You are here

  • Home
  • Archive
  • Volume 10, Issue 6
  • Tumor immune microenvironment of primary prostate cancer with and without germline mutations in homologous recombination repair genes

Article Text

Article menu
Basic tumor immunology
Original research
Tumor immune microenvironment of primary prostate cancer with and without germline mutations in homologous recombination repair genes
  1. Anna Sofia Trigos1,2,
  2. Anupama Pasam1,
  3. Patricia Banks3,
  4. Roslyn Wallace3,
  5. Christina Guo4,5,
  6. Simon Keam1,2,
  7. Heather Thorne1,2,
  8. kConFab1,2,
  9. Catherine Mitchell6,
  10. Stephen Lade6,
  11. David Clouston7,
  12. Alexander Hakansson8,
  13. Yang Liu8,
  14. Benjamin Blyth1,2,
  15. Declan Murphy2,9,
  16. Nathan Lawrentschuk2,3,
  17. Damien Bolton10,
  18. Daniel Moon3,
  19. Phil Darcy1,2,
  20. Ygal Haupt1,2,
  21. Scott G Williams11,
  22. Elena Castro12,
  23. David Olmos12,13,
  24. David Goode1,2,
  25. Paul Neeson1,2 and
  26. Shahneen Sandhu1,2,3
  1. 1Division of Cancer Research, Peter MacCallum Cancer Centre, Melbourne, Victoria, Australia
  2. 2Sir Peter MacCallum Department of Oncology, The University of Melbourne, Parkville, Victoria, Australia
  3. 3Department of Medical Oncology, Peter MacCallum Cancer Centre, Melbourne, Victoria, Australia
  4. 4Institute of Cancer Research Sutton, Sutton, Surrey, UK
  5. 5Royal Marsden Hospital Sutton, Sutton, London, UK
  6. 6Department of Pathology, Peter MacCallum Cancer Centre, Melbourne, Victoria, Australia
  7. 7TissuPath, Melbourne, Victoria, Australia
  8. 8Veracyte Inc, South San Francisco, California, USA
  9. 9Division of Cancer Surgery, Peter MacCallum Cancer Centre, Melbourne, Victoria, Australia
  10. 10Austin Hospital, Heidelberg, Victoria, Australia
  11. 11Division of Radiation Oncology and Cancer Imaging, Peter MacCallum Cancer Centre, Melbourne, Victoria, Australia
  12. 12Instituto de Investigacion Biomedica de Malaga, Malaga, Spain
  13. 13Medical Oncology Department, Instituto de Investigación Hospital 12 de Octubre, Madrid, Spain
  1. Correspondence to Dr Shahneen Sandhu; Shahneen.Sandhu{at}petermac.org

Abstract

Background Aberrations in homologous recombination repair (HRR) genes are emerging as important biomarkers for personalized treatment in prostate cancer (PCa). HRR deficiency (HRD) could affect the tumor immune microenvironment (TIME), potentially contributing to differential responses to poly ADP-ribose polymerase (PARP) inhibitors and immune checkpoint inhibitors. Spatial distribution of immune cells in a range of cancers identifies novel disease subtypes and is related to prognosis. In this study we aimed to determine the differences in the TIME of PCa with and without germline (g) HRR mutations.

Methods We performed gene expression analysis, multiplex immunohistochemistry of T and B cells and quantitative spatial analysis of PCa samples from 36 patients with gHRD and 26 patients with sporadic PCa. Samples were archival tumor tissue from radical prostatectomies with the exception of one biopsy. Results were validated in several independent cohorts.

Results Although the composition of the T cell and B cells was similar in the tumor areas of gHRD-mutated and sporadic tumors, the spatial profiles differed between these cohorts. We describe two T-cell spatial profiles across primary PCa, a clustered immune spatial (CIS) profile characterized by dense clusters of CD4+ T cells closely interacting with PD-L1+ cells, and a free immune spatial (FIS) profile of CD8+ cells in close proximity to tumor cells. gHRD tumors had a more T-cell inflamed microenvironment than sporadic tumors. The CIS profile was mainly observed in sporadic tumors, whereas a FIS profile was enriched in gHRD tumors. A FIS profile was associated with lower Gleason scores, smaller tumors and longer time to biochemical recurrence and metastasis.

Conclusions gHRD-mutated tumors have a distinct immune microenvironment compared with sporadic tumors. Spatial profiling of T-cells provides additional information beyond T-cell density and is associated with time to biochemical recurrence, time to metastasis, tumor size and Gleason scores.

  • gene expression profiling
  • genetic markers
  • prostatic neoplasms
  • tumor microenvironment

Data availability statement

Data are available upon reasonable request. Our patient consent does not allow depositing data online. Data will be made available to other researchers upon reasonable request.

http://creativecommons.org/licenses/by-nc/4.0/

This is an open access article distributed in accordance with the Creative Commons Attribution Non Commercial (CC BY-NC 4.0) license, which permits others to distribute, remix, adapt, build upon this work non-commercially, and license their derivative works on different terms, provided the original work is properly cited, appropriate credit is given, any changes made indicated, and the use is non-commercial. See http://creativecommons.org/licenses/by-nc/4.0/.

https://doi.org/10.1136/jitc-2021-003744

Statistics from Altmetric.com

    Request Permissions

    If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.

    WHAT IS ALREADY KNOWN ON THIS TOPIC

    • Prostate cancer (PCa) is generally an immunologically cold tumor. Patients with PCa with homologous recombination repair deficiency (HRD) have shown improved responses to immune checkpoint inhibition (ICI) compared with non-HRD cancers, but to date, the molecular basis of this difference has not been elucidated.

    WHAT THIS STUDY ADDS

    • We profiled the tumor immune microenvironment of PCas with germline mutations in homologous recombination repair genes for the first time, and show that these cancers have a more T-cell inflamed microenvironment than sporadic tumors. Further, our gene expression signature was associated with longer time to biochemical recurrence and metastasis.

    HOW THIS STUDY MIGHT AFFECT RESEARCH, PRACTICE, OR POLICY

    • Immune spatial profiling of the microenvironment in PCa may provide prognostic information and define a subset of patients that may benefit from ICI. Our findings warrant further validation in prospective studies.

    Background

    Homologous recombination repair deficiency (HRD), predominantly caused by BRCA1/2 alterations, are enriched in metastatic castration resistant prostate cancer (mCRPC) and serve as important biomarkers for personalized therapy. The PROfound phase III trial established a survival advantage for the poly ADP-ribose polymerase (PARP) inhibitor olaparib in mCRPC with HRD leading to regulatory approval and implementation of mainstream molecular testing and personalized treatment in a subset of mCRPC patients. Notably, the response rate for PARP inhibitors in BRCA1/2 altered tumors was consistently 40%–50% across multiple trials,1–3 implying the presence of genomic alterations in the homologous recombination repair (HRR) pathway alone may be insufficient to predict responses. Biallelic deletions, loss of heterozygosity of the wild type allele,4 functional implications of the specific variant, co-occurring genomic alterations and other factors such as the tumor immune microenvironment (TIME) may potentially contribute to the differential responses observed. In breast and ovarian cancer, response to programmed death 1 (PD-1/PD-ligand 1 (PD-L1)) inhibition is enriched in tumors with germline (g) BRCA1/2 mutations and these tumors have been shown to harbor higher tumor mutational load, lymphocyte infiltration and tumor-specific neoantigens for immune activation.5–8 In the KEYNOTE-199 and Checkmate 650 trials, HRD in prostate cancer (PCa) were also associated with higher and more durable responses to immune checkpoint inhibitors (ICI)9 10 although the biological basis for this observation is poorly understood. Currently, we lack an understanding of the TIME of PCa in patients with gHRD mutations, and how this differs from that of sporadic tumors.

    PCa is generally considered an immunologically cold tumor.11 While a higher tumor immune infiltration has been associated with better immune control and prognosis in other cancers,12 studies in PCa have revealed conflicting results, with higher T-cell levels in tumor areas associated with better prognosis in some studies,13 14 but worse in others.15–19 Possible explanations include heterogeneity between cohorts, the complexity of the spatial interactions between immune and tumor cells and generally low levels of immune infiltration that make the application of the classical definitions of immune infiltrated or excluded tumors challenging. In melanoma, sarcoma and breast cancer, complex spatial patterns of distinct immune subsets in the stroma and tumor-immune interactions have been linked to overall survival and response to immunotherapy.20–26 Further studies of the interface between tumor genomics and the TIME are warranted to understand how best to personalize therapies. Herein we profiled the density and spatial distribution of T and B cells in primary PCa with and without gHRD in tumor and surrounding stroma using gene expression and quantitative spatial analysis.

    Methods

    Sample cohort

    Our cohort consisted of 62 primary PCa samples. Twenty-six samples were from patients with gBRCA2 mutations, five from patients with gBRCA1 mutations, a double gBRCA2+gMSH2 carrier and four with other gDNA repair mutations (ATM, FANCI, PALB2, CHEK2, n=1 of each) (figure 1, online supplemental methods). We refer to the cohort with germline alterations in HRR genes hereafter as the HRD cohort/tumors. An additional 26 were primary PCa samples from patients without gDNA repair mutations based on germline testing using an established targeted DNA repair panel of 26 genes and are referred hereafter as ‘sporadic’. All samples were radical prostatectomies, except for the gFANCI sample which was a prostate biopsy, thereby enabling a comprehensive evaluation of the TIME in whole prostates. Clinical characteristics of the cohort and germline variants are outlined in online supplemental tables S1 and S2.

    Supplemental material

    Supplemental material

    Figure 1
    Figure 1

    Study flow chart. Our study includes 36 samples from primary PCa with germline mutations in HRR genes and 26 from patients with sporadic PCa. All samples underwent OPAL mIHC. Thirty-three HRD and 26 sporadic samples passed quality control for NanoString gene expression analysis. HRD, homologous recombination repair deficiency; HRR, homologous recombination repair; g, germline; n, number of samples; mIHC, multiplex immunohistochemistry; PCa, prostate cancer.

    Gene expression

    We used the NanoString nCounter PanCancer Immune Profiling Panel comprizing of 730 immune-related genes and 40 housekeeping genes. RNA was extracted from formalin fixed paraffin embedded (FFPE) sections using the RNAeasy FFPE kit (Qiagen). Fifty-nine samples were evaluable. One gBRCA1 and two gBRCA2 samples were excluded from the gene expression analyses due to poor RNA quality. We performed differential expression analyses and investigated the expression of specific signatures such as the tumor inflammation signature (TIS)27 which measures a T-cell inflamed tumor microenvironment that has previously been shown to correlate with response to PD-1/L1 blockade. We also evaluated the tissue-resident memory (TRM) T-cell signature which is implicated in maintaining immunity.28

    Multiplex immunohistochemistry

    Given the prevalence of TIS and TRM signatures in the HRD cohort, seven-color OPAL multiplex immunohistochemistry (mIHC) was used to quantify CD3+CD4+ (helper T cells), CD3+CD8+ (cytotoxic T cells), CD3+CD4+FOXP3+ (regulatory T cells (Tregs)), tumor cells (AMACR+) and PD-L1+ cells (online supplemental methods). Tumor areas were marked by a pathologist. Fifteen representative multispectral images of the tumor area and five of normal stromal tissue outside of the tumor area (adjacent to the tumor and distant from the tumor) were obtained to gain a comprehensive representation of the tumor and stroma within whole prostatectomy (n=62) and tumor biopsy (n=1) samples. These selected regions are referred to as regions of interest (ROI).

    CD20 immunohistochemistry

    To better understand the spatial distribution of B cells and the formation of tertiary lymphoid structures (TLS), we performed 3,3'-diaminobenzidine (DAB)-staining of B-cells with CD20 (Clone L26, Agilent Technologies) and dendritic cells with CD21 (Clone 1G9, Novocastra antibody, Leica) in serial sections of the same sample set used for OPAL mIHC. Slides were scanned on the VS120 slide scanner (Olympus) and analyzed for the presence of DAB-positive cells using the CellSens Dimension Desktop software (Olympus). After co-registering the DAB-stained and OPAL-stained images, we matched the 15 fields in the tumor area and 5 in the normal stromal tissue to those selected in the OPAL T cell panel. The presence and number of B cells per ROI was then indexed to the mIHC data. Identification of TLS was carried out with the aid of an expert hematopathologist based on morphology and size (an area of at least 2000 µm²) of the B cell aggregates and the CD21 staining to define the follicular architecture.

    Spatial analysis

    Spatial analysis of the immune microenvironment was carried out using the SPIAT R package, which was developed inhouse for the spatial image analysis of cells in tissues.29 Only images with at least 100 tumor cells were considered for the tumor area analyses. Clusters of CD3+CD4+, CD3+CD8+, CD3+CD4+FOXP3+, CD3+CD4−CD8− and AMACR−PD-L1+ cells were defined using a previously described custom algorithm.29 To identify clusters, first we calculated the distance between all T and AMACR−PD-L1+ cells in all images. The top 0.5% pairs of closest cells were regarded as being neighbors. Groups of neighboring cells within the same image were defined as aggregates. Visual inspection confirmed such aggregates were clearly evident. Each of these individual aggregates were considered ‘clusters’ in our analysis if comprised of at least 50 cells. Cells not in clusters or in aggregates of less than 50 cells were considered ‘free’. A cut-off of 50 was selected based on the distribution of the composition of cell types, where cells in aggregates of less than 50 cells had a similar composition to those that were not in aggregates, whereas those of more than 50 cells had a composition more consistent with those in larger clusters. Visual inspection also supported that intuitively 50 cells corresponded to a clearly identifiable cluster of cells. Mann-Whitney tests were used to test for statistical significance. One-sided tests were used to test for differences in a specified direction.

    Validation cohorts

    The association between gene expression signatures, tumor size, Gleason score and Kaplan-Meier analysis for time to biochemical recurrence and metastasis were validated in four large independent cohorts of primary PCas with gene expression data: 497 samples from The Cancer Genome Atlas (TCGA),30 73 from the Fraser et al cohort,31 8635 from the Decipher GRID database,32 including 855 from a meta-analysis of radical prostatectomy (RP) patients from the Spratt et al cohort,33 545 of RP patients from the Erho et al cohort34 (both with available survival outcomes data) and 7235 from RP patients obtained from clinical use of the Decipher test ordered by physicians between 2013 and 2017 (with baseline pathology information) as previously described.32

    Results

    HRD tumors have a more inflamed TIME than sporadic tumors

    We performed differential expression analysis between HRD and sporadic tumors to understand differences in the TIME profile. We obtained 190 differentially expressed genes (adjusted p value <0.05) (figure 2A). Upregulated genes in the HRD cohort included B-cell markers (CD79A, CD79B, MS4A1), the immune checkpoint CD96, as well as cytokines and chemokines (CXCL13, CXCL9, CXCL10) (online supplemental table S3). The most highly differentially expressed gene in the HRD cohort was the Major Histocompatibilty Complex (MHC) Class I molecule Human Leukocyte Antigen (HLA)-A, which is required for immune recognition by cytotoxic effector T cells (fold change=2.60, p value adjusted for multiple testing using Benjamini-Hochberg Procedure=3.66×10–8) (figure 2A,B). To investigate whether there was a T-cell–inflamed TIME in the HRD cohort, we used the TIS, which has previously been shown to measure tumor inflammation and enhanced responses to ICI.27 HRD samples had higher TIS levels than sporadic tumors (one-sided Mann-Whitney test p value=8.12×10–5) (figure 2C, online supplemental figure S1A). HRD samples also had higher levels of a TRM T-cell signature28 (one-sided Mann-Whitney test p value=0.00051) (figure 2D, online supplemental figure S1B), which is linked to immunosurveillance, improved prognosis and ICI modulation.28 35 Overall, the HRD tumors had a more T-cell–inflamed TIME than sporadic tumors.

    Supplemental material

    Figure 2
    Figure 2

    Gene expression analysis of HRD and sporadic tumors. (A) Differential gene expression between HRD and sporadic tumor samples. There were 190 differentially expressed genes, including 125 upregulated in HRD tumors, and 65 upregulated in sporadic tumors. Red dots correspond to genes with adjusted p values<0.05. The horizontal line corresponds to an adjusted p value of 0.05. (B) Gene expression levels of HLA-A in HRD and sporadic tumors. HLA-A were higher in HRD tumors, although gATM, gFANCI and gPALB2 samples had lower levels than gBRCA2 and gBRCA1 tumors (fold change=2.60, adjusted p value=3.66×10–8). (C) Tumor inflammation signature (TIS) in HRD and sporadic tumors. HRD tumors had significantly higher TIS levels (one-sided Mann-Whitney test p value=8.12×10–5), indicating a more inflamed T-cell microenvironment. (D) Tissue-resident memory T-cell (TRM) signature in HRD and sporadic tumors. HRD tumors had significantly higher TRM signature levels (one-sided Mann-Whitney test p value=0.00051), indicating a more inflamed T-cell microenvironment. g, germline; HRD, homologous recombination repair deficiency; HLA-A, Human Leukocyte Antigen A.

    T-cell densities in the TIME of HRD and sporadic tumors

    To investigate the composition of T cells within tumors, we used seven-color OPAL mIHC to characterize helper (CD3+CD4+), cytotoxic (CD3+CD8+), regulatory (CD3+CD4+FOXP3+) and double negative (CD3+CD4–CD8–) T-cell populations (figure 3A). T-cell density of the tumor area was heterogeneous across cohorts (figure 3B,C and online supplemental figure S2) (43.93 to 1089.01 cells/mm2 in HRD samples and 48.06 cells/mm2 to 2,161.14 cells/mm2 in sporadic samples) (figure 3D). The tumor area of HRD samples had lower T-cell densities (median of 303.91 cells/mm2) compared with those of sporadic samples (433.99 cells/mm2), although this difference was not statistically significant (Mann-Whitney test p value=0.14). There was enrichment of T cells in the tumor area compared with non-tumor areas in 18/26 gBRCA2 samples, 20/26 sporadic, 4/5 gBRCA1 and in the gATM, gCHEK2, gFANCI, gPALB2, gBRCA2+gMSH2 samples (figure 3E).

    Figure 3
    Figure 3

    The T-cell and PD-L1 composition of the microenvironment of primary prostate tumors with and without HRD. (A) OPAL mIHC of tissue sections. Tumor and stromal regions were marked by a pathologist. Fifteen representative tumor areas (blue squares) and five representative normal stromal areas (red squares) of 1.3 mm2 were selected from each tissue section. (B and C) OPAL mIHC of gBRCA2 tumors with high (B) and low levels (C) of T cell, indicating a range of T cell densities in the microenvironment. (D) Density of T cells across the cohorts. HRD tumors tended to have lower T-cell densities, although the difference was not statistically significant. (E) Enrichment of T cells in the tumor area. Values greater than 0 indicate a higher density of T cells in the tumor area, whereas values lower than 0 indicate depletion. The majority of samples in both cohorts (77% of sporadic samples and 75% of HRD samples), are enriched in T cells in the tumor area. (F and G) Composition of T-cell populations in HRD (F) and sporadic cohorts (G). CD3+CD4+ cells were the most common, followed by CD3+CD8+ cells and CD3+CD4–CD8– cells. CD3+CD4+FOXP3+ were rare in our cohort. Values depicted are medians±SE. g, germline; HRD, homologous recombination repair deficiency; mIHC, multiplex immunohistochemistry.

    T-cell density heterogeneity was associated with the type of the mutated gene in HRD tumors. gBRCA2 tumors had a median of 368.90 cells/mm2 (range=52.57–1089.00 cells/mm2). gBRCA1 tumors clustered towards the lower range of the spectrum in the HRD cohort (median=184.28 cells/mm2, range=43.93–288.01 cells/mm2), as did tumors with germline mutations in gFANCI, gATM, gPALB2 and gCHEK2 (54.47, 79.80, 151.10 and 347.17 cells/mm2, respectively). Levels of T cells in the gBRCA2+gMSH2 tumor was similar to gBRCA1 and gBRCA2 samples (173.35 cells/mm2) (figure 3D).

    The composition of the T cell population in the tumor areas was similar across both cohorts (figure 3F,G and online supplemental figure S3). CD4+ T cells were the most common subtype, representing a median of 69.78% and 70.08% of the T-cell population in the HRD and sporadic cohorts, respectively. This was followed by CD8+ T cells (25.47% for HRD, 26.49% for sporadic), double negative T cells (3.22% for HRD, 1.65% for sporadic), and Tregs (0% for HRD, 0.46% for sporadic). gBRCA2 had similar percentages of CD8+ T cells to sporadic tumors (Mann-Whitney test p value=0.82), and similar CD8+/CD4+ ratios (Mann-Whitney test p value=0.96). However, gBRCA1 tumors had a higher percentage of CD8+ T cells than gBRCA2 tumors (one-sided Mann-Whitney test p value=0.018), and a higher CD8+/CD4+ ratio (one-sided Mann-Whitney test p value=0.013). The double carrier gBRCA2+gMSH2 tumor had the highest proportion of CD8+ T cells (50.77%), consistent with previous reports for tumors with mismatch repair mutations.36

    PD-L1+ cells were rare across cohorts

    Since tumor PD-L1 expression is currently approved as a predictive biomarker for PD-L1 blockade in several cancers, and HRD tumors had higher TIS scores, a potential predictor of response to PD-1/PD-L1 blockade,27 we investigated the presence PD-L1+ cells. PD-L1+ tumor and non-tumor cells were rare across both cohorts, consistent with previous reports.37 Only 1/26 gBRCA2 and 4/26 sporadic samples had more than 1% PD-L1+ stromal cells (online supplemental figure S4A,B). Similarly, only 2/26 sporadic and 2/26 gBRCA2 samples had more than 1% PD-L1+ tumor cells (online supplemental figure S4C,D).

    Spatial distribution of tumor-infiltrated T cells

    Despite similar T-cell density and composition between the two cohorts, T cells displayed distinct spatial profiles across tumors. In some tumor areas, T cells were aggregated into clusters (figure 4A), while in others T cells were individually scattered across the tumor area (figure 4B). We refer to these patterns as the cluster immune spatial (CIS) profile characterized by ‘clustered’ T cells largely in the stromal area, and the free immune spatial (FIS) profile, characterized by ‘free’ T cells largely in the tumor regions. To quantify these patterns, we performed spatial analysis using SPIAT,29 which allowed identifying immune cells forming clusters from those that were freely distributed across the tissue (see Methods) and characterizing their composition.

    Figure 4
    Figure 4

    Clustered and free immune spatial profiles. (A) mIHC image depicting an example of a CIS profile within the tumor region in a sporadic tumor. T cells are aggregated in a cluster away from tumor cells. (B) An example of a FIS profile within the tumor region of a gBRCA2 tumor. T cells are freely distributed and interacting with tumor cells. (C) Density of clusters of T cells in HRD and sporadic tumors. HRD tumors are depleted of clusters compared with sporadics. (D) Ratio of free to clustered cells per mm2. HRD tumors have overall higher levels of the free/clustered cell ratio. The horizontal line indicates the maximum ratio in sporadic. (E) Percentage of free CD8+ T cells in HRD and sporadic tumors. The percentage of free CD8+ T cells is higher in HRD tumors compared with sporadic tumors (one-sided Mann-Whitney test p value=0.015). (F) Percentage of clustered CD8+ T cells in HRD and sporadic tumors. The percentage of clustered CD8+ T cells is higher in sporadic tumors compared with HRD tumors (one-sided Mann-Whitney test p value=0.015). FIS, free immune spatial; CIS, clustered immune spatial; g, germline; HRD, homologous recombination repair deficiency; mIHC, multiplex immunohistochemistry.

    The predominant pattern of CIS and FIS profiles differed between HRD and sporadic cohort, implying that T-cell density alone may be insufficient to understand the biological implications of T cells in PCa. To quantify the CIS profile in each patient, we averaged the number of clusters across each of the ROIs selected in the tumor area for each patient. HRD tumors had fewer clusters, with a median of 0.48 clusters/mm2 compared with 1.06 clusters/mm2 in sporadic tumors (one-sided Mann-Whitney test p value=0.024) (figure 4C). In contrast, HRD tumors had significantly more free T cells infiltrating the tumor compared with sporadic tumors. The median ratio of the free/clustered cells per mm2 was 67.71 in HRD tumors compared with 48.34 in sporadic tumors (one-sided Mann-Whitney test p value=0.0039). This ratio may infer the degree to which T cells are available to interact with tumor cells (figure 4D). In 13/36 (36.11%) of HRD tumors (10/26 gBRCA2, 2/5 gBRCA1 and the gBRCA2+gMSH2 tumor) the free/clustered ratio was higher than in any of the sporadic tumors (figure 4D). Notably, HRD tumors also had higher percentages of free CD8+ T cells compared with sporadic tumors (one-sided Mann-Whitney test p value=0.015) (figure 4E), whereas sporadic tumors had higher levels of clustered CD8+ T cells (one-sided Mann-Whitney test p value=0.015) (figure 4F).

    Formation of TLS

    Given the higher expression of B cell markers in the HRD cohort, we next investigated how the T cell spatial profiles (CIS and FIS) associated with B cells and if these B cells formed TLS within the tumor and/or stromal regions. We performed immunohistochemistry (IHC) staining for CD20 to investigate the density of B-cells across HRD and sporadic samples. Overall, B cells were significantly rarer than T cells in the tumor areas of both cohorts (medians of 341.99 T cells/mm2, 32.61 B cells/mm2, one-sided Man-Whitney test p=1.16×10–9 in HRD tumors, and 416.74 T cells/mm2, 36.99 B cells/mm2, p=5.96×10–8 in sporadic tumors) (figure 5A,B). There were no statistical differences in the B-cell density in tumor areas of gBRCA2 and sporadic samples (median density in gBRCA2 samples=46.31 cells/mm2, sporadic tumors=36.99 cells/mm2, Mann-Whitney test p value=0.81) (figure 5C).

    Figure 5
    Figure 5

    B cells and TLS in the HRD and sporadic cohorts. (A) Example multiplex IHC of the OPAL T cell panel (left) and IHC of CD20, B cells (right). T cells are more common than B cells. (B) Density of T cells and B cells in HRD and sporadic tumors. B cells are rarer than T cells in both cohorts (Mann-Whitney p value=1.16×10–9 in HRD tumors and 5.96×10–8 in sporadic tumors). (C) Density of B cells in the tumor area of HRD and sporadic samples. gBRCA2 tumors have similar B cell densities to sporadic tumors (two-sided Man-Whitney test p value=0.81). (D) Density of B cells in the stromal area of HRD and sporadic samples. gBRCA2 tumors had higher levels of B cells in the stromal area compared with sporadic tumors (one-sided Man-Whitney test p value=0.032). (E) Example TLS in a sporadic sample, showing the classical pattern of T cells (left) surrounding dense B-cell regions (middle), with a core of dendritic cells (right) forming the follicular structure. (F) Density of T cell clusters and TLS across the cohorts. The density of T cell clusters is consistently larger than that of TLS. (G) Clusters of T cells with no accompanying B cell staining. g, germline; HRD, homologous recombination repair deficiency; IHC, immunohistochemistry; TLS, tertiary lymphoid structures.

    There was a higher density of B cells in the tumor area of sporadic samples compared with the stroma (one-sided Man-Whitney paired test p value=0.00065). In contrast, the B cell density in the surrounding stroma of gBRCA2 samples was similar to that found in its tumor areas (median in stroma=36.79 cells/mm2, median in tumor area=46.31 cells/mm2, Man-Whitney paired test p value=0.56). In fact, the density of B cells in the stroma of gBRCA2 samples was higher than that in sporadic samples (median density in gBRCA2 samples=36.79 cells/mm2, median density in sporadic samples=11.75 cells/mm2, one-sided Man-Whitney test p value=0.032) (figure 5D). This higher density of B-cells in the stroma of gBRCA2 samples might account for the overexpression of B cell markers in the HRD cohort (figure 2A).

    A similar proportion of sporadic and gBRCA2 samples displayed no B cells in their tumor area (3/25 sporadic and 4/26 gBRCA2 samples), but had no unique clinical characteristics. One of the two profiled gBRCA1 samples did not display B cells, and the other only had a density of 23.56 cells/mm2 in the tumor area and 14.92 cells/mm2 in the stromal area. The gCHEK2, gFANCI and gPALB2 tumors had no B cells, and the gATM and gBRCA2+gMSH2 samples had similarly low levels (24.57 cells/mm2 in the tumor area and 7.46 cells/mm2 in the stromal area of the gATM sample, and 21.93 cells/mm2 in the tumor area and 23.58 cells/mm2 in the stromal area of the gBRCA2+gMSH2 sample). Overall, the tumor areas of HRD tumors were not enriched in B cells, although these tumors had higher levels of B cells in the surrounding stroma compared with sporadic tumors.

    We identified TLS based on CD20 staining and used CD21 to verify follicular architecture (figure 5E). Despite HRD samples displaying a FIS profile, we observed the presence of TLS in the tumor area of 65.38% (17/26) of gBRCA2 samples and in the gATM and gBRCA2+gMSH2 tumors. This was similar to the percentage found in sporadic samples (76.00%, 19/25). However, the median density across the cohorts was low (0.057 TLS/mm2 in the tumor area and 0 TLS/mm2 in the stroma area of HRD samples, 0.079 TLS/mm2 in the tumor area and 0 TLS/mm2 in the stroma area of sporadic samples). There were no differences in the TLS density of sporadic and HRD samples in the tumor area (Mann-Whitney test p value=0.40) or stromal areas (Mann-Whitney test p value=0.61).

    Clusters of T cells associated with a CIS profile were significantly more common than TLS (figure 5F) (one-sided paired Mann-Whitney p value=1.35×10–6 for HRD tumors and 9.70×10–6 for sporadic tumors), with the majority of T cell clusters not co-localizasing with B cells (figure 5G). T cell clusters were more common than TLS in sporadic and HRD tumors with TLS (median of 7.20 and 4.67 times, respectively). This points towards the likely formation of a CIS profile as a distinct process to the formation of TLS.

    Finally, we investigated whether B cell or TLS density were associated with clinical outcome. We found no significant association with time to biochemical recurrence, time to metastasis, or survival with either density of B cells or density of TLSs (p values=0.99, 0.5, 0.97, and 0.88, 0.78, 0.87, respectively).

    The free and clustered T-cell populations

    The distribution of T-cell subtypes was distinct between the clustered and free populations. As T cell clusters got bigger (included more cells), there was a higher proportion of CD4+ T cells (one-sided Jonckheere Telspra (JT) test p=3.74×10–4) (figure 6A), whereas the percentage of CD8+ T cells decreased (one-sided JT test p=2.25×10–8) (figure 6B). In contrast, a higher percentage of free T cells within the tumor were CD8+ compared to the clustered T cell population (one-sided paired Mann-Whitney test p value=7.75×10–10) (figure 6C), and a lower percentage were CD4+ (p value=5×10–8) (figure 6D). The CD8+/CD4+ cell ratios were higher in free cells compared with clustered cells (JT one-sided test p<2.2×10–16) (figure 6E). Given the established roles of CD4+ T and CD8+ T cells, free and clustered T cell populations likely have distinct biological implications.

    Figure 6
    Figure 6

    Clustered and free immune spatial profiles. (A) Percentage of CD4+ T cells in immune clusters. Larger clusters have higher percentages of CD4+ cells, suggesting functional aggregation of CD4+ T cells. (B) Percentage of CD8+ T cells in immune clusters. Larger cluster have lower percentages of CD8+ cells, indicating a depletion of these cells. (C and D) Comparison of CD4+ and CD8+ T cells in the clustered and free populations. There are higher levels of CD8+ cells in the free T cell populations, but higher levels of CD4+ cells in the clustered population. (E) Ratio of CD8+/CD4+ cells (log2) in free and clustered cells of different size. Larger clusters tended to have a lower CD8+/CD4+ ratio. (F) Average minimum distances between CD4+ and CD8+ T cells with Tregs and PD-L1+ stromal cells. CD4+ and CD8+ T cells are more closely interacting with Tregs and PD-L1+ stromal cells in clusters than when they are freely distributed, pointing towards an inhibitory environment in the immune clusters. (G) Percentage of PD-L1+ stromal cells in clusters and in freely distributed cells. Clusters are enriched in PD-L1+ cells compared with free cells. (H) The average minimum distance from CD8+ T cells to tumor cells is shorter in the free T cell population compared with the clustered population, indicating greater levels of tumor recognition. (I) Association between levels of HLA-A and the ratio of free/clustered CD8+ T cells, showing a positive correlation (Spearman correlation=0.42, p value=9.7×10–4). PD-L1, programmed death ligand 1; Treg, regulatory T cells; HLA-A, Human Leukocyte Antigen A.

    We hypothesized that a CIS profile could be linked to a limited immune-tumor cell interaction, whereas FIS profile could be linked to higher levels of tumor immune recognition. To investigate this, we measured the distance of CD4+ and CD8+ T cells to Tregs and PD-L1+ cells, which suppress T cell responses. The distances were shorter in the clustered compared with free T-cell populations (one-sided Mann-Whitney test p values=5.36×10–10 for CD4+ to CD4+FOXP3+, 3.02×10–16 for CD4+ to PD-L1+, 7.23×10–9 for CD8+ to CD4+FOXP3+, 3.05×10–16 for CD8+ to PD-L1+ cells) (figure 6F). Furthermore, PD-L1+ cells were more commonly found in the clustered T-cell population (4.97% of clustered cells) than the free population (2.20%) (one-sided paired Mann-Whitney test p value=1.51×10–7) (figure 6G). These results may suggest that the clustering of T cells which occurs predominantly in the stomal regions may be indicative of an immunosuppressive microenvironment.

    To investigate whether the free T cell population represented a population of T cells likely involved in tumor immune recognition, we calculated the distance of free and clustered CD8+ cells to tumor cells. Free CD8+ cells were indeed closer to tumor cells than clustered CD8+ cells (one-sided Mann-Whitney p value=0.0074) (figure 6H). Furthermore, expression of the MHC Class I molecule HLA-A required for immune recognition by cytotoxic effector T cells was positively correlated with the ratio of free to clustered CD8+ T cells (Spearman correlation=0.42, p value=9.7×10–4) (figure 6I), consistent with higher immune recognition.

    FIS profile gene signatures predict disease aggressiveness and survival outcomes

    We next hypothesized that a FIS profile might be linked to clinical characteristics of tumors. We derived a gene expression signature for the FIS profile by calculating the Spearman correlation between the ratio of free and clustered CD8+ cells and the genes available in the NanoString platform and selecting the top five most highly correlated genes (IRF7, CEACAM1, ITGAM, LILRA1 and BAX). There was a positive correlation between the levels of this signature and the ratio of free to clustered CD8+ T cells in our cohort (online supplemental figure S5) (Spearman correlation=0.45, p value=4.07×10–4), suggesting it captures a FIS profile.

    Consistent with our mIHC analyses, HRD tumors had higher FIS profile signatures than sporadic samples (one-sided Mann-Whitney test p value=3.04×10–4) (figure 7A). While the FIS of HRD tumors remained consistent across grade groups, there was a decreasing trend in sporadic tumors, although not significant in our cohort (decreasing JT test=0.42) (online supplemental figure S6). Using data from 497 tumors from TCGA,30 we found that FIS profile signature levels were higher in smaller tumors (decreasing JT p value=0.0037) and tumors with a lower Gleason score (p value=0.0046) (figure 7B,C). These trends were also found in the Fraser cohort31 of 73 samples (online supplemental figure S7) and in the Erho et al cohort34 of 545 samples (p value=10–5) (figure 7D). In this cohort the FIS profile signature was higher in tumors of patients that remained free of regional or distant metastasis after RP (one-sided Mann-Whitney p value=2.06×10–4) (figure 7E) and patients with a FIS profile signature above the median had longer time to metastasis (p value=0.02) (figure 7F). A similar trend was observed in the Spratt et al cohort of 855 patients33 (p value=0.010) (figure 7G). Time to biochemical recurrence was longer in patients with tumors with a higher FIS profile signature in the Fraser cohort31 (log-rank test p value=0.041) (figure 7H). Finally, in a cohort of 8635 RP patients we found higher FIS profile scores associated with lower Glinsky signature scores,38 corresponding to the better prognosis group (JT test p<2×10–4) (online supplemental figure S8). These results support our hypothesis of better accessibility of CD8+ cells to tumor cells in PCas with a FIS profile that may translate into better disease control.

    Figure 7
    Figure 7

    FIS profile gene expression signature is a marker for low grade tumors, longer time to biochemical recurrence and metastasis. (A) FIS profile signature scores across our cohort of HRD and sporadic tumors. Consistent with the multiplex immunohistochemistry results, HRD tumors have higher FIS profile scores. gBRCA1, gBRCA2 and the gFANCI and gBRCA2+gMSH2 tumors have the highest scores. gATM, gCHEK2, gPALB2 have similar levels to those found in sporadic tumors. (B, C) Association of the FIS profile signature with tumor size (B) and Gleason score (C) from 497 primary tumors of the The Cancer Genome Atlas data set. Smaller tumors and of lower Gleason score had higher FIS profiles scores. (D, E) Association of the FIS profile signature with the Gleason score (D) and the development of metastasis after radical prostatectomy (E) from 545 samples of the Erho et al cohort. (F, G) Kaplan-Meier curve of time to metastasis in the Erho et al cohort of 545 patients (F) and the Spratt et al cohort of 855 patients (G). Patient tumors with a FIS profile signature above the median had longer times to metastasis. (H) Time to biochemical recurrence in the data set of 73 non-indolent primary prostate cancers from Fraser et al showing that the time to biochemical recurrence was longer for tumors with a stronger FIS profile signature. FIS, free immune spatial; g, germline; HRD, homologous recombination repair deficiency.

    Discussion

    To our knowledge this the first study to deeply profile the spatial distribution of the TIME in up to 15 distinct areas in whole primary PCas with and without germline HRD mutations. Our results reveal several novel findings including immune gene expression signatures linked to HRD status and a complex spatial structure of the T-cell microenvironment linked to prognostic factors.

    We identified a more inflamed T-cell immune microenvironment in PCas with germline HRD mutations, including cytokines, chemokines, and higher levels of the TIS and TRM gene expression signatures. Despite no significant differences in the density or composition of the tumor T-cell microenvironment between cohorts, spatial analysis revealed higher levels of free CD8+ T cells that were closer to tumor cells and higher levels of HLA-A expression in the HRD cohort, potentially suggestive of better immune tumor recognition in this subset of tumors. Data from several clinical trials (Checkmate 65039 and KEYNOTE 1999) have shown improved responses to immune check point inhibition in the HRD cohort.

    To date most immune microenvironment studies in PCa have focused on density estimations, with studies reporting conflicting results regarding the significance of tumor immune infiltration levels,13–19 including in tumors with BRCA2 mutations.40 Our quantitative spatial analyses reveal a free CD8+ T cell spatial profile in the tumor area associated with positive factors for prognosis, including smaller tumor size, lower Gleason score, longer time to biochemical recurrence and onset of metastasis. This profile was more prominent in gHRD tumors, despite gHRD mutations generally being associated with more aggressive disease.41 One potential hypothesis is that a FIS profile can emerge from distinct underlying biological processes, and these may be distinct in HRD and sporadic tumors. For example, we have shown that the FIS is consistent across grade groups in HRD tumors, whereas it decreases with increasing grade group in sporadic tumors.

    While we identified TLS in both the HRD and sporadic cohort, the densities of B cells and TLS were substantially lower than that of T cells and T cell clusters. TLS have been associated with better response to immunotherapy in multiple cancers, including melanoma,20 21 renal cell carcinoma,20 head and neck,42 although this has not been described for PCa. In the KEYNOTE-199 and Checkmate 650 trials, HRD was associated with higher and more durable responses to ICI.9 39 We did not find enrichment of TLS in the tumor area of the HRD cohort compared with the sporadic cohort, although there was higher density of B cells in the stroma of HRD samples, thereby accounting for the increased B cell markers seen in the gene expression data for the HRD cohort.

    A major unanswered question is the role of the TIME in HRD cancers in mediating durable responses to PARP inhibitors and/or ICIs. Exploratory assays such as FIS profile may complement established genomic assays and warrants further investigation in the context of prospective trials of ICI.

    We acknowledge several inherent limitations of our study. We have only analyzed gHRD mutations and have not evaluated somatic events in HRR genes that may also impact the TIME. Recent reports suggest that most germline BRCA2 events have corresponding heterozygous loss of the second allele.4 Second, there were only five BRCA1 cases and one case each of ATM, CHECK2, PALB2, FANCI, and MSH2, making it impossible to draw definitive conclusions in these smaller subsets. Our findings should be considered hypothesis-generating and should be further validated in larger prospective studies.

    Conclusions

    HRD tumors have a more inflamed TIME than sporadic tumors. A free spatial profile of CD8+ T cells may be linked with better disease control. Spatial profiling and the FIS profile signature also provides prognostic information that warrants further investigation in prospective studies.

    Supplemental material

    Data availability statement

    Data are available upon reasonable request. Our patient consent does not allow depositing data online. Data will be made available to other researchers upon reasonable request.

    Ethics statements

    Patient consent for publication

    Ethics approval

    This study involves human participants and was approved by Peter Mac HREC committee ID 97/27 and HREC/16/PMCC/59 (31457). Participants gave informed consent to participate in the study before taking part.

    Acknowledgments

    We thank the patients and their families. The authors would like to thank Drs Robert B Den, Felix FY Feng, Stephen Freedland, Robert B Jenkins, R Jeffery Karnes, Eric A Klein, Ashley E Ross, Edward M Schaeffer, and Daniel E Spratt for providing access to outcomes data for retrospective cohorts analyzed in the GRID registry. We acknowledge the Centre for Advanced Histology and Microscopy (CAHM) at the Peter MacCallum Cancer Centre for their support of this work.

    References

    1. ↵
      1. de Bono J,
      2. Mateo J,
      3. Fizazi K, et al
      . Olaparib for metastatic castration-resistant prostate cancer. N Engl J Med 2020;382:2091–102.doi:10.1056/NEJMoa1911440pmid:http://www.ncbi.nlm.nih.gov/pubmed/32343890
      OpenUrlCrossRefPubMed
    2. ↵
      1. Mateo J,
      2. Porta N,
      3. Bianchini D, et al
      . Olaparib in patients with metastatic castration-resistant prostate cancer with DNA repair gene aberrations (TOPARP-B): a multicentre, open-label, randomised, phase 2 trial. Lancet Oncol 2020;21:162–74.doi:10.1016/S1470-2045(19)30684-9pmid:http://www.ncbi.nlm.nih.gov/pubmed/31806540
      OpenUrlCrossRefPubMed
    3. ↵
      1. Smith MR,
      2. Scher HI,
      3. Sandhu S, et al
      . Niraparib in patients with metastatic castration-resistant prostate cancer and DNA repair gene defects (GALAHAD): a multicentre, open-label, phase 2 trial. Lancet Oncol 2022;23:362–73.doi:10.1016/S1470-2045(21)00757-9pmid:http://www.ncbi.nlm.nih.gov/pubmed/35131040
      OpenUrlPubMed
    4. ↵
      1. Carreira S,
      2. Porta N,
      3. Arce-Gallego S, et al
      . Biomarkers associating with PARP inhibitor benefit in prostate cancer in the TOPARP-B trial. Cancer Discov 2021;11:2812–27.doi:10.1158/2159-8290.CD-21-0007pmid:http://www.ncbi.nlm.nih.gov/pubmed/34045297
      OpenUrlAbstract/FREE Full Text
    5. ↵
      1. Clarke B,
      2. Tinker AV,
      3. Lee C-H, et al
      . Intraepithelial T cells and prognosis in ovarian carcinoma: novel associations with stage, tumor type, and BRCA1 loss. Mod Pathol 2009;22:393–402.doi:10.1038/modpathol.2008.191pmid:http://www.ncbi.nlm.nih.gov/pubmed/19060844
      OpenUrlCrossRefPubMedWeb of Science
    6. ↵
      1. McAlpine JN,
      2. Porter H,
      3. Köbel M, et al
      . BRCA1 and BRCA2 mutations correlate with TP53 abnormalities and presence of immune cell infiltrates in ovarian high-grade serous carcinoma. Mod Pathol 2012;25:740–50.doi:10.1038/modpathol.2011.211pmid:http://www.ncbi.nlm.nih.gov/pubmed/22282309
      OpenUrlCrossRefPubMed
    7. ↵
      1. Nolan E,
      2. Savas P,
      3. Policheni AN, et al
      . Combined immune checkpoint blockade as a therapeutic strategy for BRCA1-mutated breast cancer. Sci Transl Med 2017;9:eaal4922.doi:10.1126/scitranslmed.aal4922pmid:http://www.ncbi.nlm.nih.gov/pubmed/28592566
      OpenUrlAbstract/FREE Full Text
    8. ↵
      1. Strickland KC,
      2. Howitt BE,
      3. Shukla SA, et al
      . Association and prognostic significance of BRCA1/2-mutation status with neoantigen load, number of tumor-infiltrating lymphocytes and expression of PD-1/PD-L1 in high grade serous ovarian cancer. Oncotarget 2016;7:13587–98.doi:10.18632/oncotarget.7277pmid:http://www.ncbi.nlm.nih.gov/pubmed/26871470
      OpenUrlPubMed
    9. ↵
      1. Antonarakis ES,
      2. Piulats JM,
      3. Gross-Goupil M, et al
      . Pembrolizumab for treatment-refractory metastatic castration-resistant prostate cancer: Multicohort, open-label phase II KEYNOTE-199 study. J Clin Oncol 2020;38:395–405.doi:10.1200/JCO.19.01638pmid:http://www.ncbi.nlm.nih.gov/pubmed/31774688
      OpenUrlCrossRefPubMed
    10. ↵
      1. Sharma P,
      2. Pachynski RK,
      3. Narayan V, et al
      . Nivolumab plus ipilimumab for metastatic castration-resistant prostate cancer: preliminary analysis of patients in the CheckMate 650 trial. Cancer Cell 2020;38:489–99.doi:10.1016/j.ccell.2020.08.007pmid:http://www.ncbi.nlm.nih.gov/pubmed/32916128
      OpenUrlPubMed
    11. ↵
      1. Berger MF,
      2. Lawrence MS,
      3. Demichelis F, et al
      . The genomic complexity of primary human prostate cancer. Nature 2011;470:214–20.doi:10.1038/nature09744pmid:http://www.ncbi.nlm.nih.gov/pubmed/21307934
      OpenUrlCrossRefPubMedWeb of Science
    12. ↵
      1. Fridman WH,
      2. Zitvogel L,
      3. Sautès-Fridman C, et al
      . The immune contexture in cancer prognosis and treatment. Nat Rev Clin Oncol 2017;14:717–34.doi:10.1038/nrclinonc.2017.101pmid:http://www.ncbi.nlm.nih.gov/pubmed/28741618
      OpenUrlCrossRefPubMed
    13. ↵
      1. Yang Y,
      2. Attwood K,
      3. Versaggi C, et al
      . Association of high CD8+ tumor infiltrating lymphocytes at prostatectomy with improved survival of prostate cancer patients. JCO 2018;36:5068–68.doi:10.1200/JCO.2018.36.15_suppl.5068
      OpenUrl
    14. ↵
      1. Nardone V,
      2. Botta C,
      3. Caraglia M, et al
      . Tumor infiltrating T lymphocytes expressing FoxP3, CCR7 or PD-1 predict the outcome of prostate cancer patients subjected to salvage radiotherapy after biochemical relapse. Cancer Biol Ther 2016;17:1213–20.doi:10.1080/15384047.2016.1235666pmid:http://www.ncbi.nlm.nih.gov/pubmed/27791459
      OpenUrlPubMed
    15. ↵
      1. Kärjä V,
      2. Aaltomaa S,
      3. Lipponen P, et al
      . Tumour-infiltrating lymphocytes: a prognostic factor of PSA-free survival in patients with local prostate carcinoma treated by radical prostatectomy. Anticancer Res 2005;25:4435–8.pmid:http://www.ncbi.nlm.nih.gov/pubmed/16334122
      OpenUrlAbstract/FREE Full Text
    16. ↵
      1. Ness N,
      2. Andersen S,
      3. Valkov A, et al
      . Infiltration of CD8+ lymphocytes is an independent prognostic factor of biochemical failure-free survival in prostate cancer. Prostate 2014;74:1452–61.doi:10.1002/pros.22862pmid:http://www.ncbi.nlm.nih.gov/pubmed/25111810
      OpenUrlCrossRefPubMed
    17. ↵
      1. Leclerc BG,
      2. Charlebois R,
      3. Chouinard G, et al
      . CD73 expression is an independent prognostic factor in prostate cancer. Clin Cancer Res 2016;22:158–66.doi:10.1158/1078-0432.CCR-15-1181pmid:http://www.ncbi.nlm.nih.gov/pubmed/26253870
      OpenUrlAbstract/FREE Full Text
    18. ↵
      1. Petitprez F,
      2. Fossati N,
      3. Vano Y, et al
      . PD-L1 expression and CD8+ T-cell infiltrate are associated with clinical progression in patients with node-positive prostate cancer. Eur Urol Focus 2019;5:192–6.doi:10.1016/j.euf.2017.05.013pmid:28753812
      OpenUrlPubMed
    19. ↵
      1. Zhao SG,
      2. Lehrer J,
      3. Chang SL, et al
      . The immune landscape of prostate cancer and nomination of PD-L2 as a potential therapeutic target. J Natl Cancer Inst 2019;111:301–10.doi:10.1093/jnci/djy141pmid:http://www.ncbi.nlm.nih.gov/pubmed/30321406
      OpenUrlCrossRefPubMed
    20. ↵
      1. Helmink BA,
      2. Reddy SM,
      3. Gao J, et al
      . B cells and tertiary lymphoid structures promote immunotherapy response. Nature 2020;577:549–55.doi:10.1038/s41586-019-1922-8pmid:http://www.ncbi.nlm.nih.gov/pubmed/31942075
      OpenUrlCrossRefPubMed
    21. ↵
      1. Cabrita R,
      2. Lauss M,
      3. Sanna A, et al
      . Tertiary lymphoid structures improve immunotherapy and survival in melanoma. Nature 2020;577:561–5.doi:10.1038/s41586-019-1914-8pmid:http://www.ncbi.nlm.nih.gov/pubmed/31942071
      OpenUrlPubMed
    22. ↵
      1. Petitprez F,
      2. de Reyniès A,
      3. Keung EZ, et al
      . B cells are associated with survival and immunotherapy response in sarcoma. Nature 2020;577:556–60.doi:10.1038/s41586-019-1906-8pmid:http://www.ncbi.nlm.nih.gov/pubmed/31942077
      OpenUrlCrossRefPubMed
    23. ↵
      1. Keren L,
      2. Bosse M,
      3. Marquez D, et al
      . A structured tumor-immune microenvironment in triple negative breast cancer revealed by multiplexed ion beam imaging. Cell 2018;174:e19:1373–87.doi:10.1016/j.cell.2018.08.039pmid:http://www.ncbi.nlm.nih.gov/pubmed/30193111
      OpenUrlPubMed
    24. ↵
      1. Huang Y-K,
      2. Wang M,
      3. Sun Y, et al
      . Macrophage spatial heterogeneity in gastric cancer defined by multiplex immunohistochemistry. Nat Commun 2019;10:3928.doi:10.1038/s41467-019-11788-4pmid:http://www.ncbi.nlm.nih.gov/pubmed/31477692
      OpenUrlCrossRefPubMed
    25. ↵
      1. Saltz J,
      2. Gupta R,
      3. Hou L, et al
      . Spatial organization and molecular correlation of tumor-infiltrating lymphocytes using deep learning on pathology images. Cell Rep 2018;23:181–93.doi:10.1016/j.celrep.2018.03.086pmid:http://www.ncbi.nlm.nih.gov/pubmed/29617659
      OpenUrlPubMed
    26. ↵
      1. Gong C,
      2. Anders RA,
      3. Zhu Q, et al
      . Quantitative characterization of CD8+ T cell clustering and spatial heterogeneity in solid tumors. Front Oncol 2018;8:649.doi:10.3389/fonc.2018.00649pmid:http://www.ncbi.nlm.nih.gov/pubmed/30666298
      OpenUrlPubMed
    27. ↵
      1. Ayers M,
      2. Lunceford J,
      3. Nebozhyn M, et al
      . IFN-γ-related mRNA profile predicts clinical response to PD-1 blockade. J Clin Invest 2017;127:2930–40.doi:10.1172/JCI91190pmid:http://www.ncbi.nlm.nih.gov/pubmed/28650338
      OpenUrlCrossRefPubMed
    28. ↵
      1. Savas P,
      2. Virassamy B,
      3. Ye C, et al
      . Single-cell profiling of breast cancer T cells reveals a tissue-resident memory subset associated with improved prognosis. Nat Med 2018;24:986–93.doi:10.1038/s41591-018-0078-7pmid:http://www.ncbi.nlm.nih.gov/pubmed/29942092
      OpenUrlPubMed
    29. ↵
      1. Yang T,
      2. Ozcoban V,
      3. Pasam A
      . SPIAT: an R package for the spatial image analysis of cells in tissues. bioRxiv2020.
    30. ↵
      1. Cancer Genome Atlas Research Network
      . The molecular taxonomy of primary prostate cancer. Cell 2015;163:1011–25.doi:10.1016/j.cell.2015.10.025pmid:http://www.ncbi.nlm.nih.gov/pubmed/26544944
      OpenUrlCrossRefPubMed
    31. ↵
      1. Fraser M,
      2. Sabelnykova VY,
      3. Yamaguchi TN, et al
      . Genomic hallmarks of localized, non-indolent prostate cancer. Nature 2017;541:359–64.doi:10.1038/nature20788pmid:http://www.ncbi.nlm.nih.gov/pubmed/28068672
      OpenUrlCrossRefPubMed
    32. ↵
      1. Ben-Salem S,
      2. Hu Q,
      3. Liu Y, et al
      . Diversity in androgen receptor action among treatment-naïve prostate cancers is reflected in treatment response predictions and molecular subtypes. Eur Urol Open Sci 2020;22:34–44.doi:10.1016/j.euros.2020.10.002pmid:http://www.ncbi.nlm.nih.gov/pubmed/33299986
      OpenUrlPubMed
    33. ↵
      1. Spratt DE,
      2. Yousefi K,
      3. Deheshi S, et al
      . Individual patient-level meta-analysis of the performance of the decipher genomic classifier in high-risk men after prostatectomy to predict development of metastatic disease. J Clin Oncol 2017;35:1991–8.doi:10.1200/JCO.2016.70.2811pmid:http://www.ncbi.nlm.nih.gov/pubmed/28358655
      OpenUrlCrossRefPubMed
    34. ↵
      1. Erho N,
      2. Crisan A,
      3. Vergara IA, et al
      . Discovery and validation of a prostate cancer genomic classifier that predicts early metastasis following radical prostatectomy. PLoS One 2013;8:e66855.doi:10.1371/journal.pone.0066855pmid:http://www.ncbi.nlm.nih.gov/pubmed/23826159
      OpenUrlCrossRefPubMed
    35. ↵
      1. Edwards J,
      2. Wilmott JS,
      3. Madore J, et al
      . CD103+ tumor-resident CD8+ T cells are associated with improved survival in immunotherapy-naïve melanoma patients and expand significantly during anti-PD-1 treatment. Clin Cancer Res 2018;24:3036–45.doi:10.1158/1078-0432.CCR-17-2257pmid:http://www.ncbi.nlm.nih.gov/pubmed/29599411
      OpenUrlAbstract/FREE Full Text
    36. ↵
      1. Guedes LB,
      2. Antonarakis ES,
      3. Schweizer MT, et al
      . MSH2 loss in primary prostate cancer. Clin Cancer Res 2017;23:6863–74.doi:10.1158/1078-0432.CCR-17-0955pmid:http://www.ncbi.nlm.nih.gov/pubmed/28790115
      OpenUrlAbstract/FREE Full Text
    37. ↵
      1. Haffner MC,
      2. Guner G,
      3. Taheri D, et al
      . Comprehensive evaluation of programmed death-ligand 1 expression in primary and metastatic prostate cancer. Am J Pathol 2018;188:1478–85.doi:10.1016/j.ajpath.2018.02.014pmid:http://www.ncbi.nlm.nih.gov/pubmed/29577933
      OpenUrlCrossRefPubMed
    38. ↵
      1. Glinsky GV,
      2. Glinskii AB,
      3. Stephenson AJ, et al
      . Gene expression profiling predicts clinical outcome of prostate cancer. J Clin Invest 2004;113:913–23.doi:10.1172/JCI20032pmid:http://www.ncbi.nlm.nih.gov/pubmed/15067324
      OpenUrlCrossRefPubMedWeb of Science
    39. ↵
      Initial results from a phase II study of nivolumab (NIVO) plus ipilimumab (IPI) for the treatment of metastatic castration-resistant prostate cancer (mCRPC; CheckMate 650). Genitourinary Cancers Symposium; 2019 Presented February 14, 2019; San Francisco, CA.
    40. ↵
      1. Jenzer M,
      2. Keß P,
      3. Nientiedt C, et al
      . The BRCA2 mutation status shapes the immune phenotype of prostate cancer. Cancer Immunol Immunother 2019;68:1621–33.doi:10.1007/s00262-019-02393-xpmid:http://www.ncbi.nlm.nih.gov/pubmed/31549213
      OpenUrlPubMed
    41. ↵
      1. Pritchard CC,
      2. Mateo J,
      3. Walsh MF, et al
      . Inherited DNA-repair gene mutations in men with metastatic prostate cancer. N Engl J Med 2016;375:443–53.doi:10.1056/NEJMoa1603144pmid:http://www.ncbi.nlm.nih.gov/pubmed/27433846
      OpenUrlCrossRefPubMed
    42. ↵
      1. Ruffin AT,
      2. Cillo AR,
      3. Tabib T, et al
      . B cell signatures and tertiary lymphoid structures contribute to outcome in head and neck squamous cell carcinoma. Nat Commun 2021;12:3349.doi:10.1038/s41467-021-23355-xpmid:http://www.ncbi.nlm.nih.gov/pubmed/34099645
      OpenUrlCrossRefPubMed

    Supplementary materials

    Footnotes

    • Twitter @anna_t_g

    • Contributors Concept and design of study and guarantor: SS. Provision of study materials or patients: SS, DMu, NL, DB, DMo, HT, kConFab. Collection and assembly of data: AP, PB, RW, CG, CM, SLa, DC. Design of data analysis: AST. Data analysis and interpretation: AST, AP, CM, SLa, DC, BB, AH, YH, SPK, DG, PJN, SS. Manuscript writing and critical evaluation: All authors. Final approval of manuscript: All authors. Accountable for all aspects of the work: All authors.

    • Funding This study was funded by a Peter MacCallum Foundation grant, a Tolmar-ANZUP fellowship grant, and National Health and Medical Research Council grants (No. 2003115, 2003887).

    • Competing interests Grant support to the institution for unrelated work from Merck Sharp & Dohme, Merck Serono, Amgen, Pfizer, AstraZeneca, Endocyte and AAA (a Novartis company), Genentech and Pfizer, Bayer, Janssen. Honorarium to the institution for consulting/advice from AstraZeneca, Merck Sharp & Dohme, Merck Serono, Roche/Genetech, Bristol Myers Squibb and Novartis, Ipsen, Ferring, Clovis, Bayer, Janssen, Roche, Atellas, Pfizer.

    • Provenance and peer review Not commissioned; externally peer reviewed.

    • Supplemental material This content has been supplied by the author(s). It has not been vetted by BMJ Publishing Group Limited (BMJ) and may not have been peer-reviewed. Any opinions or recommendations discussed are solely those of the author(s) and are not endorsed by BMJ. BMJ disclaims all liability and responsibility arising from any reliance placed on the content. Where the content includes any translated material, BMJ does not warrant the accuracy and reliability of the translations (including but not limited to local regulations, clinical guidelines, terminology, drug names and drug dosages), and is not responsible for any error and/or omissions arising from translation and adaptation or otherwise.