ALS-Causing Protein TDP-43 Impairs SREBP2 Cholesterol Regulation

Dyslipidemia is correlated with the prognosis of amyotrophic lateral sclerosis (ALS), a fatal motor neuron disorder characterized pathologically by TAR DNA binding protein (TDP-43) inclusions. We investigated molecular mechanisms of lipid metabolism regulated by TDP-43 in ALS. Expression microarray and RNA deep sequencing (RNA-Seq) were performed using cell lines expressing doxycycline-inductive TDP-43. Gene expression and transcriptome proling identied 434 signicantly altered genes under high levels of TDP-43 (Tukey’s test, P < 0.05) including sterol regulatory element-binding protein 2 (SREBP2), a master regulator of cholesterol homeostasis and its downstream genes. TDP-43 overexpression impaired SREBP2 transcriptional activity, leading to inhibition of cholesterol biosynthesis via the nutrient- and growth factor-responsive kinase mTOR Complex I (mTORC1). The amount of cholesterol was signicantly decreased in motor neuronal cultures derived from ALS-patient induced pluripotent stem cells (iPSCs) as well as in the spinal uids of ALS patients and in the spinal cord of ALS model mice. Impairment of cholesterol synthesis is caused by an increase in toxic function of TDP-43 in ALS. and that Spinocerebellar Ataxia type 2 model mice harboring TDP-43 pathology exhibit concomitant cholesterol biogenesis suppression(37). Our observations suggest that TDP-43 inhibited the maintenance of the cellular cholesterol level by SREBP2, and that this impaired cholesterol metabolism, which is presumably involved in other neurodegenerative diseases including Alzheimer’s disease (AD) and Huntington’s disease(16, 38, 39), is crucial in the ALS pathophysiology. A clinical study reported that there was a signicant correlation between the CSF level of 24S-hydroxycholesterol and total tau protein in AD patients(40). CYP27A1 associated with cholesterol metabolism has been identied as a susceptible gene for sporadic ALS, and CYP27A1 mutation leading to cerebrotendinous xanthomatosis manifests as an upper motor neuron symptom(41). We found a modest yet signicant difference in total and free cholesterol in the spinal cord between control and ALS model mice, but only in female mice. A previous study revealed that male mice die of bowel obstruction due to gut dysfunction(33), which could have masked the metabolic phenotypes in our study, and with this the cause of death. induction. A, A heat TDP-43-overexpressed the microarray Rex stably doxycycline (Dox)-induced DAP-TDP-43 (Dox on (+, N = 3) vs. off (-, N = 3) condition, fold change (F.C.) < -1.2). B, QRT-PCR showed that mRNA levels of target genes of SREBP2 were decreased under TDP-43-overexpressed condition. *P < 0.05, **P < 0.01, t-test, error bars = s.d. For each independent experiment, N = 3. C, Deep RNA sequencing analysis (each N = 3) of genes related to cholesterol metabolism. Tukey, *P value < 0.05. under Dox on or off condition.


Introduction
Amyotrophic lateral sclerosis (ALS) is a fatal neurodegenerative disease characterized by selective motor neuronal cell death with cytosolic aggregates(1) mainly consisting of TAR-DNA binding protein of 43 kDa (TDP-43) (2,3). Clinical studies have shown that dyslipidemia and presymptomatic body fat are associated with the risk and prognosis of ALS patients (4)(5)(6)(7). High serum levels of cholesterol, or hyperlipidemia, have been reported to be protective against the prognosis of ALS(8- 11), and the presence of hypolipidemia precedes disease progression in ALS model mice (12,13). In addition, glycolysis upregulation was reported to be protective against ALS progression in a Drosophila model of TDP-43 proteinopathy (14). Furthermore, the onset of ALS in patients with antecedent hyperlipidemia is delayed by 6 months (15), and dyslipidemia is presumed to be involved in other neurodegenerative diseases such as Huntington's disease(16).
Past studies that analyzed motor neurons (MNs) derived from induced pluripotent stem cells (iPSCs) generated from familial ALS patients with TDP-43 mutation (17,18) have reported that TDP-43 was elevated in both RNA and protein levels and that the expression of genes involved in sterol biosynthesis is decreased in ALS MNs with ALS-related mutant TDP-43. Cellular models of TDP-43 proteinopathy exhibited TDP-43 mislocalization in mitochondria and disruption of mitochondrial complex I assembly, suggesting that TDP-43 may play a role in cholesterol metabolisms in ALS (19,20). In the current study, we analyzed cholesterol metabolisms and associated gene expression pro les regulated by TDP-43 using cellular and mouse models and patient samples.

Construction of doxycycline-inducible cell lines
To establish a cell line stably expressing DAP-TDP-43 upon doxycycline application, we used an Flp-In T-Rex Expression System (Invitrogen). Flp-In T-Rex 293 cells (293T Rex) were transfected with pOG44 (Invitrogen) and DAP-TDP-43 pcDNA5/FRT/TO and cultured for 48 hours in culture medium containing 200 µg/ml of hygromycin B (Invitrogen) (19).

Gene expression analysis
One hundred ng of total RNA was processed by using the Ambion WT Expression Kit (Life Technologies, Carlsbad, CA, USA) and WT Terminal Labeling and Controls Kit (Affymetrix, Santa Clara, CA, USA) according to the manufacturers' instructions. The sample was then hybridized onto GeneChip Human Transcriptome Array 2.0 (Affymetrix). After washing and staining, the microarray was scanned by GeneChip Scanner 3000 7G (Affymetrix). The sample was re-hybridized onto a GeneChip Human Gene 1.0 ST Array (Affymetrix), and the microarray was washed, stained and scanned. Signal data were analyzed by using Affymetrix Transcriptome Analysis Console software (Affymetrix) for the Human Transcriptome Array and Partek Genomic Suite software (Partek Inc., St. Louis, MO, USA). IPA analysis and gene set analysis for genes found with a fold-change greater than 1.2 were performed using Benjamini Hochberg FDR (p < 0.05). The data were analyzed by Genespring GX software (Agilent Technologies, La Jolla, CA, USA) for the gene set analysis.

RNA sequence analysis (RNA-seq)
We extracted total RNA using the RNeasy Plus Kit (QIAGEN, Hilden, Germany). After the depletion of ribosomal RNA by Ribo-Zero Gold (Illumina, San Diego, CA, USA), we prepared libraries using the Illumina TruSeq Stranded Total RNA Sample Prep Kit (Illumina). The libraries were sequenced in the 100 cycle Single-Read mode of HiSeq2500 (Illumina). All sequence reads were extracted in FASTQ format using BCL2FASTQ Conversion Software 1.8.4 in the CASAVA 1.8.4 pipeline. The number of sequence reads is listed in Additional le 1: Table S4. The sequence reads were mapped to hg19 reference genes, downloaded on 10 December 2012, using Tophat v2.0.8b, and quanti ed by RPKMforGenes, downloaded on 19 October 2012. Gene Ontology analysis was performed by GOstats and GOdb v2.14.0 in R package 3.1.0. To estimate transcript variants, Partek Genomics Suite v6.6 with Gencode v19 reference annotation was used. Mapping was performed using TopHat (CCB at JHU, USA).
We used a human control iPSC line (HPS0063) and a familial ALS iPSC line with TDP-43 mutation M337V (A3416). Several days after the transduction, the broblasts were harvested and replated on an SNL feeder layer. On the next day, the medium was changed to primate embryonic stem cell medium (Reprocell, Kanagawa, Japan) supplemented with 4 ng/ml bFGF (Wako Chemicals, Osaka, Japan). The medium was changed every other day. Thirty days after the transduction, iPSC colonies were gathered.

Immunoblots
Differentiated cells were treated with cytosine arabinoside (AraC) (Sigma-Aldrich). Three days after the removal of AraC, the cells were harvested and lysed in TS buffer (50 mM Tris-HCl buffer, pH 7.5, 0.15 M NaCl, 5 mM ethylenediaminetetraacetic acid, 5 mM ethylene glycol bis (β-aminoethyl ether)-N,N,N,Ntetraacetic acid, and protease inhibitor cocktail (Roche, Basel, Switzerland)] containing 1% Triton X-100 on ice for 10 minutes. After sonication, samples were centrifuged at 10,000 x g for 15 minutes at 4°C. The supernatant was saved as the soluble fraction, and the pellet was resuspended, sonicated in 2% SDS buffer and saved as the insoluble fraction. Each 10-µg sample of protein was subjected to SDS-PAGE (10-20% polyacrylamide gels, BIO CRAFT, Tokyo, Japan), with or without 2-ME, and separated proteins were transferred to PVDF. The membranes were incubated with primary antibodies, followed by appropriate secondary antibodies, and then visualized using ECL plus or ECL chemiluminescence (GE Healthcare, Chicago, IL, USA). The images were acquired on an LAS 4000 (GE Healthcare). The following primary antibodies were used in this assay: TDP-43 (Protein Tech, 1:1,000), β-actin (Sigma-Aldrich

Mice
We purchased Prp-TDP43A315T mice (B6Cg-Tg (Prnp-TARDBP*A315T) 95Balo/J, 010700) from The Jackson Laboratory (location?). As described previously (21), the asymptomatic stage is between 1-2 months of age. The onset of gait disorder appeared at ∼3 months of age in transgenic mutant TDP-43 mice (symptomatic stage) and increased slightly over the next several months. The animals became paralyzed at ∼5 months of age, corresponding to the end stage (average including males and females); they were unable to move their hindlimbs or right themselves when placed on their backs. For cholesterol measurements, the mice were perfused with cold PBS. All surgical procedures were performed according to the rules set by the Ethics Committee of Kyoto University.

Human spinal uid samples
For disease diagnosis, we obtained spinal uid samples by lumbar puncture; initial pressure, cell count, total protein level, and glucose level were measured. The remaining samples were frozen at 80℃ until further investigation. We collected spinal uids from ALS patients (N = 20) and control patients (N = 20).

Cholesterol measurement
Cells were homogenized in PBS, and the lipids were extracted by modi ed Bligh and Dyer extraction method for the quantitative analysis of cholesterol (22,23). Brie y, 100 µl of chloroform plus methanol (2:1) was added to 100 µl of PBS including homogenized cells after counting the cell number. After centrifugation, the chloroform layer was collected and dried with a centrifugal dryer. The dry matter was lysed in 100 µl of ethanol to measure free cholesterol using a Cholesterol Assay kit (Cayman Chemical Company, Ann Arbor, MI, USA). Fluorescence signals were read with an Envision Multilabel Reader (PerkinElmer). Total lipids including sterols in spinal uids from patients were extracted by the same method, with the extracts being subjected to LC-MS/MS-based quanti cation (Agilent 6400). Brie y, 100 µl of chloroform plus methanol (2:1) was added to 10 µl of spinal uids mixed with 90 µl of PBS. After centrifugation, the chloroform layer was collected and dried with a centrifugal dryer. The dry matter was lysed in 100 µl of ethanol and ltered with a 0.2-µm centrifugal lter tube, and measured for cholesterol by using Agilent 6400.

Statistics
All data are shown as mean ± s.e.m. or ± s.d. The comparison of two groups was analyzed using unpaired two-tailed Student's t-test or paired t-test. One-way ANOVA was performed for each comparison followed by Tukey's post hoc tests for the evaluation of pair-wise group differences. A P value < 0.05 was considered statistically signi cant. Analyses were performed by JMP (SAS Institute Inc., Cary, NC, USA) and Excel Tokei (Social Survey Research Information, Tokyo, Japan).

Gene expression and transcriptome pro ling under high levels of TDP-43
A past study using 293T Rex cells stably expressing doxycycline-inductive DAP-TDP-43 reported that mislocalized TDP-43 bound to L-strand transfer RNA of mitochondria and mRNA coding respiratory complex I in mitochondria, disrupting complex I assembly (19). To further investigate the pathogenic role of TDP-43 in 293T Rex cells, we performed gene expression and transcriptome pro ling under an overexpressed condition of TDP-43 using expression microarray and RNA deep sequencing. Microarray showed that elevated TDP-43 downregulated the expression of SREBP2 and its downstream genes including HMGCS1, HMGCR and LDLR with signi cantly decreased cholesterol metabolic process as a gene ontology (GO) term ( Fig. 1A and Additional le 1: Table S1). SREBP2 was a top upstream transcriptional regulator of the altered gene expressions caused by TDP-43 (Additional le 1: Table S2). QRT-PCR showed that TDP-43 reduced the mRNA levels of downstream genes of SREBP2 by altering SREBP2 mRNA levels (Fig. 1B). We further performed RNAseq and identi ed 434 genes that were signi cantly altered under doxycycline-inductive TDP-43 (Tukey's test, P < 0.05, Additional le 1: Table  S3). Among them, elevated TDP-43 signi cantly decreased the expression levels of HMGCS1, HMGCR, DHCR7 and SREBP2 without biasing the splicing variants of SREBP2 (Fig. 1C-E). Taken together, TDP-43 could regulate SREBP2 by regulating its mRNA expression level.

TDP-43 decreases cholesterol biogenesis by inhibiting SREBP2 activity
Next, we investigated whether TDP-43 could affect the level of SREBP2 protein and its function. We overexpressed or knocked-down TDP-43 in the DAP-TDP-43 293T cell line using small interfering RNA (siRNA). The overexpression of TDP-43 signi cantly reduced the mRNA level of SREBP2 ( Fig. 2A, B), and western blotting analysis revealed that TDP-43 persistently decreased the amount of the cleaved N-active form of SREBP2 (N-SREBP2) in a TDP-43 expression-dependent manner (Fig. 2C, D). SREBP2 is a critical transcriptional factor that regulates various enzymes in the cholesterol biosynthetic pathway by binding sterol-regulatory elements (SREs) in the promoters of cholesterol-related genes (24,25). In accordance with the suppressive effect on SREBP2 quantity, TDP-43, through overexpression and knockdown, negatively regulated the endogenous transcriptional activity of SREBP2 as measured with a transfected SRE-luciferase construct (SRE-luc) (Fig. 2E, F). Elevated TDP-43 could inhibit SRE sensitivity to decreased cholesterol level in a fashion similar to authentic sterol-regulated conditions such as cholesterol excess by cholesterol addition or cholesterol depletion by the addition of methyl-beta-cyclodextrin (MbCD) and lovastatin, two HMGCoA reductase inhibitors (Fig. 2G). Subsequently, TDP-43 altered the amount of cholesterol (Fig. 2H).

TDP-43 regulates SREBP2 activity via the mTORC1 signaling pathway
We next analyzed the effect of the regulatory mechanism of TDP-43 on SREBP2 expression. SREBP2 expression is regulated by mTOR Complex 1 (mTORC1) mainly consisting of mTOR, Raptor(26-28), an inhibitor of autophagy, via p70S6 kinase1 (p70S6K), and Lipin-1 (29,30), since a past study demonstrated that TDP-43 could regulate the autophagy-lysosome pathway by mTORC1 activity (31). The mTORC1 pathway is in uenced by adenosine triphosphate (ATP) concentration (32). Therefore, we tested the mTORC1 contribution to the TDP-43 regulation of SREBP2 expression as well as the contribution of the activation-status of mTORC1 and its mediators to SREBP2 expression, p70S6K and Lipin-1. We examined the expression level of mTORC1 under doxycycline-induced TDP-43 and found that overexpressed TDP-43 inhibited the expression level of mTOR, Raptor and Lipin-1 48 hours after the TDP-43 induction (Fig. 3A-C). The transcriptional activity of SREBP2 was inhibited by mTOR knockdown or by the mTOR inhibitor rapamycin, but these inhibitory effects were nulli ed by TDP-43 (Fig. 3D). Finally, an activator of mTORC1, the MEK inhibitor dorsomorphin, rescued the inhibitory effect of TDP-43 on SREBP2 transcriptional activity (Fig. 3E). These results indicated that the inhibitory effect of TDP-43 on SREBP2 is mediated by the mTORC1 signaling pathway.

Decreased cholesterol levels in spinal cord tissue of ALS model mice, ALS-patient iPSC-derived motor neurons and cerebrospinal uids in ALS patients
Next, we investigated the cholesterol level of spinal cord tissue from ALS model mice overexpressing A315T mutant TDP-43 (N = 10)(33). Whole spinal cord tissues from transgenic mice or littermate control mice were fractionated to measure the amount of cholesterol per unit weight. The amounts of both total and free cholesterol were signi cantly decreased in the spinal cord tissue of transgenic female mice compared to their littermate controls at the pre-symptomatic stage (Fig. 4A), suggesting that TDP-43 could impair cholesterol biosynthesis in vivo. To analyze the effect of TDP-43 regulation on cholesterol homeostasis in human motor neurons, we employed ALS motor neuronal cultures derived from patient iPSC lines with ALS-related TDP-43 mutation M337V (17). As previously reported, the amount of TDP-43 was increased in motor neuronal cultures differentiated from ALS patient iPSCs with TDP-43 mutation compared to control (Fig. 4B). Further, the expression levels of SREBP2 and mTOR were decreased (Fig. 4B), as also were the mRNA levels of SREBP2 and downstream genes (Fig. 4C). The amounts of cholesterol were also decreased in ALS motor neuronal cultures compared to control (Fig. 4D), suggesting that ALS-related mutant TDP-43 decreased cholesterol biogenesis regulated by SREBP2 in human motor neurons. To reveal aberrant cholesterol metabolism in sporadic ALS pathophysiology, we examined the amount of cholesterol in the cerebrospinal uid (CSF) of sporadic ALS patients (N = 20) and diseasecontrol patients (N = 20) with similar characteristics ( Table 1). The cholesterol amounts were signi cantly decreased in CSF of ALS patients, but not in their serum (Fig. 4E and Tables 2, 3). De novo synthesis is responsible for almost all cholesterol in the central nervous system (34). CSF is in indirect contact with interstitial uids and re ects biochemical changes in the central nervous system (35), suggesting that the amount of cholesterol in CSF should re ect lipid metabolism in the central nervous system. Collectively, cholesterol metabolism was impaired by TDP-43 in ALS.

Discussion
By translating ALS clinical ndings reversely into basic research analysis, we found that TDP-43 protein inhibited the expression of SREBP2 and its transcriptional activity through the mTORC1 signaling pathway, suggesting a potential mechanism for neurodegenerative diseases.
Sterol regulation maintaining cellular cholesterol is critical for cell growth and survival. In general, the SREBP2 pathway is robustly activated during sterol depletion. It has been reported that the level of HMG-CoA reductase is elevated in ALS spinal cord grey matter, that the SREBP2 expression level is declined in ALS model mice with mutant SOD1(36), and that Spinocerebellar Ataxia type 2 model mice harboring TDP-43 pathology exhibit concomitant cholesterol biogenesis suppression (37). Our observations suggest that TDP-43 inhibited the maintenance of the cellular cholesterol level by SREBP2, and that this impaired cholesterol metabolism, which is presumably involved in other neurodegenerative diseases including Alzheimer's disease (AD) and Huntington's disease (16,38,39), is crucial in the ALS pathophysiology. A clinical study reported that there was a signi cant correlation between the CSF level of 24Shydroxycholesterol and total tau protein in AD patients (40). CYP27A1 associated with cholesterol metabolism has been identi ed as a susceptible gene for sporadic ALS, and CYP27A1 mutation leading to cerebrotendinous xanthomatosis manifests as an upper motor neuron symptom (41).
We found a modest yet signi cant difference in total and free cholesterol in the spinal cord between control and ALS model mice, but only in female mice. A previous study revealed that male mice die of bowel obstruction due to gut dysfunction (33), which could have masked the metabolic phenotypes in our study, and with this the cause of death.
We performed a solvent extraction method to measure the amount of free cholesterol in the CSF of ALS patients, nding major metabolites such as 24S-hydroxycholesterol in the brain, which was reported to decrease in the CSF of ALS patients by another study using a solid-phase extraction method (42).
We speculate that a cholesterol complement is not enough to rescue ALS MNs. Olesoxime, which has a cholesterol-like structure, has been thought to be neuroprotective for ALS MNs (43). However, this compound failed in a phase 3 clinical trial for ALS (44), which could underlie the requirement for a concomitant therapy that compensates for cholesterol in ALS MNs. Furthermore, as shown in Huntington's disease, SREBP2 gene therapy may be promising for ALS (45). Taken together, this reversetranslational study provides a molecular basis for the future ALS therapy targeting cholesterol metabolism.

Declarations
Ethics approval and consent to participate This study was approved by the Institutional Review Boards of Kyoto University and Tokushima University, and all surgical procedures for animal studies were performed according to the rules set forth by the Ethics Committee of Kyoto University. Written informed consent was received from the participants prior to inclusion in the study. Samples from the participants were identi ed by numbers, not by names.      (N = 3). The expression level of N-SREBP2 at the indicated times (C) was dependent on the TDP-43 expression level 48 hours after Dox (1.5 μM) treatment (D). E-G, 293T Rex cells were transfected with pSyn-SRE-Luc and pRL-SV40 and lysed for the dual-luciferase assay to detect sterolregulatory element (SRE)-activity. **P < 0.01, one-way ANOVA (E), two-way ANOVA (F, G). H, Amounts of free cholesterol in 293T Rex cells (1×106) 48 hours after transfection with or without TDP-43 siRNA. *P < 0.05, **P < 0.01, two-way ANOVA, error bars = s.d. For each assay performed, N = 3.