Neurological manifestations in adult patients with the m.3243A>G variant in mitochondrial DNA

Subjects

We have previously ascertained 17 pedigrees with m.3243A>G in the province of North Ostrobothnia, Finland. Eleven probands were identified in a population-based epidemiological survey, the setting of which has been described elsewhere.10 In addition, six other probands were encountered in the following years at the Neurology Outpatient Clinic of Oulu University Hospital that is the only provider of neurological services in the province and that has had a specific interest in mitochondrial diseases. The probands, their mothers and one complete sibship from each pedigree were intended to be included in the present study. There were 122 adult subjects in the matrilineal sibships. Non-participants numbered 71 subjects (35 men and 38 women) including 39 subjects who had died before the start of the study and 32 subjects who were excluded from the study (11 subjects not living in the province; 5 subjects not harbouring m.3243A>G; 16 subjects not volunteering to the study). Finally, the 17 probands and 34 adult maternal relatives (18 men and 33 women) aged 18–72 years (45±14 years) with the m.3243A>G variant were recruited prospectively to the study. The maternal relatives included both symptomatic and asymptomatic carriers.

Clinical evaluation

The subjects were thoroughly interviewed, and they underwent a complete neurological examination. The phenotype was ascertained by clinicians with expertise in mitochondrial disorders (KM and MK). Standard clinical evaluations were performed on all patients included in the study. The age of onset of symptoms was determined based on medical history or appropriate medical files. The date of the first abnormal finding in clinical examination was used to define the age of onset if the history was inconclusive. The diagnoses were established or confirmed on clinical grounds for for myopathy, ophthalmoplegia, retinopathy, ataxia, epilepsy, migraine-like headache and stroke. The diagnoses of peripheral neuropathy,11 SNHI12 and hypertrophic cardiomyopathy13 were based on ancillary examinations as described previously. Trail making test was used to assess cognitive impairment that was diagnosed if the time to complete the test exceeded the mean of age-stratified normative data by two SDs.14 Short stature was defined as a height below two SDs of the mean height of the Northern Finnish population (men <165 cm, women <152 cm).15 Diabetes mellitus was defined according to WHO recommendations for the diagnostic criteria for diabetes and intermediate hyperglycaemia.16 Severity of disease was estimated using modified Rankin Scale (mRS, 0–5)17 and Clinical Global Impression of Severity of Illness Rating (CGISIR grade 1–7).18

Molecular methods

The m.3243A>G heteroplasmy was determined in the buccal epithelium of 41 subjects, in skeletal muscle of 37 subjects, and in blood of two subjects by using an Apa I restriction digestion of a 390 bp mitochondrial DNA (mtDNA) fragment (spanning between positions m.3150 and m.3550) amplified by PCR in the presence of ³⁵S-ATP.19 The intensities of the fragments were determined by using autoradiography. Heteroplasmy was determined both in the muscle and buccal epithelium of 31 subjects, and it was found to be correlated, so that muscle heteroplasmy=22.3+0.86×buccal epithelium heteroplasmy (R²=0.709, p<0.0001). This equation was used to estimate muscle heteroplasmy given the buccal heteroplasmy in the 12 cases, where only buccal sample was available. The estimated values were then used to impute the 12 missing values and together with the 37 values of muscle heteroplasmy they composed the variable of imputed muscle heteroplasmy that was used in subsequent statistical analyses.

mtDNA haplogroups were determined by restriction fragment analysis of informative polymorphisms.20 Twenty-eight patients belonged to mtDNA haplogroup UK, 12 to haplogroup H or V, six to haplogroup T, four to haplogroup I and one patient to haplogroup Z. The D loop was amplified in a fragment spanning nucleotides 15 975–725, and the sequence between nucleotides 16024 and 400 was determined by use of forward primers with their 5’ nucleotides at positions 15975 and 16449, respectively.20

Meta-analysis

The identification and review of publications for the meta-analysis were carried out according to the guidelines in the PRISMA 2020 checklist (https://www.prisma-statement.org/prisma-2020-checklist). We searched articles until 31 January 2023 in PubMed database (figure 1). The following text words were used to search: (1) “A3243G”; (2) “m.3243A>G”; (3) “(3243 clinical) OR (3243 phenotype) AND mitochondrial NOT (case report) NOT m.3243A>G” and (4) “MELAS AND clinical”. In addition, a search was performed using the terms: (5) “MELAS(title) NOT (case report)(text word)”. In searches (1)–(3) the date limit of 1 January 1990 was applied, and in searches (4) and (5), the limit was set at 1 January 1984. Two authors (KM and MK) were blinded to each other’s work and independently assessed the titles and abstracts of the retrieved articles for eligibility.

Request permission

We included studies that reported case series with more than 10 individuals who were 18 years or older. We set the limit at 10 patients in order to avoid inclusion of case reports and small case series that may bring descriptions of more extreme phenotypes into the dataset. Studies were excluded if the cases were paediatric, or if cases had been ascertained based on the MELAS phenotype. Studies reporting phenotypes from samples combining children and adults were included only if the mean age of the cases was more than 18 years or if adult cases could be selected from the data. In addition, it was required that the study reported at least 6 out of the 13 phenotypes listed above. The lower limit was set at six phenotypes in order to avoid inclusion of publications that have examined selected phenotypes, such as diabetes and deafness, or stroke-like episodes. Focusing on a single phenotype carries the risk of ignoring other phenotypes. No restrictions based on sex or ethnicity were applied. Twenty-five studies (Dataset-25) were identified (table 1). Stroke-like episodes and epilepsy are frequent manifestations in the MELAS syndrome5 and, therefore, we created a second dataset, where studies were excluded if the frequency of either phenotype was ≥50%. This exclusion was aimed to avoid preferential inclusion of the severe MELAS phenotype. Seventeen studies were included in Dataset-17.

Publication bias across individual studies was assessed using funnel plot inspection and the Begg’s rank test.21 Heterogeneity between included studies was assessed with I² statistics.22 The random effects model was employed in the meta-analysis.

Statistical analyses

Demographic and clinical data were summarised using descriptive statistics, such as Mann-Whitney U test and Spearman’s rank correlation test. χ² test was used to analyse associations between categorical variables. Binary logistic regression analysis was used to model the relationship between an outcome variable and independent variables using the enter method. Kaplan-Meier estimator was used to analyse the probability of phenotypes over time, where onset of the phenotype was the outcome, and subjects were censored at the age of death or at present age. Statistical analyses were performed with the SPSS Statistics V.29 software package for Windows, and the package meta in R (V.4.3.1) was used for meta-analysis.23 Difference was considered significant if p<0.05.

Neurological manifestations

Fifty-one patients with m.3243A>G (18 men) were examined neurologically (table 2). Forty-two patients had at least one neurological manifestation, while nine patients (18 %, mean age 39 years, range 28–56 years) were free of neurological manifestations. The mean age of onset of the first neurological manifestation was 31 years (range, 0–70 years). The most common neurological phenotypes were SNHI, cognitive impairment and myopathy, whereas only six patients (12 %) had suffered from stroke-like episodes (table 2). None of the patients was diagnosed with dementia, but 53% of the patients were diagnosed with cognitive impairment. In addition, cardiomyopathy, short stature and diabetes were common. Diabetes and SNHI were associated (p=0.0002) and other significant associations were between cognitive impairment and SNHI (p=0.003), between stroke-like episodes and epilepsy (p=0.006) and between cognitive impairment and retinopathy (p=0.009).

Functional capacity

Twenty-three patients with m.3243A>G had no significant disability (mRS 0 or 1), 18 patients had mild disability (mRS 2) and 10 patients had moderate disability (mRS 3). The clinical impression of severity of illness gave grades 1 or 2 (normal or borderline ill) for 17 patients, grade 3 (mildly ill) for 15 patients and grades 4–5 (moderately to markedly ill) for 19 patients. There was a good correlation between the two assessments (p<0.001). Two patient groups with mRS 0–1 or mRS 2–3 were formed for binary logistic regression analysis, where the variables HI, myopathy and stroke-like episodes were selected to represent the correlating phenotypes. The analysis yielded a model with R² 0.477 and classification 72.5%.

Factors contributing to the clinical severity of disease

Imputed muscle heteroplasmy, age, sex and mtDNA haplogroup were found to contribute to the clinical severity of disease. Heteroplasmy correlated with Rankin scores (p<0.001) and CGISIR grading (p<0.001) and patients with mild or moderate disability had higher heteroplasmy than those with no significant disability (p=0.009). The two groups did not differ in age (p=0.52).

Imputed muscle heteroplasmy was higher in patients with encephalopathic phenotypes compared with those without the phenotype (table 1 in online supplemental file 1). Differences between the groups in age were not conspicuous, but the penetrance of m.3243A>G increased with age. Late-onset phenotypes included myopathy, cognitive impairment and peripheral neuropathy, while SNHI manifested earlier in life. Kaplan-Meier analysis showed that the estimated proportion of myopathy and SNHI at age 70 years was 100 %, while that of epilepsy was 22% and that of stroke-like episodes were 11% (figure 2).

Request permission

Men were more severely affected than women as estimated by means of Rankin score (p=0.003, χ² test) and CGISIR grade (p=0.001, χ² test). All the 18 men had at least one neurological manifestation, while nine women (27%) did not have any manifestations. Among the phenotypes, however, only the frequency of cognitive impairment (p=0.022, χ² test) and stroke-like episodes (p=0.017, χ² test) were higher among men. mtDNA haplogroup UK was associated with no significant disability (p=0.004, χ² test).

Neurological phenotype as a dependent variable and imputed muscle heteroplasmy, age, sex and mtDNA haplogroup as covariates were imported in binary logistic regression analysis. The ORs suggested that higher heteroplasmy more than higher age was a modest explanatory covariate for all phenotypes (table 2 in online supplemental file 1), whereas male sex was a significant covariate for stroke-like episodes and mild or moderate disability and female sex for migraine. Haplogroup UK was an explanatory variable for the low frequency of most neurological phenotypes, significant only for epilepsy, and the phenotype of no significant disability. Regression analyses using muscle heteroplasmy and buccal epithelium heteroplasmy separately as covariates decreased the analytical power and did not provide additional information.

Meta-analysis

Twenty-five studies including 1314 patients with m.3243A>G (table 1) were identified for meta-analysis. The I² values were 0.5≤I²≤0.9 suggesting substantial heterogeneity between the studies in all phenotypes and inspection of the phenotype frequencies revealed that stroke-like episodes or epilepsy were reported at a frequency of ≥50% in eight studies. Such high proportions suggested preferential inclusion of patients with the MELAS phenotype and, hence, the eight studies were removed creating a second dataset of 17 studies with 996 patients for the meta-analysis. Inspection of funnel plots and subsequent Begg’s rank test revealed that possible bias remained in stroke-like episodes (p=0.048) and in epilepsy (p=0.182), and substantial heterogeneity prevailed between the 17 studies according to the I² values.

Meta-analysis was performed separately on Dataset-25 and Dataset-17. Peripheral neuropathy and retinopathy were rare phenotypes with estimated frequencies lower than 20%, and HI and myopathy belonged to the common phenotypes with frequencies higher than 35% (table 3; Forest plots are shown in online supplemental file 2). Except for ataxia, the estimated frequencies of the phenotypes were higher in Dataset-25 than in Dataset-17. The difference was less than 10% in most phenotypes, but the frequency of epilepsy was 45% and that of stroke-like episodes was 49% higher in Dataset-25.

We then estimated confident ranges for the reported proportions of affected patients in Dataset-25 by defining limits that were set at mean proportion±1 SD. Fifty-two of the 186 proportions (28 %) were outliers, and deviation was most common in the phenotypes of migraine, stroke-like episodes, ataxia and epilepsy. In our study, we found deviation above the upper limit in the proportion of cognitive impairment and deviation below the lower limit in that of stroke-like episodes and ataxia.