HOXA13 regulates Aldh1a2 expression in the autopod to facilitate interdigital programmed cell death

Authors


Abstract

Background: Retinoic acid (RA), plays an essential role in the growth and patterning of vertebrate limb. While the developmental processes regulated by RA are well understood, little is known about the transcriptional mechanisms required to precisely control limb RA synthesis. Here, Aldh1a2 functions as the primary enzyme necessary for RA production which regulates forelimb outgrowth and hindlimb digit separation. Because mice lacking HOXA13 exhibit similar defects in digit separation as Aldh1a2 mutants, we hypothesized that HOXA13 regulates Aldh1a2 to facilitate RA-mediated interdigital programmed cell death (IPCD) and digit separation. Results: In this report, we identify Aldh1a2 as a direct target of HOXA13. In absence of HOXA13 function, Aldh1a2 expression, RA signaling, and IPCD are reduced. In the limb, HOXA13 binds a conserved cis-regulatory element in the Aldh1a2 locus that can be regulated by HOXA13 to promote gene expression. Finally, decreased RA signaling and IPCD can be partially rescued in the Hoxa13 mutant hindlimb by maternal RA supplementation. Conclusions: Defects in IPCD and digit separation in Hoxa13 mutant mice may be caused in part by reduced levels of RA signaling stemming from a loss in the direct regulation of Aldh1a2. These findings provide new insights into the transcriptional regulation of RA signaling necessary for limb morphogenesis. Developmental Dynamics 242:687–698, 2013. © 2013 Wiley Periodicals, Inc.

INTRODUCTION

The formation of the autopod requires the integration of signaling factors such as bone morphogenetic proteins (BMPs), fibroblast growth factors (FGFs), hedgehog proteins, and retinoic acid (RA) to facilitate limb outgrowth, skeletal element formation, and digit separation (Zuzarte-Luis and Hurle, 2002; Zhao et al., 2010; Rosello-Diez et al., 2011; Probst et al., 2011). The coordination of these signals is controlled, in part, by the Group 13 HOX transcription factors including HOXA13 and HOXD13 whose target gene products regulate the formation of the digits, phalangeal joints, and carpal/tarsal elements (Fromental-Ramain et al., 1996; Stadler et al., 2001; Knosp et al., 2004, 2007; Kuss et al., 2009; Perez et al., 2010; Villavicencio-Lorini et al., 2010).

Digit separation, a process facilitated by interdigital programmed cell death (IPCD), is also regulated by HOXA13, as mice lacking HOXA13 exhibit reduced levels of IPCD as well as severe syndactyly (Stadler et al., 2001; Knosp et al., 2004; Perez et al., 2010). Our investigations of the mechanisms underlying digit syndactyly in the Hoxa13 mutant limb revealed that HOXA13 promotes IPCD by directly regulating Bmp2 and Bmp7 in the interdigital tissues (Knosp et al., 2004). Interestingly, while BMP2 or BMP7 supplementation increased IPCD in the mutant autopod, these treatments could not fully restore apoptosis to the levels exhibited by littermate controls, suggesting that additional targets of HOXA13 may be necessary for digit separation (Knosp et al., 2004).

Recently, several studies suggest that fibroblast growth factor 8 (FGF8) and RA signaling play important roles in regulating IPCD and limb growth. For IPCD, FGF8 down-regulation in the overlying ectoderm appears to provide an essential signal to the underlying mesenchyme to undergo programmed cell death (Hernandez-Martinez et al., 2009). Once initiated, IPCD is sustained by RA, as deficiencies in the retinoic acid receptors or the metabolic enzymes necessary to convert vitamin A to RA profoundly affect IPCD, digit separation, and forelimb growth (Zhao et al., 1996; Cash et al., 1997; Dupe et al., 1999; Niederreither et al., 2002; Zuzarte-Luis and Hurle, 2002; Zhao et al., 2010). In the developing embryo, HOX proteins appear to be one of the coordinating factors necessary for RA signaling. Indeed studies by Vitobello et al. (2011) demonstrate that ALDH1A2 (RALDH2), the key enzyme necessary to convert retinaldehyde to RA, is regulated by HOXA1 to facilitate segmentation of the early hindbrain. ALDH1A2 is also the primary retinaldehyde converting enzyme in the limb, where it produces the RA necessary for forelimb outgrowth and IPCD in the hindlimb (Cash et al., 1997; Dupe et al., 1999; Niederreither et al., 2002; Kuss et al., 2009; Hernandez-Martinez et al., 2009; Zhao et al., 2010).

Recognizing that Aldh1a2 can be regulated by HOX proteins during embryogenesis as well as its essential role in IPCD, we hypothesized that the syndactyly exhibited by humans and mice lacking HOXA13 function may be caused by a loss in Aldh1a2 expression in the interdigital tissues, resulting in insufficient levels of RA signaling necessary for IPCD (Stern et al., 1970; Poznanski et al., 1975; Innis, 1993; Fromental-Ramain et al., 1996; Mortlock and Innis, 1997; Stadler et al., 2001; Perez et al., 2010). Testing this hypothesis, the present study confirms that Aldh1a2 expression and RA signaling are delayed and ultimately reduced in Hoxa13 homozygous mutant limb. Moreover, we identified direct binding of a cis-regulatory element in the Aldh1a2 locus by HOXA13 in the limb mesenchyme and that HOXA13 can use this cis-regulatory element to direct gene expression in vitro. Finally, supplementation with sub-teratogenic dosages of RA was successful at improving RA signaling and IPCD in the Hoxa13 mutant limb, confirming that reduced levels of Aldh1a2 expression and RA signaling contribute to the loss of IPCD seen in Hoxa13 homozygous mutants. Taken together, these results provide new insight into the hierarchy of genes regulated by HOXA13 in the distal limb, as well as the transcriptional components necessary for IPCD and digit separation.

RESULTS

Aldh1a2 Expression is Delayed and Reduced in the Hoxa13 Mutant Limb

Analysis of Aldh1a2 and Hoxa13 expression indicates that Hoxa13 expression precedes Aldh1a2 expression in the embryonic day (E) 10.5 limb (Fig. 1A–C). By E11.5 Aldh1a2 expression is readily detected in three limb regions: a proximal zone at the base of the forelimb bud, an intermediate autopod zone, and in a distal autopod zone in wild-type embryos (Fig. 1D). In contrast, while homozygous mutants also exhibited Aldh1a2 expression in the proximal and intermediate autopod zones, no expression was detected in the distal autopod zone at E11.5 (Fig. 1E). In the hindlimb, faint Aldh1a2 expression was also faintly detected in the distal autopod and medial autopod domains in wild-type embryos; whereas only the medial autopod expression domain was detected in the homozygous mutant limb. By E12.5, Aldh1a2 expression is present but reduced in the mutant autopod interdigital zones II–III and III–IV, suggesting both a delay and reduction in Aldh1a2 expression in the autopod (Fig. 1G–I). These same interdigital domains also exhibited lower levels of detected Aldh1a2 transcripts in the E13.5 limb compared with littermate controls (Fig. 1J,K). Quantification of Aldh1a2 expression in the hindlimb and forelimb autopods at E13.5 by quantitative reverse transcriptase-polymerase chain reaction (qRT-PCR) confirmed a consistent reduction in expression with the mutant hindlimb exhibiting the greatest reduction in expression (2.5-fold reduced) compared with the mutant forelimb (1.6-fold reduced) (Fig. 2).

Figure 1.

Analysis of Aldh1a2 expression in Hoxa13-GFP mutant and wild-type littermates. A–C: Aldh1a2 expression is not detected in the forelimb or hindlimb of Hoxa13-GFP wild-type or homozygous mutant littermates at embryonic day (E) 10.5 (A,B), whereas Hoxa13 expression (C) is present in the forelimb and hindlimb as this same developmental stage. F, forelimb; H, hindlimb. D: Aldlh1a2 expression is present in the distal autopod, medial autopod, and proximal forelimb regions of wild-type E11.5 embryos. Arrowheads denote the sites of expression along the limb proximal–distal axis. E: Aldh1a2 expression is absent in the distal autopod region (asterisk) of E11.5 Hoxa13-GFP homozygous mutant embryos whereas expression appears to be maintained in the medial autopod and proximal limb regions (arrowheads). F: Hoxa13 is strongly expressed throughout the autopod at E11.5, placing the transcription factor in the limb domain exhibiting perturbed Aldh1a2 expression. H: Aldh1a2 expression is restricted to the interdigital limb tissues in wild-type E12.5 embryos. H: Aldh1a2 expression is either absent or severely reduced in the interdigital tissues of E12.5 Hoxa13-GFP homozygous mutant littermates. Roman numerals depict the digit condensations. I: Interdigital expression of Hoxa13 (arrowheads) in E12.5 limbs. Roman numerals depict the digit condensations. J: Aldh1a2 expression is restricted to the outer interdigital tissues at E13.5 in wild-type limbs. Roman numerals depict the digit condensations. K: Interdigital expression of Aldh1a2 is reduced in the E13.5 Hoxa13-GFP homozygous mutant limb. Roman numerals depict the digit condensations. L: Hoxa13 is co-expressed in the same outer interdigital tissue as Aldh1a2 in the E13.5 limb.

Figure 2.

Quantification of Aldh1a2 expression in the Hoxa13-GFP mutant limb. Quantitative reverse transcriptase-polymerase chain reaction (qRT-PCR) characterizing fold-change differences in Aldh1a2 expression between embryonic day (E) 13.5 mutant and wild-type limb tissues. Note that mutant hindlimbs (Mut HL) and mutant forelimbs (Mut FL) exhibited reduced levels of Aldh1a2 expression relative to wild-type controls. Error bars depict the standard deviation of the average fold change values for six independent assays. GFP, green fluorescent protein.

A Conserved cis-Regulatory Element in the Aldh1a2 Locus is Bound by HOXA13 in the Developing Limb

Recognizing the overlap in expression between Hoxa13 and Aldh1a2 in the developing autopod, we hypothesized that HOXA13 may directly regulate Aldh1a2 facilitate IPCD. Sequence analysis of the murine Aldh1a2 locus revealed a highly conserved 556-bp region upstream of the Aldh1a2 initiation codon (UCSC Genome browser, Chr9:71062106–71062662) that contained multiple copies of the previously characterized HOXA13 binding site (Fig. 3) (Knosp et al., 2007). Chromatin immunoprecipitation (ChIP) using a HOXA13 antibody and chromatin derived from wild-type E12.5 autopods confirmed that HOXA13 directly binds this conserved region in the developing murine limb (Fig. 3). Parallel ChIP assays using autopod chromatin derived from Hoxa13 homozygous mutants did not detect the same Aldh1a2 DNA region, a finding consistent with the Hoxa13-GFP (GFP, green fluorescent protein) mutation that replaces the DNA binding domain with green fluorescent protein yet allows for immunoprecipitation using the HOXA13 antibody which recognizes the N-terminal region of the mutant HOXA13-GFP protein (Fig. 3) (Stadler et al., 2001; Knosp et al., 2004). Analysis of the bound 556-bp region for transcriptional regulation by HOXA13 revealed significant dosage-dependent reporter activation by HOXA13 (P < 0.05), confirming that the DNA region bound by HOXA13 in the autopod functions as a cis-regulatory element in vitro (Fig. 4).

Figure 3.

Identification of a conserved Aldh1a2 DNA region bound by HOXA13 in the developing autopod. A,B: Analysis of the murine Aldh1a2 locus on mouse chromosome 9 identified a 556 bp region containing multiple HOXA13 binding sites that are conserved in several tetrapod species. HOXA13 binding sites are underscored in panel (B). C: Chromatin immunoprecipitation (ChIP) using a HOXA13-specific antibody and limb bud chromatin from embryonic day (E) 12.5 Hoxa13-GFP wild-type and homozygous mutant embryos revealed binding of the HOXA13 protein to the conserved 556 bp region in the Aldh1a2 locus in wild-type limb tissues. Parallel assays using limb chromatin derived from homozygous mutant littermates showed no enrichment of the 556 bp region, which is consistent with the absence of the DNA binding domain in the mutant HOXA13 protein. Two primer sets were used to confirm the binding of HOXA13 to this region in wild-type chromatin.

Figure 4.

The conserved Aldh1a2 region functions as a HOXA13-regulated cis-acting element. Blue bars represent the response of the Aldh1a2 cis-acting element in response to increasing doses of HOXA13. Gray bars represent control transfections using the same dosages of the Hoxa13 expression vector and the pGL3-promoter luciferase vector lacking the Aldh1a2 cis-acting element. Error bars represent the standard deviation of the averaged and normalized luciferase readings for six independent assays. Luciferase expression was significantly increased in response to HOXA13 expression for all dosages P ≤ 0.05. A Renilla luciferase expression vector was included in all transfections to normalize for transfection efficiency.

Decreased RA Signaling in the Mutant Limb Can Be Restored by Supplementation of the Maternal Diet

Recognizing that Aldh1a2 is the primary retinaldehyde dehydrogenase that converts retinaldehyde to RA in the limb, we hypothesized that decreased Aldh1a2 expression in the Hoxa13 mutant autopod would result in decreased RA signaling in this region (Zhao et al., 1996; Niederreither et al., 2002). To test this hypothesis, an RA signaling reporter mouse (RARE-Hspa1b/lacZ; Rossant et al., 1991) was bred into the Hoxa13-GFP mouse colony. Mice heterozygous for the RARE-Hspa1b/lacZ reporter and the Hoxa13-GFP mutant allele were intercrossed and the limbs were examined for LacZ staining as an indicator of active RA signaling. No expression of the RARE-Hspa1b/lacZ reporter was detected in the autopod region of mutant or wild-type E11.5 limbs (n = 4) (Fig. 5A,B). By E12.5, LacZ expression was detected in wild-type forelimb interdigital zones I–II, II–III, and III–IV, and IV–V (n = 8) (Fig. 5C). In contrast, LacZ expression was only detected in interdigital zones II–III, and III–IV (n = 6) in the homozygous mutant littermate forelimb (Fig. 5D). In the wild-type E12.5 hindlimb, LacZ staining was also present in the II–III and III–IV interdigital zones (n = 8), whereas no LacZ staining was detected in the mutant hindlimb autopod (n = 6) (Fig. 5F).

Figure 5.

Analysis of active retinoic acid signaling in the Hoxa13 mutant limb. A,B: Active retinoic acid (RA) signaling as indicated by expression of the transgenic RARE-LacZ reporter is absent in the distal fore- and hindlimbs at embryonic day (E) 11.5 in Hoxa13 wild-type and homozygous mutants. C,E: Retinoic acid signaling is present in the interdigital tissues in wild-type fore- and hindlimbs at E12.5. D,F: Retinoic acid signaling is reduced in the interdigital tissues of E12.5 Hoxa13 homozygous mutant forelimb and not detected in the homozygous mutant hindlimb.

In the E13.5 autopod, robust LacZ staining was detected in the interdigital tissues of wild-type forelimbs and hindlimbs (n = 6) (Fig. 6). In contrast, LacZ staining was reduced and limited to interdigital zones II–III, III–IV, and IV–V of homozygous mutant littermates in both the forelimbs and the hindlimbs (n = 6) (Fig. 6). Next, recognizing that reductions in Aldh1a2 expression and its subsequent effect on RA levels in the Hoxa13 mutant autopod could be by-passed by maternal dietary supplementation, we examined whether sub-teratogenic RA supplementation in the maternal diet could improve activation of the RARE-LacZ reporter in embryonic interdigital tissues (Niederreither et al., 2002). A comparison of LacZ staining in E13.5 homozygous mutant autopods revealed a consistent expansion of LacZ expression in hindlimb regions II–III and III–IV of embryos receiving maternal RA supplementation compared with normal diet controls (n = 4) (Fig. 6).

Figure 6.

Response of the limb interdigital retinoic acid (RA) signaling to dietary RA supplementation. A,B: RA signaling in the embryonic day (E) 13.5 Hoxa13 homozygous mutant left and right forelimb derived from pregnant females fed a normal diet lacking RA. C,D: RA signaling in the E13.5 Hoxa13 homozygous mutant left and right forelimb derived from pregnant females fed a diet supplemented with 0.5 mg RA/g food. E,F: RA signaling in E13.5 wild-type forelimbs derived from pregnant females fed a normal diet lacking RA. G,H: RA signaling in the E13.5 Hoxa13 homozygous mutant left and right hindlimbs derived from pregnant females fed a normal diet lacking RA. I,J: RA signaling in the E13.5 Hoxa13 homozygous mutant left and right hindlimb derived from pregnant females fed a diet supplemented with 0.5 mg RA/g food. Note the elevated levels of LacZ staining in the RA-supplemented limbs compared with mutant hindlimbs receiving no RA supplementation. K,L: RA signaling in E13.5 wild-type hindlimbs derived from pregnant females fed a normal diet lacking RA. Roman numerals denote the respective digit condensations. The digit I region, which is missing in the Hoxa13 mutant limb, is indicated by an asterisk.

Finally, because overexpression of the RA inactivating enzyme Cyp26B1 could also account for changes in limb RA signaling, we examined whether Cyp26B1 expression was affected in the Hoxa13 mutant limb. Characterization of the Cyp26B1 expression pattern revealed high levels of expression throughout the distal limb bud at E11.5 with no differences in expression between Hoxa13 mutant and wild-type limbs (Fig. 7A,B). By E12.5, Cyp26B1 expression was restricted to the distal digit condensations in the wild-type forelimbs with higher levels being present in the mutant condensations (Fig. 7C,D). In the wild-type hindlimb, Cyp26B1 expression was also restricted to the digit condensations at E12.5, whereas homozygous mutants only exhibited diffuse expression in the distal autopod due to the delayed formation of the hindlimb condensations at this same developmental stage (Fig. 7E,F).

Figure 7.

The expression and localization of Cyp26B1 is unaffected by the loss of HOXA13 function in the embryonic day (E) 11.5 and 12.5 limbs. A,B: Cyp26B1 expression is distributed throughout the distal limb in E11.5 Hoxa13 wild-type and homozygous mutant littermates. C,D: Cyp26B1 expression in E12.5 wild-type and mutant littermate forelimbs is restricted to the digit condensations. E,F: Cyp26B1 expression in E12.5 wild-type hindlimb is also restricted to the digit condensations which are delayed in their maturation in the mutant littermate hindlimb.

Effect of RA Supplementation on IPCD in Hoxa13 Mutant Limbs

Recognizing the inductive role of RA in IPCD, we hypothesized that decreased RA signaling may contribute to reduced IPCD seen in Hoxa13 mutant limbs (Stadler et al., 2001; Knosp et al., 2004). Testing this hypothesis, we examined IPCD in E14.5 Hoxa13 mutant limbs that received maternal RA supplementation starting at E8.5. Interestingly, while homozygous mutant limbs from mothers treated with 2.5 μg/g RA exhibited some improvement in IPCD between digits III and IV (compare Figs. 8A and 8E to 8B and 8F) (n = 4), the response of the mutant limbs to maternal RA supplementation was not improved by increasing the RA dosage. Indeed homozygous mutant limbs from mothers treated with RA dosages from 3.0–4.0 μg/g exhibited less IPCD between digits III and IV in the forelimb and the absence of digit formation in the hindlimbs (Fig. 8G,H) (n = 4). Of interest, heterozygous mutant control hindlimbs which exhibit normal levels of IPCD (Fig. 8I,M) also responded to the 2.5 μg/g RA dosage, showing elevated IPCD between digits III and IV and IV and V (Fig. 8N) (n = 6). Higher RA dosages (3.0–4.0 μg/g) also affected digit formation in the control limbs, causing reductions in digit I and shortening of the remaining digits (Fig. 8K–P) (n = 8), suggesting that RA dosages higher than 2.5 μg/g function as a general antagonist to the normal growth of the digit and interdigit zones, a process required to facilitate IPCD.

Figure 8.

Response of Hoxa13 homozygous mutant and heterozygous control limbs to retinoic acid (RA) supplementation by oral gavage at embryonic day (E) 14.5. A: Programmed cell death in homozygous mutant forelimb fed only the corn oil control vehicle. Note that the Hoxa13 homozygous mutant limbs typically exhibit severe reductions in interdigital programmed cell death (IPCD) with the exception of interdigital region IV–V at E14.5. Arrow denotes absence of IPCD in interdigital zone III–IV in the vehicle control treated mice. I* denotes the site of the anterior first digit which is absent in the Hoxa13 homozygous mutant forelimb. V denotes the fifth digit. B: Homozygous mutant forelimb from mothers fed 2.5 μg/g body weight RA by oral gavage. Arrows denotes increased IPCD in interdigital zones III–IV, and IV–V. I* denotes the site of the anterior first digit which is absent in the Hoxa13 homozygous mutant limb. V denotes the fifth digit. C,D: Homozygous mutant forelimbs from mothers fed 3.0 and 4.0 μg/g body weight RA by oral gavage respectively. Note that dosages higher than 2.5 μg/g result in shortening of the digits, and loss of IPCD. I* denotes the site of the anterior first digit which is absent in the Hoxa13 homozygous mutant limb. V denotes the fifth digit. E: Programmed cell death in homozygous mutant hindlimb fed only the corn oil control vehicle. Hoxa13 homozygous mutant hindlimbs typically exhibit severe reductions in IPCD with the exception of the anterior digit I region and interdigital region IV–V at E14.5. Arrow denotes absence of IPCD in interdigital zone III–IV in the vehicle control treated mice. I* denotes the site of the anterior first digit which is typically absent in the Hoxa13 homozygous mutant hindlimb. V denotes the fifth digit. F: Homozygous mutant hindlimb from mothers fed 2.5 μg/g body weight RA by oral gavage. Arrow denotes increased IPCD in interdigital zones III–IV. I* denotes the site of the anterior first digit which is absent in the Hoxa13 homozygous mutant limb. V denotes the fifth digit. G,H: Homozygous mutant hindlimbs from mothers fed 3.0 and 4.0 μg/g body weight RA by oral gavage, respectively. Note that dosages higher than 2.5 μg/g cause nearly a complete loss of hindlimb digit formation. I* and V* denote the sites where the first and fifth digits should have developed. I,M: IPCD present in heterozygous control fore- and hindlimbs from mothers fed only the vehicle control. I and V denote the first and fifth digit. J,N: IPCD present in heterozygous control fore- and hindlimbs from mothers fed 2.5 μg/g body weight RA by oral gavage. Note that IPCD is elevated in interdigital zones III–IV and IV–V of the hindlimb (arrows). Hindlimbs from mothers fed 2.5 μg/g RA exhibit a reduction in digit I (arrowhead), whereas digit V appears normal. K,L,O,P: IPCD present in heterozygous control fore- and hindlimbs from mothers fed 3.0- and 4.0 μg/g body weight RA by oral gavage. Note that heterozygous hindlimbs from mothers receiving RA dosages of 3.0–4.0 μg/g exhibit reduction in the size of digit I (arrowheads) as well as shortening of digits II–V. I,M: Hindlimb IPCD is also reduced in the digit III–IV regions of embryos from mothers receiving 3.0–4.0 μg/g RA by oral gavage compared with the vehicle control.

Finally, recognizing that perturbations in Fgf8 down-regulation could also affect Aldh1a2 expression (Hernandez-Martinez et al., 2009) and IPCD, we also examined whether Fgf8 expression and down-regulation was affected in the Hoxa13 mutant limb. Analysis of Fgf8 expression in the developing autopod revealed strong expression in the apical ectodermal ridge (AER) in wild-type and homozygous mutant forelimbs and hindlimbs at E11.5 (Fig. 9A–D). Reduction in Fgf8 expression in the AER was also detected in wild-type and mutant limbs at E12.5 (Fig. 9E–H). By E13.5 Fgf8 transcripts could not be detected in the AER of wild-type and mutant limbs (Fig. 9I–L) with the exception of a small patch of expression in the ectoderm overlying digit II in the mutant forelimb (Fig. 9J).

Figure 9.

Analysis of fibroblast growth factor-8 (Fgf8) expression and down-regulation in the Hoxa13 mutant limb. A–D: Initial expression of Fgf8 in the embryonic day (E) 11.5 apical ectodermal ridge is not affected by the loss of HOXA13 function in the forelimb or hindlimb. E–H: At E12.5, Fgf8 expression is reduced in the remaining apical ectodermal tissues with no differences in expression between wild-type and homozygous mutant limbs. I–L: At E13.5, Fgf8 down-regulation is nearly complete with only minor amounts of transcript being detected (arrowhead) on the digit ectoderm in the homozygous mutant forelimb.

DISCUSSION

A common function of HOX proteins is their capacity to precisely control tissue formation by regulating multiple steps of a developmental genetic cascade (Weatherbee et al., 1998; Hittinger et al., 2005; Walsh and Carroll, 2007). Mechanistically, HOX proteins achieve this level regulation by binding unique DNA sequences present in target gene cis-regulatory elements. For HOXA13, loss of function analyses indicate key roles in the regulation mesenchymal cell adhesion, chondrogenesis, skeletal element formation, and IPCD suggesting the capacity to regulate multiple steps within individual developmental cascades (Fromental-Ramain et al., 1996; Stadler et al., 2001; Knosp et al., 2004, 2007; Perez et al., 2010). In its regulation of IPCD, HOXA13 appears to be functioning as canonical HOX protein, regulating the expression of BMPs and RA to facilitate the removal of the interdigital tissues (Francis et al., 1994; Cash et al., 1997; Merino et al., 1999; Knosp et al., 2004, 2007; Hernandez-Martinez et al., 2009).

HOXA13 is Required for the Maintenance of IPCD but not its Initiation

Previous studies indicate that FGF-, BMP-, and RA signaling play predominant roles in the IPCD genetic cascade (Francis et al., 1994; Dupe et al., 1999; Merino et al., 1999; Guha et al., 2002; Niederreither et al., 2002; Hernandez-Martinez et al., 2009). Key to the initiation of IPCD is the down-regulation of FGF8 (Hernandez-Martinez et al., 2009). Here, HOXA13 does not appear to be playing a direct role, as the HOXA13 and Fgf8 expression domains are separated in the developing autopod (this work; Stadler et al., 2001). Moreover, because Fgf8 expression and down-regulation proceed normally in the Hoxa13 mutant limb, it is also unlikely that HOXA13 indirectly regulates Fgf8 expression.

While the initiation of IPCD proceeds normally in the Hoxa13 mutant limb, little IPCD is detected as the limb matures, suggesting that the maintenance of ICPD requires HOXA13 function. In this role, HOXA13 appears to coordinate the expression of several genes whose products are required for IPCD progression including Aldh1a2, Bmp2, and Bmp7 (Dupe et al., 1999; Merino et al., 1999; Stadler et al., 2001; Niederreither et al., 2002; Guha et al., 2002; Knosp et al., 2004). The addition of Aldh1a2 as target of HOXA13 in the interdigital tissues also provides mechanism to explain why BMP supplementation in the Hoxa13 mutant limb only partially rescues the loss of IPCD, as RA produced by ALDH1A2 is required to induce Caspase 3 expression, a primary effector of IPCD (Kuida, 1997; Woo et al., 1998; Ali-Khan and Hales, 2003; Knosp et al., 2004).

Impact of Aldh1a2 and Cyp26B1 Expression on RA Signaling in the Hoxa13 Mutant Autopod

The absence of Aldh1a2 expression in the mutant E11.5 autopod domain suggests that IPCD could be affected by reduced levels of RA signaling. Testing this hypothesis revealed a surprising result, as no RA signaling was detected in the autopod domain at E11.5 in either wild-type or homozygous mutants using the RARE-LACZ reporter. This lack of signaling indicates that little RA signaling is occurring in the distal autopod at E11.5 and can be explained by the corresponding peak of Cyp26B1 expression in this same region. Here, CYP26B1 metabolizes RA to less active forms such as 5, 8-epoxy RA, 4-hydroxy RA, and 18-hydroxy RA which modulates RA signaling to facilitate embryogenesis (White et al., 1996; Fujii et al., 1997; Sakai et al., 2001). As the limb matures however, Cyp26B1 expression is reduced in the interdigital tissues allowing active RA signaling and IPCD providing a mechanism to temporally regulate digit separation. Here, changes in Aldh1a2 expression in the mutant interdigital tissues would affect RA signaling and IPCD, a conclusion supported by the reduced levels of RARE-LACZ reporter activity in the mutant limbs from E12.5 to E14.5 as well as decreased IPCD in the mutant limbs not receiving RA supplementation.

Role of RA Supplementation, Hoxd13, and Hoxa13 in IPCD and Limb Development

Aldh1a2 expression and RA signaling are both reduced Hoxa13 homozygous mutant limbs. Because RA is required for the expression of tissue remodeling genes Hmgn1 and Fgf18 in the digit-interdigit regions to facilitate digit separation, it is likely that the reduction in RA production could contribute, in part, to the loss in IPCD seen in the mutant limb (this work; Zhao et al., 2010). From this work, it is clear that low RA dosages partially rescue the loss in IPCD caused by the loss of HOXA13 function. However, it is unlikely that maternal RA supplementation could be developed as a therapy to restore normal IPCD in the mutant limb, as the teratogenic effects caused by higher RA dosages appear to be epistatic to the effects of RA on IPCD. Moreover, because Bmp2 and Bmp7 are also regulated by HOXA13 in the interdigital tissues, it is possible that these tissues are less capable to respond to RA as an additional pro-apoptotic signal (Knosp et al., 2004).

HOXD13 also appears to directly regulate Aldh1a2 in the developing limb. However, its function in the limb is different from HOXA13. Here, HOXD13, appears to regulate Aldh1a2 to facilitate RA signaling needed to modulate the formation of cartilage and bone (Kuss et al., 2009). In the presence of the Spd allele of Hoxd13, Aldh1a2 expression is significantly reduced, resulting in the formation of supernumerary digits and ectopic condensations of cartilage and bone (Kuss et al., 2009). What is interesting about the co-regulation of Aldh1a2 by HOXA- and HOXD13 is that the levels of RA needed to suppress ectopic bone growth in Spd mutants, 5 μg/g, would cause severe malformations in Hoxa13 mutant limbs. This result suggests that the levels of RA produced in the Spd (Hoxd13) mutant limb may be lower than what is present in the Hoxa13 mutant limb, which would allow for higher dosages to be given to Spd mutants before causing a teratogenic effect. Spd limbs responding to 5 μg/g dosages do appear to be smaller in size, suggesting that some general inhibition of autopod growth may be present at this RA dosage. Alternatively, the limb tissues affected by the loss of HOXA13 function may be more sensitive to fluctuations in RA signaling than the cell types impacted by the Spd mutation resulting in more significant teratologic effects being elicited by RA in the Hoxa13 mutant limb. Finally, the regulation of Aldh1a2 by HOXA13 may also have a more global role, as functional HOXA13 is required for its own down-regulation and RA is known to suppress Hoxa13 expression (Knosp et al., 2004; Shou et al., 2005).

In summary, these studies identify Aldh1a2 as a novel target of HOXA13 in the developing autopod. In the absence of HOXA13 function, Aldh1a2 expression is reduced causing a subsequent reduction in active RA signaling in the interdigital tissues. Dietary supplementation with RA successfully bypasses the loss in Aldh1a2 expression, resulting in increased RA signaling and IPCD in the HOXA13 mutant hindlimb, confirming that HOXA13's regulation of Aldh1a2 plays essential role in mediating hindlimb IPCD. Previously, we identified Bmp2 and Bmp7 as pro-apoptotic factors regulated by HOXA13 in the interdigital mesenchyme (Knosp et al., 2004). These results in conjunction with the present study suggest that HOXA13 is functioning in the interdigital tissues to regulate multiple factors necessary for IPCD and digit separation. Moreover, these results provide supporting evidence that multi-step regulation of appendicular developmental cascades may be a canonical function for HOX proteins conserved in both invertebrate and vertebrate species.

EXPERIMENTAL PROCEDURES

Mouse Strains

The Hoxa13-green fluorescent protein (GFP) mouse line was maintained as heterozygous mutants on a C57BL/6J genetic background as described (Stadler et al., 2001; Perez et al., 2010). To assess retinoic acid signaling in the Hoxa13-GFP mutant limb, the Hoxa13-GFP mouse line was crossed three to five generations onto the CD1 genetic background (Charles River Labs, Wilmington, MA). Heterozygous Hoxa13-GFP-CD1 mice were then crossed with the transgenic RARE-Hspa1b/lacZ reporter mice from Jackson Laboratories (Bar Harbor, ME). Hoxa13-GFP-CD1; RARE-Hspa1b/lacZ compound heterozygous mice were intercrossed to produce Hoxa13-GFP-CD1 homozygous mutants embryos that were also transgenic for the RARE-Hspa1b/lacZ reporter. All experimental procedures, handling, and housing of the Hoxa13-GFP mouse colony were done in accordance to an approved animal protocol (OHSU protocol number: IS00001648 to HSS).

In Situ Hybridization

The mouse Aldh1a2 riboprobe was generated from a RNA expression plasmid containing nucleotides 1086–1717 of the murine Aldh1a2 gene (NM_009022). The mouse Cyp26b1 riboprobe was generated from a RNA expression plasmid containing nucleotides 1821–2304 of the murine Cyp26b1 gene (NM_175475.3). The analysis of Hoxa13 expression in the limb was accomplished using a Hoxa13 riboprobe as previously described (Morgan et al., 2003). Riboprobe synthesis and hybridization were performed on littermate or age-matched embryos as described by Manley and Capecchi (1995). For all analyses, control and homozygous mutant limbs were treated identically using the same preparation and concentration of riboprobe, hybridization buffer, temperatures, and colorometric detection times. Embryos were photographed using a Canon EOS 40D digital camera attached to a Leica MZFLIII stereo microscope. Individual images were assembled into limb photo montages using Adobe Photoshop CS4.

qRT-PCR Analysis

RNA was isolated from E13.5 Hoxa13-GFP mutant and control limb buds using an RNeasy RNA extraction kit as described by the manufacturer (Qiagen, Valencia, CA). Reverse transcription of the limb RNA and subsequent cDNA synthesis was accomplished using the IMPROM II cDNA synthesis kit following the manufacturer's instructions (Promega, Madison, WI). Quantification of Aldh1a2 expression in E13.5 Hoxa13-GFP wild-type and homozygous mutant limbs was accomplished using an IQ5 optical thermocycler and IQ SYBR® Green Supermix as recommended by the manufacturer (Bio-Rad, Hercules, CA). The following primers were used for the qRT-PCR reactions:

  • GAPDH FWD1 CCA CCC AGA AGA CTG TGG AT
  • GAPDH REV1 TTC AGC TCT GGG ATG ACC TT
  • Aldh1a2-FWD CAG CAA GGA ACA TGG CAA GGA G
  • Aldh1a2-REV TCA GCT TCT CCA GCA CAG CA

For each genotype, six independent qRT-PCR analyses were performed. Average fold change differences and standard deviation were calculated using Microsoft Excel.

Chromatin Immunoprecipitation

Chromatin immunoprecipitation (ChIP) was performed using a HOXA13 antibody and limb autopods dissected from E12.5 embryos as described (Knosp et al., 2004). Autopod tissues were dissected in phosphate buffered saline (PBS) containing 15 μL/ml protease inhibitor cocktail (PIC) (Sigma, St. Louis, MO.) and fixed in 1% formaldehyde/PBS and rocked at room temperature for 10 min. Protein-DNA cross-linking was stopped by the addition of glycine to a final concentration of 0.125 M for 5 min. Next, the samples were homogenized using a 2-ml dounce homogenizer followed by low speed centrifugation and a pellet wash using ice-cold PBS containing PIC. The pellet was re-suspended in 100 μl cell lysis buffer (5 mM PIPES, pH 8.0 / 85 mM KCl / 0.5% NP40) plus PIC and incubated on ice for 10 min. Next, the lysate suspension was micro-centrifuged at 5,000 rpm for 5 min at 4°C, followed by resuspension in 50 μl nuclear lysis buffer (50 m Tris-HCl, pH 8.1, 10 mM ehtylenediaminetetraacetic acid [EDTA], 1% sodium dodecyl sulfate [SDS]) plus PIC and incubation for 10 min on ice. The lysed nuclei were sonicated for 20 periods of 30 sec ON and 1 min OFF at 4°C using a Bioruptor (Diagenode, Delville, NJ) to produce sheared chromatin of an average length of 200–1,000 bp. The sheared chromatin was micro-centrifuged at 13,000 rpm for 10 min at 4°C and the supernatant was transferred to a new tube. ChIP was performed using a ChIP Assay Kit as described by the manufacturer (Upstate Biotechnologies/Millipore). Each chromatin supernatant was precleared with 40 μl of Salmon Sperm DNA/ Protein A agarose (Upstate Biotechnology). The chromatin samples were incubated with the Hoxa13 antibody or IgG control antibody on a rotating platform at 4°C for 3 hr. Washes, DNA elution, and reverse cross-linking were performed as described in the Upstate ChIP Assay Kit. Samples were ethanol precipitated, resuspended in 100 μl of TE, and DNA purified using the QIAquick PCR Purification kit (Qiagen).

The eluted DNA from samples incubated with the HOXA13 antibody or negative control IgG and were assessed for the presence of a conserved Aldh1a2 cis-element containing several high affinity HOXA13 binding sites (UCSC Genome Browser, Santa Cruz, CA) by PCR using two primer sets. Primer Set 1: (Amplicon: Chr9:71062106–71062662)

5′-CCCTCGCTCCTTCTCTCCTGTA-3′ and 5′-ACTTCTTCTTCCTAGAGCAACA-3′ and Primer set 2: (Amplicon: Chr9:71062106–71062443) 5′-CCCTCGCTCCTTCTCTCCTGTA-3′ and 5′-CCAAGCAACTGCTTCGCTCTT-3′. PCR conditions used 50 ng of template chromatin DNA and 35 cycles of 94 degrees Celsius for 30 sec, 56 Celsius for 30 sec and 72 Celsius for 30 sec. ChIP-positive PCR amplicons were fractionated on a 2% NuSieve GTG Agarose gel (Lonza) and photographed using a Li-Cor Biosciences (Lincoln, NE) gel documentation system.

Luciferase Assays

NG108-15 cells (mouse neuroblastoma and rat glioma somatic cell hybrid, ATCC) were grown in 60-mm tissue culture dishes containing DMEM without sodium pyruvate (CellGro) supplemented with 5% fetal bovine serum (Invitrogen), 100 U/ml penicillin, 100 μg/ml streptomycin (Invitrogen), and 1× HAT (ICN). At 90% confluence, the cells were passaged into 12-well dishes (Costar) and grown at 37°C, 10% CO2 for 24 hr. After 24 hr, the experimental transfections consisted of: 0.5 μg of a pGL3-promoter plasmid (Promega) containing the 550 bp Aldh1a2 region detected by ChIP to be bound by HOXA13, 1, 2, or 4 μg of the HOXA13 expression plasmid pCAGGSHOXA13 previously described (Knosp et al., 2007), and 0.1 μg of a Renilla Luciferase expression plasmid (Promega) to normalize for transfection efficiency. Control transfections consisted of: 0.5 μg of the empty pGL3-promoter vector, 1, 2, or 4 μg of the HOXA13 expression plasmid, and 0.1 μg of the Renilla luciferase expression plasmid to normalize for transfection efficiency. All transfections used the Superfect transfection reagent as recommended by the manufacturer (Qiagen). Forty-eight hours after transfection, cells were rinsed with 1× PBS and lysed with 150 μl/well M-Per lysis reagent (Pierce, Rockford, IL). Cell lysates were processed to detect luciferase activity using the Dual-Glo Luciferase Assay System (Promega) in OptiPlate-96F black plates (Perkin-Elmer). Luciferase activity was determined as the average of three separate readings of each well (1 sec/read) using a Packard Fusion Universal Microplate Analyzer (Perkin Elmer, Waltham, MA). For each experimental condition, four separate transfections were performed. Results were normalized for transfection efficiency using relative Renilla luciferase expression levels as described by the manufacturer (Promega). A student's t-test was used to determine statistically significant differences in normalized luciferase activity. Data analysis and plotting were accomplished using Microsoft Excel.

RA Supplementation

RA (R2625 Sigma, St. Louis, MO) was solubilized in absolute anhydrous ethyl alcohol and diluted to dosages of 2.5–4.0 μg/g mouse body weight as described by Kuss et al. (2009) with the exception that corn oil was used as the vehicle. The RA dosages were administered by oral gavage once daily until the embryos were collected and processed for the analysis of interdigital programmed cell death using the Lysotracker reagent as recommended by the manufacturer (Invitrogen/Molecular Probes, Eugene, OR). To assess changes whether RA supplementation could affect RA signaling, RA (R2625 Sigma, St. Louis, MO) was solubilized in absolute, anhydrous ethyl alcohol and coated onto PicoLab® Diet 5053 food pellets at a final concentration of 0.5mg RA per gram of food. The PicoLab® Diet 5053 contained no reported RA but was supplemented with 25 IU Vitamin A per gram of food. RARE-LacZ pregnant females with embryos at E 8.5 consumed the RA-treated pellets ad libitum with fresh RA-treated pellets being provided each evening for four-six days depending on the final gestational age needed for analysis. At the end of the treatment, the embryos were collected and processed for LacZ staining.

β-Galactosidase (LacZ) Staining

In vivo retinoic acid signaling was detected using Hoxa13-GFP wild-type or homozygous mutant mice bearing the transgenic Hspa1b/lacZ reporter allele. Embryos collected at E11.5–E13.5 were fixed on ice in 2% paraformaldehyde, 2 mM MgCl2, 2 mM EGTA, and 0.1 M PIPES for 25 min. After fixing, the embryos were washed 3 × 10 min in phosphate buffered saline containing 2 mM MgCl2, 0.02% NP-40 (Tergitol), and 0.01% sodium deoxycholate. After washing, the embryos were placed in phosphate buffered saline staining solution a containing 1 mg/ml X-gal, 5 mM K3Fe(CN)6, 5 mM K4Fe(CN)6, 2 mM MgCl2, 0.02% NP-40, 0.01% sodium deoxycholate and allowed to stain for 6–24 hr at 37°C.

Mutant and wild-type embryos were stained for identical times. After staining, the embryos were fixed in 2% paraformaldehyde /PBS for 2 hr followed by a PBS rinse and clearing with 50% glycerol/PBS. After clearing, the embryos were photographed using a Leica MZFLIII stereoscope fitted with a Canon EOS 40D camera.

ACKNOWLEDGMENT

H.S.S. was funded by the Shriners Hospital for Children.

Ancillary