Single cell analyses reveal the PD-1 blockade response-related immune features in hepatocellular carcinoma

Abstract

Background: Immunotherapy has demonstrated its potential to improve the prognosis of patients with hepatocellular carcinoma (HCC); however, patients' responses to immunotherapy vary a lot. A comparative analysis of the tumor microenvironment (TME) in responders and non-responders is expected to unveil the mechanisms responsible for the immunotherapy resistance and provide potential treatment targets. Methods: We performed sequencing analyses using 10x Genomics technology on six HCC patients who responded to anti-PD-1 therapy and one HCC patient who did not respond. Additionally, we obtained single cell data from untreated, responsive, and nonresponsive HCC patients from public databases, and used part of the datasets as a validation cohort. These data were integrated using algorithms such as Harmony. An independent validation cohort was established. Furthermore, we performed spatial transcriptomic sequencing on the tumor adjacent tissues of three HCC responsive patients using 10x Genomics spatial transcriptomic technology. Additionally, we analyzed data about three HCC patients obtained from public databases. Finally, we validated our conclusions using immunofluorescence, flow cytometry, and in vivo experiments. Results: Our findings confirmed the presence of "immune barrier" partially accounting for the limited efficacy of immunotherapy. Our analysis revealed a significant increase in TREM2⁺ Macrophages among non-responsive patients expressing multiple immunosuppressive signals. anti-Csf1r monoclonal antibodies effectively eliminated these macrophages and augmented the therapeutic effects of anti-PD-1 therapy. TCR⁺ Macrophages possessed direct tumor-killing capabilities. IL1B⁺ cDC2 was the primary functional subtype of cDC2 cells. Absence of THEMIS^hi CD8⁺ T subtypes might diminish immunotherapeutic effects. Furthermore, CD8⁺ T cells entered a state of stress after anti-PD-1 treatment, which might be associated with CD8⁺ T cell exhaustion and senescence. Conclusions: The profiles of immune TMEs showed differences in HCC patients responsive, non-responsive and untreated. These differences might explain the discounted efficacy of immunotherapy in some HCC patients. The cells and molecules, which we found to carry unique capabilities, may be targeted to enhance immunotherapeutic outcomes in patients with HCC.

Keywords: PD-1; hepatocellular carcinoma; immune microenvironment; immunotherapy; single cell.

Figure 1

Single cell atlas of HCC samples responsive and non-responsive to PD-1 blockade. (A) Workflow of this study. (B) MRI images of patients with and without response to treatment. (C) Cell type maps for different response conditions. (D) Cell subgroups and their corresponding gene markers. (E) Box plots showing the proportions and statistics of various cell types in response and non-response patients in the GSE206325 cohort (Wilcoxon test. *P < 0.05, **P < 0.01, ***P < 0.001, ****P < 0.0001). (F) The proportion of cell types across different response types in tumor tissue with our cohort. (G) UMAP plot illustrating myeloid cell subpopulations. (H) UMAP plot showing NK/T cell subpopulations.

Figure 2

Spatial transcriptomic features of responsive and non-responsive HCC adjacent tissues. (A-C) Cell types and corresponding markers in patients P1, P6, and P8. (D) Distribution images of POSTN, TREM2, CD8A, GZMK, and PD-1 in the tumor margin of patients and density distribution maps of POSTN and TREM2 expression, indicating the presence of immune barriers in both responsive and non-responsive patients. (E) TREM2 and CD68 represent TREM2⁺ Macrophages, and POSTN represents POSTN⁺ CAFs). Multicolor immunofluorescence staining of the tumor margin in patient P8 further demonstrates the existence of immune barriers composed of TREM2⁺ Macrophages and POSTN⁺ CAFs in responsive patients.

Figure 3

TREM2⁺ Macrophages represent a predominant immunosuppressive subset within the macrophage population. UMAP plot illustrating subpopulations of macrophages in our cohort. (B) UMAP plot showing subpopulations of macrophages in different tissue sites. (C) Bar graph depicting the distribution of macrophage subgroups across different tissue types. (D) UMAP plot illustrating subpopulations of macrophages in the GSE206325 cohort. (E) Similarities between macrophage subpopulations in our cohort and the GSE206325 cohort. (F) Boxplot showing the distribution of macrophage subpopulation proportions in treatment response type and tissue type (Wilcoxon test. *P < 0.05, **P < 0.01, ***P < 0.001, ****P < 0.0001). (G) Multicolor immunofluorescence showing TCR⁺ Macrophages in HCC-mouse module tumor tissue. (H) Multicolor immunofluorescence showing TREM2⁺ Macrophages in human HCC tumor tissue.

Figure 4

TREM2⁺ macrophages represent a predominant immunosuppressive subset within the macrophage population. Marker genes of TREM2⁺ Macrophages (top50). (B) Marker genes of TCR⁺ Macrophages (top50). (C) GO enrichment results for the signature gene of TREM2⁺ Macrophages. (D) GO enrichment results for the signature gene of TCR⁺ Macrophages. (E) Correlation of TREM2 expression with HAVCR2, CTLA4, PDCD1, ENTPD-1 in the TCGA-LIHC cohort (analyzed using Log2(TPM+1) values) and spatial correlation of TREM2 expression with CD8⁺ Tex signatures in P11T. (F) The dot-plot shows the results of CD8⁺ Tex and macrophage subgroups pair-receptor combinations calculated by CellphoneDB software in the tumor tissue of our cohort (after filtering with p-values < 0.05 condition). (G) Kaplan-Meier survival analysis Trem2^-/- and WT HCC-mouse module. (H) tumor numbers and liver-body weight ratio in Trem2^-/- and WT HCC-mouse module (T-test. *P < 0.05, **P < 0.01, ***P < 0.001, ****P < 0.0001). (I) tumor numbers and liver-body weight ratio in Trem2^-/- and WT HCC-mouse module after treated with anti-PD-1 (T-test. *P < 0.05, **P < 0.01, ***P < 0.001, ****P < 0.0001). (J) tumor numbers and liver-body weight ratio in Trem2^-/- and WT HCC-mouse module after treated under different conditions (T-test. *P < 0.05, **P < 0.01, ***P < 0.001, ****P < 0.0001). (K) The difference of TREM2⁺ Macrophages ratio in anti-Csf1r and Isotype treatment condition (T-test. *P < 0.05, **P < 0.01, ***P < 0.001, ****P < 0.0001). (L-N) The difference in cell ratio of different treatment conditions (T-test. *P < 0.05, **P < 0.01, ***P < 0.001, ****P < 0.0001).

Figure 5

IL1B⁺ cDC2s are the main executor on cDC2s. UMAP plot showing DC subgroups of the discovery cohort. (B) Bubble plot displaying unique markers of different DC subgroups. (C) Differential expression of genes related to DC function across DC subgroups. (D) UMAP plot illustrating the distribution of DC subgroups across different tissues and under different treatment conditions. (E) Bar graph showing the proportion of DC subgroups across different tissues and under different treatment conditions. (F) Volcano plot depicting differential genes between two cDC2 subgroups. (G) Multicolor immunofluorescence staining confirms the presence of two types of cDC2 in human HCC tissues. (H) Bubble plot shows the differences in the co-receptor pairing between the two types of cDC2 and CD4⁺ T cell subgroups (p-values < 0.05, IL1B⁺ cDC2 vs DPYD⁺ cDC2). (I) Cell interactions between IL1B⁺ cDC2 dendritic cell and CD4⁺ T cell subgroups in the tumor tissue of non-responsive patients. (J) Cell interactions between DPYD⁺ cDC2 dendritic cell and CD4⁺ T cell subgroups in the tumor tissue of non-responsive patients. (K) Communication of the CD40 signaling pathway between cDC2 dendritic cells and CD4⁺ T cell subgroups. (L) Communication of the TGF-β signaling pathway between cDC2 dendritic cells and CD4⁺ T cell subgroups.

Figure 6

Transcriptional changes of CD8⁺ T cells after anti-PD-1 treatment. (A) UMAP plot showing CD8⁺ T cell subgroups. (B) Differential expression of genes related to CD8⁺ T cell function across subgroups. (C) CD8⁺ T cell developmental trajectory simulated with monocle2. (D) IL7R, GZMK, PDCD1, and CBLB expression levels along the developmental trajectory. (E) UMAP plot for re-clustering and defining CD8⁺ T cells from the discovery cohort. (F) Feature plot showing expression levels of GZMB, PDCD1, TOX, and MKI67. (G) Gene Set Enrichment Index of all CD8⁺ T cells from responsive, non-responsive, and untreated patients. (H) Volcano plot showing differential genes in CD8⁺ T cells from responsive and non-responsive patients. (I) GO enrichment of differential genes between responsive and non-responsive groups. (J) Volcano plot showing differential genes in CD8⁺ T cells from responsive and untreated patients. (K) GO enrichment of up-regulated differential genes in CD8⁺ T cells from responsive patients. (L) Scoring of the TSTR signature in CD8⁺ T cells from non-responsive, responsive, and untreated patients. (M) Violin and box plots showing TSTR scores in CD8⁺ T cells from responsive, non-responsive, and untreated groups (Wilcoxon test. *P < 0.05, **P < 0.01, ***P < 0.001, ****P < 0.0001). (N) Multicolor immunofluorescence shows a statistically significant difference in the abundance of HSPA1B⁺ CD8⁺ T cells between responders and non-responders in the treatment (T-test. *P < 0.05, **P < 0.01, ***P < 0.001, ****P < 0.0001).

Figure 7

Transcriptional changes of CD8⁺T cells after anti-PD-1 treatment. (A) Grouping of CD8⁺ T cells based on TSTR scores, shown with box and violin plots for exhaustion scores (Wilcoxon test. *P < 0.05, **P < 0.01, ***P < 0.001, ****P < 0.0001). (B) Trends in cytotoxic, exhaustion, and senescence scores about TSTR scores. (C) Trends in glucose metabolism, oxidative phosphorylation, and fatty acid metabolism about TSTR scores. (D) UMAP plot showing the distribution of CD8⁺ T cell subgroup in responsive, non-responsive, and untreated patients. (E) Bar graph showing the proportion of CD8⁺ T cell subgroup in responsive, non-responsive, and untreated patients. (F) Kaplan-Meier survival analysis of the signature of C3_Tn_THEMIS in the TCGA-LIHC cohort. (G) Bubble plot showing Top30 marker genes of C3_Tn_THEMIS. (H) GO enrichment results for marker genes of C3_Tn_THEMIS. (I) Correlation of CD8A and THEMIS in the TCGA-LIHC cohort. (J) UMAP plot illustrating the developmental trajectory of non-responsive CD8⁺ T cells simulated with monocle3. (K) UMAP plot showing the developmental trajectory of responsive CD8⁺ T cells simulated with monocle3. (L) Radar plot showing cell type inclination scores for C1_Tn and C3_Tn_THEMIS subgroups. (M) Multicolor immunofluorescence staining confirms the presence of THEMIS⁺ CD8⁺ T cells in human HCC tissues. (N) tumor characteristics of HCC-mouse models after treatment with anti-PD-1 and anti-PD-1 combined with AAV-Themis (T-test. *P < 0.05, **P < 0.01, ***P < 0.001, ****P < 0.0001). (O) Bubble plot showing cell communication between cDC1, LAMP3⁺ cDC3, and various types of CD8⁺ T cells in the tumor tissue of responsive patients. (P) Communication of TIGIT, PDL1, PDL2, and CD40 signaling pathways between cDC1 and CD8⁺ T cells.

Figure 8

Transcriptional changes of CD4⁺T cells after anti-PD-1 treatment. (A) The UMAP plot illustrates the subpopulations of CD4⁺ T cells in the discovery cohort. (B) The marker genes associated with CD4⁺ T cell subpopulations are identified in the discovery cohort. (C) Comparison of differentially expressed genes in CD4⁺ T cells between responders and non-responders to anti-PD-1 treatment in tumor and normal tissues. (D) The figure shows the distribution of CD4⁺ T cell subpopulations in the discovery cohort across different tissues and treatment conditions. (E) The distribution of CD4⁺ T cell subpopulations in tumor tissue across responders and non-responders in the discovery cohort. (Wilcoxon test. *P < 0.05, **P < 0.01, ***P < 0.001, ****P < 0.0001). (F) The bubble plot shows the top 50 genes of the CD4_THEMIS subpopulation. (G) GO enrichment results for marker genes of CD4_THEMIS. (H) Kaplan-Meier survival analysis of the signature of CD4_THEMIS in the TCGA-LIHC cohort. (I) The differentially expressed genes between the high and low infiltration levels of CD4_THEMIS in TCGA-LIHC. (J-K) The legend illustrates the main findings of this study.