TRIM28 drives immune evasion via PARP1 SUMOylation and NAD+ depletion in clear cell renal cell carcinoma

Abstract

Background: Immune checkpoint blockade (ICB) therapy has demonstrated significant clinical potential in a variety of cancers; however, its efficacy in clear cell renal cell carcinoma (ccRCC) remains suboptimal. In ccRCC, an increased infiltration of CD8⁺ T cells does not necessarily correlate with improved prognosis, indicating the presence of unique immune evasion mechanisms within the tumor microenvironment (TME).

Methods: Tripartite motif-containing 28 (TRIM28) was identified as a potential therapeutic target through single-cell transcriptomics (GSE159115) and Geneformer-based perturbation screening. Functional validation was performed by constructing shTRIM28 and overexpression cell models to assess tumor proliferation, CD8⁺ T cell co-cultures, flow cytometry, and patient-derived xenograft models. Co-immunoprecipitation and GST pull-down assays were used to analyze the TRIM28-poly (ADP-ribose) polymerase 1 (PARP1) interaction. SUMOylation/ubiquitination studies elucidated the mechanism regulating PARP1 stability, and chromatin immunoprecipitation-quantitative PCR identified the transcriptional regulation of programmed death-ligand 1 (PD-L1). High-throughput screening was conducted with RNA-seq, liquid chromatography-tandem mass spectrometry, and metabolomics. Virtual screening identified the TRIM28 inhibitor Eltrombopag, which was tested in combination with anti-programmed cell death protein-1 (PD-1) therapy for in vivo efficacy and metabolic reprogramming.

Results: We identified TRIM28 as a central regulator of immune evasion in ccRCC. Using high-throughput gene knockout screening, we demonstrated that TRIM28 depletion reprograms malignant epithelial cells toward a less aggressive phenotype and significantly enhances tumor cell susceptibility to cytotoxic T lymphocyte killing. Mechanistically, TRIM28 promotes immune resistance through dual immunometabolic mechanisms: first, by stabilizing PARP1 and promoting its SUMOylation, which in turn amplifies PD-L1 expression via NAD⁺-SIRT1-p65 signaling; second, by depleting NAD⁺ in the TME, limiting NAD⁺ availability for CD8⁺ T cells and impairing their respiration and effector function. These findings provide a novel mechanistic framework for TRIM28-driven immune suppression, integrating tumor-intrinsic metabolic reprogramming with CD8⁺ T cell dysfunction. Notably, we identified Eltrombopag as a candidate TRIM28 inhibitor, which synergized with anti-PD-1 therapy to enhance antitumor immunity and overcome ICB resistance in murine models.

Conclusions: This study reveals that TRIM28 is a key regulator of PD-L1 expression and T cell dysfunction in ccRCC through PARP1 stabilization and NAD⁺ metabolic reprogramming. Targeting TRIM28/PARP1/PDL1 with Eltrombopag reshapes the immunosuppressive TME and enhances checkpoint blockade efficacy, providing a novel combinatorial strategy for ccRCC immunotherapy.

Keywords: Immunosuppression; Immunotherapy; Kidney Cancer; T cell.

Figure 1. TRIM28 is a potential therapeutic vulnerability in ccRCC identified through in silico gene perturbation analysis. (A) Schematic overview of the Geneformer-based in silico knockout workflow using scRNA-seq data from ccRCC samples (GSE159115). (B–C) Transcriptional shift of malignant epithelial cells toward a normal phenotype following TRIM28 knockout. (D–E) TRIM28 mRNA expression is significantly upregulated in tumor tissues compared with adjacent normal tissues in the TCGA and GSE36895 cohorts. (F) Western blot analysis of TRIM28 protein levels in 12 paired ccRCC and adjacent normal tissues. (G) Representative IHC staining images showing elevated TRIM28 expression in tumor tissues. (H) TRIM28 expression stratified by tumor stage. (I) IHC H-score quantification reveals that TRIM28 expression is significantly associated with advanced clinical features including higher T stage, N1 status, and distant metastasis (M1). (J–K) Kaplan-Meier survival analysis indicates that high TRIM28 expression correlates with poorer overall survival in both ICGC and TCGA cohorts. ccRCC, clear cell renal cell carcinoma; ICGC, the International Cancer Genome Consortium; IHC, immunohistochemical; Kidney Renal Clear Cell Carcinoma; mRNA, messenger RNA; NK, natural killer; NAT, Normal Adjacent Tissue; scRNA-seq, single-cell RNA sequencing; TCGA, The Cancer Genome Atlas; TRIM28, tripartite motif-containing 28; UMAP, Uniform Manifold Approximation and Projection. **:P<0.01;***:P<0.001.

Figure 2. TRIM28 modulates CD8⁺ T cell-mediated antitumor immunity in ccRCC. (A–B) Correlation between TRIM28 expression and T cell infiltration in the TCGA-KIRC cohort. High TRIM28 expression was significantly associated with reduced CD8⁺ T cell infiltration (A) and lower overall T cell abundance scores (B). (C–D) Relationship between TRIM28 expression and tumor microenvironment scores. High TRIM28 expression was linked to a higher stromal score (C) and a lower immune score (D), suggesting a more immune-excluded tumor phenotype. (E–F) Cytotoxicity assays in co-cultures of 786-O RCC cells with activated CD8⁺ T cells. Tumor cells with TRIM28 knockdown (sh-TRIM28) showed increased susceptibility to CD8⁺ T cell-mediated cytotoxicity, as demonstrated by higher proportions of dead tumor cells compared with control groups (sh-Control). (G) Cytokine release assays in co-culture systems. TRIM28 knockdown enhanced CD8⁺ T cell activation, as evidenced by significantly elevated IFN-γ and TNF-α levels in supernatants from sh-TRIM28 tumor cell co-cultures. (H,I) In vivo tumor growth assessment in nude and BALB/c mice. No significant difference in tumor volume or weight was observed between sh-Control and sh-TRIM28 groups in immunodeficient nude mice. However, in immunocompetent BALB/c mice, TRIM28 knockdown led to significantly reduced tumor growth. (J–K) Enhanced T cell effector function in sh-TRIM28 tumors in vivo. CD8⁺ T cells isolated from Renca tumors in BALB/c mice exhibited increased cytotoxic activity (J) and enhanced IFN-γ production (K) when co-cultured with TRIM28 knockdown tumors. (L) CD8⁺ T cell infiltration was significantly higher in sh-TRIM28 tumors in BALB/c mice, while no significant change was observed in CD4⁺ T cell infiltration between groups. ccRCC, clear cell renal cell carcinoma; IFN-γ, interferon-gamma; KIRC, Kidney Renal Clear Cell Carcinoma; MFI, Mean Fluorescence Intensity; ns, not significant; TCGA, The Cancer Genome Atlas; TNF-α, tumor necrosis factor alpha; TRIM28, tripartite motif-containing 28. *:P<0.05;**:P<0.01;***:P<0.001.

Figure 3. TRIM28 interacts directly with PARP1. (A) Co-immunoprecipitation (co-IP) followed by silver staining of Flag-tagged TRIM28 expressed in 786-O cells. Distinct protein bands in the immunoprecipitated complex suggest specific interacting proteins. (B) Liquid chromatography–tandem mass spectrometry (LC–MS/MS) analysis of co-IP eluates identified PARP1 as the most significantly enriched interactor of TRIM28. Representative data from mass spectrometry analysis are shown in online supplemental figure 4A,B. (C–D) Reciprocal co-IP assays in HEK293T cells stably expressing Flag-tagged TRIM28 and Myc-tagged PARP1. Co-IP experiments using anti-Flag (C) and anti-Myc (D) antibodies confirm the physical interaction between TRIM28 and PARP1. (E–H) Endogenous co-IP assays in HEK293T, 786-O, and CAKI-1 cells using anti-TRIM28 and anti-PARP1 antibodies. The co-IP results confirm the interaction between endogenous TRIM28 and PARP1 under physiological conditions. Representative data are shown for all three cell lines. (I) GST pull-down assays demonstrate direct binding between recombinant GST-tagged TRIM28 and PARP1 in a cell-free system, supporting a direct interaction between the two proteins. (J) Immunofluorescence microscopy reveals colocalization of TRIM28 and PARP1 within the nucleus of 786-O and CAKI-1 cells, suggesting potential collaboration in nuclear processes. (K–N) Mapping of the interaction domains between TRIM28 and PARP1. Truncation mutants of both proteins were generated based on their known domain structures. Co-IP assays (L and M) in HEK293T cells expressing these mutants showed that the RING domain of TRIM28 specifically interacts with the catalytic domain of PARP1. (O) GST pull-down assays with truncated PARP1 proteins confirm that the TRIM28 RING domain binds directly to the PARP1 catalytic core, further establishing a domain-dependent interaction between TRIM28 and PARP1. DAPI:4′,6-diamidino-2-phenylindole; IB, immunoblotting; PARP1, poly (ADP-ribose) polymerase 1; TRIM28, tripartite motif-containing 28.

Figure 4. TRIM28 promotes SUMO2-dependent SUMOylation of PARP1 in ccRCC. (A–B) Ni²⁺-NTA pull-down assays in HEK293T cells cotransfected with His-tagged SUMO isoforms (SUMO1–4), Flag-tagged TRIM28, Myc-tagged PARP1, and HA-tagged UBC9. TRIM28 specifically enhances SUMO2 modification of PARP1, as shown by increased PARP1 SUMOylation in the presence of TRIM28. (C–D) Identification of SUMOylation sites on PARP1 using the GPS-SUMO prediction tool. Mutation of lysine residues K203 and K486 to arginine (K→R) reduces SUMOylation of PARP1, with simultaneous mutation of both sites completely abrogating SUMO2 modification, indicating their critical role in SUMOylation. (E) Co-IP assays demonstrate that mutation of K203 and K486 on PARP1 significantly diminishes the interaction between PARP1 and TRIM28, suggesting a SUMO-dependent mechanism for the TRIM28-PARP1 complex formation. (F) Overexpression of TRIM28 carrying a C651A mutation (E3 ligase-deficient) fails to promote PARP1 SUMOylation to the same extent as wild-type (WT) TRIM28, confirming that TRIM28 functions as an E3 SUMO ligase for PARP1. (G) Comparison of the effects of overexpressing various SUMO E3 ligases (PIAS1–4 and TRIM28) on PARP1 SUMOylation in 786-O cells. Only TRIM28 robustly enhances PARP1 SUMOylation, as confirmed by immunoprecipitation of endogenous PARP1. (H–I) TRIM28 knockdown in 786-O and CAKI-1 ccRCC cells significantly reduces PARP1 SUMOylation, while overexpression of TRIM28 (but not its E3-deficient mutant) leads to a dose-dependent increase in PARP1 SUMO2 modification. Representative data are shown for both cell lines. (J–K) Immunofluorescence staining of 786-O and CAKI-1 cells demonstrates substantial nuclear colocalization of PARP1 with SUMO2/3. This colocalization is markedly reduced on stable knockdown of TRIM28, supporting a role for TRIM28 in mediating nuclear SUMOylation of PARP1. ccRCC, clear cell renal cell carcinoma; co-IP, co-immunoprecipitation; DAPI, 4′,6-diamidino-2-phenylindole; IB, immunoblotting; PARP1, poly (ADP-ribose) polymerase 1; TRIM28, tripartite motif-containing 28. ***:P<0.001.

Figure 5. TRIM28 SUMOylation regulates PARP1 stability by inhibiting RNF114-dependent K63-linked ubiquitination. (A) Treatment with the SUMOylation inhibitor 2-D08 leads to a significant decrease in PARP1 protein levels in both 786-O and CAKI-1 cells, indicating a role for SUMOylation in PARP1 stabilization. (B–C) Cycloheximide (CHX) chase assays in 786-O cells demonstrate that TRIM28 knockdown accelerates PARP1 degradation, while overexpression of wild-type (WT) TRIM28 significantly prolongs PARP1 half-life. In contrast, the SUMOylation-deficient TRIM28^C651A mutant exhibits reduced capacity to stabilize PARP1. (D) Overexpression of the non-SUMOylatable PARP1 mutant (PARP1^2KR) in 293 T cells leads to accelerated degradation compared with wild-type PARP1 in CHX assays, suggesting that SUMOylation contributes to PARP1 stability. (E–F) Ubiquitination assays in 786-O cells show that TRIM28 knockdown results in increased PARP1 ubiquitination (E). Conversely, overexpression of TRIM28^WT significantly reduces PARP1 ubiquitination, whereas TRIM28^C651A has no effect (F). (G) Ubiquitination assays in 293 T cells cotransfected with HA-Ub and either PARP1^WT or PARP1^2KR reveal increased ubiquitination of the non-SUMOylatable PARP1^2KR mutant, further supporting the role of SUMOylation in PARP1 stabilization. (H) TRIM28 specifically inhibits K63-linked ubiquitination of PARP1, as shown by ubiquitination assays using various lysine-specific ubiquitin mutants (HA-Ub-K6, HA-Ub-K11, HA-Ub-K27, HA-Ub-K33, HA-Ub-K48, and HA-Ub-K63) in 293 T cells. (I) Overexpression of a K63R ubiquitin mutant abolishes TRIM28’s ability to reduce PARP1 ubiquitination, confirming that TRIM28 specifically targets K63-linked ubiquitin chains. (J–K) TRIM28 overexpression abrogates RNF114-mediated PARP1 degradation (J). Co-IP assays show that TRIM28 reduces the interaction between RNF114 and PARP1, suggesting that TRIM28 inhibits RNF114-mediated ubiquitination of PARP1 (K). (L) Exogenous validation in 293 T cells reveals that RNF114 increases K63-linked ubiquitination of Myc-PARP1, and overexpression of TRIM28 reverses this RNF114-mediated K63-linked ubiquitination of PARP1. co-IP, co-immunoprecipitation; IB, immunoblotting; PARP1, poly (ADP-ribose) polymerase 1; TRIM28, tripartite motif-containing 28. *:P<0.05;***:P<0.001.

Figure 6. TRIM28 and PARP1 cooperatively regulate PD-L1 expression via NF-κB signaling in ccRCC. (A–B) Gene Ontology (GO) and Kyoto Encyclopedia of Genes and Genomes (KEGG) enrichment analyses of differentially expressed genes following TRIM28 or PARP1 knockdown in 786-O cells. The “PD-L1 expression and PD-1 checkpoint pathway in cancer” is significantly enriched, suggesting a role for TRIM28 and PARP1 in immune regulation. (C) RNA-seq peak profiles demonstrate a significant reduction in PD-L1 transcription on knockdown of either TRIM28 or PARP1 in 786-O cells. (D) Western blot analysis confirms a significant decrease in PD-L1 protein expression on TRIM28 or PARP1 knockdown. (E) Overexpression of PARP1 in TRIM28-depleted 786-O cells partially restores PD-L1 protein levels, suggesting functional cooperation between TRIM28 and PARP1 in regulating PD-L1. (F–G) TRIM28 knockdown reduces PD-L1 expression in 786-O cells under both IFN-γ–stimulated and unstimulated conditions, as shown by Western blotting (F) and flow cytometry (G). (H–I) Immunohistochemistry analysis of ccRCC clinical samples reveals a positive correlation between TRIM28 expression and PD-L1 positivity in tumor tissues, indicating clinical relevance of the TRIM28-PARP1-PD-L1 axis. (J) Gene Set Enrichment Analysis (GSEA) of differentially expressed genes following TRIM28 or PARP1 knockdown identifies multiple signaling pathways involved in PD-L1 regulation. (K) ChIP-Atlas enrichment analysis reveals significant occupancy of the PD-L1 promoter by key transcription factors, including NF-κB (p65), HIF1A, STAT1, and STAT3. (L) ChIP-qPCR analysis of PD-L1 promoter regions shows that TRIM28 knockdown significantly reduces p65 binding to the PD-L1 promoter, an effect that is reversed by wild-type PARP1 overexpression. (M) TRIM28 overexpression enhances p65 recruitment to the PD-L1 promoter, while a SUMOylation-deficient TRIM28 mutant fails to do so, indicating the importance of TRIM28-mediated PARP1 modification for p65 activity at the PD-L1 locus. (N) In vivo pharmacological inhibition of NF-κB signaling with bortezomib significantly impairs tumor growth in a subcutaneous xenograft model using Renca murine RCC cells in BALB/c mice. (O) Flow cytometric analysis of tumor-infiltrating cells shows that bortezomib treatment reduces PD-L1 expression and the percentage of PD-1+CD8⁺ T cells, highlighting the role of NF-κB–driven PD-L1 expression in immune evasion in vivo. ccRCC, clear cell renal cell carcinoma; ChIP, chromatin immunoprecipitation; IFN-γ, interferon-gamma; MFI, Mean Fluorescence Intensity; ns, not significant; PARP1, poly (ADP-ribose) polymerase 1; PD-1, programmed cell death protein-1; PD-L1, programmed death-ligand 1; qPCR, quantitative PCR; TRIM28, tripartite motif-containing 28. **:P<0.01;***:P<0.001.

Figure 7. TRIM28 promotes PD-L1 expression and tumor immune evasion via modulation of NAD⁺ metabolism. (A) Untargeted metabolomic profiling of 786-O cells stably overexpressing TRIM28 reveals increased levels of nicotinic acid and butyrylcarnitine, key intermediates in NAD⁺ biosynthesis. (B–C) Overexpression of TRIM28 in 786-O and CAKI-1 cells results in a significant reduction in intracellular NAD⁺ and NADH levels. This decrease is partially rescued by silencing PARP1 expression. (D–E) Overexpression of a SUMOylation-deficient TRIM28 mutant fails to reduce NAD⁺ and NADH levels, suggesting that the metabolic regulation by TRIM28 is dependent on its SUMOylation activity. (F) TRIM28 knockdown in 786-O cells reduces P65 acetylation at lysine 310 (P65(K310)), while co-depletion of SIRT1 restores acetylation levels and increases PD-L1 expression, indicating the role of TRIM28 in regulating SIRT1-mediated P65 acetylation. (G) TRIM28 overexpression increases P65(K310) acetylation, an effect that can be reversed by supplementation with the NAD⁺ precursor nicotinamide mononucleotide (NMN). NMN supplementation also attenuates TRIM28-induced upregulation of PD-L1. (H–I) Co-culture of TRIM28-overexpressing tumor cells with primary human CD8⁺ T cells results in reduced NAD⁺ and NADH levels in T cells, suggesting metabolic competition between the tumor and immune cells. (J–K) TRIM28-overexpressing tumor cells suppress glycolytic activity and mitochondrial respiration in co-cultured CD8⁺ T cells, indicating the impact of TRIM28 on T cell metabolism. (L) Flow cytometric analysis of co-culture supernatants shows that TRIM28 overexpression significantly reduces tumor cell death in CD8⁺ T cell co-cultures, an effect that can be reversed by NMN supplementation. (M) In vivo subcutaneous xenograft experiments in BALB/c mice show that tumors derived from TRIM28-overexpressing cells grow faster than control tumors, and NMN supplementation mitigates this effect. (N) NAD⁺ and NADH levels are significantly reduced in TRIM28-overexpressing tumor tissues compared with controls, consistent with the in vitro findings. (O) Immunohistochemical staining of tumor tissues reveals increased PD-L1 expression and reduced CD4⁺ and CD8⁺ T cell infiltration in TRIM28-overexpressing tumors, both of which are partially reversed by NMN treatment, confirming that TRIM28 modulates the tumor immune microenvironment. ECAR, extracellular acidification rate; FCCP, Carbonyl cyanide-p-trifluoromethoxyphenylhydrazone; FITC, Fluorescein Isothiocyanate; NC, negative control; ns, not significant; OCR, oxygen consumption rate; OE, overexpression; PE, Phycoerythrin; PI, Propidium Iodide; PARP1, poly (ADP-ribose) polymerase 1; PD-L1, programmed death-ligand 1; TRIM28, tripartite motif-containing 28. *: P<0.05; **: P<0.01; ***: P<0.001.

Figure 8. Structure-based screening identifies Eltrombopag as a functional TRIM28 inhibitor that enhances antitumor immunity. (A–C) Structure-based virtual screening of FDA-approved and de novo compound libraries identified three potential small-molecule binders of TRIM28, including Eltrombopag (ZINC000011679756), Ergotamine (ZINC000052955754), and VX-809 (ZINC000064033452), selected based on AutoDock Vina docking affinity scores. (D–F) Immunoblot analysis shows that treatment of 786-O cells with Eltrombopag (ZINC000052955754) for 24 hours significantly reduces PARP1 SUMOylation levels, confirming its functional inhibition of TRIM28. (G) Molecular dynamics simulations show a stable interaction between Eltrombopag and TRIM28 over a 100 ns trajectory, supporting the potential binding compatibility of the compound. (H) In vivo efficacy of Eltrombopag was evaluated in subcutaneous xenograft tumor models in BALB/c mice, where treatment with Eltrombopag significantly inhibited tumor growth compared with control. (I,J) Flow cytometry analysis of tumor-infiltrating lymphocytes revealed increased levels of IFN-γ, TNF-α, GZMB, and perforin in the Eltrombopag-treated group, indicating enhanced immune activation. (K) Increased CD8⁺ T cell infiltration was observed in the Eltrombopag-treated tumors, suggesting enhanced antitumor immune response. (L) A preclinical RCC model using immunocompromised NCG mice engrafted with RCC patient-derived xenografts (PDXs) was used to assess the impact of Eltrombopag on anti-PD-1 therapy. Tumor-specific CD8⁺ T cell transfer was used to replicate the tumor microenvironment. (M–P) Combination therapy with Eltrombopag and anti-PD-1 significantly reduced tumor burden and improved overall survival (OS) compared with either therapy alone in the humanized immune system model. The combination treatment demonstrated a more substantial therapeutic effect. FDA, Food and Drug Administration; IB, immunoblotting; IFN-γ, interferon-gamma; IP, immunoprecipitation; MFI, Mean Fluorescence Intensity; ns, not significant; RMSD, Root Mean Square Deviation; PARP1, poly (ADP-ribose) polymerase 1; PBS, phosphate-buffered saline; PD-1, programmed cell death protein-1; RCC, renal cell carcinoma; TNF-α, tumor necrosis factor alpha; TRIM28, tripartite motif-containing 28. *: P<0.05; **: P<0.01; ***: P<0.001.