Cholesterol activates the G-protein coupled receptor Smoothened to promote morphogenetic signaling

Cholesterol is necessary for the function of many G-protein coupled receptors (GPCRs). We find that cholesterol is not just necessary but also sufficient to activate signaling by the Hedgehog (Hh) pathway, a prominent cell-cell communication system in development. Cholesterol influences Hh signaling by directly activating Smoothened (SMO), an orphan GPCR that transmits the Hh signal across the membrane in all animals. Unlike most GPCRs, which are regulated by cholesterol through their heptahelical transmembrane domains, SMO is activated by cholesterol through its extracellular cysteine-rich domain (CRD). Residues shown to mediate cholesterol binding to the CRD in a recent structural analysis also dictate SMO activation, both in response to cholesterol and to native Hh ligands. Our results show that cholesterol can initiate signaling from the cell surface by engaging the extracellular domain of a GPCR and suggest that SMO activity may be regulated by local changes in cholesterol abundance or accessibility.

concentration was varied. Under these conditions, Hh signaling activity increased in proportion to the amount of cholesterol in the MβCD:cholesterol complexes ( Figure 2A). Thus, cholesterol must be the active ingredient in these complexes that activates Hh signaling.
To define the structural features of cholesterol required to activate Hh signaling, we used MβCD to deliver a panel of natural and synthetic analogs ( Figure 2B). This experimental approach was inspired by previous studies of the cholesterol sensor SREBP cleavage-activating protein (SCAP) (Brown et al., 2002). The Hh signaling activity of cholesterol was exquisitely stereoselective-neither its enantiomer (ent-cholesterol) nor an epimer with an inverted configuration only at the 3-hydroxy position (epi-cholesterol) could activate Hh target genes ( Figure 2C). Enantioselectivity is consistent with cholesterol acting through a chiral binding pocket on a protein target, rather than by altering membrane properties (Covey, 2009). Hh signaling activity was also lost when either the number or the position of double bonds in the tetracyclic sterol nucleus were altered in 7-dehydrocholesterol (7-DHC) and lathosterol, two endogenous biosynthetic precursors of cholesterol. Interestingly, desmosterol, another immediate biosynthetic precursor of cholesterol that contains an additional double-bond in the iso-octyl chain, retained signaling activity. This structure-activity relationship points to the tetracyclic ring, conserved between cholesterol and desmosterol, as the critical structural element required for activity. We cannot exclude the possibility that desmosterol activated signaling because it was rapidly converted to cholesterol in cells. These strict structural requirements suggest a specific, protein-mediated effect of cholesterol on the Hh signaling pathway and further exclude the possibility that signaling activity is due to extraction of an inhibitor from cells by MβCD (present at the same concentration in all the sterol complexes tested in Figure 2C).
MβCD:sterol inclusion complexes have been suggested to potentiate Hh signaling by depleting an inhibitory molecule through an exchange reaction (Sever et al., 2016). This model cannot explain our results because the concentration ( Figure 2A) and structure ( Figure 2C) of the sterol in the inclusion complex, despite an unchanging MβCD concentration, can modulate Hh signaling activity.

Cholesterol functions at the level of Smoothened to activate Hedgehog signaling
A simplified schematic of the Hh signaling pathway is provided in Figure 3A (Briscoe and Therond, 2013). The receptor for Hh ligands, Patched 1 (PTCH1), inhibits signaling by suppressing the activity of SMO, a member of the GPCR superfamily. SHH binds and inhibits PTCH1, thereby allowing SMO to adopt an active conformation and transmit the Hh signal across the plasma membrane. Cytoplasmic signals from SMO overcome two negative regulators of the pathway, protein kinase A (PKA) and suppressor of fused (SUFU), ultimately leading to the activation and nuclear translocation of the GLI family of Hh transcription factors.
To pinpoint the site of cholesterol action within this sequence of signaling events, we conducted a series of epistasis experiments (Figure 3). The addition of forskolin (Fsk), which leads to an increase in the activity of PKA, blocks Hh signaling at a step between SMO and the GLI proteins. Fsk inhibited MβCD:cholesterol-mediated signaling, placing the site of cholesterol action at the level of or upstream of PKA ( Figure 3B). Two direct SMO antagonists, the steroidal natural product cyclopamine and the anti-cancer drug vismodegib, blocked Gli1 activation by MβCD:cholesterol ( Figure 3B) (Sharpe et al., 2015). This pharmacological profile established that MβCD:cholesterol requires SMO activity to promote signaling. Indeed, MEFs completely lacking SMO (Smo -/cells) failed to respond to MβCD:cholesterol, and the stable reexpression of wild-type (WT) SMO, but not a point mutant locked in an inactive conformation , rescued signaling ( Figure 3C) (Varjosalo et al., 2006;Wang et al., 2014). Thus, cholesterol must activate the Hh pathway at the level of PTCH1, SMO or an intermediate step.
We evaluated the possibility that MβCD:cholesterol interferes with the function of PTCH1 by using Ptch1 -/-MEFs, which completely lack PTCH1 protein and have high levels of Hh target gene induction driven by constitutively activated SMO (Taipale et al., 2002). MβCD:cholesterol activated signaling in Ptch1 -/cells treated with cyclopamine to partially suppress SMO activity, showing that cholesterol signaling activity did not depend on the presence of PTCH1 ( Figure   3D). MβCD:cholesterol behaved much like the direct SMO agonist SAG, since both could overcome SMO inhibition by cyclopamine in the absence of PTCH1.
Our epistasis experiments pointed to SMO as the target of cholesterol. However, compared to treatment with the native ligand SHH, SMO did not accumulate to high levels in primary cilia in  Figure 4A) (Nachtergaele et al., 2012;Sharpe et al., 2015). Agonistic oxysterols, such as 20(S)-hydroxycholesterol (20(S)-OHC), engage a hydrophobic groove on the surface of the extracellular cysteine-rich domain (CRD) of SMO (Myers et al., 2013;Nachtergaele et al., 2013;Nedelcu et al., 2013). We recently reported that cholesterol could also occupy this CRD groove.
A cholesterol molecule was resolved in this groove in a crystal structure of SMO. Furthermore, purified SMO bound to beads covalently coupled to cholesterol and this interaction could be blocked by free 20(S)-OHC, consistent with the view that both 20(S)-OHC and cholesterol occupy the same binding site . In addition, the extracellular end of the SMO 7TMD binds to the steroidal alkaloid cyclopamine, as well as to several non-steroidal synthetic agonists and antagonists (Chen et al., 2002a;Chen et al., 2002b;Frank-Kamenetsky et al., 2002;Khaliullina et al., 2015).
In order to distinguish if the activating effect of cholesterol is mediated by the cholesterol binding groove in the SMO CRD or the cyclopamine binding site in the 7TMD, we asked whether MβCD:cholesterol could activate signaling in Smo -/cells stably reconstituted with wild-type SMO (SMO-WT) or SMO variants carrying mutations in gatekeeper residues that have been shown to disrupt these two ligand-binding sites. The Asp477Gly mutation in the 7TM bindingsite of SMO ( Figure 4A), initially isolated from a patient whose tumor had become resistant to vismodegib, reduces binding and responsiveness to a subset of 7TM ligands, including SAG and vismodegib (Yauch et al., 2009). In the CRD, Asp99Ala/Tyr134Phe and Gly115Phe are mutations at opposite ends of the shallow sterol-binding groove that block the ability of 20(S)-OHC to both bind SMO and activate Hh signaling ( Figure 4A) (Nachtergaele et al., 2013). The Asp99Ala and Tyr134Phe mutations disrupt a hydrogen-bonding network with the 3β-hydroxyl group of sterols ( Figure 4A, inset) .
The Asp477Gly mutation in the 7TMD domain had no effect on the ability of MβCD:cholesterol to activate Hh signaling ( Figure 4B). SMO bearing a bulkier, charge-reversed mutation at this site (Asp477Arg) that increases constitutive signaling activity also remained responsive to  (Dijkgraaf et al., 2011). In contrast, the Asp99Ala/Tyr134Phe mutation in the CRD reduced the ability of MβCD:cholesterol to activate Hh signaling ( Figure 4C). The Asp99Ala/Tyr134Phe SMO mutant was also impaired in its responsiveness to SHH and to 20(S)-OHC, but remained responsive to the 7TMD ligand SAG  (Myers et al., 2013;Nedelcu et al., 2013).
This mutational analysis supports the model that the CRD binding-site, rather than the 7TMD binding-site, mediates the effect of cholesterol on SMO activity and thus on Hh signaling.
Interestingly, a mutation in Gly115, which is located on the opposite end of the CRD ligandbinding groove ( Figure 4A), did not alter the response to MβCD:cholesterol, even though it diminished the response to 20(S)-OHC as previously noted ( Figure 4D) (Nachtergaele et al., 2013). The SMO-Gly115Phe mutant also responded normally to the native ligand SHH ( Figure   4D). Gly115 is located near the iso-octyl chain of cholesterol in the SMO structure ( Figure 4A).
The introduction of a bulky, hydrophobic phenyl group at residue 115 may prevent the hydroxyl in the iso-octyl chain of 20(S)-OHC from being accommodated in the binding groove, but not disrupt binding of the purely hydrophobic iso-octyl chain of cholesterol. The ability of mutations to segregate 20(S)-OHC responses from cholesterol responses is consistent with solution-state small-angle X-Ray scattering data showing distinct conformations for SMO bound to these two steroidal ligands .
The ability of the Gly115Phe mutation to distinguish between cholesterol and 20(S)-OHC responses allowed us to address an important outstanding question: could cholesterol activate SMO only after being oxidized to a side-chain oxysterol? In addition to 20(S)-OHC, oxysterols carrying hydroxyl groups on the 25 and 27 positions can bind and activate SMO (Corcoran and Scott, 2006;Dwyer et al., 2007;Myers et al., 2013;Nachtergaele et al., 2012). However, 20(S)-

Cholesterol can drive the differentiation of spinal cord progenitors
Our mechanistic experiments in cultured fibroblasts led us to ask whether cholesterol could also promote Hh-dependent cell differentiation decisions. In the developing vertebrate spinal cord, the Hh ligand Sonic Hedgehog (SHH) acts as a morphogen to specify the dorsal-ventral pattern of progenitor subtypes ( Figure 5A) (Jessell, 2000). This spatial patterning process can be recapitulated in vitro. Mouse neural progenitors exposed to increasing concentrations of SHH will express transcription factors that mark differentiation towards progressively more ventral neural subtypes: low, medium and high Hh signaling will generate progenitor subtypes positive for Nkx6.1, Olig2, and Nkx2.2, respectively (Dessaud et al., 2008;Gouti et al., 2014;Kutejova et al., 2016).
MβCD:cholesterol induced the formation of both Nkx6.1 + and Olig2 + progenitor subtypes at a low frequency in cultures of mouse spinal cord progenitors (Figures 5B, 5C) and also activated the transcription of Gli1 ( Figure 5D). The efficacy of both Gli1 induction and ventral neural specification induced by MβCD:cholesterol were significantly less than those produced by a saturating concentration of SHH. However, we note that MβCD:cholesterol inclusion complexes could not be delivered at higher concentrations due to deleterious effects on the adhesion and viability of neural progenitors. Taken together, these observations suggest that MβCD:cholesterol is sufficient to activate low-level Hh signals in neural progenitors and consequently to direct differentiation towards neural cell types that depend on such signals.

Discussion
To establish a causal or regulatory role for a component in a biological pathway, experiments should demonstrate that the component is both necessary and sufficient for activity. Cholesterol has been shown to be necessary for SMO activation, based on experiments using inhibitors of cholesterol biosynthesis and high concentrations (~10 mM) of naked MβCD to strip the plasma membrane of cholesterol (Cooper et al., 2003). Impaired SMO activation caused by cholesterol deficiency has also been noted in Smith-Lemli-Opitz syndrome (SLOS), a congenital malformation syndrome caused by defects in the enzyme that converts 7-dehydrocholesterol to cholesterol (Blassberg et al., 2016;Cooper et al., 2003). In contrast to our results, the SMO CRD is dispensable for this permissive role of cholesterol. The depletion of cholesterol reduces signaling by SMO mutants lacking the entire CRD (Myers et al., 2013) or carrying mutations in the CRD binding-groove (Blassberg et al., 2016). By analogy with other GPCRs, these permissive effects are likely to be mediated by the SMO 7TMD.
We now find that cholesterol is also sufficient to activate Hh signalling in a dose-dependent manner. This instructive effect is mediated by the Class F GPCR SMO and maps to its extracellular CRD. Cholesterol engages a hydrophobic groove on the surface of the CRD, a groove that was previously shown to mediate the activating influence of oxysterols (Myers et al., 2013;Nachtergaele et al., 2013;Nedelcu et al., 2013) and represents an evolutionarily conserved mechanism for detecting hydrophobic small-molecule ligands (Bazan and de Sauvage, 2009). An analogous mechanism is present in the Frizzled family of Wnt receptors, where the Frizzled CRD binds to the palmitoleyl group of Wnt ligands, an interaction that is required for Wnt signaling (Janda et al., 2012). Thus, the instructive effects of cholesterol revealed in our present study and the permissive effects of cholesterol reported previously map to distinct, separable SMO domains.
There are many reasons why this activating effect of cholesterol on Hh signalling may not have been appreciated previously despite the fact that the activating effects of side-chain oxysterols have been known for a decade (Corcoran and Scott, 2006;Dwyer et al., 2007). First, the method of delivery, as an inclusion complex with MβCD, is critical to presenting cholesterol, a profoundly hydrophobic and insoluble lipid, in a bioavailable form capable of activating Smo.
Even clear solutions of cholesterol in the absence of carriers like MβCD contain microcrystalline deposits or stable micelles that sequester cholesterol (Haberland and Reynolds, 1973). In contrast, side-chain oxysterols, which harbor an additional hydroxyl group, are significantly more hydrophilic and soluble in aqueous solutions, shown by their ~50-fold faster transfer rates between membranes (Theunissen et al., 1986). Second, cholesterol levels in the cell are difficult to manipulate because they are tightly controlled by elaborate homeostatic signalling mechanisms (Brown and Goldstein, 2009). MβCD:cholesterol inclusion complexes have been shown to be unique in their ability to increase the cholesterol content of the plasma membrane rapidly at timescales (~1-4 hours) at which cytoplasmic signaling pathways operate (Christian et al., 1997;Yancey et al., 1996). Other methods of delivery using low density lipoprotein particles and lipid dispersions, or mutations in genes regulating cholesterol homeostasis, function on a much slower time scale and are thus more likely to be confounded by indirect effects given the myriad cellular processes affected by cholesterol (Christian et al., 1997). Finally, the bell-shaped Hh signal-response curve ( Figure 1A) implies that MβCD:cholesterol must be delivered in a relatively narrow, intermediate concentration range (1-2 mM) to observe optimal activity, with higher (>5 mM) concentrations commonly used to load cells with cholesterol producing markedly lower levels of signaling activity.
Our results are particularly informative in light of the recently solved crystal structure of SMO, unexpectedly found to contain a cholesterol ligand in its CRD groove ( Figure 4A) . Molecular dynamics simulations showed that cholesterol can stabilize the extracellular domains of SMO , but the function of this bound cholesterol, whether it is an agonist, antagonist or co-factor, remains an important unresolved question in SMO regulation.
Structure-guided point mutations in CRD residues that form hydrogen-bonding interactions with the 3β-hydroxyl of cholesterol, reduced signaling by cholesterol ( Figure 4C) making it likely that cholesterol activates SMO by binding to the CRD in the pose revealed in the structure ( Figure   4A). Thus, the cholesterol-bound SMO structure may very well represent an active-state conformation of the CRD.
A surprising feature of the structure is that CRD-bound cholesterol is located at a considerable distance (~12 Å) away from the membrane, which would require a cholesterol molecule to desolvate from the membrane and become exposed to water in order to access its CRD binding pocket   (Figure 6). The kinetic barrier, or the activation energy ( , for this transfer reaction is predicted to be high (~20 kcal/mole), based on the for cholesterol transfer between two acceptors through an aqueous environment (Yancey et al., 1996). The unique ability of MβCD to shield cholesterol from water while allowing its rapid transfer to acceptors would allow it to bypass this kinetically unfavorable step by delivering it to the CRD binding site ( Figure 6). These considerations present a regulatory puzzle for future research: how does cholesterol gain access to the CRD-binding pocket without MβCD and is this process regulated by native Hh ligands? Indeed, the kinetic barrier for cholesterol transfer to the CRD pocket makes it an ideal candidate for a rate-limiting, regulated step controlling SMO activity in cells.
MβCD:cholesterol was consistently less active than the native ligand SHH in our assays which has led to view that this site does not regulate physiological signaling (Myers et al., 2013;Yauch et al., 2009). In contrast, mutations in the cholesterol-binding site impaired responses to SHH . Hence, a putative alternate ligand would have to engage a third, undefined site. Lastly, the presence of active PTCH1 is a major difference between SHH-and MβCD:cholesterol-induced signaling. The biochemical activity of PTCH1 (which is inactivated by SHH) may oppose the effects of MβCD:cholesterol, limiting signaling responses.
Interestingly, MβCD:cholesterol was able to restore maximal Hh responses in the absence of PTCH1 ( Figure 3D).
Our results may have implications for understanding how PTCH1 inhibits SMO, a longstanding mystery in Hh signaling. The necessity and sufficiency of cholesterol for SMO activation, mediated through two different regions of the molecule, means that SMO activity is likely to be highly sensitive to both the abundance and the accessibility of cholesterol in its membrane environment. Furthermore, PTCH1 has homology to a lysosomal cholesterol transporter, the Niemann-Pick C1 (NPC1) protein (Carstea et al., 1997), and PTCH1 has been purported to have cholesterol binding and transport activity (Bidet et al., 2011). Thus, our work supports a model where PTCH1 may inhibit SMO by reducing cholesterol content or cholesterol accessibility (or chemical activity) in a localized membrane compartment (such as the base of primary cilia) that contains SMO, leading to alterations in SMO conformation or trafficking (Bidet et al., 2011;Incardona et al., 2002;Khaliullina et al., 2009). Further tests of this hypothesis will require analysis of the biochemical activities of purified SMO and PTCH1 reconstituted into cholesterol-containing membranes. While cholesterol is an abundant lipid, clearly critical for maintaining membrane biophysical properties and for stabilizing membrane proteins, our work suggests that it may be also used as a second messenger to instruct signaling events at the cell surface through GPCRs and perhaps other cell-surface receptors.

Competing Interests
None

Reagents and Cell Lines
NIH/3T3 and 293T cells were obtained from ATCC, Smo -/fibroblasts have been described previously (Varjosalo et al., 2006)  NHS ester). Ent-cholesterol was synthesized as described previously (Jiang and Covey, 2002).
Antibodies against GLI3 and GLI1 were from R&D Systems (AF3690) and Cell Signaling Technologies (Cat#L42B10) respectively. Human SHH carrying two isoleucine residues at the N-terminus and a hexahistidine tag at the C-terminus was expressed in Escherichia coli Rosetta(DE3)pLysS cells and purified by immobilized metal-affinity chromatography followed by gel-filtration chromatography as described previously (Bishop et al., 2009).

Methyl-β-Cyclodextrin Sterol Complexes
Sterols were dissolved in a mixture of chloroform-methanol (

Data Analysis
Each experiment shown in the paper was repeated at least three independent times with similar results. All data was analyzed using GraphPad Prism. All points reflect mean values calculated from at least 3 replicates and error bars denote standard deviation (SD). The statistical tests used to evaluate significance are noted in the figure legends. Statistical significance in the figures is denoted as follows: ns: p>0.05, *: p≤0.05, **p≤0.01, ***p≤0.001, ****p≤0.0001.

NIH/3T3 cells were cultured in Dulbecco's modified Eagle's Medium (DMEM) containing 10%
Fetal Bovine Serum (FBS, Optima Grade, Atlanta Biologicals) in 24-well plates at an initial density of 7.5 x 10 4 on acid-washed glass cover-slips that were pre-coated with poly-L-lysine.
Confluent cells were exchanged into 0.5% FBS DMEM to induce ciliogenesis for 24 hours.

Image Analysis
Image processing for ciliary SMO levels was carried out using maximum projection images of the acquired Z-stacks using ImageJ. For quantification of ciliary Smo, first a mask was constructed using the Arl13b image (primary cilia marker), and then the mask was applied to the corresponding Smo image where the integrated fluorescence was measured. An identical region outside the cilia was measured to determine background fluorescence. Background correction was applied on a per cilia basis by subtracting the background fluorescence from the cilia fluorescence.
For neural differentiation experiments, fluorescent images were collected on a Leica TCS SP8 confocal imaging system equipped with a 40x oil immersion objective using the Leica Application Suite X (LASX) software. For each experiment, coverslips from each condition were grown, collected, and processed together to ensure that the cells were fixed and stained for the same duration of time. To ensure uniformity in imaging, the gain, offset, and laser power settings on the microscope were held constant for each antibody. At least 15 images were taken per condition. To ensure all cells were represented, z-stacks were acquired and counts were performed on the compressed images. Cell counts were conducted using the NIH ImageJ software suite with cell counter plugin. In total, 5000-6000 cells were analyzed per condition.
The experiment was conducted independently a total of three times. Representative images shown in Figure 5 were processed equally using Adobe Photoshop, Adobe illustrator, and CorelDraw software.

Cholesterol Quantification
Cells were cultured in Dulbecco's modified Eagle's Medium (DMEM) containing 10% Fetal Bovine Serum (FBS, Optima Grade, Atlanta Biologicals) in 6-well plates at an initial density of 3 x 10 5 cells / well. Confluent cells were switched into 0.5% FBS DMEM to induce ciliogenesis for 24 hr. Cells were treated with indicated drugs dissolved in 0.5% FBS DMEM in duplicate. One sample was used to measure total protein by bicinchoninic acid assay (BCA), and the second for total lipid extraction and subsequent cholesterol quantification. Cells were washed once with Phosphate Buffered Saline (PBS), and harvested using a Corning cell lifter in PBS. The cell suspension was transferred to a 1.5 mL ependorf tube, centrifuged at 1000 x g and the PBS aspirated. Total lipids were extracted from the cell pellet by the addition of 200 μL of chloroformmethanol (2:1 vol/vol). To induce phase separation, 100 μL of PBS was added to the lipid extract and the sample was centrifuged at 5,000 x g for 5 minutes. The organic layer was transferred to a fresh 1.5 mL ependorf tube and the solvent removed under reduced pressure.
Relative total free cholesterol was measured using the Amplex Red Cholesterol Assay Kit (Thermo Fisher Scientific) following the manufacturer's instructions. Lysis buffer containing 50 mM Tris pH 7.4, 150 mM NaCl, 2% Nonidet P-40, 0.5% sodium deoxycholate, 0.1% sodium dodecyl sulfate, 1 mM dithithreitol, and Sigma Fast protease inhibitor cocktail (Sigma-Aldrich) was used to disrupt the cell pellet. A ratio of total free cholesterol to total protein was used as a normalization method.   Asterisks denote statistical significance for difference from the untreated sample using one-way ANOVA with a Holm-Sidak post-test.      subsequently require a second transfer step from the membrane to the CRD. The activation energy for the direct delivery mechanism on the left (<10 kcal/mole) is much lower than for the mechanism on the right (~20 kcal/mole), where cholesterol has to desolvate from the membrane without a carrier to access the CRD site (Lopez et al., 2011;Yancey et al., 1996).  Asterisks denote statistical significance for difference from the untreated sample using one-way ANOVA with a Holm-Sidak post-test.