Introduction

Mitochondrial DNA (mtDNA) is 16.6 kilobases in length, encoding for 37 genes, including 13 polypeptides which form core subunits of 4 out of the 5 oxidative phosphorylation (OXPHOS) protein complexes. MtDNA population variants have been implicated in multifactorial diseases, including hypertension, diabetes, atherosclerosis and Alzheimer’s disease (AD) [1–3]. Hypotheses linking mtDNA variation to the underlying neuropathology of AD include excessive oxidative stress due to mildly deleterious variants [4]. Reduced activity in Complex IV has also been highlighted by a number of investigators [5]. MtDNA is maternally inherited, meaning that the evolution of mtDNA is defined by the emergence of distinct global lineages called haplogroups. Different common mitochondrial haplogroups have been associated with increased (and decreased) risk of AD in various studies; some studies have failed to find any associations at all, including a recent meta-analysis [6]. The most likely reason for these conflicts in the literature is population stratification that occurs with common mtDNA variants [7].

In response to the conflicting data generated by the haplogroup association model, the variant load model (VLM) has recently been introduced [1, 8–10]. The VLM investigates the combined effects of population variants predicted to be mildly deleterious, most of which are rare. Rare variants do not show the same geographical stratification as common variants. We recently applied the VLM to investigate AD patients compared to aged, but non-neurodegenerative disease patients, using sequences from two independent cohorts, the MitoKor (MK-US) and Medical Research Council (MRC-UK) cohorts [1, 11]. The study found no association between mtDNA variant load and the onset of AD. Surprisingly, the variant load was lower in the US controls, indicating a potential association between low VL and healthy ageing. In order to validate these results in a different population, the current analysis explored the role of rare mtDNA variants in Japanese cohorts, consisting of both AD patients and the aged, but otherwise healthy individuals and a younger cohort.

Methods

Results

Variant load comparison

The calculated mean “all scoring variants”, “MutPred >0.5 variants”, and “MitoTIP tRNA variants” variant loads for each group are shown in Table 1. MutPred >0.5 Variant Load Comparison.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 1. Mean variant loads (with standard deviations) for Japanese centenarian, AD patient and healthy young cohorts, calculated using: (a) all MutPred scoring variants, (b)variants scoring >0.5 on MutPred and (c) MitoTIP tRNA variant scores.

https://doi.org/10.1371/journal.pone.0276169.t001

When considering all genes at a MutPred variant load setting of >0.5, a statistically significant difference was found between the three groups (Fig 1) (F (2.12, 88.03) = 3.43, p = 0.034). A Tukey post-hoc test revealed a statistically significantly higher all genes MutPred >0.5 variant load in centenarians (mean variant load = 0.754, SD +/- 0.641, p = 0.054) compared with the AD group (mean variant load = 0.567, SD +/- 0.508); with a difference in the healthy young group (mean variant load = 0.577, SD +/- 0.508, p = 0.07). This finding, of higher variant loads in centenarians (albeit of marginal significance) opposes our earlier findings [11]. Refining the analysis to utilise just the MutPred scores of variants in genes affecting Complex IV scoring >0.5, again found no statistically significant difference between the groups (F (0.09, 7.71) = 1.60, p = 0.203). Supplemental Tables (a-c) provide all raw variant load values by sequence.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 1. Means and 95% confidence intervals of variant loads calculated using MutPred.

All genes variants scoring >0.5 on MutPred (top left), variants scoring >0.5 on MutPred affecting Complex IV (top right) and MitoTIP tRNA variants (bottom left).

https://doi.org/10.1371/journal.pone.0276169.g001

Haplogroup context of high scoring variants

Interrogation of the centenarian sequences revealed 9 high-scoring variants (>0.5), which occurred more than twice in this cohort. A number of these variants (3391A, 5460A, 8414T, 8794T, 12811C and 13651G) appeared to be present in higher frequencies in the centenarian cohort than in the other 2 groups. Table 2 shows the variants, along with their more common haplogroup associations using MITOMAP data from 2019. Importantly, none of these high-scoring variants appeared to be present out of the normal haplogroup context. In the majority of cases, the variants were present in >98% of the analysed sequences of the corresponding haplogroup branch, as seen in GenBank. The remaining variants were not particularly uncommon in their haplogroup contexts, being present in >10% of the relevant GenBank sequences.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 2. Haplogroup context and common GenBank 2019 Haplogroup associations of variants scoring >0.5 in MutPred found in the study population.

https://doi.org/10.1371/journal.pone.0276169.t002

Discussion

Several studies have implicated mtDNA variation in the onset and progression of AD. Here we explored the role of mtDNA variation in the onset of AD using a new association strategy, the VLM, applied to an East-Asian population dataset [15]. Our results do not provide any evidence for an association between levels of mildly deleterious mtDNA and the onset of AD. This finding also supports our prior observations, made in UK and US-based populations [1, 11]. In contrast to our previous study, our present results suggest that those living to advanced old age may harbour more mildly deleterious mtDNA variations. While, intuitively, it seems logical that fewer mildly deleterious mutations might associate with healthier ageing, some experimental evidence suggest that a limited degree of mitochondrial functional decline can lead to a lengthened lifespan in a range of organisms, including C. Elegans, Drosophila and mice [16, 17].

Our current study cohort and that used in our previous study, have some notable differences, as shown in S1 Table. Firstly, the elderly group differed in age, with the average age of the aged controls in the study by Pienaar and others (2017) that were 77.2 years (± 9.6, Standard Deviation (SD)) for the UK cohort and 83.4 years (± 5, SD) in the US cohort. In contrast, the Japanese centenarians were, by definition, more than a decade older. As little information about the sequenced individuals is available from either the database website or the two original study papers, little is known about the health status of the centenarians or the criteria used for AD diagnosis. This suggests that a better approach might be to apply such a model in more comprehensively described ageing cohort. There are a number of cohorts including those of octogenarians from Newcastle (UK) and Leiden (Netherlands) with detailed phenotypic data gathered at baseline then 3 and 5 years. As being able to correlate with the specific aspects of the health, status of populations with mtVLM would provide a much more satisfactory link of genotypes to phenotype.

The definitions used to classify the healthy non-obese young men are also unclear, although their mean body mass index (BMI) was 20.2 (+/- 2.3).

Although the aim in using the VLM was to ameliorate the effects of population stratification, a residual stratification effect may influence the results of VLM studies, due to a very small number of common haplogroup-defining variants remaining, even when only using variants with MutPred scores >0.5 for the analysis [3]. In the present work, we also demonstrated the presence of haplogroup-defining variants in this Asian population with high MutPred scores, suggesting that they may be mildly deleterious. In the centenarian cohort, who had the highest mean variant load scores, nine individuals had frequently occurring (more than twice) variants, with high MutPred scores. Of these variants, five are present in over 98% of the GenBank haplogroup sequences corresponding to the most common haplogroups of the study sequences in which they were found. Two of these (8414T and 8794T) are listed as markers for their respective haplogroups on MITOMAP, while another (8584A) is listed as a marker of the major branch of the mtDNA phylogenetic tree. The other cohorts also had frequently occurring high-scoring variants; ten in the AD group and five in the healthy young group.

Overall, our analysis suggests that our finding of a marginally higher variant load in centenarians is not primarily being driven by many rare population variants, but by a few common variants that have high MutPred scores. We predict that these variants do not exert a deleterious effect due to the haplogroup context in which they are found. In support of this a number of studies have been completed that have found mutations shown to cause inherited mitochondrial disease in humans to be present in other species in the absence of disease [18–20]. These investigations made suggestions as to what elements of the haplogroup context of the human mutations that allowed them to exist as benign variants in these species. Here we make the speculation that if the deleterious effect of very rare population variants that have been casually linked to inherited disease can be compensated by haplogroup background, so can the effect of mildly deleterious population variants in the context of common complex disease. Additional support of this notion comes from work to suggest common population variants seen in a different haplogroup context (in humans) might be associated with the onset of common disease [21]. We explored this in the current work with none of the high-scoring variants in centenarians being seen out of their usual haplogroup context. Population variants that are haplogroup associated when seen in a different haplogroup context have been called private variants in the past and are an area of interest in the context of mtDNA population variation and common disease [1]. The use of this terminology again might be of help in the development of new models. Hence, a “shadow” of the haplogroup stratification effect may endure in VLM-based studies, potentially contributing to inconsistent results. Taken together this suggests that refinement to the VLM might be required if it is to achieve its aim. But this will may not be as simple as the elimination of frequently occurring variants with high MutPred scores. As recent paper, patients with atherosclerosis were found to have a higher likelihood of having a common variant scoring >0.5 in MutPred, alongside a rarer variant with a similarly high score. This suggests a two-hit hypothesis and [8].

In summary, a consistent null picture in the context of the role of mtDNA variation in AD is emerging [11, 22], after application of a number of methods for investigating this notion. Further exploration of mtDNA variant load association in the context of ageing would benefit from larger, dedicated cohorts, consisting of individuals with detailed clinical and demographic information this approach has been used in other phenotypes [2, 10] such cohorts exist in the ageing context but as yet mtDNA sequence data is unavailable [23, 24].

Supporting information