Background
Advanced urothelial carcinoma (UC) of the bladder has a poor prognosis. At presentation, 15–30 % of UC patients are diagnosed with muscle-invasive tumors, and these carry a 5-year risk of death ranging from 33 to 73 % [
1]. The standard treatment for localized invasive UC, is surgical removal of the bladder and regional lymph nodes, but metastatic disease is a major cause of death in these patients. The addition of cisplatin-containing combination neoadjuvant therapy has been shown to improve outcomes following cystectomy [
2,
3], but metastasis remains common and although treatment with cisplatin-containing chemotherapy is beneficial in some cases, median survival for metastatic UC is only 13–15 months. As no significant improvements in survival have been achieved in recent years, new approaches to therapy, particularly second line therapies for metastatic disease, are urgently needed. Detailed molecular information on UC is now available [
4,
5], but targeted agents have not yet been widely applied [
6].
The phosphatidylinositol 3-kinase (PI3K) signaling pathway plays a critical role in regulation of cell metabolism, migration, proliferation and survival [
7] and mutations that lead to aberrant activation of the pathway are found in virtually all types of cancer. In bladder cancer, 50–70 % of tumors contain mutations that are predicted to activate this pathway. These include activating mutations in
PIK3CA, [
8,
9] and
AKT1 [
10], and inactivating mutations of
PTEN [
11,
12],
PIK3R1 [
13],
TSC1 and
TSC2 [
9,
14]. Assessment of the phosphorylation status of key pathway proteins confirms that pathway activation is present in bladder tumors of all grades and stages [
15]. These tumors may benefit from PI3K-targeted therapy. Clinical trials of mTORC1 inhibitors in patients with bladder cancer have been initiated in recent years. In trials of the mTOR inhibitor Evirolimus, exceptional responses have been reported in patients with advanced UC whose tumors contained
TSC1 or
mTOR mutations [
16,
17]. In general however, responses to mTOR inhibitors have not been impressive [
18], and indeed not all UC patients with tumors containing
TSC1 mutations have shown responses [
16]. A potential reason is that mTOR inhibition triggers feedback loops that activate AKT [
19]. Inhibitors of AKT have therefore been examined in preclinical studies of UC [
20,
21]. Importantly, these studies revealed that sensitivity to AKT inhibition was strongly related to the presence of
PIK3CA mutation. Taken together, it is clear that a thorough understanding of the signaling events initiated by the PI3K pathway is required in order to maximize clinical benefit.
Inhibition of PI3K as a potential therapeutic approach in UC has not previously been examined, though mutations in
PIK3CA represent the most frequent PI3K pathway mutations in this cancer type, including 12–20 % of muscle-invasive tumors [
14,
22]. Preclinical studies and early clinical trials indicate sensitivity to inhibitors of PI3K in several cancers including breast, ovarian, endometrial, lung and multiple myeloma [
18,
23‐
29]. The majority of these studies highlight the Class 1 PI3K inhibitor, GDC-0941, as a good therapeutic drug for solid tumors. Furthermore, a phase I dose-escalation study of GDC-0941 has recently been completed and reports good tolerability of the drug with confirmed target modulation in tumor tissues [
30]. Several studies in non-bladder cell lines have sought predictive biomarkers of sensitivity to PI3K inhibitors and it has been suggested that mutation of
PIK3CA or loss of PTEN function are related to sensitivity to inhibitors of class I PI3K and that mutations in RAS genes are associated with resistance (Reviewed in [
31]), though prediction based on these biomarkers is not absolute.
Previously we examined the effect of ectopic expression of mutant PIK3CA in telomerase-immortalized normal human urothelial cells (TERT-NHUC) and showed that this induces cell proliferation and migration [
32]. In bladder tumors, more than one lesion in the PI3K pathway is commonly present [
9] and this could potentially lead to distinct types of pathway dependence and response to specific therapeutic agents. Therefore, we have examined the consequences of specific inhibition of mutant PIK3CA in UC cells using stable knockdown, and treatment of a panel of UC cell lines containing a range of PI3K pathway alterations with the class I PI3K inhibitor, GDC-0941. Our findings strongly suggest that targeting of PIK3CA maybe a valid therapeutic approach in advanced bladder cancer.
Discussion
Activating mutations of
PIK3CA are found in bladder tumors of all grades and stages. Whilst these are more common in tumors of low grade and stage (26–34 %) they are also found at significant frequency (12–20 %) in advanced UC (
> stage T2) [
8,
9,
22] (Hurst, Platt, and Knowles, unpublished data), which are commonly treated with systemic therapies and for which novel therapeutic approaches are urgently needed. Our objective was to examine the effects of specific inhibition of mutant PIK3CA in bladder tumor cells to determine whether mutant PIK3CA can be considered a valid therapeutic target in bladder cancer.
Stable knockdown of mutant E545K/E545G PIK3CA in three UC cell lines reduced PIK3CA protein levels by up to 92 % and was associated with reduced AKT activation, proliferation and
in vivo tumor growth. This is consistent with observations reported in colon, gastric and ovarian cancer cell types [
45‐
47], and with our previous data on the effects of expression of mutant PIK3CA in TERT-NHUC [
32] and in other cell types [
48‐
50]. Importantly, the intensity of reduction of these phenotypes was linked to the level of PIK3CA protein knockdown and the related reduction in AKT phosphorylation. We only observed inhibition of directional migration in VM-CUB-3 cells, which suggests that additional molecular alterations may contribute to this phenotype. It is possible that invasion may also be affected by PIK3CA and further investigation is warranted to investigate this. It is also important to note that the mice used for the
in vivo work were immunodeficient. Whilst the
in vivo results show a clear effect, and using immunodeficient mice allows the use of human tumor cells, there is a caveat in that interfering with the PI3K pathway can influence the immune system, and examination in a syngeneic model system could be relevant.
The class IA PI3K inhibitor GDC-0941 greatly reduced cell viability of UC cell lines with hotspot
PIK3CA mutation status, similar to findings in other preclinical models [
25‐
27,
41,
51]. The two cell lines with rare
PIK3CA mutations had minimal response to the drug and contained either homozygous deletion of
PTEN (J82) or mutation of
TSC1 (639V), suggesting that co-existing
PTEN or
TSC1 mutations confer some resistance to PI3K inhibition. Cell lines with wildtype
PIK3CA were less sensitive or lacked sensitivity to GDC-0941 completely, in agreement with studies in breast cancer, multiple myeloma, lung cancer, and endometrial cancer, where
PIK3CA mutations are also frequent [
25,
26,
51]. Interestingly, our data support the findings of a study that examined the effect of the pan-AKT inhibitor, MK-2206, on the viability of UC cell lines [
20]. Hotspot mutant PIK3CA expressing cell lines, 253J, HT1197 and VMCUB-1 were sensitive to AKT inhibition, whereas cell lines with rare
PIK3CA mutations (J82 and 639V) or those with
PTEN and RAS gene or
TSC1 mutations (UMUC3 and RT4) had IC
50 values above 2 μM (classed as resistant).
As reported in several other tumor types, additional information on PI3K pathway mutational status is needed to correctly predict the response to PI3K inhibition. There have been conflicting reports of the effect of
PTEN mutation on sensitivity to class I PI3K inhibition. For example, in non-small cell lung carcinoma, endometrioid and breast carcinoma cell lines, PTEN loss of expression or mutation was associated with sensitivity to GDC-0941 [
25,
26,
51] but in multiple myeloma cell lines [
27], PTEN loss had no predictive value. In the present study, all UC lines with PTEN loss as a single alteration showed resistance. PI3K pathway-dependent cells with mutations in genes that act below PIK3CA in the pathway are predicted to be resistant to PI3K inhibition. Indeed, our data showed that
PTEN, AKT1 and
TSC1 mutant UC cell lines were less sensitive than those with
PIK3CA mutations, and in the case for RT4 (
TSC1 mutant), resistant to PI3K inhibition.
Similarly, in some studies RAS mutation has been reported to predict resistance to PI3K inhibition and in others to have no impact on sensitivity [
25,
52]. Here we found that HT-1197, which harbors both
PIK3CA and
NRAS mutations, was sensitive to GDC-0941 but other lines with RAS mutations co-occurring with
PTEN or
AKT1 mutations were less sensitive. This is likely to be related to relative dependence on RAS-MAPK and PI3K pathways in individual cases. Recent observations have highlighted the effect of context-dependent crosstalk on MEK signaling associated with inhibition of PI3K in breast cancer [
53,
54]. Interestingly, a phase I clinical trial of GDC-0941 in solid cancers showed that a melanoma patient with good response had a BRAF mutation and wild-type
PIK3CA [
30]. As many bladder cancers have alterations that are known to activate the MAPK pathway [
4], it is likely that dual MEK and PI3K inhibition may be advantageous. Results from ongoing clinical trials of GDC-0941 in combination with drugs including GDC-0973 (MEK inhibitor), erlotinib (EGFR inhibitor), fulvestrant, and cisplatin are awaited.
In this study, shRNA knockdown of PIK3CA in three cell lines clearly demonstrated dependence on this p110 isoform. However, as many of the cell lines examined expressed PIK3CB, we cannot rule out that the effects of GDC-0941 on PIK3CB may have contributed to the observed effects of this inhibitor. As p110α-specific drugs e.g. NVP-BYL719 [
55], are now in clinical trials, it will be important to assess the relative dependence of each isoform in UC, prior to consideration of potential clinical studies.
Levels of pAKT (Ser473) were decreased by GDC-0941 treatment, independent of whether a cell line was sensitive or resistant, as previously reported [
26,
27,
56]. Resistant cell lines may not be dependent on AKT signaling or may have additional mechanisms to activate the PI3K pathway. Interestingly, RT4, the most resistant cell line in this study, had an undetectable basal level of AKT activation compared to the other cell lines studied. As this cell line has mutant
TSC1, this implies that only the mTOR branch of the pathway is active in this case, as shown previously [
20]. As many
PIK3CA mutant cancers rely on effectors other than AKT, such as PDK1 and its substrate SGK3 [
57], phospho-AKT may not be an ideal pharmacodynamic biomarker for relevant PI3K inhibition.
The effect of GDC-0941 in the majority of sensitive bladder cancer cells was via induction of both G1 cell cycle arrest and some apoptosis. Thus in some contexts, single agent GDC-0941 may exert cytotoxic and cytostatic effects in UC as seen in other tumor models [
26,
27,
58]. Dual inhibition of mTOR and PI3K has been shown to be a promising approach in cell lines from other cancer types [
29] and may be particularly efficacious in bladder where molecular lesions are found in multiple genes in the PI3K pathway, often concurrently [
4].