Regeneration-specific promoter switching facilitates Mest expression in the mouse digit tip to modulate neutrophil response

Introduction

Humans and mice can regenerate the distal region of the digits after amputation^1,2. This process involves the formation of a unique structure known as the blastema, which is the cellular source for the regenerated tissue and is a hallmark of epimorphic regeneration. The blastema is composed of numerous cell types, the majority of which are a heterogeneous population of fibroblasts, as we define by co-expression of markers Prx1, Pdgfrα, and Lum. Our prior study showed that while closely related by transcriptomics, the fibroblast population was composed of distinct subtypes suggesting separate biological functions³. We previously identified 67 genes enriched in the blastema fibroblasts³ and hypothesize that these are candidate pro-regenerative factors. Of these, we have prioritized Mest (mesoderm specific transcript), a member of the alpha/beta hydrolase family with structural similarity to epoxide hydrolases^4,5, though its substrate(s) is not defined.

Mest is a maternally imprinted gene first identified as a marker of mouse embryonic mesoderm and was originally named Peg1 (paternally expressed gene 1)^6,7; however, its exact biological function in development is still unclear. Beyond embryogenesis, Mest has been most studied in the context of adipogenesis and fat mass expansion in mouse obesity models^6,8,9. Roles for Mest in tissue regeneration have recently been reported: Mest expression is associated with de novo adipogenesis in regenerative wound healing¹⁰, Mest is necessary for skeletal muscle regeneration following CTX-induced injury¹¹, and expression of Mest can induce multipotency and increase differentiation potential in several cell types^12,13,14. Corroborating these results, our blastema single-cell RNA sequencing (scRNAseq) data show Mest is expressed in fibroblasts with no defined computational lineage trajectory¹⁵. This suggests that Mest may function in defining early regenerative cell states during digit tip regeneration.

As has been found in both mouse and human, the Mest genomic locus is imprinted via methylation of the maternal allele, resulting in exclusive expression of the paternal allele^16,17. Monoallelic expression resulting from this imprinting is well-described for embryonic Mest expression and persists in adult tissues including blood, intestine, and adipose tissue^7,18. Biallelic expression of Mest has been reported in human blood lymphocytes and sporadically in the mouse spleen^7,19. However, for the Mest locus, this phenomenon is better studied in human cancers, specifically in breast, colorectal, and lung carcinomas and cell lines^20,21,22,23. While Mest is strictly imprinted in embryos and often has biallelic expression in cancers, it is unknown how it is regulated during regeneration.

In this paper, we investigate the role that Mest, a maternally imprinted gene upregulated in the digit blastema, plays in mouse digit tip regeneration. We find that in vitro overexpression of Mest promotes bone differentiation, and in vivo Mest genetic loss of function results in delayed digit tip bone regeneration. Interestingly, paternal heterozygous knockout mice do not phenocopy the Mest homozygous knockout bone regeneration phenotype, but we establish that methylation of the imprinted region persists in the blastema. Instead, we find maternal allele expression arises via promoter switching to produce an alternate Mest transcript in regeneration, which does not occur during embryogenesis. Furthermore, through scRNAseq and FACs analyses, we find that neutrophil recruitment and clearance in the wounded tissue is impaired in the homozygous knockout mice, supporting a role for Mest in the inflammatory response. Ultimately, our data demonstrate that Mest is genetically necessary for proper digit tip regeneration and blastema Mest expression is facilitated by biallelic expression, despite genomic imprinting.

Results

Increased Mest expression is regeneration-specific in the mouse digit

Our previous digit tip blastema single-cell RNA sequencing (scRNAseq) study identified Mest as a putative pro-regenerative factor³. We have now generated a 28 day post-amputation (dpa) digit tip scRNAseq sample (Supplementary Fig. 1A, B), the stage where regeneration is complete, and integrated the data with our original unamputated (UA), 11, 12, 14, and 17 dpa scRNAseq dataset (Supplementary Fig. 1C–E). Mest expression was highly specific to fibroblasts (Supplementary Fig. 1F) with 94% of all cells expressing Mest as fibroblasts; thus, we computationally isolated and re-clustered the fibroblasts based on expression of Pdgfrα and Lum (Supplementary Fig. 1D and Fig. 1A). Mest expression was found in a subset of the fibroblast subpopulations (Fig. 1A’) and is highly upregulated in the blastemal fibroblasts from 11 to 17 dpa compared to the unamputated and regenerated (28 dpa) fibroblasts (Fig. 1B).

**Fig. 1: Mest is a candidate pro-regenerative factor.**

To assess whether Mest upregulation is specific to blastema-mediated regeneration or more broadly in response to wound healing, we evaluated Mest expression in regenerative distal post-amputation digits (terminal phalanx; P3) compared to non-regenerative proximal post-amputation digits (second phalanx; P2) which undergo fibrotic wound healing^24,25. By qPCR analysis, Mest expression is increased 15X in the 12 dpa blastema as compared to the UA digit mesenchyme (p = 7.0e-4; Fig. 1C). No statistically significant increase in Mest expression was found in 12 dpa P2 fibrosing tissue as compared to the UA digit (p = 0.095), and P2 expression was 2.5X less than P3 blastemal expression (p = 0.007). These findings were corroborated by hybridization chain reaction RNA fluorescence in-situ hybridization (HCR-FISH)²⁶; no appreciable Mest expression was detected in the P3 UA digit or the 12 dpa P2 fibrosing digit while significant Mest expression was found in the blastema³ (Fig. 1D–F, Supplementary Fig. 2A).

To determine the timing of Mest expression in the regenerating digit and if it is expressed earlier in the P2 non-regenerating tissue, we performed HCR-FISH at additional timepoints throughout regeneration and fibrosis. In the P3 digit tip, Mest is first expressed in a few cells adjacent to the amputation site between 3 and 4 dpa (arrowheads, Fig. 1G, H). The expression domain broadens by 6 dpa, continues to increase through the blastema stages, then decreases as the blastema matures and regeneration is completed (Fig. 1E, G–J, Supplementary Fig. 2A–C). In contrast, no Mest expression was found by HCR-FISH during early wound healing in the P2 non-regenerating digit at 3, 6, or 9 dpa (Supplementary Fig. 2D–F), correlating upregulation of Mest with blastema-mediated regeneration, not wound healing.

Mest does not induce multipotency for mesenchymal stem cell lineages in vitro

We performed computational lineage trajectory analysis with the integrated and subsetted fibroblast scRNAseq data using SPRING¹⁵ (Fig. 2A). From this, we can see that the fibroblasts in the UA digit tip project away from the blastema stage cells, indicating significant transcriptomic differences between homeostatic and regenerative fibroblasts. Additionally, the newly added 28 dpa fibroblasts project similar to the UA fibroblasts, supporting a return to a homeostatic state (Fig. 2A’). Several differentiation trajectories are predicted, including osteoblastic and other musculoskeletal and connective tissue lineages (Fig. 2A” and Supplementary Fig. 2G–J), consistent with our previous report³. Mest expression is associated with cells of early blastema stages, not the differentiating cells (Fig. 2A”), suggesting Mest expression is upregulated in undifferentiated and progenitor-like cells.

**Fig. 2: No Mest-induced in vitro MSC multipotency.**

Recent reports find in vitro overexpression of Mest in mouse adipocytes or human periodontal ligament cells induces transdifferentiation^13,14. Thus, we hypothesized that Mest expression in digit tip fibroblasts could induce multipotency for fibroblastic lineages. To test this, we utilized immortalized pre-osteoblasts (MC3T3) to determine if Mest overexpression could induce multipotency for adipogenesis and chondrogenesis, mesenchymal stem cell differentiation lineages distinct from osteogenesis (Supplementary Fig. 3A). MC3T3 cells were transfected with a constitutively expressing Mest construct (pCAG-Mest) or a control GFP construct (pCAG-GFP) and were compared with immortalized mouse mesenchymal stem cells (MSCs; C3H10T1/2) transfected with pCAG-GFP as a multipotency control. Mest RNA and protein expression were verified 24–48 h post Mest transfection (Supplementary Fig. 3B–D). Oil Red O staining for terminal adipogenesis 12 days post-induction (dpi) indicated robust differentiation of the control GFP transfected C3H10T1/2 MSCs and no significant differentiation in the control GFP MC3T3 pre-osteoblasts (Fig. 2B, C). Similarly, no significant terminal adipogenesis occurred in the context of Mest overexpression (Fig. 2D, E). To determine if Mest overexpression induced early adipocyte identity, 1, 7, and 12 dpi RNA was collected and evaluated for adipogenesis marker genes Perilipin-1 and PPARγ²⁷. Perilipin-1 expression in control C3H10T1/2 cells at 7 dpi increased 2000X (p = 2.4e-5) and significantly decreased by 12 dpi (p = 3.3e-5; Supplementary Fig. 3E); PPARγ expression increased by 30X at 12 dpi (p = 0.001; Supplementary Fig. 3F). While both GFP and Mest transfected MC3T3 cells expressed Perilipin-1 by 7 dpi, this increase was significantly less than the 2000X increase seen in C3H10T1/2 cells. Interestingly, while Perilipin-1 expression decreased by 12 dpi in GFP MC3T3 cells, it was maintained in Mest transfected cells, suggesting persistence of an early adipogenic-like state (Fig. 2F). The expression of PPARγ was similar in both GFP and Mest expressing MC3T3 cells with only a 7.5X increase in gene expression at 12 dpi (Fig. 2G). The absence of a significant difference in PPARγ expression between the Mest and GFP overexpressing cells at 12 dpi suggests the cells express low levels of adipogenic genes in response to the differentiation media, but do not truly differentiate.

To assess if Mest overexpression in pre-osteoblasts can generate progenitors competent for chondrogenesis we assessed terminal differentiation by Alcian Blue staining at 21 dpi. Robust chondrogenesis of the control C3H10T1/2 cells was found (Fig. 2H) though minimal differentiation of either GFP or Mest transfected MC3T3 cells occurred (Fig. 2I–K). RNA isolated at 7, 14, and 21 dpi was assessed for early and late chondrogenesis marker genes Col2a1 and Col10a1²⁸. Control C3H10T1/2 cells had peak expression of Col2a1 at 7 dpi that decreased 3X by 21 dpi while Col10a1 expression exhibited a 7.5X increase at 14 dpi (p = 0.02) and persisted through until 21 dpi (Supplementary Fig. 3G, H). Both GFP and Mest overexpressing MC3T3 cells had a peak expression of Col2a1 at 7 dpi with a sharp decline in expression at 14 and 21 dpi (Fig. 2L). However, neither had increased Col10a1expression by 21 dpi, supporting the negligible chondrogenesis found by Alcian Blue staining (Fig. 2M, Supplementary Fig. 3H). These data demonstrate that overexpression of Mest in pre-osteoblasts does not induce a mesenchymal stem cell-like multipotency competent for adipogenesis or chondrogenesis in vitro.

Mest expression modulates osteogenesis in vitro and in vivo

While Mest over-expression did not induce multipotency of pre-osteoblasts for other MSC lineages, we found it changed the timing and amount of osteogenesis in vitro (Supplementary Fig. 3I–L). 28 dpi cells stained with Alizarin Red revealed a 6.8X increase in ossification in the Mest versus GFP transfected C3H10T1/2 cells (p = 0.004; Fig. 3A–C), which was also observed for MC3T3 cells (Supplementary Fig. 3J–L). Consistent with these findings, Runx2 expression at 7 dpi was 2.5X higher in Mest expressing cells as compared to GFP (p = 0.04; Fig. 3D), demonstrating that Mest promotes osteogenesis in both mouse pre-osteoblasts and MSCs.

The ability of Mest to modulate in vitro osteogenesis suggests the expression of Mest in the blastema may regulate digit tip bone regeneration. To determine if genetic deletion of Mest yields a digit tip regeneration phenotype, we utilized a characterized Mest knockout allele²⁹. Mest is a maternally imprinted gene^4,16, thus we analyzed digit tip regeneration in the following genotypic cohorts: Mest wildtype (WT), heterozygous maternal knockout (mKO), heterozygous paternal knockout (pKO), and homozygous knockout (KO). Weights of 8–10 week-old mice did not differ between genotype cohorts within each sex (Supplementary Fig. 4A, B) supporting no gross developmental differences. Masson trichrome histology of 12 dpa digits revealed that all Mest genotypes formed blastemas that did not statistically differ in 2D area when averaged across at least 6 blastemas per genotype (Fig. 3E–H, Supplementary Fig. 4C). Masson trichrome histology at 28 dpa revealed that all tissues regenerated for all genotypes, with digits consisting of a full-length nail, bone, epithelium, and mesenchymal tissue surrounding the bone (Supplementary Fig. 4D–G). Additional cohorts were analyzed by whole mount Alizarin Red staining at 28 dpa, for which we measured the 2D area and length of each P3 bone (Fig. 3I–L, Supplementary Fig. 5A–D). While the unamputated digits across genotypes did not significantly differ in either measurement (Supplementary Fig. 5E, F), 28 dpa Mest-KO regenerated digit tip bones were significantly shorter (WT vs KO p = 4.30e-8) and exhibited less total bone regeneration (WT vs KO p = 2.3e-8) as compared to all other genotypes (Fig. 3M, N, Supplementary Fig. 5G, H). Interestingly, digits analyzed at 56 dpa, a stage well-beyond the completion of regeneration, revealed that Mest-KO digits had caught up with the other genotypes as assessed by percent bone area regeneration, and their percent bone length regeneration was significantly more than wildtype digits (Fig. 3O–T). This time point also revealed that the unamputated 56-day Mest-KO bones were significantly smaller and shorter than the other genotypes, suggesting a potential defect in attaining peak bone mass and/or in digit tip periosteal maintenance between 12 and 16 weeks (Supplementary Fig. 5I–P). Importantly, this phenotype is separable from regeneration; while the 2D bone area of the 56 dpa Mest-KO regenerated bones was significantly smaller than the other genotypes (Supplementary Fig. 5O), the KO digits regenerate comparably to the WT controls in bone area (Fig. 3S) and more than WT controls with respect to bone length (Fig. 3T, WT vs KO p = 0.014). Taken together, these genetic experiments indicate that Mest is not necessary for blastema formation, but may function in osteogenic differentiation pertaining to bone homeostasis and regeneration, corroborating our in vitro findings.

Genomic imprinting of the Mest promoter persists in the digit tip blastema

Previous studies have established that the Mest genomic locus is maternally imprinted via methylation around the first exon during embryonic development as well as in adult Mest expressing tissues^7,16 (Fig. 4A). Biallelic expression of Mest has been reported in primary human breast and colorectal cancers and lung cancers, both primary and cell lines, and at low frequencies in another mouse Mest knockout allele^{20,21,22,23,30,31}. Within this framework, we anticipated the Mest-pKO digits to phenocopy the Mest-KO cohort; however, no significant difference was found in percent regeneration or bone length at 28 dpa among the WT, mKO, and pKO groups (Fig. 3M, N). We first verified this was not secondary to expression changes in miR-335, a microRNA embedded in intron 2 of the Mest locus shown to promote osteogenesis^32,33 (Fig. 4A, B). With no significant difference in miR-335 expression among genotypes (ANOVA p = 0.122), the absence of Mest-pKO bone regeneration phenotype instead suggests continued expression of Mest in the paternal knockout regenerating digits. qPCR analysis confirms that there is no significant difference in WT and mKO 12 dpa Mest expression, and that KO blastema expression is 16.7X lower than WT (p = 8.4e-5; Fig. 4C). The pKO blastema expression is 3X less than that of WT blastemas (p = 8.7e-4) and 5.7X higher than KO, though not statistically significant (p = 0.19; Fig. 4C), which does not align with expectations for an imprinted gene. Moreover, the qPCR primers are specific to sequences within the knockout allele deletion (Fig. 4A); thus, any amplification is de facto from the wildtype allele. Therefore, Mest expression in the Mest-pKO 12 dpa blastema is derived from the maternal allele, supporting biallelic expression of Mest during digit tip regeneration.

**Fig. 4: Genomic imprinting of Mest persists in the blastema.**

To explore potential loss of imprinting of the maternal allele in digit tip regeneration, we used targeted bisulfite sequencing to assess methylation surrounding the canonical promoter (Fig. 4A). A differentially methylated region (DMR) surrounds exon 1, also referred to as a CpG island, due to hypermethylation of the maternal allele and hypomethylation of the paternal allele (Fig. 4A)¹⁶. We assayed the upstream and downstream regions of this DMR, Exon1-5’ and Exon1-3’ (Fig. 4A), using bisulfite converted genomic DNA isolated from Mest-WT, mKO, pKO, and KO 12 dpa digit tip blastemas, as well as from Mest-WT and pKO e12.5 embryonic hindlimbs as controls. Cloned PCR amplicons were sequenced and scored for observed to predicted methylated cytosine ratios, whereby a ratio nearing 0 reflects hypomethylation and a ratio nearing 1 reflects hypermethylation. Consistent with imprinting of the Mest locus during development, cytosine ratios for Exon1-5’ and Exon1-3’ clones from e12.5 hindlimb DNA were either hypo- or hypermethylated, even in the context of the paternal knockout (Fig. 4D, F). Importantly, this bimodal methylation of alleles was also found for Mest-WT 12 dpa digit tip blastemas, demonstrating that genomic imprinting of Mest continues during digit tip regeneration (Fig. 4E, G). Moreover, Mest genetic loss of function does not change methylation of the canonical promotor as seen by Mest-mKO, pKO, and KO 12 dpa blastemas (Fig. 4E, G, Supplementary Fig. 6A, B). These data demonstrate persistent genomic imprinting of Mest when it is highly expressed during regeneration, even under the pressure of genetic loss of function.

With genomic imprinting intact, we reasoned that biallelic Mest expression found in the pKO blastema could be regulated by an alternate promoter. Indeed, an alternate transcription start site exists 4.5 kb upstream of the Mest canonical promoter (Fig. 4A)^18,34. We focused on two regions to characterize DNA methylation: one within a CTCF binding element conserved with the human locus and one covering exon A, where another CpG island exists in humans but not mice³⁵. We evaluated methylation of these two regions, CTCF and ExonA (Fig. 4A, Supplementary Fig. 6C), by bisulfite sequencing. Clones through both regions in Mest-WT and pKO e12.5 hindlimbs revealed a uniform distribution of allelic methylation patterns, indicating that this upstream promoter is not imprinted or strictly regulated by methylation during limb development (Fig. 4H, J). Methylation of ExonA in the Mest-WT 12 dpa blastema was similarly uniform, and all amplicons through the CTCF element were significantly hypermethylated (Fig. 4I, K). However, unlike the embryonic limb, methylation of the upstream Mest promoter in the 12 dpa blastema changes in the context of Mest genetic loss of function. The CTCF region becomes increasingly hypomethylated in the Mest-mKO, pKO, and KO blastemas, and this demethylation is most dramatic in the KO blastema with an average cytosine ratio of 0.61 compared to 0.84 in the WT blastema (Fig. 4I, Supplementary Fig. 6D). The ExonA region becomes increasingly hypermethylated in Mest-mKO, pKO, and KO blastemas (Fig. 4K), specifically in CpG positions directly upstream of the transcription start site for the Mest210 variant (Supplementary Fig. 6C, E). However, with only 5 predicted methylated CpGs for the ExonA amplicon, interpretation of this finding is challenging. Collectively, these data highlight a change in gene regulation at an upstream promoter as a mechanism to compensate for Mest genetic loss of function during regeneration, suggesting that Mest-pKO blastema biallelic expression may result from this change in upstream regulation.

Biallelic Mest expression during regeneration occurs via promoter switching

To determine whether this shift in methylation of the Mest upstream promoter impacts transcription, we analyzed three Mest transcript variants: Mest202, Mest210, and Mest211 (NM_001252292, NM_001252293, and NM_008590). Mest202 and Mest211 are both regulated by the imprinted exon 1 promoter region and Mest210 is regulated by the alternate upstream exon A promoter region (Figs. 4A, 5A)^16,18. These variants differ in transcription and translation start sites, though they all encode the same protein with minor variation in the N-terminus (Fig. 5A and Supplementary Fig. 7A). Importantly, the MEST N-terminus is not predicted to contain any functional domains and all three variants encode the downstream enzymatic domain, supporting conserved function among them (Supplementary Fig. 7B).

**Fig. 5: Biallelic expression of Mest via promoter switching.**

Using variant-specific qPCR, we assessed Mest transcript variant expression in 12 dpa blastemas isolated from the Mest genotypic cohorts. We found that in the wildtype blastema, Mest211 is the predominant isoform, with significantly less Mest210 and Mest202 expression (3.7X and 47X lower, respectively; Fig. 5B). This variant expression profile was highly similar in Mest-mKO blastemas, which only express Mest from the paternal allele (Fig. 5C). In contrast, Mest-pKO blastemas predominantly express the Mest210 variant, with a 3.3X increase compared to canonical Mest211 (p = 5.3e-5; Fig. 5D), and 3.5X more Mest210 than Mest-WT blastemas (p = 0.008; Fig. 5E). The increased Mest210 expression in Mest-pKO blastemas is consistent with maternal allele expression derived from exon A that undergoes epigenetic changes (Figs. 4K, 5A), circumventing the imprinted Mest211 promoter (Figs. 4D–G, 5A). Additionally, the relative levels of Mest210 to Mest211 were 3-fold higher in the WT blastema (0.27, Fig. 5B) than the mKO blastema (0.08, Fig. 5C), supporting promoter switching on the maternal allele occurs in wildtype digits during regeneration. These analyses also revealed that Mest202 is the minor Mest variant, with extremely low expression in the blastema for all genotypes (Fig. 5B–D).

To determine if Mest promoter switching is unique to the blastema, we assessed these transcript variants in Mest-WT and pKO e12.5 embryos, when Mest is highly expressed and strictly imprinted. As found for the blastema, canonical Mest211 is the predominant Mest transcript in the wildtype embryos, with significantly less Mest210 expression (1000X lower) and Mest202 expression (59X lower) (Fig. 5F). In stark contrast to the Mest-pKO 12 dpa blastema, Mest210 transcription is not upregulated in pKO e12.5 embryos. Mest210 and Mest202 exhibit 105X and 48X lower expression relative to Mest211, respectively (Fig. 5G), and there is no significant difference in Mest210 expression between WT and pKO e12.5 embryos (p = 0.367, Fig. 5H). These data align with the lack of difference in Exon1, ExonA, and CTCF methylation profiles between Mest-WT and pKO embryos (Fig. 4D, F, H, J), ultimately resulting in negligible Mest expression in pKO embryos (Fig. 5I). These findings support previous studies establishing strict imprinting during development^7,19 and further illustrate promoter switching to Mest210 is specific to regeneration.

Mest knockout blastemas have impaired neutrophil response during inflammation

To begin to understand the function of Mest and the need for its high expression during digit tip regeneration, we performed single-cell RNA transcriptomics on Mest-WT and Mest-KO 12 dpa blastemas. Post-quality control and filtering, our dataset included 8802 wildtype cells and 6,898 knockout cells. All digit tip blastema cell types previously identified by scRNAseq^3,36 were present in both the WT and KO samples (Supplementary Fig. 8A). Integration of the WT and KO data revealed the two genotypes were well-mixed, confirming that Mest-KO indeed form a blastema (Fig. 3H, Supplementary Fig. 8A, B). All major cell types have proportional population sizes except for epithelial cells, though this is secondary to our dissection technique to minimize epithelium (Supplementary Fig. 8C). We subsetted and re-clustered the integrated fibroblasts for more detailed analysis since this cell type highly expresses Mest (Fig. 6A, Supplementary Fig. 8D). We looked for differentially expressed genes (DEGs) between the Mest-WT and KO fibroblasts with an average log2FC > 0.58 and expressed in at least 25% of the cells, and found 119 significantly upregulated genes in the Mest-WT fibroblasts and 22 upregulated genes in the Mest-KO fibroblasts (Fig. 6B, Supplementary Tables 2, 3). Of the top 10 upregulated genes in the wildtype fibroblasts, four are chemokines involved in leukocyte recruitment, especially neutrophils (Ccl2, Cxcl5, Cxcl1, and Cxcl2)³⁷, and other genes including Tnfaip3 and Nfkbia are involved in regulating inflammation^38,39 (Supplementary Table 2). Gene set enrichment analysis of the 119 upregulated Mest-WT DEGs established the IL-17 and TNF signaling pathways were significantly upregulated, consistent with leukocyte recruitment (Fig. 6C). To assess which fibroblast subpopulation(s) may be driving the inflammation associated DEGs, we used differential proportion analysis and found significantly fewer Mest-KO fibroblasts in cluster-0 as compared to Mest-WT (p = 0.007; Fig. 6D). Indeed, KEGG pathway analysis of cluster-0 marker genes filtered by average log2FC > 1.0 and expressed in at least 25% of cells revealed enrichment of IL-17 and TNF signaling among other pathways (Fig. 6E), and the chemokine DEGs we previously identified (Fig. 6B) are differentially expressed between Mest-WT and KO cluster-0 fibroblasts (Supplementary Fig. 8E, Supplementary Table 4). These findings suggest that Mest-KO blastema fibroblasts have reduced neutrophil chemoattractant production, primarily due to a reduction in the cluster-0 subpopulation.

**Fig. 6: Single-cell transcriptomic analysis of Mest wildtype and knockout blastemas.**

Finding differential inflammatory signaling between Mest-WT and KO fibroblasts led us to characterize the immune cells at a higher resolution. We subsetted and re-clustered the macrophages and monocytes, neutrophils, T cells, and pre-osteoclasts and found that the largest subpopulation change was in the neutrophils, in which the Mest-KO blastema had fewer neutrophils than the Mest-WT blastema (Supplementary Fig. 8F, G; cluster 2 red arrow). This reduction was not statistically significant by differential proportion analysis (p = 0.117) though it conceptually aligns with our finding of reduced expression of neutrophil recruitment chemokines in the Mest-KO fibroblasts. We used cell-cell communication computational tools CellChat⁴⁰ and NICHES⁴¹ to identify signaling relationships between fibroblasts and neutrophils. Both tools corroborated the decreased cytokine interactions between fibroblasts and neutrophils in the Mest-KO blastema (Fig. 6F, Supplementary Fig. 9A, B). In addition, we found that Annexin (Anxa1) signaling, a key pathway involved in the resolution of inflammation⁴², occurs in the wildtype blastema between many cell types including fibroblasts signaling to neutrophils (Fig. 6G). However, Anxa1 signaling is absent in the knockout blastema (Supplementary Fig. 9A–D, red arrows), which may be due to the lack of Fpr2, one of the main receptors for Anxa1, expression on KO neutrophils rather than a global reduction in Anxa1 expression (Supplementary Fig. 9E). Protein expression by blastema western analysis confirms this finding; no difference in ANXA1 expression is detected between Mest genotypes, yet we find a 1.4X reduction in FPR2 expression in Mest-KO blastemas (imageJ quantification normalized to Beta-tubulin) (Fig. 6H). This finding further underscores a disrupted immune response in Mest-KO digits.

To evaluate these findings in vivo, we performed flow cytometry for neutrophils in Mest-WT and KO blastemas at 9, 12, and 15 dpa (Fig. 6I, Supplementary Fig. 10A, B, Table 1). Consistent with our scRNAseq finding of reduced production of leukocyte chemoattractants, there were significantly fewer CD11b⁺ /Ly6g⁺ neutrophils at 9 and 12 dpa in the Mest-KO blastemas as compared to Mest-WT (p = 0.000 and p = 0.005, respectively; Fig. 6I). However, by 15 dpa there were significantly more CD11b⁺ /Ly6g⁺ neutrophils in the Mest-KO blastemas as compared to Mest-WT (p = 0.003), suggesting neutrophil clearance as required for the resolution of inflammation is delayed⁴³. Interestingly, there were significantly more CD11b⁺ /Ly6g⁻ non-neutrophil myeloid cells in the KO tissue at 9 and 12 dpa (Supplementary Fig. 10B, C, Table 2), the inverse of what was found for neutrophils suggesting a change in immune cell ratios. Taken together, these data indicate that the Mest-KO blastema has an impaired immune response due to reduced cytokine production in a subpopulation of fibroblasts, which prolongs the inflammatory process and ultimately delays digit bone regeneration.

Table 1 Neutrophil flow cytometry results

Full size table

Table 2 Non-neutrophil myeloid flow cytometry results

Full size table

Discussion

We began this study with the hypothesis that Mest is a pro-regenerative gene that can drive plasticity of fibroblast lineages during digit tip regeneration. However, our in vitro experiments show that Mest overexpression in pre-osteoblasts is not sufficient to induce a multipotent progenitor state capable of differentiating into adipogenic or chondrogenic lineages (Fig. 2). This data conflicts with other reports where Mest did induce lineage transdifferentiation, though the differences may be due to distinct species, cell types, or lineages assayed^13,14. Interestingly, the control in vitro osteogenic differentiation assays in our experiments revealed Mest overexpression significantly increased osteogenesis (Fig. 3A–D, Supplementary Fig. 3J–L), which ultimately complemented the delayed digit tip bone regeneration we found in the Mest-KO mice (Fig. 3I–T). This finding is nuanced because it is the composite of two distinct bone phenotypes: failure to attain peak bone mass/length and reduced bone regeneration. At 28 dpa (12 weeks old), Mest-KO digit tips have reduced bone regeneration, but the UA digits support normal bone development among genotypes. However, by 56 dpa (16 weeks old), the Mest-KO UA digits are significantly smaller in area and length than the other genotypes, suggesting a failure to reach peak bone mass/length which occurs around this time on the C57Bl6/J background^44,45, and/or reduced maintenance of the digit tip periosteum. Within the context of a smaller digit tip bone, the Mest-KO bones have fully regenerated by 56 dpa, consistent with delayed regeneration.

The mechanism of how Mest expression regulates osteogenesis and bone regeneration remains to be determined, though our scRNAseq experiments suggest this is an indirect effect. We found that Mest expression in a subpopulation of fibroblasts mediates cytokine production and ultimately neutrophil recruitment and clearance following digit tip amputation (Fig. 6, Table 1), likely resulting in an environment that promotes regeneration. Broadly speaking, the resolution of inflammation, which includes clearance of neutrophils, is a critical phase of wound-healing and tissue repair^43,46. Our scRNAseq cell-cell communication analyses highlight Annexin signaling, a key pathway in initiating the resolution of inflammation^42,46, is absent in Mest-KO blastemas (Fig. 6G, Supplementary Fig. 9A), which aligns with more neutrophils in Mest-KO 15 dpa blastemas (Fig. 6I). Intriguingly, MEST is structurally conserved with epoxide hydrolases, a class of proteins involved in the biosynthesis of specialized proresolving mediators (SPMs)^47,48. SPMs are signaling molecules in the resolution of inflammation and include maresins, resolvins, lipoxins, and protectins, which broadly aid in tissue regeneration^46,49. Taken together, these data suggest Mest is critical for resolution of inflammation and moving forward it will be important to experimentally anchor MEST to this process.

A key finding in this study is that regulation of Mest transcription in the digit tip blastema can escape genomic imprinting via promoter switching. A priori, we anticipated persistence of maternal imprinting in the blastema as has been defined for Mest during embryogenesis and in many adult tissues^7,16. Consistent with this hypothesis, we found that the canonical variant Mest211, regulated by the DMR surrounding the canonical promoter, contributed to the majority of Mest expression in Mest-WT and mKO blastemas (Figs. 4E, G,5B, C). Moreover, no significant expression of any Mest transcript variant was found in Mest-KO blastemas (Fig. 4C), as would be expected for a genetic knockout. However, while Mest-pKO blastemas maintain maternal imprinting of the Mest canonical promoter and have negligible Mest211 expression, Mest expression was unexpectedly found in the form of the Mest210 variant (Figs. 4E, G, 5D, E). These findings align with altered regulation via the upstream promoter, whereby Mest-pKO and KO blastemas have decreased methylation of the CTCF consensus region (Fig. 4I), which is associated with increased CTCF binding to modulate DNA conformational changes^50,51. While we found increased methylation surrounding exon A (Fig. 4K), this region can be defined as a low CpG promoter (five CpGs in 245 bp, Supplementary Fig. 6C, E), in which DNA methylation has been shown to increase transcription in various contexts^52,53,54.

These data provide strong evidence that reduced expression of Mest during digit tip regeneration results in Mest upregulation via promoter switching, not loss of imprinting. Why there is a demand to maintain high levels of Mest in the blastema remains to be understood, though it is clear this is not simply an artifact of our genetic system. We find almost no Mest210 transcription in e12.5 embryos and no significant change in the canonical or alternative Mest promoter methylation between Mest-WT and pKO embryos (Figs. 4, 5), indicating there is no compensation for genetic loss of Mest during embryogenesis. Since promoter switching does not occur during development but does in regeneration, this suggests different functions for Mest in these two processes, further supporting recent data that limb development genes are not re-deployed for the same function in the regenerating mouse digit tip⁵⁵. However, promoter switching^30,35 and biallelic expression at the Mest locus have been reported, primarily in human cancers^20,21,22,30. The mechanism behind the promoter switching, and whether a discrete set of transcription factors is required, is not yet known. While our study focuses on the regulation and function of a single gene, the connection we find between the blastema and cancer is a classical concept that has not been experimentally resolved^56,57. We have shown here that Mest is poised to shed light on this topic, and in future work it will be important to determine how many genes in the blastema share regulatory mechanisms with cancers when under genetic pressure.

Methods

Mice

All mice were housed in the Brigham and Women’s Hospital Hale BTM vivarium, maintained by the Center for Comparative Medicine. Breeding and surgery protocols were approved by BWH IACUC. Male and female wildtype 8-week-old FVB/NJ mice (JAX #001800) were used for HCR FISH expression analyses and 28 dpa scRNAseq. Mest genetic analyses used the existing Mest knockout allele, B6-Mest^tm1.2Rkz (courtesy of Dr. Robert Koza, MMCRI)²⁹. Mest allele genotyping primers from the original publication were used (Supplementary Table 1)²⁹.

Mouse digit tip amputation surgeries

Male and female 8 to 10-week-old mice were used for digit amputation surgeries and subsequent tissue collection. Mice were anesthetized with inhaled 1–2% isoflurane in oxygen during surgeries. For both hindlimbs, digits 2, 3, and 4 were amputated using a sterile disposable scalpel and data was pooled from all three digits for analyses. Distal (P3), regenerative amputations were made midway through the third phalangeal bone; proximal (P2), nonregenerative amputations were made midway through the second phalangeal bone. For analgesia, subcutaneous buprenorphine HCl (0.05 mg/kg) was given peri- and post-surgically. Mice were euthanized by carbon dioxide inhalation for post-amputation tissue collections.

Hybridization chain reaction RNA fluorescent in situ hybridization (HCR-FISH)

Post-amputation digits were fixed in 4% PFA at 4 °C overnight, followed by PBS washing and decalcification (Decalcifying Solution-Lite, Sigma-Aldrich). Tissues were cryopreserved through a sucrose gradient, embedded in OCT (Tissue-Tek), and sectioned at 20 µm with a Leica CM3050s cryostat. HCR-FISH protocol was performed as previously reported^3,26, using a commercial Mest probe (Genbank: NM_001252292.1; Molecular Instruments), and were counterstained with 1 ng/µL DAPI. TrueVIEW autofluoresence quenching kit (Vector Laboratories) was used for background reduction. Slides were imaged on a Zeiss LSM880 confocal microscope and z-stack images were processed using Fiji⁵⁸.

Masson Trichrome histology and Alizarin Red bone staining

For Masson Trichrome histology, post-amputation digits were fixed in 4% PFA at 4 °C overnight, decalcified with Surgipath Decalcifier I (Leica Biosystems), and cryopreserved through a sucrose gradient to OCT (Tissue-Tek). 14 µm sections made with a Leica CM3050s cryostat were stained using a Masson Trichrome Staining Kit (Sigma-Aldrich). Manufacturer’s protocol was followed except for the omission of deparaffinization and microwaving. Stained sections were imaged with a Leica DM2000 upright microscope and images were processed and analyzed with Fiji⁵⁸. 2D blastema area was quantified using Fiji and one-way ANOVA followed by post-hoc t-test with Bonferroni correction was used to assess statistical significance of the measurements among genotypes.

Whole mount Alizarin Red stained digit tip bones were generated as previously reported⁵⁵. Post-amputation digits and contralateral unamputated control digits were collected for skeletal staining at 28 dpa and 56 dpa; 6–8 mice were used for each genotype. Stained bones were imaged on a Leica M165FC stereomicroscope using the LAS X imaging software. Images were analyzed in Fiji to measure the 2D area and length of the P3 bone. Percent area regeneration was calculated for each regenerated digit as 2D area normalized to the average area of the unamputated digits for its respective genotype. Length was measured from the midpoint at the base of the P3 bone to the tip of the bone and percent length regeneration was calculated using the same method as for percent area regeneration. Statistical analysis among the genotypes was performed with one-way ANOVA followed by post-hoc t-test with Bonferroni correction.

Cell culture and in vitro MSC differentiation

To generate the Mest overexpression plasmid, we utilized the pCAG-GFP vector⁵⁹ (pCAG-GFP was a gift from Connie Cepko; Addgene #11150). GFP was excised with EcoRI and NotI digestion and a synthesized Mest211 cDNA gBlock (Genbank: NM_008590; Integrated DNA Technologies) was inserted with NEBuilder HiFi DNA assembly (New England Biolabs). The final construct was verified by plasmid DNA sequencing (Plasmidsaurus).

Mouse multipotent C3H10T1/2 cells (ATCC #CCL-226) and mouse pre-osteoblastic MC3T3-E1 cells (ATCC #CRL-2593) were maintained in αMEM media with 2mM L-glutamine, 1% penicillin-streptomycin and 10% FBS. 20,000 cells were seeded in 12-well plates and transfected with 300 ng of pCAG-Mest or pCAG-GFP control plasmid using FuGENE HD transfection reagent (Promega) per manufacturer’s protocol. Lineage-appropriate differentiation media was added 24 h post-transfection. For osteodifferentiation, media composition was αMEM with 10% FBS, 1% penicillin-streptomycin, 10 mM β-glycerophosphate, 50 μg/mL ascorbic acid, 100 nM dexamethasone⁶⁰. Cells were fixed with 4% PFA and stained with 40 mM Alizarin Red to visualize mineralization which was quantified using Fiji⁵⁸. For adipogenic differentiation, media was composed of DMEM with 10% FBS, 1% penicillin-streptomycin, 100 nM dexamethasone, 0.5 µM isobutylmethylxanthine, and 10 ng/mL insulin⁶¹. Terminal differentiation was assessed by Oil Red O staining which was quantified by OD490 nm reading of isopropanol extracted dye. For chondrogenic differentiation, tissue culture plates coated with 10% gelatin were used. Media was composed of DMEM with 10% FBS, 1% penicillin-streptomycin, 100 nM dexamethasone, 1% insulin-transferrin-selenium, 50 µM L-ascorbic acid 2-phosphate, 50 µg/mL proline, and 20 ng/mL TGFβ3⁶¹. Differentiated chondrocytes were visualized with Alcian Blue staining which was quantified by OD595nm reading of hydrochloric acid extracted dye.

Protein purification and western blot

For validation of pCAG-Mest construct expression, cells were harvested 2 days post transfection in RIPA buffer with protease inhibitor. For ANXA1 and FPR2 protein expression, 12 dpa blastemas from 6 to 8 mice for each Mest genotypic cohort were harvested with RIPA buffer with protease inhibitor. 25–30 µg of total protein lysate was loaded onto 4–20% Tris-Glycine polyacrylamide gels and transferred onto 45 µm nitrocellulose membrane. Primary antibodies were anti-β-tubulin (1:1000, ThermoFisher 32-2600), anti-Mest (1:1000, Abcam ab151564), anti-Annexin-1 (1:1000, Invitrogen 71-3400), and anti-Fpr2 (1:1000, Thermo Scientific 720293) used with HRP conjugated secondary antibodies (Jackson ImmunoResearch). Uncropped, full blot images are available in Supplementary Fig. 11.

RNA purification and qPCR

For in vitro experiments, cells were collected from plates on the day noted and RNA was purified using the RNeasy mini prep kit (Qiagen), following the manufacturer’s protocol. For mouse experiments, 30–48 12 dpa blastemas from each genotype were dissected and pooled. Timed pregnant mice were used to collect e12.5 Mest-WT and Mest-pKO embryos. RNA extraction for blastema and embryo tissues was performed using TRIzol reagent (Invitrogen), following the manufacturer’s protocol; tissues were homogenized using Navy Eppendorf RNA lysis tubes with the Bullet Blender Storm24 (Next Advance) at 4 °C. cDNA was synthesized with SuperScript IV First Strand Synthesis kit (Invitrogen) primed with oligo-dT. Quantitative PCRs for Perilipin-1, PPARγ, Col2a1, Col10a1, Runx2, Mest(total), Mest202, Mest210, and Mest211 were performed using SsoAdvanced Universal SYBR green supermix (BioRad) on QuantStudio 5 Real-time PCR machine. ΔCt was calculated for each well using GAPDH as a housekeeping gene. In-plate technical triplicates allowed for calculation of average relative expression using the 2^(-ΔCt) method⁶². All qPCR experiments were performed in triplicate; qPCR primers used for all genes, including those previously published^{29,61,63,64,65}, are detailed in Supplementary Table 1. For miRNA expression, gene-specific cDNA was synthesized using U6 and miR-335 primers with the Taqman MicroRNA Reverse Transcription kit (ThermoFisher). qPCR was performed using U6 and miR-335 TaqMan assays (001973, 000546; ThermoFisher). Statistical analyses were performed by two-sample t-test, one-way ANOVA followed by post-hoc t-tests with Bonferroni correction, or two-way ANOVA followed by post-hoc t-tests with Tukey HSD correction.

Bisulfite sequencing

12–18 12 dpa blastemas were collected and pooled for each Mest genotype group and hindlimbs were dissected from two e12.5 Mest-WT and Mest-pKO embryos. Genomic DNA was purified with the DNeasy Blood and Tissue Kit (Qiagen). 500 ng of genomic DNA was used for bisulfite conversion with the EZ DNA Methylation Kit (Zymo Research), following the manufacturer’s protocol. Bisulfite converted DNA was used for PCR amplification for the four regions of interest: CTCF, ExonA, Exon1-5’, and Exon1-3’, using ZymoTaq DNA polymerase (Zymo Research). Primers used for amplification were designed with the Bisulfite Primer Design Tool (www.zymoresearch.com/pages/bisulfite-primer-seeker) and MethPrimer (www.urogene.org/methprimer/) (Supplementary Table 1). PCR amplicons were TA cloned into pGEM-T-Easy (Promega) to facilitate allele-specific DNA sequencing from the Sp6 promoter (Eton Biosciences). A minimum of 10 clones were sequenced per genotype for each genomic region. Each sequencing read was scored for number of cytosines present in the appropriate amplicon region and compared to the predicted, fully methylated bisulfite converted sequence. Thus, an observed:predicted cytosine ratio of 0 indicates hypomethylation while a ratio of 1 indicates hypermethylation. Statistical analysis was performed using the Chi-square discrete test for uniformity, using the number of predicted methylated cytosines as the denominator for the test. One-way ANOVA followed by post-hoc t-tests with Bonferroni correction was used for statistical analysis.

Single cell isolation, library construction and RNA sequencing

All single cell transcriptomics were performed using the 10X Chromium single cell gene expression platform (10X Genomics) at the BWH Center for Cellular Profiling. The 28 dpa single cell isolation, library construction, and sequencing were performed as previously published reported³. For Mest-WT and Mest-KO scRNAseq, 36 12 dpa blastemas were dissected and pooled for each genotype. To minimize any potential batch effect, these were isolated and underwent library construction on the same day. Single-cell suspensions were generated for each genotype as previously described³, with the addition of FACS for live cells using propidium iodide. In short, tissue was digested with 2.5% trypsin-EDTA and 10% collagenase, followed by mechanical dissociation by pipetting. Cells were strained through 35 µm filters and red blood cells were removed with ACK lysing buffer (Gibco). Cells were resuspended in 0.4% BSA/PBS and 1 µg/µL propidium iodide. 10,000 propidium iodide negative cells were collected into 0.4% BSA/PBS and immediately used for 10x Genomics droplet encapsulation. cDNA libraries were made using the Single Cell 3’ v3 chemistry (10x Genomics). All libraries were sequenced at the DFCI Molecular Biology Core Facility by Illumina NextSeq 500.

For neutrophil flow cytometry, Mest-WT and KO regenerating digit tips were harvested at 9, 12, and 15 dpa. Single-cell suspensions were generated as described above with an additional incubation with anti-CD11b-BV421 (1:200, Biolegend #101251) and anti-Ly6g-AF647 (1:200, Biolegend #127609) prior to the addition of propidium iodide. Samples were sorted on a BD FACSAria, gating to remove doublets based on FSC-A and FSC-H, isolating PI⁻/CD11b⁺/Ly6g⁺ cells. Analysis was performed using FloJo version 10.8.1 (BD Biosciences). Percentages of CD11b⁺ /Ly6g⁺ double positive cells and CD11b⁺ /Ly6g⁻ of all PI⁻ cells were calculated for each genotype and time point. Cell numbers for PI⁻ and CD11b⁺/Ly6g⁺ and CD11b⁺ /Ly6g⁻ cells were used for differential proportion analysis between Mest-WT and KO at each time point.

scRNAseq computational analyses

High-performance computing was done with the O2 cluster supported by the Harvard Medical School Research Computing Group (it.hms.harvard.edu/service/high-performance-computing-hpc). FASTQ files for all scRNAseq experiments were aligned to the mouse mm10-2020-A transcriptome with introns using CellRanger version 7.1.0 (10x Genomics). Cell by gene matrices for each dataset were individually imported into R version 4.3.1⁶⁶ for analysis using Seurat version 5.0.1⁶⁷. Cells were filtered based on percent mitochondrial genes, UMI count, and doublet prediction as previously reported³. Post-filtering, the UA, 11, 12, 14, 17, and 28 dpa datasets contained 11499, 6918, 3039, 5414, 7947, and 8591 singlets, respectively, which we utilized for further analyses. The Mest-WT and Mest-KO 12 dpa datasets contained 8805 and 6898 singlets, respectively, used for analyses. All individual datasets were processed via normalization, identification of variable features, scaling, linear dimensional reduction, and clustering as previously described³.

Integration of the UA, 11, 12, 14, 17, and 28 dpa data was performed using the CCA Integration method in Seurat followed by re-normalization, scaling, and clustering. Fibroblast clusters were identified and subsetted by high expression of marker genes Prx1, Pdgfrα, and Lum. These subsetted fibroblasts were used for lineage trajectory, using the web interface (https://kleintools.hms.harvard.edu/tools/spring.html) for SPRING¹⁵ analysis. Integration of Mest-WT and Mest-KO datasets was performed using the CCA Integration method in Seurat followed by re-normalization, scaling, and clustering. Fibroblast clusters were identified and subsetted based on high expression of Pdgfrα and Lum. Immune cell clusters were identified and subsetted based on expression of Lyz2 (macrophages/monocytes), Cd3g (T-cells), Ctsk (pre-osteoclasts), and S100a9 (neutrophils). Differentially expressed genes (DEGs) for fibroblast subpopulations and between Mest genotypes were determined using Seurat FindAllMarkers and FindMarkers, respectively, with the following parameters: min.pct = 0.25 and avg.log2FC > 1.00 (for fibroblast cluster DEGs) or avg.log2FC > 0.58 (for Mest-WT and KO DEGs) and adj.pval < 0.05 (Supplementary Tables 2–4). ClusterProfiler version 4.10.0 was used to perform gene set enrichment analysis of DEGs with KEGG pathway⁶⁸. To determine statistically significant changes in populations between Mest-WT and Mest-KO cell types, fibroblast and immune cell clusters, we utilized differential proportion analysis as previously reported^3,69. Signaling changes were assessed using our defined cell types with CellChat version 2.1.1⁴⁰ and NICHES version 1.0.0⁴¹, utilizing the mouse CellChatDB and omnipath ligand-receptor databases, respectively.