Abstract
Using a panel of non–small cell lung cancer (NSCLC) lines, we show here that MAP-ERK kinase (MEK) and RAF inhibitors are selectively toxic for the KRAS-mutant genotype, whereas phosphoinositide 3-kinase (PI3K), AKT, and mTOR inhibitors are not. IGF1 receptor (IGF1R) tyrosine kinase inhibitors also show selectivity for KRAS-mutant lung cancer lines. Combinations of IGF1R and MEK inhibitors resulted in strengthened inhibition of KRAS-mutant lines and also showed improved effectiveness in autochthonous mouse models of Kras-induced NSCLC. PI3K pathway activity is dependent on basal IGF1R activity in KRAS-mutant, but not wild-type, lung cancer cell lines. KRAS is needed for both MEK and PI3K pathway activity in KRAS-mutant, but not wild-type, lung cancer cells, whereas acute activation of KRAS causes stimulation of PI3K dependent upon IGF1R kinase activity. Coordinate direct input of both KRAS and IGF1R is thus required to activate PI3K in KRAS-mutant lung cancer cells.
Significance: It has not yet been possible to target RAS proteins directly, so combined targeting of effector pathways acting downstream of RAS, including RAF/MEK and PI3K/AKT, has been the most favored approach to the treatment of RAS-mutant cancers. This work sheds light on the ability of RAS to activate PI3K through direct interaction, indicating that input is also required from a receptor tyrosine kinase, IGF1R in the case of KRAS-mutant lung cancer. This suggests potential novel combination therapeutic strategies for NSCLC. Cancer Discov; 3(5); 548–63. ©2013 AACR.
See related commentary by Chen and Sweet-Cordero, p. 491
This article is highlighted in the In This Issue feature, p. 471
Introduction
Activating point mutations in the genes encoding the RAS subfamily of small GTP-binding proteins contribute to the formation of a large proportion of human tumors. In lung cancer, one of the most prevalent cancer types worldwide (1), KRAS is mutationally activated in approximately 25% of adenocarcinomas (2, 3). This poses a significant therapeutic challenge, as KRAS mutations are generally associated with resistance to existing therapies (4, 5). Targeting RAS itself presents an attractive approach to this issue, as RAS-mutant tumors have been shown to exhibit oncogene addiction (6, 7). However, in contrast to the efficacy of tyrosine kinase inhibitors in patients with mutant receptor tyrosine kinases (RTK), pharmacologic targeting of activated RAS proteins has been unsuccessful to date. Thus, efforts have shifted toward targeting pathways acting downstream of RAS. Indeed, combined inhibition of extracellular signal-regulated kinase (ERK) and phosphoinositide 3-kinase (PI3K) signaling, 2 well-described RAS-controlled pathways, has shown some efficacy in mutant Kras-driven mouse lung tumor models (8). This combination of pathway inhibitory drugs is being explored in a number of early-phase clinical trials, but so far, both the toxicity and efficacy of this approach is unclear.
Tumors with RAS mutations can also show selective dependencies on activities that are not regulated directly by RAS. To identify factors or pathways necessary for survival and proliferation of cells harboring KRAS mutations, several groups have conducted synthetic lethal RNA interference (RNAi) screens. The list of candidates obtained thus far includes the TANK-binding kinase 1 (TBK1; ref. 9), the transforming growth factor β–activated kinase 1 (TAK1; ref. 10), the transcription factor GATA2 (11, 12), the G1–S regulator cyclin-dependent kinase 4 (CDK4; ref. 13), mitotic regulators (14), and proteasome components (12, 14). Differences in cell type and in specific assay conditions may help explain some of the variability across these different datasets and deeper investigation is required to understand the broader significance of these factors in RAS-driven tumors. Crucially, most of these screens have identified candidate novel targets for drug development, meaning that a significant period must inevitably elapse until any such potential therapy reaches clinical trials. Thus, a complementary approach is to identify targets that are necessary for survival of RAS-mutant cells using compounds that are already available and/or in clinical use. The use of drugs in RAS synthetic lethal screening can permit the analysis of a larger panel of cells, help avoid some of the off-target effects associated with RNA interference (RNAi) and, more importantly, identify immediately applicable therapeutic strategies to treat RAS-mutant tumors.
In this study, we have assayed a collection of small-molecule inhibitors on a panel of human lung cancer cell lines to identify drugs that show selectivity for the KRAS-mutant genotype. Cells harboring KRAS mutations were found to be more sensitive than KRAS wild-type cells to inhibition of the RAF/mitogen-activated protein (MAP)-ERK kinase (MEK)/ERK pathway, whereas no KRAS genotype selectivity was observed when the PI3K/AKT/mTOR pathway was inhibited. Interestingly, however, KRAS-mutant cells exhibit increased dependence on the activity of the insulin-like growth factor 1 receptor (IGF1R). Mechanistically, we show that the ability of KRAS to directly activate the PI3K activity of the p110 catalytic subunit requires a coordinate input from an RTK, IGF1R in the case of lung cancer, acting via the p85 regulatory subunit. These findings suggest potential therapeutic strategies for lung tumors harboring KRAS mutations while avoiding the potential toxicities of direct PI3K inhibition.
Results
KRAS-Mutant NSCLC Cell Lines Are Selectively Sensitive to MEK, RAF, and IGF1R Inhibitors
Using a collection of small-molecule inhibitors, we aimed to identify pathways that are critical for the maintenance and survival of tumor cells carrying an activating KRAS mutation, but not to those lacking this oncogene. For this purpose, we assembled a panel of 25 non–small cell lung cancer (NSCLC) cell lines, 13 of which are KRAS mutant and 12 KRAS wild-type (Supplementary Table S1). Cell lines known to harbor epidermal growth factor receptor (EGFR) mutations were purposely excluded from the selection. To conduct an initial characterization of the dependence of the 2 groups on the expression of KRAS for cell survival, we used RNA interference to deplete endogenous levels of KRAS acutely. As anticipated, KRAS knockdown using 2 different siRNA pools led to a notable selective increase in apoptosis in most of the KRAS-mutant, but not wild-type, cells and an accompanying decrease in cell viability (Fig. 1A and B). This effect was more statistically significant using siRNAs that have been chemically modified to reduce off-target effects (OTP; ref. 15) and indicates that most of the KRAS-mutant cell lines in this panel show some evidence of RAS oncogene addiction.
Next, we used the panel of 25 NSCLC cell lines to assess the effect on cell viability of more than 50 small-molecule inhibitors targeting pathways directly controlled by RAS, such as RAF/MEK/ERK or PI3K/AKT/mTOR, as well as drugs directed against other less direct targets such as HSP90 or NF-κB. Figure 1C–J and Supplementary Fig. S1A and S1B illustrate the effect on cell viability of several selected inhibitors. To identify those drugs achieving statistical significance in discriminating between KRAS-mutant and wild-type cells, we conducted two-way ANOVA (Table 1). The analysis revealed that cells bearing KRAS mutations tended to be, as expected (16), significantly more sensitive to RAF and MEK inhibitors than KRAS wild-type cells. Of the RAF inhibitors, AZ628 showed the greatest selectivity; this is a pan-RAF inhibitor with somewhat more potency towards CRAF (29 nmol/L) than BRAF (110 nmol/L; ref. 17). However, no significant KRAS genotype selectivity was observed when the PI3K/AKT/mTOR pathway was inhibited by any of a range of targeted molecules, with considerable loss of cell viability seen on most cell lines irrespective of genotype. Intriguingly, KRAS-mutant cells exhibited enhanced sensitivity to a different class of drugs, 3 of the 5 tested IGF1R inhibitors. Indeed, p values associated with these 3 drugs were among the most significant, comparing favorably with those produced by the most potent MEK inhibitors. In contrast, although values failed to reach statistical significance, KRAS wild-type cells tended to show increased sensitivity toward EGFR inhibition compared with mutant cells. Finally, cells carrying KRAS mutations also responded slightly more strongly to the HSP90 inhibitors 17-AAG and 17-DMAG and to the MET/ALK kinase inhibitor PF-02341066, although the magnitude of these effects was considerably less than for the best MEK, RAF, and IGF1R inhibitors. Rho kinase (ROCK) and proteasome inhibitors did not show selectivity as single agents, although combination inhibition of these pathways is selectively toxic for KRAS-mutant cells, especially in vivo (11, 12). As illustrated in the viability graphs in Fig. 1 and Supplementary Fig. S1, drugs directed against the same target tend to cluster together in a heatmap analysis (Supplementary Fig. S1C), providing a degree of reassurance with respect to the reproducibility and on-target nature of these differential effects.
. | . | . | Two-way ANOVA (MUT vs. WT) . | ||
---|---|---|---|---|---|
. | Drug . | Target . | Dose range (nmol/L) . | P . | Significance . |
Trametinib | MEK | 1,250-9.76 | 0.0031 | ** | |
PD-0325901 | MEK | 1,250-9.76 | 0.0053 | ** | |
Selumetinib | MEK | 1,250-9.76 | 0.0181 | * | |
CI-1040 | MEK | 1,250-9.76 | 0.0123 | * | |
AZ628 | RAF | 2,500-19.5 | 0.0037 | ** | |
RAF/MEK/ERK | L779450 | RAF | 20,000-156 | 0.0274 | * |
PLX4720 | RAF | 20,000-156 | 0.0116 | * | |
GDC-0879 | RAF | 20,000-156 | 0.0483 | * | |
Sorafenib | RAF | 10,000-78 | 0.0285 | * | |
ZM336372 | RAF | 20,000-156 | 0.0924 | ns | |
GW5074 | RAF | 10,000-78 | 0.1282 | ns | |
SB590885 | RAF | 2,0000-156 | 0.2135 | ns | |
GDC0941 | PI3K | 5,000-39 | 0.6078 | ns | |
PIK-90 | PI3K | 2,500-19.5 | 0.7292 | ns | |
PIK-75 | PI3K (p110α) | 250-1.95 | 0.2477 | ns | |
NVP-BEZ235 | PI3K/mTOR | 250-1.95 | 0.2202 | ns | |
PF-04691502 | PI3K/mTOR | 5,000-39 | 0.9707 | ns | |
PI3K/AKT/mTOR | PP242 | mTOR (kinase) | 2,0000-156 | 0.6741 | ns |
AZD8055 | mTOR (kinase) | 5,000-39 | 0.5811 | ns | |
Everolimus | mTOR (rapalog) | 5,000-39 | 0.8585 | ns | |
Temsirolimus | mTOR (rapalog) | 5,000-39 | 0.9338 | ns | |
Akti- 1/2 | AKT | 20,000-156 | 0.2065 | ns | |
MK-2206 | AKT | 20,000-156 | 0.9727 | ns | |
NVP-AEW541 | IGF1R | 5,000-39 | 0.0042 | ** | |
OSI-906 | IGF1R | 10,000-78 | 0.0041 | ** | |
BMS-754807 | IGF1R | 5,000-39 | 0.0014 | ** | |
RTK | Picropodophyllin | IGF1R | 1,000-7.8 | 0.4921 | ns |
IGF-1R Inhibitor II | IGF1R | 20,000-156 | 0.5752 | ns | |
Erlotinib | EGFR | 5,000-39 | 0.1073 | ns | |
Gefitinib | EGFR | 5,000-39 | 0.0139 | ns | |
17-AAG | HSP90 | 500-3.9 | 0.0355 | * | |
17-DMAG | HSP90 | 500-3.9 | 0.0401 | * | |
BIIB021 | HSP90 | 500-3.9 | 0.1456 | ns | |
NVP-AUY922 | HSP90 | 500-3.9 | 0.5857 | ns | |
BX-795 | TBK1 | 5,000-39 | 0.1786 | ns | |
PF-02341066 | c-Met | 5,000-39 | 0.0479 | * | |
SU11274 | c-Met | 5,000-39 | 0.4032 | ns | |
Bortezomib | Proteasome | 250-1.95 | 0.406 | ns | |
MG-132 | Proteasome | 5,000-39 | 0.3896 | ns | |
PSI | Proteasome | 2,500-19.5 | 0.8714 | ns | |
Doxorubicin | Topoisomerase | 1,125-8.9 | 0.1158 | ns | |
Topotecan | Topoisomerase | 2,500-19.5 | 0.4927 | ns | |
OTHER | BMS-345541 | IKK-β | 20,000-156 | 0.3172 | ns |
SC-514 | IKK-2 | 20,000-156 | 0.9998 | ns | |
CAY10576 | IKK-e | 20,000-156 | 0.5216 | ns | |
5Z-7-Oxozeaenol | TAK1 (NF-κB) | 20,000-156 | 0.2505 | ns | |
Fasudil | ROCK | 20,000-156 | 0.8516 | ns | |
Y-27632 | ROCK | 20,000-156 | 0.9011 | ns | |
Docetaxel | Microtubule | 10-0.078 | 0.3609 | ns | |
MK2a Inhibitor | MK2 | 5,000-39 | 0.1352 | ns | |
Deguelin | MT-bioenergetics | 20,000-156 | 0.7901 | ns | |
10058-F4 | c-Myc | 20,000-156 | 0.2072 | ns | |
PF-573,228 | FAK | 20,000-156 | 0.4752 | ns | |
GDC-0449 | Hedgehog pathway | 20,000-156 | 0.6057 | ns | |
Dasatinib | SRC | 5,000-39 | 0.1236 | ns |
. | . | . | Two-way ANOVA (MUT vs. WT) . | ||
---|---|---|---|---|---|
. | Drug . | Target . | Dose range (nmol/L) . | P . | Significance . |
Trametinib | MEK | 1,250-9.76 | 0.0031 | ** | |
PD-0325901 | MEK | 1,250-9.76 | 0.0053 | ** | |
Selumetinib | MEK | 1,250-9.76 | 0.0181 | * | |
CI-1040 | MEK | 1,250-9.76 | 0.0123 | * | |
AZ628 | RAF | 2,500-19.5 | 0.0037 | ** | |
RAF/MEK/ERK | L779450 | RAF | 20,000-156 | 0.0274 | * |
PLX4720 | RAF | 20,000-156 | 0.0116 | * | |
GDC-0879 | RAF | 20,000-156 | 0.0483 | * | |
Sorafenib | RAF | 10,000-78 | 0.0285 | * | |
ZM336372 | RAF | 20,000-156 | 0.0924 | ns | |
GW5074 | RAF | 10,000-78 | 0.1282 | ns | |
SB590885 | RAF | 2,0000-156 | 0.2135 | ns | |
GDC0941 | PI3K | 5,000-39 | 0.6078 | ns | |
PIK-90 | PI3K | 2,500-19.5 | 0.7292 | ns | |
PIK-75 | PI3K (p110α) | 250-1.95 | 0.2477 | ns | |
NVP-BEZ235 | PI3K/mTOR | 250-1.95 | 0.2202 | ns | |
PF-04691502 | PI3K/mTOR | 5,000-39 | 0.9707 | ns | |
PI3K/AKT/mTOR | PP242 | mTOR (kinase) | 2,0000-156 | 0.6741 | ns |
AZD8055 | mTOR (kinase) | 5,000-39 | 0.5811 | ns | |
Everolimus | mTOR (rapalog) | 5,000-39 | 0.8585 | ns | |
Temsirolimus | mTOR (rapalog) | 5,000-39 | 0.9338 | ns | |
Akti- 1/2 | AKT | 20,000-156 | 0.2065 | ns | |
MK-2206 | AKT | 20,000-156 | 0.9727 | ns | |
NVP-AEW541 | IGF1R | 5,000-39 | 0.0042 | ** | |
OSI-906 | IGF1R | 10,000-78 | 0.0041 | ** | |
BMS-754807 | IGF1R | 5,000-39 | 0.0014 | ** | |
RTK | Picropodophyllin | IGF1R | 1,000-7.8 | 0.4921 | ns |
IGF-1R Inhibitor II | IGF1R | 20,000-156 | 0.5752 | ns | |
Erlotinib | EGFR | 5,000-39 | 0.1073 | ns | |
Gefitinib | EGFR | 5,000-39 | 0.0139 | ns | |
17-AAG | HSP90 | 500-3.9 | 0.0355 | * | |
17-DMAG | HSP90 | 500-3.9 | 0.0401 | * | |
BIIB021 | HSP90 | 500-3.9 | 0.1456 | ns | |
NVP-AUY922 | HSP90 | 500-3.9 | 0.5857 | ns | |
BX-795 | TBK1 | 5,000-39 | 0.1786 | ns | |
PF-02341066 | c-Met | 5,000-39 | 0.0479 | * | |
SU11274 | c-Met | 5,000-39 | 0.4032 | ns | |
Bortezomib | Proteasome | 250-1.95 | 0.406 | ns | |
MG-132 | Proteasome | 5,000-39 | 0.3896 | ns | |
PSI | Proteasome | 2,500-19.5 | 0.8714 | ns | |
Doxorubicin | Topoisomerase | 1,125-8.9 | 0.1158 | ns | |
Topotecan | Topoisomerase | 2,500-19.5 | 0.4927 | ns | |
OTHER | BMS-345541 | IKK-β | 20,000-156 | 0.3172 | ns |
SC-514 | IKK-2 | 20,000-156 | 0.9998 | ns | |
CAY10576 | IKK-e | 20,000-156 | 0.5216 | ns | |
5Z-7-Oxozeaenol | TAK1 (NF-κB) | 20,000-156 | 0.2505 | ns | |
Fasudil | ROCK | 20,000-156 | 0.8516 | ns | |
Y-27632 | ROCK | 20,000-156 | 0.9011 | ns | |
Docetaxel | Microtubule | 10-0.078 | 0.3609 | ns | |
MK2a Inhibitor | MK2 | 5,000-39 | 0.1352 | ns | |
Deguelin | MT-bioenergetics | 20,000-156 | 0.7901 | ns | |
10058-F4 | c-Myc | 20,000-156 | 0.2072 | ns | |
PF-573,228 | FAK | 20,000-156 | 0.4752 | ns | |
GDC-0449 | Hedgehog pathway | 20,000-156 | 0.6057 | ns | |
Dasatinib | SRC | 5,000-39 | 0.1236 | ns |
NOTE: Cell viability was measured across an 8-point titration range. Two-way ANOVA was used to examine significant differences in sensitivity between KRAS-mutant (MUT) and wild-type (WT) cells. Only primary drug targets are indicated.
Abbreviation: ns, not significant.
In summary, we found that NSCLC cells harboring a KRAS-mutant allele are, in general, more sensitive to MEK, RAF, and IGF1R inhibitors than cells with wild-type KRAS. No obvious differences were seen in this between the different amino acid changes at codons 12, 13, or 61 in the KRAS-mutant cell lines used.
IGF1R Inhibitors Selectively Inhibit AKT Activation in KRAS-Mutant NSCLC Cells
To investigate the mechanistic basis for the different response of NSCLC cell lines to MEK and IGF1R inhibitors, we examined the effect of these compounds on the activity of the MEK/ERK and PI3K/AKT pathways. As expected, we observed efficient reduction of ERK phosphorylation upon treatment with the MEK inhibitor PD-0325901 across the entire cell panel (Fig. 2A and Supplementary Fig. S2). In addition, there was a modest and persistent increase in AKT phosphorylation in both genotypes, probably due to suppression of well-characterized negative feedback loops (18–20). Interestingly, MEK inhibition in KRAS-mutant, but not wild-type, cells produced a striking reduction in ribosomal protein S6 phosphorylation, an indirect measure of mTOR complex 1 (mTORC1) activity, which became evident at later time points, possibly indicating a more indirect mechanism. Consistent with this finding, we also found reduced phosphorylation on Thr389 of the direct mTORC1 substrate p70 ribosomal protein S6 kinase (p70S6K) after MEK inhibitor treatment of KRAS-mutant cells.
In response to IGF1R inhibition by NVP-AEW541, cells harboring a KRAS mutation showed an early, marked suppression of AKT phosphorylation that was sustained at 24 hours (Fig. 2B and Supplementary Fig. S3A). Consistent with this finding, there was a strong reduction in phosphorylation of the AKT substrate PRAS40 on Thr246. Notably, these effects were not evident in KRAS wild-type cells, even though treatment with AKT or PI3K inhibitors produced the same level of reduction in AKT phosphorylation in both KRAS-mutant and wild-type cells (Supplementary Fig. S3B). These data suggest that inhibition of IGF1R has a clear impact upon the reduction of PI3K activity only in the cells carrying a KRAS mutation. Moreover, the change in AKT phosphorylation seen at 4 hours after NVP-AEW541 treatment correlated strongly with the effect on cell viability after a 72-hour treatment (Fig. 2B, right). Thus, the differences in the reduction of AKT phosphorylation may provide an explanation as to why KRAS-mutant NSCLC cells are more sensitive to IGF1R inhibition.
Combining IGF1R Inhibitors with MEK Inhibitors Enhances Their Differential Impact upon Mutant KRAS–Driven Lung Cancer
The data presented above show that KRAS-mutant NSCLC cells are preferentially sensitive to inhibition of both MEK and IGF1R, and that IGF1R inhibition reduces AKT phosphorylation only in KRAS-mutant cells. Thus, a combination of both drugs would allow for simultaneous inhibition of the PI3K/AKT and MEK/ERK pathways selectively in KRAS-mutant cells and might be expected to increase the differential sensitivity between KRAS-mutant and wild-type cells.
To explore this possibility, we examined the effect of a combination of NVP-AEW541 with PD-0325901 upon the activity of MEK/ERK and PI3K/AKT signaling pathways after a 4-hour treatment (Fig. 2C and Supplementary Fig. S4). As expected, this combination decreased ERK phosphorylation in both mutant and wild-type cells with no differences as compared with the effect of MEK inhibitor alone. Moreover, the combination reduced AKT phosphorylation only in KRAS-mutant cells with the effects being comparable to those seen with the IGF1R inhibitor alone. Phosphorylation on Tyr612 of the adaptor protein insulin receptor substrate 1 (IRS1) served as an additional monitor of IGF1R pathway inhibition by NVP-AEW541 both alone and in combination. Intriguingly, combined inhibition of MEK and IGF1R led to a more robust inhibition of S6 phosphorylation in KRAS-mutant cells. Consistent with this, a corresponding effect was also evident when we looked at phosphorylation of the S6 upstream kinase p70S6K. These data indicate that the combination of MEK and IGF1R inhibitors in KRAS-mutant cells causes not only a combined inhibition of PI3K/AKT and MEK/ERK pathways, but also a stronger inhibition of mTORC1 activity.
To assess the effect of drug combinations further, we augmented NVP-AEW541 with low doses of PD-0325901 and found that this reduced cell viability more strongly than single agent in KRAS-mutant cells but not in wild-type cells (Supplementary Fig. S5A). This synergistic effect was associated with an increased induction of apoptosis, at least in some cell lines (Supplementary Fig. S5B). Comparison of the IC60 values (drug dose leading to 60% survival relative to untreated cells) showed that in most KRAS-mutant cells, the combination of NVP-AEW541 with PD-0325901 clearly reduced the IC60 value, whereas no significant differences were observed in most KRAS wild-type cells (Supplementary Fig. S6A). This increase in the differential effect between KRAS-mutant and wild-type cells could be seen across a range of doses of NVP-AEW541 and was also evident when we compared the average response of each KRAS genotype (Fig. 3A). Interestingly, the combination of NVP-AEW541 with low doses of the potent pan-RAF inhibitor AZ628 showed similar effects (Fig. 3B). These results could be replicated with an alternative IGF1R inhibitor, OSI-906 (Supplementary Fig. S6C–S6E) and with trametinib (GSK1120212), an alternative MEK inhibitor (Supplementary Fig. S6B and S6F). Furthermore, the combination of IGF1R and MEK inhibitors in a long-term cell-growth assay also showed a strong relative reduction of cell viability in KRAS-mutant cells (Supplementary Fig. S6G).
Combination treatment with PI3K and MEK inhibitors has previously shown efficacy in Kras-mutant lung tumor mouse models (8). We therefore decided to assess the effect of combining a PI3K inhibitor with low doses of a MEK or RAF inhibitor in the panel of NSCLC cell lines. Although treatment with PI3K inhibitors alone showed no selectivity between wild-type and mutant cells, KRAS-mutant cells exhibited enhanced sensitivity to the combination of PI3K and MEK inhibitors (Fig. 3C). The addition of a MEK or RAF inhibitor to the PI3K inhibitor GDC0941 increased the sensitivity of KRAS-mutant but not KRAS wild-type cells (Fig. 3C and D and Supplementary Fig. S6H–I), but the enhanced genotype-specific differential effect was, in general, less striking than that seen with IGF1R and MEK inhibitor combinations, due mainly to the stronger impact of direct PI3K inhibition on KRAS wild-type cells.
The fact that the IGF1R inhibitors used in this study are known to inhibit the closely related insulin receptor (INSR) to varying degrees prompted us to use siRNAs directed against IGF1R or INSR as a means to assess the effects of abrogating the activity of each receptor individually. Silencing of IGF1R expression in the panel of NSCLC cells led to a significant loss of viability of KRAS-mutant cells as compared with KRAS wild-type counterparts, whereas knockdown of INSR produced rather minor effects (Fig. 3E). In keeping with our observations using IGF1R inhibitors, IGF1R knockdown strikingly reduced AKT phosphorylation in KRAS-mutant cells, with INSR silencing producing no such response (Supplementary Fig. S7A), and the combination of IGF1R knockdown with MEK inhibition augmented the KRAS-mutant genotype-specific effect on cell viability (Fig. 3F).
To investigate the possible utility of drug combinations in an in vivo setting, we sought to assess the impact of MEK and IGF1R inhibition on the maintenance and progression of Kras-driven lung tumors in 2 different autochthonous genetically engineered mouse models. We elected to use trametinib for MEK inhibition due to both its potency at low concentrations in vitro (Fig. 1D and Supplementary Fig. S6B, S6F, and G) and its long half-life in vivo (21). In addition, alone of the MEK inhibitors, this drug has proven to be effective in a clinical trial on BRAF-mutant melanoma (22). Accordingly, KrasLA2-G12D/+ mice (23) were allowed to develop lung tumors that could be readily detected by micro-computerized tomography (CT) scanning. Animals were then treated daily with either vehicle, IGF1R inhibitor NVP-AEW541, MEK inhibitor trametinib, or a combination of both inhibitors, for 6 weeks and were scanned again at the end of the treatment period. The change in volume of individual tumors over time was then evaluated. Individual lung tumors arising in KrasLA2-G12D/+ mice tend to grow relatively slowly and, as anticipated, tumors that were longitudinally tracked in vehicle control-treated animals generally exhibited a modest increase in size over the treatment period. Nevertheless, we observed that tumors in mice treated with individual MEK or IGF1R inhibitors showed a small decrease in mean tumor volume and that this effect was exacerbated when the inhibitors were combined (Fig. 3G). The efficacy of each inhibitor in this in vivo context is illustrated in Supplementary Fig. S7B. Analysis of individual tumor nodules at the conclusion of the treatment regime showed that IGF1R inhibition had produced a clear, albeit incomplete, reduction in AKT phosphorylation, and MEK inhibition resulted in the total abrogation of ERK phosphorylation (Supplementary Fig. S7B). To evaluate the effect of MEK and IGF1R inhibition in a more aggressive Kras-driven mouse lung tumor model, we inoculated the lungs of KrasLSL-G12D; Trp53Flox/Flox mice with adenovirus-expressing Cre recombinase to induce concomitant activation of oncogenic KRAS and deletion of the tumor suppressor p53 (24). Mice were scanned by micro-CT to identify development of individual lung tumors, and tumor-bearing animals were then treated daily with either vehicle, MEK inhibitor trametinib, IGF1R inhibitor OSI-906, or a combination of both inhibitors for 2 weeks. After rescanning at the end of the treatment period, changes in the volume of individual tumors over this time frame were calculated for each group (Fig. 3H). Although tumors that develop in this mouse model tend to grow more rapidly than those in the KrasLA2-G12D/+ model, we observed a similar response to MEK and IGF1R inhibition. Targeting each pathway individually provided some reduction in tumor growth, but inhibiting both pathways simultaneously had a considerably stronger impact. Taken together, our results suggest the combination of IGF1R and MEK inhibitors as a novel potential therapy for KRAS-mutant NSCLC.
KRAS-Mutant NSCLC Cells Exhibit Increased Dependence on IGF1R Signaling
The IGF1R pathway is activated by insulin-like growth factors binding to the heterotetrameric IGF1 RTK, resulting in receptor autophosphorylation, binding to the IRS adaptor proteins, IRS protein tyrosine phosphorylation, and subsequent binding to effector enzymes such as the regulatory p85 subunit of PI3K. To investigate the differential effect of IGF1R inhibition on PI3K activity in NSCLC cells, we analyzed the activity of the IGF1R pathway in 12 cell lines, 6 of which are KRAS-mutant and 6 KRAS wild-type. Cells were serum-starved overnight and then stimulated for 30 minutes with either IGF1 or EGF. A phosphospecific antibody recognizing Tyr612 of the IGF1R adaptor protein IRS1 (and also equivalent Tyr653 on IRS2) was used to measure activation of the IGF1R pathway; these sites, when phosphorylated, bind to p85, leading to PI3K activation. IGF1 stimulation induced a strong increase in phospho-IRS and phospho-AKT in all 6 KRAS-mutant cell lines tested, whereas only 3 of 6 wild-type cells showed activation of the IGF1R pathway (Fig. 4A and Supplementary Fig. S8A). As described above, cells carrying KRAS mutations showed a marked suppression in steady-state AKT phosphorylation in response to IGF1R inhibition by NVP-AEW541; in contrast, treatment with the EGFR inhibitor erlotinib did not affect AKT phosphorylation (Fig. 4B and Supplementary Fig. S8B). KRAS wild-type cells showed a higher degree of variability in their responses to IGF1R and EGFR inhibition. IGF1R inhibition decreased phospho-AKT only in the 3 cell lines that were responsive to IGF1 stimulation, although the magnitude of this effect was much less pronounced than in KRAS-mutant cells. Moreover, the wild-type cells in general also showed a more prominent decrease in AKT phosphorylation in response to EGFR inhibition. In keeping with these observations, KRAS-mutant cells generally expressed higher steady-state levels of phospho-IRS1, whereas KRAS wild-type cells had higher levels of phospho-EGFR (Supplementary Fig. S8C). To explore further the activation of PI3K in this collection of NSCLC cell lines, we analyzed the binding of IRS adaptor proteins to p85α. Immunoprecipitation of p85α led to the clear coprecipitation of IRS1 and/or IRS2 in the KRAS-mutant cells, whereas coprecipitation of either of these IRS proteins from KRAS wild-type cells was barely detectable (Fig. 4C). Taken together, these results suggest that cells harboring KRAS mutations have an IGF1R pathway with strong basal activity and that this pathway is critical for PI3K activation.
To assess the relative expression levels of known regulators of the IGF1R pathway between the KRAS-mutant and wild-type genotypes, we isolated mRNA from the large NSCLC cell panel and conducted quantitative PCR analysis on several components of the pathway, including the genes encoding the receptors (IGF1R, IGF2R, INSR), ligands (IGF1, IGF2), IGF-binding proteins (IGFBPs 1–6), and adaptors (p85α, GRB10, IRS1, and IRS2). The results showed that while levels of most mRNAs were very similar across the different genotypes, KRAS-mutant cells expressed modestly higher levels of IRS1 than wild-type cells. Moreover, although the values did not reach statistical significance, KRAS-mutant cells also exhibited increased levels of IRS2 (Fig. 4D and data not shown). Interestingly, analysis of publicly available gene expression data emerging from 2 independent large-scale cancer cell line projects (25, 26) indicates that, in general, expression levels of IRS1 are elevated in KRAS-mutant lung cancer cell lines relative to KRAS wild-type comparators (Supplementary Fig. S8D and S8E). In addition, KRAS-mutant lung adenocarcinoma tissue samples (27) exhibited increased expression of both IRS2 and IGF1R (Supplementary Fig. S8E). Finally, we analyzed the dependence of the NSCLC cell line panel upon IRS1 and/or IRS2 expression by conducting siRNA-mediated gene knockdown. Depletion of IRS1, IRS2, or both together produced a selective decrease in cell viability, accompanied by an increase in apoptosis in the KRAS-mutant cells that was comparable to the effects elicited by control KRAS siRNA treatment (Fig. 4E and see also Fig. 1A). These data are consistent with the higher degree of sensitivity of KRAS-mutant NSCLC cells to IGF1R inhibition by targeted small molecules and support the notion that KRAS-mutant cells display an increased reliance upon IGF1R signaling for their survival.
KRAS Depletion Attenuates AKT Activation in KRAS-Mutant NSCLC Cells
To investigate whether the loss of KRAS expression in lung cancer cells leads to the suppression of PI3K as well as ERK pathway activation, we assessed the impact of KRAS knockdown using 2 different siRNA pools in 12 cell lines, 6 of which are KRAS mutant and 6 KRAS wild-type. We observed that acute loss of KRAS expression led to a striking reduction in ERK phosphorylation that was much more evident in KRAS-mutant cells. In addition, the mutant cells exhibited a similarly strong and selective reduction in S6 phosphorylation. Moreover, we found that KRAS depletion also significantly diminished AKT activation, monitored by phosphorylation of AKT on either Ser473 or Thr308 or PRAS40 on Thr246, preferentially in KRAS-mutant NSCLC cells, albeit to a lesser extent than its impact upon phospho-ERK and phospho-S6 (Fig. 5A and Supplementary Fig. S9A).
The fact that mTORC1 activity, as indicated by S6 phosphorylation, is sensitive to MEK inhibition (Fig. 2A) and to KRAS knockdown (Fig. 5A and Supplementary Fig. S9A) in KRAS-mutant NSCLC cells suggested that the established negative regulatory feedback loop involving phosphorylation of IRS1 by mTORC1 directly or via S6K1 (28–30) may play a significant role in the control of PI3K activity in these cells. Thus, when MEK and S6K are inhibited following KRAS knockdown, loss of negative feedback means there is a tendency to increase IGF1R signaling via IRS to PI3K/AKT, which counteracts any possible direct impact of KRAS loss on PI3K activation. We therefore sought to assess the effect of inhibiting this feedback loop upon AKT phosphorylation by treating cells with rapamycin in both the presence and absence of KRAS expression. As illustrated in Fig. 5B and Supplementary Fig. S9B, rapamycin treatment of control siRNA-transfected KRAS-mutant NSCLC cells increased the levels of phospho-AKT, indicating the presence of an intact feedback loop. Nevertheless, rapamycin was clearly unable to enhance AKT activation following acute depletion of KRAS expression, emphasizing the extent of the KRAS knockdown-induced decrease in AKT activation, even in cell lines such as H1792 where the effect of KRAS knockdown alone is less striking. Taken together, these data suggest that direct interaction of KRAS with p110 may play a critical role in the control of PI3K signaling in NSCLC cells.
Activation of PI3K by Acute Oncogenic RAS Signaling Is Sensitive to IGF1R Inhibition
To look further into the influence of oncogenic RAS activity on IGF1R-mediated survival signaling, we sought to analyze the effect of acute oncogenic RAS activation in untransformed human epithelial cells. To this end, we stably introduced a 4-hydroxytamoxifen (4-OHT)-regulatable oncogenic RAS chimeric protein, ER:HRAS V12 (31), into the spontaneously immortalized breast epithelial cell line MCF10A. The addition of 4-OHT to these cells led to the activation of RAS downstream signaling in a time-dependent fashion, as evidenced by the sustained increase in ERK and AKT phosphorylation (Supplementary Fig. S10A). As anticipated, pretreatment of MCF10A/ER:HRAS V12 cells with MEK inhibitors led to the abrogation of ERK phosphorylation in response to short-term 4-OHT stimulation, with no effect on AKT phosphorylation (Fig. 6A). More notably, pretreatment of the cells with IGF1R inhibitors led to the ablation of residual and 4-OHT–inducible IRS1 phosphorylation, along with a striking inhibition of AKT phosphorylation in response to RAS activation (Fig. 6A). To rule out possible RAS isoform-specific effects, we first established that these observations could be replicated in the same cell system expressing a 4-OHT–activatable ER:KRAS V12 chimeric protein (ref. 10; Supplementary Fig. S10B). Next, to extend our findings to an untransformed lung epithelial cell context, we stably expressed ER:KRAS V12 in NL-20 (32) and type II pneumocyte (33) cells, immortalized human cell lines derived from bronchial and alveolar epithelia, respectively. Figure 6B and C show that the short-term activation of an oncogenic KRAS signal in each of these cell lines led to the marked increase in phosphorylation of ERK and AKT, albeit from a higher basal level than seen in the MCF10A cells. Importantly, as in the MCF10A cell background, pretreatment of the cells with IGF1R inhibitors effectively blocked the 4-OHT–induced phosphorylation of AKT. Finally, to investigate the acute activation of oncogenic RAS signaling in a cancer cell context, we stably expressed ER:HRAS V12 in the NSCLC cell line SK-MES-1, which is wild-type for KRAS and only very modestly sensitive to IGF1R inhibitors. A short 4-hour stimulation of SK-MES-1/ER:HRAS V12 with 4-OHT was also able to induce both ERK and AKT phosphorylation. Moreover, the activation of AKT was again sensitive to prior inhibition of IGF1R, although not completely blocked, whereas ERK activation remained unaffected (Fig. 6D). As shown in Fig. 4B, the phosphorylation of AKT in SK-MES-1 NSCLC cells was also sensitive to inhibition of EGFR by erlotinib. We therefore assessed the effect of pretreating SK-MES-1/ER:HRAS V12 cells with the EGFR inhibitor erlotinib, or a combination of NVP-AEW541 and erlotinib, before 4-OHT induction. Figure 6D illustrates that erlotinib inhibited RAS-induced AKT activation to a similar level as NVP-AEW541, implying a significant input from EGFR as well as IGF1R in these cells. Furthermore, the combination of both of these targeted inhibitors was able to provide a near-complete blockade of AKT phosphorylation in response to 4-OHT. In sum, these observations confirm that inhibition of IGF1R is able to blunt the activation of AKT elicited by acute induction of RAS signaling and further suggest that context-dependent input from other RTKs can also play a notable role. As a whole, our data support the contention that PI3K activation is controlled by coordinate input from RAS proteins and RTKs, and that in KRAS-mutant NSCLC the predominant RTK in this regard is the IGF1R (Fig. 6E).
Discussion
In the standard model of RAS-driven tumorigenesis, oncogenic RAS protein is thought to induce the activity of a number of downstream effector enzyme families by direct interaction of GTP-bound RAS with its targets, including RAF kinases, PI3K isoforms, and guanine nucleotide exchange factors for RAL GTPases (4, 34). In the case of type I PI3K, GTP-bound RAS can interact directly with the RAS-binding domain (RBD) on the catalytic p110 subunits (35–39), leading to enzymatic activation. The interaction of RAS.GTP with p110 promotes allosteric activation of PI3K in a manner that is cooperative with signal inputs from RTKs, which act through the binding of tyrosine phosphorylated sequences to the p85 regulatory subunit, relieving its autoinhibitory function (37, 38, 40). The ability of RAS to interact with p110α has been shown to be essential for mutant Kras-induced lung cancer formation and mutant Hras-induced skin cancer formation in mouse models (41).
The ability of RAS to activate both RAF and PI3K directly has led to great interest in the possibility of treating RAS-mutant tumors by inhibiting both pathways in combination. The use of PI3K and MEK inhibitors in a mouse model of Kras-induced lung cancer has provided support for this idea (8). However, although it has been shown that, once established, RAS-mutant cancers show dependence on PI3K signaling for tumor maintenance (42), it is not yet clear whether this is due to direct RAS–PI3K interaction or some more indirect mechanism. It is also not certain that RAS-mutant cancer cells show any greater degree of dependence on PI3K signaling than do cells with other genotypes, raising the issue of whether or not PI3K inhibitors will have a useful therapeutic window in the treatment of RAS-mutant cancers. We therefore undertook the drug screening approach described here to look for agents with selectivity for RAS-mutant relative to RAS wild-type lung cancer cell lines. The results show that while PI3K inhibition was toxic to cultured RAS-mutant cells, it was not obviously any more selective for cells with RAS mutations compared to cells with other genotypes. This is in contrast to the finding that RAF/MEK/ERK pathway function is indeed selectively required by RAS-mutant cells, as has been described with increasing certainty by others in recent years (16, 25, 26, 43, 44). In addition, we unexpectedly found that RAS-mutant lung cancer cell lines very clearly showed heightened sensitivity to RTK inhibitors targeting the IGF1 receptor. It is worth noting that these KRAS-mutant genotype-specific effects of RAF/MEK and IGF1R inhibition are also present in data available from the Genomics of Drug Sensitivity in Cancer project from the Wellcome Trust Sanger Institute (Cambridgeshire, United Kingdom; ref. 26), based on large-scale drug screening of several hundred cell lines derived from a broad range of tissue types: mutant KRAS selectivity is seen with AZ628 (RAF inhibitor), PD-0325901, selumetinib and RDEA119 (MEK inhibitors), and BMS-754807 and OSI-906 (IGF1R inhibitors).
A study of KRAS-mutant colon cancer cell lines recently reported a clear tendency toward sensitivity to IGF1R inhibition (45). In this work, as in our work on KRAS-mutant lung cancer cell lines, RAS-mutant cells showed good sensitivity to combinations of MEK and IGF1R inhibitors, and there were indications that basal PI3K signaling was dependent on signaling flux through IGF1R to IRS1/IRS2 to p85/p110. However, while the therapeutic implications of our work and that of Ebi and colleagues (46) are similar, different mechanistic interpretations were made. In contrast to our analysis here, Ebi and colleagues did not see a negative impact of removal of KRAS by RNAi knockdown on PI3K activity in KRAS-mutant cells. The basis for this difference is unclear. One possibility is that it reflects the differing tissue types of origin of the cells; the frequency of coincident mutation of KRAS and PIK3CA in colon but not lung cancer suggests that there might be significant differences in the interplay between these signaling systems in the 2 tissues. A quantitative model of RAS signaling to PI3K concludes that the relative contributions of RAS and RTKs to PI3K activation depend strongly on the quantities and binding affinities of the interacting proteins, which are likely to vary greatly across different cell types and stimuli (46). Alternatively, this might reflect differences in the efficiency of KRAS knockdown between the short hairpin RNA (shRNA) and siRNA approaches used. It is possible that RAS protein expression has to be reduced below different thresholds to have an impact on RAF and on PI3K signaling. The tendency of MEK and mTOR inhibition to cause PI3K activation due to relief of negative feedback onto IRS1 can also obscure the direct impact of loss of RAS expression on PI3K activity, which can be revealed when mTOR activity is artificially inhibited by rapamycin, as shown in Fig. 5.
The use of a posttranslationally activatable form of oncogenic RAS allows more precise probing of the role of RAS in PI3K regulation, including in a time frame that will be minimally affected by RAS pathway–induced changes in gene expression. From this, it is clear that short-term RAS activation can result in stimulation of PI3K, but that this is dependent on input from the IGF1R tyrosine kinase. It is thus likely that RAS requires relief of the inhibitory effect of the unliganded p85 regulatory subunit of PI3K (47) to be able to effectively activate its lipid kinase activity through direct RAS–p110 interaction, and that, in KRAS-mutant lung cancer, this signaling input into p85 is provided by basal IGF1R signaling. This effect was seen in untransformed immortalized breast epithelial cells and also in 2 different cultures of normal immortalized lung epithelial cells with posttranslationally inducible RAS activity. We also tested this in an NSCLC line lacking KRAS mutation. Although this showed dependence of RAS-induced PI3K pathway activation on IGF1R function, there was also a component of EGFR dependence. It is likely that this reflects the mixed IGF1R and EGFR dependence of the parental KRAS wild-type SK-MES-1 cell line, whereas the KRAS-mutant NSCLC lines seem to be much more dependent on IGF1R rather than EGFR signaling (Fig. 4B). We speculate that in this inducible system, acutely activated RAS will use input from whatever basally active RTK is present in the cells to relieve p85 mediated autoinhibition of PI3K activity; in KRAS-mutant NSCLC this is predominantly IGF1R, whereas in KRAS wild-type NSCLC both IGF1R and EGFR contribute.
The findings described here using cultured lung cancer cell lines and also mouse lung cancer models suggest that there may be value to the use of combinations of MEK and IGF1R inhibitors to treat patients with KRAS-mutant lung cancer. The work reported here has used small-molecule kinase inhibitors that target both IGF1R and the related insulin receptor; further work will be required to determine the relative merits in this context of these inhibitors compared with IGF1R-directed monoclonal antibodies, which generally do not target the insulin receptor. In comparison with PI3K inhibitors, IGF1R inhibitors seem to have less single-agent impact on KRAS wild-type cells, suggesting that these agents might show less toxicity in vivo. However, to date, IGF1R inhibitors have not shown great promise as single agents in clinical trials, with the exception of on some sarcomas (48). With the MEK inhibitor trametinib clearly now an attractive candidate for the treatment of KRAS-mutant NSCLC, our work suggests that early combination with an IGF1R inhibitor may be beneficial.
Methods
Cell Lines and Culture
MCF10A/ER:HRAS V12 and SK-MES-1/ER:HRAS V12 cells were constructed by transducing parental MCF10A breast epithelial cells or SK-MES-1 NSCLC cells with a bleocin-resistant retrovirus encoding the murine ecotropic receptor. Selected cells were subsequently infected with puromycin-resistant ER:HRAS V12 retrovirus (31). MCF10A/ER:KRAS V12, NL-20/ER:KRAS V12 and TypeII/ER:KRAS V12 were constructed by transducing parental MCF10A, NL-20, or TypeII cells with pLenti-PGK-ER-KRAS(G12V) (Haber lab Addgene plasmid no. 35635) and selecting under hygromycin. Detailed origin and growing conditions of all cell lines used are given in the Supplementary Material. Cell lines were authenticated by the Cancer Research UK Central Cell Services facility using short tandem repeat profiling.
siRNA Reagents and Cell Viability Assays
All siRNAs were obtained from Dharmacon and were used as “SMARTpools” according to the manufacturer's instructions. Viability assays following siRNA transfection experiments or the addition of small-molecule inhibitors were conducted in 96-well format as previously described (12). Starting cell density was optimized to produce an 80% confluent monolayer in mock-treated cells at the conclusion of the experiment. Cell viability was determined using Cell Titer Blue (Promega), and apoptosis induction was recorded using a caspase-3/7 consensus site peptide (Z-DEVD)2 conjugated to rhodamine 110 (Invitrogen; ref. 12). For long-term drug treatments, cells were seeded in 12-well format for 24 hours and treated with drugs for 12 days. Cells were fixed with 2% paraformaldehyde, stained with 0.2% crystal violet, and finally dissolved with acetic acid. Absorbance was measured at 595 nm. Detailed information on the small inhibitors used is given in the Supplementary Material.
Western Blotting
For quantitative Western blotting, bound primary antibodies were detected by secondary conjugates compatible with infrared detection at 700 nm and 800 nm, and membranes were scanned using the Odyssey Infrared Imaging System (Odyssey, LICOR). Alternatively, membranes were incubated with horseradish peroxidase-conjugated secondary antibody, detected using chemiluminescence (Millipore), and quantified using Image Quant LAS4000 (GE Healthcare). Detailed information on the antibodies used is given in the Supplementary Material.
Co-Immunoprecipitations
Cells growing under steady-state conditions were scraped into ice-cold lysis buffer comprising 25 mmol/L Tris pH 7.6, 150 mmol/L NaCl, 0.5% Nonidet P-40, 0.5 mmol/L DTT, 1 mmol/L EDTA, 1 mmol/L EGTA, 0.5 mmol/L phenylmethylsulfonylfluoride, 10 μg/mL leupeptin, 5 μg/mL aprotinin, 50 mmol/L NaF, 1 mmol/L sodium vanadate, 10 mmol/L β–glycerophosphate, and 10 mmol/L sodium pyrophosphate. Following a short incubation on ice, lysates were centrifuged at 20,000 × g for 5 minutes at 4°C and the supernatants used for immunoprecipitation using anti-p85α antibody. Immunoprecipitates were washed 3 times with ice-cold lysis buffer before boiling in sample buffer.
Quantitative RT-PCR
RNA was isolated (Qiagen) and reverse transcription was conducted (Applied Biosystems) using standard methods. Quantitative real-time PCR was conducted using gene-specific primers (QuantiTect Primer Assays, Qiagen) for IGF1R, IRS1, IRS2, p85α, or 18S with Fast SYBR Green Master Mix (Applied Biosystems).
Mouse Experiments
KrasLSL-G12D; Trp53Flox/Flox mice and KrasLA2-G12D/+ mice were from the Mouse Models of Human Cancer Consortium (23, 49). KrasLSL-G12D; Trp53Flox/Flox mice were infected with adenovirus expressing Cre recombinase as described (50). Sixteen-week-old KrasLSL-G12D;Trp53Flox/Flox mice; and 12-week-old KrasLA2-G12D/+ mice were treated for 2 or 6 weeks, respectively, by oral gavage delivery of vehicle, MEK inhibitor (2.5 mg/kg/d trametinib), IGF1R inhibitor (40 mg/kg/d OSI-906 or 50 mg/kg/d NVP-AEW541), or both drugs together. Micro-CT analysis was conducted using the SkyScan 1176. Mice were scanned pre- and postdrug treatment regimes. Micro-CT data were sorted, processed, and reconstructed using the N-Recon (SkyScan). Reconstructed data were subsequently imaged using DataViewer, and tumor volumes were calculated using the CTan program (SkyScan).
Data Analysis
Data are presented as mean ± SD unless otherwise stated. For viability and Western blot quantifications, significance was assessed with the 2-tailed unpaired t test. For apoptosis and gene expression analysis, significance was determined using the Mann–Whitney U test. For correlation analyses, Pearson coefficient was used. Comparison between 2 viability curves was done using 2-way ANOVA. The level of significance was set at P < 0.05 (*), P < 0.01 (**), and P < 0.001 (***). The CalcuSyn program (Biosoft), which uses the combination index equation of Chou–Talalay, was used to determine likely synergy of drug combinations using fixed drug ratios.
Disclosure of Potential Conflicts of Interest
No potential conflicts of interest were disclosed.
Authors' Contributions
Conception and design: M. Molina-Arcas, D.C. Hancock, J. Downward
Development of methodology: D.C. Hancock, M.S. Kumar
Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): M. Molina-Arcas, D.C. Hancock, C. Sheridan, M.S. Kumar
Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): M. Molina-Arcas, D.C. Hancock, C. Sheridan
Writing, review, and/or revision of the manuscript: M. Molina-Arcas, D.C. Hancock, J. Downward
Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): D.C. Hancock
Study supervision: J. Downward
Acknowledgments
The authors thank Markus Diefenbacher, Ralph Fritsch, Michael Steckel, Gavin Kelly, Emma Nye, and Gordon Stamp for assistance and discussion.
Grant Support
This work was supported by Cancer Research UK and by the European Union Framework Programme 7 “LUNGTARGET” consortium grant.