Purpose:

We explored the prognostic effect of PIK3CA mutation in HER2+ patients enrolled in the ShortHER trial.

Patients and Methods:

The ShortHER trial randomized 1,253 patients with HER2+ breast cancer to 9 weeks or 1 year of adjuvant trastuzumab combined with chemotherapy. PIK3CA hotspot mutations in exon 9 and 20 were analyzed by pyrosequencing. Expression of 60 genes, including PAM50 genes was measured using the nCounter platform.

Results:

A mutation of the PIK3CA gene was detected in 21.7% of the 803 genotyped tumors. At a median follow-up of 7.7 years, 5-year disease-free survival (DFS) rates were 90.6% for PIK3CA mutated and 86.2% for PIK3CA wild-type tumors [HR, 0.84; 95% confidence interval (CI), 0.56–1.27; P = 0.417]. PIK3CA mutation showed a favorable prognostic impact in the PAM50 HER2-enriched subtype (n = 232): 5-year DFS 91.8% versus 76.1% (log-rank P = 0.049; HR, 0.46; 95% CI, 0.21–1.02). HER2-enriched/PIK3CA mutated versus wild-type tumors showed numerically higher tumor-infiltrating lymphocytes (TIL) and significant upregulation of immune-related genes (including CD8A, CD274, PDCD1, and MYBL2, a proliferation gene involved in immune processes). High TILs as well as the upregulation of PDCD1 and MYBL2 were associated with a significant DFS improvement within the HER2-enriched subtype (HR, 0.82; 95% CI, 0.68–0.99; P = 0.039 for 10% TILs increment; HR, 0.81; 95% CI, 0.65–0.99; P = 0.049 for PDCD1 expression; HR, 0.72; 95% CI, 0.53–0.99; P = 0.042 for MYBL2 expression).

Conclusions:

PIK3CA mutation showed no prognostic impact in the ShortHER trial. Within the HER2-enriched molecular subtype, patients with PIK3CA mutated tumors showed better DFS versus PIK3CA wild-type, which may be partly explained by upregulation of immune-related genes.

Translational Relevance

In early HER2+ breast cancer, PIK3CA mutation is associated with reduced rate of pathologic complete response after neoadjuvant chemotherapy and anti-HER2 agents without any impact on disease-free or overall survival. This study evaluated the prognostic role of PIK3CA mutation in the largest cohort from a randomized trial of HER2+ patients treated with adjuvant chemotherapy and trastuzumab. Beyond showing no prognostic effect in the whole study cohort, our study provides unique data describing for the first time a potential prognostic effect of PIK3CA mutation in the HER2-enriched subtype. We also provide data supporting the association of PIK3CA mutation and immune activation in the HER2-enriched cohort, which might at least in part explain the favorable prognostic effect. These data warrant further validation and may contribute to the generation of an integrated multiple biomarker score for HER2+ early breast cancer.

HER2+ breast cancer is a heterogeneous entity with regards to gene expression and gene mutation profile. Such heterogeneity may affect both prognosis and treatment efficacy.

The PI3K pathway is frequently aberrantly activated in breast cancer through activating mutations in the helical (exon 9) or kinase (exon 20) domain of the PIK3CA gene. The rate of tumors harboring a PIK3CA gene mutation varies according to tumor subtype and has been reported in the range of 20%–25% in early HER2+ breast cancer (1–3). In vitro studies have shown that the activation of the PI3K pathway, being downstream to HER2 and hub of multiple intracellular signaling, may drive escape from upstream inhibition of HER2 (4). In the metastatic setting, patients with HER2+ and PIK3CA mutated breast cancer treated with chemotherapy and anti-HER2 therapy experience poorer prognosis as compared with PIK3CA wild-type patients (5, 6). In early breast cancer, a large pooled analysis of individual patients data from prospective trials have demonstrated that PIK3CA mutation is associated with reduced rate of pathologic complete response after chemotherapy and anti-HER2 agents (3). However, investigation of PIK3CA mutation in HER2+ patients enrolled in adjuvant trials have not demonstrated any prognostic effect of this molecular alteration, nor a predictive effect for trastuzumab added to adjuvant chemotherapy (7–9).

However, because HER2+ disease is heterogeneous, one could hypothesize that the effect of PIK3CA mutation may depend on the tumor subtype. For example, in their pooled analysis of neoadjuvant trials for HER2+ breast cancer, Loibl and colleagues showed that PIK3CA mutation was associated with an improved long-term outcome in hormone receptor–negative patients and to a worse outcome in hormone receptor–positive patients (3). It is now clear that the simple dichotomization of HER2+ breast cancer in two subtypes according to hormone receptor status is oversimplistic. PAM50 molecular intrinsic subtypes more accurately recapitulate the complexity of HER2+ breast cancer biology (10). All relevant intrinsic subtypes are represented within HER2+ disease, with a distribution that varies according to hormone receptor coexpression (10–12).

Currently, there are no data on the impact of PIK3CA mutation within molecular intrinsic subtypes of HER2+ breast cancer.

In this study, we investigated the prognostic role of PIK3CA mutation in HER2+ patients with breast cancer enrolled in the prospective randomized ShortHER trial of adjuvant chemotherapy and trastuzumab. Our aim was to evaluate the prognostic effect of PIK3CA mutation in the overall population and according to molecular intrinsic subtype.

Patients

The ShortHER trial is a phase III multicentric trial of adjuvant therapy that randomized (1:1) 1,253 patients with HER2+ early breast cancer to:

Arm A (long): AC (adriamycin 60 mg/sqm plus cyclophosphamide 600 mg/sqm) or EC (epirubicin 90 mg/sqm plus cyclophosphamide 600 mg/sqm) every 3 weeks for four courses followed by paclitaxel 175 mg/sqm or docetaxel 100 mg/sqm every 3 weeks for four courses combined with trastuzumab every 3 weeks for 1 year starting concomitant with taxane (8 mg/kg loading dose followed by 6 mg/kg thereafter) or Arm B (short): docetaxel 100 mg/sqm every 3 weeks for three courses with concomitant trastuzumab every week for 9 weeks (4 mg/kg loading dose followed by 2 mg/kg thereafter) followed by FEC (5-Fluorouracil 600 mg/sqm, epirubicin 60 mg/sqm, and cyclophosphamide 600 mg/sqm) every 3 weeks for four courses.

The aim was to demonstrate the noninferiority of short (Arm B) versus long (Arm A) treatment in terms of disease-free survival (DFS). Further details and results of the primary study endpoint are reported elsewhere (13). Briefly, the study failed to demonstrate the noninferiority of 9 weeks of the short treatment: 5-years DFS rates were 88% in the long and 85% in the short arm [HR, 1.13; 90% confidence interval (CI), 0.89–1.42, with a predefined noninferiority margin of 1.29]. This analysis was approved by the competent Ethical Committee in November 2014, patients provided signed informed consent for tumor sample centralization and use for research purpose. The study was conducted in accordance with ethical guidelines (1964 Helsinki declaration and its later amendments or comparable ethical standards).

The consort diagram for the analyses described in this work is shown in Supplementary Fig. S1.

PIK3CA gene status

Formalin-fixed paraffin-embedded (FFPE) tumor blocks were reviewed for quality and tumor content. Macro dissection from 5-μmol/L FFPE sections of primary tumor lesions containing at least 20% of tumor cells were carried out to obtain tumor DNA. DNA was extracted with the Maxwell 16 FFPE Tissue LEV DNA Purification Kit (Promega) in the Maxwell 16 Instrument (Promega), according to the manufacturer's instructions. PIK3CA status was analyzed by pyrosequencing using anti-EGFR MoAb response (PIK3CAstatus) Kit (Diatech Pharmacogenetics), according to the manufacturer's instructions. Reactions were run on a PyroMark Q96 ID (Qiagen). This kit allows identification of the most important mutations in exon 9 (E542K, E545K, E545A, E545G, Q546E, and Q546K) and exon 20 (M1043I, H1047Y, H1047R, H1047L, G1049R, and G1049S) of the PIK3CA gene.

Tumor-infiltrating lymphocytes

Methods for tumor-infiltrating lymphocytes (TIL) assessment and TILs data have been described previously (14). TILs data included in this article derive from previous publication (14).

Gene expression

First, a section of centralized FFPE surgical tumor sample was examined with hematoxylin and eosin staining to confirm diagnosis and determine tumor surface area. If needed, microdissection was then performed to avoid normal breast contamination. RNA was then extracted from FFPE material using the High Pure FFPET RNA Isolation Kit (Roche) following manufacturer's protocol and quantified using the NanoDrop spectrophotometer (Thermo Fisher Scientific).

A minimum of approximately 100 ng of total RNA was used to measure the expression of 55 BC-related genes, including the PAM50 genes, androgen receptor and some immune-related genes (e.g., CD8A, CD4, PD1, and PD-L1), and five housekeeping genes (ACTB, MRPL19, PSMC4, RPLP0, and SF3A1) using the nCounter platform (Nanostring Technologies; ref. 15). Data were log base 2–transformed and normalized using the housekeeping genes. The complete list of genes can be found in Supplementary Table S1. Intrinsic molecular subtyping at baseline was determined using the previously reported PAM50 subtype predictor (16).

Statistical analysis

Statistical analysis was carried out using IBM SPSS Version 25 and R project software 3.4.4 (17).

Association between variables was evaluated by the Pearson χ2 test or the Mann–Whitney test, according to the nature of the variables.

DFS was defined as the time between randomization and any of the following events, whichever first: local, regional, and distant recurrence; contralateral breast cancer, excluding in situ carcinoma; other second invasive primary cancer; death before recurrence or second primary cancer. Patients without event were censored at the date of last follow-up. The Kaplan–Meier method was used to estimate survival curves, the log-rank test was used to compare between groups. Cox proportional regression models were used to calculate HRs and 95% CIs. The likelihood ratio test was used to evaluate the amount of prognostic information provided by single variables when added to integrated prognostic models within the HER2-enriched subtype. Level of significance was P < 0.05.

To identify genes whose expression was significantly different according to PIK3CA mutational status, we used a two-class unpaired significance of microarrays (SAM) analysis with a FDR <10%.

No formal sample size calculation was performed, because the sample population was based on the number of cases with centralized tumor sample that was suitable for the present analysis.

The analyses described in this work were not prespecified in the study protocol.

Patients’ characteristics

Centralized tumor samples were available for 913 of the 1,253 randomized patients. A total of 803 cases (64% of all randomized patients) had tumor samples suitable for PIK3CA assessment and were included in this analysis (Supplementary Fig. S1). The characteristics of the 803 patients analysed for PIK3CA and the comparison with the 450 patients from the ShortHER trial not included in this analysis are shown in Supplementary Table S2. As compared with patients not assessed for PIK3CA, those included in the PIK3CA cohort were older [median age 56 (Q1–Q3: 48–64) vs. 54 (Q1–Q3: 46–62), P = 0.012], presented more often hormone receptor positive tumors (71% vs. 64%, P = 0.010), and showed lower TILs levels [median 5 (Q1–Q3: 1–15) vs. 5 (Q1–Q3: 1–30), P = 0.040].

PIK3CA

A mutation of the PIK3CA gene was detected in 21.7% of the 803 genotyped patients (n = 174 mutated; n = 629 wild-type). Mutations in exon 9 and 20 occurred in 78 (9.7%) and 95 (11.8%) cases, respectively (Supplementary Fig. S2 shows the pattern of mutations).

Association between PIK3CA and clinicopathologic characteristics is shown in Table 1. PIK3CA mutation occurred more frequently in hormone receptor–positive versus hormone receptor–negative cases (23.5% vs. 17.4%, P = 0.057) and in postmenopausal vs. premenopausal patients (23.8% vs. 17.7%, P = 0.050). No association with stage, age, grade, TILs, and treatment arm was observed. Although the distribution of PIK3CA mutation according to molecular subtype was not significant, the rate of PIK3CA mutation was numerically higher in Luminal A (22/86, 26%), followed by HER2-enriched (49/232, 21%), Luminal B (8/42, 19%), and Basal-like (2/27, 7%).

Table 1.

Patients’ characteristics in the PIK3CA cohort according to PIK3CA gene status.

PIK3CA mutatedPIK3CA wild-typeTOT (PIK3CA cohort)
CharacteristicsN (%)N (%)N (%)P
Total  174 (22%) 629 (78%) 803 (100%) — 
Age (years)      
 <60 104 (60) 393 (62) 497 (62)  
 ≥60 70 (40) 236 (38) 306 (38) 0.515 
 Median (Q1–Q3) 57 (50–64) 56 (48–64) 56 (48–64) 0.169 
Menopausal status      
 Premenopausal 49 (28) 227 (36) 276 (34)  
 Postmenopausal 125 (72) 401 (64) 526 (66) 0.050 
AJCC Stage      
 70 (40) 259 (41) 329 (41)  
 II 77 (44) 277 (44) 354 (44)  
 III 27 (16) 91 (15) 118 (15) 0.936 
N stage      
 N0 85 (49) 343 (55) 428 (53)  
 N1-N2 61 (35) 191 (30) 252 (31)  
 N3 28 (16) 95 (15) 123 (15) 0.393 
Hormone receptors      
 Negative 41 (24) 195 (31) 236 (29)  
 Positive 133 (76) 434 (69) 567 (71) 0.057 
Histologic grade      
 Grade 1–2 54 (31) 172 (28) 226 (28)  
 Grade 3 118 (69) 448 (72) 566 (72) 0.348 
 
TILs Median (Q1–Q3) 5 (1–15) 5 (1–15) 5 (1–15) 0.637 
 
PAM50 Intrinsic subtype      
 LumA 22 (24) 64 (19) 86 (20)  
 LumB 8 (9) 34 (10) 42 (10)  
 HER2-enriched 49 (53) 183 (54) 232 (54)  
 Basal-like 2 (2) 25 (7) 27 (6)  
 Normal-like 11 (12) 35 (10) 46 (11) 0.358 
Treatment arm      
 A long 93 (53) 312 (50) 405 (50)  
 B short 81 (47) 317 (50) 398 (50) 0.369 
PIK3CA mutatedPIK3CA wild-typeTOT (PIK3CA cohort)
CharacteristicsN (%)N (%)N (%)P
Total  174 (22%) 629 (78%) 803 (100%) — 
Age (years)      
 <60 104 (60) 393 (62) 497 (62)  
 ≥60 70 (40) 236 (38) 306 (38) 0.515 
 Median (Q1–Q3) 57 (50–64) 56 (48–64) 56 (48–64) 0.169 
Menopausal status      
 Premenopausal 49 (28) 227 (36) 276 (34)  
 Postmenopausal 125 (72) 401 (64) 526 (66) 0.050 
AJCC Stage      
 70 (40) 259 (41) 329 (41)  
 II 77 (44) 277 (44) 354 (44)  
 III 27 (16) 91 (15) 118 (15) 0.936 
N stage      
 N0 85 (49) 343 (55) 428 (53)  
 N1-N2 61 (35) 191 (30) 252 (31)  
 N3 28 (16) 95 (15) 123 (15) 0.393 
Hormone receptors      
 Negative 41 (24) 195 (31) 236 (29)  
 Positive 133 (76) 434 (69) 567 (71) 0.057 
Histologic grade      
 Grade 1–2 54 (31) 172 (28) 226 (28)  
 Grade 3 118 (69) 448 (72) 566 (72) 0.348 
 
TILs Median (Q1–Q3) 5 (1–15) 5 (1–15) 5 (1–15) 0.637 
 
PAM50 Intrinsic subtype      
 LumA 22 (24) 64 (19) 86 (20)  
 LumB 8 (9) 34 (10) 42 (10)  
 HER2-enriched 49 (53) 183 (54) 232 (54)  
 Basal-like 2 (2) 25 (7) 27 (6)  
 Normal-like 11 (12) 35 (10) 46 (11) 0.358 
Treatment arm      
 A long 93 (53) 312 (50) 405 (50)  
 B short 81 (47) 317 (50) 398 (50) 0.369 

Abbreviations: AJCC, American Joint Committee on Cancer; LumA, luminal A; LumB, luminal B; TILs, tumor-infiltrating lymphocytes.

PIK3CA and DFS

At a median follow-up of 7.7 years (95% CI, 7.5–7.9), 146 DFS events occurred: 28 in patients with PIK3CA mutated tumor (28/174, 16%) and 118 in patients with PIK3CA wild-type tumor (118/629, 19%; P = 0.419). Figure 1 shows the Kaplan–Meier DFS curves according to PIK3CA gene status. DFS rates at 5 years were 90.6% for PIK3CA mutated and 86.2% for PIK3CA wild-type groups (HR, 0.84; 95% CI, 0.56–1.27; P = 0.417). Similar results were obtained in patients randomized to receive 1 year of trastuzumab (HR, 0.86; 95% CI, 0.48–1.55) and in patients randomized to 9 weeks of trastuzumab (HR, 0.84; 95% CI, 0.47–1.50).

Figure 1.

Kaplan–Meier curves for DFS according to PIK3CA gene mutation (n = 803 patients from the ShortHER trial).

Figure 1.

Kaplan–Meier curves for DFS according to PIK3CA gene mutation (n = 803 patients from the ShortHER trial).

Close modal

We explored the prognostic effect of PIK3CA gene status according to hormone receptor expression and intrinsic molecular subtype. Because of limited sample size and number of events within individual intrinsic molecular sutbypes, we analyzed the impact of PIK3CA mutations in two groups: HER2-enriched and non-HER2-enriched (including Luminal A, Luminal B, Basal, and Normal-like). Kaplan–Meier curves are shown in Fig. 2. PIK3CA mutation had no significant effect on DFS in hormone receptor–positive and hormone receptor–negative subgroups. According to intrinsic subtype, patients with PIK3CA tumor experienced better DFS as compared with PIK3CA wild-type in the HER2-enriched subgroup: 5-year DFS rate was 91.8% for PIK3CA mutated versus 76.1% for PIK3CA wild-type groups, log-rank P = 0.049 (HR, 0.46; 95% CI, 0.21–1.02; P = 0.055). No difference in DFS was observed in non-HER2-enriched patients according to PIK3CA gene status. The test for interaction was not significant (P = 0.269). Within the HER2-enriched group, the positive prognostic impact of PIK3CA mutation was particularly evident in patients with hormone receptor–negative tumor (n tot = 88; PIK3CA mutated n = 14, 0 events; PIK3CA wild-type n = 74, 21 events): 5-year DFS rate was 100% versus 75.6%, log-rank P = 0.029. In HER2-enriched/hormone receptor–positive patients (n tot = 144, PIK3CA mutated n = 35, 7 events; PIK3CA wild-type n = 109, 30 events), 5-year DFS rate in the PIK3CA mutated and wild-type groups was 88.6% and 76.5%, respectively (log-rank P = 0.329).

Figure 2.

Kaplan–Meier curves for DFS according to PIK3CA gene mutation in subgroups: hormone receptor–positive (A), hormone receptor–negative (B), HER2-enriched (C), and non–HER2-enriched (D).

Figure 2.

Kaplan–Meier curves for DFS according to PIK3CA gene mutation in subgroups: hormone receptor–positive (A), hormone receptor–negative (B), HER2-enriched (C), and non–HER2-enriched (D).

Close modal

Clinicopathologic characteristics of PIK3CA mutated, HER2-enriched tumors

To explore the potential reasons for favorable prognostic effect of PIK3CA mutation within the HER2-enriched molecular subtype, we first analyzed the association between PIK3CA gene status and classic clinicopathologic factors in this patients’ subgroup (Supplementary Table S3). PIK3CA gene mutation was numerically associated with hormone receptor–positive status (71% vs. 60%, P = 0.128) and higher TILs [median 8 (Q1–Q3:2–30) vs. median 5 (Q1–Q3:1–20), P = 0.164]. Figure 3 shows box plot for TILs levels according to PIK3CA mutation in the entire HER2-enriched cohort as well as in HER2-enriched/hormone receptor–positive and HER2-enriched/hormone receptor–negative patients. In the HER2-enriched/hormone receptor–negative subgroup, TILs were significantly higher in PIK3CA mutated tumors [median 24% (Q1–Q3:10–50) for PIK3CA mutated vs. median 7 (Q1–Q3:2–20) in PIK3CA wild-type, P = 0.005]. We have previously demonstrated the independent prognostic role of TILs in the ShortHER trial (14). TILs maintained a significant association with DFS in the subgroup of HER2-enriched patients analyzed for PIK3CA in this study: (HR, 0.82; 95% CI, 0.68–0.99; P = 0.039 for each 10% TILs increment].

Figure 3.

Box plot showing TILs levels according to PIK3CA mutation in HER2-enriched patients: all patients (A), hormone receptor–positive (B), and hormone receptor–negative (C).

Figure 3.

Box plot showing TILs levels according to PIK3CA mutation in HER2-enriched patients: all patients (A), hormone receptor–positive (B), and hormone receptor–negative (C).

Close modal

Gene expression profile of PIK3CA mutated, HER2-enriched tumors

We analyzed differences in gene expression among HER2-enriched tumors according to PIK3CA gene status (Fig. 4). As reported in Supplementary Table S4, the following genes were overexpressed in PIK3CA mutated tumors (two-class unpaired SAM analysis, FDR < 0.10): genes tracking luminal properties (ESR1 and PGR), genes tracking proliferation and cell-cycle processes (MYC, MKI67, CEP55, MYBL2), immune-related genes (CD8A, CD274 encoding for PD-L1, PDCD1 encoding for PD-1), and a gene encoding for the microtubule-associated protein tau (MAPT). Genes that resulted downregulated in PIK3CA mutated tumors were: ERBB2, GRB7 (which is one of the 105 protein-encoding genes located in the same amplicon as ERBB2) and TMEM45B [encoding for a member of the transmembrane (TMEM) family, which includes proteins that span biological membranes]. The same analysis was conducted stratified by hormone receptor status (Supplementary Table S4): the only differentially expressed genes were ERBB2 and GRB7 that resulted downregulated in PIK3CA mutated tumors in the HER2-enriched/hormone receptor–positive subgroup.

Figure 4.

Unsupervised clustering of 55 genes across HER2-enriched tumors from the ShortHER trial (N = 232). Hormone receptor status (positive - blue; negative - yellow) and PIK3CA mutational status (mutated - black; gray – wild-type) of each sample is also presented.

Figure 4.

Unsupervised clustering of 55 genes across HER2-enriched tumors from the ShortHER trial (N = 232). Hormone receptor status (positive - blue; negative - yellow) and PIK3CA mutational status (mutated - black; gray – wild-type) of each sample is also presented.

Close modal

We then explored the association between differentially expressed genes in PIK3CA mutated versus PIK3CA wild-type tumors and DFS, in the cohort of HER2-enriched patients. Univariate cox regression analysis is shown in Table 2. Two genes found to be upregulated in HER2-enriched/PIK3CA mutated had a significant association with improved DFS: MYBL2 (HR, 0.72; 95% CI, 0.53–0.00; P = 0.042) and PDCD1 (HR, 0.81; 95% CI, 0.65–0.99; P = 0.049). Interestingly, as further described in the Discussion, MYBL2 may be implicated in processes linked to tumor immune activation. Among the genes found to be downregulated in HER2-enriched/PIK3CA mutated tumors, TMEM45B was associated with a worse DFS (HR, 1.31; 95% CI, 1.02–1.69; P = 0.037).

Table 2.

HER2-enriched patients: univariate DFS Cox regression analysis of genes found to be differentially expressed according to PIK3CA gene status.

All patients
GeneHR (95% CI)P
Upregulated in PIK3CA mut 
ESR1 1.06 (0.95–1.19) 0.300 
PGR 1.04 (0.96–0.93) 0.490 
MYC 0.84 (0.64–1.09) 0.193 
MKI67 0.75 (0.49–1.17) 0.203 
CEP55 0.64 (0.39–1.04) 0.073 
MYBL2 0.72 (0.53–0.99) 0.042 
CD8A 0.84 (0.68–1.05) 0.118 
CD274 0.80 (0.57–1.12) 0.197 
PDCD1 0.81 (0.65–0.99) 0.049 
MAPT 0.98 (0.88–1.19) 0.801 
Downregulated in PIK3CA mut 
ERBB2 0.98 (0.81–1.20) 0.873 
GRB7 0.97 (0.80–1.18) 0.773 
TMEM45B 1.31 (1.02–1.69) 0.037 
All patients
GeneHR (95% CI)P
Upregulated in PIK3CA mut 
ESR1 1.06 (0.95–1.19) 0.300 
PGR 1.04 (0.96–0.93) 0.490 
MYC 0.84 (0.64–1.09) 0.193 
MKI67 0.75 (0.49–1.17) 0.203 
CEP55 0.64 (0.39–1.04) 0.073 
MYBL2 0.72 (0.53–0.99) 0.042 
CD8A 0.84 (0.68–1.05) 0.118 
CD274 0.80 (0.57–1.12) 0.197 
PDCD1 0.81 (0.65–0.99) 0.049 
MAPT 0.98 (0.88–1.19) 0.801 
Downregulated in PIK3CA mut 
ERBB2 0.98 (0.81–1.20) 0.873 
GRB7 0.97 (0.80–1.18) 0.773 
TMEM45B 1.31 (1.02–1.69) 0.037 

Abbreviations: CI, confidence interval; HR, hazard ratio.

Integrated prognostic models within the HER2-enriched subtype

We assessed the amount of prognostic information provided by PIK3CA mutation, TILs, PDCD1 expression, and MYBL2 expression when added to integrated prognostic models for the HER2-enriched subtype. Among classic clinicopathologic factors, stage was significantly associated with DFS (stage II vs. stage III: HR, 0.48; 95% CI, 0.26–0.88; P = 0.019; stage I vs. stage III: HR, 0.37; 95% CI, 0.18–0.75; P = 0.006). Age (HR, 1.02; 95% CI, 0.99–1.04; P = 0.264), histologic grade (grade 1–2 vs. 3: HR, 1.03; 95% CI, 0.56–1.91; P = 0.920), and hormone receptor status (positive vs. negative: HR, 1.11; 95% CI, 0.65–1.90; P = 0.703) were not associated with DFS. When TILs, PIK3CA mutation, PDCD1 expression, or MYBL2 expression were added as single variables to a model containing stage, all provided a significant amount of prognostic information (Supplementary Table S5). The inclusion of either TILs or PDCD1 expression added significant prognostic information when included to a model containing stage and PIK3CA status (Supplementary Table S5). To the opposite, the inclusion of PIK3CA mutation to a model containing stage and TILs, stage and PDCD1 expression, or stage and MYBL2 expression was not significantly prognostic (Supplementary Table S5).

This study represents the largest cohort of patients with HER2+ early breast cancer homogeneously treated with adjuvant chemotherapy and trastuzumab in the context of a clinical trial that were evaluated for PIK3CA mutation (n = 803) and the first study to date to investigate the prognostic role of PIK3CA mutation within molecular intrinsic subtypes of HER2+ breast cancer (7–9).

We did not find any prognostic impact of PIK3CA gene mutation in the overall study cohort. This result is concordant with data from other adjuvant trials showing no association between PIK3CA mutation and long-term outcome (7–9). Moreover, in the NSABP B-31 and FinHER studies, PIK3CA mutation was not predictive of reduced benefit from adjuvant trastuzumab (7, 8). More recently, PIK3CA mutation did not result associated with differential benefit from adjuvant neratinib in the ExteNET trial (9). Two pooled analyses also confirmed no robust impact of PIK3CA mutation on long-term outcome. Zardavas and colleagues conducted a pooled analysis of more than 10,000 patients with early breast cancer (18). In the HER2+ subgroup, they described a numerically worse overall survival for patients with PIK3CA mutation (HR, 1.17; 95% CI, 0.94–1.46). However, there was no impact of PIK3CA mutation on invasive DFS (HR, 0.98) or distant DFS (HR, 0.93). Moreover, not all the patients in this cohort received adjuvant trastuzumab. In the pooled analysis of neoadjuvant trials in HER2+ disease, although pathologic complete response rates after chemotherapy and anti-HER2 treatment were lower in patients with PIK3CA mutated tumor, this did not translate into a worse DFS (3). Some hypotheses may explain this discrepancy in the effect of PIK3CA mutation on pathologic complete response and long-term outcome. First, PIK3CA mutation might confer a more indolent biology that may result in lower rates of pathologic complete response without affecting long-term outcome, similarly to what is observed in case of HER2+/hormone receptor–positive breast cancer (19). In our cohort, patients with PIK3CA mutation were more frequently postmenopausal and showed more frequently HR-positive disease. However, the association with hormone receptor status is not consistent across studies (3, 8, 9). According to intrinsic subtype, in our study, Luminal A tumors showed the highest frequency of PIK3CA mutation, although the distribution of PIK3CA mutation across intrinsic subtypes was not statistically significant. In the CALGB 40601 neoadjuvant study, the highest rate of PIK3CA mutation was detected in Luminal B/HER2+ tumors (20). A second potential explanation for discrepant data may be linked to the observation that PIK3CA mutation not necessarily determines downstream PI3K pathway activation (21), a factor that may act as a confounder in studies looking at PIK3CA mutation. Finally, one could also hypothesize a different role of PIK3CA mutation in macroscopic versus microscopic disease. It has been previously suggested that antibody-dependent cell-mediated cytotoxicity (ADCC), which is known as the main mechanisms of action of trastuzumab, may be more effective, in the presence of PIK3CA mutation, in case of microscopic disease, whereas in case of macroscopic disease it may be more difficult for ADCC to overcome resistance due to downstream pathway activation (22, 23).

However, another hypothetic explanation may be that the role of PIK3CA mutation depends on tumor subtype. A first hint was suggested by the pooled analysis by Loibl and colleagues. In this study, there was a significant interaction between PIK3CA mutation and hormone receptor status for both pathologic complete response and DFS: the reduction in pathologic complete response rate was significant in hormone receptor–positive disease and not in hormone receptor–negative disease (Pinteraction = 0.036), PIK3CA was associated with significantly worse DFS in hormone receptor–positive patients and with a trend for a better DFS in hormone receptor–negative patients (Pinteraction = 0.021; ref. 3).

In our study, we did not observe any difference in the impact of PIK3CA mutation on outcome according to hormone receptor status. However, we found an association between the presence of PIK3CA mutation and improved outcome in HER2-enriched patients (5-year DFS rate 91.8% vs. 76.1%, log-rank P = 0.049; HR, 0.46; 95% CI, 0.21–1.02; P = 0.055; test for interaction between molecular subtype and PIK3CA mutation not significant). In this subgroup, PIK3CA mutation was significantly prognostic beyond stage. On the basis of this result, we investigated potential biological differences between PIK3CA mutated and wild-type tumors within the HER2-enriched subtype. We observed that HER2-enriched/PIK3CA mutated tumors were numerically more frequently hormone receptor–positive, showed increased expression of ESR1 and PGR and a decreased expression of ERBB2 as compared with HER2-enriched/PIK3CA wild-type tumors. These results suggest that PIK3CA mutation may be associated, even within the HER2-enriched subgroup, to a more luminal-like and less HER2-addicted gene expression profile as compared with PIK3CA wild-type tumors that might account for a more indolent behavior. We also found that HER2-enriched/PIK3CA-mutated tumors showed numerically higher TILs levels (statistically significant in hormone receptor–negative patients), and upregulation in genes involved in proliferation and immune response. In particular, when exploring, within the HER2-enriched subtype, the association with DFS of the features and genes found to be differentially expressed between PIK3CA mutated and PIK3CA wild-type tumors, the biological processes that seemed to affect prognosis to a larger extent were those attributable to immune pathways. Indeed, increase TILs levels and increased expression of PDCD1 and MYBL2 were all significantly associated to improved prognosis. In multivariable models, TILs and PDCD1 added significant prognostic information beyond stage and PIK3CA mutation, whereas PIK3CA mutation did not add a significant amount of prognostic information beyond models containing stage and TILs or stage and PDCD1 expression. This observation corroborates the hypothesis that immune features associated with PIK3CA mutation, rather than PIK3CA mutation itself, may be the main biological driver of the observed prognostic effect on DFS of PIK3CA in univariate analysis. TILs are known to be a strong prognostic factor in HER2+ breast cancer and reflect a general state of immune activation. The significant prognostic effect of TILs in patients with early HER2+ breast cancer treated with adjuvant therapy has been previously demonstrated by our group and others (14, 24). PDCD1 expression is dynamic and tends to increase to counterbalance an antitumor immune activation; the prognostic role of PD-L1 gene expression in HER2-enriched breast cancer has not been clearly established thus far (25). MYBL2 is a transcription factor involved in cell cycle; however, there is evidence that it may also be implicated in immune activation or in immune response–promoting processes. Indeed, MYBL2 may mediate increase mutational load by inducing APOBEC expression (26), may promote XCL1 expression in breast cancer which is associated with a type 1 dendritic cell signature and improved survival (27) and may be implicated in chromosomal instability (28). In our article, MYBL2 provided significant prognostic information beyond stage but not to a model containing stage and PIK3CA mutation.

Interestingly, other works have linked PIK3CA mutation to immune infiltration. A recent study by Sobral-Leite and colleagues focused on hormone receptor–positive patients (also including HER2+ cases) found that tumors with PIK3CA mutation tend to have more CD8 cells and tumors enriched for FOXP3-positive cells show downstream activation of the PI3K pathway (29). More functional data are needed to better understand the interactions between the PI3K pathway and the immune microenvironment.

The strengths of our study are: the large prospective cohort of patients from a randomized trial, the large number of samples suitable for molecular analyses, and the study design. Our study has limitations, including: the fact that these analyses were not prespecified in the protocol and the lack of statistical power that precluded the possibility to evaluate the association of PIK3CA with DFS within each single molecular intrinsic subtype separately. Another limitation is the lack of correction for multiple comparisons, due to the exploratory nature of the study, raising the possibility that some of the findings might be due to chance. Therefore, our results should not be regarded as conclusive but rather as hypothesis generating. However, they add novel findings to the knowledge in the field that deserve validation.

In conclusion, our study provides unique data exploring the role of PIK3CA mutation in HER2+ breast cancer by integrating the classification into molecular intrinsic subtype; we describe for the first time a potential prognostic effect of PIK3CA mutation in the HER2-enriched subtype; we provide data supporting the association of PIK3CA mutation and immune activation in the HER2-enriched cohort, which might at least in part explain the favorable prognostic effect. These data warrant further validation and may contribute to the generation of an integrated multiple biomarker score for HER2+ early breast cancer. Indeed, beyond classic clinicopathologic factors, it is now clear that intrinsic subtypes and immune features represent independent prognostic factors that could be combined into a prognostic score for patients’ stratification. How and if PIK3CA mutation should be integrated in prognostic models for HER2-enriched tumors needs to be further evaluated.

V. Guarneri reports personal fees from Eli Lilly (advisory board, speaker's bureau), Novartis (advisory board, speaker's bureau), and grants from Roche (institutional research grant) outside the submitted work. M.V. Dieci reports personal fees from Eli Lilly, Genomic Health, Novartis, and Celgene outside the submitted work, and is listed as a coinventor on a patent regarding a prognostic index for HER2+ patients with early breast cancer that will be licensed to the University of Barcelona and University of Padova. A. Frassoldati reports personal fees from Roche (nonfinancial support), Novartis (nonfinancial support), Pfizer (nonfinancial support), and Lilly (nonfinancial support) outside the submitted work. A. Musolino reports grants and personal fees from Roche and Eisai, as well as personal fees from Macrogenics, Lilly, Novartis, and Merck-MSK outside the submitted work. O. Garrone reports personal fees from Eisai, Amgen, Eli Lilly, Novartis, and Pfizer outside the submitted work. G. Griguolo reports nonfinancial support from Pfizer (travel support). N. Chic reports other from Novartis (travel expenses) and Eisai (travel expenses) outside the submitted work. A. Prat reports personal fees from Nanostring Technologies during the conduct of the study; grants and personal fees from Roche and Novartis; and personal fees from Daiichi Sankyo, AstraZeneca, Oncolytics Biotech, Pfizer, and MSD outside the submitted work, and is listed as a coinventor of a patent regarding predicting prognosis in early HER2+ disease, owned by IDIBAPS and the University of Padova. P. Conte reports grants from AIFA (institutional research grant) during the conduct of the study; grants from Merck KGA (institutional research grant); grants and personal fees from Bristol-Myers Squibb (institutional research grant; advisory board), Novartis (institutional research grant; advisory board), and Roche (institutional research grant; advisory board); and personal fees from Eli Lilly (advisory board), Tesaro (advisory board), and AstraZeneca (advisory board) outside the submitted work, and is listed as a coinventor on a patent regarding a prognostic index for HER2+ patients with early breast cancer that will be licensed to the University of Barcelona and University of Padova. No potential conflicts of interest were disclosed by the other authors.

V. Guarneri: Conceptualization, supervision, funding acquisition, writing-review and editing. M.V. Dieci: Conceptualization, data curation, formal analysis, writing-original draft. G. Bisagni: Data curation, writing-review and editing. A.A. Brandes: Data curation, writing-review and editing. A. Frassoldati: Data curation, writing-review and editing. L. Cavanna: Data curation, writing-review and editing. A. Musolino: Data curation, writing-review and editing. F. Giotta: Data curation, writing-review and editing. A. Rimanti: Data curation, writing-review and editing. O. Garrone: Data curation, writing-review and editing. E. Bertone: Data curation, writing-review and editing. K. Cagossi: Data curation, writing-review and editing. O. Nanni: Data curation, writing-review and editing. F. Piacentini: Data curation, writing-review and editing. E. Orvieto: Data curation, writing-review and editing. G. Griguolo: Data curation, formal analysis, writing-review and editing. M. Curtarello: Data curation, writing-review and editing. L. Urso: Data curation, writing-review and editing. L. Paré: Data curation, writing-review and editing. N. Chic: Data curation, writing-review and editing. R. D'Amico: Data curation, formal analysis, methodology, writing-review and editing. A. Prat: Data curation, writing-review and editing. P. Conte: Conceptualization, data curation, supervision, funding acquisition, writing-review and editing.

This work was supported by research grants from Agenzia Italiana del Farmaco (AIFA, grant FARM62MC97; to P. Conte), Italian Association for Cancer Research (AIRC, project MFAG 2014 – 15938; to V. Guarneri), Veneto Institute of Oncology (5 × 1000 program; to M.V. Dieci), Instituto de Salud Carlos III (P116/00904; to A. Prat), and Breast Cancer Now (2018NOVPCC1294; to A. Prat).

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.

1.
Dieci
MV
,
Guarneri
V
. 
PIK3CA: a target or a marker in breast cancers
.
Curr Breast Cancer Rep
2015
;
7
:
161
9
.
2.
Cancer Genome Atlas Network
. 
Comprehensive molecular portraits of human breast tumors
.
Nature
2012
;
490
:
61
70.
3.
Loibl
S
,
Majewski
I
,
Guarneri
V
,
Nekljudova
V
,
Holmes
E
,
Bria
E
, et al
PIK3CA mutations are associated with reduced pathological complete response rates in primary HER2-positive breast cancer: pooled analysis of 967 patients from five prospective trials investigating lapatinib and trastuzumab
.
Ann Oncol
2016
;
27
:
1519
25
.
4.
Berns
K
,
Horlings
HM
,
Hennessy
BT
,
Madiredjo
M
,
Hijmans
EM
,
Beelen
K
, et al
A functional genetic approach identifies the PI3K pathway as a major determinant of trastuzumab resistance in breast cancer
.
Cancer Cell
2007
;
12
:
395
402
.
5.
Baselga
J
,
Cortés
J
,
Im
SA
,
Clark
E
,
Ross
G
,
Kiermaier
A
, et al
Biomarker analyses in CLEOPATRA: a phase III, placebo-controlled study of pertuzumab in human epidermal growth factor receptor 2-positive, first-line metastatic breast cancer
.
J Clin Oncol
2014
;
32
:
3753
61
.
6.
Baselga
J
,
Phillips
GDL
,
Verma
S
,
Ro
J
,
Huober
J
,
Guardino
AE
, et al
Relationship between tumor biomarkers and efficacy in EMILIA, a phase III study of trastuzumab emtansine in HER2-positive metastatic breast cancer
.
Clin Cancer Res
2016
;
22
:
3755
63
.
7.
Loi
S
,
Michiels
S
,
Lambrechts
D
,
Fumagalli
D
,
Claes
B
,
Kellokumpu-Lehtinen
PL
, et al
Somatic mutation profiling and associations with prognosis and trastuzumab benefit in early breast cancer
.
J Natl Cancer Inst
2013
;
105
:
960
7
.
8.
Pogue-Geile
KL
,
Song
N
,
Jeong
JH
,
Gavin
PG
,
Kim
SR
,
Blackmon
NL
, et al
Intrinsic subtypes, PIK3CA mutation, and the degree of benefit from adjuvant trastuzumab in the NSABP B-31 trial
.
J Clin Oncol
2015
;
33
:
1340
7
.
9.
Chia
SKL
,
Martin
M
,
Holmes
FA
,
Ejlertsen
B
,
Delaloge
S
,
Moy
B
, et al
PIK3CA alterations and benefit with neratinib: analysis from the randomized, double-blind, placebo-controlled, phase III ExteNET trial
.
Breast Cancer Res
2019
;
21
:
39
.
10.
Prat
A
,
Carey
LA
,
Adamo
B
,
Vidal
M
,
Tabernero
J
,
Cortés
J
, et al
Molecular features and survival outcomes of the intrinsic subtypes within HER2-positive breast cancer
.
J Natl Cancer Inst
2014
;
106
:
1
8
.
11.
Cejalvo
JM
,
Pascual
T
,
Fernández-Martínez
A
,
Brasó-Maristany
F
,
Gomis
RR
,
Perou
CM
, et al
Clinical implications of the non-luminal intrinsic subtypes in hormone receptor-positive breast cancer
.
Cancer Treat Rev
2018
;
67
:
63
70
.
12.
Dieci
MV
,
Prat
A
,
Tagliafico
E
,
Paré
L
,
Ficarra
G
,
Bisagni
G
, et al
Integrated evaluation of PAM50 subtypes and immune modulation of pCR in HER2-Positive breast cancer patients treated with chemotherapy and HER2-targeted agents in the CherLOB trial
.
Ann Oncol
2016
;
27
:
1867
73
.
13.
Conte
P
,
Frassoldati
A
,
Bisagni
G
,
Brandes
AA
,
Donadio
M
,
Garrone
O
, et al
Nine weeks versus 1 year adjuvant trastuzumab in combination with chemotherapy: Final results of the phase III randomized Short-HER study
.
Ann Oncol
2018
;
29
:
2328
33
.
14.
Dieci
MV
,
Conte
P
,
Bisagni
G
,
Brandes
AA
,
Frassoldati
A
,
Cavanna
L
, et al
Association of tumor-infiltrating lymphocytes with distant disease-free survival in the ShortHER randomized adjuvant trial for patients with early HER2+ breast cancer
.
Ann Oncol
2019
;
30
:
418
23
.
15.
Geiss
GK
,
Bumgarner
RE
,
Birditt
B
,
Dahl
T
,
Dowidar
N
,
Dunaway
DL
, et al
Direct multiplexed measurement of gene expression with color-coded probe pairs
.
Nat Biotechnol
2008
;
26
:
317
25
.
16.
Parker
JS
,
Mullins
M
,
Cheang
MCU
,
Leung
S
,
Voduc
D
,
Vickery
T
, et al
Supervised risk predictor of breast cancer based on intrinsic subtypes
.
J Clin Oncol
2009
;
27
:
1160
7
.
17.
The R foundation
. 
R: The R project for statistical computing
. Available from: https://www.r-project.org/.
18.
Zardavas
D
,
Te Marvelde
L
,
Milne
RL
,
Fumagalli
D
,
Fountzilas
G
,
Kotoula
V
, et al
Tumor PIK3CA genotype and prognosis in early-stage breast cancer: a pooled analysis of individual patient data
.
J Clin Oncol
2018
;
36
:
981
90
.
19.
Goel
S
,
Krop
IE
. 
PIK3CA mutations in HER2-positive breast cancer: an ongoing conundrum
.
Ann Oncol
2016
;
27
:
1368
72
.
20.
Carey
LA
,
Berry
DA
,
Cirrincione
CT
,
Barry
WT
,
Pitcher
BN
,
Harris
LN
, et al
Molecular heterogeneity and response to neoadjuvant human epidermal growth factor receptor 2 targeting in CALGB 40601, a randomized phase III trial of paclitaxel plus trastuzumab with or without lapatinib
.
J Clin Oncol
2016
;
34
:
542
9
.
21.
Beelen
K
,
Opdam
M
,
Severson
TM
,
Koornstra
RHT
,
Vincent
AD
,
Wesseling
J
, et al
PIK3CA mutations, phosphatase and tensin homolog, human epidermal growth factor receptor 2, and insulin-like growth factor 1 receptor and adjuvant tamoxifen resistance in postmenopausal breast cancer patients
.
Breast Cancer Res
2014
;
16
:
R13
.
22.
Barok
M
,
Isola
J
,
Pályi-Krekk
Z
,
Nagy
P
,
Juhász
I
,
Vereb
G
, et al
Trastuzumab causes antibody-dependent cellular cytotoxicity-mediated growth inhibition of submacroscopic JIMT-1 breast cancer xenografts despite intrinsic drug resistance
.
Mol Cancer Ther
2007
;
6
:
2065
72
.
23.
Cescon
DW
,
Bedard
PL
. 
PIK3CA genotype and treatment decisions in human epidermal growth factor receptor 2-positive breast cancer
.
J Clin Oncol
2015
;
33
:
1318
21
.
24.
Kim
RS
,
Song
N
,
Gavin
PG
,
Salgado
R
,
Bandos
H
,
Kos
Z
, et al
Stromal tumor-infiltrating lymphocytes in NRG oncology/NSABP B-31 adjuvant trial for early-stage HER2-positive breast cancer
.
J Natl Cancer Inst
2019
;
111
:
867
71
.
25.
Matikas
A
,
Zerdes
I
,
Lövrot
J
,
Richard
F
,
Sotiriou
C
,
Bergh
J
, et al
Prognostic implications of PD-L1 expression in breast cancer: Systematic review and meta-analysis of immunohistochemistry and pooled analysis of transcriptomic data
.
Clin Cancer Res
2019
;
25
:
5717
26
.
26.
Chou
WC
,
Chen
WT
,
Hsiung
CN
,
Hu
LY
,
Yu
JC
,
Hsu
HM
, et al
B-Myb induces APOBEC3B expression leading to somatic mutation in multiple cancers
.
Sci Rep
2017
;
7
:
44089
.
27.
Chou
WC
,
Hsiung
CN
,
Chen
WT
,
Tseng
LM
,
Wang
HC
,
Chu
HW
, et al
A functional variant near XCL1 gene improves breast cancer survival via promoting cancer immunity
.
Int J Cancer
2020
;
146
:
2182
93
.
28.
Pfister
K
,
Pipka
JL
,
Chiang
C
,
Liu
Y
,
Clark
RA
,
Keller
R
, et al
Identification of drivers of aneuploidy in breast tumors
.
Cell Rep
2018
;
23
:
2758
69
.
29.
Sobral-Leite
M
,
Salomon
I
,
Opdam
M
,
Kruger
DT
,
Beelen
KJ
,
Van Der Noort
V
, et al
Cancer-immune interactions in ER-positive breast cancers: PI3K pathway alterations and tumor-infiltrating lymphocytes
.
Breast Cancer Res
2019
;
21
:
90
.