TBC1 domain family member 23 is essential for STING-mediated anti-melanoma effect

TL;DR

TBC1D23 is crucial for the STING pathway's ability to fight melanoma by supporting immune cell activation and tumor suppression.

Contribution

This study establishes TBC1D23 as a critical regulator of STING-mediated antitumor immunity.

Findings

High TBC1D23 expression correlates with immune infiltration and better melanoma prognosis.

TBC1D23 deficiency impairs STING signaling, reduces chemokine expression, and weakens CD8⁺ T cell function.

Loss of TBC1D23 accelerates melanoma progression by suppressing antitumor immunity.

Abstract

The innate immune cyclic GMP–AMP synthase–stimulator of interferon genes (cGAS–STING) pathway plays a central role in antitumor immune responses. Cytosolic DNA recognition by cGAS leads to cGAMP production, activating STING and downstream effectors TANK-binding kinase 1 (TBK1) and Interferon regulatory factor 3 (IRF3). This signaling induces the production of type I interferons (IFN-I) and chemokines, facilitating CD8⁺ T cell activation and recruitment to enhance antitumor immunity. The cGAS–STING pathway is critically regulated by vesicular trafficking, which governs its spatially resolved activation across multiple subcellular compartments, thereby facilitating precise control of innate immune signaling. TBC1 domain family member 23 (TBC1D23), a member of the Tre2–Bub2–Cdc16 (TBC) family, is a key regulator of intracellular vesicle transport and is required for trafficking TBK1 from endosomes to the trans-Golgi network. Although Tbc1d23 deletion impairs STING-induced IFN-I production, its physiological role in antitumor immunity remains unclear. Here, we report that high TBC1D23 expression correlates with enhanced immune infiltration and favorable prognosis in melanoma. Conversely, TBC1D23 deficiency severely impairs the STING-dependent expression of key IFN-inducible chemokines, suppresses immunostimulatory macrophage maturation, and diminishes the infiltration, activation, and cytotoxic function of tumor-antigen-specific CD8⁺ T cells. Consequently, loss of TBC1D23 accelerates melanoma progression. Our findings establish TBC1D23 as an essential regulator of STING signaling and a critical mediator of antitumor immunity.

The online version contains supplementary material available at 10.1186/s43556-026-00418-3.

Figures

High TBC1D23 expression is correlated with greater immune cell infiltration in melanoma. a Expression level of TBC1D23 in melanoma (n = 558) and adjacent normal (n = 461) tissues from TCGA database was analyzed using the GEPIA 2 on 12 October, 2024 ([http://gepia2.cancer-pku.cn/#analysis](http://gepia2.cancer-pku.cn/#analysis)). Bars indicate the median expression level in each group. Statistical analysis was performed using t test. b Prognostic value of TBC1D23 in melanoma was analyzed using the Kaplan–Meier Plotter database ([http://kmplot.com/analysis/](http://kmplot.com/analysis/)), which

Loss of TBC1D23 impairs STING-mediated antitumor effects in melanoma models. a Experimental workflow to determine how TBC1D23 cooperates with the STING pathway to regulate antitumor immunity. WT and *Tbc1d23* KO mice were inoculated with 2 × 10^5^ B16F10 cells subcutaneously. Six days after inoculation, the mice were treated with vehicle or DMXAA. Seven days later, tumor tissues were collected and analyzed. DMXAA were generated by Targetmol Biotechnology Co., Ltd(T6273). b-d Representative images of tumors (b), tumor volume (c) and tumor weights (d) in WT and *Tbc1d23* KO mice treated with veh

Loss of TBC1D23 comprises STING-mediated anti-tumor immunity. a,b Heatmap of differentially expressed genes (DEGs) of CD45^+^ tumor-infiltrating immune cells between tumor bearing WT and *Tbc1d23* KO mice treated with vehicle or DMXAA. B16F10 tumors were isolated as in Fig. [2](#Fig2)a, and CD45^+^ tumor-infiltrating immune cells were isolated for bulk RNA-seq analysis. b The indicated dynamic DEGs with six different expression patterns and showed as heatmap via ClusterGVis package. The corresponding gene expression trajectories and representative GO terms were shown in the right panels

TBC1D23 is critical for STING-dependent infiltration of immune cells. a, b Representative FACS data of frequency (a) and quantification (b) of tumor infiltrating CD45^+^ T cells or CD8^+^ T cells (n = 5 per group) of the mice as in Fig. [3](#Fig3)a. c B16F10 tumors were isolated as in Fig. [2](#Fig2)a, then stained with anti-CD45-cy3, anti-F4/80-cy5, anti-CD8-spAqua or anti-GZMB-FITC, and counterstained with DAPI. F4/80 is marker of macrophages. Scale bar, 50 μm. d Quantitative analysis of the numbers of CD45^+^ cells, macrophages, CD8^+^ T cells and GZMB^+^ CD8.^+^ T cells within same area of

Deletion of TBC1D23 in macrophages impairs STING signaling -dependent chemokine production. a Volcano plot of differentially expressed genes. WT and *Tbc1d23* KO BMDMs were treated with DMXAA for 4 h, RNA was extracted and subjected to microarray analysis. Red and green dots show upregulated and downregulated genes in *Tbc1d23* KO BMDM, respectively. b KEGG pathway enrichment analysis of differentially expressed genes as in (a), and the top 20 pathways were shown. Interferon related pathways were shown in red color. (c-f) WT and *Tbc1d23* KO BMDMs were treated with DMXAA (25 μg/mL) for 4 h, an

Deletion of TBC1D23 in macrophages impairs STING-mediated antigen presentation. a, b Mean fluorescent intensity (MFI) of MHC-II (A) and CD86 (B) in BMDMs stimulated with or without DMXAA. WT and *Tbc1d23* KO BMDMs were stimulated with or without DMXAA for 24 h, then subjected to flow cytometry analysis (n = 3 biologically independent samples). Left and right panels show representative histograms and quantitative data, respectively. c, d Flow cytometric analysis of expression of H2-k^b^ and H2-k^b^-SIINFEKL on the surface of B16F10-OVA. WT and *Tbc1d23* KO BMDMs were treated with or without DMX

Model of TBC1D23-dependent regulation of STING-mediated antitumor immunity. TBC1D23 coordinates endosomal trafficking of TBK1 to the Golgi via interactions with the WASH complex, golgin-97/245, and WDR11. a TBC1D23 facilitates endosome-to-Golgi trafficking of TBK1, activating the IFN-I signaling axis. This enhances secretion of IFN-I and chemokines, promoting macrophage maturation, antigen presentation, and CD8⁺ T cell infiltration/activation to suppress tumor growth. b TBC1D23 deficiency impairs TBK1 translocation to the Golgi, inhibiting IFN-I pathway activation. Consequently, reduced IFN-I/

Full text

Introduction

The cyclic GMP–AMP synthase–stimulator of interferon genes (cGAS–STING) pathway represents a central cytosolic sensor system for DNA. Through direct binding to cytosolic double-stranded DNA (dsDNA), cGAS triggers a strong inflammatory signaling cascade [1, 2]. Detection of aberrant or exogenous DNA facilitates the generation of the second messenger 2′3′-cyclic GMP–AMP (2′3′-cGAMP), which in turn engages STING and activates downstream cascades, including the TANK-binding kinase 1 (TBK1)–IRF3 axis and the NF-κB signaling module. These signaling events lead to the induction and release of type I interferons (IFN-I) and/or multiple proinflammatory cytokines [3]. In addition its canonical role in antiviral defense, the cGAS–STING signaling influences a wide range of physiological and pathological processes, including tumorigenesis [4, 5]. Within the tumor microenvironment (TME), activation of the cGAS–STING signaling augments antigen presentation, supports the infiltration and cytolytic function of T lymphocytes and natural killer (NK) cells, and remodels immunosuppressive niches [6–9]. Such immunomodulatory capacity has established cGAS–STING as an attractive candidate for therapeutic intervention in oncology [10–14]. Thus, comprehensive characterization of its spatiotemporal control mechanisms is essential for unlocking its full clinical application value.

The activation of the cGAS–STING signaling cascade involves multiple subcellular compartments [15, 16]. STING, an endoplasmic reticulum (ER)-resident transmembrane protein [17], relocates to trans-Golgi networks (TGN). At TGN membranes, it recruits TBK1. TBK1 becomes activated via autophosphorylation. and, in turn, phosphorylates interferon regulatory factor 3 (IRF3) [18–20]. The phosphorylated IRF3 forms dimers that translocate to the nucleus, where they induce transcription of IFN-I and interferon-stimulated genes (ISGs) [5, 21]. To switch off the signal, phosphorylated STING is sorted to endolysosomes for degradation through an AP-1–dependent process [22]. The elucidation of the spatiotemporal trafficking network of the STING pathway provides a new dimension for therapeutic development. For instance, disrupting lysosomal targeting of STING may enhance antitumor immunity [23].

TBC1D23, a Tre2–Bub2–Cdc16 (TBC) family GTPase-activating protein, has emerged as an important regulator of membrane transport [24–26]. Acting in concert with the Golgi-associated proteins golgin-97 and golgin-245, the WASH complex, and the WDR11 complex, it mediates endosome-to-Golgi trafficking [27–30]. Our recent work identified TBC1D23 as indispensable for the retrograde transport of TBK1 from endosomes to the trans-Golgi network (TGN), a trafficking step required for optimal STING signaling [31]. Mechanistically, the WASH complex subunit FAM21 acts as a molecular bridge linking TBK1 to TBC1D23, enabling their coordinated functional activity. Loss of Tbc1d23 in mice profoundly suppresses STING-dependent type I interferon induction, yet the full extent of its physiological repercussions has not been definitively determined.

In this work, we identified a strong positive correlation between high TBC1D23 expression in melanoma and increased immune cell infiltration, as well as improved clinical prognosis. Loss of Tbc1d23 significantly compromises the antitumor potency of STING agonists, accompanied by reduced macrophage and CD8⁺ T cell infiltration. Mechanistically, TBC1D23 augments type I interferon–driven chemokine production, facilitates macrophage maturation, and elevates tumor cell MHC-I expression, thereby activating CD8⁺ T cell–mediated immune elimination and intensifying STING-dependent antitumor immunity. These results indicate TBC1D23 is a pivotal regulator of STING-dependent immune surveillance in the tumor microenvironment.

Result

High TBC1D23 expression correlates with increased immune cell infiltration in melanoma

To uncover the role of TBC1D23 in cancer immunology, we chose melanoma, a highly immunogenic tumor, and analyzed the expression level of TBC1D23 in melanoma and adjacent normal tissues. Analysis of the TCGA database revealed a trend toward lower TBC1D23 expression in melanoma tissues compared with adjacent normal tissues, although the difference did not reach statistical significance (Fig. 1a). Furthermore, melanoma patients with high expression of TBC1D23 (TBC1D23^high^) displayed significantly better overall survival, event-free survival and post-progression survival than those with low expression of TBC1D23 (TBC1D23^low^) (Fig. 1b).Fig. 1. High TBC1D23 expression is correlated with greater immune cell infiltration in melanoma. a Expression level of TBC1D23 in melanoma (n = 558) and adjacent normal (n = 461) tissues from TCGA database was analyzed using the GEPIA 2 on 12 October, 2024 (http://gepia2.cancer-pku.cn/#analysis). Bars indicate the median expression level in each group. Statistical analysis was performed using t test. b Prognostic value of TBC1D23 in melanoma was analyzed using the Kaplan–Meier Plotter database (http://kmplot.com/analysis/), which integrates multiple GEO datasets. Overall survival, event free survival and post progression survival were analyzed in all patients with melanoma. Log-rank tests were performed. HR below 1 implies greater survival probabilities for TBC1D23^high^ group compared with TBC1D23^low^ group. Detailed information on tumor samples can be found in the individual GEO dataset in the NCBI GEO website. c The estimated enrichment of 7 immune cells in the TBC1D23^high^ and TBC1D23^low^ groups was calculated by CIBERSORT. *P < 0.05; **P < 0.01; ***P < 0.001; ****P < 0.0001. d Kaplan–Meier plots for CD8^+^ T cells infiltrates and TBC1D23 mRNA expression to visualize survival differences in melanoma. e Representative images of mIHC staining of the expression pattern of TBC1D23 and CD8a in human melanoma tissue microarrays. Scale bar 50 μm. f Correlation analysis between the number of TBC1D23 positive cells and the number of CD8^+^ T cells in human melanoma tissue microarrays. Human melanoma tissue microarrays were generated by SHANGHAI OUTDO BIOTECH CO.LTD. Data were analyzed by Pearson’s correlation

Immunotherapy has emerged as a promising therapeutic strategy for melanoma [32]. A critical determinant of therapeutic efficacy, the tumor immune microenvironment orchestrates antitumor immune responses, with cytotoxic CD8⁺ T lymphocytes and natural killer (NK) cells serving as the principal effectors in this intricate niche [33, 34]. To investigate the role of TBC1D23, we quantified immune infiltration within melanoma tumors from TBC1D23^high^ and TBC1D23^low^ individuals, and employed CIBERSORT for analysis. CIBERSORT analysis showed that the TBC1D23^high^ group displayed a higher infiltration of CD4^+^ T, CD8^+^ T and NK cells, than the TBC1D23^low^ group (Fig. 1c). High intratumoral infiltration of CD8⁺ T cells is known to be closely associated with favorable prognosis and improved clinical outcomes in cancers. We therefore analyzed the correlation of TBC1D23 expression, infiltration of CD8⁺ T cells, and survival time of melanoma patients. Patients with both high TBC1D23 expression and high CD8⁺ T cells infiltration exhibited prolonged survival compared to other groups (Fig. 1d). Multiplex immunofluorescence staining on human melanoma tissue microarrays revealed a positive correlation between the number of TBC1D23 positive cells and the number of CD8^+^ T cells (Fig. 1e, f). Together, these data indicated that high TBC1D23 level is associated with increased immune cell infiltration in melanoma, and predicts favorable clinical outcomes.

TBC1D23 is critical for STING-mediated antitumor effect

As patient data indicates that TBC1D23 expression positively correlates with the extent of immune cell infiltration in melanoma, we next tested this idea using a syngenic model of melanoma. Mouse melanoma B16F10 cells were subcutaneously inoculated into wild-type (WT) or Tbc1d23 KO C57BL/6 mice. Around 7 days, the tumors reached a volume of 100–150 mm^3^. Tumor-bearing WT or Tbc1d23 KO mice were further divided into two groups, with each group bearing a similar size of tumors. These mice were then intraperitoneally injected once with either vehicle or the STING agonist DMXAA. Seven days after DMXAA treatment, tumors were harvested (Fig. 2a). In the absence of DMXAA treatment, tumor progression did not differ significantly between WT and Tbc1d23 KO mice (Fig. 2b-d). Following DMXAA treatment, Tbc1d23 KO mice exhibited significantly increased tumor growth relative to WT mice (Fig. 2b-d).Fig. 2. Loss of TBC1D23 impairs STING-mediated antitumor effects in melanoma models. a Experimental workflow to determine how TBC1D23 cooperates with the STING pathway to regulate antitumor immunity. WT and Tbc1d23 KO mice were inoculated with 2 × 10^5^ B16F10 cells subcutaneously. Six days after inoculation, the mice were treated with vehicle or DMXAA. Seven days later, tumor tissues were collected and analyzed. DMXAA were generated by Targetmol Biotechnology Co., Ltd(T6273). b-d Representative images of tumors (b), tumor volume (c) and tumor weights (d) in WT and Tbc1d23 KO mice treated with vehicle or DMXAA. Groups: WT- Vehicle, Tbc1d23 KO- Vehicle, WT-DMXAA, Tbc1d23 KO-DMXAA (n = 5 for each group). e Representative images of IHC staining of the expression pattern of pSTING and pSTAT1 in B16F10 melanoma specimens. Scale bar 100 μm. f Quantitative analysis of the numbers of pSTING^+^ cells and pSTAT1.^+^ cells within same area of tumors as in (e). means ± SEM. Data were analyzed by two-way ANOVA (b) or one-way ANOVA (d and f)

To evaluate the role of TBC1D23 in STING signaling, we performed immunohistochemistry (IHC) on B16F10 tumors. In WT mice, DMXAA treatment significantly promoted the proportions of pSTING^+^ (Ser366) and pSTAT1^+^ (Tyr701) cells, indicating activation of the STING pathway (Fig. 2e, f). In contrast, Tbc1d23 deficiency almost completely reversed the proportions of pSTING^+^ and pSTAT1^+^ cells to levels comparable to vehicle-treated WT mice. Collectively, these results demonstrated that TBC1D23 is essential for STING-mediated antitumor effect.

TBC1D23 is critical for STING-mediated activation and recruitment of CD8+ T cells

We next assessed how TBC1D23 contributed to STING-mediated antitumor effect. We sorted tumor-infiltrating CD45^+^ immune cells from WT or Tbc1d23 KO mice and performed transcriptomic sequencing. In WT mice, 1065 differentially expressed genes (DEGs) were identified in the DMXAA-treated group compared to the control group. Remarkably, over 85% of them were overlapping the DEGs between WT and Tbc1d23 KO mice upon DMXAA treatment (Fig. S1a).

In WT mice, DMXAA treatment enhanced antigen presentation responses and upregulated interferon and chemokine-related genes, including H2-Eb1, H2-k1, H2-d1, B2m, Irf7, Tap1, Ifi47, and Cxcl9. These genes correlated with the significantly reduced tumor burden compared to vehicle-treated controls (Fig. 3a). However, the upregulation of these genes was largely reversed upon TBC1D23 depletion (Fig. 3a). DMXAA treatment also increased the expression of multiple genes involved in maturation, activation, and cytotoxicity of CD8 + T cells and NK cells, such as Cd69, Cd28, Gzmb, Prf1, and Nkg7. Analogously, deletion of TBC1D23 significantly reduced their expression (Fig. 3a). We classified the overlapping DEGs into three distinct expression patterns which were associated with chemotaxis, T cell activation, and antigen presentation, respectively, followed by GO enrichment analysis. DMXAA treatment in WT mice promoted T cell-mediated antitumor immune responses and antigen presentation signaling pathways, and reduced neutrophil infiltration (Fig. 3b). However, these effects were largely abolished upon TBC1D23 deletion (Fig. 3b). A similar conclusion was obtained by analyzing the ssGSEA (single-sample Gene Set Enrichment Analysis) scores of multiple tumor immune-related pathways in different subgroups (Fig. S1b) [35].Fig. 3. Loss of TBC1D23 comprises STING-mediated anti-tumor immunity. a,b Heatmap of differentially expressed genes (DEGs) of CD45^+^ tumor-infiltrating immune cells between tumor bearing WT and Tbc1d23 KO mice treated with vehicle or DMXAA. B16F10 tumors were isolated as in Fig. 2a, and CD45^+^ tumor-infiltrating immune cells were isolated for bulk RNA-seq analysis. b The indicated dynamic DEGs with six different expression patterns and showed as heatmap via ClusterGVis package. The corresponding gene expression trajectories and representative GO terms were shown in the right panels

CD8^+^ T cell-mediated immune responses are crucial for limiting tumor progression. Complementary to RNA-seq data, we performed flow cytometry analysis of tumors from WT or Tbc1d23 KO mice with or without DMXAA treatment. In the absence of DMXAA treatment, deletion of Tbc1d23 did not affect the infiltration of CD45^+^ or CD8^+^ T cells (Fig. 4a, b). DMXAA treatment of WT mice increased the proportion of CD45^+^ cells within live cells and the proportion of CD8^+^ T cells within CD45^+^ cells (Fig. 4a, b). Deletion of Tbc1d23 almost completely abrogated the DMXAA-induced increase in the infiltration of CD45^+^ and CD8^+^ T cells (Fig. 4a, b). These results were further confirmed by multiplex immunohistochemistry (mIHC) staining, which demonstrated that Tbc1d23 deficiency markedly attenuated the DMXAA-triggered influx of immune cells, including CD45⁺ cells (70% reduction), macrophages (60%), and CD8⁺ T cells (90%) (Fig. 4c, d). Moreover, deletion of Tbc1d23 also reduced the number of GZMB^+^CD8^+^ T cells, which indicate the activation of CD8^+^ T cells (Fig. 4c, d). Consistently, Immunohistochemistry (IHC) analysis demonstrated that STING activation by DMXAA promoted the redistribution of CD8^+^ T cells from the tumor periphery to the tumor interior in WT sample, while TBC1D23 deficiency nearly abolished this effect (Fig. S2a,b).Fig. 4TBC1D23 is critical for STING-dependent infiltration of immune cells. a, b Representative FACS data of frequency (a) and quantification (b) of tumor infiltrating CD45^+^ T cells or CD8^+^ T cells (n = 5 per group) of the mice as in Fig. 3a. c B16F10 tumors were isolated as in Fig. 2a, then stained with anti-CD45-cy3, anti-F4/80-cy5, anti-CD8-spAqua or anti-GZMB-FITC, and counterstained with DAPI. F4/80 is marker of macrophages. Scale bar, 50 μm. d Quantitative analysis of the numbers of CD45^+^ cells, macrophages, CD8^+^ T cells and GZMB^+^ CD8.^+^ T cells within same area of tumors as in c. Data are representative of three independent experiments. means ± SEM. Data were analyzed by one-way ANOVA (c and d)

Finally, to further evaluate the role of TBC1D23 in the STING activation-induced CD8^+^ T cell infiltration, we labeled CD8^+^ T cells with CFSE [36], a cell-permeable fluorescent dye, and intravenously injected these cells into tumor-bearing mice. Flow cytometric analysis of tumor digests revealed that DMXAA treatment increased the infiltration of CFSE^+^CD8^+^ T cells in tumors in WT mice, while TBC1D23 deletion significantly reduced the infiltration (Fig. S2c,d). Taken together, our results demonstrate that TBC1D23 is critical for STING-mediated activation and recruitment of CD8^+^ T cells.

TBC1D23 promotes STING-dependent chemokine production

Activation of the STING signaling triggers secretion of key chemokines, including CCL5, CXCL9, and CXCL10, to promote CD8⁺ T cell recruitment [37, 38]. As our prior experiments demonstrated that TBC1D23 is critical for STING-induced IFN-I production, we next assessed whether TBC1D23 regulates STING-mediated chemokine secretion. We isolated bone marrow-derived macrophages (BMDMs) from WT and Tbc1d23 KO mice, treated with DMXAA, and determined the global gene expression pattern by RNA-sequencing. Volcano plot analysis revealed 110 upregulated and 39 downregulated differentially expressed genes (DEGs) in Tbc1d23 KO versus WT BMDMs under DMXAA treatment, with Cxcl9 expression significantly reduced (Fig. 5a). Subsequent Gene Ontology (GO) enrichment analysis of these DEGs showed that the top 20 enriched biological pathways were primarily associated with type I IFN signaling pathways (Fig. 5b).Fig. 5. Deletion of TBC1D23 in macrophages impairs STING signaling -dependent chemokine production. a Volcano plot of differentially expressed genes. WT and Tbc1d23 KO BMDMs were treated with DMXAA for 4 h, RNA was extracted and subjected to microarray analysis. Red and green dots show upregulated and downregulated genes in Tbc1d23 KO BMDM, respectively. b KEGG pathway enrichment analysis of differentially expressed genes as in (a), and the top 20 pathways were shown. Interferon related pathways were shown in red color. (c-f) WT and Tbc1d23 KO BMDMs were treated with DMXAA (25 μg/mL) for 4 h, and the expression levels of Ccl5 (c), Cxcl9 (d), *Cxcl10 *(e), *Cxcl11 *(f) were analyzed by qPCR. Data are representative of three independent experiments. means ± SEM. Data were analyzed by one-way ANOVA (c-f)

To validate the transcriptomics data, we next assessed transcript levels of interferon-induced chemokines that are critical for CD8⁺ T cell tumor infiltration, including Ccl5, Cxcl9, Cxcl10, and Cxcl11. Quantitative PCR (qPCR) analysis showed that DMXAA treatment in WT BMDMs significantly induced the expression of these chemokines compared to untreated control. Critically, TBC1D23 deficiency substantially attenuated the induction of all chemokines analyzed, with transcript levels reduced by 47% (Ccl5), 63% (Cxcl9), 66% (Cxcl10), and 25% (Cxcl11) relative to WT controls (Fig. 5c-f). Collectively, these findings demonstrate that TBC1D23 exerts a positive regulatory role in the STING-dependent chemokine network. Thus, deletion of TBC1D23 disrupts STING-mediated chemokine signaling, leading to impaired CD8⁺ T cell tumor infiltration.

TBC1D23 enhances STING-mediated antigen presentation

Activation of the STING pathway promotes the differentiation and maturation of antigen-presenting cells (APCs), such as dendritic cells or macrophages [39–43]. To assess the impact of TBC1D23 on APC maturation and activation, WT and Tbc1d23 KO BMDMs were treated with or without the STING agonist DMXAA. Flow cytometric revealed that DMXAA treatment significantly upregulated surface expression of MHC-II and CD86 which are macrophage activation marker on WT BMDMs (Fig. 6a, b). Deletion of TBC1D23 markedly inhibited the upregulation, indicating that TBC1D23 is critical for macrophage maturation and activation (Fig. 6a, b).Fig. 6. Deletion of TBC1D23 in macrophages impairs STING-mediated antigen presentation. a, b Mean fluorescent intensity (MFI) of MHC-II (A) and CD86 (B) in BMDMs stimulated with or without DMXAA. WT and Tbc1d23 KO BMDMs were stimulated with or without DMXAA for 24 h, then subjected to flow cytometry analysis (n = 3 biologically independent samples). Left and right panels show representative histograms and quantitative data, respectively. c, d Flow cytometric analysis of expression of H2-k^b^ and H2-k^b^-SIINFEKL on the surface of B16F10-OVA. WT and Tbc1d23 KO BMDMs were treated with or without DMXAA, supernatants were collected and applied to B16F10-OVA cells pre-treated with either isotype control or anti-Ifnar1. Expression of H2-k^b^ and H2-k^b^-SIINFEKL on the surface of B16F10-OVA was analyzed by flow cytometry. Left and right panels show representative histograms and quantitative data, respectively. e–g Flow cytometric analysis of antigen-specific cytotoxicity in CD8⁺ T cells. WT and Tbc1d23 KO BMDMs were treated with or without DMXAA, supernatants were collected and then applied to B16F10-OVA cells. CD8^+^ T cells were isolated from the spleen and lymph node of OT-I mice and co-cultured with B16F10-OVA cells pretreated with supernatants for 48 h. Percentages of GZMB^+^CD8^+^ T cells in CD8^+^ T cells (e), and CD107a^+^CD8^+^ T cells in CD8^+^ T cells (f), Zombie NIR.^+^ cells in tumor cells (g) were analyzed by flow cytometry. Data are representative of three independent experiments. Data were analyzed by one-way ANOVA (a–d, g) or by unpaired t test (e–f)

To further evaluate TBC1D23's role in MHC-I antigen presentation, we assessed its impact on surface expression of the OVA peptide-MHC-I complex (H2-Kᵇ-SIINFEKL) in B16F10-OVA cells. These cells ectopically express chicken ovalbumin (OVA), enabling intracellular processing to generate the SIINFEKL peptide for MHC-I loading and subsequent recognition by OVA-specific CD8⁺ T (OT-I) cells. We isolated BMDMs from WT or Tbc1d23 KO mice, and treated with DMXAA. The supernatants were then collected and applied to B16F10-OVA cells that were pretreated with isotype control antibody or anti-Ifnar1 blocking antibody. Flow cytometry revealed that DMXAA treatment significantly upregulated H2-Kᵇ-SIINFEKL surface expression compared to controls (Fig. 6c, d). Notably, Tbc1d23 deficiency substantially impaired this response (Fig. 6c, d). Critically, anti-Ifnar1 pretreatment abolished the differential expression of H2-Kᵇ-SIINFEKL between WT and Tbc1d23 KO groups, suggesting that Tbc1d23 regulates H2-Kᵇ-SIINFEKL surface expression via* mediating* IFN-I production (Fig. 6c, d).

We next determined whether attenuated antigen presentation in Tbc1d23 KO BMDMs impacts CD8⁺ T cell cytotoxicity. B16F10-OVA cells were first incubated with supernatants from DMXAA-treated WT or Tbc1d23 KO BMDMs. We then co-cultured these cells with OT-I CD8⁺ T cells that were isolated from spleen and lymph nodes. Tbc1d23 deficiency significantly reduced the frequency of activated cytotoxic CD8⁺ T cells, as evidenced by the decreased populations of GranzymeB⁺ (GZMB^+^) and CD107a⁺CD8⁺ T cells. Consequently, Tbc1d23 deficiency also markedly impaired T cell-mediated tumor cell killing relative to WT controls (Fig. 6e-g). Collectively, these data demonstrate that TBC1D23 potentiates DMXAA-triggered IFN-I signaling, thereby amplifying antigen presentation and enhancing CD8⁺ T cell-dependent antitumor immunity**.**

Discussion

Understanding STING–IFN-I regulatory mechanisms is critical for developing effective cancer immunotherapies [44, 45]. In this study, we show that high TBC1D23 levels associate with increased immune infiltration and improved prognosis in melanoma. We further demonstrate that TBC1D23 deletion suppresses STING–IFN-I signaling, reduces type I IFN and chemokine secretion, and diminishes CD8⁺ T cell infiltration and activation. Ultimately, TBC1D23 deficiency leads to accelerated melanoma progression(Fig. 7). These findings establish TBC1D23 as an essential mediator of the STING signaling and a critical regulator of anti-tumor immunity.Fig. 7. Model of TBC1D23-dependent regulation of STING-mediated antitumor immunity. TBC1D23 coordinates endosomal trafficking of TBK1 to the Golgi via interactions with the WASH complex, golgin-97/245, and WDR11. a TBC1D23 facilitates endosome-to-Golgi trafficking of TBK1, activating the IFN-I signaling axis. This enhances secretion of IFN-I and chemokines, promoting macrophage maturation, antigen presentation, and CD8⁺ T cell infiltration/activation to suppress tumor growth. b TBC1D23 deficiency impairs TBK1 translocation to the Golgi, inhibiting IFN-I pathway activation. Consequently, reduced IFN-I/chemokine secretion compromises macrophage maturation, antigen presentation, and CD8⁺ T cell recruitment/activation, accelerating melanoma progression

STING agonists have been actively developed for cancer immunotherapy [46, 47]. Current STING agonists all target the ligand-binding pocket of the STING protein. These agonists stimulate an acute and potent response, but are often associated with tissue toxicities that limit their efficacy. Our results herein suggest an alternative approach to specifically enhance the STING-IFN-I pathway while minimally impacting the STING-NF-κB signaling: enhancing the TBC1D23-FAM21 or FAM21-TBK1 interaction. Enhancing these interactions could increase TBK1 flux to the TGN, thus amplify STING-IFN-I-mediated antitumor regulation. Supporting this concept, a recent study demonstrates the SNACIP (Specific molecule-nanobody conjugate induced proximity) platform, which utilizes a nanobody targeting a protein of interest (POI) conjugated to a small molecule binding motif for a second target to induce intracellular proximity and modulate signaling pathways [48].

In this study, we broadened the biological significance of the TBC1D23–STING signaling axis by highlighting its role in anti-tumor immunity. In contrast to prior work that has largely emphasized pharmacological activation of STING or characterization of downstream signaling components, our findings identify TBC1D23 as a critical determinant of the endogenous regulation of STING–IFN-I–mediated anti-tumor responses. Notably, TBC1D23 expression levels markedly influence the composition and functional state of the tumor immune microenvironment, suggesting that loss or impairment of TBC1D23 activity may represent a previously unrecognized mechanism of immune evasion. Moreover, compared with therapeutic strategies that directly target STING using agonists, our results support an alternative concept: potentiating endogenous STING–IFN-I signaling through modulation of TBC1D23. This approach may provide a theoretical basis for developing more selective and potentially less toxic STING-centered immunotherapeutic interventions.

Although we observed that TBC1D23 deletion suppresses the maturation of immunostimulatory macrophages and is accompanied by reduced infiltration, activation, and cytotoxic function of tumor antigen–specific CD8⁺ T cells, we cannot exclude the possibility that TBC1D23 functions in additional immune subsets or in tumor cells themselves, thereby contributing to these phenotypes. Consequently, we did not establish direct causal evidence demonstrating that the CD8⁺ T-cell dysfunction is strictly dependent on macrophage alterations. Future studies employing cell type–specific genetic approaches or immune cell reconstitution experiments will be essential to delineate the cellular hierarchy and intercellular interactions governed by the TBC1D23–STING axis within the tumor immune network.

Previous studies have linked homozygous mutations of TBC1D23 with pontocerebellar hypoplasia (PCH), a group of rare neurological disorders characterized by pons and cerebellum maldevelopment [49, 50]. In addition to abnormal brain development, PCH patients often suffer recurrent respiratory infections as the predominant extra-neurological symptom. Our identification of TBC1D23 as a critical regulator of STING–IFN-I signaling implicates attenuated antiviral immunity due to TBC1D23 deficiency in this clinical phenotype, providing a mechanistic basis for improved diagnosis and targeted therapeutic strategies.

Materials and methods

Mice

All mice were housed under specific pathogen-free (SPF) conditions at the Laboratory Animal Center of West China Second University Hospital, Sichuan University. Tbc1d23 (flox/flox) mice on a C57BL/6 background were kindly provided by Professor Lu Chen (Sichuan University). To generate a systemic conditional knockout of Tbc1d23, Tbc1d23 (flox/flox) mice were crossed with ROSA26-CreERT2 (Cre-ERT) mice. The ROSA26-CreERT2 line carries a tamoxifen-inducible CreERT2 recombinase inserted into the ubiquitously expressed Rosa26 safe-harbor locus, enabling temporal control of gene deletion upon tamoxifen administration. Tbc1d23 (flox/flox) mice were first crossed with ROSA26-CreERT2 mice to obtain Tbc1d23 (flox/+; Cre-ERT) offspring, which were subsequently crossed with Tbc1d23 (flox/flox) mice to generate Tbc1d23 flox/flox; Cre-ERT) mice (hereafter referred to as Tbc1d23 KO). Tamoxifen was administered by intraperitoneal injection to 4-week-old mice to induce Tbc1d23 deletion.

Tumor inoculation and surface staining

For the B16F10 tumor model, 2 × 10^5^ B16F10 cells were subcutaneously inoculated into the right flank of WT or Tbc1d23 KO mice. After approximately 6 days, the tumors reached a volume of 100–150 mm^3^. Tumor-bearing WT or Tbc1d23 KO mice were divided into two groups, with each group bearing tumors of similar size. These mice were then intraperitoneally injected once with either vehicle or 25 mg/kg of the STING agonist DMXAA(Targetmol #T6273). Tumor sizes were measured every two days after treatment. Seven days after DMXAA treatment, tumors were harvested. For tumor dissection, tumors were weighed, and the remaining tumor chunks were digested using collagenase type IV (ThermoFisher #17,104,019, 100 μg/mL) and DNase I (Invitrogen #EN0521, 100 μg/mL). Digested tumor tissues were treated with red blood cell lysis buffer (BioLegend #420,301) and filtered through 70 μm strainers. To assess tumor-infiltrating immune cells, mouse TruStain FcX™ (BioLegend) was diluted 1:100 at room temperature to block Fc receptors for 15 min. Flow cytometric analysis was performed to assess the infiltration of CD8⁺ T cells using BV421 anti-mouse CD45, APC anti-mouse CD3, and PE anti-mouse CD8. Zombie NIR was used to exclude dead cells. Data were acquired using a BD FACSAria III and analyzed with FlowJo v10.10.0.

Immunohistochemical and multiplex immunofluorescence staining

B16F10 tumor tissues from WT or Tbc1d23 KO mice treated with vehicle or DMXAA were fixed and paraffin-embedded. Endogenous peroxidase activity was blocked. For immunohistochemical (IHC) staining, the following antibodies were used: anti-pSTING (Ser366) (CST) and anti-pSTAT1 (Tyr701) (CST). Multiplex immunofluorescence staining was performed by Hubei BIOSSCI Biotechnology Co., Ltd (Hubei, China) using anti-CD45-Cy3, anti-F4/80-Cy5, anti-CD8-SpAqua, anti-GZMB-FITC, and DAPI for nuclear visualization. Slides were scanned using a VS200 scanner under brightfield and fluorescence conditions.

Flow cytometry

WT and Tbc1d23 KO BMDMs were treated with DMSO or 25 μg/mL DMXAA for 24 h, followed by washing with PBS containing 2 mM EDTA and 2% FBS. Mouse TruStain FcX™ was used for Fc receptor blocking at a 1:100 dilution at room temperature for 15 min. Then, BMDMs were stained with the following fluorescently labeled antibodies: FITC anti-mouse I-A/I-E (MHC-II) and PE-Cy7 anti-mouse CD86 for 40 min at 4 °C.

B16F10-OVA cells were generated by infection with lentivirus obtained by Jingruibio Biotechnology Co.,Ltd (Guangdong, China) Lentivirus Packaging Kit (L002C) for OVA transduction and cultured with 3 μg/mL puromycin for 1 week. WT and Tbc1d23 KO BMDMs were treated with DMSO or 25 μg/mL DMXAA for 4 h. Then, supernatants were collected and applied to B16F10-OVA cells pre-treated with either isotype antibody or 10 μg/mL anti-Ifnar1 antibody to block Ifnar1 signaling. Expression of H2-Kb and SIINFEKL-H2-Kb on the surface of B16F10-OVA cells was analyzed by flow cytometry. Data were recorded on a BD FACSAria III and analyzed using FlowJo v10.10.0.

In vivo chemotaxis assay of CD8+ T cells

For the in vivo chemotaxis assay, 2 × 10^5^ B16-F10 cells were subcutaneously inoculated into the right flank of WT or Tbc1d23 KO mice. Tumor-bearing mice were divided into 2 groups with comparable tumor sizes and intraperitoneally injected with either vehicle or 25 mg/kg DMXAA (STING agonist). CD8⁺ T cells were isolated from spleens and lymph nodes of wild-type C57BL/6 mice using CD8⁺ T cell isolation beads (MCE #HY-K0310), then activated by stimulation for 72 h on plates coated with 1 μg/mL anti-CD3 and 1 μg/mL anti-CD28, supplemented with 20 ng/mL mIL-2. Activated CD8⁺ T cells were labeled with 5 μM CFSE (BioLegend #423,801) for 20 min at room temperature and washed three times. Labeled cells were intravenously injected into tumor-bearing mice. After 48 h, tumors were dissected and digested. CFSE-labeled tumor-infiltrating cells were detected in the FITC channel, with 7-AAD used to exclude dead cells. Data were acquired on a BD FACSAria III and analyzed using FlowJo v10.10.0.

Isolation and culture of bone marrow-derived macrophages (BMDMs)

Primary BMDMs were isolated from femurs and tibiae of 8–10-week-old WT or Tbc1d23 KO mice. Bone marrow cells were flushed with ice-cold PBS, filtered through a 70-μm cell strainer, and centrifuged at 300 × g for 5 min. Erythrocytes were lysed with red blood cell lysis buffer (BioLegend #420,301) for 5 min at room temperature. After washing, cells were resuspended in DMEM-based medium containing 50% DMEM, 20% heat-inactivated FBS, 1% penicillin–streptomycin (Gibco), and 30% L929 cell-conditioned supernatant. Cells were seeded at 1 × 10^6^ cells/mL in Petri dishes and incubated at 37 °C with 5% CO₂. Fresh medium was replenished on day 3. Differentiated BMDMs were harvested on day 7 by scraping and re-seeded for experimental assays.

qPCR and RNA sequencing

WT and Tbc1d23 KO BMDMs were treated with or without 25 μg/mL DMXAA for 4 h, and then total RNA was extracted using FastPure Complex Cell/Tissue Total RNA Isolation Kit (Vazyme). RNA integrity was verified by Nanodrop analyzer. First-strand cDNA synthesis was executed using HiScript IV RT SuperMix for q-PCR (+ gDNA wiper) (Vazyme) in accordance with the manufacturer's protocol. qPCR were analyzed by ChamQ Universal SYBR qPCR Master Mix (Vazyme). Bulk RNA Sequencing was performed on Illumina NovaSeq 6000 platform (Illumina). To investigate the transcriptional changes in CD45^+^ tumor-infiltrating immune cells between tumor bearing WT and Tbc1d23 KO mice treated with vehicle or DMXAA., CD45^+^ immune cells from B16F10 tumors isolated from WT or Tbc1d23 KO mice were sorted by BD Aria lIII for RNA extract and RNA sequencing. Zombie NIR was used to exclude cells.

B16F10-OVA and OT-1 T cell co-culture assay

WT and Tbc1d23 KO BMDMs were treated with DMSO or 25 μg/mL DMXAA for 4 h, after which the supernatants were collected. B16F10-OVA cells were then cultured with the supernatants for 48 h, followed by removal of the supernatants and two washes with PBS. Murine CD8⁺ T cells were isolated from the spleens and lymph nodes of OT-1 mice using CD8⁺ T cell sorting beads (MCE,HY-K0310), and then co-cultured with B16F10-OVA cells that had been pretreated with supernatants for 48 h. After co-culture, CD8⁺ T cells and tumor cells were harvested, and flow cytometry was performed to assess CD8⁺ T cell activation and tumor cell viability. CD8⁺ T cells and tumor cells were stained with FITC anti-mouse CD8a and PE anti-mouse CD107a for 40 min at 4 °C, followed by fixation and permeabilization (BD Biosciences,554,722). After washing with Perm/Wash buffer (BD Biosciences, 554,723), cells were stained with BV421 anti-mouse GZMB. Zombie NIR was used to identify dead tumor cells and exclude dead CD8⁺ T cells. Data were acquired using a BD FACSAria III and analyzed with FlowJo v10.10.0.

Volcano map, gene ontology and GSEA analyses

Differential gene expression heatmaps were generated using pheatmap (version 1.0.12) and ClusterGVis (version 0.1.2). Differential genes were clustered using the Mfuzz algorithm, and functional enrichment analysis for distinct gene clusters was performed using clusterProfiler (version 4.12.6).

The expression of TBC1D23 and its association with prognosis in melanoma were evaluated using GEPIA2 and the Kaplan–Meier Plotter. The estimated enrichment of ten immune cell types in the TBC1D23^high^ and TBC1D23^low^ groups was calculated using CIBERSORT [51]. Kaplan–Meier survival curves stratified by M1 macrophage infiltration and TBC1D23 mRNA expression were generated using TIMER to visualize survival differences in melanoma.

Volcano plots were generated using the ggplot2 package in R, and the indicated genes were annotated using ggrepel. Dynamic differentially expressed genes (DEGs) with six distinct expression patterns were visualized as heatmaps using the ClusterGVis package. Genes with |log₂FoldChange|> 1 and FDR < 0.05 were defined as differentially expressed. Specific gene expression patterns were plotted and normalized using the pheatmap function from the pheatmap R package.

Gene Ontology (GO) enrichment was visualized with a bubble plot generated using ggplot2, where dot size indicates the number of enriched genes, color represents the enrichment ratio, and the x-axis shows -log₁₀(p-value). GSEA results were visualized using the gseaNb function from the GseaVis package [52]. DEG heatmaps were generated with pheatmap (v1.0.12) and ClusterGVis (v0.1.2). Differential genes were clustered using the Mfuzz algorithm, and functional enrichment of each gene cluster was performed using clusterProfiler (v4.12.6) [53].

Statistical analysis

GraphPad Prism 10.4.2 software was used for data analysis. Depending on the number of samples, and as specified in the figure legends, we used either a Student’s t-test or a one-way with appropriate multiple comparison tests. Differences were considered significant when P < 0.05. All data were represented as mean ± SEM.

Supplementary Information

Supplementary Material 1.