PDZD8-deficient mice accumulate cholesteryl esters in the brain as a result of impaired lipophagy

Summary Dyslipidemia including the accumulation of cholesteryl esters (CEs) in the brain is associated with neurological disorders, although the underlying mechanism has been unclear. PDZD8, a Rab7 effector protein, transfers lipids between endoplasmic reticulum (ER) and Rab7-positive organelles and thereby promotes endolysosome maturation and contributes to the maintenance of neuronal integrity. Here we show that CEs accumulate in the brain of PDZD8-deficient mice as a result of impaired lipophagy. This CE accumulation was not affected by diet, implicating a defect in intracellular lipid metabolism. Whereas cholesterol synthesis appeared normal, degradation of lipid droplets (LDs) was defective, in the brain of PDZD8-deficient mice. PDZD8 may mediate the exchange of cholesterol and phosphatidylserine between ER and Rab7-positive organelles to promote the fusion of CE-containing LDs with lysosomes for their degradation. Our results thus suggest that PDZD8 promotes clearance of CEs from the brain by lipophagy, with this role of PDZD8 likely contributing to brain function.


INTRODUCTION
Dyslipidemia including cholesterol storage disease is associated with neurodegenerative disorders such as Alzheimer's disease, Huntington's disease, and Parkinson's disease [1][2][3][4][5] Intracellular cholesterol and its derivatives including cholesteryl esters (CEs) are either supplied by low-density lipoprotein (LDL) or synthesized in the endoplasmic reticulum (ER), and they are stored in lipid droplets (LDs). The abnormal accumulation of CEs has been attributed to the dysfunction of late endosomes and lysosomes (LEs/Lys). 6 However, the detailed mechanism underlying dyslipidemia in the brain has remained unclear.
Membrane contact sites (MCSs) of intracellular organelles are regions where different organelles are closely associated, and they play an important role in the interorganelle communication. The main functions of MCSs include lipid transfer between organelles, calcium regulation, and control of organelle dynamics. [7][8][9] In particular, MCSs between ER and LEs/Lys play a key role in mediating the intracellular distribution of cholesterol. [10][11][12][13] We have now found that PDZD8 knockout (KO) mice manifest abnormal CE accumulation in the brain that results from an impairment of lipophagy. This impairment is itself due to defective lysosomal maturation dependent on Rab7 and cholesterol. Our results thus suggest that PDZD8 plays an essential role in maintaining the integrity of brain function by regulating cholesterol metabolism.

RESULTS
Abnormal cholesteryl ester accumulation in the brain of PDZD8-deficient mice Lipidosis or abnormal lipid accumulation in the brain is associated with neurological disorders. Given that PDZD8 possesses lipid transfer activity that promotes endosome maturation and thereby maintains neuronal integrity, we investigated lipid abnormalities in the brain of PDZD8-KO mice by lipidome analysis. For this analysis, we applied supercritical fluid chromatography-triple-quadrupole mass spectrometry (SFC/QqQMS), a novel method for quantitative lipid analysis that allows the comprehensive measurement of lipid molecules with high sensitivity. The basal ganglia (BG) and cortex (Cx) of the brain were sampled ( Figure 1A). PDZD8-KO mice showed a marked abnormal accumulation specifically of CEs in the BG, with this effect being less pronounced in the Cx, compared with wild-type (WT) mice ( Figure 1B). Comparison of brain versus liver revealed abnormal CE accumulation in the BG but not in the liver of PDZD8-KO ( Figures 1C and 1D).
To investigate whether this abnormal CE accumulation in the brain of PDZD8-KO mice was affected by diet, we compared the lipidomes of mice fed either a normal diet (ND) or a high-fat diet (HFD). The KO/WT ratio for CE content showed a similar increase for the BG, but not for the liver, of mice fed either diet ( Figure 1D). Whereas both WT and PDZD8-KO mice showed marked increases in both CE and triglyceride (TG) content of the liver in response to feeding with the HFD, no such changes were apparent for the BG ( Figure 1E). The CE content of the brain thus appeared to be largely unaffected by diet.

Diet-independent accumulation of cholesteryl esters in the brain of PDZD8-KO mice
The lipidome data for the brain and liver ( Figures 1B-1E) were then analyzed in more detail. The total amount of CEs in the BG of PDZD8-KO mice showed an $3-fold increase relative to WT mice at 7 months of age ( Figure 2A). Similar increases of $4-fold and $2-fold were apparent for the BG of the mutant mice at 5 months ( Figure 2B) and 3 months ( Figure 2C) of age, respectively, whereas no differences were apparent for the liver between the two genotypes ( Figures 2B and 2C). The increase in CE content in the BG of PDZD8-KO mice relative to WT mice was not affected by feeding with the HFD compared with the ND (Figure 2C). Together, these results revealed the abnormal accumulation of CEs in the brain, in particular in the BG, of PDZD8-KO mice at all ages examined (3 to 7 months).
We next investigated why only CEs out of all lipids accumulated in the brain of PDZD8-KO mice and why such an effect was apparent in the brain but not in the liver. Reverse transcription (RT) and real-time polymerase chain reaction (PCR) analysis revealed that the abundance of PDZD8 mRNA was much higher in the brain than in the liver of WT mice ( Figure S1A), suggesting that PDZD8 function with regard to CE metabolism may be more important in the brain. We also found that the amount of CEs was markedly lower in the brain than in the liver, whereas the abundance of other lipids was higher in the brain or similar between the two organs ( Figure S1B), suggesting that the CE content of the brain might need to be maintained low by constitutive regulation. Classification of CEs according to the type of fatty acid (FA) constituent revealed that the amounts of all types of CE in the BG were higher for PDZD8-KO mice than for WT mice at 5 months of age, with CE (22:6) showing the largest absolute increase ( Figures 2D and S2A). No such large difference in the amount of any type of CE was apparent for the liver of PDZD8-KO versus WT mice ( Figure 2D). All types of CE were also more abundant in the BG of PDZD8-KO mice than in the BG of WT mice at 7 months of age, with CE(22:6) again showing the largest absolute increase ( Figures S2B and S2C). The most abundant types of CE in WT differed between the BG and the liver, with CE(22:6) and CE(20:4) being most abundant in the BG and CE(18:1) and CE(18:2) being most abundant in the liver ( Figure 2E). In the liver, unlike BG, the amount of CE(18:1) showed a large increase in response to HFD feeding compared with ND feeding, with no substantial difference being apparent between WT and PDZD8-KO animals ( Figure 2E). Whereas lipid content in the liver is well known to be affected by diet, the insensitivity of that in the brain to food type has not previously been described. Our results thus suggested that the abnormal CE accumulation in the brain of PDZD8-KO mice is not dependent on cellular uptake of diet-derived LDL, but rather is related to either excessive synthesis or impaired metabolism of CE-containing LDs within cells. iScience Article Defective lipophagy in the brain of PDZD8-KO mice We then examined whether PDZD8 might play a role in cholesterol synthesis. Cholesterol is synthesized and modified in ER, with cholesterol-related gene expression in the nucleus being regulated by a feedback system ( Figure S3). Hydroxymethylglutaryl-CoA reductase (HMGCR) catalyzes the conversion of HMG-CoA to mevalonate in the cholesterol biosynthetic pathway, and cholesterol is subsequently converted to A B C D E Figure 1. Abnormal CE accumulation in the brain of PDZD8-deficient mice (A) Schematic representation of mouse brain regions subjected to lipidome analysis. (B-D) Heat maps of lipid amount ratios for PDZD8-KO relative to WT mice as determined by lipidome analysis. The ratios are shown according to the indicated color scale (B, left). The tissues compared in each analysis were obtained from the same mice, but the mice fed the ND or the HFD in (D) were different. (B) BG and Cx for three mice at 7 months (M) of age. (C) BG and liver from two mice at 5 months of age, with the ratios for each type of CE being shown to the right of the heat map. (D) BG and liver for three ND-or HFD-fed mice at 3 months of age (HFD feeding for 1 month).
(E) Heat maps of lipid amount ratios for HFD-fed mice relative to ND-fed mice determined by lipidome analysis. The ratios are shown according to the indicated color scale (left) and were determined for the mice analyzed in (D iScience Article derivatives such as CEs and oxysterol (24-OHC) by acyl-CoA:cholesterol acyltransferase 1 (ACAT1) and cholesterol 24-hydroxylase (CYP46A1), respectively. Intracellular cholesterol inhibits the expression of the gene for SREBP2, a transcription factor that regulates expression of genes including those for HMGCR and the LDL receptor (LDLR). However, the abundance of mRNAs for SREBP2, HMGCR, LDLR, ACAT1, and CYP46A1 in the BG did not differ between PDZD8-KO and WT mice ( Figure 3A).
Next, in order to examine the degradation of CE-containing LDs in the brain, we attempted to identify brain regions in which PDZD8 is highly expressed. The brain of WT mice was divided into the following regions: the anterior portion of the BG (a), the posterior portion of the BG (b), the hindbrain (c), and the Cx (d) (Figure 3B). RT and real-time PCR analysis of regional marker gene expression confirmed the appropriate dissection of these brain regions ( Figure 3C). Similar analysis of the same samples revealed that the abundance of PDZD8 mRNA was similar in all four brain regions ( Figure 3D), suggesting that the accumulation of CEs to a greater extent in the BG than in the Cx of PDZD8-KO mice was not due simply to the expression level of PDZD8, with the presence or absence of a functionally complementary molecule possibly playing a role.
We also examined the regional expression of PDZD8 in the brain by immunohistofluorescence staining. This analysis detected PDZD8 in the striatum, medial habenula (MHb), amygdala, ventral tegmental area (VTA)/substantia nigra pars reticulata (SNr), and trigeminal mesencephalic nucleus (Vme) ( Figures 3E  and 3F). Among these regions, despite its small size, the MHb was the most suitable for examination by transmission electron microscopy (TEM), given that PDZD8 was highly expressed in almost all its component cells ( Figure 3F). The specificity of the PDZD8 antibody signal in the MHb was confirmed by its loss in PDZD8-KO mice, and the PDZD8-expressing cells were positive for acetylcholine transferase (AChT) ( Figure 3G). We then examined lysosomes and LDs, or lipophagy, in MHb neurons by TEM. LDs (which appeared white) seemed to make contact or to undergo fusion with lysosomes (which appeared black) in PDZD8-KO and WT neurons, respectively ( Figure 3H). Whereas most LDs in WT neurons appeared to undergo degradation, as suggested by an overall gray coloration, segmentation, and indistinct boundaries with lysosomes, those in PDZD8-KO neurons seemed not to undergo degradation, remaining distinct and intact ( Figures 3H-3J). These results suggested that the abnormal accumulation of CEs in the brain of PDZD8-KO mice may result from a failure of LD degradation due to insufficient lipophagy.

PDZD8 possesses phosphatidylserine and cholesterol transfer activity
We previously showed that PDZD8 possesses lipid extraction activity with donor liposomes but not lipid insertion activity with acceptor liposomes. 16 We here modified our in vitro fluorescence resonance energy transfer (FRET)-based liposome assay to examine further this activity of PDZD8. In the new version of the assay, PDZD8 was anchored to the donor membrane in order to increase the efficiency of lipid extraction by including in the donor liposomes DGS-NTA(Ni), a lipid conjugated to nitrilotriacetic acid (Ni 2+ salt) that allows the docking of hexahistidine (His 6 )-tagged recombinant proteins, 38 and the measurement time was extended from 300 to 1800 s ( Figure S4A). In this assay, donor liposomes containing rhodamine-labeled phosphatidylethanolamine (PE) and nitrobenzoxadiazole (NBD)-labeled phospholipid give rise to FRET that is abrogated by lipid transfer ( Figure S4A). NBD fluorescence would be expected to saturate rapidly if the test protein mediates only lipid extraction, whereas it would be expected to continue to increase if the protein mediates lipid transfer (including both lipid extraction and insertion). Although phosphatidic acid (PA), phosphatidylcholine (PC), and phosphatidylethanolamine (PE) were only extracted from donor liposomes, phosphatidylserine (PS) was both extracted from and inserted into liposomes by His 6 -tagged PDZD8(DTM), a mutant form of PDZD8 lacking the transmembrane domain ( Figures 4A, 4B, and S4B). In this setting, however, it would be possible for PS to be transferred in both directions, given that the donor and acceptor liposomes have the same base lipid composition, including PC and PE ( Figure S4A). The PS transfer activity of various His 6 -tagged forms of PDZD8 was calculated by subtracting the assay value obtained with His 6 -tagged glutathione S-transferase (GST) as a negative control from that obtained with each Figure 2. Diet-independent dyslipidemia in the brain of PDZD8-deficient mice (A-C) Lipid amount ratios (PDZD8-KO/WT) for lipid classes in the BG and Cx as in Figure 1B, in the BG and liver as in Figure 1C, or in the BG and liver of ND-or HFD-fed mice as in Figure 1D, respectively. The ratios were determined as the mean + SD in each case.
(D and E) Amount of each type of CE in the BG and liver of WT and PDZD8-KO mice as in Figure 1C or in Figure 1D Cholesterol-binding proteins, such as GRAMD1s, are able to access cholesterol in a PS-rich membrane domain. 39 Thus, we predicted that the SMP domain of PDZD8 might also be more accessible to cholesterol in the PS-rich membrane domain ( Figure 4E). We therefore next investigated the potential cholesterol transfer activity of PDZD8 by further modifying the liposome-FRET assay to include PS in the donor liposomes ( Figure 4F). Comparison of fluorescence values obtained with His 6 -PDZD8(DTM) in the absence or presence of acceptor liposomes revealed that the fluorescence saturated rapidly in the former instance and continued to increase in the latter, suggesting that PDZD8 possesses both cholesterol extraction and cholesterol insertion activity-that is, cholesterol transfer activity ( Figure 4G). Cholesterol transfer is unidirectional in this system as the base lipids of the donor and acceptor liposomes have different compositions ( Figure 4F). Examination of the cholesterol transfer activity of various His 6 -tagged PDZD8 mutants revealed that such activity was markedly reduced by the deletion of the SMP domain and completely lost after the deletion of both the SMP and PDZ domains ( Figures 4D and 4H). In contrast, SMP and SMP + PDZ fragments showed high activity, almost equal to that of PDZD8(DTM). These results thus implicated the SMP domain as the region responsible for the cholesterol transfer activity of PDZD8, with the PDZ domain playing a supporting role.
PDZD8 is an ER-resident protein and its C1 domain binds to phosphatidylinositol 4-phosphate [PI(4)P], 16 which is abundant in the endosome membrane. 40 We, therefore, examined the possible role of the C1 domain and PI(4)P in cholesterol transfer by PDZD8. The activity of PDZD8(DTM) was markedly increased by the addition of PI(4)P to acceptor liposomes, whereas it was attenuated by the deletion of C1 (Figure S4C), suggesting that the association of the C1 domain of PDZD8 in the ER with PI(4)P in endosomes facilitates cholesterol transfer by PDZD8.
These findings and previous observations [16][17][18][19] provide the basis for a model of the mechanism underlying PS and cholesterol transfer by PDZD8 ( Figure 4I). The ER-resident protein PDZD8 thus tethers ER to Rab7-positive organelles through the binding of its CC domain to Rab7. PDZD8 may then transfer PS from Rab7-positive organelles to ER as well as cholesterol from PS-rich domains of the ER membrane to the Rab7-positive organelles in a manner dependent on its SMP domain ( Figure 4I).

PDZD8 regulates the subcellular localization of phosphatidylserine and cholesterol
To validate this model of lipid transfer by PDZD8, we next examined the subcellular distribution of PS and cholesterol in HeLa cells depleted of PDZD8 by transfection with a small interfering RNA (siRNA). The PS probes Lactadherin-C2 and Evectin2-PH have different targeting properties, detecting PS predominantly in the plasma membrane and in endosomes, respectively. 41, 42 We, therefore, examined PS dynamics between ER and Rab7-positive organelles with Evectin2-PH. We found that overlap of the ER marker mCherry-KDEL and green fluorescent protein (GFP)-tagged Evectin2-PH was lower in PDZD8depleted cells than in control cells ( Figures 5A and 5B). In addition, the overlap of mCherry-Rab7a(Q67L) and GFP-Evectin2-PH was higher in PDZD8-depleted cells than in control cells ( Figures 5C and 5D). We next examined the localization of enhanced GFP (EGFP)-tagged KDEL and the cholesterol probe filipin in cells transfected with control or PDZD8 siRNAs. Overlap of EGFP-KDEL and filipin was higher in the PDZD8-depleted cells than in the control cells ( Figures 5E and 5F). The cholesterol probe D4H can detect both cholesterol and its derivatives, [43][44][45] and we detected mCherry-D4H signals within endolysosomes of HeLa cells in the steady state ( Figure S5A). The localization of EGFP-Rab7a(Q67L), filipin, and mCherry-D4H overlapped considerably in control cells but not in PDZD8-depleted cells ( Figure 5G). Statistical analysis with Pearson's correlation coefficient showed that the overlap of Rab7a(Q67L) with either filipin or D4H was significantly lower in PDZD8-depleted cells than in control cells ( Figures 5H  and 5I). These results were thus consistent with the notion that PDZD8 transfers cholesterol from ER to Rab7-positive organelles. Together, our data thus provided support for the model proposed in Figure 4I, showing that PDZD8 promotes the exchange of PS and cholesterol between ER and Rab7-positive organelles. However, given that it is possible that PDZD8 transfers lipids in both directions between ER and Rab7-positive organelles, the amounts of lipids we detected in organelle membranes might reflect the net outcome of bidirectional transfer. In other words, at ER-endosome MCSs, at which PDZD8 is localized, cholesterol transport from ER to the Rab7-positive organelles may be preferentially enhanced.
Localization of the cholesterol probes filipin and mCherry-D4H overlapped substantially in control cells but hardly at all in PDZD8-depleted cells ( Figures 5G, 5J, and 5K). Whereas mCherry-D4H showed an endolysosome-like distribution in control cells, it showed an ectopic distribution in the plasma membrane and intracellular fibrous structures as well as localization to endolysosome-like organelles in PDZD8-depleted cells ( Figures 5K, 5L, and S5B). In addition, mCherry-D4H and EGFP-KDEL showed little overlap in either control or PDZD8-depleted cells ( Figure S5C). These results suggested that PDZD8 transfers cholesterol from ER to Rab7-positive organelles and that such transfer might in turn promote the incorporation of CEs into endolysosomes.
PDZD8 promotes fusion of Rab7-positive organelles and D4H-positive organelles Cholesterol promotes membrane fusion between Rab7-positive organelles (endolysosomes) and autophagosomes. 33 Furthermore, active Rab7 facilitates the incorporation of LDs into lysosomes by lipophagy. 34 PDZD8, which binds to active Rab7, might therefore play a role in cholesterol-dependent membrane fusion of Rab7-positive organelles and in lipophagy. Overexpression of PDZD8 in HeLa cells resulted in the occasional appearance of abnormal organelles consisting of double-or triple-layered structures with membranes positive for mCherry-D4H, EGFP-Rab7a, or both markers ( Figures 6A and 6B). The outermost layer of the multilayered organelles was positive for both mCherry-D4H and EGFP-Rab7a and appeared to surround a D4H-positive organelle, consistent with the presence of an LD within a lysosome ( Figure 6C). These abnormal structures, which were detected only in PDZD8-overexpressing cells, might result from a delay in the degradation of excessively incorporated LDs within lysosomes, with such excessive incorporation possibly being caused by the presence of an excessive amount of cholesterol in the lysosome membrane. These results thus indicated that PDZD8 facilitates the incorporation of LDs by and their fusion with lysosomes-that is, PDZD8 promotes LD degradation by lipophagy. We then examined lipophagy activity with the use of EGFP-tagged perilipin 2 (PLIN2), a marker of LDs. PC12 rat pheochromocytoma cells were transfected with control or PDZD8 siRNAs as well as with an expression vector for EGFP-PLIN2 and were also labeled with the lysosome marker LysoTracker Red. Fluorescence microscopic analysis revealed a significant aggregation of LDs in the PDZD8-depleted cells ( Figures 7A, 7B, and S6A). Quantitative analysis revealed that, whereas most control cells did not contain LD aggregates, LDs in $70% of PDZD8-depleted cells were present in aggregates of two or more droplets ( Figure 7C). The number of LDs per aggregate in individual PDZD8-depleted cells varied widely, with some aggregates containing >10 LDs ( Figure 7D). The overlap of EGFP-PLIN2 and LysoTracker Red was also significantly reduced in PDZD8-depleted cells compared with control cells ( Figure 7E). Furthermore, the size of LDs was significantly greater in PDZD8-depleted cells than in control cells ( Figure 7F).
Endosomal maturation involves the conversion of LEs to lysosomes as well as lysosomal activation. 12,30,32 Lysosome activity is dependent on a reduction in internal pH and can be measured with LysoTracker Red, which fluoresces maximally at low pH ($pH 4-5). Lysosome activity was markedly reduced in PDZD8depleted PC12 cells compared with control cells (Figures 7A, 7G, and S6B), consistent with PDZD8 promoting endosome maturation. 16 We then examined lysosomal activity during lipophagy with the use of monomeric red fluorescent protein (mRFP)-EGFP-PLIN2, with the fluorescence of mRFP being stable and that of EGFP being reduced on exposure to the low pH of the lysosome lumen. 46 The EGFP/mRFP fluorescence intensity ratio for tagged PLIN2 within lysosomes is thus inversely correlated with the activity of lysosomes and lipophagy ( Figure 7H). PC12 cells transfected with control or PDZD8 siRNAs as well as with expression vectors for an enhanced blue fluorescent protein (EBFP)-tagged form of the lysosome marker protein LAMP1 and for mRFP-EGFP-PLIN2 were analyzed for the EGFP/mRFP ratio per EBFP-positive organelle. The ratio was found to be significantly increased in PDZD8-depleted PC12 cells compared with control cells (Figures 7I and 7J). These results thus suggested that PDZD8 contributes to the progression of lipophagy by promoting lysosome maturation.

DISCUSSION
We have here revealed the abnormal accumulation of CEs in the brain of PDZD8-KO mice and characterized its underlying mechanism. PDZD8 transfers cholesterol from ER to LEs/Lys and thereby promotes endosome maturation, leading to lysosome maturation and fusion with LDs and consequent CE degradation by lipophagy. CE accumulation in the PDZD8-KO brain is therefore the result of defective CE degradation as a consequence of impaired lipophagy ( Figure S7). Our study thus provides insight into how the lipid transfer activity of PDZD8 might contribute to brain function and indicates that cholesterol storage disease is due to a defect in lipophagy. iScience Article Although our previous study suggested that PDZD8 possesses cholesterol extraction activity, 16 with an improved liposome-FRET assay we now show that PDZD8 actually mediates PS and cholesterol transfer, including both extraction and insertion-that is, it possesses PS and cholesterol exchange transfer activity. However, NBD is a bulky tag and NBD-cholesterol may possible to change the physical property of cholesterol. Our present results then could be further confirmed by lipid transfer assays using dehydroergosterol (DHE), a naturally occurring fluorescent sterol analog. 47 Intracellular cholesterol is transferred between several organelles including between ER and LEs/Lys as well as between ER and the plasma membrane through the action of various LTPs. In addition to PDZD8, LTPs such as NPC1/2, ORP1L, and STARD3 transfer cholesterol between ER and LEs/Lys. [48][49][50] How these proteins might share the task of cholesterol transfer remains unknown, but PDZD8 and ORP1L are similar in that they both bind to Rab7 as well as to VAP.
LDs consist of a core of lipid esters surrounded by a single layer of phospholipids and cholesterol at the surface. 51 The lipid esters in adipocytes consist mostly of TG, whereas those in nonadipocytes comprise mostly CEs. 52 We found that only CEs manifest abnormal accumulation in the PDZD8-deficient brain, with the amounts of other lipids such as TG and FAs being unaffected. Specific accumulation of CEs has also been detected in induced pluripotent stem cell-derived neurons from individuals with Alzheimer's disease, in Trem2-or ApoE-deficient glia, and in the brain of individuals with Huntington's disease 3,5,53 Furthermore, the accumulation of CEs in Trem2-or ApoE-deficient glial cells was enhanced by clozapine, a selective inhibitor of the mesolimbic dopaminergic system. 4,5 Such association between specific accumulation of CEs in the central nervous system and neurological disease suggests that CEs must be maintained at low levels to ensure normal brain function. However, the reason for the accumulation of only CE but not of other lipids such as TG in PDZD8-deficient brain is unclear at this time.
We have demonstrated that lipophagy was impaired in the cholinergic neurons of the MHb in PDZD8-deficient mouse brain (Figures 3G-3J) and in the cultured dopaminergic neurons of PDZD8-deficient PC12 cells (Figure 7). On the other hand, we do not rule out the possibility of CE accumulation in PDZD8-deficient glial

Limitations of the study
This study demonstrates for the first time that PDZD8 plays a role in promoting lipophagy and that deficiency of PDZD8 leads to dyslipidemia in the mammalian brain. However, these evidences were limited to neuronal cell lines and mouse brains, and further confirmation of findings in humans is needed. Increased inflammation due to lipid accumulation in the brain has been suggested as a causative mechanism of neurological diseases in humans. Mutations in human PDZD8 have also been reported to be associated with abnormal brain development and intellectual disability. PDZD8 may therefore be a key molecule in elucidating the relationship between dyslipidemia and inflammation-dependent neurological disorders in the brain.

STAR+METHODS
Detailed methods are provided in the online version of this paper and include the following:

Lead contact
Further information and requests for resources and reagents should be directed to and will be fulfilled by the lead contact, Michiko Shirane (shiram@phar.nagoya-cu.ac.jp).

Materials availability
This study did not generate new unique reagents.

Data and code availability
All data reported in this paper will be shared by the lead contact upon request. This paper does not report original code.
Any additional information required to reanalyze the data reported in this paper is available from the lead contact upon request.

Mouse diets
The ND (Rodent Diet CE-2) and HFD (High Fat Diet-32) were obtained from CLEA Japan. The HFD contains a crude fat content of $32% including powdered beef tallow and high oleic-type safflower oil. Mice were fed the HFD (or ND) for 1 month before analysis.

Antibodies and reagents
Rabbit polyclonal antibodies to PDZD8 were from Sigma-Aldrich, and guinea pig polyclonal antibodies to AChT were from Sigma-Aldrich. Alexa Fluor 488 goat anti-rabbit IgG and Alexa Fluor 555 donkey anti-goat IgG as well as LysoTracker Red were from Thermo Fisher Scientific. Hoechst 33342 and filipin were from Sigma-Aldrich. Digitonin was from Tokyo Chemical Industry. HeLa cells were cultured under a humidified atmosphere of 5% CO 2 at 37 C in Dulbecco's modified Eagle's medium (DMEM, Wako) supplemented with 10% fetal bovine serum (FBS, Nichirei). They were transfected with expression vectors for 24 h with the use of the ExtremeGENE9 reagent (Roche). PC12 cells were cultured under a humidified atmosphere of 5% CO 2 at 37 C in DMEM supplemented with 10% FBS and on plates coated with poly-L-lysine (150-300 kDa, Sigma) before exposure to mouse submaxillary gland nerve growth factor (Merck Millipore) at 100 ng/mL in RPMI 1640 supplemented with 1% horse serum (Thermo Fisher Scientific). They were transfected with siRNAs or expression vectors with the use of a Nucleofector system (Lonza).

RNA interference
Stealth siRNAs targeted to human or rat PDZD8 mRNA were obtained from Invitrogen-Life Technologies. The siRNAs were introduced into HeLa cells with the use of the Lipofectamine RNAiMAX reagent (Invitrogen-Life Technologies), and the cells were then cultured for 72 h before experiments. The siRNAs were introduced into PC12 cells with the use of a Nucleofector instrument (Lonza), and the cells were then cultured for 72 h before experiments. Results are shown for human and rat siRNAs #1, but similar findings were obtained with the other two siRNAs for each species. Stealth siRNA sequences are provided in Table S2.

RT and real-time PCR analysis
Total RNA was isolated from mouse brain with the use of an RNeasy Kit (Qiagen) and was subjected to RT with the use of ReverTra Ace qPCR RT Master Mix with gDNA Remover (Toyobo). The resulting cDNA was subjected to real-time PCR analysis with Thunderbird Next SYBR qPCR Mix (Toyobo) in a StepOnePlus Real-Time PCR System (Thermo Fisher Scientific). The amounts of target mRNAs were normalized by that of HPRT mRNA for relative quantification. Plasmid cDNA was used as a standard for absolute quantitative analysis. Primer sequences for PCR are provided in Table S1.

Fluorescence imaging of live cells
Cells that had been transfected with plasmids encoding fluorescently tagged proteins or metabolically labeled with fluorescent probes were observed with an LSM800 confocal microscope (Zeiss), and the images were processed for calculation of fluorescence intensity with ZEN imaging software (Zeiss). LysoTracker Red (0.5 mM) was added to cells for 1 h at 37 C. Filipin (5 mg/mL) was added to cells for 1 h at 37 C after permeabilization with digitonin (3 mg/mL) for 10 min at 37 C.

Immunohistofluorescence analysis
Mouse brain was fixed with 3.7% paraformaldehyde in phosphate-buffered saline (PBS, Wako), exposed consecutively to 15% and 30% sucrose in PBS for 2 days, and embedded in Surgipath FSC22 (Leica). Thin sections were prepared with a cryostat microtome (CM1850 UV, Leica) and stained consecutively with primary antibodies and Alexa Fluor-labeled secondary antibodies in PBS containing 0.5% Triton X-100. Nuclei were stained with Hoechst 33342 (Wako) as indicated. The sections were observed with an LSM800 confocal microscope (Zeiss).

TEM
Mouse brain was fixed with 2% glutaraldehyde (Nisshin EM) in 0.05 M cacodylate buffer containing 5 mM CaCl 2 . The fixed tissue was washed in cacodylate buffer with CaCl 2 , exposed to buffered 1% OsO 4 plus 0.8% K 4 [Fe(CN) 6 ]$3H 2 O for 1 to 2 h, dehydrated with acetone or ethanol, and embedded in Epon-Araldite or Spurr's medium (Nisshin EM). Thin sections were stained with uranyl acetate and lead citrate and observed with a JEM-1400 Plus instrument (JEOL).

Expression and purification of recombinant proteins
Recombinant His 6 -tagged proteins were expressed in and purified from Escherichia coli. The BL21(DE3) pLys bacterial cells were transformed with pET30-based vectors, cultured, and then exposed to 0.5 mM isopropyl-b-D-thiogalactopyranoside for 16 h at 10 C. The cells were then subjected to ultrasonic treatment, and the soluble fraction of the cell lysates was isolated. The expressed His 6 -tagged proteins were purified with the use of Ni-NTA agarose (Wako) and were eluted with imidazole (Wako