Sphingomyelin suppresses Hedgehog signaling by restricting cholesterol accessibility at the ciliary membrane

Transmission of the Hedgehog signal across the plasma membrane by Smoothened is proposed to be triggered by its direct interaction with cholesterol. But how is cholesterol, an abundant lipid, regulated tightly enough to control a signaling system that can cause birth defects and cancer? Using toxin-based sensors that distinguish between distinct pools of cholesterol, we find here that Smoothened activation and Hedgehog signaling are driven by a biochemically defined fraction of membrane cholesterol, termed accessible cholesterol. Increasing accessible cholesterol levels by depletion of sphingomyelin, which sequesters cholesterol in complexes, potentiates Hedgehog signaling. By inactivating the transporter-like protein Patched 1, Hedgehog ligands trigger an increase in cholesterol accessibility in the ciliary membrane, the subcellular location for Smoothened signaling. Thus, compartmentalization of Hedgehog signaling in the primary cilium may allow cholesterol accessibility to be used as a second messenger to mediate the communication between Patched 1 and Smoothened, without causing collateral effects on other cellular processes.


Introduction
A long-standing mystery in Hedgehog (HH) signaling is how Patched 1 (PTCH1), the receptor for HH ligands, inhibits Smoothened (SMO), a G-protein-coupled receptor (GPCR) family protein that transduces the HH signal across the membrane (Kong, Siebold, and Rohatgi 2019) . The observation that cholesterol can directly bind and activate SMO has led to the proposal that PTCH1 regulates SMO by restricting its access to cholesterol (E. F. Byrne et al. 2016;Luchetti et al. 2016;P. Huang et al. 2016) . Structural and biochemical studies have confirmed that PTCH1 could be a cholesterol transporter; however, transport activity has not yet been demonstrated either in a purified system or at endogenous expression levels in cells Gong et al. 2018;X. Qi et al. 2018;Bidet et al. 2011;C. Qi et al. 2018;Qian et al. 2019) . Furthermore, the resolution of the PTCH1 cryo-EM structures is not high enough to distinguish cholesterol from other sterol lipids as PTCH1 substrates.
A challenge to this model is presented by the fact that cholesterol constitutes up to 50% of the lipid molecules in the plasma membrane (Y. Lange et al. 1989; A. Das et al. 2013;Touster et al. 1970;Colbeau, Nachbaur, and Vignais 1971) : how can such an abundant lipid be kept away from SMO to prevent inappropriate pathway activation? Indeed, other less abundant lipids can bind and regulate SMO activity, including oxysterols, phosphoinositides, endocannabinoids and arachidonic acid derivatives (Nachtergaele et al. 2012;Khaliullina et al. 2015;Arensdorf et al. 2017;Jiang et al. 2016) . Side-chain oxysterols, synthesized through the enzymatic or non-enzymatic oxidation of cholesterol, are appealing alternatives to cholesterol because of their lower abundance, higher hydrophilicity and structural similarity to cholesterol (Corcoran and Scott 2006;Dwyer et al. 2007) . To address the issue of the endogenous lipidic activator of SMO, we took an unbiased genetic approach to identify lipid-related genes whose loss influences the strength of HH signaling.

A focused CRISPR screen targeting lipid-related genes
Using our previously described strategy (Pusapati, Kong, Patel, Krishnan, et al. 2018) to identify positive and negative regulators of the HH pathway, we conducted focused loss-of-function CRISPR screens using a custom library targeting 1,244 lipid-related genes compiled by the LIPID MAPS consortium ( Supplementary Table 1 provides a list of all genes and guide RNAs in the library). This CRISPR library targeted all annotated genes encoding enzymes involved in the synthesis or metabolism of lipids as well as proteins that bind to or transport lipids. We used a previously characterized NIH/3T3 cell line (NIH/3T3-CG) that expresses C as9 and G FP driven by a HH-responsive fluorescent reporter (GLI-GFP) (Pusapati, Kong, Patel, Krishnan, et al. 2018) .
To ensure that HH signaling in these cells would be sensitive to perturbations of endogenous lipid metabolic pathways, the entire population of mutagenized cells was grown in lipoprotein-depleted media for one week prior to the screen and then treated with U18666A, a drug that prevents absorption of exogenous cholesterol from the lysosome, to minimize any sterol uptake from the media ( Fig.1A ). In the screen for positive regulators (hereafter the "HiSHH-Bot10%" screen) , we treated cells with a saturating concentration of the ligand S onic H edge h og (HiSHH) and used Fluorescence Activated Cell Sorting (FACS) to collect poor responders, those with the lowest 10% of GLI-GFP fluorescence ( Figs.1A and 1B ; full screen results in Supplementary Table   2 ). In the screen for negative regulators, we treated cells with a low, sub-saturating concentration of SHH (LoSHH) that activated the reporter to <10% of maximal strength and selected super-responders, cells with the top 5% of GLI-GFP fluorescence ("LoSHH-Top5%" screen) ( Figs.1A and 1C; full screen results in Supplementary Table 3 ).
The screens correctly identified all four positive controls included in the library: Smo and Adrbk1 (or Grk2 ) as positive regulators and Ptch1 and Sufu as negative regulators ( Figs.1B and 1C ). In addition, genes previously known to influence HH signaling ( Gnas ) and protein trafficking at primary cilia ( Inpp5e ) were amongst the most significant hits (Regard et al. 2013;Garcia-Gonzalo et al. 2015;Chávez et al. 2015) . In addition to Inpp5e , other genes involved in phosphoinositide metabolism ( Mtmr3 and Plcb3 ) were also significant hits. Pla2g3 , which encodes a secreted phospholipase, was identified as a negative regulator of HH signaling, an effect that may be related to its known role as a suppressor of ciliogenesis ( Fig.1D ) (Gijs et al. 2015; Kim et al. 2010) .
To identify lipid species that influence HH signaling, we separately analyzed the intersection of all genes expressed in NIH/3T3-CG cells based on RNAseq and annotated as part of a lipid metabolic pathway in the Kyoto Encyclopedia of Genes and genomes (KEGG) ( Fig.1D ; gene lists used for each pathway are shown in Supplementary Table 4 ; RNAseq data in Supplementary Table 5 ). Statistically significant hits clustered in two major pathways: (1) genes encoding enzymes in the cholesterol biosynthesis pathway were positive regulators of HH signaling and (2) genes encoding enzymes in the sphingolipid biosynthesis pathway were negative regulators (positive regulators are shown in blue and negative regulators in orange in Fig.1D ). We focused on these two pathways for the work described in the rest of this study.

Enzymes in the cholesterol biosynthesis pathway positively regulate Hedgehog signaling
Mutations in Dhcr7 and Sc5d , which encode enzymes that catalyze the terminal steps in cholesterol biosynthesis, impair HH signaling in target cells and cause the congenital malformation syndromes Smith-Lemli-Opitz and lathosterolosis, respectively (Cooper et al. 2003;Blassberg et al. 2016;Porter and Herman 2011;Horvat, McWhir, and Rozman 2011) . In addition to these genes, most of the genes encoding enzymes in the pathway that converts squalene to cholesterol were statistically significant hits with a FDR-corrected p -value threshold of 0.1 in the HiSHH-Bot10% screen ( Fig.2A 1D ). A lone oxysterol synthesis enzyme (CYP7A1) was implicated in an opposite role, a negative regulator, in the LoSHH-Top5% screen ( Fig.1D ).
In summary, the data from our genetic screen supports the view that cholesterol itself, rather than a precursor or a metabolite, is the endogenous sterol lipid that regulates SMO activation. Caveats of genetic screens include their inability to identify genes or pathways that are (1) redundant, (2) required for cell viability or growth, or (3) dependent on non-enzymatic reactions or exogenous molecules supplied by the media.

Cellular sphingomyelin suppresses Hedgehog signaling
Multiple enzymes in the sphingolipid synthesis pathway were statistically significant hits in the LoSHH-Top5% screen, indicating these enzymes are negative regulators of HH signaling strength ( Fig.3A ). Top hits from this screen include Sptlc2 , the first committed step in sphingolipid synthesis from serine and palmitoyl-CoA, as well as Sgms1 , which converts ceramide to s phingo m yelin (SM). The identification of Sgms1 suggests that SM is the relevant product of the sphingolipid pathway that attenuates HH signaling.
Because we were unable to isolate viable NIH/3T3 cell lines entirely depleted of SPTLC2 or SGMS1 protein using CRISPR editing, we used an established pharmacological strategy. Myriocin is a fungal antibiotic that potently inhibits SPTLC2 ( Fig.3A ) and is commonly used to deplete SM in cells (Courtney et al. 2018;Tafesse et al. 2013Tafesse et al. , 2015 . SM depletion by myriocin in NIH/3T3 cells was confirmed using both thin-layer chromatography ( Supplementary Fig.3A ) and flow cytometry of intact cells stained with a fluorescent protein probe (OlyA_E69A) that binds to total SM on the outer leaflet of the plasma membrane ( Supplementary   Fig.3B ) (Endapally et al. 2019) . Myriocin treatment markedly potentiated the response to SHH in NIH/3T3 cells, as measured by the transcriptional induction of Gli1 ( Fig.3B ). This effect was also observed in two additional cell types. In mouse embryonic fibroblasts (MEFs), myriocin was sufficient to activate HH signaling even in the absence of added HH ligands ( Supplementary Fig.3C ). Mouse spinal neural progenitor cells (NPCs) differentiate into Olig2 -expressing motor neuron progenitors in response to moderate concentrations of SHH. Myriocin potentiated the effect of SHH on NPCs, substantially reducing the concentration of SHH required to drive motor neuron differentiation ( Fig.3C ).
Several control experiments established that the potentiating effect of myriocin on HH signaling was caused by the depletion of SM, rather than an unrelated effect. Fumonisin B1, a mycotoxin structurally distinct from myriocin that inhibits a different step in SM synthesis ( Fig.3A ), also amplified HH signaling ( Supplementary Fig.3D ). Second, the potentiating effect of myriocin on HH signaling could be reversed by the exogenous administration of SM ( Fig.3D ). Finally, increasing SM levels in cells using low-dose staurosporine had the predicted opposite effect: reduction of HH signaling strength ( Fig.3E and Supplementary Fig.3E ) (Maekawa et al. 2016) .

Sphingomyelin restrains Hedgehog signaling at the level of Smoothened
SM is mostly localized in the outer leaflet of the plasma membrane where it plays key roles in its lateral organization, including the formation of ordered membrane microdomains that can influence protein trafficking, signaling and other processes Ikonen 2000, 1997) . Therefore, we looked broadly at the effects of myriocin on HH-relevant phenotypes, paying particular attention to primary cilia, organelles that are required for HH signaling in vertebrates (Huangfu et al. 2003) . Myriocin did not significantly alter the abundances of the HH pathway proteins GLI3, SMO, SUFU or PTCH1 ( Supplementary Fig.4A ) and also did not change either the frequency or the length of primary cilia ( Supplementary Fig.4B ). Sensitivity of target cells to HH ligands can be influenced by the ciliary abundances of PTCH1, the receptor for all HH ligands that inhibits SMO, and by GPR161, a G-protein coupled receptor (GPCR) known to negatively regulate HH signaling (Rohatgi, Milenkovic, and Scott 2007;Pusapati, Kong, Patel, Gouti, et al. 2018;Mukhopadhyay et al. 2013) . However, both proteins were properly localized in the ciliary membrane in myriocin-treated cells and were cleared (as expected) from cilia in response to SHH addition ( Supplementary Figs.4C and 4D ). Thus, myriocin does not seem to significantly alter ciliary biogenesis, ciliary morphology or ciliary trafficking events of receptors that negatively regulate HH signaling.
The SHH-triggered accumulation of SMO in primary cilia is required for initiation of HH signaling in the cytoplasm (Corbit et al. 2005) . Myriocin potentiated SMO ciliary accumulation in NIH/3T3 cells, suggesting that SM depletion enhances SMO activation ( Supplementary Fig.4E ). HH signaling in cells treated with myriocin was blocked by the SMO antagonist Vismodegib (  Fig.3C ), we were able to assess the effects of these mutations without the confounding effects of HH ligands or SMO agonists. Previously defined mutations in the sterol-binding CRD site (D99A/Y134F, Fig.4B ), which abrogate cholesterol binding by disrupting a key hydrogen bond with the 3β-hydroxyl of cholesterol, prevented myriocin from activating signaling ( Fig.4C ). In contrast, a mutation (D477G, Fig.4B ) in the TMD site failed to diminish myriocin-induced signaling ( Fig.4C ). In control experiments, SMO-D99A/Y134F and SMO-D477G were responsive to SAG and SHH, respectively, demonstrating protein integrity ( Fig.4C ). The fact that point mutations in SMO abrogated the effect of SM depletion suggests that myriocin influences HH signaling at the level of SMO.

Sphingomyelin restrains Hedgehog signaling by sequestering cholesterol
The observation that mutations in the CRD of SMO, a well-defined binding site for cholesterol (E. F. Byrne et al. 2016) , abrogated the effects of SM depletion ( Fig.4C ) suggested that SM regulates SMO activity by controlling the availability of cellular cholesterol that can bind to SMO. There is a precedent for SM regulation of cholesterol availability in another cellular signaling context, namely the control of cholesterol synthesis (Slotte and Bierman 1988;Scheek, Brown, and Goldstein 1997;Akash Das et al. 2014) . These studies have led to a proposal that plasma membrane cholesterol is organized in three pools: a fixed pool essential for membrane integrity, a SM-sequestered pool with low chemical activity and a third ("accessible") pool with higher chemical activity that is available to interact with proteins ( With these tools in hand, we sought to test the model that SM depletion by myriocin potentiates HH signaling by increasing the pool of accessible cholesterol. Reducing accessible cholesterol in myriocin-treated cells with methyl-β-cyclodextrin (MβCD), measured by PFO* staining, decreased SHH-induced activation of the GLI-GFP reporter back to levels seen in untreated cells ( Fig.5F ). This rescue is not consistent with the alternative possibility that SM negatively regulates SMO either directly or through a different mechanism.
These results, together with the requirement of the cholesterol-binding CRD for the potentiating effect of myriocin ( Fig.4C ), support the model that SM impairs HH signaling by sequestering cholesterol into complexes where it is inaccessible to SMO. This conclusion is also consistent with the observation that purified SMO is

The ciliary membrane is a compartment with low cholesterol accessibility
The regulation of SMO by PTCH1 occurs at primary cilia, the only post-Golgi compartment in the cell where both proteins can be found localized together (Rohatgi et al. 2009;Rohatgi, Milenkovic, and Scott 2007) . Since the ciliary membrane is thought to have a different lipid (and protein) composition than the plasma membrane (recently reviewed by (Nachury and Mick 2019) ), we compared PFO*, OlyA and OlyA_E69A staining in the ciliary membrane relative to the plasma membrane using confocal microscopy. Controls confirmed that these probes could be used to measure levels of cholesterol, SM-cholesterol complexes and total SM in the ciliary membrane using quantitative fluorescence microscopy ( Supplementary Figs.6A-6C ), analogous to how we used them to measure these species at the plasma membrane by flow cytometry ( Figs.5B-5E ).
For each cilium imaged, we calculated the ratio of mean ciliary fluorescence to mean plasma membrane fluorescence in a region surrounding the cilium (hereafter the "C/P ratio") (Geneva, Tan, and Calvert 2017) . We used this metric because myriocin treatment or cholesterol loading will lead to changes in probe staining at both the plasma membrane and the ciliary membrane. The C/P ratio reflects changes in the ciliary membrane relative to changes in the plasma membrane: if probe staining increases by the same factor in the plasma membrane and the ciliary membrane the C/P ratio will remain unchanged.
In untreated cells, the C/P ratio was significantly higher for OlyA_E69A staining compared to OlyA or PFO* staining ( Fig.6A ). This suggests that the ratio of SM to cholesterol, which determines the abundance of accessible cholesterol, is higher in the ciliary membrane compared to the plasma membrane. Indeed, while OlyA_E69A staining of total SM was readily detectable in cilia, most cilia did not show distinctive staining for ). Indeed, myriocin had a greater effect on the C/P ratio of accessible cholesterol compared to even cholesterol loading, suggesting that SM provides the major restraint on cholesterol accessibility at cilia ( Fig.6C ). Taken together, we conclude that the ratio of sphingomyelin to cholesterol is higher in the ciliary membrane compared to the plasma membrane, leading to reduced cholesterol accessibility. This may be critical for keeping SMO, which is cycling through the ciliary membrane even in the absence of SHH, in an inactive state by restricting its access to cholesterol (Ocbina and Anderson 2008) . Reducing SM levels with myriocin (or increasing cholesterol levels by cholesterol loading) increases accessible cholesterol in the ciliary membrane and, consequently, potentiates SMO activation.
Hedgehog ligands cause an increase in cholesterol accessibility at primary cilia PTCH1, the receptor for HH ligands, is thought to inhibit SMO by reducing its access to cholesterol using its transporter-like activity. Since PTCH1 is localized in and around the cilium, we have proposed that it could function to inhibit SMO by reducing accessible cholesterol in the ciliary membrane  Siebold, and Rohatgi 2019; P. Huang et al. 2016) . This model predicts that SHH, which inhibits PTCH1 activity, should lead to an increase in accessible cholesterol and PFO* staining at the ciliary membrane. Given the potential artifacts associated with overexpressing a transporter protein, we sought to measure changes in accessible cholesterol at endogenous PTCH1 expression levels.
SHH did not induce much of a change in PFO* staining of the bulk plasma membrane, measured by FACS ( Fig.7A ). A lack of an effect is not surprising because changes in overall cholesterol accessibility would influence many other cellular processes, including the signaling system that maintains cholesterol homeostasis  Fig.7A ). In addition, activation of signaling with SAG, which bypasses PTCH1 and directly activates SMO, also did not cause significant changes in accessible cholesterol at the ciliary membrane ( Supplementary Fig.7B ). Both controls show that the SHH-induced changes in accessible cholesterol at primary cilia are dependent on PTCH1 activity. In the absence of myriocin, we detected only minimal changes in PFO* staining in cilia ( Fig.7B ). We believe that this is related to a well-known property of PFO*: it only binds to membranes when cholesterol rises above a specific threshold and becomes accessible (A. Das et al. 2013;Flanagan et al. 2009;Sokolov and Radhakrishnan 2010) . In untreated cells, accessible cholesterol levels at cilia are too low (even with SHH) to allow sufficient probe binding. However, SM depletion induced by myriocin raised accessible cholesterol levels above the threshold levels required for PFO* binding, thus allowing us to detect increases in ciliary accessible cholesterol induced by SHH.

Discussion
The results of our unbiased screen for lipid-related genes that influence the strength of HH signaling uncovered two pathways--cholesterol and SM synthesis--that both converge on accessible cholesterol as the critical species that regulates the interaction between PTCH1 and SMO. The potentiating effect of SM depletion on HH signaling points to cholesterol itself as the regulatory sterol, since side-chain oxysterols do not form We end with a speculative answer to an enigma in HH signaling: why is the HH pathway dependent on primary cilia in vertebrates but not in Drosophila ? The predominant sterol in Drosophila is ergosterol, with cholesterol itself representing <5% of membrane sterols (Rietveld et al. 1999) . In addition, flies are cholesterol auxotrophs: they acquire cholesterol from the diet and have lost many of the genes for cholesterol biosynthesis (Vinci, Xia, and Veitia 2008;Carvalho et al. 2010) . Thus, flies do not need the regulatory machinery that monitors accessible cholesterol in the plasma membrane and adjusts the transcription of cholesterol biosynthetic genes. Indeed, the SREBP pathway, which monitors accessible cholesterol in vertebrates, has been repurposed in Drosophila to respond to phosphatidylethanolamine (Dobrosotskaya et al. 2002) .
Cholesterol levels in Drosophila are instead sensed by a very different nuclear receptor-based mechanism (Bujold et al. 2010) . We propose that the lack of a need to regulate cholesterol biosynthetic pathway genes abrogates the need to sequester HH signaling in primary cilia in insects. Reagents used for cell culture are discussed in the methods for Cell Culture and Drug Treatments.

Guide RNA library targeting lipid-related genes
In order to generate a lipid-library, human lipid-modifying genes/proteins were downloaded from the LIPID MAPS Proteome Database ("LIPID MAPS Proteome Database," n.d.  (Joung et al. 2017) .

CRISPR/Cas9 screen targeting lipid-related genes
Our reporter-based screening platform has been described previously in detail (Lebensohn et al. 2016;Pusapati, Kong, Patel, Krishnan, et al. 2018) . NIH/3T3-CG cells were used because they respond to SHH in a concentration-dependent manner, carry stably integrated Cas9, and carry fluorescence-based, quantitative reporter of HH signaling (GLI-GFP) (Pusapati, Kong, Patel, Krishnan, et al. 2018) . This reporter allows the isolation of cell populations with enhanced or reduced HH signaling phenotypes by FACS. CRISPR library amplification, lentiviral production, functional titer determination and transduction were carried out as previously described in detail (Joung et al. 2017;Pusapati, Kong, Patel, Krishnan, et al. 2018) .
To prepare the library of cells for screening, 4x 15 cm plates were seeded with 5 million cells each. were sorted for the lowest 10% (HiSHH-Bottom10% screen) or highest 5% of GFP fluorescence (HiSHH-Top5% screen). Finally, genomic DNA was extracted from the unsorted and sorted cells and the sgRNA library was amplified by nested PCR, subjected to Illumina sequencing, and analyzed using the MAGeCK algorithm as described previously (Pusapati, Kong, Patel, Krishnan, et al. 2018;Li et al. 2014) .

Kyoto Encyclopedia of Genes and genomes (KEGG) analysis of Lipid Pathways
In order to determine which lipids influence Hedgehog signaling, mouse-specific genes were manually curated into lists for each lipid metabolic pathway identified on the Kyoto Encyclopedia of Genes and genomes (KEGG)  Table 5 ). If a gene was not expressed (a value of 0 in both RNAseq data sets), it was not included in the pathway analysis of Fig1D. Methyl-β-cyclodextrin complexes were generated as described previously ) . Egg sphingomyelin: fatty acid free BSA complexes were made by dissolving in Optimem (Gibco) at a molar ratio of 1000:1 (sphingomyelin:BSA) followed by water-bath sonication.

Neural Progenitor Differentiation Assays
To assess the effect of myriocin on Hedgehog (HH) signaling in a more physiological, differentiation based assay, we used the HM1 mESC line described previously (Pusapati, Kong, Patel, Gouti, et al. 2018) . This cell line harbors the GLI1-Venus and OLIG2-mKate dual reporter system to evaluate the strength of HH signaling output both through Gli1 target gene induction and the Olig2 differentiation marker for motor neuron progenitors. After growth and maintenance of mESC on feeder cells, the cells were plated on 6-well gelatin-coated CellBIND plates (Corning) at a density of 100,000 cells/well for flow cytometry analysis.

Hedgehog signaling assays
Gli1 mRNA transcript levels were measured using the Power SYBR Green Cells-to-CT kit (Thermo Fisher Scientific). Gli1 levels relative to Gapdh were calculated using the Delta-Ct method (CT( Gli1 ) -CT( Gapdh )).
The RT-PCR was carried out using custom primers for Gli1 (forward primer: 5′-ccaagccaactttatgtcaggg-3′ and reverse primer: 5′-agcccgcttctttgttaatttga-3′), and Gapdh (forward primer: 5′-agtggcaaagtggagatt-3′ and reverse primer: 5′-gtggagtcatactggaaca-3′). For analysis of GLI1-GFP in NIH/3T3-CG cells by flow cytometry, cells were harvested by trypsinzation followed by quenching in low (0.5%) serum media at 4°C. Cells were either analyzed by flow cytometry immediately or spun and processed for lipid probe staining. All lipid probes were labeled with red or far-red fluorescent dyes, so GLI-GFP reporter expression and probe staining could be measured simultaneously by two-channel flow cytometry.

Purification and labeling of lipid probes
Mutant Perfringolysin O (PFO*) was purified as previously described (Akash Das et al. 2014) and covalently labeled with Alexa Fluor 647 following the manufacturer's instructions. Expression, purification and labeling of OlyA and OlyA_E69A variants containing a single cysteine has been described (Endapally et al. 2019) . Both proteins were expressed in Escherichia coli Rosetta(DE3)pLysS cells, purified by metal-affinity and gel-filtration chromatography and finally labelled with Atto-647 maleimide dye following the manufacturer's instructions (Sigma Aldrich, product #05316).

Measurement of whole-cell lipid probe staining by flow cytometry
To stain cells with these labeled probes, they were harvested for flow cytometry by trypsinization followed by quenching in ice-cold low (0.5% FBS) serum supplemented DMEM. All subsequent steps were completed on ice. Cells were spun at 1000g and then resuspended in Probe Blocking Buffer (PBB, 1x PBS with 10mg/mL BSA). After 10 minutes in PBB, cells were spun and then resuspended in PBB containing desired probes at the following final concentrations: 5 µg/mL PFO*, 2 µM OlyA and 2 µM OlyA_E69A. Cells were stained for 1 hour and then washed 3 times in PBB before flow cytometry.

Measurement of ciliary probe staining by microscopy
Probe staining at cilia was carried out using NIH/3T3 or Ptch -/cells stably expressing ARL13B-GFP as a cilia marker to avoid the use of detergents for permeabilization. For PFO* probe staining, the coverslips were transferred to an ice-cold metal rack and intact cells were stained in PFO* (at a final concentration of 5 µg/mL) diluted into ice cold low (0.5% FBS) serum supplemented DMEM for 30 minutes. For OlyA and OlyA_E69A probe staining, live cells were stained at room temperature in probe (at a final concentration of 2 µM for both OlyA and OlyA_E69A) diluted in room temperature low (0.5% FBS) serum supplemented DMEM for 10 minutes. After staining, cells were washed with 1x PBS and then immediately fixed in 4% PFA in 1xPBS for 10 minutes. Coverslips were then washed three times with 1x PBS and mounted on glass slides in antifade mountant media (ProLong Diamond Antifade Mountant, Thermo Fisher Scientific) where they cured overnight at room temperature before imaging.

Thin layer chromatography
Cells grown in the presence or absence of myriocin (40 μM for 3 days) were harvested by washing with 4°C 1x PBS, scraping and spinning at 1000g. Lipids were extracted using chloroform/methanol/water (2:2:1). Lipid extracts were loaded onto a TLC plate (Millipore) along with egg and milk sphingomyelin as lipid standards.

Analysis of Lipid Probe Staining at Cilia by Immunofluorescence
Images were obtained using a Leica TCS SP5 confocal imaging system containing a 63x oil immersion objective. Quantification of probe staining at cilia was performed using code written in MATLAB R2014b using the following steps. Leica Image Files (LIF) were converted into matrices using the bfmatlab toolbox. Following a max-z-projection, cilia were identified by applying a two-dimensional median filter followed by a high-pass user-defined threshold for signal versus noise to generate a cilia mask in the cilia channel. Any group of contiguous pixels that had signal was labeled as a "potential cilium." Each potential cilium was then subjected to a series of tests (measuring area, eccentricity, solidity, intensity, and length). If each test was passed, the "true cilium" pixels were then mapped to a matrix containing the pixels in the lipid probe channel. In the lipid probe channel, each focal plane was measured independently to avoid noise caused by probe staining outside of the focal plane of a given cilium. The average intensity of pixels falling within a given cilia mask were measured for each focal plane and these values were recorded in a matrix. In order to measure the local plasma membrane fluorescence around each cilium, the cilia mask was dilated to a user-defined size, and the initial cilia mask was subtracted from the dilated cilia mask in order to create the "plasma membrane mask." Plasma membrane probe staining was then measured by averaging the pixel intensities within the plasma membrane mask in each focal plane. To generate the final intensity value for each cilium, the focal plane containing the highest average ciliary probe staining was normalized (either by subtraction or division when specified) to the focal plane containing the highest average plasma membrane probe staining. This MATLAB code is available on GitHub ( https://github.com/mkinnebr/lipids-HH ).

Statistics
The statistical significance of differences between two groups was determined by an unpaired, nonparametric t-test (Mann-Whitney test). When three or more groups were compared a nonparametric ANOVA (Kruskal-Wallis test) was used. All experiments were repeated at least 3 times.   Figure 1. CRISPR screens identify lipid-related genes that influence Hedgehog signaling. (A) Flowchart summarizing the screening strategy. Inset highlights the dual strategy of lipid depletion and lysosomal cholesterol import inhibition used during the screen.

Figure 1
(B and C) Volcano plots of the HiSHH-Bot10% (B) screen for positive regulators and the LoSHH-Top5% (C) screen for negative regulators. Enrichment is calculated as the mean of all sgRNAs for a given gene in the sorted over unsorted population, with the y-axis showing significance based on the false discovery rate (FDR)-corrected p-value.
(D) Screen results analyzed by grouping genes based on the core lipid biosynthetic pathways in KEGG. In all panels, genes identified as positive and negative regulators are labeled in blue and orange respectively.  (C) Differentiation of spinal Neural Progenitor Cells (NPCs) into OLIG2-postive motor neuron progenitors exposed to either LoSHH (5 nM) or HiSHH (25 nM) was assessed using flow cytometry to measure the fluorescence of a OLIG2-mKate differentiation reporter (n > 5000 cells for each treatment). (D) HH signaling strength in NIH/3T3-CG Reporter cells treated with LoSHH (5 nM) or HiSHH (50 nM) after treatment with myriocin alone or myriocin followed by addition of exogenous egg SM. Each data point represents the mean GLI1-GFP fluorescence derived from ~250 cells. (E) HH signaling strength measured in NIH/3T3 cells treated with either LoSHH (5 nM) or HiSHH (25 nM) in the presence or absence of 50 nM staurosporine to increase SM. Statistical significance was determined by the Mann-Whitney test (C, D and E); *p-value≤ 0.05, ***p-value≤ 0.001, ****p-value≤ 0.0001.