Endocannabinoids are conserved inhibitors of theHedgehog pathwayHelena Khaliullinaa,1, Mesut Bilginb, Julio L. Sampaiob, Andrej Shevchenkob, and Suzanne Eatonb,1

aDepartment of Physiology, Development and Neuroscience, University of Cambridge, CB2 3DY Cambridge, United Kingdom; and bMax Planck Institute ofMolecular Biology and Genetics, 01307 Dresden, Germany

Edited by Brian K. Kobilka, Stanford University School of Medicine, Stanford, CA, and approved February 2, 2015 (received for review August 27, 2014)

Hedgehog ligands control tissue development and homeostasis byalleviating repression of Smoothened, a seven-pass transmem-brane protein. The Hedgehog receptor, Patched, is thought toregulate the availability of small lipophilic Smoothened repressorswhose identity is unknown. Lipoproteins contain lipids required torepress Smoothened signaling in vivo. Here, using biochemicalfractionation and lipid mass spectrometry, we identify these re-pressors as endocannabinoids. Endocannabinoids circulate in humanand Drosophila lipoproteins and act directly on Smoothened atphysiological concentrations to repress signaling in Drosophila andmammalian assays. Phytocannabinoids are also potent Smo inhibi-tors. These findings link organismal metabolism to local Hedgehogsignaling and suggest previously unsuspected mechanisms for thephysiological activities of cannabinoids.

endocannabinoids | smoothened | hedgehog | lipids | lipoproteins

Hedgehog (Hh) signaling regulates growth and differentiationduring embryonic development and adult tissue homeostasis(1). It is inappropriately activated in many tumors (2) and alsoinfluences physiological functions including lipid and sugar me-tabolism (3, 4), nocioception (5), the response to ischemia (6),and immune activation (7, 8). Different classes of compoundswith activity toward Hh signaling could have broad therapeuticapplications.Hh proteins are lipid-modified secreted ligands that can associate

with lipoprotein particles (9–12). Hh signals by binding the 12-transmembrane domain protein Patched (Ptc), which prevents Ptcfrom repressing the seven-pass transmembrane protein Smooth-ened (SMO) (13). Ptc-dependent SMO repression involves alter-ations to SMO trafficking. In vertebrates, Ptc activity prevents SMOaccumulation in the primary cilium (14, 15), whereas in Drosophila,it blocks accumulation on the basolateral membrane (16). However,Ptc must exert additional effects on SMO that are independent ofSMO trafficking; localization to primary cilia, or to the basolateralmembrane, is insufficient for full activation of SMO signaling (11,17, 18). Once it is active, SMO signaling influences transcription byregulating the processing of Gli family proteins and exerts addi-tional effects on cell migration and metabolism by Gli-independentmechanisms (1, 19–21).Hh signaling can be targeted at different levels (2). Antibodies

to mammalian Sonic Hedgehog (Shh) block ligand activity, andchemical inhibitors of Gli function block its transcriptional output.SMO is especially sensitive to chemical inhibition, and its activitycan be modulated by a variety of exogenous and endogenous smallmolecules that interact with SMO at different binding sites. Ptc issimilar to resistance-nodulation-division (RND) family transporters,which use proton gradients to transport lipophilic molecules acrossmembranes, and exogenous SMO inhibitors are thought to mimicthe activity of endogenous molecules whose availability is regulatedby Ptc. However, the identity of these endogenous SMOmodulatorsis not clear.SMO activity can be regulated by different small molecules that

bind to distinct sites. Many of these molecules target a pocketformed by the SMO transmembrane domains (22). The steroidalalkaloid cyclopamine (derived from Veratrum californicum) was the

first such inhibitor to be identified (23). Chemical library screens(24, 25) subsequently found many other molecules with positiveand negative effects. Many of these compete with cyclopamine forbinding to SMO, suggesting they target the same site, or influencethe cyclopamine binding site allosterically. Interestingly, both theSMO agonist (SAG) and the SMO antagonist 1 (SANT-1) competewith cyclopamine for SMO binding but compete less well with eachother (26). Thus, the influence of these small molecules on theconformation of the SMO transmembrane domains is likely to becomplex. The endogenous ligands for this site are unclear, althoughvitamin D is a candidate (27). Another binding site in the SMOextracellular domain is targeted by oxysterols, which activate SMO(28, 29). Competition studies and genetic analysis suggest there maybe additional regulatory binding sites on SMO (2, 29). A betterunderstanding of the endogenous compounds that regulate SMOsignaling would provide insights into the logic of the pathway andprovide important clues for drug design.Genetic studies in Drosophila have played a major role in

identifying Hh pathway components and elucidating theirmechanism of action: The wing disk has been a particularlypowerful system for understanding the Hh pathway (30). Recently,we discovered that one or more lipids present in Drosophila lipo-protein particles are required in vivo to keep Hh signaling off inwing discs in the absence of Hh ligand. These lipids destabilizeDrosophila Smoothened (Smo) and promote processing of theDrosophila Gli homolog cubitus interruptus (Ci) (17). Hh associateswith lipoproteins via its lipid anchors, and this association blockstheir repressive function. Mammalian lipoproteins have a similar

Significance

Hedgehog proteins regulate development and tissue homeo-stasis. They signal through activation of the transmembraneprotein Smoothened. Smoothened hyperactivation underliesdevelopment of many tumors. Smoothened activity can bemodulated by several synthetic small molecules, which haveshown promise in the clinic. However, occurrence of resistance-inducing mutations limits their effectiveness. Little is knownabout endogenous small molecules that may inhibit Smooth-ened in vivo. Previous work suggested that lipids present inlipoproteins are required for Smoothened inhibition in vivo.Here, we use biochemical fractionation and lipidomics toidentify these lipids as endocannabinoids and show that theiractivity as Smoothened inhibitors has been conserved from fliesto mammals. Endocannabinoids may provide useful templatesfor the design of new therapeutic Smoothened antagonists.

Author contributions: H.K., A.S., and S.E. designed research; H.K., M.B., J.L.S., and S.E.performed research; H.K., M.B., J.L.S., A.S., and S.E. analyzed data; and H.K. and S.E. wrotethe paper.

The authors declare no conflict of interest.

This article is a PNAS Direct Submission.1To whom correspondence may be addressed. Email: hk402@cam.ac.uk or eaton@mpi-cbg.de.

This article contains supporting information online at www.pnas.org/lookup/suppl/doi:10.1073/pnas.1416463112/-/DCSupplemental.

www.pnas.org/cgi/doi/10.1073/pnas.1416463112 PNAS | March 17, 2015 | vol. 112 | no. 11 | 3415–3420

CELL

BIOLO

GY

Dow

nloa

ded

by g

uest

on

June

28,

202

1

http://crossmark.crossref.org/dialog/?doi=10.1073/pnas.1416463112&domain=pdfmailto:hk402@cam.ac.ukmailto:eaton@mpi-cbg.demailto:eaton@mpi-cbg.dehttp://www.pnas.org/lookup/suppl/doi:10.1073/pnas.1416463112/-/DCSupplementalhttp://www.pnas.org/lookup/suppl/doi:10.1073/pnas.1416463112/-/DCSupplementalwww.pnas.org/cgi/doi/10.1073/pnas.1416463112

repressive activity toward Shh signaling, and association of Shh withthese particles reverses their inhibitory activity in the ShhLIGHT2reporter assay (11). In this study, we use biochemical fractionationand lipid mass spectrometry to identify these inhibitory lipids fromextracts of human very low-density lipoprotein (VLDL).

Results and DiscussionInitial experiments showed that VLDLs carry lipids that represssignaling in both Drosophila discs and ShhLIGHT2 cells (Fig. S1A–C). Therefore, we used VLDL as a starting material to identifythese inhibitory lipids. To reduce the complexity of the lipid pool,we first saponified VLDL extracts to deplete glycerolipids andremoved the resulting free fatty acids. Residual nonsaponifiablelipids retain inhibitory activity in Hh signaling assays (Fig. S1 D′–I).We fractionated saponified extracts by reversed phase high-per-formance liquid chromatography (HPLC), assaying elution frac-tions for inhibitory activity in ShhLIGHT2 cells. This fractionationrevealed several clusters of elution fractions that inhibited Shhsignaling and, surprisingly, one region with stimulatory activity(Fig. 1A).We used shotgun analyses by Fourier transform mass spec-

trometry (FTMS) and tandem mass spectrometry (MS/MS) tosearch for signaling lipids whose concentration peaked in activeHPLC elution fractions (Table S1). Active fractions did not co-incide with peaks of known pathway regulators, such as vitaminD3, 7-dehydrocholesterol, or hydroxysterols (Fig. S2). However,MS/MS analysis identified endocannabinoids and endocannabinoid-related molecules that cofractionated with each region of ac-tivity (Fig. 1B and Table S1). Endocannabinoids consist of fattyacids or alcohols linked to various polar head groups. Arach-idonoyl derivatives of ethanolamine, dopamine, and glycerol arepotent ligands for the cannabinoid receptors CB1 and CB2 (31).

Related molecules with different fatty acid moieties and headgroups have biological activities exerted through a variety of otherreceptors (32). We identified peaks of different N-acylethano-lamide, N-acyldopamine, and 2-alkylglycerol species in repressiveelution fractions and detected peaks of N-acylserines in fractionswith stimulatory activity (Fig. 1 A and B).To explore effects of different endocannabinoid classes and

species on Shh signaling, we assayed synthetically producedcannabinoids in ShhLIGHT2 cells. N-acylserine 16:0 (present instimulatory fractions) increased signaling (Fig. 2A). Surprisingly,N-acylserine 20:4 inhibited signaling (Fig. 2A). We observedsimilar dependence on fatty acid chain length/unsaturation forseveral other classes of endocannabinoids. N-acyl ethanolamides18:2 and 20:4 inhibited Shh signaling, whereas those with shortersaturated fatty acids (16:0 or 18:0) were much less effective(Fig. 2B). Shh signaling was repressed by 2-alkylglycerol 20:4 atconcentrations similar to N-acylethanolamide 20:4 (Fig. 2C). Shhsignaling was also inhibited by 2-acylglycerol 20:4 (2-arachi-donoylglycerol, 2-AG), but not by the 2-acylglycerol speciescontaining the 18:2 fatty acid (Fig. 2E). Thus, N-acylserines,N-acylethanolamides, 2-acylglycerols, and 2-alkylglycerol in-hibit Shh signaling with a potency that depends on the structureof their fatty acid moieties. Species with shorter fatty acids areless active, or even stimulatory in the case of N-acylserine.N-acyldopamines also repress signaling in the ShhLIGHT2 cellassay. Interestingly, however, their activities show the oppositecorrelation with fatty acid chain length and unsaturation; 16:0and 18:0 species are potent inhibitors, but N-acyldopamine20:4 is much less active (Fig. 2D). Cannabis-derived phyto-cannabinoids mimic the effects of endocannabinoids on a vari-ety of receptors. These compounds also robustly suppressShh signaling at concentrations at least 10-fold lower than

Fig. 1. Endocannabinoids in whole serum and active VLDL lipid fractions. (A) Ratio of reporter activity in ShhLIGHT2 cells treated as indicated. HPLC elutionfractions with stimulatory/inhibitory activities are shaded in green/red. Error bars indicate SDs of three independent experiments. n = 3. (B) Distribution ofendocannabinoid species, detected by MS/MS in the HPLC elution fractions tested in A. Structures of each class are shown in boxes next to distributionprofiles. R, fatty acid/alcohol moiety. (C) Concentration (in nanomolars) of endocannabinoids and related molecules in human serum quantified by LC-MS/MSusing the method of multiple reaction monitoring (MRM). AU, arbitrary units.

3416 | www.pnas.org/cgi/doi/10.1073/pnas.1416463112 Khaliullina et al.

Dow

nloa

ded

by g

uest

on

June

28,

202

1

http://www.pnas.org/lookup/suppl/doi:10.1073/pnas.1416463112/-/DCSupplemental/pnas.201416463SI.pdf?targetid=nameddest=SF1http://www.pnas.org/lookup/suppl/doi:10.1073/pnas.1416463112/-/DCSupplemental/pnas.201416463SI.pdf?targetid=nameddest=SF1http://www.pnas.org/lookup/suppl/doi:10.1073/pnas.1416463112/-/DCSupplemental/pnas.201416463SI.pdf?targetid=nameddest=SF1http://www.pnas.org/lookup/suppl/doi:10.1073/pnas.1416463112/-/DCSupplemental/pnas.201416463SI.pdf?targetid=nameddest=SF1http://www.pnas.org/lookup/suppl/doi:10.1073/pnas.1416463112/-/DCSupplemental/pnas.201416463SI.pdf?targetid=nameddest=SF1http://www.pnas.org/lookup/suppl/doi:10.1073/pnas.1416463112/-/DCSupplemental/pnas.201416463SI.pdf?targetid=nameddest=SF1http://www.pnas.org/lookup/suppl/doi:10.1073/pnas.1416463112/-/DCSupplemental/pnas.201416463SI.pdf?targetid=nameddest=ST1http://www.pnas.org/lookup/suppl/doi:10.1073/pnas.1416463112/-/DCSupplemental/pnas.201416463SI.pdf?targetid=nameddest=SF2http://www.pnas.org/lookup/suppl/doi:10.1073/pnas.1416463112/-/DCSupplemental/pnas.201416463SI.pdf?targetid=nameddest=ST1www.pnas.org/cgi/doi/10.1073/pnas.1416463112

endogenous cannabinoids (Fig. 2F). Taken together, these dataindicate that specific endocannabinoids account for the in-hibitory activity in VLDL. The dependence of inhibitory ac-tivity on fatty acid chain length suggests they influence Shhsignaling through specific receptor ligand interactions, and thatN-acyldopamines may act by a different mechanism than otherendocannabinoids.

The most potent endocannabinoids were active at low micro-molar concentrations (Fig. 2 A–E). To assess whether endogenouscirculating endocannabinoids are sufficiently abundant to suppressHh signaling in vivo, we quantified them in human serum by liquidchromatography-tandem mass spectrometry (LC-MS/MS). As re-ported, anandamide was present at nanomolar concentrationsin human serum (33); however, other endocannabinoid specieswere more abundant. Total concentrations of N-acylethanolamides,2-acylglycerols, and N-acylserines were each in the micromolarrange (Fig. 1C). Thus, endocannabinoids circulate at levels suffi-cient to regulate Hh signaling in vivo.Endocannabinoids are rapidly metabolized by intracellular

enzymes, including fatty acid amide hydrolase (FAAH) (34). Wewondered whether endocannabinoids repress the Hh pathway atlower concentrations if their rate of metabolism were reduced.Therefore, we assayed endocannabinoids in the presence of se-lective inhibitor of FAAH licensed by Pfizer (PF-3845), a selec-tive inhibitor of mammalian FAAH. PF-3845 potentiated pathwayinhibition by N-acylethanolamide 18:2 and N-acyldopamine 18:0,allowing them to act at 10-fold lower concentrations (Fig. 2 B andD). Thus, rapid FAAH-dependent endocannabinoid metabolismnormally limits their effects on Shh signaling.Endocannabinoids and related molecules modulate the activ-

ity of many receptors including CB1, CB2, transient receptorpotential vanilloid channel (TRPV) channels, GPR55, peroxi-some proliferator-activated receptor (PPAR)γ, and PPARα. Τoask whether endocannabinoids influenced Shh signaling indirectlythrough these pathways, we assayed the activity of specific agonistsand antagonists of these receptors in ShhLIGHT2 cells (Fig. 3A).None repressed signaling by Shh. Furthermore, comparing ourresults with published values for the activities of different can-nabinoids toward these pathways indicates different patternsof potency. For example, CB1 and TRPV1 are activated byN-acyldopamine 20:4 and not by N-acyldopamine 18:0 or 16:0,whereas the reverse is true for inhibition of Shh signaling(Fig. S3). Moreover, N-acylethanolamide 18:1 and 16:0 activatePPARα better than N-acylethanolamide 20:4, but these com-pounds act in reverse order on Shh signaling. Thus, there arespecific structural features of endocannabinoids that optimizetheir repressive activity toward Shh signaling compared withother pathways. These observations suggest that endocannabi-noids exert their effects directly on one or more components ofthe Shh pathway.To investigate the step at which endocannabinoids act, we

asked whether they inhibited pathway activation by SMO agonist(SAG). SAG binds to SMO, promotes its ciliary translocation,and increases SMO signaling (18, 24, 35). Activation of SMOsignaling peaks at SAG concentrations of approximately 100 nMand decreases thereafter—no pathway activation is observed at10 μM SAG. We assayed pathway activation by 100 nM SAG inthe presence of increasing concentrations of N-acyldopamine16:0, N-acyldopamine 18:0, N-acylethanolamide 18:2, N-acyle-thanolamide 20:4, Cannabinol, and Cannabidiol (Fig. 3B). Allreduced pathway activation by SAG at concentrations similar tothose that were effective against Shh (Fig. 2). To further examinethe interaction between endocannabinoids and SAG, we assayedthe effectiveness of different SAG concentrations in the pres-ence of a constant amount of different endocannabinoids. In-terestingly, endocannabinoids both reduce maximal pathwayactivation by SAG and shift peak pathway activation to lowerSAG concentrations (Fig. 3C). These findings suggest that endo-cannabinoids act at the level of SMO or at a subsequent step toinhibit the Shh pathway.To investigate whether cannabinoids or endocannabinoids might

bind SMO, we asked whether they could compete with BODIPY-cyclopamine for binding to fixed SMO-overexpressing cells. Can-nabinol, Cannabidiol, and N-acylethanolamide 20:4 strongly reducebinding of BODIPY-cyclopamine, and N-acylethanolamide 18:2

Fig. 2. Synthetic endocannabinoids and phytocannabinoids repress Shhsignaling. Effects of different cannabinoids on signaling by nonlipoprotein-associated Shh in ShhLIGHT2 cells: N-acylserines (A), N-acylethanolamides (B),2-alkylglycerol 20:4 (C), N-acyldopamines (D), 2-acylglycerols (E), and phy-tocannabinoids (F). Different endocannabinoid species are depicted in theindicated colors. The pink line in B and blue line in E show activities ofN-acylethanolamide 18:2 and N-acyldopamine 18:0 in the presence of FAAHinhibitor PF-3845. The maximum ratio of reporter activity is normalized to100 and indicated by the orange line. In each case, this maximum ratio (i.e.,fold stimulation by ShhN* over control) ranged between 8- and 10-fold.Renilla luciferase activity (reflecting cell number and viability) was un-affected at all endocannabinoid concentrations shown. Error bars indicateSDs of five independent experiments. n > 5 for each experiment.

Khaliullina et al. PNAS | March 17, 2015 | vol. 112 | no. 11 | 3417

CELL

BIOLO

GY

Dow

nloa

ded

by g

uest

on

June

28,

202

1

http://www.pnas.org/lookup/suppl/doi:10.1073/pnas.1416463112/-/DCSupplemental/pnas.201416463SI.pdf?targetid=nameddest=SF3

has a weaker effect. In contrast, N-acyldopamines 16:0 and 18:0,N-acylserine 16:0, and 2-AG do not (Fig. 3D). This suggests thatphytocannabinoids and N-acylethanolamides either bind SMO inthe transmembrane pocket or influence the structure of this siteallosterically. Thus, these molecules are direct inverse agonistsor negative allosteric modulators of mammalian SMO. 2-AGand N-acyldopamines may repress SMO activity indirectlythrough “entourage” effects—i.e., by competing with the actualendocannabinoid ligand for endocannabinoid transporters andmetabolic enzymes (36).Activation of SMO signaling is regulated both by ciliary ac-

cumulation and also by a subsequent step that takes place inprimary cilia. For example, some SMO inhibitors block ciliaryaccumulation, whereas cyclopamine promotes ciliary accumula-tion but represses the activity of SMO in the cilium (14, 35). Wetherefore asked whether phytocannabinoids or N-acylethanolamide18:2 might block SAG-induced ciliary accumulation of SMO. Evenat concentrations that inhibit SAG-mediated pathway activation,none of these compounds prevented ciliary SMO accumulation

(Fig. S4). Thus, cannabinoid binding represses SMO activity ata step subsequent to SAG-dependent ciliary translocation.To explore whether endocannabinoids might account for the

repressive activity of lipoproteins toward the Hh pathway inDrosophila, we quantified endocannabinoid levels in larval he-molymph by LC-MS/MS. Indeed, hemolymph contains micro-molar amounts of ethanolamides and acylglycerols (Table S2).Furthermore, resupplying endocannabinoids to wing imaginaldiscs reverses the ectopic Hh pathway activation caused byloss of lipoproteins. Knockdown of the Drosophila lipoproteinLipophorin (Lpp) elevates accumulation of Smo on the basolateralmembrane and increases the amount of full-length Ci155 (thesingle Drosophila Gli homolog) (Fig. 4 A′, A′′, B′, and B′′; ref.17). Both effects are reversed by addition of either 2-AG orN-acylethanolamide 16:0 to explanted discs (Fig. 4 A–B** andFig. S5 A–B*). Interestingly, phytocannabinoids mimic the re-pressive activity of endocannabinoids in wing discs, just as theydo in mammalian ShhLIGHT2 cells (Fig. S5E). These data sug-gest that loss of endocannabinoid delivery is responsible for theectopic stabilization of Smo and full length Ci155 that occurs uponlipoprotein knockdown, and that endocannabinoids thereforeregulate these processes in vivo in Drosophila.To ask whether endocannabinoid catabolism was important to

maintain Hh pathway activity in wing discs, we sought to inhibitDrosophila FAAH activity. The Drosophila genome encodes sixproteins similar to mammalian FAAH (Flybase). Preparing he-molymph in the presence of PF-3845 increases the yield of bothethanolamides and 2-acylglycerols (Table S2), suggesting thatthis compound is active against Drosophila enzymes. We there-fore treated explanted wild-type discs with PF-3845 and moni-tored effects on Smo and Ci155 2 hours later. In the anteriorcompartment, PF-3845 treatment depletes Smo from the baso-lateral membrane and reduces Ci155 accumulation (Fig. 4 C–D*).Interestingly, PF-3845, like 2-AG, also lowered Smo levels in theposterior compartment where Ptc is not present (Fig. 4 C–C*and Fig. S5 A–A**). Thus, inhibiting endocannabinoid degra-dation appears to circumvent the requirement for Ptc in Smorepression. The importance of FAAH-like enzymes for normalSmo trafficking and Ci155 processing supports the idea thatendocannabinoids regulate Hh signaling in the wing disk.These experiments demonstrate that lipoprotein-derived

endocannabinoids are endogenous SMO inhibitors that areconserved across phyla. Whereas oxysterols and vitamin D haveclear effects on mammalian SMO signaling (Fig. S6 A–D andrefs. 27 and 37), cannabinoids are the first compounds to showsuch conserved activity. This suggests that cannabinoids inhibitan important basic step in SMO activation that is difficult to alterduring evolution, making these compounds a particularly in-teresting starting point for drug development. Our results in-dicate that phytocannabinoids and N-acylethanolamides bindmammalian SMO directly. They act at a step subsequent to SMOciliary translocation. Whereas cannabinoids do not obviouslyinfluence trafficking of mammalian SMO—at least into the pri-mary cilium—they influence trafficking of Drosophila Smo, pre-venting its accumulation on the basolateral membrane. It ispossible that cannabinoids act by different mechanisms in fliesand mammals. Alternatively, ciliary and basolateral membranelocalization may not be functionally analogous. However, in bothcases, full SMO activation appears to be a two-step process, withonly one step that is influenced by endocannabinoids. Endo-cannabinoids regulate the ability of Drosophila Smo to stabilizethe full-length form of Ci, but lipoprotein knockdown experi-ments show that additional steps are required for target geneactivation (11, 17).Endocannabinoids and phytocannabinoids have many phys-

iological activities that are not completely understood. Someof these may reflect their activity toward the Hh pathway.Cannabinoids and Hh signaling regulate many of the same

Fig. 3. Some endocannabinoids and phytocannabinoids are direct agonists/allosteric modulators of SMO. (A) Signaling by nonlipoprotein-associatedShh in ShhLIGHT2 cells in the presence of specific agonists (green-shadedregions) or antagonists (red-shaded regions) of CB1 and CB2 receptors,GPR55, PPAR-alpha and PPAR-gamma receptors, and TRPV channels. Errorbars are SDs of three independent experiments. n = 3. (B) Hh pathwaystimulation by varying concentrations of different cannabinoids in thepresence of 100 nM SAG. Error bars are SDs of three independent experi-ments. Dotted line indicates the firefly:Renilla luciferase ratio in unstimu-lated cells; green line indicates the ratio observed in cells treated with100 nM SAG alone. n > 3. (C) Hh pathway stimulation by varying concentrationsof SAG in the presence of indicated endocannabinoids. Error bars are SDs ofthree independent experiments. Dotted line indicates the firefly:Renilla lu-ciferase ratio in unstimulated cells, and the orange line shows the ratioobserved in a parallel experiment by using Shh alone. n > 3. (D) Fluorescenceof BODIPY-cyclopamine (B-C) bound to tetracycline-induced SMO in 293Scells, as monitored by FACS. Cells were treated with 5 nM B-C alone (blacktracing) or with 5 nM B-C and 1 μM Cannabinol (violet), 1 μM Cannabidiol(dark red), 10 μM N-acylethanolamide 20:4 (dark green), 100 nM SANT-1(yellow), 100 nM SAG (turquoise), 20 μM N-acyethanolamide 18:2 (pink),20 μM N-acyldopamine 16:0 (orange), 20 μM N-acylglycerol (light green), 5 μMN-acylserine 16:0 (red), or 20 μM N-acyldopamine 18:0 (blue). As negativecontrol, the light gray tracing shows fluorescence of cells not treated withB-C. n = 2.

3418 | www.pnas.org/cgi/doi/10.1073/pnas.1416463112 Khaliullina et al.

Dow

nloa

ded

by g

uest

on

June

28,

202

1

http://www.pnas.org/lookup/suppl/doi:10.1073/pnas.1416463112/-/DCSupplemental/pnas.201416463SI.pdf?targetid=nameddest=SF4http://www.pnas.org/lookup/suppl/doi:10.1073/pnas.1416463112/-/DCSupplemental/pnas.201416463SI.pdf?targetid=nameddest=ST2http://www.pnas.org/lookup/suppl/doi:10.1073/pnas.1416463112/-/DCSupplemental/pnas.201416463SI.pdf?targetid=nameddest=SF5http://www.pnas.org/lookup/suppl/doi:10.1073/pnas.1416463112/-/DCSupplemental/pnas.201416463SI.pdf?targetid=nameddest=SF5http://www.pnas.org/lookup/suppl/doi:10.1073/pnas.1416463112/-/DCSupplemental/pnas.201416463SI.pdf?targetid=nameddest=SF5http://www.pnas.org/lookup/suppl/doi:10.1073/pnas.1416463112/-/DCSupplemental/pnas.201416463SI.pdf?targetid=nameddest=ST2http://www.pnas.org/lookup/suppl/doi:10.1073/pnas.1416463112/-/DCSupplemental/pnas.201416463SI.pdf?targetid=nameddest=SF5http://www.pnas.org/lookup/suppl/doi:10.1073/pnas.1416463112/-/DCSupplemental/pnas.201416463SI.pdf?targetid=nameddest=SF5http://www.pnas.org/lookup/suppl/doi:10.1073/pnas.1416463112/-/DCSupplemental/pnas.201416463SI.pdf?targetid=nameddest=SF6http://www.pnas.org/lookup/suppl/doi:10.1073/pnas.1416463112/-/DCSupplemental/pnas.201416463SI.pdf?targetid=nameddest=SF6www.pnas.org/cgi/doi/10.1073/pnas.1416463112

processes—e.g., angiogenesis (6, 38), hair follicle development(39, 40), nocioception (5, 41), bone formation (42, 43), andenergy metabolism (3, 4, 44). Indeed, cannabinoids block growthof tumors known to depend on Hh signaling (45–50). Interestingly,Cannabis exposure in utero inhibits fetal growth and alters braindevelopment (51); whether impaired Shh signaling contributesto these problems is worth investigating.We have shown that many different species of endocannabi-

noids are present in circulation. Their ability to repress Hh sig-naling suggests a previously unidentified mechanism by whichsystemic metabolism could influence development, tissue homeo-stasis, and cancer. The fact that lipoprotein-derived endocan-nabinoids repress such an important tumor-promoting pathwaymight help explain the link between disturbed lipoprotein me-tabolism and cancer risk (52). Our findings forge a link betweencannabinoids and Hh signaling, opening new research avenuesfor both important classes of signaling molecules.

Materials and MethodsHh Signaling Assays. Drosophila wing disk assay was performed as described(17). ShhLIGHT2 signaling assay, SAG competition assay, ciliary translocationassay, and BODIPY-cyclopamine binding assay were described (23, 24).

Column Chromatography. Extracts containing saponification-resistant lipidsfrom VLDL were loaded onto a C18 column (25 mm × 4.5 mm, 5 μMparticles).Five hundred microliters of extract was loaded in 60% aqueous methanol andeluted with a step gradient of 80% methanol in H2O (60 min) and a lineargradient of 80–100%methanol in water (40 min) at the flow rate of 1 mL/min.The eluate was collected into 40 fractions that were dried down and subjectedto the mass spectrometric analysis and ShhLIGHT2 signaling assay.

Screening of Endocannabinoid Fractions by Shotgun Lipidomics. Fifty micro-liters (1/20) of each HPLC elution fraction was mixed with sixty-five microlitersof 13 mM ammonium acetate in isopropanol. Twenty microliters of samplewere loaded onto 96-well plate and centrifuged for 5 min at 4,000 rpm.Mass spectrometric analyses were performed on a Q Exactive instrumentwith robotic nanoflow ion source TriVersa NanoMate controlled by the

Chipsoft 6.4 software (nanoflow chips: 4-μm spraying nozzle diameter;ionization voltage: ±1.25kV; gas back pressure: 0.95 psi). Mass spectrawere acquired in positive and negative ion mode with the target massresolution of Rm/z 200 = 140,000 within m/z range of 100–1,000. Precursorswithin m/z range of 200–500 were fragmented in a data-dependent acquisi-tion mode with the target resolution of Rm/z 200 = 17,000.

Extraction of Endocannabinoids from Human Blood Serum and DrosophilaLarval Hemolymph. At 4 °C, 500 μL of human serum or 500 μL of dilutedlarval hemolymph (equivalent to 13.5 larvae) were supplemented with10 μM PF-3845. Then, 600 μL of ethyl acetate/n-hexane 9:1 and 17.5 μL of theinternal standard mixture (14 pmol/mL d4-N-acylethanolamide 16:0,7.39 pmol/mL d8-N-acylethanolamide 20:4, 100 pmol/mL N-acylserine 20:4,100 pmol/mL 2-alkylglycerol 20:4, 90.3 pmol/mL d8-2-acylglycerol 20:4, and100 pmol/mL d5-1-acylglycerol 20:4) were added. After 15 min of centri-fugation at 15,000 × g and 10 min of incubation on dry ice, the organicphases were dried and reconstituted in 35 μL of water/acetonitrile/iso-propanol/formic acid (5:4:0.5:0.1), centrifuged for 5 min at 10,000 × g andtransferred for LC-MS/MS analysis.

LC-MS/MS Quantification of Endocannabinioids. The analysis was performed onAgilent LC1100 system with a C18 column (5 μm, 0.5 × 150 mm); flow rate of20 μL/min; injection volume of 3 μL interfaced on-line to a triple quadrupolemass spectrometer TSQ Vantage. The elution gradient comprised solvent A:0.1% formic acid in water and solvent B: acetonitrile/isopropanol/formic acid(9:1:0.1) run with the profile: 0 min, 50% B; 0–2 min, 50–66.4% B; 2–8 min,66.4–73% B; 8–10 min, 73–95% B; 10–14 min, 95% B; 14–15 min, 95–50% B;15–21 min, 50% B. Endocannabinoids were detected as [M+H]+ ions by MRMtransitions in the positive mode; S-lens voltages and collision energies wereoptimized individually for each standard compound in direct infusion mode.The transfer capillary temperature was 275 °C and the ions isolation widthwas 0.7 amu.

ACKNOWLEDGMENTS. We thank P. Beachy for providing ShhLIGHT2,HEK293S-TetR, and NIH3T3/Smo-mEos2 cells; B. Borgonovo for assistancewith column chromatography; and T. Mitchison, P. Born, N. Dye, A. Sagner,W. Palm, K. Simons, J. Rodenfels, M. Zerial, S. Bray, and W. Harrisfor critical comments on the manuscript. Support was provided byMax-Planck-Gesellschaft, Deutsche Forschungsgemeinschaft EA4/2-4, andSonderforschungsbereich TRR83 Projects A14 and A17.

Fig. 4. FAAH-dependent endocannabinoid homeostasis regulates Smo and Ci155 levels in Drosophila wing imaginal discs. (A–B**) depicts wing discs fromeither wild-type animals (A’, A*, B’, and B*) or animals depleted of Lpp (A’’, A**, B””, and B**) that were mock-treated (A’, A”, B’, and B’’) or treated with50 μM N-acylethanolamide 16:0 (A’’, A**, B’’, and B**), stained for Smo (A’–A**) and Ci155 (B’–B**). A and B show the average values of Smo (A) and Ci155(B) staining intensities calculated from 10 wing imaginal discs for each condition. (C–D**) Wild-type wing discs either mock treated (C’ and D’) or treated with10 μM FAAH inhibitor PF-3845 (C* and D*), stained for Smo (C’ and C*) and Ci155 (D’ and D*). Corresponding average staining intensities calculated from10 wing imaginal discs are shown in C and D. To estimate the significance of the difference between staining intensities in mock-treated and N-acylethanolamide16:0- or PF-3845-treated discs, we used the mean intensity and SD values at each point along the curves in the anterior compartment. P values were calculated byapplying the Student’s t test. The double-headed arrows indicate the curves that were compared. In A, **P = 1.3 × 10−37; in B, *P = 3.4 × 10−2. In C, **P = 2.56 × 10−14;in D, **P = 4.82 × 10−26. Anterior is to the right. AP, anteroposterior. (Scale bars: 10 μm.) n > 10 discs for each quantification.

Khaliullina et al. PNAS | March 17, 2015 | vol. 112 | no. 11 | 3419

CELL

BIOLO

GY

Dow

nloa

ded

by g

uest

on

June

28,

202

1

1. Ingham PW, Nakano Y, Seger C (2011) Mechanisms and functions of Hedgehog sig-nalling across the metazoa. Nat Rev Genet 12(6):393–406.

2. Amakye D, Jagani Z, Dorsch M (2013) Unraveling the therapeutic potential of theHedgehog pathway in cancer. Nat Med 19(11):1410–1422.

3. Pospisilik JA, et al. (2010) Drosophila genome-wide obesity screen reveals hedgehogas a determinant of brown versus white adipose cell fate. Cell 140(1):148–160.

4. Teperino R, et al. (2012) Hedgehog partial agonism drives Warburg-like metabolismin muscle and brown fat. Cell 151(2):414–426.

5. Babcock DT, et al. (2011) Hedgehog signaling regulates nociceptive sensitization. CurrBiol 21(18):1525–1533.

6. Byrd N, Grabel L (2004) Hedgehog signaling in murine vasculogenesis and angio-genesis. Trends Cardiovasc Med 14(8):308–313.

7. Crompton T, Outram SV, Hager-Theodorides AL (2007) Sonic hedgehog signalling inT-cell development and activation. Nat Rev Immunol 7(9):726–735.

8. de la Roche M, et al. (2013) Hedgehog signaling controls T cell killing at the immu-nological synapse. Science 342(6163):1247–1250.

9. Pepinsky RB, et al. (1998) Identification of a palmitic acid-modified form of humanSonic hedgehog. J Biol Chem 273(22):14037–14045.

10. Porter JA, Young KE, Beachy PA (1996) Cholesterol modification of hedgehog sig-naling proteins in animal development. Science 274(5285):255–259.

11. Palm W, et al. (2013) Secretion and signaling activities of lipoprotein-associatedhedgehog and non-sterol-modified hedgehog in flies and mammals. PLoS Biol 11(3):e1001505.

12. Panáková D, Sprong H, Marois E, Thiele C, Eaton S (2005) Lipoprotein particles arerequired for Hedgehog and Wingless signalling. Nature 435(7038):58–65.

13. Rohatgi R, Scott MP (2007) Patching the gaps in Hedgehog signalling. Nat Cell Biol9(9):1005–1009.

14. Rohatgi R, Milenkovic L, Scott MP (2007) Patched1 regulates hedgehog signaling atthe primary cilium. Science 317(5836):372–376.

15. Corbit KC, et al. (2005) Vertebrate Smoothened functions at the primary cilium. Na-ture 437(7061):1018–1021.

16. Denef N, Neubüser D, Perez L, Cohen SM (2000) Hedgehog induces opposite changesin turnover and subcellular localization of patched and smoothened. Cell 102(4):521–531.

17. Khaliullina H, et al. (2009) Patched regulates Smoothened trafficking using lipopro-tein-derived lipids. Development 136(24):4111–4121.

18. Rohatgi R, Milenkovic L, Corcoran RB, Scott MP (2009) Hedgehog signal transductionby Smoothened: Pharmacologic evidence for a 2-step activation process. Proc NatlAcad Sci USA 106(9):3196–3201.

19. Briscoe J, Thérond PP (2013) The mechanisms of Hedgehog signalling and its roles indevelopment and disease. Nat Rev Mol Cell Biol 14(7):416–429.

20. Teperino R, Aberger F, Esterbauer H, Riobo N, Pospisilik JA (2014) Canonical and non-canonical Hedgehog signalling and the control of metabolism. Semin Cell Dev Biol 33:81–92.

21. Jenkins D (2009) Hedgehog signalling: Emerging evidence for non-canonical path-ways. Cell Signal 21(7):1023–1034.

22. Wang C, et al. (2013) Structure of the human smoothened receptor bound to anantitumour agent. Nature 497(7449):338–343.

23. Chen JK, Taipale J, Cooper MK, Beachy PA (2002) Inhibition of Hedgehog signaling bydirect binding of cyclopamine to Smoothened. Genes Dev 16(21):2743–2748.

24. Chen JK, Taipale J, Young KE, Maiti T, Beachy PA (2002) Small molecule modulation ofSmoothened activity. Proc Natl Acad Sci USA 99(22):14071–14076.

25. Frank-Kamenetsky M, et al. (2002) Small-molecule modulators of Hedgehog signal-ing: Identification and characterization of Smoothened agonists and antagonists.J Biol 1(2):10.

26. Rominger CM, et al. (2009) Evidence for allosteric interactions of antagonist bindingto the smoothened receptor. J Pharmacol Exp Ther 329(3):995–1005.

27. Bijlsma MF, et al. (2006) Repression of smoothened by patched-dependent (pro-)vi-tamin D3 secretion. PLoS Biol 4(8):e232.

28. Nedelcu D, Liu J, Xu Y, Jao C, Salic A (2013) Oxysterol binding to the extracellulardomain of Smoothened in Hedgehog signaling. Nat Chem Biol 9(9):557–564.

29. Myers BR, et al. (2013) Hedgehog pathway modulation by multiple lipid binding siteson the smoothened effector of signal response. Dev Cell 26(4):346–357.

30. Hartl TA, Scott MP (2014) Wing tips: The wing disc as a platform for studyingHedgehog signaling. Methods 68(1):199–206.

31. Pertwee RG, Ross RA (2002) Cannabinoid receptors and their ligands. ProstaglandinsLeukot Essent Fatty Acids 66(2-3):101–121.

32. Pertwee RG, et al. (2010) International Union of Basic and Clinical Pharmacology.LXXIX. Cannabinoid receptors and their ligands: Beyond CB₁ and CB₂. Pharmacol Rev62(4):588–631.

33. Fanelli F, et al. (2012) Estimation of reference intervals of five endocannabinoids andendocannabinoid related compounds in human plasma by two dimensional-LC/MS/MS.J Lipid Res 53(3):481–493.

34. De Petrocellis L, Cascio MG, Di Marzo V (2004) The endocannabinoid system: Ageneral view and latest additions. Br J Pharmacol 141(5):765–774.

35. Wang Y, Zhou Z, Walsh CT, McMahon AP (2009) Selective translocation of in-tracellular Smoothened to the primary cilium in response to Hedgehog pathwaymodulation. Proc Natl Acad Sci USA 106(8):2623–2628.

36. Ben-Shabat S, et al. (1998) An entourage effect: Inactive endogenous fatty acidglycerol esters enhance 2-arachidonoyl-glycerol cannabinoid activity. Eur J Pharmacol353(1):23–31.

37. Nachtergaele S, et al. (2012) Oxysterols are allosteric activators of the oncoproteinSmoothened. Nat Chem Biol 8(2):211–220.

38. Zhang X, Maor Y, Wang JF, Kunos G, Groopman JE (2010) Endocannabinoid-likeN-arachidonoyl serine is a novel pro-angiogenic mediator. Br J Pharmacol 160(7):1583–1594.

39. Telek A, et al. (2007) Inhibition of human hair follicle growth by endo- and exo-cannabinoids. FASEB J 21(13):3534–3541.

40. St-Jacques B, et al. (1998) Sonic hedgehog signaling is essential for hair development.Curr Biol 8(19):1058–1068.

41. Cravatt BF, Lichtman AH (2004) The endogenous cannabinoid system and its role innociceptive behavior. J Neurobiol 61(1):149–160.

42. Bab I, Smoum R, Bradshaw H, Mechoulam R (2011) Skeletal lipidomics: Regulation ofbone metabolism by fatty acid amide family. Br J Pharmacol 163(7):1441–1446.

43. Lai LP, Mitchell J (2005) Indian hedgehog: Its roles and regulation in endochondralbone development. J Cell Biochem 96(6):1163–1173.

44. Guzmán M, et al. (2004) Oleoylethanolamide stimulates lipolysis by activating thenuclear receptor peroxisome proliferator-activated receptor alpha (PPAR-alpha).J Biol Chem 279(27):27849–27854.

45. Hamtiaux L, et al. (2012) The association of N-palmitoylethanolamine with the FAAHinhibitor URB597 impairs melanoma growth through a supra-additive action. BMCCancer 12:92.

46. Yun JI, Kim HR, Park H, Kim SK, Lee J (2012) Small molecule inhibitors of thehedgehog signaling pathway for the treatment of cancer. Arch Pharm Res 35(8):1317–1333.

47. Miyato H, et al. (2009) Pharmacological synergism between cannabinoids and pacli-taxel in gastric cancer cell lines. J Surg Res 155(1):40–47.

48. Velasco G, et al. (2007) Cannabinoids and gliomas. Mol Neurobiol 36(1):60–67.49. Bifulco M, Malfitano AM, Pisanti S, Laezza C (2008) Endocannabinoids in endocrine

and related tumours. Endocr Relat Cancer 15(2):391–408.50. Santini R, et al. (2012) Hedgehog-GLI signaling drives self-renewal and tumorigenicity

of human melanoma-initiating cells. Stem Cells 30(9):1808–1818.51. Morris CV, DiNieri JA, Szutorisz H, Hurd YL (2011) Molecular mechanisms of maternal

cannabis and cigarette use on human neurodevelopment. Eur J Neurosci 34(10):1574–1583.

52. Braun S, Bitton-Worms K, LeRoith D (2011) The link between the metabolic syndromeand cancer. Int J Biol Sci 7(7):1003–1015.

3420 | www.pnas.org/cgi/doi/10.1073/pnas.1416463112 Khaliullina et al.

Dow

nloa

ded

by g

uest

on

June

28,

202

1

www.pnas.org/cgi/doi/10.1073/pnas.1416463112