Digit regeneration is regulated by Msx1 and BMP4 in fetal mice

The amphibian limb has been the prototypical organ for the study of epimorphic regeneration both in adults and during development. These studies indicate a number of requirements for successful limb regeneration, including:(1) a wound healing response with the formation of an apical epithelial cap(AEC); (2) a cellular contribution, via the de-differentiation of mature tissues and/or stem cells, to form the blastemal mesenchyme; (3) a proliferative response regulated by multiple tissue interactions, including a neurotrophic effect, an AEC effect and a positional effect; and (4)morphogenesis and re-differentiation to restore the limb pattern (reviewed by Stocum, 1995; Tsonis, 1996). In general, the regeneration response can be separated into those events that are injury related and crucial for blastema formation, and those that represent a`redevelopment' response whereby regeneration recapitulates limb development(see Bryant et al., 2002). It is clear that the complexity of the regeneration response lends itself to multiple avenues whereby regeneration can be inhibited, and indeed numerous requirements for limb regeneration have been identified in this way. In parallel, there is a growing database of genes that are likely to play key roles, based on their regulated expression during regeneration, yet in most cases a functional understanding of how these genes act remains largely unexplored (see Gardiner et al.,2002).

The regeneration of amputated distal digit/finger tips has been reported in various mammals, including humans and rodents, and represents a mammalian system where successful epimorphic regeneration can occur (see Muller et al., 1999). In humans, regenerative potential is restricted to the distal tip of the finger in a region associated with the nail organ(Douglas, 1972; Illingworth, 1974). Experimental studies using rodent models demonstrate that regenerative capacity is level specific and, similarly, is associated with the proximal extent of the nail forming organ (Borgens,1982; Zhao and Neufeld,1995). A role for the nail organ in the regeneration response is supported by experiments showing that excluding the nail organ from the amputation wound results in no regenerative response; however, nail organ grafts result in ectopic bone formation and do not induce regeneration(Zhao and Neufeld, 1995; Mohammad et al., 1999). These experimental results, together with clinical studies in humans, provide evidence that the nail organ plays some role in the stimulation of adult digit tip regeneration. Genes specifically expressed in, or associated with, the nail organ are candidates for regulating this regenerative response. In previous studies, we demonstrated that the homeobox-containing genes Msx1 and Msx2 are expressed in association with the nail organ of neonatal digits, and at the apex of developing digits in the nail-forming region (Reginelli et al.,1995). Furthermore, mapping the regenerative ability of embryonic and fetal digit tips demonstrated that regenerative capacity correlated with amputation within the Msx1, but not the Msx2, expression domain in developing digits. These studies suggest a role for MSX genes in the digit regeneration response.

In all tetrapod vertebrates, Msx1 and Msx2 are co-expressed in the apical mesenchyme during limb formation. In animals that regenerate their appendages, MSX genes are upregulated during the regeneration response and downregulated in association with re-differentiation(Crews et al., 1995; Simon et al., 1995; Yokoyama et al., 2001). In the regenerating urodele limb, Msx2 is rapidly induced in the healing epidermis and subjacent tissues following amputation or simple wounding,whereas Msx1 expression is restricted to blastemal cells(Koshiba et al., 1998; Carlson et al., 1998). During fetal digit tip regeneration in the mouse, both Msx1 and Msx2 are expressed in the regenerating digit mesenchyme, whereas neither is expressed following proximal amputations that fail to regenerate(Reginelli et al., 1995). Beyond these descriptive studies, the role that MSX genes play in regeneration of fish fins, amphibian limbs or mammalian digits is largely unexplored. However, the roles of MSX genes during limb development and in cultured cells have been extensively studied. Expression of MSX genes in the apical mesenchyme of the limb bud is dependent on signaling from the apical ectodermal ridge (AER), and also on interactions with neighboring mesenchymal cells (Davidson et al., 1991; Ros et al., 1992; Wang and Sassoon, 1995). A number of factors crucial for limb formation have been shown to regulate MSX gene expression, including members of the FGF(Watanabe and Ide, 1993; Fallon et al., 1994; Wang and Sassoon, 1995; Vogel et al., 1995), BMP(Wang and Sassoon, 1995; Ganan et al., 1996; Marazzi et al., 1997; Hofmann et al., 1996; Pizette and Niswander, 1999; Pizette et al., 2001) and TGFβ (Ganan et al., 1996)signaling families, and retinoic acid(Yokouchi et al., 1991; Wang and Sassoon, 1995). One function of Msx1 in early limb development is to mediate a BMP signaling pathway that leads to the induction of the AER(Pizette et al., 2001). Nevertheless, mice carrying a targeted deletion of the Msx1 gene form normal limbs, thus indicating that AER formation can occur in the absence of Msx1 function (Satokata and Maas,1994). Because Msx1 and Msx2 are co-expressed in the apical mesenchyme it remains possible that these genes function redundantly during limb formation.

MSX gene function is implicated in the control of cellular differentiation during embryogenesis (see Bendall and Abate-Shen, 2000). The expression pattern of MSX genes during limb development is consistent with a role in the control of cell proliferation and/or cell differentiation: Msx1 and Msx2 are expressed in the apical mesenchyme in association with undifferentiated proliferating cells, whereas proximal tissues where MSX genes are not expressed are associated with reduced growth and tissue differentiation. Forced expression implicates Msx1 in the inhibition of myogenesis(Song et al., 1992; Woloshin et al., 1995), and there is evidence that this inhibitory activity can be generally extended to the differentiation of a number of mesenchymal (e.g. adipose, cartilage and bone) and epithelial (e.g. mammary) cell types(Hu et al., 2001). A similar conclusion can be drawn from the results of loss-of-function studies that indicate that both Msx1 and Msx2 mutants display defects in the formation of certain skeletal elements and ectodermally derived organs(Satokata and Maas, 1994; Satokata et al., 2000). An intriguing discovery that is potentially relevant to limb regeneration is the demonstration that regulated Msx1 expression in differentiated C2C12 myotubes induces a de-differentiation response resulting in the establishment of multi-potent progenitor cells (Odelberg et al., 2000). In addition, an extracellular activity derived from newt blastema extract possesses a similar de-differentiation activity, thus suggesting that an intercellular signal, presumably acting through MSX1, is involved in de-differentiation during the early stages of limb regeneration(McGann et al., 2001).

The availability of mice carrying targeted deletions of the Msx1and Msx2 genes, and the established regenerative response of fetal digit tips, allowed us to carry out a functional analysis of the MSX genes in regeneration. In this study, we have established a fetal digit tip regeneration model in cultured explanted autopods of E14.5 digits. Using this model, we discovered that Msx1, but not Msx2, mutant digits displayed a regeneration defect. Gene expression studies demonstrated that the Msx2 expression domain was expanded into the Msx1 domain in the Msx1 mutant digit, thus showing that Msx1 and Msx2 are not functioning in a redundant manner in digit regeneration. By contrast, the Bmp4 expression domain, which coincides with that of Msx1 in wild-type digits, was apically restricted in the Msx1 mutant digit, and Bmp4 transcripts were not detected in the Msx1/Msx2 double-mutant digit tip. Exogenous application of BMP4 was found to rescue the Msx1 mutant-digit regeneration defect in a dose-dependent manner, and exogenous noggin application inhibited the regeneration response in wild-type digits. These studies provide functional evidence linking Msx1 function and BMP signaling to the control of digit tip regeneration in the mammalian fetus.

Mice carrying a targeted deletion of the Msx1 gene(Satokata and Maas, 1994) or the Msx2 gene (Satokata et al.,2000) were kindly provided by Dr Richard Maas. Homozygous Msx1 or Msx2 mutant embryos were obtained by heterozygote mating. Msx1/Msx2 double-mutant embryos were obtained by double heterozygotes mating. At E14.5, Msx1 and Msx2 mutants cannot be distinguished from wild-type mice, so each experiment was carried out blind and tissues were collected for subsequent PCR genotyping as described (Satokata and Maas,1994; Satokata et al.,2000). Msx1/Msx2 double-mutant embryos are morphologically distinct at E14.5 and their genotype was verified by PCR. The limb buds of double mutants appear to be developmentally delayed; the morphology of E14.5 hindlimb (stage 11) is similar to the E13.5 wild-type hindlimb (stage 10). Limb stages are described by Wanek et al.(Wanek et al., 1989).

For in vivo studies, stage 11 fetal digit tips were amputated at E14.5 using exo utero surgical techniques as described(Reginelli et al., 1995; Ngo-Muller and Muneoka,2000a). Briefly, timed-pregnant mice were anesthetized with sodium pentobarbital (60 μg/g body weight), fentany (1.6 μg/animal) and droperidol (80 μg/animal). The abdomen was opened with a mid-ventral incision and fetuses were exposed by incision of anti-placental uterine wall. Access to the hindlimb was through an incision in the extraembryonic membranes and the hindlimb was teased out with a blunt probe. The three central hindlimb digits, digits 2, 3, and 4, were amputated at either a distal level,approximately 75 μm from the digit tip (see Fig. 1H), or a proximal level through the presumptive terminal interphalangeal joint. The uterus with fetuses attached was positioned within the abdominal cavity and the abdominal wall of female mouse was closed. Operated fetuses were allowed to develop for 2 to 4 days exo utero (Muneoka et al.,1986) after which the hindlimbs were collected for analysis of the digits.

For organ culture, stage 11 hindlimbs were collected from E14.5 fetuses and transferred to a dish containing lactated Ringer's solution, where digits were amputated as described above. Hindlimbs were trimmed proximally at the level of the ankle and were cultured in Dulbecco's Modified Eagle Medium supplemented with 10% fetal bovine serum on filters (Millipore, 0.45 μm pore size, 13 mm diameter) supported by a metal grid(Zhang et al., 2000). Hindlimbs with amputated digits were cultured either with or without added growth factors for 2 or 3 days with daily changes of the medium. Growth factors used in these studies included recombinant human BMP2 and BMP4 (kindly provided by Genetics Institute), and noggin (R&D). BMP2 activity was tested independently, based on the induction of Nodal during left-right axis formation following BMP2 bead implantation into the chick embryo (Schlange et al.,2002).

For in situ hybridization, digoxigenin-labeled riboprobes complementary to Msx1 (Ngo-Muller and Muneoka,2000b), Msx2(Ngo-Muller and Muneoka,2000b), Bmp4 (Jones et al., 1991), Hoxc13(Godwin and Capecchi, 1998)and Ihh (St-Jacques et al.,1999) were used in whole-mount or paraffin-sectioned preparations as described (Schaller and Muneoka,2001; Omi et al.,2002). Fetal digit tissues were fixed by immersion in 4%paraformaldehyde and post-natal digits were fixed by injection of fixative into the digits followed by immersion in 4% paraformaldehyde. For tissues processed in parallel, all aspects of the in situ hybridization staining protocol were carried out in synchrony and imaged identically. For Msx1/Msx2 double-mutant tissue that is negative for Bmp4expression, we used the expression of Bmp2 in limb tissues as a positive control for tissue viability.

For differentiating skeletal analysis in the developing digits, tissues were stained with Alcian Blue/Alizarin Red S according to methods described by McLeod (McLeod, 1980). For BrdU incorporation studies, BrdU was added to the culture 1 hour prior to tissue fixation following a protocol recommended by the manufacturer (Roche). Paraffin-sectioned tissue samples were incubated with anti-BrdU and differentiated with anti-mouse Ig-alkaline phosphatase. Incorporated BrdU was detected using NBT and X-phosphate as substrates for alkaline phosphatase.

In a previous study we established a correlation between the expression domains of Msx1 and Msx2 and the level-specific regenerative capacity of mature mouse digits (Reginelli et al., 1995). In situ hybridization studies of neonatal digits indicates that Msx1 is expressed in loose connective tissue that is sandwiched between the nail bed and the terminal phalangeal skeletal element(Fig. 1B). We observed that Msx2 expression is restricted to the cells of the nail bed(Fig. 1C), so that, in the mature digit, Msx1 and Msx2 are both expressed in a similar region of the digit tip but their expression domains are non-overlapping. Our earlier studies showed that various stages of fetal digit tip development have the capacity to regenerate in a level-specific manner(Reginelli et al., 1995). We have since focused on the regeneration of the stage 11 digit tip because at this stage the digit tip has initiated differentiation, and it retains the ability to undergo a rapid regeneration response. Overt differentiation of the digit tip is initiated by this stage as the condensing terminal phalangeal element is apparent in whole-mount Alcian Blue stained digits(Fig. 1A). At this stage, Msx1 and Msx2 are expressed in distinct domains that overlap in the distal mesenchyme. Msx1 is expressed in all mesenchymal cells at the digit tip and forms a domain that is complementary to the central chondrofying terminal phalangeal element(Ngo-Muller and Muneoka,2000b) (Fig. 1G,H). Msx2 is expressed in the apical ectoderm and in distal mesenchymal cells subjacent to the ectoderm (Fig. 1I), thus the region of Msx1 and Msx2co-expression is restricted to the apical mesenchyme. During the maturation of the digit, the Msx1 expression domain becomes restricted to the dorsal loose connective tissue that lies between the terminal phalanx and the nail organ. During that same timeframe, Msx2 expression is downregulated in the distal mesenchyme and the ectodermal domain shifts to the dorsally located nail organ.

In addition to Msx1 and Msx2 expression, Bmp4 is expressed in the stage 11 distal digit mesenchyme in a pattern similar to that of Msx1 (Fig. 1J). At this stage, the onset of endochondral differentiation of the terminal phalanx is indicated by the expression of Ihh at the distal tip of the digit(Fig. 1K), and the initiation of nail organ formation is indicated by the expression of Hoxc13 on the dorsal surface of the digit (Fig. 1L). In neonatal digits (E18.5 and postnatal), the anatomy of the terminal digit region is complete. In the dorsal mesenchyme separating the nail organ and the terminal phalanx, Msx1 is expressed prominently,both at birth (not shown) and postnatally(Fig. 1B), whereas Bmp4 is expressed weakly at birth (data not shown) but is expressed strongly postnatally (Fig. 1D). Msx2 expression is downregulated in the dorsal mesenchyme and its expression in the epidermis is restricted to the nail bed(Fig. 1C). Hoxc13expression remains associated with the forming nail and is prominent in the neonatal nail bed (Fig. 1F). The formation of the terminal phalanx itself is characterized by the distal expression of Ihh (Fig. 1E).

In vivo amputation of the stage 11 digit tip of E14.5 fetuses results in a wound healing response that is followed by the regeneration of the digit tip blastema and the eventual formation of an anatomically complete digit. This process is completed in a 4 day period so that at birth (E18.5) the digit tip has a relatively normal appearance, albeit somewhat shorter by comparison to non-amputated control digits (compare the regenerated central digits with the non-amputated peripheral digit in Fig. 2A). Gene expression studies of the 4-day-regenerated digit tip corroborate anatomical observations and indicate that the regenerate is normal. Msx1 is expressed in the dorsal mesenchyme between the nail organ and the terminal phalanx (Fig. 2B). Msx2 is downregulated in the dorsal mesenchyme, but is expressed in the epidermis associated with the nail organ(Fig. 2C). Bmp4 is weakly expressed in the dorsal mesenchyme(Fig. 2D). The differentiation markers, Ihh and Hoxc13, are expressed in the terminal phalanx and the forming nailbed, respectively(Fig. 2E,F). Analysis of the regenerating digit tip 2 days post-amputation shows that the response involves the reformation of a digit blastema distal to, and surrounding, a central cartilaginous element. Within this digit blastema marker genes are expressed largely in a manner similar to those of the developing digit tip(Fig. 2H-L). Thus, Msx1,Msx2 and Bmp4 are all expressed in the regenerating digit blastema mesenchyme, Msx2 and Hoxc13 are expressed in the apical epidermis, and Ihh is expressed in the central cartilaginous element. However, one difference is that the Msx2 expression domain in the distal mesenchyme is expanded so as to coincide with that of Msx1 (Fig. 2I). Amputation of the digit tip at a more proximal level does not result in a regenerative response, and distal digit marker genes (Msx1, Msx2 and Bmp4) are not expressed at the site of amputation injury(Reginelli et al., 1995)(Fig. 2G).

We used techniques for organ culture to determine whether the regeneration of stage 11 digits can occur in vitro. Stage 11 autopods were isolated from E14.5 fetuses, and the tips of 2-3 central digits were amputated and cultured for up to 4 days. After 2 days of culture a clear regeneration response is evident based on the reformation of the digit blastema, and after 3-4 days of culture the digit blastema elongates distally(Fig. 3A). Based on external anatomy, the in vitro regeneration response is comparable to the in vivo response. The cultured autopods appear morphologically different than in vivo limbs, primarily because the re-fusion of digits, which occurs late in limb development (Maconnachie,1979), does not occur in cultured autopods. In addition to the anatomical response, we characterized the expression of genes associated with digit regeneration in vivo. In 4-day-cultured regenerates, we observed robust expression of Msx1 and Bmp4 in mesenchymal cells surrounding the terminal phalanx, and Msx2 expression restricted largely to the apical epidermis (Fig. 3B-D). Expression of Ihh in 4-day-cultured regenerates is associated with skeletogenesis of the terminal phalanx(Fig. 3E), whereas Hoxc13 expression is associated with the apical epidermis, and is spatially more variable than in in vivo regenerates(Fig. 3F). In 2-day-cultured regenerates, an apical blastema of mesenchymal cells is present and Msx1 is expressed strongly throughout the regenerating digit tip(Fig. 3H). Msx2 is expressed strongly in the basal layer of the apical epidermis and is expressed weakly in the most distal mesenchyme (Fig. 3I). Like Msx1, Bmp4 is expressed in the distal mesenchyme (Fig. 3J). We also investigated the expression of the differentiation marker genes Ihhand Hoxc13, associated with the terminal phalanx and the forming nailbed, respectively (Fig. 3K,L). Overall, we observed an anatomical regeneration response in more than 90% of cultured wild-type digits(Table 1), and we found perfect correlation between a regeneration response and the expression of Msx1,Msx2 and Bmp4.

We used the regeneration of cultured digits as a screen to test the hypothesis that MSX genes play a role in the regeneration response. E14.5 Msx1 and Msx2 mutants were generated by heterozygote cross,and all resulting fetuses were collected and tested for digit tip regeneration. To eliminate bias, anatomical scoring for a regeneration response was carried out without knowledge of genotype. Thus, in each experiment the regenerative ability of wild-type, heterozygote and mutant digits was tested. Our studies show that Msx2 mutants regenerate in a manner that is indistinguishable from wild-type or Msx2^+/-digits, thus demonstrating that Msx2 function is not required for the regeneration response (Table 1). However, regeneration studies with Msx1 mutant digits indicate a regeneration defective phenotype, in which mutant digits regenerated at a frequency (28%) that was much lower than their heterozygote and wild-type counterparts (Table 1, Fig. 3G). This finding was verified in vivo, where Msx1 mutant digits were found to regenerate at a similarly low frequency (37%; Table 1, Fig. 3M). An analysis of gene expression in in vivo and in vitro mutant digits revealed that expression of nonfunctional Msx1 transcripts was detected at the amputation wound of 2-day regenerates (Fig. 3N),whereas Msx2 and Bmp4 transcripts were not detected(Fig. 3O,P). We were also unable to detect expression of either Ihh or Hoxc13 in 2-day non-regenerating Msx1 mutant digits(Fig. 3Q,R). In the few Msx1 mutant digits that did regenerate, we found expression of Msx2 and Bmp4 associated with the regeneration response, and we were unable to distinguish this mutant regeneration response from regenerating wild-type digits (data not shown).

The variability of the Msx1 mutant digit regeneration phenotype suggests the existence of a redundant activity that might replace Msx1 function in the regenerative response. An obvious candidate in the forming digit is Msx2 because: (1) its expression is associated with digit regeneration, (2) Msx1 and Msx2 display similar biochemical characteristics(Bendall and Abate-Shen, 2000),and (3) the Msx2 expression domain overlaps with Msx1 in the distal digit mesenchyme. To explore this possibility, we carried out an analysis comparing Msx2 expression in wild-type and Msx1mutant digits. In initial studies, we processed wild-type and Msx1mutant digits side by side for whole-mount in situ hybridization for Msx2 expression, and discovered that Msx2 expression is visibly enhanced in the Msx1 mutant digit as compared with wild-type(Fig. 4A,B), thus suggesting that Msx2 expression was compensating for the absence of Msx1. When the spatial pattern of Msx2 expression was analyzed in the Msx1 mutant digit, we discovered that the mesenchymal Msx2 expression domain was expanded to encompass the wild-type Msx1 expression domain (Fig. 4C). These observations suggest that Msx1 functions to restrict the Msx2 expression domain during digit formation, although this inhibition must involve additional activities, as both genes are co-expressed in the distal mesenchyme. These data are consistent with a redundancy in Msx1 and Msx2 function during digit development; however, because the Msx1 mutant digit displays a regeneration defect, the evidence suggests that Msx1 and Msx2 are functioning in a partially redundant way in regeneration.

To explore the defective regeneration response in Msx1 mutant digits, we analyzed the expression of Bmp4. By processing Msx1 mutant and wild-type digits for whole-mount in situ hybridization in parallel, we found that Bmp4 expression was visibly reduced in the Msx1 mutant (Fig. 4D,E). Analyzing for changes in spatial expression of Bmp4 indicated that the Bmp4 expression domain changed from being coincident with the Msx1 expression domain to being coincident with the wild-type Msx2 expression domain in Msx1 mutant digits (Fig. 4F). We next analyzed Bmp4 expression in the tip of Msx1/Msx2double-mutant digits and found no detectable transcripts by in situ hybridization (data not shown). These data are consistent with the conclusion that Bmp4 is regulated by both Msx1 and Msx2, and raises the possibility that BMP4 may be a limiting factor in Msx1mutant digit regeneration.

To test this hypothesis, we carried out rescue experiments with exogenous BMP4. Amputated hindlimb digits obtained from Msx1^+/-crosses were cultured in medium containing recombinant human BMP4 at two different concentrations, 200 ng/ml and 1000 ng/ml. Treatment at the lower BMP4 concentration resulted in a partial rescue of the regeneration response(73%, Table 1), whereas treatment at the higher concentration enhanced the regeneration response to a level comparable to that seen in wild-type controls (86%, Table 1). In parallel studies,application of BMP2 (200 ng/ml) had no effect on Msx1 mutant digit regeneration (27.8%, Table 1),indicating that the rescue effect is specific to BMP4. The BMP4 rescue of the Msx1 regeneration defect is associated with a striking upregulation of Msx2 expression in the apical epidermis(Fig. 5A), and of Bmp4expression in the induced digit blastemal mesenchyme(Fig. 5B). Hoxc13 is also expressed in the dorsal epidermis in the BMP4-induced regenerate(Fig. 5C), and Ihh is expressed in the forming terminal skeletal element (data not shown). In summary, these results show that BMP4 functions in a dose-dependent manner downstream of Msx1, and identifies BMP4 as an essential regulator of the regeneration response.

As an alternative approach to investigate the role of BMP4 in digit tip regeneration, we treated cultures of amputated wild-type and mutant digits with the BMP-binding protein noggin. As Bmp2 is not expressed during later stages of mouse digit formation (M.H., unpublished), and as exogenous BMP2 does not rescue the Msx1 mutant regeneration defect, noggin treatment is likely to act as a specific inhibitor of BMP4 in our digit regeneration studies. The influence of noggin (200 ng/ml) on regeneration was tested in wild-type digits, and in Msx1 heterozygote and Msx1 mutant digits resulting from Msx1^+/-crosses. Overall, noggin treatment caused a reduction of regeneration frequency in all three groups. Msx1 mutant digit regeneration was reduced to 8.7% (Table 1), a 3-fold reduction compared with the 27.8% regeneration observed without noggin treatment. This finding provides support for the conclusion that reduced levels of BMP4, as indicated by in situ hybridization analyses, is responsible for the low level regeneration response in Msx1 mutant digits. This observation also implicates Bmp4 as the downstream target gene responsible for the partial redundancy displayed by Msx1 and Msx2 in digit regeneration. Msx2 or Bmp4transcripts were not detected by in situ hybridization in noggin-inhibited Msx1 mutant digits (data not shown). Noggin treatment of Msx1 heterozygote and wild-type digits reduced the regeneration frequency 5-fold, from 90.9% to 18.3%(Table 1). Unlike the Msx1 mutant digits, noggin-inhibited amputated wild-type digits do express Msx1, Msx2 and Bmp4 transcripts at the wound site(Fig. 5D-F). This data suggests that Msx2 and Bmp4 are functionally downstream of Msx1, and that BMP4 signaling is not a prerequisite for the maintenance of either Msx1 or Msx2 expression. In summary,these data provide additional support for the conclusion that BMP4 signaling is essential for the fetal digit tip regeneration response.

In tetrapod vertebrates, epimorphic regeneration of adult and developing limb tissues is associated with a localized growth response at the site of injury. To analyze cell proliferation during the fetal digit tip regeneration response, we carried out BrdU incorporation studies of amputated wild-type and Msx1 mutant digits after treatments to modulate the regeneration response. After 2 days in culture, ectodermal wound closure is complete and a morphological response is evident. In wild-type digits, we found localized incorporation of BrdU associated with outgrowth of the regenerating digit blastema (Fig. 6A). By contrast, we observed little BrdU incorporation at the injury site in Msx1 mutant digits that fail to mount a regeneration response(Fig. 6B). BMP4-treated Msx1 mutant digits displayed BrdU-labeled cells associated with the rescued regeneration response (Fig. 6C), providing evidence that BMP4 induces a proliferative response in regenerating digit blastema cells. Consistent with this conclusion, BrdU incorporation was reduced in noggin-treated wild-type digits in which regeneration is inhibited (Fig. 6D), which indicates that BMP signaling is required for proliferation. These studies provide evidence that fetal digit tip regeneration is an epimorphic response associated with an apical growth zone,and suggests a role for Msx1 and BMP4 in the control of cell proliferation following amputation injury.

Amputation of the murine fetal digit tip results in an epimorphic regenerative response that restores the forming digit blastema, which subsequently undergoes differentiation to form a normal digit tip. This regenerative response occurs both in vivo and in vitro, and we have used this model to provide functional evidence that both MSX1 and BMP4 are required for successful regeneration. Regeneration rescue experiments show that regeneration-defective Msx1 mutant digits are induced to regenerate by the exogenous application of BMP4, and that this response is dose-dependent. Gene expression studies suggest that regeneration may be regulated by both MSX1 and MSX2, via their combined involvement in controlling Bmp4 expression during digit development. Experiments on wild-type digits showing that regeneration is inhibited by application of the BMP-binding protein noggin confirm a role for BMP signaling in the control digit regeneration. These studies provide both loss- and gain-of-function evidence that MSX1 and BMP signaling are required for the successful regeneration of the fetal mouse digit tip.

The capacity of embryonic/fetal tissues to undergo enhanced regenerative repair identify developing tissues as relatively simple models for investigating regenerative responses (see Muller et al., 1999). The regenerating mammalian digit tip is no exception to this rule. In mice, mature digit tip regeneration is level-specific, associated with the nail organ, and proceeds very slowly, requiring weeks to months to complete. Fetal digit tip regeneration is also level-specific and associated with the nail anlagen, but is completed within a 4 day period. Because the regeneration response occurs so rapidly in fetal digits, we have been able culture this regenerating structure in vitro, where the regeneration environment can be manipulated. The fact that a regeneration response can be elicited in explanted limb tissue indicates that components that are likely to play crucial roles in mature digit regeneration, such as functional innervation, vascularization and the availability of circulating hormones, are not required for fetal regeneration. Thus, the simplicity of the regeneration response provides a means to experimentally dissect underlying regulatory mechanisms that might otherwise be inaccessible in the mature digit.

Regeneration studies on the mature mammalian digit tip have focused almost exclusively on the role of the nail organ and the distal growth of the terminal phalanx (see Zhao and Neufeld,1995). This association is derived from both clinical studies in humans and experimental studies on rodents(Douglas, 1972; Illingworth, 1974; Borgens, 1982). By studying both developing and mature digits, we found that regenerative potential is associated with the expression domain of Msx1(Reginelli et al., 1995), but that Msx1 expression in mature digits is restricted to loose connective tissue fibroblasts subjacent to the nail, and is not found in the nail organ itself. The demonstration that Msx1 mutant digits display a regeneration defect indicates that Msx1-expressing cells play a crucial role in the response. These findings suggest that digit tip regeneration may not be dependent on the nail organ, but rather on the connective tissue cells underlying the nail organ. Because this population of cells is closely associated with the nail organ it would be difficult to distinguish between a nail organ effect and the influence of the underlying connective tissues. The importance of this cell population in mature digit regeneration is also supported by histological observations noting a streaming of fibroblastic cells toward the regenerating digit tip(Revardel and Chebouki, 1987; Muller et al., 1999), and by the observation that a regeneration response itself need not include the nail plate (Reginelli et al.,1995).

The expression patterns of the MSX genes at the digit apex represent a simple nested relationship that partitions the digit apex into a distal domain, in which both MSX genes are expressed, and a sub-distal domain, in which only Msx1 is expressed. Digit formation occurs normally in Msx1 and Msx2 mutant mice, indicating that individual MSX gene function is not essential during outgrowth(Satokata and Maas 1994; Satokata et al., 2000). The co-expression of Msx1 and Msx2 in apical mesenchymal cells raises the possibility that these two genes function redundantly during digit development. The expansion of the Msx2 expression domain in Msx1 mutant digit tips suggests a compensatory response by Msx2 in the absence of Msx1, and is consistent with a redundancy hypothesis. This response also suggests that Msx1 plays an inhibitory role in Msx2 expression within the sub-distal cells where only Msx1 is expressed in wild-type digits. However, if MSX1 inhibits Msx2 expression, its activity must be repressed in the distal zone where both genes are co-expressed. Dlx5 is also expressed in the distal digit mesenchyme (Robledo et al.,2002) (M.H., unpublished), and has been shown to compete with and/or repress the activity of MSX genes (see Bendall and Abate-Shen, 2000),thus it represents a candidate for modulating MSX activity in the digit tip.

Our studies indicate that MSX1 functions in a regeneration-specific manner. In the absence of MSX1, digit formation is normal yet digit regeneration is defective, thus MSX1 function is necessary for regeneration but not for development. In addition, Msx2 compensation in the Msx1mutant suggests an incomplete or partial redundancy of function that is restricted to digit regeneration. Studies on cultured cells indicate that one activity of MSX1 involves the control of cell differentiation (see Hu et al., 2001). Based on amphibian limb regeneration studies, a significant regeneration-specific event is the de-differentiation of cells at the wound site to form the blastema. Cellular de-differentiation has been best documented for multinucleated amphibian myotubes that are induced to form individual cells in vitro and,after grafting, in vivo (Lo et al.,1993). Studies using differentiated C2C12 myotubes and regulated Msx1 expression provide evidence that MSX1 induces de-differentiation in vitro, and that subsequent suppression of Msx1 expression can lead to transdifferentiation to multiple cell types, including cartilage, bone,adipose and muscle (Odelberg et al.,2000). Although these de-differentiation studies are specific to muscle tissue, forced Msx1 expression studies demonstrate that multiple mesenchymal and epithelial cell types are inhibited from differentiation in culture, and that differentiation of mammary epithelial tissue is impaired in vivo (Hu et al.,2001). Thus, the available evidence suggests that MSX1 functions in multiple cell types to control differentiation and, in the context of regeneration, de-differentiation. As regenerative potential is restricted to domains of Msx1 expression in the mature and fetal digit tip, we hypothesize that MSX1 functions to maintain a population of undifferentiated mesenchymal cells that can participate in a regeneration response. One significant difference between amphibian limb regeneration and regeneration of the mammalian digit tip is that Msx1 expression is induced in response to amputation in amphibian limbs, whereas, in mammals, regenerative potential is linked to regions where Msx1 expression is maintained in the mature digit.

BMP signaling is crucial for fetal digit regeneration. A number of lines of evidence support this conclusion. First, BMP4 rescues the Msx1 mutant regeneration defect in a dose-dependent manner. Second, Bmp4expression is downregulated in the Msx1 mutant digit, which is consistent with the idea that residual regenerative capability is associated with this reduced level of BMP availability. Third, noggin-treated Msx1 mutant digits display a more severe phenotype associated with the absence of Bmp4 expression. Fourth, noggin treatment of wild-type digits results in a more than 5-fold reduction in regenerative capability. Noggin is not expressed within the forming digit tip, but is expressed in chondrofying skeletal elements at proximal digit levels that lack regenerative potential (Brunet et al.,1998; Capdevila and Johnson,1998; Merino et al.,1998). Other BMP signaling antagonists, such as follistatin,gremlin (Cktsf1b1 - Mouse Genome Informatics) and chordin, are also expressed in developing limbs(D'Souza and Patel, 1999; Merino et al., 1999; Zhang et al., 2002). The expression of multiple BMP signaling antagonists associated with digit formation suggests that the inability of proximal digit amputations to regenerate may be a consequence of regulated BMP activity associated with skeletal differentiation.

The sole function of MSX1 in digit regeneration lies in the regulation of Bmp4 expression. BMP4 rescue of Msx1 mutant digits indicates that MSX1 function is not required downstream of BMP signaling. Successful regeneration of a low percentage of Msx1 mutant digits correlates with a reduced level of Bmp4 expression and provides definitive evidence that successful digit regeneration can occur in the absence of MSX1 function. During digit formation, we have identified Bmp4 as functionally downstream of both Msx1 and Msx2. The shift of the Bmp4 expression domain to coincide with the wild-type Msx2 expression domain is consistent with the hypothesis that Bmp4 is regulated by both Msx1 and Msx2 in the distal digit domain, but by only Msx1 in the sub-distal domain. However, the compensatory response of Msx2 in the Msx1mutant complicates this interpretation and indicates the presence of an Msx2-dependent Bmp4 regulatory component that is co-expressed in the distal digit compartment. It is interesting that in vitro studies of the transcriptional regulator Runx2 (previously known as Cbfa1) indicate that it is regulated by MSX2(Shirakabe et al., 2001) and that it regulates Bmp4 (Helvering et al., 2000), and that we find it co-expressed with Bmp4in the developing digit tip (M.H., unpublished). Thus, Runx2represents a candidate for mediating the regulation of Bmp4expression by MSX genes. Unravelling the details underlying the regulation of BMP4 signaling in the digit tip should provide key insights into the control of fetal digit regeneration.

We thank R. Anderson, S. Bryant, C. Burdsal, Y. Chen, D. Gardiner, L. Ma,M. Omi, S. Schaller and J. Tasker for discussions and comments on the manuscript; R. Maas for the Msx1 and Msx2 mutant lines; L. Ma for Msx1/Msx2 double-mutant limb buds; and the Genetics Institute for providing BMP growth factors. Special thanks to Y. Chen, T. Muller and X. Yu for supporting data. This work was supported in part by HD35245 from the NIH and by CCR419466 from the CDC.