Introduction

Embryonic surface epithelium appears first as a single layered ectodermal sheet covered by a periderm that is shed towards the end of embryogenesis. While much of the surface ectoderm stratifies to become the epidermis with its protective, cornified outer layer, subsets of surface ectodermal cells undergo complex reciprocal interactions with underlying mesenchymal cells to form a variety of appendages, including hair follicles and teeth as well as mammary glands (Fig. 1). While timing and site of induction for these various appendageal structures differ, early steps in their formation exhibit remarkable similarities at the morphogenetic and molecular levels. In this review, we will compare the cellular and signaling events that shape ectodermal appendages. We will attempt to unravel both common themes and unique aspects in the morphogenesis of diverse appendages that have implications not only for understanding appendage development, but also for elucidating common disease mechanisms and designing strategies for regeneration of certain appendages, such as hair follicles and teeth, in cases of congenital absence or loss through disease.

Figure 1
figure 1

Schematic depiction of the development of stratified epidermis, mammary gland, hair follicle and tooth from embryonic skin.

Full size image

Overview of Embryonic Mammary Gland Development

Mammary gland development is confined to two stripes of ventral lateral surface ectoderm, known as the mammary lines, that appear in response to mesenchymal signals [1, 2]. In embryos of some species, such as human and rabbit, mammary lines can be visualized by scanning electron micrography (SEM) as ridges of raised ectoderm. These are not apparent in SEM of mouse embryos; however, they can be identified histologically as lines of thickened surface ectoderm that appear at approximately embryonic day (E) 10.5 and overlie condensed mesenchymal cells [3]. Ectodermal thickening appears to be the result of pseudostratification of the surface ectoderm [3]. Within the next 24 h, the mammary line resolves into discrete lens-shaped thickenings, known as placodes, at sites of future mammary gland development [2] (Fig. 2a). Scanning electron microscopy of rabbit embryos reveals apparently motile cells atop the mammary lines, suggesting that mammary placodes may arise from migration and accumulation of motile cells at defined locations along the lines [4]. Between E11.5 and E12.5 in mice, the placodes invaginate to form bud structures (Fig. 2a; Fig. 3a). At the same time, mesenchymal cells surrounding the buds condense further to form a dense mammary mesenchyme (Fig. 3a) that expresses the androgen receptor, and epithelial cells overlying the bud differentiate into specialized nipple skin. Until this stage the mitotic index in mammary buds is lower than in the surrounding epidermis, consistent with the idea that displacement of epithelial cells, rather than increased proliferative activity, accounts for early morphological changes [1, 5]. However, definitive proof of cell displacement in the mammary region is lacking.

Figure 2
figure 2

Schematic depiction of stages in the embryonic development of mammary gland (a), hair follicle (b) and tooth (c). Black ovals, mesenchymal cells; yellow oval, fat pad; pink, brown and red, epithelial cells; purple, enamel knot.

Full size image
Figure 3
figure 3

Comparison of the histology of embryonic mammary gland (a), hair follicle (b) and tooth (c) at the bud stage. Note that the mammary bud is shown at a more advanced stage than the hair follicle and tooth buds. All three embryonic buds show invagination of epithelial cells into the underlying mesenchyme, and condensation of adjacent mesenchymal cells.

Full size image

In male embryos of many mouse strains, the mammary buds become separated from the surface ectoderm between E14.5 and E15.5 under the influence of fetal androgens, and eventually degenerate. In female embryos the buds remain quiescent from E14.5 until approximately E16, when the epithelial cells begin to proliferate and each bud extends through the mesenchyme into the developing fat pad located in the dermis, forming a mammary sprout. Once the sprout reaches the fat pad it starts to branch and subsequently a lumen forms within the epithelium (Fig. 2a). Normal development of the epithelium depends on interactions with the fat pad [6]. At birth the mammary gland is composed of a primary duct and 15–20 secondary branches invested in the mammary fat pad.

Molecular Regulators of Embryonic Mammary Development

Formation of the Mammary Line

The mammary line is marked in mouse embryos at E10.5 by expression of a TOPGAL transgene [7] that is sensitive to activation of the WNT/β-catenin signaling pathway [8]. TOPGAL is activated both in the epithelial line and in underlying mesenchymal cells. Ectopic expression of the secreted WNT inhibitor Dickkopf1 (Dkk1) in the surface ectoderm abolishes TOPGAL expression as well as morphologic and molecular markers of mammary placode development, indicating that WNT/β-catenin signaling plays an essential role in placode formation [7]. Of those Wnt genes whose expression in the mammary region has been examined to date, Wnt10b shows the earliest localized expression, appearing as a series of discontinuous streaks along the mammary line [3, 7]. Ectopic Dkk1 inhibits localized expression of Wnt10b, indicating that Wnt10b upregulation requires an earlier WNT signal.

Fgf10 is expressed in ventral somites prior to appearance of the mammary line, is required for formation of four of the five pairs of mammary placodes in mouse embryos, and is unaffected by surface ectodermal expression of Dkk1 [7, 9]. Fgf10 is required for TOPGAL and Wnt10b expression in the mammary line except for the region encompassing placode 4, whose development is unaffected in Fgf10-null embryos. These observations suggest that Fgf10 signaling acts upstream of WNT activation in most of the mammary line [7, 10]. FGF10 secreted from somitic cells may interact directly with FGFR2B expressed in the surface epithelium, as FGF receptor expression was not detected in somites or dermal mesenchyme, and Fgfr2b-null embryos have mammary phenotypes similar to those of Fgf10-null mutants [2, 10]. Defects in the mammary line are detected in Gli3 and Pax3 mutants that display altered somitic expression of Fgf10 [10].

Another potential regulator of WNT pathway activity in the mammary line is the ERBB4 ligand Neuregulin 3 (NRG3). Aberrant expression of Nrg3 causes altered mammary specification, and exogenous Nrg3 can induce ectopic mammary bud formation [11]. The timing and localization of initial expression of Nrg3 and Erbb4 are similar to that of TOPGAL, and exogenous Nrg3 can enhance expression of another WNT reporter transgene, BAT-gal [12], in the mammary line [11]. It is not yet known whether TOPGAL or BAT-gal activity is affected in Nrg3 mutants, or how Fgf10 signaling impacts on Nrg3 expression.

Development of Mammary Placodes and Buds

Formation of four of the five pairs of mammary placodes in mouse embryos requires the FGF receptor 2b, the WNT pathway transcription factor LEF1, and the T-box transcription factor TBX3, all of which are expressed in placodes (reviewed in [13]). In addition, genes encoding another FGF receptor, FGFR1b, and several FGF ligands are expressed in mammary placodes [13]. Forced activation of WNT/β-catenin signaling in cultured mouse embryos by incubation in LiCl or WNT3A causes accelerated development of expanded placodes; however expression of the TOPGAL WNT reporter transgene is restricted to regions close to the mammary line, underscoring the importance of additional factors that must control competence for response to WNT [7]. WNT inhibition blocks localized expression of both Tbx3 and Lef1 in the mammary region, indicating that these genes lie downstream of an initial WNT signal [7, 14]. Loss of Tbx3 function prevents localized expression of Wnt10b and Lef1 [15], and treatment of cultured embryos with an FGF inhibitor blocks expression of Tbx3 and Lef1 [14].

Another important regulator of mammary placode development is p63, a member of the p53 tumor suppressor gene family. p63 is expressed in the embryonic ectoderm and in self-renewing basal layers of all stratified epithelia in the adult. Two classes of p63 transcripts are produced that either code for, or lack a potent N-terminal transactivation domain (TA and ΔN isoforms, respectively) [16]. Ablation of p63 in mice leads to total lack of multilayered epidermis and its derivatives including teeth, hair follicles, and mammary glands [17, 18]. Localized Lef1 expression is not detected in the mammary region of p63-null embryos suggesting that mammary placodes fail to form [17]. Interestingly, p63 seems to regulate splicing of Fgfr2 transcripts such that the 3b isoform is preferentially produced in p63 mutants [19]. Absence of Fgfr2b may account for failed mammary bud development in p63-null embryos [9]. However, whether the mammary line is properly defined in the absence of p63, and the precise relationship of this factor to WNT activation, are currently unknown. A model of possible interactions between p63, Fgf10, Nrg3, WNT, Tbx3 and Lef1 in mammary placode formation is depicted in Fig. 4a.

Figure 4
figure 4

Models of possible regulatory factor interactions in mammary gland, hair follicle and tooth placode induction.

Full size image

Invagination of mammary placodes to form buds, and maintenance of buds, appears to require WNT/β-catenin signaling, as embryos lacking Lef1 form four pairs of small placodes that degenerate prior to bud development [13]. The homeodomain-containing transcription factor gene Msx1 is expressed in mammary bud epithelium, and the related gene Msx2 is expressed in both epithelial and mesenchymal cells. While loss of function of either of these genes alone does not affect bud formation, in Msx1-/-; Msx2-/- double mutants mammary development is arrested at the placode stage, revealing an essential role(s) for these factors in bud morphogenesis [13].

Development and Branching of the Mammary Sprout

Elongation of mammary buds to form a sprout requires parathyroid hormone-related protein (PTHrP), expressed in the epithelium, and its receptor, PTH1R, expressed in the mesenchyme [13]. Mammary buds form in embryos lacking either of these proteins, but sprout development, differentiation of specialized mammary mesenchyme, and formation of the nipple do not occur. Remarkably, when PTHrP is overexpressed in the epidermis, the entire ventral surface of the embryo is transformed into nipple skin [20, 21].

WNT/β-catenin signaling is activated in a subset of epithelial cells accompanying initial branch formation [7], but its functional significance is not yet known. Failure of branching morphogenesis in embryonic mammary glands is observed in mice null for the homeodomain transcription factor gene Msx2 [22], or lacking EGF receptor, the EGFR ligand amphiregulin (AREG) or the transmembrane metalloproteinase ADAM17 that releases AREG from its inactive transmembrane form [23]. In addition, mammary buds from RhoGAP p190B-null embryos fail to show ductal outgrowth when transplanted into cleared fat pads, indicating an essential role for this negative regulator of Rho in ductal development [24].

Overview of Hair Follicle Development

Like mammary glands, hair follicles develop via extensive interactions between cells of the surface ectoderm and underlying dermal cells [25–27]. The first morphological sign of hair follicle formation is the development of a regular array of placodes, or local thickenings, in the surface ectoderm. Placode formation is thought to require signals from the dermis [25] (Fig. 2b). Signaling from each placode to the dermis causes the formation of a cluster of mesenchymal cells, known as a dermal condensate (Fig. 3b) [25]. Reciprocal signaling from the dermal condensate to the epithelium results in the proliferation and downgrowth of epithelial cells into the dermis, forming a hair germ. The hair germ elongates, and the epithelial cells surround the dermal condensate, known from this stage on as the dermal papilla. The follicular epithelium differentiates to form several concentric cell layers of the hair shaft and inner root sheath [28]. An outer root sheath that is contiguous with the epidermis surrounds the inner root sheath, and the follicle is bound by a connective tissue sheath. Production of the hair shaft and inner root sheath is dependent on signaling between the dermal papilla and the follicular epithelium [29], and also requires communication between cells in different epithelial layers of the follicle [26].

Molecular Regulators of Hair Follicle Development

The Nature of the “First Dermal Signal” Initiating Hair Follicle Development

The nature of the first dermal signal that initiates hair follicle development is not known. Experiments in the chick suggest that the dermal signal that initiates feather formation is uniform, rather than periodically localized [30]. Further experiments in the chick suggest Fgf10 as a component of the dermal signal, as this gene is expressed in the dermis and a blocking antibody to FGF10 suppresses morphological and molecular indicators of placode formation [31]. Consistent with an early role for FGF signaling in hair follicle morphogenesis, disruption of the mouse Fgfr2b gene causes defective skin development as well as decreased initiation of placode formation [32]. Fgf10-/- embryos do not display this defect, indicating that, if it plays a role in hair follicle initiation, FGF10 must act redundantly with one or more other FGF family members. Activin may also be a component of the dermal signal, based on its mesenchymal site of expression and its ability to induce the placode promoting factor Edar (see below) in embryo skin explants [33]. Nuclear localization of β-catenin, an indication of WNT/β-catenin signaling pathway activity, is detected in the upper dermis 2 days before initiation of feather placode development in the chick, suggesting that this pathway may also play a role in generating the dermal signal [34]. Consistent with this, Lef1 is expressed in the mesenchyme of the mouse vibrissa pad prior to vibrissa follicle development, and initiation of vibrissa follicle development is dependent on mesenchymal Lef1 expression [35].

Development of Hair Follicle Placodes

The initiating dermal signal has been suggested to trigger expression of molecules that either promote or repress follicle fate and compete with one another, resulting in the establishment of a regular array of follicles [36, 37]. Consistent with this model, several positive and negative regulators of hair follicle fate are initially expressed uniformly in the epidermis and subsequently become localized to placodes [26]. In TOPGAL WNT reporter transgenic skin, expression of the reporter is detected in placodes and dermal condensates, providing evidence that the WNT pathway is active in both epithelial and mesenchymal components of developing follicles (reviewed in [26]). Expression of a stabilized form of β-catenin in chick or mouse epidermis causes formation of ectopic follicles, indicating that activation of the WNT signaling pathway in the epithelium is sufficient to direct follicle development [38]. Conversely, loss of function of the β-catenin gene in mouse epidermis results in defective placode development [39], and a null mutation in the mouse gene encoding WNT effector protein LEF1, causes failure of formation of vibrissae and two-thirds of the body hair follicles [35, 40]. The P63 transcription factor is also required for hair follicle placode formation; in mice lacking this factor, hair placodes are not morphologically visible, and localized expression of placodal marker genes is not detected [41].

Ectodysplasin (EDA), a molecule related to tumor necrosis factor, and its receptor ectodysplasin receptor (EDAR) act early in the development of several classes of hair follicle in the mouse, including the earliest developing (primary or guard), and late developing zigzag underfur follicles, but are not required for the formation of awl and most vibrissa (whisker) follicles (reviewed in [26, 27]). Histological changes suggesting pre-placode initiation can be detected transiently in Eda mutant (Tabby) and Edar mutant (downless) embryos, but most placode markers are not expressed [33, 42]. Conversely, over-expression of the A1 isoform of EDA results in formation of hair follicle placodes of increased size, and treatment of embryonic skin with recombinant EDA-A1 in vitro promotes placodal cell fate in a dose-dependent manner [43]. The Eda-null and Edar-null phenotypes are mimicked by ectopic expression of a super-repressor of NF-κB signaling [44], and NF-κB signaling is suppressed in skin epithelium, developing teeth and mammary glands of Eda-null and Edar-null mutants [42] (M. Mikkola, unpublished observations). These findings suggest that Eda and Edar act via NF-κB at early stages of primary and zigzag hair follicle formation, and that alternate signaling pathways control development of awl and whisker follicles. The homeobox-containing genes Msx1 and Msx2 are expressed in placodes [45], and mutant mice lacking both of these genes have reduced numbers of hair follicles [22], indicating that MSX transcription factors also promote placode fate. However, their relationship with other molecular regulators of hair follicle placode formation is not yet clear.

Regulation of Hair Follicle Placode Spacing

The WNT inhibitor Dkk1 is endogenously expressed in dermal cells surrounding developing hair follicles. Endogenous Dkk1 is suppressed by ectopic keratin 14 promoter-driven Dkk1, suggesting that WNT signaling may normally induce Dkk1 expression as part of a negative feedback loop to limit hair follicle growth and perhaps regulate follicular spacing [46]. Members of the bone morphogenetic protein (BMP) family of secreted signaling molecules can also act to inhibit hair follicle formation. Several Bmp family members are endogenously expressed in hair and feather follicle placodes and dermal condensates, and their ectopic expression suppresses feather bud formation [26]. A null mutation in the gene encoding the BMP inhibitor Noggin causes formation of fewer hair follicles than normal and retarded follicular development; conversely, ectopic expression of Noggin in chick or mouse embryonic skin causes enlarged and ectopic follicles [47, 48]. Consistent with an inhibitory role for BMP signaling in embryonic hair follicle development, loss of function of BMP receptor IA in the epithelium leads to accelerated hair follicle formation [49]. A model for possible interactions of WNT, EDA, SHH and BMP factors in hair follicle placode development is depicted in Fig. 4b.

Formation of Hair Follicle Buds

Induction of the dermal condensate fails in the absence of epithelial β-catenin [39] suggesting a requirement for epithelial WNT signaling in generating secreted signals that direct clustering of mesenchymal cells. In contrast, the secreted signaling factor Sonic hedgehog (SHH) is not required for formation of the dermal condensate, but instead acts within the follicular epithelium via Gli2 and N-myc to control proliferation and downgrowth of follicular cells at the bud stage [50–53]. Expression of Shh and its target/effector Gli1 is absent from the skin of mice ectopically expressing Dkk1, or lacking epithelial β-catenin, Eda or Edar, and is detected upon rescue of Eda-null skin with recombinant EDA, indicating that Shh lies downstream of WNT and EDA/ EDAR signaling [33, 39, 46, 54, 55].

Mice lacking a functional Tgfβ2 gene have reduced numbers of hair follicles and a delay in follicular morphogenesis [56]. Tgfβ2 is required for expression of Snail and down regulation of E-cadherin expression, both of which are correlated with hair follicle bud formation [57]. Down regulation of E-cadherin also requires a combination of Noggin, which activates Lef1 expression, and stabilized beta-catenin, induced by a WNT signal [58]. These observations suggest that several different signals converge to control E-cadherin expression, and provide a possible link between transcription and cell adhesion in the bud.

Differentiation of the Hair Shaft and Inner Root Sheath

As the hair follicle bulb appears, at least seven different epithelial cell layers constituting components of the mature hair follicle are formed, including the three layers of the inner root sheath, as well as the cuticle, cortex and medulla of the hair shaft [28]. Signaling by epithelial BMP receptor 1A is essential for matrix cells in the follicular bulb to cease proliferating and differentiate into any of these structures [49, 59–61]. The transcription factor GATA3 is required for inner root sheath differentiation [62], and Notch signaling is essential for maintenance of this structure [63], while differentiation of the outer root sheath requires the HMG-box-containing gene Sox9 [64]. Other transcription factors required for hair shaft differentiation include a homeobox protein HOXC13 [65], and a winged-helix/forkhead transcription factor FOXN1 (formerly WHN) [66–68], which is mutated in nude mice and regulates expression of an acidic hair keratin gene [69].

Overview of Tooth Development

Tooth morphogenesis is governed by interactions between the oral ectoderm and neural crest-derived ectomesenchyme (for reviews see [70–72]). Initiation of individual teeth is preceded by the formation of dental lamina, a horseshoe shaped stripe of thickened epithelium marking the future dental arch (E11 in mouse embryos). All mammalian teeth bud from the dental lamina, and in this respect dental lamina resembles the mammary line. In mouse, where a toothless diastema region separates the incisor from three molar teeth in each jaw quadrant, the anatomical nature of the dental lamina has been debated [73, 74]. The odontogenic field can, however, be visualized by epithelial molecular markers such as Pitx2 [75, 76], a homeobox transcription factor required for tooth morphogenesis to proceed beyond the placode/early bud stage [77, 78], and Shh [79].

By E12 in mouse embryos, one incisor and one molar placode emerge from the dental lamina. The second molar placode buds from the posterior end of the first molar placode, and the third molar buds off from the second one. Between E12 and E13, dental placodes invaginate further into the mesenchyme which condenses around the nascent tooth bud (Fig. 2c; Fig. 3c). After the bud stage, cells at the tip of the bud form the primary enamel knot, a histologically recognizable aggregate of cells that controls the shape of the tooth crown [70, 80] (Fig. 2c). The enamel knot is a transient signaling center consisting of non-proliferating cells expressing more than ten different growth factors, most of which are also expressed in the nascent dental placodes [81]. The dental epithelium lateral to the enamel knot expands and folds, becoming cap-shaped by E14–E15, and enveloping mesenchymal cells which form the dental papilla and later give rise to dentin-secreting odontoblasts and the tooth pulp. At this stage the epithelial cells differentiate into outer enamel epithelium, stellate reticulum, stratum intermedium, and inner enamel epithelium, lying adjacent to the papilla. By the bell stage (E16 onwards) the enamel knot is removed by apoptosis, and secondary enamel knots form sequentially within the epithelium marking the tips of future cusps and thereby patterning the tooth crown. Dental papilla cells adjacent to the epithelium differentiate into odontoblasts, while cells of the inner enamel epithelium differentiate into enamel-producing ameloblasts [82].

Molecular Regulators of Tooth Development

Establishment of the Dental Field

Classic tissue recombination experiments indicate that before the formation of individual tooth primordia (E9–E11), odontogenic potential resides in the presumptive dental epithelium that can induce tooth formation in neural-crest mesenchyme even from regions normally not participating in tooth development [74, 83]. From the placode stage (E12) onwards this capacity shifts from the epithelium to the mesenchyme. This transition is preceded by a period when expression of many mesenchymal transcription factors such as Pax9 (see below) becomes independent of the epithelium [84–86], and coincides with a shift in expression of Bmp4 from epithelium to mesenchyme [87].

Regional specification of the tooth area is thought to be dependent on antagonistic effects of Fgf8, Fgf9 and Bmp4 expressed in the epithelium [70, 71]. At early stages of development, their spatial distribution is believed to instruct regionalized expression of mesenchymal transcription factors and thereby pattern the jaws into proximal (molar forming) and distal (incisor forming) domains, possibly even prior to dental lamina formation [88–90]. Fgf8 can induce expression of the homeobox gene Pitx2 while BMP4 represses Pitx2 expression [91]. Slightly later, but still prior to morphological signs of odontogenesis, interplay of FGFs and BMPs positions the expression of Pax9 in prospective molar and incisor mesenchyme [84]. Although tooth development is arrested at the bud stage in Pax9 deficient mice, tooth primordia develop in their correct locations implying that other genes are also involved in defining the location of teeth [92].

Development of the Dental Lamina

The cellular mechanisms of dental lamina formation are poorly understood. Interestingly, as seen in the mammary line, it appears that dental lamina formation does not result from local increase in mitotic activity, but rather is associated with changes in orientation of mitotic spindles such that they become perpendicular to oral epithelium instead of the generally parallel orientation [93]. Although this might suggest a stratification mechanism similar to that occurring in the epidermis [94], loss of p63 function prevents epidermal stratification but does not affect formation of the dental lamina [41].

Formation of Dental Placodes and Buds

Although the dental lamina forms in p63 mutants, dental placode formation is abolished in these embryos [17, 18, 41]. p63 is essential for expression of several signaling molecules in the ectoderm, including Fgfr2b, Bmp7, Jagged1, and Notch1, indicating that these lie downstream of p63 and suggesting that p63 activates multiple signaling pathways involved in ectodermal organogenesis [41]. In double knockout mice for transcription factors Msx1 and Msx2, or Dlx1 and Dlx2, tooth development is arrested at the placode-early bud stage (in compound Dlx mutants only upper molars are affected) [95, 96]. Both FGFs and BMPs induce expression of Msx1 and Dlx2; Msx2 is induced by BMP4, and Dlx1 by FGF, highlighting the importance of these signaling pathways in regulating invagination of dental lamina [87, 96, 97]. A loss of function mutation in Pitx2 causes arrest of tooth development at the placode-early bud stage, and is associated with failure to maintain ectodermal Fgf8 expression, with expanded Bmp4 expression [78]. Mice ectopically expressing the secreted WNT inhibitor Dkk1 under control of a keratin 14 promoter also display arrested dental development at the dental lamina-early bud stage, indicating an essential role for WNT/β-catenin signaling in bud development [46]. Mice deficient in Eda do not display altered dental placode development [98, 99]; however forced expression of Eda-A1 promotes survival of an extra dental placode normally found anterior to the first molar in wild type embryos suggesting that this molecule can function to promote dental placode survival [43, 100]. Likewise, in mice deficient for ectodin, a secreted modulator of BMP and WNT pathways, an extra tooth develops from a similar ectopic placode [80]. Whether Eda and ectodin function in the same or in parallel pathways to stimulate placode formation is currently unknown. Interestingly, an extra tooth develops in this location in embryos carrying a hypomorphic mutation in polaris, a protein required for the assembly of cilia [101].

Embryos null for two downstream mediators of Shh, Gli2 and Gli3, show no sign of molar development, revealing an important role for Shh signaling in molar development; however aberrant incisor buds are formed in Gli2/3 double mutants [102]. It is not clear whether formation of incisor buds is due to signaling through Gli1 in these mutants or through Shh-independent signaling pathways. A schematic summary of the roles of signaling and transcription factors in dental lamina and placode development is shown in Fig. 4c.

Shh acts within the epithelium to regulate proliferation during odontogenesis [103–105], analogous to its role in developing hair follicles. Mutation of Shh or its effector Smoothened causes abnormalities in tooth shape, possibly due to defective epithelial proliferation [104, 105]. Intriguingly, in embryos null for the protein kinase IKKα, incisor and whisker epithelia grow outwards instead of invaginating into the mesenchyme [106], indicating that IKKα regulates the direction of epithelial growth. The mechanism by which this occurs does not involve NFκB signaling, but is not yet understood.

Transition from the Bud to Cap Stage

Folding at the base of the tooth bud and formation of the enamel knot begin at the late bud stage [74, 107]. These processes fail to occur in mice lacking the transcription factors LEF1, MSX1 and PAX9, which are targets of WNT, BMP and FGF pathways, respectively [35, 92, 96]. Interestingly, dental mesenchymal BMP4 expression is lost in Msx1 and Pax9 mutants. Msx1 deficiency can be rescued, at least to the cap stage, by exogenous BMP4 emphasizing the importance of BMP4 for development of the cap stage and induction of the enamel knot [108–110]. Consistent with a role for BMP signaling from the mesenchyme to the epithelium in regulating transition from the bud to the cap stage, epithelial-specific deletion of Bmpr1a causes arrested tooth development at the bud stage [49]. FGF4, a direct transcriptional target of LEF1 in the enamel knot, can overcome developmental arrest in Lef1 null teeth [111]. As LEF1 is required only at a single step in tooth morphogenesis [111] it may act redundantly with other TCF molecules in mediating canonical WNT signaling at other developmental stages. Runx2, encoding a runt domain transcription factor essential for bone development, functions in the mesenchyme and is essential for mediating FGF signaling during the transition from bud to cap stages [112, 113].

Tooth Shape

At the bell stage, induction of secondary, non-proliferating, enamel knots together with proliferation of the rest of the epithelium results in formation of multiple cusps. The species and tooth type specific cusp patterns are thought to be fixed by the positions of secondary enamel knots that express multiple signaling factors [70]. Normal cusp patterning requires the TNF family member Eda acting via the NFκB pathway [98, 106, 114]. In mice lacking the BMP inhibitor ectodin, enamel knots are expanded and cusp patterns irregular, revealing a key role for BMP inhibition in cusp patterning [80].

Common and Unique Strategies in Appendage Development

Initiation of Appendage Development

Induction of the various types of skin appendage is strictly regulated with respect to developmental stage and body region. Thus, mammary gland initiation is confined to two bilateral stripes of cells running from forelimb to hindlimb; teeth develop along the dental laminae of the maxilla and mandible; vibrissae form in a specific region of the snout; and hair follicles are initiated at a later stage and develop all over the body with the exception of footpad/palm regions. As discussed earlier, current evidence suggests that for hair follicle and mammary gland development, initiating signals arise in the dermis, whereas an early epithelial signal is thought to initiate tooth development and can induce formation of dental structures in combination with neural crest derived mesenchyme or aggregated embryonic or adult stem cells [115]. Whether these apparent differences in the site of initiating signal reflect regional differences in signaling capacities of surface epithelium versus mesenchyme, or are due to the use of different embryonic stages in recombination studies, is not entirely clear.

Regional specification of the tooth area is thought to depend on antagonistic effects of epithelial BMP and FGF factors [70, 71]. Less is understood about initiating signals for mammary gland and hair follicle development, although FGF signaling may play a role in both processes [7, 9, 32], and activation of dermal WNT/β-catenin signaling is observed under both mammary line [7] and dental lamina (E. Y. Chu and S. E. Millar, unpublished data).

Do Mammary, Tooth and Hair Follicle Placodes Arise through Similar Mechanisms?

The mammary line and the dental lamina are morphologically similar, and in both cases these thickened epithelial lines resolve to form a series of localized placodes. As discussed above, circumstantial data suggest that in the case of the mammary line, the formation of placodes could involve directed cell movement within the mammary line. Such evidence is not available for the dental lamina, and prevailing models favor the idea that positive and negative regulators of dental placode induction compete to establish final positions for these structures. In contrast, hair follicle development is first apparent in epithelium at the placode stage. Morphologically, the placodes first appear as broad circular thickenings, and gradually become refined to circles of smaller diameter as they begin to invaginate. Gradual restriction of placode fate regulator expression is observed, consistent with the competitive model described above (reviewed in [26, 27]). However, cell displacement from a broad placode into an invaginating bud of smaller diameter may contribute to hair follicle bud formation.

Ectopic expression of the Dkk1 WNT inhibitor under control of a keratin 14 promoter blocks formation of hair follicle and mammary gland placodes, indicating that for these organs, placode development requires WNT signaling [7, 46]. In contrast, tooth development is blocked at the dental lamina-early bud stage in K14-Dkk1 mice [46]. It is currently not clear whether this reflects a relatively later requirement for WNT signaling in dental development, or is due to inefficient inhibition of WNT activity in the oral cavity at the stage when tooth morphogenesis is initiated. FGF signaling is also required at early stages in the formation of mammary, tooth and hair follicle placodes [9, 32, 86], although the specific FGF ligands, or combinations of ligands, involved may differ. p63 is required for development of mammary, hair follicle and dental placodes [17, 18, 41]. It is not yet known whether this is also true for the mammary line. Finally, MSX transcription factors are required for hair follicle, dental and mammary bud formation [22, 96], although how Msx gene expression is controlled has not been investigated in any detail for developing hair follicles and mammary glands. Thus, from what is known to date, several signaling pathways and transcription factors appear to play conserved roles in placode and bud formation of diverse appendages. The effects of mutations in regulatory factors on early stages of embryonic mammary gland, hair follicle and tooth development are summarized in Table 1.

Table 1 Mouse mutants displaying defects at early stages of embryonic mammary gland, hair follicle or tooth morphogenesis.
Full size table

Despite these similarities in molecular regulation of early stages of appendage formation, some striking differences in pathway utilization are noted for each organ type (see Table 1). Sonic hedgehog signaling plays key roles in regulating proliferation in both hair follicle and dental epithelia, and yet is dispensable for development of primordial mammary glands [116]. It is possible that Shh acts redundantly with other hedgehog family members in embryonic mammary development, but this has yet to be demonstrated. Alternatively, distinct requirements for Shh may be in part responsible for differences in the ways that hair follicle, dental and mammary buds proliferate. Indeed, current evidence suggests that Shh causes expansion of a sub-population of epithelial cells in the tooth bud, contributing to the non-uniform shape of molar teeth [105]. EDA/EDAR signaling is necessary for maintenance of primary hair follicle placodes [42] but is not essential for formation or maintenance of awl hair, vibrissa, mammary or dental placodes [37, 98–100]. Forced expression of Eda-A1 in surface ectodermal cells causes formation of an ectopic mammary placode and dental placode within the mammary line and dental lamina regions respectively, eventually producing an ectopic mammary gland and tooth [43, 100]. These data indicate that Eda-A1 is capable of promoting formation of these structures. Eda does not stimulate cell proliferation but rather appears to expand placode cell fate by promoting incorporation of neighboring cells into nascent placodes [43]. It is possible that related TNF and TNF receptor family members (such as Troy and XEDAR) might act redundantly with EDA/EDAR in development of mammary, dental and awl hair follicle placodes. However, expression of an NF-kB super-repressor recapitulates Eda-null and Edar-null phenotypes [44], and NF-kB reporter activity is absent from awl follicle placodes of Eda-null and Edar-null mutant embryos [27], as well as from dental and mammary placodes in these mutants (M. Mikkola, unpublished data). Thus it is more likely that an alternate, parallel signaling pathway is responsible for maintenance and further development of placodes that are unaffected by loss of EDA/EDAR/NFκB signaling.

In addition to these differences in pathway utilization, a number of transcription factors have been found to function differently in the development of each appendage. For instance, Tbx3 is required for formation of mammary but not vibrissa buds [15], and Pitx2-null mice show defects in tooth bud formation [117], but defects in mammary or hair follicle buds have not been described. Further work will be needed to determine whether these phenotypes are due to functional redundancy with other Tbx and Pitx family members, or reflect differences in gene activation that could explain some of the morphogenetic differences between mammary, tooth and hair follicle buds.

Alterations in cell adhesion that allow controlled and limited dissociation and re-association of epithelial cells are likely to be essential for bud development. As discussed above, cell adhesion in hair follicle buds may be regulated in part via control of E-cadherin expression in response to cooperating TGF-β, Noggin and WNT signals [57, 58]. Analysis of conditional mutant mice has revealed that epithelial expression of β1 integrin is required for remodeling the basement membrane and hair follicle downgrowth [118, 119], and that α-catenin is also required for normal follicular morphogenesis [120]. Other studies suggest that remodeling of desmosomal and hemidesmosomal components is associated with both hair follicle and mammary bud formation [121, 122]. Detailed comparative studies of expression and function of cell adhesion molecules in hair follicle, mammary and tooth buds will be important for understanding mechanisms by which signaling molecules effect the morphological changes in epithelium and mesenchyme that are necessary for formation of diverse appendages.

Not surprisingly, molecular variations between developing mammary glands, hair follicles and teeth become more striking beyond the bud stage, as these organs undergo clearly distinct patterns of branching or folding morphogenesis and initiate unique programs of differentiation. How differences in signaling and transcription factor expression/activation result in the development of distinct classes of skin appendage will be a fascinating area for future study.

Human Diseases Linked to Appendage Formation

Several inherited and sporadic diseases in humans are caused by loss or gain of function mutations in signaling pathways that regulate appendage formation. These include Gardner’s syndrome, which is associated with loss of function mutation of the WNT pathway antagonist Adenomatous Polyposis Coli (APC), resulting in inappropriate WNT/β-catenin signaling and development of supernumerary teeth and odontomas as well as colon cancer [123, 124]. Loss of function of another WNT antagonist, AXIN2, results in hypodontia as well as colon cancers [125], suggesting that WNT activity must be tightly controlled for normal tooth development. Stabilizing mutations in the β-catenin gene are found in human pilomatricoma, a tumor of hair follicle matrix cells [126], and elevated WNT signaling is also associated with breast cancer in humans and in mouse models [127]. Similarly, deregulation of several other pathways required for embryonic appendage development, including the FGF, EGF and neuregulin signaling pathways, is oncogenic in mammary epithelium [128, 129]. Mutations of human P63 result in a number of dominant syndromes that affect development of the skin and its appendages including hair follicles, teeth and nails. These include ectrodactyly-ectodermal-dysplasia-clefting (EEC), ankyloblepharon-ectodermal dysplasia-clefting or Hay–Wells (AEC), acro-dermato-ungual-lacrimal-tooth (ADULT), and limb-mammary (LMS) syndromes, as well as non-syndromic split-hand/split-foot malformation (SHFM), all of which show a striking genotype-phenotype correlation [130, 131]. Mammary gland abnormalities are a consistent feature of LMS, and are occasionally observed in patients with EEC, AEC and ADULT. The EDA and EDAR genes are mutated in human anhidrotic ectodermal dysplasias, causing a decreased number of hair follicles, defects of the teeth and sweat glands and occasional mammary hypoplasia [132, 133]. Mutations in the SHH pathway target and repressor PTC1 cause Gorlin syndrome, characterized by developmental defects and basal cell carcinoma (BCC) [134–136], a tumor that has several characteristics in common with immature hair follicles, including similar histology, ultrastructure, and patterns of keratin gene expression [136–138]. The majority of sporadic BCCs are associated with PTC1 mutations [139], and inappropriate activation of hedgehog signaling is also associated with odontogenic tumors [140]. Consistent with its functions in the mouse, mutation of human TBX3 causes defects in mammary glands, teeth and limbs, but not hair follicles (Ulnar-mammary Syndrome) [141], and mutation of PITX2 (Rieger Syndrome) causes defects in teeth but is not associated with abnormal breast or hair development [142]. These findings reveal both shared and unique aspects of various human appendage developmental programs, and underscore the relevance of mouse models for dissecting mechanisms underlying human developmental syndromes and appendage tumors including breast cancer.