Background

Hormone receptor-positive(HR+)/HER2-negative (HER2−) breast cancer (BC) encompasses almost two-thirds of all breast tumours.1 2 It is now widely accepted that HR+/HER2− BC represents a highly heterogeneous disease, where the clinical complexity reflects an even greater—and only partially uncovered—complexity at a molecular and biological level. Indeed, in the past years, major efforts have been made to unravel the mechanisms underlying different clinical behaviours in terms of prognosis and treatment response for apparently clinicopathologically similar diseases,3 with PI3K/PTEN/mTOR pathway alterations emerging as one of the possible players within this complex landscape.4 5 Moreover, the role of this major intracellular network has gained increasingly appeal in the light of the progressive availability of treatment strategies selectively targeting and inhibiting this pathway.6–9 Within this framework, available evidence is consistent in highlighting a crucial role of PI3K/PTEN/mTOR pathway in shaping the resistance to endocrine-based treatments, mostly in the advanced setting. In the early setting (including locally advanced BC, LABC), accumulating evidence suggests PI3K-mediated signalling as possibly implicated in both chemoresistance10 11 and endocrine resistance,10–13 with conflicting data regarding the prognostic impact.5 14 However, a comprehensive assessment of the dynamic behaviour of PIK3CA mutations and its major downstream effectors in HR+/HER2− early/LABC (EBC/LABC) is lacking.

Neoadjuvant treatment, besides the undeniable benefits from a purely clinical perspective, represents a strategical research platform,15 offering a unique translational window of opportunity, allowing to dynamically assess tissue biomarkers immediately before and after the exposure to systemic treatment. In HR+/HER2− EBC/LABC, neoadjuvant chemotherapy (NACT) represents the standard of care in case of locally advanced/large tumours or when optimal surgery is not feasible upfront.16

The main aim of the present work was to longitudinally track PI3K/PTEN/mTOR pathway dynamic behaviour from baseline biopsy to residual disease (RD) and metastases in case of relapse, and assess its prognostic value in HR+/HER2− BC patients undergoing NACT.

ResultsPatient characteristics

Ninety-two patients with HR+/HER2− BC failing to achieve pCR after NACT were included. Clinicopathological features of the overall cohort are shown in table 1.

Table 1

Clinicopathological features of the overall cohort

Briefly, median age at diagnosis was 51 years. Approximately half of the patients were premenopausal. The majority of patients had ductal histology, grade 3 and clinical stage III tumours. In the great majority of cases, NACT consisted of anthracycline-taxane-based treatment. In addition, almost a half of patients exhibited RCB class III aftyer NACT. Almost all patients received adjuvant endocrine therapy, consisting on tamoxifen in one-third of cases and aromatase inhibitor in two third of cases. Finally, less than 25% of patients underwent further chemotherapy in the adjuvant setting.

Almost a half of the patients experienced disease relapse, mostly consisting of distant relapse.

PIK3CA mutational status at baseline was available for 74 patients. A PI3KCA mutation was detected in 25.7% of cases (n=19/74). The spectrum of PIK3CA mutation type is reported in table 2.

Table 2

Spectrum of PIK3CA mutation type

PIK3CA mutational status at baseline was significantly associated with tumour grade and proliferation index, with PIK3CA mutated cases being enriched for G2 (p=0.017) and less proliferating (lower ki67) tumours (p=0.017).

Pten loss status at baseline was available for 57 cases. Loss of Pten was detected in 40.3% of cases (n=23). mTOR, 70SK6 and 4EBP1 phosphorylation levels at baseline were available for 49, 50 and 55 cases, respectively. mTOR dephosphorylation status was significantly associated with lobular histology (p=0.045), with a similar association of borderline statistical significance observed for dephosphorylated-4EBP1 (p=0.054).

PIK3CA mutational status on RD was available for 87 patients. A PI3KCA mutation was detected in 25.2% of cases (n=22/87). The spectrum of PIK3CA mutation type is reported in table 2. Two patients harboured a double PIK3CA mutation.

Pten loss status on RD was available for 89 cases. Loss of Pten was detected in 69.7% of cases (n=62/89).

mTOR, 70SK6 and 4EBP1 phosphorylation levels on RD were available for 88, 87 and 88 cases, respectively.

PI3K/PTEN/mTOR pathway dynamics

PIK3CA mutational status evolution from baseline biopsy to RD is shown in figure 1 (n=70). In particular, 5/53 (total: 7.1%) and 4/17 (total: 5.7%) patients gained and lost PIK3CA mutation, respectively, with a total conversion rate of 12.8%. In addition, two patients harbouring a PIK3CA mutation at baseline (one patient with H1047x and one patient with E545x mutation), acquired a second mutation on RD after NACT (C420R and H1047x, respectively). No statistically significant association was observed between the acquisition of double PIK3CA mutation and clinicopathological features.

Figure 1

PIK3CA mutational status evolution from baseline biopsy to RD to metastatic tissue (in case of disease relpase). RD, residual disease.

We tracked PIK3CA status on metastatic tissue in patients experiencing disease relapse (n=12), as shown in figure 1. In particular, 1/12 (8.3%) patient gained a PIK3CA mutation (C420R), while no patients experienced PIK3CA loss. The association between PIK3CA mutational status and the metastatic site and the distribution of PIK3CA mutations across different metastatic sites are shown in online supplemental table 1.

Regarding Pten status evolution from baseline biopsy to RD (n=57), 2/23 patients (total: 3.5%) lost Pten expression, while 17/34 patients (total: 29.8%) showing Pten-loss at baseline gained Pten expression on RD, with a total rate of Pten instability of 33.3%.

mTOR phosphorylation levels significantly increased from baseline biopsy to RD (median ph-mTOR delta=+30, p<0.001), while both its substrates significantly decreased (median ph-70S6k delta=−47, p<0.001; median ph-4EBP1 delta=−44, p<0.001).

During NACT, phosphorylation dynamics of mTOR and 4EBP1 were found to be independent from PIK3CA mutational status, while phosphorylation dynamics of 70S6K showed an association with PIK3CA mutational status, with a greater decrease of 70S6K phosphorylation levels in PIK3CA mutated tumours (p=0.05). We then deepened 70S6K phosphorylation level dynamics according to PIK3CA evolution under NACT exposure and we found that both the subgroup of patients with concordant mutated PIK3CA status and that of patients gaining PIK3CA mutation on RD showed significantly greater magnitude of 70S6K phosphorylation level reduction after NACT as compared with those losing PIK3CA mutation or with concordant PIK3CA wild-type status (p=0.001), as shown in figure 2.

Figure 2

70S6K phosphorylation level dynamics according to PIK3CA mutational status evolution from baseline biopsy to RD. RD, residual disease WT, wild-type.

No association was found between phosphorylation dynamics of mTOR, 4EBP1 and 70S6K and Pten loss.

Correlation between PI3K/PTEN/mTOR alterations

At baseline, PI3K/PTEN/mTOR alterations were not significantly correlated with each other, with the exception of a moderate positive correlation between PIK3CA mutation and ph-70S6K (Spearman coefficient 0.366, p=0.05).

On RD, ph-70S6K was moderately negatively correlated with Pten-loss (Spearman coefficient −0.331, p=0.002) and weakly positively correlated with ph-4EBP1 (Spearman coefficient 0.275, p=0.10).

Finally, delta ph-70S6K and delta ph-4EBP1 were moderately positively correlated with each other (Spearman’s coefficient 0.476, p<0.001).

Prognostic role of PI3K/PTEN/mTOR pathway alterations

The median follow-up of the present patient cohort was 9.3 years.

At baseline, PIK3CA mutational status was not significantly associated with RFS.

At baseline, ph-mTOR showed a negative association of borderline significance with RFS (HR 2.39, 95% CI 0.94 to 6.02, p=0.064) (figure 3A). When testing ph-mTOR prognostic association according to PIK3CA mutational status, ph-mTOR negative prognostic impact was statistically significant in patients with PIK3CA wild-type status (HR 3.78 95% CI 11.67 to 12.23, p=0021) (figure 3B), while no prognostic association was seen in patients harbouring a PIK3CA mutation at baseline.

Figure 3

Kaplan-Meier curves of RFS according to mTOR phosphorylation status. (A) Overall; (B) PIK3CA-wt cohort. RFS, recurrence-free survival.

At baseline, ph-70S6K was significantly and positively associated with RFS (HR 0.311, 95% CI 0.11 to 0.87, p=0.026) (figure 4).

Figure 4

Kaplan-Meier curves of RFS according to 70S6K phosphorylation status. RFS, recurrence-free survival.

When assessing the prognostic impact of ph-70S6K according to upstream players, the prognostic impact appeared independent from PIK3CA mutational status.

At baseline, 4EBP1 phosphorylation status was not significantly associated with RFS overall nor in patients stratified according to mTOR phosphorylation status.

On RD, none of the PIK3CA/PTEN/mTOR alterations was significantly associated with RFS.

Discussion

In the present study, we dynamically/longitudinally tracked PI3K/PTEN/mTOR pathway alterations on tumour tissue and assessed their prognostic role in HR+/HER2− BC patients with RD after NACT.

In our cohort, the rate of PIK3CA mutation was slightly more than 25% both at baseline and on RD, which was relatively lower than expected based on previous studies. Indeed, the rate of PIK3CA mutation in BC patients has been reported as ranging from ≈30% to 60% across studies. In addition, a negative association between the presence of PIK3CA mutation at baseline and pCR after NACT has been consistently reported.10 11 17–19 For this reason, being our population entirely represented by patients with RD after NACT, we would have expected an enrichment for PIK3CA mutated cases. However, it should be acknowledged that data regarding PIK3CA prevalence in high-risk EBC/LABC are limited as compared with those gathered in the advanced setting, and mostly derived from small retrospective series, thus making cross-study comparisons of limited value.

Interestingly, we found two patients harbouring a N345K PIK3CA mutation, which would have been missed by applying the FDA-approved companion diagnostic Therascreen-PIK3CA-RGQ PCR test. This observation adds to the accumulating evidence, which overall highlights a suboptimal coverage of the Therascreen panel in terms of PIK3CA mutational status detection.20 21 In this context, the N345x mutation represents the most prevalent among the Therascreen-neglected PIK3CA alterations.20 21 Within this framework, if from one hand the implementation of more comprehensive PIK3CA panels is warranted, on the other it is imperative to make additional research efforts aiming at unravelling the heterogeneity of PIK3CA mutational landscape.

Regarding Pten expression, we detected Pten loss in more than 40% of patients at baseline and almost 70% of cases on RD, thus adding to available evidence overall suggesting Pten loss as a relatively frequent phenomenon in BC patients.22

Regarding the clinicopathological landscape of PIK3/PTEN/mTOR pathway alterations, PIK3CA mutational status was significantly associated with features of restricted biological aggressiveness, consistently with previous findings.23 Another intriguing observation was the significant association between dephosphorylation (as a surrogate of inactivated status) of both mTOR and its substrate 4EBP1, and lobular histology, which appears consistent with a previous report where lobular cancers were associated with significantly lower frequency of phosphorylated-mTOR as compared with ductal tumours.24

The main aim of the present work was to evaluate PI3K/PTEN/mTOR pathway dynamics during NACT. Regarding PIK3CA mutational landscape evolution, we observed a 12.8% rate of instability from baseline biopsy to matched samples of RD after NACT, with 7% of patients with wild-type PIK3CA at baseline gaining PIK3CA mutation on RD and 26% of PIK3CA-mutated patients at baseline exhibiting PIK3CA loss.

Notably, we also tracked PIK3CA mutational landscape in the subgroup of patients experiencing disease relapse and undergoing metastatic tissue resampling, with only one patient gaining PIK3CA mutation, with no PIK3CA loss events observed. These observations raise two orders of considerations. First, the relatively low rate of PIK3CA mutational shift under NACT exposure (and at disease relapse) solidifies data, mostly gathered in the advanced setting, overall highlighting the propensity of PIK3CA mutational phenotype to be clonally dominant during BC evolution.25 Second, we confirmed in a homogenous population of HR+/HER2− BC mostly treated with anthracycline-taxane-based NACT, that, of patients showing PIK3CA mutational status instability, PIK3CA mutation loss among PIK3CA mutated patients was more frequent than PIK3CA mutation gain among PIK3CA wild-type patients, as suggested in prior studies conducted in unselected BC patients, receiving heterogeneous NACT regimens.11 26

Notably, in our study, two patients with PIK3CA single mutation at baseline, acquired a second PIK3CA mutation on RD. Although the acquisition of double PIK3CA mutations has been already described in previous studies,27 28 the actual biological and clinical value of such phenomenon is not fully understood. A multihistology study reported that double PIK3CA mutations may determine an enhanced activation of PI3K pathway as compared with single PIK3CA mutations, causing increased PIK3CA-driven cell proliferation and tumour growth.27 Interestingly, the direct consequence of such phenomenon may be the association between multiple PIK3CA mutations, enhanced PIK3CA oncogene addiction and subsequent greater sensitivity to PI3K inhibition in HR+/HER2− advanced BC patients receiving fulvestrant+placebo/taselisib (PI3K inhibitor) in the context of the SANDPIPER phase III trial.27 In our study, the presence of double PIK3CA mutations did not appear as a parental phenomenon but rather an acquired one, triggered by NACT exposure. Although the small number of patients with double PIK3CA mutations precluded the possibility to deepen this association and to properly assess the role of chemotherapy in enhancing this phenomenon, it may be speculated that patients acquiring a second PIK3CA mutational event after NACT may represent ideal candidates for postneoadjuvant treatment with PI3K inhibitors+endocrine therapy given the potential enhanced sensitivity to PI3K-inhibition.

When exploring Pten landscape dynamics, we observed a distinct behaviour as compared with PIK3CA, both in terms of higher instability and the direction of such instability, with a substantially higher tendency to Pten loss rather than Pten restoration after NACT. It should, however, be acknowledged that Pten expression evaluation may be limited in scope by the lack of consistency and reproducibility of Pten assessment across studies, with regard to type of assay, scoring system, cutoffs and interpathologists’ agreement.22

Interestingly, we also observed a significant increase of mTOR functional status after NACT, while that of its substrates (4EBP1 and 70S6K) significantly decreased. Interestingly, the magnitude of decrease of 70S6K phosphorylation levels from baseline to RD appeared to differ according to PIK3CA mutational status evolution, with a significantly greater decrease in patients either maintaining or gaining PIK3CA mutation during NACT as compared with those losing PIK3CA mutation or with concordant PIK3CA wild-type status. This observation may suggest that NACT-induced impairment of 70S6K functional status may be prerogative of tumours with intrinsically PIK3CA-mutated tumours or made them so under NACT exposure. PI3K/PETN/mTOR functional status in EBC/LABC and its evolution during NACT represents an unexplored phenomenon so far. Although our findings need to be interpreted as only descriptive, they generate the hypothesis that PI3K-related pathway is highly dynamic under NACT exposure, possibly affected by multiple cross-talking pathways, with—so far—unpredictable biological consequences, and such phenomenon may not be properly recapitulated or surrogated by only focusing on PIK3CA mutational status. This consideration finds a possible corroboration from the observation of lower/weaker than expected correlations between the different players of the PI3K/PTEN/mTOR pathway. In particular, at baseline, PIK3CA was positively and weakly associated with 70S6K phosphorylated status, thus indirectly implying the engagement of mTOR. On RD, under the effect of NACT exposure, while the two classical mTOR substrates (4EBP1 and 70S6K) were positively correlated with each other, 70S6K was no longer correlated with PIK3CA mutational status. These findings may potentially imply that, while in treatment-naïve HR+/HER2− EBC/LABC the biological hubs represented by PI3K and mTOR may work, at least in part, as a series circuit, under the exposure of NACT they seem to rather function as parallel circuits, disengaging from the mutual biological dependence. Another piece of the puzzle is the observation of an opposite dynamic behaviour of mTOR and its substrates after NACT exposure. This observation may suggest that NACT-induced mTOR activation may be driven by alternative pathways, both upstream and downstream, with mTOR relying on alternative triggers (PI3K-independent) and substrates (different from 70S6K and 4EBP1) to mediate its action. This hypothesis may be further strengthened by our exploratory survival analyses, which revealed appealing hints. In particular, we observed baseline phosphorylated-mTOR being significantly associated with poorer RFS and this negative prognostic impact seemed to be restricted to patients with PIK3CA wild-type status. This observation, although exploratory, may suggest that in HR+/HER2− BC patients with RD after standard NACT, the activation of mTOR may be clinically relevant in terms of prognostic impact only when it reflects PIK3CA-independent mechanisms. Conversely, we observed a positive prognostic impact of phosphorylated-70S6K, which was independent from PIK3CA mutational status. Although the opposite prognostic value of mTOR and its classical substrate 70S6K may appear counterintuitive, it might further underly the complexity of PI3K/PTEN/mTOR interaction and crosstalk with other pathways. Of course, it should be acknowledged that the interpretation of our survival data may be complicated by the fact that almost all patients received adjuvant endocrine therapy and no definitive and solid conclusions may be drawn in this regard.

The present work has several strengths. First, this represents one of the first attempts to uncover the role of PI3K/PTEN/mTOR dynamic behaviour under NACT exposure in HR+/HER2− BC patients with evidence of RD at surgery. In addition, the majority of patients received anthracycline-taxane-based NACT and subsequent adjuvant endocrine therapy, thus representing a homogenously treated population. In addition, the long follow-up period (>9 years) allowed us to explore the long-term prognostic value of PI3K pathway-related biomarkers, thus increasing the reliability of our exploratory survival analyses, if considering that HR+/HER2− BC recurrence pattern is typically characterised by a sustained risk even after several years from diagnosis.29–31

Some limitations should be acknowledged as well. In particular, the retrospective nature may have been responsible for selection bias. In addition, only patients with RD after NACT were included, thus precluding the possibility to assess the role of baseline PI3K/PTEN/mTOR pathway alterations in patients achieving pCR. Moreover, none of the patients included received adjuvant abemaciclib as treatment escalation, thus making our patient population not entirely representative of the contemporary therapeutic landscape of high-risk HR+/HER2 EBC/LABC patients. Finally, although in our study, PIK3CA mutational status tended to be stable under NACT exposure, we cannot exclude that the potential suboptimality of core biopsy-based sampling in capturing tumour heterogeneity, could have been responsible for hindering the presence of a PIK3CA mutation at baseline among the five patients exhibiting PIK3CA mutation gain on RD. However, it has been reported that the genotypic intratumoural heterogeneity of PIK3CA mutations within the primary tumour is rare,32 thus reassuring on the reliability of core biopsy in the assessment of PIK3CA mutational status within primary BC.

In conclusion, we observed in a population of HR+/HER2− BC with RD after NACT that PI3K/PTEN/mTOR pathway is highly dynamic under NACT exposure with PIK3CA-dependent signalling possibly capturing only a small fraction of such complexity. In this context, mTOR activation trough alternative pathways may have a crucial role in shaping the molecular landscape of this challenging patient population. Within this framework, it is imperative to further elucidate the role of PIK3CA and mTOR-dependent pathways in shaping chemoresistance and endocrine resistance in order to improve therapeutic selection and prognostic stratification of high-risk HR+/HER2− EBC/LABC patients.