Bile acid synthesis impedes tumor-specific T cell responses during liver cancer

Abstract

The metabolic landscape of cancer greatly influences antitumor immunity, yet it remains unclear how organ-specific metabolites in the tumor microenvironment influence immunosurveillance. We found that accumulation of primary conjugated and secondary bile acids (BAs) are metabolic features of human hepatocellular carcinoma and experimental liver cancer models. Inhibiting conjugated BA synthesis in hepatocytes through deletion of the BA-conjugating enzyme bile acid-CoA:amino acid N-acyltransferase (BAAT) enhanced tumor-specific T cell responses, reduced tumor growth, and sensitized tumors to anti-programmed cell death protein 1 (anti-PD-1) immunotherapy. Furthermore, different BAs regulated CD8⁺ T cells differently; primary BAs induced oxidative stress, whereas the secondary BA lithocholic acid inhibited T cell function through endoplasmic reticulum stress, which was countered by ursodeoxycholic acid. We demonstrate that modifying BA synthesis or dietary intake of ursodeoxycholic acid could improve tumor immunotherapy in liver cancer model systems.

Fig. 1.. Conjugated BAs are elevated in human liver tumors, and their synthesis promotes tumorigenesis in experimental models.

(A and B) Primary BAs were measured in liver tissues from patients with HCC and non-HCC by LC-MS (n = 45). Graphs show the amounts of total detectable BAs (A) and individual BAs (B) as normalized to tissue weight. CA, cholic acid; GCA, glycocholic acid; GDCA, glycodeoxycholic acid. (C) Formalin-fixed paraffin-embedded sections from the same patient biospecimens (available) (n = 39) in (A) were stained with hematoxylin and eosin (H&E) and BAAT and imaged at 20×. Representative brightfield images are shown on the left, and dot plots on the right show the percentage of BAAT expression as measured by the ratio of BAAT⁺ area to the total tissue area. Scale bars, 1 mm. (D and E) AST mice were infected with AAV8-TBG-CRE (1 × 10¹¹ Genome copies (GC) per mouse), and 6 weeks later, BA levels were enumerated by LC–MS. Plots in (D) show the concentrations of total (left) and individual (right) primary unconjugated and conjugated BAs in the tumor interstitial fluid (TIF) of AST^AAV8-TBG-CRE and AST^WT mice (n = 5 to 9). MCA, muricholic acid; HCA, hyocholic acid; THCA, taurohyocholic acid. Mass spectrometry imaging (E) was performed on frozen liver sections, and shown are representative images of TCA and TβMCA in AST^AAV8-TBG-CRE and AST^WT mice (n = 4). The yellow dashed outlines indicate regions of interest. m/z, mass/charge ratio. (F) Schematic representation of BA synthesis pathways; gray boxes highlight enzymes that were knocked out in our experiments. (G to K) AST-CAS9 mice were separately infected with AAV-TBG-CRE expressing guide RNAs (gRNAs) targeting each of the BA synthesis enzymes shown in (F), and livers were profiled 6 weeks later. (G) The heatmap in (G) represents fold changes in BAs in the TIF of various KO tumors in infected mice compared with uninfected AST^WT mice, scaled across columns. Plots in (H) show total levels of primary conjugated BAs within the TIF of various KO tumors relative to sgSCR controls (n = 6 to 16). Shown in (I) are representative images of livers from different KOs. Bar graphs show the number of visible or surface tumors (n = 5 to 20) (J) and the weights of livers at 6 weeks after infection or KO (n = 5 to 13) (K). Data points in each panel represent pooled experiments unless otherwise specified. Data are presented as means ± SEM. Statistical significance was determined by unpaired Welch’s t test for (A) and (C) or one-way analysis of variance (ANOVA) with Dunnett’s multiple comparisons test for (D), (G), (H), (J), and (K). *p < 0.05; **p < 0.01; ***p < 0.001; ****p < 0.0001.

Fig. 2.. Production of conjugated BAs represses antitumor T cells and confers resistance to anti–PD-1 immunotherapy in liver cancer models.

(A to G) AST-CAS9 mice were infected with AAV8-TBG-CRE-gRNA to simultaneously induce SV40 expression and knock out different BA-synthesis enzymes (synthetases). At 3 weeks after AAV infection, Thy1.1⁺ Tag⁺ T cells (1 × 10⁵) were adoptively transferred and analyzed in the liver 15 days later (~5 weeks after tumor induction). In (A), fluorescence-activated cell sorting (FACS) plots on the left show the representative frequency of Thy1.1⁺ Tag⁺ T cells within livers 15 days after transfer, and the scatterplot on the right shows the cumulative results (n = 5 to 16 per group). In (B), representative immunofluorescence images (scale bars, 200 μM) of livers stained with anti-Thy1.1 (yellow) and anti-SV40 (magenta) show Thy1.1⁺ Tag⁺ T cells and SV40⁺ tumor cells within control (sgSCR) and BAAT KO (sgBaat) livers (n = 4). The plot in (C) shows the function (percentage IFN-γ⁺) of Tag⁺ T cells within control (sgSCR) livers or those lacking the various BA synthetases (n = 5 to 16 per group). In (D), the heatmap illustrates the abundance of cytokines within liver TIF isolated from control and various BA synthetase KOs (row Z-score averaged across n = 4 per group). In (E), CITE-seq was performed on CD8⁺ and CD4⁺ T cells isolated from healthy AST^WT or tumor-burdened AST^AAV8-TBG-CRE livers (either from sgSCR control or sgBaat livers). Each dot in the uniform manifold approximation and projection (UMAP) plot represents a single cell, color coded by the cluster type. Stacked bar graphs in (F) show the proportion of different T cell clusters within AST^WT, AST^AAV8-TBG-CRE sgSCR, and sgBaat livers. In (G) is a UMAP plot showing Thy1.1⁺ Tag⁺ T cells based on Thy1.1⁺ protein abundance, as detected using CITE-seq. In (H), AST-Cas9 mice were infected with AAV8-TBG-CRE expressing sgBaat or co-infected with AAVs expressing sgBaat or sgB2m to knock out major histocompatibility complex (MHC) class I. Representative liver images and the number of visible tumors at 7 weeks after infection (n = 7 per group) are shown. (I and J) AST^AAV8-TBG-CRE sgSCR or sgBaat tumor-burdened mice were treated with or without anti–PD-1 for 2 weeks, starting at 3 weeks after infection and coinciding with transfer of Tag⁺ T cells. In (I), representative images of livers (left) and the number of visible tumors (n = 6 to 12 per group) (right) are shown. The data in (J) represent the percentage of Tag⁺ T cells in the liver (n = 6 to 12 per group). (K and L) Human HCC biospecimens were stained with BAAT and CD8 antibody and analyzed by QuPath. Shown in (K) are representative immunohistochemical images showing CD8 and BAAT expression in “high” and “low” BAAT-expressing HCC tissues (n = 20). Insets show a zoomed in view of the indicated regions. Black indicates selected area for zoomed in image. Scale bars, 200 μm; inset scale bars, 50 μm. Shown in (L) are two-dimensional histograms of the CD8⁺ T cell spatial density and BAAT 3,3′-diaminobenzidine (DAB) optical density across all tiles from all samples, colored by cluster, as grouped by k-means clustering. The colored bars at the top show the number of tiles. (M) Heatmap showing normalized counts of T cell and BA synthesis genes for 348 liver hepatocellular carcinoma samples from TCGA and correlation analysis between these genes, including p values. Data points in each panel represent pooled experiments unless otherwise specified. Data are presented as means ± SEM. Statistical significance was determined by unpaired Welch’s t test for (H) or one-way ANOVA with Dunnett’s multiple comparisons test for (A), (C), (I), and (J). *p < 0.05; **p < 0.01; ***p < 0.001.

Fig. 3.. Tumor-specific T cells fail to efflux conjugated BAs, resulting in cell death.

(A) Tag⁺ T cells and PD-1⁻ bystander CD8⁺ T cells were sorted by FACS from tumor-burdened AST^AAV8-TBG-CRE liver tumors at 1 week after transfer, lipids were extracted, and primary BAs were quantified by LC-MS. Data show all detectable BAs normalized to cell number (n = 6). (B) Naïve Tag⁺ T cells were in vitro–activated for 5 days and then cultured overnight (16 hours) with the indicated conjugated and unconjugated primary BAs at increasing concentrations (100, 250, and 500 μM for conjugated BAs and 10, 50, and 100 μM for unconjugated BAs). The bar graph shows the fold change in the percentage of live CD8⁺ T cells compared with phosphate-buffered saline (PBS) and are representative of three independent experiments. (C and D) Tag⁺ CD8⁺ T cells were in vitro–activated and left untreated or cultured with TCA (1 mM) or TCDCA (250 and 500 μM) overnight. In (C), oxygen consumption rates (OCRs) were measured using a Seahorse extracellular flux analyzer (means ± SEM with n = 5 to 8 per group). Data are representative of three independent experiments. FCCP, carbonyl cyanide-p-trifluoromethoxyphenylhydrazone. In (D), ROS levels were measured by CellROX Green staining and flow cytometry (n = 9 to 13 per group). In (E), the FACS plot histogram and data represent changes in ROS levels of Tag⁺ T cells and bystander PD-1⁻ CD8⁺ T cells from liver tumor (sgSCR) at day 7 after transfer. Data are representative of at least three repeats (n = 3). FMO, fluorescence minus one; MFI, mean fluorescence intensity. (F) Tag⁺ T cells (1 × 10⁶ cells per mouse) transduced with MigR1 empty vector (EV) or catalase overexpressing (CAT-OE) retroviruses were adoptively transferred (co-transfer) into tumor-bearing AST^AAV-TBG-CRE mice and then enumerated 7 days later (n = 6 per group). Data shows the percentage of transferred cells gated on Tag⁺ Thy1.1⁺ CD8⁺ T cells (G) UMAP of a feature plot from the Seurat package showing the distribution of gene expression of Slco3a1 within T cell clusters. (H) Tag⁺ T cells and PD-1⁻ CD8⁺ bystander T cells isolated from liver tumors were pulse-labeled with Rh123 in the presence or absence of 100 nM elacridar (an efflux inhibitor), rested for 1 hour to efflux, and then analyzed by flow cytometry. Efflux rates were calculated by comparing minimum/maximum ratios of Rh123 MFI for untreated T cells (Rh123_min) with those of T cells treated with elacridar (Rh123_max) (n = 6 per group). (I and J) Slco3a1 was deleted (sgSlco3a1) or left intact (sgSCR) in naïve Tag⁺ T cells using a CRISPR-Cas9 ribonucleoprotein electroporation method. In (I), T cells were activated in vitro for 5 days, and Rh123 efflux rates were assessed as in (H). Data are representative of at least three repeats (n = 3). In (J), naïve control and Slco3a1 KO Tag⁺ T cells were adoptively transferred into tumor-bearing AST^AAV8-TBG-CRE mice (3 weeks after AAV infection), and their frequencies and numbers were measured within the liver 2 weeks later by flow cytometry (n = 14 per group). (K) Thy1.1⁺ Tag⁺ T cells transduced with control MigR1 retroviruses or those that overexpress SLCO3A1 (isoforms 1 and 2) were adoptively transferred into AST^AAV8-TBG-CRE mice (3 weeks after AAV infection) and enumerated 2 weeks later by flow cytometry. Flow plots (left) show representative frequencies, and the scatter plot (right) shows cumulative numbers of Tag⁺ T cells within livers (n = 9 per group). Data points in each panel represent pooled experiments unless otherwise specified. Data are presented as means ± SEM. Statistical significance was determined by unpaired Welch’s t test for (E), (F) and (H), one-way ANOVA with Dunnett’s multiple comparisons test for (A), and (D), or two-way ANOVA with Tukey’s multiple comparisons test for (B). *p < 0.05; **p < 0.01; ***p < 0.001; ****p < 0.0001.

Fig. 4.. Secondary BAs LCA and isoalloLCA promote T cell dysfunction by regulating ER stress.

(A and B) Secondary BAs were measured in TIF by LC-MS from tumor-bearing AST^AAV8-TBG-CRE and healthy AST^WT mice (n = 5 to 9) (A) and sgSCR and sgBaat tumors (n = 12) (B). (C) Naïve Tag⁺ T cells were in vitro activated for 5 days and then cultured overnight (16 hours) with the indicated secondary BAs at concentrations of 10, 50, and 100 μM or ethanol (control). The bar graph shows the fold change in the percentage of IFNγ⁺ TNF-α⁺ T cells compared with control. Data are representative of three repeats (n = 3 per group). (D) Human peripheral blood mononuclear cells (PBMCs) were activated in vitro with anti-CD2, -CD3 and -CD28 beads for 10 days and later exposed to secondary BAs (UDCA and isoalloLCA) overnight; secreted cytokines were measured by Isoplexis. The heatmap (scaled by row) illustrates the average concentration of secreted cytokines across the samples (n > 400 cells per condition) with statistical significance (p value) indicated. (E) Bulk RNA-seq of in vitro–activated Tag⁺ T cells cultured overnight with different secondary (LCA, isoalloLCA, dehydroLCA, isoLCA, UDCA) and primary (TCA, TCDCA) BAs identified DEGs in T cell activation and ER stress pathways. The heatmaps (scaled by row) represent averaged fragments per kilobase of transcript per million mapped reads (FPKM) values across three samples per condition. (F) Naïve Tag⁺ T cells were CRISPR-Cas9–edited to delete Nr4a1, Nr4a1/2, Atf6, Ern1, Eif2ak3, or scramble control (sgSCR) and then adoptively transferred into tumor-bearing AST^AAV8-TBG-CRE mice (at 3 weeks after AAV infection). The bar graph shows function (percentage IFNγ⁺) in donor Tag⁺ T cells 7 days after transfer (n = 3 to 8 per group). (G to J) Tumor-bearing AST mice at 3 weeks after infection were fed a 0.5% LCA diet or control chow diet at the same time of Tag⁺ T cell transfer. Shown in (G) are representative images of the liver. The number of visible tumor nodules were measured 1 week post LCA diet. Data are representative of three repeats (n = 4 or 5). The FACS plot in (H) represents function (percentage IFN-γ⁺ TNF-α⁺) of Tag⁺ T cells within liver tumors 1 week after adoptive transfer and upon restimulation with phorbol myristate acetate (PMA) and ionomycin. In (I), plots show fold change in expression of NUR77, XBP1, and ATF3 in intratumoral Tag⁺ T cells from mice fed the LCA diet compared with mice fed the control diet. Data are representative of two independent experiments (n = 3 to 5 per group). In (J), the plot shows fold change in IFN-γ⁺ Tag⁺ T cells in various KO or sgSCR Tag⁺ T cells from mice fed the LCA diet compared with mice fed the control diet (n = 4 or 5 per group). Note, Tag⁺ T cells lacking Atf6, Ern1, or Eif2ak3 were not as inhibited by the LCA diet as sgSCR control cells or those lacking Nr4a1/2. (K) The heatmap shows ER stress response gene expression within different CD8⁺ T cell clusters from scRNA-seq data comparing T cells from sgSCR livers with those from sgBaat livers (as described in Fig. 2E). Data points in all panels represent pooled experiments unless otherwise specified. Data are presented as means ± SEM. Statistical significance was determined by unpaired Welch’s t test for (G), one-way ANOVA with Dunnett’s multiple comparisons test for (F) and (J), or two-way ANOVA with Tukey’s multiple comparisons test for (A) to (C) and (I). *p < 0.05, **p < 0.01, ***p < 0.001; ****p < 0.0001.

Fig. 5.. Secondary BA UDCA promotes T cell function and inhibits tumor growth.

(A) The plot shows the concentration of UDCA within liver TIF from different BA-synthesis KOs as measured by LC-MS (n = 6 to 16). (B) The plot illustrates the ratio of UDCA to the sum of all of the LCA-related BA (LCA, isoLCA, isoalloLCA, dehydroLCA) conjugates within liver TIF from the various BA-synthetase KOs as measured by LC-MS (n = 6 to 16). (C to E) At 4 weeks after AAV infection, tumor-bearing AST^AAV8-TBG-CRE mice were either fed a 1% UDCA, 0.5% LCA, or control chow diet concurrent with the transfer of Tag⁺ T cells. In (C), the plot illustrates the ratio of UDCA to the sum of all of the LCA-related BA (LCA, isoLCA, isoalloLCA, dehydroLCA) conjugates within liver TIF as measured by LC-MS (n = 4 or 5 per group). Shown in (D) are representative liver images (left) and a plot of liver weights (right) two weeks after feeding UDCA diet. Data shown are representative of three independent experiments (n = 4 per group). The plot in (E) shows the percentage of Gzmb⁺ Tag⁺ T cells, 2 weeks after transfer. Panels (A), (B) and (E) represent data from pooled experiments. Data are presented as means ± SEM. Statistical significance was determined by unpaired Welch’s t test for (D) and (E) or one-way ANOVA with Dunnett’s multiple comparisons test for (A) to (C). *p < 0.05; **p < 0.01; ****p < 0.0001.