Cux2 refines the forelimb field by controlling expression of Raldh2 and Hox genes

Limb buds emerge as small bulges that protrude from the body trunk at discrete positions along the rostral-caudal axis. It has been proposed that the nested expression of Hox genes in the lateral plate mesoderm is related to the specification of position along the rostral-caudal axis to generate forelimb, interlimb and hindlimb regions (Burke, 2000; Burke et al., 1995; Cohn et al., 1995, 1997). Such a role of Hox genes in limb positioning is supported by mice lacking Hoxb5, which show a rostral shift of the forelimb buds (Rancourt et al., 1995). Recent studies provided evidence for a role for Hox genes in the specification of the forelimb field via regulation of the transcription of T-box 5 (Tbx5). Tbx5 is expressed in the anterior paired appendages of zebrafish, chick and mouse embryos (Gibson-Brown et al., 1996; Isaac et al., 1998; Tamura et al., 1999) and plays dispensable roles in the initiation of limb development (Ahn et al., 2002; Garrity et al., 2002; Naiche and Papaioannou, 2003; Ng et al., 2002; Rallis et al., 2003; Takeuchi et al., 2003). Developmental and molecular analyses of chick and mouse embryos showed that, in the forelimb field, rostrally expressed Hox genes directly activate Tbx5 transcription and thus control the position of the forelimb field (Minguillon et al., 2012). Furthermore, caudally expressed Hoxc9, which is expressed in the interlimb and hindlimb fields, represses expression of Tbx5 in the caudal lateral plate mesoderm, possibly by recruiting co-repressors (Nishimoto et al., 2014). Rostrally expanded distribution of Hoxc8 is present in the body trunk of the python snake (Cohn and Tickle, 1999), supporting the view that Hoxc8 represses expression of Tbx5 (Nishimoto et al., 2014). In fact, the position of the hindlimbs shift posteriorly in Hoxc8 null mutants (van den Akker et al., 2001). These results suggest that a combination of collinearly expressed Hox genes dictates the position of forelimbs along the rostral-caudal axis (Nishimoto et al., 2014).

Recent analyses of mouse mutants revealed that Hox9 and Hox5 genes are involved in establishing the posterior and anterior field of the forelimb, respectively (Xu et al., 2013; Xu and Wellik, 2011). The early polarity in the limb field is established by antagonistic interaction between Hand2 in the posterior mesenchyme and Gli3 in the anterior mesenchyme (Welscher et al., 2002b), prior to the initiation of Shh expression, which marks the zone of polarizing activity in the posterior margin of the limb buds (Riddle et al., 1993). An analysis of Hox9 quadruple mutants revealed that axial Hox9 paralogs are involved in the establishment of the posterior forelimb field by triggering the posteriorly restricted expression of Hand2, which regulates Shh directly to initiate its expression in the posterior margin of the limb bud (Xu and Wellik, 2011). In contrast, deletion of all three Hox5 genes suggests that Hox5 proteins interact with promyelocytic leukemia zinc finger (Plzf) and cooperatively mediate repression of Shh expression in the anterior part of the forelimb buds (Xu et al., 2013).

Past and recent studies indicated the involvement of retinoic acid in the initiation and specification of the forelimb field. Administration of disulfiram, an inhibitor of retinoic acid synthesis, to chick embryos prior to limb bud outgrowth leads to hypoplasia or a caudal shift of the forelimb bud (Stratford et al., 1996). In mouse retinaldehyde dehydrogenase 2 (Raldh2) mutants, the heart-forming field is expanded posteriorly and forelimb initiation fails (Ryckebusch et al., 2008; Sirbu et al., 2008; Zhao et al., 2009). Similarly, zebrafish mutants for raldh2 fail to initiate pectoral fin formation (Begemann et al., 2001), and zebrafish embryos treated with the retinoic acid inhibitor 4-diethylaminobenzaldehyde (DEAB) show a posterior expansion of the heart field and lack pectoral fin buds (Waxman et al., 2008). Several lines of evidence indicate that retinoic acid signaling regulates the transcription of Hox genes and leads to the regionalization of the lateral plate mesoderm along the anterior-posterior axis (Lo and Frasch, 2003; Niederreither et al., 1999; Waxman et al., 2008; Xavier-Neto et al., 2000). More recently, developmental analyses of chick and mouse embryos revealed that retinoic acid signaling and Hox proteins cooperatively activate Tbx5 transcription to induce forelimb bud formation (Nishimoto et al., 2015).

Cut/Cux transcription factors have four conserved DNA binding domains, three cut repeats and a homeodomain (Gingrasa et al., 2005; Hulea and Nepveu, 2012; Sansregret and Nepveu, 2008). In Drosophila, cut is expressed in the dorso-ventral boundary cells of the forelimb disc (Blochlinger et al., 1993; Buceta et al., 2007; Micchelli et al., 1997), and depletion of cut function disrupts the formation of the forelimb margin, suggesting that cut is required for dorso-ventral boundary formation of the developing forelimb margin (Blochlinger et al., 1993; Buceta et al., 2007; Micchelli et al., 1997). In mouse and chicken, two orthologs of cut, Cux1 and Cux2, have been identified (Tavares et al., 2000; Valarche et al., 1993). In chick embryos, Cux1 is expressed in the ectoderm adjacent to the apical ectodermal ridge and restricts its position within the limb buds (Tavares et al., 2000). While Cux2 is initially expressed in the presumptive forelimb field, its expression becomes restricted to the posterior part of the limb buds and to the interlimb flank of chick embryos (Tavares et al., 2000). The function of Cux2 in the lateral plate mesoderm, however, remains to be identified.

Here we explored the function of Cux2 during chick embryogenesis. We show that Cux2 is involved in the specification of the forelimb field. Further functional analyses revealed that Cux2 directly activates the expression of Raldh2 and Hoxb genes, which are involved in the specification of the limb-forming fields. These results suggest that Cux2 in the lateral plate mesoderm refines the forelimb-forming fields via regulation of transcription of Raldh2 and Hoxb genes in chicken embryos.

First, we investigated the expression of Cux2 during development of chick embryos (Fig. S1A–E). Transcripts of Cux2 were detected throughout the lateral plate mesoderm at Hamburger-Hamilton stage (HH) 13 (Fig. S1A), and subsequently expression was increased in the forelimb-forming fields (at the level of somites 14–18) at early HH 14 (20-somite stage; Fig. S1B). Expression of Cux2 in the anterior part of the forelimb field gradually decreased, and Cux2 expression became restricted to the posterior part of the forelimb buds and interlimb flank region by HH 17; weak expression was also detected in the posterior part of the hindlimb buds at the same stage (Fig. S1C). Cux2 was expressed in the posterior limb buds and interlimb flank region by HH 19 (Fig. S1D). By HH 26, transcripts of Cux2 were undetectable in the interlimb flank and were restricted to the posterior limb buds (Fig. S1E).

Expression analyses of Cux2 showed that it is expressed in the limb-forming region prior to the outgrowth of limb bud. To investigate whether Cux2 has a role in limb development, we downregulated endogenous Cux2 using siRNAs that targeted chick Cux2 (Cux2-siRNA). To evaluate the effect of Cux2-siRNA in vitro, we transfected COS7 cells with Cux2-siRNA or control-siRNA together with pCMV-Cux2-EGFP (Fig S1F–H). At 24 h after transfection with Cux2-siRNA, the number of EGFP-positive cells was reduced by 40% relative to the control siRNA (P<0.00001, Student's t-test; Fig. S1H).

We then investigated the effect of depleting Cux2 in limb development (Fig. S1I,J). At 22–24 h after electroporation of Cux2-siRNA into the presumptive forelimb region of HH 15–17 chick embryos, expression of Cux2 was relatively downregulated in the EGFP-positive region (2/6; Fig. S1I). In contrast, Cux2 expression was not affected in the forelimb buds of embryos electroporated with control-siRNA (0/9; Fig. S1J).

To investigate the role of Cux2 in forelimb development, we then co-electroporated Cux2-siRNA or control siRNA with pCAGGS-EGFP into the presumptive forelimb field on the right side of HH 15–17 embryos (Fig. 1A,B). The embryos were fixed 22–24 h after electroporation and examined for expression of Fgf8, a marker for the apical ectodermal ridge of limb buds (Crossley et al., 1996). When Cux2-siRNA was introduced into the presumptive forelimb field, 8 of 19 embryos showed changes in the posterior extent of the electroporated forelimb bud (Fig. 1A; Fig. S2). Embryos in which the control-siRNA was electroporated did not show any morphological changes in forelimb formation (0/9; Fig. 1B; Fig. S2). These results suggest that Cux2 is involved in the development of the forelimb buds.

We then electroporated constitutively active human Cux2 (hCux2)-VP16 or full-length hCux2 constructs into the presumptive forelimb field at HH 13–14 to further investigate the function of Cux2 in limb development. The morphology of the limb buds was evaluated based on the expression of Fgf8 22–24 h after electroporation (Fig. 1C–E). Misexpression of hCux2-VP16 led to an anterior shift of the forelimb buds, or a reduction in the forelimb field along the anterior-posterior axis (7/14; Fig. 1C; Fig. S2). Similarly, electroporation of full-length hCux2 led to an anterior shift of the posterior boundary of the forelimb buds (4/10; Fig. 1D). We also found one embryo with posteriorly extended forelimb bud, after electroporation of full-length hCux2. Control forelimb buds electroporated with pCAGGS-EGFP did not show altered limb bud morphology (0/10; Fig. 1E; Fig. S2). These results indicate that Cux2 is involved in specification of the forelimb field.

Despite the anterior shift of forelimb buds at an early stage, we could not detect any gross changes in the cartilage patterns for hCux2-VP16-misexpressing forelimbs 8 days after electroporation (n=11; Fig. S3). Thus, the effects of Cux2 misexpression are transient, and the forelimb buds normalize.

Signals involved in pre-patterning along the anterior-posterior axis of the limb and establishing the polarizing region have been proposed to also be involved in limb positioning along the rostral-caudal body axis (Rallis et al., 2003). To investigate whether Cux2 is involved in the anterior-posterior patterning of the limb, we examined the expression of Hand2 and Shh after Cux2 misexpression (Fig. 2). hCux2-VP16 was mixed with pCAGGS-EGFP, and electroporated into the right side of the forelimb field at HH 13-14, and embryos were fixed at 22–24 h after electroporation for gene expression analysis. Misexpression of hCux2-VP16 led to the anterior expansion of Hand2 throughout the forelimb field on the injected side (3/9) only in embryos with shifted limbs (three out of three shifted limbs), whereas Hand2 expression was restricted to the posterior domain on the uninjected side (3/9; Fig. 2A). No changes in Hand2 expression were observed after misexpression of the control pCAGGS-EGFP construct alone (0/5; Fig. 2B). Additionally, misexpression of hCux2-VP16 led to a rostral shift in the expression of Shh (3/6) only in embryos with shifted limbs (two out of three shifted limbs; Fig. 2C) and, of these, two samples showed reduced Shh expression (2/3; Fig. 2C), whereas misexpression of the control pCAGGS-EGFP construct did not alter Shh expression (0/6; Fig. 2D). These results suggest that Cux2 may be involved in establishment of the position of Shh expression in the polarizing region, at least in part via regulation of Hand2 expression.

Our results suggest the possible function of Cux2 in positioning of forelimb field, upstream of Hand2 expression, along the anterior-posterior axis. We then explored whether Cux2 controls factors regulating Hand2 transcription. In mouse forelimb buds, Hox9 is involved in the positioning of Shh expression via regulation of Hand2 expression (Xu and Wellik, 2011). Furthermore, Hoxb5 is involved in forelimb positioning (Rancourt et al., 1995). Thus, we examined the possibility that Cux2 regulates expression of Hoxb genes (Fig. 3A–F). First, we examined whether expression of Hoxb genes in the lateral plate mesoderm of HH 14 and 19 chick embryos (Fig. S4), to see whether their expression overlapped with that of Cux2. Hoxb3 was expressed in the anterior lateral plate mesoderm including the forelimb field at HH 14 (Fig. S4A) and in the proximal part of the limb buds as well as in the interlimb flank lateral plate mesoderm at HH 19 (Fig. S4E). Hoxb5 transcripts were detected in the anterior lateral plate mesoderm at HH 14 (22-somite stage; Fig. S4B) and were expressed in the proximal part of the limb buds and in the interlimb flank region at HH 19 (Fig. S4F). Hoxb9 was expressed in the posterior lateral plate mesoderm including the posterior part of forelimb field at HH 14 (Fig. S4C) and in the interlimb flank region at HH 19 (Fig. S4G). Thus, expression domains of Hoxb3, Hoxb5 and Hoxb9 are at least partially overlapping with those of Cux2 in the lateral plate mesoderm at HH 14 and 19 (Fig. S4).

Retinoic acid is known to control expression of Hox genes (Conlon and Rossant, 1992; Kessel and Gruss, 1991), and the specification of forelimb field (Begemann et al., 2001; Niederreither et al., 1999; Nishimoto et al., 2015; Stratford et al., 1996; Zhao et al., 2009). We, thus, examined the possibility that Cux2 controls expression of Raldh2, encoding an enzyme that catalyzes the retinoic acid synthesis from retinaldehyde (Fig. 3G,H). Raldh2 was expressed in the anterior part of the lateral plate mesoderm including the presumptive forelimb field at HH 14 (Fig. S4D) and in the interlimb flank region at HH 19 (Fig. S4H) as previously reported (Swindell et al., 1999), suggesting its expression is also overlapping with Cux2 expression in the lateral plate mesoderm at these stages.

Next, we investigated whether Hoxb genes and Raldh2 are downstream targets of Cux2 in the lateral plate mesoderm. For this purpose, we misexpressed constitutively active Cux2 in the right side presumptive forelimb field at HH 13–14 and examined the expression of Hoxb genes and Raldh2 (Fig. 3; Fig. S5). Misexpression of Cux2-VP16 in the mesenchyme of the presumptive forelimb field induced anterior expansion of Hoxb3 (3/5; Fig. 3A; Fig. S5A), whereas misexpression of pCAGGS-EGFP did not alter Hoxb3 expression (0/5; Fig. 3B). In addition, we examined expression of Hoxb5 after misexpression of constitutively active Cux2 in the presumptive forelimb field (Fig. 3). Misexpression of Cux2 induced an ectopic expression of Hoxb5 in the mesenchyme of the forelimb bud (5/10; Fig. 3C; Fig. S5B), whereas introduction of control pCAGGS-EGFP did not alter its expression pattern (0/5; Fig. 3D). Misexpression of Cux2 led to the weak upregulation of the Hoxb9 expression throughout the mesenchymal region of the forelimb bud (9/16; Fig. 3E; Fig. S5C), while control pCAGGS-EGFP introduction did not change its expression (7/7; Fig. 3F). Raldh2 expression was ectopically activated in the anterior and posterior parts of forelimb buds 22–24 h after electroporation of pCAGGS-Cux2-VP16 (10/15; Fig. 3G), whereas introduction of pCAGGS-EGFP did not cause any changes in Raldh2 expression (0/4; Fig. 3H). These results suggest that Cux2 activates expression of Raldh2 and Hoxb3 in the forelimb field and probably also controls other Hoxb genes.

To find the active enhancers, the possible target sequences of Cux2 in the lateral plate mesoderm, close to the loci of HoxB clusters and Raldh2, we dissected the lateral plate mesoderm from 40 chick embryos at HH 15 and performed ChIP-Seq using antibodies against a histone acetylation marker (H3K27ac). Immunoprecipitated DNA fragments were analyzed using massively parallel sequencing, and the resulting 50-bp sequence reads were aligned with the reference chicken genome (galGal3). This analysis allowed us to identify sequences enriched with H3K27ac within ±50 kb of the transcriptional start sites (TSSs) of Hoxb3 and Raldh2 (Fig. 4A; SRA accession number, SRP075943). We then compared these data with the putative regulatory elements enriched in H3K4me1 previously identified in HH 16 chick embryos (Seki et al., 2017). This allowed us to identify the possible enhancer sequences, 10 kb downstream of Hoxb3 (BS-Hoxb3; Fig. 4A) and within the first intron (23 kb downstream of the TSS) of Raldh2 (BS-Raldh2; Fig. 4A).

We then assessed the enhancer potential of BS-Raldh2 and BS-Hoxb3 within the forelimb bud by in ovo reporter analysis (Fig. 4B,C). These sequences were cloned upstream of a basal promoter followed by an EGFP reporter, and the resulting constructs were introduced into HH 13-14 chick presumptive forelimb fields with a RFP vector, and reporter expression was examined at HH 19. Twenty-four hours after electroporation, both BS-Raldh2-EGFP and BS-Hoxb3-EGFP drove EGFP expression in forelimb buds (arrowheads in Fig. 4B,C), suggesting that these sequences have the enhancer activity in forelimb fields.

Finally, we carried out ChIP-qPCR analysis with stage 15 chicken lateral plate mesoderm at the level of somites 21–26 using anti-Cux2 (Fig. 4D,E). ChIP-qPCR analysis showed a distinct degree of enrichment of Cux2 binding to BS-Raldh2 and BS-Hoxb3 (Fig. 4D,E). These results suggest that Cux2 directly binds to the BS-Raldh2 and BS-Hoxb3 putative enhancers, positively regulating the expression of Raldh2 and Hoxb3 in the lateral plate mesoderm.

Taken together, our results indicate that Cux2 refines the forelimb field by controlling transcription of Hoxb genes as well as the synthesis of retinoic acid in the lateral plate mesoderm in chicken embryos.

In this study, we have described a novel function for Cux2 in the refinement of the forelimb field in chicken embryos. Developmental analyses of chick embryos revealed that Cux2 refines the axial position of the forelimb field. Furthermore, Cux2 directly activates expression of Hoxb3, other Hoxb genes to a lesser degree, and the retinoic acid synthesis enzyme Raldh2 in the forelimb field. These results suggest that Cux2 refines the forelimb position along the rostral-caudal body axis via regulation of Hoxb transcription and retinoic acid synthesis (Fig. 4F).

It has been proposed that the nested expression of Hox genes in the lateral plate mesoderm is somehow related to the regionalization of the forelimb, interlimb flank and hindlimb fields (Cohn et al., 1997; Nelson, 1994). Anterior Hox genes were recently shown to directly activate transcription of Tbx5 in the forelimb field (Minguillon et al., 2012), whereas posterior Hox genes repress Tbx5 expression to restrict its expression to the forelimb field (Nishimoto et al., 2014). Furthermore, a more recent study showed that in chicken embryos the lateral plate mesoderm compartment, such as the forelimb, interlimb and hindlimb fields, is progressively formed by sequential collinear activation of Hox genes during gastrulation (Moreau et al., 2019). Thus, the regulation of Hox genes by Cux2 may further restrict Tbx5 expression to the forelimb field as a late step of positioning of the forelimb. In addition, retinoic acid signaling regulates the expression of Hox genes and leads to the regionalization of the lateral plate mesoderm along the rostral-caudal axis (Niederreither et al., 1999; Waxman et al., 2008). Moreover, modulation of retinoic acid signaling during gastrulation affects the axial extent of Hox gene expression and leads to changes in the forelimb field in chicken embryos (Moreau et al., 2019). Therefore, retinoic acid, the synthesis of which is activated by Cux2, may also be involved in regulation of Hox gene expression in the lateral plate mesoderm during a late step of forelimb positioning. Modulation of Cux2 expression causes only a slight shift in forelimb position. The data presented here are consistent with previous studies indicating that certain Hox gene mutation results in a shift in limb position of only one to a few somite lengths (Favier et al., 1996; Jurberg et al., 2013; McIntyre et al., 2007). Multiple sequences enriched for H3K27ac (SRA accession number, SRP075943) and H3K4me1 (Seki et al., 2017) were identified near the HoxB clusters, and thus we cannot exclude the possibility that Cux2 regulates multiple Hoxb genes. Consistent with this view, electroporation of pCMV-hCux2 caused not only the anterior shift of the posterior boundary of the forelimb bud, but also caused posterior extension of the forelimb bud in a single case. Importantly, Drosophila, cut is involved, directly or indirectly, in the control of expression and/or function of at least two homeotic genes, proboscipedia and Antennapedia (Johnston et al., 1998).

Therefore, we should also consider the possibility that Cux2 regulates expression of other Hox cluster genes and/or posterior Hox genes as well.

In this study, we also showed the involvement of Cux2 in the specification of the posterior forelimb field. Recent analyses of mouse mutants revealed that Hox9 and Hox5 genes are involved in the establishment of the posterior and anterior domain, respectively, of the forelimb field (Xu et al., 2013; Xu and Wellik, 2011). The Hox9 paralogous genes trigger posteriorly restricted expression of Hand2 to establish the posterior forelimb field (Xu and Wellik, 2011). In contrast, Hox5 genes interact with Plzf to cooperatively restrict Shh expression to the posterior domain of the forelimb bud (Xu et al., 2013). In this study, misexpression of Cux2 resulted in the expansion of Hand2 expression and an anterior shift in Shh expression and, in a few cases, reduction of the Shh expression domain. This reduction in the region that expresses Shh could be caused by ectopically expressed Hoxb5 genes after Cux2 misexpression. In addition, we observed two types of transient morphological changes after the misexpression of Cux2-VP16: an anterior shift of the forelimb bud and a reduction in the width of the forelimb bud. Because misexpression of Cux2-VP16 reduced the level of Shh expression in a few cases, the width of the shifted forelimb bud may be regulated by the range of signaling from the polarizing region. It is also possible that mosaic and transient expression of Cux2-VP16 may not be able to continue altering the expression of target genes during the late stages of limb development; instead, the limb bud normalizes. In the early limb field, genetic antagonism between Hand2 and Gli3 establishes an anterior-posterior pre-pattern (te Welscher et al., 2002a), and such genetic interplay is also involved in refining the limb position (Rallis et al., 2005). Thus, modification of Hox gene expression caused by misexpression of Cux2 may have caused the anterior expansion of Hand2 expression and thereby led to the anterior shift in Shh expression and of the forelimb bud. In addition, misexpression of Cux2 showed only mild effects on the expression of Hand2 and Shh. These results are consistent with our view that modification of Hox gene expression, which causes only a slight shift of limb position in mutants (Favier et al., 1996; Jurberg et al., 2013; McIntyre et al., 2007), leads to a change in Hand2 expression. In addition, based on the posteriorly restricted expression pattern of Cux2 in the limb buds, a potential role for Cux2 in the establishment of Shh expression in the posterior domain of limb buds has been proposed (Tavares et al., 2000).

Retinoic acid signaling is correlated with forelimb initiation and positioning in zebrafish, chick and mouse embryos (Mic et al., 2002; Stratford et al., 1996; Waxman et al., 2008; Zhao et al., 2009). Treatment of chick embryos with the retinoic acid inhibitor disulfiram leads to a disruption of limb formation or a shift in limb position (Stratford et al., 1996). In zebrafish, raldh2 mutants lack pectoral fin buds (Gibert et al., 2006; Grandel et al., 2002), and embryos treated with retinoic acid inhibitor show a downregulation of hoxb5b expression and a failure to induce the formation of pectoral fins (Waxman et al., 2008). Retinoic acid signaling was recently shown to directly activate transcription of Tbx5 in the forelimb-forming fields (Nishimoto et al., 2015). Thus, we should also consider the possibility that Cux2 contributes to the establishment of the Tbx5 expression domain in the forelimb field via activation of retinoic acid synthesis. In this study, we showed that Cux2 directly binds to the enhancer of Raldh2. However, in certain cases, Raldh2 expression was altered in non-Cux2-transfected regions of the limb buds, suggesting that Raldh2 expression can also be changed by unknown genes modified after misexpression of Cux2. In mouse Raldh2 mutants, heart-forming fields expand posteriorly and forelimb initiation fails (Mic et al., 2002; Niederreither et al., 1999; Zhao et al., 2009). Thus, retinoic acid signaling seems to have a role in the regionalization of the lateral plate mesoderm into the cardiac and the posterior lateral plate mesoderm. In this study, we introduced Cux2 constructs into the coelom at the level of the forelimb field, which appears after HH 14. Thus, it is unlikely that the shift in the forelimb position was caused by an alteration of the regionalization into the cardiac and the posterior lateral plate mesoderm, although we cannot exclude the possibility that Cux2 is involved in this process at much earlier stages.

In this study, we found that Cux2 expressed in the lateral plate mesoderm refines the position of the forelimb field in chicken embryos. Notably, however, the expression of Cux2 differs between chicken and mouse embryos, at least in late limb buds (Iulianella et al., 2003). In mice embryos, in contrast to chicken embryos, Cux2 is expressed in the mesenchyme underlying the apical ectodermal ridge in late limb buds and subsequently in the interdigital region (Iulianella et al., 2003). Although no publications have described the expression or function of Cux2 in the lateral plate mesoderm prior to the initiation of limb buds, it is possible that the role of Cux2 is not the same in mouse embryos as in chicken embryos with respect to limb positioning.

Importantly, expression of Cux2 in the hindlimb-forming field is not as strong as in the forelimb field at the pre-limb bud stage. As we mentioned above, the factors involved in specification of the hindlimb field are different from those that specify the forelimb field (Kawakami et al., 2001; Logan and Tabin, 1999; McPherron et al., 1999; Szeto et al., 1999; Xu et al., 2013; Xu and Wellik, 2011). It is, however, possible that factor(s) with similar roles to Cux2 may be distributed in the hindlimb-forming field.

In conclusion, we revealed a novel function for Cux2 on refining the forelimb-forming field via regulation of transcription of Hoxb gene(s) and retinoic acid synthesis in chicken embryos. As discussed here, retinoic acid and Hox play multiple roles during specification of limb field. Thus Cux2 is likely to be involved in the forelimb specification via multiple pathways, including the specification of the polarizing region and the activation of forelimb initiation gene(s), which future studies should aim to address.

No statistical methods were used to predetermine sample size. The experiments were not randomized and the investigators were not blinded to allocation during experiments and outcome assessment. The age of each specimen is noted in the figure legends. The sex of the embryos is unknown.

All experiments were performed in accordance with guidelines for animal experiments of Tokyo Institute of Technology.

Fertilized White Leghorn chicken (Gallus gallus) eggs were incubated at 38°C in a humidified incubator and staged according to Hamburger-Hamilton's staging (Hamburger and Hamilton, 1951). Gallus gallus Cux2 [730 base pairs (bp)], Hoxb3 (734 bp), Hoxb5 (717 bp) and Raldh2 (1500 bp) were amplified from cDNA pools prepared from HH 20 chick embryos using the following primers, which hybridized to the indicated published sequences: Cux2 (GenBank accession number, XM415167.4), 5′-CCAGGGCAGTGTGAGTGACATGC-3′ and 5′-CCCTTGGCTTCTTGATCTGCAGG-3′; Hoxb3 (GenBank accession number, NM_204743), 5′-GTCAAAAGGGATGGGCTCTT-3′ and 5′-CTTGGAAACTGTGCCAAACAG-3′; Hoxb5 (GenBank accession number, NM_001025355), 5′-AGGACAGCGTACACTCGCTAC-3′ and 5′-ACTGCGACTGTAGTGCAGGAAC-3′; Raldh2 (GenBank accession number, NM_204995), 5′-ATGGCATCTCTGCATCTGCTG-3′ and 5′-TTAGGAATTCTTCTGAGGGATC-3′. Wholemount in situ hybridization was carried out essentially as described (Izpisua-Belmonte et al., 1993). Probe templates for Fgf8, Shh and Hoxb9 were described previously (Burke et al., 1995; Crossley et al., 1996; Riddle et al., 1993). Some wholemount in situ samples were embedded in 2% agarose in phosphate-buffered saline (PBS), and sections were cut at 100 μm thickness with a MicroSlicer ZERO1N (Dosaka EM).

siRNA specific for chicken Cux2 were obtained from Sigma-Aldrich (siRNA Duplex), and a control siRNA were obtained from Invitrogen (Stealth^TM RNAi). The targeted sequence was 5′-CCUACCUGAAGCGUCGGUAUGGGCU-3′ of chick Cux2, which corresponds to nucleotides 3215–3239, the homeodomain-encoding region (GenBank accession number, XM415167.4). For negative control, siRNA Negative Control Hi GC Duplex #2 (Invitrogen) was used.

For pCMV-Cux2-EGFP, the cDNA sequence encoding amino acid residues 870–1113 of chick Cux2 (XP_415167.4) was amplified with the polymerase chain reaction (PCR) using the primers 5′-CCGCTCGAGCCCACCATGGGCAGTGTGAGTGACATGCTG-3′ and 5′-CGGGATCCGGCTTCTTGATCTGCAGGAGG-3′, inserted into the pGEM-T easy vector (Promega) and then cloned into the XhoI and BamI sites of pAcGFP-N1 (Clontech). For pCMV-hCux2, the cDNA for the entire human Cux2 coding region was obtained from pF1KSDA0293 (Kazusa DNA Research Institute, human clone KIAA0293) and cloned into the SgfI and PmeI sites of the pF4A-pCMV Flexi vector (Promega). For pCAGGS-hCux2-VP16, RCASBP(A)-Tbx5-VP16 (CT#630), a gift from Cliff Tabin (Addgene, Plasmid #13969) (Rallis et al., 2003), was digested with ClaI, cloned into the ClaI site of the SLAX12NCO vector (Morgan and Fekete, 1996) and then digested with EcoRI and HindIII; the resulting 2×VP16 fragment was cloned into the pBKKS vector (pBKKS-2×VP16). The cDNA sequence encoding amino acid residues 889–1228 of human Cux2 was cloned into the SpeI and EcoRI sites of pBKKS- duplex VP16 (pBKKS-ΔhCux2-2×VP16). The cDNA fragment of ΔhCux2-2×VP16 was amplified using primers including the XhoI sequence (5′-TTTGGCAAAGAATTCCTCGAGCCCACCATGGAGCTGTACAT-3′ and 5′-CTGAGGAGTGAATTCCTCGAGGTCGACGGTATCGATAAGC-3′) and inserted into the XhoI site of pCAGGS vector (Niwa et al., 1991) via the In-Fusion reaction (Clontech). For enhancer analysis, Binding Sequence (BS)-Raldh2 and BS-Hoxb3 were isolated from the chick genome by PCR. The following forelimb forward and reverse primers were used: BS-Raldh2, 5′-GCACATACATGACACACCGT-3′ and 5′-CATAGATATTCCTACCACTAAG-3′; BS-Hoxb3, 5′-CATAGATCTGCAACACTTCA-3′ and 5′-CCGCTTCGATTCCTTTCC-3′. These sequences were subcloned in front of a chicken β-actin basal promoter (Matsuo et al., 1991), that is followed by a GFP reporter (Ogino et al., 2008).

COS7 cells provided by M. Komada (Tokyo Institute of Technology, Yokohama, Japan) (Reincke et al., 2015), were maintained in DMEM (Nacalai tesque) supplemented with 10% fetal bovine serum (Gibco Invitrogen), 100 U/ml penicillin and 100 μg/ml streptomycin (Wako). Co-transfection of target plasmid pCMV-Cux2-EGFP (encoding enhanced green fluorescent protein, EGFP) and siRNAs was carried out with Polyethylenimine ‘MAX’ (Polysciences Inc.) as per the manufacturer's protocol. To assess the efficiency of the siRNAs on pCMV-Cux2-EGFP expression in COS7 cells, the number of EGFP-positive cells was counted at 24 h after transfection. Student's t-test was performed to assess differences in the number of EGFP-positive cells between cells transfected with Cux2 and control siRNAs (P<0.00001).

Plasmid or siRNA solution was colored with 1–5% Fast Green and co-electroporated with pCAGGS-EGFP (a gift from Dr Miyazaki and Dr Ogura) (Niwa et al., 1991) or pCAGGS-RFP (Das et al., 2006) (a gift from Dr Tickle) into the presumptive limb mesenchyme as described (Onimaru et al., 2015; Suzuki and Ogura, 2008). Briefly, the solution was injected into the lateral plate mesoderm of HH 13–14 chick embryos, and electroporated with two pulses of 8–12V, 85 ms by CUY21 EDIT (BEX Co., Ltd), or five pulses of 6V, 20 ms after a poration pulse of 25V, 0.05 ms by CUY21 EDIT II (BEX Co., Ltd.). The concentrations of the prepared plasmid solutions or siRNAs were as follows: pCAGGS-EGFP, 1–5 mg/ml; pCMV-hCux2, 5 mg/ml; pCAGGS-hCux2-VP16, 5 mg/ml; control siRNA, 5 μM and Cux2 siRNA, 250 μM. For in ovo enhancer analysis, 5–10 mg/ml of BS-Hoxb3 or BS-Raldh2 was electroporated into the lateral plate mesoderm of HH 13–14 chick embryos.

Measurements were made using ImageJ (https://imagej.net/Downloads). The length of both the AER and the region between seven somites including the limb field were measured in embryos stained with fgf8 RNA probes. Half of the distance between the seven somites was defined as ‘zero’ and used as a reference to measure the relative position of the limbs.

The morphology of the limbs that were subjected to experimental manipulation was studied after cartilage staining with Alcian Blue (Tanaka et al., 2000).

ChIP was performed as described (Nakato et al., 2013; Suda et al., 2014; Visel et al., 2009). Briefly, the lateral plate mesoderm (at the level of somites 21–26) from 40 HH-15 chick embryos was 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 H3K27ac (ab4729, Abcam) or Cux2 antibodies (Eurofins Genomics K. K., Tokyo, Japan). To generate polyclonal antibodies against chicken Cux2, two synthetic peptides, N-SAGSDSESPGARSEC-C and N-C+RLSTSVQRRHEKMA-C, derived from estimated amino acids 1080–1094 and 1346–1359, respectively (GenBank accession number, XP_415167.4), were synthesized, conjugated to keyhole limpet hemocyanin and used to immunize rabbits; the resulting antisera were purified by affinity chromatography (Eurofins Genomics K.K., Tokyo, Japan).

DNA samples from the whole-cell extract (WCE) and ChIP fractions were further sheared with an ultra sonicator (Branson Sonifire 250D), ligated to sequencing adapters and amplified according to the manufacturer's instructions (Applied Biosystems SOLID 5500). Gel-purified amplified DNA (100–150 bp) was sequenced on an Applied Biosystems SOLID 5500 platform to generate 50-bp reads. Sequence reads were aligned with the G. gallus reference genome (galGal3) with Bowtie version 1.1.2 (Langmead et al., 2009), allowing three mismatches per read and outputting only uniquely mapped reads (-n3 -m1 option). The mapping statistics are summarized in Table S1. More than 10 million reads were mapped for each sample. For peak calling and data visualization, we used DROMPA (Nakato et al., 2013) with a parameter set that identified the regions that satisfied the following criteria: >3.0-fold enrichment (ChIP/WCE), P<1×10^–4 (one-sided Wilcoxon test) and a normalized peak intensity of >3.0. ChIP-Seq from this study are available from the Sequence Read Archive (SRA) database (http://www.ncbi.nlm.nih.gov/sra) under the accession number SRP075943.

For ChIP-qPCR, DNA samples from the WCE and ChIP fractions were sheared with the ultra sonicator, incubated with Cux2 antibodies (Eurofins Genomics K. K., Tokyo, Japan) and purified. Putative binding site sequences were amplified with the following primers (BS-Raldh2, 5′-GTAGCATGATTTACATGGAAGC-3′ and 5′-GTGACCGTGGTAAAGGCTAAC-3′; BS-Hoxb3, 5′-TCTCAGGAATCAGAATGAGCC-3′ and 5′-GCGCTTCCCTCGGTTTTATA-3′). The mean±standard deviation (s.d.) was calculated, and a statistical analysis was performed using Student's t-test.

We thank Yasuko Kanazawa for technical assistance and critical comments. We also thank Cheryll Tickle for critical comments, plasmids and use of her facilities, Takayuki Suzuki, Yuji Watanabe and Yuji Yokouchi for technical advice, Natsuno Suda, Ryo Yamada, Hisanobu Shimamura for technical assistance, Stephen Hughes, Toshihiko Ogura, Junichi Miyazaki, Malcolm Logan, Cliff Tabin, Ann Burke, Hajime Ogino and Masayuki Komada for plasmids and a cell line and the Biotechnology Center of Tokyo Institute of Technology for sequencing services.