Heteroplasmic variants of mitochondrial genome in tumor and normal samples from patients with hematologic malignancies

Mitochondria are maternally inherited organelles involved in energy metabolism, proliferation, apoptosis, and immune regulation [ 1 ]. Due to its high copy number and mutation rate, mitochondrial DNA (mtDNA) frequently exhibits heteroplasmy, a state in which wild-type and mutant mtDNA molecules coexist within the same cell [ 2 ]. While heteroplasmic mtDNA mutations have been extensively characterized across human solid cancers, their relevance in hematologic malignancies remains poorly investigated [ 3 ]. Oxidative stress is increasingly recognized as a contributor to cancer pathogenesis, and the mitochondrial genome is particularly vulnerable to reactive oxygen species (ROS)-induced DNA damage. In fact, alterations of the mitochondrial genome may be cause and consequence of tumor development, and may cooperate with potent oncogenes such as MYC , regulating metabolic fitness and chemoresistance [ 4 ]. This could be particularly relevant in diffuse large B-cell lymphoma (DLBCL), acute myeloid leukemia (AML) and chronic lymphocytic leukemia (CLL), in which a pathogenic role of oxidative stress has been described [ 5 , 6 , 7 ]. Here, we explored the mitochondrial mutational landscape of DLBCL, AML, and CLL and identified recurrent heteroplasmic mtDNA variants that extended beyond tumor cells and were also detected in normal tissues. We analyzed publicly available whole genome (WGS) and whole exome sequencing (WES) datasets from DLBCL, AML, and CLL patients alongside two independent healthy donor cohorts. Specifically, we analyzed 80 matched tumor–normal WGS (40 patients) and 74 WES samples (37 patients) from DLBCL, 86 WGS (43 patients) and 298 WES samples (159 patients) from AML, and 60 WGS samples (30 patients) from CLL, together with WGS data from peripheral blood mononuclear cells (PBMCs) of 130 healthy donors (80 adults and 50 children). After correction for nuclear-embedded mitochondrial sequences (NUMTs), WGS achieved high mtDNA coverage (mean ≈ 4280×), enabling reliable detection of low-heteroplasmy variants (>0.1%), whereas WES provided substantially lower mtDNA coverage. Analyses were therefore primarily based on WGS data, with WES cohorts used for validation (Fig. 1a and Supplementary Fig. S1a–d ). Fig. 1: Landscape of heteroplasmic mtDNA variants across DLBCL, AML and CLL. Full size image a Representative circos plot showing mean mtDNA coverage across each locus before NUMT correction (outer circle) and after NUMT correction (inner circle) for matched tumor and normal samples coming from WGS platform. b Boxplots illustrating the distribution of mtDNA variants per sample across all analyzed datasets, stratified by disease type (DLBCL, AML, CLL) and healthy donors (adults and children). Variants are further classified according to their genomic location as coding or non-coding. Statistical comparisons between groups are indicated above the plots (* p < 0.05, ** p < 0.01, *** p < 0.001, **** p < 0.0001). c Oncoplots displaying mtDNA variants identified in DLBCL, AML, and CLL tumor samples together with their matched normal counterparts ( d) . Each row corresponds to a distinct mitochondrial mutation, with the percentage of affected samples indicated at the right margin. Each column represents an individual patient sample. The bar plots above the oncoplots summarize the mutational burden per sample (number of mtDNA variants detected). e Kaplan–Meier overall survival analysis in DLBCL (WES), comparing patients harboring both mtDNA mutations and MYC alterations versus others. f Kaplan–Meier overall survival analysis in AML (WGS), comparing patients with and without mtDNA variants. Survival differences were assessed using the log-rank test. g Multivariate Cox regression analysis of overall survival in AML, including mtDNA status, age, genomic rearrangements and recurrent nuclear mutations. mtDNA mitochondrial DNA, NUMT nuclear-embedded mitochondrial sequence, WGS whole-genome sequencing, DLBCL diffuse large B-cell lymphoma, AML acute myeloid leukemia, CLL chronic lymphocytic leukemia, WES whole-exome sequencing. Across the analyzed WGS datasets, mtDNA variants were predominantly missense transitions, mainly T > C and A > G, consistent with previous reports [ 3 ] (Supplementary Fig. S2a ). Overall, by using a standard reference mitochondrial genome [Revised Cambridge Reference Sequence (rCRS) [ 8 ]], we detected a mean of 79 variants per sample. As expected from the known accumulation of mtDNA mutations with age, adult healthy donors showed a higher number of variants compared to children [ 9 ]. In patients, variants were detected in both tumor and matched normal samples (PBMCs in DLBCL, skin in AML, saliva in CLL). DLBCL, AML and CLL patients showed increased variant counts relative to healthy controls; however, patients affected by DLBCL exhibited a higher burden of mitochondrial variants in both tumor and normal compartments compared with all other subsets (Fig. 1b ). To identify disease-specific mtDNA variants, we filtered out mutations shared with healthy donors, identifying 21, 19 and 16 recurrent variants in DLBCL, AML, and CLL, respectively (Fig. 1c and Supplementary Tables S1 – S3 ). In DLBCL, 18 out of 21 variants occurred in protein-coding genes, particularly MT-ND5 and MT-ND4L , encoding subunits of respiratory chain complex I. Nearly all DLBCL patients ( n = 39, 98%) harbored at least one mutation, often with multiple co-occurring variants (median = 4). All recurrent mutations were detected in a heteroplasmic state, with levels ranging from 0.2 to 65% (mean = 4.1%). Among these, 7 variants were tumor-specific, mainly involving MT-ND1 , and were sporadic across patients. In contrast, 14 variants were also detected in matched normal samples at comparable heteroplasmy levels (mean = 3.6%) (Fig. 1d ). Of note, 11 of 21 variants were validated in an independent WES cohort (Supplementary Fig. S2b ). Given the role of the MYC oncogene in DLBCL pathogenesis and mitochondrial fitness, we investigated the combined effect of MYC genomic alterations and mtDNA mutations on patient outcomes. We found that co-occurrence of mtDNA mutations and MYC alterations (rearrangements, mutations or amplifications) was associated with a significantly inferior overall survival (OS) (Fig. 1e and Supplementary Fig. S2c, d ). The presence of mitochondrial variants was not associated with specific patterns of nuclear mutations (Supplementary Fig. S2e ). In AML, we identified 19 recurrent variants in 32 patients (74%), most frequently affecting MT-ND5 and MT-ATP6 , with 5 mutations shared with DLBCL. All mutations were heteroplasmic with levels ranging from 0.2% to 77% (mean = 2%). Six variants were tumor-specific, including two co-occurring MT-ATP mutations observed in four patients (9%). In contrast, 13 variants were also detected in matched normal samples at lower heteroplasmy levels (mean = 0.7%), likely reflecting the differences in normal tissue source. All variants were validated in an independent WES cohort (Supplementary Fig. S2f ). Of the 19 variants, 9 occurred in rRNA/tRNA-coding genes and in the D-Loop region. In particular, 2 mutations were located in tRNA-coding genes ( MT-TF and MT-TL2 ). Given the lack of redundancy of mitochondrial tRNAs, these mutations may impact structural integrity and translation (Supplementary Fig. S2g ). In AML, patients harboring at least one mtDNA variant showed significantly worse OS compared with non-mutated patients, which was confirmed in multivariate analysis (Fig. 1f, g ). Notably, the stratification based on common nuclear mutations ( NPM1 , FLT3 , DNMT3A ) did not result in significant differences in OS (Supplementary Fig. S2h ). In CLL, we identified 16 recurrent heteroplasmic variants that were also detected in matched normal samples but absent in healthy donors. To validate the identified variants and their heteroplasmy levels, we performed targeted mitochondrial sequencing on mtDNA extracted from isolated mitochondria of 13 DLBCL and 10 AML cell lines. In DLBCL, all 21 recurrent variants were confirmed across the analyzed cell lines, whereas 17 of 19 variants were validated in AML models (Fig. 2a, c ). Consistent with the in-silico analyses, most variants showed comparable heteroplasmy levels; however, some displayed increased heteroplasmy in purified mitochondrial DNA. In DLBCL cell lines, this was observed for variants in MT-CO3 (m.9824 T > A), MT-ND1 (m.4051 G > A), and MT-ATP6 (m.8854 G > A), with mean heteroplasmy increasing from 66.8 to 74%, 0.27 to 7.41%, and 0.22 to 13.2%, respectively (Fig. 2b ). This finding could suggest improved detection accuracy when sequencing purified mitochondrial DNA. Notably, the MT-ATP6 m.8854 G > A mutation also showed increased heteroplasmy in AML cell lines, suggesting a shared feature across malignancies (Fig. 2d ). Fig. 2: Validation of mtDNA variants across tumor cell lines and immune cells. Full size image Oncoplot showing the mtDNA variants validated by targeted mitochondrial sequencing across 13 DLBCL cell lines ( a ) and 10 AML cell lines ( c ). The right panels compare mean heteroplasmy levels of the mutations as detected in silico from WGS data (left side) and experimentally validated by mtDNA-seq (right side) in DLBCL cell lines ( b ) and AML cell lines ( d ). Connecting lines indicate the same mutation across the two platforms, highlighting differences in heteroplasmy detection. e Oncoplot showing the mtDNA variants validated by mtDNA-seq performed on normal T lymphocytes from 5 DLBCL patients. The right panel compares mean heteroplasmy levels of MT-ATP6 m.8854 and MT-ND1 m.4051 mutations as detected in silico from WGS data (left side), in DLBCL cell lines (middle) and in normal T lymphocytes of DLBCL patients (right side) ( f ). DLBCL diffuse large B-cell lymphoma, AML acute myeloid leukemia, WGS whole-genome sequencing, mtDNA-seq mitochondrial DNA sequencing, mtDNA mitochondrial DNA. Since many recurrent variants were also detected in normal counterparts of DLBCL patients at similar heteroplasmic levels, we next performed targeted mitochondrial sequencing in purified T lymphocytes from 5 DLBCL patients. Thirteen of the 21 disease-specific variants were detected in these cells, including the MT-ATP6 m.8854 G > A and MT-ND1 m.4051 C > T mutations, which showed higher heteroplasmy levels (19% and 10%, respectively) (Fig. 2e, f ). These findings confirm that recurrent mtDNA variants in DLBCL are also present in non-malignant immune populations. To explore the potential relationship between mitochondrial and nuclear genomic alterations, we analyzed the co-occurrence between mtDNA variants and recurrent nuclear DNA (nDNA) mutations in the WGS cohorts. In DLBCL, mtDNA variants were broadly distributed across patients and showed no significant association with mutations in major lymphoma-associated genes, indicating that DLBCL is uniformly characterized by high burden of mitochondrial mutations irrespective of the pattern of genomic alterations in their nuclear genome (Supplementary Fig. S3a ). In AML, mutations in DNMT3A were observed to co-occur with several mtDNA variants, including alterations in MT-ND5 , MT-CO2 , MT-RNR2 and MT-TF , suggesting an interaction between nuclear and mitochondrial genomic alterations (Supplementary Fig. S3b ). No significant associations between mitochondrial and nuclear alterations were observed in CLL (Supplementary Fig. S3c ). Overall, our findings reveal a previously underappreciated landscape of heteroplasmic mitochondrial variants in hematologic malignancies, with potential prognostic impact. While functional consequences have been mostly described for homoplasmic mtDNA variations, recent literature confirms that mitochondrial heteroplasmic SNVs may serve as a prognostic marker for cancer [ 10 ]. The importance of mtDNA alterations in hematologic malignancies is being increasingly recognized, with recent studies linking mitochondrial genomic alterations to clonal evolution, metabolic vulnerabilities and disease progression in leukemia [ 11 , 12 , 13 ]. However, a comprehensive characterization of recurrent mtDNA variants across hematologic malignancies is still lacking, especially for low-heteroplasmy mutations. In this study, we analyzed whole-genome and whole-exome sequencing samples from DLBCL, AML and CLL patients, alongside two cohorts of healthy donors. We identified recurrent heteroplasmic mtDNA mutations that were absent in healthy donors. The presence of mtDNA mutations was associated with inferior OS in AML and MYC -altered LBCL, which are still considered unmet medical needs, providing prognostic information not captured by established nuclear genomic features. A subset of low-heteroplasmy mutations was detected exclusively in tumor samples; although rare, these tumor-restricted events may represent early alterations that arise during tumor evolution. By contrast, the detection of recurrent mtDNA mutations in both malignant and non-malignant cellular compartments suggests that mitochondrial genomic alterations may extend beyond tumor cells and potentially influence the broader cellular environment. The presence of these variants in effector T cells of DLBCL patients supports a recently described mechanism of immune evasion by cancer cells through mitochondrial transfer [ 14 ]. These mtDNA alterations could potentially impair the metabolic fitness of T cells and ultimately reduce their antitumor activity, which may be particularly relevant in the context of T-cell-redirecting therapies currently used in DLBCL [ 15 ]. Overall, these findings highlight the clinical and biological relevance of mtDNA heteroplasmy in hematologic malignancies. Future studies integrating larger cohorts and functional analyses will be necessary to clarify the biological and clinical implications of mtDNA heteroplasmy in hematologic malignancies.