Defining human MC phenotypes in nasal polyposis. To gain deeper insights into the developmental signals directing MC differentiation and heterogeneity in severe T2 inflammation, MCs were flow sorted from the nasal polyps of 4 patients with aspirin exacerbated respiratory disease (AERD), a disease associated with refractory eosinophilic nasal polyposis and asthma. MCs were sorted based on canonical surface markers (CD45+, CD117+, FcɛR1α+, lineage [CD11b–, CD11c–, CD3–, CD19–]) for single-cell RNA-Seq (scRNA-Seq) using the 10x Genomics, version 3, chromium platform (Supplemental Figure 1A; supplemental material available online with this article; https://doi.org/10.1172/JCI174981DS1), taking advantage of the increased transcript detection and recovery available with the platform chemistry relative to earlier platforms. MCs were identified through protease transcript expression (TPSAB1, CPA3), computationally separated from contaminating populations and a small donor-restricted cluster enriched for interferon signature genes (Supplemental Figure 1, B and C), and integrated using the Harmony software package, version 1.2 (12). We identified 6 MC clusters, including 2 MCTC and 2 MCT clusters designated based on expression of subset-associated transcripts (CTSG and CMA1 for MCTC,IL17RB and GPR183 for MCT), a transitional cluster, and proliferating MCs (Figure 1, A and B, and Supplemental Figure 2) (3). All clusters were detected in all donors; however, the proliferating cluster exhibited high variability (Supplemental Figure 2, A and B). Ribosomal content was elevated in MCT1, while mitochondrial content was highest in transitional MCs (Supplemental Figure 2C).

Phenotypic heterogeneity of MCs is marked by expression of distinct transcriptional cassettes. (A) Uniform manifold approximation and projection (UMAP) depiction of 6 MC clusters identified through scRNA-Seq analysis of MCs sorted from 4 AERD patients (top), with expression patterns for select subset-associated transcripts (bottom). (B) Row-normalized heatmap of common genes expressed by MCTC and MCT clusters (FDR < 0.05, log2FoldChange > 0.5, DESeq2) (C) Row-normalized heatmap of top differentially expressed genes across clusters (FDR < 0.05, log2FoldChange > 0.5, DESeq2), with representative cluster-enriched genes highlighted. (D–G) Enrichment of biological processes in (D) MCTC1, (E) MCTC2, and (F) MCT1, and (G) MCT2 clusters with row-normalized heatmaps showing expression of select process-associated genes. Heatmap columns indicate average cluster expression for each of n = 4 individuals; scale bars denote z score.

Differential expression analysis indicated common core sets of transcripts distinguishing MCTC (CMA1, CTSG, FCER1A, GPR65, C3AR1) from MCT (CPA3, TPSAB1, GPR183, CD38, IL17RB) (Figure 1B). However, each MCTC and MCT subcluster was also associated with distinct transcriptional cassettes (Figure 1C and Supplemental Table 1). These transcripts included genes encoding granule factors, the γ subunit of FcεRI, and hematopoietic prostaglandin D2 (PGD2) synthase in MCTC1 (FCER1G, NDST2, HPGDS), inflammation-associated factors in MCTC2 (CSF2, IL13, PTGS2), ribosome and translation initiation factors in MCT1 (RPL10A, RPS27, EIF3E), a separate set of inflammation-associated factors in MCT2 (PTGS1, LTC4S, IL5), semaphorins and DNA-binding proteins in transitional cells (SEMA4A, SEMA7A, POLG2), and cell cycle–associated genes in the proliferating cluster (Figure 1C).

To gain further insight into the biological processes associated with each cluster, we conducted gene ontology (GO) pathway enrichment analysis (Supplemental Table 2). MCTC1 exhibited enrichment for electron transport chain and protein processing (Figure 1D), while MCTC2 was enriched in inflammation and adhesion-associated transcripts (Figure 1E). Transitional MCs were enriched for chromatin and histone modification and chromatin remodeling pathways, suggesting ongoing epigenetic reprogramming (Supplemental Figure 3), consistent with prior characterization of this population as immature MCs undergoing polarization (3).

MCT1 exhibited limited enrichment for biological processes, suggesting MCT-specific pathways are not well represented within the GO database. Among the pathways identified, MCT1 was enriched for translation and ribosome biogenesis (Figure 1F), whereas MCT2 showed enrichment for cell activation, the MHC-I pathway, and SMAD phosphorylation (Figure 1G). Several TGF-β target transcripts previously identified in murine inflammation-expanded airway MCs were significantly elevated in the transitional and both MCT clusters (SKIL, SMAD7, and LDLRAD4), while ITGAE was restricted to the MCT clusters (Supplemental Figure 2E) (11), suggesting a broader role for TGF-β signaling within the intraepithelial MC compartment.

Intraepithelial MCs reside in a TGF-β–rich tissue niche. MCT enrichment for the SMAD phosphorylation pathway and murine MC TGF-β target genes was of particular interest based on our prior findings linking TGF-β to lower airway intraepithelial MC differentiation in vivo (11). To assess potential sources of TGF-β signaling in nasal polyps, we used a prior transcriptional atlas of sinonasal tissues, identifying basal epithelial cells (EpCs) as a predominant source for transcript encoding TGF-β2, while TGF-β1 was more broadly expressed (Supplemental Figure 4A) (4, 13). Reanalysis of a prior bulk RNA-Seq dataset of flow-sorted basal EpCs (13) indicated upregulation of both transcripts in subjects with CRSwNP (inclusive of AERD) relative to CRS without nasal polyps (CRSsNP), a milder clinical phenotype lacking an expanded MCT population (Supplemental Figure 4B and Supplemental Table 3). Notably, although TGF-β2 binds the TGF-β receptor with lower affinity than TGF-β1, both isoforms direct human MC chemotaxis and inhibit their proliferation with similar potency (14, 15).

Transcript encoding integrin β6 (ITGB6), which pairs with integrin αV to mediate latent TGF-β activation and regulate murine mucosal MC development (16, 17), showed a trend toward upregulation in CRSwNP EpC compared with CRSsNP (Supplemental Figure 4B). Flow cytometric evaluation identified a significant increase of cell-surface αVβ6 on EpCs from both CRSwNP and AERD polyps relative to CRSsNP, suggesting increased availability of activated TGF-β within or adjacent to the epithelium in nasal polyps (Supplemental Figure 4C). Analysis of lung EpCs from a recent scRNA-Seq study similarly indicated enrichment of TGFB1 and TGFB2 in basal EpCs, which significantly upregulated both ITGB6 and ITGAV in asthmatic patients but not allergic nonasthmatic patients following allergen challenge (Supplemental Figure 4D and Supplemental Table 4) (18).

To determine the functional consequences of EpC αVβ6 expression, we conducted histologic assessment of SMAD2/3 phosphorylation within the epithelium, subepithelium, parenchyma, and glandular regions of tissue sections from CRSsNP and CRSwNP donors, normalized by cell density (Supplemental Figure 5). In CRSwNP, we observed significantly increased pSMAD2/3 in the epithelium relative to the parenchyma, with similar pSMAD2/3 levels in the epithelium and subepithelium (Padj < 0.05). CRSsNP samples showed a trend toward elevated pSMAD2/3 in the epithelium and glandular regions of the tissue but no significant differences across compartments, with low phosphorylation in the subepithelium and parenchyma. These observations suggest that EpC αVβ6 expression contributes to TGF-β signaling in proximity to the epithelium, indicating intraepithelial MCs reside within a TGF-β–rich tissue niche. Thus, together with our prior observations in mice, we hypothesized TGF-β signaling could direct the human intraepithelial MCT phenotype.

TGF-β directs an MCT-like transcriptional phenotype. To test the relationship between TGF-β signaling and the MCT transcriptional phenotype, PB-MCs were treated for 24 hours or 6 days with TGF-β1 or, in separate experiments, MCs were differentiated from CD34+ cells in the presence or absence of TGF-β1 for the full duration of culture (7 weeks) and assessed via bulk RNA-Seq. TGF-β1 treatment exerted discrete, time-dependent effects on the MC transcriptome, differentially regulating 414 genes at 24 hours, 1,013 genes at 6 days, and 1,903 genes at 7 weeks (Figure 2, A and B, and Supplemental Table 5). A core set of 149 genes were significantly altered by TGF-β1 across all time points, including upregulation of IL4R, SIGLEC6, and LTC4S and downregulation of CMA1, CTSG, and MRGPRX2, with approximately half of the transcripts changing to a similar degree and the other half showing a gradient of upregulation or downregulation with time (Figure 2C). Transcripts enriched in the MCT1 and MCT2 clusters in vivo were predominantly upregulated in PB-MCs treated with or differentiated in TGF-β1 (Figure 2D).

TGF-β signaling during MC development elicits an MCT-like transcriptional phenotype. (A) Venn diagram showing common versus timepoint-specific differentially expressed in PB-MC treated with TGF-β1 for 24 hours (green) and 6 days (blue) or PB-MCs differentiated in TGF-β1 for 7 weeks (pink) (FDR < 0.05, DESeq2). (B and C) Heatmap of (B) all differentially expressed genes and (C) transcripts showing gradients of upregulation (left) or downregulation (right) following TGF-β1 treatment (FDR < 0.05, DESeq2); color indicates log2fold change versus untreated samples. (D) Heatmap of differentially expressed genes associated with MCT1 (left) and MCT2 (right) clusters. (E) Violin plots showing per-cell expression as a percentage of all transcripts for timepoint-specific TGF-β1 target genes in PB-MCs treated with TGF-β1 for 24 hours, 6 days, or differentiated in TGF-β1 across AERD MC clusters (Padj matrix shown in Supplemental Figure 6A). (F) GSEA for 7-week, 6-day, and 24-hour TGF-β in vitro signatures across AERD MC clusters, showing normalized enrichment score and adjusted P values. Gray color indicates statistically insignificant positive or negative enrichment. (G and H) Heatmap of (G) selected differentially expressed genes and (H) transcription factors both restricted to PB-MCs differentiated in TGF-β1 across time points (left) and differentially expressed across AERD MC clusters (right). Heatmaps show log2FoldChange versus untreated for each donor (bulk) or donor-averaged expression values (scRNA-Seq); scale bars show log2FoldChange versus untreated (bulk) or z score (scRNA-Seq).

We next constructed TGF-β–upregulated gene signatures unique to each time point to probe our in vivo dataset (Supplemental Table 5). The 24-hour and 6-day gene signatures showed low-level expression across MC clusters, with lowest expression in the MCTC1 and proliferating clusters but only modest differences between the remaining MC clusters (Figure 2E, P value matrix in Supplemental Figure 6A). The 7-week developmental TGF-β1 signature had substantially higher expression in both MCT clusters and the transitional cluster with significant elevation relative to both MCTC clusters (Figure 2E). The core TGF-β1 target genes included MC-restricted proteases and cell-surface receptors (Figure 2C), indicating a cell type–specific effect on human MCs and potentially explaining the limited detection of TGF-β signaling within the MCT clusters through pathway enrichment (Figure 1, F and G). When we instead used our MC-derived gene signatures for gene set enrichment analysis (GSEA), we identified significant elevation of the 7-week signature in both MCT clusters and significant decreases in the MCTC clusters (Figure 2F and Supplemental Table 2). The 6-day signature was significantly enriched in the transitional and MCTC1 clusters and decreased in MCTC2, while only the transitional cluster was enriched for the 24-hour signature (Padj < 0.05).

Notably, a subset of the genes differentially expressed between MCT and MCTC clusters in vivo were only significantly altered in PB-MCs differentiated in TGF-β1, including several MCT-enriched transcription factors (HES1, AEBP1, IKZF3) (Figure 2, G and H). PB-MC differentiation in TGF-β further drove upregulation of transcripts encoding ribosomal components, mirroring MCT ribosomal enrichment in vivo (Supplemental Figure 6B). CRSwNP MCT and MCTC exhibited similar cell-surface expression of the TGF-β receptor R2 subunit (TGF-β R2), which pairs with the R1 subunit to recognize TGF-β1 and TGF-β2 (Supplemental Figure 6, C and D). As additional controls, we confirmed that flow sorting did not elicit degranulation or impact viability in PB-MCT or PB-MCTC (Supplemental Figure 6E). Together with the pSMAD2/3 patterns in CRSwNP, these observations strongly suggested that differences in developmentally associated exposure to TGF-β across tissue microenvironments direct MC polarization in vivo and potentially implicated shorter-term TGF-β signaling as a factor differentiating MCTC1 and MCTC2 clusters.

To explore whether the TGF-β developmental signature was restricted to MCTs in AERD or represented a broader feature of MCTs, we expanded our analysis to other diseases and tissues. We first assessed our prior sinonasal polyp MC dataset (3), finding that MCTs in both AERD and aspirin-tolerant CRSwNP were enriched for the TGF-β1 developmental signature relative to MCTCs (Supplemental Figure 6F). We next evaluated MCs from an scRNA-Seq study of ulcerative colitis that fractionated the intestinal epithelium from lamina propria (19). MCs from the lamina propria fraction were significantly enriched for core sinus MCTC transcripts (CMA1, GPR65, ICAM1) and expressed inflammation-associated sinus MCTC2 transcripts (PTGS2, CSF1) (FDR < 0.05). Expression of IL13 was only detected in the lamina propria, although this did not reach statistical significance (Supplemental Figure 6G). MCs from the epithelial fraction were significantly enriched for sinus MCT core transcripts (TPSAB1, CTSW, CD9) and for sinus MCT1-enriched transcripts encoding eicosanoid biosynthetic enzymes (PTGS1, LTC4S, ALOX5) (FDR < 0.05). Epithelial MCs further exhibited significant enrichment for the TGF-β developmental signature compared with lamina propria MCs (Supplemental Figure 6H and Supplemental Table 6). Collectively, these observations supported a role for TGF-β in directing the MCT transcriptome across mucosal tissues and suggested a link between TGF-β developmental signaling and MC effector function.

Defining the influence of TGF-β on the MC granule phenotype. As the MCTC-associated proteases CMA1 and CTSG (encoding chymase and CTSG, respectively) were strongly downregulated by TGF-β treatment at all 3 time points (Figure 2C and Supplemental Figure 7A), we next assessed granule-associated transcripts. Nasal polyp MCTs were enriched for CHSY1, encoding chondroitin sulfate synthase, genes encoding tryptases (TPSAB1, TPSB2, TPSD1, TPSG1), CPA3, encoding carboxypeptidase A3, and PRSS21, a serine protease not previously associated with MCs, suggesting a greater degree of protease heterogeneity in human MCs than previously appreciated (Figure 3A). MCTCs were enriched for NDST2, encoding a central enzyme required for synthesis of heparin sulfate, a major proteoglycan component of MCTC granules (20, 21), CMA1, and several cathepsin-encoding transcripts (CTSC, CTSD, CTSG). TGF-β drove similar expression patterns to those observed in vivo, downregulating CMA1, CTSG, and NDST2 while upregulating transcripts encoding CPA3 and α, β, and δ tryptases (TPSAB1, TPSB2, TPSD1) (Figure 3A and Supplemental Figure 7A). Thus, TGF-β profoundly altered expression of transcripts regulating MC granule composition and protease content.

TGF-β signaling directs the MCT protease phenotype during early development. (A) Differentially expressed genes encoding granule components in vivo (left), with violin plots for select proteases (center), and in vitro gene expression for PB-MC stimulated with TGF-β1 for 24 hours, 6 days, or differentiated in TGF-β1 (right). Columns indicate donor-averaged cluster expression (scRNA-Seq) or log2FoldChange versus untreated for each donor (bulk); scale bars denote z score (scRNA-Seq) or log2FoldChange (bulk), FDR < 0.05, log2FoldChange > 0.5 for scRNA-Seq and FDR < 0.05 for bulk (DESeq2). (B) Chymase expression and quantification in PB-MCs treated with (red) or without (blue) TGF-β1 for 6 days versus isotype control (gray). n = 6 individual donors (t test). (C–E) Expression and quantification of (C) chymase, (D) CTSG, and (E) CPA3 in PB-MCs differentiated with (purple) or without (blue) TGF-β1 versus isotype control (gray). n = 7–8 individual donors. **P < 0.01; ***P < 0.001; ****P < 0.0001 (t test). (F) One-week CPA3 and tryptase β2 release in PB-MCT versus PB-MCTC supernatants. n = 6 and 5 individual donors, respectively. *P < 0.05; **P < 0.01 (Mann-Whitney). (G) Chymase expression and quantification for PB-MCs differentiated in TGF-β1 and subsequently cultured with (purple) or without (orange) TGF-β1 for 2 weeks. n = 8 individual donors. ***P < 0.001 (t test). (H) Chymase expression and quantification for PB-MCT cocultured with EpCs for 2 weeks supplemented with SCF (100 ng/mL), IL-6 (50 ng/mL), and the indicated concentration of LY2109761. n = 8. *Padj < 0.05, **Padj < 0.01 (ANOVA) (I) Gating strategy to isolate nasal polyp MCTs and MCTCs. (J) Chymase expression and quantification for primary nasal polyp MCTs (top) or MCTCs (bottom) maintained in culture media with (red) or without (blue) TGF-β1 for 2 weeks. n = 6–7 for nasal polyp MCs. **P < 0.01 (t test). Box-and-whisker plots show median, interquartile range, and minimum/maximum values observed.

PB-MC differentiated in stem cell factor (SCF) and IL-6 alone contained high levels of intracellular chymase, indicative of an MCTC-like protease phenotype content (PB-MCTC) (Figure 3B). Intracellular chymase content was unaffected by a 6-day TGF-β1 treatment despite the striking change in transcript expression, indicating a potential “uncoupling” of the MC transcriptional and protease phenotype. We hypothesized that chymase stability was due to slow turnover of MC granule–associated proteins. Supporting this, PB-MC differentiation in TGF-β1 significantly reduced intracellular chymase content (Figure 3C), directing an MCT-like protease phenotype (PB-MCT). PB-MC differentiated in TGF-β1 further exhibited significant reductions in intracellular CTSG (Figure 3D), a protease robustly expressed by MCTC in vivo but absent from MCT (Supplemental Figure 8). While CPA3 was transcriptionally upregulated by TGF-β1 (Figure 3A), intracellular CPA3 protein was lower in PB-MCTs relative to PB-MCTCs (Figure 3E). Instead, we observed significant elevations in extracellular CPA3 in PB-MCT culture supernatants relative to PB-MCTC (Figure 3F). Tryptase β2, encoded by TPSB2, was similarly elevated in PB-MCT supernatants (Figure 3F). In stark contrast to the stability of the PB-MCTC protease phenotype, intracellular chymase content of PB-MCT significantly increased following TGF-β1 removal (Figure 3G). Coculture with EpCs maintained the PB-MCT chymase-low phenotype in the absence of exogenous TGF-β, while treatment of cocultures with the TGF-β receptor kinase inhibitor LY2109761 increased intracellular chymase in a dose-dependent manner (Figure 3H), further implicating EpC-derived TGF-β as responsible for maintaining the MCT protease phenotype in vivo.

Following our observations with in vitro–differentiated PB-MCT and PB-MCTC, primary MCTs and MCTCs were flow-sorted from nasal polyps using previously defined surface markers (Figure 3I) and cultured for 2 weeks supported by SCF either alone or supplemented with TGF-β1 (3). Consistent with in vitro–differentiated MCs, primary tissue MCTs maintained a chymase-low phenotype in TGF-β–supplemented media, while culture in SCF alone led to increased chymase content (Figure 3J). In contrast, the protease phenotype of primary MCTCs was highly stable, with TGF-β supplementation having no significant impact on chymase content (Figure 3J). TGF-β signaling during MC differentiation was required to establish the MCT protease phenotype, while interruptions in TGF-β1 signaling increased intracellular chymase levels in both PB-MCTs and polyp MCTs. Thus, our in vitro and ex vivo findings indicate a 1-way MC granular plasticity, with MCTCs having a stable granule phenotype but MCTs increasing intracellular chymase content if TGF-β signaling is interrupted. These observations further suggest that chymase expression is a “default” pathway for MCs and must be actively suppressed to maintain the MCT phenotype.

TGF-β regulation of MC surface receptor expression. MCs express a broad repertoire of cell surface receptors responsible for fine tuning their responses to microenvironmental cues. Across sinonasal MCs, we observed differential expression of cytokine and growth factor receptors (KIT, CSF2RB, IL17RB, NTRK1) and receptors associated with MC activation or inhibition (FCER1A, C3AR1, CD33, SIGLEC6) (Supplemental Figure 9A). TGF-β1 treatment in vitro influenced a subset of these receptors, upregulating transcripts encoding receptors for IL-3 (IL3RA) and IL-9 (IL9R), which can regulate MC growth and proliferation, and NTRK1, encoding a receptor for nerve growth factor (Figure 4A and Supplemental Figure 7B), a growth factor elevated in sinus mucosa and asthmatic bronchial epithelium (22–24). TGF-β1 upregulated expression of the α subunit of the receptor for IL-4 (IL4R) at all time points, a signal previously found to drive ex vivo proliferation of human intestinal MCT (25). Further, TGF-β1 upregulated inhibitory receptors (CD22, CD33, SIGLEC6, FCGR2B) while downregulating the transcripts encoding IL-1 family cytokine receptors (IL18R1, IL1RL1, IL1RN) and receptors linked with IgE-independent MC degranulation, including the platelet-activating factor receptor (PTAFR), the complement component C3a receptor (C3AR1), and MRGPRX2, which mediates MC degranulation in response to neuron-derived substance P and during pseudoallergic drug reactions (Figure 4B) (26–28). Many of these receptors were differentially expressed between polyp MCT and MCTC clusters in vivo, including enrichment of SIGLEC6 and NTRK1 in MCTs and IL18R1, IL1RL1, PTAFR, and C3AR1 in MCTCs (Supplemental Figure 9A).

TGF-β selectively regulates MC expression of activating receptors. (A–C) Heatmaps of select differentially expressed transcripts encoding (A) cytokine, chemokine, and growth factor receptors, (B) activating and inhibitory receptors, and (C) FcɛR1 signaling pathway components following PB-MC stimulation with TGF-β1 for 24 hours or 6 days or differentiation of MCs in TGF-β1 for 7 weeks. Heatmaps show log2FoldChange expression of genes versus unstimulated cells at each time point. Columns indicate individual donors, FDR < 0.05 (DESeq2). (D) Representative flow plot and quantification of FcɛR1α expression by PB-MCTCs treated with (red) or without (blue) TGF-β1 for 6 days. n = 12 (t test). (E) Expression and quantification of FcɛR1α in PB-MCTs (purple) versus PB-MCTCs (blue). n = 8. **P < 0.01 (t test). (F) Degranulation of PB-MCTCs treated with or without TGF-β1 for 6 days (left) or PB-MCTs (right) at 1 hour after activation with anti-IgE. n = 5 individual donors (ANOVA). (G) Intracellular chymase content of PB-MCTC (light blue) and PB-MCT (purple) at baseline and following degranulation with anti-IgE. n = 7 individual donors. **Padj < 0.01; ***Padj < 0.001 (ANOVA). (H) MRGPRX2 expression and quantification for PB-MCTCs following 6-day treatment with (red gradient) or without (blue) TGF-β1 and PB-MCTs (purple). n = 8 individual donors. **Padj < 0.01; ***Padj < 0.001; ****Padj < 0.0001 (ANOVA). (I) Degranulation of PB-MCTCs treated with (red) or without (blue) TGF-β1 for 6 days, and PB-MCTs (purple) following 1 hour stimulation with MRGPRX2 ligands compound 48/80 (left) and substance P (right). n = 5–6 individual donors. **Padj < 0.01; ***Padj < 0.001; ****Padj < 0.0001 (ANOVA). (J) Representative flow plot and quantification of MRGPRX2 expression by PB-MCTs maintained in culture media with (purple) or without (orange) TGF-β1 for 2 weeks. n = 8 individual donors. **P < 0.01 (t test). Box-and-whisker plots show median, interquartile range, and minimum/maximum values observed.

TGF-β1 treatment impacted several transcripts associated with the high-affinity IgE receptor signaling pathway, including upregulating transcripts encoding the FcεR1 α chain (FCER1A) while downregulating transcripts encoding the β chain (MS4A2). TGF-β1 further influenced key downstream signaling pathway components, downregulating SYK and PLCG1 while upregulating LAT, LYN, and PLCG2 (Figure 4C and Supplemental Figure 7B). Despite prior reports of TGF-β1 downregulating murine MC FcεR1α expression and inhibiting IgE-mediated activation (29, 30), TGF-β1 treatment did not alter FcεR1α surface expression on PB-MCTC (Figure 4D and Supplemental Figure 9B), whereas FcεR1α was moderately elevated on PB-MCTs relative to PB-MCTC (Figure 4E). No differences were observed in IgE/anti-IgE-driven degranulation between either PB-MCT and PB-MCTC or PB-MCTCs treated with or without TGF-β1 (Figure 4F and Supplemental Figure 9C). However, intracellular staining for chymase indicated a significantly larger MFI reduction in PB-MCTCs (2938 ± 844.7) compared with PB-MCT (719.3 ± 182.5) (Figure 4G). Thus, although degranulation levels were similar between the 2 subsets, due to differences in granule composition, the mediators released during the degranulation process likely differ considerably.

As MRGPRX2 is a well-established driver of MCTC degranulation (31), we evaluated MRGPRX2 surface expression and response to its ligands. TGF-β1 treatment moderately decreased PB-MCTC surface expression of MRGPRX2 at 24 hours, while substantial dose-dependent downregulation was observed at 6 days (Supplemental Figure 9D and Figure 4H). PB-MCTs displayed minimal MRGPRX2 expression (Figure 4H). Consequently, PB-MCTC degranulation in response to activation with the MRGPRX2 ligands compound 48/80 and substance P was significantly reduced in a dose-dependent manner following 6-day stimulation with TGF-β1, while degranulation was abolished in PB-MCTs (Figure 4I). As with intracellular chymase content, removal of TGF-β1 from PB-MCT culture media for 2 weeks increased MRGPRX2 expression (Figure 4J). Collectively, these findings establish that MRGPRX2 protein expression and activity are dynamically regulated by TGF-β, which may have a similar effect on other IgE-independent MC activating receptors. Further, in contrast to the MCTC granule phenotype, these observations indicate that the MCTC cell-surface phenotype exhibits substantial plasticity.

TGF-β facilitates a distinctive MCT effector phenotype. MC effector functions are mediated by both release of preformed mediators and de novo production of protein and lipid mediators. Extending our prior observations, we noted differential expression of many cytokine and chemokine transcripts across polyp MC subsets in vivo. MCTCs showed elevated expression of chemokines for monocytes and/or T cells (CCL2, CCL4, CCL23, CXCL16), the cytokine IL13, and the monocyte/macrophage lineage-associated growth factors CSF1 and CSF2. MCTC2 additionally expressed neutrophil chemokines (CXCL2, CXCL3), several cytokines (IL3, LIF), and the growth factor VEGFA (Figure 5A and Supplemental Figure 10). All MCTs were transcriptionally enriched for a separate set of cytokines (IL18, MIF, TGFB1, TNFSF10) (Figure 5A), while MCT2 were further enriched for IL5 and CCL1. Thus, MC subsets within human nasal polyps likely carry out distinct microenvironmental-associated effector functions, potentially including regional recruitment and retention of other leukocyte subsets.

TGF-β selectively reshapes MC proinflammatory cytokine, chemokine, and growth factor production following IgE crosslinking. (A) Row-normalized heatmap of differentially expressed genes associated with cytokine, chemokine, and growth factors across nasal polyp MC clusters. Columns show averaged expression by donor; scale bar denotes z score. FDR < 0.05, log2FoldChange > 0.5 (DESeq2). (B) Heatmap showing differentially expressed transcripts in TGF-β1–stimulated cells. Columns show individual donors; scale bars indicate log2FoldChange versus unstimulated controls. FDR < 0.05 (DESeq2) (C and D) Row-normalized average data of n = 6 individual donors showing protein secretion of cytokines, chemokines, and growth factors at 6 hours following anti-IgE activation by (C) PB-MCTCs versus PB-MCTs and (D) PB-MCTCs cultured with or without TGF-β1 for 6 days. (E and F) Row-normalized average data of n = 6 individual donors showing release of cytokines, chemokines, and growth factors at 6 hours after IL-33 stimulus by (E) PB-MCTCs versus PB-MCTs and (F) PB-MCTCs cultured with or without TGF-β1 for 6 days. Scale bars denote z score. *Padj < 0.05 between activated PB-MCT and PB-MCTC (ANOVA).

Following our initial characterization of TGF-β1 regulation of the MC granule and surface phenotypes, we hypothesized it could further direct the discrete cytokine, chemokine, and growth factor profiles enriched in nasal polyp MCTs in vivo. PB-MCT downregulated a subset of MCTC-associated transcripts (LIF, CXCL16, CCL2, CSF1) and upregulated MCT-associated ones (MIF, CCL1) (Figure 5B and Supplemental Figure 7C). PB-MCTC treatment with TGF-β1 had minimal impact on inflammatory mediators. Transcripts encoding cytokines such as IL-5 and IL-13 are typically upregulated in response to MC activation (32). Thus, we examined whether TGF-β1 modified activation-induced MC mediator production by activating PB-MCT and PB-MCTC with IL-33 or IgE crosslinking. PB-MCTCs treated with TGF-β1 for 6 days were evaluated in parallel. Minimal baseline secretion was observed for any mediators measured, despite constitutive transcript expression for several (CCL2, CSF1, VEGFA, IL10) (Supplemental Figures 11 and 12). Following IgE cross-linking, PB-MCTs secreted significantly elevated levels of IL-5, IL-10, CCL4, and VEGF, while PB-MCTCs preferentially secreted IL-13 and CSF1 (M-CSF) (Padj < 0.05). No significant differences were observed in TNF-α, PDGF-A, CCL2, CSF2 (GM-CSF), or CXCL8 (IL-8) secretion following IgE crosslinking. PB-MCTC treatment with TGF-β for 6 days increased CCL4 secretion while reducing CSF1 and PDGF-A, but no significant differences were observed for IL-5, IL-10, or IL-13 (Figure 5, C and D, and Supplemental Figure 11B).

In response to IL-33 treatment, PB-MCTs again produced significantly more IL-5, while PB-MCTCs preferentially produced IL-13 (Figure 5E and Supplemental Figure 12A). In contrast to IgE crosslinking, following IL-33 activation, PB-MCTs produced significantly less IL-10, IL-8, and CCL2 (Figure 5E and Supplemental Figure 12A). No significant differences were observed between PB-MCTs and PB-MCTCs for CCL4, VEGF, CSF1, PDGF-A, or TNF-α following IL-33 stimulus (Figure 5E and Supplemental Figure 12A). Six-day treatment of PB-MCTC with TGF-β1 significantly increased IL-5 and CXCL8 secretion, while inhibiting production of IL-10 and CCL2, but again had no impact on IL-13 production (Padj < 0.05) (Figure 5F and Supplemental Figure 12B). Notably, the PB-MCT mediator secretion pattern following IL-33 activation (enhanced IL-5 with reduced IL-13 and CCL2) mirrored the differences in transcripts encoding these T2-associated factors in polyp MCTs versus MCTCs (Figure 5A), while the secretion pattern following the 6-day stimulation (increased IL-5 but no difference in IL-13 or CSF2) was suggestive of the differential expression patterns observed between MCTC2 and MCTC1. Moreover, IL-33 was far more potent than IgE crosslinking for driving cytokine production, while IgE crosslinking drove increased chemokine secretion (Supplemental Figures 11 and 12). Overall, this suggested a major role for TGF-β in shaping MC cytokine and chemokine production following activation by diverse stimuli in vivo.

Enhanced eicosanoid production by MCT. As activated MCs are well-characterized sources of arachidonic acid metabolites, we evaluated the impact of TGF-β on eicosanoid biosynthetic enzyme expression. Notably, both at 6 days and 7 weeks, TGF-β1 upregulated the expression of transcripts encoding 5-lipoxygenase (ALOX5), 5-lipoxygenase activating protein (ALOX5AP), and leukotriene C4 synthase (LTC4S), proteins required for LTC4 synthesis (Figure 6A and Supplemental Figure 7D). PTGS1, encoding the enzyme cyclooxygenase (COX) 1, a key component of PGD2 production, was upregulated at both 6 days and 7 weeks (Figure 6A and Supplemental Figure 7D), while a trend toward PTGS2 upregulation (encoding COX-2) was also observed at 7 weeks (FDR < 0.1). Thus, we hypothesized that TGF-β treatment would enhance MC eicosanoid production.

TGF-β enhances MC lipid mediator production. (A) Heatmap showing genes associated with eicosanoid biosynthesis differentially regulated by TGF-β1. n = 3 individual donors. FDR < 0.05, log2FoldChange > 0.5 (DESeq2). Scale bar indicates log2FoldChange versus unstimulated control samples. (B) CysLTs (left) and PGD2 (right) production by PB-MCTCs versus PB-MCTs and activated with anti-IgE for 1 hour. n = 8–9 individual donors. *Padj < 0.05; **Padj < 0.01; ****Padj < 0.0001 (ANOVA). (C) CysLTs (left) and PGD2 (right) production by PB-MCTCs treated with or without TGF-β1 for 6 days and activated with anti-IgE for 1 hour. n = 15 individual donors. * Padj < 0.05; *** Padj < 0.001; ****Padj < 0.0001 (ANOVA). (D and E) Effects of selective inhibitors for COX-1 (SC560) and COX-2 (SC236) on PGD2 production by (D) PB-MCTCs versus PB-MCTs or (E) PB-MCTCs treated with TGF-β1 for 6 days prior to activation with anti-IgE. n = 6 individual donors. * Padj < 0.05; ** Padj < 0.01; *** Padj < 0.001 (ANOVA). (F) Violin plots of differentially expressed genes associated with CysLTs and PGD2 biosynthesis across polyp MC clusters, FDR< 0.05, log2FoldChange > 0.5 (DESeq2). (G) Gating strategy to isolate nasal polyp MCT and MCTCs (left). (H) Eicosanoid production following activation with anti-IgE for 1 hour (right). n = 9 individual donors. * Padj < 0.05; **Padj < 0.01; *** Padj < 0.001; ****Padj < 0.0001 (ANOVA). Box-and-whisker plots show median, interquartile range, and minimum/maximum values observed.

Consistent with their increased expressions of ALOX5, ALOX5AP, LTC4S, and COX enzymes, PB-MCTs produced significantly more cysteinyl leukotrienes (CysLTs) and PGD2 than PB-MCTCs following IgE crosslinking (Figure 6B), while 6-day treatment of PB-MCTCs with TGF-β1 also increased secretion of both eicosanoids (Figure 6C). As COX-1 and COX-2 can both convert arachidonic acid into PGH2, the precursor of PGD2, we assessed the role of each enzyme in driving the increased PGD2 production observed in vitro. Treatment of PB-MCTCs with the COX-1–specific inhibitor SC560 significantly reduced FcεR1-induced PGD2 synthesis in both groups (Figure 6, D and E), indicating a central role for this enzyme in driving PGD2 production, while the COX-2-specific inhibitor SC236 marginally decreased PGD2 production only in TGF-β1–treated cells (Figure 6, D and E). Together, these findings suggest that MCTs are poised for rapid PGD2 synthesis primarily due to TGF-β1–driven COX-1 upregulation.

Mirroring the effects of TGF-β in vitro, both in vivo MCT clusters were enriched for LTC4S, ALOX5AP, and PTGS1, with MCT2 further enriched for ALOX5 (Figure 6F). MCTC1 was enriched for HPGDS, encoding hematopoietic PGD2 synthase, while MCTC2 was further enriched for PTGS2. To test the functional consequences of these differential expression patterns, primary MCTCs and MCTs were flow sorted from nasal polyps (Supplemental Figure 6C and Figure 6G) and activated using anti-IgE. As with their in vitro counterparts, activated MCTs generated significantly higher quantities of CysLTs compared with MCTCs and trended toward generating more PGD2 than MCTCs on a per-cell basis (Figure 6H). Thus, our findings highlight MCTs as a dominant source of CysLT production both in vitro and in vivo, underscoring the discrete effector functions for MCTs and MCTCs, and identify TGF-β1 as the likely driver of the MCT effector phenotype.