Protein Acyl Thioesterases (Review)

Protein acyl thioesterases (Review)

Running title: Protein acyl thioesterases

Ruth Zeidman, Caroline S. Jackson and Anthony I. Magee

Address:

Molecular Medicine

National Heart & Lung Institute

SirAlexanderFlemingBuilding

South Kensington

ImperialCollegeLondon

London SW7 2AZ

UK

Corresponding author: Anthony I. Magee

E-mail:

Telephone: +44 20 7594 3135

Keywords

APT1

Palmitoylation

Protein acyl thioesterase

Protein acyl transferase

S-acylation

Abstract

Many proteins are S-acylated, affecting their localization and function. Dynamic S-acylation in response to various stimuli has been seen for several proteins in vivo. The regulation of S-acylation is beginning to be elucidated. Proteins can autoacylate or be S-acylated by protein acyl transferases (PATs). Deacylation, on the other hand, is an enzymatic process catalyzed by protein thioesterases (APT1 and PPT1) but only APT1 appears to be involved in the regulation of the reversible S-acylation of cytoplasmic proteins seen in vivo. PPT1, on the other hand, is involved in the lysosomal degradation of S-acylated proteins and PPT1 deficiency causes the disease infant neuronal ceroid lipofuscinosis.

S-acylation

Proteins are modified in many ways as a means to control their localisation and function. Many proteins that exert their function at or near a cellular membrane are modified by the covalent attachment oflipids. S-acylation is the post-translational attachment of fatty acids to proteins through a thioester bond, usually on a cysteine residue. Palmitic acid, a 16-carbon saturated fatty acid, is the most common fatty acid to be linked to proteins in this manner, explaining why S-acylation is often referred to as palmitoylation, although proteins can also be modified in this manner by many other fatty acids, both saturated and unsaturated, including myristic, oleic, arachidonic and stearic acids.

A wide variety of proteins are acylated, including transmembrane proteins like G-protein coupled receptors, T cell co-receptors CD4 and CD8 and ion channels, cytosolic and membrane-associated proteins like members of the Ras and Src families, heterotrimeric G-proteins, endothelial nitric oxide synthase (eNOS), SNARE vesicle fusion proteins, as well as secreted and viral proteins.Addition of fatty acids to a protein has various effects on signalling, trafficking, protein stability, protein-protein interaction and membrane and membrane subdomain association[1-3].

In the cases where the acylation takes place on an N-terminal cysteine, the fatty acid is often found linked to the protein through an amide bond to the N-terminus. This type of acylation is referred to as N-acylation and often occurs in secreted proteins, for example Hedgehog/Sonic hedgehog [4], but also in Gαs[5]. This may well occur in two steps; S-acylation via a thioester bond between the fatty acid and the sulphur in the cysteine, followed by an internal S to N acyl shift to the α-amino group.

S-acylation is a reversible modification

One interesting difference between S-acylation and N-acylation is the reversibility of the S-acylation. In fact, S-acylation is the only lipid modification of proteins that is readily reversible, seen as the half time of theattachment of the fatty acidside chainbeing significantly shorter than the life time ofthe protein.Many proteins have been shown to be dynamically S-acylated in vivo with different turnover rates.For instance, S-acylation of Fyn, a member of the Src kinase family, is required for its plasma membrane localisation and has been shown to be reversible, with a half life of 1.5-2 h[6]. The unpalmitoylated pool of Fyn was found in intracellular membranes. Another Src family kinase, Lck, has also been demonstrated to be reversibly palmitoylated in pervanadate-stimulated human T blasts, with a half life of 15-30 minutes (Fig. 1B; unpublished work, C.S. Jackson and A.I. Magee). Ras proteins were one of the first proteins to be reported to have dynamic S-acylation[7] and different H-Ras variants are also deacylated at variable rates. Oncogenic H-Ras mutants, which are GTP-bound to a higher proportion than wild-type H-Ras, were determined to have shorter palmitate half lives than the wild type H-Ras, even though the steady state levels of palmitoylation did not differ[8].

(Insert Fig 1 near here)

For some proteins, theincreased acyl turnover seen on many proteins in vivois in response to various stimuli. Onesuch example is the acylation turnover of the neuronal protein PSD-95, which is increased upon neuronal activity and causes a clustering of PSD-95 at the synapse[9]. eNOS deacylation is more efficient in the presence of Ca2+-calmodulin, an eNOS activator [10]. Many G-protein coupled receptors - including the β2-adrenergic, α2A-adrenergic, m2muscarinic acetylcholine, 5-HT4a serotonin, V1a vasopressin and δ opioid receptors - are also dynamically acylated after agonist binding[11-16]. The effects of increased fatty acyl turnover on the receptors are not clear. The steady state levels of acylation will depend on the relative reacylation and deacylation rates. Even if the steady state levels are unchanged, under conditions with of fatty acyl turnover there will be a proportion of transiently deacylated receptors. The relative size of this pool and the time the receptors stay deacylated might be a mechanism for controlling signalling through the receptors. In fact, receptor desensitization by phosphorylation is increased in depalmitoylated receptors compared to their palmitoylated counterparts [17-19].

The turnover of palmitate on Gα subunits is also regulated by stimulation of their associated receptors[20-24] as activation of the receptor leads to the dissociation of the α subunit from the βγ subunits, making the cysteine on the α subunit available for depalmitoylation by the thioesterase APT1[25].

Autoacylation

Even though proteins have been known to be modified by fatty acids and the effects of acylation on localization and signalling have been known for many years, the mechanism behind the actual process of attaching the acyl chains has not been very clear. Many proteins and peptides can autoacylate when incubated with the appropriate acyl-CoA in vitro. Myristoylated Giα1 is autopalmitoylated in the presence of palmitoyl-CoA at the same cysteine resdidue that is acylated in vivo[26].Similarly, the autoacylation of other proteins have been reported, including rhodopsin [27], SNAP-25 [28], carbamoyl-phosphate synthetase 1 [29], Bet3 [30],β2-adrenergic receptor [31] and c-Yes [32]. Autoacylation does not seem to be a universal mechanism as, for example, GAP-43 and Fyn do not autoacylate under the same conditions asGiα1does [26]. In addition, not all of these in vitro reactions take place at physiological pH and/or acyl-CoA concentrations. In some cases the reaction times are also too long to accommodate the fast palmitate cycling seen on many proteins in vivo.

DHHC PATs

It seems more likely that, at least for most proteins, S-acylation is an enzymatic process. In recent years there has been much progress in identifying protein-acyl transferases (PATs). Two families of these enzymes are the subjects of other reviews in this issue. Briefly, one such group of PATs is the DHHC family, consisting of multi-spanning transmembrane proteins that have a cysteine-rich domain containing a conserved asparatate-histidine-histidine-cysteine (DHHC) motif, with is essential for the PAT activity[33, 34]. The PAT activity of the DHHC proteins was originally discovered in yeast, where the DHHC protein Erf2, together with Erf4, mediates the acylation of yeast Ras [33];another DHHC protein, Akr1, is a PAT for yeast casein kinase 2 [34].The palmitoylation of specific proteins could not be detectedin yeast strains deficient in different combinations of the yeast DHHC proteins, confirming the in vivo acyl-transferase activity of the DHHC proteins[35]. In a screen on potential substrates, the 23 DHHC proteins predicted in the mammalian genome had variable substrate specificities, with some overlap [36].The DHHC PAT substrates are intracellular proteins[1, 37, 38] including: H-and N-Ras which are S-acylated by DHHC9 together with a Golgi-associated protein GCP16, echoing the yeast Erf2/Erf4 complex [39];the DHHC21 substrate eNOS[40];huntingtin which is acylated by HIP14/DHHC17[41, 42]; theγ2 subunit of the GABAA receptor which is acylated by GODZ/DHHC3 [43]and the neuronal PSD-95 protein, which in slightly contradictory reports has been suggested to be the substrate of DHHC2, 3, 7 and 15 (but not DHHC17) [36] or of HIP14/DHHC17, in addition to other neuronal substrates[41].

MBOAT PATs

A different group of PATs appear to have secreted proteins and peptides as their substrates. Hedgehog (Hh) proteins e.g. Sonic hedgehog (Shh), Spitz, Wnts and ghrelin require acylation for their function and are acylated by members of the MBOAT (membrane-bound-O-acyltransferase) family [44]. A common feature of this group of proteins is that they have several membrane-spanning domains and that they transfer organic acids,typically fatty acids, onto hydroxyl groupsof membrane-embedded targets[45]. Only a small subset of members is known to transfer fatty acids and other lipids to proteins. Rasp/Skinny hedgehog/Hedgehog acyl transferase (Hhat) is the PAT for Hh/Shh [46]and Spitz [47], Porcupine (Porc) is the PAT for Wnts[48] and GOAT is the PAT for ghrelin[49].The acylation of secreted proteins is reviewed elsewhere in this issue.

Ykt6-type PATs

A third mechanism for fatty acidattachment to proteins incorporates elements of both autoacylation and enzymatic acyl transfer. The yeast SNARE protein Ykt6binds palmitoyl-CoA at an N-terminal longin domain and autoacylates cysteines at its C-terminus[50] in addition to mediating the palmitoylation of Vac8, presumably in a non-enzymatic way as equimolar amounts of Ykt6 and Vac8 are needed [51]. The exact mechanism behind Vac8 palmitoylation is not clear, and further complicated by the fact that the DHHC protein Pfa3 can palmitoylate both Vac8 [52] and Ykt6 [53]. One possibility is that Ykt6 could act as a co-factor for Pfa3 in the way that Erf4 does for Erf2, although this remains to be elucidated.

Regulation of deacylation

It is thus not completely clear if all fatty acylations are enzymatic and if they are regulated.This raises the possibility that the level of regulation of acylation lies on the deacylation step, especially for those proteins where autoacylation has been suggested. Deacylation of proteins appears to be an enzymatic process. Despite the fast progress that has been made, and is continously being made, in identifying and characterizing PATs, much less is known about the thioesterases that deacylate proteins. So far, only two thioesterases have definitively been shown to catalyze the removal of fatty acids from proteins, and out of those two, only one appears to be involved in regulation of the dynamic acylation cycles seen for many intracellular proteins.

Acylprotein thioesterases

Acyl protein thioesterase 1 (APT1) was originally isolated from rat liver as a lysophospholipase [54], but was later shown to have a preferred substrate specificity for palmitoylated proteinsover lysophosphatidylcholine and palmitoyl-CoA [25].It is a cytosolic protein [25] with a widespread tissue distribution [55]. Several proteins have been identified as APT1 substrates inin vitroassays, like Ras [25], various heterotrimeric G protein α subunits[25, 56], eNOS [10], RGS4 [25] andSNAP-23 [57]. Viral proteins, like the spike glycoprotein E2 from the Semliki Forest virus, HEF and HA from influenza viruses and the Gprotein from vesicular stomatitis virus (VSV) are also deacylated by APT1 in in vitro assays [58], although whether this has any biological significance remains to be elucidated.

Cell-basedexperimentshave validated APT1 as a thioesterase. In yeast cells deficient in Apt1p, the yeast homologue of APT1,no Gαi1deacylationtakes place [56], confirming its status as an APT1 substrate. In permeabilized platelets,infusion of APT1 protein causes an almost 50% reduction of the steady state levels of acylation and, more specifically,a reduction of the acylation of Gαq and concomitant reductionof membrane association[59]. Co-expression of APT1 and eNOS increases the level of eNOS depalmitoylation compared with expressing eNOS alone [10]. Furthermore, the dislocation of the APT1 substrates SNAP-23 and a mutant Gsα from the plasma membrane has also been seen on elevating APT1 levels in cells [59, 60].

There is nospecific consensus sequence for S-acylation, although some patterns of S-acylation exist, for instance N-terminal dual S-acylation and myristoylation, C-terminal dual S-acylation and prenylation, and multiple S-acylation at cysteine string motifs [1]. Following on from the lack of distinct S-acylation motifs, there is also no clear consensus for the sequences surrounding the thioacyl group recognized by APT1. In fact, APT1 appears very promiscuous in its substrate specificity. In vitro, APT1 can depalmitoylate structurally different proteins, both soluble intracellular proteins, like eNOS and Gαi1, which are S-acylated oncysteinesin proximity to an N-terminal myristoylation site [61, 62] and Ras, which is modified at its C-terminus by both acylation and prenylation [63], as well as integral membrane proteins. SNAP-23 is S-acylated at multiple cysteines located in the transmembrane domain [64, 65] and the viral proteins E2, HEF, HA and VSVG protein areS-acylated at the interface between the transmembrane domain and the intracellular part of the proteins [66, 67].

APT1 does not, however, deacylate proteins without any discrimination. Caveolin, for instance, an integral membrane protein which is acylatedon cysteine residues in the intracellular C-terminal domain [68] is not deacylated by recombinant APT1 under conditions where eNOS deacylation is readily seen [10]. Moreover, not all substrates are deacylated with the same efficiency. Rat APT1 is 10-fold more efficient in deacylating Gαi1compared to Ras. This preference for Gα subunits over Ras is even more pronounced in yeast APT1, where the difference is 70-fold [56].Also, the activational status of the substrates may influence the catalytic efficiency displayed by APT1 towards it; free Gαi1 is a more readily deacylated by APT1 than heterotrimeric Gαi1 [56].

Another explanation for the variable substrate preference could be the possibility that APT1 could haveselectivity for different fatty acids linked to its substrates. This is,however, contradicted by the fact that two good viral APT1 substrates, the Semliki Forest virus (SFV) proteinE2 and the influenza virus protein HEF,are acylated almost exclusively with palmitic and stearic acid, respectively, whereas the SFV E1 protein, which is mainly palmitoylated [69], is a poor APT1 substrate[58].

The crystal structure of human APT1 has been solved and shows that the enzyme is a member of the large enzyme group α/βhydrolases, which also includeslipases, esterases and dehalogenases, and hasa classical catalytic triad made up of Ser-114, His-203, and Asp-169in APT1 [70]. A Blast search of the APT1 sequence reveals homologues in a wide variety of species, including humans and other mammals as well as lower organisms such asDrosophila melanogaster, Saccharomyces cerevisiae, Caenorhabditis elegans, Arabidopsis thaliana and Mycobacterium leprae, implying anessential role for ATP1 as it is so well conserved throughout evolution.

APT1 inhibitors have been developed based on the structure of the H-Ras C-terminus and in vitro assays show promise for some of the compounds, with IC50 values in the low nanomolar ranges [71]. Their effects in vivo have not yet been determined but they could provide a useful tool for studying the role of APT1 in cells. In general, the field of protein acylation has suffered from a lack of specific potent inhibitors so these compounds could be an important addition to the toolbox.

A homologue of APT1, called lysophospholipase II or APT2 with 64% identity to APT1, has been cloned[55]. It was reported to have activity against several lipid substrates, with varying efficiencies. APT2 mRNA transcripts can be found in a wide variety of tissues, suggesting it is ubiquitously expressed[55]. An interesting question that has not yet been answered is whether APT2, just like APT1, is a cytosolic protein thioesterase involved in protein acylation regulation. In fact, in extracts from yeast strains where the APT1 gene is disrupted, the thioesterase activity against acylated H-Ras is similar to that of extracts from wild-type yeast, despite deacylation of Gαi1 being almost completely abolished [56]. It is tempting to speculate that APT2 could be responsible for this residual thioesterase activity. However, Blast searches using either yeast APT1 or human APT2 sequences do not suggest that there is an APT2 protein in yeast (R. Zeidman and A.I. Magee). Instead, the Ras deacylation could be caused by another still unidentified thioesterase activity.

A Blast search of the human APT1 sequence reveals another homologue of APT1, with 31% identity to APT1 and with the catalytic triad conserved (R. Zeidman and A.I. Magee). This protein was named lysophospholipase-like 1when its gene was discovered during the sequencing of chromosome 1 [72]. As of yet there are no reports of this protein’s biological activity and substrates, but if any thioesterase activity of this protein will be identified, we suggest that it should be called acyl protein thioesterase like 1 (APTL1) in accordance with the naming of APT1.

APT1 was purified from the soluble S100 fraction of a rat liver homogenate [25] and it does not contain any predicted transmembrane domains or obvious sites for lipid modification. The substrates of APT1 are, however, membrane-associated, which raises the question of how APT1 can access its substrates efficiently. One answer could be that APT1 might interact with membraneproteins, either its direct substrates or other proteins that bring APT1 in close proximity to its substrates. Our own data suggest that a significant proportion of APT1 and APT2 sediment with cellular membranes from cultured cells (R. Zeidman and A.I. Magee, unpublished work), supporting this speculation.

Protein-palmitoyl thioesterases

Palmitoyl-protein thioesterase 1 (PPT1) was first isolated from bovine brain extracts based on its ability to depalmitoylate [3H]-palmitate-labeled H-Ras [73] and was also found to depalmitoylate Gα subunits and acyl-CoAsin vitro, with a preference for a chain length between 14 and 18 carbons[74]. Further expression studies revealed that PPT1 is a lysosomal enzyme[75]and it is therefore unlikely to play a role in regulated deacylation of cytoplasmic proteins. The gene encoding PPT1 was found to be located on chromosome 1p32, the region linked to theneurodegenerative disease infantile neuronal ceroid lipofuscinosis (INCL), and as mutations in the PPT gene were found in INCL patients, mutated PPT1 was identified as the underlying cause of INCL [76]. One of the features of INCL is the accumulation of granular deposits inside the cells. [35S]cysteine-labeled lipid thioesters accumulate in immortalized lymphoblasts from patients with INCL, and this accumulation can be reversed by adding recombinant PPT1 to the cells [77], which is taken up and trafficked to lysosomes[78]. The normal function of PPT1 is thus to remove the acyl chains from proteins being degraded in the lysosome.

A homologue of PPT1, called PPT2, has been cloned. It is also a lysosomal thioesterase, but has a substrate specificity for palmitoyl-CoA and not palmitoylatedproteins [79] and therefore likely does not act as a thioesterase for acylated proteins during the degradation process.

Deregulation of S-acylation implicated in diseases