Lineage Tracing Reveals Distinctive Fates for Mesothelial Cells and Submesothelial Fibroblasts during Peritoneal Injury

Abstract

Fibrosis of the peritoneal cavity remains a serious, life-threatening problem in the treatment of kidney failure with peritoneal dialysis. The mechanism of fibrosis remains unclear partly because the fibrogenic cells have not been identified with certainty. Recent studies have proposed mesothelial cells to be an important source of myofibroblasts through the epithelial–mesenchymal transition; however, confirmatory studies in vivo are lacking. Here, we show by inducible genetic fate mapping that type I collagen–producing submesothelial fibroblasts are specific progenitors of α-smooth muscle actin–positive myofibroblasts that accumulate progressively in models of peritoneal fibrosis induced by sodium hypochlorite, hyperglycemic dialysis solutions, or TGF-β1. Similar genetic mapping of Wilms’ tumor-1–positive mesothelial cells indicated that peritoneal membrane disruption is repaired and replaced by surviving mesothelial cells in peritoneal injury, and not by submesothelial fibroblasts. Although primary cultures of mesothelial cells or submesothelial fibroblasts each expressed α-smooth muscle actin under the influence of TGF-β1, only submesothelial fibroblasts expressed α-smooth muscle actin after induction of peritoneal fibrosis in mice. Furthermore, pharmacologic inhibition of the PDGF receptor, which is expressed by submesothelial fibroblasts but not mesothelial cells, attenuated the peritoneal fibrosis but not the remesothelialization induced by hypochlorite. Thus, our data identify distinctive fates for injured mesothelial cells and submesothelial fibroblasts during peritoneal injury and fibrosis.

Many patients with kidney failure rely on the peritoneal membrane to perform life-saving dialysis.^1–3 In addition to changes in permeability of the peritoneal membrane, the dialysis process itself frequently triggers a fibrosing process that progressively reduces membrane function resulting in dialysis failure, sometimes with high patient mortality.^4–7 In a small percentage of patients, severe fibrosis occurs primarily in the visceral peritoneum, resulting in encapsulating peritoneal sclerosis (EPS), a catastrophic complication with obscure pathogenesis and a high mortality rate.^6,7 Such dialysis failure is characterized by progressive peritoneal fibrosis that can be seen as with thickening of basal lamina and accumulation of α-smooth muscle actin (αSMA)⁺ myofibroblasts.^4,5,8–10

The peritoneum is composed of mesothelium, basal lamina, and submesothelial (SM) connective tissue.^11–13 The mesothelium consists of a single layer of flattened mesothelial cells (MCs) that lines the peritoneal cavity and internal organs.^12,14–16 In many circumstances such as organ development or tissue injury repair, MCs are the cellular source of growth factors including TGF-β1, PDGF, and vascular endothelial growth factor, which support cell proliferation and differentiation of parenchymal and stromal cells as well as angiogenesis.^8,17–22 By contrast, SM connective tissue containing plexuses of blood vessels, lymphatic channels, and scattered fibroblasts has drawn much less attention.^22–24 A number of studies suggested MCs as the major source of myofibroblasts through the epithelial–mesenchymal transition (EMT) during peritoneal fibrosis.^8,18,25–27 However, these studies relied predominantly on in vitro experiments to show that MCs can be stimulated to express αSMA and produce matrix proteins outside of the body under the influence of profibrotic agents such as TGF-β1.^8,25–28 Nevertheless, confirmatory studies of mesothelial EMT in vivo are lacking even though costaining of cytokeratin and αSMA was previously shown.^8,18 The contribution of SM fibroblasts to myofibroblasts in vivo is not clear despite some studies in vitro have suggested.^22–24

A conditional cell lineage analysis using WT1^CreERT2/+ mice demonstrated that Wilms’ tumor-1 (WT1)⁺ septum transversum mesenchyme gives rise to MCs, SM fibroblasts covering the liver, and hepatic stellate cells within the liver during hepatic development.^12,13 Using WT1^CreERT2/+ mice, a recent study reported that WT1⁺ MCs may differentiate into myofibroblasts in liver injury.²⁹ WT1 expression in MCs is observed in both embryonic development and adult peritoneum; however, the expression of WT1 by SM fibroblasts is not clearly defined in the adult peritoneum.^{12,13,21,30–32} Hence, the progenitors of myofibroblasts in the injured liver and the peritoneum remain controversial despite these studies.

Because efforts to design new antifibrotic therapies require a rigorous understanding of the cellular origin of myofibroblasts in vivo, we performed lineage tracing of both MCs and SM fibroblasts in models of peritoneal fibrosis induced by sodium hypochlorite solution, hyperglycemic dialysis solution, or adenovirus-expressing TGF-β1 (AdTGF-β1). Although these models are more akin to EPS than the progressive thickening peritoneum seen in humans on peritoneal dialysis, they represent robust tools to study the pathogenesis of peritoneal fibrosis in the laboratory.^33–36 Contrary to the prevailing model, our findings indicate that peritoneal myofibroblasts derive from SM fibroblasts and peritoneal membrane disruption is repaired by surviving MCs.

Results

Expanded Population of Myofibroblasts in Models of Peritoneal Fibrosis

Using Col1a1-GFP transgenic reporter mice (Col1a1-GFP^Tg) expressing enhanced green fluorescent protein (GFP) under the regulation of the Col1a1 promoter and enhancers, we studied the collagen-producing cells in normal and injured peritoneum. Col1a1-GFP–positive cells were cytokeratin^–, PDGF receptor-β (PDGFRβ)⁺, and vimentin⁺, and lay beneath the nidogen⁺ basal lamina (Figure 1, A–C). We called these cells SM fibroblasts. Seven days after hypochlorite injury, the population of Col1a1-GFP⁺ cells expanded in the peritoneum of the liver, omentum, and abdominal wall (Supplemental Figure 1). In addition, we noted diffuse thickening of the basal lamina to become scar tissue (Figure 1, D and E). Because the preparation of peritoneum covering solid organs for immunofluorescence study was easier, we showed the findings of peritoneal covering of liver unless otherwise specified. After hypochlorite injury, cytokeratin⁺ staining was not detected at many surfaces of the peritoneum, suggesting the loss of MCs. At peritoneal surfaces in which MCs remained attached after injury, the cytokeratin⁺ MCs could now be seen to express Col1a1-GFP, in sharp contrast with the healthy state (Figure 1, A, D, and F). Cytokeratin⁺ MCs did not express detectable αSMA (Figure 1F). Myofibroblasts, defined by coexpression of αSMA and Col1a1-GFP, accumulated markedly within the thickened basal lamina between MCs and SM fibroblasts (Figure 1, E and F), and expressed both PDGFRβ and vimentin (Figure 1, G and H). The observations that injured cytokeratin⁺ MCs generated Col1a1 transcripts and that αSMA⁺ myofibroblasts accumulated in the thickened laminin⁺ scar were reproduced in Col1a1-GFP^Tg mice 10 days after intraperitoneal injection of AdTGF-β1 (Supplemental Figure 2).

Figure 1.

Col1a1-GFP identifies collagen-producing cells in normal and injured peritoneum. (A) Collagen-producing cells expressing enhanced GFP under the regulation of the collagen type I (α1) (Col1a1) promoter and enhancers are SM fibroblasts (arrows), not cytokeratin⁺ MCs (arrowheads), in normal peritoneum of Col1a1-GFP^Tg mice. (B and C) Col1a1-GFP⁺ SM fibroblasts (arrows) express PDGFRβ and vimentin in normal peritoneum. Cells (arrowheads) above Col1a1-GFP⁺ SM fibroblasts, suggesting MCs are PDGFRβ^– and weak vimentin⁺. (D) Cell numbers of Col1a1-GFP⁺ collagen–producing cells and thickness of the nidogen⁺ scar in Col1a1-GFP^Tg mice increase within 4 days after intraperitoneal injection of hypochlorite. Cytokeratin⁺ MCs after hypochlorite injury (arrowheads) also express Col1a1-GFP. (E) Myofibroblasts, characterized by αSMA⁺ and Col1a1-GFP⁺ coexpression, accumulate in the thickened laminin⁺ scar 7 days after hypochlorite injury. Col1a1-GFP⁺ cells on the peritoneal surface, suggesting injured MCs (arrowheads), do not express αSMA. (F) Cytokeratin⁺ MCs (arrowheads) express Col1a1-GFP, not αSMA, 7 days after hypochlorite injury. (G) Col1a1-GFP⁺ cells on the peritoneal surface, suggesting injured MCs (arrowheads), do not express PDGFRβ 7 days after hypochlorite injury. (H) Vimentin is expressed in Col1a1-GFP⁺ cells 7 days after hypochlorite injury. Arrowheads indicate Col1a1-GFP⁺ cells on the peritoneal surface, suggesting injured MCs. All images are taken at the original magnification from the peritoneal covering of the liver unless otherwise specified. Bar, 20 μm. Original magnification, ×630.

The Fate Marker Activated in WT1^CreERT2/+ Mice Identified MCs and a Small Population of SM Fibroblasts

Permanent expression red fluorescence protein (RFP) was activated in WT1-expressing cells by somatic DNA recombination in adult WT1^CreERT2/+;ROSA26^fstdTomato (WT1-RFP) mice. Activation was induced conditionally by oral tamoxifen administration so that a cohort of cells was permanently labeled only during tamoxifen exposure (Figure 2A). On the peritoneal surface, 83.3% of the cytokeratin⁺ MCs underwent somatic recombination and were referred to as WT1-RFP⁺ MCs (Figure 2, B and C, Supplemental Figure 3). In addition, WT1-RFP⁺;cytokeratin^– cells, representing 23.6% of all WT1-RFP⁺ cells, were seen beneath the mesothelium in keeping with labeling of SM fibroblasts (Figure 2B). In addition to notable expression in SM fibroblasts, vimentin was detectable in normal WT1-RFP⁺ MCs (Figures 1C and 2D), suggesting that MCs normally expressed this mesenchymal protein. To confirm that WT1-RFP⁺ cells beneath the mesothelium were SM fibroblasts, we generated WT1^CreERT2/+;ROSA26^fstdTomato;Col1a1-GFP^Tg mice. Of the Col1a1-GFP⁺ SM fibroblasts, 17.6% expressed WT1-RFP, indicating that CreERT2 at the WT1 locus activated ROSA26^fstdTomato in a small population of SM fibroblasts (Figure 2E). Furthermore, 74.8% of all WT1-RFP⁺ cells were Col1a1-GFP^– and were on the peritoneal surface, in keeping with them being MCs (Figure 2E). FACS analysis showed similar results (Figure 2F).

Figure 2.

WT1 expression in the adult peritoneum enables efficient labeling of MCs as well as a minor population of SM fibroblasts. (A) Experimental schema for cohort labeling in WT1^CreERT2/+;ROSA26^fstdTomato and WT1^CreERT2/+;ROSA26^fstdTomato;Col1a1-GFP^Tg mice from 10 weeks of age. Analysis is performed on the peritoneal covering of liver 2 weeks after cohort labeling. (B) Cytokeratin⁺;WT1-RFP⁺ (arrowheads) and cytokeratin⁺;WT1-RFP^– MCs (asterisks) are shown in normal peritoneum of WT1^CreERT2/+;ROSA26^fstdTomato mice. Cytokeratin^–;WT1-RFP⁺ cells (arrows) are occasionally seen. The graph shows the percentage of cytokeratin⁺ MCs labeled with WT1-RFP and the percentage of WT1-RFP⁺ cells without cytokeratin expression (mean±SEM, n=6). (C) Three-dimensional images with XZ stacks show cytokeratin⁺;WT1-RFP⁺ MCs (arrowheads) on the surface of normal peritoneum of WT1^CreERT2/+;ROSA26^fstdTomato mice. (D) Vimentin is expressed by both WT1-RFP⁺ cells (arrowheads, suggesting MCs) and WT1-RFP^– cells (arrows, suggesting SM fibroblasts) in normal peritoneum of WT1^CreERT2/+;ROSA26^fstdTomato mice. (E) WT1-RFP⁺;Col1a1-GFP^– MCs (arrowheads) and WT1-RFP^–;Col1a1-GFP⁺ SM fibroblasts (asterisks) separated by the laminin⁺ basal lamina are shown in normal peritoneum of WT1^CreERT2/+;ROSA26^fstdTomato;Col1a1-GFP^Tg mice. WT1-RFP⁺;Col1a1-GFP⁺ SM fibroblasts (arrow) are occasionally seen. The graph shows the percentage of Col1a1-GFP⁺ SM fibroblasts labeled with WT1-RFP and the percentage of WT1-RFP⁺ MCs without Col1a1-GFP expression (n=6). (F) Representative plot shows FACS analysis of peritoneal cells prepared from WT1^CreERT2/+;ROSA26^fstdTomato;Col1a1-GFP^Tg mice after intraperitoneal retention of trypsin-EDTA. The graph shows the percentage of Col1a1-GFP⁺ SM fibroblasts labeled with WT1-RFP and the percentage of WT1-RFP⁺ MCs without Col1a1-GFP expression (n=3). Bar, 20 μm. Original magnification, ×630.

Injured Peritoneum Was Remesothelialized by WT1-RFP⁺ MCs

Peritoneal fibrosis was initiated by hypochlorite in WT1^CreERT2/+;ROSA26^fstdTomato mice 2 weeks after the last dose of tamoxifen and the fate of the cohort of WT1-RFP⁺ cells was mapped (Figure 3A). Within 10 days after injury, peritoneal surfaces were partially covered by WT1-RFP⁺ or cytokeratin⁺ MCs (Figure 3, B and C), but many surfaces were devoid of MCs (Figure 3B). WT1-RFP expression was noted in 85.6% of cytokeratin⁺ MCs (Figure 3B). There was no dilution of the proportion of cytokeratin⁺ MCs with the fate marker WT1-RFP before and after injury (Figures 2B and 3B). Active proliferation in WT1-RFP⁺ MCs was confirmed by positive Ki67 staining (Figure 3D). No recombination of somatic DNA in the absence of tamoxifen administration was detected before and after injury (Supplemental Figure 4).

Figure 3.

Injured peritoneum is remesothelialized by surviving MCs. (A) The experimental schema shows cohort labeling followed by hypochlorite injury in WT1^CreERT2/+;ROSA26^fstdTomato mice. Analysis is performed on the peritoneal covering of liver 10 days after injury. (B) Low-powered (upper panel) and high-powered (lower panel) images show cytokeratin⁺;WT1-RFP⁺ MCs (arrows) on the surface of the thickened nidogen⁺ scar. Arrowheads indicate the denuded peritoneum. Only rare WT1-RFP⁺ cells are noted within or beneath the nidogen⁺ scar. The graph beneath shows the percentage of cytokeratin⁺ MCs labeled with WT1-RFP at this time point (n=6). (C) Three-dimensional images with YZ and XZ stacks show WT1-RFP⁺ MCs (arrowheads) on the surface of the thickened nidogen⁺ scar. Only rare WT1-RFP⁺ cells (arrows) are noted within the nidogen⁺ scar. (D) Active proliferation of WT1-RFP⁺ MCs after injury is shown by Ki67 expression (asterisks). (E) Images show that most of WT1-RFP⁺ cells are found above (MCs, arrowheads) or below (SM fibroblasts, arrows) the thickened laminin⁺ scar after injury. Extremely rare WT1-RFP⁺;αSMA⁺ myofibroblasts (asterisks) are within the thickened laminin⁺ scar. The graph beneath shows the percentage of αSMA⁺ myofibroblasts coexpressing WT1-RFP (n=6). (F) Three-dimensional images with YZ and XZ stacks show that most of WT1-RFP⁺ cells are on the peritoneal surface (MCs, arrowheads) after injury. WT1-RFP⁺;αSMA⁺ myofibroblasts (asterisks) are extremely rare. Bar, 20 μm. Original magnification, ×200 in B upper panel; ×630 in B lower panel and C–F.

In contrast with the active involvement of surviving MCs in remesothelialization after hypochlorite injury (Figure 3, B–D), only a few WT1-RFP⁺ cells coexpressing αSMA were found within the thickened laminin⁺ scar (Figure 3, E and F, Supplemental Figure 5). Most of the WT1-RFP⁺;αSMA^– cells were found on the peritoneal surfaces, indicating that they were MCs as those observed in Figures 2 and 3C, the others were beneath the thickened scar and therefore SM fibroblasts (Figure 3, E and F, Supplemental Figure 5). The percentage of αSMA⁺ cells expressing WT1-RFP was 15.9%, representing a proportion similar to that of Col1a1-GFP⁺ SM fibroblasts expressing WT1-RFP (Figures 2, E and F, and 3, E and F). Furthermore, WT1-RFP⁺ cells within and beneath the thickened scar expressed PDGFRβ, but WT1-RFP⁺ MCs did not (Supplemental Figure 6), collectively suggesting that the minor population of WT1-RFP⁺ SM fibroblasts is the source of the minor population of WT1-RFP⁺ myofibroblasts.

To validate the fate of WT1 lineage of cells in peritoneal injury, we studied a second model of peritoneal fibrosis induced by daily intraperitoneal injection, for 2 weeks, of dialysis solution containing 4.25% glucose and 40 mM of the glucose degradation product methylglyoxal (Supplemental Figure 7A). Although the injury was much milder than that induced by hypochlorite, αSMA⁺ myofibroblasts was detectable (Supplemental Figure 7B). Similar to the observation in the hypochlorite-induced model, WT1-RFP⁺ cells were noted on the peritoneal surface or beneath the thickened scar (Supplemental Figure 7, B and C). A low percentage (16.5%) of αSMA⁺ myofibroblasts coexpressing WT1-RFP was detected within the laminin⁺ scar (Supplemental Figure 7B). Similar to the findings in the hypochlorite-induced model, >85% of the cytokeratin⁺ MCs at the peritoneal surface were WT1-RFP⁺ before and after injury (Supplemental Figure 7C).

TGF-β1 Upregulated αSMA in WT1-RFP⁺ MCs In Vitro but Not In Vivo

We induced a third model of peritoneal fibrosis by intraperitoneal injection of AdTGF-β1 in WT1^CreERT2/+;ROSA26^fstdTomato mice (Figure 4A). Before AdTGF-β1 administration, 82.6% of cytokeratin⁺ MCs were labeled with the WT1-RFP fate marker. Ten days after AdTGF-β1 administration, marked accumulation of αSMA⁺ myofibroblasts was seen in the peritoneum (Figure 4, B and C). In addition to WT1-RFP⁺;αSMA^– SM fibroblasts beneath the thickened laminin⁺ scar, occasional scattered WT1-RFP⁺;αSMA⁺ myofibroblasts were noted within the thickened scar but these amounted to only 14.6% of αSMA⁺ myofibroblasts (Figure 4, B and C). αSMA was not detected in WT1-RFP⁺ MCs on the peritoneal surface (Figure 4C). In peritoneal areas in which coverage by MCs remained intact, >85% of cytokeratin⁺ MCs were WT1-RFP⁺ (Figure 4D).

Figure 4.

Overexpression of TGF-β1 in the peritoneum induces peritoneal fibrosis in WT1^CreERT2/+;ROSA26^fstdTomato mice. (A) Experimental schema for cohort labeling in WT1^CreERT2/+;ROSA26^fstdTomato mice followed by injection of AdTGF-β1 transgenic expression. Analysis is performed on the peritoneal covering of liver 10 days after AdTGF-β1. (B) Low-powered (upper panel) and high-powered images (lower panel) show large areas of the peritoneal surface devoid of WT1-RFP⁺ MCs. Occasional WT1-RFP⁺ cells (SM fibroblasts, arrows) are found beneath the accumulated αSMA⁺ myofibroblasts. Rare WT1-RFP⁺;αSMA⁺ myofibroblasts (asterisks) are identified. (C) Surviving WT1-RFP⁺ MCs (arrowheads) on the surface of injured peritoneum do not express αSMA. Rare WT1-RFP⁺;αSMA⁺ myofibroblasts (asterisks), amounting to 14.6% of αSMA⁺ myofibroblasts, are seen exclusively within the thickened laminin⁺ scar. WT1-RFP⁺;αSMA^– SM fibroblast (arrows) is located beneath the scar. (D) Images show WT1-RFP⁺;cytokeratin⁺ MCs on the peritoneal surface after AdTGF-β1 (arrowheads). A WT1-RFP⁺;cytokeratin^– SM fibroblast (arrows) is seen. (E) MCs isolated and cultured from peritoneal WT1-RFP⁺;PDGFRβ-APC^– cells of WT1^CreERT2/+;ROSA26^fstdTomato mice after cohort labeling lack expression of αSMA in culture medium alone (CON), whereas activate expression of αSMA in the presence of TGF-β1 for 2 days. Scale bar, 20 μm. Original magnification, ×200 in B upper panel; ×630 in B lower panel and C–E.

To determine whether primary MCs had the capacity to express αSMA in vitro, which some investigators have used to define the mesothelial EMT process, we isolated and cultured WT1-RFP⁺;PDGFRβ-APC^– MCs from WT1^CreERT2/+;ROSA26^fstdTomato mice after tamoxifen pretreatment (Supplemental Figure 8). MCs in culture medium alone lacked detectable αSMA (Figure 4E). In the presence of TGF-β1 for 2 days, cultured MCs activated expression of αSMA (Figure 4E), indicating that although MCs in vivo do not express αSMA, they are capable of activating this protein in vitro. In contrast with the upregulation of mesenchymal genes (Acta2 and Col1a1) by TGF-β1 in cultured MCs, TGF-β1 suppressed Gpm6a (glycoprotein m6a, a MC marker) (Supplemental Figure 9). However, Krt8 (cytokeratin8) was upregulated by TGF-β1 (Supplemental Figure 9).

SM Fibroblasts Were the Major Precursors of Peritoneal Myofibroblasts

To determine whether SM fibroblasts were the primary source of αSMA⁺ myofibroblasts, we generated Col1a2-CreERT^Tg;ROSA26^fstdTomato;Col1a1-GFP^Tg mice in which cells expressing the Col1a2 chain of collagen I could undergo somatic recombination to express the tdTomato RFP permanently after tamoxifen administration (Col1a2-RFP⁺ cells). Dynamic expression of the Col1a1 chain reported by the Col1a1-GFP transgene was used to define SM fibroblasts. The cohort of Col1a1-GFP⁺ SM fibroblasts induced by tamoxifen to express Col1a2-RFP was 64.2% in Col1a2-CreERT^Tg;ROSA26^fstdTomato;Col1a1-GFP^Tg mice (the numbers of Col1a2-RFP⁺ cells and Col1a1-GFP⁺ cells were 2.8 and 4.4 cells per field at ×630 magnification, respectively) (Figure 5, A, B, and D). Two weeks after cohort labeling, hypochlorite injury was induced. After 1 week, the numbers of Col1a2-RFP⁺ cells and Col1a1-GFP⁺ cells increased (11.7 and 18.5 cells per field at ×630 magnification, respectively); the proportion of Col1a1-GFP⁺ cells that coexpressed Col1a2-RFP was 63.0% (Figure 5, A, C, and D). The absence of a significant decrease in the proportion of Col1a1-GFP⁺ cells coexpressing Col1a2-RFP indicated that SM fibroblasts, rather than other cell sources, were the major precursors of collagen-producing cells in the injured peritoneum. To be sure that these SM fibroblasts became αSMA⁺ myofibroblasts, we induced peritoneal fibrosis by hypochlorite injection in Col1a2-CreERT^Tg;ROSA26^fstdTomato mice 2 weeks after tamoxifen treatment to label a 64.2% cohort of SM fibroblasts (Figure 6A). Before hypochlorite injury, Col1a2-RFP⁺ SM fibroblasts were beneath cytokeratin⁺ MCs (Supplemental Figure 10). Seven days after hypochlorite injection, many Col1a2-RFP⁺ cells entered the cell cycle evidenced by expression of Ki67 (Figure 6B). Col1a2-RFP⁺ cells accumulated within the thickened peritoneum and accounted for >60% of αSMA⁺ myofibroblasts, confirming SM fibroblasts as the major precursors of peritoneal myofibroblasts (Figure 6, C and D, Supplemental Figure 11) and also supporting WT1-RFP⁺ SM fibroblasts as the progenitors of WT1-RFP⁺;αSMA⁺ myofibroblasts in WT1^CreERT2/+;ROSA26^fstdTomato mice after peritoneal injury (Figures 3, E and F, and 4, B and C, Supplemental Figures 5–7). Although remesothelialization occurred, no cytokeratin⁺ MCs coexpressed Col1a2-RFP, indicating that the injured peritoneum is not remesothelialized by SM fibroblasts (Figure 6, E and F, Supplemental Figure 12) and also excluding the possible contribution of WT1-RFP⁺ SM fibroblasts to remesothelialization (Figure 3, B and C). Without cohort labeling by tamoxifen, <0.01% of cells expressed RFP before or after hypochlorite injury, indicating that “leaky” somatic recombination could not be responsible for the appearance of Col1a2-RFP⁺ cells after injury (Supplemental Figure 13).

Figure 5.

SM fibroblasts are the major precursors of collagen-producing cells during peritoneal fibrosis after hypochlorite injury. (A) Experimental schema for cohort labeling of Col1a2⁺ cells and hypochlorite peritoneal injury in Col1a2-CreERT^Tg;ROSA26^fstdTomato;Col1a1-GFP^Tg mice. Analysis is performed on the peritoneal covering of liver before and 7 days after hypochlorite injury. (B and C) Images show tamoxifen-induced cohort labeling of Col1a1-GFP⁺ SM fibroblasts with Col1a2-RFP before (B) and 7 days after hypochlorite injury (C). Arrowheads and arrows indicate Col1a1-GFP⁺ with and without coexpression of Col1a2-RFP, respectively. Numerous DAPI⁺ nuclei without labeling of RFP or GFP noted at the surface of normal peritoneum suggest MCs (asterisks) (B). (D) The graph shows the percentages of Col1a1-GFP⁺ cells that coexpress the fate reporter Col1a2-RFP and the proportion of Col1a2-RFP⁺ cells that coexpress Col1a1-GFP before (Con) and 7 days (Hypochlorite) after injury (n=6 per group). DAPI, 4′,6-diamidino-2-phenylindole. Bar, 20 μm. Original magnification, ×630.

Figure 6.

Col1a2-RFP⁺ fate-mapped SM fibroblasts are the major precursors of myofibroblasts in the peritoneum after hypochlorite injury. (A) Experimental schema for cohort labeling and hypochlorite injury in Col1a2-CreERT^Tg;ROSA26^fstdTomato mice. Analysis is performed on the peritoneal covering of liver 7 or 10 days after injury. (B) Active proliferation of Col1a2-RFP⁺ SM fibroblasts after injury is shown by Ki67 expression (arrows). (C) Images show numerous Col1a2-RFP⁺;αSMA⁺ myofibroblasts (arrowheads) after injury. Col1a2-RFP⁺;αSMA^– SM fibroblasts and DAPI⁺;Col1a2-RFP^–;αSMA^– MCs at the peritoneal surface are indicated by arrows and asterisks, respectively. The graph beneath shows the percentage of αSMA⁺ myofibroblasts coexpressing the fate marker Col1a2-RFP (n=6 per group). (D) Three-dimensional images with YZ and XZ stacks show numerous Col1a2-RFP⁺;αSMA⁺ myofibroblasts after injury. Cells are indicated as in C. (E) Images show that cytokeratin⁺ MCs (arrowheads) do not express the fate marker Col1a2-RFP. (F) Three-dimensional images with YZ and XZ stacks show numerous Col1a2-RFP⁺ cells beneath cytokeratin⁺ MCs after injury. Cytokeratin⁺ MCs (arrowheads) do not express the fate marker Col1a2-RFP. DAPI, 4′,6′-diamidino-2-phenylindole. Bar, 20 μm. Original magnification, ×630.

In separate cohorts of Col1a2-CreERT^Tg;ROSA26^fstdTomato mice, we induced a second model of fibrosis by AdTGF-β1 injection (Supplemental Figure 14A). Ten days after viral administration, we observed marked expansion of the Col1a2-RFP⁺ SM fibroblasts, which again accounted for 62.5% of myofibroblasts, indicating in this second model that SM fibroblasts are the major source of myofibroblasts during peritoneal fibrosis (Supplemental Figure 14B).

To determine whether primary SM fibroblasts had the capacity to express αSMA in vitro, we isolated and cultured Col1a1-GFP⁺ SM fibroblasts from normal Col1a1-GFP^Tg mice (Figure 7A). SM fibroblasts in culture medium alone lacked detectable αSMA (Figure 7B). In the presence of TGF-β1 for 2 days, SM fibroblasts activated expression of αSMA (Figure 7B). Quantitative PCR (QPCR) showed the increase of Acta2 and Col1a1 by TGF-β1 (Figure 7C).

Figure 7.

TGF-β1 induces αSMA expression in cultured SM fibroblasts. (A) Peritoneal Col1a1-GFP⁺ SM fibroblasts are isolated and cultured from normal Col1a1-GFP^Tg mice after intraperitoneal retention of trypsin-EDTA. The bright-field image shows SM fibroblasts at passage two cultured in the treated cell culture dish. (B) Col1a1-GFP⁺ SM fibroblasts cultured in the chamber slide lack expression of αSMA in culture medium alone (CON), whereas they activate expression of αSMA in the presence of TGF-β1 for 2 days. (C) Col1a1-GFP⁺ SM fibroblasts are cultured in the presence or absence of TGF-β1 for 24 hours. Quantitative PCR shows that Acta2 and Col1a1 are upregulated by TGF-β1. The expression levels are normalized by Gapdh and expressed as the mean±SEM (n=3). ^†P<0.01. Bar, 25 μm in A; 20 μm in B. Original magnification, ×200 in A; ×630 in B.

Imatinib Reduced Peritoneal Myofibroblasts and Fibrosis

Because SM fibroblasts, not MCs, express PDGFRβ and SM fibroblasts are the precursors of scar-forming myofibroblasts, we blocked PDGFR signaling during the hypochlorite model using the PDGFR tyrosine kinase inhibitor imatinib (Supplemental Figure 15A). Imatinib significantly attenuated peritoneal adhesion, thickening of fibrotic peritoneum, and accumulation of αSMA⁺ myofibroblasts despite no change in the remesothelialization induced by hypochlorite (Supplemental Figure 15, B–E).

Discussion

These studies report an extensive population of Col1a1⁺;Col1a2⁺;PDGFRβ⁺ SM fibroblasts lying below the basal lamina of the normal peritoneal membrane, which produce collagen protein in healthy states. By mapping the fate of cohorts of conditional, somatically labeled SM fibroblasts, our studies show that they transdifferentiate into an expanding population of Col1a1^+;Col1a2⁺;PDGFRβ⁺;αSMA⁺ myofibroblasts in the region of pathologic matrix production. These cells thus are a major source of myofibroblasts in three different models of peritoneal fibrosis and are an important new cellular target for fibrosing diseases of the peritoneum. Our labeling strategies did not label all SM fibroblasts, but we successfully labeled cohorts of approximately 65% of these discrete SM fibroblast cells and they expanded in disease settings to become approximately 65% of all myofibroblasts. The fact that the proportion of cells did not change from healthy to disease states suggests that the vast majority of myofibroblasts in these peritoneal fibrosis models derive from SM fibroblast precursors, rather than from an alternate cell precursor. Although conditional labeling of all SM fibroblasts would be desirable, the labeling of 65% of cells is the technical limit of labeling using these genetic tools in adult mice.

We mapped the fate of MCs using a similar strategy to that for mapping the fate of SM fibroblasts. The transcriptional regulator WT1 is expressed by adult MCs and therefore the conditional Cre enzyme knocked into the WT1 locus was used to map MCs. Our conditional labeling strategy resulted in labeling a cohort of >80% of all MCs. It also labeled a minor population (approximately 15%) of SM fibroblasts. In response to peritoneal injury, the cohort of somatically labeled WT1⁺ cells did not expand. In fact, there was a reduction in cell number. The proportion of myofibroblasts that derived from WT1-labeled cells was approximately 15%, a number consistent with these myofibroblasts deriving from the WT1-labeled SM fibroblasts in the normal adult, rather than from the cohort of WT1-labeled MCs.

Although our data did not support the transition of MCs to myofibroblasts in vivo, MCs were found to express Col1a1 after peritoneal injury. Therefore, the role of MCs in peritoneal fibrosis cannot be excluded because they produce Col1a1 after injury. Although EMT results in differentiation of epithelial cells to migratory mesenchymal cells in the setting of cancer metastasis and development, recent fate mapping experiments in mice using conditional somatic recombination techniques led to a reappraisal of the importance of EMT as an explanation for the appearance of myofibroblasts in multiple organs as well as to a new appreciation for discrete mesenchymal cells known as resident fibroblasts or pericytes as the major origin of myofibroblasts in many tissues, including the liver, muscle, skin, intestine, lung, spinal cord, and kidney.^37–44 Our studies stand in contrast with previous studies that reported MCs as an important source of peritoneal myofibroblasts^8,25–27; rather, our studies support SM fibroblasts as the major source of myofibroblasts during peritoneal fibrosis. Our studies represent an advance of the previous studies that analyzed MC responses to peritoneal injury for two reasons. First, we used conditional labeling of cohorts of cells with robust Cre-expressing mouse lines, whereas previous studies used less robust methods including immunohistochemistry and cell culture in vitro. Second, we described a poorly appreciated population of SM fibroblasts that were not considered in those previous studies. Although variations in experimental conditions cannot be excluded as a cause for differing results, our mapping strategy provides the most robust approach reported to date.

PDGFRβ signaling by myofibroblasts was shown to be pathologic in models of kidney, liver, and lung fibrosis.^43,45,46 The restricted expression of PDGFRβ to SM fibroblasts and myofibroblasts in peritoneal fibrosis suggested that PDGFRβ blockade might be an attractive therapeutic strategy. Blockade of PDGFRβ signaling by the tyrosine kinase–inhibitor imatinib provides supportive evidence that SM fibroblast–derived myofibroblasts are critical cells in fibrogenesis, and provides new avenues for therapeutic discovery. However, we emphasize that the result is far more applicable to EPS than to the vast majority of peritoneal dialysis patients in whom fibrosis is a late clinical event.

Our data indicate that the injured peritoneum was remesothelialized by surviving MCs, not by SM fibroblasts. These studies did not support WT1-labeled SM fibroblasts as stem cells for mesothelial repair. This mechanism of mesothelial regeneration is similar to kidney epithelial regeneration, which was previously shown to be from intrinsic surviving tubular epithelial cells rather than epithelial progenitors.⁴⁷ Because MCs express different growth factor receptors from SM fibroblasts and myofibroblasts (e.g., EGF receptor, as previously reported²⁹), interventions that activate specific receptor signaling for mesothelial repair may promote more effective remesothelialization after injury.

In conclusion, we used comprehensive genetic lineage analysis to clarify the role of MCs and SM fibroblasts in the repair of mesothelium and generation of myofibroblasts during peritoneal fibrosis. We provide lineage tracing evidence that SM fibroblasts are the major myofibroblast precursors in peritoneal fibrosis and surviving MCs are the principal cells for remesothelialization after injury (Figure 8). These findings need to be recapitulated in more clinically relevant models.

Figure 8.

The illustration indicates the major fates of MCs and SM fibroblasts after peritoneal injury.

Concise Methods

Animals

Col1a1-GFP transgenic mice were generated and validated as previously described on the C57BL/6 background whose Col1a1-expressing cells expressed GFP.³⁷ Col1a2-CreERT transgenic mice were generated using a 6-kb Col1a2 enhancer to drive the expression of a cDNA encoding CreERT.⁴⁸ WT1^CreERT2/+ mice obtained from The Jackson Laboratory (Bar Harbor, ME) were generated by knocking a cDNA encoding CreERT2 into the WT1 locus.⁴⁹ B6.Cg-Gt(ROSA)26Sor^{tm14(CAG-tdTomato)Hze}/J reporter mice (referred as ROSA26^fstdTomato) were obtained from The Jackson Laboratory. All studies were carried out under a protocol approved by the Institutional Animal Care and Use Committee of National Taiwan University College of Medicine.

Mouse Models of Peritoneal Fibrosis

Adult mice (aged 14 weeks) were used for induction of peritoneal fibrosis. Hypochlorite-induced peritoneal fibrosis was induced by an intraperitoneal injection of 100 ml/kg body weight of normal saline with 0.05% sodium hypochlorite.^33,34 Dialysis solution–induced peritoneal fibrosis was induced by a daily intraperitoneal injection of 50 ml/kg body weight of dialysis solution by mixing Dianeal containing 4.25% glucose (Baxter Healthcare, Singapore) with 40 mM methylglyoxal (Sigma-Aldrich, St. Louis, MO) to accelerate fibrosis within 2 weeks.³⁵ On the basis of evidence that TGF-β1 was the major profibrotic cytokine in tissue fibrosis, TGF-β1–induced peritoneal fibrosis was induced by an intraperitoneal injection of cesium chloride gradient purified AdTGF-β1 at a dose of 1.5×10⁸ plaque forming units diluted in 100 μl PBS.^17,36,39,50 Control virus at the same dose was used as the control.

Imatinib Administration in a Mouse Model of Peritoneal Fibrosis

Wild-type C57BL/6 mice were administered PBS or imatinib mesylate (50 mg/kg; Novartis Pharmaceuticals Co., Basel, Switzerland) through oral gavage 2 hours before an intraperitoneal injection of 100 ml/kg body weight of normal saline with 0.05% sodium hypochlorite, and then once a day until analysis on day 10 (n=6 for each group). The peritoneal adhesion score was quantified by scoring 1 point for each presence of an adhesion between the abdominal wall and intestine, intestine and intestine, intestine and omentum, omentum and kidney, and kidney and liver according to a previously described method.³⁴ The peritoneal adhesion score was expressed as the mean of total scores for each mouse. Peritoneal membrane thickness and cell numbers of αSMA⁺ myofibroblasts were quantified using Masson’s trichrome stain and immunofluorescence, respectively, in the peritoneal covering of liver using the method described below in the subsection on tissue preparation and histology.

Temporal Induction of Cre Activity by Tamoxifen Administration

A tamoxifen base in olive oil (10 mg/ml) (Sigma-Aldrich) was prepared by sonication. Mice were administered 1 mg daily through oral gavage for 5 days every week at the age of 10 and 11 weeks.^29,48 After washing period for 2 weeks, mice were subjected to peritoneal injury.

Tissue Preparation and Histology

Mouse tissues including the liver, omentum, and anterior abdominal wall were prepared and stained as previously described.^37,38 Primary antibodies against the following proteins were used for immunolabeling in 5 μm-thick cryosections: αSMA-Cy3, αSMA-FITC, laminin, and cytokeratin (Sigma-Aldrich), as well as Ki67 (Abcam, Inc., Cambridge, UK), vimentin and nidogen (Santa Cruz Biotechnology, Santa Cruz, CA), and PDGFRβ (gift from Dr. Stallcup). Fluorescence conjugated secondary antibody labeling (Jackson Immunoresearch Laboratories, West Grove, PA), colabeled with 4′,6-diamidino-2-phenylindole, and mounting with Vectashield were carried out as previously described.^37,38 Conventional and confocal images were taken with an Zeiss Axio Imager A1 Microscope with AxioVision Software and a Zeiss Laser Scanning 780 Microscope with Zen 2011 Software, respectively (Carl Zeiss, Jena, Germany). Images were processed using Adobe Photoshop software. GFP⁺ and RFP⁺ cells were identified by positive nuclear and cytoplasmic fluorescence; Ki67⁺ cells was identified by positive nuclear fluorescence; and αSMA⁺, cytokeratin⁺, vimentin⁺, and PDGFRβ⁺ cells were identified by >75% of the cell area immediately surrounding the nuclei (detected by 4′,6′-diamidino-2-phenylindole) staining positive with fluorescence. To avoid bias caused by focal alterations, the numbers of cells with specific staining in the peritoneal membrane of livers were quantified in 10 sections of ×630 magnification from every 10th section of each mouse (five randomly selected images per section). The peritoneal surface was divided into 25 equal parts and the percentage of the peritoneal surface with cytokeratin⁺ MCs was calculated in 10 sections from every 10th section for each mouse (five randomly selected images per section). Masson’s trichrome stain was performed in 4-μm–thick paraffin sections and the peritoneal membrane thickness on the liver surface was measured in 10 sections from every 10th section for each mouse (five randomly selected images per section). The cell numbers of αSMA⁺ myofibroblasts, percentage of peritoneal surface with cytokeratin⁺ MCs, and peritoneal membrane thickness of each group were averaged and expressed as the mean±SEM.

Purification and Culture of MCs and SM Fibroblasts

Two weeks after the last dose of tamoxifen, WT1^CreERT2/+;ROSA26^fstdTomato mice were injected intraperitoneally with 10 ml of 0.125% trypsin and 0.05% EDTA (Life Technologies, Carlsbad, CA). After digestion for 30 minutes, the intraperitoneal cell suspension was collected by syringe. After centrifugation, cells were resuspended in 5 ml of PBS/1% BSA, and filtered (40 μm). MCs were purified by isolating WT1-RFP⁺;PDGFRβ-APC^– cells (eBioscience, San Diego, CA) using a FACSAria cell sorter (BD Biosciences, San Jose, CA). Col1a1-GFP⁺ SM fibroblasts were purified from peritoneal cells of normal Col1a1-GFP^Tg mice after intraperitoneal injection with trypsin/EDTA as described above. Purified cells were then cultured in DMEM containing 10% FBS. Cells at passages two to four were used for experiments. Cultured cells were stimulated with TGF-β1 (5 ng/ml) or no additional treatment. Two days later, cells were fixed with 4% paraformaldehyde for 15 minutes, washed with PBS, and then labeled with antibodies for αSMA. Total RNA was extracted from cells in the presence or absence of TGF-β1 for 24 hours and QPCR was performed using previously described methods.⁵⁰ The specific primer pairs used in QPCR are listed in Supplemental Table 1.

Detection of TGF-β1 in Peritoneal Effluent

Ten days after intraperitoneal injection of adenovirus, 3 ml of PBS was instilled in the peritoneum, and samples were collected 15 minutes later. The fluid was centrifuged at 1500 rpm for 5 minutes and the supernatant was frozen at −20°C. This effluent was assayed for total TGF-β1 concentration using ELISA according to the manufacturer’s instructions (R&D Systems, Minneapolis, MN).

Statistical Analyses

Data are expressed as the mean±SEM. Statistical analyses were carried out using GraphPad Prism software (GraphPad Software, La Jolla, CA). The statistical significance was evaluated by one-way ANOVA. P<0.05 was considered significant.

Disclosures

J.S.D. is on the Scientific Advisory Board for Regulus Therapeutics and Promedior Inc., a cofounder of Muregen LLC.