A conserved degron containing an amphipathic helix regulates the cholesterol-mediated turnover of human squalene monooxygenase, a rate-limiting enzyme in cholesterol synthesis

Cholesterol biosynthesis in the endoplasmic reticulum (ER) is tightly controlled by multiple mechanisms to regulate cellular cholesterol levels. Squalene monooxygenase (SM) is the second rate-limiting enzyme in cholesterol biosynthesis and is regulated both transcriptionally and post-translationally. SM undergoes cholesterol-dependent proteasomal degradation when cholesterol is in excess. The first 100 amino acids of SM (designated SM N100) are necessary for this degradative process and represent the shortest cholesterol-regulated degron identified to date. However, the fundamental intrinsic characteristics of this degron remain unknown. In this study, we performed a series of deletions, point mutations, and domain swaps to identify a 12-residue region (residues Gln-62–Leu-73), required for SM cholesterol-mediated turnover. Molecular dynamics and circular dichroism revealed an amphipathic helix within this 12-residue region. Moreover, 70% of the variation in cholesterol regulation was dependent on the hydrophobicity of this region. Of note, the earliest known Doa10 yeast degron, Deg1, also contains an amphipathic helix and exhibits 42% amino acid similarity with SM N100. Mutating SM residues Phe-35/Ser-37/Leu-65/Ile-69 into alanine, based on the key residues in Deg1, blunted SM cholesterol-mediated turnover. Taken together, our results support a model whereby the amphipathic helix in SM N100 attaches reversibly to the ER membrane depending on cholesterol levels; with excess, the helix is ejected and unravels, exposing a hydrophobic patch, which then serves as a degradation signal. Our findings shed new light on the regulation of a key cholesterol synthesis enzyme, highlighting the conservation of critical degron features from yeast to humans.


Cholesterol biosynthesis in the endoplasmic reticulum (ER) is tightly controlled by multiple mechanisms to regulate cellular cholesterol levels. Squalene monooxygenase (SM) is the second rate-limiting enzyme in cholesterol biosynthesis and is regulated both transcriptionally and post-translationally. SM undergoes cholesterol-dependent proteasomal degradation when
cholesterol is in excess. The first 100 amino acids of SM (designated SM N100) are necessary for this degradative process and represent the shortest cholesterol-regulated degron identified to date. However, the fundamental intrinsic characteristics of this degron remain unknown. In this study, we performed a series of deletions, point mutations, and domain swaps to identify a 12-residue region (residues Gln-62-Leu-73), required for SM cholesterol-mediated turnover. Molecular dynamics and circular dichroism revealed an amphipathic helix within this 12-residue region. Moreover, 70% of the variation in cholesterol regulation was dependent on the hydrophobicity of this region. Of note, the earliest known Doa10 yeast degron, Deg1, also contains an amphipathic helix and exhibits 42% amino acid similarity with SM N100. Mutating SM residues Phe-35/Ser-37/Leu-65/Ile-69 into alanine, based on the key residues in Deg1, blunted SM cholesterol-mediated turnover. Taken together, our results support a model whereby the amphipathic helix in SM N100 attaches reversibly to the ER membrane depending on cholesterol levels; with excess, the helix is ejected and unravels, exposing a hydrophobic patch, which then serves as a degradation signal. Our findings shed new light on the regulation of a key cholesterol synthesis enzyme, highlighting the conservation of critical degron features from yeast to humans.
The regulation of the multistep cholesterol biosynthesis pathway is a key process in cholesterol homeostasis (1). Squalene monooxygenase (SM) 4 is one of two rate-limiting steps in cholesterol biosynthesis (2) and is implicated in various human diseases, including metabolic and neurological diseases (3,4), as well as certain cancers (5,6). Excess cholesterol accelerates the degradation of SM by the ubiquitin-proteasome system as an example of end-product inhibition (2). Through this regulatory mechanism, SM is well-placed as a key control point in the pathway to limit cholesterol synthesis and regulate cellular cholesterol levels.
Regulated turnover of proteins via the proteasome is determined by degradation signals within the proteins themselves. These intrinsic protein elements are referred to as degrons (7). We previously showed that the first 100 amino acids of SM (SM N100) represent the region responsible for its cholesterol-mediated turnover (2). Fusing SM N100 to GFP (SM N100-GFP) causes the heterologous protein to be degraded in a cholesteroldependent manner, indicating this region is a transferable cholesterol-regulated degron (2). To date, this represents the shortest degron known to respond to cholesterol. However, the inherent elements of these 100 amino acids that permit the cholesterol-regulated turnover event remain unknown. Examining the components of this cholesterol-regulated degron is fundamental to understand how SM fulfills its function in responding to excess cholesterol and limiting cholesterol synthesis.
We recently characterized the membrane topology of SM N100 (8). Both the N and C termini are cytosolic, and a re-entrant loop spanning 24 (between residues 18 and 41) to 33 residues (between residues 11 and 43) anchors the protein to the membranes of the endoplasmic reticulum (ER) (Fig. 1A). Excess cholesterol induces a conformational change in SM N100, detected at residues 9, 11 and 41, bordering the ER membranecytosol interface of the re-entrant loop (8). We also demonstrated that membrane-associated RING finger protein 6 (MARCH6, also called TEB4) is the E3 ubiquitin ligase respon-sible for the cholesterol-dependent degradation of SM (9). Similarly, Erg1 is degraded by Doa10, the yeast homologs of SM and MARCH6, respectively (10). However, Erg1 lacks the critical N100 region from human SM (2), suggesting different interaction sites with the E3 ubiquitin ligase between organisms.
In this study, we aimed to identify the critical regions or residues required for cholesterol regulation of SM N100, and we have discovered a key amphipathic helix located in the second half of this degron, which shares remarkable sequence homology with Deg1.

C-terminal half of SM N100 expresses poorly but retains cholesterol regulation
We previously showed that the N-terminal half of SM N100 (SM N49) was not cholesterol-regulated, and we showed higher basal protein expression than SM N100 (8). Following this finding, we considered the possibility that the C-terminal half of SM N100 is critical for cholesterol regulation. We cloned the second half of SM N100 (SM N100 ⌬48), fused to GFP, to test alongside the SM N49 and wild-type (WT) SM N100 constructs (Fig. 1B). SM N100 ⌬48 protein expression was lower compared with WT SM N100 (Fig. 1, C and D), suggesting that the second half of SM N100 contains a degron element.
We also compared the cholesterol regulation of the N-and C-terminal halves of SM N100, relative to WT SM N100. We found that cholesterol regulation in SM N49 was blunted, whereas SM N100 ⌬48 remained cholesterol-regulated ( Fig. 1, C and D), indicating that both halves of SM N100 are required for normal cholesterol regulation and protein expression.

Cholesterol-regulated degron requires a 20-residue region in the second half of SM N100
Because cholesterol regulation was observed in the low-expressing SM N100 ⌬48, and SM N49 showed blunted cholesterol regulation (Fig. 1, C and D), we predicted there was a region in the second half of SM N100 needed for cholesterolmediated turnover. To test this, we created four constructs with successive 10-amino acid truncations of the C-terminal half of SM N100, generating SM N90, SM N80, SM N70, and SM N60, all fused to GFP ( Fig. 2A). We tested the cholesterol regulation of these mutants and compared them with WT SM N100. Both SM N70 and SM N60 showed little to no cholesterol regulation contains a re-entrant loop, which is embedded in the ER membrane. Residues between 11-18 and 41-43 (shown in red) are noted to be on the ER membranecytosol interface. The black dashed lines at the C terminus of SM N100 refer to the continuation of the protein with the linker region and GFP (not shown). The C-terminal half of SM N100 is cytosolically exposed along with GFP. B, schematic of truncation mutants of SM N100. All constructs contain GFP fused to the C-terminal end of SM N100 and contain a V5 epitope tag. C, CHO-7 cells were transfected for 24 h with the indicated SM N100 constructs (0.25 g of expression plasmid and 0.75 g of pTK-empty vector). Cells were then pre-treated overnight with compactin (5 M) and mevalonate (50 M). The next day, cells were treated for 8 h with or without Chol/CD (20 g/ml). The Western blot is representative of at least three independent experiments. D, densitometric analyses of the Western blot in C. Data are presented as mean ϩ S.E. from at least three independent experiments. Significant changes in cholesterol regulation or protein levels compared with WT SM N100 are indicated: *, p Յ 0.05, or **, p Յ 0.01.
Cholesterol-regulated degron of squalene monooxygenase (Fig. 2,B and C). Cholesterol regulation of SM N80 reached the same level as WT SM N100 (Fig. 2, B and C), indicating that a region between residues 60 and 80 is necessary for cholesterolmediated turnover of SM N100.

Alanine-scanning mutagenesis reveals a residue required for cholesterol regulation
To locate short linear motifs required for the cholesterolregulated degron, we performed systematic site-directed mutagenesis of SM N100 by mutating every second amino acid to alanine and testing each mutant for cholesterol regulation (Fig. 3A). We expected point mutations that diminished cholesterol-mediated turnover would reveal important residues or motifs. Such approaches have been used to identify important residues in other cholesterol-related proteins such as Insig-2 (23), Scap (24), and Niemann-Pick C2 (25).
We quantified the protein expression ( Fig. 3B) and cholesterol regulation (Fig. 3C) of all 47 alanine mutants, normalized to WT SM N100. To increase stringency and reduce false discovery rates, we further imposed an arbitrary threshold, with relative protein expression levels below 0.5 or above 1.5 and relative cholesterol regulation below 0.8 or above 1.2 required to accept any observed statistical significance. We noted V32A did not express, whereas W2A and L70A expressed very poorly (Fig. 3B). Val-32 is located within the known re-entrant loop (8), so the mutation could be a destabilizing mutation that affects the native structure of the re-entrant loop.
One mutant, L42A, containing a mutated residue at the boundary of the exiting re-entrant loop (Fig. 1A), showed significantly higher protein levels and blunted cholesterol regulation (Fig. 3, B and C). However, alanine mutations of the neighboring residues (V41A and S43A) had normal cholesterol regulation (data not shown). Taken together, our alanine-scanning approach failed to identify short linear motifs required for the cholesterol-mediated turnover of SM N100, indicating that other approaches are needed to elucidate the degron.

SM N100 varies in protein expression and cholesterol regulation between species
Next, we took a comparative biology approach to identify motifs necessary for cholesterol regulation. SM N100 is wellconserved in mammals, but this region is lacking in lower eukaryotes such as yeast (2). We have also demonstrated that hamster full-length SM undergoes cholesterol-mediated turnover, highlighting the mammalian conservation of SM regulation (2). We hypothesized that SM N100 cholesterol regulation differs in lower order vertebrates due to evolutionary divergence giving rise to different primary amino acid sequences.
We aligned the first 100 amino acids of human, chicken, zebrafish, and lamprey SM (Fig. 4A). These animals were chosen to represent different evolutionary time points, with the lamprey, a jawless fish, being one of the earliest known vertebrates. Our multiple sequence alignment showed 14 identical residues, which were mostly in the first half of SM N100 (Fig.  4A). Cholesterol regulation might be expected to emerge with species like chicken, which are more closely conserved with human SM N100 (62% similarity), whereas the most divergent sequences such as lamprey (43% similarity) (Fig. 4B) may not be cholesterol-regulated. To test this, we cloned SM N100-GFP constructs for each of these species and tested their protein expression and cholesterol regulation. Protein expression of chicken SM N100 was significantly higher than human SM N100, whereas both zebrafish and lamprey had low expression (Fig. 4, C and D). Unexpectedly, chicken SM N100, which had the highest similarity and identity to human SM N100 (Fig. 4B), showed blunted cholesterol regulation (Fig. 4, C and D). In contrast, zebrafish SM N100 remained cholesterol-regulated, and lamprey SM N100 was stabilized by cholesterol (Fig. 4, C and D). Therefore, these divergent sequences appear to affect both protein expression and cholesterol regulation in an unpredictable manner.
We then investigated the overall protein sequence variability. We observed high sequence variability between residues 41-70 and 91-100 (Fig. 4E), consistent with the lack of conservation

Cholesterol-regulated degron of squalene monooxygenase
from the sequence alignment (Fig. 4A). The high variability and low conservation prompted us to determine whether disordered regions were present in SM N100, as disordered regions tend to be evolutionarily less conserved, exhibiting higher variability (26). More importantly, disordered regions are recognized as essential features in degrons (18, 20 -22). Next, we obtained disordered predictions from 13 predictors. In both human and chicken SM N100, residues 91-100 were most frequently predicted to be disordered, followed by residues 81-90 and 51-60 ( Fig. 4F). As expected, the disordered distribution was different in the two more distant species, zebrafish and lamprey. Taken together, we propose that the different choles-terol regulation and protein expression is not due to primary sequence divergence alone. Instead, structural differences, such as disordered regions, may explain the different responses observed in each degron variant.

Conserved yeast degron residues are important for SM N100 cholesterol regulation
We next considered known degron examples in the literature. One of the earliest-identified and most well-studied degrons is Deg1, which contains an amphipathic helix necessary for degradation (12), and it is degraded by Doa10, the yeast homolog of MARCH6 (14). We compared Deg1 with our SM  Red bars indicate pairwise identity below 30%. B, pairwise comparisons of sequence identity and similarity in A were determined using Geneious 9.1.5. Boldface red numbers indicate percentage similarity and boldface black numbers refer to percentage identity. C, CHO-7 cells were transfected for 24 h with the indicated SM N100 species constructs (0.25 g of expression plasmid and 0.75 g of pTK-empty vector). Cells were then pre-treated overnight with compactin (5 M) and mevalonate (50 M). The next day, cells were treated for 8 h with or without Chol/CD (20 g/ml). The Western blot is representative of at least three independent experiments. D, densitometric analyses of Western blot in C. Data are presented as mean ϩ S.E. from at least three independent experiments. Significant changes in cholesterol regulation or protein levels compared with human SM N100 are indicated: *, p Յ 0.05, or **, p Յ 0.01. E, heat map showing the SM N100 protein sequence variability using the Wu-Kabat variability coefficient. F, heat map shows frequency of disordered regions in 10-residue windows for each SM N100 species variant.

Cholesterol-regulated degron of squalene monooxygenase
N100 degron through a pairwise alignment, and we observed SM N100 has 27% identity and 42% similarity with the 62 residues of Deg1. Moreover, we observed that four critical amphipathic helix residues of Deg1 (12) aligned with SM N100. These residues corresponded to Phe-35, Ser-37, Leu-65, and Ile-69 in SM N100 (Fig. 5A). Using site-directed mutagenesis, we produced a quadruple alanine mutant F35A/S37A/L65A/ I69A in both SM N100 and full-length SM. In both cases, protein levels increased for the mutants, whereas the cholesterol regulation was blunted (Fig. 5, B-E). This suggests that these four conserved degron residues play an important role in SM cholesterol-mediated turnover.

A helical structure is present within the second half of SM N100
Amphipathic helices have been implicated as degrons of other proteins (12, 16 -19). We hypothesized that there may be an amphipathic helix in the second half of SM N100, which is required for SM cholesterol-mediated turnover. We first determined any secondary structural elements in full-length SM with PSIPRED version 3.3 and focused on residues 50 -100. We identified secondary helix structures from residues 62 to 73 and 90 to 99 (Fig. 6A). We also used PEP-FOLD3 and identified a helical structure between residues 61 and 71 (Fig. 6B), but no other secondary structure existed for the rest of the second half of SM N100 (data not shown).
To verify this, we performed molecular dynamics simulations of SM N100 in an all-atomistic representation. Two independent molecular dynamics simulations with different ab ini-tio modeled starting structures were performed (Fig. 6, C and D). The total computing time was 2.4 s. Simulation 1 showed residues 59 -61 and 68 -71 formed a 3 10 -helix (Fig. 6C), whereas simulation 2 showed residues 63-75 formed the longest and most stable ␣-helix (Fig. 6D). The data from the molecular dynamics simulations and secondary structure prediction servers support the presence of a helical structure within the second half of SM N100.

Amphipathic helix of SM N100 is necessary for cholesterol regulation
Previously, we ruled out the existence of any hydrophobic transmembrane regions in the second half of SM N100 (8).
Here, we considered the possibility of an amphipathic helix that is weakly associated with the ER membrane instead of a hydrophobic transmembrane domain. A helical wheel representation of SM N100 residues 62-73 revealed the hydrophobic moment, a measure of helix amphipathic character, for SM N100 ( H ϭ 0.54) (Fig. 7A) was comparable with that of Deg1 ( H ϭ 0.51) (Fig. 7D). We next removed the SM N100 amphipathic helix (⌬62-73) and found that this deletion abolished cholesterol regulation (Fig. 7, B and C), suggesting a crucial role of this amphipathic helix in cholesterol regulation of SM. We then showed that Deg1 fused to GFP is not cholesterol-regulated (Fig. 7, E and F), demonstrating that cholesterol regulation of a protein is not reliant on an amphipathic helix alone. Furthermore, Deg1-GFP levels were significantly lower than SM N100, consistent with Deg1 being intrinsically unstable as would be

Cholesterol-regulated degron of squalene monooxygenase
expected for a degron responsible for the turnover of a shortlived protein, MAT␣2 (12). We proposed the amphipathic helix is membrane-associated, allowing the helix to sense membrane cholesterol. Amphipathic helices have been reported to be involved in sensing membrane lipids or introducing membrane curvature, both of which are important in mediating proteinlipid interactions (27). Thus, it is likely that a cytosolically exposed helix cannot substitute the SM N100 amphipathic helix. We replaced the amphipathic helix with an ␣-helix (EAEIKPLAQSHAT) of myoglobin (Protein Data Bank code 3RGK) (28), one of the earliest and best studied globular cytosolic proteins with a high percentage of helices (29). As expected, SM N100 showed blunted cholesterol regulation when the highly charged ␣-helix (Fig. 7G) substituted the SM N100 amphipathic helix (Fig. 7, H and I). Hence, we have identified a critical amphipathic helix required for the cholesterolregulated turnover of SM N100.

Hydrophobicity is a key feature involved in cholesterol regulation of SM N100
As mentioned, SM N100 lacks any obvious membrane attachment point in the second half, and it is possible the amphipathic helix is only partially embedded in the membrane as exemplified by other known amphipathic helices (27). Increased cholesterol levels in the membrane can thicken the membrane (30) and perhaps induce dissociation of the partially associated amphipathic helix, leading to proteasomal degradation. If the attachment were stronger, increased cholesterol would be unlikely to eject this region from the membrane. We decided to test this by replacing the SM N100 amphipathic helix with a re-entrant loop that would retain the cytosolic orientation of the C-terminal end of SM N100. To this end, we replaced the amphipathic helix of SM N100 with re-entrant loops deduced from membrane topological studies of two ER membrane proteins, GOAT (YFSYLLFFPALLGGSL) and HHAT (WMLAYVFYYPVLHNG) (31)(32)(33). Both loops are very hydrophobic with minimal amphipathic character, as shown by their helical wheel representations (Fig. 8A).
We next tested the cholesterol regulation of these two new constructs. Surprisingly, SM N100 remained cholesterolregulated when the two hydrophobic loops replaced the amphipathic helix, despite having very low expression (Fig. 8, B and C). Because these hydrophobic loops permit cholesterol regulation, perhaps the overall hydrophobicity is more important in determining cholesterol regulation than the amphipathic character. From the 11 human SM N100 constructs cloned in this study with variations in this region (Fig.  8D), we observed a linear relationship between cholesterol regulation and hydrophobicity of the amphipathic helix region (Fig. 8E). Single point alanine substitutions in the amphipathic helix retained cholesterol regulation (Fig. 3), possibly due to hydrophobicity being unchanged. In contrast, large reductions of hydrophobicity, such as introducing a hydrophilic myoglobin helix, blunt cholesterol regulation (Fig. 7, G-I). Overall, we calculated that at least two-thirds of the variation (R 2 ϭ 0.70) (Fig. 8E) in cholesterol regulation are determined by hydrophobicity of the helical region between residues 62 and 73.

Cholesterol-regulated degron of squalene monooxygenase Amphipathic helix forms in a membrane-like matrix
We performed circular dichroism on a synthetic peptide made up of residues from the amphipathic helix of SM N100 (residues Gln-62-Leu-73). Addition of a non-ionic detergent, n-dodecyl ␤-D-maltoside as a membrane-like matrix, to the buffer promoted its transition from a disordered state to a helical structure, as indicated by the local minimum at 206 nm (Fig. 9A).
Therefore, the amphipathic helix is likely to form in contact with the ER membrane, as modeled here by the detergent micelles. We also analyzed a synthetic peptide from residues 50 to 60 of SM N100, the predicted disordered region (Fig. 4F). The peptide remained disordered with or without the addition of n-dodecyl ␤-D-maltoside (Fig. 9B).

Disordered region of SM N100 promotes basal turnover
Degrons are often characterized by disordered regions as these stretches are necessary for feeding into the proteasome (20 -22). We hypothesized that the amphipathic helix undergoes a transition to a disordered region when cholesterol ejects it from the membrane. This gives a 12-residue hydrophobic stretch, and together with the disordered stretch before the helix (residues 50 -60) (Fig. 4F), there should be sufficient length for engagement with the proteasome for degradation (21).
To test this, we replaced the amphipathic helix with a disordered flexible linker sequence containing glycine, alanine, and serine residues (34). Replacing the amphipathic helix with this disordered stretch blunted cholesterol regulation but did not affect basal protein turnover (Fig. 10, A and B). In contrast, deletion of the disordered region from residues 50 to 60 (⌬50 -60) increased basal protein levels, but cholesterol regulation remained the same as WT (Fig. 10, A and B). Although the disordered replacement failed to promote degradation, it is possible that the difference in hydrophobicity of the disordered linker and a disordered amphipathic helix accounted for differences in degradation (Fig. 8D).

Cholesterol dissociates the second half of SM N100 from the membrane
We now have evidence that an amphipathic helix in the second half of SM N100 is necessary for cholesterol regulation (Fig.  7, A-C) and that it only forms within a membrane-like environment (Fig. 9A). To further examine the involvement of an amphipathic helix in the second half of SM N100, we determined the effects of in vitro cholesterol treatment of microsomes prepared from cells transfected with either WT SM N100 or SM N100 ⌬48. SM N100 ⌬48 lacks the re-entrant loop but contains the amphipathic helix. We predicted that this truncated protein would be less membrane-associated and that

Cholesterol-regulated degron of squalene monooxygenase
cholesterol treatment would reduce this membrane association further.
On fractionation of the cell lysate to membrane pellet and supernatant, WT SM N100 was found almost exclusively in the pellet in agreement with our previous observations (8), whereas soluble GFP was found mostly in the supernatant as expected (Fig. 11A). Consistent with earlier experiments (Fig. 1, C and D), SM N100 ⌬48 had lower protein expression (Fig. 11A). We accounted for the higher protein expression of WT SM N100, by loading a quarter of these samples to match the protein levels of SM N100 ⌬48 (Fig. 11B). This allowed more direct comparison of pellet and supernatant fraction levels of the two proteins. We normalized the supernatant fraction to the pellet for each construct (Fig. 11C) and found that SM N100 ⌬48 was more enriched (ϳ6-fold) in the supernatant than WT SM N100. In vitro cholesterol treatment of the microsomes increased the proportion of SM N100 ⌬48 levels in the supernatant, whereas WT SM N100 did not change significantly (Fig. 11C). These data demonstrated that the second half of SM N100 is more loosely associated with the membrane than WT SM N100 and is more readily dissociated in the presence of cholesterol.

Discussion
As a key control point in cholesterol biosynthesis, SM must be tightly regulated to control the rate of cholesterol biosynthe-

Cholesterol-regulated degron of squalene monooxygenase
sis in cells. At a post-translational level, our previous work demonstrated that cholesterol induces the degradation of SM via the ubiquitin-proteasome system, which requires the first 100 amino acids as the cholesterol-regulated degron (2). Therefore, two requirements of this homeostatic system are as follows: (i) adequate enzyme expression in low cholesterol conditions to enable sufficient cholesterol synthesis, and (ii) cholesterol responsiveness so that the enzyme is degraded in high cholesterol conditions.
In this study, we have examined these two parameters in more than 60 variants of SM N100, and we identified a key degron element required for its cholesterol-responsive degradation. We mapped this element through progressive deletions and assorted mutations to a critical region between residues 62 and 73. Molecular dynamics (Fig. 6, C and D) and circular dichroism studies (Fig. 9A) confirmed secondary structural predictions that this is an amphipathic helix, which is stabilized by a hydrophobic matrix representing ER membrane. Removing this region severely blunted cholesterol regulation, as did replacing it with a hydrophilic ␣-helix (Fig. 7). However, replacing this region with highly hydrophobic loops maintained cholesterol regulation but compromised protein expression (Fig. 8, A-C). Our initial intention with the GOAT and HHAT constructs was to swap the amphipathic helix with a re-entrant loop that irreversibly anchors the second half of SM N100 to the ER membrane, but it appears that we may have simply introduced very hydrophobic domains lacking in amphipathic character. Indeed, we found that the hydrophobicity of this domain was a key determinant of cholesterol regulation, accounting for 70% of the variability in response to 12 constructs tested with alterations in this domain (Fig. 8E). However, protein expression did not correlate with hydrophobicity of this domain. This is likely affected by changes along the length of SM N100, with certain single point mutations (Fig. 3B) compromising protein expression as much as large deletions (Fig. 1D).
Amphipathic helices are becoming a common feature of degrons (12, 16 -19) and, as exemplified by one of the first degrons characterized, Deg1, part of the yeast transcription factor Mat␣2 (12). From our comparison of SM N100 with Deg1 (Fig. 5A), we identified four key residues required for normal cholesterol regulation of SM (Fig. 5, B-E). The four most important residues of Deg1 (12), all within the helical domain, aligned with two residues of SM N100 (Phe-35 and Ser-37) in the re-entrant loop (8) and two within the newly identified amphipathic helix (Leu-65 and Ile-69). The structure of SM is currently unknown, but the membrane-buried Phe-35 and Ser-37 are likely to be part of secondary structures, namely a hydrophobic helix within the ER. We previously showed that a cholesterol-responsive conformational change can be detected at the boundaries of the re-entrant loop, with flanking residues becoming less exposed with addition of cholesterol (8), perhaps due to membrane thickening (30).
As in its native yeast environment (12), Deg1 is highly unstable in a mammalian system, and unsurprisingly it is not cholesterol-responsive (Fig. 7, E and F). The SM N100 degron may be an example of evolutionary co-option, with natural selection repurposing an existing yeast degron, Deg1. Further refinement may then have followed to ensure that SM adequately expresses in low cholesterol conditions but is degraded in the presence of high cholesterol, thus providing mammals with another posttranslational control point in cholesterol synthesis, beyond HMG-CoA reductase. How this evolutionary co-option may have occurred is obscure, especially considering that SM N100 in lamprey, a more primitive organism, has less similarity to Deg1 (13%) than the more recently evolved human SM N100 (42%).
An emerging theme in protein-membrane interactions is the ability of amphipathic helices to sense membrane lipid composition (27). For example, the transducer of ER stress, Ire1, uses an amphipathic helix to sense membrane irregularities (35). The important phosphatidylcholine-synthesizing enzyme, CTP:phosphocholine cytidylyltransferase, also has an amphipathic helix that detects packing defects in the ER membrane at low levels of phosphatidylcholine, and then when phosphatidylcholine levels are sufficient, it is pushed out of the membrane, losing its secondary structure (36). Could an analogous situation occur with SM? By thickening (30) or even flattening the membrane (37), could cholesterol help eject the amphipathic helix of SM N100, enabling its degradation? In support of this contention, we have shown that the amphipathic helix preferably forms in the presence of membranes, as mimicked by a detergent in our circular dichroism Cholesterol-regulated degron of squalene monooxygenase studies (Fig. 9A). Moreover, we found that the second half of SM N100 that contains the amphipathic helix, its only obvious means of membrane anchorage, is more loosely associated with the membrane and that this association is further loosened when cholesterol is added (Fig. 11).
Besides having an amphipathic helix, another common feature of a degron is a disordered region. Degrons can contain an intrinsically disordered region of 20 -30 amino acids that can feed into the 26S proteasome (21), but the disordered stretch between residues 50 and 60 of SM N100 is too short. We propose that the neighboring amphipathic helix in SM N100 unravels after cholesterol-induced ejection from the membrane (analogous to CTP:phosphocholine cytidylyltransferase in response to sufficient phosphatidylcholine) to give the disordered length needed for the initiation of degradation (Fig. 12). Exposure of the unraveled helix, a hydrophobic region, may also promote degradation, given that exposed hydrophobic patches occur in many degrons (11, 38 -40). Furthermore, degradation of proteins with amphipathic helical degrons have been shown to be dependent on their hydrophobicity (12,17,19). This is likely the case for SM N100, because hydrophobicity of the amphipathic helix was a strong determinant of the cholesterolresponsiveness of SM N100 (Fig. 8E). Without association of the helix to the membrane, the exposure of a highly hydrophobic surface could register as a sign of misfolding, possibly leading to recognition by the degradation machinery, including MARCH6, the E3 ubiquitin ligase for SM N100 (9). In line with this, Doa10, the yeast homolog of MARCH6, is proposed to degrade substrates by recognizing hydrophobic patches on its substrates (11).
More work is needed to test this hypothetical model, and the cholesterol-unresponsive mutants identified in this study will be important to add more mechanistic detail. For example, we aim to determine whether these mutants still interact with MARCH6 and become ubiquitinated. To conclude, our studies on the regulatory domain of SM, a flux-controlling enzyme in cholesterol synthesis, have identified key features that define this as the shortest known cholesterol-responsive degron and provide new insights into its molecular mechanism of action. Furthermore, we have discovered a remark-able congruence between this degron and the archetypal yeast degron, Deg1.

Experimental procedures
Cell culture CHO-7 cell lines (gifts of Drs. Goldstein and Brown, University of Texas Southwestern) were maintained in DMEM/F-12 media supplemented with 5% (v/v) newborn calf lipoproteindeficient serum (LPDS), penicillin (100 units/ml), and streptomycin (100 g/ml).  . SM N100 has a characteristic re-entrant loop and an amphipathic helix that is proposed to associate transiently with the ER membrane. In theory, the ER membrane may exhibit membrane curvature because of protein insertions and phospholipid asymmetry (upper part of diagram). Cholesterol can thicken and condense the ER membrane, flattening the membrane, resulting in the amphipathic helix being displaced from the ER membrane (lower part of diagram). The unraveling of the amphipathic helix upon displacement exposes the hydrophobic patch that constitutes the amphipathic helix. This extends the disordered region length, resulting in a sufficient number of disordered residues to facilitate the cholesterol-mediated turnover of SM N100.

Transfection
Prior to transfection, cells were refreshed with DMEM/F-12 media supplemented with 5% (v/v) LPDS without antibiotics. For cholesterol regulation assays, CHO-7 cells were seeded in 6-well plates at 2.5 ϫ 10 5 cells per well and transfected the following day. Cells were transfected with 0.75 g of pTKempty vector DNA, 0.25 g of expression plasmid DNA, and 4 l of Lipofectamine-LTX for 24 h. For membrane in vitro cholesterol treatments, CHO-7 cells were seeded in 150-mm dishes and transfected with 12.7 g of pTK-empty vector plasmid DNA, 4.2 g of expression plasmid DNA, and 67.6 l of Lipofectamine-LTX for 24 h.

Cholesterol regulation assays
Transfected CHO-7 cells were pre-treated in maintenance media containing compactin (5 M) and mevalonate (50 M) overnight to reduce basal cholesterol status. Cells were then treated with cholesterol complexed with methyl-␤-cyclodextrin (CD) (20 g/ml) for 8 h before harvesting cell lysates. The complexing process was performed as described previously (43).

Cell fractionation
Transfected CHO-7 cells were pre-treated in maintenance media containing compactin (5 M) and mevalonate (50 M) overnight to reduce basal cholesterol status. Microsomal membranes were harvested and prepared as described previously with modifications (8,44). Briefly, cell lysates were centrifuged at 1,000 ϫ g at 4°C for 5 min, and the supernatant was subsequently centrifuged at 20,000 ϫ g at 4°C for 30 min to separate the cytosol and membrane fractions. Microsomal membrane pellets were resuspended in Buffer A (10 mM Hepes-KOH, pH 7.4, 10 mM KCl, 1.5 mM MgCl 2 , 5 mM sodium EDTA, 5 mM sodium EGTA, and 250 mM sucrose). For in vitro cholesterol treatment, equal amounts of microsomal membrane proteins were treated with cholesterol complexed with CD (20 g/ml) in Buffer A for 30 min at 37°C. After treatment, microsomal membranes were centrifuged at 20,000 ϫ g at 4°C for 30 min to remove excess cholesterol. Equal proportions of pellet and supernatant were loaded for Western blotting.

Molecular dynamics simulations
Because SM N100 does not share homology with any known experimentally resolved structure, we carried out two independent microsecond-time scale molecular dynamics simulations to extract detailed structural information. The starting structure of two molecular dynamics simulations was modeled using different methodologies. First, unbiased random dihedral assignment was performed to obtain a nearly linear chain of SM N100, and second, ab initio modeled structure using the on-line server QUARK (45). These two modeled structures enabled us to start two independent molecular dynamics simulations (simulation 1 for 1.3 s and simulation 2 for 1.1 s, respectively) for better sampling of folded conformations.
Molecular dynamics simulations were performed using the program GROMACS (46) by employing OPLS all-atom force field (47). The system was solvated using the TIP4P water representation (48). Periodic boundary conditions were used, and long-range electrostatic interactions were treated with the Particle Mesh Ewald summation (49). The real-space cutoff distance was set to 1.0 nm, and the van der Waals cutoff was set to 1.2 nm. A time step of 2 fs for numerical integration of the equations of motion was used. The protein was placed in a dodecahedral water box, large enough to contain the protein and at least 1.0 nm of solvent on all sides. The starting structures were subjected to energy minimization using the steepest descent method. Systems were simulated at 300 K and maintained by a V-rescale thermostat (50) with a time constant of 0.1 ps. Pressure coupling was done employing a Parrinello-Rahman (51) barostat using a 1-bar reference pressure and a time constant of 2 ps.

Circular dichroism
The synthetic peptides were diluted to 60 M for circular dichroism measurements. The Asn-50 -Gly-60 peptide was initially dissolved in water and diluted with buffer (final concentration of 10 mM potassium phosphate, 50 mM sodium sulfate buffer, pH 7.0). The Gln-62-Leu-73 peptide was initially dissolved in 100% (v/v) DMSO and further diluted with buffer (final concentration of 10 mM potassium phosphate, 50 mM sodium sulfate buffer containing 0.01% (v/v) DMSO, pH 7.0). Circular dichroism spectra were recorded with a 0.5-mm path length cuvette and collected on the Chirascan circular dichroism spectrometer (Applied Photophysics) with constant nitrogen stream at ambient temperature of 25°C. Each spectrum was recorded with a scan speed of 20 nm/min, bandwidth of 2 nm, and step of 1 nm. Spectrum presented is an average of three scans recorded from 190 to 260 nm with buffer subtracted from the peptide spectra in Pro-Data Viewer (Applied Photophysics).

Data presentation and statistical analysis
Relative protein levels for all SM N100, full-length SM, and Deg1 constructs were determined by normalizing to WT (human SM N100 or human full-length SM) as appropriate, which was set to 1. So for instance, a value of 2 would indicate the construct had double the protein level of WT. Cholesterol regulation was given as the proportion of protein degraded for each construct, normalized to the proportion of protein degraded for WT, which was set to 1. Values approaching zero on the cholesterol regulation scale indicate little or no cholesterol regulation, whereas a value of 1 would be the same as wild type. All Western blottings are representative of at least three independent experiments. Densitometry data from at least three independent experiments are presented as bar or line graphs. Densitometry data in bar graphs are presented as mean ϩ S.E., whereas data in line graphs are presented as mean Ϯ S.E. Statistical differences were determined by the Student's paired t test (two-tailed), where p values of p Յ 0.05 (*) and p Յ 0.01 (**) were considered statistically significant.
For the heat map of disordered region frequency distribution, the SM N100 protein for each species was divided into 10 parts, each containing 10-residue windows. The frequency distribution was determined by the cumulative number of predicted disordered residues within each 10-residue window from 13 predictors. The heat map for protein sequence variability was presented as the average of variability scores in each region as indicated in the figure.
Author contributions-N. K. C. and V. H. performed the experiments, analyzed, and interpreted the data. N. K. C. also generated the bulk of the bioinformatic analyses. N. J. and L. T. performed and provided data from molecular dynamics simulations. A. J. B. conceived the idea for the project and interpreted the data. All authors contributed to writing the paper, experimental designs, and approved the final version of the manuscript.