p38MAPK Plays a Crucial Role in Stromal-Mediated Tumorigenesis

The critical role the tumor microenvironment (TME) plays in disease is underscored by findings that changes within stromal cells can predict clinical outcome (1–3). For this reason, many groups have focused on how various stromal cell types affect tumorigenesis. For example, activated fibroblasts isolated from carcinomas (cancer-associated fibroblasts, or CAF) promote preneoplastic cell growth and increase tumor cell migration, invasion, and angiogenesis (4). Likewise, senescent fibroblasts, which are also found in human tissue (5), support tumorigenesis through the promotion of growth, invasion, and angiogenesis (6–8). Intriguingly, both senescent fibroblasts and CAFs express a plethora of protumorigenic factors, and in senescent cells this is referred to as the senescence-associated secretory phenotype (SASP; refs. 6, 9).

There is significant overlap between the protumorigenic factors expressed in CAFs and senescent cells. Expression array analyses of human fibroblasts treated with granulin, which renders a CAF-like phenotype (10), and fibroblasts isolated from human tumors reveal that both populations express SASP factors (refs. 11, 12, and reviewed in ref. 13). In addition, CAFs isolated by laser capture microdissection (LCM) or via cell surface marker expression similarly display SASP factor expression (1–4, 14). Finally, cells that fail to enter senescence following exposure to a senescence-inducing stress robustly express SASP factors (15, 16), indicating that entrance into senescence is not a prerequisite for SASP expression. Together, these observations raise the possibility that the mechanisms that govern SASP expression are conserved in many tumor-promoting fibroblasts and are not dependent upon the induction of senescence. Thus, identifying mechanisms that activate and sustain SASP expression will have a profound impact on our understanding of the development of a protumorigenic TME and the identification of novel therapeutic targets.

Despite the profound impact the protumorigenic SASP has on tumor cell growth and progression, the mechanisms that lead to its activation and maintenance remain poorly understood. The majority of regulatory pathways elucidated thus far have focused on SASP factor transcription, specifically by NF-κB and CAAT/enhancer binding protein β (C/EBPβ; refs. 15–19). NF-κB's transcriptional activation of the SASP is dependent on the mitogen-activated protein kinase p38 (p38MAPK) and the DNA-damage response protein ATM (19). However, in other systems, p38MAPK facilitates expression of cytokines, including IL6, by affecting posttranscriptional mRNA stability, possibly through the RNA-binding protein AUF1 (20–22). Posttranscriptional regulation of the SASP by p38MAPK is yet to be investigated.

Given the importance of the SASP on stromal-supported tumorigenesis, we investigated the impact of p38MAPK on SASP-mediated tumor promotion. We demonstrate that inhibition of p38MAPK activity abrogates the tumor-promoting capacity of senescent fibroblasts. Furthermore, inhibiting p38MAPK in CAFs inhibits their tumor-promoting abilities, demonstrating for the first time that regulatory mechanisms elucidated in senescent stroma are applicable in CAFs. Finally, we elucidate a p38MAPK-dependent posttranscriptional SASP regulatory pathway that modulates RNA-binding protein activity.

SASP factors promote preneoplastic cell growth (6–8, 23, 24) and p38MAPK contributes significantly to the initiation of SASP factor expression (19). To confirm this, senescent fibroblasts [fibroblasts staining positive for senescence-associated β-galactosidase (SA-β-gal); Supplementary Fig. S1A] were treated with a highly specific small-molecule inhibitor of p38MAPK (SB203580; ref. 25). Heat-shock 27 KDa protein (HSP27) is a direct downstream target of p38MAPK. Therefore, to confirm that our treatment inhibited the kinase activity of p38MAPK, we measured HSP27 phosphorylation by Western blot analysis. We found that SB203580 treatment led to a reduction in HSP27 phosphorylation, indicating successful inhibition of p38MAPK activity (Fig. 1A). As expected, SB203580 treatment of senescent fibroblasts resulted in a significant reduction in the expression of SASP factors IL6, IL8, and granulocyte macrophage colony-stimulating factor (GM-CSF; Fig. 1B). To determine whether p38MAPK activity was responsible for the tumor-promoting activities of senescent cells, we performed coculture experiments with normal human fibroblasts induced to senesce by treatment with bleomycin [referred to as stress-induced premature senescence (SIPS)] and preneoplastic HaCaT keratinocyte cells expressing click beetle red (CBR) luciferase (HaCaT-CBR; ref. 23). Before the addition of HaCaT-CBR cells, fibroblasts were treated with vehicle or SB203580 as indicated in Fig. 1C. Senescent fibroblasts treated with vehicle increased the growth of HaCaT-CBR cells compared with HaCaT-CBR cells cultured with young fibroblasts (Fig. 1D), recapitulating our previously published observations (15, 23). However, though inhibition of p38MAPK had no effect on HaCaT-CBR cells grown in the absence of fibroblasts (Supplementary Fig. S1B), we found that p38MAPK inhibition reduced the protumorigenic activity of senescent fibroblasts by significantly reducing HaCaT-CBR cell growth (Fig. 1D).

Given the potent impact of p38MAPK inhibition in coculture experiments, we next examined the impact of p38MAPK depletion on preneoplastic cell growth in xenograft experiments. p38MAPK was depleted from senescent fibroblasts (Fig. 1E), resulting in a significant reduction in the level of p38MAPK-dependent SASP factor IL8 (Fig. 1F). To assess the impact of p38MAPK loss in vivo, young, senescent, or p38MAPK-depleted senescent fibroblasts were admixed with the preneoplastic epithelial cell line BPH1 expressing CBR luciferase (BPH1-CBR) and injected subcutaneously into nude mice. Tumor growth was analyzed by bioluminescence imaging. As expected, senescent fibroblasts increased BPH1-CBR cell growth relative to young fibroblasts (Fig. 1G). However, depletion of p38MAPK and subsequent reduction in p38MAPK-dependent SASP factor expression reduced tumor growth to the level observed when BPH1-CBR cells were coinjected with young fibroblasts (Fig. 1G). These results indicate that expression of p38MAPK-dependent SASP factors within the TME plays a pivotal role in preneoplastic cell growth in vivo.

We next sought to elucidate the mechanism by which p38MAPK regulates protumorigenic SASP factor expression. Previous work demonstrated that p38MAPK modulates NF-κB–driven transcription of SASP factors including IL6and IL8 (19). To determine that the effects of p38MAPK inhibition were transcriptionally based, senescent fibroblasts were treated with the transcription inhibitor actinomycin D at several time points following bleomycin treatment. SASP factor expression was significantly inhibited when cells were treated with actinomycin D 24 hours after bleomycin treatment (Fig. 2A), a time point at which SASP factor mRNA was increased (Supplementary Fig. S2A) but cells were not yet senescent (Supplementary Fig. S1A). These results indicate that at this time point SASP factor expression is dependent on transcription. Surprisingly, at 96 hours after bleomycin treatment, when cells displayed morphologic features characteristic of senescence, including staining positive for SA-β-gal (Supplementary Fig. S1A), treatment with actinomycin D failed to reduce SASP factor mRNA levels (Fig. 2A). These changes in mRNA were also reflected at the protein level. Indeed, we found that IL6 protein levels in conditioned medium collected from cells treated with actinomycin D at 24 hours fell drastically compared with untreated cells. In contrast, when cells were treated with actinomycin D at 96 hours, IL6 protein levels remained high (Fig. 2B). Given that p38MAPK inhibition at the later time point significantly reduced SASP expression (Fig. 1B), these findings raised the possibility that p38MAPK affects SASP factor mRNA stability rather than NF-κB–driven transcriptional activation upon the acquisition of senescence. To confirm that p38MAPK had no effect on NF-κB–driven transcription at the later time point, normal human fibroblasts were transduced with an NF-κB transcription reporter plasmid driving expression of luciferase (NF-κB-luc). Transduced cells were treated with bleomycin, and 72 hours later, senescent cells were treated with the p38MAPK inhibitor SB203580 for an additional 48 hours. As expected, when SB203580 treatment was initiated 72 hours after bleomycin treatment, there was no significant effect on NF-κB transcriptional activity (Fig. 2C). These results indicate that after the establishment of senescence, p38MAPK has a profound effect on SASP factor mRNA stability.

To address whether SASP factor mRNA stability was affected in cells undergoing replicative senescence or other types of stress-induced senescence, normal human fibroblasts were induced to senesce through telomere dysfunction (replicative senescence) or treatment with the histone deacetylase inhibitor sodium butyrate (NaB). Cells undergoing replicative senescence or NaB-induced senescence robustly induced expression of SASP factors, including IL6 and IL8 (Supplementary Fig. S2B and S2C, respectively). Furthermore, we found that SASP factor mRNAs were significantly stabilized in cells that had undergone replicative senescence or NaB-induced senescence (Fig. 2D and Supplementary Fig. S2D, respectively). Significantly, SASP factor mRNA stabilization was not limited to skin fibroblasts; when IMR90 human lung fibroblasts were treated with bleomycin, they displayed a similar increase in SASP factor mRNA stability 96 hours after bleomycin treatment (Supplementary Fig. S2E). Together, these data indicate that SASP factor mRNAs are stabilized by a posttranscriptional regulatory program that is active in fibroblasts from diverse tissues, regardless of the mechanism through which senescence is induced.

Our results indicate that p38MAPK inhibition reduces SASP expression and TME-dependent promotion of tumorigenesis but does not affect the activity of the primary transcriptional regulator of the SASP, NF-κB, following induction of senescence. Interestingly, p38MAPK posttranscriptionally regulates IL6 and IL8 in other contexts (20, 21). Thus, we investigated p38MAPK's role in stabilizing SASP factor mRNA. We first examined whether p38MAPK was active throughout the time course under investigation. To assess p38MAPK activation, lysates were prepared from cells 24 or 96 hours after bleomycin treatment and examined for phosphorylated p38MAPK (p-p38) by Western blot analysis. In agreement with previous findings (19), we observed that p-p38 increased from 24 to 96 hours following bleomycin treatment (Supplementary Fig. S2F). These kinetics were consistent with SASP factor mRNA stabilization, suggesting that p38MAPK activation regulates SASP factor mRNA stability.

To elucidate p38MAPK's role in regulating SASP factor mRNA stabilization, normal human fibroblasts depleted of p38MAPK (shp38) were treated with actinomycin D 24 or 96 hours after bleomycin treatment (Fig. 2E). When treated with actinomycin D 24 hours after bleomycin treatment, cells expressing shSCR or shp38 displayed decreased SASP mRNA stability, indicating that p38MAPK does not posttranscriptionally regulate SASP mRNAs at this time point. As expected, both IL6 and IL8 mRNA stability increased when shSCR control cells were treated with actinomycin D 96 hours after bleomycin treatment, although not to the same extent as that observed in nontransduced fibroblasts. In contrast, when shp38 cells were treated with actinomycin D 96 hours after bleomycin treatment, they displayed significantly reduced IL6 and IL8 mRNA stability when compared with cells expressing the control hairpin (shSCR; Fig. 2E). Similar results were obtained with a second independent shRNA targeting p38MAPK (data not shown).

We next examined the mechanisms by which SASP factor mRNA was stabilized. The 3′-untranslated region (UTR) of many mRNAs contains protein-binding motifs that alter mRNA stability under diverse biologic stimuli (26). To determine whether the 3′-UTRs of SASP factor mRNAs govern posttranscriptional regulation, we used a luciferase reporter cDNA fused to the 3′-UTR of IL6 (lucIL6) or GM-CSF(lucGM-CSF). A luciferase reporter cDNA fused to the 3′-UTR of GAPDH (lucGAP) was used as a control. Normal human fibroblasts were stably transduced with the luciferase reporter constructs, and luciferase mRNA levels were monitored in response to actinomycin D at the time points indicated (Fig. 3A). Similar to our observations with the endogenous IL6 and GM-CSF transcripts, we observed that the stability of the lucIL6 and lucGM-CSF transcripts increased significantly when treated with actinomycin D 96 hours compared with 24 hours after bleomycin treatment (Fig. 3A). As expected, there was no significant change in the stability of the lucGAP transcript (Fig. 3A), indicating that 3′-UTR–dependent mRNA stabilization was specific for SASP factor mRNA and did not extend to all mRNAs in response to senescence. These results indicate that the 3′-UTR of SASP factor transcripts mediates increases in mRNA stability.

AUF1 is a protein that binds to the 3′-UTRs of many mRNAs including IL6, IL8, and GM-CSF and reduces their stability (27–29). Furthermore, p38MAPK is known to affect AUF1 activity in other settings (22), although a link between AUF1 and p38MAPK in the posttranscriptional regulation of IL6 and IL8 has not been demonstrated. To examine AUF1 binding to SASP factor mRNA in response to senescence, we used RNA-binding protein immunoprecipitation (RIP). Cell lysates were collected 24 and 96 hours after bleomycin treatment and subjected to immunoprecipitation with either an AUF1-specific antibody or a nonspecific IgG; mRNA levels were normalized to the levels of each transcript measured in the input fractions. We observed that AUF1 occupancy on IL6 and IL8 mRNAs significantly decreased from 24 to 96 hours after bleomycin treatment (Fig. 3B), corresponding with the increase in mRNA stability observed in Fig. 2. We observed similar results for GM-CSF and CCL20 mRNA, indicating that this mechanism affects many SASP factor mRNAs (Fig. 3B). This observation suggests that decreased AUF1 binding leads to increased mRNA stability once senescence is established.

We next sought to determine whether AUF1 was required to destabilize SASP mRNAs. To address this question, we stably transduced normal human fibroblasts with two independent shRNA constructs targeting AUF1 (shAUF1a and shAUF1b; Fig. 3C). AUF1-depleted cells were treated with bleomycin and 24 hours later treated with actinomycin D, a time at which AUF1 is bound to SASP factor mRNAs displaying reduced stability (Fig. 3B). In contrast to control cells, we found that AUF1 depletion significantly increased the stability of IL6 and IL8 mRNA at the early time point when these mRNAs are normally unstable (Fig. 3D). These data demonstrate that before senescence is established, AUF1 destabilizes SASP mRNAs by binding to their 3′-UTRs.

To address whether the impact of p38MAPK on SASP mRNA stabilization was due to modulation of AUF1–SASP mRNA binding, we carried out RIP analysis. Following bleomycin treatment, normal human fibroblasts were treated with SB203580 or vehicle control as described in Fig. 3E. In contrast to control cells in which AUF1 occupancy decreased at the late time point, there was no decrease in AUF1 occupancy on IL8 mRNA in p38MAPK-inhibited cells collected 96 hours after bleomycin treatment (Fig. 3E). Similar results were obtained for IL6 (data not shown). These observations indicate that p38MAPK activation is required to release AUF1 from SASP factor mRNA. Furthermore, these studies suggest that loss of SASP factor mRNA stabilization in p38MAPK-inhibited cells (Fig. 2) is the result of a failure to remove AUF1 from SASP factor transcripts.

The TME plays a pivotal role in tumor progression, and recent expression analyses indicate that TME-specific expression changes are predictive of clinical outcome (1–3). Both senescent fibroblasts and CAFs express protumorigenic SASP factors, raising the intriguing possibility that the regulatory mechanisms that control SASP expression in senescent cells also operate in cancer-associated stroma. Given the importance of p38MAPK in SASP factor expression, we carried out a meta-analysis to establish a list of p38MAPK-regulated genes in senescent fibroblasts and evaluated their expression in the TME of human breast cancers (Fig. 4A). We performed RNA sequence (RNA-seq) analysis of young, senescent, and p38MAPK-inhibited senescent human fibroblasts and observed that IL6 and IL8 expression was p38MAPK-dependent. Along with previously identified factors, we found that 50 additional SASP factors were p38MAPK-dependent, including GM-CSF, granulocyte CSF (G-CSF), IL1α,IL1β, CXCL1, CXCL2, CXCL5, CCL20, MMP1, and MMP7 (Supplementary Table S1 and Fig. 4A). A subset of these factors was validated by quantitative real-time PCR (qRT-PCR; Supplementary Fig. S3A). Gene Ontology (GO) process analysis performed on the p38MAPK-dependent factors demonstrated that genes related to the regulation of inflammation, chemotaxis, cell adhesion, angiogenesis, and proliferation were significantly enriched in this gene set (Fig. 4B).

We next compared our p38MAPK-dependent SASP list with factors significantly overexpressed in the TME of breast cancer lesions. We examined three datasets generated from microarray analyses of normal stroma versus cancer-associated stroma that had been obtained by LCM of breast tissue (Fig. 4A; refs. 1–3). Of the 50 p38MAPK-dependent factors identified in senescent fibroblasts, we found that 29 were expressed in the stroma of the Finak breast cancer dataset (1), including CXCL2 and IL24. Seventeen factors were expressed in the TME of the Ma breast cancer dataset (2), including IL1β. Finally, seven factors overlapped with the Karnoub breast cancer dataset (3), including CCL20 (Supplementary Table S1). Furthermore, CCL20, CXCL5, IL11, IL1β, IRAK3, MMP1, MPP7, and SOD2 were expressed in the breast cancer–associated stromal compartment of at least two studies (Fig. 4C). Of note, CCL20, CXCL5, IL11, IL1β, and MMP1 are factors with known protumorigenic activities (30–34). We observed that breast cancer–associated stromal genes compose a large percentage of the total number of p38MAPK-dependent SASP factors involved in the regulation of inflammation, chemotaxis, angiogenesis, and cell adhesion based on GO process analyses (Fig. 4B). Given that these factors are associated with disease progression, our findings raise the possibility that anti-p38MAPK therapy could significantly affect tumor progression in humans.

Expression of p38MAPK-dependent factors within the stroma of breast cancer lesions raised the possibility that they contribute to the tumor-promoting activities of CAFs. Therefore, we examined whether inhibition of p38MAPK would abrogate the tumor-promoting activities of CAFs as it did for senescent fibroblasts. To generate CAFs, we obtained normal human mammary fibroblasts (NMF) from reduction mammoplasty, admixed them with MCF7-RAS breast carcinoma cells, and injected the cell mixture into immunocompromised mice and allowed tumors to grow. Human CAFs were isolated from these tumors and we assessed their tumor-promoting potential by coculturing them with preneoplastic HaCaT skin keratinocytes expressing CBR luciferase (HaCaT-CBR). As expected, CAFs significantly stimulated HaCaT-CBR cell growth compared with HaCaT-CBR cells cultured with parental NMFs (Fig. 4D). To investigate the importance of p38MAPK-dependent CAF factors, we inhibited p38MAPK in CAFs with SB203580 and assessed their ability to promote preneoplastic cell growth. Similar to what we observed when senescent fibroblasts were treated with the p38MAPK inhibitor SB203580 (Fig. 1C), CAFs treated with SB203580 failed to promote HaCaT-CBR cell growth (Fig. 4D). These results indicate that p38MAPK regulates the tumor-promoting activity of CAFs. Together with our meta-analysis and expression of p38MAPK-dependent genes in the stromal compartment of human breast cancer lesions, these observations suggest that p38MAPK plays a central role in sustaining the expression of tumor-promoting factors. Thus, stromal p38MAPK represents a novel therapeutic target for senescent and nonsenescent cancer-associated stromal compartments in breast cancer.

The critical importance of SASP factor expression in our tumor models and our work to uncover the mechanisms that sustain SASP factor expression identified p38MAPK as a central player in SASP expression in senescent cells as well as in CAFs. Given our findings that p38MAPK-dependent factors are expressed in human breast cancer lesions, we evaluated the feasibility of targeting p38MAPK in a preventative and therapeutic setting. Several p38MAPK inhibitors have entered phase II clinical trails for rheumatoid arthritis and thus have proved safe in a nonlethal disease (35, 36). We obtained a p38MAPK inhibitor (CDD-111, also referred to as SD-0006; ref. 37; Confluence Life Sciences, Inc.) and compounded it into mouse chow. CDD-111 was chosen because it can be orally administered and shows high specificity for the p38MAPKα subunit (37). Indeed, extensive analysis of CDD-111 revealed that it is selective for p38MAPKα over 50 other kinases including p38MAPKβ, p38MAPKγ, and p38MAPKδ. Furthermore, the IC₅₀ for inhibiting TNFα release in vitro and in vivo was less than 200 nmol/L (37). Treatment of senescent cells with CDD-111 in vitro revealed that it effectively reduced SASP expression as evidenced by a significant reduction in IL6 and IL8 levels (data not shown).

To establish the impact of orally administered CDD-111 on p38MAPK activity in our system, mice were placed on CDD-111 (p38i) or control chow for 3 days, challenged with lipopolysaccharide (LPS), and serum TNFα levels were measured. We found that mice receiving oral p38i failed to mount a robust TNFα response following an LPS challenge compared with animals receiving control chow (Supplementary Fig. S3B). We also verified that p38i inhibited SASP expression in vivo. Senescent normal human fibroblasts were injected subcutaneously into the rear flanks of nude mice maintained on control or p38i chow. Ten days after injection, the cells were removed, RNA was isolated, and the levels of human IL8 were analyzed by qRT-PCR. Senescent fibroblasts isolated from mice on p38i had significantly less IL8 mRNA than senescent fibroblasts isolated from mice on control chow (Fig. 4E), demonstrating that CDD-111 inhibited SASP expression in vivo.

We next evaluated the p38MAPK inhibitor's efficacy in a xenograft setting. BPH1-CBR cells admixed with young or senescent fibroblasts were subcutaneously injected into mice maintained on control or p38i chow (Fig. 4F). Bioluminescence analysis of tumor growth revealed that p38i significantly reduced the growth of BPH1-CBR cells coinjected with senescent fibroblasts (Fig. 4G and H). Analysis of cellular proliferation (Ki67 staining) revealed that senescent fibroblasts significantly increased BPH1 cell proliferation compared with when BPH1 cells were coinjected with young fibroblasts (Fig. 5A). Importantly, the increase in BPH1 proliferation that was noted in the presence of senescent fibroblasts was markedly reduced when mice were maintained on p38i versus control chow (Fig. 5A). These data demonstrate that the reduced tumor size observed in response to p38i administration was a result of decreased epithelial cell proliferation. Importantly, the difference in epithelial cell proliferation between tumors containing senescent stroma from p38i- and control-fed mice was not due to differences in stromal composition between these tumor types, as staining for a senescence marker, p16, and a fibroblast marker, vimentin, demonstrated that (i) senescent fibroblasts persisted throughout the time course of the experiment regardless of p38MAPK inhibition and (ii) the stromal composition of treated and untreated tumors was similar. Vascularity and myeloid infiltration were also investigated in these tumors. No significant differences in either vascularity or leukocyte infiltration were noted (data not shown). Administration of CDD-111 to mice injected with BPH1 cells admixed with young fibroblasts also resulted in a decrease in epithelial cell growth, although not to the same extent as that observed in tumors containing senescent fibroblasts. Oral administration of CDD-111 had no significant impact on BPH1-CBR cells injected alone (data not shown).

To address the effectiveness of p38MAPK inhibition in a therapeutic setting, mice were injected with BPH1-CBR cells admixed with senescent fibroblasts, and tumors were allowed to grow for 1 week until the average tumor volume reached 74 mm³. Mice were then administered control or p38i chow and bioluminescence imaging was used to monitor tumor growth. Significantly, tumor growth was arrested in mice receiving p38i. In contrast, tumors in mice receiving control chow continued to show significant growth (Fig. 5B).

To investigate the applicability of orally administered p38i in CAF-containing microenvironments, we obtained primary CAFs (pCAF) from a lesion removed from a patient with invasive breast cancer. We subcutaneously injected BPH1 cells alone or BPH1 cells admixed with pCAFs into nude mice fed either control or p38i chow as described for the experiments in Fig. 4F. As expected, there was no difference in BPH1 cell growth whether mice were fed control or p38i chow (Fig. 5C). Importantly, in mice receiving control chow, BPH1 cells admixed with pCAFs grew significantly more than BPH1 cells injected alone, verifying that our patient-derived fibroblasts were bona fide CAFs (Fig. 5C). pCAF-mediated BPH1 growth was significantly inhibited in mice receiving p38i (Fig. 5C), similar to what was observed with senescent fibroblast-mediated BPH1 growth (Fig. 4G and H). These findings, combined with those from p38MAPK inhibition of senescent-fibroblast driven tumors, suggest that p38MAPK is a viable, stromal-specific therapeutic target that may show efficacy in diverse TMEs and diverse tumor types.

The regulation of SASP expression is complex, involving the DNA damage response (16), HDAC1 activity (15), and transcriptional regulation by NF-κB and C/EBPβ (17–19). p38MAPK perhaps best exemplifies the complexity of SASP regulation. Previous reports have shown that p38MAPK affects NF-κB–driven transcriptional control of SASP expression immediately following exposure to a senescence-inducing signal (19). In our system, p38MAPK inhibition had no effect on NF-κB transcriptional activity when it was initiated after cells acquired the senescent phenotype as evidenced by SA-β-gal staining. However, p38MAPK inhibition did have a significant impact on SASP factor mRNA stability. Our data are consistent with p38MAPK playing a dual role in SASP factor expression. We hypothesize that SASP factor expression is achieved through early rounds of transcription followed by posttranscriptional mRNA stabilization, both of which require distinct p38MAPK functions.

Inhibiting the SASP represents a novel stromal-specific therapeutic cancer modality that could be beneficial at multiple stages of tumorigenesis. We have demonstrated that senescent cells are present in the microenvironment before the formation of preneoplastic lesions and that SASP factors promote preneoplastic cell growth (15, 23). The SASP also promotes more aggressive malignancies by increasing angiogenesis and invasion (9, 38). Finally, the SASP is hypothesized to promote later events in cancer progression including metastasis and recurrence through its promotion of cancer stem cell formation and chemoresistant niches (7, 39, 40). Together, these findings suggest that inhibition of the SASP will prevent the development and/or progression of malignancies. p38MAPK could provide an ideal target as it affects both the transcriptional and posttranscriptional regulation of SASP (19) and may be particularly effective because it can inhibit SASP expression after the stabilization of SASP mRNAs has already occurred.

Our findings that oral administration of a p38MAPK inhibitor dramatically inhibits SASP-mediated tumor growth driven by senescent fibroblasts and CAFs indicate for the first time that the tumor-promoting capabilities of senescent fibroblasts and CAFs are mediated through similar signaling pathways. Furthermore, these findings suggest that p38MAPK is an important therapeutic target with wide applicability in a variety of tumor-promoting microenvironments. This is strengthened by our in silico analysis of the stromal compartment of breast cancer lesions, which we show express many p38MAPK-dependent genes. These data are intriguing in light of the fact that p38MAPK inhibitors have moved into phase II and III clinical trials for inflammatory diseases, including rheumatoid arthritis, Crohn disease, and psoriasis, demonstrating their tolerability in patients (35, 36). Given our findings, we suggest that p38MAPK inhibitors warrant investigation for use as antineoplastic therapy.

BJ human foreskin fibroblasts were obtained from Dr. Robert Weinberg (Massachusetts Institute of Technology, Cambridge, MA) and were cultured as previously described (23). IMR90 human lung fibroblasts were purchased from the American Type Culture Collection and were cultured in DMEM supplemented with 10% FBS (Sigma) and 1% penicillin/streptomycin. Patient-derived breast CAFs were purchased from Asterand and cultured in DMEM supplemented with 10% FBS, 1 μg/mL hydrocortisone, 5 μg/mL transferrin, 5 μg/mL insulin, and 1% penicillin/streptomycin. Fibroblasts were treated with bleomycin sulfate (100 μg/mL; Sigma) for 24 hours, followed by incubation in normal culture medium for the time points indicated. Fibroblasts were treated with actinomycin D (10 μg/mL; Sigma) for 24 hours, SB203580 (10 μmol/L; Millipore) for 48 hours, or CDD-111 (also referred to as SP-006; 1 μmol/L; Confluence Life Sciences, Inc.) for 48 hours, unless indicated otherwise. SB203580 and CDD-111 were replenished daily. Fibroblasts were treated with two fresh changes of 4 mmol/L NaB (Sigma) for 72 or 120 hours. RNA was isolated using TRI Reagent (Life Technologies) at the time points indicated. HaCaT preneoplastic keratinocyte cells (obtained from Dr. Norbert E. Fusenig, German Cancer Research Center, Heidelberg, Germany) stably expressing CBR luciferase (HaCat-CBR; ref. 16) were grown in DMEM supplemented with 10% heat-inactivated FBS and 1% penicillin/streptomycin (Sigma). BPH1 preneoplastic prostate epithelial cells (obtained from Dr. Robert Weinberg, Massachusetts Institute of Technology, Cambridge, MA) stably expressing CBR luciferase (BPH1-CBR) were grown in DMEM supplemented with 10% non–heat-inactivated FBS and 1% penicillin/streptomycin. All cells were cultured at 37°C in 5% carbon dioxide and 5% oxygen. No cell lines used were authenticated.

The luciferase reporter construct fused to the 3′-UTR of IL6 (lucIL6) was a gift from Dr. Nicholas Davidson (Washington University School of Medicine, St. Louis, MO) and was subcloned into the EcoRI site of pBABE-hygro. Luciferase reporter constructs fused to the 3′-UTR of GM-CSF or GAPDH were purchased from Switch Gear Genomics and were subcloned into pBABE-hygro using the SnabI and SalI restriction sites. shRNA sequences targeting human AUF1 (shAUF1A: 5′-AGAGTGGTTATGGGAAGGTAT-3′, shAUF1B: 5′-AGTAAGAACGAGGAGGATGAA-3′), p38 (5′-GCCGTATAGGATGTCAGACAA-3′), and Hsp27 (5-CCCGGACGAGCTGACGGTCAA-3′) were obtained from the Children's Discovery Institute's (St. Louis, MO) viral vector–based RNA interference (RNAi) core at Washington University (St. Louis, MO), and were supplied in the pLKO.1-puro backbone. Luciferase reporter assays were performed using a plasmid containing an NF-κB–responsive promoter driving expression of firefly luciferase (NF-κB-luc) and a plasmid encoding Renilla luciferase driven by the thymidine kinase promoter, obtained from Dr. David Piwnica-Worms (Washington University School of Medicine).

SA-β-gal staining was carried out as described previously (23).

cDNA synthesis and quantitative real-time PCR (qRT-PCR) were performed using previously published protocols and manufacturers' instructions (ref. 41; SYBR Green; Life Technologies). Primers for GAPDH (F: 5′-GCATGGCCTTCGGTGTCC-3′, R: 5′-AATGCC AGCCCCAGCGTCAAA-3′), IL6 (F: 5′-ACATCCTCGACGGCATCT CA-3′, R: 5′-TCACCAGGCAAGTCTCCTCA-3′), IL8 (F: 5′-GCTCTGT GTGAAGGTGCAGT-3′, R: 5′-TGCACCCAGTTTTCCTTGGG-3′), MMP3 (F: 5′-GTTTTGGCCCATGCCTATGCCCC-3′, R: 5′-GGAGTCAGGGGGAGGTCCATAGAGG-3′), CCL20 (F: 5′-CTGCGGCGAATCAGAAGCAGC-3′, R: 5′-CCTTCATTGGCCAGCTGCCGT-3′), lucIL6 (F: 5′-CGGGCGCGGTCGGTAAAGTT-3′, R: 5′-AAACAACAACGGCGGCGGGA-3′), lucGM-CSFand lucGAP (F: 5′-GAGAAACATGCGGAGAACGC-3′, R: 5′-AGCATGCACGATAGCCTTGA-3′) were purchased from IDT. GM-CSF cDNA was amplified using a TaqMan probe/primer set (catalog number Hs00929873_m1; Life Technologies).

Conditioned medium was generated by incubating cells for 24 hours in serum-free medium. Following collection, secreted IL6 protein levels were measured using the human IL6 Quantikine ELISA Kit (catalog number D6050; R&D Systems).

Cell pellets were lysed in buffer containing 50 mmol/L Tris pH 8.0, 5 mmol/L EDTA, 0.5% NP-40, and 100 mmol/L sodium chloride for 20 minutes at 4°C. Protein concentration was quantified using the Bradford Protein Assay (Bio-Rad). The primary antibodies used were polyclonal AUF1 (Millipore; catalog number 07260MI) diluted at 1:3,000, polyclonal p-p38 (Phospho Solutions; catalog number p190-1802) diluted at 1:1,000, polyclonal p38 (Cell Signaling Technology; catalog number 9218) diluted at 1:1,000, monoclonal β-catenin (BD Biosciences; catalog number 610153) diluted at 1:5,000, and monoclonal α-tubulin (Abcam; product number ab6160) diluted at 1:1,000. All secondary antibodies from the appropriate species were horseradish peroxidase–conjugated (The Jackson Laboratory) and diluted at 1:10,000.

Virus was produced as described previously (23).

Cell pellets from 7 × 10⁷ BJ fibroblasts were lysed in the same buffer used for Western blot analysis. Protein concentration was analyzed using the Bradford Protein Assay (Bio-Rad); 3 mg of protein was used for each immunoprecipitation. The following primary antibody was used: 30 μg of polyclonal AUF1 (Millipore; catalog number 07260MI). Equivalent amounts of normal IgG antibody (Cell Signaling Technology) were used to control for specific immunoprecipitation. Cell lysates were precleared with 20 μL protein A Dynabeads (Life Technologies) for 30 minutes at 4°C before immunoprecipitation. Protein A Dynabeads (100 μL) were used for each immunoprecipitation. Beads were washed three times in 0.1 mol/L monosodium phosphate and then incubated in 0.1 mol/L monosodium phosphate with the appropriate antibody for at least 1 hour at room temperature. Beads were then washed three times in Buffer A (1× PBS, 0.1% SDS, 0.3% sodium deoxycholate, and 0.3% NP-40), followed by incubation for 30 minutes at room temperature in NT2 buffer (50 mmol/L Tris pH 7.4, 150 mmol/L sodium chloride, and 1 mmol/L magnesium chloride). Antibody-bound beads were then added to precleared cell lysates, and immunoprecipitated overnight at 4°C. Cell lysate (100 μL) was removed from the IgG immunoprecipitation to be used for input controls. Immunoprecipitated beads were washed two times with each of the following buffers: Buffer A, Buffer B (5× PBS, 0.1% SDS, 0.5% sodium deoxycholate, and 0.5% NP-40), and Buffer C (50 mmol/L Tris pH 7.4, 10 mmol/L magnesium chloride, and 0.5% NP-40). Beads were then resuspended in NT2 containing 0.1% SDS, 80 U RNase OUT (Life Technologies), and 30 μg Proteinase K and incubated at 55°C for 30 minutes. RNA was isolated from the beads by adding 1 mL of TRI Reagent (Life Technologies). Following cDNA synthesis, mRNA levels of SASP factors were analyzed by qPCR using the primers and procedures described above.

BJ fibroblasts were transiently transfected with plasmids encoding NF-κB-luc and Renilla luciferase. Renilla luciferase expression was used to standardize for transfection efficiency. Transfection was performed using manufacturer's protocol for the Lipofectamine 2000 reagent (Life Technologies). Luciferase activities were measured 48 hours posttransfection using live-cell imaging as described previously (42).

Coculture experiments were performed as previously described with the following modifications (23). A total of 1.3 × 10⁴ fibroblasts were plated in black-walled 96-well plates (Fisher Scientific). Cells were incubated in starve medium (DMEM + 1% penicillin/streptomycin) for 3 days before the addition of HaCat-CBR cells. SB203580 was refreshed daily until HaCaT-CBR plating. HaCat-CBR cells were cultured in starve medium for 24 hours before plating on fibroblasts. A total of 1.0 × 10³ HaCat-CBR cells were plated on fibroblasts and incubated for the indicated length of time. At the times indicated, d-luciferin (Biosynth) was added to a final concentration of 150 μg/mL. After 10 minutes, plates were imaged using an IVIS 100 camera (PerkinElmer) using the following settings: exposure = 10 seconds–5 minutes, field of view = 15, binning = 16, f/stop = 1, open filter.

A total of 1 × 10⁶ BPH1-CBR preneoplastic prostate epithelial cells were coinjected with 1 × 10⁶ BJ human foreskin fibroblasts. Cells were injected subcutaneously in a 50:50 mixture of DMEM:growth factor–reduced Matrigel (BD Biosciences) into the rear flanks of female NcR nude mice (Taconic). In vivo bioluminescence imaging was performed on the days indicated on an IVIS 100 (PerkinElmer; Living Image 3.2; 1–60-second exposures; binning, 4, 8, or 16; field of view, 15 cm; f/stop, 1; open filter) following intraperitoneal injection of d-luciferin (150 mg/kg; Biosynth). For analysis, total photon flux (photons/s) was measured from a fixed region of interest over the xenografts using Living Image 2.6 (PerkinElmer).

Total RNA was isolated using TRI Reagent (Life Technologies) and the RiboPure RNA Isolation Kit (Life Technologies) following the manufacturer's instructions. rRNA was removed by poly-A selection using oligo-dT beads. mRNA was then fragmented and reverse transcribed to yield double-stranded cDNA using random hexamers. cDNA was blunt ended, had an A base added to the 3′ ends, and then had Illumina sequencing adapters ligated to the ends. Ligated fragments were then amplified for 12 cycles using primers incorporating unique index tags. Fragments were sequenced on an Illumina HiSeq-2000 using single reads extending 50 bases. Raw data were de-multiplexed and aligned to the reference genome using TopHat. Transcript abundances were then estimated from the alignment files using Cufflinks. EdgeR was used for differential expression analysis.

Primary breast tissue was collected without patient identifiers in compliance with a protocol approved by the Brigham and Women's Hospital (Boston, MA; Institutional Review Board 93-085). Fibroblasts were isolated (10, 11) and immortalized through expression of hTERT-GFP (43) as previously described.

To generate CAFs, 3 × 10⁶ human mammary fibroblasts were coinjected with 1 × 10⁶ MCF7-Ras tumor cells subcutaneously into nude mice. After tumors reached 1 cm, mice were euthanized and CAFs were re-isolated by digesting tissues in 1 mg/mL collagenase A for 1 to 4 hours at 37°C with continuous rotation. Resulting cell suspensions were dispersed with an 18-gauge needle, washed two times with resuspension buffer [2% heat-inactivated fetal calf serum in sterile Hank's Balanced Salt Solution (HBSS)], and filtered through 70-μm nylon mesh. GFP⁺ CAFs were then isolated by fluorescence-activated cell sorting and maintained under their standard culture conditions. CAFs were confirmed to be human by staining with human-specific mitochondrial DNA (data not shown).

The p38MAPK small-molecule inhibitor CDD-111 (Confluence Life Sciences, Inc.) was compounded at 516 ppm with Purina Rodent Chow #5001 to generate a daily exposure of 80 mg/kg/d. Female NcR nude mice (Taconic) were fed ad libitum.

Female NcR nude mice (Taconic) were fed ad libitum for 3 days. LPS (100 ng) (Sigma) was then administered by intraperitoneal injection. Serum was collected 1 hour after LPS dosage. TNFα levels were analyzed by ELISA (R&D Systems).

Following excision, tumors were fixed in 10% formalin and embedded in paraffin for sectioning. Standard hematoxylin and eosin (H&E) technique was used for all sections. Serial sections were stained for Ki67 (1:50; catalog number 550609; BD Bioscience), p16 (1:100; catalog number sc-1661; Santa Cruz Biotechnology), and vimentin (1:700; catalog number ab45939; Abcam).

Data are presented as the mean ± SEM. Statistical significance was determined using the Student t test, with a P < 0.05 considered significant. Percentage of mRNA remaining was calculated as the fold mRNA in actinomycin D–treated SIPS cells over untreated SIPS cells. Overrepresented GO terms in the expression data were identified using a Fisher exact test, with a significance threshold of P < 0.05 as implemented in GOstat (44).

No potential conflicts of interest were disclosed.

Conception and design: E. Alspach, K.C. Flanagan, E. Pazolli, J. Monahan, S.A. Stewart

Development of methodology: E. Alspach, K.C. Flanagan, M.K. Ruhland, E. Pazolli, T. Marsh, S.S. McAllister, S.A. Stewart

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): E. Alspach, K.C. Flanagan, H. Huang, D. Piwnica-Worms, J. Monahan, S.S. McAllister, S.A. Stewart

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): E. Alspach, K.C. Flanagan, M.J. Donlin, D. Piwnica-Worms, J. Monahan, D.V. Novack, S.A. Stewart

Writing, review, and/or revision of the manuscript: E. Alspach, K.C. Flanagan, M.K. Ruhland, E. Pazolli, M.J. Donlin, D. Piwnica-Worms, J. Monahan, D.V. Novack, S.S. McAllister, S.A. Stewart

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): E. Alspach, X. Luo, M.K. Ruhland, H. Huang, T. Marsh, S.A. Stewart

Study supervision: S.A. Stewart

The authors thank Dr. Nicholas Davidson for the lucIL6 construct and advice on the RIP protocol; Julie Prior and The BRIGHT Institute at Washington University School of Medicine for live animal imaging and advice, and Amey Barakat for technical assistance with mitochondrial DNA staining. RNAi constructs for this project were obtained through the Broad Institute and funded in part by the Children's Discovery Institute of Washington University in St. Louis. The authors also thank the Genome Technology Access Center in the Department of Genetics at Washington University School of Medicine for help with genomic analysis. The Center is supported by NCI Cancer Center Support Grant P30 CA91842 to the Siteman Cancer Center and by ICTS/CTSA Grant UL1RR024992 from the National Center for Research Resources (NCRR). The authors further thank Dr. John Edwards for assistance in RNA-seq analyses; Mr. Daniel Teasley, Drs. Michelle Hurchla, Katherine Weilbaecher, Gregory Longmore, Andrey Shaw, and Kendall Blumer for critical reading of the article; Dr. Craig Allred for advice on human expression; and Mr. Adnan Elhammali for assistance with GO processes analysis.

This study was financially supported by NIH 5 R01 CA130919 (to S.A. Stewart), NIH Cellular Biochemical and Molecular Sciences Predoctoral Training Grant T32 GM007067 (to E. Alspach and K.C. Flanagan), an American Cancer Society Research Scholar Award (to S.A. Stewart and S.S. McAllister), and NIH RO1 CA166284-01 (to S.S. McAllister). Molecular imaging was funded by NIH P50 CA94056 (to D. Piwnica-Worms). Histologic analysis was supported by NIH P30 AR057235 to the Washington University Musculoskeletal Research Center (to D.V. Novack). The RNAi constructs were obtained from the Viral-vector–based RNAi Core at Washington University, which is supported by a grant from the Children's Discovery Institute and Broad Institute.