The G-patch protein family is defined by a conserved glycine-rich structural motif. Several members, including GPATCH1, GPATCH8, and GPATCH11, play critical roles in RNA splicing (21, 24, 29). However, the function of GPATCH3 remains poorly characterized. To explore whether GPATCH3 is involved in RNA splicing, we performed co-immunoprecipitation coupled with mass spectrometry (Co-IP/MS) to identify GPATCH3-interacting proteins ( Figures 1A, B ; Supplementary Data Sheet S1 ). Gene Ontology (GO) enrichment analysis revealed that the GPATCH3-associated proteins are significantly enriched in splicing-related biological processes ( Figure 1C ). Interaction network analysis further demonstrated that GPATCH3 is primarily associated with components of the U2 snRNP and the PRPF19 complex, both of which are core elements of the spliceosome ( Figures 1D, E ).

To validate the Co-IP/MS findings, we immunoprecipitated Flag-tagged GPATCH3 and detected the co-precipitated proteins via western blotting ( Figure 2A ). The results confirmed that GPATCH3 interacts with U2 snRNP and associated proteins, such as DHX15 and SNRPA1, and the PRPF19 complex, including PRPF19 and HSPA8 ( Figure 2A ). Immunofluorescence-based subcellular localization analysis in multiple cell lines showed that GPATCH3 is predominantly distributed in the nucleus ( Figure 2B ), further supporting its potential involvement in splicing regulation.

We next sought to determine whether GPATCH3 directly interacts with its binding partner. GST pulldown assays confirmed a direct interaction between GPATCH3 and DHX15 ( Figure 2C ). To map the structural domains responsible for this interaction, we generated a series of GPATCH3 truncation mutants guided by structural predictions from AlphaFold ( Figure 2D ). Deletion of either the N-terminal region (amino acids 1–236) or the G-patch domain (amino acids 410–458) abrogated the interaction with DHX15 ( Figure 2E ), indicating that both regions are required for their association.

Taken together, these findings suggest that GPATCH3 may participate in pre-mRNA splicing through its physical and functional interaction with DHX15.

Previous studies have shown that certain G-patch proteins, such as GPATCH2, GPATCH4, and RBM5, can directly bind to DHX15 via their G-patch domains and markedly enhance its ATPase activity, thereby contributing to spliceosome disassembly and intron lariat resolution (22, 30, 31). We therefore investigated whether GPATCH3 exerts a similar regulatory effect. In vitro ATPase assays revealed that GPATCH3 significantly stimulated the ATPase activity of DHX15 in a dose-dependent manner ( Figures 3A, B ). These results indicate that GPATCH3 can function as a cofactor to potentiate DHX15 enzymatic activity.

Given that DHX15 activity governs the dynamic remodeling of the spliceosome and affects pre-mRNA splicing fidelity, we next assessed whether GPATCH3 influences alternative splicing outcomes. Using an E1A minigene reporter system (32), we showed that GPATCH3 overexpression reduced usage of the most proximal 5′ splice site, resulting in decreased levels of the 12S/13S isoforms and increased levels of the 9S isoform ( Figures 3C, D ). These data demonstrate that GPATCH3 modulates pre-mRNA alternative splicing.

Pan-cancer transcriptomic analysis using GEPIA2 (33) revealed that GPATCH3 is significantly upregulated in a broad spectrum of malignancies compared to corresponding normal tissues ( Figure 4A ). To explore the clinical relevance of this upregulation, we performed survival analyses using TCGA datasets. Kaplan–Meier curves demonstrated that high GPATCH3 expression is associated with reduced overall survival in multiple cancer types, including lower-grade glioma (LGG), liver hepatocellular carcinoma (LIHC), and prostate adenocarcinoma (PRAD) ( Figures 4B–D ).

Consistent with these findings, in vivo functional studies showed that depletion of GPATCH3 significantly suppressed tumor growth in mouse xenograft models ( Figures 4E–H ). Together, these data highlight GPATCH3 as a potential oncogenic factor that promotes tumor progression and may serve as a prognostic biomarker in certain cancers.

The dynamic interplay between tumors and their surrounding stromal and immune constituents—collectively termed the tumor microenvironment (TME)—has emerged as a critical determinant of cancer progression and therapy response (34–36). Given the well-established role of immune cell infiltration in modulating antitumor immunity, we hypothesized that GPATCH3, a poorly characterized gene in cancer biology, might influence TME composition. To test this, we employed TIMER2.0 (37) to assess the immunomodulatory impact of GPATCH3. Notably, GPATCH3 expression was significantly negatively correlated with cytotoxic CD8⁺ T cells and NK cell infiltration, while showing positive correlations with immunosuppressive populations such as myeloid-derived suppressor cells (MDSCs) and cancer-associated fibroblasts (CAFs) across multiple cancer types ( Figures 5A–C ).

Since the extent of immune cell infiltration is closely influenced by the dynamic balance of immunoregulatory factors, we next evaluated the associations between GPATCH3 and key classes of immune modulators, including immunoinhibitors, immunostimulators, receptors, and chemokines. Heatmap analyses revealed that GPATCH3 expression was broadly negatively correlated with chemokines such as CXCL5 and CXCL8, as well as immunostimulatory molecules including CD40LG and TNFSF15, in a pan-cancer context ( Figures 6A–D ). These alterations may contribute to the reprogramming of immune infiltration patterns within the TME.

Together, these findings from multiple lines of analysis support a model in which GPATCH3 functions as a key orchestrator of the immunosuppressive TME, likely through regulating the splicing and expression of immune-related genes involved in these pathways.

Given the role of GPATCH3 in alternative splicing regulation, we investigated whether its deficiency alters the splicing patterns of key immunoregulatory genes known to generate functionally distinct isoforms. For example, CXCR3 produces two major isoforms—CXCR3-A and CXCR3-B—through alternative 3′ splice site selection within exon 2 (38). CXCR3-B utilizes a more proximal 3′ splice site, resulting in a longer extracellular N-terminus compared to CXCR3-A. Functionally, CXCR3-A promotes cell proliferation, survival, and chemotaxis, whereas CXCR3-B is associated with growth inhibition and apoptosis (39). We observed an increased ratio of the CXCR3-A isoform upon GPATCH3 depletion ( Figure 7A ).

Similarly, alternative splicing of CCL5 was also perturbed in the absence of GPATCH3 ( Figure 7B ). Given that non-canonical CCL5 isoforms may influence mRNA stability or translational efficiency, these alterations suggest a potential mechanism by which GPATCH3 modulates chemokine signaling. We further identified aberrant splicing events in additional immune-related genes, including CD44 and FOXP3 ( Figures 7C, D ). Together, these results suggest that GPATCH3 modulates alternative splicing of immuneregulatory genes, thereby may affect immune infiltration.

Because alternative splicing can influence gene expression levels either directly or indirectly, we further investigated the impact of GPATCH3 deficiency on the gene expression. Notably, depletion of GPATCH3 led to significant upregulation of CXCL8, CCL5, LAG3, and PVRL2 ( Figure 7E )—genes with well-established roles in immune cell recruitment, immune checkpoint signaling, and tumor–immune interactions.

Collectively, our findings demonstrate that loss of GPATCH3 disrupts both the splicing and expression of genes involved in immune modulation, thereby potentially reshaping the immune landscape of the TME.

Primary antibodies used for immunoblotting included: anti-PRPF19 (Cat. # A12590), anti-SART3 (Cat. # A12124), anti-SNRPA (Cat. # A6410), anti-HSPA8 (Cat. # A2487) from ABclonal; anti-DHX15 (Cat. # L2221) from Santa Cruz Biotechnology; anti-FLAG (clone M2, Cat. # F1804), anti-GST (Cat. # G1160) from Sigma-Aldrich; anti-His (clone 27E8, Cat. # 2366) from Cell Signaling Technology. Secondary antibodies were Peroxidase-Conjugated AffiniPure Goat Anti-Rabbit IgG (H+L) (Cat. # 111-035-003; Jackson Lab) and Goat Anti-Mouse IgG (H+L) (Cat. # 115-035-003; Jackson Lab).

A549, HEK293T, MC38, and B16/F10 cell lines were obtained from the American Type Culture Collection (ATCC) and maintained in Dulbecco’s modified Eagle’s medium (DMEM; Gibco) supplemented with 10% heat-inactivated fetal bovine serum (Lonsera), 100 U/mL of penicillin and 100 µg/mL of streptomycin at 37°C in a humidified incubator with 5% CO₂. Prior to experimentation, cell line authenticity was verified by short tandem repeat (STR) profiling, and routine mycoplasma testing using PCR-based detection confirmed the absence of mycoplasma contamination throughout the study period.

Six- to eight-week-old C57BL/6 mice were randomly assigned to two groups (5 mice per group) and injected subcutaneously with either GPATCH3-depleted MC38 or B16/F10 tumor cells, or corresponding control cells. At the experimental endpoint, mice were anesthetized and euthanized. Tumors were excised and measured for volume and weight.

Flag-tagged GPATCH3 plasmid was generated by cloning the human GPATCH3 coding sequence (CDS) into pcDNA6-N-Flag vectors. Plasmids expressing GST-tagged and His-tagged proteins were generated by inserting the CDS into pGEX-4T-1 and pET-28a (+) vectors, respectively. Lentiviral knockdown plasmids were constructed by inserting the annealed shRNAs into pLV-H1-EF1α-puro vector (Cat. # SORT-B19; Biosettia). The shRNA sequence specifically targeting GPATCH3 was: 5′–GGACATGAGTGTGTACTATGA–3′ (shGPATCH3). The nontargeting control shRNA (shCtrl) sequence was 5′–CAACAAGATGAAGAGCACC–3′. All constructs were validated by Sanger sequencing.

Lentiviral particles were produced in HEK293T cells by co-transfecting the lentiviral vector with three packaging plasmids (pVSV-G, pMDLg/pRRE, and pRSV-Rev) using calcium phosphate precipitation. Target cells were transduced with the resulting viral supernatant by spinoculation (1,000× g, 30 min at 32°C) in the presence of 8 μg/mL polybrene (Cat. # H9268; Sigma-Aldrich). Stable cell lines were generated by selecting transduced cells with puromycin (1–2 μg/mL) starting 48 hours post-transduction.

Cells were washed with ice-cold phosphate-buffered saline (PBS) and lysed with lysis buffer (25 mM Tris-HCl, pH 7.4; 150 mM NaCl; 1% NP-40; 1 mM EDTA; 1 mM EGTA; 1× protease and phosphatase inhibitor cocktail). Samples were incubated on ice for 30 min and sonicated for 15 seconds. Cell lysates were centrifuged at 12,000 × g for 15 min at 4°C. Ten percent of the supernatant was prepared as the input. For immunoprecipitation of Flag-tagged proteins, the supernatant was incubated with anti-Flag M2 magnetic beads (Cat. # M8823; Sigma-Aldrich) at 4°C for 4 h on a rotor. After immunoprecipitation, the beads were washed four times with lysis buffer. Then, 3× Flag peptide (Cat. # F4799; Sigma-Aldrich) was added to elute the bound proteins from beads. Proteins in the supernatants were resolved on SDS–PAGE and detected by immunoblotting or silver staining (Cat. # C510027; Sangon Biotech). To identify GPATCH3-interacting partners, the eluted proteins were subjected to silver staining and semi-quantitative mass spectrometry (liquid chromatography-tandem mass spectrometry) analysis by the Proteomics Core Facility at the Institutes of Biomedical Sciences, Fudan University.

Proteins were quantified by BCA assay (Cat. # ZJ102; Epizyme), separated by SDS–polyacrylamide gel electrophoresis (SDS–PAGE) and transferred to Nitrocellulose membranes (Cat. # 10600002; Cytiva). The membranes were blocked with 5% nonfat milk in Tris-buffered saline at room temperature for 1 h and then probed with primary antibodies overnight at 4°C, followed by HRP-conjugated secondary antibodies at room temperature for 1 h. Protein signals were finally detected using an enhanced chemiluminescence (ECL) system (Cat. # SB-WB012; Share-Bio) and captured with a digital imaging system (Tanon).

GST- or His-tagged proteins were expressed in BL21 (DE3) bacteria and grown in LB media until the OD600 reaches 0.8~1. Then, the cultures were induced at 18°C overnight with 0.2 mM isopropyl β-D-thiogalactopyranoside (IPTG; Cat. # A100487; Sangon Biotech). Bacterial extracts were prepared by sonication in PBS supplemented with protease inhibitors, followed by centrifugation at 12,000 rpm at 4°C for 15 min. The supernatants were incubated with glutathione agarose beads (Cat. # 16100; Thermo Fisher) on a rotating mixer for 2 h at 4°C. The beads were washed three times and the purified proteins were ready for GST pulldown assay. His-tagged proteins were purified similarly using Ni-NTA Magarose Beads (Cat. # SM008025; Smart-Lifesciences) and eluted with imidazole (Cat. # A600277; Sangon Biotech). His-tagged proteins were incubated with GST or GST-tagged GPATCH3 proteins bounded to glutathione agarose beads in binding buffer (0.1% NP-40 in PBS). The binding reaction was performed 2 h with a rotating mixer at 4°C, after which the beads were washed three times. GST proteins were eluted with elution buffer (50 mM Tris-HCl, pH 8.0, 150 mM NaCl, 10 mM reduced glutathione). The samples were boiled at 95°C for 5 min and subjected to Western blot analysis.

GST-GPATCH3 proteins were purified from E.coli and Flag-tagged DHX15 proteins were purified from transfected HEK293T cells. Different doses of GPATCH3 were incubated at 37°C with 100 ng DHX15 for 30 min in the reaction buffer (50 mM HEPES, pH 7.4, 150 mM NaCl, 5% Glycerol, 10 mM MgCl_2, 0.1% NP-40, 1 mM DTT). Following incubation, ATP was added at a final concentration of 100 μM and incubated for another 30 min at 37°C. Then the reaction was detected by Malachite Green Phosphate Detection Kit (Cat. # S0196M; Beyotime) according to the manufacturer’s protocol.

Total RNA was extracted using TRIzol reagent (Cat. # R401-01; Vazyme) and reverse transcribed into cDNA using HiScript III RT SuperMix (Cat. # R323-01; Vazyme). Semi-quantitative RT–PCR and a Tanon Bio-Fragment Analyzer (BiOptic) were used to analyze alternative spliced products. The oligonucleotides used for RT–PCR were as follows: ElA569, 5′–ATTATCTGCCACGGAGGTGT–3′; ElA1315, 5′–GGATAGCAGGCGCCATTTTA–3′; CXCR3-RT-F, 5′–CCCAGCAGCCAGAGCACCAGC–3′; CXCR3-RT-R, 5′–CAGGAAGGCCCGGTCGAAGTTC–3′; CCL5-RT-F, 5′–CACCCTGCTGCTTTGCCTACA–3′; CCL5-RT-R, 5′–AGGTTCAAGGACTCTCCATCC–3′; CD44-RT-F, 5′–GGAAGAAACAGCTACCCAGA–3′; CD44-RT-R, 5′–CATTGAAAGAGGTCCTGTCC–3′; FOXP3-RT-F, 5′–TCCAGGGCCGAGATCTTCGAG–3′; FOXP3-RT-R, 5′–TGAGGGAGAAGACCCCAGTGG–3′.

The oligonucleotides used for RT–qPCR were as follows: CXCL8-RT-F, 5′–ATACTCCAAACCTTTCCACCC–3′; CXCL8-RT-R, 5′–TCTGCACCCAGTTTTCCTTG–3′; LAG3-RT-F, 5′–CCTACACCTGCCATATCCATC–3′; LAG3-RT-R, 5′–AGCGTTCTTGTCCAGATACTG–3′; CCL5-RT-F, 5′–TGCCCACATCAAGGAGTATTTC–3′; CCL5-RT-R, 5′–CCATCCTAGCTCATCTCCAAAG–3′; PVRL2-RT-F, 5′–GCATGAGAGCTTCGAGGAAC–3′; PVRL2-RT-R, 5′–GGAGATGGACACTTCAGGAG–3′.

Data in the study are presented as means ± S.D. and analyzed statistically using GraphPad Prism 8.0 software. P values were calculated using unpaired two-tailed Student’s t-test between two groups and one-way or two-way ANOVA for multiple groups, respectively. P < 0.05 was considered statistically significant.