Research ArticleFibrosis

Inhibition of hyperglycolysis in mesothelial cells prevents peritoneal fibrosis

See allHide authors and affiliations

Science Translational Medicine  05 Jun 2019:
Vol. 11, Issue 495, eaav5341
DOI: 10.1126/scitranslmed.aav5341

Preventing peritoneal fibrosis

Peritoneal dialysis is used to treat patients with end-stage renal disease, but many of these patients develop peritoneal fibrosis that limits treatment efficacy. Si et al. studied cells from patients undergoing peritoneal dialysis and used a mouse model to understand the mechanism underlying progression of peritoneal fibrosis. They found that transforming growth factor β1 stimulated hyperglycolysis in mesothelial cells, contributing to a mesothelial-to-mesenchymal transition phenotype and peritoneal fibrosis. Manipulating the expression of three microRNAs using adeno-associated viruses could inhibit fibrosis in the mouse model, suggesting that correcting the altered metabolic state in mesothelial cells could be therapeutic for peritoneal fibrosis.

Abstract

Progressive peritoneal fibrosis affects patients receiving peritoneal dialysis (PD) and has no reliable treatment. The mechanisms that initiate and sustain peritoneal fibrosis remain incompletely elucidated. To overcome these problems, we developed a strategy that prevents peritoneal fibrosis by suppressing PD-stimulated mesothelial-to-mesenchymal transition (MMT). We evaluated single-cell transcriptomes of mesothelial cells obtained from normal peritoneal biopsy and effluent from PD-treated patients. In cells undergoing MMT, we found cellular heterogeneity and intermediate transition states associated with up-regulation of enzymes involved in glycolysis. The expression of glycolytic enzymes was correlated with the development of MMT. Using gene expression profiling and metabolomics analyses, we confirmed that PD fluid induces metabolic reprogramming, characterized as hyperglycolysis, in mouse peritoneum. We found that transforming growth factor β1 (TGF-β1) can substitute for PD fluid to stimulate hyperglycolysis, suppressing mitochondrial respiration in mesothelial cells. Blockade of hyperglycolysis with 2-deoxyglucose (2-DG) inhibited TGF-β1–induced profibrotic cellular phenotype and peritoneal fibrosis in mice. We developed a triad of adeno-associated viruses that overexpressed microRNA-26a and microRNA-200a while inhibiting microRNA-21a to target hyperglycolysis and fibrotic signaling. Intraperitoneal injection of the viral triad inhibited the development of peritoneal fibrosis induced by PD fluid in mice. We conclude that hyperglycolysis is responsible for MMT and peritoneal fibrogenesis, and this aberrant metabolic state can be corrected by modulating microRNAs in the peritoneum. These results could provide a therapeutic strategy to combat peritoneal fibrosis.

INTRODUCTION

Nearly 15% of patients with end-stage renal disease (ESRD) are treated with peritoneal dialysis (PD) and achieve outcomes similar to those of hemodialysis patients as long as the dialysis achieved by the peritoneal membrane is adequate (1). Unfortunately, peritoneal damage and loss of dialysis function can occur. It is suspected that the glucose-based dialysate is responsible for the loss of mesothelial cells, the development of inflammation and angiogenesis, and the presence of fibrosis in peritoneal membranes (2). These responses are relevant because peritoneal fibrosis leads to loss of peritoneal function and affects about 50% of patients undergoing PD (3). Methods that successfully prevent and treat progressive peritoneal fibrosis are lacking (4).

During the development of peritoneal fibrosis, a particular form of cell transition in mesothelial cells that is comparable to epithelial-to-mesenchymal transition (EMT) called mesothelial-to-mesenchymal transition (MMT) can be detected (5). The process of MMT enables mesothelial cells to proliferate and migrate more efficiently as part of mesothelial repair (6); however, excessive MMT can also suppress the expression of adhesive and tight junction proteins, leading to mesothelial cell denudation that is aggravated by processes occurring during PD (5, 7). The loss of mesothelial cells exposes the basement membrane, stimulating fibrogenesis through activation of resident fibroblasts, and recruits inflammatory cells. MMT also produces myofibroblast-like cells, which may invade the underlying mesenchyme to promote fibrogenesis (8). It has been proposed that mesothelial cells arising from MMT are the major source of myofibroblasts (9). However, lineage-tracing experiments performed in mice suggest that most activated myofibroblasts in fibrotic peritoneum originated from resident fibroblasts rather than mesothelial cells (10). Thus, the source of myofibroblasts arising during peritoneal fibrosis remains under debate.

It has been suggested that organ fibrosis is a reactive response to damage of parenchymal cells, such as epithelial cells (11). Alternatively, the activation and proliferation of resident fibroblasts in the peritoneal membrane could be a reactive process that occurs in response to mesothelial cell damage. Autocrine/paracrine factors such as transforming growth factor–β1 (TGF-β1), released by mesothelial cells undergoing MMT or inflammatory cells, can activate resident fibroblasts as well (12). Given this information, we hypothesized that targeting the molecular alterations in damaged mesothelial cells might be a more appropriate strategy to combat PD-related peritoneal fibrosis than targeting fibroblast activation.

One such molecular alteration in mesothelial cells could be metabolic reprogramming, which describes how cellular energy metabolism changes in response to nutrient deprivation, tissue damage, and/or increase in growth factors or cytokines. For example, in chronic kidney disease, renal tubular cells switch from fatty acid oxidation to more glycolytic metabolism, and preventing this metabolic reprogramming suppresses renal tubulointerstitial fibrosis (13, 14). During PD, there is a substantial increase in nutrient supplementation along with peritoneum damage and increased production of cytokines (15). These factors could cause metabolic reprogramming in peritoneal mesothelial cells. The metabolic state of peritoneal mesothelial cells in vivo, particularly in patients undergoing PD, has not been investigated.

Here, we used single-cell RNA sequencing (scRNA-seq) analysis to characterize the transcriptome of peritoneal cells obtained from normal peritoneum and from the dialysates of patients undergoing PD. We demonstrate that mesothelial cells develop hyperglycolysis, a metabolic alteration that is correlated with the development of MMT. Using a mouse model, we confirmed that blocking hyperglycolysis by 2-deoxyglucose (2-DG) or a triad of adeno-associated viruses (AAVs) encoding three microRNAs (miRNAs) prevented the development of MMT and peritoneal fibrosis.

RESULTS

scRNA-seq identifies hyperglycolysis in mesothelial cells from patients undergoing PD

To define the metabolic landscape of human mesothelial cells in response to PD, we analyzed 96,446 single-cell transcriptomes including cells that dissociated from normal peritoneum (hernia surgery, n = 3) and peritoneal cells from effluent of short-term patients undergoing PD (PD less than 2 weeks, n = 6) and from long-term patients undergoing PD (PD more than 6 years, n = 4; fig. S1A). We first cataloged peritoneal cell types of normal peritoneum in an unbiased manner and assigned peritoneal cells to seven distinct cell clusters (Fig. 1A and fig. S1B). We generated cluster-specific marker genes by performing differential gene expression analysis and identified these clusters by matching cluster-specific marker genes with established cell markers (Fig. 1B and fig. S1C). Mesothelial cells were identified as the major component of peritoneum (61.5%), according to the previously reported markers for mesothelial cells. For example, cytokeratins (5, 16, 17), intercellular adhesion molecule 1 (ICAM1) (5), calbindin 2 (CALB2) (18), Wilms tumor 1 (WT1) (19), podoplanin (PDPN), mesothelin (MSLN) (16), uroplakin 3B (UPK3B), leucine-rich repeat neuronal 4 (LRRN4) (20), and glycoprotein M6A (GPM6A) (17) were all predominately expressed in the mesothelial cluster compared with other clusters (fig. S2A). In addition, we found that intelectin 1 (ITLN1), kallikrein-related peptidase 11 (KLK11), and small proline-rich protein 2F (SPRR2F) were also specifically expressed by the mesothelial cell cluster (fig. S2B). On the basis of these cell identities in normal peritoneal cells, we then cataloged effluent-derived peritoneal cells from patients undergoing PD. A total of 77,217 cells (51,965 of short-term PD and 25,252 of long-term PD) were classified into five groups: (i) peritoneal cells (clusters 1 and 2), (ii) mononuclear phagocytes (clusters 3 and 4), (iii) T lymphocytes (clusters 5, 6, and 7), (iv) B lymphocytes (cluster 8), and (v) unclassified cells (cluster 9) (Fig. 1, C and D). We further separated peritoneal cells into mesothelial cells (96.81%, using UPK3B and cytokeratins as identities), myofibroblasts (1.04%, coexpressing COL1A1, α-SMA, and PDGFRβ), and vascular endothelial cells (2.15%, KDR- and PECAM1-positive cells) (fig. S3, A and B). To understand which cell populations are able to produce extracellular matrix, we calculated the fraction of COL1A1+ and COL4A1+ cells of all clusters and found that more than 87% of collagen-producing cells were located in the mesothelial cell population (Fig. 1E). The expression of mesenchymal markers increased in long-term PD (cluster 2) compared with short-term PD (cluster 1), whereas the expression of CDH1 (E-cadherin), an epithelial signature gene, was reduced in long-term PD group (Fig. 1F).

Fig. 1 Identification of cell populations in normal human peritoneum and effluent of patients undergoing PD.

(A) Unsupervised scRNA-seq analysis distinguishing seven different cell clusters in normal human peritoneum tissue presented by t-distributed stochastic neighbor embedding (t-SNE) plot (n = 3). (B) Violin plots demonstrating the expression of representative marker genes across the seven cell clusters. The y axis shows the log2 scale–normalized read count. (C) Unsupervised scRNA-seq analysis distinguishing nine different cell clusters in effluent-derived peritoneal cells from patients undergoing PD (t-SNE) (n = 10). Clusters and corresponding cell counts are listed in the table below the map. (D) Heat map showing expression of canonical marker genes in the cell clusters. Color indicates the gene expression value scaled by Z score. (E) t-SNE maps demonstrating the expression of extracellular matrix proteins. The color gradient is based on the log2-transformed read count. (F) t-SNE maps demonstrating expression of mesenchymal marker genes and epithelial marker genes in mesothelial cells from the long-term PD and the short-term PD groups. The color gradient is based on the log2-transformed read count. Jitter plots on the bottom panel of each t-SNE plot show the quantification results based on normalized read counts. *P < 0.05 and ***P < 0.001 by Kruskal-Wallis test.

To identify the gene expression signature of mesothelial cells, we performed gene set enrichment analysis (GSEA) on single-cell transcriptomes of normal mesothelial cells and mesothelial cells from long-term patients undergoing PD. As shown in Fig. 2A, the genes changed by long-term PD were related to EMT and TGF-β signaling, confirming that EMT plays an indispensable role in PD-related peritoneal fibrosis. We also noted that glycolysis was one of the most enriched gene sets associated with PD (Fig. 2B and fig. S4A), whereas in the seven nonperitoneal cell clusters, there was no significant (P > 0.06) enrichment of glycolysis (fig. S4B). GSEA between normal and short-term PD mesothelial cells (fig. S4C) identified marginally significant (P = 0.053) enrichment of glycolysis. When analyzing the key glycolysis enzymes in the mesothelial population, we found that the expression of hexokinase 1 (HK1), HK2, phosphofructokinase (PFK), and fructose-2, 6-biphosphatase 3 (PFKFB3) markedly increased in long-term PD compared with short-term PD or normal mesothelial cells (Fig. 2C). These data suggested that PD-induced hyperglycolysis was associated with MMT in mesothelial cells. To confirm the correlation between MMT and hyperglycolysis, we performed a Spearman’s correlation analysis using the transcriptomic data in normal, short-term PD, and long-term PD mesothelial cells. We found that the expression of key glycolysis enzymes was correlated with the expression of EMT markers including COL1A1, FN1, and α-SMA (Fig. 2D).

Fig. 2 scRNA-seq reveals enhanced glycolytic enzyme expression associated with the development of MMT.

(A and B) GSEA plots demonstrating enrichment score (ES) of gene sets in the scRNA-seq data of mesothelial cells. Genes in each gene set are ranked by signal-to-noise ratio according to their differential expression between normal mesothelial cells and the long-term PD group. FDR, false discovery rate. (C) Heat map showing the expression of key glycolytic enzymes in mesothelial cells of normal peritoneum, short-term PD, and long-term PD groups. (D) Scatter plots of the correlation between expression of glycolytic enzymes and mesenchymal marker genes. Spearman’s correlation analysis was performed, with ρ (Spearman’s correlation coefficient) and P values labeled in the plots. (E) Ordering of scRNA-seq expression data of peritoneal mesothelial cells according to the pseudotime produced by Monocle analysis. (F) Expression of mesenchymal marker genes, epithelial marker genes, and glycolytic enzyme–encoding genes demonstrated along the trajectory that depicts MMT. (G and H) Immunoblots and quantification of the expression of glycolytic enzymes (G) and fibrotic proteins (H) in mesothelial cells sorted by fluorescence-activated cell sorting using UPK3B antibody. Data are presented as means ± SEM; *P < 0.05, **P < 0.01, or ***P < 0.001 versus normal cells and #P < 0.05, ##P < 0.01, or ###P < 0.001 versus short-term PD by one-way analysis of variance (ANOVA) with Bonferroni’s multiple comparison test or Kruskal-Wallis test, n = 6 in each group.

EMT is a dynamic but asynchronous process. To demonstrate cell states during this transition, we performed cell trajectory analysis. Mesothelial cells were ordered along the pseudotime axis on the basis of the top 1000 differentially expressed genes among normal, short-term PD, and long-term PD groups. This trajectory revealed that mesothelial cells underwent transition from epithelial-like to completely mesenchymal states, passing through intermediate hybrid states. In addition, the transition from epithelial to mesenchymal states branched off into several different mesenchymal cell fates, demonstrating the extent of transcriptional heterogeneity in mesothelial cells undergoing EMT (Fig. 2E). Gain of mesenchymal markers coincided with an increase in expression of glycolytic enzymes (Fig. 2F).

After we validated the specificity of UPK3B as a mesothelial marker (fig. S4D), we isolated mesothelial cells from normal peritoneum and from dialysate effluent of short-term and long-term patients undergoing PD. Immunoblot results demonstrated that the expression of key glycolytic enzymes increased in mesothelial cells from patients with long-term PD compared to normal mesothelial cells (Fig. 2G). The increase in these enzymes was associated with enhanced expression of mesenchymal markers and a decrease in E-cadherin protein expression (Fig. 2H). In summary, our single-cell transcriptomic data demonstrated that the development of MMT is associated with hyperglycolysis, supporting that hyperglycolysis is involved in peritoneal fibrogenesis.

PD fluid enhances glycolysis in mouse peritoneum

We created a mouse model of peritoneal fibrosis by daily intraperitoneal injection of Dianeal PD fluid containing 4.25% glucose for 6 weeks. The peritoneum in PD fluid–treated mice thickened significantly (P < 0.001) and showed increased submesothelial collagen deposition compared to that of saline-injected control mice (Fig. 3A and fig. S5A). Using real-time quantitative polymerase chain reaction (RT-qPCR) and immunoblot analysis, we confirmed that the expression of FN1, COL1A1, and α-SMA increased in peritoneal tissues from mice treated with PD fluid (fig. S5, B and C). PD fluid suppressed the expression of epithelial marker E-cadherin (fig. S5, C and D), indicating the occurrence of MMT. We also assessed peritoneum function using a modified peritoneal equilibration test and found an increase in peritoneal permeability of glucose and blood urea nitrogen (fig. S5E).

Fig. 3 PD fluid induces fibrosis and metabolic reprogramming of the peritoneum in mice.

(A) Masson’s trichrome staining of peritoneal tissue from mice treated with saline or PD fluid. Representative images derived from six animals per group. Scale bar, 100 μm. (B) Heat map of PCR-based mRNA array showing the significantly differentially expressed genes involved in glucose metabolism (P < 0.05 by t test; fold change, >2). (C) KEGG pathway enrichment analysis of the genes significantly differentially expressed due to PD fluid in mouse peritoneum. (D) Real-time PCR analysis of glycolytic enzymes in the peritoneum of mice treated with PD fluid or saline. Data are presented as means ± SEM; *P < 0.05 and **P < 0.01 by t test, n = 6). (E) Immunoblots and quantification of glycolytic enzymes in peritoneum of mice treated with PD fluid or saline. Data are presented as means ± SEM; **P < 0.01 and ***P < 0.001 by t test, n = 6). (F) Volcano plot of metabolomics analysis (PD fluid versus saline, n = 6 in each group). (G) Statistical analysis of metabolites in the peritoneum from PD fluid–treated mice compared to the control group. Data are presented as means ± SEM; *P < 0.05, **P < 0.01, and ***P < 0.001 by t test, n = 6. Right: Pathway map showing up-regulated enzymes and metabolites in red color. G6P, glucose 6-phosphate; F1,6BP, fructose 1,6-bisphosphate; DHAP, dihydroxyacetone phosphate; G3P, glyceraldehyde 3-phosphate; R5P, ribose 5-phosphate; IMP, inosine monophosphate; AMP, adenosine monophosphate; UMP, uridine monophosphate; CMP, cytidine monophosphate; GMP, guanosine monophosphate; Leu, leucine; Phe, phenylalanine; Tyr, tyrosine; GPI, glucose-6-phosphate isomerase; ALDO, fructose-bisphosphate aldolase; TPI, triosephosphate isomerase. (H) Representative micrographs showing mitochondrial morphologic changes in response to PD fluid. Scale bars, 10 μm. Blue, nuclei. Red, cytokeratin 5. Green, mitochondria. Boxed regions areas are presented at higher magnification on the right. Scale bars, 5 μm.

To investigate whether energy metabolism is altered in the peritoneum after exposure to PD fluid, we used a PCR-based messenger RNA (mRNA) array to determine the expression of genes involved in energy metabolism. Six weeks of PD fluid treatment caused significant changes (P < 0.05) of expression in 17.8% of the tested genes (Fig. 3B). A Kyoto encyclopedia of genes and genomes (KEGG) pathway analysis of those genes indicated that most of them were involved in the glycolysis pathway (Fig. 3C). We confirmed that the expression of key enzymes in the glycolysis pathway, including HK1, HK2, PFK, and PFKFB3, increased in the peritoneum of mice treated with PD fluid (Fig. 3, D and E). These results suggested that PD fluid induced hyperglycolysis in the peritoneum. To confirm this conclusion, we assayed stable intermediate metabolites using liquid chromatography–mass spectrometry. Metabolic profiling analysis showed that PD fluid induced distinct, differential changes in cellular metabolites (Fig. 3F). Intermediate metabolites of glycolysis, the pentose phosphate pathway, and the nucleotide synthesis pathway were significantly (P < 0.05) elevated (Fig. 3G and fig. S5F), indicating a hyperglycolytic and proliferative state in PD fluid–treated peritoneum. In addition, cellular amino acids were also elevated, suggesting a high rate of protein metabolism.

Because metabolic reprogramming is often associated with dynamic changes in mitochondria, we examined the mitochondrial morphology of peritoneal mesothelial cells from PhaMexcised mice, which express mitochondrial-specific Dendra2 green fluorescent protein (GFP) (21). We found that the mitochondria of peritoneal mesothelial cells from saline-treated mice had a coarse particle morphology with clustered distribution. Mitochondria in PD fluid–treated mice shifted to a punctate morphology with sparse distribution (Fig. 3H), suggesting that mitochondrial fission occurred during PD fluid–induced metabolic reprogramming. These results demonstrate that PD fluid causes hyperglycolysis in the peritoneum, which is accompanied by dynamic mitochondrial changes.

TGF-β1 induces hyperglycolysis and profibrotic phenotype in human mesothelial cells

High-glucose and glucose degradation products of PD fluid can activate TGF-β1 signaling, which is an important event in the initiation of peritoneal fibrogenesis. Using RT-qPCR, we confirmed that TGF-β1 mRNA increased in fibrotic peritoneal tissue of mice treated with PD fluid (fig. S6A). This induction of TGF-β1 was associated with activation of Smad2/3 (fig. S6B). Because TGF-β1 reportedly induces metabolic changes in many types of cells (14, 22), we presumed that TGF-β1 was a major factor inducing metabolic reprogramming in the peritoneum. We tested this hypothesis first in a human mesothelial cell line, MeT-5A cells. A PCR-based mRNA array in MeT-5A cells treated with TGF-β1 for 48 hours showed that the expression of ~10% of the tested genes significantly (P < 0.05) changed. In contrast, only ~1% of tested genes significantly (P < 0.05) changed in response to PD fluid (Fig. 4A). KEGG pathway analysis indicated that most of the differentially expressed genes were involved in the glycolysis pathway (Fig. 4B). Using RT-qPCR, we confirmed that the expression of key enzymes in the glycolysis pathway, including HK1, HK2, PFK, and PFKFB3, was up-regulated by TGF-β1 (Fig. 4C).

Fig. 4 TGF-β1 stimulates hyperglycolytic metabolism in human mesothelial cells.

(A) Volcano plots showing the results of PCR-based mRNA array from MeT-5A cells treated with TGF-β1 (5 ng/ml; left panel) or medium containing PD fluid (right panel), n = 3 in each group. (B) Heat map showing the significant gene expression changes for genes involved in energy metabolism (P < 0.05 by t test; fold change, >2) in MeT-5A cells treated with or without TGF-β1. The right panel shows the KEGG pathway enrichment analysis of these significantly changed genes. CTL, control group. AA, amino acids. (C) RT-PCR analysis of the expressions of glycolytic enzymes in MeT-5A cells treated with or without TGF-β1 (means ± SEM; *P < 0.05 and ***P < 0.001 by t test, n = 6). (D) Assessment of cellular respiration with Mito Stress assay in MeT-5A cells treated with or without TGF-β1. Results are normalized to protein expression; data are presented as means ± SEM. (E and F) Quantification of measurements in (D). **P < 0.01 and ***P < 0.001 by t test, n = 6 in each group. (G) Glycolysis stress test in MeT-5A cells treated with or without TGF-β1. Results are normalized to protein expression; data are presented as means ± SEM. (H and I) Quantification of measurement in (G). ***P < 0.001 by t test, n = 6. (J) Immunoblots and quantification of the expression of fibrotic proteins and E-cadherin in MeT-5A cells treated with or without TGF-β1 (means ± SEM; *P < 0.05 and **P < 0.01 by t test). (K) Cellular respiration of human primary mesothelial cells assessed by Mito Stress test. Results are normalized to protein expression and are presented as means ± SEM, n = 7. (L) Quantification analysis of measurement in (K), ***P < 0.001 by t test, n = 7. (M) Immunoblots and quantification of the expression of FN1, COL1A1, α-SMA, and E-cadherin in human primary mesothelial cells treated with or without TGF-β1 (means ± SEM; *P < 0.05 and **P < 0.01 by t test, n = 6).

We further examined cellular respiration based on the extracellular acidification rate (ECAR; a surrogate for glycolytic rate) and oxygen consumption rate (OCR; a representation of mitochondrial activity) in MeT-5A cells that were treated with or without TGF-β1 for 48 hours. Treatment with TGF-β1 increased ECAR (Fig. 4, D and E) and reduced OCR (fig. S6C and Fig. 4F), supporting our hypothesis that TGF-β1 induces a glycolytic phenotype in mesothelial cells. Moreover, we performed a glycolysis stress test by sequential addition of glucose, oligomycin, and 2-DG (a glucose analog that inhibits glycolysis) in MeT-5A cells treated with or without TGF-β1 (Fig. 4G). After glucose injection, TGF-β1–treated cells exhibited higher ECAR compared to control cells (Fig. 4H), confirming that TGF-β1 stimulates glycolytic metabolism in mesothelial cells. After the addition of oligomycin, which blocks mitochondrial adenosine triphosphate (ATP) generation, ECAR significantly (P < 0.001) increased in TGF-β1–treated cells compared with controls (Fig. 4I), indicating that TGF-β1 also increases glycolytic capacity, which can compensate for the loss of mitochondrial ATP generation. After these changes in metabolism, the expression of fibrotic proteins was up-regulated in TGF-β1–treated cells (Fig. 4J). Expression of E-cadherin was significantly suppressed by TGF-β1 (P < 0.05), suggesting that TGF-β1 induced EMT in these cells.

Because MeT-5A cells show genetic similarities to malignant mesothelioma cells, which may exhibit high glycolysis rate, we tested our conclusion in primary cultured human mesothelial cells. Compared with the results from MeT-5A cells, primary cultures had a lower basal ECAR rate, indicating lower glycolysis. In response to TGF-β1, the primary cultures exhibited an increase in ECAR similar to the increase achieved in MeT-5A cells (Fig. 4, K and L). This enhanced glycolysis was associated with increased expression of fibrotic proteins (Fig. 4M). Together, these results indicate that TGF-β1 is a potent factor for inducing hyperglycolysis in mesothelial cells, reinforcing that hyperglycolysis is involved in the development of peritoneal fibrosis.

Blocking hyperglycolysis suppresses MMT and fibrogenesis in mice

To investigate whether hyperglycolysis mediates the TGF-β1–induced fibrotic phenotype in mesothelial cells, we determined whether blocking hyperglycolysis suppressed the expression of fibrotic genes stimulated by TGF-β1. Incubating MeT-5A cells with 0.2 mM 2-DG for 1 hour before treatment with TGF-β1 blocked increases in glycolysis and glycolytic capacity (Fig. 5A). As expected, the mRNA and protein expression of profibrotic genes including FN1, COL1A1 and α-SMA were also significantly (P < 0.01) repressed by 2-DG (Fig. 5, B and C). We also examined the effect of 2-DG in primary cultured human mesothelial cells and found that, similar to the results obtained using MeT-5A cells, 2-DG blocked the induction of both glycolysis and glycolytic capacity stimulated by TGF-β1 (Fig. 5D). The expression of fibrotic proteins induced by TGF-β1 was also suppressed by 2-DG treatment in these cells (Fig. 5E).

Fig. 5 Inhibition of glycolysis attenuates the profibrotic phenotype of MeT-5A cells and peritoneal fibrosis in mice.

(A) ECAR measurements of glycolysis stress test in MeT-5A cells treated with TGF-β1 and 2-DG (0.2 mM) or TGF-β1 (5 ng/ml) alone. Data are presented as means ± SEM. Right: Quantification analysis of basal ECAR, glycolysis, and glycolytic capacity. (B and C) RT-qPCR analysis and immunoblots of FN1, COL1A1, and α-SMA in MeT-5A cells with or without 2-DG treatment followed by TGF-β1 treatment. Quantitative analysis results are demonstrated as bar graphs. (D) ECAR measurements of glycolytic stress test in human primary mesothelial cells. Right: Quantification analysis of basal ECAR, glycolysis, and glycolytic capacity. (E) Representative immunoblots of human primary mesothelial cells. Quantitative analysis results are presented as a bar graph. Data (A to E) are presented as means ± SEM (*P < 0.05, **P < 0.01, or ***P < 0.001 for TGF-β1 versus CTL and #P < 0.05, ##P < 0.01, or ###P < 0.001 for TGF-β1 plus 2-DG versus TGF-β1, one-way ANOVA with Bonferroni’s multiple comparison test, n = 6 to 8). (F) Masson’s trichrome staining of peritoneal tissue from mice treated with or without 2-DG and PD fluid. Representative images derived from five animals per group. Scale bar, 100 μm. (G) Quantitative analysis of the thickness of the peritoneal submesothelial compact zone. (H) Immunoblots and quantification of protein expression of FN1, COL1A1, and α-SMA in peritoneal tissue from mice treated with PD fluid and 2-DG. Data (G and H) are presented as means ± SEM (**P < 0.01 or ***P < 0.001 versus saline; #P < 0.05 or ###P < 0.001 versus PD fluid, one-way ANOVA with Bonferroni’s multiple comparison test, n = 5).

We next examined the effect of 2-DG treatment in vivo. We intraperitoneally injected mice with PD fluid daily for 6 weeks. At the start of the fifth week, mice were treated with 2-DG (80 mg/kg) in addition to PD fluid for the two remaining weeks. Histology examination revealed that the PD fluid–induced increase in peritoneum thickness was reduced in mice treated with 2-DG (Fig. 5, F and G). Immunoblot indicated that the expressions of FN1, COL1A1, and α-SMA were suppressed by 2-DG (Fig. 5H).

Our results demonstrated that hyperglycolysis is a critical factor leading to mesothelial EMT and peritoneal fibrosis and that inhibition of glycolysis could attenuate the development of peritoneal fibrosis caused by PD fluid. However, because of the concern that permanent in vivo inhibition of the rate-limiting glycolysis enzyme by 2-DG treatment would cause detrimental effects (23), we developed an miRNA-based therapeutic approach to target the key enzymes of glycolysis.

Identification of miRNAs involved in glycolysis during peritoneal fibrogenesis

To identify miRNAs potentially involved in the regulation of glycolysis pathways during peritoneal fibrosis, we performed miRNA microarray analysis in fibrotic peritoneum from mice treated with PD fluid. Compared with results from saline-treated control mice, we found that 41 miRNAs had significantly (P < 0.05) altered expression in PD fluid–treated peritoneum, of which 23 miRNAs were up-regulated and 18 were down-regulated (Fig. 6A). Heat map and one-way hierarchical clustering analysis revealed that miRNAs in PD fluid–treated and saline-treated groups were classified into two distinct clusters, indicating differences in miRNA expression signatures (Fig. 6B). We collected predicted target genes of the differentially expressed miRNAs using TargetScan (www.targetscan.org/vert_72) and miRDB (www.mirdb.org/). Nineteen miRNAs were involved in the process of fibrosis, of which eight targeted key enzymes in glucose metabolism (Fig. 6C). For example, miR-200a and miR-26a target PFK and PFKFB3, respectively, two critical enzymes in the glycolysis pathway. In addition, miR-21a was predicted to target pyruvate dehydrogenase phosphatase catalytic subunit 2 (PDP2; Fig. 6D), which plays a crucial role in switching metabolic flux from glycolysis to aerobic oxidation (24). Up-regulation of miR-21a suppresses PDP2, leading to anaerobic glycolysis. KEGG pathway analysis also showed that these eight miRNAs were highly related to both fibrosis signaling pathways and glucose metabolism pathways (Fig. 6E). These miRNAs were further evaluated on the basis of whether they exhibited evolutionary conservation and whether their expression in mice was comparable to human subjects. On the basis of these criteria, we focused on miR-21a, miR-26a, and miR-200a as potential therapeutic targets and proposed that correction of abnormal expression of these miRNAs could suppress hyperglycolysis and inhibit fibrogenesis, thereby exerting a synergistic effect on peritoneal fibrosis.

Fig. 6 Identification of miRNAs involved in glycolysis during peritoneal fibrogenesis.

(A) Volcano plot of miRNA microarray data of peritoneum from mice treated with PD fluid or saline. The red or blue dots represent miRNAs in peritoneum that are significantly (P < 0.05) up-regulated or down-regulated, respectively (n = 3 in each group). (B) Heat map showing relative expressions of the 41 miRNAs, which significantly changed (P < 0.05 by t test; fold change, >2) in peritoneum from PD fluid–treated mice compared to control group. Color gradient indicates expression value scaled by Z score. One-way hierarchical clustering of the 41 miRNAs performed (n = 3 in each group). (C) miRNA targets analysis for miRNAs involved in glycolysis and fibrogenesis. The up-regulated miRNAs are indicated by a red arrow, and down-regulated miRNAs are indicated with a blue arrow. (D) Table listing the eight miRNAs and their predicted target genes involved in glucose metabolism. (E) KEGG analysis of target genes of eight miRNAs, which are predicted to target both glycolysis pathways and fibrogenesis pathways. (F) The expression of miR-21a, miR-26a, and miR-200a in peritoneum from mice treated with PD fluid or saline. (G) The expressions of miR-21a, miR-26a, and miR-200a in PD fluid effluent-derived mesothelial cells from short-term and long-term patients undergoing PD. Results (F and G) are normalized to U6 small nuclear RNA (snRNA) and are presented as means ± SEM (**P < 0.01 or ***P < 0.001 by t test, n = 6). (H) Representative immunoblots and quantification of the expressions of PDP2, PFKFB3, and phosphofructokinase, muscle (PFKM) in MeT-5A cells after transfection of mimics of miR-21a, miR-26a, and miR-200a, respectively (data are presented as means ± SEM; *P < 0.05 by t test, n = 6).

Using miRNA-specific TaqMan RT-qPCR assays, we confirmed the expression of miR-21a, miR-26a, and miR-200a in both fibrotic peritoneum from a mouse model and effluent-derived mesothelial cells from patients undergoing PD. Expression of miR-21a was elevated, whereas the expression of miR-26a and miR-200a was down-regulated in mouse fibrotic peritoneum compared to the control group, which is consistent with the results of our miRNA microarray (Fig. 6F). We also confirmed up-regulation of miR-21a and down-regulation of miR-26a and miR-200a in mesothelial cells from patients with long-term PD, compared with patients with short-term PD (Fig. 6G).

To validate that glycolytic enzymes are the targets of these selected miRNAs, we transfected MeT-5A cells with mimics of miR-21a, miR-26a, or miR-200a and evaluated their effects on the expressions of predicted target enzymes, namely, PDP2, PFKFB3, and PFKM, respectively. Immunoblot showed that the expression of PDP2, PFKFB3, and PFKM significantly reduced after transfection with mimics of miR-21a, miR-26a, and miR-200a, respectively (Fig. 6H), indicating that these three miRNAs regulate the expression of targeted glycolytic enzymes in mesothelial cells. Together, these results suggested that miR-21a, miR-26a, and miR-200a could be used as therapeutic targets to suppress both hyperglycolysis and fibrotic gene expression in the peritoneum.

An miRNA triad suppresses TGF-β1–induced metabolic reprogramming and profibrotic phenotype in mesothelial cells

Because altered expression of miR-21a, miR-26a, and miR-200a is linked to hyperglycolysis and fibrosis in peritoneum, we presumed that correction of aberrant expression using miRNA mimic and/or inhibitor approaches might block PD fluid–induced fibrogenesis. To test this possibility, we determined the effects of individual miR-21a inhibitor, miR-26a mimic, or miR-200a mimic treatment in MeT-5A cells with or without TGF-β1 stimulation. As expected, MeT-5A cells singly transfected with the miR-21a inhibitor or the mimic of miR-26a exhibited lower glycolysis compared to cells treated with TGF-β1 and scrambled control miRNA (fig. S7, A and B). Transfection of the miR-200a mimic alone did not alter glycolysis (fig. S7C). Immunoblotting confirmed that the expression of FN1, COL1A1, and α-SMA was reduced by individual transfection of miR-21a inhibitor, miR-26a mimic, or miR-200a mimic (fig. S7D). Combining all three miRNAs together as an miRNA triad significantly suppressed the induction of ECAR stimulated by TGF-β1 (Fig. 7A). Both TGF-β1–induced glycolysis and glycolytic capacity were largely blunted by the miRNA triad (Fig. 7, B and C). Concomitant with these changes in glycolytic metabolism, the expression of PFKM and PFKFB3 was repressed, whereas the expression of PDP2 was restored by miRNA triad treatment despite stimulation with TGF-β1 (Fig. 7D). Subsequently, the expression of fibrotic proteins (FN1, COL1A1, and α-SMA) was also reduced by the miRNA triad (Fig. 7E). Because the antifibrotic effects of the miRNA triad were greater than those of any individual miRNA mimic or inhibitor (fig. S7D), these results suggest that the triad had synergistic effects.

Fig. 7 miRNA triad suppresses glycolysis and profibrotic phenotype induced by TGF-β1 in MeT-5A cells.

(A) ECAR measurements of glycolysis stress test performed in MeT-5A cells treated with TGF-β1 plus miRNA triad or scrambled, control miRNA (miRNA CTL; data are represented as means ± SEM, n = 6). Because the glycolysis stress test for miR-21a inhibitor, miR-26a mimic, or miR-200a mimic and the triad of these microRNAs were performed in the same experiment, the data of miRNA CTL and TGF-β1 plus miRNA CTL group shown in (A) were the same as shown in fig. S7, A to C. (B and C) Quantitative analysis of glycolytic stress test data reported in (A) (***P < 0.001 TGF-β1 + miRNA CTL versus miRNA CTL and ##P < 0.01 TGF-β1 + miRNA triad versus TGF-β1 + miRNA CTL, one-way ANOVA with Bonferroni’s multiple comparison test, n = 6). (D) Representative immunoblots and quantification of the expressions of PFKM, PFKFB3, and PDP2 in mesothelial cells treated with miRNA triad plus TGF-β1 or scrambled control. (E) Representative immunoblot and quantification of FN1, COL1A1, and α-SMA stimulated by TGF-β1 with or without miRNA triad treatment. Data (D and E) are presented as means ± SEM (*P < 0.05, **P < 0.01, or ***P < 0.001 for TGF-β1 + miRNA CTL versus miRNA CTL and #P < 0.05 or ##P < 0.01 for TGF-β1 + miRNA triad versus TGF-β1 + miRNA CTL, one-way ANOVA with Bonferroni’s multiple comparison test, n = 6 in each group).

An AAV1-miRNA triad blocks hyperglycolysis and fibrogenesis in mouse peritoneum

To efficiently deliver the miRNA triad in vivo while minimizing unfavorable effects, we developed an AAV-mediated therapy. AAV isotype 1 (AAV1) was selected as the vector because of its high affinity for mesothelial cells. We validated the transfection efficiency of AAV1 in mouse peritoneum by intraperitoneal injection of AAV1-GFP. Seventy-two hours after injection of the virus, more than 95% of peritoneal cells expressed GFP (Fig. 8A). We generated three AAV1s encoding the miR-21a inhibitor, miR-26a mimic, and miR-200a mimic, respectively, and mixed equal amounts of these three AAV1s to yield a viral miRNA triad. Using the PD fluid–induced mouse model of peritoneal fibrosis, we showed a more than fivefold increase in miR-26a and miR-200a expression and miR-21a suppression by 90% in peritoneal tissue 2 weeks after AAV1-miRNA triad treatment [single-dose intraperitoneal injection of triad (body weight, 1 × 1011 gv/g) given on the fifth week of PD fluid; Fig. 8, B and C]. PD fluid caused significant (P < 0.001) thickening of both mesenteric and parietal peritoneal tissue, accompanied by an increase in collagen deposition (Fig. 8, D and E). This thickening was largely reduced in mice treated with the AAV1-miRNA triad (Fig. 8, D and E). Peritoneal function also improved in miRNA triad–treated mice (Fig. 8F). miRNA triad–treated mice had lower levels of active TGF-β1 in peritoneal effluent (Fig. 8G). Immunoblotting revealed that phosphorylated Smad2/3 in peritoneum was suppressed by miRNA triad treatment, and the expression of ZEB1 and SNAI1, which are transcription factors known to promote EMT, was also reduced by miRNA triad injection (Fig. 8H). The induction of glycolysis-related enzymes in the fibrotic peritoneum was suppressed, whereas the expression of PDP2 was restored in the peritoneum of mice treated with the AAV1-miRNA triad (Fig. 8I). Concomitantly, the induction of FN1, COL1A1, and α-SMA in fibrotic peritoneum was markedly reduced by miRNA triad treatment (Fig. 8J). These results indicate that AAV1-mediated miRNA triad treatment efficiently inhibits peritoneal hyperglycolysis and fibrogenesis in a mouse model of PD fluid–induced fibrosis.

Fig. 8 miRNA triad treatment alleviates hyperglycolysis and fibrogenesis in the peritoneum of a mouse model of peritoneal fibrosis.

(A) Representative microscopic image of mouse peritoneum transfected with AAV1-GFP. Green, GFP. Red, nuclei of peritoneal cells. Scale bar, 100 μm. (B) Schematic graph of experimental design. Arrows denote the time point of PD fluid injection, AAV1-miRNA triad treatment, and peritoneum sample collection. (C) RT-qPCR showing the expression of miR-21a, miR-26a, and miR-200a in mouse peritoneum after the infection of AAV1-miRNA triad (means ± SEM; **P < 0.01 or ***P < 0.001 versus saline + AAV1 CTL and ###P < 0.001 for PD fluid + AAV1-miRNA triad versus PD fluid +AAV1 CTL; one-way ANOVA with Bonferroni’s multiple comparison test, n = 6). (D) Masson’s trichrome staining and collagen 1 staining of peritoneal tissue from mice treated with PD fluid plus miRNA CTL or miRNA triad. Representative images derived from six mice of each group. Scale bars, 100 μm. (E) Quantitative analysis of the thickness in mesentery and parietal peritoneal submesothelial compact zone of mice treated with PD fluid plus miRNA CTL or miRNA triad. (F) Peritoneal permeability of mice treated with PD fluid plus miRNA CTL or miRNA triad examined by modified peritoneal equilibration test. (G) Active TGF-β1 in effluent of peritoneal dialysate measured by enzyme-linked immunosorbent assay. (H) Immunoblots of phosphorylated and total SMAD2/3, ZEB1, and SNAI1 in peritoneum from mice treated with PD fluid plus AAV1-miRNA triad. (I) Immunoblots of PFKM, PFKFB3, and PDP2 in peritoneum from mice treated with PD fluid plus AAV1-miRNA triad. (J) Immunoblots of FN1, COL1A1, and α-SMA in peritoneal tissue from mice treated with PD fluid plus AAV1-miRNA triad. Data (E to J) were presented as means ± SEM (*P < 0.05, **P < 0.01, or ***P < 0.001 for PD fluid + AAV1 CTL versus saline + AAV1 CTL and #P < 0.05, ##P < 0.01, or ###P < 0.001 for PD fluid + AAV1-miRNA triad versus PD fluid + AAV1 CTL; one-way ANOVA with Bonferroni’s multiple comparison test, n = 6).

DISCUSSION

The results of our scRNA-seq analysis demonstrated that effluent-derived mesothelial cells could become an important source of extracellular matrix contributing to fibrosis because they robustly produce collagens (COL1A1 and COL4A1). Mesothelial cells are embryologically derived from the mesoderm (25). They share some features of epithelial cells because they express not only cytokeratins and tight junction proteins including ICAM1 and TJP1 but also vimentin (VIM), an intermediate filament protein, which is considered a mesenchymal marker (26). Because mesothelial cells are not typical epithelial cells, they may exhibit distinct biological characteristics during the transition to a mesenchymal state. Several transcriptomic studies have profiled gene expression of mesothelial cells. Namvar et al. (19) performed RNA-seq on HBME1-sorted mesothelial cells from rodent omentum, in which they detected a “mesenchymal/extracellular matrix signature” that increased after TGF-β1 treatment. Similarly, we found that the expression of VIM and COL4A1 increased in mesothelial cells from the long-term PD group compared to those of the short-term PD group. Namvar et al. (19) noted very low read numbers for CDH1. In our scRNA-seq using human mesothelial cells, we found that the expression of CDH1 was relatively low in mesothelial cells from both normal peritoneum and patients undergoing PD; however, the protein E-cadherin was relatively abundant. This discrepancy between mRNA and protein expression may indicate posttranslational modification of E-cadherin (27). Mesothelial cells also exhibited heterogeneity during MMT. Ruiz-Carpio et al. (28) profiled whole-genome microarray on mesothelial cells from patients undergoing PD, finding different gene expression between early/epithelioid and advanced/non-epithelioid mesothelial cells. These data suggest heterogeneity and high plasticity of mesothelial cells. Using scRNA-seq, we confirmed heterogeneity of mesothelial cells during MMT and demonstrated the trajectory of mesothelial transition. We found that MMT was accompanied by increased expression of glycolytic enzymes.

Our data demonstrated that exposure to PD fluid leads to hyperglycolysis in mesothelial cells, facilitating not only energy demand but also cell phenotype transition and proliferation. This finding suggests that an intervention inhibiting MMT could successfully prevent peritoneal fibrogenesis. Metabolomics analysis of a mouse model of peritoneal fibrosis revealed changes in metabolites involved in major energy metabolic pathways, including evidence that PD fluid can induce hyperglycolysis in peritoneal cells. Besides the responses stimulated by PD fluid, we showed that TGF-β1 induces hyperglycolysis in cultured human primary mesothelial cells and the MeT-5A cell line. Blocking hyperglycolysis largely suppressed MMT in both types of mesothelial cells. This response was confirmed in the mouse model of peritoneal fibrosis in which administration of 2-DG efficiently blocked hyperglycolysis and the development of peritoneal fibrosis.

We did not anticipate that short-term exposure (≤48 hours) of cultured mesothelial cells to PD fluid would not change the expression of enzymes mediating glycolysis, whereas exposure to TGF-β1 rapidly induced the expression of these enzymes. These differences suggest that exposure to PD fluid does not directly cause hyperglycolysis in mesothelial cells. Mesothelial cells in the peritoneum are known to produce TGF-β1 even in a homeostatic state (29). The production of TGF-β1 can be amplified in response to advanced glycation end products, glucose degradation products, or low pH in the PD fluid (3), which stimulates MMT through Smad-dependent and Smad-independent pathways (30). TGF-β1 stimulates both glycolysis and the expression of fibrotic proteins, indicating that the profibrotic potency of TGF-β1 in mesothelial cells may partially depend on its ability to induce hyperglycolysis.

Many miRNAs reportedly participate in fibrogenesis (31, 32). There is emerging evidence that miRNAs regulate cellular metabolism by directly targeting key metabolic enzymes and/or transcription regulators (33). Knowing that one miRNA can regulate multiple genes (34), we designed an miRNA triad (an miRNA-21a inhibitor and mimics of miRNA-26a and miRNA-200a) to target glycolysis-related enzymes and fibrogenesis-related signaling pathways. AAV1 was used as the delivery vector because of its safety and efficacy for gene therapy in preclinical and clinical trials (35). Compared with individual miRNA, our AAV1-miRNA triad achieved a more reliable therapeutic effect including marked repression of hyperglycolysis and amelioration of peritoneal fibrosis induced by PD fluid in mice.

There are limitations in our study. First, mesothelial cells collected from the effluent PD fluid of patients may mainly represent mesothelial cells that shed from the peritoneum due to MMT. This fluid may not include some mesothelial cells that invade into the underlying tissue. Tissue invasion may be due to chemokines released by inflammatory cells in the peritoneal tissue, and the difference between these shed cells collected in effluent and invading mesothelial cells not captured in effluent needs further study. Second, the peritoneal membrane consists of various types of cells. Consequently, the AAV1-miRNA triad we used to block peritoneal fibrogenesis may influence functions of other nonmesothelial types of cells that contribute to the development of peritoneal fibrosis, for example, resident fibroblasts and macrophages.

Our results provide evidence that hyperglycolysis is a mechanism that promotes EMT in mesothelial cells. Blocking hyperglycolysis suppresses mesothelial EMT and peritoneal fibrosis. This therapeutic strategy may have implications for fibrogenesis in other organs.

MATERIALS AND METHODS

Study design

Our study had three main objectives. The first objective was to identify the metabolic state of peritoneal mesothelial cells in response to PD fluid exposure in patients receiving PD. The second objective was to investigate the role of metabolic reprogramming in pathogenesis of peritoneal fibrosis. The third objective was to develop a triad of AAV1-encoded miRNAs and to evaluate therapeutic potential to treat peritoneal fibrosis. The use of human tissue for this study was approved by the Ethics Committee of the Third Affiliated Hospital of Sun Yat-sen University, China. Written consent was obtained from all patients. For experiments in mouse models, mice were randomly assigned to control or treatment groups with equivalent numbers. For biochemical and histological analysis in vitro and in vivo, at least three biological replicates were performed.

Human sample collection

Peritoneal cells digested from normal peritoneal tissues of patients. Normal parietal peritoneal tissues were taken from six patients by a surgeon during laparoscopic inguinal hernia repair. Written consents were obtained from the patients. Tissues were cut into about 4-mm2 pieces and digested for 15 min at 37°C with agitation in a digestion solution containing collagenase D (10 mg/ml; Roche), pronase (Roche), collagenase II (Sigma-Aldrich), dispase II (Sigma-Aldrich), and deoxyribonuclease I (100 μg/ml; Sigma-Aldrich) in Hank’s balanced salt solution (HBSS; Gibco). Samples were added into cold HBSS with 10% fetal bovine serum and kept on ice for 10 min to stop reaction. The resulting suspension was passed through a 70-μm cell strainer (Falcon), followed by centrifugation for 500g for 5 min. The pellet was incubated with red blood cell (RBC) lysate for 3 min at room temperature, centrifuged at 500g for 5 min, and resuspended in a solution containing 0.4% bovine serum albumin (BSA) in phosphate-buffered saline (PBS). The yielded single-cell suspension (10 μl) was washed and counted using a hemocytometer with trypan blue. Overall, it took 2 to 4 hours from obtaining biopsies to generating single-cell suspension runs on the Chromium 10x device. A total of three samples were qualified for single-cell sequence library preparation.

Effluent-derived peritoneal cell from patients undergoing PD. We collected PD fluid samples from 12 patients who were diagnosed with ESRD and received PD. Six patients had received PD for less than 2 weeks, and the other six patients had received PD for more than 6 years. Written consents were obtained from the patients undergoing PD. Peritoneal dialysate (Dianeal PD solution contains 1.5% dextrose; Baxter Healthcare) was drained after overnight (10 hours) dwell exchange. The effluent of peritoneal dialysate was obtained within 1 hour of the patient’s dialysate outflow. After centrifugation at 3000 rpm for 10 min, cell pellet was incubated with 2 ml of RBC lysis buffer on ice for 1.5 min. Cells were washed then centrifuged at 1000 rpm for 5 min and resuspended at a concentration of 2000 cells/μl in PBS containing 0.5% BSA. Cell number and viability were analyzed using Cellometer AutoT4 (Nexcelom Bioscience) to ensure single-cell suspension with more than 80% viability. Each sample was split for Western blot and scRNA-seq analysis. For scRNA-seq analysis, four samples from long-term patients undergoing PD and six samples from short-term patients undergoing PD were qualified for single-cell sequence library preparation.

Mouse model and sample collection. Eight- to ten-week-old male C57BL/6 mice were purchased from the laboratory animal center of the Shandong Experimental Animal Center. PhAMexcised mice (stock no.108397) were purchased from the Jackson Laboratory. All animal procedures are approved by the Ethics Committee of the Third Affiliated Hospital of Sun Yat-sen University. All animal experiments followed the animal research: reporting of in vivo experiments (ARRIVE) guideline. A mouse model of peritoneal fibrosis was established by daily intraperitoneal injection of 4.25% PD fluid (Dianeal containing 4.25% glucose; Baxter Healthcare) at 0.1 ml/g body weight for 6 weeks. For 2-DG treatment, mice were treated with daily intraperitoneal injection of 2-DG (80 mg/kg; Sigma-Aldrich) from the fifth week of PD fluid injection and for a total of 2 weeks. For miRNA triad treatment, AAV1 miRNA viruses (mixture of AAV1-miR-21a inhibitor, AAV1-miR-26a mimic, and AAV1-miR-200a mimic) or AAV1-miR-blank virus (abmGood) was intraperitoneally injected at 1 × 1011 gv/g body weight at the beginning of the fifth week of PD fluid injection.

mRNA preparation and quantitative real-time RT-PCR. Total RNA was extracted from snap-frozen mesenteric tissues with QIAzol lysis reagent (QIAGEN) and precipitated in isopropanol. Complementary DNA (cDNA) was synthesized using qScript cDNA SuperMix (Quanta Bio). SYBR Green RT-qPCR was performed with CFX96 Touch Real-Time PCR Detection System (Bio-Rad) according to the manufacturer’s instructions. The specificity of RT-PCR was confirmed using melting-curve analysis. The expression levels of the target genes were normalized by β-actin level in each sample.

For the quantitative analysis of miRNAs, TaqMan miRNA assay (Applied Biosystems) was performed following manufacturer’s instructions. U6 snRNA was used as a loading control.

The qPCR-based mRNA array was performed using mouse glucose metabolism (PAMM-006z, QIAGEN) following the manufacturer’s protocol. The primers for qPCR are listed in table S1.

Western blots. Total protein from tissues or cells was extracted by radioimmunoprecipitation assay lysis and extraction buffer (G-Biosciences) containing protease and phosphatase inhibitor (Thermo Scientific, Rockford) on ice. The lysates were collected after centrifugation at 13,600 rpm at 4°C for 15 min and heated with sample buffer at 100°C for 5 min before being separated by SDS–polyacrylamide gel electrophoresis on gradient gels. The proteins were electrotransferred to polyvinylidene difluoride membranes (Merck Millipore Ltd). The membranes were blocked with 5% nonfat milk–tris-buffered saline and incubated overnight with primary antibodies at 4°C, followed by 1 hour of incubation with DyLight 680 or 800 or horseradish peroxidase–conjugated secondary antibodies (Cell Signaling Technology) at room temperature. The bands were visualized with ChemiDoc MP Imaging System (Bio-Rad). To ensure equal protein loading, β-actin protein was used as the endogenous control.

The anti–collagen I antibody was purchased from Santa Cruz Biotechnology. The anti-fibronectin antibody was purchased from Abcam. The anti–α-SMA antibody was purchased from Sigma-Aldrich. Anti-PDP2 antibody was purchased from LSBio. The E-cadherin antibody was purchased from BD Biosciences. Other antibodies were purchased from Cell Signaling Technology.

Statistical analysis. Data are shown as means ± SEM and analyzed using the IBM Statistical Product and Service Solutions (SPSS) software (version 23). For experiments comparing two groups, we analyzed results by the two-tailed, unpaired Student’s t tests. When more than two groups were compared, parametric data were analyzed by ANOVA methods, followed by Bonferroni’s multiple comparison tests to analyze differences between two interested groups. Nonparametric data were analyzed by a Kruskal-Wallis test. A value of P < 0.05 was considered statistically significant. Primary data are reported in data file S1.

SUPPLEMENTARY MATERIALS

stm.sciencemag.org/cgi/content/full/11/495/eaav5341/DC1

Materials and Methods

Fig. S1. Quality control and clustering of single cells.

Fig. S2. Mesothelial cell marker gene expression in the single-cell maps.

Fig. S3. Subclustering of peritoneal cells derived from PD effluent.

Fig. S4. GSEA of cell clusters and immunohistochemical staining of human peritoneum.

Fig. S5. PD fluid induces fibrosis in mouse peritoneum.

Fig. S6. PD fluid activates TGF-β1 in mouse peritoneum.

Fig. S7. miR-21a inhibitor, miR-26a mimic, and miR-200a mimic suppress hyperglycolysis and fibrogenesis in human mesothelial cells.

Table S1. RT-qPCR primer sequences.

Data file S1. Primary data.

REFERENCES AND NOTES

Acknowledgments: We thank the Guangzhou Qianyang Biomedical Co. Ltd. for assistance in sequencing and bioinfomatics analysis. Funding: This work was supported by the National Natural Science Foundation of China (NSFC84170955 and NSFC81670675 to H.P.), the National Natural Science Foundation of Guangdong, China (2017A030313714 to H.P.), and the Science and Technology Program of Guangzhou, China (International Science & Technology Cooperation Program; 201807010037 to H.P.). Z.H. is supported by an NIH grant (AR063686). Author contributions: H.P., M.S., and Z.H. designed the study and wrote the manuscript. M.S., Q.W., Y.L., H.L., D.L., W.Z., X.D., and Y.H. carried out experiments and analyzed results. M.S., J.L., and Z.H. performed experiments of scRNA-seq. H.Z. and T.L. interpreted the data. Competing interests: The authors declare that they have no competing interests. Data and materials availability: All data associated with this study are present in the paper or Supplementary Materials. The miRNA microarray data have been submitted to Gene Expression Omnibus (GEO) with accession no. GSE130387. scRNA-seq data are available at GEO with accession number GSE130888.
View Abstract

Stay Connected to Science Translational Medicine

Navigate This Article