A systematic review of mitochondrial abnormalities in myalgic encephalomyelitis/chronic fatigue syndrome/systemic exertion intolerance disease

Patients with Myalgic Encephalomyelitis/Chronic Fatigue Syndrome (ME/CFS) or Systemic Exertion Intolerance Disease (SEID) present with a constellation of symptoms including debilitating fatigue that is unrelieved by rest. The pathomechanisms underlying this illness are not fully understood and the search for a biomarker continues, mitochondrial aberrations have been suggested as a possible candidate. The aim of this systematic review is to collate and appraise current literature on mitochondrial changes in ME/CFS/SEID patients compared to healthy controls. Embase, PubMed, Scopus and Medline (EBSCO host) were systematically searched for articles assessing mitochondrial changes in ME/CFS/SEID patients compared to healthy controls published between January 1995 and February 2020. The list of articles was further refined using specific inclusion and exclusion criteria. Quality and bias were measured using the Joanna Briggs Institute Critical Appraisal Checklist for Case Control Studies. Nineteen studies were included in this review. The included studies investigated mitochondrial structural and functional differences in ME/CFS/SEID patients compared with healthy controls. Outcomes addressed by the papers include changes in mitochondrial structure, deoxyribonucleic acid/ribonucleic acid, respiratory function, metabolites, and coenzymes. Based on the included articles in the review it is difficult to establish the role of mitochondria in the pathomechanisms of ME/CFS/SEID due to inconsistencies across the studies. Future well-designed studies using the same ME/CFS/SEID diagnostic criteria and analysis methods are required to determine possible mitochondrial involvement in the pathomechanisms of ME/CFS/SEID.

Page 2 of 16 Holden et al. J Transl Med (2020) 18:290 presentation is heterogeneous ranging from mild to severe, even leaving some patients bed bound [1]. The underlying pathomechanisms of ME/CFS/SEID are nebulous and the search for standardised biomarkers continues, so diagnosis entirely depends upon symptom specific case criteria following the exclusion of any other explanatory diagnosis [1][2][3][4].

Literature search
This systematic review was conducted according to the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) (Fig. 1) and Cochrane guidelines. PRISMA and Cochrane guidelines were used to ensure international standards were maintained and used for reporting information contained in this systematic review. The databases EMBASE, PubMed, Scopus and Medline (EBSCO host) were systematically searched using full-text and Medical Subject Headings (MeSH) terms. Mitochondrial search terms and ME/CFS/SEID search terms are presented in Table 1. Boolean operators 'OR' and ' AND' were used to expand the search to include all relevant key terms and to specify articles containing both a ME/CFS/SEID search term and a mitochondrial search term. Full code can be found in Additional File 1. Literature searches were conducted independently by authors SH and RM on February 18th, 2020. Reference list checking and citation searching was carried out, no additional papers were found. Unpublished literature was not searched. No additional papers were identified in the final search or through alternative databases such as Griffith University institute library or Google Scholar.

Inclusion and exclusion criteria
Studies were included for review if at least one mitochondrial search term AND at least one ME/CFS/SEID search term (Table 1) were found in the title or abstract and the study complied with the following inclusion criteria: (i) published after 1994 (ii) research conducted on human participants only, aged 18 years or older; (iii) full-text article was available in the English language; (iv) original research only was reported; (v) ME/CFS/SEID was diagnosed using: FC (1994), CCC (2003), ICC (2011) or IOMC (2015); (vi) investigation was conducted on mitochondrial aberrations in ME/CFS/SEID patients compared to a HC group.
Articles that did not contain at least one mitochondrial search term AND at least one ME/CFS/SEID search term in the title or abstract were excluded from the review ( Table 1). Articles were excluded if any of the following applied: (i) written prior to the introduction of the FC on December 15th 1994 (all studies 1994 or earlier were excluded considering time was required to be aware of the FC); (ii) conducted in non-human participants or those under the age of 18; (iii) articles not written in English or not available as full-text; (iv) studies that reported on non-original data including: duplicate studies, case reports or review articles; (v) use of criteria other than FC, CCC, ICC or IOMC; (vi) Comparison with a patient group (e.g.,) fibromyalgia or depression, without comparison to HC participants; (vii) studies that were not within the scope of this review.

Selection of studies
Following retrieval of articles from the databases all articles were stored in the reference management software package Endnote X9.2. Duplicates were manually removed and all articles that did not contain the listed key words in title or abstract were omitted. The remaining articles were reviewed and those that followed the eligibility criteria were selected. This process was conducted by authors SH and RM independently. Final papers to be included in this review were then reassessed by all other listed authors.
During title and abstract screening, we performed manual and automated screening. The automated screening was achieved by building an algorithm in Endnote 9.2 using the grouping function with AND/OR logic gates. When we compared the manual screening result to automated screening, one article was retained by automated screening which manual screening had excluded, that article was included in final analysis. Two articles were found by manual screening which automated screening had excluded, both of which were included in final analysis. Authors SH and RM then checked discrepancies, the final list of articles retained after screening was deemed accordant and correct by both authors.

Data extraction
Following selection of papers, data was extracted including: (i) study design; (ii) diagnostic criteria used; (iii) sample size; (iv) method of analysis.

Quality assessment
The Joanna Briggs Institute (JBI) Critical Appraisal Checklist for Case Control Studies (CACCCS) checklist was used to evaluate quality and bias (Additional File 2). CACCCS checklist was selected based on it being an internationally recognized, validated, rigorous process to evaluate study quality and bias. Quality assessment was conducted by authors SH and RM independently. Each checklist item assesses the following: [1] group matching, [2] source population, [3] criteria, [4] method of exposure, [5] assessment of exposure, [6] confounding variables identification, [7] mitigation of confounding variables, [8] measurement of outcomes, [9] exposure period selection, [10] statistical analysis. Items four, five and nine of the JBI CACCCS were not assessed as intervention-based studies were excluded from analysis.

Literature reporting changes in mitochondrial deoxyribonucleic acid (DNA)/ribonucleic acid (RNA)
One study found no significant single nucleotide polymorphisms (SNPs) between ME/CFS/SEID patients and HC participants. There, however, was a positive correlation between haplotypes J, U and H as well as eight other SNPs and symptoms, symptom clusters or symptom severity in ME/CFS/SEID patients. Heteroplasmy frequency was low [6].
Another study assessed DNA variants in patient groups and HC participants from two distinct locations: South Africa and the United Kingdom. Similarly, most patients with severe or moderate ME/CFS did not have a mildly deleterious population variant [20].
A study on RNA from PBMCs found that there were significantly increased gene transcripts important for mitochondrial function in ME/CFS/SEID patients compared with HC participants including PMAIP1, PMPCB and JUN [18].

Literature Reporting structural changes in mitochondria
Structural abnormalities relating to mitochondria in ME/ CFS/SEID patients including sub-sarcolemmal mitochondrial aggregates, intermyofibrillar mitochondrial aggregates, mitochondrial circumference, mitochondrial area, mitochondrial pleomorphism, or compartmentalization of the inner mitochondrial membrane were examined in one of the studies included in this review. That study did not identify any significant structural changes in ME/CFS/SEID patient mitochondria compared with HC participants [16].

Literature reporting changes in mitochondrial respiratory function
Mitochondrial respiratory function was investigated in five of the included studies. There were four studies that identified significant differences in mitochondrial respiratory function between ME/CFS/SEID patients and HC participants. One of these studies found that there was lower mitochondrial membrane potential, lower proton leak (oligomycin, ATP synthase inhibitor, resulting in the depletion of proton motive force), lower ATP production in CD8 + T cells from ME/CFS/SEID patients compared with HC participants, also lower glycolysis at rest in both CD8 + and CD4 + cells from ME/CFS/SEID patients [12,27].
Another study [22] examining lymphoblasts also showed lower mitochondrial membrane potential in ME/ CFS/SEID patients compared to HC participants. There was also lowered activation of ATP synthesis by complex V and hyperactivated target of rapamycin (TOR) complex 1 stress signaling. In contrast to the aforementioned study, in Mandarano et al's study there was greater proton leak, greater complex 1 oxygen consumption rate (OCR), greater maximum OCR (OCR is an indicator of cellular metabolism and fitness) and greater spare respiratory capacity (excess respiratory electron transport chain capacity not being used in basal respiration) [12,22,28]. Additionally, in Mandarano et al's study there was also greater nonmitochondrial OCR (oxygen consuming process) activity, greater number of enzymes of β-oxidation and greater tricarboxylic acid cycle (TCA) activity [12].
The Missailidis et al. study also found no difference in mitochondrial mass, genome copy number, glycolytic rates, or steady state ATP levels between ME/CFS/SEID patients and HC participants [22]. These findings were validated in an associated study that found cells belonging to ME/CFS/SEID patients can be differentiated from HC participant cells based on respiratory function parameters [23].
One study examining PBMCs [19] found all oxidative phosphorylation parameters including basal respiration, ATP production, proton leak, maximal respiration, reserve capacity, non-mitochondrial respiration, and coupling efficiency were significantly lower in ME/CFS/ SEID patients compared to HC participants. Similarly to the Missailidis et al. study there was no significant difference in glycolytic activity between both groups [22]. One study contrasted this by finding no difference in mitochondrial respiration in NK cells, however, glycolytic function was significantly lower in ME/CFS/SEID patients compared to HC participants [15].

Literature Reporting Changes in metabolites
From the 19 studies, 12 investigated metabolic changes in ME/CFS/SEID patients compared to HC participants [5, 7-9, 12, 13, 15, 17, 19, 21-23]. One study investigating metabolic changes in cerebral spinal fluid found no differences in levels of high energy phosphate metabolites including ATP, creatine phosphate, and inorganic phosphate between ME/CFS/SEID patients and HC participants. This study used arterial spin labelling reporting higher ventricular lactate and lower glutathione in the ME/CFS/SEID patient group [17]. As previously mentioned, mitochondrial respiratory tests found that there was lower ATP production in three of the studies [8,12,19]. One study, however, reported that there was no difference in steady state ATP levels between groups [22]. An ATP profile study found that there was reduced ATP availability in neutrophils from ME/CFS/SEID patients in comparison to HC participant neutrophils [7]. This finding was supported by another study conducted on PBMCs [8].
A study investigating metabolite concentrations in tricarboxylic acid (TCA) and urea cycles found: ornithine, citrulline, pyruvate and isocitrate ratios were significantly higher in ME/CFS/SEID patients compared to HC participants [21]. Another study found that ME/CFS/SEID patients had higher blood glucose levels and lower blood lactate, urine pyruvate and urine alanine compared to HC participants [5]. In plasma, abnormalities in 20 out of 63 metabolic pathways were found in ME/CFS/SEID patients compared to HC participants: this includes purine, cholesterol, pyrroline-5-carboxylate, riboflavin, and branch chain amino acid [13].

Literature reporting changes in coenzymes
One study found significantly lower levels of CoQ 10 in PBMCs from ME/CFS/SEID patients compared to HC participants [8]. Another study supported this, finding FibroFatigue scale scores, fatigue, and autonomic symptoms negatively correlated with CoQ 10 levels [11].

Literature reporting changes in mitochondrial signaling pathways
One study investigated cytosol and mitochondrial calcium (Ca 2+ ) influx following stimulation with 2-aminoethoxydiphenyl borate (2-APB), and thapsigargin in ME/CFS/SEID patients compared to HC participants. There was significantly lower cytosolic Ca 2+ ion concentration in CD19 + B lymphocytes and CD56 bright NK cells in ME/CFS patients compared to HC participants. In ME/CFS/SEID patients, however, there was no significant difference in mitochondrial Ca 2+ concentration compared to HC participants [14].

Quality assessment
The JBI CACCCS checklist was used to assess each article for quality and bias (Additional File 2). Items four, five and nine were omitted due to exclusion of interventional studies from this review. Item eight was the most addressed item where 19 (100%) of included studies assessed outcomes in a standard, valid, reliable way. Secondly, 15 (78.9%) of included studies selected appropriate criteria for ME/CFS/SEID and HC participants. From all included studies, 14 (73.7%) appropriately matched patient and HC groups (item one), identified confounding factors (item six), and utilized appropriate statistical tests (item 10). Identified confounding factors were effectively controlled for in 11 (57.9%) of the studies. Checklist item two was least addressed, only 10 studies (52.6%) adequately matched source population.

Discussion
The aim of this systematic review was to collect and analyze current research on the role of mitochondria in ME/CFS/SEID pathomechanisms. A total of 19 studies met the inclusion criteria and were included in this review. Changes in mitochondrial structure, DNA/ RNA, respiratory function, metabolites, and coenzymes have been reported. The results from this systematic review indicate a significant amount of variability across all studies, where in many cases standardized protocols were not apparent or similar outcomes were not assessed, thus making comparisons between studies difficult and ultimately providing little consistency for the analysis of ME/CFS/SEID pathomechanisms. A systematic review on mitochondrial dysfunction and fatigue was previously published by Filler et al. (2014). That paper used the keyword "fatigue", however most of the included articles were reporting on ME/CFS/SEID. Since then, numerous papers have been published in the ME/CFS/SEID field. The Filler et al. paper also had less stringent exclusion criteria than this review, included articles that did not have HC participants, had participants under the age of 18 and used criteria beyond the three main established FC, CCC and ICC [29]. The IOMC was not yet established when Filler et al. published in 2014 and has been incorporated to this review. To ensure comparability of the studies examined in this review, stricter exclusion criteria were put in place. This systematic review was the first to assess mitochondrial changes in case control cohorts.
A novel computerized title and abstract screening filter was programmed and implemented in this study by SH. The process of title and abstract screening was made more robust by comparing the automated output against the result from manual title and abstract screening as an additional process verification step, incorporating the advantages of computer processing in a way which did not compromise manual screening. A logically correct computerized title and abstract filter might provide an additional verification layer to reduce human error during title and abstract screening in future studies. Computerized screening may be ideal if the screening process exclusively consists of predefined logical steps without interim decision making, provided the computing architecture is properly designed and operates without error.
In this review, the average ages for ME/CFS/SEID patients and HC participants were 43.7 and 42.4, respectively. Most of the participants were female (77%), this is consistent with literature stating the illness most commonly affects females aged between 35 and 45 [15,30,31]. Selected studies matched participants by age and sex [5, 6, 8, 9, 11-13, 15-18, 22, 23]. Matching these characteristics in mitochondrial studies is essential to account for age-related mitochondrial decline and sex-associated differences including uncoupled respiration, citrate synthase activity and ATP levels [32]. Six of the studies included data on race or ethnicity, reporting the majority Page 13 of 16 Holden et al. J Transl Med (2020) 18:290 of participants as Caucasian; this feature is also consistent with literature [15,30,31]. The predominant use of the FC across the studies is a significant limitation considering the broad nature of this criteria and considerable overlap with other illnesses [3]. This may explain some inconsistencies found between studies. With the more recent publications, a greater number of papers incorporated the later, more stringent CCC, ICC or IOMC into their recruitment decisions [5,9,12,13,15,18,22,23]. The inclusion of more stringent criteria allows for a more homogenous subset of patients [3].
Analysis was conducted on samples from a variety of different sources including urine, whole blood, plasma, PBMCs, NK cells, B cells, and T cells. The selection of immune cells resulted from published data demonstrating immune involvement in ME/CFS/SEID pathology [12]. The different cell types make it difficult to establish comparisons between the studies. Cells were from different sources (frozen and fresh), further complicating comparison. For experiments such as measuring mitochondrial respiration, the cellular stress arising from the freezing process significantly impacted cellular bioenergetics including ATP production, maximal respiration, and reserve capacity of the PBMCs in both groups [19].
Only one study included in this review reported on ultrastructural aberrations. These included: extramitochondrial aggregates, mitochondrial circumference, mitochondrial area, and mitochondrial pleomorphism in ME/CFS/SEID patients compared to HC participants [16]. Reports not included in this review described mitochondrial changes in 70% of ME/CFS/SEID patients, involving changes in mitochondrial size as well as cristae branching and cristae fusion resulting in a compartmentalized appearance [33,34]. Plioplys and Plioplys used similar protocol, differing by criteria used to define ME/ CFS/SEID patients, yet reported contradictory findings. This difference may have arisen as a result of inconsistencies in quantification methods [16].
All included DNA studies in this review suggested that ME/CFS/SEID patients do not exhibit disease causing variants [20]. This finding was consistent across both moderately and severely affected ME/CFS patient groups. Additionally, there was no difference found in DNA copy number. A limitation to our search strategy was that some potentially relevant articles were excluded based on not having a HC group. Further analysis of those excluded articles supported these findings (e.g., [35]). These findings suggest that ME/CFS/SEID is not a primary mitochondrial disorder. Altered leukocyte gene expression levels have been identified in fatigue-related pathways [10]. Transcriptome analyses found significantly increased transcription of three genes in ME/CFS/ SEID patients: PMAIP1, PMPCB and JUN [18]. These genes have an important role in mitochondrial function and apoptosis, and may contribute to the neuroinflammatory processes, oxidative stress, increased ventricular lactate, imbalanced metabolites, disrupted circadian rhythm, and impaired respiratory function described across multiple studies [5,12,13,17,18,[21][22][23]35].
Results from respiratory function and glycolysis studies lacked consistency. Interestingly, two papers reported decreased proton leak with decreased mitochondrial efficiency parameters in ME/CFS/SEID patients [12,19]. An increase in proton leak would usually correspond to decreased mitochondrial efficiency, as observed in two other studies by Missailidis et al. [12,22,23,36]. Missailidis et al. showed that respiratory function parameters can be part of an effective method to distinguish between ME/CFS/SEID and HC participant cells, however these findings require further validation in a larger cohort [23]. It is difficult to make reliable conclusions based on the glycolytic activity and respiratory function described in these featured studies, further research is required.
Nguyen et al. found no significant mitochondrial Ca 2+ concentration changes in the presence of stimulants [14]. That study, however, reported a reduction of cytoplasmic Ca 2+ concentration in CD19 + B lymphocytes and CD56 bright NK cells in the presence of stimulants. Mitochondrial processes including respiratory function are Ca 2+ -dependent, cytosolic Ca 2+ levels influence uptake by mitochondria through Ca 2+ -dependent channels [37,38]. Disruption of Ca 2+ channel function, specifically transient receptor potential melastatin 3, has been implicated in ME/CFS/SEID patient NK cell pathology resulting in decreased Ca 2+ mobilization [39,40]. As Ca 2+ is fundamental for many NK cell processes including cytotoxicity, NK cell function is consequently impaired [39,40]. This impairment may aggravate the production of reactive oxygen species and contribute to the decline of mitochondrial processes, both observed in other studies [41]. Impaired NK cell cytotoxicity is the most consistently described feature in ME/CFS/SEID [42]. Mitochondrial dysfunction may therefore be a consequence, rather than a primary causative factor, in ME/CFS/SEID [44].
Twelve of the studies investigated changes in metabolites including ATP, TCA, and urea cycles [5, 7-9, 12, 13, 15, 17, 19, 21-23]. In this review, discrepancies were observed in ATP levels. The reduction of ATP levels was observed in three studies [8,12,19]. Three studies reported no changes in ATP levels [9,17,22]. Another study not included in the final review found an increase in externally-derived mitochondria [43]. Missailidis et al. reported that ATP levels were able to be maintained despite inefficiency of Complex V, due to signaling networks being able to homeostatically respond to cellular stresses [17]. Conditions might account for varied results, for example cellular glucose levels [19]. Citrulline, pyruvate, and isocitrate ratios were significantly higher in ME/CFS/SEID patients compared to HC participants. Succinate (which is downstream of the other metabolites) does not significantly differ in level between ME/CFS/SEID patients and HC participants. Disruption of early stages of the TCA cycle is suggested to be present in ME/CFS/SEID patients [21]. Abnormalities in 20 out of 63 metabolic pathways that were tested included: purine, cholesterol, and pyrroline-5-carboxylate. It has been described that these differences make ME/CFS/SEID patients chemically distinguishable from respective HC participants [13]. This is only a small representation of available metabolite studies, a detailed investigation on metabolomic dysregulation in ME/CFS/SEID patients compared to healthy controls has been conducted by Huth et al. [44].
The two studies that investigated CoQ 10 found a decrease of this compound in ME/CFS/SEID patients compared to HC participants [8,11]. This reduction could result in an upregulation of nuclear factor kappalight-chain-enhancer in activated B cells (NFκB) and also induction of oxidative and nitrosative stress pathways [11]. An interventional study has shown potential benefits of CoQ 10 plus nicotinamide adenine dinucleotide supplementation, finding reduced maximum heart rate and perceptions of fatigue post exercise testing in ME/CFS/SEID patients [45]. Interventional studies, however, were not in the scope of this review.
An article published by Gorman et al. identified overlapping features in classic forms of mitochondrial disease and ME/CFS/SEID, perceived fatigue was a particular feature [46]. Molecular analysis of mitochondrial dysfunction in ME/CFS/SEID has not identified any characteristic mitochondrial gene variants in mitochondrial disease [20]. Smits et al. compared mitochondrial respiratory chain complex activity between ME/CFS/SEID, known mitochondrial disorder and HC participants [47]. Due to the inclusion of inappropriate HC participants, that paper was not included in the final review, it identified distinct differences in ATP production rate and respiratory chain complex activity between ME/CFS/SEID and known mitochondrial disorder participants [47].
An additional study investigating the presence of autoreactive antibodies in ME/CFS/SEID patients has been released. That article, though following all our inclusion criteria, was not included in the final analysis because it was published after we screened for papers. Only one out of 161 ME/CFS/SEID patients were positive for anti-pyruvate dehydrogenase complex antibodies. Anti-mitochondrial antibodies in general were negative in ME/CFS/SEID populations. This research suggests that mitochondrial dysfunction in ME/CFS/ SEID patients cannot be explained by the presence of circulating anti-mitochondrial autoantibodies [48].
All included studies, while evidence of mitochondrial pathway disruption is present, are limited by small sample sizes and experiments are not standardized, having low reproducibility. Deriving appropriate conclusions is therefore difficult. Additionally, these studies are case-controlled so only reflect singular points in time. Longitudinal study design may be an important consideration for future studies as differences in mitochondrial function over time may relate to changes in disease activity, symptom presentation, and symptom severity [35].

Quality assessment
Quality levels varied across studies. All studies assessed outcomes in a standard, valid and reliable way (JBI CACCCS item eight) for ME/CFS/SEID and HC participants using a variety of different methods such as Seahorse, MitoTracker, nuclear magnetic resonance (NMR) spectroscopy, magnetic resonance spectroscopy, DNA/RNA sequencing and electron microscopy. JBI CACCCS item three was the second-most addressed checklist item, selection criteria used to define ME/ CFS/SEID and HC cohorts were provided and consistent between both groups. Absent criteria for HC participants was the main reason why some studies did not adhere to this item. JBI CACCCS items one, six and 10 followed. Groups were comparable (other than the presence of disease in ME/CFS/SEID participants and absence of disease in HC participants) through age and sex matching. Identified confounding factors included obesity, sampling time, medication and/or dietary supplements, smoking status, and shift work. Studies that adhered to JBI CACCCS item 10 conducted appropriate statistical analyses including adjusting for multiple comparisons where required and performing normality checks. Mitigation of confounding factors was variously achieved by inclusion of a sampling time range, exclusion criteria (such as shift work, obesity, and smoking status) and ceasing medications and/or dietary supplements for an advised duration prior to study participation. In some cases, studies identified confounding factors but did not adequately control for them. JBI CACCCS item two was least addressed, involving appropriate matching of ME/CFS/SEID patient source population to HC participants included in the study. In all these cases, source population information was not