Mitochondrial DNA mutations in head and neck squamous cell carcinoma: a systematic review and meta-analysis.

Introduction
Squamous cell carcinoma (SCC) comprises 95% of all malignant lesions found in the head and neck. SCC development is a multistep process modulated by genetic predisposition, chronic inflammation, tobacco and alcohol abuse, and viral infections [1].
Head and neck squamous cell carcinoma (HNSCC) is the sixth most common cancer worldwide, with 890,000 new cases and 450,000 deaths in 2018. The incidence of HNSCC continues to increase and is anticipated to rise by 30% (that is, 1.08 million new cases annually) by 2030 [2], which makes it a major world health problem [3]. Tumor volume, grade, and TNM stage are the most important factors predicting the outcome of SCC. However, these clinicopathologic factors are not sufficient explanations for the biological behavior or therapeutic response of HNSCC [4].
Despite significant improvements in therapeutic modalities, the 5-year post-therapeutic survival rate remains among the lowest for major cancers. Locoregional relapse is the major cause of death in patients with SCC [5, 6]. Consequently, there is an urgent need for reliable biomarkers for early detection, monitoring, and margin evaluation [5]. A deeper understanding of the molecular characteristics of SCC may ultimately facilitate the development of more suitable treatments [4]. The identification of prognostic and predictive biomarkers is a crucial approach to enhance patient management [3].
Mitochondrial DNA (mtDNA) is a small, circular, double-stranded genome comprising 37 genes, including 13 protein-coding subunits of the electron transport chain and the process of oxidative phosphorylation (OXPHOS). The remaining 24 genes are responsible for producing the ribosomal RNAs (rRNAs) and transfer RNAs (tRNAs) necessary for the translation of these proteins within the mitochondria [7]. However, unlike nuclear DNA, mtDNA lacks protective histones and robust DNA repair systems, rendering it approximately ten times more susceptible to oxidative stress. Reactive oxygen species (ROS) play a vital role in cancer pathogenesis [8]. This vulnerability is critically relevant in HNSCC where established risk factors for HNSCC, such as tobacco use, generate reactive oxygen species (ROS). Since mitochondria are themselves a primary source of ROS, this can initiate a vicious cycle of mtDNA damage and genomic instability, thereby driving HNSCC pathogenesis [9]. This cycle of damage and dysfunction promotes physio-dynamic alterations in tumor cells (manifesting as the Warburg effect), evasion of apoptosis, and ultimately uncontrolled cellular growth [8].
Although mutations occur throughout the mitochondrial genome, the most frequently detected place is in the displacement-loop (D-loop) region; around a third of HNSCC samples harbor D-loop mutations [3]. The D-loop region is important because it is the major control site for mtDNA expression and replication [10].
Somatic mtDNA mutations have been reported mostly in primary SCC patients [5]. Moreover, recent studies have further illustrated that certain mtDNA single-nucleotide polymorphisms (SNPs) may correlate with the prognosis of certain cancers, such as esophageal squamous cell carcinoma [4]. On the other hand, no significant correlation between mtDNA mutations and content alteration was found in esophageal cancer [11]. To date, no study has extensively analyzed the regional distribution of mtDNA mutational burden in HNSCC. Due to conflicting reports and uncertainty in the literature, the current study aimed to quantify the relative contributions of several mitochondrial genome regions to the overall mutational burden in HNSCC to contextualize their potential biological importance.

Materials and methods

Study registration
This systematic review and meta-analysis follow the guidelines established by Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) [12]. The review protocol has been registered with PROSPERO under the registration number CRD42025641447.

Search strategy
Two independent reviewers conducted a comprehensive search of four electronic databases in May 2025. The databases included PubMed, EMBASE, Scopus, and Web of Science. No filters were applied during the search process. The search terms concentrated on Head and Neck Squamous Cell Carcinoma, Mitochondrial DNA, and Mutation. The search query is shown in Table 1.

Study selection
The results obtained were subsequently imported into EndNote 20 to remove duplicates and conduct a screening process. Studies were included according to the following PICO/PECO framework:Population (P): Patients diagnosed with HNSCC, including subsites such as oral squamous cell carcinoma (OSCC) and laryngeal squamous cell carcinoma (LSCC).

Intervention/Exposure (I): Identifying and reporting somatic mtDNA mutations employing molecular methodologies, such as DNA extraction, PCR amplification, sequencing, or variant detection techniques.

Comparator (C): Not limited; studies lacking explicit comparators were included if they reported mtDNA mutations. Comparisons may encompass tumor tissue to nearby normal tissue, blood samples, or healthy controls when applicable.

Outcome (O): Information regarding the frequency, distribution, or burden of mtDNA mutations throughout various parts of the mitochondrial genome (e.g., D-loop, ND, COX, CYTB, rRNA, tRNA).

The subsequent criteria for exclusion from this study are as follows:Non-English studies

Animals and laboratory studies

Reviews, systematic reviews, meta-analyses, case reports, editorials, conference abstracts, or grey literature

Studies that do not explicitly document somatic mtDNA mutations in HNSCC patients.

Studies assessing mtDNA mutations in HNSCC patients lack sufficient data on sufficient methodological detail or mutation data for extraction.

Screening process
Initially, two authors (FF and FJ) independently screened the articles based on titles and abstracts. In cases of disagreement, a final decision was reached after discussion with another author (ZK). The reviewers were not blinded to the authors' names, institutions, or journal titles during their evaluation. Afterwards, the potentially relevant articles were assessed independently at the full-text level (FF, FJ, and ZK). Finally, we searched the databases again to identify any potentially missing articles, and one author (MM) performed a final review of the included articles.

Data extraction
Two reviewers (FJ and MN) independently collected all relevant data, and any discrepancies in their findings were resolved through discussion. If necessary, a third person (MM) was involved to reach a consensus. No raw or individual participant data were used. The meta-analysis utilized aggregated data obtained from previously published studies. The data extracted included: Title, authors, year, country, type of study, objective, inclusion criteria, exclusion criteria, sample size, location of DNA extraction for both tumor and normal samples, control group, DNA extraction method, amplification, mutation detection methodology and tools, total mutations identified with their type, location, frequency based on location and status, p value, somatic mutation prevalence, statistical analysis, key findings, limitations, subgroup analysis, effect modifiers, and outcomes.

Quality assessment
The risk bias in the included studies was evaluated using the appropriate Joanna Briggs Institute (JBI) critical assessment instruments tailored for each study type [13]. These tools assess many bias domains, including selection bias, performance bias, detection bias, and reporting bias. Each study was graded by two authors (MN, FJ). Disagreement between reviewers were resolved by consensus with the involvement of a third reviewer (MM) if necessary.

Data analysis
All analyses were conducted in R (version 4.5.1; R Foundation for Statistical Computing, Vienna, Austria) using the meta package.
A mutation-burden meta-analysis was conducted to characterize the distribution of somatic mtDNA mutations across six regions: D-loop, ND, COX, CYTB, tRNA, and rRNA. The focus of analysis was on mutation events rather than on patients. In each study, the numerator represented the number of mutations within a specific region, while the denominator indicated the total number of mutations identified across all areas within that study. The result was the pooled proportion of mutation events per region, representing the relative contribution to the overall mutational burden.
Random-effects meta-analyses were conducted using R (version 4.5.1; meta package). Heterogeneity across studies was evaluated using Q, I2, τ2, and 95% prediction intervals. Sensitivity analyses encompassed leave-one-out diagnostics and various pooling methods. Funnel plots and Peters’ test were utilized when a minimum of 10 studies were present, noting that these methods are less established for proportional event-share data.

Results

Study selection
A total of 92 articles were identified by a comprehensive search across PubMed, EMBASE, Scopus, and Web of Science. 44 distinct records were kept for screening after 48 duplicates were removed. After removing 18 records through title and abstract screening, 26 full-text publications remained for in-depth analysis. Finally, 17 articles were included in the qualitative and quantitative synthesis after meeting the inclusion criteria (Fig. 1).

Quality assessment
The Joanna Briggs Institute (JBI) critical appraisal tools were used to assess risk of bias, and the results showed that the overall quality of the evidence was moderate. The majority of studies provided sufficient descriptions of mutation detection, DNA extraction, and patient selection. However, typical limitations were the possibility of misclassifying population variants, the absence of blinded mutation calling, and the insufficient reporting of sequencing coverage.
Overall, while methodological quality was adequate, heterogeneity in detection techniques and reporting likely contributes to the observed variation in mutation prevalence across genomic regions.

Pooled mutation burden by genomic region
The meta-analysis assessed the proportional impact of six mitochondrial genomic areas on the mutational burden in HNSCC (Table 2).
Mutations were distributed unevenly across all studies, with the D-loop and ND genes identified as predominant hotspots. Moderate contributions were observed from COX, rRNA, and tRNA genes, while CYTB exhibited a smaller yet consistent pattern.

D-loop mutations
Several studies have consistently shown that the mitochondrial D-loop is a mutational hotspot in HNSCC. Zhou et al. [26] detected D-loop mutations in 29% of cancer cases, with mutation frequency rising alongside histological severity. Tan et al. [24] indicated that 76.9% of somatic mutations were located within the D-loop. In contrast, Lai et al. [4] noted that 55% of all mutations were concentrated in this region, exhibiting a frequency almost 15 times greater than that in coding sections. Mondal et al. [20] identified mutations primarily within nucleotide positions 51–595 and linked their occurrence with a significantly elevated risk of OSCC. Additional studies corroborated the biological significance of this region, observing elevated mtDNA copy numbers in samples containing D-loop mutations [11], correlations with hypopharyngeal cancers and tobacco consumption [17], and recurrent hotspots in hypervariable regions HVR1 and HVR2 (Mondal et al. [19]). Schubert et al. [23] emphasized the D-loop as the most mutation-rich area of the mitochondrial genome, indicating its role in early carcinogenic processes, but Palodhi et al. [22] reported a significant level of heteroplasmy, with both recurrent and unique variants found in more than half of the patients.
According to Fig. 2, the pooled share of reported mutations in the D-loop across 12 studies was 0.67 (95% CI: 0.28–0.91). There was significant heterogeneity among studies (I2 = 93.2%, p < 0.0001) (Table 3), with individual research estimates varying from 3 to 100%. The funnel plot (Fig. 3) exhibited visual asymmetry, indicative of possible small-study effects. The data collectively indicate that the D-loop serves as the primary site for somatic mtDNA mutations in HNSCC. At the same time, the significant variability underscores both methodological discrepancies and possible biological subtypes.

ND mutations
Mutations in ND genes, which encode Complex I subunits of the mitochondrial genome (ND1–ND6, ND4L), are among the most frequently documented mtDNA changes in HNSCC and have been associated with mitochondrial dysfunction and tumor growth. Dasgupta et al. [3] identified five ND mutations, including a germline nonsense variation in ND4, which is anticipated to terminate the protein and compromise Complex I integrity. Mondal et al. [19] discovered that 47% of somatic mutations in OSCC pertained to ND genes, including a nonsynonymous ND1 substitution (Cys → Tyr), and observed that all changes were transitions, indicating a mutational bias. Evidence for early involvement was presented by Mondal et al. [19], who detected ND4L mutations in both dysplastic margins and tumor tissues, supporting persistence from precancerous stages. Zhou et al. [26] additionally documented nonsynonymous changes in ND2 and ND5, with several cancers including over 20 ND mutations.
Further research confirms these results. Uzawa et al. [25] discovered homoplasmic mutations in ND2 and ND3 within OSCC-derived cell lines, correlating with increased circulating mutant mtDNA and unfavorable prognosis. Mutlu et al. [21] detected ND4 deletions in patient tumors, with structural modeling suggesting functional implications. Tan et al. [24] documented recurrent ND2 missense mutations, demonstrating a transition from heteroplasmy to homoplasmy, which is indicative of clonal selection and associations with apoptotic dysregulation. Schubert et al. [23] highlighted a substantial burden of mutations in ND4, ND6, ND1, ND3, and ND4L, particularly when normalized for gene length; meanwhile, Palodhi et al. [22] identified 12 ND5 mutations among oral cancer patients, reinforcing the recurrent involvement of ND genes.
In six included studies assessing ND-region mutations (n = 522 mutation events), the pooled mutations share in ND genes was 29% (95% CI: 0.20–0.40) (Fig. 4). The 95% prediction interval was broad (0.09–0.63), suggesting significant variability in mutation proportions anticipated in future studies. Substantial heterogeneity was observed between studies (I2 = 69.3%, τ2 = 0.241, p = 0.006) (Table 3), indicating variability in study populations and methodologies. The examination of the funnel plot (Figure S1) indicated mild asymmetry, as one smaller study reported a higher mutation burden relative to the others. Overall, these findings suggest that ND mutations constitute a moderate yet variable aspect of the mitochondrial mutational burden in HNSCC. Their regular presence in ND1, ND2, ND4, ND5, and ND4L, along with indications of functional impairment in oxidative phosphorylation, prevention of apoptosis, and resistance to therapy, highlights their potential as therapeutic targets. In comparison to the D-loop, ND mutations represent a smaller but biologically significant aspect of the mtDNA mutational spectrum, indicating a balance between instability in regulatory regions and disruption in functional coding regions.

COX mutations
Mutations in the COX genes, which encode subunits of Complex IV (COX1–COX3), have been reported in various studies of HNSCC and OSCC. The alterations are primarily nonsynonymous or frameshift, indicating potential functional implications for oxidative phosphorylation. Dasgupta et al. [3] identified 12 mutations in COX genes, comprising eight in COX1, three in COX2, and one in COX3, from a total of 20 coding-region mutations, of which 13 were classified as nonsynonymous. Complex IV demonstrated the highest mutational load relative to other mitochondrial complexes (p < 0.001), indicating selective pressure on tumor cells. Mondal et al. [19] reported a single somatic mutation (5.8%) in Complex IV among 17 coding-region mutations, suggesting a more sporadic role in OSCC.
Additional studies highlighted the repeated engagement of COX1 and COX3. Palodhi et al. [22] identified 12 COX1 mutations in 10 patients, frequently associated with alterations in ND5 and CYTB, highlighting their role in the mitochondrial mutational landscape of oral cancer. Tan et al. [24] identified a frameshift deletion in COX3 that results in a truncated protein, likely impairing its function and correlating with the dysregulation of apoptotic genes. These findings indicate that mutations in the COX gene may hinder Complex IV activity, thereby influencing apoptosis and tumor progression in HNSCC.
Seven studies (n = 686 mutation events) revealed the prevalence of mtDNA mutations in the COX region. The proportion of patients exhibiting at least one COX mutation varied from 0.04 (95% CI: 0.00–0.20) in Mondal et al. [19] and Tan et al. [24] to 0.47 (95% CI: 0.32–0.62) in Mutlu et al. [21].
The pooled mutation share in COX genes was 0.12 (95% CI: 0.06–0.25) (Figure S2), indicating that approximately 12% of the total mutational burden originates from this region. There was significant between-study heterogeneity (I2 = 88.1%, τ2 = 1.0622, χ2 = 50.33, p < 0.0001) (Table 3). The 95% prediction interval (0.01–0.69) indicates that the true prevalence in future studies may vary from 1 to 69%. The funnel plot exhibited a mild asymmetry, indicating potential small-study effects or publication bias. The funnel plot exhibited a mild asymmetry, indicating potential small-study effects or publication bias (Figure S3). These findings indicate that the pooled prevalence of COX mutations is relatively modest, yet significant variability exists across studies, suggesting potential methodological and biological differences. Collectively, these reports indicate that COX mutations, although less prevalent than D-loop or ND mutations, could have a more significant functional impact related to their direct involvement in oxidative phosphorylation and the regulation of apoptosis.

CYTB mutations
The CYTB gene, the only mitochondrially encoded subunit of Complex III, has been frequently reported in HNSCC and OSCC. These alterations are often nonsynonymous, indicating potential functional implications for mitochondrial respiration. Dasgupta et al. [3] identified three CYTB mutations among 20 total coding-region alterations in HNSCC, with most coding-region variants classified as nonsynonymous, suggesting potential deleterious effects on protein function. Mondal et al. [19] identified three mutations in Complex III, including CYTB, which accounted for 17.6% of the somatic coding-region mutations detected. All mutations were transition substitutions (T • C and C • T), aligning with established tumor-associated mutational signatures. Palodhi et al. [22] identified 12 CYTB mutations among 12 patients, positioning it as one of the most frequently mutated mitochondrial genes in their dataset, alongside COX1 and ND5, thereby demonstrating its potential significance in tumor development.
Five studies (n = 620 mutation events) were included in the meta-analysis of the CYTB region. The pooled mutation share was 0.08 (95% CI: 0.06–0.10) (Figure S4) as determined by a random-effects GLMM, accompanied by a narrow 95% prediction interval (0.05–0.11). Between-study heterogeneity was minimal (I2 = 0.0%, τ2 = 0, p = 0.95) (Table 3), indicating consistency among the studies. The forest plot indicated that individual study estimates were largely consistent and closely grouped around the pooled effect size. The funnel plot (Figure S5) showed no significant asymmetry; however, the reliability of formal small-study effect testing was compromised due to the inclusion of fewer than 10 studies. The findings indicate that CYTB mutations represent a modest yet consistent portion of the mitochondrial mutational burden in HNSCC. Overall, CYTB mutations constitute a minor yet consistent aspect of the mtDNA mutational burden in HNSCC, with increasing evidence indicating their functional significance via the disruption of Complex III activity. Compared to the D-loop and ND regions, CYTB mutations are less frequent yet consistently observed across studies, suggesting a conserved function in mitochondrial dysfunction and potentially in maintaining tumor metabolic adaptation.

rRNA Mutations
Mutations in mitochondrial rRNA genes, specifically RNR1 and RNR2, have been frequently documented in HNSCC. Dasgupta et al. [3] identified seven rRNA mutations, comprising four in RNR1 and three in RNR2, indicating alterations in both the small and large subunits of the ribosome. Challen et al. [5] identified two RNR1 mutations, suggesting that rRNA mutations may arise early or independently of wider mutational patterns. Fendt et al. [14] identified heteroplasmy at position 709 in rRNA genes. In contrast, Kloss-Brandstätter et al. [1] reported that 18.3% of mitochondrial mutations in their dataset were located in rRNA genes, including the haplogroup-associated A663G variant, indicating a possible overlap between somatic and population-specific variants. Lai et al. [15] found that 13.8% of somatic mtDNA mutations in OSCC samples were located in rRNA genes, highlighting their significance. Mondal et al. [19] identified alterations in rRNA in OSCC, including mutations in RNR1 and RNR2. Schubert et al. [23] and Palodhi et al. [22] have validated the importance of this region, recording 47 and 29 rRNA mutations, respectively. Palodhi et al. [22] also noted a heteroplasmy rate of 40.7%, suggesting the possibility of clonal selection in tumor progression.
A total of 1,270 mutation events from seven studies were included in the meta-analysis of the mutation burden in the rRNA region. The pooled mutation share was 13% (95% CI: 0.08–0.21) (Figure S6). Between-study heterogeneity was very high (I2 = 91%, τ2 = 0.4, Q = 66.93, p < 0.0001) (Table 3), and the 95% prediction interval (0.03–0.46) indicated wide variability across studies. The funnel plot (Figure S7) showed some asymmetry, suggesting possible small-study effects or publication bias. Overall, rRNA mutations contribute a modest but highly variable fraction of the mitochondrial mutational burden in HNSCC.
The findings indicate that rRNA mutations contribute a modest portion of the overall mitochondrial mutational burden. Their frequent detection and variability suggest a potential interplay between somatic instability and inherited haplogroup background.

tRNA Mutations
Somatic mutations in mitochondrial tRNA genes are frequently observed in HNSCC and OSCC, potentially causing significant clinical and functional implications. Numerous studies demonstrate that these mutations occur repeatedly, are non-random, and may be associated with tumor biology and environmental factors. Dasgupta et al. [3] reported that 53% (9 of 17) of non-coding mtDNA mutations were found in tRNA genes, indicating their significant prevalence. Fendt et al. [14] noted that tRNA regions accounted for 0.99% of heteroplasmic sites, suggesting a degree of genomic stability and infrequent mutation events, similar in frequency to rRNA and coding genes. Kloss-Brandstätter et al. [1] reported that 6.9% of all somatic mutations are located in tRNA genes, highlighting their role in the mitochondrial mutational landscape. Lai et al. [15] identified 20 tRNA mutations, representing 8.3% of the total mutations, which showed a significant association with tumor differentiation (p = 0.02) and possible associations with alcohol consumption (p = 0.05) and cigarette smoking (p = 0.07). Mondal et al. [19] reported a tRNA-Glu mutation that accounts for 5.8% of the mutations observed. Schubert et al. [23] identified 12 distinct tRNA mutations, whereas Palodhi et al. [22] validated tRNA variants among 46 non-coding somatic mutations, additionally observing heightened heteroplasmy within tRNA regions.
Five studies, encompassing 905 identified mutations, have been included in the mutation-burden meta-analysis of the tRNA region. The pooled mutation share was 9% (95% CI: 0.03–0.22) (Figure S8), accompanied by a broad 95% prediction interval (0–0.76), indicating variability across studies. Significant heterogeneity was observed (I2 = 95.3%, τ2 = 1.27, p < 0.0001) (Table 3), reflecting considerable variability in mutation proportions among the studies. The funnel plot (Figure S9) indicated a nearly symmetrical distribution.. Altogether, tRNA mutations constitute a small but biologically significant segment of the mtDNA mutational spectrum, exhibiting considerable heterogeneity and possible associations with tumor differentiation and metabolic stress.

Comparative summary
When all regions are analyzed collectively, the D-loop and ND genes represent approximately 70–80% of documented mtDNA mutations, whereas COX, rRNA, tRNA, and CYTB together account for around 20–30%. This skewed distribution highlights the significant impact of regulatory and Complex I-associated genes on the mitochondrial genomics of HNSCC. The considerable heterogeneity observed across regions (I2 ranging from 0% to > 90%) indicates the presence of both technical and biological variability sources, such as differences in tumor subsites, sequencing depth, and ethnic background. The data suggest a hierarchical pattern of mtDNA involvement, characterized by an initial accumulation of D-loop instability, achieved by functionally significant ND and COX mutations that may contribute to or maintain mitochondrial dysfunction and carcinogenic progression.

Discussion
This systematic review and meta-analysis provides a comprehensive quantitative overview of the distribution of somatic mitochondrial DNA (mtDNA) mutations in Head and Neck Squamous Cell Carcinoma (HNSCC), revealing a distinct pattern of genomic vulnerability across the mitochondrial genome. By pooling data from 17 studies, we present an overview of the regions that most significantly contribute to the mutational burden, regardless of patient-specific prevalence.
The pooled data revealed a highly skewed mutational landscape, sharply polarized between the non-coding D-loop (67%, 95% CI: 0.28–0.91) and the functional, coding regions of the mitochondrial genome. Specifically, the D-loop's dominance, coupled with the substantial contribution from the ND genes (29%, 95% CI: 0.20–0.40), establishes a critical dichotomy: high-frequency genomic instability markers versus lower-frequency, yet functionally selected, drivers of carcinogenesis (COX, rRNA, tRNA, and CYTB contributed 8 − 13% each). This foundational distinction informed our hypothesis of a tiered clinical utility.
Several prior studies have recognized the non-coding mitochondrial D-loop as a mutational hotspot in bladder, lung, and head and neck tumors, indicating its possible role in carcinogenesis [27]. This area, encompassing the leading-strand origin of replication and essential promoters for mitochondrial gene transcription, is pivotal in governing mtDNA replication and expression. Mutations in the D-loop could potentially disrupt these processes, although their exact biological implications remain ambiguous [28]. Our findings identified D-loop as the primary hotspot, representing the greatest proportion of documented mtDNA mutations in HNSCC; however, significant diversity exists among studies. On the other hand, Challen et al. [5], showed that the D-loop mutations are predominantly random occurrences with no functional significance in HNSCC etiology. Potential explanations for their occurrence encompass random mutational accumulation, tissue-specific variability, or discrepancies in detection methodologies. Dasgupta et al. [3] noted a single mutation in the D-loop region, which regulates the transcription of all 13 mtDNA-encoded genes, suggesting that during tumor progression, cells may prioritize the preservation of a functional D-loop while acquiring growth-promoting mutations in coding regions. Despite these uncommon cases, the comprehensive body of evidence and our pooled analysis clearly identify the D-loop as the primary locus of somatic mtDNA mutations in HNSCC.
However, its precise functional value has yet to be thoroughly explained. This contrast between high prevalence and disputed functional significance positions the D-loop primarily as a highly sensitive biomarker of genomic instability, in contrast to the coding region mutations discussed below, which exhibit clearer functional consequences.
To place the spectrum of mtDNA mutations into clinical context, we hypothesize a tiered application based on their distinct biological roles. The highly prevalent D-loop (≈67% pooled burden) is primarily suited as a highly sensitive detection marker in non-invasive screening, offering a frequent, shed signal of tumor presence due to its nature as a hotspot for genomic instability [5, 22]. Conversely, mutations in the coding regions appear to be functionally selected and hold greater potential for prognostic and predictive utility. Specifically, ND genes (Complex I; ≈29%) are validated by experimental evidence to drive carcinogenesis via oxidative stress and metabolic reprogramming [24], supporting their feasibility as pathogenic biomarkers that may predict response to metabolic therapies. Similarly, the enrichment of COX mutations (Complex IV; ≈12%) in recurrent tumors suggests they may serve as markers for treatment resistance or aggressive disease progression [3]. Finally, the modest but consistent presence of mutations in CYTB (Complex III) and the contributions of tRNA and rRNA genes point to a specialized, context-dependent utility, as their associated anti-apoptotic [29] and translational deficiencies [30, 31] may identify a subset of HNSCC susceptible to therapies targeting apoptosis evasion or mitochondrial ribosome vulnerability [32].
Our findings indicate that ND genes, which encode subunits of Complex I (NADH dehydrogenase), represent a moderate yet variable portion of the mitochondrial mutational burden in HNSCC. Complex I serves as the initial entry point for electrons in the respiratory chain and is essential in oxidative phosphorylation and the metabolic processing of carcinogens. Deficiencies in Complex I disrupt electron transport, resulting in the excessive generation of reactive oxygen species (ROS) and the accumulation of mutagenic damage [21]. Mutations in specific subunits, such as ND4 and ND5, have been associated with tumorigenesis across various cancer types [21, 33]. Numerous studies highlight the significance of ND mutations in HNSCC. Mondal et al. [19] indicated that 47% of somatic mtDNA coding mutations in OSCC patients were related to Complex I genes, including a harmful ND1 substitution, highlighting Complex I as a significant mutational target in these cancers. Zhou et al. [26] similarly noted a nonrandom distribution of mtDNA mutations, predominantly affecting ND subunits. They discovered an identical ND4L mutation in both tumor and adjacent dysplastic tissue, indicating that ND alterations may emerge early and clonally develop during field cancerization. They also conducted functional validation by introducing a representative ND2 mutation from an HNSCC tumor into cultured cells, which facilitated both anchorage-dependent and -independent growth, increased ROS, and induced a glycolytic shift accompanied by the stabilization of hypoxia-inducible factor-1α (HIF-1α), a characteristic of the Warburg effect. The oncogenic alterations were reversible using ROS scavengers, directly associating the phenotype with mitochondrial dysfunction-driven oxidative stress [26].
Overall, these findings indicate that ND mutations in HNSCC are not merely random passenger events; instead, they represent recurrent, functionally significant alterations that facilitate carcinogenesis via oxidative stress, metabolic reprogramming (Warburg effect), and clonal selection. The functional validation supporting a causal role for ND-driven ROS strongly supports their utility as functional biomarkers.
According to our results, mutations in the cytochrome COX genes (MT-CO1, MT-CO2, and MT-CO3) encoding Complex IV subunits represent a modest but variable component of the mitochondrial mutational burden in HNSCC, with potential functional roles in disrupting Complex IV activity and apoptotic regulation. Complex IV is the terminal enzyme of the respiratory chain, responsible for transferring electrons to molecular oxygen. It plays a critical role in minimizing ROS generation by ensuring efficient electron flow. Thus, defects in COX subunits can lead to electron leakage and heightened ROS levels [21]. Somatic mutations in COX genes have been documented in multiple cancer types (including pancreatic, prostate, colon, and thyroid malignancies) [34]. However, their involvement in head and neck cancer was not clearly established until recently [21]. Our findings, combined with emerging data, suggest that COX alterations do occur and may be important in HNSCC. In the study by Dasgupta et al. [3] on recurrent HNSCC, 75% of the identified coding-region mtDNA mutations were nonsynonymous, with the majority located in COX genes (Complex IV). This implies strong selection for COX dysfunction in the tumors of patients who experienced locoregional recurrence. In contrast, Mondal et al. [19] reported that COX mutations were relatively infrequent in primary OSCC tumors from a different cohort. Accounting for only one out of 17 coding mtDNA mutations detected (≈5–6%). This variation could reflect differences in disease context (recurrent vs. primary tumors), patient populations, or simply the limited sample sizes of individual studies. Nonetheless, the evidence of enrichment in recurrent tumors suggests that COX mutations, while present, may be under strong positive selection during disease progression or resistance to therapy, conferring a selective advantage in specific settings or subsets of HNSCC.
Recent evidence also indicates that inherited polymorphisms in COX genes may be potential risk factors. A case–control study in a Turkish population identified a particular germline variant in MT-CO1 (an adenine deletion at nucleotide position 6272) that was significantly more frequent in HNSCC patients than in healthy controls. The authors postulated that this CO1 variant could predispose to HNSCC, although further validation is required [21]. Altogether, the available data indicate that COX subunit mutations can arise during head and neck tumorigenesis and possibly confer growth or survival advantages via increasing oxidative stress or adaptation to hypoxic tumor environments. However, their exact role may vary between patient groups. Additional research is needed to determine the functional impact of COX mutations in HNSCC and to clarify whether certain COX changes might predict prognostic outcomes (e.g., recurrence) or resistance to specific treatment modalities.
The current study shows that CYTB mutations represent a small but consistent component of the mitochondrial mutational burden in HNSCC. Complex III (cytochrome bc1 complex) connects electron flow from Complex I/II to Complex IV, and mutations in CYTB can impair this process, leading to inefficiencies in electron transport, reduced ATP production, and increased ROS generation. Although CYTB mutations in HNSCC have not been as extensively studied as those in Complex I and IV genes, several studies have noted their occurrence in head and neck tumors. Dasgupta et al. [3] reported three somatic CYTB mutations among the mtDNA alterations found in HNSCC tumors, and most of the coding-region mutations in those tumors (65%) were nonsynonymous, suggesting that many CYTB variants are not neutral but potentially deleterious [35]. Some deep-sequencing analyses have likewise pointed to CYTB as a recurrently mutated gene in OSCC, labeling it a possible mutational hotspot in the mitochondrial genome of oral cancers [27]. Perhaps most compelling, experimental evidence from other models suggests that CYTB mutations can confer a survival advantage to cells. In an in vitro study by Dasgupta et al. [3], expression of a pathogenic 7-amino-acid deletion in the CYTB gene (originally identified in a bladder cancer) in a human uroepithelial cell line led to a striking mitochondrial phenotype: the mutant CYTB induced an increase in mitochondrial copy number and COX I protein levels, while markedly inhibiting apoptosis (as evidenced by cytoplasmic sequestration of Bax and lack of cytochrome c release or PARP cleavage).
The authors concluded that the CYTB mutation triggered mitochondrial proliferation and an anti-apoptotic cascade that favored sustained cell growth [29]. This mechanistic demonstration supports the notion that CYTB mutations, when present, can confer resistance to apoptosis and a growth advantage on cells [36]. In the context of HNSCC, our finding of a pooled positive association suggests that CYTB mutations, though less frequent than some D-loop or ND mutations, may be functionally important in a subset of tumors. Given the central position of Complex III in the electron transport chain, a deleterious CYTB mutation could elevate mitochondrial ROS production and contribute to the pro-survival, pro-proliferative milieu in cancer cell progression. This evidence calls for further investigation into the frequency of CYTB mutations in larger HNSCC cohorts and their potential therapeutic targets. The near-zero heterogeneity observed for CYTB in our synthesis indicates a stable, yet small, signal across studies, in contrast to the broad variability seen in D-loop and certain coding regions.
Our analysis also demonstrated that rRNA mutations represent a modest but highly variable component of the mitochondrial mutational burden in HNSCC, with evidence for both recurrence and functional relevance. The mitochondrial rRNAs (12S rRNA and 16S rRNA, encoded by MT-RNR1 and MT-RNR2) are integral components of the mitochondrial ribosome, and they are essential for translating the 13 mtDNA-encoded proteins of the oxidative phosphorylation system. Mutations in these rRNA genes can compromise mitochondrial ribosome assembly and function, leading to defective synthesis of respiratory chain subunits and, consequently, impaired oxidative phosphorylation and altered cellular energy homeostasis. Several studies have highlighted the occurrence and potential impact of mt-rRNA mutations in head and neck cancers. Dasgupta et al. [3] identified somatic mutations in both the 12S and 16S rRNA genes in HNSCC patient tumor samples. Schubert et al. [23] and Palodhi et al. [22] have reported that mitochondrial rRNA mutations are not only prevalent in HNSCC tumors but often present with high heteroplasmy, indicating that these mutations can clonally expand within tumor cell populations [21, 24]. The high heteroplasmic levels suggest that cells carrying rRNA mutations may gain a selective advantage, possibly inducing a state of mitochondrial dysfunction that cancer cells can tolerate or exploit (for example, shifting energy production toward glycolysis or activating stress signaling pathways). Lai et al. [15] observed that approximately 13.8% of all somatic mtDNA mutations in OSCC involved the rRNA genes, underlining that rRNA alterations comprise a substantial subset of mitochondrial mutations in these cancers [37]. Thoroughly, these data portray mt-rRNA mutations as an important element of the mitochondrial genomic aberrations in HNSCC. Their presence across multiple studies, combined with the known essential role of rRNAs in mitochondrial protein synthesis, raises the possibility that rRNA mutations contribute to tumor biology by diminishing mitochondrial translational output and forcing cells to adapt metabolically. It is also worth noting that mitochondrial rRNA mutations (especially in MT-RNR1) have clinical relevance beyond cancer, as certain mutations predispose individuals to ototoxicity from aminoglycoside antibiotics; this intersection of mitochondrial genetics and clinical phenotype reinforces the need to understand rRNA variants in cancer contexts as well. In the future, mt-rRNA mutations should be further explored as potential biomarkers (for early detection or risk stratification) or as mechanistic factors in HNSCC pathogenesis. The consistent presence of these mutations in tumors suggests they could help identify dysregulation of mitochondrial translation in cancers, and potentially guide therapeutic strategies that target mitochondrial ribosomes or the metabolic vulnerabilities of cells with compromised oxidative phosphorylation. Due to the significant heterogeneity observed across studies regarding rRNA, it is crucial to accurately differentiate between genuine somatic variants and haplogroup-associated polymorphisms through the use of matched normals and ancestry-aware filtering in future research.
Somatic mutations in mitochondrial tRNA genes showed that tRNA mutations represent a modest but highly variable fraction of the mitochondrial mutational burden in HNSCC. Mitochondrial tRNAs (mt-tRNAs) are crucial for intramitochondrial protein synthesis, and their unique evolutionary history has rendered them especially susceptible to mutation. Compared to canonical cytosolic tRNAs, mt-tRNAs have a reduced and relatively fragile secondary structure due to streamlined evolution and a higher baseline mutation rate [30]. Mitochondrial aminoacyl-tRNA synthetases recognize these unusual tRNAs via induced-fit interactions to ensure proper charging. Still, this mechanism only partially compensates for structural deficiencies, meaning mt-tRNAs remain highly vulnerable to deleterious sequence changes. Over 200 distinct mt-tRNA mutations have been linked to human diseases, many of which impair the tRNA’s ability to be aminoacylated, thereby disrupting mitochondrial protein translation [31]. In the context of cancer, a loss of mt-tRNA function can lead to ribosomal stalling and mistranslation of respiratory chain proteins, faulty assembly of electron transport complexes, and an increase in electron leak and ROS production due to inefficient respiration. Cells may attempt to compensate for these deficiencies by increasing mitochondrial biogenesis and producing more mitochondrial copies to boost overall protein output. Still, when this response is insufficient or dysregulated, influenced by nuclear genetic background and environmental factors, it may create conditions that favor tumor initiation and progression [38]. Our findings align with this understanding that the prevalence of mt-tRNA gene mutations in HNSCC tissue was significantly higher than expected by chance, suggesting that these mutations are not randomly distributed but rather may be clonally selected during tumor development. Numerous independent investigations support this. In the recurrent HNSCC study by Dasgupta et al. [3], over half (53%) of the non-coding mtDNA mutations identified were in tRNA genes, underscoring that mt-tRNA alterations constituted a major portion of the mutational burden in those tumors. Lai et al. [15] identified 20 different somatic mt-tRNA mutations (accounting for 8.3% of all mtDNA mutations) in a cohort of OSCC patients, further indicating that mt-tRNA alterations are a consistent feature of the mitochondrial mutational landscape in head and neck cancers. Moreover, high-throughput sequencing studies have revealed that many of these tRNA mutations are heteroplasmic (present in a fraction of mitochondrial genomes) rather than homoplasmic, implying ongoing clonal evolution within tumors. Schubert et al. [23] (using ultra-deep sequencing) and others have documented substantial mt-tRNA mutation loads and heteroplasmy levels in HNSCC tumors, suggesting possible clonal expansion of cells harboring advantageous tRNA mutations. We did detect some asymmetry in the funnel plot and Egger’s test for this category, indicating a degree of publication bias (smaller studies with null results might be underreported). Nevertheless, the overall consistency of evidence and the clear functional rationale for mt-tRNA mutations affecting mitochondrial physiology lend credibility to the idea that these mutations play a contributory role in HNSCC oncogenesis. Mt-tRNA mutations appear to impair mitochondrial protein synthesis and bioenergetics in ways that cancer cells can exploit, and they merit further study as potential biomarkers of disease or even targets for interventions that restore mitochondrial translation fidelity. According to our pooled estimates, tRNA mutations constitute a minor yet biologically plausible factor in mitochondrial dysfunction, with their clinical relevance dependent on standardized detection of heteroplasmy and comprehensive functional annotation.
The current study has several limitations. First, the included studies varied in their mutation detection methods, ranging from Sanger sequencing to next-generation sequencing (NGS), which led to inconsistencies in identifying low-frequency (heteroplasmic) variants. In addition, many studies had small, retrospective cohorts, which may bias results and reduce generalizability. Publication bias is a possibility, as studies with insignificant findings may be underreported. Some mutations labeled as somatic might be rare germline polymorphisms, especially when matched normal tissue was not analyzed. Moreover, the functional or clinical impact of detected mutations could not be assessed due to a lack of patient-level data. Furthermore, the overall ratio of mutant mtDNA in HNSCC relative to non-tumor tissues remains undetermined, as the majority of the included studies didn't provide comprehensive data on the entire mtDNA mutational load. The lack of this information hinders direct comparisons of global mutation loads between tumor and normal tissues and must be considered when interpreting the results.The analysis relied on single-arm proportions, limiting causal inference about the role of mtDNA mutations in HNSCC. Another limitation of this systematic review is the variability of HNSCC subsites. HNSCC encompasses anatomically and etiologically diverse areas, including the oral cavity, larynx, hypopharynx, and oropharynxfi [39]. Oropharyngeal squamous cell carcinoma is commonly associated with human papillomavirus (HPV) infection, whereas oral and laryngeal carcinomas are generally associated with tobacco and alcohol consumption [40, 41]. The significant heterogeneity evident between studies—arising from methodological and biological variances—must also be taken into account when assessing the pooled mutation estimates. The majority of the included studies did not stratify their subjects based on viral status or subsite, potentially introducing biological heterogeneity into the pooled mutation estimates [42]. Consequently, the identified mtDNA mutational patterns might suggest a combination of effects from both viral and environmental carcinogenesis. Future research should distinctly examine virally related and non-virally associated HNSCC to elucidate the distinctive mitochondrial mutational fingerprints of each subgroup. Despite these limitations, the selected articles aimed to explore and provide valuable insights into mitochondrial alterations in HNSCC. To improve future syntheses, standardized reporting of sequencing coverage, variant calling thresholds, and matched-normal usage would significantly decrease methodological variation and facilitate more detailed subgroup analysis (e.g., HPV status, subsite, exposure).
Future research should address several gaps identified in this meta-analysis. Large-scale, multi-center studies using uniform and high-resolution sequencing methods (e.g., ultra-deep NGS) are needed to accurately assess the spectrum and frequency of mtDNA mutations in HNSCC. Such efforts will reduce inter-study variability and clarify the prevalence of heteroplasmic variants. Studies should also consistently include matched normal tissue to distinguish true somatic mutations from inherited polymorphisms, especially considering population-specific haplogroup variation. Functional investigations are crucial to elucidate how specific mtDNA mutations (e.g., in ND, COX, CYTB, tRNA, and rRNA genes) influence tumorigenesis, bioenergetics, ROS generation, and therapy resistance.
Additionally, integrating mtDNA mutational data with nuclear genomic, transcriptomic, and metabolic profiles could provide a more holistic view of mitochondrial contributions to cancer. Future clinical studies should evaluate whether mtDNA mutations have prognostic or predictive utility.). Finally, the therapeutic potential of targeting mitochondria through metabolic inhibitors, ROS modulators, or strategies to restore mitochondrial translation. Ultimately, integrating quantitative mtDNA measures (mutation burden, heteroplasmy, copy number) with clinicopathologic characteristics and outcomes in prospective cohorts will be crucial to determine whether mtDNA changes might function as reliable biomarkers or actionable therapeutic targets in HNSCC.
Our study suggests a staged model in which significant D-loop instability interferes with mtDNA replication and transcription, leading to functionally relevant coding mutations—primarily in ND and, to a lesser extent, COX/CYTB/rRNA/tRNA—that increase oxidative stress and promote metabolic adaptation. This integrated perspective elucidates the predominance of D-loop/ND signals while integrating the diverse yet significant contributions of other regions, providing a coherent framework for future mechanistic and translational research.

Conclusion
In conclusion, this meta-analysis confirms that somatic mtDNA mutations are a ubiquitous characteristic of HNSCC, but their significance is tiered by genomic region. The highly prevalent D-loop (pooled prevalence ≈67%) is established as the primary mutational hotspot, positioning it as a potential highly sensitive detection marker for early mitochondrial genomic instability. Conversely, mutations in the coding regions of the Electron Transport Chain and translation apparatus appear to be functionally selected and offer greater potential for prognostic and predictive utility. Specifically, ND genes (Complex I; ≈29% pooled burden) are validated by experimental evidence to drive carcinogenesis via oxidative stress and metabolic reprogramming, supporting their feasibility as pathogenic biomarkers. Finally, the modest but consistent presence of mutations in CYTB (Complex III) and the variable contributions of COX (Complex IV), rRNA, and tRNA genes warrant further focused investigation to exploit their context-dependent roles in treatment resistance, apoptosis evasion, or metabolic vulnerabilities in specific subsets of HNSCC.