The emerging role of immune checkpoint inhibitors for the treatment of breast cancer

ABSTRACT

Introduction

Breast cancer has traditionally been viewed as immunogenically ‘cold,’ but two immune checkpoint inhibitors have been approved in combination with chemotherapy for PD-L1 positive advanced triple-negative breast cancer (TNBC), and pembrolizumab was also recently approved for early stage TNBC. As the landscape is rapidly evolving, a comprehensive review of checkpoint inhibitors in breast cancer is needed to aid clinicians in selecting appropriate candidates for therapy, and to highlight ongoing promising studies in this area and topics in need of further investigation.

Area covered

This review summarizes the latest evidence from completed and ongoing trials of immune checkpoint inhibitors. Ongoing studies were identified using a search of ClinicalTrials.gov with the term ‘breast cancer’ along with specific checkpoint inhibitor agents.

Expert opinion

A number of novel combination strategies are under investigation to enhance response and overcome resistance to immunotherapy, with promising preliminary data from checkpoint inhibitors targeting TIGIT, combinations with small molecule inhibitors such as lenvatinib, and injectable agents directly influencing the immune microenvironment. As immunotherapy enters into the curative setting, biomarkers predictive of immunotherapy benefit are needed, as PD-L1 status has not been a helpful discriminator in completed trials in early-stage breast cancer.

KEYWORDS:

• Atezolizumab
• breast cancer
• checkpoint inhibitors
• durvalumab
• immunotherapy
• small-molecule inhibitors
• PARP inhibitors
• pembrolizumab
• triple negative breast cancer

1. Introduction

Immune checkpoint inhibitors have transformed the therapeutic landscape of oncology, with anti-PD-1/PD-L1-based therapies approved for nearly every solid malignancy [1]. The initial therapeutic success of checkpoint inhibition was observed in melanoma, followed shortly thereafter by approvals for other T-cell inflamed tumors. However, breast cancers have historically been viewed as immunogenically silent, and were believed to be resistant to immunotherapy [2]. The first trials of immunotherapy in breast cancer were disappointing, demonstrating limited efficacy and a high level of toxicity due to the narrow therapeutic index of interferon and interleukins [3,4]. Vaccine-based immunotherapy demonstrated favorable toxicity profiles with some immune stimulatory effects such as antibody production or T-cell proliferation [5,6], but large randomized trials of vaccines have failed to demonstrate significant clinical activity [7]. As distinct molecular profiles of breast cancer became clearly defined over the last two decades, triple-negative breast cancer (TNBC) emerged as an appealing subtype to evaluate novel immunotherapy approaches [8]. TNBC is defined by the lack of expression of the hormone receptors (HR) and the human epidermal growth factor receptor 2 (HER2), and exhibits an aggressive clinical course and unique disease biology, disproportionately affecting young women, BRCA mutation carriers, as well as west African and Hispanic populations [9]. The tumor microenvironment in patients with TNBC is uniquely characterized by higher levels of both stromal and tumor infiltrating lymphocytes (TILs) [10], higher levels of PD-L1 expression [11], and greater mutational variability and genomic instability that can then give rise to tumor-specific neoantigens [12]. Thus, immunotherapy offered a promising therapeutic strategy in TNBC, especially with the relative dearth of effective treatment options beyond chemotherapy. Checkpoint inhibitor immunotherapy has been most extensively evaluated in TNBC, but efficacy has now been described in HR+[13] and HER2+ tumors [14], and novel combination strategies promise to alter the tumor microenvironment and induce responses in a greater proportion of patients. This review summarizes the current and future landscape of checkpoint inhibitor immunotherapy across all subtypes of breast cancer, for both early and advanced disease. Relevant clinical trials were identified through a search of ClinicalTrials.gov on 1 March 2021 with the terms ‘breast cancer’ and ‘pembrolizumab,’ ‘atezolizumab,’ ‘durvalumab,’ ‘nivolumab,’ ‘avelumab,’ ‘ipilimumab,’ or ‘tremelimumab.’

2. Advanced triple-negative breast cancer

2.1. Checkpoint inhibitor monotherapy

Early studies investigating the role of immune checkpoint inhibitor monotherapy demonstrated modest clinical activity with single agents, but found responses were enriched in patients with higher PD-L1 expression and fewer previous lines of treatment. A phase I study of single agent atezolizumab in 115 patients with metastatic TNBC (mTNBC) documented an overall response rate (ORR) of 24% in patients with previously untreated disease, compared to 6% in the second line setting [15]. Moreover, patients with immune cell (IC) PD-L1 expression ≥ 1% had higher ORRs and longer overall survival (OS) compared to those with expression <1%. Similarly, the phase Ib KEYNOTE-012 trial of pembrolizumab monotherapy in 32 patients with heavily pre-treated and treatment-naïve mTNBC patients demonstrated an ORR of 18.5%[16]. On the basis of these early phase data, the phase II KEYNOTE-086 trial evaluated pembrolizumab monotherapy in 2 cohorts: those with previously treated disease and those with untreated disease and PD-L1 combined positive score (CPS) ≥ 1. In the previously untreated cohort, ORR was 21.4% compared to 5.3% in previously treated patients and 5.7% in previously treated patients with PD-L1 CPS ≥ 1 [17,18].

To definitively evaluate the role of checkpoint inhibitor monotherapy in advanced TNBC, the phase III KEYNOTE-119 trial randomized 622 patients in a 1:1 fashion to pembrolizumab or single agent chemotherapy of investigator’s choice: capecitabine, gemcitabine, eribulin or vinorelbine (Table 1) [19]. Eligible patients had 1–2 prior lines of therapy for advanced TNBC, with prior treatment with a taxane or anthracycline in the neoadjuvant, adjuvant, or metastatic setting. The study had three co-primary endpoints: OS in patients with CPS ≥ 10 (31% of patients), CPS ≥ 1 (65% of patients) and the overall population. There was no significant difference in OS in the CPS ≥ 10 (median 12.7 months with pembrolizumab versus 11.6 months with chemotherapy), CPS ≥ 1 (median 10.7 months versus 10.2 months), or overall populations (median 9.9 months versus 10.8 months). An exploratory analysis in patients with CPS ≥ 20 (18% of patients) showed improved progression free survival (PFS) and OS for patients treated with pembrolizumab, with ORR of 26.3% compared to 11.5% with chemotherapy. Median duration of response (DOR) was 3.9 months longer with pembrolizumab compared to chemotherapy in the overall population, and median DOR was not reached in the CPS ≥ 10 subgroup with a median study follow-up of 31.4 months. In an analysis of a subset of 26 patients with high tumor mutational burden (TMB), there was a trend toward improved ORR (14.3% with pembrolizumab versus 8.3% with chemotherapy) in the 14 patients who received pembrolizumab [20]. The encouraging efficacy versus single agent chemotherapy and the potential for prolonged response supports the role of immunotherapy in mTNBC.

Table 1. Randomized clinical trials of immune checkpoint inhibitors in advanced/metastatic breast cancer. Outcomes listed for immunotherapy containing arms vs. non-immunotherapy arms

Study	Patient Population	Sample Size	Treatment	ORR	Median PFS (months)	Median OS (months)	Comment
NCT02555657 KEYNOTE-119 Phase 3 Open-label	TNBC 2^nd or 3^rd line Prior anthracycline and taxane	n = 312 n = 310	Pembrolizumab Capecitabine, Eribulin, Gemcitabine, or Vinorelbine	9.6% 10.6%	2.1 3.3 (not formally tested)	9.9 10.8 (not formally tested)	No OS difference in ITT (not formally tested), CPS ≥ 1 (p = 0.073), and CPS ≥ 10 (p = 0.057) subgroups. Post-hoc analysis suggest OS benefit in CPS ≥ 20 (median 14.9 mo vs 12.5 mo)
NCT02425891 IMpassion130 Phase 3 Double-blind Placebo-controlled	TNBC 1^st line TFI ≥12 months	n = 451 n = 451	Atezolizumab + nab-paclitaxel Placebo + nab-paclitaxel	56% 46%	7.2 5.5 (p = 0.002)	21.0 18.7 (p = 0.08)	OS difference seen in PD-L1+ subgroup (median 25.4 mo vs 17.9 mo) but not statistically assessed due to hierarchical design.
NCT03125902 IMpassion131 Phase 3 Double-blind Placebo-controlled	TNBC 1^st line TFI ≥12 months	n = 431 n = 220	Atezolizumab + Paclitaxel Placebo + Paclitaxel	54% 47%	5.7 5.6 (not formally tested)	19.2 22.8 (not formally tested)	No PFS (p = 0.20) or OS benefit seen in the PD-L1+ subgroup
NCT02819518 KEYNOTE-355 Phase 3 Double-blind Placebo-controlled	TNBC 1^st line TFI ≥ 6 months	n = 566 n = 281	Pembrolizumab + nab-paclitaxel, Paclitaxel, or Gemcitabine/Carboplatin Placebo + nab-paclitaxel, Paclitaxel, or Gemcitabine/Carboplatin	41.0% 35.9%	7.5 5.6 (not formally tested)	NA	PFS benefit reached significance only in the CPS ≥ 10 subgroup (median 9.7 mo vs 5.6 mo, p = 0.0014)
NCT02299999 SAFIR02 Breast Phase 2 Open label	≤2 prior CTs Received 6–8 cycles CT without PD No targetable molecular alteration	n = 131 n = 68	Maintenance Durvalumab Maintenance CT	NA NA	2.7 4.6 (p = 0.047)	21.7 17.9 (p = 0.423)	OS prolonged in TNBC subgroup (median 21.2 mo vs 14.0 mo, p = 0.0377)
NCT03051659 Phase 2 Open label	HR+, HER2- 2+ prior ET	n = 44 n = 44	Pembrolizumab + Eribulin Placebo + Eribulin	27% 34%	4.1 4.2 (p = 0.33)	13.4 12.5 (p = 0.65)	PFS and OS not significantly prolonged in PD-L1 positive patients
NCT02924883 KATE2 Phase 2 Double-blind Placebo-controlled	HER2+ Prior trastuzumab Prior taxane Any PD-L1 status	n = 133 n = 69	Atezolizumab + T-DM1 Placebo + T-DM1	45% 43%	8.2 6.8 (p = 0.33)	NA	1 year OS 89% in both arms

HER2 = human epidermal growth factor receptor 2. HR = hormone receptor. ORR = overall response rate. PFS = progression-free survival. OS = overall survival. NR = not reached. NA = not available. PD-L1 = programmed death ligand-1. CPS = combined positive score. CT = chemotherapy. ET = endocrine therapy.

In patients with tumors characterized by PD-L1 expression and no prior therapy in the advanced setting, there may be a role for immunotherapy alone, although no agent is FDA approved as monotherapy specifically for breast cancer. However, pembrolizumab monotherapy does carry tumor agnostic indications for patients with TMB of ≥ 10 mutations/mb or microsatellite instability-high status as a result of the KEYNOTE-158 trial [21,22], warranting consideration in the minority of breast cancer patients eligible per these indications.

2.2. Chemotherapy and immunotherapy combinations

While single agent checkpoint inhibition has not definitively outperformed chemotherapy in the treatment of mTNBC, several chemoimmunotherapy regimens are now approved for the treatment of PD-L1 positive mTNBC. Preceded by a phase Ib trial demonstrating promising preliminary efficacy and tolerability [23], the double-blind phase III IMpassion130 trial randomized 902 patients with untreated advanced TNBC to atezolizumab or placebo in combination with nab-paclitaxel [24]. IMpassion130 had four co-primary endpoints – PFS and OS in both the intention-to-treat (ITT) and PD-L1 IC ≥ 1% population (the latter group constituting 41% of patients), with statistical testing performed hierarchically in the ITT and then PD-L1 subgroups for both outcomes. PFS was significantly higher in both populations with atezolizumab (ITT: median 7.2 months versus 5.5 months; PD-L1+: median 7.5 months versus 5.0 months). Although OS did not significantly differ in the ITT population (median 21.0 months with atezolizumab versus 18.7 months with placebo), a clinically meaningful 7.5 month median OS benefit was seen with atezolizumab in the PD-L1+ population (25.4 months versus 17.9 months), although formal statistical analysis of this latter result was not performed due to the prespecified testing hierarchy [25]. These results led to the first FDA and European Commission accelerated approval of an immune checkpoint inhibitor combination, atezolizumab with nab-paclitaxel, for PD-L1+ (as assessed by the VENTANA SP142 companion diagnostic assay) advanced TNBC.

The similarly designed IMpassion131 double-blind, placebo-controlled phase III trial randomized 651 patients with untreated advanced TNBC in a 2:1 fashion to atezolizumab versus placebo, administered in combination with paclitaxel [26]. Despite the similar trial design, there was no statistically significant difference in PFS or OS with the addition of atezolizumab to paclitaxel in either the ITT or PD-L1 IC-positive subgroup. The reasons behind the difference in outcome compared to those seen in IMpassion130 remain unclear. Steroid premedication for paclitaxel and differences in tissue distribution of the two formulations could theoretically impact immune response, the increasing availability of immunotherapy may influence OS results if immunotherapy was given as a later line treatment. The results underscore the imperfection of PD-L1 status, assessed by immunohistochemistry, as a biomarker for selecting patients for immunotherapy – and other unmeasured confounders such as proportion of TILs may have influenced results. Given the conflicting results of these trials, the FDA indication for atezolizumab with chemotherapy for advanced breast cancer was voluntarily withdrawn, although the European Commission approval remains in place as of this publication.

The double-blinded, placebo controlled, phase III KEYNOTE-355 trial randomized 847 patients in a 2:1 fashion to the combination of pembrolizumab plus investigator’s choice chemotherapy (nab-paclitaxel, paclitaxel, or gemcitabine plus carboplatin) versus chemotherapy alone in the first line setting for advanced or mTNBC [27]. Definitive PFS assessment was performed at second interim analysis (December 2019), revealing a significant improvement with pembrolizumab in the CPS ≥ 10 population (median 9.7 months versus 5.6 months). However, the PFS improvement with pembrolizumab was not significant in the CPS ≥ 1 (median 7.6 months versus 5.6 months) and not tested in the ITT population due to hierarchical testing design (median 7.5 months versus 5.6 months). Moreover, the patients who respond to combination chemoimmunotherapy appear to have remarkably durable responses, with median DOR of 19.3 months in the combination arm compared to 7.3 months in chemotherapy alone in the CPS ≥ 10 population [28]. Notably, as opposed to IMpassion131, responses did not differ in patients treated with different taxanes, and pembrolizumab with paclitaxel remains a viable treatment option in this setting. KEYNOTE-355 led to the approval of pembrolizumab in combination with chemotherapy in PD-L1 positive (CPS ≥ 10 as assessed by the Dako 22C3 companion diagnostic) advanced TNBC.

These studies also illustrate the divergence in PD-L1 assays and methods of quantification (evaluating only immune cells versus both tumor and immune cells), which are correlated but not synonymous [29,30]. The specific cells on which PD-L1 is assessed, the cutoff for positivity, and the antibody used for PD-L1 detection [31], all influence the subset of patients selected. A post-hoc analysis of IMpassion130 compared PD-L1 positivity with the VENTANA SP142 used in IMpassion130 and the Dako 22C3 assay used in KEYNOTE-355 [30]. Among 614 evaluable patients, 46% were PD-L1+ (IC ≥ 1%) via the SP142 assay, and 81% were PD-L1+ (CPS ≥ 1) via the 22C3 assay – and a greater PFS/OS benefit was seen in those positive by both assays, rather than those positive only by the 22C3 assay. Further comparison with the CPS cutoff of 10 used in KEYNOTE-355 is needed to determine if one assay more reliably predicts treatment benefit.

In summary, two immunotherapy agents, atezolizumab and pembrolizumab, have been approved in combination with chemotherapy for PD-L1 positive advanced TNBC. Pembrolizumab was granted regular approval by the FDA, but the U.S. approval of atezolizumab was withdrawn given the conflicting results of IMpassion130 and IMpassion131. The optimal chemotherapy backbone for such regimens also remains to be determined. Given the efficacy of the recently approved antibody-drug conjugate sacituzumab govitecan in the third line setting, it is now under evaluation as a frontline treatment with or without pembrolizumab in the randomized Saci-IO TNBC trial (NCT04468061); the study will help determine if sacituzimab govitecan has synergistic effects with immunotherapy.

2.3. Sequencing of therapy

Chemotherapy has long been posited to have an immunomodulatory role that may improve responses to immunotherapy – with evidence that cyclophosphamide depletes circulating regulatory T cells [32], and anthracyclines induce type I interferon signaling in tumors [33]. Preclinical work has also shown that taxanes can promote the differentiation of pro-inflammatory macrophages [34], and platinum agents can promote neoantigen release upon cancer cell death, influencing immune response [35]. Two randomized phase II trials have examined the role of immune checkpoint inhibition after induction treatment with chemotherapy or radiation. The adaptive phase II randomized TONIC trial evaluated nivolumab monotherapy preceded by a two week induction with radiotherapy (3 fractions of 8 Gy) or chemotherapy (cyclophosphamide, cisplatin, doxorubicin) versus no induction in 67 patients with metastatic TNBC [36]. The highest responses were observed in patients who received induction chemotherapy with cisplatin and doxorubicin, with ORR 23% and 35%, respectively. Biomarker analysis in these cohorts found that patients who received cisplatin or doxorubicin induction had upregulation of immune-related genes involved in PD-1/PD-L1 and cytotoxic T cell pathways, and increased levels of proinflammatory genes in the JAK-STAT and TNF-α signaling pathways with doxorubicin induction.

The SAFIR02-BREAST IMMUNO substudy evaluated the role of maintenance durvalumab in 82 patients with mTNBC and 108 patients with metastatic HR+/HER2- breast cancer resistant to endocrine therapy [37]. Patients from the SAFIR02-BREAST study were included in this arm if they had no targetable mutations, and had completed 6–8 cycles of induction chemotherapy without progressive disease. Patients were randomized 2:1 to maintenance durvalumab or maintenance chemotherapy of physicians choice [37]. While there was no significant difference in PFS or OS in the overall population with durvalumab versus chemotherapy maintenance, an exploratory subgroup analysis showed improved OS in 82 patients with TNBC treated with durvalumab maintenance (median 21.2 months versus 14.0 months). This benefit was more pronounced in the 32 patients with PD-L1 IC ≥ 1% TNBC (median 27.3 months versus 12.1 months) compared to the 50 PD-L1-negative patients (median 19.5 months versus 14.0 months). No subgroup (even TNBC) demonstrated a significant PFS benefit with durvalumab, and PFS was longer with chemotherapy in the HR+ subgroup.

Although the preclinical models support the immunomodulatory effects of chemotherapy and results from these early stage clinical trials are intriguing, immunotherapy should currently be prioritized as part of first line treatment as per the IMpassion130 and KEYNOTE-355 trials. The data from SAFIR02-BREAST suggest a potential role for immune checkpoint inhibition in the maintenance setting, but will require further evaluation in phase III trials. Ongoing trials of immunomodulatory chemotherapy given with immunotherapy may strengthen the evidence for this approach, including the TONIC-2 trial (NCT04159818) evaluating nivolumab with cisplatin or doxorubicin, and the ALICE trial (NCT03164993) of liposomal doxorubicin and cyclophosphamide with or without atezolizumab.

3. Early stage triple-negative breast cancer

With standard neoadjuvant chemotherapy (NACT), only about a third of patients with TNBC will have a pathologic complete response (pCR) [38]. As such, a number of trials have evaluated the combination of checkpoint inhibitors with standard neoadjuvant chemotherapy in an attempt to improve rates of response, and ultimately the number of patients cured of disease (Table 2) [39]. With NACT, pCR is a robust early surrogate for the clinically important outcomes of event free survival (EFS) and OS [38]. Accumulating long-term survival outcomes suggest that pCR to immunotherapy-based neoadjuvant regimens also portends a good prognosis, but further study is needed to confirm the validity of this early endpoint for immunotherapy.

Table 2. Randomized clinical trials of neoadjuvant/adjuvant immunotherapy for breast cancer. Outcomes listed for immunotherapy containing arm vs non-immunotherapy arm

Study	Patient Population	Sample Size	Treatment	pCR Rate	Comment
NCT03036488 KEYNOTE-522 Phase 3 Double-blind Placebo-controlled	T1cN1-2 or T2-4N0-2	n = 784 n = 390	Paclitaxel + Carboplatin + Pembrolizumab → AC/EC + Pembrolizumab → Surgery → Pembrolizumab Paclitaxel + Carboplatin + Placebo → AC/EC + Placebo → Surgery → Placebo	63.0% 55.6% (not formally tested)	Results listed for IA3. Significant pCR difference at IA1 (p = 0.00055) and IA2 (p = 0.00221), not tested at IA3. At IA4, 3-year EFS 84.5% vs 76.8% (p = 0.00031)
NCT03197935 IMpassion031 Phase 3 Double-blind Placebo-controlled	Unilateral tumor >2 cm	n = 165 n = 168	nab-paclitaxel + Atezolizumab → EC + Atezolizumab nab-paclitaxel + Placebo → EC + Placebo	58% 41% (p = 0.0044)	PD-L1+ pCR rates: 69% vs 49% (p = 0.021)
NCT03726879 IMpassion050 Phase 3 Double-blind Placebo-controlled	T2-4N1-3 disease HER2+	n = 226 n = 228	AC + Atezolizumbab → Paclitaxel + HP + Atezolizumab → Surgery → HP + Atezolizumab AC + Placebo → Paclitaxel + HP + Placebo → Surgery → HP + Placebo	62.4% 62.7% (p = 1.0)	PD-L1+ pCR rates: 64.2% vs 72.5% (p = 0.2)
NCT02620280 NeoTRIPaPDL1 Phase 3 Open label	T1cN1+ or T3N0+ Unilateral IDC High Ki-67 or Grade	n = 142 n = 138	nab-paclitaxel + Carboplatin + Atezolizumab → Surgery → AC/EC/FEC nab-paclitaxel + Carboplatin → Surgery → AC/EC/FEC	43.5% 40.8% (p = 0.66)	Difference in pCR not significant
NCT01042379 I-SPY2 Phase 2 Adaptively randomized Open label	HER2- Tumor > 2 cm on imaging/2.5 cm on exam MammaPrint low-risk excluded	n = 69 n = 181	Paclitaxel + Pembrolizumab → AC Paclitaxel → AC	44% 17% (not formally tested)	TNBC pCR rates: 60% vs 22% HR+ pCR rates: 30% vs 13%
		n = 73 n = 295	Paclitaxel + Pembrolizumab → Pembrolizumab Paclitaxel → AC	15% 15% (not formally tested)	Protocol amended so patients no longer received pembro monotherapy due to 3 progressions
		n = 73 n = 299	Paclitaxel + Olaparib + Durvalumab → AC Paclitaxel → AC	37% 20% (not formally tested)	TNBC pCR rates: 47% vs 27% HR+ pCR rates: 28% vs 14%
		n = 73 n = 299	Paclitaxel + SD-101 + Pembrolizumab → AC Paclitaxel → AC	34% 20% (not formally tested)	TNBC pCR rates: 44% vs 28% HR+ pCR rates: 26% vs 14%
NCT02685059 GeparNuevo Phase 2 Double-blind Placebo-controlled	Tumor ≥ 2 cm	n = 88 n = 86	nab-paclitaxel + Durvalumab → EC + Durvalumab nab-paclitaxel + Placebo → EC + Placebo	53.4% 44.2% (p = 0.224)	Improvement seen in DDFS (HR 0.37, p = 0.0148) and OS (HR 0.26, p = 0.0076) with addition of durvalumab

pCR = pathologic complete response. HER2 = human epidermal growth factor receptor 2. HR = hormone receptor. TNBC = triple negative breast cancer. A = adriamycin. C = cyclophosphamide. E = epirubicin. F = 5-fluorouracil. PD-L1 = programmed death ligand-1. IDC = invasive ductal carcinoma. IA = interim analysis. HP = trastuzumab + pertuzumab.

A number of immunotherapy strategies have been evaluated for TNBC and HR+/HER2- patients in the adaptively randomized phase II I-SPY2 trial. Enrolled patients are required to have tumors of at least 2.5 cm in size, and are assigned to a treatment arm based on Bayesian probability of a treatment achieving a higher rate of pCR than the control. Treatments are compared to a continuously enrolling contemporary control arm, in which patients are treated with 12 doses of weekly paclitaxel followed by four cycles of doxorubicin and cyclophosphamide (AC) [40,41]. An investigational treatment graduates when the predicted probability of success in a confirmatory phase III trial reaches 85%. The pembro4 regimen was the first immunotherapy combination evaluated in I-SPY2, consisting of four cycles of pembrolizumab concurrently with weekly paclitaxel, followed by standard AC [42]. The 29 patients enrolled in this arm with TNBC had a pCR rate of 60%, comparing favorably to 22% in the contemporary controls. Rates of pruritus (31.9%) and hypothyroidism (10.1%) were as expected, but adrenal insufficiency occurred in 8.7% of patients, higher than expected from prior trials in metastatic disease. An exploratory analysis demonstrated numerically similar EFS in patients who had a pCR in both the experimental and control arms, with a median follow-up for 2.8 years. Similar results were seen in an arm evaluating durvalumab and olaparib given concurrently with paclitaxel followed by AC [43]. Rates of pCR were 47% in 21 patients enrolled with TNBC, compared to 27% in historical controls.

The pembro8-noAC arm consisted of four cycles of pembrolizumab with weekly paclitaxel followed by four additional cycles of pembrolizumab alone as an anthracycline free treatment strategy [44]. However, the protocol was amended after three patients progressed on pembrolizumab monotherapy, and investigators were given the option to move forward with surgery after completion of pembrolizumab with paclitaxel, or to administer pembrolizumab concurrent with AC prior to surgery. For the primary efficacy analysis, patients receiving AC were considered not to have a pCR. In a preliminary analysis, estimated pCR rate was 27% in both treatment and control patients with TNBC in the pembro8-noAC arm, failing to reach the target threshold of 60% – although not all patients were evaluable and pCR was estimated by MRI in some cases. Adverse events were again notable for an 8.3% rate of adrenal insufficiency. A third arm of the I-SPY2 trial evaluated durvalumab, olaparib, and paclitaxel followed by AC (with inclusion independent of BRCA mutational status) [45]. A pCR rate of 47% was reported in patients with TNBC, compared to 27% in contemporary controls. Grade 3+ immune-related adverse events (irAEs) of note included 7% of patients experiencing adrenal insufficiency as well as a 7% rate of colitis.

Other early randomized trials failed to demonstrate a clear benefit to immunotherapy combinations in TNBC. The phase III randomized, open label NeoTRIPaPDL1 trial evaluated the addition of atezolizumab to neoadjuvant carboplatin and nab-paclitaxel [46]. The pCR rate was 43.5% in 138 patients treated with atezolizumab, compared to 40.8% in 142 patients receiving chemotherapy alone. Responses were higher with both treatment strategies in the PD-L1 IC ≥ 1% subgroup, but not significantly different between the two arms (51.9% with atezolizumab versus 48.0% without). Possible explanations for the lack of pCR benefit in NeoTRIPaPDL1 include a higher proportion of TILs in the chemotherapy arm (given that TILs are a known predictor of chemotherapy response) [47], and the administration of anthracyclines after surgery given the potential immunomodulatory effects of doxorubicin illustrated in the aforementioned TONIC trial [36]. It remains to be seen if the interplay between pre-operative immunotherapy and post-operative anthracycline is reflected in the EFS data of NeoTRIPaPDL1.

The phase II double-blind GeparNuevo trial also initially reported equivocal results with the addition of durvalumab or placebo to nab-paclitaxel, followed by epirubicin and cyclophosphamide [48]. The protocol initially included a window-phase for the first 117 of 174 patients randomized, in which durvalumab or placebo was administered two weeks prior to starting chemotherapy – but this was later eliminated due to concerns about delaying definitive therapy with placebo treatment. There was a significant increase in pCR with durvalumab in patients treated with a window-phase (61.0% versus 41.4%), but the pCR benefit of durvalumab was not significant in the intention to treat population (53.4% versus 44.2%). However, durvalumab was associated with significantly improved three-year DDFS (91.4% versus 79.5%) and OS (95.1% versus 83.1%) [49]. PD-L1 expression, TMB, and stromal TILs from pre-treatment biopsies were associated with response in both arms – further reinforcing that many standard immunotherapy biomarkers are not predictive in the neoadjuvant setting [48,50]. Increase in intratumoral TILs during the window phase was associated with response in the durvalumab arm alone – suggesting that successful recruitment of TILs is essential to immunotherapy response.

The randomized phase III KEYNOTE-522 trial is the largest completed study of immunotherapy in addition to NACT in early stage breast cancer, enrolling 1,174 patients [51]. Patients with untreated stage II or III TNBC received neoadjuvant therapy with paclitaxel and carboplatin followed by doxorubicin/epirubicin and cyclophosphamide. Patients were randomized 2:1 to pembrolizumab or placebo for 8 cycles concurrent with neoadjuvant treatment, with adjuvant pembrolizumab or placebo continued for up to 9 cycles post-operatively. Although first interim analysis demonstrated a significant improvement in pCR, the benefit has nearly halved on the third interim analysis inclusive of all enrolled patients, with a pCR rate of 63.0% with pembrolizumab versus 55.6% with placebo [52]. Immune-related toxicities included thyroid dysfunction (20.2%), severe skin toxicity (5.8%) – adrenal insufficiency was seen less frequently (2.6%) than in earlier neoadjuvant immunotherapy trials. In recently reported results from the July 2021 ESMO Virtual Plenary, KEYNOTE-522 met its EFS endpoint at the fourth interim analysis, with a three-year EFS of 84.5% in the pembrolizumab arm versus 76.8% in the placebo arm [53]. The EFS benefit was consistent across subgroups, including patients with or without lymph node involvement, and with PD-L1 positive or negative disease. Based on these results, the FDA approved pembrolizumab in combination with neoadjuvant chemotherapy, and as single agent adjuvant treatment for high-risk early stage TNBC. Although KEYNOTE-522 included carboplatin for all participants, the optimal chemotherapy backbone and role of platinum agents in neoadjuvant immunotherapy is still under further exploration. The phase Ib open label multicohort KEYNOTE-173 trial has explored treatment with a taxane with or without carboplatin, followed by standard AC, with pembrolizumab administered throughout neoadjuvant treatment [54]. Response rates were similar in platinum-containing and platinum-free arms, although sample size was small.

The phase III double blind IMpassion031 trial data provides further promising data for immunotherapy in early TNBC [55]. In this study, 333 patients with stage II–III TNBC were randomized in a 1:1 fashion to atezolizumab or placebo with neoadjuvant chemotherapy, consisting of 12 weeks of weekly nab-paclitaxel followed by AC. Rates of pCR were significantly higher with atezolizumab in the intention to treat population (58% versus 41%). Benefit was magnified in PD-L1 IC positive patients (69% versus 49%) although this comparison did not reach the predefined cutoff for significance. Additionally, lymph node positive patients derived greater benefit from atezolizumab in an exploratory subgroup analysis (pCR rate 57% versus 31%). EFS and OS data are currently immature, although the study was not formally powered for these endpoints. Atezolizumab had a tolerable safety profile and did not compromise the ability to receive chemotherapy, with rates of irAEs in line with other studies.

In summary, the large randomized phase III KEYNOTE-522 and IMpassion031 trials have demonstrated a pCR benefit with the addition of immunotherapy to standard neoadjuvant chemotherapy, and emerging data from GeparNuevo and KEYNOTE-522 suggest such approaches improve long-term outcomes. This has led to the groundbreaking approval of pembrolizumab in the neoadjuvant setting in combination with chemotherapy and in the adjuvant setting as a single agent. Although the inclusion criteria for the KEYNOTE-522 trial can serve as a guide for candidacy for neoadjuvant immunotherapy, only a subset of patients benefit, but no clear biomarker has emerged to identify this population. Lymph node positivity has been associated with greater improvements in pCR with neoadjuvant immunotherapy in some studies, but has not yet served as a clear discriminator of long-term outcomes, and further study is needed.

4. Hormone receptor positive breast cancer

Although immunotherapy has shown the most promise in patients with triple-negative disease, early immunotherapy approaches targeted HR positive disease (Table 1). Pre-clinical data demonstrated that interferon upregulated ER and PR expression, leading to trials of interferon in combination with tamoxifen, although results were disappointing with manageable but frequent toxicity [3,56]. The success of immune checkpoint inhibition in TNBC has led to the advent of similar approaches in HR+/HER2-breast cancer.

4.1. Advanced hormone receptor positive breast cancer

Early studies in HR+/HER2- breast cancer evaluated the use of checkpoint inhibitor monotherapy in metastatic breast cancer, with limited efficacy but occasional durable responses. The phase Ib JAVELIN trial evaluated the PD-L1 antagonist avelumab in a mixed cohort of metastatic triple-negative breast cancer patients, including 72 ER+ or PR+ and HER2- patients [57]. Only two patients (2.8%) responded in the HR+/HER2- subgroup. Response was not correlated with tumor cell PD-L1 expression, and but was correlated with PD-L1 expression of tumor associated immune cells. A better result was seen in the PD-L1 CPS ≥ 1% selected population included in HR+/HER2- cohort of the phase Ib KEYNOTE-028 study of pembrolizumab [13]. The 25 included patients were heavily pre-treated with a median of nine prior lines of therapy, and an overall response rate of 12% was seen, with an impressive median duration of response of 12 months.

Subsequent studies in the hormone receptor positive space have evaluated combination therapies, to capitalize on the potential benefit of checkpoint inhibition while attempting to improve the rate of responders. Two studies have evaluated pembrolizumab in combination with eribulin. An investigator initiated open-label phase II trial randomized 88 patients with ER+ breast cancer in a 1:1 fashion to eribulin plus pembrolizumab or eribulin alone [58]. Patients were required to have received at least two lines of endocrine therapy and up to two lines of chemotherapy. No significant difference was seen in ORR (27% vs 34%), DOR (1.5 months vs 2.1 months), or PFS (4.2 months vs 4.3 months). Crossover to pembrolizumab monotherapy was allowed, but of the 14 patients who elected to crossover, only one patient had clinical benefit, with stable disease for 4.2 months. Exploratory evaluation of PD-L1 status, TILs, and TMB did not identify a subgroup that benefitted from the addition of pembrolizumab, although only 24 patients had PD-L1 positive disease. Rates of grade 3 liver enzyme elevations were seen in 14% of patients treated with combination versus 7% of patients receiving eribulin monotherapy; otherwise immunotherapy adverse events consisted of rash and hypothyroidism at expected frequencies. The single arm phase IIa KELLY evaluated the same combination in 44 anthracycline, taxane, and hormone therapy pretreated patients who had received one or two chemotherapy regimens for metastatic disease [59]. ORR was 36%, with a median PFS of 6 months, slightly better than the results in the aforementioned randomized trial, although analysis of exploratory biomarkers is pending.

Murine models have demonstrated synergy between checkpoint inhibitors and CDK4/6 inhibitors – with the latter leading to pro-inflammatory changes to the tumor microenvironment, and with combination of the two agents leading to complete tumor regressions [60]. As a result, the combination has been tested in several studies, unfortunately with significant toxicity. A multicenter phase Ib study evaluated the safety and efficacy of abemaciclib and pembrolizumab in two cohorts [61,62]. In a cohort of 28 patients who had prior systemic chemotherapy for metastatic disease (but no more than two prior lines of treatment), a response was seen in 29% of patients, with a clinical benefit rate of 46%. Grade 3–4 AST elevations were seen in 18% of patients, and grade 3–4 ALT elevations in 11% of patients. A second cohort enrolled postmenopausal patients with untreated metastatic disease, and also included anastrozole in the treatment regimen [62]. In 26 enrolled patients, 19% had a confirmed response with a 77% disease control rate. However, 31% of patients experienced grade 3–4 transaminase elevations, and two fatal events of pneumonitis occurred, higher than would be expected with checkpoint inhibitor or CDK4/6 inhibition alone. Similarly, the phase II NEWFLAME trial enrolled patients with untreated metastatic disease, or those who progressed on one line of endocrine therapy, and evaluated treatment with nivolumab, abemaciclib, and letrozole or fulvestrant [63]. Despite regimen activity, 59% of patients experienced grade 3–4 liver toxicity, and three patients experienced pneumonitis with one fatal event. CheckMate 7A8 (NCT04075604) is investigating the combination of the CDK4/6 inhibitor palbociclib and anastrozole with or without nivolumab, although the concerning safety signals discussed above warrant careful monitoring to see if palbociclib based combinations carry the same risk of hepatitis and pneumonitis [64].

In summary, it is clear that a small proportion of HR+/HER2- patients benefit from immunotherapy given the double-digit response rates seen in KEYNOTE-028. However, for HR+/HER2- disease, immunotherapy is currently only approved for patients with high TMB or microsatellite instability as per tumor agnostic indications for pembrolizumab. Combination approaches with chemotherapy or CDK4/6 inhibitors have thus far been disappointing, but ongoing trials may identify promising combinations for appropriately selected patients. The Saci-IO HR+ (NCT04448886) trial is evaluating sacituzumab govitecan with or without pembrolizumab in patients with PD-L1+ disease, and the AMBITION (NCT04732598) trial is evaluating paclitaxel and bevacizumab with or without atezolizumab in patients with no prior chemotherapy for metastatic disease.

4.2. Early hormone receptor positive breast cancer

Although early stage HR+/HER2- disease carries a better prognosis than triple-negative cancers, rates of pCR to NACT are low. Several trials have evaluated the impact of neoadjuvant immunotherapy combinations on pathologic response, with data on long-term outcomes still accumulating (Table 2). The phase II GIADA evaluated the interplay between induction chemotherapy, immunotherapy, and endocrine therapy, with a novel regimen consisting of three cycles of neoadjuvant epirubicin and cyclophosphamide followed by eight cycles of exemestane and nivolumab. The study enrolled 43 premenopausal patients with stage II–IIIA breast cancer with a luminal B phenotype (Grade 3 or Ki-67 > 20%) [65]. All patients also received ovarian suppression with triptorelin concomitantly with neoadjuvant therapy. The primary outcome of pCR was not met, with a pCR rate of 16.3%. pCR was associated with higher pre-immunotherapy levels of CD3+/PD-1+, CD4+/FOXP3+, and CD68+/CD163+ cells, as well as an increase in CD8+ cells from before chemotherapy to after chemotherapy.

As described above, several reported arms of the I-SPY2 trial enrolled HR+/HER2- patients, all of whom had high-risk disease by the MammaPrint assay. In the pembro4 arm, the 40 patients with HR+/HER2- disease had an estimated pCR rate of 30%, compared to 13% in the 96 HR+/HER2- patients treated with standard chemotherapy, and the regimen graduated in the HR+/HER2-subgroup with predicted success in a phase III confirmatory trial [42]. In the pembro8-noAC arm, rates of pCR were identical between treatment and control patients with HR+/HER2-disease (15%) – representing preliminary promise of an anthracycline-free strategy, although disappointing single agent activity in early stage breast cancer [44]. In the durvalumab and olaparib arm of I-SPY2, 52 patients with HR+/HER2-disease received the experimental regimen, resulting in a pCR rate of 28% compared to 14% in control patients [43]. A pre-specified subgroup analysis suggested that HR+/HER2- patients in the MammaPrint ultra-high-risk group derived the most benefit from the addition of durvalumab and olaparib, with a pCR rate of 64% in 28 patients in this subgroup, compared to 22% in 49 controls. This MammaPrint subgroup is characterized by high proliferation gene expression and low ER pathway gene expression, and thus may be more immunologically similar to TNBC. It remains to be seen if the MammaPrint ultra-high-risk signature predicts immunotherapy benefit in other populations of HR+/HER2- disease.

Although neoadjuvant immunotherapy has shown promising improvements in response rates in several arms of I-SPY2, further study is needed to confirm this benefit and the impact on long-term outcomes. The randomized phase III KEYNOTE-756 (NCT03725059) and CheckMate 7FL (NCT04109066) trials are evaluating neoadjuvant and adjuvant pembrolizumab and nivolumab respectively in combination with standard chemotherapy, and may clarify the utility of immunotherapy in HR+/HER2- disease [66,67].

5. HER2+ breast cancer

Due to the observation that HER2+ breast cancer often has prominent immune infiltrates, several trials have evaluated checkpoint inhibition in HER2+ patients. Two trials have assessed PD-1 inhibitors in combination with trastuzumab. The phase Ib/II, single-arm PANACEA trial evaluated the combination of pembrolizumab and trastuzumab in 58 women with HER2+ breast cancer that had progressed on prior trastuzumab based therapy [14]. The trial enrolled PD-L1 positive patients but was later amended to include PD-L1 negative tumors. PD-L1 status was initially assessed with the QualTek PD-L1 assay and subsequently with the Dako 22C3 PD-L1 assay (with positivity defined as PD-L1 CPS ≥ 1). Of 46 PD-L1 positive patients, ORR was 15%, with a disease control rate (DCR) of 24%; none of the PD-L1 negative patients responded or had disease control. With re-analysis of PD-L1 status, seven patients who had tested positive with the QualTek assay were negative by the 22C3 assay; and none of these patients experienced a response. Correlative studies also found an association with stromal TILs and overall response. Rates of irAEs were as expected, although one patient passed away from treatment-related Lambert-Eaton syndrome. In the phase Ib CCTG IND 229 trial, 15 patients with prior treatment with trastuzumab, pertuzumab, and a taxane were treated with durvalumab and trastuzumab. Treatment was well tolerated aside from an episode of durvalumab-associated diabetic ketoacidosis, but no responses were seen. Other trials of checkpoint inhibitors in combination with trastuzumab are ongoing, including a randomized phase III trial of paclitaxel, trastuzumab, and pertuzumab with or without atezolizumab (NCT03199885).

Two trials have also investigated checkpoint inhibitors in combination with trastuzumab emtansine (T-DM1). The phase II double-blind KATE2 trial randomized patients in a 2:1 fashion to atezolizumab or placebo along with T-DM1 [68] (Table 2). Patients had previously received trastuzumab and a taxane, and were stratified by PD-L1 status, world region, and presence or absence of liver metastases. The primary outcome of PFS was not significantly different between arms, with a median of 8.2 months in the 133 patients receiving the combination and 6.8 months in the 69 patients receiving T-DM1 alone. A trend toward prolonged PFS was seen with atezolizumab in the PD-L1 IC+ patients (median 8.5 months versus 4.1 months), and a significantly improved PFS was seen in preplanned subgroups of patients with high PD-L1 gene expression as well as CD8+ cells at the invasive margin. Aside from expected adverse events, notable toxicities occurring more frequently with combination therapy included pyrexia in 35% of patients, leading to hospitalization in seven, grade 3 transaminitis in 8% of patients, and one episode of fatal hemophagocytic lymphohistiocytosis. A phase Ib trial of pembrolizumab and T-DM1 enrolled a similar cohort of 20 patients [69]. Transaminitis was again common, with 10% of patients experiencing grade 3 AST or ALT elevations. Overall response rate was 20% with a median progression free survival of 9.6 months. Exploratory biomarker analysis showed a trend toward higher response in PD-L1 CPS < 1 and TIL ≤ 10% populations, although the small sample sizes and unintuitive results prevent meaningful interpretation of this finding.

Several trials are evaluating checkpoint inhibitors in combination with neoadjuvant chemotherapy and HER2-directed therapy. The IMpassion050 trial randomized patients in a 1:1 fashion to atezolizumab or placebo, given concurrently with neoadjuvant AC followed by paclitaxel, trastuzumab, and pertuzumab [70]. Atezolizumab or placebo was continued adjuvantly along with either trastuzumab and pertuzumab or T-DM1 to complete 52 weeks total treatment. However, enrollment was stopped after 454 patients due to an unfavorable risk-benefit profile. Response rates favored the placebo arm in both the ITT (62.4% with atezolizumab versus 62.7% with placebo) and the PD-L1 IC ≥ 1% population (64.2% with atezolizumab versus 72.5% with placebo). Serious adverse events were more common with atezolizumab (19.5% versus 13.3%), including five deaths in the atezolizumab arm, two of which were deemed treatment related.

Immunotherapy has demonstrated activity in select patients with HER2+ disease, but further clinical study is needed before such treatment can be implemented in practice. In the metastatic setting, the KATE2 trial demonstrated trend toward PFS benefit with the addition of atezolizumab to T-DM1 in PD-L1 positive subgroup, and the KATE3 trial (NCT04740918) will evaluate the same regimen in PD-L1 positive patients. For early stage disease, the ongoing phase III APTneo (NCT03595592) study is comparing paclitaxel, carboplatin, trastuzumab, and pertuzumab (TCHP), atezolizumab with TCHP, and atezolizumab with the regimen of AC followed by TCHP. However, given the results of IMpassion050, immunotherapy appears to have minimal benefit in early stage HER2+ disease, and further biomarker analysis is needed to identify if a select subgroup might derive greater benefit.

6. Novel checkpoint inhibitor strategies

6.1. Other checkpoint inhibitor agents

Dual checkpoint inhibition with PD-1/PD-L1 and cytotoxic T-lymphocytic antigen 4 (CTLA-4) inhibition has demonstrated improved efficacy but increased toxicity as compared to single agent immunotherapy in melanoma; studies of combinations of checkpoint inhibitors are limited in breast cancer. A pilot study of durvalumab with the CTLA-4 inhibitor tremelimumab demonstrated an ORR of 43% in the seven patients with previously treated mTNBC, and no response in 11 patients with metastatic ER+ disease [71]. Nearly all patients experienced some degree of hepatotoxicity with the combination. All responding patients had a DOR of at least 10 months at the time of data cutoff. One cohort of the phase II DART trial evaluated dual checkpoint inhibition with ipilimumab and nivolumab in 17 patients with metaplastic breast cancer, with responses seen in three patients [72]. All responses have been ongoing for at least 11 months, with the most common adverse events being hepatotoxicity and fatigue.

A number of novel checkpoint inhibitors are in development for breast cancer. Eftilagimod alpha is a LAG-3 immunoglobulin fusion protein, which stimulates dendritic cells to boost T-cell immune response. Eftilagimod alpha was first tested in a phase I/II trial in combination with paclitaxel as first line chemotherapy for metastatic breast cancer with no significant adverse events, and a 50% objective response rate in 30 patients with predominantly HR+/HER2- disease [73]. A subsequent double blind phase II trial, AIPAC, randomized 227 patients with metastatic HR+ breast cancer to paclitaxel + eftilagimod alpha or placebo in a 1:1 fashion [74]. Although six-month PFS was 63% with the combination versus 54% with paclitaxel alone, neither PFS nor OS were significantly prolonged with the combination. Several predefined subgroups had more meaningful benefit from the combination, including patients younger than 65, those with luminal B cancers, and those with lower monocyte counts at baseline. Another LAG-3-directed therapy, REGN3767, is currently under investigation in combination with cemiplimab as neoadjuvant treatment in one arm of the I-SPY2 trial (NCT01042379). Although much of the success of immunotherapy has focused on the blockade of inhibitory immune checkpoints, agonists of costimulatory checkpoints are also under investigation. The I-SPY2 trial investigated SD-101, an injectable intratumoral toll like receptor-9 agonist in combination with pembrolizumab. In the recently presented data, this arm of I-SPY2 reported similar rates of pCR with this combination of 44% in TNBC and 26% in HR+/HER2-, which did not exceed the rates seen in the pembro4 arm [75].

Although at this point checkpoint inhibitors directed toward targets other than PD-1/PD-L1 have no clear clinical indication in breast cancer treatment, the number of promising agents in this space is rapidly growing. Several investigational agonists of 4–1BB, a stimulatory T-cell receptor, are being evaluated in breast cancer, including the monoclonal antibody utomilumab (AVIATOR – NCT03414658), and the bifunctional HER2/4-1BB fusion protein PRS-343 (NCT03650348). T cell immunoglobulin and ITIM domain (TIGIT) is an inhibitory checkpoint expressed on T-cells, which has shown promising activity in lung cancer, and is being evaluated in TNBC (NCT04584112).

6.2. Combination with poly (ADP-ribose) polymerase (PARP) inhibitors

Aside from the durvalumab and olaparib arm of I-SPY2, several other clinical trials have evaluated the combination of immune checkpoint and PARP inhibition in breast cancer. Preclinical evidence from BRCA-deficient models of TNBC suggest that PARP inhibitors increase cytoplasmic DNA and activate STING (stimulator of interferon genes), thus increasing type 1 interferon and T cell intratumoral infiltration [76]. PARP inhibitors also upregulate PD-L1 expression regardless of underlying homologous recombination deficiency [77]. The open-label, single arm phase Ib/II TOPACIO/KEYNOTE-162 trial evaluated the safety and efficacy of the combination of the PARP inhibitor niraparib and pembrolizumab in 55 patients with previously treated advanced or metastatic TNBC [78]. The combination was safe, and in 47 patients evaluable for efficacy, the ORR was 21% and DCR was 49%. In a subset of 15 patients with tumor BRCA mutations, seven patients (47%) had at least a partial response and 12 (80%) had stable disease or better. PFS was prolonged in patients with BRCA mutations (median 8.3 months versus 2.1 months in BRCA wild-type patients).

MEDIOLA was a multicenter open-label phase I/II basket study of olaparib plus durvalumab in patients with germline BRCA mutations, which included 34 patients with breast cancer [79]. Similar response rates were seen in the 17 TNBC (ORR 59%) and 13 HR+ (ORR 69%) patients with evaluable disease. Median PFS was longer in the 13 patients with HR+ disease (9.9 months versus 4.9 months in TNBC) although OS was similar in both subgroups (22.4 months versus 20.5 months). The combination was well tolerated, with a similar profile of immunotherapy-related adverse events as in prior studies of durvalumab monotherapy, with no reports of pneumonitis or myelodysplastic syndrome.

It is still uncertain whether the addition of immunotherapy to PARP inhibitors improves outcomes, given comparable response rates and PFS with the above studies and the OlympiAD trial of olaparib monotherapy in patients with a germline BRCA mutation [80]. However, cross trial comparisons can be highly flawed, and the additive benefit of immunotherapy in combination with PARP inhibition in patients with germline or somatic BRCA mutations is under evaluation in a randomized phase II trial (NCT02849496) [81]. The combination is also under evaluation as maintenance therapy after induction chemotherapy regardless of mutational status in the randomized phase II/III KEYLYNK-009 trial (NCT04191135) [82].

6.3. Combination with small molecule inhibitors

A variety of small molecule inhibitors have been evaluated in combination with immunotherapy, due to preclinical evidence suggesting modulatory effects on the tumor microenvironment. Based on the hypothesis that alterations in the PI3K/AKT/PTEN pathway are linked to resistance to immunotherapy [83], a phase Ib trial (CO40151) investigated the combination of atezolizumab plus ipatasertib and paclitaxel or nab-paclitaxel as first line therapy for 114 patients with advanced TNBC [84]. An ORR of 54% was seen with a median PFS of 7.2 months, comparable to the results of IMpassion130 – with no consistent trend when patients analyzed by PD-L1 status or presence of PI3K/AKT/PTEN mutation.

Data from preclinical models has shown that MAPK/extracellular signal-related kinase (MEK) inhibition can overcome innate resistance to taxanes [85] as well as upregulate PD-L1 [86]. The randomized phase II COLET trial investigated the addition of the MEK inhibitor cobimetinib to atezolizumab plus paclitaxel or nab-paclitaxel in patients with advanced or metastatic TNBC [87]. Response rates were similar in patients who received cobimetinib with paclitaxel (ORR 38.3%), cobimetinib with atezolizumab and paclitaxel (ORR 34.4%), or cobimetinib with atezolizumab and nab-paclitaxel (ORR 29.0%). A retrospective pooled biomarker analysis in patients with PD-L1+ disease in the atezolizumab cohorts demonstrated a trend toward improved PFS and ORR.

Treatment with the multikinase inhibitor lenvatinib may induce immune modulatory effects in the tumor microenvironment, reducing inhibitory tumor associated macrophages and increasing type 1 interferon signaling [88]. The phase II LEAP-005 trial evaluated the combination of lenvatinib and pembrolizumab in 31 patients with previously treated mTNBC, with an encouraging ORR of 29% for a chemotherapy free regimen with no unexpected toxicities [89].

6.4. Combination with local therapies

Radiation has long been considered an immune stimulator, by generating neoantigens and promoting the presentation of antigens to immune effector cells, which can result in responses in non-irradiated disease, termed the abscopal effect [90]. This has been evaluated in combination with checkpoint inhibition for breast cancer in a handful of trials. The most promising data comes from a single arm phase II trial of pembrolizumab within three days of the first fraction of 30 Gy radiotherapy for mTNBC, resulting in an objective response in unirradiated lesions of 17.6% with three complete responses [91]. A randomized phase II trial of pembrolizumab administered 2–7 days prior to 5 fractions of 4 Gy radiotherapy to patients with metastatic HR+ breast cancer failed to demonstrate objective responses in the first 8 patients enrolled [92]. Cryoablation similarly stimulates an immune response through local tumor disruption, and is posited to more effectively release intact neoantigens. Two pilot studies have examined pre-operative cryoablation in combination with ipilimumab with or without nivolumab, finding that combination therapy did not delay surgery, and resulted in elevations in peripheral activated and proliferating CD4 and CD8 + T-cells [93,94]. A randomized phase II trial (NCT03546686) is further assessing the efficacy of this treatment approach [95].

6.5. Other combination treatments

Although checkpoint inhibitor immunotherapy can produce durable responses in a minority of patients with breast cancer, strategies to promote responses in a greater proportion of patients are under active investigation. Cabiralizumab is a novel antibody directed toward colony stimulating factor-1 receptor, theoretically blocking the activation and survival of inhibitory tumor-associated macrophages, allowing for a more robust immune response by altering the tumor microenvironment. Cabiralizumab is being studied in combination with nivolumab and neoadjuvant carboplatin and paclitaxel in a phase I/II study (NCT04331067). Hypomethylating agents have the potential to increase antigen expression on cancer cells and suppress immune regulatory cells and have been studied in combination with checkpoint inhibition. In a phase II multicohort study, 9 patients with heavily pre-treated ER+/HER2- breast cancer received CC-486 (oral azacitidine) and durvalumab [96]. No responses were seen in any cohort, and this study was terminated early due to lack of efficacy. The VEGF inhibitor bevacizumab may have immunomodulatory effects through normalization of tumor vasculature. A phase II study of bevacizumab, nivolumab, and paclitaxel in 56 patients with untreated metastatic HR+ or triple-negative breast cancer demonstrated an ORR of 70% and median PFS of 14.8 months [97].

Several oncolytic viruses have been developed that selectively infect tumor cells after intratumoral injection, promoting tumor lysis and adaptive anti-tumor immunity, which may be synergistic with checkpoint inhibitors. Talimogene laherparepvec was evaluated in combination with atezolizumab in a phase Ib trial, which enrolled 8 patients with TNBC and hepatic lesions, with treatment complicated by one grade 3 hepatic hematoma and abdominal infection, and resulting in one confirmed partial response [98]. A window of opportunity study of atezolizumab and the oncolytic virus pelareorep demonstrated improvement in a combined metric of decrease in tumor cellularity and increase in TILs, as well as improvement in PD-L1 expression in 7 ER+ HER2- patients treated with the combination [99]. The DNA plasmid tavokinogene telseplasmid induces local interleukin-12 production within the tumor microenvironment after intratumoral injection and electroporation. Preliminary results from a phase II study documented an ORR of 27.3% in 11 patients with advanced TNBC treated with tavokinogene telseplasmid in combination with pembrolizumab [100].

7. Conclusion

Checkpoint inhibitor immunotherapy has been evaluated in all subtypes of breast cancer, as well as part of treatment for both early and advanced disease. Initial studies of checkpoint inhibitor monotherapy in patients with advanced disease demonstrated responses in a small fraction of patients, enriched in patients with PD-L1 positive TNBC and no prior chemotherapy. However, the randomized KEYNOTE-119 trial failed to demonstrate a long-term outcome benefit of pembrolizumab monotherapy compared to physician’s choice of chemotherapy for advanced TNBC in the second or third line setting. The role of monotherapy with checkpoint inhibitors outside of patients with tumor-agnostic indications of high TMB or microsatellite-instability remains uncertain.

Other studies in the advanced setting have evaluated combinations of chemotherapy and immunotherapy. Atezolizumab was the first checkpoint inhibitor to receive FDA and European Commission accelerated approval for advanced TNBC in combination with nab-paclitaxel for PD-L1 positive disease, but approval has been withdrawn in the U.S. given the conflicting results seen in a similar population treated with atezolizumab and paclitaxel in IMpassion131. Conversely, pembrolizumab has regular approval for PD-L1 positive advanced TNBC in combination with nab-paclitaxel, conventional paclitaxel, or gemcitabine and carboplatin – with similar efficacy regardless of chemotherapy partner seen in KEYNOTE-355. Identifying appropriate candidates for treatment with immunotherapy remains challenging, with the availability of multiple approved companion assays for PD-L1 testing, differing rates of PD-L1 expression at different biopsy sites, and the manual interpretation of staining required to assess positivity which leaves room for ambiguity. In advanced HR+/HER2- and HER2+ disease, the addition of checkpoint inhibitors to chemotherapy or HER2-directed therapy has not yet yielded conclusive improvements in outcome, and further study is ongoing to identify treatment regimens that may potentiate immunotherapy response in these populations.

Checkpoint inhibitors have also been studied extensively in combination with standard neoadjuvant chemotherapy, particularly in patients with triple-negative disease, and most trials have demonstrated an improvement in pCR with immunotherapy. The largest completed trial, KEYNOTE-522, has demonstrated a decline in the pCR benefit of pembrolizumab at subsequent interim analyses, but has recently met its EFS endpoint, leading to the FDA approval of pembrolizumab for high-risk early stage TNBC. The use of carboplatin in KEYNOTE-522 adds further complexity to selecting an appropriate treatment strategy – as other studies of carboplatin in early stage TNBC have failed to demonstrate an EFS benefit. Preliminary data from the I-SPY2 trial suggest that similar approaches may also have promise for HR+/HER2- breast cancer, and randomized trials are ongoing. However, unlike in the metastatic setting, PD-L1 status has not selected for responders in neoadjuvant trials. Identification of patients who truly benefit from immunotherapy is paramount given the potential for permanent toxicity of checkpoint inhibitors such as adrenal insufficiency and hypophysitis that can permanently affect the quality of life in patients with a potentially curable cancer.

Although a number of trials have demonstrated improved responses and survival with immunotherapy in breast cancer, only a minority of patients benefit from PD-1/PD-L1 blockade. Combination of immunotherapy with radiotherapy, small molecule inhibitors, or injectable treatments that directly alter the tumor microenvironment may lead to responses in additional patients. Targeting other immune checkpoints, including LAG-3, CTLA-4, and TIGIT, may also potentiate responses. Nonetheless, checkpoint inhibitors have begun to shift the landscape of breast cancer treatment, contributing to steady improvements in prognosis for TNBC where the need for novel therapies has been the greatest.

8. Expert opinion

The checkpoint inhibitors pembrolizumab and atezolizumab both received accelerated approval for the treatment of advanced PD-L1 positive TNBC in combination with chemotherapy, but atezolizumab’s approval in the United States has been voluntarily withdrawn. The conflicting results of the IMpassion130 and IMpassion131 trial results raise questions about about the efficacy of paclitaxel in combination with immunotherapy, but the equivalence of chemotherapy partners (paclitaxel, nab-paclitaxel, and gemcitabine/carboplatin) in KEYNOTE-355 should give clinicians flexibility to personalize treatment. Alternative chemotherapeutic agents or dosages may better potentiate immune responses, and combinations with the antibody-drug conjugate sacituzumab govitecan are of particular interest given its efficacy in refractory TNBC. All patients with advanced TNBC should have PD-L1 testing of their tumor with the Dako 22C3 assay to determine if they are candidates for pembrolizumab-based treatment. Aside from tumor biology, treatment line also impacts response to immunotherapy, and every effort should be made to include checkpoint inhibitors as part of frontline treatment for eligible patients.

Not all patients respond to immunotherapy, and a number of strategies are under evaluation to enhance responses to checkpoint inhibition. Novel checkpoint inhibitors/agonists are in development to overcome resistance to anti-PD-1/PD-L1 immunotherapy. Tiragolumab, an anti-TIGIT monoclonal antibody, is under investigation in breast cancer, and has shown particular promise in non-small cell lung cancer, where response rates doubled with tiragolumab and atezolizumab compared to atezolizumab alone [101]. Several intratumoral injections that directly alter the microenvironment are under investigation, and preliminary data on Il-12 encoding tavokinogene telseplasmid suggest the agent can induce responses in combination with pembrolizumab even in patients with PD-L1 negative disease. Initial data on the combination of lenvatinib and pembrolizumab in the phase II LEAP-005 trial are also intriguing, with an ORR of 29% in patients with heavily pre-treated disease with a chemotherapy-free regimen. Further study is ongoing with these agents in advanced TNBC, although hopefully success in this space will translate to other breast cancer subtypes, as there is a need for effective immunotherapy approaches in HR+ and HER2+ breast cancer.

The use of checkpoint inhibitors in combination with standard neoadjuvant therapy has yielded encouraging results in some but not all trials, with a few notable exceptions. The lack of benefit seen in NeoTRIPaPDL1 may possibly be driven by the lack of anthracycline in the pre-operative setting – and an interplay between topoisomerase inhibitors and immunotherapy is also suggested by the TONIC trial in the metastatic setting. However, KEYNOTE-522 has now documented improvements in both co-primary endpoints of pCR and EFS, leading to the approval of pembrolizumab as part of neoadjuvant and adjuvant treatment, perhaps soon to be followed by atezolizumab as results from IMpassion031 mature. Given the consistent benefit across tumor stage, lymph node status, and PD-L1 status in KEYNOTE-522, this therapy should be considered for most patients with stage II or III TNBC. Further study is also needed to determine the role if any of immunotherapy in Stage I disease, and as responses continue to improve in TNBC, it will be important to identify if select subsets of patients can receive immunotherapy as a way to de-escalate the chemotherapy backbone of treatment. A number of adjuvant trials of immunotherapy are ongoing, and treatment based on residual disease may allow for treatment intensification in patients at highest risk of recurrence, rather than carte blanche application prior to neoadjuvant therapy. Given the excellent prognosis of patients with pCR, extended adjuvant immunotherapy can likely be reserved for patients with residual disease at the time of surgery. Since KEYNOTE-522 did not allow adjuvant capecitabine or olaparib, it will be important to answer how to best integrate immunotherapy with other promising adjuvant treatments for patients with residual disease. Although immunotherapy is most well-studied in early stage TNBC, two arms of the I-SPY2 trial have shown promising results in HR+/HER2- disease – especially in those with the MammaPrint ultra-high-risk signature, perhaps due to homology with basal-like TNBC. The phase III KEYNOTE-756 and CheckMate 7FL trials will clarify the role of immunotherapy for early stage HR+/HER2- cancers.

The treatment paradigm for breast cancer has shifted dramatically over the past few years, and immunotherapy will continue to play a key role in providing durable responses with manageable toxicity for an increasing number of patients. A number of large, randomized trials will answer questions regarding the optimal chemotherapy partners, and the benefit of immunotherapy in both early and metastatic breast cancer (Table 3). Response prediction in both the early and metastatic setting will likely be refined as ongoing translational work refines our understanding of immunotherapy. Given the potential for irreversible lifelong irAEs, predictive biomarkers are urgently needed for early-stage disease. Quantification of the spatial relationship between immune cell subsets using multiplex immunohistochemistry may better predict response than traditional biomarkers that can fail to capture intratumoral heterogeneity [102]. Artificial intelligence approaches have also been applied to identify features of the tumor microenvironment evident from digital histology samples that predict immunotherapy responsiveness in melanoma [103]. Similar approaches can predict gene expression and other latent features in breast cancer histology and may prove useful for biomarker development [104–106], although artificial intelligence must be applied equitably to avoid recapitulating the disparities present in breast cancer care [107]. Furthermore, as highly accurate models of response are developed for standard neoadjuvant chemotherapy using clinical data and imaging features [108,109], immunotherapy can be reserved for patients predicted to do poorly with standard regimens, avoiding potential lifelong toxicities in patients who do not require treatment escalation. Circulating tumor DNA is also prognostic in the neoadjuvant setting, and could be leveraged to better tailor therapies to patient risk [110]. The explosion of novel agents and combinations will lead to beneficial therapeutics for more patients, as we usher in an era of individualized strategies based on multi-omic data for true personalized medicine.

Table 3. Select ongoing randomized clinical trials of checkpoint inhibitors in breast cancer. TNBC = triple negative breast cancer. HR = hormone receptor. HER2 = human epidermal growth factor receptor 2. RT = radiation therapy. NAC = neoadjuvant chemotherapy

Study	Treatment	Patient Population	Phase
NCT01898117 Triple-B	Carboplatin + Cyclophosphamide ± Atezolizumab versus Paclitaxel ± Atezolizumab	Advanced TNBC	Phase II
NCT03164993 ALICE	Cyclophosphamide + Pegylated Liposomal Doxorubicin ± Atezolizumab	Advanced TNBC	Phase II
NCT04177108	Paclitaxel ± Atezolizumab ± Ipatasertib	Advanced TNBC	Phase III
NCT03371017 IMpassion132	Gemcitabine + Carboplatin + Capecitabine ± Atezolizumab	Advanced TNBC	Phase III
NCT04191135 KEYLYNK-009	Pembrolizumab + Carboplatin + Gemcitabine → Olaparib versus Chemotherapy Maintenance	Advanced TNBC	Phase II/III
NCT04683679	RT + Pembrolizumab ± Olaparib	Advanced TNBC	Phase II
NCT04468061 Saci-IO TNBC	Sacituzumab govitecan ± Pembrolizumab	Advanced TNBC	Phase II
NCT03567720	Tavokinogene telseplasmid + Pembrolizumab ± nab-paclitaxel	Advanced TNBC	Phase II
NCT02755272	Carboplatin + Gemcitabine ± Pembrolizumab	Advanced TNBC	Phase II
NCT03167619 DORA	Olaparib ± Durvalumab	Advanced TNBC	Phase II
NCT03616886 SYNERGY	Paclitaxel + Carboplatin + Durvalumab ± Oleclumab	Advanced TNBC	Phase I/II
NCT03742102 BEGONIA	Durvalumab + Paclitaxel ± Experimental Agents	Advanced TNBC	Phase Ib/II
NCT04159818 TONIC-2	Nivolumab ± Induction Cisplatin or Doxorubicin	Advanced TNBC	Phase II
NCT02849496	Olaparib ± Atezolizumab	Advanced TNBC, germline/somatic BRCA	Phase II
NCT03281954 GeparDouze	NAC ± Atezolizumab	Early TNBC	Phase III
NCT03498716 IMpassion030	Adjuvant Chemotherapy ± Atezolizumab	Early TNBC	Phase III
NCT02883062	NAC ± Atezolizumab	Early TNBC	Phase II
NCT04443348	Pembrolizumab + NAC ± RT	Early TNBC	Phase II
NCT02954874	Adjuvant Pembrolizumab versus Observation	Early TNBC	Phase III
NCT03872505 PANDoRA	Carboplatin + Paclitaxel + Durvalumab ± RT	Early TNBC	Phase II
NCT03818685 BreastImmune03	Adjuvant Capecitabine versus Nivolumab + Ipilimumab	Early TNBC	Phase II
NCT04331067	NAC + Nivolumab ± Cabiralizumab	Early TNBC	Phase Ib/II
NCT03487666 OXEL	Adjuvant Nivolumab, Capecitabine or Both	Early TNBC	Phase II
NCT04732598 AMBITION	Paclitaxel + Bevacizumab ± Atezolizumab	Advanced HR+	Phase III
NCT04448886 Saci-IO HR+	Sacituzumab Govitecan ± Pembrolizumab	Advanced HR+	Phase II
NCT03409198 ICON	Pegylated Doxorubicin + Cyclophosphamide ± Ipilimumab and Nivolumab	Advanced HR+	Phase II
NCT03095352	Carboplatin ± Pembrolizumab	Advanced HR+ or TNBC	Phase II
NCT04740918 KATE3	T-DM1 ± Atezolizumab	Advanced HER2+	Phase III
NCT03199885	Paclitaxel + Trastuzumab + Pertuzumab ± Atezolizumab	Advanced HER2+	Phase III
NCT03515798 PELICAN	NAC ± Pembrolizumab	HER2- Inflammatory Breast Cancer	Phase II
NCT03725059 KEYNOTE-756	NAC ± Pembrolizumab	Early HR+	Phase III
NCT04109066 CheckMate 7FL	NAC ± Nivolumab	Early HR+	Phase III
NCT04075604 CheckMate 7A8	Neoadjuvant Anastrazole + Pablociclib ± Nivoumab	Early HR+	Phase II
NCT03875573 Neo-CheckRay	NAC + RT ± Durvalumab ± Oleclumab	Early Luminal B	Phase II
NCT03356860 B-IMMUNE	NAC ± Durvalumab	Early Luminal B	Phase II
NCT03726879 IMpassion050	NAC ± Atezolizumab	Early HER2+	Phase III
NCT03595592 APTneo	NAC ± Atezolizumab	Early HER2+	Phase III
NCT03747120 neoHIP	NAC + Pertuzumab, Pembrolizumab, or Both	Early HER2+	Phase II

Article highlights

• Response to checkpoint inhibition in advanced breast cancer is enriched in patients with untreated, PD-L1 positive, triple-negative disease

• Atezolizumab and pembrolizumab were approved for the treatment of PD-L1 positive triple-negative breast cancer in combination with chemotherapy on the basis of the IMpassion130 and KEYNOTE-355 trials, although the U.S. approval for atezolizumab was withdrawn given the conflicting results of IMpassion131

• Pembrolizumab was recently approved in combination with neoadjuvant chemotherapy and as a single agent in the adjuvant setting for early-stage high-risk triple-negative breast cancer, based on improvements in pathologic complete response and event-free survival seen in KEYNOTE-522

• Unlike in the metastatic setting, PD-L1 status and other immunotherapy biomarkers have not identified patients with early-stage disease who benefit from immunotherapy

• Ongoing studies are evaluating PD-1/PD-L1 blockade in combination with novel immune checkpoint inhibitors, PARP inhibitors, small-molecule inhibitors, and intratumoral injections of immunomodulatory therapeutics

This box summarizes key points contained in the article.

Declaration of interest

R Nanda is on Advisory Boards for Aduro, Cardinal Health, Clovis, Fujifilm, G1 Therapeutics, Genentech, Immunomedics/Gilead, Ionis, iTeos, MacroGenics, Merck, Oncosec, Pfizer, and Seattle Genetics. N Nanda is also on the data safety monitoring board for G1 Therapeutics.

R Nanda also receives funding from Arvinas, AstraZeneca, Celgene, Corcept Therapeutics, Genentech/Roche, Immunomedics/Gilead, Merck, OBI Pharma, Odonate Therapeutics, OncoSec, Pfizer, Seattle Genetics, Taiho

A Pearson is on the Advisory board for Prelude Therapeutics

The authors have no other relevant affiliations or financial involvement with any organization or entity with a financial interest in or financial conflict with the subject matter or materials discussed in the manuscript. This includes employment, consultancies, honoraria, stock ownership or options, expert testimony, grants or patents received or pending, or royalties.

Additional information

Funding

All authors received salary support from the University of Chicago. F Howard also reports support from NIH grants 1F32CA265232-01 and 5T32CA009566-33.