Combined MEK and CDK4/6 inhibition (MEKi + CDK4i) has shown promising clinical outcomes in patients with NRAS-mutant melanoma. Here, we interrogated longitudinal biopsies from a patient who initially responded to MEKi + CDK4i therapy but subsequently developed resistance. Whole-exome sequencing and functional validation identified an acquired PIK3CAE545K mutation as conferring drug resistance. We demonstrate that PIK3CAE545K preexisted in a rare subpopulation that was missed by both clinical and research testing, but was revealed upon multiregion sampling due to PIK3CAE545K being nonuniformly distributed. This resistant population rapidly expanded after the initiation of MEKi + CDK4i therapy and persisted in all successive samples even after immune checkpoint therapy and distant metastasis. Functional studies identified activated S6K1 as both a key marker and specific therapeutic vulnerability downstream of PIK3CAE545K-induced resistance. These results demonstrate that difficult-to-detect preexisting resistance mutations may exist more often than previously appreciated and also posit S6K1 as a common downstream therapeutic nexus for the MAPK, CDK4/6, and PI3K pathways.
Significance: We report the first characterization of clinical acquired resistance to MEKi + CDK4i, identifying a rare preexisting PIK3CAE545K subpopulation that expands upon therapy and exhibits drug resistance. We suggest that single-region pretreatment biopsy is insufficient to detect rare, spatially segregated drug-resistant subclones. Inhibition of S6K1 is able to resensitize PIK3CAE545K-expressing NRAS-mutant melanoma cells to MEKi + CDK4i. Cancer Discov; 8(5); 556–67. ©2018 AACR.
See related commentary by Sullivan, p. 532.
See related article by Teh et al., p. 568.
This article is highlighted in the In This Issue feature, p. 517
NRAS mutations are able to induce constitutive RAF–MEK–ERK signaling cascade activation, promoting survival, proliferation, and tumor progression. Activating NRAS mutations occur in 15% to 20% of melanomas (1, 2), and effective therapeutic approaches to this genetic subtype of tumors are still needed. MEK inhibitors (MEKi) have emerged as a likely cornerstone therapy for NRAS melanoma, as it inhibits the main signal transduction pathway: The compound MEK162 has shown activity in a phase III clinical trial, inducing a 15% overall response rate (3). However, a major obstacle for single-agent MEKi is the ability of NRAS signaling to induce multiple pro-oncogenic signaling networks (4–6), making it a daunting task to select appropriately complementary partner therapies for combination. In a previous study using genetically engineered mouse models, systems biology analyses pinpointed CDK4 as a critical node that, when targeted in combination with MEKi, approximates the full shutdown of NRAS signaling (6). This manifests as the synergistic induction of apoptosis and cell-cycle arrest, leading to frank regression in multiple models of NRAS as well as BRAF melanoma (6–8). A phase Ib/II clinical trial (NCT01781572) using MEK162 and LEE011 against NRAS melanoma was designed based on these findings, and interim results show promising response rates and progression-free survival (9). The combination is also under active investigation in KRAS-mutant lung (NCT03170206), pancreatic, and colon cancers (NCT02703571). Anticipating mechanisms of resistance is pivotal for the clinical evolution of this therapeutic regimen.
In this study, we analyze for the first time a case of clinical acquired resistance to MEKi + CDK4i combined therapy in a patient enrolled in clinical trial NCT01781572. The patient showed initial response to the therapy but eventually developed resistance. Our longitudinal biospecimen investigation revealed that a PIK3CAE545K mutation preexisted in a rare subpopulation whose expansion upon therapy was responsible for the observed drug resistance. Through functional studies we demonstrate that the key resistance mechanism conferred by PIK3CAE545K is sustained cell proliferation via the pharmacologically targetable S6K1–S6 signaling axis. Our study provides the first evidence that subclonal populations of cells harboring resistance mechanisms can evade detection in single pretreatment biopsies due to a combination of their rarity and nonuniform distribution across the tumor.
A 59-year-old female patient presented with recurrent, unresectable stage IIIC malignant melanoma with in-transit metastases in the left popliteal fossa and groin lymphadenopathy (Supplementary Fig. S1A–S1B). She had a history of a nodular pigmented melanoma on the left calf which was resected 7 years prior. Clinical targeted next-generation sequencing (NGS) of a 46-gene panel of a popliteal fossa lesion identified an NRAS mutation (p.G13R, NRASG13R) and wild-type PIK3CA (Supplementary Table S1) with an average read depth of ≥3,500× per gene. The patient underwent 2 cycles of high-dose IL2 (HDIL2), which resulted in progression of disease. She was then enrolled to clinical trial NCT01781572 of MEK162 (MEKi) dosed at 45 mg by mouth twice daily continuously and LEE011 (CDK4i) 200 mg daily for 21 days of a 28-day cycle (Fig. 1A). The patient had a confirmed partial response (−39%) to therapy per RECIST 1.1 as well as a decrease in serum lactate dehydrogenase (LDH; Fig. 1B and C), but was eventually taken off protocol because of progression of the lesions (Fig. 1C). She was then treated with successive single-agent anti-CTLA4 and single-agent anti–PD-1 therapies with a lack of response to anti-CTLA4 and an initial mixed response to anti–PD-1 followed by progression (Fig. 1C and Supplementary Fig. S1), including the development of distant metastases to the bowel and brain. Five longitudinal biopsies were collected from the patient for whole-exome sequencing (WES): pretreatment, on-MEKi + CDK4i (day 16), postprogression (day 151), and on-anti–PD-1 (one from a responding lesion and one from a nonresponding lesion; Fig. 1A and Supplementary Fig S1C). All biopsies were of the in-transit metastases in the left popliteal fossa other than the anti–PD-1 responding lesion biopsy which was of a groin lymph node.
PIK3CAE545K Preexisted in a Rare Subpopulation and Expanded Early in MEKi + CDK4i Treatment
WES was performed on the 5 biopsies and matched normal tissue, and the Mutect program was used to identify somatic mutations among the samples. Single-nucleotide variant (SNV) analysis revealed that all 5 biopsies were clonally related, sharing a minimum of 63 trunk mutations, including the driver oncogenic mutation NRASG13R (Fig. 1D and Supplementary Table S2). We noted that one of the samples, the on-anti–PD-1 responding lymph node lesion, was an outlier with a much higher mutational load (Fig. 1D). This observation is in line with highly mutated tumors being more responsive to checkpoint blockade (10).
In order to identify candidate mutations involved in MEKi + CDK4i resistance, we focused on the comparison between pretreatment and post-resistance samples, aided by their high tumor purities (Supplementary Table S3). Between these two samples, 303 mutations were shared whereas 53 were private to the post-resistance sample, of which 16 were nonsynonymous coding mutations (Fig. 1D and E). This indicates a strong evolutionary continuity between the two samples and suggests that mutations private to the post-resistance sample may include causative resistance mutation(s). We hypothesized that cells with the causative mutation(s) would be a dominant clone; therefore, we prioritized mutations that were present with an allelic fraction greater than 25% and not detected before the start of treatment (Supplementary Table S3). Only one mutation of this set has been reported in the Catalogue of Somatic Mutations in Cancer (COSMIC) database and was known from the literature to be an oncogenic mutation: PI3KCAE545K (11), at an allelic fraction of 33%. PI3KCAE545K (c.1633G>A, p.E545K) is a hotspot oncogenic mutation, known to be involved in breast, GI tract, bladder, and other cancers’ development, but has rarely been reported in melanoma (∼1%; refs. 4, 12, 13). In the current patient, PIK3CAE545K was detected as early as 16 days after the initiation of MEKi + CDK4i treatment (Supplementary Table S3). Moreover, PIK3CAE545K persisted in all of the biopsies including the one from the lymph node (Supplementary Table S3 and Supplementary Fig S2). Because liquid biopsies offer the capacity to nonintrusively test for acquired resistance–causing mutations (14–17), we isolated circulating cell-free DNA and tested it by digital-droplet PCR (ddPCR). Indeed, the PIK3CAE545K mutation was readily detected in the plasma after the acquisition of resistance (Supplementary Fig S3).
The detection of PIK3CAE545K at only 16 days on therapy is consistent with the hypothesis that drug-resistant clones may exist as rare subpopulations prior to therapy which are then positively selected for (18). To our knowledge, however, although this hypothesis has been validated in vitro through an elegant barcoding strategy that identified preexisting EGFR or ABL1 inhibitor–resistant subclones that became dominant upon therapy (19), it has not been explicitly quantified in longitudinal solid-tumor biopsies (20, 21). Therefore, we performed blocker displacement amplification (BDA) to enrich for the detection of rare variants (22) on 7 different regions of the pretreatment sample (Fig. 1F). BDA is capable of detecting rare variants to a limit of ∼0.1%. Indeed, we could confidently detect the presence of PIK3CAE545K in 3 of 7 pretreatment regions as rare variants (Fig. 1G), compared with 4 of 4 post-resistance regions comprised of 1 of 1 region from early on anti–PD-1 therapy and 3 of 3 regions from a bowel metastasis biopsy at 103 days on anti–PD-1 therapy (272 days post-MEKi + CDKi therapy; Fig. 1A and 1G). By weighing the BDA data of each region for the relative tumor area they covered, we estimated that PIK3CAE545K existed as approximately 0.13% of the entire pretreatment tumor and 10.2% of the entire post-resistance bowel metastasis tumor. Moreover, the PIK3CAE545K-positive and -negative pretreatment regions were spatially segregated rather than uniformly distributed (Fig. 1F and G), offering an explanation of why single-region sampling for the WES data at 200× read coverage missed detecting this subpopulation. Indeed, both the clinical and research NGS samples came from the same area as the blocks that tested PIK3CAE545K-negative (Fig. 1F). In sum, our patient sample analyses indicate that a rare PI3KCAE545K subpopulation rapidly expanded upon MEK + CDK4i therapy, possibly due to positive selection for its putative drug resistance properties (Fig. 1H).
PIK3CAE545K Confers Resistance to NRAS-Mutant Melanoma Cells and Increases S6K1 and S6 Phosphorylation
In order to validate and explore the mechanisms and therapeutic vulnerabilities of the PIK3CAE545K mutation in MEKi + CDK4i resistance, we overexpressed GFP, wild-type PIK3CA, or PIK3CAE545K in two NRAS-mutant cutaneous melanoma cell lines, WM1366 and SB2. The cells were then assayed for sensitivity to MEKi + CDK4i treatment (Fig. 2A–F). Consistent with previous studies, MEKi + CDK4i showed a stronger inhibitory effect in the control cell lines than either drug alone (Fig. 2A and D). By contrast, the cells that expressed PIK3CAE545K were significantly more resistant to the MEKi + CDK4i combination (Fig. 2A and D), as well as to single-agent MEK162 or LEE011 (Fig. 2B and C; 2E and F).
In order to better understand the molecular mechanisms underlying the observed resistance, reverse phase protein array (RPPA) analysis of >300 proteins was performed on the cell lines administered with vehicle, MEKi, CDK4i, or their combination, and Western blot validation of selected proteins. First, to validate our model systems, we examined expected on-target changes of the various perturbations. We found that the expression of PIK3CAE545K increased AKT phosphorylation in both SB2 and WM1366 cells; this was not affected by MEKi + CDK4i, as expected (Fig. 2G and Supplementary Table S4). Next, we noted that pERK and DUSP4 were decreased by MEK162, both markers of MAPK pathway inhibition. Finally, we observed decreases in the cell-cycle markers pRB, PLK1, and CCNB1 by either drug alone, and synergistically by MEKi + CDKi in both cell lines (Fig. 2G and Supplementary Table S4), consistent with our previous study in colon cancer (23).
Next, we asked which proteins are responsible for MEKi + CDK4i efficacy in control cells that are then subverted by PIK3CAE545K signaling. We reasoned this could done by first identifying proteins altered by the combination therapy, then intersecting them with proteins consistently altered by PIK3CAE545K in both cell lines (Fig. 2G and Supplementary Table S4). The goal of this approach was to identify potentially druggable target(s) at the intersection of all three oncogenic cascades (MAPK, CDK4/6, and PI3K). This approach identified a single protein state: S6 phosphorylated at the S235/S236 and S240/S244 residues (Fig. 2G). The perturbation of this mTOR effector was validated through immunoblotting (Fig. 2H): MEKi + CDK4i administration was able to efficiently impair S6 phosphorylation in control cells, but not in PIK3CAE545K cells which showed a consistent residual S6 activation in both SB2 and WM1366 (Fig. 2G and H). This increase by PIK3CAE545K was accompanied by an enhanced phosphorylation of the direct S6 upstream kinase S6K1 (Fig. 2H), which was not affected by drug treatments. No change in phosphorylation level of the mTOR effector 4EBP1 was detected. Mechanistically, we highlight that MEK + CDK4i was able to abolish RB phosphorylation in all cell lines, meaning that PIK3CAE545K did not rescue loss of pRB (Fig. 2H and Supplementary Fig. S4).
Specific S6K1 Inhibition Resensitizes PIK3CAE545K Cells to MEKi + CDK4i
S6 is a ribosomal protein, part of the pre-40S subunit. S6 phosphorylation by S6K1 is involved in a plethora of biological processes, including cell proliferation, translation initiation, and metabolic regulation, and is known to be activated downstream of the PI3K–mTOR axis (24–26); very little has been studied regarding the specific involvement of S6K1 in drug resistance. As PIK3CAE545K induced an increase in S6K1 activity and consequent S6 phosphorylation, we investigated the effect of pharmacologically inhibiting S6K1 signaling. We first tested inhibition of its upstream regulator mTOR. Consistent with our observations, the combination of MEKi + CDKi and an mTORi (rapamycin) significantly reduced PIK3CAE545K cell growth, compared with the single and even double combination treatments (MEKi + CDK4i, MEKi + mTORi, and CDK4i + mTORi), which were able to achieve only a partial inhibition of cell growth (Fig. 3A–D). Single-agent mTORi did not show any effect (Fig. 3C and D; Supplementary Fig. S5A). As S6K1/S6 activity but not p4EBP1 was shown to be increased by PIK3CAE545K, we next asked whether specific inhibition of S6K1 was able to phenocopy the mTORi effect. PF-4708671 is a highly specific S6K1 inhibitor (S6K1i) with a 400-fold selectivity over S6K2 and no activity against 73/75 other kinases; the two showing minimal activity, MSK1 and RSK1, did not show modulation of substrate phosphorylation in cellular assays up to 10 μmol/L (27). We found that PF-4708671 was sufficient to resensitize PIK3CAE545K cells to combined MEKi + CDK4i, at a GI50 in the double-digit nanomolar range for all three drugs (Fig. 3A–D). PF-4708671 administration as single agent did not elicit any significant effect on any of the tested cell lines (Supplementary Fig. S5B). We also noted that neither mTORi nor S6K1i elicited an additional effect to MEKi + CDK4i in control GFP or PIK3CAWT cells (Fig. S5C–S5F). Moreover, MEKi + CDK4i + S6K1i displayed a GI50 >3 μmol/L against nontumorigenic hTERT-immortalized human melanocytes (Supplementary Fig. S6). Together, these results demonstrate that the increased efficacy of the triple combination is specific for PIK3CAE545K-expressing cells and is not generically toxic.
Molecular analyses confirmed that mTORi administration was effective and on-target, as both pS6K and pS6 levels were abrogated, as well as p4EBP1; by contrast, S6K1i increased pS6K1, as expected due to feedback loops (27, 28) and achieved extinction of pS6 but not p4EBP1 in combination with MEKi + CDK4i (Fig. 3E). Interestingly, neither mTORi nor S6K1i affected pRB levels with MEKi + CDK4i (Fig. 3E and Supplementary Fig. S7). Mechanistically, PIK3CAE545K expression suppressed MEKi + CDK4i-induced cell-cycle arrest (Fig. 4A and B), but did not prevent the induction of apoptosis (Fig. 4C and D). The addition of S6Ki to MEKi + CDK4i enhanced cell-cycle arrest and decreased cell proliferation by propidium iodide and CFSE analysis, respectively, in PIK3CAE545K but not PIK3CAWT-expressing cells (Fig. 4A and B; Supplementary Figs. S8 and S9A–S9B; Supplementary Table S5); by contrast, the addition of S6Ki did not increase apoptosis by caspase 3/7 activation, Annexin V, or propidium iodide assays in any line (Fig. 4C and D; Supplementary Figs. S9C–S9D and S10; Supplementary Table S5). These data indicate that PIK3CAE545K induces resistance primarily through abrogating the cell-cycle arrest triggered by MEKi + CDK4i, and that adding specific S6K inhibition is sufficient to reinstate the arrest to levels similar to those of the sensitive parental cells (Fig. 4A and B). However, we note that neither pS6 inhibition alone by mTORi/S6Ki nor pRB inhibition alone by MEKi + CDK4i was sufficient to reinduce cell-cycle arrest, but only when both targets were inhibited by the triple combination (Figs. 3E and 4A and B; Supplementary Figs. S7 and S11).
S6K1 Inhibition Reverts PIK3CAE545K-Induced Resistance to MEKi + CDK4i In Vivo
To determine the in vivo translatability of our findings, we next conducted drug-dosing experiments in the SB2 xenograft model. SB2-GFP tumors treated with MEKi + CDK4i showed rapid tumor regression leading to multiple complete responses (red line), whereas SB2-PIK3CAE545K tumors did not regress and eventually resumed growth (Fig. 5A, purple vs. red lines), thus confirming the ability of the PIK3CAE545K mutation to confer resistance to MEKi + CDK4i in vivo. Consistent with our in vitro findings, the addition of S6K1i to the MEKi + CDK4i combination reverted this drug resistance, leading to frank regression and multiple complete responses (Fig. 5A, black line). Importantly, none of the drug combinations caused significant weight loss, indicating a lack of generalized toxicity (Fig. 5B).
To determine the antioncogenic mechanism in vivo, we conducted phospho-Histone H3 (pH3) immunohistochemistry and TUNEL assays on tumor sections to assess the impact on cell cycle and apoptosis, respectively. The results confirmed our in vitro observations that PIK3CAE545K subverts MEKi + CDKi-induced cell-cycle arrest but not apoptosis and that the addition of S6K1i restores sensitivity primarily by reinducing cell-cycle arrest rather than by inducing apoptosis (Fig. 5C and D; Supplementary Figs. S12–S13).
Finally, to confirm clinical relevance, we determined that 5 of 5 post-resistance patient biopsies at two different time points showed increased levels of pS6 by immunohistochemistry compared with pretreatment samples (Fig. 6A and B). Consistent with our findings, pretreatment samples from E545K+ regions showed intermediate levels of pS6, whereas E545K− biopsies were virtually negative (Fig. 6B) Together, these results suggest that S6 phosphorylation downstream of mTOR and S6K1 could lie at a critical nexus of multiple oncogenic signaling cascades, and that S6K1i could be broadly indicated for resistance mechanisms affecting the MAPK, CDK4, and/or PI3K pathways (Fig. 6C).
In this work, we present the first study of clinical resistance to combined MEK and CDK4/6 inhibition. We provide evidence that PIK3CAE545K preexisted in a rare, spatially segregated subpopulation which expanded upon treatment, conferred resistance to the drug combination, and presented a therapeutically vulnerable node at the S6K1–S6 signaling axis. These discoveries were made possible through the longitudinal biopsy of patient tissues and blood, including multiregion tumor sampling allowing for a comprehensive and quantitative assessment of the pretreatment state. We found that PIK3CAE545K evaded detection in both the clinical and research NGS assays by virtue of both a low prevalence (0.15% of pretreatment tumor cells) and a nonuniform distribution across the tumor, leading to 4 of 7 pretreatment regions being negative for PIK3CAE545K, even with a ×1,000-fold mutant allele enrichment by BDA. To our knowledge, this is the first work in a solid tumor to quantitatively track a rare preexisting resistant subclone through an extended longitudinal analysis, and confirms that such subclones can rapidly expand to become the dominant resistant clone. Specifically, after the acquisition of drug resistance, we estimated the PIK3CAE545K population expanded from a 0.13% to a 10% to 33% variant allele frequency, as early as 16 days of MEKi + CD4Ki treatment. We note that this detection level is an order of magnitude more sensitive than previous studies of preexisting resistance mutations in solid cancers (29–31). All together, these findings strongly suggest that single-region assays of pretreatment samples likely often miss the detection of rare resistant subpopulations, particularly when their prevalence is very close to the reliable detection limit of even high-coverage assays. Therefore, our data suggest that multiregion sampling within a tumor—which is not routinely performed in pathology or research laboratories—has the potential to reveal the intratumoral heterogeneity of candidate preexisting resistant subclones (32).
Mechanistically, unbiased RPPA analysis of our cell lines revealed S6 but not 4EBP1 phosphorylation as one of the most strongly downregulated targets downstream of MEKi + CDK4i, being modulated by both drugs singly and synergistically by the combination, consistent with our previous study in colon cancer (23). We show here that PI3KCAE545K subverts this S6K1–S6 signaling axis, leading to revocation of the MEKi + CDK4i-mediated cell-cycle arrest—though not the MEKi-mediated apoptosis—ultimately causing drug resistance via continued cell proliferation. Interestingly, PI3KCAE545K did not restore RB phosphorylation levels abolished by MEKi + CDK4i despite the resumption of proliferation, which was surprising given the central role of pRB in governing entrance into the cell cycle (6). Consistently, inhibition of pS6 by either the mTOR inhibitor rapamycin or the specific S6K1 inhibitor PF-4708671 in the double-digit nanomolar range was sufficient to reinstate cell-cycle arrest—but only when pRB was also inhibited—and resensitize PI3KCAE545K cells to MEKi + CDK4i (Fig. 4A and B and Supplementary Fig. S8). In vivo studies further confirmed that PF-4708671 was able to revert the resistance via the restoration of cell-cycle arrest and without obvious toxicity. This mechanism is consistent with the known phenotype of liver-specific S6 mouse knockouts having defective liver regeneration via initiation of cell-cycle arrest (24). Relevantly, pS6 has previously been described as a key intrinsic resistance biomarker during early BRAF or MEK inhibition in melanoma (33). Our findings on acquired resistance involving the MEK, CDK4/6, and PI3K pathways and sufficiency of the understudied S6K1i drug class (28) to reverse the resistance expand our understanding of the role of S6 in targeted therapy. Together, these findings suggest that pRB and pS6 may cooperate in controlling cancer cell proliferation downstream of multiple pathways (Fig. 6C) and posit the S6K1–S6 axis as a critical drug target node at their convergence, enabling simultaneous inhibition of all three oncogenic pathways at the level of a single shared downstream effector. We also note that activating PIK3CA mutations occur in 20% to 30% of colon cancer, 1% to 2% of pancreatic cancer, and 1% to 3% of non–small cell lung cancer, cancers that are just beginning MEKi + CDK4i clinical trials.
In summary, our findings shed light on the evolution of tumor resistance from a rare preexisting subpopulation and lay a blueprint for the functional identification of countertherapeutic vulnerabilities. Currently, it is unknown the ratio at which acquired resistance mutations in patients arise from de novo versus preexisting mutations (18, 32); this knowledge would be critical as they have different implications for therapeutic management. Our finding that a very low frequency, spatially segregated preexisting resistance subclone can rapidly expand in response to therapy suggests that such rare subclones may exist more often than previously appreciated. It also argues that comprehensive, high-resolution, multiregion sampling is crucial for uncovering such rare subclones; determining how often they exist in the general cancer patient population would provide a foundation from which to rationally strategize detection and differential counterresistance therapy options. Similarly, a thorough cataloging of potentially effective counterresistance therapies like MEKi + CDKi + S6K1i would enable the design of preclinical and clinical trials with highly personalized, precision medicine. As targeted therapy in melanoma and other cancers moves steadily toward second- and third-line treatments to overcome acquired resistance to previous regimens, the existence of rare resistant subpopulations prior to therapy argues that improving first-line combination treatments informed by functional studies may pay up-front dividends and extend patient survival.
Patient Tumor Samples and Imaging
The patient was treated at The University of Texas (UT) MD Anderson Cancer Center between October 2013 and January 2015. The patient was enrolled to clinical trial NCT01781572, which was conducted in accordance with the Declaration of Helsinki. Tumor samples were obtained from the MD Anderson Cancer Center Department of Pathology archive and Institutional Tumor Bank with appropriate written informed consent. Biopsy collection and analyses were approved by the MD Anderson Cancer Center Institutional Review Board (PA12-0305). LDH analysis and PET and spiral CT scans were obtained using standard procedures in the UT MD Anderson Cancer Center as part of the routine clinical care of the patient. RECIST 1.1 measurements were performed by a physician formally trained in tumor metrics. An NGS-based analysis for the detection of frequently reported (hotspot) mutations in a total of 46 genes, including the genes requested clinically, was performed on the DNA extracted from the sample in a Clinical Laboratory Improvement Amendments–certified molecular diagnostics laboratory.
After fixation and mounting, 5 to 10 slices with 5-μm thickness were obtained from formalin-fixed, paraffin-embedded (FFPE) tumor blocks. Tumor-enriched tissue was macrodissected, and xylene (EMD Millipore) was used for deparaffinization, followed by two ethanol washes. Reagents from the Qiagen QIAamp DNA FFPE Tissue Kit (#56404) were used in conjunction with an overnight incubation at 55°C to complete tissue lysis. Next, samples were incubated at 90°C for 1 hour to reverse formaldehyde modification of nucleic acids. After isolation by QIAamp MinElute column, variable amounts of buffer ATE were added to each column to elute the DNA. Germline DNA was obtained from peripheral blood mononuclear cells. The post-resistance (“post-ipi”) biopsy was exhausted during this procedure, making slides unavailable for IHC or other procedures.
The initial genomic DNA input into the shearing step was 250 ng in 55 μL of low tris-EDTA buffer. Forked Illumina paired-end adapters with random 8–base pair indexes were used for adapter ligation. All reagents used for end repair, A-base addition, adapter ligation, and library enrichment polymerase chain reaction (PCR) were from the KAPA Hyper Prep Kit (#KK8504). Unligated adapter and/or adapter-dimer molecules were removed from the libraries before cluster generation using solid-phase reverse immobilization bead cleanup. The elution volume after post-ligation cleanup was 25 μL. Library construction was performed following the manufacturer's instructions. Sample concentrations were measured after library construction using the Agilent Bioanalyzer. Each hybridization reaction contained 650 to 750 ng of the prepared library in a volume of 3.4 μL. Samples were lyophilized and reconstituted to bring the final concentration to 221 ng/μL. After reconstitution, the Agilent SureSelectXT Target Enrichment (#5190-8646) protocol was followed according to the manufacturer's guidelines. The libraries were then normalized to equal concentrations using an Eppendorf Mastercycler EP Gradient instrument and pooled to equimolar amounts on the Agilent Bravo B platform. Library pools were quantified using the KAPA Library Quantification Kit (#KK4824). On the basis of quantitative PCR quantification, libraries were then brought to 2 nmol/L and denatured using 0.2 N NaOH. After denaturation, libraries were diluted to 14 to 20 pmol/L using Illumina hybridization buffer. Next, cluster amplification was performed on denatured templates according to the manufacturer's guidelines (Illumina), HiSeq v3 cluster chemistry and flow cells, as well as Illumina's Multiplexing Sequencing Primer Kit. The pools were then added to flow cells using the cBot System, sequenced using the HiSeq 2000/2500 v3 sequencing-by-synthesis method, and then analyzed using RTA v.1.13 or later. Each pool of whole-exome libraries was subjected to paired 76-base pair runs. An 8–base pair index-sequencing read was used to meet coverage and to demultiplex the pooled samples. The mean coverage for exome data was 177 ×. Sequence data have been deposited at the European Genome–phenome Archive (EGA), which is hosted by the European Bioinformatics Institute and the Centre for Genomic Regulation, under accession number EGAS00001002846.
Exome sequencing data were processed using Saturn V, the NGS data processing and analysis pipeline developed and maintained by the bioinformatics group of the Institute for Applied Cancer Science and Department of Genomic Medicine at the UT MD Anderson Cancer Center. BCL files (raw output of Illumina HiSeq) were processed using Illumina CASAVA (Consensus Assessment of Sequence and Variation) software (v1.8.2) for demultiplexing/conversion to FASTQ format. The FASTQ files were then aligned to the hg19 human genome build using BWA (v0.7.5). The aligned BAM files were subjected to mark duplication, realignment, and recalibration using the Picard tool and GATK software tools. The BAM files were then used for downstream analysis. MuTect (v1.1.4) was applied to identify somatic point mutations. ABSOLUTE algorithm was used to estimate sample purity, ploidy, and absolute somatic copy numbers.
BDA qPCR assays were performed on a CFX96 Touch Real-Time PCR Detection System using 96-well plates (Bio-Rad). In a typical Taq polymerase-based assay, PowerUp SYBR Green Master Mix (Thermo Fisher) was used for enzymatic amplification and fluorescence signal generation; primer concentrations were 400 nmol/L each, and blocker concentration was 4 μmol/L. gDNA sample input was 10 ng per well; reactions were performed in triplicate, and the total volume was 20 μL in each well. Thermal cycling started with a 3-minute incubation step at 95°C for polymerase activation, followed by 66 repeated cycles of 10 seconds at 95°C for DNA denaturing and 30 seconds at 60°C for annealing/extension. All amplicon sequences were verified by Sanger sequencing (GeneWiz) to verify mutation identity.
Cell Lines and Drugs
SB2 and WM1366 human melanoma cell lines were maintained in RPMI-1640 supplemented with 10% FBS and penicillin–streptomycin (100 U/mL—100 μg/mL). Cells were acquired in 2012. SB2 cells are a kind gift of Dr. Menashe Bar-Eli; WM1366 cells were acquired from MD Anderson IACS cell bank. Cells were last authenticated in 2016 by T200 sequencing platform. SB2 harbors the following notable mutations: NRASQ61K, DDX3XK35X, RAC1P34H, and a TERT promoter mutation. WM1366 harbors the following mutations: NRASQ61L, DDX3XG123X, RAC1V51M, CDKN2AP24L, NF1P2242L, and a TERT promoter mutation. hTERT-immortalized human melanocytes were purchased from ABMGood and maintained in Prigrow II medium supplemented with 10% FBS and penicillin–streptomycin (100 U/mL—100 μg/mL). Drugs used were MEK162 (MEK inhibitor; Chemietek), LE0011 (CDK4 inhibitor; Chemietek), rapamycin (mTOR inhibitor; Selleckem), and PF-4708671 (SK61 inhibitor; Selleckem) at the indicated concentrations.
Cell Titer, Proliferation, and Apoptosis Assessment
For cell titer, apoptosis and CSFE analysis, the time-lapse fluorescence microscope system IncuCyte was used (Essen Bioscience); 10,000 cells per well were seeded in a 96-well plate, allowed to adhere for 24 hours and then administered with drugs. Cell confluence and fluorescence were recorded at 2-hour intervals and analyzed using IncuCyte Zoom software; cell titer was expressed as the cell confluence measured at the 72-hour time point after treatment initiation normalized to control treatment (DMSO 0.01%).
For apoptosis assays, IncuCyte Annexin V Red Reagent and IncuCyte Caspase 3/7 Reagent were used according to the manufacturer's instructions. Briefly, caspase 3/7 reagent (5 μmol/L) and Annexin V (1:200) were added to cells simultaneously with drug treatments; green and red fluorescence were measured every 2 hours for 72 hours, and the percentage of green or red cells was normalized to the percentage of cell confluence and expressed as fluorescence units. For CFSE cell proliferation analysis, cells were treated with 5 μmol/L CFSE and incubated at 37°C for 30 minutes. After two washings, cells were seeded and let adhere 24 hours. Green fluorescence was monitored every 2 hours. The percentage of CFSE retaining cells was normalized to the percentage of cell confluence.
For propidium iodide assays, cells were detached using trypsin and immediately fixed in 70% v/v ethanol at −20°C (24 hours after treatment start). Cells were then treated with DNAse free RNAse (200 μg/mL) and stained with propidium iodide (20 μg/mL). Fluorescence was assessed using BD Accuri. Data were analyzed using FlowJo Software (v10.2).
Plasmids and Viral Infections
pBabe-puro-HA-PIK3CA and pBabe-puro-HA-PIK3CA-E545K were purchased from Addgene (#12522 and #12525, respectively). Retrovirus was produced by transfecting 293T cells with the expression plasmids and the two packaging vectors VSVG and pHIT60, using PEI (3 μg per μg of DNA). Viral supernatants were collected after 72 hours, filtered through a 0.45-μm filter and used to infect target cells in addition with polybrene (8 μg/mL, SantaCruz). After 24 hours, cells were washed and put under antibiotic selection for 96 hours (puromycin 1 μg/mL).
RPPA and Immunoblotting
Melanoma cell lines were lysed using RIPA buffer added with Protease and Phosphatase Inhibitor Cocktail (Thermo Fisher Scientific) 24 hours after the initiation of treatments. For molecular analysis, drugs were used at a concentration of 300 nmol/L. Protein content was estimated using BCA protein assay (Thermo Fisher Scientific) and resolved by SDS-PAGE on precast mini gels (Invitrogen). Blotting was performed using the Trans Blot Turbo Transfer System (Bio-Rad). After blocking, nitrocellulose membranes were incubated with primary antibody. Primary antibodies used were anti-mTOR (#2983), anti–phospho-S2448-mTOR (#5536), anti–p70-S6 Kinase (#9202), anti–phospho-T389-p70-S6 kinase (#9234), anti-AKT (#4691), anti–phospho-S473-AKT (#4060), anti-Ribosomal-Protein-S6 (#2317), anti–phospho-S235/236-Ribosomal-Protein-S6 (#4858), anti-RB (#9309), anti–phospho-S780-RB (#8180), anti-4EBP1 (#9644), anti-phopsho-T37/46-4EBP1 (#2855), anti-ERK1/2 (#9102), anti-phospho-T202/Y204-ERK1/2 (#4370; 1:1,000; Cell Signaling Technology), and anti-GAPDH (1:5,000; Millipore). The appropriate HRP-conjugated secondary antibody was then applied; chemiluminescent reaction was induced by SuperSignal West Dura Extended Duration Substrate (Thermo Fisher Scientific) and then impressed on autoradiographic film. Secondary antibodies used were anti-rabbit IgG, HRP-linked antibody (#7074), anti-mouse IgG, and HRP-linked antibody (#7076; 1:5,000; Cell Signaling Technology). RPPA was performed by MD Anderson RPPA core.
DNA was extracted from patient plasma using Qiagen's QiAmp circulating nucleic acid kit per the manufacturer's instructions. Extracted DNA was eluted in 30-μL low-TE buffer and quantified using Qubit fluorometer (ThermoFisher Scientific). Each sample was run in duplicate. The maximum volume of sample input was used for setting up droplet PCR reaction. The droplets were automatically generated using Bio-Rad's automatic droplet generator, after which they were amplified in a deep-well thermocycler. The droplets were read with the Bio-Rad droplet reader and analyzed using the Quantasoft software (Bio-Rad).
Immunohistochemistry and TUNEL Assay
FFPE tumor specimens were sectioned with a microtome (5-μm sections) and put on slides. After deparaffinization, citrate-based antigen retrieval was performed. Blocking was performed using Normal Horse Serum (ImmPRESS Reagent Kit, Vector Labs). Slides were then incubated with anti–phospho-S235/236-Ribosomal-Protein-S6 (#4858, Cell Signaling Technology) or anti–phospho-Histone H3 (#9701, Cell Signaling Technology) primary antibody overnight at 4°C in a humid chamber, avoiding drying of specimens. Slides were then incubated with HRP-conjugated anti-rabbit secondary antibody. Processed slides were finally added with 3,3′-diaminobenzidine (DAB; Vector Laboratories) to develop a chromogenic reaction or TSA Plus Fluorescein System (Perkin-Elmer) to allow flourophore deposition and counterstained with hematoxylin (Vector Laboratories) or DAPI respectively. Coverslips were mounted after dehydration of the sections, using Permanent Mounting Medium (Vector Laboratories). TUNEL assay (Biotium) was performed following the manufacturer's instructions. Images were digitally acquired with ScanScope (Aperio). Images were digitally acquired with ScanScope (Aperio) and Nikon Eclipse T2 microscope connected to DS-Ri2 camera (Nikon). Quantification of IHC and TUNEL assay signal, expressed as number of positive cells/number of total cells, was performed using ImageJ (Rasband, W.S., ImageJ, U.S. National Institutes of Health, Bethesda, MD, https://imagej.nih.gov/ij/, 1997–2016.).
Data are presented as means ± SEM. For paired or unpaired t test, a P value <0.05 was considered statistically significant. For RPPA expression dataset analysis, multiple t tests were performed with an FDR of 1%. Statistical analysis was performed using GraphPad Prism (GraphPad Software).
SB2-GFP or SB2-E545K cells (5 × 106) were intradermally injected in the flanks of athymic nude female mice (Taconic). Tumor volumes were calculated using electronic calipers to measure the length (l), width (w), and height (h) and using the formula (l × w × h) × π/6. When the tumors reached the average volume of 100 mm3, animals were distributed among the treatment groups. MEK162 and PF-4708671 were resuspended in 1% carboxymethylcellulose/0.5% Tween 80; LEE011 was resuspended in 0.5% methylcellulose. Because of different solubility properties of the drugs, LEE011 was always dosed 4 hours apart from MEK162 and PF-4708671. MEK162 (10 mg/kg, twice a day), LEE011 (100 mg/kg, daily), and PF-4708671 (50 mg/kg, daily) were administered by oral gavage using sterile flexible plastic adapters in a volume of 100 μL. Body mass was measured using an electronic scale. Animals were euthanized when the tumor burden reached >1,000 mm3 or when the tumor became ulcerated. Animals used for molecular analysis were euthanized 4 days after treatment start. All animal experiments were performed according to protocols approved by the Institutional Animal Care and Use Committees of The University of Texas MD Anderson Cancer Center.
Disclosure of Potential Conflicts of Interest
A.E. Aplin reports receiving a commercial research grant from Pfizer. M.A. Davies reports receiving commercial research grants from GSK, Roche/Genentech, AstraZeneca, Merck, Oncothyreon, and Sanofi Aventis, and is a consultant/advisory board member for GSK, Roche/Genentech, Novartis, Sanofi, Vaccinex, and Syndax. R.N. Amaria reports receiving commercial research support from Array Pharmaceuticals. D.Y. Zhang is co-founder and consultant at Nuprobe Global, has received honoraria from the speakers bureau of New England Biolabs, has ownership interest (including patents) in Nuprobe Global, and is a consultant/advisory board member for Nuprobe Global. J.A. Wargo reports receiving commercial research support from GSK. L.N. Kwong reports receiving a commercial research grant from Array Biopharma. No potential conflicts of interest were disclosed by the other authors.
Conception and design: G. Romano, L.N. Kwong
Development of methodology: G. Romano, P. Song, A.J. Lazar, D.Y. Zhang
Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): G. Romano, P.-L. Chen, P. Song, J.L. McQuade, R.J. Liang, M. Liu, D.Y. Duose, F.CL. Carapeto, M. Chen, A.J. Lazar, P.A. Futreal, R.N. Amaria, L.N. Kwong
Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): G. Romano, P. Song, J.L. McQuade, W. Roh, D.Y. Duose, J. Li, J.L.F. Teh, A.E. Aplin, J. Zhang, A.J. Lazar, R.N. Amaria, D.Y. Zhang, J.A. Wargo, L.N. Kwong
Writing, review, and/or revision of the manuscript: G. Romano, J.L. McQuade, R.J. Liang, W. Roh, J.L.F. Teh, A.J. Lazar, M.A. Davies, P.A. Futreal, R.N. Amaria, D.Y. Zhang, J.A. Wargo, L.N. Kwong
Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): J.L. McQuade, J.A. Wargo, L.N. Kwong
Study supervision: G. Romano, L.N. Kwong
We thank the Melanoma informatics, tissue resource, and pathology core (MELCORE) for patient sample retrieving and processing. We also thank Gao Zhang and Kate Nathanson for help with WM1366. The RPPA studies performed at the Functional Proteomics Core Facility at MD Anderson Cancer Center were supported by a National Cancer Institute (NCI) Cancer Center Support Grant (CA-16672). W. Roh was supported by a CPRIT Graduate Scholar Award. M.A. Davies is supported by the Dr. Miriam and Sheldon G. Adelson Medical Research Foundation. M.A. Davies and A.J. Lazar are supported by the MD Anderson Melanoma Moon Shots Programs. P. Song and D.Y. Zhang were supported by grant R01CA203964 from the NCI. A.E. Aplin was supported by NIH/NCI R01 grant CA182635. L.N. Kwong was supported by the University of Texas Rising STARS award and the Melanoma Research Alliance Young Investigator Award 508743. L.N. Kwong and P.A. Futreal were supported by NIH/NCI grant 4P01 CA163222-04.