Retinoic acid (RA) plays a role in cell–cell signaling during organogenesis by functioning as a ligand for nuclear RA receptors (RAR) that control transcription of key genes (Duester, 2008). Recent studies indicate that RA is essential for induction of forelimb buds, but that RA is unnecessary for both hindlimb induction and establishment of anteroposterior or proximodistal patterning of limb buds (Zhao et al., 2009). However, at later stages of limb development when the autopod begins to form digits, a role for RA signaling in interdigital morphogenesis has been proposed based on studies of mouse RARb/RARg double null embryos that exhibit interdigital webbing at embryonic day (E) 15.5 that would normally have been lost by this stage (Ghyselinck et al., 1997). Further studies on autopods of RARb/RARg double mutants revealed a loss of apoptosis in the interdigital necrotic zone as well as down-regulation of tissue transglutaminase (Tgm2; tTG), a cross-linking enzyme implicated in apoptosis, and down-regulation of matrix metalloproteinase-11 (Mmp11; stromelysin-3) (Dupé et al., 1999). These findings suggest that RA is required for cell death in the interdigital mesenchyme to prevent interdigital webbing.
Digit formation in mouse forelimbs and hindlimbs is also associated with complementary expression of two high mobility group (HMG) domain transcription factors encoded by Sox9, expressed in digit cartilage, and Hmgn1, expressed at the digit–interdigit boundary between E12.5 and E14.5 (Furusawa et al., 2006). Analysis of null mutants in mice has demonstrated that Sox9 is essential for cartilage formation (Bi et al., 1999). Analysis of Hmgn1−/− limb bud micromass cultures has shown that HMGN1 inhibits Sox9 expression and retards differentiation, thus suggesting a role for HMGN1 in controlling digit formation (Furusawa et al., 2006). An additional gene of note, Fgf18, is expressed in the perichondrium (outer surface of the developing bone) and interdigital mesenchyme where it is proposed to function as a regulator of chondrogenesis and osteogenesis (Liu et al., 2002). Thus, Hmgn1 and Fgf18 play roles in tissue remodeling at the digit–interdigit junction during digit formation.
Retinaldehyde dehydrogenase-2, encoded by Raldh2 (Aldh1a2), oxidizes the vitamin A metabolite retinaldehyde to RA, and thus controls much of the endogenous synthesis of RA during vertebrate organogenesis (Duester, 2008). Loss of RA signaling in Raldh2−/− mouse embryos leads to defects in body axis extension and heart development that result in major developmental abnormalities beyond E8.5 with no sign of limb development (Niederreither et al., 1999; Mic et al., 2002). Limited maternal dietary RA supplementation of Raldh2−/− embryos rescues early lethality (Niederreither et al., 2001), with forelimbs consistently exhibiting retarded growth compared with hindlimbs which appear normal at E10.5 (Zhao et al., 2009). These differences between forelimb and hindlimb development in rescued Raldh2−/− embryos demonstrate that RA action in the body axis from E7.5–E8.5 is required for induction of forelimbs (most likely through RA repression of axial Fgf8 expression), but that RA is not required for hindlimb induction at E9.5, nor for patterning of either forelimbs or hindlimbs (Zhao et al., 2009). As Raldh2 expression later initiates in the interdigital mesenchyme at E12.5 (Niederreither et al., 1997; Kuss et al., 2009), examination of hindlimbs in Raldh2−/− rescued mutants at later stages establishes a genetic loss-of-function model for examining a later role of RA in the interdigital zones.
Here, we use rescued Raldh2−/− embryos to examine the late effects of RA signaling in the E13.5 mutant hindlimb autopod, which appears relatively normal from an external view, but which completely lacks RA synthesis. Our findings demonstrate that RA generated by Raldh2 is required for interdigital apoptosis. Our studies also show that RA induces expression of Hmgn1 and Fgf18 specifically at the digit–interdigit junction, thus placing RA upstream of genes that define the digit periphery during interdigital tissue remodeling.
Raldh2 Null Model for Studying Interdigital RA Signaling
Expression of Raldh2 in mouse hindlimb autopods was examined by whole-mount in situ hybridization at E12.5–E13.5. At E12.5, Raldh2 mRNA was detected throughout the interdigital zone, whereas at E13.5 Raldh2 mRNA was detected mostly at the digit–interdigit junction (Fig. 1A,B). These findings suggest that, in addition to playing a role in the induction of apoptosis throughout the interdigital mesenchyme as previously suggested (Dupé et al., 1999), RA signaling may play another role specifically at the digit–interdigit junction.
To examine the function of RA during interdigital development, we examined rescued Raldh2−/− embryos by performing maternal dietary RA supplementation from E6.75–E9.25 followed by return to a normal diet from E9.25 onward. At E12.5–E13.5, rescued Raldh2−/− embryos consistently exhibited a stunted forelimb and a relatively normal-sized hindlimb (Fig. 1C,D). The RARE-lacZ RA-reporter transgene present in these embryos demonstrated that rescued mutants lack expression of RARE-lacZ in the interdigital zones of both forelimbs and hindlimbs (Fig. 1C,D). As mutant forelimbs always exhibited growth retardation, all further studies were performed on hindlimbs from wild-type and rescued Raldh2−/− embryos, with the latter referred to simply as Raldh2−/− embryos.
Effect of Raldh2 Loss-of-Function on Interdigital RA Signaling, Mmp11 Expression, and Apoptosis
At E13.5 we observed that hindlimb autopodal RARE-lacZ expression is normally very high in the interdigital zones, but this expression was completely absent in hindlimbs from Raldh2−/− embryos (Fig. 2A,B). We also examined expression of Cyp26b1 encoding a P450 enzyme that catalyzes RA degradation in several regions of the embryo including limb buds (MacLean et al., 2001). Cyp26b1 mRNA was localized in developing digit cartilage but not interdigital mesenchyme in wild-type E13.5 hindlimbs, and this pattern was still observed in Raldh2−/− embryos although expression was somewhat reduced (Fig. 2C,D). These findings suggest that RA synthesis controlled by Raldh2 in the interdigital mesenchyme coupled with RA degradation controlled by Cyp26b1 in the digits, results in the confinement of autopodal RA signaling to the interdigital zones.
As previous studies on RARb/RARg double mutants had demonstrated that loss of these two RA receptors results in loss of interdigital expression of Mmp11, we examined this gene in autopods lacking RA synthesis. In Raldh2−/− hindlimbs at E13.5, we observed loss of interdigital Mmp11 expression (Fig. 2E,F). Thus, our findings are consistent with the RARb/RARg loss-of-function model, providing evidence that Mmp11 is an important target gene for RA action in the interdigital zone.
As our rescued Raldh2−/− mutant embryos do not survive beyond E13.5, we were unable to determine if older embryos retain interdigital webbing as shown for E15.5 RARb/RARg double mutants (Ghyselinck et al., 1997). The RARb/RARg phenotype is also associated with a large reduction in interdigital apoptosis at E13.5 (Dupé et al., 1999). Terminal deoxynucleotidyl transferase–mediated deoxyuridinetriphosphate nick end-labeling (TUNEL) assays of wild-type and Raldh2−/− hindlimbs at E13.5 show that a loss of RA also results in a large reduction in interdigital apoptosis (Fig. 2G,H).
Loss of Interdigital RA Synthesis Results in Loss of Hmgn1 Expression but Sox9 Expression Is Unchanged
Our observation that Raldh2 expression progressively becomes limited to the digit–interdigit junction by E13.5 led us to examine expression of genes known to play a role in digit formation. As previously reported, Hmgn1 and Sox9 exhibit complementary expression in the autopod, with Hmgn1 limited to the digit–interdigit boundary and Sox9 found in the digits (Furusawa et al., 2006). Here, we demonstrate that the E13.5 hindlimb autopod normally exhibits very high expression of Hmgn1 limited to the digit–interdigit junction (not throughout the interdigital mesenchyme), but that Raldh2−/− hindlimbs exhibit greatly reduced Hmgn1 mRNA in this domain (Fig. 3A,B). In contrast, Sox9 mRNA normally expressed at high levels in the digits is unchanged in the Raldh2−/− hindlimbs (Fig. 3C,D). Alcian blue staining of cartilage also shows that E13.5 wild-type and Raldh2−/− hindlimbs have comparable digit cartilage formation (Fig. 3E,F). Together, these findings suggest that RA acts at the digit–interdigit junction to maintain a high level of Hmgn1 expression, but loss of Hmgn1 in Raldh2 mutants does not appear to allow Sox9 expression to expand into the interdigital zone.
Effect of RA Signaling on Interdigital Expression of RARb and Fgf18
To extend our gene expression analyses at the digit–interdigit junction, we examined expression of two additional genes previously shown to be expressed in the interdigital zone, i.e., RARb (Mendelsohn et al., 1991) and Fgf18 (Liu et al., 2002). Previous studies have shown that RARb is RA inducible and contains an RA response element in its promoter (Hoffmann et al., 1990), and Fgf18 expression in the caudal body axis is lost in Raldh2−/− embryos (Vermot et al., 2005; Zhao and Duester, 2009). Here, we found that RARb mRNA localized primarily to the digit–interdigit junction (rather than throughout the interdigital zone) in E13.5 hindlimbs, and this domain was lost in Raldh2−/− embryos (Fig. 4A,B). Concurrently, we found that Fgf18 mRNA was expressed predominately in the developing joints and at lower levels in the interdigital zone in E13.5 hindlimbs; although interdigital expression was low it is interesting to notice that such expression is localized primarily to the digit–interdigit junction rather than being spread evenly across the interdigital zone (Fig. 4C). In E13.5 Raldh2−/− hindlimbs, the digit–interdigit expression domain of Fgf18 mRNA was lost, whereas expression in joints was unchanged (Fig. 4D). These observations provide evidence that a major function of interdigital RA signaling is regulation of fibroblast growth factor (FGF) signaling at the digit–interdigit junction during tissue remodeling.
Bmp Expression in Raldh2 Mutant Hindlimbs
Bone morphogenetic protein (BMP) signals are thought to control interdigital cell death perhaps by regulating events in the apical ectodermal ridge such as FGF signaling or Msx2 expression (Pajni-Underwood et al., 2007; Maatouk et al., 2009). Msx2, expressed in the apical ectodermal ridge and interdigital mesenchyme, has been found to be required for regression of interdigital mesenchyme (Marazzi et al., 1997; Satokata et al., 2000). Here, we found that expression of Fgf8 and Msx2 in autopods of E13.5 Raldh2−/− hindlimbs is not significantly altered (Fig. 4E–H). Thus, loss of RA signaling may not control interdigital apoptosis through effects on the apical ectodermal ridge.
We also examined expression of several BMPs in the autopod at E13.5. We found that Bmp2, expressed primarily in the interdigital region, was unaffected by loss of RA in the Raldh2−/− mutant (Fig. 5A,B). We also found that Bmp4 expression in the apical ectodermal ridge and digit mesenchyme was unaffected by loss of RA in the Raldh2−/− mutant, and expression in the interdigital region was maintained at the same level in both wild-type and mutant (Fig. 5C,D). However, Bmp7 expression in the interdigital region was completely lost in the Raldh2−/− mutant (Fig. 5E,F). Similar results for Bmp2/4/7 expression were obtained for RARb/RARg mutants (Dupé et al., 1999). Together, these findings suggest that RA signaling results in the induction of interdigital Bmp7 expression which may be important for inducing interdigital cell death. Although our results suggest that RA does not regulate Bmp2 and Bmp4, these genes have also been shown to be important for interdigital cell death (Maatouk et al., 2009) by means of their receptor Bmpr1a expressed in the apical ectodermal ridge (Pajni-Underwood et al., 2007).
Following limb induction and patterning, RA synthesis initiates in the autopod by E12.5 and RA signaling becomes confined to the interdigital zones by a combination of interdigital Raldh2 expression and Cyp26b1 expression in the developing digits. Analysis of Raldh2−/− autopods has identified several genes whose interdigital expression is either completely lost (Mmp11, RARb, Fgf18, Bmp7) or greatly down-regulated (Hmgn1). In contrast, expression of Sox9 and Bmp4 in digit cartilage is unaffected by a loss of RA. Similarly, while Fgf18 expression is lost in the interdigital zone of Raldh2−/− autopods, we find that it is maintained in digits (particularly joints) consistent with RA signaling not occurring in digit cartilage due to Cyp26b1 expression. A precedent exists for RA regulation of certain Fgf18 expression domains as it has been shown that Fgf18 expression at the somite/presomitic mesoderm border requires RA synthesis controlled by Raldh2 (Vermot et al., 2005; Zhao and Duester, 2009). Our observation that RA signaling is normally absent in developing digit cartilage and apparently unnecessary for skeletogenesis is consistent with previous observations that addition of RA or loss of Cyp26b1 inhibits cartilage formation or ossification (Biddulph et al., 1988; Laue et al., 2008; Kuss et al., 2009; Maclean et al., 2009). In addition, recent studies have demonstrated that, although RARg expression in the bone growth plate is required for bone growth, endogenous RA is undetectable in the growth plate, and further functional experiments demonstrated that RARg functions as a ligand-less repressor during bone growth (Williams et al., 2009).
RA synthesis controlled by Raldh2 expression initially occurs throughout the interdigital zone by E12.5. At E13.5, we find that RARE-lacZ (an RA-reporter transgene) and Mmp11 (an endogenous RA target gene) are expressed throughout the interdigital zone and both require RA synthesis for expression. From previous studies on RARb/RARg mutant mice it appears that RA must act throughout the interdigital zone to induce Mmp11 and to initiate apoptosis that will eventually eliminate this tissue by E15.5 (Dupé et al., 1999). We found that Raldh2−/− autopods also have a large reduction in interdigital apoptosis, thus providing evidence that both ligands and receptors for RA signaling are crucial for interdigital apoptosis. Mmp11 presumably functions in tissue remodeling during the apoptotic process through degradation of extracellular matrix as do other members of the matrix metalloproteinase family (Bai et al., 2005). We also found that Raldh2−/− autopods have a large reduction in interdigital Bmp7 expression which may be crucial for BMP-mediated induction of apoptosis. However, we found that endogenous RA does not regulate expression of Bmp2 or Bmp4 in the apical ectodermal ridge which has also been associated with interdigital apoptosis (Pajni-Underwood et al., 2007; Maatouk et al., 2009), and we observed no requirement for endogenous RA to regulate Msx2 or Fgf8 expression in the apical ectodermal ridge.
The findings presented here suggest that in addition to induction of interdigital apoptosis, RA also functions to define the digit–interdigit junction where tissue remodeling occurs. By E13.5, we observed that expression of Raldh2 is limited to the digit–interdigit junction rather than being expressed throughout the interdigital zone. In contrast to Mmp11, interdigital expression of RARb, Hmgn1, and Fgf18 in E13.5 autopods occurs specifically at the digit–interdigit junction rather than throughout the interdigital zone, thus correlating with Raldh2 expression in this domain. Hmgn1 and Fgf18 have previously been shown to be expressed in the interdigital zone and both are implicated in regulation of digit formation (Liu et al., 2002; Furusawa et al., 2006). We propose that RA synthesis becomes more predominately localized at the digit–interdigit junction as autopod development proceeds, leading to RARb up-regulation which intensifies RA signaling at this junction, resulting in up-regulation of Hmgn1 and Fgf18. Thus, RA-induced FGF signaling specifically at the digit–interdigit interface may be necessary for normal tissue remodeling during digit formation. In summary, our new model for interdigital RA action suggests that RA signaling acts upstream of Hmgn1 and Fgf18 to initiate the tissue remodeling process that will lead to formation of the digit–interdigit junction, ultimately defining the edge of the domain that is slated for elimination during interdigital apoptosis.
Generation of Rescued Raldh2−/− Embryos
Generation of heterozygous Raldh2+/− adult mice were previously described (Mic et al., 2002). Rescue of Raldh2−/− early lethality (which occurs at E9.5) was performed by supplementing the maternal diet with RA as follows. From E6.75 to E9.25 pregnant females were provided food with 0.1 mg of all-trans-retinoic acid (catalog no. R2625, Sigma Chemical Co., St. Louis, MO) per gram of standard mouse chow. RA-supplemented food was prepared by mixing 50 μl of a 50 mg/ml suspension stock solution of RA in 100% ethanol with 450 μl corn oil, then evenly distributing 200 μl RA/ethanol/oil suspension into 10 g of mouse chow (ground with mortar and pestle) to achieve a final RA concentration is 0.1 mg/g mouse chow. The food was put in a deep glass Petri dish and changed twice a day (once in the morning and once in the evening) in a cage where all other food has been removed. RA was provided at E6.75 (evening of day 6), E7.25 (morning of day 7), E7.75, E8.25, and E8.75. At E9.25, mice were put back on normal food until embryos were collected at E12.5–E13.5. Raldh2−/− embryos were identified by the stunted forelimb (hindlimb normal) or by genotyping using polymerase chain reaction analysis of yolk sac DNA. All mouse studies conformed to the regulatory standards adopted by the Animal Research Committee at the Burnham Institute for Medical Research.
Whole-Mount In Situ Hybridization, lacZ Detection, Cartilage Staining, and TUNEL Assay
Whole-mount in situ hybridization (ISH) was used to detect Raldh2 mRNA and other mRNAs as previously described (Mic et al., 2002). To assess RARE-lacZ expression, β-galactosidase activity was detected by performing X-gal (5-bromo-4-chloro-3-indolyl-β-D-galactopyranoside) staining for 6–24 hr as previously described (Rossant et al., 1991). Alcian blue staining of cartilage was performed as described (Hogan et al., 1994). Apoptosis was detected using the DeadEnd Colorimetric TUNEL System (Promega Corp.) following the manufacturer's protocol. Wild-type and Raldh2−/− embryos were treated under identical hybridization or staining conditions and stained for the same length of time. For each analysis (ISH, β-galactosidase stain, Alcian blue stain, or TUNEL), at least three Raldh2−/− embryos and three wild-type embryos at a similar stage were examined, all with comparable results.
We thank the following individuals for mouse cDNAs used to prepare in situ hybridization probes: M. Bustin (Hmgn1 and Sox9), V. Giguère (RARb), B. Harfe (Msx2), B. Hogan (Bmp4, Bmp7), J. C. Izpisua-Belmonte (Bmp2), N. Itoh (Fgf18), G. Martin (Fgf8), M. Petkovich (Cyp26b1), M.-C. Rio (Mmp11). We also thank J. Rossant for providing RARE-lacZ reporter mice. G.D. was funded by a National Institutes of Health grant.