Inhibition of mTORC1/2 and DNA-PK via CC-115 Synergizes with Carboplatin and Paclitaxel in Lung Squamous Cell Carcinoma

Abstract

Only a small percentage (< 1%) of patients with late-stage lung squamous cell carcinoma (LUSC) are eligible for targeted therapy. Because PI3K/AKT/mTOR signaling, particularly PIK3CA, is dysregulated in two-thirds of LUSC, and DNA damage response pathways are enriched in LUSC, we tested whether CC-115, a dual mTORC1/2 and DNA-PK inhibitor, sensitizes LUSC to chemotherapy. We demonstrate that CC-115 synergizes with carboplatin in 6 of 14 NSCLC cell lines, primarily PIK3CA-mutant LUSC. Synergy was more common in cell lines that had decreased basal levels of activated AKT and DNA-PK, evidenced by reduced P-S473-AKT, P-Th308-AKT, and P-S2056-DNA-PKcs. CC-115 sensitized LUSC to carboplatin by inhibiting chemotherapy-induced AKT activation and maintaining apoptosis, particularly in PIK3CA-mutant cells lacking wild-type TP53. In addition, pathway analysis revealed that enrichments in the interferon (IFN)α and IFNγ pathways were significantly associated with synergy. In multiple LUSC patient-derived xenograft and cell line tumor models, CC-115 plus platinum-based doublet chemotherapy significantly inhibited tumor growth and increased overall survival as compared to either treatment alone at clinically relevant dosing schedules. Immunohistochemistry and immunoblot analysis of CC-115-treated tumors demonstrated decreased P-Th308-AKT, P-S473-AKT, P-S235/236-S6, and P-S2056-DNA-PKcs, showing direct pharmacodynamic evidence of inhibited PI3K/AKT/mTOR signaling cascades. Because PI3K pathway and DNA-PK inhibitors have shown toxicity in clinical trials, we assessed toxicity by examining weight and numerous organs in PRKDC-wild type mice, which demonstrated that the combination treatment does not exacerbate the clinically accepted side effects of standard-of care-chemotherapy. This preclinical study provides strong support for the further investigation of CC-115 plus chemotherapy in LUSC.

Introduction

Despite the advances in targeted and immunotherapy in non-small cell lung cancer (NSCLC) treatment, the overall 5-year survival rate of patients diagnosed with advanced disease remains below 6% (1). NSCLC is the major contributor to lung cancer cases around the world and can be further divided based on histological features, including lung squamous cell carcinoma (LUSC; ~30%) and lung adenocarcinoma (LUAD; ~50%), although some tumors fall into more than one histological category (2,3). Not only are these NSCLC subtypes histologically distinct, but they also harbor unique molecular changes that lead to their malignant transformation. LUAD pathogenesis has been extensively studied, and driver events are well-documented with a high prevalence of mutations in EGFR (11%), KRAS (32%), and BRAF (7%), as well as fusions involving several genes including ALK (1.3%), ROS1 (1.7%) and RET (0.9%) (4,5). LUSC, on the other hand, is host to frequent amplifications of numerous genes (PIK3CA, CCND 1–3, CDK4, FGFR 1–3, MET, PDGFRA, SOX2), point mutations, and gene fusions, often in combination, but has fewer actionable driver alterations, despite a higher mutational burden when compared to LUAD (6–8).

Several common gene alterations in LUSC lead to constitutive activation of the PI3K/AKT/mTOR pathway, making it an attractive target for cancer therapy (9). Phosphatidylinositol 3-kinase CA (PIK3CA), one of the most commonly amplified (~47%) and mutated (10–16%) genes in LUSC, encodes isoform-α of the class IA PI3K catalytic subunit, p110α (7,10–12). Activating mutations in PIK3CA cluster around two hotspots, amino acids E542K or E545K in the helical domain, and amino acid H1047R in the kinase domain of the catalytic subunit (13,14). Several important pathways directly converge on proteins downstream of PI3K, including AKT, mTORC1 and mTORC2. Inactivating mutations in PI3K’s main negative regulator, PTEN, or activating mutations in its downstream signaling molecule, AKT, also occur. In total, it has been estimated that up to 69% of LUSC have aberrant activation of the PI3K/AKT/mTOR/RAS axis, leading to altered metabolism and increased proliferation, protein synthesis, evasion of apoptosis, cell migration and invasion (8,15).

Although numerous genetic drivers of each histological subtype of NSCLC have been identified, there is a disparity in the number of available targeted therapies for LUSC compared to LUAD. Currently, nearly 25% of LUAD patients benefit from targeted therapy, while only a mere 2% of LUSC patients, those with BRAF V600E or KRAS G12C oncogenic driver mutations, as well as other LUAD-associated mutations that are rare in LUSC, are eligible for targeted therapy (8,16). Because of this, cytotoxic platinum-based doublet chemotherapy, and/or, immune checkpoint inhibition, remain the standard of care in late-stage LUSC (17). These treatment disparities, along with other clinicopathologic characteristics of LUSC translate to a median survival time 30% shorter than other NSCLC histological subtypes (18). Given the frequent alterations resulting in PI3K/AKT/mTOR pathway activation in LUSC, there have been immense efforts in the last few years to target it therapeutically (19). Although these studies offered promise, they have yet to result in the approval of any PI3K/AKT/mTOR targeted agents in LUSC (10,20,21).

Because platinum-based chemotherapies and immune therapies remain the standard of care in LUSC, numerous groups have devoted efforts to understanding DNA damage response and repair (DDR) in NSCLC (22–24). A recent study revealed a significant correlation between DDR pathways and LUSC, in addition to 10.8% of LUSC patients in TCGA harboring mutations in DDR genes, perhaps revealing a therapeutic vulnerability (25). DNA-PK, a holoenzyme formed by DNA-PKcs, Ku70 and Ku80, is responsible for the initiation of non-homologous end joining (NHEJ) repair following DNA double-strand breaks (DSBs) and is increasingly being implicated in the process of carcinogenesis of many organs (26–28). Combination of DDR-inhibitors with chemotherapies and radiation, and with other kinase inhibitors has been explored (29,30). Given the dependence on genotoxic therapies in LUSC, and the high number of oncogenic alterations in the PI3K/AKT/mTOR pathway, we investigated the chemosensitization capabilities of CC-115, a dual mTORC1/2 and DNA-PK inhibitor, in LUSC cell lines and patient-derived xenografts (PDXs).

Materials and Methods

Cell culture and reagents

All cell lines were purchased from ATCC, DSMZ, JCRB or MD Anderson (Minna Lab), except for TM00244, which was purchased as a PDX from Jackson Labs. We generated a stable cell line from TM00244 and confirmed that it retained the PIK3CA mutation. All cell lines were maintained at 37°C in 5% CO₂ in RPMI (Gibco; Catalog No. 11875–093), DMEM (Gibco; Catalog No. 11995–065), EMEM (ATCC; Catalog No. 30–2003), or McCoy’s 5A (ATCC; Catalog No. 30–2007) media and supplemented with 10% fetal bovine serum (FBS), 1% penicillin/streptomycin, and 1% glutamine, except for RERFLCSQ1 and LOU-NH91, which were supplemented with 10% or 20% heat-inactivated fetal bovine serum, respectively. Cells were used within 20 passages from thaw for experiments and were STR profiled and routinely tested for mycoplasma (Lonza MycoAlert; Catalog No. LT07–703) every 6 months. For in vitro use, CC-115 (Selleck Chemicals; Catalog No. S7891) was dissolved in DMSO and clinical grade carboplatin (Rutgers Cancer Institute of New Jersey Pharmacy) was diluted in media. The molecular structure of CC-115 was previously published (31).

In vitro cell viability assays

Cells were seeded in 96-well plates, with 3,600 – 6,000 cells per well, for 24 hours. MTS assays were performed, and combination index (CI) values for combined drug treatments were based on the method of Chou–Talalay and calculated using CalcuSyn, as we previously described (32). Cells were treated with serially diluted CC-115, carboplatin, or both compounds. Reported IC₅₀ and CI values are the average of two or more independent experiments. CI values < 0.7 were considered synergistic, with decreasing CI values indicating greater synergy (32).

Crystal violet colony formation assays

Cells were seeded at 800 cells per well in 6-well plates for 24 hours, and then treated with CC-115, carboplatin, or both for 72 hours at the indicated concentrations in triplicates. Media were replaced after 72 hours, and cells were incubated at 37°C in 5% CO₂ and monitored for 3 weeks. Plates were washed with PBS, fixed with 100% methanol for 30 minutes, and stained with 0.005% crystal violet (Sigma; Catalog No. C3886) in 25% methanol. Excess stain was removed, and plates were washed gently with deionized water and air dried before imaging on a Bio-Rad Chemidoc machine. Colonies were quantified using ImageJ and a minimum of six replicates were analyzed to assess total colony area against total well area.

Immunoblotting and antibodies

Snap-frozen tumors were subjected to cryogenic grinding using a CyroMill (Retsch). Resulting tumor lysate or whole cell pellets were lysed on ice in cold NETN buffer (0.5% NP-40, 1 mM EDTA, 150 mM NaCl and 20 mM Tris Buffer) supplemented with cOmplete, Mini EDTA-free protease inhibitor cocktail (Millipore Sigma; Catalog No. 11836170001) and phosphatase inhibitor cocktail 2 and 3 (Sigma Aldrich; Catalog No. P5726 and Catalog No. P0044), then sonicated (Branson 450 Sonifier). Protein was quantified using Bradford Protein Assay Dye Reagent (Bio-Rad; Catalog No. 5000006). Western blots were run on 15 – 20 μg protein per lane, and then imaged and quantified as we previously described (5,33). Antibodies used were anti-GAPDH (Millipore #MAB374), anti-Vinculin (Santa Cruz SC-73614), anti-DNA-PKcs (Santa Cruz SC-5282), anti-P-S2056-DNA-PKcs (Abcam ab18192), anti-AKT1 (Cell Signaling Technology (CST) #2938), anti-P-S473-AKT (CST #4060), anti-P-Th308-AKT (CST #4056), anti-P-S139-H2AX (AbCam #26350), anti-PARP1 (CST #9542), anti-P-S235/236-S6 (CST #4858) and anti-S6 (CST #2217).

Pathway Enrichment Analysis

Cell line data were acquired as CCLE RNAseq gene expression data (read count) (CCLE_DepMap_18Q2_RNAseq_reads_20180502.gct.txt) obtained from the DepMap portal (https://depmap.org/portal/). The expression data consisted of raw counts segregated for synergistic and non-synergistic cell lines. Gene Set Enrichment Analysis (GSEA) (34) was performed using the curated gene sets of the Molecular Signature Database v7.0. The gene lists of Hallmark pathways (h.all.v7.4.symbols.gmt), Biocarta pathways (c2.cp.biocarta.v7.4.symbols.gmt), Kegg pathways (c2.cp.kegg.v7.4.symbols.gmt), Reactome pathways (c2.cp.reactome.v7.4.symbols.gmt), and Wikipathways (c2.cp.wikipathways.v7.4.symbols.gmt) were used for GSEA ranked analysis following the standard procedure described by GSEA user guide (http://www.broadinstitute.org/gsea/doc/GSEAUserGuideFrame.html ). Significantly enriched terms with similar descriptions and functions were further grouped into distinct biological categories to better reflect the biology in context and top categories were schematically projected on the network of enriched terms.

In vivo compound preparation and administration

Lyophilized CC-115 was reconstituted weekly in 0.5% carboxymethyl cellulose with 0.25% Tween-80 and kept under constant stirring conditions at 4°C and delivered to mice via oral gavage on a 5 day-on and 2 day-off clinically relevant schedule. Carboplatin and paclitaxel were mixed at a constant ratio and diluted with sterile PBS, delivered once weekly via intra-peritoneal (IP) injection, at least 1 hour prior to CC-115 administration, on the first day of each week. Drug concentrations were determined from titration experiments of carboplatin plus paclitaxel or CC-115 in multiple PDXs. The dose which decreased tumor volume by approximately 50%, as compared to vehicle control, was used in the future combination experiments. Dose-scheduling graphic shown in Supplemental Figure S3A was created with BioRender.com.

Patient-derived xenograft models and in vivo studies

All animal studies were performed in accordance with the Institutional Animal Care and Use Committee (IACUC) guidelines of Rutgers University. Established PDX models (TM00233, J000094918) were purchased from Jackson Laboratory, arriving in NOD scid gamma (NSG) mice (NOD.Cg-Prkdc^scid Il2rg^tm1Wjl/SzJ; Jackson Laboratories; Catalog No. 005557). The T-050 PDX was previously generated in the Pine lab (32,35). PDX fragments (2mm³) were implanted, or cells were injected (2 × 10⁶) into one or both flanks of 6–8-week-old NSG mice with 50% matrigel (Corning; Catalog No. 356231). Treatment began when average tumor size reached ~150–250 mm³ and mice were randomized based on both tumor size and sex, ensuring equivalent numbers of male and females were used in every treatment group. Body weight was collected daily and tumor volume was blindly measured triweekly using a digital caliper and calculated using the following formula: V = (0.52)(width²)(length). Mice were euthanized once tumor size reached IACUC standards. Survival was defined as the time between first day of treatment and when the tumor volume reached > 550 mm³. For immunohistochemistry (IHC) analysis, formalin-fixed tumors were embedded in paraffin, stained with hematoxylin and eosin (H&E), and IHC was performed by the Rutgers Cancer Institute of New Jersey Biospecimen Repository and Histopathology Service Shared Resource (BRHS). For assessing any potential toxicity, FVB/NJ mice (Jackson Laboratories; Catalog No. 001800) underwent three weeks of treatment, then fresh blood was collected in EDTA-coated tubes (BD; Catalog No. 365974) from four mice per group and immediately subjected to CBC blood counts (HESKA Element HT5). Organs were paraffin-embedded, and tissue slides were subjected to H&E staining by Rutgers Cancer Institute of New Jersey BRHS. H&E slides were reviewed by a board-certified pathologist for signs of toxicity.

Statistical Analysis

All data were presented as the mean ± standard error of the mean (SEM). Graphs and statistical analyses were done as we previously described (32), except a Gehan-Breslow-Wilcoxon for Median Survival Time Analysis was done for the survival analyses. P-values < 0.05 were considered significant.

Data Availability

The data generated in this study are available within the article and its supplementary data files. The data utilized in the pathway enrichment analysis came from the publicly available database, DepMap (https://depmap.org/portal/).

Results

Selection of models testing CC-115 plus chemotherapy in pre-clinical studies

We selected a panel of 14 NSCLC lines, focusing primarily on LUSC (Table 1). The number of commercially available PIK3CA-mutant LUSC lines was a limiting factor; however, we obtained five for this study, two from the US (ATCC and John Minna, UTSW), one from Japan (JCRB), and one from Germany (DSMZ). We generated the fifth cell line from a PIK3CA-mutant PDX. The cell line from ATCC, H596, was derived from a PIK3CA-mutant adenosquamous tumor, though it is generally accepted to be LUSC based on molecular characteristics (36). We selected five PIK3CA-wild type (WT) LUSC cell lines, two of which had a PIK3CA copy number gain, as well as one PIK3CA-mutant and two PIK3CA-WT LUAD cell lines. The cell lines were also chosen to represent the heterogeneity observed across human LUSC tumors. For example, given the frequency of TP53 mutations in LUSC, our cell line panel had a high prevalence of alterations in TP53 (8 cell lines, 57%). Additionally, CDKN2A alterations were common, observed in 6 out of 14 lines. Other alterations included PTEN loss, and mutations in KRAS or other proteins in the PI3K pathway.

Table 1.

Defining the Sensitivity of NSCLC Lines to Carboplatin Alone, CC-115 alone, or in Combination.

Cell Line	Histology	PIK3CA	Other Alterations	Carboplatin IC50 (μM)	CC-115 IC50 (μM)	Combination Index (CI)
LOU-NH91	LUSC A	E726K	TP53	291.1	25.10	0.21
H596	LUSC	E545K	TP53	60.8	0.68	0.50
H1975	LUAD B	G118D	TP53, EGFR, TSC2 & CDKN2A	63.13	0.31	0.63
TM00244	LUSC	E545K	TP53, EGFR, CDKN2A, & PTEN	91.94	0.43	0.41
RERFLCSQ1	LUSC	E545K (CNGC)	RICTOR (CNG)	94	0.7	1.1
HCC2450	LUSC	H1047R (CNG)	TP53 & TSC2	20.68	0.58	1.2
H226	LUSC	WT D	PIK3R6 & CDKN2A	66.69	0.95	0.38
Calu-1	LUSC	WT	PIK3C2G (CNG) & KRAS	57.35	1.38	0.21
HCC827	LUAD	WT	EGFR	133.9	0.69	0.81
SK-MES-1	LUSC	WT	PIK3C2B, TP53, EGFR & CDKN2A	28.55	0.61	1.27
H1650	LUAD	WT	EGFR & CDKN2A	46.60	0.89	1.29
H1703	LUSC	WT	PIK3C2G, TP53 & CDKN2A	119	0.46	0.98
HCC95	LUSC	WT (CNG)	TSC2	65.15	0.85	0.95
HCC2814	LUSC	WT (CNG)	AKT2 (CNG), TP53 & PTEN	37.95	1.57	4.99

Bold indicates a synergistic CI value between compounds for that cell line

^ALUSC: Lung squamous cell carcinoma

^BLUAD: Lung adenocarcinoma

^CCNG: Copy number gain

^DWT: Wild-type

CC-115 is a potent DNA-PK/mTOR inhibitor and sensitizes LUSC cell lines to platinum-based chemotherapy

To test if CC-115 sensitizes NSCLC to carboplatin, we first determined the IC₅₀ values for carboplatin or CC-115 via MTS assays (Fig. 1A and B). We chose carboplatin because it is a standard-of-care drug used for LUSC, and it causes the formation of DSBs, which are repaired by NHEJ (37). Carboplatin has also been shown to induce phosphorylation of DNA-PKcs at S2056, indicative of activated DNA-PK (38,39). We observed a wide range of sensitivities to the individual agents, with IC₅₀ values ranging from ~ 20 – 291 μM for carboplatin and ~ 0.31 – 25 μM for CC-115 (Table 1). We confirmed that downstream signaling of mTORC1 and mTORC2 was inhibited following two hours of CC-115 treatment, demonstrated by decreased P-S473-AKT, P-S474-AKT2, and P-S6. As expected based on the mechanism of action (30), CC-115 also decreased P-S2056-DNA-PKcs in two of the three cell lines tested (Fig. 1C).

Figure 1. CC-115 is a potent mTORC1/2 and DNA-PK inhibitor and sensitizes LUSC cell lines to platinum-based chemotherapy.

A, Carboplatin and B, CC-115 IC₅₀ values for each cell line were assessed 72 hours after compound treatment via MTS assay. IC₅₀ values shown are the mean of two or more independent experiments ± SEM. C, Immunoblot showing that DNA-PK, mTORC1, and mTORC2 down-stream signaling were inhibited 2 hours after treatment with the targeted inhibitor, CC-115, at 0.3 μM or 1 μM. D, Experimental Combination Index (CI) values were calculated using the Chou-Talalay method via Calcusyn at the ED₅₀ of both carboplatin and CC-115 in 14 NSCLC lines. CI values < 0.7 (dotted line) were considered synergistic, where values closer to 0 demonstrate stronger synergy, and those above 1 show antagonism between the compounds. Reported CI values are the mean of two or more independent experiments ± SEM. E, Two representative 72-hour dose-response curves for the cell lines demonstrating the highest degree of synergy between carboplatin and CC-115 in LOU-NH91 and Calu-1 LUSC cell lines.

We observed synergy between CC-115 and carboplatin (CI < 0.7) in 6 of the 14 cell lines examined, four of which were LUSC with PIK3CA mutations and TP53 alterations (Fig. 1D, Table 1). At a less stringent cut-off of CI < 1.0, 64% of the cell lines (9 of 14) demonstrated synergy. The cell lines with the highest degree of synergy between carboplatin and mTOR/DNA-PK inhibition were LOU-NH91 and Calu-1 (Fig. 1E). There was no significant difference between the CI values in PIK3CA-WT or –mutant cell lines (P = 0.36) or based on PTEN expression. The only cell line in our study with full PTEN loss, HCC2814, lacked PTEN expression as expected (Supplementary Fig. S1A and S1B). It is also noteworthy that the PIK3CA-mutant cell lines had less vinculin and more PTEN expression comparted to the PIK3CA-WT cell lines, which is associated with a more epithelial-like phenotype.

To confirm that cell toxicity induced by carboplatin is enhanced by CC-115 in LUSC, we performed colony formation assays in three LUSC cell lines in which we had observed synergy in the cell viability assays. Combination of CC-115 and carboplatin at concentrations far below their respective IC₅₀ values significantly reduced clonogenic ability in all three cell lines compared to the individual drugs (P < 0.005) (Fig. 2A and B). These results confirm that inhibition of mTOR/DNA-PK by CC-115 enhances the toxicity induced by carboplatin in LUSC.

Figure 2. CC-115 combined with chemotherapy can synergistically inhibit colony formation in LUSC cell lines.

A, Representative image of crystal violet stained colonies formed on dish after three weeks in each condition. H226, LOU-NH91, and Calu-1 cells were treated with 0.2 μM, 0.4 μM, and 0.4 μM CC-115 and 2 μM, 5 μM, and 2 μM Carboplatin, respectively. B, Quantification of six independent replicates of normalized total colony area (px) quantified by ImageJ software. The findings were replicated in at least two independent experiments. Error bars represent SEM and significance was calculated via unpaired t-test where ** P < 0.005 and *** P < 0.0005.

Decreased basal AKT and DNA-PK activation are biomarkers of efficacy for carboplatin plus CC-115

Because PIK3CA mutations and PTEN expression were not associated with synergy between carboplatin and CC-115, we explored other potential biomarkers that might predict efficacy, by interrogating relevant signaling pathways in the commercially available cell lines via immunoblot under basal conditions. When we grouped the cell lines based on whether CC-115 and carboplatin were synergistic, unique patterns emerged. Cell lines in which the drugs were synergistic had decreased P-S473 and P-Th308 AKT, although this did not reach statistical significance. However, P-S2056 DNA-PKcs was significantly decreased (P = 0.04) (Fig. 3A and 3B). These data suggest that low basal activation of these pathways might be associated with synergy, and that decreased basal activation of AKT and DNA-PK may be biomarkers for sensitization of LUSC to carboplatin by CC-115.

Figure 3. Signaling cascade assessment demonstrates unique networks in LUSC cell lines under basal conditions.

A, Immunoblotting across cell lines under basal conditions demonstrates unique patterns in key signaling molecules when samples are separated by their response to combination treatment. Cell lines in which carboplatin and CC-115 are synergistic demonstrate decreased levels of basal P-AKT and P-DNA-PKcs B, Quantification of P-DNA-PK (S2056) levels, normalized to total DNA-PK and vinculin and quantification of P-AKT (Th308) or P-AKT(S473) levels, normalized to total AKT1 and vinculin. Statistical analysis used in B was a Wilcoxon-Mann-Whitney test where * P < 0.05.

CC-115 sensitizes LUSC to carboplatin by inhibiting chemotherapy-induced AKT activation and enhancing apoptosis

Because synergy between CC-115 and carboplatin was potentially associated with low basal activation of AKT and DNA-PK, and carboplatin treatment induced the activation of AKT and DNA-PK (Fig. 4A), we next asked if carboplatin-induced PI3K pathway or DNA-PK activation are stronger in cell lines in which the drugs are synergistic. Indeed, cell lines in which CC-115 and carboplatin were synergistic demonstrated a significantly greater induction of P-AKT at S473, a downstream marker of mTORC2 activation, upon carboplatin treatment compared to cell lines in which synergy was lacking (P = 0.03) (Fig. 4A and 4B). However, synergy was not associated with a significant increase in carboplatin-induced P-Th308-AKT or P-S2056-DNA-PKcs, even though there was a nearly two-fold average increase in P-S2056-DNA-PKcs among the cell lines that demonstrated synergy. After three days of treatment, carboplatin increased DNA double strand breaks, evidenced by higher levels of γ-H2AX, and also increased cleaved PARP1, indicative of apoptosis (Fig. 4A). Interestingly, in two of the LUSC cell lines we examined, the induction of P-S473-AKT upon treatment with carboplatin was diminished when CC-115 was applied (Fig. 4C). These data suggest that CC-115 sensitizes LUSC to carboplatin at least partly by blocking carboplatin-induced AKT activation, mediated by mTORC2. In p53-null H596 and LOU-NH91 cells, apoptosis, as measured by cleaved PARP1, increased upon treatment with carboplatin, which was maintained following combination treatment (Fig. 4C). However, cleaved PARP1 was unaffected in the p53-WT cell lines H226 or Calu-1. Thus, the synergy may also be attributed to increased apoptosis caused by CC-115-induced inhibition of NHEJ which is more evident in LUSCs lacking p53, representing a potential enhanced therapeutic vulnerability amongst this subset of LUSC.

Figure 4. CC-115 inhibits chemotherapy-induced AKT activation and enhances apoptosis.

A, Immunoblotting demonstrates increased activation of P-AKT at S473 and Th308 and P-DNA-PKcs in cell lines which display synergy between carboplatin and CC-115 following treatment with carboplatin at the cell line-specific IC₂₅ values. Numbers on the blot represent quantification of protein levels normalized to loading control vinculin, and then to the basal level of P-AKT to assess induction. B, Quantification of protein levels for either P-AKT S473, P-AKT Th308, or P-DNA-PKcs (S2056) normalized to loading control vinculin. Significance was calculated via a t-test where * P < 0.05 C, Immunoblot assessment of single compound treatment, or combination treatment at the IC₂₅.

Cell lines who responded synergistically to CC-115 plus carboplatin have enrichment in IFNα and IFNγ pathways

Gene set enrichment analyses were conducted by comparing gene expression profiles of 5 of the cell lines which responded to the combination synergistically, against 8 cell lines which did not respond synergistically (Fig. 5A). This analysis showed that Calu-1, LOU-NH91, H1975, H226 and H596, all synergistic-responders, had a significant enrichment in both IFNα (P < 0.00001; FDR q-value < 0.00001) (Fig. 5B and Supplementary Fig. S2A) and IFNγ (P < 0.00001; FDR q-value < 0.00001) (Fig 5B and Supplementary Fig. S2B) pathway genes within the Hallmark defined gene set. Consistent with this result demonstrating enhanced immune and inflammatory signaling, the next most enriched pathway in the cell lines which responded synergistically was tumor necrosis factor-α (TNF-α) signaling (P < 0.00001; FDR q-value = 0.02) (Supplementary Fig. S2C). This analysis gives further insight into what other subtypes of LUSC may respond to the combination of CC-115 with chemotherapy.

Figure 5. Cell Lines who responded synergistically to mTOR/DNA-PK inhibition plus carboplatin have enrichment in IFNα and IFNγ pathways.

A, Gene set enrichment analysis (GSEA) of gene expression changes in synergistic cell lines compared to non-synergistic cell lines, represented as normalized enrichment scores (NES) for Hallmark Pathways where * P < 0.00001. B, Enrichment plot for Hallmark Interferon Alpha Response and Hallmark Interferon Gamma Response.

CC-115 sensitizes LUSC patient-derived xenografts to carboplatin and paclitaxel in vivo

Given our in vitro findings suggesting that mTOR/DNA-PK inhibition via CC-115 can effectively sensitize LUSC to chemotherapy, we tested this combination in three LUSC PDX models, and one cell line xenograft model, in immunocompromised mice. We included paclitaxel because it is a common standard-of-care drug that is used as part of platinum-based doublet chemotherapy in LUSC, and therefore would likely be used in combination with carbotaxol as a treatment arm in a future clinical trial. Treatment schedules were designed to match the patients’ clinical schedule, with weekly intraperitoneal (I.P.) injection of carboplatin and paclitaxel, and daily oral gavage of CC-115 (Supplementary Fig. S3A) (31). Doses were determined by individual titration of each compound (Supplementary Fig. S3B and S3C). Combination of CC-115 with carboplatin and paclitaxel was significantly better at inhibiting tumor growth (Fig. 6A, 6B and 6C) in two PIK3CA-mutant PDXs, and one PIK3CA-WT LUSC xenograft model. In contrast, in a PIK3CA-WT LUSC PDX model, there was no response to the targeted inhibitor, and no significant inhibition of tumor growth in the combination-treated group compared to chemotherapy alone (Supplementary Fig. S3D). There was no significant change in weight in each of the treatment groups (Supplementary Fig. S3E–3H), suggesting lack of toxicity. In addition, to mimic the clinical scenario, we followed the Calu-1 xenograft cohort after treatment was removed (last day of treatment Day 28) and compared survival times via Kaplan Meyer analysis. The median survival time of combination-treated mice was significantly longer and extended to 55.5 days, whereas median survival in the carboplatin and paclitaxel treated mice was only 37 days (P = 0.05) (Fig. 6D). Immunoblot of tumor lysates collected after four weeks of treatment (Fig. 6E) confirmed inhibition of P-DNA-PKcs, P-AKT and P-S6 in the CC-115 and combination-treated groups as compared to the vehicle control. Furthermore, when tumors were examined via IHC (Fig. 6F), Ki67 and P-S6 were significantly decreased in the combination group (P = 0.05) (Fig. 6G).

Figure 6. CC-115 synergizes with carboplatin and paclitaxel in lung squamous cell carcinoma in vivo models.

A, T050 LUSC PDX model (6 mice per treatment arm) and B, Calu-1 LUSC xenografts (5 mice per treatment arm) were treated with double vehicle (PBS via I.P and 0.5% carboxymethyl cellulose with 0.25% Tween-80 via oral gavage), 1 mg/kg CC-115 via oral gavage (Green arrowhead), 40 mg/kg carboplatin and 15 mg/kg paclitaxel (carbotaxol) via I.P. injection (Blue arrow) or a combination of the three. Tumors were blindly measured 3 times weekly, and each treatment arm contained a minimum of 8 tumors. C, J000094918 LUSC PDX model mean tumor volume (8 mice per treatment arm) during treatment. D, Kaplan-Meyer Survival Analysis of Calu-1 xenografts, and median survival time by groups (days). E, Immunoblot of tumors collected 1.5 hours after compound administration following four weeks of complete treatment in J000094918 LUSC PDX. F, H&E and Immunohistochemical staining of Ki67 and P-S6 in J000094918 LUSC PDX following four weeks of treatment. Tumors were collected 1.5 hours after treatment. Scale bar indicates 200 for overview picture (left) or 50 μM for zoom-in (right) G, Quantification of Ki67 and P-S6 immunohistochemistry (minimum 3 tumors per group) analyzed via QuPath software with slides scanned at 20X, showing decreased Ki67 and P-S6 in the combination group. Statistical analyses were performed in GraphPad Prism using a t-test where * P < 0.05, ** P < 0.005, and *** P < 0.0005.

Combination treatment of carboplatin, paclitaxel and CC-115 is well-tolerated in vivo

Because CC-115 plus chemotherapy was nontoxic in NSG mice which are null for PRKDC, and CC-115 targets DNA-PK, we assessed any potential toxicity in PRKDC-WT FVB/NJ mice, the parental strain from which NSG mice were derived. Small intestines, hearts, lungs, livers, and spleens were stained with H&E and confirmed by a board-certified pathologist to show no signs of organ-specific toxicity and there was less than 10% weight loss (Supplementary Fig. S4A, S4B). Peripheral blood samples showed no changes in the CC-115-treated group, and as expected, a mild to moderate anemia, but no other bone marrow toxicity in the combination group (Supplementary Fig. S4C), demonstrating a commonly documented side effect experienced with carboplatin and paclitaxel combination treatment in NSCLC (40). Overall, these results suggest that the combination of mTOR/DNA-PK inhibition with carboplatin and paclitaxel is potentially safe in humans because it does not exacerbate the clinically accepted side effects of the standard of care chemotherapy.

Discussion

In this study, we provide preclinical in vitro and in vivo evidence that the dual mTORC1/2 and DNA-PK inhibitor, CC-115, sensitizes LUSC to standard chemotherapy. Four of six PIK3CA-mutant NSCLC cell lines responded synergistically to the combination in vitro, though effectiveness of this combination was not limited to the presence of an activating PIK3CA mutation. This result was expected given the extensive alterations in cell lines, as well as the high mutational burden common in LUSC, suggesting that classification by PIK3CA status is unlikely the only genomic alteration to predict excellent response to this combination (8). Interestingly, however, was that both of the synergistic LUSC cell lines which lack PIK3CA mutations, contain alterations in other PI3K genes, including one with a copy number gain in the catalytic subunit, PIK3C2G, and the other with an inactivating mutation in the regulatory subunit, PIK3R6.

Various compounds ranging from pan-PI3K, isoform-specific PI3K, AKT, mTOR inhibitors, to dual PI3K/mTOR inhibitors have been developed and tested for cancer treatment. Alpelisib (BYL719), a PIK3CA-specific inhibitor, is currently in clinical use for PIK3CA-mutant metastatic breast cancer in combination with fulvestrant, and a phase II clinical trial has recently been completed for this compound in advanced NSCLC patients harboring PIK3CA mutations (NCT02276027). Another phase II trial under Lung-MAP, focused specifically on recurrent or stage IV LUSC patients with PIK3CA alterations, tested the daily administration of taselisib (GDC-0032), another PIK3CA-specific inhibitor (NCT02785913). Although these clinical trials offered promise, they have yet to result in the approval of any PI3K/AKT/mTOR targeted agents in LUSC (10,20,21), thus, the efficacy of CC-115 plus chemotherapy in our study is a timely advancement that may provide an additional avenue for a clinical trial.

DNA-PK is a multifunctional kinase whose dysregulation in cancer pathogenesis is becoming increasingly evident (26,28,41). Not only is DNA-PK protumorigenic in numerous cancer types including prostate, breast, colon, and chronic leukemias, its heightened activity is associated with poor clinical outcomes and increased resistance to both chemo- and radiation-therapies (27,42–44). Consistent with our results, Sirzen et al. demonstrated that NSCLC cell lines with the lowest basal DNA-PKcs activity were most sensitive to inhibition via radiotherapy (43). Given this increasing body of evidence, efforts have been made to pharmacologically inhibit DNA-PK in cancer therapy, including dual-inhibitors of DNA-PK and PI3K, with some candidates advancing to clinical trials, but none receiving clinical approval as of yet (26). With a wider inhibitory coverage as a dual mTORC1/2 and DNA-PK inhibitor, CC-115 remains one of the most attractive compounds due to its excellent bioavailability, tolerable toxicity, and extensive clinical research in numerous ongoing trials across cancer types (NCT01353625; NCT0283388; NCT02977780) (31,45).

It is well known that carboplatin causes extensive DNA damage, including DSBs, which was confirmed by the presence of increased γ-H2AX in our study, making it a logical partner for combination treatment with a DDR-inhibitor in LUSC. This rationale is further strengthened by transcriptomic data demonstrating a significant correlation between DDR pathways and LUSC (25). We show that carboplatin specifically induces activation of DNA-PKcs and AKT. In the cell lines in which the drugs were synergistic, we observed a relatively lower degree of basal P-S473 and P-Th308-AKT activation, in addition to decreased P-S2056-DNA-PKcs, all of which were markedly increased upon carboplatin treatment. The reason for this association with synergy could be that cell lines with higher steady state levels of activated DNA-PK and/or AKT more robustly resist the effects of CC-115, though this would need to be tested experimentally. Indeed, the most significant change following carboplatin treatment in the LUSC cell lines that showed synergy was seen in phosphorylation of AKT at S473. Activation of AKT is a well-documented mechanism of resistance to chemo- and radiation-therapy across cancer types (46). In our study, synergy between CC-115 and standard chemotherapy could be at least partly explained by the robust carboplatin-induced activation of AKT, which was prevented by inhibition of mTORC2 via CC-115 in both in vitro and in vivo experiments. In addition, in a subset of LUSC, CC-115 increased carboplatin-induced apoptosis by inhibiting the DDR. Indeed, CC-115 was shown to inhibit activation of DNA-PK and mTOR signaling, causing a decrease in P-S2056 in DNA-PKcs, and P-S235/236 in ribosomal protein S6. This diminished signaling along the PI3K/AKT/mTOR pathway caused by our combination treatment provides a strong rationale for biomarker driven clinical studies built on this defined crosstalk between DNA-PK and PI3K signaling by combining mTOR/DNA-PK inhibition with platinum-based doublet chemotherapy.

Gene set enrichment analysis demonstrated that cell lines which responded synergistically to the combination of mTOR/DNA-PK inhibition with carboplatin had significant enrichment of IFNα and IFNγ pathway genes. In NSCLC, IFNγ signaling is of much interest in the context of immune checkpoint inhibition. Following treatment with either PD-1 of PD-L1 inhibitors, NSCLC patients with increased IFNγ signaling were found to have higher overall response rates and longer progression-free survival than those with low IFNγ signaling (47,48). This inflammatory signature enrichment present in the synergistic cell lines begs the question as to whether CC-115 would synergize with immunotherapy in combination with carboplatin. Given the increasing use of immune checkpoint inhibition in NSCLC as a standard of care, it is imperative that we continue this preclinical work by assessing the potential of CC-115 in combination with both chemotherapy and immunotherapy.

The present study has many strengths, including that it surveys a large, difficult-to-obtain panel of LUSC cell lines representing the heterogeneity of human LUSC which would be encountered in a clinical trial setting. In addition, this panel of cell lines includes 6 PIK3CA-mutant cell lines, a highly prevalent alteration for which there is no targeted therapy currently available. This includes the stable PIK3CA-mutant cell line that our lab derived and that has not been previously investigated. Our efficacious findings of the combination therapy in our multiple PDX and cell-line xenograft models was strengthened by the tolerable toxicity profile demonstrated in our PRKDC-WT mouse experiments.

Despite demonstrating the potential use of CC-115 and chemotherapy in LUSC, some questions still remain unanswered. Although we demonstrate that cell lines which responded synergistically had lower basal levels of P-S473-AKT, a specific genetic alteration as a biomarker for efficacy was not identified. In order to move this therapeutic strategy to the clinic, further studies across a wider range of LUSC genomic alterations should be performed, and additional toxicity studies to compare chemotherapy to CC-115 plus chemotherapy should be carefully examined. This effort would also be enhanced by further mechanistic studies, interrogating gene expression changes following individual and combination treatment to offer additional insight into the mechanism underlying the synergy between CC-115, carboplatin, and paclitaxel in LUSC.