Macro CD5L+ deteriorates CD8+T cells exhaustion and impairs combination of Gemcitabine-Oxaliplatin-Lenvatinib-anti-PD1 therapy in intrahepatic cholangiocarcinoma

Abstract

Intratumoral immune status influences tumor therapeutic response, but it remains largely unclear how the status determines therapies for patients with intrahepatic cholangiocarcinoma. Here, we examine the single-cell transcriptional and TCR profiles of 18 tumor tissues pre- and post- therapy of gemcitabine plus oxaliplatin, in combination with lenvatinib and anti-PD1 antibody for intrahepatic cholangiocarcinoma. We find that high CD8 GZMB⁺ and CD8 proliferating proportions and a low Macro CD5L⁺ proportion predict good response to the therapy. In patients with a poor response, the CD8 GZMB⁺ and CD8 proliferating proportions are increased, but the CD8 GZMK⁺ proportion is decreased after the therapy. Transition of CD8 proliferating and CD8 GZMB⁺ to CD8 GZMK⁺ facilitates good response to the therapy, while Macro CD5L⁺-CD8 GZMB⁺ crosstalk impairs the response by increasing CTLA4 in CD8 GZMB⁺. Anti-CTLA4 antibody reverses resistance of the therapy in intrahepatic cholangiocarcinoma. Our data provide a resource for predicting response of the combination therapy and highlight the importance of CD8⁺T-cell status conversion and exhaustion induced by Macro CD5L⁺ in influencing the response, suggesting future avenues for cancer treatment optimization.

Fig. 1. The GOLP regimen showed high therapeutic efficacy in iCCA.

A Bar plot showing the overall response rate (ORR) in clinical trials of different first-line therapeutic strategies for iCCA patients. B Workflow of combination therapy of Gemcitabine, Oxaliplatin, Lenvatinib, and anti-PD1 antibody (GOLP) for iCCA patients. Created with BioRender.com. C Representative results of ‘stable disease’ (SD) and ‘partial response’ (PR) iCCA patients at baseline and after 3 cycles of GOLP treatment. SD and PR were evaluated according to the Response Evaluation Criteria in Solid Tumors version 1.1 (RECIST1.1). The arrows indicate the tumors of interest. D Percentage of tumor size reduction after 3 cycles of GOLP treatment in FDU-ZS-ICC-T cohort. Patients evaluated as SD were defined as ‘Poor response’ and the patients evaluated as PR were defined as ‘Well response’. E Workflow of GOLP and other control groups for the treatment of iCCA-bearing mouse with mIC-22 or AY-LTC2. Created with BioRender.com. F Tumor mass after three cycles of different treatments (t test, two-sided). N= 6 biologically independent animals for each group. Source data are provided as a Source Data file.

Fig. 2. Single-cell profiling of the iCCA microenvironment during GOLP treatment.

A Study diagram and experimental workflow. Created with BioRender.com. B Uniform manifold approximation and projection (UMAP) plot depicting 12 cell lineages from the 18 samples (N = 131,139 cells). C Dot plot showing the average expression levels of marker genes across the 12 cell types. The dot size indicates the fraction of cells expressing the marker gene in each cluster, and the dot color represents the average expression level of the marker gene in each cluster. The dendrogram on the right represents the hierarchical clustering of different cell types based on the expression levels of the indicated marker genes. D Heatmap showing the relationship among the 12 cell types based on the correlation of gene expression between each pair of cell types as assessed by Pearson’s correlation analysis. Blue and red indicate negative and positive correlations, respectively. E Bar plots show the distribution of immune and nonimmune cells in each patient. The samples are presented in pre- and post-GOLP treatment pairs for the 8 patients (Pts. 1–8), and 2 patients underwent surgery without pre-GOLP treatment (Pts. 9 and 10). Pt., patient. For every sample, annotated cell types were divided into immune cell type (T cells, NK cells, pDCs, neutrophils, B cells, and plasma cells: Left) and non-immune cell type (fibroblasts, epithelial cells, endothelial cells, and hepatocytes: Right). For every sample, the percentage of immune cell type plus percentage of non-immune cell type is 100%. F The enrichment scores of different cell types in the pre- and post-GOLP treatment groups. The colors represent different cell types. G, H The Pi (G) and Ti (H) values of major cell types in iCCA patients treated with GOLP (N = 8 paired samples). Pi, predictive index; Ti, therapeutic index. The dot size represents the significance evaluated as -log₁₀ (P value). *: P < 0.05, Wald test, two-sided. Source data are provided as a Source Data file.

Fig. 3. Tumor features correlated with iCCA prognosis and GOLP response.

A Bar plot showing the top enriched pathways of differentially expressed genes between pre- and post-GOLP tumor cells (N = 18 samples). A hypergeometric test was used. A Benjamini–Hochberg adjusted P < 0.05 (two-sided) was used as a significance cutoff. B Correlation between gene features of tumor cells before GOLP treatment and tumor shrinkage (N = 8 patients). Gene features above zero were favorably associated with tumor shrinkage (P < 0.05: red); gene features below zero were unfavorably associated with tumor shrinkage (P < 0.05: blue); Pearson’s correlation test (two-sided); exact P values were listed in Supplementary Data 3. C GSEA showed that PD1 signaling, positive regulation of vasculature development and signaling by KIT in disease were enriched in the pre-GOLP tumor cells in samples with a good response. Empirical phenotype-based permutation test (one-sided). D UMAP plot showing the expression profiles of tumor cells from the 18 samples (N = 14,411 cells). E Heatmap showing the 4 meta-clusters (C1-C4) of tumor cells identified using consensus non-negative matrix factorization (cNMF). The terms on the left represent the pathways enriched in each meta-cluster. F Dot plot showing the average expression levels of representative effector genes in the 4 meta-clusters between the PR and SD groups (N = 8 patients). The dot size indicates the fraction of cells expressing the specific gene in the group, and the dot color represents the average gene expression level in the group. G Pi values of the 4 meta-clusters (C1-C4). N = 8 patients. The dot size represents the significance evaluated as -log₁₀ (P value). *: P < 0.05. Wald test, two-sided. P values of C1, C2, C3, and C4 are 0.0011, 0.57, 0.27 and 0.9, respectively. (H) Heatmap showing the expression profiles of the 4 meta-clusters in the large GOLP treatment-naive iCCA cohort (FU-iCCA, N = 262) with bulk RNA-seq data. The representative genes of each meta-cluster are highlighted on the right. Source data are provided as a Source Data file.

Fig. 4. High proportion of Macro CD5L⁺ in tumors with poor response to GOLP.

A Dot plot showing the top pathways significantly enriched in pre-GOLP and post-GOLP myeloid cells (N = 18 samples). A hypergeometric test was used. Benjamini–Hochberg (BH) adjusted P < 0.05 (two-sided) was used as a significance cutoff. B Dot plot showing the top pathways significantly enriched in myeloid cells of patients with SD or PR (N = 8 patients). A hypergeometric test was used. Benjamini–Hochberg (BH) adjusted P < 0.05 (two-sided) was used as a significance cutoff. C UMAP plot showing myeloid cells clustered into 14 subtypes (N = 20,825 cells). Macro, macrophage; DC, dendritic cell; Mono, monocyte. D Violin plot showing the marker gene expression levels of the 14 myeloid subtypes. The colors in the violin plot represent the median gene expression levels in the subtypes. E Pairwise comparison of overall myeloid cell proportions between the pre-GOLP and post-GOLP groups. Paired Wilcoxon test, two-sided. N = 8 biologically independent patients. The results are depicted in boxplots: center line indicates median, box represents first and third quantiles, and whiskers indicate maximum and minimum values. F Bar plot showing the occupancy score of each myeloid subtype (N = 8 patients). G UMAP plot (N = 20,825 cells) showing the dynamics of Mono CCL20⁺ and Macro FBP1⁺ upon GOLP treatment (N = 8 patients). H Pi values of myeloid subtypes in iCCA patients treated with GOLP. N = 8 patients. The arrow points to Macro CD5L⁺ (P = 0.046), the baseline proportion of which was a strong predictor of unfavorable GOLP response (Wald test, two-sided). I Representative samples from patients with SD and PR (N = 2 samples) stained by IHC with anti-CD5L (green) and anti-CD68 (white) antibodies. Scale bar, 40 μm. J Dotplot showing the baseline Macro CD5L⁺ percentages of the SD and PR groups in three GOLP treatment cohorts (NCT03951597 and FDU-ZS-ICC-T). N = 55 biologically independent patients. Mann–Whitney test, two-sided, P = 0.035. The results are depicted with Mean ± SEM. K Heatmap showing previously defined macrophage signatures across all myeloid subtypes. Macro CD5L⁺ were highlighted by arrow as M2-polarized and anti-inflammatory macrophages with mostly phagocytosis-related functions. Source data are provided as a Source Data file.

Fig. 5. High baseline levels of CD8 GZMB⁺ predicted a favorable response to GOLP treatment.

A UMAP plot (N = 49,053 cells) showing the 12 subtypes of T and NK cells (N = 8 patients). B Heatmap showing the normalized expression of marker genes in lymphoid subtypes. C Proportions of three CD4 subtypes between the pre-/post-GOLP. Paired Wilcoxon test, two-sided. N = 8 biologically independent patients. Depicted in boxplots: center line indicates median, box represents first and third quantiles, and whiskers indicate maximum and minimum values. D The Pi and Ti of CD4⁺ T and NK subtypes in iCCA treated with GOLP. NK GNLY⁺ showed a negative Pi (P = 0.014), CD4 CXCL13⁺ showed a positive Pi (P = 0.047), and CD4 Treg showed a negative Ti (P = 0.033). N = 8 paired samples. Wald test, two-sided. *: P < 0.05. E PDCD1 expression among CD4⁺ T-cell subtypes. N = 8 paired samples. Depicted in violin plots the dot indicates median. F Pi and Ti of CD8⁺ T-cell subtypes in iCCA tumors treated with GOLP. CD8 proliferating (P = 0.0006) and CD8 GZMB⁺ (P = 0.016) showed significantly positive Pi values. CD8 GZMK⁺ showed a positive Ti (P = 0.001), whereas CD8 proliferating (P = 0.0035) and CD8 GZMB⁺ (P = 0.23) showed negative Ti values. N = 8 paired samples. Wald test, two-sided. *: P < 0.05. G Relationship between tumor shrinkage and the log₁₀ normalized alteration ratio of the CD8 GZMK⁺ proportion after GOLP treatment (N = 8 paired samples, Wald test, two-sided, P = 0.001, R = 0.92). H Survival plot showing that high GZMK expression was associated with better prognosis in both the FU-iCCA validation cohort (N = 262, P = 0.034) and the FDU-ZS-ICC validation cohort (N = 329, P = 0.011). Log-rank test, two-sided. I, J Flow cytometry plot and analysis showing the percentages of the three CD8⁺ T-cell subtypes from patients with PR or SD pre-GOLP. t test, two-sided. N = 10 biologically independent patients. Depicted in violin plots: the center line indicates median; the dotted lines represent the first and third quantiles. K, L Representative mIHC and statistics of CD8⁺ T-cell subtypes from patients with PR or SD pre-GOLP. Scale bar, 50 μm. N = 55 biologically independent patients, Mann–Whitney test, two-sided. Depicted in dotplots with Mean ± SEM. Source data are provided as a Source Data file.

Fig. 6. TCR clonotype expansion and influence of CD8 proliferating to CD8 GZMB⁺ to CD8 GZMK⁺ transition on GOLP response.

A TCR clonal types classified by the change upon GOLP therapy (N = 25,893 types, N = 8 patients, Fisher’s exact test, two-sided). See exact P in Supplementary Data 8. B Correlation between tumor change and clonotype status in different T-cell types. N = 8 paired samples, Wald test, two-sided. *: P < 0.05. See exact P in Source Data. C, D Distribution of contracted/expanded clonotypes among the T-cell subtypes and proportions of CD8⁺T subtypes with clonal expansion between the pre- and post-GOLP groups (Wilcoxon rank-sum test, two-sided. N = 8 independent patients). E Exhausted scores of CD8 GZMB⁺ cells with clonal expansion pre- and post-GOLP. Gray dashed line cutoff: exhausted score calculated from CD4 SOCS3⁺ and CD4 naive cells (N = 8 paired samples, 99% confidence level, one-tail test). F Distribution of CD8 GZMB⁺ and CD8 GZMK⁺ in the pre- or post-GOLP groups between patients with PR and SD (N = 21) derived from mIHC analysis in GOLP treatment cohorts. Fisher’s exact test, two-sided. G Flow cytometry analysis showing the alteration ratio of CD8⁺T-cell subtypes from patients with PR or SD post-/pre-GOLP (N = 9). Mann–Whitney U test, two-sided. H TCR clonal similarity across T/NK subtypes. I RNA velocity analysis showing the transition among the three CD8⁺T cells (N = 19,128 cells). J Distribution of exhausted and cytotoxic scores of the three CD8⁺T-cell subtypes pre-/post-GOLP treatment (Reference: CD4⁺T-cell). K Exhausted score of CD8⁺ T cells between PR and SD patients upon GOLP treatment. Mann–Whitney U test, two-sided, N = 8 independent patients. L Change in exhausted score along the trajectory in two representative patients pre-/post-GOLP. M Overall survival with different CD8 GZMK⁺/CD8 proliferating ratio (N = 262, log-rank test, two-sided). N Alterations in ratios of three CD8⁺ T cells in the pre- or post-GOLP groups between patients with PR and SD. Mann–Whitney U test, two-sided, N = 8 independent patients. O Ti of changes in the CD8 GZMK⁺/CD8 proliferating (P = 0.023), CD8 GZMK⁺/CD8 GZMB⁺ (P = 0.385), and CD8 GZMB⁺/CD8 proliferating ratios (P = 0.031). N = 8 paired samples, Wald test, two-sided. *: P < 0.05. P The status and phenotypic shift in CD8⁺ T cells. For all boxplots: center line indicates median, box represents first and third quantiles, and whiskers indicate maximum and minimum values. Source data are provided as a Source Data file. Created with BioRender.com.

Fig. 7. Macro CD5L⁺ exacerbated the exhaustion of CD8 GZMB⁺ in patients with a poor response to GOLP.

A The transition trajectory of CD8 proliferating, CD8 GZMB⁺ and CD8 GZMK⁺ inferred by RNA velocity analysis of samples from patients with SD and PR (N = 19,128 cells). B Proportion of CD8 GZMB⁺ with clonal expansion in SD and PR samples. Fisher’s exact test, two-sided. C Distribution of exhausted scores of CD8 GZMB⁺ in PR and SD samples taken before and after GOLP treatment. Mann–Whitney U test, two-sided, 5631 cells from N = 8 biologically independent patients. D Differentially expressed genes of CD8 GZMB⁺ between the PR and SD group. Benjamini–Hochberg adjusted P < 0.05 was used as a significance cutoff. t test, two-sided. E Top enriched pathways in CD8 GZMB⁺ in the PR and SD groups. Benjamini–Hochberg adjusted P < 0.05 was used as a significance cutoff. Hypergeometric test, two-sided. F Dot plot showing the cytokine signaling strength of CD8 GZMB⁺, CD8 GZMK⁺, CD8 proliferating and the other 14 myeloid subtypes and tumor clusters (C1-C4), N = 8 paired samples. G CellChat analysis showing the network of cytokine signaling pathways in CD8 GZMB⁺, CD8 GZMK⁺, CD8 proliferating and the other 14 myeloid subtypes and tumor meta-clusters (C1-C4). The signal level from Macro CD5L⁺ was highest among the signals to CD8 GZMB⁺, and the signal level to CD8 GZMB was highest among the signals from Macro CD5L⁺. H Scatter plot showing correlations between CD8 GZMB⁺ and Macro CD5L⁺ in the FU-iCCA cohort (N = 262) based on signature gene expression. The regression line ± the 95% confidence interval was displayed, Wald test, two-sided. I A representative sample (N = 1 patient) showing crosstalk between CD8 GZMB⁺ and Macro CD5L⁺, which were stained by mIHC with anti-CD8 (yellow), anti-GZMB (red), anti-CD68 (white) and anti-CD5L (green) antibodies. Scale bar, 20 μm. J Co-culture strategy of CD8⁺ T and Macro CD5L⁺ or the control (Left). The expression of CTLA4, KLRB1 and LAG3 in CD8 GZMB⁺ after co-culture (Right, t test, two-sided, N = 4 biologically independent experiments for each group). For all boxplots: center line indicates median, box represents first and third quantiles, and whiskers indicate maximum and minimum values. Source data are provided as a Source Data file.

Fig. 8. Anti-CTLA4 antibody reversed GOLP resistance in iCCA.

A Workflow of mouse iCCA treatment and the analysis of tumor mass between Macro CD5L⁺ injection group and the control group (N = 5 biologically independent animals for each group, t test, two-sided). Created with BioRender.com. B scRNA-seq confirming the absence of Macro CD5L⁺ in the control group, while the Macro CD5L⁺ injection group had colonization of Macro CD5L⁺ in the tumor environment (N = 32,285 cells). C scRNA-seq showing the six unsupervised clustering group 0-5 of mouse CD8⁺T cells (Top, N = 572 cells); Similar score between the six unsupervised clustering group in mouse CD8⁺T cells and the three key human CD8⁺T cells (Bottom). D The gene expression of Ctla4, Pdcd1, Tox, Tox2, Klrb1 and Lag3 in the human-like CD8 GZMB⁺ clusters (mouse CD8⁺T Cluster0, 1) by scRNA-seq (N = 305 cells, Wilcoxon test, two-sided). E, F Workflow of mouse iCCA treatment and the analysis of tumor mass between the anti-CTLA4 group and the control group (N = 5 biologically independent animals for each group, t test, two-sided). Created with BioRender.com. G Representative samples (N = 2 samples) from patients with PR or SD after GOLP stained by mIHC with anti-CD8 (yellow), anti-CTLA4 (red) antibodies. Scale bar, 20 μm. H Quantitative comparison of CTLA4 expression in CD8⁺T cells between the PR and SD groups using mIHC results in the GOLP treatment cohorts with post-GOLP surgery (NCT03951597 and FDU-ZS-ICC-T) (total of 55 patients, including 21 with post-GOLP surgery; Mann–Whitney U test, two-sided). The results are depicted with Mean ± SEM. Source data are provided as a Source Data file.

Fig. 9. Summary of the findings of this study.

This study highlights: (1) High proportions of CD8 GZMB⁺ and CD8 proliferating cells predicted good responses to GOLP treatment in iCCA; (2) Transition of CD8 proliferating to CD8 GZMK⁺ facilitated a favorable GOLP response in iCCA; (3) Macro CD5L⁺ and CD8 GZMB⁺ crosstalk impaired GOLP response by CTLA4 up-regulation. Created with BioRender.com.