Cancer-associated mutations in the p85α N-terminal SH2 domain activate a spectrum of receptor tyrosine kinases

Significance

Phosphoinositide 3-kinase activation typically occurs following stimulation by upstream receptor tyrosine kinases (RTKs), which alleviate p110α inhibition by p85α. p85α and p110α driver mutations have been reported to activate p110α by disrupting the inhibitory interface between p85α and p110α. This study revealed that driver mutations in the p85α N-terminal SH2 domain can enhance p110α activity by inducing the activation of multiple RTKs. Furthermore, combination treatment with RTK and AKT inhibitors provides synergistic therapeutic efficacy. This previously uncharacterized oncogenic mechanism presents the exploitable vulnerability of a class of p85α mutant tumors.

Abstract

The phosphoinositide 3-kinase regulatory subunit p85α is a key regulator of kinase signaling and is frequently mutated in cancers. In the present study, we showed that in addition to weakening the inhibitory interaction between p85α and p110α, a group of driver mutations in the p85α N-terminal SH2 domain activated EGFR, HER2, HER3, c-Met, and IGF-1R in a p110α-independent manner. Cancer cells expressing these mutations exhibited the activation of p110α and the AKT pathway. Interestingly, the activation of EGFR, HER2, and c-Met was attributed to the ability of driver mutations to inhibit HER3 ubiquitination and degradation. The resulting increase in HER3 protein levels promoted its heterodimerization with EGFR, HER2, and c-Met, as well as the allosteric activation of these dimerized partners; however, HER3 silencing abolished this transactivation. Accordingly, inhibitors of either AKT or the HER family reduced the oncogenicity of driver mutations. The combination of these inhibitors resulted in marked synergy. Taken together, our findings provide mechanistic insights and suggest therapeutic strategies targeting a class of recurrent p85α mutations.

PIK3R1, which encodes the p85α regulatory subunit of phosphoinositide 3-kinases (PI3Ks), is frequently mutated in cancers. PIK3R1 mutations are associated with poor survival of cancer patients [the Genomic Data Commons (GDC) Data Portal] (1). Cancer-associated mutations have been detected in all five protein domains of p85α, namely, the Src homology 3 (SH3) domain, GTPase-activating protein (GAP) domain, N-terminal SH2 (nSH2) domain, inter-SH2 (iSH2) domain, and C-terminal SH2 (cSH2) domain (2). Hotspot PIK3R1 mutations cluster in the iSH2 and SH2 domains in agreement with the primary roles of these domains in stabilizing and inhibiting p110 in the p85α–p110 heterodimer (3).

The first reported and characterized clusters of cancer patient–derived mutations were located in two regions of the iSH2 domain (i.e., the E439–K459 and D560–W583 regions) (4–6). Driver mutations in these regions can disrupt the inhibitory interaction between the iSH2 and p110 C2 domains (7), thereby alleviating the inhibition of p110 kinase activity by p85α (8). The other driver mutations target the inhibitory interactions between the nSH2 domain and p110 helical domain (9). The nSH2 domain driver mutations (G376R and K379E) have been suggested to play oncogenic roles by weakening this inhibitory interface (4). Importantly, all these p85α mutants retain the ability to physically bind to p110 and stabilize it. The cSH2 domain interacts with the p110 kinase domain and contributes to p110 inhibition (10). A cSH2 domain driver mutation, K674R, elevates AKT phosphorylation (11). However, the effect of this mutant on p110 remains to be elucidated. Apart from binding to p110, the nSH2 and cSH2 domains bind to phosphotyrosine (pY)-containing consensus sequences (pYXXM) in pY-phosphorylated receptor tyrosine kinases (RTK) or adaptor proteins (3). Engineered p85α mutations in the nSH2 (R358A, S361D) and cSH2 (R649A, S652D) domains impair binding to pY peptides by either removing charge pairing (R358A and R649A) or introducing charge repulsion (S361D, S652D) to the pY phosphate group (12, 13). Binding to the pY motifs upon RTK stimulation is incompatible with the p110-inhibiting interactions of the p85α SH2 domains. Therefore, binding of p85α to pY motifs allows the controlled activation of p110.

In this study, we revealed an oncogenic mechanism evoked by a group of driver mutations in the nSH2 domain. These p85α mutations promote the stabilization of the HER3 protein, thereby activating multiple RTKs. The activation of both RTKs and PI3K/AKT by these nSH2 domain driver mutations should be considered to achieve therapeutic efficacy.

Results

nSH2 Domain Driver Mutations Promote Malignant Phenotypes In Vitro and In Vivo.

A previous functional genomic screen annotated 12 patient-derived PIK3R1 mutations in the nSH2 domain (14). Ten of these mutations were oncogenic that promoted the survival of growth factor–dependent cells in the absence of a growth factor (SI Appendix, Fig. S1A). The remaining two mutations were functionally neutral passengers. Among the 10 driver mutations, four were truncated mutants that do not bind to p110 and have acquired a neomorphic function of activating the MAPK pathways (15). In this study, we characterized the other six driver mutations, namely, W333R, G353R, LR372del, G376R, K379E, and L380del. Endometrial cancer cells were used as models, because 1) endometrial cancer patients with PIK3R1 mutations had significantly worse survival outcomes than patients with wild-type (WT) PIK3R1 (P = 0.035; SI Appendix, Fig. S1B) (1), and 2) nSH2 domain mutations are frequent in the disease.

To assess the functional consequences driven by the six nSH2 domain mutations, endometrial cancer cells KLE and EFE-184 stably expressing WT or mutant PIK3R1 were subjected to the cell viability and colony formation assays. The passenger mutation R386G was used as the control. The KLE and EFE-184 cells harbored WT PIK3R1 and genes of the major members of the PI3K pathway. Overexpression of HA-tagged p85α WT and mutants was confirmed (SI Appendix, Fig. S1C). Cells stably expressing LacZ were included as a control. Consistent with the initial screening, we observed marked increases in the cell viability and clonogenic growth capacity of cells expressing W333R, G353R, LR372del, G376R, K379E, or L380del compared with those expressing PIK3R1 WT or the passenger R386G (Fig. 1 A and B). In addition to cell growth, the six driver mutations significantly enhanced cell migration (Fig. 1C and SI Appendix, Fig. S1D) and invasion (Fig. 1D and SI Appendix, Fig. S1E). To investigate the effect of the driver mutations on tumorigenic progression in vivo, the EFE-184 stable cells were subcutaneously injected into female nude mice. The weights of the tumors expressing the driver mutations were significantly higher than those of the tumors expressing WT PIK3R1 or the passenger R386G (Fig. 1E).

Fig. 1.

PIK3R1 nSH2 domain driver mutations enhance oncogenic phenotypes. EFE-184 or KLE cells stably expressing WT PIK3R1 or mutations or LacZ were examined for (A) cell viability, (B) colony formation, (C) migration, and (D) invasion. (A) Cell viability was measured over a period of 6 d. (B) Representative images (Left) and mean areas (Right) of the colonies formed. (C and D) Representative images (Upper) and mean numbers (Lower) of migrated or invaded EFF-184 cells in five random fields at 100× magnification (Scale bar, 200 μm). Data are mean ± SD (n = 3). (E) Stable EFE-184 cells were subcutaneously injected into nude mice (n = 5). Tumor nodule weights (Upper) and collected nodules (Lower) are shown. Error bars represent SD. *P < 0.05; **P < 0.01; ***P < 0.001; and n.s., no significant difference compared with WT using t test.

p110α and the AKT Pathway Are Activated by nSH2 Domain Driver Mutations.

We investigated the physical association between the p85α nSH2 domain driver mutations and p110α. Consistent with the notion that the interaction between p85α and p110 is primarily mediated by the iSH2 domain, all p85α nSH2 domain mutants could bind to p110α in immunoprecipitation experiments using anti-HA or anti-p110α antibody (Fig. 2A and SI Appendix, Fig. S2). The physical association between p85α and p110α protects p110α from degradation (16). Accordingly, we observed elevated p110α protein levels in WT p85α-overexpressing cells than LacZ-expressing cells (Fig. 2B). Elevated p110α levels were found in cells expressing all mutations, including the passenger, suggesting that the mutations do not disrupt the physical association of p85α with p110α.

Fig. 2.

PIK3R1 nSH2 domain driver mutations activate p110α and AKT signaling. (A) Protein lysates of cells stably expressing HA-tagged WT PIK3R1 or mutations or LacZ were immunoprecipitated (IP) with anti-HA antibody prior to Western blotting (IB). Immunoprecipitation with rabbit IgG was negative control. (B) Protein levels of p110α were examined with ERK2 as loading control. (C) p110α protein was IP, and were subjected to PI3K activity assay. (D) The lipids in the stable cells were collected for detection. PIP3 levels were normalized with those of PIP2. Data are mean ± SD (n = 3). *P < 0.05; **P < 0.01; ***P < 0.001; and n.s., no significant difference compared with WT using t test. (E) Protein levels of AKT signaling molecules were examined. The bar graphs or numbers below the blots indicate the mean densitometry values normalized to those of HA, ERK2, or the corresponding total protein of three independent experiments.

In the PI3K heterodimer, inhibitory contacts are present between the p85α nSH2 domain and p110 helical domain (9). To examine whether nSH2 domain driver mutations altered p110α kinase activity, we measured the production of phosphatidylinositol 3,4,5 trisphosphate (PIP3) catalyzed by immunoprecipitated p110α. p110α exhibited higher kinase activity in the presence of oncogenic mutants (Fig. 2C), with concomitant increase in PIP3 levels (Fig. 2D). G376R, K379E, and L380del are located on the β-strand 3 that is in direct contact with p110α. These mutations compromise the inhibitory association of nSH2 with p110α, explaining the increased p110α activity (8). W333R, G353R, and LR372del are expected to destabilize the three-dimensional structure of nSH2 and thereby its association with p110α, through the loss of hydrophobic interactions and the introduction of clashes, strains, and unfavorable charges (SI Appendix, Fig. S3A). Differential scanning fluorimetry (DSF) on purified recombinant nSH2 domains confirmed that W333R, G353R, and LR372del were not stably folded (SI Appendix, Fig. S3B).

PTEN is another key regulator of cellular PIP3 levels by dephosphorylating PIP3. The p85α SH3 and GAP domains involve in p110α-independent regulation of PTEN (17, 18). PIK3R1 mutations in these domains reduced PTEN protein levels (18, 19). Therefore, we investigated the potential effect of nSH2 domain oncogenic mutants on PTEN protein. As shown in SI Appendix, Fig. S4 A and B, WT p85α increased the protein level and activity of PTEN. All the mutants had the same effect on PTEN as WT p85α. It is unlikely that the oncogenic mechanisms of the tested mutants involve PTEN deregulation.

Next, we determined the phosphorylation levels of signaling proteins of the PI3K pathway. In line with increased PI3K activity, AKT, mTOR, p70-S6K, and S6 were activated in EFE-184 and KLE stable cells (Fig. 2E). Unlike the nSH2 truncation mutants that were shown to activate the MAPK pathways in endometrial and ovarian cancer cells (15), this subset of driver mutations had no effect on ERK1/2, JNK, and p38 MAPK (SI Appendix, Fig. S4C).

nSH2 Domain Driver Mutations Activate Multiple RTKs.

Given that the p85α nSH2 domain interacts with pY motifs in RTKs, we investigated the potential effects of nSH2 domain driver mutations on RTK activity. An RTK array that measured the pY levels of 49 human RTKs was performed using EFE-184 cells expressing the driver mutations W333R, G353R, or LR372del (Fig. 3A). Interestingly, the pY levels of EGFR, HER2, HER3, c-Met, and IGF-1R were markedly increased in cells expressing W333R, G353R, or LR372del (Fig. 3B). In contrast, the phosphorylation levels of HER4 remained unaltered (Fig. 3C). A previous study demonstrated that the PIK3R1 GAP domain mutation R274A activated PDGFR by deregulating Rab5-mediated PDGFR trafficking in mouse fibroblast NIH 3T3 (20). However, our data showed that nSH2 domain driver mutations had no effect on PDGFR levels (Fig. 3C) or Rab5 activity (SI Appendix, Fig. S4D).

Fig. 3.

RTK phosphorylation arrays revealed increased phosphorylation of multiple RTKs by PIK3R1 nSH2 domain driver mutations. Protein lysates of EFE-184 cells stably expressing WT PIK3R1 or mutations were subjected to RTK phosphorylation arrays. (A) Pictures of the arrays at different exposures. (B and C) Densitometry values of the duplicate spots on the slides of two different exposures were obtained. Selected spots of the arrays (Upper) and fold changes in densitometry values (Lower) of the (B) activated or (C) unaltered RTKs of the mutant cells compared with those of WT are shown. The W333R slide after long exposure had high background and was excluded from analysis. Error bars represent SD. *P < 0.05; **P < 0.01; ***P < 0.001; and n.s., no significant difference compared with WT using the t test.

Western blotting confirmed the effects of nSH2 domain driver mutations on RTK activation. EGFR, HER2, HER3, c-Met, and IGF-1R were activated in EFE-184 and KLE cells stably expressing W333R, G353R, LR372del, G376R, and K379E (Fig. 4A). Because the signal of phosphorylated IGF-1R was weak, we included phosphorylated insulin receptor substrate 1 (IRS-1), a downstream substrate of IGF-1R, as a readout of IGF-1R activity (21). IRS-1 phosphorylation was markedly enhanced by W333R, G353R, LR372del, G376R, and K379E. In contrast, RTKs and IRS-1 were neither activated by the passenger R386G nor by the oncogenic mutant L380del (Fig. 4A). Surprisingly, in addition to tyrosine phosphorylation levels, the total protein levels of HER3 and IGF-1R were increased by driver mutations (Fig. 4A). In correlation with the lack of an effect on RTK activities, L380del did not increase the levels of HER3 or IGF-1R protein (Fig. 4A). These data suggest that the increase in the levels of phosphorylated HER3 and IGF-1R could be due to the increased total HER3 and IGF-1R protein levels, respectively. PIK3CA small interfering RNA (siRNA) silencing of p110α neither affected the increased phosphorylation of RTKs and IRS-1 nor total HER3 and IGF-1R levels (SI Appendix, Fig. S5), indicating that RTK activation by these driver mutations is p110α-independent.

Fig. 4.

PIK3R1 nSH2 domain driver mutations activate EGFR, HER2, HER3, c-Met, and IGF-1R. (A) Phosphorylated or total protein levels of the RTKs and IRS-1 in cells stably expressing WT PIK3R1 or mutations or LacZ were examined by Western blotting. Data of some phosphorylation sites in KLE were unavailable due to undetectable levels. (B and C) Lysates of parental, CRISPR WT unedited control, or CRISPR PIK3R1 W333R knock-in HCT-116 cells were subjected to Western blotting. ERK2 was loading control. The numbers below the blots indicate the mean densitometry values of three independent experiments.

We established a CRISPR/Cas9-mediated W333R knock-in in the human colon cancer cell line HCT116, because PIK3R1 mutations frequently manifest in colon cancer patients. W333R was selected for the knock-in due to the availability of the nearby protospacer adjacent motif recognized by Cas9. Successfully edited clones were confirmed via sequencing (SI Appendix, Fig. S6A). As expected, the three knock-in clones with a heterozygous W333R mutation (designated as CRISPR-W333R-1, -2, and -3) displayed higher viability and clonogenic growth than parental HCT116 or unedited control clones that retained WT PIK3R1 after the same gene editing procedure (CRISPR-WT-1 and -2) (SI Appendix, Fig. S6B). Activation of the AKT signaling pathway was evident in the edited cells (Fig. 4B). Interestingly, consistent with the endometrial cancer cell models, W333R knock-in cells had elevated levels of phosphorylated RTKs and IRS-1 as well as total levels of HER3 and IGF-1R (Fig. 4C).

nSH2 Domain Driver Mutations Stabilize the HER3 Protein, Leading to the Activation of EGFR, HER2, and c-Met.

The HER3 kinase domain has minimal catalytic activity and HER3 signaling primarily occurs through heterodimerization with other RTKs. Upon binding, HER3 acts as an “activator” that allosterically activates its partner RTKs (the “receiver”), which subsequently phosphorylates the receptor tails (22–24). RTKs that dimerize with HER3 include HER family members as well as c-Met and IGF-1R (25–28). We speculated that the increase in HER3 protein levels leads to increased heterodimerization with receiver RTKs, thereby inducing RTK activation. First, HER3 siRNA abolished the induced phosphorylation of EGFR, HER2, and c-Met, confirming that HER3 mediates the activation of these RTKs (Fig. 5A). Moreover, immunoprecipitation revealed the higher abundance of HER3 dimerization with EGFR, HER2, or c-Met (Fig. 5B). The induced phosphorylation of IGF-1R and IRS-1 and total IGF-1R was not altered by HER3 knockdown, suggesting that the activation of IGF-1R/IRS-1 is a consequence of an HER3-independent increase in total IGF-1R levels. We also investigated the possibility that IGF-1R/IRS-1 signaling contributes to the RTK activation. Intriguingly, as shown in SI Appendix, Fig. S7A, IRS1 siRNA inhibited the activation of c-Met, concordant with previous observation that IGF-1R induces HGF-independent c-Met activation (29). In contrast, IRS1 knockdown had no impact on the phosphorylation of HER family.

Fig. 5.

PIK3R1 nSH2 domain driver mutations enhance the protein stability of HER3, which mediates the phosphorylation of the other RTKs. (A) EFE-184 cells stably expressing WT PIK3R1 or mutations transfected with HER3 siRNA or mock for 48 h were subjected to Western blotting. (B) Dimerization of HER3 with EGFR, HER2, or c-Met was analyzed by IP using anti-HER3 antibody and Western blotting (IB). Equal amounts of input protein were used for each IP. (C) EFE-184 cells were treated with 25 µg/mL cycloheximide (CHX) for the indicated time course. Some blots are shown in SI Appendix, Fig. S7C. (D) Lysates were immunoprecipitated with anti-HER3 antibody. HER3 protein levels were normalized prior to IP by using proportionally different amounts of lysates. Molecular weight standards (kilodalton) and the location of HER3 are indicated. The numbers or bar graphs below the blots indicate the mean densitometry values normalized to those of the corresponding total protein, immunoprecipitated protein, or ERK2 loading control of three independent experiments. Error bars represent SD. **P < 0.01; ***P < 0.001; and n.s., no significant difference compared with WT using the t test.

Next, we investigated the mechanisms by which the driver mutations increased total HER3 protein levels. The driver mutations had no impact on HER3 messenger RNA (mRNA) levels (SI Appendix, Fig. S7B). Instead, the half-life of HER3 protein was significantly prolonged in the presence of driver mutations (Fig. 5C and SI Appendix, Fig. S7C). This stabilization is likely due to reduced protein degradation because the driver mutations inhibited HER3 ubiquitination (Fig. 5D). Further, we observed that immunoprecipitated HER3 displayed increased interaction with the p85α driver mutants (SI Appendix, Fig. S8A). Reciprocal immunoprecipitation using anti-HA antibody to immunoprecipitate p85α confirmed this finding (SI Appendix, Fig. S8A). p85α binds to multiple pYXXM motifs of HER3 directly (30, 31), we therefore examined the binding between the mutants with a doubly tyrosine-phosphorylated (2pY) HER3 peptide (DGGGPGGD(pY)AAMGACPASEQG(pY)EEMRAFQG). Our peptide pull-down assay revealed that the 2pY HER3 peptide bound stronger to W333R, G353R, LR372del, G376R, and K379E than to WT p85α (SI Appendix, Fig. S8B).

To test whether the mutated nSH2 domains have stronger pY peptide interaction, we repeated this binding assay with purified recombinant nSH2 domains. In contrast to mutants in the context of full-length p85α protein, the recombinant domains either had an unchanged affinity for the 2pY HER3 peptide (G376R, K379E, and L380del), or had markedly lost binding affinity (W333R, G353R, and LR372del) compared with WT p85α (SI Appendix, Fig. S8C). These results were confirmed by isothermal titration calorimetry (ITC) (SI Appendix, Fig. S8D). The nSH2 mutants with the most markedly lost affinity for the HER3 peptides were those identified as structurally destabilized using DSF (SI Appendix, Fig. S3B). Conversely, G376R and K379E affect surface residues that may tolerate these substitutions. The observation that L380del neither decreased the nSH2 stability nor the pY peptide affinity was unexpected, given its location within the peptide binding site (SI Appendix, Fig. S3A). Taken together, our data suggest that those driver mutants that increase HER3 protein levels do so by protecting HER3 from degradation. However, the effect does not result from an increased affinity of the mutated nSH2 domains toward the HER3 pY motifs.

Driver Mutations Sensitize Cancer Cells to AKT and HER Family Inhibitors.

Cells expressing the six p85α driver mutations had higher AKT phosphorylation levels and were more sensitive to AKT inhibitor (MK-2206) than those expressing WT PIK3R1 or the passenger R386G (Fig. 6 A and B). Because the half inhibitory concentration could not be achieved in some conditions, the area under the dose–response curve (AUC) was calculated to represent drug response (Fig. 6B). The nSH2 domain mutations do not confer sensitivity to EGFR inhibitor erlotinib (Fig. 6A). Building on our observations of HER2 and HER3 activation in driver mutation-expressing cells, we performed drug-response assays using lapatinib (EGFR and HER2 inhibitor) and AZD-8931 (EGFR, HER2, and HER3 inhibitor). Cells with the five driver mutations that caused RTK activation (W333R, G353R, LR372del, G376R, and K379E) were more sensitive to lapatinib and AZD-8931 than cells with LacZ, WT, and the passenger R386G (Fig. 6 A and B). In line with minimal changes in RTK signaling, L380del-expressing cells did not show better response to HER family inhibitors (Fig. 6 A and B).

Fig. 6.

PIK3R1 nSH2 domain driver mutations render cells sensitive to AKT inhibitor and HER family inhibitors. EFE-184 cells stably expressing WT PIK3R1 or mutations or LacZ were treated with the inhibitors for 72 h. (A) Dose–response curves of the inhibitors. P values were calculated using values at the highest doses. Data are mean ± SD (n = 3). *P < 0.05; **P < 0.01; ***P < 0.001; and n.s., no significant difference compared with WT using the t test. (B) Heatmap of the AUC values of each inhibitor.

Dual Inhibition of AKT and the HER Family Results in Synergism in Driver Mutation-Expressing Cells.

We assessed the combined effects of inhibiting both the AKT and HER family pathways. Driver mutation-expressing cells or WT cells were treated with AKT or HER family inhibitors alone or in combination with serial concentrations of these inhibitors at a fixed molar ratio. A pronounced synergistic antitumor effect was achieved following combined treatment with MK-2206 and AZD-8931 in cells expressing the five driver mutations (W333R, G353R, LR372del, G376R, and K379E) (Fig. 7 A and B). We also observed a similar synergistic effect in cells cotreated with MK2206 and lapatinib (Fig. 7 A and B and SI Appendix, Fig. S9). The synergistic antitumor effects of these drug combinations were not observed in cells expressing LacZ, WT, L380del, or R386G.

Fig. 7.

Cotargeting AKT and HER family induces synergism in cells expressing the PIK3R1 nSH2 domain driver mutations. (A) EFE-184 or (B) KLE cells stably expressing WT PIK3R1 or mutations or LacZ were treated with MK-2206, AZD-8931, or lapatinib alone or in combination with serial concentrations at a fixed molar ratio. (Left) Dose–response curves of MK-2206 and AZD-8931. Data are mean ± SD (n = 3). (Right) Combination Index values of the drug combination were calculated using the Chou–Talalay method, and values at the highest doses are presented. (C) Proposed mechanistic model underlying the activation of p110α by the PIK3R1 nSH2 domain driver mutations: apart from weakening the inhibitory interaction with p110α, the mutants lead to increased abundance of HER3 protein, which dimerizes with and activates EGFR, HER2, and c-Met.

Discussion

The p85α nSH2 domain establishes an inhibitory contact with p110, which is outcompeted by binding to pYXXM motifs in RTKs (3, 8). Herein, we identify six nSH2 variants that weaken these inhibitory contacts, either by directly affecting p110 binding (G376R, K379E, and L380del) or by destabilizing the nSH2 domain fold (W333R, G353R, and LR372del). In line with our results, K379E has been previously proposed to be oncogenic by disrupting the inhibitory salt bridge interaction between p85α and E545 in the p110α helical domain (4).

In addition, we observed that five of the six nSH2 domain driver mutations (W333R, G353R, LR372del, G376R, and K379E) activated multiple RTKs through p110α-independent increases in the protein levels of HER3 and IGF-1R (Fig. 7C). Among the five activated RTKs, HER3, c-Met, and IGF-1R contain YXXM motifs that bind to the p85α nSH2 domain. In contrast, EGFR and HER2 lack the YXXM motif and do not directly bind to p85α. Interestingly, the binding between p85α and HER3 was shown to be stronger than that between p85α and other HER family members or c-Met (32), inferring that HER3 is a preferred partner of p85α. The driver mutants enhanced HER3 protein levels through inhibiting HER3 degradation. At present, our data cannot explain the regulation of IGF-1R protein by the mutations. Although heterotrimers comprising HER2, HER3, and IGF-1R have been observed in trastuzumab-resistant breast cancer cells (28), our data suggest that HER3 does not mediate IGF-1R regulation in p85α mutant cells. Further, the regulation of RTKs by the nSH2 mutants is selective because the mutants did not stabilize c-Met despite the fact that p85α can directly bind to c-Met.

Our data show that the increase in HER3 protein stability induced by driver mutations elevated the abundance of heterodimeric EGFR/HER3, HER2/HER3, and c-Met/HER3. Our observation that HER3 did not activate HER4 can be explained by previous observations that HER3 is a poor activator of HER4 (33) and that HER3/HER4 heterodimer is an inhibitory interaction (34). Thus, our data present the general mechanism to explain the mutants’ effect on the RTKs. However, the molecular details remain to be elucidated, because increased HER3 binding was not observed with recombinant nSH2 domain mutants. The difference in HER3 binding between mutants in the context of the full-length p85α and the isolated nSH2 domains may indicate the involvement of other p85α domains. For examples, the cSH2 domain also binds to pYXXM motifs (35) and the p85α SH3 domain, which itself does not bind HER3, enhances HER3 binding by p85α (30, 31). The presence of the nSH2 mutations in the context of cell-derived full-length p85α may lead to alterations of intramolecular or intermolecular association of p85α that favor HER3 binding. The mutations’ effect on RTKs cannot be simply explained by the relief of p110α inhibitory interaction, because this effect was independent of p110α and it was not produced by the L380del mutant.

HER3 overexpression is frequently observed in malignancies, including endometrial and colon cancers (36–38). HER3 is an oncogene associated with tumorigenesis and poor patient prognosis (37, 39–41). The nSH2 domain driver mutant cells displayed characteristics resembling HER3-overexpressing cells. First, cells with high HER3 expression showed activation of the PI3K pathway but not the MAPK pathway (42, 43). Second, therapeutically, HER3 overexpression correlates with responses to the dual EGFR/HER2 inhibitor lapatinib across 19 endometrial cancer cell lines (44). We also observed that nSH2 driver mutations did not alter the sensitivity of cells toward the EGFR inhibitor erlotinib. In line with this finding, HER3 activation might cause erlotinib resistance by dimerizing with other RTKs to induce compensatory signaling (26).

In summary, this study revealed the oncogenic mechanisms of several patient-derived nSH2 domain driver mutations and identified responses to drugs targeting the activated molecules. Five (W333R, G353R, G376R, K379E, and L380del) of the six driver mutations were detected in more than one cancer patient (SI Appendix, Fig. S1A). The cancer types with these mutations detected include that of breast, colon, and skin and most frequently endometrial cancer and glioblastoma [Catalogue Of Somatic Mutations In Cancer (COSMIC) database] (45). Exploitation of these mutations for cancer therapeutics warrants further investigation.

Materials and Methods

A detailed description of the materials and methods can be found in SI Appendix.

Cell Lines and Transfection.

KLE and HCT-116 cells were obtained from American Type Culture Collection (Manassas), and EFE-184 cells were obtained from DSMZ-German Collection of Microorganisms and Cell Cultures (Braunschweig, Germany). The cell culture conditions, generation of stable cells and knock-in cells, and information of plasmids and siRNA are provided in SI Appendix.

Phenotypic Assays.

In colony formation assay, 800 cells were seeded into 6-well plate and were cultured up to 21 d. In the viability assay, 1,000 cells were seeded in triplicate into a 96-well plate. Cell viability was measured at indicated time points with resazurin. In cell migration and invasion assays, 3 × 10⁴ cells were seeded into Matrigel-coated (invasion) or -uncoated (migration) inserts. After 24 h, cells were fixed and stained with crystal violet. In the in vivo tumorigenicity experiment, 5 × 10⁶ cells were subcutaneously injected into female athymic nude mouse (n = 5 each group; Charles River Lab). Tumor nodules were collected after 8 wk (details in SI Appendix).

Western Blotting, Immunoprecipitation, Peptide Pull-Down Assay, and Phospho-RTK Array.

Protein lysates were prepared using the corresponding lysis buffers. Equal amounts of protein were subjected to Western blotting or phospho-RTK array. In immunoprecipiation or pull-down experiments, protein lysates were incubated with primary antibody or peptide. The eluents were analyzed by Western blotting. Detailed procedures and information of the antibodies are provided in SI Appendix.

Production of Recombinant Proteins, DSF, and ITC.

Recombinant nSH2 domain proteins were expressed in Escherichia coli Bl21 (DE3) cells. The purified proteins were subjected to DSF and ITC. DSF experiments were performed on C1000 Touch thermal cycler CFX96 Real-Time system (Bio-Rad) in conjunction with SYPRO Orange 5000X dye (Invitrogen). ITC experiments were performed on a MicroCal ITC200 (GE HealthCare). Refer to SI Appendix for details.

Drug Response Assay.

Cells seeded at a density of 1,000 cells/well in a 96-well plate were treated with serial diluted single or combined inhibitors for 72 h. Viable cells were detected with resazurin (details in SI Appendix). The AUC values were obtained using the GraphPad Prism 8 software (La Jolla). The synergistic effects of the drug combinations were quantified using the CompuSyn Software (Biosoft, Version 2.1)

PI3K Activity Assay and PIP3/PIP2 Quantification.

To measure in vitro kinase activity, the immunoprecipitated p110α protein was incubated with PIP2 substrate to generate PIP3. The amount of generated PIP3 was detected at 450 nm. To measure cellular PIP3 and PIP2 levels, total lipids were extracted using ice-cold 0.5 M trichloroacetic acid prior to extraction of acidic lipids in the organic phase (details in SI Appendix).

Statistical Analyses.

All experiments were performed three times. Results are presented as mean ± SD. P values were calculated using the Student’s unpaired two-tailed t test with GraphPad Prism. P values < 0.05 were considered statistically significant.

Data Availability

All study data are included in the article and/or SI Appendix.