Abstract

Tumor-infiltrating PD-1hi dysfunctional CD8+ T cells have been identified in several tumors but largely unexplored in breast cancer (BC). Here we aimed to extensively explore PD-1hiCD8+ T cells in BC, focusing on the triple-negative BC (TNBC) subtype. Flow cytometry was used to study the phenotypes and functions of CD8+ T-cell subsets in peripheral blood and surgical specimens from treatment-naive BC patients. RNA-seq expression data generated to dissect the molecular features of tumoral PD-1neg, PD-1lo and PD-1hi CD8+ T cells. Further, the associations between tumoral PD-1hi CD8+ T cells and the clinicopathological features of 503 BC patients were explored. Finally, multiplexed immunohistochemistry (mIHC) was performed to evaluate in situ PD-1hiCD8+ T cells on the tissue microarrays (TMAs, n=328) for prognostic assessment and stratification of TNBC patients. PD-1hiCD8+ T cells found readily detectable in tumor tissues but rarely in peripheral blood. These cells shared the phenotypic and molecular features with exhausted and tissue-resident memory T cells (TRM) with a skewed TCR repertoire involvement. Interestingly, PD-1hiCD8+ T cells are in the state of exhaustion characterized by higher T-BET and reduced EOMES expression. PD-1hiCD8+ T cells found preferentially enriched within solid tumors, but predominant stromal infiltration of PD-1hiCD8+ T subset was associated with improved survival in TNBC patients. Taken together, tumoral PD-1hiCD8+ T-cell subpopulation in BC is partially exhausted, and their abundance signifies ‘hot’ immune status with favorable outcomes. Reinvigorating this population may provide further therapeutic opportunities in TNBC patients.

Introduction

Breast cancer (BC) is the most commonly occurring cancer in women worldwide, estimated over 2 million new cases in 2018 [1]. Classically, BC can be categorized into several subtypes based on the expression of estrogen receptor (ER), progesterone receptor (PR) and human epidermal growth factor receptor 2 (HER2), among which triple-negative BC (TNBC) is defined by the negative for ER and PR and a lack of HER2 overexpression. TNBC comprises 10–20% of all BC and represents the most vicious disease due to lack of appropriate therapeutic targets [2]. Despite initial theories suggesting that BC is not an immunogenic disease, now ample evidence confirms and consolidates the idea that the immune system plays a critical role in checking BC development [3]. Interestingly, several studies have uncovered a close link between the presence of tumor-infiltrating lymphocytes (TILs) [4] or CD8+ T cells [5] and favorable clinical outcomes in BC (particularly in TNBC), which offer therapeutic opportunity to potentiate immune response to treat this notorious disease.

Immunotherapy, by targeting the programmed cell death protein-1 (PD-1) or cytotoxic T lymphocyte antigen 4 (CTLA4), has made substantial achievements in the treatment of melanoma, non-small-cell lung cancer (NSCLC) and prostate cancer [6–9]. However, the overall response rate was merely 18.5% for pembrolizumab (anti-PD-1) in treating advanced TNBC patients in one phase Ib clinical study [10]. Polk et al. [11] reviewed the current status of checkpoint inhibitors in BC and found that the overall response rate of PD-1/PD-L1 monotherapy varied from 5 to 30% in heavily pretreated TNBC, indicating an urgent need to dissect the underlying mechanisms of immune responses and fill in thesesubstantial knowledge gaps.

Tumor-infiltrating cytotoxic CD8+ T cells could potently suppress tumor growth, however, CD8+ T cells can turn to a state called ‘exhaustion’ or ‘dysfunction’ within tumor tissues. Exhausted T cells are characterized by high expression of inhibitory receptors such as PD-1, CTLA4 and mucin domain-containing protein 3 (TIM3), poor effector function and distinct epigenetic and transcriptional signature compared with effector or memory T cells [12,13]. PD-1hi exhausted CD8+ T cells have been reported in several human tumors, including NSCLC [14], hepatocellular carcinoma (HCC) [15], renal cell carcinoma (RCC) [16], glioblastoma [17] and head and neck cancer patients [18]. Interestingly, the prognostic values of PD-1hiCD8+ T cells seem to vary across tumors. For instance, higher frequencies of PD-1hiCD8+ T cells correlated with poor clinical outcome in RCC [16] and head and neck cancer patients [18]. By contrast, PD-1hiCD8+ T cells were strongly predictive for both response and survival in a small cohort of NSCLC patients treated with PD-1 blockade [14]. The above studies highlight a complex scenario of PD-1hiCD8+ T cells within human tumors.

In the present study, we aim to perform a comprehensive study of PD-1hiCD8+ T cells in BC at molecular, cellular, functional and clinical levels, with a purpose to define the features of this complex subset and provide critical clues to potentially target these cells in clinics, particularly for TNBC.

Materials and methods

Patients’ samples

All studies were conducted with approval from the Fudan University Shanghai Cancer Center Institutional Review Board in accordance with recognized ethical guidelines (1404134-8-1611A). Total 503 treatment-naïve BC patients undergoing primary surgical resection of tumor from July 2015 to April 2017 included in the prospective collection of tumor tissue. Besides, matched peripheral blood also collected from 12 cases. Informed consent was obtained from all subjects. Blood specimens were collected into ethylene diamine tetraacetic acid (EDTA)-containing tubes. Peripheral blood mononuclear cells (PBMCs) were freshly isolated from venous blood over LymphoPrep™ (StemCell Technologies) gradient centrifugation. Tumor specimens were prepared as follows: we minced tumor tissues into small pieces followed by digestion in RPMI-1640 (GIBCO Inc.) cocktail supplemented with hyaluronidase (0.5 mg/ml), collagenase type-IV (1 mg/ml), DNase I solution (0.01 mg/ml) (Sigma–Aldrich). For the preparation of single cell suspension, tissues were digested with intermittent shaking (1200 rpm for 1 h at 37°C) and finally passed through a 70-μm filter (BD Biosciences) to remove clots.

Flow cytometry and cell sorting

Human-specific flow cytometry antibodies were purchased from the following providers, and used according to the manufacturer’s recommended concentrations, routinely based on a test (for instance, a given antibody, 100 tests, 5 µl/test, we took 5 μl to prepare the staining cocktail). BD Biosciences (CD3: SK7; CD8: HIT8a; CD45: HI30; CD69: FN50; HLADR: G46-6; Perforin: dG9; Granzyme B: GB11; 2B4: 2-69; interferon (IFN)-γ: 4S.B3), BioLegend (CD27: O323; CD38: HIT2; CD57: HNK-1; CD103: Ber-ACT8; CD127: A019D5; CD160: BY55; BTLA: MIH26; CD45RA: HI100; CCR7: G043H7; PD-1: E12.2H7; PDL1: 29E.2A3; TCRαβ: IP26; Tim3: F38-2E2; Granzyme A: CB9; Granulysin: DH2; KLRG1: 14C2A07; Bcl-2; Ki67; tumor necrosis factor (TNF)-α: MAb17), eBioscience (CD4: SK3; ICOS: ISA-3; CTLA4: 14D3; T-bet: eBio4B10; Eomes: WD1982; LAG3: 3DS223H; TIGIT: MBSMA23; interleukin-2 (IL-2): MQ1-17H12) or Immuno-tools (Granzyme K). For surface staining, the antibodies were prepared and stained in an FACS buffer (phosphate buffer saline (PBS) supplemented with 2% fetal bovine serum (FBS) and 2.5 mM EDTA). For intracellular staining, the antibodies were prepared in a Permeabilization Buffer from a Fixation and Permeabilization Buffer kit (eBioscience). Appropriate isotype controls were used where applicable. The LIVE/DEAD Zombie Yellow™ Fixable Viability Kit (Biolegend) was used to exclude dead cells, as indicated by the provider. For intracellular cytokine staining, cells were stimulated with phorbol 12-myristate 13-acetate (PMA, 50 ng/ml, Sigma) and ionomycin (1 μg/ml, Sigma) for 5 h at 37°C, 5% CO2, in the presence of brefeldin A (1/1000, Biolegend). Surface staining was performed and cells were fixed and permeabilized with BD Cytofix/Cytoperm Kit and stained for IFN-γ, IL-2 and TNF-α. Cells were washed, fixed and analyzed on an LSRII FORTESSA flow cytometer (BD Bioscience) and data analyzed in FlowJo software (v 9.3.2).

Fresh TILs were labeled, and populations of interest were purified after reaching cell sorting purity to 99% on a MoFlo Astrios (Beckman Coulter). Briefly, CD8 subsets from TILs were sorted as CD3+CD8+PD-1(PD-1neg), CD3+CD8+PD-1lo(PD-1lo) and CD3+CD8+PD-1hi (PD-1hi).

RNA-Seq data acquisition and analysis

For RNA-sequencing, RNA was extracted from 10000 sorted CD8+ T-cell subsets using an miRNeasy Micro Kit (Qiagen), and the quality was assessed with an Agilent RNA 6000 Pico Kit (Agilent, cat. no. 5067-1513) on an Agilent 2100 Bioanalyzer. Each library was prepared from 2 ng of total RNA. After reverse transcription, cDNA was amplified for eight cycles followed by Agencourt AMPure XP purification. Then, cDNA was sheared to a 200–500 bp size range using the Covaris AFA system. The library was generated by using the NEB Next ULTRA II DNA Prep Kit, and the quality and size were checked using the Agilent High Sensitivity DNA Kit. The libraries were sequenced on a HiSeq 4000 system (Illumina) using PE150 mode. At least 10 million read pairs were obtained for each sample.

The principal component analysis (PCA) for all the genes was generated using the DESeq2 version of plotPCA. Differentially expressed genes were identified with the cutoff threshold of adjusted P-value less than 0.01 and at least one fold change between the PD-1hi and the PD-1neg groups by using the R package DEseq2 [19]. For gene set enrichment analysis (GSEA), the expression levels of all genes were put into GSEA package [20] to find enriched pathways with the normalized enrichment score > 1 and q-value <0.05. The tissue-resident memory T cell (TRM) up and down signature gene sets were constructed from the union set of up and down genes in three different studies [21–23], respectively. The expression of T-cell repertoire genes in CD8+ T cells was calculated using the TRUST algorithm [24]. The usage for each TCR gene was calculated as the expression of a V gene in each sample divided by the total expression of all V genes in the same sample.

TIL evaluation

TIL evaluation was performed on Hematoxylin and Eosin-stained (H&E) sections. The percentage of stromal TILs for each patient was evaluated by two pathologists according to the instructions of an international TILs working group 2014 and the threshold of stromal lymphocytes for lymphocyte-predominant (LP) BC is 60% of the stromal surface area [25].

Patient cohort and tissue microarray construction

For the present study, paraffin-embedded pretherapeutic core biopsies from 328 TNBC patients, who underwent surgical resection in Shanghai Cancer Center, Fudan University from 2008 to 2014, were used. The median follow-up time for these patients was 61.5 months. Tissue microarrays (TMAs) comprised representative areas of the tumor which were selected based on the review of H&E staining. Two 1.0-mm tissue cores collected from respective formaldehyde-fixed and paraffin-embedded (FFPE) blocks in order to capture the contextual heterogeneity.

Multiplexed immunohistochemistry and quantitative pathology

We aimed to visualize and quantitate exhausted signature in CD8+ T cells with PD-1 assessment, particularly PD-1hi estimation in tissue cores from TMAs by quantitative immunofluorescence (QIF) assay. Briefly, 4–5 μm FFPE tissue containing TMA slides were pre-warmed at 58°C prior to deparaffinization and followed by dewaxing in three changes of xylene and two changes of 100% ethanol and subsequent gradation of 95, 80 and 70% alcohol for 3 min each. TMAs were briefly fixed in 10% neutral buffered formalin (NBF) before heat-induced epitope retrieval (HIER). Antigen retrieval was performed by steaming the slides with a preheated AR6 buffer (pH 6.0, PerkinElmer, Inc., U.S.A.) at 98°C for 20 min in a microwave (Sanyo, Japan). In tyramide signal amplification (TSA)-based multiplexing, a subsequent round of antigen retrieval steps is followed, and here AR6 buffer is used after each round of chromogenic detection to remove unbound Abs–dye complex of the previous staining. After retrieval, sections were allowed to cool and washed in 0.03% Tris-buffered saline–Tween-20 (TBST, Amresco) three times at 5 min each with gentle agitation. Slides were blocked for 7 min with freshly made 3% H2O2 following another rinsing step and application of protein block (ADI-950-230, Enzo Life Sciences, Inc., U.S.A.) for 10 min. Sections were then incubated using mAbs directed against CD8 (C8/144B, DAKO, 1/2000, incubation for 45 min), PD-1 (NAT105, Abcam, 1/2000, incubation for 60 min) and Cytokeratin Pan (AE1/AE3, GeneTex, 1/2000, 30-min incubation) followed by corresponding isotype matched polymer based horseradish peroxidase–conjugated secondary Abs (Vector Laboratories, CA). After washing, TSA step was performed to mediate the horseradish peroxidase (HRP)-covalent binding of different fluorophore (Opal-520, Opal-570 and Opal-690) sequentially, as specified by the manufacturer (NEL810001KT, PerkinElmer Inc). After all sequential reactions, tissues were counterstained with DAPI (Sigma–Aldrich), mounted in Vectashield hardset® fluorescence mounting medium (Vector Labs, Burlingame, CA) and left to dry flat for 30 min in the absence of light and imaged in Vectra 3 digital imaging platform.

QIF assessment was performed using inForm (v2.3, PerkinElmer Inc.) based single-cell mean pixel fluorescence intensity aided with an active machine learning algorithm. For accuracy over visual assessment, especially for PD-1 ‘hi’ threshold cutoff determination, each target was analyzed in FCS Express 6 Plus v6.04.0034 (De Novo Software) and displayed in contour plot with multiple back gating in DAPI+ nucleated cells to remove noise. Scoring of PD-1hi T cells was done with HALO-AI based algorithm (v2.0, Indica Labs, NM), and cytokeratin positivity accounted for distinguishing tumors from the stroma. A dual histospot score was averaged for each patient to achieve better heterogeneity.

Multispectral image acquisition, spectral unmixing and analysis

Multiplex-stained TMA slides were loaded into an integrated robotics tray (Prior) of Vectra3.0 Multispectral Imaging System (PerkinElmer, U.S.A.) and image acquired with Vectra software after creating user-defined protocol with each channel exposure adjustment. Regions of interest (ROIs, 4× magnification) were considered for high power (20×) multispectral imaging format (.im3). Each raw image comprising four stitched 200× multispectral image cubes was obtained for each TMA core. Each 200× multispectral image cube was created by combining images obtained every 10 nm of the emission light spectrum across the range of 420–720 nm through corresponding filter cube. We build a single channel compensation library using the single control slides by inForm (v 2.2.0, PerkinElmer).

Spectral unmixing was undertaken using the inForm Advanced Image Analysis Software (v2.3 PerkinElmer) with earlier built spectral libraries from single stained slides along with unstained tissue auto-fluorescence references and data were exported in multi-image TIFF integrated with x,y coordinates for each pixel. The resulted images were analyzed using HALO, digital image analysis software (v2.0, Indica Labs, NM). Randomly selected 10–15% mixed phenotype cores from each slide were chosen for active learning algorithm buildup, which consists of DAPI-positive nuclear segmentation, cytoplasm/membrane mapping, tissue training, and intensity-based threshold positivity. Tumor (cytokeratin+) and stromal (cytokeratin−) areas within each core were defined using software, and associated phenotype was obtained. The positivity cutoff based on visual assessment of immune cells at different staining intensities was verified in FCS Express 6 Plus v6.04.0034 (De Novo Software) to set algorithms for scoring. In particular, PD-1hi/lo was defined based on mean pixel intensity (per cell) from a cumulative 105 CD8+ T-cell data exported in .csv file enumerated by inForm v2.3 (PerkinElmer) based machine learning algorithm; normalized to control. Further, threshold values defined by plotting FACS alike-dot plot fashion using FCS Express (De Novo Software) and linear cut-off values implemented to all samples. Component TIFF images were put in batch analysis using HALO for each subset frequency enumeration. Acquired score data are merged, exported in MS Excel, and processed further in SPSS (IBM, U.S.A.) for the clinicopathological outcome.

Variable definition

Disease-free survival (DFS) was defined as the time from the date of surgery to the date of the first relapse, second primary malignancyor death resulting from any cause. Overall survival (OS) was defined as the time elapsed from the date of surgery to death by any cause or the date of last follow-up. The cut-off point of prognostic variables was confirmed using the package of ‘X-tile’ [26], which can best discriminate patient survival. It was 12.4, 0.6 and 7.6%, respectively, for the percentage of CD8+ cells in total stromal cells, PD-1hi CD8 cells in total stromal cells and PD-1hi cells among CD8+ T cells.

Statistics

The Mann–Whitney U test or Student’s t test was used to compare two groups as appropriate; for more than two groups, one-way analysis of variance with post hoc Tukey’s multiple comparison test was used (GraphPad Prism V.6.0f).

Kaplan–Meier survival curves were compared using the log-rank test. Univariate and multivariate analyses were performed using Cox proportional hazards regression model. Statistical Package for the Social Sciences software (version 17.0, SPSS Inc, Chicago, IL, U.S.A.) was used to analyze all the survival data.

Results

PD-1hiCD8+ T cells display mixed exhausted and resident memory T cell gene signatures

PD-1hiCD8+ T cells have been found in several human tumor tissues, but their clinical significance shows high variability [14–18]. To investigate PD-1hiCD8+ T cells in BC, we first checked the PD-1 expression on CD8+ T cells from peripheral blood and paired BC tumor tissues. The results showed that while peripheral CD8+ T cells were positive for PD-1, a distinct PD-1hi subset was almost exclusively observed in BC tumor tissues (Figure 1A,B).

Molecular characterization of tumoral PD-1hiCD8+ T cells from BC patients

Figure 1
Molecular characterization of tumoral PD-1hiCD8+ T cells from BC patients

(A) Representative flow cytometric analysis to show the PD-1 expression in CD8+ T cells from one matched peripheral blood and tumor specimen. (B) Comparison of the frequency of PD-1hiCD8+ T cells from 12 matched blood and BC tumor tissues. (C) PCA to show the global gene expression difference among PD-1hi, PD-1lo and PD-1neg CD8+ T cells from BC tumors. Dots indicate samples of the three different populations from a total of five patients. Each symbol represents one patient. (D) Unsupervised clustering of the 954 differentially expressed genes between PD-1hi and PD-1neg CD8+ T cells with an FDR < 1% among all three subsets. Blue color indicates down-regulated genes, and orange indicates up-regulated genes. (E) GSEA identified that ‘Exhausted signature (up gene)’, ‘enhancer of zeste homolog 2 (EZH2) targets’ and ‘TRM signatures (both up and down genes)’ were enriched in PD-1hi CD8 T cells compared with PD-1lo CD8+ T cells. (F,G) Examples of differentially expressed genes between PD-1hi CD8 and PD-1neg CD8+ T cells present in the ‘exhausted signature” (green) (F) or the ‘TRM’ (blue) (G) gene sets. Each symbol/line represents one patient.***P<0.001.

Figure 1
Molecular characterization of tumoral PD-1hiCD8+ T cells from BC patients

(A) Representative flow cytometric analysis to show the PD-1 expression in CD8+ T cells from one matched peripheral blood and tumor specimen. (B) Comparison of the frequency of PD-1hiCD8+ T cells from 12 matched blood and BC tumor tissues. (C) PCA to show the global gene expression difference among PD-1hi, PD-1lo and PD-1neg CD8+ T cells from BC tumors. Dots indicate samples of the three different populations from a total of five patients. Each symbol represents one patient. (D) Unsupervised clustering of the 954 differentially expressed genes between PD-1hi and PD-1neg CD8+ T cells with an FDR < 1% among all three subsets. Blue color indicates down-regulated genes, and orange indicates up-regulated genes. (E) GSEA identified that ‘Exhausted signature (up gene)’, ‘enhancer of zeste homolog 2 (EZH2) targets’ and ‘TRM signatures (both up and down genes)’ were enriched in PD-1hi CD8 T cells compared with PD-1lo CD8+ T cells. (F,G) Examples of differentially expressed genes between PD-1hi CD8 and PD-1neg CD8+ T cells present in the ‘exhausted signature” (green) (F) or the ‘TRM’ (blue) (G) gene sets. Each symbol/line represents one patient.***P<0.001.

To gain more insight into this PD-1hi CD8 T-cell population, we sorted PD-1neg, PD-1lo and PD-1hi CD8+ T cell subsets from five BC tumors to perform an RNA-seq-based gene expression profile analysis (Supplementary Figure S1A and Table S1). PCA and unsupervised hierarchical clustering revealed that the gene signature of PD-1hi was far different from PD-1lo and PD-1neg CD8+ T cells, with the latter two highly similar (Figure 1C,D). GSEA identified that PD-1hiCD8+ T cells were enriched with exhausted T cell and tissue resident memory T cell (TRM) gene signatures (Figure 1E and Supplementary Figure S2). Gene transcripts associated with T-cell exhaustion such as HAVCR2 (TIM-3), ENTPD1 (CD39) related genes were significantly up-regulated in PD-1hi as compared with PD-1lo and PD-1neg CD8+ TILs (Figure 1F). PD-1hi CD8+ T cells also demonstrated a TRM gene signature with higher expression of ITGAE (CD103) and lower expressions of genes KLF2, S1PR1, SELL (CD62L) and TCF7 compared with PD-1lo and PD-1neg CD8+ TILs (Figure 1G). Besides, PD-1hi CD8+ T cells also showed higher expression of enhancer of zeste homolog 2 (EZH2) target genes (Figure 1E and Supplementary Figure S2), implicating a modification at epigenetic level. Collectively, these results demonstrate that PD-1hiCD8+ T cells are highly enriched within BC tumor tissues and transcriptionally different from PD-1lo and PD-1neg CD8+ T cells with exhausted and TRM gene signatures.

PD-1hiCD8+ T cells are phenotypically and functionally dysfunctional

To validate the above gene expression data, we analyzed several representative molecules at the protein level by flow cytometry (Figure 2A, Supplementary Figures S1B and S3). PD-1hiCD8+ T cells expressed higher levels of exhausted molecules TIM3, CLTA4, TIGIT, LAG3 and CD39 when compared with PD-1lo or PD-1neg CD8 T cells. PD-1hiCD8+ T cells also highly expressed activation markers CD38, CD69, HLADR and ICOS. For differentiation markers, PD-1hi CD8+ T cells displayed lower expression of CCR7, CD127 and CD45RA, indicative of an effector memory phenotype [27,28]. The expression of KLRG1 and CD57 was significantly lower on PD-1hi CD8+ T cells compared with PD-1lo and PD-1neg CD8+ T cells. KLRG1 in mice and CD57 in humans have been used to identify terminally differentiated or senescent T cells [29,30], thus, PD-1hiCD8+ T cells seem not to be terminally differentiated. The PD-1hiCD8+ T cells also had increased frequencies of Ki-67, suggesting that these cells might recently encounter with the cognate antigen and were in a state of proliferation. In addition, TRM marker CD103 [31] was highly expressed in the PD-1hiCD8+ T-cell population, and CD103+PD-1hiCD8+ T cells expressed higher exhausted markers TIM3, CTLA4 and LAG3 than CD103PD-1hiCD8+ T cells (Supplementary Figure S4). These results indicate that within PD-1hiCD8+ T cells, exhausted T cells and TRM are not two distinct subsets; instead, the features of exhaustion and tissue residency seems to be co-acquired during differentiation of PD-1hiCD8+ T cells.

Phenotypic and functional properties of PD-1hi, PD-1lo and PD-1neg CD8+ T cells from BC tumors

Figure 2
Phenotypic and functional properties of PD-1hi, PD-1lo and PD-1neg CD8+ T cells from BC tumors

(A) Representative flow cytometry analysis to show the expression of surface and intracellular markers. One representative out of at least three was shown. MFI ± standard deviation was marked in the figure. (B,C) Representative flow cytometry analysis (B) and accumulated MFI data (n=13) (C) of the expressions of EOMES and T-BET of PD-1hi, PD-1lo and PD-1neg CD8+ T cells from BC tumors. (D) Representative flow cytometric analysis of TNF-α, IL-2 and IFN-γ produced by the three CD8 TIL subsets stimulated for 5 h with PMA/ionomycin. Numbers in each box indicate percent cells positive for TNF-α, IL-2 and IFN-γ for each subset. (E) Accumulated data of TNF-α, IL-2 and IFN-γ-positive cells by each CD8 TIL subsets. Data are from five BC patients. Mean percentage ± standard deviation was marked in the figure. *P<0.05; **P<0.01; ***P<0.001. P-values were determined by one-way analysis of variance with Tukey’s post hoc testing. Abbreviation: MFI, median fluorescence intensity.

Figure 2
Phenotypic and functional properties of PD-1hi, PD-1lo and PD-1neg CD8+ T cells from BC tumors

(A) Representative flow cytometry analysis to show the expression of surface and intracellular markers. One representative out of at least three was shown. MFI ± standard deviation was marked in the figure. (B,C) Representative flow cytometry analysis (B) and accumulated MFI data (n=13) (C) of the expressions of EOMES and T-BET of PD-1hi, PD-1lo and PD-1neg CD8+ T cells from BC tumors. (D) Representative flow cytometric analysis of TNF-α, IL-2 and IFN-γ produced by the three CD8 TIL subsets stimulated for 5 h with PMA/ionomycin. Numbers in each box indicate percent cells positive for TNF-α, IL-2 and IFN-γ for each subset. (E) Accumulated data of TNF-α, IL-2 and IFN-γ-positive cells by each CD8 TIL subsets. Data are from five BC patients. Mean percentage ± standard deviation was marked in the figure. *P<0.05; **P<0.01; ***P<0.001. P-values were determined by one-way analysis of variance with Tukey’s post hoc testing. Abbreviation: MFI, median fluorescence intensity.

A striking finding was the expression pattern for transcription factors T-BET and EOMES in PD-1hi CD8+ T cells. It was reported that T-cell exhaustion during chronic viral infection was associated with the loss of PD-1loTbethiEomeslo T cells, and the accumulation of terminally differentiated PD-1hiTbetloEomeshi exhausted T cells [32]. However, we found that PD-1hiCD8+ T cells in BC tumors expressed the highest levels of T-BET and a somewhat reduced EOMES compared with the other two CD8+ T-cell subsets (Figure 2B,C). This suggests that PD-1hi CD8 T cells from BC tumors might be at an early stage of exhaustion or partially exhausted.

Next, the effector function of the CD8+ T-cell subsets was assessed by analyzing the capacity of cytokine production and granzyme expression. Following stimulation with PMA and ionomycin, intracellular IFN-γ, IL2 and TNF-α were assessed by flow cytometry. Though the sample size was small, obvious differences could be observed. We found that significantly fewer IL2 and TNF-α positive cells were observed in PD-1hiCD8+ T cells compared with the other two T cell subsets, while the defect of IFN-γ production by PD-1hiCD8+ T cells was not dramatic (Figure 2D,E). There was almost no difference for the expression of perforin, granzyme A, granzyme K and granulysin among the CD8+ T-cell subsets (Supplementary Figure S3). Of note, the PD-1hiCD8+ T cells seemed to have an increased cytotoxic potential, as demonstrated by an increased expression of granzyme B (Figure 2A). Thus, despite that PD-1hiCD8+ T cells from BC tumors expressed certain exhausted markers, these cells were not terminally exhausted and retained specific cytotoxic potentials.

PD-1hiCD8+ T cells have a different TCR repertoire usage

Mapped reads from RNA-seq were used to calculate the usage of different TCRα variable (TRAV) and TCRβ variable (TRBV) genes across the three CD8+ T subsets from five BC tumors. Interestingly, PD-1hiCD8+ T cells exhibited detectable repertoire skewing including increased usages of TRAV19 (logFC = 2.05, P=0.027), TRAV29/DV5 (logFC = 1.835, P=0.029) and TRBV30 (logFC = 2.73, P=0.02), while the usage of TRBV18 (logFC = −2.04, P=0.047) was significantly reduced (Figure 3A,B). The repertoire skewing of TRAV and TRBV gene segments observed here may suggest that certain PD-1hiCD8+ T cells were activated with cognate antigens and undergone expansion, which is consistent with an increased frequency of Ki-67+ cells observed before.

TCR repertoire analysis of CD8+ T cell subsets from BC tumor tissues

Figure 3
TCR repertoire analysis of CD8+ T cell subsets from BC tumor tissues

TCR Vα (A) and TCR Vβ (B) analyses were performed on the PD-1hi, PD-1lo and PD-1neg CD8+ T cells isolated from human BC TILs (n=5). Significance was assessed between PD-1hi and PD-1neg CD8+ T cells using Student’s t test. *P<0.05.

Figure 3
TCR repertoire analysis of CD8+ T cell subsets from BC tumor tissues

TCR Vα (A) and TCR Vβ (B) analyses were performed on the PD-1hi, PD-1lo and PD-1neg CD8+ T cells isolated from human BC TILs (n=5). Significance was assessed between PD-1hi and PD-1neg CD8+ T cells using Student’s t test. *P<0.05.

PD-1hiCD8+ T cells are enriched in the tumor tissues of TNBC

TILs abundance has been widely evaluated in previous studies and was correlated with higher pathological complete response (pCR) rate and better patient survival in BC [33–35]. Next, we sought to explore the relationship between PD-1hiCD8+ T-cell percentage and TIL infiltration. The result showed that 84 among 503 samples were LPBC and the percentage of PD-1hiCD8+ T cells was significantly higher in the LPBC group than that in the non-LPBC group (P<0.0001, Figure 4A,B). When compared with different pathological characteristics, the percentage of PD-1hiCD8+ T cell was significantly higher in the patients with high histological grade and elevated Ki-67 proliferative index (Figure 4C,D). Furthermore, the percentage of PD-1hiCD8+ T cells in TNBC was significantly higher than other molecular subtypes except the Her2+ group, for which a tendency of increase was also observed (Figure 4E). Clinicopathologic features of the 503 patients were detailed in Supplementary Table S2. Collectively, by using a relatively large number of tumor samples, we observed that high infiltration of PD-1hiCD8+ T cells is linked with LPBC, high histological grade, highly proliferative status and the TNBC subtype.

Association between PD-1hiCD8+ T-cell percentage and clinicopathological features of BC patients

Figure 4
Association between PD-1hiCD8+ T-cell percentage and clinicopathological features of BC patients

(A) Stromal TILs were evaluated by HE stain in BC and divided into two groups (LPBC and non-LPBC). Representative images of non-LPBC (left) and LPBC (right) were shown. Upper photomicrographs, 100× magnification; lower photomicrographs, 200× magnification. (BE) Comparison of tumoral PD-1hiCD8+ T-cell percentage among total CD8+ T cells measured by flow cytometry in BC patients with different TIL infiltration (B), histological grade (C), proliferative index Ki67 (D) and molecular subtypes (E). Histological grades I and II are defined as ‘Low’ and histological grade III is defined as ‘High’, respectively. The proliferative index Ki67 more than 20% was defined high Ki67 and the proliferative index Ki67 less than 20% was defined low Ki67. *P<0.05; **P<0.01; ***P<0.001; ****P<0.0001. t test was used for (B–D) and one-way analysis of variance with Tukey’s post hoc testing was used for (E).

Figure 4
Association between PD-1hiCD8+ T-cell percentage and clinicopathological features of BC patients

(A) Stromal TILs were evaluated by HE stain in BC and divided into two groups (LPBC and non-LPBC). Representative images of non-LPBC (left) and LPBC (right) were shown. Upper photomicrographs, 100× magnification; lower photomicrographs, 200× magnification. (BE) Comparison of tumoral PD-1hiCD8+ T-cell percentage among total CD8+ T cells measured by flow cytometry in BC patients with different TIL infiltration (B), histological grade (C), proliferative index Ki67 (D) and molecular subtypes (E). Histological grades I and II are defined as ‘Low’ and histological grade III is defined as ‘High’, respectively. The proliferative index Ki67 more than 20% was defined high Ki67 and the proliferative index Ki67 less than 20% was defined low Ki67. *P<0.05; **P<0.01; ***P<0.001; ****P<0.0001. t test was used for (B–D) and one-way analysis of variance with Tukey’s post hoc testing was used for (E).

PD-1hiCD8+ T cells serve as a favorable prognostic biomarker in TNBC

Since the PD-1hiCD8+ T cells were enriched in TNBC tumor tissues and the TNBC was considered to have the worst prognosis among different subtypes of BC. We explored the possible relationship between the PD-1hiCD8+ T cells infiltration with DFS and OS in a Chinese TNBC cohort (n=328). PD-1hiCD8+ T cells in the stroma were determined within primary tumors by multispectral image acquisition, spectral unmixing and analysis (Figure 5A–D). PD-1 ‘high’ threshold was defined with mean pixel intensity plotting, gated on CD8+ cell population (Figure 5B). The patients have received standard of care treatments and their characteristics were detailed in Supplementary Table S3. Consistent with the flow cytometry data, the multiplexed immunohistochemistry (mIHC) analysis revealed a positive association between histological grade and PD-1hiCD8+ T-cell frequency of TNBC patients.

Prognostic value of PD-1hiCD8+ T cells in TNBC

Figure 5
Prognostic value of PD-1hiCD8+ T cells in TNBC

(A) Representative multicolor image to show the expression of CD8, PD-1 and Pan-CK from one BC tumor (×200). Unmixing image of single markers were shown on the right. (B) Flow cytometry alike density plot is defining PD-1hi-threshold among CD8+ T cells based on quantitative mean pixel fluorescence intensity. (C) Representative phenotype map to show PD-1hiCD8+ T cells and other cells distribution in TME. (D) Representative cores to show high (left) and low (right) infiltrations of PD-1hiCD8+ T cells in TNBC tumors. Kaplan–Meier curves to show 5-year DFS (E,G,I) and OS (F,H,J) for the percentage of CD8+ cells within total stromal cells (E,F), the percentage of PD-1hiCD8+ cells among total stromal cells (G,H) and the percentage of PD-1hi cells among stromal CD8+ cells (I,J) (n=328).

Figure 5
Prognostic value of PD-1hiCD8+ T cells in TNBC

(A) Representative multicolor image to show the expression of CD8, PD-1 and Pan-CK from one BC tumor (×200). Unmixing image of single markers were shown on the right. (B) Flow cytometry alike density plot is defining PD-1hi-threshold among CD8+ T cells based on quantitative mean pixel fluorescence intensity. (C) Representative phenotype map to show PD-1hiCD8+ T cells and other cells distribution in TME. (D) Representative cores to show high (left) and low (right) infiltrations of PD-1hiCD8+ T cells in TNBC tumors. Kaplan–Meier curves to show 5-year DFS (E,G,I) and OS (F,H,J) for the percentage of CD8+ cells within total stromal cells (E,F), the percentage of PD-1hiCD8+ cells among total stromal cells (G,H) and the percentage of PD-1hi cells among stromal CD8+ cells (I,J) (n=328).

Previous study demonstrated that CD8+ T-cell infiltration was positively correlated with patient survival in BC [5] and we also checked the relevance of CD8+ T cells. Patients with higher frequency of CD8+ T cells at the time of surgery had better DFS (5-year survival: 88.4 vs. 73%; P-value <0.0001) (Figure 5E) and OS (5-year survival: 90.9 vs. 79.9%; P-value =0.002) (Figure 5F). Furthermore, we explored the prognostic value of the frequency of PD-1hiCD8+ T cells among total stromal cells and PD-1hi cells within CD8+ T cells. Patients with higher frequency of PD-1hiCD8+ T cells in total stromal cells at the time of surgery had better DFS (5-year survival: 86.1 vs. 74.6%; P-value =0.002) (Figure 5G) and OS (5-year survival: 89.3 vs. 80.9%; P-value =0.007) (Figure 5H). Similarly, the results showed that patients with higher density of PD-1hi cells in CD8+ T cells at the time of surgery had better DFS (5-year survival: 87.4 vs. 74.3%; P-value =0.001) (Figure 5I) and OS (5-year survival: 90.7 vs. 80.4%; P-value =0.001) (Figure 5J).

To account for confounding variables when evaluating the association of PD-1hiCD8+ T-cell frequency with DFS, we took patient age as well as histology type, histological grade, tumor size, lymph node status into consideration for the multivariate analysis. The Cox proportional hazards multivariate regression analysis showed that higher frequency of PD-1hi cells in CD8+ T cells measured by mIHC at the time of surgery was associated with better DFS (hazard ratio [HR] = 0.484; 95% confidence interval (CI) = 0.279–0.840; P-value =0.01) while positive lymph node status at the time of surgery was associated with worse DFS (HR = 2.447; 95% CI = 1.426–4.198; P-value =0.001) (Supplementary Table S4). Similarly, higher frequency of PD-1hi cells within CD8+ T cells measured by mIHC at the time of surgery was associated with better OS (HR = 0.419; 95% CI = 0.221–0.795; P-value =0.008) while positive lymph node status at the time of surgery was associated with worse OS (HR = 2.413; 95% CI = 1.310–4.443; P-value =0.005) (Table 1).

Table 1
Cox proportional hazard model showing hazard ratios for OS conferred by variables in TNBC patients
OS
VariableCategoryUnivariate analysisMultivariate analysis
HR95% CIP-valueHR95% CIP-value
Age (years) ≥50 vs <50 0.630 0.348–1.140 0.127 0.688 0.374–1.266 0.230 
Histology type IDC vs others 1.041 0.143–7.566 0.968 2.023 0.261–15.666 0.500 
Grade III vs I+II 0.785 0.424–1.452 0.440 0.718 0.376–1.371 0.315 
Tumor size ≥2 vs <2 cm 2.158 0.771–6.043 0.143 1.886 0.665–5.352 0.233 
Lymph node Positive/negative 2.686 1.483–4.863 0.001 2.413 1.310–4.443 0.005 
PD-1hi/CD8+High vs low 0.365 0.193–0.688 0.002 0.419 0.221–0.795 0.008 
OS
VariableCategoryUnivariate analysisMultivariate analysis
HR95% CIP-valueHR95% CIP-value
Age (years) ≥50 vs <50 0.630 0.348–1.140 0.127 0.688 0.374–1.266 0.230 
Histology type IDC vs others 1.041 0.143–7.566 0.968 2.023 0.261–15.666 0.500 
Grade III vs I+II 0.785 0.424–1.452 0.440 0.718 0.376–1.371 0.315 
Tumor size ≥2 vs <2 cm 2.158 0.771–6.043 0.143 1.886 0.665–5.352 0.233 
Lymph node Positive/negative 2.686 1.483–4.863 0.001 2.413 1.310–4.443 0.005 
PD-1hi/CD8+High vs low 0.365 0.193–0.688 0.002 0.419 0.221–0.795 0.008 

Abbreviation: IDC, invasive ductal carcinoma.

Collectively, we have shown that the frequency of PD-1hiCD8+ T cells in primary tumors of TNBC correlates with increased survival and could serve as an independent prognostic factor in TNBC.

Discussion

Tumor-infiltrating CD8+ T cells play a pivotal role in exerting profound anti-tumor immunity. However, they often shift to a status of ‘exhaustion’ or ‘dysfunction’, which has posed a hurdle to generate effective anti-tumor immune responses [36,37]. Thus, understanding the functional status of CD8+ T cells in a given tumor is a prerequisite in the development of a novel therapeutic approach to target these cells. In the present study, we have extensively characterized tumoral PD-1hiCD8+ T cells at phenotypic, molecular, functional and clinical levels in BC. We provide the evidence that PD-1hiCD8+ T cells within BC tumors share the features with tissue resident memory T cells and are not terminally exhausted. Particularly, PD-1hiCD8+ T cells are more enriched within TNBC tumor milieu, and their abundance correlates with a significantly lower risk of death in TNBC patients.

A striking feature of PD-1hiCD8+ T cells from BC tumors is their less exhaustion status. T-cell exhaustion has been described during chronic viral infection, where CD8+ T cells gradually acquire exhausted phenotype, from less exhausted TbethiEomeslo to terminally exhausted TbetloEomeshi [32]. Similarly, terminally exhausted TbetloEomeshi T cells have been identified in human HCC [15] and glioblastoma [17], and these PD-1hiCD8+ exhausted T cells also show impaired capacity to produce IL-2, TNF-a and IFN-g. While here, we observed that tumoral PD-1hiCD8+ T cells highly expressed canonical exhausted markers include TIM3, CTLA4, TIGIT and CD39; these cells did not express high levels of EOMES; instead, they expressed significantly higher T-BET than PD-1lo and PD-1neg CD8+ T cells. Furthermore, PD-1hiCD8+ T cells retained IFN-γ producing capacity and even expressed a higher level of cytotoxic Granzyme B molecule than PD-1lo and PD-1neg CD8+ T cells. All of these indicated that PD-1hiCD8+ T cells from BC tumors are at an indeterminate state of exhaustion. As the differentiation of exhausted T cells is controlled by TCR signaling as well as environmental factors [12], we believe that the tumor microenvironment shapes the uniquely differentiated status of PD-1hiCD8+ T cells in BC. The underlying mechanism for this phenomenon warrants further exploration.

Another distinct feature of tumoral PD-1hiCD8+ T cells of BC tumors is their positive association with heavy lymphocyte infiltration and signifies a favorable prognosis in TNBC patients. It is well established that high levels of lymphocytic infiltration are associated with a more favorable prognosis in TNBC and HER2-positive BC patients [3]. Along this line, PD-1hiCD8+ T cells were preferentially enriched in TNBC and, to a less extent, in HER2-positive BC patients. Importantly, we demonstrated that PD-1hiCD8+ T cells predict a better survival for TNBC patients. In melanoma, colorectal, lung and other solid tumors, it has shown that tumor-reactive T cells are enriched within the exhausted T-cell compartment, particularly those T cells expressing the exhausted marker CD39 [38–40]. Even we did not perform adequate tumor-reactive experiments; we did see a TCR repertoire skewing of PD-1hiCD8+ T cells in BC tumors. Taken together, we propose that the differentiation of PD-1hiCD8+ T cells is driven by a persistent immune response involving tumor antigens. In addition, the abundance of PD-1hiCD8+ T cells reflects a ‘hot’ immune status, which constitutes a novel prognostic marker for TNBC patients.

Given the association of active immune response and favorable outcome, there has been intense interest in using immune checkpoint blockade (ICB) therapy to treat TNBC patients. The efficacy of ICB largely depends on the pre-existing anti-tumor CD8+ T cells. Even with pre-selection of eligible patients (for instance, positive PD-L1 expression in tumors), the response rates are generally low [11]. In a recent report of the phase II KEYNOTE-086 study, an objective response rate is 21.4% for previously untreated, PD-L1-positive, metastatic TNBC patients [41]. Suggested evidence indicates that TNBC patients deserve further stratified in order to achieve the optimum therapeutic benefit. Of note, the current study provides a means for this purpose. Considering the less exhausted nature and a proposed tumor-reactive feature, the TNBC patients showing an increased prevalence of PD-1hiCD8+ T cells in tumors could be most benefited from immediate ICB therapy by targeting PD-1/PD-L1 or relevant pathways.

The differentiation of exhausted T cells is also closely linked with epigenetic modifications, including chromatin accessibility, DNA methylation and demethylation, histone modifications, which may constitute cell-intrinsic barriers for ICB therapy [42]. Indeed, we observed that the epigenetic regulator EZH2 and its target genes were strongly enriched in exhausted PD-1hiCD8+ T cells. EZH2 is the catalytic subunit of Polycomb repressive complex 2 (PRC2), which is a histone methyltransferase that targets Lys27 of histone H3. The methylated H3-K27 chromatin mark is commonly associated with silencing of target genes [43]. High expression of EZH2 in exhausted T cells also reported in two other studies [44,45]. Interestingly, EZH2 plays a vital role in mediating the effect of transcription factor YY1 to down-regulate the expression of IL-2 [45], and thus suggesting EZH2 is a negative functional regulator of exhausted T cells. In a murine model, it was shown that inhibition of EZH2 expression in T cells increased the effectiveness of anti-CTLA-4 therapy [46]. Moreover, EZH2 is crucial in the maintenance of regulatory T cell (Treg) identity [47], and interruption of EZH2 activity in Tregs, remodel the tumor microenvironment and enhances anti-tumor immunity [48]. Altogether, these above studies suggest that including an EZH2 inhibitor may potentiate the efficacy of ICB therapy by dampening Treg activity and enhancing exhausted T-cell functions in TNBC, a hypothesis to be tested in the future.

In summary, our study has provided a comprehensive picture of exhausted PD-1hiCD8+ T cells in BC. The unique feature of these T cells expands our understanding of the diversity of human exhausted T cells, as recent studies indicate that human exhausted T cells constitute a broad spectrum of dysfunctional states [49,50]. Furthermore, our study highlights that PD-1hiCD8+ T cells could be integrated as both a prognostic and a stratifying marker for TNBC patients. Targeting PD-1hiCD8+ T cells by ICB therapy and epigenetic modifier may provide long-lasting therapeutic benefits for TNBC patients.

Clinical perspectives

  • Tumor-infiltrating PD-1hi CD8+ T cells were largely unexplored in BC, especially in TNBC.

  • Tumoral PD-1hiCD8+ T cells in BC are partially exhausted, and their abundance signifies ‘hot’ immune status with favorable outcomes.

  • Reinvigorating this PD-1hiCD8+ T-cell subpopulation may provide further therapeutic opportunities in TNBC patients.

Competing Interests

The authors declare that there are no competing interests assocaited with the manuscript.

Funding

This work was supported by the Strategic Priority Research Program [grant number XDB29030302]; the Frontier Science Key Research Project, Chinese Academy of Sciences [grant number QYZDB-SSW-SMC036]; the National Natural Science Foundation of China [grant numbers 31770960, 81602488, 81861138010]; the National Key R&D Program of China [grant number 2017YFC1311004]; and the PIFI project [grant number 2016PB076].

Author Contribution

Conception and design: Jiong Wu and Xiaoming Zhang. Development of methodology: Liang Guo, Shyamal Goswami and Chunmei Cao. Acquisition of data (acquired and managed patients, provided facilities etc.): Liang Guo, Shyamal Goswami, Chunmei Cao, Teng Li, Xiaoyan Huang and Linxiaoxi Ma. Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): Liang Guo, Shyamal Goswami, Yicheng Guo, Jiong Wu and Xiaoming Zhang. Writing, review and/or revision of the manuscript: Liang Guo, Shyamal Goswami, Jiong Wu and Xiaoming Zhang. Administrative, technical or material support (i.e., reporting or organizing data, constructing databases): Chunmei Cao, Teng Li, Xiaoyan Huang, Benlong Yang and Yayun Chi. Study supervision: Jiong Wu and Xiaoming Zhang.

Data Availability

The NCBI accession number for the RNA-seq experiments reported in this manuscript is BioProject PRJNA529949.

Acknowledgements

We thank the bioinformatics support from Dr. Yun Liu from Fudan University, Dr. Zizhang Sheng from Columbia University and Dr. Yuanhua Liu from Institut Pasteur of Shanghai.

Abbreviations

     
  • BC

    breast cancer

  •  
  • CI

    confidence interval

  •  
  • CTLA4

    cytotoxic T lymphocyte antigen 4

  •  
  • DAPI

    4′,6-diamidino-2-phenylindole

  •  
  • DFS

    disease-free survival

  •  
  • EDTA

    ethylene diamine tetraacetic acid

  •  
  • ER

    estrogen receptor

  •  
  • EZH2

    enhancer of zeste homolog 2

  •  
  • FFPE

    formaldehyde-fixed and paraffin-embedded

  •  
  • GSEA

    gene set enrichment analysis

  •  
  • HCC

    hepatocellular carcinoma

  •  
  • HER2

    human epidermal growth factor receptor 2

  •  
  • HR

    hazard ratio

  •  
  • H&E

    Hematoxylin and Eosin

  •  
  • ICB

    immune checkpoint blockade

  •  
  • IFN

    interferon

  •  
  • IL-2

    interleukin-2

  •  
  • LP

    lymphocyte-predominant

  •  
  • mAb

    monoclonal antibody

  •  
  • mIHC

    multiplexed immunohistochemistry

  •  
  • NBF

    neutral buffered formalin

  •  
  • NSCLC

    non-small-cell lung cancer

  •  
  • OS

    overall survival

  •  
  • PBS

    phosphate buffer saline

  •  
  • PCA

    principal component analysis

  •  
  • PD-1

    programmed cell death protein-1

  •  
  • PD-L1

    programmed cell death 1 ligand 1

  •  
  • PMA

    phorbol 12-myristate 13-acetate

  •  
  • PR

    progesterone receptor

  •  
  • QIF

    quantitative immunofluorescence

  •  
  • RCC

    renal cell carcinoma

  •  
  • RNA-seq

    RNA sequencing

  •  
  • TCR

    T cell receptor

  •  
  • TIL

    tumor-infiltrating lymphocyte

  •  
  • TMA

    tissue microarray

  •  
  • TNBC

    triple-negative BC

  •  
  • TNF

    tumor necrosis factor

  •  
  • TRAV

    TCRα variable

  •  
  • TRBV

    TCRβ variable

  •  
  • Treg

    regulatory T cell

  •  
  • TRM

    tissue-resident memory T cell

  •  
  • TSA

    tyramide signal amplification

References

References
1.
Bray
F.
,
Ferlay
J.
,
Soerjomataram
I.
,
Siegel
R.L.
,
Torre
L.A.
and
Jemal
A.
(
2018
)
Global cancer statistics 2018: GLOBOCAN estimates of incidence and mortality worldwide for 36 cancers in 185 countries
.
CA Cancer J. Clin.
68
,
394
424
[PubMed]
2.
Dent
R.
,
Trudeau
M.
,
Pritchard
K.I.
,
Hanna
W.M.
,
Kahn
H.K.
,
Sawka
C.A.
et al.
(
2007
)
Triple-negative breast cancer: clinical features and patterns of recurrence
.
Clin. Cancer Res.
13
,
4429
4434
[PubMed]
3.
Savas
P.
,
Salgado
R.
,
Denkert
C.
,
Sotiriou
C.
,
Darcy
P.K.
,
Smyth
M.J.
et al.
(
2016
)
Clinical relevance of host immunity in breast cancer: from TILs to the clinic
.
Nat. Rev. Clin. Oncol.
13
,
228
241
[PubMed]
4.
Dushyanthen
S.
,
Beavis
P.A.
,
Savas
P.
,
Teo
Z.L.
,
Zhou
C.
,
Mansour
M.
et al.
(
2015
)
Relevance of tumor-infiltrating lymphocytes in breast cancer
.
BMC Med.
13
,
202
[PubMed]
5.
Ali
H.R.
,
Provenzano
E.
,
Dawson
S.J.
,
Blows
F.M.
,
Liu
B.
,
Shah
M.
et al.
(
2014
)
Association between CD8+ T-cell infiltration and breast cancer survival in 12,439 patients
.
Ann. Oncol.
25
,
1536
1543
[PubMed]
6.
Ribas
A.
,
Kefford
R.
,
Marshall
M.A.
,
Punt
C.J.
,
Haanen
J.B.
,
Marmol
M.
et al.
(
2013
)
Phase III randomized clinical trial comparing tremelimumab with standard-of-care chemotherapy in patients with advanced melanoma
.
J. Clin. Oncol.
31
,
616
622
[PubMed]
7.
Robert
C.
,
Thomas
L.
,
Bondarenko
I.
,
O’Day
S.
,
Weber
J.
,
Garbe
C.
et al.
(
2011
)
Ipilimumab plus dacarbazine for previously untreated metastatic melanoma
.
N. Engl. J. Med.
364
,
2517
2526
[PubMed]
8.
Kwon
E.D.
,
Drake
C.G.
,
Scher
H.I.
,
Fizazi
K.
,
Bossi
A.
,
van den Eertwegh
A.J.
et al.
(
2014
)
Ipilimumab versus placebo after radiotherapy in patients with metastatic castration-resistant prostate cancer that had progressed after docetaxel chemotherapy (CA184-043): a multicentre, randomised, double-blind, phase 3 trial
.
Lancet Oncol.
15
,
700
712
[PubMed]
9.
Fehrenbacher
L.
,
Spira
A.
,
Ballinger
M.
,
Kowanetz
M.
,
Vansteenkiste
J.
,
Mazieres
J.
et al.
(
2016
)
Atezolizumab versus docetaxel for patients with previously treated non-small-cell lung cancer (POPLAR): a multicentre, open-label, phase 2 randomised controlled trial
.
Lancet
387
,
1837
1846
[PubMed]
10.
Nanda
R.
,
Chow
L.Q.
,
Dees
E.C.
,
Berger
R.
,
Gupta
S.
,
Geva
R.
et al.
(
2016
)
Pembrolizumab in patients with advanced triple-negative breast cancer: phase Ib KEYNOTE-012 Study
.
J. Clin. Oncol.
34
,
2460
2467
[PubMed]
11.
Polk
A.
,
Svane
I.M.
,
Andersson
M.
and
Nielsen
D.
(
2018
)
Checkpoint inhibitors in breast cancer - current status
.
Cancer Treat. Rev.
63
,
122
134
[PubMed]
12.
Wherry
E.J.
and
Kurachi
M.
(
2015
)
Molecular and cellular insights into T cell exhaustion
.
Nat. Rev. Immunol.
15
,
486
499
[PubMed]
13.
Wang
C.
,
Singer
M.
and
Anderson
A.C.
(
2017
)
Molecular dissection of CD8(+) T-cell dysfunction
.
Trends Immunol.
38
,
567
576
[PubMed]
14.
Thommen
D.S.
,
Koelzer
V.H.
,
Herzig
P.
,
Roller
A.
,
Trefny
M.
,
Dimeloe
S.
et al.
(
2018
)
A transcriptionally and functionally distinct PD-1(+) CD8(+) T cell pool with predictive potential in non-small-cell lung cancer treated with PD-1 blockade
.
Nat. Med.
24
,
994
1004
[PubMed]
15.
Kim
H.D.
,
Song
G.W.
,
Park
S.
,
Jung
M.K.
,
Kim
M.H.
,
Kang
H.J.
et al.
(
2018
)
Association between expression level of PD1 by tumor-infiltrating CD8(+) T cells and features of hepatocellular carcinoma
.
Gastroenterology
155
,
1936
1950.e17
16.
Granier
C.
,
Dariane
C.
,
Combe
P.
,
Verkarre
V.
,
Urien
S.
,
Badoual
C.
et al.
(
2017
)
Tim-3 expression on tumor-infiltrating PD-1(+) CD8(+) T cells correlates with poor clinical outcome in renal cell carcinoma
.
Cancer Res.
77
,
1075
1082
[PubMed]
17.
Woroniecka
K.
,
Chongsathidkiet
P.
,
Rhodin
K.
,
Kemeny
H.
,
Dechant
C.
,
Farber
S.H.
et al.
(
2018
)
T-cell exhaustion signatures vary with tumor type and are severe in glioblastoma
.
Clin. Cancer Res.
24
,
4175
4186
[PubMed]
18.
Kansy
B.A.
,
Concha-Benavente
F.
,
Srivastava
R.M.
,
Jie
H.B.
,
Shayan
G.
,
Lei
Y.
et al.
(
2017
)
PD-1 status in CD8(+) T cells associates with survival and anti-PD-1 therapeutic outcomes in head and neck cancer
.
Cancer Res.
77
,
6353
6364
[PubMed]
19.
Love
M.I.
,
Huber
W.
and
Anders
S.
(
2014
)
Moderated estimation of fold change and dispersion for RNA-seq data with DESeq2
.
Genome Biol.
15
,
550
[PubMed]
20.
Subramanian
A.
,
Tamayo
P.
,
Mootha
V.K.
,
Mukherjee
S.
,
Ebert
B.L.
,
Gillette
M.A.
et al.
(
2005
)
Gene set enrichment analysis: a knowledge-based approach for interpreting genome-wide expression profiles
.
Proc. Natl. Acad. Sci. U.S.A.
102
,
15545
15550
[PubMed]
21.
Kumar
B.V.
,
Ma
W.
,
Miron
M.
,
Granot
T.
,
Guyer
R.S.
,
Carpenter
D.J.
et al.
(
2017
)
Human tissue-resident memory T cells are defined by core transcriptional and functional signatures in lymphoid and mucosal sites
.
Cell Rep.
20
,
2921
2934
[PubMed]
22.
Mackay
L.K.
,
Minnich
M.
,
Kragten
N.A.
,
Liao
Y.
,
Nota
B.
,
Seillet
C.
et al.
(
2016
)
Hobit and Blimp1 instruct a universal transcriptional program of tissue residency in lymphocytes
.
Science
352
,
459
463
[PubMed]
23.
Hombrink
P.
,
Helbig
C.
,
Backer
R.A.
,
Piet
B.
,
Oja
A.E.
,
Stark
R.
et al.
(
2016
)
Programs for the persistence, vigilance and control of human CD8(+) lung-resident memory T cells
.
Nat. Immunol.
17
,
1467
1478
[PubMed]
24.
Li
B.
,
Li
T.
,
Pignon
J.C.
,
Wang
B.
,
Wang
J.
,
Shukla
S.A.
et al.
(
2016
)
Landscape of tumor-infiltrating T cell repertoire of human cancers
.
Nat. Genet.
48
,
725
732
[PubMed]
25.
Salgado
R.
,
Denkert
C.
,
Demaria
S.
,
Sirtaine
N.
,
Klauschen
F.
,
Pruneri
G.
et al.
(
2015
)
The evaluation of tumor-infiltrating lymphocytes (TILs) in breast cancer: recommendations by an International TILs Working Group 2014
.
Ann. Oncol.
26
,
259
271
[PubMed]
26.
Camp
R.L.
,
Dolled-Filhart
M.
and
Rimm
D.L.
(
2004
)
X-tile: a new bio-informatics tool for biomarker assessment and outcome-based cut-point optimization
.
Clin. Cancer Res.
10
,
7252
7259
[PubMed]
27.
Willinger
T.
,
Freeman
T.
,
Hasegawa
H.
,
McMichael
A.J.
and
Callan
M.F.
(
2005
)
Molecular signatures distinguish human central memory from effector memory CD8 T cell subsets
.
J. Immunol.
175
,
5895
5903
[PubMed]
28.
Woroniecka
K.
,
Chongsathidkiet
P.
,
Rhodin
K.
,
Kemeny
H.
,
Dechant
C.
,
Farber
S.H.
et al.
(
2018
)
T-cell exhaustion signatures vary with tumor type and are severe in glioblastoma
.
Clin. Cancer Res.
24
,
4175
4186
[PubMed]
29.
Brenchley
J.M.
,
Karandikar
N.J.
,
Betts
M.R.
,
Ambrozak
D.R.
,
Hill
B.J.
,
Crotty
L.E.
et al.
(
2003
)
Expression of CD57 defines replicative senescence and antigen-induced apoptotic death of CD8+ T cells
.
Blood
101
,
2711
2720
[PubMed]
30.
Joshi
N.S.
and
Kaech
S.M.
(
2008
)
Effector CD8 T cell development: a balancing act between memory cell potential and terminal differentiation
.
J. Immunol.
180
,
1309
1315
[PubMed]
31.
Corgnac
S.
,
Boutet
M.
,
Kfoury
M.
,
Naltet
C.
and
Mami-Chouaib
F.
(
2018
)
The emerging role of CD8(+) tissue resident memory T (TRM) cells in antitumor immunity: a unique functional contribution of the CD103 Integrin
.
Front. Immunol.
9
,
1904
[PubMed]
32.
Paley
M.A.
,
Kroy
D.C.
,
Odorizzi
P.M.
,
Johnnidis
J.B.
,
Dolfi
D.V.
,
Barnett
B.E.
et al.
(
2012
)
Progenitor and terminal subsets of CD8+ T cells cooperate to contain chronic viral infection
.
Science
338
,
1220
1225
[PubMed]
33.
Denkert
C.
,
Loibl
S.
,
Noske
A.
,
Roller
M.
,
Muller
B.M.
,
Komor
M.
et al.
(
2010
)
Tumor-associated lymphocytes as an independent predictor of response to neoadjuvant chemotherapy in breast cancer
.
J. Clin. Oncol.
28
,
105
113
[PubMed]
34.
Denkert
C.
,
von Minckwitz
G.
,
Brase
J.C.
,
Sinn
B.V.
,
Gade
S.
,
Kronenwett
R.
et al.
(
2015
)
Tumor-infiltrating lymphocytes and response to neoadjuvant chemotherapy with or without carboplatin in human epidermal growth factor receptor 2-positive and triple-negative primary breast cancers
.
J. Clin. Oncol.
33
,
983
991
[PubMed]
35.
Salgado
R.
,
Denkert
C.
,
Campbell
C.
,
Savas
P.
,
Nuciforo
P.
,
Aura
C.
et al.
(
2015
)
Tumor-infiltrating lymphocytes and associations with pathological complete response and event-free survival in HER2-positive early-stage breast cancer treated with lapatinib and trastuzumab: a secondary analysis of the NeoALTTO Trial
.
JAMA Oncol.
1
,
448
454
[PubMed]
36.
Zarour
H.M.
(
2016
)
Reversing T-cell dysfunction and exhaustion in cancer
.
Clin. Cancer Res.
22
,
1856
1864
[PubMed]
37.
Thommen
D.S.
and
Schumacher
T.N.
(
2018
)
T cell dysfunction in cancer
.
Cancer Cell
33
,
547
562
[PubMed]
38.
Simoni
Y.
,
Becht
E.
,
Fehlings
M.
,
Loh
C.Y.
,
Koo
S.L.
,
Teng
K.W.W.
et al.
(
2018
)
Bystander CD8(+) T cells are abundant and phenotypically distinct in human tumour infiltrates
.
Nature
557
,
575
579
[PubMed]
39.
Duhen
T.
,
Duhen
R.
,
Montler
R.
,
Moses
J.
,
Moudgil
T.
,
de Miranda
N.F.
et al.
(
2018
)
Co-expression of CD39 and CD103 identifies tumor-reactive CD8 T cells in human solid tumors
.
Nat. Commun.
9
,
2724
[PubMed]
40.
Li
H.
,
van der Leun
A.M.
,
Yofe
I.
,
Lubling
Y.
,
Gelbard-Solodkin
D.
,
van Akkooi
A.C.J.
et al.
(
2019
)
Dysfunctional CD8 T cells form a proliferative, dynamically regulated compartment within human melanoma
.
Cell
176
,
775
789.e18
41.
Adams
S.
,
Loi
S.
,
Toppmeyer
D.
,
Cescon
D.W.
,
De Laurentiis
M.
,
Nanda
R.
et al.
(
2019
)
Pembrolizumab monotherapy for previously untreated, PD-L1-positive, metastatic triple-negative breast cancer: cohort B of the phase II KEYNOTE-086 study
.
Ann. Oncol.
30
,
405
411
[PubMed]
42.
Ghoneim
H.E.
,
Zamora
A.E.
,
Thomas
P.G.
and
Youngblood
B.A.
(
2016
)
Cell-intrinsic barriers of T cell-based immunotherapy
.
Trends Mol. Med.
22
,
1000
1011
[PubMed]
43.
Cao
R.
,
Wang
L.
,
Wang
H.
,
Xia
L.
,
Erdjument-Bromage
H.
,
Tempst
P.
et al.
(
2002
)
Role of histone H3 lysine 27 methylation in Polycomb-group silencing
.
Science
298
,
1039
1043
[PubMed]
44.
Mognol
G.P.
,
Spreafico
R.
,
Wong
V.
,
Scott-Browne
J.P.
,
Togher
S.
,
Hoffmann
A.
et al.
(
2017
)
Exhaustion-associated regulatory regions in CD8(+) tumor-infiltrating T cells
.
PNAS
114
,
E2776
E2785
[PubMed]
45.
Balkhi
M.Y.
,
Wittmann
G.
,
Xiong
F.
and
Junghans
R.P.
(
2018
)
YY1 upregulates checkpoint receptors and downregulates type I cytokines in exhausted, chronically stimulated human T cells
.
iScience
2
,
105
122
[PubMed]
46.
Goswami
S.
,
Apostolou
I.
,
Zhang
J.
,
Skepner
J.
,
Anandhan
S.
,
Zhang
X.
et al.
(
2018
)
Modulation of EZH2 expression in T cells improves efficacy of anti-CTLA-4 therapy
.
J. Clin. Invest.
128
,
3813
3818
[PubMed]
47.
Arvey
A.
,
van der Veeken
J.
,
Samstein
R.M.
,
Feng
Y.
,
Stamatoyannopoulos
J.A.
and
Rudensky
A.Y.
(
2014
)
Inflammation-induced repression of chromatin bound by the transcription factor Foxp3 in regulatory T cells
.
Nat. Immunol.
15
,
580
587
[PubMed]
48.
Wang
D.
,
Quiros
J.
,
Mahuron
K.
,
Pai
C.C.
,
Ranzani
V.
,
Young
A.
et al.
(
2018
)
Targeting EZH2 reprograms intratumoral regulatory T cells to enhance cancer immunity
.
Cell Rep.
23
,
3262
3274
[PubMed]
49.
Bengsch
B.
,
Ohtani
T.
,
Khan
O.
,
Setty
M.
,
Manne
S.
,
O’Brien
S.
et al.
(
2018
)
Epigenomic-guided mass cytometry profiling reveals disease-specific features of exhausted CD8 T cells
.
Immunity
48
,
1029
1045.e5
50.
Miller
B.C.
,
Sen
D.R.
,
Al Abosy
R.
,
Bi
K.
,
Virkud
Y.V.
,
LaFleur
M.W.
et al.
(
2019
)
Subsets of exhausted CD8(+) T cells differentially mediate tumor control and respond to checkpoint blockade
.
Nat. Immunol.
20
,
326
336
[PubMed]

Author notes

*

These authors contributed equally to this work.