Dynamic regulation of myofibroblast phenotype in cellular senescence

Abstract Cellular senescence is an antiproliferative response with a critical role in the control of cellular balance in diverse physiological and pathological settings. Here, we set to study the impact of senescence on the regulation of cell plasticity, focusing on the regulation of the myofibroblastic phenotype in primary fibroblasts. Myofibroblasts are contractile, highly fibrogenic cells with key roles in wound healing and fibrosis. Using cellular models of fibroblast senescence, we find a consistent loss of myofibroblastic markers and functional features upon senescence implementation. This phenotype can be transmitted in a paracrine manner, most likely through soluble secreted factors. A dynamic transcriptomic analysis during paracrine senescence confirmed the non‐cell‐autonomous transmission of this phenotype. Moreover, gene expression data combined with pharmacological and genetic manipulations of the major SASP signaling pathways suggest that the changes in myofibroblast phenotype are mainly mediated by the Notch/TGF‐β axis, involving a dynamic switch in the TGF‐β pathway. Our results reveal a novel link between senescence and myofibroblastic differentiation with potential implications in the physiological and pathological functions of myofibroblasts.


| INTRODUC TI ON
Cell senescence is a stable form of cell cycle arrest linked to diverse pathological and physiological situations (Chan & Narita, 2019;Di Micco et al., 2021). The senescence program typically involves a blockade to proliferation followed by clearance mediated by immune cells, leading to the elimination of cells that undergo an unrepairable level of damage or are no longer required for normal tissue function. Senescence plays an essential role in the control of cell balance in diverse physiological settings such as embryonic development, wound healing, or regeneration. In turn, senescence dysfunction has been associated with a large number of diseases, in particular agerelated diseases, such as cancer, fibrosis, diabetes, and atherosclerosis (Munoz-Espin & Serrano, 2014). Senescence is a complex, finely regulated cellular program that is connected with other essential cell processes. Among these, a link between senescence and cell plasticity has emerged recently, which appears to be markedly contextdependent. For example, senescence can antagonize reprogramming to iPS in a cell-autonomous manner (Banito et al., 2009) and diminish self-renewal in aged adult stem cells (Sousa-Victor et al., 2014).
Paradoxically, senescence can also promote cell plasticity in vivo in a paracrine process mediated by the SASP (Chiche et al., 2017;Demaria et al., 2014;Mosteiro et al., 2016;Ritschka et al., 2017). In addition, the expression of a differentiation-related gene signature is a typical feature in diverse senescence cellular models (Adrados et al., 2016;Storer et al., 2013). Here, we set to explore the link between senescence and cell plasticity, focusing on myofibroblastic differentiation. Myofibroblasts are a specialized cell type with fibroblastic and muscle cell features, which play key roles in several physiological and pathological contexts (Hinz et al., 2012). Myofibroblast function is required for tissue repair after injury, facilitating contraction and replenishing of extracellular matrix. The dysregulation of the physiological repair function of myofibroblasts can cause their aberrant accumulation leading to fibrosis that compromises organ function. While the origin of myofibroblasts in vivo is controversial, differentiation of resident fibroblasts is considered a major source of myofibroblasts (Hinz et al., 2012). Interestingly, previous studies have suggested a link of cell senescence to myofibroblast differentiation and fibrosis, but this relation seems complex and strongly context-dependent (Schafer et al., 2018). For instance, the accumulation of senescent cells promotes lung fibrosis (Schafer et al., 2017), but an antifibrotic role of senescence has been described in liver or heart fibrosis (Krizhanovsky et al., 2008;Meyer et al., 2016) or during wound healing (Jun & Lau, 2010). In order to gain insights into the role of senescence in the differentiation and the physiological and pathological functions of myofibroblasts, here we have studied the regulation of myofibroblast phenotype in cellular models of fibroblast senescence. Our work shows that the myofibroblast phenotype is dynamically regulated during fibroblast senescence in a process that involves the senescence-associated secretory phenotype (SASP), with differential regulation of the TGFβ pathway.

| Downregulation of myofibroblast phenotype during senescence
To investigate the impact of cell senescence on the regulation of the myofibroblastic phenotype, we analyzed the expression of myofibroblast markers in different cellular models of fibroblast senescence. Myofibroblasts are typically characterized by the expression of α-smooth muscle actin (α-SMA, ACTA2), associated with increased contractile force. They also show high expression of collagen genes, indicative of active deposition of extracellular matrix (ECM;Hinz et al., 2012). Notably, control non-senescent fibroblasts showed detectable basal levels of α-SMA, consistent with previous reports of expression of myofibroblast markers in fibroblasts in standard tissue culture conditions (Ehler et al., 1996). We found a consistent reduction in α-SMA and collagen genes by Western blot and qPCR in two strains of primary human fibroblasts undergoing different forms of senescence, such as OIS (oncogene-induced senescence) caused by chronic or inducible RAS expression, or DIS (DNA damage-induced senescence) caused by bleomycin (Figure 1a,b, Figure S1a). Similar results were obtained in mouse embryo fibroblasts (MEFs) made senescent with the CDK inhibitor palbociclib (Figure 1c). Senescence induction was validated by reduced proliferation, increased SA-BetaGal (senescence-associated beta-galactosidase) activity, and induction of the senescence effectors p16INK4A or p21CIP1, as well as IL-8, a member of the inflammatory SASP (Acosta et al., 2008;Coppe et al., 2008) (Figure 1a,c and Figure S1b,c). Immunofluorescence analysis in the inducible ER:RAS model showed that α-SMA was present in characteristic stress fibers and α-SMA positivity showed a significant cell-to-cell inverse correlation with senescence markers such as the senescence-associated heterochromatin foci (SAHF; Narita et al., 2003) and IL-8 (Figure 1d), reinforcing the notion that myofibroblast markers are downregulated during senescence in fibroblasts. Of note, we did not detect a general correlation of α-SMA levels with proliferation status (Figure S1d), suggesting that the observed phenotype was specifically associated with senescence.
Since myofibroblasts are associated with active ECM remodeling, we asked whether the observed changes in myofibroblast phenotype during senescence may have an impact in the ECM. To this end, we studied type I collagen and fibronectin by immunofluorescence in monolayers of growing and senescent ER:RAS fibroblasts, as well as in ECM after cell removal (Figure 1e). These stainings revealed clear changes in matrix patterns. The ECM from non-senescent fibroblasts displayed a regular, compact pattern with parallel fibers, whereas senescent ECM showed a more irregular, lattice-like pattern with large open spaces between fibers. This phenotype could be explained by the combined effect of altered matrix deposition and degradation mediated by matrix metalloproteases (Basisty et al., 2020;Birch & Gil, 2020). Of note, we also observed reduced protein levels of collagen and fibronectin in senescent fibroblasts ( Figure 1e, left panel), consistent with the RNA expression data. The myofibroblastic phenotype is also linked to increased migration and invasion (De Wever et al., 2004). In vitro wound healing and Transwell Matrigel invasion assays revealed significantly reduced migration and invasion in OIS fibroblasts (Figure 1f,g). Collectively, these results indicate that the induction of senescence in cultured fibroblasts is accompanied by a loss of myofibroblastic phenotype.

| The regulation of myofibroblast phenotype during senescence is mediated by the SASP
To further characterize the regulation of myofibroblast phenotype in senescence, we took advantage of the IMR90 ER:RAS-inducible OIS model to examine the expression kinetics of myofibroblasts markers during senescence induction. Using Western blot and qPCR ). Normal fibroblasts exposed to conditioned medium from OIS cells, but not to control medium, showed a significant drop in myofibroblast markers, similar to that of senescent cells that coincided with a marked increase in the inflammatory SASP marker IL-8 (Figure 2f,g). Taken together, these results clearly indicate that the senescent-associated change in myofibroblast phenotype can be transmitted in a non-cell-autonomous manner, presumably via soluble SASP factors released to the medium.

| Contribution of the fibrogenic and inflammatory phases of the SASP
Recent evidence has shown that the composition of the SASP changes dynamically during senescence implementation, with a Notch/TGFβ-dependent fibrogenic SASP active at early time points, followed by an inflammatory SASP at later times (Hoare et al., 2016;Ito et al., 2017). Given the kinetics of myofibroblast marker expression during senescence and the link to the SASP shown above, we decided to study the role of the different waves of the SASP in the regulation of myofibroblast phenotype. To determine the contribution of the early Notch-dependent SASP, we wondered whether the effects observed upon modulation of the NF-κB pathway in control cells could be linked to TGFβ. Indeed, we found that the TGFβ target TGFBI was significantly upregulated by the NF-κB super-repressor and downregulated by TNFα in control non-senescent cells (Figure 3e and Figure S3d,e). This result suggests that the observed impact on myofibroblast markers under both conditions could, at least in part, be mediated by TGFβ. This is consistent with previous reports of functional crosstalk between the NF-κB and TGFβ pathways (Freudlsperger et al., 2013). Consistent with the above findings, the combined manipulation of the Notch/TGFβ and NF-κB pathways in nonsenescent cells, using inducible NICD and TNFα, showed that TNFα strongly opposed the effects of NICD on ACTA2 and IL-8 expression, at least in part, by inhibition of the TGFβ pathway ( Figure S3f). Collectively, this set of results suggests that the regulation of myofibroblast markers during senescence is mostly linked to the modulation of TGFβ signaling, which may involve both the fibrogenic and inflammatory SASP.

| Dynamic regulation of paracrine transmission of myofibroblast phenotype
To gain a more complete picture of the transmission of the senescence-associated changes in myofibroblast phenotype, in connection with the SASP, we carried out a dynamic transcriptomic analysis of paracrine senescence using RNA-Seq. To this end, normal IMR90 human fibroblasts were exposed to conditioned representative of the successive phases of the SASP (Figure 4a and Figure S4a). As expected (Acosta et al., 2013), exposure to late SASP induced a senescent phenotype in normal IMR90 fibroblasts ( Figure S4b). For simplicity, normal fibroblasts exposed to medium obtained at day 0 post-induction will be designated D0, and so on for the rest of samples (D4, D7, and D10). Volcano plots showed that the intensity of expression changes increases along the kinetics of senescence induction, with the most significant variations detected at D10, especially among upregulated genes ( Figure 4b). Next, we sought to identify sets of genes with common expression patterns, using gene clustering (Figure 4c).  Figure S4d). Collectively, these results confirm the paracrine transmission of the senescence-associated changes in myofibroblast phenotype and suggest that these changes may be associated with the differential regulation of the BMP and TGFβ branches of the TGFβ superfamily.
We next sought to determine whether the cellular response to TGFβ was affected by senescence. To this end, control and senescent fibroblasts were treated with TGFβ. As expected, ectopic TGFβ efficiently induced myofibroblast markers in growing fibroblasts (Hinz et al., 2012); however, this effect was significantly reduced in senescent ER:RAS cells (Figure 5a-c). Similar results were obtained in OIS fibroblasts due to constitutive RAS expression, and in IMR90 shp53/p16 RAS fibroblasts (SASP induction without senescent arrest) ( Figure S5a,b). Additional TGFβ-inducible genes, such as TGFBI and TGFB1I1, also showed blunted induction by TGFβ in ER:RAS-senescent fibroblasts, consistent with defective TGFβ signaling in senescence (Figure 5d). Of note, treatment with TGFβ alone did not cause significant induction of senescence markers ( Figure S7a). The phosphorylation and nuclear translocation of the SMAD2/3 transcription factors are well-established readouts of active TGFβ signaling. Using immunoblot and immunofluorescence, we observed that the total protein levels, phosphorylation, and nuclear accumulation of SMAD2 in response to TGFβ were reduced in OIS fibroblasts relative to non-senescent controls (Figure 5e,f, Figure S6). These results further support the notion that TGFβ signaling is impaired in our cellular model of senescence (Figure 5e,f) providing a potential mechanism for the observed downregulation of myofibroblast phenotype in fibroblast senescence.  Narita, 2019;Di Micco et al., 2021). Here, we have used cellular models of senescence to study the impact of senescence in the regulation of the myofibroblastic phenotype in fibroblasts. This process has important biological implications, given the essential role of myofibroblasts in key physiological and pathological settings, such as wound healing or fibrosis (Hinz et al., 2012). In our studies, we have taken advantage of the fact that primary fibroblasts in standard tissue culture conditions display features of the canonical myofibroblast phenotype. These include the expression of key myofibroblast markers, such as α-SMA, contractile properties, and increased invasion and migration (Ehler et al., 1996). Our data clearly show that the implementation of different forms of senescence consistently leads to the repression of markers and functional features characteristic of myofibroblasts.

| DISCUSS ION
Mechanistically, our co-culture and conditioned medium assays indicate that the senescence-associated changes in myofibroblast phenotype can be transmitted in a paracrine manner. Although a role for extracellular vesicles cannot be formally excluded, our data are consistent with transmission through soluble SASP factors. Recent results have identified the existence of successive waves of SASP along senescence induction (Hoare et al., 2016;Ito et al., 2017). Interestingly, the downregulation of myofibroblast markers in senescence follows a kinetics that is reminiscent of this temporal shift in SASP composition. Our results from the transcriptomic analysis of paracrine senescence and manipulation of signaling pathways collectively suggest that the senescenceassociated changes in myofibroblast phenotype are most likely caused by the temporal changes in the TGFβ pathway observed during senescence. Notably, there is evidence that the two main branches of the TGFβ pathway can have antagonistic effects in myofibroblast differentiation and fibrosis. It is well established than the TGFβ branch is a potent inducer of myofibroblast differentiation with strong profibrotic action (Hinz et al., 2012;Zent & Guo, 2018). Thus, the reduced TGFβ activity in late senescence reported here could clearly contribute to blunting the myofibroblast phenotype. Conversely, members of the BMP branch have been shown to be antifibrotic, including BMP2 and BMP7, upregulated during senescence in our study (Dituri et al., 2019). At this stage, the potential involvement of the BMP branch in control of the myofibroblast phenotype is less clear. Preliminary experiments with recombinant BMP2 and BMP7 in control fibroblasts showed a pro-senescence effect, in line with previous reports (Acosta et al., 2013;Kaneda et al., 2011). However, we have not been able to detect a significant impact of individual BMP factors in myofibroblast marker expression, at least in the conditions used ( Figure S7b). Further studies are clearly warranted to clarify the role of the BMP branch in this context.
The communication of the senescent cell with its environment, including the ECM, is an emerging feature of the senescent phenotype (Fafian-Labora & O'Loghlen, 2020). It has been proposed that interactions between cellular and matrix proteins can regulate senescence (Hiebert et al., 2018;Rapisarda et al., 2017). Our results identify a potential novel layer of regulation of the senescent phenotype related to the crosstalk with the ECM. Mechanical stress from the cell microenvironment is recognized as a major determinant of the myofibroblast phenotype. Our data, in line with previous reports (Hiebert et al., 2018;Mavrogonatou et al., 2017;Mellone et al., 2016), clearly indicate that senescent cells may influence the composition and, presumably, physical properties of the ECM. Moreover, the storage of latent TGFβ or other signaling molecules in the matrix could be similarly affected by senescencespecific changes in the ECM. Thus, it is conceivable that the senescent ECM may provide mechanical and chemical signals critical for the modulation of myofibroblastic traits, and perhaps other features of the senescent phenotype. Previous studies indicate that the link of senescence to myofibroblast differentiation and fibrosis is complex and strongly context-dependent (Schafer et al., 2018). On the one hand, senescence may act as a mechanism limiting fibrosis in liver, heart, and wound healing (Jun & Lau, 2010;Krizhanovsky et al., 2008;Meyer et al., 2016). On the other hand, it has also been proposed that senescence can promote fibrosis in other settings, including lung, kidney, and pancreas fibrotic diseases (Kellogg et al., 2021;Schafer et al., 2017). Our results clearly show the downregulation of myofibroblast markers in cellular models of senescence. Despite differences between our in vitro model and the physiological context, these observations might reflect an alternative mechanism by which senescence could limit fibrosis. In addition to the canonical antiproliferative effect of senescence, which would restrain the expansion of myofibroblasts (Krizhanovsky et al., 2008), our data suggest that senescence could also lead to the reversal of the specialized myofibroblast phenotype, which may synergistically contribute to resolution of fibrosis (Jun & Lau, 2018). Interestingly, a similar differentiation-limiting effect of senescence as the one shown here has been described for muscle differentiation in fibroblasts (Latella et al., 2017). Such a role for senescence in modulating differentiation is in line with previous reports linking senescence to cell plasticity in different contexts (Chiche et al., 2017;Mosteiro et al., 2016;Ritschka et al., 2017) and highlights the complex crosstalk of senescence with cell plasticity regulation. In summary, our results unveil a novel link between cellular senescence and myofibroblastic differentiation, associated with dynamic changes in the TGFβ pathway. These findings may have important implications in the role of senescence in the function of this cell type in physiology and disease.

| Limitations of the study
In this report, we have used different cellular models of fibroblast senescence to address the connection between senescence and specification of the myofibroblast phenotype. Cellular models of se-

| Senescence-associated betagalactosidase staining
Cells were plated at a density of 3 × 10 4 cells per well in six-well plates. The following day, they were fixed and stained as described (Adrados et al., 2016).

| Immunofluorescence
Cells were plated on glass coverslips in six-well dishes (1.5 × 10 5 cells/well). The day after, they were fixed with 4% paraformaldehyde for 20 min, permeabilized with 0.1% Triton X-100 for 15 min, blocked with 2% BSA or 5% serum from the same species as the secondary antibody, and incubated overnight with primary antibodies (see Table S1). Except for primary antibodies coupled to fluorochrome, cells were then incubated with fluorochrome-conjugated secondary antibodies for 2 h at room temperature in the dark. Finally, cells were mounted in Prolong with 1:500 DAPI (D1306; Invitrogen). Images were captured in a confocal microscope (LSM710; Zeiss).

| EdU incorporation
Cells were plated on glass coverslips in six-well dishes (1.5 × 10 5 cells/well). The day after, EdU was added at 10 µM for 6 h (IMR90), After washing, control or senescent fibroblasts were seeded (10 5 cells per 24-well dish). At days 1, 3, and 5, DMEM containing 50 µg/ ml ascorbic acid was added. At day 7, cells were left untreated or removed adding decellularization buffer (20 mM NH 4 OH, 0.5% Triton X-100) for 15 min at room temperature. After three washes with PBS, cells were processed for immunofluorescence as described above.

| Western blot
Protein lysates were prepared with medium salt buffer ( Table S1.

| RNA extraction and quantitative PCR
Total RNA was extracted with TRI Reagent (AM9738; Thermo Fisher Scientific) as described (Gomez-Cabello et al., 2010). RT-PCR was performed with the Applied Biosystems 7900HT PCR System using SYBR Green, at the Genomics Service of the Instituto de Investigaciones Biomédicas (Madrid). 18s ribosomal RNA was used as a reference. Primer sequences are shown in Table S1.

| RNA sequencing
Normal IMR90 fibroblasts were incubated for 3 days with conditioned medium obtained from IMR90 ER:RAS fibroblasts at days 0, 4, 7, and 10 of induction with 4-OHT. Total RNA from two independent experiments (RNA integrity number between 9.5 and 10) was used for RNA sequencing at the Genomics Unit of CNIO (Madrid), essentially as described (De Lope et al., 2019). Differential expression was calculated for each time point (days 4, 7, and 10) relative to day 0, using DESeq2.
Clustering of genes with similar expression kinetics was performed using MeV (Multiple Experiment Viewer) with genes with significant (p < 0.05) differential expression in at least one time point. Twenty clusters were built with K-means clustering using Euclidean distance.
Two clusters with only one gene were discarded and the rest used for further study. Functional enrichment was done with Enrichr software using the following libraries: Transcription, Pathways, and Ontologies.
Heatmaps were built with the Heatmapper tool.

ACK N OWLED G M ENTS
We thank Masashi Narita for reagents and discussions, Miguel

CO N FLI C T O F I NTE R E S T
The authors declare no conflicting interests.