The methyltransferase SET9 regulates TGFB1 activation of renal fibroblasts via interaction with SMAD3

The earliest histological feature of chronic kidney disease (CKD) is the aberrant accumulation of activated myofibroblasts, the cell type responsible for replacement of normal kidney tissue with fibrosis (Meran and Steadman, 2011). TGFB1 stimulates myofibroblast migration to the site of tissue injury and is a key driver of myofibroblast differentiation, characterised by their expression of α-smooth muscle actin (α-SMA; also known as ACTA2) (Bottinger and Bitzer, 2002; Massagué, 2012). Upon TGFB1 stimulation, α-SMA is incorporated into intracellular stress fibres, conferring contractile properties upon the myofibroblast (Hinz et al., 2001). TGFB1 also stimulates fibrillar collagen-rich extracellular matrix (ECM) production from the myofibroblast (Wynn and Ramalingam, 2012; Zeisberg and Kalluri, 2013). Stress fibres containing α-SMA indirectly transmit mechanical tension through this ECM, maintaining myofibroblast differentiation and facilitating tissue retraction (Tomasek et al., 2002). During normal healing after tissue injury, these events allow wound contraction and resolution to occur. However, in the context of CKD and other fibrotic diseases, persistent myofibroblast activity, driven by TGFB1, results in pathological development of fibrosis.

TGFB1 is able to signal through serine-threonine kinase transmembrane receptors (type I and type II receptor heterodimers), which directly phosphorylate the downstream intracellular receptor-regulated SMAD proteins (R-SMADs) SMAD2 and SMAD3 (Shi and Massagué, 2003). R-SMAD heterocomplexes, along with SMAD4, are then translocated to the nucleus in order to regulate transcription of fibrotic genes such as α-SMA and collagens, by recruitment of transcriptional co-activators (Inoue et al., 2007). This pathway can be suppressed by ubiquitin-mediated destruction of TGFB receptors and R-SMADs, including SMAD7 (Fukuchi et al., 2001; Kavsak et al., 2000; Yan et al., 2016). Recent data show that SMAD ubiquitylation is reversible (Inui et al., 2011; Nan et al., 2016). Several studies have shown that TGFB1–SMAD3 signalling is required for renal fibrosis in mouse models of CKD (Fujimoto et al., 2003; Inazaki et al., 2004; Liu et al., 2012; Sato et al., 2003; Wang et al., 2007), raising the possibility of targeting this pathway to treat CKD.

SET9 (also known as SETD7) belongs to the evolutionarily conserved Su(var) enhancer of zeste and trithorax domain family containing lysine methyltransferases that are important in transcriptional regulation, and was originally shown to methylate lysine 4 in isolated histone H3 (H3K4) (Wang et al., 2001). Interestingly, however, SET9 only weakly methylates H3K4 in intact nucleosomes, and cells from the SET9-knockout mouse contain unaltered total methyl H3K4 levels (Lehnertz et al., 2011) pointing towards functional redundancy between methyltransferases and hinting at a role for SET9 elsewhere. SET9 was subsequently shown to regulate various transcription factors involved in differentiation, including p53, AR, ERα, NFκB, DNMT1, MyoD, FoxO3 and TAF10 (Calnan et al., 2012; Chuikov et al., 2004; Ea and Baltimore, 2009; Esteve et al., 2009; Gaughan et al., 2011; Kouskouti et al., 2004; Subramanian et al., 2008; Tao et al., 2011) and non-transcriptional proteins (Oudhoff et al., 2016; Shan et al., 2013). Nevertheless, because SET9 appears to be dispensable for p53 function in vivo and SET9-deficient mice undergo normal development (Campaner et al., 2011; Lehnertz et al., 2011) questions have again been raised over the true role of this methyltransferase (Keating and El-Osta, 2013).

Recent work has demonstrated that SET9 is involved in TGFB1-mediated ECM production in pulmonary fibrosis by a mechanism involving direct lysine methylation, and ubiquitin-mediated degradation of SMAD7 (Elkouris et al., 2016). Complementary work has shown that SET9 regulates ECM gene expression at a more downstream transcriptional level, through histone H3 monomethylation (Sun et al., 2010) and that SET9 inhibition may reduce fibrosis in a mouse model of ureteric obstruction (Sasaki et al., 2016). Whilst these studies define an important role for SET9 in ECM production, the role of SET9 in governing key fibroblast activities, including myofibroblast differentiation, wound healing and collagen contractility, remains unexplored. Additionally, the possibility that SET9 regulates R-SMADs also remains unexplored.

Here, we demonstrate a novel interaction between SET9 and SMAD3, the key downstream R-SMAD in the TGFB1 pathway that is required for renal fibrosis. We show that SET9 regulates SMAD3 through multiple molecular mechanisms and influences important TGFB1-mediated processes including collagen contractility and α-SMA-driven differentiation of primary human kidney fibroblasts. We propose that pharmacological inhibition of SET9 could be important in the treatment of CKD.

Given the uncertainty over SET9 function, we examined global gene expression in Set9^−/− mouse embryonic fibroblasts (MEFs) compared with MEFs from wild-type (Set9^+/+) mice. Under normal growth conditions we observed that 760 genes were expressed at 1.5-fold or higher (range, 1.5- to 67-fold) in Set9^+/+ compared with Set9^−/− MEFs (Table S2). Gene set enrichment analysis (GSEA) of these 760 genes produced cytoskeleton, wound healing and extracellular matrix amongst the 6 highest-ranked annotations – processes that are known to be regulated by TGFB1 (Table 1). We additionally treated MEFs with TGFB1 for 48 h. Under these conditions the number of differentially expressed genes in Set9^+/+ compared with Set9^−/− MEFs increased to 2434 (range 1.5- to 66.53-fold, Table S3). Of the 760 genes differentially expressed under normal growth conditions 644 genes were also found to be differentially expressed among the 2343 genes altered upon TGFB1 treatment, demonstrating consistent effects of loss of SET9. GSEA of the 2343 SET9-regulated genes increased gene numbers under the previously listed annotations, in particular cytoskeleton, ubiquitin conjugation and extracellular matrix (Table 1). Additional data from a 24 h TGFB1 time point indicated that the highest ranking annotations remain enriched in response to TGFB1 treatment, including extracellular matrix, ubiquitin conjugation (the highest two ranking of all annotations at 24 h TGFB1 treatment) and cytoskeleton (Tables S1,S4). Overall, we conclude that SET9 is required for the fibroblast response to TGFB1.

We next examined whether SET9 could regulate TGFB1 signalling through downstream signalling mechanisms. We studied SMAD3, the principal factor that propagates TGFB1 signalling to the nucleus, in HKC-8 cells that are known to have an intact TGFB1 pathway. Immunoprecipitation of endogenous SET9 resulted in co-immunoprecipitation of SMAD3, in contrast to immunoprecipitation with control immunoglobulins (Fig. 1A). We observed enrichment of the SMAD3–SET9 interaction upon exogenous TGFB1 treatment. In a reciprocal experiment, immunoprecipitation of endogenous SMAD3 resulted in co-immunoprecipitation of SET9 (Fig. S1A). Even upon adjustment of cell lysates to 600 mM sodium chloride, we could still co-immunoprecipitate SET9 in a specific manner (Fig. S1A), demonstrating a strong SMAD3–SET9 interaction.

To exclude the possibility that antisera cross-reactivity caused these results, we immunoprecipitated overexpressed FLAG-tagged SET9 via the FLAG epitope. This resulted in co-immunoprecipitation of endogenous SMAD3, again with enrichment upon TGFB1 treatment (Fig. 1B). We next tested the SET9–SMAD3 interaction in vitro, free of other factors such as SMAD7, which might mediate indirect interactions. Immunoprecipitation with SMAD3 antibody led to co-immunoprecipitation of purified recombinant full-length SET9, only in the presence of SMAD3 (Fig. 1C). Conversely, immunoprecipitation of purified full-length SET9 led to co-immunoprecipitation of an N-terminal SMAD3 fragment (residues 1–211), containing the conserved MH1 domain, but not a C-terminal SMAD3 fragment (residues 199–425) (Fig. S1B). Control immunoglobulins did not immunoprecipitate significant quantities of any protein. Finally, we examined whether SET9 could interact with the transcriptionally active form of SMAD3, phosphorylated at C-terminal serine residues 423 and 425. Immunoprecipitation with a different SET9 antibody to that used in Fig. 1A resulted in co-immunoprecipitation of phosphorylated SMAD3 (Fig. S1C). We conclude for the first time that SET9 interacts directly with SMAD3, in a manner that is enhanced by TGFB1.

Given its role in transcriptional regulation, we examined the effects of SET9 upon SMAD3 activity, using reporter gene assays. SET9 overexpression resulted in a significant enhancement of luciferase expression driven by a synthetic promoter containing 12 basal SMAD3 CAGA binding sites (CAGA-luc), in a dose-dependent fashion (Fig. 2A). Conversely, introduction of a SET9 mutant (H297A) lacking methyltransferase activity suppressed reporter activity, in a dominant-negative manner. We also reduced SET9 levels by means of siRNA knockdown. This resulted in a significant reduction in CAGA-luc activity, even upon exogenous TGFB1 treatment (Fig. 2B). Finally, introduction of the characterised SET9 inhibitor R-PFI-2 also resulted in a significant reduction in CAGA-luc activity, suggesting that SET9 methyltransferase activity is required, at least in part, for the effects upon SMAD3 (Fig. 2C).

We also examined the effects of SET9 on a natural PAI1 cis element driving luciferase reporter gene expression, arguably of more physiological relevance and known to be responsive to SMAD3. Overexpression of SET9 again significantly increased reporter gene expression, whereas addition of SET9 siRNA reduced reporter activity (Fig. S2). Overall, we conclude that SET9 is a SMAD3 co-activator.

To further understand how SET9 controls SMAD3 activity, we examined the key regulatory steps in TGFB1 signalling: SMAD3 nuclear translocation and protein turnover. We firstly reduced SET9 levels by using siRNA, then assessed SMAD3 immunofluorescence. Compared with cells transfected with control siRNA, in which we observed principally nuclear staining for SMAD3, we noted a marked redistribution of SMAD3 to the cytoplasm upon knockdown of SET9 (Fig. 3A). Additionally, subcellular fractionation demonstrated that SMAD3 shifted from the cytoplasm in response to TGFB1, whereas SET9 siRNA that achieved a median 75% knockdown produced a persistent cytoplasmic pool of SMAD3 (Fig. 3B); in the absence of exogenous TGFB1, the percentage of total SMAD3 in the nucleus changed from 58% to 49% upon SET9 knockdown but in the presence of exogenous TGFB1, the percentage of total SMAD3 in the nucleus was reduced from 90% to 49% upon knockdown of SET9. We conclude that SET9 regulates TGFB1-driven nuclear import of SMAD3. We noted lower SMAD3 levels upon SET9 knockdown, which we investigated further.

To understand how SET9 alters SMAD3 levels, we firstly measured SMAD3 half-life, using cycloheximide (CHX) to block de novo protein synthesis. We observed a SMAD3 half-life of approximately 118 min in control cells (Fig. 3C). However, upon SET9 knockdown there was a lower steady state level of SMAD3, and increased SMAD3 turnover, with a half-life of approximately 48 min, which is less than half that of the control transfectants (Fig. 3C). No such effect could be seen upon SMAD2 (Fig. 3C). To exclude a confounding effect of transfection, we also examined SMAD3 half-life in Set9^−/− MEFs. Compared with wild-type MEFs, SMAD3 was present at a similar steady state level (Fig. S3A), but exhibited a shorter half-life in Set9^−/− cells upon CHX treatment (Fig. S3B). Finally, in order to understand whether or not SET9 methyltransferase activity was required to modulate SMAD3 turnover, we examined SMAD3 half-life in stably transfected cells expressing either wild-type or the H297A methyltransferase-deficient SET9. Cells expressing wild-type SET9 demonstrated entirely stable SMAD3 levels, whereas those expressing H297A mutant SET9 exhibited loss of SMAD3 upon CHX treatment (Fig. 3D). We could reverse the stabilising effect of wild-type SET9 upon SMAD3 by addition of the SET9 inhibitor R-PFI-2 (Fig. 3E). We conclude that SET9 stabilises SMAD3 in a methyltransferase-dependent manner.

We have observed a direct interaction between SET9 and SMAD3, and effects of SET9 upon SMAD3, whereas others have shown that SET9 regulates SMAD7. We therefore examined the effects of loss of SET9 upon SMAD7, in the Set9^−/− MEF cells exposed to TGFB1. Compared with Set9^+/+ cells, the Set9^−/− cells exhibited markedly increased levels of SMAD7, a failure to induce SMAD3 phosphorylation in response to exogenous TGFB1, and lower levels of total SMAD3 protein (Fig. 3F). These data confirm the important regulatory effect of SET9 upon SMAD7 levels described by Elkouris et al. (2016), but also suggest that SET9 directly influences SMAD3 levels.

To explore SMAD3 turnover in more detail we examined ubiquitylation, a process known to target SMAD3 for proteasomal destruction. We established a nickel agarose (Ni-NTA) pull-down assay to recover ubiquitylated SMAD3 by means of transfection of HIS-tagged ubiquitin. Addition of the proteasomal inhibitor MG-132 resulted in enrichment of high molecular weight SMAD3 species, characteristic of ubiquitylated SMAD3 (data not shown). Given the effects of SET9 methyltransferase activity on SMAD3 turnover, we incorporated the SET9 inhibitor R-PFI-2 (Barsyte-Lovejoy et al., 2014). Treatment of cells with low doses of MG-132 and R-PFI-2 resulted in an enrichment of ubiquitylated SMAD3 compared with that of cells treated with a low dose of MG-132 alone (Fig. 4A). We also examined ubiquitylation in cells stably expressing wild-type SET9 or methyltransferase-deficient H297A SET9. We observed a relative enrichment in ubiquitylated SMAD3 from cells expressing the H297A mutant in comparison to cells expressing wild-type SET9 (Fig. 4B). We conclude that SET9 methyltransferase activity is involved in regulating SMAD3 stability, in a manner that influences SMAD3 ubiquitylation and proteasomal destruction.

To test whether or not SET9 can directly methylate SMAD3, we used both recombinant full-length SMAD3 and a series of SMAD3 synthetic peptides as substrates for in vitro radioisotope-based methylation assays. Whilst we could readily detect methylation of histones by SET9, we failed to demonstrate methylation of either full-length SMAD3 or SMAD3 peptides. Additionally, we failed to detect methylation of the synthetic SMAD3 peptides by mass spectrometry (data not shown).

We evaluated the biological significance of SET9 in fibroblasts, which are known to be dependent upon TGFB1 and SMAD3 for differentiation to myofibroblasts (Evans et al., 2003). We again utilised MEFs, but also diseased human primary renal (myo)fibroblasts (PRFs) from fibrotic kidneys (three patients) and human mesangial cells. Whereas PRFs are responsible for tubulointerstitial fibrosis, mesangial cells cause glomerulosclerosis, the two entities constituting renal fibrosis in CKD.

Firstly, we examined fibroblast morphology. Whilst normal MEFs adopt a spindle-shaped appearance in response to TGFB1, in keeping with myofibroblast differentiation, Set9^−/− MEFs failed to undergo this characteristic change (Fig. 5A), a phenotype also shared with Smad3^−/− MEFs. Next, we examined fibroblast migration in wound healing assays in response to TGFB1, as a model of fibroblast recruitment in tissue injury. Whilst normal MEFs were capable of migrating to allow wound closure, Set9^−/− MEFs were unable to do so (Fig. 5B), which is another property shared by Smad3^−/− MEFs. Addition of the SET9 inhibitor R-PFI-2 to PRFs also significantly retarded the wound healing assay (Fig. 5C and Fig. S4A). This effect was reproducible in PRFs from all three patients (Fig. S4A). Normal MEFs and mesangial cells also exhibited equally poor wound closure upon R-PFI-2 treatment (data not shown). In the absence of TGFB1, R-PFI-2 had minimal effect on wound closure (Fig. S4B). We conclude for the first time that SET9 is important in fibroblast migration into the wound in response to TGFB1 – a key process in fibrosis.

We next assessed the role of SET9 in collagen contractility, an established methodology for studying myofibroblast–ECM interactions, and a model of tissue retraction in fibrosis (Grinnell, 2000). Compared with wild-type MEFs, Set9^−/− MEFs exhibited significantly reduced collagen contractility (Fig. 6A), which was phenotypically similar to Smad3^−/− cells. This could be observed by measuring either collagen disc diameter or weight (Fig. 6A and data not shown). Addition of R-PFI-2 to PRFs from the three patients also showed a marked and reproducible reduction in collagen contractility (Fig. 6B). Application of R-PFI-2 to human mesangial cells or normal MEFs also resulted in a significant reduction in contractility (data not shown). In the absence of TGFB1, R-PFI-2 had minimal effect on collagen contractility (Fig. S4C). We infer that SET9 is required for fibroblast-mediated collagen contractility in response to TGFB1.

Next, we investigated the expression of ECM proteins. Previous studies have suggested that SET9 contributes to ECM production in lung fibrosis and mesangial cells (see Introduction) and collagens I and III are known to be key in fibrotic replacement of normal parenchyma. We observed significantly lower expression of collagens I and III and fibronectin-1 in Set9^−/− MEFs versus wild-type MEFs in the presence of TGFB1 (Fig. 6C and Fig. S5A for example images). Importantly, treatment with R-PFI-2 significantly decreased the expression of ECM proteins in the diseased PRFs (Fig. 6C and Fig. S5B for example images). We deduce that SET9 is required for PRF expression of ECM in response to TGFB1, in keeping with previous reports (Elkouris et al., 2016; Sun et al., 2010).

Having shown altered myofibroblast contractility upon SET9 inhibition, we finally examined α-SMA, the most widely accepted marker of TGFB1-induced myofibroblast differentiation and contractility, but not previously known to be regulated by SET9. We established a chromatin immunoprecipitation assay to examine recruitment to the SMAD3 binding site within the α-SMA gene. Firstly, we observed a 2.6-fold enrichment of SMAD3, and a 3.8-fold enrichment of SET9, to the α-SMA gene but not the GAPDH gene, in response to TGFB1 in PRFs (Fig. 7A and Fig. S6A). These are the first data describing regulation of α-SMA by SET9. We then examined SET9 knockdown; we observed an ∼80% reduction in SMAD3 recruitment upon transfection of SET9 siRNA or SMAD3 siRNA (Fig. 7B and Fig. S6B). In keeping with a role for SET9 in SMAD3 protein turnover, the addition of MG-132 partly reversed the effect of SET9 knockdown on SMAD3 recruitment to the α-SMA gene. We went on to observe a ∼60% reduction in α-SMA expression at the protein level in PRFs from all three patients in the presence of R-PFI-2, suggesting that SET9 inhibition prevents these cells from undergoing myofibroblast differentiation in response to TGFB1 (Fig. 7C). We also noted that addition of R-PFI-2 also reduced basal levels of α-SMA expression, which are likely to be stimulated by endogenous TGFB1 production. We conclude that SET9 is key to PRF differentiation and contractility in response to TGFB1.

Here, we show that SET9 interacts with SMAD3 and increases SMAD3 transcriptional activity, protein turnover and nuclear import. These are the first data showing this direct interaction and its effects on TGFB1 signalling.

Prior work has shown that the upstream regulatory protein SMAD7 is a target for SET9 methylation, resulting in its destruction through Arkadia-mediated ubiquitylation (Elkouris et al., 2016). SMAD3 protein levels, localization and transcriptional activity were not examined in that study. We confirm here that loss of SET9 stabilises SMAD7 in MEF cells. Interestingly, SMAD7 has recently been shown to interact with the conserved MH2 domain of SMAD3, resulting in SMAD3 destruction (Yan et al., 2016). Given that SET9 methylates SMAD7, it is possible that some of the effects of SET9 that we describe here are mediated, in part, through this indirect mechanism. For example, the stabilization of SMAD3 by SET9 we observed was dependent upon SET9 methyltransferase activity, but we could not demonstrate methylation of SMAD3 by SET9. This effect might therefore be indirectly mediated by SMAD7 methylation.

Nevertheless, we show that SET9 can interact directly with SMAD3 in the absence of SMAD7, and that SET9 interacts with the SMAD3 N-terminal MH1 domain, whereas SMAD7 was shown to interact with the C-terminal MH2 domain of SMAD3 (Yan et al., 2016). In keeping with SET9 having direct regulatory effects upon SMAD3, treatment of cells with the SET9 inhibitor R-PFI-2 or transfection of methyltransferase-deficient SET9 did not inhibit SMAD3 nuclear import (data not shown). If this effect was SMAD7 dependent, we predict that SET9 methyltransferase activity would be required. Additionally, the very low levels of total SMAD3 protein observed in Set9^−/− cells suggest that SET9 has a direct influence on SMAD3 stability. We suggest that the different regulatory mechanisms upon SMADs – methylation and destruction in the case of SMAD7 but stabilization, nuclear import and transcriptional activation in the case of SMAD3 – makes SET9 a unique regulatory component of TGFB1 signalling. It is currently unclear whether SET9 interacts with other SMAD proteins aside from SMAD3 and SMAD7; however, we failed to identify an interaction between SMAD2 and SET9 by co-immunoprecipitation using antibodies against SMAD2 (data not shown).

Fibrosis is driven by TGFB1, which stimulates myofibroblast activities (Duffield, 2014). Whilst the origin of fibroblasts is debated, it is accepted that histologically confirmed fibrosis predicts poorer clinical outcomes in CKD (Menn-Josephy et al., 2016). There are no specific treatments to retard fibrosis (Tampe and Zeisberg, 2014). Previous work has shown that SET9 is important in ECM production in fibrosis (Elkouris et al., 2016; Sasaki et al., 2016) but the role of SET9 in other fibrotic processes is unclear. We demonstrate for the first time that SET9 is recruited to the α-SMA gene and that SET9 is required for SMAD3 occupancy at the α-SMA gene in PRFs, in response to TGFB1. These events coincide with TGFB1-mediated differentiation through expression of α-SMA, and increased collagen contractility. Taken together, our data are consistent with a model in which SET9 mediates SMAD3 stabilization, nuclear import and subsequent recruitment to upregulate the α-SMA gene. Myofibroblast differentiation, α-SMA expression and collagen contractility are key features of fibrosis. Our results therefore suggest that SET9 inhibition might retard these myofibroblast activities in order to treat CKD (Duscher et al., 2014; Liu et al., 2003).

We additionally show for the first time that migration, which accounts for the presence of myofibroblast accumulation at the site of injury (Bochaton-Piallat et al., 2016), and perpetuation of fibrosis elsewhere within the organ, is dependent upon SET9. Previous work has shown that SET9 promotes ECM deposition in fibrosis; we also show that SET9 is required for ECM protein production, specifically in PRFs. Taken together, we propose that inhibition of myofibroblast accumulation at the site of injury, through targeting SET9, should reduce subsequent ECM deposition, as an additional means to treat CKD.

One limitation of our study is that it is currently unclear whether changes in gene expression, resulting from SET9 acting as a SMAD3 co-activator, account for all the phenotypic effects we observe in PRF cells. It is possible that some of the non-genomic effects of SET9 upon TGFB1–SMAD3 signalling we have observed might be responsible for the phenotypic effects of SET9 observed in myofibroblasts. We have not examined what proportion of genes altered in Set9^−/− MEF cells in response to TGFB1 are co-regulated by SMAD3. It is possible that some of the gene expression changes we observe at the 48 h TGFB1 time point could result from activation of factors additional to SMAD3. Nevertheless, the stark differences in gene expression observed between the wild-type versus Set9^−/− MEF cells, supports our original conclusion that SET9 is required for the global TGFB1 response, whether that is mediated solely through SMAD3 or via SMAD3 plus additional factors.

Recent data has suggested that loss of TGFB2 receptors does not prevent renal fibrosis in mouse models of CKD and that targeting upstream TGFB1 signalling in kidney disease may not be effective (Voelker et al., 2016). Nevertheless, SMAD3-deficient mice are resistant to renal fibrosis. Additionally, SMAD3 can be activated in a manner that is independent of TGFB1 to cause fibrosis (Chung et al., 2010; Li et al., 2004). It therefore remains likely that downstream signalling molecules such as SMAD3 and SET9 remain valid targets in CKD and other fibrotic diseases.

Set9^+/+ and Set9^−/− MEFs were cultured in the presence or absence of TGFB1 before two-step total RNA extraction using Trizol and PureLink Micro scale columns (Life Technologies). RNA quality and cRNA synthesis was assessed and performed at The Wellcome Trust Centre for Human Genetics, as was cRNA hybridisation to Illumina Mouse WG-6 arrays. Quality control checks and normalisation was performed as described (Jones et al., 2015), again at that facility. Gene sets showing 1.5-fold or more upregulation between wild-type and Set9^−/− MEFs were used for GSEA with the functional annotation tool of the DAVID Bioinformatics Resource v.6.8, default settings (Huang da et al., 2009a; Huang da et al., 2009b). The top six enriched annotations, ranked by fold enrichment over genes returned by chance from a random gene set of identical size are listed. Microarray data, including the full data gene sets can be found in Tables S2–S4.

For IP, cells were lysed in 50 mM Tris-HCl, pH 7.4, 150 mM NaCl, 1 mM EDTA, 1% Triton X-100 with cOmplete protease inhibitors (Roche). Antibodies against SMAD3 (Abcam, Ab28379), SET9 (Millipore, EPR5574 or Cell Signaling Technology C24B1), FLAG (Sigma, F3165) or isotype control were crosslinked to Protein A Dynabeads using BS3 (Fisher). Where stated, lysates were adjusted to the required concentration of NaCl prior to immunoprecipitation with crosslinked antibodies overnight. For immunoblotting, SET9 rabbit monoclonal was used throughout (Cell Signaling Technologies, C24B1, 1:1000 dilution) and Veriblot secondary antibodies (Abcam, Ab131366, 1:1000 dilution) were used to avoid obscuring immunoglobulin chains. The two SMAD3 antibodies used, used at 1:1000 dilution, Ab28379 and phospho-SMAD3 423/425 (Cell Signaling Technology, C25A9) do not cross-react with other SMAD family members. Full-length recombinant SMAD3 (50991-M20B) and SET9 (14-469) were purchased from Sino Biologicals and Millipore, respectively. GST-tagged SMAD3 fragments were expressed from pGEX plasmids (Addgene) transformed into BL21 E. coli (Zhang et al., 1998) and purified over GST SpinTrap columns (GE Healthcare). Equimolar concentrations of recombinant proteins were mixed in PBS for 2 h prior to immunoprecipitation with the indicated antibodies, followed by extensive washing in PBS containing 0.05% Tween-20 and were then used for immunoblotting.

Diseased renal fibroblasts from three patients were used in each assay described (DV Biologics). They were characterised by positive expression of vimentin, α-SMA, collagen I and III, electron microscopy and lack of tubular cell markers (not shown). For immunofluorescence, cells were seeded into glass chamber slides, treated as indicated, fixed with methanol and stained using antibodies listed (1:100 dilution). Fluorescence from Alexa Fluor-conjugated secondary antibodies (1:1000 dilution) was captured using a Zeiss Axioimager fluorescence microscope across four separate fields. Fluorescence in each field was then divided by the number of cells present (indicated by DAPI staining) to determine fluorescence per cell. Mean fluorescence across the four fields was then calculated and expressed as percentage of the highest values and subject to ANOVA statistical analysis. For western blot densitometry, ImageJ was used to subtract background signal from the band of interest prior to correction for signal arising from the relevant loading control.

Cells were transfected using Lipofectamine or RNAiMAX (Life Technologies) for plasmids or siRNA (30 nM), respectively. siRNA sequences have been described previously (Chuikov et al., 2004; Gaughan et al., 2011). Reporter assays were performed as described (Logan et al., 2009). Nuclear and cytoplasmic fractions were prepared using IGEPAL CA-630 (Sigma). For protein half-life assessment, cells transfected with siRNA or the indicated plasmids were treated with 50 µg/ml cycloheximide (Sigma) or DMSO vehicle for the indicated times prior to direct lysis into Laemmli buffer then immunoblotting. R-PFI-2 (Sigma) was used at the concentrations indicated.

To examine ubiquitylation, cells were transfected overnight with the indicated plasmids, treated with 10 ng/ml TGFB1 and low-dose 1 µM MG-132 then used for Ni-NTA chromatography as described (Logan et al., 2016). Wound healing assays were performed using two-chamber cell culture inserts (Ibidi). After allowing cells to adhere, they were treated as indicated then chambers removed and serial imaging performed to monitor wound closure, as stipulated by the manufacturer. For collagen contractility assays, 50,000 cells per sample were suspended into 0.5 ml of 0.65 mg/ml collagen I (BD Biosciences) in DMEM with 3.7 mM NaOH, per well of a 24-well plate then allowed to set for 30 min; 0.5 ml DMEM was then overlaid and matrices allowed to stress for 48 h. Matrices were then released from the well with a needle and serial diameter measurements taken. Sample weight was measured at the assay end point.

Chromatin from PRFs was prepared, after cross-linking cells in 1% formaldehyde for 10 min at room temperature, using ChromaFlash High-Sensitivity ChIP Kit (Epigentek); 20 µg chromatin was used, with 1 µg of the respective antibody. Oligonucleotide sequences 5′-TTTCAGCTTCCCTGAACACC-3′ and 5′-CGGGTAATTAAAAGAGCCACTG-3′ were used to amplify the human sequence corresponding to the Rat SBE2 SMAD binding element, identified by sequence alignment (Hu et al., 2003), with 19 cycles of conventional PCR or SYBR green-based quantitative real-time PCR with ROX reference dye, incorporating melt curve analysis (Sigma S4438). GAPDH oligonucleotides were used as recommended in the ChIP kit.

We thank Guiseppe Testa for SET9 MEFs, Jill Norman for human mesangial cells, Peter ten Dijk for FLAG-SMAD3 and PAI1-luc, Caroline Hill for CAGA-luc, Erwin Bottinger for Smad3^−/− cells and Adam Pickard for collagen contractility advice.