. (1994) Msx1 deficient mice exhibit cleftpalate and abnormalities of craniofacial and toothdevelopment. Nature Genet. 6, 348-356.

Savage R.J.G., Long M.R. (1986) Mammal evolution: anillustrated guide. A copublication of Facts On File andThe British Museum (Natural History), UK.

Schmitt R., Lesot H., Vonesch J.-L., Ruch J.-V. (1999)Mouse odontogenesis in vitro: the cap-stagemesenchyme controls individual molar crownmorphogenesis. Int. J. Dev. Biol. 43, 255-260.

Schwartz J.H. (1982) Morphological Approach toHeterodonty and Homology. In: Teeth: Form, Functionand Evolution. eds. Kurtén B. Columbia UniversityPress, New York, USA. pp. 123-144.

Seale P., Rudnicki M.A. (2000) A New Look at the Origin,Function , and “Stem Cell” Status of Muscle SatelliteCells. Dev. Biol. 218, 115-124.

Semina E.V., Reiter R., Leysens N.J., Alward W.L.M., SmallK.W., Datson N.A., Siegel-Bartelt J., Bierke-Nelson D.,Bitoun P., Zabel B.U., Carey J.C., Murray J.C. (1996)Cloning and characterization of a novel bicoid-relatedhomeobox transcription factor gene, RIEG, involved inRieger syndrome. Nature Genet. 14, 392-399.

Sharpe P.T. (1995) Homeobox genes and orofacialdevelopment. Connect. Tissue Res. 32, 17-25.

Simpson G.G. (1944) The Tempo and Mode of Evolution.Columbia University Press, NewYork, USA.

Simpson P. (1996) Drosophila development: A prepatternfor sensory organs. Curr. Biol. 6, 948-950.

Simpson P., Woehl R., Usui K. (1999) The development andevolution of bristle patterns in Diptera. Development126, 1349-1364.

Smith M.M. (1995) Heterochrony in the evolution of enamelin vertebrates. In: Evolutionary Change andHeterochrony. eds. McNamara K.J., John Wiley & SonsLtd

Smith M.M., Coates M.I. (1998) Evolutionary origins of thevertebrate dentition: phylogenetic patterns anddevelopmental evolution. Eur. J. Oral Sci. 106 Suppl 1,482-500.

Srivastava A.K., Pispa J., Hartung A.J., Du Y., Ezer S., JenksT., Shimada T., Pekkanen M., Mikkola M.L., KoM.S.H., Thesleff I., Kere J., Schlessinger D. (1997) TheTabby phenotype is caused by mutation in a mousehomologue of the EDA gene that reveals novel mouseand human exons and encodes a protein (ectodysplasin-A) with collagenous domains. PNAS 94, 13069-13074.

ibid. Stanley, S. (1979) Macroevolution: Pattern and Process,San Francisco, W.H. Freeman and Company. p. 41

Štorba O. (1981) Prenatal development of dentition inMicrotus arvalis. Folia Zool. 30, 331-337.

St-Jacques B., Dassule H.R., Karavanova I., BotchkarevV.A., Li J., Danielian P.S., McMahon J.A., Lewis P.M.,Paus R., McMahon A.P. (1998) Sonic hedgehogsignaling is essential for hair development. Curr. Biol. 8,1058-1068.

Stock D.W., Ellies D.L., Zhao Z., Ekker M., Ruddle F.H.,Weiss K.M. (1996) The evolution of vertebrate Dlx genefamily. PNAS 93, 10858-10863.

Stock D.W., Weiss K,M., Zhao Z. (1997) Patterning of themammalian dentition in development and evolution.Bioessays 19, 481-490.

Stockton D.W., Das P., Goldenberg M., D’Souza R.N., PatelP.I. (2000) Mutation of PAX9 is associated witholigidontia. Nature Genet. 24, 18-19.

Strumpf D., Volk T. (1998) Kakapo, a Novel Cytoskeletal-associated Protein Is Essential for the RestrictedLocalization of the Neuregulin-like Factor, Vein, at theMuscle-Tendon Junction Site. J. Cell Biol. 143, 1259-1270.

Tabata M.J., Kim K., Liu J.G., Yamashita K., Matsumura T.,Kato J., Iwamoto M., Wakisaka S., Matsumoto K.,Nakamura T., Kumegawa M., Kurisu K. (1996)Hepatocyte growth factor is involved in themorphogenesis of tooth germ in murine molars.Development 122, 1243-1251.

Tabin C.J. (1992) Why we have (only) five fingers per hand:Hox genes and the evolution of paired limbs.Development 116, 289-296.

Thesleff I. (1995) Differentiation of Ameloblasts and ItsRegulation by Epithelial-Mesenchymal Interactions. In:Dental Enamel Formation to Destruction. eds. RobinsonC., Kirkham J., Shore R. CRC Press, Boca Raton, USA.pp. 1-19.

Thesleff I., Vaahtokari A., Kettunen P., Åberg T. (1995a)Epithelial-Mesenchymal Signaling during ToothDevelopment. Conn. Tissue Res. 32, 9-15.

Thesleff I., Vaahtokari A., Partanen A.-M. (1995b)Regulation of organogenesis. Common molecularmechanisms regulating the development of teeth andother organs. Int. J. Dev. Biol. 39, 35-50.

Thesleff I., Jernvall J. (1997) The Enamel Knot: A PutativeSignaling Center Regulating Tooth Development. in:Symposium 62: Pattern formation during development.Cold Spring Harbour Laboratory, Cold Spring Harbour,pp. 257-267.

Thesleff I., Sharpe P. (1997) Signalling networks regulatingdental development. Mech. Dev. 67, 111-123.

Thesleff I., Sahlberg C. (1999) Organ Culture in the Analysisof Tissue Interactions. Methods Mol. Biol. 97,Molecular Embryology Methods and Protocols, eds.Sharpe P.T., Mason I. pp. 23-31.

Thomas B.L., Tucker A.S., Qui M., Ferguson C.A.,Hardcastle Z., Rubenstein J.L., Sharpe P.T. (1997) Roleof Dlx-1 and Dlx-2 genes in patterning of the murinedentition. Development 124, 4811-4818.

Thomas B.L., Tucker A.S., Ferguson C., Qiu M., RubensteinJ.L.R., Sharpe P.T. (1998) Molecular control ofodontogenic patterning: positional dependent initiationand morphogenesis. Eur. J. Oral Sci. 106 Suppl 1, 44-47.

Thomas B.L., Sharpe P.T. (1998) Patterning of the murinedentition by homeobox genes. Eur. J. Oral Sci. 106Suppl 1, 48-54.

Tiffee J.C., Xing L., Nilsson S., Boyce B.F. (1999) DentalAbnormalities Associated with Failure of ToothEruption in src Knockout and op/op Mice. Calcif. TissueInt. 65, 53-58.

Tissier-Seta J.-P., Mucchielli M.-L., Mark M., Mattei M.-G.,Goridis C., Brunet J.-F. (1995) Barx1, a new mouse

89

homeodomain transcription factor expressed in cranio-facial ectomesenchyme and the stomach. Mech. Dev. 51,3-15.

Travis A., Amsterdam A., Belanger C., Grosschedl, R.(1991) LEF-1, a gene encoding a lymphoid-specificprotein with an HMG domain, regulates T-cell receptorα enhancer function. Genes Dev. 5, 880-894.

Tucker A.S., Al Khamis A., Sharpe PT. (1998a) InteractionsBetween Bmp-4 and Msx-1 Act to Restrict GeneExpression to Odontogenic Mesenchyme. Dev. Dyn.212, 533-539.

Tucker A.S., Matthews K.L., Sharpe P.T. (1998b)Transformation of Tooth Type Induced by Inhibition ofBMP Signaling. Science 282, 1136-1138.

Tucker A.S., Yamada G., Grigoriou M., Pachnis V., SharpeP.T. (1999) Fgf-8 determines rostral-caudal polarity inthe first branchial arch. Development 126, 51-61.

Tuisku F., Hildebrand C. (1994) Evidence for a NeuralInfluence on Tooth Germ Generation in a PolyphyodontSpecies. Dev. Biol. 165, 1-9.

Turecková J., Sahlberg C., Åberg T., Ruch J.V., Thesleff I.,Peterková R. (1995) Comparison of expression of themsx-1, msx-2, BMP-2 and BMP-4 genes in the mouseupper diastemal and molar tooth primordia. Int. J. Dev.Biol. 39, 459-468.

Vaahtokari A., Åberg T., Thesleff I. (1996) Apoptosis in thedeveloping tooth: association with an embryonicsignaling center and suppression by EGF and FGF-4.Development 122, 121-129.

Vainio S., Karavanova I., Jowett A., Thesleff I. (1993)Identification of BMP-4 as a Signal MediatingSecondary Induction between Epithelial andMesenchymal Tissues during Early Tooth development.Cell 75, 45-58.

van Genderen C., Okamura R.M., Farinas I., Quo R.G.,Parslow T.G., Bruhn L., Grosschedl R. (1994)Development of several organs that require inductiveepithelial-mesenchymal interactions is impaired in LEF-1-deficient mice. Genes Dev. 8, 2691-2703.

Vastardis H., Karimbux N., Guthua S.W., Seidman J.G.,Seidman C.E. (1996) A human MSX1 homeodomainmissense mutation causes selective tooth agenesis.Nature Genet. 13, 417-421.

Vogel A., Rodriguez C., Izpisuá-Belmonte J.-C. (1996)Involvement of FGF-8 in initiation, outgrowth andpatterning of the vertebrate limb. Development 122,1737-1750.

von Koenigswald W. (1982) Enamel Structure in the Molarsof Arvicolidae (Rodentia, Mammalia), a Key toFunctional Morphology and Phylogeny. In: Teeth: Form,Function and Evolution. eds. Kurtén B. ColumbiaUniversity Press, New York, USA. pp. 109-122.

Wall N.A., Hogan B.L.M. (1995) Expression of bonemorphogenetic protein-4 (BMP-4), bone morphogeneticprotein-7 (BMP-7), fibroblast growth factor-8 (FGF-8)and sonic hedgehog (SHH) during branchial archdevelopment in the chick. Mech. Dev. 53, 383-92.

Wang J., Walsh K. (1996) Resistance to ApoptosisConferred by Cdk Inhibitors during MyocyteDifferentiation. Science 273, 359-361.

Warren R., Nagy L., Selegue J., Gates J., Carroll S. (1994)Evolution of homeotic gene regulation and function inflies and butterflies. Nature 72, 458-461.

Wasserman J.D., Freeman M. (1998) An AutoregulatoryCascade of EGF Receptor Signaling Patterns theDrosophila Egg. Cell 95, 355-364.

Weatherbee S.D., Nijhout H.F., Grunert L.W., Halder G.,Galant R., Selegue J., Carroll S. (1999) Ultrabithoraxfunction in butterfly wings and the evolution of insectwing patterns. Curr. Biol. 9, 109-115.

Weil M., Itin A., Keshet E. (1995) A role for mesenchyme-derived tachykinins in tooth and mammary glandmorphogenesis. Development 121, 2419-2428.

Weiss K.M., Stock D.W., Zhao Z. (1998a) DynamicInteractions and the Evolutionary Genetics of DentalPatterning. Crit. Rev. Oral Biol. Med. 9, 369-398.

Weiss K., Stock D., Zhao Z., Buchanan A., Ruddle F.,Shashikant C. (1998b) Perspectives on genetic aspects ofdental patterning. Eur. J. Oral Sci. 106 Suppl 1, 55-63.

Wessels W., de Bruijn H., Hussain S.T., Leinders J.J.M.(1982) Fossil rodents from the Chinji Formation, BandaDaud Shah, Kohat, Pakistan. Proceedings of theKoninklike Nederlandse Akademie van Wetenschappen(B) 85(3):337-364.

Westergaard B., Ferguson M.W.J. (1986) Development ofthe dentition in Alligator mississippiensis. Earlyembryonic development in the lower jaw. J. Zool. (A)210, 575-597.

Westergaard B., Ferguson M.W.J. (1987) Development ofthe dentition in Alligator mississippiensis. Laterdevelopment in the lower jaws of embryos, hatchlingsand young juveniles. J. Zool. 212, 191-222.

Westergaard B., Ferguson M.W.J. (1990) Development ofthe Dentition in Alligator mississippiensis: Upper JawDental and Craniofacial Development in Embryos,Hatchlings, and Young Juveniles, With a Comparison toLower Jaw Development. Amer. J. Anat. 187, 393-421.

Wilson E.O. (1992) The Diversity of Life. Penguin Books,London, England.

Winnier G., Blessing M., Labosky P.A., Hogan B.L.M.(1995) Bone morphogenetic protein-4 is required formesoderm formation and patterning in the mouse. GenesDev. 9, 2105-2116.

Witter K., Míµek I., Peterka M., Peterková R. (1996) Stagesof odontogenesis in the field vole (Microtus agrestis,Rodentia) – a pilot study. Acta Vet. Brno 65, 285-296.

Wolpert L., Beddington R., Brockes J., Jessell T., LawrenceP., Meyerowitz E. (1998) Principles of Development.Current Biology Publications, London, UK.

Wozney JM. (1998) The bone morphogenetic protein family:multifunctional cellular regulators in the embryo andadult. Eur. J. Oral Sci. 106 Suppl 1, 160-166.

Xu P.-X., Zhang X., Heaney S., Yoon A., Michelson A.M.,Maas R.L. (1999) Regulation of Pax6 expression isconserved between mice and flies. Development 126,383-395.

Yang Y., Drossopoulou G., Chuang P.-T., Duprez D., MartiE., Bumcrot D., Vargesson N., Clarke J., Niswander L.,McMahon A., Tickle C. (1997) Relationship betweendose, distance and time in Sonic Hedgehog-mediated

90

regulation of anteroposterior polarity in the chick limb.Development 124, 4393-4404.

Zeller R., Haramis A.G., Zuniga A., McGuigan C., Dono R.,Davidson G., Chabanis S., Gibson T. (1999) Formindefines a large family of morphoregulatory genes andfunctions in establishment of the polarising region. CellTissue Res. 296, 85-93.

Zhang H., Bradley A. (1996) Mice deficient for BMP2 arenonviable and have defects in amnion/chorion andcardiac development. Development 122, 2977-2986.

Zhang Y., Zhao X., Hu Y., St. Amand T., Zhang M.,Ramamurthy R., Qiu M., Chen Y. (1999) Msx1 Is

Required for the Induction of Patched by SonicHedgehog in the Mammalian Tooth Germ. Dev. Dyn.215, 45-53.

Zhang Y., Zhang Z., Zhao X., Yu X., Hu Y., Geronimo B.,Fromm S.H., Chen, Y. (2000). A new function ofBMP4: dual role for BMP4 in regulation of Sonichedgehog expression in the mouse tooth germ.Development 127, 1431-1443.

Zúñiga A., Haramis A.-P.G., McMahon A.P., Zeller R.(1999) Signal relay by BMP antagonism controls theSHH/FGF4 feedback loop in vertebrate limb buds.Nature 401, 598-602.