Defective somite patterning in mouse embryos with reduced levels ofTbx6

One of the defining features of vertebrate embryology is the segmentation of paraxial mesoderm into somites, which lie on both sides of the neural tube. Subsequently, in response to signals from the surrounding tissues, the somites undergo regional differentiation to form the dermomyotome dorsally and the sclerotome ventrally. These will later differentiate to form the dermis,skeletal muscle of the back, body wall and limbs, and the vertebrae and ribs(reviewed by Christ et al.,2000). During their formation, each somite is also divided into rostral and caudal (R-C) halves, such that cells in the rostral half are different molecularly and physically from cells in the caudal half (reviewed by Pourquie and Kusumi, 2001;Saga and Takeda, 2001). These R-C differences are initially required for the establishment of somite boundaries and later for the process of resegmentation. During resegmentation,the sclerotome region of the somites is divided and reorganized such that the caudal region of one somite fuses with the rostral portion of the next more posterior somite to ultimately form the vertebrae and ribs of the adult organism. Resegmentation results in the connection of a skeletal muscle group(which is derived from a single somite) to two consecutive vertebrae, thereby allowing for vertebral movement (Huang et al., 1996).

Notch signaling is essential for establishing and maintaining somite boundaries by setting up differences between R-C halves of somites; this occurs within the anterior presomitic mesoderm (PSM), prior to somite formation. Mutations in various components of the Notch signaling pathway disrupt the process of somite segmentation and the latter process of resegmentation (reviewed by Pourquie and Kusumi, 2001). For example, mutations in the human DLL3,a Notch ligand, results in spondylocostal dysostosis, which is characterized by hemivertebrae, rib fusions and deletions that together lead to short stature in the affected individuals (Bulman et al., 2000). Improper subdivision of the somites into R-C halves results most noticeably in the formation of fused ribs and vertebrae. It may also result in defective patterning of other tissues, in particular the peripheral nervous system (PNS). The neural crest-derived spinal ganglia migrate through the rostral halves of the somites, thus giving the PNS a segmental arrangement, so that disruption of R-C somite patterning disrupts the segmental arrangement of the PNS (reviewed byBronner-Fraser, 2000).

The mouse T-box transcription factor Tbx6 is expressed in the primitive streak and PSM, and is downregulated as the paraxial mesoderm segments to form the somite (Chapman et al., 1996). Gene targeting studies in mice revealed that Tbx6 is necessary for the formation of paraxial mesoderm posterior to the forelimb bud, which is replaced by ectopic neural tubes in the Tbx6-null mutant embryos (Chapman and Papaioannou,1998). Primitive streak markers T, Wnt3a andFgf8 are all expressed in the Tbx6-null mutant, while PSM markers such as Notch1 and paraxis are not, suggesting that differentiation of primitive streak into PSM is blocked(Chapman and Papaioannou, 1998)(D.L.C., unpublished).

The Notch ligand, delta-like 1 (Dll1) is expressed in the primitive streak and PSM, in a domain that overlaps with Tbx6(Bettenhausen et al., 1995;Chapman et al., 1996). Like other genes that are normally expressed in the PSM, Dll1 is not expressed in the Tbx6-null mutant embryos(Chapman and Papaioannou,1998), suggesting that it might be a target of Tbx6. Alternatively, its expression may be lost simply because the PSM is not present in the mutant. Herein, however, we present data that supports Tbx6 functioning through Dll1 in patterning the somites. First, we show that reduction of Tbx6 expression in embryos below heterozygous levels leads to fusions of the ribs and vertebrae, a characteristic shared by Notch signaling mutants. We further show that these malformed vertebrae are due to improper patterning of the somites along their R-C axis. Second, we demonstrate, not only that Tbx6 genetically interacts with the classic mouse mutant rib-vertebrae (rv), but also thatrv is a hypomorphic mutation in Tbx6. Finally, we show thatTbx6 genetically interacts with Dll1, and that this,together with the absence of Dll1 expression in Tbx6-null mutant embryos, suggest that Tbx6 is upstream of Dll1 in the pathway leading to somite formation and patterning.

An 8.8 kb XbaI fragment from the Tbx6 genomic locus was cloned upstream of the lacZ reporter gene, p1229. This fragment contains the entire coding region of the Tbx6 gene, along with 2.3 kb upstream and 2 kb downstream sequences. The reporter plasmid, p1229 (gift of M. Maconochie), consists of the β-globin minimal promoter, thelacZ gene and the polyadenylation sequence from SV40. The insert was excised from the plasmid, purified and injected into FVB/N fertilized eggs according to standard techniques (Hogan et al., 1994). Eggs were then transferred to recipient females and were either dissected at embryonic day (E) 10.5 to E12.5 or allowed to go to term. Two permanent transgenic lines, Tg(Tbx6)46Dlc and Tg(Tbx6)130Dlc (hereafter referred to as Tg46and Tg130), were established with this transgene.

The Tbx6^tm1Pa mice have previously been described and are maintained on a mixed C57Bl6/J 129Sv/Ev genetic background(Chapman and Papaioannou,1998). Dll1^tm1Gos and rib-vertebrae(rv) mutant mice were obtained from Jackson Laboratories and are maintained on a C57Bl6/J background(Beckers et al., 2000b;Hrabe de Angelis et al., 1997;Nacke et al., 2000). The twoTbx6 transgenic lines (Tg46 and Tg130) are maintained on both FVB/N and a mixed genetic background. Genotyping of theTbx6 and Dll1 mutant mice was performed as previously described (Chapman and Papaioannou,1998; Hrabe de Angelis et al.,1997). The rv allele was genotyped by Southern blot analysis using a Tbx6-specific genomic probe that detects a polymorphism between the Tbx6 wild-type and rv mutant alleles. Similarly the Tbx6 transgene was detected using aTbx6-specific probe by Southern blot analysis.

Genomic DNA from rv/rv and wild-type C57Bl6/J mice was used for PCR using the following primers designed from sequences obtained at Ensembl Mouse Genome Server: (forward) 5′-CATTCCCAAAACCCCATTGC -3′,(reverse) 5′-TGCCCCTTCACTCTCTCCATC -3′; and (forward)5′-GTGTAGTTGAAATGTTCTCGGCG -3′, (reverse)5′-CACAGTTCCTGGTTCTCCAAGC -3′. The PCR products were cloned and sequenced as described above.

Whole-mount in situ hybridization was performed as previously described by Wilkinson (Wilkinson, 1992)using antisense riboprobes for Tbx6, Tbx18, Dll1, Dll3. Mesp2, Uncx4.1,Cer1, neurofilament L and myogenin. Hybridization and washes were performed at 63°C.

Embryos were dissected at the indicated times and stained forβ-galactosidase activity with X-gal as previously described(Ciruna et al., 1997).

Skeletons from embryos dissected at E14.5 and E15.5 were stained with Alcian Blue (staining the cartilage) and Alizarin Red (staining the bone) as described by Hogan et al. (Hogan et al.,1994), except that staining was performed simultaneously at 37°C. After staining, embryos were cleared in 1% KOH and stored in 1%KOH:glycerol (90:10).

To understand the genetic pathway that leads to PSM formation, we are using a transgenic approach to identify the cis-acting regulatory regions of Tbx6 and thereby the gene(s) regulating Tbx6 expression. During these studies, we identified a Tbx6 genomic fragment capable of driving the expression of a minimal promoter-lacZ reporter gene in a Tbx6-specific manner (Fig. 1). This Tbx6 transgene contains the entire coding region of Tbx6, as well as upstream and downstream flanking regions(Fig. 1A). Embryos carrying this transgene had Tbx6-specific β-galactosidase staining in the primitive streak and PSM (Fig. 1B-D) along with staining extending into the somites, whereTbx6 transcripts are not normally detected (compareFig. 1D with normalTbx6 expression in Fig. 1E). This somite staining was almost certainly due to the perdurance of the β-galactosidase, as whole-mount in situ hybridization showed that lacZ transcripts were downregulated in the first somite formed (Fig. 1F).

The Tbx6 transgene described above contained the entireTbx6-coding region along with sequences capable of driving reporter gene expression in a Tbx6-specific spatial and temporal pattern. Two transgenic lines, Tg(Tbx6)46Dlc and Tg(Tbx6)130Dlc (hereafter referred to as Tg46and Tg130), were established and used in attempts to rescue theTbx6 mutant phenotype (Fig. 2). Tbx6-null embryos(Tbx6^tm1Pa/Tbx6^tm1Pa) can be distinguished at E9.5 by the lack of somites posterior to the forelimb bud and a bulbous tailbud (Chapman and Papaioannou,1998). Both transgenic lines drove reporter gene expression inTbx6-null embryos – hence in the absence of endogenous Tbx6(Fig. 2B,C and data not shown). A single copy of the Tg46 transgene was able to rescue posterior somite formation in Tbx6 mutant embryos. A noticeable feature of theTbx6^tm1Pa/Tbx6^tm1Pa Tg46/+ mutants was an enlarged tailbud (albeit smaller than the tailbud of a Tbx6 null mutant), which tended to branch at later times in development(Fig. 2B-D). Interestingly,β-galactosidase staining accumulated in the dorsal region of the enlarged tailbud (Fig. 2B,C). Incomplete rescue of the Tbx6 mutant phenotype suggested that the levels ofTbx6 expressed from the transgene might be reduced compared with normal embryos. Examination of Tbx6 expression in theTbx6^tm1Pa/Tbx6^tm1Pa Tg46/+ embryos revealed that although transcripts were present in both the primitive streak and PSM,they were found at lower than heterozygous levels(Fig. 2E), which could account for the incomplete rescue observed for this transgene. To examine the fate of somites formed in Tg46 rescued embryos, litters were dissected fromTbx6^tm1Pa/+ Tg46/+ × Tbx6^tm1Pa/+matings at E15.5. At the level of gross morphology, Tg46 rescued embryos can be distinguished by their short stature and short tails compared with control embryos (Fig. 2F,G). Skeletal preparations revealed severe fusions of the ribs and vertebrae along the axis, which may account for the short stature of the embryos (Fig. 2I). These embryos were still alive at E15.5, indicating that the transgene rescues the early embryonic lethality observed in theTbx6^tm1Pa/Tbx6^tm1Pa embryos, which are normally dead by E12.5. The Tg46 rescued mutants, however, are dead at birth,perhaps due to the failure of the malformed ribs to protect internal organs during birth.

The second transgenic line, Tg130, was much less efficient at restoring somite formation in the Tbx6 null. Approximately half of the Tbx6^tm1Pa/Tbx6^tm1Pa Tg130/+ embryos were similar in phenotype to theTbx6^tm1Pa/Tbx6^tm1Pa embryos and were resorbed by E15.5, while the remainder formed only a few fused ribs and vertebrae and died later in utero as they were never born(Fig. 2K). Tbx6 is expressed at very low levels, if at all, in theTbx6^tm1Pa/Tbx6^tm1Pa Tg130/+ embryos (data not shown). In attempts to increase the level of Tbx6 expression, we also generated Tbx6^tm1Pa/Tbx6^tm1Pa that were homozygous for the Tg130 transgene. TheTbx6^tm1Pa/Tbx6^tm1Pa Tg130/Tg130 embryos displayed a range of phenotypes, some similar in phenotype to theTbx6^tm1Pa/Tbx6^tm1Pa Tg130/+ embryos, forming only a few ribs and vertebrae, while others resembled the Tg46rescued Tbx6 mutants, forming fused vertebrae and ribs and found still born (data not shown). Because the Tg46 transgene is lethal when homozygous, we were unable to determine the effect of increasingTbx6 expression by generatingTbx6^tm1Pa/Tbx6^tm1Pa Tg46/Tg46 embryos. Instead,we increased Tbx6 expression levels by combining both theTg130 and Tg46 transgenes hemizygously in the Tbx6null. Although the Tbx6^tm1Pa/Tbx6^tm1Pa Tg46/+Tg130/+ embryos still showed signs of rib fusions(Fig. 2L,M) they were much less severe than in embryos with only one copy of either transgene (compareFig. 2L,M with 2I,K).

Through our attempts to rescue the Tbx6 mutant using a transgene that expresses Tbx6 at low levels, we have established a series ofTbx6 phenotypes characterized by rib and vertebral fusions, which suggests improper patterning of the somites. To explore this possibility more carefully, we performed whole-mount in situ hybridization using markers of somite R-C patterning on Tg46 rescued embryos and their littermates derived from crosses of Tbx6^tm1Pa/+ Tg46/+ ×Tbx6^tm1Pa/+ mice.

The paired homeobox gene Uncx4.1 is normally expressed in the caudal halves of somites (Mansouri et al.,1997; Neidhardt et al.,1997) (Fig. 3A). In the Tbx6^tm1Pa/Tbx6^tm1Pa Tg46/+ embryo,Uncx4.1 transcripts are detected throughout the somitic region with no apparent stripes (Fig. 3B,C). In some of these embryos, Uncx4.1 appears as darker stripes in the most recently formed somites; however, these stripes are found amidst uniform, low-level Uncx4.1 expression(Fig. 3D). Thus, somites in these partially rescued Tbx6 mutants appear to be caudalized.

To explore further whether rostral somite compartments form in theTg46 rescued embryos, we examined the expression of several markers of rostral somite compartment identity. The bHLH transcription factor,Mesp2, is normally expressed in the anterior PSM, initially in a domain that spans the width of a somite, but rapidly becoming localized to the rostral half of the forming somite (Saga et al., 1997). Mesp2-null embryos lose rostral somite compartment identity and the caudalized paraxial mesoderm fails to segment. Although Mesp2 was not expressed in the Tbx6-null mutant,transcripts were present in the anterior PSM of both control and Tg46rescued embryos (Fig. 3E; data not shown), suggesting that rostral identity is not completely lost in these embryos. In some of the Tg46 rescued embryos, a low level ofMesp2 transcripts was detected in the most recently formed somite, in addition to expression in the anterior PSM. As Mesp2 is required to repress its own expression (Takahashi et al.,2000), this ectopic expression suggests that in at least some of these Tg46 rescued embryos, Mesp2 fails to downregulate its own expression.

Mouse cerberus-related gene 1, Cer1, is expressed as two stripes in the anterior PSM and in the rostral region of the most recently formed somite (Biben et al., 1998;Shawlot et al., 2000)(Fig. 3F). As expected,Cer1 is not expressed in the Tbx6-null mutant (data not shown). Cer1 is expressed in the PSM of Tg46 rescued embryos, but levels are reduced compared with control embryos, and no expression is detected in the newly formed somite(Fig. 3F). This result also suggests that rostral somite compartment identity of the PSM is initially observed; however, it is not maintained in these embryos.

Neural crest-derived spinal ganglia migrate through the rostral half of the sclerotome, therefore loss or improper formation of this somite compartment will lead to alterations in the appearance of the spinal ganglia. To investigate this, we examined the expression of neurofilament L, which is segmentally expressed in the spinal ganglia along the anteroposterior (AP)axis of normal embryos (Lewis and Cowan,1985) (Fig. 3G). The rostral compartment of the somites is clearly disrupted in theTg46 rescued embryos as shown by the expression pattern of neurofilament L. In contrast to the segmented appearance of the neurofilament L-expressing spinal ganglia in the normal embryo, the segments are disorganized and often joined in rescued embryos(Fig. 3H).

The T-box gene Tbx18 is expressed in the rostral region of somites along the AP axis of the embryo, in addition to the heart and limb buds(Kraus et al., 2001)(Fig. 3I,K). A range of expression patterns was observed for this gene in the Tg46 rescued embryos. Some rescued embryos failed to express Tbx18 in the somites(Fig. 3J), while other embryos expressed reduced levels of Tbx18 transcripts in the rostral region of the newly formed somites, but this expression was not maintained in more anterior somites (Fig. 3L). In between these two extremes were embryos that expressed Tbx18 at reduced levels and in smaller domains in the most newly formed somites. However, it was no longer expressed in more anterior regions of the embryo(data not shown). Altogether these marker gene studies show that R-C somite patterning is clearly disrupted in the Tg46 rescued mutant embryos,with somites losing their rostral compartment identity.

The defect in R-C somite patterning in theTbx6^tm1Pa/Tbx6^tm1Pa Tg46/+ embryos suggests that expression of one or more components of the Notch signaling pathway is disrupted in these embryos. Two Notch ligands, Dll1 and Dll3, are known to be involved in R-C somite patterning(Barrantes et al., 1999;Dunwoodie et al., 2002).Dll3 is normally expressed in the primitive streak, PSM and later in the rostral somite compartment of the two most recently formed somites(Dunwoodie et al., 1997). In contrast to most PSM expressed genes, including Dll1, Dll3 is expressed in the Tbx6-null mutant and its expression in the tailbud of the Tg46 rescued embryos is comparable with controls(Fig. 3M). Dll3expression in the rostral regions of the newly formed somites of theTg46 rescued embryos is not apparent; however, expression in this domain was not seen in control embryos either(Fig. 3M).

Dll1 is normally expressed in the primitive streak and PSM, then localizing to the caudal halves of somites(Bettenhausen et al., 1995)(Fig. 3N,Q). AlthoughDll1 is not expressed in the tail mesoderm of the Tbx6-null embryo (Chapman and Papaioannou,1998), Dll1 expression is restored in theTbx6^tm1Pa/Tbx6^tm1Pa Tg46/+ embryos, with transcripts present in the PSM with a slight upregulation in the somite that is next to form (Fig. 3O,P).Dll1 expression levels, however, were reduced in the Tg46rescued embryos compared with Tbx6 heterozygotes (compareFig. 3N,Q with 3O,P,R,S). Similar to β-galactosidase staining, Dll1 transcripts localize to the dorsal region of the expanded tail region and are absent from more ventral areas (Fig. 3R,S). Low levels of Dll1 expression in the caudal halves of somites can be seen in both the Tbx6 heterozygotes and in the partially rescued mutants(Fig. 3N-S). In addition to this mesoderm expression, Dll1 is also expressed in neural tissue(Bettenhausen et al., 1995).Dll1 expression is detected in the flanks of some Tg46rescued embryos, marking the ectopic neural tissue, which still forms in the paraxial region of some of these mutants despite the presence of paraxial somites (Fig. 3P).

We have also examined expression of other Notch signaling pathway genes in the Tbx6 nulls and Tg46 rescued embryos, specificallyNotch1 and lunatic fringe (Lfng). Although neither gene is expressed in the Tbx6-null embryos, the expression of both is restored in the PSM of Tbx6^tm1Pa/Tbx6^tm1PaTg46/+ embryos; however, this expression is reduced compared with normal embryos (data not shown). The dynamic expression of Lfng in the PSM was observed in the Tg46 rescued embryos, but transcripts were present at reduced levels. The reductions in expression of the Notch signaling pathway genes Dll1, Notch1 and Lfng in the Tg46rescued embryos compared with normal littermates could individually or together account for the R-C somitic defects observed.

The phenotype of the partially rescued Tbx6 mutant embryos was reminiscent of the of the rib-vertebrae (rv) mutant phenotype. rv is a spontaneous recessive mutation characterized by fusions of ribs and vertebrae, shortened trunk, kinked tail, formation of a single kidney, and reduced fertility(Beckers et al., 2000b;Nacke et al., 2000;Theiler and Varnum, 1985).rv homozygous embryos are recognized by an enlarged tailbud, which later in development has multiple outgrowths [which are characteristics shared with the Tg46 rescued embryos(Fig. 2D)], and by caudal duplications of the neural tube [which is a characteristic shared with theTbx6-null mouse (Chapman and Papaioannou, 1998; Theiler and Varnum, 1985)]. Interestingly, rv had been mapped to position 62 cM on mouse chromosome 7(Beckers et al., 2000b), which was close to the mapped location of Tbx6 at position 61 cM(Chapman et al., 1996). Altogether, these data suggest that Tbx6 is the gene affected in therv mutation. To determine whether Tbx6 interacts genetically with rv, embryos were dissected from Tbx6^tm1Pa/+× rv/rv crosses at E15.5 and their skeletons were stained with Alcian Blue and Alizarin Red (Fig. 4A-E). At the level of gross morphology, compound heterozygous embryos can be distinguished by their short stature and short tails(Fig. 4B), similar to theTg46 rescued embryos. The genetic interaction between these two mutations is further illustrated by skeletal preparations of these embryos,which clearly show fusions of ribs and vertebrae along the entire AP axis(Fig. 4D,E).Tbx6^tm1Pa/+ rv/+ pups are either resorbed during development or are found dead at birth. Examination of Tbx6expression in rv/rv and rv/+ embryos revealed that althoughTbx6 was expressed in the correct spatial pattern, transcripts were present at lower levels in the tailbuds of rv/rv compared withrv/+ (Fig. 4F). In addition, Tbx6 expression is reduced in the compound mutant embryos compared with Tbx6^tm1Pa/+ (data not shown). These results indicate that Tbx6 expression is directly affected in the rvmutant, and that the genetic interaction observed between the two mutants is probably due to reduced levels of Tbx6 expression.

The strong genetic interaction between Tbx6 and rv led us to identify the molecular lesion responsible for the rv mutation,reasoning that it was a mutation in Tbx6. Genomic DNA prepared from both wild-type and rv/rv mice was analyzed by Southern blot hybridization using a number of probes derived from the Tbx6 locus. Polymorphisms were detected between wild-type and rv/rv genomic DNA(data not shown), indicating that Tbx6 is the affected gene in therv mutation. The region of Tbx6 affected in the rvmutation is upstream of the initiating methionine within 1.2 kb of the start of transcription. PCR was used to amplify the affected region and sequencing revealed that the rv mutation is caused by a 185 bp insertion of a duplicated region of the Tbx6 gene(Fig. 3G). This duplication,which includes a region of the first exon, the entire first intron and a region of the second exon, is inverted with respect to the endogenous gene and inserted into the presumed enhancer region of Tbx6. The location of the rv lesion, together with the reduced Tbx6 expression in tailbuds of rv/rv versus rv/+ embryos, suggests that this mutation somehow affects the ability of this enhancer element to function properly, and therefore rv is a hypomorphic allele ofTbx6.

The defect in somite formation in rv/rv mice is due to disruption of R-C patterning of the somites and this ultimately leads to the fusions of ribs and vertebrae (Beckers et al.,2000b; Nacke et al.,2000). rv genetically interacts with Dll1(Beckers et al., 2000b;Nacke et al., 2000). As we have identified rv as an allele of Tbx6, the targeted mutant allele of Tbx6 (Tbx6^tm1Pa) should also interact genetically with Dll1. We confirmed this by generatingTbx6^tm1Pa/+ Dll1^tm1Gos/+ embryos and mice. Examination of the Tbx6^tm1Pa/+ Dll1^tm1Gos/+embryonic skeletons revealed a variety of vertebral defects, including fusions of the neural arches in cervical vertebrae, missing regions of ribs and abnormally formed vertebrae along the AP axis(Fig. 4I-K).Tbx6^tm1Pa/+ Dll1^tm1Gos/+ mice are born and can be distinguished by their kinked tails (not shown). The phenotypes of these compound heterozygotes were much milder than those seen for either theTg46 rescued embryos or the Tbx6^tm1Pa/rvmutant embryos described above.

Whether reduced Tbx6 expression in the embryo is achieved by transgene rescue of the Tbx6-null or by a bona fide Tbx6hypomorphic mutation, patterning of the myotome compartment of the somite is also clearly disrupted. Marker gene analysis of theTbx6^tm1Pa/Tbx6^tm1Pa Tg46/+ andTbx6^tm1Pa/rv embryos revealed irregular patterning of the myotome in these mutant embryos (Fig. 5). Myogenin is normally expressed in distinct stripes along the AP axis, marking the myotome (Cheng et al., 1992; Sassoon et al.,1989; Wright et al.,1989) (Fig. 5A,C). In both the Tbx6^tm1Pa/Tbx6^tm1Pa Tg46/+ and theTbx6^tm1Pa/rv embryos, myogenin expression no longer has an evenly spaced striped pattern, but instead has areas of fused adjacent myotomes, as well as isolated pools of myogenin-expressing cells along the AP axis (Fig. 5B,D). Defects in these Tbx6 hypomorphs therefore affect multiple compartments of the somite.

The process of somitogenesis in mice can be temporally divided into two phases: first the specification of paraxial mesoderm from cells within the primitive streak; and, second, the division of paraxial mesoderm into somites and the generation of R-C patterning within them. Previous studies have shown that Tbx6 is required for the specification of posterior paraxial mesoderm,while other studies have shown that the correct spatial and temporal expression and activity of Notch signaling pathway components is required for somite formation and R-C patterning.

In Tbx6-null mutant embryos, somites posterior to the forelimb bud fail to form, and mesodermal tissue in the tails of these mutants appears to be arrested at the primitive streak stage of differentiation. It was not clear from these studies, however, if Tbx6 played any role in patterning of the somites or if it was simply required for establishing the paraxial mesoderm fate. We show that embryos with Tbx6 expression levels intermediate between null homozygotes and heterozygotes develop somites but that these are incorrectly patterned along their R-C axis, which leads to abnormalities in somite-derived tissue, most notably the axial skeleton. These embryos were generated in two ways: first, by using a rescuing transgene containing theTbx6-coding region and cis-regulatory elements in a Tbx6null mutant; and, second, by identifying a previously known mutation,rv, as a hypomorphic mutation in Tbx6.

Expression studies in embryos with reduced Tbx6 expression reveal that the R-C division is clearly disrupted. In both the rv/rv andTbx6^tm1Pa/Tbx6^tm1Pa Tg46/+ embryos, the caudal somite compartment marker Uncx4.1 is expressed in a wider somite domain compared with normal embryos, suggesting that somites in these embryos have adopted a caudal phenotype (Fig. 3A-D) (Beckers et al.,2000b). The degree of caudalization is correlated with the severity of the phenotypes and hence with the level of Tbx6expression. The low level of Tbx6 expression observed in theTg46 rescued embryos results in almost complete caudalization of the somites (as assessed by Uncx4.1 expression) and to severe fusions of the ribs and vertebrae along the entire AP axis of the embryo. Inrv/rv embryos, intermediate levels of Tbx6 (between theTg46 rescued embryos and Tbx6 heterozygotes) result in somites with wider than normal domains of Uncx4.1 expression;however, unstained regions in these somites were also observed(Beckers et al., 2000b). Consequently, the rib and vertebral fusions were not as severe in therv/rv embryos compared with Tg46 rescued embryos(Fig. 2I)(Beckers et al., 2000b). Although somites in the Tg46 rescued embryos appear caudalized,marker genes specific for the rostral compartment of the forming somite, such as Mesp2, Cer1 and Tbx18, are still present. Mesp2marks the rostral half of the forming somite and plays a role in specifying rostral somite compartment identity, as paraxial mesoderm in theMesp2 homozygous mutant embryos becomes caudalized(Saga et al., 1997).Mesp2 is still expressed in Tg46 rescued embryos and suggests that rostral somite identity is initially established in these embryos; however, it is eventually lost, based on Uncx4.1 expression. This result is supported by the expression of another rostral somite marker,Cer1. Cer1 is expressed in the anterior PSM of the Tg46rescued embryos, but is expressed at low levels and is not expressed in a third stripe in the most recently formed somite, again supporting an eventual loss of rostral compartment identity. In this regard, our data forCer1 expression in the Tbx6^tm1Pa/Tbx6^tm1PaTg46/+ embryos differs from that previously reported for therv/rv mutant embryos, where Cer1 expression was shown to be absent (Beckers et al., 2000b). It is unclear why this difference was observed between the rv/rv andTbx6^tm1Pa/Tbx6^tm1Pa Tg46/+ embryos, especially as the rostral marker Mesp2 was expressed in the rv/rvembryos, indicating that the rostral compartment is formed. Further analysis of rostral somite compartment identity revealed that Tbx18, which is normally expressed in the rostral halves of somites along the AP axis of the embryo, is either not expressed or is expressed at lower levels in the most recently formed somites, and is lost in more anterior somites of theTg46 rescued embryos. The irregular R-C somite patterning is further exemplified by neurofilament L expression in the neural crest-derived spinal ganglia, which normally appear as discrete segments along the AP axis because of their migration through the rostral halves of somites. In our Tg46rescued embryos, neurofilament L expression appears disorganized with segments fusing.

Segmentation of the PSM is believed to be dependent on these R-C differences (Stern and Keynes,1987). Clear somite boundaries appear to be established in embryos with reduced levels of Tbx6, suggesting that some differences in R-C compartments are initially established in these embryos. These R-C differences, however, are not sufficiently maintained, as demonstrated byTbx18 and neurofilament L expression patterns. The failure to establish firmly and/or maintain R-C differences lead to later defects in resegmentation, which can account for the vertebral defects observed. Although the myotome is not directly dependent on R-C patterning of the somite, the myotome is affected in the Tbx6 hypomorphs. Our results suggest therefore that the failure to maintain these R-C differences will indeed affect other compartments of the somite, perhaps because of a failure to maintain somite boundaries.

The disruption in somite patterning observed in embryos with reduced levels of Tbx6 is similar to that of Notch signaling component mutants,suggesting that the expression or activity of one or more genes involved in this pathway is disrupted in the presomitic mesoderm of these embryos. The Notch ligand Dll1 is an obvious candidate, because expression is completely lost in the Tbx6-null embryo(Chapman and Papaioannou,1998). Dll1 expression is restored in Tg46rescued embryos, but the level of Dll1 transcripts is clearly reduced compared with that in Tbx6 heterozygotes. Furthermore, previous studies with rv and our studies here with the Tbx6-null allele have revealed a strong genetic interaction between Dll1 andTbx6: Tbx6^tm1Pa/+ Dll1^tm1Gos/+ embryos have defects in rib and vertebral patterning, although this phenotype is much weaker than rv/rv or Tbx6^tm1Pa/Tbx6^tm1PaTg46/+. Thus, one likely cause for the defective somite patterning in embryos with reduced levels of Tbx6 is a reduction in Dll1expression. As demonstrated by gene targeting experiments, Dll1 is required for proper R-C patterning of the somites and for epithelialization of the somites (Hrabe de Angelis et al.,1997).

Dll3 expression in embryos with reduced Tbx6 expression is indistinguishable from wild type, indicating that there is not a simple reduction in expression of all presomitic-specific genes in these embryos. Consistent with this, no genetic interaction between Dll3 andrv was detected (Beckers et al.,2000b). Notch1 and Lfng, which are also required for somite formation and patterning, are expressed at reduced levels in the PSM of these embryos, but in the correct spatial patterns. The reduced expression levels of these genes might contribute to the observed phenotypes. However, no genetic interaction between Notch1 and rv was detected (Beckers et al.,2000b), again implicating Dll1 as the primary factor disrupted in these embryos.

One problem with directly linking Tbx6 and Dll1 is that the mutants appear to have opposite phenotypes in terms of their effects on compartmentalization of the somites. In Dll1-null mutants, somites appear to be rostralized (Barrantes et al.,1999), in contrast to their caudalization in embryos with reduced levels of Tbx6. At present, the reason for this is not clear. It is likely that Tbx6 is regulating the expression of other genes besidesDll1, as would be expected based on the differences in theTbx6 and Dll1 null phenotypes. Other likely candidates areNotch1 and Lfng, the expression of which was restored in embryos expressing lowered levels of Tbx6 compared with their lack of expression in the homozygous null. The combined effects of lowered Dll1,Notch1 and Lfng may account for the differences seen in compartment phenotypes. A complete explanation awaits the identification of these additional Tbx6 targets. Another explanation is that reduction rather than complete loss of Dll1 may result in caudalization rather than rostralization: this can only be tested directly with Dll1hypomorphs.

Tbx6 is a transcription factor and thus could be directly activatingDll1. The regulatory elements necessary for Dll1 expression have been identified through transgenic studies(Beckers et al., 2000a). We have examined these regulatory elements for T-box-binding sites using the binding site identified for T (Kispert and Herrmann, 1993); however, no palindromic or half sites were identified. As the binding site for Tbx6 has not yet been identified, further analysis along these lines awaits the identification of a Tbx6-specific binding site.

Mouse embryos that bear homozygous Tbx6-null mutations result in a lethal phenotype by embryonic day 12.5(Chapman and Papaioannou,1998). It is likely that similar mutations in humans would also lead to early embryonic lethal phenotypes that would probably appear as spontaneous abortions. Mice heterozygous for the Tbx6 mutation are normal and fertile, with no signs of skeletal or muscle abnormalities,suggesting that humans heterozygous for Tbx6-null mutations would presumably be normal. This paper identifies phenotypes that lie between these two extremes, with hypomorphic levels of Tbx6 expression resulting in R-C patterning defects in the somites. Thus far, no human syndromes that would result from improper R-C somite patterning map to human TBX6 on chromosome 16. However, several syndromes like Klippel-Feil and spondylocostal and spondylothoracic dysostosis, which display a number of clinical defects including rib and vertebral fusions, may in the future be linked to hypomorphic mutations in TBX6 or to genes functioning in this pathway.

We thank S. Dunwoodie, Y. Saga, A. Kispert, M. Maconochie, R. Behringer, A. Gossler, Anita Myer and W. Klein for gifts of plasmids, and Jeffrey Hildebrand and Gerard Campbell for helpful discussions. This work was supported by a grant from the NIH (HD38786).