Adaptive RSK‐EphA2‐GPRC5A signaling switch triggers chemotherapy resistance in ovarian cancer

Abstract Metastatic cancers commonly activate adaptive chemotherapy resistance, attributed to both microenvironment‐dependent phenotypic plasticity and genetic characteristics of cancer cells. However, the contribution of chemotherapy itself to the non‐genetic resistance mechanisms was long neglected. Using high‐grade serous ovarian cancer (HGSC) patient material and cell lines, we describe here an unexpectedly robust cisplatin and carboplatin chemotherapy‐induced ERK1/2‐RSK1/2‐EphA2‐GPRC5A signaling switch associated with cancer cell intrinsic and acquired chemoresistance. Mechanistically, pharmacological inhibition or knockdown of RSK1/2 prevented oncogenic EphA2‐S897 phosphorylation and EphA2‐GPRC5A co‐regulation, thereby facilitating a signaling shift to the canonical tumor‐suppressive tyrosine phosphorylation and consequent downregulation of EphA2. In combination with platinum, RSK inhibitors effectively sensitized even the most platinum‐resistant EphA2high, GPRC5Ahigh cells to the therapy‐induced apoptosis. In HGSC patient tumors, this orphan receptor GPRC5A was expressed exclusively in cancer cells and associated with chemotherapy resistance and poor survival. Our results reveal a kinase signaling pathway uniquely activated by platinum to elicit adaptive resistance. They further identify GPRC5A as a marker for abysmal HGSC outcome and putative vulnerability of the chemo‐resistant cells to RSK1/2‐EphA2‐pS897 pathway inhibition.

4. The inhibitor studies by using the pharmacological RSK inhibition to demonstrate the relevance of RSK-EphA2-GPRC5A signaling, should be complemented by the use of RNAi approaches (knock-down of EphA2 or GPCR5A or RSK). 5. To demonstrate that targeting RSK could improve platinum therapy response, the authors should analyze the effect of RSKi in vivo, in HG-SOC xenografts and PDX, providing indication whether combination with platinum can counteract HG-SOC drug resistance. Similarly, as suggested in point 2-3, the combination of RSKi and platinum should be evaluated in terms of apoptosis. 6. Moreover what happens in the experiment where "sensitive" HG-SOC cells are used and EphA2 or GPCR5A or RSK are overexpressed? Would these GPCR5A overexpressing cells now be less sensitive to platinum? This experiment might add further support for targeting this signaling pathway in HG-SOC. 7. The acquisition of drug resistance can be explained in part by intrinsic properties of cancer cells, but could be dependent also by the tumor microenvironment (TME). In order to demonstrate that targeting RSK can disable critical signaling in HGSOC cells, but not alter cell vitality in cancerassociated fibroblasts (CAF), the authors should evaluate the heterogeneous expression of EphA2 or GPCR5A or RSK on tumor cells and CAF. Moreover, it should be interesting to evaluate host component effects in HG-SOC treated with RSKi in mono and combination therapy with platinum co-cultured in the absence and in the presence of CAF. 8. To better define GPRC5A as a predictive marker the authors should provide more details using larger cohort of platinum-sensitive and resistant HG-SOC patients. Moreover the authors should evaluate the expression of RSK-EphA2-GPRC5A axis as predictor signature of poor platinum-based therapy responses and shorter survival in HG-SOC patients.
1st Revision -authors' response 16th January 2020 ***** Reviewer's comments ***** To study the effects of cisplatin and RSKi in OC-CAF co-culture, we first generated red fluorescent OVCAR8-RFP cells, and allowed them to form spheroids as mono-or co-cultures with green fluorescent BjhTERT-GFP fibroblast cell line in non-adherent conditions, followed by treatment with cisplatin and RSKi alone as well as in combination. Using RFP signal intensity at the starting point as well as after 48h and 72h of treatments as a measure of OC cell content, we validated the activity of RSKi, cisplatin and their combination in the 3D cell spheroids. In OVCAR8-RFP monocultures, single treatments with RSKi or cisplatin inhibited RFP signal relative to untreated control, but most notably, combination treatment with RSKi and cisplatin reduced OC cells (RFP intensity) even relative to the treatment start, suggestive of effective cell killing. The combination treatment reduced OVCAR8-RPF most effectively also in co-culture with BjhTERT-GFP fibroblasts. Please see these results below in Figure 1 for review. This data, although further supporting the efficacy of cisplatin-RSKi combination, did not allow us to separate apoptosis and proliferation. Moreover, the strong induction of OVCAR8 cell content by BjhTERT fibroblasts compared to monocultures appeared to contrast with the effect of OC patient-derived CAFs on OVCAR8 in our following experiments now included in the revised manuscript (see below). Therefore, we decided not to include this BjhTERT-OVCAR8 co-culture data in the manuscript. It is included here, however, to further support the suggestion that the cisplatin-RSKi combination can sensitize the cells to platinum-based therapy, as well as to show how differential impact different fibroblasts can have on the OC cells.

Mock Cisplatin+LJH685
Cisplatin LJH685 * ***  As suggested by both Reviewer#1 and #2, to analyze in a highly relevant 3D microenvironment the specific effects of platinum and RSKi treatments in OC apoptosis and proliferation, we prepared spheroids of OVCAR8-RFP and OC patient-derived CAFs under non-adherent conditions, embedded them in 3D collagen and allowed the cells to grow for 4 days. After subsequent treatments with cisplatin and RSKi alone and in combination for 20 h, apoptosis was assessed by immunofluorescence for cleaved caspase-3 (clCasp3). In OVCAR8-RFP 3D monocultures clCasp3/apoptosis was increased by cisplatin alone and by combining LJH685 with cisplatin (new Fig 7A-B). In patient CAF monocultures, cisplatin alone or in combination with LJH685 did not affect clCasp3/apoptosis. Notably, in the 3D co-culture with CAFs, cisplatin or LJH685 alone also failed to enhance OVCAR8-RFP apoptosis significantly, whereas the combined LJH685-cisplatin treatment increased clCasp3 in OVCAR8-RFP (new Fig 7A-B; p = 0.022).
To validate these findings in an even more clinically relevant model, we generated co-cultures of patient-derived OCKI_p13 HGSC cells and OCKI_p22 CAFs and treated them as described above. Notably, combination of LJH685 with cisplatin further enhanced the platinum-induced cancer cell apoptosis (clCasp3) in the co-culture (p = 0.002). These results, indicating that in the relevant TME, RSKi-cisplatin combination can sensitize OC cells to platinum-based therapy, have now been included in the new Fig  7A-D, Appendix Fig S6A-B and described in Results on p. 12 of the revised manuscript. Investigation of this possibility would strengthen the translational impact of the paper. Response: We appreciate the importance of providing experimental evidence to discuss and suggest such translational implications. In the originally submitted manuscript, we had only used MEKi UO126 (at a concentration of 10 µM), showing efficient inhibition of the cisplatin-induced EphA2-pS897 in TYK-nu cells ( Figure 4E). To address this relevant point, we first investigated the in vitro efficacy of two additional MEKi, Refametinib and Trametinib, the latter being an FDA approved drug for solid tumors (1,2). Both these MEKi showed good inhibition of the EphA2 oncogenic EphA2-pS897, while UO126 (now used at a lower concentration of 1 µM based on literature search) was inefficient, and thus not used for further experiments. Please see new Appendix Fig S3E-F and Results on p. 10: "Coincident with ERK1/2-pT202/Y204 inhibition, Trametinib decreased viable OVCAR4 by over 40% compared to untreated control, whereas the relative cell cisplatin sensitivity remained unaltered". To further address this point, we examined the effects of Trametinib-carboplatin combination treatment in cell death (TUNEL) and proliferation (Ki67) in the model of OVCAR4 xenografts in vivo. Trametinib in combination with carboplatin did not induce more apoptosis or significantly reduce proliferation than carboplatin treatment alone. These results have now been included in Appendix Fig S3G-J and described in the revised manuscript Results on p. 10. Altogether, this data suggests that the inhibition of broadly acting, proliferation-driving MEK-ERK1/2 pathway has a different mode of action that may not be directly relevant and related to the specific inhibition of the RSK-EphA2-GPRC5A axis affected by combinatorial treatment with RSKi and platinum. To address this consideration, we have now revised this aspect in Discussion in p. 17.

Were all the cell lines used validated as HGSC origin?
Response: This is an important question, since recent studies have brought into light striking findings on the use of cells that do not recapitulate the mutational landscape of specific OC subtypes, such is the case of HGSC: up to 90% of published studies using "HGSC" cell lines are actually based on cells that are TP53 wild-type and instead have characteristic mutations of other OC subtypes (3). In this study we used TP53 mutant OVCAR3, OVCAR4, OVCAR8, TYK-nu and TYK-nu.R cell lines, from which OVCAR3, OVCAR4, TYK-nu and TYKnu.R have been ranked based on the genetic alterations as highly likely HGSC, while OVCAR8 cell line has been described as likely HGSC (4). This reference and information have now been included in the cell line description in Materials and Methods on p. 19 of the revised manuscript. The HGSC patient-derived passaged cells OCKI_p01 -OCKI_p11 used in the experiments have also been analyzed by TP53 sequencing and tested for nutlin insensitivity as an indication of TP53 mutation status as shown in Appendix Fig S2C. The description of patient-derived cells has also been revised in Materials and Methods on p. 19-20 to clarify the presentation.

Referee #2 (Remarks for Author):
In this article, Moyano-Galceran L et al. investigated the effect of targeting RSK to overcome resistance to platinum-based therapy in high-grade serous ovarian cancers (HG-SOC). The current work provides new insight indicating that chemotherapy-induced RSK-EphA2-GPRC5A signaling switch is associated to resistance to the platinum and that pharmacological RSK inhibition prevented platinum-induced oncogenic EphA2-S897 phosphorylation and EphA2-GPRC5A co-regulation, and in combination with platinum can counteract HG-SOC drug resistance. Further studies should offer more definite insight into the specific mechanisms through which RSK inhibition might represent a possible novel therapeutic strategy for HG-SOC patients. These findings need to be strengthened by additional experiments.

Response:
We thank Reviewer #2 for the interest in our manuscript and for the constructively critical comments that helped us to improve our manuscript. To better define the specific mechanism and strengthen the manuscript, we have performed various additional experiments during the revision, included the results and addressed the points as described below.

The hypothesis that platinum treatment leads to EphA2 upregulation and
EphA2-pS897 phosphorylation in patient-derived HG-SOC cells ex vivo and in HG-SOC cell lines is intriguing, but the data are insufficient to offer more definite insight into the specific mechanism that regulate oncogenic EphA2 phosphorylation switch by platinum chemotherapy in HG-SOC cells. Response: We thank the Reviewer for this constructive criticism, which motivated us to conduct several new experiments and allowed important improvements to the manuscript in terms of the specific mechanism underlying the oncogenic EphA2 phosphorylation switch linked to chemoresistance in HGSC cells.
The following specific insights on these mechanisms have now been included in the revised version of our manuscript: A) Clarified presentation of our original results showing that platinum induces ERK1/2-RSK pathway activation, which correlates with the EphA2-S897 phosphorylation ( Figure 4H-I in the original and revised manuscript). B) New results of siRNA experiments demonstrating that the depletion of specific activity of RSK2 in OVCAR8/4, as well as of RSK1 or RSK2 in TYKnu and TYK-nu.R, will lead to the tumor suppressive EphA2 serine-totyrosine reversal (new Fig 6D- 6E-F; new Fig EV3). D) New results highlighting exclusive GPRC5A induction in HGSC cells, but not in corresponding tumor stroma or CAFs (new Fig 8A-B; Appendix Fig  S7A).
The results C) and D) combined, i.e. the specific function of RSK1 in GPRC5A regulation coupled to cancer-specific GPRC5A induction, will help to explain the specific sensitization of the malignant cells to platinuminduced apoptosis, while stroma remains protected. Platinum chemotherapy is known to induce oxidative stress/ROS-related ERK1/2 activation in different types of malignant and non-malignant cells (5). This notion has now also been included in the revised Discussion. In the originally submitted manuscript, we showed that both ERK1/2 and RSK were activated upon platinum treatment in our OC cell models ( Figure 4H and I). Further, we showed that the EphA2-pS897 phosphorylation was repressed by pharmacological inhibition of either MEK-ERK1/2 axis or its downstream target RSK (RSK1-3 members/proteins inhibited by both the inhibitors; Figure 4E-I).
To clarify specifically the mechanism of the broadly acting ERK1/2-RSK axis in the EphA2 phosphorylation switch, we have now additionally assessed the effects of MEKi in OC platinum responses (new Appendix Fig 3E-J, see also response to comment 2 of Reviewer #1) and RSK expression of RSK1-4 in the OC cells used in the experiments (new Fig  EV3A) Fig EV1D-E).
Further, to assess the effects of platinum alone or combined with RSKi on apoptosis in another highly relevant cell culture model, we stained for clCasp3 3D collagen OVCAR8-RFP spheroids treated with cisplatin alone or in combination with LJH685. When used alone, cisplatin treatment induced apoptosis, while LJH685 alone had no significant effect on apoptosis. When used in combination, LJH685-cisplatin significantly induced apoptosis/clCasp3 (p = 0.002). Likewise, in a 3D co-culture model of patient-derived HGSC cells and CAFs, cisplatin-LJH685 combination further enhanced platinum-induced apoptosis. These results have now been included in new Fig 7A-D, Appendix Fig S6A-B and described in the revised manuscript Results p. 12.
To directly associate the RSK1/2 activities and EphA2-GPRC5A coregulation to apoptosis and proliferation, we further detected cleaved PARP as a marker of apoptosis along with the proliferation marker PCNA by immunoblotting in RSK1/2 siRNA transfected TYK-nu.R. In the platinumtreated cells, siRNA-mediated RSK1 depletion led to GPRC5A suppression coincident with increased cleaved PARP (Fig 6E). Instead, RSK2 depletion increased cleaved PARP in the absence of cisplatin, whereas the proliferation marker PCNA was generally less affected by RSK1/2 knockdown, and even increased after cisplatin treatment in the resistant cells (Fig 6E).
Altogether, these results indicate that both platinum and RSKi-platinum combination treatments primarily decrease cell viability via increased apoptosis.
3. The authors should provide a better understanding of the supportive role of RSK-EphA2-GPRC5A in vivo. How does this work? Is this via a decrease in apoptosis (TUNEL) or an increase in proliferation rate (Ki67)?
Response: This is a valid point. To further address the role of RSK-EphA2-GPRC5A in vivo, we analyzed the OVCAR4 xenograft tumors presented in the originally submitted manuscript for EphA2, EphA2-pS897, GPRC5A, cleaved caspase-3 (clCasp3), TUNEL and Ki67. Carboplatin treatment increased EphA2, EphA2-pS897, clCasp3 and TUNEL, but did not alter Ki67. Notably, EphA2-pS897 and clCasp3 localized to different tumor cells and areas in the carboplatin treated tumors. Therefore, we conclude that the treatment-escaping HGSC cells activated oncogenic EphA2 signaling to evade apoptosis in response to platinum chemotherapy in vivo. These results have now been included in Fig 3E-I, Fig EV2B-D and described in Results p. 8. Further, in two independent in vivo experiments, cisplatin treatment likewise induced apoptosis (detected by TUNEL) but had no major effect on proliferation (assessed by Ki67). In one of these experiments, BI-D1870 was used in combination with platinum for 48 h, significantly increasing TUNEL/apoptosis (2.5 ± 1.8 fold, p = 0.029) but not affecting Ki67/proliferation. These results (presented in Fig 7E-H, Appendix Fig 3G-J) suggest that both platinum and RSKi-platinum combination treatments decreased OC cell viability via increased apoptosis.  (Fig 6E). These results have now been included in Fig 6D- In the originally submitted manuscript we had performed siRNA-mediated silencing of EphA2 and GPRC5A (see Appendix Figure S3C-D and S5A-C).
We showed minor effect of siEphA2 on cell viability, which was consistent with our conclusion "Rather than blocking the entire signaling duality by EphA2 knockdown, the specific RSK-EphA2-pS897 inhibition and reversal to tumor-suppressive EphA2-pY588 correlated with the effective OC cell sensitization to platinum". We also noted an induction of GPRC5A upon siEphA2 as well as induction of EphA2 (total and pS897) upon siGPRC5A in TYK-nu cells. We have now clarified these results in manuscript p. 11. with carboplatin resulted in severe liver toxicity, icterus, weight loss and increased red blood cell sedimentation rate in 50% of the treated mice (see below in Figure 2 for Review) forcing us to quit the dosing. From the shorter dosing scheme, we analyzed apoptosis by TUNEL staining, which revealed a significant induction of apoptosis compared to the control group (p = 0.029). We think that these results serve as convincing proof-of-principle for the mode of action by apoptosis induction. They have now been included in Fig 7E-H and described in the revised manuscript Results on p. 13. To be successful in the in vivo tumor models, the inhibitor molecules need to entail well-optimized pharmacokinetic properties. To best of our knowledge, no good RSKi with favorable pharmaco-kinetic/-dynamic (PK/PD) characteristics in vivo has yet been developed. The commonly used in vitro inhibitors of RSK, LJH685 and BI-D1870, have both been preliminarily tested in PK/PD studies, showing poor drug stability, high clearance and short plasma half-life (6)(7)(8)(9). Although the poor PK of RSKi could be overcome via comprehensive compound optimization, it is far beyond the scope of this study. However, as mentioned above, we conducted small RSKi pilot experiments in our orthotopic model of metastasized ovarian cancer in female SCID mice. The number of mice in these experimental groups was limited to 4-6, as we firmly think that the ethical 3R (replace, reduce, refine) principle governing all animal work should be the principal guideline when working with compounds with known suboptimal PK. This 3R principle also guided our decision to not generate the PDX models suggested by the reviewer for treatment with these compounds. Moreover, the extensive efforts and time frames required for a PDX study would be out of the scope for the revision of this study and will, in our opinion, require another independent study. Response: To address this interesting suggestion by the reviewer, we have now assessed the expression of RSK, EphA2 and GPRC5A in HGSC frozen tissue sections by immunofluorescence and in patient-derived HGSC cells and CAFs by immunoblotting. Immunofluorescence showed that RSK and EphA2 expression was higher in the cancer cells than in the stroma and that GPRC5A localized exclusively in the areas with cancer cells. Immunobloting for RSK, EphA2 and GPRC5A also showed that despite variable expression of these proteins in different patient-derived cancer cells, they were notably more expressed in cancer cells than in CAFs. These results have now been included in Fig 8A-B, Appendix Fig S7A and described in the revised manuscript Results on p. 13. To assess the host component effect in HGSC cells apoptosis, we generated mono-and co-culture spheroids of OVCAR8-RFP/OCKI_p13 cancer cells and patient-derived CAFs and embedded them in 3D collagen.
Please see the answer to Reviewer #1, question 1 for detailed description of these results. We think that these results altogether have markedly improved the revised manuscript. Firstly, the cancer-specific GPRC5A induction, coupled with the direct link between RSK function and platinum induced GPRC5A regulation identified during revision when performing the siRNA experiments (Fig 6,  Fig EV3B), provide a more specific mechanistic insight into the revised study. Secondly, our OC-CAF co-culture results highlight that even in a culture where CAFs seem to further inhibit OVCAR8 platinum response, the RSKi-cisplatin combination can induce apoptosis (new Fig 7).

Response:
We have now clarified the definition of platinum sensitive and resistant HGSC patients (N=136) according to clinical standards regarding the treatment (one should not consider platinum as single agent when defining patient groups, but rather group together patients with platinum single and double treatments). We have accordingly modified the Appendix Materials and Methods section in p. 17, Fig 9E and Appendix Table S7 as well as the Results in p. 14. Moreover, we have analyzed two independent HGSC cohorts (TCGA dataset for OC with N = 578, http://cancergenome.nih.gov/; GSE4997 dataset with N = 204, (10)) to validate the power of the RSK-EphA2-GPRC5A signaling axis as predictor signature. Survival analysis of the TCGA dataset validated our findings on GPRC5A association with worse overall survival in this case at the mRNA level. Moreover, survival analysis of the GSE49997 cohort further uncovered the potential of the combination of EphA2+GPRC5A mRNA expression as an approach to predict progression-free survival of the patients (p = 0.020 for EphA2+GPRC5A with and without RSK1 and RSK2). These results have now been included in Fig 9F-H, Fig EV5 and  Do the data meet the assumptions of the tests (e.g., normal distribution)? Describe any methods used to assess it.
Is there an estimate of variation within each group of data?
Is the variance similar between the groups that are being statistically compared?
For animal studies, the number of mice per treatment group was limited to 4-6 to minimize the number of animals used yet provide adequate statistical power.
For the first animal study, additional RSK inhibitor-treated group (N = 6) were excluded from the final analysis due to extensive liver toxicity. Mice were sacrificed on week 7. For the second animal study, spontaneous death or reaching humane endpoint during the treatment course was used as criteria to exclude individual mice. Analysis and data presented in the paper includes only the individuals surviving until the day of sacrifice.
Mice were randomly assigned into groups while simultaneously ensuring equal distribution of tumor burden (based on bioluminescent signal) within each group. Kolmogorov-Smirnov and Shapiro-Wilk normality tests together with histogram analysis were used for determining the distribution of data. When data did not meet the assumption of normal distribution, appropriate statistical test was used (specific method used for each assay is indicated in corresponding figure legend).
Yes, SD was used.
Yes in most cases, when not, unequal variance test was performed.
Mice were randomly assigned into groups while simultaneously ensuring equal distribution of tumor burden (based on bioluminescent signal) within each group.
Separate or partially separate group of researchers conduted the collection of the patient samples / preparation of TMA /animal studies to those researchers analyzing the results obtained from the experiments. Additionally, see the answers above.
No specific blinding was done (see statement about randomization).

Data
the data were obtained and processed according to the field's best practice and are presented to reflect the results of the experiments in an accurate and unbiased manner. figure panels include only data points, measurements or observations that can be compared to each other in a scientifically meaningful way.
The data shown in figures should satisfy the following conditions: Source Data should be included to report the data underlying graphs. Please follow the guidelines set out in the author ship guidelines on Data Presentation.
Please fill out these boxes ê (Do not worry if you cannot see all your text once you press return) a specification of the experimental system investigated (eg cell line, species name).
Minimum number of technical and biological replicates was set to 3 based on researchers experience. For animal studies, the number of mice per treatment group was limited to 4-6 to minimize the number of animals used yet provide adequate statistical power. For patient studies, sample size was dependent on availability and increased during the revision to improve accuracy.
graphs include clearly labeled error bars for independent experiments and sample sizes. Unless justified, error bars should not be shown for technical replicates. if n< 5, the individual data points from each experiment should be plotted and any statistical test employed should be justified the exact sample size (n) for each experimental group/condition, given as a number, not a range; Each figure caption should contain the following information, for each panel where they are relevant:

B-Statistics and general methods
the assay(s) and method(s) used to carry out the reported observations and measurements an explicit mention of the biological and chemical entity(ies) that are being measured. an explicit mention of the biological and chemical entity(ies) that are altered/varied/perturbed in a controlled manner.
a statement of how many times the experiment shown was independently replicated in the laboratory.
Any descriptions too long for the figure legend should be included in the methods section and/or with the source data.
In the pink boxes below, please ensure that the answers to the following questions are reported in the manuscript itself. Every question should be answered. If the question is not relevant to your research, please write NA (non applicable). We encourage you to include a specific subsection in the methods section for statistics, reagents, animal models and human subjects.
definitions of statistical methods and measures: a description of the sample collection allowing the reader to understand whether the samples represent technical or biological replicates (including how many animals, litters, cultures, etc.).

Reporting Checklist For Life Sciences Articles (Rev. June 2017)
This checklist is used to ensure good reporting standards and to improve the reproducibility of published results. These guidelines are consistent with the Principles and Guidelines for Reporting Preclinical Research issued by the NIH in 2014. Please follow the journal's authorship guidelines in preparing your manuscript.