Preserved striatal innervation maintains motor function despite severe loss of nigral dopaminergic neurons

Abstract

Degeneration of dopaminergic neurons in the substantia nigra and their striatal axon terminals causes cardinal motor symptoms of Parkinson’s disease. In idiopathic cases, high levels of mitochondrial DNA alterations, leading to mitochondrial dysfunction, are a central feature of these vulnerable neurons.

Here we present a mouse model expressing the K320E variant of the mitochondrial helicase Twinkle in dopaminergic neurons, leading to accelerated mitochondrial DNA mutations. These K320E-Twinkle^DaN mice showed normal motor function at 20 months of age, although ∼70% of nigral dopaminergic neurons had perished. Remaining neurons still preserved ∼75% of axon terminals in the dorsal striatum and enabled normal dopamine release. Transcriptome analysis and viral tracing confirmed compensatory axonal sprouting of the surviving neurons.

We conclude that a small population of substantia nigra dopaminergic neurons is able to adapt to the accumulation of mitochondrial DNA mutations and maintain motor control.

Introduction

Parkinson’s disease (PD) is the most common neurodegenerative motor disease and affects millions of people worldwide.¹ Furthermore, the incidence of idiopathic cases rises with the demographic increase of life expectancy. Diagnosis of patients only occurs with the onset of cardinal motor symptoms, which are caused by the lack of striatal dopamine. At this point, already ∼30%–50% of dopamine-producing neurons in the substantia nigra (SN)^1-3 and ∼50%–70% of their axon projections in the striatum are lost.^4-7

Other dopaminergic neuron populations, however, are spared.^8-10 This selective vulnerability is linked to cell type-specific factors: SN dopaminergic neurons possess extraordinary long, extensively branched and unmyelinated axons,¹¹ which form an enormous number of synapses¹² and thus have high energetic demands.¹³ In addition, pacemaker activity of SN dopaminergic neurons generates oscillatory increases in free cytosolic Ca²⁺ levels, which are linked to cell death in the presence of mitochondrial dysfunction,^14,15 especially disturbing mitochondrial redox homeostasis.¹⁶

To explain the increased, ageing-related incidence of idiopathic PD, two major hypotheses were put forward: (i) the oligomerization and aggregation of α-synuclein, leading to Lewy body formation; and (ii) mitochondrial dysfunction. α-Synuclein pathology may result from its modest, long-term upregulation in idiopathic PD^17,18; however, the exact mechanisms for its elevation in monogenic PD remain unknown. On the other hand, mitochondrial respiratory chain deficiency is a well known age-related feature of SN dopaminergic neurons in PD patients.¹⁹ This defect is caused by the accumulation of mitochondrial DNA (mtDNA) deletions and gene duplications (indels) over time. Using the newly developed PCR method, two groups showed 30 years ago that such indels accumulate during ageing, and especially in brain regions rich in dopaminergic neurons.^20,21 More convincingly, analysis of single SN dopaminergic neurons confirmed these findings²² and showed that mtDNA deletions accumulate even more in PD.²³ We demonstrated in vivo and in vitro that dopamine metabolism is highly mutagenic for mtDNA.^24,25 Previously and astoundingly, the C. Tzoulis research group found that the total copy number of mtDNA in dopaminergic neurons of healthy individuals rises during normal ageing, thus compensating the effect of defective molecules. In PD patients, on the other hand, this compensatory response does not occur.²⁶

Mitochondrial complex I inhibitors, like MPTP and rotenone or genetic models, have been widely used to mimic mitochondrial dysfunction exclusively in dopaminergic neurons. However, these approaches usually lead to rapid neuron death and do not reflect the slowly progressing pathology in PD, including ageing-associated mtDNA alterations. We have developed a mouse model enabling the conditional overexpression of a dominant-negative mutant (K320E) of the mitochondrial helicase Twinkle (Rosa26-STOP-loxP-Cre System). Expression of K320E-Twinkle accelerates the accumulation of mtDNA indels in non-dividing cell types, such as cardiomyocytes²⁷ and skeletal muscle cells,²⁸ while it leads to a loss of mtDNA in rapidly replicating cells like the epidermis²⁹ or in B cells.³⁰ Here, we have expressed K320E-Twinkle in dopaminergic neurons using DAT-Cre expression as specific driver. Surprisingly, these animals do not develop a PD-like motor phenotype although 70% of SN dopaminergic neurons have perished at an advanced age of 20 months. This is due to compensatory branching of axon terminals by the surviving SN dopaminergic neuron population and adaptions in dopamine metabolism enabling normal dopamine signalling. Thus, we show that a small population of these neurons is able to compensate for harmful mtDNA mutations and escape degeneration, holding possibilities for new treatment strategies in patients with Parkinson’s disease.

Material and methods

Animals

K320E-Twinkle^DaN and control mice were bred at the University of Cologne. Heterozygous (K320E/Wt^DaN) and homozygous (K320E/K320E^DaN) K320E-Twinkle^DaN animals were generated by crossing DAT-cre mice [Cre gene inserted upstream of the translation start codon in exon 2 of the DAT gene³¹ with K320E-Twinkle transgenic mice (point mutation K320E; Rosa26-Stop-construct) downstream GFP]. DAT-cre mice were kindly provided by Nils-Göran Larsson (Karolinska Institutet, Sweden). K320E-Twinkle animals were generated previously by our group.²⁷ For the additional knockout (KO) of parkin, K320E-Twinkle^DaN mice were crossed with parkin KO animals, kindly provided by Olga Corti (Paris Brain Institute, Sorbonne University, France). Control mice were littermates harbouring the K320E-Twinkle construct (heterozygous or homozygous) in absence of the Cre allele (except viral tracing experiment). Mice were kept in individually ventilated cages at 23°C, 12:12 h light–dark cycle, with specified pathogen-free hygiene levels, free access to water and a regular chow diet ad libitum. All animals were regularly monitored for potential signs of pain and suffering. Breeding and experiments were conducted in agreement with European and German guidelines and approved by local authorities (LANUV, Landesamt für Natur, Umwelt und Verbraucherschutz Nordrhein-Westfalen, Germany; approval licenses: 84-02.04.2013-A141, 84-02.04.2016.A410, 81-02.04.2018-A210, 81-02.04.2021-A256). For all experiments, female and male animals were used. The experimental unit within each figure is presented by single animals. Sample size for animal experiments was calculated a priori together with the Institute for Medical Statistics and Bioinformatics of the University of Cologne (IMSB). In particular, the number of animals was determined by using G*Power (version 3.1.4; expected difference between groups by 80% power and 0.05 significance level). For behavioural assessments, allocation of different groups, i.e. the genotype, was unknown to the experimenter. In addition, only group-housed animals were used to exclude the influence of social deprivation on behavioural tests. Mice were genotyped by PCR using genomic DNA isolated from ear punches. Primer sequences and PCR conditions are available upon request. In total, 224 animals were used in this study.

Behavioural tests

Motor activity, motor coordination, depressive-like behaviour and social interaction were tested in K320E-Twinkle^DaN mice. For detailed set-up descriptions and protocols, see the Supplementary material.

Positron emission tomography

Mice were between 19 and 21.5 months old at the time of PET imaging. Tracers for ¹⁸F-FDOPA, ¹⁸F-FDG and ¹⁸F-MNI1126 were used to monitor dopamine synthesis and storage capacity, neuronal metabolic activity and presynaptic density in the striatum of K320E-Twinkle^DaN mice. For a detailed description, refer to the Supplementary material.

Viral tracing

For viral labelling of remaining SN dopaminergic neurons, 20-month-old K320E/K320E^DaN and control mice (R26-K320E-Twinkle^+/+; DAT^cre/+) were used. For anterograde tracing, Cre-dependent AAV-pCAG-FLEX-EGFP-WPRE (1.8 × 10¹³ particles/ml; Addgene viral prep #51502-AAV1) was unilaterally injected into the SN of anaesthetized animals at 3.0 mm posterior to bregma, 4.2 mm deep from dura and 1.6 mm lateral to midline.³² AAV-pCAG-FLEX-EGFP-WPRE was a gift from Hongkui Zeng.³³ For retrograde tracing, Cre-dependent pAAV-FLEX-tdTomato (2.1 × 10¹³ particles/ml; Addgene viral prep #28306-AAVrg) was injected into the same hemisphere at +0.7 mm anterior to bregma, 3.0 mm deep from dura and 1.8 mm lateral to midline. pAAV-FLEX-tdTomato was a gift from Edward Boyden. For all intracerebral injections, mice were initially anaesthetized with 4% isoflurane and fixed in a stereotaxic frame (#504926, WPI) at 2% isoflurane. Lubricating ointment was applied to the animals’ eyes to prevent them from drying (Bepanthen, Bayer). Body temperature was monitored by a rectal thermometer (RET-3, WPI) and kept constant by using a heating mat (MediHeat, V1200) and temperature control device (ATC2000-220, WPI). After reaching surgical tolerance, the incision site was shaved (HS 61 Contura, Wella) and cleaned with povidone iodine. Incision was made ∼1.5 cm centrally along the skull bone and craniotomy was performed with a micro drill system (#503599, WPI) above the target regions. By using a micro-injector (#UMP3T 1, MPI), a target volume of 0.5 μl was applied at 100 nl/min for anterograde tracing. For adequate delivery of the virus, the cannula remained at the injection site for another 5 min. Subsequently, a target volume of 2.0 μl was applied at 200 nl/min for retrograde tracing. After removing the cannula, the wound was sutured, and animals could recover in a heat box. Mice were subsequently provided with analgesics and carefully screened for signs of atypical behaviour and pain, according to the European animal welfare guidelines. Three weeks after viral injection, mice were perfused and brains prepared for downstream histological analysis.

dLight1.1

dLight1.1 measurement is described in detail in the Supplementary material.

Striatal tissue analysis

Striatal tissue for protein and RNA analyses was precisely collected from cryosections of 20-month-old K320E/K320E^DaN and control animals. Sample preparation and subsequent analysis via western blot and RT-PCR are described in detail in the Supplementary material.

RNA sequencing and mtDNA analysis of laser-captured substantia nigra dopaminergic neurons

Isolation of dopaminergic neurons and downstream analysis of RNA and mtDNA are extensively described in the Supplementary material.

Histological analysis and immunostaining

In-depth protocols for immunohistological and biochemical stainings can be found in the Supplementary material.

Data analysis

Data analysis was performed using GraphPad Prism 8 (GraphPad Software, Inc.). Quantified data are presented as mean + standard error of the men (SEM). For all box plots used in this work, the centre line indicates the median, the mean is shown as ‘+’, the box borders present the first and third quartiles, and the whiskers indicate the data range. Relative data are shown as percentage of control animals. Values of sample size (n) refer to actual mouse numbers. Unpaired t-test, one-way and two-way ANOVA with post hoc comparisons (Bonferroni post hoc test) were used to define differences between groups. The cumulative frequency of heteroplasmic mitochondrial single nucleotide variants (mtSNVs) between genotypes and ages was computed by two-sample Kolmogorov-Smirnov test. A significance level of 0.05 was accepted for all statistical tests.

Graphical illustrations were generated using Photoshop CS2 (Adobe Systems Software). Schematics created with BioRender.com are indicated in the Supplementary material.

Results

Severe degeneration of dopaminergic midbrain neurons in K320E-Twinkle^DaN mice

To investigate the impact of slowly accumulating mtDNA alterations on dopaminergic neurons in vivo, mice harbouring a mutant variant of the mitochondrial helicase Twinkle (K320E²⁷) were crossed with mice expressing Cre recombinase under control of the dopamine transporter promoter (DAT; Fig. 1A). Expression of K320E-Twinkle was confirmed by immunofluorescent detection of green fluorescent protein (GFP), cloned downstream, in coronal midbrain sections (Fig. 1B and C and Supplementary Fig. 1A). Approximately 98% of tyrosine hydroxylase (TH)-positive neurons in the SN and ventral tegmental area (VTA) expressed GFP in 5-month-old mice (Fig. 1D). Conversely, ∼98% of GFP-positive neurons expressed TH, showing the specificity and effectiveness of Cre-mediated recombination of K320E-Twinkle^DaN in dopaminergic neurons. While no difference in the total amount of TH-positive dopaminergic neurons was observed at 5 months of age, 10-month-old K320E-Twinkle^DaN mice revealed a significant loss (Fig. 1E and F). Mice carrying two copies of K320E-Twinkle (K320E/K320E^DaN) presented a higher degree of neuron death (∼57%) than animals with only one copy (∼33%, K320E/Wt^DaN). At the well advanced age of 20 months, degeneration was even more pronounced: K320E/K320E^DaN mice had lost ∼77% and K320E/Wt^DaN mice ∼41% of dopaminergic neurons. To examine whether neuron loss in K320E-Twinkle^DaN mice was restricted to a specific midbrain region, dopaminergic neurons were assigned to the SN and VTA, respectively. Neurons were lost to a rather similar extent in heterozygous (∼33%) and homozygous K320E-Twinkle^DaN mice (∼57%) at 10 months of age (Fig. 1G). From 10 to 20 months, the amount of dopaminergic neurons did not further change in K320E/Wt^DaN mice (for time line and statistics, see Supplementary Fig. 1B). In K320E/K320E^DaN mice, neuron number further decreased in the VTA (from ∼55% to ∼84% loss), while there was no additional significant loss in the SN (∼59% to ∼70% loss) (Fig. 1G and Supplementary Fig. 1B).

Neurodegeneration is preceded by reduced levels and activity of mitochondrial complex IV

To evaluate mitochondrial respiratory chain abundance in dopaminergic neurons, we performed quantitative analysis of the mtDNA encoded subunit I assembled into cytochrome c oxidase (COX; complex IV) using immunofluorescent staining (Fig. 2A, representative micrographs display SN neurons). At 5 months of age, before the onset of neurodegeneration, COXI levels were significantly decreased in both SN and VTA dopaminergic neurons of hetero- and homozygous K320E-Twinkle^DaN mice (Fig. 2B). Noteworthy, intensity of COXI was not uniformly affected in all investigated neurons. While the mean COXI level was reduced by ∼36% (K320E/Wt^DaN) and ∼49% (K320E/K320E^DaN), single cell analysis revealed different populations of mutant dopaminergic neurons, including cells with much lower (up to ∼70%) as well as normal COXI levels when compared to controls (Fig. 2B). In line, reduced levels of the complex I subunit NDUFB11 were found in both SN and VTA dopaminergic neurons of 5-month-old K320E-Twinkle^DaN mice (Supplementary Fig. 2A). In addition, we analysed COX enzymatic activity by histochemical staining in frozen sections. Blue, COX-deficient cells were readily detectable in both SN and VTA of K320E-Twinkle^DaN mice at 5 months (Fig. 2C) and 10 months of age (Supplementary Fig. 2B and C). At 20 months, remaining neurons revealed normal COXI levels (Supplementary Fig. 2E) and accordingly, only a few cells with deficient COX activity were apparent (Supplementary Fig. 2D). Thus, either surviving neurons have upregulated respiratory chain complexes or only cells with high respiratory chain content have survived.

Substantia nigra neurons accumulate mtDNA single nucleotide variants following K320E-Twinkle expression

With increasing age, SN dopaminergic neurons accumulated mtDNA indels in both control and K320E/K320^DaN mice (Fig. 3A), however expression of K320E-Twinkle did not change mtDNA indels number relative to control mice (Fig. 3B). Thus, we next examined the burden of mtSNVs of potential somatic origin (heteroplasmy fraction, HF, 1%–5%) or after clonal expansion (HF 5%–90%). Pooled SN dopaminergic neurons from 5-month-old K320E/K320E^DaN mice revealed a higher cumulative frequency of somatic mtSNVs compared to control mice and K320E/K320E^DaN animals at 20 months (Fig. 3C). In addition, the cumulative frequency of clonally expanded mtSNVs in 5-month-old K320E/K320E^DaN mice was comparable to the mtSNVs frequency of their 20-month-old counterparts, which suggests an early acquisition of mtSNVs resulting in a state of accelerated mtDNA ageing. mtSNV accumulation was apparently compensated by a pronounced increase of mtDNA copy number at 5 months, and the 30% surviving mutant neurons showed equally high copy numbers compared to the intact pool of control neurons at 20 months (Fig. 3D).

Normal motor behaviour despite severe loss of substantia nigra dopaminergic neurons

Next, we explored the impact of SN neurodegeneration on motor function of K320E-Twinkle^DaN mice. Spontaneous motor activity was analysed using the open field BeamBreak test (Fig. 4A). Surprisingly, neither horizontal locomotion nor vertical rearing of K320E-Twinkle^DaN mice differed from control littermates at 20 months of age (Fig. 4B and C). In addition, motor coordination was examined by the RotaRod test (Supplementary Fig. 4A). Ten- and 20-month-old K320E/K320E^DaN mice showed similar latency periods to fall from the rotating rod when compared to age-matched control mice (Supplementary Fig. 4B).

Dopaminergic innervation is preserved in the dorsal but not ventral striatum

Absent motor impairment in K320E-Twinkle^DaN mice suggests that, despite severe degeneration of SN dopaminergic neurons, sufficient levels of dopamine are released in the dorsal striatum maintaining normal motor function. To test this hypothesis, we first analysed the density of dopaminergic axon terminals in the striatum by TH immunohistochemistry (Fig. 4D). TH levels were determined in the dorsal and ventral striatum, which harbour axonal projections of dopaminergic neurons from SN and VTA, respectively. At 5 months, TH signal did not differ between K320E/Wt^DaN mice and control littermates (Fig. 4E), while K320E/K320E^DaN mice revealed a slight reduction of TH-positive fibres in the dorsal striatum (∼17%), whereas the ventral striatum was unchanged. Interestingly, at 10 months of age, similar TH levels were detected in the dorsal striatum in all groups. At that age, numbers of dopaminergic SN neurons, in turn, were already diminished by ∼41% (K320E/Wt^DaN) and ∼59% (K320E/K320E^DaN), respectively (Fig. 1). In the ventral striatum, however, TH-positive fibre density decreased by ∼15% in heterozygous and by ∼30% in homozygous K320E-Twinkle^DaN animals (Fig. 4E). After 20 months the discrepancy in TH immunoreactivity between the dorsal and ventral striatum continued to grow. Whereas TH-positive fibre density in the ventral striatum was decreased by ∼46% (K320E/Wt^DaN) and ∼67% (K320E/K320E^DaN), respectively, TH immunoreactivity in the dorsal striatum was preserved at ∼94% in heterozygous and ∼77% in homozygous K320E-Twinkle mice. Similar results were obtained by DAT immunohistochemistry (Supplementary Fig. 4C–F). As VTA neurons projecting to the ventral striatum are involved in reward and social behaviour, we next tested if loss of the dopaminergic mesolimbic system has consequences in K320E-Twinkle^DaN mice. While K320E-Twinkle^DaN animals showed normal immobility times during the tail suspension test (Supplementary Fig. 4G and H), they consumed significantly lower amounts of sucrose water in the sucrose preference test (Supplementary Fig. 4I and J). Moreover, K320E/K320E^DaN mice showed a tendency for increased social affiliation but no abnormalities when exposed to new social stimuli in the three chamber test (Supplementary Fig. 4K–N).

Decreased dopamine synthesis capacity but normal transients of striatal dopamine release

Next, we explored the in vivo dopamine synthesis and storage capacity in the striatum by PET using the radioligand ¹⁸F-FDOPA (Fig. 5A). Cerebellar ¹⁸F-FDOPA uptake mostly reflects unspecific accumulation and was used as background signal for intensity normalization, as its dopaminergic innervation is minimal and dopamine binding sites are ∼10 times less abundant than in the striatum.³⁴ While 20-month-old control mice showed pronounced ¹⁸F-FDOPA uptake, it was considerably reduced in K320E/K320E^DaN mice (Fig. 5B). Volume of interest analysis revealed a mean value of 1.12 ± 0.02 in controls and 1.06 ± 0.01 in K320E/K320E^DaN mice. As cerebellar ¹⁸F-FDOPA uptake was set to 1.00, the striatal ¹⁸F-FDOPA uptake in K320E/K320E^DaN mice can be interpreted as a ∼50% reduction compared to control animals. PET tracers reflecting presynaptic density (¹⁸F-MNI1126) and general neuronal metabolic activity (¹⁸F-FDG) revealed no difference between control and K320E/K320E^DaN mice (Supplementary Fig. 5A and B).

To examine striatal dopamine release in freely moving mice, we next performed fibre photometry using the genetically encoded extracellular dopamine optical sensor dLight1.1.³⁵ An adeno-associated virus (AAV) encoding the dLight1.1 construct was unilaterally injected into the dorsal striatum of 20-month-old mice (Fig. 5C), followed by implantation of an optical fibre for induction and collection of the dLight1.1 signal (Fig. 5D). Dopamine release was triggered in awake, freely behaving mice by three salient stimuli, which are well characterized to evoke dopaminergic activity: sudden door-opening of the experimental cage (Fig. 5E), delivery of a palatable sucrose pellet (Fig. 5H) and change to a new cage (Fig. 5K).³⁶ K320E/K320E^DaN mice revealed attenuated raw dLight1.1 signals compared to control animals, suggesting lower basal and absolute striatal dopamine levels (Supplementary Fig. 5C). Importantly, however, the relative dopamine response was indistinguishable in both dynamics and intensity between K320E/K320E^DaN and control mice following all three stimuli (Fig. 5E–M). Taken together, dopamine photometry analyses revealed that K320E/K320E^DaN mice may have lower basal and absolute striatal dopamine levels, which is in agreement with our ¹⁸F-FDOPA results. However, the relative increase in striatal dopamine in response to different salient stimuli is unaffected.

Western blot analysis of striatal samples, including both dorsal and ventral regions, revealed low TH and DAT levels at 20 months, reflecting the loss of dopaminergic axon projections, especially in the ventral region (Fig. 5N and O). It is noteworthy that relative levels of DAT were more reduced than TH in K320E/K320E^DaN (refer to the ‘Discussion’ section). K320E/K320E^DaN striata showed enhanced levels of both dopamine 1 (D1R) and 2 receptors (D2R) and decreased levels of catechol-O-methyltransferase (COMT), which is responsible for dopamine degradation and predominantly represented in the brain by its membrane-bound isoform (MB-COMT).³⁷ There were no changes in protein levels of monoamine oxidase A (MAO-A) (Supplementary Fig. 5D and E), which was recently shown to be the monoamine oxidase mainly contributing to dopamine degradation.³⁸ Levels of MAO-B, conversely, were increased but this was not due to an enhanced presence of astrocytes, where MAO-B is primarily localized,^39,40 according to the astrocytic marker glial fibrillary acidic protein (GFAP). These findings point to a compensatory adaptation of enhanced dopamine signalling and reduced dopamine degradation in the striatum of K320E/K320E^DaN mice.

Finally, it has been shown that in striatal dopaminergic axons, only ∼30% of varicosities contain active zone-like sites, which can be identified by the expression of the scaffolding protein bassoon.⁴¹ Co-staining for TH and bassoon showed similar numbers of bassoon expressing varicosities in TH-labelled dopaminergic axon terminals in both genotypes (Supplementary Fig. 5F–I), excluding an increased recruitment of active zone-like sites as a mechanism for sufficient dopamine release.

Surviving neurons upregulate pathways for axon guidance, calcium signalling and ligand-receptor interaction

To unmask changes in gene expression following K320E expression, cDNA from control and K320E/K320E^DaN SN dopaminergic neurons was sequenced at 5 and 20 months of age (Supplementary Fig. 6A). SN neurons showing TH immunoreactivity were precisely isolated by laser capture microscopy⁴² (Fig. 6A and Supplementary Fig. 6B–D) and revealed enrichment of genes characteristically expressed in dopaminergic neurons (Fig. 6B). Genes differentially expressed in 5-month-old (Supplementary Fig. 6F and H) and 20-month-old K320E/K320E^DaN mice (Fig. 6C and Supplementary Fig. 6E) were clustered by pathway and process enrichment analysis. Among the most powerfully represented pathways, surviving SN dopaminergic neurons revealed upregulation of calcium signalling and neuroactive ligand-receptor interaction, as well as axon guidance and regulation of actin cytoskeleton (Fig. 6C). Protein-protein interaction analysis of upregulated genes identified calcium and acetylcholine signalling, Ras, MAPK and TGFb signalling, and axon guidance as the most strongly pronounced networks (Fig. 6D). Axon guidance and actin skeleton organization were also identified as upregulated pathways in SN dopaminergic neurons of 5-month-old K320E/K320E^DaN animals (Supplementary Fig. 6F and G). Due to the striking upregulation of receptors for factors modulating axon growth in surviving SN dopaminergic neurons, we examined expression of neurotrophic factors and guidance molecules by RT-qPCR in the striatum of 20-month-old K320E/K320E^DaN mice (Fig. 6E). While there was no difference in mRNA levels for the conventional neurotrophic factors Ngf, Bdnf and Gdnf, K320E/K320E^DaN mice revealed upregulation of netrin 1 (Ntn1) and ephrin- A2 (Efna2) (Fig. 6F), which have both been reported to positively influence the innervation of the dorsal striatum by SN dopaminergic neurons.^43,44 This was accompanied by downregulation of Sema3A and Slit2, which in turn, are linked to the inhibition of axon branching.^45-47

Surviving substantia nigra dopaminergic neurons preserve dorsal striatum innervation

Our results indicate that the remaining population of SN dopaminergic neurons in K320E/K320E^DaN mice preserve dopamine supply in the dorsal striatum by compensatory axonal sprouting despite the presence of K320E-Twinkle. Although we have shown that almost 100% of SN dopaminergic neurons expressed the construct before the onset of neuron loss (Fig. 1B–D), a small proportion of the surviving cells could still lack the mutant Twinkle. It should be noted that previous studies reported on the existence of TH-positive neurons with little or no DAT expression,^48-50 which in our case, would affect the efficiency of DAT-Cre-mediated recombination. Immunofluorescent detection of TH and DAT, however, revealed a high level of co-localization in remaining SN neurons of 20-month-old K320E/K320E^DaN mice (Supplementary Fig. 7A–C). To collectively confirm K320E expression and the compensatory axonal sprouting of surviving SN dopaminergic neurons in K320E/K320E^DaN mice, we performed anterograde viral tracing, using an AAV that expresses enhanced green fluorescent protein (EGFP) under control of the Cre/loxP system through a modified Flex switch (AAV-FLEX-EGFP; Fig. 7A). AAV-FLEX-EGFP was unilaterally injected into the SN of 20-month-old K320E/K320E^DaN mice and corresponding control animals expressing DAT-Cre (Fig. 7B). Viral injection led to a robust expression of EGFP in TH-labelled dopaminergic midbrain neurons (Fig. 7C and Supplementary Fig. 7D). The infection efficiency of TH-labelled dopaminergic neurons in the SN was ∼90% for both K320E/K320E^DaN and control animals and significantly higher than in the VTA (∼45%; Fig. 7D). Conversely, ∼99% of EGFP-positive neurons expressed TH, which excludes potential off-target viral infection (Supplementary Fig. 7E). The number of remaining midbrain neurons in K320E/K320E^DaN animals that express TH and EGFP was consequently reduced to a similar extent (Fig. 7E and Supplementary Fig. 7F). In line with the high infection efficiency of SN dopaminergic neurons, analysis of the axon projection area revealed EGFP expression preferentially in the dorsal striatum for both K320E/K320E^DaN and control mice (Fig. 7F and G). Non-injected hemispheres were devoid of EGFP signal (Supplementary Fig. 7G and H). In comparison to the dorsal striatum of control animals, EGFP-positive fibre density of K320E/K320E^DaN mice was decreased by ∼59% (Fig. 7G), but importantly, matched with TH-labelled axon terminals by ∼91% (Fig. 7H).

Discussion

In idiopathic PD, a high mtDNA mutation load²³ meets low wild-type copy numbers in dopaminergic neurons, in contrast to unaffected individuals in which copy number increases during ageing.²⁶ This renders impaired mtDNA homeostasis an important puzzle piece for age-dependent SN dopaminergic neuron loss. The K320E-Twinkle mutation slows down mtDNA replication in dividing cells⁵¹ and induces indels in non-dividing cells.^27,28 Our data show that in SN dopaminergic neurons, neither deletions nor duplications of mtDNA were further increased by K320E expression. Instead K320E-Twinkle led to an early accumulation of single nucleotide variants (SNVs) before the onset of neurodegeneration, resulting in a state of accelerated mtDNA ageing, which has not been reported before. As a response, SN dopaminergic neurons from K320E/K320E^DaN mice revealed an early compensatory upregulation of mtDNA copy number to levels present in control animals after 20 months, similar to what is observed in healthy ageing humans, but not in idiopathic PD patients.²⁶ High levels of wild-type mtDNA have consistently been shown to counteract the consequences of mtDNA mutations.^52-54 In the POLG mutator mouse, which accumulates both SNVs and deletions in mtDNA, similar results were reported. MtDNA deletions in the midbrain were associated with an upregulation of mtDNA copy number.⁵⁵ Dawson and colleagues found an increase of D-loop sequence amplicons but no change in other mtDNA sequences, probably indicative of deletions in those regions.⁵⁶ In the work of Youle and colleagues, the measured decrease in COXI amplicons does not necessarily reflect a decrease in total copy number, since deletions in this large region have not been analysed and would explain the decrease as well.⁵⁷

However, the early accumulation of SNVs was associated with the reduction of respiratory chain complexes and caused degeneration of a large proportion of neurons in both SN and VTA starting between 5 and 10 months. In contrast to SN dopaminergic neurons from K320E/K320E^DaN mice at 5 months of age, after 20 months the surviving population showed normal ageing-associated levels of SNVs, which was accompanied by normal respiratory chain function, suggesting efficient mitochondrial quality control. The E3 ubiquitin ligase parkin regulates the clearance of dysfunctional mitochondria via mitophagy.⁵⁸ Mutations in the encoding PARK2 gene cause autosomal recessive PD with early disease onset⁵⁹ and the ligase activity of parkin is assumed to be affected in idiopathic cases as well.⁶⁰ We therefore tested if, in the absence of parkin, the phenotype of K320E/K320E^DaN animals would get worse and motor deficits might emerge. Parkin deficiency neither changed the number of remaining dopaminergic neurons nor the density of dopaminergic axon terminals or motor behaviour (Supplementary Fig. 8A–F). We hence conclude that in surviving SN dopaminergic neurons of K320E/K320E^DaN mice, mitochondrial quality control occurs in a parkin-independent manner, which is in line with previous studies questioning the relevance of parkin for the survival of dopaminergic neurons in rodents.^61,62 Besides the wide variety of mitophagy pathways that can act in the absence of parkin,⁶³ our group has recently described a new pathway for the selective turnover of mutation-bearing mtDNA involving VPS35,⁶⁴ mutations in which are linked to autosomal dominant PD.⁶⁵

Although ∼70% of SN dopaminergic neurons had perished, K320E/K320E^DaN mice did not show any signs of motor impairment at the late age of 20 months, which is in stark contrast to other genetically or chemically induced PD mouse models.^31,66-71 Additional immunohistochemical approaches, as well as viral tracing, confirmed the severe loss of dopaminergic neurons (Fig. 7E and Supplementary Fig. 7A–C and E). Normal motor behaviour is explained by the maintenance of ∼75% of dopaminergic axon terminals in the dorsal striatum. In the ventral striatum, on the other hand, only ∼30% of dopaminergic projections originating from the VTA remained, which caused depressive traits and social abnormalities in K320E/K320E^DaN animals. The largely preserved dopaminergic innervation in the dorsal striatum of K320E/K320E^DaN mice was accompanied by a ∼50% reduction in ¹⁸F-FDOPA uptake in the entire striatum, indicating a decreased capacity of L-DOPA conversion to dopamine and/or its storage.^72,73 However, normal uptake of ¹⁸F-MNI1126 and ¹⁸F-FDG suggests a compensatory upregulation of presynaptic density and, consequently, neuronal metabolic activity. Nevertheless, both tracers are not specific for dopaminergic axons and the dopaminergic terminal fraction might be too small to see significant differences in tracer uptake. In line with this, an analogous tracer to ¹⁸F-MNI1126 revealed no difference in the striatum of PD patients.⁷⁴ As ¹⁸F-MNI1126 only detects synaptic vesicle glycoprotein 2A (SV2A),⁷⁵ it is also possible that SV2A is not the predominant SV2 protein in dopaminergic axon terminals, as reported recently.⁷⁶ Finally, K320E/K320E^DaN mice did not show an upregulation of active-zone release sites at dopaminergic axon terminals.

Noteworthy, PET scanning was performed in anaesthetized animals. To analyse in vivo dopamine transients in freely behaving, awake K320E/K320E^DaN mice, we used the extracellular dopamine sensor dLight1.1 expressed in the dorsal striatum. In line with our ¹⁸F-FDOPA results, K320E/K320E^DaN mice showed lower baseline dopamine levels. Most importantly, however, the relative dopamine response following different tasks was indistinguishable from control animals. In addition, K320E/K320E^DaN mice show a compensatory upregulation of striatal D1Rs and D2Rs, indicating an increased sensitivity of medium spiny neurons to dopamine, as well as downregulation of DAT and COMT, suggesting decreased dopamine breakdown. Upregulation of D2Rs⁷⁷ and downregulation of DAT^6,78,79 are phenomena also found in the early stage of idiopathic PD. In contrast to K320E/K320E^DaN mice, in PD patients this compensatory response is preceded by a ∼50%–70% loss of dopamine axon projections in the striatum.^4-7

Preserved dopamine axon terminals in the dorsal striatum of K320E/K320E^DaN mice is accompanied by upregulation of pathways promoting axonal growth and branching in surviving SN dopaminergic neurons at 20 months. Among others, we identified genes operating in Ras signalling, which is involved in nearly all stages of axiogenesis (reviewed by Hall et al.⁸⁰) and TGFβ and MAPK signalling, promoting dopaminergic neuron survival and branching in vitro and in vivo.^81,82 Among the upregulated genes involved in axon guidance, we found Pak4, which has previously been shown to prevent degeneration of SN dopaminergic neurons and motor impairment in rat models of PD.⁸³ The upregulation of receptors for cues modulating axon growth, including ephrin, prompted us to also examine expression of genes for axon guidance molecules in the striatum. K320E/K320E^DaN mice showed elevated mRNA levels of Ntn1 and Efna2. Netrin-1 is known to promote axon growth and branching in vitro,^45,46 and more importantly, it restored dopaminergic axon projections in the striatum of pharmacologically induced PD mouse models.⁴³ Ephrin-A signalling contributes to the guidance of dopaminergic projections towards the dorsal striatum during development,⁴⁴ a potential link for Ephrin-A2 to dopaminergic axon sprouting at advanced age, however, has not been reported before. K320E/K320E^DaN mice further showed downregulation of striatal Sema3A and Slit2, which are both described to inhibit axonal growth and branching.^45,46,84,85 While gene expression analysis in PD patients suggests a rather chemorepulsive environment, which might contribute to the denervation of striatal dopaminergic projections,⁸⁶ transcriptional changes in both the striatum and surviving SN dopaminergic neurons of K320E/K320E^DaN mice indicate predominant signalling towards the attraction and stabilization of dopaminergic axon terminals. The candidates presented here could hence be of relevance for the induction of dopaminergic axon sprouting, as supported by preclinical studies.^43,83 Genome wide analyses of single nucleotide polymorphisms in PD patients showed axon guidance among the most significant pathways affected, including polymorphisms for the Netrin-1 receptor DCC, several Ephrin-A and -B receptors, as well as the Slit receptor ROBO3.^87,88

Intriguingly, calcium and acetylcholine signalling were also upregulated in surviving SN dopaminergic neurons of K320E/K320E^DaN mice, which could not only support axon growth.^89-91 The group of Pascal Kaeser recently presented a mechanism for striatal dopamine release supported by acetylcholine, enabling its local control independent of electrical input from the midbrain.⁹² Regarding the low number of SN dopamine neurons left and the enhanced metabolic burden associated with axon sprouting,⁹³ it might thus be possible that normal dopamine release in K320E/K320E^DaN mice is additionally supported by local cholinergic transmission.

We also determined the transcriptome of 5-month-old animals (extended data Fig. 6F–H) and found axon guidance, actin skeleton organization, but also apoptosis as the most upregulated pathways. However, we anticipated that in-depth analysis of the transcriptome at this stage would yield ambiguous results due to the heterogeneity between neurons doomed to die and those that will survive, respectively. Therefore, we decided to extensively analyse the transcriptome of the surviving neurons at 20 months only, knowing that this will not allow us to distinguish between surviving neurons being a preadapted population from the beginning or a population that is able to mount adaptive processes following mitochondrial dysfunction.

In line with the transcriptional adaptations, we finally showed, by viral tracing, that the preserved dopaminergic innervation in the dorsal striatum of 20-month-old K320E/K320E^DaN mice indeed came from the small surviving population of SN neurons. It is worthy to note that the proportion of remaining dopaminergic projections in the dorsal striatum was lower compared to our previous densitometric analyses using TH and DAT immunohistochemistry (Fig. 4D and E and Supplementary Fig. 4C–G). This was certainly due to the stereotaxic injection procedures (Supplementary Fig. 7G), most likely into the dorsal striatum (data not shown), which are known to locally cause tissue damage.^94,95 In the non-injected hemisphere, in turn, the density of dopaminergic fibres in the dorsal striatum of K320E/K320E^DaN mice was similar to the results of prior densitometric analyses. Importantly, TH-labelled axon terminals matched with the EGFP signal originating from SN dopaminergic neurons by ∼91%.

Together, our data show that upon accelerated accumulation of mtDNA mutations in mice, a subset of SN dopaminergic neurons adapts to this insult and is able to fully compensate the severe reduction of the entire population at an advanced age. This is mediated by axonal sprouting through guidance molecules and adaptation of dopamine metabolism, enabling normal dopamine release and thus motor performance. These findings may be of utmost importance for the development of new treatment strategies.

Data availability

All data generated or analysed in this study are included in this published article and its Supplementary material, or available from the authors upon request.

Acknowledgements

We acknowledge Steffen Koch, Katrin Wollenweber, Robin Wolter and especially Katrin Lanz for their extensive help by preparing paraffin-embedded brain sections and genotyping of the used animal cohorts. We would further like to thank Prof. Olga Corti (Paris Brain Institute, Sorbonne University) for sharing the parkin KO mouse, the Eva Hedlund lab (Stockholm University) for sharing protocols and experience on the quick TH stain and Prof. Matteo Bergami (CECAD, Cologne) for sharing his expertise on viral tracing. We wish to thank the Cologne Center for Genomics (CCG) for RNA sequencing, especially Janine Altmüller, and our mechanical engineers under former leadership of Harald Metzner and Hans-Josef Reimer for the in-house construction of the equipment used for behavioural assessments. For their continuous support concerning microscopy, we conclusively would like to acknowledge the CECAD Imaging Facility, especially Dr Christian Jüngst.

Funding

T.P and R.J.W received funds from the Center of Molecular Medicine Cologne (CMMC, C17); K.M.R. received funding from Parkinson Canada Basic Research Fellowship (BRF-2021-0000000048); A.C. received support by the Cologne Graduate School of Ageing Research; R.S.C. and J.P. received funds from the Medical Research Council UK (MC_UU_00028/5). Y.N. and P.C. were supported by a Wellcome Collaborative Award (224486/Z/21/Z), the Medical Research Council Mitochondrial Biology Unit (MC_UU_00028/7), the Medical Research Council (MRC) International Centre for Genomic Medicine in Neuromuscular Disease (MR/S005021/1), the Leverhulme Trust (RPG-2018-408), an MRC research grant (MR/S035699/1), an Alzheimer's Society Project Grant (AS-PG-18b-022), and the NIHR Cambridge Biomedical Research Centre (BRC-1215-20014); M.A. acknowledges financial support by the Friebe Foundation (T0498/28960/16) and the Deutsche Forschungsgemeinschaft (DFG, German Research Foundation)—Project-ID 431549029—SFB 1451.

Competing interests

The authors report no competing interests.

Supplementary material

Supplementary material is available at Brain online.

References