Abstract
Heterochromatin disorganization is a key hallmark of aging and DNA methylation state is currently the main molecular predictor of chronological age. The most frequent neurodegenerative diseases like Parkinson disease and Alzheimer’s disease are age-related but how the aging process and chromatin alterations are linked to neurodegeneration is unknown. Here, we investigated the consequences of viral overexpression of Gadd45b, a multifactorial protein involved in active DNA demethylation, in the midbrain of wild-type mice. Gadd45b overexpression induces global and stable changes in DNA methylation, particularly on gene bodies of genes related to neuronal functions. DNA methylation changes were accompanied by perturbed H3K9me3-marked heterochromatin and increased DNA damage. Prolonged Gadd45b expression resulted in dopaminergic neuron degeneration accompanied by altered expression of candidate genes related to heterochromatin maintenance, DNA methylation or Parkinson disease. Gadd45b overexpression rendered midbrain dopaminergic neurons more vulnerable to acute oxidative stress. Heterochromatin disorganization and DNA demethylation resulted in derepression of mostly young LINE-1 transposable elements, a potential source of DNA damage, prior to Gadd45b-induced neurodegeneration. Our data implicate that alterations in DNA methylation and heterochromatin organization, LINE-1 derepression and DNA damage can represent important contributors in the pathogenic mechanisms of dopaminergic neuron degeneration with potential implications for Parkinson disease.
Introduction
Epigenetic marks separate chromatin into actively transcribed euchromatin and repressive heterochromatin domains, and participate in the spatial organization of the genome into highly structured 3D domains 1. These epigenetic signatures include DNA methylation, various post-translational modifications of histones, and attraction forces between different types of genomic repeat elements, including transposable elements (TEs) 1. Changes in chromatin structure regulate DNA accessibility to transcription factors and the physical proximity of enhancers to promoters, thereby regulating gene expression.
It is well established that DNA methylation regulates various important cellular processes during development and cell differentiation. In recent years, however, perturbations of chromatin organization have been linked to the aging process and global changes in DNA methylation are currently the main molecular predictor of chronological age (reviewed in 2). Aging-induced epigenetic remodeling of chromatin can impact on genomic stability 3,4, and vice versa 2. Thus chromatin states and genomic stability are important, interdependent factors associated with the aging process 5. One emerging culprit related to both processes is the un-silencing of transposable elements (TEs) with age. Around half of the human genome is comprised of TEs. The evolutionary most successful TEs in mammals are long interspersed nuclear element-1 (LINE-1 or L1). Mostly fossilized and a few remaining active LINE-1 sequences (around 100 in humans and 3000 in mice) represent around 17% of the human 6 and 21% of the mouse genome 7. Young and full-length LINE-1 elements are autonomous retrotransposons, expanding in the genome through a “copy and paste” retrotransposition mechanism and encoding the two necessary proteins, namely ORF1p and ORF2p, required for their mobilization. ORF1p is an RNA binding protein with strong “cis” preference 8–10 and ORF2p encodes an endonuclease and a reverse transcriptase 11,12. Several repressive cellular mechanisms, including DNA methylation and heterochromatinization, limit their expression 13. When these fail with age, TEs can become derepressed 14. An increased activity of LINE-1 is associated with genomic instability through the induction of DNA damage 15–18.
How aging and neurodegeneration are linked at the molecular level remains widely unknown. This question is, however, of high relevance as age is the primary risk factor for the most common neurodegenerative diseases (NDs) including Parkinson disease (PD) and Alzheimer disease (AD) 19. Some of the cellular processes defined as the hallmarks of aging 20 overlap with pathways shown to be dysfunctional in NDs. This is the case for impaired proteostasis, mitochondrial dysfunction, deregulated nutrient sensing, increased oxidative stress and neuroinflammation 21, 22. Whether two other important nominators of aging, namely perturbations in the chromatin organization and genomic instability, are associated with neuronal aging and neurodegeneration has not been unequivocally proven yet.
A cardinal feature of PD is the degeneration of midbrain dopaminergic (mDA) neurons in the substantia nigra pars compacta (SNpc). These neurons project to the striatum and their loss leads to a striatal deficiency in the neurotransmitter dopamine, inducing the typical motor symptoms of PD. Decreased global DNA methylation with age in the SNpc has been observed 23 and DNA methylation changes, mostly on specific genetic risk loci, have been linked to several NDs 24 including PD 25. Alterations in histone modifications have also been observed in PD 26. However, the possible contribution of age-related epigenetic alterations to the pathogenesis of PD and the onset of neurodegeneration has not been demonstrated yet.
Here, we investigated how SNpc mDA neurons react to perturbations of chromatin organization. For this purpose, we overexpressed Gadd45b in the SNpc of wild-type mice using an adeno-associated virus (AAV) vector. GADD45B is a multifunctional protein which coordinates in the nucleus an active DNA demethylation pathway involving cytidine deaminases and DNA glycosylases 27 in association with the base excision repair (BER) pathway 28. We chose GADD45B because it is a known DNA demethylase in postmitotic neurons 27, 29 and it is highly inducible in conditions of oxidative stress in the SNpc 30. We show that overexpression of Gadd45b in the SNpc of wild-type mice leads to widespread perturbations of DNA methylation, heterochromatin disorganization, increased vulnerability of mDA neurons to oxidative stress, activation of LINE-1 elements, DNA strand breaks and neuronal death. Our data reinforces the hypothesis that aging-induced global chromatin disorganization initiates neurodegeneration, possibly via the derepression of LINE-1 elements.
Results
Gadd45b overexpression in the SNpc of wild-type mice leads to early and stable DNA methylation perturbations in gene bodies of genes related to neuronal functions
Wild-type mice littermates (6 weeks old) were injected unilaterally in the SNpc using AAV8-mCherry or AAV8-mGadd45b. The animals were sacrificed after 14 or 90 days post-injectionem (p.i.) and either perfused or dissected as schematized in Figure 1A. The AAV8-mCherry control virus efficiently infected tyrosine hydroxylase positive (TH+) neurons of the SNpc as shown by mCherry expression at 14d p.i. (Fig. 1B, upper panel). Due to the lack of a good antibody against GADD45B, expression of Gadd45b was verified at 14d p.i. by in situ hybridization (Fig1. B, lower panel) and RT-qPCR after manual micro-dissection of the SNpc at 14d and 90d p.i. (Fig. 1C). Gadd45b transcripts on the injected ipsilateral side were increased up to 79-fold on average (11.02±4.83; 869.1±283; n=6) at 14d p.i. and up to 178-fold (7.89±1.52; 1402±683; n=4) at 90d p.i. compared to the endogenous transcript levels of the non-injected contralateral side.
Having verified the efficient overexpression of Gadd45b in the SNpc, we asked whether this perturbs DNA methylation patterns in the SNpc. To this end, we injected wild-type mice with AAV8-mGadd45b (n=6) or AAV8-mCherry (n=6) and simultaneously extracted DNA and RNA from manually micro-dissected SNpc biopsies at 14d p.i. (Scheme Fig. 1A). The DNA of two mice per condition, selected for high expression by RT-qPCR of Gadd45b or mCherry, respectively, was then subjected to reduced representation bisulfite sequencing (RRBS) for DNA methylation analysis. Bioinformatic analysis of RRBS data detected 809117 CpGs that were common between both conditions. Of those, 76185 individual CpGs were differentially methylated (DMCs) with a q-value smaller than, or equal to, 0.01 and at least 25% difference. Using a window and step size of 1000 bp, differentially methylated regions (DMRs) were defined and 16702 regions passed the defined significance threshold. A majority of DMCs and DMRs were hypomethylated throughout time, located in open sea regions (defined as regions outside CpG islands, CpG shores or CpG shelves) and localized in gene bodies, particularly introns. The Volcano plot in Figure 1D illustrates the DMC methylation pattern at 14d p.i.. Upon Gadd45b overexpression, 46508 DMCs (61.05 %) were hypomethylated and 29677 DMCs (38.95%) were hypermethylated. The percentage of hypo- and hypermethylated regions per chromosome was similar (Suppl. Fig. 1G). We then interrogated DMCs in relation to their distance to a CpG island and found 69.73% to be overlapping with open sea regions (Fig. 1E), defined as regions outside of a known CpG island, CpG shore (2000 bp flanking the CpG island) or CpG shelves (2000 bp flanking the CpG shores). Analysis of the genomic context revealed that 41.36% of DMCs were located in introns (Fig. 1F), followed by 22.3% in the intergenic space, 23.6% in exons and 12.5% in promoter regions. This analysis indicates early and widespread methylation changes primarily in gene bodies upon Gadd45b overexpression.
In order to understand the long-term changes in methylation patterns induced by Gadd45b overexpression, we extracted DNA from mice 90 days after injection of AAV8-mGadd45b or AAV8-mCherry control. Analysis revealed a similar distribution of hypo- and hypermethylated DMRs at 14d p.i. (Suppl. Fig. 1A) and at 90d p.i. (Suppl. Fig. 1C). The genomic context of DMCs and DMRs, both at 14d p.i. (Suppl. Fig. 1B) and at 90d p.i. (Suppl. Fig. 1D) was very similar as well, the majority of differential methylation concerning open sea regions and gene bodies, particularly introns (summarized in Table 1). The overall numbers of DMRs and DMCs with AAV8-mGadd45b also resembled that at 14d p.i., but common DMRs or DMCs examination revealed an overlap of only 15% of DMRs (2517) and 18% of DMCs (13920) (Fig. 1G). While this indicates that the specific regions with methylation changes induced by Gadd45b overexpression are not stable over time, the general location in open sea regions and in introns of genes of DMRs and DMCs is maintained as a specific and stable footprint of Gadd45b overexpression, which we termed “Gadd45b-regulon”. Furthermore, there was an important overlap of more than half of the genes containing intronic hypomethylated or hypermethylated DMCs at 14d and at 90d (Fig. 1H). The extent of overlap was similar for intronic DMRs (hypoDMRs at 14d: 3937 genes; hypoDMRs at 90d: 3467 genes, overlap (1712 common genes); Suppl. Fig. 1E). Of those common genes, 2353 genes contained at least one intronic hypo- and one intronic hypermethylated DMC (Fig. 1H) and 447 genes at least one intronic hypo- and one intronic hypermethylated DMR at both 14d and 90d p.i. (Suppl. Fig. 1E). Thus, Gadd45b overexpression induces stable changes in methylation patterns in gene bodies, particularly introns.
The results of the RRBS analysis prompted us to identify functional categories associated with this GADD45B-induced methylation footprint corresponding to the 2353 genes stably containing DMCs over time (Fig. 1H). We used Gene ontology analysis (PANTHER version 15.0; http://pantherdb.org) and the PANTHER overrepresentation test with the GO-Slim annotation data set ‘biological process’. 145 GO categories were significantly overrepresented with a fold change >2 and an FDR <0.05. The first fifteen significantly overrepresented GO categories are displayed in Figure 1I. Of those, 10 GO categories are explicitly related to neuronal functions with two prevailing categories, namely synaptic function and organization, and neurodevelopment and neurogenesis. This is very similar to what we found for DMRs (Suppl. Fig. 1F) and suggests that Gadd45b is involved in the specific regulation of gene body methylation of neuron-related genes.
Gadd45b overexpression leads to heterochromatin disorganization in mDA neurons
Members of GADD45 protein family have been described to promote heterochromatin relaxation 31. We therefore examined whether Gadd45b overexpression, in addition to global methylation changes, would also alter chromatin organization, in particular the organization of heterochromatin. To do so, we stained mDA neurons for histone H3 lysine 9 trimethylation (H3K9me3), a repressive heterochromatin mark. Immunostaining for H3K9me3 shows a perinucleolar pattern composed, on average, of 3 or 4 foci (3.64±0.12) in TH+ neurons in the SNpc of AAV8-mCherry injected mice (Fig. 2A). This pattern becomes disorganized in AAV8-mGadd45b injected mice at 14d p.i.. Semi-automated quantification of H3K9me3 staining specifically in TH+ neurons identified a 1.13-fold increase in the number of H3K9me3 foci (4.10±0.12) scattered across the nucleus, an increase by 1.24-fold of the average H3K9me3 foci volume (4.78±0.15; 5.91±0.22 μm3), a reduction by 1.21 fold in the intensity of the diffuse nucleoplasmic H3K9me3 staining (2.37×107±557961; 1.96×107±441292), but no difference in foci intensity (1.32×106±54236; 1.41×106±64942) (Fig. 2A, quantification in Fig. 2B-E). This shift in heterochromatin organization is still detectable at 90d p.i. (Fig. 2F-J). These results show that a de-structuration of heterochromatin is already detectable 14d and stable up to 90d after the injection of AAV8-mGadd45b indicating an early and stable perturbation of global heterochromatin organization upon Gadd45b overexpression.
We next examined whether Gadd45b overexpression leads to perturbations in the pattern of the DNA methylation marker MeCP2. Immunostaining of sections from mice injected with AAV8-mGadd45b or AAV8-mCherry did not show any difference in the number of MeCP2 foci in TH+ neurons (Suppl. Fig. 2A-D), the intensity of the diffuse nucleoplasmic MeCP2 staining nor the volume and intensity of MeCP2 foci, neither at 14d nor at 90d p.i. (Suppl. Fig. 2E,G-J). There was a slight increase in foci intensity at 14d p.i. with AAV8-mGadd45b (Suppl. Fig. 2F). This analysis indicates that DNA methylation perturbation as detected by RRBS does not imply a global change in the organization of MeCP2 distribution.
Gadd45b overexpression leads to the loss of TH+ neurons
To examine the effect of Gadd45b overexpression on the survival of mDA neurons in the SNpc, we quantified the number of TH+ neurons in mice injected with AAV8-mCherry or AAV8-mGadd45b at 14d and 90d p.i. (Fig. 3A). There was no difference between TH+ cell numbers comparing AAV8-mCherry- and AAV8-mGadd45b-injected (ipsilateral) sides to the non-injected side (contralateral) at 14d p.i. (Fig. 3B: AAV8-mCherry ipsi/contra: 1.04±0.02; AAV8-mGadd45b ipsi/contra: 0.97±0.04). However, 90d after injection of AAV8-mGadd45b, this ratio shifted to an average of 0.82±0.04, indicating a specific loss of 18% of TH+ neurons on the AAV8-mGadd45b injected side compared to the contralateral side (Fig. 3C). The TH+ neurons ratio between the ipsi- and the contralateral side remained unchanged in AAV8-mCherry injected mice (Fig. 3C: ipsi/contra: 0.97±0.02). These results show that Gadd45b overexpression in the SNpc can trigger degeneration of mDA neurons in the long-term. In addition, the absence of a significant TH+ cell loss at 14d p.i. suggests that alterations in the distribution of methylation at CpGs and in the organization of chromatin precede neuronal death.
Enhanced vulnerability of mDA neurons overexpressing Gadd45b to oxidative stress
To examine whether the heterochromatin de-structuration observed in TH+ neurons at 14d p.i. upon Gadd45b overexpression could render these neurons more vulnerable to oxidative stress, mice were first injected with AAV8-mCherry or AAV8-mGadd45b and 14d later with 6-hydroxy-dopamine (6-OHDA, 2 μl; 0.5 μg/μl) in the ipsilateral striatum. When injected into the striatum, 6-OHDA induces a specific and retrograde death of mDA neurons in the SNpc and is frequently used to model PD in rodents 32 including mice 33. Three days after the unilateral, striatal injection of 6-OHDA, immunostainings of striatal sections show a 29% loss (mCherry ipsi/contra: 0.71±0.04) of TH intensity in the ipsilateral striatum compared to the contralateral side in mCherry expressing mice (illustrated in Fig. 3D, quantified in Fig. 3E). This decrease in TH staining intensity reaches 44% (mGadd45b ipsi/contra: 0.56±0.04) in AAV8-mGadd45b injected mice (illustrated in Fig. 3D, quantified in Fig. 3E), suggesting that Gadd45b overexpression increases the axonal degeneration of mDA neurons induced by 6-OHDA. However, this experimental paradigm did not lead to any significant loss of TH cell bodies in the SNpc (Fig. 3F), potentially due to the time of analysis. We analyzed mice 3 days rather than 6 days after 6-OHDA injection, the time point normally used to induce mDA cell death in the SNpc 32, to identify early events in TH+ cell bodies. An increased vulnerability to oxidative stress of mDA neurons overexpressing Gadd45b was also reflected at the heterochromatin level. The injection of 6-OHDA changed the nuclear localization and increased the number of H3K9me3 positive foci (Fig. 3G, H) compared to AAV8-mCherry. Under these conditions, the number of MeCP2 positive foci in mDA neurons expressing AAV8-mGadd45b was also increased as compared to AAV8-mCherry (Fig. 3I, J). This presumably reflects more significant DNA methylation changes upon Gadd45b overexpression under oxidative stress.
Chromatin de-structuration is accompanied by increased DNA damage
The heterochromatin loss model of aging stipulates that heterochromatin de-condensation is a driving force of cellular aging 34. Loss of proteins involved in heterochromatin maintenance has been shown to lead to increased DNA damage and accelerated aging 35, 36. In the context of NDs, recent studies have reported that chromatin relaxation in the brain could also result in increased DNA damage and genome instability 18, 37–39. Since Gadd45b overexpression led to chromatin changes, we investigated whether it might also induce DNA damage. We therefore performed immunostainings for phosphorylated histone H2AX (γ-H2AX), a marker for DNA strand breaks (Fig. 4A). TH+ neurons contained either a single perinucleolar focus or a diffuse nucleoplasmic staining without any foci. After injection of AAV8-mGadd45b (14d p.i.), the majority (65.94±2.10 %) of TH+ neurons displayed a diffuse, intense nucleoplasmic staining against only 39.81±2.40 % of TH+ neurons after AAV8-mCherry injection, the majority showing one or more prominent perinucleolar γ-H2AX-positive foci (Fig. 4B, C). The quantification of the diffuse γ-H2AX staining in the nucleus revealed a more intense staining after Gadd45b overexpression, indicating widespread DNA damage.
LINE-1 methylation is affected by Gadd45b overexpression
Heterochromatin alterations can unsilence normally repressed TEs, including LINE-1 elements 14. LINE-1 are a potential source of DNA damage 16-18, 40. Repeat elements are detected by RRBS and the Bismark software used for mapping of the RRBS reads only considers uniquely mapped reads to avoid any bias during the methylation calling. Thus, reads that are mapping with a same mapping score to multiple locations on the reference genome, which will be the case for most reads derived from repetitive elements, will not be considered. We exploited these facts to explore the methylation status of LINE-1 elements in our experimental conditions. Using the RRBS data, we used the LINE-1 annotation included in the software HOMER, which contains its own database of LINE-1 elements for the mouse genome to interrogate the overlap of DMRs and DMCs with LINE-1 elements 14 days after injection of AAV8 viruses. This analysis showed that 3530 DMCs and 1030 DMRs overlapped with an annotated LINE-1 element, which we termed L1-DMCs and L1-DMRs respectively. 794 (22.5%) L1-DMCs (Fig. 5A) and 264 (25.6%) L1-DMRs (Suppl. Fig. 3A) were located in intronic regions, and 2734 (77.5%) L1-DMCs (Fig. 5A) and 766 L1-DMRs (Suppl. Fig. 3A) in intergenic regions. Of all the L1-DMCs the majority was hypomethylated (1933, 54,8%, Fig. 5B), similarly to L1-DMRs (595, 57,8%).
To determine which LINE-1 families containing at least one L1-DMC upon Gadd45b overexpression were overrepresented, we counted the number of families and ordered them by frequency. Figure 5C shows the ten most represented LINE-1 families with a L1-DMC. Interestingly, of 94 different families present, three LINE-1 family members, namely L1-Md-F2 (975 L1-DMCs), L1Md-T (349 L1-DMCs) and L1Md-A (280 L1-DMCs), were the three most frequently represented. L1Md-T, L1Md-A and L1-Md-F2 elements are young LINE-1 elements 41, which contain full-length, retrotransposition-competent LINE-1 copies. Of LINE-1 associated intronic DMCs (L1-iDMCs, 794) at 14d p.i., the majority (54.7%) was hypomethylated (434 hypomethylated vs 360 hypermethylated, Fig. 5D). L1-iDMCs were mostly located in protein-coding genes (82.8%, Fig. 5E) and the GO analysis of biological processes of 483 genes, containing at least one L1-iDMC, identified seven enriched categories. Among those categories, three were neuron-related (Fig. 5G). The most frequent LINE-1 families in intronic L1-iDMCs were, again, the young L1Md-F2 family (139 of 794) and L1Md-T (57 L1-iDMCs) followed by the old Lx8 family (57 L1-iDMCs) (Fig. 5F). Notably, other young LINE-1 families were also overlapping with iDMCs (L1Md-A: 27 L1-iDMCs, L1Md-Gf: 6 L1-iDMCs, not shown) (Fig. 5F). Similarly to L1-iDMCs, the majority of LINE-1 associated DMRs in introns (L1-iDMRs) was hypomethylated (63,3 %, 167 L1-iDMRs; Suppl. Fig. 3A) and belonged to the L1Md-F2 family (39 of 264) (Suppl. Fig. 3B). Hypo- and hypermethylated iDMRs were also found in LINE-1 elements of the active L1Md-T (3 and 2, respectively) and L1Md-A families (3 and 3 elements, respectively). This prompted to examine a data-base annotating full-length LINE-1 elements (L1Basev2 42) to see if intronic and intergenic DMRs and DMCs could coincide with possibly active LINE-1. Ten DMRs and 62 DMCs overlapped with a full-length LINE-1. More than half of them (7 out of 10 DMRs and 38 out of 62 DMCs, Fig. 5H, Suppl. Fig. 3C), were hypomethylated. This data indicates a widespread change in the methylation status of LINE-1 elements upon Gadd45b overexpression (14d p.i.). The most differentially methylated LINE-1 elements are hypomethylated and belong to young L1 families (L1Md-F2, L1Md-T, L1Md-A) suggesting a possible expression of these individual LINE-1 loci. Interestingly, L1-iDMCs are frequently located in introns of genes related to neuronal functions (Fig. 5G).
LINE-1 transcripts are increased upon overexpression of Gadd45b
Having established a change in the methylation pattern of LINE-1 elements after injection of AAV8-mGadd45b, some of which were full-length, we evaluated the expression of the youngest L1 families in mice, namely L1Md-A and L1Md-Tf/Gf. The analysis by RT-qPCR with specific primers located in the 5’UTR of the L1Md-A (Fig. 5I, left panel) and L1Md-Gf/Tf families (Fig. 5I, right panel) showed a 1.5-fold increase in LINE-1 transcripts in the SNpc after 14d of Gadd45b overexpression.
Gadd45b overexpression is accompanied by expression changes in genes with DMCs
Expression levels of candidate genes with intronic DMCs at either 14d or 90d p.i. were then analyzed by RT-qPCR. We selected 13 genes based on their known function either in chromatin remodeling (Satb1, Setdb1, Wapl), DNA methylation (Tet2, Tet3, Dnmt3a), PD relevance (Lrrk2, Park2), synaptic remodeling (Sorcs2), DNA damage (Xpa), or in aging and senescence (Cdkn2a-p19, Cdkn2d, Sirt1). None of the candidate genes showed a change in expression at 14d p.i. However, at 90d p.i., the expression of Satb1, Setdb1, Dnmt3a, Tet3 and Park2 decreased significantly in the ipsilateral SNpc injected with AAV8-mGadd45b compared to AAV8-mCherry. (Fig. 6A). Of note, 4 out of these 5 are genes belonging to the Gadd45b-DMC-regulon (Fig. 1H). Using RNA-seq data of laser-microdissected SNpc from wild-type mice (GEO GSE72321;30 and unpublished), we compared the expression levels of the selected candidate genes to those of dopaminergic neuron-specific markers such as TH, dopamine transporter Slc6a3 or the homeogene, Engrailed 1 (En1)43. We found high levels of expression of Satb1 compared to the other selected genes, indicating that Satb1 is strongly expressed in the SNpc (Suppl. Fig. 4). Exonic and intronic DMCs in downregulated genes comprised hypo-as well as hypermethylated CpGs. Disruption of the normal gene body methylation state of several genes of the GADD45B-DMC-regulon might thus provoke their dysregulation over time but is not associated with a loss of a particular methylation pattern.
Discussion
Epigenetic alterations and chromatin relaxation are hallmarks of aging 2, 3, but evidence of age- and disease-related global changes in the epigenetic landscape of different neuronal populations is still scarce. This information, however, holds important promises for understanding the implication of aging in the neuronal degeneration characterizing age-related NDs and might thus foster the understanding of the pathogenesis of NDs.
Here, we show that Gadd45b overexpression induces early changes in DNA methylation, particularly in introns of genes related to neuronal functions and on young and potentially active LINE-1 elements. This, accompanied by perturbations of heterochromatin organization and increased DNA damage, culminate in neuronal cell death after several weeks. At an early time point, before the onset of neurodegeneration, Gadd45b overexpression induces a vulnerable state in mDA neurons, increasing their sensitivity to oxidative stress induced by the striatal injection of 6-OHDA. This vulnerability is characterized by an amplified striatal dopaminergic axon terminal loss and an accentuation of perturbations in the organization of heterochromatin and DNA methylation in mDA cell bodies of the SNpc. We also demonstrate that LINE-1 transcripts are increased early on after Gadd45b overexpression. This increase in LINE-1, which are potent inducers of DNA damage in mDA neurons 18, could explain the cell death of mDA neurons we observe when overexpressing Gadd45b long-term.
A number of recent studies indicate that GADD45 proteins play a key role in active DNA demethylation in post-mitotic neurons in the brain 27, 28 by serving as scaffolding proteins to recruit DNA repair enzymes such as the thymidine DNA glycosylase (TDG) to the site of DNA demethylation 44. So far, GADD45B-regulated DNA demethylation has been described in the context of adult neurogenesis 29, depressive-like behavior in mice 45, major psychosis in humans 46 and cerebral cortex plasticity 47 on specific promoters, mostly on promoters of the Bdnf gene. Unexpectedly, when overexpressing Gadd45b in the SNpc of wild-type mice, we observed very limited changes in methylation on gene promoters, but rather widespread methylation changes on gene bodies with both, hypermethylated and hypomethylated CpGs. This apparent discrepancy with previous studies in terms of extent, localization and pattern of the DNA methylation changes induced by GADD45B might be due to methodology. Indeed, most studies in the brain did not use techniques allowing for unbiased global DNA methylation surveillance.
GADD45B has been shown to associate with TET (Ten-eleven translocation) proteins 48, 49 which transform 5-methylcytosines (5mC) through a series of sequential oxidations. Those modified cytosines are excised by TDG and then replaced by non-methylated cytosines through BER-dependent mechanisms 50. A recent study has shown that Tet2 expression is increased in PD patients leading to altered 5mC patterns in enhancers of neuronal genes. Conversely, TET2 loss in mDA neurons was neuroprotective 51. Tet2 did not show any changes in expression at either time point after AAV8-mGadd45b injection. This might be explained by the fact that Tet2 has many splice variants, which were not all covered in our RT-qPCR assay. However, Tet3 is down-regulated 90d after AAV8-mGadd45b injection. TET3 has been shown to interact with transcriptional regulators and histone writers such as H3K36 methyltransferases and to allow the active transcription of certain neuronal genes 52. TET3 has also been reported to bind DNA and prevent aberrant methylation at the transcription start site of genes involved in lysosomal functions, mRNA processing and the BER pathway, pointing to a possible relevance of TET3 in the pathogenesis of NDs 53. While our knowledge of how genome-wide DNA methylation patterns or epigenetic changes are correlated with PD pathogenesis is still scarce, there is evidence supporting that methylation changes in the SNpc correlate with aging 23 and with cognitive impairment in PD 54. Most of the existing data concerns methylation patterns of specific genes associated with NDs. Expression of the SNCA gene, encoding α-synuclein and mutated in familial PD, for example, is under the control of DNA methylation 55,56. Chromatin modifications on specific genes have also been reported, and the expression of several genes mutated in PD are modulated by histone modifications, including SNCA and MAPT encoding Tau 25. Differentially methylated enhancers have been reported in AD patients, and a dysregulation of histone acetylation in both, AD and PD patients 57–60.
By serving as an adaptor between repair factors and chromatin, GADD45B can be seen as a communicating platform between DNA repair and epigenetics 28, 61. This could provide an explanation to the fact that we observe global changes in heterochromatin organization in addition to changes in DNA methylation. In line with this, GADD45 proteins have been reported to induce heterochromatin relaxation during cellular reprogramming 31. GADD45B might thus be part of protein coordinators which link DNA methylation and histone modifications 60. Perturbations of DNA methylation and global changes in heterochromatin organization, as we show here, induce neurodegeneration of mDA neurons in the SNpc with time and might thus be primary drivers of neurodegeneration.
Another primary hallmark of aging is genomic instability and some data suggest that this is also true in neuronal aging 22. In this context it is important to note, that DNA strand breaks are physiologically occurring in post-mitotic neurons. This physiological process, however, needs to be tightly regulated, since the induction of DNA strand breaks is pathologically exacerbated in AD 62. We have shown earlier, that DNA damage in mDA neurons can be induced by acute and chronic oxidative stress 18, 30 and that LINE-1 activity participates in DNA damage 18. In the latter study, DNA damage was prevented either by siRNA against LINE-1 ORF2, the LINE-1 repressive protein Piwil1 or a nucleoside analogue reverse transcriptase inhibitor in vitro 18. This LINE-1 induced DNA damage is dependent on young and active LINE-1 copies. These young LINE-1 elements belong mainly to the L1Md-Tf/Gf, A and F families. In this context it is interesting to note that we observe the most pronounced changes in DNA methylation upon Gadd45b overexpression on these young LINE-1 families (Fig. 5). Intronic DMCs and DMRs, intronic L1-associated DMCs (L1-iDMCs), and DMCs or DMRs overlapping with full-length LINE-1 (flL1-iDMCs/flL1-iDMRs), are majorly hypomethylated and iDMCs are preferentially located in genes related to neuronal functions. It is therefore conceivable that the changes in methylation patterns we observe upon Gadd45b overexpression, particularly on young LINE-1 elements, are functionally linked to the increase in LINE-1 transcripts we observe 14 days after injection of AAV8-mGadd45b. Furthermore, based on previous evidence summarized above, this increase in expression of young L1 elements of the L1Md-Tf/Gf, A and F families might be at the origin of the DNA damage in mDA neurons upon Gadd45b overexpression.
Several lines of evidence suggest that the activation of TEs might be associated with aging and NDs. The expression of LINE-1 in wild-type mice increases in neurons 63, liver and muscle during aging 64. The activation of TEs with aging leads to neuronal decline and shorter lifespan in drosophila 65, 66 and increased TE expression has been reported in brain tissue from PD, AD and ALS patients 37, 67. Recently, elevated transcripts of repetitive sequences have been found in the blood of PD patients 68. Heterochromatin de-structuration and increased TE activity lead to an AD-like phenotype in a mouse model with targeted disruption of Bmi1, a gene involved in heterochromatin maintenance and altered expression in AD patients 39. In AD and in ALS, the Tau protein as well as TDP-43 can induce heterochromatin relaxation, especially at the level of LINE-1 elements in the case of TDP-43, leading to increased TE activity and neurotoxicity 37,67,69. An aging-induced activation of TEs, which are intrinsic components of the genomes of virtually all eukaryotes, might link genomic instability and epigenetic changes to the aging process.
Changes in chromatin states lead to changes in gene expression as exemplified during the transition from neuronal progenitors to adult neurons 70 and during aging 71, and recent data suggests that this could also be the case in NDs 59, 72. In several NDs (AD and HD 73, 74; ALS 75), disease-specific gene expression changes are increasingly recognized, but whether they overlap with expression profiles characteristic to aging is not known yet. In AD patients, widespread loss of heterochromatin was accompanied by a transcriptomic profile resembling the one of a fetal brain 38, suggesting that the relaxation of chromatin allows the expression of normally repressed genes which alters several biological processes and leads to neurodegeneration 76. This is also in line with our data showing the preferential location of iDMCs and iDMRs in genes involved in neurogenesis but it remains to be seen whether their expression is altered. Interestingly, it has been shown that Gadd45b activity promotes adult neurogenesis 29. The majority of these genes with iDMCs and iDMRs upon overexpression of Gadd45b belonged to neuronal categories, particularly synapse-related and neurodevelopmental categories. In this context it is interesting to note that synaptic homeostasis is an emerging key player in the pathogenesis of PD 77 and synaptic dysfunction is an early event in neurodegeneration 78. Gadd45b overexpression induced DMCs and DMRs preferentially in introns of genes. The role of gene body or intronic methylation on gene expression is not completely understood 79. In recent years, several studies have reported that gene body methylation influences gene expression levels and/or alternative splicing 80,81,82. Among the list of genes underlying a direct or indirect regulation through methylation changes triggered by Gadd45b overexpression, the decrease in expression did not correlate with a specific direction of methylation changes towards hypo- or hypermethylated CpGs but rather with a change in the methylation state throughout the gene body. Satb1 has been described as a dopaminergic-specific regulator of senescence 83, a dopaminergic neuron cell survival factor 84 and a regulator of global chromatin structuration 85, 86. The decline in Satb1 expression at 90d p.i. might help to explain, at least partly, the neurodegeneration of mDA neurons upon Gadd45b overexpression. Setdb1 is a histone-lysine-methyltransferase that specifically trimethylates lysine-9 of histone H3. It is tempting to speculate that the decrease in expression might relate to the change in the organization of H3K9me3 we observe upon Gadd45b overexpression. Dnmt3a, a genome-wide de novo DNA methyltrasferase, and Tet3, are involved in DNA methylation or demethylation, respectively. Interestingly, inactivation of TET and/or DNMT proteins causes gains and losses of DNA methylation, suggesting that the loss of one regulator can lead to the redistribution of other regulators and of DNA modifications 87. Park2, encoding the ubiquitin protein ligase Parkin, is a PD-related gene and loss-of-function mutations in this gene are responsible for familial forms of PD. It is thus possible, that a decrease in the expression of Park2 in the context of Gadd45b overexpression might contribute to the degeneration of mDA neurons. Overall, the expression changes might be due to the Gadd45b-induced epigenetic dysregulation of the neuronal genome and participate in the Gadd45b overexpression phenotype and the challenge will be to correlate one with the other.
Altogether, our data are in line with an emerging concept on a new pathogenic pathway initiating age-related neurodegeneration. Recent evidence, including from our group, suggests that aging-induced chromatin reorganization triggers the activation of LINE-1 retrotransposons and subsequent LINE-1 induced DNA damage cumulating in neuronal cell death. Our group has shown that acute and chronic oxidative stress leads to heterochromatin relaxation and LINE-1 activation in mDA neurons in vivo 18, 30. Aging-induced epigenetic alterations might produce a vulnerable pre-ND state. Combining this pre-ND state with a particular genetic susceptibility, a familial gene mutation or an accelerating environmental trigger, could initiate a cascade of secondary events including protein aggregation, metabolic dysregulation and mitochondrial dysfunctions. NDs share several common pathological features and despite extensive investigation, no disease-modifying treatment is available. Acknowledging aging as a vulnerability factor for neurodegeneration is important, not only for understanding the pathogenesis of NDs, but also for modeling, testing and developing therapeutics for crucially lacking disease-modifying treatments. Our study suggests two novel therapeutic targets for neuroprotection. Drugs restoring chromatin structure and/or repressing LINE-1 transcription or activity might hold promise for the prevention of age-related neurodegeneration.
Materials and Methods
Animals
All animal treatments followed the guidelines for the care and use of laboratory animals (US National Institutes of Health), the European Directive number 2010/68/UE (EEC Council for Animal Protection in Experimental Research and Other Scientific Utilization). This project was validated by the competent ethical committee (CEA 59) and authorized by the Minister of Higher Education, Research and Innovation (n° 00703.01 and n° APAFIS#6605-2016090209434306 v3). For surgical procedures, animals were anesthetized with Xylazine (Rompun 2%, 5 mg/kg) and Ketamine (Imalgene 1000, 80 mg/kg) by intraperitoneal injection and a local subcutaneous injection of lidocaine (0.5%, 3mg/kg) on the incision site. Post-chirurgical analgesia was insured by an injection of the analgesic Meloxicam (Metacam, 0,5mg/kg) s.c.. Swiss OF1 wild-type mice (Janvier) were maintained under a 12 h day/night cycle with ad libitum access to food and water. A maximum of 5 mice were housed in one cage, and cotton material was provided for mice to build a nest. Experimental groups consisted of five to seven male mice of 6 weeks of age. Sample size calculations were based on previous experiments.
AAV8 vectors to overexpress Gadd45b
Forced expression of Gadd45b in mDA neurons was achieved using an AAV8 viral vector. The constructs contained cDNAs for either mouse Gadd45b (AAV8-mGadd45b) or mCherry (AAV8-mCherry) under the control of the ubiquitous EF1a promotor. Gadd45b cDNA was flanked by the cognate 5’ and 3’ UTRs (Fig. 1A). AAV8 was chosen because it has previously been shown to efficiently infect mDA neurons in the midbrain.
Brain injections
For injections, mice were placed in a stereotaxic instrument, and a burr hole was drilled into the skull 3.3 mm caudal and 1.3 mm lateral to the Bregma. The needle was lowered 3.8 mm from the surface of the skull, and AAV8-Ef1a-mCherry or AAV8-Ef1a-mGadd45b (Vector Biolabs; 2 μl; 4.8×1013 GC/ml suspended in NaCl 0,9% with 5% glycerol) injections were performed over 4 min. Where indicated 6-OHDA (2 μl; 0.5 gg/gl; Sigma) injections were performed in the same manner 0.4 mm rostral, 1.8 mm lateral and 3.8 mm ventral to the bregma, over 4 min.
Tissue dissection
For RNA and DNA analyses, biopsies of the SNpc were performed. Brains were put into a custom-made brain slicer for adult mice brain. A coronal slice of ≈2mm encompassing the SNpc was excised (Bregma −3.26 mm to −5.2mm) and placed on a cold cover slide with the rostral side facing the experimenter. Dissection of the SNpc was then done following anatomical landmarks: a sagittal cut to separate the two hemisphere, a second parasagittal cut through the fasciculus retroflexus and the mammillothalamic tract (about 2/3 starting from the midline of the distance between the midline and the rostral end of cerebral peduncle) to remove the VTA, a transversal section from the ventral part of the lateral geniculate complex to the midline, a second transversal cut from the ventral end of the cerebral peduncle to the midline. The cerebral cortex was then removed to only keep the midbrain part containing the SNpc and immediately frozen on dry ice and kept at −80°C until extraction.
RT-qPCR
Total RNA from dissected SNpc was extracted using the AllPrep DNA/RNA Micro Kit (Qiagen 80284) adding an on-column DNase I treatment (Qiagen 79256), followed by RT-qPCR. RNA (200 ng) was reverse-transcribed using the QuantiTect Reverse Transcription kit (Qiagen 205313). Quantitative PCR reactions were carried out in duplicates with SYBR Green I Master Mix (Roche S-7563) on a LightCycler 480 system (Roche Applied Science). The following primers were used: Hprt (sense: AGCAGGTGTTCTAGTCCTGTGG, antisense: ACGCAGCAACTGACATTTCTAA); LINE-1 A (sense: TTCTGCCAGGAGTCTGGTTC, antisense: TGAGCAGACCTGGAGGGTAG); LINE-1 Tf/Gf (sense: CTGGGAACTGCCAAAGCAAC, antisense: CCTCCGTTTACCTTTCGCCA); Gadd45b (sense: ACTGATGAATGTGGACCCCG, antisense: CCTCTGCATGCCTGATACCC); Satb1 (sense: TCTTTTACCCCCTCCTCCCA, antisense: TCACCTGCCAGAACACTTCA); Tet3 (sense: CTCGGCGGGGATAATGGGAG, antisense: AGCCTGTCTTGACAGTCGCC); Dnmt3a (sense: GCCGAATTGTGTCTTGGTGGATGACA, antisense: CCTGGTGGAATGCACTGCAGAAGGA), Setdb1 (sense: GTTTGCCTGGGTTTGGCAAG, antisense: CTTTGGCCCTCAGTCCGTC); Park2 (sense: GCTCAAGGAAGTGGTTGCTAAG, antisense: CAATACTCTGTTGTTCCAGGTCA). Primer efficiencies were tested using 10-fold dilution series of cDNA spanning at least three orders of magnitude. Data were analyzed using the ddCt method and values normalized to hypoxanthine-guanine phosphoribosyl transferase (Hprt).
DNA extraction and quantification
DNA was also extracted during the same process of RNA extraction using the AllPrep DNA/RNA Micro Kit (Qiagen 80284). The DNA was then purified, treated with RNase H (ThermoFischer, 18021071) and Proteinase K (PCR grade, Roche, 3115836001) and concentrated with the DNA Clean & Concentrator-5 kit (Zymo, D4013). DNA concentration of each sample was measured using the Qubit® fluorometer with dsDNA BR Assay Kit (Thermo Fisher Scientific).
RRBS
RRBS was performed by Diagenode. DNA quality of the samples was assessed with the Fragment AnalyzerTM and the DNF-488 High Sensitivity genomic DNA Analysis Kit (Agilent). DNA was slightly more fragmented than defined by the quality control standards but this fragmentation was minor. RRBS libraries were prepared using the Premium Reduced Representation Bisulfite Sequencing (RRBS) Kit (Diagenode) which uses the Mspl restriction enzyme and size selection to enrich for CpG-rich regions (coverage of about 4 million CpGs). 100ng of genomic DNA were used to start library preparation for each sample. Following library preparation, samples were pooled together by 8. PCR clean-up after the final library amplification was performed using a 1.45x beads:sample ratio of Agencourt® AMPure® XP (Beckman Coulter). DNA concentration of the pools was measured using the Qubit® dsDNA HS Assay Kit (Thermo Fisher Scientific). The profile of the pools was checked using the High Sensitivity DNA chip for 2100 Bioanalyzer (Agilent). RRBS library pools were sequenced on a HiSeq3000 (Illumina) using 50 bp single-read sequencing (SR50). Bisulfite conversion and amplification were performed using Diagenode’s Premium RRBS Kit. After conversion, the pooled samples were analyzed by qPCR. Sequencing was performed in single-end mode, generating 50 bases reads (SE50) on an Illumina HiSeq 3000/4000. Quality control of sequencing reads was performed using FastQC version 0.11.8 (https://www.bioinformatics.babraham.ac.uk/projects/fastqc/). Adapter removal was performed using Trim Galore (https://www.bioinformatics.babraham.ac.uk/projects/trim_galore/) version 0.4.1. Reads were then aligned to the murine reference genome mm10/GRCm38 using bismark v0.20.0 88. Bismark is a specialized tool for mapping bisulfite-treated reads such as the ones generated in RRBS-seq experiments. Bismark requires that the referenced genome first undergoes an in-silico bisulfite conversion while transforming the genome into forward (C → T) and reverse strand (G → A) versions. The reads producing a unique best hit to one of the bisulfite genomes were then compared to the unconverted genome to identify cytosine contexts (CpG, CHG or CHH - where H is A, C or T). The cytosine2coverage and bismark_methylation_extractor modules of Bismark were used to infer the methylation state of all cytosines (for every single uniquely mappable read) and their context, and to compute the percentage methylation. The reported cytosines were filtered to get only the CpGs covered in each sample. The spike-in control sequences were used at this step to check the bisulfite conversion rates and to validate the efficiency of the bisulfite treatment. Methylkit v1.7.0 89, a R/Bioconductor package, was used to perform the differential methylation analysis between the two groups of samples. The CpG data set was filtered for low coverage (CpGs with coverage less than 10X in all samples per comparative group were discarded) and for extremely high coverage to discard reads with possible PCR bias (CpGs with coverage higher than the 99.9th percentile were discarded). The data was then normalized for read coverage distribution between samples. A pairwise comparison was performed for first versus second group of samples to identify differentially methylated CpGs (DMCs) and differentially methylated regions (DMRs), the latter with a window and step size of 1000bp. Methylkit uses logistic regression to compare the methylation percentages between groups at a given CpG/region. All DMCs and DMRs were annotated with the R/Bioconductor package annotation 90, with the refGene and CpG island annotations from UCSC. The annotation comprised two categories: (i) distance to a CpG island and (ii) regional annotation. The distance related annotation classified DMCs and DMRs whether they overlapped a known CpG island, 2000 bp of the flanking regions of the CpG islands (shores), 2000 bp of the flanking regions of the shores (shelves) or outside these regions (open sea). The regional analysis classified DMCs and DMRs in four groups, namely, exons, introns, promoters and intergenic regions.
Immunostaining
For immunostaining, animals received a lethal intraperitoneal injection of 1 μl/g body weight dose of Euthasol (150mg/kg) and were then perfused with 8 mL of Phosphate Buffer Saline (PBS) then 8 mL of 4% Paraformaldehyde (PFA) at a rate of 300 ml/h using a syringe pump. Brains were then post-fixated 1 hour at room temperature (RT) in 4% PFA, washed in PBS three times for 30 minutes at RT and placed in PBS with 20% sucrose overnight at 4°C. After cryoprotection, the brains were embedded in Tissue Freezing Medium (TFM, Microm Microtech), frozen on dry ice and 30 μm sections of mouse striatum and ventral midbrains encompassing the SNpc were prepared using an HM 560 Microm cryostat (Thermo Scientific).
Slides with 30 μm striatum or midbrain sections were washed in PBS and permeabilized with 1% Triton X-100. After 30 minutes at RT in 100 μM glycine buffer (for TH/mCherry and TH/ORF1p) or 30 minutes at 100°C in demasking citrate buffer (10 mM, pH 6, 0.05% Tween) (for TH/MeCP2, H3K9 or γ-H2AX), sections were first blocked in 10% Fetal Bovine Serum (FBS, Gibco) in the presence of 0.5% Triton X-100 for 1 hour at RT and incubated with primary antibodies overnight at 4°C, washed and further incubated with secondary antibodies for 1 hour at RT. The following primary antibodies used: anti-γ-H2AX (mouse, 1/200, Millipore, clone JBW301), anti-TH (chicken, 1/500, Abcam, ab76442), anti-ORF1p (guinea pig, 1/200, in-house, clone 09 as in 83, anti-mCherry (mouse, 1/200, Clontech 632543), rabbit anti-H3K9me3 (rabbit, 1/200, Abcam, ab8898) and anti-MeCP2 (rabbit, 1/200, Millipore MABE328). Sections were incubated with appropriate secondary antibodies (488 anti-chicken, 546 anti-mouse, 647 anti-guinea pig, 647 anti-rabbit, 647 anti-mouse, Alexa Fluor, Life Technologies)
In situ hybridization
Mice were perfused with PBS in RNase-free conditions, and frozen in isopentane (embedded in TissueTek O.C.T). Brain slices (20 μm) were fixed in 4% PFA in PBS for 10 min at RT and then permeabilized twice for 10 min in RIPA buffer (150 mM NaCl, 1% NP-40, 0.5% Na deoxycholate, 0.1% SDS, 1 mM EDTA, 50 mM Tris–HCl pH 8). Brain sections were fixed again for 5 min in 4% PFA, acetylated for 10 min in 0.25% acetic anhydride (in 0. 1 M triethanolamine, pH 8). Sections were then permeabilized for 30 min in PBS with 1% Triton X-100, 10 min in 10mM citrate buffer pH6 at a 100°C and then pre-incubated for 1 h at 70°C in hybridization buffer (50% formamide, 5× SSC, 5× Denhardt (1% Ficoll, 1% SSC, 1% Tween-20), 500 μg/ml Salmon sperm DNA, 250 μg/ml yeast tRNA). Sections were then hybridized overnight at 70 °C with a digoxigenin (DIG)-labeled RNA probes (DIG RNA labeling kit, Roche 11277073910) for Gadd45b mRNA. Sections were washed with FAM/SSC (50% formamide, 2× SSC, 0.1% Tween-20) twice 30 min at 37°C, then twice in 0.2× SCC at 42°C, blocked in B1 buffer (100 mM maleic acid pH 7.5, 150 mM NaCl) with 10% fetal bovine serum (FBS) for 1 h, and incubated overnight at 4°C in B1 buffer with alkaline phosphatase-conjugated anti-DIG (1/2000; Roche 11633716001). After three washes in B1 buffer and one wash in B3 buffer (100 mM Tris–HCl pH 9, 50 mM MgCl2, 100 mM NaCl, 0.1% Tween-20) for 15 min, slides were stained using the NBT/ BCIP kit (Vector labs, SK5400), stopped with ddH20 and slides were mounted with DAKO-mounting medium.
Imaging/ Microscopy
All large field images used for TH+ neuron quantification, level of viral infection quantification and striatal intensity quantification were made on widefield microscope (Axio zoom V16 – Zeiss – Apotome.2) at 2,3 magnification with a zoom factor of 100. ISH image was taken by upright widefield microscope equipped with a color CCD camera (Nikon 90i microscope) at 20x magnification in brightfield. H3K9me3, MeCP2 and γ-H2AX foci quantification as well as ORF1p intensity quantification were made on images taken by spinning disk microscopy (Yokogawa W1 Spinnning-disk head mounted on an inverted Zeiss AxioObserver Z1) at 63x magnification.
The images in the figures are for illustration purposes and were taken by confocal microscopy (LSM 980 with Airyscan 2, Zeiss) at 63x magnification with a zoom factor of 1.4, except for the MeCP2 and ORF1p images (performed on the spinning-disk microscope).
Cell counting and image quantification
TH cell counting in conditions comparing ipsi- (treated) and contralateral (non-treated) sides were done as follows: For every brain, a minimum of four serial sections were stained, and the number of TH cells was counted in the SNpc of both ipsi- and contralateral sides. An ipsi/contra ratio was calculated for each section, and the resulting mean of four sections was used to quantify the difference between the TH cell number of the ipsi- and contralateral side of the same animal. The counting was done blindly.
The quantification of axonal degeneration in the striatum comparing ipsi- and contralateral sides was done as follows: For every brain, a minimum of seven serial sections were stained, and the integrated density of TH staining intensity was measured in ImageJ by determining the entire contralateral striatum as region of interest (ROI) and conserving area for the measurement of ipsilateral sides. An ipsi/contra ratio was calculated for each section, and the mean ratio of sections containing the striatum was used to quantify the difference between TH striatal intensity of the ipsi- and contralateral side of the same animal. The quantification was done blindly.
Quantifications of foci were performed using a 63× magnification and 0.3 μm - thick successive focal planes except for γ-H2AX foci quantification, which was made using 0.2 μm-thick successive focal planes. Immunostainings of one parameter (H3K9, MeCP2, Etc.) were all done in one experiment for all the conditions. For each immunostaining, 3 images of the SNpc per side were taken on 3 sections per mouse, thus 18 images per mouse (n=3 per condition), or 54 images per condition. The same parameters were set-up on the spinning disk microscope to allow for comparison between experimental conditions for the same staining. Images were analyzed by the same experimenter using ImageJ software 84. For foci quantifications the foci counting Fiji Plug-in was used 85. In addition, an image analysis plug-in was developed for the ImageJ/Fiji software, using Bio-Format (openmicroscopy.org), mcib3D 86 and GDSC (Alex Herbert from Sussex University) libraries. First, nuclei that belonged to TH+ neurons were manually marked with the plug-in Cell Counter, an xml file for each image containing 3D nuclei coordinates was saved. Then, nucleus channel was filtered with a median filter (radius = 4) and a Difference of Gaussian (DOG) (sigma1 = 30, sigma2 = 15), a binary mask was done with an Otsu threshold. Only nuclei that are associated to the nuclei defined in the xml file were kept. MeCP2 and H3K9me3 foci detections were performed using a median filter (radius=2), DOG (sigma1 =10, sigma2=2), binary mask was done with a MaxEntropy threshold, then 3D objects (foci) that had a volume comprised between 1.5 and 40 μm3 and were inside TH+ nuclei or at a 2 μm distance to nucleus was associated to nuclei. γ-H2AX foci detection was performed using a median filter (radius=2), DOG (sigma1=7, sigma2=3), binary mask was done with a Moments threshold, then 3D objects (foci) that had a volume comprised between 0.5 and 100 μm3 were inside TH+ nuclei or at a 2 μm distance to nucleus was associated to nuclei. For each nucleus, foci number, average foci volume, average foci integrated intensity and average nucleus integrated intensity was computed.
Gene ontology analysis
Gene ontology analysis (PANTHER version 15.0; 87) was done using the PANTHER overrepresentation test with the GO-Slim annotation data sets ‘biological process’, ‘molecular function’ and cellular component’ and the ‘mus musculus’ gene set as the reference list. Fisher’s exact test was used to compute statistical significance of overrepresentation with the false discovery rate (FDR) set at p < 0.05. The first 15 categories, ordered by fold enrichment, are displayed along with the corresponding FDR value.
Statistics
Unless otherwise stated, the graphs represent each replicate and the error bar the SEM of the mean of replicates. Error bars, values of n and mean ± SEM are as stated in the figure legends. Results were considered as statistically significant for P-value <0.05; in some cases, the exact P-value is given. Normality test were performed prior to the statistical test and unless stated otherwise, the nonparametric Wilcoxon-Mann-Whitney test was applied. All statistical analysis was done with the software Prism.
For the bioinformatic analysis of the RRBS, we formulated the null hypothesis that there are no differences in methylation between the two groups. After the p-values have been computed, Methylkit, an R package for DNA methylation analysis and annotation, uses the sliding window model (SLIM) to correct P-values to q-values for multiple comparison tests. Statistically significant DMCs and DMRs were identified with a q-value cutoff <0.01 and a methylation difference higher than 25%.
Author contributions
CRG performed most of the experiments and participated in the writing of the manuscript, OMB contributed experimentally, PM designed the semi-automated image analysis workflow, AP co-supervised the beginning of the study, RLJ co-supervised the study and contributed to the writing of the manuscript, JF co-supervised the study, analyzed the RRBS data, wrote the manuscript and received the funding.
Acknowledgements
This work was supported by the Fondation de France (00086320 to J.F.) and the Fondation du Collège de France (to J.F.). We thank all primary donors for their financial contributions to this work. We thank the animal facility members for their essential contributions. We gratefully acknowledge Julien Dumont and the Collège de France Orion imaging facility (IMACHEM-IBiSA), member of the French National Research Infrastructure France-BioImaging (ANR-10-INBS-04), which received support from the program «Investissements d’Avenir» ANR-10-LABX-54 MEMOLIFE. We thank Yves Dupraz for the manufacturing of a customized mouse brain slicer.
References
- 1.↵
- 2.↵
- 3.↵
- 4.↵
- 5.↵
- 6.↵
- 7.↵
- 8.↵
- 9.
- 10.↵
- 11.↵
- 12.↵
- 13.↵
- 14.↵
- 15.↵
- 16.↵
- 17.
- 18.↵
- 19.↵
- 20.↵
- 21.↵
- 22.↵
- 23.↵
- 24.↵
- 25.↵
- 26.↵
- 27.↵
- 28.↵
- 29.↵
- 30.↵
- 31.↵
- 32.↵
- 33.↵
- 34.↵
- 35.↵
- 36.↵
- 37.↵
- 38.↵
- 39.↵
- 40.↵
- 41.↵
- 42.↵
- 43.↵
- 44.↵
- 45.↵
- 46.↵
- 47.↵
- 48.↵
- 49.↵
- 50.↵
- 51.↵
- 52.↵
- 53.↵
- 54.↵
- 55.↵
- 56.↵
- 57.↵
- 58.
- 59.↵
- 60.↵
- 61.↵
- 62.↵
- 63.↵
- 64.↵
- 65.↵
- 66.↵
- 67.↵
- 68.↵
- 69.↵
- 70.↵
- 71.↵
- 72.↵
- 73.↵
- 74.↵
- 75.↵
- 76.↵
- 77.↵
- 78.↵
- 79.↵
- 80.↵
- 81.↵
- 82.↵
- 83.↵
- 84.↵
- 85.↵
- 86.↵
- 87.↵
- 88.↵
- 89.↵