Apoptosis-inducing factor mitochondria-associated 2 (AIFM2) promotes tumor progression and predicts poor prognosis in oral squamous cell carcinoma

Abstract

Background/purpose: Head and neck squamous cell carcinoma (HNSCC), including oral squamous cell carcinoma (OSCC), remains a major malignancy with limited therapeutic efficacy. Apoptosis-inducing factor mitochondria-associated 2 (AIFM2), also known as ferroptosis suppressor protein 1, regulates ferroptosis and tumor progression. This study investigated the oncogenic function, clinical relevance, and regulation of AIFM2 in OSCC.

Materials and methods: Transcriptomic data from TCGA HNSCC and in-house OSCC RNA-Seq datasets were analyzed to assess AIFM2 expression and its association with clinicopathological features and outcomes. Functional assays evaluated the effects of AIFM2 knockdown or overexpression on OSCC cell proliferation, migration, invasion, and therapeutic response. MicroRNAs targeting AIFM2 were identified through bioinformatics, luciferase reporter, and mimic assays. A Light Gradient Boosting Machine (LGBM) model was used for prognostic prediction.

Results: AIFM2 overexpression was associated with advanced stage, poor tumor differentiation, and unfavorable survival in HNSCC/OSCC. AIFM2 knockdown suppressed, whereas its overexpression enhanced, OSCC cell proliferation, migration, and invasion, while exerting minimal effects on cisplatin, palbociclib, or cold atmospheric plasma sensitivity. miR-32-5p and miR-432-5p directly targeted AIFM2 and were downregulated in tumors. AIFM2-associated transcripts were enriched in pathways related to oxidative stress, lipid metabolism, and E2F targets. The LGBM-derived AIFM2 gene signature demonstrated strong prognostic predictive power.

Conclusion: AIFM2 acts as an oncogenic driver in OSCC, regulated by tumor-suppressive miR-32-5p and miR-432-5p, and serves as a potential prognostic biomarker and therapeutic target.

Keywords: AIFM2; Head and neck carcinoma; Oral carcinoma; Prognosis.

Figure 1

The association between AIFM2 expression and clinicopathological states in HNSCC/OSCC. (A, C) HNSCC. (B, D) OSCC. (A, B) Clinicopathological variants as related to AIFM2 expression. (C, D) Kaplan–Meier survival curve. T, tumor size; N, nodal metastasis; PNI, perineural invasion; LVI, lymphovascular invasion; ENE, extranodal extension; TPM, transcripts per million. ∗, ∗∗, and ∗∗∗, P < 0.05, P < 0.01, and P < 0.001, respectively.

Figure 2

The AIFM2 expression in cell lines. (A) NOK and HNSCC/OSCC cell lines. Lt, qPCR analysis reveals the higher AIFM2 expression in cancer cell lines relative to NOK, except for OECM1. Rt, Western blot analysis reveals the AIFM2 expression in SAS and FaDu cells. NOK and OECM1 exhibits scanty and absent AIFM2 expression. (B) AIFM2 knockdown. Cells are treated with si-AIFM2 oligonucleotide or si-RNA-A oligonucleotide. Lt, qPCR analysis. Treatment with 60 or 100 nM si-AIFM2 for 24 h or 48 h decreases AIFM2 mRNA expression in SAS cells. Rt, Western blot analysis. Treatment with 100 nM for 24 h si-AIFM2 decreases AIFM2 protein expression in SAS and FaDu cells. (C, D) Wound healing assay and invasion assay, respectively. Knockdown of AIFM2 decreases the competence of wound healing (in C) and invasion (in D) of SAS and FaDu cells. (E) Western blot analysis of the SAS cell. It reveals AKT, ERK, and p38 activation following plasmid transfection for 24 h. The slight upregulation of AIFM2 and COX2 follows the ERK and p38 activation, respectively. Values below the Western blot diagram denote normalized expression levels. oe, transient overexpression. ∗∗, and ∗∗∗, P < 0.01, and P < 0.001, respectively.

Figure 3

Transient AIFM2 expression in cell lines. (A, B) SAS and OECM1 cells, respectively. Lt, proliferation; middle, wound healing; Rt, invasion. Transient AIFM2 expression increases cell proliferation, wound healing, and invasion. (C, D) FaDu cells. (C) Upper, proliferation. Lower, wound healing. Transient AIFM2 expression does not affect such phenotypes in FaDu cells. (D) Invasion assay. Upper, quantification of invaded cells; Lower, representative fields of invaded cells on the transwell membrane. x200. oe, transient overexpression; va, vector alone. ns, not significant. ∗∗, and ∗∗∗, P < 0.01, and P < 0.001, respectively.

Figure 4

Stable AIFM2 overexpression in cell lines. (A, B) Western blot analysis of SAS and OECM1 cells. (A) GFP OE cell subclones. The Doxy treatments at the doses more than 1000 ng/mL for 48 h drastically induce GFP expression. (B) AIFM2 OE cell subclones. The Doxy treatments at the doses more than 500 ng/mL for 48 h drastically induce AIFM2 expression. (C–G) Lt, SAS cell subclones; Rt, OECM1 cell subclones. (C, D) Proliferation assays. Doxy+, 2000 ng/mL doxycycline treatment for 48 h; Doxy -, no treatment. (C) Comparison between Doxy treatment and control in OE cell subclones. (D) Comparison between OE cell subclones and VA cell subclones in the presence of Doxy treatment. (E–G) Dose response curves of CDDP, CAP, and Palbociclib, respectively. par, parental; OE, stable overexpression; VA, vector alone; CDDP, cisplatin; CAP, cold atmospheric plasma. ∗∗∗, P < 0.001.

Figure 5

The targeting of suppressor miRNAs on AIFM2. (A) A Venn diagram illustrates the retrieval of potential miRNA repeatedly predicted by 4 or 5 multiple algorithms. (B) Schematic diagram of AIFM2 3′UTR (1905-bp) and the two reporters being constructed to prove the potential targeting of miR-32-5p, miR-150-5p, and miR-432-5p on the AIFM2 3′UTR region. The color boxes and arrows indicate the predicted targeting miRNAs. H and T, the head part and tail part of 3′UTR, respectively. @, the number of the first nucleotide in the binding sites. (C) Reporter assays. Following the treatment of 60 nM miRNA mimics for 24 h, reporter assays are performed to disclose the repression of reporter activity. H, AIFM2 3′UTR-H reporter; T, AIFM2 3′UTR-T reporter; VAR, vector alone reporter. Scr, scramble mimic. Note that miR-150-5p expression represses both H and T reporters, while miR-32-5p and miR-432-5p expression only represses the T reporter. (D) Western blot analysis. The treatment with 100 nM mimics for 24 h represses AIFM2 protein expression. The repression of miR-150-5p mimic lasts to 48 h. Values below the Western blot diagram denote normalized expression levels. (E) miR-32-5p (upper) and miR-432-5p (lower) expression are downregulated in the TCGA HNSCC tumor cohort. RPM, reads per million; Arrows, downregulation. ns, not significant. ∗, ∗∗, and ∗∗∗, P < 0.05, P < 0.01, and P < 0.001, respectively. (For interpretation of the references to color in this figure legend, the reader is referred to the Web version of this article.)

Figure 6

Prediction of patient survival based on the AIFM2-associated gene signature. (A) Schematic illustration of the analytical workflow used for survival prediction. (B) Volcano plot showing significantly dysregulated transcripts correlated with AIFM2 expression in TCGA HNSCC tumors. Red dots indicate upregulated genes, blue dots indicate downregulated genes, and grey dots represent genes without significant change. Representative genes with prominent differential expression are labeled. (C) Bubble plot depicting the major functional pathways enriched among AIFM2-associated transcripts in accordance with Normalized Enrichment Score (NES), False Discovery Rate (FDR)-q value and size of hallmarks. Detailed enrichment results are provided in Supplementary Table S7 and Fig. S1. (D, E) Survival analysis of the TCGA HNSCC training cohort (D; n = 156) and validation cohort (E; n = 104) stratified by risk groups predicted using the Light Gradient Boosting Machine (LGBM) model. Upper left: Kaplan–Meier survival curve; upper right: receiver operating characteristic (ROC) curve; lower left: distribution of risk scores; lower right: event status plot. (F) Survival analysis of the OSCC validation cohort (n = 100). Upper panel: Kaplan–Meier survival curve; lower panel: ROC curve. HR, hazard ratio. (For interpretation of the references to color in this figure legend, the reader is referred to the Web version of this article.)