Discussion
Several mutations and genes have been identified that affect the pattern of hair follicle development, but Tbx15 is the only gene of which we are aware that affects the pattern of hair pigmentation in different body regions. Ventral areas that normally produce yellow hair in the trunk, limbs, and craniofacial regions are expanded in deH/deH mice and, in the trunk at least, represent inappropriate dorsal expression of an Agouti mRNA isoform that is normally restricted to ventral skin. The deH allele is caused by a large deletion that removes most of the Tbx15 coding sequence, but the pleiotropic phenotype is caused by a simple loss of function for Tbx15 rather than a dominant-negative or contiguous gene effect. In particular, there is no heterozygous phenotype, no other genes lie within or close to the deletion breakpoints, and the expression pattern of Tbx15 is consistent with the spectrum of phenotypic abnormalities in both the original de allele and the deH allele. Finally, a Tbx15 targeted allele has the same phenotype as deH.
Our results suggest that patterned expression of Tbx15 provides an instructional cue required to establish the future identity of dorsal dermis with regard to pigmentary and hair length patterning. The ventral edge of Tbx15 expression in the developing flank does not correspond to a known lineage compartment, but, like limb development, occurs within lateral plate mesoderm. These findings represent a novel role for T-box gene action in embryonic development and provide evidence for a previously unappreciated complexity to acquisition of dorsoventral positional identity in mammalian skin.

Distinct Morphologic Regions Represent the Sum of Different Gradients
The visual boundary between dorsal and ventral skin in at/at mice is reminiscent of other systems in which adjacent compartments enforce a binary choice between alternative patterns of gene expression and cell fate (reviewed in Dahmann and Basler 1999). However, Agouti mRNA in both embryonic and postnatal skin is distributed along a gradient whose dorsal boundary is indistinct and overlaps with two additional gradients recognized by their effects on hair length and histochemical staining for melanocytes. The three gradients are close but not congruent, and it is their proximity that gives rise to the superficial distinction between dorsal and ventral skin of at/at mice. Indeed, slight differences between the regions of transition for pigment-type switching and pigment content give rise to a subtle yellow stripe along the flank (see Figures 1, 2, and 9A). Levels of Agouti mRNA remain high throughout the entire ventrum, but hair pigment content is reduced, giving rise to a cream-colored region in the ventrum that, depending on age and genetic backgrounds, may appear more or less distinct from the yellow flank stripe.
Figure 9  Model for Acquisition of Dorsoventral Patterning in the Trunk and the Role of Tbx15
(A) A tricolor pigmentation pattern is generated by the combination of distinct mechanisms that affect distribution of Agouti mRNA and histochemical staining for melanocytes; effects of the latter mechanism by itself are evident in ae/ae mice (see Figure 1). In at/at mice, reduced hair melanocyte activity and high levels of Agouti mRNA in the ventrum lead to a cream color; as melanocyte activity gradually increases towards the dorsum, a lateral stripe is apparent on the flank. The distributions of Agouti mRNA and histochemical staining for melanocytes are both affected by Tbx15 and are externally evident by a widening of the lateral stripe and an increased proportion of total skin occupied by the cream-colored area.
(B) The lateral yellow stripe in at/at mice lies at the same level as the limb dorsoventral boundary. As described in the text, we propose that distinct dorsoventral compartments in ectoderm of the trunk provide an instructional cue to the mesoderm, leading to expression of Tbx15 in dorsal trunk mesenchyme and acquisition of dorsal dermis character. In the absence of Tbx15, dorsal mesenchyme assumes ventral characteristics instead. Loss of Tbx15 affects dorsoventral transitions of hair length, pigment content, and expression of the ventral-specific Agouti isoform; however, the former two effects are subtle and contribute little, if at all, to the abnormal pigmentation of adult deH/deH mice. Thus, despite the abnormal pattern of dark skin in neonatal deH/deH mice (e.g., Figure 2D), the most obvious feature in adults is dorsal displacement of the “boundary” between black and yellow hair (Figure 9A).

Genetics of Tbx15
Named for the presence of a DNA-binding domain first identified in the mouse Brachyury gene (haploinsufficiency causes a short tail), T box–containing genes have been identified as developmental regulators in a wide spectrum of tissues and multicellular organisms (reviewed in Papaioannou 2001). The Tbx15 subfamily, which also includes Tbx18 and Tbx22, is likely to have arisen during early chordate evolution since there is a single gene in amphioxus but no obvious homolog in the fly genome (Ruvinsky et al. 2000). Consistent with this relationship, the three genes are expressed in partially overlapping patterns that include anterior somites (Tbx18 and Tbx22), limb mesenchyme (Tbx15 and Tbx18), and craniofacial mesenchyme (all three genes, Tbx15 more broadly than Tbx18 or Tbx22) (Agulnik et al. 1998; Kraus et al. 2001; Braybrook et al. 2002; Bush et al. 2002; Herr et al. 2003). These observations suggest that an ancestral gene for Tbx15, Tbx18, and Tbx22 may have been important for craniofacial development in cephalochordates, with acquisition of additional expression patterns and developmental functions in the limb and the trunk during early vertebrate evolution. Expression of Tbx18 and Tbx22 has not been reported in embryonic flank mesenchyme, which suggests that Tbx15 is the only family member involved in establishing the dorsoventral identity of the trunk. However, it would not be surprising to find some degree of functional redundancy in animals mutated for two or three of the subfamily members in other body regions, particularly the limbs and the head. For example, mutations in Tbx22 cause the human syndrome X-linked cleft palate and ankyloglossia (Braybrook et al. 2001). Despite high levels of Tbx22 expression in periocular embryonic mesenchyme (Braybrook et al. 2002; Bush et al. 2002; Herr et al. 2003), the condition does not affect the eye, perhaps because residual activity is provided by Tbx15 in the same region.
In an initial description of the expression and map location of mouse Tbx15, Agulnik et al. (1998) suggested human Tbx15 that lies on Chromosome 1p11.1 as a candidate for acromegaloid facial appearance (AFA) syndrome, for which there is a weak positive LOD score to Chromosome 1p (Hughes et al. 1985). Originally described as a rare autosomal-dominant syndrome with progressive facial coarsening, overgrowth of the intraoral mucosa, and large, doughy hands, more recent case reports describe macrosomia, macrocephaly, or both and generalized hypertrichosis with progressive coarsening (Dallapiccola et al. 1992; Irvine et al. 1996; da Silva et al. 1998; Zelante et al. 2000). The deH phenotype exhibits little overlap with these features; instead, we suggest a more likely candidate for mutations of human TBX15 would be frontofacionasal syndrome, an unmapped autosomal recessive condition characterized by brachycephaly, blepharophimosis, and midface hypoplasia (Reardon et al. 1994).
Two of us (S. Kuijper and F. Meijlink) became interested in the deH mutation because of its effects on skeletal development (Curry 1959) and the possibility that the aristaless-related gene Alx3 might be allelic with droopy ear (ten Berge et al. 1998). In spite of similarities between skeletal phenotypes of deH and Alx3 or Alx4 mutants, subsequent experiments (unpublished data) excluded allelism of Alx3 and deH, and a full description of the Tbx15 skeletal phenotype will be published elsewhere.

Developmental Mechanism of Tbx15 Expression and Action in the Skin
Our attention to the role of Tbx15 in pigment patterning was motivated by the effects of Agouti in postnatal animals. However, Agouti is also expressed in the embryo, where it provides a convenient marker of ventral dermis identity. Because an expanded domain of embryonic Agouti expression in deH/deH animals is detectable by E14.5, the effects of Tbx15 on dorsoventral patterning must occur prior to this time. Among other T-box genes whose developmental actions are at least partially understood, two general themes have emerged, one focused on the ability to specify alternative fates for an undifferentiated group of precursor cells and another focused on the ability to support proliferative expansion of a cell population whose fate is already determined (reviewed in Tada and Smith 2001). Either mechanism may apply to the apparent dorsal-to-ventral transformation in deH/deH mice. For example, while the expanded domain of Agouti expression in postnatal deH/deH animals can be traced to events that occur between E11.5 and E13.5, the underlying cause may be that embryonic cells in dorsolateral mesenchyme acquire a ventral rather than dorsal identity or that those cells fail to proliferate normally, followed by compensatory expansion of ventral cells. Cell lineage studies should provide a definitive answer, but we favor the latter hypothesis, because measurements of dorsoventral regions according to hair color in deH/deH mice revealed a small increase of the cream-colored ventral region in addition to the approximate doubling of the yellow flank region (see Figure 2).
In embryonic mesenchyme, expression of Tbx15 and Agouti are complementary, and it is possible that Tbx15 acts directly to inhibit Agouti transcription in dorsolateral mesenchyme. However, the ability of Tbx15 to suppress expression of the ventral-specific Agouti isoform in postnatal mice is likely to be indirect, since postnatal expression of Tbx15 occurs broadly along the dorsoventral axis and overlaps extensively with that of Agouti. In either case, the targets of Tbx15 action in the skin include genes in addition to Agouti, since hair length and melanocyte distribution exhibit a demonstrable, albeit subtle, alteration in animals that carry a null Agouti allele. One potential target is Alx4, which, like Agouti, is expressed in ventral embryonic mesenchyme, and, when mutated, affects hair-follicle as well as limb and craniofacial development (Qu et al. 1997, 1998; Wu et al. 2000; Wuyts et al. 2000; Mavrogiannis et al. 2001). However, expression of ventral markers such as Alx4, as well as Alx3 and Msx2, appears to be unaffected at E11.5 in deH/deH embryos (data not shown).

Differences and Similarities to Dorsoventral Limb Patterning
Loss of Tbx15 also affects regional distribution of hair color in the limbs, with areas that would normally produce black hair giving rise to yellow hair instead. However, neither normal patterns of pigment-type synthesis in the limb nor their disruption in deH/deH mice correspond to obvious developmental compartments. Furthermore, losses of function for En1 or Wnt7a, which cause a partial transformation of the distal limb from dorsum to ventrum (Loomis et al. 1996) or ventrum to dorsum (Parr and McMahon 1995), respectively, have no effect on regional patterns of Agouti expression or distribution of hair-color regions (Y. Chen, unpublished data). (Ectopic pigmentation of the ventral footpads that develops in En1 mutant mice is unrelated to pigment-type synthesis and instead likely reflects a requirement for En1, independent of Wnt7a, to repress migration or proliferation (or both) of pigment cells in ventral epidermis [Cygan et al. 1997; Loomis et al. 1998].)
These considerations notwithstanding, control of dorsoventral trunk pattern by Tbx15 shares certain features with control of dorsoventral limb patterning by Lmx1b, a LIM domain transcription factor that acts downstream of Wnt7a and En1 (Riddle et al. 1995; Vogel et al. 1995; Cygan et al. 1997; Logan et al. 1997; Loomis et al. 1998; Chen and Johnson 2002). Both Tbx15 and Lmx1b act autonomously in mesenchymal cells to promote a dorsal identity, yet have expression domains that do not correspond to cell lineage compartments in the flank (Tbx15) or the limb (Lmx1b) (Altabef et al. 1997; Michaud et al. 1997). In the case of Lmx1b, its expression in the distal limb depends on Wnt7a produced in the overlying dorsal ectoderm (Riddle et al. 1995; Cygan et al. 1997; Loomis et al. 1998). Wnt7a, in turn, is restricted to dorsal ectoderm by En1 in the ventral ectoderm (Loomis et al. 1996; Cygan et al. 1997; Logan et al. 1997), whose expression marks a lineage boundary coincident with the dorsoventral midline of the apical ectodermal ridge (Altabef et al. 1997; Michaud et al. 1997; Kimmel et al. 2000). As described above, En1 or Wnt7a mutations have not been reported to affect patterns of hair-color distribution (C. Loomis, personal communication; Parr and McMahon 1995; Loomis et al. 1996). However, the essential theme that ectodermal lineage compartments control the fate of underlying mesenchyme in developing limbs may apply to the trunk as well as the limb. The mammary glands also develop at a stereotyped dorsoventral position and depend on epithelial–mesenchymal interactions. However, the number and apparent position of the mammary glands are normal in deH/deH animals, indicating the existence of additional mechanisms that control dorsoventral patterning in the trunk as well as in the limbs.
These ideas are summarized in the model shown in Figure 9B. We speculate that a diffusible signal from dorsal trunk ectoderm, at or prior to E11.5, promotes expression of Tbx15 in dorsal trunk mesenchyme, which then establishes dorsal positional identity of those cells as manifested by differences in Agouti expression, pigment-cell development, and hair growth. Because the ventral limit of Tbx15 expression corresponds to the dorsal limit of En1 expression and because the normal position of the pigmentation boundary lies approximately in register with the limb-bud outgrowths, we depict the position of a putative dorsoventral boundary in trunk ectoderm as coincident with the limb dorsoventral boundary. This model is consistent with studies in the chick, where distinct dorsal and ventral lineage compartments exist for ectoderm in both the limb (Altabef et al. 1997, 2000; Michaud et al. 1997; Kimmel et al. 2000) and interlimb regions (Altabef et al. 1997, 2000), but not for limb mesoderm (Altabef et al. 1997; Michaud et al. 1997). In fact, the same mechanism that determines dorsoventral position of the limbs and the apical ectodermal ridge may also act on expression of Tbx15 in the trunk, since ectopic limbs induced in the interlimb region by application of FGF beads develop along a single line that is coincident with normal limb buds (and the future pigmentation boundary) (Cohn et al. 1995; Crossley et al. 1996; Vogel et al. 1996; Altabef et al. 1997, 2000).
Our model predicts that ectopic expression of Tbx15 in ventral mesenchyme should give rise to a dorsalized pigmentation phenotype and could be tested with gain-of-function approaches. However, Tbx15 expression is very dynamic and is restricted to dorsal mesoderm only from E11.5 to E13.5. It is possible that Tbx15 influences skin patterning in a very narrow window of development; alternatively, establishment of dorsal identity by Tbx15 may require another as-yet-unidentified factor that is only present in the mesenchyme underlying dorsal ectoderm.

Pigmentation Patterns and Tbx15 in Other Mammals
The lateral somitic frontier, defined as the lineage boundary between somite-derived versus lateral plate-derived mesoderm, is established during somitogenesis early in development (Mauger 1972; Christ et al. 1983; Olivera-Martinez et al. 2000; Nowicki et al. 2003), but remains distinct in postnatal animals despite the potential for extensive cell mixing (see Figure 8). However, our transplantation and fate-mapping studies demonstrate that the lateral somitic frontier lies dorsal to the pigmentation boundary and does not obviously correlate with a difference in skin morphology. An additional dorsoventral domain that is not externally apparent has emerged from studies of Msx1, whose expression marks a subgroup of somite-derived mesenchymal cells that contribute to dermis in a narrow stripe along the paraspinal region (Houzelstein et al. 2000). Thus, there exist at least three distinct boundaries in postnatal mammalian skin that are parallel to the sagittal plane, marked by differences in pigment-type synthesis, differences in cell lineage, and differences in expression of Msx1.
In rodents, only the pigmentation boundary is evident externally, but many mammals have more complicated patterns of hair type, length, and/or color that vary along the dorsoventral axis. Raccoons, squirrels, skunks, and many different ungulates exhibit lateral stripes whose developmental origins have not been investigated, but may correspond to the lateral somitic frontier, the paraspinal Msx1 compartment, or an interaction between these domains.
The effect of Tbx15 on pigmentation in laboratory mice is reminiscent of coat-color patterns in both selected and natural populations of other mammals. Saddle markings are common in some dog breeds, such as German shepherds, and in certain populations of Peromyscus polionotus, in which a dorsal extension of ventral depigmentation provides an adaptive advantage to subspecies that live on white sand reefs (Blair 1951; Kaufman 1974; Belk and Smith 1996). Neither German shepherds nor deer mice have craniofacial characteristics similar to the deH mutation, but the pigmentation patterns in these animals could represent alterations in the regulation or action of Tbx15 activity. From the opposite perspective, the effects of Tbx15 on coat color are only apparent in certain genetic backgrounds and may not be evident at all in mammals that lack dorsoventral pigmentation patterns. Studying the sequence and expression of Tbx15 in other vertebrates may provide additional insight into patterns that affect the skeleton as well as the pigmentary system.