The immune-dependent and SERT-independent anti-tumor effects of citalopram in HCC. (A, B) Mouse HCC cells, Hepa1-6 or Hep53.4, were subcutaneously injected into Rag1-/- C57BL/6 or immunocompetent C57BL/6 mice (n = 5-7 per group). When bore visible tumors, 5 mg/kg citalopram was treated daily for 15 or 25 days. Tumors were excised after mice were sacrificed, and the tumor weight was measured. (C, D) Western blotting showed the knockdown efficiency of SERT in Hepa1-6 and Hep53.4 cells. (E, F) shControl and shSlc6a4 Hepa1-6 and Hep53.4 cells were subcutaneously injected into Rag1-/- C57BL/6 or immunocompetent C57BL/6 mice (n = 5-6 per group). When bore visible tumors, 5 mg/kg citalopram was treated daily for 15 or 25 days. Tumors were excised after mice were sacrificed, and the tumor weight was measured. In all panels, *p < 0.05, **p < 0.01, ***p < 0.001. Values as mean ± SD and compared by one-way ANOVA multiple comparisons with Tukey’s method (for bar chart comparison) and two-way ANOVA with Dunnett’s multiple comparisons (for survival curve comparison).

C5aR1 is a direct target of citalopram.

(A) Gene Set Enrichment Analysis (GSEA) identified GLUT1 and C5aR1 as two top hits related to SSRI-induced gene changes. (B) GSEA of HCC RNAseq data (TCGA cohort) with the SSRI-related gene signature. Sample grouping was made based on the median expression of C5aR1. (C) Representative immunohistochemical images showed the expression pattern and cellular distribution of C5aR1 in human HCC tissues. Scale bar, 50 μm. (D) Single cell RNA sequencing analysis showed the expression pattern of C5aR1 with the immune microenvironment of HCC. (E) Co-immunofluorescence of C5aR1 (green) with CD163 (red) in HCC samples. Scale bar, 10 μm. (F) The DARTS assay and immunoblot analysis showed C5aR1 protein stability against 5 μg/mL pronase in the presence and absence of 100 μM citalopram treatment. (G) The DARTS assay and immunoblot analysis showed C5aR1 protein stability against 5 μg/mL pronase in the presence of different concentration of citalopram treatment. (H) The overall conformation of citalopram binding to C5aR1. (I) Representative models of citalopram in pose-1 (left), pose-2 (middle) and allosteric site (right). Several polar interactions were indicated by black dashed lines. (J) HEK293T cells were transfected with either WT or mutant C5aR1 expression plasmids for 48 h, followed by DARTS assay with immunoblotting analysis of C5aR1 protein levels. In all panels, *p < 0.05, **p < 0.01. Values as mean ± SD and compared by the Student’s t test (F) or one-way ANOVA multiple comparisons with Tukey’s method among groups (G, J).

Citalopram targets C5aR1+ TAMs

(A) Western blotting showed the knockdown efficiency of GLUT1 in mouse Hepa1-6 cells. (B) GLUT1KD Hepa1-6 cells were subcutaneously injected into the Rag1-/- or immunocompetent C57BL/6 mice, and mice were treated with 5 mg/kg citalopram when bore visible tumors; three weeks later, tumor burden was examined (n = 6-7 per group). (C) The growth kinetics of GLUT1KD Hepa1-6 tumors in C5ar1+/- and C5ar1-/- C57BL/6 host (n = 7). (D) Immunofluorescence analysis of C5a deposition in GLUT1KD Hepa1-6 tumors from C5ar1+/- and C5ar1-/- C57BL/6 host. Scale bar, 50 μm. (E, F, H) GLUT1KD Hepa1-6 cells admixed with C5ar1+/- and C5ar1-/- BMDMs from donor (d) mice were subcutaneously implanted into syngeneic recipient (r) mice. The therapeutic effect of citalopram (E), C5a deposition (F), and macrophage phagocytosis (H) in this model were analyzed. Scale bar, 50 μm. (G) The phagocytic capacity of macrophages isolated from GLUT1KD Hepa1-6 tumors in C5ar1+/- and C5ar1-/- C57BL/6 host. (I-K) Flow cytometry showed the infiltration of CD45+CD11b+F4/80+ macrophages (I), CD206+ TAMs and CD11b+ TAMs (J), tumor-infiltrating lymphocytes (K) in tumor tissues from orthotopic xenograft model, which generated in immunocompetent C57BL/6 mice with Hepa1-6 cells (n = 5 per group). (L, M) Measurement of CD8+ T cell function in tumor tissues from the groups mentioned in C and E. (N) The growth kinetics of GLUT1KD Hepa1-6 tumors in C5ar1+/- and C5ar1-/- C57BL/6 host upon CD8+ T cell deletion (n = 7). (O) Correlation analysis of C5aR1 expression and immune checkpoint molecules, gene signatures of TAMs, exhausted T cells, and effector Tregs in the TCGA cohort (n = 371). In all panels, *p < 0.05, **p < 0.01, ***p < 0.001; ns, non-significant. Values as mean ± SD and compared by two-way ANOVA with Dunnett’s multiple comparisons (B, C, E, N), Student’s t test (G, H, I-L), one-way ANOVA multiple comparisons with Tukey’s method (B, M), and the Spearman’s rank correlation methods (O).

Citalopram activates CD8+ T cells.

(A, B) Measurement of CD8+ T cell function and glycolysis in orthotopic tumor tissues from WT C57BL/6 mice (n = 5 per group). (C, D) Measurement of CD8+ T cell function and glycolysis in orthotopic tumor tissues from MASH mice (n = 5 per group); Basal ECAR indicates glycolysis after the addition of glucose, and ΔECAR represents the difference between oligomycin-induced ECAR and 2-DG-induced ECAR. (E) Serum 5-HT levels in GLUT1KD Hepa1-6 tumor-bearing mice fed with chow diet or CDAHFD, with the presence or absence of citalopram treatment (n = 5 per group). (F) Serum TNF-α, IL-1β, and IL-6 levels in GLUT1KD Hepa1-6 tumor-bearing mice fed with chow diet or CDAHFD, with the presence or absence of citalopram treatment (n = 5 per group). (G) Serum 5-HT levels in WT C57BL/6 and Tph1-/- mice, with the presence or absence of citalopram treatment (n = 5 per group). (H) Tumor growth of WT and Tph1-/- mice after subcutaneous injection of Hepa1-6 cells and treatment with citalopram. (I) Measurement of CD8+ T cell function in tumor tissues from the groups mentioned in H. (J) The therapeutic effect of citalopram on GLUT1KD Hepa1-6 tumor was tested in the presence or absence of CD4+ T or CD8+ T cell deletion. In all panels, *p < 0.05, **p < 0.01, ***p < 0.001; ns, non-significant. Values as mean ± SD and compared by the Student’s t test (A-F), one-way ANOVA multiple comparisons with Tukey’s method (I), and two-way ANOVA with Dunnett’s multiple comparisons (H, J).

Mechanism model.

Model depicting the molecular mechanism by which citalopram inhibit the Warburg effect and elicit an anti-tumor response in HCC. Citalopram not only inhibits glycolytic metabolism of cancer cells by targeting GLUT1 (Ref 16) but also enhances macrophage-driven anti-tumor immunity and induces a systemic immunostimulatory effect on CD8+ T cell functions through serotonergic mechanisms.

Citalopram inhibits HCC cell proliferation and promotes cell apoptosis in immune-competent and immune-deficient mouse models.

(A, B) Immunohistochemical analysis of cleaved caspase-3 (CCS3) and Ki67 in Hepa1-6-bearing subcutaneous xenograft tumors from Rag1-/- or immunocompetent C57BL/6 mice, treated with DMSO or 5 mg/kg citalopram. (C, D) Immunohistochemical analysis of CCS3 and Ki67 in Hep53.4-bearing subcutaneous xenograft tumors from Rag1-/- or immunocompetent C57BL/6 mice, treated with DMSO or 5 mg/kg citalopram. In all panels, **p < 0.01, ***p < 0.001. Scale bar, 50 μm. Values as mean ± SD and compared by the Student’s t test.

Citalopram suppresses HCC cell proliferation and promote cell apoptosis in a SERT-independent manner.

(A) Plate colony formation assay revealed the effect of SERT knockdown alone or combined with citalopram treatment on the long-term cell proliferation of Hepa1-6 and Hep53.4 cells (n = 6 per group). (B) Caspase-3/7 activity in Hepa1-6 and Hep53.4 cells upon SERT knockdown or combined treatment with 5 μM citalopram (n = 6 per group). (C, D) Immunohistochemical analysis of CCS3 and Ki67 in shControl and shSlc6a4 Hepa1-6-bearing subcutaneous xenograft tumors from Rag1-/- or immunocompetent C57BL/6 mice, treated with DMSO or 5 mg/kg citalopram. Scale bar, 50 μm. (E, F) Immunohistochemical analysis of CCS3 and Ki67 in shControl and shSlc6a4 Hep53.4- bearing subcutaneous xenograft tumors from Rag1-/- or immunocompetent C57BL/6 mice, treated with DMSO or 5 mg/kg citalopram. Scale bar, 50 μm. In all panels, *p < 0.05, **p < 0.01, ***p < 0.001. Scale bar, 50 μm. Values as mean ± SD and compared by one-way ANOVA multiple comparisons with Tukey’s method among groups.

Identification of C5aR1 as a direct target for citalopram. (A) The DARTS assay and immunoblot analysis showed C5aR1 protein stability against 5 μg/mL pronase in the presence of different concentrations of SSRIs treatment (0, 1, 10, 50, and 100 μM). (B) For C5aR1, the predicted binding energy distribution of the clusters with poses more than 50. (C) Sequencing analysis showed the successful generation of six C5aR1 mutants. (D) The best-scored complex models of C5aR1 with other four different SSRIs. (E) HEK293T cells were transfected with either WT or mutant C5aR1 (D282A) expression plasmids for 48 h, followed by DARTS assay with immunoblotting analysis of C5aR1 protein levels. In all panels, *p < 0.05. Values as mean ± SD and compared by one-way ANOVA multiple comparisons with Tukey’s method among groups (E).

Macrophage depletion mitigates the anti-tumor effects of citalopram.

(A) Schematic depicting macrophage blockade with clodronate liposomes. 12 days before tumor inoculation, C57BL/6 mice were pretreated with clodronate liposomes or phosphate-buffered saline (PBS) liposomes. Subsequently, GLUT1KD Hepa1-6 cells were subcutaneously injected into the C57BL/6 mice. The effect of citalopram (5 mg/kg) on the tumor burden was evaluated after 18 days of drug treatment. (B) Immunofluorescence analysis of F4/80+ macrophages in the liver and tumor tissues of indicated groups. (C) In C57BL/6 mice, the effect of citalopram on the GLUT1KD Hepa1-6 xenograft tumors was measured in the presence of macrophage deletion (n = 7 per group). (D) Immunohistochemical analysis of CCS3 and Ki67 in GLUT1KD Hepa1-6-bearing subcutaneous xenograft tumors, treated with DMSO or 5 mg/kg citalopram (n = 7 per group). Scale bar, 50 μm. Values as mean ± SD and compared by the Student’s t test (C, D) and two-way ANOVA with Dunnett’s multiple comparisons (C).

Citalopram reverses C5a-mediated macrophage phagocytic impairment via targeting C5aR1.

(A) Western blotting and immunofluorescence analysis showed SERT protein levels in Cas9-sgControl, -sgC5ar1 THP-1 subclones. (B) Effects of C5aR1 deficiency on the macrophage phagocytosis of HCC-LM3 in the presence or absence of C5a stimulation. (C) Effects of different SSRIs on the macrophage phagocytosis of HCC-LM3 in the presence of C5a stimulation. (D) Reconstituted expression of WT and D282A mutant C5aR1 in C5aR1KO THP-1 cells. (E) The effects of citalopram on macrophage phagocytosis in the absence of C5aR1 with reconstituted expression of C5aR1WT or C5aR1D282A. In all panels, *p < 0.05, **p < 0.01, ***p < 0.001. Values as mean ± SD and compared by one-way ANOVA multiple comparisons with Tukey’s method among groups. Data are representative of three independent experiments (B, C, E).

Predicted binding modes of citalopram to human and mouse C5aR1. (A) Conformations of orthosteric binding sites in human (light blue) and mouse (orange) C5aR1. The conformation of human C5aR1 was obtained from the crystal structure (PDB id: 6c1q). The structure of mouse C5aR1 was predicted using the ColabFold (AlphaFold2) software. (B) The predicted binding modes of citalopram to human (light blue) and mouse (orange) C5aR1. The conformations of citalopram were shown in pink (binding mode 1) or deep green (binding mode 2) sticks. For mouse C5aR1, green sticks indicate residues set to flexible in the molecular docking process.

Flow cytometry analysis of immune cells in tumor tissues and spleen.

(A, B) Gating strategies used for flow cytometry analysis of tumor and splenic lymphocytes. Panel A: Identification of CD4+ T cells, CD8+ T cells, and dendritic cells (DC). Panel B: Identification of B220+ B cells, tumor-associated macrophages (TAMs), and nature killer (NK) cells. (C) Flow cytometry showed the infiltration of CD45+CD11b+F4/80+ macrophages, CD4+ T cells, CD8+ T cells, B220+ B cells, CD11c+ DC cells, and NK1.1+ NK cells in spleen tissues from orthotopic xenograft model, which generated in immunocompetent C57BL/6 mice with Hepa1-6 cells (n = 5 per group). Values as mean ± SD and compared by the Student’s t test.

The expression pattern of Glut1 and Glut3 in CD8+ T cells.

(A) Real-time qPCR analysis of Glut1 and Glut3 expression in intratumoral CD8+ T cells (n = 3 per group). (B) Single cell analysis of the expression of glucose transporter members in two HCC cohorts. Values as mean ± SD and compared by the Student’s t test (A), **p < 0.01.

The in vitro effects of citalopram treatment on CD8+ T cell function.

(A) Splenic CD8+ T cells isolated from C57BL/6 mice were stimulated with plate-bound α-CD3 plus α-CD28 for 72 h, and T-cell proliferation (CFSE staining) upon citalopram treatment was determined by flow cytometry (n = 3 per group); ACT: activated T cells. (B) Splenic CD8+ T cells from C57BL/6 mice were stimulated with plate-bound α-CD3 and α-CD28 for 72 h, and CD44 and CD62L levels in activated CD8+ T cells upon citalopram treatment were determined by flow cytometry (n = 3 per group). (C-E) Intracellular GZMB, IFN-γ, and TNF-α were detected in activated CD8+ T cells upon citalopram treatment (n = 3 per group). In all panels, **p < 0.01, ***p < 0.001. Values as mean ± SD and compared by the Student’s t test.