Mutationally Activated PIK3CA^H1047R Cooperates With BRAF^V600E To Promote Lung Cancer Progression

Abstract

Adenocarcinoma of the lung, a leading cause of cancer death, frequently displays mutational activation of the KRAS proto-oncogene but, unlike lung cancers expressing mutated EGFR, ROS1 or ALK, there is no pathway-targeted therapy for patients with KRAS mutated lung cancer. In pre-clinical models, expression of oncogenic KRAS^G12D in the lung epithelium of adult mice initiates development of lung adenocarcinoma through activation of downstream signaling pathways. By contrast, mutationally activated BRAF^V600E, a KRAS effector, fails to initiate lung carcinogenesis despite highly efficient induction of benign lung tumorigenesis. To test if PI3′-kinase-α (PIK3CA), another KRAS effector, might cooperate with oncogenic BRAF^V600E to promote lung cancer progression, we employed mice carrying a conditional allele of Pik3ca that allows conversion of the wild-type catalytic subunit of PIK3CA to mutationally activated PIK3CA^H1047R. Whereas expression of PIK3CA^H1047R in the lung epithelium, either alone or in combination with PTEN silencing, was without phenotype, concomitant expression of BRAF^V600E and PIK3CA^H1047R led to dramatically decreased tumor latency and increased tumor burden compared to BRAF^V600E alone. Most notably, co-expression of BRAF^V600E and PIK3CA^H1047R elicited lung adenocarcinomas in a manner reminiscent of the effects of KRAS^G12D. These data emphasize a role for PI3′-kinase signaling, not in lung tumor initiation per se, but in both the rate of tumor growth and the propensity of benign lung tumors to progress to a malignant phenotype. Finally, biological and biochemical analysis of BRAF^V600E/PIK3CA^H1047R expressing mouse lung cancer cells revealed mechanistic clues as to cooperative regulation of the cell division cycle and apoptosis by these oncogenes.

INTRODUCTION

Non-small cell lung cancer (NSCLC) is a leading cause of cancer mortality (1). Recently identified mutations in proto-oncogenes in NSCLC have provided strategies for the deployment of pathway-targeted therapies (2). However, despite the success of such agents in the treatment of genetically-defined subsets of lung cancer, most NSCLC patients do not have recourse to the use of such agents.

Of the proto-oncogenes mutated in NSCLC, KRAS remains the most common (≥25%) and the most pharmacologically intractable (2). Consequently, attention has turned to inhibiting key downstream signaling pathways required for maintenance of KRAS mutated NSCLC. Moreover, KRAS effectors such as BRAF, PI3′-kinase and AKT are also directly implicated as bona fide human lung cancer genes (2, 3, 4, 5). However, a key outstanding question is, what KRAS^G12D regulated pathways are essential for lung carcinogenesis in GEM models? Although KRAS^G12D-driven tumors are MEK1/2 dependent, expression of BRAF^V600E in the lung epithelium elicits benign lung tumors that rarely progress to malignancy due to a senescence-like growth arrest (6, 7, 8). By contrast, although transgenic expression of mutationally activated PIK3CA^H1047R in the lung epithelium was reported to promote lung tumorigenesis, combined inhibition of PI3′-kinase/mTorc had no effect on established KRAS^G12D-initiated lung tumors (9). Consequently, we tested whether mutationally activated BRAF^V600E and PIK3CA^H1047R might cooperate in promoting lung carcinogenesis using GEM models (6, 11).

Perhaps surprisingly, we could not substantiate reports that PIK3CA^H1047R can initiate lung tumorigenesis. However, concomitant expression of both BRAF^V600E and PIK3CA^H1047R led to rapid onset of lung tumorigenesis with evidence of malignant progression observed six months after tumor initiation. The cooperative effects of PIK3CA^H1047R in vivo were AKT dependent and were modeled in vitro. Finally, we demonstrate that both BRAF^V600E and PIK3CA signaling is necessary for proliferation and survival of BRAF^V600E/PIK3CA^H1047R expressing lung cancer cells and demonstrate that these pathways cooperatively regulate the cell division cycle and apoptosis.

RESULTS

Expression of PIK3CA^H1047R combined with PTEN silencing fails to induce lung tumors in mice

To test whether PIK3CA^H1047R could initiate lung tumorigenesis, we employed Pik3ca^lat-H1047R mice (Pik3ca^lat) carrying a modified Pik3ca allele that expresses normal PIK3CA prior to Cre-mediated recombination after which PIK3CA^H1047R is expressed under control of the gene’s chromosomal regulatory elements (9, 11). Cre-mediated genetic alterations were initiated using an adenovirus encoding Cre recombinase (Ad-Cre) (6). Since expression of PIK3CA^H1047R from a single allele might be insufficient for lung tumorigenesis, we generated mice with every possible heterozygous or homozygous combination of Pik3ca^lat in combination with a conditional null allele of Pten (Pten^lox) (12).

The lungs of adult mice of the various Pik3ca^lat/Pten^lox genotypes were infected with Ad-Cre and euthanized for analysis three, six or ≥12 months post-infection. As a control for successful tumor induction, BRAF^V600E expression was initiated in the lung epithelium of BRaf^CA mice, which developed benign lung adenomas at high multiplicity requiring euthanasia ~12 weeks post-initiation (Fig. 1A, right panel). By contrast, we failed to detect lung tumors in mice carrying conditional alleles of Pik3ca or Pten, either alone or in combination, at three or six months after Ad-Cre infection (Fig. 1A, left and middle panels). That PTEN silencing was insufficient for lung tumorigenesis was consistent with previous observations, but the lack of lung tumorigenesis in the compound Pik3ca^lat/Pten^lox mice was surprising (13). At ≥12 months we detected benign adenomas in Pik3ca^lat/lat; Pten^lox/lox mice however these were rare (< one tumor/mouse). Most of these tumors stained positive for PTEN expression and negative for phospho (p)-AKT suggesting that they may be spontaneously arising lung tumors unrelated to mouse genotype. However, we also detected very rare lung tumors in Pik3ca^lat/lat; Pten^lox/lox mice that stained brightly for pAKT and may therefore due to this combination of genetic modifications (Fig. 1T). Immunofluorescence analysis of Ad-Cre infected Pik3ca^lat or Pik3ca^lat/lat mice failed to detect evidence of cells with pAKT or Ki67 expression above that detected in normal mouse lung (Fig. 1F–M). However, in Pik3ca^lat mice carrying either one or two Pten^lox alleles, small numbers of airway epithelial hyperplasias were detected (Fig. 1N–U). Most prevalent in Pik3ca^lat/lat; Pten^lox/lox mice, these lesions comprised small numbers of cells that stained negative for PTEN and positive for pAKT, with the strongest pAKT signal detected in lesions in Pik3ca^lat/lat; Pten^lox/lox mice (Fig. 1R). These lesions were ≤50μm in diameter and not proliferative (Fig. 1S). Hence, by contrast to the effects of KRAS^G12D or BRAF^V600E, there was no combination of Pik3ca^lat and Pten^lox that elicited lung tumors in mice within 6 months after Ad-Cre (6, 9, 14).

Figure 1. PI3′-kinase pathway activation is insufficient to initiate tumorigenesis in the mouse lung.

A: Lungs of mice of the indicated genotypes were infected with Ad-Cre and monitored for three or ≥12 months as indicated at which time they were euthanized and their lungs processed for H&E staining. Representative H&E-stained tissue sections from Pik3ca^lat/lat (left) or Pik3ca^lat/lat;Pten^lox/lox (middle) animals euthanized 6 months after infection. BRaf^CA mice analyzed 12 weeks following Ad-Cre initiation are presented as a positive control (right).

B–U: Mouse lung sections from control (uninfected) or mice of the indicated genotypes treated as described in A were stained with DAPI, antisera against phospho-(p)-AKT or Ki67 as indicated. Insets indicate higher magnification of regions of airway hyperplasia.

PIK3CA^H1047R dramatically accelerates BRAF^V600E driven lung tumorigenesis

To determine if PIK3CA^H1047R cooperates with BRAF^V600E in lung tumorigenesis, we infected adult BRaf^CA or compound BRaf^CA; Pik3ca^lat mice with either 5x10⁶ or 10⁷ pfu Ad-Cre. Kaplan-Meier survival analysis of mice infected with 5x10⁶ pfu Ad-Cre indicated that all BRaf^CA; Pik3ca^lat mice reached end-stage by ~50 days based on their body conditioning score, a time at which all BRaf^CA mice remained healthy (Fig. 2A) (15). Overall, BRaf^CA mice lived twice as long as BRaf^CA; Pik3ca^lat littermates after initiation (median survival 100 vs. 39 days, p = 1.25 × 10⁻⁵, Fig. 2B). Furthermore, three weeks after infection with 10⁷ pfu Ad-Cre, many BRaf^CA; Pik3ca^lat mice displayed labored breathing and reduced body weight indicating onset of lethal lung tumorigenesis at which time representative mice were euthanized for necropsy (Fig. 2B). Compared to control BRaf^CA mice, BRaf^CA; Pik3ca^lat mice presented with more tumors (96 vs. 54, p=0.025) and BRAF^V600E/PIK3CA^H1047R expressing tumors were larger than their BRAF^V600E expressing counterparts (52,516 μm² vs. 6,782.5μm², p=0.004, Figs. 2B&C). Three weeks post-initiation, BRAF^V600E/PIK3CA^H1047R expressing lung lesions manifested as large adenomas, whereas BRaf^CA mice displayed only alveolar and/or airway hyperplasias (Figs. 2B&C). Overall, BRaf^CA; Pik3ca^lat mice displayed a 14-fold higher tumor burden at three weeks post-initiation compared to BRaf^CA mice (16.8% vs. 1.2%, p=0.01, Fig. 2C).

Figure 2. PIK3CA^H1047R accelerates lung tumorigenesis initiated by BRAF^V600E.

A: BRaf^CA and BRaf^CA; Pik3ca^lat mice were infected with Ad-Cre and monitored prospectively over ~120 days. Mice were euthanized at end-stage per IACUC regulations and a Kaplan-Meier survival curve was plotted.

B: BRaf^CA and BRaf^CA; Pik3ca^lat mice were infected with 10⁷ pfu Ad-Cre to initiate expression of BRAF^V600E either alone or in combination with PIK3CA^H1047R. Mice were euthanized three weeks later with lungs processed for H&E staining. Scale bar = 100μm.

C: Tumor number, size and burden were calculated as described in Materials and Methods in mice carrying either BRAF^V600E or BRAF^V600E/PIK3CA^H1047R expressing tumors.

PTEN silencing enhances BRAF^V600E-induced lung tumorigenesis

To complement analysis of the effects of PIK3CA^H1047R on BRAF^V600E-induced lung tumorigenesis, we tested if deregulation of PI3′-kinase signaling through PTEN silencing might have similar effects. Indeed, this is a combination of alterations commonly found in human melanoma and which cooperate in GEM models of the disease (5, 16). We infected BRaf^CA, BRaf^CA; Pik3ca^lat/+ or compound BRaf^CA; Pten^lox/lox with 10⁶ pfu Ad-Cre with mice euthanized four weeks after initiation. Compared to control BRaf^CA mice, BRaf^CA; Pik3ca^lat and BRaf^CA/+; Pten^lox/lox mice displayed a ≥10-fold and a 4-fold increase in overall tumor burden respectively (2.8% vs. 31.2% vs. 10.8%, Figs. 3A & B). Whereas most lung tumor cells arising in BRaf^CA mice displayed readily detectable PTEN expression, the vast majority of tumor cells in BRaf^CA; Pten^lox/lox mice were PTEN negative (Fig 3C). These data indicate that the statistically significant difference between lung tumorigenesis in BRaf^CA; Pik3ca^lat vs. BRaf^CA/+; Pten^lox/lox mice is not due to inefficient PTEN silencing. Overall, these data support the hypothesis that activation of PI3′-kinase signaling, either by co-expression of PIK3CA^H1047R or PTEN silencing, cooperates with BRAF^V600E in lung tumorigenesis (13).

Figure 3. PTEN silencing accelerates BRAF^V600E-initiated lung tumorigenesis.

A: Mice of the indicated genotypes were infected Ad-Cre and euthanized four weeks later for analysis of tumor burden, which was quantified as described previously. Scale bars represent 100μM.

B: Lung tumors arising in BRaf^CA or BRaf^CA; Pten^lox/lox mice were stained with antisera against PTEN.

Pharmacological blockade of MEK or AKT prevents the growth of BRAF^V600E/PIK3CA^H1047R lung tumors

To assess the role of downstream signaling components on the cooperation between BRAF^V600E and PIK3CA^H1047R, we employed pharmacological inhibitors of either MEK1/2 (PD325901) or AKT (MK-2206) (17, 18). These agents are non-substrate competitive, allosteric inhibitors with high specificity and selectivity. BRaf^CA or BRaf^CA; Pik3ca^lat mice were infected with 10⁶ Ad-Cre and then separated into two groups for treatment with either vehicle control or MK-2206 (120 mg/kg, q.d., po) to inhibit AKT for an additional 4–5 weeks at which time mice were euthanized.

MK-2206 treatment had no effect on BRAF^V600E-initiated lung tumor burden (vehicle - 25% vs. MK-2206 - 22%, p=0.63, Fig. S1A). By contrast, consistent with previous results, MEK1/2 inhibition largely abolished BRAF^V600E-induced lung tumorigenesis (not shown) (6, 7). These data strongly suggest that BRAF^V600E-induced lung tumors are AKT independent for their initial growth.

By contrast to BRAF^V600E driven tumors, MK-2206 had a significant effect on BRAF^V600E/PIK3CA^H1047R driven lung tumorigenesis (Figs. 4A&B). Although four weeks of MK-2206 dosing had no effect on overall tumor number (vehicle = 31.8 vs. MK-2206 = 28.7; p=0.76), both tumor size (vehicle = 121,413 μm² vs. MK-2206 = 30,083 μm², p=<0.0001) and overall tumor burden (vehicle = 20.7%, MK-2206 = 5%, p=0.004), were significantly reduced following AKT inhibition (Figs. 4B). Indeed, tumor sizes and overall tumor burden that developed in BRaf^CA; Pik3ca^lat mice in the presence of MK-2206 were roughly equivalent to that observed in BRaf^CA mice. These data support the hypothesis that AKT activity is required for the cooperation between BRAF^V600E and PIK3CA^H1047R in lung tumorigenesis but largely dispensable for the growth of BRAF^V600E-induced lung tumors. Inhibition of pS473-AKT by MK-2206 was confirmed by immunoblot analysis of lysates of BRAF^V600E or BRAF^V600E/PIK3CA^H1047R expressing lung tumors treated with either vehicle or MK-2206 two hours prior to euthanasia (Figs. S1B & S1C, respectively).

Figure 4. Pharmacological inhibition of PI3′K/AKT and MAPK signaling inhibits growth of BRAF^V600E/PIK3CA^H1047R tumors.

A:BRAF^V600E or BRAF^V600E/PIK3CA^H1047R expressing lung tumors were initiated in mice of the appropriate genotype using Ad-Cre. Two weeks later mice were dosed for a further four weeks with vehicle control, MK-2206 or PD325901 as described in Materials and Methods at which time mice were euthanized and lungs prepared for H&E staining. Scale bar represents 500μM.

B: The effects of MK-2206 or PD325901 on lung tumor number, size and overall burden in mice bearing BRAF^V600E or BRAF^V600E/PIK3CA^H1047R expressing lung tumors was estimated as described in Materials and Methods. ** represents p values less than or equal to 0.005.

C&D: Lung tumors were initiated in BRaf^CA; Pik3ca^lat mice and five weeks later they were either untreated or treated with inhibitors of BRAF^V600E (LGX-818) or class 1 PI3′-kinases (BKM-120) either alone or in combination as indicated. Mice were euthanized when they developed end-stage disease and this was used to calculate a Kaplan-Meier survival curve (C). In addition, the weight of every mouse in each group was measured and normalized to its starting weight, which was set at 100%, and the average normalized weight of all mice in a treatment group was used to assess their response to treatment (D).

E&F: Lung tumor burden in BRaf^CA; Pik3ca^lat treated as described in C&D was assessed by H&E staining of lung tumor sections. Lung tumor specimens from untreated or BKM-120 treated mice came from mice with end-stage disease whereas the specimens from the LGX-818 or combined LGX-818 plus BKM-120 treatments were derived from mice at the end of the 4 week dosing schedule. *** represents a p value of 0.0003, **** represents a p value of less than 0.0001.

Previous studies have indicated the exquisite sensitivity of BRAF^V600E driven lung tumors to MEK1/2 inhibition with PD325901 (6, 7). However, analysis of lung, pancreas and colon cancer cell lines suggest that PI3′-kinase signaling, either by mutation of PIK3CA or PTEN silencing, can render cells less sensitive to MEK inhibition possibly by sustaining the expression of D-type cyclins (19, 20). Hence, we sought to determine if BRAF^V600E/PIK3CA^H1047R lung tumors remained sensitive to MEK1/2 inhibition in vivo. Lung tumors were initiated in BRaf^CA; Pik3ca^lat mice and two weeks later mice were dosed with vehicle or PD325901 (12.5mg/kg, q.d. po) for a further four weeks at which time mice were euthanized (Figs. 4A&B). Compared to controls, MEK1/2 inhibitor treated mice displayed a 15-fold reduction in tumor number (31.8 vs. 2.0, p=0.005). Lesions detected in PD325901 treated BRaf^CA; Pik3ca^lat mice were largely small alveolar or airway hyperplasias and no fully formed adenomas were present in these mice. This resulted in a ~12-fold reduction in average tumor size (vehicle = 121,413 μm² vs. PD325901 = 10,914μm² for drug treated mice, p<0.0001) and a >20,000-fold reduction in overall tumor burden (vehicle = 20.7% vs. PD325901 = 0.001%, p=0.0004) compared to controls (Figure 4B). Consequently, these data indicate that BRAF^V600E/PIK3CA^H1047R lung tumors remain sensitive to the anti-tumor effects of MEK1/2 inhibition and highlight the central importance of BRAF^V600E→MEK→ERK signaling in the growth of BRAF^V600E/PIK3CA^H1047R-induced lung tumors.

Enhanced therapeutic benefit of combined BRAF^V600E plus PIK3CA^H1047R inhibition against BRAF^V600E/PIK3CA^H1047R expressing lung tumors

To examine the requirement for sustained signaling in maintenance of BRAF^V600E/PIK3CA^H1047R expressing lung tumors, we initiated tumorigenesis in 34 BRaf^CA; Pik3ca^lat mice and monitored them for five weeks until they were close to end-stage. At this time two mice were euthanized with the remaining mice divided into four cohorts for the following interventions: 1. No treatment; 2. Class 1 PI3′-kinase inhibitor (BKM-120, 60mg/kg, q.d., po); 3. BRAF^V600E inhibitor (LGX-818, 30mg/kg, q.d., po) or; 4. Combination of BKM-120 and LGX-818 (60 and 30mg/kg respectively, q.d., po) (21, 22). Drug benefit was assessed by Kaplan-Meier analysis with mice euthanized when their body conditioning score (BCS) was ≤2 (15) (Fig. 4C). In addition we monitored mouse body weight, which correlates with changes in tumor burden (Fig. 4D).

There was no statistical difference in survival between untreated (n=6) or BKM-120 treated (n=8) mice, all of which lost weight requiring euthanasia within five days after dosing (Figs. 4C&D). By contrast, mice treated with LGX-818 (n=8) or combined LGX-818 plus BKM-120 (n=8) displayed increased survival (Fig. 4C) such that no mice in the combination arm required euthanasia. Although LGX-818 treated mice continued to lose weight, this was at reduced rate and only 1/8 mice developed end-stage disease. Finally, mice treated with combined LGX-818 plus BKM-120 displayed immediate signs of improvement in their BCS that continued throughout the treatment period (Fig. 4D).

Analysis of mouse lungs from each group indicated that untreated and BKM-120 treated mice displayed a substantial tumor burden (average of ~28% and ~23%) with no difference between them, consistent with the lack of benefit afforded by BKM-120 (Figs. 4E&F). By contrast, LGX-818 treated mice displayed reduced tumor burden (~10%) compared to control or BKM-120 treated mice. The therapeutic benefit of LGX-818 treated mice was further enhanced by combination with BKM-120 (~3% tumor burden) consistent with the survival and health benefits afforded these mice. Taken together, combined blockade of BRAF^V600E and PIK3CA^H1047R signaling in BRAF^V600E/PIK3CA^H1047R expressing lung tumors was superior to blockade of BRAF^V600E or PIK3CA^H1047R alone.

PIK3CA^H1047R promotes malignant progression of BRAF^V600E-induced tumors

KRAS^G12D-induced lung tumorigenesis in GEM models depends on RAF→MEK→ERK, PI3′-kinase, RAC1 and NF-κB signaling (7, 9, 23–26). However, to date, activation of single nodes downstream of KRAS.GTP fails to recapitulate the malignant progression displayed by KRAS^G12D-initiated lung tumors. Although BRAF^V600E induces benign lung tumors, such lesions are low grade and fail to progress to lung cancer (6, 7). As demonstrated here, although PIK3CA^H1047R is unable to initiate lung tumorigenesis, it promotes the growth BRAF^V600E-initiated lung tumors. This prompted us to evaluate whether BRAF^V600E and PIK3CA^H1047R might cooperate for malignant lung cancer progression. To that end, tumors were initiated in either BRaf^CA or BRaf^CA; Pik3ca^lat mice (5x10⁵ pfu Ad-Cre) with mice euthanized 25 weeks later for evaluation (7, 27).

By contrast with early time points, overall tumor number was not different between BRaf^CA and BRaf^CA; Pik3ca^lat mice with an average of 30 tumors per lobe recorded in both genotypes (Figs. 5A&B). However, individual lung tumors in BRaf^CA; Pik3ca^lat mice were larger than those in BRaf^CA mice (35,000μm² vs. 5,000μm², p=0.0005, Fig. 5B). Differences in tumor size were reflected by overall increased tumor burden in BRaf^CA; Pik3ca^lat mice (24% vs. 6%, p=0.0001, Fig. 5B). These data suggest that PIK3CA^H1047R does not influence the frequency of BRAF^V600E initiated cells in the lung epithelium, but decreases the latency period prior to tumorigenesis and also increases their overall proliferative potential.

Figure 5. PIK3CA^H1047R promotes malignant progression of BRAF^V600E-initiated lung tumors.

A: BRAF^V600E or BRAF^V600E/PIK3CA^H1047R expressing lung tumors were initiated in mice of the appropriate genotype using Ad-Cre with mice monitored for 25 weeks at which time they were euthanized and their lungs processed for H&E staining. Scale bar represents 50μM. Lung tumor number, size and overall burden was assessed as described in Materials and Methods.

B: The grade of BRAF^V600E or BRAF^V600E/PIK3CA^H1047R expressing lung tumors was assessed on a 4 point scale using the grading criteria described in Supplemental Figure S3.

One possible explanation for the larger tumors observed in BRaf^CA; Pik3ca^lat mice is that each tumor arises in a polyclonal manner from more than one initiated cell. To test the clonality of BRAF^V600E/PIK3CA^H1047R lung tumors we generated BRaf^CA; Pik3ca^lat mice containing the “Confetti” reporter (Gt(ROSA)26Sor^{tm1(CAG-Brainbow2.1)Cle}, R26^Confetti hereafter) (Supp. Fig. 2) (28). The action of Cre on the R26^Confetti reporter elicits a recombination event that stochastically places one of four different fluorescent proteins (FP) downstream of the CAG promoter. Hence, if BRAF^V600E/PIK3CA^H1047R-induced lung tumors are clonally derived, each tumor should express a single FP. However, if tumors are polyclonal in origin, they will contain a mixture of cells with different FP expression. BRaf^CA; Pik3ca^lat; R26^Confetti mice were infected with 10⁷ pfu Ad-Cre with lung tumors analyzed 9 weeks later. Individual BRAF^V600E/PIK3CA^H1047R lung tumors were uniformly positive for a single FP, indicating that each tumor is clonally derived from a single initiated cell (Supp. Fig. 2).

To assess malignant progression, we employed a grading scheme based on the classification of lung lesions in GEM models to assess the grade of >250 tumors from mice of each genotype (Fig. 5C and Supp. Figs. 3A&B) (14, 27). Hyperplasia was designated Grade 1, benign adenomas as Grade 2, larger adenomas as Grade 3 and adeno-squamous or adenocarcinomas as Grade 4. BRAF^V600E-induced lung lesions were predominantly Grade 1 or 2 with rare examples of Grade 3 detected. By contrast, BRAF^V600E/PIK3CA^H1047R-induced lesions were generally Grade 2 or 3 adenomas but ~3% of tumors were malignant Grade 4 adeno-squamous or adenocarcinoma lesions. These data indicate that BRAF^V600E cooperates with PIK3CA^H1047R to both accelerate early-stage lung tumorigenesis and to promote late-stage malignant progression (14).

To determine if the cell of origin might influence the spectrum of lung tumors initiated in these mice, we employed adenovirus vectors with lung cell specific expression of Cre recombinase (29). Ad-SPC-Cre, Ad-CCSP-Cre and Ad-CGRP-Cre direct Cre expression to Alveolar type II pneumocytes, Clara or neuroendocrine cells respectively. Of these, we noted that Ad-SPC-Cre was the most potent inducer of lung tumorigenesis in either BRaf^CA or BRaf^CA; Pik3ca^lat mice and was equivalent in potency to Ad-CMV-Cre. Moreover, tumors emerging in mice initiated with either Ad-CCSP-Cre or Ad-CGRP-Cre, which were less potent in inducing lung tumorigenesis, displayed the cuboidal morphology and expression of SP-C characteristic of AT2 cells (data not shown).

In these experiments we assessed tumor grade at 22 weeks in groups of BRaf^CA or BRaf^CA; Pik3ca^lat mice initiated with Ad-SPC-Cre to confirm the ability of PIK3CA^H1047R to promote malignant progression of BRAF^V600E-initiated lung tumors (Supp. Fig. 4). As before, the majority of BRAF^V600E expressing lung tumors were small, grade 2 adenomas. By contrast, >50% of lung tumors in the BRaf^CA; Pik3ca^lat mice were high-grade adenomas (Grade 3) and 8.5% were either grade 4 adeno-squamous cancers or adenocarcinomas confirming observations made with Ad-CMV-Cre (30).

TP53 constrains malignant progression of BRAF^V600E/PIK3CA^H1047R lung tumors

TP53 mutations frequently coexist with mutations in KRAS, BRAF or PIK3CA mutations in human lung adenocarcinomas and TP53 mutation or silencing promotes malignant progression of either KRAS^G12D or BRAF^V600E-driven lung tumors (6, 14). Consequently, we determined if TP53 loss would enhance lung tumorigenesis initiated by combined expression of BRAF^V600E and PIK3CA^H1047R. Tumorigenesis was initiated in a BRaf^CA/CA; Pik3ca^lat, Trp53^lox/lox mouse with a BRaf^CA; Trp53^lox/lox mouse as a littermate control. Within four weeks the former mouse displayed end-stage disease requiring euthanasia. Whereas the BRaf^CA; Trp53^lox/lox mouse had a tumor burden of 0.2% made up of mainly small hyperplasias with a solitary benign adenoma detected (Supp. Fig. 5A), the BRaf^CA/CA; Pik3ca^lat, Trp53^lox/lox mouse had a tumor burden of 41% made up of high grade, anaplastic adenocarcinoma. Moreover, these high-grade tumors stained positive for Vimentin, a marker of epithelial-mesenchymal transition, lacked TTF-1/NKX2-1 expression, a marker of the distal lung epithelium and displayed a high proliferative index (Ki67, Supp. Fig. 5B). Similar results were obtained using a second BRaf^CA/+; Pik3ca^lat, Trp53^lox/lox mouse.

PIK3CA^H1047R promotes BRAF^V600E-induced oncogenic transformation in vitro

An advantage of GEM models cancer is the ability to engineer defined genetic alterations and then generate cell lines for in vitro analysis. To that end, we generated BRAF^V600E/TP53^Null and BRAF^V600E/PIK3CA^H1047R/TP53^Null mouse lung cancer derived cell lines (BT and BPT cells hereafter) from suitably manipulated mice (7).

Since BT cells displayed a limited capacity for anchorage independent growth we engineered them to express either mCherry (control), PIK3CA^H1047R or a myristoylated, constitutively activated AKT1 (M⁺AKT1) to test whether activation of PI3′-kinase signaling might cooperate with BRAF^V600E to influence the behavior of these cells (Fig. 6). Cell extracts were analyzed by immunoblotting for the activation status of key nodes in the PI3′-kinase signaling pathway (Fig. 6A and Supp. Fig. 8A). Compared to control, cells expressing either PIK3CA^H1047R or M+AKT1 displayed elevated pPRAS40 and pRP-S6 while M⁺AKT1 expressing cells displayed elevated levels of pAKT, p4E-BP1, pFOXO1 and pGSK3β. PIK3CA^H1047R and M⁺AKT1 cells also showed a decrease in p27^KIP1 expression. Consistent with the ability of PIK3CA^H1047R to cooperate with BRAF^V600E in lung tumorigenesis in vivo, expression of either PIK3CA^H1047R or M⁺AKT1 led to increased anchorage-independent colony formation in vitro (Fig. 6B). Indeed, the parental and mCherry control BT lung cancer cells failed to form colonies in agarose even when cultured for ~8 weeks. By contrast, the PIK3CA^H1047R or M+AKT1 expressing cells cloned with an efficiency of 1% and 1.4%, respectively, after three weeks. These data indicate that PI3′-kinase→AKT signaling cooperates with BRAF^V600E in oncogenic transformation of cultured lung cancer cell lines.

Figure 6. Ectopic expression of either activated PIK3CA or AKT1 promotes anchorage independent growth to BRAF^V600E; TP53^null lung cancer derived cells.

A: BRAF^V600E/TP53^null lung cancer derived cells were infected with retroviruses as described in the text. Cell extracts were analzed by immunoblotting as indicated. Quantification of blots is included as Supp. Fig. 8A.

B: Anchorage-independent growth of BRAF^V600E/TP53^null lung cancer cells engineered to express mCherry, PIK3CA^H1047R or M+AKT1 was quantified over three weeks.

C: BRAF^V600E/PIK3CA^H1047R/TP53^null lung cancer cells were plated in soft agar in the absence (vehicle) or presence of inhibitors of MEK1/2 (1μM PD325901), class 1 PI3′-kinase (5μM GDC-0941) or AKT (5μM MK-2206). Anchorage-independent growth was assessed over three weeks.

The ability of BPT lung cancer cells to form colonies in agarose was assessed in the presence of various pathway-targeted inhibitors. Unlike BT cells, BPT cells are capable of anchorage-independent colony formation (Vehicle, Fig. 6C). However, when cultured in the presence of a MEK inhibitor, a class 1 PI3′-kinase inhibitor or an AKT inhibitor, colony formation was abolished even when cultured for 8 weeks (Fig. 6C) (31). These data emphasize the ability of PI3′-kinase→AKT signaling to cooperate with BRAF^V600E in promoting the transformed phenotype on lung cancer cells.

BRAF^V600E/PIK3CA^H1047R/TP53^Null lung cancer cell lines are sensitive to combined pharmacological inhibition of RAF or PI3′-kinase pathway inhibition

To assess the mechanism(s) of cooperation between BRAF^V600E and PIK3CA^H1047R on the behavior of BPT cells they were treated with various pathway-targeted inhibitors of BRAF^V600E or PIK3CA^H1047R signaling, either alone or in combination, at which time effects on cell viability or the cell division cycle were assessed (Figs. 7A–C).

Figure 7. Cooperative effects of combined pharmacological blockade BRAF^V600E and PIK3CA^H1047R signaling on BRAF^V600E/PIK3CA^H1047R/TP53^null lung cancer derived cells.

A&B: BRAF^V600E/PIK3CA^H1047R/TP53^null lung cancer cells were treated with a range of concentrations of MEKi (PD325901), PI3′Ki (GDC-0941), ERKi (SCH772984) or AKTi (MK-2206) either alone or in the indicated combinations for 72 hours when cell viability was measured using a Cell-Titre-Glo assay (Promega) and normalized to zero concentration values.

C: BRAF^V600E/PIK3CA^H1047R/TP53^null lung cancer cells were treated with inhibitors of MEK1/2, ERK1/2, AKT1-3 or class 1 PI3′-kinases either alone or in the indicated combinations for 24 hours at concentrations listed in Table S1. Cell cycle status was assessed by FACs analysis of cells labeled with BrdU and Propidium Iodide.

D: Unsupervised clustering of RPPA analysis of cell extracts derived from cells treated in biological triplicates with inhibitors of MEK1/2 (PD), ERK1/2 (SCH), AKT1-3 (MK) or class 1 PI3′-kinases (GDC) either alone or in the indicated combinations for 24 hours. Full dataset is included as Supp. Fig. 7 and an Excel file.

E: Cell extracts prepared from BPT lung cancer cells treated with inhibitors of MEK1/2 (PD), ERK1/2 (SCH), AKT1-3 (MK) or class 1 PI3′-kinases (GDC) either alone or in the indicated combinations for 24 hours were analyzed by immunoblotting and quantified against β-actin (Supp. Fig. 8B).

F: Mutationally activated KRAS^G12D promotes lung hyperplasia (H) then benign adenomas (BA), some of which progress to non-small cell lung cancers (NSCLC). Similarly initiated expression of BRAF^V600E leads to more rapid and numerous hyperplasia and benign adenomas but without progression to NSCLC, most likely due to a senescence-like growth arrest. Although PIK3CA^H1047R fails to initiate pathology in the lung, when combined with BRAF^V600E, it promotes rapid emergence of hyperplasias and benign adenomas, some of which progress to NSCLC. At the molecular level, BRAF^V600E cooperates with PIK3CA^H1047R to regulate the cell division cycle and suppress apoptosis.

As expected, single agent inhibition of MEK1/2 or class 1 PI3′-kinases effectively decreased cell proliferation. Moreover, the combination of these agents had the strongest anti-proliferative effects, demonstrating at least an additive effect particularly in the mid-range of drug (Fig. 7A). Despite its potency in vivo, single agent AKT inhibition (MK-2206) was ineffective at preventing the growth of these cells (Fig. 7B) even at 5μM. Single agent ERK1/2 inhibition (SCH772984) also exerted a cytostatic effect, only decreasing cell viability at higher concentrations (>250nM) (32). However, combined ERK1/2 plus AKT inhibition displayed cooperativity such that concentrations of inhibitors that had no single agent effect had striking inhibitory effects on cell proliferation when combined (Fig. 7B), which was further confirmed by Crystal Violet staining (Supp. Fig. 6A).

To assess the effects of the various agents on the cell division cycle, BPT cells were incubated with the various agents either alone or in combination and then labeled with BrdU to assess S phase progression (Fig. 7C, see Table 1 for drug concentrations). Vehicle-treated cells displayed a high proliferative index with >75% of cells incorporating BrdU. As expected, single-agent blockade of MEK1/2, ERK1/2, PI3′-kinase or AKT led to decreased BrdU incorporation and increased the G1 population with the PI3′-kinase inhibitor most potent, the MEK and ERK inhibitors roughly equipotent and the AKT inhibitor least potent. Cells treated with combined MEK1/2 plus PI3′-kinase or ERK1/2 plus AKT inhibitors displayed the most profound effects on the cell division cycle in addition to displaying evidence of cell death as assessed by the number of non-adherent cells and cells with a sub-G1 DNA content (Fig. 7C and Table 2). Taken together, these data indicate that BRAF^V600E and PIK3CA^H1047R cooperatively regulate the cell division cycle and cell death in vitro.

BRAF^V600E and PIK3CA^H1047R likely regulate many biochemical processes in the cell to influence cell proliferation in vitro and tumorigenesis in vivo. To assess the molecular mechanism(s) of such oncogenic cooperation we employed reverse phase protein array (RPPA) analysis to quantify protein expression/modification in BPT cells treated with inhibitors (Fig. 7D). Triplicate cultures of BPT cells were treated with inhibitors of MEK1/2 (PD), ERK1/2 (SCH), class 1 PI3′-kinases (GDC) or AKT1-3 (MK) either alone or in combination as indicated (Fig. 7D). Prior to RPPA analysis, we confirmed that each of the agents had the anticipated biochemical effects (Supp. Fig. 6B). As expected, inhibition of MEK1/2 or ERK1/2 led to strongly diminished pERK1/2 and promoted a modest increase in pAKT. Inhibition of PI3′-kinases or AKT led to strongly diminished pAKT with no obvious effects on pERK. Finally, the combinations of inhibitors potently suppressed both pERK1/2 and pAKT.

The triplicate samples were subject to RPPA analysis using antisera against >150 proteins or protein modifications (33, 34). Unsupervised hierarchical clustering of the data indicated that samples clustered based on the pathway inhibited such that MEKi or ERKi treated samples clustered together as did PI3Ki and AKTi treated samples (Fig. 7D and Supp. Fig. 7). Samples from cells treated with combined pathway inhibitors clustered separately from samples from cells treated with the various single agents with the samples from cells treated with combined MEK1/2 plus class PI3′-kinase showing the greatest separation.

RPPA analysis generated predictions with regards to the consequences of inhibition of BRAF^V600E or PIK3CA^H1047R signaling on regulators of cell proliferation, some of which were then validated in an independent sample set by immunoblotting (Fig. 7E and Supp. Fig. 8). As predicted by RPPA analysis, inhibition of MEK1/2→ERK1/2 signaling led to increased expression of the pro-apoptotic BCL-2 family protein BIM and decreased expression of c-MYC as well as cyclins D1 and A. Inhibition of PIK3CA^H1047R, but not AKT, also promoted BIM expression and inhibited c-MYC, cyclins D1 and A and led to elevated p27^KIP1 expression and suppression of pRP-S6. Consistent with the lack of effect of MK-2206 against the proliferation of cultured BPT cell, we noted that inhibition of AKT had no single-agent effect on BIM, c-MYC, cyclins D1 or A, p4E-BP1 or pRP-S6 but largely extinguished phosphorylation of PRAS40, a known AKT substrate. Finally, combined inhibition of both MEK and PI3′-kinases displayed additive effects on expression of c-MYC, cyclins D1 and A and BIM, the latter correlating with increased cleaved Caspases 3 (CC3) and 7 (CC7) that were also predicted by RPPA analysis. These analyses indicate the complexity with which these pathways cooperatively regulate apoptosis and the cell division cycle, which in turn influences tumor latency, burden and malignant progression. Although these experiments do not identify a single mechanism of BRAF^V600E - PIK3CA^H1047R cooperation in lung carcinogenesis, they suggest numerous areas of future investigation.

DISCUSSION

The mechanism(s) by which KRAS^G12D promotes lung carcinogenesis in humans and mice remain incompletely understood. However, here we demonstrate that two well-credentialed KRAS.GTP effectors, BRAF and PIK3CA, cooperate to mimic the effects of oncogenic KRAS^G12D in GEM models of lung cancer. Whereas BRAF^V600E expression elicits benign lung tumorigenesis, such tumors rarely progress to lung cancer, which contrasts with KRAS^G12D-initiated lung tumors (Fig. 7F). Unexpectedly, we noted that PIK3CA^H1047R failed to initiate lung tumorigenesis, even when combined with PTEN silencing, a result at-odds with a previous report (9) (Fig. 7F). Indeed, no combination of heterozygous or homozygous Pik3ca^lat and Pten^lox initiated lung tumorigenesis, contrasting with the effects of a tetracycline inducible PIK3CA^H1047R transgene. It remains possible that strength of pathway activation, genetic background or the promoter system may influence these disparate results as suggested by similar observations in pancreatic cancer (24, 35). However, the lack of PIK3CA^H1047R effects in lung is consistent with the effects of PIK3CA^H1047R in ovarian epithelial cells (11). Furthermore, others have demonstrated that PI3′-kinase activation through PTEN silencing failed to elicit lung tumorigenesis (13). It is unclear whether the inability of PIK3CA^H1047R to initiate lung tumorigenesis will be predictive of anti-tumor effects of pharmacological inhibition targeting of PI3′-kinase in the clinic (9).

Despite its inability to initiate tumorigenesis, PIK3CA^H1047R cooperated with BRAF^V600E in dramatic fashion to promote lethal lung tumorigenesis, which occurred with short latency and complete penetrance (Fig. 7F). That PTEN silencing also cooperated with BRAF^V600E, albeit less strikingly, confirmed these observations indicating an important role for PI3′-lipid signaling. Although at early time points, BRAF^V600E/PIK3CA^H1047R-induced tumors were benign, six months post-initiation there was evidence of cancer progression in a manner reminiscent of KRAS^G12D. Finally, TP53 silencing in conjunction with combined BRAF^V600E/PIK3CA^H1047R expression, led to rapid lung cancer development, which showed signs of de-differentiation and epithelial-to-mesenchymal transition, hallmarks of advanced disease (36, 37, 38).

In conjunction with previous reports of the MEK1/2 and PI3′-kinase dependency of KRAS^G12D-induced lung tumors (39), the simplest explanation is that combined activation of RAF→MEK→ERK and PI3′-kinase→AKT signaling cooperates to accelerate benign tumorigenesis and to promote malignant lung cancer progression. Although activation of BRAF and PI3′-kinase signaling accounts for much of the phenotype associated with KRAS^G12D expression, a role for other KRAS^G12D effectors remains possible since ~10% of KRAS^G12D initiated tumors display malignant progression whereas only ~3–8% of BRAF^V600E/PIK3CA^H1047R tumors do so (7). Indeed, since ≤10% of KRAS^G12D- or BRAF^V600E/PIK3CA^H1047R-initiated lung tumors display malignant progression, there must be additional stochastic genetic/epigenetic changes in benign tumor cells that promote malignant progression as emphasized by the effects of TP53 silencing on the progression of KRAS^G12D- or BRAF^V600E/PIK3CA^H1047R-initiated lung tumors.

The absence of malignant progression of BRAF^V600E-initiated lung tumors has been ascribed to a senescence-like proliferative arrest, which is due to feedback inhibition of PI3′-lipid production (10). This hypothesis is supported by our data since PIK3CA^H1047R expression enhanced the proliferative capacity of BRAF^V600E-initiated lung tumors. However, it remains unclear how PI3′-kinase signaling sustains the proliferation of these tumors. Previous data indicate cooperative regulation of cyclins cyclin-dependent kinase inhibitor expression by these pathways (40). Analysis of BPT cells suggests that the cell division cycle is indeed under the coordinate control of both BRAF^V600E and PIK3CA^H1047R. However, future experiments are required to determine which of the nodes of cell cycle regulation are essential to the observed cooperation.

BPT cells were sensitive to combined inhibition of BRAF^V600E and PIK3CA^H1047R with pathway-targeted agents. In general, single-agent inhibition of either pathway had inhibitory effects on the cell division cycle without promoting apoptosis. By contrast, combined pathway inhibition had stronger inhibitory effects on the cell cycle and also promoted cell death in vitro and in vivo. That AKT inhibition was without effect against lung tumors initiated by BRAF^V600E but potently suppressed the cooperation between BRAF^V600E and PIK3CA^H1047R suggests a genotype-drug response phenotype with regards to AKT inhibition in GEM models. Whether these data will translate into the clinical utility of single or combined pathway-targeted interventions in humans with BRAF mutated lung cancer remains an open question. Whereas KRAS mutation is frequent in lung cancer, expression of BRAF^V600E or PIK3CA^H1047R accounts for a smaller percentage of NSCLC patients (2). Since mutated KRAS has proven an intractable pharmacological target, it will be interesting to determine whether combined pathway-targeted inhibition of RAF and PI3′-kinase signaling will promote regression of KRAS mutated lung cancers.

MATERIALS AND METHODS

Mice and Adenovirus delivery

Animal experiments were conducted in accordance with protocols approved by the UCSF Institutional Animal Care and Use Committee (IACUC). BRaf^CA (Braf^tm1Mmcm), Trp53^lox (Trp53^tm1Brn), PIK3CA^lat-H1047R and Pten^lox mice were bred and genotyped as previously described (6, 11, 12, 14, 41). All mice were back-crossed onto the FVB/NJ genetic background for at least 10 generations. Experiments were performed using littermates as controls whenever possible. Stocks of Adenovirus encoding Cre recombinase were purchased from Viraquest (North Liberty, IA) or the University of Iowa, Gene Transfer Vector Core (29). Adenoviruses with cell specific expression of Cre recombinase were generously provided by Kate Sutherland and Anton Berns (NKI, Amsterdam) and purchased from the University of Iowa, Gene Transfer Vector Core (29). Intranasal instillation of adenovirus for infection of the mouse lung epithelium was performed as previously described (42). Tumor bearing mice were euthanized for analysis either at a pre-determined time point or when their body conditioning score (BCS) was ≤2 (15).

Histology and quantification of lung tumor burden

Lungs were removed, fixed in zinc buffered formalin (ZBS) and stored in 70%(v/v) ethanol prior to paraffin embedding. 6μm sections were stained with Hematoxylin and Eosin (H&E) on slides, scanned with an Aperio ScanScope scanner. Quantification of tumor number, size, and overall burden of individual lung sections was conducted using Aperio Spectrum ImageScope software. Overall tumor burden was calculated as (area of lung section occupied by tumor)/(total area of section) in μm².

Treatment of mice with pathway-targeted therapeutics

PD325901 (Hansun Trading Co.) was formulated in 0.5%(w/v) Hydroxy-Propyl-Methylcellulose (HPMT, Sigma) and administered daily (q.d) by oral gavage (po) at 12.5 mg/kg. MK-2206 (Merck) was formulated in 30%(w/v) Captisol (Ligand Technology, La Jolla, Ca.) and dosed at 120 mg/kg q.d. po. LGX-818 and BKM-120 (Novartis) were formulated in 0.5% carboxymethylcellulose (Sigma) plus 0.5% Tween 80 (Sigma) and dosed q.d. po at 30 mg/kg and 60 mg/kg, respectively. Mice were dosed with drug for pre-determined time points or until their body conditioning score (BCS) was ≤2, at which point they were euthanized for analysis (15).

Immunostaining of mouse lung tissue and immunoblotting

ZBS fixed sections of mouse lungs were subject to citrate-mediated antigen retrieval and then probed with: anti-pS473-AKT, anti-Vimentin, anti-PTEN (Cell signaling technology), anti-NKX2.1, anti-Ki67, anti-SP-C and anti-p63 (Santa Cruz Biotechnology). Immunoblotting was performed on 50μg of extracted protein and probed with antisera against various phospho- or backbone specific antisera as described. Immunoblots were quantified using an Odyssey FC system and Image Studio software.

Lung tumor cell isolation, culture and analysis

Single cell suspensions were generated from BRAF^V600E/TP53^Null- or BRAF^V600E/PIK3CA^H1047R/TP53^Null-induced lung cancers and cultured in Hams-F12/Glutamax media with 10%(v/v) fetal bovine serum. Following outgrowth, single cell-derived clones were isolated. Recombination of BRaf^CA, Trp53^lox and Pik3ca^lat alleles was verified by PCR (11). Expression of oncogenic BRAF^V600E was determined by immunoblotting with a mAb (VE1) specific to BRAF^V600E (43).

Cell proliferation was assessed in triplicate cultures of cells treated for 72 hours with: DMSO control, MEK1/2 inhibitor (PD325901, 15.6nM-1μM), ERK1/2 inhibitor (SCH772984, 7.8nM-5μM), AKT1-3 inhibitor (MK-2206, 80nM-5μM) or class 1 PI3′-kinase inhibitor (GDC-0941, 80nM-5μM) alone or in combination as indicated using a Cell Titre Glow assay (Promega). Cell cycle status was assessed in fixed floating and adherent cells stained with anti-BrdU-FITC (Becton-Dickinson) and propidium iodide (Sigma). Sorting of mCherry encoding ecotropic retroviruses was performed using a FACS-Calibur (Becton Dickinson) (44, 45). Anchorage-independent colony formation was performed by plating 5x10³ cells in 1.4%(w/v) Sea-Plaque low melting temperature agarose (Cambrex, Rockland, ME) and culturing them for three weeks at which time colony formation was enumerated using dissecting microscope.

Reverse Phase Protein Array (RPPA) Analysis

Reverse phase protein array (RPPA) analysis (Functional Proteomics Core Facility at M.D. Anderson Cancer Center) was performed as previously described (33, 34) and in supplementary information.