Low-dose radiotherapy combined with dual PD-L1 and VEGFA blockade elicits antitumor response in hepatocellular carcinoma mediated by activated intratumoral CD8+ exhausted-like T cells

Abstract

Atezolizumab (anti-PD-L1) combined with bevacizumab (anti-VEGFA) is the first-line immunotherapy for advanced hepatocellular carcinoma (HCC), but the number of patients who benefit from this regimen remains limited. Here, we combine dual PD-L1 and VEGFA blockade (DPVB) with low-dose radiotherapy (LDRT), which rapidly inflames tumors, rendering them vulnerable to immunotherapy. The combinatorial therapy exhibits superior antitumor efficacy mediated by CD8⁺ T cells in various preclinical HCC models. Treatment efficacy relies upon mobilizing exhausted-like CD8⁺ T cells (CD8⁺ Tex) with effector function and cytolytic capacity. Mechanistically, LDRT sensitizes tumors to DPVB by recruiting stem-like CD8⁺ Tpex, the progenitor exhausted CD8⁺ T cells, from draining lymph nodes (dLNs) into the tumor via the CXCL10/CXCR3 axis. Together, these results further support the rationale for combining LDRT with atezolizumab and bevacizumab, and its clinical translation.

Fig. 1. Combined LDRT and DPVB (LR-DPVB) confers tumor responsiveness.

a Schematic illustration of the in vivo study evaluating different regimes in Hepa1-6 HCC model. b Representative images of Hepa1-6 HCC model liver MRI scanning. c Waterfall plot of the percentage change of the tumor volume for Hepa1-6 HCC model with different treatments. n = 14 mice/group. d Kaplan–Meier curve of Hepa1-6 HCC model mice treated with 4 different regimes (n = 14 mice/group). e Schematic illustration of the in vivo study evaluating different regimes in DEN + CCI₄ HCC model. f Representative image of DEN + CCI₄ HCC model. g The liver weight of DEN + CCI₄ HCC model. h The tumor number of DEN + CCI₄ HCC model. i H&E staining for livers from DEN+CCl₄ HCC model (top), representative images (bottom) of CD8 IHC staining in sections of tumor tissue in indicated groups. Scale bars: 50 µm. j Quantification of tumor cells (X10³) / tissue area (mm²) and CD8⁺ T cells (left) and CD8⁺ T cells (X10³) / tissue area (mm²) (right) of sections in indicated groups of DEN+CCl₄ tumors. k Schematic illustration of the in vivo study evaluating different regimes in Trp53^KO/MYC^OE HCC model. l Representative image of Trp53^KO/MYC^OE HCC model. m The liver weight of Trp53^KO/MYC^OE HCC model. n Kaplan–Meier curve of Trp53^KO/MYC^OE HCC model (n = 10 mice/group). o H&E staining for livers from Trp53^KO/MYC^OE HCC model (top), representative images (bottom) of CD8 IHC staining in sections of tumor tissue in indicated groups. Scale bars: 50 µm. p Quantification of tumor cells (X10³) / tissue area (mm²) (left) and CD8⁺ T cells (X10³) / tissue area (mm²) (right) of sections in indicated groups of Trp53^KO/MYC^OE tumors. Data shown as means ± SD. g–h, j, l–m and p, n = 5 mice/group. P values of d and n were determined by log-rank test (Mantel-Cox). P values of g, h, j and p were calculated using a two-sided unpaired Student’s t test. Source data are provided as a Source Data file.

Fig. 2. The antitumor effect of LR-DPVB is mediated by CD8⁺ T cells.

a, b tSNE maps of scRNAseq data from TIICs of Hepa1-6 HCC tumors (n = 3 mice/ group). c Fold-change of TIICs populations from LR-DPVB and CON in Hepa1-6 HCC tumors. n = 3 mice/group. d Representative flow cytometry plots of % CD4⁺ and % CD8⁺ in CD45⁺ cells of Hepa1-6 HCC tumor TIICs. e Percentage quantification (left panel) of % CD4⁺ and % CD8⁺ in CD45⁺ cells and immune cell count/tumor weight (right panel) in Hepa1-6 HCC tumor TIICs. n = 3 mice/group. f, g Representative images f and quantification g of CD8 IHC staining in serial sections of Hepa1-6 HCC tumor tissue in indicated groups. Scale bars: 500 µm. n = 5 mice/group. h tSNE maps of scRNAseq data from CD45⁺ TIICs of DEN + CCI₄ HCC tumors (n = 3 per group). i Fold-change of CD45⁺ TIICs populations from LR-DPVB and CON groups in DEN + CCI₄ HCC tumors. n = 3 mice/group. j Representative flow cytometry plots of % CD4⁺ and % CD8⁺ in CD45⁺ cells in DEN + CCI₄ HCC tumor TIICs. k Percentage quantification (left panel) of % CD4⁺ and % CD8⁺ in CD45⁺ cells and immune cell count/liver weight (right panel) in DEN + CCI₄ HCC tumor TIICs. n = 3 mice/group. l Representative flow cytometry plots of % CD4⁺ and % CD8⁺ in CD45⁺ cells in Trp53^KO/MYC^OE HCC TIICs at the end of the therapeutic cycle. m Percentage quantification (left panel) of % CD4⁺ and % CD8⁺ in CD45⁺ cells and immune cell count/liver weight in Trp53^KO/MYC^OE HCC TIICs. n = 3 mice/group. Data of e, k and m shown as means ± SD derived from tumor mouse models (n = 3 mice/group). Data of g shown as means ± SD derived from tumor mouse models (n = 5 mice/group). P values of e, g, k and m were calculated using a two-sided unpaired Student’s t test. Source data are provided as a Source Data file.

Fig. 3. LR-DPVB expands antitumor CD8⁺ Tex exhibiting states of effector function and cytolytic capacity in Hepa1-6 tumors.

a UMAP plots of T cells scRNA-seq data (n = 3 per treatment, collected at the end of therapeutic cycle 2). Left and right: contour plots reveal cell density of indicated groups. Using the ProjecTILs method, T cells were subdivided into CD8_Tex, CD8_Tpex, CD8_ effector memory (CD8_EM), CD8_ early activated (CD8_EA), CD8_ naïve like (CD8_NL), CD4_ naïve like (CD4_NL), Treg, T helper 1 (Th1) and T follicular helper (Tfh). b Hepa1-6 HCC tumors T cells expressing indicated genes across major cell subsets and their corresponding average expression. c Fold-change of T cell subsets of Hepa1-6 HCC tumors following LR-DPVB vs DPVB. d Violin plots representing the expression of various effected function and cytotoxicity markers in CD8_Tex cell subsets of Hepa1-6 HCC tumors with indicated treatment. e Density profiles show the distribution of CD8_Tex and CD8_Tpex of Hepa1-6 HCC tumors along the pseudotime trajectory. f Pseudotime trajectory analysis of CD8_Tpex and CD8_Tex clusters was performed by the Monocle tool. Left: Reference trajectory map of Hepa1-6 HCC tumor in all groups; right: pseudotime trajectory maps of indicated groups. g Fold-change of exhausted CD8_T cell subsets in Hepa1-6 HCC tumors following LR-DPVB vs DPVB. h Violin plots representing the expression of various effected function and cytotoxicity markers in CD8_Tef cell subsets of Hepa1-6 HCC tumors with indicated treatment. i Quantification of flow cytometry plots of % Prf1⁺, % Gzmb⁺, % Tnf-α⁺ and % Ifn-γ⁺ in TCF1⁺ PD-1⁺ CD8⁺ T cells in Hepa1-6 HCC tumors of the indicated groups. n = 3 per group. P values were calculated using a two-sided unpaired Student’s t test. j, k Table j and histogram k describes Treg: CD8 or CD4 ratios in Hepa1-6 HCC tumors of LR-DPVB and DPVB groups. Data shown as means ± SD derived from tumor mouse models (n = 3 mice/group). Source data are provided as a Source Data file.

Fig. 4. The stem-like CD8⁺ Tpex enriched by LDRT expanded the antitumor effect of DPVB.

a Fold-change of T cell subsets following LDRT vs CON in Hepa1-6 tumors. n = 3 mice/group. b Representative flow cytometry plots (left panel) and quantification (right panel) of % TCF1⁺ PD-1⁺ in CD8⁺ T cells of the indicated groups. n = 3 samples/group. c, d Illustration of the in vivo adoptive transfer experiment, as described above in methods. n = 3 mice/group. e Representative image of Hepa1-6 liver orthotopic tumors at the end of the second therapeutic cycles (circled by yellow lines). n = 3 mice/group. f Representative flow cytometry plots of % Gzmb⁺ in CD8⁺ T cells (upper panel) and TUNEL assay (lower panel) of the indicated groups. n = 3 samples/group. Scale bars: 100 µm. g Quantification of % Gzmb⁺ in CD8⁺ T cells (left panel) and (%) apoptosis cells (right panel) of the indicated groups. n = 3 samples/group. h Tumor growth curves of Hepa1-6 bearing mice of the indicated days. n = 5 mice/group. P values were calculated using One-way repeated-measures ANOVA test. Data shown as means ± SD. i Graphical abstract of PDTF collection, culture and treatment strategy. j Flow cytometry assays (left panel) and histopathological analysis (H&E staining and TUNEL assay, right panel) of the PDTFs after 48 h of ex vivo T + A therapy. n = 3 samples/group. Scale bars: 100 µm. k Quantification of % TCF1⁺ PD-1⁺ CD8⁺ in CD8⁺ T cells (left panel) and (%) apoptosis cells (right panel) of the indicated groups (n = 3 samples/group). P values of b, g and k were calculated using a two-sided unpaired Student’s t test. Data of b and g shown as means ± SD derived from tumor mouse models (n = 3 mice/group). Source data are provided as a Source Data file.

Fig. 5. Intratumoral infiltration of stem-like CD8⁺ Tpex from dLNs after LDRT through the CXCL10/CXCR3 axis.

a Bubble chart of DEG pathway enrichment of TILs in LDRT and CON groups. b Anatomical view of mouse liver dLNs (left panel, Scale bars: 1 mm) and H&E staining (right panel, Scale bars: 100 µm) of the dLNs sections (n = 3 mice/group). c Representative flow cytometry plots (left panel) and quantification (right panel) of % TCF1⁺ PD-1⁺ in CD8⁺ T cells and % Ki-67 in TCF1⁺ PD-1⁺ CD8⁺ T cells in dLNs of the indicated groups. d Schematic illustration of peripheral drainage of lymphoid tissues blockade experiment in Hepa1-6 bearing LDRT treated mice. e Representative flow cytometry plots (left panel) and quantification (right panel) of % TCF1⁺ PD-1⁺ in CD8⁺ T cells in the tumors of the indicated groups. f Bubble chart of KEGG enrichment pathways of RNA-seq in LDRT vs CON groups. n = 3 mice/group. g Heatmap of the DEGs in the cytokine-cytokine receptor interaction pathway. n = 3 mice/group. h Representative flow cytometry plots (left panel) and quantification (right panel) of % CXCR3⁺ in CD8⁺ T cells, % TCF1⁺ PD-1⁺ in CXCR3⁺ CD8⁺ T cells and CXCR3^- CD8⁺ T cells in the Hepa1-6 tumors of the indicated groups. Data of c, e and h shown as means ± SD derived from tumor mouse models (n = 3 mice/group). P values of c, e and h were calculated using a two-sided unpaired Student’s t test. Source data are provided as a Source Data file.

Fig. 6. The relevance of stem-like CD8⁺ Tpex infiltration in the response T + A-treated HCC patients and the Treg/Tef ratio in the prognosis of T + A-treated HCC patients and all HCC patients.

a (Left panel) Representative MRI of pre and post T + A in T + A response and non-response HCC. (Right panel) Representative mIHC image of stem-like CD8⁺ Tpex in T + A response and nonresponse HCC. Scale bars: 50 µm. b Quantification of mIHC of the TCF1⁺ PD-1⁺ CD8⁺ cells of CD8⁺ cells (%) in the indicated groups. T + A non-response: n = 6. T + A response: n = 3. c Graphs depict correlation between the Treg/Tef ratio and tumor diameter change from baseline (%). p value and R² value were calculated using linear regression analysis (n = 9). d Quantification of flow cytometry plots of the Treg/Tef ratio in the PBMCs and TILs of the indicated groups. Relapse: n = 4. non-relapse: n = 16. e Representative IF image of Treg/Tef in HCC. Scale bars: 100 µm. f, g Kaplan–Meier curve of 5-year OS (upper panel) and 5-year RFS (lower panel) for HCC patients with low Treg/Tef ratio (Treg/Tef -low; n = 60) versus those with high Treg/Tef ratio (Treg/Tef - high; n = 60). P values of b and d were calculated using a two-sided unpaired Student’s t test. P values of f, g were determined by two-sided log-rank test (Mantel-Cox). Data of b and d shown as means ± SD. Source data are provided as a Source Data file.

Fig. 7. The overview of the mechanism and experimental model investigating the synergistic antitumor effect of combined LDRT with T + A.

The first panel: Schematic diagram showing the mechanism antitumor effect of HCC by LDRT combined with T + A activate the intratumoral CD8⁺ Tex. The second panel: The scRNA-seq of the HCC mouse model with different treatments. The third panel: Relationship of stem-like CD8⁺ Tpex and Treg/Tef with the efficacy of T + A treatment and the prognosis of HCC.