1 Introduction

Immunotherapy has changed the paradigm of cancer treatment and provided a new avenue for research endeavors. Over the last decade, several immune checkpoint inhibitors have been approved by the US Food and Drug Administration (FDA) to treat cancer, with many ongoing clinical trials assessing the efficacy of next-generation immunotherapy agents, new indications, and combination therapies. Currently approved checkpoint inhibitors include programmed cell death 1 (PD-1) inhibitors pembrolizumab, nivolumab, and cemiplimab; programmed death-ligand 1 (PD-L1) inhibitors atezolizumab, durvalumab, and avelumab; and the cytotoxic T lymphocyte–associated protein 4 (CTLA-4) inhibitor ipilimumab. These agents can provide a durable response in tumors characterized by tumor cell or infiltrating immune cell PD-L1 positivity, a ligand which typically binds to PD-1 on T lymphocytes to inhibit immune clearance of tumor cells. By blocking this interaction with PD-1/PD-L1 inhibitors, the immune system becomes able to recognize and eliminate tumor cells. Another mechanism for tumor immune evasion is the interaction between CTLA-4 on T lymphocytes and the ligands CD86 and CD80 on dendritic antigen-presenting cells, which normally inhibits T lymphocyte response. Following exposure to the CTLA-4 inhibitor ipilimumab, priming and activation of T cells to recognize tumor cells becomes possible.

In breast cancer, the FDA granted accelerated approval to the anti-PD-L1 agent atezolizumab and the anti-PD-1 agent pembrolizumab, each in combination with chemotherapy, for treatment of metastatic PD-L1-positive triple-negative breast cancer (TNBC) based on progression-free survival (PFS). These regimens were evaluated in the first-line setting. These therapies are available following years of prior study of immunotherapy in TNBC (Table 1), which will be outlined in this review.

Table 1 Select trials of checkpoint inhibitor therapy in metastatic triple-negative breast cancer
Full size table

2 Immune checkpoint inhibitors as monotherapy

Several monotherapy phase I trials were designed after the identification of TNBC as a tumor subtype that would derive clinical benefit from immunotherapy. In the phase I KEYNOTE-012 trial, 111 patients with metastatic TNBC were screened for PD-L1 expression and 32 were enrolled to receive the PD-1 inhibitor pembrolizumab 10 mg/kg IV every 2 weeks. PD-L1 positivity was defined as staining in the stroma or in 1% or more of tumor cells by immunohistochemistry. Among the 27 patients evaluable, the overall response rate (ORR) was 18.5% with a median PFS of 1.9 months. One patient (3.7%) had a complete response (CR), 4 (14.8%) had a partial response (PR), and 7 (25.9%) had stable disease with a median time to response of 17.9 weeks and median duration of response not reached [1, 2]. With promising single-agent activity, a phase II trial was designed to evaluate the activity of pembrolizumab as monotherapy. In KEYNOTE-086, pembrolizumab 200 mg IV every 3 weeks was evaluated in metastatic TNBC in two cohorts: (A) an unselected patient population later line and (B) PD-L1-positive tumors first-line. Flat dosing of pembrolizumab was chosen for the phase II study, following population pharmacokinetic analyses in melanoma that showed flat dosing provided exposures that maintained maximal efficacy response and tolerability as compared to the previously established 2 mg/kg dose every 3 weeks [3]. In cohort A which included 170 patients, the ORR was 5.3% overall with a median PFS of 2.0 months and median overall survival (OS) of 9.0 months. In PD-L1-positive patients, the ORR was 5.7% [4]. In cohort B which included 84 patients, the ORR was 21.4%, including 4 patients with a CR. The median PFS was 2.1 months and the median OS was 18.0 months [5]. Herein, the importance of immunotherapy administration in an earlier line setting is highlighted, as response rates in the PD-L1-positive populations differed significantly, dependent on the line of therapy.

KEYNOTE-119 was a phase III trial that enrolled 622 metastatic TNBC patients to assess pembrolizumab 200 mg IV every 3 weeks versus physician’s choice chemotherapy (capecitabine, eribulin, gemcitabine, or vinorelbine) following 1–2 prior lines of treatment. PD-L1 positivity was measured by a combined positive score (CPS), which is the number of PD-L1 staining cells (tumor cells, lymphocytes, macrophages) divided by the total number of viable tumor cells, multiplied by 100. In the overall population, there were no significant differences in outcomes, including OS, ORR, and duration of response. However, in tumors classified as PD-L1 positive, a trend toward improved OS was noted, particularly as PD-L1 enrichment increased. Furthermore, for patients with tumors classified by CPS ≥10, OS was 12.7 months with pembrolizumab versus 11.6 months with chemotherapy, compared to 9.9 months with pembrolizumab versus 10.8 months with chemotherapy for the overall population [6]. Given these results, pembrolizumab as monotherapy is not recommended as a second- or third-line treatment for metastatic TNBC.

In another phase Ia trial with an expansion cohort that enrolled 115 unselected patients with metastatic TNBC, treatment with the PD-L1 inhibitor atezolizumab (three different schedules employed) led to an ORR of 10.0% overall with a median duration of response of 21 months. Responses were improved in patients whose tumors were characterized by 5% or greater PD-L1 expression, with an ORR of 13.0% in PD-L1-positive tumors versus 5.0% in tumors with less than 5% PD-L1 expression. The median previous lines of treatment were seven (range 0–21). For patients receiving atezolizumab in the first-line setting, the ORR was 24.0%, versus 6.0% for subsequent lines. Median OS overall was 8.9 months, and 100% of responders were alive at year 2. Additionally, the median OS in the first-line setting was 17.6 months versus 7.3 months in those receiving second-line and beyond atezolizumab (p=0.01) [7, 8]. In the phase Ib JAVELIN study, unselected patients with metastatic breast cancer were treated with avelumab, an anti-PD-L1 monoclonal antibody, 10 mg/kg IV every 2 weeks. Of the 168 patients enrolled, 58 patients had TNBC of which 68.8% were determined to have PD-L1 -positive tumors. The ORR was just 3.0% overall, 5.2% in patients with TNBC overall, and 22.2% in TNBC tumors that were PD-L1 positive [9].

Taken together, checkpoint inhibitors as single agents in the treatment of metastatic TNBC demonstrated modest response rates overall, with more favorable responses seen in the first-line setting and in immune enriched tumors. Additionally, some patients achieved a sustained duration of response that was longer than typically achieved with chemotherapy.

3 Immune checkpoint inhibitors combined with chemotherapy

With modest response rates seen with immune checkpoint inhibitors as monotherapy, and with benefits only in a select group of patients, efforts have shifted toward evaluating these agents with chemotherapy. Synergistic efficacy is desired, and with combination therapies, the additional agent may help create an inflamed tumor microenvironment that stimulates responses to immunotherapy in otherwise non-responders [10].

Chemotherapy is an attractive paired agent, as following its use, high levels of antigens are released by tumor cells, suppressive immune cells are depleted, and PD-L1 is up-regulated [11]. For example, when anthracyclines are employed, immunogenic cell death ensues, which activates dendritic cells and can increase the proliferation of CD8+ T cells [12]. Cyclophosphamide has also been shown to induce immunogenic cell death, suppress T regulatory cells, and increase the proliferation of CD8+ T cells and natural killer cells [13, 14]. Taxanes decrease T regulatory and myeloid-derived suppressor cells (MDSCs) and increase tumor infiltrating lymphocyte (TIL) recruitment [15,16,17,18]. Additionally, platinum therapy induces immunogenic cell death and can increase major histocompatibility complex (MHC) class I complex presentation on tumor cells, which promotes T cell activation and downregulates MDSC function [19,20,21]. Gemcitabine decreases MDSCs and increases proliferation of CD8+ T cells [22, 23]. Fortunately, these findings have translated to improved clinical outcomes when chemotherapy is combined with immunotherapy in metastatic TNBC.

Based on the results of the IMpassion130 phase III trial, atezolizumab and nab-paclitaxel is a treatment option for patients with metastatic TNBC whose tumors have PD-L1 expression. In this study, patients with untreated metastatic TNBC were randomized 1:1 to receive atezolizumab 840 mg IV every 2 weeks plus nab-paclitaxel 100 mg/m2 IV on days 1, 8, and 15 of a 28-day cycle compared with placebo plus nab-paclitaxel [24]. The chemotherapy agent nab-paclitaxel was chosen to mitigate the need for pre-treatment steroids which can cause immunosuppression. All patients had to complete previous chemotherapy ≥ 12 months before randomization. PD-L1 expression was defined as >1% staining positive on tumor-infiltrating immune cells with the Ventana SP142 assay. In both treatment arms, 41% of patients had PD-L1-positive tumors. At the first interim analysis, at a median follow-up of 12.9 months, the median PFS was 7.2 months with atezolizumab and nab-paclitaxel, as compared with 5.5 months with placebo plus nab-paclitaxel, at a hazard ratio for progression or death of 0.80 (p=0.0025). Among patients with PD-L1-positive tumors, median PFS was 7.5 months and 5 months, respectively, which was statistically significant (p ≤ 0.001). The trial met its PFS endpoint in the intent-to-treat (ITT) and PD-L1-positive group. The ORR was also significantly higher in PD-L1 -positive patients treated with atezolizumab (59% vs 43%, p=0.002). The median OS in the ITT was 21.3 months in the atezolizumab and nab-paclitaxel arm versus 17.6 months in the placebo plus nab-paclitaxel arm, but this was not significant (p = 0.08). In the PD-L1-positive subgroup, median OS was longer in the atezolizumab arm by almost 10 months (25.0 vs 15.5 months; HR 0.62 [95% CI, 0.45–0.86]); however, formal statistical testing could not be conducted due to the pre-specified hierarchical statistical analysis plan requiring a statistically significant median OS in the ITT population. At the second interim analysis, again there was no statistically significant improvement observed in the median OS in the ITT population. Although an improved median OS was observed in the atezolizumab arm of the PD-L1-positive subgroup of 7 months (25 months with combination therapy vs 18 months with nab-paclitaxel and placebo), this positive result could not be formally tested [25]. In an updated survival analysis with a median follow-up of 18.8 months, the PD-L1-positive group achieved a higher median overall survival of 25.4 months with atezolizumab plus nab-paclitaxel versus 17.9 months with placebo plus nab-paclitaxel (final OS improvement of 7.5 months). The pre-specified testing hierarchy precluded a statistical assessment of this difference in the absence of a statistically significant OS benefit in the ITT population [26]. A patient with metastatic TNBC that has a tumor with PD-L1 expression ≥ 1%, and previous curative chemotherapy completed ≥ 12 months, would be a reasonable candidate for atezolizumab and nab-paclitaxel as first-line therapy given the modest PFS benefit.

In the phase III IMpassion131 trial, atezolizumab or placebo was combined with paclitaxel in patients with metastatic, previously untreated TNBC. Primary results were presented at the 2020 ESMO Congress, with no improvements in PFS or OS in the patients randomized to receive atezolizumab in addition to paclitaxel, for reasons that are not clear. These findings persisted regardless of PD-L1 positivity [27]. Subsequently, the FDA issued an alert to healthcare professionals that paclitaxel should not replace nab-paclitaxel when combined with atezolizumab for the treatment of metastatic TNBC. The phase III trial IMpassion132 is ongoing and is assessing first-line atezolizumab or placebo in metastatic TNBC patients recurring ≤12 months after completing standard (neo)adjuvant anthracycline and taxane chemotherapy, combined with other chemotherapeutic agents including gemcitabine/carboplatin or capecitabine. Results are not available at the time of this publication.

KEYNOTE-355 is a phase III first-line study in patients with metastatic TNBC, which randomized eligible patients 2:1 to receive chemotherapy with or without pembrolizumab 200 mg IV every 3 weeks. This study evaluated the efficacy of combination chemotherapy-immunotherapy with several regimens, namely nab-paclitaxel, paclitaxel, and gemcitabine/carboplatin. Eligible patients were treatment naïve in the advanced setting and had a disease-free interval (DFI) of ≥ 6 months from completion of adjuvant chemotherapy. A statistically significant PFS benefit was reported with the addition of pembrolizumab to chemotherapy in the 323 patients who had a CPS ≥10 (9.7 months vs 5.6 months, HR 0.65, p=0.0012) [28]. While both taxane and gemcitabine/carboplatin treatment groups were shown to benefit from the addition of pembrolizumab, the PFS benefit was more pronounced with a taxane regimen (HR 0.51 [95% confidence interval (CI) 0.33–0.78] with taxane, HR 0.77 [95% CI 0.53–1.11] with gemcitabine/carboplatin). Thirty-eight percent of patients had tumors characterized by CPS ≥10 and 75% had tumors characterized by CPS ≥1. The magnitude of benefit from the addition of pembrolizumab was greater for patients with CPS ≥10 than CPS ≥1 (CPS ≥1 median PFS 7.6 months with the addition of pembrolizumab vs 5.6 months, HR 0.74, p=0.0014). OS data is not yet available. Following these promising results, pembrolizumab combined with chemotherapy was granted accelerated approval by the FDA in November of 2020 for the treatment of metastatic PD-L1 positive (CPS ≥10) TNBC. Of note, this trial supports the use of a non-taxane regimen combined with a checkpoint inhibitor that can be used in the first-line setting for metastatic TNBC and can be considered to treat patients with a DFI as short as 6 months.

When comparing IMpassion130 to KEYNOTE-355, PD-L1 positivity was defined by two different tests—Ventana SP142 assay on tumor-infiltrating immune cells for IMpassion130 (the companion diagnostic for atezolizumab) and CPS via the Dako 22C3 assay for KEYNOTE-355 (the companion diagnostic for pembrolizumab). In a recent analysis of the performance of various PD-L1 assays on the biobank from IMpassion130, patients with CPS ≥10 via the Dako 22C3 assay did not achieve as significant of an OS benefit with the addition of atezolizumab, compared to patients with PD-L1-positive tumors by SP142. Utilizing the Dako 22C3 assay to select PD-L1-positive tumors (CPS ≥10), median OS was 22 months with atezolizumab versus 18.7 months (HR 0.77), compared to 25 months with atezolizumab versus 18 months (HR 0.71) via SP142 [29]. It is recommended that if a TNBC patient’s tumor is PD-L1 positive by the Ventana SP142 assay with a ≥ 1% score, then the patient is eligible for treatment with atezolizumab plus nab-paclitaxel and if a TNBC patient’s tumor is PD-L1 positive by the Dako 22C3 assay with a CPS ≥10, then a pembrolizumab-based regimen is appropriate.

In the phase Ib ENHANCE-1 study, pembrolizumab 200 mg IV every 3 weeks was combined with eribulin for 167 patients with metastatic TNBC, with 44% being characterized as PD-L1 positive. Eribulin is a microtubule inhibitor that decreases TGF-β production from cancer cells and surrounding stromal cells, a factor that causes immunosuppression and angiogenesis [30]. The combination of eribulin with a checkpoint inhibitor was anticipated to exert favorable antitumor effects, due to a combination of cytotoxic effect from the chemotherapeutic agent and restoration of tumor-specific T cell response. In ENHANCE-1, clinical activity was most pronounced in patients with PD-L1-positive tumors and treated first line (ORR CPS ≥ 1 34.5% vs 16.1%; CPS ≥ 10 30.8% vs 23.4%) [31].

Metastatic TNBC remains a challenging disease to treat without targeted therapy options. However, the combination of chemotherapy with checkpoint inhibitor therapy has shown great promise, particularly for patients with immune enriched tumors, and provides a new standard of care for first-line treatment of patients with metastatic PD-L1-positive TNBC.

4 Immune checkpoint inhibitors in combination with targeted therapy

In addition, a variety of targeted agents have also been evaluated in combination with immunotherapy in the treatment of metastatic TNBC, with encouraging preclinical rationale. The following combination approaches are described.

4.1 PARP inhibitors

A group of patients with TNBC who may particularly benefit from immunotherapy are patients with germline BRCA1/2 mutations. Tumors characterized by a BRCA1 mutation were shown to have a higher somatic mutational load, a greater number of TILs, and increased PD-L1 and CTLA-4 expression, suggesting immune activation [32]. In another evaluation, tumors characterized by a somatic mutation in the homologous recombination DNA damage repair system (including BRCA1/2) were more likely to have a higher mean tumor mutational burden (TMB), increased PD-L1 overexpression, and increased frequency of mutations in chromatin remodeling genes [

5 Immunotherapy toxicity

The immune checkpoint proteins, such as PD-L1, are involved in suppressing T cell function. When immunotherapy is employed (i.e., anti-PD-L1 therapy), tumor suppression of T cell function is relieved. This immune activation may not be tumor-specific, and consequently, a patient treated with immunotherapy may experience toxicities related to immune activation. The spectrum of toxicities from immunotherapy is broad, including inflammation of the skin, colon, thyroid gland, liver, pancreas, lung, and central nervous system.

The incidence of immune-related adverse events (irAEs) is significant. In a recent meta-analysis, irAEs were appreciated in 66% of all patients treated with anti-PD-1 or anti-PD-L1 therapies, including some fatal toxicities [57]. Previous experience with irAEs has informed management guidelines, allowing for recovery for most patients and often immunotherapy rechallenge (Table 2). Some studies have suggested the development of irAEs may correlate with an improved response to therapy [58,59,60]. Limited data exists for the safety and efficacy of immunotherapy in immune-vulnerable populations, such as patients with pre-existing autoimmune conditions or HIV/AIDS, though generally it is favored to pursue immunotherapy in these patients if clinically appropriate [61].

Table 2 Immune-related adverse events
Full size table

While the risks of checkpoint inhibitor therapy should be considered for each patient, the benefits of this therapy can be paramount. Moreover, durable responses can be achieved in patients with metastatic TNBC. Patients with immune-enriched advanced TNBC should be considered for immunotherapies, with close monitoring and multi-disciplinary management of irAEs should they arise.

6 Biomarkers of response to immunotherapy

It has been well-established that certain cancers, such as lung cancer and melanoma, respond to immunotherapy treatment algorithms, whereas other cancers do not achieve a robust response. Extensive work has been carried out to determine why certain cancers respond, and which patients within traditionally non-responsive cancers will achieve a response. An active and enriched tumor immune microenvironment is important for predicting response to immunotherapy, with higher rates of response in tumors with elevated TILs, PD-L1 positivity, and an increased density of tumor-infiltrating CD8+ T cells [62]. Such tumors have been identified as “inflamed” or “hot” tumors.

Conversely, tumors with low levels of TILs, PD-L1 positivity, and CD8+ T cells are considered “non-inflamed” or “cold,” with a lower likelihood of response to immunotherapy. Breast cancer tumors more commonly fall in this category. In the initial phase I studies of immunotherapy as monotherapy, triple-negative breast tumors of the breast cancer subtypes were identified as being more likely to achieve a response. Subsequently, it has been shown that TNBC patients have tumors that are “hotter” than other breast cancer subtypes, with higher expression levels of PD-L1 [63], increased genomic instability and higher TMB leading to increased neoantigen presentation [64,65,66], and higher levels of TILs [67].

Exploratory analysis has confirmed these inflammatory markers can identify which breast tumors will be more likely to respond to immunotherapy. Biomarker evaluation as part of the IMpassion130 trial (evaluating nab-paclitaxel with or without atezolizumab first-line in metastatic TNBC) determined PD-L1 expression on immune cells by the VENTANA SP142 IHC assay to be highly predictive of response, with significant improvements seen in both PFS and OS in the PD-L1-positive patient cohort [24]. TILs, and specifically a higher density of CD8+ cytotoxic TILs, are also positively correlated with immunotherapy response. KEYNOTE-173 (a phase Ib trial evaluating neoadjuvant chemotherapy with or without pembrolizumab in early TNBC) showed that a higher combined score evaluating levels of stromal TILs and PD-L1 expression was significantly associated with higher pathologic complete response and overall response rates in patients with early-stage TNBC [68]. Furthermore, in a retrospective tissue evaluation of early-stage breast tumors, a high number of CD8+ T cells with features of tissue-resident memory T (TRM) cell differentiation were significantly associated with improved survival in TNBC [69]. In the metastatic setting, a higher density of CD8+ TILs was associated with increased PFS and OS in the IMpassion130 trial [70]. Stromal TILs were associated with PFS benefit only.

In addition to an assessment for levels of TILs, PD-L1 expression, and TMB, gene signatures may also identify breast tumors primed for an immune response. In a comprehensive genomic analysis of TNBC, four distinct gene signatures were identified—luminal androgen receptor (LAR), mesenchymal (MES), basal-like immunosuppressed (BLIS), and basal-like immune-activated (BLIA)—of which the BLIA subtype is considered “inflamed” and “hot” and primed for response to immunotherapy. In the BLIA subtype, tumors tend to upregulate genes controlling B cell, T cell, and natural killer cell functions, whereas in the BLIS subtype, tumors tend to decrease expression of molecules controlling antigen presentation, immune cell differentiation, and innate and adaptive immune cell communication [71]. While hypothesis-generating, this subclassification has not been correlated with treatment response, but has been associated with outcome, with the best prognosis achieved with the BLIA subtype and worst disease-free survival in the BLIS subtype. In another evaluation where breast cancer tumors were classified into one of four groups based on their immune-related gene signatures, patients with tumors classified as highly immunogenic (with upregulation of immunoregulatory transcripts including PD-L1, PD-1, FOXP3, IDO1, and CTLA-4) achieved prolonged survival [72]. Again, this evaluation did not include a correlation with immunotherapy treatment response.

Immunotherapy biomarker expression is influenced by tissue source. Moreover, immune-enriched tumors are more commonly identified from the primary breast tumor, compared to a metastatic tumor. In a paired analysis of 76 TNBC tumors for TIL count and PD-L1 positivity (using the E1L3N XP rabbit monoclonal antibody, with positivity defined as ≥1% positive tumor, or stromal cells), both biomarkers were significantly decreased among metastatic tumors compared to primary tumors [73]. Of note in this study, tissue samples derived from tissue microarray compared to the whole section (whether metastatic or primary) had overall lower biomarker expression as well. In another study of the Foundation Medicine clinical database including 340 TNBC cases, PD-L1 positivity (as defined by the SP142 assay [≥1% immune cells stained positive]) was significantly more common among primary tumors compared to metastatic tumors (63.7% vs 42.4%, p<0.0001) [74]. In the majority of the clinical trials referenced in Table 1, tissue was allowable from both the primary breast or a metastatic site, with fresh tissue and a metastatic site preferred, though archived and/or primary breast tissue allowable. Limited analyses exist from these trials regarding source of tissue and biomarker expression. In a post hoc analysis of the IMpassion130 trial that assessed atezolizumab with nab-paclitaxel in advanced TNBC, PD-L1 positivity was more common if testing was performed on breast tissue (comprising 62% of available tissue) compared to a metastatic site, 44% versus 36% [29]. Given these data, PD-L1 testing ideally can be performed on a metastatic tumor and if not available, then it is reasonable to test on primary breast tissue.

7 Concluding remarks

With currently available and evolving data, checkpoint inhibitor therapy should be offered to metastatic TNBC patients with an identified biomarker predictive of response (i.e., PD-L1 overexpression or an elevated CPS). Timing is important as patients should be evaluated for biomarkers of response at time of initial diagnosis, with immunotherapy utilized in the first-line setting when possible. Several studies have highlighted the best response to immunotherapy, and a durable response is achieved when therapy is used in the first-line setting. It is not clear if one checkpoint inhibitor is better than another, nor which biomarker assay most reliably predicts for response, and additional work is needed in this area to help select the best treatment for an individual patient. Many exciting combination approaches are being evaluated, and patients should consider clinical trial opportunities when available.