Mitochondrial DNA and Alzheimer’s disease: a first case–control study of the Tunisian population

Alzheimer’s disease (AD) is the most common neurodegenerative disorder in humans and presents a major health problem throughout the world. The etiology of AD is complex, and many factors are implicated, including mitochondria. Mitochondrial alteration has been proposed as a possible cause of AD. Therefore, several studies have focused on finding an association between inherited mitochondrial DNA variants and AD onset. In this study, we looked, for the first time, for a potential association between mitochondrial haplogroups or polymorphisms and AD in the Tunisian population. We also evaluated the distribution of the major genetic risk factor for AD, the apolipoprotein E epsilon 4 (APOE ε4), in this population. In total, 159 single-nucleotide polymorphisms (SNPs) of mitochondrial DNA haplogroups were genotyped in 254 individuals (58 patients and 196 controls). An additional genotyping of APOE ε4 was performed. No significant association between mitochondrial haplogroups and AD was found. However, two individual SNPs, A5656G (p = 0.03821, OR = 10.46) and A13759G (p = 0.03719, OR = 10.78), showed a significant association with AD. APOE 4 was confirmed as a risk factor for AD (p = 0.000014). Our findings may confirm the absence of a relation between mitochondrial haplogroups and AD and support the possible involvement of some inherited variants in the pathogenicity of AD.


Introduction
Alzheimer's disease (AD; MIM 104,300) is the major cause of dementia and the most common neurodegenerative disorder [1,2]. AD is manifested by a progressive memory loss and a decline in cognitive function [2,3]. According to the age of onset, two forms of AD can be distinguished: earlyonset AD (EOAD) and late-onset AD (LOAD). About 1-6% of AD cases are of the early-onset type, with 60% of them involving familial aggregation [4,5].
Most AD cases have the LOAD phenotype, also called sporadic AD. Sporadic AD is considered a multifactorial disease with a non-Mendelian (complex) transmission. To date, AD pathogenesis is not well understood; only some risk factors are known to be associated with this disease. AD is caused by both genetic and non-genetic factors. The genetic cause is still not known, especially because patients with the LOAD phenotype show no mutations in presenile 1 3 genes and amyloid precursor protein (APP) [6,7]. The apolipoprotein E (APOE) ε4 allele is regarded as the strongest risk factor [8][9][10][11]. Besides mutations in the APOE e4 allele, genetic variations in more than 20 susceptibility genes have been identified by genome-wide association studies and next-generation sequencing. These genes are involved in many biological functions, such as immune response, synapse function, and lipid metabolism. Each of these genes has a small effect on the risk of developing AD [12][13][14][15][16][17].
The lack of a clear and established etiology of AD has prompted the emergence of the "mitochondrial hypothesis" for explaining AD. Mitochondrial abnormalities have been found in the brain and other tissues of AD patients, reinforcing the potential association between mitochondrial dysfunction and neurodegeneration [18][19][20][21]. In addition, a relationship between a maternal history of AD and an increased risk of AD onset has been found, supporting the mitochondrial theory of AD [22]. In that regard, over 20% of inherited LOAD cases have a maternal history, and the maternal influence seems to be more powerful than the paternal one [23,24].
Mitochondrial DNA (mtDNA) is highly polymorphic and has more mutations than nuclear DNA. These mutations are dispersed across the whole mtDNA. The mutation rate in mtDNA is 10 times higher than that in nuclear DNA [25,26]. In addition to uniparental transmission, mammalian mtDNA does not undergo recombination, so the evolution of mtDNA shows stable variations that are inherited from the same ancestor and that are grouped into specific lineages called mitochondrial haplogroups [27][28][29][30][31]. Every haplogroup is composed of a combination of characteristic single-nucleotide polymorphisms (SNPs) or signature mtDNA mutations, representing the major branch points in the human phylogenetic tree [32,33]. That is, every haplogroup has accumulated mutations over time, leading to the divergence of maternal lineages. Consequently, haplogroups tend to be specific for populations and continents [34]. In these haplogroups, some additional mutations may have arisen, creating young branches in the human mtDNA tree. As a result, haplogroups can be divided into smaller groups called sub-haplogroups [35]. The classification of sub-haplogroups is still in progress because of the increase in complete mtDNA sequences of many populations from different ethnic groups [36]. According to the latest classification, individuals with haplogroup L are sub-Saharan Africans, with seven major families (L0, L1, L2, L3, L4, L5, L6) [36,37]. East Asian populations have haplogroups A, B, C, D, and E, while haplogroups H, I, J, K, T, U, V, W, and X are specific to Eurasian populations [34]. Thanks to a huge number of studies on mitochondrial genome diversity, specific mitochondrial haplogroups have been identified for many populations [38,39]. A hypothesis was proposed that some mitochondrial haplogroups are not neutral and are involved in many mechanisms. Some haplogroups may have been selected during evolution as an adaptation to particular environmental conditions [40]. Accordingly, several studies have explored the association of mitochondrial haplogroups with parameters such as longevity [40], age-related diseases [41,42], and multifactorial diseases such as AD [43]. For example, in the Italian population, sub-haplogroup H5 was described as a risk factor for AD [35], while in the Japanese population, the haplogroups G2a, B4c1, and N9b1 seem to be associated with an increased risk of AD [26]. However, no definitive results can confirm or prove the involvement of mtDNA in AD onset. The results cannot be generalised to other world populations because of the differences in mitochondrial haplogroups from different regions. In this context, it would be interesting to study the association between haplogroups or polymorphisms of mtDNA and AD in populations founded by mixtures, such as the Tunisian population. The prevalence of AD in Tunisia is 3.7% [44].
Tunisia is located in northwest Africa. It is bordered by Algeria to the west and southwest, by Libya to the southeast, and by the Mediterranean Sea to the north and east. The settlement history of Tunisia goes back to prehistoric times. The country has successively witnessed the flowering of the Aterian civilisation (20,000 YBP), the Iberomaurusian civilisation (12,000 YBP), and the Capsian civilisation (9000 YBP). The Berbers, the original people of Tunisia, are a heterogeneous group of humans who have spread uniformly across North Africa for about 4000 years. This Berber population has been modified by the arrival of the Phoenicians, Roman Byzantines, Arabs and Bedouins, Andalusians, Ottomans, and French. Genetic studies conducted on the Tunisian population targeting mtDNA have shown a significant diversity of the mitochondrial haplogroup landscape in relation to the country's rich history of complex and eventful settlements [45][46][47][48][49][50]. The Tunisian populations host a Eurasian component represented by haplogroups originating from Europe and the Middle East (H, V, U5, J, T, R0a), with a frequency of about 70%; a sub-Saharan component represented by L haplogroups and their subtypes, with a global frequency of 25%; and a North African component represented by haplogroup U6 and its subtypes, with an overall frequency of around 5%. In Tunisia, these three components vary in their frequency from one group to another, reflecting the mosaic character of the Tunisian population [51]. This mix of West European, Middle Eastern, and sub-Saharan mtDNA lineages is highly informative for our study, especially the presence of the sub-Saharan haplogroups, which is an advantage over previously studied populations in relation to AD. This rich population encouraged us to investigate the involvement of the Tunisian mtDNA genome in AD. First, we evaluated mitochondrial haplogroups, and then we chose to simplify the research by focusing on some single mitochondrial mutations instead of a set of SNPs. We adopted this approach because polymorphisms in mtDNA can affect encoded proteins, inducing changes in mitochondrial respiratory activity and increasing the number of free radicals, resulting in early apoptosis [36].

Samples
DNA samples from 254 subjects (58 AD patients and 196 healthy controls) were recruited from the Department of Neurology, University Hospital of Fattouma Bourguiba, Monastir, Tunisia; the Department of Neurology, University Hospital of Mahdia, Tunisia; and the Alzheimer Family Assistance Center in Tunisia (AFA center). The control group was comparable in age, gender, and ethnicity to the patient group (Supporting Information; Demographic information's).
Written informed consent was obtained from all participants or their representatives. All participants were unrelated over three generations. For each participant, age, demographic information, and family and medical histories were collected. In addition, for AD patients, a detailed clinical assessment was carried out, including a Mini Mental State Examination (MMSE), a standard brain MRI, neuropsychological tests, and laboratory studies. All AD patients were included after at least 2 years of follow-up.
Diagnoses of AD were made according to the criteria of the National Institute of Neurological and Communicative Disorders and Stroke and the Alzheimer's Disease and Related Disorders Association (NINCDS-ADRDA) for possible or probable AD [52].
Controls were selected after passing an MMSE test and according to the following criteria: no family member with AD or any type of dementia or psychiatric disorder, a mentally healthy status, and no history of alcohol or drug use.

Mitochondrial DNA genotyping
Genomic DNA was extracted from fresh blood (2 ml) by the standard phenol-chloroform protocol. Then, genotyping was performed on the Axiom 815 K Spanish Biobank Array (Thermo Fisher). This array has polymorphic sites (SNPs) covering the whole genome, including variations with common and low frequencies and 159 mtDNA sequence polymorphic sites. A data set of mitochondrial SNPs, dispersed across the coding region and representing the most common branch of the mtDNA phylogeny, was obtained. Genotyping was performed in the Spanish National Center for Genotyping (CeGEN, Santiago de Compostela, Spain). The genotyping was conducted according to the manufacturer's instructions (Axiom™ 2.0 Assay Manual Workflow). We performed a quality control (QC) for the plates and then for the samples using the Affymetrix Power Tool (APT) 1.15.0 software and according to Axiom's data analysis instructions. Only samples with a call rate above 97% were considered. The data passed through many QC checks, including checks for discordant sex, duplicated samples, and call rate (samples with a call rate < 0.95 were eliminated), and underwent an identity by descent (IBD) analysis. Duplicated and genetically related individuals (PIHAT > 0.1875) were excluded. Individuals above six standard deviations from the European population mean were regarded as outliers of the population and removed from our cohort. Then, the data of 159 mitochondrial SNPs were extracted from the whole array (Supporting Information; table s1). Haplotyping and the construction of a phylogenetic tree were done using Hap-loGrep 2 following the updated nomenclature proposed in PhyloTree Build 17 [32].

Mitochondrial DNA HVS-I sequencing
The amplification of HVS-I (16,024-16,383) was performed on 64 samples (25% of our cohort). The choice of the amplified samples was done randomly to only confirm the haplogroup's identity already obtained when exploiting the array's results. This step was to check and confirm primary obtained results from the array. The HVS-I region of mtDNA was amplified by using specific primers [53]: The amplified product was purified with the QIAGEN kit. The sequencing reaction was performed using the Big-Dye Terminator Cycle Sequencing Kit with AmpliTaq DNA polymerase (Applied Biosystems). The sequencing products were run in the automatic sequencer ABI377 (Applied Biosystems). The obtained HVS-I sequences were checked manually and then aligned to the revised Cambridge Reference Sequence (rCRS [54,55]; using BioEdit version 7.2.5. As with the previous results obtained by the array, the haplotyping of the sequences was done using HaploGrep2.

Statistical analyses
Descriptive statistics-means, frequencies, and standard deviations (SDs)-were calculated. For continuous measures, the comparison of the mean between patients and controls was performed using Student's t-test. Odds ratios (ORs) were calculated with 95% confidence intervals.
Mitochondrial sub-haplogroups, APOE genotypes, and gender and genotype frequencies were compared between AD patients and controls using Pearson's chi-square test or Fisher's exact tests with SPSS. A p value of 0.05 was considered significant with 95% confidence intervals (CIs). The obtained results from HaploGrep2 showed a large variability in mitochondrial sub-haplogroups (52 subgroups). Mitochondrial sub-haplogroups were grouped into 41 subhaplogroups according to their phylogenetic relationships. The association between mitochondrial SNPs and AD was tested using Fisher's exact test provided by PLINK 1.9.
Since we observed very low frequencies of some mitochondrial sub-haplogroups, which might disturb the analysis of our results, we decided to reduce the number of haplogroups by eliminating all sub-haplogroups with a frequency of less than 0.5%.
We grouped every sub-haplogroup into the haplogroup it belongs to. Thus, the new classification contains a group called H*, which includes H4, H5, H9, H11, and H30. Another group called J* includes J1, J1b, and J1c. The group U* includes U1, U4, U5, and U7, and N* includes N1 and N9a. After this classification, some heterogenous haplogroups with low frequencies remained, which we classified into "others". This group contains haplogroups X2, X3, M73a, I2a, W, and G. This step reduced the number of sub-haplogroups from 41 to 27 haplogroups or sub-haplogroups. Then, each of the sub-haplogroups was evaluated separately to look for a significant association. The comparisons between each mitochondrial sub-haplogroup were conducted using Bonferroni's adjustments.

The apolipoprotein E (APOE) genotyping
The APOE status was also examined in our study. APOE genotyping was performed using data of two SNPs (rs7412 and rs429358) from the array. SNP genotypes for both polymorphisms were combined for each sample. The classification of the participants' genotypes was based on the presence of APOE alleles. APOE genotypes were determined according to Table 1 [56]. Patients and controls were classed into two groups, taking the presence of the APOE4 allele as a criterion. The first group (APOE4 +) had at least one copy of APOE4. The second group (APOE4−) did not carry any copy of the APOE4 allele.

The apolipoprotein E (APOE) study
Since APOE is the major genetic risk factor of AD and became a routine examination in all genetic AD studies, we evaluated the allele frequencies with the covariates gender and age. APOE4 + subjects (with one or two copies of APOE4) represented 13.2% of our total cohort. Of all AD patients, 46.6% are APOE4 + , and 53.4% are APOE4−. The effect size (OR) for APOE4 was 0.39. It is important to mention that only 4.3% of our samples presented the ε2 allele of APOE, while a large part of the samples (82.5%) presented the ε3 allele.
The comparison between AD patients and controls without taking gender into consideration showed a significant correlation for APOE4 carriers (p = 0.000014). We conducted a separate comparison according to gender to evaluate APOE4 values in AD patients and controls. Results showed that 48.1% of Female patients (mean age 80.25 ± 10.88) are APOE4 + (having at least one copy of APOE4) and 51.9% are APOE4-. However, female controls (mean age 69.66 ± 9.05) showed a high percentage (83.3%) of APOE4 + whereas only 16.7% were APOE4−. Almost the same distribution was noticed for males in our population: male patients (mean age77.03 ± 7.60) were grouped as 54.8% of APOE4 + and 45.2% of APOE4−. While, in the male control group (mean age 66.68 ± 9.24), 79% were APOE4 + and only 21% were APOE4−. Results showed that APOE4 carriers had a significant p value, which confirmed our previous findings ( Table 2).

Comparison between the two cohorts: AD patients and controls
The comparison of the distribution of mitochondrial haplogroups between the AD and the control cohort was first carried out by considering the 27 sub-haplogroups described previously in the two cohorts. No difference was found between the two cohorts (p = 0.774). Moreover, by comparing each of the 27 sub-haplogroups separately between the two cohorts, we did not detect any significant differences between the AD and the control cohort, suggesting the absence of an association between a specific lineage and AD in the Tunisian population. A particular significant p value (p = 0.0450) was noticed for the Sub-Saharan haplogroup L3b. This Sub-haplogroup was present only in the controls' group and totally absent in patients' group. The percentage of this sub-haplogroup is 6.6% in patients' group and present only 5.1% of our studied population.
It is necessary to mention that "Others" group is an heterogenous group composed by minor mitochondrial subgroups with a frequency less than 0.5%. It enfolds the haplogroups X2, X3, M73a, I2a, W, and G. It is not a specific mitochondrial haplogroup and all observed p values with reference to this group were not considered (Table 4).
Then, we grouped the haplogroups and sub-haplogroups into 10 lineages (Supporting Information; Tables 5 and 6) in the two cohorts. None of these haplogroups showed an association with AD, although some haplogroups showed differences in their distribution between the two cohorts, such as haplogroup H (p = 0.0906, OR = 1.6758), which was more frequent in AD patients (44.5%) than in controls (32.4%), and the haplogroups L (p = 0.1296, OR = 0.551) and U6 (p = 0.1506, OR = 0.247), which, on the contrary, were more frequent in controls (24.8%, 6.6%) than in AD patients (15.3%, 1.7%) (Fig. 1).
To obtain a clear and global picture of the composition of mitochondrial haplogroups in our samples, we focused on grouping all haplogroups according to their ethnic lineages: North African, sub-Saharan, and Eurasian lineages. The major component in our population was the Eurasian lineages (70.5%), followed by the sub-Saharan lineages (22.8%) and the North African lineages (6.7%).
Almost the same profile of these three principal lineages was observed when we compared the mitochondrial haplogroups separately between patients and controls. Again, the dominant component belonged to the Eurasian lineages, with percentages of 67.3% and 81% for controls and AD patients, respectively. Sub-Saharan haplogroups were present in 25% of the control group and in 15.5% of AD patients. Finally, the North African lineages were found in 3.5% of AD patients and 7.7% of controls (Figs. 2). The frequencies of mitochondrial haplogroups were compared separately for gender between AD patients and controls. The results showed no significant differences between both groups: p = 0.439 for males and p = 0.515 for females.

Search for an association between mitochondrial SNPs and AD
The second part of our investigation was focused on individual mtDNA SNPs without considering haplogroups. We analysed all 159 mitochondrial SNPs in order to test a possible association with AD. The association was tested using Fisher's exact test. Of all the mitochondrial polymorphisms studied, only two SNPs showed a significant result among patients: A5656G (p = 0.03821, OR = 10.46) and A13759G (p = 0.03719, OR = 10.78). The mitochondrial SNP A13759G is located in NADH dehydrogenase subunit 5. This SNP appears in the branch point of two sub-haplogroups, H11 and j2a2, according to the human phylogenetic tree (see Supporting Information; the phylogenetic tree). The mitochondrial SNP A5656G is located in a non-coding region separating two structural genes of transfer RNAs (tRNAs): tRNA (Ala) and tRNA (Asn).

Discussion
Sporadic AD is a complex disease with many risk factors, including genetic, environmental, and lifestyle factors (higher education, reading, taking care of family, smoking, etc.) [57]. Whether mitochondrial polymorphisms are also risk factors is largely unknown. Recent studies suggest that  Multiple studies have been conducted to inspect the potential causative relation between mtDNA and AD [58]. In this study, we evaluated the possible association between AD and mitochondrial haplogroups or polymorphisms in the Tunisian population by genotyping mitochondrial polymorphisms dispersed across the whole mitochondrial genome, using a cohort of 58 AD patients and 196 controls comparable in gender, age, and ethnicity. To our knowledge, this investigation is the first case-control study related to AD and mtDNA in Tunisia, home to an admixed population from North Africa. The results showed that none of the observed mitochondrial haplogroups in our population had a significant association with AD. However, we observed that two mitochondrial variants, A5656G and A13759G, were statistically associated with AD.
In this study, we explored Tunisian mtDNA in relation to AD. The results showed a large variety of haplogroups, creating a mosaic structure composed of 41 haplogroups in our cohort (254 individuals only) and representing a high level of genetic diversity in the Tunisian population. This genetic diversity in Tunisia can be explained by a heterogeneous gene pool caused by invasions from different populations: Phoenicians, Byzantines, Arabs, Romans, Ottomans, Vandals, Andalusians, etc. Most of these populations have left a genetic imprint on the Tunisian population. In addition, the arrival of thousands of Europeans (from France, Italy, Spain, Austria, Belgium, Greece, Russia, and Malta) in Tunisia during the colonisation (about 1881-1921) further enriched the genetic pool [59]. As has been noted in previous studies, the long and rich history of Tunisia created a specific genetic pool. This genetic heterogeneity was observed both in different Tunisian regions and in populations of the same ethnic background. In our study, we found the same haplogroups as previously described in other studies [60,61]. The major haplogroups found belonged to the West Eurasian lineages (H, T, U, K, N, V, I, W, J, X), with a predominance of the sub-haplogroup H2a. In general, haplogroup H is the most frequent European lineage in North Africa. The highest frequencies were found in Berber-and Arab-speaking Moroccans (37% and 34%, respectively) [62]. Approximately the same percentage was found in our study, as 35.4% of our cohort had haplogroup H. Some previous studies have reported an association between haplogroup H and AD. For example, in Asturias (northern Spain), a significant association was found between haplogroup H and sporadic AD [41]. In Italy, a specific sub-haplogroup of H (H5) was identified as a risk factor for AD [35]. Despite the predominance of sub-haplogroup H2a in our population, we did not find any significant value for an association with AD.
The Eurasian lineage U was also present in our population (10.6%). This mitochondrial haplogroup has a large distribution in Europe and southwest Asia [63]. In Caucasian Americans, haplogroup U (p = 0.007) showed a differential effect in AD patients according to gender; it had a protective effect in females, and it increased the risk of developing AD in males [3]. The North African-specific lineage U6 was present (5.5%) and constituted, together with the North African haplogroup M1, 6.7% of our population.
In agreement with Frigi et al. [51], who described the Tunisian genetic pool by studying 15 Tunisian population groups distributed across the country, our cohort contained two North African haplogroups, M1 and U6, with a percentage of 7.1%. For the sub-Saharan lineages (haplogroup L), we found almost the same percentage (22.8%) as Frigi et al. [51] (23%). Sub-Saharan haplogroup L1 showed an increased risk of developing dementia in African Americans with lower plasma levels of Aβ [64]. Sub-Saharan haplogroup L3b was absent in patients' group and showed a significant p value for the control group. According to this observed value, Sub-haplogroup L3b could be considered as a protective factor in AD. This hypothesis could not be confirmed based on this small frequency of L3b in our samples. Further studies on largest cohort are mandatory to either confirm or decline this speculation.
Mitochondrial haplogroup T was among the Eurasian lineage in our cohort (9.8%) and was presented by both sub-haplogroups T1 and T2. In conflicting reports, The major observed component was the Eurasian lineages, followed by the sub-Saharan lineages and the North African lineages mitochondrial haplogroups U and T have been implicated as conferring both a decreased and an increased risk of developing AD [65]. In our case, the results showed no huge difference between patients and controls regarding haplogroup T (T1 + T2), which was present in 10.3% of patients and 9.6% of controls.
In many studies, evidence was found for an association between some specific haplogroups and AD, either as a risk factor or as a protective one. In our cohort, we have a mix composed of North African, sub-Saharan, and Eurasian lineages. This mixed population facilitates the evaluation of mitochondrial haplogroups in relation to AD. In our cohort, we have almost all mitochondrial haplogroups that were indicated in previous studies as being associated with AD. However, in our case, none of the haplogroups had a significant association with AD. These findings support the idea that haplogroup implication in AD is not strong enough to explain the risk of AD. Our study is in agreement with previous studies showing no definitive association of AD with a specific mitochondrial haplogroup [8,36,43,[65][66][67][68][69]. In fact, investigating the implication of mitochondrial haplogroups in AD pathology is an alternative way of determining the involvement of the mitochondrial genome. It is an approach to analysing the frequencies of mitochondrial polymorphisms that could affect AD pathology by altering mitochondrial respiration and causing overproduction of reactive oxygen species (ROS). This means that a group of individuals or a population with the same mitochondrial genotype has the same predisposition to apoptotic processes [28,29]. The analysis of the distribution of haplogroups in controls and AD patients is a way to determine whether a particular mitochondrial haplogroup is associated with the risk of AD by being more vulnerable to oxidative stress than other mitochondrial haplogroups.
However, the real involvement of mtDNA and its haplogroups in AD is still under discussion. Two theories are possible. The first one is that oxidative stress and mitochondrial dysfunction are responsible for neurodegenerative pathologies and have a primary role in AD. The second theory is that neurodegenerative pathologies are the consequence of cell death and that the decline in mitochondrial function is caused by other factors [70]. To our knowledge, no previous findings, except one study [41], were replicated to date.
Classically, it has been difficult to study the direct implication of mtDNA mutations in AD patients, and many studies have produced non-conclusive results [66]. Therefore, genetic investigations started to focus on single polymorphisms rather than haplogroups [36]. Since some mitochondrial mutations were identified in association with AD [71], we changed the strategy from haplogroups to mitochondrial SNPs. Our findings showed two large effects: A5656G (p = 0.03821, OR = 10.46) and A13759G (p = 0.03719, OR = 10.78); both SNPs were not mentioned before as a risk factor for AD.
According to the human phylogenetic tree, A13759G is located on the branch point of two sub-haplogroups, H11 and j2a2. A13759G is located in subunit 5 of the NADH dehydrogenase locus. This enzyme belongs to complex I (NADH dehydrogenase, NDU) of the electron transport chain (ETC) and represents one of the entry enzymes of cellular respiration and contributes to the production of ATP. The NADH dehydrogenase gene was mentioned before in relation to AD. In 1992, Lin et al. identified two mutations in the mtDNA of AD patients, located in subunit 2 of the NADH dehydrogenase gene, but these results were not confirmed by other studies [72,73]. Recently, a meta-analysis was carried out by Holper et al. [74] of both complex I and complex IV activities in many disorders, including AD. They provided strong evidence for impairments in both enzymes' activities in the blood and some brain regions. These mitochondrial enzymes are well documented in AD and seem to be affected by amyloid β peptide (Aβ) accumulation in mitochondria, inducing neural toxicity and inhibiting mitochondrial enzymes, finally leading to mitochondrial dysfunction [75][76][77]. Thus, it is clear from the literature that complex I has been implicated in neurodegenerative diseases, including AD, but no previous findings were found regarding A13759G in association with AD.
The second positive association was with A5656G. This polymorphism is the characteristic SNP of sub-haplogroup U5b1. Ridge et al. [78] showed that four clades (U5B1, U5B1B2, K1A1B, and K1A1B2A1) were associated with different endophenotypes of AD patients. Our results confirm and extend these previous studies, proving that sub-haplogroup U5b1 is probably implicated in AD. Furthermore, our findings show that the A5656G polymorphism may be the main factor in this sub-haplogroup's involvement in AD. This SNP is located in the non-coding region separating two structural genes of tRNAs: tRNA (Ala) and tRNA (Asn). Interestingly, a similar transition in position 3302 has been reported to disturb the processing of primary transcripts, causing a functional alteration [79]. Thus, the mechanism is still not clear, but it seems probable that this polymorphism located between two tRNA genes might cause a functional defect [80]. This theory is plausible because the A5656G transition was found in many diseases: tubulointerstitial nephritis [80], occipital stroke [81], and Parkinson's disease [82].
An increased mutation frequency was observed in the brains of AD patients [83]. Also, it has been suggested that mutations in mtDNA might change the age of AD onset and contribute to the process of neurodegeneration [72]. Unfortunately, only a few SNPs were described as being linked to AD risk; an example is the variant at position 4336 in the tRNA (Gln) gene [71,84]. The associated SNPs (A5656G 1 3 and A13759G) found in our study were not mentioned before as a risk factor for AD in previous studies. Since mtDNA encodes for mitochondrial subunits, transfer RNAs, and ribosomal RNAs, it is logical to think that mutations or clusters in the mtDNA might alter mitochondrial metabolism and lead to AD [85]. However, oxidative stress can explain the high level of mitochondrial mutations.
In other studies, brains of elderly subjects and AD patients showed an increased mutation rate in comparison with young subjects, indicating that mutations in mtDNA are not specific to the pathogenesis of AD. Mutations are also observed in ageing [86], so it is unclear whether an increased mutation rate is caused by AD or ageing.
At the same time, evidence of maternal transmission of AD was mentioned in many studies. Honea et al. [22] reported that persons with a maternal history of dementia showed an increased presence of a biomarker of AD compared with persons with paternal AD or no history of AD. Since mtDNA is maternally transmitted, it is logical to think about the implication of mtDNA in the pathogenesis of AD.
In view of all these hypotheses, many studies have been undertaken to understand the role of mtDNA in AD. Unfortunately, there are no conclusive results regarding the exact role of mtDNA in AD pathogenesis. Mutations in mtDNA can be a primary cause of AD, or the damages observed in mitochondria might be the result of AD.
Finally, in the second part of our study, we found an association of the APOE e4 allele (p = 0.000014) with AD, confirming previous findings indicating that the APOE e4 allele is a major genetic risk factor for AD in Tunisia. The frequencies of observed genotypes were 4.3%, 82.5%, and 13.2% for the APOE alleles ε2, ε3, and ε4, respectively. These values are almost the same as those reported by previous studies on the APOE allele frequencies in the Tunisian population [87].
The APOE gene is the strongest genetic risk factor for AD (R 2019; [88]. This gene encodes the apolipoprotein E protein, which is implicated in cholesterol and lipid transport. Several hypotheses have been put forward to explain the relation between APOE and AD risk. One idea is that there is a connection between APOE and mitochondrial dysfunction. This hypothesis was based on the observation that the apolipoprotein E4 isoform is differently folded compared with other isoforms. This alternative folding makes it more susceptible to proteolysis, leading to a small peptide. The resulting fragments constitute toxic bioactive elements that enter the cytosol and lead to cell death by disturbing the mitochondrial energy balance [9,89]. In addition, the interaction of APOE ε4 and mtDNA could be of interest for AD pathology. This idea was put forward in a cognitive study on the association of parental dementia with AD. The results revealed that non-demented APOE4 carriers with an AD-affected mother showed less memory performance than individuals with an AD-affected father with a higher rate of brain atrophy [90]. This finding could point to the possibility of an interaction between APOE4 and certain mitochondrial regions or haplogroups in modulating AD risk. Indeed, several previous reports have mentioned that some mitochondrial haplogroups (U and K) seem to neutralise the harmful impact of APOE4 [91]. One study proposed that some mitochondrial haplogroups have an influence on allele 4 of APOE, the major risk factor of AD. In particular, haplogroups K and U have been suggested to neutralise the effect of APOE4 [91]. This hypothesis could be investigated in future studies with a larger number of samples of the Tunisian population.

Conclusion
Recently, the implication of mtDNA in AD has been widely studied. Despite many findings indicating the presence of some possible risk factors for AD, such as mitochondrial haplogroups or SNPs, none has been proven so far. This difficulty of obtaining conclusive results might be due to differences between populations in the frequencies of mitochondrial polymorphisms, as caused by the variety of ethnic groups. Therefore, in our case, it was interesting to study the association of mitochondrial haplogroups with AD at the population level. However, we did not find any associated haplogroup. Then, in the second part of our investigation, we looked for the association of specific mitochondrial SNPs with AD and found two significant associations (A5656G and A13759G). This finding provides additional information on the role of mtDNA in the risk of AD and may contribute to the diagnostic process. Finally, our results were based on a small sample, so future investigations need to replicate our results in larger samples with the sequencing of the whole mtDNA.