Mammary gland, limb and yolk sac defects in mice lacking Tbx3,the gene mutated in human ulnar mammary syndrome

Mutations in the human T-box transcription factor gene TBX3underlie the autosomal dominant disorder known as ulnar-mammary syndrome (UMS;OMIM 181450) (Bamshad et al.,1997). Human TBX3 was identified during an attempt to positionally clone the gene responsible for another developmental disorder,Holt-Oram syndrome, which had been mapped to chromosome 12q. Although it was later found that Holt-Oram syndrome is caused by mutations in the closely linked human TBX5 gene (Basson et al., 1997; Li et al.,1997), mutations in TBX3 were identified in individuals with ulnar-mammary syndrome, a developmental disorder that was also mapped to this chromosomal region (Bamshad et al.,1995; Bamshad et al.,1997). A number of different mutations in TBX3 have been identified in families with ulnar-mammary syndrome, most of which are predicted to encode truncated proteins or missense mutations resulting in loss of function (Coll et al.,2002) (for a review, see Papaioannou, 2001). UMS is fully penetrant but with a highly variable clinical presentation, even among family members with the same mutation. As the name suggests, the syndrome is predominately characterized by posterior forelimb deficiencies or duplications, involving the ulna and little finger (with rare involvement of the hindlimb), and apocrine/mammary gland hypoplasia or dysfunction with absent or abnormal nipples. Other common features are abnormal dentition,delayed puberty, genital abnormalities and growth retardation(Bamshad et al., 1995; Bamshad et al., 1999; Bamshad et al., 1996). There have been no reports of humans homozygous for TBX3 mutations.

TBX3 and the mouse ortholog, Tbx3, which was discovered by Bollag et al. (Bollag et al.,1994), are members of the Tbx2 subfamily of the T-box family of transcription factor genes. The family is characterized by a highly conserved sequence encoding a DNA-binding domain, called the T-domain, that binds specific sequences in the promoters of target genes. The Tbx2 subfamily, which includes Tbx2, Tbx3, Tbx4 and Tbx5, evolved by tandem duplication of an ancestral locus to form a linked pair of genes, followed by duplication of the gene pair and dispersal to two chromosomal locations. In both mouse and human, Tbx2 and Tbx4 are linked, and Tbx3 and Tbx5 are linked(Agulnik et al., 1996; Papaioannou, 2001). The relatively recent separation of the paralogs, Tbx2 and Tbx3,is reflected in ∼90% identity in amino acid sequence in the T-box and similarities in expression patterns and function of the two genes(Agulnik et al., 1996; Brummelkamp et al., 2002; Chapman et al., 1996b; Gibson-Brown et al., 1998a; Gibson-Brown et al., 1998b; Jacobs et al., 2000; Lingbeek et al., 2002). Both of the mouse genes and their human counterparts can act as transcriptional repressors (Carlson et al.,2001; Carreira et al.,1998; He et al.,1999), and although no downstream target genes have yet been identified for mouse Tbx3, the human genes TBX3 and TBX2 were both identified in functional screens as negative regulators of the cell cycle control gene Cdkn2a. TBX2 and TBX3 repress the Cdkn2a transcript p14^ARF in humans (known as p19^ARF in mouse) through an interaction with a variant T-domain binding site that is specific to these two T-box genes(Brummelkamp et al., 2002; Jacobs et al., 2000; Lingbeek et al., 2002). Recently, the crystal structure of TBX3 in complex with DNA has been reported and suggests that TBX3 binds as a monomer to its natural target sites(Coll et al., 2002).

The highly specific patterns of T-box gene expression during development implicate them in many developmental processes(Papaioannou, 1997; Papaioannou, 2001; Smith, 1999). In addition to the UMS mutations in TBX3, mutations in several other T-box genes confirm their vital developmental roles in a variety of metazoan species,including human and mouse (Basson et al.,1997; Bruneau et al.,2001; Chapman et al.,1996a; Jerome and Papaioannou,2001; Russ et al.,2000). In general, the phenotypic abnormalities observed in these mutants correlate with areas of gene expression. In mouse, Tbx3 is first expressed in the inner cell mass (ICM) of the blastocyst, and later in the mesoderm and endoderm layers of the yolk sac, and in the mesoderm of the amnion, chorion and allantois during gastrulation. During organogenesis, Tbx3 is expressed in the primordia of many tissues and organs,notably in the apical ectodermal ridge (AER) and the anterior and posterior margins of the developing limbs, and in the epithelium of the mammary buds(Chapman et al., 1996b; Gibson-Brown et al., 1996; Gibson-Brown et al., 1998b; Yamada et al., 2000). Tbx3 is also expressed in various adult organs(Bollag et al., 1994). Less is known about the spatial pattern of expression in humans, but the gene is expressed in a variety of adult organs, as well as in fetal lung, kidney,heart, liver, spleen and pituitary gland(Bamshad et al., 1999; Bamshad et al., 1997). It has been postulated that the abnormalities seen in individuals with UMS result from a haploinsufficiency of TBX3 that leads to hypoplasia of tissues through an interference with apoptosis or differentiation during critical stages of embryogenesis (Bamshad et al.,1997; Carlson et al.,2002). To provide a mouse model of UMS in order to understand the developmental role of Tbx3, we have generated a mutation in mouse Tbx3 by gene targeting in embryonic stem cells. In contrast to human UMS mutations, the mouse Tbx3 mutation is recessive on several genetic backgrounds, with the exception of a mild defect in female genitalia. However, the homozygous mutant phenotype reflects the predominant features of UMS with severe defects in forelimb and mammary gland development and, in addition, reveals a novel role for Tbx3 in the yolk sac.

Mouse Tbx3 genomic clones were isolated from a genomic λphage 129/SvJ library (Stratagene, La Jolla, CA). The targeting construct,which contains 7.2 kb of total homology was generated by replacing a 3.2 kb SpeI fragment containing three coding exons of the T-box with a loxP-flanked selection cassette containing the herpes simplex virus thymidine-kinase gene (hsv-tk) under the control of its own promoter and the neomycin-resistance gene under the control of the phosphoglycerate kinase promoter (pgk-neo). The diptheria toxin gene under the control of theβ-actin promoter was placed 5′ of the upstream homology for negative selection and the construct was linearized at a unique NotI site. The linearized targeting construct(Fig. 1A) was electroporated into R1 ES cells (Nagy et al.,1993) and drug-resistant clones were screened for homologous recombination using Southern blot analysis(Fig. 1A,B). One targeted cell line was recovered and was then transiently transfected with a circular pIC-Cre plasmid (Gu et al.,1993) and grown in media containing FIAU to select for removal of the thymidine kinase gene (Fig. 1A). The targeting strategy deletes three exons encoding 158 amino acids of the T domain and, if splicing were to occur between exons 1 and 4 flanking the deleted region, a frameshift would lead to a premature stop codon at the next codon. Three Cre-excised cell clones were recovered and injected into C57BL/6Tac blastocysts (Taconic Farms, Germantown, NY) to produce chimeras (Papaioannou and Johnson,2000), which were bred to C57BL/6Tac and 129 mice. Germline transmission of the targeted, Cre-excised allele, Tbx3^tm1Pa, was obtained from all three cell lines. Subsequent analysis using each of these three lines gave the same range of phenotypes and data from these lines have been combined.

Analyses of neonatal and adult heterozygous mice was performed on the following genetic backgrounds: (C57BL/6TacX129)F₁, F₂and F₂ intercross, and 129 inbred. Analysis of embryos was performed on these backgrounds, as well as on random-bred Swiss-Webster and ICR mice (Taconic Farms, Germantown, NY). Because no differences in phenotype were seen on these different backgrounds, the data have been combined. Heterozygous mice and wild-type littermates were examined for limb and tooth defects, and weighed at 2, 4 and 6 weeks of age. Females were examined for the presence and number of nipples, and the day of vaginal opening was scored. External genitalia were examined at birth, weaning and in adults. Embryos were collected from timed matings (day of the plug=0.5 days post coitus, dpc) and fixed in either Bouin's fixative for histology, or 4% paraformaldehyde for in situ hybridization or immunohistochemistry. Mutant and wild-type embryos, with their yolk sacs and placentae, were histologically examined after Hematoxylin and Eosin staining at 8.5, 10.5, 12.5 and 13.5 dpc. Offspring and embryos were genotyped using a three-primer PCR protocol designed to amplify wild-type and mutant bands using the following primers: (1)5′-CTAAGCCTGATGGTGTGAGA-3′; (2)5′-GATAAACCTCAATGCTACCTAC-3′; and (3)5′GACTTATCATTCAGCCCAGCAG-3′(Fig. 1C). PCR conditions were 3 minutes at 94°C, 40 cycles (30 seconds at 94°C, 30 seconds at 59°C, 30 seconds at 72°C), and 10 minutes at 72°C.

Whole-mount in situ hybridization was performed as described previously(Wilkinson, 1992). PECAM-1 antibody staining (PharMingen; San Diego, CA) on whole yolk sacs was performed using standard protocols (Davis et al.,1991). Apoptosis was examined in yolk sacs from 9.5 dpc and 10.5 dpc embryos that were living at the time of recovery (two mutant and two wild-type at each age) using immunohistochemistry for activated caspase (Cell Signaling technology, NEB, Beverley, MA)(Di Cunto et al., 2000). For detection of mammary bud induction in homozygous mutant embryos, embryos alive at the time of recovery were examined by in situ hybridization for Wnt10b (11.5 and 12.5 dpc) and Lef1 (12.5 and 13.5 dpc),along with stage and age-matched controls. Embryos (12.5 and 13.5 dpc) were examined histologically in serial sections as described above for evidence of mammary bud development. The following markers of limb development were examined by in situ hybridization on mutants and age/stage matched controls: Shh (10.5, 11.5 and 12.5 dpc); Fgf8, engrailed 1(En1), Wnt7a and Tbx2 (10.5 and 11.5 dpc); Msx1, Tbx4 and Tbx5 (10.5 dpc); Hand2 (9.5 and 10.5 dpc).

Inguinal mammary glands were collected from six-week-old virgin females,pregnant females (12.5 days of gestation), lactating females (2 weeks after parturition) and postlactation females (1 week after weaning). The virgin and pregnant females were from both 129 inbred and C57BL/6Tac-129 mixed background mice (F₂ and F₂ intercross), whereas the lactating and postlactation glands were from the mixed background only. Three or four glands were examined from each genotype and background. In each case, the right inguinal mammary gland (number 4) was isolated from the skin by blunt dissection and scraping with a scalpel, and then spread onto a glass slide for overnight fixation in Carnoy's fixative. Fixed glands were stained in Carmine Alum, using standard procedures (Kordon and Smith, 1998). Glands were scored using a semi-quantitative measure for the extent of ductal development and the ratio of terminal end buds to end buds.

Standard methods were used for Alcian Blue/Alizarin Red skeletal preparations of embryos (Mallo and Brändlin, 1997). For measurements of the posterior zeugopod bones, late fetal stage embryos (17.5 dpc for ulna measurements and 18.5 dpc for fibula measurements) were stained and the limb bones were measured using a grid under a dissecting microscope.

Mice heterozygous for the Tbx3^tm1Pa allele are born at the expected frequency (72/163 from 34 litters of heterozygous × wild type matings; χ²=2.21, P>0.05), and are viable and fertile. On different genetic backgrounds, no differences were observed in growth rates between heterozygous (n=44) and wild type(n=64) mice (data not shown), in the number and location of nipples in females, or the age of puberty in females, as measured by the day of vaginal opening (average for heterozygotes, 26.3 days, n=21; average for wild type, 25.8 days, n=29; Mann-Whitney U test, z=0.934; P>0.17). Dentition was grossly normal. Skeletal preparations of late fetal stages were carried out for the examination of the bones of the fore- and hindlimbs. The number of digits and digit bones were all normal and there was no fusion of carpal bones, a common feature of UMS. There was no difference in the lengths of the posterior zeugopod elements, the ulna and fibula, between heterozygous and wild-type mice(Table 1).

No difference was seen between wild-type and heterozygous females in the extent of ductal development of the virgin inguinal (number 4) mammary glands(Fig. 2A,B) or in the ratio of end buds to terminal end buds. Pregnant, lactating and postlactation involuting inguinal mammary glands were indistinguishable between heterozygous and wild-type females (Fig. 2C,D), and heterozygous mice had no apparent difficulty in nursing and rearing offspring.

The only trait that differentiated Tbx3 heterozygous from wild-type mice was a minor difference in external genitalia in adult mice. In a sample of 85 heterozygous and 82 wild-type mice from mixed genetic backgrounds, all of the heterozygotes and none of the wild-type females had a small ventral split in the glans clitoris, resulting in a distal bifurcation of the prepuce. In addition, a small proportion of these heterozygous females(9/85) showed a failure of vaginal opening, whereas this was not observed in the wild-type mice.

No Tbx3^tm1Pa homozygous mutants were recovered at term in 18 litters from inter se matings of Tbx3^tm1Pa/+ mice. Examination of embryos during gestation revealed that homozygous mutants were present at the expected Mendelian ratios(Table 2A;χ ²=0.0318, P>0.05), but about half of them (17/35)were dead by 11.5 dpc and none survived beyond about 16.5 dpc(Table 2B). The occasional abnormalities seen in +/+ or Tbx3^tm1Pa/+ embryos ranged from open brains to a placenta with no embryonic remains, whereas the phenotype of Tbx3^tm1Pa/Tbx3^tm1Pahomozygous mutant embryos consists of three major abnormalities: yolk sac defects, lack of mammary glands and limb defects. Prior to the onset of lethality at 10.5 dpc, some homozygous mutant embryos have normal yolk sacs(Fig. 3E,F), but others (6/24)show a reduction in yolk sac vasculature(Fig. 3A). In these mutant embryos, yolk sac blood vessels were small or absent, as confirmed by immunostaining for an endothelial marker, PECAM(Fig. 3B-D), and sometimes contained no detectable blood, although blood was present in the heart and embryonic circulation. At 9.5 and 10.5 dpc, no difference in the amount of apoptosis was detected between mutant and wild-type yolk sacs of living embryos using activated caspase immunohistochemistry (data not shown). By 12.5 dpc, however, most mutant embryos were dead (21/33) and the yolk sacs of many of the surviving mutants were abnormal. Histologically, the endoderm layer of abnormal-appearing yolk sacs consisted mostly of dead cells with pyknotic nuclei, and the blood vessels had begun to deteriorate(Fig. 3G,H). No obvious abnormalities were seen in placental morphology or morphology of fetal organs and tissues with the exception of a smaller than normal liver (see Fig. 2M-P). No surviving homozygous mutant embryos were observed beyond 13.5 dpc, although two dead fetuses were recovered near term that had survived to about 16 dpc, judging from their developmental stage (Fig. 4E-H).

Mammary gland development begins with the induction of localized ectodermal placodes along the lateral body wall at 10.5 dpc that results in the formation of discrete, lens-shaped epithelial mammary buds by 12.5 dpc. Tbx3expression first begins to appears in the epithelium of the nascent mammary bud at 11.5 dpc and is present in the buds of all male and female embryos between 12.5 and 14.5 dpc (Chapman et al.,1996b) (data not shown). Tbx3 expression is also seen in adult mammary glands and mammary tumors by Northern analysis (T.G.D. and V.E.P., unpublished). Induction of the mammary bud epithelium was examined in mutant embryos that were still alive at 11.5-13.5 dpc using histology and in situ hybridization for Wnt10b and Lef1. Wnt10b is normally expressed in a raised streak of lateral body wall epithelium at 11.5 dpc, just before the formation of the mammary buds, and then becomes restricted to epithelial mammary buds at 12.5 dpc(Christiansen et al., 1995). Lef1 is also expressed in the epithelial buds between 11.5 and 13.5 dpc, and is then downregulated (Mailleux et al., 2002). In Tbx3 mutant embryos, neither Wnt10b nor Lef1 was observed between 11.5 and 13.5 dpc(Fig. 2E-L). This lack of mammary bud induction is not due to a general developmental delay, which is sometimes observed in later stage mutants, as wild-type controls clearly showed evidence of mammary bud induction at comparable developmental stages(e.g. compare Fig. 2I with 2L). Seven 12.5-13.5 dpc mutant embryos were examined histologically in serial sections for evidence of mammary bud induction(Fig. 2M-P). No buds were detected at 12.5 dpc, although at 13.5 dpc one mutant has a single, small pair of mammary buds at the level corresponding to bud number 2, and a second mutant has a single, unilateral bud in a similar position(Fig. 2P). These embryos were somewhat smaller than littermate controls but were judged by the development of the ear pinnae and the vibrissa to be of a comparable developmental stage,appropriate to their chronological age. No cell death was observed within these buds, although they were less well organized than the buds from wild-type embryos (Fig. 2O,P).

Homozygous mutant embryos are normal at 8.5 dpc, but about 70% of mutants are identifiable by reduced hindlimb bud development by 9.5-10.5 dpc, whereas the forelimb bud appears to be normal in size and shape at this stage. By 11.5 dpc, all homozygous embryos that were alive, or were dead but sufficiently intact to be scored, displayed both fore- and hindlimb abnormalities, with irregularities in the hand plate and either very little development or an irregular shape to the foot plate that resembled a posterior deflection of the autopod (Table 2B; Fig. 4A-D). The AER is present in both fore- and hindlimbs, although it appears reduced in the forelimb and more severely reduced in the hindlimb as indicated by Fgf8 expression(Fig. 5A). Msx1, which marks cells in the progress zone, was expressed in a normal graded pattern in the fore- and hindlimbs of Tbx3 homozygous mutant embryos, although the expression in the forelimb appeared to be restricted posteriorly (data not shown). Dorsoventral patterning of the limb is normal, as indicated by the expression of Wnt7a in the dorsal ectoderm and engrailed 1 in the ventral ectoderm (data not shown). The expression of two T-box genes implicated in the specification of fore-versus hindlimb, Tbx5 and Tbx4, respectively, is normal(Fig. 5C,D). However, Tbx2, a closely related T-box gene with normal expression largely overlapping that of Tbx3 in the limb bud margins, shows reduced expression in the posterior margin of the forelimb and no expression in the posterior margin of the hindlimb of Tbx3 homozygous mutants(Fig. 5B). Sonic hedgehog(Shh) expression, which marks the zone of polarizing activity (ZPA)at the posterior limb margin, was present in fore- and hindlimbs of three out of four wild-type embryos at 10.5 dpc, and in both limbs of all wild-type embryos at 11.5 dpc. In mutant embryos, however, there is no expression in stage-matched mutants at 10.5 dpc (Fig. 5E) and only one out of four mutants showed any Shhexpression at 11.5 dpc. In this embryo, expression was reduced in the forelimbs and limited to a very small area in the hindlimbs(Fig. 5F). Hand2,which is initially expressed in the flank and then becomes localized to the posterior limb buds (Charite et al.,2000), is normally expressed in the flank and posterior forelimb bud at 9.5 dpc in mutant and wild-type embryos. However, by 10.5, mutant embryos (n=4) showed a reduced domain of posterior forelimb bud expression compared with stage-matched controls, and none has expression in the hindlimb buds (Fig. 5G).These early limb abnormalities are reflected in corresponding abnormalities of the limb skeleton of the small number of homozygous mutants that reach late fetal stages. In one fetus of ∼16.5 dpc(Fig. 4E-G), the right forelimb had a shortened ulna and was missing the metacarpals and phalanges of the fifth digit, and phalanges of the fourth digit. The left forelimb lacked an ulna and carpals as well as metacarpals and phalanges of the fourth and fifth digits (Fig. 4G). In the more severely affected hindlimb, the femur was present, but a single misshapen element represented the zeugopod. A single digit was attached to the zeugopod element of the right hindlimb. In a second fetus of the same approximate age,the right forelimb was shortened but contained an ulna, radius and five digits. The left forelimb was missing the ulna, and the metacarpals and phalanges of the fourth and fifth digits. In the hindlimbs, the femur was present bilaterally; the left hindlimb resembled that of the fetus described above, with a single digit attached to the zeugopodal element; the right hindlimb had a femur, a tibia and a single digit attached to the tibia(Fig. 4H). The pelvic bones were small and poorly formed in both fetuses(Fig. 4G).

We have produced a mutation in the mouse T-box gene Tbx3 that closely corresponds to truncation mutations seen in some individuals with UMS that are predicted to lead to haploinsufficiency through nonsense-mediated mRNA decay (Bamshad et al.,1999). The targeting strategy we used to produce a nonfunctional allele of Tbx3 deletes three exons encoding most of the DNA-binding T-domain and introduces a frame shift that is predicted to lead to a premature stop codon. If any protein is produced from this allele, it would not be capable of binding DNA (Coll et al.,2002). In spite of this similarity with UMS mutations, however,there are differences in the phenotype of the mutation between human and mouse. First, only a minor heterozygous phenotype is observed in the external genitalia of adult mice, whereas human TBX3 mutations appear to be completely penetrant with variable expressivity. A spectrum of abnormalities,predominantly in the forelimb and apocrine glands, but also including abnormal external genitalia such as imperforate hymen, micropenis and imperforate anus,is seen in heterozygous individuals(Bamshad et al., 1999; Bamshad et al., 1996). Second,in human heterozygotes, the forelimb is more frequently and more severely affected than the hindlimb, whereas in homozygous mutant mice, the hindlimb is always the more severely affected limb. Although the phenotype of human TBX3 mutant homozygotes is unknown, it seems likely that the forelimb will be more severely affected because of the heterozygous phenotype.

Several possibilities could explain these species differences: they could be the result of differences in genetic background modifiers, species-specific differences in dose sensitivity between mice and humans, or alternatively,differences in the developmental role of Tbx3 between mouse and human. With respect to genetic background effects, humans are generally considered an outbred population and could be segregating for genetic modifiers, which might explain the variable phenotype seen in different individuals with identical TBX3 mutations. However, in mice with uniform genetic backgrounds (inbred mice and F₁ mice), we have observed variable expressivity. We have also observed a similar variability and range of expressivity in several mixed genetic backgrounds, indicating that phenotypic variability is not due solely to genetic background differences but must also be caused by stochastic events during development. Species differences in dose sensitivity between humans and mice have been observed for a number of other genes [e.g. SHH(Chiang et al., 1996; Ming et al., 1998; Roessler et al., 1996) and PAX9 (Peters et al.,1998; Stockton et al.,2000)], including another T-box gene, TBX1, which produces a severe haploinsufficient phenotype in humans that is reflected in a mild and transient heterozygous phenotype in mice(Jerome and Papaioannou, 2001; Lindsay et al., 2001; Merscher et al., 2001). Finally, although it appears that the human mutation is completely penetrant in the pedigrees studied, it could be that other families harbor mutations that go undetected in the heterozygous condition.

The difference in severity between the forelimb and hindlimb is somewhat harder to explain and could represent functional differences in the developmental role of Tbx3 in limb growth and patterning, or species differences in the expression of Tbx3 as have been observed between chick and mouse (Gibson-Brown et al.,1998a). Tbx2 and Tbx3 are normally expressed in virtually identical patterns in the margins of early limb buds. It is possible that these closely related genes have partially overlapping functions, and that human and mouse have evolved a different balance of function between the two. As yet, no mutations in Tbx2 have been reported in either species to test this idea. Notwithstanding these species differences, the major aspects of the Tbx3 mutant phenotype in limb, genitalia and mammary gland in mice correspond to abnormalities characteristic of UMS, and,in addition, a yolk sac defect not reported in humans has been uncovered.

The mammary gland hypoplasia seen in individuals with UMS is reflected in the deficiency in mammary placode induction in homozygous mutant mice and the absence or reduction of mammary buds in mutant embryos. Although there appears to be no haploinsufficiency effect on mammary development or function in heterozygotes, in Tbx3 homozygous mutant mice, a severe deficiency of mammary development is seen at the earliest stages of mammary bud induction,as indicated by the absence of the two earliest markers of the mammary placode, Wnt10b and Lef1, and lack of morphological evidence of placodes. Normally, five pairs of mammary placodes form between 11 and 12.5 dpc, but in 13 Tbx3 homozygous mutants examined either histologically or with molecular probes between 11.5 and 13.5 dpc, only three buds were detected, all of them histologically: one 13.5 dpc embryo had a single, small mammary bud and a second 13.5 dpc embryo had a pair of buds. These embryos were smaller than their normal littermates, but were at comparable developmental stages, as judged by developmental landmarks such as the development of the pinnae of the external ear.

Studies using targeted gene mutations have begun to yield information on the earliest stages of mammary placode induction, and several signaling pathways have been implicated in its control. The five mammary placodes are induced asynchronously between 11 and 12.5 dpc with placode 3 appearing first and placode 2 being the last to form(Mailleux et al., 2002). In addition, induction of the inguinal bud (placode 4) is different from the others in that it is independent of the FGF10/FGFR2 pathway, although its later maintenance depends on FGFR2 activity. Mice with a null mutation of either Fgf10 or Fgfr2 develop only the inguinal bud, which is maintained in the absence of FGF10 but not in the absence of FGFR2(Mailleux et al., 2002). Induction of this inguinal bud also appears to be independent of WNT signaling, as mice lacking LEF1, an effector of WNT/β-catenin signaling,also develop only a single pair of inguinal mammary buds(van Genderen et al., 1994). Mice that lack Msx2 expression show variable arrest of mammary gland development, but Msx1, Msx2 double homozygotes reveal redundancy for these two transcription factors in early mammary gland induction; placodes form but fail to undergo further development into mammary bud(Satokata et al., 2000),although the status of each individual placode was not reported in this study. The deficiency of mammary placodes in mice lacking Tbx3 argues that Tbx3 is required for mammary bud induction and acts upstream of both the FGF and WNT signaling pathway because all of the placodes, including the FGF- and WNT-signaling independent placode 4, are affected. It may be that Tbx3 is required for mammary epithelium to become competent for continued development, although the rare, late appearance of placode 2 indicates that this requirement is not absolute.

The deficiency in the development of posterior limb elements in individuals with UMS is reflected in limb abnormalities in mutant mice, including deficiency of the forelimb digits and ulna, as well as the foot and fibula,that result from a failure in the development of posterior limb elements. Limb development and patterning depends on reciprocal interactions between the AER and the ZPA, involving complex signaling through the FGF and SHH signaling pathways (Martin, 2001). FGF signaling from the limb mesenchyme is essential for the formation of the AER,which then induces the ZPA in the posterior part of the limb. Shhsignaling from the ZPA is primarily responsible for anteroposterior (AP)patterning of the limb and is regulated by a complex feedback loop with FGFs in the AER (Sun et al., 2000). Tbx3 is normally expressed both in the AER and in the limb mesenchyme at the anterior and posterior margins of the limbs(Chapman et al., 1996b; Gibson-Brown et al., 1996);therefore, it could be involved in the initiation or maintenance of Fgf gene expression in the AER, and/or Shh expression in the ZPA.

The abnormalities in anteroposterior patterning seen in Tbx3homozygous mutant limbs closely resemble the limb phenotype seen in mice with homozygous mutations in Shh, or with a limb-specific reduction of Shh expression (Chiang et al.,2001; Kraus et al.,2001; Lewis et al.,2001), suggesting that Tbx3 normally functions in the SHH pathway, and that loss or reduction of Shh is responsible for the limb phenotype in Tbx3 mutants. In both Tbx3 and Shh mutants, the posterior tissue of the fore- and hindlimbs is reduced and the AER appears weak and disorganized, as shown by FGF8 expression. In addition, only digit one, a Shh-independent digit,forms in the hindlimb of Tbx3 homozygous mutants.

The similarities in the limb abnormalities observed in Tbx3 and Shh mutant embryos combined with the absence or reduction of Shh expression in Tbx3 homozygous mutants suggest a role for Tbx3 in the initiation and/or maintenance of Shh expression. Tbx3 is not absolutely required for Shh expression because a small proportion of Tbx3 homozygous mutant embryos express Shh, as shown by the presence of Shh in one out of four mutants at 11.5 dpc, and by the normally patterned right forelimb in one late stage mutant fetus. We propose that Tbx3 functions with Hand2 during early limb patterning to regulate Shhexpression and set up the initial anteroposterior patterning of the limb. Shh expression is then required to maintain and expand the early anteroposterior pattern, and to maintain Hand2 and Tbx3expression in the posterior limb, as the posterior domain of Tbx3expression has been shown to be Shh dependent(Tumpel et al., 2002).

Alternatively, or in addition to this role in regional signaling in limb patterning, Tbx3 may play a role in maintaining proliferation in the posterior limb margin. Both Tbx3 and the closely related Tbx2 have recently been identified in senescence bypass screens to identify genes that allow the bypass of proliferation arrest(Brummelkamp et al., 2002; Jacobs et al., 2000),implicating these genes in the maintenance of cellular proliferation. A deficiency in the number of posterior limb margin cells in mutants could result in a reduced ZPA, with fewer cells producing SHH. The feedback loop maintaining the AER would consequently be affected. The posterior reduction in Msx1 expression, which is an AER-independent marker of the progress zone, would fit with a reduction in posterior tissue. Whether the lower expression of Tbx2 in the posterior margins of the limbs of mutant mice is due to loss of posterior tissue or an interaction between Tbx2 and Tbx3 has yet to be determined. This decrease in posterior Tbx2 expression could also be a direct result of decreased Shh expression, as it has been shown that Shh can regulate Tbx2 (Gibson-Brown et al.,1998a).

No homozygous TBX3 humans have been reported. Our mouse model provides a tentative explanation for this in that the yolk sac, which is vital for hematopoiesis and maternal-fetal exchange during gestation, is severely compromised in mutant mice. Although the yolk sac vasculature initially forms normally in the majority of mutant embryos and blood is produced, massive cell death of the yolk sac endoderm occurs at a variable time during gestation,even though Tbx3 expression in the yolk sac and other fetal membranes appears to be transitory in the early stages of yolk sac development(Chapman et al., 1996b). In the absence of the detection of other life-threatening abnormalities in fetal membranes or in the fetus, and because some embryos still have a beating heart at the same time as the yolk sac defects are evident, we attribute embryonic lethality to the yolk sac failure, which includes, but is not limited to the vasculature. If there is a similar role for TBX3 in human fetuses,they too would be lost during gestation. At present, we have no explanation for the variable time of death or the variable expressivity of the yolk sac phenotype, and it is not clear what developmental pathways are affected. It is possible that some mutant embryos survive with a threshold level of yolk sac development only to succumb to delayed effects of an earlier deficiency in the metabolic function of the yolk sac or deficiencies in the vasculature, blood islands or hematopoietic precursors. The small size of the liver in mutant embryos might be a reflection of this deficiency.

The putative function of Tbx3 in the p19^ARF/Mdm2/p53 pathway (Brummelkamp et al.,2002; Carlson et al.,2002; Jacobs et al.,2000; Lingbeek et al.,2002) offers a possible explanation for the cell death observed in the yolk sac, as well as the hypoplasia of the mammary bud and posterior limb elements. The failure of Tbx3 to suppress p19^ARFexpression in these areas of Tbx3 embryonic expression could lead to the upregulation of this gene and the subsequent upregulation of p53. This in turn could lead to cessation of proliferation and/or increased apoptosis in the yolk sac and eventual death of the fetus. This possibility is under investigation.

In summary, the homozygous Tbx3 mutant mouse recapitulates two of the major features of UMS, the posterior limb defect and mammary gland hypoplasia, as well as uncovering an additional role for the gene in the yolk sac. Tbx3 mutant mice provide a means of investigating the involvement of this gene in specific signaling pathways during development of the yolk sac, limb and mammary gland.

We thank V. Aita, D. Chapman, T. Bestor and members of the Papaioannou laboratory for helpful discussion; P. Rodriguez and U. Saha for technical assistance; J. Gao for performing the fibula measurements; B. Hogan and J. Gearhart for mouse embryo cDNA libraries; and R. Axel, R. Grosschedl, A. Joyner, R. Maas, G. Martin, A. McMahon, G. Shackelford and D. Srivastava for generously supplying in situ hybridization probes. This work was funded by a grants from the NIH (RO1 HD33082) and the United States Army Medical Research and Materiel Command Breast Cancer Program (DAMD17-96-1-6090).