Targeting the CDK4/6-Rb Pathway Enhances Response to PI3K

Published OnlineFirst August 9, 2018; DOI: 10.1158/1078-0432.CCR-18-0717

Translational Cancer Mechanisms and Therapy Clinical Cancer Research Targeting the CDK4/6-Rb Pathway Enhances Response to PI3K Inhibition in PIK3CA-Mutant Lung Squamous Cell Carcinoma Ruoshi Shi1,2, Ming Li1, Vibha Raghavan1, Shirley Tam1, Michael Cabanero1, Nhu-An Pham1, Frances A. Shepherd1,3, Nadeem Moghal1,2, and Ming-Sound Tsao1,2,4

Abstract

Purpose: Lung squamous cell carcinoma (LUSC) is a major of LUSC PDX models identified PI3K pathway alterations in subtype of non–small cell lung cancer characterized by mul- over 50% of the models. In vivo screening using PI3K tiple genetic alterations, particularly PI3K pathway alterations inhibitors in 12 of these models identified PIK3CA mutation which have been identified in over 50% of LUSC cases. Despite as a predictive biomarker of response (<20% tumor growth being an attractive target, single-agent PI3K inhibitors have compared with baseline/vehicle). Combined inhibition of demonstrated modest response in LUSC. Thus, novel combi- PI3K and CDK4/6 in models with PIK3CA mutation resulted nation therapies targeting LUSC are needed. in greater antitumor effects compared with either mono- Experimental Design: PI3K inhibitors alone and in com- therapy alone. In addition, the combination of the two bination with CDK4/6 inhibitors were evaluated in previously drugs achieved targeted inhibition of the PI3K and cell-cycle established LUSC patient-derived xenografts (PDX) using an pathways. in vivo screening method. Screening results were validated with Conclusions: PIK3CA mutations predict response to PI3K in vivo expansion to 5 to 8 mice per arm. Pharmacodynamics inhibitors in LUSC. Combined PI3K and CDK4/6 inhibition studies were performed to confirm targeted inhibition of enhances response to either single agents alone. Our findings compounds. provide a rationale for clinical testing of combined PI3K Results: Consistent with results from The Cancer Genome and CDK4/6 inhibitors in PIK3CA-mutant LUSC. Clin Cancer AtlasanalysisofLUSC,genomicprofiling of our large cohort Res; 24(23); 5990–6000. 2018 AACR.

Introduction (2–4). However, to date, there has yet been very little successful clinical translation of these discoveries. Although preclinical work Lung squamous cell carcinoma (LUSC) is a common subtype of has been reported on these molecular targets (2), none of the non–small cell lung cancer (NSCLC) affecting over 300,000 inhibitors tested in early-phase clinical trials has demonstrated individuals worldwide each year (1). The current standard of care high response rates. The major shortcomings of these clinical trials for LUSC is surgical resection for early stages and cytotoxic could include drug toxicities and the lack of robust predictive chemotherapy and immunotherapy for inoperable disease. To biomarkers to select patients who may benefit from these date, the identification of treatment strategies for targetable dri- therapies. vers in LUSC has remained an elusive goal (2), with the possible One particular pathway whose alteration has been investigated exception of immunotherapy, although its long-term effects is the PI3K pathway. PI3K belongs to a family of lipid kinases remain to be seen. The Cancer Genome Atlas and other consortia involved in the regulation of cell proliferation, survival, and have molecularly characterized LUSC in large cohort studies, metabolism (5). This pathway has been explored in LUSC largely identifying numerous potentially targetable alterations including due to the high frequency of perturbations, with PIK3CA muta- FGFR1 amplification, PI3K pathway mutations/copy-number tions, amplifications, and PTEN loss being identified in 10% to changes, DDR2 mutation, and NRF2-KEAP1 pathway aberrations 15%, 50%, and 20% to 30% of LUSC cases (3). These PI3K pathway alterations putatively confer constitutive activation of 1University Health Network, Ontario Cancer Institute/Princess Margaret Cancer the pathway, resulting in oncogenic transformation in normal Centre, Toronto, Ontario, Canada. 2Department of Medical Biophysics, breast epithelial cell lines in soft-agar colony formation assays, University of Toronto, Toronto, Ontario, Canada. 3Department of Medical increased tumorigenic potential in mice, and predictive response Oncology and Hematology, University of Toronto, Toronto, Ontario, Canada. to PI3K inhibition in in vitro preclinical models (6–8). Inhibitors 4 Department of Laboratory Medicine and Pathobiology, University of Toronto, of PI3K such as BKM120 and BYL719 have demonstrated efficacy Toronto, Ontario, Canada. in PIK3CA-mutant pan-cancer cell lines, and these agents now are Note: Supplementary data for this article are available at Clinical Cancer being evaluated in clinical trials (9). Research Online (http://clincancerres.aacrjournals.org/). LUSC patients harboring PI3K-activated tumors have poorer Corresponding Author: Ming-Sound Tsao, University Health Network, Princess prognosis than patients without PI3K aberrations (10). Unfortu- Margaret Cancer Centre, 101 College Street 14-305, Toronto, Ontario M5G 1L7, nately, despite the potential driver role of PI3K alterations in Canada. Phone: 416-340-4737; Fax: 416-340-5517; E-mail: [email protected] LUSC, results from the BASALT-1 clinical trial (NCT01820325) of doi: 10.1158/1078-0432.CCR-18-0717 BKM120 in stage IV NSCLC patients were not as promising as 2018 American Association for Cancer Research. anticipated (11). Thus, reliable predictive biomarkers of response

5990 Clin Cancer Res; 24(23) December 1, 2018

Dual Targeting of PI3K and CDK4/6 in PIK3CA-Mutant LUSC

filtering implemented for samples without a matched normal Translational Relevance using dbSNP, ExAC, and ESP (17, 18). Annovar was used to There is currently a lack of targeted therapies that have annotate all the final calls (19). For CNV, DNA from each demonstrated robust efficacy in lung squamous cell carcinoma sample was hybridized to either the HumanOmni 2.5 (LUSC). In our study, we determined through an in vivo BeadChip or the HumanOmni 1-Quad BeadChip SNP arrays screen that PIK3CA-mutated LUSC may better respond to (Illumina). The LogR and B allele frequency values were PI3K inhibition. In addition, combined PI3K and CDK4/6 extracted at each probe locus for all the samples, which was inhibition enhanced response to PI3K monotherapy in then used by Allele-Specific Copy Number Analysis of Tumors PIK3CA-mutated tumors compared with PI3K wild-type (20) to identify copy-number gains and losses within each LUSC. Our data support further clinical evaluation of com- sample. GISTIC v2.0.23 (21) was used to determine regions bined PI3K and CDK4/6 inhibitors in PIK3CA-mutated LUSC. that are significantly amplified or deleted across the samples. For IHC, PTEN IHC tissue microarray (TMA) study was reported previously (22).

In vivo and combination therapies are needed to improve LUSC response drug treatment studies in vivo to PI3K inhibitors. In addition, a combination therapy strategy in The 1 1 1 drug screening method was adopted from a previous study (23). Drug treatment began when tumors reached LUSC could potentially mitigate toxicity resulting from single- 3 agent PI3K inhibitor by reducing drug doses. Therefore, preclin- 200 to 300 mm . Tumors for each model were implanted into ical testing of PI3K inhibitors and combination therapy using 2 NOD/SCID mice, one treated with vehicle and the other with clinically relevant LUSC models is needed. drug via daily oral gavage for 21 days. Tumors were harvested at fi fi In the present study, we used chemotherapy-na€ve LUSC the experimental endpoint, formalin- xed paraf n-embedded patient-derived xenografts (PDX) as preclinical models to screen and snap frozen for subsequent histologic analysis and protein for sensitivity to PI3K inhibitors and potential combination isolation. Waterfall plots of response were generated by calculat- targeted drugs. We determined that PIK3CA mutation is a pre- ing the slope of growth at 10-day intervals compared with base- fi dictor of PI3K inhibitor response (<20% tumor growth compared line. Drug response was determined by the Modi ed RECIST (23). with baseline/vehicle). In addition, combined inhibition of Drug response was validated with an expansion study of 6 to CDK4/6 with PI3K inhibitors in PDXs harboring PIK3CA muta- 8 mice per arm in which tumors were treated for 25 to 30 days. tion and p16 loss resulted in enhanced tumor shrinkage which BKM120 (25 and 50 mg/kg), BYL719 (25 and 50 mg/kg), was not observed with single-agent treatments. We provide a palbociclib (150 and 75 mg/kg), and abemaciclib (50 mg/kg) rationale for clinical testing of combined PI3K and CDK4/6 were purchased from University Health Network Shanghai, Inc. inhibition in LUSC. BKM120, BYL719, palbociclib, and abemaciclib were dissolved in PEG300/NMP, 0.5% carboxymethylcellulose, 0.05N sodium ¼ Materials and Methods lactate (pH 4), and 1% carboxyethylcellulose, respectively (Sigma). Patient samples and PDX establishment Establishment of PDXs from surgically resected primary LUSC FISH tumors from early-stage LUSC patients was approved by the The Vysis LSI CDKN2A SpectrumOrange/Vysis CEP 9 University Health Network Human Research Ethics and Animal SpectrumGreen probes were used for FISH analysis of CDKN2A Care Committee (REB #09-0510-T). All patients provided deletion (Abbott Molecular). PTEN FISH probes were described informed written consent. All studies were performed in accor- in a previous publication (22). Tissue sections were incubated dance with TRI-Council Policy Statement: Ethical Conduct for at 56C for 4 hours, deparaffinized in CitriSolv, and air dried. Research Involving Humans. Patient samples and the protocol for The slides were incubated in 2x saline sodium citrate (SSC) at establishing LUSC PDXs were previously described (12). The 75C for 15 minutes and 0.6 mg/mL Proteinase K at 45Cfor identities of LUSC PDXs were verified by short tandem repeat 18 minutes prior to probe hybridization. DNA was denatured at analysis. 85C for 15 minutes prior to hybridization at 37C for 40 hours. The slides were then washed in 2x SSC/0.3% NP-40 at 74Cand Molecular characterization of PDXs 2x SSC at room temperature for 2 minutes each and dehy- GenomicDNAandRNAwereextractedaccordingtoa drated. Then, 14 mL of DAPI/anti-fade was used to mount the protocol mentioned previously (12). Whole-exome enrich- slides. FISH results were analyzed using Cyto Vision software ment was performed using the Agilent SureSelect Human All (Leica Biosystems). Exon V4 and SureSelect Human All Exon V4þUTR capture kits. Captured libraries were then sequenced using a HiSeq 2000 Immunohistochemistry sequencer, and paired-end sequence reads were generated for Formalin-fixed paraffin-embedded tissues were sliced into each sample. Subsequently, Xenome (13) was used to remove 4-mm-thick sections and processed for hematoxylin and eosin any contaminating reads pertaining to mouse stroma, follow- staining and IHC. All IHC analyses were performed using a ing which, Burrows–Wheeler Aligner v0.7.12 was used to align BenchMark XT autostainer (Ventana Medical Systems). the reads to the human reference genome (GRCh37; ref. 14). Phospho-S6 Ser235/236 (Cell Signaling Technology; #2211), The standard GATK pipeline, involving Picard v1.140, was p16 (Ventana; #725–4713), and Rb (Cell Signaling then employed to perform quality control and further proces- Technology; #9309) IHC were performed using an optimized sing of the mapped data. Somatic mutations were called using laboratory protocol. Loss of PTEN protein expression was defined Mutect v1.1.5 (15) and Varscan v2.3.8 (16), with additional by complete loss of cytoplasmic and nuclear expression (22).

www.aacrjournals.org Clin Cancer Res; 24(23) December 1, 2018 5991

Shi et al.

Western blot analysis NSCLC patient tumors. Specific details regarding patient charac- Snap-frozen tumors were homogenized in RIPA buffer, and teristics, PDX engraftment rate, and histology have been reported protein was quantified using BCA assay according to the manu- previously (30). Of our 61 established LUSC PDX models, facturer's instructions (Thermo Fisher Scientific; #23225). Protein 51 models were profiled for mutations, copy-number aberrations, was resolved on a 4% to 20% Midi gel (Bio-rad; #5671094) and and gene expression. PTEN protein expression loss was deter- transferred on nitrocellulose membrane. pAkt (S473, #9271), mined via IHC on TMA based on the criteria of a complete absence total Akt (#9272), pS6 (Ser235/236, #2211), total S6 (#2217), of cytoplasmic and nuclear staining. PIK3CA mutations, ampli- pRb (S780, #9307), and total Rb (#9309) antibodies were fications, and PTEN loss were examined in our cohort as potential obtained from Cell Signaling Technologies. b-Actin (#A1978) biomarkers for PI3K inhibitors due to the high frequency of was obtained from Sigma. these alterations in LUSC. Of the 51 profiled PDX models, 8 of 51 (16%) harbored PIK3CA mutations, (19/51) 37% harbored Differential gene expression analysis PIK3CA amplification, and 22 of 51 (43%) harbored PTEN loss Aliquots of the RNA samples were provided to the University (Fig. 1A). Not all of the models were mutually exclusive in terms Health Network Princess Margaret Genomics Centre for analysis of PI3K alterations as overlaps among PIK3CA mutation, ampli- using the Illumina Whole-Genome Gene Expression Direct fication, and PTEN loss were identified in a subset of LUSC PDXs Hybridization assay or the Whole-Genome DASL HT Assay. In (Fig. 1B). Of the 8 PIK3CA-mutated PDXs, 6 of 8 (75%) harbored brief, 200 ng of total RNA from each sample was labeled and the hotspot mutations E545K and E542K, whereas 2 of 8 (25%) amplified using the Illumina Whole-Genome Gene Expression harbored unreported mutations S438C and W446C (refs. 31, Direct Hybridization or Whole-Genome DASL assay kit according 32; Fig. 1C). to the manufacturer's protocol (Illumina). The labeled samples were hybridized to Human HT-12 v4.0 BeadChips, incubated at PIK3CA-mutant LUSC PDXs are sensitive to pan-PI3K inhibitor 58C for hybridization for 18 hours. The BeadChips were then To validate the general BKM120 response rates of PDX har- washed and stained as per the Illumina protocol, and scanned boring PI3K pathway alterations, we selected 12 LUSC PDX on the iScan (Illumina). The data files were quantified in models harboring PIK3CA E545K or E542K mutations, PIK3CA GenomeStudio (Illumina) and loaded into the R statistical envi- amplification, PTEN loss, or a combination of two alterations ronment (v3.3.0) using the lumi package (v2.24.0; ref. 24), for in vivo BKM120 testing (Fig. 2A and B). We adopted the 1 implemented in the Bioconductor libraries (25). Data from the 1 1(n ¼ 1) method of using 1 mouse per model per drug individual platforms were log2 transformed and normalized initially to screen BKM120 response in PDXs (23). Response was using quantile normalization. No background correction was determined in a waterfall plot by averaging the slope of the growth performed. Common probes detected across the two platforms rate from treatment day 0 (baseline) to day 10, and days 10 to 20. were selected, and the data were combined. Batch effects were A Modified RECIST proposed in prior work (23) was used to adjusted using ComBat (26), implemented in the sva package categorize tumor response, in which tumors that grew more than (v3.20.0; ref. 27). Differential gene expression was performed 20% above baseline size were considered nonresponders and using the limma (Bioconductor) package (28). Top 200 upregu- tumors that did not increase more than 20% were considered lated and downregulated genes (P < 0.01) were submitted to responders. Among the 12 models tested, 4 of 12 (33%) were DAVID for gene set enrichment analysis (29). Top upregulated responders, whereas 8 of 12 (67%) were nonresponders (Fig. 2C; and downregulated genes were clustered using the heatmap2 Supplementary Fig. S1A). To verify whether growth rate variation package in R. Gene expression of the 9 LUSC PDXs was deposited affected response in our PDX models, the growth response of the in the Gene Expression Omnibus with GSE109679. BKM120-treated tumor was normalized to that of the vehicle. The responses before and after normalization were similar (Supple- Methylation analysis mentary Fig. S1B). Thus, the four responding models and eight Methylation was performed using Infinium MethylationEPIC nonresponding models were appropriately categorized. BeadChip arrays (Illumina). Note that 250 ng of DNA extracted This screen revealed that 4 of 5 PIK3CA mutants responded to from fresh-frozen tissue samples were subjected to bisulfite con- BKM120, whereas PDXs harboring PIK3CA amplification, PTEN version using the EZ DNA Methylation Kit (Zymo Research). The loss, or wild-type PI3K did not respond to treatment. BKM120 DNA (4 mL) was then subjected to whole-genome amplification sensitivity in the four responders from the n ¼ 1 study was further and loaded onto BeadChips, followed by single-base extension validated by expanding the BKM120 drug testing to 6 mice per and fluorophore labeling. The BeadChips subsequently were arm. Consistent with the screening results, the 4 responders from scanned using an iScan instrument (Illumina). Data processing the initial study demonstrated a durable response with 20 to and analysis were described previously (12). Methylation scores 40 days of BKM120 treatment and reflected the n ¼ 1 study were compared between p16 IHC-positive and -negative samples. (Fig. 2D–G). To determine whether BKM120 targeted the PI3K pathway, a pharmacodynamics study of phospho-S6 (pS6) and Statistical analysis phospho-Akt (pAkt) was performed with Western blotting and The Student t test and two-way ANOVA were used to calculate IHC at 2 and 16 hours after BKM120 dosing. Targeted inhibition P values for in vivo drug studies at selected time points. of pAkt and pS6 was achieved in the 4 responders at the 2-hour time point, with a rebound of these markers observed at 16 hours, which is consistent with previous pharmacodynamics studies of Results BKM120 (ref. 7; Supplementary Fig. S2). To determine whether PI3K pathway alterations in LUSC PDX models the response to BKM120 was due to the inhibition of PI3Ka, the LUSC PDX models, we refer to as patient human lung cancer PI3Ka isoform–specific inhibitor BYL719 was tested in the four (PHLC), were established from surgically resected early-stage responders. Using the n ¼ 1 method, three of the four models

5992 Clin Cancer Res; 24(23) December 1, 2018 Clinical Cancer Research

Dual Targeting of PI3K and CDK4/6 in PIK3CA-Mutant LUSC

A Oncoprint for LUSC PHLC (n = 51)

47% PIK3CA

43% PTEN 50 62 55 86 57 77 85 422 356 377 353 268 255 322 451 225 272 286 321 274 132 152 410 267 188 106 123 299 462 247 369 122 271 328 252 149 193 204 205 139 428 220 223 200 365 273 417 242 171 113 156 Deep deletion Amplification Mutation Downregulaton (IHC)

B C PHLC PIK3CA mutation PIK3CA mutation 188 E542K 223 E542K 3 267 E545K 356 E542K 1 4 410 E545K PTEN PIK3CA 422 E545K loss 8 amplification 152 S438C 13 7 220 W446C

Figure 1. PIK3CA and PTEN alterations in LUSC PDX. A, Oncoprint of PIK3CA mutation, amplification, and PTEN loss in 51 LUSC PDXs. PIK3CA mutations were detected by exome sequencing followed by sanger sequencing for validation. PIK3CA amplifications/loss were detected by SNP array and FISH. PTEN loss was identified by IHC assessment. B, Venn diagram of PIK3CA mutation, amplification, and PTEN loss overlaps in PI3K aberrant PDXs. C, Table summarizing specific PIK3CA mutations in 8 PDX models. PHLC, patient human lung cancer, nomenclature for our lung PDX models. responded to BYL719 (Supplementary Fig. S3A). Although were significantly elevated in samples with PIK3CA mutation PHLC410 did not respond in the n ¼ 1 study, it did show a compared with those with PI3K wild-type status and PTEN loss response to both BKM120 and BYL719, along with targeted (32). Taken together, these findings suggest PI3K signaling in inhibition, in an expanded study involving 4 mice per arm the 12 PDXs is more robustly activated in the PIK3CA mutants (Supplementary Fig. S3B and S3C). Overall, our screen demon- than in PI3K wild-type or PTEN-deleted models. Furthermore, strated that PIK3CA mutation may predict response to PI3K pS6 may be used as a biomarker to identify PIK3CA mutants inhibitors in chemo-na€ve LUSC. and PI3K inhibitor sensitivity.

PIK3CA-mutant PDXs express high levels of pS6 p16 loss is present in PIK3CA-mutant LUSC PDXs Due to the BKM120 response variation among our 12 LUSC Although our findings suggest that PDXs harboring PIK3CA PDX models in which an enrichment of PIK3CA mutants mutation are sensitive to BKM120, further investigation, partic- was identified in the responder group, we investigated why ularly strategies for combination therapy, is needed to improve on PIK3CA-mutant models were sensitive, whereas wild-type PDX this response. A global gene expression analysis was performed to models were resistant. Among the 12 models that were selected identify differentially expressed genes between responders and for BKM120 testing, the downstream PI3K activation target pS6 nonresponders (see Materials and Methods). We identified over was significantly more elevated in the PIK3CA mutants than in 279 downregulated and upregulated genes between the two PTEN-inactivated PDXs and PI3K wild-type PDXs (Fig. 3A–D). groups. Gene set enrichment analysis using DAVID tool revealed Notably, the percentage of pS6-positive cells between PDXs enrichment of downregulated genes in the responders involved in with PIK3CA mutation and PTEN loss were similar. However, apoptosis, cell–cell junction, and cell-cycle arrest. These categories all five PIK3CA mutants exhibited high expression of pS6, were examined due to their relevancy in the PI3K pathway whereas only two of four models with PTEN loss expressed (33, 34). Notably, CDKN2A, a gene encoding for the cell-cycle high pS6 levels. These data were concordant with another study regulator p16, was found to be downregulated in the responders done in a large pan-cancer patient cohort where pAkt and pS6 (Fig. 4A and B). p16 is a well-established tumor suppressor

www.aacrjournals.org Clin Cancer Res; 24(23) December 1, 2018 5993

Shi et al.

A B PHLC Mutation CNV IHC 61 LUSC PDX PIK3CA PIK3CA PTEN 10 excluded 149 - - - (molecular data 152 - - + unavailable) 188 E542K - - 51 LUSC PDX 267 E545K AMP + Excluded some slow 271 - AMP - growing models and 322 - - - those lacking PI3K 328 - AMP - pathway alterations 356 E542K AMP + 12 selected for 410 E545K AMP + in vivo BKM120 422 E545K AMP + Figure 2. screen 205 --+ In vivo BKM120 screen in LUSC 377 - AMP + PDX. A, Consort chart of the LUSC ﬂ C *AMP = amplification, (+) = present, (−) = loss PDX platform work ow. A total of in vivo 12 PDX models of 51 were selected BKM120 drug screen for PI3K inhibitor screen. Each 100 model was implanted into 2 mice, PIK3CA WT (n = 2) one treated with vehicle and the other with BKM120 for 21 days. B, n PIK3CA amp ( = 1) PI3K status of the 12 selected PDX 50 PIK3CA amp and PTEN loss (n = 2) models for BKM120 screening. Only PIK3CA PTEN loss (n = 2) hotspot mutations were in n reported. C, Waterfall plot of the PIK3CA mut and PTEN loss ( = 1) vivo BKM120 screen in LUSC PDXs. 0 PIK3CA mut and amp (n = 4) Models were treated with BKM120 (% baseline) (50 mg/kg) for 21 days, and change NR R in tumor volume was determined

Change in tumor volume NR = nonresponder by calculating the tumor volume −50 R = responder difference in 10-day intervals and 2 1 15 205 328 27 422 149 322 377 410 356 188 267 dividing by the baseline tumor volume (either on day 0 or day 10). PHLC Error bars are determined by the D PHLC188 E PHLC267 standard deviation of the mean. 1,000 Models that had tumor growth 1,500 n n ) Vehicle ( = 6) ) Vehicle ( = 6) 3 3 below 20% were considered BKM120 (n = 6) BKM120 (n = 6) 800 responders. D–G, Validation of the one screen in four BKM120- 1,000 600 Treat sensitive PDX models. Six mice Treat 400 * were included in each arm. 50 * P < P < ** , 0.05 and , 0.0001 ** 200 compared with vehicle using the Tumor volume (mm Tumor Tumor volume (mm Tumor Student t test at selected time 0 0 0204060 020406080 points. Day Day F PHLC356 G PHLC410 1,000 1,000 Vehicle (n = 6) Vehicle (n = 6) ) ) 3 BKM120 (n = 6) 3 800 BKM120 (n = 6) 800 600 600 Treat Treat 400 400 * * 200 ** ** 200 Tumor volume (mm Tumor 0 volume (mm Tumor 0 01020304050 01020304050 Day Day

and plays a major role in inhibiting the activity of the cyclin- levels were compared between models with low and high p16 dependent kinases CDK4/6. Loss of p16 is a recognized major protein expression by IHC. PDX models with low p16 protein oncogenic event in cancer such as in the pancreas and lung expression exhibited higher methylation proﬁle in the CDKN2A (35, 36). Furthermore, CDKN2A deletion is a common event gene, suggesting that although p16 loss generally is driven by that has been identiﬁed in over 20% of LUSC cases (3), suggesting copy-number deletion (37), methylation may also be involved in its downregulation is important in disease pathogenesis. To silencing this gene (Supplementary Fig. S4A and S4B). To deter- correlate p16 protein expression with methylation, methylation mine whether CDKN2A copy-number deletions can be detected in

5994 Clin Cancer Res; 24(23) December 1, 2018 Clinical Cancer Research

Dual Targeting of PI3K and CDK4/6 in PIK3CA-Mutant LUSC

A B 205 152 Table 1. pS6 expression in LUSC PHLC % Tumor area PHLC PIK3CA PTEN positive for pS6 *152 - + 1.5 *205 - + 5 C - *271 AMP 20 271 328 *328 AMP - 15 *149 - - 40 *322 - - 60 188 E542K - 30 267 E545K/AMP + 30 356 E542K/AMP + 65 149 322 410 E545K/AMP + 20 *422 E545K/AMP + 30 *BKM120 Resistant, (+) = positive, (−) = loss, AMP = amplification D 267 356 188 410 422

Figure 3. Phospho-S6 expression in LUSC PDX. A, pS6 protein expression level was assessed by percent-positive cells via IHC. PIK3CA mutants exhibited stronger pS6 staining than in models with wild-type PI3K and PTEN loss. Representative pS6 IHC images in (B) PIK3CA wild-type models, (C) PTEN loss, and (D) PIK3CA mutants. Scale bar, 300 mm.

the four responders, copy-number variation analysis and valida- (PHLC205), but the effect was greater in the PIK3CA mutants tion by FISH were performed (Supplementary Table S1). Two of (Fig. 5A). Experimental expansion to 5 to 8 mice per arm con- the four responders harbored 100% homozygous deletion of firmed the greater antitumor effects of the combination treatment CDKN2A, whereas the other two harbored heterozygous deletions in three PIK3CA mutants but not in the two PI3K wild-type PDXs of CDKN2A. p16 and Rb1 protein expressions were further (Fig. 5B–F). Furthermore, the effect of combination treatment at validated by IHC, which confirmed complete absence of p16 half doses for BKM120 and palbociclib was superior to the staining and presence of Rb1 staining in the responders (Supple- antitumor effects of single agents at full dose in PIK3CA mutants mentary Fig. S4C). (Supplementary Fig. S6). Combination of BKM120 or BYL719 with a more specific CDK4/6 inhibitor abemaciclib was also Combined PI3K and CDK4/6 inhibition elicits tumor effective in suppressing the growth of PIK3CA mutants (Supple- regression in PIK3CA-mutant LUSC PDXs mentary Fig. S7). PI3K is known in multiple cancer types to function upstream of To verify targeted inhibition of BKM120 and palbociclib, a CDK4/6 and thus potentially could regulate its activity through pharmacodynamics study was performed in three PIK3CA cyclin D1 (38). p16 functions as a suppressor of CDK4/6, and loss mutants. The mice were dosed with either monotherapies or of this protein could result in greater activation of the cell cycle. combination therapy, and the tumors were harvested at three Through our gene expression analysis, p16 loss was identified in time points (2, 8, and 24 hours). Targeted inhibition of pAkt and the four BKM120 responders. Due to our findings of p16 loss in pS6 was observed in all models with single BKM120 treatment at responders and prior reports of PI3K promoting CDK4/6 activity 2 and 8 hours after dosing with a rebound at 24 hours, whereas (39), we hypothesized that LUSC PDXs would be especially interestingly, this rebound 24 hours after dosing appeared dimin- sensitive to the combination of PI3K and CDK4/6 inhibition. ished in the combination setting (Fig. 6A). Although we have Utilizing the n ¼ 1 drug screening method, we tested BKM120 (at made this observation in our PDX models, whether inhibition of a reduced dose) either alone or in combination with palbociclib, a the cell cycle together with the PI3K pathway enhances suppres- CDK4/6 inhibitor, in both PIK3CA mutant with p16 loss and PI3K sion of pAkt and pS6 remains to be an area of investigation. In wild-type PDX models. To assess the toxicity of combining PI3K addition, with single-agent palbociclib treatment and in combi- inhibitors with CDK4/6 inhibitors, mouse weight during combination with BKM120, pRb was suppressed in all of the models nation treatment was analyzed. The combination of BKM120 and (Fig. 6A). Thus, our results suggest that targeted inhibition with palbociclib was well tolerated in mice, and no adverse events were BKM120 and palbociclib treatment was achieved, with enhanced noted (Supplementary Fig. S5). Our results demonstrated greater inhibition of PI3K downstream targets observed in the combi- efficacy of combination therapy (at half doses) compared with nation setting (Fig. 6B). Overall, these findings provide evidence single-agent BKM120 and Palbociclib alone in both PIK3CA that CDK4/6 inhibitors in combination with PI3K inhibitors may mutants (PHLC356, 410, and 188) and a wild-type PDX improve LUSC response to PI3K monotherapy.

www.aacrjournals.org Clin Cancer Res; 24(23) December 1, 2018 5995

Shi et al.

A Color Key

−2 −1 0 1 2 Row Z−Score

CDKN2A IRF6 TGFB1 Cell-cycle arrest TBRG4 BNIP1 FASTK MKNK2 NUAK2 Apoptosis STEAP3 NOL3 LMNA OCLN TGFB1 Cell−cell junction TBX19 TBX3 YY2 BPTF Figure 4. HSF5 Transcription HEY2 Differential gene expression analysis NFIB ZNF234 between 9 PI3K inhibitor responders and ZNF33B nonresponders. A, Differential gene ACVRL1 ACVR2A BMP binding expression analysis was performed in LUSC ABCA10 BMP2K PDXs, and 279 upregulated and N4BP2 downregulated genes were identiﬁed. The NLRP2 WNK4 expression of 38 of these genes was ASCC3 visualized on a heatmap between BKM120 ACSL3 DNAH17 responders and nonresponders. ETNK1 ATP binding LATS1 R, responders; NR, nonresponders. B, The MAP3K4 gene list was submitted to DAVID to MSH4 PIK3C2A identify gene sets enriched in cell-cycle RIMKLB arrest, apoptosis, cell–cell junction, transcription, ATP binding, and BMP – P 149 422 328 356 188 267 410 322 271 binding as shown by their log2 ( value). CDKN2A encoding for p16 was identiﬁed to PIK3CA mut be downregulated in the responders. PTEN loss R R R R NR NR NR NR NR BKM120 sensitivity

B 15 Downregulated genes Upregulated genes

10 ( P value) 2 5 − Log

0 g

Apoptosis cell junction Transcription ATP binding BMP signalin Cell-cycle arrestCell−

Discussion better to this combination than PI3K wild-type PDXs, and targeted inhibition of the combination therapy was achieved In this study, we have demonstrated that PIK3CA mutation in representative PDX models. Our ﬁndings thus provide the predicts response to PI3K inhibitors BKM120 and BYL719 in our rationale for using the combination of PI3K and CDK4/6 inhi- large cohort of chemotherapy-na€ve LUSC PDX models. Although bitors in PIK3CA-mutant LUSC patients in view of the poor PI3K inhibitor monotherapy is effective in PIK3CA mutants, response rate observed in the single-agent PI3K inhibitor combining PI3K inhibitors with CDK4/6 inhibitors can lead to therapy trial. greater antitumor effects than monotherapy alone. We also dem- Targeted therapies that have demonstrated efﬁcacy in cell lines onstrated that PIK3CA-mutant PDXs with p16 loss responded often fail to progress through clinical trials. This discordance

5996 Clin Cancer Res; 24(23) December 1, 2018 Clinical Cancer Research

Dual Targeting of PI3K and CDK4/6 in PIK3CA-Mutant LUSC

A Vehicle Palbociclib B BKM120 Combination

1.0 PHLC356 Vehicle )

3 BKM120 0.5 1,000 Palbociclib Combination 0.0 500 Treat -0.5 PIK3CA (normalized to vehicle) to (normalized MUT MUT MUT MUT WT volume (mm Tumor ** Fold change in tumor volume Fold change -1.0 0 0204060 356 267 410 188 205 PHLC Day C D PHLC410 PHLC188 1,000 1,000 ) ) 3 3 800 800

600 600 Treat Treat 400 400

200 *** 200 Tumor volume (mm Tumor volume (mm 0 0 ** 0204060 020406080 Day Day EF 1,500 PHLC205 2,000 PHLC152 ) ) 3 3 1,500 1,000 Treat 1,000 * 500 Treat 500 * Tumor volume (mm Tumor volume (mm 0 0 0 20406080 010203040 Day Day

Figure 5. Combination of BKM120 and palbociclib treatment in PIK3CA wild-type and mutated LUSC PDX. A, Waterfall plot of monotherapy and combination therapy of BKM120 (25 mg/kg) and palbociclib (75 mg/kg) in 4 models with PIK3CA mutation and p16 loss and one model with wild-type PIK3CA. Change in tumor volume (relative to vehicle) was determined by calculating the growth rate from days 0 and 20 relative to the vehicle group. B–F, Combination treatment validation in 3 PIK3CA mutants and 2 PIK3CA wild-type PDXs. Five to 8 mice per arm were used for this study. , P > 0.05; , P < 0.01; and , P < 0.001 compared with BKM120 single treatment. potentially could be explained by the lack of clinically relevant PIK3CA mutation. Overlaps of PI3K pathway alterations, as was models amenable to drug testing. We utilized PDXs as preclinical observed in our LUSC PDX models such as PIK3CA mutations models of LUSC for drug testing because the molecular profiles of with amplifications, or mutations with PTEN loss, are common the PDXs are known to reflect those of their patient tumors (40). given that PIK3CA amplification occurs in over 60% of LUSC Our finding that PIK3CA mutation predicts BKM120 sensitivity in cases, and PTEN loss and PIK3CA mutation occur in 20% to 30% chemo-na€ve LUSC PDXs suggests that PI3K inhibitors may be and 10% to 15% of cases, respectively (3). This suggests that potential therapeutic options for those whose tumors harbor patients with tumors harboring overlaps of PI3K pathway

www.aacrjournals.org Clin Cancer Res; 24(23) December 1, 2018 5997

Shi et al.

A PHLC410 PHLC267 PHLC356 BKM120 Palbociclib Combo BKM120 Palbociclib Combo BKM120 Palbociclib Combo

0 2 8 24 2 8 24 2 8 24 0 2 8 24 2 8 24 2 8 24 0 2 8 24 2 8 24 2 8 24 Hours pRb (S780) Rb pAkt (S473) Akt pS6 (S235/236)

β-Actin

B BKM120 Palbociclib Combination

BKM120 BKM120 *PI3K *PI3K *PI3K

Akt Akt Akt

mTOR Palbociclib mTOR Palbociclib mTOR

p16 Cyclin D1 p16 Cyclin D1 p16 Cyclin D1 CDK4/6 CDK4/6 CDK4/6

pRb pRb pRb p pRb p pRb p pRb 9:;p pRb 9:;p pRb p9:; p pRb p E2F E2F

Figure 6. Pharmacodynamics (PD) study of BKM120 and palbociclib combination treatment in LUSC PDX. A, A total of 10 mice were used per model for the PD analysis. Mice were treated either with vehicle, BKM120, or palbociclib alone, or the combination of the two drugs for 0, 2, 8, and 24 hours. Tumors were harvested at the end of the time points and snap-frozen for protein isolation. Western blot was used to determine expression of selected proteins. B, Schematic of the mechanism of inhibition with either BKM120, palbociclib, or combination treatment in PIK3CA-mutated PDX.

alterations could potentially have greater activation of the PI3K to understand why PIK3CA amplification does not exhibit the pathway and respond differently to PI3K inhibitors. Thus, careful same level of PI3K activation, as indicated by pS6 expression, evaluation for this subset of patients should be taken into con- compared with PIK3CA mutation. sideration when selecting patients for clinical trials of PI3K This is the first preclinical report investigating the efficacy of inhibitor testing. combined PI3K and CDK4/6 inhibitors in LUSC. Although we In addition, our study suggests that PDXs harboring PIK3CA have shown in our initial BKM120 screen that the drug is effective mutation may exhibit greater PI3K pathway activation, as in PIK3CA-mutant LUSC, tumor regression was not observed. This indicated by pS6 protein levels. Phospho-S6, a downstream target result is consistent with the findings in the BASALT-1 trial in which of PI3K/mTOR pathway, was shown to be particularly enriched in 5 of the 6 patients with PIK3CA mutation showed tumor size PDXs harboring PIK3CA mutations in our 12 PDX models eval- reduction, although not sufficient to reach partial response. uated for BKM120 sensitivity. Thus, it would be worthy to Through our differential gene expression analysis, we identified investigate in a large cohort study whether tumors harboring high p16 loss in the majority of our PIK3CA mutants. p16 is an levels of pS6 would respond to PI3K inhibitors. Furthermore, a inhibitor of CDK4/6, and loss of expression results in cell-cycle large-scale proteomics study has profiled pS6 and pAkt levels in progression. Clinically, CDK4/6 inhibitors palbociclib and abe- >10,000 cancer samples and correlated their expression with PI3K maciclib as single agents are currently being evaluated in pathway status (32). Samples with PIK3CA mutation were shown advanced-stage LUSC (NCT02664935, NCT02785939, and to harbor high pS6 and pAkt protein levels, whereas no such NCT02450539; ref. 41). Based on our findings of p16 loss in the conclusions were made in tumors with PIK3CA amplification and responding PDXs, we posited that LUSC driven by PIK3CA muta- wild-type status. Although PIK3CA copy number correlates well tion and p16 loss may benefit from the combination of PI3K and with mRNA expression (3, 32), further investigations are needed CDK4/6 inhibitors. We tested the combination of PI3K and

5998 Clin Cancer Res; 24(23) December 1, 2018 Clinical Cancer Research

Dual Targeting of PI3K and CDK4/6 in PIK3CA-Mutant LUSC

CDK4/6 inhibitors in PIK3CA mutants with p16 loss and found Combining therapies with clinically existing or novel therapies at that all these models were sensitive to combination treatment, reduced dosages while optimizing efficacy in biomarker selected whereas the PI3K wild-type PDXs were resistant. This shows that patients may overcome these challenges. In the present preclinical PIK3CA mutation may predict response to combination therapy, study, we identified the combination of PI3K and CDK4/6 and whether p16 status affects sensitivity to treatment requires inhibitors as a viable therapeutic option in PIK3CA-mutant LUSC. further investigation with more preclinical models. Currently, Future investigations of PI3K and CDK4/6 inhibitor combination LUSC patients are being recruited in a phase I clinical trial warrant clinical testing in this disease setting. of the PI3K/mTOR inhibitors gedatolisib and palbociclib (NCT03065062); assessment of the role of PIK3CA mutation and Disclosure of Potential Conflicts of Interest p16 loss may be warranted. In terms of the mechanism of No potential conflicts of interest were disclosed. sensitivity to combination treatment, one proposal is that the PI3K pathway positively regulates cyclin D1 expression through Authors' Contributions b b downregulation of glycogen synthase-3 (GSK3 ), which would Conception and design: R. Shi, M.-S. Tsao normally degrade cyclin D1 via ubiquitination (42). Cancer cells Development of methodology: R. Shi, M. Li, N.-A. Pham, M.-S. Tsao dependent on PI3K activation could potentially upregulate cyclin Acquisition of data (provided animals, acquired and managed patients, D1 and feeds into the cell-cycle pathway. Thus, inhibition of both provided facilities, etc.): R. Shi, M.-S. Tsao PI3K and CDK4/6 could potentially result in greater suppression Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): R. Shi, V. Raghavan, S. Tam, M. Cabanero, N. Moghal, of this pathway. M.-S. Tsao Hyperglycemic toxicities associated with PI3K inhibitors, par- Writing, review, and/or revision of the manuscript: R. Shi, V. Raghavan, ticularly with BKM120, have been well documented in clinical S. Tam, M. Cabanero, N.-A. Pham, F.A. Shepherd, N. Moghal, M.-S. Tsao trials (43, 44). In fact, drug-related toxicity was also one of the Administrative, technical, or material support (i.e., reporting or organizing reasons for the failure of the BASALT-1 trial. Although MTDs can data, constructing databases): R. Shi, N.-A. Pham, F.A. Shepherd be adjusted to mitigate toxicity, the optimal doses required to Study supervision: M.-S. Tsao achieve antitumor effects of these compounds may not be Acknowledgments achieved. Thus, our study proposes the use of combination We would like to thank Jing Xu, Wendy So, and Jian Zhou for all IHC staining therapy of PI3K and CDK4/6 inhibitors for the purpose of and Olga Ludkovski for the FISH analysis. We thank Sebastiao Martins and Lili Ji reducing the dosage of both compounds while maximizing for PTEN IHC interpretation. We acknowledge Dr. Mamatjan Yasin for the antitumor efficacy. In our in vivo drug studies, we reduced single methylation analysis. We acknowledge Dr. Quan Li for preparing the drug doses by half in the combination setting compared with the gene expression files for GEO submission. We thank Dr. Vuk Stambolic and monotherapy setting. In addition, due to the nonoverlapping Dr. Trevor Pugh for advising and overseeing the project. This work was supported by the Canadian Institute of Health Research toxicities of PI3K and CDK4/6 inhibitors in which the former is Foundation grant FDN-148395, the Canadian Cancer Society Research Institute associated with hyperglycemia whereas the latter is associated grant 701595, and the Ontario Ministry of Long Term Health. R. Shi is funded by with hematologic effects (39), the combination of these inhibitors a University of Toronto Ontario Student Opportunity Trust Fund and Ontario could potentially be tolerated in LUSC patients. Careful moni- Graduate Scholarship. M. Cabanero is supported by the Terry Fox Foundation toring of drug combination doses would be necessary to control Training Program in Molecular Pathology of Cancer at CIHR (STP 53912). for toxicity should this combination be implemented in the clinic. Dr. F.A. Shepherd is the Scott Taylor Chair in Lung Cancer Research. Dr.M.-S.TsaoistheM.QasimChoksiChair in Lung Cancer Translational Suitable MTDs for this combination will have to be determined in Research. phase I trials of PI3K and CDK4/6 inhibitors in LUSC. Results from preclinical and clinical studies of PI3K inhibitors fi The costs of publication of this article were defrayed in part by the in lung cancer have varied in terms of ef cacy, particularly the lack payment of page charges. This article must therefore be hereby marked of translatability to the clinic. Early preclinical reports of PI3K advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate inhibitors have been efficacious in PIK3CA-mutant cell lines this fact. (45, 46, 47, 48), but efficacy in LUSC patients has been modest. The lack of predictive biomarkers and high toxicity profiles of Received March 2, 2018; revised July 18, 2018; accepted July 30, 2018; compounds targeting LUSC are the current challenges in this field. published first August 9, 2018.

References 1. Cruz C, Tanoue LT, Matthay RA. Lung cancer: epidemiology, etiology, and 6. Dogruluk T, Tsang YH, Espitia M, Chen F, Chen T, Chong Z, et al. prevention. Clin Chest Med 2011;32:605–44. Identification of variant-specific functions of PIK3CA by rapid phenotyping 2. Perez-Moreno P, Brambilla E, Thomas R, Soria JC. Squamous cell carci- of rare mutations. Cancer Res 2015;75:5341–54. noma of the lung: molecular subtypes and therapeutic opportunities. Clin 7. Maira SM, Pecchi S, Huang A, Burger M, Knapp M, Sterker D, et al. Cancer Res 2012;18:2443–51. Identification and characterization of NVP-BKM120, an orally available 3. TCGA. Comprehensive genomic characterization of squamous cell lung pan-class I PI3-kinase inhibitor. Mol Cancer Ther 2012;11:317–28. cancers. Nature 2012;489:519–25. 8. Fritsch C, Huang A, Chatenay-Rivauday C, Schnell C, Reddy A, Liu M, et al. 4. Hammerman PS, Sos ML, Ramos AH, Xu C, Dutt A, Zhou W, et al. Characterization of the novel and specific PI3Ka inhibitor NVP-BYL719 Mutations in the DDR2 kinase gene identify a novel therapeutic and development of the patient stratification strategy for clinical trials. Mol target in squamous cell lung cancer. Cancer Discov 2011;1: Cancer Ther 2014;13:1117–29. 78–89. 9. Lee JJ, Loh K, Yap Y. PI3K/Akt/mTOR inhibitors in breast cancer. 5. Thorpe LM, Yuzugullu H, Zhao JJ. PI3K in cancer: divergent roles of Cancer Biol Med 2015;12:342–54. isoforms, modes of activation and therapeutic targeting. Nature Rev 10. Paik PK, Shen R, Won H, Rekhtman N, Wang L, Sima CS, et al. Next- 2015;15:7–24. generation sequencing of stage IV squamous cell lung cancers reveals an

www.aacrjournals.org Clin Cancer Res; 24(23) December 1, 2018 5999

Shi et al.

association of PI3K aberrations and evidence of clonal heterogeneity in 31. Gabelli SB, Echeverria I, Alexander M, Doung-Ly KC, Chaves-Moreira D, patients with brain metastases. Cancer Discov 2015;5:610–21. Brower ET, et al. Activation of PI3Ka by physiological effectors and by 11. Vansteenkiste JF, Canon J, De Braud F, Grossi F, De Pas T, Gray JE, et al. oncogenic mutations: structural and dynamic effects. Biophys Rev 2014; Safety and efficacy of buparlisib (BKM120) in patients with PI3K pathway- 6:89–95. activated non-small cell lung cancer: results from the phase II BASALT-1 32.ZhangY,NgPK,KucherlapatiM,ChenF,LiuY,TsangYH,etal. study. JTO 2016;10:1319–27. A pan-cancer proteogenomic atlas of PI3K/AKT/mTOR pathway 12. Wang D, Pham NA, Tong J, Sakashita S, Allo G, Kim L, et al. Molecular alterations. Cancer Cell 2017;31:820–32. heterogeneity of non-small cell lung carcinoma patient-derived xenografts 33. Gonzalez-Mariscal L, Tapia R, Chamorro D. Crosstalk of tight junction closely reflect their primary tumors. Int J Cancer 2017;140:662–73. components with signaling pathways. BBA 2008;1778:729–56. 13. Conway T, Wazny J, Bromage A, Tymms M, Sooraj D, Willams ED, et al. 34. Franke TF, Hornik CP, Segev L, Shostak GA, Sugimoto C. PI3K/Akt and Xenome–a tool for classifying reads from xenograft samples. Bioinformat- apoptosis: size matters. Oncogene 2003;22:8983–98. ics 2012;28:172–8. 35. Sterlacci W, Tzankov A, Veits L, Zelger B, Bihl MP, Foerster A, et al. 14. Li H, Durbin R. Fast and accurate short read alignment with Burrows- A comprehensive analysis of p16 expression, gene status, and promoter Wheeler transform. Bioinformatics 2009;25:1754–60. hypermethylation in surgically resected non-small cell lung carcinomas. 15. Cibulskis K, Lawrence MS, Carter SL, Sivachenko A, Jaffe D, Sougnez C, et al. JTO 2011;6:1649–57. Sensitive detection of somatic point mutations in impure and heteroge- 36. McWilliams RR, Wieben ED, Rabe KG, Pedersen KS, Wu Y, Sicotte H, et al. neous cancer samples. Nat Biotechnol 2013;31:213–9. Prevalence of CDKN2A mutations in pancreatic cancer patients: implica- 16. Koboldt DC, Zhang Q, Larson DE, Shen D, McLellan MD, Lin L, et al. tions for genetic counseling. Eur J Hum Genet 2011;19:472–8. VarScan 2: somatic mutation and copy number alteration discovery in 37. TamKW,ZhangW,SohJ,StastnyV,ChenM,SunH,etal.CDKN2A/ cancer by exome sequencing. Genome Res 2012;22:568–76. p16 inactivation mechanisms and their relationship to smoke expo- 17. Sherry ST, Ward MH, Kholodov M, Baker J, Phan L, Smigielski EM, et al. sure and molecular features in non-small cell lung cancer. JTO 2013; dbSNP: the NCBI database of genetic variation. Nucleic Acids Res 8:1378–88. 2001;29:308–11. 38. Averous J, Fonseca BD, Proud CG. Regulation of cyclin D1 expression by 18. Lek M, Karczewski KJ, Minikel EV, Samocha KE, Banks E, Fennell T, et al. mTORC1 signaling requires eukaryotic initiation factor 4E-binding protein Analysis of protein-coding genetic variation in 60,706 humans. Nature 1. Oncogene 2008;27:1106–13. 2016;536:285–91. 39. Vora SR, Juric D, Kim N, Mino-Kenudson M, Huynh T, Costa C, et al. CDK 19. Wang K, Li M, Hakonarson H. ANNOVAR: functional annotation of genetic 4/6 inhibitors sensitize PIK3CA mutant breast cancer to PI3K inhibitors. variants from high-throughput sequencing data. Nucleic Acids Res Cancer Cell 2014;26:136–49. 2010;38:e164. 40.GandaraDR,MackPC,BultC,LiT,LaraPN,RiessJW,etal. 20. Loo PV, Nordgard SH, Lingjaerde OC, Russnes HG, Rye IH, Sun W, et al. Bridging tumor genomics to patient outcomes through an integrated Allele-specific copy number analysis of tumors. PNAS 2010;107:16910–5. patient-derived xenograft (PDX) platform. Clin Lung Cancer 2015;16: 21. Beroukhim R, Getz G, Nghiemphu L, Barretina J, Hsueh T, Linhart D, et al. 165–72. Assessing the significance of chromosomal aberrations in cancer: meth- 41. Patnaik A, Rosen LS, Tolaney SM, Tolcher AW, Goldman JW, Gandhi L, odology and application to glioma. PNAS 2007;104:20007–12. et al. Efficacy and safety of abemaciclib, an inhibitor of CDK4 and CDK6, 22. Yanagawa N, Leduc C, Kohler D, Saieg MA, John T, Skyes J, et al. Loss of for patients with breast cancer, non-small cell lung cancer, and other solid phosphatase and tensin homolog protein expression is an independent tumors. Cancer Discov 2016;6:740–53. poor prognostic marker in lung adenocarcinoma. JTO 2012;7:1513–21. 42. Diehl JA, Cheng M, Roussel MF, Sherr CJ. Glycogen synthase kinase-3 23. Gao H, Korn JM, Ferretti S, Monahan JE, Wang Y, Singh M, et al. High- regulates cyclin D1 proteolysis and subcellular localization. Genes Dev throughput screening using patient-derived tumor xenografts to predict 1998;12:3499–511. clinical trial drug response. Nature Med 2015;21:1318–25. 43. Dreyling M, Cunningham D, Bouabdallah K, Assouline S, Van den Neste E, 24. Du P, Kibbe WA, Lin SM. lumi: a pipeline for processing Illumina micro- Vitolo U, et al. Phase 2A study of copanlisib, a novel PI3K inhibitor, in array. Bioinformatics 2008;24:1547–8. patients with indolent lymphoma. Blood 2014;124:1701. 25. Gentleman RC, Carey VJ, Bates DM, Bolstad B, Dettling M, Dudoit S, et al. 44. Younes A, Salles G, Martinelli G, Bociek G, Barrigon DC, Barca EG, et al. An Bioconductor: open software development for computational biology and open-label phase II study of buparlisib (BKM120) in patients with relapsed bioinformatics. Genome Biol 2004;5:R80. and refractory diffuse large B-cell lymphoma, mantle cell lymphoma or 26. Johnson WE, Li C, Rabinovic A. Adjusting batch effects in microarray follicular lymphoma. Blood 2014;124:1718. expression data using empirical Bayes methods. Biostatistics 2007;8: 45. Furet P, Guagnano V, Fairhurst RA, Imbach-Weese P, Bruce I, Knapp M, 118–27. et al. Discovery of NVP-BYL719 a potent and selective phosphatidylino- 27. Leek JT, Johnson WE, Parker HS, Jaffe AE, Storey JD. The sva package for sitol-3 kinase alpha inhibitor selected for clinical evaluation. Bioorg Med removing batch effects and other unwanted variation in high-throughput Chem Lett 2013;23:3741–48. experiments. Bioinformatics 2012;28:882–3. 46. Spoerke JM, O'Brien C, Huw L, Koeppen H, Fridlyand J, Brachmann RK, 28. Ritchie ME, Phipson B, Wu D, Hu Y, Law CW, Shi W, et al. limma powers et al. Phosphoinositide 3-kinase (PI3K) pathway alterations are asso- differential expression analyses for RNA-sequencing and microarray ciated with histologic subtypes and are predictive of sensitivity to PI3K studies. Nucleic Acids Res 2015;43:e47. inhibitors in lung cancer preclinical models. Clin Cancer Res 2012; 29. Huang DW, Sherman BT, Tan Q, Collins JR, Alvord WG, Roayaei J, et al. 18:6771–83. The DAVID gene functional classification tool: a novel biological module- 47. O'Brien C, Wallin JJ, Sampath D, GuhaThakurta D, Savage H, Punnoose EA, centric algorithm to functionally analyze large gene lists. Genome Biol et al. Predictive biomarkers of sensitivity to the phosphatidylinositol 30 2007;8:R183. kinase inhibitor GDC-0941 in breast cancer preclinical models. Clin 30. John T, Kohler D, Pintilie M, Yanagawa N, Pham N, Li M, et al. The ability to Cancer Res 2010;16:3670–83. form primary tumor xenografts is predictive of increased risk of disease 48. Weigelt B, Warne PH, Downward J. PIK3CA mutation, but not PTEN loss of recurrence in early-stage non–small cell lung cancer. Clin Cancer Res function, determines the sensitivity of breast cancer cells to mTOR inhib- 2009;17:134–41. itory drugs. Oncogene 2011;30:3222–33.

6000 Clin Cancer Res; 24(23) December 1, 2018 Clinical Cancer Research

Targeting the CDK4/6-Rb Pathway Enhances Response to PI3K Inhibition in PIK3CA-Mutant Lung Squamous Cell Carcinoma

Ruoshi Shi, Ming Li, Vibha Raghavan, et al.

Clin Cancer Res 2018;24:5990-6000. Published OnlineFirst August 9, 2018.

Updated version Access the most recent version of this article at: doi:10.1158/1078-0432.CCR-18-0717

Supplementary Access the most recent supplemental material at: Material http://clincancerres.aacrjournals.org/content/suppl/2018/08/09/1078-0432.CCR-18-0717.DC1

Cited articles This article cites 48 articles, 14 of which you can access for free at: http://clincancerres.aacrjournals.org/content/24/23/5990.full#ref-list-1

Citing articles This article has been cited by 1 HighWire-hosted articles. Access the articles at: http://clincancerres.aacrjournals.org/content/24/23/5990.full#related-urls

E-mail alerts Sign up to receive free email-alerts related to this article or journal.

Reprints and To order reprints of this article or to subscribe to the journal, contact the AACR Publications Department at Subscriptions [email protected].

Permissions To request permission to re-use all or part of this article, use this link http://clincancerres.aacrjournals.org/content/24/23/5990. Click on "Request Permissions" which will take you to the Copyright Clearance Center's (CCC) Rightslink site.