Structural Study of the Family

Alexei Gorelik

Department of Biochemistry McGill University, Montreal August 2017

A thesis submitted to McGill University in partial fulfillment of the requirements of the degree of Doctor of Philosophy

© Alexei Gorelik, 2017 Abstract

The acid sphingomyelinase (ASMase) converts the lipid (SM) to . This protein participates in lysosomal lipid and plays an additional role in signal transduction at the cell surface by cleaving the abundant SM to ceramide, thus modulating membrane properties. These functions are enabled by the ’s lipid- and membrane- interacting saposin domain. ASMase is part of a small family along with the poorly characterized ASMase-like 3A and 3B (SMPDL3A,B). SMPDL3A does not hydrolyze SM but degrades extracellular nucleotides, and is potentially involved in purinergic signaling. SMPDL3B is a regulator of the innate immune response and podocyte function, and displays a partially defined lipid- and membrane-modifying activity. I carried out structural studies to gain insight into substrate recognition and molecular functions of the ASMase family of . Crystal structures of SMPDL3A uncovered the helical fold of a novel C-terminal subdomain, a slightly distinct catalytic mechanism, and a nucleotide-binding mode without specific contacts to their nucleoside moiety. The ASMase investigation revealed a conformational flexibility of its saposin domain: this module can switch from a detached, closed conformation to an open form which establishes a hydrophobic interface to the catalytic domain. This open configuration represents the active form of the enzyme, likely allowing lipid access to the . The SMPDL3B structure showed a narrow, boot-shaped substrate that accommodates the head group of SM. However, no in vitro lipid hydrolysis could be detected; further work is required to identify potential bona fide substrates. In summary, these studies illustrate how an enzyme family can adapt a conserved architecture and mechanism to perform divergent functions.

2 Résume

La sphingomyélinase acide (ASMase) produit le lipide céramide à partir de la sphingomyéline (SM). Cette protéine contribue au métabolisme des lipides dans le et participe à la transduction de signaux sur la membrane plasmique en altérant les propriétés de cette dernière par son action. Ces fonctions sont accomplies avec l’aide du domaine saposine de cette enzyme, qui interagit avec les lipides et les membranes. ASMase fait partie d’une famille de protéines dont les deux autres membres sont mal caractérisés : les phosphodiestérases SMPDL3A et B. SMPDL3A n’hydrolyse point la SM mais dégrade par contre les nucléotides extracellulaires, ce qui implique potentiellement cette enzyme dans la signalisation purinergique. SMPDL3B régule les réponses immunitaires innées ainsi que le fonctionnement des podocytes par l’entremise d’une activité incomplètement définie de modification de lipides et de membranes. J’ai mené des études structurales sur cette famille de protéines afin de mieux comprendre leurs fonctions moléculaires et la reconnaissance des substrats par ces . Les structures cristallines de SMPDL3A ont révélé le repliement hélicoïdal d’un nouveau sous-domaine C- terminal, un mécanisme catalytique légèrement modifié, et un mode de liaison aux nucléotides dénué de contacts spécifiques avec leur portion nucléoside. L’investigation de l’ASMase a démontré la flexibilité conformationnelle de son domaine saposine : ce module alterne entre une configuration fermée détachée, et une forme ouverte qui interagit avec le domaine catalytique via une interface hydrophobe. Cette conformation ouverte représente la forme active de la protéine, probablement en facilitant l’accès des lipides au site actif. La structure de SMPDL3B a dévoilé un site de liaison au substrat étroit en forme de botte, où le groupe de tête de la SM peut se positionner. Par contre, aucune hydrolyse de lipides n’a pu être démontrée in vitro; des travaux additionnels sont requis afin d’identifier des substrats authentiques. En résume, ces études illustrent la capacité d’une famille d’enzymes à adapter une structure et un mécanisme communs vers des fonctions variées.

3 Table of contents

Abstract 2 Résumé 3 Table of contents 4 List of figures 7 List of tables 9 List of abbreviations 10 Preface 12 Author contributions 13 Acknowledgements 14

Chapter 1 – General introduction 15 1.1 The -like phosphoesterases 15 1.1.1 Shared structural features 17 1.1.2 PPP-type 20 1.1.3 Other serine/threonine phosphatases 20 1.1.4 21 1.1.5 21 1.1.6 Enzymes of other or unknown function 22 1.1.7 Bacterial enzymes 23 1.1.8 Acid sphingomyelinase-like proteins 23 1.2 Acid sphingomyelinase 24 1.2.1 24 1.2.2 Sphingomyelinases 26 1.2.3 Niemann-Pick disease types A and B 27 1.2.4 Lysosomal roles of ASMase 28 1.2.5 Saposins 29 1.2.6 Roles of ASMase at the cell surface 32 1.2.7 ASMase in the circulation 34 1.3 SMPDL3B 34

4 1.3.1 Lipid rafts in toll-like receptor signaling 35 1.3.2 ASMase in TLR signaling 35 1.3.3 SMPDL3B in TLR signaling 36 1.3.4 Lipid rafts in podocytes 38 1.3.5 SMPDL3B in podocyte function 38 1.3.6 SMPDL3B outside of the cell 40 1.4 SMPDL3A 41 1.4.1 SMPDL3A in the cholesterol response 41 1.4.2 Purinergic signaling 42 1.5 Conclusion 43

Chapter 2 – Study of SMPDL3A 45 Introductory transition 45 2.1 Abstract 46 2.2 Introduction 47 2.3 Results 49 2.4 Discussion 65 2.5 Experimental procedures 67 Concluding transition 69

Chapter 3 – Study of ASMase 74 Introductory transition 74 3.1 Abstract 75 3.2 Introduction 76 3.3 Results 78 3.4 Discussion 103 3.5 Experimental procedures 105 Concluding transition 109

Chapter 4 – Study of SMPDL3B 116 Introductory transition 116

5 4.1 Abstract 117 4.2 Introduction 118 4.3 Results 121 4.4 Discussion 134 4.5 Experimental procedures 138 Concluding transition 141

Chapter 5 – Conclusion 144 References 148

6 List of figures

Figure 1.1 Mammalian members of the calcineurin-like phosphoesterase superfamily 16 Figure 1.2 Structural superimposition of calcineurin-like phosphoesterases 18 Figure 1.3 Active site of calcineurin-like phosphoesterases 19 Figure 1.4 Chemical structures of representative lipid species 25 Figure 1.5 Structural flexibility of saposins 31

Figure 2.1 Sequence alignment of SMPDL3A and ASMase 50 Figure 2.2 Structure of SMPDL3A 52 Figure 2.3 Active site of SMPDL3A 54 Figure 2.4 Inhibition of SMPDL3A by excess zinc 56 Figure 2.5 Ligands bound in the active site of SMPDL3A 58 Figure 2.6 Proposed reaction mechanism for SMPDL3A 61 Figure 2.7 Nucleotide binding modes of human and murine SMPDL3A 70

Figure 3.1 Structural overview of ASMase 79 Figure 3.2 The ASMase fold 81 Figure 3.3 Comparison and mutation of the ASMase active site and proposed catalytic mechanism 84 Figure 3.4 ASMasesap – ASMasecat interactions 87 Figure 3.5 Purification of murine ASMase and bNPP hydrolysis assay 90 Figure 3.6 Purification of human ASMase and activity assays 92 Figure 3.7 Main chain B-factor analysis of catalytic domain interface loops 94 Figure 3.8 Saposin dimers in the crystal lattice 96 Figure 3.9 Electrostatic surface and substrate binding site of ASMase 97 Figure 3.10 Liposomal activity assay of ASMase in the presence of cationic amphiphilic drugs (CADs) 98 Figure 3.11 Electron density and binding mechanism for the co-crystallized inhibitor AbPA 100 Figure 3.12 Structural mapping of disease mutations 102 Figure 3.13 Schematic for ASMase activation 104

7 Figure 3.14 Representative electron density maps of the ASMase structures 108 Figure 3.15 PC recognition by human ASMase 110 Figure 3.16 Lipid delivery to GALC by SapA 112

Figure 4.1 Crystal structure of SMPDL3B 122 Figure 4.2 Structural comparison of SMPDL3B, SMPDL3A and ASMase 124 Figure 4.3 Region around active site 126 Figure 4.4 Phosphocholine bound in the active site of SMPDL3B 128 Figure 4.5 In vitro enzymatic activity against various substrates 130 Figure 4.6 Functional impact of SMPDL3B mutants on LPS-induced IL-6 release in macrophages 133 Figure 4.7 Proposed binding modes of potential substrates 136 Figure 4.8 Putative rituximab epitope on SMPDL3B 142

8 List of tables

Table 2.1 SMPDL3A X-ray data collection and structure refinement statistics 71 Table 2.2 Michaelis-Menten parameters of ATP hydrolysis by SMPDL3A 73 Table 3.1 ASMase X-ray data collection and structure refinement statistics 78 Table 3.2 Predicted effects of ASMase mutations found in Niemann-Pick patients 101 Table 3.3 Predicted effects of ASMase variants of unknown significance 115 Table 4.1 SMPDL3B X-ray data collection and structure refinement statistics 143

9 List of abbreviations

AbPA – 1-aminodecylidene bis-phosphonic acid alk-SMase – alkaline sphingomyelinase AMPCPP – ɑ,β-methylene ATP AP4A – diadenosine tetraphosphate ASMase – acid sphingomyelinase BMP – bis(monoacylglycero)phosphate bNPP – bis(4-nitrophenyl) phosphate CAD – cationic amphiphilic drug CTD – C-terminal domain / subdomain DKD – diabetic kidney disease E-NPP – ecto-nucleotide pyrophosphatase / E-NTPD – ecto-nucleoside triphosphate diphosphohydrolase ERT – enzyme replacement therapy FSGS – focal segmental glomerulosclerosis GPI – glycosylphosphatidylinositol IL-6 – interleukin-6 LDL – low density lipoprotein LPS – lipopolysaccharide M6P – mannose-6-phosphate NP – Niemann-Pick NPPC – p-nitrophenylphosphorylcholine PC – phosphocholine pNP-PC – p-nitrophenyl phosphorylcholine pNP-TMP – p-nitrophenyl thymidine 5'-monophosphate SM – sphingomyelin SMase – sphingomyelinase SMPD1 – sphingomyelin phosphodiesterase 1 SMPDL3A – acid sphingomyelinase-like phosphodiesterase 3a SMPDL3B – acid sphingomyelinase-like phosphodiesterase 3b

10 suPAR – soluble urokinase-type plasminogen activator TLR – toll-like receptor TM – transmembrane TNFα – tumor necrosis factor α uPAR – urokinase-type plasminogen activator

11 Preface

This is a manuscript-based thesis consisting of three published articles [211,221,228], in chronological order:

Chapter 2 Gorelik A, Illes K, Superti-Furga G, Nagar B. Structural Basis for Nucleotide Hydrolysis by the Acid Sphingomyelinase-like Phosphodiesterase SMPDL3A. J Biol Chem. 2016 Mar 18;291(12):6376-85.

Chapter 3 Gorelik A, Illes K, Heinz LX, Superti-Furga G, Nagar B. Crystal structure of mammalian acid sphingomyelinase. Nat Commun. 2016 Jul 20;7:12196.

Chapter 4 Gorelik A, Heinz LX, Illes K, Superti-Furga G, Nagar B. Crystal Structure of the Acid Sphingomyelinase-like Phosphodiesterase SMPDL3B Provides Insights into Determinants of Substrate Specificity. J Biol Chem. 2016 Nov 11;291(46):24054-24064.

12 Author contributions

- Dr. Leonhard X. Heinz provided expression plasmids for all three projects. He also designed, carried out and analyzed the IL-6 release experiments in macrophages for the SMPDL3B project (Fig. 4.6).

- Dr. Giulio Superti-Furga supervised Dr. Leonhard X. Heinz.

- Dr. Bhushan Nagar supervised Alexei Gorelik, helped with the initial structure determination of SMPDL3A, provided advice for all three projects and participated in manuscript preparation.

- Katalin Illes performed baculovirus generation and insect cell culture, helped with the initial purifications of SMPDL3A and provided advice for all three projects.

- Alexei Gorelik designed the projects, carried out the experimental work (besides what is listed above), analyzed the data and participated in manuscript preparation.

13 Acknowledgements

- Dr. Bhushan Nagar, for accepting me into his lab, always being available for questions and advice, and giving me the freedom to work on any interesting projects I want.

- Katalin Illes, for making this work possible, and for being the scientific Yin to my Yang.

- Our collaborators Dr. Leonhard X. Heinz and Dr. Giulio Superti-Furga, for bringing the SMPDL3A, ASMase and SMPDL3B projects to us, thus changing the course of my graduate studies.

- Dr. Shaun Labiuk along with the rest of the CMCF staff for crystallographic data collection at the CLS, Dr. Guennadi Kozlov for crystallographic data collection at CHESS, and Dr. John Silvius for help with liposome preparation.

- Lab members from whom I learned so much: Dr. Yazan Abbas and Rose Szittner.

- My research advisory committee: Dr. Kalle Gehring, Dr. John Silvius and Dr. Morag Park.

- Funding: Dr. Bhushan Nagar, CIHR, FRSQ, the Chemical Biology program, the CREATE Training Program in Bionanomachines, GRASP, the Department of Biochemistry, and McGill University.

14 Chapter 1 – General introduction

The premise of structural biology is that knowledge of the three-dimensional structure of proteins helps understanding their functions. The work covered in this thesis was centered on a family of three proteins, the acid sphingomyelinase (ASMase) and the acid sphingomyelinase- like phosphodiesterases 3A and 3B (SMPDL3A, SMPDL3B). Whereas the roles of ASMase are relatively well-characterized, the specific purposes of its two paralogs are still uncertain. I investigated the structure of these three enzymes in order to learn the molecular details of their activity and ultimately contribute to our understanding of their physiological functions. In this introduction, the of which these enzymes are members is first described to illustrate the functional diversity achieved with a shared and conserved structural scaffold. Then, the current knowledge of ASMase, SMPDL3B and SMPDL3A is summarized and information on specific topics is provided as context for the roles of these proteins. The subsequent three chapters present the results of my studies, as published.

1.1 The calcineurin-like phosphoesterases Sequence analysis reveals that certain proteins are evolutionarily related and allows classifying them into structural families. Similar molecules in different organisms descended from a common ancestor are called orthologs, whereas homologous within a single organism are paralogs. ASMase, SMPDL3A and SMPDL3B form a small family within the superfamily of calcineurin-like phosphoesterases (PFAM code PF00149). These three enzymes are paralogs, as they are all present in mammals including humans and are related to each other. In addition, at least 26 other members of PF00149 are found in mammals (Fig. 1.1), displaying lower but still identifiable . Although this superfamily also encompasses proteins from bacteria, these will be mentioned only briefly. The conserved structural features of PF00149 are first described, followed by a summary of the mammalian members grouped by functional class.

15 Figure 1.1 Mammalian members of the calcineurin-like phosphoesterase superfamily

A phylogram of the known mammalian enzymes part of the PF00149 superfamily is shown ( names), with the three proteins at the center of this thesis in blue. Relevant gene names are ASMase (SMPD1) and calcineurin (PPP3CA, PPP3CB, PPP3CC).

16 1.1.1 Shared structural features Sequence similarity usually entails structural homology, and the multiple crystal structures of proteins within PF00149 revealed a shared fold (Fig. 1.2). The phosphoesterase domain (as some of these enzymes contain additional domains) consists of a four-layer α-β-β-α sandwich. The highest conservation is observed at the active site, which consists of two metal ions coordinated by the side chains of seven residues (Fig. 1.3). These proteins catalyze hydrolysis of phosphoester or phosphodiester bonds; the phosphate group of their substrates invariably binds at the metal ions, which also activate a nearby water molecule to act as nucleophile [1,2]. After nucleophilic attack on the phosphorus atom, a leaving group dissociates from the latter, facilitated by protonation by an adjacent histidine residue. Whereas this mechanism is conserved in the superfamily, the identity of the phosphate-containing substrate varies widely. These can be categorized into several classes: serine- or threonine-phosphorylated proteins, nucleic acids, nucleotides, and sphingomyelin.

17 Figure 1.2 Structural superimposition of calcineurin-like phosphoesterases

The phosphoesterase domain of representative enzymes part of the PF00149 superfamily is superimposed to highlight the two conserved β-sheets (blue) and surrounding α-helices (red). Figure generated using the superimposition available in the CATH database [186].

18 Figure 1.3 Active site of calcineurin-like phosphoesterases

The active site of representative enzymes part of the PF00149 superfamily is superimposed to highlight the conserved metal ions and their coordinating residues. Figure generated using the superimposition available in the CATH database [186].

19 1.1.2 PPP-type phosphatases Protein phosphorylation regulates multiple cellular functions in eukaryotes. It is estimated that over half of human proteins can be phosphorylated [3], most commonly on serine but also on threonine and tyrosine residues. This modification is carried out by kinases, with 428 serine/threonine kinases found in humans [4]. Regulation by phosphorylation is reversible by the action of phosphatases. Interestingly, only about 40 serine/threonine phosphatases are responsible for this process [5], implying that each one must recognize multiple protein substrates. This group mainly consists of two classes: protein metal (PPM) and phosphoprotein phosphatase (PPP). The PPPs are members of the PF00149 structural superfamily; therefore protein dephosphorylation is one function performed by enzymes related to ASMase. The thirteen mammalian PPPs (Fig. 1.1) are: the PP1α (PPP1CA), PP1β (PPP1CB) and PP1γ (PPP1CC) isoforms; the PP2Aα (PPP2CA) and PP2Aβ (PPP2CB) isoforms; the PP2Bα (PPP3CA, CALNA, CNA), PP2Bβ (PPP3CB, CALNA2, CALNB, CNA2) and PP2Bγ (PPP3CC, CALNA3, CNA3) isoforms (calcineurin, for which the PF00149 superfamily is named); PP4C (PPP4C, PPX), PP5 (PPP5C), PP6 (PPP6C), PPEF1 (PPEF1, PPP7C) and PPEF2 (PPEF2). These are the catalytic subunits, which associate with a large number of regulatory subunits and other interacting proteins to form different holoenzymes. Substrate selectivity and specificity is partially achieved by this arrangement. For example, almost 200 such interactors have been identified for PP1 [6]. Numerous crystal structures of PPPs revealed a Y-shaped groove extending across the active site [7] that accommodates the phosphorylated substrate, with two conserved arginine side chains contacting the phosphate group. However, selectivity is conferred by interactions with other regions of the substrate protein, away from the active site [8].

1.1.3 Other serine/threonine phosphatases In addition to the PPP family, three other enzymes in PF00149 are likely serine/threonine phosphatases: the tartrate-resistant (TRAP) type 5 (ACP5) and the acid phosphatase type 7 (ACP7, PAPL), both members of the purple acid phosphatase subfamily which has mainly been characterized in plants, and CPPED1 (CPPED1, CSTP1). The owe their color to a Fe(III) – tyrosine coordinate bond [9]. The physiological substrates of TRAP are not fully defined, but the enzyme can dephosphorylate the bone matrix

20 proteins osteopontin and bone sialoprotein; this action reduces attachment and may regulate bone resorption [10]. Furthermore, TRAP is able to produce reactive oxygen species, and its high expression in macrophages lead to a proposed role in bacterial killing [11]. On the other hand, ACP7 is more similar to the plant purple acid phosphatases [12], but its functions and substrates are unknown. The other recently identified serine/threonine phosphatase, CSTP1, dephosphorylates Akt at a specific site, inhibits cell cycle progression and promotes apoptosis [13].

1.1.4 Nucleases DNA and RNA constitute a completely different class of substrates for two members of the PF00149 superfamily, DBR1 and MRE11. DBR1 is the enzyme responsible for intron debranching, a step necessary for their degradation or generation of various small RNAs [14]. During mRNA splicing, excised introns form branched lariat structures that contain a 2’,5’- phosphodiester connection in addition to the regular 3’,5’ linkages. Structures of the enzyme in complex with RNA ligands revealed a recognition mode with the 2’,5’ phosphate at the metal site, specific interactions with the branchpoint adenine and the first nucleotide of the 2’ arm, and electrostatic contacts with phosphate groups in the 3’ and 5’ arms [14]. The entire branchpoint is housed in a wide rectangular cavity on the protein. MRE11, participates in repair of DNA double-strand breaks. It is at the core of the MRN complex, along with RAD50 and NBS1 [15]. The MRN complex is responsible for sensing, signaling, and repairing breaks and also serves a role in telomere maintenance. In addition to the catalytic domain, MRE11 contains two DNA-binding domains and possesses endo- and activities against single- and double-stranded DNA [15]. Although crystal structures of the enzyme in complex with other MRN components or with DNA are available, its mechanism of action is not firmly established. One model proposes ATP hydrolysis-coupled DNA unwinding by Rad50 and entry of the single-stranded substrate into the MRE11 active site [16].

1.1.5 Nucleotidases Whereas multiple members of the PF00149 superfamily are able to hydrolyze various nucleotides in vitro, two among them are bona fide nucleotidases, meaning that these molecules

21 were demonstrated to be their physiological substrates: 5’- (NT5E, or CD73) and manganese-dependent ADP-ribose/CDP-alcohol diphosphatase (ADPRM). CD73 degrades extracellular AMP to adenosine; these and other nucleotides mediate purinergic signaling, a system that is described later in the introduction. The protein comprises two domains and exists in open and closed conformations [17]. The substrate-binding site is located between the two domains, with the nucleobase moiety bound to the C-terminal region, and the phosphate group, to the phosphoesterase domain metal ions [17]. Large inter-domain rotations occur to allow substrate binding, formation of the active site, and product release. In the cytoplasm, ADP-ribose and its cyclic counterpart are second messenger molecules involved in calcium regulation; ADPRM hydrolyzes these compounds [18]. Its substrate binding mode is unknown, although the active site is located at the bottom of a long, narrow U-shaped cavity [19].

1.1.6 Enzymes of other or unknown function Certain enzymes within the PF00149 superfamily have not been extensively characterized or were demonstrated to be inactive; or their substrates fall outside of the classes described above. For instance, the two related phosphoesterases MPPED1 and MPPED2 (or 239AB and 239FB) are expressed predominantly in the adult and fetal brain, respectively [20], but their specific roles are unknown. MPPED1 cleaves cyclic AMP (a second messenger molecule) in vitro, whereas MPPED2 contains a metal binding residue substitution that negatively affects catalysis [20]. The crystal structure of MPPED2 contained a naturally bound GMP nucleotide which also acts as inhibitor [21], and a structural or scaffold function has been proposed for this protein. Another PF00149 member devoid of enzymatic activity is the vacuolar protein sorting- associated protein 29 (VPS29), part of the retromer complex. Retromer mediates vesicular transport of transmembrane receptors from endosomes to the trans-Golgi network and cell membrane [22] and includes the VPS35-VPS29-VPS26 subcomplex at its core. Although VPS29 possesses an intact set of metal-binding residues, it lacks the conserved histidine side chain that would protonate the leaving group during catalysis [23]; this protein instead acts as a scaffold for retromer assembly.

22 The enzymes enumerated so far are all soluble with the exception of 5’-nucleotidase (NT5E, or CD73), which exists either in a soluble form, or attached to the cell membrane via a glycosylphosphatidylinositol (GPI) anchor [17]. Two transmembrane (TM) proteins also contain the calcineurin-like phosphoesterase domain: TM protein 62 (TMEM62) and TM protein with metallophosphoesterase domain (TMPPE). In the former, this module is connected to an N- terminal TM helix and a C-terminal five-pass TM domain, while in the latter, the phosphoesterase portion is attached to an N-terminal five-pass TM domain. Virtually nothing is known about these proteins. Lastly, one or two TM helices are also found in metallophosphoesterase 1 (MPPE1, PGAP5), which participates in GPI remodeling. Exit of GPI- anchored proteins from the endoplasmic reticulum requires removal of a mannose-linked phosphoethanolamine moiety from the anchor, a step catalyzed by MPPE1 [24]. The crystal structure of this enzyme has not yet been determined.

1.1.7 Bacterial enzymes The PF00149 superfamily extends to bacteria. In E. coli, members of known function include serine/threonine protein phosphatases and enzymes that hydrolyze nucleotides (3’,5’- cAMP to 5’-AMP; 2’,5’-cNMPs to 3’-NMPs; 5’-NMPs or 3’-NMPs to nucleosides; 5’,5’- diadenosine tetraphosphate to ADP; and UDP-sugars to UMP). In addition, the enzyme SbcD is part of the SbcCD complex which cleaves DNA hairpin structures to assist in replication [25].

1.1.8 Acid sphingomyelinase-like proteins Besides the functionalities listed so far – serine/threonine phosphatases, nucleases and nucleotidases – a group within the PF00149 superfamily is associated with . They are the acid sphingomyelinase-like proteins: ASMase, SMPDL3A and SMPDL3B. The ASMase gene is present from mammals down to unicellular eukaryotes, whereas its paralogs SMPDL3A and SMPDL3B are only found in multicellular organisms, in the kingdom Animalia. The latter two proteins likely originated from ASMase, although it is unclear which one of them appeared first. They display 39% identity to each other and about 28% to ASMase. They are single-domain enzymes, as opposed to ASMase which also contains a saposin module. Saposins are small lipid- and membrane-binding proteins and are discussed later. ASMase requires its

23 saposin domain to degrade the lipid sphingomyelin to ceramide. The roles of the paralogs SMPDL3A and SMPDL3B, however, are not yet clearly defined. SMPDL3B may also be involved in lipid metabolism, as it was found to affect cell membrane lipid composition and Toll- like receptor signaling in macrophages [26]. On the other hand, SMPDL3A was proposed to be an extracellular nucleotide with an anti-inflammatory function in macrophages [27].

1.2 Acid sphingomyelinase Cellular membranes consist of lipid bilayers with embedded proteins, and the specific lipid composition affects membrane protein function and impacts cellular processes including signaling. Here, one major class of membrane constituents – the sphingolipids – is introduced. The various mammalian sphingomyelinases are listed, and areas of the current knowledge of ASMase are summarized. The saposin proteins are also described, as well as the lysosomal lipid degradation process in which they take part.

1.2.1 Sphingolipids Two general types of lipids form cellular membranes: cholesterol and fatty acid- containing polar lipids. The latter are subdivided into glycerol-based phospholipids, and sphingolipids. In mammals, the most abundant phospholipid is phosphatidylcholine, comprising about 50% of the non-cholesterol lipids, whereas the major is sphingomyelin (SM) [28] (Fig. 1.4). These two species are chemically similar, as they share a phosphocholine head group and two acyl or alkyl chains. However, fatty acids in phospholipids tend to be more cis- unsaturated, which decreases packing tightness and results in a liquid-disordered membrane phase [28]. Conversely, sphingolipids contain longer, saturated (or trans-unsaturated) acyl chains and adopt the solid gel phase in isolation [28]. Furthermore, their backbone bears hydroxy and amide groups, enabling formation of extensive hydrogen bonds with each other and with cholesterol [29,30].

24 Figure 1.4 Chemical structures of representative lipid species

Sphingomyelin consists of a ceramide backbone and a phosphocholine group, which is also found in phosphatidylcholine.

25 Owing to their high abundance, phosphatidylcholine and SM are considered structural lipids [28]. In response to cellular and extracellular events, these molecules can be enzymatically converted into new species that serve direct signaling roles or affect membrane properties. For example, removal of the phosphocholine (PC) head group from SM by sphingomyelinases (SMases) produces ceramide (Fig. 1.4). The levels of this sphingolipid in resting cells is extremely low, but can increase rapidly or in a sustained manner in response to cellular stress or various stimuli [29]. Because of its small head group (hydroxyl) and resulting high hydrophobicity, ceramide displays different physicochemical properties relative to its parent SM. Mainly, this lipid tends to segregate into ceramide-enriched lipid domains separate from the bulk of the membrane due to its immiscibility with phospholipids [31]. In turn, properties of these domains, such as thickness and order, differ from other areas of the membrane, and favor partitioning and clustering of certain membrane proteins, thus impacting cellular processes including signaling [31,30]. These domains have been termed lipid rafts and are described in more details later.

1.2.2 Sphingomyelinases Ceramide is synthesized via multiple routes, including acylation of or dihydrosphingosine by ceramide synthases and deglycosylation of by glucosidase or galactosidase [32]. These pathways may be mostly relevant for sustained production or catabolism, whereas rapid generation of ceramide at the plasma membrane in response to stimuli is accomplished by the action of SMases [29,31]. Six SMase genes are present in mammals, and these proteins have been categorized according to their optimal activity pH: ASMase (SMPD1), the alkaline SMase (ENPP7), and the four neutral SMases – nSMase1 (SMPD2), nSMase2 (SMPD3), nSMase3 (SMPD4) and MA-nSMase (SMPD5) [33]. Although they catalyze the same reaction, these enzymes belong to different structural families, except for nSMase1, nSMase2 and MA-nSMase which are part of one superfamily of nucleases and phosphatases (PFAM code PF03372), along with bacterial SMases-C [34]. The neutral SMases are located intracellularly. nSMase2 is found at the inner leaflet of the plasma membrane, or can translocate there from the Golgi subsequent to various stresses and stimuli [33]. The protein contains an N-terminal domain with two hydrophobic segments, which serves as membrane anchor and allosteric activator of the catalytic domain in a manner involving

26 anionic phospholipids [35]. The enzyme is also palmitoylated – a modification that could assist its proper orientation at the membrane [35]. nSMase1 is localized in the Golgi and endoplasmic reticulum, as well as in the nuclear matrix [36], and contains two TM segments [33]. Phosphorylation also plays a role in regulating nSMase2 and nSMase1 activity [35,36]. Meanwhile, MA-nSMase is a mitochondrial enzyme [37,38] with a TM helix, but has not been extensively characterized. Lastly, nSMase3 shares little homology with its counterparts [39], and the function of this endoplasmic reticulum protein is not clear. SM is also a dietary lipid. Its absorption requires initial degradation to ceramide, a process carried out by the alkaline SMase (alk-SMase, ENPP7) [40]. alk-SMase is part of a structural family of extracellular nucleotidases and [41]. It is expressed in the small intestine, attached to the outer surface of microvilli by a C-terminal TM segment – although intestinal trypsin can release alk-SMase into the lumen [40]. The enzyme acts on dietary SM solubilized in bile salt micelles, and a hydrophobic loop adjacent to the active site facilitates access to the substrate [42]. The best characterized of the mammalian SMases is the lysosomal acid sphingomyelinase. First identified over fifty years ago [43], purified forty years ago [44] and cloned almost thirty years ago [45], the enzyme was associated early on with the genetic disease Niemann-Pick (NP) types A and B [46] where sphingomyelin accumulates in organs. In the last twenty years however, interest in ASMase has shifted to its role in signaling, as the enzyme can translocate to the surface of the cell [47] in response to various stimuli and modulate cellular events such as apoptosis [48]. Here, NP disease is briefly described, and lysosomal and other functions of ASMase are listed, as well as its molecular properties.

1.2.3 Niemann-Pick disease types A and B Lysosomal storage diseases are rare genetic recessive disorders caused by mutations in one of about seventy lysosomal that act on sphingolipids, glycoproteins, or polysaccharides, or transport proteins. They result in accumulation of un-degraded molecules, causing a variety of phenotypic defects. Inactivating mutations in the ASMase gene (SMPD1), of which over 185 have been identified so far [49], give rise to Niemann-Pick (NP) disease types A and B. The former is more severe, characterized by hepatosplenomegaly (enlarged liver and spleen) and neurological problems, and fatal within the first three years of life [49]. On the other

27 hand, NP type B (NPB) is not neurodegenerative but causes organ dysfunction, especially of the liver, spleen, lungs and heart [50]. NPB patients often survive into adulthood, but respiratory and liver failure remain a common cause of morbidity and mortality [50]. Foam cells (lipid-laden macrophages) infiltrate into multiple tissues, and inflammation is the primary cause of respiratory problems [50]. Residual ASMase activity (less than 10% versus close to 0%) determines the type and severity of NP disease [51,52]. Although SM is the primary accumulating lipid in NPA/B, cholesterol and are also elevated [53], as are serum LDL cholesterol and triglycerides [54]. The SM imbalance thus impacts metabolism of other lipid classes. Several lysosomal storage disorders are treated by enzyme replacement therapy (ERT), whereby the deficient protein is administered by injection and taken up from the circulation into via mannose or mannose-6-phosphate receptors [55]. For instance, Gaucher disease can be managed by recombinant acid beta-glucosidase [56], whereas Fabry disease is treated with recombinant alpha-galactosidase A [57]. Human ASMase is currently being developed for ERT (drug name olipudase alfa) [50]. However, although providing the protein could improve organ dysfunction and SM storage in peripheral tissue, it is unlikely to remediate neurological problems in NPA [50], since most enzymes used in ERT are unable to cross the blood-brain barrier efficiently [58]. Other therapeutic avenues for lysosomal storage diseases include inhibition of substrate formation [59], stabilization of the defective enzyme by small molecule chemical chaperones [60], and gene therapy [61].

1.2.4 Lysosomal roles of ASMase Lysosomes are organelles where proteins, lipids and other macromolecules are degraded and recycled. They originate by endocytosis, phagocytosis or autophagy and undergo maturation; their luminal pH gradually decreases to 4.5 [62] and hydrolytic enzymes are delivered from the Golgi. Lysosomal hydrolases, including ASMase, are targeted for transport by phosphorylation of mannose residues on their asparagine-linked glycans [63]. Metabolism of membrane components does not occur on the “limiting” (surrounding) lysosomal membrane, which is protected from enzymes by a glycocalyx layer [64] composed of glycoproteins. Instead, intra- lysosomal vesicles, originating from the plasma membrane or other fragments to be digested, are the sites of lipid degradation [62]. The lipid composition of these vesicles evolves during

28 lysosomal maturation: glycosphingolipids and SM are hydrolyzed to ceramide, cholesterol is exported out of the lysosome by the NPC1 and NPC2 proteins [65], and the anionic lipid bis(monoacylglycero)phosphate (BMP) is generated [62]. This metabolite is important for certain lysosomal hydrolases; its presence stimulates ASMase activity [66] by favoring binding of the enzyme, which is cationic [67], to anionic vesicles [66,68]. The contribution of ASMase to lysosomal function can be studied using cationic amphiphilic drugs (CADs). These compounds are hydrophobic and weakly basic, enter lysosomes, become protonated and trapped [69]. High CAD concentrations dissociate the enzyme from anionic vesicles [68], and leads to its degradation by lysosomal proteases [70]. One effect of impaired ASMase activity is defective export of cholesterol, as evidenced by the secondary accumulation of this lipid in NPA/B [53]. In membranes, SM interacts favorably with cholesterol, whereas ceramide competes with the latter for association with SM and displaces it [71]. Failure to metabolize SM impairs cholesterol removal [67], which is required to decrease membrane rigidity [62]. Cholesterol depletion is necessary for degradation of glycosphingolipids by hydrolases assisted by saposin proteins [72]. Highlighting the housekeeping role of lysosomal ASMase, this enzyme was also reported to contribute to lysosome stability [73,74]. Cells lacking the protein are extremely sensitive to photo-oxidation-induced lysosomal damage, and it was concluded that the stabilizing effects of the Hsp70 chaperone on lysosomes were mediated by its interaction with the lipid BMP and ASMase [73]. In addition, cancer cells in general display enlarged lysosomes and reduced expression of SMases [75]. Targeting of tumor cells with CADs results in ASMase degradation and elevated SM levels, which disrupt membrane homeostasis and cause lysosomal membrane permeabilization, thus preferentially killing these cells [74].

1.2.5 Saposins In order to metabolize lipids, hydrolases must be able to access their membrane- embedded substrates. Some of these enzymes are integral membrane proteins, whereas others possess membrane-binding domains or features. In the lysosome, enzymatic degradation of SM, ceramide, and glycosphingolipids bearing less than four sugar residues [62] is assisted by several protein cofactors – the saposins. They are small lipid- and membrane-binding proteins, and act in one of two general ways: the “solubilizer” method consists of extraction and formation of a

29 soluble saposin-lipid complex followed by substrate presentation to a hydrolase, whereas the “liftase” mode involves facilitation of substrate access to the enzyme at the membrane surface by the disruption or remodeling action of the saposin [76]. Mammals contain four saposin polypeptides, A though D, products of a single gene (PSAP), as well as another of similar function, the GM2 activator (GM2A). SapA – D originate from proteolytic cleavage of the precursor in the lysosome [62] and are structurally homologous, consisting of four or five α-helices stabilized by three disulfide bonds. Multiple crystal and NMR structures reveal high conformational flexibility in their fold. Depending on the environment (pH, presence of detergents or lipids), these proteins adopt either an open conformation which exposes a large hydrophobic surface, or a closed form; half-open states have also been observed (Fig. 1.5). In these structures, hydrophobicity of the wide open conformation invariably entails dimerization of the protein, occasionally with a layer of detergent in the middle [77,78]. In contrast, the half-open saposins are sometimes bound to individual lipid molecules [79], as is the case with the structure of GM2A [80].

30 Figure 1.5 Structural flexibility of saposins

Saposins can adopt an open (left), half-open (middle) or closed conformation (right). AOAH and ASMase refer to the saposin domains of and acid sphingomyelinase. Structures shown are: open SapA (PDB code 4DDJ), SapC (2QYP), SapD (unpublished data) and ASMase (this thesis); half-open AOAH (unpublished data), SapB (1N69) and SapC (1SN6); closed SapA (2DOB), SapC (1M12), SapD (2R0R) and ASMase (this thesis).

31 These structures can be correlated with the modes of action of individual saposins. GM2A assists β- A in processing of the gangliosides (glycosphingolipids) GM2 and GA2 [81], and SapB enables hydrolysis of sulfates by A, of globotriaosylceramide and digalactosylceramide by α-galactosidase A, and of other [82]. Both cofactors display the half-open conformation and act as “solubilizer” by delivering individual lipids to their respective enzymes [62,83]. SapA forms particles consisting of a protein dimer sandwiching multiple lipid molecules, suggested to be the relevant form of galactosylceramide presentation to galactocerebrosidase (GALC) [78]. On the other hand, SapC functions as “liftase”, allowing degradation of glucosylceramide by (GBA) [84]. This saposin locally alters regions of bilayer for subsequent attack by GBA [76]. Lastly, SapD facilitates (but is not necessary for) cleavage of ceramide to sphingosine by acid [85]; however, its mode of action is unclear. Unlike all the other lysosomal hydrolases involved in sphingolipid metabolism, ASMase possesses a built-in saposin domain [86], linked to the C-terminal phosphoesterase domain by a proline-rich segment [87]. This module allows SM cleavage in absence of external saposins [87], although SapA and especially SapC enhance its hydrolysis rate [88,87]. The internal saposin domain is necessary for ASMase function [89], although the precise manner in which it enables activity is unknown. SapC forms a relatively stable dimer with its cognate enzyme GBA [84], possibly analogous to the ASMase inter-domain interaction, but neither complex has been structurally characterized.

1.2.6 Roles of ASMase at the cell surface Due to its association with NP disease, initial interest in ASMase was focused on its lysosomal function. Nevertheless, the enzyme had also been detected in the extracellular space, including in urine [90] and serum [91], and was found to be secreted by multiple cell types in culture [92]. In eukaryotes, proteins destined towards the secretory pathway, including lysosomal enzymes, contain an N-terminal hydrophobic signal peptide that targets the nascent protein for translocation into the endoplasmic reticulum during translation. After glycosylation and folding, many proteins then continue through the default secretory pathway. However, lysosomal hydrolases are recognized in the Golgi apparatus by the sensor UDP-N-acetylglucosamine 1- phosphotransferase and phosphorylated on mannose residues for delivery to lysosomes via

32 mannose-6-phosphate (M6P) receptors [93]. ASMase undergoes mannose phosphorylation [63], but a fraction of the enzyme is also secreted; this could be due to its exceptionally long signal peptide [94]. The enzyme may additionally be transported to the cell surface in secretory lysosomes [95] – specialized organelles found in certain cell types [96]. In either case, the protein ends up on the outer leaflet of the cell membrane, as opposed to the neutral SMase nSMase2 which is located on the inner leaflet. Despite its name, the acid SMase also has appreciable activity at neutral pH; whereas its affinity for SM (Km) is optimal at pH 4.6, its maximal velocity (Vmax) is pH-independent [97]. Differences in lipid composition between the plasma membrane and intra-lysosomal vesicles likely affect its activity rate as well [98]. Lysosomal and secreted forms of the enzyme are distinguished by their type of asparagine-linked glycans (high-mannose versus complex) [99] and proteolytic processing. In addition to signal peptide cleavage of the first 48 residues [100], a pro-peptide of variable length is removed, resulting in His62 as the N-terminal residue for the secreted version or Gly68 for the lysosomal form [99]. However, Gly85 was reported as the N-terminus of lysosomal ASMase from placenta [101]. Furthermore, lysosomal cleavage of a C-terminal segment (last nine amino acids) [102] is required to relieve the inhibition imparted by the C-terminal cysteine residue, which is hypothesized to block the active site [103]. In contrast to the housekeeping role of ASMase in lysosomes, it participates in signal transduction at the cell surface. In response to various stresses and stimuli, the enzyme translocates to the cell membrane [104]. This may be carried out through phosphorylation of ASMase by protein kinase Cδ on Ser510 [105]. The manner in which these proteins come in contact, and the subcellular location involved, are still unclear. Ceramide formation on the plasma membrane outer face affects receptor-mediated signaling. As previously mentioned, the cell membrane contains sphingolipid- and cholesterol-rich, ordered rafts [30]; the size and lifetime of these domains varies with lipid composition. Certain proteins, including glycosylphosphatidylinositol (GPI)-anchored or palmitoylated ones, or receptors with relatively long transmembrane domains, preferentially localize to rafts [30]. Raft localization can affect signaling by excluding or including specific effectors in the membrane domain, inducing conformational changes in proteins, or increasing the local receptor concentration, thus facilitating oligomerization [30]. The resting level of ceramide in the cell membrane is extremely low [29]; presence of this lipid modifies raft properties and, notably, enhances receptor

33 clustering [106]. Formation of ceramide-rich domains was observed following a variety of receptor and non-receptor stimuli, so this appears to be a general mechanism of signaling modulation [106]. However, it has especially been associated with cell death [52]. ASMase- deficient cells are resistant to radiation-induced apoptosis [48], and inhibiting this enzyme in conditions of excessive undesired apoptosis may be beneficial. For example, certain antidepressant drugs happen to be CADs – they lead to dissociation and degradation of lysosomal ASMase [68]. Depression has been proposed to involve an imbalance between neurogenesis and cell death in the hippocampus [107]. Blockage of ASMase-generated ceramide improves survival of neurons and remediates depression symptoms in mouse models, positing this as one mode of action of these compounds [107].

1.2.7 ASMase in the circulation In addition to its functions at the plasma membrane, secreted ASMase displays activity on lipoproteins in the circulation. Low density lipoproteins (LDL) consist of an apoB-100 protein molecule wrapped around a particle containing about 3000 lipid molecules and covered in a monolayer of phosphatidylcholine, SM and cholesterol [108] – similar to the cell membrane outer face. LDL from the circulation can become trapped below the endothelial layer of arteries; this is the initial step of . ASMase has been implicated in retention and aggregation of LDL, and atherogenesis [109]. The enzyme is secreted in large amounts by vascular endothelial cells, increasing in response to inflammatory cytokines [110]. It can then hydrolyze local LDL SM in the acidic environment of the atherosclerotic lesion [98]. ASMase is also able to act on LDL that was oxidized or modified by A2 – events associated with atherogenesis [98]. This action alters LDL properties and promotes further LDL aggregation [111], notably by facilitating its ingestion by macrophages that attempt to remove the particle [112]. ASMase-deficient mice thus display reduced atherogenesis [113].

1.3 SMPDL3B As mentioned earlier, ASMase is part of the calcineurin-like phosphoesterases structural superfamily, and forms a small family along with the acid sphingomyelinase-like phosphodiesterases 3A and 3B (SMPDL3A, SMPDL3B). In contrast to the abundance of scientific literature covering the properties and functions of this enzyme, information on its two

34 paralogs is only beginning to emerge. Here, the current knowledge on SMPDL3B is summarized. This GPI-anchored [114,26] enzyme is found on the cell surface [26,115] and has been implicated in two processes thus far: regulation of innate immune responses in macrophages, and maintenance of podocyte integrity in the kidney. These two roles appear to involve membrane- modulating activities of SMPDL3B.

1.3.1 Lipid rafts in toll-like receptor signaling Toll-like receptors (TLRs) are central to innate immunity. These proteins recognize characteristic bacterial, viral or other foreign molecules and initiate signaling cascades that result in production of pro-inflammatory factors. TLR1, 2, 5, 6 and 10 function at the plasma membrane, whereas TLR3, 7, 8 and 9 signal from endosomes; TLR4 acts at both locations [116]. These single-pass transmembrane receptors bind their ligands via their ectodomain, inducing dimerization (or dimer rearrangement) necessary for signal transmission into the cytoplasm [116]. Membrane composition and lipid rafts modulate TLR function, as exemplified by the lipopolysaccharide (LPS) receptor, TLR4. Its activation requires association with accessory factors in rafts [117]. LPS is first bound by the co-receptor CD14, a GPI-anchored protein localized in cholesterol- and sphingolipid-rich rafts in the outer leaflet [118] due to the affinity of the GPI moiety for this environment [30]. TLR4 itself is not initially present in rafts, but is recruited there after LPS stimulation [119,120], a process required for efficient dimerization [121]. Other TLRs undergo such a translocation as well [119]. Depletion of cholesterol decreases inflammatory signaling downstream of TLRs [119]. Sphingomyelin is also necessary for recruitment of TLR4, as deficiency of SM synthases disrupts this response [122,123].

1.3.2 ASMase in TLR signaling Stimulation of macrophages with LPS results in rapid accumulation of ASMase in plasma membrane rafts [124] and generation of ceramide [125,120]. However, the effects of this on downstream events, such as tumor necrosis factor α (TNFα) production, are ambiguous. On one hand, the ASMase inhibitor SMA-7 suppresses ASMase activation, ceramide production and inflammatory cytokine release in stimulated macrophages [126]. Pre-treatment of macrophages with the CAD imipramine – which causes lysosomal ASMase degradation – decreases ASMase

35 activation, ceramide production, raft clustering of TLR4 and TNFα production in response to LPS [120]. ASMase action alters the plasma membrane biophysical properties [127] and redistributes rafts [120]; they merge into larger, more stable platforms and facilitate certain receptor interactions – possibly including that of CD14 with TLR4 [117]. Addition of exogenous short chain (C2) ceramide reverts all the effects of ASMase inhibition in the context of LPS stimulation [127]. In fact, treatment of TLR4-expressing cells with bacterial SMase or exogenous short chain alone can trigger associated downstream signaling [128]. It should be noted that addition of exogenous short chain ceramides does not necessarily mimic ASMase action at the cell membrane, as this enzyme natively generates longer chain products with different properties [117]. On the other hand, ceramide generation by ASMase also has negative regulatory roles in TLR signaling. Exogenous C8-ceramide reduces TNFα production in LPS-stimulated macrophages [129]. Liposomal administration of C6-ceramide decreases bacteria-induced inflammation [130]. Further evidence for an inhibitory role of this lipid in these processes was reported: inhibition or silencing of ASMase increases TNFα release, whereas treatment of cells with exogenous C8 ceramide or bacterial SMase has the opposite effect [131]. ASMase-deficient mice display a normal increase in serum TNFα in response to LPS [132]. ASMase-deficient macrophages respond with increased amounts of TNFα; altered post-translational processing of this cytokine could explain some of these observations [133]. ASMase could therefore participate both in initiation and downregulation of TLR signaling. On the plasma membrane inner leaflet, nSMase1 is also activated by LPS stimulation, with a potential negative regulatory role [131]. Other lipid-modifying enzymes at the cell surface could therefore modulate innate immune responses given the importance of lipid composition and rafts. The following section summarizes a study [26] reporting on the ASMase-like protein SMPDL3B in this context.

1.3.3 SMPDL3B in TLR signaling Lipid rafts were characterized early on as detergent-resistant membrane fractions. This property allows isolation of proteins potentially present together in the same membrane compartment by differential detergent solubility. SMPDL3B was thus identified by co- purification with endosomal TLRs 3, 7, 8 and 9 from murine macrophages. This was confirmed

36 by co-immunoprecipitation with TLRs 4 (which can also signal from endosomes), 7, 8 and 9. This finding does not necessarily implicate direct inter-protein interactions, as these proteins could simply be located in the same detergent-resistant membrane fractions. As mentioned earlier, TLRs localize to lipid rafts upon ligand stimulation, whereas other associated factors including CD14 are constitutively present there due to their GPI anchor. SMPDL3B was enriched in raft fractions along with CD14 from macrophages, and was detected on the surface of transfected cells. Treatment with a GPI-cleaving enzyme and deletion of a C-terminal peptide containing the putative anchor attachment site revealed that SMPDL3B is also GPI-anchored. The protein is expressed in macrophages and dendritic cells; its transcription is further induced upon TLR stimulation, and protein levels are elevated in response to ligands of different TLRs and to interferon γ. A negative regulatory role in TLR signaling was reported for SMPDL3B. Its depletion from macrophages upregulates NF-κB activation and interleukin-6 (IL- 6) release after TLR stimulation by various ligands. SMPDL3B-deficient mice react to intraperitoneal LPS or bacteria with an increased immune cell influx, serum levels of IL-6, TNF and other cytokines, indicating an aberrantly strong immune response. As this protein is related to ASMase, its anti-inflammatory effects could be carried out through action on membrane lipids. Although SMPDL3B hydrolyzes a generic phosphodiesterase small molecule substrate (most efficiently at neutral pH), and is a major source of phosphodiesterase activity on macrophages, the study was unable to demonstrate activity on SM in vitro. Nevertheless, SMPDL3B-deficient intact macrophages display a strong decrease of cell membrane order (increased fluidity). Analysis of total cellular lipids – which is not ideal given the protein’s plasma membrane localization – revealed changes in most lipid classes. C16:0, C22:0 and C24:1 ceramide levels are reduced, two SM species are increased while five are decreased, and cholesterol is lowered by about 35%. In terms of glycerol-based phospholipids, phosphatidylcholine species are overall elevated, whereas most phosphatidylethanolamine, -inositol and -serine are decreased. These last three classes are mainly found in the plasma membrane’s inner leaflet [134], suggesting that if SMPDL3B is indeed a lipid-modifying enzyme, its action indirectly affects non-substrate lipids as well, as observed in certain lysosomal storage diseases [53]. These results were compared to a related lipidomics study [135] and were indicative of a pro-inflammatory lipid state. Pre-treatment of SMPDL3B-deficient macrophages with the “missing” C16:0, C22:0 and C24:1 ceramides –

37 which are predicted to dampen TLR-dependent IL-6 release [135] – reverts the post-LPS- stimulation IL-6 release down to wild-type levels. As SMPDL3B deficiency lowers overall plasma membrane order, the integrity of rafts was examined. Total surface concentration of the ganglioside GM1, a sphingolipid normally enriched in rafts [136], is unaffected. However, this approach would not be able to detect differences such as distribution or local clustering of this marker. Lastly, since the enzyme is associated with the endosomal TLRs, and because exofacial ceramide can promote inward membrane curvature and endocytic vesicle budding [137], the integrity of endocytic functions was evaluated. Surface expression and endocytosis of TLR4 upon LPS stimulation is unaffected by SMPDL3B deficiency. In summary, this enzyme is a negative regulator of TLR signaling, but the specific molecular mechanisms involved remain to be clarified.

1.3.4 Lipid rafts in podocytes In addition to its role in macrophages, SMPDL3B has been identified in podocytes – specialized epithelial kidney cells that filter blood plasma. They constitute the last layer of the glomerular filtration barrier and contain extensions called foot processes interlinked by slit diaphragms [138]. The slit diaphragm is a 40 nm gap bridged by overlapping protein molecules (nephrin and others), preventing plasma protein loss into the urine. This structure is attached to the cell membrane via podocin, CD2AP and other factors, which also anchor the complex to the actin cytoskeleton [138,139]; this is crucial for foot process stability. Podocyte stress or injury can result in slit diaphragm loss and proteinuria [139]. The base of the slit diaphragm is located in lipid rafts [140], mediated by podocin [141]. Proper lipid composition, including cholesterol, is required for the integrity of these structures [142]. Accumulation of various sphingolipids in the kidney in models of nephropathy and in several genetic lipid storage disorders is associated with glomerular disease [143]. For instance, accumulation of the ganglioside Gb3 in Fabry disease causes podocyte injury [144]. The ganglioside Gm3, on the other hand, is found in lipid rafts of the slit diaphragm and is essential for autocrine preservation of the podocyte actin cytoskeleton [145]; its levels are increased or decreased in kidney disorders [142].

1.3.5 SMPDL3B in podocyte function

38 The participation of SMPDL3B in podocyte function was discovered via an off-target effect of the therapeutic antibody rituximab. This monoclonal anti-CD20 antibody targets B-cells in certain cancers and autoimmune diseases. It was found that rituximab additionally recognizes a peptide present in SMPDL3B [146]. The antibody has been used to treat nephrological conditions, where it was hypothesized to act in a B-cell-independent manner possibly involving this enzyme [115]. Focal segmental glomerulosclerosis (FSGS) is one such disorder; it is characterized by damage to the glomerulus and proteinuria, and can reoccur after kidney transplantation. A study investigated the action of rituximab in this setting [115]. SMPDL3B is expressed in human podocytes and localized on the cell membrane in flotillin-1-positive lipid raft fractions. Rituximab binds to the enzyme, and co-precipitates a GFP-tagged transfected form of it. This was prevented by a SMPDL3-mimicking short peptide, suggesting a direct interaction between the two proteins. In patients who experienced post-transplant recurrence of FSGS, SMPDL3B protein levels were decreased relative to the non-recurring group. Serum from recurring patients uniquely induced downregulation of SMPDL3B protein and mRNA in cultured podocytes. Rituximab treatment at the time of transplant reduced FSGS recurrence rate, and prevented loss of SMPDL3B protein and mRNA in culture. It was proposed that antibody binding to the enzyme might avert its degradation in FSGS. This could be mediated via alteration of endocytosis, but the effect on mRNA levels suggests a more complex mechanism. Recurring FSGS serum also decreased total cellular ASMase protein amounts and activity – changes preventable by rituximab. However, only the levels of the enzyme’s 52 and 54 kDa isoforms were altered, which do not correspond to the full-length protein (70 kDa). ASMase was reported to undergo cleavage by caspase-7 to a 57kDa form in internalized TNF receptosomes in Jurkat cells, although this processing actually increased its activity [147]. The putative rituximab epitope on SMPDL3B is only partially conserved in ASMase. The involvement of this enzyme in this setting is thus unclear. On the other hand, SMPDL3B may contribute to actin cytoskeleton stability in podocytes, as its overexpression, or preservation by rituximab treatment, prevents loss of stress fibers triggered by recurrent FSGS serum. Interestingly, SMPDL3B knockdown itself does not cause a phenotype in podocytes, but predisposes them to a more profound serum-induced disruption of actin fibers relative to wild-type. Rituximab cannot prevent this disruption in

39 knockdown cells, confirming the involvement of the enzyme in these processes. In summary, serum factors likely initiate actin remodeling and SMPDL3B downregulation, resulting in proteinuria, whereas increased SMPDL3B levels protect podocytes from these disruptions. The mechanisms and causative links between these events remain to be determined. A subsequent study confirmed SMPDL3B expression in pig glomeruli, interaction with rituximab and protective effect of the antibody in xenotransplants: rituximab prevents SMPDL3B downregulation, podocyte morphological disruption and decreased proliferation caused by anti-pig antibodies [148]. Similar results were obtained with rituximab in an adriamycin-induced rat model of nephrotic syndrome, with the additional observation that the antibody increases SMPDL3B mRNA in normal rats [149]. In radiation nephropathy, podocytes undergo cytoskeletal alterations, reduction in SMPDL3B protein level (but not mRNA), elevation of total ceramide and reduction of neutral ceramidase activity, total sphingosine and sphingosine-1-phosphate [150]. Again, rituximab treatment or SMPDL3B overexpression prevents these changes and protects podocytes. Interestingly, overexpression does not affect baseline levels of ceramide or sphingosine, but increases sphingosine-1-phosphate. These lipid alterations were not proposed to directly result from SMPDL3B action. A limitation of the study is the analysis of total cellular lipids as opposed to the plasma membrane where this enzyme is located. Lastly, the role of SMPDL3B has been investigated in diabetic kidney disease (DKD), also characterized by podocyte injury [151]. The enzyme’s expression is increased in DKD; it was demonstrated to interact with the urokinase-type plasminogen activator receptor (uPAR) or its soluble form (suPAR). However, this result may not indicate direct binding, but merely co- localization in lipid rafts, as uPAR is also GPI-anchored [152]. Elevated SMPDL3 was proposed to sequester suPAR from its binding partner integrin β3, ultimately resulting in susceptibility to apoptosis. Knockdown of the enzyme protects podocytes from DKD serum-induced apoptosis.

1.3.6 SMPDL3B outside of the cell In addition to the described roles of this enzyme in macrophages and podocytes, SMPDL3B may have functions in the extracellular space. It was detected as one of 300 proteins in the milk proteome [153], and is present in low abundance in the cerebrospinal fluid along with about 700 other proteins [154]. It is unknown whether these consist of the GPI-anchored enzyme

40 or a potential soluble form. Furthermore, SMPDL3B is found in pancreatic zymogen granules– special secretory organelles – along with about 400 other proteins [155], and is one of the 200 proteins in exosomes [156]. None of the known SMases and were identified in these studies. SMPDL3B could participate in the biogenesis of these vesicles, given its putative membrane-modifying functions.

1.4 SMPDL3A The third member of the ASMase-like proteins family is the acid sphingomyelinase-like phosphodiesterase 3A (SMPDL3A). Little is known about its functions. As its paralogs, it contains an N-terminal signal peptide which targets nascent polypeptides to the secretory pathway. Lack of membrane-interacting features such as saposin domain or GPI anchor results in its release into the extracellular space. The enzyme is found in the circulation [157], bound to serum apolipoproteins [158,159,160]. It is secreted by [161], adipocytes [162,163] and astrocytes [164,165,166]. However, the mannose-6-phosphate modification, which targets proteins to the lysosome, has been detected on SMPDL3A [161,167,168,169,170,171,172,173], suggesting additional lysosomal roles. Recently, this enzyme has been implicated in the response to cholesterol in macrophages. Unexpectedly, it does not appear to act on lipids, but hydrolyzes extracellular nucleotides, thus associating it to purinergic signaling.

1.4.1 SMPDL3A in the cholesterol response A study identified SMPDL3A as a gene induced by liver X receptors (LXRs) [174]. LXRs are transcription factors directly activated by oxidized cholesterol derivatives and serve to protect cells from cholesterol overload [175], with further functions in lipogenesis and glucose homeostasis [176]. In peripheral tissue including macrophages, LXR targets promote cholesterol efflux and other responses. In atherosclerosis, cholesterol-laden macrophages initially aggregate in the artery wall; LXR agonists inhibit development of these legions whereas macrophages lacking LXRs accumulate cholesteryl esters and rapidly become foam cells [176]. These transcription factors also play a role in suppression of inflammatory responses by preserving co- repressors on the promoters of NF-κB target genes [177]. In this context, SMPDL3A was identified as one of 18 targets upregulated (five-fold) by synthetic LXR agonists in human macrophages [174], but not in mice. Expression of all other

41 SMases or SMPDL3B is unaltered. Absolute levels of SMPDL3A transcript are highest in the kidney and colon, followed by placenta, lung and liver. However, LXRs do not induce this gene in these other tissues, suggesting a macrophage-specific role in the response to cholesterol. A subsequent study expanded these findings [27]. SMPDL3A mRNA, intracellular and secreted protein levels are increased by loading of human macrophages with acetylated LDL or cholesterol-cyclodextrin, or an LXR ligand. Intracellular cAMP also upregulates this gene. ASMase and SMPDL3B are practically unaffected by these treatments. In vitro, SMPDL3A catalyzes cleavage of the phosphodiester compounds bis(4-nitrophenyl) phosphate (bNPP), p- nitrophenyl phosphorylcholine (pNP-PC) – with the same head group as SM – and p-nitrophenyl thymidine 5'-monophosphate (pNP-TMP), but had no monophosphoesterase (phosphatase) activity. The optimal pH for catalysis is between 4 and 6. Interestingly, the enzyme is unable to cleave SM or phosphatidylcholine in Triton X-100 micelles. A lipidomics analysis of 300 individual species revealed no significant differences upon SMPDL3A overexpression in cells. On the other hand, the protein hydrolyzes the terminal phosphate of ATP, CTP, GTP and UTP with equal affinity, as well as the terminal phosphate of their NDP counterparts at a lower rate. No activity was detected against AMP, 3’,5’-cAMP or 3’,5’-cGMP. These results are comparable to the recently identified ASMase-like enzyme from the bacterium Ralstonia solanacearum (24% identity) that converts ATP and ADP to AMP, but is inactive on SM [178]. SMPDL3A appears to be the only nucleotide phosphodiesterase secreted from macrophages under basal conditions, although other such enzymes are present on the surface of these cells. Its expression is increased in atherosclerotic lesions in vivo [179], and this enzyme was thus proposed to participate in a macrophage response to modulate inflammation via extracellular nucleotides as part of purinergic signaling.

1.4.2 Purinergic signaling Most mammalian cell types release ATP in response to certain stimuli, and almost all cells possess surface receptors activated by extracellular nucleotides [180]; this is the basis for purinergic signaling, a form of auto- and paracrine inter-cellular communication mediating various processes. Neurons store and release ATP in synaptic vesicles for neurotransmission [181]. Other cell types rely on vesicles or channels such as pannexin 1 and others [182]. ATP can also escape from damaged cells [183]. Besides ATP, other nucleotides including ADP, UTP,

42 UDP, UDP-glucose, dinucleotide polyphosphates and adenosine participate in purinergic signaling [41], either directly released from cells, or as secondary extracellular metabolites. In humans, nineteen transmembrane receptors for these compounds are present. P2X1-7 are ion channels activated by ATP. P2Y1,2,4,6,11-14 are G protein-coupled receptors with partially overlapping specificities for ATP, ADP, UTP, UDP or UDP-glucose. P1 receptors A1, A2a, A2b and A3 are also G protein-coupled receptors and recognize adenosine [180]. P2-type receptors elicit pro-inflammatory immune responses – in macrophages, chemotaxis and activation of NF- κB – whereas adenosine is generally anti-inflammatory [183]. Multiple extracellular enzymes metabolize secreted nucleotides in order to modulate or terminate purinergic signaling. They are members of four structural families: the ecto-nucleoside triphosphate diphosphohydrolases (E-NTPDs), ecto-nucleotide pyrophosphatases / phosphodiesterases (E-NPPs), alkaline phosphatases, and the lone ecto-5′-nucleotidase (CD73) [41]. NTPDs hydrolyze NTPs or NDPs to NMPs as terminal product. NTPD1,2,3,8 are transmembrane proteins on the cell surface, whereas NTPD4,7 are found in organelles; transmembrane organellar or soluble forms of NTPD5,6 exist [184]. For instances, macrophages express NTPD1 (CD39) on the plasma membrane [185]. Key differences between this enzyme and SMDPL3A include topology (membrane-bound versus secreted) and substrate preference (ADP versus ATP) [27]. Another family, the NPPs, contains both nucleotidases and proteins with unrelated functions. The former class consists of NPP1,3,4 [41], which are single-pass transmembrane proteins – although secreted forms of NPP1,3 occur – and generate NMPs from various substrates including NTPs, NDPs, dinucleotide polyphosphates and UDP-glucose [41]. Finally, the NMP products of all these enzymes are further metabolized to adenosine by the GPI- anchored proteins ecto-5′-nucleotidase and the four alkaline phosphatases which have broad substrate specificity [41]. SMDPDL3A could thus be a new player modulating purinergic signaling.

1.5 Conclusion The acid sphingomyelinase family of proteins (ASMase, SMPDL3A and SMPDL3B) is part of the calcineurin-like phosphoesterases superfamily, which encompasses diverse functions. One common element to these enzymes is recognition and cleavage of a phosphoester bond in their substrates. Phosphate groups are found in a wide range of small molecules and

43 macromolecules however, and for the family members whose role is unknown or unclear, it cannot be guessed based on primary sequence; besides the PPP-type phosphatases, sequence conservation in this superfamily is low (Fig. 1.1). Residues that form the active site and segments that maintain the structure or fold of the proteins are preserved, whereas elements that participate in substrate recognition vary more. Three-dimensional structures of several of these enzymes have contributed to our understanding of their function, and such studies would shed new light on the acid sphingomyelinase family of proteins.

44 Chapter 2 – Structural Basis for Nucleotide Hydrolysis by the Acid Sphingomyelinase-Like Phosphodiesterase SMPDL3A

SMPDL3A was proposed to be a novel ectonucleotidase with no activity against SM. Diverse functions within the calcineurin-like enzymes structural superfamily had already been characterized, so this was not a total surprise. However, the 3D structures of ASMase, SMPDL3B and SMPDL3A had not been determined, so the basis for this substrate specificity was unknown. Here I present my structural and enzymatic study of SMPDL3A.

45 2.1 Abstract

Sphingomyelin phosphodiesterase, acid-like 3A (SMPDL3A) is a member of a small family of proteins founded by the well-characterized lysosomal enzyme, acid sphingomyelinase (ASM, ASMase). ASMase converts sphingomyelin (SM) into the signaling lipid, ceramide. It was recently discovered that, in contrast to ASMase, SMPDL3A is inactive against SM and, surprisingly, can instead hydrolyze nucleoside diphosphates and triphosphates, which may play a role in purinergic signaling. As none of the ASMase-like proteins have been structurally characterized to date, the molecular basis for their substrate preferences are unknown. Here we report crystal structures of murine SMPDL3A, which represent the first structures of an ASMase-like protein. The catalytic domain consists of a central mixed β-sandwich surrounded by ɑ-helices. Additionally, SMPDL3A possesses a unique C-terminal domain formed from a cluster of four ɑ-helices that appears to distinguish this from other phosphoesterases. We show that SMDPL3A is a di-zinc-dependent enzyme with an active site configuration that suggests a mechanism of phosphodiester hydrolysis by a metal-activated water molecule and protonation of the leaving group by a histidine residue. Co-crystal structures of SMPDL3A with AMP and α,β-methylene ADP (AMPCP) reveal that the substrate binding site accommodates nucleotides by establishing interactions with their base, sugar and phosphate moieties, with the latter being the major contributor to binding affinity. Our study provides the structural basis for SMPDL3A substrate specificity and sheds new light on the function of ASMase-like proteins.

46 2.2 Introduction

Sphingomyelin is a major lipid constituent of mammalian cellular membranes. Hydrolysis of sphingomyelin (SM) to ceramide and phosphocholine by sphingomyelinase enzymes (SMases) alters membrane properties and can result in the formation of signaling platforms, thus facilitating signal transduction [29]. Mammalian SMases are classified into three categories based on their pH optimum [187]. The lone alkaline SMase metabolizes SM in the intestinal tract [41] whereas neutral SMases are active on the cytosolic side of the plasma membrane and in various organelles [33]. The acid sphingomyelinase (ASMase) is found mainly in lysosomes, however, a fraction of this protein is secreted where it can act on the extracellular side of the plasma membrane and on lipoproteins in the circulation [109]. ASMase is a member of a small family of three paralogs that also includes SMPDL3A and SMPLD3B. The latter two proteins are poorly characterized compared to ASMase. SMPDL3B was recently shown to be a negative regulator of the immune signaling response induced by Toll-like receptors, and deficiency of this enzyme changes the cellular lipid composition and membrane fluidity [26]. Given its role in membrane modulation and the fact that SMPDL3B is anchored to the membrane via a glycosylphosphatidylinositol (GPI) link [114], its substrate is likely to be a membrane constituent like SM. The substrate and function of SMPDL3A, however, are less clear. This enzyme is secreted by osteoclasts [161], adipocytes [162], astrocytes [166] and macrophages [27], where it is up-regulated by the activation of the liver X receptor [188,174]. In the serum, SMPDL3A interacts with apolipoprotein A1 [158,159]. Its expression is increased in the vasculature in obesity [189], in atherosclerotic lesions [179] and in drug-induced phospholipidosis [190,191]. These findings suggested a lipid-related role for SMPDL3A in the circulation. Additionally, the mannose-6-phosphate (M6P) lysosomal targeting signal has been detected on this protein from multiple tissues [192], showing that SMPDL3A is transported to lysosomes [170]. However, its lysosomal targeting is reduced during pathological accumulation of cholesterol and glycolipids in that compartment [172]. More recently, SMPDL3A was reported to be a phosphoesterase up-regulated by cholesterol in macrophages [27]. Surprisingly, the enzyme was found to be inactive against SM or other lipids, but instead active in the hydrolysis of nucleotide diphosphates and triphosphates [27]. An anti-inflammatory role as a moderator of nucleotide-based purinergic signaling was

47 therefore proposed for this protein. Furthermore, SMPDL3A was shown to be inhibited by zinc, whereas its close homolog ASMase, is a zinc-dependent enzyme [109]. A recently discovered bacterial ASMase-like protein, RsASML, also displayed no activity against SM but was able to hydrolyze ATP and ADP [178]. This bacterial enzyme shares 23% sequence identity with human SMPDL3A, which is in turn 39% identical to SMPDL3B and 29% identical to ASMase. These reports revealed an unexpected divergence of substrate specificity within the ASMase-like family. As neither ASMase nor its paralogs have been structurally characterized, the basis for their substrate preferences remains unknown. Here we present high resolution crystal structures of murine SMPDL3A bound to several small molecules. The protein adopts a β-sandwich fold surrounded by ɑ-helices and has a unique C-terminal domain. Structures of the enzyme bound to AMP and an ADP analog explain its ability to accommodate nucleotide substrates and its lack of activity against SM. Based on the active site arrangement and enzymatic assay results, a catalytic mechanism for the hydrolysis of nucleotides is proposed.

48 2.3 Results

Overall structure of SMPDL3A — We determined the crystal structure of full-length murine SMPDL3A, which shares 80% identity with human SMPDL3A (Fig. 2.1). Our construct excluded the N-terminal signal peptide, which is normally not present in the mature protein, and extends from residues 23 to 445. The structure was solved at 1.8 Å resolution using SAD phasing from two tightly bound zinc ions in the active site. All subsequent structures with ligands were solved with molecular replacement (see Table 2.1 for data collection and refinement statistics). The enzyme adopts a relatively spherical shape with overall dimensions of 42 Å x 46 Å x 48 Å (Fig. 2.2A). The structure consists of two mixed β-sheets at its core flanked by ɑ-helices (Fig. 2.2B); this architecture classifies it as a member of the calcineurin-like phosphoesterase structural superfamily (PFAM code PF00149, Fig. 2.2C) which also contains nucleotidases and, most notably, the PPP family of serine/threonine phosphatases [7]. Both β-sheets contain six strands and form a mixed β-sandwich closed at one end and open at the other. The closed end as well as both longitudinal sides of the sandwich are surrounded by a total of seven ɑ-helices. In addition, the structure includes a helical C-terminal domain (CTD). The CTD forms a cluster of four ɑ-helices packed up against the open end of the β-sandwich, burying ~1900 Å2 of surface area, and is unique to the three ASMase-like proteins in mammals (Fig. 2.1). The interface between the domains is formed mainly by hydrophobic contacts as well as multiple hydrogen bonds. Two disulfide bonds stabilize the CTD while another disulfide bridge is present in the catalytic domain. The structure is also decorated by six N-linked glycans. Human SMPDL3A [27] shares five of these glycosylation sites and has two additional potential sites, which correspond to surface-exposed locations in the structure.

49 Figure 2.1 Sequence alignment of SMPDL3A and ASMase

The amino acid sequences of the catalytic and C-terminal domains (CTD) of human and murine SMPDL3A (80% identity) as well as of human ASMase (29% identity to SMPDL3A) and

50 SMPDL3B (39% identity) are aligned. Cysteines involved in conserved disulfide bonds are highlighted in yellow, metal-coordinating residues, in green and potential N-glycosylation sites, in gray. The two histidines and the aspartic acid involved in phosphate binding or catalysis are colored in pink. The tyrosine and the glutamine forming the sides of a cleft which accommodates the adenine base of nucleotides are shown in blue, as are the corresponding residues of human SMPDL3A. The CTD is delimited by blue brackets. Secondary structure elements of murine SMPDL3A are shown as red cylinders for helices and yellow arrows for strands.

51 Figure 2.2 Structure of SMPDL3A

(A) The core catalytic domain is shown in green and the CTD in blue. Disulfide bonds appear as yellow sticks. N-linked glycans (white sticks) are only partially displayed for clarity. The two zinc ions in the active site are represented by spheres. N and C-termini as well as secondary structure elements are labeled. (B) Topology diagram of SMPDL3A. The central β-sheets are shown in gray rectangles. Disulfide bonds are represented by yellow connectors. Red circles indicate locations of metal binding or catalytic residues. (C) The β-sandwich fold of SMPDL3A (green) is comparable to that of related phosphoesterases including the purple acid phosphatase (PAP, PDB code 1WAR, overall sequence identity = 11%, r.m.s.d. = 1.5 Å for 46 corresponding ɑ-carbons from the conserved β-strands only), shown in violet, and calcineurin (PDB code 3LL8, overall sequence identity = 11%, r.m.s.d. = 2.6 Å for 46 corresponding ɑ-carbons from the conserved β-strands only), displayed in yellow. The loop regions and the CTD of SMPDL3A are hidden for clarity.

52 Active site configuration — The active site of SMPDL3A is in a depression on the surface of the protein, formed by loops extending from the central β-sheets. At its core are found two metal ions coordinated by the side chains of seven conserved residues (Fig. 2.3). A sulfate ion from the crystallization solution is positioned above the metals and coordinates them with two of its oxygen atoms. The third sulfate oxygen forms hydrogen bonds with two nearby histidines, H149 and H111. An additional interaction is formed between the sulfate ion and one of the metal-binding residues, N148. A water molecule also bridges both metals.

53 Figure 2.3 Active site of SMPDL3A

Residues that coordinate the two zinc ions (gray spheres) are shown as sticks, as are the two histidines (H111 and H149) interacting with a sulfate ion (sticks). The electron density for the water molecule (red sphere) located midway between the zinc ions and for the sulfate ion is displayed as a Fo – Fc simulated annealing omit map contoured at 8σ. Distances are indicated (Å).

54 SMPDL3A is part of a superfamily of phosphoesterases that utilize two metal cations for catalysis. While ASMase is known to require zinc for activity [109], SMPDL3A was shown to be inhibited by zinc [27]. As all metal-coordinating residues are conserved between both proteins (Fig. 2.1), this finding was rather perplexing. Thus, we sought to further investigate the cation dependency of SMPDL3A. When purified from insect cell media, the recombinant enzyme is catalytically active, and X-ray absorption scans on the protein crystals indicated the presence of zinc (data not shown). We found that low concentrations of zinc ranging from 5 to 100 μM caused the hydrolysis rate of the generic phosphodiesterase substrate bNPP by SMPDL3A to double (data not shown). However, increasingly higher concentrations of zinc did indeed inhibit the enzyme as reported [27], most noticeably at neutral pH. To understand the structural basis for this inhibition, we determined the structure SMPDL3A in presence of supraphysiological levels of zinc (10 mM). Under these conditions, we found that a third zinc ion becomes bound by the protein, coordinated by the side chains of H149 and H111 (Fig. 2.4). The third zinc actually blocks the active site by occupying the location where the sulfate or substrate phosphate group binds under normal conditions. This inhibition could be reversed by mild EDTA treatment to remove the loosely bound third zinc ion. These results suggest that, like ASMase, SMPDL3A is a zinc-dependent enzyme.

55 Figure 2.4 Inhibition of SMPDL3A by excess zinc

The active site of the protein exposed to a high zinc concentration reveals a third zinc ion bound to the active site. Zinc ions (spheres) with their corresponding anomalous difference electron density map peaks (yellow mesh) are shown contoured at 10σ.

56 Substrate recognition — A variety of substrate specificities exists within the phosphoesterase superfamily (PFAM code PF00149), including nucleic acids and phosphoproteins. However, as SMPDL3A is most similar to the well-studied sphingomyelinase ASMase, the finding that the former hydrolyzes nucleotides rather than lipids [27] was unexpected. To address why SMPDL3A can hydrolyze nucleotides but not SM, we determined its structure in the presence of several possible substrates and reaction products (Fig. 2.5). It should be noted that the crystallization solution contained 200 mM sulfate, a competitor for the active site and therefore, substrates may not be present at full occupancy in the crystals. The substrate-binding site is a large water-filled T-shaped cavity that emanates from the zinc ions. The crystallization of phosphocholine (PC), a product of SM hydrolysis, with SMPDL3A revealed that it can bind to the active site (Fig. 2.5A, 2.5B), but the interaction is mediated mainly by the phosphate group. The positively charged choline moiety does not form any specific interactions with the protein that could stabilize it, such as cation-π [193] or electrostatic contacts with negatively charged residues. By contrast, the structure of SMPDL3A bound to AMP, shows that in addition to specific phosphate interactions, the adenine ring is positioned in a cleft between the side chains of Q324 and Y257 (Fig. 2.5C). This arrangement allows for π stacking between the tyrosine and the adenine base. However, the ribose group of AMP does not stably interact with the protein, and its electron density is fragmented, suggesting conformational flexibility in this region (Fig. 2.5D).

57 Figure 2.5 Ligands bound in the active site of SMPDL3A

The structures of ligands present in the active site are displayed along with their corresponding electron density Fo – Fc simulated annealing omit maps contoured at 3σ. (A, B) Phosphocholine (PC). (C, D) AMP. (E, F) ADP analog (AMPCP). Zinc ions and water molecules are represented by gray and red spheres, respectively. The two residues interacting with the adenine base are shown as sticks and labeled.

58 We also crystallized the protein in the presence of the ATP analog AMPCPP, which is uncleavable between the ɑ- and β-phosphates. However, the electron density map revealed only the presence of AMPCP in the active site, indicating that the γ-phosphate was hydrolyzed by the protein (Fig. 2.5E). The position of the adenine ring of AMPCP is similar to that of AMP and the electron density map around it is more clearly defined in this case (Fig. 2.5F). One notable difference between the two is the orientation of their terminal phosphate: in AMPCP, the three oxygen atoms coordinate the zinc ions, while in AMP, one oxygen points away from them (discussed further below). The terminal phosphate group of both ligands forms hydrogen bonds with H111, N148 and H149, although in slightly different manners due to its shift in position. Again, there are no specific interactions between the ribose moiety and the protein. To assess the importance of the tyrosine involved in stacking interactions with the nucleotide base, kinetic parameters of the wild-type protein were compared to those of the Y257A mutant (Table 2.2). Activity was assayed at pH 7.5 and 5, reflecting the localization of

SMPDL3A in the extracellular space as well as in lysosomes. First, we measured the Km values for ATP hydrolysis by the wild-type human and murine enzymes at neutral and acidic pH to be in the range of 0.1 – 1 mM. Affinity for ATP is three to four times higher at acidic pH for both murine and human proteins, possibly due to electrostatic attraction between the negatively charged phosphate groups and protonated histidines H111 and H149. Surprisingly, we found that mutation of Y257 in murine SMPDL3A did not affect the Km for ATP and lowered the turnover number only slightly. These results suggest that Y257 does not contribute significantly to the binding affinity of SMPDL3A for nucleotide substrates. Instead, the cleft formed by Q324 and Y257 may simply provide shape complementarity to the base moiety to allow entry of nucleotide substrates whilst excluding other types of phosphodiester-containing molecules. In this manner, SMPDL3A would be able to accommodate a variety of different nucleotide substrates by forming relatively weak non-specific interactions with their non-phosphate portions, with strong binding affinity arising mainly from contacts between their phosphate groups and H111, N148, H149 as well as the zinc ions. Proposed catalytic mechanism — Comparison of the high resolution structures of SMPDL3A with those of other phosphoesterases from the calcineurin-like superfamily allowed us to propose a nucleotide hydrolysis mechanism for it. An overlay of the active sites of SMPDL3A and calcineurin is shown in Fig. 2.6A. The majority of the metal-coordinating

59 residues are conserved, with the exception of H292 which is unique to SMPDL3A. Calcineurin is a Ser/Thr phosphatase and its mechanism of protein dephosphorylation involves a nucleophilic attack by a metal-activated water on the phosphate [7]. Importantly, the serine or threonine leaving group is protonated by a histidine residue paired with an aspartate, both highly conserved in the phosphoesterase superfamily [194]. In SMPDL3A, H149 is the corresponding histidine and forms a hydrogen bond to the phosphate (or sulfate) group (Fig. 2.6A). However, the matching aspartate is missing and instead, an adjacent H111-D79 pair is present, which also hydrogen bonds to the phosphate group (Fig. 2.6A). Interestingly, this second histidine is replaced by arginine in calcineurin and suggests that the identity of histidine protonating the leaving group could be different in SMPDL3A. To determine which histidine is responsible for leaving group protonation, kinetic parameters of ATP hydrolysis by corresponding glutamine or alanine mutants were determined. Mutation of either H149 or H111 lead to a roughly tenfold decrease in substrate affinity and lowered the turnover rate by at least two orders of magnitude (Table 2.2). We can therefore only conclude that both histidines are essential for the catalytic activity of SMPDL3A, with H111 assisted by D79 most likely responsible for leaving group protonation.

60 Figure 2.6 Proposed reaction mechanism for SMPDL3A

(A) The active site of SMPDL3A is compared with that of calcineurin (PDB code 3LL8). Residues in beige, one zinc ion and one iron ion (orange sphere) form the calcineurin catalytic center, occupied by a phosphate ion (orange central atom). Residues in green and a sulfate ion (yellow central atom) are part of the SMPDL3A active site. Active site residue labels in parentheses refer to calcineurin. (B) A comparison of SMPDL3A bound to AMP and AMPCP is shown (only the ligands are depicted). The structures are in the identical orientation to illustrate the difference in the phosphate group between the two ligands. Zinc ions and the nucleophilic water are labeled. (C) Model of ATP substrate bound to SMPDL3A. An ATP molecule was manually docked into the substrate binding site using the AMP complex structure as a guide. The β-phosphate is positioned for nucleophilic attack by a water molecule (red sphere) while the γ - phosphate acts as the leaving group. Potential interactions with protein residues are indicated.

61 (D) Schematic diagram of the proposed ATP hydrolysis mechanism for SMPDL3A, which involves a nucleophilic attack by a zinc-activated water molecule on the β-phosphate. This is followed by protonation of the departing γ-phosphate by one of the two nearby histidines, likely H111 with assistance from D79.

62 In calcineurin, the attacking water molecule is located midway between the zinc ions, which are thought to increase its nucleophilicity as well as the electrophilicity of the phosphate moiety [195]. An analogous water molecule is visible in the active site of SMPDL3A bound to sulfate, PC or AMP (Figs. 2.3, 2.5B, 2.5D). For the hydrolysis of nucleoside triphosphate substrates into nucleoside diphosphates, the scissile bond is between the γ-phosphorus atom and the oxygen atom bridging it to the β-phosphorus atom (Pγ-O-Pβ). This bond cleavage can in theory occur by attack of the nucleophilic water molecule onto either the γ-phosphorus with ADP acting as the leaving group, or the β-phosphorus atom with the γ-phosphate as the leaving group. Our structures of SMPDL3A in complex with AMP and AMPCP help us discern between these two possibilities. In the AMP structure, the nucleotide appears to be bound as bona fide substrate with the nucleophilic water molecule poised for attack 2.9 Å away from the AMP phosphorus atom (Fig. 2.5D, 2.6B). Two of the oxygen atoms of the phosphate coordinate with the zinc ions, while the third oxygen points in the direction opposite the incoming nucleophilic water. However, in the AMPCP structure, the nucleotide appears to be bound as a hydrolysis product of AMPCPP. In this case, the positions of two zinc-coordinating oxygen atoms are essentially the same, but in stark contrast, the orientation of the third oxygen atom is flipped as in an inverted configuration, and replaces the position of the attacking water molecule (Fig. 2.6B). This analysis allowed us to construct a manually docked model of SMPDL3A bound to ATP – the preferred substrates for the enzyme being nucleoside triphosphates [27] (Fig. 2.6C). Importantly, if the γ-phosphate of ATP is placed in the active site then the ADP leaving group cannot be positioned appropriately to become protonated by H149/H111. Only when the β- phosphate is positioned in the active site, is the γ-phosphate leaving group hydrogen bonded by H149 and H111 for protonation. With the β-phosphate of ATP positioned in the active site, the sugar and base moieties are pushed back further into the cleft between Q324 and Y257 than in the case of the AMP/AMPCP co-crystal structures. This difference could enable hydrogen bonding between the ribose O4’ and Y257 as well as between O2’ and Q324, although the wide binding site precluded modeling the orientation of the ribose with great certainty. These additional potential interactions may in part explain the preference for nucleotide triphosphate substrates by SMPDL3A. In this arrangement, the β-phosphate makes contacts with the zinc ions and can undergo nucleophilic attack by the activated water molecule with a concomitant inversion of its configuration (Fig. 2.6D). The γ-

63 phosphate leaving group is protonated by either H149 or H111 and departs. Similarly, for the conversion of ADP to AMP, the active site would be occupied by the α-phosphate while the terminal β-phosphate would act as the leaving group.

64 2.4 Discussion

This study reports the first structural characterization of a member of the ASMase-like family of enzymes. The architecture of the catalytic domain is similar to other phosphoesterases, whereas the small CTD is unique to the family. The purpose of the CTD is unknown, but based on its tight association with the catalytic domain, it may simply be an extension of the latter. Mutations in the CTD of ASMase have been shown to inactivate the protein [49], but whether this is due to effects on trafficking, activity or protein stability is unclear. Unlike SMPDL3A and SMPDL3B, ASMase contains a saposin domain that facilitates binding to membranes and lipids [109]. SMPDL3B, in turn, is attached to the membrane via a GPI anchor [26]. The lack of any obvious membrane-interacting features in the SMPDL3A structure, such as the hydrophobic hairpin of bacterial sphingomyelinases [196], may be related to its evolutionary specificity towards the hydrolysis of nucleotides rather than membrane lipids. It should be noted that a number of lipid-degrading enzymes require the assistance of protein cofactors for access to lipids [109], and we cannot exclude this possibility for SMPDL3A. Although SMPDL3A can accommodate the choline head group of SM in its active site, binding is mediated mainly by the phosphate moiety. Interestingly, in the case of nucleotide substrates, binding affinity also seems to stem mainly from electrostatic and hydrogen bond interactions involving the β- and possibly γ-phosphates (or ɑ- and β-phosphates in nucleotide diphosphates). Interactions with parts of the nucleotide beyond the phosphates are weak and do no contribute significantly to binding as demonstrated by the Y257A mutant. Indeed, this tyrosine is replaced by a serine residue in human SMPDL3A (Fig. 2.1), which is also active against nucleotides, further confirming that Y257 is not important for substrate binding. We suggest that the relative lack of specific recognition beyond the phosphate group is what allows this protein to hydrolyze different types of nucleotide triphosphates and diphosphates [27]. Our work reconciles previous findings of inhibition of SMPDL3A by zinc, with the known dependency of ASMase on this metal. Millimolar concentrations of zinc block the active site of the enzyme and optimal activity is only possible at lower concentrations in the micromolar range, consistent with the physiological serum levels of zinc [109]. Excessive zinc concentrations are known to also inhibit ASMase [99]. This artificial inhibition is less

65 pronounced at acidic pH, perhaps due to electrostatic repulsion between the metal ions and the partially protonated coordinating histidine residues. The catalytic mechanism proposed here is similar to that of other members of the phosphoesterases superfamily, with the identity of the catalytic histidine as a possible key difference. Our mechanism accounts for the ability of SMPDL3A to hydrolyze ATP and ADP while remaining inactive against AMP [27], as there is no possible leaving group when AMP is bound in the same orientation as the other nucleotides. Additionally, it accounts for the much higher Kcat of the enzyme at pH 5, since the catalytic histidine would be a better proton donor in an acidic environment. All of the residues involved in the proposed mechanism are conserved in ASMase (Fig. 2.1). In addition, disulfide-forming cysteines are conserved, and sequence insertions occur only outside of secondary structure elements, suggesting that ASMase adopts the same overall fold, including the CTD. However, there is no sequence conservation with ASMase in residues that interact with the substrate portions beyond the phosphate group (Fig. 2.1), which could explain the different substrate specificities of these two proteins. This study therefore sheds new light on the function of ASMase-like enzymes.

66 2.5 Experimental procedures

Constructs — SMPDL3A from mouse (residues 23-445; Uniprot code P70158) and human (residues 23-453; Uniprot code Q92484) were subcloned into a derivative of pFastBac 1 (Invitrogen). The vector contained the melittin signal peptide MKFLVNVALVFMVVYISYIYA followed by a hexahistidine tag DRHHHHHHKL. Protein expression and purification — Recombinant baculoviruses were generated according to the Bac-to-Bac Baculovirus Expression System protocol (Invitrogen) with minor modifications: DH10MultiBac cells were used [197] and viruses were added to Sf9 cells grown in I-Max medium (Wisent Bioproducts). Protein expression took place at 27°C for 64 hours. Subsequent steps were carried out at 4°C. Cells were removed by centrifugation at 1,000g then at 9,000g and supernatants containing the secreted protein were incubated with HisPur Ni-NTA resin (Thermo Fisher Scientific). The beads were washed with buffer A (50 mM Tris-HCl pH

7.5, 500 mM NaCl, 1 mM MgCl2) and eluted with buffer A containing 250 mM imidazole. Proteins were concentrated and loaded on a Superdex 200 10/300 GL size-exclusion column (GE Healthcare) equilibrated with buffer B (15 mM Tris-HCl pH 7.5, 100 mM NaCl). The protein was concentrated to 10 mg mL-1 and flash-frozen. Enzymatic assay with generic phosphodiesterase substrate — Protein at 50 nM was incubated with 1.7 mM bis(p-nitrophenyl) phosphate (bNPP) in assay buffer (100 mM NaCl and 100 mM Tris-HCl pH 7.5 or sodium acetate pH 5) at 37°C. 100 mM NaOH was then added before measuring absorbance at 405 nm. The change in absorbance after 30 min was used for obtaining the normalized activity rates. Enzymatic assay with nucleotide substrate — The Biomol Green phosphate detection method (Enzo Life Sciences) was slightly modified. Protein at 1 nM to 2 µM was incubated with

4.35 µM to 2 mM ATP in assay buffer supplemented with 5 µM ZnCl2 at 37°C. Four volumes of Biomol Green reagent were added followed by incubation and absorbance measurement at 620 nm. The difference in absorbance after 15 min of ATP hydrolysis by SMPDL3A was used to interpolate the amount of phosphate produced via a standard curve. Linearity of the initial rates was ensured by comparing the result with those of a 30 min hydrolysis reaction. Michaelis- Menten parameters were obtained by non-linear regression of the initial rates at varying ATP concentration.

67 Crystallization and data collection — Crystals were grown by hanging drop vapor diffusion at 22°C. Protein at 10 mg mL-1 in buffer B was mixed with an equal volume of well solution. Murine SMPDL3A was crystallized in 0.2 M (NH4)2SO4, 0.1 M sodium acetate pH 4.6 and 25% PEG 4000. For anomalous diffraction experiments, crystals were soaked in well solution supplemented with 20 mM ZnCl2. Ligand-bound structures were obtained by crystallizing the protein in the presence of 50 mM phosphocholine (PC), AMP, or ɑ, β-methylene ATP (AMPCPP). All crystals were briefly soaked in well solution supplemented with 20% glycerol before flash-freezing. Diffraction data were collected at 100 K on beamlines 08ID-1 and 08B1-1 at the Canadian Macromolecular Crystallography Facility, Canadian Light Source, and on beamline MacCHESS F1, Cornell High Energy Synchrotron Source. Data were processed with HKL2000 [198]. Structure determination and refinement — The structure of murine SMPDL3A was solved by zinc single-wavelength anomalous diffraction using Autosol in Phenix [199] and manually rebuilt in Coot [200]. Subsequent structures were obtained by molecular replacement using Phaser in Phenix. Refinement was performed with phenix.refine. Data processing and refinement statistics are presented in Table 2.1.

68 Chapter 2 concluding transition

My structural and enzymatic study of murine SMPDL3A revealed a zinc-dependent enzymatic mechanism with a distinct His-Asp catalytic pair and helical C-terminal subdomain – features likely shared with its paralogs ASMase and SMPDL3B. Its nucleotide binding mode was also determined. However, the protein does not establish strong contacts with any parts of the substrate besides the phosphate moieties. For instance, the π-stacking interaction between the nucleobase moiety and Tyr257 seems dispensable for activity. At the same time my study was published, the crystal structure of human SMPDL3A was also reported [201]. The enzyme was demonstrated to additionally hydrolyze the substrates CDP-choline, CDP-ethanolamine and ADP-ribose with efficiencies comparable to ATP and ADP. The former two compounds are intracellular building blocks for phospholipid synthesis [202], whereas ADP-ribose could play a role in neurotransmission [181]. The crystal structure of the protein in complex with a CMP product was also determined. Interestingly, the nucleotide is bound in different orientations in the three copies of the protein in the asymmetric unit of the crystal (Fig. 2.7). Moreover, these orientations differ from that of AMP in complex with the murine enzyme. This apparent lack of specificity could reflect a “maintenance” role to recycle various nucleotides in the extracellular space [201]. Further in vivo studies are needed to clarify the functions of SMPDL3A.

69 Figure 2.7 Nucleotide binding modes of human and murine SMPDL3A

The human enzyme (green) and its CMP ligand (yellow tones, three copies from three protein molecules in the asymmetric unit of the crystal, PDB code 5EBE) is compared to its murine counterpart (cyan) bound to AMP (gray, this thesis).

70 Table 2.1 SMPDL3A X-ray data collection and structure refinement statistics

Parameters Sulfate-bound PC-bound AMPCP-bound Sulfate-bound Zinc-inhibited AMP-bound Crystal used (tetragonal) for Zn-SAD PDB code 5FC1 5FC5 5FC6 5FC7 5FCA 5FCB

Data Collection

Beamline 08ID-1, CMCF, F1, MacCHESS 08ID-1, CMCF, 08ID-1, CMCF, 08B1-1, CMCF, 08ID-1, 08ID-1, CLS CLS CLS CLS CMCF, CLS CMCF, CLS Wavelength (Å) 0.97949 0.91790 0.97949 0.97949 1.28137 0.97949 1.28223

Space group C 2 2 21 C 2 2 21 C 2 2 21 P 41 P 21 21 21 C 2 2 21 C 2 2 21

Unit cell dimensions 123.65 131.87 123.734 132.90 124.35 132.89 87.39 87.39 80.14 123.59 124.03 132.38 123.70 132.63 (Å) 80.06 80.14 80.52 79.84 131.61 79.74 80.36 Unit cell angles (°) 90 90 90 90 90 90 90 90 90 90 90 90 90 90 90 90 90 90 90 90 90

Resolution range 48.93 - 1.389 28.25 - 1.678 45.4 - 1.658 36.31 - 1.456 39.31 - 1.924 39.87 - 1.550 50.00 - 1.770 (Å) (1.439 - 1.389) (1.738 - 1.678) (1.717 - 1.658) (1.508 - 1.456) (1.993 - 1.924) (1.605 - 1.550) (1.83 - 1.770) Resolution at 2.88 2.37 CCano = 30 % (Å) Total reflections 1,606,469 952,273 567,906 450,079 2,339,663 582,620 824,926

Unique reflections 124,089 (7898) 68,190 (3859) 78,590 (7430) 100,829 (7376) 93,218 (6018) 89,133 (6001) 63,961 (5816)

Multiplicity 12.9 (2.8) 14.0 (11.5) 7.2 (5.6) 4.5 (1.9) 25.1 (6.7) 6.5 (2.6) 12.9 (5.6)

Completeness (%) 94.7 (61.9) 90.4 (52.8) 99.6 (97.4) 96.4 (72.2) 93.6 (61.7) 93.7 (63.8) 99.0 (91.2)

I/σ(I) 23.5 (0.5) 17.1 (0.4) 15.8 (0.8) 15.0 (0.3) 23.8 (0.5) 19.7 (0.6) 15.7 (1.0)

Wilson B-factor 20.24 29.44 22.78 25.35 37.92 23.5 (Å2)

Rmeas (%) 9.4 (> 100) 12.2 (> 100) 10.6 (>100) 7.4 (> 100) 11.9 (> 100) 8.6 (> 100) 13.4 (> 100)

CC1/2 (%) (15) (17) (24) (16) (39) (20) (17)

Refinement

Reflections used in 123,471 (7329) 66,735 (3350) 78,503 (7350) 98,417 (5318) 92,449 (5359) 88,924 (5829) refinement Reflections used for 6150 (348) 3355 (176) 3932 (377) 1957 (114) 1983 (112) 4411 (264) R-free R-work 0.1293 (0.4420) 0.1734 (0.3815) 0.1622 (0.3534) 0.1539 (0.4304) 0.1754 (0.3889) 0.1592 (0.3815) R-free 0.1609 (0.4896) 0.2034 (0.3868) 0.1913 (0.3923) 0.1827 (0.4473) 0.2034 (0.4227) 0.1871 (0.4004) Number of non- 4298 4165 4208 3979 8138 4198 hydrogen atoms macromolecules 3485 3457 3470 3401 6904 3463

ligands 268 220 218 172 426 223

solvent 545 488 520 406 808 512

Protein residues 426 426 426 423 852 426

RMS(bonds) (Å) 0.02 0.005 0.01 0.009 0.012 0.007

RMS(angles) (°) 1.66 0.98 1.15 1.09 1.23 1.05

Ramachandran 96 96 96 96 96 96 favored (%) Ramachandran 4.3 4.3 4.1 3.7 4.2 4.1 allowed (%) Ramachandran 0 0 0 0 0 0 outliers (%) Rotamer outliers 0.5 0.51 0.51 0.52 0.77 0.51 (%) Clashscore 4.02 1.92 1.37 2.7 2.75 1.51

Average B-factor 30.33 40.26 31.32 33.88 48.38 33.06 (Å2) macromolecules 25.54 36.72 27.44 31.35 45.27 29.13

ligands 69.11 80.77 71.9 66.63 94.13 70.67

solvent 41.91 47.1 40.17 41.21 50.87 43.22

71

Values in parentheses refer to the highest resolution shell. High resolution cutoff was guided by the correlation coefficient CC1/2 [203]. Zn-SAD, zinc single wavelength anomalous diffraction. CCano, anomalous correlation coefficient. CMCF, Canadian Macromolecular Crystallography Facility. CLS, Canadian Light Source. MacCHESS, Macromolecular Diffraction Facility at the Cornell High Energy Synchrotron Source.

72 Table 2.2 Michaelis-Menten parameters of ATP hydrolysis by SMPDL3A pH 5 pH 7.5 -1 -1 Protein km (μM) kcat (min ) km (μM) kcat (min ) human WT 174 ± 11 3258 ± 160.3 735 ± 14 168 ± 4.2 murine WT 107 ± 9 1542 ± 63.6 330 ± 10 260 ± 3.8 murine Y257A 99 ± 2 860 ± 8.9 310 ± 26 172 ± 6.7 murine H149Q 466 ± 49 11 ± 0.2 2468 ± 285 2 ± 0.1 murine H149A 1099 ± 96 1 ± 0.0 * murine H111Q 1163 ± 107 1 ± 0.1 * murine H111A 1221 ± 141 3 ± 0.3 *

Activity of wild-type proteins as well as murine mutants against ATP was assayed. Values are the mean and standard deviation of a representative of two experiments performed in triplicates. Asterisks indicate rates too low for parameter determination.

73 Chapter 3 – Crystal Structure of Mammalian Acid Sphingomyelinase

ASMase was cloned almost thirty years ago and extensively characterized, so the lack of structural studies of this enzyme was surprising. Although several of the features observed in the SMPDL3A structures can be extended to its paralog, ASMase is a more complex protein, bearing a saposin domain for membrane or lipid binding. At the time, no structure of a saposin cofactor in complex with an enzyme had been reported. Here I present my structural and enzymatic study of ASMase.

74 3.1 Abstract

Acid sphingomyelinase (ASMase, ASM, SMPD1) converts sphingomyelin into ceramide, modulating membrane properties and signal transduction. Inactivating mutations in ASMase cause Niemann-Pick disease, and its inhibition is also beneficial in models of depression and cancer. To gain a better understanding of this critical therapeutic target, we determined crystal structures of mammalian ASMase in various conformations. The catalytic domain adopts a calcineurin-like fold with two zinc ions and a hydrophobic track leading to the active site. Strikingly, the membrane interacting saposin domain assumes either a closed globular conformation independent from the catalytic domain, or an open conformation which establishes an interface with the catalytic domain essential for activity. Structural mapping of Niemann-Pick mutations reveals that most of them likely destabilize the protein’s fold. This study sheds light on the molecular mechanism of ASMase function, and provides a platform for the rational development of ASMase inhibitors and therapeutic use of recombinant ASMase.

75 3.2 Introduction

Acid sphingomyelinase (ASMase, ASM, SMPD1) is an enzyme found in lysosomes and in the extracellular space where it catalyses the conversion of sphingomyelin, a major component of membranes, into ceramide and phosphocholine [204,109]. Ceramide is a key signalling lipid that modulates membrane biophysical properties and is involved in numerous cellular processes as well as diseased states. Hereditary mutations of ASMase result in the toxic accumulation of sphingomyelin and are the cause of the lysosomal storage disease Niemann-Pick type A, which is neuropathic and fatal, and type B, which causes visceral anomalies [49]. Clinical trials for treating this disease have been carried out using recombinant ASMase as enzyme replacement therapy [204]. More recently, it was discovered that inhibition of ASMase activity mediates the effects of antidepressant drugs in hippocampal neurons [107,205], selectively kills cancer cells by destabilizing their fragile lysosomes [74,206], reduces inflammation associated with cystic fibrosis [207], decreases atherosclerotic lesions [208], and diminishes symptoms associated with Alzheimer’s disease [209]. These findings have established ASMase as a critical therapeutic target. However, while its activity has been shown to be indirectly inhibited by certain antidepressants [107] and other cationic amphiphilic drugs (CADs) that lead to its degradation, no direct inhibitor of ASMase is currently in medical use. Although the protein was characterized over thirty years ago, knowledge of its three-dimensional structure for aid in inhibitor development is still lacking. ASMase consists of an N-terminal saposin domain and a C-terminal catalytic domain. The saposin domain is found in small lysosomal proteins that generally act in isolation by binding to or extracting lipids from the membrane for subsequent breakdown by hydrolytic enzymes [210,79,78,76], and the structures of several saposins have been determined. Sequence analysis of the catalytic domain classifies it as a phosphoesterase enzyme and, recently, the crystal structure of the ASMase paralog, acid sphingomyelinase-like phosphodiesterase 3A (SMPDL3A) was reported [201,211], which provided some insight into the catalytic mechanism. However, SMPDL3A cannot cleave sphingomyelin and hydrolyzes nucleotides instead [27]. Moreover, it lacks a saposin domain, which limits its usefulness for understanding how ASMase works.

76 To gain insight into the molecular mechanism of ASMase function and the interplay between its domains, we determined crystal structures of the full-length protein from mouse that reveal a conformational flexibility of the saposin domain. In one conformation, the saposin domain adopts a globular closed form independent of the catalytic domain, while in the other conformation, it assumes an open V-shaped fold that establishes an extended interface with the catalytic domain. Activity assays demonstrate that this interface is indispensable for substrate hydrolysis.

77 3.3 Results

ASMase is synthesized and differentially modified post-translationally in the endoplasmic reticulum and Golgi to yield distinct lysosomal and secreted forms. Our crystallized construct (murine residues 84 – 611 corresponding to human residues 88 – 615, 88% identity to human ASMase; Fig. 3.1A) produced in insect cells most closely resembles the well- characterized lysosomal form with regards to glycosylation pattern, cleavage of a nine-residue C- terminal peptide that renders the enzyme fully active [102,103], and N-terminal processing up to residue 84 (human residues 85 to 622) [101]. We determined several structures of ASMase (Table 3.1) in various forms including two distinct conformations in which the N-terminal saposin domain (ASMasesap) changes its fold and position relative to the catalytic domain (ASMasecat, Fig. 3.1B).

78 Figure 3.1 Structural overview of ASMase

(a) Domain organization of ASMase. SP, signal peptide; PL, proline rich linker; CTD, C- terminal subdomain. (b) Closed and open ASMase coloured as in (a). Consecutive grey spheres indicate the disordered linker between closed ASMasesap and ASMasecat in one molecule of the asymmetric unit. Boxed region magnified in (c) is not in the identical orientation. Glycans omitted for clarity. Zincs are indicated as purple spheres. (c) ASMase active site. Zinc interactions are shown as dashes. Zinc-zinc distance = 3.6 Å. Zinc-ligand distance range from 2.0 Å to 2.1 Å. Phosphate is coloured orange and red. Shown are zinc interacting residues (beige) and residues important for leaving group protonation and substrate binding (green).

79 Catalytic domain and active site — We begin by describing ASMasecat, which was also crystallized in isolation. Its core, composed of two six-stranded mixed β-sheets surrounded by eight ɑ-helices, belongs to the calcineurin-like phosphoesterase structural family (Pfam code PF00149) and resembles recent structures of the ASMase-like protein, SMPDL3A [201,211] (Fig. 3.2B). Like SMPDL3A, ASMasecat possesses an additional C-terminal subdomain (CTD) consisting of four ɑ-helices that pack up against the core burying 1795 Å2 of surface area, thus distinguishing it from most other phosphoesterases. The linker connecting ASMasecat to ASMasesap is proline-rich and mostly rigid, wrapping itself around the catalytic domain as an L- shaped strap (Fig. 3.1B). The protein has five sites of N-glycosylation (three in ASMasecat, one in the linker, and one in ASMasesap) (Fig. 3.2C) and eight disulfide bonds (five in ASMasecat and three in ASMasesap, Fig. 3.2A).

80 Figure 3.2 The ASMase fold

(a) Left, topology diagram of ASMase. The central β-sheets of the catalytic domain are highlighted with gray shading. Strands are represented by arrows and ɑ-helices, by cylinders. The topology of the open form of the saposin domain (pink) is shown; in the closed form, ɑ-helix 3 is divided into two helices. Disulfide bonds are represented by yellow connectors. Red circles indicate regions of metal binding or catalytic residues. CTD helices are coloured lime green. Right, Structure of ASMase with secondary structure elements labelled. (b) Comparison of the ASMase structure with related phosphodiesterases. The β-sandwich fold of the ASMase catalytic domain (green) is comparable to that of related phosphodiesterases including calcineurin (PDB code 3LL8, overall sequence identity = 10%, r.m.s.d. = 1.9 Å for 40 corresponding ɑ-carbons from the conserved β-strands only), displayed in gold. In addition, the catalytic and C-terminal domains of ASMase are conserved in SMPDL3A (PDB code 5FC1, overall sequence identity =

81 29%, r.m.s.d. = 0.9 Å for 281 corresponding ɑ-carbons), which is displayed in blue. (c) Glycosylation sites (coloured sticks) on ASMase produced from insect cells. Asn N-glycan attachment residues are shown as black sticks.

82 Together, these elements form a spherical domain with a shallow depression on one side, at the base of which resides a two-zinc ion centre. In the structure of isolated ASMasecat, a phosphate ion is bound to the active site and forms several interactions with the protein and zinc ions, mimicking the scissile phosphate from sphingomyelin. As in other phosphoesterases the di- metal active site has octahedral coordination geometry from seven highly conserved protein residues, a water molecule, and completed by the phosphate group (Fig. 3.1C and 3.3A). Additionally, His317 and His280 are positioned near the bound phosphate as potential proton donors for the ceramide leaving group (Fig. 3.1C). Mutation of these histidines and assessment of enzymatic activity at acidic pH with the generic small molecule phosphodiesterase substrate bis(p-nitrophenyl) phosphate (bNPP) revealed that His280 assisted by nearby Asp249 is the likely proton donor, while His317 is probably important for substrate binding (Fig. 3.3C). Together, these findings suggest that ASMasecat catalyzes sphingomyelin hydrolysis by the canonical mechanism used by phosphoesterases in which nucleophilic attack of a zinc-activated water molecule and protonation of the leaving group by His280 releases phosphocholine and ceramide (Fig. 3.3D).

83 Figure 3.3 Comparison and mutation of the ASMase active site and proposed catalytic mechanism

84 (a) Active site of ASMase (green) is superimposed with that of calcineurin (PDB code 3LL8, beige) or SMPDL3A (PDB code 5FC1, blue). The active sites are very well conserved and likely use a similar mechanism of catalysis. However, there are slight differences. For example, in calcineurin the His-Asp proton donating pair actually corresponds to His317 in ASMase, which is more likely involved in substrate binding. Additionally, the His280 proton donor position in calcineurin is replaced with an arginine residue. (b) Electron density map around the zinc ions

(blue spheres), bound phosphate and nucleophilic water (red sphere). Shown is an Fo-Fc simulated annealing omit map contoured at 3σ (blue mesh) and 25σ (black mesh). The zinc activated nucleophilic water molecule, which is part of the zinc octahedral coordination shell, could not be resolved separately in the electron density because of the resolution limit (2.6 Å). Thus, we based its position on the much higher resolution structure of SMPDL3A (PDB code 5FC1, 1.4 Å). (c) Top, substitution of His280 with alanine completely inactivates the enzyme, while the His317 to alanine substitution only slightly reduces activity against the non-lipid substrate bNPP. In absolute numbers, the activity of the wild-type enzyme is 1.76 μM bNPP hydrolyzed per nM protein per hour. Data are the means and standard deviations of two to five experiments performed in triplicates. Bottom, by contrast, using liposomes as substrate instead of bNPP, both His280 and His317 mutants abrogate enzymatic activity, suggesting that His317 is also important, probably for substrate binding or orientation. (d) The catalytic mechanism involves a nucleophilic attack by a zinc-activated water molecule on the phosphate group of sphingomyelin. This is followed by protonation of the ceramide leaving group by His280 with assistance from Asp249 and release of ceramide and phosphocholine.

85 Interface between saposin and catalytic domains — ASMasesap consists of four ɑ-helices and belongs to the saposin family of proteins (Saposins A, B, C and D) [82]. As observed with other saposins, ASMasesap adopts either a closed or open conformation depending on environmental conditions (Fig. 3.1B). In the closed form, ASMasesap is compact, globular and interacts mainly with the CTD (Fig. 3.1B), although this interaction appears to be a consequence of the crystal lattice since the two molecules in the asymmetric unit occupy slightly different positions, with the linker between the domains disordered in one molecule (Fig. 3.4A).

86 Figure 3.4 ASMasesap – ASMasecat interactions

(a) Superposition of the two molecules in the asymmetric unit of the closed form. The reference portion for the superposition is the catalytic domain (light green surface). The saposin domains (worm representation) are shifted by roughly 8 Å and the linker connecting the saposin domain to the catalytic domain is disordered in one of the molecules (blue). (b) Close-up of the interface formed between open ASMasesap (pink) and ASMasecat (green). Inset, yellow shading marks the surface buried upon interface formation. Active site loops in contact with the interface are coloured black. Interface mutations tested are marked by an asterisk. (c) Activity at pH 5 of wild- type (WT) and mutants on liposomes containing sphingomyelin (SM). Bar colours correspond to ASMasecat (green) or ASMasesap (pink) mutants. Activity is normalized to WT enzyme. 100% activity = 537 μM SM hydrolyzed per μM protein per hour on anionic liposomes. Data are the means and standard deviations of two independent experiments performed in triplicates. (d) Activity at pH 5 of WT and mutants on the small-molecule substrate bNPP. Activity is

87 normalized to the WT enzyme. 100% activity = 1.76 μM bNPP hydrolyzed per nM protein per hour. Data are from two to five experiments performed in triplicates. (e) Effect of detergents on activity against bNPP. Data are from two independent experiments performed in triplicates. OGP, octyl-β-D-glucopyranoside. TX-100, Triton X-100. NP-40, Tergitol-type NP-40. For latter two detergents, 0.1 mM and 1 mM represent concentrations below and above their critical micellar concentrations.

88 By contrast, the open form of ASMasesap assumes a V-shaped conformation with its hydrophobic core exposed and its position shifted on ASMasecat (Fig. 3.1B) such that one face of helix α3 forms an extensive 1318 Å2, hydrophobic interface with ASMasecat (Fig. 3.4B). To assess the significance of this interface for sphingomyelin hydrolysis, we measured the activity of interface mutants (Fig. 3.5A and 3.5B) on anionic liposomes containing sphingomyelin at acidic pH to mimic the lysosomal environment [82]. Deletion of ASMasesap almost completely abrogated activity and six interface mutants substantially reduced activity, verifying that both the saposin domain and the interface are required for membrane sphingomyelin hydrolysis (Fig. 3.4C). The interface presumably stabilizes the open conformation of ASMasesap, allowing it to dock onto membrane surfaces, and possibly deliver lipids to the active site.

89 Figure 3.5 Purification of murine ASMase and bNPP hydrolysis assay

(a) Size exclusion chromatography elution profiles of murine ASMase and point mutants. UV absorbance is normalized. The elution volume of the wild-type protein (vertical marker) corresponds to a molecular weight 55 kDa, as extrapolated from a standard curve. (b) SDS- PAGE of purified mutants. (c) Multi-angle light scattering measurements of ASMase and the isolated catalytic domain. (d) Activity measurements of bNPP hydrolysis by ASMase interface mutants with added detergent (dark grey bars). 100% activity corresponds to 1.76 μM bNPP hydrolyzed per nM protein per hour for ASMase. Addition of 0.2 mM TX-100 had no effect on the isolated catalytic domain. However, all interface mutants which initially reduced enzymatic activity due to the mutation, had their activity restored to above wild-type levels. Data are the means and standard deviations of triplicates (e) We noted that the mutations on the saposin domain, V128E and V143R, had weaker inhibitory effects than those on the catalytic domain. This is likely because V128 and V143 are also important for stabilizing the closed conformation as shown (red sticks).

90 Closer examination of the interface revealed direct contacts between ASMasesap and active site loops of ASMasecat (Fig. 3.4B), hinting at a coupling between lipid binding, interface formation and catalytic activity. We assessed this hypothesis by reverting to the small molecule bNPP substrate, which should depend only on catalytic domain attributes. Unexpectedly, we found that isolated ASMasecat had no activity compared to wild-type, and interface mutants also diminished activity on bNPP (Fig. 3.4D). Conversely, the addition of various detergents which should bind to and favour open ASMasesap, substantially increased activity on bNPP (Fig. 3.4E), even in the context of the interface mutants (Fig. 3.5D). Analogous results were obtained for human ASMase (Fig. 3.6). The structure of isolated ASMasecat is unchanged relative to the catalytic domain in the open form full-length protein, but analysis of their crystal temperature factors indicated that interface formation reduces catalytic loop dynamics (Fig. 3.7). This suggests that the interface is important not only for stabilizing open ASMasesap for membrane docking, but also for stabilizing ASMasecat and activating catalysis.

91 Figure 3.6 Purification of human ASMase and activity assays

(a) Size exclusion chromatography elution profiles of human ASMase and point mutants. UV absorbance is normalized. The elution volume of the wild-type protein (vertical marker) corresponds to a molecular weight of 57 kDa, as extrapolated from a standard curve. (b) SDS- PAGE of purified mutants. (c) Activity measurements of wild-type and mutant proteins on liposomes containing sphingomyelin (SM). Activity is normalized to the wild-type enzyme. 100% activity corresponds to 534 μM SM hydrolyzed per μM protein per hour on anionic liposomes. Data are representative of two independent experiments performed in triplicates. (d) Activity measurements of wild-type and mutant proteins on the small molecule substrate bis(p- nitrophenyl) phosphate (bNPP). Activity is normalized to the wild-type enzyme. 100% activity corresponds to 1.11 μM bNPP hydrolyzed per nM protein per hour. Data are the means and standard deviations of two to five experiments performed in triplicates. (e) Effect of detergents on activity of wild-type protein against the non-lipid substrate bNPP. Data are representative of two independent experiments performed in triplicates. OGP, octyl β-D-glucopyranoside. TX- 100, Triton X-100. NP-40, Tergitol type NP-40. For the latter two detergents, 0.1 mM and 1 mM represent concentration below and above their critical micellar concentrations. (f) Activity

92 measurements of bNPP hydrolysis by ASMase interface mutants with added detergent. Data are the means of triplicates. In all panels, error bars represent the standard deviation.

93 Figure 3.7 Main chain B-factor analysis of catalytic domain interface loops

The two histidines of the active site are displayed as sticks, and the loops at the inter-domain interface leading to these histidines are marked with asterisks. (a) In the structure with the open form of the saposin domain, catalytic domain loops have relatively lower B-factors due to the interactions with the saposin domain (white helix). (b) In the structure of the isolated catalytic domain, B-factors are relatively higher since the interface is absent. Low temperature factors are colored in blue, while high temperature factors, indicative of relative higher mobility, are shown in red. Note that the color and thickness are for the relative B-factors within each structure, and not absolute B-factor values across both structures. The range of α-carbon B-factors is 21 to 50 Å2 in (a) and 37 to 83 Å2 in (b).

94 Membrane docking — The location on open ASMasesap responsible for membrane lipid docking under physiological conditions likely includes its hydrophobic surface, which in the crystal is buried by a symmetry-related saposin domain reminiscent of the ‘handshake’ clasp observed in Saposin B dimers [79] (Fig. 3.8A). Indeed, crystallization in the presence of lipids rearranges the saposin dimer and reveals lipid molecules bound within it (Fig. 3.8B). Additionally, the top edge of helix α2, which is lined with a ridge of positively charged residues, is appropriate for interaction with anionic intra-lysosomal vesicles [82] (Fig. 3.9A). Consistent with this, mutation of Lys120 (murine Lys116) to glutamate decreases ASMase activity [89], and cationic amphiphilic drugs (CADs) interfere with ASMase membrane attachment, thereby indirectly inhibiting its activity [68]. Accordingly, inclusion of CADs in our liposomal assay dramatically reduced ASMase activity (Fig. 3.10), presumably by blocking the electrostatic interaction between helix α2 and the anionic liposomes. CADs are critical for the treatment of patients with major depressive disorder where ASMase is hyperactive [205], and can kill several types of cancer cells where ASMase activity is already debilitated [74]. However, their indirect mechanism of action makes them promiscuous and consequently, it is imperative to develop direct ASMase inhibitors with better selectivity.

95 Figure 3.8 Saposin dimers in the crystal lattice

(a) Rearrangement to the V-shaped conformation of ASMasesap exposes its hydrophobic core, which in the crystal is buried by a symmetric dimerization with another saposin domain. (b) Co- crystallization in the presence of lipid causes the dimer to shift such that a cavity is created at its center, and a molecule of octadecylphosphonic acid (ODPA) is bound between the saposin monomers. (c) Fo-Fc electron density map contoured at 2.5 σ before inclusion of the lipid molecule. Density for a second lipid molecule is also present within the dimer; however, it is fragmented and its last carbon atom overlaps with density for the first molecule, indicating likely partial occupancy of the lipid molecules in different positions. (d) Comparison of the saposin domains from the four protein chains of the lipid-bound structure (blue) and from the two chains of the apo structure (red). Proteins were superimposed via their catalytic domain (not shown).

96 Figure 3.9 Electrostatic surface and substrate binding site of ASMase

(a) Electrostatic surface representation of open ASMase contoured at +/- 3 kT. The calculation was done at pH 7 to highlight relative charge differences, which are otherwise masked at pH 5. Positively charged residues proposed to interact with anionic membrane lipids are labelled. A black line demarcates the ASMasesap:ASMasecat boundary. A green outline marks the lipid interacting surface on ASMasesap. (b) AbPA inhibitor bound to ASMase. Left, inhibitor shown as CPK spheres. Right, chemical structure of AbPA. Close-up, electrostatic interactions (red dashes) between AbPA (purple tail, orange and red phosphate) and zincs (spheres) and protein residues (green). (c) ASMase surface coloured by polar (white) to hydrophobic (dark green). A white line demarcates the ASMasesap:ASMasecat boundary. Sphingomyelin (yellow sticks) was docked manually into the active site. For comparison, the end of the AbPA lipid tail is shown as black sticks.

97 Figure 3.10 Liposomal activity assay of ASMase in the presence of cationic amphiphilic drugs (CADs)

Error bars represent the standard deviation of triplicates.

98 Inhibitor bound ASMase — To further these efforts, we determined the structure of ASMase bound to a known high affinity inhibitor, 1-aminodecylidene bis-phosphonic acid

(AbPA; IC50 20 nM) [212] (Fig. 3.9B and 3.11). One phosphate group of AbPA completes the coordination shell for the zinc ions as described above, whereas its 9-carbon lipid tail sits in a relatively featureless shallow groove. This structure and the position of the leaving group protonating His280 allowed us to propose a model for how sphingomyelin might bind to the active site (Fig. 3.9C). The fatty acid chains of sphingomyelin sit along a hydrophobic track that extends from the edge of the active site to the saposin domain and surprisingly, do not correspond to the position of the AbPA lipid tail, which instead coincides with the position of the choline head group. These results will help direct the rational design of new ASMase inhibitors that better exploit the shape of the active site.

99 Figure 3.11 Electron density and binding mechanism for the co-crystallized inhibitor AbPA

(a) Simulated annealing omit Fo-Fc electron density map contoured at 3 σ. (b) Comparison of the phosphate-bound active site of ASMase (below) vs that of AbPA-bound (above). Interestingly, in the AbPA-bound structure the phosphate group appears to be positioned as a product as if its configuration has been inverted after nucleophilic attack by the zinc-activated water molecule. Thus for AbPA, the phosphate group not only blocks the substrate binding site, but also completes the octahedral zinc coordination shell by excluding the nucleophilic water molecule. In the phosphate-bound structure, the phosphate group is positioned as a substrate with O3 pointed in a direction opposite to that of AbPA. The structures are in the identical orientation to illustrate the difference in the phosphate group between the two ligands. Zinc ions and the nucleophilic water are labeled.

100 Structural mapping of Niemann-Pick mutations — ASMase is best known for its causative role in Niemann-Pick disease types A and B. Over 130 missense mutations and short indels have been identified in the ASMase gene of patients suffering from this autosomal recessive disease. According to our structure, 103 mutations reside in the catalytic core, 19 in the CTD, 8 in the saposin domain and 4 in the Pro-rich linker (Fig. 3.12A). Most mutations are predicted to affect the fold or stability of ASMase (Table 3.2). However, about 12 are surface mutations that could affect the interaction of ASMase with membranes or other proteins, and 5 are located in the ASMasesap-ASMasecat interface, which could impair the regulatory mechanism described here. Interestingly, 80 ASMase variants of unknown significance have also been reported [49]. Half of these are surface exposed with no obvious deleterious effects; however, several are predicted to significantly affect the fold or stability of ASMase (Fig. 3.12B and Table 3.3). Further characterization of these potentially harmful mutations will have important implications for Niemann-Pick carrier screening programs [213] on higher risk populations.

101 Figure 3.12 Structural mapping of disease mutations

(a) Mapping of Niemann-Pick mutations on ASMase. Mutations are indicated by spheres coloured according to their predicted effect. (b) Locations of additional ASMase variants of unknown significance.

102 3.4 Discussion

Taken together, we propose a model whereby ASMase in solution exists in equilibrium between open and closed forms of the saposin domain (Fig. 3.13). In the absence of membranes, closed ASMasesap decoupled from ASMasecat would predominate and render the enzyme inactive. In the presence of anionic membranes, open ASMasesap becomes prevalent, docks onto the membrane surface and concomitantly forms an interface with the catalytic domain activating it for sphingomyelin hydrolysis. Our structural analysis of ASMase is an important step towards rationally designed ASMase inhibitors for treating the myriad of pathologies this enzyme mediates, predicting the effects of newly discovered mutations, as well as developing recombinant ASMase replacement therapy for hereditary diseases [204].

103 Figure 3.13 Schematic for ASMase activation

Membrane lipids that have red head groups indicate negative charges. Released products are ceramide (yellow worm) and phosphocholine (orange and blue circles).

104 3.5 Experimental procedures

Constructs — ASMase was cloned into a derivative of pFastBac™ 1 (Invitrogen). The vector contained the cleavable melittin signal peptide MKFLVNVALVFMVVYISYIYA followed by a hexahistidine tag DRHHHHHHKL. Constructs of human (RefSeq NP_000534.3) and murine (RefSeq NP_035551.1) ASMase extended from residues 84 – 611 and 88 – 615, respectively. The crystallized construct of the isolated murine ASMase catalytic domain extended from residues 165 – 627. For activity assays, isolated catalytic domain constructs encompassed residues 169 – 615 of human ASMase and 165 – 611 of its murine counterpart. All constructs and mutants were sequenced. Protein expression and purification — Recombinant baculoviruses were generated according to the Bac-to-Bac® Baculovirus Expression System protocol (Invitrogen) with minor modifications: DH10MultiBac cells were used [197] and viruses were added to Sf9 cells (Invitrogen) grown in I-Max medium (Wisent Bioproducts). Proteins were expressed at 27° C for 64 hours. Subsequent steps were carried out at 4° C. Cells were removed by centrifugation at 1,000g then at 9,000g and supernatants were incubated with HisPur™ Ni-NTA resin (Thermo Fisher Scientific). The beads were washed with buffer A (50 mM Tris-HCl pH 7.5, 500 mM

NaCl, 1 mM MgCl2) and eluted with buffer A containing 250 mM imidazole-HCl. Proteins were concentrated and loaded on a Superdex™ 200 10/300 GL size-exclusion column (GE

Healthcare) equilibrated with buffer B (15 mM Tris-HCl pH 7.5, 100 mM NaCl, 10 µM ZnCl2). Fractions containing ASMase were applied to a Mono Q™ anion exchange column (GE Healthcare) in buffer B. The flow-through was concentrated to 10 mg mL-1 and flash-frozen. Enzymatic assay with non-lipid substrate — Protein at 100 nM (1 µM for the H280A mutant) was incubated with 2 mM bis(p-nitrophenyl) phosphate in assay buffer (100 mM NaCl and 20 mM sodium acetate pH 5) at 37° C for one hour. 100 mM NaOH was then added before measuring absorbance at 405 nm. Activity was quantified by a p-nitrophenol standard curve. Liposome-based enzymatic assay — Liposomes were prepared by extrusion through 0.1 μm polycarbonate filters. Negatively charged liposomes consisted of 10% egg sphingomyelin, 55% dioleoylphosphatidylcholine (DOPC), 20% cholesterol and 15% bis(monooleoylglycero)phosphate (BMP). The Amplex® Red Sphingomyelinase Assay (Thermo Fisher Scientific) for phosphocholine detection was slightly modified. ASMase at 100 to 1000

105 nM was incubated with liposomes at 3 mM total lipids in assay buffer for one hour at 37° C. The reaction was terminated at 95° C for five minutes and an equal volume of the second step solution was added as recommended. The second step was carried out at 37° C and the change in fluorescence (560 nm excitation, 590 nm emission) after 30 minutes was used to quantify product formation with the help of a phosphocholine standard curve. Crystallization and data collection — Crystals were grown by sitting drop or hanging drop vapor diffusion at 22° C. Protein at 10 mg mL-1 in buffer B was mixed with an equal volume of well solution. Crystals of murine ASMase with the saposin domain in an open conformation (approximately 200 x 25 x 25 μm in size) were obtained in 100 mM sodium MES pH 6.5 and 1.5 M ammonium sulfate. The same protein with the saposin domain in a closed conformation was crystallized (50 x 50 x 50 μm) in 0.2 M lithium acetate and 20% PEG 3350 by first incubating it with 5 mM 1-aminodecylidene bis-phosphonic acid (AbPA, Cayman Chemical) at 22° C for 12 hours. The protein was also incubated with Triton X-100 and sphingomyelin or octadecylphosphonic acid (ODPA, Sigma-Aldrich) at 22° C for 12 hours and crystals with the saposin domain in a slightly different open conformation (100 x 100 x 25 μm) were grown in 100 mM NaH2PO4, 100 mM KH2PO4, 100 mM sodium HEPES pH 6.5 and 2 M NaCl. The same crystallization condition also yielded crystals of the isolated catalytic domain (25 x 25 x 25 μm). All crystals were briefly soaked in well solution supplemented with 20% glycerol before flash-freezing. Diffraction data were collected at 100 K on beamlines 08ID-1 equipped with a Rayonix MX300 CCD detector at a wavelength of 0.97949 Å (beam size of 130 x 30 μm), or 08B1-1 with a Rayonix MX300HE CCD detector at 1.28154 Å (230 x 195 μm) in the case of the AbPA-bound crystal, at the Canadian Macromolecular Crystallography Facility, Canadian Light Source. Data collection parameters were as follows: for the open conformation crystal, 1 s 0.45° images covering 119.7°; for the AbPA-bound crystal, 15 s 0.5° images covering 270°; for the open conformation crystal in the presence of lipid, 1 s 0.25° images covering 180°; for the catalytic domain crystal, 2 s 0.85° images covering 378.25°. Data was processed with HKL2000 [198] or XDS [214]. Structure determination and refinement — All structures of ASMase were solved with the Phaser/Phenix [199] molecular replacement package using a 1.4 Å resolution experimentally determined structure of the ASMase paralog, SMPDL3A lacking the saposin domain (PDB code 5FC1). After rebuilding in Coot [200] (Fig. 3.14) and refinement in Phenix [199] using metal

106 coordination restraints but no non-crystallographic symmetry restraints, the saposin domain could be located and built from difference electron density. The crystal of ASMase bound to ODPA had a twin fraction of 50% and was refined with the twin law h,-k,-l. Structural images were generated with PyMOL (The PyMOL Molecular Graphics System, Version 1.3 Schrödinger, LLC).

107 Figure 3.14 Representative electron density maps of the ASMase structures

A wall-eyed stereo view of the 2Fo-Fc electron density maps, contoured at 1.5 σ. (a) Isolated catalytic domain structure showing a disulfide-rich loop. (b) A glycan on the structure with the saposin domain in a closed form. (c) The open form of the saposin domain; the latter is stabilized by disulfide bonds. (d) The inter-domain interface in the structure with the saposin domain in an open form in the presence of lipid.

108 Chapter 3 concluding transition

My structural and enzymatic study of murine ASMase revealed a conformational flexibility of its saposin module, with an inactive closed form or a catalytically competent open configuration featuring a hydrophobic interface to the catalytic domain. We proposed a model of initial membrane interaction through a ridge of cationic residues lining the saposin domain, and funneling of SM molecules to the active site via a continuous hydrophobic surface. Several experiments that were not included in the article are listed here. Additional activity assays were carried out at neutral pH on neutral liposomes, or with SM in Triton X-100 micelles, with similar outcomes, highlighting the importance of the inter-domain interface. The reported inhibition of the enzyme by its C-terminal cysteine residue [103] was studied with short peptides or by x-ray absorption spectroscopy, but the results were inconclusive. Proteolytic activation of ASMase by caspase-7 [147] was also examined, as the reported Asp253 cleavage site is not surface-exposed in the crystal structures. In our in vitro experiments with caspase-7, only proteolysis between Asp220 and Cys221 was observed, with a concomitant decrease in ASMase enzymatic activity; this was not further investigated. Lastly, the protein’s conformation in solution was probed by size exclusion chromatography-coupled small-angle X-ray scattering (SEC-SAXS) at neutral and acidic pH. ASMase is monomeric and appears to adopt the closed saposin form, as expected. Around the same time my study was published, crystal structures of the human ASMase were also determined [215,216]. Both studies reported only the open configuration of the protein. Molecular docking of SM identified Asn325 as potential interactor for the hydroxyl or the amide group of the substrate, and suggested only superficial contacts between the PC head group and the enzyme [215]. Molecular dynamics simulations of the protein’s interaction with an anionic phosphatidylglycerol bilayer produced two membrane-binding modes: partial insertion of the saposin domain into the head group region, or complete presentation of its inner hydrophobic surface to the bilayer [215]. In both cases, the active site is facing the lipid head groups, but membrane contacts with the catalytic domain itself are weak and transient. The structure of ASMase in complex with PC was also reported [216]. As expected, no interactions are established with the choline moiety (Fig. 3.15), suggesting that selectivity for SM is achieved through its phosphate group and possibly hydrogen bonds with the ceramide portion.

109 Figure 3.15 PC recognition by human ASMase

Residues adjacent to the PC ligand (yellow) are displayed as sticks (PDB code 5I85).

110 Very recently, the crystal structure of SapA in complex with its cognate hydrolase, β- (GALC), was determined [217]. A SapA dimer is bound to GALC via a polar and charged interface with one SapA monomer, and a small non-essential secondary interface with the second monomer. SapA side chains contacting the enzyme are not conserved in SapB-D, thereby providing specificity to the system. A hydrophobic cavity in the saposin dimer opens onto the GALC active site; a galactosylceramide molecule was modeled based on previous galactose-bound GALC structures (Fig. 3.16). This arrangement is proposed to be the functional substrate delivery mechanism. This “solubilizer” model is in contrast with the “liftase” mode of lipid access by ASMase, which does not require complete lipid extraction from the membrane or dimerization of its saposin module.

111 Figure 3.16 Lipid delivery to GALC by SapA

The crystal structure of GALC (blue) in complex with a SapA dimer (yellow and orange) (PDB code 5NXB) is displayed. A galactosylceramide molecule (sticks) is manually docked into the hydrophobic cavity leading to the active site. Figure adapted with permission from [217].

112 Table 3.1 ASMase X-ray data collection and refinement statistics

Crystal Closed form with AbPA Open form Open form in presence of lipid Catalytic domain PDB code 5FI9 5FIB 5FIC 5HQN Data collection

Space group P1 P4212 P41 C2 Cell dimensions a, b, c (Å) 57.39, 72.02, 74.42 181.17, 181.17, 109.90 101.75, 101.75, 401.83 124.84 127.30 102.52 ɑ, β, γ () 80.47, 71.52, 83.60 90, 90, 90 90, 90, 90 90, 121.68, 90 Resolution (Å) 46.65 ‒ 2.54 (2.63 ‒ 2.54) 45.70 ‒ 2.80 (2.90 ‒ 2.80) 49.32 ‒ 2.80 (2.90 ‒ 2.80) 37.67 ‒ 2.60 (2.69 ‒ 2.60) Rmeas 0.12 (0.67) 0.13 (1.62) 0.21 (1.26) 0.19 (2.00) I/σI 9.7 (1.6) 17.3 (1.7) 10.2 (1.46) 10.0 (1.44) Completeness (%) 97.2 (90.9) 100 (100) 100 (100) 100 (100) Redundancy 2.9 (2.5) 9.7 (9.0) 6.7 (4.9) 7.7 (6.5) Wilson B-factor 38.0 40.7 48.4 52.9 Refinement Protein molecules per ASU 2 2 4 2 Resolution (Å) 46.65 ‒ 2.54 45.70 ‒ 2.80 49.32 ‒ 2.80 37.67 ‒ 2.60 Number of unique reflections 33,629 39,766 99,738 41,929 Rwork / Rfree 20.2 / 25.4 18.8 / 23.9 19.0 / 23.9 21.3 / 25.8 Number of atoms Protein 8,294 8,474 16,675 7,050 Glycans, ligands, ions 410 521 286 226 Water 54 30 39 74 B-factors Protein 41.7 41.9 57.5 51.2 Glycans, ligands, ions 64.4 96.2 76,1 88.8 Water 29.6 25.4 31.6 44.5 Ramachandran statistics Favored (%) 93.3 94.0 95 94.5 Allowed (%) 6.6 5.8 4.9 5.5 Outliers (%) 0.1 0.2 0.1 0.0 Rotamer outliers (%) 0.8 0.6 0.4 0.8 R.m.s deviations Bond lengths (Å) 0.003 0.003 0.004 0.005 Bond angles (º) 0.72 0.82 0.72 0.67

Values for the highest resolution shell are shown in parentheses.

113 Table 3.2 Predicted effects of ASMase mutations found in Niemann-Pick patients

Mutations that disrupt the active site

H321Y D280A H427R H461P D253E D253H

Mutations that alter the saposin domain hydrophobic surface or its interface with the catalytic domain

L105P V114M L163P F392del V132A L139P P325A W393G

Mutations that prevent disulfide bond formation or glycosylation

C91H C94W C159R C228R C387R C433R C596Y N522S

N522D

Mutations predicted to affect the folding or stability of the protein

F482del T594del R610del G168R I178N A198P R202C W211R

L227M L227P R230C R230H A243V G244R W246C G247D

G247S S250R T258I A283T R291H Q294K L304P V314M

Y315H P332R W342C L343P A359D A359V Y369C P373S

R378H R378L S381F S381P L382F M384I N385K N385S

N391T A415V H423R H423Y P429L P430S L434P W437C

W437R S438R Y448C T451P L452P A453D (G)A454V G458D

T460P F465S Y469S R476W R476Q P477L F482L F482S

A484E A484V S486R T488A (F)Y490N G496C G496S R498C

R498H R498L Y500H H516Q E517V Y519C P533L W535R

Y539H A541T (M)L551P V559L F572L H577L H577R H577D

K578N G579S Q598R L599F R602H R602P S436_W437dup

Surface mutations without clear deleterious effect

P186L G234D E248K E248Q R296Q G336S E471K (N)G492S

Q525H H556Y D565Y R610C

Mutations were classified according to their predicted effect, based on the ASMase structure. When the human and murine amino acid sequences differ, the latter is indicated in parentheses.

114 Table 3.3 Predicted effects of ASMase variants of unknown significance

Variants that prevent disulfide bond formation or glycosylation

Probably harmful: C433W (P)S507F

Variants that alter the saposin domain hydrophobic and positively charged surfaces or its interface with the catalytic domain

Probably benign: (L)I101V V117M (A)I136V R113H

Uncertain: A158T R113C V318E T324I V132M

Probably harmful: V145del P325delinsLS

Other non-surface variants

Probably benign: N102S G115S L256P G270S (I)V301I P331S

A346V (R)Y374F A446V N450S A487V* Y519F I520V (A)T550A

V559I R591H L599I

Uncertain: L266V A269V P282T R296W A297V V305M

P332L R378C V409G P430A E449Q (G)A454S V512M R591C

Probably harmful: L379F S510F

Surface variants, probably benign

S192del G104R R150C (T)S175P P187S (Q)R240Q (F)Y245H (Q)H288D

(S)T290N (D)A303T P313S (Q)R341C (Q)R341H (Q)R341P E352D (H)R361C

(H)R361H R389C (K)R443G (K)R443Q E472D G508R* G530A R542Q

N557K R561H D565N P582S S583L T588M R610H M613I

Variants were classified according to their predicted effect, based on the ASMase structure. Reported polymorphisms are marked by an asterisk. When the human and murine amino acid sequences differ, the latter is indicated in parentheses.

115 Chapter 4 – Crystal Structure of the Acid Sphingomyelinase-Like Phosphodiesterase SMPDL3B Provides Insights into Determinants of Substrate Specificity

The crystal structures of SMPDL3A and ASMase revealed a relative lack of selectivity towards the nucleobase or choline moieties of their respective substrates. Instead, other characteristics of the proteins, such as localization and membrane-interacting features, appear to play important roles in determining the types of molecules that access the active site. The third ASMase-like protein, SMPDL3B, is a GPI-anchored enzyme on the cell surface, but its physiological substrate is unknown – although a lipid-modifying function was suggested. Here I present my structural and enzymatic study of SMPDL3B.

116 4.1 Abstract

The enzyme acid sphingomyelinase-like phosphodiesterase 3B (SMPDL3B) was shown to act as a negative regulator of innate immune signaling, affecting cellular lipid composition and membrane fluidity. Furthermore, several reports identified this enzyme as an off-target of the therapeutic antibody rituximab, with implications in kidney disorders. However, structural information for this protein is lacking. Here we present the high-resolution crystal structure of murine SMPDL3B which reveals a substrate binding site strikingly different from its paralogs. The active site is located in a narrow boot-shaped cavity. We identify a unique loop near the active site which appears to impose size constraints on incoming substrates. A structure in complex with phosphocholine indicates that the protein recognizes this head group via an aromatic box, a typical choline-binding motif. Although a potential substrate for SMPDL3B is sphingomyelin, we identify other possible substrates such as CDP-choline, ATP and ADP. Functional experiments employing structure-guided mutagenesis in macrophages highlight amino acid residues potentially involved in recognition of endogenous substrates. Our study is an important step towards elucidating the specific function of this poorly characterized enzyme.

117 4.2 Introduction

Sphingomyelins (SM) represent a major class of phospholipids found in cellular membranes and are particularly enriched in the outer leaflet of the plasma membrane [218]. Removal of the phosphocholine (PC) head group from SM leads to the generation of ceramide sphingolipids, whose resting levels in the plasma membrane are extremely low, and which have been shown to alter biophysical properties of the membrane and affect cellular processes [29]. This reaction is carried out by enzymes called sphingomyelinases (SMases). In mammals, three SMase families can be distinguished: The alkaline SMase is expressed on microvilli in the intestines, where it metabolizes dietary SM [219]. The four neutral SMases are localized in the plasma membrane inner leaflet, Golgi, endoplasmic reticulum, nucleus or mitochondria [33]. The acid SMase (ASMase, ASM, SMPD1) is found in the lysosome where it serves catabolic purposes, and can also be secreted to act on plasma lipoproteins or the cell membrane [109]. This latter function occurs in response to numerous stresses and modulates signaling events via ceramide production [109]. ASMase is the founding member of a family of three paralogs, along with the relatively poorly characterized enzymes SMPDL3A and SMPDL3B. Interestingly, SMPDL3A is not a SMase but hydrolyzes various nucleotide triphosphates and diphosphates [27]. In this regard, an anti-inflammatory role in purinergic signaling in the extracellular space was proposed for SMPDL3A [27], although it was also reported to be found in lysosomes [192]. However, its function is still linked with lipid metabolism, since its expression and secretion is stimulated by cholesterol and liver X receptor ligands in macrophages [27,174]. On the other hand, SMPDL3B is a glycosylphosphatidylinositol (GPI)-anchored protein [220,114,26] with reported roles in inflammatory processes as well as in kidney diseases. A negative regulatory role in innate immune signaling was uncovered in murine macrophages, where the enzyme is found on the cell surface and upregulated upon inflammatory stimuli [26,135]. Its knockdown results in enhanced responsiveness to Toll-like receptor (TLR) stimulation and Smpdl3b-deficient mice display an increased inflammatory response. Furthermore, deficiency of this protein in macrophages is associated with alterations in the cellular lipid composition and changes in membrane fluidity.

118 In the kidney disease focal segmental glomerulosclerosis expression of SMPDL3B on podocytes is reduced; overexpression of the enzyme prevents actin cytoskeleton disruption and apoptosis associated with this condition [115]. Rituximab, a therapeutic anti-CD20 monoclonal antibody targeting B cells but also used in certain nephrological conditions, binds to SMPDL3B [115,148,149] and remediates these defects in a SMPDL3B-dependent manner [115]. Rituximab similarly protects podocytes from morphological disruption, decreased proliferation and SMPDL3B protein loss in xenotransplants [148]. In contrast, this protein is elevated in glomeruli from patients with diabetic kidney disease where it binds to the soluble urokinase plasminogen activator receptor on podocytes; SMPDL3B knockdown protects podocytes from apoptosis in this condition [151]. The identification of this enzyme at other locations suggests additional functions. For instance, SMPDL3B has been found in pancreatic zymogen granules, an organelle containing digestive enzymes to be released into the pancreatic duct [155]; in saliva exosomes which are secreted vesicles with various functions [156]; in the plasma protein-depleted cerebrospinal fluid [154]; and in milk [153]. While these findings provide clues as to the potential roles of SMPDL3B, its precise function and substrate specificity remain unknown. That it is membrane-anchored, bears similarity to ASMase, and possesses membrane-modulating activities, suggests that it may act on lipid substrates. On the other hand, its other homolog, SMPDL3A, is a nucleotide phosphodiesterase. The crystal structures of ASMase and SMPDL3A were recently reported [201,211,215,221], but no structural information is as yet available for SMPDL3B. Here we present the high resolution crystal structure of murine SMPDL3B, which revealed that its substrate binding site is a narrow pocket, quite distinct from its paralogs, a consequence of a unique conformation of a surface loop. The structure of the enzyme in complex with PC showed that this potential product is accommodated inside a narrow pocket surrounded by a “cage” of aromatic side chains, a common feature of choline-binding proteins [193]. Although no in vitro enzymatic activity against lipids could be detected, SMPDL3B can cleave other choline-containing substrates including CDP-choline, and releases phosphate from ATP and ADP. Functional experiments in murine macrophages further revealed amino acid residues found on the channel leading to the active site of SMPDL3B as potentially important for

119 recognition of endogenous substrate molecules involved in the modulation of macrophage function.

120 4.3 Results

Overall structure — We determined the crystal structure of full-length murine SMPDL3B, which shares 78% identity with its human counterpart. Our construct excluded the N-terminal signal peptide as well as the C-terminal GPI attachment signal peptide which is normally cleaved off to generate a new C-terminus – the ω-site of GPI attachment. The ω-site is Ala-431 for human SMPDL3B [114], corresponding to Gly-431 in the murine protein. Although the crystallized construct contained four additional residues beyond the ω-site (residues 19 to 435), Gly-431 is the last residue visible in the crystal structure. The structure was solved by single-wavelength anomalous diffraction from two zinc ions in the active site and refined to a resolution of 1.14 Å (Table 4.1). The enzyme comprises a catalytic domain affixed with a small C-terminal subdomain (CTD) (Fig. 4.1). The catalytic domain consists of two central mixed β-sheets of six strands each, surrounded by eight α-helices. This architecture classifies SMPDL3B in the calcineurin- like phosphoesterase structural superfamily (PFAM code PF00149). One end of the β-sandwich has a wide opening, against which the CTD, consisting of a globular cluster of four α-helices, is packed up against. The fourth helix of the CTD is terminated by Gly-431. The CTD appears to be a defining feature of the ASMase-like proteins as it is also found in ASMase [109] and SMPDL3A [27], but not in other phosphoesterases. It displays lower sequence conservation than the catalytic domain and its exact function is unknown, but it may enhance the stability of these proteins in their lysosomal and extracellular environments.

121 Figure 4.1 Crystal structure of SMPDL3B

The catalytic domain is shown in green and the C-terminal subdomain, in blue. Zinc ions are represented by gray spheres. N-linked glycans (white sticks) are only partially displayed for clarity. Disulfide bonds are shown as yellow sticks. The N-terminus and the putative C-terminal ω-site of GPI anchor attachment are indicated.

122 The protein fold is stabilized by a total of three disulfide bridges - two in the CTD and a third at the base of a large loop in the catalytic domain. Although murine SMPDL3B has five potential sites of N-linked glycosylation, only three glycans are discernable in the structure. The two other sites are located on the surface of the protein suggesting that they potentially can be glycosylated. Human SMPDL3B shares only two of these five sites and has a third potential one, also on the surface of the protein. Structural superimposition of SMPDL3B with the ASMase catalytic domain and with SMPDL3A (Fig. 4.2A) shows that the overall fold and all secondary structure elements are well conserved within this subfamily. The main differences arise in the lengths and compositions of several loops on the “top” face of the proteins which harbors the active site.

123 Figure 4.2 Structural comparison of SMPDL3B, SMPDL3A and ASMase

(A) Superimposition of the ASMase family of proteins. SMPDL3B is shown in green, SMPDL3A, in blue (PDB code 5FC6), and ASMase, in yellow (PDB code 5HQN). Loops on the face of the protein containing the active site that diverge between the family members are displayed as thick lines. The loop L9-10 of SMPDL3B (described later) is marked with an asterisk. SMPDL3B shares 28% identity with ASMase (root-mean-square deviation of 0.851 Å over 284 alpha carbons) and 38% identity with SMPDL3A (root-mean-square deviation of 0.606 Å over 295 alpha carbons). The saposin domain of ASMase is not shown. (B) Comparison of the SMPDL3B active site with that of SMPDL3A and ASMase. Active site residues of SMPDL3B are shown as green sticks. The active site of SMPDL3A (transparent blue sticks) and ASMase (transparent yellow sticks) are overlaid. Zinc ions are represented by gray spheres. The bridging water molecule is displayed as a red sphere. Distances are indicated (Å).

124 Active site and substrate recognition — At the core of SMPDL3B are found two zinc ions coordinated by the side chains of seven residues located on loops extending from the central β-sheets (Fig. 4.2B). Two histidines, His-135 along with the His-97 – Asp-65 pair, complete the active site of the enzyme. All of these residues are perfectly conserved in SMPDL3A and ASMase (Fig. 4.2B), suggesting that all three family members utilize the same general catalytic mechanism for substrate hydrolysis. Briefly, a phosphate-containing substrate would bind in the active site mainly via strong electrostatic interactions between the zinc ions and the phosphate moiety. A zinc-activated nucleophilic water molecule would then attack the phosphate, resulting in departure of the leaving group, protonated by one of the neighboring histidines [201,211,215,221]. The nucleophilic water molecule coordinated midway between the two zinc ions is clearly visible in the high-resolution structure of SMPDL3B (Fig. 4.2B). Although the three phosphodiesterases likely employ the same catalytic mechanism, the identity of the phosphate-containing substrate varies widely within this subfamily. ASMase hydrolyzes SM to ceramide and PC. The related lipid phosphatidylcholine, bearing the same PC head group, may also serve as substrate to a lower extent [109]. On the other hand, SMPDL3A is inactive against SM and instead hydrolyzes various nucleotide triphosphates and diphosphates [27]. The physiological substrate of SMPDL3B, however, is not known. Interestingly, despite the overall structures of the three enzymes being highly conserved (Fig. 4.2A), the shapes of the regions surrounding their active sites are markedly different (Fig. 4.3A). The structure of ASMase in complex with sulfate [221] displays a wide depression emanating from the sulfate and zinc ions at the active site. In contrast, murine SMPDL3A [211] accommodates its nucleotide substrate in a roughly T-shaped cleft, which restricts the type of substrates that can enter its active site. Human SMPDL3A displays a similar T-shaped configuration [201]. In all cases, the regions adjacent the active site of these proteins are surface exposed depressions.

125 Figure 4.3 Region around active site

(A) The pocket at the active site of SMPDL3B (green). Zinc ions are represented by black spheres. The substrate binding site of SMPDL3A in complex with a nucleotide (blue, PDB code 5FC6) and ASMase bound to phosphate (yellow, PDB code 5HQN) are shown for comparison. (B) Boot-shaped cavity at the active site of SMPDL3B. (C) The loop L9-10 of SMPDL3B (green) is compared with the corresponding loops of SMPDL3A (blue) and ASMase (yellow). Sequences of the loops of human and mouse homologs are aligned. (D) Conformational differences between the L9-10 loop of SMPDL3B and SMPDL3A are highlighted.

126 Strikingly, the structure of SMPDL3B revealed a relatively closed off active site with a narrow boot-shaped cavity accessible through a small circular opening on the protein surface (Fig. 4.3, A and B). The active site zinc ions are located at the heel of the boot indicating that substrates must reach deep into this cavity. The unique nature of the SMPDL3B substrate binding site is mainly due to a loop insertion between strand β9 and strand β10 (L9-10), which assumes a distinct conformation relative to its paralogs (Fig. 4.3C), and forms the “tongue” of the boot-shaped substrate-binding cavity. In ASMase this insertion is essentially absent, leading to its more expansive substrate binding site. In SMPDL3A, this region is shorter by three residues and is pinned back by hydrophobic interactions between Val-322 at the tip of the loop and Ala-65 on a neighboring loop (Fig. 4.3D). This creates space for the accommodation of nucleobases in this region of SMPDL3A. In SMPDL3B, however, the sequence of L9-10 differs from SMPDL3A and this loop is oriented toward the substrate binding site, leaving a small, 9 Å opening for substrate entry. The shape of the substrate binding site provides some clues as to the nature of the phosphate-containing substrate of this enzyme. Given that SMPDL3B could potentially act on lipids at the membrane surface [26], we wondered whether it can accommodate the small PC head group of SM. To assess this, we crystallized the protein in presence of PC and found that it is indeed located in the substrate binding cavity (Fig. 4.4A). Its phosphate group forms numerous electrostatic interactions with the zinc ions and several active site residues as observed previously in structures of ASMase and SMPDL3A (Fig. 4.4B). The choline moiety extends into the cavity and interacts with the backbone carbonyl of His-277, and the side chains of Leu-310 and Thr-308, which emanate from L9-10 of SMPDL3B. This space is occupied by a glycerol molecule from the crystallization condition in the apo structure. Additionally, the choline moiety is surrounded by five aromatic residues: Phe-242, Trp-250, Trp-306, His-277 and His-279 – the latter two of which also participate in zinc coordination. This arrangement of several aromatic residues within 4 – 5 Å of the cationic quaternary amine has been termed a “cation-π box” and is a common choline recognition motif [193].

127 Figure 4.4 Phosphocholine bound in the active site of SMPDL3B

(A) PC (sticks) in the active site of SMPDL3B, in proximity to loop L9-10 (pink). (B) Residues that interact with the choline group of PC are shown as sticks (left). Distances between the choline nitrogen atom and the closest protein non-hydrogen atoms are indicated (Å). Residues that form contacts with the phosphate portion of PC, including zinc ions and water, are shown as sticks (right). Inter-atomic distances are indicated (Å). The electron density Fo – Fc simulated annealing omit map around PC and the nucleophilic water molecule is contoured at 3σ.

128 The dimensions of the PC group just about match the size of the substrate binding site opening, suggesting that it may have to adjust slightly to allow PC access to the buried active site. Comparison of the PC-bound structure with the apo form, however, revealed no conformational changes, although binding of PC seems to increase the disorder in L9-10 as interpreted from higher crystal temperature factors (an increase by 18 – 64 % in alpha carbon temperature factors normalized to the entire protein). Thus, L9-10 may have some malleability to allow entry of small substrates like PC, or it can perhaps move out of the way altogether to accommodate larger substrates. To assess these possibilities, we turned to mutagenesis and enzymatic assays. Enzymatic activity in vitro — It had been previously shown that SMPDL3B possesses phosphodiesterase activity against the artificial generic substrate bis(4-nitrophenyl) phosphate (bNPP) [26]. Based on its relatedness to ASMase and its effects on cellular lipid composition and membrane biophysical properties [26], we initially sought to study the activity of the purified enzyme in vitro against PC-containing lipids. Interestingly, no head group hydrolysis was detected with the following substrates: SM in Triton X-100 micelles, SM in neutral liposomes (also comprising phosphatidylcholine and cholesterol), the soluble micelle-forming lipids sphingosylphosphorylcholine (lyso-SM) and platelet-activating factor. These assays, however, were carried out with the soluble form of the protein used for crystallization that lacks the GPI moiety responsible for anchoring SMPDL3B to the cell surface [220,114,26]. Thus, it may be that the membrane tethered form of the enzyme is capable of hydrolyzing lipids. On the other hand, SMPDL3B was able to cleave CDP-choline (Fig. 4.5C), an essential building block for the biosynthesis of PC-containing phospholipids in the Kennedy pathway [222]. Additionally, as with SMPDL3A, we found that SMPDL3B can release phosphate from ATP and ADP (Fig. 4.5, D and E), but not from AMP.

129 Figure 4.5 In vitro enzymatic activity against various substrates

The activity profiles of purified SMPDL3B are shown for the wild type protein (solid line) as well as for loop swaps of L9-10 by glycine, GG or GSG (dashed lines). The chemical structures of substrates are depicted. In (F), activity of purified murine ASMase [221] and SMPDL3A [211] is also shown. Values are the mean and standard deviation of triplicates representative of one of two experiments.

130 To assess the role of L9-10 in substrate binding, this loop was replaced by a glycine residue, GG or GSG. However, none of these modifications enabled hydrolysis of lipid substrates. Notably, shortening of L9-10 did increase activity against CDP-choline as well as bNPP and the artificial substrate p-nitrophenylphosphorylcholine (NPPC) (Fig. 4.5, C, A and B), suggesting that it does regulate access to the active site. In theory, CDP-choline could be recognized by the protein either with its choline group positioned at the bottom of the cavity, as in the PC bound structure (Fig. 4.4A), or in the reverse orientation with the nucleotide moiety occupying that location, as occurs with SMPDL3A substrates (21, 22). The latter would require a stable rearrangement of L9-10 since the cytidine moiety from CDP-choline is too large to fit in the observed substrate cavity. To ascertain between these possibilities, the enzyme was challenged with diadenosine tetraphosphate (AP4A), a substrate with nucleosides at both ends of a tetraphosphate group. Purified SMPDL3A can cleave this molecule, but SMPDL3B is completely inactive against it (Fig. 4.5F). Moreover, shortening of L9-10 did not allow hydrolysis of AP4A (although the “G” loop-swap did have a very low rate of activity). This suggests that L9-10 cannot stably rearrange into a different conformation to accommodate larger substrates like nucleosides. In the case of CDP-choline, its PC moiety would be bound to the active site as observed in the PC bound structure, and for ATP/ADP, only the terminal phosphate would enter the active site. Thus, L9- 10 appears to act as a relatively fixed selectivity filter, that through slight ‘breathing motions’, can allow entry of smaller substrates like PC and isolated phosphate ends. Enzymatic activity in cells — The observed activity on nucleotides, including CDP- choline, and absence of knowledge of the precise physiological substrate, lead us to examine other residues besides the aromatic cage and the L9-10 loop that may interact with the substrate. The walls of the channel leading to the active site are lined with amino acids involved in catalysis but also with four other residues (Lys-140, Asn-141, Tyr-198 and Asn-200, Fig. 4.6A), two of which – Lys-140 and Tyr-198 – are highly conserved in vertebrates. We introduced mutations in these residues, as well as in His-135, one of the conserved amino acids contacting the substrate’s phosphate group. These mutations should not disrupt the fold of the protein, but could potentially abrogate polar contacts with the physiological substrate. SMPDL3B was reported to negatively regulate TLR signaling in murine macrophages, affecting the release of pro-inflammatory cytokines from these cells [26]. In order to test the impact of the mutants in

131 such an experimental system, we generated a series of stable RAW264.7 macrophage cell lines, overexpressing an empty vector control (mock), wild type or mutant forms of SMPDL3B. Western blot analysis revealed that introduction of all SMPDL3B constructs lead to similarly increased protein expression as compared to the endogenous level observed in control cells (mock) (Fig. 4.6B). The cell lines were subsequently stimulated with the TLR ligand lipopolysaccharide (LPS) and evaluated for cell-associated enzymatic activity towards the experimental substrate bNPP as well as for release of the TLR-induced cytokine interleukin-6 (IL-6). Overexpression of wild type SMPDL3B lead to an increase in cell-associated enzymatic activity towards bNPP (Fig. 4.6C) and concomitantly to a significant reduction of IL-6 release in comparison to control (Fig. 4.6D), confirming its negative impact on TLR-dependent processes. Individual mutants differed in their ability to hydrolyze the experimental substrate bNPP, with KN140MA and H135A almost completely failing to increase detectable activity (Fig. 4.6C). Interestingly, these are also the only two mutants tested, that failed to significantly reduce LPS- induced IL-6 release from RAW264.7 cells in this experimental system (Fig. 4.6, D and E). The molecular position of His-135, Lys-140 and Asn-141 in relation to the PC group in the structure, along with the functional data, suggest that these residues are indeed important features for recognition and binding of endogenous substrates affecting the activity of macrophages.

132 Figure 4.6 Functional impact of SMPDL3B mutants on LPS-induced IL-6 release in macrophages

(A) Residues lining the wall of the channel leading to the active site are displayed in orange along with their degree of conservation in vertebrates. Other adjacent amino acids, including the catalytic machinery, are in green. (B) Western blot analysis for SMPDL3B, HA and Tubulin in samples from RAW264.7 macrophages stably transduced with an empty vector (mock), or HA- tagged wild type or mutant murine SMPDL3B constructs as indicated. (C, D) RAW264.7 macrophages stably transduced with an empty vector (mock), or HA-tagged wild type or mutant murine SMPDL3B constructs were stimulated with 100 ng/ml LPS for 8h. Cells were probed for enzymatic activity by incubation with bNPP (C), supernatants were analyzed for IL-6 release by ELISA (D). Experimental data were normalized to mock and represent mean ± SEM from five independent biological replicates. Statistical significance was tested using a one sample t test against a hypothetical mean value of 1. * P ≤ 0.05; n.s. not significant. (E) Relative enzymatic activity and relative IL-6 release shown in (C, D) plotted against each other.

133 4.4 Discussion

Structure and activity of SMPDL3B — The overall structure of SMPDL3B determined here is similar to that of its paralogs SMPDL3A and ASMase. All three possess a highly conserved active site that hydrolyzes phosphate-containing substrates via a zinc-activated water molecule. Despite this, differences in protein localization and the shape of the substrate binding site appear to confer functional differences in these proteins. ASMase has an additional N- terminal saposin domain – a small membrane- and lipid-binding module which enables it to act on its SM substrate on the outer leaflet of the plasma membrane, on intra-lysosomal vesicles, on lipoprotein particles in the circulation and at other locations [109]. Additionally, ASMase has a relatively wide and accessible substrate binding site with optimal activity coupled to the conformation and position of the saposin domain relative to the catalytic domain [221]. Conversely, SMPDL3A is not known to be membrane-interacting and has a T-shaped substrate binding site. These attributes alter its substrate specificity to soluble nucleoside- diphosphates and triphosphates and implicate it in purinergic signaling pathways [27]. In contrast to its paralogs, SMPDL3B is constitutively attached to the cell membrane exoplasmic face via a GPI anchor [220,114,26] and the structure reveals that its substrate binding site is different from either paralog. An extended and conformationally distinct loop, L9-10, closes off a large portion of the surface near the active site, leaving a small restricted opening. This configuration of the active site might suggest an altogether different substrate specificity for SMPDL3B from either ASMase or SMPDL3A. However, the point of GPI attachment of SMPDL3B is located on the “top” face of the enzyme (Fig. 4.1) and places the active site opening in proximity to the membrane lipid head groups. Moreover, previous studies implicated SMPDL3B in TLR signaling, where its action on macrophage membranes modulated lipid composition, affecting the levels of SM, other PC-based lipids, and ceramide, concomitantly altering membrane fluidity [26]. This notion is further bolstered by the structure of SMPDL3B bound to PC. The small molecule head group of SM binds in the tight cavity of the active site, and the chemical surroundings are appropriate for choline recognition as multiple aromatic residues around the quaternary ammonium group establish cation-π interactions with it. A more anionic environment would have been expected for phosphoethanolamine or other cationic or polar head groups,

134 which would be inappropriate for choline since it cannot form hydrogen bonds or salt bridges with protein residues. The difference in conformation of L9-10 from the corresponding loops in SMPDL3A or ASMase, and that it was slightly more mobile in the presence of PC, raised the possibility of a switch-like regulatory role for this segment. Although replacement of L9-10 does increase activity on some substrates tested, there is no indication that a stable conformational change occurs, at least in solution. Instead, the conformation of L9-10 appears to be more or less fixed, allowing entry of only appropriately sized substrates through small breathing motions. Alternatively, the conformation of L9-10 may be impacted at the membrane by presence of lipid head groups, since this loop is on the side of the protein likely to face the membrane. This could also apply to the other loops that diverge between SMDPL3B and its paralogs and that are found on the same face (Fig. 4.2A). Based on the crystal structure, we were able to identify amino acid residues potentially required for recognition of endogenous substrates and could evaluate their importance for SMPDL3B-dependent effects in a cellular model of LPS-induced IL-6 release. In addition to His-135, a conserved residue mediating interactions with the phosphate group, we found that a mutant affecting both Lys-140 and Asn-141, two amino acids situated in the channel leading to the active site of SMPDL3B, failed to significantly reduce LPS-induced IL-6 release. The location of these amino acids relative to PC seen in the structure, along with the functional information, makes it likely that these residues indeed specifically participate in substrate recognition. Taken together, these findings suggest that SM with its small PC head group is a viable candidate substrate for SMPDL3B. Manual modeling of SM based on previous docking of SM onto ASMase [215,221] shows that in both cases, SM is positioned similarly and can insert into the active site with few clashes (Fig. 4.7A). With SMPDL3B, only the headgroup of SM would be stably bound, while with ASMase, portions of the ceramide moiety may also interact with the protein with the aid of is accompanying saposin domain. While we did not observe any in vitro hydrolysis of SM in detergent micelles or liposomes, or of other soluble lipids even when the loop L9-10 is replaced, this is likely because the protein requires proper attachment to membranes for substrate hydrolysis to occur.

135 Figure 4.7 Proposed binding modes of potential substrates

(A) SM (dark gray sticks) was manually placed in the substrate binding site using bound PC as a guide. A GPI anchor (stylized gray lines) is attached to the putative C-terminal ω-site. The surface of the plasma membrane is represented by a gray plane. (B) Comparison of potential substrate positioning in SMPDL3B and ASMase for SM, and in SMPDL3B and SMPDL3A for nucleotide triphosphates. Zinc ions (gray spheres) and the nucleophilic water molecule (black sphere) are shown.

136 Alternatively, the physiological substrate could in fact be a non-lipid molecule. We found that the enzyme can hydrolyze CDP-choline, the activated metabolite of PC used in lipid biosynthesis, in vitro. Even though this molecule is usually found intracellularly, it might encounter SMPDL3B under specific physiological conditions such as inflammatory tissue damage or digestive processes [223]. However, the affinity and activity rate of the protein on CDP-choline are rather low (Fig. 4.5C). We also found activity against ATP and ADP. In all of these cases, though, the means by which substrate molecules are hydrolyzed would be different from SMPDL3A, which also acts on nucleoside diphosphates and triphosphates (Fig. 4.7B). In SMPDL3A, during the hydrolysis of ATP or ADP, the phosphate preceding the terminal phosphate (β-phosphate for ATP, α-phosphate for ADP) must be coordinated by the zinc ions for attack by the nucleophilic water molecule, and the terminal phosphate (γ-phosphate for ATP, β- phosphate for ADP) is the leaving group due to the constraints imposed by the position of the protonating histidine side chain and orientation of the nucleobase in the T-shaped substrate binding site [211]. Conversely, in SMPDL3B, because of the block imposed by L9-10 preventing accommodation of large moieties like complex glycolipids or nucleobases inside the cavity, ATP or ADP phosphate release likely proceeds in a “tail-first” manner where the terminal phosphate contacts the zinc ions in the active site for nucleophilic attack, while the rest of the molecule remains in the channel leading to them and is the leaving group upon hydrolysis. The suitability of nucleotides as physiological substrates of SMPDL3B, however, is unclear due to their weak, millimolar Km (Fig. 4.5, D and E) and relatively low activity rate - about tenfold slower than SMPDL3A at neutral pH [211]. In summary, the structure of murine SMPDL3B highlights molecular details important for catalysis and substrate recognition. Knowledge of the molecular details will further contribute to the understanding of which substrates are bound and hydrolyzed under physiological conditions and, as a consequence, how SMPDL3B influences cellular properties such as lipid composition or inflammatory signaling. The tight cavity forming its substrate binding site, differing from its paralogs, also provides a promising feature for the identification of enzyme-specific inhibitors that could be exploited for modulation of SMPDL3B-dependent functions in relevant experimental systems, including immunity and kidney biology.

137 4.5 Experimental procedures

Constructs — For expression in Sf9 insect cells, murine SMPDL3B (residues 19-435; UniProt code P58242) was subcloned into a derivative of pFastBac 1 (Invitrogen). The vector contained the melittin signal peptide MKFLVNVALVFMVVYISYIYA followed by a hexahistidine tag DRHHHHHHKL. Swaps of the loop L9-10 consisted of residues 307-316 (KTTLPGVVDG) replaced by a glycine, GG or GSG. Expression constructs for mammalian cells were based on a derivative of the entry vector pEntry1A (Invitrogen) encoding the signal peptide of influenza hemagglutinin (HA) followed by an HA tag and the cDNA for murine SMPDL3B (residues 19-456) as reported previously [26]. Individual point mutations corresponding to H135A, K140M, N141A, KN140MA, Y198F, N200A and YSN198FSA were introduced by site-directed mutagenesis using the QuikChange II system (Agilent). For retroviral gene transduction, cDNAs were shuttled to a derivative of pMSCV-GW containing a blasticidin resistance gene using the Gateway LR Clonase (Invitrogen) system. All constructs and mutants were verified by sequencing. Protein production — Soluble SMPDL3B was expressed as a secreted protein in Sf9 cells and purified as previously described for SMPDL3A [211]. The protein eluted as a monomer in size exclusion chromatography. Crystallization and data collection — Soluble SMPDL3B at 10 mg mL-1 in buffer (15 mM Tris-HCl pH 7.5, 100 mM NaCl) was mixed with an equal volume of well solution consisting of 0.2 M NH4NO3 and 25% PEG 3350. Crystals were grown by hanging drop vapor diffusion at 22 °C. For anomalous diffraction experiments, crystals were soaked in well solution supplemented with 10 mM ZnCl2. PC-bound crystals were obtained by pre-incubating and crystallizing the protein with 20 mM PC. Diffraction data were collected at 100 K on beamline 08ID-1 at the Canadian Macromolecular Crystallography Facility, Canadian Light Source, and on beamlines MacCHESS A1 and F1, Cornell High Energy Synchrotron Source. Data were processed with HKL2000 [198]. Structure determination and refinement — The structure of SMPDL3B was solved by zinc single-wavelength anomalous diffraction using Autosol in Phenix [224,199] and manually rebuilt in Coot [225]. The PC-bound structure was obtained by molecular replacement using Phaser [226] in Phenix. Refinement was performed with phenix.refine [227]. Data processing

138 and refinement statistics are presented in Table 4.1. Structural images were generated with PyMOL (The PyMOL Molecular Graphics System, Version 1.3 Schrödinger, LLC). In vitro enzymatic activity assays — Proteins for activity assays were purified as described above, but the final buffer included 10 μM ZnCl2. To detect hydrolysis of the choline- containing substrates SM in Triton X-100 micelles, SM in neutral liposomes (also comprising phosphatidylcholine and cholesterol), sphingosylphosphorylcholine (lyso-SM), platelet- activating factor and CDP-choline, the Amplex Red sphingomyelinase assay (Thermo Fisher Scientific) was slightly modified. Proteins at 5 to 25 nM for CDP-choline or up to 10 μM otherwise were incubated with substrate in buffer (20 mM Tris-HCl pH 7.5, 100 mM NaCl) for one hour at 37 °C. The reaction was terminated at 95 °C for five minutes and an equal volume of the second step solution was added as recommended. The second step was carried out at 37 °C and the change in fluorescence (560 nm excitation, 590 nm emission) after 30 minutes was used to quantify product formation with the help of a PC standard curve. To detect hydrolysis of bNPP and NPPC, proteins at 1 to 50 nM were incubated with substrate in buffer (100 nM NaCl, Tris-HCl pH 7.5 at 20 mM for NPPC or 100 mM for bNPP) for thirty minutes at 37 °C. NaOH was added to 100 mM before measuring absorbance at 405 nm. Product formation was quantified with a p-nitrophenol standard curve. Enzymatic activity on ATP, ADP, AMP and AP4A was measured by a slightly modified Biomol Green phosphate detection method (Enzo Life Sciences). Proteins at 10 nM to 1 μM were incubated with substrate in buffer (20 mM Tris- HCl pH 7.5, 100 mM NaCl) for thirty minutes at 37 °C; four volumes of Biomol Green reagent were added followed by incubation for twenty minutes and absorbance measurement at 620 nm. Product formation was quantified with a phosphate standard curve. Mammalian cell culture and retroviral gene transduction — HEK293T and RAW264.7 cells (ATCC) were maintained in DMEM (Gibco) supplemented with 10% FCS (Gibco) and antibiotics (100 U/ml penicillin and 100 μg/ml streptomycin; Gibco) at 37°C, 5% CO2. Stable expression constructs were introduced to RAW264.7 cells by retroviral gene transduction. Briefly, HEK293T cells were transiently transfected with packaging plasmids pGAG-POL, pVSV-G, pADVANTAGE as well as the individual pMSCV expression constructs using PolyFect (Qiagen). 24 h after transfection, the medium was replaced. 48 h after transfection, the virus-containing supernatants were harvested, filtered through 0.45µm polyethersulfone filters (GE Healthcare), supplemented with 5 μg/ml polybrene (Sigma) and added to 40–60% confluent

139 RAW264.7 cells. 24 h after infection, cells were replated in fresh medium, 48 h after infection, transduced cells were selected by addition of 25 µg/ml blasticidin (InvivoGen). Western blot analysis — Protein was extracted from stably transduced RAW264.7 cells using RIPA buffer (10 mM Tris-HCl pH 7.5, 150 mM NaCl, 1% NP-40, 1% sodium deoxycholate, 0.1% SDS, 1 mM EDTA, Complete protease inhibitor cocktail (Roche) and 250 u/ml Benzonase (Novagen). Protein concentration in cleared lysates was determined using the Bradford protein assay (Bio-Rad) and normalized using RIPA buffer. Samples were mixed with Laemmli sample buffer, resolved by SDS-PAGE and analyzed by western blot. Custom rabbit anti-SMPDL3B antibodies raised against recombinant murine SMPDL3B (aa1-281) were reported previously [26]. Mouse monoclonal anti-HA.11 antibodies were from Covance; anti- Tubulin ones were from Abcam. HRP-conjugated secondary antibodies were from Jackson ImmunResearch. LPS stimulation, enzymatic activity measurement and IL-6 release from RAW264.7 cells — RAW264.7 cells stably transduced with empty vector, wild type or mutant SMPDL3B constructs were harvested in serum-free DMEM medium and counted. In technical triplicates, 5x104 cells in 100 µl DMEM per well were seeded in tissue culture-treated 96-well plates

(Sarstedt) and adhered for 2 h at 37°C, 5% CO2. Cells were stimulated for 8 h with 100 ng/ml ultrapure LPS from E. coli 0111:B4 (InvivoGen). Supernatants were harvested and analyzed for mouse IL-6 by ELISA (eBioscience) according to the manufacturer’s instructions. In parallel, cells were gently washed with PBS and incubated with 100 µl per well of 1 mM bNPP (Sigma) in isotonic TBS buffer (G-Biosciences) at 37°C, 5% CO2 as described previously [26]. 80 µl from each well were transferred to new 96-well plates and combined with 20 µl of 2M NaOH per well. Generation of p-nitrophenol from bNPP by phosphodiesterase activity was determined as absorbance at 405 nm in a microplate reader (SpectraMax).

140 Chapter 4 concluding transition

My structural and enzymatic study of SMPDL3B revealed a narrow substrate-binding cavity that can accommodate a PC head group inside an aromatic box. However, no hydrolysis of SM was detected in neutral liposomes or Triton X-100 micelles, and the enzyme was inactive on the soluble micelle-forming lipids sphingosylphosphorylcholine (lyso-SM) and platelet- activating factor. In subsequent experiments, the protein was produced and purified with a GPI anchor or with an artificial C-terminal hydrophobic helix, and mixed with SM in liposomes or detergent micelles – still without measurable head group cleavage. Future work includes assaying activity on lipids bearing a simple phosphate head group, such as ceramide-1- phosphate, sphingosine-1-phosphate, phosphatidic acid and lysophosphatidic acid, since our small molecule hydrolysis assay results did not demonstrate a definite preference for PC- containing substrates. The reported SMPDL3B binding by the therapeutic antibody rituximab [146,115] will also be investigated. The putative epitope is located far from the active site and, interestingly, is only partially solvent-accessible (Fig. 4.8). We were unable to recombinantly produce the antibody, but recently obtained it from the manufacturer. Initial experiments will seek to confirm a direct interaction between these proteins in vitro, with the end goals of determining the structure of this complex and its effect on enzymatic activity.

141 Figure 4.8 Putative rituximab epitope on SMPDL3B

The SMPDL3B segment corresponding to a peptide recognized by rituximab is displayed in red (left) and black (right).

142 Table 4.1 SMPDL3B X-ray data collection and structure refinement statistics Parameters SMPDL3B SMPDL3B with PC Crystal used for Zn-SAD PDB code 5KAR 5KAS Data Collection Beamline F1, MacCHESS A1, MacCHESS 08ID-1, CMCF, CLS Wavelength (Å) 0.91790 0.97700 1.28206 Space group P 2 P 2 P 2 Unit cell dimensions (Å) 50.89 47.03 93.86 50.90 47.03 93.87 50.98 46.00 93.77 Unit cell angles (°) 90 98.07 90 90 98.07 90 90 97.87 90 Resolution range (Å) 31.99 - 1.14 (1.18 - 1.14) 31.28 - 1.62 (1.68 - 1.62) 50.00 - 1.72 (1.78 - 1.72) Anomalous resolution (Å) 2.81 Total reflections 874,062 341,710 265,057 Unique reflections 147,638 (8,714) 50,662 (2,654) 38,784 (2,258) Multiplicity 5.9 (2.7) 6.7 (3.9) 6.8 (4.8) Completeness (%) 92.9 (56.0) 90.5 (51.3) 84.8 (50.0) I/σ(I) 17.4 (0.7) 12.8 (2.4) 17.9 (1.7) Wilson B-factor (Å2) 13.57 20.28 Rmerge (%) 8.6 (> 100) 11.1 (56.9) 7.9 (88.7) CC1/2 (%) (38) (83) (81) Refinement Reflections used in refinement 147,420 (8,531) 50,343 (2,646) Reflections used for R-free 7,396 (425) 2,537 (130) R-work (%) 13.46 (35.46) 16.92 (23.51) R-free (%) 15.18 (35.88) 19.32 (26.76) Protein molecules per ASU 1 1 Number of non-hydrogen atoms 4,019 3,887 Macromolecules 3,380 3,344 Ligands, carbohydrates 143 140 Solvent 496 403 Protein residues 412 412 RMS(bonds) (Å) 0.009 0.005 RMS(angles) (°) 1.07 0.8 Ramachandran favored (%) 96 96 Ramachandran allowed (%) 4.2 4.1 Ramachandran outliers (%) 0 0 Rotamer outliers (%) 0.27 0.55 Clashscore 1.88 1.62 Average B-factor (Å2) 20.07 29.69 Macromolecules 17.63 27.48 Ligands, carbohydrates 33.11 53.57 Solvent 32.96 39.73

Values in parentheses refer to the highest resolution shell. Zn-SAD, zinc single wavelength anomalous diffraction. Anomalous resolution refers to the resolution limit for substructure determination by SAD. ASU, asymmetric unit. CMCF, Canadian Macromolecular Crystallography Facility. CLS, Canadian Light Source. MacCHESS, Macromolecular Diffraction Facility at the Cornell High Energy Synchrotron Source.

143 Chapter 5 – Conclusion

In summary, I determined the crystal structures of the mammalian ASMase-like proteins: SMPDL3A, ASMase and SMPDL3B. The SMPDL3A project uncovered the helical fold of the novel C-terminal subdomain and a zinc-dependent catalytic mechanism slightly distinct from that of other calcineurin-like phosphoesterases. Structures in complex with reaction products revealed a nucleotide-binding mode without specific contacts to the nucleoside moiety. The ASMase study illustrated the conformational flexibility of its saposin domain: this module can switch from a detached, closed conformation to an open form which establishes a hydrophobic interface to the catalytic domain. I demonstrated that this open configuration represents the active form of the enzyme, likely allowing substrate access to the active site. Finally, our structural investigation of SMPDL3B showed that the protein accommodates a PC head group inside a narrow, boot-shaped substrate binding site. However, further work is required to identify potential bona fide substrates. Structural biology seeks to reveal how proteins work. In this regard, my results clarified certain aspects of the functions of these enzymes, although not as much as initially hoped for. The ASMase-like proteins are members of the calcineurin-like phosphoesterase superfamily which act on a wide variety of molecules, and multiple crystal structures of these diverse enzymes have shed light on the determinants of their substrate preferences. The roles of several family members are still unknown or unclear (as is the case for SMPDL3A and 3B), so structural biology is one approach to answer these questions. Three-dimensional structures can provide information that cannot be obtained from primary sequences alone, for several reasons. First, sequence conservation among the calcineurin-like phosphoesterases is low (Fig. 1.1), and whereas the PPP phosphatase group are relatively closely related, other superfamily members are distant from each other. Although the ASMase-like proteins display most sequence similarity to each other, their functions are not conserved. In fact, the best-preserved portions of these enzymes are not highly informative because they either form the structural core of the molecule – and so are important in maintaining the fold – or constitute the active site. On the other hand, the substrate-binding area is often constituted by loops and other elements which cannot easily be aligned across family members, for example between SMPDL3A and other nucleotide-

144 cleaving calcineurin-like enzymes. In summary, primary sequence alignments allows structural but not functional classification of these proteins. The core of the active site consists of two metal ions, which serve multiple roles. They contribute to binding affinity through electrostatic interactions with the phosphate moiety of the substrate. They also facilitate hydrolysis by increasing electrophilicity of the phosphorus atom via polarization of the coordinating oxygen atoms, and generating a nucleophilic water molecule for the reaction. The enzyme also provides a set of residues to protonate the leaving group. The structure of ASMase and its paralogs revealed that the histidine-aspartate pair involved in this step is different from that in other calcineurin-like phosphoesterases. One remaining question is the factors that influence their optimal pH; SMPDL3B is more active at neutral pH whereas SMPDL3A and ASMase prefer acidic conditions, however no obvious differences are found in their catalytic machinery. The contribution of the zinc ions to substrate affinity was, in a way, a confounding factor in these studies. Due to the strength of these electrostatic interactions, various phosphate- containing molecules are able to bind to the active site, without providing much information on substrate specificity. This was exemplified by the crystal structures these enzymes in complex with PC. In SMPDL3A, the choline moiety did not form specific contacts with the protein, which was expected due to its lack of activity on PC-containing compounds. On the other hand, no specific interactions between choline and ASMase were observed either, even though its substrate bears this head group. Finally, SMPDL3B harbors an aromatic “cage” which is likely selective for choline-containing compounds, but no activity was detected on sphingomyelin or other tested lipids, with relatively weak activity (high km) on CDP-choline. These cases illustrate the notion that complex formation in a crystal does not necessarily mean productive interactions with the compound in solution. Furthermore, detectable activity in an in vitro assay does not necessarily entail physiological relevance of the substrate – for instance, both SMPDL3A and SMPDL3B can hydrolyze ATP. Comparison of kinetic parameters helps assessing the significance of these results, but this analysis becomes complicated with lipids, for which the type of assay system (the environment in which these molecules are found) can greatly influence the outcome. In addition to the structural and in vitro activity data, it is therefore necessary to consider the cellular location of the enzyme under study, with respect to presence and abundance of the

145 substrates of interest. For example, intracellular members of the calcineurin-like family should be highly selective against hydrolysis of ATP, since such action would be harmful to the intracellular pool of this nucleotide and its use as energy source. Conversely, the extracellular ASMase-like enzymes might cleave ATP, as extracellular nucleotides serve signaling (transient) roles, but detection of this activity does not necessarily posit it as their primary function. Localization of a protein at membranes should also be considered, in terms of the high local concentration of potential lipid substrates, while keeping in mind that most membrane-associated enzymes are not . Also, as seen with ASMase and potentially SMPLD3B, lipases need a way to access their substrates from their hydrophobic environment; simple proximity of the protein to the membrane is not sufficient. Due to all these factors, the findings of my structural studies are a complement to, and must be validated by, investigations of these proteins in a physiological context, such as in cells or whole organisms. Structural biology is thus not as suitable for discovering what proteins do, as for determining how they do it, with a great example being the saposin domain of ASMase. It was known from biochemical work and disease-causing mutations that saposins assist enzymes in sphingolipid hydrolysis, with various functional models being proposed, but their detailed mechanisms were clarified by structural studies of saposins and of their complexes with enzymes, including my ASMase investigation. The structure-function relationships found among the calcineurin-like phosphoesterases are not unique to this superfamily. As conclusion, I will draw a parallel with another enzyme family I studied, the nucleotide pyrophosphatases / phosphodiesterases (NPPs), which were briefly mentioned throughout the thesis, and to which many concepts relevant to the ASMase- like proteins also apply. The seven mammalian NPPs are extracellular phosphoesterases harboring a di-zinc active site that carry out various functions [229]. They can be divided into two general classes: nucleotidases, and enzymes that act on lipids or related compounds. Structural biology studies elucidated the bases of this substrate diversity. Besides their phosphate – zinc coordination, NPP1 and NPP4 bind to nucleotides via π-stacking [230,231,232], whereas a disruption of this interaction prevents nucleotide hydrolysis for NPP5 [233]. On the other hand, NPP2 adopts an “inverted” binding mode to its phospholipid substrates, with a hydrophobic tunnel to accommodate its acyl chain [234,235]. Finally, NPP6 and NPP7 recognize the choline

146 moieties of lipids or other metabolites through a “cation-π box” [236,42], with a hydrophobic loop found near the active site of NPP7 involved in accessing lipid substrates from micelles. These studies illustrate how enzyme families can preserve structure and catalytic machinery while evolving widely divergent functions and substrates.

147 References

1. Mitić N, Smith SJ, Neves A, Guddat LW, Gahan LR, Schenk G. The catalytic mechanisms of binuclear metallohydrolases. Chem Rev. 2006 Aug;106(8):3338-63. 2. Schenk G, Mitić N, Gahan LR, Ollis DL, McGeary RP, Guddat LW. Binuclear metallohydrolases: complex mechanistic strategies for a simple chemical reaction. Acc Chem Res. 2012 Sep 18;45(9):1593-603. 3. Vlastaridis P, Kyriakidou P, Chaliotis A, Van de Peer Y, Oliver SG, Amoutzias GD. Estimating the total number of phosphoproteins and phosphorylation sites in eukaryotic proteomes. Gigascience. 2017 Feb 1;6(2):1-11. 4. Manning G, Whyte DB, Martinez R, Hunter T, Sudarsanam S. The protein kinase complement of the . Science. 2002 Dec 6;298(5600):1912-34. 5. Li X, Wilmanns M, Thornton J, Köhn M. Elucidating human phosphatase-substrate networks. Sci Signal. 2013 May 14;6(275):rs10. 6. Heroes E, Lesage B, Görnemann J, Beullens M, Van Meervelt L, Bollen M. The PP1 binding code: a molecular-lego strategy that governs specificity. FEBS J. 2013 Jan;280(2):584-95. 7. Shi Y. Serine/threonine phosphatases: mechanism through structure. Cell. 2009 Oct 30;139(3):468-84. 8. Li H, Zhang L, Rao A, Harrison SC, Hogan PG. Structure of calcineurin in complex with PVIVIT peptide: portrait of a low-affinity signalling interaction. J Mol Biol. 2007 Jun 22;369(5):1296-306. 9. Gaber BP, Sheridan JP, Bazer FW, Roberts RM. Resonance Raman scattering from uteroferrin, the purple glycoprotein of the porcine uterus. J Biol Chem. 1979 Sep 10;254(17):8340-2. 10. Ek-Rylander B, Flores M, Wendel M, Heinegård D, Andersson G. Dephosphorylation of osteopontin and bone sialoprotein by osteoclastic tartrate-resistant acid phosphatase. Modulation of osteoclast adhesion in vitro. J Biol Chem. 1994 May 27;269(21):14853-6. 11. Räisänen SR, Alatalo SL, Ylipahkala H, Halleen JM, Cassady AI, Hume DA, Väänänen HK. Macrophages overexpressing tartrate-resistant acid phosphatase show altered profile of

148 free radical production and enhanced capacity of bacterial killing. Biochem Biophys Res Commun. 2005 May 27;331(1):120-6. 12. Flanagan JU, Cassady AI, Schenk G, Guddat LW, Hume DA. Identification and molecular modeling of a novel, plant-like, human purple acid phosphatase. Gene. 2006 Aug 1;377:12- 20. 13. Zhuo DX, Zhang XW, Jin B, Zhang Z, Xie BS, Wu CL, Gong K, Mao ZB. CSTP1, a novel , blocks cell cycle, promotes cell apoptosis, and suppresses tumor growth of bladder cancer by directly dephosphorylating Akt at Ser473 site. PLoS One. 2013 Jun 17;8(6):e65679. 14. Montemayor EJ, Katolik A, Clark NE, Taylor AB, Schuermann JP, Combs DJ, Johnsson R, Holloway SP, Stevens SW, Damha MJ, Hart PJ. Structural basis of lariat RNA recognition by the intron debranching enzyme Dbr1. Nucleic Acids Res. 2014;42(16):10845-55. 15. Lamarche BJ, Orazio NI, Weitzman MD. The MRN complex in double-strand break repair and telomere maintenance. FEBS Lett. 2010 Sep 10;584(17):3682-95. 16. Liu Y, Sung S, Kim Y, Li F, Gwon G, Jo A, Kim AK, Kim T, Song OK, Lee SE, Cho Y. ATP-dependent DNA binding, unwinding, and resection by the Mre11/Rad50 complex. EMBO J. 2016 Apr 1;35(7):743-58. 17. Knapp K, Zebisch M, Pippel J, El-Tayeb A, Müller CE, Sträter N. Crystal structure of the human ecto-5'-nucleotidase (CD73): insights into the regulation of purinergic signaling. Structure. 2012 Dec 5;20(12):2161-73. 18. Cabezas A, Ribeiro JM, Rodrigues JR, López-Villamizar I, Fernández A, Canales J, Pinto RM, Costas MJ, Cameselle JC. Molecular bases of catalysis and ADP-ribose preference of human Mn2+-dependent ADP-ribose/CDP-alcohol diphosphatase and conversion by mutagenesis to a preferential cyclic ADP-ribose phosphohydrolase. PLoS One. 2015 Feb 18;10(2):e0118680. 19. Markley JL, Aceti DJ, Bingman CA, Fox BG, Frederick RO, Makino S, Nichols KW, Phillips GN Jr, Primm JG, Sahu SC, Vojtik FC, Volkman BF, Wrobel RL, Zolnai Z. The Center for Eukaryotic Structural Genomics. J Struct Funct Genomics. 2009 Apr;10(2):165- 79.

149 20. Tyagi R, Shenoy AR, Visweswariah SS. Characterization of an evolutionarily conserved metallophosphoesterase that is expressed in the fetal brain and associated with the WAGR syndrome. J Biol Chem. 2009 Feb 20;284(8):5217-28. 21. Dermol U, Janardan V, Tyagi R, Visweswariah SS, Podobnik M. Unique utilization of a phosphoprotein phosphatase fold by a mammalian phosphodiesterase associated with WAGR syndrome. J Mol Biol. 2011 Sep 23;412(3):481-94. 22. Liu JJ. Retromer-Mediated Protein Sorting and Vesicular Trafficking. J Genet Genomics. 2016 Apr 20;43(4):165-77. 23. Collins BM, Skinner CF, Watson PJ, Seaman MN, Owen DJ. Vps29 has a phosphoesterase fold that acts as a protein interaction scaffold for retromer assembly. Nat Struct Mol Biol. 2005 Jul;12(7):594-602. 24. Fujita M, Maeda Y, Ra M, Yamaguchi Y, Taguchi R, Kinoshita T. GPI glycan remodeling by PGAP5 regulates transport of GPI-anchored proteins from the ER to the Golgi. Cell. 2009 Oct 16;139(2):352-65. 25. Connelly JC, Kirkham LA, Leach DR. The SbcCD nuclease of Escherichia coli is a structural maintenance of (SMC) family protein that cleaves hairpin DNA. Proc Natl Acad Sci U S A. 1998 Jul 7;95(14):7969-74. 26. Heinz LX, Baumann CL, Köberlin MS, Snijder B, Gawish R, Shui G, Sharif O, Aspalter IM, Müller AC, Kandasamy RK, Breitwieser FP, Pichlmair A, Bruckner M, Rebsamen M, Blüml S, Karonitsch T, Fauster A, Colinge J, Bennett KL, Knapp S, Wenk MR, Superti- Furga G. The Lipid-Modifying Enzyme SMPDL3B Negatively Regulates Innate Immunity. Cell Rep. 2015 Jun 30;11(12):1919-28. 27. Traini M, Quinn CM, Sandoval C, Johansson E, Schroder K, Kockx M, Meikle PJ, Jessup W, Kritharides L. Sphingomyelin phosphodiesterase acid-like 3A (SMPDL3A) is a novel nucleotide phosphodiesterase regulated by cholesterol in human macrophages. J Biol Chem. 2014 Nov 21;289(47):32895-913. 28. van Meer G, Voelker DR, Feigenson GW. Membrane lipids: where they are and how they behave. Nat Rev Mol Cell Biol. 2008 Feb;9(2):112-24. 29. Castro BM, Prieto M, Silva LC. Ceramide: a simple sphingolipid with unique biophysical properties. Prog Lipid Res. 2014 Apr;54:53-67.

150 30. Sezgin E, Levental I, Mayor S, Eggeling C. The mystery of membrane organization: composition, regulation and roles of lipid rafts. Nat Rev Mol Cell Biol. 2017 Jun;18(6):361- 374. 31. Stancevic B, Kolesnick R. Ceramide-rich platforms in transmembrane signaling. FEBS Lett. 2010 May 3;584(9):1728-40. 32. Hannun YA, Obeid LM. Principles of bioactive lipid signalling: lessons from sphingolipids. Nat Rev Mol Cell Biol. 2008 Feb;9(2):139-50. 33. Airola MV, Hannun YA. Sphingolipid metabolism and neutral sphingomyelinases. Handb Exp Pharmacol. 2013;(215):57-76. 34. Openshaw AE, Race PR, Monzó HJ, Vázquez-Boland JA, Banfield MJ. Crystal structure of SmcL, a bacterial neutral sphingomyelinase C from Listeria. J Biol Chem. 2005 Oct 14;280(41):35011-7. 35. Airola MV, Shanbhogue P, Shamseddine AA, Guja KE, Senkal CE, Maini R, Bartke N, Wu BX, Obeid LM, Garcia-Diaz M, Hannun YA. Structure of human nSMase2 reveals an interdomain allosteric activation mechanism for ceramide generation. Proc Natl Acad Sci U S A. 2017 Jul 11;114(28):E5549-E5558. 36. Yabu T, Shiba H, Shibasaki Y, Nakanishi T, Imamura S, Touhata K, Yamashita M. Stress- induced ceramide generation and apoptosis via the phosphorylation and activation of nSMase1 by JNK signaling. Cell Death Differ. 2015 Feb;22(2):258-73. 37. Yabu T, Shimuzu A, Yamashita M. A novel mitochondrial sphingomyelinase in zebrafish cells. J Biol Chem. 2009 Jul 24;284(30):20349-63. 38. Wu BX, Rajagopalan V, Roddy PL, Clarke CJ, Hannun YA. Identification and characterization of murine mitochondria-associated neutral sphingomyelinase (MA- nSMase), the mammalian sphingomyelin phosphodiesterase 5. J Biol Chem. 2010 Jun 4;285(23):17993-8002. 39. Krut O, Wiegmann K, Kashkar H, Yazdanpanah B, Krönke M. Novel tumor necrosis factor- responsive mammalian neutral sphingomyelinase-3 is a C-tail-anchored protein. J Biol Chem. 2006 May 12;281(19):13784-93. 40. Duan RD. Alkaline sphingomyelinase: an old enzyme with novel implications. Biochim Biophys Acta. 2006 Mar;1761(3):281-91.

151 41. Zimmermann H, Zebisch M, Sträter N. Cellular function and molecular structure of ecto- nucleotidases. Purinergic Signal. 2012 Sep;8(3):437-502. 42. Gorelik A, Liu F, Illes K, Nagar B. Crystal structure of the human alkaline sphingomyelinase provides insights into substrate recognition. J Biol Chem. 2017 Apr 28;292(17):7087-7094. 43. GATT S. ENZYMIC HYDROLYSIS AND SYNTHESIS OF CERAMIDES. J Biol Chem. 1963 Sep;238:3131-3. 44. Yamaguchi S, Suzuki K. Purification and characterization of sphingomyelinase from human brain. J Biol Chem. 1977 Jun 10;252(11):3805-13. 45. Quintern LE, Schuchman EH, Levran O, Suchi M, Ferlinz K, Reinke H, Sandhoff K, Desnick RJ. Isolation of cDNA clones encoding human acid sphingomyelinase: occurrence of alternatively processed transcripts. EMBO J. 1989 Sep;8(9):2469-73. 46. Brady RO, Kanfer JN, Mock MB, Fredrickson DS. The metabolism of sphingomyelin. II. Evidence of an enzymatic deficiency in Niemann-Pick diseae. Proc Natl Acad Sci U S A. 1966 Feb;55(2):366-9. 47. Gulbins E. Regulation of death receptor signaling and apoptosis by ceramide. Pharmacol Res. 2003 May;47(5):393-9. 48. Santana P, Peña LA, Haimovitz-Friedman A, Martin S, Green D, McLoughlin M, Cordon- Cardo C, Schuchman EH, Fuks Z, Kolesnick R. Acid sphingomyelinase-deficient human lymphoblasts and mice are defective in radiation-induced apoptosis. Cell. 1996 Jul 26;86(2):189-99. 49. Zampieri S, Filocamo M, Pianta A, Lualdi S, Gort L, Coll MJ, Sinnott R, Geberhiwot T, Bembi B, Dardis A. SMPD1 Mutation Update: Database and Comprehensive Analysis of Published and Novel Variants. Hum Mutat. 2016 Feb;37(2):139-47. 50. Schuchman EH, Desnick RJ. Types A and B Niemann-Pick disease. Mol Genet Metab. 2017 Jan - Feb;120(1-2):27-33. 51. Wasserstein MP, Schuchman EH. Acid Sphingomyelinase Deficiency. GeneReviews® [Internet]. Seattle (WA): University of Washington, Seattle; 1993-2017. 2006 Dec 7 [updated 2015 Jun 18] 52. Smith EL, Schuchman EH. The unexpected role of acid sphingomyelinase in cell death and the pathophysiology of common diseases. FASEB J. 2008 Oct;22(10):3419-31.

152 53. Schuchman EH. Acid sphingomyelinase, cell membranes and human disease: lessons from Niemann-Pick disease. FEBS Lett. 2010 May 3;584(9):1895-900. 54. McGovern MM, Pohl-Worgall T, Deckelbaum RJ, Simpson W, Mendelson D, Desnick RJ, Schuchman EH, Wasserstein MP. Lipid abnormalities in children with types A and B Niemann Pick disease. J Pediatr. 2004 Jul;145(1):77-81. 55. Neufeld EF. Enzyme replacement therapy - a brief history. Fabry Disease: Perspectives from 5 Years of FOS. Oxford: Oxford PharmaGenesis; 2006. 56. Stirnemann J, Belmatoug N, Camou F, Serratrice C, Froissart R, Caillaud C, Levade T, Astudillo L, Serratrice J, Brassier A, Rose C, Billette de Villemeur T, Berger MG. A Review of Gaucher Disease Pathophysiology, Clinical Presentation and Treatments. Int J Mol Sci. 2017 Feb 17;18(2). pii: E441. 57. El Dib R, Gomaa H, Carvalho RP, Camargo SE, Bazan R, Barretti P, Barreto FC. Enzyme replacement therapy for Anderson-Fabry disease. Cochrane Database Syst Rev. 2016 Jul 25;7:CD006663. 58. Urayama A. Toward the successful delivery of lysosomal enzymes across the blood--brain barrier. Clin Exp Neuroimmunol. 2013 Aug;4(2):228-38. 59. Cox T, Lachmann R, Hollak C, Aerts J, van Weely S, Hrebícek M, Platt F, Butters T, Dwek R, Moyses C, Gow I, Elstein D, Zimran A. Novel oral treatment of Gaucher's disease with N-butyldeoxynojirimycin (OGT 918) to decrease substrate biosynthesis. Lancet. 2000 Apr 29;355(9214):1481-5. 60. Fan JQ, Ishii S, Asano N, Suzuki Y. Accelerated transport and maturation of lysosomal alpha-galactosidase A in Fabry lymphoblasts by an . Nat Med. 1999 Jan;5(1):112-5. 61. Worgall S, Sondhi D, Hackett NR, Kosofsky B, Kekatpure MV, Neyzi N, Dyke JP, Ballon D, Heier L, Greenwald BM, Christos P, Mazumdar M, Souweidane MM, Kaplitt MG, Crystal RG. Treatment of late infantile neuronal ceroid lipofuscinosis by CNS administration of a serotype 2 adeno-associated virus expressing CLN2 cDNA. Hum Gene Ther. 2008 May;19(5):463-74. 62. Schulze H, Kolter T, Sandhoff K. Principles of lysosomal membrane degradation: Cellular topology and biochemistry of lysosomal lipid degradation. Biochim Biophys Acta. 2009 Apr;1793(4):674-83.

153 63. Newrzella D, Stoffel W. Functional analysis of the glycosylation of murine acid sphingomyelinase. J Biol Chem. 1996 Dec 13;271(50):32089-95. 64. Fürst W, Sandhoff K. Activator proteins and topology of lysosomal sphingolipid catabolism. Biochim Biophys Acta. 1992 Jun 5;1126(1):1-16. 65. Infante RE, Wang ML, Radhakrishnan A, Kwon HJ, Brown MS, Goldstein JL. NPC2 facilitates bidirectional transfer of cholesterol between NPC1 and lipid bilayers, a step in cholesterol egress from lysosomes. Proc Natl Acad Sci U S A. 2008 Oct 7;105(40):15287- 92. 66. Linke T, Wilkening G, Lansmann S, Moczall H, Bartelsen O, Weisgerber J, Sandhoff K. Stimulation of acid sphingomyelinase activity by lysosomal lipids and sphingolipid activator proteins. Biol Chem. 2001 Feb;382(2):283-90. 67. Oninla VO, Breiden B, Babalola JO, Sandhoff K. Acid sphingomyelinase activity is regulated by membrane lipids and facilitates cholesterol transfer by NPC2. J Lipid Res. 2014 Dec;55(12):2606-19. 68. Kölzer M, Werth N, Sandhoff K. Interactions of acid sphingomyelinase and lipid bilayers in the presence of the tricyclic antidepressant desipramine. FEBS Lett. 2004 Feb 13;559(1- 3):96-8. 69. Anderson N, Borlak J. Drug-induced phospholipidosis. FEBS Lett. 2006 Oct 9;580(23):5533-40. 70. Hurwitz R, Ferlinz K, Sandhoff K. The tricyclic antidepressant desipramine causes proteolytic degradation of lysosomal sphingomyelinase in human fibroblasts. Biol Chem Hoppe Seyler. 1994 Jul;375(7):447-50. 71. Peter Slotte J. Molecular properties of various structurally defined -- correlation of structure with function. Prog Lipid Res. 2013 Apr;52(2):206-19. 72. Sandhoff K. Neuronal : Membrane lipids and sphingolipid activator proteins regulate lysosomal sphingolipid catabolism. Biochimie. 2016 Nov;130:146-151. 73. Kirkegaard T, Roth AG, Petersen NH, Mahalka AK, Olsen OD, Moilanen I, Zylicz A, Knudsen J, Sandhoff K, Arenz C, Kinnunen PK, Nylandsted J, Jäättelä M. Hsp70 stabilizes lysosomes and reverts Niemann-Pick disease-associated lysosomal pathology. Nature. 2010 Jan 28;463(7280):549-53.

154 74. Petersen NH, Olsen OD, Groth-Pedersen L, Ellegaard AM, Bilgin M, Redmer S, Ostenfeld MS, Ulanet D, Dovmark TH, Lønborg A, Vindeløv SD, Hanahan D, Arenz C, Ejsing CS, Kirkegaard T, Rohde M, Nylandsted J, Jäättelä M. Transformation-associated changes in sphingolipid metabolism sensitize cells to lysosomal cell death induced by inhibitors of acid sphingomyelinase. Cancer Cell. 2013 Sep 9;24(3):379-93. 75. Gulbins E, Kolesnick RN. It takes a CAD to kill a tumor cell with a LMP. Cancer Cell. 2013 Sep 9;24(3):279-81. 76. Alattia JR, Shaw JE, Yip CM, Privé GG. Molecular imaging of membrane interfaces reveals mode of beta-glucosidase activation by saposin C. Proc Natl Acad Sci U S A. 2007 Oct 30;104(44):17394-9. 77. Rossmann M, Schultz-Heienbrok R, Behlke J, Remmel N, Alings C, Sandhoff K, Saenger W, Maier T. Crystal structures of human saposins C andD: implications for lipid recognition and membrane interactions. Structure. 2008 May;16(5):809-17. 78. Popovic K, Holyoake J, Pomès R, Privé GG. Structure of saposin A lipoprotein discs. Proc Natl Acad Sci U S A. 2012 Feb 21;109(8):2908-12. 79. Ahn VE, Faull KF, Whitelegge JP, Fluharty AL, Privé GG. Crystal structure of saposin B reveals a dimeric shell for lipid binding. Proc Natl Acad Sci U S A. 2003 Jan 7;100(1):38- 43. 80. Wright CS, Zhao Q, Rastinejad F. Structural analysis of lipid complexes of GM2-activator protein. J Mol Biol. 2003 Aug 22;331(4):951-64. 81. Conzelmann E, Sandhoff K. Purification and characterization of an activator protein for the degradation of glycolipids GM2 and GA2 by hexosaminidase A. Hoppe Seylers Z Physiol Chem. 1979 Dec;360(12):1837-49. 82. Kolter T, Sandhoff K. Lysosomal degradation of membrane lipids. FEBS Lett. 2010 May 3;584(9):1700-12. 83. Remmel N, Locatelli-Hoops S, Breiden B, Schwarzmann G, Sandhoff K. Saposin B mobilizes lipids from cholesterol-poor and bis(monoacylglycero)phosphate-rich membranes at acidic pH. Unglycosylated patient variant saposin B lacks lipid-extraction capacity. FEBS J. 2007 Jul;274(13):3405-20.

155 84. Gruschus JM, Jiang Z, Yap TL, Hill SA, Grishaev A, Piszczek G, Sidransky E, Lee JC. Dissociation of glucocerebrosidase dimer in solution by its co-factor, saposin C. Biochem Biophys Res Commun. 2015 Feb 20;457(4):561-6. 85. Azuma N, O'Brien JS, Moser HW, Kishimoto Y. Stimulation of acid ceramidase activity by saposin D. Arch Biochem Biophys. 1994 Jun;311(2):354-7. 86. Ponting CP. Acid sphingomyelinase possesses a domain homologous to its activator proteins: saposins B and D. Protein Sci. 1994 Feb;3(2):359-61. 87. Jenkins RW, Canals D, Hannun YA. Roles and regulation of secretory and lysosomal acid sphingomyelinase. Cell Signal. 2009 Jun;21(6):836-46. 88. Ferlinz K, Linke T, Bartelsen O, Weiler M, Sandhoff K. Stimulation of lysosomal sphingomyelin degradation by sphingolipid activator proteins. Chem Phys Lipids. 1999 Nov;102(1-2):35-43. 89. Kölzer M, Ferlinz K, Bartelsen O, Hoops SL, Lang F, Sandhoff K. Functional characterization of the postulated intramolecular sphingolipid activator protein domain of human acid sphingomyelinase. Biol Chem. 2004 Dec;385(12):1193-5. 90. Quintern LE, Weitz G, Nehrkorn H, Tager JM, Schram AW, Sandhoff K. Acid sphingomyelinase from human urine: purification and characterization. Biochim Biophys Acta. 1987 Dec 14;922(3):323-36. 91. Spence MW, Byers DM, Palmer FB, Cook HW. A new Zn2+-stimulated sphingomyelinase in fetal bovine serum. J Biol Chem. 1989 Apr 5;264(10):5358-63. 92. Schissel SL, Schuchman EH, Williams KJ, Tabas I. Zn2+-stimulated sphingomyelinase is secreted by many cell types and is a product of the acid sphingomyelinase gene. J Biol Chem. 1996 Aug 2;271(31):18431-6. 93. Coutinho MF, Prata MJ, Alves S. Mannose-6-phosphate pathway: a review on its role in lysosomal function and dysfunction. Mol Genet Metab. 2012 Apr;105(4):542-50. 94. Rhein C, Reichel M, Mühle C, Rotter A, Schwab SG, Kornhuber J. Secretion of acid Sphingomyelinase is affected by its polymorphic signal peptide. Cell Physiol Biochem. 2014;34(4):1385-401. 95. Zhang Y, Li X, Grassmé H, Döring G, Gulbins E. Alterations in ceramide concentration and pH determine the release of reactive oxygen species by Cftr-deficient macrophages on infection. J Immunol. 2010 May 1;184(9):5104-11.

156 96. Blott EJ, Griffiths GM. Secretory lysosomes. Nat Rev Mol Cell Biol. 2002 Feb;3(2):122-31. 97. Callahan JW, Jones CS, Davidson DJ, Shankaran P. The active site of lysosomal sphingomyelinase: evidence for the involvement of hydrophobic and ionic groups. J Neurosci Res. 1983;10(2):151-63. 98. Schissel SL, Jiang X, Tweedie-Hardman J, Jeong T, Camejo EH, Najib J, Rapp JH, Williams KJ, Tabas I. Secretory sphingomyelinase, a product of the acid sphingomyelinase gene, can hydrolyze atherogenic lipoproteins at neutral pH. Implications for atherosclerotic lesion development. J Biol Chem. 1998 Jan 30;273(5):2738-46. 99. Schissel SL, Keesler GA, Schuchman EH, Williams KJ, Tabas I. The cellular trafficking and zinc dependence of secretory and lysosomal sphingomyelinase, two products of the acid sphingomyelinase gene. J Biol Chem. 1998 Jul 17;273(29):18250-9. 100. Schuchman EH, Suchi M, Takahashi T, Sandhoff K, Desnick RJ. Human acid sphingomyelinase. Isolation, nucleotide sequence and expression of the full-length and alternatively spliced cDNAs. J Biol Chem. 1991 May 5;266(13):8531-9. 101. Lansmann S, Ferlinz K, Hurwitz R, Bartelsen O, Glombitza G, Sandhoff K. Purification of acid sphingomyelinase from human placenta: characterization and N-terminal sequence. FEBS Lett. 1996 Dec 16;399(3):227-31. 102. Jenkins RW, Idkowiak-Baldys J, Simbari F, Canals D, Roddy P, Riner CD, Clarke CJ, Hannun YA. A novel mechanism of lysosomal acid sphingomyelinase maturation: requirement for carboxyl-terminal proteolytic processing. J Biol Chem. 2011 Feb 4;286(5):3777-88. 103. Qiu H, Edmunds T, Baker-Malcolm J, Karey KP, Estes S, Schwarz C, Hughes H, Van Patten SM. Activation of human acid sphingomyelinase through modification or deletion of C-terminal cysteine. J Biol Chem. 2003 Aug 29;278(35):32744-52. 104. Grassme H, Jekle A, Riehle A, Schwarz H, Berger J, Sandhoff K, Kolesnick R, Gulbins E. CD95 signaling via ceramide-rich membrane rafts. J Biol Chem. 2001 Jun 8;276(23):20589- 96. 105. Zeidan YH, Hannun YA. Activation of acid sphingomyelinase by protein kinase Cdelta- mediated phosphorylation. J Biol Chem. 2007 Apr 13;282(15):11549-61. 106. Henry B, Ziobro R, Becker KA, Kolesnick R, Gulbins E. Acid sphingomyelinase. Handb Exp Pharmacol. 2013;(215):77-88.

157 107. Gulbins E, Palmada M, Reichel M, Lüth A, Böhmer C, Amato D, Müller CP, Tischbirek CH, Groemer TW, Tabatabai G, Becker KA, Tripal P, Staedtler S, Ackermann TF, van Brederode J, Alzheimer C, Weller M, Lang UE, Kleuser B, Grassmé H, Kornhuber J. Acid sphingomyelinase-ceramide system mediates effects of antidepressant drugs. Nat Med. 2013 Jul;19(7):934-8. 108. Kumar V, Butcher SJ, Öörni K, Engelhardt P, Heikkonen J, Kaski K, Ala-Korpela M, Kovanen PT. Three-dimensional cryoEM reconstruction of native LDL particles to 16Ã… resolution at physiological body temperature. PLoS One. 2011 May 9;6(5):e18841. 109. Kornhuber J, Rhein C, Müller CP, Mühle C. Secretory sphingomyelinase in health and disease. Biol Chem. 2015 Jun;396(6-7):707-36. 110. Marathe S, Schissel SL, Yellin MJ, Beatini N, Mintzer R, Williams KJ, Tabas I. Human vascular endothelial cells are a rich and regulatable source of secretory sphingomyelinase. Implications for early atherogenesis and ceramide-mediated cell signaling. J Biol Chem. 1998 Feb 13;273(7):4081-8. 111. Schissel SL, Tweedie-Hardman J, Rapp JH, Graham G, Williams KJ, Tabas I. Rabbit aorta and human atherosclerotic lesions hydrolyze the sphingomyelin of retained low-density lipoprotein. Proposed role for arterial-wall sphingomyelinase in subendothelial retention and aggregation of atherogenic lipoproteins. J Clin Invest. 1996 Sep 15;98(6):1455-64. 112. Tabas I, Williams KJ, Borén J. Subendothelial lipoprotein retention as the initiating process in atherosclerosis: update and therapeutic implications. Circulation. 2007 Oct 16;116(16):1832-44. 113. Devlin CM, Leventhal AR, Kuriakose G, Schuchman EH, Williams KJ, Tabas I. Acid sphingomyelinase promotes lipoprotein retention within early atheromata and accelerates lesion progression. Arterioscler Thromb Vasc Biol. 2008 Oct;28(10):1723-30. 114. Masuishi Y, Nomura A, Okayama A, Kimura Y, Arakawa N, Hirano H. Mass spectrometric identification of glycosylphosphatidylinositol-anchored peptides. J Proteome Res. 2013 Oct 4;12(10):4617-26. 115. Fornoni A, Sageshima J, Wei C, Merscher-Gomez S, Aguillon-Prada R, Jauregui AN, Li J, Mattiazzi A, Ciancio G, Chen L, Zilleruelo G, Abitbol C, Chandar J, Seeherunvong W, Ricordi C, Ikehata M, Rastaldi MP, Reiser J, Burke GW 3rd. Rituximab targets podocytes in recurrent focal segmental glomerulosclerosis. Sci Transl Med. 2011 Jun 1;3(85):85ra46.

158 116. Ruysschaert JM, Lonez C. Role of lipid microdomains in TLR-mediated signalling. Biochim Biophys Acta. 2015 Sep;1848(9):1860-7. 117. Płóciennikowska A, Hromada-Judycka A, Borzęcka K, Kwiatkowska K Co-operation of TLR4 and raft proteins in LPS-induced pro-inflammatory signaling. Cell Mol Life Sci. 2015 Feb;72(3):557-81. 118. Triantafilou M, Gamper FG, Haston RM, Mouratis MA, Morath S, Hartung T, Triantafilou K. Membrane sorting of toll-like receptor (TLR)-2/6 and TLR2/1 heterodimers at the cell surface determines heterotypic associations with CD36 and intracellular targeting. J Biol Chem. 2006 Oct 13;281(41):31002-11. 119. Triantafilou M, Miyake K, Golenbock DT, Triantafilou K. Mediators of innate immune recognition of bacteria concentrate in lipid rafts and facilitate lipopolysaccharide-induced cell activation. J Cell Sci. 2002 Jun 15;115(Pt 12):2603-11. 120. Cuschieri J, Bulger E, Billgrin J, Garcia I, Maier RV. Acid sphingomyelinase is required for lipid Raft TLR4 complex formation. Surg Infect (Larchmt). 2007 Feb;8(1):91-106. 121. Finberg RW, Kurt-Jones EA. CD14: chaperone or matchmaker? Immunity. 2006 Feb;24(2):127-9. 122. Li Z, Fan Y, Liu J, Li Y, Huan C, Bui HH, Kuo MS, Park TS, Cao G, Jiang XC. Impact of sphingomyelin synthase 1 deficiency on sphingolipid metabolism and atherosclerosis in mice. Arterioscler Thromb Vasc Biol. 2012 Jul;32(7):1577-84. 123. Gowda S, Yeang C, Wadgaonkar S, Anjum F, Grinkina N, Cutaia M, Jiang XC, Wadgaonkar R. Sphingomyelin synthase 2 (SMS2) deficiency attenuates LPS-induced lung injury. Am J Physiol Lung Cell Mol Physiol. 2011 Mar;300(3):L430-40. 124. Dhungana S, Merrick BA, Tomer KB, Fessler MB. Quantitative proteomics analysis of macrophage rafts reveals compartmentalized activation of the proteasome and of proteasome-mediated ERK activation in response to lipopolysaccharide. Mol Cell Proteomics. 2009 Jan;8(1):201-13. 125. MacKichan ML, DeFranco AL. Role of ceramide in lipopolysaccharide (LPS)-induced signaling. LPS increases ceramide rather than acting as a structural homolog. J Biol Chem. 1999 Jan 15;274(3):1767-75. 126. Sakata A, Ochiai T, Shimeno H, Hikishima S, Yokomatsu T, Shibuya S, Toda A, Eyanagi R, Soeda S. Acid sphingomyelinase inhibition suppresses lipopolysaccharide-mediated

159 release of inflammatory cytokines from macrophages and protects against disease pathology in dextran sulphate sodium-induced colitis in mice. Immunology. 2007 Sep;122(1):54-64. 127. van Blitterswijk WJ, van der Luit AH, Veldman RJ, Verheij M, Borst J. Ceramide: second messenger or modulator of membrane structure and dynamics? Biochem J. 2003 Jan 15;369(Pt 2):199-211. 128. Fischer H, Ellström P, Ekström K, Gustafsson L, Gustafsson M, Svanborg C. Ceramide as a TLR4 agonist; a putative signalling intermediate between sphingolipid receptors for microbial ligands and TLR4. Cell Microbiol. 2007 May;9(5):1239-51. 129. Chiba N, Masuda A, Yoshikai Y, Matsuguchi T. Ceramide inhibits LPS-induced production of IL-5, IL-10, and IL-13 from mast cells. J Cell Physiol. 2007 Oct;213(1):126-36. 130. Sun Y, Fox T, Adhikary G, Kester M, Pearlman E. Inhibition of corneal inflammation by liposomal delivery of short-chain, C-6 ceramide. J Leukoc Biol. 2008 Jun;83(6):1512-21. 131. Józefowski S, Czerkies M, Łukasik A, Bielawska A, Bielawski J, Kwiatkowska K, Sobota A. Ceramide and ceramide 1-phosphate are negative regulators of TNF-α production induced by lipopolysaccharide. J Immunol. 2010 Dec 1;185(11):6960-73. 132. Haimovitz-Friedman A, Cordon-Cardo C, Bayoumy S, Garzotto M, McLoughlin M, Gallily R, Edwards CK 3rd, Schuchman EH, Fuks Z, Kolesnick R. Lipopolysaccharide induces disseminated endothelial apoptosis requiring ceramide generation. J Exp Med. 1997 Dec 1;186(11):1831-41. 133. Rozenova KA, Deevska GM, Karakashian AA, Nikolova-Karakashian MN. Studies on the role of acid sphingomyelinase and ceramide in the regulation of tumor necrosis factor alpha (TNFalpha)-converting enzyme activity and TNFalpha secretion in macrophages. J Biol Chem. 2010 Jul 2;285(27):21103-13. 134. Fadeel B, Xue D. The ins and outs of phospholipid asymmetry in the plasma membrane: roles in health and disease. Crit Rev Biochem Mol Biol. 2009 Sep-Oct;44(5):264-77. 135. Köberlin MS, Snijder B, Heinz LX, Baumann CL, Fauster A, Vladimer GI, Gavin AC, Superti-Furga G. A Conserved Circular Network of Coregulated Lipids Modulates Innate Immune Responses. Cell. 2015 Jul 2;162(1):170-83. 136. Dietrich C, Volovyk ZN, Levi M, Thompson NL, Jacobson K. Partitioning of Thy-1, GM1, and cross-linked phospholipid analogs into lipid rafts reconstituted in supported model membrane monolayers. Proc Natl Acad Sci U S A. 2001 Sep 11;98(19):10642-7.

160 137. Li R, Blanchette-Mackie EJ, Ladisch S. Induction of endocytic vesicles by exogenous C(6)- ceramide. J Biol Chem. 1999 Jul 23;274(30):21121-7. 138. Saleem MA, Ni L, Witherden I, Tryggvason K, Ruotsalainen V, Mundel P, Mathieson PW. Co-localization of nephrin, podocin, and the actin cytoskeleton: evidence for a role in podocyte foot process formation. Am J Pathol. 2002 Oct;161(4):1459-66. 139. Grahammer F, Schell C, Huber TB. The podocyte slit diaphragm--from a thin grey line to a complex signalling hub. Nat Rev Nephrol. 2013 Oct;9(10):587-98. 140. Schwarz K, Simons M, Reiser J, Saleem MA, Faul C, Kriz W, Shaw AS, Holzman LB, Mundel P. Podocin, a raft-associated component of the glomerular slit diaphragm, interacts with CD2AP and nephrin. J Clin Invest. 2001 Dec;108(11):1621-9. 141. Huber TB, Simons M, Hartleben B, Sernetz L, Schmidts M, Gundlach E, Saleem MA, Walz G, Benzing T. Molecular basis of the functional podocin-nephrin complex: mutations in the NPHS2 gene disrupt nephrin targeting to lipid raft microdomains. Hum Mol Genet. 2003 Dec 15;12(24):3397-405. 142. Fornoni A, Merscher S, Kopp JB. Lipid biology of the podocyte--new perspectives offer new opportunities. Nat Rev Nephrol. 2014 Jul;10(7):379-88. 143. Walia HK, Li H, Rueschman M, Bhatt DL, Patel SR, Quan SF, Gottlieb DJ, Punjabi NM, Redline S, Mehra R. Association of severe obstructive sleep apnea and elevated blood pressure despite antihypertensive medication use. J Clin Sleep Med. 2014 Aug 15;10(8):835-43. 144. Najafian B, Svarstad E, Bostad L, Gubler MC, Tøndel C, Whitley C, Mauer M. Progressive podocyte injury and globotriaosylceramide (GL-3) accumulation in young patients with Fabry disease. Kidney Int. 2011 Mar;79(6):663-70. 145. Jin J, Sison K, Li C, Tian R, Wnuk M, Sung HK, Jeansson M, Zhang C, Tucholska M, Jones N, Kerjaschki D, Shibuya M, Fantus IG, Nagy A, Gerber HP, Ferrara N, Pawson T, Quaggin SE. Soluble FLT1 binds lipid microdomains in podocytes to control cell morphology and glomerular barrier function. Cell. 2012 Oct 12;151(2):384-99. 146. Perosa F, Favoino E, Caragnano MA, Dammacco F. Generation of biologically active linear and cyclic peptides has revealed a unique fine specificity of rituximab and its possible cross- reactivity with acid sphingomyelinase-like phosphodiesterase 3b precursor. Blood. 2006 Feb 1;107(3):1070-7.

161 147. Edelmann B, Bertsch U, Tchikov V, Winoto-Morbach S, Perrotta C, Jakob M, Adam- Klages S, Kabelitz D, Schütze S. Caspase-8 and caspase-7 sequentially mediate proteolytic activation of acid sphingomyelinase in TNF-R1 receptosomes. EMBO J. 2011 Jan 19;30(2):379-94. 148. Tasaki M, Shimizu A, Hanekamp I, Torabi R, Villani V, Yamada K. Rituximab treatment prevents the early development of proteinuria following pig-to-baboon xeno-kidney transplantation. J Am Soc Nephrol. 2014 Apr;25(4):737-44. 149. Takahashi Y, Ikezumi Y, Saitoh A. Rituximab protects podocytes and exerts anti-proteinuric effects in rat adriamycin-induced nephropathy independent of B-lymphocytes. Nephrology (Carlton). 2017 Jan;22(1):49-57. 150. Ahmad A, Mitrofanova A, Bielawski J, Yang Y, Marples B, Fornoni A, Zeidan YH. Sphingomyelinase-like phosphodiesterase 3b mediates radiation-induced damage of renal podocytes. FASEB J. 2017 Feb;31(2):771-780. 151. Yoo TH, Pedigo CE, Guzman J, Correa-Medina M, Wei C, Villarreal R, Mitrofanova A, Leclercq F, Faul C, Li J, Kretzler M, Nelson RG, Lehto M, Forsblom C, Groop PH, Reiser J, Burke GW, Fornoni A, Merscher S. Sphingomyelinase-like phosphodiesterase 3b expression levels determine podocyte injury phenotypes in glomerular disease. J Am Soc Nephrol. 2015 Jan;26(1):133-47. 152. Wei C, El Hindi S, Li J, Fornoni A, Goes N, Sageshima J, Maiguel D, Karumanchi SA, Yap HK, Saleem M, Zhang Q, Nikolic B, Chaudhuri A, Daftarian P, Salido E, Torres A, Salifu M, Sarwal MM, Schaefer F, Morath C, Schwenger V, Zeier M, et al. Circulating urokinase receptor as a cause of focal segmental glomerulosclerosis. Nat Med. 2011 Jul 31;17(8):952- 60. 153. Picariello G, Ferranti P, Mamone G, Klouckova I, Mechref Y, Novotny MV, Addeo F. Gel- free shotgun proteomic analysis of human milk. J Chromatogr A. 2012 Mar 2;1227:219-33. 154. Thouvenot E, Urbach S, Dantec C, Poncet J, Séveno M, Demettre E, Jouin P, Touchon J, Bockaert J, Marin P. Enhanced detection of CNS cell secretome in plasma protein-depleted cerebrospinal fluid. J Proteome Res. 2008 Oct;7(10):4409-21. 155. Rindler MJ, Xu CF, Gumper I, Smith NN, Neubert TA. Proteomic analysis of pancreatic zymogen granules: identification of new granule proteins. J Proteome Res. 2007 Aug;6(8):2978-92.

162 156. Ogawa Y, Miura Y, Harazono A, Kanai-Azuma M, Akimoto Y, Kawakami H, Yamaguchi T, Toda T, Endo T, Tsubuki M, Yanoshita R. Proteomic analysis of two types of exosomes in human whole saliva. Biol Pharm Bull. 2011;34(1):13-23. 157. Wicher KB, Fries E. Prohaptoglobin is proteolytically cleaved in the endoplasmic reticulum by the complement C1r-like protein. Proc Natl Acad Sci U S A. 2004 Oct 5;101(40):14390- 5. 158. Zhou M, Lucas DA, Chan KC, Issaq HJ, Petricoin EF 3rd, Liotta LA, Veenstra TD, Conrads TP. An investigation into the human serum “interactome”. Electrophoresis. 2004 May;25(9):1289-98. 159. Veenstra TD, Conrads TP, Hood BL, Avellino AM, Ellenbogen RG, Morrison RS. Biomarkers: mining the biofluid proteome. Mol Cell Proteomics. 2005 Apr;4(4):409-18. 160. Zhou M, Conrads TP, Veenstra TD. Proteomics approaches to biomarker detection. Brief Funct Genomic Proteomic. 2005 May;4(1):69-75. 161. Czupalla C, Mansukoski H, Riedl T, Thiel D, Krause E, Hoflack B. Proteomic analysis of lysosomal acid hydrolases secreted by osteoclasts: implications for lytic enzyme transport and bone metabolism. Mol Cell Proteomics. 2006 Jan;5(1):134-43. 162. Lim JM, Sherling D, Teo CF, Hausman DB, Lin D, Wells L. Defining the regulated secreted proteome of rodent adipocytes upon the induction of insulin resistance. J Proteome Res. 2008 Mar;7(3):1251-63. 163. Lim JM, Wollaston-Hayden EE, Teo CF, Hausman D, Wells L. Quantitative secretome and glycome of primary human adipocytes during insulin resistance. Clin Proteomics. 2014 May 12;11(1):20. 164. Keene SD, Greco TM, Parastatidis I, Lee SH, Hughes EG, Balice-Gordon RJ, Speicher DW, Ischiropoulos H. Mass spectrometric and computational analysis of cytokine-induced alterations in the astrocyte secretome. Proteomics. 2009 Feb;9(3):768-82. 165. Moore NH, Costa LG, Shaffer SA, Goodlett DR, Guizzetti M. Shotgun proteomics implicates extracellular matrix proteins and protease systems in neuronal development induced by astrocyte cholinergic stimulation. J Neurochem. 2009 Feb;108(4):891-908. 166. Greco TM, Seeholzer SH, Mak A, Spruce L, Ischiropoulos H. Quantitative mass spectrometry-based proteomics reveals the dynamic range of primary mouse astrocyte protein secretion. J Proteome Res. 2010 May 7;9(5):2764-74.

163 167. Sleat DE, Zheng H, Qian M, Lobel P. Identification of sites of mannose 6-phosphorylation on lysosomal proteins. Mol Cell Proteomics. 2006 Apr;5(4):686-701. 168. Sleat DE, Wang Y, Sohar I, Lackland H, Li Y, Li H, Zheng H, Lobel P. Identification and validation of mannose 6-phosphate glycoproteins in human plasma reveal a wide range of lysosomal and non-lysosomal proteins. Mol Cell Proteomics. 2006 Oct;5(10):1942-56. 169. Sleat DE, Zheng H, Lobel P. The human urine mannose 6-phosphate glycoproteome. Biochim Biophys Acta. 2007 Mar;1774(3):368-72. 170. Qian M, Sleat DE, Zheng H, Moore D, Lobel P. Proteomics analysis of serum from mutant mice reveals lysosomal proteins selectively transported by each of the two mannose 6- phosphate receptors. Mol Cell Proteomics. 2008 Jan;7(1):58-70. 171. Sleat DE, Della Valle MC, Zheng H, Moore DF, Lobel P. The mannose 6-phosphate glycoprotein proteome. J Proteome Res. 2008 Jul;7(7):3010-21. 172. Sleat DE, Wiseman JA, Sohar I, El-Banna M, Zheng H, Moore DF, Lobel P. Proteomic analysis of mouse models of Niemann-Pick C disease reveals alterations in the steady-state levels of lysosomal proteins within the brain. Proteomics. 2012 Dec;12(23-24):3499-509. 173. Sleat DE, Sun P, Wiseman JA, Huang L, El-Banna M, Zheng H, Moore DF, Lobel P. Extending the mannose 6-phosphate glycoproteome by high resolution/accuracy mass spectrometry analysis of control and acid phosphatase 5-deficient mice. Mol Cell Proteomics. 2013 Jul;12(7):1806-17. 174. Noto PB, Bukhtiyarov Y, Shi M, McKeever BM, McGeehan GM, Lala DS. Regulation of sphingomyelin phosphodiesterase acid-like 3A gene (SMPDL3A) by liver X receptors. Mol Pharmacol. 2012 Oct;82(4):719-27. 175. Zhao C, Dahlman-Wright K. Liver X receptor in cholesterol metabolism. J Endocrinol. 2010 Mar;204(3):233-40. 176. Hong C, Tontonoz P. Liver X receptors in lipid metabolism: opportunities for drug discovery. Nat Rev Drug Discov. 2014 Jun;13(6):433-44. 177. Kidani Y, Bensinger SJ. Liver X receptor and peroxisome proliferator-activated receptor as integrators of lipid homeostasis and immunity. Immunol Rev. 2012 Sep;249(1):72-83. 178. Airola MV, Tumolo JM, Snider J, Hannun YA. Identification and biochemical characterization of an acid sphingomyelinase-like protein from the bacterial plant pathogen

164 Ralstonia solanacearum that hydrolyzes ATP to AMP but not sphingomyelin to ceramide. PLoS One. 2014 Aug 21;9(8):e105830. 179. Perisic L, Hedin E, Razuvaev A, Lengquist M, Osterholm C, Folkersen L, Gillgren P, Paulsson-Berne G, Ponten F, Odeberg J, Hedin U. Profiling of atherosclerotic lesions by gene and tissue microarrays reveals PCSK6 as a novel protease in unstable carotid atherosclerosis. Arterioscler Thromb Vasc Biol. 2013 Oct;33(10):2432-43. 180. Junger WG. Immune cell regulation by autocrine purinergic signalling. Nat Rev Immunol. 2011 Mar;11(3):201-12. 181. Mutafova-Yambolieva VN, Durnin L. The purinergic neurotransmitter revisited: a single substance or multiple players? Pharmacol Ther. 2014 Nov;144(2):162-91. 182. Bao L, Locovei S, Dahl G. Pannexin membrane channels are mechanosensitive conduits for ATP. FEBS Lett. 2004 Aug 13;572(1-3):65-8. 183. Cekic C, Linden J. Purinergic regulation of the immune system. Nat Rev Immunol. 2016 Mar;16(3):177-92. 184. Knowles AF. The GDA1_CD39 superfamily: NTPDases with diverse functions. Purinergic Signal. 2011 Mar;7(1):21-45. 185. Kansas GS, Wood GS, Tedder TF. Expression, distribution, and biochemistry of human CD39. Role in activation-associated homotypic adhesion of lymphocytes. J Immunol. 1991 Apr 1;146(7):2235-44. 186. Orengo CA, Michie AD, Jones S, Jones DT, Swindells MB, Thornton JM. CATH--a hierarchic classification of protein domain structures. Structure. 1997 Aug 15;5(8):1093- 108. 187. Ong WY, Herr DR, Farooqui T, Ling EA, Farooqui AA. Role of sphingomyelinases in neurological disorders. Expert Opin Ther Targets. 2015;19(12):1725-42. 188. Pehkonen P, Welter-Stahl L, Diwo J, Ryynänen J, Wienecke-Baldacchino A, Heikkinen S, Treuter E, Steffensen KR, Carlberg C. Genome-wide landscape of liver X receptor chromatin binding and gene regulation in human macrophages. BMC Genomics. 2012 Jan 31;13:50. 189. Padilla J, Jenkins NT, Lee S, Zhang H, Cui J, Zuidema MY, Zhang C, Hill MA, Perfield JW 2nd, Ibdah JA, Booth FW, Davis JW, Laughlin MH, Rector RS. Vascular transcriptional

165 alterations produced by juvenile obesity in Ossabaw swine. Physiol Genomics. 2013 Jun 3;45(11):434-46. 190. Sawada H, Takami K, Asahi S. A toxicogenomic approach to drug-induced phospholipidosis: analysis of its induction mechanism and establishment of a novel in vitro screening system. Toxicol Sci. 2005 Feb;83(2):282-92. 191. Anthérieu S, Rogue A, Fromenty B, Guillouzo A, Robin MA. Induction of vesicular steatosis by amiodarone and tetracycline is associated with up-regulation of lipogenic genes in HepaRG cells. Hepatology. 2011 Jun;53(6):1895-905. 192. Sleat DE, Jadot M, Lobel P. Lysosomal proteomics and disease. Proteomics Clin Appl. 2007 Sep;1(9):1134-46. 193. Cheng J, Goldstein R, Gershenson A, Stec B, Roberts MF. The cation-Ï€ box is a specific phosphatidylcholine membrane targeting motif. J Biol Chem. 2013 May 24;288(21):14863- 73. 194. Rusnak F, Mertz P. Calcineurin: form and function. Physiol Rev. 2000 Oct;80(4):1483-521. 195. Andreini C, Bertini I, Cavallaro G, Holliday GL, Thornton JM. Metal ions in biological catalysis: from enzyme databases to general principles. J Biol Inorg Chem. 2008 Nov;13(8):1205-18. 196. Ago H, Oda M, Takahashi M, Tsuge H, Ochi S, Katunuma N, Miyano M, Sakurai J. Structural basis of the sphingomyelin phosphodiesterase activity in neutral sphingomyelinase from Bacillus cereus. J Biol Chem. 2006 Jun 9;281(23):16157-67. 197. Berger I, Fitzgerald DJ, Richmond TJ. Baculovirus expression system for heterologous multiprotein complexes. Nat Biotechnol. 2004 Dec;22(12):1583-7. 198. Otwinowski Z, Minor W. Processing of X-ray diffraction data collected in oscillation mode. Methods Enzymol. 1997;276:307-26. 199. Adams PD, Afonine PV, Bunkóczi G, Chen VB, Davis IW, Echols N, Headd JJ, Hung LW, Kapral GJ, Grosse-Kunstleve RW, McCoy AJ, Moriarty NW, Oeffner R, Read RJ, Richardson DC, Richardson JS, Terwilliger TC, Zwart PH. PHENIX: a comprehensive Python-based system for macromolecular structure solution. Acta Crystallogr D Biol Crystallogr. 2010 Feb;66(Pt 2):213-21. 200. Emsley P, Cowtan K. Coot: model-building tools for molecular graphics. Acta Crystallogr D Biol Crystallogr. 2004 Dec;60(Pt 12 Pt 1):2126-32.

166 201. Lim SM, Yeung K, Trésaugues L, Ling TH, Nordlund P. The structure and catalytic mechanism of human sphingomyelin phosphodiesterase like 3a--an acid sphingomyelinase homologue with a novel nucleotide hydrolase activity. FEBS J. 2016 Mar;283(6):1107-23. 202. Gibellini F, Smith TK. The Kennedy pathway--De novo synthesis of phosphatidylethanolamine and phosphatidylcholine. IUBMB Life. 2010 Jun;62(6):414-28. 203. Karplus PA, Diederichs K. Linking crystallographic model and data quality. Science. 2012 May 25;336(6084):1030-3. 204. Schuchman EH, Wasserstein MP. Types A and B Niemann-Pick disease. Best Pract Res Clin Endocrinol Metab. 2015 Mar;29(2):237-47. 205. Gulbins E, Walter S, Becker KA, Halmer R, Liu Y, Reichel M, Edwards MJ, Müller CP, Fassbender K, Kornhuber J. A central role for the acid sphingomyelinase/ceramide system in neurogenesis and major depression. J Neurochem. 2015 Jul;134(2):183-92. 206. Saftig P, Sandhoff K. Cancer: Killing from the inside. Nature. 2013 Oct 17;502(7471):312- 3. 207. Becker KA, Riethmüller J, Lüth A, Döring G, Kleuser B, Gulbins E. Acid sphingomyelinase inhibitors normalize pulmonary ceramide and inflammation in cystic fibrosis. Am J Respir Cell Mol Biol. 2010 Jun;42(6):716-24. 208. Leger AJ, Mosquea LM, Li L, Chuang W, Pacheco J, Taylor K, Luo Z, Piepenhagen P, Ziegler R, Moreland R, Urabe A, Jiang C, Cheng SH, Yew NS. Adeno-associated virus- mediated expression of acid sphingomyelinase decreases atherosclerotic lesion formation in apolipoprotein E(-/-) mice. J Gene Med. 2011 Jun;13(6):324-32. 209. Lee JK, Jin HK, Park MH, Kim BR, Lee PH, Nakauchi H, Carter JE, He X, Schuchman EH, Bae JS. Acid sphingomyelinase modulates the autophagic process by controlling lysosomal biogenesis in Alzheimer's disease. J Exp Med. 2014 Jul 28;211(8):1551-70. 210. Olmeda B, García-Álvarez B, Pérez-Gil J. Structure-function correlations of pulmonary surfactant protein SP-B and the saposin-like family of proteins. Eur Biophys J. 2013 Mar;42(2-3):209-22. 211. Gorelik A, Illes K, Superti-Furga G, Nagar B. Structural Basis for Nucleotide Hydrolysis by the Acid Sphingomyelinase-like Phosphodiesterase SMPDL3A. J Biol Chem. 2016 Mar 18;291(12):6376-85.

167 212. Roth AG, Drescher D, Yang Y, Redmer S, Uhlig S, Arenz C. Potent and selective inhibition of acid sphingomyelinase by bisphosphonates. Angew Chem Int Ed Engl. 2009;48(41):7560-3. 213. Zlotogora J, Grotto I, Kaliner E, Gamzu R. The Israeli national population program of genetic carrier screening for reproductive purposes. Genet Med. 2016 Feb;18(2):203-6. 214. Kabsch W. XDS. Acta Crystallogr D Biol Crystallogr. 2010 Feb;66(Pt 2):125-32. 215. Xiong ZJ, Huang J, Poda G, Pomès R, Privé GG. Structure of Human Acid Sphingomyelinase Reveals the Role of the Saposin Domain in Activating Substrate Hydrolysis. J Mol Biol. 2016 Jul 31;428(15):3026-42. 216. Zhou YF, Metcalf MC, Garman SC, Edmunds T, Qiu H, Wei RR. Human acid sphingomyelinase structures provide insight to molecular basis of Niemann-Pick disease. Nat Commun. 2016 Oct 11;7:13082. 217. Hill CH, Cook GM, Spratley SJ, Graham SC, Deane JE. The mechanism of sphingolipid processing revealed by a GALC-SapA complex structure. bioRxiv doi: 10.1101/112029 218. Slotte JP. Biological functions of sphingomyelins. Prog Lipid Res. 2013 Oct;52(4):424-37. 219. Wu J, Hansen GH, Nilsson A, Duan RD. Functional studies of human intestinal alkaline sphingomyelinase by deglycosylation and mutagenesis. Biochem J. 2005 Feb 15;386(Pt 1):153-60. 220. Cortes LK, Vainauskas S, Dai N, McClung CM, Shah M, Benner JS, Corrêa IR Jr, VerBerkmoes NC, Taron CH. Proteomic identification of mammalian cell surface derived glycosylphosphatidylinositol-anchored proteins through selective glycan enrichment. Proteomics. 2014 Nov;14(21-22):2471-84. 221. Gorelik A, Illes K, Heinz LX, Superti-Furga G, Nagar B. Crystal structure of mammalian acid sphingomyelinase. Nat Commun. 2016 Jul 20;7:12196. 222. Fagone P, Jackowski S. Phosphatidylcholine and the CDP-choline cycle. Biochim Biophys Acta. 2013 Mar;1831(3):523-32. 223. Cornell RB, Ridgway ND. CTP:phosphocholine cytidylyltransferase: Function, regulation, and structure of an amphitropic enzyme required for membrane biogenesis. Prog Lipid Res. 2015 Jul;59:147-71. 224. Terwilliger TC, Adams PD, Read RJ, McCoy AJ, Moriarty NW, Grosse-Kunstleve RW, Afonine PV, Zwart PH, Hung LW. Decision-making in structure solution using Bayesian

168 estimates of map quality: the PHENIX AutoSol wizard. Acta Crystallogr D Biol Crystallogr. 2009 Jun;65(Pt 6):582-601. 225. Emsley P, Lohkamp B, Scott WG, Cowtan K. Features and development of Coot. Acta Crystallogr D Biol Crystallogr. 2010 Apr;66(Pt 4):486-501. 226. McCoy AJ, Grosse-Kunstleve RW, Adams PD, Winn MD, Storoni LC, Read RJ. Phaser crystallographic software. J Appl Crystallogr. 2007 Aug 1;40(Pt 4):658-674. 227. Afonine PV, Grosse-Kunstleve RW, Echols N, Headd JJ, Moriarty NW, Mustyakimov M, Terwilliger TC, Urzhumtsev A, Zwart PH, Adams PD. Towards automated crystallographic structure refinement with phenix.refine. Acta Crystallogr D Biol Crystallogr. 2012 Apr;68(Pt 4):352-67. 228. Gorelik A, Heinz LX, Illes K, Superti-Furga G, Nagar B. Crystal Structure of the Acid Sphingomyelinase-like Phosphodiesterase SMPDL3B Provides Insights into Determinants of Substrate Specificity. J Biol Chem. 2016 Nov 11;291(46):24054-24064. 229. Zimmermann H, Zebisch M, Sträter N. Cellular function and molecular structure of ecto- nucleotidases. Purinergic Signal. 2012 Sep;8(3):437-502. 230. Jansen S, Perrakis A, Ulens C, Winkler C, Andries M, Joosten RP, Van Acker M, Luyten FP, Moolenaar WH, Bollen M. Structure of NPP1, an ectonucleotide pyrophosphatase/phosphodiesterase involved in tissue calcification. Structure. 2012 Nov 7;20(11):1948-59. 231. Kato K, Nishimasu H, Okudaira S, Mihara E, Ishitani R, Takagi J, Aoki J, Nureki O. Crystal structure of Enpp1, an extracellular glycoprotein involved in bone mineralization and insulin signaling. Proc Natl Acad Sci U S A. 2012 Oct 16;109(42):16876-81. 232. Albright RA, Ornstein DL, Cao W, Chang WC, Robert D, Tehan M, Hoyer D, Liu L, Stabach P, Yang G, De La Cruz EM, Braddock DT. Molecular basis of purinergic signal metabolism by ectonucleotide pyrophosphatase/phosphodiesterases 4 and 1 and implications in stroke. J Biol Chem. 2014 Feb 7;289(6):3294-306. 233. Gorelik A, Randriamihaja A, Illes K, Nagar B. A key tyrosine substitution restricts nucleotide hydrolysis by the ectoenzyme NPP5. FEBS J. 2017 Nov;284(21):3718-3726. 234. Hausmann J, Kamtekar S, Christodoulou E, Day JE, Wu T, Fulkerson Z, Albers HM, van Meeteren LA, Houben AJ, van Zeijl L, Jansen S, Andries M, Hall T, Pegg LE, Benson TE, Kasiem M, Harlos K, Kooi CW, Smyth SS, Ovaa H, Bollen M, Morris AJ, et al. Structural

169 basis of substrate discrimination and integrin binding by . Nat Struct Mol Biol. 2011 Feb;18(2):198-204. 235. Nishimasu H, Okudaira S, Hama K, Mihara E, Dohmae N, Inoue A, Ishitani R, Takagi J, Aoki J, Nureki O. Crystal structure of autotaxin and insight into GPCR activation by lipid mediators. Nat Struct Mol Biol. 2011 Feb;18(2):205-12. 236. Morita J, Kano K, Kato K, Takita H, Sakagami H, Yamamoto Y, Mihara E, Ueda H, Sato T, Tokuyama H, Arai H, Asou H, Takagi J, Ishitani R, Nishimasu H, Nureki O, Aoki J. Structure and biological function of ENPP6, a choline-specific glycerophosphodiester- phosphodiesterase. Sci Rep. 2016 Feb 18;6:20995.

170