Prognostic landscape of mitochondrial genome in myelodysplastic syndrome after stem-cell transplantation

Myelodysplastic neoplasia (also known as myelodysplastic syndromes, MDS) are a heterogenous group of clonal hematopoietic cell disorders characterized by blood cytopenias and a tendency to progress to acute myeloid leukemia (AML) [1]. Despite treatment advances, allogeneic hematopoietic stem-cell transplantation (allo-HCT) remains the only potentially curative therapy for MDS. However, mortality after allo-HCT is high due to disease relapse and transplant-related complications [1]. Deciding which MDS patients will most likely benefit from allo-HCT is challenging given the clinical and biological heterogeneity of the disease [1]. Previous genomic analyses have shown that mutations in specific nuclear genes (e.g., TP53) could inform prognostic stratification of MDS undergoing allo-HCT [2]. Surprisingly, despite the critical role of mitochondria in energy production, heme biosynthesis and metabolism [3], there has not yet been a comprehensive evaluation of mutations in the mitochondrial genome on transplant outcomes for MDS.

To address this knowledge gap, we performed whole-genome sequencing (WGS) on whole blood samples obtained before allo-HCT from 494 patients with MDS and analyzed their mitochondrial genomes (Additional file 2: Table S1). All patients were of European ancestry. Details of mtDNA analysis are provided in Additional file 3. We identified 2666 mtDNA variants (2542 substitutions and 124 small indels), 250 (9.4%) of which have not been reported in the MITOMAP database (Fig. 1A). Among them, 411 variants are putative pathogenic, with the majority (95%) having low allele frequency (AF < 1%) in our cohort. This is consistent with the ACMG/AMP assessment that pathogenic variants tend to be rare [4]. The mitochondrial control region (also known as the D-Loop) was the most frequently mutated region with 14.4% of the variants located in this region (Fig. 1B and C). The most frequently mutated protein-coding gene was MT-ND5 (16%), followed by MT-CO1 (14%). Eight percent of the variants were in the mitochondrial MT-tRNA genes. Mutational signature analysis showed that T > C and C > T transitions were the predominant substitutions in MDS (Additional file 1: Fig. S1). These results are consistent with previous reports from small-scale MDS studies [5].

The median number of mtDNA mutations identified per patient was 37 (ranging from 4 to 157). Presence of mtDNA mutations was not associated with patient age at transplant, Karnofsky Performance Score (KPS), the Revised International Prognostic Score (IPSS-R), MDS types, or cytogenetic abnormalities (all P > 0.05, Fig. 1D). Overall, an increase in the number of mitochondrial mutations was significantly associated with a decrease in overall survival (OS) (HR, 1.11; 95% CI, 1.01–1.22; P = 0.029), higher risk of relapse (HR, 1.13; 95% CI, 1.00–1.27; P = 0.049) and shorter relapse-free survival (RFS) (HR, 1.13; 95% CI, 1.04–1.24; P = 0.007). Analyses for each putative pathogenic variant observed 13 rare variants (11 missense and 2 stop-gain) that were associated with at least one of the four post-transplant outcomes (i.e., OS, relapse, RFS, and transplant-related mortality (TRM)) after Bonferroni correction (P < 0.05/411 = 1.22 × 10^–4) (Additional file 2: Table S2). All these 13 variants were present in a heteroplasmic state, with HF levels ranging from 1% to 68.2%. Common variants (AF ≥ 1%) that were associated with posttransplant outcomes at P values < 0.05 were listed in Additional file 2: Tables S3–S6, 23 of which were associated with more than one outcome. Among them, MT-CO3 m.9656 T > C, MT-ND2 m.5495 T > C, and MT-CYB m.15607A > G reached Bonferroni-adjusted threshold (P < 0.05/27 = 1.85 × 10^–3) in the conditional analysis. The Fine-Gray model derived similar results for relapse and TRM, with minor differences (Additional file 2: Tables S5–S6).

Gene-based analyses yielded significant associations with OS for MT-CYB (P = 1.04 × 10^–3), MT-ND2 (P = 3.06 × 10^–3) and MT-ND4 (P = 1.41 × 10^–3) after Bonferroni correction (P < 0.05/16 = 3.13 × 10^–3). For RFS, significant associations were observed for MT-CYB (P = 2.78 × 10^–3) and MT-ND4 (P = 7.75 × 10^–4). Four mtDNA genes were significantly associated with relapse, including MT-CYB (P = 7.03 × 10^–4), MT-ND2 (P = 5.92 × 10^–4), MT-ND5 (P = 1.91 × 10^–3) and MT-tRNA (P = 1.99 × 10^–4). Two mtDNA genes (MT-CYB and MT-ND4L) were significantly associated with TRM, with P values of 2.24 × 10^–3 and 4.91 × 10^–4 for MT-CYB and MT-ND4L, respectively (Additional file 2: Table S7). Additional significantly associated mitochondrial genes were also observed in burden test and/or SKAT (Additional file 2: Table S8 and Additional file 1: Figs. S2–S5). Most of these associated genes are located on the mitochondrial electron transport chain (ETC). ETC function is coupled to oxidative phosphorylation and the production of metabolites by the tricarboxylic acid cycle [6]. Mutations on these genes could result in complex dysfunction and abnormal reactive oxygen species production, which further promotes tumorigenesis and tumor progression [7].

Sixteen haplogroups were predicted in our cohort. Consistent with previous studies in non-Hispanic whites, haplogroup H was the most common haplogroup in our cohort [8]. Compared to haplogroup H, haplogroup I was significantly associated with worse OS (HR, 2.32; 95% CI, 1.23–4.35; P = 0.01) and shorter RFS (HR, 2.04; 95% CI, 1.09–3.83; P = 0.03). Haplogroup K was significantly associated with increased risk of relapse (HR, 1.70; 95% CI, 1.03–2.81; P = 0.04) (Additional file 2: Table S9).

To investigate whether mtDNA mutations could improve the prognostic stratification of MDS receiving allo-HCT, we fitted random survival forest models with and without inclusion of mtDNA mutations in the models (Additional file 1: Fig. S6). The model based only on mtDNA genes had a c-index of 0.58 to predict OS, which was slightly higher than the IPSS-R (c-index = 0.48) and clinical model (IPSS-R plus MDS type and pre-transplantation treatments, c-index = 0.57). Adding mtDNA genes improved the predictive performance of the model, with the c-index increasing from 0.48 to 0.63 for the IPSS-R model and from 0.57 to 0.66 for the clinical model (Fig. 2). Similar results were observed for the models in predicting RFS, relapse and TRM. Because IPSS-R is the currently available clinical scoring system to predict OS and leukemic transformation of MDS, we further examined the reclassification properties of adding mtDNA mutations to the IPSS-R model in predicting OS and relapse (Additional file 1: Fig. S7). The model with mtDNA mutations up-staged 18.6%, 41.7%, and 52.5% of the patients having lower-risk IPSS-R scores and down-staged 35.1%, 14.7% and 1.6% of the patients having higher-risk IPSS-R scores for 1-, 2-, and 5-year OS, respectively. For relapse, the model with mtDNA mutations up-staged 18.8%, 22.0%, and 22.0% of the patients having lower-risk IPSS-R scores and down-staged 26.2%, 35.6% and 35.6% of the patients having higher-risk IPSS-R scores for 1-, 2-, and 5-year risk of relapse, respectively.

Most recently, Bernard et al. developed a clinical-molecular prognostic model, termed the IPSS-Molecular (IPSS-M) model, which combines somatic mutations of 31 genes with hematologic and cytogenetic parameters [9]. The IPSS-M improved prognostic discrimination across all clinical end points and reclassified 46% of MDS patients as compared to the IPSS-R model, demonstrating the importance and clinical utility of recurrent mutations in MDS risk stratification [9]. Although our current analysis focused on the prognostic significance of mtDNA mutations, to investigate whether mtDNA mutations could provide additional prognostic stratification to the recurrent mutations, we further conducted stratified analysis by TP53 mutation status. We chose TP53 mutation as an example because our group and others have repeatedly identified TP53 mutation as a powerful predictor of MDS survival after transplantation [2, 9‐11]. However, more than 80% of MDS patients do not carry TP53 mutations. In these patients, additional prognostic markers are needed to further stratify their posttransplant outcomes [2, 10, 11]. In our MDS cohort, TP53 mutations were present in 11% of the patients and as expected, were associated with shorter OS, increased risk of relapse, shorter RFS and worse TRM than those without TP53 mutations (all log-rank P < 0.01, Additional file 1: Fig. S8). The presence of TP53 mutations was not correlated with the number of mtDNA mutations, nor the mutation status of each mtDNA gene (all P > 0.05, Fig. 1D). Of note, mtDNA mutations could provide additional prognostic stratification for patients who don’t carry TP53 mutations and improved the predictive performance of the models based on the IPSS-R and clinical factors (Additional file 2: Table S10). These findings suggest that mtDNA mutations could provide additional prognostic stratification information to the recurrent nuclear DNA mutations. However, we did not systematically evaluate the accumulated effects of nuclear DNA mutations and mtDNA mutations on patient survival, which requires further investigation.

In conclusion, this was the first attempt to characterize mtDNA genomic landscape in MDS receiving allo-HCT. Our results provide novel insights into the clinical utility of mtDNA mutations and could serve as a proof of concept that integration of mtDNA mutations into the scoring system to improve the clinical decision-making. Further studies are needed to better understand the biological mechanisms, and new techniques for mitochondrial genome editing are required to help restore mitochondrial function and develop mitochondria-targeted therapy.