SCD Inhibition Protects from α-Synuclein-Induced Neurotoxicity But Is Toxic to Early Neuron Cultures

Abstract Here, we report the independent discovery and validation of stearoyl-CoA desaturase (SCD) as a modulator of α-synuclein (αSyn)-induced pathology and toxicity in cell-based Parkinson’s disease (PD) models. We identified SCD as top altered gene from transcriptional profiling in primary neurons exogenously expressing αSyn with the amplified familial PD mutation 3K. Thus, we sought to further explore SCD as a therapeutic target in neurodegeneration. We report that SCD inhibitors are toxic to early human and rat neuron cultures while displaying minimal toxicity to late cultures. The fatty acid product of SCD, oleic acid (OLA), fully rescues this toxicity in early cultures, suggesting on-target toxicity. Furthermore, SCD inhibition rescues αSyn 3K-induced toxicity in late primary neurons. We also confirm that SCD inhibitors reduce formation of αSyn accumulations, while OLA increases these accumulations in an αSyn 3K neuroblastoma model. However, we identify a caveat with this model where αSyn 3K levels can be suppressed by high SCD inhibitor concentrations, obscuring true effect size. Further, we show that both SCD1 or SCD5 knock-down reduce αSyn 3K accumulations and toxicity, making both a putative drug target. Overall, we confirm key findings of published data on SCD inhibition and its benefits in αSyn accumulation and stress models. The differential neurotoxicity induced by SCD inhibition based on neuron culture age must be accounted for when researching SCD in neuron models and has potential clinical implications. Lastly, our gene profiling studies also revealed novel putative genes connected to αSyn neurotoxicity that are worth further study.


Introduction
As the population becomes increasingly aged it is ever more important to develop therapeutics for age-associated neurodegenerative diseases such as Parkinson's disease (PD). PD is pathologically characterized by neuronal inclusions, called Lewy bodies (LBs) and Lewy neurites (LNs), and neuronal loss, preferentially of dopaminergic neurons in the substantia nigra pars compacta. LB/LNs are mainly composed of aggregated a-synuclein (aSyn) and lipid membranes (Fares et al., 2021). Reduction of these hallmark structures are considered a therapeutic approach for a disease modifying therapy for PD and other synucleinopathies. Recent studies have shown that inhibition of stearoyl-CoA desaturase (SCD) ameliorates aSyn-related pathology and neurotoxicity in both in vitro and in vivo models (Vincent et al., 2018;Fanning et al., 2019;Imberdis et al., 2019;Maulik et al., 2019;Terry-Kantor et al., 2020;Nuber et al., 2021). SCD is the rate limiting enzyme in the production of monounsaturated fatty acids (MUFAs), such as oleic acid (OLA). aSyn can directly bind OLA (Sharon et al., 2001) and the presence of OLA in membranes can enhance their interaction with aSyn (Kubo et al., 2005). aSyn's membrane binding properties are believed to play a functional role in synaptic vesicle trafficking and neurotransmission (Runwal and Edwards, 2021). These aSyn-membrane interactions can alter aSyn aggregation kinetics based on lipid membrane composition (Galvagnion et al., 2016). Recent studies have implicated OLA in promoting aSyn-lipid accumulation pathology Imberdis et al., 2019). Altering cellular lipid profiles, and more specifically lowering MUFAs such as OLA through SCD inhibition, may thus be a viable therapeutic approach to target aSyn pathology in PD. Yumanity Therapeutics is testing a SCD inhibitor in clinical trials for PD and has another SCD inhibitor in their pipeline for LB dementia (Yumanity Therapeutics, 2021). Scientists from Yumanity initially discovered SCD inhibition as protective from aSyn-induced toxicity in yeast (Vincent et al., 2018).
Many of the studies demonstrating benefits of SCD inhibition in PD used an "amplified" familial aSyn E46K model, which harbors E35K 1 E46K 1 E61K mutations (aSyn 3K). This model is advantageous in its ability to recapitulate features of human PD pathology (Ericsson et al., 2021). Transgenic mice with aSyn 3K display dopaminergic neurodegeneration and develop PD like motor symptoms that are responsive to L-Dopa treatment (Nuber et al., 2018). LBs are not only loaded with aSyn but are full of membranous lipids and vesicular structures (Shahmoradian et al., 2019). Transgenic mice and cellular models that express aSyn 3K recapitulate this lipid rich vesicular clustering during accumulation formation and exhibit cytotoxicity (Dettmer et al., 2017;Nuber et al., 2018;Terry-Kantor et al., 2020). E35K, E46K, and E61K aSyn mutations alone and in combination have been shown to have altered lipid binding properties (Rovere et al., 2019), which likely contributes to their pathogenicity.
Here, we report the independent discovery of SCD as a top hit identified in a transcriptional profiling study investigating aSyn neurotoxicity. This internal discovery and the active clinical interest spurred us to validate key published findings and expand data on SCD inhibition in PD models. In addition to supporting much of the published data on SCD inhibition in cell-based models, we also identified important caveats that must be considered when studying this target in neurons and aSyn overexpressing systems.

SCD inhibitors and OLA
MF-438 (MF; EMD Millipore, 569406) and CAY10566 (CAY; Ambeed, A698046) were resuspended in DMSO and dispensed to cell cultures via a HP D300e digital dispenser. OLA BSA conjugate (O3008-5ML) and fatty acidfree BSA control (A8806) were purchased from Sigma. OLA and BSA control were diluted in culture media before addition to cell culture (50% exchange).

AAV generation and transduction
cDNAs for human aSyn, aSyn 3K (aSyn E35K 1 E46K 1 E61K) and aSyn 3K S129A were synthesized (Geneart, ThermoFisher Scientific) and subcloned into AAV entry vectors under the CAG promoter. Empty vector without insert control (EV) vectors were packaged into adeno-associated viruses (AAVs) using AAV9 capsid serotype (Packgene Biotech). Control AAV (EV) was generated using entry vector without cDNA insert. Purity was determined using SDS-PAGE followed by Coomassie staining. AAV genome copies were quantified by determining copy numbers of AAV2 ITRs via qRT-PCR. Cells were transduced at a multiplicity of infection (MOI) of 80K for primary cortical neuron cultures and 250K for iPSC neurons, unless stated otherwise.

Primary neuron cultures
Mouse cortical neuron cultures were isolated and homogenized from timed-pregnant C57BL/6J mouse embryos on embryonic day (E)17 according to Institutional Animal Care and Use Committee (IACUC) guidelines. Rat cortical neurons cultures were generated from timed pregnant Sprague Dawley female rats at gestational day E16. Briefly, cortex tissues were dissected on ice in HBSS without calcium or magnesium (Invitrogen, catalog #21-022-CV), carefully separated from the meninges, washed 3Â in ice-cold HBSS, and then incubated at 37°C with 0.25% Trypsin-EDTA and 1 mg/ml DNase I (Sigma, catalog #DN25-10MG) for 15 min. Cell were resuspended in Neurobasal media (Invitrogen, catalog #21103) containing 10% fetal bovine serum (FBS) and 1Â GlutaMAX (Invitrogen, catalog #35050) and filtered through a 100-mm cell strainer. Neurons were plated at a density of 16,000 (imaging) or 50,000 (CellTiter-Glo) cells per well in Corning Biocoat Poly-D-Lysine 96-well plate Cellware (ThermoFisher Scientific, catalog #356640) with black wells and clear bottom with lid; 1.5 h later, plating media were aspirated and replaced with neurobasal feeding media supplemented with 2% B27 supplement (Invitrogen, catalog #17504-044), 1Â GlutaMAX, and 1Â pen/strep (Invitrogen, catalog #15140122). SCD1 inhibitors, OLA, and AAVs were given on day in vitro (DIV)7 for early cultures and DIV18 for late cultures. OLA and AAVs were added with fresh media. Cultures were fixed (4% PFA in PBS for 15 min) for imaging or lysed for analysis 12 d after treatments.

RNA sequencing (RNA-Seq)
Rat cortical neuron cultures were treated with AAV9 EV or AAV9-aSyn 3K on DIV5. RNA was isolated 6, 12, and 19 d after AAV treatment with a RNeasy plus kit with DNase digestion. RNA quality was measured using the RNA 6000 Pico kit (Agilent, 5067-1513) on a Bioanalyzer 2100, all RNA integrity scores were !9. There were five replicates per treatment and for every time point (30 samples total). cDNA was prepared with the SMART-Seq v4 Ultra Low Input RNA kit (Takara, 634891) according to manufacturer's instructions. A total of 8 ng of input RNA with 1.6 ml of added 1:10,000 ERCC RNA (ThermoFisher, 4456740) was used for cDNA generation, followed by cDNA amplification with seven cycles of PCR using CB PCR Buffer (Takara, 638526). Amplified cDNA was quantified using the High Sensitivity DNA kit (Agilent, 5067-4626) on a Bioanalyzer 2100. Libraries were prepared from amplified cDNA using the Nextera XT DNA Library Prep kit (Illumina, FC-131-1096) according to manufacturer's instructions; 150 pg of cDNA was used as input, 12 cycles of PCR were used to amplify tagmented cDNA with Nextera Index kit Set A (Illumina, FC-131-2001), and library cleanup was performed as recommended for 300-to 500-bp libraries. Libraries were quantified and QCed on the Lab Chip GX DNA HS Chip (PerkinElmer, CLS760672). Libraries were normalized to 10 nM, pooled, and sequenced on a HiSeq 2500 in paired end mode 2 Â 50 bp to an average depth of 8.8 million reads per sample.
RNA-Seq read data were analyzed using a pipeline that used STAR (version 2.5.2b; Dobin et al., 2013) for read alignment and RSEM (version 1.2.31; Li and Dewey, 2011) for transcript read count quantification on the genome (Rattus norvegicus.Rnor_6.0.89) with aSyn 3K vector added. All samples passed quality checks on raw reads with FASTQC (Wingett and Andrews, 2018) and read alignment quality (.90% uniquely mapped paired reads). Normalization and differential expression analysis were conducted with the Bioconductor package DESeq 2 (Love et al., 2014) using a statistical model formulated with time and treatment factors. Sample expression variation was quantified using PCA on gene TPM data which showed sample clustering per factor (time, treatment, no outliers). Thresholds of false discovery rate (FDR) , 0.05 were consistently applied to identify differentially expressed genes. To identify differential expression changes consistent across aSyn 3K treatments and across time points, a rank-based meta-analysis was performed with Rankprod (Del Carratore et al., 2017). Rat genes were mapped to the mouse genome according to the orthologous gene mapping from Ensembl biomart for analysis. Pathway enrichment analysis was performed using Fisher exact tests using the Reactome database. RNA-Seq data were deposited at the Gene Expression Omnibus accession #GSE172385 Neuroblastoma M17 aSyn 3K-GFP model Human full-length aSyn 3K with C-terminal GFP (aSyn 3K-GFP) was cloned into pLVX-TetOne-Puro lentiviral vector (Clontech 631849). aSyn 3K-GFP lentivirus was produced in 293T cells by co-transfecting the lentiviral plasmid and lentiviral packaging mix (Virapower, Thermo Scientific) using Lipofectamine 2000 (Invitrogen 11668-500). Stable inducible BE(2)-M17 cells were generated by three consecutive rounds of lentiviral spinfection and selection in complete growth medium containing 1 mg/ml puromycin (Invitrogen A1113803). After third round of transduction, cells were selected in complete growth medium containing 10 mg/ml puromycin and single-cell clones were generated by limiting dilution. aSyn 3K-GFP expression and accumulation formation was induced by addition of 2 mg/ml doxycycline to the culture medium. Multiple single-cell clones were assessed by counting the GFP positive cells and then quantifying the percentage of cells with accumulations. This was accomplished by high-content imaging using the Opera Phenix. The line with the most accumulations was chosen for subsequent assays.
aSyn 3K-GFP accumulations were assessed after 48-h induction with doxycycline (2 mg/ml) by counting the number of cells that were positive, giving a % of cells that harbored accumulations. This was accomplished with live imaging using PerkinElmer Opera Phenix high-content screening system.

iPSC neuron culture
Human iPSC source was a Lymphoblastoid Cell Line ND07189 obtained from Coriell Cell Repository. The donor is a non-Hispanic/Latino, white male (46 XY). iPSCs were generated using the Epi5 Episomal iPSC Reprogramming kit (ThermoFisher Scientific). iPSCs were confirmed to be (1) grossly karyotypically normal (46 XY); (2) pluripotent; (3) myoplasma and human pathogen-free (except for EBV1). Neurons were differentiated from 7189L-NGN2-H2 iPSCs which carry a tet-inducible NGN2 expression cassette targeted into the AAVS1 safe harbor locus. Three days after doxycycline treatment to induce NGN2 expression and differentiation, neurons were frozen.
Previously frozen NGN2 neurons were partially thawed in 37°C water bath. The exterior of vial was sterilized with 70% EtOH, then placed inside the TC hood. Cells were transferred to sterile 50-ml tubes, and slowly resuspended with 5 ml of room temperature NB/B27 plating media [Neurobasal:DMEM/F12 (1:1 vol) 1 200 mM ascorbic acid 1 1 mM dbcAMP 1 10 ng/ml BDNF 1 10 ng/ml GDNF 1 1Â pen/ strep 1 1 mg/ml doxycycline 1 CultureOne]. Cells were spun down at 800 rpm for 4 min at room temperature, and cell pellet was gently resuspended in NB/B27 plating media. Cells were counted on Vi-Cell and plated on PDL-coated microplates. To minimize edge effect, microplates were left for 30 min inside the TC hood to allow cells to attach, before placing inside the 37°C/CO 2 TC incubator. Doxycycline treatment continued until DIV7, at which point half the media were exchanged with [Neurobasal:DMEM/F12 (1:1 vol) 1 200 mM ascorbic acid 1 1 mM dbcAMP 1 20 ng/ml BDNF 1 20 ng/ml GDNF 1 1Â pen/strep].
SCD inhibitor, AAV, and or OLA treatments began on DIV7 and DIV21 for early and late cultures, respectively.

Cell viability measurements
Cytotoxicity in neuroblastoma M17 aSyn 3K-GFP model was assessed by caspase 3 and 7 activity via a commercial kit (Promega, G8090). Before lysis and caspase measurement cell counts were taken by staining with Hoechst and Opera Phenix high-content imaging. Caspase activity was normalized to cell counts. Viability in rodent cortical neurons and in NGN2 neurons was assessed by CellTiter-Glo (Promega, G7571). In addition, neuronal and astrocytic viability in rat cortical neurons was determined after fixation, permeabilization (0.1% Triton X-100 for 20 min), and staining with MAP2 (Abcam, ab92434) and GFAP (Abcam, ab7260) by neuron and astrocyte counting in the Opera Phenix high-content imaging system. MAP2 and GFAP were visualized by Alexa Fluor 647 and 488 secondary antibodies.

qRT-PCR
mRNA isolation was conducted one of two ways. For primary and iPSC neurons, the Cells-to-CT kit was used for mRNA isolation and cDNA synthesis (TaqMan Fast Advanced Cells-to-CT kit, A35374). For neuroblastoma cells, the Cells-to-CT kit was used or the combination of the RNeasy Plus Micro kit (QIAGEN, 74134) with iScript Reverse Transcription Supermix for RT-qPCR (Bio-Rad, 1708840) for mRNA isolation and cDNA synthesis. qRT-PCR was then performed using the TaqMan Fast Advanced Master Mix (ThermoFisher Scientific, 4444556) with QuantStudio 7 Pro Real-Time PCR System.
For qRT-PCR, target genes were normalized and multiplexed with HPRT gene expression (neuroblastoma and iPSC neurons) or with ACTB (rat primary neurons). Reference genes used VIC and target genes FAM dye signal.

Human tissue
Frozen postmortem amygdala tissues of a non-demented control donors (female, 70 years of age) and a donor with clinical diagnosis of PD (male, 87 years of age, Braak stage 6) were obtained from the Netherlands Brain Bank (NBB; Netherlands Institute for Neuroscience, Amsterdam, open access; www.brainbank.nl). Written informed consent for the use of the samples for research purposes and clinical information of donors was obtained by the NBB.

Western blotting and dot blot analysis
Cells were lysed in RIPA buffer with protease and phosphatase inhibitors. Lysates were protein normalized using BCA assay. Where indicated proteins were sequentially extracted from tissue or cells in high salt (750 mM), 1% Triton X-100 buffer and 1% SDS buffer as described previously (Weihofen et al., 2019). For Western blotting, proteins were then resolved on Bolt gels (4-12%) under reducing conditions using MES SDS running buffer system, followed by transfer to PVDF membrane with the iBlot 2 Gel Transfer system. For dot blot analysis, total extracts were spotted onto 0.45-mm nitrocellulose membranes (normalized for total aSyn expression). Membranes were blocked in either 5% milk in TBS (0.05% Tween) or LICOR blocking buffer (0.05% Tween 20) and probed with primary antibodies diluted in respective blocking buffer. Bound antibodies were detected by enhanced chemiluminescence or LICOR infrared imaging system.

Statistical analysis
Analyses were performed with GraphPad Prism 9. Significant level was defined as p , 0.05. Intergroup differences were tested using two-tailed t test, one-way ANOVA or two-way ANOVA. Multicomparisons were corrected with Dunnett's post hoc test when comparing three or more groups (for details, see Extended Data Fig.  1-2). Data were displayed as boxplot or line charts showing error bars with SDs. Statistical analyses for RNA-Seq data are described in its own section above.

Results
To reveal mechanistic insights into aSyn neurotoxicity, we performed gene expression profiling on neurons overexpressing cytotoxic aSyn. To do so, we first overexpressed human wild-type aSyn (aSyn WT) and aSyn 3K in rat primary cortical neuron cultures using AAV9. Equivalent overexpression was confirmed by Western blotting for aSyn WT and 3K (Fig. 1A). Cell viability was assessed by measuring ATP levels and was significantly reduced by ;40% 19 d after AAV9 aSyn 3K compared with EV AAV9 transduction (Fig. 1B). A clear trend toward cytotoxicity was observed after 12 but not 6 d with aSyn 3K (Fig. 1B). In contrast, only minor or no cytotoxicity was observed with aSyn WT in this system (Fig. 1B). Dose-dependent and time-dependent cytotoxicity was observed in rat and mouse cortical neuron cultures, where mouse cultures showed greater sensitivity to aSyn 3K stress (Extended Data Fig. 1-1). Of note, we found aSyn 3K to be hyperphosphorylated at S129 in comparison to aSyn WT, and mutation of the phosphorylation site (S129A) did not reduce cytotoxicity (Fig. 1A,  B). This suggests that phosphorylation at S129 is not required for the cytotoxic properties of aSyn 3K.
Having found aSyn 3K to elicit robust toxicity in neuron cultures we then performed RNA-Seq on rat cortical neuron cultures treated with AAV9 aSyn 3K or EV with isolations at 6, 12, and 19 d after AAV transduction. Principal component analysis of the gene expression profiling in our primary rat neuron cultures showed the greatest transcriptional changes between different culture ages (6 vs 12 vs 19 d after AAV9; Fig. 1C). Furthermore, the expression of endogenous aSyn increases in the neuron cultures through time (Fig. 1E), which has been reported in primary neuron maturation previously (Courte et al., 2020). The PCA and aSyn expression data combined suggest that our primary cultures are maturating in vitro. Separation of EV and aSyn 3K groups began at 12 d after transduction (Fig. 1E). Table 1 shows a list with the top ten significantly changed genes when including all three time points. We focused on gene hits found in all three time points to ensure robustness of the hits and to uncover pathologic changes occurring before obvious cytotoxicity was observed. Many of these hits were lipid regulatory genes, and indeed pathway analysis showed enrichment in lipid processes and fatty acid metabolism (Extended Data Table 1-2). Among these top hits was SCD1, which was downregulated by aSyn 3K (Fig. 1D). We also confirmed SCD1 downregulation by aSyn 3K by qPCR (Fig. 1F). The SCD2 isoform, which has high expression in the rodent brain, was also down regulated by aSyn 3K (Extended Data Fig. 1-1E; Kaestner et al., 1989).
Next, we wanted to assess the effects of SCD manipulation on aSyn accumulation formation and cytotoxicity. To do so, we developed a M17 human neuroblastoma cell model with doxycycline-inducible aSyn 3K-GFP expression. Similar aSyn 3K models, which display cytotoxicity and form lipid rich aSyn accumulations that are a potential early pathogenic step toward LBs, had previously been developed in the same neuroblastoma cell line (Dettmer et al., 2017;Imberdis et al., 2019;Terry-Kantor et al., 2020). To highlight differences between LBs and aSyn 3K accumulations, we show that pS129 aSyn is mostly soluble in our neuroblastoma aSyn 3K model while it is almost exclusively Triton X-100 insoluble and displays a ladder-like pattern on SDS PAGE in PD patient brain (Extended Data Fig. 2-2A,B). Similarly, no signal for oligomeric/fibrillar aSyn was detected in a dot blot analysis with aSyn 3K neuroblastoma extract (Extended Data Fig.  2-2C). For quantifying aSyn 3K-GFP accumulations we applied high-content imaging and used caspase 3/7 activation for assessing aSyn 3K-induced cytotoxicity. aSyn 3K accumulations were reduced by ;80% if cells where fixed in presence of 1% Triton X-100, showing that accumulations are not composed of fibrillar aSyn (Extended Data Fig. 2-2D,E). We then recapitulated a dose-dependent increase in accumulation formation from treatment with the SCD product OLA ( Fig. 2A,H ;Fanning et al., 2019). We also found that the formation of accumulations coincides with increased caspase 3/7 activity and that OLA treatment can further increase it (Fig. 2B).
Humans have two SCD isoforms (SCD1 and SCD5), and we tested whether their suppression influenced accumulation formation with siRNA knock-down. mRNA reduction of both isoforms was robust (;90%) and in the case of SCD1 led to a near complete loss of protein (Extended Data Fig. 2-1B-D). No antibody for measuring SCD5 protein levels is readily available. siRNA knockdown of both SCD1 and SCD5 led to a suppression of accumulations, albeit to a lesser degree by SCD5 knockdown (;20% vs 10% suppression; Fig. 2C). Of note, SCD5 is expressed ;80% less than SCD1 in our neuroblastoma model but still influenced accumulation formation (Fig. 2D). SCD1 knock-down also rescued aSyn 3K-GFP accumulation-induced toxicity as assessed by caspase 3/7 activity (Extended Data Fig. 2-1A). We also recapitulated a dose-dependent decrease in accumulations from the SCD inhibitors CAY and MF (Fig. 2E,F,H). SCD inhibition suppressed accumulations prophylactically and also therapeutically as when inhibitors were given after accumulation formation (Extended Data Fig.  2-1F). The observed decreases in accumulations at higher concentrations were much larger compared with complete knock-down of SCD1 or SCD5 by siRNA. At lower concentrations, the reduction of accumulations from SCD inhibition was reestablished by OLA treatment (Fig. 2E,F,H). Moreover, SCD inhibitors also reduced aSyn 3K-induced caspase 3/7 activation, and OLA reestablished caspase 3/7 activity too at lower concentrations (Fig. 2G).
Interestingly, at higher concentrations of SCD inhibition (1 and 10 mM) we observed that OLA did not reestablish aSyn 3K-GFP-induced accumulations or caspase 3/7 activity ( Fig. 2E-G), suggesting off-target effects. To explore this further we measured mRNA and protein levels of aSyn 3K-GFP with CAY and MF inhibition from 0.01 to 10 mM. We discovered that at !1 mM, there was a dose-dependent suppression of aSyn 3K-GFP mRNA and protein ( Fig. 3A-C), 1 mM produced a ;50% reduction and 10 mM a ;90% reduction. In contrast to exogenous aSyn 3K-GFP, endogenous aSyn was unaffected by SCD inhibition (Extended Data Fig. 3-1A). Additionally, robust siRNA knock-down of SCD1 or SCD5 did not alter aSyn 3K-GFP protein levels, suggesting it is not a specific phenomenon related to SCD inhibition (Extended Data Fig. 2-1D,E). Overall, we validate that SCD inhibition lowers aSyn 3K accumulations and reduces aSyn 3K-induced cytotoxicity but add caution to any users of this type of aSyn 3K-GFP model (or similar overexpression system) as high inhibitor concentrations suppressed exogenous mRNA which led to off-mechanism (non-MUFA related) lowering of aSyn 3K accumulations.
Next, we wanted to assess the effects of SCD manipulation in primary neuron cultures. We treated 7 DIV cultures with 0.01 and 0.1 mM SCD inhibitors and assessed viability with ATP levels 12 d after treatment. We found that DIV7 cultures showed dose-dependent cytotoxicity from both CAY and MF compounds (Fig. 4A,C), 0.01 mM led to a ;50% drop and 0.1 mM a ;75% drop in viability. Importantly, the SCD inhibitor-induced toxicity was fully rescued by treatment with OLA, suggesting the cytotoxicity is on-mechanism. The finding of SCD inhibitor-induced toxicity in primary neurons was surprising as one study showed tolerance of SCD inhibitors on primary rat neurons . SCD1 expression decreases overtime in these cultures suggesting that they become less dependent on SCD and OLA as they mature (Fig. 1D). To further explore SCD inhibitor toxicity we repeated the experiment starting at DIV18, hypothesizing more established cultures may have increased tolerance. Indeed, we found no significant SCD inhibitor-induced cytotoxicity in these late cultures (Fig. 4B,C). To confirm that neuron death was occurring in early DIV7 cultures with SCD inhibition, we repeated the viability experiment and counted neurons and astrocytes present in these cultures by staining with MAP2 and GFAP markers, respectively. We found that both neurons and astrocytes were dying from SCD inhibition to a similar degree (Fig. 5B,C). Both neuron and astrocyte populations were also rescued by OLA treatment. Overall, these data show that early (;DIV7) primary cultures are susceptible to SCD inhibition toxicity but that late (;DIV18) cultures are resistant.
Having demonstrated that SCD inhibition toxicity was dependent on primary culture age (DIV7 vs DIV18), we then wanted to assess SCD inhibition effects on aSyn-induced cytotoxicity. We treated both early and late cultures with CAY, AAV9-aSyn 3K or EV, and OLA for 12 d  Fig. 2-1F) reduce accumulations. G, Relative aSyn 3K-GFP accumulation-induced caspase activity with increasing concentration of CAY or MF with and without 100 mM OLA for 48 h (* denotes comparison between non-OLA and OLA treatments, # denotes respective inhibitor concentration vs DMSO control). Note that OLA does not reestablish the aSyn 3K accumulation-induced caspase activity under SCD inhibition !1 mM. H, Representative high-content images of aSyn 3K-GFP neuroblastoma model with induced aSyn 3K accumulations showing OLA-induced increase, SCD inhibitor-induced decrease, and OLA rescue of accumulations from SCD inhibition. Additional aSyn 3K accumulation characterization including comparison to LBs is shown in Extended Data Figure 2-2. One-way or two-way ANOVA run with Dunnett's multiple test correction, data displayed as boxplots or as line charts showing error bars with SD, *p , 0.05, **p , 0.01, ***p , 0.001. All plots n ! 3 independent experiments.
Research Article: New Research and then counted neurons and astrocytes (Fig. 5A). Early cultures were more resistant to aSyn 3K stress overall (;20% neuron loss vs ;50% in late neurons) and did not receive any protection from SCD inhibition (as observed before SCD inhibition was toxic; Fig. 5B,C). Although late neuron cultures were more susceptible to aSyn 3K toxicity, they were rescued by SCD inhibition (Fig. 5D,E). Importantly, OLA treatment ablated the SCD inhibitor rescue of aSyn 3K toxicity in late neurons and astrocytes, suggesting the rescue was on-mechanism ( Fig.  5D-F). OLA itself exacerbated neuron and astrocyte death under aSyn 3K stress in both early and late cultures, supporting its role in promoting aSyn pathology (Fig. 5B-E). Astrocytes showed similar trends to neurons where aSyn 3K toxicity was more severe in the late cultures (;60% vs ;90% astrocyte loss). Viability results with MF was equivalent to CAY (Extended Data Fig. 5-1). Taken together, these data demonstrate that the suppression of MUFAs through SCD inhibition can alleviate aSyn 3K stress in late primary neuron cultures but that early cultures receive no protection.
Notably, SCD inhibition by CAY or MF caused an increase in SCD1 mRNA and protein levels in our neuroblastoma model and our primary neuron cultures ( Fig. 3; Extended Data Fig. 3-1). The increase in mRNA and protein levels following SCD inhibition was reversed by OLA treatment in both models, suggesting it is a compensatory mechanism. SCD upregulation on its inhibition by small molecules has been observed previously (Vincent et al., 2018). These data further suggest mechanism and robust SCD inhibition by CAY and MF in our in vitro models.
Human iPSC neurons have been shown to tolerate SCD inhibition and even be protected from aSyn toxicity (Vincent et al., 2018;Fanning et al., 2019;Nuber et al., 2021). Based on our differential response to SCD inhibition in our rat primary neuron cultures, we tested early versus late human iPSC neuron cultures under SCD inhibition. Early iPSC neuron cultures started treatment on Licor western blot depicting aSyn 3K-GFP, bactin, and SCD1 protein levels from samples treated with SCD inhibitors (0.01 to 10 lM), aSyn 3K-GFP induction and SCD inhibitor treatment was 48 hours before protein isolation. B, Quantification of aSyn 3K protein levels as in A, CAY10566 and MF-438 data was equivalent and merged for this plot. C, mRNA levels of aSyn 3K-GFP at equivalent time points and treatments as in C. Note the drop in aSyn 3K protein and mRNA under SCD inhibition ! 1 lM. D, Quantification of SCD1 protein levels under increasing concentrations of SCD inhibitor, note the dose dependent increase. Please see Extended Data Figure 3-1 for further SCD mRNA and protein characterization under inhibitor treatment. One-way ANOVA run with Dunnett's multiple test correction, all data displayed as boxplots, *** p , 0.001. All plots n ! 3 independent experiments. DIV7 and late on DIV21. As before, SCD inhibitors and OLA treatment were done simultaneously, and cultures were assessed for ATP level viability 12 d later (Fig. 6A). Both early and late iPSC neuron cultures showed dosedependent SCD inhibition-induced cytotoxicity. Similar to our primary neuron cultures, the late iPSC neuron cultures were more resistant to SCD inhibitor-induced toxicity. Late iPSC cultures were resistant up to 0.1 versus 0.01 mM for early cultures (Fig. 6B). For example, at 0.1 mM, early cultures had ;20% less viability where late cultures had negligible loss on average. OLA rescued SCD inhibition toxicity in the early iPSC neuron cultures but appeared to have minimal effect in the older cultures.
Having established that iPSC human neurons display different sensitivity to SCD inhibition-induced toxicity based on in vitro age, we then compared early and late iPSC neuron cultures under 0.1 mM (a concentration that was non-toxic to late but toxic to early) under aSyn 3K stress. Even at higher AAV titers than that used in the primary neuron cultures (MOI 250K vs 80K), we only observed between 5% and 20% ATP loss in the iPSC neurons cultures with late cultures appearing more sensitive to 3K stress. Despite the minimal aSyn 3K-indued neurotoxicity, 0.1 mM SCD inhibition could rescue viability in late iPSC neuron cultures (Fig. 6C). OLA treatment reversed this rescue, suggesting on-mechanism protection. CellTiter-Glo analysis on primary rat cortical neuron cultures after 12-d treatment 6 SCD inhibitors (0.01 or 0.1 mM) 6 OLA (100 mM) starting at DIV7 (A, early) or DIV18 (B, late). C, Representative high-content images of both early and late cultures as in A, B with OLA, CAY, or both CAY and OLA. Note the SCD inhibitor induced toxicity in early cultures that is rescued by OLA, while there is no toxicity in the late cultures. Two-way ANOVA run with Dunnett's multiple test correction, all data displayed as boxplots, ***p , 0.001. All plots n ! 3 independent experiments.
As expected, because of their toxicity, SCD inhibitors exacerbated aSyn 3K-induced toxicity in early iPSC neurons (Fig. 6D). Viability results with MF were equivalent to CAY (Extended Data Fig. 6-1). These data demonstrate that the differential response, based on in vitro age, to SCD inhibition-induced toxicity also occurs in human neurons and that non-toxic concentrations can protect neurons from aSyn stress.

Discussion
We report that a significant factor in primary neuronal susceptibility to toxicity induced by SCD inhibition is their maturity or in vitro age. Our data suggest that there is early dependence on MUFAs and OLA that reduces with culture maturation. This is supported by our finding of early neuron and astrocyte populations being susceptible to SCD inhibition-induced toxicity that can be rescued by OLA treatment (Figs. 4A, 5B,C). Late neurons and astrocytes display no significant susceptibility to SCD inhibition toxicity (Figs. 4B, 5D,E), presumably because they are less reliant on MUFA synthesis and OLA. Further supporting this differential requirement of MUFAs is that the expression of SCD1 reduces through time in primary neuron cultures, dropping by ;50% every 6 d between early and late cultures (Fig. 1D). Furthermore, SCD is a target of Figure 5. Late primary neuron cultures are protected from aSyn 3K toxicity by SCD inhibition (CAY). A, Schematic of treatment schedule in early and late cultures. Primary rat cortical neuron cultures were treated for 12-d 6 SCD inhibitors (0.01 or 0.1 mM) 6 OLA (100 mM) 1 AAV9-EV or AAV9-aSyn 3K starting at DIV7 (early) or DIV18 (late). Neurons and astrocytes were counted by staining with MAP2 and GFAP markers, respectively. B, Neuron counts from early primary cultures. C, Neuron counts from late primary cultures. D, Astrocyte counts from early cultures. E, Astrocyte counts from late cultures. F, Representative high-content images of late cultures as in B, D. Note that late cultures are rescued from aSyn 3K toxicity by SCD inhibition, which can be reversed by OLA treatment. Extended Data Figure 5-1 shows results with MF. Two-way ANOVA run with Dunnett's multiple test correction, all data displayed boxplots, *p , 0.05, **p , 0.01, ***p , 0.001. All plots n ! 3 independent experiments. interest in cancer because it is upregulated because of high lipogenic and MUFA demand (from increased metabolic activity and division; Tracz-Gaszewska and Dobrzyn, 2019). A similar mechanism may be at play in early neuron cultures that also have a high metabolic demand and require OLA for synaptic development and neurotrophic support (Medina and Tabernero, 2002;Polo-Hernández et al., 2014) . Any investigator looking to use SCD inhibitors without toxicity side effects in primary neuron models should establish when they no longer induce significant cytotoxicity or when OLA is no longer critical for their survival.
SCD inhibition, when not toxic, rescues aSyn 3K toxicity in primary cultures which can be reversed by OLA treatment (Fig. 5D). OLA treatment itself exacerbated neuron and astrocyte death in our late cultures under aSyn 3K stress (Fig. 5D,E). This demonstrates suppression of MUFAs and OLA are key to SCD inhibition-induced protection. OLA levels have specifically been found to be upregulated by excess aSyn and to exacerbate aSyn stress . The down regulation of SCD1 and SCD2 ( Fig. 1D; Extended Data Fig. 1-1E) under aSyn 3K stress in primary cultures is likely a compensatory process to lower MUFAs and OLA, where significant SCD inhibition accomplishes this and provides protection (as with CAY and MF SCD inhibitors). Conversely, SCD1 was upregulated in our aSyn 3K neuroblastoma model (Fig.  2D). A transformed cell line likely processes aSyn 3K stress differently than primary cultures. In either model, the inhibition of SCD and MUFAs is beneficial.
Astrocytes in both the early and late primary cultures were more susceptible to aSyn 3K stress than neurons, this is clear in the early cultures where most death is in the astrocyte population (Fig. 5A,B). OLA synthesized by astrocytes has been shown to be important for neurotrophic support and synaptic development (Medina and Tabernero, 2002;Polo-Hernández et al., 2014). Astrocytes are also reported to have enhanced local translation of SCD and other fatty acid synthesis genes in their peripheral processes (Sakers et al., 2017). It is possible that astrocyte death occurs first from SCD inhibition and promotes neuron death from the loss of OLA and neurotrophic support in early cultures. Mature neuron death from aSyn 3K stress may similarly be Figure 6. Late human iPSC neurons are protected from aSyn 3K toxicity by SCD inhibition (CAY) and are less sensitive to SCD inhibition toxicity. A, Schematic of treatment schedule in early and late cultures. B, CellTiter-Glo analysis on human iPSC neuron cultures after 12-d treatment 6 CAY (0.01, 0.1, 1, or 10 mM) 6 OLA (10 mM) starting at DIV7 (early) or DIV21 (late). C, D, Same as B but only with 0.1 mM CAY and with AAV9-EV or AAV9-aSyn 3K for 12-d in late (C) or early (D) cultures. Note the rescue of viability in late iPSC neuron cultures by 0.1 mM CAY (a concentration non-toxic to late neurons but toxic to early neurons). Extended Data Figure 6-1 shows results with MF, and Extended Data Figure 6-2 shows SCD single-cell expression in human substantia nigra. Two-way ANOVA run with Dunnett's multiple test correction, all data displayed as boxplots, *p , 0.05, **p , 0.01, ***p , 0.001, ns = not significant. All plots n ! 3 independent experiments. promoted by the dysfunction and death of the astrocyte population. This is not without precedent as, when present, astrocytes can protect dopaminergic neurons from aSyn spreading and aggregation (Tsunemi et al., 2020). Furthermore, it has been shown that excess aSyn increases OLA, triglycerides, and their storage in lipid droplets in neurons . Lipid droplets can sequester the build-up of fatty acids, however, neurons generally do not make lipid droplets where astrocytes have an enhanced ability for droplet formation and fatty acid mitigation (Bélanger and Magistretti, 2009). In fact, astrocytes have been shown to protect neurons from toxic fatty acid build-up with detoxification and consumption through B-oxidation after their delivery from neurons (Ioannou et al., 2019). The interplay of astrocytes and neurons under aSyn 3K stress should be explored further. Additionally, the aSyn 3K model may be useful in the study of synucleinopathies with glial focused pathology such as multiple system atrophy.
Our results with human iPSC neuron cultures were similar to that of our primary rat neuron cultures. Older iPSC neuron cultures were more resistant to SCD inhibition-induced toxicity but did show some sensitivity (Fig. 6B) in contrast to late primary rat neuron cultures which displayed no significant toxicity at the same concentrations. One possible explanation is the that the primary rat neuron cultures are mixed and contain glia, where astrocytes and potentially microglia exert a protective effect. Another being the different SCD isoforms that are present in rodents versus humans, which is discussed below. Additionally, toleration of SCD inhibition has been shown in human iPSC neurons previously but it is unclear at what DIV inhibitors were administered (Vincent et al., 2018;Fanning et al., 2019;Nuber et al., 2021). Presumably in these studies SCD inhibitors were given in primary or iPSC neuron cultures that were matured sufficiently where MUFAs are no longer required for development and survival. To our knowledge this is the first-time showing iPSC neurons are protected specifically from aSyn 3K stress with non-toxic concentrations of SCD inhibitors. Researchers must consider culture age carefully when designing experimental paradigms with SCD inhibitors with in vitro iPSC neuron cultures.
Our independently generated aSyn 3K-GFP accumulation forming cell line confirmed many published results. However, our model showed that high inhibitor concentrations (!1 mM) suppressed the expression and protein levels of aSyn 3K-GFP in an off-target manner (Fig. 3). In a similar neuroblastoma model it was found that aSyn 3K protein levels were unchanged up to 10 mM SCD inhibitor concentrations, although the CAY inhibitor showed a trend of protein lowering . It is important to note that the IC50s of CAY and MF are ;0.005-0.01 mM in cell culture (Liu et al., 2007;Uto, 2016) and much higher concentrations are not significantly blocking SCD activity further. Moreover, maximal SCD inhibition at 0.1 mM and robust siRNA knock-down of SCD1 both produced ;25% reduction in accumulations (Fig. 2). Together this suggests that maximal on target suppression of accumulations by SCD inhibition is ;25%. Interestingly, a 25% reduction in accumulations fully rescued caspase 3/7 activity, suggesting a threshold for accumulation-induced cytotoxicity. Thus, there may be a smaller true effect of MUFA down regulation (through SCD inhibition) on aSyn 3K accumulation formation than previously reported. It is also possible our model's aSyn 3K-GFP expression may be more sensitive to experimental perturbations because of expression construct design or potential parental cell line differences. Despite the issues with high inhibitor concentrations, we found similar results in the suppression of aSyn 3K accumulation formation and cytotoxicity with SCD inhibition in our neuroblastoma model.
Rats have two SCD isoforms (SCD1 and SCD2) which share ;87% sequence homology. In cell culture SCD1 and SCD2 are both inhibited by CAY (Masuda et al., 2012). Taken together, it is likely that both isoforms were blocked by SCD inhibitors in our primary rat neuron culture experiments. Human SCD1 shares ;85% sequence homology with all rodent SCD isoforms. But the second human SCD isoform, SCD5, is unique to primates, has limited homology to other isoforms, and has high expression in brain tissue (Wang et al., 2005). We examined published single-cell RNA-Seq data from the substantia nigra region in humans (Agarwal et al., 2020) and found SCD5 to be overall higher in human neurons and astrocytes (Extended Data Fig. 6-2), although SCD1 is higher in oligodendrocytes. Yumanity's SCD inhibitor compound YTX-7739, currently in clinical development, targets both SCD1 and SCD5 (Cross, 2018). We believe targeting SCD5 is a good strategy to inhibit SCD activity more effectively in the brain and not elsewhere in the body. This is important because of the connection between SCD and MUFAs to proper immune function and other potential negative side effects of broad SCD inhibition (Meingassner et al., 2013;Zhou et al., 2021).
Out of the top 10 hits in our aSyn 3K toxicity gene expression profiling study, three genes have been previously linked to PD, namely SCD, regulators of G-protein signaling (RGS4) and acyl-CoA synthetase bubblegum family member 1 (ACSBG1). Inhibitors of RGS4 have been shown to reverse D2 antagonist-induced bradykinesia in rats (Blazer et al., 2015), generating interest in their use as novel therapeutics for PD. An RNAi screen discovered that ACSBG1 regulates aSyn levels, which was confirmed in human neurons and mouse brain tissue (Rousseaux et al., 2018). Many of the transcriptional profiling hits were lipid regulatory genes and pathway analysis indeed found several lipid metabolic process enrichments (Extended Data Table 1-2). For instance, sterol regulatory element binding factor 1 (SREBF1) is a transcription factor that regulates SCD and other lipogenic genes. SREBF1 has been identified in a PD genome wide association study and a functional screen for PD (Do et al., 2011;Ivatt et al., 2014). Overall, our transcriptional profiling supports the interconnection between aSyn pathology, MUFAs, and lipids more generally. Investigations of our most significant hit LPCAT1, which is also involved in lipid metabolism, is under way. The fact that many of the top gene hits are connected to PD supports the pursuit of investigating LPCAT1 and other novel hits previously unconnected to PD and supports the utility of the aSyn 3K model.
Overall, our study supports the use of SCD inhibitors in PD to ameliorate aSyn pathology and neurotoxicity. We demonstrate differential SCD inhibition toxicity based on neuron culture age which must be considered when researching SCD in vitro and has possible clinical implications. It is also important to note that a SCD inhibitor has been assessed in a preclinical mouse model, where it did not induce neurodegeneration and in fact protected from WT and 3K-induced neuron loss . Since those who develop PD are overwhelmingly older adults, toxicity to developing or early neurons would not be a major issue; the principal concern is saving established neurons. However, the use of SCD inhibitors in younger people and children with more active neurogenesis should be carefully considered and investigated because of the concerns of neurotoxicity.