Data-independent acquisition-based quantitative proteomic analysis of m.3243A>G MELAS reveals novel potential pathogenesis and therapeutic targets

The pathogenesis of mitochondrial myopathy, encephalopathy, lactic acidosis and stroke like episodes (MELAS) syndrome has not been fully elucidated. The m.3243A > G mutation which is responsible for 80% MELAS patients affects proteins with undetermined functions. Therefore, we performed quantitative proteomic analysis on skeletal muscle specimens from MELAS patients. We recruited 10 patients with definitive MELAS and 10 age- and gender- matched controls. Proteomic analysis based on nanospray liquid chromatography-mass spectrometry (LC-MS) was performed using data-independent acquisition (DIA) method and differentially expressed proteins were revealed by bioinformatics analysis. We identified 128 differential proteins between MELAS and controls, including 68 down-regulated proteins and 60 up-regulated proteins. The differential proteins involved in oxidative stress were identified, including heat shock protein beta-1 (HSPB1), alpha-crystallin B chain (CRYAB), heme oxygenase 1 (HMOX1), glucose-6-phosphate dehydrogenase (G6PD) and selenoprotein P. Gene ontology and kyoto encyclopedia of genes and genomes pathway analysis showed significant enrichment in phagosome, ribosome and peroxisome proliferator-activated receptors (PPAR) signaling pathway. The imbalance between oxidative stress and antioxidant defense, the activation of autophagosomes, and the abnormal metabolism of mitochondrial ribosome proteins (MRPs) might play an important role in m.3243A > G MELAS. The combination of proteomic and bioinformatics analysis could contribute potential molecular networks to the pathogenesis of MELAS in a comprehensive manner.


Introduction
The mitochondrial diseases are one of the most common congenital metabolic defects, with a minimum prevalence of 1:5000 1, 2 . Mitochondrial myopathy, encephalopathy, lactic acidosis and stroke like episodes (MELAS) is one of the relatively common mitochondrial diseases affecting the central nervous system. MELAS can affect multiple systems, with a wide range of clinical manifestations including stroke, recurrent headaches, epilepsy, hearing impairment, myopathy, dementia, ataxia and diabetes 3 . Almost 80% of MELAS patients are caused by the m.3243A > G mutation in the MT-TL1 gene 4,5 . The mutation disturbs mitochondrial protein synthesis and ultimately affect the assembly of Complexes I, III and IV on oxidative respiration chains 6 . The molecular mechanism underlying the pathogenesis of MELAS is not fully understood. Finding out more effective diagnostic methods, and fully understanding the pathogenesis at molecular level, can prevent and reduce the complications of MELAS in a more effective manner.
Mitochondrial proteome is co-encoded by mtDNA and nuclear genome. The mtDNA only encodes 13 OXPHOS proteins; the nuclear genome encodes the remaining roughly 1500 protein and these proteins are transferred to the mitochondria through complex import systems 7 . The mutations of mitochondrial diseases affect proteins with undetermined functions. These proteins might shed light on the pathogenesis of mitochondrial diseases. Finding these proteins could help us gain valuable insight into pathophysiology of mitochondrial diseases. Recently, the combination of high-throughput omics techniques and complex bioinformatics analysis have brought new hope for revealing the pathogenesis and therapeutic targets about mitochondrial diseases. Increasingly, the quantitative proteomics and bioinformatics analysis can improve our understanding to the role of post-translational modi cations 8 . Compared to data-dependent-acquisition (DDA) method, the use of data-independent acquisition (DIA) method can obtain more reliable results and require a smaller samples size. So, the DDA-based proteomics is gradually replaced by DIA-based proteomics. In recent years, as a new tool, DIA-based proteomics is increasingly applied to the study of various diseases 9 .
In this study, we performed nanospray liquid chromatography-mass spectrometry (LC-MS) based proteomic analysis in the DIA modes on skeletal muscle specimens from patients with MELAS. The purpose of this study was to reveal the differentially expressed proteins in skeletal muscle tissue. Subsequently, we performed bioinformatics analysis for potential signaling pathways in the hope of gaining a better understanding of the pathogenesis.

Characteristics of individuals and muscle biopsy
The diagnostic criteria of MELAS study committee in Japan were adopted in this study 10 . In this study, all subjects underwent muscle biopsy for diagnostic purposes. We selected the MELAS patients positive point mutation for m.3243 A > G in genetic screening. We selected the age-and gender-matched subjects who were ultimately deemed to be free of neuromuscular diseases through muscle biopsies as controls. All collecting samples of muscle biopsies were informed consent of patients and controls. All procedures were approved by the Ethics committee of First Hospital of Shanxi Medical University (Taiyuan, China) and carried out in accordance with the principles of Helsinki Declaration.

Mitochondrial sequencing analysis
The entire mitochondrial genome was isolated and mitochondrial m.3243A > G mutation was detected by nextgeneration sequencing (NGS). Brie y, DNA was isolated from 200 mg muscle tissue by use of a DNA Kit CWE9600 (Beijing ComWin Biotech Co., Ltd), and mtDNA was ampli ed by long-range PCR using speci c primer. Next-generation sequencing libraries were constructed by KAPA LTP Library Preparation Kit (KK8234, VWR, USA), and ampli ed libraries were used in Illumina NovaSeq High-throughput sequencing system (Illumina, USA). Sequenced data was assessed to be quali ed by Sequencing Control Software (Illumina, USA) and analyzed by bioinformatic analysis referring human mitochondrial genome database. Mitochondrial m.3243A > G mutation was later con rmed by Sanger Sequencing using speci c DNA primers anking MT-TL1 gene.

Protein Digestion
100 µg of protein from each sample was solubilized in a new Eppendorf tube containing 8 M urea to achieve a nal volume of 100 µL. Then, 2 µL of 0.5 M TCEP was added for incubation for 1 h at 37℃. After 40 minutes of incubation with 4 µL of 1 M iodoacetamide in the dark at room temperature, samples were then precipitated by addition of -20℃ pre-chilled acetone at the ration of 1:5 at -20℃ overnight and centrifuged for 20 min at 12,000 G. Precipitates were washed and centrifuged again treated with 1 mL pre-chilled 90% acetone aqueous solution. Samples were re-dissolved in 100µL 100 mM TEAB and digested by Sequence grade modi ed trypsin (Promega, Madison, WI) containing 1:50 enzyme : protein (weight : weight) at 37℃ overnight. The peptide mixture was desalted by C18 ZipTip followed by quanti cation with Pierce™ Quantitative Colorimetric Peptide Assay (23275). SpeedVac was used for lyophilization.

High PH reverse phase separation
The mixed peptides were re-dissolved in buffer A (buffer A: 20 mM ammonium formate in water, pH 10.0, adjusted with ammonium hydroxide) and then loaded onto a reverse phase column (XBridge C18 column, 4.6 mm x 250 mm, 5 µm, Waters Corporation, MA, USA) using Ultimate 3000 HPLC system (Thermo Fisher scienti c, MA, USA). High pH separation was achieved in 40 min with a linear gradient starting from 5% B to 45% B (B: 20 mM ammonium formate in 80% ACN, pH 10.0, adjusted with ammonium hydroxide). The column was re-equilibrated at 30℃ for 15 min ( ow rate: 1 mL/min). A total of 10 fractions were obtained and dried by vacuum concentrator for future use.

nano-HPLC-MS/MS Analysis
The peptides were re-dissolved in solvent A (A: 0.1% formic acid in water) and on-line nanospray LC-MS/MS was performed for analysis on an Orbitrap Lumos coupled to EASY-nLC 1200 system (Thermo Fisher Scienti c, MA, USA). 4 µL peptide sample was loaded onto analytical column (Acclaim PepMap C18, 75 µm x 25 cm) and eluted with 120-min gradient from 4-32% B (B: 0.1% formic acid in ACN). We set the column ow rate at 300 nL/min and the capillary voltage at 2 kV. We run the mass spectrometer under DDA mode with MS and MS/MS mode automatically switched. The parameters were as follows: (1) MS: scan range (m/z) = 350-1500; resolution = 60,000; AGC target = 4e5; maximum injection time = 50 ms; dynamic exclusion = 30 s; (2) HCD-MS/MS: resolution = 15,000; AGC target = 5e4; maximum injection time = 38 ms; collision energy = 32.

Search the database
All Raw data were processed and analyzed by Spectronaut X (Biognosys AG, Switzerland) with default settings. Trypsin was assumed as the digestion enzyme with Spectronaut set up to search the database of swissprot_human_201902. The parameters were as follows: carbamidomethyl (C) was set as the xed modi cation, while oxidation (M) was speci ed as the variable modi cations. Q value (FDR) cutoff on precursor and protein level was set to 1%. Ultimately, an initial target list of 53,148 precursors, 37,666 peptides, 4,415 proteins and 4,301 protein group was obtained.

Data Analysis
We used Spectronaut X (Biognosys AG, Switzerland) with default settings to process and analyze the raw Data of DIA. Retention time prediction type was set to dynamic iRT and automatically determine the ideal extraction window. The criteria for protein identi cation were as follows: Q value (FDR) cutoff on precursor and protein level were both applied to 1%. The Decoy database was generated using a mutated strategy similar to scrambling a random number of amino acids (min = 2, max = length/2). All selected precursors that meet the screening conditions were used for quanti cation. All interfering fragment ions were excluded by MS2 interference unless less than 3 were found. The average value of the peak areas of the rst 3 peptides less than 1.0% FDR was used to quantify the protein group. Different expressed proteins were ltered if their p value < 0.05 and fold change > 1.3 with Welch's ANOVA Test performed.

Bioinformatics Analysis
The partial least-squares-discriminant analysis (PLS-DA) is a classic PLS regression for solving discrimination and classi cation problems. Relevant analysis was performed by mixOmics package (https://CRAN.Rproject.org/package=mixOmics). Hierarchical cluster analysis is an unsupervised method of clustering algorithms. The pheatmap package (https://CRAN.R-project.org/ package = pheatmap) was used in this analysis. The volcano plot is a kind of scatter-plot used to visualize changes in a data set within a given comparison. The X-axis and Y-axis represents the fold change and signi cance respectively. The ggplot2 package (http://ggplot2.org) was used in this study. Functional annotation was performed by Blast2GO version 5 and gene ontology (GO) analysis was performed using GOATOOLS 11,12 . Kyoto Encyclopedia of Genes and Genomes (KEGG) pathway enrichment analysis was performed by KOBAS (http://kobas.cbi.pku.edu.cn/) 13 .

Clinical data
We recruited 10 patients (5 males, 5 females) with de nitive MELAS from specialist clinics according to the diagnostic criteria in our hospital. The median age of the patients was 38.5 years (range 11-52). A tabulated description of clinical features of MELAS patients was shown in Table 1. The common symptoms were seizures and stroke. Other symptoms included headache, hearing loss, ataxia and weakness. The lactate levels was higher than 2 mmol/L in plasma at rest in 8 patients. The mutation heteroplasmy ranged from 56-89% in muscle tissue. These controls were healthy individuals without neuromuscular diseases and examined by the authors. A total of 10 controls (5 males, 5 females) with the median age of 39.5 years (range 12-53) were matched with patients for age and gender.

Muscle biopsy
The patients: multiple RRFs were found upon HE (Fig. 1A) and MGT staining (Fig. 1B). Some bers exhibited elevated levels of succinate dehydrogenase even ragged blue bers (RBFs) were found upon SDH staining (Fig. 1C). SDH/COX double staining revealed that SDH was strongly positive of COX negative/decreased bers (Fig. 1D). There were unremarkable ndings upon ATP staining, oil red O (ORO) staining and periodic acidschiff (PAS) staining.
The controls: no obvious abnormality were found in the muscle biopsy upon HE, MGT, SDH, COX, ATP, PAS and ORO staining.

Mutational analysis
The mitochondrial m.3243A > G mutation was found by NGS and later con rmed by conventional Sanger Sequencing in MELAS patients (Fig. 2a). This mutation was absent in the controls by Sanger Sequencing (Fig. 2b).

Identi cation of the differences between MELAS and controls
The PLS-DA model showed obvious difference, indicating a good class difference between the two groups (Fig. 3). The Volcano plots of differential proteins revealed 128 differential proteins (Supplementary 1) were ultimately obtained between MELAS and controls, including 68 down-regulated proteins and 60 up-regulated proteins (Fig. 4). Heat map analysis exhibited levels of differential proteins in every samples and provided information on the underlying metabolic disorders caused by m.3243A > G in muscle (Fig. 5). These results suggested that although donor variation existed, the differential proteins were generally MELAS-relevant.

Proteins signature of glycometabolism
We studied the expression of proteins involved in glycogen synthesis (Glycogenin-1, GYG1), glycolysis (glyceraldehyde-3-phosphate dehydrogenase, GAPDH; enolase 1, ENO1 and lactate dehydrogenase A, LDHA) and glycogenolysis (muscle isoform of glycogen phosphorylase, PYGM). The GYG1 expression was signi cantly increased in MELAS patients compared with controls. Interestingly, there were no differences in the other proteins signature between the two groups in Table 2.

The mitochondrial structural proteins
Subsequently, we analyzed the differences in Hsp60 (a mitochondrial structural protein) and mitochondrial inner membrane protein (calcium uptake protein 2, MICU2 and choline dehydrogenase, CHDH). We also focused on the expression of the enzymes related to the decarboxylation of pyruvate (pyruvate dehydrogenase, PDH-E1α), tricarboxylic acid cycle (citrate synthase, CS), oxidation of fatty acids (hydroxyacyl-CoA dehydrogenase/3-ketoacyl-CoA thiolase/enoyl-CoA hydratase, alpha subunit, HADHA) and mitochondrial translation (pentatricopeptide repeat domain-containing protein 3, PTCD3). No differences in expression of Hsp60, PDH-E1α and CS were observed between the two groups (Table 3). Interestingly, the expression of HADHA was increased but CHDH was decreased when compared to controls. The GYG1 expression was signi cantly increased in MELAS patients. The PTCD3 and MICU2 were not expressed at all in MELAS patients (Table 3).

Proteins involved in the oxidative stress and antioxidant defense
Finally, we analyzed the differences in proteins involved in oxidative stress: glucose-6-phosphate dehydrogenase (G6PD), selenoprotein P (SEPP1), heat shock protein beta-1 (HSPB1), alpha-crystallin B chain (CRYAB) and heme oxygenase 1 (HMOX1). At the same time, we focused on the differences in proteins associated with antioxidant defense: superoxide dismutase 1 (SOD1), superoxide dismutase 2 (SOD2) and protein S100-A9 (S100A9). The results indicated that the G6PD expression in the skeletal muscle was obviously reduced in MELAS patients. The SEPP1 was not expressed at all in MELAS patients. On the contrary, the expression of HSPB1 and CRYAB were obviously up-regulated in patients. The S100A9 was up-regulated in patients when compared to controls. The HMOX1 was not expressed at all in controls. There were no differences in the expression of SOD1 and SOD2 between the two groups (Table 4).

Gene ontology (GO) analysis
The 128 identi ed differential proteins in the two groups were categorized into three groups: CC, MF and BP. The intracellular part, membrane-bounded organelle and extracellular organelle were the three most enriched CC (Fig. 6A). The results indicated that protein binding, iron binding and lipid binding were the three most enriched MF (Fig. 6B). The response to stress, response to chemical, response to external stimulus, immune response and interspecies interaction between organisms were the most enriched BP (Fig. 6C).

KEGG pathway enrichment and PPI network analysis
We performed KEGG pathway enrichment analysis to deepen the understanding to the functions of the differential proteins. We observed that these pathways, including hematopoietic cell lineage, Epstein-Barr virus infection, phagosome, systemic lupus erythematosus, proliferator-activated receptors (PPAR) signaling pathway, ribosome, viral myocarditis and vitamin B6 metabolism showed signi cant enrichment (Fig. 7). We constructed PPI networks by using the STRING database to help us better understand the interactions of differential proteins. A total of 128 proteins were used as search inputs, 93 of which were matched in the database, including 43 up-regulated proteins and 50 down-regulated proteins. The PPI network of differential proteins was performed as shown in Figs

Discussion
So far, this is the rst proteomes analysis of skeletal muscle collected from m.3243A > G MELAS patients compared with controls using the method of the DIA. We had found 128 differential proteins, including 68 proteins for down-regulation and 60 proteins for up-regulation. We studied the expression involved in oxidative stress and found a highly signi cant up-regulation including HSPB1, CRYAB, HMOX1 but a decrease in G6PD and SEPP1. The most key proteins were HSPB1 and CRYAB, which were small heat shock proteins (sHsp) that combined misfolded proteins to prevent them from denaturation, inhibited cellular apoptosis and regulated the intracellular redox state as molecular chaperones 15,16 . In mutant HSPB1 expressing motor axons, the researchers observed the decreased mitochondrial Complex I activity and increased mitochondrial vulnerability, resulted in increased superoxide release and decreased mitochondrial glutathione levels 17 . The CRYAB could protect cells from hypoxia and maintain mitochondrial integrity 18,19 and also had other roles in vascular biology 20 . The CRYAB was crucial for endothelial cell survival in hypoxia 19,21,22 . sHsps could attenuate mitochondrial dysfunctions, block oxidative stress and minimize neuronal apoptosis, so sHsps were promising protectants in some neurodegenerative diseases 23 . G6PD was the rst rate-limiting enzyme in the pentose phosphate pathway which was necessary to maintain oxidation-reduction equilibrium in cells 24 . There were some studies found that sHsp signi cantly increased G6PD activity and played the role of anti-oxidative stress through a G6PD-dependent ability 16 . Our results showed that G6PD was down-regulated, different from another study 25 . We considered that there were some reasons: rstly, the pathogenesis of MELAS was extremely complex; secondly, the muscle tissue only provided the eeting information about a certain stage in the course of MELAS; thirdly, sHsps played the anti-oxidative stress role through other pathways rather than regulating G6PD-dependent ability. In our study, the HMOX1 was highly signi cant up-regulation in MELAS but not expression at all in controls. The HMOX-1 over-expression associated with intracellular oxidative stress, mitochondrial dysfunction and mitophagy 26 . Increased levels of HMOX1 had been observed in many neurodegenerative diseases including Alzheimer's, Parkinson's diseases 26 . Thus, we speculated that these oxidative-related proteins might be associated to the development and severity of clinical symptoms of MELAS.
We had discovered these pathways, including phagosome, PPAR signaling pathway, ribosome and vitamin B6 metabolism showed signi cant enrichment. There was a study that the authors had found accumulation of autophagosomes in MELAS broblasts 27 . The enhancement of autophagosome content might be an adaptive response to intracellular OXPHOS and mitochondrial de ciency. As compensatory response, the increased antophagosome could clear damaged mitochondria to avoid further cellular dysfunction 28,29 . On the contrary, the mitochondrial dysfunction blocked both autophagy induction and autophagic ux 30 . So it was controversial whether autophagy played protective or harmful effects. Defects in phagosome was related to a variety of cellular metabolic process, including abnormal lipid metabolism, oxidative stress injure and mitochondrial dysfunction 31-33 but autophagosome accumulation could impair cell bioenergetics and also induce cell death 34 . We proposed that the response to autophagosome varied widely, depending on the severity of mitochondrial de ciency. The phagosome pathway may be one of the improtant molecular networks involved in the pathogenesis of MELAS.
PPAR signaling pathway is a crucial regulator for the β oxidation of fatty acid 35,36 and promote transcription of nuclear-encoded mitochondrial genes. Beza brate, as a pan-PPAR activator, could up-regulate the expression of Peroxisome proliferator-activated receptor gamma coactivator 1-alpha (PGC-1α), which promoted mitochondrial biogenesis and improved mitochondrial de ciency of mitochondrial diseases 37 . The PPAR signaling pathway may be a potentially disease-modifying therapeutic target.
The mitochondrial ribosome proteins (MRPs) were synthesized in the cytoplasm, introduced into mitochondria for assembly and then responsible for the translation of 13 mitochondrial mRNAs 38 . In our study, MRPL22, MRPL18 were up-regulated and MRPL4, PTCD3 were down-regulated. MRPL18 facilitates ribosome assembly to participate in selected mRNAs translation and stress regulation to achieve a strong cytoplasmic stress response under stress conditions 39 . So MRPL18 is one of a key regulator to cytoplasmic stress response. PTCD3, also known as MRPS39, is one of the mitochondrial ribosomal supernumerary proteins unique to mammals 40 . A study of PTCD3 knockdown showed protein synthesis was seriously disturbed to result in OXPHOS de ciency, although no effect on RNA metabolism 41 . PTCD3 mutations led to impaired translation of mtDNA-encoded proteins, resulting in combined defects of Complex I and IV, and decreased ATP production 42 . Although MRPL4 had not been reported in the mitochondrial diseases, this molecule had attracted increasing attention from researchers because it had been reported to be downstream of hypoxia-inducible factor-1α (HIF-1α) 43 . From the above, we speculated that the MRPs played a key role in the occurrence, development and prognosis of MELAS.
Of course, this study also had certain limitations: rst of all, limited by the sample size, the results could not accurately re ect the complete picture of MELAS. The Second, we tried to minimize the individual differences, but the skeletal muscle only provided eeting information about a certain stage in the MELAS process. The third, the DIA protein library was established on the basis of the DDA atlas database and the might lead to the loss of some peptides in the process of analysis. Therefore, the larger sample size and more quantitative studies were needed to verify our results.

Conclusion
Our results indicated that a large number of differential proteins were repeatedly recognized in skeletal muscle of MELAS patients. In addition, the results revealed that the imbalance between oxidative stress and antioxidant defense, activation of autophagosomes and abnormal metabolism of mitochondrial ribosome proteins played a critical role in m.3243A > G MELAS patients. Finally, further investigation of these proteins could contribute novel insights to the pathogenesis of MELAS in a comprehensive manner.

Declarations
Ethics approval and consent to participate All collecting samples of muscle biopsies were informed consent of patients and controls. All procedures were approved by the Ethics committee of First Hospital of Shanxi Medical University (Taiyuan, China) and carried out in accordance with the principles of Helsinki Declaration.
provided some muscle tissue samples. Juan Wang, Jing Zhang and Wenqu Yang interpreted the results and revised the manuscript.

Consent for publication
All authors read and approved the nal version of this manuscript for submission.