Neuronal Oxidative Stress Promotes α-Synuclein Aggregation In Vivo

Both genetic and environmental factors increase risk for Parkinson’s disease. Many of the known genetic factors influence α-synuclein aggregation or degradation, whereas most of the identified environmental factors produce oxidative stress. Studies using in vitro approaches have identified mechanisms by which oxidative stress can accelerate the formation of α-synuclein aggregates, but there is a paucity of evidence supporting the importance of these processes over extended time periods in brain. To assess this issue, we evaluated α-synuclein aggregates in brains of three transgenic mouse strains: hSyn mice, which overexpress human α-synuclein in neurons and spontaneously develop α-synuclein aggregates; EAAT3−/− mice, which exhibit a neuron-specific impairment in cysteine uptake and resultant neuron-selective chronic oxidative stress; and double-transgenic hSyn/EAAT3−/− mice. Aggregate formation was evaluated by quantitative immunohistochemistry for phosphoserine 129 α-synuclein and by an α-synuclein proximity ligation assay. Both methods showed that the double transgenic hSyn/EAAT3−/− mice exhibited a significantly higher α-synuclein aggregate density than littermate hSyn mice in each brain region examined. Negligible aggregate formation was observed in the EAAT3−/− mouse strain, suggesting a synergistic rather than additive interaction between the two genotypes. A similar pattern of results was observed in assessments of motor function: the pole test and rotarod test. Together, these observations indicate that chronic, low-grade neuronal oxidative stress promotes α-synuclein aggregate formation in vivo. This process may contribute to the mechanism by which environmentally induced oxidative stress contributes to α-synuclein pathology in idiopathic Parkinson’s disease.


Introduction
Several lines of evidence indicate that α-synuclein aggregates or oligomers contribute to Parkinson's disease [1][2][3][4]. Evidence also indicates that oxidative stress contributes to the pathogenesis of Parkinson's disease [5]. While α-synuclein aggregation and oxidative stress may act by independent mechanisms, they may also converge on a common final pathway. One potential mode of convergence involves the effects of oxidative stress on α-synuclein aggregation. Studies using in vitro preparations have identified mechanisms by which oxidative stress can accelerate the formation of toxic aggregates [6][7][8][9][10][11][12][13] but there is a paucity of evidence that chronic oxidative stress contributes to aggregate formation in vivo. In this study we address this issue using a transgenic mouse strain that exhibits chronic neuronal oxidative stress.
α-synuclein exists in neurons as monomeric species in equilibration with ordered tetramers, and in both forms the protein conformation is likely dependent upon association with lipid membranes [1,14,15]. Higher order soluble α-synuclein aggregates are cytotoxic, and likely a major factor driving cell death and disease progression in Parkinson's

Mouse Breeding and Husbandry
Line 61 Thy-1 α-synuclein (hSyn) mice on the DBA/2 strain were obtained from Dr. Elezier Masliah, UCSD. The hSyn mice carry the human α-synuclein transgene on the X chromosome (X hSyn ). We initially backcrossed these mice onto the C57BL/6 strain (Jackson Labs), but due to poor breeding on that background the mice were re-crossed with the DBA/2 parental strain to generate a colony on a 1:1 DBA/2: C57BL/6 background, as described previously [26][27][28]. EAAT3 −/− mice on the C57BL/6 background were then likewise crossed with DBA/2 mice to generate a colony on a 1:1 DBA/2: C57BL/6 background. The hSyn and EAAT3 −/− mice were then interbred with repeated intercrosses for 6 generations to generate a colony with a homogeneous 1:1 DBA/2: C57BL/6 background. Female XX hSyn mice have a very attenuated phenotype [26], and so only male mice were used for experiments. Mice used for experiments were generated by crossing EAAT3 -/+ , XY males with EAAT3 -/+ , X hSyn X females to produce male mice with the four genotypes of interest; wild type (EAAT3 +/+ , XY); hSyn (EAAT3 +/+ , X hSyn Y); EAAT3 −/− (EAAT3 −/− , XY); and hSyn/EAAT3 −/− (EAAT3 −/− , X hSyn Y). The numbers of each genotype used for the experiments are listed on Table 1, along with weight and mortality data. Mice of the hSyn/EAAT3 −/− genotype exhibited greater mortality over the 8 month experimental period than the other mouse genotypes (Table 1).  16 4 Weights are means ± s.e.m. from 6 mice of each genotype at age 4 months. Mortality is the number of mice that died before the 8-month behavioral assessment.

Experimental Design
Studies were approved by the San Francisco Veterans Affairs Medical Center animal studies committee and performed in adherence to the Guide to the Care and Use of Experimental Animals and The Ethics of Animal Experimentation and the 2020 ARRIVE guidelines [29]. A total of 71 mice were assessed, divided into 4 cohorts of 16-20 mice, each with a roughly equal distribution of genotypes. Each cohort of mice underwent behavioral assessments first at age 3-4 months and then again at age 7-8 months. At age 8 months the mice were euthanized and brains removed for histology. All brains from each cohort were fixed, immunostained, and analyzed in a single batch.

Behavioral Assessments
Each cohort of mice was assessed using the rotarod test and the pole test as described [19,30], with minor modifications. The individuals handling and scoring the mice were blinded to mouse genotype. For habituation to the rotarod test, mice were placed for 90 s on a horizontal rod (San Diego Instruments Rota-rod) rotating at 4 rpm (1st day) and 5.5 rpm (2nd day), 5 times each day. For testing, mice were placed on the horizontal rod at a start speed 0 rpm and an acceleration rate of 6 rpm/minute, and the time until fall from the rod was recorded. Tests were repeated 6 times each day for 2 consecutive days. The lowest and highest time for each mouse was discarded and the average time from the remaining 10 trials was calculated for each mouse. For the pole test, mice were placed face-up at the top of a pole (45 cm length, 1.2 cm diameter), wrapped in tape for grip, and scored for turn-around time (time taken to face downward) and descent time (time taken to come down the pole). A cardboard square was affixed to the top of the pole to prevent mice from crossing up and over the pole to descend. Each assessment involved 2 days of habituation followed by 3 days of testing. Testing was performed twice per day every other day and video recorded. The recordings were later analyzed by an observer who was also blinded to the treatment conditions. Timing was begun with the first mouse movement, and any time exceeding 3 s during which the mouse was motionless was subtracted from the total time. The longest and shortest times were discarded, and the average time from the remaining 10 trials was calculated for each mouse. Descent-time was not used for data analysis because some impaired mice were unable to descend the pole normally and instead either slid down or used their tails to assist. Results obtained from mice that died before the second behavioral assessment were also excluded from the data analysis.

Immunohistochemistry and Image Analysis
Deeply anesthetized mice were perfused with cold 0.9% NaCl followed by perfusion with freshly depolymerized paraformaldehyde (PFA; 4%) in 0.1 M sodium phosphate buffer (PB) pH 7.0. The brains were removed and post-fixed in the same PFA/PB solution for 18 h, followed by incubation for 24 h in 20% sucrose for cryoprotection. The brains were then frozen and coronal sections (40 µm thickness) were prepared with a cryostat. For immunohistochemistry, the sections were incubated in blocking buffer (2% donkey serum and 0.1% bovine serum albumin in 0.1 M PB) at room temperature for 30 min, and then incubated with the primary antibodies overnight at 4 • C. The antibody sources and dilutions used are listed in Table 2. After washing, antibody binding was detected using fluorescent secondary antibodies. The sections were then mounted on glass slides in DAPI-containing mounting medium (Vector laboratories, Burlingame, CA). Controls were prepared omitting either primary or secondary antibodies. Immunostaining for phosphoserine-129 α-synuclein (pSyn) was quantified in cortex and hippocampus in evenly spaced sections spanning bregma −1.5 to −2.3. The sections were immunostained with anti-pSyn and anti-NeuN. Confocal images were taken from both the right and left hemisphere of a pre-determined region of cortex (layer 5) and hippocampus CA1 on each of three sections from each animal. Two optical sections through each region were stacked in Image J software. Neuronal perikarya as identified by NeuN staining were used as a mask to identify the regions of interest, and the mean intensity of anti-pSyn immunofluorescence within these regions was measured. For analysis of the substantia nigra, sections were additionally immunostained for tyrosine hydroxylase (TH) to permit assessment specifically in dopaminergic neurons. The substantia nigra pars compacta was defined by the pattern of the TH-positive neurons, as previously described [19,31]. Confocal images were taken at both right and left hemisphere of substantia nigra from three sections spaced 240 µm apart (bregma −2.8 to −3.6). Two optical sections through each region were stacked in Image J software. The TH-positive neurons were identified in the confocal images and NeuN-demarcated perikarya of the TH-positive cells was used as a mask to identify the regions of interest. The measurements made in the left and right cortex of all three sections were then averaged to generate a single value for the cortex of each brain, and the same was done for hippocampus and substantia nigra. To minimize potential differences introduced by variations in fixation, antibody binding, or photography, all values from each cohort were normalized to the mean of the wild-type values measured in the cohort. Both the persons taking the photographs and analyzing the images were blinded to the mouse brain genotypes.

Proximity Ligation Assay
In two cohorts of mice, α-synuclein density was additionally analyzed using a proximity ligation assay (PLA), as described [32]. The experiments were performed using Duolink PLA Brightfield kits from Millipore Sigma according to the manufacturer's manual. In brief, brain sections mounted on glass slides were incubated with 10 mM sodium citrate, pH 6.0 for 10 min at 95 • C for antigen retrieval, then incubated in hydrogen peroxide to quench endogenous peroxidase activity. Antibody-oligonucleotide conjugates were prepared using the Duolink Probemaker kit by incubating mouse anti-α-synuclein with plus-or minussense oligonucleotide probes. The brain sections were then immersed sequentially in Duolonk ligase, polymerase and amplification reagents, and Duolink detection solution. The sections were counterstained to identify cell nuclei and dehydrated in graded ethanol and citrisolve.
In a separate study, the PLA method was used to identify α-synuclein-4-hydroxynonenal conjugates. For these studies the sections were incubated with mouse antibody to αsynuclein and rabbit antibody to 4-hydroxynonenal, as previously described [33]. After washing, the sections were incubated with anti-rabbit plus sense and anti-mouse minus sense oligonucleotide probes. The brain sections were then immersed sequentially in Duolink blocking solution, the antibody-oligonucleotide conjugates, Duolonk ligase, polymerase and amplification reagents, and Duolink detection solution.
PLA signals were quantified using a stereological microscope (Stereo Investigator, MicroBrightField, Williston, VT). Every sixth coronal section spanning the substantia nigra (Bregma level -2.5 to -3.5) was used, which includes cortex and hippocampal regions. A counting frame (25 × 25 × 20 µm) was placed at each intersection of a 200 × 200 µm sampling grid superimposed onto regions of interest manually outlined on each section. For cortex, this was a 1 mm 2 area polygon on the dorsal surface (including layers 2-6); for hippocampus, this was a 1 mm 2 polygon encompassing dorsal CA1, stratum oriens and stratum radiatum; and for substantia nigra the polygon was drawn using morphological landmarks [19,31]. PLA puncta in the counting frames were manually counted using 40x objective (Axioskop 2 plus; Carl Zeiss Microscopy, LLC, White Plains, NY, USA). In many instances the PLA puncta were overlapping. For this reason, the analysis was performed two ways; with only the discrete single puncta counted and then with only the overlapping clusters of puncta counted. The tissue preparation and puncta counts were performed by observers blinded to the mouse genotype. Data analysis was performed separately for single puncta and multi puncta counts, using the same method as for the pSyn immunostaining.

Western Blots
Brain wedges containing cortex and hippocampus were homogenized in RIPA buffer (Thermo Scientific, Rockford, IL, USA) containing Halt protease and phosphatase inhibitor cocktail (Thermo Scientific). The tissue lysate was centrifuged at 12,000× g for 10 min at 4 • C and 30 µg of protein was subjected to electrophoresis on NuPAGE 4-12% Bis-Tris gel (Invitrogen, Carlsbad, CA, USA) and transferred to a PVDF membrane. The blot was incubated in 10% BSA in Tris-Buffered Saline with Tween 20 (TBST) for 1 h and then incubated overnight with a primary antibody ( Table 2). The blot was washed with TBST and reacted with a horseradish peroxidase secondary antibody for 1 h. The immunoreactivity was visualized by a chemiluminescent substrate (Thermo Scientific) and the signals were acquired by using a C300 imager (Azure Biosystems, Dublin, CA, USA).

Statistics
All data are expressed as means ± s.e.m., with the "n" of each study defined as the number of mice. Data were analyzed using ANOVA and Sidak's test for multiple comparisons, or by t-tests where only two groups were compared. Normality of each data set was assessed by the Shapiro-Wilk test, and all data sets passed (did not significantly deviate from normality) at the α = 0.05 level.

Results
The four genotypes of mice generated by the transgenic crosses were not grossly distinguishable from one another and did not significantly differ in size over the age range used for these studies (Table 1). Immunostaining and Western blotting for the glutamate/cysteine transporter EAAT3 confirmed expression to be absent in the EAAT3 −/− and hSyn/EAAT3 −/− genotypes, whereas immunostaining and Western blots for human α-synuclein confirmed it to be strongly expressed in the hSyn and hSyn/EAAT3 −/− genotypes ( Figure 1A and Figure S1). Oxidative stress was confirmed in both the EAAT3 −/− and the hSyn/EAAT3 −/− mouse strains by immunostaining for the lipid peroxidation marker 4-hydroxynonenal ( Figure 1B). range used for these studies (Table 1). Immunostaining and Western blotting for the glutamate/cysteine transporter EAAT3 confirmed expression to be absent in the EAAT3 −/− and hSyn/EAAT3 −/− genotypes, whereas immunostaining and Western blots for human α-synuclein confirmed it to be strongly expressed in the hSyn and hSyn/EAAT3 −/− genotypes ( Figures 1A and S1). Oxidative stress was confirmed in both the EAAT3 −/− and the hSyn/EAAT3 −/− mouse strains by immunostaining for the lipid peroxidation marker 4-hydroxynonenal ( Figure 1B). α-synuclein aggregates were assessed by quantitative immunohistochemistry using an antibody that recognizes both mouse and human phosphoserine-129 α-synuclein [34]. In cortical neurons, the immunostaining for phosphoserine-129 α-synuclein (abbreviated as pSer129 α-syn) revealed no detectable signal in the WT or EAAT3 −/− strains, but showed a robust signal in the hSyn strain as previously reported [26,27]. Notably, the pSer129 α-syn signal was increased in the hSyn/EAAT3 −/− mice relative to that observed in the hSyn mice (Figure 2A,C). As a confirmatory approach, aggregate formation was also assessed using a proximity ligation assay with antibodies to human α-synuclein. Aggregates are visualized as small puncta by this approach (Figure 2B, inserts). In some cases the puncta occurred in clusters, making it difficult to count the individual dots. For this reason, the data were quantified in two ways: with only discrete puncta counted, and with only clusters of puncta counted. As shown in Figure 2, the two quantification methods provided similar results and showed the same pattern as observed with the pSer129 α-syn immunostaining: a negligible signal in WT or EAAT3 −/− mouse cortical neurons, a robust signal in the hSyn neurons, and a significantly greater signal in the hSyn/EAAT3 −/− neurons ( Figure 2B,D,E). A modification of the PLA method was used to determine if αsynuclein undergoes increased oxidative modification in the hSyn/EAAT3 −/− neurons. In this method, 4-hydroxynonenal-α-synuclein conjugates were detected using an antibodies to these two epitopes [33]. The studies revealed a greater 4-hydroxynonenal-α-synuclein PLA signal in hSyn/EAAT3 −/− than in hSyn brain sections ( Figure S2).
hSyn mice also develop α-synuclein aggregates in neurons of the hippocampus [26], and we assessed aggregate formation in the hippocampal CA1 neurons using the same methods as for cortical neurons. We again found a good correspondence between results obtained by the two methods, and the same pattern of results was observed in the cortical neurons: a negligible signal in the WT and EAAT3 −/− mice, a robust signal in the hSyn mice, and a significantly greater signal in the hSyn/EAAT3 −/− mice ( Figure 3).
Prior reports using this hSyn mouse strain have noted less aggregate formation in the substantia nigra pars-compacta than in hippocampus or cortex, as assessed using pSer129 α-syn immunoreactivity [26,27]. Our results arere similar in that the pSyn129 signal was less robust in dopaminergic neurons of the substantia nigra pars compacta than in cortex or hippocampus (Figure 4). Nevertheless, the pattern of aggregates formation across the four mouse genotypes was the same as observed in the other structures: the pSyn129 signal was negligible in the WT or EAAT3 −/− mice, present in the hSyn/EAAT3 −/− mice, and significantly greater in the hSyn/EAAT3 −/− mice ( Figure 4A,C). Aggregate formation assessed by the PLA assay in the substantia nigra showed the same pattern of results ( Figure 4B,D,E). Of note, the magnitude of PLA signal in the substantia nigra was comparable to that observed in the cortex and hippocampus, unlike the pSyn 129 signal which was markedly less prominent in the substantia nigra than in those structures.
These histological assessments were accompanied by two measures of motor function, the pole test and the rotarod test. The pole test revealed a small but significantly greater impairment in the hSyn/EAAT3 −/− mice relative to the hSyn mice at age 4 months, and a somewhat larger impairment at age 8 months ( Figure 5A). Results of the rotarod test similarly showed a trend toward greater impairment in hSyn/EAAT3 −/− mice, though these differences did not achieve statistical significance at the p < 0.05 level ( Figure 5B).

Discussion
These studies compared α-synuclein aggregate formation in hSyn mice, which overexpress human α-synuclein, to hSyn mice crossed with EAAT3 −/− mice, which in addition exhibit chronic neuron-specific oxidative stress. Both quantitative pSyn129 immunohistochemistry and a proximity ligation assay showed that the double transgenic hSyn/EAAT3 −/− mice exhibited significantly higher α-synuclein aggregate densities than littermate hSyn mice in each brain region examined. Negligible aggregate formation was observed in the parental EAAT3 −/− mouse strain, suggesting a synergistic rather than additive interaction between the two genotypes. A similar pattern of results was observed in assessments of motor function.
Many of the known genetic variants imparting risk for Parkinson's disease affect either α-synuclein per se or aspects of protein quality control [35], while almost all environmental factors known to contribute to Parkinson's disease generate intracellular oxidative stress [36,37]. A link between α-synuclein aggregation and oxidative stress has been established by the observations that α-synuclein aggregates themselves can induce oxidative stress [19,38], and that mice deficient in α-synuclein exhibit less dopaminergic neuronal loss when exposed to the oxidant-producing compounds MPTP or rotenone [39,40]. Moreover, cell culture studies have identified several mechanisms by which oxidative stress can accelerate α-synuclein aggregation [6][7][8][9], including conjugation of α-synuclein with the lipid peroxidation product 4-hydroxynonenal [41][42][43]. In line with those findings, the PLA assessment of 4-hydroxynonenal-α-synuclein conjugates performed here showed increased signal in the hSyn/EAAT3 −/− mice. However, the interpretation of this PLA finding is complicated by the fact that the hSyn/EAAT3 −/− neurons also contain an increased density of α-synuclein aggregates, and this increased aggregate density may contribute to the overall increase in the 4-hydroxynonenal-α-synuclein conjugate signal.
Despite the several possible mechanisms by which oxidative stress may promote α-synuclein aggregation, there have been few prior demonstrations of chronic oxidative stress influencing α-synuclein aggregation in vivo. Pharmacological studies using rotenone or MPTP (1-methyl-4-phenyl-1,2,3,6-tetrahydropyridine) to induce oxidative stress and substantia nigra dopaminergic cell loss show generally sparse evidence of α-synuclein aggregate formation [44][45][46][47][48], possibly because of the relatively short duration of these studies. A genetic approach using SOD2 +/mice crossed with mice expressing the human A30P mutant α-synuclein observed an increased in α-synuclein aggregation relative to the parental A30P α-synuclein strain, as assessed by poteinase-K resistant inclusions [49]. This study differed from the present work in that the mice expressed a mutant rather than wild-type α-synuclein, and the method of aggregate detection was different. Given that SOD2 haploinsufficiency causes oxidative stress in all cell types, it is also possible that systemic or specifically microglial effects of the oxidative stress [50] rather than a direct intraneuronal mechanism may contribute to the findings in these mice. Despite these differences, the study using SOD2 +/mice nevertheless comports with the present results indicating that chronic oxidative stress promotes α-synuclein aggregate formation in the intact brain.
There is evidence that EAAT3, in addition to serving as a cysteine transporter, may also function to take up glutamate at glutamatergic synapses [51]. To the extent this occurs, neurons in EAAT3 −/− mice could also undergo increased oxidative stress as a consequence of increased activation of NMDA-type glutamate receptors [52,53]. Additionally, elevated extracellular glutamate levels could deplete intracellular cysteine levels (and hence glutathione levels) via glutamate cystine exchange [54]. These possibilities do not significantly alter the interpretation of our results, however, because both of these potential effects of EAAT3 −/− would also lead to increased neuronal oxidative stress.
The patterns of α-synuclein aggregation we observed in the hSyn (Line 61) mice corresponds with prior studies using this mouse strain. As in prior reports [26,27], we found aggregate accumulation as detected by pSyn129 immunostaining to be less robust in the substantia nigra pars reticulata than in cortex or hippocampus. However, by the PLA method we found aggregates to be comparably dense in the three brain regions, consistent with aggregate assessments using other methods [55,56]. Human α-synuclein is expressed to similar levels in all three regions in hSyn mice [26]. The reason for the reduced amount of pSyn129 immunostaining signal in the substantia nigra is thus unclear, but may be attributable to cell-type differences in α-synuclein phosphorylation.
Results of our behavioral assessments also correspond to prior studies using the hSyn mice [28,57]. On the pole test, turnaround time was greater in the hSyn/EAAT3 −/− than in the hSyn mice at both the four-month and eight-month time points, and this difference was further increased at the eight-month time point. The double transgenic mice also did most poorly on the rotarod test, at both time points assessed, though this difference did not reach statistical significance at the p < 0.05 level. It is possible that the increased mortality in the hSyn/EAAT3 −/− mice led to an underestimate of the behavioral impairment in that genotype. Unexpectedly, all groups did slightly better at eight months than at four months on this test. The reason for this is not clear, though possibly this was a learning effect with repeated testing over time.
A limitation of our study is that accelerated neuronal death does not occur in the hSyn mice, in the substantia nigra or elsewhere, thus obviating use of that experimental endpoint. An additional limitation is that only male mice could be used, because the female mice have an attenuated phenotype [26]. However, there are only small sex differences in the incidence and progression of human Parkinson's disease [58], and the absence of female mice in the current study should not negate the central finding that aggregate formation is promoted by chronic oxidative stress.

Conclusions
The findings of this in vivo study extend prior observations obtained using cell culture and cell-free systems, and support the contention that chronic, low-grade oxidative stress promotes the formation of α-synuclein aggregates in brain neurons. This process may contribute to the mechanism by which environmentally induced oxidative stress contributes to the α-synuclein pathology in idiopathic Parkinson's disease.
Supplementary Materials: The following supporting information can be downloaded at: https:// www.mdpi.com/article/10.3390/antiox11122466/s1, Figure S1: Western blots of EAA3 and human αsynuclein from brains of the transgenic mouse strains, Figure S2: Proximity ligation assay assessments of 4-hydroxynonenal-α-synuclein conjugates in neurons of cortex and hippocampus CA1 cell field

Data Availability Statement:
The raw data and images that were analyzed for this publication will be provided upon request to the corresponding author: raymond.swanson@ucsf.edu