RNA splicing regulator EIF3D regulates the tumor microenvironment through immunogene-related alternative splicing in head and neck squamous cell carcinoma

Data collection and analysis

First, we acquired mRNA expression data for EIF3D from the TIMER database (https://cistrome.shinyapps.io/timer/) encompassing both pan-cancer cohorts and samples from normal individuals. Additionally, transcriptome sequencing data, along with corresponding clinical information for 522 HNSC samples and 44 normal samples, were procured from The Cancer Genome Atlas (TCGA) database (https://portal.gdc.cancer.gov). Subsequently, we categorised HNSC patients into EIF3D high-expression group and the EIF3D low-expression group based on the median EIF3D expression level. To describe differences in the composition of the TME between these groups, we employed ESTIMATE to compute ImmuneScores, StromalScores, and ESTIMATEScores. Furthermore, we conducted an analysis of immune cell infiltration within the tumour based on CIBERSORT software (https://cibersort.stanford.edu/). In order to compare immuno-therapy scores between the high and low expression EIF3D groups, we obtained immuno-therapy scores for samples within the TCGA database from TCIA (https://tcia.at), specifically for anti-CTLA4 and anti-PD1 inhibitor treatments.

For a comprehensive understanding of mRNA alternative splicing (AS) patterns, we retrieved detailed AS data from TCGA SpliceSeq (https://bioinformatics.mdanderson.org/TCGASpliceSeq/). This data encompassed seven primary AS event types, including Alternate Acceptor site (AA), Alternate Donor site (AD), Alternate Promoter (AP), Alternate Terminator (AT), Exon Skipping (ES), Mutually Exclusive Exons (ME), and Retained Intron (RI). AS patterns of protein-coding genes in HNSC samples were delineated based on specific criteria, requiring a Percent Spliced In (PSI) value of ≥75% and a minimum standard deviation of ≥0.10. To delve deeper into the role and underlying mechanisms of EIF3D in HNSC, this study subsequently conducted a series of in vitro cellular experiments.

Lentivirus information

We obtained all siRNA duplexes from Genepharma (Suzhou, China). The non-targeting control siRNA had the following sequence: 5′-UUCUCCGAACGUGUCACGUTT-3′ (sense). The siRNA targeting EIF3D (siEIF3D) featured the following sequence: 5′-CCUAGAAUACUACGACAAATT-3′ (sense).

Cell culture and transfections

FaDu cell line (CL-0083, Procell Life Science and Technology Co., Ltd., China) was maintained in MEM (PM150410, Procell Life Science and Technology Co., Ltd., China) with 10% fetal bovine serum (FBS) (10091148, Gibco, China), 100 μg/mL streptomycin, and 100 U/mL penicillin (SV30010, Hyclone, USA) at 37°C with 5% CO₂. Transfection of siRNA into the cells was carried out using Lipofectamine™ RNAiMAX Transfection Reagent (13778030, Invitrogen, Carlsbad, CA, USA) following the manufacturer’s instructions. After 48 hours, transfected cells were collected for RT-qPCR analysis.

Quantitative real-time PCR

Glyceraldehyde-3-phosphate dehydrogenase (GAPDH) served as the control gene for assessing the impact of EIF3D knockdown. Standard procedures were followed for cDNA synthesis, and RT-qPCR was executed using the Bio-Rad S1000 system with HieffTMqPCR SYBR^® Green Master Mix (Low Rox Plus; Yeasen, Shanghai, China). Supplementary Table 1 contains the information of the primers used. To standardize the concentration of each transcript, we normalised them to the GAPDH mRNA level using the 2^−ΔΔCT method [24]. Statistical comparisons were conducted using the paired Student’s t-test, facilitated by GraphPad Prism software (Version 8.0, San Diego, CA, USA).

Western blot

FaDu cells were lysed in ice-cold Wash Buffer (1× PBS, 0.1% SDS, 0.5% NP-40, and 0.5% sodium deoxycholate), supplemented with a protease inhibitor cocktail (Roche, Basel, Switzerland), and incubated on ice for 30 minutes. Subsequently, samples were boiled for 10 minutes in boiling water with 1X SDS sample buffer and separated on 10% SDS-PAGE. Membranes were then subjected to incubation with primary antibodies: EIF3D antibody (1:1,000, 66024-1-Ig, Proteintech, China) and GAPDH (1:1,000, 60004-1-Ig, Proteintech), followed by exposure to HRP-conjugated secondary antibodies. Detection of the bound secondary antibody (anti-rabbit, 1:10,000, SA00001-2, Proteintech, China, or anti-mouse, 1:10,000, AS003, ABclonal, China) was carried out using enhanced chemiluminescence (ECL) reagent (170506, Bio-Rad, Hercules, CA, USA).

Cell proliferation assay

We conducted the cell proliferation assay employing the Cell Counting kit-8 (CCK-8, 40203ES76, Yeasen, Shanghai, China). Briefly, FaDu cells were seeded at a density of 10,000 cells per well in 96-well culture plates. The cells in both the control and experimental groups received their respective treatments, while vials without cells served as blank controls. Following incubation for 0, 24, 48, and 72 hours at 37°C with 5% CO₂, we introduced 10 μl of CCK-8 solution into the culture medium and incubated it for an additional 3 hours at 37°C. We measured the optical density of the cells at an absorbance of 450 nm using a Microplate Reader (ELX800, Biotek, USA). The cell proliferation rate was calculated using the formula: proliferation rate = (experimental OD value - blank OD value)/(control OD value - blank OD value) × 100%.

Apoptosis assay

For the detection of cell apoptosis, we employed an Annexin V-APC/7-ADD apoptosis detection kit (KGA1026, KeyGEN BioTECH, China) following the manufacturer’s instructions. Specifically, FaDu cells were seeded into six-well plates and cultured for 24 hours before transfection with the plasmid for 48 hours. Subsequently, both treated and control cells were mixed with 5 μl Annexin V-APC and incubated at room temperature in the dark for 5 minutes, followed by incubation with 5 μl of 7-AAD reagents for another 5 minutes in the dark. The samples were then subjected to flow cytometry analysis using FACSCanto (BD, La Jolla, CA, USA).

Invasion assay

In vitro invasion assays were performed using transwell chambers (3422, Corning, USA). These chambers featured an 8 μm filter precoated with a thin layer of Matrigel (356234, BD Biosciences, USA), diluted at a ratio of 1:8 using serum-free medium. A total of 100 μl of the diluted matrigel was incubated in the chambers for 1 hour at 37°C with 5% CO₂, after which unsolidified supernatant was removed. Subsequently, 5 × 10⁴–10⁵ cells in 0.2 ml of serum-free medium were added to the inserts, while the transwell chambers were placed in medium containing 600 μl of 10% FBS (10091148, Gibco, China) as a chemoattractant in the lower chamber. The cells were then incubated for 24–48 hours at 37°C with 5% CO₂. After the incubation period, cells remaining on the upper membrane surface of the inserts were removed using a cotton swab, and the total number of invading cells that had entered the lower chamber were fixed with 4% paraformaldehyde (P0099, Beyotime, China) for 20 minutes. These cells were subsequently stained with 0.1% crystal violet (C0121, Beyotime, China) and observed and counted under an inverted microscope (MF52-N, Mshot, China) at 200× magnification.

Cell migration assay

In vitro migration assays were performed using transwell chambers (3422, Corning, USA). 5 × 10⁴–10⁵ cells in 0.2 ml serum-free medium were added to the transwell chambers with 8 μm filter, then the chambers inserted in medium with 600 ul 10% FBS (10091148, Gibco, China) served as a chemo-attractant in the lower chamber, and incubated for 24–48 h at 37°C and 5% CO₂. Cells remaining on the upper membrane surface of the inserts were then removed with a cotton swab, and the total number of cells that migrated into the lower chamber were fixed by 4% paraformaldehyde (P0099, Beyotime, China) for 20 min, then stained with 0.1% crystal violet (C0121, Beyotime, China). The migration cells were observed and counted under inverted microscope (MF52-N, Mshot, China) at 200× magnification.

RNA extraction and sequencing

Total RNA was treated with RQ1 DNase (Promega, Madison, WI, USA) to remove DNA. The quality and quantity of the purified RNA were determined by the absorbance at 260 nm/280 nm (A260/A280). RNA integrity was further verified by 1.5% agarose gel electrophoresis. For each sample, 1 μg of total RNA was used for RNA-seq library preparation. mRNAs were captured by VAHTS mRNA capture Beads (N401, Vazyme, China). The purified RNA was treated with RQ1 DNase (Promega) to remove DNA before used for directional VAHTS Universal V8 RNA-seq Library Prep Kit for Illumina (NR605) Polyadenylated mRNAs were purified and fragmented. Fragmented mRNAs were converted into double strand cDNA. Following end repair and A tailing, the DNAs were ligated to Adaptor (N323). After purification of ligation product and size fractioning to 300–500 bps, the ligated products were amplified and purified, quantified and stored at −80°C before sequencing. The strand marked with dUTP (the 2nd cDNA strand) is not amplified, allowing strand-specific sequencing. For high-throughput sequencing, the libraries were prepared following the manufacturer’s instructions and applied to Illumina Novaseq 6000 system for 150 nt paired-end sequencing.

RNA-seq raw data cleaning and alignment

Raw reads containing more than 2-N bases were first discarded. Then adaptors and low-quality bases were trimmed from raw sequencing reads using FASTX-Toolkit (Version 0.0.13). The short reads less than 16 nt were also dropped. After that, clean reads were aligned to the GRCh38 genome by HISAT2 [25] allowing 4 mismatches. Uniquely mapped reads were used for gene reads number counting and FPKM calculation (fragments per kilobase of transcript per million fragments mapped) [26].

Alternative splicing analysis

We defined and quantified alternative splicing events (ASEs) and regulated alternative splicing events (RASEs) between the samples using the ABLas pipeline, as previously described [27, 28]. In summary, ABLas detects ten types of ASEs based on splice junction reads. These include exon skipping (ES), alternative 5′ splice site (A5SS), alternative 3′ splice site (A3SS), mutually exclusive exons (MXE), mutually exclusive 5′ UTRs (5pMXE), mutually exclusive 3′ UTRs (3pMXE), cassette exon, A3SS&ES, and A5SS&ES.

To assess EIF3D-regulated ASEs, we conducted a Student’s t-test to evaluate the significance of the ratio alteration of ASEs. Events that reached significance at the P-value cutoff, corresponding to a false discovery rate cutoff of 5%, were considered EIF3D-regulated ASEs.

Functional enrichment analysis

For sorting out functional categories of Differentially Expressed Genes (DEGs), we identified Gene Ontology (GO) terms and KEGG pathways using the KOBAS 2.0 server [29]. The hypergeometric test and the Benjamini-Hochberg FDR controlling procedure were employed to determine the enrichment of each term.

Statistical analysis

Gene expression levels were represented as mean ± standard deviation and analyzed using R software (v4.1.3). The cut-off value for EIF3D expression was determined using the median method of gene expression. We employed a Student’s t-test to compare the two groups and utilized Kaplan-Meier curves with the log-rank test to assess overall survival in different groups. Furthermore, univariate and multivariate Cox regression analyses were conducted to evaluate the significance of EIF3D expression and clinicopathologic features for overall survival. Statistical significance was defined as a P-value < 0.05.

Data availability

The data utilized in this study are accessible on the TCGA website, and for experimental data inquiries, please contact the corresponding author.

EIF3D expression is upregulated in HNSC and associated with poor prognosis

In our pan-cancer analysis, significant differential expression of EIF3D was observed in breast cancer (BRCA), bile duct cancer (CHOL), colon and rectal cancer (COAD), and HNSC, with a noteworthy overexpression observed across most tumor types (P < 0.01, as depicted in Figure 1A). To delve deeper into the role of EIF3D in HNSC, we analysed its mRNA expression profile in 522 HNSC samples along with 44 normal samples from TCGA database. This analysis revealed a substantial upregulation of EIF3D in HNSC (P < 0.001, Figure 1B, 1C), and heightened EIF3D levels were linked to shorter overall survival (OS) (P = 0.006, Figure 1D) and progression-free survival (PFS) (P < 0.001, Figure 1E). Univariate Cox regression analysis (P = 0.010, HR = 1.562, 95% CI 1.114–2.189) demonstrated a significant association between high EIF3D expression and reduced OS in HNSC patients (Figure 1F). Multivariate Cox regression analysis further established EIF3D (P = 0.031, HR = 1.491, 95% CI 1.038–2.142) as an independent prognostic parameter for HNSC patients (Figure 1G). These findings underscore the upregulation of EIF3D in HNSC and its correlation with adverse prognosis in these patients.

Figure 1. Expression level and prognosis of EIF3D in HNSC. (A) Box plot illustrating the expression level of EIF3D in pan-cancer dataset. (B, C) Box plots displaying EIF3D expression levels in a cohort of 566 HNSC samples retrieved from the TCGA database, encompassing 522 tumor samples and 44 normal samples. Error bars indicate the mean ± SEM. Significance denoted by P < 0.05. (D, E) Survival curves depicting the impact of EIF3D expression on HNSC patient survival, as determined from the TCGA databases. (F, G) Univariate and multivariate Cox regression analyses. Significance levels: ^*P-value < 0.05, ^**P-value < 0.01, ^***P-value < 0.001.

EIF3D has great potential in the regulation of tumor microenvironment

Comparing patients with high EIF3D expression to those with low EIF3D expression, results revealed significantly elevated StromalScores, ImmuneScores, and ESTIMATEscores (P < 0.001, Figure 2A), indicating that EIF3D plays a pivotal role in regulating the TME. Employing the CIBERSORT algorithm, we analysed the proportions of 22 immune cell types in HNSC patients and presented the relative proportions in EIF3D high and low expression groups using box plots (Figure 2B). Among these, the EIF3D-low expression group exhibited higher proportions of plasma cells, CD8T cells, Tregs cells, and neutrophils compared to the EIF3D-high expression group. Conversely, the EIF3D-high expression group demonstrated a higher proportion of M0 macrophages, activated mast cells, and eosinophils. Correlation analysis revealed positive correlations between EIF3D and M0 macrophage infiltration, eosinophils, and activated mast cells, while negative correlations were observed with naive B cells, resting mast cells, follicular helper T cells, plasma cells, Tregs, resting dendritic cells, neutrophils, and CD8T cells (Figure 2C).

Figure 2. Correlation between the EIF3D and the tumor microenvironment and immune checkpoint genes. (A) Correlation analysis of ImmuneScore, StromalScore, and ESTIMATEScore with EIF3D. (B) Box plots illustrating variations in immune cell infiltration proportions between high-EIF3D and low-EIF3D patient groups. (C) Lollipop graph depicting the correlation between immune infiltration and EIF3D. (D) Correlation analysis between EIF3D and immune checkpoint genes. (E–H) Evaluation of immunotherapy outcomes involving anti-CTLA4 and anti-PD1 inhibitors in EIF3D high and low expression groups. Error bars denote mean ± SEM. Significance levels: ^*P-value < 0.05, ^**P-value < 0.01, ^***P-value < 0.001.

We further explored the correlation between EIF3D and immune checkpoint inhibitors using the expression levels of EIF3D and 47 immune checkpoint-associated genes. EIF3D exhibited significant associations with 13 immune checkpoint-related genes (P < 0.001). Notably, EIF3D expression displayed positive correlations with CD276, VTCN1, and CD70, while showing negative correlations with PD1, CD244, and TIGIT (Figure 2D). ImmuneScores were also calculated for anti-CTLA4 and anti-PD1 inhibitors, with higher ImmuneScores indicating enhanced treatment efficacy. Significant differences were observed in CTLA4-negative-PD1-positive, CTLA4-positive-PD1-negative, and CTLA4-positive-PD1-positive samples (Figure 2E–2H), indicating that anti-PD1, anti-CTLA4, and the combination of anti-PD1 and anti-CTLA4 therapies are more effective in patients with low EIF3D expression compared to those with high EIF3D expression.

EIF3D affects the differentially spliced IGAS event

In our analysis, we uncovered a total of 42,849 ASEs affecting 10,124 genes in HNSC patients. ES emerged as the most prevalent ASE type, followed by AT, AP, AA, AD, RI and ME (Supplementary Figure 1). Our investigation into DSAS between the EIF3D high and low expression groups unveiled 5,916 such events impacting 3,385 genes. Once again, ES constituted to be the majority of aberrantly regulated AS events (Figure 3A). Notably, in these groups, we identified 747 DEIGs through differential expression analysis, among them 105 DSAS parent genes belonged to the DEIGs category, which referred to as IGAS parent genes (Figure 3B). These findings strongly suggest that EIF3D may function as an RNA splicing regulator, influencing the prognosis of HNSC patients through its modulation of IGAS events.

Figure 3. Overview of IGAS events and parent genes in HNSC. (A) UpSet plot illustrating interactions among the seven AS types of DSAS parent genes detected. (B) Venn plot displaying the intersection between DSAS parent genes and DEIGs.

Knockdown of EIF3D inhibits cellular proliferation, migration and invasion, and promotes apoptosis in FaDu cells

To discern the biological role of EIF3D in HNSC, this study established an EIF3D knockdown cell model using FaDu cell lines. The mRNA levels of EIF3D were notably lower in the EIF3D knockdown cell samples compared to the siRNA control (P < 0.001, Figure 4A). Similarly, in the cell model with reduced EIF3D expression, protein levels of EIF3D exhibited a significant decrease (Figure 4B). These results affirm the successful attenuation of EIF3D expression in FaDu cells. Subsequent analysis revealed that following EIF3D knockdown, the proliferation of FaDu cells was significantly curtailed at 48 and 72 hours (Figure 4C, 4D). Flow cytometry demonstrated that down-regulation of the EIF3D gene substantially increased the overall apoptosis rate of FaDu cells (Figure 4E). Furthermore, in vitro EIF3D downregulation significantly hampered the migration and invasion abilities of FaDu cells (Figure 4F–4I). Collectively, these results underscore the role of EIF3D in promoting FaDu cell proliferation, migration, and invasion while inhibiting apoptosis, providing further evidence that EIF3D plays a pro-cancer role in the progression of HNSC.

Figure 4. Knockdown of EIF3D inhibits cellular proliferation and migration, and promotes apoptosis in FaDu cells. (A) Bar plot depicting the RT-qPCR results for control and treatment samples (NC represents the control sample, and Si represents the experimental sample). (B) Western blot results. (C, D) Proliferation results of FaDu cells after EIF3D knockdown. (E) Apoptosis results of FaDu cells after EIF3D knockdown. (F, G) Cell migration results of FaDu cells after EIF3D knockdown. (H, I) Cell invasion results of FaDu cells after EIF3D knockdown. ^*P-value < 0.05, ^**P-value < 0.01, ^***P-value < 0.001.

EIF3D regulates immune-related gene expression in FaDu cells

Moreover, we explored the impact of down-regulating EIF3D on gene expression using RNA-seq (Figure 5A). The scatterplot displayed distinctive clustering patterns of the siEIF3D group and the normal control (NC) group after Principal Component Analysis (PCA) analysis (Figure 5B). Leveraging RNA-seq data, we uncovered potential gene targets influenced by EIF3D at the transcriptional level. Employing the R package “DESeq2” to screen for Differentially Expressed Genes (DEGs) between the two groups with the criteria (|log₂FC| > 3/2 or ≤ 2/3 and P < 0.01), our analysis revealed that silencing the EIF3D gene in FaDu cells led to differential expression in 531 genes, comprising 182 up-regulated DEGs and 349 down-regulated DEGs (Figure 5C). Furthermore, the heatmap depicted the expression patterns of DEGs in the siEIF3D and NC groups (Figure 5D). Notably, over 65% of the genes among the 531 DEGs were down-regulated.

Figure 5. EIF3D regulates gene expression in FaDu cells. (A) Boxplot illustrating the expression levels of EIF3D in RNA-seq data. (B) PCA based on FPKM values of all detected genes. The ellipse for each group represents the confidence ellipse. (C) Volcano plot displaying all differentially expressed genes (DEGs) between treatment and control samples with DEseq2. P-value < 0.01 and FC (fold change) ≥ 1.5 or ≤ 0.67. (D) Hierarchical clustering heatmap depicting expression levels of all DEGs. (E) Bubble diagram presenting the most enriched GO biological process results of the up-regulated and down-regulated DEGs. (F) Boxplot showing the expression pattern and statistical differences of immune-related DEGs. Error bars represent mean ± SEM. ^*P-value < 0.05, ^**P-value < 0.01, ^***P-value < 0.001.

To gain insights into the potential biological roles of EIF3D, we conducted GO and KEGG enrichment analyses based on the 531 DEGs. Among the upregulated 182 DEGs, notable enrichments included processes such as positive regulation of cell population proliferation, activation of MAPK activity, regulation of the cell cycle, oxidation-reduction processes, apoptotic pathways, Wnt signaling, heart development, spermatogenesis, extracellular matrix organization, and positive regulation of gene expression. Conversely, the significantly down-regulated 349 DEGs were predominantly associated with multicellular organism development, epidermal development, angiogenesis, drug response, keratinocyte differentiation, signal transduction, positive regulation of apoptosis, ion transmembrane transport, cell differentiation, and regulation of catalytic activity, among other biological processes (Figure 5E). Moreover, we identified 52 immune-related genes among the 531 DEGs and selected 12 of these for further analysis. Within this subset, the expression of TNFSF10, NEDD4, HSPA2, MMP12, CDKN1A, and NGFR decreased in siEIF3D-treated FaDu cells, while FABP5, ISG20L2, OGFR, AEN, CCN2, and GSEC showed increased expression (Figure 5F). Remarkably, this study also showed a significant down-regulation of the alternative splicing factor NOVA2 among the DEGs. Taken together, these results imply that EIF3D may influence HNSC cell proliferation, apoptosis, angiogenesis, and cell cycle regulation by modulating the expression levels of immune-related genes and AS factors.

EIF3D regulates IGAS event in FaDu cells

In recent years, the close association between AS and the TME has garnered attention [21, 22]. To investigate the mechanistic role of EIF3D in HNSC, we conducted a comprehensive analysis of EIF3D-regulated ASEs. We detected a total of 114,901 ASEs, encompassing 53,373 known ASEs and 61,528 new ASEs. These AS events could be categorized into nine main types, including MXE, ES, Cassette Exon, A5SS&ES, A3SS&ES, A3SS, 5pMXE, 3pMXE (Figure 6A).

Figure 6. EIF3D influences the alternative splicing of genes in FaDu cells. (A) Bar plot displaying the number of all significantly regulated alternative splicing events (RASEs). The X-axis represents the RASE number, while the Y-axis represents the different types of AS events. "Up" indicates that the splicing pattern occurs in a higher proportion in the siEIF3D group than in the NC group, while "down" indicates the opposite. (B) Bubble diagram presenting the most enriched GO biological process results of RASGs. (C) Venn diagram illustrating the overlap in gene numbers between RASGs and DEGs. (D) Chord diagram demonstrating the co-expression of EIF3D with immune-associated RASGs. (E) Schematic diagrams depicting the structures of ASEs. RNA-seq validation of ASEs is displayed at the bottom of the right panel. Error bars represent mean ± SEM. ^*P-value < 0.05, ^**P-value < 0.01, ^***P-value < 0.001.

For the analysis of EIF3D-regulated ASEs, we employed a t-test to compare the changes in AS levels of genes with the same splice phenotype between two different samples, applying a significant difference ASE screening criterion of p-value ≤ 0.05. The results unveiled 998 DSAS events that were more prevalent in the siEIF3D group compared to the NC group, with higher proportions of 3pMXE, 5pMXE, A3SS, A3SS&ES, A5SS&ES, ES, and MXE events in the siEIF3D group.

Subsequently, we conducted GO enrichment analysis of EIF3D-regulated Differential Alternative Splicing Genes (RASG). The outcomes indicated that EIF3D-regulated ASEs are linked to various biological processes, including translation, viral processes, DNA repair, cellular response to DNA damage stimulus, chromatin organization, translation initiation, vesicle-mediated transport, translation regulation, DNA recombination, and regulation of circadian rhythm (Figure 6B).

Moreover, overlapping analysis of DEGs and RASGs identified 28 genes that were both DEGs and RASGs (Figure 6C). Additionally, we detected 129 IGAS events out of 80 immune-related genes in FaDu cells. Figure 6D illustrates that EIF3D exhibits significant co-expression with IGAS parent genes such as S100A3, RFXANK, PPIA, IFNGR2, ILF3, MDK, and RPS19, where red indicates positive correlation and green indicates negative correlation (|correlation coefficient| > 0.8, P < 0.05).

Within the list of ASEs, we selected events with significantly different ratios between treatment and control groups, leading to the identification of four IGAS events: EZH2, STAT3, TRPM2, and MFF. To validate EIF3D-regulated IGAS, we further investigated these four IGAS events using RNA-Seq techniques. EIF3D exhibited a downregulating effect on A3SS events, which potentially led to altered EZH2 expression. Additionally, EIF3D influenced the A5SS events of STAT3 (P < 0.05, Figure 6E). Furthermore, EIF3D modulated the 5pMXE events of MFF and the ES events of TRPM2 (P < 0.05, Supplementary Figures 2 and 3). These findings, consistent with the results of bioinformatics analysis, underscore EIF3D’s significant potential in regulating IGAS in the context of HNSC.