Higher SBP in female patients with mitochondrial disease

Background: Previous research suggests that hypertension is more prevalent among patients with mitochondrial diseases. Blood pressure (BP) is linearly related to increased cardiovascular risk, and this relationship is strongest for SBP; nevertheless, studies on SBP and DBP in mitochondrial diseases have not yet been performed. Method: In a retrospective case–control study design, BP in mitochondrial disease patients was compared with BP in a population cohort. Secondly, using multiple linear regression, we examined blood pressure differences in various genetic mitochondrial diseases. Lastly, we explored additional predictors of BP in a subgroup with the m.3243A > G variant. Results: Two hundred and eighty-six genetically confirmed mitochondrial disease patients were included. One hundred and eighty of these patients carried the m.3243A>G mitochondrial DNA variant. SBP was 9 mmHg higher in female mitochondrial disease patients than in the general female population (95% CI: 4.4–13.3 mmHg, P < 0.001), whereas male patients had similar BP compared with controls. BP was not significantly different in patients with m.8344A>G and m.8363G>A, a mtDNA deletion or a nuclear mutation compared with m.3243A>G patients. Higher SBP was a predictor for left ventricular hypertrophy in the m.3243A>G subgroup (P = 0.04). Conclusion: Novel aspects of the role of mitochondrial dysfunction in blood pressure regulation are exposed in this study. Compared with the general population, female mitochondrial disease patients have a higher SBP. Left ventricular hypertrophy is more prevalent in patients with higher SBP. Clinicians should be aware of this to prevent hypertensive complications in mitochondrial disease patients.

INTRODUCTION P rimary mitochondrial diseases are a heterogenous group of disorders, both genetically and clinically. Mitochondrial diseases can be caused by pathogenic variants in either mitochondrial DNA (mtDNA) or nuclear DNA (nDNA) [1]. In all patients affected, the origin of the disease can be localized to the mitochondria, which are present in every human cell except red blood cells and play a central role in energy production and cellular metabolism [2]. Multiple organs can be affected, including skeletal muscles, the brain, heart, liver, kidney, eye, ear, pancreas and more [3]. Several factors influence the age of onset, severity, organ involvement and symptomatology of the disease, including the mutated gene, the base position affected and the heteroplasmy level (for mtDNA variants).
Several studies have identified pathogenic mtDNA variants as likely cause for hypertension in hypertensive family or cohort studies [4][5][6][7]. Interestingly, most of these studies found homoplasmic variants in patients with no other disease manifestations. More recent studies have suggested that hypertension might also be an underappreciated symptom in 'classical' mitochondrial syndromes, such as mitochondrial-inherited diabetes and deafness (MIDD, OMIM #520000), mitochondrial myopathy, encephalopathy, lactic acidosis, and stroke-like episodes (MELAS, OMIM #540000) and myoclonic epilepsy associated with ragged-red fibers (MERRF, OMIM #545000) [8][9][10]. The latter three studies included a heterogeneous group of patients but most had one of four common genetic defects: the heteroplasmic mtDNA variant m.3243A>G (located in MT-TL1, OMIM Ã590050), the m.8344A>G variant (in MT-TK, OMIM Ã590060), a mitochondrial DNA deletion or a variant in the mitochondrial DNA polymerase, polymerase g (encoded by the nuclear gene POLO, OMIM Ã 174763). All three studies described an increased prevalence of hypertension compared with the general population, when matched by age. Standardized prevalence ratios in these studies ranged between 1.3 and 2.5 (Supplementary Table 1, http://links.lww.com/HJH/B864). On the other hand, an earlier study among patients with MIDD, reported that blood pressure (BP) and hypertension prevalence were actually lower in mitochondrial disease patients than in age-matched and sex-matched controls with other causes of diabetes mellitus [11].
In these pioneering studies, some questions remain unanswered. Are SBP and DBP both elevated in mitochondrial diseases? Is BP elevated to the same degree in genetically different mitochondrial disease, when comparing by type of variant or by heteroplasmy level? Furthermore, studying hypertension rather than BP wastes power [12] and might lead to bias because of spurious identification of hypertension [13], especially in mitochondrial patients who visit a physician more often than the general population [14].
We hypothesized that mitochondrial disease patients have higher SBP and DBP in general, and that this effect would be more pronounced for a genetic subgroup of mitochondrial diseases. Using a retrospective study design, we analyzed BP in 286 patients with genetically proven mitochondrial diseases. BP was compared with a national reference cohort using case-control matching. Secondly, we studied BP differences among distinct mitochondrial diseases grouped by genotype. Finally, we explored the correlation between BP and mitochondrial disease characteristics in the m.3243A>G subgroup, including heteroplasmy levels, presence of diabetes mellitus and presence of left ventricular hypertrophy (LVH).

Patient inclusion
We used a retrospective design to study two patient cohorts. For the first cohort, we retrospectively identified and included patients of at least 16 years of age with genetically proven mitochondrial disease, who visited the department of Internal Medicine of the Radboudumc between January 2005 and January 2020. Patients who were not capable of making an informed decision about consent for study participation were excluded. BP was measured at the moment of their first referral to the Radboudumc. Data on all other variables was collected from the same date in most cases, and if not possible, the measurement closest to the BP measurement was taken (maximized to 1 year). For the second cohort (acronym Mitostraat), we collected BP data and all other data during an in-patient, 4-day multidisciplinary evaluation program (as part of usual care, usually planned at or shortly after the first presentation). These patients visited the Radboudumc between March 2015 and March 2020. For both cohorts, we collected data on the following variables: sex, age in year of measurement, age at diagnosis, clinical diagnosis, the exact pathogenic mtDNA or nDNA variant, smoking status, heteroplasmy levels in leucocytes and urine epithelial cells, diabetes status, hypertension status, presence of kidney disease, presence of (micro)albuminuria, SBP, DBP, height, weight, BMI, estimated glomerular filtration rate (calculated with CKD-EPI), antihypertensive drug treatment at the time of the BP measurement and presence of LVH on electrocardiogram/echocardiogram.
As a reference cohort, we used anonymized BP data obtained by NLdeMaat, a national public health survey among 3847 men and nonpregnant women aged 30-70 years, executed by the National Institute for Public Health and the Environment (RIVM) [15].

Measurement of variables
In the out-patient cohort, manual sphygmanometer BP measurements were performed according to standard guidelines [16]. During the visit, three measurements were performed with a manual BP monitor (Maxi Stabil 3, Welch Allyn GmbH & Co. KG, Germany) and the average of the last two was recorded in the patient record. In the in-patient cohort, BP was recorded with an automated BP monitor (Langzeitblutdruckmessgerät Mobil-O-Graph NG, Köln, Germany). In the national reference cohort, BP was recorded with an automated monitor, Omron M6 (Omron, Kyoto, Japan).
Data was collected from electronic health records and entered in Castor EDC [17].

Genetic assessment
Mitochondrial DNA was obtained from urine epithelial cells by centrifuging urine sediments for 10 min at 3000 rpm, the pellet was washed with a phosphate-buffered saline. Additionally, DNA was isolated from leukocytes in peripheral venous blood using a salting-out method. A commercially available DNA isolation kit (Gentra Puregene Blood kit, #158389; Qiagen, Venlo, the Netherlands) was used to extract the DNA and the percentage of mutant mtDNA was determined using Pyrosequencing technology (Pyrosequencing, Uppsala, Sweden) [18]. The pyrosequence reaction of the m.3243A>G variant had a precision of 1.5%, and the mutation was detected from a heteroplasmy level of 5%. The detection limit for the m.3243A>G variant was determined by serial dilution of a sample containing this mutation with wild type mtDNA [19]. We evaluated pathogenicity by taking into account segregation information, heteroplasmy levels, variant allele frequency, in silico pathogenicity predictions and biochemical and histochemical results.

Definitions
Hypertension was defined as SBP at least 140 mmHg, DBP at least 90 mmHg [16], or if the medical patient record mentioned a personal history of hypertension. A mitochondrial disease was considered genetically proven when a pathogenic or likely pathogenic mtDNA or nDNA variant was identified. Mean arterial pressure (MAP) was calculated from SBP and DBP by the following formula: MAP ¼ (1/3ÃSBP) þ (2/3ÃDBP). We defined proteinuria/(micro-)albuminuria as an albumin/creatinine ratio of greater than 3. 5 mg/mmol or a urinary albumin excretion of more than 30 mg/l in spot urine. Individuals were counted as smokers if they reported smoking at the time of presentation, regardless of frequency. Past smokers were counted as nonsmokers.

Statistical analyses
We described clinical characteristics with descriptive statistics. Our primary analyses were comparison of BP in a national reference cohort versus mitochondrial disease patients and among genetically different mitochondrial disease subgroups. As secondary analyses, we explored whether heteroplasmy levels and several mitochondrial disease manifestations (e.g. syndrome classification, presence of diabetes, presence of LVH and presence of [micro-) albuminuria] were correlated with BP in the m.3243A>G subgroup. We imputed missing data under the assumption of data missing at random (Supplementary Material, http:// links.lww.com/HJH/B864). We performed sensitivity analyses on the original, unimputed data to check the effect of the imputation.
In our first primary analysis, we control-matched mitochondrial disease patients by age, gender and BMI (Supplementary Material, http://links.lww.com/HJH/B864). We performed a two-tailed, unpaired t-test assuming unequal variances to test for any significant differences in BP, age and BMI between cases and controls. We used a Pearson chi-square test to explore differences in hypertension rates, cardiovascular disease, diabetes and (micro) albuminuria.
For the second primary analysis and secondary analyses, we used multiple linear regression analyses. We included known risk factors for high BP in each model as covariates regardless of significance; we, therefore, included age, agesquared (age 2 ) and BMI as continuous variables, and included sex and smoking status as binary variables. Age 2 was included as age is known to have a strong but not linear relationship with BP. We also included the cohort as binary variable (automatic measurement cohort or manual measurement cohort). Dummy variables were created for each genotype subgroup (analysis 2), the largest subgroup (m.3243A>G) was chosen as reference group. We assessed significance of the correlation between each subgroup and BP by stepwise selection, and subsequently entered all subgroups together to display an estimate of effect sizes and directions in nonsignificant subgroups.
For each regression analysis, before interpreting results, we examined whether the dataset met basic assumptions for correct interpretation of multiple regression. We, thus assessed normality of the residuals, homoscedasticity, absence of multicollinearity and in case of a continuous independent variables, linearity of the relationship with the outcome.
As part of sensitivity analyses, we examined whether our results were different from the same analysis on the unimputed (original) data.
As the observed BP cannot be used for the analysis directly, we performed a transformation on the BP values of all patients who were treated with antihypertensives, as suggested by Tobin et al. [12]. Thus, we obtained the imputed BP, X i , from the observed BP, Y i , by applying the following formula: On the basis of previous research [12,20], we set the constant to 10 mmHg for imputation of the underlying SBP and 5 mmHg for imputation of the underlying DBP. For sensitivity analyses, we varied the constant factor between 6 and 15 mmHg for SBP, and between 3 and 10 mmHg for DBP.
The regression formulas can be found in the

Cohort description
Two hundred and eighty-six patients with genetically proven mitochondrial diseases were included, as well as 3847 individuals from a national reference population (Supplementary Table 2

Female patients have higher blood pressure than female controls
To correct for the marked differences in age, gender and BMI between the two cohorts, case-control matching was performed. No significant differences were detectable between the two cohorts after matching for age, gender and BMI (Table 1). SBP was 5 mmHg higher in mitochondrial disease patients than in individuals from the national population cohort (95% CI 2-8.9 mmHg; Table 1). MAP was significantly higher too in mitochondrial disease patients (3 mmHg, 95% CI 0.7-5.7 mmHg) and DBP showed a similar trend (2 mmHg, 95% CI -0.2 to 4.2 mmHg). When exploring sex-differences, we found that this was solely attributable to higher BP in female patients (Fig. 1 We also explored differences in hypertension prevalence, antihypertensive use and several hypertensionrelated comorbidities ( Table 1). The hypertension prevalence was 44% in mitochondrial disease patients versus 37% in control individuals (prevalence ratio 1.2, x 2 test P ¼ 0.09). Use of at least one antihypertensive, presence of diabetes and presence of (micro-)albuminuria were all significantly higher in mitochondrial disease patients (respective odds ratios: 5.2, 19 and 6.3; for all P < 0.001; Table 1). Mitochondrial disease patients also had a higher incidence of cardiovascular disease (odds ratio 2.3; 95% CI 1.2-4.6), specifically female patients (Supplementary Table 6, http://links.lww.com/HJH/B864).
As sensitivity analyses, the t tests and x 2 test were repeated on the unimputed data, yielding very similar results (data not shown). Also, we studied the effect of choosing different constant factors as treatment correction (between 6 and 15 mmHg for SBP and between 3 and 10 mmHg for DBP). Higher constant factors increased the difference between the two cohorts (Supplementary Table  7, http://links.lww.com/HJH/B864).

Genotype is not correlated with blood pressure
Linear regression analyses were performed to study the effect of the genetic mitochondrial disease classification; potentially confounding variables were included. Higher age was significantly associated with higher SBP, DBP and MAP (Fig. 1, Table 2 and Supplementary Table 8, http:// links.lww.com/HJH/B864). Higher BMI was also significantly associated with higher BP (0.36-1.17 mmHg per 1point increment in BMI), with the strongest relationship for SBP (Fig. 1). Male sex was associated with higher BP but this was only significant for DBP (P ¼ 0.034). A difference was observable between the manual measurement cohort and automatic measurement cohort, with the manual measurement cohort having a higher BP (7 mmHg difference for SBP). Stepwise selection of the genetic diagnosis as additional covariate left only the 'mtDNA mutation, remainder' subgroup as a significant variable for BP. Analysis ofBP residuals per individual could not pinpoint, which DNA mutations were most associated with higher SBP or lower

DBP* (mmHg)
FIGURE 1 Blood pressure rise with age and BMI. Mitochondrial disease patients were matched 1:1 with individuals from a national population survey (n ¼ 231 each) by sex, age and BMI before plotting average blood pressure against age (panels a and b) and BMI (panels c and d) with a simple linear regression. ÃBlood pressure was corrected for use of antihypertensives [12]. Table 9, http://links.lww.com/ HJH/B864). As sensitivity analyses, the same linear regression was performed on the unimputed original data. This gave similar results (data not shown). Varying the constant factor to correct for antihypertensive treatment effect did not affect the outcomes (data not shown).

Higher blood pressure is associated with left ventricular hypertrophy in the m.3243A>G subcohort
To exclude any possible genotype -phenotype interaction effects, we further studied the subcohort of patients with the m.3243A>G variant for predictors of higher blood pressure. Positive correlations of BP with age, BMI and male sex were observed (Table 3 and Supplementary Table  10, http://links.lww.com/HJH/B864). These were similar in size and direction as the associations observed in the total group of mitochondrial disease patients. None of the subgroups with one of three predefined clinical diagnoses (MELAS, myopathic or dormant carrier) had a significantly different blood pressure from the MIDD subgroup. The comorbidities diabetes, (micro)albuminuria, heteroplasmy and cardiovascular disease were not significantly associated with blood pressure either, nor were urinary or blood heteroplasmy levels. Following a stepwise approach that would include significant predictors, only LVH -as assessed by electrocardiogram or cardiac ultrasound -was retained in the model. Entering LVH as separate variable in the model showed 6.6 mmHg higher SBP in patients with LVH (95% CI 0. 24-12.9). A logistic regression to investigate factors that determine LVH in our cohort indeed classified higher SBP as significant predictor, as well as lower BMI (Supplementary Table 11, http://links. lww.com/HJH/B864). The odds ratio of having LVH was 1.32 for a 10 mmHg rise in SBP (95% CI 1.01-1.72).
As part of sensitivity analyses, we performed the same analyses on the unimputed data. Results were very similar (data not shown). Also, we investigated whether changing the constant factor added in case of antihypertensive treatment influenced the results. Higher estimations of the treatment effect (up to 15 mmHg) yielded slightly larger and more significant effect sizes (up to þ8.2 mmHg for LVH as predictor of SBP, P ¼ 0.015, data not shown).

DISCUSSION
Using a retrospective case -control design, this study provides evidence that female patients with a mitochondrial disease have a higher SBP, compared with the general female population. Multiple linear regression analyses show that BP differences between different mitochondrial diseases are small, and in our study, nonsignificant. Lastly, SBP was associated with LVH in patients with the m.3243A>Gvariant.
Analysis of SBP and DBP separately allowed to discern a divergent relationship for SBP and DBP. In our cohort, female mitochondrial disease patients have a higher SBP than the general female population. SBP elevation is strongly associated with cardiovascular outcomes, even more than DBP elevation [21]. Furthermore, the WHO has estimated that worldwide, 13% of all mortality can be  attributed to high blood pressure [22]. Female mitochondrial patients in our study had a higher incidence of cardiovascular disease in comparison to the general female population. Careful blood pressure monitoring and treatment of elevated blood pressure would thus be expected to prevent morbidity and mortality, particularly in women. Novel hypotheses on the pathophysiology of hypertension in mitochondrial disease can be drawn from this study. The more pronounced effect on SBP suggests an increased stiffness of the large conduit arteries or narrowing of smaller, muscular arteries and arterioles [23]. Studies on vascular function in mitochondrial disease are limited but some provide data to support such a hypothesis. For instance, POLO-deficient mice exhibited an increased arterial stiffness [24], whereas hypertensive mitochondrial disease patients showed an increased total peripheral resistance when values were normalized to body surface area [10]. Furthermore, pulse wave contour analysis in a patient with the m.3243A>G variant demonstrated decreased capacitive and oscillatory compliance [25], suggesting both stiffness of the large conduit arteries and narrowing of small arteries and arterioles [26]. Lastly, MELAS patients appear to have a deficiency of nitric oxide [27], which would be expected to lead to increased resistance at muscular arteries.
Why only female patients would suffer from increased stiffness or impaired vascular relaxation is unclear. Sex differences in the pathophysiology of hypertension has been a subject of extensive study and debate in the normal population [28], but whether this would translate to the mitochondrial disease population remains to be determined. For a link between sex differences and altered mitochondrial function specifically, only two rodent studies are to our knowledge available [24,29]. Studies that investigate the pathophysiology of high BP in mitochondrial disease patients are, therefore, needed.
A recent study has suggested that the type of genetic defect plays a significant role in conferring the risk of hypertension in mitochondrial disease patients [9]. We wanted to reinvestigate this after careful correction for con-founders, such as age, sex, BMI and smoking, to prevent identification of spurious correlations. Multiple linear regression with correction for such confounders showed that BP did not correlate significantly with the genetic diagnosis. Patients with the m.8344A>G and m.8363G>A, a mtDNA deletion or a nuclear mutation did not have significantly different BP compared with patients with the m.3243A>G variant. Only patients from a group with the 'remainder' of pathogenic mtDNA variants (i.e. with a variant not classifiable in one of the other groups) had nominally significant higher SBP and lower DBP.
The absence of further correlations between BP and genetic diagnosis is unexpected in the light of the findings by Chong-Nguyen et al. For example, they reported a hypertension prevalence of AE 60% in patients with the m.3243A>G and m.8344A>G variant while observing only a AE 25% hypertension prevalence for patients with single deletions (reported P < 0.001) [9]. A possible explanation is that no correction was performed for differences in age, sex and BMI in that specific analysis. Furthermore, studying hypertension as outcome variable might overestimate the true effect because of spurious identification of hypertensive cases in patients frequently visiting a doctor [13].
Interestingly, we found an independent association between LVH and SBP in patients with the m.3243A>G variant; a 10 mmHg increase in BP was associated with an odds ratio of 1.3 for having LVH. Our retrospective study design was not equipped to analyze the nature of this relationship, but in general, LVH is considered to be a result of hypertension [30]. Given the high incidence of LVH in mitochondrial diseases, adequate monitoring and treatment of high BP might thus be a valuable strategy to lower the incidence of LVH in mitochondrial disease patients.
We did not find a correlation with BP for several other factors associated with BP in the normal population, including diabetes, albuminuria or lactate [31][32][33].
The correlation of heteroplasmy levels with hypertension in patients with the m.3243A>G variant has been a matter of debate. In our study, BP was not significantly correlated with heteroplasmy levels in leukocytes or urine epithelial cells. Two previous studies [8,9] have reported a small and marginally significant correlation of heteroplasmy levels with hypertension whereas another study was not able to confirm this [10]. A difficulty shared by all studies, including ours, is that heteroplasmy was not measured in the vasculature or kidneys, that is, in tissues most relevant for hypertension. Nevertheless, our data suggest that urinary or blood heteroplasmy levels cannot be used as a predictor for blood pressure.
Important strengths of our study are the careful correction for confounders, the investigation of BP instead of hypertension as outcome variable, the detailed characterization of the patients and the large sample size of the m.3243A>Gsub-cohort. The retrospective nature of the study also limits expectation bias.
The most important limitation is that the type of BP device could not be standardized because of the retrospective nature of the study. As manual BP measurements generally yield higher BP values in our center, we might have overestimated the BP in the mitochondrial disease patient group. Furthermore, antihypertensive use by a subset of patients necessitated us to estimate the underlying blood pressure in these patients [12]. Additionally, we cannot exclude a modest temporal bias as it was not possible to collect the patient cohort at exactly the same time as the population survey. A last limitation is the relatively small sample size of some subgroups in our study. An example is the small number of patients with nuclear POLO variants compared with the study by Pauls et al. [10].
In conclusion, our study reveals novel aspects of the role of mitochondrial dysfunction in blood pressure regulation. It brings nuance to the observations in earlier studies by showing that specifically female mitochondrial disease patients have a higher SBP compared with the general female population. Furthermore, our data suggest a possible role for elevated BP in the higher incidence of left ventricular hypertrophy often observed in mitochondrial disease patients. Clinicians should be aware of this and consider routine screening of BP in mitochondrial disease patients.