Mitochondrial DNA heteroplasmy distinguishes disease manifestation in PINK1- and PRKN-linked Parkinson’s disease

Introduction

Parkinson’s disease (PD) is a neurodegenerative disorder affecting over seven million patients world-wide[9], of which ∼10% can be genetically explained [3, 9]. Multiple genes have been found to be implicated in familial and sporadic PD, including pathogenic mutations and strong risk factors. Biallelic PINK1 or PRKN mutations are a well-known cause of early-onset PD [15, 37]. Interestingly, monoallelic PINK1/PRKN mutations have been considered a PD risk factor [20]. Based on a conservative estimate of a frequency of 2% of monoallelic PINK1 and PRKN mutations among all PD patients, one can expect ∼140,000 carriers worldwide, this warrants a deeper understanding of the possible functional consequence of PINK1 or PRKN monoallelic mutations. Affected monoallelic PINK1 mutation carriers have a ∼10 year later age at onset (AAO of 43.2 years) than carriers of biallelic mutations (AAO 32.6 years), but were on average still younger than idiopathic PD (iPD) patients (AAO 70 years) used for comparison in the meta-analysis [19]. Furthermore, neuroimaging studies have demonstrated PD-like changes in humans with monoallelic PINK1/PRKN mutations [16, 46]. However, the possible role of monoallelic pathogenic PINK1 and PRKN mutations in PD remains a matter of vivid debate: a recent study found 0.089% of monoallelic PINK1 mutation carriers in cases compared to 0.071% in controls in a large population screen (n=376,558)[22]. Another study did not find any associations between PRKN variants and risk of PD. Pathogenic and likely-pathogenic heterozygous single nucleotide variants and copy-number variations were less frequent in PD patients (1.52%) compared to controls (1.8%) [50]. These studies suggest that PINK1/PRKN monoallelic mutations may not contribute to risk, albeit with limited power and no possibility to assess (subtle) signs of parkinsonism in the seemingly healthy controls or without information on the much more common (monoallelic) PRKN mutations. Still, when deep clinical phenotyping was applied, patients with PD harboring PINK1 or PRKN monoallelic pathogenic variants were at a significantly higher rate than healthy controls [20].

Biologically, PINK1 and Parkin are important in regulating mitochondrial quality control. Activation of the PINK1/Parkin mitophagy pathway has been shown to result in the selective removal of dysfunctional mitochondria and their damaged DNA (mtDNA) molecules [12, 17, 42]. In a recent study, Parkin-deficient mice were crossed with “mutator mice”, which harbor a proof-reading defective mitochondrial polymerase gamma. In the resulting animals, mtDNA disintegration and impaired mitophagy led to the release of mtDNA molecules into the extracellular space, where they act as damage-associated molecular patterns (DAMPs) triggering inflammatory pathways [41]. However, whether accumulation of mtDNA damage has an influence on the disease manifestation of PINK1 and PRKN-associated PD still needs to be elucidated.

Herein, we present an in-depth characterization of the mitochondrial genome in blood cells as a potential disease marker for genetic PD. We use four cohorts, including two large and homogeneous founder populations (South Tyrolean and Tunisian Arab Berber) carrying PINK1 and PRKN mutations, neuronal models and postmortem brain tissue to assess mitochondrial dynamics as a marker of disease manifestation.

Materials and Methods (max 3000)

Study individuals and genetic analysis

Deep mitochondrial sequencing and real-time PCR assays using TaqMan technology were performed on a total of 411 individuals (Table 1a and S1). Patients were recruited from Germany, Italy, Tunisia and Serbia fulfilling the UK Brain Bank Criteria for PD. Demographic and basic clinical data from PINK1 and PRKN mutation carriers (biallelic and monoallelic, n=254) are shown in Table 1b. The AAO of disease for individuals with PINK1 mutations, PRKN mutations, and iPD is reported in Table 1a, 1b and S1. MtDNA was extracted from whole blood using standard protocols for extraction of genomic DNA. PINK1 and PRKN genetic screening was performed with Sanger sequencing and MLPA (multiplex ligation probe-dependent amplification) as previously described[43]. All individuals with PD were negatively screened for other pathogenic variants in known PD genes (SNCA, LRRK2, VPS35, and DJ1) with gene panel sequencing using either the Illumina NextSeq500 or HiSeq4000 sequencers and MLPA analysis. PINK1 and PRKN mutations refer to ‘possibly, probably or definitely’ pathogenic variants and variants of uncertain significance (VUS, e.g. PINK1 p.G411S, p.N451T) were not included based on the Movement Disorder Society Genetic mutation database (MDSGene) [18]. PINK1 and PRKN variants were labelled as “pathogenic mutations” based on the criteria of ‘possibly, probably or definitely’ pathogenic with clear guidelines from the MDSGene consortium[18, 27]. All individuals provided informed consent prior to their participation. Specific approvals were obtained from the local ethics committees at the Research Ethics Boards of the Universities of Lübeck and Luxembourg and the respective local ethics committees from the cohort centers. Clinical assessment of all patients was performed by movement disorders specialists. Population-matched healthy controls were used in the study. The AAO was self-reported by patients and based on the occurrence of the first motor symptoms. Familial relationships were defined based on self-report. An overview of individuals in the study is depicted in Figure 1.

View this table:

• View inline
• View popup
• Download powerpoint

Table 1a.

Demographics of all cohorts: individuals with PINK1/PRKN mutations, idiopathic Parkinson’s disease and controls.

View this table:

• View inline
• View popup
• Download powerpoint

Table 1b.

Demographics of cohorts stratified by genotype

• Download figure
• Open in new tab

Figure 1. Overview of study.

Flow chart showing individuals incorporated in this study, biospecimen, rationale and techniques. mtDNA: mitochondrial DNA, NGS: next-generation sequencing, NM: neuromelanin, TGS: third-generation sequencing.

Deep mitochondrial sequencing

Illumina NextSeq short-read sequencing

Deep mitochondrial sequencing was performed with bait enrichment with IDT lockdown probes specific for the entire mitochondrial genome to generate libraries of 6-plex or long-range PCR. The two primer sets used for the long-range PCR are described in Table S2. Subsequent next-generation sequencing was performed on Illumina, using the NextSeq500 (Illumina, Inc.) to produce 2 × 150bp reads. Raw sequencing reads were converted to standard FASTQ format using bcl2fastq software 2.17.1.14 (Illumina, Inc.). Raw FASTQ files were analyzed with FastQC and aligned with BWA MEM[25] to the mitochondrial reference genome (rCRS) [1]. The resulting SAM files were processed with SAMtools to sorted BAM files [26]. Subsequently, the BAM files were processed with GATK [31] and duplicates removed with MarkDuplicates. Additional quality control was performed with QualiMap 2 [33], and subsequently MultiQC [10] was applied on the resulting reports, including the FastQC reports. The mtDNA-Server [48] successor Mutserve 2 [47] was used for variant calling on the mitochondrial genome to detect homoplasmic and heteroplasmic variants in the sequence data. After additional quality control with Haplocheck [47], we set the minimum heteroplasmy frequency (HF) level to 0.05 [48] and kept all other parameters with the default parameters (per base quality 20, mapping quality 30, alignment score 30), as some samples showed contamination either from different samples or from Nuclear Mitochondrial DNA segments (NUMTs) below that threshold. Mitochondrial genomes with a mean depth of 5,538X (SD:4,919X) coverage were obtained. A cut-off of >1,000X coverage was applied and samples that were contaminated or had an unusually high number of heteroplasmic variants that may be related to NUMTs (>40) were removed (n=86). We defined variants as homoplasmic (>95%) or heteroplasmic (<95% and >5%) based on allele frequency.

Oxford Nanopore long-read single-molecule sequencing

To obtain more even coverage across the mtDNA genome and eliminate NUMTs, a long-read sequencing approach was additionally used to detect mtDNA heteroplasmy in blood-derived (n=16) and neuronal-derived DNA (n=10) of PRKN mutation carriers and controls. The same two primer sets from Illumina sequencing were used for the long-range PCR. Samples were barcoded and multiplexed, and subsequent sequencing was performed on an R9 flow cell. Mitochondrial genomes with a mean depth of 3433.2X (SD:±2299.9X) coverage were obtained. Base-calling was performed with Guppy v5.0.11 with the super-accurate model. The base-called reads were filtered with Filtlong (v0.2.0) to only include the best 50% of the reads, based on Phred quality scores (q-score), with a minimum read length of 9kb. Then, the reads were trimmed with NanoFilt (v2.5.0). Subsequently, the Nanopore reads were aligned against the mitochondrial genome reference sequence using Minimap2 (v2.17) [24]. The alignments were sorted and indexed with Samtools (v1.3.1). Mutserve2 was used to call variants and the same default parameters were applied in the same fashion as with our short-read sequencing. However, different thresholds for HF levels were set at 0.05, 0.02 and down to 0.01. The per base quality scores allowed for more accurate identification of heteroplasmic variants. All variants from third and second generation sequencing were annotated with a predefined file provided by Mutserve2, with allele frequencies from the 1000 Genomes Project [11], HelixMtDB [4], NUMTs defining variants [7] as well as pathogenicity scores [6, 35].

MtDNA deletion and 7S DNA analysis

Real-time PCR assays were performed using TaqMan technology. MtDNA deletion rates were derived from the ratio of the mtDNA-encoded genes NADH:ubiquinone oxidoreductase core subunit 4 and NADH-dehydrogenase 1 (ND4:ND1). ND1 is a gene located in the minor arc of the mitochondrial genome, in an area spared from deletions. ND4 is located in the major arc, in a region commonly affected by somatic deletions. MtDNA transcription initiation was assessed by calculating the 7S DNA:ND1 ratio, as the relative abundance of 7S DNA in the D-Loop region allows for an estimation of the mitochondrial transcription initiation rate[32]. Forward and reverse primers as well as probes coupled with a non-fluorescent quencher described in Table S2 were used to quantify the concentrations of ND1, ND4, and D-loop, where 7S DNA is integrated during the initial phase of mtDNA transcription. The reactions were performed in 384-well plates in a final reaction volume of 5 µL using a LightCycler480(Roche) as previously described[14]. To minimize inter-experimental variability, reagents for each real-time PCR analysis were pipetted using an automated liquid handler (BioMek FX). A serial dilution of the plasmid p7D1 encoding the target sequence of ND1, ND4 and the D-Loop region was used to quantify the samples. All samples were measured in triplicates. A mixture of all samples was used to normalize across plates. The average and standard deviation of the normalized samples were used for analysis.

Generation of iPSC-derived neurons from PRKN-mutant PD patients and controls

iPSCs were generated from fibroblasts of four PD patients (mean age±SD: 57.75±12.03 years; three females and one male) with biallelic PRKN mutations (homozygous c.1072Tdel; compound-heterozygous c.1072Tdel+delEx7; compound-heterozygous delEx4+c.924C>T; compound-heterozygous delEx1+c.924C>T) and three age-matched controls (mean age±SD: 45.33±9.74 years, one female and two male) as previously described[44] and cultured in mTeSR1™ complete medium (StemCell Technologies). First, iPSCs were differentiated into small molecule neural precursor cells (smNPCs) using an established protocol[38]. In short, smNPCs were cultured in N2B27 medium: Neurobasal (Gibco)/DMEM-F12 (Gibco) 50:50 and supplemented with 1% B27 without vitamin A (ThermoFisher), 0.5% N2 (Life Technologies), 1% penicillin-streptomycin (ThermoFisher) and 1% 200 mM glutamine (Westburg). During smNPC cultivation, N2/B27 medium was additionally supplemented with 3 µM CHIR 99021 (Sigma), 150 µM ascorbic acid (Sigma) and 0.5 µM purmorphamine (PMA) (Sigma). The medium was changed every other day.

The induction of midbrain neurons began after smNPCs reached the 15^th passage. Once ∼75% confluent, cells were detached by subjection to Accutase (Merck Millipore) for 5 min at 37°C and then centrifuged for 3 min at 300 x g at room temperature. Cells were then resuspended and counted via the Countess II FL Automated Cell Counter (ThermoFisher). 750,000 smNPCs per well were seeded onto Matrigel-coated 6-well plates in N2/B27 medium with 1 µM PMA, 200 µM ascorbic acid and 100 ng FGF8 (PeproTech) for 8 days. Next, cells were cultured in N2/B27 medium with 0.5 µM PMA and 200 µM ascorbic acid for two days. For the next 22 days, cells were cultured in N2/B27 medium with 200 µM ascorbic acid, 500 µM dibutyryl-cAMP (Applichem), 1 ng/mL TGF-β3 (Peprotech), 10 ng/mL GDNF (Peprotech) and 20 ng/mL BDNF (PeproTech). The medium was changed every second day. Finally, DNA was prepared from previously characterized iPSC-derived neurons (n=5 million) [44] with the QIAAmp Minikit (Qiagen) and was quantified using standard methods.

Laser-microdissected midbrain tissue from PRKN-mutant PD patients and controls

Frozen human postmortem midbrain tissue sections and associated clinical and neuropathological data were supplied by Juntendo University and the Parkinson’s UK brain bank at Imperial College London. One PD patient (age at death: in their 70s, male, compound-heterozygous PRKN delEx2 + delEx3) and one control (age at death: in their 70s, male) was used for this analysis. Frozen midbrain blocks were cryosectioned at ∼15 μm thickness in the transverse plane. For the isolation of neuromelanin-positive neurons from the substantia nigra, sections were gradually thawed and fixed with 4% paraformaldehyde for 10 min. After three brief washes in TBST buffer, sections were left to dry on room temperature. Laser capture microdissection (LCM) was performed with the PALM MicroBeam (Zeiss) and isolated neurons were lysed in a buffer (50 mM Tris-HCl, pH 8.5, 1 mM EDTA, 0.5% Tween-20, 200ng/mL proteinase K) for 3hrs at 55°C and 10min at 90°C. In addition, an area without visible neuromelanin deposits surrounding the substantia nigra was cut-out and lysed in the same manner.

RNA Ampliseq Transcriptome

RNA was extracted from blood of PRKN biallelic and monoallelic mutation carriers (PRKN/all) (n=23, mean age±SD: 56.2±13.2, mean AAO±SD: 38.8±18.2) and controls (n=6, mean age±SD: 71.7±9.1) with RNeasy Qiagen Minikit and DNase I digested prior to assessing the concentration, quality and RNA integrity with Agilent 2100 bioanalyzer, RNA LabChip kit and associated software (Agilent). Ampliseq™ whole transcriptome analysis was performed with the Ion Proton (Life Technologies, Inc) with average of 9.96M±1.14M reads. Transcriptome reads were aligned to the GRCh37/hg19 reference genome and analyzed by RNAseq Analysis plugin (V5.0.0.2) using default analysis settings. DESeq2 R package within Bioconductor was used to test for differential expression by use of negative binomial generalized linear models; the estimates of dispersion and logarithmic fold changes incorporate data-driven prior distributions. The non-normalized counts of sequencing reads/fragments were used in DESeq2’s statistical model that accounts for library size differences internally. Differentially expressed genes (DEGs) with nominally-significant p-values (<0.05) and Log2fold (L2F) changes (>|0.2|) were mapped on KEGG pathways with pathview [28].

IL6 measurements in serum

To investigate interleukin 6 (IL6) levels in PINK1/PRKN biallelic and monoallelic mutation carriers (n=20, mean age±SD: 52.9 years ±14.3 years, 9 females), venous blood was collected in serum tubes and processed within one hour. Serum samples were centrifuged at 4°C for 10 min at 3000g, immediately frozen at -80°C and transferred to a certified diagnostic laboratory, where IL6 levels were measured using the Elecsys IL6 assay (Cobax).

Statistical Analyses

The affection status (whether the individual was diagnosed with PD), PINK1/PRKN mutational dosage or heteroplasmic variant load were used as outcome variables in this study. JMP software, Version 10 (SAS Institute Inc, Cary, NC USA) and R, Version 4.1 were used to compare the clinical observations across different groups (i.e. PD vs asymptomatic; AAO; genotypes) and number of heteroplasmic mtDNA variants. For analyses in the blood-derived samples from the cohorts, Kruskal-Wallis tests (followed by post-hoc pairwise comparisons if p<0.05) and Mann-Whitney U-tests were reported. For distinguishing affection status with heteroplasmic mtDNA variants in monoallelic PINK1/PRKN mutation carriers, multivariable logistic regression models were used to classify predict status by mtDNA variants and age at examination. Then, the corresponding receiving operating characteristic (ROC) curves were reported. We performed one statistical test in the narrow sense to test the specific hypothesis regarding the PINK1/PRKN affection status with mtDNA heteroplasmic variant load. All other tests were performed exploratorily only, so that reported p-values are not corrected for multiple testing but to be interpreted descriptively or exploratorily. DESeq2 R package within Bioconductor was used to describe differences in gene expression using negative binomial distribution based on estimate variance-mean dependence in count data from high-throughput sequencing assays. The p-values were attained by the Wald test for the gene expression differences. Pearson correlation coefficient was reported for the correlation of IL6 and heteroplasmic variants. For analyses in iPSC-derived neurons, t-tests were performed.

Graphical output

All graphics and figures were created with GraphPad Prism and BioRender.com.

Results

Heteroplasmic mtDNA variants are associated with disease manifestation in PINK1/PRKN mutation carriers

We explored whether an association between mtDNA variant burden and PD status exists by comparing the cumulative mtDNA variant load between symptomatic (PD+) and asymptomatic carriers (PD-) of PINK1 and PRKN monoallelic mutation carriers.

This analysis revealed a higher number of heteroplasmic variants in symptomatic (PINK1+/PRKN+/PD+, n=29), compared to asymptomatic monoallelic carriers (PINK1+/PRKN+/PD-, n=109) (Mann-Whitney U-test, Z=2.93, p=0.0033, n=138). The resulting area under the curve indicated good discrimination based on low-level mtDNA variants and age at blood drawn for comparison between symptomatic (PD+) and asymptomatic carriers (PD-) of PINK1 and PRKN monoallelic mutation carriers (AUC=0.83, CI:0.74-0.93) (Figure 2a and b).

• Download figure
• Open in new tab

Figure 2. Disease manifestation of PRKN and PINK1 mutations are associated with mtDNA heteroplasmy in blood-derived DNA.

(a) The sensitivity and specificity of the heteroplasmic mtDNA variants (HF>5%) as a biological marker plus age at examination were assessed with ROC analysis. The area under the curve (AUC) indicated good discrimination for affected (n=29) vs. unaffected (n=109) monoallelic PRKN/PINK1 mutation carriers (AUC=0.83, CI:0.74-0.93). (b) Scatter plot showing the number of heteroplasmic mtDNA variants for affected (n=29) vs. unaffected (n=109) monoallelic PRKN/PINK1 mutation carriers. Mann-Whitney U-test was performed. Bars indicate means and 95%CI. PINK1+:PINK1 monoallelic mutation; PRKN+=PRKN monoallelic mutation; PD+: patient with Parkinson’s disease; PD-: individuals without Parkinson’s disease.

By contrast, a comparison of homoplasmic mtDNA variants did not show differences between symptomatic (PINK1+/PRKN+/PD+, n=29) and asymptomatic PINK1 or PRKN monoallelic mutation carriers (PINK1+/PRKN+/PD-, n=109) (Mann-Whitney U-test, Z=0.43, p=0.66, data not shown).

Heteroplasmic mtDNA variants increase with number of mutant PINK1/PRKN alleles

To assess whether the number of PINK1 or PRKN mutations has an impact on the mtDNA variant burden, we compared the heteroplasmic variant load between affected and unaffected individuals with biallelic or monoallelic mutations, idiopathic PD (iPD) patients and controls.

The number of heteroplasmic mtDNA variant load differed between patients with iPD (n=54) and patients with PINK1/PRKN biallelic or monoallelic mutations (PINK1/PRKN/PD+, n=78) (Mann-Whitney U-test, p=0.0441) (Figure 3a). Separating patients with monoallelic (PINK1+/PRKN+/PD+, n=29) and biallelic PINK1/PRKN mutations (PINK1++/PRKN++/PD+, n=49), no overall difference was seen to iPD patients (Kruskal-Wallis test, p=0.104) (Figure 3b).

• Download figure
• Open in new tab

Figure 3. PRKN/PINK1 status is associated with number of heteroplasmic mtDNA variants in blood-derived DNA.

Scatter plot showing number heteroplasmic mtDNA variants (HF>5%) for (a) idiopathic PD (iPD, n=54) vs PINK1/PRKN/PD+ (n=78); (b) iPD (n=54) vs PINK1+/PRKN+/PD+ (n=29) vs PINK1++/PRKN++/PD+ (n=49); (c) controls (n=67) vs iPD (n=54) vs all monoallelic mutation carriers (n=150) vs all biallelic mutation carriers (n=52). Mann-Whitney U-test was performed for (a) and Kruskal-Wallis tests for (b) and (c) shown in bold. Post-hoc analyses are not bolded. Bars indicate means and 95%CI. PINK1: monoallelic and biallelic PINK1 mutations; PRKN: monoallelic and biallelic PRKN mutations; PINK1+:PINK1 monoallelic, PINK1++: PINK1 biallelic mutations; PRKN+=PRKN monoallelic, PRKN++:PRKN biallelic mutations; PD+:patient with Parkinson’s disease; all=individuals in study with or without Parkinson’s disease.

Heteroplasmic mtDNA variant load differed between biallelic and monoallelic PINK1/ PRKN mutation carriers regardless of affection status (PINK1/PRKN/all) and controls (n=67), higher mtDNA variant load was observed for biallelic compared to monoallelic mutation carriers (Kruskal-Wallis test, p=0.0006) (Figure 3c). Post-hoc pairwise analyses revealed higher heteroplasmic mtDNA variant load for biallelic compared to monoallelic PINK1/ PRKN mutation carriers (p=0.002) and higher variant load for biallelic PINK1/PRKN/ mutation carriers compare to controls (p=0.001).

The number of homoplasmic variants also increased with the number of mutant PINK1/PRKN alleles: mtDNA variants burden differed between patients with iPD (n=54) and PINK1 or PRKN carriers regardless of affection status (n=78) (Figure S2, Kruskal-Wallis test, p<0.001). As the homoplasmic mtDNA variants burden should not be influenced by age, we did not perform the age adjustment. These results suggest that heteroplasmic and homoplasmic mtDNA variant load are associated with PINK1/PRKN genotype.

MtDNA variants were also compared between iPD patients and controls in the same manner. The median values for the number of heteroplasmic variants and homoplasmic variants was not found to differ between iPD patients and controls (Mann-Whitney U-test p=0.0846 and p=0.4668, respectively).

Quantitative analyses of mtDNA-associated 7S DNA and major arc deletions

To extend our mtDNA variant burden analysis to include large-scale deletions, we investigated ND4:ND1 ratios. After filtering for quality control and removing outliers, n=302 individuals were included for further analyses.

Patients with biallelic PINK1 or PRKN mutations (PINK1++/PRKN++/PD+; n=57) had more mtDNA deletions than patients with monoallelic PINK1 or PRKN mutations (PINK1+/PRKN+/PD+; n=25) or iPD cases (n=50) (Kruskal-Wallis test, p=0.009) (Figure 4a). Disregarding the disease status, mtDNA major arc deletions were more abundant in biallelic mutation carriers (n=57) than in monoallelic mutation carriers (n=130) or controls (n=65) (Kruskal-Wallis test, p=0.003) (Figure 4b).

• Download figure
• Open in new tab

Figure 4. Analysis of mtDNA major arc deletions and mtDNA transcription-associated 7S DNA by real-time PCR in blood-derived DNA.

Simultaneous real-time PCR quantification of the mtDNA genes ND4 relative to ND1 or detection of 7S DNA relative to ND1. Scatter plots showing ND4:ND1 ratios for (a) patients with idiopathic PD (iPD, n=50) vs patients with PD due to monoallelic (n=25) or biallelic (n=57) mutations in PINK1 or PRKN. (b) Controls (n=65) vs iPD (n=50) vs all PINK1 or PRKN monoallelic mutation carriers (n=130) vs all PINK1 or PRKN biallelic mutation carriers (n=57). Scatter plot showing 7S DNA:ND1 ratios for (c) patients with iPD (n=49) vs patients with PD due to monoallelic (n=25) or biallelic (n=57) mutations in PINK1 or PRKN. (d) Controls (n=65) vs iPD patients (n=49) vs all monoallelic PINK1/PRKN mutation carriers (n=131) vs all PINK1 or PRKN biallelic mutation carriers (n=57). Kruskal-Wallis tests were performed and in bold. Post-hoc analyses are not bolded. Bars indicate means and 95%CI; PINK1+:PINK1 monoallelic, PINK1++: PINK1 biallelic mutations; PRKN+: PRKN monoallelic, PRKN++: PRKN biallelic mutations; PD+: patient with Parkinson’s disease.

MtDNA maintenance impairments may also extend to changes in transcription and replication of the mitochondrial genome. Events of mtDNA transcription/replication initiation were measured by determining 7S DNA:ND1 ratios (Figure 4c and 4d). Biallelic mutation carriers (PINK1++/PRKN++/all; n=57) showed lower 7S DNA:ND1 ratios than controls (n=65); a smaller reduction in 7S DNA:ND1 ratios was seen for monoallelic mutation carriers (PINK1+/PRKN+/all; n=131) compared to controls, and the 7S DNA:ND1 ratios were different between biallelic (PINK1++/PRKN++/all) and monoallelic (PINK1+/PRKN+/all) PINK1 or PRKN mutation carriers; lastly, compared to controls, iPD patients (n=49) also showed a similar reduction in 7S DNA:ND1 ratios (Kruskal-Wallis test p<0.001) like biallelic PINK1 or PRKN mutation carriers (PINK1++/PRKN++/all) (Figure 4d).

MtDNA variants in iPSC-derived dopaminergic neurons and midbrain tissue with Nanopore sequencing

Next, we tested whether alterations in mtDNA variant load in PINK1/PRKN-linked PD are a phenomenon restricted to peripheral tissues or whether they can also be observed in patient-derived neurons and midbrain tissue. To assess whether heteroplasmic mtDNA variants are present in neurons, we generated iPSC-derived midbrain neurons from four PD patients with biallelic PRKN mutations and three age-matched controls. MtDNA Illumina short-read deep-sequencing of these neuronal cultures revealed an increase in heteroplasmic variants in patient cells (p=0.0389) (Figure 5a). These findings suggests that somatic mtDNA mutation accumulation not only occurs in peripheral tissues but also in iPSC-derived midbrain neurons from PD patients with biallelic PRKN mutations.

• Download figure
• Open in new tab

Figure 5. Heteroplasmic mtDNA variants are associated with PRKN mutation carriers in iPSC-derived and postmortem neurons and inflammation.

Scatter plot showing number of heteroplasmic mtDNA variants in iPSC-derived midbrain neurons from 3-4 differentiations of controls (n=4) and biallelic PRKN-PD patients (n=4) performed with (a) Illumina short-read sequencing at >5%, (b) Nanopore long-read sequencing at >5%, (c) Nanopore long-read sequencing at >2%, and (d) Nanopore long-read sequencing at >1%. T-tests were performed. PRKN++: PRKN biallelic mutations; PD+: patients with Parkinson’s disease. (e) Line plot showing number heteroplasmic mtDNA variants for control-vs PRKN++/PD+-derived nigral laser-microdissected neuromelonin (NM)-positive neurons and non-nigral midbrain tissue at >5%, >2%, and >1%. (f) Correlation of serum IL6 and total mtDNA variant load in PINK1 or PRKN monoallelic (n=14) and biallelic (n=6) mutation carriers. Bars indicate means and 95% CI.

Due to limitations in short-read sequencing, we utilized Nanopore long-read sequencing for a more even coverage within amplicons, better mapping quality and single-molecule technology. We first tested the technology in blood-derived PRKN monoallelic mutation carriers (n=16 select samples). Nanopore sequencing at >1,000X coverage confirmed 28 out of 32 (87.5%) variants (HF>0.05) in the Illumina short-read dataset. Besides data from our own study, we followed the benchmarking protocol established for Nanopore long-read sequencing in low-frequency variants [29]. When applying 5%, 2% and 1% heteroplasmy thresholds on long-read sequencing data in the neurons, we found that the number of heteroplasmic variants was higher in PD patients with biallelic PRKN mutations compared to controls (HF>5%, p=0.0633, HF>2%, p=0.0024, HF>1%, p=0.0466) (Figure 5b-d).

Finally, in light of recently reported limitations of mtDNA studies in iPSC-derived models [45], we compared mtDNA variants in laser-microdissected neuromelanin-positive neurons isolated from postmortem substantia nigra and non-nigral midbrain tissue of a PD patient with biallelic PRKN mutations and one control. As we investigated lower HF levels, the difference between the nigral neurons (HF>5%, n=4; HF>2%, n=21; HF>1%, n=33) and the non-nigral midbrain tissue (HF>5%, n=5; HF>2%, n=7; HF>1%, n=13) increases in the PRKN mutation carrier. This phenomenon was not as pronounced in nigral neurons (HF>5%, n=7; HF>2%, n=10; HF>1%, n=18) compared to non-nigral midbrain tissue (HF>5%, n=5; HF>2%, n=6; HF>1%, n=14) in the control (Figure 5e).

Discussion

In this study, we investigated the mtDNA variant burden as a marker of disease state in patients with PINK1 and PRKN mutations. Deep sequencing of blood-derived mtDNA revealed an accumulation of heteroplasmic variants, which are likely to be of somatic origin[34] that could influence the clinical manifestation of PINK1 or PRKN pathogenic mutations. Although the effect size of mtDNA heteroplasmy as a disease marker is relatively low, the ROC curve analyses showed predictive potential after including age (AUC>0.83). Unfortunately, only for a subset of individuals the exact age at blood-draw was available (n=231/325 total individuals). Moreover, the number of affected carriers was too sparse for a continuous AAO trait analysis, and may have been further hampered by the linear association of mtDNA damage and age[21]. However, being mindful of these limitations, the fact that mtDNA heteroplasmy could distinguish between symptomatic and asymptomatic monoallelic carriers of PINK1 or PRKN mutations but not between iPD and controls, still suggests that the effect is specific for individuals with PINK1 or PRKN mutations. As we found that mtDNA heteroplasmy is associated with disease manifestation, we may extrapolate mtDNA heteroplasmic variant load as a penetrance modifier. However, as PINK1 or PRKN monoallelic mutations are still debatable as risk factors of PD, and biallelic mutations are fully penetrant, there are still caveats to this terminology. Interestingly, the analysis of heteroplasmic mtDNA variant load revealed a mutation dosage effect for patients with PINK1 or PRKN mutations. A similar dosage effect was also observed for major arc deletions and 7S DNA. Thus, this finding highlights a mechanistic link between impaired mtDNA maintenance and increased mtDNA variant load. Parkin has been shown to modulate mtDNA transcription by ubiquitination of PARIS, which represses the mitochondrial biogenesis master regulator PGC1-alpha [13]. In addition, our own recent work revealed that a metabolic shift in Parkin-deficient neurons is sensed by SIRT1, which in turn causes a depletion of PGC1-alpha and mtDNA dyshomeostasis [44]. Accordingly, Parkin protects mtDNA from oxidative damage and stimulates mtDNA repair [39].

An increase in mtDNA heteroplasmy likely indicates a development during an individual’s lifetime (somatic variants) that can influence phenoconversion. While somatic mtDNA changes are possibly a direct result of a disturbance in mitochondrial quality control in carriers of PINK1 or PRKN mutations, homoplasmic mtDNA variants are likely to be inherited. The accumulation of homoplasmic variants could be explained by transmission of mtDNA variants across multiple generations over decades in families, a generational conversion of heteroplasmic into homoplasmic variants in populations with PINK1 or PRKN mutations, though this can be influenced by specific mtDNA haplotypes.

To investigate whether the observed mtDNA phenotypes in carriers of PINK1 or PRKN mutations are only a feature detectable in peripheral tissue or whether mtDNA integrity contributes to neurodegeneration in these individuals, we additionally studied neuronal samples. Comparing iPSC-derived midbrain neurons from four PD patients deficient of Parkin and three controls, we confirmed an upregulation of somatic mtDNA mutations in the patient cells. By contrast, recent sequencing analyses in iPSC lines implicated a high mtDNA mutation rate during reprogramming [45]. While this technical limitation alone would not explain the genotype-specific differences in mtDNA mutational load observed here, it motivated us to additionally study postmortem nigral neurons from a biallelic PRKN-PD patient and a matched control. Indeed, in laser-microdissected dopaminergic neurons (which were identified based on their neuromelanin deposits) from the PRKN-mutant patient, we again detected elevated numbers of somatic mutations. Given that previous deep-sequencing analyses in single postmortem dopaminergic neurons from nigral tissue of iPD patients and controls did not reveal differences in terms of mtDNA somatic mutational load [8], this finding further strengthens the role of Parkin in regulating mtDNA homeostasis. In addition, in line with previous results from iPD patient and control postmortem sections [2, 8], our comparison of the mtDNA status in neuromelanin-positive nigral neurons versus non-nigral midbrain tissue revealed that dopaminergic neurons are particularly vulnerable to this kind of damage. This phenomenon may be explained by the autoxidation properties of dopamine [23], which facilitate the formation of free radicals. ROS, in turn, can cause single- and double-strand breaks, create abasic sites and oxidize purines and pyrimidines in the mitochondrial genome [49].

Apart from its role in mtDNA maintenance, Parkin deficiency has been shown to trigger mtDNA disintegration and dyshomeostasis as a result of impaired mitochondrial clearance [13]. Upon depolarization, oxidative stress or protein misfolding, PINK1 recruits Parkin to the mitochondria to initiate mitophagy [13, 36]. If mitochondrial clearance fails in the absence of PINK1 or Parkin, pro-inflammatory signalling is induced. PRKN knockout mice with proof reading-deficient mitochondrial polymerase gamma (POLG) as well as PD patients with PINK1 or PRKN mutations showed elevated circulating cell-free mtDNA levels in serum [5], which escapes from cells as consequence of impaired mitophagy. In the extracellular space, mtDNA can act as a damage-associated molecular pattern that is recognized by cGAS/STING and leads to inflammasome activation. Accordingly, PRKN-knockout mutator mice and PINK1 or PRKN-mutant PD patients showed an upregulation of the cytokine IL6 in serum [5, 40]. To further investigate the relationship between mtDNA integrity and inflammation, we quantified serum IL6 levels in a small subset of our PINK1- and PRKN mutation carriers [5, 40]. Remarkably, IL6 concentrations correlated with mtDNA heteroplasmic variant load in these samples. In addition, we observed DEGs enriched in inflammatory pathways when comparing the gene expression profiles of PRKN mutation carriers with differential mtDNA variant load. These results concur with the above-mentioned findings in mice and human biosamples[5, 41]. However, cellular studies will be needed to conclusively clarify if inflammation is a downstream effect triggered by accumulation of mtDNA variants and subsequent mtDNA release from damaged mitochondria.

The strength of our study lies in the inclusion of two large founder populations: South Tyroleans and North Africa Arab Berbers, where the frequency of biallelic PINK1 and PRKN variants are higher than in other ethnic groups. These populations provide genetic and environmental homogeneity to increase power for discovery and comprise the majority of our study cohort. Furthermore, our analysis compared specific controls from these founder populations, which also avoids genetic heterogeneity and background bias. However, the disease severity of patients and affection status of currently unaffected carriers may change and longitudinal follow up of our study cohort is warranted. We only had access to a small number of the corresponding mothers of investigated individuals to unequivocally exclude inherited mtDNA variants. In our study, heteroplasmic variants from deep sequencing served as a surrogate marker for somatic mtDNA variation[34]. Using this approach, we provided functional evidence for a role of monoallelic mutations in PINK1/PRKN as PD risk factors or pathogenic variants of highly reduced penetrance. In addition, we applied single-molecule sequencing to postmortem midbrain tissue and confirmed an upregulation of somatic mutations in nigral dopaminergic neurons from a PRKN PD case. Finally, our data provides further evidence for the recently discovered link between the PINK1/Parkin mitophagy pathway and inflammation in the context of increased mtDNA heteroplasmy in human samples.

In summary, our findings strengthen the relevance of the PINK1/Parkin pathway in maintaining a healthy mitochondrial pool. In addition, our study highlights mtDNA heteroplasmy as a potential disease manifestation marker for PINK1/PRKN PD. Thus, patient blood-based deep mtDNA sequencing may be considered for genetic counseling and future clinical trials. This is important for monitoring disease onset in asymptomatic carriers, as mtDNA heteroplasmy is reliably measurable. Furthermore, longitudinal prospective studies can aim to prioritize and follow monoallelic mutation carriers for neuroprotective trials before phenoconversion.

Competing interests

Authors report no competing interests relevant to the manuscript.

Author Contributions

AG, CK and JT designed the study and wrote the manuscript. JT, AG, SD, SI, KW and NO performed the genetic experiments. JT and CVG performed transcriptomic analyses. JT, AH, IK, SD, MW, AG, KL, ELG, and MF performed statistical analyses and data interpretation. JT, HW, PM performed mtDNA Illumina sequencing analysis. JT, TL, SS performed Nanopore sequencing analysis. MB, FH, PB, KKK, SD, AR, VK, AEL, PP, PS, AR, RY, ELG, KL, RY, NB, CK and AG were responsible for clinical recruitment, data acquisition and analyses. AG and CK supervised the study. All authors contributed to critical revision of the manuscript.