Dermal adipogenesis is dynamically coupled with the initiation and resolution of neutrophilic skin inflammation in the imiquimod-induced psoriasis mouse model
Infiltration of neutrophils is considered the hallmark of the onset of psoriatic inflammation21. Immunostaining of human skin samples showed that myeloperoxidase-positive (MPO+) neutrophils were found in close proximity to vimentin-positive (VIM+) fibroblasts in the dermis of patients with plaque psoriasis (PV) (Fig. 1A). Furthermore, the dermis from patients with generalized pustular psoriasis (GPP) exhibited an excessive accumulation of neutrophils around fibroblasts (Fig. 1A). These findings suggest that aberrant activation of dermal fibroblasts may contribute to the activation and accumulation of neutrophils, potentially driving the pathogenesis of psoriasis.
To gain insight into the dermal mechanism of neutrophil activation, we utilized the imiquimod (IMQ)-induced psoriasis-like mouse model, wherein neutrophil infiltration plays a pivotal role in the pathogenesis of skin inflammation 13,22,23. As shown in Fig. 1B-C, daily application of IMQ led to the progression of skin erythema, thickening, and scaling continued steadily up to 6 days post-treatment (p.t.), whereas these clinical manifestations began to resolve from 6 ~ 10 days p.t., indicating the development of self-tolerance to IMQ over prolonged exposure.
Histological analysis (Fig. 1C-D, S1A-C) revealed a rapid reduction of the dWAT layer by day 2 p.t., followed by a pronounced re-expansion of dWAT, where neutrophils characterized by their distinctive multi-lobulated nuclei24 were detected at day 3 p.t.; re-growth of mature adipocytes occurred from day 6 to 10 p.t., coinciding with inflammation resolution. Analysis of neutrophils, marked by high level of Ly6G and CD11B expression25,26 (Fig. S1D-F), confirmed that neutrophil infiltration peaked between 3 and 6 p.t.. Immunostaining to ascertain the spatial relationship between dWAT cells and neutrophils found that FABP4+ adipocytes were rapidly lost by 3 p.t., and concurrently Ly6G+ neutrophils and PDGFRA+ fibroblasts specifically co-populated the dWAT layer, with a peak in their presence observed at day 3 p.t. (Fig. 1F-G). Subsequently, from day 6 to 10 p.t., there was a repopulation of FABP4+ adipocytes within the dWAT, coinciding with a clearance of neutrophils from the skin (Fig. 1F-G).
Lipid staining analyses, conducted to evaluate changes of lipogenesis (Fig. 1H, Fig. S1G), indicated a reduction in lipid-laden adipocyte size on day 3, with a subsequent re-expansion by day 6 p.t.. Quantitative analysis of CAV1+PLIN1+ adipocytes revealed a transient decrease in the number of large, mature adipocytes (> 1000 µm2) on day 3 p.t., accompanied by a rise in the number of small adipocytes (< 500 µm2) observed on both days 3 and 6 p.t. (Fig. S1H-I, Fig. 1I). Moreover, there was a notable regeneration of PDGFRA+ fibroblasts within the dWAT region (Fig. 1J). Additionally, fibroblasts derived from IMQ-treated skin samples (p.t. day 3 or 6) exhibited an enhanced adipogenic potential in vitro (Fig. 1J-K). Collectively, these findings depict a two-step adipogenic response triggered by the IMQ application. In the early phase of inflammation progression, dWAT is repopulated with PDGFRA+ preadipocytes (pAds), along with neutrophils. Subsequently, a re-expansion of lipid-ladened adipocytes coincides with the regression of neutrophils as the inflammation enters its resolution phase.
Defining the immune response of Pdgfra+ dermal fibroblasts by scRNAseq
To elucidate the cell type-specific immune responses underlying IMQ-induced skin inflammation, we performed single-cell RNA sequencing (scRNA-seq) analysis on both control and IMQ-treated skin samples. The analysis led to the identification of 27 distinct cell clusters, which were categorized into various cell types such as dFBs, keratinocytes (KC), neutrophils (NEU), macrophages (MAC), and T cells, based on the expression of established marker genes (Fig. 2A-B, Fig. S2A-C). Through differential gene expression analysis27, we identified the most differentially upregulated genes across these cell types. Notably, in dermal fibroblasts, Saa3 and Prg4 were markedly upregulated; in keratinocytes, S100a8 and Krt6a were the top upregulated genes; T cells exhibited heightened expression of Il22 and Il17a/f; neutrophils displayed increased expression of Cxcl2; and macrophages exhibited elevated expression of Chil3 (Fig. 2C-D).
To further characterize the immune response of fibroblasts, Pdgfra+ dFBs were classified into seven sub-clusters (r1 ~ r7), denoted as reticular (RET), papillary (PAP), and follicular dFBs, as well as various adipocyte-lineage cell clusters, including adipose regulatory cells (Areg), adipocyte progenitors (AP), and pAds, based on the expression of established dFB marker genes (Fig. 2E, S2D)15. Among all dFB sub-clusters, application of IMQ notably increased the relative abundance of adipocyte-lineage cell clusters (Fig. 2F). Multiple independent cellular differentiation trajectory analyses, including Monocle trajectory analysis, CytoTRACE, and RNA velocity, consistently predicted that dFB_r2 APs were differentiating into dFB_r3 pAds (Fig. 2G, S2E-F). Furthermore, both dFB_r2 and r3 cells showed a significant enrichment of preadipocyte or inflammatory gene signatures after IMQ treatment (Fig. 2H-I, S2G-H), suggesting that IMQ application may promote the accumulation of proinflammatory AP/pAds in the skin. Following IMQ application, Saa3, Prg4, Lcn2, Cxcl1, and Cxcl12 were identified as the most prominently upregulated genes in dFB_r2 and/or r3 cells, while Tnc was uniquely upregulated in dFB_r6 PAP/follicular cells (Fig. 2J, S2I-J). Notably, Prg4 and Fabp4 were co-induced in the dFB_r2 cells, whereas Saa3 showed higher levels of induction in r3, r1, and r6 cells (Fig. 2J, S2I-J). This differential expression pattern of Prg4 and Saa3 may reflect distinct activation states of pAds.
Immunostaining analysis confirmed the presence of PRG4+ cells, which were exclusively localized to the dWAT layer, and these cells co-expressed PDGFRA and PLIN1 (Fig. 2K, S2K), indicating that PRG4 marks activated pAds that are undergoing differentiation into PLIN1+ adipocytes. Additionally, SAA3+ cells, which also co-expressed PDGFRA, were abundantly detected not only within the dWAT but also in the reticular dermis and at the dermal-epidermal junction (DEJ) region in skin treated with IMQ (Fig. 2L). Within the dWAT, SAA3+ cells co-expressed medium-to-low levels of PRG4 (Fig. S2L-M). In contrast, TNC expression was predominantly induced by IMQ in the DEJ and follicular regions, but not in the dWAT (Fig. 2M). These immunostaining findings align with the gene expression pattern observed in the scRNAseq results (Fig. 2J and S2J).
Together, our findings suggest that IMQ application triggers a cascade of events that promote dermal adipogenesis. Specifically, Anxa3+ APs appear to be primed to differentiate into proinflammatory Saa3+ pAds or Prg4+ pAds that have the potential to differentiate into Fabp4+ adipocytes. Additionally, Saa3+ pAds may further differentiate into Saa3+Tnc+ dFBs within the DEJ region (Fig. 2N). It is important to note that further research is required to validate these proposed differentiation trajectories of adipocyte-lineage cells during psoriasis pathogenesis.
Neutrophils trigger the inflammatory response of dFBs through the IL1β-IL1R signaling axis
GO pathway analysis identified IL1 and neutrophil chemotaxis as the most significantly upregulated pathways in dFB_r3 pAds following IMQ application (Fig. 3A). Cell-chat analysis of the IL1b-IL1r1 signaling network pinpointed neutrophils as the primary source of the IL1 ligand, acting on IL1r expressed on dFB clusters (Fig. S3A). Violin plots further illustrated that neutrophils expressed the highest level of Il1b among all major cell types examined, whereas dFBs expressed the highest level of Il1r1 (Fig. 3B). Immunostaining data revealed that IMQ application specifically induced IL1R1 expression in PDGFRA+ dFBs within dWAT (Fig. 4C, S4B). In vitro study also identified Saa3 and Cxcl1 as the most prominent inducible genes by IL1β in primary neonatal dFBs (Fig. 3D), which are composed of over 90% PDGFRA+Ly6A+THY1+ APs/pAds (Fig. S3C). These findings imply that IL1β, released from neutrophils, may play a crucial role in activating dFBs during psoriasis pathogenesis.
To validate the role of neutrophils in activating dFB, we systemically depleted neutrophils via intravenous injection of the Ly6G antibody (Fig. S3D-E). This depletion significantly reduced neutrophil infiltration into the skin and suppressed the development of psoriatic phenotypes induced by IMQ (Fig. 3E-H, S3F). Moreover, IL1b expression was substantially reduced upon neutrophil depletion (Fig. 3I), supporting the idea that neutrophils are a primary source of IL1β in IMQ-treated skin. Consequently, the IMQ-induced mRNA expression of Saa3 and Cxcl1 (Fig. 3I), as well as protein expression of SAA3 in PDGFRA+ dFBs (Fig. 3J-K) were significantly diminished upon neutrophil depletion.
To further validate the role of IL1 signaling in dFB activation, IL1r1−/− (KO) and wildtype (WT) littermate mice were subjected to IMQ-application (Fig. 3L, S3I). IL1r1 deficiency not only inhibited the progression of the psoriatic phenotype (Fig. 3L-M), but also blocked IMQ-induced dWAT expansion (Fig. S3J, Fig. 3N), SAA3 protein expression in PDGFRA+ dFBs (Fig. 3O-P, S3K), and mRNA expression of Saa3 and Cxcl1 (Fig. 3Q). Collectively, these results underscore the critical function of neutrophils in the immune activation of pAds via the IL1β-IL1R1 signaling pathway during the progression of IMQ-induced skin inflammation.
Neutrophils and dFBs engage in a bidirectional IL1β-IL1R and CXCL1-CXCR2 signaling circuit
We identified Cxcl1, a pivotal neutrophil chemotactic gene28,29, and Saa3 as the most highly expressed inflammatory genes inducible by IL1β in dFBs (Fig. 3D). This prompted us to investigated the role of dFB-derived CXCL1 and/or SAA3 in driving neutrophil activation. scRNA-seq analysis showed that dFBs were the primary source of Cxcl1, acting on Cxcr2 expressed by neutrophils, which reciprocally expressed Cxcl2 (Fig. 4A-B). Immunostaining confirmed that CXCL1 was predominantly expressed in PDGFRA+ dFBs located in the dWAT following IMQ treatment (Fig. 4C, S4A).
To further explore the cellular interactions between dFBs and neutrophils, we developed an in vitro co-culture system, in which conditioned medium was collected from IL1β-primed or control dFB (dFBIL1β-CM or dFBctrl-CM) to assess its effect on neutrophils (Fig. 4D). Our findings revealed that dFBIL1β-CM significantly upregulated the expression of Il1b, Cxcl2, and Nos2 in neutrophils (Fig. S4B). Notably, the addition of an anti-CXCL1 antibody reduced the induction of Il1b and Cxcl2, though not Nos2, in neutrophils (Fig. 4E-F, S4C). In contrast, when dFBIL1β-CM was obtained from Saa3 knockdown dFBs, it led to a decrease in Nos2 expression with no significant effect on Il1b or Cxcl2 Nos2 (Fig. 4G, Fig. S4F-G).
Furthermore, dFBIL1β-CM was found to enhance neutrophil migratory activity in a transwell assay, and this effect was significantly attenuated by the anti-CXCL1 antibody (Fig. 4H-I, S4H). In contrast, primary keratinocytes, while responsive to IL17A with an inflammatory response, did not show induced expression of Il1b, Cxcl1, or other keratinocyte-specific inflammatory genes following treatment with either IL1β nor dFBIL1β-CM (Fig. S4I-L). These findings suggest that the CXCL1-IL1β signaling circuit is specific to the interaction between dFBs and neutrophils, rather than with keratinocytes.
In vivo, the administration of an anti-CXCL1 antibody concurrently with IMQ treatment in mice substantially alleviated the manifestation of IMQ-induced psoriatic phenotypes (Fig. 4J-L), reduced the recruitment of Ly6G+ neutrophils to dWAT (Fig. 4M), and diminished the expression of inflammatory genes associated with neutrophils or dFBs (Fig. 4N, S4M). These results collectively indicate that IL1β-activated pAds can augment neutrophil chemotaxis and activation through the secretion of CXCL1 and/or SAA3, contributing to the establishment of a self-perpetuating inflammatory response in in the skin dermis.
Prolonged skin inflammation prompts the differentiation of Preadipocytes to Adipocyte
We next explored the mechanisms underlying the resolution of inflammation during extended imiquimod (IMQ) treatment. First, to determine whether Pdgfra+ pAds can differentiate into adipocytes during the resolution phase of IMQ-induced skin inflammation, we induced CRE activity in Pdgfra-ERT2cre;mTmG mice by tamoxifen application during the initial days of IMQ application to label Pdgfra+ dFBs with GFP (Fig. 5A). At day 6 p.t., we observed co-expression of GFP and FABP4 in the dWAT of IMQ-treated skin (Fig. 5A), supporting the differentiation of PDGFRA+ pAd into FABP4+ adipocytes.
Building upon our previous work that IL1R signaling activation in dFBs triggers a dermal adipogenesis response crucial for during skin development and wound regeneration15, we have now observed that while brief IL1β exposure induced an immediate inflammatory response in dFBs, extended treatment significantly upregulated Pparg, the key transcription factor driving adipogenesis30, along with other adipocyte-related genes (Fig. 5B, S5A). Analysis of the transcriptomic changes during the in vitro adipogenesis process from dFBs/APs to adipocytes revealed a distinct sequence of molecular events. This included the commitment of Anxa3+Thy1hi dFBs to PdgfrahiLy6ahi pAds post-confluency, followed by differentiation into Camp+ early adipocytes (eAd), and ultimately to Pparghi mature adipocytes (Fig. 5C). During this process, there was a rapid decline in the expression of IL1 pathway-related genes, Il1r1 and Cxcl1, and a transient expression of Prg4 by differentiating pAds (Fig. 5C).
In vivo IMQ application mirrored these in vitro changes, showing an upregulation of Prg4, identified as a marker for differentiating pAds, which preceded the induction of Camp, Pparg, and other adipocyte-related genes between days 6 and 10 of IMQ application. Concurrently, there was a significant suppression of inflammatory genes, including Cxcl1, Saa3, Il1b, and Cxcl2 (Fig. 5D-E, S5B-C). Immunostaining confirmed robust PPARγ induction in dermal white adipose tissue (dWAT) 6 days post-IMQ application (Fig. 5F). These findings suggest that a PPARγ-dependent adipocyte differentiation program is a critical component of the resolution phase of skin inflammation.
Characterization of the immune response of Adipocytes by single-nuclei RNA sequencing
To maximally captured adipocytes, which are largely lost through enzymatic digestion in scRNA-seq31, we next performed single-nuclei RNA sequencing (Sn-RNAseq) of IMQ-treated skin samples. Sn-RNAseq identified a distinct cluster of Pparg+ adipocyte, representing 5% of the total cells (Fig. S5D-F). Further reclustering of fibroblast and adipocyte clusters delineated six sub-clusters (r0–r5), including PparghiLpl+Adipoq+ adipocytes (r2), Lpl+Pparglo−med preadipocytes (r1), and various other dermal fibroblast clusters (Fig. 5G-H, S5G). Pseudotime analysis predicted the differentiation trajectory from r1_pAds to r2_adipocytes (Fig. 5I, S5H). Notably, r1_pAds exhibited the highest inflammatory scores and expressed high levels of Cxcl1 and Saa3, whereas r2_adipocytes showed the lowest inflammatory scores and expressed genes associated with adipogenesis (Camp, Adipoq, and Fabp4) but not inflammatory genes (Fig. 5J-K), suggesting that the inflammatory response is suppressed in differentiating adipocytes.
PPARγ-mediated preadipocyte differentiation is necessary for the resolution of skin inflammation
To determine the role of PPARγ in mediating neutrophil clearance, we administered BADGE, a selective pharmacological inhibitor of PPARγ19,32,33, via intraperitoneal injection during IMQ application (Fig. 6A). BADGE treatment led to an exacerbation of the psoriatic phenotype (Fig. 6B-C), inhibited the formation of FABP4+ adipocytes in dWAT, and resulted in increased neutrophil infiltration and upregulation of neutrophil-associated inflammatory genes (Il1b and Cxcl2) in IMQ-treated skin (Fig. 6D-F, S6A-B).
To block adipogenesis by deleting PPARγ in PDGFRA+ pAds, we next generated tamoxifen (TAM)-inducible fibroblast-specific Pparg knockout mice, termed PpargFB−iKO, by crossing Ppargflox/flox mice with Pdgfra-cre/ERT mice (Fig. 6G). TAM application to PpargFB−iKO mice during IMQ-application specifically ablated the expression of PPARγ in dWAT (Fig. S6C), leading to an exacerbation of psoriatic clinical phenotypes (Fig. 6H-I), inhibition of adipogenesis (Fig. 6J, S6D) and Camp expression (Fig. 6K), and increased expression of inflammatory genes (Cxcl2, IL1b, Nos2) (Fig. 6K, S6E). These results highlight the critical role of PPARγ in preadipocyte differentiation and the resolution of neutrophilic skin inflammation during topical IMQ application.
Early adipocytes exhibit anti-inflammatory effects against myeloid cell activation
We next investigated the therapeutic potential of adipocytes in countering inflammation-medidated by myeloid cells, including neutrophils and macrophages. Conditioned medium (CM) was collected from three stages of adipocyte differentiation: undifferentiated (undif) dFB/pAd, early adipocytes (eAd) secreting CAMP, and mature adipocytes (mAd) with elevated FABP4 secretion (Fig. 7A). These CM samples were then used to treat neutrophils or peritoneal macrophages activated by FSL or LPS (Fig. 7A and Fig. S7A). Notably, CM from eAd, but not undif or mAd, significantly reduced the expression of proinflammation genes Cxcl2, IL1b, and Nos2 in activated neutrophils and/or macrophages (Fig. 7B-D, S7B), and induced anti-inflammatory M2-macrophage-associated genes, such as Chil3 and Cd16334, in activated macrophages (Fig. S7C-D).
Subsequently, we explored the in vivo therapeutic effects of eAd-CM against IMQ-induced skin inflammation. eAd-CM i.d. injections substantially alleviated the development of psoriatic phenotypes (Fig. S7E-G), reduced the expression of inflammatory genes (Fig. S7H), and decreased infiltration of Ly6G+ neutrophils in dWAT (Fig. S7I-J) in IMQ-treated skin. Moreover, eAd-CM suppressed IMQ-induced epidermal hyperplasia and the presence of Ki67+ proliferative epidermal cells (Fig. S7J-K), and inhibited the IMQ-mediated induction of Krt6a, Defb14, and Il17a (Fig. S7H). Hence, eAd-CM’s inhibitory effect on epidermal cell activation is likely due to its suppressive effect on myeloid cell activation.
Early adipocyte-derived lipids are anti-inflammatory
We observed that the anti-inflammatory substances in eAd-CM were heat-stable (Fig. 7E, S7L-M), suggesting they are not protein-based. In addition, only the lipid fraction, but not the protein fraction of eAd-CM, exhibited inhibitory effects against LPS-induced Il1b, Cxcl2, and Nos2 expression (Fig. 7F-G, S7N), demonstrating that the anti-inflammatory substances in eAd-CM are lipids. To develop a topical approach to deliver eAd_lipids, we utilized Haliclona sp. spicules (SHS), microneedles derived from marine sponges, facilitating skin penetration of therapeutics even nanoparticles by overcoming barriers and creating nano-pores across the skin epithelium35,36. Topical application of lipids from eAd, in combination with SHS, effectively prevented the onset of IMQ-induced psoriatic features (Fig. 7H-J), blocked the infiltration of Ly6G+ neutrophils (Fig. 7K), and reduced the number of epidermal Ki67+ cells (Fig. 7K-L). Furthermore, it suppressed pro-inflammatory gene expression and restored the expression of claudin genes important for epidermal tight junctions (Fig. 7M, S7O).
These results underscore the role of early adipocytes in secreting lipids with anti-inflammatory properties to attenuate myeloid cell activation, thereby aiding in the resolution of inflammation during IMQ application.
Preadipocyte Signature and IL1 Pathway Enrichment Characterize Proinflammatory Fibroblasts in Human Psoriasis
Next, we sought to establish the relevance of our mouse findings to human psoriasis by reanalyzing single-cell RNA sequencing (sc-RNA-seq) data from patients with psoriasis and healthy controls37. Aligning with mouse data, PDGFRA+ dFBs were identified as the main producers of IL1R1, CXCL1, and PRG4, with myeloid cells predominantly expressing IL1B in psoriatic lesions (Fig. 8A). Reclustering of PDGFRA+ dFBs resulted in nine dFB subclusters (r0–r8), and pseudotime analysis predicted that r1, r2, and r7 clusters were in the terminal state of cellular differentiation (Fig. 8B-C, S8A). Notably, correlation analysis revealed that the human r1 and r2 dFB clusters were closely related to the murine IMQ-induced dFB_r3 cluster, which represents the pro-inflammatory pAds (Fig. 8D). Additionally, dFB_r2 cells exhibited enrichment in pAd and inflammatory gene-set signatures (Fig. 8E).
Gene Ontology (GO) analysis identified pathways, including responses to IL1, WNT, and TGFβ, as the top enriched pathways associated with dFB_r2 cells in human psoriasis (Fig. 8F). Given our previous findings that TGFβ and WNT are key inhibitors of dermal adipogenesis15,18, their overactivation may impede the transition of pro-inflammatory pAds into anti-inflammatory adipocytes, thereby potentially exacerbating dermal inflammation.
To gain insight into the increased neutrophil infiltration in generalized pustular psoriasis (GPP) compared to plaque psoriasis (PV), we analyzed transcriptomic data from normal, PV, and GPP human skin samples (GSE79704)38. We found that GPP samples expressed higher levels of IL1B, CXCL1, and key adipogenesis inhibitor genes WNT3 and TGFB1, compared to normal and/or PV samples. In contrast, genes related to adipogenesis, including PPARG, PRG4, CEBPB, and FABP4, were downregulated in GPP samples (Fig. 8G, S8B). Furthermore, correlation analyses showed that IL1B expression positively correlated with CXCL1 and negatively with PPARG and PRG4, while PPARG and PRG4 exhibited a positive correlation (Fig. 8H, S8C).