|Home | About | Journals | Submit | Contact Us | Français|
PI3K is an important therapeutic target. Mutations in PIK3CA, which encodes p110α, the catalytic subunit of PI3K, occur in endometrioid and non-endometrioid endometrial cancers (EECs and NEECs). The goal of this study was to determine whether PIK3R1, which encodes p85α, the inhibitory subunit of PI3K, is mutated in endometrial carcinoma. We performed exonic sequencing of PIK3R1 from 42 EECs and 66 NEECs. The pattern of PIK3R1 mutations was compared to the patterns of PIK3CA, PTEN and KRAS mutations. The biochemical effect of seven PIK3R1 mutations was examined by stable expression in U2OS cells, followed by coimmunoprecipitation analysis of p110α, and Western blotting of phospho-AKTSer473. We found that PIK3R1 was somatically mutated in 43% of EECs and 12% of NEECs. The majority of mutations (93.3%) localized to the p85α-nSH2 and -iSH2 domains. Several mutations were recurrent. PIK3R1 mutations were significantly (P=0.0015) more frequent in PIK3CA-wild type EECs (70%) than in PIK3CA-mutant EECs (18%). Introduction of wild type p85α into U2OS cells reduced the level of phospho-AKTSer473 compared to the vector control. Five p85α mutants, p85αdelH450-E451, p85αdelK459, p85αdelY463-L466, p85αdelR574-T576, and the p85αN564D positive control, were shown to bind p110α and led to increased levels of p-AKTSer473. The p85αR348X and p85αK511VfsX2 mutants did not bind p110α and showed no appreciable change in p-AKTSer473 levels. In conclusion, our study has revealed a new mode of PI3K alteration in primary endometrial tumors and warrants future studies to determine whether PIK3R1 mutations correlate with clinical outcome to targeted therapies directed against the PI3K pathway in EEC and NEEC.
Endometrial cancer kills approximately 74,000 women worldwide each year (1). Tumors are classified into two major subtypes, endometrioid endometrial cancers (EECs) and non-endometrioid endometrial cancers (NEECs) (2). At diagnosis, the vast majority of endometrial tumors are EECs. Although many EECs are detected at an early stage and can be treated effectively with surgery, improved therapeutic strategies are needed for the treatment of recurrent and advanced stage EEC (3, 4). NEECs represent a minority of tumors at presentation (4), but they are the most clinically aggressive subtype and cause a disproportionate fraction of all endometrial cancer related deaths (5). Therefore new therapeutic approaches to treat NEEC are also needed.
The PI3K signal transduction pathway represents an important therapeutic target (6). PI3K is a heterodimer comprised of a catalytic subunit (p110α) encoded by PIK3CA, and a regulatory subunit (p85α) encoded by PIK3R1. In quiescent cells, p85α binds to p110α and causes both stabilization and catalytic inhibition of p110α. Somatic mutations in PIK3CA occur in many tumor types, including endometrial cancer (7, 8), whereas somatic PIK3R1 mutations are restricted to a few tumor types (9–12).
We recently showed that the ABD and C2 domains of p110α, which mediate binding to p85α, are frequently mutated in endometrial carcinomas (13). We therefore hypothesized that PIK3R1 (p85α) itself might be mutated in endometrial tumors. Herein, we report that PIK3R1 is somatically mutated in 43% of EECs and 12% of NEECs. Mutations preferentially localized to the p85α-iSH2 domain, which mediates binding to p110α. Several PIK3R1 mutations promoted increased phosphorylation of AKTSer473. Collectively, our findings reveal a new mechanism by which the PI3K pathway is activated in endometrial cancer.
Primary tumor (42 EECs and 66 NEECs) and matched normal tissues were collected at resection, prior to treatment, and obtained with appropriate IRB approval (13). A pathologist reviewed H&E sections of tumors to verify histology and delineate regions of tissue comprised of >70% tumor cells for macrodissection.
Genomic DNA was isolated from macrodissected tumor tissue or normal tissue using the PUREGENE kit (Gentra Systems). Matched tumor and normal DNAs were genotyped using the Coriell identity testing kit (Coriell).
All coding exons of PIK3R1 were amplified from tumor DNA, using the polymerase chain reaction (PCR), followed by nucleotide sequencing (see Supplementary Methods). Purified tumor cell populations were isolated from three tumors using laser capture microdissection (LCM), followed by RT-PCR and sequencing to determine whether there was monoallelic or biallelic expression of mutations (See Supplementary Methods).
A retroviral expression construct containing full-length, wild type, PIK3R1 cDNA in the pBABE vector (Addgene) was used to generate a series of PIK3R1 mutant constructs by site directed mutagenesis using the QuikChange II XL Site-Directed Mutagenesis Kit (Stratagene). Inserts were excised using BamHI and SalI and subcloned into the MYC-tagged pCMV-3Tag-7 expression vector (Agilent Technologies). The integrity of inserts was confirmed by Sanger sequencing.
The U2OS osteosarcoma cell line was provided by Sean Lee (NIH); it was not subjected to an authentication test. U2OS cells were transfected with vector, wild type, or mutant p85α expression constructs using FuGENE-6 (Roche). Following hygromycin selection, pools of stably selected cells were serum-starved in DMEM/0.5%FBS for 16hr, followed by lysis and Western blotting (details in Supplementary Methods). For immunoprecipitation, lystates were incubated with MYC-tag sepharose bead conjugates (Cell Signaling) overnight at 4°C. All Western blots were repeated in triplicate.
PIK3R1 was somatically mutated in 43% (18 of 42) of EECs and 12% (8 of 66) of NEECs (P= 0.0004, 2-tailed Fisher’s exact test) (Table 1 and Supplementary Figure S1). Within the NEECs, 8% (4 of 46) of serous tumors and 20% (4 of 20) of clear cell tumors were mutated. We observed no significant correlations between PIK3R1 mutations and tumor stage or grade (Supplementary Table S3 and Supplementary Table S4).
The distribution of PIK3R1 mutations was non-random; 93.3% (28 of 30) of PIK3R1 mutations, including three recurrent mutations, localized to the nSH2 and iSH2 domains of p85α which mediate binding to p110α (Figure 1). Fifty percent (15 of 30) of all coding mutations localized within the proximal region (residues 434–475) of the iSH2 domain, including a series of ten overlapping in-frame deletions defined by three shortest regions of overlap (SRO1-SRO3) (Figure 1).
All somatic PIK3R1 mutations appeared to be heterozygous. To determine if the mutations were truly heterozygous or if the wild type allele was contributed by contaminating normal cells, we used LCM to isolate purified tumor cell populations from three cases (T88, T100, and T120) followed by RT-PCR and sequencing. Expression of both mutant and wildtype alleles was observed, confirming heterozygosity in tumor cells (Supplementary Figure S2).
PIK3R1 and PIK3CA mutations are mutually exclusive in glioblastoma multiforme but coexist in colorectal cancer (10, 11, 14). We previously determined the mutational status of PIK3CA, PTEN, and KRAS in our endometrial tumors (13). When merged with our analysis of PIK3R1 mutations, we found that 95% (40 of 42 cases) of EECs, and 41% (27 of 66) of NEECs had somatically mutated one or more of the four genes (Figure 2). We evaluated the patterns of mutations among EECs because all four genes were mutated at high frequency in these tumors. There was no significant difference in the frequency of PIK3R1 mutations between PTEN-mutant (48%, 16 of 33 tumors) and PTEN-wild type EECs (22%, 2 of 9 tumors), or between KRAS-mutant (50%, 9 of 18 tumors) and KRAS-wild type EECs (37%, 9 of 24 tumors). In contrast, PIK3R1 mutations were significantly (P=0.0015) more frequent in PIK3CA-wild type EECs (70%, 14 of 20) than in PIK3CA-mutant EECs (18%, 4 of 22).
Ten tumors (4 EECs and 6 NEECs) had co-existing PIK3R1 and PIK3CA mutations (Table 2). Strikingly, the proportion of truncating mutants of PIK3R1 that co-existed with PIK3CA mutants (63%, 7 of 11 truncations) was significantly higher (P=0.010) than the proportion of missense mutations or in-frame insertions/deletions of PIK3R1 (12%, 2 of 17 mutations) that coexisted with PIK3CA mutations.
We transfected U2OS osteosarcoma cells with constructs expressing either wild type or mutant PIK3R1 to determine the biochemical effects of p85α mutants on phosphorylation of AKTSer473, an important PI3K substrate. We used U2OS cells because they express low endogenous levels of phospho-AKTSer473 (15). Seven p85α mutants present in endometrial tumors were analyzed: p85αdelK459 and p85αdelY463-L466, which define SRO2 and SRO3; p85αR348X and p85αdelR574-T576 which were recurrent mutations in our study; p85αK511VfsX2, the most carboxy-terminal truncation mutant; and p85αdelH450-E451 and p85αN564D, which were present here in endometrial tumors and are also present in other tumor types (10, 11). p85αN564D served as a positive control because it is known to increase PI3K activity (10).
Coimmunoprecipitation of p110α and MYC-tagged p85α mutants showed that all mutants retained the ability to bind p110α except for p85αR348X and p85αdelK511VfsX2 (Figure 3A). Western blotting demonstrated the expected low endogenous level of phospho-AKTSer473 in U2OS cells transfected with the vector control (Figure 3B). As noted previously (10), and consistent with the inhibitory effect of wild type p85α on the PI3K pathway, introduction of wild type p85α into U2OS cells reduced the level of phospho-AKTSer473 compared to vector alone (Figure 3B). In contrast to wild type p85α, stable expression of five p85α mutants (p85αdelH450-E451, p85αdelK459, p85αdelY463-L466, p85αdelR574-T576, and p85αN564D) led to increased levels of p-AKTSer473 compared with vector control (Figure 3B). Only the p85αR348X and p85αK511VfsX2 mutants did not exhibit appreciable changes in phospho-AKTSer473 levels compared with vector control. Ribosomal protein S6, an important downstream target of AKT and mTOR, exhibited a similar phosphorylation pattern to AKTSer473 (Figure 3B).
To our knowledge this is the first report of somatic PIK3R1 (p85α) mutations in endometrial carcinoma. The high frequency and nonrandom distribution of these mutations strongly suggests that mutations of PIK3R1 may be examples of “driver” mutations (16) that confer a selective advantage in endometrial tumorigenesis. In support of this idea, we have shown that stable expression of several p85α mutants leads to functional activation of the PI3K pathway, as evidenced by increased phosphorylation of AKTSer473. Our present findings have relevance not only to endometrial cancer but also to other tumor types; one of the mutants (p85αdelH450-E451) that we have shown to promote AKTSer473 phosphorylation, has also been found in a glioblastoma (14).
Analysis of two in-frame deletion mutants that correspond to two shortest regions of overlapping deletion within the proximal p85α-iSH2 domain (p85αdelK459 and p85αdelY463-L466) demonstrated that each promotes phosphorylation on AKTSer473. Based on this finding we predict that the additional overlapping in-frame deletions are also likely to have altered biochemical properties. Although we have not determined the mechanism by which these deletions promote AKT phosphorylation, we speculate that it might result from altered interactions between the mutant forms of p85α and the cell membrane, since structural studies have suggested that residues 447–561 of p85α form contact with lipid membranes (17), and/or from altered interactions between the p85-iSH2 domain and the p110α-ABD domain (18).
An excess of p85α-nSH2 and -iSH2 truncation mutants was observed in endometrial tumors that had coexisting mutations in PIK3CA. We therefore hypothesize that these truncating mutants of p85α are not functionally equivalent to p110α mutants. In support of this idea, the p85αR348X and p85αK511VfsX2 truncations, which coexist with p110α mutations, did not bind to p110α nor increase p-AKTSer473 levels when stably expressed in U2OS cells; in the case of the p85αR348X mutant, this is consistent with previous observations that this protein fails to bind p110α (10). Since the p85αR301X and p85αY334X mutants, which also coexisted with p110α mutations, truncate p85α amino terminal to residue 348, we predict that these mutants also do not bind p110α or hyperphosphorylate AKT. Exactly how the truncating mutants of p85α, that coexist with p110α mutants, affect p85α function remains to be determined. Nonetheless, their effect on structurally important domains, their preferential co-occurrence with p110α mutations, and the recurrent nature of the R348X mutant here and in colorectal cancers (10), strongly suggest that these are likely to be driver mutations that contribute to endometrial tumorigenesis. Because the majority of somatic p85α mutations uncovered in NEECs were truncation mutants, of uncertain functional significance, future studies will be critical to elucidate the contribution of p85α disruption to this tumor subtype.
In conclusion, we have identified a new mode of PI3K alteration in primary endometrial tumors. Targeted therapies directed against the PI3K pathway have already entered clinical trials for patients with endometrial cancer (19–21). Our findings indicate that it will be important to consider the mutational status of PIK3R1 as molecular correlates associated with clinical outcome are sought. Finally, given our observation that not all p85α mutants are functionally equivalent, future studies will be critical to understand the biochemical properties of the complete spectrum of PIK3R1 mutations present in endometrial carcinoma.
We thank our colleagues for careful reading of the manuscript and insightful discussions.
Grant Support: Funded by the Intramural Program of the NIH/National Human Genome Research Institute (DWB); and in part by grants from the NIH R01 CA140323 (AKG), U01 CA113916 (AKG), NIH RO1-1CA112021-01 (DCS), the Ovarian Cancer Research Fund (AKG), the NCI SPORE in breast cancer at Massachusetts General Hospital (DCS), and the Avon Foundation (DCS).