Parkin regulates microglial NLRP3 and represses neurodegeneration in Parkinson's disease

Abstract Microglial hyperactivation of the NOD‐, LRR‐, and pyrin domain‐containing 3 (NLRP3) inflammasome contributes to the pathogenesis of Parkinson's disease (PD). Recently, neuronally expressed NLRP3 was demonstrated to be a Parkin polyubiquitination substrate and a driver of neurodegeneration in PD. However, the role of Parkin in NLRP3 inflammasome activation in microglia remains unclear. Thus, we aimed to investigate whether Parkin regulates NLRP3 in microglia. We investigated the role of Parkin in NLRP3 inflammasome activation through the overexpression of Parkin in BV2 microglial cells and knockout of Parkin in primary microglia after lipopolysaccharide (LPS) treatment. Immunoprecipitation experiments were conducted to quantify the ubiquitination levels of NLRP3 under various conditions and to assess the interaction between Parkin and NLRP3. In vivo experiments were conducted by administering intraperitoneal injections of LPS in wild‐type and Parkin knockout mice. The Rotarod test, pole test, and open field test were performed to evaluate motor functions. Immunofluorescence was performed for pathological detection of key proteins. Overexpression of Parkin mediated NLRP3 degradation via K48‐linked polyubiquitination in microglia. The loss of Parkin activity in LPS‐induced mice resulted in excessive microglial NLRP3 inflammasome assembly, facilitating motor impairment, and dopaminergic neuron loss in the substantia nigra. Accelerating Parkin‐induced NLRP3 degradation by administration of a heat shock protein (HSP90) inhibitor reduced the inflammatory response. Parkin regulates microglial NLRP3 inflammasome activation through polyubiquitination and alleviates neurodegeneration in PD. These results suggest that targeting Parkin‐mediated microglial NLRP3 inflammasome activity could be a potential therapeutic strategy for PD.


| INTRODUC TI ON
Parkinson's disease (PD) is a common neurodegenerative disease characterized by progressive dopaminergic neuron loss (Balestrino & Schapira, 2020). PD pathogenesis has not been clearly elucidated; however, among the existing theories, the interplay between genetic susceptibility and environmental effects is widely accepted (Kalia & Lang, 2015). The PRKN gene is the most common autosomal recessive gene in PD; it encodes the Parkin protein, a RINGtype E3 ubiquitin ligase. Parkin has a recognized role in mitophagy and quality control in the mitochondria (Pickrell & Youle, 2015).
Therefore, mice with PINK 1 deficiency mice have similar phenotypes as mice with Parkin deficiency mice (Paul & Pickrell, 2021).

Matheoud et al. demonstrated that intestinal infection triggers
motor symptoms in PINK1 − mice (Matheoud et al., 2019). Frank-Cannon et al. reported that a low-dose continuous injection of LPS induces neurodegeneration in Parkin-deficient mice (Frank-Cannon et al., 2008), revealing details about the interplay between Parkin deficiency and neuroinflammation.
Central and peripheral inflammation plays an important role in driving PD pathology (Nguyen & Palm, 2022;Pajares et al., 2020).
Briefly, gut microbiota dysbiosis and immune system alterations have an influence on the central nervous system, leading to microglial activation, inflammatory response, and subsequent neuronal death (Lazdon et al., 2020;Sampson et al., 2016). Among the numerous inflammatory pathways, NOD-, LRR-and pyrin domain-containing 3 (NLRP3) inflammasome activation is a main source of inflammatory regulation in microglia (Haque et al., 2020). NLRP3 expression is induced after stimulation, after which the complex incorporates the apoptosis-associated speck-like protein containing a CARD (PYCARD/ASC) adaptor and the effector pro-caspase-1 into the inflammasome assembly and induces protein cleavage and cytokine secretion (Huang et al., 2021). Elevated NLRP3 activation was found in patients with PD and in various animal models of PD, driving dopaminergic neuron death. Inhibition of NLRP3 prevents PD pathology in mice, indicating a crucial role of the NLRP3 inflammasome in the onset of PD (Gordon et al., 2018;Lee et al., 2019;von Herrmann et al., 2018). Recent studies have shown that Parkin may exhibit broad substrate selectivity (Shires et al., 2017), and inflammatory pathways are possible targets of Parkin (Quinn et al., 2020;Sliter et al., 2018), although the underlying mechanism remains unclear.
Increasing evidence suggests that Parkin may regulate inflammation through the ubiquitination of NLRP3. NLRP3 is a protein with multiple ubiquitination sites, which induce NLRP3 degradation through autophagy (Han et al., 2019). Polyubiquitination of NLRP3 by tripartite motif containing 31 (TRIM31), another RING-type E3 ubiquitin ligase, induces ubiquitin-proteasome degradation (Song et al., 2016). De-ubiquitination of NLRP3 facilitates its activation; conversely, ubiquitination of NLRP3 inhibits its activation (Juliana et al., 2012;Tang et al., 2021). Ubiquitination is a vital form of posttranslational modification, the status of which determines whether NLRP3 becomes activated or degraded. Recently, NLRP3 was reported to be a substrate of Parkin in neurons and BEAS-2B epithelial cells (Ge et al., 2021;Panicker et al., 2022). However, no experimental evidence has shown whether this occurs in the microglia. The total RNA sequence obtained from Genevestigator software revealed significantly higher PRKN RNA expression levels in microglia than in other neural cells, indicating that Parkin may play an important role in microglia ( Figure S5) open field test 1-week post-injection; Rotarod test 1-, 2-, 4-, and 6-months post-injection; and pole test 6 months post-injection.
Two mice in the KO-LPS group died after LPS injection, and one mouse in the KO-PBS group died from fighting with other mice; they were excluded from the study.

| Cellculture
Mouse primary microglia (PM) were isolated from WT and Parkin KO mice on a C57BL6/J background. Newborn mice (P0) were decapitated, and cortices, without the meninges or blood, were collected and digested in 0.25% trypsin at 37°C for 15 min. Cells from three cortices were filtrated through a 70μm-pore-size filter and seeded

| Co-immunoprecipitation(Co-IP)
Co-IP was conducted according to a commercial instructions of Co-IP kits (Thermo Scientific™,88,804). Briefly, two 10 cm dishes of BV2 cells were transfected with flag-Parkin plasmid for 48 h and then treated with LPS for 4 h. Cells from each 10 cm dish were harvested with 500 μL NP-40 detergent containing protease inhibitors.
Five percent of cell lysate was used for the input; the remainder was incubated overnight with magnetic beads coated with antibodies at 4°C. Cell lysate was incubated with 2 μg target antibody or IgG control. NLRP3 or Flag antibodies were used for Co-IP and reverse Co-IP. Beads were boiled at 100°C for 10 min in 20 μL loading buffer to elute protein.

| Behavioraltesting
For the open field test, mice were left undisturbed in the testing room for 30 min for adaptation, then placed in the middle of the field and allowed to roam freely for 5 min. The distance traveled was measured by SMART video tracking software (Smart 3.0). The field was equally divided into 16 areas; the sum of the distances travelled in all 16 areas indicated the total distance, and the sum of the distance travelled in the four middle areas represented the "middle distance." The activity of rearing was recorded by an observer during the 5-min experiment.
For the Rotarod test, mice were placed on a rotating rod accelerating from 4-40 rpm over a 5 min period. Before testing, mice were pre-trained for 5 days, and the average fall latency was recorded as the baseline. In each experiment, mice were tested five separate times, and the average latency to fall was recorded. The cut-off latency was 150 s. The Rotarod apparatus was provided by Panlab, Barcelona,Spain (LE8205).
For the pole test, a wooden instrument, 50-cm in length and a pole with a 1-cm diameter, was used. A ball with a 2.5-cm diameter was situated at the top of the pole. During the test, the pole was raised at a 90° angle to the ground. Mice were placed on the ball and allowed to climb down spontaneously. The time taken to reach the bottom from the top was recorded as the climb time. Five independent tests were conducted for each mouse to determine the average.

| Immunofluorescence
Mice were anesthetized with 1% pentobarbital through intraperitoneal injection. Anesthetized mice were perfused intracardially with PBS first and then 4% paraformaldehyde to fix brain tissue. Brains were fixed in 4% paraformaldehyde for 24 h then dehydrated and cryoprotected in 30% sucrose until tissues sunk to the bottom of the container. Frozen brains embedded in Optimal Cutting Temperature (OCT) compound (Sakura, 4583) were sliced on a cryostat microtome into 30μm coronal sections (for neuron counting) or 12μm coronal sections (for microglia and protein labeling).
Brain slices and fixed cell climbing slices were incubated in 5% bovine serum albumin (BSA) containing 0.3% Triton X-100 for 1 h, followed by an overnight primary antibody incubation at 4°C. The next day, slices were washed three times in PBS and incubated in a secondary antibody solution for 1.5 h at 37°C. Subsequently, slices were washed five times and mounted with anti-fade agents containing 4′, 6-diamidino-2-phenylindole (DAPI). Images were captured using a fluorescence microscope (Leica DM6B) and confocal laser endomicroscopy (Olympus FV1200). The number of tyrosine hydroxylase-positive (TH + ) neurons were counted via stereology.
Thirtyμm coronal sections of midbrain tissue containing substantia nigra pars compacta (SNc) were used for neuronal counting. We examined 6-7 sections per brain at similar layer and calculated the average number of cells.

| Westernblotting(WB)
Cell lysates were collected using radioimmunoprecipitation assay (RIPA) buffer containing a protease inhibitor cocktail and phosphatase inhibitor mixture and then submitted to ultrasonification for 5 s. Brains were homogenized using a tissue grinder in RIPA buffer containing a protease inhibitor cocktail. Cell lysates or tissue homogenates were then centrifuged at 13,800 g at 4°C for 15 min, and the supernatant was boiled in 1x loading buffer for 8 min at 100°C.
WB was conducted as previously described .
MitoSOX was diluted to a 10 μmol/L working solution in DMEM, then added to cell culture plates and incubated for 30 min. PM were washed twice with PBS, then digested in trypsin at 37°C for 15 min. ROS levels were quantified by measuring fluorescence intensity using a flow cytometer (CytoFLEXLX, Beckman). Data processing was conducted using CytExpert Software (V2.0).

| Statisticalanalysis
WB and immunofluorescence images were quantized using FIJI ImageJ software based on pixel intensity (https://imagej.net/softw are/fiji). Co-localization analysis and calculation of Pearson's R values were conducted using the coloc-2 plugin. Raw data process was conducted using Microsoft Excel 2013. Statistical analysis was conducted using GraphPad Prism 8.0 software. Student's t-tests, oneway analysis of variance (ANOVA), and two-way ANOVA were used for significance testing. Tukey's multiple comparison post hoc test was used for comparisons between groups when appropriate.

| ParkinregulatesNLRP3inflammasome activationinBV2cellsandPM
BV2 cells were stimulated with LPS and ATP, then NLRP3 inflammasome activation was confirmed by a significant increase in NLRP3 expression and IL-1β secretion. Parkin protein levels did not decrease until 24 h post-stimulation (pS), and they were dramatically downregulated at 48 h pS ( Figure S1a

| InteractionbetweenParkinandNLRP3after LPS stimulation
To determine whether the ~180 kDa band of NLRP3 was a posttranslational modification band mediated by Parkin, we tested whether Parkin and NLRP3 interacted. We performed protein-protein affinity prediction using the PPA-Pred2 website (www.iitm.ac.in/bioin fo/ PPA_Pred/predi ction.html). The predicted value of Delta G (binding free energy) was −12.73 kcal/mol, and the Kd (dissociation constant) was 4.59e-10 M. Confocal immunofluorescence microscopy of LPStreated PM revealed Parkin and NLRP3 co-localization (Figure 2a).
Two-photon fluorescence imaging was also used for BV2 cells after

| Parkindeficiencyinducesgreatermicroglial activationandneuroinflammationinmiceafteracute LPS treatment
Iba-1 fluorescent labeling was conducted to evaluate microglial activation in brain slices at post-injection day 1. Confocal imaging showed significant microglial activation, characterized by larger cell sizes and amoeboid morphology, after intraperitoneal LPS injections in WT and KO mice. The differences were seen in the corpus striatum and substantia nigra (SN) areas (Figure 4a). In the KO-PBS group, microglial activation was also evident in the absence of inflammatory stimulation (Figure 4a). Iba-1-positive cells were counted to evaluate the degree of microglial proliferation in response to different inflammatory triggers. The KO-PBS group exhibited a significantly higher number of microglia than the WT-PBS group. The same trend was found between the KO-LPS and WT-LPS groups, but the difference was not significant (Figure 4b).
NLRP3 levels were measured from confocal images. As expected, the mean NLRP3 fluorescence intensity was five times higher in the KO-PBS group than the WT-PBS group. NLRP3 levels were significantly upregulated in the KO-LPS group compared to those in the WT-LPS group (Figure 4c). Co-labeling of NLRP3 and Iba-1 revealed that the NLRP3 inflammasome was mainly activated in microglia, although some signals did not colocalize with microglial markers (Figure 4d). WB analyses showed significantly increased NLRP3 expression throughout the brain and ele-

| Parkindeficiencyinduceslong-term microglialactivation,chronicneuroinflammation,and greaterdopaminergicneuronloss6 monthsafter LPS treatment
To determine the long-term effect of LPS treatment in Parkin KO mice, we conducted behavioral testing and pathological analyses.
Rotarod testing was conducted at baseline and 1, 2, 4, and 6 months after intraperitoneal injection (Figure 5a). All four groups exhibited a decreased fall latency over this period due to aging and weight gain. No difference was seen between the WT-LPS and WT-PBS groups, whereas the KO-LPS group performed worse than the KO-PBS group beginning in the second month; however, the results at The mean CD11b fluorescence intensity was distinctly higher in the Parkin KO-LPS and KO-PBS groups, indicating chronic microglial activation (Figure 5c). Co-immunolabeling of CD11b and NLRP3 showed significant hyperexpression of NLRP3 in the SN, mostly within microglia ( Figure S2e,f). Correspondingly, the numbers of TH + neurons were significantly decreased in the KO-LPS group, but not in the WT-LPS group, which is consistent with the previous low-dose LPS exposure model (Frank-Cannon et al., 2008). The loss of TH + neurons did not differ between male and female mice (Figure 5f-h). Nissl staining was also conducted to confirm neuron loss ( Figure S3a,b). No significant difference of TH staining was seen in the corpus striatum ( Figure S3c). Microglial inflammation, TH + neuron loss, and the behavioral testing results were corroborative.
To investigate whether the neuronal death was caused by NLRP3 activation, we conducted an in vitro experiment by co-culturing primary microglia and SH-SY5Y cells. Parkin KO PM was treated with either LPS for 4 h followed by ATP for 30 min or MCC950 for 4 h together with LPS before ATP treatment. The supernatant was collected and added into SH-SY5Y cells and incubated for 48 h.
CCK-8 assay did not show a significant cell death in different groups ( Figure S3d); however, a significant decrease of Bcl-2/Bax was observed in LPS group, and such decrease was rescued by NLRP3 inhibitor MCC950 ( Figure S3e,f), indicating that activation of NLRP3 inflammasome in microglia leads to apoptotic execution propensity in neurons.

| HSP90α is a potential regulator of Parkininduced NLRP3 degradation
Molecular chaperones are vital regulators of protein degradation (Margulis et al., 2020). Among all kinds of molecular chaperones, HSP90 was reported to inhibit NLRP3 inflammasome activation  (Nizami et al., 2021). We performed protein-protein interaction (PPI) network analysis and constructed the PPI network using the STRING database (https://cn.strin g-db.org/) ( Figure S4a). The molecular chaperone protein, HSP90α, was a potential mediator between NLRP3 and Parkin. Furthermore, confocal fluorescence analysis of PM revealed moderate co-localization between HSP90α and NLRP3 F I G U R E 4 Parkin deficiency induced stronger microglia activation and heavier inflammatory response in mice. 12 WT and 12 Parkin KO male mice were divided into four groups (n = 6): WT-PBS, WT-LPS, KO-PBS, and KO-LPS. In each group, three were used for immunofluorescence and three were used for WB. (a, b) Iba-1 labeling of brain sections from SNc and striatum regions and Iba-1-positive cell counts. Each dot represents the SNc or striatum region of one mouse. (c, d) NLRP3 and Iba-1 co-labeling of brain sections of mice and statistical analysis of the integrated density of NLRP3 levels. Each dot represents one mouse. (e-g) Motion trajectories of mice (e) and distance traveled in the whole open field (f) or the middle area of the open field apparatus (g). (h-j) WB of NLRP3 inflammatory proteins in the brains of mice. Scale bar: 50 μm (a), 30 μm (d). KO, knock; LPS, lipopolysaccharide; NLRP3, NOD-, LRR-, and pyrin domain-containing 3; SNc, substantia nigra pars compacta; WB, Western blot; WT, wield type. *p < 0.05, **p < 0.01, ***p < 0.001, ****p < 0.0001.
( Figure S4b). The co-IP results also showed an interaction between HSP90α and both Parkin and NLRP3 ( Figure S4c,d).
To explore the regulatory role of HSP90α in Parkin-mediated NLRP3 degeneration, we used a specific inhibitor of HSP90, 17-AAG. NLRP3 content decreased in microglia under 17-AAG treatment, accompanied by a decreased downstream inflammatory response ( Figure S4e). 293 T cells overexpressing NLRP3 and Parkin also exhibited increased NLRP3 degradation after 17-AAG treatment ( Figure S4f,g). IP of NLRP3 was conducted after 4-h incubation with 17-AAG, and NLRP3 levels were slightly decreased in the input sample, accompanied by a significant elevation of the ubiquitination band ( Figure S4h,i). Therefore, HSP90α inhibition facilitates NLRP3 degradation through the UPP via Parkin mediation.
Intraperitoneal injection of LPS is a classical peripheral inflammation model that is also used as a chronic model of PD. Both singlefactor models, Parkin KO or LPS injection, require a long period to trigger the onset of PD. A previous study assessing a chronic inflammation model on Parkin KO mice showed similar neurodegeneration to our experiments, consistently proving that Parkin regulates inflammation (Frank-Cannon et al., 2008). Our research reveals that a single high-dose of LPS results in long lasting neuroinflammation in Parkin knockout mouse, suggesting that the combination of acute inflammatory response and gene mutation could be a trigger of late-life neurodegeneration. These results support the two-hit theory of PD pathogenesis, which involves synergy between gene mutations and environmental stress, and may support the gut-brain axis theory (Avagliano et al., 2022;Gao et al., 2011). However, in-

Previous research demonstrated that NLRP3 acts as a Parkin
substrate in neurons. Our results confirmed this mechanism also occurs in microglia and that Parkin induces NLRP3 degradation through the UPP, which is mediated by K48-linked polyubiquitination. Ubiquitination is a main regulatory mechanism of NLRP3 activation. Ubiquitination of NLRP3 by a range of E3 ubiquitin ligases, including tripartite motif containing 65 (TRIM65), and Ariadne homolog 2 (ARIH2) inhibits inflammasome activation (Kawashima et al., 2017;Tang et al., 2021). However, other E3 ubiquitin ligases such as Pellino2, TNF receptor-associated factor 6 (TRAF6), and HUWE1 were reported to facilitate NLRP3 priming (Guo et al., 2020;Humphries et al., 2018;Xing et al., 2017). These results uncover the complexity of NLRP3 ubiquitination and should be considered in future research. Our results showed that Parkin KO induces a significant shift in microglial activation markers ARG-1 and iNOS, indicating that Parkin deficiency promotes microglial neurotoxicity.
Moreover, microglial activation drives neuronal death in multiple neurodegenerative diseases, and drugs that target neuroinflammation and regulate the microglial activation state could effectively relieve symptoms and pathology. Recent research involving small molecule drug-targeting of ubiquitin-specific protease 7 showed that inhibiting deubiquitination in microglia could alleviate inflammation and attenuate PD pathogenesis (Zhang et al., 2022). Collectively, the findings suggest that the ubiquitination level plays an important role in microglial activation, and focusing on UPP degradation of inflammatory proteins may have potential therapeutic value in treating PD.

F I G U R E 6
Schematic diagram of the regulatory mechanism through which Parkin and HSP90 modulate NLRP3-associated inflammation. Microglia receive stimulation through the activation of surface receptors and then express NLRP3 and pro-inflammatory proteins. NLRP3 can either assemble into an inflammasome or degrade through the ubiquitin-proteasome pathway. Parkin is responsible for NLRP3 degradation by mediating K48-linked polyubiquitination. The combination of NLRP3 with HSP90 could prevent NLRP3 from becoming degraded. NLRP3 inflammasome activation in microglia could create a cytotoxic environment for neurons and lead to neurodegeneration in Parkinson's disease. HSP90, heat shock protein 90; NLRP3, NOD-, LRR-and pyrin domain-containing 3.
The specific inhibitor of HSP90, 17-AAG, is a mature product that is already in phase III clinical trials for cancer treatment (Pillai et al., 2020). Recent studies reported that the protein quality control function of HSP90 participates extensively in PD pathogenesis (Pratt et al., 2015), as multiple pathological proteins involved in PD, including α-synuclein, are HSP90 clients (Burmann et al., 2020). The increased HSP90 expression in PD is positively correlated with α-synuclein aggregation, and inhibition of HSP90 attenuates NLRP3 inflammasome activation (Nizami et al., 2021;Uryu et al., 2006). Our results revealed that NLRP3 also acts as an HSP90 client and further support the value of HSP90 inhibition in treating PD at the anti-inflammatory level; they also highlight the importance of molecular chaperones in regulating Parkin function and PD pathogenesis. However, the effects of HSP90 inhibitors in PD remain controversial, with some studies showing that HSP90 inhibition could accelerate α-synuclein aggregation (Bohush et al., 2019). More investigations are needed to evaluate the functions of two HSP90 isoforms, HSP90α and HSP90β, in PD pathogenesis. This research has some major limitations. First, the mouse model we used knocked out Parkin in all kinds of cells and not in microglia conditionally, and we therefore could not rule out the involvement of other nerve cells in this process. Second, we did not induce Parkin overexpression in microglia of Parkin KO mice; therefore, further studies should explore the therapeutic role of Parkin supplementation. Third, an inhibition of MCC950 should be used in vivo to certify causal relationship between NLRP3 inflammasome activation and neurodegeneration. Finally, the HSP90 inhibition experiments were only performed in vitro, and in vivo investigations are required for confirmation.

| CON CLUS ION
Ultimately, our study revealed that Parkin regulates microglial NLRP3 degradation and protects neurons by mediating microglial activation.

ACK N OWLED G M ENTS
Thanks to the helpful staff members from the Core Facilities of Zhejiang University School of Medicine for their guidance on instrument operation. We would like to thank Editage (www.edita ge.cn) for English language editing.

FU N D I N GI N FO R M ATI O N
This study was supported by the National Natural Science Foundation of China (No. 82271268, No. 82271444, and No. 82001346), and the Key Research and Development Program of Zhejiang Province (No. 2020C03020).

CO N FLI C TO FI NTE R E S TS TATE M E NT
All authors claim that there are no conflicts of interest.

DATAAVA I L A B I LIT YS TATE M E NT
The datasets used and/or analyzed during the current study are available from the corresponding author on reasonable request.