Categories
Prostate cancer

Single cell analysis of cribriform prostate cancer reveals cell intrinsic and tumor microenvironmental pathways of aggressive disease

Altered epithelial and microenvironmental cell types in prostate ICC/IDC

To comprehensively analyze all cell types in ICC/IDC and the associated TME, paired benign-enriched and ICC/IDC-enriched prostate tissue from RP were isolated from 7 patients for scRNAseq (Fig. 1a, b). Benign-enriched and ICC/IDC-enriched regions were verified by obtaining a rapid frozen H&E of prostate tissue for histologic examination prior to processing (Fig. 1b and Supplementary Fig. 1). Overall, patients had Grade Group 2-5 prostate cancer that was either stage pT3aN0/X or pT3bN0 (Fig. 1c and Supplementary Table 1). FFPE sections from RP were examined by immunohistochemistry (IHC) for High Molecular Weight Cytokeratin (HMWCK), TP63, AMACR, AR, ERG, and PTEN (Fig. 1d, e and Supplementary Fig. 2a). In 6 patients, IDC was intermixed with ICC to varying proportions as determined by IHC staining for TP63 (Fig. 1d and Supplementary Fig. 2a). Cancer glands in all patients stained positive for AMACR (Fig. 1d, e and Supplementary Fig. 2a). Although ERG genomic rearrangements have not been previously associated with ICC/IDC16, ERG overexpression by IHC was detected in ICC/IDC from 5 patients (Fig. 1d, e and Supplementary Fig. 2a). Interestingly, ERG was overexpressed in Gleason pattern 3 from an additional patient but not in adjacent ICC/IDC. Consistent with prior findings25,26, homogenous PTEN loss by IHC was detected in 5 of the 6 patients with IDC (83%).

Fig. 1: ScRNAseq of ICC/IDC-enriched and benign-enriched prostate.
figure 1

a Schematic of scRNAseq protocol of ICC/IDC-enriched and benign-enriched prostate. b Representative rapid frozen H&E of benign-enriched and ICC/IDC-enriched prostate isolated for scRNAseq at 100x, bar = 200 µm, (n = 7 biologically independent samples). This image and the six additional representative images are in Supplementary Fig. 1. c Patient clinical characteristics. d Representative HMWCK (High Molecular Weight Cytokeratin), TP63, AMACR, AR, ERG, and PTEN expression by IHC on patient FFPE prostate tissue from RP at 100x, bar = 100 µm, (n = 7 biologically independent samples). This image and the six additional representative images are in Supplementary Fig. 2a. e AMACR, AR, ERG, and PTEN expression by IHC per patient. f, g The number (f) and percent (g) of benign-enriched and ICC/IDC-enriched cells per patient. h, i Unsupervised graph-based clustering of all samples visualized by UMAP delineated by cluster (h) and cell-type (i). j Bubble plot of representative cell-type specific markers across all clusters. Source data are provided as a Source Data file.

To facilitate analyses of all cell types found in the prostate, including cell types that were less abundant or difficult to isolate, cells were sorted for live cells and broad cell types (immune, epithelial, and other) by flow cytometry and then recombined for scRNAseq. Following tissue isolation and histological confirmation by H&E, paired benign-enriched and ICC/IDC-enriched samples were single-cell disassociated, stained with DAPI and antibodies against CD45 (pan-immune marker) and EpCAM (pan-epithelial marker), tagged for multi-plex sequencing, and flow-sorted into three DAPI (live) populations that were CD45+ (immune cells), EpCAM+ (epithelial cells), or CD45EpCAM (cells other than immune and epithelial cells such as endothelial cells, smooth muscle cells, fibroblasts, and nerve cells). DAPI cells from benign-enriched and ICC/IDC-enriched tissue were then recombined at a ratio of 30% CD45+, 25% EpCAM+, and 45% CD45EpCAM (Fig. 1a). Benign-enriched and ICC/IDC-enriched cells were then mixed at a 30:70 ratio, respectively, for scRNAseq and TCR VDJ sequencing using 10X genomics (Fig. 1a). After filtering low-quality cells and doublets, over 57,000 cells in total from 7 patients were analyzed by scRNAseq and over 15,000 of these cells were additionally analyzed by TCR sequencing (Fig. 1f, g and Supplementary Tables 25). The mean number of total cells analyzed per patient was 8242.

Unsupervised graph-based clustering and accompanying visualization with the Uniform Manifold Approximation and Projection (UMAP) algorithm yielded 26 clusters encompassing multiple cell types, including immune, endothelial, SMC, fibroblasts, and epithelial (Fig. 1h–j). Sample contribution to each cluster was variable, but most clusters were derived from a relatively even distribution of samples (Supplementary Fig. 3a, b, Supplementary Tables 6 and 7). Within the epithelial clusters, cluster 6 was significantly increased in ICC/IDC-enriched tumors compared to benign-enriched prostate, while clusters 12 and 21 were significantly decreased (Fig. 2a–f, and Supplementary Fig. 3b). Amongst the non-immune TME (CD45/EpCAM), cluster 22 (endothelial cells) was increased in ICC/IDC-enriched tumors compared to benign-enriched prostate (Fig. 2d).

Fig. 2: Epithelial and microenvironmental cell types were altered in prostate ICC/IDC.
figure 2

a, b Unsupervised graph-based clustering of all samples visualized by UMAP delineated by benign-enriched (a) and ICC/IDC-enriched (b) prostate. ce Percent total of EpCAM+ (c), EpCAM/CD45 (d), and CD45+ (e) benign-enriched and ICC/IDC-enriched prostate cells per cluster. Graphs shown as mean ± SEM and analyzed by Wilcoxon matched-pair signed rank two-tailed test; n = 7 biologically independent samples. The partial graph (endothelial cells) in d is also shown in Fig. 4 (g). f Percent of cell type analyzed per patient delineated by benign-enriched and ICC/IDC-enriched prostate. Source data are provided as a Source Data file.

Increased SCHLAP1 and JAG1 in prostate ICC/IDC

The heterogeneity of ICC/IDC cells has not been well established. Unsupervised graph-based clustering of all cells generated 7 epithelial clusters: 2, 5, 6, 10, 11, 12, and 21, as well as a small ciliated epithelial cluster (cluster 25) (Figs. 1j and 3a). Clusters 12 and 21 were significantly decreased in ICC/IDC-enriched tumors compared to benign-enriched prostate (Fig. 2c). While benign-enriched prostate cells from all patients contributed to clusters 12 and 21, fewer cells were from ICC/IDC-enriched tumors (Fig. 2c and Supplementary Fig. 3a, b). Clusters 12 and 21 were positive for acinar luminal epithelial markers (MSMB) but were negative for cancer cell markers (ERG and AMACR) (Fig. 3b and Supplementary Fig. 4a, genes for clusters 12 and 21 in Source Data)36. While cluster 12 was positive for AR and AR-induced genes (KLK3), cluster 21 had diminished expression of these genes (Fig. 3b). These findings support that both the AR high, and AR low populations of benign luminal epithelial cells were decreased in the ICC/IDC TME.

Fig. 3: Increased SCHLAP1 and JAG1 in prostate ICC/IDC.
figure 3

a Unsupervised graph-based clustering of epithelial cell clusters (2, 5, 6, 10, 11, 12, 21) separated by benign-enriched and ICC/IDC-enriched prostate. b, c Violin (b) and feature (c) plots of gene expression in epithelial cell clusters. d, e Representative images at 20x, bar = 10 µm (d) and quantification (e) of SCHLAP1 and JAG1 expression by RNAscope in ICC1-7 at RP for benign prostate luminal epithelial cells, Gleason pattern 3 prostate cancer, and ICC (n = 7 biologically independent samples). Quantification of SCHLAP1 and JAG1 in an extended validation RP cohort of benign prostate luminal epithelial cells (n = 20) as well as Gleason pattern 3 (n = 18), Gleason pattern 4 non-ICC (NC) (n = 14), ICC/IDC (n = 11), and Gleason pattern 5 (n = 6) prostate cancer. A total of n = 23 biologically independent samples were assessed for SCHLAP1 and JAG1 expression with samples having more than one histology. Quantification of SCHLAP1 and JAG1 expression by H-score (intensity x percent expression). Graphs are shown as mean ± SEM and analyzed by one-way ANOVA with Tukey’s multiple comparisons. f PssGSEA of hallmark pathways altered in ICC/IDC cancer cells (clusters 5, 6, and 11) compared to benign luminal epithelial cells in cluster 12. g Kaplan–Meier and log-rank test of progression-free survival in the TCGA PanCancer Atlas prostate adenocarcinoma for JAG1 by median expression (n = 492). h Percent of patient cells per cluster after re-clustering clusters 5, 6, and 11 into 7 clusters (CRIB-0 through CRIB-6). i, j UMAP visualization of re-clustering of clusters 5, 6, and 11 (CRIB-0 through CRIB-6) color-coded by cluster (i) and by patient (j). k Violin plots of gene expression in CRIB-0 through CRIB-6. l PssGSEA of hallmark pathways altered in benign epithelial clusters in ICC/IDC-enriched prostate compared to benign-enriched prostate. Source data are provided as a Source Data file.

In contrast to benign luminal epithelial cells, clusters 5, 6, and 11 were principally composed of cells from ICC/IDC-enriched prostate with very minimal contribution from benign-enriched prostate. Clusters 5, 6, and 11 expressed luminal epithelial markers, AR, AR-induced genes, and cancer cell markers (ERG and AMACR), supporting their identity as cancer cells (Fig. 3b, c). ICC/IDC-enriched regions sampled for scRNAseq contained varying levels of ICC, IDC, Gleason pattern 4 non-ICC, and Gleason pattern 3 (Supplementary Fig. 4b, c). Collectively, data suggest clusters 6 and 11 were enriched for ICC/IDC cells while cluster 5 was enriched for Gleason pattern 3 cells and potentially a subset of Gleason pattern 4 non-cribriform cells. Cluster 11 was spatially distinct from clusters 5 and 6, and nearly all cells in cluster 11 were from one patient (ICC4) whose sampled region for scRNAseq was predominantly ICC with minimal to no adjacent IDC, Gleason pattern 4 non-ICC, or Gleason pattern 3, thereby supporting ICC as the principal cellular identity of cluster 11. Conversely, cluster 5 predominantly consisted of cells from 5 patients (ICC1, ICC2, ICC3, ICC5, and ICC6), all of whom had adjacent Gleason pattern 3 (>5%) in the region isolated for scRNAseq, whereas the other 2 patients (ICC4 and ICC7) had minimal contribution to cluster 5 and had minimal Gleason pattern 3 (<5%) detected in their isolated tissue (Supplementary Fig. 4b–d). Cells in cluster 5 were ERG positive, which is consistent with isolated tissue for scRNAseq containing adjacent Gleason pattern 3 (ICC1, ICC2, ICC3, ICC5, and ICC6) but not with ICC/IDC which was ERG negative in ICC/IDC from 2 of the 7 patients (Figs. 1e and 3c). Cluster 6 was composed of cells from all patients and consistent with ICC/IDC, had ERG+ and ERG subpopulations (Fig. 3c). Compared to cluster 5, clusters 6 and 11 had increased expression of FOLH1, which has been shown to be overexpressed in ICC (Fig. 3b)37. Similarly, SCHLAP1, a lncRNA associated with ICC/IDC and adverse outcomes20,31,38,39, was increased in clusters 6 and 11 compared to cluster 5 and benign epithelial cells (Fig. 3b). RNAscope of RP tissue from ICC1-7 and an additional extended independent cohort showed increased SCHLAP1 expression in ICC/IDC compared to benign prostate epithelial cells and Gleason pattern 3, Gleason pattern 4 non-ICC (NC) and Gleason pattern 5 prostate cancer (Fig. 3d, e). Collectively, pathology and gene expression data support that clusters 6 and 11 were enriched for ICC/IDC while cluster 5 was likely enriched for Gleason pattern 3 prostate cancer.

Compared to benign luminal epithelial cells, cancer cells in all clusters (5, 6, and 11) were enriched for potential therapeutic targets and/or biomarkers, including FOLH1 (PSMA)40,41,42 and PCA343,44 (Fig. 3b). APOD, an oxidative stress response gene increased in ETS+ prostate cancers45, and CD276 (B7-H3), an immune checkpoint associated with adverse prostate cancer outcomes46, were also elevated in cancer cell clusters (Fig. 3b). Single-sample gene-set enrichment analysis with paired comparisons (pssGSEA) was used to test for enrichment of hallmark pathways. PssGSEA indicated that the MYC Targets VI hallmark was increased in clusters 5, 6, and 11, suggesting that this alteration may be common among several Gleason patterns of prostate cancer (Fig. 3f). In support, elevated MYC expression was detected in both ICC/IDC and Gleason pattern 3 prostate cancer by IHC (Supplementary Fig. 2b). In contrast, the TNFα signaling via NFκB hallmark was increased in ICC/IDC-enriched cells in cluster 6 compared to benign luminal epithelial cells. Of the top 5 ranked TNFα signaling via NFκB hallmark genes, JAG1, a Notch ligand correlated with prostate cancer metastasis, angiogenesis, and reactive stroma formation47,48,49, was distinctly increased in clusters 6 and 11 compared to clusters 5 and 12 (Fig. 3b and Supplementary Fig. 4e). RNAscope of RP tissue from ICC1-7 and from an extended independent validation cohort confirmed higher JAG1 expression in ICC/IDC compared to benign luminal epithelial cells, Gleason pattern 3, and Gleason pattern 4 non-ICC (NC) prostate cancer (Fig. 3d, e). Increased JAG1 showed a significant, but modest association with worse prostate cancer progression-free survival in the TCGA PanCancer Atlas prostate adenocarcinoma cohort (Fig. 3g).

To define further the heterogeneity of prostate ICC/IDC cells, clusters 5, 6, and 11 were re-clustered, yielding 7 distinct clusters: CRIB-0 through CRIB-6 (Fig. 3h–k, Supplementary Fig. 4f, genes for CRIB0-6 in Source Data). Cells predominantly clustered by patient except for cells from ICC1 which were split between two clusters (Fig. 3h–j). Patient-based clustering did not occur in benign epithelial clusters (clusters 2, 10, 12) when re-clustered individually using similar parameters (Supplementary Fig. 4g). CD276 was expressed in all clusters while SCHLAP1 and JAG1 expression was heterogenous between clusters with high expression in most patients (Fig. 3k). Collectively, these findings support that ICC/IDC cancer cells have high inter-patient heterogeneity, but commonly upregulate SCHLAP1 and TNFα signaling via NFĸB pathway member JAG1.

Increased inflammatory pathways in benign epithelial cells in the ICC/IDC TME

How the development of ICC/IDC impacts adjacent benign epithelial cells in the TME is not fully known. In addition to luminal epithelial cells, adult human prostate consists of several other cell types, including basal cells, rare neuroendocrine (NE) cells, and the recently described club and hillock cells50. Clusters 10 and 2 consisted of cells from all patients, and the relative abundance of clusters 10 and 2 were not significantly altered in ICC/IDC-enriched tumors compared to benign-enriched prostate (Fig. 2c and Supplementary Fig. 3a). Overall, both Clusters 10 and 2 had low expression of prostate cancer-associated genes, AR, and AR-induced genes (Fig. 3b and Supplementary Fig. 4a). Cluster 10 was enriched for KRT5+TP63+ cells with distinct subclusters of KRT5+ cells enriched for either KRT14+ basal cells or KRT13+ hillock cells (Fig. 3b, c and Supplementary Fig. 4a)50. Similar cell type heterogeneity was detected in cluster 2. A subcluster of cells in cluster 2 was enriched for KRT7 and RARRES1, two markers of ductal luminal epithelial cells36, while a distinct small subcluster of cells expressed club cell markers, SCGB1A1 and SCGB3A1 (Fig. 3b, c, Supplementary Fig. 4a, genes for cluster 2 in Source Data). Consistent with young adult benign prostate50, KRT13+ (hillock) and SCGB1A1+ (club) cells were infrequently clustered in benign epithelial glands and were only rarely interspersed among cancer cells (Supplementary Fig. 5). Interestingly, cluster 2 also contained rare cells that strongly expressed NE markers (CHGA, SCG2, ASCL1, and GRP)50, thereby, supporting cluster 2 identity as a heterogenous cluster of largely ductal luminal epithelial cells and club cells but also rare NE cells (Fig. 3c)50.

PssGSEA showed increased androgen response in cluster 2 (club/ductal luminal) and cluster 10 (basal/hillock) cells from ICC/IDC-enriched regions compared to benign-enriched regions (Fig. 3l), which is consistent with a recent report showing increased androgen response in prostate cancer-associated club and basal cells compared to the normal club and basal cells, respectively51. Thus, increased androgen response in prostate cancer-associated club and basal cells may be common across multiple prostate cancer subtypes. Interestingly, club/basal/hillock cells in ICC/IDC-enriched regions had increased inflammatory hallmarks, including TNFα signaling via NFκB, IFNγ response, and IL6/JAK/STAT3 signaling compared to these cell types in benign-enriched prostate (Fig. 3l). Inflammatory hallmarks were also increased in benign luminal epithelial cells (clusters 12 and 21) from ICC/IDC-enriched regions compared to benign-enriched prostate (Fig. 3l). Collectively, these findings support that benign epithelial cells in the ICC/IDC TME had differential gene expression reflecting increased inflammatory response and signaling compared to these cell types in the benign prostate environment.

Increased JAG1/NOTCH signaling and angiogenesis in prostate ICC/IDC

How ICC/IDC impacts non-epithelial cells in the TME has not been well established. JAG1 is a cell surface ligand that activates NOTCH receptors through cell-to-cell contact with adjacent cells. Due to elevated JAG1 in ICC/IDC cancer cells, NOTCH signaling may be increased in cells directly adjacent to ICC/IDC. Expression analyses revealed that NOTCH receptors were distinctly enriched in PECAM1+ endothelial cells (clusters 0 and 22) and in BCAM+ vascular SMC (cluster 13) (Fig. 4a–d, Supplementary Fig. 6a, b) in both the ICC/IDC TME and the benign prostate microenvironment. Specifically, NOTCH4 was highly expressed by endothelial cells in clusters 0 and 22, NOTCH1 was also expressed by endothelial cells in cluster 22, and NOTCH3 was expressed by vascular SMC (cluster 13) (Fig. 4d). NOTCH2 was enriched in cluster 10 (hillock/basal) cells from benign-enriched prostate. Consistent with elevated JAG1 in ICC/IDC cancer cells, NOTCH target genes were significantly increased in endothelial cells in clusters 0 (HES1) and 22 (HES1 and HEY1) and SMC (HES4) located the ICC/IDC TME compared to benign prostate (Fig. 4e).

Fig. 4: Increased JAG1/NOTCH signaling and angiogenesis in the prostate ICC/IDC TME.
figure 4

a Unsupervised graph-based clustering of all samples visualized by UMAP highlighted for endothelial clusters 0 and 22 and SMC cluster 13 delineated by benign-enriched and ICC/IDC-enriched prostate. b Violin plots of endothelial, blood, and lymphoid marker expression in clusters 0 and 22. c Violin plots of SMC and pericyte markers in clusters 13 and 20. d Violin plots of NOTCH receptor expression in clusters 0–25. e Violin plots of NOTCH-induced genes in endothelial clusters 0 and 22 and SMC cluster 13 delineated by benign-enriched and ICC/IDC-enriched prostate. f Violin plots of markers differentially enriched in cluster 0 compared to cluster 22. g Percent total of EpCAM/CD45 benign-enriched and ICC/IDC-enriched prostate cells per endothelial cluster (0 and 22). The graph shown as mean ± SEM and analyzed by Wilcoxon matched-pair signed rank two-tailed test; n = 7 biologically independent samples. The graph is also shown as part of Fig. 2d. h PssGSEA of hallmark pathways in ICC/IDC-enriched prostate compared to benign-enriched prostate in clusters 0 and 22. i Violin plots of markers in ICC/IDC-enriched prostate compared to benign-enriched prostate in clusters 0 and 22. j UMAP visualization of cluster 13 after re-clustering. k Percent of cells in clusters 13–0 through 13–7. The graph is shown as mean ± SEM and analyzed by Wilcoxon matched-pair signed rank two-tailed test; n = 7 biologically independent samples. l Violin plots of markers in clusters 13–0 through 13–3. m Representative images of CD31 expression by IHC in ICC/IDC prostate cancer from RP (n = 7 biologically independent samples) at 100x (bar = 100 µm) and 400x. Source data are provided as a Source Data file.

Marker gene expression analysis was used to delineate endothelial identity in clusters 0 and 22. Both endothelial clusters were positive for blood markers (BCAM) and had a minimal expression of lymphatic markers (PROX1), supporting that they predominantly consisted of blood endothelial cells (Fig. 4b and Supplementary Fig. 6a). Cells within cluster 0 were characterized by the expression of endothelial markers found in postcapillary veins (ACKR1, VWF), immature cells (PLVAP, IGFBP4), and quiescent cells (SPARCL1), while cells within cluster 22 were characterized by the expression of markers found in arteries (CXCL12, ENPP2), capillaries (ICAM2, IFI27, TIMP3), and immature (A2M, SLC9A3R2, CRIP2) endothelial cells (Fig. 4f and Supplementary Fig. 6c, d, and genes for clusters 0 and 22 in Source Data)52. The abundance of endothelial cells in cluster 22, but not cluster 0, was significantly increased in ICC/IDC-enriched regions compared to benign-enriched regions (Fig. 4g). PssGSEA showed increased angiogenesis and markers of hypoxia in endothelial cells from ICC/IDC-enriched regions compared to benign-enriched prostate (Fig. 4h, i).

Consistent with increased endothelial cells and NOTCH signaling, scRNAseq analyses support that ICC/IDC-enriched regions also had increased vascular SMC. Cluster 13 expressed vascular SMC (BCAM) and pericyte (RGS5) markers (Fig. 4c). To assess for differences in these cell types between ICC/IDC-enriched and benign-enriched prostate, cluster 13 was re-clustered into 8 clusters (clusters 13–0 through 13–7) (Fig. 4j, genes for 13-0 to 13-7 in Source Data). Clusters 13–4 through 13–7, however, were small or were contributed to by only a minority of patients. Cluster 13-0 was significantly enriched for pericyte markers, while clusters 13–1 and 13–3 were significantly enriched for the expression of several transcription factors (JUN and ATF3) (Supplementary Fig. 6e). While the relative abundance of clusters 13–0, 13–1, and 13–3 were similar, cluster 13-2 was significantly increased in ICC/IDC-enriched regions compared to benign-enriched regions (Fig. 4k). Cluster 13–2 was enriched for multiple vascular SMC genes as well as the NOTCH target gene HES4 (Fig. 4l).

IHC for CD31 (PECAM1) on RP sections indicated that ICC/IDC foci were associated with adjacent external vessels (Fig. 4m). However, some ICC/IDC foci had limited tumor endothelial cell (TEC) infiltration, but consistent with histologic features diagnostic of cribriform, the majority of intraglandular cells were not in contact with stroma53. Collectively, these findings support a model in which increased JAG1 expression in ICC/IDC cancer cells induced angiogenesis through NOTCH signaling in vascular endothelial and SMC cells.

CAFÉ CAF are enriched in ICC/IDC and are associated with worse outcomes

JAG1-NOTCH2 signaling between breast cancer cells and fibroblasts was shown to impact CAF phenotypes54, however, minimal expression of NOTCH and NOTCH-induced genes by CAF in the ICC/IDC TME indicates that alternative mechanisms drive their activation (Fig. 5a, b). Instead, ligand/receptor analyses suggest that increased PDGFA and FGF13 expression by ICC/IDC cancer cells (cluster 6) may impact CAF phenotypes through PDGFRα and FGFR1 in fibroblasts (cluster 20)55 (Fig. 5c, d).

Fig. 5: CAFÉ CAF are enriched in ICC/IDC and are associated with worse outcomes.
figure 5

a Violin plots of fibroblast marker expression in clusters 13 and 20. b PssGSEA of hallmark pathways altered in ICC/IDC-enriched compared to benign-enriched prostate in clusters 13 and 20. c Violin plots of ligand expression in epithelial clusters. d Violin plots of receptor expression in clusters 13 and 20. e Violin plots of peri-epithelial (APOD) and interstitial (C7) fibroblast markers in ICC/IDC-enriched compared to benign-enriched prostate in cluster 20. f UMAP visualization of unsupervised graph-based re-clustering of cluster 20 (F0–F3) separated by benign-enriched and ICC/IDC-enriched prostate. g Cell percentage per cluster from ICC/IDC-enriched and benign-enriched regions after re-clustering cluster 20 (F0–F3). The graph is shown as mean ± SEM and analyzed by Wilcoxon matched-pair signed rank two-tailed test; n = 7 biologically independent samples. h Violin plots of marker gene expression in clusters F0-F3. i, j, k Violin plots of CAFÉ CAF gene expression in cluster 20 differentiated by ICC/IDC-enriched and benign-enriched prostate (i, k) and in clusters F0–F3 (j, k). l Kaplan–Meier and log-rank test of progression-free survival in the TCGA PanCancer Atlas Prostate Adenocarcinoma (n = 492) for the CAFÉ CAF signature. Kaplan–Meier and log-rank test of disease-free survival in the MSKCC Prostate Adenocarcinoma for the CAFÉ CAF signature (n = 131). m, n Representative images at 400x, bar = 10 µm (m) and quantification (n) of expression in a combined RP prostate cancer cohort of new and historical33 samples adjacent to benign prostate (CTHRC1 n = 17, ASPN n = 40, FAP n = 27, and ENG n = 26), Gleason pattern 3 (CTHRC1 n = 6, ASPN n = 24, FAP n = 15, and ENG n = 16), Gleason pattern 4 non-ICC (G4 NC; CTHRC1 n = 9, ASPN n = 24, FAP n = 15, and ENG n = 15), ICC/IDCC (CTHRC1 n = 7, ASPN n = 21, FAP n = 12, and ENG n = 12), and Gleason pattern 5 (G5; CTHRC1 n = 6, ASPN n = 6, FAP n = 6, and ENG n = 6) prostate cancer. N = 42 biologically independent samples were used to assess CTHRC1, ASPN, FAP, and ENG expression with samples having more than one histology for assessment and overlap between markers. Graphs are shown as mean ± SEM and analyzed by one-way Anova with Tukey’s Multiple Comparisons. Source data are provided as a Source Data file.

Multiple studies have begun to elucidate fibroblast heterogeneity in several cancer types, including breast and pancreatic cancer33,56,57,58,59,60. A recent study identified two fibroblast subtypes in benign human prostate: an APOD+ peri-epithelial subtype and a C7+ interstitial subtype61; however, CAF heterogeneity has not been fully delineated in prostate cancer33. ICC/IDC-enriched CAF in cluster 20 were significantly increased for APOD and significantly decreased for C7 expression compared to benign-enriched fibroblasts (Fig. 5e). To determine if peri-epithelial fibroblasts were indeed the fibroblast subtype enriched in ICC/IDC regions, cluster 20 was re-clustered into four clusters (F0-F3) (Fig. 5f, g, genes for F0-F3 in Source Data). Cluster F0 was delineated by higher APOD expression and was significantly increased in ICC/IDC-enriched regions compared to benign-enriched prostate (Fig. 5f–h). In contrast, cluster F1 was marked by higher C7 expression and was significantly decreased in ICC/IDC-enriched regions compared to benign-enriched prostate.

CAF from ICC/IDC-enriched regions examined both prior to and after re-clustering were significantly elevated for the expression of genes associated with adverse pathology, poor outcomes, and/or immunosuppression, including TNC62, TGFB163, SFRP464, CCL265, CTHRC166,67, ASPN33,68,69, FAP34,35,58,70, and ENG71,72,73 (Fig. 5i, j and Supplementary Fig. 7a). A 4-gene signature based on ICC/IDC CAF markers: CTHRC1, ASPN, FAP, and ENG (CAFÉ CAF), showed a significant association with worse prostate cancer progression-free survival in the TCGA PanCancer Atlas prostate adenocarcinoma cohort74 and worse disease-free survival in the MSKCC Prostate Adenocarcinoma cohort75 (Fig. 5k, l). CTHRC1+, ASPN+, FAP+, and ENG+ CAF spatial dynamics and associations with other prostate cancer grades and/or histological subtypes were examined by RNAscope in an independent extended RP prostate cohort of combined new and historical samples33. CTHRC1+, ASPN+, and FAP+ CAF were located peri-epithelial to ICC/IDC and were significantly enriched in ICC/IDC compared to benign prostate as well as Gleason pattern 3 and Gleason pattern 4 non-ICC prostate cancer (Fig. 5m, n and Supplementary Fig. 7c, d). While CTHRC1+ CAF were slightly elevated, ASPN+ and FAP+ CAF were comparable between ICC/IDC and Gleason pattern 5 prostate cancer. ENG+ CAF were significantly elevated in all cancer grades/histological subtypes examined compared to benign prostate. These results support that CTHRC1+, ASPN+, and FAP+ CAF were increased in ICC/IDC and Gleason pattern 5 prostate cancer, while ENG+ CAF were increased in cancer. Collectively, these findings support that CAF in the ICC/IDC TME express peri-epithelial fibroblasts markers, have common gene expression as CAF adjacent to Gleason pattern 5 prostate cancer, and are associated with worse outcomes.

Immune exclusion and reduced T cell fraction and clonality in the prostate ICC/IDC TME

Expansion of a CAF subtype expressing immunosuppressive markers, including FAP34,35 suggests that the ICC/IDC TME may be associated with dysfunctional T cells; however, little has been reported about the immune TME associated with prostate ICC/IDC. To better determine the immune repertoire and heterogeneity in the ICC/IDC TME compared to benign regions, CD45+ cells were analyzed by scRNAseq and T cells were analyzed by TCR sequencing. Flow cytometry analysis prior to sequencing showed a significant decrease in CD45+ cells in ICC/IDC-enriched regions compared to benign-enriched regions (Fig. 6a, b). Normalization of CD45+, EpCAM+, and CD45/EpCAM fractions detected by flow cytometry prior to sequencing with the number of TCR+ cells after sequencing, indicated that within the immune fraction, significantly fewer T cells were detected in ICC/IDC-enriched tumors compared to benign-enriched prostate (Fig. 6c, d). In addition to fraction, T cells in the ICC/IDC TME were examined for diversity by analyzing clonotype richness (percent of different clonotypes) and evenness (percent distribution of each clonotype) by Simpson clonality. Simpson clonality was significantly decreased in ICC/IDC-enriched compared to benign-enriched prostate, thereby indicating a more even distribution of clonotypes in ICC/IDC-enriched regions (Fig. 6e, f). Richness (percent of different clonotypes), however, was similar between ICC/IDC-enriched and benign-enriched prostate (T cell richness in Source Data). Differences in TCR clonotype repertoire were detected and approximately 5-15% of clonotypes had a two-fold or greater expansion/contraction in ICC/IDC-enriched compared to benign-enriched regions (Fig. 6g, h). These data support that ICC/IDC-enriched prostate had diminished immune infiltration, and of the immune cells, the T cell fraction was reduced and had decreased clonality.

Fig. 6: Immune exclusion, reduced T cell fraction and clonality, and increased T cell dysfunction in the prostate ICC/IDC TME.
figure 6

a Quantification of percent CD45+ cells by flow cytometry from paired samples of benign-enriched and ICC/IDC-enriched prostate. Graph shown as mean ± SEM analyzed by paired two-tailed t-test; n = 4 biologically independent samples. FACS gating strategies shown in Supplementary Fig. 10a. b Representative images of inflammatory cells (red arrow) in benign-enriched regions and IDC/ICC-enriched regions by H & E at 100x, bar = 50 µm, (n = 7 biologically independent samples). c, d Percent T cells of the immune fraction by individual patient (c) and collectively (d) for benign-enriched and ICC/IDC-enriched prostate. Graph in d shown as mean ± SEM analyzed by Wilcoxon matched-pair signed rank two-tailed test; n = 7 biologically independent samples. e, f Simpson Clonality by individual patient (e) and collective (f) for benign-enriched and ICC/IDC-enriched prostate. Graph in f shown as mean ± SEM analyzed by Wilcoxon matched-pair signed rank two-tailed test; n = 7 biologically independent samples. g Percentage of T cells contracted or expanded between benign-enriched and ICC/IDC-enriched prostate. h Clonotype frequency between benign-enriched and ICC/IDC-enriched prostate. i Unsupervised graph-based clustering of all samples visualized by UMAP highlighted for T cell clusters 1, 3, 7, 9, 17, and 23 delineated by benign-enriched and ICC/IDC-enriched prostate. j, k Violin (j) and feature plots (k) of clusters 1, 3, 7, 9, 17, and 23 for immune and T cell markers. l Violin plots of markers in ICC/IDC-enriched prostate compared to benign-enriched prostate in clusters 1, 3, 7, 9, 17, and 23. m PssGSEA of hallmark pathways in ICC/IDC-enriched prostate compared to benign-enriched prostate in T cell clusters. n Pseudotime trajectory analysis for clusters 1, 3, 7, 9, 17, and 23. o Representative clonotype cluster location in benign-enriched and ICC/IDC-enriched prostate and quantification. The graph is shown as mean ± SEM analyzed by Wilcoxon matched-pair signed rank two-tailed test; n = 7 biologically independent samples. Source data are provided as a Source Data file.

Increased dysfunctional markers in CD8+ T cells in the prostate ICC/IDC TME

Recent single-cell analyses have provided insight into the substantial heterogeneity in intratumoral T cell states that likely exist along a continuum76,77,78. Due to the lack of a consensus nomenclature for human T cell states analyzed by scRNAseq, Van der Leun et al. integrated multiple scRNAseq studies to broadly categorize CD8+ T cell states as naïve-like, predysfunctional (effector memory, memory, and transitional), cytotoxic (effector), and dysfunctional (exhausted)77. Unsupervised graph-based clustering of all cells generated six CD45+CD3+ clusters with contributions from all patients (Fig. 6i, k and Supplementary Fig. 3a). Mapping showed a heterogenous cluster (cluster 1) that expressed naïve-like markers including CD3, IL7R, CCR7, and SELL, which were similar between ICC/IDC-enriched and benign-enriched T cells (Fig. 6j and Supplementary Fig. 8a, genes for cluster 1 in Source Data)76,77,78,79.

CD8 + T cells mapped to 5 of the 6 T cell clusters: 3, 7, 9, 17, and 23 (Fig. 6j, k). Both ICC/IDC-enriched and benign-enriched CD8+ T cells in clusters 3, 7, and 17 expressed the predysfunctional marker GZMK (Supplementary Fig. 8a). Compared to clusters 3 and 17, CD8+ T cells in cluster 7 expressed lower levels of granzymes and perforin but were notable for high levels of human stress-activated protein (Supplementary Fig. 8a). ICC/IDC-enriched CD8+ T cells in cluster 7 had decreased expression of IFNG and increased expression of PDCD1 (Fig. 6l, DEG analysis for cluster 7 in Source Data). Cluster 3 cells expressed PRF1, GZMA, GZMH, TNF, and IFNG; however, expression of IFNG and TNF were lower in ICC/IDC-enriched compared to benign-enriched CD8+ T cells (Fig. 6l and Supplementary Fig. 8a)77. CD8+ T cells in cluster 17 were distinguished by high CCL4 expression and additionally expressed GZMA, GNLY, PRF1, TNF, and IFNG suggesting that cluster 17 cells fell on the spectrum toward cytotoxic CD8+ T cells. Cluster 17 CD8+ T cells from ICC/IDC-enriched prostate had increased expression of dysfunctional markers PDCD1 and LAG3 and decreased TNF and IFNG expression (Fig. 6l, DEG analysis for cluster 17 in Source Data). Like cluster 17 cells, cluster 9 cells also expressed GZMA, GZMB, GZMH, GNLY, and PRF1. CD8+ T cells in ICC/IDC-enriched prostate had a higher expression of PDCD1 and LAG3 and decreased IFNG, thereby suggesting that cluster 9 CD8+ T cells in benign-enriched regions were more cytotoxic while cluster 9 CD8+ T cells in ICC/IDC-enriched regions were more dysfunctional (Fig. 6l, DEG analysis for cluster 9 in Source Data). PssGSEA indicated that ICC/IDC-enriched T cells in cluster 9 had decreased TNFα signaling via NFκB, thereby supporting their reduced effector activity in the TME (Fig. 6m). Cells in cluster 23 also expressed markers of dysfunctional CD8+ T cells (PDCD1 and LAG3) that were slightly higher in ICC/IDC-enriched T cells. Consistent with dysfunctional cells, cluster 23 cells expressed lower levels of granzymes, TNF, and IFNG (Fig. 6l and Supplementary Fig. 8a). Pseudotime trajectory analysis showed progression of cells from naïve cells in cluster 1 to predysfunctional cells in clusters 7/3/17, to cytotoxic cells in cluster 9 and lastly to dysfunctional cells in cluster 23 (Fig. 6n). Mapping of the top 20 TCR clonotypes showed a significant shift in several CD8+ T clonotypes found in benign-enriched clusters 3, 7, and/or 17 to cluster 9 in ICC/IDC-enriched prostate (Fig. 6o and Supplementary Fig. 9a–c). Overall, these findings support that CD8+ T cells in the ICC/IDC TME expressed decreased effector cytokines and increased dysfunctional markers.

In contrast to CD8+ T cells, CD4+ T cells mapped to three of the six T cell clusters. CD4+ T cells mapped to clusters 1, 7, and 23 (Fig. 6j, k). Treg markers (CD4, FOXP3, IL2RA, CTLA4) mapped to a distinct subset of these cells in cluster 1 as well as in cluster 23 (Fig. 6k and Supplementary Fig. 8b). These findings support heterogeneity in CD4+ Treg cells in the prostate with some having features closer to naïve T cells while others having features closer to dysfunctional T cells. Expression of TIGIT was increased in cluster 1 T cells in ICC/IDC compared to benign-enriched prostate (Supplementary Fig. 8a, b). Collectively, these findings indicate that ICC/IDC-enriched tumors had fewer infiltrating immune cells, and of the immune cells, the T cell fraction was lower, had less clonality, and had higher expression of exhausted markers compared to benign-enriched prostate.

Increased C1QB
+
TREM2
+
APOE
+ M2 macrophages in prostate ICC/IDC TME

While our data indicate that T cells were largely excluded or suppressed, it is not known if myeloid cells also contribute to a pro-tumorigenic immune microenvironment in ICC/IDC. CTHRC1, which is highly expressed by CAF in the ICC/IDC TME, has been shown to polarize macrophages to the M2, pro-tumorigenic, phenotype through TGF-β signaling67. To determine if ICC/IDC was associated with increased M2 macrophages, clusters were analyzed for myeloid lineage markers, including CD68 (Fig. 7a, b). While monocyte markers (VCAN and S100A9) mapped to cluster 18, cluster 4 cells expressed dendritic cell markers (CD1C and CLEC10A) as well as macrophage markers associated with disease recurrence in clear cell renal cell carcinoma (C1QB, TREM2, and APOE)80 (Fig. 7b). A subcluster of cells within cluster 4 was notably increased in ICC/IDC-enriched prostate cancer compared to benign-enriched prostate. These cells were C1QB+, TREM2+, and APOE+ and expressed the anti-inflammatory M2 macrophage markers CD163, MSR1, and MRC1 (Fig. 7c). C1QB, TREM2, APOE, CD163, and MSR1 were significantly increased in ICC/IDC-enriched compared to benign-enriched cluster 4 cells (Fig. 7d and DEG analysis for cluster 34 in Source Data). Consistent with a M2 anti-inflammatory phenotype, cells in cluster 4 from the ICC/IDC TME had decreased inflammatory-related hallmarks by pssGSEA (Fig. 7e). To better delineate myeloid heterogeneity in the ICC/IDC TME, clusters 4 and 18 were re-clustered to 7 clusters (Mac0-6) with Mac1 and Mac4 almost entirely derived from cluster 18 while Mac0, Mac2, Mac3, Mac5, and Mac6 were largely derived from cluster 4 (Fig. 7f, genes for Mac0-Mac6 in Source Data). Monocyte markers mapped to cluster Mac1, while dendritic cell markers mapped to cluster Mac2 with a proliferative subset (STMN1+MKI67+) in Mac5 (Fig. 7f, g). C1QB+TREM2+APOE+ macrophages mapped to Mac0, which were increased along with CD163 and MSR1 in the ICC/IDC TME (Fig. 7h, i). A gene signature based on these cells (C1QB, TREM2, APOE, CD163, MRC1, and MSR1) showed a significant association with worse prostate cancer progression-free survival in the TCGA PanCancer Atlas prostate adenocarcinoma cohort74 and worse disease-free survival in the MSKCC Prostate Adenocarcinoma cohort75 (Fig. 7j). Overall, C1QB+TREM2+APOE+ macrophages that express M2 macrophage markers, CD163 and MSR1, were increased in the ICC/IDC TME.

Fig. 7: Increased C1QB+TREM2+APOE+ M2 macrophages in prostate ICC/IDC.
figure 7

a Unsupervised graph-based clustering of all samples visualized by UMAP highlighted for myeloid clusters 4 and 18 delineated by benign-enriched and ICC/IDC-enriched prostate. The dotted area demarks a subcluster of cells increased in ICC/IDC-enriched regions compared to benign-enriched regions. b Violin plots of myeloid, monocyte, dendritic cell, and macrophage marker expression in clusters 4 and 18. c Feature plots of C1QB, TREM2, APOE, and M2 macrophage markers (CD163, MSR1, and MRC1) in clusters 4 and 18. d Violin plots of C1QB, TREM2, APOE, and M2 macrophage markers (CD163, MSR1, and MRC1) in clusters 4 and 18 separated by ICC/IDC-enriched and benign-enriched. e PssGSEA of hallmark pathways in ICC/IDC-enriched prostate compared to benign-enriched prostate cells in clusters 4 and 18. f Re-clustering of clusters 4 and 18 into six clusters (Mac0-Mac6) separated by ICC/IDC-enriched and benign-enriched prostate and demarked by the original clusters 4 and 18. gi Violin plots of markers in Mac0-Mac6. j Kaplan–Meier and log-rank test of progression-free survival in the TCGA PanCancer Atlas Prostate Adenocarcinoma for C1QB, TREM2, APOE, CD163, MRC1, and MSR1 signature (C1QB+TREM2+APOE+ M2 Signature) by median expression (n = 492). Kaplan-Meier and long-rank test of disease-free survival (DFS) in the MSKCC Prostate Adenocarcinoma for C1QB, TREM2, APOE, CD163, MRC1, and MSR1 signature (C1QB+TREM2+APOE+ M2 Signature) by median expression (n = 131).