An immortalized cell line derived from renal erythropoietin-producing (REP) cells demonstrates their potential to transform into myofibroblasts

The erythroid growth factor erythropoietin (Epo) is produced by renal interstitial fibroblasts, called REP (renal Epo-producing) cells, in a hypoxia-inducible manner. In chronic kidney disease (CKD), REP cells lose their Epo-production ability, leading to renal anaemia. Concurrently, REP cells are suggested to be transformed into myofibroblasts, which are the major player of renal fibrosis. Although establishment of cultured cell lines derived from REP cells has been a long-term challenge, we here successfully established a REP-cell-derived immortalized and cultivable cell line (Replic cells) by using a genetically modified mouse line. Replic cells exhibited myofibroblastic phenotypes and lost their Epo-production ability, reflecting the situation in renal fibrosis. Additionally, we found that cell-autonomous TGFβ signalling contributes to maintenance of the myofibroblastic features of Replic cells. Furthermore, the promoters of genes for Epo and HIF2α, a major activator of Epo gene expression, were highly methylated in Replic cells. Thus, these results strongly support our contention that REP cells are the origin of myofibroblasts in fibrotic kidneys and demonstrate that cell-autonomous TGFβ signalling and epigenetic silencing are involved in renal fibrosis and renal anaemia, respectively, in CKD. The Replic cell line is a useful tool to further investigate the molecular mechanisms underlying renal fibrosis.


Cultivation and immortalization of REP cells isolated from ISAM-REC mice.
We previously established a gene-modified mouse line in which REP cells are efficiently labelled with tdTomato red fluorescent protein expression 13 . In this mouse line, referred to as ISAM-REC mice (Epo GFP/GFP :Rosa26 CAG-LSL-tdTomato/WT :Tg 3.3K-Epo :Tg EpoCre genotype) 14 , the expression of transgenic Cre recombinase under the control of the Epo gene regulatory region is highly induced by severe anaemic conditions due to Epo deficiency, and most REP cells permanently express tdTomato as a marker for Cre-mediated recombination without any treatment 14,23 . Thus, tdTomato-positive cells from ISAM-REC kidneys were applied for ex vivo cultivation and immortalization to generate cell lines derived from REP cells.
First, tdTomato-positive cells were isolated from ISAM-REC kidney cell suspensions by using a cell sorter, and the cells were incubated with mesenchymal stem cell growth medium (MSCM). However, no cells survived after the 10-day ex vivo cultivation, suggesting that cell-cell communications and/or soluble factors secreted by kidney cells other than REP cells are required for REP cell expansion and maintenance. Therefore, the cell suspensions were directly incubated without cell sorting. After a week of cultivation, we observed that tdTomato-positive cells grew with tdTomato-negative cells, attaching to the bottoms of culture dishes (Fig. 1a).
Because it was considered that the 1-week cultivation let the kidney cells adapt to the culture conditions, the cells were integrated with a lentivirus vector expressing the human HRAS gene bearing an oncogenic mutation in order to immortalize the cells. After expansion of the infected cells for 2 weeks, tdTomato-positive REP cells in the dishes were sorted and subjected to pure cultivation. Finally, a single colony of tdTomato-positive cells was developed (Fig. 1b). It was possible for the colony-derived cells in culture dishes to grow with MSCM for more than 1 year, to recover from freeze-thawing, and to be passaged more than 10 times. We thus concluded that a REP-cell-derived cell line was established and named the cells Replic (REP cell-lineage immortalized and cultivable) cells.

Growth of Replic cells. Replic cells and natural REP cells in ISAM-REC mice shared a fibroblast-like shape
that exhibited bipolar stretching with bright fluorescence of tdTomato (Fig. 1c) 12 . The cell shape was narrow and stretched with many processes when the cells were cultured with MSCM rather than with Dulbecco's modified Eagle's medium (DMEM) supplemented with 10% foetal bovine serum (FBS) (Fig. 1c). This observation suggests that MSCM is a more appropriate culture medium than DMEM for maintaining the features of in vivo REP cells under ex vivo culture conditions because the REP cells show a sharp and extended morphology with many processes in the renal interstitia 11,12 . Therefore, MSCM was used for characterization of Replic cells in this study. Replic cells proliferated significantly in MSCM rather than in DMEM and doubled daily in the logarithmic growth phase on normal polystyrene dishes without any coating (Fig. 1d). Although the contents of MSCM are veiled by the manufacturer, the cell-growth-optimized MSCM likely promotes proliferation of Replic cells compared with the basic medium.
Genotyping PCR confirmed that the immortality and proliferative activity of Replic cells were supported by integration of the oncogenic mutant HRAS gene (Fig. 1e). The ISAM-REC kidney origin of Replic cells was confirmed by PCR data demonstrating that Replic cells and ISAM-REC mice commonly harboured the Cre, tdTomato, and GFP genes but lacked the Epo gene (Fig. 1e). From these results, we concluded that the first REP-cell-derived cell line has been established and that the cells vigorously proliferate, maintaining their fibroblastic features under culture conditions optimized for mesenchymal stem cells.
Fibroblastic phenotype of Replic cells. Flow cytometry of Replic cells detected high-level tdTomato fluorescence in every Replic cell (Fig. 2a), showing the clonality of the cell line. The data also revealed that compared www.nature.com/scientificreports www.nature.com/scientificreports/ to wild-type mouse embryonic fibroblasts (MEFs), Replic cells strongly expressed CD73 (ecto-5′-nucleotidase), a fibroblastic surface marker of REP cells in mouse kidneys 12,24,25 . Consistently, gene expression analyses showed that Replic cells expressed the gene for CD73 (Nt5e) at higher levels than MEFs did (Fig. 2b).
Replic cells did not express the Kdr, Cdh1, and Nphs1 genes, which are lineage markers for vascular endothelial cells, tubular epithelial cells, and podocytes, respectively (Fig. 2c). Taken together, the findings revealed that the cell line originated from tdTomato + CD73 + REP cells of the ISAM-REC kidney but not from other lineage . On Day 0, 1.0 × 10 4 cells were seeded onto 3.5-cm dishes with 2.5 mL medium, and cell numbers were counted daily. The data are shown as the means ± standard errors. n = 3 for each group. **p < 0.01 by two-tailed and unpaired Student's t tests. (e) Genomic PCR of Replic cells compared with that of organs from wild-type C57BL/6 (WT) and ISAM-REC mice. N, a non-template negative control. The mRNA expression levels of marker genes for fibroblasts (b) and non-fibroblastic renal cells (c) in Replic cells were compared to those in wild-type (WT) kidneys and MEFs. Arrows indicate undetectable levels. Cell culture was conducted with MSCM. Ten biologically independent samples for each cell line were analysed. The average expression level of WT kidneys (n = 3) was set at 1.0, and error bars indicate standard errors. **p < 0.01 by multiple comparisons using one-way ANOVA with Tukey-Kramer tests.
Epo gene activity and hypoxic response of Replic cells. Because the Epo gene of ISAM-REC mice was homozygously replaced with the gene for green fluorescent protein (GFP), the Epo gene activity of Replic cells derived from an ISAM-REC mouse was able to be evaluated by measuring the amounts of mRNA transcribed from the EpoGFP recombinant gene 13,27 . In REP cells, Epo gene expression is induced in a hypoxia-inducible manner via stabilization and activation of HIF2α 15 . Therefore, we investigated EpoGFP mRNA levels in Replic cells incubated under hypoxic (1% O 2 ) conditions.
In ISAM-REC kidneys, EpoGFP mRNA was highly expressed by severe anaemia, in agreement with our previous report (Fig. 3a,b) 27 . However, EpoGFP mRNA expression was undetectable in Replic cells regardless of the presence of hypoxic stimuli, while the 24-hour hypoxic exposure induced expression of the Vegfa and Slc2a1 genes (encoding vascular endothelial cell growth factor A and glucose transporter 1, respectively), both of which are well known as HIF target genes (Fig. 3a) 28,29 . Additionally, expression of the gene for HIF2α (Epas1) was undetectable in both normoxic and hypoxic Replic cells, whereas Hif1a mRNA was consistently expressed under both conditions (Fig. 3a).
We then tested effects of the PHDi GSK360A 27 on EpoGFP mRNA expression of Replic cells. The results showed that, similar to hypoxic exposure, GSK360A treatment induced neither EpoGFP nor Epas1 expression (Fig. 3b). Because GSK360A enhanced the expression of the Vegfa and Slc2a1 genes (Fig. 3b), it was demonstrated that GSK360A activated HIF signalling in Replic cells. Indeed, HIF1α protein accumulated in Replic cells after GSK360A treatment, and hypoxic exposure also induced HIF1α accumulation (Fig. 3c). Additionally, GSK360A treatment resulted in the accumulation of both HIF1α and HIF2α in a Hep3B human hepatoma cell line (Fig. 3c). However, in accordance with the data showing the absence of Epas1 mRNA in Replic cells, HIF2α protein was undetectable in the cells even in the presence of GSK 360A or a hypoxic stimulus (Fig. 3c).
DNA methylation in the promoters of genes for Epo and HIF2α. We surmised that epigenetic silencing blunted Epo and Epas1 gene expression in Replic cells. It has been reported that DNA methylation in the Epo gene promoter region is involved in the loss of Epo production in fibrotic kidneys 30 . Bisulfite genomic sequence analyses of the Epo and Epas1 gene promoters in Replic cells were conducted by a comparison with ISAM-REC livers, in which the Epo gene promoter is active in the majority of cells 13 .
The results revealed that the proximal promoter regions of both genes were significantly methylated in Replic cells, whereas a few CpG methylations were detected in the promoters of the ISAM-REC livers (Fig. 4a). These data suggest that the expression of Epo and HIF2α in Replic cells was silenced by DNA methylation in their gene promoters. Additionally, it was thought that Replic cells may undergo transformation to myofibroblasts, in which the Epo gene is methylated and inactivated 30 , maintaining their fibroblastic features.
A 1-week incubation with 5-aza-2′-deoxycytidine (5-Aza), an inhibitor of DNA methyltransferase 1 (DNMT1), which is essential for maintenance of DNA methylation patterns beyond mitosis 31 , did not alter DNA methylation of the Epo gene promoter region, whereas the upstream enhancer region (the -8kHRE-flanking region) of the Epo gene 32 was significantly demethylated by the inhibitor in Replic cells ( Supplementary Fig. 1). Therefore, the contribution of continuous de novo DNA methylation to the hyper-methylation of Epo and Epas1 promoters in Replic cells was assumed 33 . We then tested whether exogenous HIF2α overexpression could activate EpoGFP gene expression in Replic cells. Constitutively active HIF2α, in which PHD-target prolyl residues were replaced with alanine residues 34 , was transiently overexpressed in Replic cells (Fig. 4b). However, EpoGFP expression was not induced (Fig. 4c), indicating that exogenous HIF2α overexpression was insufficient to reactivate the Epo-production ability in Replic cells. Thus, we propose that Replic cells lose their Epo-production ability due to continuous DNA methylation in the Epo gene promoter, which blocks the transcriptional activation signal of HIF2α.

Myofibroblastic property of Replic cells. Because REP cells transform to myofibroblasts in injured kid-
neys with loss of the Epo-production ability and DNA hyper-methylation of the Epo gene promoter 6,7,30 , the myofibroblastic property of Replic cells was investigated by a comparison to fibrotic kidneys of normal mice subjected to unilateral ureteral obstruction (UUO) for 14 days. Ureteral obstruction is an established model of renal fibrosis 7,35 and Masson's trichrome staining of kidney sections confirmed that our ureteral obstruction surgery induced renal fibrosis ( Supplementary Fig. 2a) 15 . Although MEFs were used in this experiment as the normal fibroblast control, strong expression of myofibroblastic genes was reported in MEFs 36 . Gene expression analyses of Replic cells revealed that the cells highly expressed the Acta2 and Fn1 genes, both of which are known to be expressed in myofibroblasts 37,38 , compared to the expression levels in MEFs or injured kidneys (Fig. 5a). The data clearly indicated that Replic cells exhibit myofibroblastic properties.
TGFβ is known to be secreted by fibrotic kidneys and to promote organ fibrosis 7,39-41 . In fact, our analyses clearly demonstrated that mRNA expression of the gene for TGFβ (Tgfb1) was dramatically induced by kidney injury (Fig. 5a). Notably, Replic cells expressed the Tgfb1 gene at a comparable level to that in injured kidneys (Fig. 5a). The expression levels of genes for the major TGFβ receptors TGFβR1 (Tgfbr1) and TGFβR2 (Tgfbr2) were also detected in Replic cells, while other TGFβ superfamily receptor genes (Acvrl1 and Bmpr1b) were not expressed (Fig. 5b). Flow cytometry confirmed that TGFβR2 is apparently expressed on the cell surface of the majority of Replic cells (Fig. 5c). Additionally, the gene expression levels of these receptors were significantly higher in injured kidneys than in contralateral kidneys of UUO-subjected mice ( Supplementary Fig. 2b), supporting the idea that TGFβ signalling is involved in progression of renal fibrosis 42 . These data led us to assume that cell-autonomous TGFβ signalling is involved in the myofibroblastic property of Replic cells.
Based on the cell shape described above, we proposed that Replic cells seemed to differentiate into myofibroblasts by changing the culture medium from MSCM to DMEM (see Fig. 1c). This proposal was supported by the gene expression data showing that expression levels of the myofibroblastic marker genes Acta2 and Fn1 were significantly higher in Replic cells cultured with DMEM than in Replic cells cultured with MSCM (Fig. 5d). In contrast, the expression levels of genes for TGFβ and TGFβ receptors were not responsive to changes in the www.nature.com/scientificreports www.nature.com/scientificreports/ Activation of cell-autonomous TGFβ signalling in Replic cells. In the TGFβ signal, TGFβ binding to a TGFβR1/TGFβR2 heterodimeric receptor results in phosphorylation of Smad2 and/or Smad3 transcription factors followed by induction of Acta2 and Fn1 gene expression 44 . Consistent with the myofibroblastic gene expression profiles (Fig. 5a,d), Smad2 and/or Smad3 (Smad2/3) were highly phosphorylated in Replic cells compared to MEFs when cells were cultured with DMEM, and the phosphorylation was reduced by cultivation with MSCM (Fig. 6a). Consequently, cell-autonomous TGFβ signalling through TGFβR1/TGFβR2-Smad2/3 was considered one of the major pathways promoting myofibroblastic transformation of Replic cells.
To confirm TGFβ production by Replic cells, TGFβ concentrations in the supernatant incubated with Replic cells for 24 hours were measured by enzyme-linked immunosorbent assay (ELISA). The data demonstrated that Replic cells produce significantly higher amounts of TGFβ than MEFs (Fig. 6b), although these cell lines expressed comparable levels of Tgfb1 mRNA (Fig. 5a). ELISA data from mouse plasma showed that plasma TGFβ concentrations are elevated in mice suffering from renal fibrosis ( Supplementary Fig. 2c). These data suggest that TGFβ is secreted by fibrotic kidney cells including myofibroblast-transformed REP cells.
We further performed an antibody-based cytokine array on the culture supernatants. In agreement with the ELISA data, intense signals were detected in samples from Replic cells compared to those from MEFs (Fig. 6c, www.nature.com/scientificreports www.nature.com/scientificreports/ Supplementary Fig. 3a,b). The data also showed that Replic cells produced proinflammatory cytokines such as MCP1 (monocyte chemotactic protein 1, also known as CCL2) and CXCL5 (C-X-C chemokine ligand 5) at higher levels than MEFs (Fig. 6c, Supplementary Fig. 3a,b). These results suggest that Replic cells acquire myofibroblastic properties related to cell-autonomous TGFβ signalling during isolation, cultivation, and/or immortalization of fibroblastic REP cells.

TGFβ-dependent myofibroblastic property of Replic cells.
To investigate the necessity of cell-autonomous TGFβ signalling for the myofibroblastic property of Replic cells, Replic cells were incubated with SB431542, a kinase inhibitor specific for TGFβR1, ALK4 (activin receptor-like kinase 4), and ALK7 45 . The www.nature.com/scientificreports www.nature.com/scientificreports/ mRNA expression levels of the Acta2 and Fn1 genes were markedly decreased by SB431542, while Tgfb1 mRNA expression was unaltered (Fig. 7a). In addition to SB431542, TGFβ neutralizing antibody (TGFβ N-Ab) reduced Acta2 gene expression in Replic cells (Fig. 7b). Although it was predicted that the original fibroblastic property of REP cells was, at least in part, restored by inhibition of TGFβ signalling in Replic cells, EpoGFP mRNA expression was still undetectable in Replic cells cultured with SB431542 for 24 hours (Fig. 7a). Inhibition of TGFβ signalling by SB431542 was confirmed by immunoblotting data showing a decrease in phosphorylated Smad2/3 (p-Smad2/3, Fig. 7c).
Finally, we tested whether the Epo-production ability could be restored by simultaneous supplementation with SB431542, GSK360A, and 5-Aza. As the result, the inhibitor cocktail did not activate EpoGFP mRNA expression, while the expression of other HIF target genes was significantly induced (Serpine1 and Slc2a1, Fig. 7d) 29,46 . The Epas1 gene remained silent under the cocktail presence, although Hif1a mRNA was constantly expressed in Replic cells with or without the cocktail (Fig. 7d). Therefore, expression of the Serpine1 and Slc2a1 genes in Replic cells is thought to be driven through HIF1α activation by the inhibitor cocktail, which includes a PHDi. Consequently, we propose that Replic cells would be useful for elucidating the mechanisms of the myofibroblastic transformation of REP cells even though the cells may irreversibly lose their Epo-production ability.

Discussion
Because REP cells play a central role in renal anaemia and renal fibrosis in CKD patients, cell lines derived from REP cells have long been waited for use in studies elucidating the molecular pathophysiology of REP cells. In this study, we successfully generated a REP-cell-derived cell line "Replic cells" for the first time. Because Replic cells exhibit myofibroblastic features and defects in Epo production, we are certain that Replic cells will allow us to understand the molecular mechanisms underlying the development of renal anaemia and fibrosis during kidney disease progression.
We previously reported that REP cells are inactive in proliferation and energy metabolism in healthy kidneys, whereas kidney injuries stimulate the proliferation of REP cells with Ki67 expression 7,15 . In this study, we observed that REP cells isolated from single-cell suspensions of mouse kidneys failed to survive for more than 1 week under any culture conditions tested here. Then, we tested mixed cell cultures of renal cells and found that REP cells require intercellular communication with renal cells for survival and growth in vitro. During the mixed culture, REP cells gained proliferative activity. Although our preliminary experiments showed that the www.nature.com/scientificreports www.nature.com/scientificreports/ primary REP cells in the mixed culture for 3 days expressed the EpoGFP gene in response to the PHDi, the cells lost their Epo-production ability within 1 week. Thus, it is proposed that the renal microenvironment, including tubulointerstitial and/or vasculointerstitial interactions, keeps REP cells quiescent in healthy kidneys and that damage to tubules and/or vessels in injured kidneys or in culture dishes induces the transformation of REP cells through activating proliferation and inactivating Epo production. Additionally, quiescence may be required for the Epo-producing ability of REP cells. www.nature.com/scientificreports www.nature.com/scientificreports/ Replic cells preferentially grow and maintain their sharp configuration in a special culture medium optimized for mesenchymal stem cell (MSC) growth rather than in a general medium. In addition, Replic cells and REP cells commonly express CD73 at high levels 12,24,25 ; CD73 is an MSC marker 47 . Thus, it is reasonable to conclude that REP cells include MSCs. Indeed, there are reports demonstrating that MSCs play essential roles in kidney development by secreting growth factors 48 , that renal interstitial fibroblasts in healthy kidneys and renal myofibroblasts in injured kidneys are derivatives of MSCs 5 , and that cells bearing MSC properties exist in kidneys for tissue repair 25,49,50 . Importantly, a subset of MSCs isolated from adult mouse kidneys expresses Epo 25 .
A variety of cell types have been reported as origins of myofibroblasts emerging in kidney diseases. For instance, interstitial fibroblasts such as REP cells [5][6][7] , bone marrow-derived cells 8,9 , tubular epithelial cells 3 and vascular endothelial cells 4 . The myofibroblastic features of Replic cells clearly demonstrate that REP cells are one of the sources of renal myofibroblasts. This conclusion coincides with our previous articles reporting that the majority of renal myofibroblasts in injured mouse kidneys originate from REP cells 6,7 .
TGFβ is known as a master regulator of fibrosis 2,51 , but its source is controversial in renal fibrosis. Injured tubular epithelial cells, endothelial cells, and macrophages are major TGFβ-producing cell candidates in fibrotic kidneys [52][53][54] . On the basis of data from Replic cells, we further argue that cell-autonomous TGFβ signalling promotes the myofibroblastic transformation of REP cells. We and others previously demonstrated that activation of TGFβ signalling, which is defined by the phosphorylation of Smad2/3, is detected in MF-REP cells in injured kidneys 7,30 . However, further studies using Replic cells are required to identify triggers of cell-autonomous TGFβ induction in myofibroblasts of injured kidneys. In addition to TGFβ, MCP1 and CXCL5 are secreted at high levels from Replic cells, and both chemokines have been suspected to be secreted from injured kidneys and to contribute to inflammation and fibrosis by recruiting myeloid cells from the bone marrow 55,56 . Analyses of Replic cells will allow us to deeply understand the molecular mechanism of organ fibrosis.
Hypoxia-inducible Epo production in REP cells is regulated at the gene transcription level 23 . In Replic cells exhibiting features of MF-REP cells, the promoter region of the Epo gene is highly methylated. Because DNA methylation in gene promoters blocks the association of basic transcription factors and RNA polymerases 57 , it is proposed that REP cells lose their Epo-production ability due to Epo gene silencing by DNA methylation during the transformation to myofibroblasts in injured kidneys. Similarly, a previous report demonstrated that deficient Epo production is caused by hyper-methylation of the Epo gene promoter in renal myofibroblasts 30 although this study could not exclude the possibility that H-RAS overexpression but not myofibroblastic transformation directly mediated the DNA methylation. This content is confirmed by our experiment showing that the Epo-production ability of Replic cells was not induced by overexpression of HIF2α, a major activator of Epo gene transcription in REP cells 15 . In this study, we revealed that the expression of HIF2α at the mRNA and protein levels is blunted due to hyper-methylation of the Epas1 promoter region in Replic cells. Therefore, epigenetic silencing in the Epo and Epas1 loci is involved in the development of renal anaemia.
We previously reported that the Epo-production ability is retained beyond myofibroblastic transformation by the constitutive activation of HIF2α in REP cells before injury 15 . During Replic cell establishment under normal air conditions, HIF2α was probably inactivated, and the Epo and Epas1 promoter regions were methylated. We then propose that active HIF2α or active Epo gene transcription may protect the Epo gene promoter from hyper-methylation in renal myofibroblasts. Overexpression of the constitutively active H-RAS may also contribute to the DNA-methylation-mediated gene silencing, because RAS signalling has been reported to induce DNA methylation primarily through the MAPK pathway 58,59 . Although the molecular mechanism of gene-specific methylation in renal myofibroblasts has not been elucidated, TGFβ signalling is supposed to induce DNA methylation through enhancing the expression of DNA methyltransferases in renal myofibroblasts 15,30 . This study proposes that de novo methylation of specific genomic regions, including the Epo and Epas1 gene promoters, is constitutively induced by cell-autonomous TGFβ production in renal myofibroblasts.
Using mouse models, we previously demonstrated that MF-REP cells restore REP cell properties by ameliorating kidney injury 7 . Although this study failed to recover the Epo-production ability in Replic cells by TGFβ signal blockage, the inhibition of TGFβ signalling partially debilitated the myofibroblastic phenotype of Replic cells in the gene expression profile. Further attempts to restore REP-cell properties in Replic cells are directly linked to the identification of therapeutic targets for both renal fibrosis and anaemia in CKD. To treat renal anaemia, PHDi is effective for inducing renal Epo production in CKD patients, including end-stage renal disease patients 20 . However, the Epo gene in Replic cells is not responsive to PHDi. Our preliminary data show that the Epo-induction effect of PHDi is gradually weakened with the progression of kidney disease in mice. These observations suggest that REP cells may transform into myofibroblasts in a stepwise manner during kidney disease progression and that Replic cells may correspond to severely myofibroblast-transformed REP cells, in which Epo and HIF2α expressions are epigenetically silenced. Finally, we conclude that Replic cells will provide significant information on the molecular pathology of CKD, from which more than 10% of people suffer worldwide and for which no plausible treatments have been established due to a poor understanding of the mechanisms of CKD pathogenesis.

Isolation and immortalization of REP cells.
Kidneys of ISAM-REC mice were chopped and digested with 1.5 mg mL −1 of collagenase type II (Worthington) for 60 minutes at 37 °C followed by dissociation using a gentleMACS dissociator (Miltenyi Biotec). The cell suspension was incubated for one week, and adherent cells were infected with a lentivirus vector expressing the G12V oncogenic mutant of human H-RAS (abm). From the immortalized cells, REP-cell-derived cells positive for tdTomato expression were isolated using a FACS Jazz cell sorter with FACS Diva software (Becton Dickinson). In the isolated cell culture, a single colony was developed and expanded as the Replic cell line.

Transient transfection.
A pCMX-HIF2αCA plasmid expressing mouse HIF2α, of which 2 specific prolyl residues were replaced with alanine residues to become resistant to PHD-mediated hydroxylation and constitutively active in mammalian cells 34 , and pEGFP-N1 (TaKaRa) expressing enhanced GFP were used for transient transfection assays. Cells with 80% confluency in a well of 6-well plates were transfected with a total of 600 ng of pCMX-HIF2αCA and/or pEGFP-N1 using Lipofectamine 2000 reagent (Thermo Fisher Scientific) and harvested 48 hours after transfection. Transfection efficiency was estimated by measuring the ratio of GFP-expressing cells in total cells using Tali.
Compounds. SB431542 (Adooq Bioscience), GSK360A (Toronto Research Chemicals), and 5-Aza (Sigma-Aldrich) were resolved in dimethyl sulfoxide (DMSO, Nacalai Tesque) at 10 mM, 50 mM, and 5 mM concentrations, respectively. TGFβ N-Ab (1D11; Novus Biologicals) was dissolved in culture medium at a 2 mg mL −1 concentration just before administration. The resolved compounds were administered to the cells at a 1:1000 dilution. DMSO or culture medium was used as the vehicle control.
Genomic PCR. Genomic DNA was purified with phenol/chloroform/isoamyl alcohol (Nakalai Tesque) from cells or mouse organs. Genotype PCR was performed with primers listed in Supplementary Table 1 using Ex-Taq polymerase (TaKaRa). Amplified DNA fragments were electrophoresed on 2% agarose gels containing ethidium bromide (Nakalai Tesque) and detected using an E-BOX imaging system with ultra-violet light (VILVER).

Reverse transcription-quantitative PCR (RT-qPCR).
Total RNA was extracted and purified using ISOGEN (Nippon-Gene) immediately after cells were removed from incubators. For control samples, total RNA was purified from whole mouse kidneys using a Precellys 24-bead homogenizer (Bertin) and ISOGEN (Nippon-Gene). cDNA was synthesized using the reverse transcriptase from a Superscript III kit (Thermo Fisher Scientific). Quantitative PCR was performed with primers and probes listed in Supplementary Table 1 using FastStart SYBR green master mix (Roche) or TaqMan master mix (Roche) with a LightCycler 96 system (Roche). The expression levels of housekeeping Hprt (hypoxanthine phosphoribosyl transferase) mRNA were used as internal standards.
Flow cytometry. Cells in culture plates were dissociated with accutase (Nakalai Tesque). Single-cell suspensions were stained with biotin-conjugated CD73 antibody (Thermo Fisher Scientific) or biotin-conjugated TGFβR2 antibody (BAF532; R&D Systems) for 30 minutes on ice and then washed and stained with streptavidin-APC for 30 minutes. After washing, flow cytometry was performed using FACS Jazz.
Bisulfite sequencing. Genomic DNA was purified from Replic cells and chopped mouse organs incubated with proteinase K (Nakalai Tesque) overnight, and bisulfite conversion of genomic DNA was performed using an EpiTect Bisulfite Kit (Qiagen). Bisulfite-converted DNA of the Epo and Epas1 gene promoters was amplified