Membrane bound O-acyltransferases and their inhibitors

Naoko Masumoto*†, Thomas Lanyon-Hogg*, Ursula R Rodgers‡, Antonios D Konitsiotis‡§, Anthony I Magee†‡ and Edward W Tate*†

*Department of Chemistry, Imperial College London, South Kensington Campus, London, SW7 2AZ, United Kingdom

†Institute of Chemical Biology, Department of Chemistry, Imperial College London

‡Molecular Medicine Section, National Lung & Heart Institute, Sir Alexander Fleming Building, South Kensington Campus, Imperial College London, SW7 2AZ

§Current address: Max Planck Institute of Molecular Physiology, Department of Systemic Cell Biology, Otto-Hahn-Str. 11, 44227 Dortmund, Germany.

Joint corresponding Authors: Anthony I. Magee, Molecular Medicine Section, National Heart & Lung Institute, Imperial College London, South Kensington Campus, London SW7 2AZ, UK. Email:

Edward W. Tate, Department of Chemistry, Imperial College London, South Kensington, London SW7 2AZ, UK. Email:

Abbreviations

Hh = Hedgehog, UAG = unacylated ghrelin, TMD = transmembrane domain, DAP = (S) 2,3-diaminopropionic acid, cat-ELCCA = catalytic assay using enzyme-linked click chemistry, Shh = Sonic Hedgehog

Key Words

Membrane bound O-acyltransferases (MBOATs), Octanoylation, Palmitoleoylation, Palmitoylation, Ghrelin O-acyltransferase (GOAT), Porcupine (Porcn), Hedgehog acyltransferase (HHAT)

Abstract

Since the identification of the Membrane Bound O-Acyltransferase (MBOATs) protein family in the early 2000s, three distinct members (Porcupine, Hedgehog acyltransferase and Ghrelin O-acyltransferase) have been shown to acylate specific proteins or peptides. In this review, topology determination, development of assays to measure enzymatic activities, and discovery of small molecule inhibitors are compared and discussed for each of these enzymes.

Introduction

In 2000, Hoffmann [1] reported a family of multi-domain membrane spanning acyltransferases responsible for O-acylation reactions called Membrane Bound O-Acyl Transferases (MBOATs) during analysis of the conserved sequence of Porcupine (PORCN), the activity of which is important in Wingless (Drosophila) /Wnt (vertebrates) signalling pathways. Since this discovery, more than ten genes have been identified to encode MBOATs in humans [2].

MBOATs are further categorised into three subgroups based on their biochemical reactions: lipid biosynthesis, sterol acylation, and acylation of secreted proteins/peptides, including the appetite stimulating peptide hormone ghrelin and the Hedgehog (Hh) and Wnt morphogen families [3]. These acylated polypeptides are involved in cellular signalling or subsequent protein-protein interactions which are dysregulated in a range of diseases, thus making MBOATs attractive targets for novel drug discovery.

MBOATs are predicted to have 8-12 transmembrane domains and localise in the protein secretory pathway (ER/Golgi Complex) [2]. Their acyl-CoA substrates are produced during fatty acid beta-oxidation predominantly in the mitochondria and cytosol. In order for catalysis to occur, acyl-CoAs must cross the ER membrane and MBOATs have been suggested also to act as acyl-CoA transporters [4]. Analysis of MBOATs has proven challenging due to their polytopic nature and limitations in available techniques and tools. Investigating the activities of MBOATs using knock-out or transgenic mice has resulted in developmental defects and embryonic lethality [5]. To date, five MBOAT family members have been fully mapped topologically: human ACAT1 (Acyl-CoA:cholesterol acyltransferase, also known as Sterol O-acyltransferase, SOAT), ACAT2, Ghrelin O-acyltransferase (GOAT), Hedgehog acyltransferase (HHAT) and yeast Gup1p [6-11]. Common to all MBOATs are two key residues: a highly conserved asparagine/aspartic acid and an invariant histidine residue (Figure 1) [2]. These residues have been hypothesised to be involved in catalysis; however, to date there is no conclusive evidence that defines the precise location of the catalytic centre.

In the past few years rapid progress has been made in understanding the MBOATs responsible for acylation of secreted polypeptides, aided by the development of methods to study protein acylation such as bioorthogonal ligation techniques [12-14]. GOAT is involved in octanoylation of ghrelin at Ser-3, which increases its potency as an appetite enhancing hormone [7]. Porcupine (PORCN) catalyses primarily palmitoleoylation (C16:1) of Wnt-3a proteins on Ser-209, which enhances Wnt secretion [15, 16]. Hh proteins are irreversibly palmitoylated (C16:0) by HHAT at the N-terminal Cys-24 revealed by secretory signal peptide cleavage [17], which is crucial for biological activity in vertebrates. In this review, recent progress in the understanding of MBOATs and discovery of inhibitors of GOAT, PORCN and HHAT are discussed.

GOAT

Ghrelin was identified by Kojima et al [18] during a search for a binding partner of an orphan G protein coupled receptor (GHS-R1a) which stimulates secretion of growth hormones in the pituitary gland. After cleavage of pro-ghrelin (117 amino acids), ghrelin (28 amino acids) undergoes post-translational octanoylation at Ser-3 in the ER lumen, which is thought to be required for secretion. Although octanoylation of ghrelin augments its potency 1000 fold, the dominant form of ghrelin present in plasma is the unacylated form (unacylated ghrelin, UAG). Initially, UAG was considered biologically non-functional due to its inability to bind GHS-R1a. It was later revealed that non-endocrine activities are induced by UAG upon binding to UAG receptors [19]. Ghrelin is also modified with different acyl chain lengths, such as decanoate (C10:0) and decenoate (C10:1) [20], the function of which is unknown; however, emerging evidence suggests these variations are partially due to nutrient intake. As expected from its function, expression of GOAT is highly restricted to major ghrelin releasing tissues, such as the stomach and intestine.

GOAT topology and key residues

Cole et al reported the first comprehensive study on GOAT topology [7], demonstrating GOAT has 12 distinct hydrophobic regions with 11 transmembrane domains (TMDs) and one re-entrant loop (RL), each separated by relatively short hydrophilic loops (Figure 2A). It has short terminal tails, in the lumen at the N-terminus and the cytosolic C-terminus. The invariant histidine (His-338) is located on the luminal side and the conserved asparagine (Asn-307) is on the cytosolic side. It is predicted that His-338 is likely to be involved in the active site, whereas Asn-307 is unlikely to be involved in catalysis, although it might be important for substrate interactions and transport or protein structural stability. Photocrosslinkable acyl ghrelin analogues can bind at the C-terminal region of GOAT indicating that the peptide and octanoyl-CoA interaction may occur near the C-terminus, although the identification of the exact catalytic site was unsuccessful. Although MBOATs are known to form oligomers in vitro and in cell culture, purified GOAT in detergent micelles exists as monomers and ghrelin and its analogues bind to monomeric purified GOAT.

GOAT inhibitors

It has been previously demonstrated that GOAT is regulated by nutrient availability and its activity mediates the impact on body adiposity [4, 21, 22]. So far, three different types of GOAT inhibitors have been discovered: peptide-based analogues, bisubstrate analogues and small molecules [23]. Yang et al exploited peptidomimetics by substituting the first pentapeptide of the GOAT recognition motif of ghrelin (GSSFL) with different amino acids [24]. Inhibition in vitro was significantly increased with amidated full-length ghrelin (2) (IC50=0.2 µM) or a pentapeptide containing octanoylated (S)2,3-diaminopropionic acid (DAP) in place of Ser-3 (IC50=1.0 µM) (3). However, these compounds pose pharmacologic challenges for in vivo applications and are likely to act as potent agonists of GHS-1a. A transition state mimic of Ser-3 octanoylation, BK-1114 (4), is effective in the micromolar range on isolated enzyme and in intact cells [25]. Cole et al introduced compounds inspired by bisubstrate GO-CoA-Tat inhibitors (5a-5c), on the premise that GOAT might form a ternary complex with both substrates [26], consisting of octanoate, CoenzymeA (CoA) and the first 10 amino acids of ghrelin linked irreversibly with an amide linkage (octanoate to the ghrelin) and a thioether linkage (alpha-carbon of octanoate to CoA) (Figure 3). Validated in vitro and in vivo, the outcome was especially encouraging in mice (reduction in weight gain and improvement in glucose intolerance) despite their limited pharmacological utility in vivo due to large size, polarity and cell penetration. Janda et al have developed and utilised cat-ELCCA (catalytic assay using enzyme-linked click chemistry) to identify the first small molecule inhibitors containing a naphthalene core structure (6) [27, 28]. Validation of these compounds in cell based assays is still pending; however, this assay format has potential as a platform to find new small molecule inhibitors for other MBOATs.

PORCN

Secreted Wnt proteins play key roles in embryonic development, tissue homeostasis and stem cell self-renewal; among the 19 Wnt proteins encoded in humans, Wnt3a is the most extensively studied isoform [29]. PORCN catalyses palmitoleoylation of Wnt signalling proteins at a serine residue (209 in human Wnt3a). An early report that PORCN palmitoylated Wnts at a cysteine residue in the N-terminal region (Cys-77 in Wnt3a) has been shown to be erroneous [30]. The bent conformation of palmitoleate may provide the appropriate three dimensional conformation for interaction of Wnt with Wntless (WIs) which packages Wnt into exosomes for secretion [31]. The palmitoleate acts as an anchor to a hydrophobic groove of the receptor Frizzled [32] where Wnt binds after travelling through the extracellular space. Moreover, it appears that Wnt proteins can be modified with various lengths of acyl chains (typically C13-16) with or without saturation [33], which may be involved in gradient formation or differential regulation of Wnt ligand transport [16, 34].

PORCN topology and key residues

According to topology prediction by MEMSAT-SVM, Porcn has 11 TMDs with the conserved Asn-306 in a cytosolic loop and the invariant His-341 embedded in TMD 9 (Figure 2B) [34]. Interestingly, PORCN itself is palmitoylated at cytosolic Cys-187[33]. Identification of key residues and regions involved in catalysis were accomplished using alanine-scanning mutagenesis. Within TMD 9, mutants N306A and W312A did not alter the activity of PORCN. Therefore, the conserved Asn-306 is not involved in catalysis. W305A, Y316A and Y334A showed moderate defects in activity (30-50%). S337A, L340A and H341A had little or no activity (<20% WT), indicating that these residues are critical for acyltransferase activity. Confocal imaging demonstrated that impaired enzymatic activity was not due to PORCN misfolding or mislocalisation. Co-immunoprecipitation with Wnt-3a confirmed that mutants Y334A, S337A and H341A were not able to bind the protein substrate; however, L340A was capable of binding Wnt3a, suggesting that Leu-340 may be involved in fatty acid recognition rather than Wnt binding.

PORCN inhibitors

Chen et al identified highly selective PORCN inhibitors termed ‘Inhibitors of Wnt Production’ (e.g. IWP-2, (7)) which can be used in a variety of in vitro settings including tissue engineering and stem cell biology [35]. Due to limited bioavailability, these are not compatible with in vivo studies; however, recent studies show that PORCN can accommodate chemically diverse scaffolds [36]. Among them, IWP12 (8) is effective in zebrafish as evidenced by the loss of Wnt activity, and C59 (9), disclosed in a Novartis patent [37], possesses nanomolar activity and was found to inhibit Wnt signalling and growth of a Wnt-driven breast cancer cell line [37]. Similar to the C59 scaffold, LGK974 (Novartis) (10) has entered phase I clinical trials (NCT01351103) for treatment of malignant cancers [38]. Liu et al observed the regression of Wnt-driven tumours absent from the formation of abnormal histopathological defects and delay in tumour growth [39].

HHAT

Hedgehog (Hh) proteins are involved in development and tissue homeostasis in adult organisms and are unique in that they are post-translationally modified by two lipids during their maturation process [40]. Addition of palmitate at the N-terminal cysteine is catalysed by HHAT [41] and cholesterol attachment occurs at the C-terminus after autocatalytic cleavage of a non-signalling domain [42]. It has been thought that dual lipidation of Hhs improves membrane affinity and contributes to formation of extracellular multimeric complexes, which translates to enhanced signalling activity [43]. A major role of palmitoylation is to direct Hh proteins to specific membrane domains and to establish long-range signalling upon formation of soluble multimeric complexes [5]. The importance of palmitate for Hh signalling activity was demonstrated in rodent ventral forebrain formation [44] where removal of the palmitoylation site abolished the induction of neuronal cell differentiation. On the other hand, cholesterol provides affinity for cell membranes, regulates cell surface distribution, and establishes the extracellular range and concentration gradient. Reports on the role of each type of lipid modification and compositions of multimeric complexes frequently present conflicting evidence [45-47]. Like PORCN, HHAT can accept various different lengths of acyl-CoA as a substrate [41] in vitro and in cell-based assays in our lab (unpublished data). In vitro, Hh proteins showed preference for modification by shorter acyl-CoAs [48]. The dominant form of acyl-CoA in vivo is usually palmitoyl-CoA but other lengths are present and may modify Hh depending on the local availability of different acyl-CoAs.

HHAT topology and key residues

The Tate/Magee and Resh groups recently reported the first detailed analyses of HHAT topology [8, 9], which were independently obtained by somewhat different experimental approaches but produced remarkably harmonious results. The current working topology model consists of ten TMDs and two RLs (Figure 2C). The invariant His-379 is between TM9 and 10 at the luminal side while the Asp-339 is within the cytosol. Point mutagenesis experiments showed that HHAT activity was severely lost in the D339N mutant whereas the H379A mutant retained ~50% activity, in agreement with observations from Buglino et al [49]. In our study, it was demonstrated that HHAT itself is also palmitoylated, similar to PORCN but at four distinct cytoplasmic sites. Mutagenesis experiments showed that D339 is more important than H379 for HHAT palmitoylation [8], and that the state of HHAT palmitoylation also affects HHAT activity.

HHAT Inhibitors

Resh et al identified the first candidate HHAT inhibitors (11-14) from a target-orientated high throughput screen [50, 51], opening a new avenue to explore the impact of palmitoylation inhibition on Hh transport and signal transduction. Hh pathway inhibitors targeting upstream components such as HHAT may also be useful to combat development of resistance at the downstream components during chemotherapy [52, 53]. The compound most used to date in the literature, RU-SKI-43, has been investigated in cell based assays including various cancer cell lines; however, it showed limited utility in vivo (half-life in mouse plasma of 17 minutes) [50]. To circumvent this issue, Panc-1 cells stably expressing shRNAs against HHAT, Shh and a control scrambled sequence were injected into immunocompromised mice [54] and over 72 hours of treatment tumour growth was inhibited by 70% by both HHAT and Shh depletion. Moreover, in combination with the antiproliferative compound Rapamycin, cell proliferation was inhibited further than with individual depletion. These data should be treated with caution, based on two independent studies on Hh signalling in pancreatic cancer that were reported recently in which tumour growth was substantially enhanced after inhibition of Hh signalling, using a variety of mouse models [55-57]. Although inhibition of Hh signalling leads to disruption in paracrine signalling and stroma desmoplasia, it is not therapeutically beneficial as the stroma appears to physically restrain tumour growth. In our lab, RU-SKI-43 showed a very narrow therapeutic window due to significant off-target toxicity; however, modified RU-SKI compounds were more potent in various in vitro and cell-based assays with lower cell toxicity (unpublished data).