Advancing Deep Variant Phenotyping of Mitochondrial Enzyme Complexes For Precision Medicine in Allogeneic Hematopoietic Stem-Cell Transplantation

Abstract

Allogeneic hematopoietic stem-cell transplantation (allo-HCT), an early developed methodology for precision medicine, remains the only curative therapy for myelodysplastic syndromes (MDS). However, allo-HCT carries significant risks of morbidity and mortality due to relapse and transplant-related complications. Recurrent mutations in mitochondrial DNA (mtDNA) have been identified as significant prognostic indicators for MDS outcomes following allo-HCT. However, the biological mechanisms of mtDNA mutations remain unclear. Thus, here we performed deep variant phenotyping by integrating computational biophysics and structural genomics approaches to reveal the molecular mechanisms underlying mtDNA variant dysfunction. This emerging genomics discipline employs structural models, molecular mechanic calculations, and accelerated molecular dynamic simulations to analyze gene products, focusing on their structures and motions that determine their function. We applied this methodology on the variants in the mitochondria-encoded complex I genes that are associated with MDS pathobiology and prognosis after allo-HCT. Our results demonstrate that this approach significantly outperforms conventional analytical methods, providing enhanced and more accurate information to support the potential pathogenicity of these variants and better infer their dysfunctional mechanisms. We conclude that the adoption and further expansion of computational structural genomics approaches, as applied to the mitochondrial genome, have the potential to significantly increase our understanding of molecular mechanisms underlying the disease. Our study lays a foundation for translating mitochondrial biology into clinical applications, which will advance the integration of precision medicine with allo-HCT.

INTRODUCTION

Myelodysplastic syndromes (MDS) are a heterogeneous group of clonal hematopoietic cell disorders characterized by ineffective hematopoiesis, leading to blood cytopenias, and by a tendency to progress to acute myeloid leukemia (AML) in about 30% of patients.¹ Despite advances in treatment, allogeneic hematopoietic stem-cell transplantation (allo-HCT) remains the only curative therapy for MDS. However, mortality after HCT is high due to disease relapse and transplant-related complications. Less than 50% of patients with MDS survive for five years after transplantation.¹ Identifying which MDS patients will most likely benefit from allo-HCT is challenging, given the clinical and biological heterogeneity of the disease.¹ Prognostic tools defining MDS risk groups, such as the International Prognostic Scoring System (IPSS) and the revised IPSS (IPSS-R), which have varying probabilities of survival and leukemic transformation are used to predict posttransplant outcomes for clinical decision-making. Such models rely on traditional clinical and hematologic parameters and, thus, cannot accurately predict outcomes. During the past decade, sequencing technologies have dramatically improved our understanding of the genetics underlying MDS, with more than 50 recurrently mutated genes being identified, most of which also associate with unfavorable outcomes after allo-HCT.² Incorporating somatic mutations of 31 genes, Bernard et al. developed a clinical-molecular prognostic model [IPSS-Molecular (IPSS-M)], exhibiting improved prognostic discrimination across all clinical endpoints compared to the IPSS-R.³ Around the same time, the International Consensus Classification (ICC) published a revised framework for classifying myeloid neoplasms and acute leukemias, integrating genomic data.⁴ The shift to molecular markers may allow personalized medicine and has notably influenced clinical practice. Although it is now well recognized that genomic alterations play a central role in MDS, most studies have focused on characterizing mutations in the nuclear genome, overlooking the potential critical importance of the mitochondrial genome in achieving proper immunological competence and surveillance.

Mitochondria are critical dynamic organelles maintaining hematopoietic cell homeostasis and differentiation.⁵ They produce cellular energy and reactive oxygen species via oxidative phosphorylation (OXPHOS), trigger apoptosis, and control heme biosynthesis and iron metabolism.⁶ Dysfunction in any of these processes is pathogenic for MDS.⁶ In addition, allo-HCT exacts a high metabolic demand to restore normal hematopoiesis and immunologic functions. Mitochondria are not only the “powerhouse” of the cell, but also function in innate and adaptive immune response.⁷ Alterations in mitochondrial DNA (mtDNA), including genetic mutations and epigenomic modifications, could impair mitochondrial function and contribute to disease pathophysiology and survival outcomes.^6,8 For instance, we conducted whole-genome sequencing (WGS) to characterize the prognostic landscape of mitochondrial genome in MDS patients receiving allo-HCT and in their matched donors, which yielded insights into prognosis.⁹ Our findings demonstrate that genetic variants in frequently mutated mtDNA genes, such as mitochondrial Complex I genes (e.g., MT-ND4L and MT-ND5), are highly prognostic for allo-HCT outcomes in MDS patients.⁹ However, large-scale experimental validation of these variants remains impractical. For this purpose, we developed a multi-parametric mechanistic-based assessment that integrates conventional two-dimensional (2D) sequence-based annotation with computational biophysics and structural genomics approaches to understand the molecular dysfunction of genetic variants at the atomic level.^10–13 We have successfully validated our approach by revealing the molecular mechanisms of genetic variants in several nuclear genes (e.g., KRAS, KDM6A, KMT2C, and RAG).^10–13 Here, we used our multi-parametric approaches to investigate the molecular mechanisms of mtDNA variants that were associated with posttransplant outcomes, with a specific focus on mitochondrial genes that form the OXPHOS Complex I, the largest component of the mitochondrial OXPHOS system.

METHODS

2D Sequence-based analyses using linear sequence predictors for pathogenicity

To predict the pathogenicity of mtDNA variants, we integrated information from different databases. Variants, regardless of their biotype, were classified as “pathogenic” if they (1) were listed as “confirmed” pathogenic in MITOMAP database,¹⁴ (2) had clinical significance of “pathogenic/likely pathogenic” based on the modified American College of Medical Genetics and Genomics and Association of Molecular Pathology (ACMG/AMP) guidelines for mtDNA variants (criteria provided in ClinVar),¹⁵ or (3) identified as loss of function (LoF) variants by the Ensembl Variant Effect Predictor (VEP).¹⁶ According to the mtDNA specifications of the ACMG/AMP guidelines, mtDNA variants with allele frequency in the population database < 0.00002 provide evidence of pathogenicity.¹⁷ Based on this criterion, variants with AF_hom < 0.00002 in the gnomAD v3.1 database were also considered as “pathogenic” if they were non-synonymous and predicted to be deleterious from 5 of 7 prediction tools (HmtVar, SIFT, PolyPhen-2, MutPred2, PANTHER, PhD-SNP and SNPs&GO),^18–24 or had MToolBox disease score > 0.4311.²⁵

3D Structure-Based calculations for evaluation of mechanisms of dysfunction

The Cryo-EM structure of mitochondrial respiratory supercomplex, composed of three membrane-bound members of the ETC (i.e., complex I, dimeric complex III, and complex IV), were retrieved from RCSB Protein Data Bank (PDB; entry code: 5xth)²⁶ and used in the current analysis to give the most physiologically relevant context to our 3D assessment and structure-based calculations (Figure 1). We used well-established metrics from protein structural analysis (i.e., amino acid local packing density, solvation energy, hydrogen bonds, and salt bridges) and more advanced metrics (i.e., multi-body statistical contact potentials, energetic frustration, and thermodynamic cooperativity) and made images using PyMOL v2.5.4.^27–32 We performed evolutionary coupling analysis with tools from the Marks lab,³³ allosteric path analysis with tools from the Bahar lab.^34,35 The stability of the mutated protein were assessed by variant-induced changes in folding energy (ΔΔG_fold) using FoldX.³⁶ The explicit environment was modified for each mtDNA variant using the FoldX-generated models. Frustratometer was used to calculate residual frustration in the wildtype and mutated structures and to quantify changes in local interactions caused by each mtDNA variant.³²

Figure 1: Structure of the MT super-complex.

We label each selected variant with a red sphere marking its location.

RESULTS

Assembly and sequence-based evaluation of genomic variants in mitochondrial complex I that influence allo-HCT outcomes.

Our WGS data analyses of patients undergoing allo-HCT identified a total of 159 missense variants on mitochondrial Complex I genes (i.e., MT-ND1, MT-ND2, MT-ND3, MT-ND4, MT-ND4L, MT-ND5, and MT-ND6) as potentially pathogenic (Table 1).⁹ To test the gain of information that is generated by our deep variant phenotyping algorithms, we chose the variant that had the highest deleterious predicted score according to 2D annotation methods for each gene (i.e., MT-ND1 m.3470T>C, p.L55P; MT-ND2 m.5016T>C, p.S183P; MT-ND3 m.10176G>A, p.G40S; MT-ND4 m.11423G>A, p.E222K; MT-ND4L m.10726G>A, p.G86D; MT-ND5 m.12770A>G, p.E145G; and MT-ND6 m.14601G>A, p.P25S).

Table 1.

Potentially pathogenic missense variants on mitochodnrioal Complex I genes.

Variant	Locus	Amino Acid	Variant	Locus	Amino Acid	Variant	Locus	Amino Acid
3323TC	MT-ND1	L6P	10826AG	MT-ND4	I23V	12764GA	MT-ND5	G143D
3326TC	MT-ND1	L7P	10833TC	MT-ND4	I25T	12769GA	MT-ND5	E145K
3358GA	MT-ND1	A18T	10869TC	MT-ND4	I37T	12770AG	MT-ND5	E145G
3361TC	MT-ND1	F19L	11018GA	MT-ND4	E87K	12772GA	MT-ND5	G146S
3380GA	MT-ND1	R25Q	11064TC	MT-ND4	L102P	12782TC	MT-ND5	I149T
3394TC	MT-ND1	Y30H	11079TC	MT-ND4	I107T	12800TC	MT-ND5	I155T
3454GA	MT-ND1	A50T	11087TC	MT-ND4	F110L	12811TC	MT-ND5	Y159H
3470TC	MT-ND1	L55P	11126GA	MT-ND4	E123K	12830AG	MT-ND5	N165S
3517TC	MT-ND1	Y71H	11153AG	MT-ND4	I132V	12868GA	MT-ND5	G178S
3526GA	MT-ND1	A74T	11163GA	MT-ND4	R135Q	12871GA	MT-ND5	D179N
3550GC	MT-ND1	A82P	11210TC	MT-ND4	F151L	12875TC	MT-ND5	I180T
3572TC	MT-ND1	L89P	11225GA	MT-ND4	G156S	13018GA	MT-ND5	G228S
3578TA	MT-ND1	M91K	11273GA	MT-ND4	G172S	13063GA	MT-ND5	V243I
3578TC	MT-ND1	M91T	11276TC	MT-ND4	S173P	13069GA	MT-ND5	A245T
3590TA	MT-ND1	L95Q	11295TC	MT-ND4	L179P	13093GA	MT-ND5	V253I
3599TA	MT-ND1	L98H	11327TC	MT-ND4	W190R	13112TC	MT-ND5	L259S
3602AG	MT-ND1	N99S	11354TC	MT-ND4	Y199H	13129CT	MT-ND5	P265S
3605TA	MT-ND1	L100Q	11379TC	MT-ND4	M207T	13178GA	MT-ND5	G281D
3608GC	MT-ND1	G101A	11423GA	MT-ND4	E222K	13180GA	MT-ND5	A282T
3638TC	MT-ND1	L111P	11432AG	MT-ND4	I225V	13201GA	MT-ND5	A289T
3727TC	MT-ND1	S141P	11435GA	MT-ND4	A226T	13208GC	MT-ND5	C291S
3737TC	MT-ND1	V144A	11453GA	MT-ND4	A232T	13237AG	MT-ND5	I301V
3791TC	MT-ND1	L162P	11480TC	MT-ND4	Y241H	13238TC	MT-ND5	I301T
3918GC	MT-ND1	E204D	11490TC	MT-ND4	M244T	13240GA	MT-ND5	V302M
3929TC	MT-ND1	V208A	11493GA	MT-ND4	R245H	13246TC	MT-ND5	F304L
3935GA	MT-ND1	G210D	11535AG	MT-ND4	Y259C	13247TC	MT-ND5	F304S
3958GA	MT-ND1	G218S	11538CT	MT-ND4	P260L	13277TC	MT-ND5	M314T
4028TC	MT-ND1	I241T	11556TC	MT-ND4	L266P	13367GA	MT-ND5	G344E
4136AG	MT-ND1	Y277C	11583TA	MT-ND4	I275N	13369TC	MT-ND5	S345P
4211TC	MT-ND1	M302T	11642GA	MT-ND4	A295T	13373TC	MT-ND5	I346T
4600TC	MT-ND2	L44P	11693GA	MT-ND4	A312T	13376TC	MT-ND5	I347T
4665GA	MT-ND2	A66T	11790TC	MT-ND4	L344P	13414GC	MT-ND5	G360R
4752TC	MT-ND2	S95P	11891GA	MT-ND4	E378K	13499GA	MT-ND5	G388D
4849GA	MT-ND2	G127D	11940TC	MT-ND4	L394P	13528AG	MT-ND5	T398A
4903TC	MT-ND2	I145T	11955AG	MT-ND4	N399S	13538TC	MT-ND5	M401T
4950TC	MT-ND2	S161P	11988TC	MT-ND4	M410T	13558GA	MT-ND5	A408T
4986AG	MT-ND2	T173A	10534AG	MT-ND4L	Y22C	13664TC	MT-ND5	I443T
5016TC	MT-ND2	S183P	10554TC	MT-ND4L	S29P	13691AG	MT-ND5	N452S
5032GC	MT-ND2	G188A	10663TC	MT-ND4L	V65A	13718GA	MT-ND5	S461N
5050TC	MT-ND2	L194P	10675GA	MT-ND4L	C69Y	13844AG	MT-ND5	D503G
5055TC	MT-ND2	Y196H	10726GA	MT-ND4L	G86D	13931TA	MT-ND5	I532N
5136TC	MT-ND2	S223P	10731GA	MT-ND4L	D88N	13952AG	MT-ND5	Y539C
5148AC	MT-ND2	T227P	12457GA	MT-ND5	A41T	14050TC	MT-ND5	S572P
5224GA	MT-ND2	G252E	12524TC	MT-ND5	I63T	14093TC	MT-ND5	L586P
5262GC	MT-ND2	A265P	12601TC	MT-ND5	F89L	14180TC	MT-ND6	Y165C
10138TC	MT-ND3	L27P	12613GA	MT-ND5	A93T	14231TC	MT-ND6	Y148C
10176GA	MT-ND3	G40S	12631TC	MT-ND5	S99P	14424AG	MT-ND6	S84P
10237TC	MT-ND3	I60T	12635TC	MT-ND5	I100T	14484TC	MT-ND6	M64V
10282TC	MT-ND3	L75P	12638TC	MT-ND5	M101T	14501AG	MT-ND6	I58T
10290GA	MT-ND3	A78T	12650TC	MT-ND5	L105P	14525TC	MT-ND6	Y50C
10353GC	MT-ND3	A99P	12673AG	MT-ND5	N113D	14601GA	MT-ND6	P25S
10360GA	MT-ND3	S101N	12704TC	MT-ND5	I123T	14606GA	MT-ND6	P23L
10805TC	MT-ND4	W16R	12731TC	MT-ND5	V132A	14619AG	MT-ND6	F19L

Deep variant phenotyping using computational structural genomics.

Subsequently, we performed the 3D structure-based assessment, initially of variants by examining the environments around the mtDNA variants. We found that mutations of MT-ND4:E222 (Figure 2A) and MT-ND5:E145 (Figure 2B) were likely to change the dynamics of the complex because they are internally facing charges surrounded by a mixture of positive and negative charges. The residue E222 is in a solvent exposed helix and appears to have a crucial role in holding the adjacent solvent exposed helix by hydrogen bond. Additionally, the residue E145 is located at the interface stabilizing the interaction with MT-ND4. We anticipate that such an arrangement balances architectural stability with essential flexibility. Mutation to a different charge (e.g., E222K) or that form greater internal cavities (e.g., E145G) will have dynamic effects on the structure or the electron transport process. Interestingly, these candidate variants were selected without structural data, and there are proximal relationships among them like MT-ND6:P25 being across from MT-ND4L:G86 (Figure 2C). The role of G86 is probably to provide a steric free edge at the end of helix for a sharp turn of C-terminal peptide, which could otherwise not be stabilized if placed in other direction. We anticipate that any bulky mutation at this site could be deleterious. In addition, Evolutionary coupling analysis identified all seven variants as deleterious. Allosteric path analysis identified six out of seven variants as deleterious with one neutral call (MT-ND4:E222K) but within 1% of the threshold required for assignment to this category. Structural stability analysis, which is based on molecular mechanic calculations of folding energy identified one variant with neutral (MT-ND5:E145G), one stabilizing (MT-ND4:E222K) and the remaining five variants with highly destabilizing effect (Table 2). Thus, this data indicates that the prognostic effects of these mtDNA variants on MDS patients receiving allo-HCT likely involve their deleterious impacts on mitochondrial function. Thus, when using computational biophysics and structural genomics for their analyses, mtDNA variants are a powerful prognostic tool for predicting the outcome of allo-HCT and advances the field of precision medicine for MDS.

Figure 2: 3D structure-based assessment for mtDNA variants.

A) MT-ND4:E222 and B) MT-ND5:E145 are located within internal pockets that are lined with combinations of positive and negative charges, likely facilitating the flexible conformational changes that are required for electron transport. C) Selected mutations are nearby in space and typically at the ends of alpha helices where they likely modulate flexibility.

Table 2.

Structural stability analysis for mtDNA variants.

Gene	Amino Acid	ΔΔGfold (kcal/mol)
MT-ND1	L55P	5.9 ± 0.2
MT-ND2	S183P	5.4 ± 0.3
MT-ND3	G40S	6.9 ± 0.0
MT-ND4	E222K	−1.5 ± 0.2
MT-ND4L	G86D	3.6 ± 0.2
MT-ND5	E145G	−0.2 ± 0.3
MT-ND6	P25S	4.3 ± 0.0

DISCUSSION

Considering the novelty of the results reported in this study, it becomes important to briefly highlight additional applications. Computational biophysics and structural genomics increase our ability to better predict pathogenicity than conventional methods due to the precision of these methods over the ones popularized by guidelines. More importantly, this approach provides explanations for mechanisms of dysfunction underlying the impact of variants, in particular because of its used of advanced methods of molecular dynamic simulations. Lastly, since many drugs have been developed using similar methodologies, the conformations that results from molecular dynamics simulation build the trajectory toward the identification of mutant-specific cavities for development of organic probes or small drugs. Therefore, we anticipate that the continue application to the deep variant characterization of both nuclear and Mt-DNA-encoded mitochondria proteins will significantly advance not only the field of MDS but other genomic diseases of importance to our field. In conclusion, in the current molecular era, mtDNA variants are a powerful prognostic tool for both predicting the outcome of allo-HCT and represent an advancing the field of precision medicine for MDS. Studies such as we illustrated here increase our ability to better predict pathogenicity, provide explanations for their mechanism of dysfunction, and yield structural conformation to help build novel therapeutic tools. In this manner, we can better leverage mtDNA genomics into the functional interpretation and actionable potential. Thus, we are optimistic that this study will serve as a useful framework for characterizing the spectrum of translational research in the field of HCT, from bedside (e.g., measuring individual’s genomics) to bench (e.g., computational structural genomics) then back to bedside (e.g., improved risk stratification and clinical outcomes), necessary for our precision medicine practice and research. Such work requires a multidisciplinary team with high level of expertise in HCT clinical management, molecular epidemiology, systems biology, mitochondrial biochemistry, computational biophysics, computational chemistry, and stem cell biology. Should such a proposed translational cycle be successful, it will be a paradigm shift in allo-HCT, from donor selection, recipient risk stratification to mitochondrial targeted therapying in HCT, MDS, and most likely to other diseases caused by alterations in this organelle.