The immunological profile of early-stage breast cancer treated with neoadjuvant PARP inhibitors has not been described. The aim of this study was to delineate the changes in the tumor immune microenvironment (TiME) induced by talazoparib.
Patients with operable germline BRCA1/2 pathogenic variant (gBRCA1/2+) breast cancer were enrolled in a feasibility study of neoadjuvant talazoparib. Thirteen patients who received 8 weeks of neoadjuvant talazoparib were available for analysis, including 11 paired pre- and post-talazoparib core biopsies. Treatment-related changes in tumor-infiltrating lymphocytes were examined and immune cell phenotypes and their spatial distribution in the TiME were identified and quantified by multiplex immunofluorescence using a panel of 6 biomarkers (CD3, CD8, CD68, PD-1, PD-L1, and CK).
Neoadjuvant talazoparib significantly increased infiltrating intratumoral and stromal T-cell and cytotoxic T-cell density. There was no difference in PD-1 or PD-L1 immune cell phenotypes in the pre- and post-talazoparib specimens and PD-L1 expression in tumor cells was rare in this cohort. Spatial analysis demonstrated that pre-talazoparib interactions between macrophages and T cells may correlate with pathologic complete response.
This is the first study with phenotyping to characterize the immune response to neoadjuvant talazoparib in patients with gBRCA1/2+ breast cancer. These findings support an emerging role for PARP inhibitors in enhancing tumor immunogenicity. Further investigation of combinatorial strategies is warranted with agents that exploit the immunomodulatory effects of PARP inhibitors on the TiME.
Growing preclinical evidence clearly suggests that PARP inhibitors modulate the TiME. This study further evaluates these findings in patients with gBRCA1/2-mt breast cancer treated with neoadjuvant talazoparib. Using multiplex immunofluorescence to quantitatively monitor and spatially resolve immune cell phenotypes, we demonstrated that PARP inhibition increased T-cell trafficking into the TiME. This study sets forth a rationale to further characterize PARP inhibitors novel immunomodulatory function in future translational studies and for the development of combination strategies with immunotherapy.
Mutations in BRCA1 and BRCA2 cause defects in homologous recombination repair and lead to replication stress and genomic instability. PARP enzymes detect and repair DNA damage through the base excision repair pathway and maintain genetic stability and inhibition of PARP in BRCA1/2-deficient cancer cells results in synthetic lethality (1). Preclinical evidence suggests that PARP inhibitors may enhance the antitumor immune response by increasing tumor-infiltrating lymphocytes (TIL) and more specifically cytotoxic T cells through activation of the cGAS–STING innate immune pathway (2, 3).
The importance of infiltrating immune cells and their spatial location in relation to breast cancer cells in response to PARP inhibitors remains unknown. Multiplex immunohistochemical methods allow multiparametric analysis of the dynamic immune composition and resolve spatial interactions within the tumor immune microenvironment (TiME). Comprehensive characterization of the TiME to inform strategies for harnessing the immune system for clinical benefit remains a high priority.
Talazoparib is a PARP1/2 inhibitor approved for use in patients with advanced breast cancer with a germline BRCA1/2 pathogenic variant (gBRCA1/2+). A phase 3 trial showed a significant improvement in median progression-free survival (PFS) for patients receiving talazoparib compared with physician's choice of chemotherapy (4). We previously reported a feasibility study of 8 weeks of talazoparib administered in the neoadjuvant setting (NCT02282345) to 13 patients with early-stage gBRCA1/2+ breast cancer and demonstrated a median decrease in tumor volume by 88% (range, 30%–98%) by ultrasound. These patients then received neoadjuvant chemotherapy before surgery (5). Here, we report the impact of talazoparib on the TiME and describe the spatial distribution of immune and tumor cells in pre- and post-treatment specimens from that feasibility trial.
Patients and Methods
Study design and patients
NCT02282345 was a feasibility trial of neoadjuvant talazoparib monotherapy in patients with operable breast cancer and a pathogenic gBRCA1/2+ (5). Patients received talazoparib (1 mg per day) orally for 8 weeks before starting a neoadjuvant anthracycline- and taxane-based chemotherapy regimen of the physician's choice. Core biopsies obtained at baseline (pre-talazoparib) and after 8 weeks of talazoparib (post-talazoparib) were analyzed for changes in the TiME. All patients included in the present analysis consented to evaluation of their archival tumor samples. This study was conducted under Institutional Review Board approved Protocol 2014–0045 and in accordance with relevant guidelines at The University of Texas MD Anderson Cancer Center (MDACC, Houston, TX). The Institutional Review Board at MDACC provided approval for the use of patient samples in this study (study number 2014–0045). Samples were acquired with written informed consent from all participants included in the study and this study was performed in accordance with the ethical standards outlined in the 1964 Helsinki Declaration and its later amendments.
Formalin-fixed, paraffin-embedded (FFPE) core biopsy specimens were obtained from patients at baseline and after 8 weeks of talazoparib.
TILs were quantified manually by hematoxylin–eosin (H&E)-stained FFPE sections and scored as a percentage of tumor area (tumor cells and stroma) by a breast pathologist (E.R.P. Cuentas) according to the International Immuno-Oncology Working Group method for assessing TILs (6).
Immune profiling by multiplex immunofluorescence staining
Immune profiling was performed on pre- and post-talazoparib biopsies with multiplex immunofluorescence (mIF) to simultaneously evaluate 6 biomarkers (pancytokeratin AE1/AE3, CD8, CD3, CD68, PD-1, and PD-L1). Using Vectra multispectral imaging system v3.0 (Akoya/PerkinElmer), FFPE sections stained with the mIF panel and Opal 7-color Kit were scanned as previously described (7, 8) and detailed in Supplementary Methods. Cell phenotypes were quantified as cell density (cells/mm2) and co-localization of the biomarkers was performed on invasive tumor and associated stromal regions by two pathologists (F. Yang and E.R.P. Cuentas) using Inform 2.3 image analysis software (Akoya/PerkinElmer). The following are the identification biomarkers: Total T cells, CD3+; CD8− T cells, CD3+CD8−; cytotoxic T cells, CD3+CD8+; epithelial breast cancer cells, AE1/AE3+ (CK+); and macrophages, CD68+. Co-expression of PD-1 and PD-L1 on T-cell subsets, and PD-L1 on tumor cells and macrophages were also examined.
Spatial interaction analysis
Scanned images were analyzed by Inform 2.3 (Akoya/PerkinElmer) to determine the spatial location of each cell phenotype. Pairwise combinations from spatial interactions based on the nearest neighbor distribution were computed with spatial G-function to quantify spatial interactions of cells of interest as previously described and detailed in Supplementary Methods (9). The probability of pairwise cell phenotype interactions was compared before and after 8 weeks of talazoparib and in correlation to pathologic complete response (pCR) after completion of neoadjuvant chemotherapy.
Gene expression analysis
RNA-sequencing data were generated by the Core Genomics Laboratory at MD Anderson. cDNA was generated from 150 ng of RNA using the NuGEN Ovation system and amplified using both 3′ poly(A) selection and random priming. Amplified products were shared using a Covaris E220 ultrasonicator and fragment sizes confirmed with an Agilent Bioanalyzer. Libraries were prepped using the NuGEN Ovation Ultralow library prep protocol, and sequenced on an Illumina HiSeq 2000 system.
We verified the quality of the sequencing data using FASTQC (http://www.bioinformatics.babraham.ac.uk/projects/fastqc/). Then, we mapped reads to the hg19 human genome assembly using a GTF gene model from ENSEMBL (PMID 33137190) using the STAR aligner (PMID 23104886). Gene-level counts were quantified with HTSeq-count (PMID 25260700), and transcripts per million (TPM) were calculated using RSEM (PMID 20022975). To deconvolute the immune cell phenotypes from the bulk transcriptional profiles, we applied CIBERSORT in absolute mode (10).
The Wilcoxon signed-rank test was used to evaluate associations between pre- and post-TALA biomarker changes. Unpaired patient specimens were excluded from pre- versus post-talazoparib analyses. Correlations were evaluated using Spearman rank correlation. Statistical comparison of gene expression analysis was done using a two-tailed Student t test. A nominal P value of <0.05 was considered statistically significant. All data outputs are provided as median values. Statistical calculations were performed by R v3.6.1 software (http://www.r-project.org) and GraphPad Prism v8.
The datasets used and/or analyzed during the current study are available from the corresponding author on reasonable request.
Of the 13 patients, 24 tissue biopsies were available for analysis and included 11 paired, one pre-, and one post-talazoparib patient specimens (Fig. 1A). Clinicopathologic characteristics of paired specimens are summarized in Supplementary Table S1 (5).
We first used H&E to examine the impact of talazoparib on TIL expressions. Six paired samples did not have residual tumor cells after treatment with talazoparib were excluded from the analysis. For those with residual tumor, there was a trend toward an increase in TILs after treatment. Pre-talazoparib, the median TIL level was 3.0% (range, 1%–40%); post-talazoparib, the median TIL level was 17.5% (range, 1%–90%; P = 0.058; Fig. 1B and C).
Major constituents of the TiME were determined using mIF (Fig. 2A). The complexity of the immune cell subpopulations in the intratumoral and intrastromal compartment for each patient is shown as a percentage of the total number of immune cells (CD3+ and CD63+ cells; Fig. 2B). Some samples were dominated by CD8− T cells (patient 10, pre-talazoparib), whereas others were dominated by macrophages (patient 2, pre-talazoparib). PD-1+ and PD-L1+ immune cell subsets represented a minority of the total immune cells across specimens. Total T cells, CD8− T cells, and cytotoxic T cells significantly increased after treatment in the tumor and the surrounding stroma; there was no change in macrophages (Fig. 2C). The median percentage of PD-1+ T cells and PD-1+ cytotoxic T cells out of the total number of T cells and cytotoxic T cells, respectively, was approximately 1% in pre- and post-talazoparib specimens and the percentage of PD-L1+ T cells and macrophages was similarly low (Supplementary Fig. S1). The percentage of PD-L1+ tumor cells was absent/low and detected in 1/12 pre-talazoparib specimens (0.02% PD-L1+ tumor cells) and in 2/7 post-talazoparib specimens (3.14% and 6.35% PD-L1+ tumor cells for each patient). Four post-talazoparib specimens had treatment-related effects, but no tumor cells were detected for analysis. There was no difference in the pre- and post-talazoparib immune cell phenotypes by clinical stage, subtype, or BRCA mutational status (Supplementary Table S1). Pre- and post-talazoparib macrophages correlated with a decrease in tumor volume by ultrasound after treatment (Supplementary Fig. S2). Pre-talazoparib total T-cell densities predicted pCR on surgical specimens after completion of 8 weeks of talazoparib and neoadjuvant chemotherapy (Supplementary Fig. S3). Other immune cell phenotype densities pre- or post-talazoparib did not predict response by ultrasound or pCR (Fig. 3; Supplementary Fig. S3).
Spatial analysis revealed significant heterogeneity in the probability of cell phenotype interactions. Some pre-talazoparib specimens with a low probability of tumor cells interacting with CD8− T cells, converted to a high probability after treatment with talazoparib with a representative paired specimen (Fig. 3A). The probability of spatial interactions of cell phenotypes on pre-talazoparib specimens was correlated with pCR. Of these pairwise interactions, macrophages were in closer proximity to CD8− T cells in patients who achieved pCR versus not (probability of interaction: 0.22 vs. 0.09, P = 0.017; Fig. 3B). The probability of spatial interaction of other cell phenotypes pre-talazoparib and pCR was not significant (Supplementary Fig. S4). A spatial matrix analysis of cell phenotype interactions in pre-talazoparib biopsies of individual patients is depicted by heatmap clustered by pCR response (Fig. 3C).
To further characterize the TiME changes by mIF, we analyzed bulk gene expression data. In agreement with the findings of mIF, talazoparib significantly increased in gene expression levels of T-cell markers, CD3 and CD8 subunits, and CD4 (Supplementary Fig. S5); there was no significant difference in granzyme B or FOXP3 (Supplementary Fig. S5). Post-talazoparib, there was modest enrichment in immune checkpoints ICOS, CTLA4, LAG3, PDCD1 (PD-1), and CD274 (PD-L1; Fig. 4A). To assess the differential immune cell infiltration, we applied CIBERSORT (Fig. 4B). The two main immune subpopulations were memory resting CD4+ T cells and M2 macrophages (Fig. 4C). Post-talazoparib specimens were enriched in CD4+ T memory resting, M2 macrophages, CD8+ T-cell, memory B-cell, T-cell gamma delta, and activated mast cell subsets clustered in post-talazoparib specimens. Expressions of T regulatory and other immune cell subsets were less frequent.
This study is the first description of the impact of talazoparib monotherapy, administered in the neoadjuvant setting to patients with gBRCA1/2+ breast cancer, on the TiME. Our findings show that 8 weeks of talazoparib results in increased T-cell and cytotoxic T-cell density in the tumor and the adjacent stroma.
Application of CIBERSORT to enumerate the immune cell composition within the TiME suggests that the increase in T cells by mIF is predominantly by CD4+ memory resting cells, CD8+ T cells, and gamma delta T cells. In addition, CIBERSORT revealed that talazoparib may also lead to infiltration of plasma and memory B cells, and mast cells. Post-talazoparib there was an increase in targetable immune checkpoints, ICOS, CTLA4, LAG3, PD-1, and PD-L1, possibly suggesting future combinatorial strategies after additional validation.
Correlating TiME changes and outcomes in this study cohort is limited; all patients had a decrease in tumor volume by ultrasound of at least 30% and went on to receive interval chemotherapy before surgery and determination of pCR. Nonetheless, baseline total T cells by mIF predicted pCR, consistent with prior reports (11). The effect of neoadjuvant chemotherapy on the TiME and early and late outcomes is inconsistent and limited to a few smaller patient studies. In SWOG S0800, a decrease in TILs correlated with pCR; however, in another smaller study an increase in post-treatment TILs was associated with longer 5-year recurrence-free survival. This is potentially an important observation as in early-stage gBRCA1/2+ breast cancer, baseline TILs are associated with a greater response to neoadjuvant chemotherapy and long-term survival (12) and clinical response, defined by clinical and radiographic evaluation, to short-course neoadjuvant olaparib, in unselected early-stage TNBC (13).
Interestingly, increased densities of macrophages in both pre- and post-talazoparib–treated specimens correlated with radiographic response, but not the absolute change. Application of CIBERSORT to enumerate the immune cell composition within the TiME suggests that the predominant macrophage detected by mIF is of the immunosuppressive M2 phenotype.
In our cohort of patients, PD-1 and PD-L1 expression was low in immune cell populations and tumor cells, respectively, and unchanged after treatment with talazoparib. This is in contrast with other studies of early-stage breast cancer with staining of PD-1+ mononuclear cells (40%–50%) and PD-L1+ tumor cells (20%–50%), with variations by molecular subtype, grade, and stage (14, 15). Low expression of PD-1 and PD-L1 in our analysis may be due to variation in staining intensity by different antibody clones and/or the heterogeneous staining distribution of PD-L1 in core biopsies as previously described (16). In addition, in preclinical models of gBRCA1/2+ breast cancer, there is conflicting evidence that PARP inhibitors induce PD-L1 expression on tumor cells, which may be due to transient PD-L1 induction (2, 17). Interestingly, a recent study found that olaparib increased PD-L1+ macrophages in murine BRCA-deficient xenografts, possibly counteracting PARP inhibitor T-cell–mediated immunity (18). In our study, although not statistically significant, PD-L1+ macrophage cell density either remained low (<1%) or decreased after treatment, with the exception of one outlier. The mIF biomarker panel used for our analysis did not include additional macrophage markers such as CD163 or CSFR1 to differentiate macrophages from CD63+ tumor-associated fibroblasts of monocyte-derived fibrocytes that have been described in the breast cancer TiME (19). In future studies, an expanded mIF biomarker panel should be considered in a larger cohort.
It has previously been reported that in early-stage breast cancer, cytotoxic T cells within cancer islands compared with the overall tumor tissue or stroma are associated with relapse-free survival by mIF (20). By spatial analysis, we found heterogeneity in samples exhibiting immune infiltrated and non-infiltrated phenotype in relation to tumor cells. In addition, spatial analysis suggested that pretreatment interaction of macrophages and T cells may predict pCR. Additional investigation is required to confirm this observation and better characterize macrophage phenotypes, with additional biomarkers as above (19).
Preclinical evidence has demonstrated overwhelming synergy with the combination of PARP inhibitors and PD-1/PD-L1 immune checkpoint blockade (2, 3). However, in the initial phase 1/2 single-arm trials, MEDIOLA and TAPACIO/Keynote-162, this combination demonstrated response rates similar to historical trials with PARP inhibitor monotherapy in BRCA1/2-mt metastatic breast cancer. Further studies and follow-up are ongoing to determine whether this combination leads to a longer duration of response (4, 21–23). Thus, there is an urgent need for in depth prospective investigations of the TiME. Our analysis highlights the potential of mIF to characterize the complexities of PARP inhibitor–mediated changes on immune cell phenotypes within the TiME and to inform combinations of immune-modulating agents in future studies.
In summary, our descriptive analysis of the changes in the TiME before and after PARP inhibitor treatment provides a provocative preliminary signal that talazoparib enhances tumor immunity. The small sample size precluded more rigorous statistical analysis and correlation with histologic grade, subtype, BRCA mutation status, and response. Investigation continues into the relative importance of these findings in a phase II multicenter trial (NCT03499353) underway with single-agent talazoparib for 6 months before surgery.
J.T. Chang reports grants from Cancer Prevention and Research Institute of Texas during the conduct of the study. A. Contreras reports other support from Cell IDx outside the submitted work. G.J. Whitman reports other support from Pfizer, UpToDate, and Siemens outside the submitted work. E.A. Mittendorf reports personal fees from AstraZeneca, Merck, Exact Sciences, and Roche/Genentech; non-financial support from BMS and Lilly; other support from Gilead; and grants from Genentech outside the submitted work. J.K. Litton reports grants and non-financial support from Medivation/Pfizer, as well as grants from AstraZeneca during the conduct of the study. J.K. Litton also reports grants from Genentech, Merck, and GSK outside the submitted work. In addition, J.K. Litton reports Speaker's Bureau for MedLearning, Physicians' Education Resource, Prime Oncology, Medscape, Clinical Care Options, and Medpage, as well as royalties from UpToDate. No disclosures were reported by the other authors.
T. Kumar: Conceptualization, data curation, formal analysis, visualization, methodology, writing–original draft. E. Hobbs: Conceptualization, data curation, formal analysis, visualization, writing–original draft. F. Yang: Conceptualization, resources, data curation, formal analysis, investigation, visualization, methodology, writing–original draft. J.T. Chang: Resources, data curation, formal analysis, validation, methodology, writing–review and editing. A. Contreras: Resources, data curation, investigation. E.R.P. Cuentas: Resources, data curation, investigation. H. Garber: Conceptualization, methodology, writing–review and editing. S. Lee: Data curation, writing–review and editing. Y. Lu: Data curation, supervision, writing–review and editing. M.E. Scoggins: Data curation, investigation, writing–review and editing. B.E. Adrada: Data curation, investigation, writing–review and editing. G.J. Whitman: Data curation, investigation, writing–review and editing. B.K. Arun: Investigation, methodology, writing–review and editing. E.A. Mittendorf: Writing–review and editing. J.K. Litton: Conceptualization, supervision, investigation, methodology.
The study was supported by The Toomim Family Fund. J.T. Chang was supported by the Cancer Prevention and Research Institute of Texas (RP160710, RP170668). T. Kumar is funded by the NCI T32 Translational Genomics Fellowship and Rosalie B. Hite fellowship.
The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
Note: Supplementary data for this article are available at Clinical Cancer Research Online (http://clincancerres.aacrjournals.org/).