Renal Disease

WNT-dependent interaction between inflammatory fibroblasts and FOLR2+ macrophages promotes fibrosis in chronic kidney disease – Nature Communications

Time Stamp: January 25, 2024 12:00 AM
Source Node: 38495

Single cell RNA sequencing identifies different clusters of mesenchymal cells in kidney disease

To assess fibroblastic heterogeneity during kidney fibrosis development, we took advantage of an existing human single cell RNA sequencing (scRNAseq) dataset from patients with or without chronic kidney disease (CKD)⁸. After quality control and doublet elimination, 2908 mesenchymal cells were selected for further analyses. Unsupervised graph-based clustering identified 13 clusters, visualized with the Uniform Manifold Approximation and Projection (UMAP) algorithm (Fig. S1a). To simplify our message and increase the relevance of our comparisons, we focused on the first 6 clusters, as they represented more than 85% of the total mesenchymal population (Fig. 1a), while the other clusters contained less than 80 cells per cluster (Fig. S1b). All clusters were found in at least 2 patients, albeit in different proportions from one patient to the other (Fig. 1b and Fig. S1b). We confirmed that these different clusters were differently distributed between control and CKD patients (Fig. 1c, d). Indeed, the content in clusters 0, 3, 4, and 5 was increased in CKD patients, while the clusters 1 and 2 accumulated in controls (Fig. 1d). We first confirmed that almost all cells expressed VIM (VIMENTIN) and PDGFRB, confirming their mesenchymal origin (Fig. 1e). We also observed that high expression of RGS5 (Regulator of G protein signaling 5) and NOTCH3, specific pericyte markers, were mainly detected in clusters 1 and 2, while the pan-fibroblast marker genes PDGFRA and DCN (DECORIN) were expressed in clusters 0, 3, 4, and 5 (Fig. 1e). We thus hypothesized that clusters 1 and 2 might be pericytes-like and clusters 0, 3, 4, 5 fibroblasts-like. Differential gene expression and functional enrichment analysis confirmed this assumption and revealed that each cluster was also characterized by a specific transcriptional profile (Table S1). Although all the fibroblastic clusters were associated with extracellular matrix (ECM) remodeling pathways, each cluster was associated with specific processes: cluster 0 was also characterized by cytokine signaling pathways and inflammation, cluster 3 by response to wounding, cluster 4 by complement and coagulation cascades, and cluster 5 by translation elongation and wound-healing, among top-ranked pathways in each cluster (Table S1). In addition, the pericyte-like cluster 1 was associated with IFNα/β-dependent signaling and blood vessel development, and cluster 2 with muscle system process and oxidative phosphorylation (Table S1). In agreement with these observations, CXCL12 was highly expressed in cluster 0, and the expression of COL1A1, COL3A1, and POSTN (some of the main ECM components in fibrosis and marker of myofibroblasts) was mainly detected in clusters 3 and 5 (Fig. 1f).

**Fig. 1: Identification of distinct mesenchymal clusters in CKD.**

Interestingly, we found that the aforementioned pathways detected in fibroblasts isolated from kidney diseases were highly reminiscent of those previously identified in specific CAF populations in cancer^{14,15,16,17,20,21,25,26,27,28,29,30,31,32,35}. Indeed, the CAF-S1-specific gene signature²⁹, which identifies populations of inflammatory and myofibroblastic FAP + CAF associated with metastatic spread and immunosuppression^{15,16,18,19,29} was highly detected in the fibroblastic clusters (clusters 0, 3, 4, 5) in the kidney disease dataset, while the CAF-S4 signature (perivascular-like CAF, gene signature in Table S2) highlighted the pericyte-like cells (clusters 1, 2) (Fig. 1g). On the one hand, the signatures from detox-iCAF, IL-iCAF and IFNγ-iCAF subsets previously identified by scRNAseq in the CAF-S1 (FAP + CAF) population from breast cancer²⁹ highlighted mainly the clusters 0 and 4 (Fig. 1h), thereby confirming that these clusters were inflammatory. On the other hand, the ECM-myCAF, TGFβ-myCAF and Wound-myCAF signatures from CAF-S1 highlighted the clusters 3 and 5 (Fig. 1g). Interestingly, all the gene signatures from CAF-S1 fibroblasts, but not CAF-S4, showed an increased expression in CKD compared to controls (Fig. 1g, h), in total agreement with accumulation of clusters 0, 3, 4 and 5 (fibroblast-like) but not of clusters 1 and 2 (pericyte-like) in CKD patients compared to controls (as shown in Fig. 1c, d). We next sought to focus our further analyses on these clusters 0, 3, 4, and 5, which accumulate in CKD patients. To define specific markers of these clusters, we performed a pairwise analysis of the genes differentially expressed between the different clusters and we identified SFRP1 (Secreted frizzled Related Protein 1) as significantly up-regulated in cluster 0 (although a few cells in clusters 3 and 4 also expressed SFRP1), FAP (Fibroblast Activation Protein) expressed in both clusters 3 and 5, SFRP4 (Secreted frizzled Related Protein 4) in clusters 3 and 4 and RAMP1 (Receptor Activity Modifying Protein 1) in cluster 5 (Fig. 1i and Fig. S1c). By this way, cluster 0 could be identified as SFRP1 + SFRP4- FAP- RAMP1-; cluster 4 as SFRP1 ± SFRP4 + FAP- RAMP1-; cluster 3 as SFRP1 ± SFRP4 + FAP + RAMP1- and cluster 5 as SFRP1- SFRP4- FAP + RAMP1 + . Taken as a whole, by combining the transcriptomic profiles of these different clusters, their gene similarities with CAF-S1 clusters, and their expression of specific markers, we defined the CKD clusters 0 and 4 as inflammatory (and thus referred to as iFibro), and the CKD clusters 3 and 5 as ECM-secreting myofibroblasts (then termed myoFibro). Moreover, based on their expression profiles (Table S1), we annotated CKD clusters more precisely as follows: cluster 0, CXCL-iFibro; cluster 4, Detox-iFibro; cluster 3, Wound-myoFibro and cluster 5, TGFβ-myoFibro. The pericyte-like clusters 1 and 2, which were down-regulated in CKD, were referred to as IFNα/β-Peri-like and Contractile-Peri-like, respectively (Fig. 1j). Taken as a whole, these data show an important heterogeneity of fibroblasts in CKD, which is reminiscent of the different CAF populations recently identified in cancer.

As Kuppe et al. described several clusters of mesenchymal cells using PDGFRβ+ sorted cells⁸, we sought to identify the similarities between our fibroblast annotations and those from this study (as reported in Fig. S1d). Hence, we performed a label transfer analysis on PDGFRβ+ sorted cells from⁸. The dataset of 2495 mesenchymal cells with our own annotations (described in Fig. 1j) was used as a reference, and the dataset of PDGFRβ+ sorted cells (described in Fig. S1d), with original annotations from⁸ was used as a query. Interestingly, the pericyte-like populations that we identified (IFNαβ-Peri-like and Contractile-Peri-like) showed a high degree of similarity with pericytes and vascular smooth muscle cells described by Kuppe et al (Fig. S1e and Table S3). Similarly, the Detox-iFibro we described corresponded to the Fibroblast 1 population that Kuppe described as being a SCARA5 + MEG3+ non-activated fibroblasts, and the TGFβ-myoFibro was similar to the Myofibroblast 1, the population exhibiting the highest level of ECM protein secretion. In contrast, CXCL-iFibro, and to a lesser extent Wound-myoFibro were identified as a mix of different fibroblast and/or myofibroblast populations, according to original annotations. In particular, the CXCL-iFibro population that we identified as being a homogenous population was dispersed throughout several clusters in the Kuppe et al. study and characterized as a mix of Fibroblast 2a (30.5% of cells), Myofibroblast 3a (22.1%), Myofibroblast 3b (17.2%), Fibroblast 2b (14.9%) and Myofibroblast 2b (12.3%) (Fig. S1e and Table S3). Thus, annotating mesenchymal cells based on similarities with CAF allowed us to identify a population of fibroblasts with inflammatory properties that has not been explored further previously.

CXCL-iFibro and ECM-myFibro clusters accumulate at early and late CKD stages, respectively

Inflammatory fibroblasts (iFibro) have been poorly described in kidney fibrosis. To further our understanding, we performed trajectory inference and pseudotime analysis using Monocle 3 (Fig. 2a–c). Despite a longstanding debate regarding the origins of myofibroblasts in the kidney, several reports identified pericytes, as a major source of myofibroblasts in kidney fibrosis development^3,4,5,6,7,8. Therefore we defined the root of the pseudotime in pericytes-like mesenchymal cells. These analyses indicated that CXCL-iFibro (cluster 0) might be an intermediate stage in the differentiation from pericyte-like cells (clusters 1 and 2) toward ECM-secreting myoFibro (clusters 3 and 5) (Fig. 2a, b). The expression of specific marker genes according to pseudotime confirmed this finding (Fig. 2c), and our observations in human CKD were consistent with previous lineage tracing experiments performed in mouse CKD models^3,5,6,7. To validate the temporal dynamics of inflammatory fibroblasts and ECM-secreting myofibroblasts, we took advantage of publicly available dataset. First, we analyzed bulk RNAseq data from mice undergoing unilateral ureteral obstruction (UUO) at different time points³⁸. UUO is a classical and well characterized model of kidney fibrosis in mice, and we analyzed samples from control mice, and from mice 3 days after UUO (early time point with infiltration but no fibrosis), 7 days after UUO (intermediate level of fibrosis) and 14 days after UUO (with a high level of interstitial fibrosis). First, we calculated an ECM score characterized by the expression of ECM-related genes as described in Naba et al.³⁹, which allowed us to validate that this score increased progressively from day 0 to day 14 after UUO (Fig. 2d), and confirm the progressive increase in interstitial fibrosis. Then, to calculate the proportion of each cell type according to each time point, we performed a deconvolution of the bulk RNAseq data using BayesPrism⁴⁰. To do so, we first built a comprehensive cellular atlas (Fig. 2e) based on scRNAseq dataset from CKD and normal kidney tissues⁸. This cellular atlas was composed of 49 226 cells corresponding to 9 different cell types (Fig. 2e). Interestingly, deconvolution results showed that kidneys from control mice were composed of mainly proximal tubular epithelial cells (around 80%), whereas after injury the proportion of proximal tubular cells decreased (Fig. 2f). This result validated our approach, as it is largely known that the cellular composition of a normal mouse kidney is approximately 70-80% of proximal tubular epithelial cells, and that this proportion significantly decreases after injury⁴¹. In addition, we observed a progressive increase in the proportion of mesenchymal cells, as well as of the immune compartment and injured tubules (Fig. 2g). To have more insights into the mesenchymal compartment, we focused our next analysis specifically on these cells. While pericytes were not detected in high proportion, we observed a gradual switch from inflammatory fibroblasts to ECM-secreting myofibroblasts from day 3 to day 14 (Fig. 2g). Interestingly, inflammatory fibroblasts were virtually absent at day 14 (Fig. 2g). Finally, we confirmed our observations from pseudotime analysis in patients (Fig. 2a–c) as we saw inflammatory fibroblasts arising early after injury (day 3), being maintained at intermediate stages (day 7), before disappearing at advanced stages of interstitial fibrosis (day 14) (Fig. 2h), meanwhile the proportion of ECM-secreting myofibroblasts increased progressively up to day 14 (Fig. 2i). Finally, to increase the resolution of our analysis, we analyzed a single cell RNAseq dataset from mice who underwent UUO, with available data for sham-operated mice, UUO day2, UUO day7 and reversal UUO. Reversal UUO corresponds to the reimplantation of the ureter in the bladder after 7 days of UUO to observe a healing phase, where a decrease in Collagen deposition, myofibroblast activation and macrophage infiltration was observed⁴². To assess the temporal dynamics of inflammatory fibroblasts in this model, we performed a label transfer analysis using our cellular atlas as reference. Interestingly, we observed a progressive increase in the proportion of inflammatory fibroblasts from day 2 after UUO to day 7 after UUO, followed by a decrease in the reversal UUO (Fig. 2j), confirming that inflammatory fibroblasts are associated with fibrosis expansion. Altogether, these data identify a population of inflammatory fibroblasts, which arises early after kidney injury, before decreasing at late stages when ECM-secreting myofibroblasts expand.

**Fig. 2: Accumulation of CXCL-iFibro and ECM-myFibro clusters at distinct stages of chronic kidney disease.**

We next sought to validate the presence of these different clusters of fibroblasts in human tissue sections at different stages of the disease (Fig. 2 k–n and Fig. S2a, b for specificity of staining). In order to capture the fibroblast heterogeneity highlighted in scRNAseq data, we first focused on explants from patients with polycystic kidney disease (PKD), an autosomal dominant genetic disease causing cyst development and ultimately leading to end-stage renal disease (Fig. 2k, l). PKD kidneys exhibited very large fields of fibrosis, which characterized the terminal stage of kidney fibrosis development. Immunohistochemistry (IHC) on PKD explants targeting CXCL-iFibro (SFRP1), ECM-secreting myoFibro (FAP), Wound-myoFibro (SFRP4) and TGFβ-myoFibro (RAMP1) validated the existence of these different clusters in tissue sections (Fig. 2k, l and Fig. S2a, b). Co-staining experiments between SFRP1 or FAP and αSMA showed that SFRP1+ and FAP+ cells were mainly αSMA+, consistent with them being activated fibroblasts (Fig. S2c, d). We also observed that in CKD but not in normal kidneys some tubular cells could also be positive for SFRP1 (Fig. 2k, l and Fig. S2e). Consistent with this, it is known that, during CKD, some tubular cells can express some mesenchymal markers (such as VIM or αSMA) and undergo partial epithelial to mesenchymal transition^43,44. Strikingly, CXCL-iFibro (cluster 0) was more abundant in zones with persistent epithelial structures, while ECM-secreting myoFibro (Wound-myoFibro, cluster 3 and TGFβ-myoFibro, cluster 5) were mainly observed in fibrosis-enriched zones (Fig. 2k, l). Quantification of the histological scores (H-scores) of the different fibroblast markers, confirmed that the quantity of CXCL-iFibro (SFRP1+, cluster 0) was anti-correlated with the percentage of fibrosis (evaluated by a nephrologist and a pathologist at diagnosis) within PKD sections (Fig. 2k). In contrast, we found a linear positive correlation between the percentage of fibrosis and markers of ECM-secreting myoFibro (such as FAP, as well as SFRP4 and RAMP1), and specific markers of Wound-myoFibro (cluster 3) and TGFβ-myoFibro (cluster 5), respectively in PKD patients (Fig. 2l). Because PKD explants correspond to end-stage disease, we next measured the content of these different clusters at earlier stages in kidneys with a lower percentage of fibrosis (Fig. 2m, n). In that aim, we performed IHC on kidney biopsies from patients suffering from vascular nephropathy with interstitial fibrosis ranging from 0 to 80% of the parenchyma. The CXCL-iFibro and the two ECM-secreting myoFibro clusters, including the Wound-myoFibro and TGFβ-myoFibro clusters, were detected in kidney biopsies at these early stages of chronic kidney disease (Fig. 2m, n). Interestingly, H-score quantification integrating the proportion of positive cells within the interstitium revealed that CXCL-iFibro expanded in the interstitium at initial stages of fibrosis, before decreasing when the fibrosis percentage increased (Fig. 2m). In contrast, histological scoring of the different ECM-secreting myoFibro markers were linearly correlated with the percentage of fibrosis (Fig. 2n), indicating a constant increase in their content when fibrosis gradually developed in patients. The decrease in the content of the CXCL-iFibro cluster concomitantly to the linear increase in ECM-secreting myoFibro clusters when fibrosis was above 50% was consistent with the in-silico analysis using pseudotime and trajectory inference (shown in Fig. 2b, c). Altogether, these data validate the existence of CXCL-iFibro in kidney fibrosis at early stages and suggest that these inflammatory fibroblasts might be an early state of differentiation towards ECM-secreting myofibroblasts both in human CKD patients and a representative mouse model.

The content in CXCL-iFibro correlates with FOLR2+ macrophage infiltration in kidney disease

Based on the accumulation of the CXCL-iFibro cluster at early phase of kidney fibrosis, we next aimed to investigate its functional role in CKD. CXCL-iFibro expressed high levels of several cytokines and chemokines, suggesting potential interactions with immune cells. Interestingly, we observed that the CXCL-iFibro accumulated in COL1A1-negative zones (Fig. S2f), consistent with the lack of COL1A1 expression in scRNAseq data from CXCL-iFibro (as shown in Fig. 1e). In contrast, the majority of ECM-secreting myoFibro (SFRP1- FAP+) were mainly detected in COL1A1-positive zones (Fig. S2f), as expected based on the high expression of the COL1A1 gene in these clusters (Fig. 1e).

As macrophages are well-known to be key players in the development of fibrosis^{45,46,47,48,49}, we assessed the spatial distribution of CXCL-iFibro and macrophages in PKD explants and fibrotic kidney biopsies by performing co-staining of both CXCL-iFibro (SFRP1 + FAP-) and macrophages (CD68) (Fig. 3a, b and Fig. S3a, b for corresponding low magnification). Thus, by performing IF staining using specific markers, we confirmed the identity of these cellular clusters in PKD explants and fibrotic kidney biopsies. Interestingly, we observed a proximity and a strong positive correlation between the number of CXCL-iFibro (SFRP1+ FAP-) and the number of macrophages (CD68+) per mm² in both PKD explants (Fig. 3a and Fig. S3a) and fibrotic kidney biopsies (Fig. 3b and Fig. S3b). Interestingly, within each patient tissue section, each field that showed macrophage infiltration also exhibited CXCL-iFibro accumulation in both PKD and fibrotic kidney biopsies (Fig. 3a, b, quantifications in bottom panels). Conversely, each field that did not show CXCL-iFibro infiltration exhibited poor macrophage infiltration, confirming that the colocalization between SFRP1+ and CD68+ cells was not patient-dependent (Fig. 3a, b). Importantly, this correlation was not observed between the content in ECM-secreting myoFibro (FAP+) and macrophage infiltration in PKD (Fig. 3c and Fig. S3c). Indeed, while ECM-secreting myoFibro accumulated in the highly fibrotic zones in PKD explants, very few CD68+ cells were observed in these zones and mainly in FAP-negative zones (Fig. 3c and Fig. S3c). The co-localization between pro-inflammatory CXCL-iFibro and CD68+ macrophages suggests that CXCL-iFibro could be instrumental in attracting and promoting CD68+ macrophage accumulation at early stages of fibrosis development.

**Fig. 3: Inflammatory fibroblasts are in close vicinity of FOLR2+ macrophages.**

We next sought to identify more precisely the CD68+ myeloid cells in kidney fibrosis. Recent efforts have been made, especially in cancer, to better characterize the heterogeneity of myeloid cells and particularly monocytes and macrophages. Several types of tumor-associated macrophages (TAM), such as FOLR2+ (Folate receptor beta) and TREM2+ (Triggering receptor expressed on myeloid cells 2) TAM have been identified^50,51,52. Interestingly, these subtypes of macrophages drive either pro-inflammatory response (FOLR2+) or immunosuppression (TREM2+)^50,51,52. As these macrophage subtypes have not yet been identified in kidney diseases, we first analyzed the different myeloid cell clusters detected in publicly available scRNAseq data from CKD and controls⁸. After quality control and doublet elimination, 3960 myeloid cells were conserved for further analyses. Unsupervised graph-based clustering identified 6 clusters (Fig. 3d). These myeloid clusters were differently distributed in control and CKD patients, with a significant decrease in the content of clusters 0, 2 and 5 and accumulation of clusters 1, 3, and 4 in CKD patients compared to controls (Fig. 3e, f). These clusters were represented, albeit in different proportions, in several patients (Fig. 3g). Among these myeloid clusters, 3 clusters (0, 2 and 3) were identified as monocytes based on VCAN (Versican) expression, while three others (clusters 1, 4 and 5) corresponded to macrophages based on C1QA (Complement C1q A chain) and MRC1 (Mannose receptor C type 1) expression (Fig. S3d). As previously described⁵⁰, we identified two types of monocytes: one expressing CD14 that we annotated as CD14+ monocytes (cluster 0), and the other expressing CD16, hence referred to as CD16+ monocytes (clusters 2 and 3) (Fig. S3e). More precisely, Cluster 2 expressed RUNX3 whereas cluster 3 did not (Fig. S3e). Interestingly, in the macrophage populations, we could identify specific clusters expressing some of the recently published TAM markers including. TREM2 (left part of cluster 1) and FOLR2 (clusters 4 and 5) (Fig. S3f). It has been described that kidney resident macrophages express both CD74 and CD81⁵³. Interestingly, this population corresponded mainly to cluster 5, which was also positive for FOLR2 (Fig. S3f). We thus defined FOLR2+ CD81+ macrophages (cluster 5), which decreased in CKD, as FOLR2+ resident macrophages and FOLR2+ CD81- macrophages (cluster 4), which accumulated in CKD, as FOLR2+ CKD macrophages (Fig. S3g and Fig. 3d–f). Differential gene expression and functional enrichment analysis confirmed that these 2 FOLR2+ populations were also characterized by a specific transcriptional profile. FOLR2+ resident macrophages showed activation of RHO-GTPase pathway, regulation of protein catabolic process, cytokine production and regulation of proteolysis, potentially corresponding to a scavenging-macrophage phenotype, while FOLR2+ CKD macrophages showed an enrichment in the cellular response to stress, inflammatory response and chemotaxis, corresponding to a more inflammatory phenotype (Table S4). Finally, cluster 1 was composed of TREM2+ macrophages, and dendritic cells (DC), which were positive for MRC1, CD1E and CD1C expression (Fig. S3f). We thus annotated cluster 1 as TREM2+ macrophages / DC (Fig. 3d, f and Fig. S3g).

As these different subtypes of FOLR2+ and TREM2+ macrophages have been poorly described but very recently detected in diabetic kidney disease⁵⁴, we tested whether we could detect them in PKD explants and fibrotic kidney biopsies, and we evaluated if they were localized in close vicinity to inflammatory fibroblasts (Fig. 3h, i and Fig. S3h, i for corresponding low magnification). Co-staining of SFRP1, FOLR2 and TREM2 markers by IF in these tissue samples showed a clear increase in the proportion of FOLR2+ macrophages in CKD upon fibrosis development (Fig. 3h, i). Strikingly, we observed a strong correlation between the number of SFRP1 + CXCL-iFibro and FOLR2+ macrophages per mm² (Fig. 3h, i), correlation not detected with FAP + ECM-secreting myoFibro (Fig. 3j and Fig. S3j for corresponding low magnification). On the other hand, TREM2+ macrophages were identified in kidney samples, but in different fields than CXCL-iFibro and FOLR2+ macrophages (Fig. S3k).

To further validate the spatial distribution of the different fibroblast and myeloid populations we identified in CKD, we performed spatial transcriptomics on 2 patients suffering from kidney fibrosis (Table S5 for clinical characteristics of these patients). To analyze these data, we used the comprehensive cellular atlas described in Fig. 2e, based on scRNAseq data sets from CKD and normal kidney tissues⁸. We finely annotated cell populations, resulting in 35 different cell types and states (Fig. S4a, see also Methods). We then mapped the localization and the abundance of each cell population by performing deconvolution using the Cell2location algorithm⁵⁵ with our single cell-based cellular atlas as reference. We applied non-negative matrix factorization analysis to underlie structures and patterns of co-localizing cell types and states. At low-resolution, we first confirmed some expected colocalization, such as between glomerular capillaries and podocytes (glomerular compartment), arteriolar endothelium and pericytes, or proximal tubular cells and vasa recta (Fig. 3k, l). We next observed colocalization between injured tubules, fibroblasts including CXCL-iFibro and immune cells (Fig. 3k, l). This was of interest, as it has recently been suggested that injured tubules exhibit a pro-inflammatory transcriptomic profile and could trigger immune cell infiltration during CKD progression in mice^56,57,58. To go further in our analysis of colocalization between CXCL-iFibro and immune cells, we increased the number of factors in our colocalization analysis, allowing a more resolutive discrimination between the different cell types (Fig. 3m). Strikingly, we observed that FOLR2+ −CKD macrophages mainly colocalized with CXCL-iFibro and plasma cells (Fig. 3l, m and Fig. S4b, c). On the other hand, other fibroblast subsets (IFNαβ-peri like, Detox-iFibro, Wound-myofibro and TGFβ-myofibro colocalize together with monocytes, FOLR2 resident macrophages, B cells, T cells, dendritic cells and TREM2+ macrophages (Fig. 3l, m). Altogether, these data show that CXCL-iFibro and FOLR2+ macrophages are co-localized during CKD, suggesting that a reciprocal crosstalk could exist between these two populations.

CXCL-iFibro attract CD14+ monocytes and induce their differentiation into FOLR2+ macrophages

To decipher the molecular crosstalk between CXCL-iFibro and macrophages, we performed functional assays. To do so, we first generated in vitro cellular models recapitulating the main characteristics of both CXCL-iFibro and ECM-secreting myoFibro clusters. Based on our expertize on iCAF and myCAF isolation from CAF-S1 in breast cancer²⁹, we found that kidney-derived primary fibroblasts acquired distinct phenotypes according to the coating conditions used to expand them in culture (Fig. 4a, b). Indeed, when fibroblasts were expanded on collagen-coated plates, they exhibited higher SFRP1 and lower SFRP4, FAP and αSMA protein levels than cells cultured on plastic-dishes (Fig. 4a, b), showing that collagen-cultured primary fibroblasts were reminiscent of CXCL-iFibro and plastic-cultured fibroblasts of ECM-secreting myoFibro, as previously observed for CAF-S1²⁹. In addition, as shown for iCAF and myCAF from pancreatic and breast cancer^26,29, we observed that the increase in αSMA protein level was a marker differentiating CXCL-iFibro from ECM-secreting myoFibro in vitro (Fig. 4a, b). Of note, neither collagen- nor plastic-cultured cells expressed E-Cadherin / CDH1, a marker of epithelial cells (Fig. 4a), thereby confirming that the primary cells established from human kidney are fibroblasts.

**Fig. 4: CXCL-iFibro attract CD14+ monocytes and induces a switch into FOLR2+ macrophages.**

We then performed RNA sequencing on collagen- or plastic-cultured fibroblasts to confirm the respective identity of these cells. Consistent with the inflammatory identity detected in CXCL-iFibro, we identified several inflammatory genes, including CXCL12, IL1B and IL34, upregulated in collagen-cultured compared to plastic-cultured cells (Fig. S5a, b). These proteins are of interest, because of their role in chemoattraction of immune cells (CXCL12)⁵⁹, pro-inflammatory response (IL1B)^60,61 or in promoting the differentiation and viability of monocytes and macrophages through the colony-stimulating factor-1 receptor (IL34)⁶². Similarly, consistent with their ECM-secreting myoFibro identity, plastic- cultured fibroblasts exhibited increased expression of ECM-related genes, such as COL4A1 and COL4A5 (Fig. S5d), which are known to be important in the fibrotic process and basement membrane integrity⁶³. We also identified other components of the ECM, belonging to the integrin (IGFBP1) or the laminin (LAMA3) families (Fig. S5d). Altogether, these data suggest that collagen-cultured and plastic-cultured fibroblasts are reminiscent of CXCL-iFibro and ECM-secreting myoFibro, respectively.

To investigate interactions between CXCL-iFibro and monocytes/macrophages, we performed an in-silico ligand-receptor interaction study from scRNAseq data using the CellChat algorithm, which quantitatively infers intercellular communications⁶⁴. By focusing our analysis on CXCL-iFibro and myeloid cells, we observed that the highest number and strength of interactions of CXCL-iFibro were with CD14+ monocytes (Fig. 4c, d). Interestingly, this in silico analysis suggested that 3 couples of ligand-receptor might drive this interaction, respectively CXCL12-CXCR4, ANGPTL1-LILRB3 and THBS-CD36 (Fig. 4e–g). Because we observed a close vicinity between CXCL-iFibro and FOLR2+ macrophages (Fig. 3h, i), we hypothesized that CXCL-iFibro could attract CD14+ monocytes and modify their phenotype into a FOLR2+ macrophage phenotype, hypothesis consistent with the preferential interaction detected in-silico between CXCL-iFibro and CD14+ monocytes. To test this hypothesis, we designed co-culture experiments between collagen-cultured primary fibroblasts (CXCL-iFibro) and CD14+ monocytes isolated from the peripheral blood mononuclear cells (PBMC) of healthy donors, to assess if CXCL-iFibro could indeed attract CD14+ monocytes and potentially promote their differentiation into a FOLR2+ macrophage phenotype. We first checked whether CXCL-iFibro could attract CD14+ monocytes by performing transwell migration assays. We observed that CXCL-iFibro stimulated the migration of CD14+ monocytes more efficiently than ECM-secreting myoFibro (Fig. 4h). Moreover, co-culture experiments of CXCL-iFibro or ECM-myoFibro with CD14+ monocytes showed that CXCL-iFibro significantly increased the proportion of FOLR2 + CD206+ (MRC1) macrophages from CD14+ monocytes (Fig. 4i). ECM-secreting myoFibro also induced a switch towards FOLR2+ macrophages, but to a much lesser extent than CXCL-iFibro (Fig. 4i). Thus, taken together, our data highlight that CXCL-iFibro attract monocytes and promote their differentiation into FOLR2+ macrophages.

Macrophages induce the transition from CXCL-iFibro to ECM-secreting myoFibro through the WNT/β-catenin pathway

Based on the crosstalk between CXCL-iFibro and FOLR2+ macrophages, we next tested if, in turn, macrophages could modify the phenotype of kidney-derived primary fibroblasts. We observed that upon co-culture with CD14+ cells, CXCL-iFibro experienced an increase of both αSMA and SFRP4 proteins to a level close to the one of ECM-secreting myoFibro (Fig. 5a), suggesting that myeloid cells could induce the switch of CXCL-iFibro into ECM-secreting myoFibro. To decipher the molecular pathways involved in the transition from CXCL-iFibro towards ECM-secreting myoFibro, we performed an in-silico analysis, using transcription factor inference models on the trajectory inference analysis shown in Fig. 2a, particularly looking at the node between CXCL-iFibro and ECM-secreting myoFibro (as shown Fig. S6a, Table S6). Using the Dorothea algorithm⁶⁵, we observed that TCF4 and TCF12, two key mediators of the WNT/β-catenin pathway, were in the top-10 transcription factors involved in the transition from CXCL-iFibro to ECM-secreting myoFibro (Fig. 5b and Table S7). We also validated this finding by applying another in-silico approach by using Monocle 3. Monocle 3 classifies differentially expressed genes into modules of genes, which are co-regulated along the trajectory⁶⁶. We identified 7 gene modules, which were expressed at different points of the trajectory and specifically characterized these distinct states (Fig. S6b). Interestingly, we identified that module 4 highlighted the transition from CXCL-iFibro to ECM-secreting myoFibro (Fig. 5c, d). Indeed, we observed that the module 4 signature highlighted both CXCL-iFibro and Wound-myoFibro on the UMAP, whereas module 3 was exclusively highlighted in CXCL-iFibro (Fig. 5c). Transcriptomic profiles of the gene-modules 3 and 4 through the Metascape platform revealed that module 3 exhibited an inflammatory signature, while module 4 showed an ECM-related signature (Table S8 and Table S9), in agreement with the transition from CXCL-iFibro to ECM-secreting myoFibro. By performing functional enrichment using the TRRUST database⁶⁷ on module 4 genes, we identified TCF4 as one of the main transcription factors involved in the switch from CXCL-iFibro to ECM-secreting myoFibro (Fig. 5e). Strikingly, the TCF4 transcription factor was thus identified by the 2 approaches (Fig. 5b, e), which highlighted its relevance. Moreover, TCF4 expression was more prominent in ECM-secreting myoFibro (Fig. S6c). TCF4 expression along the pseudotime defined in Fig. 2b showed that its expression increased late in the differentiation, in ECM-secreting myoFibroblasts (Fig. S6d). Finally, expression of TCF4-target genes was also increased in Wound-myoFibro, strengthening the idea that TCF4 could be instrumental in the transition from CXCL-iFibro to ECM-secreting myoFibro (Fig. S6e). Interestingly, macrophages have been shown to participate in kidney regeneration by activating the WNT/β-catenin pathway in epithelial cells^68,69. We next aimed to validate the role of the WNT/β-catenin pathway in macrophage-induced switch of CXCL-iFibro into ECM-secreting myoFibro by functional assays. First, we treated CXCL-iFibro with two different specific agonists of the WNT/β-catenin pathway (CAS 853220-52-7; SKL2001) and observed that the activation of this pathway was concomitant with the increase of αSMA and SFRP4 protein levels (Fig. 5f), suggesting that its activation could drive the phenotype from CXL-iFibro to ECM-secreting myoFibro. Second, we observed that cultured ECM-secreting myoFibro showed more nuclear β-catenin staining than CXCL-iFibro (Fig. 5g), consistent with enhanced activation of the WNT-pathway in ECM-secreting myoFibro compared to CXCL-iFibro. We next tested if the macrophage-induced switch from CXCL-iFibro to ECM-secreting myoFibro could be driven by the WNT/βcatenin pathway. To do so, we co-cultured CD14+ myeloid cells with CXCL-iFibro either treated or not with an inhibitor of the β-catenin/TCF interaction (iCRT3). We observed that addition of the β-catenin/TCF interaction inhibitor did not affect the proportion of FOLR2+ macrophages upon co-culture (Fig. S7), indicating that the treatment had no impact on the differentiation of CD14+ monocytes into FOLR2+ macrophages. Without treatment, we confirmed by both IF and western blots that co-culture of CXCL-iFibro with CD14+ monocytes promoted the nuclear translocation of β-catenin in fibroblasts (Fig. 5h, i). Moreover, this was concomitant to the up-regulation of ECM-secreting myoFibro markers (Fig. 5j, k). Interestingly, we observed that inhibition of β-catenin/TCF interaction prevented the macrophage-induced switch from CXCL-iFibro into ECM-secreting myoFibro (Fig. 5j, k). Indeed, following inhibition, CXCL-iFibro showed a significant reduction of β-catenin nuclear translocation in fibroblasts and did not experience any increase in either αSMA or SFRP4 staining when cocultured with CD14+ monocytes (Fig. 5i, k), indicating that the WNT/β-catenin pathway was required for the differentiation of CXCL-iFibro into ECM-secreting myoFibro. Finally, because macrophages have been shown to secrete WNT ligands to promote tissue repair in the intestine or kidney, we confirmed that blocking WNT ligand secretion by C-59 (a PORCN inhibitor) reduced β-catenin nuclear translocation and prevented the fibroblast phenotypic switch from CXCL-iFibro into ECM-secreting myoFibro (Fig. 5i, k). Altogether, these data show that macrophages stimulate the differentiation of the CXCL-iFibro into ECM-secreting myoFibro through activation of the WNT/β-catenin pathway.

**Fig. 5: Macrophages induce a switch from CXCL-iFibro to ECM-secreting myoFibro through a WNT/β-catenin dependent pathway.**

Accumulation of CXCL-iFibro at early stage of chronic kidney disease predicts progression and poor prognosis of CKD patients

As the role of CXCL-iFibro and FOLR2+ macrophages is poorly defined in CKD progression, we sought to assess if their presence could predict CKD progression. As CXCL-iFibro represents an intermediate state in the differentiation process into ECM-secreting myoFibro, we hypothesized that accumulation of CXCL-iFibro might be indicative of CKD progression at early phase of the disease. To address this question, we defined a CXCL-iFibro transcriptomic signature based on the differentially expressed genes in the different clusters identified from scRNAseq data (Fig. 1 and Table S10). We confirmed that this signature highlighted CXCL-iFibro in the scRNAseq-based cellular atlas we built (Fig. 6a, b), thereby validating that we could use this gene signature to specifically detect the CXCL-iFibro population in bulk RNAseq data. To evaluate the association of the CXCL-iFibro signature with patient longitudinal outcome, we took advantage of a prospective observational cohort from the Nephrotic Syndrome Study Network (NEPTUNE) with available transcriptomic profiles and longitudinal clinical data^70,71. We selected adult patients with a mild to moderate reduction in kidney function (CKD stage 1 to 3a defined by an eGFR>45 mL/min/1.73m²) at the time of the clinically indicated renal biopsy (N = 134 patients) to determine if the CXCL-iFibro signature could predict patient outcome at an early stage of kidney disease. The patients’ characteristics were summarized in Table 1. Interestingly, we observed that the expression rate of CXCL-iFibro-specific genes was associated with a composite outcome of end-stage renal disease (ESRD), or reduction of baseline estimated glomerular filtration rate (eGFR) of more than 40% (Fig. 6c and Fig. S8a). Optimal stratification of patients for survival analyses was performed using an iterative method. Patients were ranked by their expression of CXCL-iFibro transcriptomic signature and thresholds were defined for each level of expression (Fig. S8b, Table S11). For each threshold, patients were separated into low and high CXCL-iFibro expression scores and a Log-rank test was applied. We selected the threshold, which displayed the most significant p-value and separated 50 patients with low-score from 84 patients with high-score (Fig. S8b, Table S11). Indeed, patients with a high expression score of CXCL-iFibro showed significantly poorer longitudinal outcomes compared to patients with low CXCL-iFibro score (Fig. 6c). Moreover, univariate analysis showed that the eGFR at the time of biopsy, presence of arterial hypertension (HTN) at the time of the biopsy, the urinary protein to creatinine ratio (UPCR), and the CXCL-iFibro signature expression were each significantly associated with poor outcomes (Table 2). Strikingly, CXCL-iFibro expression score was one of the most predictive variables, with a hazard ratio of 2.8. We then performed a multivariate analysis using a Cox regression model integrating eGFR, UPCR, HTN, age and CXCL-iFibro expression score as covariables. We observed that UPCR at the time of biopsy, presence of HTN and high CXCL-iFibro score were independently associated with ESRD or decrease of 40% of eGFR in this cohort of mild-to-moderate CKD patients (Fig. 6d). Finally, the eGFR slope, defined by the change of eGFR per year, was available for 128 patients. We classified patients as fast or slow progressors, based on the level of the eGFR slope (more or less than −5mL/min/1.73m², respectively). We observed that fast progressor patients exhibited a higher CXCL-iFibro expression score than slow progressor patients at the time of kidney biopsy (Fig. 6e). We also defined patients with low- or high-CXCL-iFibro expression score, defined by an expression score below or above the median of the expression score, respectively. We observed that the number of fast-progressor patients was higher in the subgroup with a high CXCL-iFibro score than in the low-CXCL-iFibro score subgroup (23 vs 6 respectively, p = 0.0006, Fig. 6f). Finally, the CXCL-iFibro expression score was anti-correlated with the eGFR slope (Fig. 6g). Based on data presented above, we hypothesized that CXCL-iFibro and FOLR2+ macrophages might be interdependent variables during CKD progression. We thus analyzed the FOLR2 expression level in this same cohort (Fig. 6h–l). We found that FOLR2 expression was associated with poor patient outcome, as observed with the CXCL-iFibro score. Indeed, patients with a high FOLR2 expression level experienced poorer outcomes than those with low FOLR2 expression (Fig. 6h, Fig. S8c, d, and Table S12). Multivariate analysis using a Cox regression model showed that UPCR at biopsy, presence of HTN and FOLR2 expression were associated with poor outcomes in an independent manner (Fig. 6i). As for the CXCL-iFibro expression score, we observed that fast progressor patients exhibited a higher FOLR2 expression than slow progressor patients (Fig. 6j) and that a higher proportion of fast progressor patients exhibited a high FOLR2 expression score (defined by an expression above the median, Fig. 6k). Interestingly, Cox regression model integrating eGFR, UPCR, age, presence of HTN, CXCL-iFibro expression score and FOLR2 expression as covariables, showed that only UPCR at biopsy and the presence of HTN were associated with outcome (Fig. S8e), indicating that FOLR2 and CXCL-iFibro expression scores were not independent. Consistent with this observation, the expression of FOLR2 and of CXCL-iFibro-specific genes was tightly correlated (Fig. 6l). This finding was in agreement with data showing co-staining of SFRP1 (marker of CXCL-iFibro) and FOLR2 protein in patient tissues (as show above Fig. 3g, h). Finally, we wanted to assess if the ability of CXCL-iFibro expression score or FOLR2+ macrophages expression to predict kidney outcome was independent of the expression of other monocytes/macrophages markers. To do so, we evaluated the role on CKD progression of CD14, a wide marker of myeloid cells (monocytes, macrophages, and dendritic cells), and MRC1, a marker of M2-like macrophages, suspected to be involved in CKD progression^72,73. Interestingly, neither CD14 nor MRC1 were associated with poor patient outcome in univariate analysis (p = 0.14 for CD14, p = 0.095 for MRC1). As for MRC1, the p-value was less than 0.1, we added it in the multivariate Cox model. Interestingly, the CXCL-iFibro expression score and FOLR2 expression remained significantly associated with kidney outcomes (Fig. S8f, g). These data suggest that the early identification of CXCL-iFibro or FOLR2+ macrophages could refine the risk prediction of CKD progression, independently of the number of CD14+ cells or MRC1+ cells in the kidney. Altogether, these data show that CXCL-iFibro and FOLR2+ macrophages are interconnected biomarkers in CKD progression and demonstrate that our findings could be relevant in clinical practice.

**Fig. 6: Expression of CXCL-iFibro gene signature predicts poor outcome of early CKD patients.**

Table 1 Clinical and biological data of the 134 patients with mild to moderate CKD

Full size table

Table 2 Results of the univariate analysis of the composite outcome (ESRD or decrease of 40% of eGFR)

Full size table

The Renal Warrior Project. Join Now
Source: https://www.nature.com/articles/s41467-024-44886-z

Tags: Blood, Cancer, Chronic, Chronic Kidney, CKD, Clinical, Diseases, endothelium, hypertension, inflammation, Kidney, Kidney Disease, Kidney Diseases, Patient, Patients, Practice, protein, Renal, Renal Disease, Research, stress, Tested, Treatment

WNT-dependent interaction between inflammatory fibroblasts and FOLR2+ macrophages promotes fibrosis in chronic kidney disease – Nature Communications

Single cell RNA sequencing identifies different clusters of mesenchymal cells in kidney disease

CXCL-iFibro and ECM-myFibro clusters accumulate at early and late CKD stages, respectively

The content in CXCL-iFibro correlates with FOLR2+ macrophage infiltration in kidney disease

CXCL-iFibro attract CD14+ monocytes and induce their differentiation into FOLR2+ macrophages

Macrophages induce the transition from CXCL-iFibro to ECM-secreting myoFibro through the WNT/β-catenin pathway

Accumulation of CXCL-iFibro at early stage of chronic kidney disease predicts progression and poor prognosis of CKD patients

A guide to studying 3D genome structure and dynamics in the kidney – Nature Reviews Nephrology

Q & A: Dr. Laura Bukavina, KCA Trailblazer Award winner – Kidney Cancer Association

The Emotional and Physical Toll of CKD: From Denial to Rebirth – World Kidney Day 2025

Understanding Nephrotic Syndrome: Causes, Symptoms, Diagnosis, and Treatment – Durham Nephrology Associates, PA

Safety problems with continuous glucose monitors or insulin pumps should be reported – The Diabetes Times

The Kidney Cancer Association Announces 2024 Grant Winners – Kidney Cancer Association

It’s Time to Get Shot

Three weeks away from a (virtual) conference you should be at

First patient study investigating apelin as a target for kidney disease shows promising results – Kidney Research UK

Last Chance to Join ISN Committees and Working Groups — Submit by October 18! – International Society of Nephrology