The terminal enzymes of cholesterol synthesis, DHCR24 and DHCR7, interact physically and functionally.

Cholesterol is essential to human health, and its levels are tightly regulated by a balance of synthesis, uptake, and efflux. Cholesterol synthesis requires the actions of more than twenty enzymes to reach the final product, through two alternate pathways. Here we describe a physical and functional interaction between the two terminal enzymes. 24-Dehydrocholesterol reductase (DHCR24) and 7-dehydrocholesterol reductase (DHCR7) coimmunoprecipitate, and when the DHCR24 gene is knocked down by siRNA, DHCR7 activity is also ablated. Conversely, overexpression of DHCR24 enhances DHCR7 activity, but only when a functional form of DHCR24 is used. DHCR7 is important for both cholesterol and vitamin D synthesis, and we have identified a novel layer of regulation, whereby its activity is controlled by DHCR24. This suggests the existence of a cholesterol "metabolon", where enzymes from the same metabolic pathway interact with each other to provide a substrate channeling benefit. We predict that other enzymes in cholesterol synthesis may similarly interact, and this should be explored in future studies.

pcDNA5-DHCR7-myc/FRT contains the protein coding sequence of human DHCR7 (NM_001360.2) with a C-terminal myc epitope tag, and was cloned from cDNA from HeLa cells into pGEM T easy (Promega), and then subcloned into pcDNA5-FRT (Life Technologies) using restriction enzyme digestion.

Plasmid and siRNA transfection
For plasmid DNA transfections, CHO-EV and CHO-DHCR24-V5 cells were seeded in 10 cm dishes and transfected with 5 g DNA (pcDNA5-DHCR7-myc/FRT) for at least 4 h using 20 l Lipofectamine LTX transfection reagent, according to the manufacturer's instructions. After transfection, the cells were refreshed with medium, or statin treated overnight before harvesting for immunoprecipitation.
For siRNA transfections, cells were seeded in 6-well plates and transfected with 25 nM siRNA using 5 l Lipofectamine RNAiMAX transfection reagent, according to the manufacturer's instructions. After 24 h, cells were refreshed with medium overnight before treating/labeling/harvesting. Hamster-specifi c siRNA targeting DHCR24 [target sequence: GAGAGCCACGTGTGAAGCA; hamster versus human mRNA knockdown characterized previously ( 22 )] and DHCR7 (target sequence: GAAGCCAGGAGTAAAG-AAC) were designed by Sigma-Aldrich.

Quantitative real-time PCR
To measure mRNA levels, cells were seeded in a 12-well plate in triplicate wells per condition. The cells were transfected/ treated accordingly, and harvested for total RNA using TRI reagent (Sigma), according to the manufacturer's instructions. investigated whether mammalian cholesterol synthetic enzymes interact in the ER membrane in an organized manner as the cholestesome, initially focusing our efforts on DHCR24.
DHCR24 is involved in a remarkable diversity of cellular functions (e.g., oxidative stress, neuroprotection, cell survival), and is implicated in many diseases including cardiovascular disease, hepatitis C, certain cancers, and neurodegenerative diseases [reviewed in ( 13 )]. Defi cient/ defective DHCR24 results in a rare autosomal recessive developmental disorder called desmosterolosis, characterized by low cholesterol and high desmosterol levels ( 14 ). Thus, DHCR24 is an important enzyme in human health and disease. Here, we used DHCR24 as bait in a proteomic screen, and found that it interacts with DHCR7. Moreover, this interaction appears to have functional consequences on DHCR7 activity. Together, these results present the fi rst case of mammalian cholesterol synthesis enzymes interacting, and indicate a novel mode of regulation of DHCR7.

Materials
Chinese hamster ovary (CHO)-7 cells ( 15 )   from an inclusion list (discussed below), followed by up to the ten most abundant ions (>5,000 counts) with charge states of >2 were sequentially isolated and fragmented via collision-induced dissociation with an activation q = 0.25, an activation time of 30 ms, normalized collision energy of 30%, and at a target value of 10,000 ions. Fragment ions were mass analyzed in the linear ion trap.
To ensure that peptides associated with cholesterol synthesis proteins were preferentially targeted for fragmentation during LC-MS/MS, MS/MS inclusion lists were generated with the aid of Skyline (version 0.7.0.2494, University of Washington). Amino acid sequences for the proteins listed in supplementary Table 1 were imported into Skyline, and m/z values for doubly charged theoretical proteotypic peptide ions associated with these proteins were generated using the following parameters: enzyme, trypsin (0 missed cleavages); and structural modifi cations, carbamidomethyl cysteine. Exported m/z values were incorporated into inclusion lists using the mixed targeted and untargeted LC-MS/MS experiments described above.
Peak lists derived from LC-MS/MS were submitted to the database search program Mascot (version 2.3, Matrix Science). The following search parameters were employed: instrument type was ESI-TRAP; precursor ion and peptide fragment mass tolerances were ±5 ppm and ±0.4 Da, respectively; variable modifi cations included were acrylamide (C), carbamidomethyl (C), and oxidation (M); enzyme specifi city was trypsin with up to two missed cleavages; and rodentia taxonomies in the NCBI protein database ( 29 ) were searched . Peptide identifi cations were considered to be high confi dence if they were statistically signifi cant ( P < 0.05) according to the Mascot expect metric.

DHCR24 protein expression
CHO-EV, CHO-DHCR24, and CHO-DHCR24 Y471S cells were harvested in 10% (w/v) SDS supplemented with 2% (v/v) protease inhibitor cocktail (Sigma). Equal amounts of protein were mixed with Laemmli loading buffer, boiled for 5 min, and subjected to SDS-PAGE and Western blotting.

Lipid extraction and GC-MS
Cells were metabolically labeled with 1 g/ml [ 2 H 6 ]desmosterol or [ 2 H 7 ]7DHC for 4 h. Cells were harvested, and protein levels were normalized before adding 0.2 g 5 ␣ -cholestane as an internal standard for GC-MS. Lysates were saponifi ed in 1 ml ethanol, 75% (w/v) KOH, 20 l 20 mM EDTA, and 1 l 20 mM butylated hydroxytoluene at 70°C for 1 h. Nonsaponifi able lipids were extracted in 2.5 ml hexane, and dried using nitrogen gas. Lipids were derivatized in BSTFA at 60°C for 1 h and then analyzed using a Thermo Trace gas chromatograph coupled with a Thermo DSQIII mass spectrometer and Thermo Triplus autosampler (Thermo Fisher Scientifi c), modifi ed from a previous protocol ( 22 ). Samples (1.5 l) were injected via a heated (300°C) splitless inlet into a Trace TR-50MS GC column (60 m × 0.25 mm, 0.25 m fi lm thickness) (Thermo Fisher Scientifi c). The column oven was initially held at 70°C for 0.7 min, then heated to 250°C at 20°C min Ϫ 1 , then to 270°C at 3°C min Ϫ 1 , and then to 305°C at 1.5°C min Ϫ 1 (and held for 6 min in scan mode, or 3 min in selective ion monitoring mode). Helium was used as a carrier gas at a constant fl ow (1.2 ml min Ϫ 1 , with vacuum Total RNA (1 g) was reverse transcribed into cDNA with Superscript III First Strand cDNA synthesis kit (Life Technologies). Quantitative real-time PCR was performed using a Rotor-Gene 3000 and analyzed using Rotor-Gene 6.1 (Corbett). SensiMix SYBR No-Rox (Bioline) was used as the amplifi cation system using primers to amplify DHCR24 ( 23 ), DHCR7 ( 24 ), and the housekeeping control, porphobilinogen deaminase ( PBGD ) ( 25,26 ). Changes in gene expression levels were normalized to PBGD for each sample by the ⌬ ⌬ Ct method.

Immunoprecipitation
Cells were washed with ice-cold PBS and harvested in either CHAPS buffer [0.5% (w/v) CHAPS, 150 mM NaCl, 2 mM EDTA, 25 mM HEPES (pH 7.4), and 5% (v/v) glycerol, supplemented with Roche's protease inhibitor (one tablet per 10 ml buffer)] for LC-MS/MS experiments, or RIPA buffer [20 mM Tris-HCl (pH 7.4), 0.1% (w/v) SDS, 1% (v/v) Nonidet P-40, 0.5% sodium deoxycholate, 150 mM NaCl, 5 mM EDTA, and 1 mM sodium orthovanadate, supplemented with 2% (v/v) protease inhibitor cocktail (Sigma)] for Western blot detection (i.e., Fig. 2 ). The cell lysates were passed through a 22 gauge needle 20 times and centrifuged for 15 min at 20,000 g at 4°C. The cell lysates (supernatant) were adjusted to the same amount of protein in 1 ml buffer per sample. Dynabeads were rotated with the antibody for 1 h at room temperature, washed, and rotated with protein lysates overnight at 4°C. For the LC-MS/MS experiment, 3 mg protein lysate and 10 g anti-V5 antibody were used per sample, and 1 mg protein lysate and 5 g anti-myc antibody were used for the Western blotting experiment. The nonspecifi c proteins were then removed by a series of washes in their respective buffers (1 h wash, 30 min wash, and a 15 min wash). After the fi nal wash, all supernatant was removed from the samples, and the pellets were resuspended in 50 l immunoprecipitation loading buffer [2 vol of CHAPS or RIPA buffer, 2 vol 10% (w/v) SDS, and 1 vol fresh 5× Laemmli loading buffer (fi nal concentration : 50 mM Tris-HCl (pH 6.8), 2% (w/v) SDS, 5% (v/v) glycerol, 0.04% (w/v) bromophenol blue, and 1% (v/v) ␤ -mercaptoethanol)]. The samples were boiled at 95°C for 10 min with occasional vortexing before subjecting the supernatant to 10% (w/v) SDS-PAGE. The gels were then transferred for Western blotting or processed for LC-MS/MS analysis.

LC-MS/MS
After electrophoresis, gels were stained with EZ-Run protein gel staining solution (Thermo Fisher Scientifi c) according to the manufacturer's instructions. Gel bands were excised, destained, reduced, and alkylated following the procedure described by Shevchenko et al. ( 27 ). For each gel slice, 40 ng of trypsin in 120 l of 0.1 M NH 4 HCO 3 was used for 16 h at 37°C. The digest solutions were transferred to new microfuge tubes and the gel slices treated with the following solutions sequentially for 30 min each: 80 l of 0.1% (v/v) formic acid in 67% (v/v) acetonitrile, and 80 l of 100% acetonitrile. Peptide solutions were then dried (Savant SPD1010, Thermo Fisher Scientifi c) before resuspending in 20 l of 0.1% (v/v) formic acid. Proteolytic peptide samples were separated by nano-LC using an UltiMate 3000 HPLC and autosampler system (Dionex, Amsterdam, The Netherlands), and ionized using positive ion mode electrospray following experimental procedures described previously ( 28 ). Single-stage MS and MS/MS were performed using an LTQ Orbitrap Velos Pro (Thermo Electron, Bremen, Germany) hybrid linear ion trap and Orbitrap mass spectrometer. Survey scans m / z 350-2,000 were acquired in the Orbitrap (resolution = 30,000 at m / z 400, with an initial accumulation target value of 1,000,000 ions in the linear ion trap; lock mass applied to polycyclodimethylsiloxane background ions of exact m/z 445.1200 and 429.0887). Up to the fi ve most abundant ions 10, and noise method set to "Incos noise". Quantifi cation of DHCR24 and DHCR7 activity was based on chromatographic peak areas of the confi rmation ion measured for [ where P values of р 0.05 (*) and р 0.01 (**) were considered statistically signifi cant.

DHCR24 coimmunoprecipitates endogenous DHCR7
Interacting partner information for DHCR24 is currently unknown. To examine the potential interacting partner(s) compensation on). MS conditions were: electron energy 70 eV, ion source temperature 200°C, transfer line temperature 305°C. The emission current was set to 130 A and the detector gain to 6.0 × 10 5 . Samples were analyzed either in scan mode (60-520 Da, 2.994 scans s Ϫ 1 ) to obtain mass spectra for peak identifi cation, or in selective ion monitoring mode to monitor the following sterols:  Fig. 2). Unlabeled cholesterol and desmosterol were not monitored, although they were also detected, as selected ions were also lowly present in unlabeled cholesterol and desmosterol, which were present in high abundance. Thermo Xcalibur software (version 2.1.0.1140) was used to acquire and process the data. Chromatographic peaks were integrated using the following parameters: baseline window 40-200, area noise factor 5, peak noise factor 10, multiplet threshold The Mascot scores, protein coverage, and number of peptide matches are all derived from peptides identifi ed with Mascot Expect values <0.05, except for underlined score. For more information, see supplementary Table 2. Y, protein found; Y*, protein inferred from a peptide identifi ed with a Mascot Expect value >0.05, as MS/MS spectra derived from this peptide match those obtained in other experiments where the same peptide was identifi ed with a higher Mascot Ion Score; Y**, protein found in EV as well, but with fewer peptide hits in EV; N, protein not found; N*, protein found with more peptide hits in EV (data omitted); ID, identifi cation . a Experiments where only a selection of bands were subjected to LC-MS/MS analysis. b Experimental conditions that were statin treated overnight.
[ 2 H 7 ]cholesterol/[ 2 H 7 ]7DHC ratio as a measure for specifi c DHCR7 activity. As expected, knockdown of DHCR24 and DHCR7 decreased their respective enzymatic activity relative to the control siRNA-transfected condition, indicating that knockdown of mRNA translates to functional consequences downstream ( Fig. 3B ). Knockdown of DHCR7 had no effect on DHCR24 activity. Knockdown of DHCR24 , however, decreased DHCR7 activity by ‫ف‬ 60%. Because knockdown of DHCR24 has no effect on DHCR7 mRNA ( Fig. 3A ), this implies that this phenomenon is occurring posttranscriptionally. Thus, the interaction between the two terminal steps of cholesterol synthesis appears to have functional consequences.

Overexpression of functional DHCR24 increases DHCR7 activity
As knocking down DHCR24 decreased DHCR7 activity ( Fig. 3B ), we next examined whether the converse is true, i.e., whether overexpressing DHCR24 increases DHCR7 activity. Thus, we compared DHCR7 activity in the CHO-EV and CHO-DHCR24 cell-lines. We also included a CHO cell-line stably overexpressing DHCR24 Y471S, a desmosterolosis mutant of DHCR24 that results in a complete loss of DHCR24 activity ( 11 ), to determine whether DHCR24 function plays a role in DHCR7 activity. As a positive control, DHCR24 activity was examined under these conditions, displaying an increase in activity in CHO-DHCR24 cells compared with CHO-EV and CHO-DHCR24 Y471S cells, even though CHO-DHCR24 of DHCR24, we employed CHO-7 cells stably overexpressing V5 epitope-tagged human DHCR24 ( 19 ) or the EV ( 21 ) as a negative control cell-line. An advantage of using the Flp-In stable system is that it expresses one copy per cell at a specifi c location (FRT site), which makes it more physiological and therefore less susceptible to immunoprecipitating false positive interacting partners . Cells were untreated, or statin treated to lower cellular cholesterol status, which therefore upregulates cholesterol synthesis gene/protein expression, and DHCR24 protein was immunoprecipitated with the anti-V5 antibody. Immunoprecipitated proteins were fractionated by SDS-PAGE and trypsin digested before analysis by LC-MS/MS. This was performed using a targeted approach in which peptides of masses potentially associated with hamster cholesterol synthesis proteins were preferentially selected for MS/MS (see LC-MS/MS in the Materials and Methods). Immunoprecipitation of DHCR24 resulted in the coimmunoprecipitation of a number of proteins, with immunoprecipitated cholesterol synthesis proteins listed in Table 1 (more information available in supplementary Table 2). Of note was DHCR7, which was pulled down in all instances, with and without statin treatment, indicating that DHCR7 is very likely an interacting partner of DHCR24.

DHCR7 coimmunoprecipitates DHCR24
To confi rm the interaction between DHCR24 and DHCR7, we transiently expressed human DHCR7 containing a myc epitope tag (DHCR7-myc) in CHO-EV and CHO-DHCR24-V5 cells, immunoprecipitated DHCR7 with an anti-myc antibody, and determined whether DHCR24 was coimmunoprecipitated. As both overexpressed DHCR24 and DHCR7 contain 6×His tags, we employed an anti-5×His antibody to simultaneously detect pulled down DHCR24 and DHCR7 protein. Because the immunoprecipitation for the proteomic screen ( Table 1 ) was performed using a relatively low stringency CHAPS buffer, we used a more stringent RIPA buffer for the immunoprecipitation and wash steps, and found that DHCR7 immunoprecipitation also pulled down DHCR24, indicating that these proteins interact in cells ( Fig. 2 ).

Gene silencing of DHCR24 decreases DHCR7 activity
Next, we examined whether the interaction between DHCR24 and DHCR7 has functional consequences. We employed specifi c siRNA to knock down hamster DHCR24 and DHCR7 in CHO-7 cells, and determined their effects on DHCR24 and DHCR7 activity. DHCR24 and DHCR7 siRNA decreased their respective target mRNA by approximately 93% and 88%, while having little or no effect on nontargeted mRNA ( Fig. 3A ).
To measure the specifi c activity of DHCR24 and DHCR7, we labeled CHO-7 cells with deuterated-substrates, [ 2 H 6 ] desmosterol for DHCR24 and [ 2 H 7 ]7DHC for DHCR7 (supplementary Fig. 2), and measured their conversion to deuterated-cholesterol by GC-MS, using the [ 2 H 6 ]cholesterol/[ 2 H 6 ]desmosterol ratio as a measure for specifi c DHCR24 activity [as in our previous study ( 22 )], and the  and CHO-DHCR24 Y471S cells express similar levels of DHCR24 protein ( Fig. 4A ). As expected, DHCR24 knockdown decreased DHCR24 activity almost completely in CHO-EV and CHO-DHCR24 Y471S compared with CHO-DHCR24 cells ( Fig. 4B ), as the DHCR24 siRNA is targeted to hamster (endogenous) and not overexpressed human DHCR24 ( 22 ). As shown in Fig. 3B , knockdown of DHCR24 decreased DHCR7 activity in all cell-lines ( Fig. 4C ). Signifi cantly, DHCR7 activity was ‫ف‬ 40% higher in cells overexpressing DHCR24 compared with CHO-EV ( Fig. 4C ). But when mutant DHCR24 was overexpressed, DHCR7 activity was comparable to CHO-EV, again consistent with the contention that DHCR24 function is important for DHCR7 activity.

DHCR24 defi ciency results in lowered DHCR7 activity
As proof of concept, we employed a cell-line defi cient in DHCR24, the mouse macrophage-like cells, J774A.1 ( 30 ). As expected, no DHCR24 activity was detected in these cells ( Fig. 5A ). While J774A.1 cells were able to convert deuterated 7DHC to cholesterol, their DHCR7 activity was much lower compared with our model cell-line, CHO-7 ( Fig. 5B ).

DISCUSSION
In this study, we found that the two terminal enzymes of cholesterol synthesis interact physically ( Table 1 , Fig. 2 ) and functionally ( Figs. 3, 4 ). Knockdown of DHCR24 decreased DHCR7 activity but did not completely impair its function, suggesting that DHCR24 aids DHCR7 function, but is not essential for it. This contention was supported by overexpression of DHCR24 further increasing DHCR7 activity, but only when functional DHCR24 was used.
We have uncovered a novel layer of regulation in cholesterol synthesis. Previously, virtually nothing was known about the posttranscriptional regulation of DHCR7, although it has recently been found that vitamin D inhibits DHCR7 activity in differentiated human keratinocytes through an unknown mechanism ( 31 ). Here, we have discovered that the activity levels of the alternate terminal enzyme, DHCR24, infl uences DHCR7 activity. Moreover, this is the first time that two mammalian cholesterol synthesis enzymes have been shown to interact, but we expect that many others also interact, probably in a coordinated fashion. Because DHCR7 and DHCR24 act sequentially in the Bloch pathway ( Fig. 1 ), it is possible that their interaction provides a simple way for intermediates to be moved from one enzyme to the next. In this way, it may be expected that something of a "daisy chain" may be formed, where each enzyme interacts with both its preceding and subsequent enzymes.
This concept of enzymes from the same pathway interacting with their related enzymes (metabolon) has also been hypothesized for other metabolic pathways. For example, there is some experimental evidence for enzyme complexes forming in glycolysis, but the full details are yet to be elucidated [reviewed in ( 32 )]. One of the evolutionary advantages suggested is similar to our hypothesis, the channeling of substrates. Physical interactions between fi ve enzymes in the TCA cycle have also been observed in Bacillus subtilis ( 33 ), and again the authors suggested substrate channeling as a likely benefi t.
Although evidence points toward the existence of metabolons, it has been suggested that these interactions are hard to detect using high-throughput assays because they only weakly or transiently interact under normal circumstances, and their assembly is regulated as required ( 34 ). This helps explain why we did not observe a stoichiometric interaction between DHCR24 and DHCR7 in our coimmunoprecipitation approach (apart from the fact that only a proportion of the cells would be transfected with the DHCR7 expression plasmid). We would expect only a subset of DHCR24 to be interacting with DHCR7 at any given time.
As DHCR24 is positioned not only at the end of the Bloch pathway but also at the gateway of the Kandutsch-Russell pathway, attuning its activity with the last step of this pathway (DHCR7) would ensure concerted control of cholesterol synthesis. Moreover, as 7DHC can be converted into cholesterol or vitamin D, DHCR7 represents a switch between these vital molecules in skin cells exposed to UVB, so it seems likely that this important branch point is regulated in some way.
In addition to DHCR24 acting at the start and end of each pathway, it is also able to act on any of the intermediates in the Bloch pathway to divert them into the Kandutsch-Russell pathway ( 35 ). This promiscuity may mean that it