Immune microenvironment changes induced by neoadjuvant chemotherapy in triple-negative breast cancers: the MIMOSA-1 study

Neoadjuvant chemotherapy (NACT) is increasingly used to treat early-stage triple-negative breast cancers (TNBCs), and NACT response, in particular, the extent of residual disease post-NACT, yields important prognostic information [1, 2]. NACT treatment responses differentiate two groups of patients: those who achieve a pathologic complete response (the pCR group) and have a good prognosis and those with residual invasive disease (the no-pCR group) and a high risk of relapse. Patients in the no-pCR group are then either treated with adjuvant capecitabine [3] or considered for inclusion in a clinical trial.

Several TNBC studies have highlighted the association of tumor-infiltrating lymphocytes (TILs) with a good prognosis [4‐13] and with a higher pCR rate after NACT [4‐8, 13‐18]. Although TIL counts are standardized [19, 20] and routinely assessed, they currently do not contribute to treatment strategy.

In addition to increased TIL counts, the expression of immune checkpoint (ICP) programmed death ligand 1 (PD-L1) appears to be associated with a higher pCR rate [14, 21‐25] and better disease-free survival (DFS) [4]. Other novel immunological breast cancer (BrCa) targets, such as T-cell immunoglobulin and mucin domain-containing molecule 3 (TIM-3) [26, 27] and lymphocyte activation gene 3 (LAG-3) [28, 29], may yield additional prognostic information in TNBCs.

Although preclinical evidence supports that the anti-tumor activity of anthracyclines and alkylating agents may be partially mediated through the modulation of the anti-tumor immune response [30‐32], only few studies have addressed this issue in TNBCs [33]. Identifying a potential impact of NACT on the TNBC immune microenvironment (IME) may help optimize the design of future clinical adjuvant trials for no-pCR patients and may lead to the development of novel biomarkers.

Our study’s primary objective was to evaluate IME changes induced by NACT in TNBCs. The prognostic value of immune biomarkers (TILs, PD-L1, TIM-3, and LAG-3) in terms of overall survival (OS) and their association with pathological response were also evaluated as part of the secondary objectives.

Our study not only examined changes in TIL counts and PD-L1 expression levels in 66 TNBC patients, before and after NACT, but is also, to the best of our knowledge, the first study to evaluate how changes in TIM-3 and LAG-3 expression levels correlate with pCR and OS.

Our study detected an overall cohort median TIL count of 10% at baseline, with 15.2% of TNBCs exhibiting greater than 30% TILs. These results are consistent with several reports in the literature [12‐14], albeit that studies in the literature report higher median TIL counts [5, 7].

PD-L1 was expressed in 53% of cases, which is comparable to other reports which use the Ventana SP142 antibody [34, 35]. In our cohort, 32% and 77% of tumors expressed TIM-3 and LAG-3, respectively. Results obtained by Burugu et al. for TIM-3 (28% of TIM-3 positivity) are similar to ours [26]. However, the LAG-3 results reported by Burugu et al. and Bottai et al. (33% and 18%, respectively) differ from our results [28, 29]. This may be attributed to differences in the LAG-3 quantitation methodologies between studies. However, these differences may also reflect the very heterogeneous nature of the immune microenvironment in TNBCs which may support a rationale for immune checkpoint inhibitor and cytotoxic combination treatment to attempt to increase the pCR rate in this patient group [34‐37].

TIL counts before NACT are both a strong prognostic factor for OS [4‐13] and an established predictor of pCR [4‐8, 13‐18], which supports the notion that chemotherapy responses are at least partially immune mediated [30‐33]. Our study did not detect any significant increases in baseline TIL counts of pCR patients, which may be attributed to a lack of power.

We report a trend of PD-L1 positivity before NACT correlating with an improved prognosis, but there was no association with pCR, perhaps due to our small sample size. This is in contrast to several other studies that report a significant association of positive PD-L1 staining with DFS [4] and pCR [14, 21‐25].

Importantly, we show that positive baseline staining for LAG-3 but not for TIM-3 was associated with an increased probability of achieving pCR. We are, to the best of our knowledge, the first to report this finding. Larger studies will be required to confirm our observation.

Although several studies have previously reported changes in TIL counts before and after NACT and the prognostic significance of such changes [4, 5, 7, 12, 14], most of these studies did not specifically focus on TNBCs [12, 14]. The current literature therefore provides no definitive results regarding the prognostic value of TIL count changes in TNBCs. We report that a decrease in TIL count was significantly more frequently associated with pCR and conversely, that an increase in TIL count was more often associated with residual disease (Suppl. Table 3). We found that the number of TILs before (but not after) NACT was a significant prognostic marker of 3yr-OS, an association that persisted (close to significance) even when the pCR group was considered independently (data not shown). Our findings are consistent with observations reported by Castaneda et al. in 98 TNBC patients, of which 30% achieved pCR after NACT [7]. In a cohort of 72 TNBC patients, Dieci et al. observed an increase in TILs in residual disease after NACT which was associated with improved disease-free survival [4]. In a cohort of 104 TNBC patients, Lee et al. reported that compared to stable cases, a significant change, either a TIL count increase or decrease, between the pre-NACT biopsy and post-NACT residual tumor tissue sample, was associated with a better prognosis [5]. Despite the statistical significance of the association, this correlation may be attributed to the overrepresentation of cases with an increased post-NACT TIL count in the Lee et al. cohort. These data suggest that patients with increased TIL counts who do not achieve pCR after NACT may benefit from adjuvant ICP inhibitor therapy. Nonetheless, a better characterization of TIL levels is required to optimize therapeutic strategies in this patient group.

With regard to PD-L1 expression changes in TNBCs after NACT, Dieci et al. found a significant increase of PD-L1 expression in residual disease which was associated with an improved DFS of borderline statistical significance [4]. Pelekanou et al. detected decreased PD-L1 expression in a cohort encompassing all BrCa phenotypes and with no-pCR [12]. In another study, the same authors found that PD-L1 expression was not altered after NACT, but observed a 15% mean decrease in TILs [14]. We found no significant change in PD-L1 expression after NACT in our entire cohort, but an increase in PD-L1 expression in the no-pCR group. We also observed that PD-L1 positivity after NACT was significantly associated with poor OS in the entire cohort (p = 0.0055) and a trend of poor OS in the no-pCR group (p = 0.1186). The lack of standardized PD-L1 protocols makes it difficult to draw any conclusions from these studies and it is quite possible that the anti-PD-L1 VENTANA SP142 antibody we used identifies fewer PD-L1-positive patients [38].

Importantly, our study is the first to report changes in TIM-3 and LAG-3 expression levels before and after NACT in TNBCs and to determine their predictive and prognostic value. We found a significant increase in TIM-3 expression after NACT and strong TIM-3 expression post-NACT to be significant poor prognostic indicators in the entire population (p = 0.0274) and showed a similar trend in the no-pCR group (p = 0.0702). Interestingly, in the overall population as well as the no-pCR group, patients whose residual disease expressed both PD-L1 and TIM-3 had a worse prognosis (3yr-OS 45.5%).

These observations suggest that TNBCs remain immunogenic after NACT and may continue to be subject to antitumor immuno-surveillance, as evidenced by the increased infiltration of tumor lymphocytes in the no-pCR patient group. This hypothesis provides a rationale to explore reactivation of lymphocytes present in residual disease using adjuvant immunotherapy, such as in the current anti-PD-L1 antibody trial (NCT02954874). In light of strong TIM-3 expression being predictive of a good anti-PD-(L)1 immunotherapy response in a cohort of 30 head and neck cancers patients [39] and of anti-TIM-3 antibodies being evaluated alone or in combination with anti-PD-(L)1 therapy in several phase 1 trials [40], our own findings, should they be confirmed by larger multicentric studies, support the rational for testing immunotherapy adjuvant treatments specifically in patients that fail to achieve a pCR.

Our study is nevertheless limited by its small size and monocentric nature which reduces its statistical power and also limited more substantial subgroup and multivariate analyses. Moreover, relevant threshold for TIM-3 expression should be clarified in future studies.

In conclusion, TNBCs have heterogeneous IMEs which are profoundly altered by NACT. After NACT, patients positive for both PD-L1 and TIM-3 had significantly worse prognoses and importantly this correlation persisted in the subgroup of patients that failed to achieve a pCR.