Introduction

Advantages of using the mouse

For almost a century, mice have been used as mod-els of human physiology, disease and toxicity. Infact, mice are used more extensively than any othermammalian species in research and testing, for sev-eral reasons. Firstly, mice are relatively easy tobreed and maintain economically in captivity. Sec-ondly, they have short life-spans, so their long-term, post-experimental care is normally not anissue. Thirdly, mice are amenable to genetic analy-sis and manipulation. Lastly, they share manyphysiological features, body systems and develop-mental and cellular processes with humans. Thereis, as a result, an enormous amount of informationavailable on the biology, physiology and genetics ofthe laboratory mouse.

The completed draft sequence of the mousegenome was published in 2002 (1). A comparisonof the mouse and human genomes (2) suggeststhat they are of comparable size (around 3 billionbase pairs), encode similar numbers of genes(around 24,000), and are 99% identical, at least interms of the presence of homologous genes (equiv-alent genes). The two genomes are also 95% syn-tenic, with the same genes clustered on equivalent

chromosomes in both genomes. These observa-tions have added further impetus to efforts to gen-erate and characterise strains of geneticallyaltered (GA) mice, in order to develop models thatimprove our understanding of human physiologyand disease.

Methods of creating GA mice

GA mice are created by using targeted mutagenesis(as used in reverse genetic research), which givesrise to genetically modified (GM) mice, or by expo-sure to mutagens (as used in forward geneticsapproaches), referred to as GA mice. There are cur-rently more than 7000 publicly-available mousestrains (3), and the ease with which mice are genet-ically manipulated is reflected in the fact that theyare the only mammalian species which has beengenetically modified by homologous recombination(whereby a mouse gene is substituted with a humangene). Since the rationale for generating suchstrains is the similarity between the mouse and thehuman genomes, it is assumed that mouse andhuman genes that are homologues and occupy sim-ilar positions in the genome, have similar functionalroles. Thus, it is commonly assumed that when thefunction of a gene is altered, any changes in pheno-type or physiology that result are indicative of the

ATLA 34, 429–454, 2006 429

Comment

The Relevance of Genetically Altered Mouse Models ofHuman Disease

Nirmala Bhogal and Robert Combes

FRAME, Nottingham, UK

Summary — The impetus to develop useful models of human disease and toxicity has resulted in a num-ber of large-scale mouse mutagenesis programmes. This, in turn, has stimulated considerable concernregarding the scientific validity and welfare of genetically altered mice, and the large numbers of mice thatare required by such programmes. In this paper, the scientific advantages and limitations of geneticallyaltered mice as models of several human diseases are discussed. We conclude that, while the use of somesuch mouse models has contributed considerably to an understanding of human disease and toxicity, othergenetically altered mouse models have limited scientific relevance, and fewer have positively contributedto the development of novel human medicines. Suggestions for improving this unsatisfactory situation aremade.

Key words: disease, genetically altered, mouse, mutant, Three Rs, toxicity.

Address for correspondence: N. Bhogal, FRAME, Russell and Burch House, 96–98 North SherwoodStreet, Nottingham NG1 4EE, UK.E-mail: nirmala@frame.org.uk

roles of that specific gene in the mouse and hence,by extrapolation, in humans.

International cooperation

Several major international consortia are involvedin generating GA mice, with the collective aim ofisolating and characterising at least one mouse GAline for every gene in the mouse genome. These con-sortia include large-scale mouse mutagenesis pro-grammes at RIKEN, Japan (4), the Baylor Collegeof Medicine (5) and Jackson disorders mutagenesis(6) projects, and the National Institutes of Health(NIH) programmes (7) in the USA, and the Euro-pean ENU mutagenesis programme (Eumorphia; 8.See also Table 1).

According to the Home Office statistics (9),between 1990–2005, there has been a progressiverise in the proportion of animal procedures thatinvolve GM animals within the UK, and, by 2005,one third of all procedures conducted under theAnimals (Scientific Procedures) Act 1986 involvedthe use of GM animals, mainly mice. If this trendpersists, it is likely that, within the next few years,the use of GM mice will increase the levels of ani-mal experimentation in the UK to beyond the pre-1986 levels.

These statistics already understate the extent towhich GA mice are used, because of a difference inthe way that GM animals are defined legally andby the UK Home Office. The Royal Society hassuggested the following definition: “an organismwhose genetic material has been altered in a waythat does not occur naturally or as a result of mat-ing and/or natural recombination of its gene” (10).This definition includes transgenic and knock-outmice, as well as mutant mice resulting from ran-dom mutagenesis caused by exposure of parentmice to radiation or mutagens, or mice created byinbreeding. However, for statistical purposes, thelatter two categories of GA mice are not classifiedas GM mice, unless the genetic defect has beendefined. For the purposes of this report, in its con-sideration of all GA mice, the latter two categoriesare classified as non-GM GA mice (GA mice), whilethe categories of mice that fall within the HomeOffice definition of GM mice are referred to as GMmice.

Limitations and problems

The generation of a single GA mouse strain canrequire the use of very large numbers of animals,because: a) the mutagenesis methods are ineffi-cient; b) there is poor germline transmission of genemutations; and c) genetic modification canadversely affect fecundity and survival. These prob-lems are so acute that, in many studies, a large pro-

portion of the animals are incidental to the work,since they are merely used for breeding or are off-spring that fail to carry the desired mutation or lacka novel and detectable phenotype.

Also, even if the desired GA mice are created,they are not always good models of human disease,for several fundamental reasons. These include thefact that humans are about 3000 times larger thanmice; they possess a much greater number of cells,some of which have to communicate over long dis-tances; and the lifespans of mice and humans arevastly different. Moreover, there is a lower likeli-hood that errors within the mouse genome willeventually result in chronic diseases such as cancer,due to the greatly reduced number of cell divisionsthat occur in the body of this species over its life-time, as compared with humans (see Mouse modelsof cancer, DNA repair disorders and their use in tox-icity testing).

We have stated earlier that the mouse andhuman genomes have similar gene clusters andmice possess gene homologues to many humangenes. However, despite this, and despite the factthat the average similarity between mouse andhuman genes is 85%, with similarity ranging from70–95% for specific genes, a single nucleotide differ-ence can dramatically alter the function of the pro-tein expressed. Such subtle changes can haveimportant phenotypic consequences for a species,affecting physical characteristics, as well as bio-chemistry, physiology and pharmacology, leadingvia variations in temporal and spatial proteinexpression to species-specific metabolism, immuneresponses, sensory perception and endocrine func-tions. Therefore, assumptions about the functionalequivalence of homologous genes in mice andhumans can be erroneous.

Objectives of the review

The aim of this paper is to retrospectively assess therelevance of a number of GA mouse models ofhuman disease and their treatment, making refer-ence to specific animal welfare problems associatedwith the ways these models are created and used.The intention is not to extensively review the wel-fare implications of mouse mutagenesis-basedresearch, but to discuss specific welfare-relatedissues.

One or more mouse models of haematological,hormonal and metabolic or cardiovascular func-tion, allergy and infection, sensory or central andperipheral nervous systems, behaviour and cogni-tion, cancer, and bone, skeletal or muscle defectsare considered, alongside their value in the devel-opment and safety testing of new therapeutics.These form some of the main areas of interest ofthe major mouse mutagenesis programmes (Table1).

430 Comment

For consistency, human genes are given in upper-case and italic and mouse homologues in lower caseand italics.

Generating GA Mice by Forward andReverse Genetics Approaches

ENU mutagenesis

GA mice can be generated by forward genetics, aphenotype-driven approach first described in the1970s. It involves correlating changes in phenotypewith the underlying genetic alterations. Thismethod can involve the exposure of male or femalemice or, preferably, isolated embryonic stem (ES)cells, to radiation or chemical mutagens, in order toincrease the frequency of spontaneous mutation. Acommonly-used approach involves injecting N-ethyl-N-nitrosourea (ENU) into the abdominal cav-ities of approximately 12-week-old male mice. TheENU causes single base pair mutations by alkylat-ing nucleic acids that may eventually lead to theincorporation of the incorrect base into DNA fol-lowing replication. However, like many chemicalmutagens, ENU can also induce temporary sterilityor fatal tumorigenesis, since mice lack many of theantineoplastic mechanisms present in humans(11). Assuming that the animals survive the ENUtreatment, sterility is generally reversed by thetime the mouse is about 17 weeks old. The malecan then be mated with untreated females. Themice in the resulting generation are phenotypicallycharacterised and those displaying novel or usefultraits are selectively bred. The frequency of muta-tion at any given locus is only about 1/500–1/1500 ofthe G1 offspring (12). However, less than 2% of theresultant offspring display desired phenotypes, dueto the high incidence of recessive traits, poorgermline transmission and poor progeny survival.Thus, when using forward genetic studies, verylarge numbers of animals are used and produced,in order to obtain a few useful GA animals

The main advantage of this technique is that itis not necessary to make any prior assumptionsabout the potential functions of genes, since allgenes are potentially susceptible to random modi-fication by a mutagen. Furthermore, in theory,many GA mice can be generated by a single muta-genic treatment, including animals with muta-tions that result in congenital, biochemical,immunological and other complex traits. Never-theless, identifying the genetic basis of the pheno-type can prove difficult, especially when gametes,rather than ES cell-based methods, are used, andparticularly where there is little or no informationabout the biochemical changes that accompany anovel phenotype. Furthermore, the majority ofENU-based mouse mutagenesis studies are geared

toward identifying dominant traits, with few stud-ies being undertaken to obtain recessive mutants,since it can be difficult to identify recessive carri-ers for subsequent breeding. It is also commonpractice to selectively screen for particular pheno-types, so large numbers of animals can be wasted,because a useful trait is not detected. Conse-quently, it is likely many potentially relevantmouse models have not been identified and iso-lated, since their phenotypes would only be evi-dent during specific life stages. However, ofparticular concern is the fact that mutated ani-mals may display only very subtle changes in phe-notypes that nevertheless are able to adverselyaffect their welfare in unpredictable or undetectedways.

Knock-out mutagenesis

The second main way to generate GA mice is byreverse genetics, sometimes known as the geno-type-driven approach. This involves targetedmanipulation of the genome, to generate GM micewhich display a desired phenotype or carry a spe-cific genetic alteration. Gene knock-out can involvealtering a gene’s function by inserting a segment ofnon-coding DNA or marker protein DNA, or byusing an antisense oligonucleotide or RNA interfer-ence to disrupt gene expression. In each case, theexpression of the targeted protein is disrupted anda loss-of-function mouse strain is generated. TheNIH Knock-out Mouse Project (KOMP) is aimed atproducing a knock-out mouse for each gene withinthe mouse genome (13). This equates to a minimumof approximately 24,000 knock-out mouse strains.However, since each affected protein could havemultiple physiological roles or could play a key rolein development, knock-out and other loss-of-func-tion mutagenesis commonly results in lethal fetalabnormalities, without necessarily allowing anyclues to be discovered as to the function of the tar-geted gene in question.

Transgenesis

Knock-in mutagenesis involves the introduction ofone or more foreign genes — or transgenes — intothe mouse genome, to permit the study of the roles ofnormal or mutant forms of mouse or human genes,or genes ordinarily absent from the mouse genome.Virus-based transgenesis, the earliest developed ofthe three main transgenic methodologies, involvesthe use of viruses to deliver the required transgeneinto sperm and eggs. However, it is difficult to con-trol the number of copies and the precise location ofinsertion of the transgene when using forwardgenetic studies. This is particularly problematical,since it is now known that many non-coding

Comment 431

432 CommentTa

ble

1:

A s

elec

tio

n o

f m

ou

se m

uta

gen

esis

pro

gra

mm

es a

nd

co

nso

rtia

Res

earc

h p

rogr

amm

eT

ype

of m

uta

gen

esis

Dis

ease

or

gen

etic

com

pon

ents

Web

site

Exa

mpl

es o

f pr

ogra

mm

es a

imed

at

prod

uci

ng

GA

mic

e

Bay

lor

Col

lege

of

Med

icin

eE

NU

Dev

elop

men

tal d

efec

ts; e

ndo

crin

e de

fect

s;h

ttp:

//ww

w.m

ouse

-gen

ome.

bcm

.tm

c.n

euro

logi

cal a

nom

alie

s; b

lood

def

ects

edu/

EN

U/M

utag

enes

isP

roj.a

sp

Bay

lor

Col

lege

of

Med

icin

eG

enet

ic m

anip

ulat

ion

Ch

rom

osom

e 11

en

gin

eeri

ng

htt

p://w

ww

.mou

se-g

enom

e.bc

m.t

mc.

edu/

Ch

rEn

g.as

p

Bay

lor

Col

lege

of

Med

icin

eG

ene

trap

pin

gR

epro

duct

ive

mut

ants

htt

p://w

ww

.mou

se-g

enom

e.bc

m.t

mc.

edu/

bart

mic

e/h

ome.

asp

Cen

tre

for

Mod

ellin

g H

uman

EN

U/g

enet

rap

Gen

etic

con

trol

of

bon

e m

iner

al d

ensi

tyh

ttp:

//ww

w.c

mh

d.ca

/gen

etra

p/in

dex.

htm

lD

isea

se, T

oron

to

Cen

tre

for

Mod

ellin

g H

uman

E

NU

Neu

rolo

gica

l an

d be

hav

iour

al r

esea

rch

; h

ttp:

//ww

w.c

mh

d.ca

/en

u_m

utag

enei

s/in

dex.

htm

lD

isea

se, T

oron

toh

aem

atop

oesi

s re

sear

ch; c

ardi

ovas

cula

r re

sear

ch

Har

wel

l Mut

agen

esis

Pro

gram

me

EN

U/t

arge

ted

mut

agen

esis

Gen

ome-

wid

eh

ttp:

//ww

w.m

gu.h

ar.m

rc.a

c.uk

/

Jack

son

Lab

orat

ory

Mut

agen

esis

/bre

edin

gA

wid

e va

riet

y of

dis

orde

rs in

clud

ing

neu

rolo

gica

l, h

ttp:

//ww

w.ja

x.or

g/h

eart

, lun

g, b

lood

an

d sl

eep

diso

rder

sh

ttp:

//ww

w.ja

x.or

g/n

mf/

McL

augh

lin R

esea

rch

In

stit

ute

EN

UG

enom

e-w

ide;

rec

essi

ve s

cree

nh

ttp:

//ww

w.m

onta

na.

edu/

mri

/en

ump.

htm

l

Nat

ion

al I

nst

itut

es o

f H

ealt

hK

noc

k-ou

t m

utag

enes

isG

enom

e-w

ide

gen

e fu

nct

ion

stu

dyh

ttp:

//ww

w.n

ih.g

ov/s

cien

ce/m

odel

s/In

itia

tive

mou

se/k

noc

kout

/kom

p.h

tml

Oak

Rid

ge N

atio

nal

Lab

orat

ory

EN

UR

egio

nal

/ch

rom

osom

e sp

ecif

ic s

cree

nh

ttp:

//lsd

.orn

l.gov

/mgd

/mut

agen

esis

/M

amm

alia

n g

enet

ics

& g

enom

ics

RIK

EN

EN

UG

enom

e-w

ide;

incl

udes

scr

een

for

rec

essi

ve

htt

p://w

ww

.gsc

.rik

en.jp

/Mou

se/m

ain

.htm

mut

ants

an

d bo

th e

arly

an

d la

te-o

nse

t di

seas

es

Comment 433

Tab

le 1

: co

nti

nu

ed

Res

earc

h p

rogr

amm

eT

ype

of m

uta

gen

esis

Dis

ease

or

gen

etic

com

pon

ents

Web

site

Pro

duce

rs o

f ge

ne-

trap

ES

cel

ls t

hat

may

be

use

d to

cre

ate

mu

tan

t m

ice

Bay

gen

omic

sK

noc

k-ou

tC

ardi

ovas

cula

r an

d pu

lmon

ary

dise

ases

htt

p://b

ayge

nom

ics.

ucsf

.edu

/ove

rvie

w/

wel

com

e.h

tml

Eur

opea

n C

ondi

tion

al M

ouse

Con

diti

onal

mut

agen

esis

;G

enom

e-w

ide

No

curr

ent

web

add

ress

— n

ew in

itia

tive

Mut

agen

esis

Pro

gram

me

knoc

k-ou

t; t

ran

sgen

esis

Ger

man

Gen

etra

p C

onso

rtiu

mT

ran

sgen

esis

Gen

ome-

wid

eh

ttp:

//gen

etra

p.gs

f.de

/

Inte

rnat

iona

l Gen

e T

rap

Con

sort

ium

Tra

nsg

enes

isG

enom

e-w

ide

htt

p://w

ww

gen

etra

p.or

g/

Om

nib

ank

Gen

e tr

ap k

noc

k-ou

tG

enom

e-w

ide

for

drug

dis

cove

ryh

ttp:

//ww

w.le

xico

n-g

enet

ics.

com

/dis

cove

ry/

omn

iban

k.h

tm

San

ger

Inst

itut

eK

noc

k-ou

tG

enom

e-w

ide

htt

p://w

ww

.san

ger.

ac.u

k/P

ostG

enom

ics/

gen

etra

p/

sequences regulate gene expression, so introducing aforeign gene can disrupt normal patterns of expres-sion for other genes. Therefore, it is not surprisingthat many of the “mosaic” offspring that result fromtransgene insertions are non-viable. This techniquehas been largely supplanted by a method in whichthe transgene is incorporated into the pronuclei ofembryos or ES cells. When transgenes are insertedinto embryos in this way, the efficiency and survivalof embryos to birth is increased to around 4–6%.However, since pronuclear injection does not over-come the problems of multiple and random trans-gene insertion, targeted gene insertion into ES cellshas become the method of choice. By using thismethod, it is possible to flank the transgene withDNA sequences that correspond with those of thedesired insertion site, and then to select engineeredstem cells for propagation and injection into blasto-cysts destined for uterine implantation. A similartechnique has been developed for engineering spermstem cells, followed by their reintroduction intotestes lacking germ cells. A detailed review of themethodology can be found elsewhere (14).

The main concern with transgenic research is that,even if the transgene is expressed, the result mightnot be sufficiently informative about the role of thetargeted gene, since its expression may not be regu-lated in the same way as it is in humans. Further-more, although tissue-specific or temporalexpression can be achieved by placing transgeneexpression under the control of tissue-specific orinducible promoters, respectively, there is no guar-antee that the protein following expression of thegene is folded, targeted or regulated in the samemanner in mice and humans. This can give rise tounexpected and significant animal welfare problems.

The Relevance of GA Mouse Models:Case Studies

GA mice as partial replacement models

In a small number of cases, GA mice might eventu-ally supplant the use of other animals in researchand testing. One example is the use of transgenicTgPVR 21 mice expressing the human polio virusreceptor for the neurovirulence testing of the oralpolio vaccine (OPV), which could obviate the use ofprimates in such tests. The use of this model is sup-ported by the World Health Organisation (15), sincenot only do these mice display histological andphysical signs of motor neuron degeneration associ-ated with human forms of the disease that can bemonitored by paralysis scoring, but also becausetests of OPV lots can be conducted in 2 weeksrather than in 1.5–2 months, the time taken to con-duct primate-based neurovirulence testing. Wherethere is a need for the routine animal testing of vac-

cines for the presence of infectious viral particles, itis perhaps more ethical to use transgenic mice thanprimates, since this also avoids serious logisticalissues, such as containment of infected animals orthe need for the long-term care or re-homing of lab-oratory animals with longer life expectancies.

Are GA mice relevant and useful?

GA mouse models of human disease often lack rele-vance in the case of complex multigenic disorders.Indeed, some studies in GA mice have been lessinformative than the corresponding investigationswith less-complex organisms and cell culture sys-tems. This is particularly true for mouse modelsdeveloped by using forward genetics, where an unde-fined number of mutations may have contributed toan overall phenotype which resembles a human dis-order, but which may share few, if any, of the under-lying biochemical or genetic causes of the respectivehuman disorders. The relevance of many transgenicmouse models can be questioned on the basis that,even if a species gene homologue has the same func-tion and expression patterns and levels in humansand mice, all the remaining components of the bio-chemical pathway must be equally represented in thesurrogate animal, if relevant mouse models of humandiseases are to be created within a laboratory setting.

It is not feasible to consider the scientific rele-vance of every GA mouse model. For this reason,the limitations, including those described above, ofusing GA mice as models of human diseases will beillustrated by reference to models of specific humandiseases and to groups of diseases that share funda-mental features.

Mouse models of haematological diseases:sickle cell anaemia

Sickle cell anaemia (SCA) is one of several types ofinherited blood disorder that can affect the shapeand haemoglobin content of red blood cells, theirhalf-life in the blood, and their ability to passthrough small vessels. In unaffected individuals,95–98% of normal haemoglobin content is haemo-globin A, which is composed of two α and two βchains; a small proportion of haemoglobin is com-posed of two α and two δ chains (haemoglobin A2),or of two α and two γ chains (fetal haemoglobin). InSCA, the red blood cells contain mostly haemoglo-bin S, because of mutations in the β chain protein-encoding gene (βs mutations).

Transgenic mouse models of SCA were first createdby replacing the mouse α and βs haemoglobin genes(hbA and hbB, respectively) with the equivalenthuman genes (HbA and HbB). The resultant miceexpress haemoglobin chain proteins that comprisethe major form of haemoglobin, haemoglobin A (16).

434 Comment

These mice demonstrate limited red blood cell sick-ling. Further genetic engineering of the βs gene hastherefore been performed, in order to create SADmice that contain three point mutations, namely theβs Antilles β23Ile and D Punjab β121Gln mutationsin the HbB gene (17). These mice exhibit enhancedsickling, a feature that has been used to examine theeffects of potential anti-sickling agents. However,these mice fail to display the haemolytic anaemiacharacteristic of the disease in humans.

Other mouse models of human SCA have beencreated by deactivating the mouse hbA and hbBgenes by using homologous recombination. Thesemodels include Berkeley mice, which express thefull complement of human SCA globin α and βgenes, including the mutant βs genes (genotype:Tg[Hu-miniLCRα1gγAγδβS]Hbao//HbaoHbbo//Hbbo).Yet Berkeley mice still display only some, but notall, features of the human disease (18). The differ-ences between the Berkeley mouse model and thehuman disease have been reviewed (19). Ratherthan an increase in red blood cell haemoglobin lev-els, as seen in human SCA, Berkeley mice tend tohave reduced levels of normal haemoglobin contentin their red blood cells. This is presumably becausemice express very low levels of fetal haemoglobin, aprotein that is able to protect against reduction inhaemoglobin levels in human individuals with SCA.Indeed, the variation in the levels of fetal haemo-globin is directly related to the severity of SCA seenin these individuals — those with relatively highfetal haemoglobin levels exhibit milder forms of thedisease. Such variation is not seen between individ-ual Berkeley mice. Furthermore, the red blood cellvolumes are much smaller in mice than in humans.Presumably, these features of the Berkeley mousetogether contribute to the increased severity of dis-ease symptoms in the mouse model in comparisonwith those that develop in humans. Furthermore,differences in the main site of haematopoiesis (thespleen in the Berkeley mouse and the bone-marrowin SCA patients) might contribute to the apparentsplenic hypertrophy in 1–6 month old mice,whereas in humans, initial hypertrophy is followedby atrophy of the spleen and almost total loss ofspleen function in adult patients. Nevertheless,since many of the same organs are affected by SCAin the mouse model as are involved in the humandisease, the Berkeley mouse is likely to be useful inthe study of some features of the human disease.

Mouse models of hormonal and metabolicfunction: diabetes

Type 1 diabetes

Human diabetes is characterised by an inability toregulate blood glucose levels. However, each of the

various types of diabetes has specific aetiologies andtreatment regimes. Human type 1, insulin-depend-ent diabetes, for instance, has been attributed to anautoimmune condition. It affects children andyoung adults, and occurs when the body makes lit-tle or no insulin.

Mouse models of insulin-dependent type 1 dia-betes have been created by gene knock-out and genemanipulation, principally to increase understand-ing of the role of the immune system in the aetiol-ogy of the disease (20). The non-obese diabetic(NOD) mouse, however, was created by the selec-tive inbreeding of a specific mouse strain. Thesemice spontaneously develop insulin-dependent dia-betes, but, unlike the analogous situation inhumans, female NOD mice develop diabetes at ahigher incidence than males (21). Currently, thereis little understanding of the role of other pathwaysin the pathogenesis of diabetes. Yet, despite its lim-itations, the NOD mouse has been used to examinethe complex and multifactorial changes in theexpression and activity of a number of proteins andgenes, and also provides some useful insights intothe human disease. This is because, despite the dif-ferences between diabetes in the NOD mouse andthe human disease, there appears to be considerablesimilarity between the roles of the components ofthe major histocompatibility complex II andcytokines in the mouse model and human condi-tions (22).

Type 2 diabetes

Maturity onset diabetes of the young (MODY) andadult type 2 diabetes are no less difficult to study,since there appear to be several mutations in dif-ferent genes that can contribute to the develop-ment of these disorders. Several GM mouse modelshave been generated for studies on individual com-ponents of type 2 diabetes. Knock-out mice havenot been useful for deciphering the mechanism oftype 2 diabetes. For instance, despite the supposedroles of the equivalent proteins in human diabetes,loss-of-function mutations in the insulin receptorgene (resulting in poor insulin sensitivity), com-bined with the absence of a KATP channel (whichregulates insulin secretion), were insufficient totrigger diabetes in mice carrying both geneticdefects (23). Furthermore, it seems that, whileloss-of-function mutations in the human KATPchannel protein, SUR1 — the receptor for sulpho-nylureas used to close the KATP channels in type 2diabetes, and thus to reduce insulin secretion —cause hyperinsulinemic hypoglycaemia, Sur1 geneknock-out mice, despite dysfunctional KATP chan-nels, are able to regulate insulin secretion. Thisappears to be due to the presence of a secondinsulin regulatory mechanism that is found inmice, but not in humans (24). This suggests that

Comment 435

the usefulness of GA mice in the development oftherapeutics that target KATP channel activity islikely to be severely limited.

The phenomenon of ectopic expression, theexpression of a protein in a tissue where it is notnormally expressed, has been used to create othermodels of type 2 diabetes. In the case of ectopicexpression of dominant mutants of agouti protein(a natural ligand for melanocortin receptors thathelps to regulate appetite) in numerous tissues(25), GM mouse models have not helped in deter-mining the mechanisms of insulin resistant dia-betes and obesity. Instead, studies in these micehave simply confirmed that agouti protein has anumber of roles within mice, many of which mayonly superficially resemble the roles of the proteinin humans. One such agouti mutant strain thatexpresses the agouti transgene hemizygotically,Tg/–, starts to develop diabetes-like symptomswhen the mice are only 4–5 weeks old, in contrastto the more delayed onset in human patients. Fur-thermore, hyperglycaemia in mice appears tooccur only in males, which is not the case inhumans. This mouse model continues to produceinsulin at higher levels than normal mice, suchthat the model is only potentially relevant forstudying one type of insulin resistant diabetes.

Mouse models of infection and susceptibilityto infection

Cystic fibrosis

Perhaps the best studied human disease is cysticfibrosis (CF). In the 1980s, CF was linked to muta-tions within the gene encoding a chloride channel,which is now referred to as the CF transmembraneconductance regulator (CFTR; 26). The CFTR nor-mally facilitates the movement of chloride ionsinto and out of cells within the lungs and otherorgans. CFTR gene mutations can impair CFTRexpression, structure or function, resulting in theaccumulation of thick, sticky mucus and othersecretions that increase the individual’s suscepti-bility to infections, impair cellular secretion and/ortransport, and contribute to early death. Some70% of CF sufferers carry a deletion of threenucleotide bases corresponding to the loss of a sin-gle phenylalanine residue in the encoded protein.However, about 1300 further frameshift and site-specific mutations have been detected within thehuman CFTR gene or the upstream promoterregions (27).

CF is an autosomal and recessive disease, so twocopies of a defective CFTR gene are needed beforethe disease manifests itself. This represents thefirst limitation of the mouse models, since not allCFTR gene mutations will have the same pheno-

typic consequences, and the likelihood that a personwould inherit identical mutant alleles from bothparents is low.

The most commonly-used CF mouse model wasgenerated from ES cells, in which the mouse genethought to be homologous to the human CFTRgene, cftr, was inactivated by gene targeting. Theresultant cftr (–/–) homozygous mice exhibited someof the features of human CF (28). However, thereare important differences between this and othermouse models of the disease and CF in humans.Some of these differences stem from the fact thatthe so-called mouse CFTR is now thought not to bea species homologue for the human protein (29).The most important difference is the fact that cftr(–/–) mice do not develop excessive amounts ofmucus in the lungs, a very common complication inhuman CF. This is because mice have fewer mucus-secreting cells in their lungs than are present inhumans. In fact, whereas lung infections are themajor cause of death in CF humans, cftr (–/–) micetend to die mainly as a result of gastrointestinalobstruction, which is not a feature of the humandisease (30). Furthermore, while pancreatic disor-ders that arise due to an inability to secrete pancre-atic enzymes are seen in around 85% of CF patients,pancreatic dysfunction is rarely seen in cftr (–/–)mice. Moreover, when it does occur, its onset isdelayed in mice. A high proportion of CF patientsalso exhibit loss of fertility, and some develop livercirrhosis. Again, neither of these symptoms is seenin cftr (–/–) mice. In fact, only by considering thecontribution of other proteins to fluid compositionwithin cells of the respiratory tract to the accumu-lation of thickened mucus in the lungs, were someof these important differences between the humandisease and mouse models of the disease resolved(31). This suggests that CF is unlikely to be a mono-genic disorder, but instead, is dependent on theexpression and function of other proteins, such asprotease-activated receptor-2, found in alimentaryand respiratory membranes, where it regulatesmucus secretion (32), and on sodium ion transport.Indeed, by overexpressing an epithelial sodium ionchannel under the control of an airway-specific pro-moter, a more useful model of airway mucus hyper-secretion has recently been created. The resultantmice show increased sodium ion absorption andhigher mucous concentrations in their lungs,resulting in many of the features of CF lung diseasein humans (31).

Thus, to date, GM mice have largely proven use-ful only in terms of identifying the differencesbetween the human disease and mouse models ofthe disease. Such studies have therefore revealedsome of the underlying mechanisms of the humandisease. However, the relevance of CF mouse mod-els to the development of therapies is somewhatlimited, due to the lack of correspondence betweenCF in the mouse models and human CF.

436 Comment

Vaccine development and testing

GM mice have been developed to study infectionscaused by pathogens such as the human poliovirus(33), to develop vaccines, and to bioassay pathogens,such as BSE (bovine spongiform encephalopathy;34). Challenge with a pathogen itself can have severeimplications for the welfare of mice used in thesestudies. However, in some instances, the use of GMmice represents a refinement in the context of theThree Rs, particularly for HIV/AIDS research, wheremice have partially replaced the use of primates (seeMouse models as partial replacements). The latterhave a much longer life expectancy, and often fail todevelop symptoms of infection, in addition to beingdifficult to naturalise or re-home, once they havebeen exposed to HIV. However, the immune systemsof mice and humans are far from identical (35), andthis limits the relevance of some mouse models, notonly for modelling human autoimmune disorderssuch as type 1 diabetes and some aspects of neurode-generative disorders, but also for modelling patho-genic infections.

Xenografting of immune-compromised mice

In an attempt to overcome some of these limita-tions, severe combined immune deficiency (SCID)mice have been developed by genetic engineering,initially by the chance mating of two recessive car-riers of a mutant form of the SCID gene. Homozy-gous SCID/SCID mice lack T-cells and B-cells.Another SCID mouse, the SCID/beige mouse, addi-tionally lacks natural killer activity. These mousemodels have been used as a recipient for humantumour cell xenografts (see Immune-compromisedmice in cancer research) and to study diseases suchas hepatitis B and hepatitis C (36), Epstein-Barrvirus infection, and HIV infection (37). A recentexample of such an application is the use of SCIDmice that are also homozygous for the urokinaseplasminogen activator (denoted (uPA)-SCID mice)as hosts for primary human hepatocytes. Once thehuman hepatocytes had been transplanted into the(uPA)-SCID mice, the animals were infected withhepatitis B virus or hepatitis C virus, whichafforded a well characterised model of human hepa-titis infection. However, the mice are immune-com-promised, so this model is not able to take accountof any influence of the host immune system on theaetiology of the diseases being studied (38).

A better approach would appear to be the gener-ation of trimera mice. Such animals are created,for instance, by the total body irradiation ofBALB/c mice to destroy the haematopoietic sys-tem, followed by its replacement by engraftingbone-marrow samples from SCID mice. Humanperipheral blood mononuclear cells (PBMCs) arealso introduced into the mice, so they are capable

of mounting primary and secondary cellular andhumoral immune responses specific to a number ofinfections, including hepatitis B, hepatitis C andHIV (see 39 and the references therein). The avail-ability of such models has provided a means ofstudying infectious diseases such as HIV, whichhave largely relied on far from ideal studies in non-human primates.

The fact that SCID mice are immune deficient,raises concerns about their welfare. Furthermore,due to the need for the containment of possiblepathogens such as HIV and the hepatitis viruses,these animals require special conditions, in order toavoid the risk of contamination of co-housed,immune-compromised animals. Infection rates areminimised by using ventilated cages with filteredair, and by sterilising the cages, bedding and nest-ing material, and food and water. Attempts havebeen made to create other useful models that retainimmune-competence, but which can still be usefulto study the same range of diseases. An example ofthis is the gp120 mouse, which expresses the HIVgenome in astrocytes (40). The main drawback ofthis model is that HIV does not replicate in humanastrocytes, so it is difficult to determine whetherthis mouse model is suitable for studying HIV-asso-ciated effects on the human brain. This limitationhas been overcome by engineering HIV, to create achimeric virus which encodes viral coat glycopro-tein, gp80, from murine leukaemia virus, instead ofthe corresponding coat protein, gp120, for HIV. Theresulting engineered virus, EcoHIV, encodes HIVpackaged within a murine viral capsule. This viruswas used to successfully infect several strains ofmice without the prior need to resort to creatingGM mice or using an immune deficient model (41).Many of the major characteristics of HIV infectionin humans were reproduced. Furthermore, sincethe engineered virus is only able to infect rodents,cross-species infection is not an issue.

Mouse models of sensory function

Using GA mice to map the roles of genes involved insensory perception is particularly difficult, sinceevolutionary changes have resulted in species-spe-cific gene repertoires for olfaction, vision and taste.Thus, olfactory receptors, for instance, are encodedby the largest mammalian gene superfamily (com-prising more than 1000 genes). However, while theolfactory gene repertoires of rodents and primatesare similar in size, the proportion of functionalolfactory receptors in rodents and new world mon-keys is higher than in old world monkeys and apes,with humans possessing the lowest number of olfac-tory receptors (42). In fact, although there is syn-tenic clustering of genes that encode thesereceptors in the mouse and human genomes, manyof the human orthologues are pseudogenes, which

Comment 437

never result in the expression of functional pro-teins, but may regulate the expression of othergenes or, as recently suggested, may have con-tributed to the acquisition of trichromatic visionspecifically by humans and other apes (43). Themere fact that the functional gene repertoire maybe expanded in mice can mean that, as in the caseof olfaction, the ability of one gene or several genesto functionally substitute a gene that has been inac-tivated, can severely limit the use of functionalgenetic studies in mice to assign roles to humangenes.

Deafness

A number of GM mice have been used to study thesensorineural and conductive forms of human deaf-ness. The former type of deafness is associated withdefects in the inner ear (mainly the cochlea) thatimpair the auditory processing of a signal, while thelatter is related to structural defects in the externalor middle ear that result in impaired sound conduc-tance down the auditory canal. The atp6b1 nullmutant mouse, which cannot express the B1 sub-unit of H+-ATPase, was developed as a model ofhuman sensorineural deafness. However, this GMmouse displays distal renal tubule acidosis (44) butnormal hearing (45), even though the same geneticdefect in humans has been related to autosomalrecessive sensorineural hearing loss as well as distalrenal tubule acidosis (46). As a result, this mousemodel has been more useful as a model of distalrenal tubule acidosis than of human deafness.

The shaker1 mouse (47), on the other hand, dis-plays profound hearing loss, and was one of the firstGA mouse models of human genetic deafness. Oncethe mouse gene underlying the natural defect in theshaker1 mouse had been identified, the location ofthe corresponding gene in the human genome couldbe determined. It was found that the shaker1 locuswas encoded in mice by the myosin VIIA gene,Myo7a. As the name suggests, shaker1 mice showhyperactivity, head-tossing and circling activity, inaddition to hearing loss. It was subsequentlydemonstrated that mutations in the Myo7a lead tohearing loss in humans, symptomatic of non-syn-dromic forms of deafness and Usher’s syndrome.

Half of all human individuals who are both deafand blind suffer from Usher’s syndrome. In humans,some mutations in the myosin VIIA gene (MYO7A)can also lead to a specific form of Usher’s syndrome(type 1b syndrome), in which both hearing loss andthe retinal degenerative disorder, retinitis pigmen-tosa, occur at around seven or eight years of age. Yetnone of the Myo7a mutations in mice cause blindnessin mice, even in very old mice, despite the factmyosin VIIA is normally expressed in the retinal pig-ment epithelium and photoreceptor cells of bothhumans and mice. This may be a reflection of the

short life-span of the mouse, which prevents theretina from receiving sufficient exposure to light toelicit pathological changes (48). However, since highintensity light appears to reduce the size of a and bwaves in young shaker1 homozygotes and, to a lesserextent, in shaker1 heterozygotes (49), the shaker1mouse might act as a suitable model for the earlystages of retinitis pigmentosa, although not the laterstages of the human disease.

Visual defects

There are also a number of specific GM mouse mod-els of human retinopathies. Autosomal dominantretinitis pigmentosa (ADRP), for example, has beenstudied in rhodopsin (the visual pigment) knock-out(rho –/–) mice and transgenic mice that carry equiv-alent rhodopsin mutations to those related tohuman forms of the disorder (50). Extensive studieson both naturally occurring hereditary mutationsin genes that influence visual perception, as well ason artificially produced GM mice, have providedmuch information about the visual process (51).

Nevertheless, several important differences existbetween the organisation of the mouse and humanvisual systems, which derive from the fact that miceare nocturnal animals. The most significant ofthese are differences in visual acuity due to the sub-stantially lower density of cones (for colour vision)and the absence of a fovea (for visual acuity) in themouse retina, compared with the human retina.The mouse retina is dominated by rod cells forimproved night vision. The presence of two, ratherthan three, colour pigments in mice may also havea bearing on the relevance of certain mouse models,since, while the absorption maxima of human red,blue and green pigments are 564, 429 and 534nm,respectively, the absorption maxima of the twomouse pigments are 360 and 510nm (52). The latterwavelength represents a considerable overlap withthe absorption maximum of rhodopsin found in therod cells of humans and mice, so there is a strongpossibility that mouse colour pigments can partlycompensate for visual defects that result from loss-of-function mutations in rhodopsin in a way that isnot possible in the human visual system. This mayexplain why the rhodopsin mutants expressed insome mouse models do not result in retinopathies ofthe same severity as those seen in humans.

Mouse models of nervous and muscle systems

Muscular dystrophy

Duchenne Muscular Dystrophy (DMD) in humansis an X-linked recessive disorder due to mutationsin the dystrophin gene (DMD). This gene encodes a

438 Comment

protein that ordinarily provides structural supportfor muscles. In humans, DMD is characterised bythe rapid degeneration of muscles early in life, usu-ally with onset before the age of five. Because thedisorder is X-linked, it mainly affects males.

The most obvious problem with mouse models ofDMD, such as dy/dy2j, is that the genetic defect isnot X-linked, but is instead a recessive autosomaltrait of the mouse dystrophin gene homologue, dy.The first sign of DMD in dy/dy2j mice, muscleweakness in the hind limbs, is evident within 2weeks of birth (53), with affected individuals gen-erally not surviving beyond 6 months. Thus, theonset of the disease appears to be earlier in themouse model than in human patients, even takinginto consideration the difference between the life-spans of mice (2–3 years) and humans (70–90years). Indeed, histological examination suggeststhat the muscle membrane is more extensivelydamaged in dy/dy2j mice than in human patients,partly because of the lower extent of fat replace-ment in the skeletal muscles of affected animals. Infact, human DMD is thought to be myogenic in ori-gin, whereas in the mouse DMD model, it isbelieved to have a neurogenic origin, arising as aresult of impeded nerve conductance due to thelack of nerve myelination.

A second mouse model of DMD, the mdx mouse,mimics the X-linked recessive nature of humanDMD. This model arose as a spontaneous mutationin the mouse dystrophin gene, but this model alsofails to exhibit many of the features of the humandisease. In fact, although young mdx mice exhibitmuscle necrosis, there is a rapid recovery due to ahigh rate of phagocytic infiltration and rapid mus-cle regeneration, such that the mice are essentiallydisease-free by the age of 5 weeks (54).

Together with differences between the anatomi-cal positioning of muscles in humans and mice (55),these observations have limited the usefulness ofmouse models of DMD in determining the mecha-nistic details of the disease and therefore in drugdevelopment. Nevertheless, using the mdx mice,and also a knock-out mouse that lacks the expres-sion of dystrophin and the related muscle architec-tural protein, utrophin (the dko mouse model),potential problems of using dystrophin gene ther-apy to treat DMD patients have been identified,including difficulties with delivering large DNAmolecules (56). This has led to considerable interestin developing alternative therapies, such as the useof minigenes and antisense oligonucleotides.

Neurodegenerative disorders

Predisposition to neurodegenerative disorders,such as Parkinson’s disease and Alzheimer’s dis-ease (AD), is commonly dependent on mutations inseveral genes. For instance, mutations in genes

encoding β-synuclein, parkin and α-ubiquitinhydrolase have all been linked to Parkinson’s dis-ease, and mutations in genes encoding β-amyloid,presenilin and tau proteins, have all been linked tohuman AD. Furthermore, many of these diseasesare age-related and typically display slow onset,which is suggestive of a complex interplay betweenseveral disease susceptibility genes and non-geneticcomponents, such as infection and age. As illus-trated below, this can complicate the developmentof relevant models of these diseases.

Alzheimer’s disease

AD in humans is an age-related disease associatedwith progressive loss of cognitive function andmemory, and the development of amyloid plaques(APs) and neurofibrillary tangles (NFTs) in thebrain. These mainly comprise β-amyloid proteinand hyperphosphorylated tau protein, respectively.The human disease is also characterised by the lossof neuronal synaptic density.

GM mouse models of AD have, until relativelyrecently, been used to examine the individual rolesof the two predominant forms of β-amyloid protein,the amyloid protein precursor (APP) and proteasesthat act on APP or β-amyloid proteins or, alterna-tively, on the expression and activity of tau orapolipoprotein E (a serum protein that regulatescholesterol levels; 57). The resultant mouse modelsdisplay some, but not all, of the features of thehuman disease. However, the development of the3xTg-AD transgenic mouse model has greatlyimproved the situation. The 3xTg-AD mouse wasgenerated by using the homozygous presenilin 1mutation (PS1M146V) knock-in mouse (PS1-KImice) as a starting strain. This transgenic mousecontains a mutated form of the presenilin 1 gene,which results in the expression of a mutant form ofpresenilin 1 that causes the rapid development ofAPs in transgenic mice that also carry the APP dou-ble mutant K670-N/M671-L (APPswe) gene (58).

The resultant PS1M146V/APPswe mice displayedlimited neuronal depletion and no sign of NFTs. Bycontrast, while APs were not seen in a transgenicmouse model expressing the human P301L tau pro-tein, NFTs were evident in this model (59). Neuron-specific CNS expression was achieved by placing theexpression of P301L tau protein under the control ofThy1.2, a neuron-specific mouse promoter.

The rationale for creating the 3xTg-AD modelwas to get the transgenes for PS1M146V APPsweand P301L tau proteins expressed simultaneously inthe same model, in order to represent all the mainfeatures of the human disease. To achieve this, thegenes encoding human K670-N/M671-L APP (APP-swe) and human P301L tau proteins were insertedinto Thy1.2 cassettes, to drive CNS expression ofmutant preselinin 1 and tau proteins. The 3xTg-AD

Comment 439

mice were then created by injecting these con-structs into single cell embryos from PS1-KI mice.

This approach meant that fewer animals wereused to generate the 3xTg-AD mice than arerequired for conventional step-wise transgenicapproaches, where one transgene is inserted intothe genome at a time and its expression is con-firmed prior to the introduction of subsequenttransgenes. The resultant 3xTg-AD mice exhibitedthe progressive, brain region-specific neuropatho-logical features characteristic of the human disease,with synaptic dysfunction preceding AP formation,which, in turn, preceded NFT formation. Thismouse model is potentially the most promising onefor developing therapies for AD, since, unlike manyAD mouse models, these mice are not behaviourallyabnormal from birth, despite exhibiting subtle dif-ferences in the extent and type of DNA damage thatoccurs in mice and humans.

It has recently been discovered that keepinggenetic mice in a pathogen-free environment maylimit the development of NFTs and APs character-istic of human AD, which suggests a role forpathogen-induced inflammation in AD (60). Thisimplies that attempts to elucidate the function ofgenes or model human diseases could fail, due to anabsence from laboratory environments of environ-mental factors that ordinarily would be present,and not because of the genetic alteration itself.

Welfare of animal models of neurodegenerative diseases

Several animal welfare implications arise specificallyas a result of creating mouse models of neurodegen-erative diseases. These include the distress and suf-fering caused by neurological impairments, includingtremors and ataxia (loss of full control of bodilymovements), confusion and disorientation. The lossof neurological capacity may also influence generalbehaviour, including the ability to interact withother animals. These symptoms often have substan-tial consequences for the fecundity and viability ofthe animals, and their occurrence requires carefulmonitoring throughout their lifetimes. This is moreproblematic when neurodegenerative phenotypesare induced as a result of exposure to mutagens,rather than by genetic modification. This is because,in the former case, all or some of the relevant fea-tures of the altered phenotype arising from anymutation may not be detected. The existing modelsof neurodegenerative diseases are nevertheless ofgreat interest. Hence, we advocate the greater use oftemporal and spatial gene expression coupled withhistopathology, imaging and electrophysiologicaltechniques, which allow the progression of neurode-generative disorders in mouse models to be mappedwith better regard for animal welfare than is possi-ble by using the more traditional, invasive

approaches, and at the same time are more reflectiveof the onset of these diseases in humans.

Development and congenital defects

Cleft palate

Cleft palate is the most common form of congenitalbone dysmorphia in humans. It develops within thefirst trimester of pregnancy. A number of genes andexternal factors may play a role in the aetiology ofthis disease, so its study in GA mice has been diffi-cult. Mouse models of cleft palate have been gener-ated by ENU mutagenesis and by targetedmutagenesis. A large proportion of ENU mousestrains exhibit cleft palate, in addition to other phe-notypes, presumably because of random and unde-fined mutations within the homeobox genes thatregulate the expression of other genes (6, 61). Whilethis has contributed to the widely-held belief thatcleft palate is a multifactorial disorder caused bypoint mutations in one or more genes, it does notcircumvent the problem that the chances of devel-oping a preventative cure for cleft palate are lim-ited. The fundamental question is whether cleftpalate should be modelled at all, in the absence of atangible human therapeutic benefit from this kindof research. This is particularly important, giventhe animal welfare implications. As a result of cleftpalate, GA model offspring may not be able to feedproperly, or survive to adulthood or to a stage atwhich the underlying mechanisms of cleft deforma-tion can be studied. This is exemplified by the caseof the homozygous MSX-1 –/– null mutant mouse.MSX-1 is a homeobox gene that is expressed in anumber of developing organs in vertebrates. MSX-1–/– neonates fail to survive, because of severe cran-iofacial abnormalities, including secondary cleftpalate and abnormal tooth development (62).

Apert syndrome

Recently, a transgenic model of Apert syndromewas generated (63). In humans, Apert syndrome isan autosomal dominant disorder characterised byskull, limb and visceral defects. Two-thirds of theaffected individuals carry a single S252W mutationin fibroblast growth factor receptor 2 gene(FGFR2). The transgenic Apert mouse model wasgenerated by homologous recombination to replacea segment of the normal mouse fgfr2 gene with asegment carrying a mutation equivalent to S252W inthe human protein. Heterozygous fgfr2+/S252W micedie within 36 hours of birth, due to defects of thepalate and respiratory system, whereas mosthuman sufferers of the disease survive past thisearly postnatal stage. Nevertheless, since the

440 Comment

fgfr2+/S252W mouse model shares many featureswith the human disease, it might serve as a usefulmodel of the disease, provided that its limitationsare carefully taken into account.

Down’s syndrome

Human Down’s syndrome (DS) is classified as acongenital disorder characterised by different levelsof mental retardation. It is caused by three geneticvariations, the most common of which is the pres-ence of an extra copy of chromosome 21 in all thecells of the affected individual’s body (trisomy 21).Mosaic trisomy 21 is similar, except that the extrachromosome 21 is only present in some of the cellsof the affected individual. The third form of DS iscaused by translocational errors that result inrepeats within chromosome 21.

Trisomy (Ts) mouse models were first isolated inthe 1970s, including the Ts16 mouse model thatpossesses an extra copy of mouse chromosome 16(64). The Ts16 mouse represents a naturally occur-ring trisomy in mice, but its relevance as a model ofDS is questionable, since the human chromosome21 and mouse chromosome 16 share only 80% of thesame genes, and human and mouse chromosomalnumbers and gene arrangements differ. Even moreproblematical is the fact that Ts mice display devel-opmental cardiovascular defects that are not evi-dent in human DS (65), but which resemblesymptoms associated with DiGeorge syndrome inhumans (66). Furthermore, these mice do not sur-vive beyond birth, prohibiting the study of CNSdevelopment and of the age-related diseases associ-ated with DS.

More recent mouse models of DS include theTs65Dn mouse, which carries an extra copy of partof mouse chromosome 16 which corresponds withthe critical region of human chromosome 21 (67)involved in the aetiology of the human form of thedisease. TS65Dn mice are able to survive to adult-hood, and display some of the characteristics ofhuman DS, including delayed postnatal develop-ment, muscular trembling, male sterility, skeletalmalformations and abnormal cholinergic function(68). However, although there appears to be someevidence of elevated levels of amyloid protein in theCNS of 6 month-old mice, the animals do not dis-play the characteristic signs of the later stages ofDS such as the APs and NFTs that predispose DSpatients to AD (69).

The most recent attempt to create a mouse modelof DS involved the addition of an almost entire copyof human chromosome 21 to the mouse genome,resulting in the Tc1 mosaic mouse, which containshuman chromosome 21 DNA in some of its cells(70). Each individual mouse potentially contains adifferent number of cells with the extra chromo-some, more closely mimicking mosaic trisomy 21,

rather than trisomy 21, the predominant form ofhuman DS. Tc1 mice exhibit several features ofhuman DS, including changes in cardiac functionthat were not accurately modelled by the Ts65Dnmouse (71). Since these Ts mice are able to survivepast birth, there is now considerable scope forstudying aspects of the development and function ofthe nervous system and premature ageing associ-ated with the human disease.

The most obvious problem with mouse models ofDS stems from the fact that none of the predispos-ing factors of the human disease, such as maternalage, are represented. Furthermore, the insertion ofhuman genetic material into mice may have unfore-seen effects which do not reflect the human condi-tion of the disease. Whether studies in these micemodels will lead to the development of gene thera-pies able to treat all the major symptoms of DS, isalso debatable. This is because DS is a collection ofhealth and developmental problems, resulting froman extra copy of a chromosome containing some 225genes, and there is evidence to suggest that addi-tional genes might also be involved in its aetiology(72). Hence, as there are already treatments avail-able for some complications of the disease, such asheart defects, resources would perhaps be betterfocused on developing treatments for other compli-cations arising from DS.

Mouse models of cognitive function and psychiatric disorders

Understanding the roles of genes involved in cer-tain inherited psychiatric disorders by using studiesin GA mice is particularly difficult, since, inhumans, the onset of these disorders can be trig-gered by non-genetic factors. Indeed, simply carry-ing a genetic defect may not necessarily mean thatthe disease will ever develop. Furthermore, GAmice may not necessarily develop neuropsychiatriceffects that can readily be extrapolated to humandiseases. GM mice have also been screened for sub-tle behavioural changes, to assist in the identifica-tion of genes that may be implicated in complexbehavioural disorders in humans, such as anxiety orschizophrenia. Again, behavioural defects are likelyto manifest themselves in very different ways inmice and humans.

Schizophrenia

Schizophrenia in humans is believed to be a poly-genic disorder, and has been attributed to muta-tions in genes on chromosomes 2, 13 and 15. Itsdevelopment can be triggered by external factors, soa positive genetic diagnosis does not necessarilymean that the disease will ever develop. In humans,the various forms of schizophrenia all cause dys-

Comment 441

functional social behaviour. Several GM mousemodels of schizophrenia recreate some of thesesymptoms, but none of them have successfullyrecreated all the symptoms associated with any ofthe forms of the human disorder. This may relate tothe fact that many of the human patients who suf-fer from these disorders do not inherit themthrough simple genetic determinants, since envi-ronmental factors play an important role.

CADASIL

A second example of how human behaviouraldefects can be displayed in very different ways inmice, concerns the mutant NOTCH 3 gene knock-inmouse model for stroke and dementia syndrome,which has been given the acronym, CADASIL (cere-bral autosomal dominant arteriopathy with subcor-tical infarcts and leukoencephalopathy; 73). Inhumans, CADASIL has been attributed to a mis-sense mutation in the NOTCH 3 epidermal growthfactor receptor gene, resulting in impaired post-translational processing and trafficking of theexpressed receptor. This causes degeneration ofvascular smooth muscle, resulting in clinical signssuch as recurring strokes and dementia in humansbetween the ages of 30 and 50 years. It has beensuggested that the absence of the CADASIL-likephenotype in heterozygotes (CADASILR142C/+) orhomozygotes (CADASILR142C/R142C) could reflectdifferences in one or more of: a) the spatial or tem-poral expression of species gene homologues; b)gene regulation; c) the organisation of the mouseand human brains; and d) life-span (because theshort lifespan of the mouse does not allow the dis-ease time to develop). What is clear is that thesemice are not adequate models of the human disease.

Interpretation of information from studies in GM mice

GM mice suffer from the further limitation that,although mouse proteins may appear to have iden-tical functions in humans, many genes and theirencoded proteins play vital roles, not only in earlydevelopment, but also in the aetiology of diseaseslater on, due to changes in protein expression inspecific tissues. Thus, mice in which the NR1 sub-unit of the NMDA receptor is deleted, die shortlyafter birth (74). However, mice in which NR1 geneexpression in the hippocampus was conditionallyknocked-out by using the CRE-LoxP system, sur-vived and showed signs of impaired spatial learning,consistent with the predicted role of the hippocam-pal NMDA pathway in learning (75). Similarly, it ispossible to mimic temporal changes in expressionby using cell-specific promoters, such as the tetra-cycline-responsive element, tTA, to drive the pro-

tein expression of a downstream gene. This can beachieved by supplementing the diet of transgenicmice with tetracycline and allows the effects ofchanges in protein expression on cognitive functionat various life stages to be monitored more pre-cisely.

Mouse models of cancer, DNA repair disorders and their use in toxicity testing

Carcinogenicity testing

Since the late 1980s, a number of transgenic mousestrains have been developed that are of interest inmutagenicity and carcinogenicity testing. Modelssuch as the Eµ-pim-1 transgenic mouse, the rasH2transgenic mouse, the p53+/– heterozygous knock-out mouse, and the Tg.AC transgenic mouse, havebeen used in carcinogenicity testing. The Eµ-pim-1transgenic mouse, which contains the activatedpim-1 oncogene, has a low spontaneous tumour fre-quency, yet develops cancers upon exposure to evenlow levels of mild genotoxic carcinogens or to nor-mal levels of mild genotoxic carcinogens that targetlymphatic tissues (76). The rasH2 transgenic mousecontains a mutation in the p21 Ha-ras oncogene,which enhances susceptibility to neoplastic induc-tion — an early marker of tumorogenesis (77). Thismodel is useful for assessing a variety of genotoxicand non-genotoxic carcinogens, and has been usedto obviate the need for conventional 2-year rodentbioassays, since carcinogens can be identified dur-ing 26-week studies (78).

p53+/– heterozygous knock-out mice containonly a single functional copy of the p53 tumour sup-pressor gene. Up to 36 weeks of age, the sponta-neous tumour frequency in these mice is very low.Although, unlike the rasH2 model, p53+/– het-erozygous knock-out mice are not suited to theassessment of non-genotoxic carcinogens, whenthese animals are exposed by almost any route ofexposure to certain classes of genotoxic agents(reviewed in 79), they develop cancers. In contrast,the Tg.AC transgenic mouse, which carries the acti-vated v-Ha-ras oncogene fused to a mouse globinpromoter, must be exposed topically to the chemicalin order to trigger the development of papillomas.Indeed, the US Food and Drug Administration havesuggested that the p53+/– mouse can be used forassessing genotoxic chemicals, except in the case ofdermal drugs, where the Tg.Ac model is preferred.On the other hand, the rasH2 model is consideredto be more suited to the assessment of non-geno-toxic chemicals. The rasH2, p53+/– and Tg.ACmodels have been used since 1997 in Europe forrapid carcinogenicity testing (< 1 year), in order toreduce the need for 2-year studies in rodents. Thishas resulted in considerable benefits to animal wel-

442 Comment

fare (80), but it should be remembered that theseassays have yet to be successfully validated for reg-ulatory use. The most obvious advantage is that,since these GM mice are more susceptible to cancerthan normal mice, fewer animals should be neededin each study to achieve statistically significantdata.

Mutagenicity testing

Another example of how the use of GM mice canhave a positive impact on animal welfare is the useof reporter mice for mutagenicity (or genotoxicity)testing. The Organisation for Economic Coopera-tion and Development (OECD) is considering thebenefits of using the LacI BigBlue® mouse andLacZ Muta™mouse transgenic models for muta-genicity testing (81). The mutagenic target ofMuta™mouse is a bacterial LacZ gene, and the tar-get of Big Blue is the LacI repressor gene. Both ofthese transgenes act as reporter genes. The LacIrepressor binds to the lac operon and prevents theexpression of β-galactosidase from the LacZ gene.Hence, mutations in the LacI repressor gene resultin changes in the levels of β-galactosidase expres-sion, whereas mutations in the LacZ gene result inthe expression of mutated forms of the enzyme.This means that exposure to a mutagen is, in the-ory, able to cause mutations which can be detectedby analysing β-galactosidase activity. This is doneby extracting genomic DNA extracted from the tis-sues of exposed mice, packaging the DNA into abacteriophage, and using it to infect Escherichiacoli in culture. These bacteria express β-galactosi-dase, the activity of which is related to the expres-sion level of the gene and whether it has beenmutated (Figure 1). The latter is determined bymonitoring the ability of infected E. coli colonies tocatalyse the formation of a coloured product. TheTSG-p53/Big Blue mouse (Taconic, New York,USA) has the LacI transgene integrated into everycell, and is hemizygous for the endogenous p53tumour suppressor. Thus, this model shares fea-tures of both the p53+/– and LacI BigBlue® mice,and is suited to the more-rapid screening of muta-genic and tumorigenic compounds.

The use of such models for mutagenicity testinghas the potential to provide a means to screenmutagens which may have escaped detection in sim-pler assays, such as the mouse spot test and mousebone-marrow micronucleus test, since it is possibleto monitor the ability of mutagens to trigger muta-tions in more than one type of tissue, without usinglarge numbers of animals (81, 82). Indeed, despitedifferences between the endpoints, tests based onthe use of these transgenic models and the mousespot test or in vivo micronucleus test have similarpredictabilities. However, tests in these transgenicmice are not able to detect mutagens that cause

chromosomal aberrations. Hence, a strategy thattakes into account the information which can bereasonably obtained by using ex vivo and cell-basedassays, such as the bacterial reverse mutation test(Ames test; OECD Test Guideline [TG] 471), the invitro mammalian cell chromosomal aberration,gene mutation, sister chromatid exchange andmicronucleus tests (TGs 473, 476 and 479, anddraft TG 487, respectively), should be appliedbefore tests are conducted in mice, in order to min-imise in vivo studies, whether or not they employGM animals. Indeed, as is discussed below, thereare reasons why human risk cannot necessarily beaccurately estimated from studies in mice alone.

The relevance of using mice to study human cancers

The incidence of cancer in humans has a strong cor-relation with age and genetic damage. This is con-sistent with the theory that the greater the numberof cycles of cell division, the greater the risk of can-cer. This relationship is less obvious in laboratorymice, where, despite similar rates of cell divisionand 105 times fewer cell divisions during the life-time of a mouse, susceptibility to cancer is higherthan in humans. In fact, around 30% of rodentsdevelop cancers during their 2-year to 3-year life-spans. Admittedly, in relative terms that take intoaccount the difference between the life-spans of thetwo species, the incidence of cancer in the twospecies is similar, but because, in comparison withhumans, mice are relatively deficient in antineo-plastic defence mechanisms (11), inferences as tothe underlying mechanisms of carcinogenesis aresubject to considerable debate. Higher metabolicrates occur in mice than in humans, and differencesin detoxification between mice and humans willundoubtedly also contribute to species differencesin the types, sites and incidences of cancer.

Most human cancers arise in epithelial cell layers.This is not usually the case in mice, although thetypes of tumours that occur in p53 mutant mice domore closely resemble the epithelial cancers foundin humans (83). This seems to be related to the factthat cells from these mice contain chromosomeswith more similar telomeres to human telomeres,the latter of which appear to be more susceptible toend-to-end chromatid fusion and non-reciprocaltranslocation of chromatid DNA, resulting inabnormal karyotypes — a common feature ofhuman (particularly epithelial) cancers, but notmouse cancers (84). This suggests a role in cancerresearch for the p53 +/– heterozygous mouse andtransgenic mice that contain mutants within themouse or human p53 genes. Indeed, these micehave been used to establish a number of cancermodels for drug development and basic cancerresearch (76–80, 85), as well as models for DNArepair-related diseases (see below).

Comment 443

Fig

ure

1:

A s

chem

atic

sh

ow

ing

th

e LacI/LacZ

and

LacZ

rep

ort

er g

ene

syst

ems

Tis

sue

DN

A p

repa

rati

on

Bac

teri

oph

age

clon

ing

E. c

oli

tran

sfor

mat

ion

Pla

que

assa

y

No

exp

osu

re t

o m

uta

gen

Th

e L

acI

/La

cZsy

stem

Th

e L

acZ

sy

stem

Lac

I

Lac

Ire

pres

sor

prot

ein

Lac

Z

No

β-ga

lact

osid

ase

expr

essi

on

inh

ibit

s

= a

ctiv

e L

acI

+ X

-gal

Lac

Z

β-ga

lact

osid

ase

expr

esse

d

= a

ctiv

e en

zym

e

+ X

-gal

Lac

Z

Non

or

less

fun

ctio

nal

β-ga

lact

osid

ase

expr

esse

d

= i

nac

tive

en

zym

e;

= a

ctiv

e en

zym

e

+ X

-gal

Lac

I

No

acti

ve L

acI

repr

esso

r

Lac

Z

β-ga

lact

osid

ase

expr

essi

on

inh

ibit

s

= m

uta

nt

Lac

I

+ X

-gal

Exp

osu

re t

o m

uta

gen

444 Comment

Immune-compromised mice in cancer research

Another group of mice that have featured heavilyin cancer research are mouse models based on thexenotransplantation of human tumours andtumour cell lines into immune-compromised mice.Early models involved the use of the nude mouse(86), a hairless and athymic mouse resulting fromspontaneous mutations in a homeobox gene. Thismouse carries two copies of a mutated forkheadbox N1 (FoxN1) transcription factor. The fact thatthe nude mouse lacks a thymus gland also ren-dered it unable to produce T-cells, essential com-ponents in foreign tissue rejection (6). Therefore,the nude mouse can be used to host humantumour cells, so that some forms of human cancercould be studied in the context of a live animal.Indeed, with the more recent advent of engineeredtumour cell lines that express bioluminescent orfluorescent reporter proteins and high resolutionimaging techniques, the growth and metastasis ofhuman tumours and their regression in responseto treatment can readily be monitored in theseanimals by non-invasive methods and overextended periods of time. Although the cellimmortalisation process will almost certainlyaffect the characteristics of the tumour cells, soinformation from such studies must be inter-preted with great care, this not only means thatfewer animals are needed for each study, but alsothat studies can be terminated when the tumourreaches a size at which it is likely to cause the ani-mal pain or distress. Indeed, the use of magneticresonance imaging (MRI) or positron enhancedtomography (PET) means that it is now possibleto monitor the fate of a single cancer cell within amouse (87). Such studies are far less reliant onthe development of large tumours, so the animalscan be euthanised before they suffer detectabledistress or pain.

However, such studies do not circumvent otheranimal welfare problems in the use of immune-compromised mice. For instance, as a result of thishomozygous mutation, 55% of nude mice diewithin two weeks of birth, and the surviving micedisplay poor hair growth and thin skin, low bodyweights, retarded growth and small body size. Thismeans that the mice are often unable to maintaina steady body temperature. Nude mice also havereduced fertility, due partly to abnormal gametedevelopment, and can exhibit signs of metabolicdisorders. These features not only affect the wel-fare of the animals that survive to adulthood, butalso mean that the strain is difficult to maintain.Similar concerns are raised by more-recentimmune-compromised mouse models, such asSCID mice and nude/SCID mice, as well as theeven more immune-compromised SCID-beige micethat also lack natural killer cell function, whichare increasingly being used because of the lower

incidence of tumour rejection in these mice (seeGA mouse models of infection). These animalsmust be afforded the highest quality of animalhusbandry, if inadvertent infections are to beavoided.

DNA repair disease models

Human DNA repair diseases have also been mod-eled in GA mice. Diseases such as xeroderma pig-mentosum (XP) and Cockayne syndrome (CS), aregenetic disorders due to specific defects innucleotide excision repair (NER; 88).

In XP, mutations in one or more of eight XPgenes (XPA-XPG and XP-variant), normally XPA,XPC, XPD and/or XPF, result in distinct patterns ofdefective DNA repair. However, the most strikingcharacteristics of all forms of XP are sunlight-induced pigment changes in the skin, erythema, dryskin, extreme sensitivity to sunlight, and photopho-bia. Affected individuals may also suffer from CNSdisorders such as sensorineural deafness and move-ment problems, as well as from immunologicaldefects. Continued exposure to sunlight can giverise to a variety of skin cancers. These additionalsymptoms are often due to loss-of-function muta-tions in the XPA gene which encodes a zinc fingerprotein, which participates in photoproduct recog-nition and DNA binding.

While xpa (–/–) knock-out mice are susceptible tosunlight-induced cancers (89), they do not exhibitvisible signs of erythema, nor do they display thecharacteristics of immunological defects or neuro-logical disorders. Nevertheless, the fact that the xpa(–/–) knockout mice, and also xpa/p53+/– doubleknock-out mice, are both susceptible to skin can-cers, and given the spontaneous incidence oftumours is low in mice of less than 11 months ofage, these models are potentially useful for geno-toxic carcinogenicity testing. This is particularlytrue of the double knock-out mouse, the more sen-sitive of the two models.

In CS, mutations in the CSA and CSB proteinsimpair transcription-coupled DNA repair. Patientswith the syndrome can exhibit signs of develop-mental defects, including severe physical and men-tal retardation, microencephaly and limb and visualdefects. They are also sensitive to sunlight. How-ever, while they develop a severe rash, they differfrom XP sufferers, in that they do not manifest anysigns of pigment changes nor hypersensitivity tosunlight-induced skin cancers. The disease is inher-ited as an autosomal recessive trait, and CS type 1can result in death during early infancy. csa (–/–)and csb (–/–) knock-out mice display a similar phe-notype to control mice, but with an increased inci-dence of carcinogenesis when exposed to UV light(90), despite this not being a feature of the corre-sponding human disease.

Comment 445

Discussion

The value of GA mice to human medicine:summary

We have considered the scientific merits and limita-tions of a range of available GA mice models thatare used to model human diseases and toxicity. Theresults are summarised in Table 2. These particularmodels were selected, since they are pivotal to ongo-ing projects to develop GA mouse models.

Although we initially set out to address the ques-tion of whether research using GA mice as modelsof human diseases has directly resulted in the devel-opment of new human medicines or treatments,this is almost impossible to determine without firstconsidering the entire history of how each diseasehas been studied. We have, therefore, selected alimited number of GA mouse models of specific dis-eases, with at least one example for each of themajor biological systems, and considered whetherthe mouse models have: a) increased understandingof the human disease; and b) been useful for assess-ing the safety and efficacy of existing and new ther-apeutic agents.

In our opinion, the current extent to which GAmice are used cannot be justified on the basis thatthey are vital for the development of human med-icines, since few human medicines have so farbeen developed which were largely or exclusivelybased on the use of GA mouse models. This unsat-isfactory situation is despite nearly four decadesof studies in GA mice — the first knockout, trans-genic and trisomy mice were all produced in the1980s. Indeed, the inability of many GA mousemodels to recapitulate all the features of a humandisease has often resulted in several mouse mod-els being created, for studies on different aspectsof the disease in question. Together with problemsof differences in the genetic backgrounds of themice used in these mutagenesis studies, this hasconfused the interpretation of the informationprovided, and has potentially slowed, rather thanexpedited, the development of new medical treat-ments.

One of the first two knock-out mice were nullmutants for the hprt (hypoxanthine-guanine phos-phoribosyl transferase) gene which is apparentlyhomologous to the human gene associated withLesch-Nyhan syndrome (LNS; 91). Despite theearly link between the gene and the correspondinghuman disease, only some of the symptoms of thedisease can be treated and there is no standard orspecific treatment for the neurological symptoms ofLNS. Similarly, the first transgenic mouse con-tained a mutated form of the metallothionein-1(MT-1) gene homologous to the human gene associ-ated with Menkes disease (92). Again, despite over20 years of research, the main way of managing this

disease is by injections of copper supplements, andthere is still no specific therapeutic treatment.Finally, trisomy mouse models have still not pro-vided a route for developing realistic treatments forDS (93).

This, albeit limited, set of examples highlightsthree questions that need to be considered beforeany new GA mouse models are generated: a) willstudies in GA mice further our understanding ofthe disease in question; b) how much would the newinformation provided by a GA model facilitate thedevelopment of new treatments for this disease;and c) would any such new treatment be betterthan any existing treatments or preventive meas-ures for the disease?

In the case of disorders such as cleft palate andDS, GA mouse models have permitted some of thefeatures of the diseases to be studied. However, it isnot clear whether this will assist with the manage-ment or prevention of the diseases themselves. Inparticular, there is intensive debate as to whethertreating cleft palate and DS as deformities isdegrading to individuals affected by these condi-tions. This could mean that, even if effective thera-pies could be devised, their use might not beethically acceptable. Certainly, the existence of suchcomplex disease aetiologies somewhat limits thescope for gene therapy, although they can beaddressed by surgical and symptomatic treatments. While it is clear that some GA mouse models haveproven useful in terms of providing relevant infor-mation about specific human diseases and biologicalprocesses, many more of the models are severelycompromised with regard to their relevance, andsome can even produce misleading information.This is partly due to the fact that there are many,often undefined, differences in the anatomy, physi-ology, metabolism, disease susceptibility and pat-terns of behaviour characteristic of humans andmice. Also, human diseases can arise because of spe-cific combinations of alleles of different genes andmutations that are induced within genetic or mito-chondrial DNA, as a result of environmental fac-tors, injury or infection, in ways that cannot befaithfully reproduced by using GA mice kept in con-trolled environments.

We have concluded that many GA mouse modelsthat carry well-defined genetic modificationsunderlying a particular disease phenotype areunsatisfactory for developing effective therapies.Therefore, it follows that models (such as thoseproduced by ENU mutagenesis), which share few,if any, of the underlying biochemical mechanismsof a disease, are likely to be even less suitable forthe above purpose. A further problem with suchmodels stems from the fact that ENU preferen-tially alters A–T base pairs (12), so some parts ofthe mouse genome are more susceptible to ENUmutagenesis than others. Since ENU induces 1base mutation every 100,000 to 1,000,000 bases,

446 Comment

each mutant animal could, in theory, carry thou-sands of mutations within its genome, many ofwhich will be within regulatory elements andhighly susceptible genes, and which will remainundetected, unless full genomic sequencing isundertaken. This means that the genetic back-ground of every animal could potentially be differ-ent, with the result that a direct link between agene and its function is extremely difficult toestablish in ENU mutant mice. Moreover, ENUtends to cause loss-of-function mutations, result-ing in a high incidence of in utero and early postpartum death. Therefore, it is questionablewhether further ENU-based mutagenesis screenswill yield any new and relevant models.

A case for using alternative models of human diseases

In view of the limitations of GA mice in the model-ling of human diseases in ways that have resulted inmany effective cures, and because some of the genesinvolved have been conserved in evolution, the pos-sibility of undertaking at least preliminary studiesin organisms such as Caenorhabditis elegans (anematode) and Drosophila melanogaster (an insect)deserves more serious attention. Indeed, the highergene density, often higher natural mutation rate,and relative ease with which the genomes of suchorganisms can be manipulated and studied, empha-sises the importance of a thorough investigationinto how studies in less sentient organisms can beused to curb the increasing animal welfare burdeninherent in using GA mice. Of course, these inver-tebrate organisms do not display the complexity ofthe human body. However, the above attributes,together with the much lower gene duplicationrates in these lower organisms, means that func-tional abnormalities due to gene mutants or knock-outs of the target gene will have a reducedlikelihood of being rescued by other genes. Thissimplifies the process of assigning gene function,which, in mice, is complicated by the prevalence ofgene duplication and gene compensation. Further-more, since lower organisms reproduce more rap-idly and some, like C. elegans, have transparentbodies, they are extremely well-suited to the studyof evolutionarily conserved developmental genesand processes. Hence, such organisms might bemore useful than GA mice, where the aim is todevelop a preventive treatment that stops the dis-ease from developing, thereby obviating the need todevelop medicines to treat symptoms. Many suchhuman diseases result in prenatal, postnatal or pre-mature death and are inherently difficult to modelusing mice.

Where there are problems with establishing therelevance of studies in lower organisms, and indeedusing GA mice, the most practical way forward is to

utilise cell culture methods and population geneticsto inform the development of clinical and cell-basedresearch into human diseases. Although these top-ics are beyond the scope of this paper, we hope thatmore emphasis will be placed on finding a way toutilise information that is already available fromsuch studies to implement a move away from thelarge scale generation and use of GA mice.

Scope for the Three Rs: recommendations

The use of GA mice as models of human diseasewarrants specific attention, since the induction ofsymptoms of human diseases in laboratory animalsunavoidably involves the potential to cause themconsiderable pain and suffering.

There should be an extensive reassessment of thevalue of GA mice as models of human diseases thattakes into account:

1. The failure of many GA mice to adequatelymodel human diseases;

2. The prospects for using the information fromthe studies to design new treatments for humandiseases; and

3. The availability and relevance of alternativemethods.

Physical and physiological differences betweenhumans and mice need to be considered in detail,prior to the generation of new models of human dis-eases. Where these differences are likely to compro-mise the relevance of models for their statedpurpose, this information must be considered care-fully before project licences are granted for theirgeneration.

The prospects of assigning functions to individualgenes by methods other than by using GA animalsshould be further investigated. Examples of suchmethods include the use of human populationgenetics information and the more extensive use ofgenomic analysis, genetically engineered cells andlower organisms to identify genes that may then goon to be studied by targeted mouse mutagenesis.

If prior evidence indicates the functional non-equivalence of mouse homologues of human dis-ease-related genes, or if the expression patterns ofthe true species homologues differ in mice andhumans, alternative investigative strategies shouldbe devised. This is particularly important, sincesome genes, such as the CFTR gene, that were oncethought to encode mouse homologues for humanproteins, have later been shown not to be function-ally equivalent. Thus, it is not acceptable to assumethat a close evolutionary relationship between miceand humans can guarantee functional homologybetween genes from the two species.

Comment 447

Tab

le 2

: A

su

mm

ary

of

the

scie

nti

fic

and

an

imal

wel

fare

co

nsi

der

atio

ns

invo

lved

in u

sin

g G

A m

ice

in m

edic

al r

esea

rch

an

d t

oxi

city

tes

tin

g

Are

a of

res

earc

hR

elev

ance

to

hu

man

hea

lth

ben

efit

Sci

enti

fic

lim

itat

ion

sA

nim

al w

elfa

re i

mp

lica

tion

s

Vac

cine

tes

ting

Thi

s is

a p

arti

al r

epla

cem

ent

that

mak

es it

pos

sibl

e to

R

elie

s on

oth

er c

ompo

nent

s of

the

H

as t

he c

apac

ity

to c

ause

ext

rem

eco

nduc

t so

me

test

s w

itho

ut u

sing

hig

her

mam

mal

sbi

oche

mic

al p

athw

ay b

eing

pre

sent

suff

erin

g un

less

hum

ane

endp

oint

sen

doge

nous

to

the

mic

ear

e us

ed

Cys

tic

fibr

osis

Lim

ited

, bec

ause

of

the

subs

tant

ial d

iffe

renc

es b

etw

een

CF

TR

hom

olog

ue in

mic

e is

not

a s

tric

t M

ice

suff

er f

rom

die

tary

and

dig

esti

veth

e ae

tiol

ogy

of t

he d

isea

se in

hum

ans

and

mic

efu

ncti

onal

hom

olog

ue o

f th

e hu

man

pro

tein

syst

em c

ompl

icat

ions

No

curr

ent

trea

tmen

t fo

r th

e di

seas

e. S

ympt

oms

are

Sign

ific

ant

diff

eren

ces

betw

een

mou

se

trea

ted

wit

h ge

neri

c dr

ugs,

e.g

. bro

ncho

dila

tors

, ste

roid

s,

mod

el a

nd h

uman

dis

ease

anti

biot

ics

and

DN

Ase

s (t

o br

eakd

own

muc

us).

Gen

eth

erap

y, d

espi

te p

rom

isin

g ou

tcom

es in

mou

se m

odel

s,

Mod

els

have

fai

led

to g

ener

ate

info

rmat

ion

did

not

prov

e be

nefi

cial

for

mos

t pa

tien

tsle

adin

g to

the

dev

elop

men

t of

eff

ecti

ve t

hera

py

Duc

henn

e m

uscu

lar

Pot

enti

ally

rel

evan

t m

odel

for

dev

elop

men

t of

gen

e T

he a

etio

logy

of

the

hum

an d

isea

se d

iffe

rs f

rom

So

me

mod

els

can

seve

rely

impa

ir

dyst

roph

yth

erap

yth

at s

een

in t

he m

ouse

mod

ello

com

otiv

e be

havi

our

and

surv

ival

Cur

rent

ly n

o sp

ecif

ic t

reat

men

t fo

r th

is o

r an

y ot

her

form

of

mus

cula

r dy

stro

phy

Dow

n’s

synd

rom

eC

urre

ntly

no

trea

tmen

tL

imit

ed t

o th

e st

udy

of d

evel

opm

enta

l and

In

the

cas

e of

sev

eral

mod

els,

mic

eea

rly

life

stag

e D

own’

s sy

ndro

me

do n

ot s

urvi

ve p

ast

birt

hP

oor

pros

pect

of

usin

g in

form

atio

n to

dev

elop

eff

ecti

ve

trea

tmen

ts, e

.g. g

ene

ther

apy

Som

e m

odel

s de

velo

p si

gns

that

aff

ect

thei

r be

havi

our

and

repr

oduc

tive

cap

acit

y, t

he

latt

er w

ith

impl

icat

ions

for

the

num

ber

of

mic

e us

ed t

o m

aint

ain

a st

rain

Dia

bete

sL

imit

ed s

cope

for

usi

ng c

erta

in m

odel

s in

dru

g Sp

ecie

s-sp

ecif

ic d

iffe

renc

es in

insu

lin r

egul

atio

nSo

me

mod

els

of d

iabe

tes

are

imm

une

deve

lopm

ent

com

prom

ised

and

pro

ne t

o in

fect

ion

Seve

ral e

xist

ing

trea

tmen

ts t

hat

pred

ate

or d

id n

ot r

ely

Poo

rly

unde

rsto

od r

oles

for

som

e pu

tati

ve

on c

ruci

al in

form

atio

n fr

om s

tudi

es in

GM

mic

eta

rget

s in

dia

bete

s

Neu

rode

gene

rati

veP

oten

tial

ly u

sefu

l for

dru

g de

velo

pmen

t. F

or e

xam

ple,

A

ge-r

elat

ed a

nd m

ulti

fact

oria

l dis

orde

rs w

ith

Im

pair

ed n

euro

logi

cal f

unct

ion

may

cau

se

diso

rder

sne

w p

utat

ive

targ

ets

for

vacc

ine

and

drug

tre

atm

ent

com

plex

aet

iolo

gies

are

dif

ficu

lt t

o m

odel

in

dist

ress

, alt

er b

ehav

iour

and

red

uce

the

have

bee

n id

enti

fied

for

Alz

heim

er’s

Dis

ease

shor

t-liv

ed s

peci

esvi

abili

ty o

f th

e an

imal

s

448 Comment

Tab

le 2

: co

nti

nu

ed

Are

a of

res

earc

hR

elev

ance

to

hu

man

hea

lth

ben

efit

Sci

enti

fic

lim

itat

ion

sA

nim

al w

elfa

re i

mp

lica

tion

s

Sick

le c

ell a

naem

iaD

evel

opm

ent

of a

nti-s

ickl

ing

drug

s an

d st

em c

ell

Fun

dam

enta

l dif

fere

nces

in s

ite

of

Poo

r ox

ygen

car

riag

e re

sult

ing

in d

istr

ess

ther

apie

s ar

e cu

rren

tly

unde

rway

alt

houg

h th

ere

is

haem

atop

oies

is, b

lood

vol

ume

and

red

bloo

d an

d pa

in d

ue t

o ra

pid

phys

iolo

gica

l cu

rren

tly

no c

ure

avai

labl

ece

ll si

ze a

nd c

linic

al e

ffec

ts

dege

nera

tion

Bon

e dy

smor

phia

Lim

ited

—m

ainl

y fu

ndam

enta

l res

earc

h to

und

erst

and

Dif

fere

nces

in t

he s

kele

tal a

nd d

enta

l ana

tom

ySo

me

mic

e ar

e un

able

to

feed

, see

or

mov

e ge

ne r

egul

atio

nno

rmal

ly, l

eadi

ng t

o ex

trem

e su

ffer

ing,

al

tere

d be

havi

our

and

earl

y de

ath

Que

stio

nabl

e sc

ope

for

ante

nata

l, pr

eim

plan

tati

on

scre

enin

g fo

r so

me

dise

ases

suc

h as

cle

ft p

alat

e be

caus

e of

eth

ical

obj

ecti

ons

Ape

rt s

yndr

ome

has

no g

ene

ther

apy

or d

rug-

base

d tr

eatm

ent.

All

trea

tmen

t is

sur

gica

l

Sens

ory

func

tion

Put

ativ

e ta

rget

s fo

r cl

assi

cal d

rug

targ

etin

g m

ay b

e T

he f

unct

iona

l gen

e cl

uste

rs in

hum

ans

and

Som

e m

ouse

mod

els

exhi

bit

rapi

d an

d id

enti

fied

for

sen

sori

neur

al d

eafn

ess

but

not

mic

e di

ffer

debi

litat

ing

loss

of

sens

ory

func

tion

wit

h re

tino

path

ies

seve

re a

nim

al w

elfa

re c

onse

quen

ces

Som

e st

em c

ell-b

ased

tre

atm

ents

for

deaf

ness

tha

t ha

veT

here

is a

larg

e de

gree

of

com

pens

ator

y ge

ne

reve

rsed

sen

sori

neur

al d

eafn

ess

in m

ice

are

curr

entl

y fu

ncti

on a

nd r

edun

danc

y in

sen

sory

per

cept

ion

bein

g in

vest

igat

ed b

ut h

ave

yet

to g

ive

a tr

eatm

ent

for

gene

s w

hich

mak

e it

dif

ficu

lt t

o m

odel

hum

an

hum

an d

eafn

ess.

Sim

ilar

trea

tmen

ts f

or s

ome

form

s se

nsor

y di

seas

es in

mic

eof

blin

dnes

s ha

ve b

een

succ

essf

ul in

mic

e

At

pres

ent

mos

t st

em c

ell t

hera

pies

hav

e ye

t to

be

prov

en s

afe

and

effe

ctiv

e

Cog

niti

ve d

isor

ders

Som

e co

ncer

ns o

ver

gene

tic

scre

enin

g of

hum

ans

to

Mul

tifa

ctor

ial d

isea

ses

that

are

dif

ficu

lt t

o U

npre

dict

able

and

oft

en s

ubtl

e ch

ange

sid

enti

fy s

usce

ptib

le in

divi

dual

s ba

sed

on p

oten

tial

ly

stud

y in

mic

e be

caus

e of

the

abs

ence

of

non-

in b

ehav

iour

wit

h po

tent

ial p

robl

ems

for

unso

und

assi

gnm

ent

of g

ene

mut

atio

ns f

rom

mou

se

gene

tic

fact

ors,

dif

fere

nces

in b

ehav

iour

al

anim

al c

are

stud

ies

effe

cts

of s

peci

fic

gene

mut

atio

ns a

nd

func

tion

al n

on-e

quiv

alen

ce o

f so

me

gene

s an

dth

e sh

orte

r lif

e sp

ans

of m

ice

Tox

icit

y te

stin

gH

uman

ris

k as

sess

men

t ba

sed

onSo

me

prob

lem

s w

ith

fals

e po

siti

ves

and

Lar

ge n

umbe

rs o

f an

imal

s m

ay b

e us

edre

lati

vely

sen

siti

ve, s

hort

-ter

m s

tudi

esex

trap

olat

ing

shor

t-te

rm s

tudi

es t

o lo

ng-t

erm

in

eac

h st

udy.

The

impl

emen

tati

on o

f th

e ex

posu

reE

U C

hem

ical

pol

icy

may

mea

n th

at m

any

mor

e m

ice

are

used

in t

oxic

ity

test

ing

Comment 449

Tab

le 2

: co

nti

nu

ed

Are

a of

res

earc

hR

elev

ance

to

hu

man

hea

lth

ben

efit

Sci

enti

fic

lim

itat

ion

sA

nim

al w

elfa

re i

mp

lica

tion

s

Can

cer

rese

arch

Use

ful t

rans

plan

tati

on b

ased

mod

els

for

test

ing

Som

e pr

oble

ms

wit

h th

e la

ck o

f fu

ncti

onal

Im

mun

e co

mpr

omis

ed a

nim

als

are

the

effi

cacy

of

nove

l the

rape

utic

sco

rres

pond

ence

of

som

e ge

nes

in m

ice

and

exte

nsiv

ely

used

in t

hese

stu

dies

giv

ing

hum

ans,

the

dif

fere

nces

bet

wee

n th

e ty

pes

rise

to

prob

lem

s w

ith

infe

ctio

n. S

ome

Pot

enti

ally

use

ful a

s a

mea

ns o

f id

enti

fyin

g of

can

cer,

inci

denc

e of

can

cer

and

life-

span

mod

els

have

a lo

w s

urvi

val.

Lar

ge

nove

l dru

g ta

rget

snu

mbe

rs o

f an

imal

s m

ay b

e us

ed in

each

stu

dySo

me

trea

tmen

ts, i

nclu

ding

mon

oclo

nal a

ntib

odie

s,

firs

t te

sted

in m

ouse

mod

els

are

now

clin

ical

ly u

sed.

So

me

expe

rim

enta

l gen

e th

erap

ies

Infe

ctio

us d

isea

ses

Som

e m

odel

s ar

e us

eful

in v

acci

ne/d

rug

deve

lopm

ent

The

imm

une

syst

ems

of h

uman

s an

d m

ice

are

Exp

erim

ents

can

cau

se a

wid

e va

riet

y of

an

d te

stin

g an

d ca

n ex

pedi

te t

esti

ng t

radi

tion

ally

no

t id

enti

cal,

whi

ch s

ugge

sts

that

vac

cine

s sy

mpt

oms,

incl

udin

g pa

raly

sis,

bre

athi

ng

cond

ucte

d in

pri

mat

es a

nd o

ther

hig

her

mam

mal

sm

ay n

ot e

licit

a h

umor

al r

espo

nse

in m

ice

diff

icul

ties

, les

ions

and

cha

nges

in b

ody

tem

pera

ture

Con

tain

men

t is

eas

ier

than

for

larg

er a

nim

als

Tes

ting

of

vacc

ines

req

uire

s th

e us

e of

larg

e nu

mbe

rs o

f an

imal

s

450 Comment

As a matter of principle, and to avoid any unnec-essary wastage of animals, GM mouse diseasemodels should, wherever possible, be created byefficient processes, so that they mimic all the keycharacteristics of the clinical condition as closelyas possible. This has now been facilitated by theavailability of conditional knock-out, tissue-spe-cific and inducible expression systems. GM micecreated by using such methods have increased sur-vival rates, and are already proving to be highlyrelevant for human diseases. However, suchimproved models should not only be used in in vivostudies, but should also serve, wherever possible,as sources of cells and tissues for replacement invitro studies.

While a few important models have been discov-ered by genome-wide mutagenesis, there is limitedevidence to suggest that ENU models have con-tributed to the development of human medicines,with the possible exception of GM mice used in thestudy of infectious diseases and some types of can-cer. Hence, ENU mutagenesis should only be usedif equivalent studies in lower organisms and cell-based systems, and genomic and molecular interac-tion analysis, are unable to provide informationabout the roles of human genes that is necessary forthe development of relevant models of human dis-eases. However, the screening of existing ENU andgene trap libraries might result in the avoidance ofthe further use of ENU mutagenesis altogether.

Conclusions

Over the past decade, there has been a dramaticshift toward the use of GA mice in research andtesting, which has, in turn, prompted concernsabout the welfare of the animals used. Issues suchas reducing the number of animals wasted duringthe production of GM mice, and the effective wel-fare assessment of GA mice, remain causes for con-cern.

The main question is whether the widespread useof GA mice can be scientifically justified. Only byaddressing this question objectively, and withoutbias, can we hope to identify the scope for replacingGA mice in research and testing. It should be recog-nised that the generation of GA mice has often con-fused, rather than improved, our understanding ofthe genetic basis of human diseases. The large num-bers of only partly-relevant models available formany diseases have complicated the meaningfulextrapolation of the information they provide tohuman medicine.

There is an urgent need to re-evaluate GA miceas models of human disease, to take into accountthe availability of alternative models based on stud-ies on lower organisms, normal and diseased humancells, and the increasingly availability of othersources of human information of direct relevance.

References

1. Mouse Genome Sequencing Consortium (2002). Ini-tial sequencing and comparative analysis of themouse genome. Nature, London 420, 520–562.

2. International Human Genome Sequencing Consor-tium (2001). Initial sequencing and analysis of thehuman genome. Nature, London 409, 860–921.

3. Anon. (2006). The International Mouse StrainResource. Website http://www.informatics.jax.org/imsr/ (Accessed 4.05.06).

4. RIKEN (2006). A Large Scale Mouse MutagenesisProgram in RIKEN GSC. Website http://www.gsc.riken.go.jp/Mouse/main.htm (Accessed 3.05.06).

5. The Baylor College of Medicine (2006). The BaylorCollege of Medicine Mouse Mutagenesis Project.Website http://www.mouse-genome.bcm.tmc.edu/(Accessed 3.05.06).

6. Jackson Laboratory (2006). The Mouse Heart, Lung,Blood and Sleep Disorders Center. Website http://pga.jax.org/ (Accessed 3.05.06).

7. National Institutes of Health (2006). Trans-NIHMouse Initiatives. Website http://www.nih.gov/science/models/mouse/ (Accessed 4.05.06).

8. Anon. (2006). Eumorphia. Website http://www.eumorphia.org/ (Accessed 3.05.06).

9. Home Office (2005). Statistics of Scientific Proce-dures on Living Animals: Great Britain 2005. 94pp.Home Office. Cm 6877. Website http://www.homeoffice.gov.uk/rds/pdfs06/spanimals05.pdf (Accessed 07.08.06).

10. The Royal Society (2001). The Use of Genetically Mod-ified Animals. 51pp. London, UK: The Royal Society.Website http://www.royalsoc.ac.uk/displaypagedoc.asp?id=11513 (Accessed 7.06.06).

11. Hanahan, D. & Weinberg, R.A. (2000). The hall-marks of cancer. Cell 100, 57–70.

12. Justice, M.J., Noveroske, J.K., Weber, J.S., Zheng, B.& Bradley, A. (1999). Mouse ENU mutagenesis.Human Molecular Genetics 8, 1955–1963.

13. National Institutes of Health (2006). The NIHKnock-out Mouse Project. Website http://www.nih.gov/science/models/mouse/knockout/index.html(Accessed 4.05.06).

14. Mepham, T.B., Combes, R.D., Balls, M., Barbieri, O.,Blokhuis, H.J., Costa, P., Crilly, R.E., de Cock Bun-ing, T., Delphire, C., O’Hare, M.J., Houdebine, L-M.,van Kreijl, C.F., van der Meer, M., Reinhardt, C.A.,Wolf, E. & van Zeller, A-M. (1998). The use of trans-genic animals in the European Union. The Reportand Recommendations of ECVAM Workshop 28.ATLA 26, 21–43.

15. Dragunsky, E., Nomura, T., Karpinski, K., Furesz,J., Wood, D.J., Perivikov, Y., Abe, S., Kurata, T.,Vanloocke, O., Karganova, G., Taffs, R., Heath, A.,Ivshina, A. & Levenbook, I. (2003). Transgenic miceas an alternative to monkeys for neurovirulencetesting of live oral poliovirus vaccine: Validation by aWHO collaborative study. Bulletin of the WorldHealth Organisation 81, 251–262.

16. Greaves, D.R., Fraser, P., Vidal, M.A., Hedges, M.,Ropers, D., Luzzatto, L. & Grosveld, F. (1990). Atransgenic mouse model of sickle cell disorder.Nature, London 11, 183–185.

17. Trudel, M., Saadane, N., Garel, M.C., Bardakdjian-Michau J., Blouquit Y., Guerquin-Kern J.L., Rouyer-Fessard P., Vidaud D., Pachnis A., Romeo P.H.,Beuzard Y. & Costantini, F. (1999). Towards a trans-genic mouse model of sickle cell disease: HemoglobinSAD. EMBO Journal 10, 3157–3165.

Comment 451

18. Paszty, C., Brion, C.M., Manci, E., Witowska, H.E.,Stevens, M.E., Mohandas, N. & Rubin, E.M. (1997).Transgenic knockout mice with exclusively humansickle haemoglobin and sickle cell disease. Science,New York 278, 876–878.

19. Manci, E.A., Hilliery, C.A., Bodian, C.A., Zhang,Z.G., Lutty, G.A. & Coller, B.S. (2006). Pathology ofBerkeley sickle cell mice: similarities and differenceswith human sickle cell disease. Blood 107,1651–1658.

20. Melanitou, E. (2005). The autoimmune contrivance:Genetics in the mouse model. Clinical Immunology117, 195–206.

21. Whitacre, C.C. (2001). Sex differences in autoim-mune disease. Nature Immunology 2, 777–780.

22. Wicker, L.S., Clark, J., Fraser, H.I., Garner, V.E.S.,Gonzalez-Munoz, A., Healy, B., Howlett, S., Hunter,K., Rainbow, D., Rosa, R.L., Smink , L.J., Todd, J.A.& Peterson, L.B. (2005). Type 1 diabetes genes andpathways shared by humans and NOD mice. Journalof Autoimmunity 25, 29–33.

23. Kanezaki, Y., Obata, T., Matsushima, R., Minami,A., Yuasa, T., Kishi, K., Bando, Y., Uehara, H.,Izumi, K., Mitani, T., Matsumoto, M., Takeshita, Y.,Nakaya, Y., Matsumoto, T. & Ebina, Y. (2004). KATPchannel knockout mice crossbred with transgenicmice expressing a dominant negative form of humaninsulin receptor have glucose intolerance but notdiabetes. Endocrine Journal 51, 133–144.

24. Hiota, C., Larsson, O., Shelton, K.D., Shiota, M.,Efanov, A.M., Hoy, M., Lindner, J., Kooptiwut, S.,Juntti-Berggren, L., Gromada, J., Berggren, P-O. &Magnuson, M.A. (2002). Sulfonylurea Receptor Type1 Knock-out Mice Have Intact Feeding-stimulatedInsulin Secretion despite Marked Impairment inTheir Response to Glucose. Journal of BiologicalChemistry 277, 37176–37183.

25. Kelbig, M.l., Wilkinson, J.E., Geisler, J.G. & Woy-chik, R.P. (1995). Ectopic expression of the agoutigene in transgenic mice causes obesity, features oftype II diabetes, and yellow fur. Proceedings of theNational Academy of Sciences of the USA 92,4728–4732.

26. Kerem, B., Rommens, J.M., Buchanan, J.A., Markie-wicz, D., Cox, T.K., Chakravarti, A., Buchwald, M. &Tsui, L.C. (1989). Identification of the cystic fibrosisgene: genetic analysis. Science, New York 245,1073–1080.

27. Anon. (2006). Cystic Fibrosis Mutation Database.Website http://genet.sickkids.on.ca/CFmutations96-1.html (Accessed 4.05.06).

28. Snouwaert, J.N., Brigman, K.K., Latour, A.M., Mal-ouf, N.N., Boucher, R.C., Smithies, O. & Koller, B.H.(1992). An animal model for cystic fibrosis made bygene targeting. Science, New York 257, 1083–1088.

29. Rozmahel, E., Gyomorey, K., Plyte, S., Nguyen, V.,Wilschanski, M., Durie, P., Bear, C.E. & Tsui, L.C.(1997). Incomplete rescue of cystic fibrosis trans-membrane conductance regulator deficient mice bythe human CFTR cDNA. Human Molecular Genetics6, 1153–1162.

30. Ameen, N., Alexis, J. & Salas, P. (2000). Cellularlocalisation of the cystic fibrosis transmembraneconductance regulator in mouse intestinal tract.Histochemistry & Cell Biology 114, 69–75.

31. Mall, M., Grubb, B.R., Harkema, J.R., O’Neal, W.K.& Boucher, R.C. (2004). Increased airway epithelialNa+ absorption produces cystic fibrosis-like lung dis-ease in mice. Nature Medicine 10, 487–493.

32. Kunzelmann, K., Schreiber, R., König, J. & Mall, M.(2002). Ion Transport Induced by Proteinase-Acti-vated Receptors (PAR2) in Colon and Airways. CellBiochemistry and Biophysics 36, 209–214.

33. Boot, H.J., Kasteel, D.T.J., Buisman, A-M. & Kim-man, T.G. (2003). Excretion of wild-type and vac-cine-derived poliovirus in the faeces of poliovirusreceptor-transgenic mice. Journal of Virology 77,6541–6545.

34. Jenkins, E.S. & Combes, R.D. (1999). The welfareproblems associated with using transgenic mice tobioassay for bovine spongiform encephalopathy. Ani-mal Welfare 8, 421–431.

35. Mestas, J. & Hughes, C.C.W. (2004). Of mice and notmen: Differences between mouse and human imm-unology. Journal of immunology 172, 2731–2738.

36. Perkins, J.D. (2006). Infection by proxy: The role ofchimeric mice in HBV and HCV research. LiverTransplantation 12, 323.

37. Chang, L.J. & Zhang, C. (1995). Infection and repli-cation of tat-minus human immunodeficiencyviruses: genetic analyses of LTR and tat GAs in pri-mary and long-term human lymphoid cells. Virology211, 157–169.

38. Meuleman, P., Libbrecht, L., De Vos, R., de Hemp-tinne, B., Gevaert, K., Vandekerckhove, J., Roskams,T. & Leroux-Roels, G. (2005). Infection by Proxy: TheRole of Chimeric Mice in HBV and HCV Research.Morphological and biochemical characterization of ahuman liver in a uPASCID mouse chimera. Hepatol-ogy 41, 847–856.

39. Ayash-Rashkovsky, M., Bentwich, Z., Arditti, F.,Friedman, S., Reisner, Y. & Borkow, G. (2005). Anovel small animal model for HIV-1 infection FASEBJournal express article 10.1096/fj.04-3184fje.

40. Klotman, P.E. & Notkins, A.L. (1996). Transgenicmodels of human immunodeficiency. Current Topicsin Microbiology & Immunology 206, 197–222.

41. Potash, M.J., Chao, W., Bentsman, G., Paris, N., Saini,M., Nitkiewicz, J., Belem, P., Sharer, L., Brooks, A.I.& Volsky, D.J. (2005). A mouse model for study of sys-temic HIV-1 infection, antiviral immune responses,and neuroinvasiveness. Proceedings of the NationalAcademy of Sciences of the USA 102, 3760–3765.

42. Rouquier, S., Blancher, A. & Giorgi, D. (2000). Theolfactory receptor gene repertoire in primates andmouse: Evidence for reduction of the functional frac-tion in primates. Proceedings of the National Acad-emy of Sciences of the USA 97, 2870–2874.

43. Gilad, Y., Wiebel, V., Przeworski, M., Lancet, D. &Paabo, S. (2004). Loss of olfactory receptor genescoincides with the acquisition of full trichromaticvision in primates. PloS Biology 2, 120–125.

44. Finberg, K.E., Wagner, C.A., Bailey, M.A., Paunescu,T.G., Breton, S., Brown, D., Giebisch, G., Geibel, J.P.& Lifton, R.P. (2005). The B1-subunit of the H+

ATPase is required for maximal urinary acidifica-tion. Proceedings of the National Academy of Sci-ences of the USA 102, 13616–13621.

45. Dou, H., Finberg, K., Cardell, E.L., Lifton, R. & Choo,D. (2003). Mice lacking the B1 subunit of H+-ATPasehave normal hearing. Hearing Research 180, 76–84.

46. Ruf, R., Rensing, C., Topaloglu, R., Guay-Woodford,L., Klein, C., Vollmer, M., Otto, E., Beekmann, F.,Haller, M., Wiedensohler, A., Leumann, E., Anti-gnac, C., Rizzoni, G., Filler, G., Brandis, M., Weber,J.L. & Hildebrandt, F. (2003). Confirmation of theATP6B1 gene as responsible for distal renal tubularacidosis. Pediatric Nephrology 18, 105–109.

452 Comment

47. Gibson, F., Walsh, J., Mburu, P., Varela, A., Brown,K.A., Antonio, M., Beisel, K.W., Steel, K.P. & Brown,S.D. (1995). A type VII myosin encoded by the mousedeafness gene shaker-1. Nature, London 374, 62–64.

48. ElAmraoui, A., Sahly, I., Picaud, S., Sahel, J., Abit-bol, M. & Petit, C. (1996). Human Usher 1B mouseshaker-1: The retinal phenotype discrepancyexplained by the presence/absence of myosin VIIA inthe photoreceptor cells. Human Molecular Genetics5, 1171–1178.

49. Libby, R.T. & Steel, K.P. (2001). Electroretino-graphic anomalies in mice with mutations in myo7a,the gene involved in human Usher syndrome type1B. Investigative Ophthalmology & Visual Science42, 770–778.

50. Sung, C-H., Makino, C., Baylor, D. & Nathans, J.(1994). A rhodopsin gene mutation responsible forautosomal dominant retinitis pigmentosa results ina protein that is defective in localization to the pho-toreceptor outer segment. Journal of Neuroscience14, 5818–5833.

51. Dalke, C. & Graw, J. (2005). Mouse mutants as mod-els for congenital retinal disorders. ExperimentalEye Research 81, 503–512.

52. Jacobs, G.H., Neitz, J. & Deegan, J.F. (1991). Reti-nal receptors in rodents maximally sensitive toultraviolet light. Nature, London 353, 655–656.

53. Cooper, B.J. (1989). Animal models of Duchenne andBecker muscular dystrophy. British Medical Bul-letin 45, 703–718.

54. Daingain, J. & Vrbova, G. (1984). Muscle develop-ment in mdx GA mice. Muscle & Nerve 7, 700–704.

55. Elbrink, J., Malhotra, S.K. & Hunter, E.G. (1987).Duchenne muscular dystrophy: Assessment ofexperimental data from animals in relation to thehuman disease. Medical Hypotheses 23, 131–136.

56. Gardner, K.L., Kearney, J.A., Edwards, J.D. &Rafael-Fortney, J.A. (2006). Restoration of all dys-trophin protein interactions by functional domainsin trans does not rescue dystrophy. Gene Therapy13, 744–751.

57. Richardson, J.A. & Burns, D.K. (2002). Mouse Mod-els of Alzheimer’s Disease: A Quest for Plaques andTangles. ILAR Journal 43, 89–99.

58. Holcomb, L., Gordon, M.N., McGowan, E., Yu, X.,Benkovic, S., Jantzen, P., Wright, K., Saad, I.,Mueller, R., Morgan, D., Sanders, S., Zehr, C., O’Campo, K., Hardy, J., Prada, C-M., Eckman, C.,Younkin, S., Hsiao, K. & Duff, K. (1998). AcceleratedAlzheimer-type phenotype in transgenic mice carry-ing both GA amyloid precursor protein and pre-senelin 1 transgenes. Nature Medicine 4, 97–100.

59. Gotz, J., Chen, F., van-Dorpe, J. & Nitsch, R.M.(2001). Tau filament formation in transgenic miceexpressing P301L tau. Journal of Biological Chem-istry 276, 529–534.

60. Itzhaki, R.F., Wozniak, M.A., Appelt, D.M. & Balin,B.J. (2004). Infiltration of the brain by pathogenscauses Alzheimer’s disease. Neurobiology of Aging25, 619–627.

61. Thyagarajan, T., Totey, S., Danton, M.J.S. & Kulka-rni, A.B. (2003). Genetically altered mouse models:The good, the bad and the ugly. Critical Reviews inOral Biology & Medicine 14, 154–174.

62. Satokata, I. & Maas, R. (1994). Msx-1 deficient miceexhibit cleft palate and abnormalities of craniofacialand tooth development. Nature Genetics 6, 348–356.

63. Ang, Y., Xiao, R., Yang, F., Karim, B.O., Iacovelli,A.J., Cai, J., Lerner, C.P., Richtsmeier, J.T., Leszi,

J.M., Hill, C.A., Yu, K., Ornitz, D.M., lisseeff, J.,Huso, D.L. & Jabs, E.W. (2005). Abnormalities incartilage and bone development in the Apert syn-drome FGFR2+/S252W mouse. Development 132,3537–3548.

64. Epstein, C.J., Cox, D.R. & Epstein, L.B. (1985).Mouse trisomy 16: An animal model of human tri-somy 21 (Down syndrome). Annals of the New YorkAcademy of Sciences 450, 157–168.

65. Bacchus, C., Sterz, H., Buselmaier, W., Sahai, S. &Winking, H. (1987). Genesis and systematization ofcardiovascular anomalies and analysis of skeletalmalformations in murine trisomy 16 and 19: Twoanimal models for human trisomies. Human Genet-ics 77, 12–22.

66. Waller III, B.R., McQuinn, T., Phelps, A.L., Mark-wald, R.R., Lo, C.W., Thompson, R.P. & Wessels, A.(2000). Conotruncal Anomalies in the Trisomy 16Mouse: An Immunohistochemical Analysis withEmphasis on the Involvement of the Neural Crest.The Anatomical Record 260, 279–293.

67. Akeson, E.C., Lambert, J.P., Narayanswami, S.,Gardiner, K., Bechtel, L.J. & Davisson, M.T. (2001).Ts65Dn — localization of the translocation break-point and trisomic gene content in a mouse modelfor Down syndrome. Cytogenetics & Cell Genetics 93,270–276.

68. Holtzman, D.M., Santucci, D., Kilbridge, J., Chua-Couzens, J., Fontana, D.J., Daniels, S.E., Johnson,R.M., Chen, K., Sun, Y., Carlson, E., Alleva, E.,Epstein, C.J. & Mobley, W.C. (1996). Developmentalabnormalities and age-related neurodegeneration ina mouse model of Down syndrome. Proceedings ofthe National Academy of Sciences of the USA 93,13333–13338.

69. Hunter, C.L., Bimonte-Nelson, H.A., Nelson, M.,Eckman, C.B. & Granholm, A-C. (2004). Behavioraland neurobiological markers of Alzheimer’s diseasein Ts65Dn mice: effects of estrogen. Neurobiology ofAging 25, 873–884.

70. O’Doherty, A., Ruf, S., Mulligan, C., Hildreth, V.,Errington, M.L., Cooke, S., Sesay, A., Modino, S.,Vanes, L., Hernandez, D., Linehan, J.M., Sharpe,P.T., Brandner, S., Bliss, T.V.P., Henderson, D.J.,Nizetic, D., Tybulewicz, V.L.J. & Fisher, E.M.C.(2005). An aneuploid mouse strain carrying humanchromosome 21 with Down syndrome phenotypes.Science, New York 309, 2033–2037.

71. Marino, B. (1993). Congenital heart disease in patientswith Down’s Syndrome: anatomic and genetic aspects.Biomedicine and Pharmacotherapy 47, 197–200.

72. Olson, L.E., Richtsmeier, T., Leszl, J. & Reeves, R.H.(2004). A chromosome 21 critical region does notcause specific down syndrome phenotypes. Science,New York 306, 687–690.

73. Lundkvist, J., Zhu, S., Hansson, E.M., Schwein-hardt, P., Qing, M., Beatus, P., Dannaeus, K., Karl-strom, H., Johansson, C.B., Viitanen, M., Rozell, B.,Spenger, C., Mohammed, A., Kalimo, H. & Lendahl,U. (2004). Mice carrying a R142C Notch 3 knock-inmutation do not develop a CADASIL-like phenotype.Genesis 41, 13–22.

74. Li, Y., Erzurumi, R.S., Chen, C., Jhaveri, S. & Tone-gawa, S. (1994). Whisker-related neuronal patternsfail to develop in the trigeminal brainstem ofNMDAR1 knockout mice. Cell 76, 427–437.

75. Tsien, J.Z., Huerta, P.T. & Tonegawa, S. (1996). Theessential role of hippocampal CA1 NMDA receptor-dependent synaptic plasticity in spatial memory.

Comment 453

Cell 87, 1327–1338.76. Kroese, E.D., van Steeg, H., van Oostrom, C.T., Dor-

tant, P.M., Wester, P., van Kranen, H.J., de Aries, A.& van Kreijl, C.F. (1997). Use of E-pim transgenicmice for short-term in vivo carcinogenicity testing:Iymphoma induction by benzo[a]pyrene but notTPA. Carcinogenesis 19, 975–908.

77. MacDonald, J.S. (2004). Human Carcinogenic RiskEvaluation, Part IV: Assessment of Human Risk ofCancer from Chemical Exposure Using a GlobalWeight-of-Evidence Approach. Toxicological Sci-ences 82, 3–8.

78. Yamamoto, S., Mitsumori, K., Kodama, Y., Mat-sunuma, N., Manabe, S., Okamiya, H., Suzuki, H.,Fukuda, T., Sakamaki, Y., Sunaga, M., Nomura, G.,Hioki, K., Wakana, S., Nomura, T. & Hayashi, Y.(1996). Rapid Induction of More Malignant Tumors byVarious Genotoxic Carcinogens in Transgenic MiceHarboring a Human Prototype c-Ha-ras Gene Than inControl Non-Transgenic Mice. Carcinogenesis 17,2455–2461.

79. Jacobson-Kram, D., Sistare, F.D. & Jacobs, A.C.(2004). Use of transgenic mice in carcinogenicityhazard assessment. Toxiological Pathology 32,Suppl. 1, 49–52.

80. ICH (1997). International Conference on Harmo-nization of Technical Requirements for Registrationof Pharmaceuticals for Human Use, Guideline S1BTesting for Carcinogenicity in Pharmaceuticals.11pp. Website http://www.ich.org (Accessed 2.06.05).

81. Wahnschaffe, U., Bitsch, A., Kielhorn, J. & Mangels-dorf, I. (2005). Mutagenicity testing with transgenicmice. Part I: Comparison with the mouse bone mar-row micronucleus test. Journal of Carcinogenesis 4,doi:10.1186/1477-3163-4-3. Website http://www.carcinogenesis.com/content/4/1/3 (Accessed 25.05.06).

82. Wahnschaffe, U., Bitsch, A., Kielhorn, J. & Mangels-dorf, I. (2005). Mutagenicity testing with transgenicmice. Part II: Comparison with the mouse spot test.Journal of Carcinogenesis 4, doi:10.1186/1477-3163-4-4. Website http://www.carcinogenesis.com/content/4/1/4 (Accessed 25.05.06).

83. Artandi, S.E., Chang, S., Lee, S.L., Alson, S., Gottlieb,G.J., Chin, L. & DePinho, R.A. (2000). Telomere dys-

function promotes non-reciprocal translocations andepithelial cancers in mice. Nature, London 406,641–645.

84. Attardi, L.D. & Jacks, T. (1999). The role of p53 intumour suppression: lessons from mouse models.Cellular and Molecular Life Sciences 55, 48–63.

85. Van Dykes, J. & Jacks, J. (2002). Cancer Modellingin the Modern Era: Progress and Challenges. Cell108, 135–144.

86. Flanagan, S.P. (1966). ‘Nude’, a new hairless genewith pleiotropic effects in the mouse. GeneticResearch 8, 295–309.

87. Hudson, M. (2005). The welfare and scientific advan-tages of non-invasive imaging of animals used in bio-medical research. Animal Welfare 14, 303–317.

88. Lehmann, A.R. (2003). DNA repair-deficient disease,xeroderma pigmentosum, Cockayne syndrome andtrichothiodystrophy. Biochimie 85, 1101–1111.

89. Van Kreijl, C.F., McAnulty, P.A., Beems, R.B., Vynck-ier, A., van Steeg, H., Fransson-Steen, R., Alden, C.L.,Forster, R., van der Laan, J-W. & Vandenberghe, J.(2001). Xpa and Xpa/p53+/– Knockout mice: overviewof available data. Toxicological Pathology 29, 117–127.

90. Van der Horst, G.T.J., van Steeg, H., Berg, R.J.W.,van Gool, A.J., de Wit, J., Weeda, G., Morreau, H.,Beems, R.B., van Kreijl, C.F., de Gruijl, F.R.,Ootsma, D. & Hoeijmakers, J.H.J. (1997). Defectivetranscription-coupled repair in Cockayne syndromeB mice is associated with skin cancer predisposition.Cell 89, 425–435.

91. Mansour, S.L., Thomas, K.R. & Capecchi, M.R.(1988). Disruption of the proto-oncogene int-2 inmouse embryo-derived stem cells: a general strategyfor targeting mutations to non-selectable genes.Nature, London 336, 348–352.

92. Cox, D.R. & Palmiter, R.D. (1982). Assignment ofthe mouse metallothionein-1 gene to chromosome-8é Implications for human Menkes disease. ClinicalResearch 30, A117–A117.

93. Epstein, C.J., Cox, D.R. & Epstein, L.B. (1985).Mouse trisomy 16: An animal model of human tri-somy 21 (Down syndrome). Annals of the New YorkAcademy of Sciences 450, 157–168.

454 Comment