The Phosphatidylinositol 3-Kinase Pathway as a Potential Therapeutic Target in Bladder Cancer

Urothelial carcinoma of the bladder or bladder cancer is among the top 10 most common cancers in the United States, leading to approximately 11,820 deaths each year (1). Five to 10% of patients are already metastatic at diagnosis, and approximately 50% of patients will develop local or distant disease recurrence after radical cystectomy (2). Cisplatin-containing combination chemotherapy has been the standard of care for advanced bladder cancer since the late 1980s, but provides a median survival of only about 14 months for metastatic bladder cancer. Little improvement in survival has been attained during the last few decades (3). Because of the limited efficacy of cisplatin-based chemotherapy in the treatment of advanced bladder cancer, novel therapeutics are currently under investigation, either as single agents or in combination with standard chemotherapy regimens. The Cancer Genome Atlas Project (TCGA) identified multiple new “druggable” targets of bladder cancer, among which 42% were within the PI3K/AKT pathway (4). Despite the recurrent genetic aberrations that have been identified to date, there is no targeted therapy approved for bladder cancer by the FDA (5).

PI3K/AKT is a critical signal transduction pathway that regulates apoptosis, survival, and proliferation. Activation of the PI3K/AKT pathway frequently occurs due to multiple molecular alterations, including mutations (PIK3CA, AKT1), gene amplification (PIK3CA, AKT1, AKT2), or loss of expression of tumor suppressors (PTEN, TSC2; ref. 6). Activation of the PI3K/AKT pathway in cancer cells may overcome the proapoptotic effect of anticancer drugs, mitigate the cytotoxic effect of chemotherapy, and cause drug resistance. Therefore, inhibiting this signaling pathway may facilitate chemotherapy treatments (7, 8). Preclinical studies and early clinical trials indicate that the treatment of several cancers including breast (9), lung (10), medulloblastoma (11), and pancreatic cancer (6) could benefit from PI3K inhibitors. It is plausible that such targeted therapeutic approaches against the PI3K/AKT pathway can be used in patients with advanced bladder cancer, either as a single agent or in combination, to decrease drug resistance and augment the efficacy of chemotherapy.

According to the structural characteristics and substrate specificity of PI3K, it can be divided into three classes (I, II, and III). Of these, only the class I PI3K composed of the p110α, β, δ, and γ isoforms is able to activate the downstream effector AKT (12, 13). Pictilisib, also known as GDC-0941, is an oral class I pan-PI3K inhibitor with activity in the nanomolar range against all class I PI3K isoforms. Pictilisib binds to the ATP-binding pocket of PI3K p110, prevents activation of PI3K and formation of PIP3, and phosphorylation of downstream targets such as AKT (14, 15). Pictilisib is now in advanced stages of clinical trials (16–18). In a phase I, open-label dose-escalation clinical trial of patients with advanced solid tumors, pictilisib was found to have antitumor activity, on-target pharmacodynamics activity, and an acceptable safety profile at doses of ≥100 mg. In addition, pictilisib was well tolerated at doses up to 330 mg with mild to moderate adverse effects and no treatment-related deaths (16).

Here we report the activity of pictilisib in selected bladder cancer cell lines and patient-derived xenograft (PDX) models of human bladder cancer. PDXs retain morphologic and genetic fidelity, and therefore, more closely mimic the behavior of patient cancers (19, 20). Our research group has developed and characterized a panel of 26 bladder cancer PDX models for screening of effective drugs or drug combinations (5). We demonstrated that pictilisib blocked the PI3K/AKT pathway and sensitized bladder cancer cells to cisplatin and gemcitabine treatment in vitro and in vivo. We also demonstrated that the phosphorylation of ribosomal protein S6 (p-S6), a downstream effector of the PI3K/AKT pathway, may serve as a potential biomarker to predict the response to pictilisib.

The bladder cancer cell lines TCCSUP, T24, J82, 5637, and RT4 were purchased from the ATCC in 2007 and stored in liquid N₂. Cells were thawed and cultured with the recommended medium and condition by the ATCC. The cell lines were not tested and authenticated by the authors. Pictilisib, sorafenib, and gemcitabine were purchased from LC Laboratories; cisplatin was purchased from EMD Biosciences, Inc.; antibodies against p-AKT(S473), total AKT (t-AKT), PTEN, p-ERK(Thr202/Tyr204), total ERK (t-ERK), p-S6(S240/244), p-P70S6K(T389) p110α, cleaved caspase-3, Ki-67, and GAPDH were purchased from Cell Signaling Technology; β-actin antibody was purchased from Sigma-Aldrich. The MTS cell viability assay was purchased from Promega. IHC kits were purchased from BioGenex. Propidium iodide and Annexin V were purchased from Biolegend.

Pictilisib and sorafenib were dissolved in dimethyl sulfoxide (DMSO) as 10 mmol/L stock solutions. Cisplatin and gemcitabine stock solutions (10 mmol/L) were prepared in PBS. Cells were seeded into 96-well plates (Becton Dickinson) overnight. Drugs were diluted in culture media and added the next day, and the plates were incubated for an additional 72 hours. The control group contained 0.2% DMSO. Cell viability assays were performed according to the manufacturer's protocol. The absolute 50% inhibitory concentrations (IC₅₀) were calculated according to Sebaugh (21) and dose–response curves were generated with GraphPad Prism 5 software (GraphPad Software Inc.). The dose range of two drugs for calculating combination indices (CI) was determined on the basis of Chou (22). CI values were calculated with CompuSyn software (http://www.combosyn.com/).

The procedures for performing Western blot analysis were described previously (23), and IHC staining was conducted according to the manufacturer's standard protocol (BioGenex).

Cells were plated in 6-well plates and incubated overnight. Drugs were added the next day, and cells were incubated for an additional 48 hours. Cell apoptosis was measured by Annexin-V-fluorescein isothiocyanate (FITC) and propidium iodide (PI) staining according to the manufacturer's protocol. Cell-cycle analysis was performed as described previously (24). Flow cytometry was performed at the UC Davis Comprehensive Cancer Center Flow Cytometer Shared Resource and data were analyzed using FlowJo software (FlowJo).

All animal studies were performed according to protocols approved by the Jackson Laboratory (JAX) and UC Davis Institutional Animal Care and Use Committees (protocols no. 12027 and 17794). PDX models (BL0269, JAX#TM00015; BL0293, JAX#TM00016; BL0440, JAX#TM00024) were developed as previously described. Fresh PDX specimens (3–5mm³) were implanted subcutaneously into the flanks of 4–5 week old NOD scid gamma (NSG) mice (5). Mice were randomized into each group (eight mice per group) when tumor volumes reached 100 to 300 mm³, and followed by initiation of drug treatment. Pictilisib was dissolved in 30% PEG-300 with 1% Tween80 and 1% DMSO. Mice were dosed daily with 100 mg/kg by oral gavage during the course of study. Cisplatin was given at 2 mg/kg intravenously at days 1, 2, 3, and days 15, 16, 17. Gemcitabine was given by intraperitoneal injection with a dose of 150 mg/kg once a week for 4 weeks. Drugs were given simultaneously in the combination treatment group. In the sequential treatment group, drugs were administered similarly to the combination treatment group, but the animals were not treated with pictilisib during days 1–3 and days 15–17. The tumor volume was calculated using the formula (length × width²) × 0.5. Tumor volumes and body weight were recorded twice weekly. Mice with tumor volume of ≥2 cm³ or with losses in body weight of ≥20% were promptly euthanized per Institutional Animal Care and Use Committee policy.

Total cellular RNA was isolated from two randomly selected tumors from control, pictilisib, cisplatin, and combination groups. Tumors were regarded as resistant to treatment when the tumor growth curve reached logarithmic growth phase. RNA samples were submitted to the UC Davis Comprehensive Cancer Center's Genomics Shared Resource for RNA sequencing analysis using a “total” RNA-Sequencing approach. The protocol for RNA-sequencing and data analysis were described previously (5). The raw sequence and processed data files are publicly available through the NCBI GEO database (https://www.ncbi.nlm.nih.gov/geo) with accession number GSE101419.

At least three independent experiments were performed for each analysis described in this article. Student t test or one-way ANOVA was used to compare continuous parametric data between two groups and multiple groups, respectively. Overall survival of mice was analyzed using Kaplan–Meier survival curve and log-rank test. Statistical analysis was performed by GraphPad Prism 5 software.

We analyzed the genetic alterations of the PI3K/AKT signaling pathway in five common bladder cancer cell lines by searching the Catalogue of Somatic Mutations in Cancer (COSMIC, http://cancer.sanger.ac.uk/cell_lines) and literature review (25–27). As shown in Supplementary Table S1, TCCSUP harbors a hotspot PIK3CA mutation E545K; T24 has a rare PTEN mutation N48I; and J82 contains both a rare PIK3CA mutation P124L and homozygous deletion of PTEN. The expression and activation status of key proteins related to the PI3K/AKT signaling pathway were determined (Fig. 1A). Consistent with the genetic mutation status, basal levels of AKT phosphorylation were higher in TCCSUP and lack of PTEN protein were observed in J82. Interestingly, AKT phosphorylation was not detected in RT4, which was reported in previous studies (28, 29).

The small-molecule pan-PI3K inhibitor pictilisib was applied to test the cytotoxicity in bladder cancer cell lines (Fig. 1B; Supplementary Table S1). Consistent with the finding that TCCSUP contained an activating mutation, this cell line was the most sensitive to pictilisib treatment with an IC₅₀ of 1.1 μmol/L. The other four cell lines were less sensitive with IC₅₀ all greater than 20 μmol/L.

To further characterize how pictilisib affected downstream effectors of PI3K, we examined AKT phosphorylation in pictilisib-sensitive (TCCSUP) and resistant (J82 and T24) cell lines. As shown in Fig. 1C and D, AKT phosphorylation was decreased by pictilisib in a time- and dose-dependent manner, not only in sensitive TCCSUP cells, but also in resistant T24 and J82 cells. When these cells were treated with pictilisib at a concentration of 1μmol/L, AKT phosphorylation reached a nadir after 30 minutes of treatment with complete disappearance in the most sensitive TCCSUP cells, while measurable levels of AKT phosphorylation persisted in resistant J82 and T24 cells. These data suggested that the PI3K/AKT pathway was not completely suppressed in the two resistant cell lines. AKT phosphorylation gradually rebounded to the baseline level around 4 hours (Fig. 1C), and the rebound was not due to inactivation of pictilisib in culture medium (Supplementary Fig. S1C). The level of AKT phosphorylation also decreased with increasing concentrations of pictilisib (Fig. 1D).

Apoptosis and cell-cycle analysis were performed to evaluate the mechanism for the antiproliferative effect of pictilisib on drug-sensitive TCCSUP and drug-resistant T24 cell lines. TCCSUP showed a modest (<3%) but significantly dose-dependent increase in apoptosis (Supplementary Fig. S1B), and more dramatic (>13%) S-phase arrest (Supplementary Fig. S1A) after pictilisib treatment. In contrast, no significant effect was observed in T24 cells (Supplementary Fig. S1).

Next, we determined whether the cytotoxic effect of pictilisib could be translated into in vivo antitumor activity in bladder cancer PDXs. If so, pictilisib can potentially be developed for treatment of advanced bladder cancers that have alterations in the PI3K pathway. For this experiment, we compared the efficacy of pictilisib between PDX BL0269 and BL0293. BL0269 harbors a PIK3CA hotspot mutation (H1047R) and BL0293 harbors a rare and non-hotspot mutation (D549Y; ref. 5). Pictilisib significantly inhibited tumor growth in BL0269 (Fig. 2A). The median time to achieve a tumor volume of 10 times the baseline increased from 15.5 days in control to 27 days with pictilisib (P < 0.01). Compared with the median survival of 18.5 days for the control, pictilisib treatment prolonged overall survival to 33.5 days (P < 0.01; Fig. 2B). Tumors obtained at day 3 revealed that pictilisib treatment substantially reduced the level of p-AKT (Fig. 2C) with diminished Ki-67 (Supplementary Fig. S2A), and increased levels of cleaved caspase-3 (Supplementary Fig. S2B). In contrast, tumor growth was only slightly retarded by pictilisib in BL0293 (Supplementary Fig. S3G) with the median time to achieve a tumor volume of 7.5 times the baseline was 19 days and 24 days in control and pictilisib treatment groups, respectively (P = 0.630).

We evaluated whether pictilisib could potentiate the cytotoxicity and antitumor activity of cisplatin. First, we used the CI method to determine the drug–drug interaction between pictilisib and cisplatin in vitro. Cells were treated with various concentrations of cisplatin or with a fixed concentration of pictilisib (0.5, 1, or 2 μmol/L), which corresponded to known patient plasma levels (maximum plasma concentration = 2 μmol/L; ref. 16). Compared with treatment with cisplatin alone, pictilisib enhanced the cytotoxicity of cisplatin, especially at a lower dose of cisplatin (<1μmol/L, Fig. 2E). The CI values of the two drugs suggested a synergistic effect (CI < 0.9; Supplementary Table S2). Furthermore, we found that pictilisib could reduce the upregulation of p-AKT induced by cisplatin treatment at 24 hours and decrease the expression of the downstream effector p-S6 in TCCSUP cells (Fig. 2F).

Next, we determined whether the synergistic effect of pictilisib and cisplatin could be translated to in vivo antitumor activity with PDXs. We compared the antitumor efficacy of single-agent pictilisib, cisplatin, concurrent combination and sequential treatment with pictilisib and cisplatin in NSG mice bearing the PDX BL0269. Combination and sequential treatment groups significantly delayed the tumor growth and prolonged lifespan compared with single drug treatment with pictilisib or cisplatin (P < 0.01, Fig. 2A and B; Supplementary Fig. S3A and S3B), while there was no significant difference in overall survival between the combination and sequential treatment groups (P = 0.156; Fig. 2B; Supplementary Fig. S3B). Furthermore, we revealed that pictilisib combined with gemcitabine and cisplatin was also more effective than the gemcitabine and cisplatin regimen in inhibiting the tumor growth of BL0269 (Supplementary Fig. S3I). Body weight was slightly decreased in all treatment groups (P < 0.01 at day 20), while there was no significant difference between single-drug treatment groups and combination drug treatment groups (P = 0.053 at day 20 Supplementary Fig. S3). When tumors became resistant to any treatment, we observed upregulation of p-AKT, p-ERK (Fig. 2D), and Ki-67 (Supplementary Fig. S2A) and downregulation of cleaved caspase-3 (Supplementary Fig. S2B). These data suggested that alternative oncogenic pathways were activated to compensate for the PI3K/AKT pathway.

Gemcitabine is another drug used in first-line chemotherapy for bladder cancer, specifically used in combination with platinum agents (gemcitabine plus cisplatin/carboplatin; ref. 3). We thus investigated whether gemcitabine and pictilisib had a similar synergistic drug–drug interaction. First, we found a significant left shift of the dose–response curve with combination of pictilisib and gemcitabine in TCCSUP cells compared with gemcitabine alone (Fig. 3A). The CI values of gemcitabine and pictilisib in all treatment groups were 0.20 or less, suggesting a very strong synergistic cytotoxic effect between pictilisib and gemcitabine (Supplementary Table S2). Furthermore, we revealed that pictilisib could inhibit the upregulation of p-AKT induced by gemcitabine treatment at 24 hours and decrease the levels of p-S6 in TCCSUP cells (Fig. 2F).

Next, we determined whether the synergistic cytotoxic effect of pictilisib and gemcitabine could be translated to in vivo antitumor effect. For this experiment, we used PDX BL0440, which had PIK3CA copy number amplification and was partially resistant to gemcitabine chemotherapy. We found the in vivo results correlated well with in vitro efficacy of pictilisib and gemcitabine. As shown in Fig. 3B and Supplementary Fig. S3D, the median time of the tumor growth to ten times the baseline increased from 18.5 days for the control to 33.5 days (P < 0.01) with pictilisib, 47 days (P < 0.01) with gemcitabine, and 66.5 days (P < 0.01) with the combination group. The median survival of the combination group (77 days) was significantly longer than pictilisib (48.5 days) or gemcitabine (57 days; P < 0.01, Fig. 3C; Supplementary Fig. S3E). No significant difference (P = 0.277) was found between pictilisib (median = 48.5 days) and gemcitabine (median = 57 days) single-drug treatment groups. Body weight was slightly reduced in all treatment groups (P < 0.01 at day 30), while no significant difference was observed between single drug groups and the combination treatment groups (P = 0.327 at day 30; Supplementary Fig. S3). Western blot analysis showed that p-AKT was significantly suppressed in the pictilisib and combination groups, but not in the gemcitabine group (Fig. 3D). As pictilisib has no direct effect on the MEK/ERK pathway, no significant decrease of p-ERK was observed. In fact, there might be slight increase of p-ERK after pictilisib treatment, possibly secondary to positive feedback of the MEK/ERK pathway after inhibition of the PI3K/AKT pathway. IHC staining of tumors collected at day 3 revealed that Ki-67 (Supplementary Fig. S4A) was significantly reduced, and cleaved caspase-3 was substantially increased in treatment groups (Supplementary Fig. S4B). When tumors became resistant to different treatment drugs, the levels of p-AKT, Ki-67 and cleaved caspase-3 returned to baseline levels (Fig. 3D; Supplementary Fig. S4). These results suggested that pictilisib was not only effective in bladder cancer with PIK3CA activation mutation, but also with PIK3CA amplification, and it could potentiate the antitumor efficacy of gemcitabine.

Signaling transduction is a complicated network associated with positive and negative feedback loops, and each component of the pathways is regulated by many factors (Fig. 4A). For example, the MEK/ERK pathway was suggested to be activated by PI3K/AKT pathway inhibition through PI3K-independent or PI3K-dependent feedback loop, causing drug resistance to the PI3K inhibitor (30, 31). In this study, we evaluated several key components of the PI3K/AKT pathway (p110α, p-P70S6K, p-S6) and MEK/ERK pathway (p-ERK) to determine which component correlates with drug response and may be useful as a biomarker for guiding pictilisib treatment. After exposure to 1 μmol/L pictilisib for 1 hour, p-ERK expression increased in TCCSUP cells; p-P70S6K was downregulated in both TCCSUP and J82 cells; and only p-S6 downregulation was observed in the pictilisib-sensitive TCCSUP cells (Fig. 4B). These results prompted us to determine whether p-S6 could potentially serve as a biomarker to predict the response of pictilisib. As shown in Supplementary Fig. S5, p-P70S6K levels in TCCSUP and J82 cells were modulated in a time-dependent fashion after pictilisib treatment, as was AKT phosphorylation (Fig. 1C). The phosphorylation of S6 was reduced by pictilisib in a time- and dose-dependent manner in TCCSUP cells, which was consistent with its response to pictilisib. Although the level of p-P70S6K started to rebound after approximately 2 hours in TCCSUP cells, its downstream effector p-S6 could be suppressed for 8 hours without rebound. These data strongly suggested that p-S6 might be a potential biomarker of pictilisib efficacy. Next, we investigated the expression of these key regulators of the PI3K/AKT pathway in the BL0269 and BL0440 PDX tumors. In line with the in vitro findings, we found a rebound of p-S6 in pictilisib-resistant tumors of both PDX models, which further demonstrated that p-S6 levels correlated well, and could potentially serve as a biomarker to predict response to pictilisib (Fig. 4C and D).

It was previously reported that the anticancer activity of PI3K/AKT inhibitors could be countered by MEK/ERK pathway activation (30, 31). Here we showed that in pictilisib-sensitive TCCSUP, p-ERK was upregulated in a time-dependent, but not dose-dependent manner (Fig. 5A and B), suggesting that even at a concentration below the IC₅₀ (1.10 μmol/L), the MEK/ERK pathway was already activated. We did not observe any significant change of p-ERK in the resistant J82 and T24 cells (Supplementary Fig. S5C and S5D). As both pathways are important for many cellular functions, we hypothesized that simultaneous blockade of the PI3K/AKT and MEK/ERK pathways by pictilisib and sorafenib, respectively, might achieve improved or even synergistic antitumor activity. Sorafenib is effective in suppressing Raf/MEK/ERK pathway activity, and has already been used for the treatment of advanced hepatocellular cancer and renal cell cancer (32). Sorafenib alone was moderately effective with an IC₅₀ of 9.4 μmol/L in TCCSUP (Fig. 5C). A synergistic to additive effect of pictilisib and sorafenib was observed at a variety of concentrations in TCCSUP (Supplementary Table S2), and the activation of p-ERK induced by pictilisib could be inhibited when cells were treated with the combination of pictilisib and sorafenib (Fig. 5D).

To explore the mechanisms potentially underlying the development of drug resistance to pictilisib in bladder cancer, we performed RNA-sequencing analysis to comprehensively define the transcriptomic changes associated with the drug-resistant PDX BL0269 tumor model. For this, gene-level expression of 21,597 annotated, protein-coding transcripts was quantified for each sample and differentially expressed genes (i.e., relative to the control) were subsequently determined for each group of tumors that developed resistance to pictilisib, cisplatin, or the combination treatment. Criteria for filtering of the results included a cut-off for differential expression of an absolute log2 fold change of ≥1 (i.e., ≥2-fold change) and a minimum threshold for an elevated level of expression of per kilobase of transcript per million mapped (FPKM) >1 (equivalent to one transcript per cell; refs. 33, 34). Hierarchical clustering was performed on the resulting gene sets for the pictilisib-, cisplatin-, and combination treatment–resistant tumors, and the expression patterns visualized as heatmaps (Fig. 6A). Intersection analysis was utilized to identify common resistance-associated expression changes, and as shown in the Venn diagram (Fig. 6B), the expression of three genes, lymphocyte-specific protein 1 (LSP1), peptidase inhibitor 3 (PI3), small nucleolar RNA H/ACA box 24 (SNORA24) were altered in all three groups, and had moderate to high levels of expression (e.g., 11–519 FPKM; Fig. 6C). Compared with the control group, although the alteration of PI3 was not consistent, downregulation of LSP1 and upregulation of SNORA24 were observed in all drug-resistant groups. SNORA24 was reported to participate in the synthesis of pseudouridine residues in the human 18S and 28S rRNAs and the spliceosomal small nuclear RNAs (35). However, whether SNORA24 plays a role in carcinogenesis or drug metabolism is not yet known. LSP1 acts as tumor suppressor gene (36). Mutation or single nucleotide polymorphisms of LSP1 have been found to be associated with increased risk of breast cancer and hepatocellular carcinoma. Loss of LSP1 function may remove suppressing effects on the ERK/MAPK pathway, leading to increased proliferation and migration of tumor cells (36, 37). In fact, we observed that p-ERK expression was increased when tumors become resistant to drug treatment in the PDX BL0269 model (Fig. 2C). Consistent with the findings for LSP1 transcript levels (i.e., RNA-seq), Western blotting further confirmed the downregulation of LSP1 protein in resistant tumor tissues (Fig. 6D).

Standard treatment with cisplatin-based chemotherapy for patients with advanced bladder cancer has reached an efficacy plateau with very little improvement in survival over the last few decades (2). Gaining insight into the biology of bladder cancer reveals new “druggable” targets, which allow the development of molecularly guided targeted therapy. Whole-exome sequencing of 131 high-grade muscle-invasive urothelial bladder carcinomas identified that alterations in the PI3K/AKT pathway occurred in 42% of tumors, and therefore presents a promising therapeutic target (4). Drugs targeting PI3K may avoid the activation of negative feedback loops of PI3K/AKT pathway and suppress this pathway robustly enough to elicit a durable response (31).

This study was initiated to determine the feasibility of targeting the PI3K pathway to kill bladder cancer cells and to potentiate the antitumor activity of cisplatin and gemcitabine. Pictilisib was chosen to show the proof of principle in this study merely because it has already reached the phase II clinical trial stage (17, 18, 38). We believe other PI3K inhibitors could possibly achieve similar effects as we previously showed with another PI3K inhibitor BEZ235 (5). The growth of five bladder cancer cell lines could be inhibited by pictilisib to some extent, but only PIK3CA E545K–mutant TCCSUP cells had an IC₅₀ as low as 1 μmol/L. These observations are consistent with previous studies that suggested sensitivity to PI3K inhibitors was dependent on hotspot PIK3CA mutation status (9, 10, 29). We also revealed that AKT phosphorylation in TCCSUP cells could be inhibited by pictilisib to a nadir after 30 minutes, while remaining measurable in resistant J82 and T24 cells. It was reported that AKT could be activated by both PI3K-dependent and by the PI3K-independent signaling, such as serine/threonine (TBK1, KBKE) and tyrosine (Src, PTK6) kinases (39, 40). We speculated that PI3K might not serve as the major regulator of AKT activity in J82 and T24 cells. Meanwhile the rapid kinetics of AKT phosphorylation suggested that alternate AKT activators also need to be considered as drug targets. We further demonstrated that pictilisib effectively inhibited tumor growth and prolonged lifespan of mice bearing PDX tumors harboring the PIK3CA hotspot mutation (H1047R) or PIK3CA amplification. PIK3CA, which encodes the p110α catalytic isoform of class I PI3K, and is one of the most commonly mutated or amplified kinases in a variety of tumors (7). By collecting data from public databases (http://www.cbioportal.org/), we identified that the frequency of mutation and amplification of PIK3CA in bladder cancer ranged from 15.2% to 26.0%. Of PI3KCA mutations in bladder cancer, E545K (41.0%) and E542K (18.7%) mutations in the helical domain are the most common. In contrast, mutation at H1047R (2.2%) in the kinase domain is more common in other cancers, but is less so in bladder cancer (Supplementary Table S1; ref. 41). Our study provided the first preclinical evidence indicating patients with bladder cancer with PIK3CA mutations (E545K and H1047R) or amplification may be appropriate for future clinical trials with a PI3K inhibitor.

The activation of PI3K/AKT pathway was reported to contribute to cisplatin and gemcitabine resistance, both of which are currently the frontline choice for advanced bladder cancer (15, 42, 43). Arjumand and colleagues (15) suggested that cervical cancer cells engineered with a PIK3CA E545K mutation were more resistant to cisplatin. Previously, we observed that PDX BL0269 with a PIK3CA-H1047R mutation showed resistance to both cisplatin and gemcitabine (5). To our knowledge, combined pictilisib plus cisplatin and/or gemcitabine as a potential therapeutic approach in bladder cancer has not previously been investigated. We found pictilisib combined with cisplatin or gemcitabine resulted in an enhanced effect in decreasing proliferation of bladder cancer in vitro and in vivo. Pandey and colleagues (44) suggested that sequential administration of a PI3K/AKT inhibitor following treatment with cisplatin resulted in enhanced antitumor efficacy for breast cancer. We found that sequential treatment with pictilisib and cisplatin is as effective as the concurrent combination treatment. Given the fact that as high as 50% of patients with bladder cancer are resistant to cisplatin and gemcitabine chemotherapy (45), our concurrent combination experiment of pictilisib and cisplatin or gemcitabine may have high clinically relevance.

The OPPORTUNE clinical trial combining pictilisib and anastrozole showed significantly increased antitumor response in patients with early-stage breast cancer than anastrozole alone (38). In contrast, the phase II PEGGY clinical trial, Vuylsteke and colleagues (18) reported that patients with hormone receptor–positive, HER2-negative locally recurrent or metastatic breast cancer did not get significant benefit from pictilisib plus paclitaxel treatment compared with paclitaxel alone. In the phase II clinical trial FERGI, treatment with pictilisib alone also failed to increase progression-free survival in locally advanced or metastatic breast cancer compared with placebo, and PIK3CA mutation status was found to have no effect on predicting the benefit of pictilisib (17). Although preclinical studies have found that mutations or amplification in PIK3CA correlated with pictilisib sensitivity, clinical trials reported little or no association between PIK3CA mutation status and outcome (9, 17, 18). Currently, a major challenge is to identify biomarkers for predicting pictilisib response in cancer treatment. Prediction of sensitivity may require more complex signatures rather than single mutational event (7). Our results indicated that treatment-induced downregulation of p-S6 correlated with the response to pictilisib in vitro and in vivo, and rebound of p-S6 was observed when tumors became resistant to pictilisib in PDX models. Yang and colleagues (46) and Brien and colleagues (9) also suggested that p-S6 could be used as a pharmacodynamic biomarker to predict PI3K inhibitor response. As a result, the alteration of p-S6 after pictilisib treatment may be a candidate biomarker for evaluating pictilisib response, but needs to be validated in clinical trials.

We performed RNA sequencing and Western blot, and found that LSP1 was downregulated in all drug-resistant groups. LSP1 was reported to interact with kinase suppressor of Ras, which functions as a scaffold for MAPK and Raf kinases and is a key regulator of cellular growth (37). Loss of LSP1 function removes ERK/MAPK pathway suppression, leading to aberrant proliferation or migration (36). In line with this mechanism, we observed that the expression of p-ERK was increased in all treatment resistant groups of the PDX BL0269 model. Similarly, Carracedo and colleagues (30) reported that patients with solid tumors and treated with an mTOR inhibitor everolimus showed a marked increase in ERK activation. Rozengurt and colleagues (31) suggested that PI3K/mTOR inhibitors relieved a negative feedback on receptor tyrosine kinases that leads to RAF/MEK/ERK activation. We observed that TCCSUP cells with the PIK3CA E545K mutation showed a time-dependent increase of p-ERK after pictilisib treatment, and inhibition of an upstream factor by a RAF inhibitor sorafenib could synergize with pictilisib to inhibit tumor cell growth and abrogate the upregulation of p-ERK by pictilisib. These data implicate the importance of rational combinations of therapeutic agents to overcome drug resistance that was induced by compensatory activation of other pathways.

In summary, we demonstrated for the first time that pictilisib effectively enhanced the antitumor effect of cisplatin and gemcitabine in human bladder cancer in vitro and in vivo. Our results also revealed that the expression of p-S6 is a potential candidate biomarker for evaluating pictilisib response. Furthermore, our preclinical data provided good insight for combination of pictilisib and chemotherapeutic agents for bladder cancer treatment in further clinical trials.

No potential conflicts of interest were disclosed.

The contents do not represent the views of the U.S. Department of Veterans Affairs or the United States Government.

Conception and design: S.-X. Zeng, Y. Zhu, A.-H. Ma, J.-M. Guo, C.-L. Xu, R.W. deVere White, C.-X. Pan

Development of methodology: S.-X. Zeng, Y. Zhu, A.-H. Ma, W. Shi, C.-X. Pan

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): S.-X. Zeng, Y. Zhu, W. Yu, H. Zhang, T.-Y. Lin, S. Airhart, C.-X. Pan

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): S.-X. Zeng, Y. Zhu, A.-H. Ma, W. Yu, T.-Y. Lin, C.G. Tepper, P.T. Henderson, J.-M. Guo, C.-L. Xu, C.-X. Pan

Writing, review, and/or revision of the manuscript: S.-X. Zeng, Y. Zhu, A.-H. Ma, H. Zhang, T.-Y. Lin, C.G. Tepper, P.T. Henderson, S. Airhart, J.-M. Guo, C.-L. Xu, R.W. deVere White, C.-X. Pan

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): S.-X. Zeng, Y. Zhu, A.-H. Ma, W. Yu, H. Zhang, C.G. Tepper, C.-X. Pan

Study supervision: T.-Y. Lin, J.-M. Guo, C.-X. Pan

We thank Qilai Long and Daniel Zhu for their assistance in conducting experiments. We are also grateful to Stephenie Y. Liu and Ryan R. Davis for their expert technical assistance in performance of the RNA-sequencing analyses.

The work was supported in part by Merit Review (Award # I01 BX001784; to C.-X. Pan) from the United States (U.S.) Department of Veterans Affairs Biomedical Laboratory Research and Development Program, NCI Cancer Center Support Grant (PI: R.W. deVere White; grant no.: 2 P30 CA 0933730), and The Laney Foundation (principal investigator: R.W. deVere White).

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.