Introduction

Activating mutations in the PIK3CA are found in approximately 30–40% of patients with cancer and induce hyperactivation of the alpha isoform (p110α) of the phosphatidylinositol 3-kinase (PI3K) [1,2,3]. In patients with HR+/HER2− BC, mTOR/mTOR pathway has been associated with endocrine therapy resistance [4]. In addition, the role of this pathway is becoming increasingly important in HER2+ and TNBC [5,6,7]. Thus, inhibition of PI3K in PIK3CA-mutated BC has been a major focus in the last decade [3].

Alpelisib is an orally bioavailable, small-molecule, α-specific PI3K inhibitor that inhibits p110α approximately 50 times as strongly as other isoforms [8]. Following successful preclinical and phase 1 data [4, 9], the SOLAR-1 phase III randomized trial evaluated the efficacy of alpelisib plus fulvestrant in 572 patients with HR+/HER2− advanced BC who had received prior endocrine therapy [10]. A clinically relevant treatment benefit was only observed in the cohort of patients with PIK3CAmut disease. In May 2019, the FDA approved alpelisib for the treatment of patients with advanced PIK3CAmut HR+/HER2− BC.

Together with alpelisib, the FDA also approved the companion diagnostic therascreen® PIK3CA test (QIAGEN Manchester, Ltd.) used in SOLAR-1 to select patients who had PIK3CA mutations in tumor tissue specimens and/or in circulating tumor DNA (ctDNA) isolated from plasma specimens [11]. Therascreen PIK3CA detects 11 PIK3CA hotspot mutations, mostly found in exons 9 and 20 [11]. In SOLAR-1, the type of PIK3CA mutation did not seem to impact the main results [10].

In this context, patients and physicians might choose not to use the therascreen PIK3CA test and use other available tests, which provide a more comprehensive mutational analysis of PIK3CA as well as other genes. This might lead to the clinical situation where PIK3CA mutations not detected by the therascreen PIK3CA assay, and thus not evaluated in SOLAR-1, are used to indicate alpelisib. To define the potential frequency of this clinical situation, here we aimed to evaluate the distribution of PIK3CA mutations in BC in relation to the therascreen PIK3CA panel.

Methods

Datasets

All non-overlapping publicly available breast datasets (i.e., 12 studies and 6477 samples) (https://www.mbcproject.org/data-release [1, 12,13,14,15,16,17,18,19,20,21]) with PIK3CA mutational status were interrogated from cBio Cancer Genomics Portal (http://cbioportal.org) [22] (Fig. 1 and Additional file 1). Among them, 2 studies focused on 117 patient-derived xenografts [16] and 22 fibroepithelial lesions of the breast [17] were removed. The remaining combined dataset included 6338 invasive tumor samples of which 5535 (87.3%) originated from the METABRIC (n = 2509), the Memorial Sloan-Kettering (n = 1918), and The Cancer Genome Atlas (n = 1108) datasets. All studies analyzed performed targeted or whole exome sequencing (Table A1 Additional file 1). Only single nucleotide variants, insertions, or short deletions in PIK3CA were analyzed.

Fig. 1
figure 1

The CONSORT diagram

Full size image

Therascreen PIK3CA mutation assay

The therascreen® PIK3CA RGQ PCR Kit is a real-time qualitative PCR test for the detection of 11 mutations in PIK3CA gene (exon 7: C420R; exon 9: E542K, E545A, E545D, E545G, E545K, Q546E, and Q546R; and exon 20: H1047L, H1047R, and H1047Y) using genomic DNA extracted from formalin-fixed, paraffin-embedded breast tumor tissue or ctDNA from plasma derived from K2EDTA anticoagulated peripheral whole blood taken from patients with BC [11].

Immunohistochemistry (IHC)-based subtypes

IHC data was available from 5813 patients (92%). Tumors were classified into the 3 main clinically relevant subtypes: (1) HR+/HER2−, (2) HER2+, and (3) TNBC. Tumors identified as progesterone receptor positive and HER2− were considered HR+ regardless of estrogen receptor (ER) status. Tumors identified as ER-negative and HER2− were considered TNBC when progesterone receptor status was not available.

Distribution of PIK3CA mutations in plasma

Frozen plasma samples from 48 patients with advanced HR+/HER2− BC were obtained before initiating a CDK4/6 inhibitor and endocrine therapy. Plasma samples were sent to Guardant Health (California, USA), and the 74-gene standardized NGS-based assay, which includes all 21 exons from the PIK3CA gene, was performed.

Study end points

Primary objective was to evaluate the proportion and distribution of PIK3CA mutations in BC. Secondary objectives were (1) to evaluate in silico the spectrum of PIK3CA mutations in BC that would be captured by the therascreen panel; (2) to evaluate the proportion and distribution of PIK3CA mutations in HR+/HER2−, HER2+, and TNBC; and (3) to explore the identification of PIK3CA mutations in HR+/HER2− advanced BC by the Guardant B360 ctDNA assay.

Statistical analyses

Patient and tumor characteristics were analyzed using descriptive statistics.

Results

Distribution of PIK3CA mutations in BC

In the combined dataset, 36% of patients had PIK3CAmut tumors (Fig. 2a). From a total of 2560 PIK3CA mutations, 205 PIK3CA mutations were unique. The most frequent PIK3CA mutations (i.e., frequency ≥ 4% of all PIK3CAmut tumors) were found in exons 4, 9, and 20: H1047R (35%), E545K (17%), E542K (11%), N345K (6%), and H1047L (4%) (Table 1, Fig. 2b). These 5 mutations comprised 73% of all PIK3CA mutations identified in the combined dataset.

Fig. 2
figure 2

Proportion of PIK3CA mutations in BC in relation to the mutations detected by therascreen. a Proportion of patients with PIK3CA mutations in the combined dataset. b Distribution of the various types of PIK3CA mutations in the combined dataset. c Proportion of PIK3CA mutations detected by the therascreen assay. d Distribution of the various types of PIK3CA mutations detected by the therascreen assay in the combined dataset

Full size image
Table 1 The 20 most frequent PIK3CA mutations in BC
Full size table

PIK3CA mutations captured by the therascreen panel

In the combined dataset, the proportion of PIK3CA mutations included in the therascreen panel was 72% (Fig. 2c). The most frequent types of PIK3CA mutation (i.e., frequency ≥ 4% of all PIK3CAmut tumors) included in the therascreen panel were H1047R (35%), E545K (17%), E542K (11%), and H1047L (4%) (Fig. 2d). These 4 mutations comprised 67% of all PIK3CA mutations detected in the dataset. Of note, N345K mutation in exon 4, which represents 6% of all tumor samples with a PIK3CA mutation, is not part of the therascreen panel (Fig. 3). Although the clinical utility of non-therascreen-detected PIK3CA mutations is currently unknown, the N345K lies within the C2 PI3K-type domain of the protein and confers a gain of function on PI3K, as does C420R (a tested mutation by therascreen assay) [23]. Moreover, N345K mutation has shown increased sensitivity to PI3K pathway inhibition in preclinical models [24]. Interestingly, Q546E PIK3CA mutation included in the therascreen panel was not found in the combined dataset.

Fig. 3
figure 3

Proportion of the 18 most frequent PIK3CA mutations in PIK3CAmut BC in the combined dataset

Full size image

Detection of multiple PIK3CA mutations in a tumor sample

Among 2261 patients with PIK3CAmut tumors, 1979 (87.5%) had 1 single mutation, 267 (11.8%) had 2 mutations, and 15 (0.7%) had 3 or more mutations. Overall, patients with tumors harboring double PIK3CA mutations represented 4% of all BC (Fig. 4a). Among patients with 1 single PIK3CA mutation, 80% would have mutations represented in the therascreen mutational panel (Fig. 4b). Among patients with 2 or more PIK3CA mutations, 78% would have 1 mutation represented in the therascreen panel; 17%, no mutation represented in the therascreen panel; and 5%, 2 or more mutations represented in the therascreen panel (Fig. 4c).

Fig. 4
figure 4

Proportion of patients with PIK3CA mutations in BC in relation to the mutations detected by therascreen. a Proportion of patients with one, two, or three or more PIK3CA mutations in BC in the combined dataset. b Proportion of patients with a single PIK3CA mutation detected PIK3CAmut by the therascreen assay. c Proportion of patients with two or more PIK3CA mutations detected by the therascreen assay as either PIK3CA mutation “not detected,” single PIK3CA mutation or as harboring 2 or more PIK3CA mutations

Full size image

PIK3CA mutational distribution according to subtypes in the BC dataset

Among 5813 patients with IHC data, 4055 (70%) had HR+/HER2− disease, 995 had TNBC (17%), and 763 (13%) had HER2+ disease. PIK3CA mutations were less frequent in TNBC (16%) than in HR+/HER2− (42%) or HER2+ disease (31%) (Fig. 5a–c). However, the distribution of PIK3CA mutations was similar across subtypes (Fig. 5d–f). Seventy-one percent of mutations in HR+/HER2− BC, 75% in HER2+ BC, and 72% in TNBC would be represented in the therascreen panel.

Fig. 5
figure 5

Proportion of PIK3CA mutations across the BC subtypes. a Proportion of PIK3CA mutations in HR+/HER2-negative BC. b Proportion of PIK3CA mutations in HER2+ BC. c Proportion of PIK3CA mutations in TNBC. d Distribution of the various types of PIK3CA mutations in PIK3CAmut HR+/HER2-negative BC. e Distribution of the various types of PIK3CA mutations in PIK3CAmut HER2+ BC. f Distribution of the various types of PIK3CA mutations in PIK3CAmut TNBC

Full size image

Distribution of PIK3CA mutations in plasma ctDNA

Therascreen assay is approved for detecting PIK3CA mutations in ctDNA from plasma samples [25]. To evaluate the distribution of PIK3CA mutations in ctDNA using a highly sensitive assay that sequences all 21 PIK3CA exons, we tested plasma samples from 48 consecutive patients with metastatic HR+/HER2− BC from the Hospital Clinic of Barcelona using the Guardant B360 standardized assay [26]. All patients had recurred or progressed to prior lines and were about to initiate a CDK4/6 inhibitor and endocrine therapy. A PIK3CA mutation was detected in 17 patients (37%), and 1 patient (6%) had double PIK3CA mutation. The spectrum of PIK3CA mutations was similar to the one found in the previous in silico population analysis (Table A2 Additional file 1). More importantly, 5 patients (28%) had PIK3CA mutations not represented in the therascreen mutational panel.

Discussion

PIK3CA mutations have recently reached level 1 evidence for predicting benefit from alpelisib, an alpha-specific PI3K inhibitor, in combination with fulvestrant in patients with advanced HR+/HER2− BC previously treated with endocrine therapy [10]. In addition, several trials are now evaluating alpelisib and other alpha-specific PI3K inhibitors in other BC subtypes harboring PIK3CA mutations [27]. Thus, there is a need to better understand the heterogeneity of the mutational landscape of PIK3CA and, at the same time, relate this heterogeneity with the recently introduced therascreen PIK3CA companion diagnostic assay approved to indicate alpelisib.

To address this topic, we performed a comprehensive evaluation of the distribution of PIK3CA mutations in BC and made the following observations. First, although PIK3CA mutations are highly heterogeneous, 5 mutations (H1047R, E545K, E542K, N345K, and H1047L) represented ~ 70% of all known types of PIK3CA mutations in the dataset. Second, the therascreen PIK3CA mutational panel would represent 72% of all the known PIK3CA mutations and 80% of all patients with a known PIK3CA mutation. Third, 83% of patients with 2 or more PIK3CA known mutations would have mutations found in the therascreen panel; however, in 78% of the cases, only 1 single PIK3CA mutation would be represented in the therascreen assay. Finally, the proportion of PIK3CA mutations differed by BC subtype with HR+/HER2− disease having the highest proportion, followed by HER2+ disease and TNBC. Although less frequent in the HER2+ and TNBC, the proportion is not negligible and several studies, including pivotal or registrational clinical trials, are focusing on these two populations [5,6,7]. To our knowledge, this is the first report to perform a comprehensive analysis of PIK3CA mutations in BC and to relate these findings with the type of mutations captured by the therascreen PIK3CA assay across the three main subtypes of BC.

The SOLAR-1 phase III trial that led to the approval of alpelisib used the therascreen PIK3CA 11-mutation assay in tumor tissue to identify PIK3CA mutations [10]. From a total of 1173 patients tested for PIK3CA mutation status that had interpretable results, 341 (29%) patients had PIK3CAmut disease [10], a proportion which is very similar (28%) to our predicted results if the assay would have been performed in our combined dataset. More importantly, mutations in exon 9 versus exon 20 predicted similarly the degree of benefit to alpelisib in SOLAR-1 [10]. Thus, based on these results, the FDA approved the use of this assay in tumor and plasma samples as a companion diagnostic to indicate alpelisib. The approval of therascreen in plasma samples is based on a subanalysis of the SOLAR-1 trial which showed that PIK3CA mutations identified in plasma samples were also associated with treatment benefit [28].

Our results have important considerations for patients and physicians. In certain parts of the world, determination of PIK3CA status is commonplace using various types of sequencing-based assays. Some of these widely used assays such as Foundation One CDx or Guardant360 cover most or all exons of the PIK3CA gene. Thus, it is highly likely that mutations which are not part of the therascreen PIK3CA 11-mutation assay will be identified with other assays and treatment decisions will be made. In other parts of the world that have not yet implemented somatic genetic testing in BC, the fact that the therascreen panel misses ~ 20–30% of patients with known PIK3CA mutations might be a reason to choose more comprehensive PIK3CA panels.

Critical questions raised by our results are if patients with PIK3CA mutations which are not part of the therascreen panel, or hotspot and non-hotspot PIK3CA mutations identified using sequencing-based assays with higher sensitivities than therascreen, will benefit from alpelisib. For example, mutation N345K represented 5.5% of all PIK3CA mutations in the analyzed dataset and is not captured by the therascreen assay. This mutation was the fourth most frequent PIK3CA mutation in the BC dataset, and COSMIC [29] and OncoKB [30] datasets consider it pathogenic (score 0.95) and oncogenic. Moreover, N345K confers a gain of function and it has shown to increase sensitivity to PI3K inhibitors in preclinical models [23, 24]. A similar situation exists for the sixth most frequently observed PIK3CA mutation, E726K, although OncoKB [30] states that there is conflicting and/or weak data describing the oncogenic function of this mutation, it has been shown that as a single mutation it is weakly activating but as a double mutation (with E545K or H1047R) it is synergistically activating [31]. It is important to notice that the vast majority of E726K mutations are found precisely as double mutants in BC [31]. On the other hand, some less frequent mutations, as G1049R, have demonstrated strong driver activity in a mutation assessment platform. G1049R exhibited activity levels similar to the E542K variant with 20-fold higher frequency [24]. Thus, better functional characterization of these and other non-hotspot PIK3CA mutations together with clinical evidence that predict benefit to alpelisib and other alpha-specific PI3K inhibitors is now of uttermost importance. At the end of the day, each type of PIK3CA mutation should be considered a biomarker by itself.

Another interesting observation is that ~ 4% of all BC, or ~ 12% of all patients with PIK3CAmut BC, have double PIK3CA mutations. Preclinically, double compound PIK3CA mutations result in increased PI3K activity and downstream signaling compared to single hotspot mutants in nontransformed cells and in HR+ BC cells [31]. More importantly, these compound mutations seem to predict for increased sensitivity to PI3K alpha-specific inhibitors compared to single hotspot mutants in both preclinical models and also in selected patients with BC treated in early phase 1 trials [31]. According to our results, the therascreen panel would not capture well double PIK3CA mutations since only 5% of patients known to harbor 2 or more PIK3CA mutations would have mutations represented in the therascreen panel. Thus, if double mutations are confirmed to be a biomarker of ultra-high sensitivity to alpelisib, the therascreen assay might not be ideal for this purpose.

Our study has limitations worth noting. First, we did not evaluate the actual analytical concordance of the therascreen assay versus other sequencing assays. In other words, we assumed that the results of the combined dataset using various sequencing-based strategies was the gold standard and that the therascreen assay would identify 100% of all the PIK3CA-wild-type tumors as “no PIK3CA mutation detected” and 100% of all the PIK3CAmut tumors in the combined dataset as “PIK3CAmut” if the type of mutation was on the therascreen mutation panel. However, the differences in the sensitive and specificity of the various sequencing assays will affect the concordance rates among them [32, 33]. According to the FDA therascreen PIK3CA assay specification sheet, the overall percent agreement between the therascreen assay and an NGS-based assay in SOLAR-1 was 94.7%. Second, the next-generation sequencing assays and the methods used across the 10 studies evaluated in our study are highly heterogeneous and most are not standardized. Third, the analyzed datasets were mostly from primary tumor samples and acquisition of new PIK3CA mutations has been described in the metastatic setting in 8–10% of the cases [34]. Whether the frequency and spectrum of PIK3CA mutations would change if metastatic-only samples had been analyzed is currently unknown.

Conclusion

PIK3CA somatic mutations in BC are highly heterogenous, and the currently validated therascreen companion diagnostic test, which covers 11 hotspot mutations, might not capture up to 20% of patients with PIK3CA mutations. Thus, there is an urgent need to better understand if patients with PIK3CA mutations not detected by the therascreen assay, and predicted to be oncogenic and activating, can benefit from alpelisib or other PI3K inhibitors. Better functional characterization of these and other non-hotspot PIK3CA mutations together with further clinical studies in tumor and plasma samples from SOLAR-1 and other studies will help to better determine the population of patients who benefit from alpelisib or other alpha-specific PI3K inhibitors.