Tumor-associated macrophages and PD-L1 in prostate cancer: a possible key to unlocking immunotherapy efficacy

Patients

This retrospective study was conducted in accordance with the principles outlined in the Declaration of Helsinki and received approval from the Institutional Review Board (IRB) of the Second Hospital of Tianjin Medical University. Written informed consent was obtained from all participating patients. Our cohort included patients who underwent radical prostatectomy at the Second Hospital of Tianjin Medical University between March 2013 and December 2020. An experienced pathologist confirmed the histological diagnoses. Clinicopathological data were obtained from medical records. Formalin-fixed, paraffin-embedded tissue blocks of human tissue specimens were collected under a protocol approved by the IRB of the Second Hospital, following the principles outlined in the Declaration of Helsinki. Table 1 provides a summary of the detailed clinicopathological information for the enrolled patients.

In addition to the local cohort, this study also utilized two publicly available prostate cancer RNA-Seq datasets. Detailed information on these datasets can be found at https://portal.gdc.cancer.gov/ (TCGA-PRAD) and http://www.cbioportal.org/ (prad_su2c_2019) [16].

Tissue microarray (TMA) construction

Three morphologically representative areas were selected from tumors of PCa patients. Furthermore, each tissue microarray (TMA) included tumor-adjacent samples (52 pairs) from the same patients who underwent radical prostatectomy, serving as internal negative controls. To verify the histopathological diagnosis and ensure adequate tissue sampling, a section from each microarray was stained with hematoxylin and eosin and then examined using bright-field microscopy.

Immunohistochemistry (IHC) staining

The TMAs were sectioned into 3 μm slices and mounted on Flex IHC microscope slides (DakoCytomation, Glostrup, Denmark). The sections were deparaffinized in xylene and rehydrated through a graded alcohol series. Antigen retrieval was performed by heating the sections with Envision Flex Target Retrieval solution at high pH (Dako, Glostrup, Denmark). Staining was conducted at room temperature using an automatic staining workstation (Dako Autostainer Plus, Dako). The primary antibodies used included rabbit anti-PD-L1 antibodies (1:100 dilution, ab205921, Abcam, Cambridge, UK), rabbit anti-CD68 antibodies (1:50 dilution, BA3638, Boster, California, UK), and rabbit anti-CD163 antibodies (1:200 dilution, bs-2527R, Bioss, Beijing, China). Negative controls were prepared by omitting the primary antibody. Four observers, blinded to clinical data, independently evaluated the IHC results. The number of CD163+ cells was counted in each 1 mm² area from three independent high-power representative microscopic fields (HPFs, 400×; 0.0625 μm²), in both the tumor nests and surrounding stroma. For survival analysis, all samples were divided into low and high groups based on the number of positive cells/mm², using cut-off values of 50 (median) for both CD68+ and CD163+ cells. PD-L1 expression in more than 10% of tumor cells was associated with poorer survival [17], which was established as a cut-off point for subsequent analyses. Specimens were classified into two categories based on tumor cell proportion score (TPS): negative (<10%) and positive (10%–100%).

Publicly available datasets analysis

In this study, two independent RNA-Seq datasets were utilized. The TCGA-PRAD (prostate adenocarcinoma) dataset’s RNA-Seq FPKM data was obtained from The Cancer Genome Atlas (TCGA) database (https://portal.gdc.cancer.gov/). Batch effect analysis was conducted using TCGA Batch Viewer (https://bioinformatics.mdanderson.org/public-software/), and no significant batch effect was observed. Clinical follow-up information, as well as pathological information, was downloaded from the Xena database (http://xena.ucsc.edu/). For the prad_su2c_2019 dataset, both mRNA expression data and clinicopathological information were obtained from the cBioPortal database (http://www.cbioportal.org/). Clinical information was manually reviewed, and cases lacking essential information were excluded from the study. Detailed clinicopathologic information can be found in Supplementary Tables 1 and 2.

Gene set enrichment analysis

To investigate the hallmarks and pathways enriched in the predicted high- and low-risk groups, Gene Set Enrichment Analysis (GSEA) was performed as previously described [18]. The infiltration level of M2 macrophages was used as the ranking metric to generate the ranked gene list in the two RNA-Seq datasets. GSEA was conducted using the WebGestalt online program (http://www.webgestalt.org/), with the WikiPathway gene set employed in the analysis.

Immune infiltration analysis by CIBERSORTx

The immune infiltration levels of 22 immune cells were determined using the CIBERSORTx program (https://cibersortx.stanford.edu/) with bulk-tumor RNA-Seq data as input. CIBERSORTx was executed in both relative and absolute models, incorporating S-mode batch effect correction. The RNA-Seq FPKM data and the built-in LM22 signature matrix were used as input.

Singe cell RNA-seq data analysis

The single-cell RNA-Seq data were analyzed using the TISCH2 database (http://tisch.comp-genomics.org/search-gene/). Six single-cell RNA-Seq datasets were selected for this study: GSE137829 (Single-cell analysis supports a luminal-neuroendocrine transdifferentiation in human prostate cancer), GSE141445 (Single-cell analysis reveals the onset of multiple progression-associated transcriptomic remodellings in prostate cancer), GSE143791 (Human prostate cancer bone metastases have an actionable immunosuppressive microenvironment), GSE150692 (Single-cell RNA-seq analysis of wild-type mouse and benign human prostate), GSE172301 (Single-cell RNA-sequencing of adult human prostates from BPH patients), and GSE176031 (Single-cell analysis of human primary prostate cancer reveals the heterogeneity of tumor-associated epithelial cell states) [18–23]. The data were presented using UMAP (Uniform Manifold Approximation and Projection) for the visualization of cell clusters [19–24].

Statistical analysis

All statistical analyses were conducted using SPSS version 18 (IBM Co., Armonk, NY, USA) or R software (version 3.6.0). Chi-square tests and Fisher’s exact tests were utilized for comparisons between categorical variables. Mean value comparisons were performed using t-tests. BCR-free time was defined as a PSA ≥0.2 ng/mL followed by a subsequent confirmatory value of ≥0.2 ng/mL. Survival curves were plotted using the Kaplan-Meier method. Differences in survival times between patient subgroups were compared using Mantel’s log-rank test. GEPIA (Gene Expression Profiling Interactive Analysis, http://gepia2.cancer-pku.cn/#index) online tools were used to determine the differentially expressed genes in the TCGA database. All P values were based on two-sided statistical analysis, and a P value less than 0.05 was considered statistically significant.

Availability of data and materials

All datasets used in this study are publicly available from the corresponding database or provided as Supplementary Materials.

Expression patterns of CD68 and CD163 proteins in PCa

To analyze the infiltration levels of TAMs in PCa, we first observed the general expression patterns of CD68 (TAM-marker) and CD163 (M2-TAM marker) proteins in PCa and adjacent normal tissues using TMAs containing 96 PCa samples (52 with matched adjacent normal tissues). The IHC results demonstrated that CD68 and CD163 in PCa tissues were diffusely expressed in stromal cells with membranous and cytoplasmic staining, indicating that TAMs were indeed infiltrating PCa tissues with a dispersed distribution pattern (Figure 1A). Moreover, tumor cells exhibited less positive staining with CD68 and CD163. We further analyzed the signal intensity of CD68 and CD163 proteins in cancerous and normal tissues, revealing that the positive rate of CD68 in PCa was significantly higher than that in tumor-adjacent and normal tissues, with average percentages of 68.68 ± 14.72, 29.90 ± 13.15, and 13.33 ± 2.89, respectively (Figure 1B, 1C, P < 0.01). These findings were similar to the results for CD163 (Figure 1D, 1E, 63.07 ± 11.76 vs. 28.75 ± 10.61 vs. 6.67 ± 2.89, P < 0.05). In addition to the IHC results, CD68 and CD163 were also upregulated at the mRNA level in TCGA-PRAD dataset (Figure 1F). These results suggested that the infiltration levels of TAMs, including CD68+ TAMs and CD163+ TAMs (M2), were increased in PCa.

Figure 1. Expression patterns of CD68 and CD163 proteins in PCa and normal prostate tissue. (A) Representative IHC images of PCa tissue slides with low (left panel) or high (right panel) levels of CD68 and CD163 protein. IHC, immunocytochemistry. (B) Dot plot shows the positive rate of CD68 in PCa and para-cancer (para-PCa) tissue. (C) Bar plot shows the positive rate of CD68 in normal prostate tissue, para-PCa, and PCa. (D) Dot plot shows the positive rate of CD163 in PCa and para-PCa tissue. (E) Bar plot shows the positive rate of CD163 in normal prostate tissue, para-PCa, and PCa. (F) Box plots showed the mRNA expression levels of CD68 and CD163 in the TCGA-PRAD dataset. Student’s t-test, ^*P < 0.05, ^**P < 0.01, ^****P < 0.001.

The infiltration levels of M2 TAMs were positively correlated with Gleason score

To investigate the clinical significance of infiltrating TAMs in patients, we first analyzed the IHC signals of CD68 and CD163 in PCa tissue, along with detailed clinicopathological information. The results revealed that CD68 had no significant associations with Gleason score (Figure 2A, P = 0.10), whereas the level of CD163 protein was significantly elevated in PCa tissues with a high Gleason score (Figure 2B, Gleason score = 7 vs. Gleason score = 9, P = 0.02). To further support our findings, we analyzed CD68 and CD163 mRNA expression in the TCGA prostate cancer dataset (TCGA-PRAD), which contains 495 PCa samples. Consistent with our local cohort, a high expression of CD163 mRNA was significantly associated with an advanced Gleason score (Figure 2C); although CD68 mRNA expression also increased with Gleason score, it was less statistically significant compared to that of CD163 (Figure 2D). Additionally, we calculated the infiltration levels of M2- and M1-TAMs in the TCGA-PRAD dataset using the CIBERSORTx program. Correlation analysis indicated that the infiltration level of M2-TAMs was significantly elevated in PCa with a higher Gleason score (Figure 2F, P = 3.5e-07), while the infiltration level of M1-TAMs showed a weaker association with Gleason score (Figure 2E, P = 0.026). Collectively, these results suggested that the infiltration of CD163+ TAMs (M2) is closely linked with the risk stratification of PCa. Both our local cohort and the TCGA-PRAD dataset indicated more pronounced CD163 expression and M2-TAM infiltration in PCa with highly aggressive features.

Figure 2. The infiltration levels of M2 TAMs were positively correlated with Gleason score. (A) CD68 positive rate was moderately negatively correlated with Gleason score (GS). (B) CD163 positive rate increased with increasing GS and the difference was more obverse between GS = 9 and GS = 6/3+4 group. (C) Violin plot showed the expression level of CD68 mRNA with increasing GS. (D) The expression level of CD274 mRNA was positively correlated with GS. (E) Violin plot showed the infiltration levels of M1 macrophages with increasing GS. (F) Violin plot showed the infiltration levels of M2 macrophages with increasing GS. A, B and C. ANOVA test; C, D and E, Kruskal-Wallis test.

M2-TAMs infiltration was positively associated with PD-L1 expression

We further investigated the association between TAMs and PD-L1, a key regulator associated with the efficacy of immune-blockade therapy. IHC results demonstrated that in 14.4% (18 out of 96) of PCa samples, more than 10% of tumor cells showed a positive PD-L1 IHC signal, which is consistent with previous reports [25]. These data indicated that PD-L1 expression levels were mainly low to moderate for PCa patients (Figure 3A). The staining levels of CD68 and CD163 in the PD-L1+ group were higher than those in the PD-L1− group (CD68: 72.11 vs. 66.36, P = 0.13; CD163: 41.83 vs. 28.03, P < 0.01; Table 1 and Figure 3B). Additionally, the expression levels of CD163 and PD-L1 were positively correlated (r = 0.51, P < 0.01; Figure 3B right panel). However, no significant correlation was found between the expression levels of CD68 and PD-L1 (r = 0.10, P = 0.38, Figure 3C left panel).

Figure 3. M2-TAMs infiltration was positively associated with PD-L1 expression. (A) Representative IHC images of PCa tissue slides with low (left panel) or high (right panel) levels of PD-L1 protein. (B) Dot plot shows the expression of CD68 in PD-L1 (+) and PD-L1 (−) PCa (Student’s t-test). (C) Dot plot shows the expression of CD163 in PD-L1 (+) and PD-L1 (−) PCa (Student’s t-test). (D) Bar plot showed PD-L1 protein expression increasing with Gleason score (ANOVA test). (E) Correlation between CD274 mRNA (encoding PD-L1) and CD68 mRNA in TCGA-PRAD and prad_su2c_2019 dataset. (F) Correlation between CD274 mRNA and CD163 mRNA in the two independent PCa datasets. (G) Correlation between the infiltration levels of M2 macrophages and the expression levels of CD274 mRNA in the two independent PCa datasets.

Moreover, IHC also showed that PD-L1 protein expression was positively correlated with the Gleason score of PCa patients (Figure 3D). In addition, we analyzed the differentially expressed genes in samples with high and low PD-L1 expression. We found that among these, 713 genes were upregulated and 83 genes were downregulated (Supplementary Figure 1A, 1B). This result suggests that the genes positively correlated with PD-L1 might play a more significant biological role. We performed an enrichment analysis on these differentially expressed genes using both the HALLMARKS 50 gene set (Supplementary Figure 1C, 1D) and the KEGG gene set. The results revealed that several immune-related pathways, such as “interferon gamma”, “interferon alpha”, “Cytokine-cytokine receptor interaction”, and “Chemokine signaling pathway”, were significantly enriched. However, for genes negatively correlated with PD-L1, we used GSEA for analysis (Supplementary Figure 1E). We found pathways like oxidative phosphorylation, MYC target genes, and DNA repair to be enriched, but the statistical significance was not very pronounced. These findings suggest that the elevated expression level of PD-L1 in PCa is closely related to immune response and activation, consistent with the classic function of PD-L1.

We further analyzed the mRNA expression correlation among CD68, CD163, and CD274 (encoding the PD-L1 protein) in two independent datasets, TCGA-PRAD, and prad_su2c_2019. In line with our local cohort, the results demonstrated that CD274 mRNA expression was significantly positively correlated with CD68 (Figure 3E) and CD163 (Figure 3F) mRNA expression. Notably, the infiltration level of M2-TAMs was also strongly correlated with CD274 expression in TCGA-PRAD and prad_su2c_2019 cohorts (Figure 3G). This is in agreement with our IHC data, which indicated that PCa with higher levels of M2-TAM infiltration exhibited elevated levels of PD-L1 expression. Furthermore, CD163 mRNA expression was also significantly positively associated with CD68 mRNA expression in the two PCa RNA-Seq cohorts (Supplementary Figure 2A). Additionally, CD68 mRNA and CD163 mRNA expression were positively associated with the levels of M2-TAM infiltration (Supplementary Figure 2B, 2C). In summary, these results suggested that PD-L1 expression is associated with M2-TAM infiltration in PCa.

M2-TAMs infiltration and PD-L1 expression are associated with the prognosis of PCa patients

To further address the clinical significance of M2-TAMs and PD-L1 in PCa patients, we first analyzed the association of TAMs markers CD68 and CD163 with BCR-free time in our local cohort. The median follow-up time was 86 months (m), 68 (71.6%) patients experienced BCR after radical prostatectomy, and the median BCR-free time was 33.03 m (3 m – 88 m). The rates of 1-, 2-, and 5-year BCR-free survival rates were 78%, 53.4%, and 31%, respectively. No statistical significance was found in the BCR-free time of patients between low- and high-CD68 expression groups (P = 0.15, Figure 4A). As for CD163, the BCR-free time of the patients with low CD163 expression was significantly longer than that of the patients with high CD163 expression (mean BCR-free time: 45.6 m vs. 31.8 m, P < 0.05, Figure 4B). Notably, the BCR-free time of the high PD-L1 expression group was shorter than that of the PD-L1 low expression group (median BCR-free time: 24.67 m vs. 35.3 m, P = 0.01, Figure 4C). These results indicated that increased CD163+ TAM (M2) infiltration and high PD-L1 expression were correlated with shorter BCR-free time and could be recognized as poor prognostic factors for PCa patients. In addition, in COX regression analysis models, the CD163 IHC signal was an independent adverse factor (HR = 2.48; P = 0.04), indicating that M2-TAMs could independently predict the prognosis of patients with PCa (Table 2).

Figure 4. Impact of M2-TAMs Infiltration and PD-L1 expression on the prognosis of PCa patients. (A–C) Kaplan-Meier analysis of the protein expression levels of CD68, CD163, and PD-L1 with biochemical recurrent (BCR) free survival time. (D–F) Kaplan-Meier analysis of the mRNA expression levels of CD68, CD163, and CD274 with BCR free survival time in TCGA-PRAD dataset. (G, H) Kaplan-Meier analysis of the infiltration levels of M1- and M2- TAMs with progress-free survival (PFS) in TCGA-PRAD dataset. (I) Kaplan-Meier analysis of the infiltration levels of M2-TAMs with overall survival (OS) in prad_su2c_2019 dataset.

In support, we also analyzed the clinical significance of CD68 and CD163 mRNA expression in TCGA-PRAD and prad_su2c_2019 cohorts. Consistent with the local cohort, high CD68 (Figure 4D) and CD163 (Figure 4E) mRNA expression were associated with poor prognosis of PCa patients. However, CD274 mRNA expression showed no association with the prognosis, which is also consistent with the IHC data (Figure 4F). Notably, we found that the infiltrated M2-TAMs had significant prognosis value in predicting DFS (Figure 4G), whereas M1-TAMs were not correlated with the clinical outcome of PCa patients (Figure 4H). Consistently, the infiltration level of M2-TAMs was correlated with the prognosis in the SU2C dataset (Figure 4I). In summary, these results suggested that PD-L1 protein expression is a promising biomarker for outcome prediction, and M2-TAMs infiltration is tightly linked with the progression of PCa.

M2-TAMs infiltration is a key factor involved in immune modulation in PCa

Since we have identified that M2-TAMs infiltration is tightly linked with the prognosis of PCa, we investigated the corresponding molecular foundations. We performed GSEA using the infiltration levels of M2-TAMs as the metric in two PCa data sets, TCGA-PRAD and prad_su2c_2019 against the WikiPathway geneset. The results showed that PD-1 blockade-related genes were highly expressed in PCa samples with a high level of M2-TAMs infiltration in the two analyzed datasets, which further supports the intrinsic connections between M2 macrophages and PD-1-related immune therapy (Figure 5A–5D). We next analyzed the survival-prediction potential of the additional 22 immune cells using the CIBERSORTx-LM22 signature matrix. The results showed that, in addition to M2-TAMs, 4 immune cell types, including regulatory T cells, memory T cells, memory B cells, and plasma cells, also showed a strong association with prognosis (Figure 5E–5H).

Figure 5. M2 macrophage infiltration is a key factor involved in immune modulation in PCa. (A, B) Gene set enrichment analysis (GSEA) of PCa patients with high and low infiltration levels of M2 macrophages in the TCGA-PRAD dataset. (C, D) GSEA analysis of PCa patients with high and low infiltration levels of M2-TAMs in prad_su2c_2019 dataset. (E–H) Kaplan-Meier analysis of PCa patients with high or low infiltration levels of T regulatory cells, T memory cells, B memory cells, and B plasma cells with PFS in TCGA-PRAD dataset.

The analysis also showed that M2-TAMs were the most abundant immune cells across the 22 analyzed cell types (Figure 6A, 6B). Furthermore, we analyzed six single-cell RNA-Seq data of PCa to characterize the cellular source of the CD274 expression across multiple cell types. Notably, the results suggested that Monocells/macrophages exhibit a higher expression of CD274 compared with other cells (Figure 6C–6E), suggesting TAMs were indeed the source of PD-L1 expression in bulk tumors. We also found that M2-TAMs were significantly negatively correlated with a large fraction of other immune cells, strongly suggesting that M2-TAMs play important roles in immune modulation in PCa (Figure 6F, 6G).

Figure 6. M2-TAMs are major immune-modulator in PCa. (A, B) The landscape of 22 immune cells in PCa. The infiltration factions of the 22 analyzed cell types were calculated by CIBERSORTx program in TCGA-PRAD (A) and prad_su2c_2019 datasets (B). (C, D) UMAP showed the distribution of cell clusters in the GSE172301 dataset, and the expression levels of CD274 mRNA were shown as blue dots in log2TMP+1. (E) Heatmap showed the expression levels of CD274 mRNA across multiple cell types in TME. (F, G) Heatmap showed the correlation among the 22 immune cells in TCGA-PRAD (F) and prad_su2c_2019 dataset (G).

Together, these results suggest that M2-TAMs infiltration is significantly associated with PD-1 blockade therapy. The levels of regulatory T cells, memory T cells, memory B cells, plasma cells, as well as the M2-TAMs, may be involved in the modulation of the tumor microenvironment in PCa and can serve as targets for improving the efficacy of immunotherapy in the treatment of PCa.