Conserved cis-regulatory regions in a large genomic landscape control SHH and BMP-regulated Gremlin1expression in mouse limb buds

Identification of conserved limb bud regulatory regions within the Grem1genomic landscape

Using functional genetics in the mouse, we previously identified a ~70 kb region located downstream of Grem1 that is required for its expression in the limb bud mesenchyme (Figure 1A, [7]). This limb bud cis-regulatory region is located between coding exons 19 and 23 of the neighbouring Fmn1 gene. The Evolutionary Conserved Regions (ECR) genome browser (http://ecrbrowser.dcode.org/) was used for multiple sequence alignments to compare the mouse genome with different mammalian and the chicken genomes. This analysis revealed several blocks of highly conserved sequences among mammalian species, but only three of them were also highly conserved in the chicken and termed HMCO1, 2 and 3 (Human-Mouse-Chicken-Opossum conserved sequences 1 to 3, Figure 1A). The most conserved parts of these three HMCO core regions are ~80% identical (for details see Additional file 1 and Additional file 2). Of these three regions only HMCO1 is also present in the orthologous genomic region in frogs (Figure 1A), which express Grem1 during limb bud development [35]. Despite the fact that Grem1 is expressed during fin bud development [36], no HMCO homologies are present in the zebrafish genome (Figure 1A). ClustalW2 alignments (http://www.ebi.ac.uk/Tools/msa/clustalw2/) of the Fmn1 Grem1 regions from different species revealed the absence of HMCO homologies in other ray-finned fish species such as fugu and medaka (Figure 1B). In contrast, all three HMCO regions are present in the genome of coelacanth, a lobe-finned fish closely related to tetrapods (Figure 1B and Additional file 1) [37]. As the Grem1 and Fmn1 loci are linked in all species analysed (Figure 1B), the distance between the orthologous Fmn1 coding exon 22 (located adjacent to HMCO1) and Grem1 coding exon 2 was determined. During evolution, this distance increased as it is smallest in the genomes of ray finned fishes and at least ~10-fold larger in tetrapods and coelacanth (Figure 1B). In addition, the distance between Fmn1 exon 19 (close to HMCO3) and exon 22 correlates well with the presence or absence of the three HMCO regions (Figure 1B and data not shown). In ray-finned fishes and frogs (lacking all or specifically HMCO2, 3, respectively), the distance is ~3-7 kb, while in mammals and coelacanth it ranges from ~50-70 kb. Thus, the increase of intronic and intergenic regions in the Fmn1-Grem1 landscape correlates well with appearance of the three HMCO regions during vertebrate evolution.

To determine the requirement of each of these three HMCO regions for Grem1 expression in mouse limb buds, we used a BAC-based strategy in combination with analysis of transgenic founder embryos (Figure 2A-D and Additional file 3). A 250 kb mouse genomic BAC encoding the critical region and the Grem1 transcription unit was used to fuse the LacZ gene in frame with the Grem1 ORF [7]. We assessed the expression of the LacZ reporter transgene by analysing the spatio-temporal distribution of ß-galactosidase activity (Figure 2A-D). Expression of the Grem1-LacZ fusion protein in the forelimb bud mesenchyme was detected in a posterior-distal domain (Figure 2A), mimicking the early endogenous Grem1 expression rather accurately. Therefore, the Grem1-LacZ BAC construct was used to engineer deletions of the three HMCO regions and determine their requirement for Grem1-LacZ expression in the posterior mesenchyme. To control for reproducible generation of expressing BAC transgenic founder embryos, the Grem1-LacZ BAC (Figure 2A) was injected in parallel to the BACs with engineered deletions (Figure 2B-D). While the control Grem1-LacZ BAC was always robustly expressed in the posterior mesenchyme (Figure 2A), deletion of the 520 bp HMCO1 core region resulted in complete loss of LacZ expression from limb buds (Figure 2B, n = 5/7 embryos with LacZ expression, for details see Additional file 3). In contrast, deletion of the 1279 bp HMCO2 core region only caused partial loss of LacZ expression from the posterior limb bud mesenchyme (Figure 2C; Additional file 3). LacZ remained expressed normally in the majority of founder embryos carrying a 924 bp deletion of the HMCO3 region (Figure 2 = 3/5 embryos with LacZ expression, for details see Additional file 3). Taken together, this BAC transgenic analysis establishes HMCO1 as most critical for Grem1 expression in the posterior-distal limb bud mesenchyme. As the HMCO1 core region is highly conserved in tetrapods and lobe-finned fish but not ray-finned fish (Figure 1B), it likely represents a cis-regulatory region important to tetrapod evolution. The other two HMCO regions might contribute to robust expression of Grem1 LacZ in the posterior mesenchyme, as in particular the deletion of HMCO2 results in significantly reduced LacZ expression (Figure 2C).

In an attempt to gain further insight into the enhancer potential and possible interactions of HMCO2/3 with HMCO1, conventional transgenic approaches using a minimal human ß-globin promoter (ßglob-LacZ) were employed. However, neither individual HMCO regions (data not shown) nor in a combination of all three was able to drive robust expression of the ßglob-LacZ transgene in the posterior limb bud mesenchyme (Figure 3A, compare to Figure 2A). Expression of the ßglob-LacZ transgene under control of all three HMCO core regions resulted in scattered LacZ positive cells in the anterior-distal mesenchyme of forelimb buds (left panel, Figure 3A, n = 3/3). In contrast, the transgene was strongly expressed in the posterior embryo including hindlimb buds (right panel, Figure 3A). These results show that a transgene consisting of an array of the HMCO core regions is unable to drive LacZ expression into the posterior forelimb bud mesenchyme. This indicates that additional elements of the Grem1 genomic landscape are required for Grem1 expression in limb buds.

Next, we assessed to which extent a larger genomic fragment containing HMCO1 and flanking regions could drive ßglob-LacZ expression into the posterior limb bud mesenchyme (Figure 3B). Analysis of transgenic founder embryos revealed that this transgenic construct drives robust ßglob-LacZ expression in the posterior limb bud mesenchyme (Figure 3B, see also Figure 4), strikingly similar to the domain of the parental Grem1-LacZ BAC construct (Figure 2A). Taken together, these results indicate that this cis-regulatory region enhances Grem1 expression in the posterior limb bud mesenchyme, which prompted us to term this 9 kb region Grem1 Regulatory Sequence 1 (GRS1, Figure 4A). GRS1 encompasses the HMCO1 region next to Fmn1 coding exon 22 and a further downstream region that is highly conserved in mammals. This region, like HMCO1 overlaps with a previously identified GLI binding region (GBR, [32]). In addition, two potential GLI binding sites were identified by in silico analysis (Figure 4A).

Spatio-temporal activity of the GRS1region during mouse limb bud development

To analyze comparatively the activity of the GRS1 region with respect to the spatio-temporal regulation of Grem1 expression, transgenic mouse strains expressing the GRS1-ßglob-LacZ reporter construct were established (Figure 4). Seven transgenic founders were obtained. In three independent strains, LacZ was expressed in the posterior fore- and hindlimb bud mesenchyme in a pattern comparable to the transgenic founder embryos (Figure 3B and data not shown). Two of these three transgenic strains were analyzed in detail and used to study the spatio-temporal LacZ distribution (Figure 4B, C and Additional file 4). Initial analysis revealed that ß-galactosidase activity of the LacZ transgene was detected earlier in one strain and is significantly higher than in the other strain (Figure 4B, compare to Figure 4C). As levels did not change during subsequent generations (data not shown), this is likely due to differences in transgene copy number and/or integration site. Therefore, the copy number of both strains was determined by real-time qPCR analysis (Additional file 4). This analysis revealed that ten copies of the GRS1-ßglob-LacZ transgene were integrated into the genome of the strain expressing higher LacZ levels (GRS1 ^tg10 ), while only two copies were detected in the GRS1 ^tg2 strain (Additional file 4). In GRS1 ^tg10/+ embryos, ß-galactosidase activity is first detected at ~ E10.25 in the posterior forelimb bud mesenchyme and continuously increases until ~ E11.75 (Figure 4B). From ~ E11.0 onwards, scattered positive cells were detected in the anterior mesenchyme, and this anterior expression increased during distal limb bud outgrowth to form a crescent in the distal-anterior autopod (right panels, Figure 4B). By E11.75, ß-galactosidase activity was rather variable, such that the anterior crescent separated from the posterior domain in some forelimb buds. In GRS1 ^tg2/+ embryos, the spatio-temporal pattern is similar, but expression levels are significantly reduced due to the lower transgene copy number, resulting in detection of ß-galactosidase activity from only ~ E10.5 onwards (Figure 4C, arrowhead). In both strains, GRS1 drove expression of the ßglob-LacZ reporter specifically in the limb bud mesenchyme, despite some low ß-galactosidase activity detected in the developing eyes of GRS1 ^tg10/+ transgenic embryos (Additional file 4). This analysis reveals the robust nature of the GRS1 region, which functions in positive regulation of Grem1 expression in the limb bud mesenchyme. The posterior domain of ß-galactosidase activity at E10.5 (Figure 4B) is comparable to the Grem1 transcript distribution in wild-type limb buds at earlier stages (E9.5-E10.25, Figure 4D). This temporal delay is likely due to postponed transcriptional activation and/or up-regulation of the GRS1 transgene as the establishment of a posterior LacZ expression domain is also only apparent at E10.5 in GRS1 ^tg10/+ transgenic limb buds (Additional file 5). Therefore, additional cis-regulatory regions likely control the temporally correct early onset of Grem1 expression. Similar delays in the onset of LacZ reporter gene expression have been previously observed by analyzing the ZRS cis-regulatory region that controls Shh expression in mouse limb buds [38]. During distal progression of limb bud development, endogenous Grem1 expression expands distal-anterior within the developing handplate and begins to fade by E11.75 (Figure 4D), due to termination of the SHH/GREM1/AER-FGF feedback loop and GLI3-mediated repression in the anterior limb bud mesenchyme [11, 26, 39, 40]. In contrast, ß-galactosidase activity remains high in the posterior limb bud mesenchyme of GRS1 transgenic embryos and the distal-anterior expansion of its expression is significantly delayed and only occurs as transcription of the endogenous Grem1 locus is starting to terminate (E11.5 onwards Figure 4B, C, compare to Figure 4D).

Grem1 expression is restricted to dorsal and ventral limb bud mesenchyme and excluded from the chondrogenic core mesenchyme (Figure 5A), which is relevant to BMP-mediated induction of chondrogenesis (ref. [26, 41] and Benazet et al., submitted). Furthermore, the LIM-homeodomain transcription factors Lhx2 and Lhx9 have been implicated in regulating Grem1 expression predominantly in the ventral limb bud mesenchyme in response to SHH signalling [42]. Therefore, the extent to which the dorso-ventral transcript distribution is maintained by the GRS1 transgene was assessed (Figure 5B). Indeed, the expression of both the high (Figure 5B) and low copy transgenes (Additional file 6) remained excluded from the core mesenchyme throughout limb bud outgrowth and patterning. Similar to Grem1 transcripts, ß-galactosidase activity is higher dorsally than ventrally (Figure 5B, compare to Figure 5A). By E11.75, ß-galactosidase activity expands anteriorly in both the dorsal and ventral mesenchyme (E11.75, Figure 5B), reminiscent of the crescent of Grem1 transcript (Figure 5A and see before), which indicates that the GRS1 contains cis-regulatory regions controlled by Lhx transcription factors [42]. Taken together, the results shown in Figures 4 and 5 establish that the 9 kb GRS1 cis-regulatory region is able to recapitulate major aspects of Grem1 expression in the limb bud mesenchyme. In particular, the GRS1 also recapitulates aspects of the distal-anterior expansion of Grem1 expression in limb buds (Figure 45), which was not observed in previous attempts to identify cis-regulatory regions that control Grem1 expression in limb buds [7, 32]. The fact that the GRS1 transgene recapitulates some aspects of this distal-anterior expansion is important, as Grem1 transcription normally expands anteriorly in register with posterior AER-FGFs and allows propagation of SHH/GREM1/AER-FGF signalling [14, 30]. However, the delay in activation and lack of termination of the GRS1 transgene (Figure 4) indicates that other regulatory inputs from the Grem1 genomic landscape are required to regulate its dynamic expression in mouse limb buds.

The GRS1 cis-regulatory region integrates inputs from the SHH and BMP signalling pathways

During limb bud initiation, mesenchymal BMP4 signal transduction is likely required to activate Grem1 expression in the posterior mesenchyme, while SHH is primarily required for up-regulation and distal-anterior expansion during progression of limb bud development [11, 14, 43]. Furthermore, GLI3 in the anterior mesenchyme and AER-FGF signal transduction are required to restrict and eventually terminate Grem1 expression (starting ~ E11.5-11.75) [26, 40]. Therefore, GRS1 ^tg10/+ embryos lacking key components of both the SHH and BMP signalling pathway in their limb buds were generated to gain insight into the possible direct impact of these main signalling pathways on the GRS1 element (Figure 6). Analysis of Shh ^Δ/Δ GRS1 ^tg10/+ embryos revealed the complete absence of ß-galactosidase activity in Shh-deficient forelimb buds (Figure 6A). This contrasts with the endogenous Grem1 expression, which is activated but not maintained in Shh-deficient limb buds [11, 14]. These results show that activation of the GRS1 region depends on SHH signalling, which indicated that it could also participate in ectopic Grem1 activation due to anterior ectopic SHH signalling in mouse [14] and chicken limb buds [12]. Furthermore, GLI3 proteins interact with specific cis-regulatory regions in the Grem1 genomic landscape [32] and genetic analysis has shown that Gli3 is essential for the spatio-temporally controlled restriction and subsequent termination of Grem1 expression in the distal anterior mesenchyme [25, 26, 44]. In Gli3-deficient GRS1 transgenic (Gli3 ^Xt/Xt GRS1 ^tg10/+ ) forelimb buds, ß-galactosidase activity in the posterior limb buds is comparable to GRS1 ^tg10/+ limb buds, while expression in the anterior mesenchyme is significantly increased by E11.75 (Figure 6B, compare to right-most panels, Figure 4B). This late up-regulation of anterior ß-galactosidase activity indicates that GRS1 is required for GLI3-mediated termination of Grem1 expression in the anterior limb bud mesenchyme, which is essential for the spatio-temporally correct initiation of mesenchymal condensations and chondrogenic differentiation of anterior digits [26].

BMPs control Grem1 activation in the posterior limb bud mesenchyme and directly modulate its expression as part of the self-regulatory SHH/GREM1/AER-FGF feedback signalling system. Therefore, a conditional Smad4 loss-of-function allele [45] in combination with the Prx1-Cre recombinase strain [46] was used to inactivate Smad4 (Smad4 ^Δ/ΔM ) and thereby canonical BMP signal transduction in the limb bud mesenchyme. In Smad4 ^Δ/ΔM mutant limb buds, endogenous Grem1 expression is down-regulated but not lost, while Shh expression remains (Benazet et al., submitted). In Smad4 ^Δ/ΔM GRS1 ^tg10/+ forelimb buds, ß-galactosidase activity appears normal in the posterior mesenchyme, while distal-anterior expansion fails to occur (Figure 6C, compare to Figure 4B). This indicates that GRS1 activity is modulated by SMAD4-mediated signal transduction during the progression of limb bud development. These results indicate that also the distal-anterior expansion of Grem1 expression depends on BMP activity and agree with the observation that genetic lowering of Bmp4 results in a global reduction of Grem1 expression in the limb bud mesenchyme [11].

Finally, in Grem1-deficient embryos, BMP activity is increased due to reduced BMP antagonism, which results in up-regulation of non-functional Grem1 transcript [13]. As BMP signal transduction modulates GRS1 activity during advanced limb bud development (Figure 6C), we also determined the potential influence of Grem1 deficiency on GRS1 activity. In Grem1 ^Δ/Δ GRS1 ^tg10/+ E10.5 limb buds ß-galactosidase activity is initially similar to GRS1 ^tg10/+ limb buds (Figure 6D, compare to Figure 4B). Subsequently, ß-galactosidase activity is increased and by E11.75 the entire Grem1 ^Δ/Δ GRS1 ^tg10/+ limb bud is positive (Figure 6D). These alterations of ß-galactosidase activity in Smad4 ^Δ/ΔM and Grem1 ^Δ/Δ forelimb buds reveal that GRS1 activity is extensively modulated by changes in BMP activity, in particular also the anterior expansion of its expression. Taken together, this genetic analysis (Figure 6) shows that GRS1 activity critically depends on SHH and is modulated extensively by BMP signal transduction. This analysis identifies the GRS1, located ~70 kb downstream of the Grem1 transcription unit as a cis-regulatory region that integrates inputs by both SHH and BMP signal transduction. However, as previous analysis provided good evidence that Grem1 is activated by mesenchymal BMP signalling upstream of establishing the SHH/FGF feedback loop [11], these studies point to the existence of additional unknown BMP response regions that activate Grem1 expression in the posterior limb bud mesenchyme.

Endogenous GLI3 proteins are part of the chromatin complexes interacting with specific parts of the GRS1region in mouse limb buds

The observed loss of GRS1 activity in Shh-deficient limb buds (Figure 6A) indicated that the GRS1 region could be regulated directly by GLI proteins. Indeed, two GLI binding regions without consensus GLI binding motif (GBR1 and GBR2, [32]) were previously mapped to the 9 kb GRS1 region using transgene-mediated expression of an epitope-tagged GLI3 transgene in combination with chromatin immunoprecipitation (ChIP). To study the potential interaction of endogenous GLI3 proteins with the GRS1 region in its normal genomic context in wild-type limb buds, ChIP analyses using antibodies that detect both the GLI3 activator and repressor protein isoforms were performed [26, 47]. GLI3 ChIP using extracts of wild-type and Gli3-deficient limb buds at E11.5 revealed significant enrichment of one specific region within GRS1 by real-time qPCR analysis (Figure 7A). A significant ~4-fold enrichment of an amplicon located within HMCO1 was observed by comparing wild-type to Gli3-deficient limb buds (amplicon “d”; p = 0.008) and to a less conserved region (amplicon “e”; located outside HMCO1 close to a putative GLI binding site; Figure 7A). In contrast, no significant enrichments of the other amplicons located in GRS1 were detected (Figure 7A). As the amplicons located in HMCO2 (Figure 7B) and HMCO3 (Figure 7C) were also not significantly enriched, the GLI3-containing chromatin complexes appear to interact rather specifically with a region within the highly conserved HMCO1 in wild-type limb buds. These results not only corroborates the identity of previously identified GBR1 [32], but indicate that this region mediates the effects of SHH signal transduction on GRS1 activity in the posterior limb bud mesenchyme.

Spatio-temporal Grem1 expression is regulated by the interaction of multiple cis-regulatory regions far downstream of the transcription unit

Previous genetic analysis showed that Grem1 transcription in limb buds is initiated by BMP signalling and up-regulated under the influence of SHH and AER-FGF signaling as limb bud development progresses [11]. Finally, Grem1 expression is terminated concomitantly with the initiation of digit formation by high levels of FGF signal transduction and GLI3-mediated repression in the anterior mouse limb bud mesenchyme [26, 40]. This dynamic spatio-temporal regulation of Grem1 expression indicates that the activity of different cis-regulatory elements may change over time. The spatio-temporal activity of the GRS1 transgene (Figure 4B) indicates that it is primarily regulated by the SHH/GREM1/AER-FGF feedback signaling system as limb bud development progresses. In particular, its activity is first apparent when the SHH/GREM1/AER-FGF feedback is already established, and the expected termination does not occur, as ß-galactosidase activity remains in the posterior mesenchyme after the endogenous Grem1 transcripts have been down-regulated (Figure 4B, D). These results indicate that FGF-mediated termination of Grem1 expression and the underlying self-regulatory feedback signalling system [40] does not occur by FGF signal transduction impacting the GRS1 region. In addition, neither the 70 kb critical region of the Grem1 landscape (Grem1-LacZ, Figure 2A) nor transgenes derived from this region (Figures. 234 and ref. [32]) are able to accurately recapitulate the entire spatio-temporal distribution of Grem1 transcript in the limb bud mesenchyme. Therefore, either the interactions among these elements or additional as yet unknown cis-regulatory regions located outside the critical region must also participate in the regulation of Grem1 expression. Indeed, there is circumstantial evidence for the latter, as deletion of the GC-rich Fmn1 coding exon 9 (located ~200 kb downstream of the Grem1 transcription unit) results in non-complementation with a Grem1 null allele, causing a hypomorphic Grem1 limb skeletal phenotype, and significantly reduced Grem1 expression (ref. [29] and A.Z and R.Z., unpublished data).

Mapping of DNAse I hypersensitive (HS) sites is used to identify active regulatory regions in chromatin (reviewed in ref. [48]). Recently, the ENCODE consortium has done genome-wide mapping of HS sites in several mouse tissues including limb buds (ENCODE Group, University of Washington) [33, 34]. Interestingly, this analysis revealed three HS sites within the GBRs of the GRS1 and HMCO2, while no HS sites mapping to HMCO3 were detected in limb buds at E11.5 (Figure 8). The five HS sites mapped by the ENCODE consortium in the Grem1 genomic landscape overlap with the cis-regulatory elements required for Grem1 expression in mouse limb buds (Figure 8). The HS site mapping to HMCO1 overlaps well with the amplicon enriched by GLI3-ChIP analysis (Figure 7A). In fact, all HS sites in the Grem1 genomic landscape overlap either with the GBR regions interacting with GLI3 [26, 32] or CTCF binding sites [31]. Analysis of Ctcf-deficient limb buds revealed its requirement for up-regulation and distal-anterior expansion of Grem1 expression [31]. Interestingly, the HS site overlapping with GBR4 [32] is not conserved between mammals and birds (Figure 8 and data not shown). While the early GBR4 activity is comparable to GRS1 in the posterior mesenchyme (Figure 4B), it does not support distal-anterior expansion of the expression during subsequent development [32]. Thus, the interaction of the GBR4 and GRS1 cis-regulatory regions could provide Grem1 expression with the necessary robustness and evolutionary plasticity, as has been postulated for the transcriptional regulation of HoxD genes by multiple interacting regulatory regions during limb bud development [8]. Taken together, these studies establish that the five HS sites in the Grem1 genomic landscape are functionally relevant for Grem1 expression in mouse limb buds, but not all of them are evolutionary highly conserved. Indeed, experimental and comparative evolutionary evidence indicates that alterations in the spatio-temporal expression of Grem1, and thereby the activity of the SHH/GREM1/AER-FGF feedback loop, likely contributes significantly to variations in digit numbers and morphologies (reviewed in ref. [49]).

Ethics Statement

All the experiments were conducted in strict accordance with Swiss law following the 3R principles and the conduct defined by the Basel Declaration. All studies involving mice were classified as grade zero by the Animal Welfare and Ethics Commission of both cantons of Basel and Argovia, which implies minimal suffering.

Mouse strains

The Shh loss-of-function allele [50], the Gli3 extra-toes-J allele [51], the Grem1 null allele [13] and the GRS1 ^tg/+ alleles (this study) were maintained in an NMRI background, while the Smad4 conditional allele (Smad4 ^flox , [45]) and the Prx1-Cre transgene [46] were maintained in a C57BL/6 J background.

BAC modifications

BAC constructs were engineered using ET cloning as previously described [7]. The deletions ΔHMCO1, ΔHMCO2 and ΔHMCO3 were induced in the Grem1-LacZ BAC clone using a Kanamycin resistance selection cassette. All primer sequences are available upon request. See Additional file 8 for the genomic coordinates of HMCO1-3.

ßglob-LacZtransgene constructs

An expression cassette was constructed in pKS-Bluescript using LacZeocin reporter under control of the minimal human ß-globin promoter sequences [52]. Various combinations of HMCO sequences were inserted downstream into this cassette. The HMCO1, -2 and -3 core sequences were amplified by PCR directly from mouse genomic DNA (primers available on request). The 9 kb GRS1 region was initially subcloned as a 9.6 kb Bgl2 fragment from BAC RP23-113 H17 (chr2: 113,611,499-113,847,299 (mm10). BacPac Resources, Children’s Hospital, Oakland, USA). All constructs were linearized with Ksp1 before microinjection.

Generation of transgenic founder embryos and transgenic strains

BAC and ßglob-LacZ transgenic constructs were injected into the pronucleus of fertilized mouse eggs. Several founder embryos for each construct were scored for ß-galactosidase activity (Additional file 3). For the GRS1 ßglob-LacZ transgene, seven independent transgenic mouse strains were established by crossing founders into the NMRI background. Transgene copy numbers were determined by real time qPCR [53] using the Bio-Rad CFX96 Real-Time PCR System in combination with the iQ SYBR Green Supermix (Bio-Rad). For each mouse, 20 ng of genomic DNA were analysed in triplicate. The primers to amplify the DBH genomic region were used for normalization and non-transgenic littermate DNA was used as control. The normalized control levels were set to 2 as the primers also amplified both alleles of the endogenous HMCO1 regions. The 2^-ΔΔCt formula was used to calculate and normalize transgene copy numbers. In addition to three different F1 males, three embryos from the F2 and three embryos from the F3 generation were analysed for the two transgenic strains shown to ascertain stable transmission of the transgenes. Primers sequences are available on request.

Detection of ß-galactosidase activity

Embryos were isolated in ice-cold PBS and staged according to somites numbers, then fixed in 1% formaldehyde, 0.2% glutaraldehyde, 0.02% NP40, 0.01% sodium deoxycholate in 1x PBS for 20–30 minutes at 4°C. Subsequently, embryos were washed three times in 1x PBS for 5 minutes at room temperature and incubated overnight at 37°C and in the dark in 1 mg/mL X-Gal, 0.25 mM K3Fe(CN6), 0.25 mM K4Fe(CN6), 0.01% NP-40, 0.4 mM MgCl2 and 1% sodium deoxycholate to detect ß-galactosidase activity, which colours cells blue. To stop the reaction, embryos were washed three times in 1x PBS for 5 minutes each at room temperature.

Chromatin Immunoprecipitation (ChIP)

Forelimbs and hindlimbs of 10 wild-type or Gli3 ^Xt/Xt embryos at E11.5 were dissected and processed for ChIP as described [26] using a polyclonal anti-GLI3 antibody (#2676, Genentech, [47]). To compute the level of enrichment of a given region, the Ct values of both input and ChIP samples were compared with those of a negative control amplicon located in the mouse ß-actin locus [21]. All results (mean ± SD) were obtained by analysing three completely independent experiments per genotype. The significance of all differences was assessed using the two-tailed, non-parametric Mann–Whitney test. The coordinates of the relevant amplification are shown in Additional file 7.

All genomic sequence alignements of the Fmn1-Grem1 locus region were performed using the ECR browser [54] and the ClustalW2 program [55].