Dimeric combinations of MafB, cFos and cJun control the apoptosis-survival balance in limb morphogenesis

During vertebrate development, apoptosis is a crucial process that sculpts the shape of embryos. In this regard, vertebrate limbs provide one of the best models in which to study apoptosis, as it has major morphogenetic functions. In avian limbs, the elimination of cells is observed in four distinct mesenchymal areas: the anterior necrotic zone (ANZ), posterior necrotic zone (PNZ), opaque patch (OP) and interdigital necrotic zones (INZs). In the ectoderm, cells are eliminated in the apical ectodermal ridge (Todt and Fallon, 1986; Zuzarte-Luis and Hurle, 2002). Apoptosis in these regions is particularly important because disruptions in apoptosis affect the final morphology of the limb (Ros et al., 1997; Yokouchi et al., 1996; Zuzarte-Luis and Hurle, 2002). Numerous studies have shown that bone morphological proteins (BMPs) trigger apoptosis both in the mesodermal cells (Ganan et al., 1996; Macias et al., 1997; Tang et al., 2000; Yokouchi et al., 1996; Zou and Niswander, 1996) and in the ectoderm of the apical ectodermal ridge (Wang et al., 2004). Overexpression of a dominant-negative form of BMP receptor Ib or Ia in chick limb represses apoptosis and induces the formation of interdigital webs (Yokouchi et al., 1996; Zou and Niswander, 1996), whereas transplantation of a BMP protein-soaked bead in limb mesenchyme induces apoptosis in chick embryos (Ganan et al., 1996; Macias et al., 1997; Tang et al., 2000). However, the molecular cascades that connect BMP signalling with the expression of genes controlling apoptosis in limbs remain largely unclear.

The AP-1 transcription factor superfamily is well conserved, and its members regulate the transcription of genes involved in diverse cellular processes, including proliferation, transformation and death (Shaulian and Karin, 2002). During vertebrate embryogenesis, some AP-1 transcription factors are crucial for the proper development of bones and the lens, liver and heart (Ogino and Yasuda, 1998; Shaulian and Karin, 2002). AP-1 transcription factors are also known to modulate gene expression triggered by free radicals (Sen and Packer, 1996). Interestingly, high levels of reactive oxygen species (ROS) are present in the INZ, and antioxidants can reduce apoptosis in limb buds (Salas-Vidal et al., 1998; Schnabel et al., 2006), suggesting that ROS stress may trigger apoptosis in the INZ. Indeed, the AP-1 family members cFos and cJun are regulated by oxidants at both transcriptional and translational levels (Sen and Packer, 1996).

Here, we examine the molecular cascades that control apoptosis in developing limb buds and tested the involvement of AP-1 transcription factor members in this process. Global expression profiles of all 41 reported AP-1 superfamily members showed that expression of MafB was restricted to the area of apoptosis. Thus, we examined whether MafB, as well as its candidate heterodimer partners cJun and cFos (Deppmann et al., 2006), are involved in controlling apoptosis during limb morphogenesis. Gain- and loss-of-function experiments revealed that MafB/cFos heterodimers inhibited apoptosis, whereas MafB/cJun heterodimers promoted apoptosis for sculpting the shape of chick limb buds. Integrating chromatin immunoprecipitation sequencing (ChIP-Seq) and functional analyses showed that expression of the transcription factors p63 and p73 was directly activated by MafB and cJun. Our results suggest that the balance of MafB, cFos and cJun heterodimer combinations controls the number of apoptotic cells and that the MafB/cJun heterodimer triggers pathways that lead to apoptosis via activation of p63 and p73 during chick limb morphogenesis.

In vertebrates, AP-1 superfamily transcription factors are involved in many processes that are crucial to the function of an organism. Some are necessary for the proper development of several organs and tissues (Shaulian and Karin, 2002). We explored whether any genes that encode AP-1 transcription factors are involved in regulating apoptosis in chick limb buds. Expression patterns of genes encoding all 41 AP-1 members that are present in chickens were examined in chick wing buds at stage 26 and in leg buds at stage 31 (supplementary material Fig. S1). Of these, MafB transcripts were detected in the areas of apoptosis, including the ANZ, the PNZ and the apical ectodermal ridge at stage 26 (Fig. 1A) and the INZ at stage 31 (Fig. 1B), as previously reported (Lecoin et al., 2004). Based on their structural similarities, MafB was predicted to form heterodimers with either Fos or Jun family members (Deppmann et al., 2006). Among all Fos and Jun family members, only cFos and cJun transcripts were detected throughout the wing and leg buds of chick embryos, including apoptotic regions (Fig. 1C-F; supplementary material Fig. S1). Thus, we next examined whether MafB and its candidate heterodimer partners cFos and cJun were involved in regulating apoptosis during limb morphogenesis.

To investigate whether MafB has important roles in regulating apoptosis in chick limb buds, we introduced a retrovirus expressing the dominant-negative form of chicken MafB (DN-MafB), which lacks a transcriptional activation domain. Nile Blue staining of wing buds electroporated with the RCAS-DN-MafB construct showed that the number of apoptotic cells was reduced in the PNZ (5/6; dashed circle in Fig. 1G) and increased in the apical ridge (4/6; arrowheads in Fig. 1G), although control wing buds electroporated with a RCAS-AP vector did not display any changes in apoptosis (4/4; supplementary material Fig. S2A,B) or in the apical ridge formation (4/4; supplementary material Fig. S2C). These results suggest that AP-1 factor(s) that were inhibited by the RCAS-DN-MafB construct were involved in regulating the number of apoptotic cells in the apoptotic regions both in the mesenchyme and in the apical ectodermal ridge. By contrast, wing buds electroporated with RCAS-MafB showed almost no changes in apoptosis (6/7; supplementary material Fig. S2D,E), although they did form a considerably disturbed ridge (2/5; supplementary material Fig. S2F). Therefore, the phenotypes of wing buds electroporated with the RCAS-DN-MafB construct may be caused by suppression of multiple AP-1 factors, probably MafB and its heterodimer partner(s). We thus evaluated dimeric formation between endogenous MafB and cFos or cJun proteins using co-immunoprecipitation from lysates of stage 30 interdigital regions of distal leg buds. Maf did indeed interact with Fos and Jun in chick leg buds (Fig. 1H). We also examined whether MafB and cFos or cJun colocalized in chick limb buds by using antibodies against MafB, Fos, and Jun (supplementary material Fig. S3). Fos colocalized with MafB in interdigital regions of stage 30 chick leg buds (white arrowheads in supplementary material Fig. S3B). Likewise, Jun and MafB co-localized in interdigital regions at stage 30 (white arrowheads in supplementary material Fig. S3C). These results suggest that MafB proteins colocalize and interact with Fos and Jun proteins in interdigital regions of chick leg buds.

To investigate our hypothesis that MafB/cFos and MafB/cJun heterodimers are involved in regulating apoptosis in limb buds, we overexpressed MafB and either cFos or cJun constructs in limb buds. Electroporation of retrovirus constructs expressing MafB together with a human cFos-expressing pEF vector (Kataoka et al., 1994a) in the presumptive wing field led to the reduction of apoptotic cells in transfected areas, including the apical ridge (5/5; dashed circle in Fig. 1I), which caused the disruption of the apical ridge formation (5/7; Fig. 1J). By contrast, electroporation of RCAS-MafB, together with a human cJun-expressing pEF vector (Kataoka et al., 1994a) in the wing field caused ectopic apoptosis in the transfected regions (4/5; Fig. 1K-L′). These results suggest that, it in chick limb buds, MafB/cFos heterodimers repress apoptosis and MafB/cJun heterodimers promote it.

To confirm our hypothesis, we electroporated a Tol2-flanked construct expressing either MafB and cFos or MafB and cJun under the control of a bipolar CMV promoter (pT2AL-MafB-BI-CMV-cFos in Fig. 2A; pT2AL-MafB-BI-CMV-cJun in Fig. 2B) along with a Tol2 transposase-expressing vector (pCAGGS-T2TP) (Kawakami and Noda, 2004) and pCAGGS-EGFP in the presumptive wing field at stage 14. Apoptotic cells in wing buds were examined with LysoTracker 2 days after electroporation. In mesenchymal cells of the PNZ (dashed circle in Fig. 2C′) and in ectodermal cells of the apical ridge (dashed circle in Fig. 2E) that were electroporated with pT2AL-MafB-BI-CMV-cFos, apoptosis was clearly repressed (10/13; Fig. 2C-F). Apoptosis was, however, ectopically activated both in mesenchymal and ectodermal cells (arrowheads in Fig. 2G′) of wing buds that were electroporated with pT2AL-MafB-BI-CMV-cJun (11/17; Fig. 2G,H). Thus, the co-existence of MafB and cFos repressed apoptosis, whereas co-existence of MafB and cJun promoted apoptosis both in mesenchymal and ectodermal cells of chick limb buds.

We electroporated pEF-cFos or -cJun alone into the PNZ of presumptive wing fields in stage 14-16 chick embryos and observed the effect on apoptosis after 48 h. Interestingly, overexpression of pEF-cFos alone in the PNZ caused both mild increases and decreases in the number of apoptotic cells (n=5; supplementary material Fig. S2G,H) and did not lead to any obvious changes in the apical ridge formation (5/5; supplementary material Fig. S2I). Electroporation of pEF-cJun alone produced only a slight increase in the number of apoptotic cells in some cases (5/7; supplementary material Fig. S2J,K) and did not lead to any alterations in the apical ridge formation (6/6; supplementary material Fig. S2L). Unlike MafB, both cFos and cJun have more than 10 other possible dimeric partners (Deppmann et al., 2006). Thus, electroporated cFos or cJun might have dimerized with other AP-1 transcription factors and may have activated both pro- and anti-apoptotic genes.

To investigate the function of endogenous cFos or cJun, we used siRNAs targeted to chick cFos or cJun. We first evaluated the effect of cFos-siRNA and cJun-siRNA in vitro by transfecting COS7 cells with cFos-siRNA and vectors expressing cFos-GFP fusion protein (pAc-cFos-GFP; supplementary material Fig. S4A) or with cJun-siRNA and vectors expressing cJun-GFP fusion protein (pAc-cJun-GFP; supplementary material Fig. S4B). At 24 h after transfection, the number of GFP-positive cells transfected with cFos-siRNA or cJun-siRNA was significantly reduced when compared with control cultures (P<0.00001, Student's t-test; supplementary material Fig. S4A,B). We then investigated the effect of depleting cFos or cJun in limb development 48 h after co-electroporation of cFos- or cJun-siRNA together with pCAGGS-EGFP into the PNZ of the presumptive wing region at stage 14-16 (supplementary material Fig. S4C-H). Interestingly, electroporation of cFos-siRNA led to both mild increases and decreases in the number of apoptotic cells (n=5; supplementary material Fig. S4E,F). By contrast, electroporation of cJun-siRNA in the wing field led to minimal changes in apoptosis (n=5; supplementary material Fig. S4G,H). It is possible that the freed partners of cFos or cJun dimerized with other AP-1 transcription factors after introduction of the cFos- or cJun-siRNA, and may have upregulated both pro- and anti-apoptotic genes. Control embryos in which control siRNA and pCAGGS-EGFP were electroporated did not show any changes in apoptosis (n=3; supplementary material Fig. S4C,D).

As seen in supplementary material Fig. S1, genes encoding other AP-1 transcription factors were also expressed in the apoptotic regions (e.g. c-maf, Nfe2l1, Nfe2l2, Xbp1). We examined whether c-maf, Nfe2l1, Nfe2l2 and Xbp1 could affect the number of apoptotic cells by introducing vectors that express those genes into the PNZ of stage 14-16 chick wing buds (supplementary material Fig. S5). Misexpression of c-maf, Nfe2l2 and Xbp1 in wing buds showed hardly any changes in apoptosis (3/5, 5/6 and 4/5, respectively), whereas that of Nfe2l1 showed only a mild reduction in apoptosis in some cases (4/6; dashed circle in supplementary material Fig. S5C). Thus, it is possible that AP-1 transcription factors other than MafB, cFos and cJun, including Nfe2l1 and its unknown dimeric partners, are also involved in regulating the number of apoptotic cells in chick limb buds.

Apoptosis in limbs is controlled by BMP signalling (Ganan et al., 1996; Macias et al., 1997; Tang et al., 2000; Yokouchi et al., 1996; Zou and Niswander, 1996). We thus investigated whether expression of MafB, cFos and cJun was regulated by BMP signalling. For this purpose, we inserted a bead soaked with BMP2, Noggin or PBS into stage 28 leg buds. Twenty-four hours after implantation, expression of MafB was activated by BMP2 (5/8; arrowheads in Fig. 3A) and repressed by Noggin (6/7; dashed circle in Fig. 3B), suggesting that MafB expression in the INZ is controlled by BMP signalling. By contrast, expression of cFos and cJun showed no changes after implantation of a bead soaked with BMP2 (5/5, Fig. 3C; 5/5, Fig. 3E) or Noggin (5/5, Fig. 3D; 4/4, Fig. 3F). No changes were observed in the expression of MafB, cFos or cJun after implantation of a PBS-soaked bead (4/4, 5/5 and 5/5, respectively; supplementary material Fig. S6A-C). These results suggest that expression of MafB, but not cFos or cJun, is controlled by BMP signalling. Similarly, ectopic MafB expression was induced in the PNZ of stage 24 wing buds 24 h after implantation of a BMP-soaked bead (7/7; Fig. 3G). By contrast, no changes were seen in MafB expression after application of a PBS-soaked bead (4/4; supplementary material Fig. S6D). We then investigated expression of MafB 3 h after implantation of a BMP2-soaked bead in the INZ at stage 28. As seen in Fig. 3H, ectopic MafB expression was induced around a BMP2-soaked bead (6/6; Fig. 3H), but not around a PBS-soaked bead (5/5; supplementary material Fig. S6E). Therefore, MafB is most likely directly regulated by Smads, phosphorylated by serine/threonine kinases BMP receptors, or through only a few intermediate transcription factor(s), in this signalling pathway.

Retinoic acid signalling is specifically active in the interdigital region (Lussier et al., 1993; Rodriguez-Leon et al., 1999) and is sufficient to trigger interdigital apoptosis (Alles and Sulik, 1989; Dupe et al., 1999; Hernandez-Martinez et al., 2009). To see whether MafB expression is controlled by retinoic acid signalling in chick limb buds, we implanted a retinoic acid-soaked bead into the PNZ of stage 21-22 wing buds and into the INZ of stage 28 leg buds and examined expression of MafB (Fig. 3I,J). Twenty-four hours after application of a retinoic acid-soaked bead, expression of MafB was weakly activated around the bead in the PNZ (3/4; arrowheads in Fig. 3I) and slightly activated in a single case in the INZ (1/9; arrowheads in Fig. 3J). By contrast, no obvious changes were observed after implantation of a DMSO-soaked bead in the PNZ and in the INZ (5/5 and 6/6, respectively; supplementary material Fig. S6F,G). Thus, expression of MafB seems to be only weakly activated by retinoic acid in early limb buds.

In the hematopoietic system, MafB is expressed at high levels in macrophages and promotes macrophage differentiation (Bakri et al., 2005; Kelly et al., 2000). To investigate whether MafB expression in the limb buds is due to the presence of macrophages, we examined the distribution of macrophages and MafB proteins in interdigital regions of stage 30 leg buds (supplementary material Fig. S3D,E). Double staining of leg buds with macrophage and MafB antibodies showed that MafB proteins were distributed in macrophages (white arrowheads in supplementary material Fig. S3D). MafB proteins were also distributed in nuclei of several non-macrophage cells in interdigital regions (open arrowheads in supplementary material Fig. S3E).

We then decided to look for downstream targets of MafB/cJun heterodimers that promote apoptosis in chick limb buds. To identify target genes and regulatory elements controlled by Maf and Jun, we dissected the distal region of 60 leg buds from stage 30 chick embryos (Fig. 4A) and performed ChIP using these tissue samples with Maf, Jun or p300 antibodies (Fig. 4C-E). Epigenomic profiling of p300, a transactivational co-activator, identifies active enhancers (Visel et al., 2009). Immunoprecipitated DNA fragments were analysed using massively parallel sequencing, and the resulting 50 base pair (bp) sequence reads were aligned with the reference chick genome (galGal3).

In leg buds, we identified 26,009, 29,815 and 18,303 regions of the genome with substantial enrichment in sequences associated with Maf, Jun and p300, respectively (in all cases, the presence of binding sites was indicated by fold enrichment >2.5 and peak width >400 bp; supplementary material Table S2). To identify direct targets of MafB/cJun heterodimers in chick limb buds, we looked for Maf and Jun binding sites that were coincident with p300 binding sites and identified 1347 peaks (supplementary material Table S3). Of these, the transcriptional start sites (TSSs) of 1717 genes resided within ±50 kb of these identified peaks, and these genes were highly expressed in limb buds based on published transcriptome data (normalized counts ≥100; supplementary material Table S4) (Wang et al., 2011). We considered genes that were within ±50 kb of these identified peaks as potential target genes of MafB/cJun heterodimers in chick limb buds. To select candidates among potential target genes of AP-1 (supplementary material Table S4), we looked for genes encoding core components that regulate the cell cycle and/or apoptosis (supplementary material Table S5); we identified p73, Chk1 and DAPK1 (Fig. 4D; supplementary material Fig. S7). An AP-1 binding site was also found near the TSS of caspase 6, but we did not choose caspase 6 as a candidate because Casp6 knockout mice exhibit normal limb phenotype (Zheng et al., 1999). In addition, we selected p63 as a potential AP-1 target, as p63 belongs to the same family as p73 (Fig. 4C). Furthermore, a possible binding site for AP-1 proteins that was coincident with a p300 binding site was identified within ±50 kb of TSSs of p63, and transcription of p63 is controlled by AP-1 transcription factors (Leventaki et al., 2007; Vartanian et al., 2011; Yao et al., 2010). To see whether these candidates are expressed in the apoptotic regions, we examined their expression patterns in stage 26 chick wing buds and stage 31 chick leg buds (Fig. 4F-K; supplementary material Fig. S7). Of these, p63, p73 and Chk1, but not DAPK1, were highly expressed in chick limb buds, including apoptotic regions (Fig. 4F-K) such as the apical ridge (p63 in Fig. 4F) and the mesenchyme (p73 and Chk1 in Fig. 4H-K). Chk1 initiates cell cycle arrest when ATM and/or ATR recognizes DNA damage, and they induce transcription of p73, which mediates apoptosis (Fig. 4B; Urist et al., 2004). p63 can also induce apoptosis (Fig. 4B; Pyati et al., 2011).

To investigate whether MafB/cJun control the transcription of candidate target genes during limb development, we misexpressed RCAS-MafB together with pEF-cJun in the presumptive wing field at stage 14 and examined the expression patterns of potential target genes (Fig. 5A,B; supplementary material Fig. S8A). Thirty-six hours after misexpression, p63 and p73, but not Chk1, were ectopically expressed in the region electroporated with RCAS-MafB, pEF-cJun and pCAGGS-RFP (p63, 6/6, arrowheads in Fig. 5A; p73, 5/7, arrowheads in Fig. 5B; Chk1, 0/6, see supplementary material Fig. S8A). Expression of p63 and p73 was not altered after misexpression of control constructs (5/8 and 2/3, respectively; supplementary material Fig. S8B,C). These results suggest that expression of p63 and p73 is controlled by MafB/cJun.

Binding profiles suggest that AP-1 binds within the first intron of p63 (BS-p63; Fig. 4C) and 8.5 kb upstream of p73 (BS-p73; Fig. 4D). To confirm binding of p300 in these regions, we carried out ChIP-qPCR analysis with stage 30 chick leg buds using anti-p300 (supplementary material Fig. S9). ChIP-qPCR analyses showed distinct degrees of enrichment of p300 binding to BS-p63 and BS-p73 (supplementary material Fig. S9). These results suggest that p300 binds to these AP-1 binding sites.

To test the functions of these AP-1-binding sites, each identified site was cloned directly in front of a minimal promoter that is followed by a luciferase reporter to test whether it is responsive to MafB/cJun transactivation (Fig. 5C,D). Transient reporter assays showed that the enhancer activities of BS-p63 and BS-p73 were highly activated by co-expression of cJun and MafB (Fig. 5C,D), suggesting that BS-p63 and BS-p73 can function as enhancer elements to control gene expression in limb buds, and that their activities are dependent on MafB and cJun. These results suggest that MafB/cJun heterodimers directly activate p63 and p73 and promote apoptosis. Taken together, our data demonstrate that dimeric combinations of MafB, cFos and cJun control the apoptosis-survival balance and that MafB/cJun heterodimers trigger apoptosis by activating expression of p63 and p73 during chick limb morphogenesis (Fig. 6).

We showed that MafB/cJun promotes apoptosis both in mesenchymal cells and in the apical ectodermal ridge, although it is likely that this heterodimer acts through different targets in the two regions. For example, one target, p63, is expressed only in the ectoderm. A p63/p73-responsive element is present in the second intron of Jagged1 (Sasaki et al., 2002), which is one of a few components known to control apoptosis in the apical ridge (Francis et al., 2005). Furthermore, p63 and p73 are required for triggering p53-dependent apoptosis in response to DNA damage (Flores et al., 2002). By contrast, p63 also directly upregulates the expression of Dlx5 (Kouwenhoven et al., 2010), which promotes proliferation in the median apical ridge (Robledo et al., 2002), and p63 knockout mice fail to form the apical ectodermal ridge (Mills et al., 1999; Yang et al., 1999). p63 in the apical ectodermal ridge likely activates multiple genes, as it is known to do in other systems (Melino et al., 2003). Such genes may be involved in pro-apoptotic cascades (including Jagged1) and proliferation (including Dlx5), and the sum of all signals is likely to determine the fate of cells. Further studies are required to determine the detailed functions of p63 targets that regulate the formation of the apical ridge. p73 contributes to p53-independent apoptosis by directly activating expression of PUMA, Bax and p21 (Urist et al., 2004). p73 also activates transcription of Noxa, mutation of which causes mitochondrial dysfunction (Flinterman et al., 2005). In agreement with these p73 functions, Bax is specifically expressed in the INZ (Dupe et al., 1999), and mice deficient in both Bax and Bak show soft-tissue syndactyly (Lindsten and Thompson, 2006). Thus, in apoptotic regions of chick limb buds, p73 very likely triggers a p53-independent apoptosis cascade probably by activating Bax. However, in mice, the absence of p73 does not affect limb development (Yang et al., 2000). This may be because levels of p73 transcripts are very low in mouse limb buds (EMBRYS; http://embrys.jp/embrys/html/MainMenu.html; Yokoyama et al., 2009) when compared with those in chick limb buds (Fig. 4H,I). In this study, we looked for only direct targets of MafB/cJun, the functions of which could explain the phenotypes noted here; it is also reasonable to expect that MafB/cFos directly activate pro-apoptotic genes. Importantly, several other AP-1 family members activate both pro-apoptotic and anti-apoptotic genes (Shaulian and Karin, 2002). Indeed, Shaulian and Karin (2002) proposed that there is a ‘balance between the pro-apoptotic and anti-apoptotic target-genes that determines whether the final outcome will be cell survival or cell death’. Hence, we must consider the possibility that MafB/cFos and Maf/cJun activate both pro- and anti-apoptotic genes, and that the balance between these activated genes determines the outcome. Furthermore, MafB/cFos and/or MafB/cJun are likely also to inhibit the expression of pro-apoptotic or survival genes, and cFos/cJun combinations may also control the expression of genes throughout limb buds. For example, MafB/cJun binds to the regulatory region of Hoxb9 to repress its transcription (Mechta-Grigoriou et al., 2003), and cFos/cJun stably binds to regulatory elements in many other systems (Shaulian and Karin, 2002). Although MafB seems to dimerize with either cFos or cJun in chick limb buds (Fig. 1H; supplementary material Fig. S3B,C), both cFos and cJun have more than 10 other possible dimer partners (Deppmann et al., 2006). As we have shown in supplementary material Figs S2 and S4, overexpression or knockdown of cJun alone produced only a slight increase or decrease in the number of apoptotic cells in some cases. Moreover, overexpression or knockdown of cFos alone both increased and decreased the number of apoptotic cells (supplementary material Fig. S2G,H and Fig. S4E,F). These results suggest that cFos and cJun can dimerize with other AP-1 transcription factors and activate both pro- and anti-apoptotic genes. It is also plausible that other combinations of AP-1 transcription factors have roles that are similar to those of MafB, cFos and cJun in regulating apoptosis during limb morphogenesis. In fact, misexpression of Nfe2l1 caused mild repression of apoptosis in chick limb buds (supplementary material Fig. S5). Although overexpression of c-maf, Nfe2l2 or Xbp1 resulted in minimal changes in the number of apoptotic cells (supplementary material Fig. S5), it is still possible that they are also involved in this process by forming heterodimers or are involved in the development of different structures in the embryo. Future studies should clarify the complex regulation of AP-1 factors in various morphological processes.

In this study, we show that MafB is involved in controlling apoptosis in chick limb buds. Importantly, a recent study in mice revealed a novel function for MafB in inhibiting foam-cell apoptosis, which in turn accelerates atherogenesis (Hamada et al., 2014). By contrast, MafB-deficient mice have phenotypically normal limbs (Satoru Takahashi, personal communication). Notably, however, the expression of many AP-1 family members, including MafB and cFos, differs between the limb buds of chick and mouse embryos (supplementary material Fig. S1; EMBRYS; Yokoyama et al., 2009). In mice, MafB is not expressed in the apical ridge and cFos transcripts are not detectable in limb buds (EMBRYS; Yokoyama et al., 2009). Furthermore, mouse limb buds show very few anterior and posterior mesenchymal regions that undergo apoptosis comparable with that observed in the chick. Such differences may explain why MafB mutant mice show relatively normal limb phenotypes. It is therefore likely that other AP-1 transcription factors in mouse limb buds play roles similar to those of MafB and cFos in chick limb buds.

The dimerization of Maf with Fos or Jun probably depends on the amount of each of these proteins in a cell and their relative binding affinities. Protein dimerization rates depend on the amounts of the respective molecules and on the Arrhenius activation energies for their dimerization (Ecevit et al., 2010). In fact, dose-dependent heterodimer formation of Maf with Fos and Jun has been demonstrated with a protein dimerization assay (Kataoka et al., 1994b).

AP-1 transcription factors control gene expression induced by free radicals (Sen and Packer, 1996). Interestingly, high levels of ROS have been detected in the INZ of limb buds, and the number of apoptotic cells was reduced after treatment with antioxidants in limb buds (Salas-Vidal et al., 1998; Schnabel et al., 2006). Moreover, both cFos and cJun are regulated by oxidants at both the transcriptional and translational levels (Sen and Packer, 1996). It is thus fascinating to speculate that transcription of MafB is induced by ROS stress in chick limb apoptotic regions; future studies will be required to link ROS stress and MafB functions during apoptosis. In this study, we show that expression of MafB is regulated by BMP signalling in limb buds. As MafB is able to either activate or inhibit apoptosis depending on its heterodimer partner (i.e. cJun or cFos, respectively), BMP probably also activates other factors that trigger apoptosis in limbs. During osteoblast differentiation, BMP induces a Nox4-derived ROS production (Mandal et al., 2011). Levels of ROS and/or antioxidant enzymes may also be regulated by BMP in limb buds as suggested previously (Schnabel et al., 2006). We have thus developed a model in which AP-1 factors are involved in controlling apoptosis in developing limb buds (Fig. 6). MafB/cFos heterodimers reduced the numbers of dying cells in apoptotic regions of chick limb buds. MafB/cJun heterodimers trigger apoptosis in chick limb buds, and they directly activate expression of p63 and p73, key factors known to arrest the cell cycle, leading to the activation of apoptosis.

BMP signalling activates expression of MafB and possibly upregulates the level of ROS in apoptotic regions (Schnabel et al., 2006). ROS stress itself may activate the expression of the AP-1 factors MafB, cFos and/or cJun in limb buds, as shown in other systems (Sen and Packer, 1996). We conclude that the balance of MafB and cFos or cJun heterodimer combinations, specifically in apoptotic regions, controls the apoptosis-survival balance, and MafB/cJun heterodimers trigger apoptosis by activating expression of p63 and p73 during limb patterning in chick embryos.

The A9300001N09Rik, Atf1, Atf2, Atf3, Atf4, Atf6, Bach1, Bach2, Batf, Batf3, CEBPd, CEBPg, Creb1, Creb3, Creb3l1, Creb3l2, Creb3l3, Creb5, Crebl2, Crem, cFos, HLF, cJun, Nfe2l1, Nfe2l2, Nfil3 and Xbp1 probes were transcribed from Biotechnology and Biological Sciences Research Council (BBSRC; http://www.chick.manchester.ac.uk/) chick EST cDNA clones (ChEST595m10, ChEST547e5, ChEST913d17, ChEST927o11, ChEST390a19, ChEST222o13, ChEST926l10, ChEST1024h14, ChEST543k2, ChEST372f15, ChEST704p8, ChEST86h12, ChEST50d14, ChEST256g16, ChEST335l1, ChEST553f15, ChEST98d7, ChEST964l8, ChEST860a20, ChEST72i15, ChEST262m21, ChEST837k20, ChEST520g15, ChEST719p8, ChEST813j4, ChEST567h3 and ChEST151f18, respectively) (Boardman et al., 2002; Hubbard et al., 2005). Atf7, CEBPa, CEBPb, Fosl2 and Jund probes were transcribed from the University of Delaware Chick EST cDNA clones (pgf1n.pk008.b16, pgl1n.pk007.g3, pgm2n.pk007.i18, pgn1c.pk014.g7 and pgl1n.pk012.d17, respectively; http://www.chickest.udel.edu/). Crebzf, JDP, MafF, MafG, MafK, TEF, p63, p73, Chk1 and DAPK1 cDNAs for in situ probes were amplified from cDNA pools prepared from stage 20-28 chick embryos using primers that hybridize to the published sequences described in supplementary material Table S1. The L-Maf, MafB, c-maf and Fgf8 probes were transcribed as described previously (Crossley et al., 1996; Kataoka et al., 1994a). Whole-mount in situ hybridization was performed as described previously (Wilkinson, 1992).

For RCAS-MafB, the open reading frame (ORF) of chick MafB was amplified using primers that include the EcoRI, Kozak and FLAG epitope sequences at the N terminus and the HindIII sequence at the C terminus (5′-CCGGAATTCCCCACCATGGACTACAAGGACGATGACGATAAGAT-CGGCTCCCACCAAGTGTC-3′ and 5′-CCCAAGCTTGCACTCACATG-AAGAACTCG-3′; all PCR primers given as forward and then reverse). The resulting PCR products were inserted into the pGEM-T Easy vector (Promega; pGEMTeasy-MafB) and then cloned into the EcoRI and HindIII sites of SLAX12 NCO (Morgan and Fekete, 1996). A Clal fragment was cloned into the Clal site of the RCAS-AP retroviral vector (Homburger and Fekete, 1996; Hughes et al., 1987). For DN-MafB, the cDNA sequence encoding amino acid residues 111 to 311 of chick MafB was amplified with PCR using primers including the EcoRI, Kozak and FLAG epitope sequences at the N terminus and the HindIII sequence at the C terminus (5′-CCGGAATTCCCCACCATGGACTACAAGGACGATGACGATAAG-GGTGGTATGGCCGGAGAGCTGAGCATC-3′ and 5′-CCCAAGCTTGC-ACTCACATGAAGAACTCG-3′), inserted into the pGEM-T Easy vector and then cloned into SLAX12 NCO. A Clal fragment was cloned into the Clal site of the RCAS-AP retroviral vector. For pEF-MafB, the ORF of chick MafB was amplified using specific primers (5′-ATCCTCTAGAGTCGACCCA-CCATGGACTACAAGGA-3′ and 5′-TAAAGGGAAGCGGCCGCACT-CACATGAAGAACTCG-3′) and cloned into the Notl and Sall sites of the pHyg-EF2 vector (Nishizawa et al., 2003) via the In-Fusion reaction (Clontech). For pT2AL-MafB-BI-CMV-cFos and pT2AL-MafB-BI-CMV-cJun, pBI-CMV1 (Clontech) was digested with SacI and SalI (for pBSK-CMV1) or with XhoI and XbaI (for pBSK-CMV2) and subcloned into the pBSK⁻ vector (pBSK-CMV1 and pBSK-CMV2, respectively). Then, a polyA tail sequence was amplified with specific primers (for pBSK-CMV1, 5′-TAAAGCGGCCGCCACCGCGG-3′ and 5′-ACGCGTCGACTAGAT-CTGATCTAGAGGATC-3′; for pBSK-CMV2, 5′-CGCGGATCCAAGC-TTGATGATCCAGACATGATAAG-3′ and 5′-GGACTAGTCTCGAGG-ATCTAGAGGATCATAATCA-3′) using pT2AL200R150G (Urasaki et al., 2006) as a template and cloned into the NotI and SalI sites of pBSK-CMV1 (pBSK-CMV1-pA) and the BamHI and SpeI sites of pBSK-CMV2, to incorporate the polyA tail sequence at the C terminus (pBSK-CMV2-pA). Then, SLAX12 Nco-cFos and SLAX12 NCO-cJun were digested with HindIII and ClaI, and the ORFs of human cFos and cJun were cloned into the pBSK⁻ vector. The ORFs of human cFos and cJun were then subcloned into the NotI and ClaI sites of pBSK-CMV1-pA (pBSK-CMV1-cFos-pA and pBSK-CMV2-cJun-pA, respectively). Then, the ORF of chick MafB from pGEMTeasy-MafB was digested with EcoRI and HindIII and cloned into the EcoRI and HindIII sites of pBSK-CMV2-pA (pBSK-CMV2-MafB-pA). pBSK-CMV1-cFos-pA and pBSK-CMV2-cJun-pA were digested with Xhol and Bglll, and cloned into pT2AL200R150G. pBSK-CMV2-MafB was digested with XhoI, and the DNA fragment was cloned into the XhoI site of pT2AL200R150G-CMV1-cFos-pA and pT2AL200R150G-CMV2-cJun-pA (pT2AL-MafB-BI-CMV-cFos and pT2AL-MafB-BI-CMV-cJun, respectively) using the In-Fusion reaction. The genomic regions of the AP-1-binding site peaks were amplified with PCR using primers described in supplementary material Table S1 and were cloned into a minimum promoter containing the pGL4.23 [luc2/miniP] vector (Promega).

The plasmid solutions or small interfering RNAs (siRNAs; see below) were coloured with 2.5-5% Fast Green and co-electroporated with pCAGGS-EGFP (a gift from Dr J. Miyazaki and Dr T. Ogura) (Niwa et al., 1991) or pCAGGS-RFP (Das et al., 2006) into the presumptive limb mesenchyme as described previously (Ogura, 2002). Briefly, the solution was injected into the lateral plate mesoderm of stage 14-16 chick embryos, and two pulses of 8-12 V and 80-90 ms were applied. The concentrations of the prepared plasmid solutions and siRNAs were as follows: RCAS-MafB, 6.2 mg/ml; pEF-cFos, 6.6 mg/ml; pEF-cJun, 6.4 mg/ml; pCAGGS-EGFP, 3.5 mg/ml; pCAGGS-RFP, 6.0 mg/ml; pT2AL-MafB-BI-CMV-cFos, 5.2 mg/ml; pT2AL-MafB-BI-CMV-cJun, 6.9 mg/ml; pCAGGS-T2TP, 8.0 mg/ml; RCAS-AP, 5.5 mg/ml; pHygEF2, 4.7 mg/ml; pEF-c-maf, 9.8 mg/ml; pMSCV-Nfe2l1, 7.0 mg/ml; pcDNA3-Nfe2l2, 8.1 mg/ml; pCMV-Xbp1, 6.4 mg/ml; control siRNA, 250 μM; cFos siRNA, 250 μM; and cJun siRNA, 250 μM.

The morphology of the limbs that were subjected to experimental manipulation was studied after cartilage staining with Alcian Blue. The pattern of apoptosis was analysed with vital staining with Nile Blue and LysoTracker Red (Invitrogen) (Mariani et al., 2008).

Co-immunoprecipitation (co-IP) was carried out as described previously (Chen et al., 2004). Briefly, interdigital regions of stage 30 limb buds were dissected from 10 embryos in PBS, and lysed in lysis buffer [10 mM HEPES (pH 7.5); 420 mM NaCl; 0.5% NP-40; 0.2% protease inhibitors (BioVision)] on ice for 25 min with trituration. Lysates were centrifuged at 2500 g for 5 min, and the supernatants were transferred into new tubes. The Maf antibody (Santa Cruz; sc-7866), the Fos antibody (Santa Cruz; sc-52), the Jun antibody (Santa Cruz; sc-1694) or the normal rabbit IgG (PeproTech EC) used for co-IP was covalently cross-linked to Dynabeads (Veritas), and bound proteins were detected on western blots with affinity-purified anti-Maf.

CM Affi-Gel Blue Gel (Bio-Rad) beads were loaded with 0.1 mg/ml BMP2 (R&D Systems), 1 mg/ml Noggin (R&D Systems) or PBS (Tumpel et al., 2002). AG1-X2 resin beads (Bio-Rad) were loaded with 10 mg/ml all-trans retinoic acid (Sigma) or dimethyl sulfoxide (DMSO). Then a single bead was inserted into the PNZ of stage 21-22 chick wing buds or the interdigital area of stage 28 chick embryos. Embryos were fixed 24 h later in 4% paraformaldehyde and then subjected to in situ hybridization as described above.

Fluorescent immunostaining was performed on cryosections of limb buds at stage 30 as described previously (Bangs et al., 2011). Primary antibodies used were anti-MafB (Santa Cruz; sc-10022), anti-Fos, anti-Jun and anti-macrophage (Acris Antibodies; AM08143PU-N). Confocal images were taken on a LSM780 confocal microscope (Carl Zeiss).

siRNAs specific for chicken cFos and cJun and a control siRNA were obtained from Invitrogen (Stealth RNAi). The targeted sequence was 5′-TCAACGACTTCGACCTGATGAAGTT-3′ (coding region 127-151 of chick cFos; GenBank accession number NM_205508) and 5′-AGA-AGAGCCTCAGACTGTACCTGAA-3′ (coding region 615-639 of chick cJun; GenBank accession number NM_001031289).

COS7 cells were maintained in DMEM supplemented with 10% FBS, 100 U/ml penicillin and 100 μg/ml streptomycin. Co-transfection of target plasmids (pAc-cFos-GFP or pAc-cJun-GFP) and siRNAs was carried out with Lipofectamine 2000 (Invitrogen) following the manufacturer's protocol. To assess the efficiency of the siRNAs on pAc-cFos-GFP or pAc-cJun-GFP expression in COS7 cells, the number of GFP-positive cells was counted 24 h after transfection. Student's t-test was performed to assess the differences between numbers of cells transfected with siRNA (P<0.00001). For reporter assays, COS7 cells were plated on 24-well plates and transfected using Lipofectamine 2000 and a total of 2.4 μg of DNA. Reporter constructs were co-transfected with 0.8 μg of MafB and/or cJun expression constructs in combination with Renilla vectors. Cells were collected 24 h post-transfection, and luciferase reporter assays were performed using the Dual Luciferase Kit (Promega). Each assay was performed twice.

ChIP was performed using leg buds at stage 30. Sixty limb buds were dissected, fixed in 1% formaldehyde for 10 min at room temperature, washed with PBS, and stored at −80°C. ChIP was performed from these tissue samples using antibodies against Maf, Jun or p300 (Santa Cruz; sc-584) as described previously (Visel et al., 2009). DNA samples from the whole-cell extract (WCE) and ChIP fractions were further sheared with an ultra sonicator, ligated to sequencing adapters and amplified according to the manufacturer's instructions. Gel-purified amplified DNA (100-150 bp) was sequenced on the Applied Biosystems SOLiD 3.5 platform to generate 50 bp reads. Sequence reads were aligned with the Gallus gallus reference genome (galGal3). More than 1 million reads were mapped for each sample. Aligned reads were extended to 100 bp in the 3′ direction. The number of reads was summed up in a 150 bp window with a step size of 10 bp along the chromosome for ChIP and WCE fractions, respectively. After normalization of total reads of the ChIP fraction against the WCE fraction, enrichment values (ChIP/WCE) were calculated as described previously (Lengronne et al., 2004) for each window.

For ChIP-qPCR, DNA samples from the WCE and ChIP fractions were sheared with the ultra sonicator, incubated with p300 antibody and purified. Putative binding site sequences were amplified with specific primers (for negative control, 5′-TCTGCATGCTGTGTGTCAAA-3′ and 5′-GCAGATC-CAAAAGCATCCAT-3′; for BS-p63, 5′-CAGCAGATAAGCGTGCAAG-A-3′ and 5′-GCACAGTGACTCGACAAAGG-3′; for BS-p73, 5′-TCACAT-GTGGTTCCCAGCTA-3′ and 5′-GTTGGAAGGCATGTGGAAGT-3′). The mean±s.d. was calculated, and a statistical analysis was performed using Student's t-test.

We thank Koichi Kawakami, Stephen Hughes, Toshihiko Ogura and Junichi Miyazaki for plasmids. We also thank Takayuki Suzuki, Yuji Watanabe, Yuji Yokouchi, Bau-Lin Huang, Hidenori Nishihara, Makiha Hukuda and Tomoya Itoh for technical advice; Satoru Takahashi for sharing unpublished information; and John Fallon for critical comments.