Activated SOX9+ renal epithelial cells promote kidney repair through secreting factors

Abstract A broad spectrum of lethal kidney diseases involves the irreversible destruction of the tubular structures, leading to renal function loss. Following injury, a spectrum of tissue‐resident epithelial stem/progenitor cells are known to be activated and then differentiate into mature renal cells to replace the damaged renal epithelium. Here, however, we reported an alternative way that tissue‐resident cells could be activated to secrete multiple factors to promote organ repair. At single‐cell resolution, we showed that the resident SOX9+ renal epithelial cells (RECs) could expand in the acutely injured kidney of both mouse and human. Compared to other cells, the SOX9+ RECs overexpressed much more secretion related genes, whose functions were linked to kidney repair pathways. We also obtained long‐term, feeder‐free cultured SOX9+ RECs from human urine and analysed their secretory profile at both transcriptional and proteomic levels. Engraftment of cultured human SOX9+ RECs or injection of its conditional medium facilitated the regeneration of renal tubular and glomerular epithelium, probably through stimulating endogenous REC self‐activation and mediating crosstalk with other renal cells. We also identified S100A9 as one of the key factors in the SOX9+ REC secretome. Altogether, the abilities to extensively propagate SOX9+ RECs in culture whilst concomitantly maintaining their intrinsic secretory capacity suggest their future application in cell‐free therapies and regeneration medicine.

injury. 8 Researchers noticed that a rare population of SOX9 positive progenitor cells was represented in the proximal tubule of healthy adult kidneys, 4,9,10 whereas widely stimulated in calbindin-28dk+ distal convoluted tubules 9 and parietal epithelial cells after injury. 11 Coupling genetic lineage tracing studies revealed the identity of activated SOX9+ populations as the critical intrinsic molecular driver of early injury response. [12][13][14][15][16][17][18] For example, Kumar et al. indicated that over 40% of SOX9+ cells re-enter mitosis after ischemic injury, contributing to regenerating functional proximal tubule epithelium. 7 Besides, the findings from Kang et al. 16 and our lab 10 further proved that the descendants of SOX9 positive cells contributed to multiple segments of epithelial repair, including proximal tubule, Henle's loop, distal tubule, collecting duct and the parietal layer of glomerulus. In summary, original researchers thought SOX9+ RECs might be the core of epithelial restoration by dedifferentiation and re-differentiation. 19 However, the exact repair mechanism remains unclear.
Recently, a body of evidence has bolstered that the paracrine mechanism is also responsible for AKI therapy. 8,20,21 The secretome is defined as the complex array of soluble molecules and extracellular vesicles (EVs) that create a microenvironment for cellular renewal. 8,22 These bioactive factors have the therapeutic functions of antifibrotic, 23 angiogenic, antioxidative, mitogenic and anti-apoptotic. 24,25 Previous studies mainly focused on the secretome of mesenchymal stem cells (MSCs). For example, Tögel and colleagues demonstrated that intra-carotid administration of MSCs observably accelerated acute kidney damage recovery 48 h after ischemia, even though no transplanted cells resided in kidney tubules after 24 h. 8,26 The results were congruent with another studying showing that MSC-derived conditional medium (CM) attenuated renal tubular cell apoptosis in a cisplatin-induced kidney injury model. In those experiments, the authors indicated that MSC-derived secreted factors enhanced the migration and proliferation of the proximal tubular epithelium in vivo and in vitro. 27 Besides MSC, adult kidney-derived CD133+ cells were also shown to ameliorate tubular dilation and fibrosis by intravenous injection without being mediated by self-homing to the kidneys. 28 These findings support the idea that the secretome administration plays a critical role in renal protection and repair.
In our study, we identified the endogenous SOX9+ RECs could be activated after damage in mouse and human. Single-cell transcriptional analysis revealed the secreted signature of SOX9+ RECs, whose function contributed to renal restoration. Based on a feederfree culture system, we showed that the long-term, cultured SOX9+ RECs could participate in renal tubular and glomerular epithelium regeneration by secretory function.

| Mouse kidney injury models
All animal experiments were carried out in accordance with Chinese National Guidelines GB/T 35892-20181, as well as under the guidance of the Institutional Animal Care and Use Committee of Tongji University. To establish a unilateral partial nephrectomy (UPN) mouse model, 8-10 weeks C57BL/6 mice were euthanized by intraperitoneal injection of 3.7% chloral hydrate (0.5 g/kg body weight). The left lateral peritoneum was cut to expose the left kidney, and the renal artery was clamped with hemostatic forceps. About one-third of the left kidney was cut off from the upper pole of the kidney using a surgical blade. After cleaning off the blood, the incision was sealed with FuAiLe Medical glue (FAL), and the hemostatic forceps on the renal artery were removed. The muscle layer was closed with sutures (Ethicon, Germany), followed by the closing of the skin. For the unilateral ureteral obstruction (UUO) injury model, the abdomen was opened with a midline incision and the left kidney and upper ureter were exposed. Mice were subjected to surgical cautery of the left ureter 15 mm below the pelvis. Kidney samples were harvested on day 0, day 1, day 10 and day 20 post-surgery for analysis. Sham surgery kidneys in UPN and UUO models were also harvested.
For the unilateral ischemia-reperfusion injury (UIRI) model, 6-8 weeks NOD SCID mice were anaesthetised using 3.7% chloral hydrate (0.5 g/kg body weight), and the mice were put prostrate on a heating pad at a temperature of 37 C for the duration of the surgical procedure. The left renal pedicle was occluded for 30-40 min using a microaneurysm clamp, during which time the kidney was moistened by phosphate-buffered saline (PBS) every 3 min. Then the clamp was removed, and reperfusion was confirmed by observing tissue colour change. The kidney was returned to the abdomen with an intraperitoneal injection of 200 μl penicillin/streptomycin (P/S) to prevent infection. For sham operation, mice had only incisions in the skin and muscle layer, but the renal pedicles were not clamped.
For the Adriamycin (ADR) induced glomeruli injury model, BALB/C mice of 8 weeks were treated with a single dose of ADR (Sangon Biotech), 10.5 mg/kg, by tail vein injection. All mouse injury experiments were performed on male mice and were randomly allocated to the experimental groups.

| Mouse and human renal tissues
For mouse REC cloning, 8-10 weeks mice bred in a specific pathogen free (SPF) facility were collected to obtain renal cortex, medulla and papilla samples. For human kidney tissue sampling, percutaneous renal needle biopsies were performed to obtain patient tissue with membranous nephropathy (MN) by ultrasound-guided core tissue biopsy needles (18 gauge). All renal specimens were subjected to pathological diagnosis. All the human renal tissues were obtained following clinical standard operating procedure (SOP) under the patient's consent and approved by the Hospital Ethics Committee.

| SOX9+ RECs isolation and expansion
For RECs isolation from kidney tissues, samples were washed with cold wash buffer (F12 medium containing 5% fetal bovine serum [FBS] and 1% P/S) and minced by sterile scalpel into 0.2-0.5 mm 3 sizes to a viscous and homogeneous appearance. The minced tissue was then digested with dissociation buffer including DMEM/F12 (Gibco, USA), 2 mg/ml protease XIV (Sigma, USA), 0.01% trypsin (Gibco, USA) and 10 ng/ml DNase I (Sigma, USA) in 37 C incubator 2 h with gentle agitation. Digested cell suspensions were washed with cold-wash buffer and passed through 70 μm Nylon mesh (Falcon, USA) to remove aggregates. Cell pellets were collected by centrifuge of 200g and then seeded onto a feeder layer of lethally irradiated 3T3-J2 cells in modified SCM-6F8 medium. Human SOX9+ RECs were generated from urine samples and expanded as previously described. 29 For better visualization of colony growth, td-Tomato+ RECs derived from mT/mG mice were used. For GFP labelling of cultured SOX9+ RECs, medium containing lentivirus was added to the cell culture together with 10 μg/ml polybrene (1:2000) 29 and cell identities were analysed by CytoFLEX LX flow cytometry with appropriate markers before use (SOX9+, ATP1A1À and CDH1À). Analysis was gated using forward and side scatters characteristics and corresponding positive/negative control. Data were analysed using the flow cytometry software, FlowJo (TreeStar).

| Quantitative real time-PCR (RT-qPCR)
Total RNA was prepared from cultured RECs using a Rneasy Mini kit (QIAGEN) according to the manufacturer's instructions. All RNA was digested with DNase I (Takara, Japan). One micro gram total RNA and PrimeScript RT Master Mix (Takara, Japan) was used for reverse transcription in a SimpliAmp Thermal Cycler (Life Technologies, USA). RT-qPCR was performed in triplicate using a QuantStudio3 Sequence Detection System and SYBR Premix Ex Taq II (Takara, Japan). DNA primer pairs were designed to span exons, when possible, to ensure that the product was from mRNA. The following cycling conditions were used: 1 cycle of 95 C for 30 s, 35 cycles of 95 C for 5 s and 60 C for 34 s. The specificity of the amplified product was evaluated using the melting curve analysis. Internal control glyceraldehyde 3-phosphate dehydrogenase (GAPDH) was used to normalize the result in each reaction, and relative fold change was calculated by the 2 ÀΔΔCt method. The following primer pairs were used:

| Western blotting
Cells were digested and lysed in RIPA buffer (CST) containing protease inhibitors cocktail (Roche) followed by standard Western blotting procedure. 31 To detect the 13-kD S100A9 expression, SOX9+ RECs derived CM were centrifuged through Ultra-15 10K Centrifugal Filter Device (Amicon, Millipore) at 4000g for 45 min at 4 C to collect concentrate whose mass was more 10-kD and then overnight freezedrying. After measuring protein concentration, samples were loaded and separated on 15% precast polyacrylamide gels, and then transferred to PVDF membranes (Roche) at 300 mA for 30 min. Membranes were blocked with 5% no-fat powdered milk, and then incubated with primary antibodies overnight, followed by secondary antibodies. The specific signals were detected by Immobilon Western Chemiluminescent HRP Substrate (Millipore) and Tanon image system.

| Urine sample collection and processing
All steps were performed on ice or at 4 C. Urine samples were col-

| Intercellular crosstalk analysis
To explore potential intercellular crosstalk between SOX9+ RECs and other cell types from healthy control (HC) and AKI samples, we implied the ligand-receptor distribution and expression of SOX9+ RECs and other cell types with a standard pipeline implemented in R using Cell-Chat R package, 33 as previously reported. We chose the receptors and ligands expressed in more than 10% of the cells in the specific cluster for subsequent analysis. The interaction pairs whose ligands belonged to the epidermal growth factor (EGF), vascular endothelial growth factor (VEGF) and Complement families were selected for the evaluation of intercellular crosstalk between the SOX9+ RECs and other cell types.

| Proteomic analysis
Protein extraction and pre-treatment of enzymolysis and desalination at  Next, we tried to dissociate kidney tissues harvested at multiple dps and plated dissociated cells using a feeder-based regenerative cloning (R-Clone) culture system. [35][36][37][38] The activated mouse cells obtained from UPN kidneys, as those from normal kidneys, represented similar clonogenicity ( Figure S1B) and stained positive for SOX9 and PAX2, which are acknowledged markers of nephron progenitors 39 ( Figure 1D). Consistent with the in vivo immunostaining data, the kidney tissue subjected to UPN yielded approximately fivefold more clones, which surged as early as 1 dps ( Figure 1E).
We also examined whether similarly activated RECs can be cloned in other kidney injury models with a different damage mechanism. The UUO injured model could also induce substantial tissue damage in the cortex and medulla regions, 40  and S1C). Therefore, all these data suggested endogenous SOX9+ RECs could be stimulated to expand post different types of renal injury.

| Activation of SOX9+ RECs in injured human kidney
Previous studies by our group and others 9,10,17,41 indicated activation of SOX9+ cells in injured rodent kidneys for tissue repair purposes, however whether a similar mechanism is applied to humans remained unknown. To investigate the existence of tissue resident SOX9+ populations, human kidney sc-RNA-seq datasets (GSE171458; GSE174220) obtained from two healthy controls (HC) and two patients with AKI were analysed. 42 After data pre-processing and stringent quality control, 12 clusters from four different kidney subjects were visualized by UMAP ( Figure S2A,B) and the selected cell lineage-specific marker gene was displayed by dot plot ( Figure S2C).
As expected, we found SOX9+ cells in renal tissues (Figure 2A). There were about 4.217 ± 0.987% cells expressing SOX9 in the HC group, while the percentage of SOX9+ RECs significantly increased up to 22.71 ± 0.478% in the AKI group ( Figure 2B), suggesting the activation of SOX9+ cells in injured human kidneys. Of note, all SOX9+ cells were restricted to epithelial-related clusters. Hereafter, we referred to these SOX9+ cells as RECs. In either healthy or AKI kidneys, both SOX9+ cells were concentrated in multiple segments of tubular cells ( Figure 2C).
Gene Ontology (GO) function analysis showed that the differentially expressed genes of SOX9+ RECs were associated with not only renal development, but also the secretory regulation processes, manifested by genes including CD24, HES1, PAX2, HBEGF and VEGF  Figure 2J). Altogether, the data above showed that the SOX9+ REC could be activated and expanded in injured human kidneys, and secreted multiple factors to facilitate kidney repair.

| Identification of SOX9+ cells in urine from humans of different ages
In previous studies, we reported the single-cell atlas of human urine. 29 Here we focused on the kidney repair-related SOX9+ RECs and F I G U R E 1 Legend on next page. analysed them in the urine samples collected from younger (n = 12, median age = 24) and older (n = 40, median age = 49) healthy people.
In total, we sequenced 1104 cells from the younger age group and 2386 cells from the older age group. After stringent quality controls, we eventually analysed 1010 and 1774 cells, respectively. Our clustering indicated seven clusters under unsupervised graph-based clustering (Seurat method) of the dataset and visualization by UMAP ( Figure 3A).
We annotated the identity of the clusters from two groups based on the established cell type-specific markers. Violin plots for representa-  Figure 3E). Network analysis showed that the secretory granule lumen related genes are closely linked to the wound healing process ( Figure 3F). To further study the secretion related gene expression in SOX9+ REC, we compared our sc-RNA-seq data with a curated human secretome dataset with 2641 known genes. 46 Hierarchical clustering enriched 602 genes totally in our dataset, among which 252 secreted genes were highly expressed in SOX9+ RECs, such as S100A9, MUC1 and NOTCH2. To compare the secretory capability of different groups, we calculated the average expression level of the gene set in each group. We found that SOX9 + RECs demonstrated significantly higher expression for secretome profiles compared to SOX9-cell groups (p < 0.05) in our dataset ( Figure 3G) and others ( Figure S3C). Altogether these data indicated that a rare population of SOX9+ RECs existed in human urine, which showed extraordinary secretory capability.

| Long-term, feeder-free culture of urinederived SOX9+ RECs
Next, we try to culture the human SOX9+ RECs for further studies.
Our previous work described the successful cloning of the SOX9+ RECs from patient renal tissue biopsy using a murine feeder-cell based system (also demonstrated in Figure S4A). 29 Here we further improved the method and successfully cloned the urine-derived SOX9+ RECs in a feeder-free culture system ( Figure 4A), which allowed the xeno-free mass production of cells in the future for therapeutic purposes.
Expanded cells from urine (RECs-urine) shared consistent morphology and proliferative (SOX9+, Ki67+) characteristics with cells from kidney tissue biopsy (RECs-tissue) by immunostaining (Figures 4B and S4A). To further confirm the function mechanism without actual cells, we collected supernatant of cultured RECs as a CM. The CM was allowed to condition for 1 day before it was harvested and centrifuged at 1000g for 15 min to remove any dead cells and cell debris. proximal tubule markers (ATP1A1+/AQP1+), activated more endogenous SOX9+ cells and decreased the serum creatinine levels ( Figure S7A-D).
We also found that CM treatment could significantly enhance the proliferative capability of cultured SOX9+ RECs in vitro ( Figure S7E), suggesting the self-activation of cells by an autocrine mechanism.  Figure 5E and S5C). Consistent with the histological restoration, both the urinary protein and serum creatinine level showed significant decrease in mice receiving CM administration compared with the ADR group ( Figure 5G,H). Taken together, these findings demonstrated that the secretory factors of SOX9+ RECs could simulate endogenous SOX9+ cell self-activation, attributing to the regeneration of renal tubules and glomeruli.

| Quantitative proteome analysis of SOX9+ REC engrafted mice
To identify the secreted proteins involved in renal injury and repair, we applied a proteomic strategy to study the blood serum of SOX9+ REC engrafted mice. We used a multiplexed quantitative proteomics approach following tandem tag-based mass spectrometry, a technology that enables protein identification and quantitation from multiple   Figure 6B). The GO of each cluster was analysed in Figure S8, showing that the Cluster I proteins were closely related to wound healing while Cluster III proteins were related to neutrophil immune response. More specifically, our proteome data allowed us to localize the expression of genes associated with specific pathways and illustrate their interaction relationship by PPI network analysis.
The results indicated that the REC treatment group highly expressed many biological processes component genes of Cluster I, which formed an interaction network about 'Regulation of reactive oxygen species metabolic process', 'Positive regulation of cell proliferation' and 'Positive regulation of wound healing' ( Figure 6C). In contrast, the UIRI group highly expressed genes (Cluster III) which formed an interaction network about 'Inflammatory response', 'Regulation of fibroblast migration' and 'Regulation of cell death', which were all involved in tissues injury status ( Figure 6D). Therefore, we speculated that there were deleterious factors in Cluster III while Cluster I probably contained beneficial proteins for the kidney.
Indeed, among these proteins in clusters, we noticed a few interesting candidates for further analysis. Cystatin-C (CST3) protein in Cluster III is known as a critical biomarker for monitoring renal toxicity or harm in clinical trials, [55][56][57] which was increased in expression in our datasets after UIRI and inversely correlated with recovery post REC engraftment treatment ( Figure 6E). In contrast, S100 calcium-binding F I G U R E 7 Recombinant S100A9 protein ameliorates renal restoration. (A) Representative image of H&E-stained indicated less tubular injury post recombinant S100A9 intervention for 7 days. Sham, with no surgery; UIRI, an equal volume of saline subcutaneous injection since 1 day after UIRI; recombinant S100A9, twice, 10ug/kg recombinant S100A9 protein subcutaneous administration. Scale bar, 50 μm. (B) Representative images of the immunostained whole kidneys after recombinant S100A9 protein treatment. Scale bar, 200 μm. (C) Quantitative scores of tubular necrosis. (D) Quantification of KIM1, ATP1A1, AQP1, SOX9 and Ki67 gene expressions after UIRI or recombinant S100A9 treatment. (E) Serum creatinine level decreased after recombinant S100A9 treatment. Data shown in (C), (D) and (E) were represented as mean ± SD (*p < 0.5, **p < 0.01, ***p < 0.001, n = 3 independent biological samples per group, each group made in duplicate). (F) Assumptive model of SOX9+ RECdriven tubular restoration under injured triggers by secretome. S100A9 protein activates resident SOX9+ RECs proliferation. EGF, VEGF and complement signallings from SOX9+ RECs promote renal repair by acting on the adjacent TECs, VECs and immune cells. The components of SOX9+ REC secretome is represented by colourful graphics. TECs, tubular epithelial cells; VECs, vascular endothelial cells.
protein 9 (S100A9) increased in the UIRI group and continuously increased after REC treatment (>2.2-fold) ( Figure 6E). As mentioned in the above text, we have identified S100A9 overexpression in the sc-RNA-Seq data of urine-derived SOX9+ RECs, which was confirmed here in cultured SOX9+ RECs by RT-qPCR ( Figure 6F). S100A9 protein was also detected in CM from SOX9+ RECs ( Figure 6G). Therefore, it is very likely that SOX9+ RECs could promote kidney repair at least partially by secreting S100A9 protein.
3.7 | Recombinant S100A9 protein promotes renal tissue repair after AKI S100A9 is a member of the S100 calcium-binding protein family. S100A9 deficiency displayed a phenotype with enhanced renal damage, revealing that S100A9 could contribute to kidney repair. 58 To further predict the role of S100A9 in AKI, we monitored the reduction in renal function following renal UIRI and the effects of the human recombinant S100A9 pro- Moreover, we observed that a few SOX9+ RECs appeared after UIRI and the number of SOX9+ RECs was significantly increased after the recombinant S100A9 protein treatment. Interestingly, after S100A9 protein administration, the kidney showed a high proliferation ratio (10% Ki67+) in vivo, which could be related to the regeneration of the proximal tubular epithelium (ATP1A1+ and AQP1+) ( Figures 7B,C and S9B). We also validated the pro-proliferative effect of S100A9 on SOX9+ RECs in vitro ( Figure S9E). Mice injected with exogenous S100A9 protein showed a significant reduction in serum creatinine level compared with the UIRI group, suggesting that exogenous S100A9 can restore renal injury in both morphology and function ( Figures 7E and S9D). Taken together, the data showed that S100A9 is one of the bioactive factors, which could trigger resident SOX9+ REC proliferation. And other components in the secretome could also participate in regeneration, partially by activation of EGF, VEGF and Complement signal pathways. (summarized in Figure 7F).

| DISCUSSION
In the current study, we noted that the endogenous SOX9+ RECs could expand in the acutely damaged kidney of both mice and human.
At the single-cell resolution, the SOX9+ RECs displayed the dominantly secreted signature, which was closely linked to kidney regeneration pathways. Based on a feeder-cell-independent regenerative cloning (R-Clone) system, we successfully obtained the cultured SOX9+ RECs from human urine, maintaining stable cell characteristics and secretory function. Transplantation of cultured human SOX9+ RECs or administration with its CM ameliorated renal tubular and glomerular epithelium damage. We also confirmed S100A9 as one of the critical factors in the SOX9+ REC secretome by quantitative proteome analysis and functional studies.
The isolation and expansion system of epithelial cells from adult tissues are fundamental techniques supporting cell-based regeneration studies. In 1975, Green et al. established the first successful example of adult human epidermal stem cell culture. 35 Our previous studies have obtained SOX9+/Ki67+ populations from human urine with massive animal-derived components, relying on a mouse fibroblast feeder layer. 29 Here we improved the method and cultured SOX9+ RECs in a novel feeder-free system. Longtermed, expanded SOX9+ RECs shared similar characteristics compared to those developed in the 3T3 feeder-cell-based system. 29 This method supports the rapid acquisition of a large number of primitive cells and clinical applications for regenerative medicine with xeno-free conditions. 59 In recent decades, the mechanism of kidney regeneration in mammalians has been widely studied, and SOX9+ cells are found to be stimulated to participate in this process. 4 The SOX9-mediated mechanisms of repair are complex. 19 Previous researchers published by our group 10 and others 9,17 indicated that the proportion of SOX9+ progenitors was extensively increased after UIRI or partial nephrectomy, whose descendants contributed to multi-segment nephron restoration, including proximal tubule, loop of Henle, distal tubule, collecting duct and the parietal layer of glomerulus. Similarly, our recent report showed that cultured human SOX9+ RECs could be in situ transplanted to generate functional proximal epithelium and early distal tubules. 29 Altogether these works supported the pattern of SOX9+ RECs adopted to regenerate the renal epithelium by differentiation. In this article, we found SOX9+ RECs derived CM could replicate effective therapy of live cell transplantation, implying another pattern of renal restoration via secretory function. 5 Moreover, here we clarified the potential molecular mechanism underlying the SOX9+ REC secretome-mediated kidney repair by quantitative proteomic analysis. The experiments showed upregulation of several renoprotective factors, including S100A9. In a study by Mark C. Dessing et al., they observed that S100A9 deficiency led to sustained renal pathological state and dysfunction following renal UIRI, 58 suggesting that S100A9 could lead to an advantageous phenotype in renal injury and repair in contrast to their role in cerebral injury and repair. 60 To elucidate the benefits of S100A9 protein treatment, we performed the S100A9 protein administration in ischemic mice by I.V. injection. Our findings described that soluble S100A9 protein could replicate part of the regenerative effects of the full secretome, suggesting that a high level of S100A9 in the circulating environment contributes to renal homeostasis. 61 Of note, the limitation of the study is that SOX9+ RECs secretome does not only contain soluble proteins, but also EVs. It would be sensible for future studies to include an EV administration study to investigate the therapeutic effects of renal damage. Moreover, previous studies reported many other types of adult stem/progenitor cells which contribute to kidney repair by cell replacement, 48,62-67 it would be interesting to know whether they have a similar pro-regeneration secretory function like SOX9+ RECs.
In conclusion, our study clarified the secretory function of SOX9+ RECs to regenerate renal epithelium, which elucidated a potential organ/tissue repair mechanism and provided novel translational opportunities for cell-free strategies.

CONFLICT OF INTEREST
All authors declare that they have no conflict of interest.

DATA AVAILABILITY STATEMENT
The data that support the findings of this study are available from the corresponding author upon reasonable request.