PIK3CA missense mutations promote glioblastoma pathogenesis, but do not enhance targeted PI3K inhibition.

BACKGROUND: Glioblastoma (GBM) is the most common adult primary brain tumor. Multimodal treatment is empiric and prognosis remains poor. Recurrent PIK3CA missense mutations (PIK3CAmut) in GBM are restricted to three functional domains: adaptor binding (ABD), helical, and kinase. Defining how these mutations influence gliomagenesis and response to kinase inhibitors may aid in the clinical development of novel targeted therapies in biomarker-stratified patients.
METHODS: We used normal human astrocytes immortalized via expression of hTERT, E6, and E7 (NHA). We selected two PIK3CAmut from each of 3 mutated domains and induced their expression in NHA with (NHARAS) and without mutant RAS using lentiviral vectors. We then examined the role of PIK3CAmut in gliomagenesis in vitro and in mice, as well as response to targeted PI3K (PI3Ki) and MEK (MEKi) inhibitors in vitro.
RESULTS: PIK3CAmut, particularly helical and kinase domain mutations, potentiated proximal PI3K signaling and migration of NHA and NHARAS in vitro. Only kinase domain mutations promoted NHA colony formation, but both helical and kinase domain mutations promoted NHARAS tumorigenesis in vivo. PIK3CAmut status had minimal effects on PI3Ki and MEKi efficacy. However, PI3Ki/MEKi synergism was pronounced in NHA and NHARAS harboring ABD or helical mutations.
CONCLUSION: PIK3CAmut promoted differential gliomagenesis based on the mutated domain. While PIK3CAmut did not influence sensitivity to single agent PI3Ki, they did alter PI3Ki/MEKi synergism. Taken together, our results demonstrate that a subset of PIK3CAmut promote tumorigenesis and suggest that patients with helical domain mutations may be most sensitive to dual PI3Ki/MEKi treatment.

Introduction

Glioblastoma (GBM) is the most common primary malignant brain tumor in adults [1]. It is also the most aggressive, with a median survival of only 12–15 months [1-3]. The molecular heterogeneity of GBM has been extensively characterized [4-6]. The vast majority of GBM arise de novo and harbor frequent mutations in 3 “core” signaling pathways: RB, TP53, and receptor tyrosine kinase (RTK)/mitogen activated protein kinase (MAPK)/phosphoinositide 3-kinase (PI3K) [5]. GBM can be stratified into 4 molecular subtypes based on gene expression [6]. However, this knowledge has yet to impact patient management. First line therapy remains empiric and consists of surgical resection followed by radiation with concurrent and adjuvant temozolomide, a DNA damaging agent [3]. Clinical trials of inhibitors targeting the pathways frequently mutated in GBM have had disappointing results for a variety of reasons, including drug resistance and inclusion of molecularly heterogeneous patients [7, 8]. Preclinical modeling can aid in development of novel therapies by defining whether mutations associated with GBM drive disease pathogenesis and are predictive of drug response.

The PI3K pathway promotes many cancer hallmarks, including survival, proliferation, and migration/invasion [9-12]. PI3K is a heterodimeric lipid kinase composed of catalytic and regulatory subunits encoded by genes such as PIK3CA and PIK3R1, respectively [13, 14]. Pathway activation is mediated by the phosphorylation of PIP2 to PIP3 by the catalytic subunit, resulting in recruitment and activation of effector proteins, including AKT. PI3K signaling (hereafter PI3K) is antagonized by the tumor suppressor PTEN [13, 14]. The PI3K pathway is an attractive therapeutic target in GBM because mutually exclusive mutations in PIK3CA, PIK3R1, and PTEN occur in 46% of patients [15-17].

PI3K activation via Pten deletion, PIK3R1 mutation, or constitutive AKT promotes tumorigenesis in multiple preclinical GBM models [18-23]. For example, we found that Pten deletion cooperates with mutant Kras to activate PI3K and potentiate malignant progression in immortalized mouse astrocytes [19, 24]. Similarly, constitutive AKT cooperated with mutant RAS to promote tumorigenesis in immortalized human astrocytes (NHA) [22, 25]. However, the role of PIK3CA mutations in gliomagenesis has not been experimentally investigated.

PIK3CA is altered in 10% of GBM, mostly via missense mutations [4, 15, 16]. These mutations are generally restricted to 3 functional protein domains (adaptor binding (ABD), helical, and kinase) and are predicted to activate PI3K via distinct biochemical mechanisms [26]. Some PIK3CA found in GBM have been shown to promote tumorigenesis in non-brain tissues, particularly the most prevalent helical (E542K, E545K) and kinase (H1047R) domain mutations [27-29]. However, their role in gliomagenesis has yet to be determined. Here we defined the role of PIK3CA in GBM pathogenesis using NHA and NHARAS. Furthermore, we determined whether these mutations influenced response to single agent and combination PI3K/MEK inhibitors buparlisib and selumetinib, respectively, to elucidate the utility of PIK3CA as a predictive biomarker.

Buparlisib (BKM120) has been proposed as a potential targeted therapy for GBM [30-33]. It is currently being investigated in a Phase II clinical trial in GBM patients (ClinicalTrials.gov, NCT01339052). While there have been no GBM clinical trials of selumetinib, we and others have shown its efficacy in preclinical models [34, 35]. Enriching future clinical trials with likely responders based on mutational profiles promises to improve the chances of clinical success.

Materials and methods

Supplement

Supplemental methods, figures, and tables can be found online.

PIK3CA mutagenesis and lentivirus production

Third generation lentiviral gateway destination vector (pLenti-PGK-Hygro-DEST, #19066, a gift from Eric Campeau and Paul Kaufman), pENTR4 vector (pENTR4-no-ccDB, #17424, a gift from Eric Campeau and Paul Kaufman), hemagglutinin (HA)-tagged wild-type (WT) PIK3CA (PIK3CA, pBabe-puro-HAPIK3CA, #12522, a gift from Jean Zhao), and HA-tagged GFP (GFP, pDEST-Flag-HA-GFP, #22612, a gift from Wade Harper) plasmids were purchased from Addgene (Cambridge, MA) [36-38]. Wild-type PIK3CA (PIK3CA) and GFP were excised and inserted into pENTR4 vector by ligation. PIK3CA (R88Q, C90Y, E542K, E545K, M1043V, H1047R) were generated by point mutagenesis of PIK3CA using the Q5 Site-Directed Mutagenesis Kit (New England Biolabs, Ipswich, MA) per manufacturer’s instructions. GFP, PIK3CA, and PIK3CA were transferred from pENTR4 to pLenti-PGK-Hygro-DEST vectors by recombination as described [36]. All mutations were confirmed by Sanger sequencing (Genewiz, South Plainfield, NJ). Lentiviral particles encoding GFP, PIK3CA, or individual PIK3CA were generated in 293FT cells (Invitrogen, Grand Island, NY) per manufacturer’s instructions.

Cell culture

NHA and NHARAS lines were a kind gift from Russell O. Pieper [25]. Cells were maintained as adherent cultures at 37°C and 5% CO2 in DMEM supplemented with 5% FBS and 1% penicillin/streptomycin (complete DMEM). To generate NHA and NHARAS lines expressing GFP, PIK3CA, or PIK3CA, 135,000 and 120,000 cells respectively were plated on 60 cm2 plates. Lentiviruses were added two days after plating, then incubated with cells overnight in complete DMEM containing 8 μg/ml polybrene (Sigma-Aldrich, St. Louis, MO) at 37°C and 5% CO2. Two days post-infection, transduced cells were selected by culture in complete DMEM plus 300 μg/ml hygromycin B (Gold Biotechnology, St. Louis, MO) for 14 days. Stable gene expression was confirmed by immunoblot for the HA tag on PIK3CA and PIK3CA. All in vitro experiments were performed in DMEM with 2.5% FBS and 1% penicillin/streptomycin (low serum medium) unless otherwise stated.

Drugs

The PI3K inhibitor (PI3Ki) buparlisib (BKM120) and the MEK inhibitor (MEKi) selumetinib (AZD6244) were purchased from MedChem Express (Monmouth Junction, NJ) or Chemietek (Indianapolis, IN) and dissolved in dimethyl sulfoxide (DMSO). In single-dose pharmacokinetics studies in human patients, maximum observed plasma concentrations for buparlisib and selumetinib were 2–5 μM (1–2 μg/mL) and 1.1–2.0 μM (0.5–0.9 μg/mL), respectively [39, 40]. Depending on experimental requirements, drugs were used at or above these clinically relevant dose ranges.

Immunoblots

Control (parental, GFP, and PIK3CA) and PIK3CA NHA and NHARAS cells were treated with either vehicle control (DMSO), buparlisib and/or selumetinib, or serum starved for 24 h. Proteins were isolated and immunoblots were performed as described [19, 24, 35]. Raw immunoblot images are shown in Supplemental Immunoblots. Each blot included both a molecular weight ladder and a reference standard composed of lysates of cultured TRP astrocytes. Quantification was performed using the following formula, where x is an individual blot and i is the target in question: [19, 24, 35]. N = 1–3 blots, mean is denoted in the corresponding figure legend. Bands annotated in red are omitted from final figures.

Cell growth

NHA and NHARAS were plated in triplicate or quadruplicate in 96-well tissue culture plates and absorbance (cell growth) was assessed using CellTiter 96 Aqueous One Cell Proliferation Assay (MTS, Promega, Madison, WI) according to manufacturer’s instructions. Relative absorbance was measured daily as described and fit to an exponential growth equation to calculate rate constants (k) and doubling times [ln(2)/k]. Differences in growth rate constants (k) were compared using the extra-sum-of-squares F test [35].

Cell migration

Migration rate across a cell-free gap was determined using culture inserts according to manufacturer’s instructions (Ibidi, Munich, Germany). Briefly, cells were imaged every 2 hours for 12 hours after creation of a cell free gap using a VistaVision inverted microscope equipped with a 4X objective and a DV-300 camera (VWR, Radnor, PA). Gap closure rates were calculated from 2–12 hours using linear regression and compared via ANCOVA.

Colony formation in soft agar

Colony formation was determined as described with minor modifications [25, 41, 42]. Briefly, cells were suspended in a mixture of DMEM/0.35% agarose (Denville Scientific INC., Holliston, MA) supplemented with 2.5% FBS and 14,000 cells were plated per well in 6-well plates. Cells were maintained for 4 weeks, fixed, and stained with 0.005% crystal violet in 70% ethanol. Plates were imaged on a Typhoon Trio (GE Healthcare) and colonies ≥ 30 μm2 were counted.

Mouse use

This study was carried out in strict accordance with the recommendation of the Guide for the Care and Use of Laboratory Animals of the National Institute of Health. The Institutional Animal Care and Use Committee of the University of North Carolina (Chapel Hill, NC) approved this study (Protocol #16–112). Animals were housed in a SPF facility in IVC cages with enrichment of nestlets and a shelter on corn cob bedding at a density of 2–5 animals per cage. Animals were kept on a 12-hour light/dark cycle at a temperature of 21o +/- 2o Celsius and were monitored daily by experienced laboratory staff following experimental initiation.

Orthotopic xenografts

Control and PIK3CA NHARAS lines were harvested by trypsinization, counted, and suspended in serum-free DMEM with 5% methyl cellulose. Male and female adult athymic (Foxn1nu/nu) nude mice (Charles River, Wilmington, MA; mean age ~3 months; N = 5–10 per group, mean = 9) were anesthetized with Avertin (250 mg/kg) and 2 x 105 GFP, PIK3CA, PIK3CA, PIK3CA, or PIK3CA NHARAS cells were injected into the right basal ganglia of mice (N = 5–10 per group, mean = 9) using the coordinates 1, -2, and -4 mm (A, L, D) from bregma as previously described [19, 24]. Subjects received bupivacaine for local nerve block and a single dose of ketorolac for post-surgical analgesia. Animals were monitored daily for the onset of neurological symptoms (lethargy, hunching, seizures, paralysis, loss of righting reflex) and euthanized via CO2 asphyxiation within 24 hours of onset, immediately followed by brain tissue harvest. Symptoms were frequently severe at first observation due to rapid tumor progression. Animals did not die without euthanasia. Survival was determined by Kaplan-Meier analyses and was compared by log-rank tests.

Drug response

Dose response assays using MTS were performed and IC50 calculated as described [24, 35]. Synergy between MEKi and PI3Ki was determined by BLISS using Combenefit v1.31 [43].

Statistics

GraphPad Prism (La Jolla, CA) was used for statistical analyses. Error bars are SEM unless otherwise stated. P≤0.05 were considered significant.

Results

PIK3CA are frequent in GBM and are heterogeneously distributed across multiple encoded protein domains, including ABD, helical, and kinase () [15, 16]. PTEN deletion or activating AKT mutations cooperate with activated MAPK signaling (hereafter MAPK) to promote tumorigenesis in preclinical glioma models [19, 22]. However, the role of PIK3CA in has not been examined in these models. To this end, we examined the 2 most frequent (hotspot) helical (E542K, E545K) and kinase (M1043V, H1047R) domain PIK3CA found in GBM and other cancer types (). ABD mutations are less prevalent in most cancers. Of these, R88Q is the most common and only recurrent mutation in GBM [4, 15]. We therefore evaluated it, as well as a second, C90Y. We transduced each of 6 PIK3CA into NHA and NHARAS via lentiviral vectors. Parental, GFP, or PIK3CA-transduced lines served as controls.

PIK3CAmut induce PI3K in vitro

Expression of PIK3CA and all 6 PIK3CA was similar in both NHA and NHARAS (), suggesting that phenotypic differences would be attributable to PIK3CA examined. Neither PIK3CA nor ABD PIK3CA significantly activated proximal (pAKT) or distal (pS6) PI3K in serum-starved NHA (). In contrast, E542K and H1047R significantly induced proximal and all 4 helical/kinase mutations induced distal PI3K.

Helical and kinase PIK3CA activate proximal PI3K.

Representative immunoblots (A) and quantification showed that proximal PI3K (pAKT) was increased by helical and kinase mutants in NHA (B). Distal PI3K (pS6) was only increased by H1047R (C) (*, P≤0.02). E542K and H1047R increased proximal PI3K and all helical and kinase mutants increased distal PI3Kcompared to PIK3CA NHA (ǂ, P≤0.04). Representative immunoblots (D) and quantification showed that proximal PI3K was increased by PIK3CA and all mutants (E). Helical and kinase mutants also increased proximal PI3K compared to PIK3CA NHA (ǂ, P≤0.007). Only M1043V increased distal PI3K compared to parental (*, P = 0.03) and PIK3CA (ǂ, P = 0.03) NHARAS (F). Bar graph data are set relative to parental lines (N = 4 biologic replicates). Fold changes in pAKT and pS6 relative to PIK3CA are shown as heatmaps.

We previously showed that mutant Kras cooperates with Pten deletion to activate PI3K in immortalized mouse astrocytes [19]. Therefore, we also determined the effects of PIK3CA on PI3K in NHARAS. PIK3CA and all PIK3CA increased proximal PI3K compared to parental NHARAS (). Furthermore, helical and kinase PIK3CA potentiated proximal PI3K more than PIK3CA. However, distal PI3K was only increased by M1043V in NHARAS (). Taken together, these results suggest that mutant RAS cooperated with ectopic expression of both PIK3CA and PIK3CA to increase proximal PI3K.

There is extensive cross-talk between PI3K and MAPK pathways [44]. We therefore determined the effects of PIK3CA on MAPK. Neither PIK3CA nor any of the PIK3CA significantly altered MAPK (pERK1/2) in NHA and NHARAS (). Thus, PIK3CA activated PI3K without affecting the MAPK pathway.

PIK3CAmut induce NHA proliferation in vitro

PIK3CA, particularly helical and kinase mutants, activated PI3K, suggesting that they may also promote cell growth (). MTS assays using high-serum (10% FBS) showed that PIK3CA and a subset of PIK3CA slightly increased growth rate (≤15%) of rapidly-proliferating NHA (doubling times ≤ 1 day; ). We therefore hypothesized that growth factor concentrations in high-serum media were masking the effects of PIK3CA on NHA growth. Indeed, MTS assays using low-serum (2.5% FBS) revealed increased proliferation of both GFP and PIK3CA NHA compared to parental cells (). While all PIK3CA except C90Y increased proliferation compared to parental and PIK3CA NHA, proliferation rates were similar in all NHARAS lines (). Taken together, these data suggest that PIK3CA promote astrocyte growth in the absence, but not presence, of mutant RAS.

PIK3CA potentiate proliferation and migration in vitro.

MTS assays showed that PIK3CA and all PIK3CA decreased doubling times of NHA (A), but not NHARAS (B) (*, P≤0.02 vs parental, ). PIK3CA, except C90Y, decreased doubling times compared to PIK3CA NHA (ǂ, P≤0.03). Growth rates were analyzed by comparing k values. Error bars are 95% confidence intervals. PIK3CA and PIK3CA, except C90Y, increased migration of both NHA (C) and NHARAS (D) (*, P≤0.04, ). E542K and H1047R also potentiated migration compared to PIK3CA NHA and NHARAS (ǂ, P≤0.005). Fold changes in doubling times and migration rates relative to parental and PIK3CA lines are shown as heatmaps.

PIK3CAmut induce migration in vitro

Complete surgical resection of GBM is precluded by its diffuse brain infiltration [45]. The PI3K pathway has an established role in migration [46]. We previously showed that Pten deletion increased migration of immortalized mouse astrocytes [19]. We therefore determined the effects of PIK3CA on migration of NHA and NHARAS using an in vitro gap closure assay. GFP, PIK3CA and all PIK3CA showed increased migration compared to parental NHA (). PIK3CA and PIK3CA except C90Y migrated faster than GFP NHA (P≤0.04). Similarly, PIK3CA except C90Y migrated faster than GFP and parental NHARAS (P≤0.006, ). Moreover, E542K and H1047R PIK3CA migrated faster than NHA and NHARAS overexpressing PIK3CA ().

PIK3CAmut potentiate cellular transformation and tumorigenesis

Anchorage-independent growth (colony formation in soft agar) is an established marker of cellular transformation [25, 41]. NHARAS, but not NHA, form colonies in vitro and develop high-grade tumors in orthotopic mouse xenograft models [25]. We therefore first determined whether PIK3CA promote NHA colony formation by selecting the most potent mutant in each domain (R88Q, E542K, H1047R) based on their effect on proximal PI3K, proliferation, and migration in NHA (Figs ). Only H1047R induced colony formation relative to GFP and parental NHA (). Since this was the only mutation to potentiate NHA colony formation, we also tested its effect in NHARAS. However, no significant increase in colony formation was evident ().

Helical and kinase PIK3CA potentiate cellular transformation and tumorigenesis.

Only H1047R increased colony formation compared to parental (*, P = 0.03) and PIKCA (ǂ, P = 0.04) NHA (A). H1047R did not affect colony formation of NHARAS (P = 0.5). Orthotopic xenografts of GFP, PIK3CA, and PIK3CA NHARAS (BC). Median survival of mice with R88Q, E542K, or H1047R PIK3CA NHARAS was decreased compared to GFP control tumors (*, P≤0.003). E542K and H1047R PIK3CA also decreased survival compared to PIK3CA (ǂ, P≤0.002) and R88Q PIK3CA (P<0.0001). Fold changes in median survival relative to GFP and PIK3CA NHARAS are shown as heatmaps.

We next performed orthotopic xenografts of GFP, PIK3CA, and PIK3CA in NHARAS to determine whether PIK3CA potentiate tumorigenesis in vivo. Mice with R88Q, E542K, or H1047R PIK3CA NHARAS tumors died more quickly than mice with control GFP tumors (P≤0.003). Additionally, mice with E542 or H1047R PIK3CA tumors succumbed to disease more quickly than mice with tumors that overexpressed either PIK3CA (P≤0.002) or R88Q PIK3CA (P<0.0001; ). Upon histopathological examination, tumor morphology was consistent across genotypes (). Taken together, these results indicate that both the mutated domain and concomitant mutant RAS influence the role of PIK3CA in gliomagenesis in vitro and in vivo.

PI3Ki efficacy is similar regardless of PIK3CA status

Neuro-oncology is transitioning towards precision medicine, wherein tumor mutation profiles are utilized to tailor targeted treatments [8, 47]. Knowing which oncogenic driver mutations influence targeted inhibitor response is a prerequisite. To this end, we determined the effects of PIK3CA on efficacy of the PI3Ki buparlisib in vitro. Buparlisib induced a dose-dependent decrease in growth of control and PIK3CA NHA and NHARAS (). High nanomolar IC50 were evident regardless of the specific PIK3CA () but tended to be slightly higher in NHARAS (). Buparlisib also induced G2/M cell cycle arrest in NHARAS lines regardless of PIK3CA status ().

PI3Ki inhibits growth and ablates PI3K regardless of PIK3CA status.

Buparlisib IC50 were similar regardless of PIK3CA status in NHA (A) and NHARAS (B) (). Fold changes in IC50 relative to parental and PIK3CA are shown as heatmaps. Representative immunoblots of control and PIK3CA NHA (C) and NHARAS (F) treated with buparlisib for 24 h. Immunoblot quantification (DEGH) demonstrated dose-dependent decreases in proximal PI3K (DG), with corresponding increases in MAPK (EH) in NHA (DE) and NHARAS (GH) lines (). Western blots were performed in either biological duplicates or triplicates.

PI3Ki ablate PI3K and potentiate MAPK

While PIK3CA did not alter PI3Ki efficacy in NHA and NHARAS in vitro (), they differentially activated PI3K (). We therefore investigated whether PIK3CA influence PI3Ki-induced changes in pathway signaling. Buparlisib inhibited proximal () and distal () PI3K dose-dependently in control and PIK3CA NHA. We and others have shown that PI3Ki induce alternate MAPK activation in preclinical GBM models [35, 44, 48]. Indeed, buparlisib induced dose-dependent increases in MAPK in NHA, regardless of their PIK3CA status ().

Mutant RAS cooperated with PIK3CA and PIK3CA to potentiate activation of proximal PI3K (), so we investigated whether RAS status influences PI3Ki-induced changes in PI3K and MAPK signaling. Buparlisib induced dose-dependent PI3K inhibition and MAPK activation in both control and PIK3CA NHARAS (). Proximal PI3K inhibition was least pronounced in helical and kinase PIK3CA lines at low buparlisib concentrations, demonstrating that higher PI3Ki doses are required to ablate PI3K in the presence of PIK3CA in NHARAS. These results also indicate that PIK3CA status does not influence alternate MAPK activation.

MEKi efficacy is independent of PIK3CA in NHA

Because PI3Ki promoted MAPK regardless of PIK3CA/RAS status, we determined efficacy of the MEKi selumetinib in control and PIK3CA NHA and NHARAS lines in vitro. Selumetinib caused gradual, dose-dependent decreases in growth () and had similar IC50 in both parental NHA and NHARAS. While PIK3CA status influenced IC50 in neither NHA () nor most NHARAS lines, C90Y and H1047R were slightly more resistant than parental NHARAS. (). Thus, PIK3CA and mutant RAS had little to no effect on MEKi sensitivity.

MEKi inhibits growth and ablates MAPK regardless of PIK3CA status.

Selumetinib IC50 were similar regardless of PIK3CA status in NHA (A), but slightly higher in most PIK3CA NHARAS compared to parental cells (*, P≤0.03) (B) (). Fold changes in IC50 relative to parental and PIK3CA lines are shown as heatmaps. Representative immunoblots of control and PIK3CA NHA (C) and NHARAS (F) treated with selumetinib for 24 h. Immunoblot quantification (DEGH) demonstrated dose-dependent decreases in MAPK in NHA (D) and NHARAS (G) lines. Although proximal PI3K was induced in control and PIK3CA NHA (E), it was only potentiated in GFP and parental NHARAS (H) (). Western blots were performed either 1 or 2 times per experiment (Mean = 1.8).

PIK3CAWT and PIK3CA influence MEKi-induced PI3K activation in NHARAS

Selumetinib inhibited MAPK in both control and PIK3CA NHA () and induced dose-dependent decreases in pERK regardless of PIK3CA status (). We and others have shown that MEKi induces alternate PI3K activation in preclinical GBM models [35, 44, 49]. We extended these findings here, showing that selumetinib potentiated proximal PI3K 1.4-5-fold in control and PIK3CA NHA (). Induction in NHA was least pronounced with E542K and H1047R, the mutations that most potentiated tumorigenesis in NHARAS ().

Mutant RAS cooperated with PIK3CA and PIK3CA to promote activation of proximal PI3K (). We therefore investigated whether PIK3CA influence MEKi-induced changes in MAPK and PI3K in NHARAS. PIK3CA status did not affect MAPK inhibition in selumetinib-treated NHARAS lines (). Selumetinib induced alternate activation of proximal PI3K in GFP and parental NHARAS but ablated it in PI3KCA and all PIK3CA NHARAS (). Taken together, these results indicate that ectopic PIK3CA expression in combination with mutant RAS prevents MEKi-induced alternate PI3K activation.

Dual PI3Ki/MEKi treatment is synergistic in PIK3CA NHA and NHARAS

We and others have shown that dual PI3Ki/MEKi efficacy is increased relative to treatment with either alone [35, 44, 48, 49]. However, the effects of GBM-associated mutations on PI3Ki/MEKi synergism remain unclear. To this end, we determined whether PIK3CA influence the effects of dual PI3Ki/MEKi treatment on NHA and NHARAS growth in vitro. Buparlisib plus selumetinib synergistically inhibited growth in control and PIK3CA NHA (). Synergy was highest in R88Q and E542K NHA relative to PIK3CA and H1047R.

PI3Ki/MEKi synergism in vitro is influenced by PIK3CA and mutant RAS.

Buparlisib and selumetinib inhibited growth and were synergistic in control and PIK3CAmut NHA (A) and NHARAS (B). BLISS showed that synergy was most pronounced with high nanomolar buparlisib and low micromolar/high nanomolar selumetinib in NHA lines. In contrast, synergistic concentrations in NHARAS lines were generally most pronounced at both low micromolar buparlisib and selumetinib.

PIK3CA and PIK3CA marginally decreased MEKi efficacy in NHARAS (). They also cooperated with mutant RAS to prevent MEKi-induced potentiation of proximal PI3K (). Therefore, both PIK3CA and PIK3CA may alter PI3Ki/MEKi synergism when combined with mutant RAS. Dual buparlisib/selumetinib treatment synergistically inhibited growth of all NHARAS lines (). However, synergism was most pronounced at higher drug concentrations in NHARAS relative to NHA lines. Furthermore, synergy was highest in R88Q, E542K, and M1043V mutant NHARAS (). Taken together, these data suggest that mutant RAS and PIK3CA alter PI3Ki/MEKi synergism.

Discussion

PIK3CA differentially activate PI3K and promote gliomagenesis

The vast majority of GBM harbor mutations in core PI3K pathway genes and/or upstream RTK [4]. Activation of PI3K via Pten deletion, PIK3R1 mutations, or constitutively active AKT mutants promotes tumorigenesis in glioma models [19-22]. Here, we determined the effects of ABD, helical, and kinase PIK3CA on gliomagenesis. Both helical and kinase PIK3CA potentiated PI3K, proliferation, and migration of NHA compared to parental and PIK3CA lines (Figs ), but only H1047R kinase mutation potentiated NHA colony formation ().

We and others have shown that PI3K activation via Pten deletion or constitutively active AKT cooperates with MAPK activation to potentiate gliomagenesis [19, 22–24]. We extended these findings here by demonstrating that mutant RAS promoted PIK3CA- and PIK3CA-induced proximal PI3K (). Unlike in NHA, PIK3CA did not increase proliferation of NHARAS, likely due to the rapid proliferation rate of parental cells (). H1047R PIK3CA also did not potentiate colony formation of NHARAS, likely because NHARAS cells are more aggressive and form colonies more readily and at higher density than NHA cells (). However, the E542K and H1047R PIK3CA potentiated malignancy of NHARAS in vivo compared to GFP and PIK3CA (). These results are consistent with previous findings that constitutively active AKT does not enhance proliferation or colony formation of NHARAS in vitro but promotes tumorigenesis in vivo [22]. These data suggest that E542K and H1047R PIK3CA promote in vivo gliomagenesis equally in the presence, but not absence, of mutant RAS.

In contrast to helical and kinase PIK3CA mutations, ABD PIK3CA did not increase PI3K, migration, or colony formation of NHA more than PIK3CA (Figs ). Similar results were obtained with ABD PIK3CA in NHARAS. Moreover, R88Q PIK3CA did not promote tumorigenesis of NHARAS more than PIK3CA in vivo (). Taken together, these results demonstrate that the phenotypic consequences of ABD PIK3CA and ectopic over-expression of PIK3CA are similar. Furthermore, they suggest that ABD PIK3CA may be passenger mutations in GBM. However, ectopic expression of PIK3CA and PIK3CA may not fully recapitulate the effects of PIK3CA when expressed under its endogenous promoter. Furthermore, other cooperating mutations and/or cellular origin may influence the role of PIK3CA in gliomagenesis. Future work will be required to investigate the role of PIK3CA in other genetic and cellular contexts.

PIK3CA do not influence PI3Ki efficacy

The precision medicine initiative seeks to direct treatment with targeted inhibitors based on tumor mutation profiles [8, 47]. However, this requires an understanding of how oncogenic mutations influence drug response. Mutational activation of kinases can cause oncogene addiction, in which tumor cells become reliant upon the activated signaling pathway(s), and are thus highly sensitive to their inhibition [50]. Additionally, kinase mutations can alter drug affinity, thereby altering efficacy [51]. Buparlisib inhibits purified PIK3CAWT and the most common PIK3CAmut, E542K, E545K, and H1047R, with similar IC50 [52, 53]. Because these helical and kinase domain PIK3CA activated PI3K and promoted gliomagenesis, we hypothesized that they would also increase PI3Ki efficacy. However, higher buparlisib doses were required to ablate PI3K in cells expressing PIK3CA, particularly those in the helical and kinase domains, and these mutations did not influence PI3Ki efficacy in vitro (). These results suggest that PIK3CA neither induce oncogene addiction nor enhance PI3Ki sensitivity. Whether they influence efficacy of isoform-specific PI3Ki or inhibitors of downstream kinases, such as AKT and mTOR, remains to be determined.

PIK3CA influence MEKi-induced PI3K activation and PI3Ki/MEKi synergism

We previously found that buparlisib induced widespread kinome changes, including MAPK activation, in immortalized murine astrocytes with Pten deletion and mutant Kras [35]. We expanded these findings here by demonstrating that buparlisib potentiated MAPK regardless of RAS/PIK3CA mutation status (). PIK3CA also had minimal to no effect on sensitivity of NHA and NHARAS to MEKi in vitro ().

We and others have shown that MEKi promote PI3K in preclinical GBM models [35, 44, 49]. We found that selumetinib increased proximal PI3K in control and PIK3CA NHA, and in GFP and parental NHARAS (). Interestingly, this increase was not apparent in PIK3CA and PIK3CA NHARAS (). The mechanism by which ectopic PIK3CA expression in combination with mutant RAS alters MEKi response is unclear. A mutually inhibitory crosstalk between PI3K and MAPK is mediated by p70S6K in glioma stem cells [44]. MAPK inhibition induces PI3K in non-GBM cell lines via removal of a negative feedback loop on RTK [54]. Similarly, selumetinib induces widespread kinome changes in breast cancer models, including increased expression and activity of multiple RTK [55]. Taken together, these results suggest that PIK3CA and PIK3CA may cooperate with mutant RAS to alter MEKi-induced dynamic kinome changes, particularly as it pertains to PI3K activation.

Dual PI3Ki/MEKi treatment is effective in multiple preclinical GBM models [35, 44, 48, 49]. It remains unclear whether the underlying genetics of these models influence drug synergism. We therefore determined if PIK3CA affected PI3Ki/MEKi synergism in the presence and absence of mutant RAS. Consistent with other GBM models, we found that dual buparlisib/selumetinib treatment was synergistic in NHA and NHARAS lines (). However, RAS/PIK3CA status influenced drug response. Higher concentrations of buparlisib and selumetinib were required to maximize synergism in NHARAS lines compared to NHA. Furthermore, synergy was generally greater in R88Q and E542K NHA and NHARAS compared to those with either PIK3CA or H1047R. Taken together, these results suggest that GBM patients with helical PIK3CA may be most sensitive to dual PI3Ki/MEKi treatment. Given that ABD PIK3CA showed no significant tumorigenic effects in vitro and in vivo, their utility in predicting PI3Ki/MEKi synergy remains questionable.

Conclusion

Defining the role of frequently occurring mutations in GBM pathogenesis and drug response can aid identification of predictive biomarkers. Our results demonstrate that PIK3CA differentially promote gliomagenesis and do not predict PI3Ki sensitivity but do impact PI3Ki/MEKi synergism.

Frequency and distribution of PIK3CA.

Lollipop plot of PIK3CA missense (green), in-frame deletion (brown), and truncating mutations (black) in GBM (A) and all published TCGA datasets (B). PIK3CA missense mutations investigated here are indicated. PIK3CA point mutations were evident in 10.3% of GBM cases from the TCGA (N = 273), with each mutation investigated here representing ≤1% of total mutations. Data were downloaded from cBioPortal (http://www.cbioportal.org/) on March 10, 2017. Ribbon diagram of PIK3CA with mutations investigated highlighted (C) (R88Q = light green; C90Y = dark green; E542K = pink; E545K = red; M1043V = purple; H1047R = blue). Model was generated in PyMOL. (Schrödinger, New York City, NY) using a script downloaded from cBioPortal.13,14

(TIF)

Click here for additional data file.

PIK3CA and PIK3CA are expressed at similar levels.

Representative immunoblots (AC) and quantification (BD) of HA-tagged PIK3CA showed that PIK3CA and PIK3CA were expressed at similar levels in NHA (BC) and NHARAS (CD) (ANOVA, P≥0.3). Bar graph data were set relative to PIK3CA lines (N = 3–4 biologic replicates, Mean = 3.5).

(TIF)

Click here for additional data file.

PIK3CA do not alter MAPK.

Representative immunoblots (AC) and quantification (BD) showed that PIK3CAmut did not alter MAPK (phosphorylation of ERK1/2, pERK) in either NHA (AB) or NHARAS (CD) (P≥0.93). Bar graph data were set relative to parental lines (N = 3–4 biologic replicates, Mean = 3.5). Fold changes in pERK relative to PIK3CAWT lines are shown as heatmaps.

(TIF)

Click here for additional data file.

PIK3CA have minor effects on growth in high-serum culture.

MTS assays (A) showed that PIK3CA and a subset of PIK3CA slightly increased growth (reduced doubling times) compared to parental NHA (*, P≤0.03) when grown in media containing high (10%) FBS. (B). R88Q growth was slightly slower than PIK3CA NHA (ǂ, P = 0.01). Statistical analyses of growth rates were performed by comparing k values. Fold changes in doubling times relative to parental and PIK3CA lines are shown as heatmaps. Error bars in B are 95% confidence intervals.

(TIF)

Click here for additional data file.

Influence of PIK3CA on proliferation, but not migration, is dependent on mutant RAS in vitro.

Growth of control and PIK3CA NHA (A) and NHARAS (B) (). Growth was determined by assessing changes in relative absorbance daily by MTS. Migration of control and PIK3CA mutant NHA (C) and NHARAS (D) across a gap ().

(TIF)

Click here for additional data file.

Histopathologic features of malignancy are consistent across genotypes.

Hematoxylin and eosin staining of tumors from PIK3CA (A), R88Q (B), E542K (C), and H1047R (D) PIK3CA mice revealed malignant histopathologic features typical of human gliomas, including cytologic and nuclear atypia, tumor giant cells, and mitotic figures (white arrows). Scale bar = 100 μm.

(TIF)

Click here for additional data file.

PI3Ki inhibits growth independent of PIK3CA status in vitro.

MTS assays showed that buparlisib caused dose-dependent decreases in growth of control and PIK3CA NHA (A) and NHARAS (B). Buparlisib IC50 were similar between control and all 6 PIK3CA NHARAS (C).

(TIF)

Click here for additional data file.

PI3Ki induces G2/M cell cycle arrest in NHARAS cells regardless of PIK3CA status.

Micromolar doses of buparlisib induced G2/M cell cycle arrest within 48 h in control and PIK3CAmut NHARAS.

(TIF)

Click here for additional data file.

PI3Ki inhibits distal PI3K regardless of PIK3CA status.

Representative immunoblots of control and PIK3CA NHA (A) and NHARAS (C) 24 h after buparlisib treatment. Immunoblot quantification demonstrated dose-dependent inhibition of distal PI3K in all NHA (B) and NHARAS (D) lines (N = 1–3 biologic replicates, Mean = 1.7).

(TIF)

Click here for additional data file.

PI3Ki inhibits proximal PI3K signaling and induces MAPK signaling in all control and PIK3CA NHARAS.

Representative immunoblots (A) and quantification of proximal PI3K (B) and MAPK (C) showed that within 24 h, buparlisib induced dose-dependent inhibition of PI3K signaling, with concurrent induction of MAPK in parental, GFP, PIK3CAWT, and all 6 PIK3CAmut NHARAS lines (N = 2–3 biologic replicates, Mean = 2.7).

(TIF)

Click here for additional data file.

MEKi inhibits in vitro growth in all PIK3CA lines.

MTS assays showed that selumetinib caused dose-dependent decreases in growth of control and PIK3CA NHA (A) and NHARAS (B). Selumetinib IC50 were slightly increased by PIK3CA and all PIK3CA, except C90Y and H1047R, compared to parental NHARAS (*, P≤0.03) (C).

(TIF)

Click here for additional data file.

MEKi induction of proximal PI3K signaling is abrogated in PIK3CA and PIK3CA NHARAS.

Representative immunoblots (A) and quantification of MAPK (B) and proximal PI3K (C) showed that selumetinib caused dose-dependent inhibition of MAPK regardless of PIKCA or PIK3CA status, but concurrent induction of proximal PI3K only occurred in parental and GFP NHARAS (N = 2–3 biologic replicates, Mean = 2.7).

(TIF)

Click here for additional data file.

PI3K/MEKi synergism is differentially influenced by PIK3CA in vitro.

Buparlisib and selumetinib were synergistic in parental and PIK3CA NHARAS ()

(TIF)

Click here for additional data file.

Supporting information supplement.

(DOCX)

Click here for additional data file.

Supplementary immunoblots.

(PPTX)

Click here for additional data file.