ABHD17 proteins are novel protein depalmitoylases that regulate N-Ras palmitate turnover and subcellular localization

Dynamic changes in protein S-palmitoylation are critical for regulating protein localization and signaling. Only two enzymes - the acyl-protein thioesterases APT1 and APT2 – are known to catalyze palmitate removal from cytosolic cysteine residues. It is unclear if these enzymes act constitutively on all palmitoylated proteins, or if additional depalmitoylases exist. Using a dual pulse-chase strategy comparing palmitate and protein half-lives, we found knockdown or inhibition of APT1 and APT2 blocked depalmitoylation of Huntingtin, but did not affect palmitate turnover on postsynaptic density protein 95 (PSD95) or N-Ras. We used activity profiling to identify novel serine hydrolase targets of the APT1/2 inhibitor Palmostatin B, and discovered that a family of uncharacterized ABHD17 proteins can accelerate palmitate turnover on PSD95 and N-Ras. ABHD17 catalytic activity is required for N-Ras depalmitoylation and re-localization to internal cellular membranes. Our findings indicate that the family of depalmitoylation enzymes may be substantially broader than previously believed. DOI: http://dx.doi.org/10.7554/eLife.11306.001


Introduction
Protein S-palmitoylation involves the post-translational attachment of the 16-carbon fatty acid palmitate to cysteine residues (Conibear and Davis, 2010;Salaun et al., 2010). While a survey of palmitoylation dynamics indicated the bulk of the palmitoyl-proteome is stably palmitoylated (Martin et al., 2011), rapid and constitutive palmitate turnover has been shown for several proteins, including the Ras GTPases, heterotrimeric G proteins, the neuronal post-synaptic density protein PSD95, and the Lck kinase (Magee et al., 1987;Degtyarev et al., 1993;El-Husseini et al., 2002;Zhang et al., 2010). Dynamic changes in palmitoylation modulate protein localization and trafficking and can be regulated in response to cellular signaling (Conibear and Davis, 2010).
Palmitoylation is mediated by a family of DHHC (Asp-His-His-Cys) proteins (Greaves and Chamberlain, 2011a), whereas the only enzymes identified to date that remove palmitate from cytosolic cysteines, the acyl-protein thioesterases (APTs) APT1 and APT2, are related members of the metabolic serine hydrolase (mSH) superfamily (Duncan and Gilman, 1998;Tomatis et al., 2010;Long and Cravatt, 2011). The b-lactone core-containing compound Palmostatin B (PalmB) potently inhibits these enzymes and blocks depalmitoylation of N-Ras and other proteins (Dekker et al., 2010;Rusch et al., 2011). Hexadecyl fluorophosphonate (HDFP) inhibits a subset of mSHs including APT1 and APT2 and also suppresses palmitate turnover (Martin et al., 2011). However, it is unclear if APT1 and APT2 are the only palmitoylthioesterases responsible for the depalmitoylation of cytosolic proteins (Davda and Martin, 2014). eLife digest Proteins play important roles in many processes in cells. Some of these proteins can be modified by the addition of a molecule called palmitate. This process, termed "palmitoylation", helps direct these proteins to the compartments within the cell where they are needed to carry out their roles. One target of palmitoylation is N-Ras, which is a protein that can promote the development of cancer.
We understand quite a lot about how palmitate is added to proteins, but much less about how it is removed. So far, researchers have only identified two enzymes -known as APT1 and APT2 -that can remove palmitate from proteins, but it is possible that there are others. Identifying other "depalmitoylase" enzymes could help us find ways to block the removal of palmitate from N-Ras, which could lead to new treatments for some cancers.
Lin and Conibear used several biochemical techniques to search for depalmitoylase enzymes in human cells. The experiments reveal that although APT1 and APT2 are important for removing palmitate from some proteins, they are not needed to remove palmitate from N-Ras. Instead, Lin and Conibear found that an enzyme called ABHD17 removes palmitate from N-Ras. The next step following on from this work will be to find out what other proteins ABHD17 acts on in cells. A longer-term challenge will be to develop specific chemicals that inhibit ABHD17 activity and test if they are able to reduce the growth of cancer cells. effect on PSD95 or N-Ras depalmitoylation when used alone (Figure 2-figure supplement 1B,C) or together ( Figure 2C,D). Double RNAi knockdown of APT1 and APT2 significantly inhibited N-HTT depalmitoylation ( Figure 2B) and also reduced palmitate turnover on GAD65 (Figure 2-figure supplement 1D) but not PSD95 or N-Ras ( Figure 2C,D). These findings, which are consistent with a recent report showing APT1/2-independent depalmitoylation of R7BP (Jia et al., 2014), strongly suggest that although APT1 and APT2 are responsible for depalmitoylating some proteins (N-HTT, GAD65), depalmitoylation of other cellular substrates, including PSD95 and N-Ras, involves other enzymes.
Previous studies suggested that APT1, APT2, and PPT1 were the sole mSHs targeted by PalmB (Rusch et al., 2011), whereas HDFP inhibited additional mSHs (Martin et al., 2011). In pulse-chase experiments, we found HDFP robustly inhibited the depalmitoylation of N-Ras, PSD95, and N-HTT ( Figure 3A-C). Because palmitate removal from N-Ras and PSD95 does not require APT1 or APT2, their depalmitoylation may be mediated by a distinct mSH that is a common target of both PalmB and HDFP. To identify overlapping targets, we defined a set of 19 candidate mSHs that showed >25% inhibition by HDFP (Supplementary file 1; Martin et al., 2011) but excluded known proteases and mSHs with established luminal activity. We added to this list APT1L, which was previously implicated in BK channel depalmitoylation (Tian et al., 2012) but whose HDFP sensitivity was unknown. The PalmB sensitivity of each enzyme was evaluated by a competitive activity-based protein profiling (cABPP) assay, in which binding of an inhibitor occludes the enzyme active site and prevents labeling with the activity probe fluorophosphonate-rhodamine (FP-rho) ( Figure 3D; Kidd et al., 2001). As expected, PalmB significantly reduced FP-rho labeling of both APT1 and APT2 ( Figure 3E,H). In contrast, it had little effect on the labeling of seven candidates ( Figure 3F,H), highlighting the distinct substrate specificities of PalmB and HDFP. Four mSHs did not label with FP-Rho due to low activity or expression and could not be assessed (Supplementary file 1). Notably, PalmB potently inhibited seven candidates: FASN, PNPLA6, ABHD6, ABHD16A, and ABHD17A/B/C ( Figure 3G,H). Thus, PalmB has additional serine hydrolase targets beyond APT1 and APT2 that may function as protein depalmitoylases. The set of candidates inhibited by both PalmB and HDFP ( Figure 3G,H) includes ABHD6, which associates with PSD95-containing complexes at synapses (Schwenk et al., 2014), and FASN, which functions in palmitoyl-CoA synthesis (Wakil, 1989). However, treatment with the ABHD6 inhibitor WWL70 (Li et al., 2007) or the FASN inhibitor C75 (Kuhajda et al., 2000) did not alter PSD95 depalmitoylation ( Figure 3-figure supplement 1A,C). Palmitate turnover on PSD95 was also unaffected by RHC-80267, which moderately inhibited ABHD6 and PNPLA6 ( Figure 3-figure supplement 1B, D; Hoover et al., 2008). Thus, ABHD6, PNPLA6, and FASN are unlikely to play a primary role in PSD95 depalmitoylation.
Selective inhibitors that target the remaining four candidates have not been identified. Therefore, we used pulse-chase click chemistry to test if increased expression of these enzymes enhances palmitate turnover. High levels of ABHD16A, ABHD6, or APT1/2 had little effect on N-Ras ( Figure 4A  We focused on ABHD17A, which showed the strongest effect in promoting palmitate turnover on N-Ras and PSD95. The ABHD17 proteins are targeted to membranes by a palmitoylated N-terminal cysteine cluster (Kang et al., 2008;Martin and Cravatt, 2009). We found ABHD17A localized to the plasma membrane and to Rab5-and Rab11-positive endosomes (     the N-terminal amino acid residues 1-19 (DN; Figure 4B) shifted it to the cytosol (Figure 4-figure supplement 2B,C) and reduced its catalytic activity ( Figure 4C). Importantly, neither mutant stimulated N-Ras or PSD95 depalmitoylation ( Figure 4D, Figure 4-figure supplement 1B), suggesting both the catalytic activity and membrane localization of ABHD17A are functionally important.
We next examined the cellular consequences of ABHD17A expression. Disrupting N-Ras palmitoylation by mutating the palmitoylated residue (C181S) or treating cells with the inhibitor 2-bromopalmitate (2-BP) relocalized N-Ras from the plasma membrane to internal organelles, as previously described (Choy et al., 1999;Goodwin et al., 2005) (Figure 4E,F). Overexpression of APT1 or APT2 had little effect on N-Ras localization ( Figure 4E,F), consistent with a recent report (Agudo-Ibáñez et al., 2015). In contrast, overexpression of ABHD17A, but not catalytically dead or cytosolic mutant forms, redistributed N-Ras from the plasma membrane to intracellular compartments consistent with its altered palmitoylation status ( Figure 4E,F). Taken together, these findings demonstrate the membrane-localized pool of ABHD17A depalmitoylates N-Ras and alters its subcellular targeting.
To determine if the endogenous ABHD17 proteins regulate palmitate cycling in vivo, we investigated the effect of ABHD17 knockdown on N-Ras depalmitoylation in HEK293T cells. RT-qPCR (Reverse transcription quantitative polymerase chain reaction) showed efficient silencing of ABHD17A alone, or ABHD17A, ABHD17B, and ABHD17C in concert, after 72 hr with siRNA treatment ( Figure 5A). ABHD17A knockdown had a slight effect on N-Ras depalmitoylation (p=0.084). In contrast, N-Ras palmitate turnover was significantly inhibited when all three ABHD17 proteins were simultaneously downregulated (p=0.0083), and this was not further enhanced by the APT1 and APT2 inhibitors C83 and C115 ( Figure 5B). Knockdown was less effective than PalmB treatment, which could be due to activity of the residual ABHD17 enzymes. PalmB may also inhibit additional factors that either directly or indirectly affect N-Ras palmitate cycling. Taken together, these results demonstrate that ABHD17 proteins redundantly mediate palmitate turnover on N-Ras.
The total number of cellular depalmitoylases is not known. We identified new PalmB targets, consistent with a recent report showing PalmB inhibits ABHD12 and monoacylglyerol lipase (Savinainen et al., 2014). As the mSH superfamily consists of >110 members, only half of which are functionally annotated , a comprehensively survey the mSH proteome may uncover yet more depalmitoylases. APTs are a critical element of the dynamic palmitoylation cycle, thus it will be imperative to identify the complete set of cellular APTs and determine how they contribute to the regulation of dynamic palmitoylation.
For cloning of mSHs for activity-profiling studies, plasmids containing corresponding human ORFs were purchased from DNASU (Arizona State University, Tempe, AZ) and OpenBiosystems, or obtained as clones from the hORFeome v8.1 Collection (Yang et al., 2011). Genes of interest were amplified by PCR using oligos with flanking restriction sites (described in Supplementary file 2), and the resulting mSH-encoding PCR products were subcloned into vectors of interest (FLAG-NT, generously provided by Dr. Stefan Taubert, University of British Columbia; or pCINeo, Promega [Madison, WI]).
The ABHD17A-FLAG construct was used as the template to generate ABHD17A mutant and mCherry-tagged plasmids. S211A-FLAG in pCINeo was generated by Quikchange mutagenesis, and ABHD17A DN-FLAG was amplified by PCR then subcloned into pCINeo. ABHD17A-mCherry wild type and mutant plasmids were generated by pairing each forward oligo with the reverse ABHD17A-mCherry-Linker oligo as listed in Supplementary file 2. The resulting ABHD17A fragments were fused with the PCR-amplified C-terminal mCherry cassette by overlapping extension PCR (OEPCR) and subcloned into pCINeo vector with EcoRI and XbaI. Similarly, mCherry-APT1 and mCherry-APT2 plasmids were constructed by fusing the N-terminal mCherry cassette with PCRamplified APT1 and APT2 fragments using OEPCR and subcloning the resulting fragments into pCI-Neo vector with EcoRI and XbaI.
The pSUPER vector and the shRNA pSUPER-APT1 plasmid used in knockdown studies was a generous gift from Dr. Gerhard Schratt (University of Marburg), and ON-TARGETplus SMARTpool siRNAs targeting APT2, ABHD17A, ABHD17B, or ABHD17C, as well as Non-Targeting control siRNA, were purchased from Dharmacon (Lafayette, CO).

cDNA and siRNA transfections
For pulse-chase metabolic studies and activity-based protein-profiling studies, COS-7 cells were transfected with cDNAs as indicated in each experiment using Lipofectamine 2000 as per manufacturer's instructions. Cells were grown in six-well plates (Corning; Corning, NY) and transfected at 90% confluence with 1 mg of cDNA per well for pulse-chase analyses with inhibitors, or 2 mg cDNA per well for pulse-chase analyses with thioesterase overexpression. For immunofluorescence studies, COS-7 cells were grown on glass coverslips (Fisher; Pittsburg, PA) in 24-well plates (Corning) and transfected at 60-90% confluence with 0.5 mg of cDNA per well using Xtreme-GENE 9 according to product instructions. Experiments involving small molecules were carried out 20-24 hr following transfection, and experiments involving co-expression of candidate mSHs were carried out 24-48 hr post-transfection, as described below.
For APT1 and APT2 studies, a double knockdown approach was used (Bond et al., 2011) where COS-7 cells were transfected with siRNA (100 nM final concentration per transfection) on days 1 and 3 with 5 mL of Lipofectamine 2000 per transfection. One microgram of cDNA was co-transfected with the siRNA on day 3, and pulse-chase studies were carried out on day 4, 20 hr following the cotransfection. For ABHD17 studies, HEK293T cells were transfected on day 1 with siRNA in 9 mL Lipofectamine RNAiMax, and on day 3 with 1mg of EGFP-N-Ras in 4 mL Lipofectamine 2000. Pulse-chase and RT-qPCR studies were performed on day 4, 20 hr following cDNA transfection.

Pulse-chase metabolic labeling with inhibitors
Twenty hours following transfection, COS-7 cells or HEK293T cells were washed twice in phosphatebuffered saline (PBS) and starved in cysteine-and methionine-free DMEM containing 5% charcoal-filtered FBS (Life Technologies) for 1 hr. Cells were then labeled with 30 mM 17-ODYA and 50 mM L-AHA for 1.5 hr in this media. The labeling media was removed, and cells were briefly washed twice in PBS before chasing in complete DMEM supplemented with 10% FBS and 300 mM palmitic acid. Small molecule inhibitors or DMSO (vehicle) were added at chase time 0. At indicated time points, cells were washed twice in PBS and lysed with 500 mL triethanolamine (TEA) lysis buffer (1% TX-100, 150 mM NaCl, 50 mM TEA pH 7.4, 2ÂEDTA-free Halt Protease Inhibitor [Life Technologies]). The lysates were transferred to 1.5 mL Eppendorf tubes (Corning), vigorously shaken (3 Â 20s) while placed on ice in between each agitation. Lysates were cleared by centrifugation at 16,000Â g for 15 min at 4˚C. Solubilized proteins in the supernatant were quantified using Bicinchoninic acid (BCA) assay (Life Technologies) and subsequently used for immunoprecipitations as described below.

Immunoprecipitations
For immunoprecipitations, Protein A or Protein G sepharose beads (GE Healthcare; Mississauga, ON) were washed thrice in TEA lysis buffer. Protein A beads were pre-incubated with rabbit anti-GFP antibodies (Life Technologies) and Protein G beads were pre-incubated with FLAG M2 antibodies (Sigma-Aldrich) for 2 hr at 4˚C, before the addition 500 mg -1 mg of transfected COS-7 cell lysates containing indicated proteins. Immunopreciptations were carried out for 12-16 hr on an end-to-end rotator at 4˚C. Following immunoprecipitation, sepharose beads were washed thrice in modified RIPA buffer (150 mM NaCl, 1% sodium deoxycholate (w/v), 1% TX-100, 0.1% SDS, 50 mM TEA pH7.4) before proceeding to sequential on-bead CuAAC/click chemistry.

Sequential on-bead CuAAC/click chemistry
Sequential on-bead click chemistry of immunoprecipitated 17-ODYA/L-AHA-labeled proteins was carried out as previously described (Zhang et al., 2010), with minor modifications. After immunoprecipitation, sepharose beads were washed thrice in RIPA buffer, and on-bead conjugation of AF488 to 17-ODYA was carried out for 1 hr at room temperature in 50 mL of freshly mixed click chemistry reaction mixture containing 1 mM TCEP, 1 mM CuSO 4 Á5H 2 O, 100 mM TBTA, and 100 mM AF488-az in PBS. After three washes in 500 mL RIPA buffer, conjugation of AF647 to L-AHA was carried out for 1 hr at room temperature in 50 mL click-chemistry reaction mixture containing 1 mM TCEP, 1 mM CuSO 4 Á5H 2 O, 100 mM TBTA, and 100 mM AF647-alk in RIPA buffer. Beads were washed thrice with RIPA buffer and resuspended in 10 mL SDS buffer (150 mM NaCl, 4% SDS, 50 mM TEA pH7.4), 4.35 mL 4Â SDS-sample buffer (8% SDS, 4% Bromophenol Blue, 200 mM Tris-HCl pH 6.8, 40% Glycerol), and 0.65 mL 2-mercaptoethanol. Samples were heated for 5 min at 95˚C, and separated on 10% tris-glycine SDS-PAGE gels for subsequent in-gel fluorescence analyses.

Competitive activity-based protein profiling
Twenty-four hours following transfection with mSH constructs, COS-7 cells were washed twice in PBS, transferred to a new vial by scraping in PBS, and lysed by gentle sonication on ice. Protein was quantified by BCA assay. Thirty micrograms of total protein was incubated either with DMSO or small molecule inhibitors at indicated concentrations at room temperature for 30 min, prior to the addition of FP-Rho (2 mM final concentration). Labeling reactions were carried out at room temperature for 1 hr and quenched with 4Â SDS-sample buffer heated to 95˚C for 5 min. Samples were separated on SDS-PAGE, analyzed by in-gel fluorescence, then transferred onto nitrocellulose membrane for Western blotting.

In-gel fluorescence analyses
A Typhoon Trio scanner (GE Healthcare) was used to measure in-gel fluorescence of SDS-PAGE gels: AF488 signals were acquired using the blue laser (excitation 488 nm) with a 520BP40 emission filter, AF647 signals were acquired using the red laser (excitation 633 nm) with a 670BP30 emission filter, and rhodamine signals were acquired with the green laser (excitation 532 nm), with a 580BP30 emission filter. Signals were acquired in the linear range and quantified using the ImageQuant TL7.0 software (GE Healthcare). For pulse-chase analyses, the ratio of palmitoylated substrates were calculated as AF488/AF647 values at each time point, normalized to the value at T=0.

Confocal microscopy and EGFP-N-Ras localization
COS-7 cells were co-transfected with EGFP-N-Ras and empty vector or indicated mCherry-tagged thioesterases at a 1:1 ratio (total 0.5 mg DNA per well) in Lab-Tek 8-well chamber slides (Fisher). Twenty-four hours post-transfection, cells were imaged on a TCS SP8 confocal laser scanning microscope (Leica Microsystems; Mannheim, Germany), and EGFP-N-Ras localization was quantified by counting 100 cells per experiment.