Changes in Peripheral and Local Tumor Immunity After Neoadjuvant Chemotherapy Reshape Clinical Outcomes in Patients with Breast Cancer Margaret L

Published OnlineFirst August 21, 2020; DOI: 10.1158/1078-0432.CCR-19-3685

CLINICAL CANCER RESEARCH | TRANSLATIONAL CANCER MECHANISMS AND THERAPY

Changes in Peripheral and Local Tumor Immunity after Neoadjuvant Chemotherapy Reshape Clinical Outcomes in Patients with Breast Cancer Margaret L. Axelrod1, Mellissa J. Nixon1, Paula I. Gonzalez-Ericsson2, Riley E. Bergman1, Mark A. Pilkinton3, Wyatt J. McDonnell3, Violeta Sanchez1,2, Susan R. Opalenik1, Sherene Loi4, Jing Zhou5, Sean Mackay5, Brent N. Rexer1, Vandana G. Abramson1, Valerie M. Jansen1, Simon Mallal3, Joshua Donaldson1, Sara M. Tolaney6, Ian E. Krop6, Ana C. Garrido-Castro6, Jonathan D. Marotti7,8, Kevin Shee9, Todd. W. Miller8,9, Melinda E. Sanders2,10, Ingrid A. Mayer1,2, Roberto Salgado4,11, and Justin M. Balko1,2

ABSTRACT ◥ Purpose: The recent approval of anti-programmed death-ligand Results: In non-TNBC, no change in expression of any single 1 immunotherapy in combination with nab-paclitaxel for meta- gene was associated with RFS or OS, while in TNBC upregulation of static triple-negative breast cancer (TNBC) highlights the need to multiple immune-related genes and gene sets were associated with understand the role of chemotherapy in modulating the tumor improved long-term outcome. High cytotoxic T-cell signatures immune microenvironment (TIME). present in the peripheral blood of patients with breast cancer at Experimental Design: We examined immune-related gene surgery were associated with persistent disease and recurrence, expression patterns before and after neoadjuvant chemotherapy suggesting active antitumor immunity that may indicate ongoing (NAC) in a series of 83 breast tumors, including 44 TNBCs, from disease burden. patients with residual disease (RD). Changes in gene expression Conclusions: We have characterized the effects of NAC on the patterns in the TIME were tested for association with recurrence- TIME, ﬁnding that TNBC is uniquely sensitive to the immunologic free (RFS) and overall survival (OS). In addition, we sought to effects of NAC, and local increases in immune genes/sets are characterize the systemic effects of NAC through single-cell associated with improved outcomes. However, expression of cyto- analysis (RNAseq and cytokine secretion) of programmed toxic genes in the peripheral blood, as opposed to the TIME, may be þ death-1–high (PD-1HI)CD8 peripheral T cells and examination a minimally invasive biomarker of persistent micrometastatic dis- of a cytolytic gene signature in whole blood. ease ultimately leading to recurrence.

Introduction was recently approved for the treatment of patients with metastatic triple-negative breast cancer (TNBC) based on results from a phase III Combination of conventional chemotherapy with an immunother- clinical trial (1). Furthermore, addition of the anti–programmed apeutic targeting programmed death-ligand 1 (PD-L1), atezolizumab, death-1 (PD-1) mAb pembrolizumab to neoadjuvant chemotherapy (NAC) can significantly improve TNBC pathologic complete response (pCR) rates (2). Thus, existing clinical data indicate that chemotherapy 1 Department of Medicine, Vanderbilt University Medical Center, Nashville, Ten- combinations with immunotherapy demonstrate enhanced efficacy 2 nessee. Breast Cancer Research Program, Vanderbilt University Medical Center, compared with chemotherapy alone. However, these results suggest a Nashville, Tennessee. 3Department of Infectious Disease, Vanderbilt University Medical Center, Nashville, Tennessee. 4Department of Oncology, University of growing need to better understand how chemotherapy modulates the Melbourne and Peter MacCallum Cancer Centre, Melbourne, Victoria, Australia. tumor immune microenvironment (TIME). 5IsoPlexis Corporation, Branford, Connecticut. 6Department of Medical Oncol- High levels of stromal tumor-infiltrating lymphocytes (sTIL) in the ogy, Dana-Farber Cancer Institute and Harvard Medical School, Boston, Mas- pretreatment biopsy are predictive of pCR in patients with TNBC 7 sachusetts. Department of Pathology & Laboratory Medicine, Dartmouth- treated with NAC (3). In NAC-treated patients with TNBC with 8 Hitchcock Medical Center, Lebanon, New Hampshire. Norris Cotton Cancer residual disease (RD) at surgery or in untreated primary TNBC Center, Geisel School of Medicine at Dartmouth, Hanover, New Hampshire. 9Department of Molecular & Systems Biology, Geisel School of Medicine at tumors, higher sTILs in the resected tumor also confer improved – Dartmouth, Hanover, New Hampshire. 10Department of Pathology, Microbiology prognosis (4 7). However, the immunomodulatory effect of chemo- and Immunology, Vanderbilt University Medical Center, Nashville, Tennessee. therapy on sTILs in patients and how chemotherapy influences the 11Department of Pathology, GZA-ZNA Hospitals, Antwerp, Belgium. TIME are poorly understood. In a study of patients with non–small cell Note: Supplementary data for this article are available at Clinical Cancer lung carcinoma, tumors treated with NAC had higher expression of þ Research Online (http://clincancerres.aacrjournals.org/). PD-L1 and higher density of CD3 T cells, suggesting NAC may be M.L. Axelrod and M.J. Nixon contributed equally to this article. immunomodulatory in some tumor types (8). In addition, in a small breast cancer study, post-NAC natural killer (NK) cells and IL6 Corresponding Author: Justin M. Balko, Vanderbilt University, 2200 Pierce expression were associated with better response. However, this study Avenue, 777 PRB, Nashville, TN 37232. Phone: 615-875-8666; Fax: 615-936- ¼ 1495; E-mail: [email protected] was limited by inclusion of only a small number of TNBCs (n 4; ref. 9). Intriguingly, a larger study of patients with breast cancer Clin Cancer Res 2020;XX:XX–XX (n ¼ 60) showed that pre-NAC sTILs and higher pre-NAC expression doi: 10.1158/1078-0432.CCR-19-3685 of cytotoxic T-cell markers and cytokines were associated with higher 2020 American Association for Cancer Research. pCR rate. However, this study did not examine long-term outcomes in

AACRJournals.org | OF1

Axelrod et al.

þ but 2 estrogen receptor–positive (ER ) patients received cytotoxic Translational Relevance þ chemotherapy as part of their regimen. Four patients (ER ) Indications for immunotherapy, alone or in combination, are received courses of hormone therapy as part of their neoadjuvant expanding, including in breast cancer. However, the immunologic regimen, two of which were in conjunction with cytotoxic chemo- þ landscape of breast cancer and how chemotherapy, the current therapy. Five patients (HER2 ) received HER2-directed therapy as standard of care for triple-negative breast cancer (TNBC), influ- part of their neoadjuvant regimen. ences local and systemic immune responses are incompletely For the “Peru” cohort, clinical characteristics and molecular analysis characterized. Herein, we show that increases in expression of of the patients (n ¼ 48 with matched pretreatment tissue) were immune-related genes and gene sets in the tumor over the course of previously described at the Instituto Nacional de Enfermedades chemotherapy are associated with improved prognosis in TNBC, Neoplasicas (12). Clinical and pathologic data were retrieved from but not other subtypes of breast cancer. Conversely, a gene medical records under an institutionally approved protocol (INEN expression signature of immune activation and cytotoxicity in the IRB 10–018). For the “VICC” cohort, which included PBMC and peripheral blood was associated with persistent disease following whole blood analyses, clinical and pathologic data were retrieved from chemotherapy and disease recurrence following surgery. Examin- medical records under an institutionally approved protocol (VICC IRB ing immune-related signatures locally and systemically may serve 030747). For the DARTMOUTH cohort, patient samples were col- as biomarkers of patients likely to benefit from additional immu- lected under a protocol approved by the Dartmouth College Institu- notherapeutic approaches. tional Review Board and the waiver of the subject consent process was IRB approved (IRB 28888). Metadata for the primary cohort of patients is presented in Supplementary Table S1. For the peripheral blood study, two cohorts of patients were used (summarized patients with RD, for whom prognosis is worse than those with pCR in Table 2). For the Vanderbilt cohort, all blood was collected within and only included a limited number of TNBC samples (n ¼ 13; ref. 10). 14 days preceding definitive surgery. Metadata for the Vanderbilt Furthermore, whether immunomodulatory effects of NAC are locor- cohort is presented in Supplementary Table S7. For the DFCI cohort egional or systemic (i.e., able to be detected in the peripheral blood) is (11), all blood was collected in the interlude between completion of unknown. NAC and definitive surgery. To address this gap in knowledge, we examined expression patterns of immune-related genes before and after NAC in a series of 83 breast TILs Quantification tumors, including 44 TNBCs, from patients with RD. As patients with sTILs were analyzed using full face hematoxylin and eosin sections pCR generally experience excellent outcomes, we chose to focus on from pre-NAC diagnostic biopsies or post-NAC RD surgical speci- patients with RD, who may benefit from additional therapies. Changes mens. Samples were scored according to the International TILs in gene expression patterns in the TIME were tested for association with Working Group Guidelines (13–15). The predefined cut point of recurrence-free (RFS) and overall survival (OS). T-cell receptor sequen- 30% (4) was used for all survival analyses. cing was performed on a subset (n ¼ 15) of tumors. In addition, in 4 patients undergoing NAC, PD-1–high (PD-1HI; top 20% expressors of NanoString nCounter analysis þ þ CD8 or CD4 cells) and PD-1–negative (PD-1NEG)CD3 CD8 periph- Gene expression and gene set analysis on pre- and post-NAC eral blood mononuclear cells (PBMC) were profiled using single-cell formalin-fixed tissues were performed using the nanoString Pan- RNA sequencing (scRNAseq) and multiplexed cytokine secretion Cancer Immunology panel (770 genes) according to the manufac- assays. Finally, we used an scRNAseq-derived signature of activated turer’s standard protocol. Data were normalized according to positive cytolytic cells to measure immune activation in the peripheral blood of and negative spike-in controls, then endogenous housekeeper con- two cohorts of patients: a Vanderbilt cohort consisting of 34 patients trols, and transcript counts were log transformed for downstream after NAC (collected within 2 weeks prior to surgery) and 24 untreated analyses. Normalized linear data are presented in Supplementary patients, and a cohort from the Dana-Farber Cancer Institute (DFCI; Table S2. Gene sets were calculated by summing the log2-transformed Boston, MA) consisting of 30 hormone receptor–positive, HER2 normalized NanoString counts for all genes contained in a given gene patients treated with NAC (including bevacizumab) as part of a clinical set (Supplementary Table S3). Samples were simultaneously assayed trial (11). We then tested the association of this signature with surgical for PAM50 molecular subtyping using a custom-designed 60-gene outcome (pCR or RD burden) and postsurgical cancer recurrence. Elements panel (Supplementary Table S4). Briefly, 10 mm sections of Intriguingly, higher expression of the gene signature at the time of diagnostic biopsies or residual tumors were used for RNA preparation surgery was associated with higher disease burden [i.e., in those with [Promega Maxwell 16 RNA formalin-fixed, paraffin-embedded RD who experienced disease recurrence within 3 years following (FFPE)] and 50 ng of total RNA >300 nt (assayed on an Agilent surgery or those with the highest residual cancer burden (RCB)]. Tapestation 2200 Bioanalyzer) was used for input into nCounter Thus, peripheral cytotoxic activity, guarded by immune checkpoints, hybridizations for Pan-Cancer Immunology panels or 500–1,000 ng may reflect ongoing micrometastatic and primary disease burden, and RNA for PAM50 analysis. Data were normalized according to positive could be a useful biomarker for disease recurrence and possibly and negative spike-in controls, then endogenous housekeeper con- immune checkpoint inhibitor benefit. trols, and transcript counts were log transformed for downstream analyses. Subtype prediction was performed in R using the genefu package. Materials and Methods For the 8-gene signature analysis in whole blood, a custom Nano- Patients String Elements was constructed to measure the gene expression levels Three cohorts of patients were combined for the tumor-profiling of PDCD1, NKG7, LAG3, GZMH, GZMB, GNLY, FGFBP2, study. All included patients received neoadjuvant therapy and had HLA-DRB5, and HLA-G, as well as 3 normalization control genes RD and matched pretreatment tissue was required for inclusion. All (PTPRC, RPL13a, TBP). Probe design is shown in Supplementary

OF2 Clin Cancer Res; 2020 CLINICAL CANCER RESEARCH

Immunologic Changes with Chemotherapy in TNBC

Table S9. RNA was isolated from whole blood (Promega Maxwell 16 TCR Sequencing Simply RNA Blood) and 100–200 ng was used for input into the TCR sequencing and clonality quantification was assessed in FFPE nCounter analysis. Data were normalized as above. Linear normalized samples of breast cancer specimens or PBMCs. For FFPE tissue, DNA data are presented in Supplementary Table S8 for the Vanderbilt or RNA was extracted from 10 mm sections using the Promega Cohort. Maxwell 16 FFPE DNA or FFPE RNA kits and the manufacturer’s þ protocol. For PBMCs, PD-1HI and PD-1NEG CD8 T cells were sorted Isoplexis (single-cell cytokine profiling) by FACS from samples isolated from EDTA collection tubes and On day 1, cryopreserved PBMCs were thawed and resuspended in processed using a Ficoll gradient. At least 100 K cells were collected, complete RPMI media with IL2 (10 ng/mL) at a density of 1–5 centrifuged, and utilized for RNA purification. TCRs were sequenced 6 10 cells/mL. Cells were recovered at 37 C, 5% CO2,overnight.Plates using survey level immunoSEQ (DNA; Adaptive Biotechnologies) and were prepared by coating with anti-human CD3 (10 mg/mL in PBS, the Immunoverse assay (RNA; ArcherDX), as described previous- 200–300 mL/well) in a 96-well flat-bottom plate at 4C, O/N. On day ly (22, 23). Sequencing results were evaluated using the immunoSEQ 2, nonadherent cells for each sample were collected and viability was analyzer version 3.0 or Archer Immunoverse analyzer. CDR3 confirmed, with dead cell depletion by Ficoll. For each sample, where sequences and frequency tables were extracted from the manufac- sufficient, volume was split in half for each of the following negative turers’ analysis platforms and imported into R for analysis using the isolations: with one half of cells from each sample, CD4 T cells were Immunarch package (https://immunarch.com; ref. 24) in R. Shannon þ isolated with CD4 negative isolation kit following Miltenyi protocol entropy, a measure of sample diversity, was calculated on the clonal (130-096-533); with the other half of cells from each sample, CD8 T abundance of all productive TCR sequences in the data set. Shannon þ cells were isolated with CD8 negative isolation kit following Mil- entropy was normalized by dividing Shannon entropy by the loga- þ tenyi protocol (130-096-495). The PD-1 and PD-1 subsets were rithm of the number of unique productive TCR sequences. This from isolated CD4 or CD8 T cells by staining with PE-conjugated normalized entropy value was then inverted (1 normalized entropy) anti–PD-1 antibody using the manufacturer’s protocol (Miltenyi, to produce the “clonality” metric. TCRb clonotypes and metadata 130-096-164) as follows: (i) stain each subset with 10 mLstain:100mL based on the primary cohort (Supplementary Fig. S5; Adaptive) are Robosep buffer for every 1 107 total cells; (ii) incubate at 4Cfor10 included as Supplementary Dataset 1. TCRb clonotypes and metadata minutes; (iii) rinse cells by adding 1–2 mL of Robosep and C/F at based on prospectively collected peripheral blood and tumor 300 g for 10 minutes; (iv) aspirate supernatant and resuspend cell from Fig. 4 (Archer) are included as Supplementary Dataset 2. þ pellets in 80 mLbufferper1 107 total cells. PD-1 cells were then isolated with anti-PE microbeads following the manufacturer’s scRNAseq protocol (Miltenyi, 130-097-054). Cells were resuspended in com- PBMCs were isolated from EDTA collection tubes, processed using plete RPMI media at a density of 1 106/mL and seeded into wells of a Ficoll gradient, and cryopreserved in 10% DMSO 90% FBS. Upon the CD3-coated 96-well flat-bottom plate with soluble anti-human thaw, whole live PBMCs were prepared by depleting dead cells using a – – HI CD28 (5 mg/mL). Plates were incubated at 37 C, 5%CO2 for 24 hours. dead cell removal Kit (Miltenyi, catalog no.: 130 090 101). PD-1 þ On day 3, supernatants (100 mL/well) were collected from all wells and PD-1NEG CD8 T cells were sorted by FACS from PBMCs. Each andstoredat80C for population assays. T cells were collected and sample (targeting 5,000–10,000 cells/sample) was processed for single stained with Brilliant Violet cell membrane stain and AlexaFluor- cell 50 RNAseq utilizing the 10x Chromium system. Libraries were 647–conjugated anti-CD8 at room temperature for 20 minutes, prepared using P/N 1000006, 1000080, and 1000020 following the rinsed with PBS and resuspended in complete RPMI media at a manufacturer’s protocol. The libraries were sequenced using the density of 1 106/mL. NovaSeq 6000 with 150 bp paired end reads. RTA (version 2.4.11; Approximately 30 mL of cell suspension was loaded into the IsoCode Illumina) was used for base calling and analysis was completed using Chip and incubated at 37 C, 5% CO2 for an additional 16 hours. 10 Genomics Cell Ranger software v2.1.1. Data were analyzed in R Protein secretions from approximately 1,000 single cells were captured using the filtered h5 gene matrices in the Seurat (25, 26) package. by the 32-plex antibody barcoded chip and analyzed by fluorescence Briefly, samples were merged, and all cells were scored for mitochon- ELISA-based assay (16–21). Polyfunctional T cells that cosecreted 2þ drial gene expression (a marker of dying cells) and cell-cycle genes cytokines per cell were evaluated by the IsoSpeak software across the to determine phase. Data were transformed using SCTransform, five functional groups: Effector: Granzyme B, TNFa, IFNg, MIP1a, regressing against mitochondrial gene expression and cell-cycle Perforin, TNFb; Stimulatory: GM-CSF, IL2, IL5, IL7, IL8, IL9, IL12, phase. Dimensional reduction was performed using Harmony (22). IL15, IL21; Chemoattractive: CCL11, IP-10, MIP-1b, RANTES; Reg- Missing values were imputed using the RunALRA function in the ulatory: IL4, IL10, IL13, IL22, sCD137, sCD40L, TGFb1; and Inflam- SeuratWrappers package (23). For the whole PBMC single-cell data, matory: IL6, IL17A, IL17F, MCP-1, MCP-4, IL1b. cell types were assigned to individual cells using SingleR (27). BlueprintEncodeData was used as a reference (28, 29). TP53 Sequencing TP53 gene sequencing was performed using either the Foundation Statistical analysis Medicine assay as reported previously (12) or using the SureMASTR All statistical analyses were performed in R or Graphpad. For TP53 sequencing assay (Agilent). For the latter, purified DNA from survival curves (RFS and OS), the log-rank statistic was reported (or FFPE breast tumor sections were amplified and sequenced according trend-test for >2 groupings). For gene-level analysis, nominal P values to the manufacturer’s standard protocol. Samples were sequenced to a were calculated using a Cox proportional hazards model, and then an depth of approximately 10,000 and mutations were called using the adjusted P value (q value) was calculated on the basis of the FDR (30). SureCall software (Agilent). Mutation allele frequency was set at 5% All single-cell statistical analyses were calculated in R using the Seurat and only likely functional mutations (early stops, frameshift deletions, package (25). Shared Nearest Neighbors were calculated using the and known recurrent hotspot single-nucleotide variation mutations) Harmony reduction, and clusters were identified at a resolution of 0.3, were selected for sample annotation. defining 5 total clusters. UMAP was performed for visualization, and

AACRJournals.org Clin Cancer Res; 2020 OF3

Axelrod et al.

missing values were imputed using ALRA (23). Visualization ably due to the selection strategy of including only patients who lacked and graph generation was performed in R. Some heatmaps were made pCR. Consistent with prior literature that the prognostic and predic- using the complex heatmap package in R (31). P-value cut-offs tive effect of sTILs differs by breast cancer subtype (32), neither pre- displayed on plots correspond to “ns” equals P > 0.05, equals 0.01 NAC nor post-NAC sTILs were prognostic for OS when considering < P ≤ 0.05, equals 0.001 < P ≤ 0.01, equals 0.0001 < P ≤ 0.001, our entire cohort (n ¼ 83). However, post-NAC sTILs were prognostic equals P ≤ 0.0001. Code used to generate figures can be for RFS (P ¼ 0.031) in the whole cohort (Supplementary Figs. S2B and accessed at https://github.com/MLAxelrod/Immunologic_changes_ S3A). This effect seems primarily driven by TNBC tumors as post- with_chemotherapy_inTNBC. NAC sTILs are not prognostic of either RFS or OS in non-TNBCs (Supplementary Fig. S3B). Neither pre-NAC nor post-NAC sTILs þ þ were prognostic in either ER or HER2 patients only (Supplementary Results Fig. S4). This may be limited by our sample size as sTILs have been þ sTILs in RD prognosticate improved outcomes in patients with previously shown to be associated with longer RFS in HER2 patients TNBC with incomplete response to NAC with cancer and shorter OS in luminal/HER2 patients with cancer We procured matched archived pretreatment (diagnostic biopsy) (32). Stratifying patients with TNBC by whether sTILs were qualitatively and posttreatment (RD surgical specimen) tumor specimens from a increased or decreased/equivocal in the surgical resection compared series of 83 patients, including 44 patients with TNBC. Importantly, we with the diagnostic biopsy did not provide any prognostic capability in refined our study to include only patients who had RD at surgery for this cohort (Supplementary Fig. S5A). Interestingly, most patients, analysis, thereby excluding patients who achieved pCR. This was a regardless of clinical subtype, had a decrease in sTILs over the course purposeful selection strategy, as patients with pCR usually experience of NAC (Supplementary Fig. S5B). Importantly, our cohort includes good outcomes, and we instead chose to focus on patients with RD for only patients with RD, and there is currently not a validated method for whom additional risk stratification could identify those who are most quantifying sTILs in pCR where, by definition, a tumor is no longer likely to benefit from additional therapies. Metadata for the patients, present. Thus, in our cohort, abundance of sTILs has the strongest including treating institution, molecular subtype (PAM50), RFS, OS, prognostic effect for the post-NAC surgical resection specimen in TP53 mutation status, and other molecular and clinical data are TNBC tumors with an incomplete response to NAC. These findings, summarized in Table 1. Individual patient-level data are available in consistent with both retrospective studies and analyses from random- Supplementary Table S1. ized controlled trials, prompted us to perform more detailed molecular As sTILs have been described and rigorously validated in breast studies aimed at understanding how NAC influences the TIME. cancer as both a prognostic factor (in surgical specimens for postsur- þ gical outcomes, particularly in TNBC and HER2 cancers), and a Suppression of immunologic gene expression with NAC in TNBC predictive factor (in diagnostic biopsies for benefit from NAC), we first To measure transcriptional changes occurring in the TIME induced asked whether these findings were consistent with our study cohort. by NAC, we performed gene expression profiling for a series of 770 Using the published cutoff (30%) of sTILs (4), we found that higher immune-related genes using nanoString (Pan-Cancer Immunology abundance of sTILs in the post-NAC RD in patients with TNBC (n ¼ panel), before and after NAC in the entire cohort (n ¼ 83). Tran- 44) was significantly prognostic for both RFS (log-rank P ¼ 0.019) and scriptional patterns and hierarchical clustering for all data primarily OS (P ¼ 0.05; Fig. 1A). Representative histology of high (>30%) and segregated tumors based on receptor status (ER/PR/HER2) and/or low (<30%) sTILs are shown in Supplementary Fig. S1. Interestingly, molecular subtype, with most luminal/hormone receptor–positive þ pre-NAC sTILs in the diagnostic biopsy were not prognostic for tumors appearing in the first cluster, most HER2 tumors in the outcomes in patients with TNBC (Supplementary Fig. S2A), presum- second cluster, and most basal-like/TNBC tumors in the third cluster (Fig. 1B). Examining the data as the change in gene expression for each gene after NAC in a patient-matched fashion (D expression; post-NAC Table 1. Demographic information for tumor cohort. minus pre-NAC) yielded similar patterns, with a trend of most patients with TNBC having generalized decreased immune gene expression DART EA1311 PERU Entire cohort patterns after NAC (Fig. 1C). To test for effects driven by differences in n (%) n (%) n (%) n (%) sampling (i.e., pre-NAC samples are biopsies whereas post-NAC samples are surgical resection samples) or treating institution, we Total 24 11 48 83 PAM50 Subtype performed principal component analyses. We did not detect signif- Basal 7 (29) 4 (36) 31 (65) 42 (51) icant clustering by time of sampling or cohort (Supplementary Fig. S6). HER2 enriched 2 (8) 3 (27) 10 (21) 15 (18) Luminal A 6 (25) 1 (9) 1 (2) 8 (10) NAC-induced immunologic gene expression is a positive Luminal B 8 (33) 3 (27) 1 (2) 12 (14) predictor of outcome in TNBC Normal 1 (4) 0 (0) 4 (8) 5 (6) While the TIME change in sTIL abundance did not prognosticate NA 0 (0) 0 (0) 1 (2) 1 (1) outcome in patients with TNBC, we asked whether changes in IHC Subtype þ individual immune-related genes are associated with outcome. We ER 20 (83) 5 (45) 1 (2) 26 (31) performed iterative Cox proportional hazards models, using the delta PRþ 13 (54) 4 (36) 1 (2) 18 (22) þ (D) of each gene (post-NAC minus pre-NAC) in an independent HER2 6 (25) 2 (18) 12 (25) 20 (24) TNBC 4 (17) 5 (45) 35 (73) 44 (53) univariate analysis, for both RFS and OS. All analyses are reported – Neoadjuvant chemotherapy using a nominal P value as well as an FDR (Benjamini Hochberg Taxane 21 (88) 7 (64) 21 (44) 49 (59) method) q value for association with RFS or OS. After correction for No taxane 3 (13) 4 (36) 27 (56) 34 (41) FDR (q < 0.10), upregulation of 11 genes were associated with improved RFS, while upregulation of only one gene was significantly Abbreviations: NA, not applicable; PR, progesterone receptor. associated with worse RFS (CDH1, which encodes e-cadherin) in our

OF4 Clin Cancer Res; 2020 CLINICAL CANCER RESEARCH

Immunologic Changes with Chemotherapy in TNBC

Figure 1. Immunologic changes in breast tumors after NAC. A, High levels of sTILs are associated with RFS (left; n ¼ 41) and OS (right; n ¼ 42) after surgery in TNBC. Patients are stratiﬁed on the basis of post-NAC sTILs ≤ 30% or > 30%, scored as recommended by the International TILs Working Group (13, 14), according to the predeﬁned cut point (4). B, Heatmap demonstrating gene expression patterns for 770 immune-related genes (NanoString Pan-Cancer Immunology panel) across all patients (TNBC and non-TNBC; n ¼ 83 total patients, 166 samples). C, Heatmap of gene expression patterns as detailed in B, instead depicting the change in expression of each gene in matched paired (pre- and post-NAC; n ¼ 83) samples. Red data points represent an upregulation, while blue data points represent a downregulation in the post- NAC RD compared with the pretreatment diagnostic biopsy.

TNBC cohort. Interestingly, e-cadherin is known to interact with killer in Supplementary Table S5. Kaplan–Meier visualization examples of cell lectin-like receptor G1 (KLRG1), an inhibitory receptor expressed strongly prognostic genes (negative prognostic: CDH1; positive prog- by memory T cells and NK cells (33). In contrast, upregulation of a nostic: CD70) reinforced the prominent association of TNBC disease larger number of genes was associated with improved OS (n ¼ 189) or outcomes with changes in immune gene expression during NAC reduced OS (n ¼ 15) at FDR q < 0.10 (Fig. 2A). These genes are listed (Fig. 2B). Conversely, no changes in immune-related gene expression

AACRJournals.org Clin Cancer Res; 2020 OF5

Axelrod et al.

Figure 2. Identification of immune-associated genes associated with RFS and OS in TNBC after chemotherapy. A, Individual genes (changes pre- to post-NAC) were tested iteratively in a univariate Cox proportional hazards model for their association with RFS (left) or OS (right) after chemotherapy and surgery. Individual genes are colored for their statistical significance (red: nominal P value <0.05; green: q value (FDR) < 0.10; black: not significant). Selected top genes are labeled but are limited in number for clarity. Genes with negative coefficients (left of the center line) are associated with better outcome, while genes with positive coefficients (right of the center line) are associated with worse outcome. B, Representative Kaplan–Meier plots for selected detrimental (CDH1; e-cadherin) and beneficial (CD70) genes are shown. Strata are defined by tertiles, and generally represent upregulation during NAC (blue), no change/equivocal (green), and downregulation (red). P values represent the log-rank test for trend.

were significantly associated with RFS or OS in patients with non- others. Although manual inspection revealed some overlap in these TNBC at q < 0.10 (Supplementary Fig. S7A). gene sets, they were largely composed of signature-exclusive genes Dimensional reduction through collapsing individual genes into (Supplementary Table S3). Kaplan–Meier visualization examples of pathways or defined functions can improve interpretation of high- strongly prognostic gene sets (“NK-cell functions” and “T-cell acti- dimensional data. Thus, we collapsed the gene expression data into vation”) reinforced the considerable association of changes in immune bioinformatically categorized immune signatures (sum scores, defined gene sets during NAC with outcomes (Fig. 3B). Interestingly, no gene fi as the summation of the log2 expression values for all genes in a or gene set was signi cantly associated with RFS or OS in patients with category). Organization of the data in this manner and testing the non-TNBC at q < 0.10 (Supplementary Fig. S7A and S7B). When the þ þ signatures (n ¼ 70) for association with RFS and OS yielded a non-TNBC group was separated into ER and HER2 groups, no gene surprising finding: all significant (q < 0.10) gene sets (n ¼ 40 for RFS or gene set was significantly associated with outcome (Supplementary and n ¼ 60 for OS; listed in Supplementary Table S6) identified in this Fig. S8). Thus, these data suggest that NAC, exclusively in TNBC, could analysis were associated with good outcome (Fig. 3A). Many of the promote immunologic activity leading to improved outcomes in a top-scoring gene sets were associated with T cells, including “T-cell subset of patients. However, these effects may be related to factors þ polarization,”“T-cell immunity,” and “T-cell activation,” among beyond TNBC biology, as hormone receptor–positive and HER2

OF6 Clin Cancer Res; 2020 CLINICAL CANCER RESEARCH

Immunologic Changes with Chemotherapy in TNBC

Figure 3. Upregulation of immune-associated gene sets after chemotherapy are associated with improved RFS and OS in TNBC. A, Gene set scores were calculated by summing expression levels of all gene set member genes across each candidate gene set (n ¼ 70). Changes pre- to post-NAC was then calculated for each patient with TNBC (n ¼ 44) and each gene set score was tested iteratively in a univariate Cox proportional hazards model for association with RFS (left) or OS (right) after chemotherapy and surgery. Individual gene sets are colored for their statistical significance [red: nominal P value < 0.05; green: q value (FDR) < 0.10; black: not significant]. Selected top gene sets are labeled but are limited for clarity. Gene sets with negative coefficients are associated with better outcome, while gene sets with positive coefficients are associated with worse outcome. B, Representative Kaplan–Meier plots for selected gene set changes with beneficial associations are shown (left: T-cell activation; right: NK-cell functions). Strata are defined by tertiles, and generally represent upregulation during NAC (blue), no change/equivocal (green), and downregulated (red). P values represent the log-rank test for trend. patients receive additional endocrine or HER2-directed therapy in the ing using the ImmunoSeq assay to estimate the number of unique adjuvant setting, complicating associations with RFS and possibly OS. T-cell clones (diversity), and the presence of expanded T-cell clones in Nonetheless, immune-related signatures, particularly those derived the TIME before and after NAC. Given the breadth of sTIL fractions from T cells, appeared to be strongly associated with improved out- observed among breast tumors as well as caveats associated with comes in TNBC. comparison of samples derived from diagnostic core needle biopsies versus surgical resections, we first verified that the number of pro- Changes in T-cell clonality and function in tumors and peripheral ductive T cells was associated with estimation of sTILs determined on blood induced by NAC adjacent sections. We detected a strong association between these A robust T-cell response is characterized by oligoclonal expansion parameters (R2 ¼ 0.6; P < 0.0001; Supplementary Fig. S9A), raising of antigen-specific T cells. Therefore, we next asked whether clonality confidence in the assay results. In this sample set, NAC did not of T cells in the TIME was altered during NAC. In a subset of samples universally alter productive clonality (Supplementary Fig. S9B), a (n ¼ 15; 8 TNBC, 7 non-TNBC), we performed TCRb chain sequenc- measurement of the number of times the same (productive) TCRb

AACRJournals.org Clin Cancer Res; 2020 OF7

Axelrod et al.

Figure 4. Evidence of enhanced T-cell functionality in the CD8þ PD-1HI peripheral compartment. A, Clinical details of 4 patients analyzed prospectively for changes in peripheral blood T-cell functionality. NST indicates no special type. B, Polyfunctionality of PD-1HICD4þ and PD-1HI CD8þ T cells isolated from PBMCs in 4 patients prior and after NAC (>1,000 individual cells/sample/timepoint) was determined by Isoplexis single-cell cytokine profiling. Polyfunctionality is defined asthepercentageofcellscapableofproducing≥ 2 cytokines following CD3/CD28 stimulation. The percentage of cells in each sample capable of secreting 2, 3, 4, or 5þ cytokines are depicted in stacked bars. Characteristics of each of the 4 patients are shown above the bars. Patients with TNBC (Pt. 1 and Pt. 4) had greater increases in polyfunctionality in the CD8þ compartment with NAC. C, Heatmap representation of log cytokine signal intensity of each cell in each þ patient sample, pre- and post-NAC. Each row represents one PD-1Hi CD8 T cell. White indicates no cytokine secreted. D, TCRb chain repertoire analysis in CD8þ peripheral blood T cells. Upper plots indicate the number of individual T cells sequenced plotted by sample on the left y axis; number of clonotypes (unique CDR3 amino acid sequences) plotted by sample on the right y axis. In the lower graph, each sample is divided into the number of clonotypes comprising expanded (hyperexpanded, large, medium, small, and rare) compositions of the detected repertoire (categories divided by orders of magnitude of fraction of total repertoire). E, The fraction of repertoire clonotypes identified in PD-1HI versus PD-1NEG CD8þ T cells (before or after NAC) classified as “hyperexpanded” or “large” (comprising > 0.1% of repertoire). P value represents a two-sample two-tailed t test.

OF8 Clin Cancer Res; 2020 CLINICAL CANCER RESEARCH

Immunologic Changes with Chemotherapy in TNBC

sequence is represented in the sample, which is a descriptor of T-cell is enriched for similarity to TILs relative to the PD-1NEG peripheral clonal expansion. When stratified by breast cancer subtype, there was compartment. no significant change in productive clonality with NAC (one-sample t þ test). However, TNBC tumors demonstrated a qualitative trend toward scRNAseq of peripheral PD-1HI CD8 T cells identifies a unique decreased clonality after NAC, while non-TNBC tumors trended population of cytolytic effector cells toward increased clonality after NAC. The difference between these Next, we utilized scRNAseq to describe the post-NAC peripheral þ two subgroups approached significance (P ¼ 0.054; two-sample t -test; PD-1HI CD8 T-cell populations at the time of surgery in the blood Supplementary Fig. S9C). There was no association of change in of 2 patients with TNBC: one with RD (Pt. 1) and one with matrix- clonality with change in sTILs, suggesting that changes in sTIL producing metaplastic TNBC who experienced pCR (Pt. 4). Uni- abundance after NAC are not necessarily due to expansion of existing form Manifold Approximation and Projection (UMAP) analysis clones (Supplementary Fig. S9D). was performed on Harmony-normalized samples to adjust for To further explore changes in T-cell clonality and function in intersample technical variation, and we stratified cells based on 5 response to chemotherapy, we prospectively collected PBMCs from clusters identified through the Louvain algorithm (Fig. 5A and B). 4 patients with breast cancer (including 2 TNBCs) before and after A heatmap of the top 10 most differentially expressed genes by NAC (Fig. 4A). In addition, the post-NAC RD (or in one case, pCR clusterispresentedinFig. 5C. Although the composition of the residual scar) was analyzed in tandem. On the basis of previous cells was largely similar, we identified one cluster (cluster 0) which findings demonstrating that tumor-reactive T cells are enriched in was enriched in Pt. 4. Examination of genes differentially expressed þ the CD8 PD-1HI population of peripheral T cells (34), we purified in this cluster of cells suggested a cellular identity concordant with þ þ CD4 and CD8 cells from each sample by FACS, further strat- that of highly cytotoxic memory (TBX21-expressing) T cells, which ifying by PD-1NEG and PD-1HI status (gating scheme shown in had an abundance of MHC-I (HLA-A/B/C) and MHC-II (e.g., HLA- Supplementary Fig. S10). Using a functional fluorescence ELISA- DRA, HLA-DRB5) family member expression as well as expression based assay of cytokine (32-plex antibody barcoded chip; Supple- of cytolytic and immune checkpoint genes [e.g., LAG3, FCRL6 (35), mentary Fig. S11A) secretion following CD3/CD28 stimulation, we and higher transcriptional expression of PDCD1; Fig. 5C and D]. determined that PD-1HI peripheral T cells had functional capacity, Verification of the pattern of expression of key cytolytic and killer- secreting multiple cytokines following activation, and these effects identity genes [GNLY (granulysin), GZMB (granzyme B), and þ were particularly pronounced in CD8 T cells (Supplementary FGFBP2 (killer-secreted protein 37)] showed that these genes were Fig. S11B). In 2/2 patients with TNBC, the percentage of “poly- almost exclusively expressed in cluster 0 (Supplementary þ functional” PD-1HI CD8 T cells—those capable of expressing Fig. S13A). This analysis also demonstrated purity-of-sort in that multiple cytokines after TCR stimulation—were increased follow- all clusters expressed CD8A and PDCD1,butnotCD4 (Supple- þ ing NAC (Fig. 4B). In contrast, 2/2 ER patients with breast cancer mentary Fig. S13B). experienced a drop or stasis in the functionality of the PD-1HI These data led us to propose two competing hypotheses: (i) Cluster þ þ CD8 population of cells following NAC (Fig. 4B). Of note, the 0 genes, reflective of cytolytic CD8 T cells, are a positive prognostic þ þ patient with ER HER2 disease has a near complete loss of T-cell factor reflective of robust antitumor immunity as evidenced by their functionality after NAC. Cytokines produced by individual PD-1HI enrichment in the patient with metaplastic TNBC with pCR, or (ii) þ CD8 cells in patients with TNBC were primarily effector (e.g., Cluster 0 genes are reflective of ongoing disease including the micro- Granzyme B, IFNg,MIP-1a,TNFa,andTNFb) and chemoattrac- metastatic component that cannot be sampled from the primary þ tive (MIP-1b)cytokines(Fig. 4C). PD-1HI CD4 T cells also tumor. The second hypothesis is supported by the observations that produced primarily effector cytokines including IFNg and TNFa pCR is less prognostic of RFS and OS in metaplastic disease (36, 37) (Supplementary Fig. S11C). and that rates of recurrence following chemotherapy are higher for þ þ PD-1HI CD8 and PD-1NEG CD8 T cells from pre-NAC and metaplastic disease than nonmetaplastic TNBC (36, 38, 39). Interest- post-NAC blood (except patient 4, for whom a sufficient pre-NAC ingly, matrix-producing metaplastic breast cancer (Pt. 4) has been sample was not available) were also analyzed by TCR sequencing. shown to be associated with pCR to NAC, but often can still recur While the number of detected T cells was consistent among all despite pCR (36, 40). Follow-up for this individual patient was samples, the clonotypes detected (unique TCRs) were considerably immature at the time of reporting, and thus recurrence, and therefore þ lower in PD-1HI CD8 T cells (Fig. 4D). This suggests that there are presence of micrometastatic disease at the time of sampling, cannot be more repetitive sequences detected in the PD-1HI population, ruled out. indicating clonal expansion. Consistent with this observation, the To gain a deeper understanding of cluster 0 genes, we performed proportion of the overall TCR repertoire occupied by expanded additional scRNAseq on post-NAC whole PBMCs from 2 patients clonotypes (large or hyperexpanded clonotypes consisting of greater with TNBC with pCR (Pt. 4, described above, and Pt. 5, not used in than 0.1% or 1% of the total repertoire, respectively) was substan- any prior analysis). UMAP was used for dimensionality reduction on þ tially higher in the PD-1HI than in PD-1NEG CD8 T-cell fractions Harmony-normalized samples. SingleR was used to computationally (Fig. 4E). We additionally sequenced the TCR repertoire in the assign cell-type annotations (Fig. 5E). Analysis of the top 5 differen- post-NAC RD, although the number of T cells sequenced in these tially expressed transcripts by cluster (Supplementary Fig. S14A) and samples were limited because of fixation of tissue and small T-cell expression of key cell-type identity markers (Supplementary Fig. S14B) abundance as a function of total RNA in the bulk samples, and thus validated annotations by SingleR. Low-confidence cell-type annota- should be interpreted with caution. Nonetheless, we found that the tions were collapsed into the category “other” and make up a minor similarity (Jaccard index, normalized to size of repertoire detected) fraction of cells (Supplementary Fig. S15A). Seven genes enriched in of tumor-infiltrating TCRs in the post-NAC sample was universally cluster 0 (PDCD1, NKG7, LAG3, GZMH, GZMB, GNLY, FGFBP2) and þ more similar to the PD-1HI CD8 peripheral TCR repertoires, one gene strongly deenriched in Pt. 4 (HLA-G) were chosen as an þ compared with the PD-1NEG CD8 repertories (Supplementary 8-gene signature for downstream applications, including validation in Fig. S12). This suggests that the PD-1HI peripheral compartment a larger cohort of patients (Supplementary Fig. S15B and S15C).

AACRJournals.org Clin Cancer Res; 2020 OF9

Axelrod et al.

Figure 5. scRNAseq of CD8þ PD-1HI peripheral T cells from 2 patients with TNBC after NAC demonstrate high expression of cytolytic markers. A, UMAP plots of 1,964 PD-1HI CD8þ peripheral T cells across 2 patients (672 and 1,292 respectively) are shown. Five clusters (0–4) were defined. B, Percent of cells sequenced comprising each cluster are plotted. C, Heatmap identifying top 10 most differentially expressed transcripts across clusters. D, Aselectionofgenesdefining cluster 0 are highlighted. Data depicted include combined cells from both Pt. 1 and Pt. 4. E, UMAP plots of 7,062 PBMCs from 2 patients with TNBC (3,525 cells from Pt. 4 and 3,537 cells from Pt. 5). Cell type annotations are defined by SingleR. The 8-gene score is defined by expression of FGFBP2 þ GNLY þ GZMB þ GZMH þ NKG7 þ LAG3 þ PDCD1 – HLA-G. F, Violin plots showing expression of the 8-gene score by cell type. Overlaid box plots show mean and interquartile range for each cell type.

Expression of this 8-gene score (PDCD1 þ NKG7 þ LAG3 þ GZMH þ tary Fig. S16A and S16B). Strong expression of this signature in post- þ GZMB þ GNLY þ FGFBP2 – HLA-G) was the highest in CD8 T cells NAC whole PBMCs led us to test whether this signature was predictive þ (as expected, given the derivation from PD-1HI CD8 T cells) and NK of response in archived whole blood samples from patients with breast cells (Fig. 5F). This ﬁnding was similar in both patients (Supplemen- cancer.

OF10 Clin Cancer Res; 2020 CLINICAL CANCER RESEARCH

Immunologic Changes with Chemotherapy in TNBC

Table 2. Demographic information for peripheral blood cohorts. significantly differentially expressed or approached statistical significance across the outcome groups (Kruskal–Wallis test). Comparisons Blood cohort, n (%) were particularly striking between the group of patients with RD who Vanderbilt DFCI experienced early disease recurrence and the group with pCR following þ Total 58 30 NAC (post hoc Dunn test; Fig. 6A). A composite score of PDCD1 IHC Subtype NKG7 þ LAG3 þ GZMH þ GZMB þ GNLY þ FGFBP2 – HLA-G also ERþ 27 (47) 27 (90) demonstrated statistically significant associations with presence of PRþ 26 (45) 27 (90) ongoing disease (Fig. 6B). Interestingly, expression levels of these þ HER2 17 (29) 0 (0) genes did not always correlate with one another, indicating hetero- TNBC 21 (36) 0 (0) geneity in their expression patterns and some degree of independence Response (Fig. 6C). Trends in gene expression were similar for patients with No NAC 24 (41) 0 (0) TNBC and non-TNBC, but in-depth subgroup analyses were limited pCR 10 (17) 1 (3) RD 24 (41) 29 (97) by sample size. Thus, peripheral antitumor immunity in blood may be RCB I not recur 3 (10) a useful measure of persistent residual primary or micrometastatic RCB I recur 2 (7) disease and could identify patients likely to benefit from additional RCB II not recur 7 (23) therapy. RCB II recur 4 (13) To extend these findings, we evaluated gene expression in a RCB III 13 (43) second cohort of archival blood from patients with breast cancer NAC treated at DFCI (Boston, MA; n ¼ 30), with blood collected in the No taxane 7 (12) 0 (0) interlude between completing NAC and definitive surgery. Notably, Taxane 27 (47) 30 (100) this cohort had differing baseline characteristics (summarized No NAC 24 (41) 0 (0) in Table 2). All of the patients were hormone receptor positive (a subtype known to have a lower pCR rate than patients with TNBC), HER2 , and all were treated with bevacizumab. Intrigu- Cytolytic gene expression signatures are present in blood and ingly, expression of the 8-gene composite score was also correlated associated with increased likelihood of recurrence with enhanced disease burden in this cohort, being highest in To determine whether cytolytic signatures representative of cluster patients with RCB III (regardless of recurrence) and lowest in 0 genes were associated with disease outcome, we evaluated archived patients with RCB 0/I/II who did not have a breast cancer recur- whole blood from a series of 58 patients with breast cancer. All samples rence within 3 years (Supplementary Fig. S17). These findings were collected within 14 days preceding surgical resection for primary suggest that peripheral blood gene expression may be predictive breast cancer, with 34 samples having received NAC, in addition to 24 of response in diverse groups of patients with breast cancer. samples from untreated patients. As described above, a series of eight genes enriched in cluster 0 (PDCD1, NKG7, LAG3, GZMH, GZMB, GNLY, FGFBP2, HLA-DRB5), one gene enriched in Pt. 1 (RCB-II) over Discussion Pt. 4 (pCR; HLA-G; Supplementary Fig. S15B and S15C), and three With the recent approval of immunotherapy in combination with normalization control genes (PTPRC, RPL13a, TBP; ref. 41) were chemotherapy in mTNBC, and promising early results in the neoad- selected for a 12-gene custom NanoString gene expression analysis. juvant setting, an improved understanding of how chemotherapy HLA-G has been described as an immune checkpoint which can reshapes antitumor immunity, both in the tumor and in the peripheral dampen anti-tumor immune responses (42–44), and thus we expected compartment, is needed. Perhaps, the most widely studied marker to HLA-G expression to be inversely correlated with the other selected approximate antitumor immunity is sTILs, which are predictive of genes, as is the case in our scRNAseq dataset. One of these genes improved NAC response when measured in the primary untreated performed poorly (HLA-DRB5), likely due to frequent polymorphisms tumor and are prognostic of good outcomes in the RD of patients in the gene leading to highly variable probe binding and was therefore lacking pCR. Moreover, sTILs have primarily been a useful biomarker omitted from further analysis. Information on the presence of pCR/RD in TNBC as opposed to hormone receptor–positive cancers, but this at surgery, ER/PR/HER2 status, and clinical follow-up (recurrence at inference is complicated by the routine use of endocrine-targeted 1,000 days after surgery for patients with RD) was collected (Table 2; agents in the adjuvant setting, that can impact postsurgical outcomes. Supplementary Table S7). However, this likely confounder leaves space for a biological contri- Nearly, all tested genes demonstrated a pattern supporting the bution, as numerous differences between molecular and clinical sub- hypothesis that gene expression in whole blood is associated with types exist. Even when considering only TNBC, quantification of sTILs ongoing disease, being highest (or lowest in the case of HLA-G)in is an imperfect biomarker, which does not precisely inform on the untreated patients (who have ongoing tumor burden by virtue of not immunobiology of the tumor. having received therapy prior to surgery) and those with RD compared In this study, we present a molecular analysis of the TIME in with those with pCR. However, it is important to note that there is a response to NAC in 83 patients with breast cancer, specifically focusing high degree of heterogeneity in the untreated patients, possibly on patients lacking a pCR, as these patients have worse outcomes. We reflecting heterogeneity in tumor size, strength of antitumor immune found that changes in tumor immunity seem to be most prevalent in response at baseline, or antitumor response following systemic ther- TNBC, often resulting in decreases in expression of immune-related apy. Furthermore, among patients with RD, higher expression (or genes. However, an upregulation of immune-related gene expression lower in the case of HLA-G) tended to be observed in patients who had in tumors following NAC was associated with a strikingly improved early recurrences in the first 3 years following surgery (and thus may outcome after surgery, specifically in TNBC. Of these genes, those have had micrometastatic disease at the time of surgery). Several of involved in cytotoxic effector cells were among the most robustly these genes (FGFBP2, GNLY, PDCD1, LAG3, and NKG7) were also associated with outcome. Furthermore, we found that cytokines

AACRJournals.org Clin Cancer Res; 2020 OF11

Axelrod et al.

Figure 6. An 8-gene activated T-cell signature derived from whole blood at surgery is associated with pCR and prognosticates recurrence in patients with RD. A, Individual gene plots of 8 analyzed genes by nanoString from RNA derived from whole blood sampled within 14 days leading up to definitive surgery. Datapoints are stratified by untreated patients (No NAC), those experiencing pCR (pCR), those with RD that did not recur (RD not recur), and those with RD that recurred (RD recur) within 3 years after surgery. Box plots represent the interquartile range. P values represent Kruskal–Wallis tests. , P < 0.05 by post hoc Dunn test. B, A composite gene signature derived as PDCD1 þ NKG7 þ LAG3 þ GZMH þ GZMB þ GNLY þ FGFBP2 – HLA-G (sum of Z-scores), stratified by outcome, as in A. C, Heatmap showing row standardized (Z-score) gene expression for genes assayed across all patients.

þ expressed by PD-1HI CD8 T cells in the peripheral blood were esis was confirmed in two validation sets totaling 88 patients with increased dramatically in patients with TNBC following NAC. breast cancer and serves as a proof-of-principle for the use of this Analysis of TCR clonotypes infiltrating into tumors suggested that signature as a possible biomarker of outcome. chemotherapy may preferentially increase the recruitment of new There are several limitations of our study. We intentionally chose to T-cell clones into the tumor, rather than expanding the T cells already focus on patients lacking a pCR for the tumor study, as these patients present. This effect was consistent with that observed in the peripheral have the worst outcomes after surgery, and because of the caveats and þ blood, where PD-1HI CD8 T cells, while highly clonal compared with difficulty in defining or assessing antitumor immunity after chemo- PD-1NEG cells, did not substantially change in clonality during NAC; therapy in surgical specimens that lack tumor tissue. Future studies will these observations reflect a lack of clonal expansion in response to need to be done to determine whether alterations in immune-related NAC, as we found in the TIME. genes are seen over the course of NAC in patients experiencing pCR. Assessment of peripheral blood represents a unique opportunity to Furthermore, our sample size was limited in our exploration of þ monitor antitumor immunity through minimally invasive means. peripheral blood PD-1Hi CD8 T cells by scRNAseq to 2 patients. Using scRNAseq, we identified a population of cytolytic effector T Using this method, we identified a unique cluster of cytotoxic cells and cells in blood that expressed elevated levels of exhaustion/checkpoint derived a gene signature. We used further scRNAseq to identify that þ genes. A gene expression signature derived from this population was this gene signature is expressed predominately by circulating CD8 T used to test the hypothesis that these highly cytolytic, but potentially cells and NK cells. Therefore, expression of this gene signature in whole þ exhausted cells may be reflective of an ongoing disease process, and blood is not limited to PD-1Hi CD8 T cells. We tested expression of therefore a peripheral approximation of disease burden. This hypoth- this gene signature in whole blood of two different cohorts of patients

OF12 Clin Cancer Res; 2020 CLINICAL CANCER RESEARCH

Immunologic Changes with Chemotherapy in TNBC

with breast cancer and found that expression was highest in those with funding from Genentech to Dana-Farber Cancer Institute for the clinical trial under RD who recurred or had the largest residual cancer burden. This which samples included in the submitted work were obtained) during the conduct of provides proof-of-concept that peripheral blood gene signatures may the study. T.W. Miller reports grants from Takeda Pharmaceuticals outside the submitted work. I.A. Mayer reports grants and personal fees from Pfizer and predict response, but optimization of signature genes is likely needed Genentech (research grant and advisory board participation); and personal fees from to improve prediction. Novartis, Lilly, AstraZeneca, GSK, Abbvie, Seattle Genetics, Immunomedics, Macro- Importantly, there has been a paucity of studies looking at the effect genics, Eisai, and Puma (advisory board participation) outside the submitted work. of chemotherapy on peripheral blood (45), with little data on disease J.M. Balko reports grants from Genentech/Roche and Incyte Pharma, personal fees outcomes. This study provides a novel assessment and framework for from Novartis Inc (consulting), and grants from BMS outside the submitted work, and an improved understanding of how chemotherapy alters antitumor is listed as a coinventor on a provisional patent application on methods to predict therapeutic outcome using blood-based gene expression patterns, that is owned by immunity both in the TIME and the peripheral compartment. These Vanderbilt University Medical Center and is currently unlicensed. No potential data represent a unique opportunity to better understand patient conflicts of interest were disclosed by the other authors. populations most likely to benefit from the addition of immunotherapy to chemotherapy, particularly in the neoadjuvant setting. Fur- Authors’ Contributions thermore, our findings demonstrating the association of expression of M.L. Axelrod: Conceptualization, data curation, formal analysis, investigation, key cytolytic and immune-activation genes in the peripheral blood visualization, writing-original draft. M.J. Nixon: Conceptualization, data curation, with presence of RD and recurrence represent a possible biomarker formal analysis, investigation. P.I. Gonzalez-Ericsson: Formal analysis, writing- platform. Peripheral gene expression signatures may identify high-risk review and editing. R.E. Bergman: Data curation. M.A. Pilkinton: Investigation, writing-review and editing. W.J. McDonnell: Formal analysis, writing-review and populations with potentially exhausted T cells and either primary or editing. V. Sanchez: Investigation. S.R. Opalenik: Investigation, project fi micrometastatic disease who may bene t from additional immuno- administration. S. Loi: Resources, writing-review and editing. J. Zhou: Formal therapeutic strategies. analysis, methodology. S. Mackay: Formal analysis, methodology. B.N. Rexer: Resources, writing-review and editing. V.G. Abramson: Resources, writing-review Disclosure of Potential Conflicts of Interest and editing. V.M. Jansen: Resources, writing-review and editing. S. Mallal: M.L. Axelrod is listed as a coinventor on a provisional patent application on Resources, writing-review and editing. J. Donaldson: Resources, data curation, methods to predict therapeutic outcome using blood-based gene expression patterns, writing-review and editing. S.M. Tolaney: Resources, data curation, writing- that is owned by Vanderbilt University Medical Center, and is currently unlicensed. review and editing. I.E. Krop: Resources, writing-review and editing. M.J. Nixon reports grants from NIH/NCI and DoD during the conduct of the study. A.C. Garrido-Castro: Resources, writing-review and editing. J.D. Marotti: W.J. McDonnell reports personal fees and other from 10 Genomics (currently a Resources, writing-review and editing. K. Shee: Resources, writing-review and shareholder and employee of 10 Genomics, which occurred after publication) editing. T.W. Miller: Resources, writing-review and editing. M.E. Sanders: outside the submitted work. S. Loi reports nonfinancial support from Seattle Genetics, Formal analysis, writing-review and editing. I.A. Mayer: Resources, data curation, Pfizer, Novartis, BMS, Merck, AstraZeneca, and Roche-Genentech [consultant writing-review and editing. R. Salgado: Resources, writing-review and editing. (nonremunerated)]; other from Aduro Biotech, Novartis, GlaxoSmithKline, J.M. Balko: Conceptualization, formal analysis, supervision, funding acquisition, Roche-Genentech, and G1 Therapeutics [consultant (remunerated, paid to visualization, writing-original draft. institution)]; and other from Novartis, Bristol Meyers Squibb, Merck, Roche- Genentech, Puma Biotechnology, Pfizer, Eli Lilly, and Seattle Genetics [research Acknowledgments funding (to institution)] outside the submitted work; and is supported by the National Funding for this work was provided by Susan G. Komen Career Catalyst Grant Breast Cancer Foundation of Australia Endowed Chair and the Breast Cancer CCR14299052 (to J.M. Balko), NIH/NCI R00CA181491 (to J.M. Balko), NIH/NCI Research Foundation, New York. V.M. Jansen reports personal fees from Eli Lilly SPORE 2P50CA098131–17 (to J.M. Balko), Department of Defense Era of Hope and Company (past employee) and Mersana Therapeutics (current employee) outside Award BC170037 (to J.M. Balko), and the Vanderbilt-Ingram Cancer Center the submitted work. S.M. Tolaney reports grants and personal fees from Genentech/ Support Grant P30 CA68485. Additional funding was provided by NIH fi Roche, Merck, Eli Lilly, Novartis, AstraZeneca, Nektar, Nanostring, P zer, Immu- T32GM007347 (to M.L. Axelrod) and F30CA236157 (to M.L. Axelrod); fi nomedics, Bristol-Myers Squibb, Eisai, Sano , Odonate, and Seattle Genetics F30CA216966 (to K. Shee); R01CA200994 (to T.W. Miller) and R01CA211869 (research funds to institution; honorarium for consulting/advisory board participa- (to T.W. Miller); and Dartmouth College Norris Cotton Cancer Center Support tion); grants from Cyclacel and Exelixis (research funds to institution); personal fees Grant P30CA023108 (to T.W. Miller). from Puma, Celldex, Daiichi-Sankyo, Silverback Therapeutics, G1 Therapeutics, Abbvie, Athenex, OncoPep, Kyowa Kirin Pharmaceuticals, Samsung Bioepsis Inc., and CytomX (honorarium for consulting/advisory board participation) outside the The costs of publication of this article were defrayed in part by the payment of page submitted work. I. Krop reports grants from Genentech (to institution) during the charges. This article must therefore be hereby marked advertisement in accordance conduct of the study; personal fees from Genentech/Roche; grants from Pfizer with 18 U.S.C. Section 1734 solely to indicate this fact. (to institution); personal fees from Bristol Meyers Squibb, Daiichi-Sankyo, Macro- genics, Context Therapeutics, Taiho Oncology, Seattle Genetics, Novartis (for DSMB participation), Merck (for DSMB participation), Celltrion, and AstraZeneca outside Received November 12, 2019; revised May 21, 2020; accepted August 18, 2020; the submitted work. A.C. Garrido-Castro reports other from Genentech (institutional published first August 21, 2020.

References 1. Schmid P, Adams S, Rugo HS, Schneeweiss A, Barrios CH, Iwata H, et al. 2-positive and triple-negative primary breast cancers. J Clin Oncol 2015;33: Atezolizumab and Nab-paclitaxel in advanced triple-negative breast cancer. 983–91. N Engl J Med 2018;379:2108–21. 4. Loi S, Drubay D, Adams S, Pruneri G, Francis PA, Lacroix-Triki M, et al. Tumor- 2. Schmid P, Cortes J, Dent R, Pusztai L, McArthur HL, Kuemmel S, et al. infiltrating lymphocytes and prognosis: a pooled individual patient analysis of LBA8_PRKEYNOTE-522: phase III study of pembrolizumab (pembro) þ early-stage triple-negative breast cancers. J Clin Oncol 2019;37:559–69. chemotherapy (chemo) vs placebo (pbo) þ chemo as neoadjuvant treatment, 5. Luen SJ, Salgado R, Dieci MV, Vingiani A, Curigliano G, Gould RE, et al. followed by pembro vs pbo as adjuvant treatment for early triple-negative breast Prognostic implications of residual disease tumor-infiltrating lymphocytes and cancer (TNBC). Ann Oncol 2019;30:v853–4. residual cancer burden in triple-negative breast cancer patients after neoadju- 3. Denkert C, von Minckwitz G, Brase JC, Sinn BV, Gade S, Kronenwett R, et al. vant chemotherapy. Ann Oncol 2019;30:236–42. Tumor-infiltrating lymphocytes and response to neoadjuvant chemotherapy 6. LoiS,DushyanthenS,BeavisPA,SalgadoR,DenkertC,SavasP,etal.RAS/ with or without carboplatin in human epidermal growth factor receptor MAPK activation is associated with reduced tumor-infiltrating lymphocytes

AACRJournals.org Clin Cancer Res; 2020 OF13

Axelrod et al.

in triple-negative breast cancer: therapeutic cooperation between MEK and 23. Linderman GC, Zhao J, Kluger Y. Zero-preserving imputation of scRNA-seq PD-1/PD-L1 immune checkpoint inhibitors. Clin Cancer Res 2016;22:1499– data using low-rank approximation. bioRxiv 2018;397588. 509. 24. ImmunoMind. ImmunoMind Team. immunarch: an R package for painless 7. Dieci MV, Criscitiello C, Goubar A, Viale G, Conte P, Guarneri V, et al. analysis of large-scale immune repertoire data; 2019. Available from: https:// Prognostic value of tumor-infiltrating lymphocytes on residual disease after immunarch.com/. primary chemotherapy for triple-negative breast cancer: a retrospective multi- 25. Satija R, Farrell JA, Gennert D, Schier AF, Regev A. Spatial reconstruction of center study. Ann Oncol 2014;25:611–8. single-cell gene expression data. Nat Biotechnol 2015;33:495–502. 8. Parra ER, Villalobos P, Behrens C, Jiang M, Pataer A, Swisher SG, et al. Effect of 26. Butler A, Hoffman P, Smibert P, Papalexi E, Satija R. Integrating single-cell neoadjuvant chemotherapy on the immune microenvironment in non-small cell transcriptomic data across different conditions, technologies, and species. lung carcinomas as determined by multiplex immunofluorescence and image Nat Biotechnol 2018;36:411–20. analysis approaches. J Immunother Cancer 2018;6:48. 27. Aran D, Looney AP, Liu L, Wu E, Fong V, Hsu A, et al. Reference-based analysis 9. Kim R, Kawai A, Wakisaka M, Sawada S, Shimoyama M, Yasuda N, et al. of lung single-cell sequencing reveals a transitional profibrotic macrophage. Immune correlates of the differing pathological and therapeutic effects of Nat Immunol 2019;20:163–72. neoadjuvant chemotherapy in breast cancer. Eur J Surg Oncol 2020;46:77–84. 28. ENCODE Project Consortium. An integrated encyclopedia of DNA elements in 10. Li X, Warren S, Pelekanou V, Wali V, Cesano A, Liu M, et al. Immune profiling of the human genome. Nature 2012;489:57–74. pre- and post-treatment breast cancer tissues from the SWOG S0800 neoadju- 29. Martens JH, Stunnenberg HG. BLUEPRINT: mapping human blood cell epi- vant trial. J Immunother Cancer 2019;7:88. genomes. Haematologica 2013;98:1487–9. 11. Tolaney SM, Boucher Y, Duda DG, Martin JD, Seano G, Ancukiewicz M, et al. 30. Benjamini Y, Hochberg Y. Controlling the false discovery rate: a practical and Role of vascular density and normalization in response to neoadjuvant bev- powerful approach to multiple testing. J R Stat Soc Series B Methodol 1995;57: acizumab and chemotherapy in breast cancer patients. Proc Natl Acad Sci U S A 289–300. 2015;112:14325–30. 31. Gu Z, Eils R, Schlesner M. Complex heatmaps reveal patterns and correlations in 12. Balko JM, Giltnane JM, Wang K, Schwarz LJ, Young CD, Cook RS, et al. multidimensional genomic data. Bioinformatics 2016;32:2847–9. Molecular profiling of the residual disease of triple-negative breast cancers after 32. Denkert C, von Minckwitz G, Darb-Esfahani S, Lederer B, Heppner BI, Weber neoadjuvant chemotherapy identifies actionable therapeutic targets. KE, et al. Tumour-infiltrating lymphocytes and prognosis in different subtypes of Cancer Discov 2014;4:232–45. breast cancer: a pooled analysis of 3771 patients treated with neoadjuvant 13. Hendry S, Salgado R, Gevaert T, Russell PA, John T, Thapa B, et al. Assessing therapy. Lancet Oncol 2018;19:40–50. tumor-infiltrating lymphocytes in solid tumors: a practical review for pathol- 33. Rosshart S, Hofmann M, Schweier O, Pfaff A-K, Yoshimoto K, Takeuchi T, et al. ogists and proposal for a standardized method from the International Immuno- Interaction of KLRG1 with E-cadherin: new functional and structural insights. Oncology Biomarkers Working Group: part 2: TILs in melanoma, gastrointes- Eur J Immunol 2008;38:3354–64. tinal tract carcinomas, non-small cell lung carcinoma and mesothelioma, 34. Gros A, Parkhurst MR, Tran E, Pasetto A, Robbins PF, Ilyas S, et al. Prospective endometrial and ovarian carcinomas, squamous cell carcinoma of the head and identification of neoantigen-specific lymphocytes in the peripheral blood of neck, genitourinary carcinomas, and primary brain tumors. Adv Anat Pathol melanoma patients. Nat Med 2016;22:433–8. 2017;24:311–35. 35. Johnson DB, Nixon MJ, Wang Yu, Wang DY, Castellanos E, Estrada MV, et al. 14. Hendry S, Salgado R, Gevaert T, Russell PA, John T, Thapa B, et al. Assessing Tumor-specific MHC-II expression drives a unique pattern of resistance to tumor-infiltrating lymphocytes in solid tumors: a practical review for pathol- immunotherapy via LAG-3/FCRL6 engagement. JCI Insight 2018;3:e120360. ogists and proposal for a standardized method from the International Immu- 36. Han M, Salamat A, Zhu Li, Zhang H, Clark BZ, Dabbs DJ, et al. Metaplastic breast nooncology Biomarkers Working Group: part 1: assessing the host immune carcinoma: a clinical-pathologic study of 97 cases with subset analysis of response, TILs in invasive breast carcinoma and ductal carcinoma in situ, response to neoadjuvant chemotherapy. Mod Pathol 2019;32:807–16. metastatic tumor deposits and areas for further research. Adv Anat Pathol 37. Kaminsky A, Bhargava R, McGuire KP, Puhalla S. Single institution experience 2017;24:235–51. with neoadjuvant chemotherapy for metaplastic breast cancer (MBC). J Clin 15. Salgado R, Denkert C, Demaria S, Sirtaine N, Klauschen F, Pruneri G, et al. The Oncol 30: 15s, 2012 (suppl; abstr 1038). evaluation of tumor-infiltrating lymphocytes (TILs) in breast cancer: recom- 38. Bae SY, Lee SeK, Koo MY, Hur SMo, Choi M-Y, Cho DH, et al. The prognoses of mendations by an International TILs Working Group 2014. Ann Oncol 2015;26: metaplastic breast cancer patients compared to those of triple-negative breast 259–71. cancer patients. Breast Cancer Res Treat 2011;126:471–8. 16.MaC,FanR,AhmadH,ShiQ,Comin-AnduixB,ChodonT,etal.A 39. Aydiner A, Sen F, Tambas M, Ciftci R, Eralp Y, Saip P, et al. Metaplastic breast clinical microchip for evaluation of single immune cells reveals high carcinoma versus triple-negative breast cancer: survival and response to treat- functional heterogeneity in phenotypically similar T cells. Nat Med ment. Medicine 2015;94:e2341. 2011;17:738–43. 40. Al-Hilli Z, Choong G, Keeney MG, Visscher DW, Ingle JN, Goetz MP, et al. 17. Lu Y, Xue Q, Eisele MR, Sulistijo ES, Brower K, Han L, et al. Highly Metaplastic breast cancer has a poor response to neoadjuvant systemic therapy. multiplexed profiling of single-cell effector functions reveals deep functional Breast Cancer Res Treat 2019;176:709–16. heterogeneity in response to pathogenic ligands. Proc Natl Acad Sci U S A 41. Ledderose C, Heyn J, Limbeck E, Kreth S. Selection of reliable reference genes for 2015;112:E607–15. quantitative real-time PCR in human T cells and neutrophils. BMC Res Notes 18. Ma C, Cheung AF, Chodon T, Koya RC, Wu Z, Ng C, et al. Multifunctional T-cell 2011;4:427. analyses to study response and progression in adoptive cell transfer immuno- 42. Dumont C, Jacquier A, Verine J, Noel F, Goujon A, Wu C-L, et al. CD8(þ) therapy. Cancer Discov 2013;3:418–29. PD-1(-)ILT2(þ) T cells are an intratumoral cytotoxic population selectively 19. Rossi J, Paczkowski P, Shen Y-W, Morse K, Flynn B, Kaiser A, et al. Preinfusion inhibited by the immune-checkpoint HLA-G. Cancer Immunol Res 2019;7: polyfunctional anti-CD19 chimeric antigen receptor T cells are associated with 1619–32. clinical outcomes in NHL. Blood 2018;132:804–14. 43. Carosella ED, Rouas-Freiss N, Tronik-Le Roux D, Moreau P, LeMaoult J. 20. Xue Q, Bettini E, Paczkowski P, Ng C, Kaiser A, McConnell T, et al. Single-cell HLA-G: an immune checkpoint molecule. Adv Immunol 2015;127: multiplexed cytokine profiling of CD19 CAR-T cells reveals a diverse landscape 33–144. of polyfunctional antigen-specific response. J Immunother Cancer 2017;5:85. 44. Lin A, Yan WH. Heterogeneity of HLA-G expression in cancers: facing the 21. Parisi G, Saco JD, Salazar FB, Tsoi J, Krystofinski P, Puig-Saus C, et al. Persistence challenges. Front Immunol 2018;9:2164. of adoptively transferred T cells with a kinetically engineered IL-2 receptor 45. Foulds GA, Vadakekolathu J, Abdel-Fatah TMA, Nagarajan D, Reeder S, agonist. Nat Commun 2020;11:660. Johnson C, et al. Immune-phenotyping and transcriptomic profiling of periph- 22. Korsunsky I, Millard N, Fan J, Slowikowski K, Zhang F, Wei K, et al. Fast, eral blood mononuclear cells from patients with breast cancer: identification of a sensitive, and accurate integration of single cell data with Harmony. 3 gene signature which predicts relapse of triple negative breast cancer. Nat Methods 2019;16:1289–96. Front Immunol 2018;9:2028.

OF14 Clin Cancer Res; 2020 CLINICAL CANCER RESEARCH

Changes in Peripheral and Local Tumor Immunity after Neoadjuvant Chemotherapy Reshape Clinical Outcomes in Patients with Breast Cancer

Margaret L. Axelrod, Mellissa J. Nixon, Paula I. Gonzalez-Ericsson, et al.

Clin Cancer Res Published OnlineFirst August 21, 2020.

Updated version Access the most recent version of this article at: doi:10.1158/1078-0432.CCR-19-3685

Supplementary Access the most recent supplemental material at: Material http://clincancerres.aacrjournals.org/content/suppl/2020/08/21/1078-0432.CCR-19-3685.DC1

E-mail alerts Sign up to receive free email-alerts related to this article or journal.

Reprints and To order reprints of this article or to subscribe to the journal, contact the AACR Publications Subscriptions Department at [email protected].

Permissions To request permission to re-use all or part of this article, use this link http://clincancerres.aacrjournals.org/content/early/2020/09/23/1078-0432.CCR-19-3685. Click on "Request Permissions" which will take you to the Copyright Clearance Center's (CCC) Rightslink site.