A systematic review and meta‐analysis of studies that have evaluated the role of mitochondrial function and iron metabolism in frailty

Abstract Frailty is a condition of global impairment due to depletion of physiological reserves. However, the underlying biological mechanisms are poorly understood. The aims of the current study were to identify the differences in mitochondrial function and iron metabolism between frail and nonfrail populations, and to investigate the contribution of different methodological approaches to the results. Searches were performed, using five online databases up to November 2019. Studies reporting measurements of mitochondrial function or iron metabolism in frail and nonfrail subjects or subjects with and without sarcopenia, were included. Pooled effect estimates were expressed as Standardized Mean Differences. Heterogeneity, expressed as I 2, was explored using regression analyses. In total, 107 studies, reporting 75 measures of mitochondrial function or iron metabolism, using six different experimental approaches, in three species were identified. Significant decreases in measures of oxygen consumption were observed for frail humans but not in animal models. Conversely, no differences between frail and nonfrail humans were observed for apoptosis and autophagy, in contrast to animal models. The most significant effect of the type of frailty assessment was observed for respiratory chain complexes where only subjects categorized as frail by the Fried Frailty Index showed a significant decrease in activity. We identified iron metabolism in frailty as an important knowledge gap, highlighted the need of consistent frailty diagnostic tools, and pointed out the limited translational potential of animal models. Inconsistency between studies evaluating the molecular mechanisms underlying frailty may present a barrier to the development of effective therapies.


INTRODUCTION
Frailty is a clinical syndrome characterized by loss of cognitive function, sarcopenia, unintentional weight loss, and low energy. In clinical terms, this translates into increased vulnerability to otherwise minor stressors, that can lead to prolonged hospitalization, or the requirement for social care or even long-term convalescent care. 1 Frailty is chiefly associated with advancing age but can occur in younger people with additional risk factors, including sedentary lifestyle, poor nutrition, chronic disease, and chronic inflammatory states. As the population ages, and as the numbers of patients with multiple chronic conditions, or cardiometabolic disease, increases, frailty will present an increasing challenge to clinicians and health systems. The frailty phenotype is poorly defined, with over 25 frailty definitions in current clinical use. 2 This reflects the limited understanding of the underlying mechanisms. Existing research points toward a multifactorial pathophysiology characterized by loss of mitochondrial function in skeletal muscle, altered iron metabolism, and exposure to oxidative stress. 3,4 The primary aim of the current study was to summarize and critically review all published studies that have measured these processes in experimental or human studies of frailty. A secondary aim was to evaluate the strengths and limitations of different measures of mitochondrial function or iron metabolism with a view to application in future translational studies of the frailty phenotype.

MATERIALS AND METHODS
A systematic review of randomized controlled trials and animal studies was performed using the methods described in the Cochrane Handbook for Systematic Reviews of Interventions. 5 The study adhered to the Preferring Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) guidelines. 6

Study eligibility
We included studies reporting measurements of iron metabolism or mitochondrial function in frail and nonfrail subjects or subjects with and without sarcopenia, irrespective of blinding, date of publication, sample size, or race of subjects. Review studies, studies examining nonmammalian populations, and studies examining juvenile populations were excluded.

Information sources
PubMed, Cochrane library, Ovid Medline, Scopus, and BioRXiv databases were searched using variations of search terms: (frail* OR "frailty syndrome" OR "frail elderly" OR sarcopenia) AND (iron OR dmt1 OR ferroportin OR transferrin OR ferritin OR hepcidin OR "circulating iron" OR anemia OR mitochondri* OR gdf15 OR vimentin OR ldh). The final search was performed on November 25, 2019. A full description of the search terms is listed in the Table S1.

Study selection
Studies identified in online searches were managed using Endnote X9. Title, abstract, and full text screening were

WHAT QUESTION DID THIS STUDY ADDRESS?
In order to clarify the role of mitochondrial function and iron metabolism in frailty, we systematically reviewed and meta-analyzed all available data examining these processes in frail versus non-frail subjects.

WHAT DOES THIS STUDY ADD TO OUR KNOWLEDGE?
This review identified mitochondrial dynamics, oxygen consumption, and the activities and abundancies of mitochondrial respiratory complexes as main categories of mitochondrial function dysregulated in frailty, and we believe these provide a core dataset of outcomes for future research in this area.

HOW MIGHT THIS CHANGE CLINICAL PHARMACOLOGY OR TRANSLATIONAL SCIENCE?
This study also identified a significant difference between the dysregulated molecular pathways in animal models and human subjects, pointing out the minuscule translational potential of frailty research using current animal models of frailty. Furthermore, our findings strongly underline the need for more consistent frailty definition, as the heterogeneity of methodological approaches and frailty assessment tools were a barrier to gaining consistent and easily interpretable results. We believe that the findings and recommendations presented in this review provide novel insights into molecular mechanism of frailty and present a valuable resource for future translational studies.
carried out and excluded studies and the reason for exclusion were recorded.

Assessment of methodological quality and publication bias
The methodological quality and the risk of bias for human studies were assessed using a modified version of the Newcastle -Ottawa Quality Assessment Scale for Human Cohort Studies (Supplementary File S1). 7 For animal studies, the Animal Research: Reporting In Vivo Experiments (ARRIVE) checklist was used, as described previously. 8

Outcomes of interest
The prespecified coprimary outcomes or this review were measures of mitochondrial function and iron metabolism.

Data extraction
Data extraction was completed using a standardized pro forma as follows: authors, publication date, journal name, study title, research aim, main findings, species, and the method used for assessment of frailty. For each measured variable in a given study, the mean and the SD were extracted as well as the number of subjects in the frail and nonfrail groups. To extract the numerical data presented only graphically, the WebPlot Digitizer 4.2 software was utilized. 9 For studies assuming frailty based on age of the human/animal model, the oldest group was considered frail and the youngest adult group was considered the control/nonfrail group. Where human subjects/animal models were assessed as prefrail, these were classified as frail in the final analysis. Where a single study measured biomarkers of frailty in multiple tissue types, only measurements from one type of tissue was included, based on the most common tissue type assessed across all studies investigating this outcome. Studies reporting only median were excluded from the analysis, as the mean and SD could not be estimated from the reported data.

Data synthesis and measures of effect sizes
Standardized mean difference (SMD) with (95% confidence intervals [CIs]) and p values for effect sizes were estimated for continuous outcomes. The analysis was performed using R programming software with the "metafor" package. 10,11 Because the included studies likely do not share the same effect size, random effects models were fitted using the rma function with the restricted maximum-likelihood (REML) estimator. 12

Dealing with heterogeneity
Prespecified sources of heterogeneity included species and methods of frailty or sarcopenia assessment. The Cochran Q test was used to test for heterogeneity between studies in random models. 13 The I 2 statistic was used to estimate the percentage of total variation across studies attributed to heterogeneity, rather than chance. Heterogeneity was defined as: • I 2 = 0%-40%: no or mild heterogeneity • I 2 = 40%-80%: moderate heterogeneity • I 2 > 80%: severe heterogeneity For all variables, random models with prespecified sources of heterogeneity as moderators were fitted using the rma function of the metafor package. 11 For clarity, pooled effect estimates were reported for all prespecified subgroups.

Assessment of methodological quality
Using the Newcastle Ottawa scale for studies including humans, three out of 32 were found to be without methodological limitations (Figure 2). For the other human studies, the most common limitations were missing records of blinding and randomization (12/32, 41%) and more than 20% of participants having missing data (12/32, 41%).
Using the ARRIVE checklist for studies, including experimental animals, no study was found to fulfil all quality criteria; therefore, all studies were considered at high risk of bias ( Figure 3). The most common methodological limitations included failure to provide a sample size calculation (76/78, 96%), missing description of methods used to prove that the analyzed data met the assumptions of the statistical approach used in the study (63/78, 78%), failure to report the exact numbers of subjects included in each group in each analysis (36/78, 46%), and failure to report the absolute number of animals studied in each experiment (37/78, 47%).

Data synthesis
Multiple measures of frailty and iron metabolism were identified, and these were further categorized into the subheadings described below. The effects of frailty on the outcomes in these categories for humans, mice, and rats are shown in Figure 4.

Mitochondrial morphology
Estimates of mitochondrial number, mitochondrial size, and mitochondrial (mt) DNA copy number in frail and nonfrail humans and animals showed severe heterogeneity between studies that was not resolved by moderator and subgroup analyses (Supplementary File S5, Table S2). Mitochondrial volume density and voltage-dependent anion channel (VDAC) protein expression did not differ significantly between the frail and nonfrail groups, without heterogeneity. Studies measuring mtDNA/nuclear DNA ratio reported significantly lower values in the frail compared to the nonfrail groups, without heterogeneity (Supplementary File S5, Table  S2).

Mitochondrial dynamics
The expression of genes and proteins that regulate mitochondrial dynamics showed differential regulation within individual subgroups. The gene expression of Dynamin related protein (DRP) 1, a regulator of mitochondrial fission, was downregulated in frail humans and rats but not in genetically modified mice (Supplementary File S5, Table S3). In contrast, protein expression of DRP 1 and Mitochondrial fission 1 protein (FIS1) were increased in rats and mice, respectively, but not in humans.
Expression of mitochondrial fusion genes Mitofusin (MFN) 1, MFN 2, and Mitochondrial Dynamin Like GTPase (OPA1), were reduced in frail groups across species, although there was residual heterogeneity of the estimates for these genes in studies where frailty was defined by age. Conversely, protein expression of MFN1 was not different between frail and nonfrail humans and animals across all moderators and subgroups. MFN2 protein expression was increased in experimental models where frailty was defined by age, but not in other subgroups. OPA1 protein expression was reduced in human models of frailty, but not in other species, and the effects were inconsistent across different frailty definitions.

Mitochondrial respiratory complexes
Heterogeneity for gene and protein expression and activity of the oxidative phosphorylation complexes were explained by the frailty definition used (Supplementary File S5, Table S4). When the Fried Frailty Index was used in human studies, all outcomes (complex I-V activity, gene and protein expression), with the exception of activity of respiratory complex I, were significantly reduced in subjects classified as frail. For genetically modified animal models, only protein and gene expression of the respiratory complexes were reduced, whereas the enzymatic activities of these complexes, with the exception of complex I, were unaffected. When age or immobilization/sedentary lifestyle were used as measures of frailty, the heterogeneity for the majority of the outcomes was severe, and where the heterogeneity did not reach significant levels, there was no difference between the frail and nonfrail groups. Only a small number of outcomes in this category was examined in studies using geriatric assessment tools or sarcopenia as a measure of frailty, limiting the assessment of these studies.

Mitochondrial energy production
Severe heterogeneity for comparisons of 31P-MRS phospho-creatine recovery rate, citrate synthase protein expression, and citrate synthase activity, between frail and nonfrail could not be resolved using either species or type of frailty assessment as moderators (Supplementary File S5, Table S5). ATP levels were significantly lower in frail rats, but for mice the residual heterogeneity remained severe. Phosphate/oxygen (P/O) ratio, defined as the number of atoms of phosphorus incorporated into an ATP molecule per every two electrons used for reduction of the O 2 molecule in OXPHOS, was significantly decreased in frail humans and mice, however, only three studies contributed to this estimate.

Oxygen consumption
Frailty was associated with significantly lower oxygen consumption in both state 3 and state 4, as well as maximal oxygen uptake (VO2 max ) in humans, without heterogeneity (Supplementary File S5, Table S6). In contrast, there was no difference in oxygen consumption measured in frail versus nonfrail rats. Mouse studies demonstrated high heterogeneity for these outcomes, and even for outcomes where the heterogeneity was resolved, no consistent influence of frailty on oxygen consumption was seen.

Oxidative stress
Assessments of the role of oxidative stress in frailty demonstrated discordant observations between species. In human studies, there were no significant differences between frail and nonfrail subjects with respect to reactive oxygen species (ROS) production or protein carbonylation (Supplementary  File S5, Table S7). Antioxidant enzyme levels were unchanged or lower in frail subjects in humans. In contrast, in mouse studies, the production of ROS in respiratory state 2 and mitochondrial ROS production were increased in frail animals and the levels of catalase and SOD1 proteins as well  Table S7). The small numbers of studies for many of the examined outcomes limited the certainty of these results. For rat studies, an increase in ROS production was seen in states 3 and 4 respiration but not for state 2, or for mitochondrial ROS production. The effects of frailty on antioxidant proteins in rats could not be examined due to the high heterogeneity and the small number of reported outcomes.

Autophagy and apoptosis
Measures of apoptosis and autophagy in frailty differed significantly between species. In human studies, there were no differences between frail and nonfrail subjects for any measures of autophagy or apoptosis (Supplementary File S5, Table S8). In rat studies, the apoptotic markers apoptosis inducing factor (AIF) release, and caspase 3 activity, were increased, and the autophagy markers microtubule-associated protein 1A/1B-light chain 3 (LC3) and Beclin1 protein expressions were reduced. There was severe heterogeneity for other reported measures in rat studies. In mouse studies, no consistent differences in markers of apoptosis or autophagy were seen, although the numbers of included studies reporting these outcomes was small.

Regulators of mitochondrial function
In human studies, out of the examined outcomes, only Sirtuin 3 (SIRT3) protein expression was shown to differ between frail and nonfrail subjects (Supplementary File S5, Table S9). In rats, peroxisome proliferator-activated receptor gamma coactivator 1-alpha (PGC1α) gene expression, insulin-like growth factor 1 (IGF1) levels, nuclear respiratory factor 1 (NRF1), and SIRT3 protein expressions were lower in frail animals. However, NRF1 gene expression was unchanged and the mitochondrial transcription factor A (TFAM) protein expression was higher in frail rats. In mouse studies, the severe heterogeneity for all of the reported outcomes limited analyses.

Measures of iron metabolism
Only three measures of iron metabolism were identified in the studies included in this review. Hemoglobin concentrations in blood were significantly lower in frail human and animal subjects across all subgroups, with the exception of one study examining human subjects with sarcopenia/muscle atrophy (Supplementary File S5, Table S10).
Non-heme iron levels in tissue were significantly increased in frail human and animal subjects across all subgroups.
There was severe heterogeneity for serum transferrin levels between studies that was not resolved by subgroup analysis.

DISCUSSION
There were three major finding of the current analysis: First, there is significant heterogeneity for the reported effects of frailty on individual outcomes between species. For example, measures of oxidative stress as well as measures of autophagy and apoptosis were not different between frail and nonfrail humans, but there was evidence of changes in these processes in frail versus nonfrail rats and mice. The examination of mitochondrial dynamics markers provided the most consistent data across species, where the observed outcomes showed similar patterns of dysregulation for both humans and rats. However, even in this category, the frail mice differed significantly from frail rats and humans. This trend was consistent for the majority of the outcomes reported in this review, suggesting that there is very little homology between frailty in mice and humans. We believe that this discordance arises from the difficulty to mimic this complex syndrome with simple genetic modifications, or aging, which were the two most common models of frailty used in mouse studies.
Second, the frailty definitions utilized for characterization of frail groups have significant effects on the results. For example, frailty is associated with reduction in respiratory complex activity, gene expression, and protein expression in humans. However, in other subgroups, this dysregulation is not visible. Moreover, our analyses showed that regardless of species, in subgroups where the definition of frailty allowed for more subjective categorization of examined subjects (based on age and sedentary lifestyle), the heterogeneity for majority of outcomes was severe and could not be resolved. These findings strongly highlight the need for more consistent frailty definitions and suggest that the processes dysregulated in frailty are distinct from aging.
Third, the inconsistent experimental design, data definitions, and reporting resulted in severe heterogeneity across multiple analyses and is likely to present a barrier to clinical progress in this important area of research. The severe heterogeneity observed for almost all outcomes limits our ability to identify a set of core measures that might be put forward for consistency of reporting. However, when considering only human studies, the heterogeneity for most outcomes was nonsignificant and three categories were identified as consistently dysregulated in frailty: mitochondrial dynamics, oxygen consumption, and the activities and abundancies of respiratory complexes. We believe a semitargeted examination of these aspects of mitochondrial function is a reasonable core dataset for future research in this area.
A final comment is that our study failed to identify detailed investigations of dysregulated iron metabolism, despite