Parkin regulates neuronal lipid homeostasis through SREBP2-lipoprotein lipase pathway—implications for Parkinson’s disease

Abstract Abnormal lipid homeostasis has been observed in the brain of Parkinson’s disease (PD) patients and experimental models, although the mechanism underlying this phenomenon is unclear. Notably, previous studies have reported that the PD-linked protein Parkin functionally interacts with important lipid regulators, including Sterol Regulatory Element-Binding Proteins (SREBPs) and cluster of differentiation 36 (CD36). Here, we demonstrate a functional relationship between Parkin and lipoprotein lipase (LPL), a triglyceride lipase that is widely expressed in the brain. Using a human neuroblastoma cell line and a Parkin knockout mouse model, we demonstrate that Parkin expression level positively correlates with neuronal LPL protein level and activity. Importantly, our study identified SREBP2, a major regulator of sterol and fatty acid synthesis, as a potential mediator between Parkin and LPL. Supporting this, SREBP2 genetic ablation abolished Parkin effect on LPL expression. We further demonstrate that Parkin-LPL pathway regulates the formation of intracellular lipid droplets, and that this pathway is upregulated upon exposure to PD-linked oxidative stress induced by rotenone. Finally, we show that inhibition of either LPL or SREBP2 exacerbates rotenone-induced cell death. Taken together, our findings reveal a novel pathway linking Parkin, SREBP2 and LPL in neuronal lipid homeostasis that may be relevant to the pathogenesis of PD.


Introduction
Parkinson's disease (PD) is a prevalent neurodegenerative disease affecting >6 million individuals globally (1). Clinically, the disease is attended by a constellation of motoric deficits that progressively worsen with age, which ultimately leads to near total immobility. Although pathological changes are distributed in the PD brain (2), the principal lesion that underlies the characteristic motor phenotype of PD patients is unequivocally the loss of dopaminergic (DA) neurons in the substantia nigra pars compacta (SNpc) of the midbrain. As DA neurons are known to be highly energy-demanding and vulnerable to metabolic insults (3,4), metabolic dysregulation has been proposed as an early pathological event that precedes neurodegeneration in PD. Interestingly, several genetic risk factors for PD (e.g. SNCA, GBA and SREBF1) are associated with lipid homeostasis, suggesting a potential involvement of lipid dysregulation in PD pathogenesis (5)(6)(7)(8). Consistent with this notion, several studies have reported abnormal lipid metabolism in brain tissues from PD patients and various experimental models of PD. For example, Halliday et al. (9) reported an increase in the density of lipid-containing neuromelanin organelles in the surviving SNpc DA neurons from the brains of patients with early stage PD compared to those of healthy controls. At the cellular level, α-Synuclein-enriched Lewy bodies (LBs), the pathological hallmark of PD, are decorated with lipid-enriched, densely packed vesicular structures and dysmorphic organelles (10)(11)(12). Furthermore, α-Synuclein overexpression (O/E) or cellular exposure to PD-relevant toxins promotes the accumulation of intracellular lipid droplets (LDs) and subsequent cell death (13)(14)(15). Notably, LDs are known to interact with and regulate the function of mitochondria, the dysfunction of which is also linked to PD pathogenesis (16)(17)(18). Notwithstanding the prominent relationship between altered lipid homeostasis and PD, the underlying molecular mechanism remains not well clarified. Interestingly, Kim et al. (19) reported about a decade ago that Parkin, a ubiquitin ligase whose mutations cause recessive early-onset Parkinsonism, regulates lipid uptake in non-neuronal cells by stabilizing cluster of differentiation 36 (CD36), which is involved in transporting fatty acid (FA), via monoubiquitination. In a more recent development, Ivatt et al. (7) reported that Parkin functionally interacts with Sterol Regulatory Element-Binding Protein −1 (SREBP1) and −2 (SREBP2), which are major regulators of lipid metabolism, but in the context of mitophagy. Prompted by these observations, we were interested to pursue the role of Parkin in lipid homeostasis.
In the present study, we demonstrated by means of multiple experimental models that Parkin regulates the SREBP2lipoprotein lipase (LPL) pathway. We found that Parkin expression is positively correlated with the levels of LPL and its upstream regulator SREBP2. Moreover, CRISPR-mediated ablation of SREBP2 abolished this relationship, which links the three components in a shared pathway. Alongside this, we showed that Parkin downregulates F-box And WD Repeat Domain Containing 7 (FBXW7)-a known component of SCF Slimb E3 ubiquitin ligase complex that targets nuclear SREBPs for degradation (20). We further showed that Parkin-LPL pathway regulates the intracellular deposition of LDs. Interestingly, the Parkin-SREBP2-LPL pathway is upregulated upon cellular exposure to PD-linked oxidative stress (i.e. rotenone), which is known to elevate LD production. Inhibition of either LPL or SREBP2 disrupts this phenomenon and exacerbates rotenone-induced cell death. Taken together, our findings reveal a novel pathway implicating Parkin, SREBP2 and LPL in neuronal lipid homeostasis and LD formation in a PD-relevant context.

Parkin regulates LPL expression and activity
Given the report that Parkin modulates FA uptake in nonneuronal cells via stabilizing the expression of CD36 [a FA translocase; (19)], we were interested to examine if Parkin also exhibits a similar function in the brain. To address this, we measured the expression level of CD36 in the brains of mouse with homozygous deletion of exon 7 of PARK2 gene [herein referred to as Parkin knockout (KO) mouse; (21)] and their wild-type (WT) counterparts. In contrast to the findings by Kim et al., we found that brain CD36 protein expression level was not affected by the genetic ablation of Parkin (Supplementary Material, Fig.  S1A). Given the close functional relationship of LPL and CD36 in the regulation of FA uptake (22), we then measured LPL protein expression in Parkin WT and KO mouse brains. Interestingly, loss of Parkin results in a dramatic reduction of LPL protein amount in the Parkin KO mouse brains compared with Parkin WT brains ( Fig. 1A and Supplementary Material, Fig. S14). Moreover, LPL lipid hydrolysis assay revealed a significant decrease in LPL activity in Parkin KO mouse brain lysates compared with those of control mice (Fig. 1B). To ascertain the specificity of Parkin effect on LPL, we measured the protein expression of a related lipase, namely Endothelial Lipase (EndoL), but found no apparent difference in its expression between Parkin WT and KO mouse brain lysates ( Fig. 1C and Supplementary Material, Fig. S14). Notably, we also detected significant reductions of LPL level in the heart, muscle and liver tissues from Parkin KO mice compared with their control counterparts (Supplementary Material, Fig. S1B-D). Further experiments with primary mouse cortical neurons also showed that LPL protein expression level is markedly reduced in Parkin KO neurons ( Fig. 1D and Supplementary Material, Fig. S14).
To further investigate the effect of Parkin on LPL level, we employed SH-SY5Y neuroblastoma cell model with stable O/E of FLAG-tagged Parkin that we have previously reported (23). Immunoblotting revealed that LPL expression level is significantly elevated in Parkin O/E cells compared with control cells (Fig. 1E and Supplementary Material, Fig. S14). However, our co-immunoprecipitation experimental failed to detect any appreciable physical interaction between Parkin and LPL in SH-SY5Y cells co-expressing FLAG-Parkin and Myc-LPL (Supplementary Material, Fig. S2A and B). Immunoprecipitated Parkin was able to pull down Hsp70, a previously reported interactor (24).
Taken together, our data suggest that Parkin positively regulates LPL expression level in neuronal cells and that this functional interaction is likely to involve other player(s).

Modulation of LPL level by Parkin is SREBP2-dependent
LPL is known to be positively or negatively regulated by several molecules at the transcript or protein level (25). Therefore, we subsequently measured mRNA levels of known regulators of LPL in brain samples obtained from Parkin WT and KO mice. From this candidate screening, we found that the transcript level of sterol regulatory binding factor-2 (SREBF2) is significantly downregulated in Parkin KO brain lysates compared with those of control brains ( Fig. 2A and Supplementary Material, Fig. S3A). SREBF2 gene encodes SREBP2 protein, a key transcription factor for lipid and cholesterol biosynthesis. Notably, SREBP2 has been reported to positively regulate the expression level of LPL in preadipocyte cells (26). To further determine if Parkin affects SREBP2 activity, we quantified the mRNA levels of SREBF2 and known SREBP2 transcriptional targets, namely HMGCR and DHCR24, in SH-SY5Y cells stably O/E Parkin via RT-PCR ( Fig. 2B and Supplementary Material, Fig. S3B). Our results revealed that increase in Parkin expression significantly elevates the mRNA levels of these genes, suggesting that Parkin promotes SREBP2 downstream signaling.
nSREBP2 protein levels in this case is clearly significantly reduced in P2(−/−) and P4(−/−) clones compared with those of control and Parkin O/E parental cells (Fig. 2E), and is expected to affect the Parkin-SREBP2-LPL axis. Importantly, we found that CRISPR/ Cas9-mediated KO of FL-SREBP2 abolishes the effect of Parkin O/E on LPL levels in these edited cells (Fig. 2E). Since CRISPR-Cas9 method may cause unwanted off-target effects (32), we performed a genetic rescue experiment by exogenously expressing nSREBP2 in the SREBP2 KO clones. Reintroduction of nSREBP2 into SREBP2 KO; Parkin O/E clones restored LPL protein expression to a comparable level to that of Parkin O/E cells ( Fig. 2F and Supplementary Material, Fig. S15). Taken together, our results demonstrate that Parkin-mediated increase in LPL expression requires SREBP2, suggesting that Parkin-SREBP2-LPL are all linked in the same pathway.

Parkin regulates nSREBP2 level by downregulating FBXW7β
Transcriptional activation of SREBP2 target genes requires proteolytic cleavage of the endoplasmic reticulum (ER)-resident, FL-SREBP2 into its active form (i.e. nSREBP2) that translocates to the nucleus upon its activation (33,34). In line with the notion that Parkin promotes SREBP2 downstream signaling, our immunoblotting results revealed that Parkin O/E results in a significant increase in nSREBP2 protein level with a corresponding decrease in precursor FL-SREBP2 protein level compared with SH-SY5Y vector control cells ( Fig. 3B and Supplementary Material, Fig. S5). Given that SREBP2 signaling is tightly regulated through various mechanisms (33,34), we then investigate if Parkin increases nSREBP2 level by either: (i) promoting the cleavage of FL-SREBP2 and/or (ii) by preventing the degradation of nSREBP2. As an initial approach, we performed immunoblotting of known SREBP2 pathway regulators such as INSIG-2, Site-1-protease (S1P) and FBXW7 ( Fig. 3A and Supplementary Material, Fig. S16). We first confirmed our previous observation that Parkin O/E promotes nSREBP2 signaling as shown by the increase in the protein levels of a SREBP2 downstream target, HMGCR (Fig. 3A). Parkin O/E did not cause any significant change in the expression of INSIG-2, a protein that blocks FL-SREBPs processing (Fig. 3A). Interestingly, we found significant increase in S1P protein levels (higher MW precursor and lower MW active form) in Parkin O/E cells (Fig 3A), suggesting that Parkin may enhance nSREBP2 expression by promoting its proteolytic cleavage via S1P.
On the other hand, nSREBPs are known to be rapidly degraded by the ubiquitin-proteasome pathway via a negative-feedback loop (34). The ubiquitin ligase FBXW7 negatively regulates SREBP2 signaling by enhancing the degradation of nSREBPs upon their phosphorylation by GSK3β (20). Furthermore, Parkin was previously shown to promote the proteasomal degradation of FBXW7 isoform β (FBXW7β) via ubiquitination in neurons (35). Therefore, it is possible that Parkin indirectly promotes the stability of nSREBP2 by enhancing the proteasomal degradation of Collectively, these data demonstrate that Parkin positively inf luences nSREBP2 level by preventing its degradation via FBXW7β and, possibly, by also promoting its cleavage via S1P.

Parkin-LPL pathway regulates intracellular LDs deposition
Recent evidence suggests a strong link between intracellular LD accumulation and cell survival (17,37,38). Notably, LPL activity has been reported to modulate FA uptake and promote intracellular LD deposition (39,40). Accordingly, we performed f luorescence imaging to visualize intracellular lipid deposition via lipid-specific Nile Red dye in SH-SY5Y cells with Parkin O/E. Interestingly, our data show that Parkin O/E cells exhibit significantly lower Nile Red staining intensity compared with control cells at the basal state ( Fig. 4A and B). We also treated these cells with 200 nM rotenone, a mitochondrial complex I inhibitor that is linked to PD, to induce oxidative stress. Rotenone treatment elevated intracellular accumulation of Nile Red-positive puncta, which are reminiscent of LDs, in both control and Parkin O/E cells ( Fig. 4A and B). This observation is consistent with previous reports in which rotenone treatment induced intracellular LDs accumulation in a glia-neuron co-culture system (16,37). However, the total amount of Nile Red-positive LD-associated puncta is markedly lower in Parkin O/E cell compared with control cells, indicating that Parkin may inhibit excessive LD accumulation during oxidative stress. To test if this phenomenon is mediated by Parkin's effect on LPL, we transiently O/E myc-LPL in control SH-SY5Y cells and measured the amount of intracellular Nile Red-positive puncta. Similar to Parkin O/E cells, the amount of Nile Red-positive LDassociated puncta is significantly reduced in myc-LPL O/E cells compared with untransfected controls upon rotenone treatment ( Fig. 4C and D). Notably, we also observed a substantial colocalization of myc-LPL (∼60%) with ER structure (Supplementary Material, Fig. S9), which is an important site of LDs biogenesis (41). Collectively, these observations suggest that Parkin-LPL pathway regulates intracellular LD-associated deposition at basal condition and during PD-linked oxidative stress.
We next investigated whether Parkin regulates LD accumulation in primary cortical neurons. For this experiment, BODIPY 493/503 dye was employed to identify LDs since this dye exhibit more specificity in labelling neutral lipids (the main LD component) compared with Nile Red. Fluorescence imaging revealed that Parkin KO neurons exhibit higher number of BODIPY-positive puncta compared with Parkin WT neurons ( Fig. 4E and F). In addition, LD size was also significantly increased in Parkin KO neurons compared with WT neurons (Fig. 4G). Next, we treated both Parkin WT and Parkin KO neurons with rotenone to investigate the role of Parkin in neuronal LD formation under PD-linked oxidative stress. We used a lower dose of rotenone in primary neuron cultures (i.e. 5-100 nM) since the dose used in SH-SY5Y cells (200 nM) rapidly induced neuronal death. Our immunoblot data revealed a significant increase in the level of LD-associated protein Plin2 in primary WT neurons upon acute rotenone treatment (Supplementary Material, Fig. S10). Notably, rotenone-treated Parkin KO neurons exhibit high level of BODIPY f luorescence intensities in both intracellular and extracellular compartments (Fig. 4H). The appearance of large BODIPY-positive structures in Parkin KO neurons hampered accurate quantification of LD puncta in these cells. Thus, we quantified average BODIPY intensity from these cells as a measure of intracellular neutral lipid accumulation. This measurement revealed that Parkin genetic ablation in neurons results in a dramatic increase in intracellular lipid deposition under oxidative stress (Fig. 4I). Altogether, our findings indicate that the loss of Parkin induces aberrant lipid homeostasis and subsequent LD accumulation in neurons under basal condition and in the presence of PD-associated oxidative stress.  Data were normalized to total βIII-tubulin-positive neuronal areas. ( * * * P < 0.001 (t-test); Parkin WT + Rot (24 cells, 9 fields); Parkin KO (57 cells, 12 fields); n = 2 biological replicates). Error bars represent standard deviation between each imaging regions/fields for Fig. 4 B, D, F, G, I. Cell nuclei were stained with DAPI (blue). Scale bar = 20 μm for Fig. 4 E, H.

PD-associated oxidative stress triggers the upregulation of Parkin-SREBP2-LPL pathway in neurons
Exposure of neurons to oxidative stress, including rotenone treatment, is known to elevate Parkin expression (42,43). On the other hand, activation of SREBP due to either increase in reactive oxygen species (ROS) or mitochondrial defects has been shown to cause glial LD accumulation and neurodegeneration in Drosophila photoreceptor system (44). Along the same line, we found that the increase in LD deposition in rotenone-treated (100 nM for 24 h) primary neuron culture correlates with increase in cell death (measured by Plin2 protein and cleaved PARP level respectively; Supplementary Material, Fig. S10). Interestingly, LPL is also significantly elevated upon this high dose rotenone treatment (Supplementary Material, Fig. S10). Given our observations thus far, it is attractive to propose that neurons utilize Parkin-SREBP2-LPL pathway to regulate intracellular, lipid-based protective response against PD-relevant oxidative stress. To test this hypothesis, we treated primary cortical neurons with rotenone (10 nM or 20 nM for 20 h) and measured Parkin, LPL and nSREBP2 protein levels. We found that both Parkin and LPL levels were apparently elevated in Parkin WT neurons upon acute exposure to rotenone at 20 nM concentration ( Fig. 5A and Supplementary Material, Fig. S17), whereas the increase in nSREBP2 did not approach statistical significance (Fig. 5A). The observed increase in LPL and Parkin level in Parkin WT neurons was replicated under chronic rotenone treatment (7 days; Fig. 5B). Under this condition, nSREBP2 expression becomes significantly upregulated ( Fig. 5B and Supplementary Material, Fig. S17). In contrast, LPL and nSREBP2 levels in Parkin KO neurons remains unaffected when treated with rotenone for up to 20 h (Fig. 5A). Correlating with these observations, we observed a reduction in FBXW7β expression in rotenone-treated WT neurons but not in Parkin KO neurons (Supplementary Material, Fig. S11). Moreover, consistent with previous report (35), FBXW7β protein level was apparently upregulated in Parkin KO neurons at basal state relative to its WT counterparts, a phenomenon that persisted in the presence of rotenone treatment (Supplementary Material, Fig. S11). Taken together, our results show that rotenone treatment induces Parkin-dependent expression of neuronal LPL and nSREBP2. Thus, Parkin-SREBP2-LPL pathway can be activated during neuronal response against oxidative stress, possibly, as a cellular defence mechanism.

Inhibition of LPL or SREBP2 exacerbates cell death upon exposure to PD-associated oxidative stress
Next, we investigated if LPL inhibition is detrimental to neuronal cells under PD-linked oxidative stress. As a cancer-derived cell line, undifferentiated SH-SY5Y cells rely primarily on aerobic glycolysis (i.e. Warburg effect), instead of mitochondrial oxidative phosphorylation/OXPHOS (45). Thus, we only observed modest cell death upon rotenone treatment, even at high drug concentration (∼a few μM for 24-48 h), in these cells (data not shown). To increase the reliance of SH-SY5Y cells on mitochondria OXPHOS, we switched the growth media from glucose supplemented-to galactose-supplemented-DMEM for our subsequent experiments (46). Under this condition, we then measured the extent of rotenone-induced cell death in SH-SY5Y cells treated with an LPL inhibitor (GSK264220A, 10 μM for 16 h) via MTT assay (Fig. 6A and B). Our data show that treatment with the LPL inhibitor alone at the given doses does not affect SH-SY5Y cell or primary cortical neuron viability ( Fig. 6A and C).
However, co-treatment of GSK264220A with rotenone (50 nM for 16 h) significantly reduced cell viability compared with cells that were treated with rotenone alone (Fig. 6B). Importantly, we observed a similar exacerbation in rotenone-induced cell death in mouse primary cortical neurons that were co-treated with either GSK264220A or a CD36 inhibitor (SSO 10 μM for 16 h), indicating a potential role of LPL and its partner CD36 in PD-relevant neuronal death ( Fig. 6D and E).
To determine the potential involvement of SREBP2 signaling in Parkin-mediated neuroprotection against mitochondrial ROSinduced cell death, we then measured the cell viability of Parkin O/E SH-SY5Y cells with or without SREBP2 KO upon rotenone treatment. We first measured the extent of rotenone-induced cell death via a f luorescence imaging approach using LIVE/DEAD cell viability assay (Invitrogen™), whereby green f luorescence (calcein-AM) indicates live cells with intact intracellular esterase activity and red f luorescence (ethidium homodimer-1) indicates dead cells with loss of plasma membrane integrity. As expected from Parkin's neuroprotective function, Parkin O/E cells display a trend towards reduced cell death upon rotenone treatment (50 nM for 16 h) compared with vector control (Fig. 6F). However, loss of SREBP2 in Parkin O/E cells enhanced the extent of rotenone-induced cell death to a comparable level with that of vector control (Fig. 6F). Moreover, a similar observation was also observed using MTT-based cell viability assay (Fig. 6G), suggesting that SREBP2 is potentially involved in Parkin-mediated cellular protection against rotenone toxicity. Supporting this, we observed significant elevation of mitochondrial ROS level (measured from MitoSOX intensity) in Parkin O/E; SREBP2 KO SH-SY5Y cells compared with control and Parkin O/E parental line at basal state and upon rotenone treatment (Supplementary Material, Fig. S12). Taken together, our data supports a potential role of SREBP2-LPL pathway in Parkin-mediated cellular protection against PD-linked stress, such as mitochondrial ROS generated by rotenone.

Discussion
In this study, we found that Parkin regulates the expression and activity of LPL in neurons, an effect that is dependent on Parkin's positive regulation of SREBP2 signaling. Furthermore, we demonstrate that Parkin-SREBP2-LPL axis is activated upon cellular exposure to PD-associated oxidative stress (induced by rotenone), which apparently inf luences the abundance of LDs-cellular organelles that are thought to be involved in the intrinsic protective mechanism against cellular stress. Importantly, our data show that pharmacological inhibition of LPL exacerbates rotenone-induced cell death. Alongside this, we also demonstrate that Parkin's protective effect against rotenone-induced cell death is prevented by the genetic ablation of SREBP2. Taken together, our findings highlight a novel role of Parkin in neuronal lipid regulation that may be relevant to PD pathogenesis.
Parkin has previously been implicated in lipid metabolism by promoting the stability of the lipid transporter CD36 that regulates free FA uptake (19). Interestingly, recent metabolomic studies using blood plasma or skin fibroblasts from PD patients with PARK2 (Parkin) mutations revealed alterations in lipid profiles compared with those of healthy controls (47,48). Here, we show that Parkin positively regulates the expression level of neuronal LPL, a key triglyceride (TG) lipase that controls extracellular free FA uptake via CD36 (22). LPL has been shown to be widely expressed in the brain, including in neurons and glial cells (49).
Notably, a recent study demonstrated that circulating brain TG modulates DA transmission in the food reward-associated meso-corticolimbic circuitry via LPL (50). Given the above, it is tempting to speculate that a similar lipid modulation by LPL may be involved in nigrostriatal DA neurons responses against PDlinked neurodegenerative insults. Supporting this, we showed in this study that Parkin-dependent upregulation of LPL in neuronal cells is associated with LD homeostasis under PDassociated oxidative stress.
Although current evidence for LPL's role in neuronal survival is lacking, there is scientific evidence to support a role for SREBP signaling in neurodegeneration. In a recent study, Chali et al. (51) reported that excitotoxic neurodegeneration induced by in vivo kainate injection in mouse hippocampus acutely promotes nuclear SREBP2 translocation and cholesterol synthesis. In another study, Liu et al. (44) demonstrate that SREBP activation in photoreceptor neurons promotes LD accumulation in surrounding glial cells and subsequent neurodegeneration in Drosophila photoreceptor system in the presence of high level of ROS. In the context of PD, inhibition of Stearoyl-CoA desaturase 1 (SCD1), which is a downstream target of SREBP1, seems sufficient to suppress α-Synuclein induced-toxicity in yeast and human neurons (15,52). Interestingly, a web-based genomewide association study has identified a mutation in SREBF1 locus as a genetic risk factor for sporadic PD in a European population (6). In this study, we found that SREBP2 signaling is upregulated in primary cortical neurons upon exposure to Rotenone (Fig. 5B).
Parkin itself seems to be important in promoting nSREBP2 signaling, since Parkin O/E elevated nSREBP2 protein level in SH-SY5Y cells (Figs. 2E and 3B) and Parkin genetic ablation prevents Mevastatin-induced pharmacological activation of SREBP2 in primary neurons (Fig. 2C). Therefore, Parkin-dependent SREBP2 signaling is likely to play an important role in PD-associated neurodegeneration process. Parkin may regulate SREBP2 signaling through the following scenarios: (i) promoting maturation of FL-SREBP2 to nSREBP2; (ii) preventing degradation of nSREBP2 and/or (iii) enhancing the transcription of SREBF2 gene. For the first scenario, our data show that Parkin O/E increases the level of S1P, one of the Golgi-resident proteases that cleaves FL-SREBP2 to its nuclear form upon pathway activation (34). Notably, the protein stability of FL-SREBP2 regulators, such as SCAP and INSIGs, are regulated via ubiquitination by other E3 ligases (53)(54)(55)(56). Therefore, the possibility that Parkin directly regulates S1P protein stability via its E3 ligase activity warrants further investigation. For the second scenario, our data demonstrate that Parkin O/E downregulates the expression level of FBXW7β, a component of SCF Slimb E3 ligase complex that enhances degradation of nSREBPs upon their phosphorylation by GSK3β (20). Of note, we found that only FBXW7β, but not FBXW7α and -γ , was expressed by SH-SY5Y cells. Moreover, our immunof luorescence data demonstrate that exogeneous eGFP-FBXW7β is localized in both cytoplasm and nuclear compartment of the cells (Supplementary Material, Fig. S8). These observations suggest that FBXW7β may substitute the role of FBXW7α and -γ in these cells and this may explain the discrepancy between our data and those of Sundqvist et al study (20). The Parkin-dependent regulation of FBXW7β observed here is in an agreement with a previous study showing that Parkin promotes the proteasomal degradation of FBXW7β via ubiquitination in neurons (35). Moreover, exogenous expression of eGFP-FBXW7β (Fig. 3B) or elevation of endogenous FBXW7β due to E3-ligase-deficient Parkin T240R O/E (Supplementary Material, Fig. S7) is sufficient to negate Parkin O/E effect on SREBP2 and LPL levels. Therefore, our data support the notion that Parkin promotes nSREBP2 stability by promoting the degradation of its negative regulator FBXW7β. Finally, our data show that Parkin level positively correlates with SREBF2 transcript level ( Fig. 2A and B). Along this line, Parkin has been demonstrated to act as a transcription factor that promotes neuroprotection by downregulating p53 transcript through its DNA binding activity (57,58). Thus, future study should also investigate the possibility of Parkin acting as a transcription factor for SREBF2 gene expression.
LD-associated structures have been observed in several neurodegenerative diseases (38,59,60). In PD, lipid-containing structures have been shown to be one of the main components of LBs found in vitro (12) and PD patient brains (11). Interestingly, O/E of α-Synuclein was sufficient to trigger LD accumulation and cellular toxicity in yeast and neuronal PD models (13,15). Our findings demonstrate that rotenone-induced oxidative stress also triggered LD deposition, which was accompanied with increase in cell death ( Fig. 4 and Supplementary Material, Fig. S10). Since increase in LPL expression reduces intracellular lipid deposition during rotenone-induced stress, we posit that the activation of Parkin-SREBP2-LPL pathway might represent a novel cellular protective mechanism against PD-associated stress. Indeed, our results demonstrate that pharmacological inhibition of LPL by GSK264220A enhances the extent of rotenone-induced cell death in SH-SY5Y cells and primary cortical neuron compared to those treated with rotenone alone. Moreover, genetic ablation of SREBP2 elevates mitochondrial ROS (Supplementary Material,  Fig. S12) and abrogates cellular protection conferred by Parkin O/E upon rotenone treatment (Fig. 6F and G). Therefore, our results strongly support that activation of Parkin-SREBP2-LPL pathway in neuronal cells is important for mitigating cellular toxicity against PD-relevant stress.
Increase in LPL activity has been shown to increase extracellular free FA uptake via CD36 (22,61) and promotes LD accumulation in peripheral cells (39). In contrast, our findings revealed that Parkin-dependent increase in LPL level resulted in lower LD deposition in SH-SY5Y cells, whereas loss of Parkin leads to increased LD accumulation in primary cortical neurons at basal state and under oxidative stress. Therefore, LPL may exhibit non-canonical role in neuronal LD maintenance under certain conditions (i.e. Parkin activation or oxidative stress). Interestingly, early studies on LPL biology had reported detectable intracellular LPL activities in various study models (62,63). Consistent with this, our immunof luorescence imaging data suggest that a substantial fraction of exogenously expressed myc-LPL was colocalized with ER ( Supplementary Material, Fig. S9), the primary site of LD biosynthesis (41). Altogether, these results suggest a nonconventional role of intracellular LPL in regulating LD homeostasis in a PD-relevant context.
In summary, our study shows that Parkin promotes SREBP2 signaling by elevating basal nSREBP2 level. In turn, increase in SREBP2 activity enhances LPL expression and activity. In neuronal cells, LPL activity regulates intracellular LD deposition at basal condition and under oxidative stress. In the latter scenario, LD accumulation may initially promote cell viability by sequestering free FAs from harmful oxidation, assisting mitochondrial functions and/or provides lipid materials for autophagy process (17,38). However, excessive LD build-up and elevated peroxidated lipid level can cause detrimental effects to neurons at the later stage, such as promoting formation of insoluble α-Synuclein inclusions and inhibiting mitophagy (14,15). These events will in turn lead to toxicity cascades leading to degeneration of neurons, especially those that are vulnerable to metabolic insults, such as DA neurons (Supplementary Material, Fig. S13). Under these conditions, Parkin-SREBP2-LPL pathway is activated to suppress excessive LD formation and dysfunction of this molecular pathway may contribute to neurodegeneration seen in PD. We foresee that our study will advance our understanding on the role of dysregulated lipid homeostasis in PD pathogenesis and aid subsequent development of new therapeutic strategy against this debilitating disease.

Generation of pcDNA-myc-LPL, pEGFP-FBXW7α and pEGFP-FBXW7β constructs
Prior to cloning LPL, human cDNA library was obtained by first performing an RNA extraction from SH-SY5Y cells using the RNA isolation kit as per manufacturer protocol (Qiagen, Germany). The RNA was then reverse transcribed, by using Super-ScriptII Reverse Transcriptase (Invitrogen, CA, USA), following which, a PCR was performed using primers specific to human LPL, FBXW7α or FBXW7β cDNA. LPL cDNA was then cloned into myc-pcDNA using BamHI and XhoI restriction enzyme (NEB, USA) sites. The resulting constructs generate myc-tagged LPL at N-terminus. Meanwhile, FBXW7 cDNAs were cloned into pEGFP-Parkin WT backbone (Addgene #45875) by swapping the pre-existing Parkin cDNA with either FBXW7α or FBXW7β cDNA using EcoRI and ApaI restriction enzyme (NEB, USA) sites. The resulting constructs generate EGFP-tagged FBXW7 at N-terminus.

Parkin KO mouse models
C57BL/6J mouse model harboring deletion of exon 7 of Parkin gene was previously described (21). For the experiments, two homozygous mouse lines (i.e. Parkin (+/+) WT and Parkin (−/−) KO lines) were generated by crossing the initial heterozygous Parkin exon7/+ founder mice and maintained in homozygous genotype throughout the study. Mouse genotyping was performed according to the previous study (21). Only male mice were used for immunoblotting and LPL hydrolysis assay experiments. All mouse studies were approved and conducted according to the guidelines of the Institutional Animal Care Committee of Tan Tock Seng Hospital-National Neuroscience Institute, Singapore and NTU-LKCMedicine Animal Research Facility.

Primary mouse cortical neuron
Primary mouse cortical neuron cultures were adapted from previously described protocol (65). Brief ly, cortical tissues were dissected from Parkin WT and KO E17-E18 mice brains and their meninges were removed. The tissues were subsequently incubated with papain solution (Worthington, USA) for 45 min, washed and dissociated in DMEM with FBS. The cell suspension was spun down at 250 g for 5 min to remove excess media. Cell pellets were resuspended in Neurobasal (NB) media supplemented with 1× B27 and 1× GlutaMAX (NB/B27 media, Gibco, USA) and plated on poly-D-Lysine (PDL, Sigma-Aldrich, USA)-coated tissue culture plates (0.01 mg/ml PDL) or glass coverslips (0.1 mg/ml PDL). Primary neurons were grown NB/B27 media for 11-21 days in vitro (DIV). Additional NB/B27 media (1/3 volume) were added at DIV4 to sustain culture growth.

Preparation of cell lysates and brain homogenates
Prior to lysate collection, SH-SY5Y cells and primary neurons were brief ly rinsed with ice-cold DMEM media without FBS or PBS, respectively. Cells were lyses in SDS lysis buffer (PBS with 1% SDS, Phenylmethylsulfonyl Fluoride/PMSF (Sigma-Aldrich, Germany), Aprotinin (Sigma-Aldrich, USA), phosphatase inhibitor cocktail 2 and 3 (Sigma-Aldrich, USA) and collected in Eppendorf tubes. Whole brain tissues were harvested from Parkin WT and KO adult male mice (4-10 months) and homogenized with Dounce homogenizer inside 1% SDS lysis buffer. Both cell and whole brain lysates were sonicated (3 × 10 s, amplitude: 30-40%) and centrifuged at 13 500 rpm at RT for 15 min to remove insoluble fraction before the supernatant was collected. Protein concentrations of cell or whole brain lysates were determined by Bradford protein assay (Bio-Rad, CA, USA). Western/Immunoblotting was performed using 20-40 μg of total protein.

Immunoprecipitation
SH-SY5Y cells were harvested and subsequently subject to lysed with lysis buffer while on ice. The lysates were then centrifuged at 13 500 rpm, 15 min at 4 • C. The supernatant was collected, and a Bradford protein assay was done to determine protein concentration in the lysate. Equal amounts of protein were then used for immunoprecipitation with the desired antibodies and left to incubate at 4 • C overnight. Protein G PLUS/protein A-agarose (Calbiochem, Millipore, Merck, USA) was added the following day and left to incubate with the lysates for 2 h at 4 • C. Subsequently, the agarose beads were washed with 1% Triton X-100/PBS (Bio-Rad, CA, USA) for a total of 5 times. Proteins captured by the beads were then released by adding SDS. Immunoprecipitates and their respective cell lysates were resolved by immunoblotting, as described previously.

Reverse transcription (RT) PCR
For cell lines and primary neurons, total RNAs were extracted using Rneasy Mini Kit (Qiagen, Germany) according to manufacturer's protocol. For whole brain RNA samples, Parkin WT and KO mice brains were firstly f lash-freezed in liquid nitrogen. The brains were then homogenized using a mortar and a pestle. Thereafter, RNAs were isolated using TRIzol reagent (Invitrogen, USA) according to the instructions provided by the manufacturer. RNAs from all samples were converted to cDNA using SuperScript II First Strand Synthesis System (Invitrogen, CA, USA) according to the manufacturer's protocol. Subsequent quantifications of gene transcripts were done by PCR amplification using primers from either Integrated DNA Technologies (Singapore) or KiCqStart SYBR Green primers (Sigma-Aldrich, USA; Table 3). The PCR products were resolved using DNA agarose gel electrophoresis and visualized with XRS+ Gel Documentation system (Bio-Rad, USA).

LPL activity assay
LPL lipid hydrolysis activity in mouse brain samples were measured using LPL Activity Fluorometric Assay Kit (Biovision, USA) according to the manufacturer's protocol.

Cell viability assays
MTT (MedChemExpress, USA) (5 mg/ml in PBS) was added directly to the cell media to working concentration of 0.3 mg/ml and incubated for 3-4 h. Thereafter, the cells were lysed with pure DMSO and the lysates was transferred to clear bottom 96 well plate. MTT signal was measured at 570-nm wavelength using microplate reader (Tecan). LIVE/DEAD cell viability assay (Invitrogen, USA) was performed according to the manufacturer's protocol. Live f luorescence imaging was performed after prior incubation with calcein-AM (green f luorescence) and ethidium homodimer-1 (red f luorescence) with an epif luorescence microscope (Olympus IX70). Cell counting was performed with ImageJ.

MitoSOX live imaging experiment
SH-SY5Y control, Parkin O/E and Parkin O/E;SREBP2(−/−) cells were grown in galactose-supplemented media prior to staining with MitoSOX Red™ (Invitrogen, USA) and rotenone treatment. Cells were co-stained with 5 μM MitoSOX, Hoechst (1:1000; Enzo Life Sciences, USA) and BODIPY 493/503 (Invitrogen, USA) (1:1000) in culture media for 10 min at 37 • C. Cells were rinsed and imaged in HBSS without phenol red (Gibco, USA) supplemented with Ca 2+ and Mg 2+ . Live confocal f luorescence imaging was performed on a spinning disc confocal (SDC) setup built around a Nikon Ti2 inverted microscope equipped with a Yokogawa CSU-W1 confocal spinning head, a Plan-Apochromat objective (100×, 1.45 NA) and a back-illuminated sCMOS camera (Prime 95B; Photometrics, USA). All image acquisitions were carried out using MetaMorph (Molecular Device, USA) with exposure time 500 ms. Automated MitoSOX intensity measurements were done using MatLab algorithm (66) and is available upon request.

Fluorescence microscopy
Immunof luorescence (IF) imaging was done using paraformaldehyde (PFA )-fixed samples grown on PDL-treated glass coverslips. Brief ly, samples were washed with ice-cold PBS and fixed in 4% PFA solution (ChemCruz, TX, USA) at 4 • C overnight. Samples were then washed with PBS and permeabilized with 0.1% Triton-X solution in PBS for 5 min. Subsequently, samples were incubated with appropriate primary antibodies (Table 2) in f luorescence dilution buffer (FDB; 5% FBS, 5% goat serum (Millipore, USA) and 2% BSA (Sigma-Aldrich, USA) in PBS) at 4 • C overnight. The following day, samples were washed with PBS and incubated with f luorescence-conjugated secondary antibodies (Table 2) for 1 h at RT. For nuclei staining, samples were then incubated with DAPI solution for 5 min, washed, and mounted to glass slides with FluorSave™ mounting reagent (EMD Millipore, USA). For staining of LDs, samples were PFA-fixed, permeabilized with 0.1 mg/ml Saponin (Sigma-Aldrich, USA) and co-stained with primary-secondary antibodies as described previously. Thereafter, samples were additionally incubated with either Nile Red (Invitrogen, USA) or BODIPY493/503 (Invitrogen, USA) for 10-15 min according to published protocol (67). Both IF and dye-stained samples were visualized using Fluoview Upright Confocal Microscope (Olympus FV1000 and FV3000) using 60× or 100× objectives equipped with 405, 488, 543 and 633 nm wavelength gas state lasers. LDs were quantified using an in-house automated image analysis software (Adcount by Nair et al, unpublished. The algorithm can be accessed at https://adcount.github.io/) or Matlab algorithm (66), which are available upon request.

Statistical analysis
All data were expressed as mean with error bar (standard error of the mean) unless otherwise stated. The data was accumulated under each condition from at least two independent experiments. The datasets in Figures 1, 2, 3, 5 and 6 were normalized using 'normalization by sum of the replicate' method as it minimizes the issue related to high coefficient of variation (CV) for data with low quantified intensities when 'normalisation by fixed point' method is performed (68). The statistical analyses were performed using GraphPad Prism with student's two-tailed unpaired t-test or one-way analysis of variance (ANOVA) with Benjamini-Hochberg procedure to decrease false discovery rate (FDR). The P value of < 0.05 is considered as statistically significant.

Supplementary Material
Supplementary Material is available at HMG online.