Development of the competency-ALM (CALM) assay
To assay patterning competency, we measured the response of limb cells to exogenous RA treatment. RA is a morphogen that plays key roles in embryonic axial patterning, and treatment of limb blastemas with exogenous RA results in the reprogramming of its pattern information to a proximal and posterior limb identity, regardless of where the blastema is located on the limb axis.24,30,31 Studies on ectopic blastemas generated by deviating a nerve to either anterior (A) or posterior (P) lateral limb wounds indicate that blastemas generated in these positions respond differently to RA administration (Fig. 1c-d).9,24,27 Ectopic blastemas in A positions express the A patterning gene ALX4. Following RA administration, ALX4 expression decreases, and the expression of the P patterning gene SHH is activated.9 The final outcome is the generation of ectopic/accessory limbs with determined A and P identities, indicating that RA reprograms A blastema cells to a P identity.9,24,27 Importantly, the induction of these accessory limbs also shows that these ectopic blastemas have acquired competency to broadly respond to limb patterning cues, since the combination of A and P signals in the blastema is required for limb outgrowth.9,28,32 In contrast, ectopic blastemas in P limb positions abundantly express SHH, and paradoxically, RA treatment suppresses this expression.9 Lacking both A and P patterning cues, these blastemas regress.9,24
Here, we leverage the site-specific RA-reprogramming phenotypes described above as an assay for positional competency called the CALM. The CALM is performed in both the A and P limb positions. In the A position (CALM-A), shifts in A/P patterning gene expression and the induction of tissue with determined P identity are evaluated (Fig. 1c). One of the strengths of CALM-A is the acquisition of these phenotypic responses if the cells are responsive to RA, which is a more stringent readout than loss of phenotype, which could be the result of multiple nonrelated factors. However, relative to the timing of blastema development, CALM-A readouts take a long time (one and ten weeks, respectively). In posterior limb positions (CALM-P), the suppression of SHH expression occurs rapidly following RA treatment (Fig. 1d).9 Therefore, the CALM-P is utilized as a fast assay for competency. Combined, CALM-A and CALM-P can be used to rigorously evaluate whether cells have acquired patterning competency.
Using the CALM to characterize the induction and maintenance of patterning competency:
We first utilized the CALM to determine the minimal environmental requirements for patterning competency. Because it was previously established that mature limb tissue was resistant to RA-mediated reprogramming,9,17,33 we first evaluated whether wounding alone was sufficient. Using CALM-A as a positive control, we observed by qRT-PCR that shifts in A/P patterning gene expression did not occur in RA-treated A-lateral wounds (Fig. 2a). We next evaluated whether the treated wound cells had gained a stabilized P identity. Since limb nerves are essential for limb regeneration,34 and A-lateral wound sites are not supplemented with a nerve, quantifying accessory limb induction directly would not allow us to determine if P identity had been generated. Therefore, we grafted the treated wound-site tissue, labeled with DiI on the day of wounding as a lineage tracer, into a host ALM (anterior) to determine whether the grafted tissue can induce accessory limbs. To determine the minimal time to evaluate the A-lateral wound tissue, we first determined that P identity was stabilized in control CALM-A blastema grafts from 3 to 7 days post-RA treatment (Fig. 2b and Supplemental Table 1.1). In contrast, the A-lateral wound tissue grafts (21 days posttreatment) did not induce accessory limbs (Fig. 2b and Supplemental Table 1.1). Together, these data show that wounding alone is insufficient, and the presence of limb nerves is required to induce patterning competency.
To elucidate the upstream mechanisms underlying patterning competency, we needed to determine the timing of its induction more precisely. We leveraged the CALM-P, which allows for a faster readout of RA responsiveness than the CALM-A. The expression of posterior patterning genes is detectable by qRT-PCR in nerve deviations (ND-Ps) performed on animals 5–7 cm snout-to-tip by 6 days post-surgery (Supplemental Fig. 1a-e), which is also the first time point that we detected significant suppression of SHH expression within 24 hours following RA treatment (Fig. 2c). Since an ectopic bump is present in the surgery site days before this time (Supplemental Fig. 1c), we tested whether the cell cycle was activated prior to competency induction using the new FUCCI transgenic axolotl line, which allows us to visualize the activation of the cell cycle in real-time in vivo.35 We observed a detectable increase in mitotic cells in the blastema mesenchyme between 5 and 6 days postsurgery (Fig. 2d), indicating that the acquisition of patterning competency correlates with the activation of the cell cycle.
As the blastema develops, the cell cycle slows in the region closest to the stump where blastema cells begin to redifferentiate.35,36 Therefore, we next tested whether patterning competency was retained in the blastema as it progressed from the early bud (EB) stage (composed of undifferentiated cells) to the late bud (LB) stage (composed of spatially restricted undifferentiated and redifferentiating cells). We performed qRT-PCR to detect differences in limb patterning gene expression following a 24-hour RA treatment of mature tissue, EB, and LB that was divided into apical (undifferentiated) and basal (redifferentiated) halves (Fig. 2e). We detected significant shifts in SHH and FGF8 (A/P gene) expression in only the EB and apical-LB samples (Fig. 2f). All RA-treated tissue samples showed an increase in the expression of MEIS2 (Fig. 2f), a gene involved in proximal/distal (Pr/Di) patterning in developing and regenerating limbs.37 Recent studies have shown that the expression of TIG1, which encodes a cell-surface protein that regulates Pr/Di identity in regenerating limbs, is also stimulated by exogenous RA in mature limb tissues.10 These observations show that both mature and regenerating cells respond to exogenous RA and that alteration of A/P patterning gene expression occurs specifically in undifferentiated blastema cells.
We further investigated the negative relationship between patterning competency and the redifferentiation of blastema cells. Previous research has shown that severing the limb nerve results in premature differentiation and the determination of pattern information in the blastema.23,38 Thus, we tested whether denervation negatively impacted expressional shifts using CALM-P (Fig. 2g). Since SHH expression remains elevated in ND-P blastemas 4 days post-denervation and is undetectable by 7 days (Supplemental Fig. 1g-h), we collected denervated CALM-P blastemas 4 days post-surgery. We found that RA treatment resulted in significant changes in SHH expression in the denervated CALM-P blastemas, although the fold reduction was more modest than that of CALM-P with sham denervation surgeries (Fig. 2h). We next evaluated whether denervated CALM-A blastemas established P identity following RA treatment by quantifying accessory limb formation (Fig. 2h and Supplemental Table 1.1). Because limb nerves are required for regenerative limb outgrowth, denervation surgery was performed close to the blastema to allow for reinnervation. Using this strategy, we observed that accessory limbs formed from both the denervated and sham CALM-A blastemas. However, we observed a decrease in the % (from 86–54%) of the accessory limbs with a complex pattern in the denervated samples (P ≤ 0001; chi-squared). We interpret these results to indicate that the nerve is required to maintain full competency in the blastema. Together, the expression data from the developing blastemas (Fig. 2f) and denervated CALM-P blastemas (Fig. 2g) and the accessory limb data from the denervated CALM-A blastemas (Fig. 2h and Supplemental Table 1.1) indicate that redifferentiating blastema cells lose patterning competency.
Global changes in H3K27me3 correspond with patterning competency
Treatment of patterning competent limb cells with exogenous RA results in changes in gene expression and positional identity, and our data show that although competent to respond to exogenous RA signaling, mature and redifferentiating blastema cells do not have the same response (Fig. 2f). We reasoned that one probable explanation for these divergent responses could be linked to differences in the epigenome that regulate the accessibility of limb patterning genes because chromatin restructuring and epigenetic changes are known to occur in the limb blastema.39–41 Therefore, we next identified changes to the epigenome that correspond with the acquisition of patterning competency. We focused on changes in the inhibitory mark H3K27me3, which regulates Polycomb Repressor Complex 2 (PRC2), because of its established association with embryonic limb patterning and maintenance of positional memory.15,42,43
We validated an H3K27me3-specific antibody on the axolotl using quantitative western blot and immunofluorescence on developing embryos and normal, denervated, and EZH2 (enzymatic activity of PRC2)-inhibited limb blastemas (Supplemental Fig. 2). Using this validated antibody, we performed ChIP sequencing on mature limb, A-lateral wound, and ND-A blastema samples. To identify stepwise changes in H3K27me3 marks, we compared data from A-lateral wounds and mature limb tissue and data from ND-A to A-lateral wounds. The normalized enrichment scores of GO biological process terms formed four distinct clusters where the largest cluster (C2) corresponded with limb patterning (Fig. 2i and Supplemental Table 2.1). Multiple limb patterning genes (SHH, PTCH2, EN1, FZD5 and FGF8) were observed in the list of the top 30 increased pathways enriched for the ND-A to A-lateral wound comparison (Fig. 2j and Supplemental Table 2.4). Therefore, we further analyzed the stepwise changes in H3K27me3 on these and other limb patterning genes that had significant changes in enrichment in the samples.
We hypothesized that this mark would decrease limb patterning genes in competent cells based on the observations that 1) multiple patterning genes are re-expressed by competent limb blastema cells, and 2) the global abundance of H3K27me3 decreases in the early blastema (Supplemental Fig. 2g). Most of the limb patterning genes show lower or no change in H3K27me3 enrichment when comparing the A-lateral wound to mature tissue. Surprisingly, almost all the patterning genes showed an increase in H3K27me3 in the ND-A blastemas compared to the A-lateral wound tissue (Fig. 2k and Supplemental Table 2.6). Since some of these genes are expressed in the ND-A blastema and not in the lateral wounds (Fig. 2a), the presence of H3K27me3 does not necessarily prevent the expression of these genes. These data show that H3K27me3 enrichment on patterning genes is dynamic as wound cells transition into patterning competent cells. Finally, along with the gene groups associated with limb patterning, we identified the cell cycle and immune and stress response groups as targets of H3K27me3 regulation in patterning competent cells (Fig. 2l-m and Supplemental Table 2.7-8).
Using the CALM to identify specific nerve signals is sufficient and required for patterning competency.
The above data show that the presence of nerves in lateral limb wounds is required and sufficient to induce patterning competency. Previous studies have shown that treatment of limb wounds with gelatin beads soaked in a combination of BMP2, FGF8,
and FGF2, factors that are expressed in the limb nerves,44,45 are sufficient to make A-lateral wounds responsive to exogenous RA.9 Here, we assayed expressional changes and accessory limb formation in CALM-A (performed on A-lateral wounds without nerves) to test whether all of these factors (B2FF) are required to induce competency or whether a combination of only two factors (F2F8, B2F8) is sufficient (Fig. 3a-c). We observed that while all growth factor treatments resulted in the formation of an ectopic bump, only the B2FF and B2F8 treatments had a significant increase in ALX4 and SHH expression compared to the negative control (Fig. 3b) and resulted in the generation of accessory limbs (Fig. 3c and Supplemental Table 1.2). These accessory limbs were able to regenerate and induce ectopic limbs when grafted into an A-ALM host site, confirming that stable/determined posterior identity had been established in the CALM-A (Supplemental Fig. 3 and Supplemental Table 1.3). Together, these data show that BMP2 and FGF8 are required and minimally sufficient to induce patterning competency in A-lateral wounds.
Using an inhibitor-based approach, we next sought to identify which of the nerve-dependent H3K27me3 patterns in patterning competent cells require BMP and FGF signaling QRT-PCR analysis for expressional targets of BMP (MSX1, MSX2) and FGF (PRRX1) signaling in ND-As treated with either BMP-signaling inhibitor (LDN193189,
BMPi), FGF-signaling inhibitor (SU5402, FGFi), or both (BiFi) validated that the inhibitors suppressed the expression of their respective signaling targets compared to the DMSO control in vivo (Fig. 3d). We then performed CUT&RUN for H3K27me3 on A-lateral wounds and ND-As treated with DMSO, BMPi, and FGFi of BiFi.
We performed differential abundance and GSEA analysis and found that the normalized enrichment scores (NES) of GO biological process terms segregated into 12 clusters when the treated ND-A samples were compared with the A-lateral wounds. As in the H3K27me3 ChIP-seq analysis (Fig. 2), limb patterning, cell cycle, and immune-related terms were strongly represented in these clusters (Fig. 3e and Supplemental Table 3.1-6). Further analysis comparing inhibitor- to DMSO-treated ND-As NES identified 6 clusters, the largest of which (C2) consisted of terms with decreased enrichment in the FGFi-treated samples and increased enrichment in the BMPi-only sample (Fig. 3f and Supplemental Table 3.7–10). Identification of the top 30 significant pathway-enriched gene regions from these comparisons consisted of genes involved in limb patterning (TBX1, TBX2, TBX3, GLI2, MSX1, HOXB8), limb development (ROR2, MMP14, OSR2, CHRD, CXCL12, NR4A2) and neural-related (DMRT3, LHX3, NDNF, NTRK2, GATA3, NDNF, CNTN2, TLX3, SLC6A4, NHLH2, PTPRS, NEFL) (Fig. 3G and Supplemental Table 3.12).
Using this dataset, we performed an ENSMBL search for genes associated with various pathways that play important roles in blastema formation and development. We found that many of these genomic regions appeared to have distinct dependencies upon BMP and FGF signaling (Supplemental Fig. 5 and Supplemental Table 4). Therefore, we next identified the top 30 most significant genomic regions with nerve-dependent H3K27me3 marks that required BMP and/or FGF activity (Fig. 3h-m). PANTHER analysis of the regions that showed FGF-specific dependency (flipped enrichment trend in FGFi and BiFi samples compared to DMSO) were almost exclusively neural-related GO terms (Fig. 3h, and Supplemental Table 3.13-14). BMP-specific dependent regions included neural-related and WNT signaling-related to dorsal/ventral (D/V) limb patterning GO-Terms (Fig. 3i, and Supplemental Table 3.15-16). Since FGF or BMP signals alone are not sufficient to induce patterning competency (Fig. 3j-k),46 we next identified the regions that required both FGF and BMP signaling, which likely include the key players in competency induction. PANTHER analysis showed that codependent marks (flipped in all inhibitor-treated groups compared to DMSO) were diverse, including RA receptor-, WNT signaling-, and cell cycle regulation-related GO terms (Fig. 3j, and Supplemental Table 3.17-18). The most significant additive marks (only the flipped trend when both pathways were inhibited) were ERBB signaling and angiogenesis-related GO terms (Fig. 3k, and Supplemental Table 3.19-20). We additionally identified FGF- and BMP-independent genomic regions that are nerve-dependent, which consisted of nerve-related GO terms (Fig. 3l and Supplemental Table 3.21-22). PANTHER analysis of the genomic regions that exhibited different patterns from these previous groups (only flipped in one single inhibitor group) identified tissue morphogenesis terms (Fig. 3m and Supplemental Table 3.23-24).
Because a third of the top 20 pathway-enriched genes that have FGF- and BMP-dependent changes in H3K27me3 enrichment consisted of limb patterning genes, we performed a more in-depth analysis of the genomic regions regulated in these and other limb patterning genes that showed significant changes in this mark relative to A-lateral wounds (Fig. 4). Each gene region identified was categorized as FGF-specific, BMP-specific, codependent/additive, FGF/BMP independent, or other based on the criteria described above (Fig. 4a). This strategy allowed us to identify specific regions of patterning genes where the regulation of H3K27me3 occurred and determine whether the marks were dependent on FGF and/or BMP signaling.
All the patterning genes that had significant changes in H3K27me3 enrichment contained regions that were sensitive to FGF and/or BMP signaling (Fig. 4 and Supplemental Table 5). To determine whether there were trends in the locations where marks were identified, we documented the location of “hits” on simple maps of each gene analyzed (Fig. 4a). The largest number of FGF/BMP sensitive regions were identified in A/P patterning genes, many of which were located within the first intron (Fig. 4a). However, this trend was not unanimous. For example, the 3’ and 5’ intergenic regions of TBX2, the patterning gene with the largest number of hits in this analysis, contained the most changes in H3K27me3.
The impact of BMP and/or FGF inhibition on the enrichment of H3K27me3 differed depending on the location of the region identified in the gene (Fig. 4b and Supplemental Table 5). For example, intron one of GREM1 had increased enrichment for H3K27me3 in the inhibitor-treated samples, while many of the intergenic regions had the opposite trend (Fig. 4b and Supplemental Table 5.1). In contrast, intron one of GLI1 and GLI2 showed decreased enrichment following inhibitor treatment, and marks in regions outside of these positions did not have obvious enrichment trends. These data show that FGF and BMP signaling are required for the regulation of H3K27me3 enrichment on limb patterning genes during the induction of patterning competency in A limb cells.