Structure of a Human Intramembrane Ceramidase Explains Enzymatic

bioRxiv preprint doi: https://doi.org/10.1101/348219; this version posted June 15, 2018. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY-NC-ND 4.0 International license.

1 2 Structure of a human intramembrane ceramidase explains enzymatic

3 dysfunction found in leukodystrophy

1† 1† 1 1 5 Ieva Vasiliauskaité-Brooks , Robert D. Healey , Pascal Rochaix , Rémy Sounier , Claire

1 1 1 2 3,4 6 Grison , Thierry Waltrich-Augusto , Mathieu Fortier , François Hoh , Essa M. Saied ,

3 5 *1 *1 7 Christoph Arenz , Shibom Basu , Cédric Leyrat and Sébastien Granier .

8 Affiliations:

1 9 IGF, University of Montpellier, CNRS, INSERM, Montpellier, France.

2 10 CBS, University of Montpellier, CNRS, INSERM, Montpellier, France.

3 11 Institute for chemistry, Humboldt-Universität zu Berlin, Brook-Taylor-Str. 2, 12489 Berlin,

12 Germany.

4 13 Chemistry Department, Faculty of Science, Suez Canal University, 41522 Ismailia, Egypt

5 14 Macromolecular Crystallography, Swiss Light Source, Paul Scherrer Institut, 5232, Villigen

15 PSI, Switzerland

† 16 Equal contribution

* 17 Correspondence to: [email protected] or [email protected]

1 bioRxiv preprint doi: https://doi.org/10.1101/348219; this version posted June 15, 2018. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY-NC-ND 4.0 International license.

25 Abstract

26 Alkaline ceramidases (ACERs) are a class of poorly understood transmembrane enzymes

27 controlling the homeostasis of ceramides. They are implicated in human pathophysiology,

28 including progressive leukodystrophy, colon cancer as well as acute myeloid leukemia. We

29 report here the crystal structure of the human ACER type 3 (ACER3). Together with

30 computational studies, the structure reveals that ACER3 is an intramembrane enzyme with a

2+ 31 seven transmembrane domain architecture and a catalytic Zn binding site in its core, similar

2+ 32 to adiponectin receptors. Interestingly, we uncover a Ca binding site physically and

2+ 33 functionally connected to the Zn providing a structural explanation for the known regulatory

2+ 34 role of Ca on ACER3 enzymatic activity and for the loss of function in E33G-ACER3

35 mutant found in leukodystrophic patients.

2 bioRxiv preprint doi: https://doi.org/10.1101/348219; this version posted June 15, 2018. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY-NC-ND 4.0 International license.

55 Main Text

56 The main bioactive sphingolipids ceramide, sphingosine and sphingosine 1-phosphate (S1P)

57 play key roles in human (patho)physiology including cancer cell biology, immune,

1 58 inflammatory and metabolic functions (reviewed in ). As a result, enzymes regulating

59 sphingolipid levels constitute key therapeutic targets, particularly for the treatment of cancer

2 60 . Among these enzymes, ceramidases (CDases) are attractive targets for clinical intervention

3 61 as they directly regulate the balance between these bioactive lipids by converting ceramides

4 5 62 into free fatty acids and sphingosine which is further processed into S1P by kinases .

63 The five ceramidases cloned to date are classified into acid, neutral and alkaline groups

3 64 according to the pH optima of the hydrolysis reaction (reviewed in ). However, the three

65 groups do not display any sequence homology; the acid ceramidase (ASAH1), ubiquitously

66 expressed, is mainly present in lysosomes, its inactivation by mutation causing Farber disease

6 67 . The recent crystal structures of ASAH1 revealed a globular fold associating α-helices and

7 68 anti-parallel β-sheets . This study also showed that the ASAH1 enzymatic activity

69 necessitates an autoproteolytic-based conformational change exposing the putative substrate

7 70 binding cavity and the cystein-based catalytic center at its base . The neutral ceramidase

71 (NCDase) is also ubiquitously expressed, structurally containing one transmembrane domain

8 72 (TM) and a large soluble domain unrelated to ASAH1. The recent crystal structure of

2+ 73 NCDase soluble domain revealed a Zn -dependent catalytic site deeply buried in a

9 74 hydrophobic binding pocket which can accommodate the ceramide .

75 Alkaline ceramidases (ACER) are much less well understood, in part because of their

76 hydrophobic nature that, until now, has rendered the biochemical and structural analyses

10 11 12 77 difficult. Three different genes have been cloned; ACER1 , ACER2 and ACER3 and

78 sequence analyses suggest that they are integral membrane proteins. ACER1 and ACER2

79 expression is rather tissue specific (skin and placenta, respectively), while ACER3 is

3 bioRxiv preprint doi: https://doi.org/10.1101/348219; this version posted June 15, 2018. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY-NC-ND 4.0 International license.

10-12 80 expressed in most tissues . Very little is known at the molecular level: ACERs are

81 localized intracellularly in the membrane of the endoplasmic reticulum-Golgi apparatus

82 network and their activity, mainly directed against ceramides with long unsaturated acyl

2+ 10,12-14 83 chains (C18:1, C20:1 and C24:1), was shown to be Ca -dependent .

84 The critical role of ACERs in human physiology and, in particular ACER3, was recently

85 revealed by clinical data demonstrating that ACER3 deficiency leads to progressive

15 86 leukodystrophy in early childhood . This study demonstrated that patients were homozygous

87 for a p.E33G ACER3 mutant and that this mutation impaired the ACER3 ceramidases activity

88 in patients' cells. When compared to healthy individuals, this loss of function resulted in

89 higher level of several ceramide species in the blood, in particular for the ACER3 preferred

90 substrates, C18:1 and C20:1 ceramides. It was proposed that these aberrant levels of

91 ceramides in the brain could result in an incorrect central myelination leading to the clinical

92 phenotype associated with the ACER3 mutant. However, in mice, while ACER3 knock-out

93 results in an aberrant accumulation of various ceramides, it does not affect myelination.

94 Instead, this deficiency induces the premature degeneration of Purkinje cells and cerebellar

16 95 ataxia . In the periphery, in mice, the modulation of C18:1 ceramide levels by ACER3

96 regulates the immune response through the upregulation of cytokines while its deficiency

17 97 increases colon inflammation and its associated tumorigenesis . Moreover, in vitro results

98 obtained in human cells revealed that ACER3 contributes to acute myeloid leukemia (AML)

18 99 pathogenesis . Indeed, it was found that ACER3 expression negatively correlates with the

100 survival of AML patients, and that ACER3 is essential for the growth of AML cells as the sh-

101 RNA inhibition of its expression resulted in an increase of apoptosis and in an important

18 102 decrease in cell growth .

103 The molecular control of ACER enzymatic activity (agonists and antagonists) thus appears as

104 a possible clinical intervention for the treatment of leukodystrophy, colon cancer and acute

4 bioRxiv preprint doi: https://doi.org/10.1101/348219; this version posted June 15, 2018. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY-NC-ND 4.0 International license.

105 myeloid leukemia among other pathologies involving ceramide dyshomeostasis. This

106 endeavor is unfortunately made impossible because of the lack of knowledge on the

107 molecular basis of ACER function.

108 In an effort to determine the molecular mechanisms of the ACER enzymatic function, we

109 undertook a structural analysis of ACER3. We solved a 2.7 Å crystal structure using in meso

110 crystallography that, together with computational studies, revealed at the atomic level how

111 this integral membrane enzyme accommodates and hydrolyzes its ceramide substrate in a

2+ 2+ 112 Zn - and Ca -dependent manner and how a single point (E33G) mutant results in the

113 calcium binding site destabilization providing a molecular explanation for the ACER3 mutant

114 dysfunction leading to leukodystrophy in human.

115

116 Results

117 ACER3 is a seven transmembrane protein

118 Recent breakthroughs in the crystallization of challenging membrane proteins such as G

119 protein-coupled receptors include the use of protein engineering with a soluble module,

120 an approach pioneered in 2007 to solve the structure of the beta2-adrenergic receptor19.

121 Here, we modified ACER3 (residues 1 to 244) with the BRIL soluble module

122 (thermostabilized apocytochrome b562RIL from Escherichia coli20) fused to its C-

123 terminus (Extended Data Fig. 1 and Methods) and showed that this fusion is

124 biologically active; it hydrolyzes C18:1 ceramide substrate in a preferred manner over

125 C18:0 ceramide in detergent micelles (Extended data Fig. 1). The ACER3-BRIL fusion

126 was crystallized in a cholesterol-doped monoolein lipidic mesophase and the structure

127 determined at 2.7 Å resolution (Extended Data table 1).

128 When cloned in 2001, ACER3 was predicted to be an integral membrane protein with

129 five TMs12. In fact, the crystal structure revealed that it possesses seven alpha-helices

5 bioRxiv preprint doi: https://doi.org/10.1101/348219; this version posted June 15, 2018. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY-NC-ND 4.0 International license.

130 forming a 7TM architecture with opposite N- and C-terminus domains (Fig. 1).

131 Strikingly, the fold of ACER3 is similar to adiponectin receptors (ADIPORs) with the 7TM

132 harboring a Zn2+ binding site21 (Fig. 1a, Extended data Fig. 2) despite a very low

133 sequence identity (14% with ADIPOR1 and 10% for ADIPOR2). This similar 7TM fold

134 suggests that ACER3 and ADIPORs possess the same topology; an intracellular N-

135 terminus exposed to the cytoplasm and the C-terminus facing the lumen (Fig. 1,

136 Extended data Fig. 2). We confirmed this topology in living cells using TR-FRET

137 experiments (Extended data Fig. 2). On each side, the TMs are connected by three

138 cytoplasmic loops (ICL 1 to 3) and three extracellular loops (ECL 1 to 3) (Fig. 1a). The

139 ICL3, connecting the TM6 and 7, is the most notable loop (residues 196-216) as it forms

140 a motif composed of two short alpha-helices (one and two turns) lying parallel to the

141 membrane plane connected by a positively-charged loop (a RKKVPP sequence) that is

142 suitably positioned to interact with the polar head of negatively-charged lipids (Fig. 1b).

143 The N-terminus (1-34), devoid of secondary structure, is packed against the bottom of

144 the TM bundle (1, 2, 3, 6 and 7) and forms a spiral-like motif in which resides a Ca2+ ion.

145 A disulfide bond formed between C21 and C196 connects the N-terminus to TM6 and

146 may play an important role in stabilizing this peculiar N-terminal domain fold (Fig. 1c).

147

148 Intramembrane binding sites

149 The ACER3 7TM domain contains a large hook-shaped intramembrane pocket directly

150 accessible to the lipid leaflet through two openings between the TM5 and the TM6 (Fig.

151 2). This cavity is formed by residues belonging to all TMs except TM4 (Fig. 2a). At the

152 top, the V164TM5 side chain is capping the pocket, while V157 TM5, I161TM5, Y176TM6 and

153 L179TM6 side chains shape the side of the cavity facing the lipids (Fig. 2b,c). Opposed to

154 this hydrophobic patch and towards the inner core of the TM, there is a rather polar

6 bioRxiv preprint doi: https://doi.org/10.1101/348219; this version posted June 15, 2018. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY-NC-ND 4.0 International license.

155 environment formed by the residues S160TM5, S178TM6 and H231TM7 in which we found

156 some electron density that we modeled as a water molecule (Fig. 2b). Such a

157 hydrophilic patch might play a role in ceramide binding. Directly below this area

158 appeared a large electron density in the calculated 2Fo-Fc map as well as in the

159 (unbiased) polder OMIT map22 (see methods) (Extended data Fig. 3) that we

160 tentatively modeled as a monoolein. Indeed, monoolein by weight forms 54% of the

161 lipidic cubic phase (concentration in the mM range) and displays an acyl chain moiety

162 chemically identical to that of C18:1 ceramide (Extended data Fig. 3), the preferred

163 substrate of ACER314.

164 At the bottom of the intramembrane pocket, we unambiguously identified a Zn2+ ion

165 (Extended data Fig. 3 and methods). This Zn2+ is coordinated by three His residues

166 (H81TM2, H217TM7 and H221TM7) and a water molecule forming hydrogen bonds with

167 S77TM2 and D92TM3 (Fig. 2d). The residues forming this Zn2+ binding site are strictly

168 conserved among the ACER family, and across species from yeast to human, highlighting

169 the importance of this site in the biological function of these proteins (Extended data

170 Fig. 4). Given the well-characterized ceramidase activity of ACER3, these structural data

171 definitively establish ACERs as Zn2+-dependent intramembrane enzymes.

172

173 Substrate docking and hydrolysis mechanism

174 We then assessed the mechanism of substrate binding and hydrolysis. To this end, we

175 used computational docking and molecular dynamics (MD) simulations to calculate the

176 most favorable binding mode of a C18:1 ceramide into the ACER3 lipid binding pocket

177 (Fig. 3). The best scoring docking pose positions the ceramide in the hook-shaped cavity

178 within the 7TM (Fig. 3a). The C18:1 acyl chain is stabilized through extensive contacts

179 with hydrophobic residues lining the interior of the cavity including M43TM1, V73TM2,

7 bioRxiv preprint doi: https://doi.org/10.1101/348219; this version posted June 15, 2018. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY-NC-ND 4.0 International license.

180 M96TM3, F103TM3, V153TM5, L156TM5, V157TM5 and F182TM6 (Fig. 3b). Surprisingly, the

181 site of unsaturation with the sp2 carbons is surrounded by a polar environment

182 constituted by the sulfhydryl group of C100TM3 and three hydroxyl groups of S99TM3,

183 Y149TM5 and S228TM7 (Fig. 3c). It is not clear from the structure alone whether this

184 feature has a role in the C18:1 vs C18:0 ceramide substrate preference. The sphingosine

185 moiety remains partially accessible to the membrane leaflet and interacts with a set of

186 hydrophobic residues from TM5 and TM6 (Fig. 3b). The ceramide carbonyl group and

187 its primary alcohol directly interact with the Zn2+ ion, resulting in an octahedral

188 coordination sphere (Fig. 3d). The ceramide is further stabilized by polar contacts

189 between its carbonyl and the amine group of W220TM7 side chain, as well as its primary

190 alcohol and the carboxylate of D92TM3. Interestingly, the crystallographic water molecule

191 that bridges S77TM2, D92TM3 and the Zn2+ ion appears to be suitably placed for a

192 nucleophilic attack on the ceramide carbonyl (Fig. 3d), suggesting a general acid-base

193 catalytic mechanism in which D92TM3 acts as a proton acceptor/donor (similar to e.g.,

194 zinc-dependent proteases)23 (Extended Data Fig. 5). From a substrate specificity point

195 of view, it is clear from the structure and the docking results that substrates presenting

196 large/bulky modifications of the primary alcohol such as sphingomyelin,

197 glucosylceramide or ceramide- 1-phosphate cannot be accommodated in the pocket and

198 are unlikely to serve as ACER3 substrate (Extended Data Fig. 5). Such a steric

199 hindrance serving as the mechanism of substrate selectivity was also observed in

200 neutral ceramidases9.

201 Additional evidence supporting the biological relevance of the described ceramide

202 binding pose was obtained from multiple simulations of the ACER3-C18:1 ceramide

203 complex. Indeed, the initial binding pose was very stable with the all-atom RMSD of the

8 bioRxiv preprint doi: https://doi.org/10.1101/348219; this version posted June 15, 2018. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY-NC-ND 4.0 International license.

204 bound C18:1 ceramide remaining below 3Å in every trajectory of five independent

205 simulations (Fig. 3a and Extended Data Fig. 5).

206

207 Comparison to ADIPORs structures

208 The overall architecture of ACER3 is similar to the ADIPORs. In particular, the position of

209 the TM1, 2, 3, 6 and 7 forms a similar pattern when viewed from the cytoplasm or from

210 the lumen (Fig. 4a). In addition to the similar 7TM architecture, the Zn2+ binding sites

211 with the canonical (H)3SD-water module are almost superimposable (Fig. 4b). Of

212 interest, the sequence (H)3SD is unifying a superfamily of putative hydrolases recently

213 identified based on statistically significant sequence similarities, termed CREST (alkaline

214 ceramidase, PAQR receptor, Per1, SID-1 and TMEM8)24.

215 Significant/major differences are observed however between ACER3 and ADIPORs.

216 Unlike ADIPORs, the ACER3 intramembrane pocket is not connected to the upper leaflet

217 of the membrane nor to the cytoplasmic domain, due to a distinct N-terminus fold

218 (Extended Data Fig. 6). Conformational changes within domains accessible to the

219 hydrophilic cytoplasm must occur for water molecules to access the active site. The

220 difference between the intramembrane pockets also resulted in distinct calculated

221 ceramide binding modes differing essentially at the level of the sphingosine moiety

222 position (Extended Data Fig. 6). These differences may have some impact on the

223 intrinsic enzymatic properties of ACER3 vs ADIPORs, i.e. KM parameters and substrate

224 preference.

225 A key difference between ACER3, ADIPOR1 and ADIPOR2 is seen in the TM4-5

226 architecture. First, the ACER3 TM4 is much shorter than that of the ADIPORs, resulting

227 from a longer ECL2 connecting TM3 and TM4 (Fig. 4c). Second, as compared to the open

228 ADIPOR1 and closed ADIPOR2 TM4-TM5 structures, ACER3 presents an intermediate

9 bioRxiv preprint doi: https://doi.org/10.1101/348219; this version posted June 15, 2018. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY-NC-ND 4.0 International license.

229 TM4-TM5 conformation (Fig. 4c). These architectural differences are most likely

230 originating from the biochemical preparations as well as from crystallization conditions:

231 where a closed ADIPOR2 structure favored the binding of an oleic acid; an intermediate

232 ACER3 bound a monoolein; and an open ADIPOR1 structure was ligand-free. The three

233 distinct conformations might however represent distinct steps of the common catalytic

234 process.

235 On the other hand, the described structure of the soluble NCDase has nothing in

236 common with ACER3 structure apart from the three histidines coordinating the Zn2+

237 (Extended Data Fig. 6). In particular, catalytic sites belong to two clearly distinct

238 scaffolds: intramembranous for ACER3 and membrane-associated for NCDase with a

239 water-soluble domain predominantly constituted of beta-sheets (Extended Data Fig. 6)

240 forming the so-called beta-triangle hydrolase scaffold9. These clearly distinct structures

241 are in agreement with the different cellular localization as well as with the functional

242 specialization of these enzymes3.

243

244 A Ca2+ binding site within the N-terminus

245 Perhaps the most exciting/interesting feature of the ACER3 structure is the presence of

246 a Ca2+ binding site within the N-terminus domain (Fig. 5a). Ca2+ was shown to modify

247 the enzymatic activity of ACER3 but the molecular basis for this effect was unknown.

248 The Ca2+ ion is coordinated by six oxygens from the D19 carboxylate (bidentate), the

249 W20 backbone carbonyl, the E22 backbone carbonyl, the N24 side chain carbonyl and

250 the E33 carboxylate (monodentate) (Fig. 5b). The resulting coordination geometry

251 resembles an incomplete pentagonal bipyramid. However, during MD simulations, the

252 coordination of D19 carboxylate switches from bidentate to monodentate, and the

253 coordination sphere is completed by a water molecule resulting in an octahedral

10 bioRxiv preprint doi: https://doi.org/10.1101/348219; this version posted June 15, 2018. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY-NC-ND 4.0 International license.

254 calcium site (Extended Data Fig. 7). The loop forming the Ca2+ binding site is connected

255 to the 7TM core through a disulfide bond formed by C21 and C196TM6 positioned at the

256 bottom of TM6 (beginning of the ICL3). The sequences D19WCE(X)N24 and 32AEF34

257 constituting the Ca2+ binding domain as well as the C196TM6 forming the disulfide bond

258 are strictly conserved among the ACER family (Fig. 5c) in agreement with the Ca2+-

259 dependent ceramidase activity described for ACER1, ACER2 and ACER310,12-14.

260 Moreover, this domain is also strictly conserved across species from yeast to human

261 (Fig. 5c, Extended Data Fig. 4). This strict conservation during evolution further

262 highlights the paramount regulatory role of Ca2+ in the enzymatic function.

263 Interestingly, based on the crystal structure and MD simulations results, two residues

264 appear to play critical roles in physically linking the Ca2+ and the Zn2+ sites, W20 and

265 E33. Indeed, in the structure, the two residues interact simultaneously with both metal

266 sites (Fig. 5d). First, E33 carboxylate interacts with the Ca2+ ion and is in close proximity

267 to H81TM2 side chain, while H81TM2 coordinates the Zn2+ (Fig. 5d). Interestingly, E33

268 carboxylate rearranges during MD simulations to form a hydrogen bond with H81TM2

269 side chain (Fig. 5d,e). Second, W20 coordinates Ca2+ through its backbone carbonyl, and

270 its indole ring forms a His-aromatic complex with H81TM2 and H217TM7 side chains (Fig.

271 5d). Such His-aromatic interactions are found in other enzymes and constitute key

272 components of the enzymatic catalytic mechanism by participating in the transition-

273 state stabilization or through critical constraints on the conformation of the catalytic site

274 (described in 25). In addition, W20 forms a network of hydrophobic interactions with

275 L18, F80TM2, W189TM6 forming the bottom of the hook-shaped intramembrane pocket

276 towards the cytoplasm (Fig. 5f). We hypothesize that the functional effect of Ca2+ on the

277 catalytic activity of ACER3 originates from this direct link, with changes to the Ca2+ site

278 propagating to the Zn2+ catalytic site.

11 bioRxiv preprint doi: https://doi.org/10.1101/348219; this version posted June 15, 2018. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY-NC-ND 4.0 International license.

279 Remarkably, this discovery provides molecular insights into the E33G ACER3 mutation

280 carried by patients suffering leukodystrophy, which results in the loss of ACER3

281 ceramidase activity15. We used MD simulations to investigate the effect of this mutation

282 on the stability of the ACER3 structure relative to the CER18:1-bound wild type model

283 (Fig. 6). The effect observed on the flexibility of the Zn2+ itself is rather mild due to the

284 integrity of its coordinating residues and the presence of the stabilizing C18:1 ceramide

285 molecule (Extended data Fig. 8). Of note, unlike the wild-type ACER3, the purified E33G

286 ACER3 mutant was highly unstable during purification and yielded mostly aggregated

287 protein eluting in the void volume during size exclusion chromatography, suggesting

288 that the destabilization of the Ca2+ site might influence the overall protein stability. In

289 agreement with this experimental observation, MD simulations revealed rapid Ca2+

290 unbinding followed by an overall destabilization of the N-terminal/cytoplasmic region

291 of the protein in the mutated E33G model (Fig. 6, Extended data Fig. 8). At the level of

292 the Ca2+ and Zn2+ sites, the loss of Ca2+ results in increased flexibility of W20 and

2+ 293 motions of the Ca binding loop (D19WCE(X)N24) which deforms the substrate binding

294 pocket and increases water accessibility to the Zn2+ site (Extended data Fig. 8). These

295 calculated alterations most likely represent a molecular explanation for the observed

296 loss of E33G ACER3 ceramidase activity in patients'cells.

297

298 Discussion

299 In this study, we solved the crystal structure of ACER3 revealing a seven-transmembrane

2+ 300 domain architecture harboring a catalytic Zn binding site nearly identical to the one we

21 301 recently uncovered in ADIPORs . Considering the fact that ACER3 and AdipoRs share a

2+ 302 similar fold, a common Zn catalytic site and similar functional enzymatic capability, it is

303 highly likely that ADIPORs are indeed genuine ceramidases. These discoveries are expanding

12 bioRxiv preprint doi: https://doi.org/10.1101/348219; this version posted June 15, 2018. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY-NC-ND 4.0 International license.

304 the family of ceramidases to, at least, seven members in humans. Given the established

305 functional link between fungal progesterone adipoQ receptor (PAQR) and ceramidase activity

26 27 306 , it is then tempting to speculate that all the members of the PAQR family (11 proteins in

307 humans including ADIPOR1 and ADIPOR2) share this functional characteristic with ACER3.

308 Moreover, our data provide a structural explanation of the previously demonstrated regulatory

2+ 12 309 role of Ca on ACER3 enzymatic activity . As it is known that high pH can increase the

2+ 310 binding of Ca , this might also constitute the molecular basis for the alkaline pH optima of

12 311 ACER ceramidase activity in vitro .

312 Although ADIPORs seem to be ligand-dependent ceramidases (activated by adiponectin), the

2+ 313 intracellularly residing ACER3 might be regulated through changes in cytoplasmic Ca

314 concentration and/or local pH. In addition, the ACER3 structure and computational studies

2+ 315 highlight the role of E33 in the formation of the Ca binding site and provide insights into the

316 loss of ceramidase activity of the E33G-ACER3 mutant found in leukodystrophic patients.

317 This knowledge will enable the development of pharmacological chaperones specifically

318 designed to restore the enzymatic function of ACER3 mutant. This may constitute a potential

319 treatment option for individuals diagnosed with ACER3 mutation when leukodystrophy is

320 suspected (at the onset) and before severe clinical conditions manifest. In addition, our study

321 opens the way to the structure-based discovery of small molecules able to control the

322 ceramidase activity of ACER3. Ultimately, such modulators could have beneficial effects

323 against the pathogenesis of diseases involving C18:1 ceramide and ACER3 dysregulation

324 including colon cancer and AML.

325

326

327

328

329

13 bioRxiv preprint doi: https://doi.org/10.1101/348219; this version posted June 15, 2018. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY-NC-ND 4.0 International license.

330 Online methods

331 332 ACER3 contructs

333 The full-length synthetic gene of human ACER3 (UniProtKB-Q9NUN7) was subcloned into

334 a modified pFastBac vector resulting in ACER3-BRIL construct bearing a C-terminal BRIL

20 335 soluble module (thermostabilized apocytochrome b562RIL ) and an N-terminal Flag epitope

336 followed by a tobacco etch virus (TEV) protease site. The full-length ACER3-BRIL construct

337 yielded diffraction quality crystals, however, those crystals diffracted to maximum 5 Å

338 resolution despite extensive attempts to optimize the crystallization conditions to improve

339 crystal quality. Therefore, based on an ab-initio model of ACER3 obtained from the I-

28 340 TASSER webserver , a putatively flexible C-terminal tail of ACER3 composed of 23 amino

341 acids were truncated which yielded crystals allowing to obtain high-resolution diffraction

342 data. E33G-ACER3 mutant was generated using the full-length ACER3 cloned to pFastBac

343 vector.

344

345 Expression and purification of ACER3 contructs

346 ACER3 constructs were expressed in Sf9 insect cells (Life technologies) using the pFastBac

347 baculovirus system (ThermoFisher) according to manufacturer’s instructions. Insect cells

6 348 were grown in suspension in EX-CELL® 420 medium (Sigma) to a density of 4 × 10 cells

349 per ml and infected with baculovirus encoding ACER3-BRIL. Cells were harvested by

350 centrifugation 48 h post-infection and stored at −80 °C until purification. After thawing the

351 frozen cell pellet, cells were lysed by osmotic shock in 10 mM Tris-HCl pH 7.5, 1 mM EDTA

−1 352 buffer containing 2 mg ml iodoacetamide and protease inhibitors. Lysed cells were

353 centrifuged and the enzyme extracted using a glass dounce tissue grinder in a solubilization

354 buffer containing 20 mM HEPES (pH 7.5), 100 mM NaCl, 1 % (w/v) n-dodecyl-β-D-

355 maltoside (DDM, Anatrace), 0.1 % (w/v) cholesteryl-hemi-succinate (CHS, Sigma),

14 bioRxiv preprint doi: https://doi.org/10.1101/348219; this version posted June 15, 2018. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY-NC-ND 4.0 International license.

−1 356 2 mg ml iodoacetamide and protease inhibitors. The extraction mixture was stirred for 1 h at

357 4 °C. The cleared supernatant was adjusted to the final concentration of 20 mM HEPES (pH

358 7.4), 200 mM NaCl, 0.5 % (w/v) DDM and 0.05% CHS and loaded by gravity flow onto anti-

359 Flag M2 antibody resin (Sigma). The resin was then washed with 2 column volumes (CV) of

360 DDM wash buffer containing 20 mM HEPES (pH 7.5), 200 mM NaCl, 0.025 % (w/v) DDM,

361 and 0.0001 % (w/v) CHS. While on the M2 antibody resin, the protein was exchanged into

362 lauryl maltose neopentyl glycol (MNG) detergent-containing buffer composed of 20 mM

363 HEPES (pH 7.5), 0.5% MNG-14, 200 mM NaCl. The detergent exchange was performed by

364 washing the column with a series of seven buffers (2 CV each) made up of the following

365 ratios (v/v) of MNG exchange buffer and DDM wash buffer: 1:3, 1:1, 3:1, 9:1, 19:1, 99:1, and

366 MNG exchange buffer alone. The column was then washed with 20× critical micelle

367 concentration (CMC) MNG buffer containing 20 mM HEPES (pH 7.5), 0.02% MNG-14 and

368 200 mM NaCl and the bound enzyme was eluted in the same buffer supplemented with

−1 369 0.4 mg ml Flag peptide. The eluted protein was concentrated to 500 µl using 100 kDa spin

370 filters and further purified by size exclusion chromatography on a Superdex 200 Increase

371 10/300 column (GE Healthcare) in 20×CMC MNG buffer. Fractions containing monodisperse

-1 372 ACER3-BRIL were collected and concentrated to 20 mg ml for crystallization trials.

373

374 Crystallization, data collection and processing

29 375 Crystallization of ACER3-BRIL was performed using the in meso method . Concentrated

376 ACER3-BRIL was reconstituted into 10:1 monoolein :cholesterol (Sigma) at a ratio of 1:1.5

377 protein:lipid by weight. Reconstitution was done using the coupled two-syringe method. The

378 resulting mesophase was dispensed onto a glass plate in 50-nl drops and overlaid with 900 nl

379 precipitant solution using a Gryphon LCP robot (Art Robbins Instruments). Crystals grew in

380 precipitant solution consisting of 34-40% PEG 400, 0.1 M Hepes pH 7.5, 75 mM magnesium

15 bioRxiv preprint doi: https://doi.org/10.1101/348219; this version posted June 15, 2018. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY-NC-ND 4.0 International license.

381 sulfate and 5% DMSO. Crystals were observed after one day and grew to full size (~20 µm x

382 20 µm x 30 µm) after five days. Crystals were harvested from the lipidic mesophase using

383 mesh grid loops and directly flash-frozen in liquid nitrogen.

30 384 Diffraction data were collected using in-meso crystallographic method at X06SA beamline

385 of the Swiss Light Source (SLS), Villigen, Switzerland. Diffraction measurement from

386 ACER3-BRIL microcrystals were carried out at tolerable dose under cryogenic condition

387 (~100K) with 10×10 μm2 micro-beam, providing 4.76 × 1011 photons/sec at 12.4 keV

388 (λ=1.0 Å, i.e., native) as well as 2.47 × 1011 photons/sec at 9.67 keV (λ=1.281 Å, i.e., Zn-

389 edge) X-ray energies. Typically, 10° wedge of data was collected from each microcrystal

390 (from ACER3 BRIL-native and Zn-edge) in oscillation step of 0.2° at a speed of 2°/sec using

391 EIGER16M detector with a sample-to-detector distance of 30 cm. A mesh-loop, containing

392 LCP-bolus with many micro-crystals, was raster grid-scanned with microbeam X-ray to

31 393 locate the crystals. The rastering procedure for crystal detection is described elsewhere . In

394 order to automate identification of tiny crystals, followed by data collection, a new graphical

32 395 user-interface, called CY+ GUI, was developed as an extension of DA+ software suite for

33 396 the beamline at the SLS. Processing of individual small wedges using XDS was automated

397 in DA+ software suite. 198 mini datasets (i.e., 10° wedge/dataset) were collected from

398 ACER3 BRIL native crystals at 12.4 keV X-ray energy. Individual small wedge of dataset

399 measures weak and sparse diffraction spots, which in turn results in partial measurement of

400 each reflection. Thereby, selecting datasets, which are correctly indexed in proper space

401 group, and scaling a subset of statistically equivalent datasets are extremely important to

402 generate a meaningful complete dataset for structural solution. An automated software suite

403 for serial synchrotron crystallographic data selection and merging procedure was developed at

404 the beamline (unpublished, code availability statement: the software will be made available

405 upon request). In order to identify correctly processed datasets, individual mini dataset was

16 bioRxiv preprint doi: https://doi.org/10.1101/348219; this version posted June 15, 2018. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY-NC-ND 4.0 International license.

406 first checked for indexing consistency, against user-provided reference unit-cell parameters.

407 Then, 194 datasets, identified as correctly indexed with a = 60.8 Å, b = 68.8 Å, and c = 257.0

33 408 Å in C2221 space-group, were scaled together using XSCALE program in XDS package.

34 409 After a first round of scaling, many datasets, which have very low ISa values (described in ),

410 were rejected against a cut-off value of 3.0. Thus, 77 datasets were selected out of 194

411 datasets, and scaled using XSCALE for second round. This yielded a scaled but unmerged

412 dataset with 100% completeness at 2.7Å resolution. The resolution cutoff was decided

35 413 based on CC1/2 cut-off of 0.3 . Similar procedure was applied to 297 mini-wedges of

414 datasets (i.e., 10°/dataset) of ACER3 BRIL-Zn collected at Zn-edge (i.e., λ=1.281 Å). Out 297

415 datasets, 198 datasets were selected based on indexing consistency and ISa cutoff of 3.0.

416 These 198 statistically equivalent datasets were scaled together using XSCALE program,

417 yielding a scaled but unmerged dataset with 100% completeness and high multiplicity at 3 Å

418 resolution. Later, these scaled but unmerged HKL files (output from XSCALE) were

419 converted into mtz files for structure determination.

420

421 Structure determination and refinement

422 The structure of ACER3-BRIL was determined by molecular replacement. Initial phases were

423 obtained using the BRIL molecule extracted from PDB 4RWD (chain B) as a search model in

36 37 424 PHASER . After density improvement using the CCP4 program parrot , the electron density

425 corresponding to ACER3 was still poor and it was only possible to build the C-terminal half

1/2 426 of helix 7 (h7 ) which is attached to the BRIL. Subsequently, an ab-initio model of ACER3

28 427 obtained from the I-TASSER webserver was cut into fragments of two transmembrane

428 helices, which were used as additional search models in PHASER while keeping the BRIL-

1/2 429 ACER3 h7 fragment as a fixed solution. This procedure allowed us to successfully place

430 two, and then four transmembrane helices. This in turn provided enough phase information to

17 bioRxiv preprint doi: https://doi.org/10.1101/348219; this version posted June 15, 2018. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY-NC-ND 4.0 International license.

38 39 431 enable manual rebuilding of ACER3 in COOT , as well as autobuilding in BUCCANEER .

40 432 Subsequently, the structure was refined using AUTOBUSTER . At late stages of refinement,

433 translation libration screw-motion (TLS) parameters generated within AUTOBUSTER were

41 434 used. MolProbity was used to assess the quality of the structure and indicated that 98% of

435 residues were within favored Ramachandran regions. The data collection and refinement

436 statistics are summarized in Extended Data Table 1. Data refinement and analysis software

42 22 437 were compiled and supported by the SBGrid Consortium . The Polder OMIT map and the

43 438 zinc difference anomalous map where calculated in PHENIX .

439

440 Docking calculations

441 Computational docking of N-Oleoyl-D-sphingosine (d18:1/18:1) (C18:1 ceramide) was

442 performed with the program Protein-Ligand ANT System (version 1.2) using as a receptor

443 our 2.7 Å structure in which all non-protein atoms except zinc and calcium were removed.

444 PLANTS combines an ant colony optimization algorithm with an empirical scoring function

44 445 for the prediction and scoring of binding poses in a protein structure . Ten poses were

446 calculated and scored by the plp scoring function at a speed setting of 1. Each of the 10 ligand

447 poses actually represents the best scoring pose from a cluster of solutions. The number of

448 iterations necessary for adequate sampling of the ligand (and optionally receptor) degrees of

449 freedom is determined automatically by the program and for this particular system was 1387,

7 450 resulting in about 5x10 scoring function evaluations. The binding pocket of ACER3 was

451 defined by all residues within a 25 Å radius around the zinc atom. All other options of

452 PLANTS were left at their default settings. The top scoring pose was selected and used as

453 input for MD simulations.

454 Of note, similar top scoring C18:1 ceramide binding poses were obtained using GlideXP as

455 well as the SWISSDOCK webserver (in blind docking mode) (Extended data Fig. 5c).

18 bioRxiv preprint doi: https://doi.org/10.1101/348219; this version posted June 15, 2018. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY-NC-ND 4.0 International license.

456

457 System preparation and Molecular dynamics simulations

458 Three protein-ligand systems were subjected to MD simulations. We initially performed two

459 independent simulations of unliganded ACER3 (i.e. the crystal structure in which the BRIL

460 moiety and the monoolein were deleted - system 1) of about ~ 300 ns. However, in this

461 system the absence of a bound lipid resulted in rapid water influx in the zinc and lipid binding

462 cavity and led to destabilization of the zinc binding site (and zinc unbinding). This system

463 was not analyzed further. Subsequently, we used a model of ACER3 in complex with the top

464 scoring ceramide-C18:1 (d18:1/18:1) docking pose (system 2). The crystallographic zinc and

465 calcium ions, as well as three ordered water molecules located close to the zinc binding site

466 were included in the system. The third system was identical to system 2 except that E33 was

467 mutated to a glycine. The wild type ACER3-ceramide-C18:1 and E33G mutant MD systems

45 468 were then set up using the CHARMM-GUI membrane builder . The proteins were inserted

469 into a hydrated, equilibrated bilayer composed of 80 molecules of 2-Oleoyl-1-palmitoyl-sn-

470 glycero-3-phosphocholine (POPC) and 20 molecules of cholesterol in the upper leaflet, and

471 115 molecules of ceramide-C18:1 in the lower leaflet. 32 sodium and 38 chloride ions were

472 added to neutralize the system, reaching a final concentration of approximately 150 mM.

473 Topologies and parameters for ceramide-C18:1 (d18:1/18:1) were available in the additive

46 474 all-atom CHARMM lipid force field .

475 Molecular dynamics calculations were performed in in GROMACS 2016.3 using the

476 CHARMM36 force field and the CHARMM TIP3P water model. The input systems were

477 subjected to energy minimization, equilibration and production simulation using the

47 478 GROMACS input scripts generated by CHARMM-GUI . Briefly, the system was energy

479 minimized using 5000 steps of steepest descent, followed by 375 ps of equilibration. NVT

480 (constant particle number, volume, and temperature) and NPT (constant particle number,

19 bioRxiv preprint doi: https://doi.org/10.1101/348219; this version posted June 15, 2018. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY-NC-ND 4.0 International license.

481 pressure, and temperature) equilibrations were followed by NPT production runs for both

482 systems. The van der Waals interactions were smoothly switched off at 10–12 Å by a force-

48 483 switching function , whereas the long-range electrostatic interactions were calculated using

49 484 the particle mesh Ewald method . The temperature and pressure were held at 310.15 K and 1

485 bar, respectively. The assembled systems were equilibrated by the well-established protocol

486 in Membrane Builder, in which various restraints were applied to the protein, lipids and water

487 molecules, and the restraint forces were gradually reduced during this process. During

488 production simulations an NPT ensemble was used with semi-isotropic pressure coupling via

489 the Parrinello-Rahman barostat method while the Nose-Hoover thermostat was used to

21 490 maintain a temperature of 310.15 K (as described in ). A leapfrog integration scheme was

491 used, and all bonds were constrained allowing for a time-step of 2 ps to be used during NPT

492 equilibration and production MD simulations. For both wild type and E33G mutant systems,

493 we performed five independent production runs of ~ 200 ns each. Production runs were

494 subsequently analyzed using GROMACS tools to yield root mean square deviations

495 (R.M.S.D.) and root mean square fluctuations (R.M.S.F.).

496

497 Ceramidase activity assay

498 The enzymatic activity of ACER3-BRIL was monitored by sphingosine detection and

499 quantification using liquid chromatography MS/MS analyses. All solvents and reagents used

500 were of highest available purity and purchased from typical suppliers. Lipids were purchased

501 from Avanti Polar Lipids.

502 Ceramidase activity assays were performed by incubating purified ACER3-BRIL (1 µM) with

503 ceramide (20 µM) for 1 hour at room temperature in 20 mM HEPES (pH 7.5), 200 mM NaCl,

504 0.025 % (w/v) DDM, and 0.0001 % (w/v) CHS. Reactions were quenched by addition of

20 bioRxiv preprint doi: https://doi.org/10.1101/348219; this version posted June 15, 2018. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY-NC-ND 4.0 International license.

505 methanol (final concentration 30%) which contained 68.5 pg of the internal standard,

506 sphingosine (d17:1).

50 507 Lipids were extracted from reaction samples using previously reported method . Briefly, a

508 mixture of dichloromethane/methanol (2% acetic acid)/water (2.5:2.5:2 v/v/v) was added to

509 each reaction and the solution was centrifuged. The organic phase was collected and dried

510 under nitrogen, then dissolved in 10 µL of MeOH. The lipid extract was stored at−20 °C

511 before LC-MS/MS analysis.

512 LC-MS/MS analysis of formed sphingosine was performed using an Agilent 1290 Infinity.

513 The samples were separated on an Acquity UPLC BEH-C8 column (particle size 1.7 µm, 2.1

514 x 100 mm) (Waters) maintained at 35 °C. The mobile phases consisted of eluent A (99.9%

515 H2O: 0.1% HCOOH, v/v) and eluent B (99.9% CH3CN: 0.1% HCOOH, v/v). The gradient

516 was as follows: 50% B at 0 min, 60% B at 2 min, 60% B at 3 min, 100% B at 4 min, 100% B

517 at 8,5 min and 50% B at 9 min. The flow rate was 0.3 mL/min. The auto sampler was set at 5

518 °C and the injection volume was 5 µL. The HPLC system was coupled on-line to an Agilent

519 6460 triple quadrupole MS equipped with electrospray ionization (ESI) source operated in

520 positive ion mode. The source parameters used were as follows: source temperature was set at

521 300 °C, nebulizer gas (nitrogen) flow rate was 10 L/min, sheath gas temperature was 300 °C,

522 sheath gas (nitrogen) flow rate was 12 L/min and the spray voltage was adjusted to +4000 V.

523 The collision energy optimums for sphingosine (d17:1) and sphingosine (d18:1) were 5 eV.

524 Analyses were performed in Selected Reaction Monitoring detection mode (SRM) using

525 nitrogen as collision gas. Peak detection, integration and quantitative analysis were done

526 using MassHunter QqQ Quantitative analysis software (Agilent Technologies) and Microsoft

527 Excel software.

528 Sphingosine (d18:1): precursor ion m/z=300.5 and product ion m/z =282.5.

529 Sphingosine (d17:1): precursor ion m/z=286.5 and product ion m/z=268.5.

21 bioRxiv preprint doi: https://doi.org/10.1101/348219; this version posted June 15, 2018. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY-NC-ND 4.0 International license.

530 The presented data for enzymatic activity are representative of three experiments performed

531 on three independent ACER3-BRIL enzyme preparations, each experiment contained six

532 replicates.

533

534 TR-FRET experiments in living cells

535 YFP-ADIPOR2 was kindly provided by the laboratory of Dr Vazquez-Martinez. For the

536 ADIPOR2-YFP construct, full-length AdipoR2 was inserted in frame of the N terminus of the

537 YFP in pcDNA 3.1-YFP by PCR using the following specific oligonucleotides: forward

538 5’TAAGCAGCTAGCATGAACGAGCCAACAGAAAACC 3’ and reverse

539 5’TAAGCAAAGCTTCAGTGCATCCTCTTCAC 3’. Full-length ACER3 was inserted in

540 frame of the N terminus of the SNAP tag in the pSNAPf vector (New England Biolabs) by

541 PCR using the following specific oligonucleotides:

542 5’AAACGCTAGCGATATCATGGACTACAAGGACGACGACGACAAGGCTCCTGCAG

543 CTGACAG 3’ as forward and 5’ TATGCAACCGGTGTGCTTCCTCAAGGGCT 3’ as

544 reverse.

545 HEK293 cells were grown in DMEM supplemented with 10% FBS and

546 antibiotics at 37°C, 5% CO2. Lipofectamine 2000-based reverse transfection were performed

547 in polyornithine coated black-walled, dark-bottom, 96-well plates. We used 30 ng of ACER3-

548 SNAP tag with either 50 ng of YFP-ADIPOR2, ADIPOR2-YFP or empty PRK6 as control

5 549 and 10 cells per well.

550 24 hours after transfection cells were fixed with 3.6% PFA and permeabilised with 0.05%

551 Triton X100. Then, intracellular ACER3-SNAP was labeled with BG-Lumi4-Tb (300 nM) for

552 1 hour at 37°C and after several washes, YFP and lumi4 fluorescence signals were observed

553 on an Infinite 500 (Tecan, Seestrasse, Switzerland). The signal was collected both at 520 nm

554 and 620 nm (10 flashes, 400 µs integration time, 150 µs lag time). Background values from

22 bioRxiv preprint doi: https://doi.org/10.1101/348219; this version posted June 15, 2018. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY-NC-ND 4.0 International license.

555 non-transfected cells were removed from raw data in each channel. Subsequently, the

556 normalized time resolved FRET ratio was obtained by dividing the specific acceptor signal at

557 520 nm by the specific donor signal at 620 nm data and plotted using GraphPad Prism

558 (GraphPad Software, Inc., San Diego, CA).

559

560

561

562

563

564

565

566

567

568

569

570

571

572

573

574

575

576

577

578

579

580

581

582

583

23 bioRxiv preprint doi: https://doi.org/10.1101/348219; this version posted June 15, 2018. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY-NC-ND 4.0 International license.

584 REFERENCES 585 586 1 Hannun, Y. A. & Obeid, L. M. Sphingolipids and their metabolism in physiology 587 and disease. Nat Rev Mol Cell Biol 19, 175-191, doi:10.1038/nrm.2017.107 588 (2018). 589 2 Ogretmen, B. & Hannun, Y. A. Biologically active sphingolipids in cancer 590 pathogenesis and treatment. Nat Rev Cancer 4, 604-616, doi:10.1038/nrc1411 591 (2004). 592 3 Coant, N., Sakamoto, W., Mao, C. & Hannun, Y. A. Ceramidases, roles in 593 sphingolipid metabolism and in health and disease. Adv Biol Regul 63, 122-131, 594 doi:10.1016/j.jbior.2016.10.002 (2017). 595 4 Nikolova-Karakashian, M. & Merrill, A. H., Jr. Ceramidases. Methods Enzymol 311, 596 194-201 (2000). 597 5 Kohama, T. et al. Molecular cloning and functional characterization of murine 598 sphingosine kinase. J Biol Chem 273, 23722-23728 (1998). 599 6 Koch, J. et al. Molecular cloning and characterization of a full-length 600 complementary DNA encoding human acid ceramidase. Identification Of the first 601 molecular lesion causing Farber disease. J Biol Chem 271, 33110-33115 (1996). 602 7 Gebai, A., Gorelik, A., Li, Z., Illes, K. & Nagar, B. Structural basis for the activation 603 of acid ceramidase. Nat Commun 9, 1621, doi:10.1038/s41467-018-03844-2 604 (2018). 605 8 El Bawab, S. et al. Molecular cloning and characterization of a human 606 mitochondrial ceramidase. J Biol Chem 275, 21508-21513, 607 doi:10.1074/jbc.M002522200 (2000). 608 9 Airola, M. V. et al. Structural Basis for Ceramide Recognition and Hydrolysis by 609 Human Neutral Ceramidase. Structure 23, 1482-1491, 610 doi:10.1016/j.str.2015.06.013 (2015). 611 10 Sun, W. et al. Upregulation of the human alkaline ceramidase 1 and acid 612 ceramidase mediates calcium-induced differentiation of epidermal keratinocytes. 613 The Journal of investigative dermatology 128, 389-397, 614 doi:10.1038/sj.jid.5701025 (2008). 615 11 Xu, R. et al. Golgi alkaline ceramidase regulates cell proliferation and survival by 616 controlling levels of sphingosine and S1P. Faseb J 20, 1813-1825, 617 doi:10.1096/fj.05-5689com (2006). 618 12 Mao, C. et al. Cloning and characterization of a novel human alkaline ceramidase. 619 A mammalian enzyme that hydrolyzes phytoceramide. J Biol Chem 276, 26577- 620 26588, doi:10.1074/jbc.M102818200 (2001). 621 13 Sun, W. et al. Substrate specificity, membrane topology, and activity regulation of 622 human alkaline ceramidase 2 (ACER2). J Biol Chem 285, 8995-9007, 623 doi:10.1074/jbc.M109.069203 (2010). 624 14 Hu, W. et al. Alkaline ceramidase 3 (ACER3) hydrolyzes unsaturated long-chain 625 ceramides, and its down-regulation inhibits both cell proliferation and apoptosis. 626 J Biol Chem 285, 7964-7976, doi:10.1074/jbc.M109.063586 (2010). 627 15 Edvardson, S. et al. Deficiency of the alkaline ceramidase ACER3 manifests in 628 early childhood by progressive leukodystrophy. J Med Genet 53, 389-396, 629 doi:10.1136/jmedgenet-2015-103457 (2016). 630 16 Wang, K. et al. Alkaline Ceramidase 3 Deficiency Results in Purkinje Cell 631 Degeneration and Cerebellar Ataxia Due to Dyshomeostasis of Sphingolipids in 632 the Brain. PLoS Genet 11, e1005591, doi:10.1371/journal.pgen.1005591 (2015).

24 bioRxiv preprint doi: https://doi.org/10.1101/348219; this version posted June 15, 2018. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY-NC-ND 4.0 International license.

633 17 Wang, K. et al. Alkaline ceramidase 3 deficiency aggravates colitis and colitis- 634 associated tumorigenesis in mice by hyperactivating the innate immune system. 635 Cell Death Dis 7, e2124, doi:10.1038/cddis.2016.36 (2016). 636 18 Chen, C. et al. ACER3 supports development of acute myeloid leukemia. Biochem 637 Biophys Res Commun 478, 33-38, doi:10.1016/j.bbrc.2016.07.099 (2016). 638 19 Rosenbaum, D. M. et al. GPCR engineering yields high-resolution structural 639 insights into beta2-adrenergic receptor function. Science 318, 1266-1273, 640 doi:10.1126/science.1150609 (2007). 641 20 Chu, R. et al. Redesign of a four-helix bundle protein by phage display coupled 642 with proteolysis and structural characterization by NMR and X-ray 643 crystallography. J Mol Biol 323, 253-262 (2002). 644 21 Vasiliauskaite-Brooks, I. et al. Structural insights into adiponectin receptors 645 suggest ceramidase activity. Nature 544, 120-123, doi:10.1038/nature21714 646 (2017). 647 22 Liebschner, D. et al. Polder maps: improving OMIT maps by excluding bulk 648 solvent. Acta Crystallogr D Struct Biol 73, 148-157, 649 doi:10.1107/S2059798316018210 (2017). 650 23 Hernick, M. & Fierke, C. A. Zinc hydrolases: the mechanisms of zinc-dependent 651 deacetylases. Arch Biochem Biophys 433, 71-84, doi:10.1016/j.abb.2004.08.006 652 (2005). 653 24 Pei, J., Millay, D. P., Olson, E. N. & Grishin, N. V. CREST--a large and diverse 654 superfamily of putative transmembrane hydrolases. Biol Direct 6, 37, 655 doi:10.1186/1745-6150-6-37 (2011). 656 25 Cauet, E., Rooman, M., Wintjens, R., Lievin, J. & Biot, C. Histidine-Aromatic 657 Interactions in Proteins and Protein-Ligand Complexes: Quantum Chemical Study 658 of X-ray and Model Structures. J Chem Theory Comput 1, 472-483, 659 doi:10.1021/ct049875k (2005). 660 26 Villa, N. Y. et al. Sphingolipids Function as Downstream Effectors of a Fungal 661 PAQR. Mol Pharmacol 75, 866-875, doi:10.1124/mol.108.049809 (2009). 662 27 Tang, Y. T. et al. PAQR proteins: a novel membrane receptor family defined by an 663 ancient 7-transmembrane pass motif. Journal of molecular evolution 61, 372-380, 664 doi:10.1007/s00239-004-0375-2 (2005). 665 28 Yang, J. & Zhang, Y. I-TASSER server: new development for protein structure and 666 function predictions. Nucleic Acids Res 43, W174-181, doi:10.1093/nar/gkv342 667 (2015). 668 29 Caffrey, M. & Cherezov, V. Crystallizing membrane proteins using lipidic 669 mesophases. Nat Protoc 4, 706-731, doi:10.1038/nprot.2009.31 (2009). 670 30 Huang, C. Y. et al. In meso in situ serial X-ray crystallography of soluble and 671 membrane proteins. Acta Crystallogr D Biol Crystallogr 71, 1238-1256, 672 doi:10.1107/S1399004715005210 (2015). 673 31 Wojdyla, J. A. et al. Fast two-dimensional grid and transmission X-ray microscopy 674 scanning methods for visualizing and characterizing protein crystals. J Appl 675 Crystallogr 49, 944-952, doi:10.1107/S1600576716006233 (2016). 676 32 Wojdyla, J. A. et al. DA+ data acquisition and analysis software at the Swiss Light 677 Source macromolecular crystallography beamlines. J Synchrotron Radiat 25, 293- 678 303, doi:10.1107/S1600577517014503 (2018). 679 33 Kabsch, W. Xds. Acta Crystallogr D Biol Crystallogr 66, 125-132, 680 doi:10.1107/S0907444909047337 (2010).

25 bioRxiv preprint doi: https://doi.org/10.1101/348219; this version posted June 15, 2018. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY-NC-ND 4.0 International license.

681 34 Diederichs, K. Quantifying instrument errors in macromolecular X-ray data sets. 682 Acta Crystallogr D Biol Crystallogr 66, 733-740, 683 doi:10.1107/S0907444910014836 (2010). 684 35 Karplus, P. A. & Diederichs, K. Linking crystallographic model and data quality. 685 Science 336, 1030-1033, doi:10.1126/science.1218231 (2012). 686 36 McCoy, A. J. et al. Phaser crystallographic software. J Appl Crystallogr 40, 658-674, 687 doi:10.1107/S0021889807021206 (2007). 688 37 Zhang, K. Y., Cowtan, K. & Main, P. Combining constraints for electron-density 689 modification. Methods Enzymol 277, 53-64 (1997). 690 38 Emsley, P., Lohkamp, B., Scott, W. G. & Cowtan, K. Features and development of 691 Coot. Acta Crystallogr D Biol Crystallogr 66, 486-501, 692 doi:10.1107/S0907444910007493 (2010). 693 39 Cowtan, K. Fitting molecular fragments into electron density. Acta Crystallogr D 694 Biol Crystallogr 64, 83-89, doi:10.1107/S0907444907033938 (2008). 695 40 Bricogne G., B. E., Brandl M., Flensburg C., Keller P., Paciorek W., & Roversi P, S. A., 696 Smart O.S., Vonrhein C., Womack T.O. . BUSTER 697 version X.Y.Z. Cambridge, United Kingdom: Global Phasing Ltd. 698 . (2016). 699 41 Chen, V. B. et al. MolProbity: all-atom structure validation for macromolecular 700 crystallography. Acta Crystallogr D Biol Crystallogr 66, 12-21, 701 doi:10.1107/S0907444909042073 (2010). 702 42 Morin, A. et al. Collaboration gets the most out of software. Elife 2, e01456, 703 doi:10.7554/eLife.01456 (2013). 704 43 Adams, P. D. et al. PHENIX: a comprehensive Python-based system for 705 macromolecular structure solution. Acta Crystallogr D Biol Crystallogr 66, 213- 706 221, doi:10.1107/S0907444909052925 (2010). 707 44 Korb, O., Stutzle, T. & Exner, T. E. Empirical scoring functions for advanced 708 protein-ligand docking with PLANTS. J Chem Inf Model 49, 84-96, 709 doi:10.1021/ci800298z (2009). 710 45 Wu, E. L. et al. CHARMM-GUI Membrane Builder toward realistic biological 711 membrane simulations. J Comput Chem 35, 1997-2004, doi:10.1002/jcc.23702 712 (2014). 713 46 Klauda, J. B. et al. Update of the CHARMM all-atom additive force field for lipids: 714 validation on six lipid types. J Phys Chem B 114, 7830-7843, 715 doi:10.1021/jp101759q (2010). 716 47 Lee, J. et al. CHARMM-GUI Input Generator for NAMD, GROMACS, AMBER, 717 OpenMM, and CHARMM/OpenMM Simulations Using the CHARMM36 Additive 718 Force Field. J Chem Theory Comput 12, 405-413, doi:10.1021/acs.jctc.5b00935 719 (2016). 720 48 Steinbach, P. J. & Brooks, B. R. New Spherical-Cutoff Methods for Long-Range 721 Forces in Macromolecular Simulation. Journal of Computational Chemistry 15, 722 667-683, doi:DOI 10.1002/jcc.540150702 (1994). 723 49 Essmann, U. et al. A Smooth Particle Mesh Ewald Method. J Chem Phys 103, 8577- 724 8593, doi:Doi 10.1063/1.470117 (1995). 725 50 Bligh, E. G. & Dyer, W. J. A rapid method of total lipid extraction and purification. 726 Can J Biochem Physiol 37, 911-917, doi:10.1139/o59-099 (1959). 727

26 bioRxiv preprint doi: https://doi.org/10.1101/348219; this version posted June 15, 2018. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY-NC-ND 4.0 International license.

728

729 730 ACKNOWLEDGMENTS

731 We are grateful to Vincent Oliéric and Takashi Tomikazi from PSI for providing assistance

732 in using beamline X06SA. We thank the MetaToul-Lipidomique Core Facility (I2MC, Inserm

733 1048, Toulouse, France), MetaboHUB-ANR-11-INBS-0010 for the sphingosine

734 quantification experiments. Funding: The research leading to these results has received

735 funding from the European Union’s Horizon 2020 research and innovation programme under

736 grant agreement n.° 730872, project CALIPSOplus. This project has received funding from

737 the European Research Council (ERC) under the European Union’s Horizon 2020 research

738 and innovation programme (grant agreement No 647687).

739

740 AUTHOR CONTRIBUTIONS

741 IVB expressed, purified, characterized and crystallized ACER3-BRIL preparations with

742 the help of PR, and TWA. RDH expressed, and purified ACER3-BRIL, optimized and

743 performed all the enzymatic assays with the help of TWA and IVB. MF performed the

744 TR-FRET experiments. EMS synthesized ceramides of different chain lengths and

745 performed HPLC-MS analysis of the ceramide cleavage reactions. CA supervised EMS.

746 RDH, CL, CG collected data with the help of FH. CL performed the computational studies

747 with help from CG. SB helped with the X-ray data acquisition and processing. CL solved

748 and refined the structure. RS prepared the figures. SG wrote the manuscript with the

749 help of RS, CL, IVB and RDH. All authors contributed to the manuscript preparation. SG

750 designed research and supervised the overall project.

751

752 AUTHOR INFORMATION

27 bioRxiv preprint doi: https://doi.org/10.1101/348219; this version posted June 15, 2018. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY-NC-ND 4.0 International license.

753 These authors contributed equally to this work: Ieva Vasiliauskaité-Brooks, Robert D.

754 Healey. Data availability statement: coordinates and structure factors for the ACER3-

755 BRIL structure have been deposited in the Protein Data Bank under accession number

756 6G7O.

757 The authors declare no competing financial interests. Correspondence and requests for

758 materials should be addressed to SG ([email protected]) or CL

759 ([email protected]).

760

28 bioRxiv preprint doi: https://doi.org/10.1101/348219; this version posted June 15, 2018. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY-NC-ND 4.0 International license.

761 Figure Legends:

762

763 Figure 1 | Crystal structure of ACER3. (a) Overall view of ACER3 crystal structure at

764 2.7 Å from within the membrane plane. The zinc and calcium ions are represented as

765 spheres with a van der Waals radius of 0.88 and 1.14, respectively. Side chains of

766 residues in close proximity to both ions are shown as sticks. (b) Close-up view of the N-

767 terminus (coloured from blue to red) and intracellular loop 3 (ICL3, coloured in light

768 blue) domains highlighting the disulfide bond formed by C21 and C196 (sticks). (c)

769 Cartoon representation of the N-terminus domain coloured as in (b) (left) revealing a

770 spiral-like motif (right).

771

772 Figure 2 | ACER3 intramembrane domains. (a) Views of the large hook-shaped

773 internal cavity shown as surface (cavity mode 1) within the 7TM helix bundle (shown in

774 light blue cartoon). The cavity is coloured according to the Eisenberg hydrophobicity

775 classification. (b)(c) Close-up views of the pocket on the top (b) and on the side (c) with

776 residues lining the pocket shown as sticks. (d) Close-up view of the zinc binding site

777 highlighting the residues forming the first coordination sphere of the Zn2+ shown as

778 sticks. The modeled water molecules are shown as blue spheres.

779

780 Figure 3 | Computational analyses of ceramide 18:1 docking into ACER3. (a) Top

781 scoring C18:1 ceramide (shown as stick) docking pose (hydrogens were omitted for

782 clarity) and observed ligand trajectory during a 216 ns long MD simulations (from blue

783 to red with ligand snapshots extracted every 8 ns). A representative all-atom RMSD of

784 the C18:1 ceramide is shown below (n=5 in total). (b) View of the ceramide docking

785 pose highlighting as sticks the residues in close proximity to the ceramide (coloured in

29 bioRxiv preprint doi: https://doi.org/10.1101/348219; this version posted June 15, 2018. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY-NC-ND 4.0 International license.

786 cyan). (c) Environment around the site of unsaturation of the docked C18:1 ceramide

787 with the side chains of polar residues close to the sp2 carbons shown as sticks. (d) Close-

788 up view of the Zn2+ catalytic site (side chains represented as sticks) with the putative

789 water molecule (red sphere) ideally positioned to attack the amide bond of the C18:1

790 ceramide.

791 In all panels, the Zn2+ and Ca2+ as well as the water molecule are represented as spheres

792 (coloured orange, green and red, respectively).

793

794 Figure 4| Comparison of ACER3 and ADIPORs structures. (a) Superposition of

795 ADIPOR1 (light green), ADIPOR2 (dark yellow) and ACER3 (light blue) represented

796 with cylindrical transmembrane helices viewed from the lumen (top) or from the

797 cytoplasm (bottom). A scheme with numbered TMs is also shown at the bottom to

798 highlight the similar architecture of TM1, 2, 3, 6 and 7. (b) Superposition of the three

799 structures showing the strict structural conservation of the Zn2+ catalytic site of ACER3,

800 ADIPOR2 and ADIPOR1 (coloured as above). The primary sequence with the H3SD motif

801 is also shown. (c) Views from within the membrane plane showing the differences in the

802 architecture of TM4 and TM5 relative to the Zn2+ (orange sphere) between ACER3,

803 ADIPOR2 and ADIPOR1 (coloured as described above). The bottom scheme represents

804 views from the cytoplasm, with the red dots highlighting the constant domain. This view

805 also reveals the large TM4 and TM5 shift between the three structures.

806

807 Figure 5| Structure and function of the N-terminus Ca2+ binding site of ACER3.

808 (a) Overall view of the N-terminus Ca2+ binding site of ACER3 crystal structure from

809 within the membrane plane. Residues participating in Ca2+ and Zn2+ coordination as well

810 as the disulfide bond formed by C21 and C196TM6 are shown as sticks with oxygen,

30 bioRxiv preprint doi: https://doi.org/10.1101/348219; this version posted June 15, 2018. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY-NC-ND 4.0 International license.

811 nitrogen, and sulfur atoms red, blue, and yellow, respectively. Crystallographic water

812 molecule next to the Zn2+ ion is shown as a red sphere. (b) Arrangement of polar

813 residues and the disulfide bond formed by C21 and C196TM6 around the Ca2+ ion in

814 ACER3 crystal structure viewed from the intracellular side. The black dashed lines

815 indicate polar contacts. (c) Overall ACER3 crystal structure (left panel) and Ca2+ binding

816 site (right panel) are shown as cartoons and coloured using the Consurf colour scale

817 according to degree of conservation ranging from not conserved, cyan, to highly

818 conserved, magenta. Conserved residues around Ca2+ and Zn2+ ions (green and orange

819 sphere, respectively) are displayed as sticks. (d) The two residues, W20 and E33, are

820 linking the Ca2+ and the Zn2+ sites by interacting simultaneously with both metal sites.

821 The top and bottom panels represent the observed structure and the initial pose of the

822 MD simulation, respectively. (e) Panel showing the minimum distances between E33

823 carboxylate and H81TM2 side chain or Ca2+ during MD simulations, reflecting the

824 observed hydrogen bond network. (f) Hydrophobic network around W20 side chain

825 forming the bottom part of the putative ceramide pocket.

826

827 Figure 6| MD simulations of ACER3 wild-type and E33G mutant.

828 Snapshots of the calculated N-terminus domain trajectories during a 240 ns long MD

829 simulation (from blue to red, taken every 10 ns) for the wild-type (a) or E33G mutant

830 (b) showing that the mutation leads to a more dynamic/unstable N-terminus domain.

831 (c)(d) Snapshots of the ACER3 wt (c) or E33G mutant (d) showing the key ligands

832 (Cer18:1, Zn2+ and Ca2+) and contacting residues (shown as sticks) trajectories during a

833 200 ns long MD simulations coloured as above and highlighting the motion/instability of

31 bioRxiv preprint doi: https://doi.org/10.1101/348219; this version posted June 15, 2018. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY-NC-ND 4.0 International license.

834 the W20 side chain and of the Ca2+ in the E33G mutant (d) in contrast to the wild-type

835 ACER3 (c).

836

837

838

839

840

841

842

843

844

845

846

847

848

849

850

851

852

853

854

855

856

857

858

32 bioRxiv preprint doi: https://doi.org/10.1101/348219; this version posted June 15, 2018. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY-NC-ND 4.0 International license.

859 Extended data TABLE 1 : Data collection and refinement statistics

860

861 Extended Data Figure 1| ACER3-BRIL crystal structure.

862 (a) Snake plot presenting the crystallized ACER3-BRIL construct and highlighting the

863 residues coordinating Ca2+ and Zn2+ (green and orange circles, respectively). The BRIL

864 module fused to the C-teminus as well as N-terminal Flag-tag and TEV cleavage site are

865 shown as black rectangulars. 23 residues at the C-terminal end of ACER3 were truncated

866 to get well diffracting crystals. (b) Detected sphingosine in LC-MS/MS analysis revealed

867 that ACER3-BRIL hydrolyzes C18:1 ceramide substrate in a preferred manner over

868 C18:0 ceramide in detergent micelles. Relative sphingosine values are represented as

869 the mean ± SD of three independent measurements performed in six replicates. (c)

870 ACER3-BRIL crystal structure viewed within the membrane plane with ACER3 and BRIL

871 coloured in light blue and wheat, respectively. (d) Cartoon representation of ACER3-

872 BRIL structure coloured according to the crystallographic B factors revealing increased

873 flexibility at the N-terminus of ACER3. Regions of lowest and highest B-factor are shown

874 in blue and yellow, respectively. Lattice packing of ACER3-BRIL crystals viewed within

875 the membrane plane (e,f) and extracellular side (g). The ACER3-BRIL fusion protein

876 crystallized with an anti-parallel arrangement of the molecules. ACER3 and BRIL are

877 coloured in light blue and wheat, respectively. Only ACER3 molecules are shown in left

878 panel of (e) coloured from blue to red.

879

880 Extended Data Figure 2| Comparisons of the 7TM fold and topology of ACER3 and

881 ADIPORs.

882 (a) Crystal structures of ADIPOR1 (PDB 5LXG) and ADIPOR2 (PDB 5LX9) shown as light

883 green and dark yellow, respectively, view from the membrane side. This view shows the

33 bioRxiv preprint doi: https://doi.org/10.1101/348219; this version posted June 15, 2018. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY-NC-ND 4.0 International license.

884 similar 7TM fold and topology between ACER3 and AdipoRs. (b) Schematic

885 representation of the different transfection conditions highlighting the positions of the

886 fluorescent tags used to monitor the topology of ACER3 and ADIPOR2 by Fluorescence

887 Resonance Energy Transfer (FRET) in living cells. (c) Graph representing the

888 quantification (as described in the methods section) of the specific FRET normalized by

889 the Terbium signal alone. Signal is higher when ACER 3 and ADIPOR2 have tags in the

890 same orientation. Quantification of specific fluorescence and Terbium signals are

891 represented in (d) and (e), respectively, and do not show changes between conditions.

892 The results shown are the mean ± sem of three independent experiments performed in

893 triplicate.

894

895 Extended Data Figure 3| Calculated electron density maps and tentatively

896 modeled ligands.

897 (a) Representation of the hook-shaped cavity within the ACER3 7TM and the position of

898 the modeled monoolein (b), the calculated 2Fo-Fc map contoured at 1 σ as well as the

899 polder OMIT map (ref 21 in the main text) contoured at 2.85 σ (c).

900 (e) Close-up view of the Zn2+ binding site in ACER3 with the anomalous difference maps

901 of ACER3 calculated from the data set collected at Zn-edge (i.e., λ=1.281 Å) at a

902 resolution of 3 Å and contoured at 4 σ (f), the 2Fo-Fc map contoured at 1 σ (g) and the

903 superposition of both maps (h).

904

905 Extended Data Figure 4| ConSurf color-coded multiple sequence alignment of

906 ACER family.

907 (a) Sequence alignment of alkaline ceramidases from human (Q9NUN7, Q8TDN7,

908 Q5QJU3), mouse (Q8R4X1, Q8VD53, Q9D099), rat (M0R603, D3ZNW4, F1M6P3), bovine

34 bioRxiv preprint doi: https://doi.org/10.1101/348219; this version posted June 15, 2018. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY-NC-ND 4.0 International license.

909 (F1MGH9, E1B8N2, A7MBH7), chicken (F1NNL6, E1C990, E1C413), salmon

910 (A0A1S3T5X1, A0A1S3RGK9, A0A1S3LR05), zebrafish (Q568I2, E7FCF4, E9QEA6), fruit

911 fly (Q9VIP7) and yeast (P38298, Q02896) created using Consurf server

912 (consurf.tau.ac.il/). The color-coding bar shows coloring scheme according to degree of

913 conservation ranging from not conserved, cyan to highly conserved, magenta. Residues

914 coordinating Ca2+ and Zn2+ ions are outstandingly conserved indicating the importance

915 of Ca2+ and Zn2+ binding sites. (b) Representation of the phylogenetic tree created using

916 Consurf server.

917

918 Extended Data Figure 5| Proposed general acid-base catalytic mechanism, MD

919 simulation of C18:1 ceramide and steric hindrance for ceramide derivatives.

920 (a) Based on the interactions with docked ceramide C18:1 and the Zn2+ catalytic site

921 architecture in the structure, we propose a general acid-base catalysis mechanism for

922 the hydrolysis of the amide bond by ACER3. In this mechanism, the zinc ion activates a

923 water molecule for nucleophilic attack of the amide carbon in which D92TM3 acts as a

924 proton acceptor/donor (1 and 3). W220 side chain polarizes the amide carbonyl and

925 stabilizes the oxyanion formed in the tetrahedral transition state (2). (b) Representation

926 of all-atom R.M.S.D. of the bound C18:1 ceramide in trajectory of five independent

927 simulations. (c) Top scoring C18:1 ceramide binding poses obtained using GlideXP, the

928 SWISSDOCK webserver (in blind docking mode) or PLANTS. (d) Steric hindrance close

929 to the primary alcohol of ceramide supporting a possible mechanism for substrate

930 selectivity.

931

932 Extended Data Figure 6| Comparison of the structural features of ACER3, ADIPORs

933 and neutral ceramidases.

35 bioRxiv preprint doi: https://doi.org/10.1101/348219; this version posted June 15, 2018. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY-NC-ND 4.0 International license.

934 (a) Structures of ACER3, ADIPOR1 and ADIPOR2 viewed from within the membrane

935 plane highlighting the distinct architecture on the intramembrane cavities (dark yellow,

936 blue and purple for ACER3, ADIPOR2 and ADIPOR1, respectively). (b) Calculated

937 ceramide C18:1 (coloured in cyan) binding modes in ACER3 and ADIPOR2.

938 (c) Comparison of the ACER3 and neutral ceramidases structures coloured according to

939 secondary structures (α-helices in cyan and β-sheet in purple) and showing the cavity

940 accomodating the ceramide substrate located in the lipid bilayer for ACER3 (dark

941 yellow) in contrast to the neutral ceramidase (NCD) which contains the ceramide

942 binding pocket outside the lipid bilayer (dark yellow). (d) Zoom of the Zn2+ catalytic site

943 of the NCD.

944

945 Extended Data Figure 7| Ca2+ coordination in the crystal structure and in MD

946 simulations.

947 (a) Overall view of ACER 3 in a lipid bilayer (grey) surrounded with water (cyan). Zoom

948 close to the Ca2+ ion (green spheres) illustrating the geometry of the Ca2+ binding site

949 during MD simulations (octahedral) with the water molecule shown as white and red

950 spheres (b) and in the crystal structure (incomplete pentagonal bipyramid) (c).

951

952 Extended Data Figure 8| MD simulations of ACER3 wild-type and E33G mutant.

953 (a) Residue-averaged root mean square fluctuations (R.M.S.F.) of ACER3 wild-type

954 (blue) and E33G mutant (red), highlighting the increase in flexibility observed in the N-

955 terminus region and ICL3. A zoom of the N-terminus and ICL3 regions are shown in

956 inset. (b) Comparison of the calcium ion R.M.S.D. in five independent wild-type and

957 mutant MD trajectories. (c) R.M.S.F. of selected ligands/residues in wild-type (blue) and

36 bioRxiv preprint doi: https://doi.org/10.1101/348219; this version posted June 15, 2018. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY-NC-ND 4.0 International license.

958 mutant (red) MD trajectories. The error bars correspond to the standard error

959 calculated from five independent trajectories.

960

37 bioRxiv preprint doi: https://doi.org/10.1101/348219; this version posted June 15, 2018. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY-NC-ND 4.0 International license.

Figure 1 | Crystal structure of ACER3.

a ACER3 b TM6 C-ter ECL1 TM1 TM7 ICL3 ECL2 ECL3 P208ICL3 TM7 C21Nter P207ICL3 ICL3 Lumen TM5 TM1 TM3 V206 K205ICL3 K204ICL3 C196TM6 R203ICL3 N-terminus

Zn2+ c TM1

TM6 TM2 TM4 Cytoplasm

ICL2 Ca2+ ICL3 ICL1

N-ter

Figure 1 | Crystal structure of ACER3. (a) Overall view of ACER3 crystal structure at 2.7 Å from within the membrane plane. The zinc and calcium ions are represented as spheres with a van der Waals radius of 0.88 and 1.14, respectively. Side chains of residues in close proximity to both ions are shown as sticks. (b) Close-up view of the N-terminus (coloured from blue to red) and intracellular loop 3 (ICL3, coloured in light blue) domains highlighting the disulfide bond formed by C21 and C196 (sticks). (c) Cartoon representation of the N-terminus domain coloured as in (b) (left) revealing a spiral-like motif (right). bioRxiv preprint doi: https://doi.org/10.1101/348219; this version posted June 15, 2018. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made Figure 2 | ACER3available intramembrane under aCC-BY-NC-ND domains 4.0 International. license. a b

Lumen H231TM7 TM7 TM7 TM6 TM5 S160TM6 TM6 S178TM5 TM5 TM1 TM3 TM1 H2O TM3 TM2 TM2 c V164TM5 Y176TM6 180° I161TM5

L179TM6 V157TM5

Eisenberg Hydrophobicity d Scale H221TM7 1.38 -2.53 S77TM2

TM3 H217TM7 D92

Cytoplasm H81TM2

Figure 2 | ACER3 intramembrane domains. (a) Views of the large hook-shaped internal cavity shown as surface (cavity mode 1) within the 7TM helix bundle (shown in light blue cartoon). The cavity is coloured according to the Eisenberg hydrophobicity classification. (b)(c) Close-up views of the pocket on the top (b) and on the side (c) with residues lining the pocket shown as sticks. (d) Close-up view of the zinc binding site highlighting the residues forming the first coordination sphere of the Zn2+ shown as sticks. The modeled water molecules are shown as blue spheres. bioRxiv preprint doi: https://doi.org/10.1101/348219; this version posted June 15, 2018. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY-NC-ND 4.0 International license.

Figure 3 | Computational analyses of ceramide 18:1 docking into ACER3.

L156TM5 a b F103TM3 V157TM5

V153TM5 M96TM3

V73TM2 F182TM6

M43TM1

C100TM3 S99TM3 S228TM7

Y149TM5

W220TM7

S77TM2 0 ns 216 ns

3 D92TM3 H217TM7

2 H81TM2

1 R.M.S.D. (Å ) 0 100 200 Time (ns)

Figure 3 | Computational analyses of ceramide 18:1 docking into ACER3. (a) Top scoring C18:1 ceramide (shown as stick) docking pose (hydrogens were omitted for clarity) and observed ligand trajectory during a 216 ns long MD simulations (from blue to red with ligand snapshots extracted every 8 ns). A representative all-atom RMSD of the C18:1 ceramide is shown below (n=5 in total). (b) View of the ceramide docking pose highlighting as sticks the residues in close proximity to the ceramide (coloured in cyan). (c) Environment around the site of unsaturation of the docked

C18:1 ceramide with the side chains of polar residues close to the sp2 carbons shown as sticks. (d) Close-up view of the Zn2+ catalytic site (side chains represented as sticks) with the putative water molecule (red sphere) ideally positioned to attack the amide bond of the C18:1 ceramide. In all panels, the Zn2+ and Ca2+ as well as the water molecule are represented as spheres (coloured orange, green and red, respectively). bioRxiv preprint doi: https://doi.org/10.1101/348219; this version posted June 15, 2018. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY-NC-ND 4.0 International license.

Figure 4| Comparison of ACER3 and ADIPORs structures. a view from lumen ADIPOR2 5lx9 ADIPOR1 b 5lxg ACER3

TM7 H3 STM2

DTM3

view from 1 7 cytoplasm 2 6 TM7 H2 3 5 2+ H TM2 4 Zn 1

1 7 2 6 3 2+ TM2 TM3 TM7 4 5 Zn S77-XXX-H81 D92 H217-XXX-H221 S198-XXX-H202 D219 H348-XXX-H352 1 7 S187-XXX-H191 D208 H337-XXX-H341 2 6 3 5 2+ 4 Zn c TM5 TM4 TM5 TM4 TM5 TM4

TM6 TM6 2+ TM6 Zn2+ Zn 6 6 6 5 7 5 7 7 5 1 1 1 4 4 4 2 3 2 2 3 3 Zn2+

Figure 4| Comparison of ACER3 and ADIPORs structures. (a) Superposition of AdipoR1 (light green), AdipoR2 (dark yellow) and ACER3 (light blue) represented with cylindrical transmembrane helices viewed from the lumen (top) or from the cytoplasm (bottom). A scheme with numbered TMs is also shown at the bottom to highlight the similar architecture of TM1, 2, 3, 6 and 7. (b) Superposition of the three structures showing the strict structural conservation of the Zn2+ catalytic site of ACER3, AdipoR2 and AdipoR1 (coloured as above). The primary sequence with the

H3SD motif is also shown. (c) Views from within the membrane plane showing the differences in the architecture of TM4 and TM5 relative to the Zn2+ (orange sphere) between ACER3, AdipoR2 and AdipoR1 (coloured as described above). The bottom scheme represents views from the cytoplasm, with the red dots highlighting the constant domain. This view also reveals the large TM4 and TM5 shift between the three structures. bioRxiv preprint doi: https://doi.org/10.1101/348219; this version posted June 15, 2018. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY-NC-ND 4.0 International license.

Figure 5| Structure and function of the N-terminus Ca2+ binding site of ACER3

membrane a c

Zn2+

Ca2+

Cytoplasm

W20 E33 C21 D19

E22 C196TM6 N24

Variable Conserved

d X-ray e

TM7 2.5 H217 E33-Ca2+ H81TM2 2.0

E33 W20 Distance (Å) 1.5 E33-H81TM2 0 100 200 Time (ns)

Molecular Dynamics f W189TM6

H217TM7 H81TM2

W20 F80TM2 W20 E33

L18 bioRxiv preprint doi: https://doi.org/10.1101/348219; this version posted June 15, 2018. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY-NC-ND 4.0 International license.

Figure 5| Structure and function of the N-terminus Ca2+ binding site of ACER3 (a) Overall view of the N-terminus Ca2+ binding site of ACER3 crystal structure from within the membrane plane. Residues participating in Ca2+ and Zn2+ coordination as well as the disulfide bond formed by C21 and C196TM6 are shown as sticks with oxygen, nitrogen, and sulfur atoms red, blue, and yellow, respectively. Crystallographic water molecule next to the Zn2+ ion is shown as a red sphere. (b) Arrangement of polar residues and the disulfide bond formed by C21 and C196TM6 around the Ca2+ ion in ACER3 crystal structure viewed from the intracellular side. The black dashed lines indicate polar contacts. (c) Overall ACER3 crystal structure (left panel) and Ca2+ binding site (right panel) are shown as cartoons and coloured using the Consurf colour scale according to degree of conservation ranging from not conserved, cyan to highly conserved, magenta. Conserved residues around Ca2+ and Zn2+ ions (green and orange sphere, respectively) are displayed as sticks. (d) The two residues, W20 and E33, are linking the Ca2+ and the Zn2+ sites by interacting simultaneously with both metal sites. The top and bottom panels represent the observed structure and the initial pose of the MD simulation, respectively. (e) Panel showing the minimum distances between E33 carboxylate and H81TM2 side chain or Ca2+ during MD simulations, reflecting the observed hydrogen bond network. (f) Hydrophobic network around W20 side chain forming the bottom part of the putative ceramide pocket. bioRxiv preprint doi: https://doi.org/10.1101/348219; this version posted June 15, 2018. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made Figure 6| MD simulationsavailable of under ACER3 aCC-BY-NC-ND wild-type 4.0 International and E33G license mutant..

a ACER3-WT b ACER3-E33G

0 ns 240 ns c d

CER18:1

Zn2+ W20 W20

E33 G33

Ca2+

Figure 6| MD simulations of ACER3 wild-type and E33G mutant. Snapshots of the calculated N-terminus domain trajectories during a 240 ns long MD simulations (from blue to red, taken every 10 ns) for the wild-type (a) or E33G mutant (b) showing that the mutation leads to a more dynamic/unstable N-terminus domain. (c)(d) Snapshots of the ACER3 wt (c) or E33G mutant (d) showing the key ligands (Cer18:1, Zn2+ and Ca2+) and contacting residues (shown as sticks) trajectories during a 200 ns long MD simulations coloured as above and highlighting the motion/instability of the W20 side chain and of the Ca2+ in the E33G mutant (d) in contrast to the wild-type ACER3 (c). bioRxiv preprint doi: https://doi.org/10.1101/348219; this version posted June 15, 2018. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made Extended Data Figure 1 | available. under aCC-BY-NC-ND 4.0 International license.

a b 400 ECL2 244 ECL1 ECL3 BRIL 300 TM7 TM1 TM3 TM5 Lumen 200 ( p mo les detected)

2+ Zn 10

5 TM2 TM4 TM6 Cytoplasm

ICL1 ICL2 h ingosine S S 0 Sp Ca2+

TEV FLAG ICL3 2 ceramide C18 N-ter ceramide C18:1

c d

e f g BRIL

Zn2+ 90° ACER3 Ca2+ Zn2+ 180°

180° 90°

Zn2+

Ca2+

Zn2+ 180° bioRxiv preprint doi: https://doi.org/10.1101/348219; this version posted June 15, 2018. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY-NC-ND 4.0 International license.

Extended Data Figure 1| ACER3-BRIL crystal structure. (a) Snake plot presenting the crystallized ACER3-BRIL construct and highlighting the residues coordinating Ca2+ and Zn2+ (green and orange circles, respectively). The BRIL module fused to the C-teminus as well as N-terminal Flag-tag and TEV cleavage site are shown as black rectangulars. 23 residues at the C-terminal end of ACER3 were truncated to get well diffracting crystals. (b) Detected sphingosine in LC-MS/MS analysis revealed that ACER3-BRIL hydrolyzes C18:1 ceramide substrate in a preferred manner over C18:0 ceramide in detergent micelles. Relative sphingosine values are represented as the mean ± SD of three independent measurements performed in six replicates. (c) ACER3-BRIL crystal structure viewed within the membrane plane with ACER3 and BRIL coloured in light blue and wheat, respectively. (d) Cartoon representation of ACER3-BRIL structure coloured according to the crystallographic B factors revealing increased flexibility at the N- terminus of ACER3. Regions of lowest and highest B-factor are shown in blue and yellow, respectively. Lattice packing of ACER3-BRIL crystals viewed within the membrane plane (e,f) and extracellular side (g). The ACER3-BRIL fusion protein crystallized with an anti-parallel arrangement of the molecules. ACER3 and BRIL are coloured in light blue and wheat, respectively. Only ACER3 molecules are shown in left panel of (e) coloured from blue to red. bioRxiv preprint doi: https://doi.org/10.1101/348219; this version posted June 15, 2018. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made Extended Data Figure 2 | available. under aCC-BY-NC-ND 4.0 International license. a ADIPOR2 (5lx9) ADIPOR1 (5lxg) C-ter TM7 C-ter TM7 ECL3 ECL1 ECL3 ECL1 ECL2 TM5 ECL2 TM5

Lumen Lumen TM3 TM3 TM1 TM1

Zn2+ Zn2+

TM4 TM4 TM6 TM6 Cytoplasm Cytoplasm ICL2 TM2 ICL2 TM2 ICL3 ICL3 N-ter ICL1 N-ter ICL1 b FRET CONTROL

C C C C C

N N N N N

ACER3-SNAP ADIPOR2-YFP ACER3-SNAP YFP-ADIPOR2 ACER3-SNAP

c d e

1.0 15 000

20 000

10 000

0.5 10 000 5 000

Normalised FRET / Tb (A.U.) 0.0 0 0 Specific Terbium signal (A.U.) Specific Fluorescence signal (A.U.) Control Control

YFP-ADIPOR2 ADIPOR2-YFP YFP-ADIPOR2 ADIPOR2-YFP YFP-ADIPOR2 ADIPOR2-YFP bioRxiv preprint doi: https://doi.org/10.1101/348219; this version posted June 15, 2018. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY-NC-ND 4.0 International license.

Extended Data Figure 2| Comparisons of the 7TM fold and topology of ACER3 and AdipoRs. (a) Crystal structures of ADIPOR1 (PDB 5LXG) and ADIPOR2 (PDB 5LX9) shown as light green and dark yellow, respectively, view from the membrane side. This view shows the similar 7TM fold and topology between ACER3 and AdipoRs. (b) Schematic representation of the different transfection conditions highlighting the positions of the fluorescent tags used to monitor the topology of ACER3 and AdipoR2 by Fluorescence Resonance Energy Transfer (FRET) in living cells. (c) Graph representing the quantification (as described in the methods section) of the specific FRET normalized by the Terbium signal alone. Signal is higher when ACER 3 and ADIPOR2 have tags in the same orientation. Quantification of specific fluorescence and Terbium signals are represented in (d) and (e), respectively, and do not show changes between conditions. The results shown are the mean ± sem of three independent experiments performed in triplicate. bioRxiv preprint doi: https://doi.org/10.1101/348219; this version posted June 15, 2018. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made Extended Data Figure 3 | . available under aCC-BY-NC-ND 4.0 International license.

a b c d

2FO-FC (1 σ) polder (2.85 σ)

Lumen

Cytoplasm

e f g h

anomalous (4 σ) 2FO-FC (1 σ)

S77TM2 H221TM7

D92TM3

H217TM7 H81TM2

Extended Data Figure 3| Calculated electron density maps and tentatively modeled ligands. (a) Representation of the hook-shaped cavity within the ACER3 7TM and the position of the modeled monoolein (b), the calculated 2Fo-Fc map contoured at 1 as well as the polder OMIT map (ref 21 in the main text) contoured at 2.85 (c). (e) Close-up view of the Zn2+ binding site in ACER3 with the anomalous difference maps of ACER3 calculated from the data set collected at Zn-edge (i.e., =1.281 Å) at a resolution of 3 Å and contoured at 4 (f), the 2Fo-Fc map contoured at 1 (g) and the superposition of both maps (h). bioRxiv preprint doi: https://doi.org/10.1101/348219; this version posted June 15, 2018. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made Extended Data Figure 4 | . available under aCC-BY-NC-ND 4.0 International license. a

Variable b

Conserved bioRxiv preprint doi: https://doi.org/10.1101/348219; this version posted June 15, 2018. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY-NC-ND 4.0 International license.

Extended Data Figure 4| ConSurf color-coded multiple sequence alignment of ACER family. (a) Sequence alignment of alkaline ceramidases from human (Q9NUN7, Q8TDN7, Q5QJU3), mouse (Q8R4X1, Q8VD53, Q9D099), rat (M0R603, D3ZNW4, F1M6P3), bovine (F1MGH9, E1B8N2, A7MBH7), chicken (F1NNL6, E1C990, E1C413), salmon (A0A1S3T5X1, A0A1S3RGK9, A0A1S3LR05), zebrafish (Q568I2, E7FCF4, E9QEA6), fruit fly (Q9VIP7) and yeast (P38298, Q02896) created using Consurf server (consurf.tau.ac.il/). The color-coding bar shows coloring scheme according to degree of conservation ranging from not conserved, cyan to highly conserved, magenta. Residues coordinating Ca2+ and Zn2+ ions are outstandingly conserved indicating the importance of Ca2+ and Zn2+ binding sites. (b) Representation of the phylogenetic tree created using Consurf server. bioRxiv preprint doi: https://doi.org/10.1101/348219; this version posted June 15, 2018. The copyright holder for this preprint (which was Extendednot certified by Data peer review) Figure is the 5author/funder, | . who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY-NC-ND 4.0 International license. H217 a 1 H221 b 3 H H H81 N N N #1 N N N H 2 2 H217 W220 H221 2+ S77 OH R.M.S.D. (Å ) Zn H H H81 H N NH N O N 3 N N O H O D92 N H - W220 S77 OH Zn2+ OH 2 #2 NH O - FA NH O D92 R.M.S.D. (Å ) SPH O HO OH 3 H OH NH O FA + H O - FA 2 2 SPH + CER - SPH #3 3 OH R.M.S.D. (Å )

H217 H221 3 H H H81 N N N N N N H 2 #4 W220

S77 R.M.S.D. (Å ) Zn2+ HO NH 3 O H D92 - O O 2 O #5 OH H N FA 2 R.M.S.D. (Å )

SPH 0 100 200 OH Time (ns) c d ceramide

HO SPH

OH FA O

sphingomyelin ceramide-1-phosphate glucosylceramide

HO SPH HO SPH HO SPH

NH NH OH NH O O O FA O FA HO O FA P O P O O O O XPGlide O O CER18:1 SWISSDOCK O HO PLANTS + N

OH bioRxiv preprint doi: https://doi.org/10.1101/348219; this version posted June 15, 2018. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY-NC-ND 4.0 International license.

Extended Data Figure 5| Proposed general acid-base catalytic mechanism, MD simulation of C18:1 ceramide and steric hindrance for ceramide derivatives. (a) Based on the interactions with docked ceramide C18:1 and the Zn2+ catalytic site architecture in the structure, we propose a general acid-base catalysis mechanism for the hydrolysis of the amide bond by ACER3. In this mechanism, the zinc ion activates a water molecule for nucleophilic attack of the amide carbon in which D92TM3 acts as a proton acceptor/donor (1 and 3). W220 side chain polarizes the amide carbonyl and stabilizes the oxyanion formed in the tetrahedral transition state (2). (b) Representation of all-atom R.M.S.D. of the bound C18:1 ceramide in trajectory of five independent simulations. (c) Top scoring C18:1 ceramide binding poses obtained using GlideXP, the SWISSDOCK webserver (in blind docking mode) or PLANTS. (d) Steric hindrance close to the primary alcohol of ceramide supporting a possible mechanism for substrate selectivity. bioRxiv preprint doi: https://doi.org/10.1101/348219; this version posted June 15, 2018. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made Extended Data Figure 6 available| . under aCC-BY-NC-ND 4.0 International license. a ADIPOR1 ADIPOR2 ACER3 5lxg 5lx9

C-ter C-ter C-ter

Lumen

Zn2+ Zn2+ Zn2+

N-ter N-ter

N-ter Cytoplasm b ACER3 ADIPOR2 5lx9

CER18:1 CER18:1

Zn2+ Zn2+

c NCD d

Lumen H303 H194 Zn2+

E540

H196

Y579

Zn2+ ACER3 Ca2+ Cytoplasm

Extended Data Figure 6| Comparison of the structural features of ACER3, AdipoRs and neutral ceramidases. (a) Structures of ACER3, AdipoR1 and AdipoR2 viewed from within the membrane plane highlighting the distinct architecture on the intramembrane cavities (dark yellow, blue and purple for ACER3, AdipoR2 and AdipoR1, respectively). (b) Calculated ceramide C18:1 (coloured in cyan) binding modes in ACER3 and AdipoR2. (c) Comparison of the ACER3 and neutral ceramidases structures coloured according to secondary structures ( -helices in cyan and -sheet in purple) and showing the cavity accomodating the ceramide substrate located in the lipid bilayer for ACER3 (dark yellow) in contrast to the neutral ceramidase (NCD) which contains the ceramide binding pocket outside the lipid bilayer (dark yellow). (d) Zoom of the Zn2+ catalytic site of the NCD. bioRxiv preprint doi: https://doi.org/10.1101/348219; this version posted June 15, 2018. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY-NC-ND 4.0 International license.

Extended Data Figure 7 | .

a b c ACER3 E33 Lumen E33 W20 W20

E22 E22

N24 D19 D19 Cytoplasm N24 H2O

E33 E33 W20 E22 W20 E22 Ca2+ Ca2+ D19 N24 D19 N24 D19

H2O

Extended Data Figure 7| Ca2+ coordination in the crystal structure and in MD simulations. (a) Overall view of ACER 3 in a lipid bilayer (grey) surrounded with water (cyan). Zoom close to the Ca2+ ion (green spheres) illustrating the geometry of the Ca2+ binding site during MD simulations (octahedral) with the water molecule shown as white and red spheres (b) and in the crystal structure (incomplete pentagonal bipyramid) (c). bioRxiv preprint doi: https://doi.org/10.1101/348219; this version posted June 15, 2018. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY-NC-ND 4.0 International license. Extended Data Figure 8 | .

a N-terminus ICL3 5 5

4 4

3 3

2 2

1 1

10 20 30 190 200 210

4 ACER3-WT ACER3-E33G 3

1 residue-averaged R.M.S.F. (Å )

50 100 150 200 250 Residue number

b Ca2+: ACER3-WT c Ca2+: ACER3-E33G ACER3-WT 12 ACER3-E33G

10 4

8 3

6 2 R.M.S.D. (Å ) 2+ 4 R.M.S.F (Å ) 1 Ca 2 0 Ca2+ Zn2+ CER18:1 W20 0 100 200 Time (ns)

Extended Data Figure 8| MD simulations of ACER3 wild-type and E33G mutant. (a) Residue-averaged root mean square fluctuations (R.M.S.F.) of ACER3 wild-type (blue) and E33G mutant (red), highlighting the increase in flexibility observed in the N-terminus region and ICL3. A zoom of the N-terminus and ICL3 regions are shown in inset. (b) Comparison of the calcium ion R.M.S.D. in five independent wild-type and mutant MD trajectories. (c) R.M.S.F. of selected ligands/residues in wild-type (blue) and mutant (red) MD trajectories. The error bars correspond to the standard error calculated from five independent trajectories. bioRxiv preprint doi: https://doi.org/10.1101/348219; this version posted June 15, 2018. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY-NC-ND 4.0 International license.

Extended Data Table 1: Data collection and refinement statistics (molecular replacement)

ACER3-BRIL, native ACER3-BRIL, Zn edge Data collection No. of wedges 77 198 Space group C2221 C2221 Cell dimensions a, b, c (Å) 60.88, 68.83, 257.52 61.41, 69.69, 258.15 α, β, γ (°) 90.00, 90.00, 90.00 90.00, 90.00, 90.00 Resolution (Å) 45.60-2.70 (2.83-2.70) * 46.07-2.85 (3.00-2.85) Rmerge 0.566 (NA) 0.614 (NA) Rpim 0.110 (0.875) 0.102 (NA) CC1/2 0.964 (0.496) 0.999 (0.349) I / σI 8.1 (1.1) 7.8 (0.7) Completeness (%) 100.0 (100.0) 100.0 (100.0) Redundancy 27.6 (27.7) 68.8 (65.2)

Refinement Resolution (Å) 45.60-2.70 - No. reflections 15331 - Rwork / Rfree 24.29/24.67 - No. atoms Protein 2855 - Zn2+ 1 - Ca2+ 1 - Mg2+ 2 Na+ 3 2- SO4 30 Mono-olein 25 - Water 109 - B-factors Protein 70.39 - Zn2+ 100.30 - Ca2+ 115.34 - Mg2+ 81.07 Na+ 74.61 2- SO4 96.12 Mono-olein 81.16 - Water 59.36 - R.m.s. deviations Bond lengths (Å) 0.021 - Bond angles (°) 1.150 -

*Values in parentheses are for highest-resolution shell. NA-not applicable, Rmerge value over 1 is statistically meaningless.