Mutual dependence of the MRTF–SRF and YAP–TEAD pathways in cancer-associated fibroblasts is indirect and mediated by cytoskeletal dynamics

1The interface between mechanical force and gene expression is central to our understanding of normal and transformed cell behavior. Variation in extracellular matrix (ECM) stiffness and cell adhesion can affect gene expression programs to promote alternative cell fate choices, while, in cancer, malignant progression of solid tumors is associated with increased stiffening of the tumor microenvironment. Cancer-associated fibroblasts (CAFs) play an important role in cancer progression. These myofibroblast-like cells, which express high levels of αSMA, promote cancer cell growth, invasion, and metastasis by paracrine signaling and ECM remodeling (for review, see ).

2The MRTF and YAP transcriptional pathways contribute to the response to mechanical stress. The MRTFs—which are recruited to DNA by their partner, SRF—control expression of dozens of cytoskeletal genes, including αSMA (for review, see ). They respond to Rho-GTPase signals, directly sensing changes in G-actin concentration via their regulatory RPEL domain, and accumulate in the nucleus when G-actin levels are low. The MRTFs can also be activated by substrate stiffness or direct integrin engagement. YAP and TAZ, which bind DNA in association with members of the TEAD family of DNA-binding cofactors, were first characterized as effectors of the Hippo growth control pathway (for review, see ). In addition, they also respond to Rho-GTPase signaling, accumulating in the nucleus in response to high cytoskeletal tension induced by mechanical cues. Previous studies have shown that YAP is activated in CAFs and that YAP–TEAD signaling maintains their contractile and proinvasive properties.

3Several observations suggest that the MRTFs and YAP/TAZ may functionally interact even though they do not share a common DNA targeting factor. The multiplicity of cytoskeletal genes and components of the YAP–TAZ interactome among direct MRTF–SRF genomic targets suggests the possibility of indirect pathway cross-talk. Some genes, such as Ctgf, Cyr61, and Ankrd1, contain binding sites for both SRF and TEAD and thus represent shared targets (for example, see; ). Moreover, YAP/TAZ and the MRTFs can physically interact, although it remains unclear whether this allows their recruitment to DNA independently of their own DNA-binding partners.

4Here we investigate the relationship between MRTF–SRF and YAP–TEAD signaling in CAFs and normal fibroblasts. We show that CAFs exhibit elevated MRTF activity, which is required for their contractile and proinvasive properties. Expression of YAP–TEAD genomic targets and expression of MRTF–SRF genomic targets are mutually dependent even when they are directly targeted by only one of the pathways. Finally, we show that activation of either pathway potentiates the activity of the other indirectly and that this depends on cytoskeletal dynamics.

5A previous analysis of stromal fibroblasts associated with tumors in the MMTV-PyMT mouse mammary carcinoma model suggested that expression of MRTF–SRF target genes may be increased in CAFs. We compared that data set with our previous ChIP-seq (chromatin immunoprecipitation [ChIP] combined with high-throughput sequencing) analysis of MRTF–SRF target genes in serum-stimulated NIH3T3 fibroblasts. A substantial number of MRTF–SRF target genes, predominantly associated with cytoskeletal regulation, including the myofibroblast activation marker αSMA (Acta2), increased with severity of disease stage (Fig. 1A; Table 1A; Supplemental Table S1). The MRTF–SRF target gene data set also exhibited significant overlap with genes overexpressed in myofibroblastic CAFs associated with human pancreatic ductal adenocarcinoma (PDAC) (Table 1A; see the Discussion; ).

6In fibroblasts, extracellular signaling through the Rho-actin pathway promotes MRTF nuclear accumulation. Although normal mammary fibroblasts (NFs) exhibited predominantly cytoplasmic and pancellular localization of MRTF-A in serum-starved conditions, the majority of CAF1 and CAF2 cells exhibited nuclear MRTF-A (Fig. 1B). Consistent with this, an MRTF–SRF reporter gene exhibited increased activity in CAFs compared with NFs (Fig. 1C) even though MRTF-A and MRTF-B expression was comparable in the two cell types (Supplemental Fig. S1A). CAFs exhibited elevated expression of the major myofibroblast and contractility markers αSMA/Acta2 and MLC2/Myl9 (Fig. 1C; Supplemental Fig. S1B). Both of these genes are MRTF targets, and their expression in CAFs was MRTF-dependent (Supplemental Fig. S1B). In contrast, the TCF (ternary complex factor)–SRF target gene Egr1 was expressed at comparable levels in NFs and CAFs (Supplemental Fig. S1C). Consistent with these observations, ChIP analysis demonstrated increased MRTF-A, SRF, and RNA polymerase II (Pol II) recruitment at Acta2 but not Egr1 (Supplemental Fig. S1E).

7Previous studies have shown that MRTF activation in serum-stimulated fibroblasts is RhoA-dependent and reflects depletion of the G-actin pool as a result of formin or LIM kinase activation. Consistent with this, in CAFs, the elevated activity of Acta2 and the transfected MRTF–SRF reporter was blocked when Rho was inactivated by C3 transferase expression; upon depolymerization of F-actin by latrunculin B (LatB); by the LIMK inhibitor LIMKi3; and by the formin inhibitor SMIFH2 (Fig. 1D). Acta2 and reporter expression was also inhibited by treatment with the Rho-actin inhibitor CCG-203971 (; Supplemental Fig. S1D). These inhibitor treatments also reduced MRTF and SRF recruitment to Acta2 as assessed by ChIP (Fig. 1E; Supplemental Fig. S1E).

8The establishment and maintenance of the activated state of CAFs and myofibroblasts are promoted by substrate stiffness (for reviews, see; ) and TGFβ signaling, so we examined the role played by these stimuli in maintaining MRTF activation in CAFs. While MRTF-A remained nuclear when CAFs were plated on stiff (50-kPa) hydrogel substrates, it relocalized to the cytoplasm upon plating on soft (0.5-kPa) hydrogel (Fig. 1B; Supplemental Fig. S1F). Consistent with this, treatment of CAFs with Y27632 or blebbistatin, which impair actomyosin contractility, inhibited both Acta2 expression and MRTF ChIP at Acta2 (Fig. 1D,E). TGFβ signaling was elevated in CAFs, as assessed by levels of phosphorylated Smad2 (Supplemental Fig. S2A). Treatment of CAFs with the TGFβ family receptor kinase inhibitor SB-431542 abolished Smad2 phosphorylation and induced relocalization of MRTF-A to the cytoplasm (Supplemental Fig. S2B,C). SB-431542 treatment also reduced expression of the MRTF–SRF reporter gene and the MRTF–SRF target genes Acta2 and Myl9 (Fig. 1D; Supplemental Fig. S2C,D). Conversely, in NFs, TGFβ treatment induced MRTF-A nuclear accumulation (Supplemental Fig. S2E) and activated Acta2 and Myl9 in an MRTF-dependent fashion (Supplemental Fig. S2F). Taken together, these results establish that MRTF activation is a marker of the CAF activated state and show that TGFβ signaling and contractility are required to maintain it.

9We next tested the role of MRTF–SRF activity in matrix remodeling behavior and contractility, hallmarks of CAF function. As a result of their increased ECM remodeling activity, when embedded in collagen gels, CAFs greatly facilitate matrix invasion by 4T1 carcinoma cells. Strikingly, the proinvasive activity of CAFs was substantially impaired upon siRNA-mediated depletion of the MRTFs (Fig. 2A), as was their ability to contract collagen matrices (Fig. 2B). CAF contractility was also inhibited by SMIFH2 and LIMKi3, which inhibit F-actin assembly and MRTF activation; the MRTF inhibitor CCG-203971; and SB-431542 (Fig. 2C). Thus, MRTF activation is required for the proinvasive and contractile properties of CAFs.

10NFs were not contractile, but their contractility was activated by serum stimulation, and this was MRTF-dependent (Fig. 2D). Expression of a constitutively active derivative of MRTF-A, MRTF123-1A, was also sufficient to induce contractility (Fig. 2E). Previous studies have shown that the enhanced contractility and matrix remodeling activity of CAFs are dependent on the transcriptional coactivator YAP, depletion of its relative TAZ having no effect in this context. Consistent with this, we found that expression of the constitutively active YAP derivative 5SA-YAP was also sufficient to induce contractility in NFs (Fig. 2E). Strikingly, however, siRNA depletion experiments showed that MRTF123-1A-induced contractility was dependent on YAP and vice versa (Fig. 2E). Taken together, these data suggest that the enhanced contractility and proinvasive character of CAFs requires both MRTFs and YAP and that their activities are mutually dependent.

11Analysis of the relationship between MRTF- and YAP-dependent gene expression requires the definition of direct genomic targets for each pathway. Ideally, such definition requires identification of genomic binding sites for the factor concerned and a demonstration that transcription of genes that are in close proximity to these sites is dependent on the factor itself or the signal pathway that regulates it. Previously, we used this approach to define MRTF–SRF target genes in mouse NIH3T3 fibroblasts based on transcriptional response to MRTF-linked signals and proximity to MRTF–SRF genomic binding sites. The MRTF–SRF gene signature overlaps significantly with genes overexpressed in CAFs (Table 1A) as well as genes induced by YAP overexpression in different cell types (Table 1B,C;;;;;; ).

12Recent ChIP-seq studies have identified genomic binding sites for YAP–TEAD in normal and transformed human cells, many of which are associated with genes that are MRTF–SRF targets in fibroblasts (Table 1D;; ). We further refined this analysis by integrating the IMR90 ChIP-seq data with an IMR90 Hi-C (chromosome capture followed by high-throughput sequencing) data set to identify genes whose transcription start sites (TSSs) are in physical contact with YAP–TEAD-binding sites. This analysis defined 776 potential YAP–TEAD direct targets in IMR90 fibroblasts (Supplemental Table S2B). Both the MRTF–SRF and the YAP–TEAD target gene sets are enriched for cytoskeletal and ECM remodeling genes and many genes involved in proliferation, signaling, and transcription (Supplemental Table S3). Many were also overexpressed in CAFs from the mouse PyMT mammary and KRas/p53 pancreatic cancer models and from human pancreatic, oral, and breast carcinoma (Table 1F,G; Supplemental Tables S4–S7;;;; ).

13Many of the YAP–TEAD target genes in human IMR90 cells were identified previously as direct MRTF–SRF target genes in mouse NIH3T3 cells (Table 1E; Supplemental Table S2E). Comparison of the two data sets thus allowed us to identify potential candidate direct target genes specific for MRTF–SRF alone, specific for YAP–TEAD alone, or shared by both regulators (referred to here as “MRTF-only,” “YAP-only,” and “shared” targets) (Supplemental Table S2C–E). All MRTF-only direct genomic targets analyzed were expressed at high levels in CAFs compared with NFs (Fig. 3A); we note that many of these were not detected as elevated in the previous analysis, presumably because of the relatively insensitive Illumina array technique used. Similarly, expression of shared MRTF–SRF/YAP–TEAD genomic targets was also elevated in CAFs compared with NFs (Fig. 3B), as was expression of candidate YAP-only genomic targets (Fig. 3C). Strikingly, however, MRTF depletion impaired not only the expression of the MRTF-only genomic targets but also the YAP-only targets (Fig. 3D), and, conversely, YAP depletion impaired expression of MRTF–SRF genomic targets regardless of whether they contain YAP–TEAD-binding sites (Fig. 3E). Thus, MRTF–SRF signaling and YAP–TEAD signaling are mutually dependent.

14The mutual dependence of MRTF–SRF and YAP–TEAD target gene expression even at those targets that are directly bound by only one of the transcription factor pairs strongly suggests that their mutual dependence reflects the influence of each on the other's cognate signaling pathway. To examine their mutual dependence more directly, we studied the regulation of MRTF–SRF and YAP–TEAD reporter genes. The elevated activity of an MRTF–SRF reporter in CAFs was sensitive to depletion of not only the MRTFs or SRFs and the Rho-actin inhibitor CCG-203971 (Fig. 4A) but also both YAP and its DNA-binding cofactor, TEAD1 (Fig. 4B). Similarly, a YAP–TEAD reporter also exhibited elevated activity in CAFs, which was dependent both on YAP and TEAD and on MRTF and SRF and was inhibited by CCG-203971 (Fig. 4C,D). Furthermore, MRTF and YAP were also mutually dependent in NFs. Serum-stimulated activity of the MRTF–SRF reporter was partially inhibited by depletion of YAP (Supplemental Fig. S3A), as was induction of the MRTF–SRF direct target genes Acta2 and Tagln (Supplemental Fig. S3B,C). Conversely, the modest response to serum stimulation of the YAP–TEAD reporter and the YAP–TEAD-specific direct target genes Amotl2 and Ddah1 was inhibited by depletion of both MRTFs and YAP (Supplemental Fig. S3D–F).

15We used ChIP to investigate whether the reporter results reflected changes in recruitment of MRTF and YAP to target genes. The enhanced recruitment of MRTF to MRTF–SRF target genes was enhanced in CAFs compared with NFs (Fig. 4E). It was significantly reduced upon depletion of either YAP or its DNA-binding partner protein, TEAD1, suggesting that MRTF activity is dependent on YAP–TEAD target gene expression (Fig. 4E). Inadequate antibodies precluded analysis of recruitment of YAP to YAP–TEAD targets in CAFs, but binding of its DNA-binding partner, TEAD1, was increased in CAFs compared with NFs and sensitive to depletion of MRTF (Fig. 4F). These data suggest that YAP–TEAD binding is MRTF-dependent. We confirmed this using human MDA-MB231 breast carcinoma cells, which are also migratory and invasive and display nuclear MRTF. Here, ChIP analysis revealed specific recruitment of MRTF-A to SRF sites and of YAP to TEAD sites; again, binding of MRTF was YAP-dependent and vice versa (Supplemental Fig. S4). Taken together, these results suggest that the mutual dependence of MRTF and YAP activity arises indirectly and is likely to reflect their influence on each other's signal pathway.

16Recent studies have demonstrated physical interaction between MRTF and YAP independent of their DNA-binding cofactor proteins, SRF and TEAD. The ChIP data presented above and the fact that YAP–TEAD and MRTF–SRF ChIP-seq data exhibit only partial overlap suggest that YAP–MRTF interaction is not sufficient to explain the mutual dependence of the two systems. Nevertheless, we sought to demonstrate directly that the two pathways are independently regulated, focusing on their response to the actin-binding drugs cytochalasin D (CD) and LatB. CD directly activates the MRTFs by blocking their inhibitory interaction with G-actin, while LatB inhibits MRTF activity by increasing cellular G-actin level, but both agents inhibit YAP by inhibiting F-actin assembly.

17In CAFs, LatB treatment resulted in the redistribution of both factors to the cytoplasm; in contrast, while CD treatment induced cytoplasmic localization of YAP, it potentiated nuclear accumulation of MRTFs (Fig. 5A,B). Consistent with this, the different classes of MRTF and YAP target genes responded differentially to CD treatment. In normal fibroblasts, MRTF-only and shared targets were CD-inducible, while YAP-only targets were not (Fig. 5C; Supplemental Fig. S5). In CAF1 cells, activity of MRTF-only targets was similar to that seen in CD-treated NFs and was further potentiated by CD (Fig. 5C; Supplemental Fig. S5A). The basal activity of shared MRTF–SRF/YAP–TEAD target genes in CAFs was higher than in CD-stimulated NFs, and CD stimulation had a modest or no further effect (Fig. 5C, middle; Supplemental Fig. S5B). In contrast, although YAP-only target genes were expressed at elevated levels in CAFs, their transcription was suppressed by CD treatment (Fig. 5C; Supplemental Fig. S5B).

18Taken together, these results suggest that MRTF recruitment and YAP recruitment to DNA are independently regulated. Although our inability to detect mouse YAP by ChIP precluded a direct test of this model using CAFs, ChIP analysis in the human MDA-MB231 cell system corroborated the transcription data. At all target types, MRTF and YAP ChIP signals were abolished by LatB (Fig. 5D,E). However, CD treatment increased recruitment of MRTF and SRF to target genes (Fig. 5D) but decreased binding of YAP and TEAD (Fig. 5E). Thus, although the two factors can physically interact, their recruitment to DNA is independently regulated, and, at shared targets, expression levels represent the integration of separate signals at discreet MRTF–SRF- and YAP–TEAD-binding sites.

19The data presented so far suggest that the mutual dependence of MRTF and YAP may reflect their ability to influence each other's upstream signal pathways, both of which involve alterations in cytoskeletal dynamics. To test this idea, we asked whether constitutively active derivatives of MRTF and YAP, whose activity is uncoupled from upstream signal pathways, are mutually dependent. MRTF123-1A is nuclear and active because it cannot bind G-actin, while 5SA-YAP lacks the phosphorylation sites required for its cytoplasmic retention by 14-3-3 protein.

20In normal fibroblasts, expression of MRTF123-1A effectively activated an MRTF–SRF reporter gene and MRTF-only target genes independently of YAP, consistent with a model in which transcription activation by MRTF does not require direct recruitment of YAP (Fig. 6A; Supplemental Fig. S6A). Activation of both MRTF–SRF and YAP–TEAD reporters by MRTF123-1A was dependent on SRF (Fig. 6C), and both reporters could also be activated by expression of the constitutively active SRF derivative SRF-VP16 independently of MRTF (Fig. 6D). In contrast, MRTF123-1A also activated the YAP–TEAD reporter gene and endogenous YAP–TEAD targets, but this was YAP-dependent (Fig. 6A,B; Supplemental Fig. S6B). Together, these data strongly suggest that MRTF123-1A activates the YAP–TEAD pathway by inducing MRTF–SRF target gene expression.

21Similarly, although the constitutively active YAP derivative 5SA-YAP activated both the YAP–TEAD reporter and YAP–TEAD target genes independently of MRTF (Fig. 6E,F; Supplemental Fig. S7A), its ability to activate the MRTF–SRF reporter was dependent on TEAD (Fig. 6G; Supplemental Fig. S7B). In contrast, activation of shared target genes by MRTF123-1A and 5SA-YAP was partially dependent on YAP and MRTF, respectively (Fig. 6H; Supplemental Figs. S6C, S7C). These data are consistent with our assignment of the various target genes to MRTF-only, YAP-only, and shared classes. Moreover, they suggest that constitutively active MRTF activates YAP upstream signal pathways as a result of its ability to activate MRTF–SRF target gene expression and vice versa.

22We next investigated the mechanistic basis for the mutual dependence of MRTF–SRF and YAP–TEAD signaling. Both pathways are responsive to the state of actin dynamics, so we first investigated whether MRTF123-1A or 5SA-YAP affects F-actin assembly. Both proteins promoted F-actin assembly in NFs, as judged by F-actin pelleting assays and phalloidin staining (Fig. 7A; Supplemental Fig. S8A). Expression of MRTF123-1A in NFs induced YAP nuclear accumulation, which was inhibited upon depolymerization of F-actin by LatB (Fig. 7B). Consistent with this, in MDA-MB231 cells, MRTF123-1A expression promoted increased YAP ChIP at YAP targets, which was abolished by LatB (Supplemental Fig. S8C).

23Cell adhesion and cell contractility act via Src kinases to activate YAP by antagonizing LATS-mediated phosphorylation of YAP S127, a major target for Hippo signaling (;; for review, see ). Cell adhesion, actomyosin contractility, and Src kinases are also required for maintenance of active YAP in CAFs. Expression of MRTF123-1A in NF1 cells decreased YAP S127 phosphorylation, which was accompanied by increased Src S416/418 and MLC2 phosphorylation (Supplemental Fig. S8D). In NF1 cells, inhibition of myosin (blebbistatin), Src family kinases (dasatinib), or integrinαv (cilengitide) blocked MRTF123-1A-induced YAP S127 dephosphorylation (Fig. 7C) and inhibited activation of the YAP reporter by MRTF123-1A without affecting activation of the MRTF reporter (Fig. 7D). Activation of YAP-only and MRTF-only target genes by MRTF123-1A was affected in a similar manner (Fig. 7E–G).

24These results show that YAP activation by MRTF123-1A involves increased cell contractility, a property that is strongly dependent on MRTF–SRF target gene expression. We showed above that in NFs, MRTF123-1A activates expression of contractile genes—including αSMA/Acta2, MLC2/Myl9, and Myh9 (Fig. 6B; Supplemental Fig. S6)—and contractility itself (Fig. 2E), and, in CAFs, contractile gene expression and contractility are MRTF-dependent (Fig. 2B; Supplemental Fig. S1B). Expression of integrins Itgb3 and Itgav, specific targets for cilengitide action, was elevated in CAFs and could be potentiated in NFs by MRTF123-1A expression (Supplemental Fig. S8D,E).

25Recent studies have shown that the septin regulator Cdc42EP3 is required in CAFs for F-actin assembly, contractility, and YAP activation. In CAFs, expression of Cdc42EP3, which is a direct MRTF–SRF target in fibroblasts, was elevated and MRTF-dependent (Fig. 3D), and, in NFs, it was strongly activated by MRTF123-1A and CD (Fig. 5C; Supplemental Fig. S6C). Consistent with the reported dependence of F-actin assembly on Cdc42EP3 levels, the ability of MRTF123-1A to induce activity of the YAP–TEAD reporter and YAP–TEAD target genes in NFs was sensitive to Cdc42EP3 depletion (Fig. 7H). Taken together with the preceding results, these data show that MRTF–SRF signaling potentiates YAP activity through activation of multiple genes involved in cell contractility.

26Consistent with the ability of constitutively active 5SA-YAP to induce F-actin assembly, it also induced nuclear accumulation of MRTF; this was inhibited by LatB and was not induced by 5SA/S94A-YAP, indicating that it involves F-actin assembly and requires activation of YAP–TEAD target genes (Fig. 8A; Supplemental Fig. S9A). ChIP analysis showed that 5SA-YAP, but not 5SA/S94A-YAP, expression induced LatB-sensitive MRTF-A recruitment to MRTF target genes (Fig. 8B; Supplemental Fig. S9B). 5SA-YAP expression also induced phosphorylation of MRTF-A (Supplemental Fig. S9C), which reflects decreased G-actin binding (Supplemental Fig. S8A; ). Consistent with these results, LatB treatment abolished 5SA-YAP activation of the MRTF reporter and MRTF-only target genes such as Acta2 (Fig. 8C) but did not affect activation of the YAP reporter or YAP-only targets such as Amotl2 (Fig. 8D). LatB treatment of cells expressing 5SA-YAP led to immediate inhibition of transcription at MRTF-only and shared targets but not YAP-only targets (Supplemental Fig. S9D).

27We showed above that the elevated MRTF activity seen in CAFs is dependent on matrix stiffness, contractility, and TGFβ signaling (Fig. 1B–D; Supplemental Fig. S2A,B). Consistent with these data, the ability of 5SA-YAP to potentiate MRTF–SRF reporter activity and Acta2 transcription was sensitive to the LIMK inhibitor LIMKi3, the ROCK inhibitor Y27632, and the TGFβ family receptor kinase inhibitor SB-431542, while its ability to activate the YAP–TEAD target gene Amotl2 was not (Fig. 8E; Supplemental Fig. S9E). Strikingly, elevated TGFβ signaling in CAFs, as assessed by SMAD2 S465/S467 phosphorylation, was strongly down-regulated upon siRNA knockdown of YAP (Fig. 8F). The TGFβ component Inhba is expressed at high levels in CAFs (Supplemental Fig. S9F; ) and mediates their autocrine dependence on TGFβ signaling (C Foster, unpubl.; D Miller and C Hill, pers. comm.). Inhba is a candidate YAP–TEAD direct target gene (Supplemental Table S2), and, in NFs, its expression was potentiated by 5SA-YAP expression (Fig. 8G).

28Taken together with the results in the preceding section, these results support a model in which the potentiation of TGFβ signaling by YAP contributes to elevated MRTF–SRF activity, which in turn influences YAP–TEAD signaling through potentiation of cell contractility (Fig. 8H; see the Discussion).

29The MRTF–SRF signaling pathway allows the coordination of gene transcription and cytoskeletal dynamics. Here we investigated the role of MRTF–SRF signaling in CAFs derived from the mouse PyMT mammary tumor model. In contrast to NFs, the MRTFs are active in CAFs and are dependent on matrix stiffness and autocrine TGFβ signaling. MRTF–SRF target gene expression is elevated and is required for CAF contractility and proinvasive properties. Strikingly, MRTF–SRF target gene expression is also dependent on the transcription coactivator YAP, which is also mechano-responsive and required for CAF proinvasive properties and vice versa even at genes that are direct targets for only one of the two pathways. MRTF and YAP can each indirectly activate the other through their ability to affect actin cytoskeletal dynamics. Our results support a model in which the mutual dependence of MRTF–SRF and YAP–TEAD arises from their ability to regulate cell contractility and TGFβ signaling (Fig. 8H). YAP functions redundantly with its relative, TAZ. Although activity of the latter does not appear limiting in the CAF system used here, our findings suggest that MRTF–SRF activity will also influence TAZ-dependent transcription in appropriate contexts.

30We found that, in contrast to NFs, a substantial proportion of mammary CAFs associated with PyMT-induced carcinoma contains nuclear MRTF and that expression of numerous MRTF–SRF target genes is concomitantly elevated. Similarly, MRTF–SRF target gene expression is elevated in tumor-proximal myCAFs (myofibroblastic CAFs) in human pancreatic ductal carcinoma. Thus, constitutively nuclear localization of MRTF and activation of MRTF–SRF target gene expression can be considered as further CAF activation markers (for discussion, see ). Constitutively nuclear MRTF has been observed previously in a number of other cell types and contexts and is generally taken to be indicative of increased MRTF–SRF signaling. It should be noted, however, that nuclear confinement of the MRTFs is not necessarily sufficient for activation of MRTF–SRF target gene expression unless G-actin levels are low.

31YAP-dependent gene expression is also increased in CAFs and is required for CAF contractility and proinvasive behavior. Recent ChIP-seq analysis has shown that YAP–TEAD target genes encode proteins involved in proliferation, signaling, transcription, and cytoskeletal components and regulators. We found that the direct YAP–TEAD target gene sets contain many genes found in our previously determined MRTF–SRF gene signature. These genes, which contain binding sites for both MRTF–SRF and YAP–TEAD, include several commonly used as model YAP–TEAD targets, including Cyr61, Ctgf, and Ankrd1. In line with our finding that YAP activity and MRTF activity are mutually dependent, we note that both MRTF–SRF and YAP–TEAD signatures defined here are significantly enriched in activated stromal gene expression signatures in both mouse tumor models and human breast, oral, and pancreatic cancers.

32We found that in CAFs, but not in normal fibroblasts, MRTF-A exhibited substrate stiffness-induced nuclear accumulation. Mechanical cues, including substrate stiffness, were shown previously to promote nuclear localization of YAP and TAZ. Application of force to integrins induces MRTF-A nuclear accumulation, and substrate stiffness can activate it in lung fibroblasts and mesenchymal stem cells. It appears that development of the CAF phenotype involves increased sensitivity of YAP and MRTF to mechanical stimulation, but how this occurs remains unclear. In PDAC, contractile “myCAFs” are found close to the tumor, coexisting with a more distant “iCAF” (inflammatory CAF) population, and organoid experiments show that the myCAF phenotype requires physical contact with the tumor. This, together with the elevated expression of MRTF–SRF and YAP–TEAD target genes in myCAFs, suggests that the myCAF phenotype might be mechanically determined.

33The MRTFs compete with the other cofactor family, the TCFs, for binding to SRFs. In mouse embryonic fibroblasts (MEFs), simple deletion of all three TCFs substantially increases MRTF-dependent gene expression, resulting in greatly enhanced contractility and proinvasive behavior. This phenotype reflects increased access of the MRTFs to SRFs rather than TCF-dependent transcriptional activation and supports the notion that MRTF–SRF signaling is an important contributor to the activated CAF phenotype. However, at least at the RNA level, the relative expression of the MRTFs and TCFs did not appear to differ between NFs and CAFs.

34Recent experiments have demonstrated direct physical interactions between the YAP/TAZ WW domain and a conserved PPxY motif at the C terminus of myocardin family proteins, including the MRTFs. Our experiments with constitutively activated MRTF and YAP suggest that these interactions do not directly contribute to the transactivating functions of MRTF. Moreover, they do not appear sufficient to generate a significant ChIP signal, since, even when coincident ChIP was observed, both corresponding DNA-binding partners were also detected, consistent with either the presence of closely spaced SRF and TEAD sites or looping to remote sites. We favor the idea that rather than mediating targeting, direct MRTF–YAP interaction facilitates activation of shared targets, allowing stabilization of higher-order complexes formed between activated MRTF–SRF and YAP–TEAD bound to separate DNA regulatory elements.

35While both MRTF–SRF signaling and YAP–TEAD signaling are responsive to cytoskeletal state and mechanical stress, they are differentially regulated. MRTF–SRF signaling senses G-actin levels, whereas YAP–TEAD signaling is F-actin-dependent and sensitive to cytoskeletal tension. Moreover, even at DNA regions where both MRTF binding and YAP binding are detected, they exhibit opposite responses to the actin-binding drug CD. Nevertheless, our results show that the pathways are mutually dependent, with inhibition (or activation) of one pathway resulting in inhibition (or activation) of the other. Our results show that this interdependence requires SRF and TEAD, the DNA-binding cofactors of MRTF and YAP, and thus must involve activation of their genomic targets. Indeed, constitutively active TEAD derivatives are sufficient to induce activation of MRTF–SRF targets such as Acta2 and Tagln.

36Cross-talk between MRTF–SRF and YAP–TEAD signaling involves multiple genes that influence cytoskeletal dynamics (Fig. 8H). YAP activation in CAFs is dependent on cell adhesion and contracility. In normal fibroblasts, MRTF–SRF signaling potentiates YAP–TEAD signaling in an adhesion- and contractility-dependent manner. Major components of the contractile machinery in CAFs are MRTF–SRF targets, including Acta2, Myl9, and Myh9. MRTF–SRF signaling also potentiates expression of integrins Itgb3 and Itgav, which have been implicated previously in Src-dependent YAP activation. Both contractility and YAP activation in CAFs are also dependent on the septin regulator Cdc42ep3, an MRTF–SRF direct target gene, which was also required for cross-talk between the MRTF–SRF and YAP–TEAD pathways. Conversely, at least one route by which YAP–TEAD signaling potentiates MRTF–SRF activity appears to be through activation of autocrine TGFβ signaling, which is likely to involve elevated expression of the YAP–TEAD target Inhba. Blockade of TGFβ receptor activity restores MRTF-A cytoplasmic localization in CAFs, and TGFβ signaling has been implicated previously in the myofibroblast transition.

37We showed that the contractile and invasiveness-promoting phenotype of CAFs reflects activation of both MRTF–SRF signaling and YAP–TEAD signaling and that these two pathways are mutually dependent. The robustness of transitions between two different functional states is substantially enhanced when they are governed by interlocking positive feedback loops, such as the simultaneous activation of a pathway promoting one state and inhibition of a pathway that inactivates that state (for discussion, see; ). It is tempting to speculate that the mutual dependence of the MRTF–SRF and YAP–TEAD signal pathways contributes to the stability of the CAF activated state.

38The antibodies used were phospho-Smad2 (Cell Signaling Technology, 138D4), Smad2/3 (BD, 610842), SMA (Sigma, A2547), MLC2 (Cell Signaling Technology, 3674), SRF (sc-335), phospho(T18/S19)-MLC2 (Cell Signaling Technology, 3674S), MRTF-A (sc-21558), MRTF-B (sc-47282), pan–ERK (BD, 610124), YAP (for immunofluorescence, clone 63.7, sc-101199; for ChIP, ab52771), phospho-Ser127-YAP (Cell Signaling Technology, 4911S), TEAD1/TEF-1 (BD, 610922), TEAD4/TEF3 (sc-101184), Actin (Cytoskeleton, AAN01), pan 14-3-3 (clone H-8, sc-1657), histone H3 (Abcam, ab1791); RNA Pol II CTD S5P (Covance, H14), Integrin β3 (ab119993), Integrin αV (ab179475), phospho-(Tyr416) Src (Cell Signaling Technology, 2101), Src (2109S), HA high affinity (3F10) (Roche, 11867431001), HA-peroxidase high affinity (3F10) (Roche, 12013819001), Flag-M2-peroxidase (Sigma, A-85292), and Flag (for ChIP, Sigma, F7425; for immunofluorescence, clone M2, F1804). Immunoblotting and immunofluorescence were by standard techniques. F-actin was visualized with Texas Red-X phalloidin (Invitrogen), and nuclei were visualized with DAPI.

39ChIP was as described previously. Three independent chromatin preparations and immunoprecipitations were analyzed in each experiment with duplicate quantitative PCR (qPCR) for each using primers as in Supplemental Table S8A. Total RNA was prepared using GeneElute (Sigma) in column format. Following DNase I treatment (Ambion), cDNA synthesis used the Transcriptor first strand cDNA synthesis kit (Roche). qPCR was performed using the ABI 7900 thermocycler (Applied Biosystems) and QuantStudio QS3/QS5 with detection by SYBR Green incorporation (Invitrogen). Relative abundance of template cDNA from intronic sequences was used as a proxy for relative transcription rates and calculated using the ΔCt method, normalizing to the abundance of GAPDH cDNA. For primers, see Supplemental Table S8B.

40Organotypic invasion assays and gel contraction assays were as described. For inhibitor treatments and serum stimulation, cells were treated as appropriate prior to embedding and maintained in inappropriate medium during the assay. For siRNA depletions, cells were treated for 72 h prior to embedding. Twenty-four hours after plating, gel contraction was calculated as a percentage of well area using ImageJ. For F-actin pelleting, the F/G-actin separation kit (Cytoskeleton) was used according to the manufacturer's instructions.

41Immortalized NFs and CAFs from the PyMT model were maintained in 10% FCS or 0.3% FCS + 1% ITS (Invitrogen) for 20 h followed by stimulation with 15% serum or 2 ng/mL TGFβ. Small-molecule treatments were 10 µM SB-143542, 20 µM CCG-203971, 1 µM LatB, 2 µM CD, 20 µM SMIFH2, 50 µM LIMKi3, 0.5 µM dasatinib, and 10 µM Y27632. Cells were treated for 12 h prior to reporter analysis and for 2 h prior to intronic RNA or ChIP analysis. Cells were transfected using Lipofectamine LTX and Plus reagent (1 µL of LTX/0.5 µL of Plus or 5 µL/2.5 µL of Plus for 24-well and six-well plates, respectively, and 30 µL of LTX/15 µL of Plus or 75 µL of LTX/37.5 µL of Plus for 10-cm and 15-cm plates). Reporter assays in 24-well plates used 20 ng of 3DA.Luc MRTF–SRF or pGL3 4xGTIIC YAP–TEAD reporter, 5 ng of Renilla ptk-RL control, and 100 ng of expression plasmid with analysis by dual-luciferase reporter assay system (Promega). Expression plasmids (15 µg per 15-cm plate) were SRF-VP16, MRTF123-1A, pCMV-HA-5SA-YAP, and pCMV Flag-5SA/S94A-YAP (gift from Dr Ralph Gruber). For siRNA-mediated protein depletion, cells were reverse-transfected using RNAi Max reagent with analysis 72 h later; expression plasmids were transfected 48 h prior to analysis. siRNAs (20 nM) were mouse YAP1 (Dharmacon On-Target Plus, J-046247-10), human YAP1 (Dharmacon, L-012200), mouse TEAD1 (Dharmacon, L-048419-01), mouse Cdc42EP3 (Dharmacon, L-046421-01), mouse SRF (Dharmacon, J-050116-09), human SRF (Dharmacon M-009800), mouse MRTFA/B dual-targeting oligo 5′-UGGAGCUGGUGGAGAAGAA-3′ (25 nM), human MRTF-A (L-015434), human MRTF-B (Dharmacon, M-019279), and human nontargeting pool (Dharmacon, D-001810-10-20).

42Published YAP–TEAD ChIP-seq data from IMR90 lung fibroblasts were integrated with a previous Hi-C analysis of IMR90 cells as described previously. The different genome alignments used in the two studies were converted to the mm9 assembly using the University of California at Santa Cruz LiftOver tool (https://genome.ucsc.edu/cgi-bin/hgLiftOver). The final analysis revealed 725 TSSs that show physical linkage to remote YAP–TEAD sites that, together with 65 TSSs located within 2 kb of a YAP–TEAD site, give a total of 776 potential YAP–TEAD target TSSs.

43Ontology analysis was performed as described previously. Hypergeometric comparisons of target gene sets were performed in R using the phyper function. Numbers of overlapping genes were determined, and statistical significance was estimated with Bonferroni correction for multiple testing. Genes overexpressed in myCAFs were defined as those commonly overexpressed greater than twofold in myCAFs compared with inflammatory CAFs and quiescent fibroblasts.

44Analyses were performed using GraphPad Prism software. Mean values and standard error of the mean were generated from the indicated repeats of biological experiments. P-values were obtained from t-tests with unpaired samples, with a significance threshold of P < 0.05.