Immunological Approaches to Breast Cancer Therapy

Immunology of Breast Cancer

The association of breast tumor immune cell (IC) infiltration and prognosis has been appreciated for decades. Given the correlation between tumor infiltrating lymphocytes (TILs), treatment response, and outcome, guidelines have been developed to enable characterization (stromal vs. intratumoral) and quantification of TILs. Although cytotoxic CD8 positive T cells are perhaps the most well-recognized TIL, a number of other ICs contribute to antitumor response, including natural killer cells and CD4 positive T helper 1 (Th1) cells, which stimulate cytotoxic T cells through interferon (IFN)-γ. Conversely, several ICs inhibit the immune response, including regulatory T cells (Treg) through the secretion of cytokines transforming growth factor-beta (TGF-β) and interleukin-10 (IL-10), as well as tumor-associated macrophages (TAMs) and myeloid-derived suppressor cells (MDSCs).

Hormone receptor positive (HR+) breast cancer is the predominant subtype but features the lowest concentration of TILs, contributing to the historic notion that breast cancer would be unresponsive to immunotherapy. Indeed, early studies of immunotherapy in breast cancer used IFN and ILs in HR+ cancers based on in vitro data suggesting such treatment sensitized tumors to hormonal therapy, but with disappointing clinical efficacy. Recognition of the differences in the tumor immune microenvironment (TIME) between breast cancer subtypes has led to renewed interest in immunotherapy, especially in triple-negative breast cancer (TNBC), which is characterized by higher levels of TILs, higher programmed death ligand 1 (PD-L1) expression, and greater tumor mutational burden (TMB).

Immunotherapy Treatment Strategies

The success of checkpoint blockade in melanoma has led to the development of a number of drugs targeting inhibitory interactions between cancer cells and lymphocytes. In essence, checkpoint proteins on healthy cells and lymphocytes keep immune responses in check and prevent autoimmunity. Cancer cells commandeer this pathway and evade detection through expression of checkpoint proteins or recruitment of inhibitory cells to the tumor microenvironment. The first US Food and Drug Administration (FDA)-approved checkpoint inhibitor was ipilimumab, a monoclonal antibody targeting cytotoxic T-lymphocyte–associated antigen 4 (CTLA-4) on cytotoxic T cells, preventing the inhibitory interaction with B7-1 (CD80) and B7-2 (CD86) on antigen-presenting cells and instead promoting the costimulatory interaction of CTLA-4 with CD28 ( Fig. 63.1 ). Thus CTLA-4 blockade leads to effector T-cell proliferation and inflammatory cytokine production. Subsequently, immune-mediated cancer cell elimination is regulated by PD-L1 on inhibitory ICs or tumor cells (TCs) and PD-1 on T cells. Targeting this interaction has been widely successful across numerous cancer types with multiple FDA-approved agents, including the anti-PD-1 monoclonal antibodies pembrolizumab, nivolumab, and cemiplimab and the anti-PD-L1 antibodies atezolizumab, avelumab, and durvalumab. Another checkpoint target in breast cancer, lymphocyte activation gene-3 (LAG-3), is expressed by multiple ICs, enhancing the activity of Tregs through interaction with MHC class II on antigen-presenting cells and inhibiting the activity of cytotoxic T cells through interaction with galectin-3 and other proteins expressed by breast cancer cells and stroma. A number of other novel immune checkpoints have been under clinical and preclinical investigation in breast cancer, including inhibitors of T-cell immunoreceptor with Ig and ITIM domains (TIGIT), T-cell immunoglobulin and mucin-domain containing-3 (TIM-3), V-domain immunoglobulin suppressor of T-cell activation (VISTA), and agonists of the costimulatory receptors OX40 and 4-1BB.

Although the excitement surrounding the use of immunotherapy in breast cancer is palpable, the clinical utility of such therapy remains limited to a fraction of patients. A number of strategies have been developed to alter the homeostasis of the tumor microenvironment, promoting responses in increasing number of patients. One of the most longstanding examples of such is the abscopal effect, where a response is seen in untreated tumors after radiation to one site of disease. Radiotherapy is posited to increase the presentation of tumor neoantigens, and systemic responses are associated with a marked decline in peripheral MDSCs, highlighting the potential impact on the tumor microenvironment. Aside from the cytotoxic nature of chemotherapy, individual treatment agents may have differing impacts on the tumor microenvironment. Cyclophosphamide is known to reduce Tregs, although this effect may be transient. Anthracyclines induce a type I IFN response in tumors, in part due to release of RNA from dying cells and toll-like receptor-3 (TLR3) signaling. Preclinical data suggest that other chemotherapeutic agents have a diverse impact on various aspects of the tumor microenvironment, suggesting a strong rationale for synergistic benefits of the combination of chemotherapy with immunotherapy. This hypothesis was evaluated in the phase II TONIC trial, which explored nivolumab treatment alone or after induction with radiation, cyclophosphamide, doxorubicin, or cisplatin. The overall response rate (ORR) was 8% in the radiation and cyclophosphamide arms, 17% without induction, 23% with cisplatin, and 35% with doxorubicin. Treatment with doxorubicin and cisplatin was associated with increased T-cell infiltration, as well as upregulation of inflammatory and IFN-γ gene signatures.

A number of small-molecule inhibitors with antitumor activity may also amplify the impact of immunotherapy due to alterations in the tumor microenvironment. Poly (ADP-ribose) polymerase (PARP) inhibitors, such as olaparib and niraparib, aside from preventing PARP-mediated DNA repair, also upregulate PD-L1 on TCs through GSK3β inactivation. The multikinase inhibitor lenvatinib is known to deplete inhibitory TAMs through inhibition of FGFR and VEGFR, promoting a more robust immune response. AKT inhibitors such as ipatasertib selectively deplete Tregs and may overcome PI3K/AKT mediated resistance to cytotoxic chemotherapy.

Beyond traditional drug therapy, a number of alternative approaches to immune stimulation are under investigation. There is a long history of attempting to stimulate immune response with anticancer vaccination in breast cancer, but results have thus far been disappointing, with no product near FDA approval. Stimulators of innate immunity including TLR agonists and stimulator of interferon gene (STING) agonists have been used as vaccine adjuvants and are also under evaluation in combination with other immunotherapeutics as intratumoral injections. Oncolytic viruses, such as talimogene laherparepvec (T-VEC), which was FDA-approved for melanoma in 2015, are under investigation in breast cancer. T-VEC is an engineered weakened herpes virus with a gene added for human GM-CSF production. The viral particles are injected into a tumor, resulting in replication and eventual tumor lysis, releasing GM-CSF when TCs burst, attracting dendritic cells (DCs) to present the released tumor antigens to effector T cells. Although most immunotherapy approaches rely on TILs or other ICs in the tumor microenvironment, recent attempts at engineering T cells to target malignancy have shown success in hematologic malignancies. Chimeric antigen receptor expressing T cells (CAR-T cells) are modified T cells expressing an antitumor antibody linked to an intracellular signaling domains, including the CD3ζ subunit of the T-cell receptor and other costimulatory moieties. Recognition of the target tumor antigen leads to intracellular signaling and T-cell proliferation, TC cytotoxicity, and sustained survival.

There is increasing recognition of the microbiome as a mediator of checkpoint blockade efficacy across multiple cancer types. In melanoma patients, fecal microbiome diversity is associated with PD-1 inhibitor response, and in a case control study, postmenopausal patients with breast cancer were noted to have a lower diversity of gut microbiome compared to controls. Supplementation with Bifidobacterium animalis lactis (EDP1503) in combination with pembrolizumab is under evaluation in TNBC, and treatment was well tolerated, with a promising response rate of 33% in the first six evaluable patients treated. T-cell infiltration following the use of an over-the-counter probiotic supplement prior to surgery was investigated in an exploratory trial of seven patients from the Mayo Clinic, but the results have not yet been reported. Antibiotic therapy has the potential to alter the gut microbiome and lower bacterial diversity, and indeed studies have confirmed that antibiotic therapy can blunt immunotherapy response. Antibiotic use may also be associated with poorer response to neoadjuvant chemoimmunotherapy in breast cancer.

Biomarkers of Immunotherapy Response

The most well-evidenced biomarkers predicting response to immunotherapy are specific to PD-1/PD-L1 inhibitors. Expression of the inhibitory PD-L1 protein can be assessed by immunohistochemistry through a number of different assays, including the Dako 22C3 pharmDx assay (the FDA companion diagnostic for pembrolizumab in metastatic TNBC) and the VENTANA SP142 assay (the companion diagnostic for atezolizumab). Unfortunately, assessment of PD-L1 status through different assays is not completely interchangeable, with the SP142 assay being more sensitive for IC PD-L1 positivity and the 22C3 assay being more sensitive for TC positivity. Furthermore, different thresholds of PD-L1 positivity are used in different studies, and different subsets of cells are assessed for PD-L1 expression—either ICs, TCs, or the combination of both ICs and TCs (referred to as the combined positive score, or CPS). In an exploratory post hoc analysis of the IMpassion130 trial, 614 patient samples were evaluated for PD-L1 status using the VENTANA SP142, SP263, and Dako 22C3 assays. PD-L1 positive patients were defined as IC ≥ 1% for the VENTANA assays and CPS ≥ 1 for the Dako assay. Most of the SP142 positive patients were also positive by the SP263 and 22C3 assays, although the latter two identified approximately 30% of additional patients as PD-L1 positive. It must be noted that the CPS cutoff for the Dako assay was more liberal than the current FDA approved cutoff of 10 for the use of pembrolizumab in metastatic TNBC. Furthermore, the site chosen for testing may influence PD-L1 positivity, with lower rates of positivity seen in bone, liver, and skin compared to other sites when assessed with the SP142 assay—bone decalcification may impact assay sensitivity, and liver metastases often have less robust immune infiltrates. The optimal way to assess PD-L1 positivity to identify responders remains to be seen, and identifying the frequency of discordance between assays and the outcomes of discordant patients will allow for refined testing strategies. Although the predictive value of PD-L1 status is clear in metastatic disease, as yet no consistent biomarker for response to neoadjuvant treatment has been identified.

TMB, a quantification of nonsynonymous mutations within a tumor genome, may predict response to checkpoint blockade immunotherapy, as a higher TMB may reflect a higher immunogenicity through neoantigen production. TMB of ≥10 mutations/Mb now carries a tumor agnostic approval for pembrolizumab based on the results of the KEYNOTE-158 trial. Furthermore, in an exploratory analysis of patients from KEYNOTE-119, a high TMB (≥10 mutations/Mb) was seen in approximately 10% of patients and was associated with prolonged progression-free survival and overall survival (OS) in pembrolizumab treated patients. Although KEYNOTE-158 also identified microsatellite instability (MSI) as another biomarker of immunotherapy response leading to a tumor agnostic approval in MSI-high tumors, less than 2% of breast cancers are MSI-high, limiting the available data and utility of MSI status as a biomarker in breast cancer. The presence of TILs within a tumor may suggest an extant immune recognition of cancer that can be upregulated with immunotherapy. TIL levels were independently associated with immunotherapy response in both the KEYNOTE-086 and KEYNOTE-119 trials.

A number of novel assays are under investigation to further delineate responders from nonresponders. Gene expression profiling can more accurately characterize active immune pathways in tumors, with IFN-γ related/T-cell inflamed gene expression associated with immunotherapy benefit in multiple cancer types. Multiplex IHC or immunofluorescence may provide the potential to precisely characterize both inhibitory and proinflammatory ICs within the tumor microenvironment, as well as their spatial relationship. A meta-analysis of the predictive accuracy of biomarker modalities suggested that multiplex IHC/immunofluorescence may more accurately predict response than other biomarkers such as PD-L1 status or TMB. However, further study is needed to develop predictive biomarkers of immunotherapy response in breast cancer.

Checkpoint Blockade Monotherapy

Single-agent PD-1/PD-L1 inhibitors were first evaluated in patients with metastatic breast cancer (MBC) in a series of early phase trials, demonstrating enriched response in triple negative and PD-L1 positive disease and in patients with fewer previous lines of therapy ( Table 63.1 ). The phase Ib multicohort KEYNOTE-012 trial evaluated single-agent pembrolizumab in a select group of 32 patients with PD-L1 positive metastatic TNBC (mTNBC), with an ORR of 18.5%. The phase Ib multicohort JAVELIN Solid Tumor trial evaluated avelumab monotherapy in 168 patients with heavily pretreated MBC, with an ORR of 3%. Responses were more frequent in TNBC patients (5.2%) compared to HR+/HER2− (2.8%) or HER2+ (0%) patients, and in those with PD-L1 IC ≥ 1% (16.7%) versus PD-L1 negative disease (1.6%). Similarly, a phase I trial of atezolizumab in 116 patients with mTNBC demonstrated an encouraging response rate in patients receiving atezolizumab as first-line (ORR 24%) versus later-line (ORR 6%) treatment, and in those with PD-L1 IC ≥ 1% (ORR 12%) versus PD-L1 negative patients (ORR 0%). Although most trials of immunotherapy monotherapy have focused on TNBC, encouraging results were seen in the HR+/HER2− cohort of the KEYNOTE-028 trial, which enrolled 25 heavily pretreated patients with PD-L1 CPS ≥ 1, demonstrating an ORR of 12% with median duration of response of 1 year, with an additional 8% of patients having stable disease for at least 24 weeks.

Further nonrandomized data in support of immunotherapy come from the phase II KEYNOTE-086 trial. Cohort A enrolled 170 patients with previously treated mTNBC who had received a taxane and anthracycline in any setting, demonstrating an ORR of 5.3%, and 5.7% in the 105 patients with PD-L1 CPS ≥ 1. With a median follow-up of 9.6 months, the median duration of response was not reached in the overall or PD-L1 positive subgroups. Cohort B of KEYNOTE-086 enrolled 84 women with previously untreated mTNBC and a PD-L1 CPS ≥ 1. The ORR was 21.4%, much higher than cohort A, and similar to the difference between treated and untreated patients as described in the above phase I trial of atezolizumab. Adverse events in these early trials were similar to studies of PD-1/PD-L1 blockade in other tumor types and generally low-grade, with most common side effects including fatigue, thyroid abnormalities, adrenal insufficiency, rash, and infrequently pneumonitis or colitis.

These early studies laid the groundwork for the phase III KEYNOTE-119 trial, which randomized 622 patients with previously treated advanced TNBC 1:1 to either pembrolizumab or the investigator’s choice of capecitabine, gemcitabine, eribulin, or vinorelbine. Included patients had received prior anthracycline and taxane treatment in any setting and had no more than two lines of prior therapy for advanced disease. The primary end point of OS was not significantly higher with pembrolizumab versus chemotherapy in the overall (median 9.9 vs. 10.8 months, hazard ratio 0.97, P not reported due to statistical design), CPS ≥ 1 (median 10.7 vs. 10.2 months, hazard ratio 0.86, P = 0.073), or CPS ≥ 10 populations (median 12.7 vs. 11.6 months, hazard ratio 0.78, P = 0.057). However, in a post hoc analysis in the 18% of patients with CPS ≥ 20, OS was prolonged with pembrolizumab (median 14.9 vs. 12.5 months, hazard ratio 0.58, P not reported due to exploratory nature). ORR increased with increasing PD-L1 CPS, and was 12.3%, 17.7%, and 26.3% in the CPS ≥ 1, CPS ≥ 10, and CPS ≥ 20 populations, respectively. Rates of grade 3 to 4 adverse events were 14% in the pembrolizumab group and 36% in the chemotherapy group. Although single-agent immunotherapy is only FDA-approved for patients with high TMB or MSI-high status, there may be a role for monotherapy in the treatment of select patients with advanced TNBC.

Checkpoint Blockade Combinations