The Msx genes of vertebrates constitute a small family of homeobox-containing genes encoding transcription factors. During mouse development, Msx genes are expressed in a wide range of tissues and in several organs used as models for studying pattern formation and tissue interactions. These include neural crest, branchial arches, cranial sensory placodes, hair follicles, mammary glands, teeth, and limbs (Chen et al., 1996; reviewed in Davidson, 1995; Bendall and Abate-Shen, 2000). However, because of functional redundancy between Msx1 and Msx2, null mutants for either gene display limited developmental defects as compared to the Msx expression patterns (Satokata and Maas, 1994; Satokata et al., 2000). Accordingly, the role of Msx genes in several embryological fields can be studied only in the Msx1null/nullMsx2null/null double homozygous mutant. In particular, limb morphology is profoundly altered only when both Msx1 and Msx2 are mutated (Lallemand et al., 2005). However, as both genes are expressed in the limb mesoderm and ectoderm, the phenotype may result from defects in either germ layer, making it difficult to analyze. Ecto–mesodermal interactions are crucial for limb formation. At early stages (i.e., before 9.75 days post-coitum [dpc] in the mouse embryo), dorsoventral polarity is transferred from mesoderm to ectoderm (Altabef et al., 1997; Michaud et al., 1997) by a process that may involve bone morphogenetic protein (BMP) signaling (Ahn et al., 2001; reviewed in Robert, 2007). This transfer is requested for the formation of the apical ectodermal ridge (AER), a thickened row of ectodermal cells at the interface between the limb bud dorsal and ventral domains. The AER, in turn produces growth factors of the Fibroblast growth factor (Fgf) family that promote limb outgrowth along its proximo–distal axis (from the shoulder or hip to the extremity of the digits) by means of a regulatory loop between AER Fgfs (mainly Fgf8 and Fgf4) and Fgf10 that is produced in the mesenchyme (Xu et al., 1998). Fgf4, Fgf8, Fgf9, and Fgf17 are expressed specifically in the mouse AER. Fgf8 expression is initiated in prospective AER cells of the nascent limb bud and, subsequently, expressed throughout the AER until it regresses (Crossley and Martin, 1995). By contrast, Fgf4, Fgf9, and Fgf17 expression is initiated only after the AER is formed, is restricted to the posterior AER and is down-regulated before AER regression (Sun et al., 2000).
In the Msx1 Msx2 double null mutant, dorsoventral information transfer is altered at the anterior of the limb bud and the anterior third of the AER does not develop, which results in agenesis of the anteriormost part of the limb mesenchyme, including the presumptive territory of digit 1 (thumb or big toe; Lallemand et al., 2005). This precludes further analysis using genetic markers expressed in a polarized manner along the anteroposterior (AP) axis (from thumb to pinky). For example, an anterior specific marker like Pax9 cannot be studied as the region corresponding to its expression domain disappears early in the Msx1 Msx2 double null mutant limb. Conversely, expression of posterior specific markers such as Hand2 or genes of the Hoxd complex (Hoxd11, Hoxd12) appear artificially expanded anteriorly because of the absence of the anterior nonexpressing domain.
For these reasons, producing mutant animals in which Msx activity would be eliminated specifically from limb mesenchyme was a requirement. To achieve this goal, we made use of the Msx2Flox-GFP conditional allele we recently described (Bensoussan et al., 2008) and the Prx1-Cre transgene that drives Cre expression specifically in the limb bud mesenchyme (Logan et al., 2002), in conjunction with an Msx1LacZ null allele (Houzelstein et al., 1997). Analysis of the Prx1-Cre Msx1 Msx2 compound mutants revealed a crucial role for mesodermal expression of Msx1 and Msx2 in specification of digit number and identity along the anteroposterior axis, by means of their involvement in both Shh and BMP signaling pathways.
Generation of Prx1-Cre Msx1 Msx2 Compound Mutants With Specific Elimination of all Msx Activity in The Limb Bud Mesoderm
To investigate Msx function specifically in the limb bud mesoderm, we first combined the conditional floxed allele Msx2Flox-GFP (hereafter termed Msx2Flox) and its deleted form, Msx2null-GFP (hereafter termed Msx2null; Bensoussan et al., 2008), together with the Prx1-Cre deleter transgene (Logan et al., 2002). The latter directs high-level production of Cre in the limb bud mesenchyme, starting as early as 9.5 dpc in the forelimb and 10.5 dpc in the hindlimb. The efficiency of this strategy to specifically inactivate Msx2 in the limb bud mesoderm was evaluated by generating Prx1-Cre Msx2null/Flox embryos and monitoring Msx2 expression by in situ hybridization (ISH). As illustrated (Supp. Fig. S1A,B, which is available online), Msx2 expression in the AER was similar in control (Prx1-Cre Msx2+/Flox) and mutant (Prx1-Cre Msx2null/Flox) embryos at 10.5 dpc (Supp. Fig. S1). On the contrary, no Msx2 transcripts were detectable in the mutant mesenchyme, thus demonstrating the efficiency of this strategy to delete the Msx2Flox conditional mutant allele specifically in the limb bud mesenchyme.
We further combined the mutant Msx1LacZ allele (hereafter termed Msx1null; Houzelstein et al., 1997) with the two Msx2 alleles and the Prx1-Cre transgene described above to generate Prx1-Cre Msx1null/nullMsx2null/Flox embryos. These lack all Msx activity in the limb bud mesoderm but retain some Msx activity (a single Msx2 functional allele) in the AER. The mating protocol used to produce these mutants (referred to from here on as Msx conditional double mutants) is described in the supporting information (Supp. Fig. S1C).
Embryos With a Single Msx2 Functional Allele (Msx1null/null Msx2+/null) Display Minor Abnormalities in the Limbs
The Msx conditional double mutants retain a single Msx2 functional allele in the AER (Supp. Fig. S1C). We have previously shown that hindlimbs from mutants with such a reduced Msx gene dosage in ectoderm and mesoderm (i.e., Msx1null/nullMsx2+/null mutants) display AP abnormalities (Lallemand et al., 2009). To avoid any misinterpretation of the Msx conditional double mutant phenotype, we carefully reinvestigated the limb phenotype of the Msx1null/nullMsx2+/null compound mutants. In accordance with our previous results (Lallemand et al., 2009), hindlimbs displayed a slight anterior overgrowth of the mesenchyme at 11.75 dpc that became prominent at 12.5 dpc, whereas no such deformation was observed in forelimbs (Fig. 1A–D′). At birth, digit 1 in the mutant hindlimb bore three phalanges instead of two in approximately 70% of the animals. In the forelimb, the size and the number of phalanges of digit 1 remained normal. However, the falciform bone was fused with the thumb (Fig. 1I,I′).
In keeping with these phenotypic observations, genes normally expressed posteriorly, such as Hoxd genes, systematically displayed anteriorized expression in the hindlimbs, but not forelimbs, of the Msx1null/nullMsx2+/null embryos (Fig. 1A–D′). Similarly, genes expressed anteriorly (Pax9, Dlx5) were down-regulated in the presumptive digit 1 domain of the mutant hindlimbs, but not in the corresponding region of the forelimbs where Dlx5 was unaffected and Pax9 only slightly down-regulated (Fig. 1E–H′). In contrast, expression of Shh and Gli3, as well as Gli1 and Ptc1, two direct targets of Shh signaling, appeared normal in the hindlimbs of these compound mutants (Fig. 1J–L′ and data not shown). In conclusion, reducing Msx gene dosage to a single Msx2 functional allele leads to impairment of the AP polarity in the hindlimb, in which the anteriormost region partially loses its specific identity, whereas malformations of the forelimbs are limited to a partial fusion of the falciform bone with digit 1, with no concomitant detectable alteration in gene expression. These defects must be taken into consideration when analyzing the phenotype of the Prx1-Cre Msx1null/nullMsx2nul/Flox conditional double mutants.
AER Development Is Not Impaired in the Prx1-Cre Msx1null/null Msx2null/Flox Mutants
In the double Msxnull/nullMsx2null/null mutants, the AER does not develop anteriorly during the early steps of limb formation, which results in agenesis of the underlying mesoderm and anterior truncation of the limb. To assess AER formation in the Msx conditional double mutant embryos, we used Fgf8 as a read-out of AER activity and morphology (Fig. 2A). Because of specific properties of the Prx1-Cre transgene (see below), we focused our analysis to the forelimbs. At 11.5 dpc, in the double Msxnull/nullMsx2null/null mutant, the anterior third of the AER was missing (Fig. 2C), as previously described (Lallemand et al., 2005). On the contrary, in the Msx conditional double mutant, the AER remained intact and was even slightly expanded at the anteriormost part of the forelimb bud (Fig. 2B). This expansion is unlikely to be solely related to the diminution of Msx activity in the AER, because it was not observed in the AER of the compound Msx1null/nullMsx2+/null mutants (Fig. 2D), but suggests a role of Msx in the mesenchyme for AER formation.
The Prx1-Cre Msx1null/null Msx2null/Flox mutants display abnormal digit formation and preaxial polydactyly
Contrary to the double Msx1null/nullMsx2null/null mutants, the Msx conditional double mutants were viable until birth, allowing skeleton preparations to be performed at 18.5 dpc (Fig. 3). At this stage, the three limb segments, i.e., the stylopod (arm or thigh), zeugopod (forearm or leg), and autopod (hand or foot), were present in both fore- and hindlimbs (Fig. 3A–D), but several abnormalities were noticeable along both the proximodistal (PD) and AP axes. Along the PD axis, abnormalities were similar in fore- and hindlimbs. They were limited to the autopod where the distal phalanges (the last one and occasionally the penultimate) were absent or incompletely developed, and devoid of calcification (Fig. 3E–H). This was confirmed by the absence of nails (normally borne by the last phalanx) in the mutants (Fig. 3I–L). This reduction of digit extremities was similarly observed in the double Msx1null/nullMsx2null/null mutant (Lallemand et al., 2005).
Along the AP axis, the phenotype was much more striking in fore- than in hindlimbs. In the forelimbs, the AP defects involved the three segments (Fig. 3A,C). In the stylopod, we observed a systematic absence of the deltoid tuberosity. In the zeugopod, the anterior bone (radius) was absent in half of the specimens analyzed. In the other half, radius phenotypes ranged from a reduction in width to shortening and bending or truncation (Fig. 3C and Supp. Fig. S2A). In the autopod, preaxial polydactyly was systematic, with up to eight digits or digit-like elements (Fig. 3F and Supp. Fig. S2B). Based on their relationships to the underlying carpals, the four posteriormost digits could be considered as digits 2 to 5. Anterior to digit 2, we systematically observed a cartilaginous structure that we interpret as the nonossified metacarpal of the misdeveloped thumb. This structure was unique in half of the cases and duplicated in the other half. Finally, the anteriormost digit displayed a morphology similar to digit 2, 3, or 4. In a minority of cases (16%) two such digits were observed (Fig. 3F and Supp. Fig. S2B).
The phenotype of the hindlimbs was more discrete. The stylopod was always normal, as was the zeugopod in two thirds of the specimens analyzed (Fig. 3D and Supp. Fig. S2C). In the remaining third, the tibia was shortened or, in a few cases, absent (Supp. Fig. S2C). The AP abnormalities of the hindlimb autopod were also very mild. Because of the reduced development of the terminal phalanges in all digits, no three-phalanged big toes (halluces) could be identified, in contrast to what is seen in the Msx1null/nullMsx2+/null compound mutants (Lallemand et al., 2009). In addition, in more than 50% of the cases, the autopod was strictly pentadactylous (Supp. Fig. S2D). In the other specimens analyzed, the only abnormality observed along the AP axis was the anterior presence of a small cartilaginous formation, either independent of or branched on digit 1 (Fig. 3H and Supp. Fig. S2D). Considering this, and the fact that the three-phalanged toe of the Msx1null/nullMsx2+/null compound mutants can be considered as the mildest manifestation of an anterior polydactyly, the hindlimb phenotype of the conditional double mutants can be viewed as a slight aggravation of the hindlimb phenotype of the Msx1null/nullMsx2+/null compound mutants.
The differences in skeletal malformations seen in the double conditional mutant fore- and hindlimbs may result in part from the properties of the Prx1-Cre transgene. This transgene is known to be expressed later and to a lesser extent in the hindlimb than in the forelimb (Logan et al., 2002; see also the Discussion section). Due to these technical limitations, and because our conditional knockout strategy aimed primarily to decipher the role of Msx genes in the anterior region of the limb, we focused the rest of our study on the analysis of the AP phenotype in the forelimb.
Alteration of the Anteroposterior Polarity in the Forelimbs of the Msx Conditional Double Mutants
The phenotype observed in the autopod of the conditional double mutant forelimb can be viewed as a partial mirror-image duplication of the digits. This was confirmed by analysis of several genetic markers expressed in a polarized pattern along the AP axis. At 11.5 dpc and 12.5 dpc, Hoxd12 is normally expressed exclusively in a posterior domain of the autopod, corresponding to presumptive digits 2 to 5 (Fig. 4A,C). In the conditional mutant forelimb, this posterior domain was not modified but Hoxd12 was further expressed in an ectopic domain at the anterior margin of the deformed autopod (Fig. 4B,D). The two domains were separated by a Hoxd12 nonexpressing one. The ectopic anterior expression domain is likely to correspond to the anteriormost extra-digits with a posterior morphology, whereas the negative region would match with the presumptive territory of the thumb-like structures observed in the mutant newborns (compare Fig. 4B with Fig. 3F). Similar results were observed for Hoxd11 (Fig. 4E,F) as well as for Hand2 (Fig. 4G,H), another posterior-specific marker. It should be noted that such anterior ectopic expression domains were not observed in the hindlimbs of conditional mutants where all these genes showed only an anterior extension in expression, similar to what was observed in the Msx1null/nullMsx2+/null compound mutant (data not shown).
In wild-type embryos, Pax9 is expressed at 11.5 dpc in a domain corresponding to the presumptive digit 1 (Neubuser et al., 1995). In the forelimb of the conditional double mutant, Pax9 was dramatically down-regulated but remained expressed at a low level in a small, proximal region (Fig. 4I,J). This region corresponds to the Hoxd12 negative domain that likely gives rise to thumb-like structures in the newborn autopod.
Altogether, gene expression results confirm that the Msx conditional double mutants display a partial anterior mirror-image duplication of the forelimb digit territory. The anteriormost digits display posterior features whereas the intermediate structures, generally limited to metacarpals, are likely to be thumb-like formations.
In the Msx Conditional Double Mutant, Shh Is Ectopically Expressed in the Anteriormost Region of the Limb Bud Mesenchyme, Despite Etv4/Etv5 Up-regulation
In the Msx conditional double mutant, gene expression abnormalities as well as anterior polydactyly are reminiscent of phenotypes due to Sonic hedgehog (Shh) ectopic expression anteriorly (Sharpe et al., 1999; Lettice et al., 2003; Sagai et al., 2004; Mao et al., 2009; Zhang et al., 2009). In normal embryos, Shh expression is restricted to a small region of the posterior mesenchyme corresponding to the zone of polarizing activity (ZPA; Fig. 5A). At 11.5 dpc, we observed a Shh ectopic expression at the anterior of the forelimb buds of mutant embryos (5/5), albeit at a low level (Fig. 5B). In addition, ectopic activation of Gli1, a direct target of Shh, was detected at the anterior margin of the bud mesenchyme, confirming functional Shh signaling (Fig. 5C,D).
Etv4/Etv5 double and Msx conditional double mutants share a strikingly similar limb phenotype, with anterior polydactyly and formation of an ectopic expression domain for posterior markers such as Hand2 and Gli1. It has been demonstrated that this phenotype is due to ectopic activation of Shh anteriorly (Mao et al., 2009; Zhang et al., 2009). We, therefore, investigated a possible down-regulation of the Etvs in the Msx conditional double mutant that might explain its phenotype. ISH actually demonstrated the reverse. Etv4 and Etv5 expression was not modified at 10.75 dpc (Supp. Fig. S3) and, at 11.5 dpc, the two genes were clearly up-regulated anteriorly (Fig. 5E–H) where Shh ectopic expression takes place. Both Etv4 and Etv5 are positively controlled by FGF signaling, and their anterior up-regulation is consistent with the anterior expansion of the Fgf8 expression domains in the Msx double conditional mutants (Fig. 2B). But even at high expression levels, Etv4 and Etv5 cannot repress ectopic expression of Shh in the Msx mutant context (see the Discussion section).
Global Signaling Activity Modification in the Anteriormost Region of the Limb Bud Mesenchyme of the Msx Conditional Double Mutant
The Msx conditional double mutant phenotype displays features different from those associated with Shh misexpression, such as observed in the Hemimelic extra-toes (ShhHx) mutant. In the latter, a point mutation in the ZPA regulatory sequence (ZRS), a long-range regulatory sequence of Shh (Lettice et al., 2003; Sagai et al., 2004) leads to the formation of an ectopic anterior Shh expression domain, resulting in a mirror-image duplication of the digits in all four limbs, and a reduction or agenesis of the anterior bone of the zeugopod (radius, tibia). The latter is more severe in the hindlimbs than in the forelimbs (Knudsen and Kochhar, 1981), whereas the reverse situation is observed in the Msx conditional double mutant. In addition to these phenotypic differences, defects in AP polarity do not seem to be brought about the same way. At 10.75 dpc (40–42 pairs of somites), Shh expression could not be detected by ISH in the anterior region of the limb mesenchyme in either of the two mutants (Fig. 6A–C). Similarly Gli1, Hand2, or Hoxd12 were expressed only in their normal posterior domain (Fig. 6D,E, and data not shown). Nevertheless, ectopic Fgf4 expression in the AER indirectly demonstrated anterior polarizing activity in the Msx conditional double mutants. At 10.75 dpc, in normal embryos, Fgf4 is expressed in the posterior aspect of the AER, but not in its anterior third (Fig. 6A,D), as its expression depends on Shh signaling from the posterior mesenchyme. At this stage, Fgf4 expression appeared normal in ShhHx/+ mutant forelimbs (Fig. 6C), whereas conspicuous, ectopic expression was observed at the anterior side of the AER in the Msx conditional double mutant (Fig. 6B,E). This difference could not be explained by minor variations in embryo development, as ectopic Fgf4 expression was also detected in slightly delayed Msx mutant embryos (compare Fig. 6B and 6C). Likewise, it was not solely due to the reduced Msx activity in the AER of the Msx conditional double mutant because Fgf4 expression in the Msx1null/nullMsx2+/null compound mutant forelimb showed no difference as compared to control at 10.75 dpc (Fig. 6F).
It should be considered that ISH is a poorly quantitative technique. Thus, at 10.75 dpc, although not revealed by ISH, Shh mRNA might already accumulate anteriorly, and at a higher level in the Msx double mutants than in the ShhHx/+ mutants. This would induce Fgf4 expression earlier in the former than in the latter. To assess this point, we dissected the anterior part of limb buds from ShhHx/+ and Msx conditional double mutant embryos at 10.75 dpc, and analyzed the Shh expression by quantitative reverse transcriptase-polymerase chain reaction (qRT-PCR). In both mutants, the level of Shh RNA in the anterior limb mesenchyme was abnormally high compared with their wild-type littermates (10 to 15 times higher; Supp. Fig. S4). Using the limb posterior region of wild-type embryos as a reference, we estimated the anterior/posterior Shh RNA level ratio to be around 0.1% in normal embryos (negative controls) whereas it was around 1 to 2 % in the ShhHx/+ or the Msx conditional double mutants (data not shown). Nevertheless, this level was not significantly different between ShhHx/+ and Msx conditional double mutants at this stage (Supp. Fig. S4), thus excluding earlier activation of Fgf4 by Shh in the latter.
BMP signaling has long been recognized to negatively regulate Fgf4 expression in the AER (Pizette and Niswander, 1999; Zuniga et al., 1999; reviewed in Dahn and Fallon, 1999), and Msx genes are involved in BMP signaling at several sites during development and may even regulate Bmp4 expression (Chen et al., 1996; Marazzi et al., 1997; Bei and Maas, 1998; Zhang et al., 2002; Ogawa et al., 2006). We thus analyzed the three Bmp genes expressed in limb development (Bmp2, Bmp4, Bmp7) as well as the BMP antagonist gene Gremlin, which has been shown to play a crucial role during limb development (Khokha et al., 2003). At 11.5 dpc, expression of these four genes was modified at the anterior of the forelimb bud, but in different ways: Bmp4 was down-regulated, whereas Bmp7 was moderately up-regulated and Bmp2 and Gremlin ectopically expressed (Fig. 7). In normal embryos, at 11.5 dpc, Bmp4 is expressed in the mesenchyme underlying the AER, with a stronger level of expression posteriorly (Fig. 7A). In the Msx conditional double mutant, only the posterior domain of expression remained whereas the anterior region of the limb mesenchyme was almost entirely devoid of Bmp4 transcripts (Fig. 7B). This down-regulation cannot be attributed to the anterior ectopic expression of Shh because Bmp4 is not down-regulated in the ShhHx/+ mutant at the same stage (Fig. 7C). Bmp2 (Fig. 7D), and Gremlin (Fig. 7J) are normally not expressed in the anteriormost part of the limb bud at 11.5 dpc. In the Msx mutant, a tiny spot of Bmp2 expression was visible at the anterior tip of the AER (Fig. 7E), whereas, for Gremlin, the anterior ectopic expression was larger and located in the mesenchyme (Fig. 7K). For Bmp7, it is normally expressed in the whole limb bud mesenchyme at 11.5 dpc, but more intensively in the posteriormost region than in the anteriormost one (Fig. 7G). In the Msx conditional double mutant, Bmp7 expression was slightly up-regulated anteriorly so that the level of staining was similar at the anterior and posterior aspects of the limb bud (Fig. 7H). For all three genes (Bmp2, Bmp7, and Gremlin), the modification was similar in the Msx conditional double mutant and the ShhHx/+ mutant (compare Fig. 7E and 7F, 7H and 7I, 7K and 7L). Contrary to what was observed for Bmp4, the modifications in expression of these genes seem to be a consequence of the Shh anterior ectopic expression.
At 10.75 dpc, the situation was different. Bmp2 and Bmp4 were similarly expressed in control and Msx mutant embryos (Supp. Fig. S3E–H). Bmp7 was slightly up-regulated anteriorly compared with control embryos, but this difference was also observed in stage-matched ShhHx/+ mutants (Fig. 7M–O). For Gremlin, a clear anterior ectopic expression domain was observed but, as for Bmp7, a similar result was obtained with the ShhHx/+ mutant (Fig. 7P–R). Thus, at 10.75 dpc, Bmp7 anterior up-regulation, and Gremlin ectopic expression, must be due, in both kinds of mutants, to the low level of Shh ectopically present in the anteriormost region of the limb bud mesenchyme.
In conclusion, with regard to the BMP signaling pathway, the only noticeable difference between Msx conditional double mutant and ShhHx/+ mutant embryos is the anterior down-regulation of Bmp4 expression at 11.5 dpc (see the Discussion section). This can explain the difference of phenotype between the two types of mutants but not the anterior ectopic expression of Fgf4, which is visible, at 10.75 dpc, only in the AER of the Msx conditional double mutant (Fig. 6B,E).
Apoptosis and Cell Proliferation
To assess the respective role of apoptosis and cell proliferation in the anterior overgrowth of the mutant limb bud, we analyzed them by immunofluorescence on limb bud sections at 11.5 dpc (Fig. 8). Analysis of active Caspase-3 revealed that the region of intense apoptotic activity, that is normally observed at the anterior aspect of the limb bud of wild-type embryos (Fernández-Terán et al., 2006; Fig. 8A), was lost in the Msx conditional double mutant (Fig. 8B). Conversely, the central domain of apoptosis was present in both kind of embryos. This is similar to what was observed in the hindlimbs of the Msx1null/nullMsx2+/null compound mutant (Lallemand et al., 2009).
Analysis of cell proliferation on the same specimens, using an anti–phospho-histone H3 antibody, did not reveal differences between control (Fig. 8C) and Msx conditional double mutant embryos (Fig. 8D). Nevertheless, during this period of development, cell proliferation is so intense in the limb mesenchyme that small differences in the growth rate are difficult to assess (see also the Discussion section).
In this report, we describe the consequences of mesenchyme-specific abrogation of all Msx activity in the limb bud. Our analysis is focused on the AP development of the mutant limb and more specifically of the most anterior digit (digit 1) territory, because this region is profoundly altered in the double Msx1 Msx2 null mutant. Our results show that Msx activity in the limb bud mesenchyme is critical for the proper development of the anteriormost region and reveal new roles for Msx genes in two major signaling pathways, namely Shh and BMP.
AER Development Is Complete in the Prx1-Cre Msx1null/null Msx2null/Flox Conditional Mutant
The territory corresponding to the most anterior digit is lost early in the double Msx1 Msx2 null mutant, following failure of the AER to form anteriorly (Lallemand et al., 2005). In contrast, the AER develops along its entire length in the Msx conditional double mutant and, consequently, the anteriormost region of the mesenchyme does not regress. This may indicate that Msx genes act cell-autonomously in the ectoderm at early stages of limb development to promote AER formation anteriorly. However, alternative hypothesis may be considered, especially in the light of data showing that abrogating BMP signaling in the AER abolishes Msx2 AER-specific expression without altering AER integrity (Wang et al., 2004; Maatouk et al., 2009). Although Msx1 expression was not investigated in these studies, this may suggest that Msx genes are dispensable in the ectoderm for formation of the AER. AER forms at the boundary between dorsal and ventral domains in the ectoderm. Initially, DV polarity is established in the mesoderm, then transferred to the ectoderm (Altabef et al., 1997; Michaud et al., 1997). This transfer likely involves the BMP pathway (Ahn et al., 2001; for a discussion, see Robert, 2007). It should be noted that the Prx1-Cre transgene activity takes place progressively in both fore- and hindlimbs (Logan et al., 2002), and that, at the earliest stages of limb outgrowth, some Msx activity is likely to remain in the mesenchyme of our conditional mutant. A recent study (Zhang et al., 2010) has shown that for Twist1, a gene specifically expressed in the limb bud mesoderm, the Prx1-Cre-induced null mutation leads to a less dramatic phenotype than the constitutive null mutation. In particular, AER development is complete in the conditional null mutant and impaired in the constitutive one. This is likely due to residual Twist1 activity at 9.5 dpc in the mesoderm. We therefore cannot exclude that the anterior truncation of the AER observed in the Msx1 Msx2 double null mutant is due to a mesenchyme-bound, non–-cell-autonomous role for Msx in DV information transfer, in keeping with frequent Msx involvement in the BMP pathway. In the Msx conditional double mutant, Prx1-Cre expression would take place too late in the mesoderm to interfere with this process.
Phenotypic Alterations in the Mesenchyme-Specific Msx Mutant
Skeleton analysis of the conditional mutants clearly revealed different phenotypes between fore- and hindlimbs along the AP axis but a similar one along the PD axis. This can be explained by the expression properties of the Prx1-Cre transgene, which is active earlier in fore- than hindlimbs (Logan et al., 2002). This difference is particularly obvious before 10.5 dpc. Later on, the recombination of floxed alleles appears almost complete in both fore- and hindlimbs. Complete inactivation of the Msx2 floxed allele might therefore occur too late (after 10.5 dpc) in the hindlimb to drastically alter its development along the AP axis. Thus the AP abnormalities observed in the hindlimbs may be due primarily to the reduction of Msx activity in limb bud ectoderm and mesoderm, as in the Msx1null/nullMsx2+/null compound mutant, slightly enhanced by the progressive loss of all Msx activity in the mesenchyme. The loss of Msx activity is likely to be more complete in the forelimb because of the earlier and more extensive activity of the Prx1-Cre transgene.
Conversely, the formation of the last phalanges is a late event in limb morphogenesis (Sanz-Ezquerro and Tickle, 2003). At this stage, the Cre-induced inactivation of the Msx2 conditional allele is likely to be complete in both fore- and hindlimbs, leading to similar phenotypes in both. Nevertheless, several mutations have been reported that lead to different AP phenotypes in fore- and hindlimbs (e.g., ShhHx/+: Knudsen and Kochhar, 1981; Blanc et al., 2002; Shh−/−: Kraus et al., 2001; Bmp4+/−: Dunn et al., 1997), including the double Msx1 Msx2 null mutation (Lallemand et al., 2005). It thus remains possible that the phenotypic variations observed between fore- and hindlimb buds are due to slightly different roles for Msx genes in either of these.
The most dramatic outcome of the mesenchyme-specific Msx mutation is an anterior outgrowth of the mesoderm that results in anterior polydactyly. This raises the question of which changes in local cell physiology are underlying this morphological alteration. We observed that the apoptotic domain that normally forms anteriorly (Fernández-Terán et al., 2006) is missing in our mutant. This is in keeping with similar absence of this domain that we observed in the hindlimb of a hypomorphic Msx1null/nullMsx2+/null mutant (Lallemand et al., 2009). Conversely, we did not observe changes in cell proliferation that similarly might underlie outgrowth. However, contrary to apoptosis, which is restricted in time and space in the developing limb bud, proliferation is active in the whole limb bud at the stage of development we analyzed and lasts over several days, such that subtle changes, not detectable in our analysis, might be cumulative. It is highly plausible that proliferation plays a role in anterior polydactyly, but, to establish it, it would be necessary to conduct a systematic analysis of cell proliferation in three-dimensions at different stages of development, the kind of which is just starting to be accessible (Boehm et al., 2010; Gros et al., 2010; Wyngaarden et al., 2010).
Control of Shh Expression
The most striking result of this study is the demonstration of a role for Msx genes in the control of Shh repression anteriorly. This was suggested by our previous work on the Msx1null/nullMsx2null/null double mutants, but Shh ectopic expression in these mutants was inconsistent (Lallemand et al., 2005). On the contrary, we could detect ectopic Shh expression in the forelimbs of all Msx conditional double mutant embryos we analyzed at 11.5 dpc (5/5). This abnormal expression of Shh was confirmed by qRT-PCR at 10.75 dpc. It has been shown that, in the limb, positive and negative control of Shh transcription depends on a long-range regulatory sequence referred to as the ZPA regulatory sequence (ZRS; Lettice et al., 2003; Sagai et al., 2004). Preliminary experiments to investigate potential binding of Msx proteins to the ZRS, that would suggest direct control of Shh expression, did not give conclusive results (data not shown). This does not exclude potential Msx interaction with the ZRS, considering that this interaction may involve several proteins engaged in a complex. In this respect, it should be noted that Gli3R, the truncated form of the Gli3 protein, accumulates in the anteriormost region of the limb, where Msx activity is high, and that Gli3−/− mutant embryos also display ectopic Shh expression anteriorly in the limbs (Buscher et al., 1997). Furthermore, Gli3R has been shown to interfere with positive transcriptional regulation at the ZRS (Galli et al., 2010). Interactions between Gli3R and Msx might be required for Shh repression anteriorly. Alternatively, Shh may be activated by ectopic expression of Hand2 in the Msx double conditional mutant. To date, we could not establish a temporal hierarchy in the activation of Hand2 vs. Shh.
Etv4 and Etv5 are two transcription factors that also repress Shh anteriorly in the limb, and the phenotype of Etv4/Etv5 double null mutants strikingly resemble this of Msx conditional double mutants (Mao et al., 2009; Zhang et al., 2009), which might suggest epistatic relations between Etv and Msx. This however is unlikely, as Msx1 remains expressed in Etv4 Etv5 double null mutant limb buds (Xin Sun, personal communication). Conversely, Etv4 and Etv5 are not down-regulated in our Msx conditional mutant. On the contrary, their expression is up-regulated and expanded in the anterior limb mesenchyme, in accordance with the increase in FGF signaling from the AER. It should be noticed that, in the Twist1 null mutant, Etv4 and Etv5 are also up-regulated in the anterior limb bud mesenchyme, which does not preclude Shh derepression anteriorly. Indeed, it has been recently demonstrated that Etv and Twist1 proteins form a complex that is required to repress Shh anteriorly (Zhang et al., 2010). The similarity between these and our results may suggest a similar mechanism for anterior repression of Shh by Msx.
BMP Signaling and Control of Bmp4 Expression
Several studies have focused on the relationships between BMP signaling and the Msx gene family. These relationships appear to be complex and context-dependent, with BMP or Msx acting either upstream or downstream of each other depending on the organ and/or the process analyzed (Chen et al., 1996; Marazzi et al., 1997; Bei and Maas, 1998; Zhang et al., 2002). In the limb, most of the results published describe Msx1 and Msx2 as BMP signaling targets. Concerning Msx2, inactivation of a Bmp4 conditional mutant allele in the limb mesenchyme, using Prx1-Cre, abolishes its expression in the overlying ventral limb ectoderm (Selever et al., 2004). A similar result was observed by generating an ectoderm-specific knockout of the Bmpr1a that prevents ectodermal cells to respond to BMP signaling (Ahn et al., 2001). Reciprocally, AER-specific null mutation of Bmp2 and Bmp4 provokes Msx2 down-regulation in the underlying mesenchyme (Maatouk et al., 2009). Furthermore, a BMP-responsive enhancer, which is sufficient for Msx2-like expression of a reporter in the limb bud, has been identified at the 5′ end of the Msx2 first exon (Brugger et al., 2004).
Conversely, both Msx1 and Msx2 are up-regulated in the limb mesenchyme by a null mutation of Gremlin, a potent BMP inhibitor expressed in the limb (Khokha et al., 2003; Michos et al., 2004). Nevertheless, although a mesenchyme-specific null mutation of Bmpr1a leads to a drastic down-regulation of the two Msx genes (Ovchinnikov et al., 2006), residual expression is still detectable for both Msx1 and Msx2 anteriorly, raising the possibility that, in this region, Msx genes are controlled by other factors such as, for example, Gli3R (Lallemand et al., 2009). Indeed, it has been shown that the Talpid2 chick mutation, which precludes proper processing of Gli3 into Gli3R (Wang et al., 2000) also abolishes Msx2 expression and leads to nonpolarized Msx1 expression in the mesoderm (Krabbenhoft and Fallon, 1992). Most of the results quoted here, however, suggest that Msx might work as downstream effectors in the BMP pathway for most of the limb mesenchyme, and indeed, the Fgf4 expression domain in the AER extends anteriorly in a similar way in the Msx conditional double mutant and the Bmp4 conditional null mutant (Selever et al., 2004; this work). This suggests that BMP signaling is impaired in the Msx conditional double mutant.
However, the phenotype of the Msx conditional double mutant is quite different from those of the mesenchyme-specific Bmp4 and Bmpr1a null mutants (Selever et al., 2004; Ovchinnikov et al., 2006). In either of these mutants, the hindlimbs are more affected than the forelimbs, when the mutation is induced using the Prx1-Cre transgene. Bmp4 mutants display pre- and postaxial polydactyly and no alteration of the zeugopod. In the Bmpr1a mutant, there is a severe dysplasia of the autopod in fore- and hindlimb and, furthermore, frequent disappearance of the fibula in the hindlimb. In addition, in these two mutants, patterning along the AP axis is not affected to such an extent as in the Msx conditional double mutant. Hoxd11, Hoxd12, or Hoxd13 expression domains extend only slightly anteriorly, and no ectopic Shh expression is detected. Therefore, in the limb mesenchyme, Msx genes cannot be considered exclusively as downstream effectors in the BMP signaling pathway.
The possible control of Bmp genes by Msx activity in the limb is less documented. It has been shown that forced expression of Msx2 in the mesoderm of chick limb bud induces Bmp4 misexpression (Ferrari et al., 1998). The results described in this report show that, in the anterior region, Msx activity is not required for initial Bmp4 expression, but rather for its maintenance from 11.5 dpc on. This corroborates previous studies showing that Msx1 interacts with Pax9 to synergistically transactivate, in vitro, a Bmp4 promoter sequence (Ogawa et al., 2006). Indeed, genetic interactions between Msx1 and Pax9 were recently demonstrated to regulate morphogenesis of the tooth, where the two genes are expressed (Nakatomi et al., 2010). Furthermore, in this organ, defects induced by Msx1 Pax9 double heterozygous null mutations could be partially rescued by a Bmp4 transgene. Pax9, Msx1, and Msx2 are also coexpressed in the presumptive territory of the thumb at 11.5 dpc (Neubuser et al., 1995), and a Pax9 null mutation, similarly to Msx gene conditional mutation, leads to preaxial digit duplication (Peters et al., 1998). Our results confirm that, in the limb, Msx activity controls Pax9 expression, as we previously showed (Lallemand et al., 2009), and second, suggest that in this field too, Msx and Pax9 proteins might interact to control Bmp4 expression.
Contrary to Bmp4, Bmp2, and Bmp7, the other two Bmp genes expressed in the limb bud, were up-regulated anteriorly in the Msx conditional double mutant. However, in both cases, these modifications were very modest and, considering the concomitant anterior ectopic expression of the BMP-antagonist Gremlin, it is likely that, in the anteriormost limb region, BMP signaling is globally diminished in our Msx mutant. Moreover, all the expression differences observed, except the Bmp4 down-regulation, are similar in the Msx conditional double mutant and the ShhHx/+ mutant embryos at both 10.75 and 11.5 dpc. Thus the anterior down-regulation of Bmp4 expression at 11.5 dpc is the only process that can account for the difference of phenotype between the two kinds of mutants. Nevertheless, it does not explain the first manifestations of the abnormal phenotype in the Msx conditional double mutant, namely the anterior ectopic expression of Fgf4 in the AER, which could be detected as early as 10.75 dpc.
The AP limb phenotype of the Prx1-Cre Msx1null/nullMsx2null/Flox mutants can be explained in part by the anterior de-repression of Shh, but also by the down-regulation, in the same region, of Bmp4. The fact that Msx proteins can play a role in two different signaling pathways (Shh and BMP) in two opposite ways (repression or activation) suggests they cooperate with different protein partners. Pax9 is a likely candidate for the control of Bmp4 expression. Concerning Shh repression, Gli3R and Etv are possible candidates. Further investigation at the protein level will be necessary to clarify this issue.
Mice and Embryos
Generation of the Msx1 mutant and the Msx2 conditional mutant alleles has been described previously (Houzelstein et al., 1997; Bensoussan et al., 2008). The Prx1-Cre transgene (a kind gift of Malcolm Logan) was introduced to the Msx1+/nullMsx2+/null background. Subsequently, the combined Prx1-Cre Msx1+/nullMsx2+/null strain was maintained on an outbred (NMRI) background. Conditional mutants were obtained by crossing Prx1-Cre Msx+/nullMsx2+/null males with Msx1+/nullMsx2Flox/Flox females. Control embryos was chosen among the littermates of the conditional mutants possessing at least one wild-type allele of each Msx gene (Prx1-Cre Msx1+/+Msx2+/Flox or Prx1-Cre Msx1+/nullMsx2+/Flox genotypes). The Prx1-Cre transgene is carried by the male because of its expression in the female germline (Logan et al., 2002). Day of the plug was considered as 0.5 dpc.
Whole-Mount In Situ Hybridization
Whole-mount in situ hybridization (ISH) was performed as described previously (Houzelstein et al., 1997). RNA probes were generated by transcription of DNA templates using Sp6, T3, or T7 RNA polymerase. The following DNA templates were prepared by cDNA PCR amplification followed by cloning into the pGEM-T vector (Promega): Fgf8 (complete cDNA sequence), Msx2 (complete exon 2 sequence) and Gli1 (complete exon 13 sequence). Bmp2, Bmp4, and Bmp7 were a gift from B. Hogan, Etv4 and Etv5 from X. Sun, Fgf4 from G. Martin, Gremlin from R. Zeller, Hoxd11 and Hoxd12 from P. Dollé, Pax9 from R. Balling, Ptc1 and Shh from A. McMahon, Gli3 from U. Rüther, Dlx5 from D. Acampora and Hand2 from E. Olson.
After killing, newborns were eviscerated and their bodies placed in water at 70°C for 50 min to permit complete removal of the skin. Embryos were stained overnight in Alizarin red/Alcian blue solution, then fixed 1 hr in 95% ethanol, cleared in 1% NaOH, and stored in phosphate buffered saline at 4°C.
Cell Death and Cell Proliferation Analysis
Cell death and cell proliferation were analyzed on frozen histological sections (14 μm) using, respectively, an anti-active Caspase-3 (BD Pharmigen) and an anti-phospho-histone H3 (Cell Signaling) according to the manufacturer's protocol.
Anterior and posterior fragments of limb buds were dissected separately from 10.75 dpc embryos. For Msx1 Msx2 conditional double mutants, seven mutant and seven control littermates were used; for Hx mutants, seven mutant and five control littermates were used. Total RNA was isolated using an RNeasy Mini Kit (Qiagen) following the manufacturer's instructions. cDNA were synthesized from 100 ng of RNA using the SuperScript III reverse transcriptase (Invitrogen) and RT-PCR was carried out using the SYBRGreen PCR master mix (Applied Biosystems) and an Applied Biosystems Step One plus cycler, according to the manufacturer's instructions. PCR cycle parameters were as follows: 10 min at 95°C (initial incubation), followed by 15 sec at 95°C, 1 min at 60°C for 4 0 cycles. Posterior fragments from wild-type embryos were used as positive controls to set up the baseline for normal Shh expression in our qRT-PCR conditions. The Gapdh gene was used to normalize results. Primer sequences were as follows: qSHH-for: TGCTGGCTCGCCTGGC TG; qSHH-rev: AAACAGCCGCCGG ATTTGGC; qGAPDH-for: GGCAAAGT GGAGATTGTTGC; qGAPDH-rev: AA TTTGCCGTGAGTGGAGTC. PCR efficiency was in the range of 98% to 100% for all assays.
We thank Dr. X. Sun for fruitful discussions and sharing of unpublished results, and Drs. C. Christ and M. Lopes for critical reading of the manuscript. This work was supported by the Institut Pasteur, the Centre National de la Recherche Scientifique, and the Agence Nationale de la Recherche. V.B-T. was the recipient of a fellowship from the Institut Weizmann des Sciences France-Europe.