Lmx1b-targeted cis-regulatory modules involved in limb dorsalization

The presence of morphologically distinguishable dorsal and ventral limb asymmetry reflects the polarized pattern accomplished during development. In mouse limbs, hair only appears on the dorsal surface of the autopod and nails on the dorsal aspect of digit tips, whereas footpads and ecrine glands are restricted ventrally. Internal differences are also present. Muscles develop as dorsal extensors and ventral flexors with disparate attachments, bones show subtle asymmetry, and joint morphology supports ventral flexion.

The polarity of the limb ectoderm is achieved by the restricted expression domains of wingless-type MMTV integration site family member 7a (Wnt7a) and engrailed 1 (En1) (Parr and McMahon, 1995; Loomis et al., 1996; Cygan et al., 1997; Loomis et al., 1998). As the limb begins to emerge, Wnt7a expression wraps around the limb bud apex (Bell et al., 1998). Bmp signals from the lateral plate mesoderm trigger the activation of En1 in the ventral limb ectoderm (Pizette et al., 2001). En1 expression expands in the ventral ectoderm, restricting Wnt7a expression to the dorsal ectoderm (Loomis et al., 1996; Cygan et al., 1997; Loomis et al., 1998). The restricted dorsal secretion of Wnt7a imparts polarity to the underlying limb mesoderm by triggering the expression of Lmx1b, a LIM homeodomain transcription factor that is ultimately responsible for limb dorsalization (Chen et al., 1998; Parr and McMahon, 1995; Riddle et al., 1995; Vogel et al., 1995). Mice lacking functional Lmx1b develop a ventral-ventral limb phenotype, whereas ectopic ventral expression of Lmx1b leads to a dorsal-dorsal limb phenotype (Chen et al., 1998; Cygan et al., 1997; Vogel et al., 1995). In humans, haploinsufficiency of LMX1B is associated with nail-patella syndrome, which is characterized by nail dysplasia, absent or hypoplastic patellae, progressive renal disease and decreased bone mineral density (Chen et al., 1998; Towers et al., 2005; Dreyer et al., 2000).

Despite the striking effect of Lmx1b on dorsal limb morphology, direct targets in the limb have been elusive. A number of potential targets have been suggested by comparative gene arrays between wild-type and Lmx1b knockout mice, but none has been confirmed (Feenstra et al., 2012; Gu and Kania, 2010; Krawchuk and Kania, 2008). Therefore, in order to elucidate direct limb targets for Lmx1b, we performed chromatin immunoprecipitation against Lmx1b followed by massive parallel sequencing (ChIP-seq) of mouse limbs during dorsalization [embryonic day (E) 12.5] and compared these sites of Lmx1b binding with genes regulated by Lmx1b. In this article, we describe the identification of Lmx1b-bound potential cis-regulatory modules (PCRMs) associated with Lmx1b-regulated genes and the predicted functional pathways affected. These data suggest that Lmx1b exerts its dorsalizing effects through the targeted regulation of genes involved in bone/joint development, extracellular matrix composition, axon tracking and cell proliferation. Intriguingly, we also identified an Lmx1b-bound PCRM upstream of the Lmx1b gene that provides the capacity for autoregulation.

Feenstra and co-workers (Feenstra et al., 2012) found more limb bud genes differentially expressed in the presence of Lmx1b at E12.5 than at either E11.5 or E13.5. Therefore, we performed ChIP-seq analyses on E12.5 mouse limbs. After peak calling analysis (MACS algorithm v1.4.2, cutoff P<1e-5), we identified 735 Lmx1b-bound genomic fragments or intervals (LBIs) common to ChIP-seq replicates with an average length of 470 bp (Table S1).

Comparison of LBIs with the annotated mouse genome showed that intronic (36%) or intergenic (54%) regions were the most common sites for Lmx1b binding (Fig. 1A). Of the regions mapped, 1% were within coding regions, 2% were found within the 5′UTR and 1% within the 3′UTR. Finally, only 6% of the intervals localized to potential promoters [−2500 and +500 bp from the transcription start sites (TSSs) based on UCSC gene annotation].

MEME-ChIP (Bailey et al., 2015) analysis for de novo motif discovery retrieved TMATWA (M=C or A, W=T or A) as the most common motif found in the Lmx1b-bound genomic regions (P=3.5e-29) (Fig. 1B). The TMATWA motif includes the published Lmx1b-binding site (TAATTA) (Morello et al., 2001) and the distribution of both motifs within intervals was similar (P=4.5e-24), consistent with Lmx1b as the ChIPed transcription factor.

The biological functions of promoter-associated intervals were predominantly related to genes with general cell functions as determined by GREAT analysis (Fig. 1C) (McLean et al., 2010). The intronic-associated intervals correlated with genes involved in growth and organ development, including limb morphogenesis. Remarkably, the intergenic-associated genomic regions were enriched near genes related to limb and skeletal development, which are functions that are anticipated to be modified by Lmx1b.

Although the annotated location and GREAT analysis give some insight into the potential regulatory activity of the LBIs, cis-regulatory modules (CRMs) are characterized by histone-modifying co-factors and bound proteins, collectively called chromatin regulatory marks, that affect genome accessibility (Hardison and Taylor, 2012). p300 (Ep300 – Mouse Genome Informatics) is a ubiquitous phosphoprotein with intrinsic histone acetyltransferase activity that predicts regulatory activity in a tissue-specific manner (Visel et al., 2009). Because Lmx1b is also expressed in other tissues, including forebrain, midbrain, hindbrain, spinal cord, kidney and the eye (Chen et al., 1998; Asbreuk et al., 2002), we evaluated whether the distribution of the 735 LBIs conformed to a limb-specific pattern. We compared the 735 LBIs with p300 ChIP-seq data in available embryonic tissues, i.e. limb, forebrain and midbrain (Visel et al., 2009), and confirmed that more LBIs colocalized with p300 from the limb (288, ∼39%), than from either the forebrain (25, ∼3%) or midbrain (1, ∼0.1%) (Fig. S1A), consistent with a limb-specific pattern of regulatory activity.

Overlap of an LBI with multiple chromatin regulatory marks increases confidence in its role as a regulatory sequence (Hardison and Taylor, 2012). Thus, in addition to p300, we compared the LBIs with available ChIP-seq data for chromatin regulatory marks associated with enhancer elements (H3K27ac and H3K4me2) (Heintzman et al., 2009) and regulatory regions undergoing active transcription (RNAP2 and Med12) (Kagey et al., 2010) in E12.5 and E11.5 limbs. We recognize that regulation during these stages is dynamic and what is active at E11.5 might not be active at E12.5; however, these data would still highlight a PCRM. After peak calling analyses, the Lmx1b ChIP-seq reads exhibited a 7-fold enrichment in tagged sequences within the 735 LBIs when compared with the input DNA (Fig. 1D). We next analyzed the distribution of tagged sequences within LBIs that overlap chromatin regulatory marks. We found that Lmx1b-ChIPed DNA was enriched within enhancer-associated chromatin marks (H3K27ac and H3K4me2) and exhibited a 3-fold enrichment over input DNA, whereas the enrichment within actively transcribed regulatory regions was 5-fold (RNAP2 and Med12) (Fig. S1B). Overall, Lmx1b ChIP-seq DNA sequences show a 4-fold enrichment within chromatin regulatory marks, consistent with a role for Lmx1b as a factor involved in active regulation (Fig. 1D; Fig. S3). As both CRMs and promoters are active regulatory regions, we examined the TSSs for genes regulated by Lmx1b (Feenstra et al., 2012). Similar to the genomic distribution of LBIs, no TSS enrichment was found in the distribution of tagged sequences between Lmx1b-ChIPed DNA and input DNA (Fig. 1D) supporting the concept that Lmx1b-mediated regulation favors CRMs over promoters.

We required at least two active regulatory marks for an interval to be categorized as a PCRM. Based on this criterion, we determined that 617 of the 735 LBIs were PCRMs (Fig. 2A; Table S2). The number of PCRMs was 30 times higher than random genomic regions with comparable characteristics (n=5 groups, each with 735 random genomic intervals; one-sample t-test P<1e-4) (Fig. S2A). The PCRMs were dispersed throughout the genome with the largest number (51) being found in chromosome 1, the largest chromosome. Interestingly, no PCRMs were identified in the X chromosome, although it is nearly as large as chromosome 2 and 28 Lmx1b-regulated genes are found on the X chromosome (Fig. S3). LBIs that overlapped repressive regulatory marks (H3K27me3) were found in 50 intervals, 41 of which were also associated with two or more active regulatory marks (Table S2). Intervals associated with repressive regulatory marks alone or with fewer than two active regulatory marks (n=9) were considered inactive PCRMs or nonspecific Lmx1b binding (Table S2).

In order to determine direct limb targets, we compared Lmx1b-bound PCRMs with genes previously reported to be regulated by Lmx1b at E12.5 (Feenstra et al., 2012). We found 292 PCRMs within 1 Mb of 254 Lmx1b-regulated genes (Fig. 2A; Table S3). In contrast, when we compared random non-Lmx1b-bound regulatory sequences (n=5 groups, each with 292 H3K27ac -positive intervals) with the 254 Lmx1b-regulated genes, only 60 (s.d.=7) regulatory sequences on average associated with 75 (s.d.=11) genes. This indicates that Lmx1b-PCRMs are enriched for their associated gene targets 5-fold over random regulatory sequences (one-sample t-test analysis, P<1e-4) (Fig. S2B). We also found that some PCRMs were associated with more than one gene and some genes were associated with multiple PCRMs (Fig. 2A). Enrichment of Lmx1b-bound gene-associated PCRMs was validated by qPCR using two independently generated ChIP samples. Nine PCRMs were selected and each exhibited at least a 4-fold enrichment when compared with an unbound control region (Fig. 2B).

Ingenuity Pathway Analysis (IPA) of the 254 genes associated with PCRMs demonstrated a significant effect on eight functional categories relevant to limb development. These categories include connective tissue development/function, skeletal and muscle system development/function and cellular movement, with some genes present in more than one functional category (Fig. 2C; Table S4). Within the functional categories affected, genes were further subcategorized into limb-related annotated functions (Fig. 2D; Table S4) including 17 that mapped to the annotated biological process termed ‘limb development’. IPA analyses also predicted that canonical pathways involved in movement, proliferation, cell adhesion, cytoskeletal regulation, and vascular development were targeted (Fig. S4).

Within the connective tissue development category, we found several genes involved with extracellular matrix (ECM) composition that are associated with Lmx1b-bound PCRMs including Col1a2, Col11a2, Kera, Lum, Dcn, Matn1/4, Epyc and Has3 (Table S4). In addition, a number of genes associated with bone/joint differentiation, Osr2, Gdf5, Runx2, Sox11 and Trps1 (Table S4), were present within the skeletal and muscle systems development/function category.

We performed comparative analyses using the VISTA browser (Frazer et al., 2004) to identify conserved PCRMs that were more likely to be functional across species. A high degree of conservation (greater than 70% homology) was found in 289 LBIs. Approximately 90% of the conserved LBIs are PCRMs with 105 associated with 94 Lmx1b-regulated genes (Tables S2 and S3).

Further functional validation was performed on two conserved Lmx1b-bound PCRMs. One of the PCRMs (LBI407) is 66 kb upstream of Lmx1b on murine chromosome 2 (Fig. 3A; Table S3). In silico analysis for transcription factor-binding sites in LBI407 confirmed two potential sites for Lmx1b as well as several other transcription factors (Fig. 3B). LBI407 (1224 bp) was isolated from murine genomic DNA, linked to a GFP reporter and transfected into chick presumptive limb mesoderm by electroporation. LBI407 showed robust enhancer activity in the limb 48 h after transfection (Fig. 4A-C,E-G). Moreover, its activity was restricted to the dorsal mesoderm coincident with LMX1B expression (Fig. 4D,H) (Vogel et al., 1995; Riddle et al., 1995; Dreyer et al., 2000).

We validated the activity of another PCRM (LBI443 from Table S3) located 82 kb downstream of growth differentiation factor 5 (Gdf5) (Fig. 5A), a factor known to be involved in joint development (Settle et al., 2003). This PCRM contained multiple potential binding sites for Lmx1b, Sox and Osr2 transcription factors (Fig. 5B). A GFP reporter construct containing the LBI443 (867 bp) isolated from murine DNA was transfected into chick presumptive limb mesoderm. LBI443 exhibited enhancer activity in the elbow, wrist, and digit joints overlapping GDF5 expression (Fig. 6C-C″). Furthermore, using section in situ hybridization we demonstrate that LMX1B expression overlaps both LBI443 activity and GDF5 expression (Fig. 6E-G′) dorsally in the developing elbow joint, consistent with a role for LMX1B-mediated dorsalization.

An additional 91 Lmx1b-bound PCRMs correspond to previously published conserved non-coding DNA sequences available in the Vista Enhancer Browser, 73 of which have confirmed activity during limb development (Table S5) (Visel et al., 2007). We associated 27 of these PCRMs to 34 genes differentially expressed in the presence of Lmx1b (Table 1). Based on the expression patterns available in the Mouse Genomic Informatics database, we found that the activity of the Lmx1b-bound PCRMs associated with Osr2, Jag1, Wnt5a, Shox2 and Cbln4 genes overlapped their respective limb mRNA expression patterns, which also overlaps expression of Lmx1b (Loomis et al., 1998; Lan et al., 2001; Witte et al., 2009; Cobb et al., 2006; Haddick et al., 2014).

We mapped the genome-wide distribution of Lmx1b binding in mouse limbs during limb dorsalization (E12.5). Most of the Lmx1b-bound genomic fragments or intervals (LBIs) mapped to intergenic or intronic regions (Fig. 1A) and associate to active enhancers rather than promoters (Fig. 1D; Fig. S1A,B), a feature shared by other development-related transcription factors (Sheth et al., 2016; McAninch and Thomas, 2014). Based on chromatin regulatory marks (chromatin modifications and bound proteins), nearly 84% (n=617) of the LBIs were categorized as PCRMs.

Typically, the distribution of either repressor (H3k27me3) or active (H3K27ac, RNAP2 or Med12) chromatin regulatory marks is mutually exclusive (Ram et al., 2011; Pasini et al., 2010); remarkably, a small population of Lmx1b-bound PCRMs, such as LBI407, aligned to both active and repressor marks (Fig. 3A; Table S2). This might reflect differences in regulation within the tightly restricted dorsal and ventral compartments of the developing limb, with potentially different genomic landscapes (Arques et al., 2007; Cotney et al., 2013; Andrey et al., 2017) and the capacity for different transcription factors such as Lmx1b to modify activity of these PCRMs. As gene regulation during limb development is a temporally dynamic process, another interpretation could be that the availability or activity of the regulatory sequence is in transition.

Regulatory sequences that are foundational to the development of the vertebrate body plan show high degrees of conservation across species (de Laat and Duboule, 2013). Thus, the conserved PCRMs identified by comparative genomic analysis (n=257) are likely to play crucial roles in Lmx1b-regulated development (Table S2). We validated enhancer activity in two of these conserved PCRMs (LBI407 and LBI443; Figs 4 and 6, respectively) and further confirmed that 80% of the PCRMs that correspond to conserved CRMs have limb activity (Vista Enhancer Browser database; Visel et al., 2007) (Table S5).

The distance between a regulatory sequence and its target gene can be highly variable and can associate with targets beyond the nearest gene (Marinic et al., 2013; Carter et al., 2002). Genomic territories with enhanced chromatin interactions, called topologically associating domains, can extend over roughly 1 Mb in vertebrates (Dixon et al., 2012). Nearly half of our PCRMs (n=292) were within 1 Mb of an Lmx1b-regulated gene (Fig. 2A; Table S3). The two functionally active PCRMs (LBI407 and LBI443) validated were associated with Lmx1b-regulated genes (Lmx1b and Gdf5, respectively) with enhancer activity overlapping associated gene expression (Figs 4 and 6). Twenty-seven of the PCRMs confirmed via the Vista Enhancer Browser (Visel et al., 2007) were also linked to Lmx1b-regulated genes (n=34). Five had readily available limb expression patterns that overlapped the associated PCRM/CRM activity (Osr2, Jag1, Wnt5a, Shox2 and Cbln4) (Fig. 7) (Lan et al., 2001; Loomis et al., 1998; McGlinn et al., 2005; Witte et al., 2009; Cobb and Duboule, 2005; Haddick et al., 2014).

PCRMs that were not associated with an Lmx1b-regulated gene at E12.5 are likely to be accessible, but might not be active at this stage. Because limb development is dynamic, stage-specific factors, in addition to Lmx1b binding, might also be required for PCRM activity. Correspondingly, Feenstra and colleagues demonstrated variation in Lmx1b-regulated genes at progressive limb stages (E11.5, E12.5 and E13.5), with less than 10% present across all three stages (Feenstra et al., 2012). Some PCRM-gene associations could also have been missed because of our 1 Mb cut-off for PCRM-target interactions. Although the average CRM-promoter distance is 120 kb (de Laat and Duboule, 2013), interactions have been reported up to a distance of 1.44 Mb (de Laat and Duboule, 2013; Benko et al., 2009).

Gene ontological analysis of candidate PCRM-associated genes predicted target pathways, and tissue systems present in both dorsal and ventral aspects of the limb. Differential compartment-specific regulation of CRMs common to limb tissues is a likely mechanism for refining dorsal asymmetry. Joint development is asymmetric along the dorsal-ventral axis. Growth differentiation factor 5 (Gdf5) is a well-established marker for joint development (Settle et al., 2003) with expression spanning both the dorsal and ventral limb compartments. Lmx1b-dependent regulation of the Gdf5-associated CRM (LBI443), an enhancer reported to drive Gdf5 expression in the developing limb joints (Chen et al., 2016), favors a model in which Lmx1b modifies dorsal Gdf5 enhancer activity.

Lmx1b-dependent regulation of CRMs associated with Wnt5a (LBI252) and ECM genes such as keratocan, lumican, decorin and epiphycan (LBI89 and LBI91) (Table S3) are additional mechanisms that could alter limb asymmetry. Wnt5a and ECM components can affect cell growth, survival, differentiation, migration and morphogenesis, which could be regulated differentially along the dorsal-ventral axis (Gros et al., 2010; Rozario and DeSimone, 2010; Koohestani et al., 2013). A role for Lmx1b in ECM regulation has already been demonstrated in the kidney with direct regulation of glomerular basement membrane collagens (Morello et al., 2001).

A role for Lmx1b in axonal guidance was proposed with the identification of targets such as Cbln4 and Ntn1 (Krawchuk and Kania, 2008; Feenstra et al., 2012). We demonstrate an Lmx1b-bound CRM (LBI456) associated with Cbln4 that mimics the pattern of Cbln4 expression (Visel et al., 2007; Haddick et al., 2014) and two Lmx1b-bound PCRMs (LBI122 and LBI123) that flank Ntn1 (Table S3), collectively supporting this hypothesis.

In humans, haploinsufficiency of LMX1B results in nail-patella syndrome (NPS) characterized by nail dysplasia, absent or hypoplastic patellae, bone fragility and premature osteoarthritis (Sweeney et al., 2003). We have identified several Lmx1b-bound PCRMs associated with Gdf5, Sox11 and several ECM-related genes (Tables S3) that are linked to osteoarthritis (Kan et al., 2013; Syddall et al., 2013; Stefansson et al., 2003; Miyamoto et al., 2007).

Another interesting finding suggested by our study is the positive autoregulation of Lmx1b expression. In the Lmx1b loss-of-function mutant, the Lmx1b transcript is reduced 5- to 6-fold (Feenstra et al., 2012). Functional validation of LBI407, located 66 kb upstream of Lmx1b, showed dorsally restricted enhancer activity in limb mesoderm coincident with Lmx1b expression (Vogel et al., 1995; Riddle et al., 1995; Dreyer et al., 2000). Positive autoregulation provides a mechanism for maintenance or amplification of Lmx1b expression following initial activation. Autoregulation during development reinforces or stabilizes a transcriptional pattern of differentiation (Crews and Pearson, 2009). This could also be of clinical importance since a population of NPS patients with mutations linked to the LMX1B locus fail to demonstrate mutations in the LMX1B coding sequence (Ghoumid et al., 2016). The fact that NPS results from haploinsufficiency of LMX1B signifies that the functional level of LMX1B is crucial for normal development. Disruption of the LMX1B autoregulatory system, via mutations in the LMX1B CRMs, could therefore account for NPS in families that lack mutations in the coding sequence.

In summary, we have generated a genomic data set of Lmx1b-bound PCRMs during limb dorsalization. Moreover, we have validated and linked Lmx1b-bound cis-regulatory modules to genes differentially expressed in the presence of Lxm1b, highlighting the processes regulated. These data will allow us to characterize the different mechanisms used by Lmx1b to accomplish limb dorsalization.

All animal procedures were performed in accordance with protocols established by the Institutional Animal Care and Use Committee of Loma Linda University. The genetic background of the mouse strain used for this study was C57BL/6. The sex was not determined in the embryos used.

Limb tissue from ten E12.5 embryos was pooled for each of the biological replicates (n=2) and submerged in PBS+1% formaldehyde for 15 min. After disruption with a Dounce homogenizer, lysates were sonicated and the DNA sheared to an average length of 300-500 bp. Chromatin (30 µg) was precleared with protein A agarose beads and incubated with 20 µl antibody against Lmx1b (BMO8) (Suleiman et al., 2007) kindly provided by Dr Witzgall (Institute for Molecular and Cellular Anatomy, University of Regensburg, Germany). The protein-DNA complex was reverse-crosslinked with an overnight incubation in Proteinase K at 65°C. ChIPed DNA was purified by phenol-chloroform extraction and ethanol precipitation.

DNA libraries were quantified and sequenced on Illumina's NextSeq 500. Reads were aligned to the mouse reference genome (mm10) using the default settings for Bowtie algorithm (Li and Durbin, 2009). Lmx1b peak locations were determined using the Model based Analysis for ChIP-seq (MACS) (v1.4.2) (Zhang et al., 2008) with a cutoff of P-value=1e-5 (empiric false discovery rate=6-16%).

Chromatin immunoprecipitation for qPCR was performed using ChIP-IT Express Kit (Active Motif) following the manufacturer's recommendations with minor modifications. Lysed cells were sonicated using an Epishear Probe sonicator (Active Motif) to obtain fragments ranging between 300 and 600 bp in length. The sonicated DNA (40 µg), Lmx1b-specific rabbit polyclonal antiserum (20 µl), and 25 µl of protein G beads were incubated overnight at 4°C. DNA elution and de-crosslinking was performed following manufacturer's recommendations. DNA was purified using the QIAquick PCR purification kit (Qiagen).

qPCR (Bio-Rad, CFX96) validation was performed using SYBR-Green (Bio-Rad, 172-5270). Validation was performed in triplicate using two independently generated ChIP samples with the Lmx1b antibody. Primers used for validation of ChIP-qPCR were assayed for primer efficiency and are listed in a 5′ to 3′ orientation in Table S6. Target enrichment was determined by calculating the percentage of target precipitated over the input DNA (% input) and adjusted for primer efficiency. Subsequently, fold enrichment was determined by comparing % input of specific targets over mouse negative control provided in the ChIP-IT qPCR Analysis kit (Active Motif).

Motif discovery analyses of Lmx1b ChIP-seq-retrieved genomic intervals were performed using the online tool MEME-ChIP 4.11.0 version (http://meme-suite.org/tools/meme-chip) as described by Ma et al. (2014). Input sequences were centered within summit regions of recovered intervals and extended 250 bp in each direction. MEME runs were performed with random subsampling and retrieved motifs between 6 and 10 bp in length with an E-value cut-off of >0.5 for the discovery of enriched motifs.

Association of genomic regions to genes was performed using the online tool Genomic Regions Enrichment of Annotations Tool (GREAT; http://bejerano.stanford.edu/great/public-2.0.2/html/) (McLean et al., 2010). Parameters were set so that regulatory domains for genes extended in both directions 1 Mb from the midpoint of the gene's TSS.

Limb ChIP-seq data on H3K27ac were obtained at the National Center for Biotechnology Information (NCBI) from the Gene Expression Omnibus database (GEO; http://www.ncbi.nlm.nih.gov/geo/) under accession numbers GSE30641 and GSE42413, p300 under accession number GSE13845, and both H3K27me3 and H3K4me2 under the accession number GSE42237. RNAP2 and Med12 ChIP-seq data were obtained from Berlivet et al. (Berlivet et al., 2013). The data for comparison were converted to the mouse build mm10 using the UCSC liftover tool. Lmx1b ChIP-seq-retrieved genomic intervals were extended 250 bp in both directions for comparison with published ChIP-seq data.

Comparative analysis of Lmx1b ChIP-seq-identified intervals was performed by pairwise alignment (Vista browser; http://genome.lbl.gov/vista/). Species selected for the pairwise alignment comparison were mouse, human, horse and chicken. An interval was considered conserved when it exhibited at least a 70% homology.

Lmx1b-regulated genes associated with LBIs were classified according to Gene Ontology terms using Ingenuity Pathway Analysis (IPA) software and database (Qiagen).

The two LBIs analyzed were isolated from mouse genomic DNA by PCR using the following primer pairs:

LBI407 (1226 bp fragment), 5′-GGGGACCAGGAGAAATATTACAGTGTG-3′ and 5′-CAGAATCCCCCCAGAGATAGATGC-3′; LBI443 (867 bp fragment), 5′-CTACAGCTCAGTCTCCTTCAGGCTACAC-3′ and 5′-CCATACATACTGAGCCACCACATGG-3′.

The PCR products were cloned into pCR-II TOPO vector (Qiagen) for subsequent subcloning into a thymidine kinase (tk) promoter-driven GFP reporter construct for functional analyses (Uchikawa et al., 2003).

Minimal promoter-driven GFP reporter constructs bearing the potential regulatory region of interest were delivered into presumptive limb region of Hamburger and Hamilton stage (HH) 14 chicken embryos as previously described by Pira and colleagues (Pira et al., 2008). Transfection efficiency was assessed by co-electroporation of a β-actin promoter-driven RFP construct. Electroporation was performed using the CUY21 electroporation station (Protech International). Depending on the construct of interest, embryos were incubated for 2-5 days before harvesting for visualization of GFP activity and digital image acquisition (Sony DKC-5000).

Whole-mount in situ hybridization was performed as described by Yamada et al. (1999). Section in situ hybridization was performed as described (Moorman et al., 2001). GDF5 and LMX1B probes were generated by RT-PCR as described by Merino et al. (1998) using the following primer pairs: cGDF5, 5′-GTAAGGACGGTGACTCCAAAGG-3′ and 5′-CCTTGCCTTCAGGTTCTTACTG-3′; cLMX1B, 5′-GGATCGCTTTCTGATGAGG-3′ and 5′-GATGTCATCATTCCTTCCATTCG-3′.

The authors would like to thank Dr Ralph Witzgall for the kind gift of the Lmx1b antibody (BMO8); Drs Nadav Ahituv, Thomas Girke and Paul Labhart for assistance and suggestions regarding data analysis and standardization; and Drs Salva Soriano and Brendon Gongol for critical review of the manuscript.