Simplifying the clinical classification of polymerase gamma (POLG) disease based on age of onset; studies using a cohort of 155 cases

Variants in POLG are one of the most common causes of inherited mitochondrial disease. Phenotypic classification of POLG disease has evolved haphazardly making it complicated and difficult to implement in everyday clinical practise. The aim of our study was to simplify the classification and facilitate better clinical recognition.


Summary
Background: Variants in POLG are one of the most common causes of inherited mitochondrial disease. Phenotypic classification of POLG disease has evolved haphazardly making it complicated and difficult to implement in everyday clinical practise. The aim of our study was to simplify the classification and facilitate better clinical recognition.
Methods: A multinational, retrospective study using data from 155 patients with POLG variants recruited from seven European countries.
Results: We describe the spectrum of clinical features associated with POLG variants in the largest known cohort of patients. While clinical features clearly form a continuum, stratifying patients simply according to age of onset-onset prior to age 12 years; onset between 12 and 40 years and onset after the age of 40 years, permitted us to identify clear phenotypic and prognostic differences. Prior to 12 years of age, liver involvement (87%), seizures (84%), and feeding difficulties (84%) were the major features. For those with onset between 12 and 40 years, ataxia (90%), peripheral neuropathy (84%), and seizures (71%) predominated, while for those with onset over 40 years, ptosis (95%), progressive external ophthalmoplegia (89%), and ataxia (58%) were the major clinical features. The earlier the onset the worse the prognosis. Patients with epilepsy and those with compound heterozygous variants carried significantly worse prognosis. Conclusion: Based on our data, we propose a simplified POLG disease classification, which can be used to guide diagnostic investigations and predict disease course.

K E Y W O R D S
Alpers syndrome, epilepsy, mitochondrial disease, POLG, stroke-like episodes

| INTRODUCTION
Mitochondria are intracellular organelles found in almost all human cells. Their key function is the production of adenosine triphosphate through the process of oxidative phosphorylation performed by the mitochondrial respiratory chain (MRC). The MRC comprises more than 90 proteins organised into five complexes (I-V). Mitochondrial DNA (mtDNA) codes for 13 proteins while the remaining proteins are encoded by nuclear DNA. 1 The enzyme that replicates and repairs mtDNA, polymerase γ, 2 is a heterotrimer comprising a catalytic subunit (POLG) and two accessory subunits (POLG2). Mutations in POLG (OMIM * 174763), the nuclear gene encoding the catalytic subunit, interfere with mtDNA maintenance. 2,3 Variants in POLG are the single most common cause of inherited mitochondrial disease. 4 The first POLG variant associated with disease was described in a family with autosomal dominant progressive external ophthalmoplegia (PEO 5 ), but since then, more than 190 disease-causing variants have been identified (http://tools.niehs.nih.giv/ polg). POLG variants are associated with a wide spectrum of overlapping phenotypes ranging from devastating fatal neonatal disease to a mild late onset disease with myopathy and PEO. A summary of the major POLG-related phenotypes reported in the literature 4,6-22 is provided in Table 1.
The clinical features of POLG disease are extremely heterogeneous making early clinical recognition challenging. The increasing numbers of terms that have been used to describe the clinical phenotypes (Table 1) have added to this confusion.
The clinical reports published so far were based on small numbers of patients and did not describe the clinical spectrum through the whole life span. Longitudinal studies describing the natural history of the disease in a large cohort of patients are still lacking.
In this study, we aimed to describe the natural history of POLG disease in the largest cohort of patients with confirmed POLG variants, focusing on the clinical features and the biomarkers which may predict the longterm prognosis. We aimed to provide a simpler clinical classification to facilitate early clinical recognition of patients with POLG disease.

| Study design, population, and data collection
We performed a multinational, retrospective study of patients from 12 centres in seven European countries: Detailed clinical, biochemical, muscle biopsy, neurophysiological, neuroimaging, and genetic data were obtained by using an electronic-case report form completed by the responsible investigator(s) at each centre and reviewed by the study-principal investigator (O.H.).
The date of disease onset was defined by the date of the first symptom(s) requiring medical evaluation. End of follow up was defined as the date of the patient's last visit to the follow-up centre or death. Available longitudinal data, both at disease onset and later during the disease course, were collected. Liver involvement was defined by the presence of two or more of the following parameters in at least two different time points; elevated aspartate aminotransferase, gamma-glutamyltransferase, bilirubin or ammonia, low serum albumin, or pathological histological findings of liver biopsy. The presence of anaemia and abnormal cerebrospinal fluid (CSF) protein and/or albumin was identified as described in previous publications. 13,14,24 We use the recent International League Against Epilepsy (ILAE) classification 25 for seizure classification.

| Data and statistical analysis
Detailed descriptive data analysis was performed on the entire study cohort using SPSS (Statistical Package of Social Sciences), Version 23.0. A two sided P value less than .05 was considered to be statistically significant. Mosaic plots was performed by using R (The R foundation for statistical computing), version 3.6.1.
In order to simplify the clinical classification patients were grouped according to the age of disease onset into three groups: (a) those with disease onset prior to the age of 12 years (before puberty), (b) those with disease onset between 12 and 40 years, and (c) those with disease onset after the age of 40 years. The age of onset of each individual symptom was recorded and classified according to these three defined age-groups. Correspondence analysis was performed to examine the relationship between two variables (groups of patients who were classified according to the age of onset as described and the age of onset of each individual symptom) graphically in a multidimensional space; this allowed examination of the clustering of symptoms around each age group. Further, mosaic plots was performed to study the differences between the above mentioned groups. The study cohort was also classified according to the presence or absence of epilepsy, regardless the age of onset.
For survival analysis, the end-point was time to death which was defined as the time in months from the date of disease onset to the date of death. Univariate survival analysis was performed using log-rank test (Kaplan-Meier) to compare differences in survival time between categories.

| Demography
One hundred and fifty-five patients, (males n = 76 [49%], females n = 79 [51%]) with confirmed pathogenic POLG variants were identified. Seventy-six were diagnosed in Norway, 44 in Sweden, 19 in the United Kingdom, 8 in Finland, 5 in Denmark, 2 in The Netherlands, and 1 in Spain. The majority of patients were Northern European (n = 146), while three patients were from Iraq, two from Cyprus and one from Croatia, Pakistan, Spain, and the United Arab Emirates.

| Age-related clinical features
We found clear evidence that the clinical features of POLG disease form a continuum ( Figure 1) rather than distinct phenotypes (Table 1). Nevertheless, by grouping the patients into three groups according to the age of onset (early, juvenile/adult, and late onset groups), we could identify clear phenotypic and prognostic differences (Table 3). To confirm this finding, correspondence analysis was performed and demonstrated clear clustering of the symptoms around the three different age groups as illustrated in Figure S1. Further, mosaic plots showed there was a statistically significance differences in the phenotypes between the above mentioned groups ( Figure S2).

| Laboratory, muscle biopsy, and neurophysiological findings
The percentage of those with raised lactate in serum and CSF at disease onset was 35% (n = 29/84) and 40% (n = 19/47), respectively. Abnormal elevated CSF protein at disease onset was reported in 68% (n = 44/68) of patients. In muscle biopsies, the presence of ragged-red fibres, COX-negative fibres, and abnormal respiratory chain activities was reported in fewer than the half of those who had been investigated (Table S1). Electroencephalogram recordings showed that approximately half (54%, n = 58/107) of patients with epilepsy had epileptiform activities over the occipital lobes. Abnormal nerve conduction was observed in 70%  (n = 43/61) of the individuals, the majority of those (81%, n = 35/43) had axonal neuropathy. None had a pure demyelinating neuropathy (Table S1).

| Neuroimaging findings
General cerebral atrophy (59%, n = 35/59) and cortical focal lesions (54%, n = 59/108) manifesting as T2/FLAIR hyperintensities affecting cortical and subcortical areas were the most frequently reported magnetic resonance imaging (MRI) abnormalities in the study group as a whole. These imaging findings were more prevalent in patients with epilepsy compared to those without epilepsy. A detailed description of MRI findings is provided in Table S2.
Further analysis showed that median survival time from disease onset to death was 19 months (range 0.5-600 months, interquartile range [IQR] 111) for those with disease onset prior to the age 12 years, 151 months (range 4-487, IQR 255) for those with disease onset between 12 and 40 years, and 191 months (range 17-336, The presence of epilepsy was associated with significantly worse survival (P < .001), and the median survival time from seizure onset to death was 37 months (range <1-487). Survival analysis also showed that patients with pathogenic compound heterozygous POLG variants had significantly (P < .001) worse survival compared to those with pathogenic homozygous variants, regardless of specific variant types (Figure 2). Further analysis showed that survival after the onset of seizures in those with early onset disease was significantly worse than those who developed seizures as part of juvenile/adult onset disease. Further, patients who developed liver involvement showed a significantly worse survival than those without liver impairment ( Figure S4).

| DISCUSSION
We present the detailed description of 155 patients with confirmed pathogenic POLG variants focusing on the clinical features, but including laboratory, genetic, and neuroimaging findings. As far as we can ascertain, this is the largest cohort of patients with POLG disease so far described. In addition to the descriptive element, we have also analysed factors, which may predict the prognosis.
We defined the age of onset of each individual symptom and our data confirms that POLG disease comprises a continuum of clinical features rather than a set of separate clinical identities (Figure 1). Apart from PEO/ptosis, all other symptoms could start from infancy to adulthood. While hypotonia and feeding difficulties in infants are likely due to different pathological processes than these features appearing in adults, seizures, peripheral neuropathy, ataxia, muscle weakness, and hepatic disturbance have a similar basis and all could present at any age. PEO/ptosis starts later and appears mainly in patients with dominantly inherited disease or in those with juvenile/early adult onset disease who do not develop epilepsy or, less often, survive despite it. Strokelike episodes appear to start slightly later than most other features. This may reflect the nature of the process 26,27 namely that these represent prolonged seizure activity or status epilepticus.
If we look at the median ages of onset, instead of looking at the age range, we do see a tendency for the features to cluster according to age. We, therefore, reanalysed the data using different age groups. Based on these findings, we found that the clinical spectrum of POLG disease was best described by grouping patients into three categories of early, juvenile/adult, and late onset. Early onset disease was classified as beginning prior to the age of 12 years. In these patients, liver involvement, feeding difficulties, seizures, hypotonia, and muscle weakness were the most dominant/important clinical features and this group had the worst prognosis. The juvenile/adult onset form (12-40 years of age) was characterised by peripheral neuropathy, ataxia, seizures, stroke-like episodes and, in patients with longer survival, PEO. This group carried a better prognosis than the early onset group. Late onset disease (after the age of 40 years) was characterised by ptosis and PEO, with additional features such as peripheral neuropathy ataxia and muscle weakness occurring frequently. This group had the best prognosis. Thus, while the clinical features associated with POLG variants can present at any age, age of disease onset provides both clues to the diagnosis and information about the outcome ( Table 3).
The most frequently reported neurological features, included seizures, ataxia, and peripheral neuropathy. Focal evolving to bilateral tonic-clonic seizures were the most common seizure types, with epileptiform activities predominantly seen in occipital regions. These findings are consistent with previous reports ( 8,10,(13)(14)(15)28,29 ), however, our results also showed that seizures were the most predominant clinical feature in patients with early onset disease (<12 years), common in those with juvenile/adult onset (12-40 years), but infrequent in those with late onset disease (>40 years). Ataxia, peripheral neuropathy, and migraine-like headache were most predominant in individuals with juvenile/adult onset disease, although reported in both early and late onset disease. Ptosis and PEO were common in late onset disease as reported previously, 22 however our data showed that the onset of ptosis and PEO occurred in all age groups. The onset of gastrointestinal features such as feeding difficulties and liver involvement occurred at any age, but was predominantly seen in patients with early onset disease.
Demographic data showed that more than half of the individuals included in this study had onset during childhood (prior to the age of 12), and the incidence of the disease decreased with age. Contrary to previous publications, 9,15 which demonstrate some male predominance, we observed no gender difference.
Survival analysis demonstrated a clear correlation between the age of disease onset and the survival time; earlier onset was associated with worse prognosis (Table 3). Further analysis showed that the presence of epilepsy was significantly associated with worse prognosis regardless of the age of disease onset, and individuals harbouring compound heterozygous POLG variants had worse prognosis compared to those with homozygous variants.
Our study showed that laboratory investigations which are commonly used in the initial diagnostic work-up of mitochondrial disorders, for example, raised blood and CSF lactate, the presence of ragged-red/COX-negative fibres and abnormal respiratory chain activities in the skeletal muscle have low diagnostic sensitivity, being present in fewer than 50% of the individuals. As we showed in a previous publication, 24 elevated CSF protein was the most sensitive (68%, n = 44/65) laboratory diagnostic biomarker at disease onset.
The majority of the patients included in this study were of Northern European descent; thus, a possible limitation of this study is that it might not be possible to extrapolate our findings to other ethnic groups. However, we provide detailed description of all the known phenotypes associated with POLG disease related to more than 40 different POLG variants. Moreover, there is little evidence in the literature of marked ethnic differences in phenotypic expression of POLG disease. Based on the diverse genotypic background of our population, we consider that the findings of our study are relevant to patients with POLG disease, regardless of the ethnic background.
A simple and robust clinical classification is the cornerstone of early diagnosis. Such a classification, together with diagnostic investigations, should facilitate easy recognition of the disease and be useful for both experts and physicians with limited experience of the field. Current nomenclature describing the phenotypic spectrum of POLG disease (Table 1) is complicated and includes overlapping clinical syndromes. This makes implementation in everyday clinical practice difficult. A clear and accurate classification that describes the full spectrum of disease taking account of age-related features is essential not only for optimal management, but also for research F I G U R E 3 Diagnosing POLG disease; clinical suspicion and relevant investigations according to the age of onset. While we have shown that POLG clinical features form a continuum, but it is also clear that age plays a role in which features predominate. Based on our age groups, we can see clear clinical patterns and these will dictate which investigations are appropriate and useful. For example, in the older age category, PEO and ataxia dominate the clinical spectrum and in these cases one can choose either to screen the known genes or to take a muscle biopsy which give both structural clues (COX negative fibres) and the possibility to examine mtDNA (for multiple deletions). We also see that the typical occipital epilepsy occurs in the younger two categories and it is in these that MRI imaging also provides important clues. Peripheral neuropathy occurs in all age groups. In earlier studies, we showed that elevated CSF protein can be helpful, for example in a child with epilepsy and focal MRI changes it can be an important indicator of poor prognosis. a: Direct POLG gene sequence analysis is recommended to confirm the diagnosis in a case of strong clinical suspicion. b: Absence of these findings does not exclude the diagnosis of POLG disease. c: targeted variant analysis for the most common variants (p. Ala467Thr and p.Trp748Ser) can be performed first in juvenile and late onset disease, whole POLG gene sequence analysis is recommended for all early onset disease and those with strong clinical suspicion of POLG disease regardless of the age of onset. CSF, cerebrospinal fluid; RRF, ragged-red fibres; PEO, progressive external ophthalmoplegia; P. neuropathy, peripheral neuropathy; SLE, stroke-like episodes and, when treatments become available, for use in clinical trials.
We provide a robust and simplified clinical classification based on data from the largest cohort of patients with POLG disease published to date. This classification highlights three distinct age groups and within these groups the major clinical features. Earlier classifications of POLG disease have focused primarily on phenotypic elements; for example, the presence of ataxia with or without myoclonus or epilepsy has variously been referred to as SANDO, ANS, or MIRAS/MSCAE. Early onset diseases have been separated into Alpers or MCHS or Leigh-like syndromes. The presence of mtDNA depletion has also been used to define POLG related disease although the presence of this is known to be tissue dependent and depletion in brain and liver is found in both young and older patients. We feel that these phenotypic labels create an unnecessarily complicated classification. Age alone appears robust enough to delineate the important features of POLG disease such that we would recommend simplifying classification to early onset, juvenile onset and late onset POLG disease. The algorithm ( Figure 3) shows how recognition of these key clinical features could be used to direct clinical investigation in the different age groups.

SUPPORTING INFORMATION
Additional supporting information may be found online in the Supporting Information section at the end of this article.