Breast cancer (BC) is not traditionally considered an immunogenic tumor, especially if compared with other solid tumors as melanoma or lung cancer [1]. However, a rapidly growing body of data suggests a substantial heterogeneity in terms of immunogenicity among different BC subtypes, with hormone receptor-negative (HR-) BC being generally more immunogenic than HR-positive (HR+) tumors. Triple-negative (TN) and HER2-positive (HER2+) subtypes have been consistently reported to show higher levels of tumor infiltrating lymphocytes (TILs) as compared to the HR+/HER2-negative (HER2-) subgroup [2]. In addition, it has been shown that the presence of TILs may be capable of predicting outcome and response to treatment in early TN and HER2+ diseases, thus underlying that the biology of these aggressive BC subtypes may be deeply affected by the immune counterpart of tumor microenvironment [3–6].

Triple-negative BC (TNBC) is characterized by high genomic instability and mutational load, both of which contribute to the generation of a great amount of neoantigens, ultimately leading to increased tumor immunogenicity [7]. HER2+ disease is similarly highly immunogenic given the high proliferation activity and the overexpression of HER2, which acts as a neoantigen itself [8]. In addition, anti-HER2 monoclonal antibodies are known to mainly exert their action by activating the immune system through the mechanism of antibody-dependent cell mediated cytotoxicity (ADCC), thus enhancing the immunogenicity of this BC subtype [9, 10].

As far as HR+/HER2- BC is concerned, available evidence-although scattered-suggests that this BC subtype establishes a complex relationship with the immune system. In this context the interplay between estrogens, HR+ cancer cells, inflammatory mediators and endocrine therapies accounts, at least in part, for such complexity [11].

Based on these premises, clinical trials testing immune checkpoint inhibitors (ICIs) in BC patients, mainly focused on TN and HER2+ subtypes, have been conducted or are ongoing. Indeed, the combination of the ICI atezolizumab with chemotherapy represents the first immunotherapeutic strategy gaining the Food and Drug Administration (FDA)/European Medicines Agency (EMA) approval for the management of TN advanced BC, and other similar combinations are expected to be approved in the near future in the early setting [12]. In this context, data on potential predictive biomarkers have been generated. Notwithstanding, we currently lack a comprehensive biomarker capable of reliably selecting patients more likely to respond.

In the present review we summarized the state of art of immunotherapy for BC treatment both in early and advanced setting, focusing on activity and efficacy results of ICIs coming from clinical trials. In addition, we discussed factors possibly accounting for different results of studies conducted in similar settings, with particular emphasis on biomarker analyses.

MBC remains to date an incurable disease, whose prognosis is closely related to different factors, such as biological subtype, organ involvement, treatments options. The role of ICIs in metastatic BC is still under study, with many hopes and few sureness (Table 2).

Among different BC subtypes, TNBC has the worst prognosis and limited treatment options, mainly based on chemotherapy. On the other hand, the greater sTIL and iTIL levels and the higher TMB and PD-L1 expression, make it a particularly suitable candidate for immunotherapy.

The results of initial trials testing ICIs as single agent for TN MBC showed modest ORR that were higher in the first line setting (up to 24%) as compared to patients who received immunotherapy in later lines [42–45].

A second category of early studies evaluated combinations of chemotherapy and ICIs in advanced TNBC. The combination of atezolizumab + NabP led to 39% of overall responses, that were higher in patients treated in the first line (54%) and in PD-L1 positive patients (42%) [46].

The combination of pembrolizumab and eribulin-mesylate induced a 26% rate of overall response, which again was numerically higher in the first-line cohort [29% vs. the second/third-line cohort (22%)] [47]. Overall, the two key messages emerging from these early studies were: patients in the first-line setting may be more sensitive to immunotherapy-based treatment, and combination with chemotherapy may enhance the efficacy of immunotherapy. As a further confirmation of these early findings, the KEYNOTE-119 trial, that randomized TNBC patients previously treated with 1-2 prior treatments for metastatic disease to receive pembrolizumab vs. chemotherapy of physician’s choice, did not demonstrate any significant difference in OS between the two arms in patients with CPS > 10 or > 1 or in all patients, although the effect of pembrolizumab treatment was larger as the level of PD-L1 expression increased [48].

The TONIC trial evaluated different induction strategies to enhance the efficacy of the anti-PD1 Nivolumab given as single-agent in TN advanced BC patients. This trial enrolled patients who received no more than 3 lines for advanced disease. Patients were randomized to receive different 2-weeks induction treatments (irradiation, CTX, CDDP, doxorubicin or control) followed by nivolumab. The results showed that the cohort of patients who received induction with doxorubicin achieved the highest overall response rate (35%), confirming that chemotherapy may be able to convert a cold microenvironment into a hot one, more susceptible to the mechanism of action of immunotherapy and that anthracyclines, by inducing immunogenic cell death, are optimal candidate drugs in this context [21].

Atezolizumab was the first ICI to be approved for the treatment of MBC based on the IMpassion130 trial results. This trial enrolled about 900 patients with metastatic TNBC disease who received no prior therapy for advanced disease (prior chemotherapy for early BC was allowed if relapse occurred at least 12 months after treatment conclusion). Patients were randomized to receive NabP plus atezolizumab or NabP plus placebo. The association of the PD-L1 antibody atezolizumab with NabP significantly improved PFS in the intention-to-treat population: median PFS 7.2 months vs. 5.5 months (HR 0.80, 95% CI 0.69–0.92, P = 0.002). Among patients with PD-L1 positive tumors (at least 1% of total tumor area occupied by positive immune cells by using the Ventana SP142 IHC assay), the median PFS was 7.5 months and 5.0 months, respectively (HR 0.62, 95% CI 0.49–0.78, P < 0.001) [12]. At the last interim analysis, no significant difference in OS was observed in the intention-to-treat population between the two arms, whereas a 7-months improvement in median OS was observed in PD-L1 positive patients treated with atezolizumab plus NabP (25.0 months) vs. placebo plus NabP (18.0 months, HR 0.71, 95% CI 0.54–0.93, statistical significance was not formally tested according to the hierarchical statistical plan) [49]. Toxicities were acceptable and consistent with the expected safety profiles of each agent and previous studies. In the experimental arm, 29 (6.4%) patients had adverse events that led to the discontinuation while fatal adverse events occurred in 6 (1.3%) cases. IMpassion130 has been a practice changing study establishing the new standard of care for previously untreated PD-L1 positive TN MBC patients. This first-line combination is approved by FDA and EMA in TNBC patients who present a PD-L1 expression > 1% on immune cells.

At the recent European Society for Medical Oncology (ESMO) Virtual Congress 2020, results from the phase III IMpassion131 were presented. This is the IMpassion130 “twin study” in which authors aimed to evaluate, in a similar population, the role of backbone chemotherapy in association with atezolizumab. The trial did not meet its primary endpoint, where atezolizumab plus paclitaxel as first line treatment failed to improve outcomes in PD-L1 positive MBC patients. In detail, PFS was not significantly improved by atezolizumab plus paclitaxel vs. paclitaxel alone in both PD-L1 positive (6.0 months vs. 5.7 months, HR = 0.82, P = 0.20) and ITT population (5.7 months vs. 5.6 months, HR = 0.86). The combination also did not improve OS in PD-L1 positive group, with even a trend towards decreased OS (22.1 months vs. 28.3 months, HR = 1.12; ITT: 19.2 months vs. 22.8 months, HR = 1.11), but moderately affected RR (63.4% vs. 55.4% among PD-L1 positive patients and 53.6% and 47.5% in the overall population). Although the reasons for these divergent results from IMpassion130 and IMpassion131 are largely unknown, a possible explanation may rely on the larger use of steroid premedication in IMpassion131 as compared to IMpassion130 trial. Notably, when looking at trial population composition and features, no other relevant differences in terms of sample size, DFIs, number of metastatic sites, prior chemotherapy for EBC and/or ABC, or proportion of PD-L1 positive patients were observed. Finally, PD-L1 may be considered a suboptimal predictive factor in this setting, thus not probably representing the ideal biomarker to select patients for ICIs. Overall considered, these data underline the need to better understand the interactions cancer-immune system, the more suitable combination between ICI and backbone chemotherapy [50].

Interesting finding emerged from the KEYNOTE-355, a phase III trial investigating the combination of another ICI, pembrolizumab, plus chemotherapy (NabP; paclitaxel; or gemcitabine/carboplatin) vs. placebo plus chemotherapy in TN MBC patients. According to the inclusion criteria, patients should have not received any prior treatment for metastatic disease and relapse should have occurred at least 6 months after the conclusion of chemotherapy for early BC. Eight hundred forty-seven patients were randomized 2:1; the primary endpoint was PFS in PD-L1 positive patients (CPS > 10 and > 1, assessed with 22C3 pharmDx assay) and in the intention-to-treat population, to be tested according to a hierarchical statistical design. Notable, PD-L1 status was assessed using a different ICH test compared to IMpassion130-131, where Ventana SP142 was employed, and evaluated not only in the immune compartment but also in tumor cells (CPS). With regards to the first PFS endpoint, the statistical assumption was met and pembrolizumab plus chemotherapy significantly improved PFS vs. chemotherapy plus placebo in pts with CPS ≥ 10 tumors (median 9.7 months vs. 5.6 months, HR 0.65, 95% CI 0.49–0.86, P = 0.0012). No statistically significant benefit of pembrolizumab + chemotherapy was observed in patients with CPS ≥ 1 tumor (median PFS 7.6 months vs. 5.6 months, HR 0.74, 95% CI 0.61–0.90, P = 0.0014, with a pre-specified P value boundary of 0.00111). According to the statistical plan, PFS analysis in the intention-to-treat population was only descriptive and reported a median PFS of 7.5 with pembrolizumab vs. 5.6 with placebo [51].

Although the best results with ICIs have been achieved in TNBC, significant evidence of immunogenicity have been proven also in HER2+ MBC. As already mentioned, the hypothesis is that ICIs and anti-HER2 agents could be synergic and that their combination could overcome potential resistance. The phase Ib/II PANACEA trial tested the efficacy and safety of pembrolizumab in combination with trastuzumab, in advanced HER2+ trastuzumab-resistant MBC. In PD-L1 positive tumors the combination was active and safe with a durable clinical benefit. In this cohort an ORR of 15% and a disease control rate (DCR) of 25% were detected. Activity was higher in the ≥ 5% TILs subgroup with ORR and DCR of 39% and 47%, respectively. No objective responses were observed in the PD-L1 negative cohort. PANACEA study provides proof of principle evidence that immunotherapy can reverse trastuzumab resistance in metastatic HER2+ BC [35].

The recent KATE2 study evaluated the addition of atezolizumab to T-DM1 in 202 patients with locally advanced or metastatic HER2+ BC patients priorly treated with trastuzumab and taxane-based therapy. Overall median PFS was 8.2 months in the atezolizumab group vs. 6.8 months in the placebo group with no statistically significant difference (HR 0.82, 95% CI 0.55–1.23, P = 0.33). In the intention-to-treat population, 1-year OS was similar in both arms. In the PD-L1 positive subgroup, both PFS and OS were numerically prolonged in the atezolizumab/T-DM1 arm vs. placebo/T-DM1 arm: median PFS was 8.5 months vs. 4.1 months (HR 0.60, 95% CI 0.32–1.11), respectively, and 1-year OS was 94.3% vs. 87.9% (HR 0.55, 95% CI 0.22–1.38), respectively. Patients with PD-L1 positive tumor showed significantly higher TILs levels and the hazard ratio showed a larger benefit from atezolizumab/T-DM1 vs. placebo/T-DM1 in patients with TILs > 5% (HR 0.55, 95% CI 0.26–1.12) as compared to patients with lower TILs (HR 1.43, 95% CI 0.51–4.01). However, based on the small sample size and limited follow up these results warrant further evaluation in larger trials [37].

A plethora of other trials focused on anti PD-1/PD-L1/CTLA-4 antibodies in combination with anti-HER2 agents are ongoing and their results are awaited [52].

In HR+ MBC, the role of immunotherapy seems to be less promising and to date has not been largely explored.

In the multicohort phase Ib KEYNOTE-028 trial, pembrolizumab was investigated in 25 heavily pretreated PD-L1 positive HR+/HER2- MBC; ORR was 12% (95% CI 2.5–31.2%) with 12 months as median duration of response (range: 7.4–15.9 months). Overall, therapy was well tolerated but with modest response, even if durable in a subgroup of patients [53].

A phase II study randomizing HR+/HER2- heavily pre-treated MBC patients (≥ 2 previous lines of hormonal therapy and 0–2 previous lines of chemotherapy) to receive either the combination of eribulin-mesylate plus pembrolizumab (arm A) or eribulin-mesylate alone (arm B), reported no significant differences in terms of PFS (primary endpoint) and ORR (secondary endpoint). In particular, median PFS and ORR were 4.1 months vs. 4.2 months (P = 0.38), and 25% vs. 34% (P = 0.49), for arm A vs. arm B, respectively. Translational analyses failed to report a predictive role for PD-L1 status (PD-L1 positive tumors: > 1% with IHC 22C3 clone assay) TILs (cutoff 10%), TMB (cutoff 6), neutrophilic-lymphocyte ratio (cutoff 4) or genomic alterations with respect to immunotherapy efficacy [54].

Besides chemotherapy and anti-HER2 agents, there is growing interest in the identification of ideal partners for immune checkpoint blockade, with the aim of enhancing MBC patients’ prognosis. Many clinical trials testing combinations of ICIs with a variety of backbones agents [poly-ADP ribose polymerase (PARP)-inhibitors, anti-IDO (indoleamine 2,3 dioxygenase), CDK4/6 inhibitors], are currently ongoing. Interestingly, encouraging results have been reported with the combination of immunotherapy plus CDK 4/6 inhibitors [55, 56].

Preclinical evidence suggests that CDK 4/6 blockade may boost anti-tumor immune response by enhancing tumor antigen presentation (mediated by the activation of tumor cell expression of endogenous retroviral elements) as well as suppressing regulatory T-cell proliferation [57, 58].

In the JPCE phase 1b study, where 28 HR+/HER2- MBC with 1–2 prior chemotherapy regimens were treated with the combination of pembrolizumab plus abemaciclib, initial ORR was 14.3% and the rate of stable disease at 16 weeks was 60%. No association was observed between PD-L1 expression and response to abemaciclib plus pembrolizumab [59].

Encouraging results have also been reported from early-phase trials evaluating the combination of immunotherapy with PARP-inhibitors. Indeed, PARP-inhibition may lead to the accumulation of damages in the DNA, thus resulting in genomic instability [60].

In addition, it has been reported that PARP-inhibitors may upregulate PD-L1 expression in a preclinical model, resulting in T-cell activity attenuation. Conversely, the combination of PARP-inhibitors with PD-L1 blockade proved to restore T-cell killing activity, thus providing a strong rational to test combinations of PARP-inhibitors and ICIs in MBC patients [61].

In particular, MEDIOLA phase II basket trial included 34 HER2- (17 were TNBC) MBC patients harboring germline deleterious mutations of BRCA 1/2 genes. Patients were treated with the PARP-inhibitor olaparib for a 4-week run-in phase, which was then associated with the anti-PD-L1 agent durvalumab. Twelve-week (primary endpoint) and 28-week DCR were 80% (> target DCR of 75%) and 50%, respectively, while ORR (secondary endpoint) was 63.3% in the overall cohort. Higher ORR (70%) was observed in MBC patients treated with ≤ 1 prior line of chemotherapy than more heavily treated patients (ORR 50%). Interestingly, translational analyses revealed BRCA2 reversion, TP53BP1 mutation, lack of BRCA2 loss of heterozygosity, PD-L1 and PD-L2 amplification as possible biomarkers predicting treatment resistance [62].

In the single-arm TOPACIO phase II trial, 55 TN MBC (patients may have received ≤ 2 lines of therapy) received the combination of the PARP-inhibitor Niraparib plus pembrolizumab. ORR (primary endpoint) was 21% (90% CI 12–33%), DCR (secondary endpoint) was 49% (90% CI 36–62%), median PFS (secondary endpoint) was 8.3 months. Numerically higher response rates were observed in patients with tumor BRCA mutations (ORR 47%, 90% CI 24–70%) [63].

In both these trials the combination of PARP-inhibitors and immune checkpoint blockade was associated with a manageable safety profile.

Emerging evidence also suggests a possible role for immunotherapy as maintenance treatment, especially in TN MBC. Interestingly, the SAFIR-02 IMMUNOTRIAL trial evaluated durvalumab as maintenance therapy after 1st or 2nd line chemotherapy in 199 HER2- MBC not harboring any targetable molecular alteration [64]. Patients were randomly assigned 2:1 to either durvalumab or maintenance chemotherapy without switch. In the overall population, median PFS was 2.7 months with durvalumab vs. 4.6 months with chemotherapy HR 1.40 (95% CI 1.00–1.96, P = 0.047). In HR+/HER2- MBC patients, a detrimental effect with Durvalumab was observed as compared to chemotherapy (HR 2.08, 95% CI 1.28–3.40), whereas in TNBC patients the HR favored Durvalumab (0.87, 95% CI 0.54–1.42). In terms of OS, among TNBC patients, a significant improvement was observed with Durvalumab with respect to chemotherapy (median OS 21 months vs. 14 months, HR 0.54, 95% CI 0.30–0.97, P = 0.0377). Similarly, a trend towards a beneficial effect of Durvalumab has been observed in the PD-L1 positive cohort (> 1% of immune cells over tumor area by SP142 assay; median OS 26 months vs. 12 months, HR 0.42, 95% CI 0.17–1.05, P = 0.0552).

Immunotherapy is emerging as a new frontier in BC treatment, leading to a new standard of care for the first line treatment of TN MBC with PD-L1 positive status. Despite this breakthrough, only a minority of BC patients seems to benefit from this novel treatment strategy. In this context, there is increasing interest in identifying the ideal partner for ICIs. The most promising results have been achieved by associating immunotherapy with chemotherapy, possibly due, at least in part, to the mechanism of chemotherapy-induced immunogenic cell death [72, 73]. However, results from early-phase clinical trials testing the combination with various biological agents, like anti-HER2 agents, PARP-inhibitors and CDK 4/6 inhibitors are encouraging, thus potentially allowing to broaden the population of BC patients who may benefit from immunotherapy. In addition, several trials currently testing combinations with other targeted agents are ongoing and results are awaited [74].

Other unresolved issues need to be addressed. First, a comprehensive predictive biomarker for immunotherapy efficacy capable of reliably identifying BC patients who may benefit from immunotherapy both in advanced and early setting is still lacking. In this context, although PD-L1 expression is associated with immunotherapy benefit in different trials in the metastatic setting, several other clinical studies failed to report a similar predictive role and there is a major discrepancy with results obtained by neoadjuvant trials. Moreover, the dynamic changes and heterogeneity of PD-L1 expression in the same sample tumor and in the different metastatic sites, together with the absence of reliable detection methods, limit its use in clinical practice. Other potential biomarkers of response/resistance to immunotherapy such as TILs, TMB, characterization of the immune infiltrate or gut microbiome have been investigated and may will provide additional information [35, 43, 44, 66, 75–77]. Unfortunately, these biomarkers require specific analyses in tumor specimens, which are not always easily available. Therefore, cheap and easy-to-measure biochemical/clinical parameters could significantly help in patients’ selection, like lactate dehydrogenase (LDH) [78] and BMI [70, 71].

Lastly, although available evidence is consistent in suggesting a larger beneficial effect of immunotherapy when administered earlier in the natural history of BC, the ideal timing for immunotherapy is still object of debate [42, 44–46]. The long journey of immunotherapy in BC has just started and its full potential yet to be discovered.