Maximizing the Efficacy of MAPK-Targeted Treatment in PTEN^LOF/BRAF^MUT Melanoma through PI3K and IGF1R Inhibition

The observation that approximately 50% of malignant melanomas harbor mutations in the BRAF gene has triggered intensive efforts to develop inhibitors of mutant BRAF, including vemurafenib and dabrafenib. Although the introduction of these drugs has revolutionized the treatment of BRAF-mutant (BRAF^MUT) melanoma patients, sustained benefit is limited by intrinsic and acquired resistance. The major mechanism of acquired resistance appears to involve reactivation of MAPK signaling, and indeed dual BRAF/MEK inhibition has improved clinical responses compared with single-agent treatment (1). Nevertheless, BRAF/MEK inhibitor combination therapy fails to provide a curative treatment option for patients with BRAF-mutant melanoma most likely due to concurrent genetic alterations conferring codependencies. In accordance with a complementary role of PTEN, PTEN^LOF alterations were identified in approximately 40% of BRAF^MUT melanoma and were shown to promote Braf-driven metastatic melanoma in a genetically engineered mouse model (2). Furthermore, PTEN^LOF alterations are associated with BRAF inhibitor resistance in cellular models and in patients as well as decreased progression-free survival upon dabrafenib treatment (3–5). Similarly, decreased PTEN protein levels were found to predict for vemurafenib insensitivity, highlighting the urgent need for the optimization of targeted therapies for this genetic context (6).

PTEN encodes a lipid phosphatase that antagonizes PI3K activity by dephosphorylating phosphatidylinositol-3,4,5-trisphosphate. Class IA PI3Ks are heterodimeric lipid kinases composed of a p85 regulatory subunit and one of three catalytic subunits, namely p110α, p110β, or p110δ (PI3Kα, PI3Kβ, or PI3Kδ). Whereas PI3Kα and PI3Kβ are ubiquitously expressed, expression of PI3Kδ has been reported to be largely restricted to the hematopoietic system. In contrast with this, interrogation of the Cancer Cell Line Encyclopedia (CCLE) and associated mRNA expression profiles revealed significant expression levels of all three isoforms in the majority of cell lines derived from diverse lineages (7). Nevertheless, PTEN^LOF cancers and cell lines of different lineages have been shown to specifically depend on PI3Kβ (8–10). In this study, we demonstrate for the first time the exquisite dependence of PTEN^LOF melanoma on the PI3Kβ isoform. Although PI3Kβ inhibitors were only partially efficacious, using an unbiased RNAi approach, we identified a homogeneous pattern of chemosensitization to PI3Kβ inhibition with PI3Kα, insulin-like growth factor receptor 1 (IGF1R), and components of the MAPK pathway as top hits in PTEN^LOF/BRAF^MUT melanoma. We subsequently confirmed that dual inhibition of PI3Kα/β was required to permanently block PI3K signaling and that addition of PI3Kα/β inhibitors strongly synergized with MAPK pathway inhibitors. In comparison with direct targeting of PI3Kα, anti-IGF1R antibodies, which presumably block receptor tyrosine kinase (RTK)–mediated PI3Kα activation in the tumor tissue only, were as efficacious, but did not lead to the induction of glucose abnormalities. These data provide a strong rationale for the first-line combination of MAPK pathway inhibitors with panPI3K or PI3Kβ/IGF1R inhibitors to overcome intrinsic and acquired resistance to BRAF inhibitors in PTEN^LOF/BRAF^MUT melanoma.

Cell lines were obtained from and cultured as recommended by the ATCC, Deutsche Sammlung von Mikroorganismen und Zellkulturen GmbH, and Health Science Research Resources Bank. Authenticity was confirmed by single-nucleotide polymorphism fingerprinting, and cells were cultivated less than 20 passages before use.

The sensitivity profiles of melanoma lines used for this study were extracted from a high-throughput profiling approach described previously (7). Manual determination of the antiproliferative effect of compounds or compound combinations was performed as described previously using methylene blue or resazurin staining as readout (11, 12). Detailed information is summarized in Supplementary Materials and Methods.

Lentiviral supernatants of barcoded custom library (Cellecta) of approximately 170,000 shRNAs targeting approximately 7,500 genes (divided into three sublibraries) were produced according to the manufacturer's recommended protocols. After infection and puromycin selection, cells were cultivated ±1 μmol/L rac-KIN-193 at a minimal representation of 1,000 cells/shRNA for five population doublings. The representation of each shRNA barcode was measured by next-generation sequencing on an Illumina HiSeq platform. Hits for each panel of cell lines were identified as described in Supplementary Materials and Methods.

Immunoblotting and coimmunoprecipitation experiments were done according to standard methods. Detailed information and a comprehensive list of antibodies are provided in Supplementary Materials and Methods.

Reverse-phase protein array (RPPA) was performed as described previously (13). In brief, cells were seeded in duplicate in 96-well plates at an appropriate density, and the following day, they were treated with serial compound dilutions for 1 hour, washed with ice-cold PBS, and lysed in CLB96 lysis buffer (Bayer Technology). Pulverized tumor samples were processed accordingly.

Immunoprecipitates were separated by SDS-PAGE and in-gel digested using a Trypsin/Lys-C Mix (Promega). Peptides were analyzed by LC-MS/MS (LTQ OrbiTrap Elite; Thermo) as described previously (14), using spectral counting as a proxy for the protein quantity.

Induction of cell death was determined in triplicates on an Array Scan VTI platform (Cellomics) using 2 μg/mL propidium iodide (PI) (Life Technologies) and 1 μg/mL Hoechst33342 (Sigma). Induction of apoptosis was determined in triplicates by Annexin V/PI staining (Life Technologies) according to the manufacturer's flow cytometry protocol.

WM-266-4 cells (2 × 10⁶) were injected subcutaneously into the flanks of nude mice. At a tumor volume of 200 to 300 cm³, mice were treated with the indicated compounds. Detailed information, including analysis of glucose/insulin levels, is provided in Supplementary Materials and Methods.

To determine whether specific compound mechanisms selectively impair the growth of PTEN^LOF melanoma cells, we compared the activity of 1397 compounds for their effect on the viability of 14 PTEN^LOF and 27 PTEN^WT melanoma lines from the CCLE based on their activity area (combined measure of compound potency and efficacy) as described elsewhere (7). This analysis revealed that PI3Kβ inhibition best discriminates among 390 different modes of action tested between PTEN^LOF and PTEN^WT melanoma lines (Fig. 1A; Supplementary Table S1). In contrast, PI3Kα, PI3Kδ, PI3Kγ, or panPI3K inhibitors did not score significantly as selectively active compounds toward PTEN^LOF melanoma lines. Interestingly, sensitivity to PI3Kα-selective inhibitors was detected primarily in PTEN^WT lines, but these compounds did not pass the significance cutoff (P < 0.01; data not shown).

Accordingly, treatment of the PTEN^LOF/BRAF^MUT melanoma cell line RVH-421 with the PI3Kβi rac-KIN-193, but not with the selective PI3Kαi BYL719, led to a reduction in Akt phosphorylation within 1 hour in a dose-dependent manner (Fig. 1B). Quantitative analysis using RPPA validated this finding in three different cell lines (Fig. 1C; Supplementary Fig. S1A). Short-term treatment with two different PI3Kβi's (rac-KIN-193/TGX-221) reduced Akt phosphorylation similarly to the panPI3Ki (GDC-0941), whereas a PI3Kα inhibitor reduced Akt phosphorylation only at high concentrations where it is also active against other PI3K isoforms.

To test whether inhibition of Akt phosphorylation led to growth inhibition, the antiproliferative effect of these compounds was determined in a panel of 9 PTEN^LOF/BRAF^MUT and 11 PTEN^WT/BRAF^MUT melanoma lines (Fig. 1D). Concordance of PTEN^LOF calls according to the CCLE and absence of PTEN protein was confirmed by immunoblotting the cell lines (Supplementary Fig. S1B). Analysis of the maximum efficacy (A_max) and GI₅₀ values confirmed that PI3Kβ inhibition, but not PI3Kα or panPI3K inhibition, selectively impaired the growth of PTEN^LOF/BRAF^MUT compared with PTEN^WT/BRAF^MUT melanoma lines. However, both PI3Kβi's only partially blocked proliferation (A_max ≥ 25%) in most melanoma lines. Interestingly, the panPI3Ki was as efficient as the PI3Kβi's in reducing pAkt levels, but superior in inhibiting proliferation. This observation suggests either that short-term pAkt inhibition might not be indicative of complete or prolonged pathway inhibition or that an alternative non-Akt substrate of PI3K plays a critical role in the regulation of proliferation.

Taken together, our unbiased approach enabled us to identify PI3Kβ as the primary PI3K isoform driving Akt signaling in this genetic context.

As the treatment with PI3Kβi's yielded only partial responses in PTEN^LOF/BRAF^MUT melanoma lines, we sought to determine whether additional signaling nodes might contribute to mitigating the overall response. To discover such nodes, a deep-coverage pooled shRNA screen was performed wherein a lentiviral shRNA library targeting approximately 7,500 genes with approximately 20 shRNAs/gene was introduced into 5 PTEN^LOF/BRAF^MUT and 6 PTEN^WT/BRAF^MUT melanoma lines. After five population doublings (±PI3Kβi), the relative abundance of each individual shRNA was determined by deep sequencing (Fig. 2A). Redundant siRNA activity (RSA) scores of drug-treated and untreated control samples were compared using Limma's linear model within each genotype to identify gene products that were selectively required in the presence of PI3Kβi in the PTEN^LOF/BRAF^MUT cells only (15). PIK3CA, IGF1R, BRAF, MAP2K1, and MAPK1 were identified as top sensitizers to PI3Kβ inhibition in a genotype-selective manner with a marked distinction from any other hit in this unbiased approach (Fig. 2B).

The above data suggested that PI3Kα and IGF1R signaling were attenuating the response to PI3Kβ inhibitors. To further test this notion, and to ask whether a PI3Kαi would hence act synergistically with a PI3Kβi in PTEN^LOF/BRAF^MUT melanoma lines, two PTEN^LOF/BRAF^MUT lines were treated with a PI3Kβi and a PI3Kαi in a matrix combination setting (Fig. 2C). Although single-agent treatment with PI3Kαi did not affect proliferation, combination with PI3Kβi1 led to an enhanced antiproliferative effect at all concentrations tested (Fig. 2C). Synergy scores (SS > 2) demonstrated that the combination of both drugs acted in a robust synergistic manner. The structurally diverse PI3Kβi2 GSK-2636771 yielded similar results, indicating that these synergistic effects are likely on-target. In conclusion, the combination of specific small-molecule inhibitors recapitulated the synthetic lethal interaction identified by RNAi in the chemosensitization screen.

The fact that PI3Kαi activity was only observed in combination with a PI3Kβi raised the possibility that PI3Kβ inhibition leads to the activation of PI3Kα. Activation of PI3Kα is regulated by membrane recruitment of the p110α/p85 complex to tyrosine-phosphorylated RTKs or their adaptor proteins via the interaction of p85 SH2 domains with phosphorylated YXXM motifs (16–18). In order to examine PI3Kα recruitment to RTKs or adaptor proteins, we next studied the interaction of p85 with phosphotyrosine-containing proteins upon PI3Kβi treatment. Immunoblot analysis of p85 coimmunoprecipitates using a pTyr antibody revealed one specific signal upon PI3Kβi treatment in RVH-421 cells (Fig. 3A). Label-free mass spectrometric analysis of these immunocomplexes identified 20 proteins that were enriched upon PI3Kβi treatment (Supplementary Table S2). Among these, only the adaptor protein insulin receptor substrate 2 (IRS2) that mediates the activation of signaling pathways in response to ligand stimulation of cell surface receptors is reported to be Tyr-phosphorylated and harbors 6 YXXM motifs, therefore being a potent activator of PI3K (16, 19). Subsequent, direct coimmunoprecipitation assays confirmed that PI3Kβ inhibition induced a robust p85/IRS2 interaction (Fig. 3B). These studies also demonstrated PI3Kβi-induced tyrosine phosphorylation of IRS2 with concomitant reduction of pIRS2 S731 levels and the consequential change in IRS2 protein migration in the SDS gel (20). Time-course experiments revealed that the interaction was detectable within 30 minutes of treatment (Supplementary Fig. S2A).

To explore the correlation of PI3K pathway inhibition and p85/IRS2 interaction in more detail, RVH-421 cells were treated with a panel of PI3K inhibitors of different isoform specificities (Fig. 3C). Both the PI3Kβi and the panPI3Ki reduced phosphorylation of Akt, S6, NDRG1, a known substrate of the mTORC2-regulated kinase SGK, and IRS2 (Fig. 3C, bottom; ref. 21), whereas the PI3Kαi did not affect PI3K signaling in this cell line. In keeping with this, treatment with both the PI3Kβi and the panPI3Ki triggered the p85/IRS2 interaction, whereas the PI3Kαi did not (Fig. 3C, top). In contrast, in the PTEN^WT/BRAF^MUT cell line A-375, the PI3KαI, but not the PI3KβI, caused the same consequences as described before for the PI3Kβi in the PTEN^LOF/BRAF^MUT setting (Fig. 3D).

In conclusion, our data support the notion that PI3K signaling is reactivated upon PI3Kβ inhibition in PTEN^LOF/BRAF^MUT melanoma by recruitment of PI3K to an activated RTK via the adaptor molecule IRS2.

In addition to the role of PI3Kβ in activating the classical PI3K pathway, PI3Kβ exhibits Akt/mTOR-independent functions, including the regulation of autophagy (22). To investigate whether classical PI3Kβ downstream signaling accounts for the induction of p85/IRS2 interaction, we asked whether inhibition of mTOR, PDK1, and Akt with selective small molecules would also trigger the p85/IRS2 interaction. The catalytic mTORi AZD8055, the allosteric mTORC1i RAD001 (Fig. 4A; Supplementary Fig. S2B), the catalytic PDK1i GSK-2334470 (Fig. 4B; Supplementary Fig. S2C), and MK2206, which inhibits Akt membrane localization (Fig. 4C; Supplementary Fig. S2D), all induced p85/IRS2 complex formation in a dose-dependent manner.

These data strongly suggest that PI3Kβ inhibition leads to the loss of feedback inhibition mediated by Ser/Thr kinases, including Akt, mTORC1, and S6K and, subsequently, to the IRS2-dependent recruitment of PI3Kα through activated RTKs.

As IGF1R was found among the highest scoring sensitizers to PI3Kβ inhibition (Fig. 2B), we speculated that IGF1R was the sole RTK mediating the reactivation of PI3K signaling upon PI3Kβi treatment. In agreement with this, AEW541 treatment (IGF1Ri) blocked PI3Kβi-induced p85/IRS2 interaction in a dose-dependent manner (Fig. 5A). In keeping with previous reports showing an increase in RTK expression levels by PI3K inhibition (23), long-term PI3Kβi treatment led to an increase in IGF1R levels in the four PTEN^LOF/BRAF^MUT models tested (Supplementary Fig. S3). To further validate IGF1R inhibition as a sensitizer to PI3Kβ inhibition, cell viability was determined upon treatment with this combination. Whereas the IGF1Ri had only minor antiproliferative effects, a fixed concentration of the compound enhanced PI3Kβi efficacy at all concentrations tested (Fig. 5B; Supplementary Fig. S4A). Furthermore, matrix combination assays confirmed the synergistic activity of IGF1Ri with two structurally distinct PI3Kβi's, but not the PI3Kαi, in two PTEN^LOF/BRAF^MUT lines (Fig. 5C; Supplementary Fig. S4B). PanPI3Ki treatment also led to synergistic effects, albeit less significant due to its stronger efficacy as single agent. Importantly, the selectivity of this synergy with PI3Kβi's in the genetic context of PTEN^LOF was underscored by the lack of combination activity in the PTEN^WT/BRAF^MUT melanoma cell line A-375 (Supplementary Fig. S4C).

These data indicate that IGF1R inhibition acts in combination specifically with PI3Kβ inhibitors to block proliferation of PTEN^LOF/BRAF^MUT melanoma by interfering with the reactivation of PI3K signaling through PI3Kα.

In keeping with the known dependence of BRAF^MUT melanoma on the MAPK pathway, the pooled shRNA screen identified shRNAs targeting key nodes of MAPK signaling (BRAF, MAP2K1, and MAPK1) among the strongest sensitizers to PI3Kβ inhibition (Fig. 2B). Indeed, the two PI3Kβi's and the panPI3Ki showed strong synergy upon combination treatment with the BRAFi (LGX818) or the MEKi (MEK162) while the PI3Kαi did not (Fig. 6A and B; Supplementary Fig. S5A and S5B). As expected, synergy was not observed in the PTEN^WT background (Supplementary Fig. S5C and S5D). Thus, our data support the notion that adding PI3Kβ inhibitors to MAPK pathway inhibitors in PTEN^LOF/BRAF^MUT melanoma elicits a robust and specific combinatorial activity.

In a next step, the impact of IGF1R-mediated activation of PI3Kα on the ability of a PI3Kβi to permanently block the PI3K/mTOR pathway was assessed in two cell lines (Fig. 6C; Supplementary Fig. S6A and S6B). Although treatment with the PI3Kβi for 2 hours efficiently blocked PI3K/mTOR signaling, a rebound after 72 hours of treatment was detected (lanes 1/2 and 11/12). Loss of activity due to compound instability was excluded as the compound had been refreshed appropriately. Dual PI3Kα/β or panPI3K inhibition prevented this PI3K/mTOR signaling rebound (lane 16; Supplementary Fig. S7A). Consistent with an exclusive role for IGF1R in PI3Kα activation, the IGF1Ri (or an anti-IGF1R antibody; Supplementary Fig. S6B) substituted for the PI3Kαi in preventing this rebound (lane 17). The elevated MAPK signaling in this BRAF^MUT setting might interfere with PI3K/mTOR signaling by regulation of the activity of the TSC1/2 complex and phosphorylation of S6 at Ser235/236 by its effector RSK (24–26). Indeed, concomitant inhibition of MEK/PI3Kβ/PI3Kα led to a significant decrease not only in Erk but also in S6 phosphorylation when treatment was continued for 72 hours. This sustained MAPK/PI3K/mTOR blockade led to the induction of PARP cleavage (lanes 9 and 19). Also, combined inhibition of MEK/PI3Kβ/IGF1R restored complete pathway inhibition in long-term assays and triggered PARP cleavage (lanes 10 and 20).

Furthermore, the induction of apoptosis triggered by the MEK and BRAF inhibitors was more pronounced when the MAPK pathway inhibitors were combined with a PI3Kβi as measured by Annexin V/PI staining in four PTEN^LOF/BRAF^MUT melanoma lines (Fig. 6D; Supplementary Fig. S7B). Most importantly, however, further addition of either the PI3Kαi or the IGF1Ri was associated with an even more pronounced increase in the fraction of early and/or late apoptotic cells underscoring the need for sustained concomitant blockade of MAPK and PI3K signaling in order to maximize impact on cell viability in this genetic context (Fig. 6D; Supplementary Fig. S7B). In addition, matrix combination assays revealed that PI3K pathway inhibition using PI3Kβi/PI3Kαi or PI3Kβi/IGF1Ri combinations also enhanced the antiproliferative effect of the current BRAF/MEK inhibitor regimen in the setting of PTEN^LOF/BRAF^MUT and shifted the response toward cell death (Supplementary Fig. S8A and S8B).

Taken together, these results show that, in PTEN^LOF/BRAF^MUT melanoma, the benefit of adding PI3Kβ to MAPK pathway inhibition is limited due to reactivation of PI3Kα by relieving a feedback inhibition on IGF1R signaling (Supplementary Fig. S9). Thus, sustained and concomitant inhibition of PI3Kβ together with PI3Kα or IGF1R and MAPK signaling is necessary to achieve complete long-term pathway inhibition and maximize induction of cell death.

Due to the dominant role of PI3Kα in glucose homeostasis, the substitution of PI3Kα with IGF1R inhibitors provides the unique opportunity to circumvent dose-limiting on-target side effects of PI3Kα inhibition in insulin-sensitive tissues. Therefore, the efficacy and effects on glucose homeostasis of PI3Kαi or IGF1Ri treatments in combination with PI3Kβi/BRAFi were determined in the PTEN^LOF/BRAF^MUT xenograft model WM-266-4. Contrary to the in vitro findings, the PI3Kαi and PI3Kβi were almost equally efficient in inhibiting tumor growth, most likely due to the reported antiangiogenic effect of PI3Kα inhibition (27). The inhibition of both isoforms individually boosted the effect of the BRAFi, but dual PI3Kα/β inhibition increased the efficacy of BRAFi treatment even further. Interestingly, although IGF1R inhibition (anti-IGF1R antibody) substituted for PI3Kα inhibition in the blockade of tumor growth (Fig. 7A), there were profound differences in their effects on glucose homeostasis. As expected, PI3Kαi treatment resulted in hyperglycemia that was particularly pronounced in the triple PI3Kαi/PI3Kβi/BRAFi combination. Strikingly, no effects on insulin signaling and body weight were observed in any of the anti-IGF1R antibody treatment arms (Fig. 7B, data not shown). Pharmacokinetics analysis confirmed the expected exposures excluding possible drug–drug interactions (data not shown).

In summary, these data demonstrate a clear improvement of efficacy of sole BRAF inhibition in the PTEN^LOF/BRAF^MUT melanoma background by concomitant dual PI3Kα/β blockade using either direct PI3Kα or IGF1R inhibitors and provide a strong rationale to test such combinations in humans.

Here, we have demonstrated for the first time that PTEN^LOF in the setting of BRAF^MUT melanoma causes selective sensitivity to PI3Kβ inhibitors. However, the major limitation of PI3Kβ as a drug target in this genetic setting is the immediate rebound of PI3Kα signaling. Therefore, concomitant inhibition of PI3Kα is required to achieve sustained PI3K pathway blockade and enhance the apoptotic response upon treatment with the current BRAF/MEK inhibitor regimen.

In parallel with our findings, recent studies revealed that the contribution of individual isoforms to PI3K signaling upon PTEN^LOF is highly context dependent. Although PTEN^LOF tumors and cancer derived from prostate and breast lineage specifically depend on PI3Kβ signaling (8–10), in PTEN^LOF cell lines and tumors derived from the endometrial lineage, dual inhibition of PI3Kα/β is required to reduce cell viability or tumor growth, especially in a Kras-mutant background (28, 29). Similarly, both PI3Kα/β contribute to the PTEN hamartoma tumor syndrome (30), whereas T-cell acute lymphoblastic leukemia driven by PTEN^LOF relies on the activity of PI3Kγ/δ in line with the predominant role of these isoforms in the hematopoietic lineage (31). These observations underscore the need for the determination of isoform dependency in different indications as well as in-depth analysis of coexisting genetic alterations. Indeed, recent studies revealed a mechanism of actively induced compensation by the various PI3K isoforms (32, 33). Similarly, while we observed a primary sensitivity to PI3Kβ inhibitors in our melanoma panel and a trend toward PI3Kδ sensitivity (Fig. 1A), combination of PI3Kβi with PI3Kαi was necessary to inhibit PI3K signaling upon long-term treatment. In contrast, long-term treatment with PI3Kδi did not synergize with PI3Kβi to robustly inhibit PI3K signaling, but dual PI3Kδ/α inhibition caused mild reduction of pAkt (Supplementary Fig. S3). In contrast with our findings, a recent study using genetically engineered murine models and three cell lines distinct to ours excluded a role of the PI3Kβ isoform in PTEN^LOF/BRAF^MUT melanoma but demonstrated the sole involvement of PI3Kδ/α instead (34). Taken together, these findings indicate that the isoforms' interplay and the corresponding tumor dependency is diverse and highly context specific. Thus, depending on the adaptive input from the upstream RTKs and G protein coupled receptors, PTEN loss will amplify the signal of different class IA PI3K isoforms. This suggests clinical strategies that would either require the combined use of isoform-specific inhibitors in specifically selected subsets of PTEN^LOF/BRAF^MUT melanoma patients or the use of panPI3K inhibitors across the whole PTEN^LOF/BRAF^MUT melanoma population. However, systemic PI3Kα inhibition will ultimately lead to the induction of on-target dose-limiting toxicities that are even exacerbated when combined with MEK inhibitors. Alternative scheduling regimens or targeting the mechanism of PI3Kα reactivation in the tumor tissue specifically might circumvent this problem. Our data provide evidence that in PTEN^LOF/BRAF^MUT melanoma PI3Kβ inhibition triggers PI3Kα activation via IGF1R and that IGF1R inhibition indeed substitutes for PI3Kα inhibition to block tumor growth, providing a unique opportunity to minimize the induction of glucose abnormalities.

In contrast with several reports pointing to highly heterogeneous resistance mechanisms to inhibitors of PI3K/mTOR involving a diverse set of RTKs (35, 36), we identified exclusively and reproducibly shRNAs targeting IGF1R as sensitizers of PI3Kβ inhibition across a panel of PTEN^LOF/BRAF^MUT melanoma lines by blocking the reactivation of PI3Kα signaling.

Activation of PI3K signaling via the IGF1R has been shown to contribute to the emergence of resistance to targeted therapies in a variety of human cancers, including BRAF^MUT melanoma treated with a BRAF inhibitor (37–40). To achieve increased efficacy of therapy and prevention of resistance, it might therefore be beneficial to improve patient stratification beyond BRAF^MUT status and extend the standard treatment of MAPK inhibition upfront with PI3K pathway inhibitors, especially in a setting where PTEN is already lost. Based on the results presented here, however, we speculate that in patients with BRAF^MUT and concomitant PTEN^LOF, a combination of MAPK pathway inhibition with dual PI3Kα/β inhibitors or with panPI3K inhibitors may be more effective than sole combination of MAPK pathway with PI3Kβ inhibitors. One of the major concern with such strategies is, however, the induction of changes in glucose homeostasis due to blockade of the PI3Kα isoform in insulin-sensitive tissues (41), leading to dose-limiting hyperglycemia. Therefore, the observation that PI3Kα was exclusively activated by IGF1R in PTEN^LOF/BRAF^MUT melanoma provided the unique opportunity to successfully replace the PI3Kα inhibition by a combination of IGF1R and PI3Kβ inhibitors, thereby minimizing on-target side effects on insulin-sensitive tissues (42).

The pharmaceutical industry has heavily pursued the development of various PI3K and MAPK pathway inhibitors with distinct modes of action over the last decade. The availability of these compounds together with the data presented in this work strongly argues for the initiation of preclinical and clinical studies to evaluate the potential long-term benefit for patients with PTEN^LOF/BRAF^MUT melanoma treated with a combination of MAPK pathway inhibitors together with pan PI3K or PI3Kβ and IGF1R monoclonal antibodies.

W.R. Sellers is VP/Global Head of Oncology at and has ownership interest (including patents) in Novartis. F. Hofmann has ownership interest (including patents) in Novartis stocks. No potential conflicts of interest were disclosed by the other authors.

Conception and design: B. Herkert, C. Schnell, H. Voshol, W.R. Sellers, F. Hofmann, S.M. Brachmann

Development of methodology: B. Herkert, A. Kauffmann, H. Erasimus, E. Billy, D. Ruddy, D. Guthy

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): B. Herkert, S. Mollé, C. Schnell, T. Ferrat, H. Voshol, J. Juengert, H. Erasimus, G. Marszalek, M. Kazic-Legueux, D. Ruddy, M. Stump, D. Guthy, M. Ristov, K. Calkins

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): B. Herkert, A. Kauffmann, S. Mollé, C. Schnell, H. Voshol, M. Kazic-Legueux, E. Billy, M. Stump, S.-M. Maira, W.R. Sellers

Writing, review, and/or revision of the manuscript: B. Herkert, A. Kauffmann, C. Schnell, H. Voshol, M. Stump, K. Calkins, S.-M. Maira, W.R. Sellers, F. Hofmann, S.M. Brachmann

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): D. Ruddy

Study supervision: B. Herkert, F. Hofmann, M.N. Hall, S.M. Brachmann

The authors thank Kristine Yu, Kalyani Gampa, Iris Kao, Philippe Megel, and Odile Weber for technical assistance and Christine Fritsch, Tobias Schmelzle, Michael Schlabach, and Karen Yu for discussions.

All research costs were covered by Novartis AG. No additional external funding was received for this study.

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.