S-acylation (palmitoylation) is a poorly understood post-translational modification of proteins involving the addition of acyl lipids to cysteine residues. S-acylation promotes the association of proteins with membranes and influences protein stability, microdomain partitioning, membrane targeting and activation state. No consensus motif for S-acylation exists and it therefore requires empirical identification.
Here, we describe a biotin switch isobaric tagging for relative and absolute quantification (iTRAQ)-based method to identify S-acylated proteins from Arabidopsis. We use these data to predict and confirm S-acylation of proteins not in our dataset.
We identified c. 600 putative S-acylated proteins affecting diverse cellular processes. These included proteins involved in pathogen perception and response, mitogen-activated protein kinases (MAPKs), leucine-rich repeat receptor-like kinases (LRR-RLKs) and RLK superfamily members, integral membrane transporters, ATPases, soluble N-ethylmaleimide-sensitive factor-activating protein receptors (SNAREs) and heterotrimeric G-proteins. The prediction of S-acylation of related proteins was demonstrated by the identification and confirmation of S-acylation sites within the SNARE and LRR-RLK families. We showed that S-acylation of the LRR-RLK FLS2 is required for a full response to elicitation by the flagellin derived peptide flg22, but is not required for localization to the plasma membrane.
Arabidopsis contains many more S-acylated proteins than previously thought. These data can be used to identify S-acylation sites in related proteins. We also demonstrated that S-acylation is required for full LRR-RLK function.
S-acylation (palmitoylation) is the reversible post-translational addition of fatty acids, such as palmitate and stearate, through thioester linkages to cysteine residues of proteins. S-acylation increases membrane affinity and promotes the segregation of proteins into specific lipid environments, known as membrane microdomains (Levental et al., 2010). S-acylation is universally recognized as being important for cellular protein sorting, vesicle trafficking, activation state control, protein stability, microdomain partitioning of proteins and protein complex assembly, and there are many reviews in the literature (Greaves & Chamberlain, 2007; Baekkeskov & Kanaani, 2009; Charollais & Van Der Goot, 2009; Hemsley, 2009). The reversibility of S-acylation allows for a protein's membrane affinity or microdomain association to be altered, aiding protein complex formation or dissociation (Roy et al., 2005; Sorek et al., 2007). S-acylation is involved in various aspects of plant function, including small GTPase signalling (Lavy et al., 2002; Sorek et al., 2007, 2010), calcium signalling (Martin & Busconi, 2000; Batistic et al., 2008), disease resistance (Kim et al., 2005), heterotrimeric G-protein signalling (Adjobo-Hermans et al., 2006; Zeng et al., 2007; Hemsley et al., 2008) and the microtubule cytoskeleton (Hemsley et al., 2008).
The enzymes responsible for eukaryotic S-acylation, protein acyl transferases (PATs), have been identified only recently (Roth et al., 2002; Huang et al., 2004; Hemsley et al., 2005), and are present in the genome in greater numbers than are myristoyl transferases or prenyl transferases. For example, Arabidopsis thaliana contains 24 putative PAT genes (Hemsley et al., 2005; Batistic et al., 2012), but only two myristoyl transferase, one farnesyl and two geranylgeranyl transferase isoforms. S-acylation can occur throughout a protein, whereas other lipid modifications are restricted to specific N- or C-terminal motifs. This wide variation in the position and sequence of S-acylation sites within a protein, coupled with the varied cellular locations and timings of S-acylation, might explain why so many distinct PATs are present in eukaryotic genomes. The only PAT characterized in plants is TIP GROWTH DEFECTIVE 1 (TIP1) and mutants show pleiotropic defects (Schiefelbein et al., 1993; Ryan et al., 1998; Hemsley et al., 2005).
The knowledge of S-acylation in plants is very poor, with the few known S-acylated proteins having been identified individually. A putative motif for S-acylation has been proposed from the analysis of four human proteins (Gubitosi-Klug et al., 2005), but only the S-acylated cysteine is conserved in the proposed sequence; all other residues are highly variable and the proposed motif has not been validated experimentally. As a result, no reliable consensus sequence for S-acylation has been identified, making the prediction of S-acylation extremely difficult. Proteomic studies using spectral counting and/or multidimensional protein identification technology (MudPIT) methods identified 34 novel S-acylated proteins from yeast (Roth et al., 2006), c. 200 novel proteins from rat brain neurons (Kang et al., 2008) and 331 proteins from human prostate cancer cell lines (Yang et al., 2010), although the authors acknowledge that there are likely to be many more that were not detected. Click chemistry metabolic labelling approaches have identified c. 125 (Martin & Cravatt, 2009) and 415 (Martin et al., 2011) S-acylated proteins from human cells, although it is unknown whether and how metabolic labelling affects S-acylation dynamics and whether this approach represents in-vivo S-acylation.
Here, we demonstrate the use of an isobaric tagging for relative and absolute quantification (iTRAQ) approach, based on the selective cleavage of S-acyl groups (Drisdel & Green, 2004; Hemsley et al., 2008), to identify S-acylated proteins from plant material. We have identified c. 600 putative S-acylated proteins from Arabidopsis root-derived callus culture and demonstrate how this dataset can be used to predict the S-acylation of other proteins in Arabidopsis.
Materials and Methods
Sample preparation and analysis methods for proteomic samples are described in detail in Supporting Information Methods S1.
Plant growth and culture
Arabidopsis thaliana (L.) Heyn Col-4 ecotype wild-type (WT) and tip1-2 root cell suspension cultures were produced as described previously (Dunkley et al., 2004). Plants for S-acylation assays were grown as described previously (Hemsley et al., 2008).
Protein S-acylation assays
Assays of protein S-acylation were performed exactly as described by Hemsley et al. (2008). NON-RACE SPECIFIC 1/HARPIN INDUCED 1 (NDR1/HIN1)-LIKE 3 (NHL3), Arabidopsis heterotrimeric G-protein gamma subunit 1 (AGG1) and Arabidopsis REMORIN 1.2 (AtREM1.2) were expressed from the 35S promoter as FLAG-yellow fluorescent protein (YFP) fusions in WT Col-0, and detected using anti-FLAG M2 antibody (Sigma). FLAGELLIN SENSITIVE 2 (FLS2) was amplified from Arabidopsis Col-0 full length cDNA and cloned into pENTR D-TOPO. A fragment containing the C830,831S mutation was produced using overlapping PCR and cloned into pENTR D-TOPO FLS2 as a StuI/XbaI fragment. FLS2 and FLS2 C830,831S entry vectors were recombined into pK7FWG2 (Karimi et al., 2007) to produce green fluorescent protein (GFP) fusions under the control of the 35S promoter. FLS2-GFP and FLS2 C830,831S-GFP were detected using anti-GFP antibody. All other proteins were assayed as native proteins from WT Arabidopsis Col-0 using specific antibodies.
FLS2-GFP and FLS2 C830,831S localization
35S:FLS2-GFP, 35S:FLS2 C830,831S-GFP and 35S:GFP were transiently expressed in 5-wk-old Nicotiana benthamiana leaves with 35S:AtPIP2a-mCherry (Nelson et al., 2007) using Agrobacterium infiltration. Twenty-four hours after infiltration, images of the lower leaf epidermis were obtained using a Leica (Milton Keynes, UK) SP5 confocal microscope and a × 40 objective. Images were processed using Leica LAS AF software and cropped to size in Adobe Photoshop. The GFP fluorescence signal was false coloured green, and the mCherry fluorescence signal magenta.
flg22 treatments and real-time PCR
Fourteen-day-old seedlings grown in long days at 22°C on 1 × Murashige and Skoog (MS) medium 0.8% agar were transferred to 5 ml of 1 × liquid MS medium in six-well plates (10 seedlings per genotype per treatment) in long days at 22°C. After 48 h, flg22 peptide (5 μM final concentration) in water was added and plants were incubated in long days at 22°C for 1 h. Negative controls were treated with water instead of peptide. Samples were harvested and RNA extracted using Qiagen RNeasy Plant Kits with on-column DNAase digestion. Samples were reverse transcribed using an Applied Biosystems (Life Technologies Ltd, Paisley, UK) High Capacity RT Kit and real-time PCR was performed using Roche FastStart Universal Sybr (Rox) master mix on an Applied Biosystems 7300 real-time PCR machine. PEX4 was used as the reference gene. Relative quantification of transcript was achieved using the ΔΔCT method, and error bars represent a 95% confidence interval calculated on three technical replicates using Student's t-test (Livak & Schmittgen, 2001).
Proteomic identification of S-acylated proteins
To prepare S-acylated proteins for iTRAQ-based mass spectrometric identification, we substituted S-acyl groups for biotin after hydroxylamine cleavage (detailed in Fig. S1 and Methods S1) based on the methods of Drisdel & Green (2004). Briefly, free cysteines were blocked with N-ethylmaleimide (NEM), and then S-acyl groups were removed by hydroxylamine treatment (Hyd+ sample) and exposed cysteines were tagged with biotin. Biotinylated proteins were purified on Neutravidin beads. As a control for nonspecific biotinylation and column binding, samples were split into two after NEM treatment and one-half was processed without hydroxylamine cleavage (Hyd− sample). tip1-2 is a mutant in one of the 24 PATs in Arabidopsis, and is therefore likely to lack some S-acylated proteins (Hemsley et al., 2005). We were interested in detecting TIP1 S-acylation targets by comparing samples from identically grown and processed WT and tip1-2 mutant cells. Previous MudPIT strategies used co-purifying contaminants to standardize between samples (Roth et al., 2006; Kang et al., 2008), but there is no way of knowing whether the co-purifying contaminants occur in equal amounts across the experimental samples being compared. By using a 4-plex iTRAQ-based approach, we could compare protein abundance between experimental and control samples concurrently in both WT and tip1-2 (WT Hyd+, WT Hyd−, tip1 Hyd+, tip1 Hyd−) without the need to standardize to contaminant proteins, thereby reducing comparative errors. To increase confidence in the identification of S-acylated proteins, the entire sample preparation and mass spectrometric experiment was carried out twice, giving up to four biologically independent observations of the S-acylation state of a given protein.
Nine hundred and twenty-four proteins were identified in our unfiltered raw data using a MASCOT Percolator (Brosch et al., 2009) and iSPY (unpublished data, program freely available on request from K. S. Lilley), and are shown in Table S1. Proteins were sorted on the basis of enrichment in the WT Hyd+ samples relative to the WT Hyd− samples. Proteins proposed not to be S-acylated (Hyd+ : Hyd− ratio < 1) in either experiment were removed and are shown in Table S2. The 694 proteins from both experiments proposed to be S-acylated (Hyd+ : Hyd− ratio > 1) were sorted on the basis of the number of distinct peptides used for identification in each experiment as assigned by the MASCOT Percolator (Brosch et al., 2009) to indicate confidence in protein identification (Methods S1). This led to three groups: high confidence (Table S3a; three peptides or more in at least one experiment, 144 proteins), medium confidence (Table S3b; two peptides in at least one experiment, 115 proteins) and lower confidence (Table S3c; one peptide only in either or both experiments, 435 proteins). The confidence in S-acylation state is not related to this identification confidence, but is determined by the ratio of Hyd+ : Hyd−. A Hyd+ : Hyd− ratio > 1 indicates S-acylation for that protein in that sample, with the confidence of S-acylation increasing as the ratio increases. On the basis of these criteria, all of the 694 proteins may contain thioester linkages, but care should be taken when interpreting data from proteins with Hyd+ : Hyd− ratios close to unity without independent assessment of S-acylation, or from the lower confidence protein identification group without independent confirmation of identity. Proteins known to contain thioester linkages that are not involved in membrane anchorage were also purified by this method and are highlighted in Table S3(a–c).
Although Table S3 contains a full list of putative S-acylated proteins, it is possible that not all are truly S-acylated. The proposed S-acylated proteins were further filtered to remove common fortuitous proteomic contaminants, such as ribosomal proteins (Dunkley et al., 2004). These proteins are shown in Table S4. Proteins known to contain thioester linkages not involved in S-acylation (e.g. nitrilase, lipid biosynthesis enzymes, etc.; highlighted in Table S3a–c) were removed and are shown in Table S5 to avoid confusion with putative S-acylated proteins. The remaining 581 proteins, constituting a curated set of putative S-acylated proteins ranked as described for Table S3(a–c), are listed in Table S6(a–c).
The identified proteins include those shown to be S-acylated, such as calcineurin B-like proteins (CBLs) (Batistic et al., 2008, 2012), those hypothesized to be S-acylated, such as RPM1-INTERACTING PROTEIN 4 (RIN4) (Kim et al., 2005) and the heterotrimeric G-protein alpha subunit 1 (GPA1) (Adjobo-Hermans et al., 2006), and those that have homology to known S-acylated proteins from other organisms, such as calcium-dependent protein kinases (CDPKs) (Martin & Busconi, 2000) and BLOCKED EARLY IN TRANSPORT 3 (Turnbull et al., 2005). We also identified four DHHC S-acyl transferases, including TIP1, that form S-acylated intermediates during the catalysis of S-acylation (Hemsley et al., 2005), and may S-acylate each other in trans (Hemsley & Grierson, 2011). These data support the proposition that our method enriches and identifies S-acylated proteins. In Table S6(a–c), proteins previously reported or hypothesized to be S-acylated are highlighted in bold. As a point of reference, the lowest Hyd+ : Hyd− scores of known or hypothesized S-acylated proteins identified by our method were those of GPA1 in experiment 1 and RIN4 in experiment 2.
Verification of proteomic results
To verify our proteomic data, we performed small-scale S-acylation assays (Hemsley et al., 2008) on individual proteins detected in our analysis (Fig. 1). All proteins tested were confirmed to be S-acylated and are highlighted in bold italics in Table S6. Six nonrace-specific disease resistance 1/harpin-induced 1 (NDR1/HIN1)-like (NHL) pathogen response proteins were identified in the proteomic analysis. FLAG-YFP-NHL3 was detected as a doublet with the upper band at c. 10 kDa heavier than predicted because of glycosylation (Varet et al., 2003). The upper band showed far greater S-acylation than the nonglycosylated lower band (Figs 1, S2).
Prediction of S-acylation of related proteins
Using multiple alignments to find conserved cysteine residues in related proteins from our dataset, it is possible to identify potential S-acylation sites. Many proteins are members of multigene families, often with developmental or tissue-specific expression; therefore, it may also be possible to predict S-acylation of proteins not found in our proteomics data by alignment with family members from our dataset. We used the leucine-rich repeat receptor-like kinase (LRR-RLK) and soluble N-ethylmaleimide-sensitive factor-activating protein receptor (SNARE) families detected in our analysis to illustrate how this approach can be employed to identify S-acylation sites and S-acylated members of families.
Prediction of S-acylation sites in LRR-RLK family proteins
LRR-RLKs represent the largest family of receptor kinases in plants, with 216 members in 13 families in Arabidopsis (Shiu & Bleecker, 2001, 2003). Arabidopsis also contains 57 proteins showing homology to the LRRs (Receptor-like proteins, AtRLPs) and 129 showing homology to the kinase domains (Receptor-like cytoplasmic kinases, RLCKs) of LRR-RLKs (Shiu & Bleecker, 2001). In this study, we identified 13 LRR-RLKs, two proteins from the CLAVATA2 (CLV2)-like AtRLP subfamily and eight RLCK subfamily members (Table S7) as being S-acylated. These included well-characterized proteins such as BRASSINOSTEROID SIGNALING KINASE 3 (BSK3), INFLORESCENCE RECEPTOR-LIKE KINASE 2 (IMK2) and RECEPTOR KINASE-LIKE 1, as well as LRR-RLK family members of unknown function.
All of the LRR-RLKs, RLCKs and AtRLPs identified in our dataset contain one to two conserved cysteine residues adjacent to the cytoplasmic face of the predicted transmembrane region or myristoylation site (Fig. S3a), making these the most likely sites for S-acylation.
FLS2 is S-acylated and S-acylation is essential for signalling in response to flg22 elicitation
To determine whether S-acylation is a feature of LRR-RLKs, we examined the S-acylation state of an LRR-RLK that was not detected by our proteomic approach, but contained the conserved cysteines. The flagellin receptor FLS2 is a well-characterized LRR-RLK (Chinchilla et al., 2006) and contains the conserved cysteines predicted to be S-acylated above. We found that endogenous (not shown) and overexpressed FLS2 was S-acylated, and mutation of the conserved cysteines (C830,831) in FLS2 to serine abolished S-acylation (Fig. 2a), confirming our multiple alignment-based predictions on the location of the LRR-RLK family S-acylation sites. These data, combined with our observation of S-acylated LRR-RLK members being distributed throughout the LRR-RLK family (Table S7), identify S-acylation as a novel post-translational modification of the LRR-RLK family.
Overexpression of FLS2 leads to enhanced responses to flagellin (Gomez-Gomez & Boller, 2000), whereas loss of FLS2 in the fls2c mutant leads to reduced response to flagellin (Zipfel et al., 2004). Treatment of Arabidopsis plants with the flagellin-derived flg22 peptide leads to the induction of many genes, including WRKY53, WRKY29 and MITOGEN ACTIVATED PROTEIN KINASE 3 (MAPK3) (Navarro et al., 2004). Overexpression of FLS2 C830,831S in the fls2c mutant partially restores flagellin responsiveness, but, crucially, not to the same degree as fls2c plants overexpressing WT FLS2, as measured by the induction of WRKY53, WRKY29 and MAPK3 (Figs 2b, S3b,c). Both FLS2 and FLS2 C830,831S in these experiments are expressed at c. 100 (lines fls2c 35S::FLS2 #321 and fls2c 35s::FLS2 C830,831S #521) or c. 300 (lines fls2c 35S::FLS2 #241 and fls2c 35s::FLS2 C830,831S #413) times that of endogenous FLS2, as determined by real-time PCR (data not shown).
S-acylation frequently acts as a mechanism to relocate proteins from one membrane compartment to another (Hancock et al., 1990; Zeng et al., 2007), and we wanted to examine whether the reduction in flg22 responsiveness in FLS2 C830,831S-expressing plants relative to FLS2-expressing plants was caused by a failure of FLS2 C830,831S to be correctly delivered to the plasma membrane. GFP fusions of FLS2 and FLS2 C830,831S were infiltrated into tobacco leaves, together with the reported plasma membrane marker AtPIP2a-mCherry (Nelson et al., 2007). As reported previously, FLS2 was observed at the periphery of the cell and showed localization to the cell periphery with AtPIP2a. FLS2 C830,831S showed an identical distribution to FLS2 (Fig. 2c), suggesting that FLS2 C830,831S is correctly trafficked to the plasma membrane in the absence of S-acylation. Free GFP was also included as a control, and showed a pattern distinct from that of FLS2, FLS2 C830,831S and PIP2a, including cytoplasmic strands and intranuclear labelling (Fig. 2c).
These data combined indicate that the S-acylation of FLS2 is not required for trafficking to the plasma membrane, but is a requirement for the mediation of a full and efficient response to flagellin elicitation at the plasma membrane.
Prediction of S-acylation sites in SNARE proteins
SNAREs are important factors during membrane fusion events, and previous work in mammals and yeast has identified some SNAREs as being S-acylated (Veit et al., 2003; Roth et al., 2006; He & Linder, 2008). We identified 13 putatively S-acylated SNAREs (Tables S3, S6, S8) and performed individual S-acylation assays on three of them (SYNTAXIN OF PLANTS 71 (SYP71), SYP122 and NOVEL PLANT SNARE 11 (NPSN11)), confirming that these SNAREs are indeed S-acylated (Fig. 3). Multiple alignments of the Arabidopsis SNAREs identified in our screen showed that the only conserved cysteine residue is adjacent to the cytoplasmic face of the transmembrane domain (Fig. S4). S-acylation assays of SNARE proteins that do not contain this cysteine residue (Table S8, Fig. S4) indicate that they are not S-acylated (e.g. SECRETION 22 (SEC22), SYP31, MEMBRIN 11 (MEMB11); Fig. 3) and that this cysteine is the probable S-acylation site. Multiple alignments indicate that a cysteine residue occurs at this position in 41 of the 62 SNARES encoded by Arabidopsis, including PEN1 and KNOLLE (Table S8).
Myristoylated and S-acylated proteins
S-acylation can also occur with N-myristoylation (Martin & Busconi, 2000; Batistic et al., 2008) and computational prediction of N-myristoylation sites (Podell & Gribskov, 2004) within our dataset identified 30 potential N-myristoylated proteins (Table S9). However, given the lack of knowledge surrounding AtNMT1 (N-myristoyl transferase 1) and AtNMT2 substrate specificity and N-myristoylated proteins from plants in general, there could be more proteins that were not detected by the search algorithm used. The majority of these proteins are CDPKs or kinases of unknown function. Interestingly, many of these proteins contain cysteines adjacent to the predicted N-myristoylation site and are likely candidates for S-acylation.
Prenylated and S-acylated proteins
Prenylated proteins can also be modified by S-acylation (Sorek et al., 2007; Zeng et al., 2007). Using PrePS (Maurer-Stroh & Eisenhaber, 2005), we analysed our dataset and identified seven potentially prenylated proteins. The heterotrimeric G-protein gamma subunit AGG1 has been reported previously to be prenylated (Mason & Botella, 2001), but not S-acylated. ATJ2 and ATJ3 are Arabidopsis members of the plant membrane-associated DnaJ homologue family (Preisig-Muller & Kindl, 1993). TOBAMOVIRUS MULTIPLICATION 2A (TOM2a) is involved in Tobamovirus pathology (Fujisaki et al., 2008). AtRABG1 was also identified by our proteomic approach as S-acylated, suggesting that, although predicted to be prenylated, it may also be S-acylated.
Potential TIP1 targets
Using our methodology, it is possible to compare the relative S-acylation levels of a protein between WT and tip1-2 protein samples. Potential TIP1 substrates should show tip1-2 Hyd+ : tip 1-2 Hyd− ratios of around unity or be absent entirely, as these proteins should not be S-acylated in the tip1-2 sample and therefore should not be enriched by our method. One hundred and three proteins were identified that showed ≥ 1.5-fold under-representation in tip1-2 samples from either experiment (Table S10). Proteins detected in both experiments, but only under-represented in one of the two tip1-2 samples analysed, were not included. It should be noted that the loss of TIP1 may also cause changes in protein S-acylation independent of direct TIP1 action, and that proteins identified in this manner may not necessarily be TIP1 targets. Three proteins were under-represented in tip1-2 from both experiments, and the only characterized protein found in both experiments was A. thaliana PLANT CADMIUM RESISTANCE 2 (AtPCR2). We were unable to verify S-acylation by TIP1 as we were unable to recover tip1-2 plants expressing detectable levels of epitope-tagged AtPCR2. It is possible that S-acylation is essential for the function of AtPCR2 and that overexpression in the absence of the cognate S-acyl transferase (TIP1) could be lethal. The list of proteins under-represented in tip1-2 samples includes TIP1, indicating that TIP1 protein levels are extremely low, confirming that this is most probably a null mutant (Hemsley et al., 2005).
In this work, we have identified 582 putatively S-acylated proteins from root cell culture, although, as with any indirect proteomic study, many of these await independent confirmation, with metabolic labelling and mass spectrometry likely to be the best methods (Martin & Busconi, 2000; Hemsley et al., 2005; Sorek et al., 2007). Before this study, very few proteins were known to be S-acylated in plants. This work indicates that S-acylation in Arabidopsis is relatively common, and shows that large numbers of proteins affecting diverse cellular functions are modified by this poorly understood modification. By comparison, the numbers of N-myristoylated and prenylated proteins in the entire genome is predicted to be c. 437 (Boisson et al., 2003) and c. 542 (Maurer-Stroh et al., 2007) proteins, respectively. Almost all of the proteins identified in this work were not only unknown to be S-acylated, but had not been previously thought to be S-acylated, and no role for S-acylation in their function had been proposed. These include multiple members of the RLK superfamily, ABC transporter superfamily, AAA, ACA, VHA and AHA ATPases, protein kinases, protein phosphatase 2c (PP2c), plasma membrane intrinsic protein (PIP) water channels and sphingosine kinases (Tables S3a–c, S6a–c), in addition to those proteins already described in the Results, as well as proteins of unknown function or protein families represented in our data by only one member. Previous research has suggested that S-acylation can determine subcellular localization, membrane partitioning, recycling, complex composition, protein conformation, complex stability or sensitivity of a receptor or receptor complex to a ligand. Our results open up many exciting avenues for research and provide explanations for hitherto unexplained observations.
Although a list of TIP1 targets might have been revealed using our method, we found relatively few proteins that were strongly S-acylated in WT and not S-acylated in tip1-2 cultures. This may be because TIP1 has few targets, many TIP1 targets are not expressed in cell culture or other PATs are able to S-acylate TIP1 substrates in the absence of TIP1. Recent reports have indicated that PATs have very low substrate specificity and are relatively promiscuous enzymes (Hou et al., 2009; Rocks et al., 2010), and this may have reduced the effects caused by the loss of TIP1 and masked the changes in abundance we had hoped to observe. Recent work has shown that Arabidopsis PATs localize to various different membrane compartments (Batistic, 2012). These data, in combination with our work presented here, may be of use in trying to identify enzyme–substrate pairing based on subcellular localizations.
Prediction of the S-acylation of related proteins
Many proteins are members of multigene families, often with temporal, developmental or tissue-specific expression. By using multiple alignments to identify conserved cysteine residues in protein families from our dataset, it is possible to identify S-acylated proteins not expressed in our samples. We used the LRR-RLK/RLP/LRR and SNARE families to illustrate and validate this approach.
S-acylation of SNARE proteins
SNAREs help to regulate membrane fusion events, and previous work has identified some non-plant SNAREs as being S-acylated (Veit et al., 2003; Roth et al., 2006; He & Linder, 2008). We identified 13 SNAREs as being S-acylated, and demonstrated that a cysteine residue adjacent to the cytoplasmic face of the transmembrane domain (Fig. S4) is the probable S-acylation site. S-acylation promotes the stability of the yeast SNARE TLG1 by preventing ubiquitination and degradation (Valdez-Taubas & Pelham, 2005), and is required for the relocation of mammalian syntaxin 7 from the plasma membrane to endosomes (He & Linder, 2008).
S-acylation of LRR-RLKs, RLCK and AtRLPs
The LRR-RLK family is one of the largest in Arabidopsis with 216 LRR-RLKs, 57 AtRLPs and 129 RLCKs (Shiu & Bleecker, 2001, 2003). In this study, we identified 14 LRR-RLKs, two AtRLPs and seven RLCKs as being S-acylated (Table S7). Multiple alignments of LRR-RLK family members indicate that one to two cysteines adjacent to the cytoplasmic end of the transmembrane span or putative N-myristoylation site in LRR-RLKs are the probable S-acylation site (Fig. S3a), and point mutagenesis and biochemical assay of FLS2 confirms this (Fig. 2a). A range of non-LRR-RLKs were also identified as being S-acylated, including FERRONIA (FER), RECEPTOR-LIKE KINASE IN FLOWERS 3 (RKF3), PROLINE EXTENSIN-LIKE RECEPTOR KINASE 1 (AtPERK1) and the RESISTANCE TO P. SYRINGAE 2 (RPS2)-interacting RLK AT4G08850. It seems likely that S-acylation plays an important role in LRR-RLK/AtRLP/RLCK/RLK family interactions and signalling.
The plasma membrane is one of the first sites at which invading pathogens are perceived, and some bacterial type III effectors are also S-acylated (Nimchuk et al., 2000; Thieme et al., 2007). Many of the S-acylated proteins identified in our analysis and discussed below are found in complexes together at the plasma membrane. RPS2 is a protein involved in the monitoring of bacterial type III effectors. Four of the 10 proteins in an RPS2-containing complex (Qi & Katagiri, 2009) were identified by us as S-acylated (RIN4, LRR-RLK AT4G08850, HYPERSENSITIVE INDUCED REATION 1(AtHIR1) and AtHIR2). FLS2 is also S-acylated (Fig. 2a) and forms part of the RPS2 complex (Qi et al., 2011) with ACA8 and ACA10 calcium channels (Frei dit Frey et al., 2012), identified here as S-acylated. These data, coupled with the identification of other pathogen response proteins as being S-acylated (e.g. remorins, SNAREs/syntaxins such as PENETRATION DEFICIENT 1 (PEN1), LRR-RLKs, AtRLPs, RLCKs, heterotrimeric G-proteins, MAPKKKs, NECROTIC SPOTTED LESIONS 1 (NSL1)-like proteins, BONZAI 1 (BON1) and BON2, AtHIR proteins, NHL proteins, ENHANCED DISEASE SUSCEPTIBILITY 5 (EDS5) multidrug efflux pump and PEN3), indicate a prominent role for S-acylation in pathogen perception and resistance.
The role of FLS2 S-acylation in flagellin perception
We have demonstrated that the well-studied LRR-RLK FLS2 (Chinchilla et al., 2006) is S-acylated at a site predicted using multiple alignments of LRR-RLK family proteins from our dataset (Fig. 2a). Non-S-acylated FLS2 still traffics exclusively to the plasma membrane (Fig. 2c), indicating that S-acylation does not act as a plasma membrane targeting signal. S-acylation is, however, essential for a full flagellin response (Figs 2b, S3b,c). Non-S-acylated FLS2 is still able to activate defence responses after elicitation by flg22, indicating that FLS2 does not require S-acylation for pathogen-associated molecular pattern (PAMP) perception, but that S-acylation potentially makes signal transduction more efficient, possibly by promoting or stabilizing interactions with other components of the active FLS2 complex. As the constructs used in this study generated FLS2 expression 100–300 times that found in WT plants, but transcriptional responses to flg22 elicitation were, at most, twice that observed in WT (Figs 2b, S3b,c), the system may have been saturated and the true difference between the signalling capacity of S-acylated and non-S-acylated FLS2 may actually be greater than that reported here using overexpression constructs. Interestingly, FLS2, FER, BSK3 and avrPphB SUSCEPTIBLE-LIKE 1 (PBL1), as well as other S-acylated RLK superfamily members, are enriched in detergent-resistant membrane (DRM) microdomains in response to fls22 elicitation (Kierszniowska et al., 2009; Keinath et al., 2010). S-acylation promotes recruitment to DRM fractions (Sorek et al., 2007) and S-acylation of FLS2 (as well as FER, BSK3 and PBL1) might be one mechanism by which FLS2 signalling complex formation and/or endocytosis is regulated in response to flagellin perception.
The NHL family of integral membrane proteins contains between 28 (Varet et al., 2002) and 45 (Zheng et al., 2004) members in Arabidopsis that are involved in senescence and the hypersensitive response to pathogens (Dormann et al., 2000; Varet et al., 2002, 2003; Zheng et al., 2004). We identified six NHL proteins as being S-acylated and demonstrated that glycosylated NHL3 is preferentially S-acylated (Fig. S2). This probably indicates that S-acylation occurs after glycosylation, and might act as a mechanism to trap mature glycosylated NHL3 on the plasma membrane or to promote its export from the endoplasmic reticulum.
Remorins (REMs) are a multigene family of plant-specific membrane-associated proteins (Reymond et al., 1996) that prevent Potato virus X movement (Raffaele et al., 2009) and are required for Sinorhizobium perception during nodulation in Medicago (Lefebvre et al., 2010). FLAG-YFP-AtREM1.2 fusions were assayed and confirmed to be S-acylated (Fig. 1). S-acylation of REMs could be important for their reported localization in specific lipid microdomains (Raffaele et al., 2009) and interaction with three LRR-RLKs (Lefebvre et al., 2010) two of which, NOD FACTOR PERCEPTION (NFP) and DOESN'T MAKE INFECTIONS 2 (DMI2), contain the proposed S-acylation site detailed earlier.
Heterotrimeric G-protein subunits
Heterotrimeric G-proteins are ubiquitous eukaryotic signalling components, and Arabidopsis contains one Gα (GPA1), one Gβ (AGB1) and two Gγ (AGG1, AGG2) subunits and additional atypical Gα (XLG2) and Gγ (AGG3) subunits. GPA1 was suggested to be S-acylated (Adjobo-Hermans et al., 2006), was identified in our dataset and independent assays have shown that it is S-acylated (Fig. 1). AGG1 was also detected and, although AGG2 has been demonstrated previously to be S-acylated (Hemsley et al., 2008), AGG1 has been proposed previously not to be S-acylated (Adjobo-Hermans et al., 2006; Zeng et al., 2007). Independent assay of AGG1 indicates that it is S-acylated (Fig. 1). The differing effects of mutagenesis of putative S-acylation sites in AGG1 and AGG2, and the idea that S-acylation plays a different role in the function of each protein (Adjobo-Hermans et al., 2006; Zeng et al., 2007), correlate with existing data on the different functions of AGG1 and AGG2 (Trusov et al., 2007).
One of the major and novel findings of this study is that a large number of RLK superfamily members are S-acylated, identifying an entirely new facet of RLK functional regulation. We demonstrate that FLS2, a member of the RLK LRR-RLK subfamily involved in pathogen perception, requires S-acylation to produce a full and correct response to flagellin elicitation. RLK superfamily members affect almost all aspects of plant development (Yang & Sack, 1995; Huck et al., 2003; Chevalier et al., 2005; Abrash et al., 2011), are vital for responses to microbes during nodulation (Arrighi et al., 2006) and suppression of bacterial pathogenesis (Gomez-Gomez & Boller, 2000; Zipfel et al., 2006; Danna et al., 2011), and combine in a multitude of ways to form active homo- or hetero-RLK multimer signalling complexes at the plasma membrane (Chinchilla et al., 2007). The formation of these complexes involves the segregation between specific lipid microdomains of various components, including RLK family members (Keinath et al., 2010). S-acylation affects microdomain segregation and is also reversible, allowing for control over complex formation in microdomains in response to various signals (Sorek et al., 2007). S-acylation may therefore provide control over which complexes are formed in response to a given signal or in a particular circumstance. S-acylation is likely to play a major role in the determination of correct and efficient complex formation and in the sorting amongst these important regulators of plant responses to external and internal cues.
Our dataset raises many interesting possibilities for S-acylation in plants, as it affects a very broad range of proteins with varied functions. Both integral and peripheral membrane proteins were identified, including proteins predicted to be N-myristoylated or prenylated, as well as proteins annotated as being soluble. Cysteines are also subject to other modifications, such as nitrosylation and glutathionylation. Recent work on postsynaptic density protein 95 (PSD-95) in animals indicates that S-acylation and S-nitrosylation may occur at the same site and regulate each other's function and occurrence (Ho et al., 2011), but it is not known whether S-acylation and glutathionylation can interact in a similar manner. It will be interesting to define roles for S-acylation in the varied proteins identified in this study.
Antibodies were gifts from Alan Jones (anti-GPA1), Mike Blatt (anti-SYP122, anti-SYP71), Natasha Raikhel (anti-NPSN11), Patrick Moreau (anti-Sec22, anti-Memb11, anti-Syp31), Georg Felix (anti-FLS2) and Fumiaki Katagiri (anti-HIR1/2/4). The fls2c mutant and full-length FLS2 cDNA were gifts from Cyril Zipfel and flg22 peptide was a gift from Ari Sadanandom. We are grateful to Angus Murphy, Heather Knight and Marc Knight for critical reading of the manuscript. The work was supported by the Biotechnology and Biological Sciences Research Council (BBSRC) grant BB/D013585/1 and Lady Emily Smyth Agricultural Research Station financial support to C.S.G., and BBSRC grant BB/C507561/1 to P.D. and K.S.L.