evolution of the vertebral formulae in mammals: a...

Evolution of the Vertebral Formulae in Mammals:A Perspective on Developmental Constraints

YUICHI NARITA and SHIGERU KURATANIn

Laboratory for Evolutionary Morphology, Center for Developmental Biology,RIKEN, Kobe 650-0047, Japan

ABSTRACT Developmental constraints refer to biases that limit phenotypic changes duringevolution. To examine the contribution of developmental constraints in the evolution of vertebratemorphology, we analyzed the distribution pattern of mammalian vertebral formulae. Data onmammalian vertebral formulae were collected from the Descriptive Catalogue of the OsteologicalSeries Contained in the Museum of the Royal College of Surgeons of England by Richard Owen(1853) and were plotted onto the most reliable mammalian phylogenetic tree based on recentmolecular studies. In addition to the number of cervical vertebrae that is almost fixed to 7, we foundthat the number of thoracolumbar vertebrae tends to be 19 in many groups of mammals. Sincefidelity of the number of thoracolumbar vertebrae was also completely maintained in Monotremataand Marsupialia, we presumed that thoracolumbar vertebral number as well as cervical vertebralnumber might have been fixed in the primitive mammalian lineage. On the basis of primitivevertebral formulae, we could clarify the polarity of evolution and identify several deviations from theprimitive states during the mammalian evolution. The changes in the vertebral formulae ineutherian mammals seem to be lineage-specific, such that most species in Carnivora have 20 insteadof 19 thoracolumbar vertebrae. Because such lineage-specific vertebral formulae contrast with theestimated distribution pattern on the assumption of evolution only through the selective pressure,we concluded that developmental constraints played an important role in the evolution ofmammalian vertebral formulae. J. Exp. Zool. 304B:91–106, 2005. r 2005 Wiley-Liss, Inc.

INTRODUCTION

Evolution is most clearly perceived as changesin morphological patterns through phylogeny, andthe changes are often hierarchically distributedamong animals, conferring the hierarchy of tax-onomy (see Hall, ’94). For example, minor andrapid changes shared by a small number of speciestend to be associated with microevolution ordistinction of genera and species, whereas themajor and slow changes in global patterns of theanimal body are shared by a larger number ofspecies and are likely to be associated with thedifferences between higher taxa. These changesare hierarchical: the minor changes can beestablished on a higher level of changes. Themammalian heterodonty, for example, could ariseonly after the appearance of undifferentiateddentition on the jaw, the derivative of themandibular arch, which was once established asthe rostralmost element of the pharyngeal archesin ancestral vertebrates (Romer and Parsons, ’77;Hildebrand and Goslow, 2001; Kardong, 2002).Likewise, Prum (’99) proposed a hierarchical

model regarding the origin and evolution offeathers based on developmental pattern of birdfeathers. Such a hierarchical series of changes isisomorphic to the nested patterns of synapomor-phies that define monophyletic groups that arealso nested along the phylogenetic tree.

Importantly, morphological stasis over a longperiod also helps us perceive the taxonomic hier-archy; that is, there are patterns unchangedthrough evolutionary radiations, and these patternstend to be used as key characters to classify highertaxa. For example, all the land vertebrates are‘‘tetrapods’’ except for a few (such as snakes) thatlost their limbs secondarily after the diversificationfrom the tetrapod ancestor (Romer and Parsons,’77; Colbert et al., 2001). Here an importantquestion arises, as Raff (’96) has asked: why are

Grant sponsor: Grants-in-Aid to S.K. from the Ministry ofEducation, Science and Culture of Japan

nCorrespondence to: Shigeru Kuratani, Laboratory for Evolution-ary Morphology, Center for Developmental Biology, RIKEN, 2-2-3Minatojima-minami, Chuo-ku, Kobe, 650-0047, Japan.E-mail: [email protected]

Received 30 June 2004; Accepted 21 October 2004Published online 19 January 2005 in Wiley InterScience (www.

interscience.wiley.com). DOI: 10.1002/jez.b.21029

r 2005 WILEY-LISS, INC.

JOURNAL OF EXPERIMENTAL ZOOLOGY (MOL DEV EVOL) 304B:91–106 (2005)

there neither ‘‘hexapods’’ nor ‘‘octapods’’ but onlytetrapods among land vertebratesFwhy are certainpatterns conserved very rigidly through evolution?As has been discussed and reviewed by many

researchers, there are two possible answers to thisquestion: stabilizing selection and developmentalconstraints (Maynard-Smith et al., ’85; Alberch,’89; Arthur and Farrow, ’99). In the stabilizingselection view, it is assumed that, in addition totetrapods, hexapods and octapods may have beenproduced during evolution. However, the nonte-trapod patterns were eliminated because theywere less adaptive than tetrapods, resulting inall extant land vertebrates’ being tetrapods. Thus,in this theory, various changes are assumed at thetime of radiation and secondarily screened byselective pressures, leaving only a few survivingvariants. In contrast, in the developmental con-straint view, the production of phenotypic varia-tions is assumed to be limited by certain unknowndevelopmental restrictions: hexapod or octapodanimals are supposed to be rarely producedbecause of limitations in the possible develop-mental patterns.Although both of these concepts are important in

evolutionary studies, few reports analyze develop-mental constraints. One example for developmen-tal constraints involves the number of mammaliancervical vertebrae. The mammalian vertebralcolumn consists of morphologically differentiatedgroups of vertebrae: cervical, thoracic, lumbar,sacral, and caudal (Gaunt, ’94; Burke et al., ’95)(Fig. 1A). The number of mammalian cervicalvertebrae is almost constant at seven, irrespectiveof the neck length of different speciesFgiraffeswith long necks and whales with short necks bothpossess seven cervical vertebrae (Fig. 1), as hasdescribed in several classic textbooks (Owen, 1853,1866; Remane, ’36). In contrast, birds and reptilesoften have variable numbers of cervical vertebrae(Owen, 1853, 1866; Remane, ’36). This suggeststhat the number of mammalian cervical vertebraehas been almost fixed by certain mammalian-specific developmental constraints (Galis, ’99b).Thus, the evolution of mammalian vertebral for-mulae may serve as a valuable model for studyingthe evolution of developmental constraints.Specification of the vertebral column is devel-

opmentally regulated by the Hox code, a nestedexpression pattern of Hox genes along the ante-rior–posterior axis of the embryo. Genetic manip-ulation of the Hox genes can cause ‘‘morphologicaltransformation’’ (Rosenberg, 1875; Goodrich, ’30)of the vertebral column in the mouse (Carrasco

and Lopez, ’94; Maconochie et al., ’96). It has alsobeen reported that differences in vertebral speci-fication among animals correlate with differencesin expression patterns of the Hox genes (Gaunt,’94; Burke et al., ’95; Galis, ’99a; Cohn and Tickle,’99). For example, in both mouse and chicken, theanterior expression boundary of HoxC6 falls at thesame morphological boundary between the cervi-cal and thoracic region, even though the positionsof the boundary lie at different axial levels interms of the segmental number of somites: theseventh vertebra is the last cervical one in themouse, whereas the 14th is the last cervicalvertebra in the chicken. Thus, the homology ofvertebrae (including the occipital bone) originatesfrom the homology of Hox genes expressed at eachlevel. This rule also applies to other nonmodelspecies of vertebrates (Burke et al., ’95).

Apparently, the mechanism of somite segmenta-tion also plays a crucial role in the development ofvertebrae. It has been reported that disruption ofsome segmentation genes, such as Delta 1 andWnt-3a, leads to homeotic transformations ofvertebrae (Ikeya and Takada, 2001; Cordes et al.,2004). Importantly, however, such transforma-tions are always accompanied by changes in Hoxgene expression patterns, which are tightly linkedwith somitogenesis per se: Hox gene expression isalso cyclical as are other segmentation genes(Zakany et al., 2001).

DEVELOPMENTAL CONSTRAINTS ANDTHEIR ANALYSES

Developmental constraints are defined as biaseson the production of phenotypic variants orlimitations on phenotypic variability caused bythe structure, character, composition, or dynamicsof the developmental system (Maynard-Smithet al., ’85). Although developmental constraintsapparently play an important role in evolution,less attention has been paid to the concept than toselection theory. This might be due in part to theconceptual ambiguity as well as the lack ofestablished methods to analyze them (Maynard-Smith et al., ’85; Schwenk, ’95; Arthur, ’97;Schwenk and Wagner, 2003). Although someresearchers counted only biases that originallylimit the production of phenotypic variants duringontogeny as ‘‘real’’ developmental constraints(Gould, ’89), from the currently available evidenceit is difficult to clarify the underlying mechanismof evolutionary stasis and distinguish the ‘‘realconstraints’’ from the effect of internal selection.

Y. NARITA AND S. KURATANI92

Richardson and Chipman (2003) mentioned thatthere is no clear line between constraints andselection and that, in many instances, develop-mental constraints result from selection acting atthe embryonic stages. For example, behind thepanmammalian constraint on the number ofcervical vertebrae, a hypothetical mechanism hasbeen suggested: Galis (’99b) has pointed out thatcervical ribs would be associated with neurologicaldisorders leading to thoracic outlet syndrome andthat the increased appearance of cervical trans-formations in children with embryonal cancers.With this evidence, she has suggested that changesin the cervical vertebral number may be coupledwith the neuronal problem and an increased riskfor neonatal cancer because of the pleiotropic

function of the Hox gene. These might workagainst the survival of the individuals withcervical ribs as direct and indirect selections. Herconclusion is that the increased costs mighthave prevented the different numbers of cervicalvertebrae being fixed in a population, thus estab-lishing the mammalian-specific developmentalconstraint.

The above argument also favors the idea ofstructural constraints. For example, a highlycomplicated pattern whose components are struc-turally interrelated with each other is hard tomodify because a small change in one element isapt to destroy the whole system. Likewise, ananimal group classified by its specific body planmay possess its own structural integrity that has

Fig. 1. Skeletons of giraffe and whales. (A) Skeleton of giraffe (Giraffa camelopardalis). Long neck of the giraffe consists ofseven cervical vertebrae (Cv). Redrawn from Owen (1866). (B) Skeleton of a balaenopterid whale. Seven separate cervicalvertebrae (Cv) are seen. Redrawn from Matthews (’68). (C) Skeleton of grampus (Grampus griseus). Cervical vertebrae aresecondarily fused with each other to form a single bone. Redrawn from Howell (’30). (D) Magnification of the fused cervicalvertebrae in the bowhead whale, Balaena mystricetus. Redrawn from Owen (1866). (1–7) Numbers of cervical vertebrae.

CONSTRAINTS ON THE MAMMALIAN VERTEBRAL COLUMN 93

been conserved in a specific developmental con-text. In this article, we do not distinguish ‘‘real’’constraints from those resulting from internalselection but define developmental constraints asany kinds of bias that act during embryogenesis,including the lack of variations because of adevelopmental limitation or the elimination ofvariant embryos by internal selection againstthem, thus leading to a group-specific evolutionaryfuture (Spurway, ’49).One method of detecting the presence of

constraints is through phylogenetic approaches(McKitrick, ’93; Arthur and Farrow, ’99). In thismethod, the distribution of both traits andpossible selective pressure are plotted onto aphylogenetic tree to test the null hypothesis thatthe phylogenetic distribution of a trait is deter-mined by selection. Some studies have used thisapproach to show the presence of constraints(Wagner and Muller, 2002; Richardson and Chip-man, 2003). This approach is based on theassumption that developmental constraints shouldaffect the relative frequency of transformations(Richardson and Chipman, 2003). In this study, wehave used a similar approach to analyze theevolutionary pattern of mammalian vertebralformulae along the phylogenetic pathways and toanalyze whether hierarchical pattern can beidentified also in the evolution of the mammalianvertebral formulae. Instead of plotting the selec-tive pressure onto the tree, we deduced thepossible distribution patterns of vertebral formu-lae based on the evolution of other amniotes aswell as the reported results of artificial selectionexperiments. Then, to analyze the contribution ofdevelopmental constraints in mammalian verteb-ral column evolution, we tested whether theobserved evolutionary pattern could be broughtabout only through the effect of selective pressure,or not, by comparing the observed distributionpatterns with the estimated possible patterns.A prerequisite for this approach is a reliable

phylogeny of the groups of interest. Traditionally,phylogenetic relationships among mammals havebeen inferred by morphology and paleontology.Novacek (’92) reviewed these studies and repre-sented a morphology-based mammalian phylo-genic tree. Since then, molecular phylogeneticanalyses have been carried out enthusiastically,and recently relationships among a majority ofmammalian orders have been fairly settled (Jankeet al., ’96, ’97; Cao et al., 2000; Murphy et al.,2001; Killian et al., 2001; Lin et al., 2002) althoughthere are still some controversies between the

different approaches and thus some relationshipsstill remain unclarified (Allard et al., ’99; Cao et al.,2000). From these data, we obtained the phyloge-netic tree of mammals that is currently the mostreliable. We used this in analyzing the evolution-ary pattern of the mammalian vertebral formulae.

EVOLUTION OF THE MAMMALIANVERTEBRAL FORMULA

Primitive vertebral formula for mammals

We collected vertebral formula data for mam-mals (Table 1) from the Descriptive Catalogue ofthe Osteological Series Contained in the Museumof the Royal College of Surgeons of England byRichard Owen (1853). This catalogue containsdata of 133 species from 15 orders of mammalsand serves as a rich resource with which to surveythe phylogenetic changes in developmental con-straints. It should be noted that the catalogueincludes only a limited data for individual varia-tions within species: a certain amount of variationhas been reported within some mammalian spe-cies based on more comprehensive analyses (Sa-win, ’37; Aimi, ’94; Pilbeam, 2004), whereas in thiscatalogue, intraspecies variations were found onlyin 6 out of the 26 species from which more thantwo samples were examined (variations in caudalvertebral numbers are not counted). As anexample of problematic cases in this catalogue,gorillas generally have 13 thoracic and 4 lumbarvertebrae (Pilbeam, 2004) instead of 12 thoracicand 5 lumbar vertebrae as Owen (1853) hasdescribed. In the latter case, however, the sum ofthoracolumbar vertebrae remained 17, as gener-ally found in the subspecies Gorilla gorilla gorilla(Pilbeam, 2004; see below). It may have been thatthe caudalmost ribs were lost in that particularspecimen. Needless to say, further collection ofvertebral formulae data will be desired for futurestudy. Nevertheless, for several selected samplespecies, data were comparable to the knownstandard vertebral formulae elucidated by morerecent studies (Sawin, ’37; Aimi, ’94; Burke et al.,’95; Pilbeam, 2004). As will be shown below,numbers of thoracolumbar domain tended to beconstant at the levels of families and orders. Thusthe formulae described in this catalogue appearedtrustworthy, to a certain extent, for the purpose ofthe present study, especially to gain a glimpse atthe patterns of formulae changes on the mamma-lian phylogeny, which could be associated with theestablishment of order-specific morphological plans.


TABLE1.Vertebral

form

ulae

inmam

mals

Vertebralform

ulada

taof133speciesfrom15

mam

malianorde

rswerecollectedfrom

theDescriptiveCatalogueoftheOsteologicalSeriescontained

intheMuseum

oftheRoya

lCollegeofSurgeonsofEngland(O

wen,1853).C

lassi¢cation

andscienti¢cn

ames

fore

achspeciesw

ereu

pdated

basedon

Corbeta

ndHill

(’86).T

he‘No.’columnrepresentstheserial

numberin

theoriginal

catalogu

eforeach

specim

en.T

heblan

kcolumnrepresentslack

of,orincomplete,da

tadu

eto

thecond

itionof

thespecim

en.T

henu

mbers

sepa

ratedby

aslash

mark(/)represent

thevertebralformulae

that

di¡eram

ongindividu

als.The

cervical

columnisredwhe

nthenu

mberofcervical

vertebraeis7.The

thoracican

dlumba

rcolumns

are

yellow

whe

nthesum

is19

andblue

whe

nthesum

is20.


TABLE1.

Con

tinu

edY. NARITA AND S. KURATANI96

The vertebral numbers were plotted on aphylogenetic tree of mammalian orders basedmainly on recent molecular phylogenetic analyses(Janke et al., ’96, ’97; Cao et al., 2000; Murphyet al., 2001; Killian et al., 2001; Lin et al., 2002).There is a clear tendency for the number ofcervical vertebrae in mammals to be seven.Deviations from this number are identified in onlytwo lineages: the manatee (Trichechus manatus,Sirenia), which has six cervical vertebrae, and thethree-toed sloth (Bradypus tridactylus, Xenar-thra), which has nine (Table 1, Fig. 2). Theseanimals do not seem to represent the state beforethe fixation of the cervical number. Rather, thefixation of seven as the common number ofcervical vertebrae appears to have occurred nearthe common ancestor of mammals, includingmonotremes (Fig. 2; Owen, 1853).

In contrast to the fixed number of cervicalvertebrae, the numbers of thoracic, lumbar, sacral,and caudal vertebrae fluctuate among mammalianorders and families (Table 1). As functionallyimportant traits are likely to be highly canalized toexhibit greater developmental stability (Hall-grimsson, 2003), the higher degree of variationin the postcervical vertebral number implieseither that there are looser functional restrictionson these traits and thus less canalization of thedevelopmental process or that functionally opti-mal numbers of these vertebrae frequently chan-ged through evolution.

Interestingly, however, we found that the sumof thoracic and lumbar vertebrae tends to be 19 inmany groups of mammals (Fig. 2). This finding isconsistent with previous preliminary reports,which have implied that the number of thoraco-lumbar vertebrae tends to remain fairly constantwithin some eutherian orders (Owen, 1853; Romerand Parsons, ’77). It has been reported that theearliest known eutherian mammal, Eomania, alsohad 19 thoracolumbar (13 thoracic and 6 lumbar)vertebrae (Ji et al., 2002). Furthermore, thisnumber was completely constant in monotremesand marsupials (Table 1). In contrast, the numberof trunk vertebrae in reptiles and birds, whichcorrespond to mammalian thoracolumbar verteb-rae, is highly variable among species (Owen, 1853)(data not shown). Based on the constancy not onlyin the eutherians but also in these two relativegroups, we presumed that the number of thor-acolumbar vertebrae as well as the cervicalnumber might have already been fixed in theprimitive mammalian lineage as plesiomor-phic traits for mammals. On the basis of the

TABLE1.

Con

tinu

edY. NARITA AND S. KURATANI98

mammalian primitive vertebral formulae, wecould clarify the polarity of evolution and identifyseveral deviations from the primitive states duringthe mammalian evolution.As compared to the relatively fixed number of

thoracolumbar vertebrae, the respective numbersof thoracic and lumbar vertebrae tend to change.Some intraspecies variations seem to have re-

sulted from ‘‘developmental trade-off’’ betweenthe thoracic and lumbar vertebrae to maintain thesum of them at 19, as seen in the rat (Rattusnorvegicus, Rodentia; Table 1) and wild boar (Susscrofa, Cetartiodactyla; Table 1). However, thereare exceptions, such as the house mouse (Musmusculus, Rodentia) and the western Europeanhedgehog (Erinaceus europaeus, Eulipotyphla), in

Fig. 2. Distribution of mammalian vertebral formulae along the phylogeny. The vertebral formula of each animal wasplotted onto a mammalian phylogenetic tree based mainly on recent molecular analyses (Janke et al., ’96, ’97; Cao et al., 2000;Murphy et al., 2001; Killian et al., 2001; Lin et al., 2002). The branches with bold lines represent the lineages that retain theprimitive vertebral formula, and those with thin lines represent deviation from the primitive state.


which the number of lumbar vertebrae varies inthe species whereas the number of thoracicvertebrae remains constant. The sum of thesetwo types of vertebra appears to be highlycanalized, but the segmental level of transitionbetween the two vertebral identities is not and ismore likely to fluctuate in both development andevolution.The distribution pattern of changes in the

number of thoracolumbar vertebrae seems to favorthe idea of developmental constraint-based stasis:the changes in the vertebral formulae in eutherianmammals do not seem to occur in each speciesrandomly but are limited to certain taxa, allowingus to locate the evolutionary events at specificsegments or branches in the phylogenetic tree (Fig.2). In turn, there appear to be lineage-specificdevelopmental constraints that cannot easily beovercome, as members of the group radiate withthe same basic body plan (which does not necessa-rily include the vertebral formula per se) shared bythe group. We further discuss the contribution ofdevelopmental constraints in mammalian verte-bral formulae in ‘‘Selection or constraints?’’ below.

Deviation from the primitive formula

One group that shows deviation from the primi-tive state is Carnivora. In this group, most of thespecies that we analyzed possess 20 instead of 19vertebrae in the thoracolumbar region (Table 1).We predict that the common ancestor of carnivoresacquired a new developmental program to specifythe vertebral column, and this has been main-tained during the radiation of Carnivora (Fig. 2).In contrast, in Primates, a secondary decrease inthe number of thoracolumbar vertebrae to 17 or 18can be detected in the hominoid lineage (Hyloba-tidae + Pongidae) (Fig. 3A).Deviation from the primitive formula can also

be identified in a sublineage of Cetartiodactyla(Fig. 3B). Recent phylogenetic studies have re-solved the relationships between cetaceans and

artiodactyls by analyzing insertion patterns ofshort and long interspersed elements (Nikaidoet al., ’99), and it is now generally accepted thatcetaceans are deeply nested within the artiodac-tyls. Despite the phylogenetic position of ceta-ceans, their vertebral formulae deviate from theprimitive formula shared by all the artiodactyls:cetaceans have lost specification of the vertebralcolumn at the post-thoracic level. A similar loss ofspecification in posterior vertebrae is also encoun-tered in sirenians. These deviations from theprimitive formula may be related to the fact thatthese two groups are aquatic and possess bodyplans unique to these taxa. It is interesting thatthe pinnipeds (walrus and seals) retain theCarnivora-specific formula, although these ani-mals have also adapted to an aquatic niche. Thedifference seems to be related to the fact that thepelvic girdle is still attached to the vertebralcolumn in pinnipeds but is detached in the othertwo groups. The lack of evidence of convergencebetween the vertebral columns of pinnipeds andthose of cetaceans and sirenians might oppose theselection concept and supports the contribution ofdevelopmental constraints in the mammalianvertebral column evolution.

An outstanding increase in the number ofthoracolumbar vertebrae is found in Afrotheria.This superordinal clade was suggested by recentmolecular phylogenetic analyses (Stanhope et al.,’98; Cao et al., 2000; Murphy et al., 2001) andincludes six mammalian orders: Hyracoidea (hyr-axes), Sirenia (dugongs and manatees), Probosci-dea (elephants), Tublidentata (aardvarks),Afrosoricida (golden moles and tenrecs), andMacrocelidea (elephant shrews). We analyzed thevertebral formulae of the first four orders.Tublidentata, which is thought to have divergedfrom other afrotherians first, has a slightlyincreased number of thoracolumbar vertebrae(21). However, the other orders show a dramaticincrease in thoracolumbar vertebrae, and thecape hyrax (Hyracoidea) has 30 thoracolumbar

Fig. 3. Deviation from the primitive mammalian vertebral formula. C and TL represent the number of cervical andthoracolumbar vertebrae, respectively. (A) Evolution of vertebral formulae in Primates. Primate phylogenetic relationship wasbased on Purvis (’95) and Shoshani et al. (’96). Note that the number of thoracolumbar vertebrae was decreased specifically inhominoid lineage (Hylobatidae + Pongidae). (B) Evolution of vertebral formulae in Cetartiodactyla. Phylogenetic relationshipsamong cetaceans and between artiodactyls and cetaceans were based on SINE (short interspersed element) and LINE (longinterspersed element) insertion patterns (Nikaido et al., ’99) as well as molecular phylogenetic analyses (Cao et al., 2000;Murphy et al., 2001; Lin et al., 2002). Internal relationships among ruminants were based on Beintema et al. (2003). Note that,although cetaceans are deeply nested within artiodactyls, the vertebral formula of the cetacean deviates from those ofartiodactyls and constitutes a unique body plan. (C) Evolution of vertebral formulae in Xenarthra. Phylogenetic relationshipsamong xenarthrans were based on Delsuc et al. (2002). Note that vertebral formulae of this group have secondarily deviatedfrom the primitive state in many directions.


vertebrae. The exact number of thoracolumbarvertebrae in sirenians could not be countedbecause of the loss of specification in the posterior

vertebral column. Nevertheless, comparison withrelated animals permits us to speculate that thenumber of sirenian thoracolumbar vertebrae


would be much larger than 19, judging from theremarkable increase in their thoracic vertebrae(17 in manatee and 19 in dugong). Similarly, anincrease in the number of thoracolumbar verteb-rae can be identified in perissodactyls (22–24).However, these are likely to be independentevolutionary events, because Perissodactyla be-longs to Laurasiatheria, which is distantly relatedto the lineages in Afrotheria.In Xenarthra, the number of thoracolumbar

vertebrae tends to vary (Fig. 3C). Armadillos(Dasypodidae), for example, have a reduced num-ber of thoracolumbar vertebrae (14–15), whereasthe two-toed sloth (Megalonychidae) retains asmany as 26 vertebrae in the same region.Furthermore, in the three-toed sloth (Bradypodi-dae), the number of cervical vertebrae has in-creased to nine. Such a wide variation in thexenarthran vertebral formulae implies that theprimitive mammalian vertebral formulae mighthave been canceled specifically in this monophy-letic group, suggesting a possibility that releasefrom the primitive formula also takes place group-specifically. An analogous situation can found inlineages of birds and reptiles other than turtles(turtles constantly possess eight cervical and tenthoracic vertebrae).This study also showed that rodents maintained

the primitive feature of mammalian vertebralformulae, in which the numbers of cervical andthoracolumbar vertebrae are fixed at 7 and 19,respectively (Table 1, Fig. 2), although they arethought to be a highly derived lineage amongmammals. Therefore, the phylogenetic position ofa group is not always consistent with, or correlatedwith, the primitiveness of its body plan (known asmosaic evolution). The latter rather depends onthe changes brought into its developmental pro-gram.

DEVELOPMENTAL CONSTRAINTS ANDTHE VERTEBRAL FORMULA

Selection or constraints?

Could the lineage-specific fixation and changesin vertebral numbers be explained through thetaxon- or lineage-specific selections that wouldlead to the invariable number of certain verteb-rae? This possibility deserves to be tested. In thefollowing discussion, we define ‘‘group-specificselection’’ as the selective pressure that may becaused by a common morphological feature orhabitat of the group and leads the group to the

changes in axial morphological traits, such as therelative length of the neck. If the group-specificvertebral formula has arisen through group-specific selection, would the changes in axialmorphological traits always correlate directly withthe changes in vertebral formulae or not correlatewith the formulae at all?

In nonmammalian amniotes, such a correlationcan be encountered in some cases, as is typicallyseen in the evolution of plesiosaurs (Fig. 4A–C).This group (superfamily, Plesiosauroidea) seemsto have originated from nothosaurs in the middleTriassic, which had necks with about 17 cervicalvertebrae (Fig. 4A) (Carroll and Gaskill, ’85). Theprimitive plesiosaurs in the late Jurassic acquiredlonger necks by an increase of cervical vertebraeup to 30 (Fig. 4B) (Andrews, ’10; Brown, ’81), andmore specialized species, such as elasmosaurid inthe late Cretaceous, had extremely long neckswith as many as 70 cervical vertebrae (Fig. 4C)(de Saint-Seine, ’55; Ginsburg, ’70). Thus, in thisgroup, the elongation of the neck synchronizeswith an increased number of cervical vertebrae;the external morphology is in parallel with thechanges in vertebral formulae.

However, a change in axial morphological traitsdoes not necessarily associate with changes in thevertebral formulae in other situations, such asartificial selection. For example, Rutledge et al.(’74) analyzed the effect of experimental selectionfor increased tail length using an inbred strain ofmice. After seven generations of selection, theresearchers produced two different lines of micewith longer tails than the nonselected lines.Although the two lines did not differ in tail length,they did differ in the anatomy of elongated tails: inone line, the number of caudal vertebrae hadincreased, whereas in the other, the number ofcaudal vertebrae remained the same as in thenonselected lines but the length of each caudalvertebra had increased. Thus, the developmentalbases for increased tail length clearly differbetween these two lines. The result suggests thatthe genome can potentially respond to the sameselective pressure by at least two different path-ways of development.

It is very likely that such potential plasticity indevelopmental mechanism also played importantroles in phylogenetic evolution. Similarly to thetwo different selection responses toward tailelongation, another pattern of neck evolution isfound in the Mesozoic aquatic reptiles. Forexample, the necks of Tanystropheus from theearly to middle Triassic (Fig. 4D) had elongated by


increasing the length of each cervical vertebra(Peyer, ’31; Wild, ’73); unlike in plesiosaurs, thenumber of cervical vertebrae remained around 10.The method of neck elongation reminds us of apresumed analogous developmental response inthe emergence of the giraffe (Fig. 1A). Suchdifferent modes of changes in vertebral formulaehave also been reported in a group of salamanderthat acquired an elongated trunk during evolution(Parra-Olea and Wake, 2001).Overall, these evidences indicate that, during

the evolution of body plans in amniotes, thedevelopmental program can potentially respondto the same selection pressure in at least twodifferent modes. In other words, the developmen-tal mechanism in response to a specific selection

pressure can have a wider range of variationsthan the equally adaptive patterns of pheno-types required by the selection. Therefore, eventhough we accept the existence of ‘‘group-specificselection,’’ we can expect that selective pressuremay still create varied numbers of vertebraeand keep changing together with the axial bodyplans throughout the evolution of that lineage.The expected distribution pattern of vertebralformulae, which was estimated on the assumptionof the null hypothesis that the evolutionarypattern of mammalian vertebral formulae is deter-mined only by selective pressures, contrastswith the observed evolutionary pattern of themammalian vertebral column. Therefore, it islikely that developmental constraints played

Fig. 4. Skeletons of Mesozoic aquatic reptiles. (A) Skeleton of nothosaur, Pachypleurosaurus. Redrawn from Carroll andGaskill (’85). (B) Skeleton of the plesiosauroid Cryptoclidus. Redrawn from Andrews (’10). (C) Skeleton of elasmosauridHydrothecrosaurus. Redrawn from de Saint-Seine (’55). (D) Skeleton of Tanystropheus. Redrawn from Wild (’73). Note that thenumber of cervical vertebrae increased with neck elongation in the first three species. In Tanystropheus, however, the length ofeach cervical vertebra increased.


important roles in mammalian vertebral formulaeevolution and gave rise to the group-specificvertebral formulae.Evolutionary patterns of vertebral formulae in

marsupials also suggest the presence of develop-mental constraints. As an apt illustration forconvergence or parallel evolution, marsupials haveradiated into diversified morphologies adapted todistinct habitats in a pattern similar to that ineutherian mammals. If the evolution of vertebralformulae tends to change under taxon-specificselection toward the taxon-specific habitat andmorphology (which should be based on the samespecific vertebral formulae), the constant verteb-ral formulae in marsupials and variable vertebralformulae in placentals could not be reconciled.Thus, the stasis in the vertebral formulae ofmarsupials contrasts with the expected level ofvariation based on the assumption of selection. Itis again much more appropriate to consider thatthe stasis has occurred as a result of marsupial-specific (and mammalian primitive) developmen-tal constraints that could not be overcome in theradiation of marsupial lineages.Evolutionary events on hypothetical develop-

mental constraints identified in the present studyseem to appear hierarchically along the phyloge-netic evolution (Figs. 2 and 3). However, theevolutionary changes of vertebral formulae neverconstitute a hierarchy by themselves. Instead, theprimitive vertebral formulae are often secondarilycancelled, and a new taxon can evolve by acquiringa new body plan based on a new vertebral formula.For instance, cetaceans deviated from the primi-tive vertebral formulae that was shared byartiodactyls and acquired a unique body plan(Fig. 3B). We cannot view the cetacean vertebralformula as a sophisticated version of the artiodac-tyl formula. Therefore, evolutionary patterns ofthe mammalian vertebral column do not exhibithierarchical constitution and do not imply polarityof evolution per se but are more like the descrip-tion of body plans in taxonomy.

Evolution of the Hox code in mammals

It may be possible to regard the evolutionaryhistory of developmental constraints on mamma-lian vertebral formulae as the trail of Hox codeevolution in mammals since differences in verteb-ral specification are thought to be caused by thechanges inHox codes (Gaunt, ’94; Burke et al., ’95;Galis, ’99a; Cohn and Tickle, ’99). For instance,the decrease in cervical vertebral number in the

manatee may have been brought about by ananterior shift of the HoxC6 expression boundary.Further, triple knockout of Hox11 paralogs in themouse results in the transformation of sacral andsome caudal vertebrae into the lumbar vertebrae;i.e., complete loss of specification in the posteriorvertebral column (Wellik and Capecchi, 2003). Thelatter mutant may be viewed as a possiblephenocopy of cetaceans and sirenians in bothevolutionary and developmental senses. Thus,the unique body plans of these animals may bebrought about either by their own changes inHox11 paralogs or their regulation, or the samephenotype could have been obtained throughmodification of the post-Hox code programs indevelopment.

Furthermore, some reports have recentlysuggested the possibility that divergence incis-regulatory elements of Hox genes wouldhave played a critical role in changes of Hoxgene expression pattern among the animals(Belting et al., ’98). With regard to mammalianvertebral formulae evolution, the next and themost important step might be to analyze compa-ratively the enhancer activities of some Hoxgenes that have central functions in vertebralspecification, such as HoxC6. This may providean opportunity to analyze the mechanism of theevolution of developmental constraints in termsof gene regulation. On the other hand, preciseanalysis of the intraspecies variation in Hoxgene expression patterns may also provideclues to understanding the mechanism behindthe developmental constraints on mammalianvertebral formulae. If the Hox expression patternis completely constant and no variation is identi-fied among individuals, this would imply a limita-tion on the production of variations, leading tothe concept of a ‘‘real’’ developmental constraint.In contrast, if a significant level of variationamong individuals is identified, this may implyinternal selection that eliminates embryos withvariant characters later in ontogeny because ofdevelopmental disruptions caused by abnormalHox gene regulation.

ACKNOWLEDGMENTS

We thank Gunter P. Wagner for critical readingof the manuscript. We also appreciate HidekiEndo, Raj Ladher, and three anonymous refereesfor their many helpful comments to improve themanuscript. This work was supported by Grants-in-Aid from the Ministry of Education, Science


and Culture of Japan (Specially Promoted Re-search) to S.K.

LITERATURE CITED

Aimi M. 1994. Numerical variation of vertebrae in Japanesemacaques, Macaca fuscata. Anthropol Sci 102(Suppl):1–10.

Allard MW, Honeycutt RL, Novacek MJ. 1999. Advancesin higher level mammalian relationships. Cladistics 15:213–219.

Alberch P. 1989. The logic of monsters: evidence for internalconstraint in development and evolution. Geobios 12:21–57.

Andrews CW. 1910. A descriptive catalogue of the marinereptiles of the Oxford clay. Part I. London: Trustees of theBritish Museum.

Arthur W. 1997. The origin of animal body plans: a study inevolutionary developmental biology. Cambridge, England:Cambridge University Press.

Arthur W, Farrow M. 1999. The pattern of variation incentipede segment number as an example of developmentalconstraint in evolution. J Theor Biol 200:183–191.

Beintema JJ, Breukelman HJ, Dubois JF, Warmels HW.2003. Phylogeny of ruminants secretory ribonuclease genesequences of pronghorn (Antilocapra americana). MolPhylogenet Evol 26:18–25.

Belting HG, Shashikant CS, Ruddle FH. 1998. Modification ofexpression and cis-regulation of Hoxc8 in the evolution ofdiverged axial morphology. Proc Natl Acad Sci USA 95:2355–2360.

Brown DS. 1981. The English upper Jurassic Pleisauroidea(Reptilia), and a review of the phylogeny and classification ofthe Plesiosauria. Bull Br Mus (Nat Hist) Geol 35:253–347.

Burke AC, Nelson CE, Morgan BA, Tabin C. 1995. Hox genesand the evolution of vertebrate axial morphology. Develop-ment 121:333–346.

Cao Y, Fujiwara M, Nikaido M, Okada N, Hasegawa M. 2000.Interordinal relationships and timescale of eutherian evolu-tion as inferred from mitochondrial genome data. Gene259:149–158.

Carrasco AE, Lopez SL. 1994. The puzzle of Hox genes. Int JDev Biol 38:558–564.

Carroll RL, Gaskill P. 1985. The nothosaur Pachypleuro-saurus and the origin of plesiosaurs. Philos Trans R SocLond B Biol Sci 309:343–393.

Cohn MJ, Tickle C. 1999. Developmental basis of limblessnessand axial patterning in snakes. Nature 399:474–479.

Colbert EH, Morales M, Minkoff EC. 2001. Colbert’s evolutionof the vertebrates: a history of the backboned animalsthrough time. New York: John Wiley & Sons, Inc.

Corbet GB, Hill JE. 1986. A world list of mammalian species,2nd edition. New York: Cornell University Press.

Cordes R, Schuster-Gossler K, Serth K, Gossler A. 2004.Specification of vertebral identity is coupled to Notchsignalling and the segmentation clock. Development131:1221–1233.

de Saint-Seine P. 1955. Sauropterigia. In: Piveteau J, editor.Traite de paleontologie. Paris: Masson S.A. p 420–428.

Delsuc F, Scally M, Madsen O, Stanhope MJ, de Jong WW,Catzeflis FM, Springer MS, Douzery EJ. 2002. Molecularphylogeny of living xenarthrans and the impact of characterand taxon sampling on the placental tree rooting. Mol BiolEvol 19:1656–1671.

Galis F. 1999a. On the homology of structures and Hox genes:the vertebral column. Novartis Found Symp 222:80–94.

Galis F. 1999b. Why do almost all mammals have sevencervical vertebrae? Developmental constraints, Hox genes,and cancer. J Exp Zool 285:19–26.

Gaunt SJ. 1994. Conservation in the Hox code duringmorphological evolution. Int J Dev Biol 38:549–552.

Ginsburg L. 1970. Les reptiles fossiles. In: Piveteau J, editor.Traite de paleontologie. Paris: Masson et Cie. p 1161–1332.

Goodrich ES. 1930. Studies on the structure and developmentof vertebrates. London: McMillan.

Gould SJ. 1989. A developmental constraint in Cerion, withcomments in the definition and interpretation of constraintin evolution. Evolution 43:516–539.

Hall BK. 1994. Homology: the hierarchical basis of compara-tive biology. San Diego: Academic Press.

Hallgrimsson B. 2003. Variation. In: Hall BK, Olson WMeditors. Keywords and concepts in evolutionary develop-mental biology. Cambridge, MA: Harvard University Press.p 368–377.

Hildebrand M, Goslow JGE. 2001. Analysis of vertebratestructure. New York: John Wiley & Sons, Inc.

Howell AB. 1930/1970. Aquatic mammals. Reprinted. NewYork: Dover Publications.

Ikeya M, Takada S. 2001. Wnt-3a is required for somitespecification along the anteroposterior axis of the mouseembryo and for regulation of cdx-1 expression. Mech Dev103:27–33.

Janke A, Gemmell NJ, Feldmaier-Fuchs G, von Haeseler A,Paabo S. 1996. The mitochondrial genome of a monotreme-the platypus (Ornithorhynchus anatinus). J Mol Evol 42(2):153–159.

Janke A, Xu X, Arnason U. 1997. The complete mitochondrialgenome of the wallaroo (Macropus robustus) and thephylogenetic relationship among Monotremata, Marsupia-lia, and Eutheria. Proc Natl Acad Sci USA 94:1276–1281.

Ji Q, Luo ZX, Yuan CX, Wible JR, Zhang JP, Georgi JA. 2002.The earliest known eutherian mammal. Nature 416:816–822.

Kardong KV. 2002. Vertebrates: comparative anatomy, func-tion, evolution. New York: McGraw-Hill.

Killian JK, Buckley TR, Stewart N, Munday BL, Jirtle RL.2001. Marsupials and Eutherians reunited: genetic evidencefor the Theria hypothesis of mammalian evolution. MammGenome 12:513–517.

Lin YH, Waddell PJ, Penny D. 2002. Pika and volemitochondrial genomes increase support for both rodentmonophyly and glires. Gene 294:119–129.

Lydekker R. 1913–1916. Catalogue of the ungulate mammalsin British Museum. I–IV. London: British Museum.

Maconochie M, Nonchev S, Morrison A, Krumlauf R. 1996.Paralogous Hox genes: function and regulation. Annu RevGenet 30:529–556.

Matthews LH. 1968. The whale. London: Allen & Unwin.Maynard-Smith J, Burian R, Kauffman S, Alberch P, CampbellJ, Goodwin B, Lande R, Raup D, Wolpert L. 1985. Develop-mental constraints and evolution. Q Rev Biol 60:265–287.

McKitrick MC. 1993. Phylogenetic constraint in evolutionarytheory: has it any explanatory power? Annu Rev Ecol Syst24:307–330.

Murphy WJ, Eizirik E, Johnson WE, Zhang YP, Ryder OA,O’Brien SJ. 2001. Molecular phylogenetics and the origins ofplacental mammals. Nature 409:614–618.

Nikaido M, Rooney AP, Okada N. 1999. Phylogenetic relation-ships among cetartiodactyls based on insertions of short andlong interspersed elements: hippopotamuses are the closest


extant relatives of whales. Proc Natl Acad Sci USA96:10261–10266.

Novacek MJ. 1992. Mammalian phylogeny: shaking the tree.Nature 356:121–125.

Owen R. 1853. Descriptive catalogue of the osteological seriescontained in the museum of the Royal College of Surgeons ofEngland. London: Royal College of Surgeons.

Owen R. 1866. On the anatomy of vertebrates. Vol. II. London:Longmans, Green & Co.

Parra-Olea G, Wake DB. 2001. Extreme morphological andecological homoplasy in tropical salamanders. Proc NatlAcad Sci USA 98:7888–7891.

Peyer B. 1931. Die Triasfauna der Tessiner Kalkalpen. II.Tanystropheus longobardicus Bass sp. Abh Schweiz PalaontGes 50.

Pilbeam D. 2004. The anthropoid postcranial axial skeleton:comments on development, variation, and evolution. J ExpZool 302:241–267.

Prum RO. 1999. Development and evolutionary origin offeathers. J Exp Zool 285:291–306.

Purvis A. 1995. A composite estimate of primate phylogeny.Philos Trans R Soc Lond B Biol Sci 348:405–421.

Raff RA. 1996. The shape of life: genes, development, and theevolution of animal form. Chicago: Chicago University Press.

Remane A. 1936. Wirßealsle und ihre Abkommlinge. In: BolkL, Gopert E, Kallius E, Lubosch W, editors. Handbuch dervergleichenden Anatomie der Wirbeltiere. Berlin: Urban &Schwarzenberg. p. 1–206.

Richardson MK, Chipman AD. 2003. Developmental con-straints in a comparative framework: a test case usingvariations in phalanx number during amniote evolution.J Exp Zool Part B Mol Dev Evol 296:8–22.

Rosenberg E. 1875. Entwicklung der Wirbelsaule desMenschen. Morphol Jahrb 1.

Romer AS, Parsons TS. 1977. The vertebrate body, 5thedition. Philadelphia: Saunders.

Rutledge JJ, Eisen EJ, Legates JE. 1974. Correlated responsein skeletal traits and replicate variation in selected lines ofmice. Theor Appl Genet 45:26–31.

Sawin PB. 1937. Preliminary studies of hereditary variation inthe axial skeleton of the rabbit. Anat Rec 69:407–428.

Schwenk K. 1995. A utilitarian approach to evolutionaryconstraint. Zoology 98:251–262.

Schwenk K, Wagner GP. 2003. Constraint. In: Hall BK, OlsonWM, editors. Keywords and concepts in evolutionarydevelopmental biology. Cambridge, MA: Harvard UniversityPress. p 52–61.

Shoshani J, Groves CP, Simons EL, Gunnell GF. 1996.Primate phylogeny: morphological vs. molecular results.Mol Phylogenet Evol 5:102–154.

Spurway H. 1949. Remarks on Vavilov’s law of homologousvariation. Ricerca Scientifica Suppl:3–9.

Stanhope MJ, Waddell VG, Madsen O, de Jong W, Hedges SB,Cleven GC, Kao D, Springer MS. 1998. Molecular evidencefor multiple origins of Insectivora and for a new order ofendemic insectivore mammals. Proc Natl Acad Sci USA 95:9967–9972.

Wagner GP, Muller GB. 2002. Evolutionary innovationsovercome ancestral constraints: a reexamination of char-acter evolution in male sepsid flies (Diptera: Sepsidae). EvolDev 4:1–6.

Wellik DM, Capecchi MR. 2003. Hox10 and Hox11 genes arerequired to globally pattern the mammalian skeleton.Science 301:363–367.

Wild R. 1973. Die Triasfauna der Tessiner Lalkalpen. 23.Tanystropheus longobardicus (Bassani) (Neue Ergebnisse).Schweiz Palaontol Abh 95:1–162.

Zakany J, Kmita M, Alarcon P, de la Pompa JL, Duboule D.2001. Localized and transient transcription of Hox genessuggests a link between patterning and the segmentationclock. Cell 106:207–217.


evolution of the vertebral formulae in mammals: a...

Documents