A novel family of mammalian transmembrane proteins involved in cholesterol transport

Cholesterol is an essential compound in mammalian cells because it is involved in a wide range of functions, including as a key component of membranes, precursor of important molecules such as hormones, bile acids and vitamin D. The cholesterol transport across the circulatory system is a well-known process in contrast to the intracellular cholesterol transport, which is poorly understood. Recently in our laboratory, we identified a novel protein in C. elegans involved in dietary cholesterol uptake, which we have named ChUP-1. Insillicoanalysis identified two putative orthologue candidate proteins in mammals. The proteins SIDT1 and SIDT2 share identity and conserved cholesterol binding (CRAC) domains with C. elegans ChUP-1. Both mammalian proteins are annotated as RNA transporters in databases. In the present study, we show evidence indicating that SIDT1 and SIDT2 not only do not transport RNA, but they are involved in cholesterol transport. Furthermore, we show that single point mutations directed to disrupt the CRAC domains of both proteins prevent FRET between SIDT1 and SIDT2 and the cholesterol analogue dehydroergosterol (DHE) and alter cholesterol transport.

. Overexpression of SIDT1 or SIDT2 does not increase dsRNA uptake. HEK293 FT cells (A), PANC-1 cells (B) and S2 cells were transfected with SID-1 (from C. elegans), SDIT1 (human), SDIT2 (mouse) or empty GFP vector (Mock) and incubated with 5 ug/ml of radioactive dsRNA (700 nucleotides) for 2 hours at 37 °C. The molarity was calculated with the sequence of dsRNA and then normalized with the total protein of each sample. Error bar: SD; n > 5; *p < 0.01. (C) S2 cells from Drosophila melanogaster incubated for 4 hours at 37 °C with dsRNA (700 nt length), miRNA (pre-let-a) or the equivalent in counts per minute (c.p.m)of 32P-γ-UTP. The uptake was normalized with the total protein for each sample; a.u. arbitrary units. (D) Mixing 5 µg/ ml of radioactive dsRNA with 10 ng of cholesterol results in dsRNA uptake in HEK293 cells overexpressing SIDT1, SIDT2, SID-1 or the empty vector. Values show the mean ± standard deviation of the amount of dsRNA incorporated per mg of total protein in the cell lysates. (E) Dose-dependent uptake of radioactive dsRNA-dependent on the cholesterol content in the mix (0 to 1000 ng of cholesterol). *p < 0.001; **p < 0.01; ***p < 0.05. (F) RT-PCR of each transfection in S2 cells or HEK293 FT cells. The actin gene (S2) and GAPDH (HEK293) were used as controls. -RT control without reverse transcriptase. (G) Confocal microscopy colocalization studies for SDIT1 or SIDT2 using FM-464 as plasma membrane marker and ER-tracker for the endoplasmic reticulum (ER). Notice in the zoom that SIDT1 shows very high co-localization with FM-464 whereas SIDT2 shows an intracellular localization co-localizing with ER tracker. (H) Mean ± standard deviation for Pearson's correlation coefficient between FM-464 and GFP. Notice the very high co-localization between Orai1 channel, which was used as plasma membrane (PM) localization control with FM-464. SIDT1 shows SCIeNtIFIC RePoRTS | 7: 7450 | DOI: 10.1038/s41598-017-07077-z transporter SID-1. However, we have previously shown that the greater homology lies between these two proteins and ChUP-1, not only at the amino acid level but also they have two putative CRAC domains located in similar regions to ChUP-1 and all form proteins with putative 9-transmembrane domains.
In the present study, we have cloned human SIDT1 and SIDT2 and produced fusion proteins to the green fluorescent protein (GFP). Using a wide variety of methods, we show evidence strongly suggesting that SIDT1 and SIDT2 are involved in cellular cholesterol transport in mammalian cells. Site-directed mutagenesis directed to disrupt putative CRAC domains indicate that the domain located in the transmembrane region of both proteins is involved in cholesterol binding. Disruption of this CRAC domain prevents FRET between SIDT1 and SIDT2 proteins and the cholesterol analogue dehydroergosterol (DHE) while affecting the uptake of [ 3 H]-Cholesterol. Most interestingly, removal of cholesterol from the plasma membrane (PM) induces the translocation of SIDT1 from intracellular compartments to the PM. This translocation requires clathrin, as reducing clathrin levels with RNA interference (RNAi) results in diminished SIDT1 translocation to the PM and prevents the uptake of [ 3 H]-Cholesterol. While SIDT1 participates in the uptake of cholesterol from the extracellular medium and into the cell, the localization of SIDT2 is in intracellular compartments, most likely the endoplasmic reticulum and Golgi apparatus. SIDT2 associates to intracellular cholesterol as evidenced by FRET experiments and may be involved in the redistribution of cholesterol among organelles. Interestingly, SIDT2 does not participate in cholesterol uptake from the extracellular space, as evidenced by the lack of [ 3 H]-Cholesterol uptake.
All the evidence gathered in the present study indicates that SIDT1 and SIDT2 are members of a novel family of proteins involved in the uptake and intracellular cholesterol transport in mammals.

Results
SIDT1 and SDIT2 are involved in cholesterol but not dsRNA transport. Previous studies have identified SIDT1 and SIDT2 as homologues of the C. elegans SID1 protein, which is involved in the transport of double-stranded RNA (dsRNA) in the nematode 32,33 . One of such studies showed the transport of dsRNA in PANC-1 cells overexpressing SIDT1. Databases have both genes annotated as dsRNA transporters (NCBI, SIDT1 Gene ID:54847 and SIDT2 Gene ID: 51092). We have shown previously that both SIDT1 and SIDT2 shared a higher identity with ChUP-1 than with SID1 (31) (Supplementary Fig. 1). We have reported previously that ChUP-1 is a novel protein involved in dietary cholesterol uptake in the nematode C. elegans. We conducted experiments with radioactive dsRNA using C. elegans SID1 as positive control. Transport experiments show that neither SIDT1 nor SIDT2 transport dsRNA or microRNAs in HEK293, PANC-1 or S2 Drosophila melanogaster cells ( Fig. 1A-C). Most interestingly, mixing dsRNA with cholesterol resulted in increased uptake of dsRNA (Fig. 1D). The transport of the dsRNA was dose-dependent on the amount of cholesterol used (Fig. 1E). The expression of SID-1, SIDT1and SIDT2 was corroborated with RT-PCR (Fig. 1F).
The fusion protein SIDT1-GFP shows high co-localization with ER-tracker (a marker for the endoplasmic reticulum) and FM-464, a marker of the plasma membrane, while SIDT2 shows co-localization almost exclusively with ER-tracker (Fig. 1G). The co-localization between SIDT2 and FM-464 was insignificant (Fig. 1H). As positive control, we used the channel Orai1, which is present only at the plasma membrane 34 . Biotinylation experiments confirmed that SIDT1 is present at the plasma membrane while SIDT2 is not (Fig. 1I).
To further explore the association of SIDT1 and SIDT2 to cholesterol, we conducted FRET experiments using the cholesterol fluorescent analogue dehydroergosterol (DHE) 31 . Since both proteins carry putative CRAC domains, we produced mutants with the CRAC domains disrupted. Mutations in the transmembrane CRAC domain prevented FRET between SIDT1 and DHE (Fig. 2). The extracellular CRAC domain does not appear to be involved in DHE association ( Fig. 2A-F). Disruption of the transmembrane CRAC domain or the double mutant not only prevented FRET between SIDT1 and DHE but also prevented [ 3 H]-Cholesterol uptake (Fig. 2F). Even though SIDT2 is not involved in [ 3 H]-Cholesterol uptake from the extracellular space, we found that intracellularly located SIDT2 generates a robust FRET signal with DHE, which can be prevented when using the mutant with the transmembrane CRAC domain disrupted or the double mutant ( Supplementary Fig. 2). Furthermore, SIDT2 shows strong co-localization with golgin-97, a protein resident of the trans-Golgi network (TGN) 35 , (Supplementary Fig. 3). All these results strongly suggest that SIDT2 is involved in the intracellular transport of cholesterol.
Reducing cholesterol from the plasma membrane triggers the translocation of SIDT1 to the PM. Since SIDT1 is present mainly at the plasma membrane (PM) and involved in the uptake of cholesterol, we explored the effect of reducing PM cholesterol using the agent methyl beta cyclodextrin (MβCD). Interestingly, treating the cells for 40 minutes with MβCD resulted in increased SIDT1 at the plasma membrane (Fig. 3A). The increment of SIDT1 was evident using line analysis of the regions in cells labeled with the plasma membrane marker, FM-464 (Fig. 3A). Using total internal reflection fluorescence microscopy (TIRFM) showed increased SIDT1-GFP fluorescence near the plasma membrane after MβCD treatment (Fig. 3C). This observation was confirmed by biotinylation studies (Fig. 3D).
To study further the translocation of SIDT1 to the PM, we conducted experiments with the mutants from the transmembrane and extracellular CRAC domains and the double mutant. Most interestingly, disruption of the transmembrane CRAC domain did not alter the translocation (Fig. 4D,E), but the disruption of the extracellular very high co-localization as well, compared with the poor co-localization of SIDT2. (I) Biotinylation studies for Orai1-GFP, SIDT1-GFP and SIDT2-GFP. Notice the subtle amount of SDIT2-GFP found at the plasma membrane when compared to SIDT1-GFP and Orai1-GFP. All conditions were included in a single figure and no other bands were detected.
CRAC domain prevented translocation of SIDT1 upon removal of cholesterol with MβCD ( Fig. 4C,D). Similar results were obtained with the double mutant. The extracellular CRAC domain overlaps with a vesiculation signal (Yxxɸ), which is recognized by specific accessory proteins to form vesicles like AP-2 or Dab-2 36 . Disruption of the vesiculation signal prevents the translocation of SIDT1 to the PM. To study this transport process in greater detail, we explored several proteins involved in vesicle transport. Because the vesiculation signal (Yxxɸ) was originally identified in cargo complexes found in clathrin-based vesiculation from the plasma membrane and the trans-Golgi network 37 , we decided to explore the role of clathrin in the translocation of SIDT1.

Selective sterol binding from the CRAC domain in SIDT1.
To study the binding of different sterols to the CRAC domains from SIDT1 we produced fusion proteins between the CRAC domains and GFP. We developed a novel high-throughput screening system based on TIRFM and microarrays of different sterols spotted on the glass. We measured the amount of the CRAC-GFP peptide bound to the TIRFM microarray using the fluorescence of GFP as a reporter. Most interestingly, the extracellular CRAC domain did not bind any of the sterols tested (Fig. 6A,B). As a positive binding control, we utilized an anti-GFP antibody spotted on the microarrays. In contrast, the transmembrane CRAC domain bound cholesterol and cholic acid, only weakly dehydrocholic acid but noβ-estradiol or testosterone (Fig. 6B). These results in combination with the [ 3 H]-Cholesterol uptake and FRET interactions strongly suggest that the transmembrane CRAC domain from SIDT1 is responsible for binding cholesterol (and perhaps other sterols) and the extracellular CRAC domain is not functional but forms part of a signal required for the translocation of SIDT1 to the PM upon reduction of cholesterol at the plasma membrane.

Discussion
The in silico analysis previously performed by our group identified putative mammalian homologues to C. elegans ChUP-1 protein, which is involved in cholesterol uptake in the digestive tract of the nematode 31 . Mammalian SIDT1 (Gene ID: 54847) and SIDT2 (Gene ID: 51092) are annotated in National Center for Biotechnology Information (NCBI) as homologues of the dsRNA transporter SID1 from C. elegans. However, we showed previously that the highest identity is with ChUP-1 31 . Even though a couple of studies have shown dsRNA transport by SIDT1 32,33 . A recent study suggested that dsRNA transport by SIDT1 was attained only when the dsRNA was covalently bound to cholesterol 38 . Furthermore, this study shows that reduction of SIDT1 protein via knockdown with RNAi prevented the transport of dsRNA covalently attached to cholesterol 38 . In the present study, we show that dsRNA transport can be attained when mixed with cholesterol (Fig. 1). These evidences strongly suggest that SIDT1 may transport dsRNA associated with cholesterol, but the primary role of SIDT1 is the transport of sterols. Due to its cellular localization, SIDT1 (which is predominantly found in plasma membrane) appears to carry a role in the uptake of cholesterol from the extracellular space; such function is performed via binding of cholesterol to the transmembrane CRAC domain of the protein. This suggests that cholesterol may be incorporated into the plasma membrane by diffusion and SIDT1 binds the cholesterol to incorporate it into the cell interior via endocytosis of the SIDT1-cholesterol complex. SIDT2, on the other hand, co-localizes with the endoplasmic reticulum and Golgi apparatus. This intracellular localization suggests that SIDT2 may participate in the re-localization of cholesterol among organelles (Fig. 7). This observation is confirmed by the robust FRET signal between SIDT2 and DHE. Both proteins (SIDT1 and SIDT2) bind cholesterol via their transmembrane CRAC domains, as evidenced by single point mutations directed to disrupt these domains. The extracellular CRAC domains in SIDT1 and SIDT2 overlap with a putative vesiculation signal (Yxxɸ). This vesiculation signal was originally identified in cargo complexes found in clathrin-based vesiculation from the plasma membrane and the trans-Golgi network 37 .
Most interestingly, disruption of this vesiculation signal prevents the translocation of SIDT1 to the plasma membrane upon removal of cholesterol from the PM. Similar results are obtained if RNAi is directed to knock down clathrin. Clathrin knockdown prevents SIDT1 mediated [ 3 H]-cholesterol uptake (Fig. 5). The translocation of SIDT1 upon reduction of cholesterol is not unique, other proteins involved in cholesterol transport like the sterol regulatory element-binding protein (SREBP), which acts as a lipid sensor in the endoplasmic reticulum to maintain the homeostasis of lipids in the cell, also shows translocation 6,8,10,23,39 . This vesiculation signal (Yxxɸ) is found in SIDT2 as well, but its functional significance was not evaluated in the present study. Measuring the translocation of cholesterol between organelles in living cells is a fairly complex task.
Both SIDT1 and SIDT2 are present in a wide variety of tissues in mammals. The expression of SDIT1 seems to be more important in the late developmental stages of mice, and it continues to be highly expressed in the adult (Supplementary Fig. 4). In contrast, SDIT2 is mainly expressed in early developmental stages of the Figure 7. Cartoon illustrating the model proposed for cholesterol transport mediated by SIDT1 and SIDT2. (A) Under normal cholesterol plasma membrane levels, SIDT1 is present in the plasma membrane and endoplasmic reticulum. Upon reduction of plasma membrane cholesterol with MβCD, SIDT1 moves to the plasma membrane by a clathrin-dependent mechanism. This newly incorporated SIDT1 may come from the endoplasmic reticulum or vesicles located near the plasma membrane. (B) While SIDT1 is present in the plasma membrane, SIDT2 is found primarily in the endoplasmic reticulum and the Golgi apparatus. SIDT2 may be involved in the transport of cholesterol between these two organelles. mouse ( Supplementary Fig. 4). Our data suggests that SDIT1 is primarily expressed in the brain and thymus (Supplementary Fig. 4).
Cholesterol plays a key role in nerve regeneration in the adult brain 40 . Furthermore, reduction in SREBP protein alters oligodendrocyte differentiation 41 and may play a role in schizophrenia 42 . The role of SIDT1 in the uptake of cholesterol in the CNS is a subject of future investigations.

Materials and Methods
Reagents. All reagents used in this study were of analytical grade from Sigma (Saint Louis, MO) unless otherwise stated. Membrane marker FM4-64 was purchased from Invitrogene (Carlsbad, CA).
Plasmids. SID-1 was cloned from c elegans as previously described 31 . SIDT1 was cloned from the human cDNA clone HFLJ from Riken Consortium (Japan) using the following primers: TCAGATCTGCACCGGGCTTTGGAA and GTGGATCCCAGAAGACAGGGATCTGGTCTC. The product was cloned in pEGFP-N1 from Clontech (Mountain View, CA. USA). Details of the cloning can be found in ref. 42. The mcherry-clathrin plasmid was purchased from AddGene. pRMHA3 plasmid was used to clone each protein for the overexpression in S2 (Drosophila melanogaster) cells, EndoCyan (a Rho GTPase localized in endosomes).
Uptake of dsRNA in cells. dsRNA was obtained using a T7 polymerase(NEBL) and α32P UTP (Perkin Elmer) from a generic primer for T7 43 . Cells overexpressing each one of the proteins were incubated in Optimem with 5 µg/ml of dsRNA (700 nucleotides of length) at 37 °C for 2 hours 43 . Then the cells were washed with PBS and lisated, the radioactivity was measured using a scintillation counter and was normalized against the total amount of protein. Cell culture. Human embryonic kidney (HEK293T) cells were purchased from the American Type Culture Collection (ATCC, Manassas, VA). Cells were grown on 35 mm culture dishes using DMEM(Invitrogen, Carlsbad, CA) supplemented with 10% of heat inactivated Fetal Bovine Serum (Wisent, premium quality, Canada), Penicillin-Streptomycin (Life technologies) and glutamine (Life technologies) in an incubator with humidity control at 37 °C and 5% of CO2. 24 hours before transfection the cells were incubated with trypsin and grown on circular coverslips then the cells were transfected by CaCl 2 protocol with the plasmids for SIDT1-GFP or SIDT2-GFP as following. Using a mixture of 1 µg of DNA and 6 µl CaCl 2 reaching 60 µl of final volume with distilled H 2 O next the mixture was added by dropping to a HeBS buffer (50 mM HEPES, 280 mM NaCl, 1.5 mM Na2HPO4, pH 7.05) and the final mixture was incubated for 30 minutes prior to replacing it withDMEM. The cells were incubated overnight with the transfection mixture and then replaced with fresh medium to perform the assays.

Site
Confocal microscopy and FRET assay. The confocal images were acquired using an Olympus Fv10i confocal microscope using the corresponding filter for each fluorophore (GFP, FM-464, DHE). For the colocalization analysis, we use specific dyes for each organelle, ER tracker (Invitrogen) for endoplasmic reticulum, Lysotracker for lysosomes, immunostaining using an antibody against golgin97 for Golgi apparatus, and FM-464 for plasma membrane. The FRET efficiency was determined by the sensitized emission method 44,45 , using the confocal microscope we acquired images with the different emission filters. Cells incubated with dehydroergosterol DHE (10 µM) during 6 hours with humidity control at 37 °C and 5% of CO2 overexpressing SIDT1-GFP, SIDT-2GFP or control cells. The automatic FRET calculation was done by the microscope ASW 2.1 software. Images for TIRF measurement were acquired with an Olympus microscope with the TIRF module using blue laser (405 nm) for GFP excitation. The images were analyzed using ImageJ software and co-localization plugin to obtain the Pearson's colocalization coefficients.

Depletion of cholesterol in the plasma membrane. Overexpressing SIDT-GFP cells were incubated
in Optimem with 1% of MβCD (w/v) for 1 hour for the depletion of cholesterol in the plasma membrane of the cells. After the incubation, the cells were rinsed three times with Krebs-Ringer solution and were maintained in this solution for the duration of the experiments.  31 . Hek293 cells were plated in 35 mm Petri dishes and used at 95% confluence or higher (1.2 × 10 6 cells). Cells were incubated with 10 mM methyl beta cyclodextrin (MβCD) to remove the cholesterol from plasma membrane. Cells were cultivated for 5 hours in serum-free media and exposed to 1 µC [ 3 H]-cholesterol 31 . After 4-hour exposure, the medium was aspirated, and the cell sheet washed gently, five times, with serum-free media. At the end of the washout, the medium was aspirated, the cell sheet washed four times with PBS and harvested in 1.0 ml PBS. To assay total radioactivity 0.3 ml cell preparation was placed on a high-performance glass vial (Perkin Elmer) mixed with 1.0 ml Pico-Fluor solution (Perkin Elmer). All samples were counted in a Quantulus liquid scintillation spectrometer (Perkin Elmer). Counting rates were corrected for quenching with an external standard and are expressed in disintegrations per minute (dpm).
Western Blot for membrane proteins. The membrane proteins were isolated using cell surface isolation kit (Pierce) in accordance with the manufacturer instructions using cells overexpressing SIDT1-GFP treated with MβCD and control cells (cells expressing SIDT1-GFP but not exposed to MβCD). Then the isolated membrane fraction proteins were analyzed by western blot, and the difference was determined by densitometry analysis using ImageJ software.
Synthesis of the CRAC domains from SIDT1. The sequence containing the extracellular CRAC domain from SIDT1 was: (150)ASMAPLGAQYKLLVTKLKHFQLRTNVA(176). The sequence containing the transmembrane CRAC domain from SIDT1 was:(633)AIHVLASLALSTQIYYMGRFKIDLGI(659). We produced fusion proteins by attaching these sequences to the amino terminus of GFP by PCR. In red are highlighted the CRAC domains. The proteins were produced using the bac-to-bac baculovirus expression system (Invitrogen), purified before used in the microarray studies. Purification was attained using the Green Fluorescent Protein Chromatography Kit (Bio-Rad). Purity was confirmed by HPLC.

Sterols microarrays.
We utilized the total internal fluorescence microscopy (TIRFM) system from TIRFLabs in microarray studies 46 . Cholesterol, cholic acid, dehydrocholic acid, β-estradiol and testosterone were purchased from Sigma (St. Louis Missouri). The anti-GFP antibody was purchased from Alamone. All sterols were dissolved in ethanol to a final concentration of 10 ng. The solution was mixed with 2% of polyacrylamide and spotted manually using a multibarrel pipette on the glass of standard 1 mm thick microarray glass slides. The slide was placed on the closed perfusion chamber of the lg-TIRFM system 46 . Purified fusion proteins containing the extracellular and transmembrane CRAC domains from SIDT1 fused to GFP were used at a final concentration of 100 nM and perfused into the chamber. The reaction was allowed to take place for 5 minutes, and then the chamber was rinsed with at least 5 times its volume to remove unbound fusion proteins. As a positive control, some spots in the microarray were printed with anti-GFP antibody suspended in PBS solution and mixed with 2% polyacrylamide. Data and materials availability. All non-commercial plasmids described in this study are available through an MTA.