Preclinical evaluation and reverse phase protein Array-based profiling of PI3K and MEK inhibitors in endometrial carcinoma in vitro

The human endometrial cancer cell lines AN3CA, HEC1A, HEC1B, HEC50, ECC1, EFE184, ETN1, KLE and SKUT2 were obtained from the Department of Systems Biology, the University of Texas M.D. Anderson Cancer Center (Houston, USA), EN, MFE296 and MFE280 cell lines were purchased from the German Collection of Microorganisms and Cell Cultures (DSMZ; Braunschweig, Germany). EN-1078D cell line was kindly provided by Prof Eric Asselin (University of Quebec), where informed consent was attained and research studies approved by the Montreal University Institutional Review Board [14]. The EC cell lines AN3CA, HEC1A, KLE and SKUT2 cell lines were grown in DMEM-F12 (Sigma #D6421), with 10% FBS (Lonza #DE14-801F); EN-1078D was grown in DMEM-F12 (Sigma #D6421) and 10% FBS with the addition of 50 μg/mL gentamicin (Invitrogen #15710–049), HEC50 was grown in DMEM-F12 with the addition of 2 mM glutamine (Sigma #D8437) and 10% FBS. ECC1, EFE184, EN, ETN1 and MFE280 were grown in RPMI-1640 (Sigma #R8758) supplemented with 10% FBS. MFE296 was grown in RPMI-1640 (Sigma #R8758) with 10% FBS and insulin-transferrin-selenium-X (Invitrogen #51500). HEC1B and HEC59 were grown in MEM (Sigma # 2279) with the addition of 10% FBS, 1% non-essential amino acids and 10 mL glutamine (Sigma #D8437). Growth media were supplemented with 100 μg/ml streptomycin and 100 U/ml penicillin (Invitrogen) and all cell lines were cultured at 37 °C in 5% CO₂. Cell lines were fingerprinted at the beginning of the study to confirm their identities.

Genomic DNA was extracted using QIAamp DNA Mini Kit (Qiagen). We designed high-throughput assays for somatic mutations in KRAS, PIK3CA, PIK3R1, PIK3R2 and PTEN, and applied a mass spectroscopy–based approach to detect single nucleotide polymorphisms (MassARRAY, Sequenom) as described previously [5, 15].

The inhibitors pictilisib (GDC-0941), apitolisib (GDC-0980) and cobimetinib (GDC-0973) were obtained from Genentech, Inc. 1 × 10⁴ cells/well were plated into flat-bottomed, 96-well plates and allowed to attach overnight. All drug treatments including combinations were tested in triplicate during a 5-day incubation period with serial dilutions of drug in a final volume of 200 uL. Drug-free controls were included in each assay. DMSO controls were also performed for each assay. Plates were incubated at 37 °C in a humidified atmosphere with 5% CO₂ and cell viability was determined using an acid phosphatase assay as described previously [16, 17].

RPPA is a high through-put antibody based technique used for functional proteomic assessments of a large number of tumour samples and it offers a platform for comparison of the relative protein expression between these samples [18]. We plated cells in 6-well plates, allowed them to reach 60–80% confluence and extracted their proteins as described by us and others previously [15, 19, 20]. Cells were washed with cold PBS and lysed with ice-cold lysis buffer (1% Triton X-100, 50 mm HEPES, pH 7.4, 150 mM NaCl, 1.5 mM MgCl2, 1 mM EGTA, 100 mM NaF, 10 mM Na pyrophosphate, 1 mM Na3VO4, 10% glycerol) supplemented with proteinase inhibitors (Roche Applied Science, Indianapolis, IN) as described previously [19, 21].

For RPPA, four-fold serial dilutions of protein extracts from EC cell lines were performed. Serially diluted lysates were arrayed on Oncyte Avid nitrocellulose-coated slides (Grace Bio-Labs, Bend, OR) using a QArray 2 arrayer (Molecular Devices, Wokingham, UK). Diluted samples were robotically printed on multiple slides that included positive and negative controls. Slides were probed with primary antibodies (see Additional file 1: Table S1) followed by a secondary antibody – either goat anti-rabbit IgG (1:5000) (Vector Laboratories, Burlingame, CA) or rabbit anti-mouse IgG (1:10) (Dako) depending on the particular primary antibody. Signals were amplified using a Dakocytomation-catalysed system (Dako) and visualized by chromogenic detection (using diaminobenzidine). Slides were then scanned, analysed, and quantified using customized software (Microvigene, VigeneTech Inc.) to generate spot intensities.

Heat-maps were generated from normalized data using unsupervised hierarchical clustering analysis performed with publicly available Morpheus software (https://​software.​broadinstitute.​org/​morpheus/​).

Drug combination assays synergy was assessed using the combination index method [22] using CompuSyn software (Combosyn Inc., Paramus, NJ, USA) [12]. A combination index (CI) of < 1 is considered to indicate synergy, = 1 is considered additive and > 1 is considered antagonistic [22]. We applied the Student’s t-test to test the significance of pair-wise comparisons using GraphPad-Prism.5. P-value < 0.05 was considered statistically significant.

By applying a mass spectroscopy–based analysis (MassARRAY, Sequenom) to the 13 EC cell lines we confirmed their mutational status. As we and others have previously shown in primary endometrioid endometrial carcinomas (EECs) [23, 24], we found that multiple members of the PI3K pathway and sometimes KRAS were mutated in the same cell lines, including concomitant mutations (Tables 1 and 2).

Further, using RPPA, we found that PTEN protein expression was retained in the majority of the EC cell lines with the exception of three cell lines (EN, AN3CA, HEC59) which are all PTEN-mutated and had relatively high levels of phosphorylated AKT. We observed that the presence of a PTEN mutation was significantly associated with activation of PI3K pathway as determined by high levels of AKT phosphorylation at Thr³⁰⁸ (P = 0.005) and at Ser⁴⁷³ (P = 0.008) in EC cell lines, however no significant association was observed between PIK3CA, PIK3R1 or KRAS mutation and AKT phosphorylation levels (see Additional file 2: Table S2). These results are inconsistent with previous studies in breast cancer cell lines [15] and endometrial cancer cell lines [24], providing PTEN mutation or PTEN loss as important activators of PI3K-mediated pro-survival signalling through AKT.

Unsupervised hierarchical clustering revealed three major cell line clusters (designated as C1, C2 and C3) (Fig. 1). Cluster C1 contained one cell line only, the PTEN-mutated MFE296, distinguished by high levels of PTEN, total MAPK-ERK1/2 expression and MAPK-ERK1/2 phosphorylation at Thr²⁰²/Tyr²⁰⁴. Cluster C2 and C3 were composed of four (HEC1A, MFE280, EFE184, KLE) and five (AN3CA, EN-1078D, HEC1B, HEC50, SKUT2) cell lines, respectively (Fig. 1). Cell line cluster C2 was composed of three of the PTEN-mutated cell lines (EN, HEC59, ETN1) of which two expressed low levels of PTEN (EN, HEC59). As compared with cluster C3, cell lines in cluster C2 contained higher levels of higher levels of phosphorylated AKT at Ser⁴⁷³ and Thr³⁰⁸, phosphorylated c-Raf at Ser³³⁸, phosphorylated S6 ribosomal protein at Ser^235/236 and P70 S6 kinase, and lower levels of phosphorylated NFkβ p65^Ser536, phosphorylated MEK1/2^Ser217/221, phosphorylated MAPK-ERK1/2^{Thr202/Tyr204} and phosphorylated GSK-3β^Ser9. Thus it seems that cluster C2 is defined by activation of the PI3K pathway and cluster C3 by activation of the RAS/MAPK pathway. Unsupervised hierarchical clustering of the 66 proteins also revealed eight protein clusters (designated P1 to P8) (Fig. 1).

We observed that EC cell lines possessing a KRAS mutation (HEC50, HEC1A and HEC1B) were more sensitive to the MEK inhibitor cobimetinib (IC₅₀s from 0.001 to 0.033 μM) when compared with the other cell lines tested (IC₅₀s from 0.101 to 39.337 μM) (Table 1). Conversely, EC cell lines possessing a PTEN mutation (MFE296, EN-1078D, ETN1, AN3CA, HEC59, EN) were more resistant to MEK inhibition (IC₅₀s from 0.178 to 36.511 μM) when compared with the cell lines without a PTEN mutation (IC₅₀s from 0.101 to 0.001 μM (with the exception of the wild-type cell line KLE with a IC₅₀ of 39.337 μM).

In terms of the dual inhibition of PI3K and mTOR, we observed that EC cell lines with coexisting PTEN and PIK3R1 mutations (HEC59, AN3CA and EN-1078D) were more sensitive to the PI3K/mTOR inhibitor apitolisib (IC₅₀s from 0.018 to 0.079 μM) when compared with the others (IC₅₀s from 0.196 to 7.253 μM) (Table 1). Two wild-type cell lines (EFE184, KLE) showed different responses to all three inhibitors, whilst KLE was very resistant to pictilisib, apitolisib and cobimetinib treatments (IC₅₀s 2.796, 7.253 and 39.337 μM, respectively), the other wild-type cell line EFE184 was markedly sensitive with the IC₅₀s of 0.372, 0.368 and 0.149 μM for pictilisib, apitolisib and cobimetinib, respectively (Table 1).

Next, we examined if expression levels of PTEN protein play an important role in response to therapies targeted to the PI3K and MEK pathways. Thus, we compared EC cell lines with PTEN protein loss (AN3CA, EN and HEC59) against EC cell lines with retained PTEN protein, and found that cell lines exhibiting PTEN loss were more sensitive to the PI3K inhibitor pictilisib (IC₅₀s from 0.057 to 1.597 μM) when compared with cell lines with retained PTEN expression (IC₅₀s from 0.242 to 2.796 μM). In terms of apitolisib sensitivity, cell lines with PTEN loss were more sensitive (IC₅₀s from 0.018 to 0.548 μM) when compared with PTEN retained cell lines (IC₅₀s from 0.106 to 7.253 μM). However, these cell lines with PTEN loss were more resistant to MEK inhibition (IC₅₀s from 0.518 to 36.511 μM) when compared with cell lines with retained PTEN expression (IC₅₀s from 0.001 to 1.077 μM), (again with the exception of the MEK inhibitor resistant wild-type KLE cell line with high levels of PTEN), indicating PTEN loss as a dominant marker of sensitivity to PI3K inhibition and resistance to MEK inhibition (Fig. 2c).

Next, we sought to examine whether dual inhibition of both the PI3K and MEK pathways might result in synergistic effects on cell viability. We found that the combination of pictilisib with cobimetinib, and the combination of apitolisib with cobimetinib both resulted in synergistic growth inhibition in most of the cell lines tested (2/10), with the exception of 2 cell lines (EN, HEC50) (Table 1, Fig. 4a–b). Endometrial cancer cell lines in which synergism was observed to the combination of pictilisib (PI3Ki), apitolisib (PI3K/mTORi) and cobimetinib (MEKi) are shown in Additional file 3: Figure S1.

We observed strong antagonism as indicated by CI values greater than 1 (see Materials and Methods), for EN cell line with the combination of pictilisib and cobimetinib (CI value = 1.555), and with the combination of apitolisib and cobimetinib (CI value = 1.200). This antagonism may occur due to the high resistance to the MEK inhibitor cobimetinib in this cell line (IC₅₀: 36.511 μM) when compared with pictilisib and apitolisib therapies alone (IC₅₀s 1.597 and 0.548 μM, respectively) (Table 1). We also determined strong antagonism for HEC50 cell line with the combination of pictilisib and cobimetinib, and with the combination of apitolisib and cobimetinib (CI values 4.921 and 6.613, respectively). In contrast to EN cell line, antagonism in HEC50 cell line may occur due to the high sensitivity to MEK inhibition (IC_50: 0.001 μM) (Table 1).

HEC50 and EN cell lines in which antagonism was observed to the combination of pictilisib (PI3Ki), apitolisib (PI3K/mTORi) and cobimetinib (MEKi) are shown in Fig. 3.

We detected differential responsiveness to the PI3K inhibitors and to the MEK inhibitor between the cell lines with the presence of PTEN mutations and PTEN loss being significantly associated with PI3K pathway activation, enhanced sensitivity to PI3K inhibition and remarkable resistance to the MEK inhibitor. Additionally, based on proteomics profiling, cell lines in cluster C2 (HEC59, EN, ETN1) were more resistant to the MEK inhibitor cobimetinib (IC₅₀s 3.289, 36.511 and 1.077 μM, respectively) than cell lines in cluster C3 (IC₅₀s from 0.001 to 0.518; P = 0.04), yet again suggesting that PTEN mutations and the level of PTEN protein expression are important events in cell responsiveness to therapies targeted to the PI3K and RAS/MAPK pathways. Therefore, based on differential sensitivity to the PI3K and MEK inhibitors and differential activation of the PI3K and RAS/MAPK signalling pathways associated with the clustering in Fig. 1b which seems to be largely driven by PTEN, PIK3CA and KRAS mutations and PTEN protein levels, we classified our cell lines into these 4 variables as follows:

In this classification system, PTEN mutations regarded as dominant over PIK3CA mutations because they, and not PIK3CA mutations, are associated with AKT activation as shown earlier. The comparative responsiveness of each group to the three targeted inhibitors is shown in Fig. 2. Comparison of these groups revealed that group 2a and group 2b cell lines with PTEN mutations were most sensitive to the PI3K inhibitor pictilisib, followed by group 1 cell lines with PIK3CA mutations only (IC₅₀ = 0.43 ± 0.01). In the PTEN-mutated group 2 cell lines, pictilisib responsiveness was independent of PTEN protein expression status. KRAS-mutated cell lines in group 3 and wild-type cell lines in group 4were the most resistant to this inhibitor (IC₅₀ values 1.03 ± 0.53 and 1.58 ± 1.21, respectively). Group 1 and group 4 cell lines did not differ significantly in their sensitivities to pictilisib, despite the former having and the latter lacking PIK3CA mutations.

Responses to the PI3K/mTOR inhibitor apitolisib overall resembled those to the PI3K inhibitor pictilisib, with for example PTEN-mutated cell lines with retained PTEN (group 2a) were being more sensitive (IC50 = 0.17 ± 0.05) compared to wild-type cell lines in group 4 (IC₅₀ = 0.33 ± 0.06) (Fig. 2b).However, group 3 KRAS-mutated cell lines were more sensitive to this inhibitor (IC₅₀ = 0.19 ± 0.09 μM) compared with group 4 cell lines (IC₅₀ = 3.81 ± 3.44 μM). In general, the differences between the groups in terms of apitolisib sensitivity were not as marked as the differences between the groups in terms of pictilisib sensitivity.

In terms of MEK inhibition, cobimetinib induced a strong inhibition in all except group 2b and group 4 EC cell lines (IC₅₀ = 10.74 ± 4.47 μM and 1.28 ± 0.60 μM, respectively), particularly in group 1 cell lines with PIK3CA mutations (IC₅₀ = 0.31 ± 0.08 μM) and group 3 KRAS-mutated cell lines (IC₅₀ = 0.05 ± 0.01 μM) (Fig. 2c). However, responses to MEK inhibition varied noticeably between PTEN-mutated cell lines with retained PTEN protein in group 2a (IC₅₀ = 0.31 ± 0.08 μM) and PTEN-mutated cell lines exhibiting loss of PTEN protein in group 2b (IC₅₀ = 10.74 ± 4.47 μM).

Next, we sought to determine how PI3K and MEK inhibitors affect protein signalling in EC cell lines. We selected a group of 6 EC cell lines, one cell line from the PIK3CA-mutated group 1 (SKUT2), one cell line from the PTEN-mutated PTEN retained cell lines in group 2a (EN-1078D), one cell line from the group 2b cell lines with PTEN loss (AN3CA), two cell lines from the KRAS-mutated group 3 (HEC1B; –PIK3CA-mutated, and HEC50; –PIK3CA-wild-type), and finally one cell line from the wild-type group (KLE).

The data in Fig. 4a–e demonstrate changes in phosphorylation and expression of AKT and RAS/MAPK signalling proteins after treatment with 0.1 μM of the PI3K, PI3K/mTOR and MEK inhibitors, and the combinations for 30 min and 6 h in 6 EC cell lines. Similar to observation in patients [25], we observed decreases in phosphorylation of AKT following treatment with the PI3K inhibitors pictilisib and apitolisib when compared with control-treated cells. We also observed that reductions in AKT, S6 ribosomal protein and GSK-3β phosphorylation occurred in some cell lines. Figure 4c also demonstrates a general decrease in the phosphorylation of MAPK-ERK1/2^{Thr202/Tyr204} after treatment with the MEK inhibitor.

We found exposure time to be an important variable, particularly in some EC cells, with some pathways up-regulated immediately after drug exposure down-regulated at longer exposure. For example, 30 min exposure to each of the PI3K inhibitors and the MEK inhibitor resulted in up-regulation of p38 MAPK^{Thr180/Tyr182} phosphorylation in the PTEN mutated, PTEN retained cell line EN-1078D, with down-regulation of p38 MAPK^{Thr180/Tyr182} phosphorylation at longer exposure (Fig. 4a–c).

While treatment with the PI3K inhibitor pictilisib significantly suppressed the levels of phosphorylated S6 ribosomal protein (especially at Ser^240/244), phosphorylated AKT (at Thr³⁰⁸ and Ser⁴⁷³) and phosphorylated GSK-3-β^Ser9 in most EC cell lines compared with control-treated cells (Fig. 4a), it significantly induced phosphorylation of MEK1/2^Ser217/221, MAPK^{Thr202/Tyr204} and p38^{Thr180/Tyr182} in some cell lines, most notably in the KRAS-mutated cell lines HEC1B and HEC50 at most time point, a likely sign of a feedback loop regulation. While this occurred after 30 min in HEC1B, it was more obvious after 6 h in HEC50, possibly indicating different adaptations by feedback regulation (Fig. 4a). This data suggests that pictilisib-induced activation of the RAS/MAPK pathway is most likely to occur in KRAS-mutated cell lines, possibly underlying the resistance of these cell lines to this PI3K inhibitor.

Consistent with growth inhibition responses to the PI3K/mTOR inhibitor apitolisib resembling those to the PI3K inhibitor pictilisib in the cell line panel, the protein signalling effects of these inhibitors were also very comparable. In general, this inhibitor led to inhibition of AKT phosphorylation in the cell lines in comparison with control-treated cells. As with pictilisib, there was no obvious correlation between degree of the PI3K pathway inhibition and apitolisib sensitivity. Apitolisib induced increases in MEK, MAPK and p38 MAPK phosphorylation were also observed, in particular at 30 min in the KRAS-mutated cell lines. Of the 6 cell lines tested herein, AN3CA (PTEN-low) and KLE (wild-type) were the most resistant to the MEK inhibitor cobimetinib. While this inhibitor decreased MAPK phosphorylation at Thr²⁰²/Tyr²⁰⁴ in all cell lines (Fig. 4c), its effect seemed to be least marked in these two cell lines. Further, at 6 h, cobimetinib treatment was associated with increased phosphorylation of AKT^Ser473 (in AN3CA) and of GSK-β^Ser9 and S6 ribosomal protein^Ser240/244 (in KLE). While MEK phosphorylation at Ser^217/221 was in general increased in response to cobimetinib, this effect was most marked in the two most sensitive and KRAS-mutated cell lines HEC1B and HEC50 (Fig. 4c). This increase in MEK phosphorylation is likely to be a feedback loop-induced mechanism that seems most marked in the presence of mutated KRAS.

Dual blockage of the PI3K and RAS/MAPK pathways has been shown to synergistically inhibit tumour cell growth in different cancers [11, 26, 27]; including endometrial cancer [28, 29]. Indeed, we observed that treatment with PI3K and MEK inhibitor combinations was more effective than each inhibitor alone in most EC cell lines tested (9 out of 11) (Table 1). While the PI3K and MEK inhibitor combinations did not abrogate MEK activation induced by cobimetinib alone, they did abrogate the increases in AKT (e.g. in AN3CA) and S6 ribosomal protein^Ser240/244 (e.g. in KLE) phosphorylation seen with the MEK inhibitor alone (Fig. 4d), possibly contributing to the synergy observed in these cell lines. In HEC50, we also observed that the significant inhibition of MAPK^{Thr202/Tyr204} phosphorylation induced by cobimetinib was no longer present with the pictilisib:cobimetinib and apitolisib:cobimetinib combinations, especially at 6 h (Fig. 4d), possibly contributing to the antagonism seen in this cell line with the drug combinations.