Here we show that endometrial cancers have unique, and tissue-specific, spectrum of somatic PIK3CA
mutations. Unlike other tumor types, in which the majority of PIK3CA
mutations are within exons 9–20, endometrial cancers display a high frequency of mutations in exons 1–7 of PIK3CA
as well as in exons 9–20. Over 60% of the somatic mutations in exons 1–7 were activating. The pattern of PIK3CA
mutations that we observed in endometrial cancers is highly statistically significantly different from that of colorectal cancer, breast cancer and bladder cancer, three tumor types for which PIK3CA
has been comprehensively resequenced (20
Importantly, our genetic and functional data reveal a new subgroup of endometrial cancer patients who have activating somatic mutations within PIK3CA. Had only exons 9 and 20 been sequenced in this study, we would have missed the 16% of NEECs (11 cases) and 16% of all EECs (7 cases) that have PIK3CA mutations exclusively in exons 1–7; of these cases at least 8 NEECs and 5 EECs have an activating mutation in exons 1–7. An additional 3% of NEECs (2 cases) and 9% of EECs (4 cases) would have incompletely genotyped since they have mutations in both exons 1–7 and exon 9–20.
Exons 1–7 of PIK3CA
encode the ABD, RBD, and C2 domains of p110α whereas exons 9–20 encode the helical and kinase domains. Recent structural studies of p110α in complex with p85α have provided critical insights into the distinct properties of the p110α domains (23
). The ABD of p110α forms an interface with both the iSH2 domain of p85α and it also has complex interactions with the first alpha helix of the p110α ABD-RBD linker region as well as the first alpha helix of the kinase domain of p110α (40
). The C2 domain of p110α mediates binding to the cell membrane, the kinase domain of p110α and the iSH2 domain of p85α. The helical domain acts as a scaffold for the assembly of all other p110α domains. The catalytic activity of p110α resides within the kinase domain.
Here we found that almost all mutations in exons 1–7 localized within the ABD, the ABD-RBD linker region, and the C2 domain of p110α whereas very few mutations localized within the RBD, which binds RAS. This pattern is reminiscent of the distribution of rare exon 1–7 mutations that have been reported in colorectal, breast, and bladder cancers (20
). As was previously shown for PIK3CA
exons 9–20 mutations (26
), we found that PIK3CA
exons 1–7 mutations could co-exist with PTEN
mutants in both EECs and NEECs.
The high frequency, and non-random distribution of amino terminal p110α mutants in EECs and NEECs infers that there is a selective advantage to mutationally disrupting the ABD, the ABD-RBD linker, and the C2 domain of p110α in endometrial carcinomas. Consistent with this idea, 62% of the 29 individual mutations we found in exons 1–7 of PIK3CA
, encode gain-of-function mutants of p110α; eight mutations were shown in this study to be gain-of-function mutants that lead to increased levels of phospho-AKTSer473
, and ten additional mutations were previously shown by others to be gain-of-function mutations (33
Approximately one-third of all nonsynonymous mutations present among the 108 endometrial tumors in this study localized within the ABD and proximal ABD-RBD linker region. Strikingly, within these two regions Arg88, Arg 93, and Lys111 residues formed mutational hotspots that accounted for 21% (6 of 29) of all mutations in EECs and 24% (7 of 29) of all mutations in NEECs in our tumor series.
Each of the mutations at Arg88 resulted in an amino acid substitution of arginine for glutamine (R88Q), a known gain-of-function mutant associated with increased AKT activation in vitro
). Our finding that R88Q is a hotspot in endometrial cancer confirms previous observations by Oda et al
., and Dutt et al
., in which R88Q constituted 40% (6 of 15) of PIK3CA
mutations present among 53 endometrial tumors and cell-lines (33
). Arg88 lies on a highly conserved surface of the ABD (40
) and it forms a hydrogen bond with Asp746 in the kinase domain of p110α (23
). It has therefore been proposed that mutations at Arg88 might disrupt this interaction resulting in an altered kinase domain conformation and increased enzymatic activation of PI3K (23
Arg93 (R93) formed a second hotspot within the ABD in our tumor series. Two different mutations were found at this reside, R93W and R93Q. Here, we characterized these two mutants functionally and showed that R93W mutation is a gain-of function mutant that leads to increased phosphorylation of AKT on serine 473. In contrast, we observed no evidence for an increase in AKT phosphorylation associated with the R93Q mutant. This was somewhat unexpected because R93Q was mutated in three different endometrial tumors in our study, strongly suggesting that it would have a selective advantage. Interestingly however, each of the three tumors that harbored the R93Q mutant also had at least one other PIK3CA
mutation in a different domain of p110α (either T1025N, H1047Y/K111N, or A1066V), whereas tumors with the activating the R93W mutant had no other p110α mutation. Our observation that R93Q always occurs as a “double” or “triple” mutant could be functionally relevant because Zhao and Vogt (41
) have shown that two mutations occurring in different domains of p110α can functionally synergize and activate PI3K more potently than either mutation alone. We therefore speculate that the p110α-R93Q mutant might be only weakly activating by itself, below the level of detection in the assays performed here, but that it cooperates with kinase domain mutations to synergistically activate PI3K. Future studies examining the combinatorial effects of the R93Q and its co-occurring mutations will be required to test this hypothesis. However, consistent with the idea that mutations in different domains of p110α can functionally co-operate, the vast majority of endometrial tumors that had two or more PIK3CA
mutations in this study had mutations in different domains of p110α.
In addition to R88 and R93 hotspots, another residue in the ABD, at position 106 (G106), was recurrently mutated in endometrial cancer. We found that both mutations at this site (G106R and G106V) were gain-of-function mutants that increased AKT phosphorylation. It is currently unclear how mutations at this residue affect p110α activity since structural studies have not revealed a specific interaction mediated by residue 106 (23
The third mutation hotspot in the amino terminus of p110α occurred at lysine 111 (K111). One endometrial tumor had a K111N mutation and two additional endometrial tumors had a K111E mutation. Here we showed that the K111E mutant is activating, leading to an increase in AKT phosphorylation compared to wildtype p110α. We observed that the level of phospho-AKT associated with the K111E was lower than for the H1047R mutant. This observation is consistent with the findings of Gymnopoulous et al., that the K111N mutant of p110α is more weakly activating than the H1047R kinase domain mutant (37
The C2 domain of p110α harbored 13% of all nonsynonymous PIK3CA
mutations among the endometrial tumors analyzed in this study. Of the eight individual mutations in the C2 domain, three mutations (E365K in 2 cases, and C420R in one case) were previously shown to be activating (33
). Here we tested the functional consequences of the other five C2 domain mutants that had not been previously characterized. We showed that the delP449-L455 and E453K mutants were activating. Both of these mutants increased AKT phosphorylation at levels similar to, or greater than, the strongly activating H1047R kinase domain mutant. Both p110α-Glu453 and the adjacent residue Glu454, form hydrogen bonds with p85α-Glu348 (39
). It is therefore likely that the delP449-L455 and E453K mutants disrupt this interaction thus leading to increased catalytic activity. Interestingly we observed that the p110α-delP449-L455 deletion mutant led to much higher levels of phosphorylated AKT than the p110α-E453K point mutant. We speculate that higher level of AKT activation seen with the p110α-delP449-L455 deletion mutant might reflect an additive effect of mutating the p110α-Glu453 and Glu454, both of which form hydrogen bonds with p85α, whereas the E453K point mutant affects only one of these residues. Although we found no convincing biochemical evidence that the p110α-V344R, -G364R, and -E453A C2-domain mutants were activating as single mutants, it remains possible that they might contribute to endometrial tumorigenesis via AKT-independent mechanisms (11
), or, in the case of p110α-V344R and p110α-G364R, which co-occur with other PIK3CA
mutations, by co-operating with other p110α mutants. Alternatively, these mutants might be bystander mutations that have no selective advantage to tumorigenesis.
In conclusion, our findings revealed a distinct subgroup of endometrial cancer patients with somatic activating mutations in the amino terminus of p110α in their tumors. Molecular alterations in the PI3K pathway can point to subgroups of cancer patients who might benefit clinically from rationally designed therapies that target the PI3K signal transduction pathway. Therefore, our findings have potential clinical implications suggesting the need to comprehensively evaluate all coding exons of PIK3CA to capture the most appropriate endometrial cancer patients for inclusion in genotype-directed trials of therapeutic agents targeting the PI3K pathway.
STATEMENT OF TRANSLATIONAL RELEVANCE
PI3Kα and its downstream signaling molecules are important therapeutic targets for molecularly defined subsets of cancer patients. In endometrial carcinomas, the occurrence of somatic mutations in the helical and kinase domains of p110α, the catalytic subunit of PI3Kα, has been well documented. Here we show that somatic, activating mutations in the ABD, ABD-linker region and C2 domains of p110α are also very frequent in primary endometrial cancers. This finding identifies a novel subgroup of endometrial cancer patients, who might benefit clinically from targeted therapies directed against the PI3K-mediated pathway.