Hoxc8 initiates an ectopic mammary program by regulating Fgf10 and Tbx3 expression and Wnt/β-catenin signaling

During embryonic development, the epidermis and underlying dermis of vertebrate skin collaborate via respective epithelial and mesenchymal signals to create cutaneous appendages, such as hair and feather follicles, mammary glands, teeth and sweat glands. Despite the morphological and functional differences among mature skin organs, each begins as a placode, a raised epithelial thickening that initiates in response to a broadly expressed Wnt signal from the dermis (Mikkola, 2007). As mammary and hair placodes begin to develop and invade the mesenchyme, dermal and epidermal Wnt signaling continues, along with additional signaling molecules such as fibroblast growth factors (FGFs), bone morphogenetic proteins (BMPs), ectodysplasin (Eda-A1) and sonic hedgehog (Shh) to effect specific development of each organ type (Andl et al., 2002; Chu et al., 2004; Gallego et al., 2002; Mikkola and Millar, 2006; Mustonen et al., 2004; Petiot et al., 2003; Plikus et al., 2004; St-Jacques et al., 1998; Veltmaat, 2013; Zhang et al., 2009).

In mice, the first visible sign of mammary development is the appearance at E10.5 of two histologically distinct lines of pseudostratified ectoderm between the forelimb and hindlimb buds, marked by Wnt10b expression (Veltmaat et al., 2004). This ectoderm is a permissive region for mammary rudiments (MRs) 2, 3 and 4, joining additional streaks of mammary permissive ectoderm in the axial and inguinal regions giving rise to MRs 1 and 5 (Veltmaat et al., 2004). Ectopic mammary glands occur most commonly at inappropriate sites along these lines. Evidence in rabbits and mice suggests that mammary placodes form by migration of epithelial cells into and along the mammary lines, resulting in the five pairs of MRs developing non-sequentially at characteristic positions along the body axis (Lee et al., 2011; Propper, 1978). Molecular requirements differ among the pairs of mammary placodes, and differential gene expression profiles may underlie some of the heterogeneous attributes and susceptibilities to tumor incidence in adult mammary glands (Veltmaat et al., 2013).

Proper positioning of the mammary line along the dorsoventral axis is achieved in part by mutual antagonism between ventrally expressed Bmp4 and the more dorsally expressed T-box transcription factor Tbx3 (Cho et al., 2006; Veltmaat et al., 2006). These and additional mammary factors, such as Gli3, retinoic acid, Nrg3 and Fgf10, play important roles in the appropriate patterning of Wnt signals that are required to achieve the proper rostrocaudal positioning of placodes (Cho et al., 2012; Cowin and Wysolmerski, 2010; Hatsell and Cowin, 2006; Howard, 2008; Lee et al., 2013; Mailleux et al., 2002; Veltmaat et al., 2006).

The idea that a ‘HOX code’ (Kessel and Gruss, 1991) might underlie the regional distribution of cutaneous appendages has been around since the discovery that Hox gene expression exhibits positional variation within the skin itself (Bieberich et al., 1991; Chuong, 1993). The majority of Hox genes appear to be expressed in fetal or adult skin and hair follicles (Awgulewitsch, 2003; Johansson and Headon, 2014), and several Hox genes are expressed in developing and mature mammary glands, or become dysregulated during mammary neoplasia (Chen and Sukumar, 2003; Hayashida et al., 2010; Lewis, 2000; Wu et al., 2006). During early embryogenesis, expression of vertebrate Hox genes initiates in a rostral to caudal direction along the body axis in a sequence mirroring the linear position of each gene within the four chromosomal Hox clusters, a unique feature termed ‘spatiotemporal colinearity’. Each cell along the Hox trajectory receives a distinct combination of Hox proteins, a HOX code (Kessel and Gruss, 1991) that may uniquely specify its position, patterning individual elements from the hindbrain to the most posterior vertebrae. However, unlike axial Hox expression, only a subset of the tested Hox genes have been shown to exhibit regional restriction of expression within mouse, human or chick embryonic skin, including Hoxc8, Hoxb4, Hoxa7, Hoxd9, Hoxd11 and Hoxd13 (Kanzler et al., 1994; Reid and Gaunt, 2002). Several others, including Hoxc13, which has a crucial role in hair shaft development, are expressed broadly throughout the epidermis and/or dermis (Godwin and Capecchi, 1998; Kanzler et al., 1994; Reid and Gaunt, 2002). Therefore, Hox gene temporal and spatial expression in embryonic skin does not strictly match the colinear expression found in axial Hox domains, and the putative HOX code responsible for globally defining domains of emerging epidermal organs has proved elusive, a likely consequence of the complex combinatorial nature of Hox expression and function.

The strongest evidence for Hox-mediated regional patterning of epidermal organs comes from two thoracic Hox genes. Adult thoracic mammary glands of mice lacking functional Hoxc6 are devoid of mammary epithelium, whereas inguinal mammary glands develop ductal structures and are less severely affected (Garcia-Gasca and Spyropoulos, 2000). Hoxc8 has been indirectly implicated in the specification of feather and hair types (Kanzler et al., 1997; Mentzer et al., 2008) and, in mice, Hoxc8 shows regionally restricted expression during the first wave of hair placodogenesis, the earliest reported expression of any Hox gene in the epidermis (Johansson and Headon, 2014; Kanzler et al., 1994).

Using a Hoxc8IresCre mouse line (Chen et al., 2010), we found Hoxc8 lineage in mammary line ectoderm by E10.75 and that it was incorporated into all five MRs by E12.5. This result prompted us to carefully re-examine Hoxc8 expression in embryonic skin in order to assess the potential of this Hox gene to mediate early skin regionalization and skin appendage specification. Further analysis demonstrated transient regionally specific expression of Hoxc8 protein in the ectoderm during mammary line formation, prior to the earliest reported Hoxc8 ectodermal expression. We tested the possibility that Hoxc8 expression plays a role in mammary line specification using mice carrying a targeted allele designed to conditionally express Hoxc8. Conditional misexpression of Hoxc8 using two out of three Cre drivers consistently led to the appearance of supernumerary MRs within two distinct domains: along the normal mammary line of mutant mice, and within the cervical region anterior to the first MR. These ectopic rudiments express the placode markers Bmp2, Wnt10b and Tbx3 and are labeled by the mammary mesenchyme-specific markers ERα and androgen receptor.

This study is the first to implicate a Hox gene in rostrocaudal positioning of mammary line ectoderm and placodes. We present evidence that Hoxc8 positively regulates Tbx3 and Fgf10 expression and Wnt/β-catenin signaling and, moreover, that Tbx3 is a direct Hoxc8 transcriptional target. These data further support the existence of a HOX code underlying regional specification of embryonic skin at the earliest stages of skin placode initiation.

Hoxc8 is cited as one of the first Hox genes expressed in embryonic mouse skin, with its earliest reported expression in E14.5 epidermis (Awgulewitsch, 2003; Johansson and Headon, 2014; Kanzler et al., 1994). Lineage analysis and Hoxc8 antibody were both employed to re-examine cutaneous Hoxc8 expression to determine if it is appropriately staged to play a role in the early specification of mammary glands. Lineage was examined using a Hoxc8IresCre mouse line (Chen et al., 2010) and either RosaYFP or RosalacZ (C8cre/YFP and C8cre/lacZ) (Soriano, 1999). The C8cre/YFP lineage at E12.5 is broadly represented throughout flank caudal to the forelimb bud (n=3). The reporter additionally labels all MRs (Fig. 1A, MR1 is obscured by the forelimb), revealing that Hoxc8 protein is also expressed in early surface ectoderm and may be present in ectodermal precursors giving rise to mammary epithelium. Hoxc8 antibody did not label E13.5 mammary bud ectoderm (Fig. 1B; data not shown; n=3), indicating that ectodermal Hoxc8 is transitory, preceding the mammary bud stage.

To pinpoint the timing and extent of transient Hoxc8 expression, we next examined the Hoxc8 lineage in sectioned C8cre/lacZ embryos (in which all Hoxc8-expressing cells and descendants are labeled), and subsequently examined transient expression in wild-type embryo sections labeled with Hoxc8 antibody (Fig. 1C-F; data not shown; two or three embryos were examined at each time point indicated). The lacZ reporter extends rostrally in E10.75 ventrolateral ectoderm to the forelimb level encompassing the region of the developing mammary line between forelimb and hindlimb (Fig. 1C). At a slightly earlier stage, Hoxc8 antibody labels all surface ectoderm extending between and including the forelimb and hindlimb buds of E10.5 embryos, which necessarily includes the rostrocaudal extent of the forming mammary line (Fig. 1D; data not shown). By E11.5 and E12.5, ectodermal expression of Hoxc8 is considerably reduced, particularly in the epithelium of the forming mammary placodes and buds (Fig. 1E,F). In situ hybridization of wild-type embryos with a Hoxc8 probe (Fig. 1G; n=2) shows expression in E11.5 hypaxial extensions of thoracic somites, which include S15 and S16, underlying the portion of the mammary line specifically giving rise to MR3 (Veltmaat et al., 2006).

The paraxial and surface ectodermal expression of Hoxc8 make it an ideal candidate for a potential role in mediating mammary line and third placode specification. We shifted the domains of Hoxc8 paraxial, mesodermal, and ectodermal expression using the Hoxa3IresCre mouse (Macatee et al., 2003) to test if Hoxc8 misexpression could alter mammary line and placode positioning. The E10.5 A3cre/lacZ lineage shows widespread expression throughout lateral mesoderm and somites (the rostral expression limit of somitic Hoxa3 corresponds to the first cervical vertebrae) and much of the body ectoderm caudal to the second branchial arch (Fig. 2A; n=3). A3cre/CAGC8 embryos express ectopic Hoxc8 wherever the IresEGFP signal is present in a pattern equivalent to the A3cre/lacZ lineage. This fluorescent signal was subsequently used to genotype A3cre/CAGC8 mutants (Fig. 2B).

In contrast to the wild-type pattern of 13 thoracic ribs seen in control embryos, mutant A3cre/CAGC8 embryos exhibited well-formed ectopic ribs on all cervical, lumbar and sacral vertebrae as well as rib-like extensions on several caudal vertebrae (Fig. 2C,D; n=6). This phenotype is 100% penetrant in A3cre/CAGC8 mutants and is consistent with previous studies demonstrating a fundamental role of Hox genes in assigning anteroposterior vertebral identity (Carapuço et al., 2005; Le Mouellic et al., 1992; McIntyre et al., 2007; van den Akker et al., 2001; Wellik and Capecchi, 2003). The cervical region appears elongated in A3cre/CAGC8 mutants, which is likely to be related to its transformation to a thoracic identity. Interestingly, we found that the rib phenotype is dependent on Hoxc8 dosage, as a nearly identical construct minus the potent CAGGS promoter (also targeted to the Rosa locus, driven only by the Rosa promoter) yielded only mild skeletal phenotypes, consistent with a previously reported Hoxc8 transgenic mutant (Pollock et al., 1992) (Fig. S1; n=16).

The dermomyotome, marked by Myf5 expression, exhibited ectopic hypaxial extensions within the cervical and lumbar regions of E11.5 A3cre/CAGC8 embryos (Fig. 2E,F; n=3). We used Bmp2 as a general placode marker, as its expression focally marks the epithelium of the mammary bud (Phippard et al., 1996), hair (Suzuki et al., 2009) vibrissae, tooth (Bitgood and McMahon, 1995), and tongue papillae (Jung et al., 1999). At E11.5, we found that A3cre/CAGC8 embryos exhibited strong upregulation of Bmp2. Focal Bmp2 expression was seen within irregularly spaced placodes in the cervical ectoderm, with dark streaks of Bmp2 signal along the normal mammary line (Fig. 2G,H; n=3). Control embryos exhibited only faint streaks of ectodermal Bmp2 expression along the mammary line, punctuated by focal upregulation within MR3, the first MR to form. We next examined expression of three of the earliest known genes associated with mammary line/placode formation: Fgf10, Tbx3 and Wnt10b (Fig. 2I-R; n≥3 per genotype per time point for each probe). Formation of placodes 1, 2, 3 and 5 requires Fgf10 expression, which emanates from thoracic somites. Homozygous ablation of Fgf10, or of its ectodermal receptor Fgfr2b, results in the absence of all four placodes (Mailleux et al., 2002; Veltmaat et al., 2006). Fgf10 signal appears upregulated in cervical and thoracic somites of A3cre/CAGC8 mutant embryos at E10.5 compared with control littermates (Fig. 2I,J, Fig. S2A,B). Other domains of Fgf10 expression, including limb buds, appear unchanged in the mutants.

Tbx3 is required for the formation of mammary buds 1, 3, 4 and 5 (Davenport et al., 2003; Eblaghie et al., 2004; Veltmaat, 2013). In humans, heterozygous mutation of TBX3 causes ulnar-mammary syndrome, characterized by upper limb, genital, mammary and other glandular defects (Bamshad et al., 1999). In both control and A3cre/CAGC8 E10.5 mouse embryos, Tbx3 is expressed in a broad strip of lateral plate mesoderm underlying the mammary line, and in another broad region of lateral mesoderm extending caudally from the fourth pharyngeal arch to the forelimb (Fig. 2K,L). By E11.5, Tbx3 levels in control embryos have greatly decreased in the cervical lateral mesoderm, but remain high in the A3cre/CAGC8 mutant, as a continuum extending from the hindlimb through thoracic and cervical levels (Fig. 2M,N, Fig. S2C,D). Focal Tbx3 upregulation in mutant cervical placodes is obvious by this stage. However, in the mammary line ectoderm, mutant Tbx3 signal persists as a streak, whereas placode formation in controls is nearly complete (Fig. 2M,N). Notably, ectopic cervical placodes are not restricted to a linear pattern, in contrast to supernumerary placodes developing within the mammary line. At E13.5 strong Tbx3 expression is confined to the ten mammary buds in controls, whereas mutants (with 100% penetrance) exhibit strong Tbx3 expression in both normally positioned mammary buds and supernumerary buds along the mammary line and in the ectopic cervical zone. We generally found one or two Tbx3-expressing ectopic buds occurring within each E13.5 mammary line, and as many as 12 ectopic cervical buds in a single A3cre/CAGC8 embryo (Fig. 2O,P). Ectodermal Wnt10b (Fig. 2Q,R, Fig. S2E,F) shows a pattern of dysregulation remarkably similar to that of Tbx3 and Bmp2 in E11.5 mutant embryos. Because A3cre/CAGC8 embryos die at or around E14.5, later stages of mutant mammary development cannot be examined without orthotopic transplantation.

We next induced Hoxc8 misexpression with Pax3cre (Engleka et al., 2005) in order to determine whether ectopic somitic Hoxc8 is sufficient to generate an anterior mammary permissive zone. In E11.5 Pax3cre/lacZ embryos, β-gal marks the dorsal neural tube and strongly labels somites and hypaxial dermomyotome (Fig. 3A,B; n=2). Importantly, lacZ is not expressed in body surface ectoderm (Fig. 3B, inset), allowing us to test the competence of Hoxc8 in establishing an ectopic mammary permissive zone via its expression within somitic derivatives only. Like A3cre/CAGC8 embryos, Pax3cre/CAGC8 embryos died at or around E14.5. By E14.0, well-formed ribs were established on all cervical vertebrae and the first three lumbar vertebrae of Pax3cre/CAGC8 embryos, whereas embryos carrying only the Pax3^cre allele produced the wild-type rib formula (Fig. 3C,D; n=6). Transformation of cervical vertebrae to a thoracic identity was accompanied by upregulation of the mammary factors Fgf10 and Tbx3 in hypaxial extensions of Pax3cre/CAGC8 cervical somites (Fig. 3E-H; n=3 per genotype per probe). However, in contrast to A3cre/CAGC8 mutants, neither Tbx3 nor Bmp2 (n=2) expression showed focal upregulation in the cervical ectoderm of Pax3cre/CAGC8 mutants (Fig. 3G-J).

Wnt/β-catenin signaling was examined in E13.5 Pax3cre/CAGC8 mice using the TOPgal reporter (DasGupta and Fuchs, 1999). Consistent with the absence of Tbx3 and Bmp2 expression in cervical placodes, no focal spots of high TOPgal expression were found anterior to MR1 in Pax3cre/CAGC8 mutants at E13.5 (Fig. 3K,L; n=7), suggesting the absence of ectopic cervical MR formation in these mutants. These results indicate that although ectopic Hoxc8 within somites is sufficient to expand thoracic vertebral identity into the cervical region, it is insufficient to expand the zone of permissive mammary-forming ectoderm. On the other hand, four out of seven mutant embryos (57%) formed a single unilateral supernumerary mammary bud between MR3 and MR4 in addition to all normally positioned MRs (Fig. 3K,L). This location overlies the hypaxial domain of endogenous Hoxc8 expression (Fig. 1G) and suggests that an increased level of somitic Hoxc8 can promote mammary placode development as long as it underlies a region of ectoderm that is competent for mammary formation.

To test whether simultaneous Hoxc8 expression in ectoderm and somitic derivatives is sufficient to recapitulate the anterior zone of cervical mammary placodes found in A3cre/CAGC8 mutants, we misexpressed Hoxc8 using a Wnt6cre driver (N. Makki, PhD Thesis, University of Utah, 2010). Wnt6 is initially expressed in a broad band of ectoderm encompassing the mammary-forming region, and becomes restricted to the developing mammary placodes (Veltmaat et al., 2004). Analysis of E10.5 W6cre/lacZ embryos showed Wnt6 lineage extending across most of the surface ectoderm prior to mammary line formation (Fig. 4A and inset; n=3), although the Wnt6 lineage was considerably weaker within lateral mesoderm compared with the Hoxa3 lineage. Dermomyotomal expression of Wnt6 is restricted to the dorsomedial lip (Ikeya and Takada, 1998). Consequently, hypaxial signal (which derives from dorsolateral dermomyotome) was not detectable in Wnt6 lineage of W6cre/lacZ control embryos (Fig. 4A), or in Wnt6 lineage of W6cre/CAGC8 mutants (as visualized by the IresEGFP reporter; Fig. 4B), indicating that ectopic Hoxc8 is restricted to non-hypaxial somite in this conditional cross.

Unlike A3cre/CAGC8 and Pax3cre/CAGC8 embryos, W6cre/CAGC8 embryos survive until birth, but die perinatally. We found rudimentary or fully formed ectopic ribs on one or two cervical vertebrae of W6cre/CAGC8 mutants (Fig. 4C,D; n=6), suggesting that hypaxial Hoxc8 expression is not required for transformation of cervical somites towards a thoracic identity. Somitic Hoxc8 expression was accompanied by upregulation of Fgf10 expression in cervical somites by E10.5 (Fig. 4E,F; n=4).

Expression patterns and levels of Tbx3 were nearly equivalent between control and W6cre/CAGC8 littermates at E10.5 (Fig. 4G,H; n=4), similar to A3cre/CAGC8 embryos and littermates at this stage. By E11.5, Tbx3 expression was focally upregulated in cervical ectoderm of Wnt6cre/CAGC8 mutants with 100% penetrance, indicating the formation of ectopic placodes (Fig. 4I-L; n=4). Whereas Tbx3 expression is aberrantly maintained in the cervical mesoderm of A3cre/CAGC8 mutant embryos at E11.5 (Fig. 2K,L), Tbx3 expression in W6cre/CAGC8 cervical mesoderm is similar to that in controls at this stage (Fig. 4G-J). Within the wild-type mammary line, Tbx3 expression in control embryos was confined to placodes by E11.5, whereas W6cre/CAGC8 embryos exhibited lingering expression along the mammary line, and considerably broader, more diffuse expression within the forming placodes themselves (Fig. 4I,J). By E13.5, Tbx3 expression was focally restricted to mammary buds in both controls and mutants (Fig. 4M,N; n=4). W6cre/CAGC8 mutants had fewer ectopic buds at E13.5 than A3cre/CAGC8 mutants, with an average of one ectopic bud in each mammary line and up to eight extra buds anterior to MR1.

At E13.5, Tbx3 antibody labeled all mammary bud epithelium (including supernumerary MRs), the surrounding mammary mesenchyme, and scattered cells in the underlying mesoderm (Fig. 5A,B; n=4). In serial sections of the same E13.5 embryos, Hoxc8 antibody labeled mutant epidermis and mammary primordia, as well as scattered cells in the underlying mesoderm (Fig. 1A, Fig. 5C,D; n=4), but was only present in scattered mesoderm of controls. We verified the mammary identity of ectopic cervical placodes in E13.5 W6cre/CAGC8 embryos by labeling mammary mesenchyme with antibodies for estrogen receptor alpha (ERα; Esr1 – Mouse Genome Informatics) (n=2) and androgen receptor (n=3). All supernumerary as well as normally positioned mammary placodes expressed both markers in control and mutant embryos (Fig. 5E-H; data not shown). We found that Tbx3 expression was aberrantly upregulated in the epithelium of E13.5 W6cre/CAGC8 vibrissae at E13.5 relative to controls (Fig. 4M,N; Fig. S3A,B), leading us to speculate that vibrissal placodes might be adopting a mammary fate. However, neither ERα nor androgen receptor antibody labeled mesenchyme of the vibrissal placodes of control or W6cre/CAGC8 embryos (Fig. S3C,D; data not shown; n=3) indicating that, although vibrissal structures are incorrectly specified, ectopic Hoxc8 and consequent Tbx3 misexpression does not direct vibrissal differentiation towards a mammary program.

As Wnt signaling is essential for the initiation and subsequent development of all ectodermal organs (Boras-Granic and Hamel, 2013; Chu et al., 2004; Lim and Nusse, 2013), we performed a detailed timeline of TOPgal reporter expression at different embryonic stages to study changes in Wnt signaling that accompany Hoxc8 dysregulation. W6cre/CAGC8 embryos survive until birth, enabling us to perform TOPgal analysis during mammary ductal elongation and branching. In E10.5 embryos, TOPgal reporter expression was consistently expanded in W6cre/CAGC8 cervical lateral mesoderm relative to control littermates (Fig. 6A,B; n=6), with expression extending rostrally to the fourth branchial arch. Limb and mammary line ectoderm showed similar faint staining between mutants and controls (Fig. 6A,B). Interestingly, the cervical pattern of Wnt/β-catenin signaling at E10.5 completely overlapped with Tbx3 expression at the same embryonic stage (Fig. 4G,H). At E11.5, the cervical TOPgal signal became more localized to the neck-forelimb junction. Vibratome sections through this region show considerably stronger signal in the mutant ectoderm and mesoderm compared with controls (Fig. 6C,D; n=4). Placode patterning along the wild-type mammary lines of E11.5 control embryos appeared as focal aggregations of β-gal-positive cells. However, Wnt signaling in the mutant showed ostensibly delayed aggregation of β-gal-positive cells into mammary placodes. (Fig. 6C,D). Placode aggregation of Wnt10b-expressing cells in W6cre/CAGC8 embryos at E11.5 paralleled the delay seen with TOPgal expression (Fig. S4; n=2).

At E12.5, both mutant and control embryos displayed diffuse patches of β-gal-positive mesoderm caudal to the ear and in the cervical region. However, in the mutant, these diffusely stained patches contained focal spots stained dark blue, representing ectopic mammary primordia (Fig. 6E,F; n=6), which resolve by E13.5 into three or four ectopic mammary buds anterior to MR1 on both sides (Fig. 6G-J; n=8). At E13.5, these ectopic buds often protruded abnormally from the ectoderm (Fig. 6H, top inset), but by E15.5 have invaginated into the underlying dermis, as do MRs along the mammary line (Fig. 6K,L,Q,R; n=3). Approximately 60% of E12.5 and E13.5 W6cre/CAGC8C8 mutant mice bore one or two small supernumerary mammary buds located between MR3 and MR4 (Fig. 6E,F). Normally positioned mammary buds of E12.5 and E13.5 W6cre/CAGC8 mutants always appeared larger than those of control littermates (as opposed to the ectopic buds, which were usually smaller than endogenous buds), and by E13.5 MR3 was situated proximally and fused to MR2 in nearly half of all mutant embryos (Fig. 6H). Supernumerary, fused, and normally positioned mammary buds all expressed the downstream Wnt transcription factor Lef1 (n=3) in mammary epithelium (Fig. 6G,H insets; data not shown).

By E15.5, Wnt/β-catenin signaling in controls is downregulated in the neck and surface epithelium overlying the growing mammary sprout. However, mammary sprouts of female W6cre/CAGC8 embryos often failed to downregulate Wnt/β-catenin signaling properly, particularly in the proximal part of MRs 2 and 3 that had fused or developed in close proximity (Fig. 6K,L). In preparations of E17.5 whole skin stained for β-gal, MRs of control embryos could be seen growing into the underlying secondary mammary mesenchyme (mammary fat pad) and branching into a primitive ductal tree. The ductal systems of all examined W6cre/CAGC8 MRs failed to develop substantially beyond the mammary sprout stage (Fig. 6M,N; n=3), and appeared to degenerate by birth. However, both control and mutant females were born with external nipples at anteroposterior positions that corresponded to embryonic MR formation (Fig. 6O,P; n=3). In newborn controls, small mammary trees were associated with nipples, whereas mutant nipples were not associated with secondary mammary mesenchyme and lacked underlying ductal branches. Interestingly, nipples of mutant newborns maintained epithelial TOPgal signal and often had hair follicles associated with nipple epithelium (Fig. 6O,P). Moreover, both female and male W6cre/CAGC8 mutants maintained mammary sprout development after E14.5 within the cervical region (Fig. 6Q,R; n=3 each for females and males). By contrast, all MRs along the mammary line underwent regression at E14.5 in mutant males (n=3), as is normal in wild-type male embryos (data not shown). Pelage hair placode and follicle morphology was normal in W6cre/CAGC8 mutant embryos (our unpublished observations). However, mutant vibrissal and whisker morphology was defective and TOPgal expression was aberrantly upregulated in vibrissal placodes by E12.5 (Fig. 6E-H,S,T).

To determine the requirement of ectodermal Tbx3 in the formation of the anterior ectopic mammary zone, we misexpressed Hoxc8 while simultaneously ablating Tbx3 expression in the Wnt6 domain. Wnt6 lineage is strongly expressed in ectoderm, but is excluded from hypaxial dermomyotome, with restricted expression in lateral mesoderm prior to mammary placode formation (Fig. 4A,B). Consequently, all embryos with conditional ablation of one or both Tbx3 alleles maintained strong Tbx3 expression in hypaxial and lateral mesoderm, whereas no evidence of ectodermal Tbx3 expression was found in E11.5 W6cre/Tbx3^Δ/Δ or W6cre/CAGC8/Tbx3^Δ/Δ embryos (Fig. 7A-D; n≥4 for each genotype). One functional copy of ectodermal Tbx3 was sufficient to induce MR formation in W6cre/Tbx3^Δ/+ and W6cre/CAGC8/Tbx3^Δ/+ embryos, with cervically localized rudiments forming (with 100% penetrance) in the latter (Fig. 7C,D). Wnt/β-catenin signaling was examined in W6cre/CAGC8 E13.5 embryos with one or both Tbx3 alleles conditionally deleted. TOPgal expression was present in cervical mammary buds of all W6cre/CAGC8/Tbx3^Δ/+/TOPgal embryos (Fig. 7E,F; n=7), but was lost in all W6cre/CAGC8/Tbx3^Δ/Δ/TOPgal embryos (Fig. 7G; n=6). This establishes an ectoderm-specific requirement for Tbx3 for mammary potentiation along the entire anteroposterior axis. In contrast to MRs, facial vibrissae and whisker pads were maintained in the absence of ectodermal Tbx3 (Fig. 7E-G), affirming the divergent developmental trajectories of these two ectodermal appendages in response to ectopic Hoxc8.

We performed chromatin immunoprecipitation (ChIP) on control E11.5 embryos to determine whether Hoxc8 is capable of direct transcriptional regulation of the Tbx3 promoter. A single primer set, located 1.5 kb 5′ of the Tbx3 ATG start codon, amplified Hoxc8-bound chromatin from both dorsal tissue (somites and neural tube) and ventrolateral thoracic tissue (mammary line ectoderm and mesoderm), but failed to amplify IgG-immunoprecipitated chromatin from either (supplementary Materials and Methods, Fig. S5). The experiment was successfully repeated on equivalent tissues derived from E11.5 W6cre/CAGC8 embryos, suggesting that Tbx3 might be directly regulated by Hoxc8 during mammary development of both control and mutant animals.

We used a gene targeting approach to respecify thoracic identity in somites and surface ectoderm in order to test the competence of Hoxc8-regulated factors to potentiate mammary ectoderm and initiate placode formation. Expansion of the thoracic boundary was apparent in all of our mutant crosses by the appearance of ectopic ribs on cervical and other vertebrae. Accompanying the expansion of thoracic identity, we observed the consistent upregulation of somitic Fgf10 by ectopic Hoxc8 in all three of our conditional mutant crosses. Fgf10 expression, emanating from central (and possibly hypaxial) somites is one of the earliest crucial regulators of mammary line initiation. Interestingly, a recent genome-wide association study found that variance in teat number in pigs was significantly associated with quantitative trait loci containing genes involved in vertebral development and possibly back length (Duijvesteijn et al., 2014). This raises the possibility that a Hoxc8-induced transformation of cervical into thoracic vertebrae creates a new signaling zone for mammary gland induction. However, ectopic Hoxc8 expression in Pax3cre/CAGC8 mutants produced completely penetrant cervical ribs without accompanying placode development in the cervical region. As these mutants often developed an additional MR along the normal mammary line, this suggests that somitic Hoxc8 misexpression (even without accompanying ectodermal expression) produces alterations in signaling gradients that can be interpreted by an ectoderm that is already specified for mammary development, but that somitic factors expressed during the thoracic patterning program cannot independently potentiate mammary ectoderm in the cervical region.

Regulation of Wnt signaling by Hox factors has not been widely reported. However, as studies continue to uncover downstream targets of Hox genes, it is becoming apparent that this regulatory role of Hox genes has been overlooked. For example, using ChIP-seq, Donaldson et al. (2012) identified regions of the genome bound by Hoxa2 in the context of second branchial arch development. Of the thousands of genes identified, the majority fell within the Gene Ontology (GO) category of ‘Wnt receptor signaling’. W6cre/CAGC8 mutants show mammary phenotypes with striking similarities to those of mice carrying targeted mutations of the Wnt pathway modulators Sostdc1 (Wise or Ectodin) and Lrp4 (Ahn et al., 2013; Närhi et al., 2012). Shared features between W6cre/CAGC8, Sostdc1^−/− and Lrp4 null mice are: delayed downregulation of Wnt signaling within mammary epithelium; larger diameter of mammary placodes; supernumerary embryonic MRs developing along the mammary line (however, Sostdc1^−/− mutant nipples only appear at puberty); and increased proximity and occasional fusion of mammary buds 2 and 3, which often protrude abnormally from the ectoderm. In addition, loss of Sostdc1 results in ectopic hair follicles developing within nipple tissue (Närhi et al., 2012). Ablation of the mesodermal mammary factor Gli3 in mice also shows notable similarities to the W6cre/CAGC8 mammary phenotype, which is likely to be due to dysregulation of Wnt signaling and of the crosstalk between the Shh and Wnt pathways during mammary development (Hatsell and Cowin, 2006). Deletion of Gli3 causes inappropriate encroachment of hair follicles close to MR2, which itself protrudes abnormally from Gli3 mutant ectoderm, similar to cervical mammary placodes in W6cre/CAGC8 mutants. Interestingly, deletion of Gli3 also prevents the normal regression of mammary buds in male mice, comparable to the persistence of male MRs within the anterior ectopic zone of W6cre/CAGC8 mice (Chandramouli et al., 2013; Hatsell and Cowin, 2006; Lee et al., 2011, 2013; Ulloa et al., 2007).

Supernumerary mammary placode development has previously been reported in experimental mice, but always within or near the wild-type mammary line, never in unique regions (Ahn et al., 2013; Chu et al., 2004; Howard et al., 2005; Mustonen et al., 2003). This difference between our and previous mouse models of mammary induction underscores the ability of ectopic Hoxc8 to initiate a mammary program, thus altering mesenchymal-ectodermal communication at the earliest stages of mammary line potentiation. Hoxc8 misexpression in both somites and the overlying ectoderm enables the generation of a novel mammary zone followed by appropriate specification and early development of MRs up to the stage of ductal tree formation.

A simplified model describing the role of ectopic Hoxc8 in the potentiation of cervical mammary ectoderm is as follows (Fig. S5). Somitic Hoxc8 expression upregulates somitic Fgf10 expression, which is a likely early requirement for potentiation of the cervical mammary zone, similar to the requirement for somitic Fgf10 for potentiation of the mammary line. However, somitic Hoxc8 expression alone is insufficient to induce cervical placode development (as evidenced by the lack of cervical MRs in the Pax3cre/CAGC8 phenotype), indicating an additional requirement for Hoxc8 expression in the overlying ectoderm (as evidenced by robust cervical MR development in the W6cre/CAGC8 phenotype). Ectodermal Hoxc8 expression triggers upregulation of ectodermal Tbx3 expression, possibly via direct transcriptional activation, but only in specific regions where Tbx3 expression and Wnt signaling co-occur in the underlying mesoderm. Ectodermal Tbx3 expression maintains Wnt signaling crucial for ectodermal mammary potentiation. Within this ectopic cervical zone, Tbx3-expressing cells migrate towards placode positions based on somitic signaling gradients of Fgf10 and levels of Wnt activators.

This model is consistent with the absence of mammary programs initiated in other regions of mutant ectoderm that express Hoxc8 but lack either or both Wnt signaling and Tbx3 in the underlying mesoderm. This model also predicts other regions of the W6cre/CAGC8 embryo in which ectodermal Tbx3 upregulation was observed in association with Wnt signaling and Hoxc8 expression, such as the whisker placodes, outer ear epidermis and eyelid conjunctiva (our unpublished observations). These structures all lie in regions beyond somitic signaling gradients and, although they were all defective, none exhibited evidence of MR development.

Endogenous Hoxc8 expression in E10.5 surface ectoderm is consistent with a scenario in which Hoxc8 helps coordinate the induction of mammary line ectoderm, but must be downregulated as epithelial cells migrate into proper position with respect to signaling gradients of Fgf10 expression, and to Tbx3, Gli3 and other modulating factors that fine-tune the levels of Wnt signaling. Following mammary line potentiation, somitic Hoxc8 is well positioned to regulate specification of the third mammary placode, which develops in ectoderm directly overlying hypaxial extensions of somites 15 and 16.

Neither the Hoxc8 knockout nor Hox8 paralog knockout mice have reported mammary defects (Le Mouellic et al., 1992; van den Akker et al., 2001), although skin appendages were not specifically investigated in these mutants. Nevertheless, the scarcity of skin and skin accessory organ phenotypes exhibited by Hox deletion mutants is likely to be due to genetic compensation (Rossi et al., 2015), particularly functional rescue by other members of the Hox family of transcription factors, many of which exhibit overlapping expression patterns in fetal and adult skin (Boucherat et al., 2013; Chen and Capecchi, 1999; Wellik and Capecchi, 2003). For this reason, the complementary approach of conditional misexpression/overexpression can be essential to unraveling developmental mechanisms that involve complex transcriptional programs and signaling pathways mediated by Hox genes and members of other large gene family networks. We suspect that myriad combinations of Hox transcription factors involved in placode patterning might provide the source of copious regional flexibility of cutaneous accessory organs that we see within and across taxa. However, it remains to be determined whether other Hox genes exhibit early transient activation within relevant ectodermal domains.

CAGC8 founders, as well as Cre driver lines Hoxc8IresCre (C8cre) (Chen et al., 2010), Hoxa3IresCre (A3cre) (Macatee et al., 2003), Wnt6IresCre (W6cre) (N. Makki, PhD Thesis, University of Utah, 2010), Pax3cre (Engleka et al., 2005), the reporter lines RosalacZ, RosaYFP (Soriano, 1999) and TOPgal (DasGupta and Fuchs, 1999), and a Tbx3^flox conditional knockout line (Frank et al., 2013) were maintained on C57BL/6 or C57BL/6×CD1 genomic backgrounds. Mice and embryos were genotyped by PCR. To create experimental and control embryos for analysis, A3cre, Pax3cre or Wnt6cre males were bred to CAGC8 females to create control and double-heterozygote mutant littermates. Mutant and controls were easily distinguished by GFP signal using a fluorescent lamp. To generate control and W6cre/CAGC8 embryos with one or two conditionally ablated copies of Tbx3 in the Wnt6 domain, CAGC8/+;Tbx3^flox/+ or CAGC8/+;Tbx3^flox/flox dams were bred to W6cre/+;Tbx3^{Δ /+};TOPgal/+ males. These males were morphologically and reproductively indistinguishable from littermates carrying no Cre allele. For further information on the construction and targeting of the conditional Hoxc8 misexpression allele see the supplementary Materials and Methods. All mouse experiments were approved by the Institutional Animal Care and Use Committee of the University of Utah.

Embryos were fixed at 4°C in 4% paraformaldehyde in PBS for 1-5 days (for paraffin embedding), or up to 24 h (for OCT embedding). Primary antibodies used for this study were: Hoxc8 (1:200; Covance, MMS-286R), Tbx3 (1:200; a generous gift from A. Moon, University of Utah), Lef1 (1:1000; Cell Signaling, 2230), AR (1:200; Millipore, 06-680) and ERα (1:200; Santa Cruz Biotechnology, sc-7207). For further details, see the supplementary Materials and Methods.

Detection of β-gal reporter expression (RosalacZ or TOPgal) was performed as described in the supplementary Materials and Methods.

In situ hybridization was performed according to published protocols (Boulet and Capecchi, 1996). For details, see the supplementary Materials and Methods.

The ChIP procedure that we employed was a modification of a Jove video protocol developed for E8.5 embryos (Cho et al., 2011). For details of the protocol, see the supplementary Materials and Methods and Table S1.

We thank Sheila Barnett, Carol Lenz, Karl Lustig and the animal care staff of the Comparative Medicine Center at the University of Utah for technical assistance; Anne Moon for providing the Tbx3 conditional mutant mouse; Anne Boulet and Matthew Hockin for critical reading of earlier versions of this manuscript; and Amir Pozner, Uchenna Emechebe and members of the M.R.C. lab for insightful discussion.