|Home | About | Journals | Submit | Contact Us | Français|
The PI3K signaling pathway regulates diverse cellular processes, including proliferation, survival, and metabolism, and is aberrantly activated in human cancer. As such, numerous compounds targeting the PI3K pathway are currently being clinically evaluated for the treatment of cancer, and several have shown some early indications of efficacy in breast cancer. However, resistance against these agents, both de novo and acquired, may ultimately limit the efficacy of these compounds. Here, we have taken a systematic functional approach to uncovering potential mechanisms of resistance to PI3K inhibitors and have identified several genes whose expression promotes survival under conditions of PI3K/mammalian target of rapamycin (PI3K/mTOR) blockade, including the ribosomal S6 kinases RPS6KA2 (RSK3) and RPS6KA6 (RSK4). We demonstrate that overexpression of RSK3 or RSK4 supports proliferation upon PI3K inhibition both in vitro and in vivo, in part through the attenuation of the apoptotic response and upregulation of protein translation. Notably, the addition of MEK- or RSK-specific inhibitors can overcome these resistance phenotypes, both in breast cancer cell lines and patient-derived xenograft models with elevated levels of RSK activity. These observations provide a strong rationale for the combined use of RSK and PI3K pathway inhibitors to elicit favorable responses in breast cancer patients with activated RSK.
The PI3Ks, PKB/AKT, and mammalian target of rapamycin (mTOR) axis is integral for various physiological processes, including proliferation, growth, survival, and metabolism. Mutations of several components of the PI3K pathway that lead to constitutive activation of this pathway are found in human cancer. In particular, members of the class IA PI3K family, which are heterodimers comprising a p85 regulatory and a p110 catalytic subunit, are frequently mutated in solid tumor types, including breast, lung, ovarian, prostate, colorectal, and pancreatic cancers (1–3). Another frequent alteration leading to activation of PI3K signaling in human cancers is the inactivation of the phosphatase and tensin homolog (PTEN) tumor suppressor through somatic mutations that result in protein truncation, homozygous or hemizygous deletions, or epigenetic silencing (4, 5). Additionally, other commonly mutated and/or amplified genes are upstream regulators of the PI3K pathway, including EGFR, HER2, IGFR, MET, and RAS, and are known to promote tumorigenicity, at least in part through the upregulation of PI3K signaling (1, 2).
Due to the importance of PI3K pathway activation in human cancer, several small molecule inhibitors targeting the PI3K/AKT/mTOR pathway are currently under clinical development for treatment of cancer. The macrolide rapamycin and its analogs, such as RAD001 (everolimus), specifically inhibit mTORC1 and have profound cytostatic activity in preclinical models. Everolimus has been shown to provide clinical benefit in treatment of advanced renal cell carcinoma (6), neuroendocrine pancreatic tumors (7), and most recently, in hormone receptor–positive breast cancer, where it significantly delays disease progression when given in combination with hormonal therapy (8). Several recent reports have also demonstrated activity of PI3K inhibitors in preclinical models in specific subsets of breast cancer cells, including most notably with PI3K inhibitor monotherapy in PIK3CA-mutated and ERBB2-amplified breast cancers (9–11). In addition, clinical activity in patients with breast cancer harboring PIK3CA mutations has also been recently reported (12). However, experience with previous targeted therapy paradigms suggests that primary and acquired resistance will be a limiting factor with these agents. Therefore, a clear understanding of the mechanisms underlying PI3K inhibitor sensitivity and/or resistance will be invaluable in determining which patients are most likely to benefit. Moreover, identification of accurate biomarkers in patients who are unlikely to respond to PI3K inhibitor therapy may promote the development of rational drug combinations that will overcome this problem. Recently, a number of clinical and preclinical studies have shown that enhanced ERK signaling, either by activation of compensatory feedback loops or intrinsic KRAS mutations, limits the effectiveness of PI3K pathway inhibitors (13–20). Also, MYC amplification, hyperactivation of the WNT/β-catenin pathway, activation of NOTCH1, and amplification of the translation initiation factor eIF4E all appear able to promote PI3K inhibitor resistance to varying degrees (21–24). Here, using a systematic functional genetic screening approach, we have identified several kinases that mediate resistance to PI3K inhibition, including ribosomal S6 kinases RPS6KA2 (RSK3) and RPS6KA6 (RSK4).
RSK3 and RSK4 are members of the p90RSK family. RSKs are directly regulated by ERK signaling and are implicated in cell growth, survival, motility, and senescence (25–28). Here, we present evidence that overexpression of RSK3 and RSK4 supports cellular proliferation under PI3K pathway blockade by inhibiting apoptosis and regulating cellular translation through phosphorylation of ribosomal proteins S6 and eIF4B. We found RSK3 and RSK4 were overexpressed or activated in a fraction of breast cancer tumors and cell lines, supporting a role for these proteins in breast tumorigenesis. Furthermore, in 2 triple-negative breast cancer patient–derived primary tumor xenografts (PDX), we observed that the PDX with higher levels of phosphorylated RSK was resistant to PI3K inhibition. Importantly, we also demonstrate that by combining inhibitors of PI3K with inhibitors of MEK or RSK, we can reverse the resistance phenotype exhibited by breast cancer cell lines and PDX models with activated RSK and propose that this therapeutic combination may be clinically effective in patients with RSK-activated breast cancers.
To identify kinases whose expression can mediate resistance to PI3K inhibitors, we performed open reading frame (ORF) expression screens in breast cancer cell lines in the presence of BEZ235 (dual PI3K/mTOR inhibitor) (29) or BKM120 (pan-PI3K inhibitor) (Figure (Figure1).1). Both of these compounds are currently in clinical development (30, 31). This ORF library is composed of 597 kinases and kinase-related genes in lentiviral expression vectors containing a blasticidin resistance marker for efficient transduction and stable overexpression in target cells (32–34). We chose to perform a focused screen with kinases, as they represent a set of readily druggable targets, facilitating validation and potentially clinical translation. We screened MCF7 and BT474 cells, as they represent the 2 genotypes of breast cancer cells previously established as exhibiting sensitivity to PI3K inhibition, MCF7 (mutant PIK3CA) and BT474 (ERBB2 amplified) (9–11).
The criteria used to select kinases that support proliferation following PI3K/mTOR blockade in the ORF screen were (a) increased cell numbers in the presence of BEZ235 or BKM120 by at least 3 SD above the mean and (b) corresponding increases in the ratio of cell number in treated versus untreated wells to remove kinases that simply stimulate general proliferation (Figure (Figure1,1, Supplemental Table 1, and Supplemental Figure 1A; supplemental material available online with this article; doi: 10.1172/JCI66343DS1). We performed validation experiments on the ORFs with the strongest phenotypes in the MCF7 screens for resistance against BEZ235 and BKM120 and were able to confirm PI3K inhibitor resistance phenotype for most of these candidates using 2 independent assays for viability (Figure (Figure2,2, A and B, and Supplemental Figure 1B). Unsurprisingly, validated candidates included the receptor tyrosine kinases ERBB2 (HER2) and IGF1R, both of which are known to be upstream of PI3K-dependent signaling and PI3K-independent signaling as well as AKT1 and AKT3, key effectors of the PI3K pathway. Of the remaining candidates, we were particularly interested in RPS6KA2 (RSK3) and RPS6KA6 (RSK4), as these 2 genes provided robust resistance against PI3K inhibition (Figure (Figure2C2C and Supplemental Figure 2A).
Since RSK3- and RSK4-overexpressing cells exhibited a profound decrease in PI3K inhibitor sensitivity, we sought to determine whether other RSK family members exhibited similar properties. In contrast to RSK3 and RSK4, expression of RSK1 and RSK2 only slightly decreased the sensitivity to PI3K inhibition, while the highly related mitogen- and stress-activated protein kinases (MSKs) exhibited no activity, and this was irrespective of expression levels (Supplemental Figure 2B). We therefore chose to focus on RSK3 and RSK4 for subsequent analyses.
To determine whether the resistance phenotypes of RSK-overexpressing cell lines extended to other PI3K pathway inhibitors, we determined the sensitivity of these cells to other inhibitors currently in early stage clinical testing, including GDC-0941, a pan PI3K inhibitor, and MK-2206, an allosteric pan-AKT inhibitor. As expected, treatment with all PI3K pathway inhibitors completely inhibited the proliferation potential of GFP-expressing control cells. However, RSK3 and RSK4 overexpression in MCF7 cells counteracted the growth inhibitory properties of all PI3K pathway inhibitors tested (Figure (Figure2,2, D and E). In contrast, while AKT1-expressing cells were resistant to the PI3K/mTOR-targeted agents, they remained sensitive to treatment with the AKT inhibitor MK2206 (Figure (Figure2,2, D and E).
The RSK family of proteins comprises a group of highly related serine/threonine kinases that regulate cell growth, survival, and cellular proliferation downstream of the RAS/RAF/MEK/ERK pathway. To elucidate the mechanisms behind PI3K inhibitor resistance in RSK-overexpressing cells, we sought to uncover differences in cellular responses to PI3K/mTOR inhibition between control and RSK-overexpressing cells. Previous studies have established that BEZ235 induces apoptosis in cell lines sensitive to PI3K/mTOR inhibition (11, 35). Since both RSK and AKT overexpression lead to decreased sensitivity to PI3K inhibitors, we reasoned that these attenuated responses might be due to the inhibition of apoptosis. As expected, the addition of either BEZ235 or BKM120 substantially enhanced PARP and caspase 7 cleavage, indicative of apoptosis, in GFP-expressing control cells. In contrast, we observed reduced cleaved PARP and cleaved caspase 7 in RSK3/4- Vor AKT1-overexpressing cells upon treatment with BEZ235 or BKM120 (Figure (Figure3A).3A). Furthermore, treatment of control cells with BEZ235 led to increased PARP cleavage in a dose-dependent manner, which was again attenuated in cells expressing RSK or AKT1 (Supplemental Figure 2C). We also observed a marked decrease in the accumulation of cells in sub-G1 in the RSK4-overexpressing cells compared with control cells upon treatment with BEZ235 (Figure (Figure3,3, B and C, and Supplemental Figure 3). Similar findings were observed in RSK-overexpressing cells treated with the pan-PI3K inhibitors BKM120 and GDC0941 (Figure (Figure3C3C and Supplemental Figure 3). Taken together, these data suggest that RSK-overexpressing cells are resistant to PI3K/mTOR inhibition at least in part through decreased induction of apoptosis.
A number of recent reports have demonstrated that the anti-tumor effects of PI3K inhibition may be reduced by the activation of the ERK signaling pathway or by upregulation of protein translation (22, 24, 36–38). Likewise, we investigated the regulation of protein translation in our RSK- or AKT1-overexpressing cells. In control cells, PI3K pathway blockade with the PI3K inhibitor BKM120, the dual PI3K/mTOR inhibitor BEZ235, or the catalytic mTOR inhibitor pp242 (39) markedly reduced eIF4B and rpS6 phosphorylation, 2 key regulators of cap-dependent translation (Figure (Figure4A).4A). In contrast, dephosphorylation of ribosomal protein S6 and eIF4B by PI3K, mTOR, or dual PI3K/mTOR inhibitors was abrogated in the RSK-overexpressing cells (Figure (Figure4A).4A). We extended these analyses to other RSK family members. Although phospho-rpS6 was maintained in RSK1, RSK3, and RSK4 overexpressing cells, phospho-eIF4B was only detectable in RSK3- and RSK4-overexpressing cells following PI3K inhibition (Figure (Figure4B).4B). These results are in line with our proliferation studies suggesting that, while RSK1, RSK3 and RSK4 decrease the sensitivity of cells to PI3K inhibitors, only RSK3- and RSK4-overexpressing cells exhibit a strong resistance phenotype (Supplemental Figure 2B).
Two classes of protein kinases are known to phosphorylate rpS6 directly. The kinases primarily responsible for rpS6 phosphorylation are the mTOR-regulated S6 kinases, which are highly sensitive to PI3K/mTOR inhibition (40). The second class is the RSK family of kinases, which are regulated by ERK signaling and are activated following mitogenic stimulation (41). Based on our observation that retention of rpS6 and eIF4B phosphorylation correlates with resistance to PI3K pathway inhibitors, we hypothesized that cell lines with higher levels of activated ERK and/or RSK signaling may maintain higher levels of phosphorylated S6235/236 upon PI3K blockade and thus be relatively insensitive to PI3K inhibition. To investigate this possibility, we surveyed 27 breast cancer cell lines by immunoblotting and queried Oncomine to identify breast cancer cell lines with high levels of RSK4 (Supplemental Figure 4A) ( www.oncomine.org). Notably, the 2 breast cancer cell lines exhibiting high levels of RSK4 in Oncomine, HCC1143 and HCC38, also demonstrated resistance to the PI3K inhibitor GSK 1059615 (data not shown). As expected, when subjected to treatment with PI3K inhibitors, cell lines with high levels of RSK4 activity exhibited a decrease in sensitivity compared with the sensitive cell line MCF7 (Figure (Figure4C).4C). Furthermore, both AU565 and MDA-MB-231, but not MCF7, retained rpS6 and eIF4B phosphorylation when treated with different PI3K pathway inhibitors (Supplemental Figure 4B). These results suggest that, while rpS6 and eIF4B phosphorylation is principally regulated by the PI3K/AKT/mTOR axis, in the context of RSK overexpression or activation by upstream components, RSKs can sustain rpS6 and eIF4B phosphorylation during PI3K pathway downregulation.
In eukaryotic cells, initiation of protein translation is the major rate-limiting step in protein synthesis (42). Recent studies have suggested that phosphorylation of Ser235/236 in rpS6 and eIF4B Ser422 is required for cap-dependent translation of mRNA (41). To determine the effects of RSK4 overexpression on translation, we monitored new protein synthesis rates in vivo by labeling cells with S35 methionine. Indeed, we observed that RSK4-overexpressing cells had higher levels of total protein synthesis in both normal and PI3K inhibitor–treated conditions compared with control cells (Figure (Figure4,4, D and E). Collectively, our data suggest that RSK overexpression prevents response to PI3K inhibition through maintenance of protein translation and through the inhibition of apoptosis.
The observations described above suggest that activation of the ERK/RSK pathway serves as a mechanism to circumvent PI3K inhibitor sensitivity. Therefore, we hypothesized that the dual blockade of PI3K and RSK pathways would reverse the resistance phenotype and the molecular markers associated with resistance seen in RSK-overexpressing cells. To test this hypothesis, we combined PI3K inhibitors with the MEK inhibitor NVP-MEK162 (MEK-162) (43) or the pan-RSK–specific inhibitor dihydropteridinone (BI-D1870) (44). In MCF7 cells, RSK3 or RSK4 expression decreased response to treatment with any of the PI3K inhibitors alone. However, the combination of PI3K inhibition with MEK162 or BI-D1870 completely reversed the resistance of RSK-expressing cells (Figure (Figure5A5A and Supplemental Figure 5A). BI-D1870 has previously been demonstrated to inhibit the cell-cycle regulators PLK1 and Aurora B, albeit at much higher concentrations than RSK inhibition (45). To verify the specific efficacy of BI-D1870, we treated AKT-overexpressing cells with combined PI3K inhibitors and RSK or MEK inhibitors. As expected, MCF7 cells overexpressing AKT1 were refractory to combined PI3K and MEK/RSK inhibition, confirming the specific efficacy of this combination for cells with activation of the MEK/ERK/RSK pathway (Supplemental Figure 5B). We observed that rpS6 and eIF4B phosphorylation was completely attenuated only when MCF7-RSK cells were treated with the combination of BEZ235 and BI-D1870 or another MEK inhibitor (AZD6244) (46), in agreement with the effects on cell viability (Figure (Figure5B5B and Supplemental Figure 6A). Accordingly, we also observed an inhibition of RSK phosphorylation at Ser380, which serves as a marker of RSK activity, in MCF7-RSK4 cells upon treatment with AZD6244 or MEK162, verifying that MEK inhibition downregulates the function of overexpressed RSK (Supplemental Figure 6A). Furthermore, combined inhibition of PI3K and RSK diminished rpS6 phosphorylation levels and proliferation compared with either inhibitor alone in breast cancer cell lines with high levels of RSK (Figure (Figure5C5C and Supplemental Figure 6B). Since RSK4 overexpression renders cells resistant to the proapoptotic effects of PI3K inhibitors, we hypothesized that combined inhibition of RSK and PI3K would enhance apoptosis compared with either compound alone. Indeed, combined inhibition of PI3K and RSK significantly enhanced apoptosis to levels similar to those in control GFP-overexpressing cells compared with RSK4-overexpressing MCF7 cells and in breast cancer cell lines exhibiting elevated levels of RSK4 (Figure (Figure5D5D and Supplemental Figure 6C). Similarly, targeted knockdown of RSK4 increased the sensitivity to PI3K inhibition in multiple RSK4-overexpressing breast cancer cell lines, substantiating the role of RSK4 in mediating resistance to PI3K inhibition (Figure (Figure5,5, E and F, and Supplemental Figure 6, D and E). Importantly, the degree of apoptosis was virtually identical in RSK4 knockdown cells versus MEK inhibition when combined with a PI3K inhibitor (Supplemental Figure 7, A and B). Furthermore, combined inhibition of PI3K with either BI-D1870 or MEK inhibition inhibited protein translation specifically in RSK-expressing cells and restored inhibition of protein translation upon PI3K inhibition (Supplemental Figure 7, C and D). Collectively, our data suggest that the combination of PI3K and RSK pathway inhibitors is effective at decreasing rpS6 and eIF4B phosphorylation, overall translation, and survival in cells with altered RSK activity.
Next, we sought to analyze the tumorigenic potential of RSK4-overexpressing cells and response to BEZ235 in a xenograft model. To this end, we injected immunodeficient mice with MCF7 cells overexpressing RSK4 or GFP as a control. BEZ235 treatment at 30 mg/kg was started 7 days after injection, when tumors reached an average volume of 250 mm3. RSK4-overexpressing cells exhibited growth rates similar to those of control cells in vehicle-treated mice (Figure (Figure6A).6A). In contrast, and in consonance with previous results in vitro, RSK4 overexpression allowed tumors to progress even in the presence of BEZ235 (Figure (Figure6A).6A). Furthermore, RSK4 expression led to robust retention of rpS6 phosphorylation in tumors in the presence of BEZ235, as measured by phospho-rpS6 staining (Figure (Figure6B).6B). To determine whether the resistance phenotype of RSK-overexpressing tumors extends to other PI3K pathway inhibitors, we further determined the sensitivity of these tumors to BKM120 and MK-2206. As observed in vitro, treatment with all PI3K pathway inhibitors completely blocked the proliferation potential of control tumors. However, RSK4-overexpressing tumors decreased the growth inhibitory properties of all the PI3K inhibitors tested (Supplemental Figure 8A). Since RSK4 expression diminished the effectiveness of single-agent PI3K treatment, we explored the antitumor activity of PI3K inhibition in combination with ERK/RSK pathway inhibitors. We analyzed tumor growth inhibition of MCF7-RSK4–derived xenografts in response to the combination of BEZ235 and the MEK inhibitor MEK162. As the BEZ235 concentration had to be reduced in these experiments from 30 mg/kg to 25 mg/kg to compensate for general toxicity of the combination treatments, the difference in drug response between RSK4- and GFP-expressing animals was less pronounced than in the single-agent experiments. Nevertheless, RSK4-overexpressing cells exhibited a clear trend toward decreased responsiveness to BEZ235 as single-agent therapy compared with the control cells (Figure (Figure6C).6C). When MEK162 was combined with BEZ235, a significant reduction of tumor growth was observed (P = 0.01). This increase in antitumor activity was accompanied by a decrease in phospho-ERK and phospho-S6 staining (Figure (Figure6,6, D and E, and Supplemental Figure 8B). No significant changes were observed in phospho-4EBP1 staining, a direct target of mTOR activity (Supplemental Figure 8, C and D).
Because the intrinsic properties of artificially cultured cell lines tend to diverge from the characteristics of real tumors, we confirmed our results in PDXs. These PDXs generate tumors with the same histopathological characteristics and oncogenic mutations as found in the human patient from whom they were derived (data not shown) (47). Protein lysates of 11 triple-negative PDXs were assessed for pRSK-380 by immunoblotting (data not shown). Of the 11 models, we identified the 2 PDXs that exhibited the greatest difference in levels of activated RSK, PDX60 (high RSK) and PDX156 (low RSK) (Figure (Figure77A).
In concordance with our previous data, the PDX that exhibited hyperactivation of RSK4 remained relatively insensitive to inhibition with the PI3K inhibitor BKM120, while the PDX with low levels of RSK activity were acutely sensitive to PI3K inhibition (Figure (Figure7B).7B). Western blot and reverse phase protein analysis (RPPA) of these PDXs confirmed that following PI3K inhibitor treatment, PDX156 tumors had reduced phospho-S6235/236 levels whereas PDX60 tumors maintained high levels of phospho-S6235/236 (Figure (Figure7C7C and Supplemental Figure 9A). Furthermore, combined inhibition of PI3K and MEK in PDX60 significantly decreased phospho-S6235/236 and overall tumor volume compared with either inhibitor alone (Figure (Figure7,7, D and E). Taken together, our data suggest that hyperactivation of RSK may limit PI3K inhibitor function in breast cancer patients.
To further assess the potential clinical relevance of RSK function in breast cancer, we investigated RSK activity, as assessed by phosphorylation of Thr359/Ser363, across a panel of breast invasive tumors from the TCGA tumor bank ( http://tcga-data.nci.nih.gov/tcga/) for which RPPA data was available (48). We observed elevated levels of phospho-RSK in a subset of basal-like, HER2-enriched, luminal A, and luminal B breast tumors, suggesting RSK is hyperactivated in at least some tumors of these subtypes (Supplemental Figure 9B). Moreover, basal-like tumors as a group had significantly higher levels of phospho-RSK compared with the rest of tumor samples, in agreement with the observation that basal-like breast tumors exhibit evidence of RAS/MEK/ERK pathway activation (Supplemental Figure 9B and refs. 10, 49). We also interrogated the Human Protein Atlas ( http://www.proteinatlas.org) for expression levels of RSK3 and RSK4 based on immunohistochemical (IHC) staining of tumor samples (50). Here, we observed frequent strong staining for RSK4, and to a lesser degree RSK3, across a number of tumor types, including breast, colorectal, prostate, thyroid, urothelial, and lung cancers (RSK3: http://www.proteinatlas.org/ENSG00000071242; RSK4: http://www.proteinatlas.org/ENSG00000072133). Finally, we determined the frequency of amplification or overexpression of RSK3 and RSK4 in a panel of breast cancer cell lines, using the Broad-Novartis Cancer Cell Line Encyclopedia (CCLE) ( http://www.broadinstitute.org/ccle/home) (51). We queried 59 breast cancer cell lines and observed that RSK3 and RSK4 transcripts are upregulated in 8% and 46% of breast cancer cell lines, respectively (Supplemental Figure 9C). Taken together, these observations suggest that RSK3 and RSK4 may be functionally important in breast tumorigenesis.
Inhibitors targeting the PI3K pathway have the potential to be effective anticancer agents and, as such, are being developed at a rapid pace. However, previous experience with targeted therapies predicts that patients who initially respond invariably relapse due to acquisition of drug resistance. To anticipate mechanisms of resistance to PI3K inhibitors, we have screened a library of kinase ORFs and have identified a number of kinases that circumvent PI3K inhibitor sensitivity. Validated candidates included potent activators of PI3K and ERK signaling pathways, such as ERBB2 and IGF1R, as well as downstream effectors AKT1 and AKT3. In addition, we have identified the RSK family members RSK3 and RSK4 as repressors of PI3K inhibitor function. Functional studies have implicated RSKs in the regulation of diverse cellular processes, including transcription, translation, survival, cell-cycle progression, and migration, through phosphorylation of targets including CREB, GSK3, TSC2, rpS6, raptor, eIF4B, BAD, and p27, among others (26, 41, 52–54). The RSKs have all been linked with tumorigenesis, albeit in different contexts. RSK1 and RSK2 have been reported as overexpressed in breast and prostate cancer, while RSK3 has been proposed to be a tumor suppressor in ovarian cancer (55–57). RSK4 has previously been characterized as essential for p53-dependent proliferation arrest as well as stress- and oncogene-induced senescence (28, 58, 59). Interestingly, the RSK4 isoform exhibits constitutively high activity, is upregulated in MMTV-Myc mouse breast tumors, is aberrantly expressed in breast cancer, and has been implicated in sunitinib resistance (58, 60–63). Here, we demonstrate that RSK3 and RSK4 can also mediate resistance to PI3K inhibitors in breast cancer cells both in vitro and in vivo.
Our observations strongly support a role for retention of rpS6 and eIF4B phosphorylation in the resistance phenotype of RSK-overexpressing cells, in agreement with a previous report noting retention of rpS6 phosphorylation in breast cancer cell lines exhibiting intrinsic resistance to PI3K inhibition (9). Previous studies have suggested that RSKs directly phosphorylate rpS6 at Ser235/236 and eIF4B at Ser422. The former promotes binding of rpS6 to the 7-methylguanosine cap complex and enables cap-dependent translation to proceed, while the latter is critical for eIF4B binding to the cap complex and enhanced helicase activity of eIF4A and increased cellular translation (64–66). In agreement with these results, we observed that RSK4-overexpressing cells exhibited elevated levels of overall translation, which are maintained in the presence of PI3K inhibitors (Figure (Figure7F).7F). These results are also consistent with a previous report implicating upregulation of cap-dependent translation by eIF4E amplification in promoting resistance to BEZ235 (67).
As RSKs are directly regulated by RAF/MEK/ERK signaling, we hypothesized that inhibition of this pathway would overcome the resistance phenotype of RSK-overexpressing cells and reverse all associated cellular phenotypes. We observed that addition of MEK or RSK inhibitors restored responsiveness of RSK-expressing cells to PI3K inhibitors by all parameters analyzed, including translation, S6 phosphorylation, cell viability, and in vivo tumor formation (Figures (Figures55–7 and Supplemental Figures 5–8). Importantly, this reversal of phenotype was specific for RSKs, as AKT1-overexpressing cells remained refractory to PI3K inhibition even with the addition of MEK or RSK inhibitors.
One potential limitation of this study is the fact that we were unable to examine RSK inhibition, either through chemical inhibition or knockdown of RSK4, in relevant xenograft models. This is primarily due to the technical difficulty of the experiments and the lack of suitable chemical reagents currently available. Significantly, however, in both in vitro and in vivo experiments, MEK inhibitors inhibited RSK phosphorylation (Figure (Figure7E,7E, Supplemental Figure 6A, and data not shown), indicating that the MEK inhibitors used in our animal models effectively inhibited RSK activity. Collectively, our data suggest that RSK overexpression renders tumors insensitive to PI3K inhibition, which can be overcome by inhibiting the MEK/ERK/RSK pathway.
The observations presented here support the notion that breast cancer cells upregulate overall protein translation and cell proliferation through overlapping but parallel pathways, the PI3K/mTOR and ERK/RSK pathways (Figure (Figure7F).7F). Interestingly, another significant outlier in our screen, the protooncogene PIM2, regulates key effectors of cap-dependent translation, including eIF4E, 4EBP1, and S6K, independently of the PI3K/mTOR pathway, supporting the notion that combined pharmacological inhibition of multiple translational regulators should be explored (Figure (Figure7F)7F) (68).
A number of reports have recently shown that an elevated ERK activation signal, either through intrinsic KRAS mutations or through the activation of compensatory feedback loops observed following PI3K inhibition, limits the effectiveness of PI3K inhibitors in the clinic (13–20). Early clinical trials assessing the effectiveness of PI3K and MEK inhibitors have demonstrated some evidence of efficacy in certain tumor types (69). However, initial reports seem to suggest that the use of MEK inhibitors in the clinic results in undesired toxicities, limiting the effectiveness of this compound (70). Importantly, our studies suggest that targeted RSK inhibition is as effective as MEK inhibition when used in combination with PI3K inhibitors, resulting in similar degrees of decreased proliferation and augmented apoptosis. As RSK-specific inhibitors target only a single effector arm of MAPK signaling, they may provide a therapeutic window circumventing many of the potential toxicities associated with current MEK-PI3K inhibitor combination strategies. Moreover, we anticipate that use of this combination will also be indicated in the treatment of tumors that exhibit evidence of MEK/ERK–driven signaling.
Kinase library ORFs and GFP controls were expressed from pLX-Blast-V5 lentiviral expression vectors, which confer blasticidin resistance, as previously described (34). Virus was produced by transfecting 239T cells in 96-well plates, and screening infections were performed in 384-well plates in octuplicate, using standard spin-infection protocols with 1 ORF per well, as previously described (Figure (Figure1A1A and ref. 34). Medium was changed 24 hours after infection to 10 μg/ml blasticidin, 200 nM BEZ235, 1 μM BKM120, or no drug, with 2 replicates per condition (Figure (Figure1A).1A). Five days after medium change, cell viability was assessed with CellTiter-Glo (Promega). Duplicates were averaged for all subsequent analysis. Infection efficiency was monitored by comparing plates selected with blasticidin with untreated plates, and those wells with greater than 2-fold difference in cell number between the 2 conditions were eliminated from the analysis. By this criterion, approximately 95% of the ORF library was efficiently transduced into the target cells and thus tested for phenotype (Supplemental Figure 1A).
MCF7 and MDA-MB-231 cells were maintained in DMEM (Gibco; Invitrogen) supplemented with 10% FBS at 37°C in 5% CO2. BT474 and AU565 cells were maintained in RPMI medium supplemented with 10% FBS at 37°C in 5% CO2. All cells were obtained from ATCC. Stable cell lines were maintained in appropriate medium supplemented with 10 μg/ml blasticidin.
MCF7 cells infected as indicated were seeded in 12-well plates (2 × 104). After 24 hours, cells were treated with BEZ235 (100 or 200 nM), BKM120 (0.75 or 1 μM), GDC-0941 (1 μM), or MK2206 (2 μM) alone or in combination with MEK162 (1 μM), BI-D1870 (10 μM), or AZD6244 (1 μM), as indicated in text. Cell numbers were quantified by fixing cells with 4% glutaraldehyde or methanol, washing the cells twice in H2O, and staining the cells with 0.1% crystal violet (Sigma-Aldrich). The dye was subsequently extracted with 10% acetic acid, and its absorbance was determined (570 nm). Growth curves were performed in triplicate. Viability assays with CellTiter-Glo (Promega) were performed by plating 2,000 cells in 96-well plates, adding the drug at 24 hours, and assaying 4 to 5 days after drug addition. Cell-cycle and hypodiploid apoptotic cells were quantified by flow cytometry as described (71). Briefly, cells were washed with PBS, fixed in cold 70% ethanol, and then stained with propidium iodide while being treated with RNase (Sigma-Aldrich). Quantitative analysis of sub-G1 cells was carried out in a FACScalibur cytometer using Cell Quest software (BD Biosciences).
Cells were lysed in solubilizing buffer (50 mM Tris, pH 8.0, 150 mM NaCl, 1 % NP-40, 0.5% deoxycholic acid, 0.1% SDS, 1 mM sodium vanadate, 1 mM pyrophosphate, 50 mM sodium fluoride, 100 mM β-glycerol phosphate) supplemented with protease inhibitors (Roche). Whole-cell extracts were then separated on SDS-PAGE gels and transferred to polyvinylidene difluoride membranes (Millipore). Membranes were blocked with bovine serum albumin and probed with specific antibodies. Blots were then incubated with an HRP-linked second antibody and resolved with chemiluminescence (Pierce). Western blots were quantified using ImageJ ( http://rsb.info.nih.gov/ij/).
Antibodies against PARP, cleaved PARP, cleaved caspase 7, phospho-AKT, AKT, phospho-ERK, ERK, phospho-S6235/236, phospho-S6240/244, phospho-eIF4B-422, phospho-GSK3, phospho-p70 S6K-389, phospho-4EBP1-37/46, and phospho-RSK-380 were from Cell Signaling Technology. Antibodies against cyclin D1, GapdH, and tubulin were from Santa Cruz Biotechnology Inc. Antibodies against total V5 were from Invitrogen. BEZ235, BKM120, and MEK162 were provided by Emmanuelle Di Tomaso and Michel Maira (Novartis Institutes for Biomedical Research). GDC-0941, MK-2206, and AZD6244 were purchased from Selleck. BI-D1870 was purchased from Axon Medcam. Cycloheximide was purchased from Sigma-Aldrich. siRNA-targeting RSK4 was purchased from Dharmacon and transfected according to the manufacturer’s protocols.
MCF7 cells were grown to 70% confluence in 10-cm plates and either incubated overnight in 10% serum or exposed to BEZ235 (200 nM), BKM120 (750 nM), GDC0941 (1 μM), or cycloheximide in 10% serum. Cells were then washed once with DMEM lacking cysteine and methionine. DMEM lacking cysteine and methionine but including dialyzed serum and kinase inhibitors as indicated was added. Cells were incubated for 1 hour, 250 pCi of Expre35S35S (NEN) was added to each well, and the cells were labeled for a further 30 minutes. Cells were washed once with ice-cold PBS, and whole-cell extracts were isolated as described above and separated by SDS-PAGE. The 35S-labeled proteins were visualized by autoradiography with film. The amount of 35S incorporated into protein was measured using a Beckman LS6500 Scintillation Counter.
Six-week-old female athymic nude Foxn1nu mice were purchased from Harlan Laboratories. Mice were housed in air-filtered laminar flow cabinets with a 12-hour light/12-hour dark cycle and given food and water ad libitum. Mice were handled with aseptic procedures and allowed to acclimatize to local conditions for 1 week before the experimental manipulations. A 17β-estradiol pellet (Innovative Research of America) was implanted subcutaneously into each mouse 1 day before cell injection. 107 MCF-GFP or MCF7-RSK4 cells were resuspended in PBS/Matrigel (BD Biosciences) (1:1) and injected subcutaneously into the right flank of each mouse in 200 μl of final volume. Treatments began when tumors reached an average size of 250 mm3 and were thus considered as established growing xenografts. Mice were treated once daily with placebo, BEZ235, BKM120, MK-2206, or MEK162 by oral gavage. BEZ235 (25–30 mg/kg, 6IW [6 days on 1 day off]) and BKM120 (30 mg/kg, 6IW) were dissolved in 10% NMP-90% PEG, freshly formulated, and administrated within 30 minutes. MK-2206 (100 mg/kg, 3IW) was formulated in 30% Captisol and MEK162 (6 mg/kg, BID) in 0.5% Tween-80, 1% carboxymethyl cellulose. For tumor growth studies, mice were treated for 7–24 days, depending on the xenograft model and treatment regime. Tumor xenografts were measured with calipers 3 times a week, and tumor volume was determined using the following formula: (length × width2) × (π/6). At the end of the experiment, the animals were anesthetized with 1.5% isofluorane-air mixture and killed by cervical dislocation. Tumors were removed 2 hours following the last administration.
Tumor xenografts or human breast cancer tumors were fixed immediately after removal in a 10% buffered formalin solution for a maximum of 24 hours at room temperature before being dehydrated and paraffin embedded under vacuum conditions. Tissue microarrays (TMA) were constructed, including triplicated cores from each xenograft. TMA slides underwent deparaffinization and antigen retrieval using the PTLink system (DAKO) following the manufacturer’s instructions. Primary antibodies were phospho-rpS6-Ser235/236, phospho-ERK-Thr202/Tyr204, phopsho-4EBP1 37/46 (all from Cell Signaling Technology), or RSK4 (Millipore). Samples were incubated with a 1:40 solution of streptavidin/peroxidase for 30 minutes. Staining was developed with freshly prepared 0.05% 30,3-diaminobenzidine tetrahydrochloride, which was then counterstained with hematoxylin. No labeling was observed in control experiments when primary antibodies were omitted or, alternatively, when normal nonimmune serum was used. There was no evidence of cross-reactivity with the antibodies used in this study. For IHC evaluation, H-scores were used to quantify the expression of the phosphoproteins.
Following implantation and engraftment of a metastatic patient needle biopsy (3 × 3 mm) in 6-week-old female HsdCpb:NMRI-Foxn1nu mice (Harlan Laboratories), a PDX colony was established and checked for expression of immunohistopathological markers in agreement with the original biopsy. Animals were supplemented with 1 μM estradiol (Sigma-Aldrich) in the drinking water. Protein lysates of 11 triple-negative PDXs were assessed for pRSK-380 by Western immunoblot (not shown). Of the 11 models, the PDXs expressing the highest (PDX60) or lowest (PDX156) levels of pRSK-380 were selected for in vivo testing of BKM120 antitumor response. Briefly, following an expansion phase by sequential implantation, tumors were implanted into nude mice for experimentation. Once tumors reached a mean size of 250 mm3, mice were distributed in groups of similar mean and SEM and treated by oral gavage with BKM120 6xQD (dissolved in NMP-PEG) or AZD6244 (dissolved in Methylcellulose/Tween). Tumor grafts were measured with calipers, and tumor volumes were determined using the following formula: (length × width2) × (π//6). At the end of the experiment, animals were sacrificed by CO2 inhalation. Tumor volumes are plotted as relative to day 1 and expressed as mean ± SEM of the group.
All statistics were calculated using GraphPad Prism or Microsoft Excel. The tests used include 2-tailed t test, SD, SEM, 1-way ANOVA, and log-rank test where appropriately indicated in figure legends. P ≤ 0.05 was considered statistically significant. All P values are depicted in the figures or in the figure legends.
Patient consent for tumor use in animals was obtained under a protocol approved by the Vall d’Hebron Hospital Clinical Investigation Committee and Animal Use Committee. Mice were maintained and treated in accordance with protocols approved by the Vall d’Hebron University Hospital Care and Use Committee.
The authors thank Anna Schinzel of the DFCI RNAi facility for assistance with reagents and Sadhna Vora and Jeff Engelman for helpful discussion. The authors also thank Emmanuelle Di Tomaso and Michel Maira (Novartis Institutes for Biomedical Research, Cambridge, Massachusetts, USA, and Basel, Switzerland) for the supply of BEZ235 and BKM120. S.Y. Kim thanks Joseph Nevins for advice and generous support. This research was supported by the Breast Cancer Research Foundation (to J. Baselga), the European Research Council (AdG09250244 to J. Baselga), the Instituto de Salud Carlos III (Intrasalud PSO9/00623 to J. Baselga), a noncommercial research agreement with Novartis (to J. Baselga), FERO foundation supporting research grants (to J. Baselga), the DFCI-Novartis Drug Discovery Program (to W.C. Hahn), R01 CA130988 (to W.C. Hahn), RC2 CA148268 (to W.C. Hahn), U54CA112962 (to W.C. Hahn), and the Snyder Medical Research Foundation (to W.C. Hahn).
Conflict of interest: José Baselga and William C. Hahn consult for Novartis Pharmaceuticals.
Note regarding evaluation of this manuscript: Manuscripts authored by scientists associated with Duke University, The University of North Carolina at Chapel Hill, Duke-NUS, and the Sanford-Burnham Medical Research Institute are handled not by members of the editorial board but rather by the science editors, who consult with selected external editors and reviewers.
Citation for this article: J Clin Invest. 2013;123(6):2551–2563. doi:10.1172/JCI66343.