Lysine acetylation is a widespread protein modification for diverse proteins in Arabidopsis.

Lysine acetylation (LysAc), a form of reversible protein posttranslational modification previously known only for histone regulation in plants, is shown to be widespread in Arabidopsis (Arabidopsis thaliana). Sixty-four Lys modification sites were identified on 57 proteins, which operate in a wide variety of pathways/processes and are located in various cellular compartments. A number of photosynthesis-related proteins are among this group of LysAc proteins, including photosystem II (PSII) subunits, light-harvesting chlorophyll a/b-binding proteins (LHCb), Rubisco large and small subunits, and chloroplastic ATP synthase (β-subunit). Using two-dimensional native green/sodium dodecyl sulfate gels, the loosely PSII-bound LHCb was separated from the LHCb that is tightly bound to PSII and shown to have substantially higher level of LysAc, implying that LysAc may play a role in distributing the LHCb complexes. Several potential LysAc sites were identified on eukaryotic elongation factor-1A (eEF-1A) by liquid chromatography/mass spectrometry and using sequence- and modification-specific antibodies the acetylation of Lys-227 and Lys-306 was established. Lys-306 is contained within a predicted calmodulin-binding sequence and acetylation of Lys-306 strongly inhibited the interactions of eEF-1A synthetic peptides with calmodulin recombinant proteins in vitro. These results suggest that LysAc of eEF-1A may directly affect regulatory properties and localization of the protein within the cell. Overall, these findings reveal the possibility that reversible LysAc may be an important and previously unknown regulatory mechanism of a large number of nonhistone proteins affecting a wide range of pathways and processes in Arabidopsis and likely in all plants.

Protein Lys acetylation (LysAc) is a posttranslational modification whereby an acetyl moiety is transferred to the «-amino group of Lys residues and is distinct from the acetylation that occurs frequently on the N terminus of a protein. Importantly, acetylation of the «-N of Lys removes the positive charge of the side chain and therefore directly impacts the electrostatic status of the modified protein (Glozak et al., 2005). LysAc is best known for its regulation of histone proteins, affecting chromatin structure and gene expression (Eberharter and Becker, 2002). Recent proteo-mic analyses further extended the scope of LysAc. In Salmonella, LysAc is found to be widespread in the central metabolic enzymes and the occurrence of LysAc correlates with alterations of carbon fluxes (Wang et al., 2010). In humans, LysAc is particularly abundant in protein complexes where it regulates protein-protein interactions and enzyme activities (Choudhary et al., 2009;Zhao et al., 2010).
Although LysAc is considered to be an evolutionarily conserved modification from prokaryotes to eukaryotes (Choudhary et al., 2009;Zhang et al., 2009), little is known about LysAc beyond histone proteins in plants. In Arabidopsis (Arabidopsis thaliana), histone LysAc is essential for plant growth and development, as mutants of Arabidopsis lacking certain members of the histone acetyltransferases or deacetylases have altered levels of histone LysAc and display pleiotropic growth defects (Tian and Chen, 2001;Vlachonasios et al., 2003).
The first objective of this study was to determine if nonhistone LysAc also occurs in Arabidopsis as in other reported organisms. We used immunoblotting with generic anti-LysAc antibodies to answer this question. We further identified the sites of LysAc modification, using peptide affinity enrichment (Choudhary et al., 2009;Zhang et al., 2009) prior to liquid chromatography (LC)/mass spectrometry (MS) analysis. The second objective of this study was to determine the functional significance of LysAc on two selected, abundant LysAc proteins (LHCb and eukaryotic elongation factor-1A ).

Abundant LysAc of Nonhistone Proteins in Arabidopsis
The presence of LysAc in different organs and tissues of Arabidopsis was demonstrated by immunoblotting with anti-LysAc antibodies (Fig. 1A). The two most prominent acetylated protein bands from shoot and leaf tissues comigrated with the 50-and 25-kD standard proteins. Silique and flower tissues also contained these two dominant bands, but they were less abundant. For the root, the proteins cross-reacting with the anti-LysAc antibodies were numerous but in general less abundant. Seeds contained major Lys acetylated proteins at 25 and 18 kD that corresponded to major storage proteins previously identified as cruciferin 2 a-subunit and cruciferin 3 b-subunit (Wan et al., 2007).
To better define the composition of the major LysAc proteins at 50 and 25 kD in leaf extracts, we separated proteins by two-dimensional electrophoresis (2-DE) followed by immunoblotting with anti-LysAc antibodies (Fig. 1B). It was determined that the 50-kD LysAc band consisted of two acetylated proteins: eEF-1A, pI 9.2; and Rubisco large subunit (RBCL), pI 5.9. From the Coomassie staining intensities it was clear that there was less eEF-1A protein but greater crossreactivity with the anti-LysAc antibodies compared to RbcL, suggesting a higher stoichiometry of LysAc. The 50-kD acetylated protein spot, with an apparent pI 3.0, was identified by LC/MS as a mixture of RbcL, ATP synthase b-subunit, and other proteins that did not match the pI. We therefore reasoned that these proteins were incompletely focused during the initial isoelectric focusing step. In contrast, the 25-kD acetylated protein band consisted of three neighboring spots, identified as LHCb1 and LHCb2 by LC/MS. The surrounding small spots were other LHCb minor isoforms, consistent with the report of Galetskiy et al. (2008).
The specificity of the anti-LysAc immunoblots was demonstrated by strong competition with acetyl-BSA but not unmodified BSA (Supplemental Fig. S1), as also reported by Zhang et al. (2009) in studies of LysAc in Escherichia coli. The specificity was also supported by the observation that although the major bands in the anti-LysAc immunoblots were also abundant proteins by Coomassie staining, not all of the abundant proteins were strongly recognized by the anti-LysAc antibodies (Fig. 1A). However, one contrasting note is that the anti-LysAc antibodies appeared to recognize carbamylated-Lys residues as well. As shown in Supplemental Figure S2, extended heating of samples with SDS sample buffer containing urea resulted in strong cross-reaction with the generic antibodies. Thus, we avoided the use of urea for sample preparations in this study, except where mentioned.

Identification of Lys-Modified Proteins by LC-MS
We identified 64 different Lys modification sites from 57 proteins and each site was verified manually (Table I; Supplemental Table S2). Consistent with Figure 1, the major Lys-acetylated proteins from immunoblotting all had at least one positive site, such as Lys-306 of eEF-1A, Lys-252 of RbcL, and Lys-37 of LHCb1. However, an acetylated peptide has the mono-isotopic mass gain of 42.0105 units, which is nearly the same as trimethylation (42.0468 mass units). Thus, the two modifications differ by only 0.0363 D (36 ppm for a M r 1,000 peptide; Zhang et al., 2004), which is less than the resolution of our LC/MS analysis. Consequently, we were not able to distinguish these two modifications by MS. However, our 2-DE separation of soluble leaf proteins followed by immunoblotting with generic antibodies suggested that LysAc was much more abundant than Lys trimethylation, and with the exception of eEF-1A the proteins modified by LysAc were distinct from the proteins modified by Lys trimethylation (Supplemental Fig. S3). Hence, it is reasonable to conclude that the majority of the proteins we identified with a mass gain of 42 in LC/MS are sites of LysAc rather than Lys trimethylation.
The identified LysAc proteins consist of a diverse group representing virtually all major plant processes ( Fig. 2), including photosynthesis, protein metabolism, cell organization and biogenesis, stress responses, and secondary metabolism. Accordingly, these proteins were localized in various cellular compartments, including the chloroplast, nucleus, plasma membrane, and so on (Supplemental Fig. S4). Representative proteins and their gene ontology classifications are listed in Table I. A number of Lys-acetylated candidate proteins are involved in photosynthesis, including light-harvesting protein LHCb1, 33-KDa subunit of photoreaction center PSII, carboxylation enzyme Rubisco large and small subunits, and energy generator chloroplast ATP synthase (b-chain). This group of proteins and their LysAc are unique to plants.
The localization of Lys-acetylated proteins was further characterized in immunolocalization experiments (Fig. 3). As shown in Figure 3, A to D, robust anti-LysAc signals were observed in mesophyll cell chloroplasts (short arrowhead) with enriched labeling on the thylakoid membranes that were visualized with FM143 membrane stain (Fig. 3). Starch grains inside the chloroplast were not labeled, which further indicated the specificity of the labeling. The section analyzed in Figure 3, A to D also included the epidermis and it was clear that the plasma membrane of epidermal cells also contained LysAc proteins (long arrow), suggesting that LysAc is a common modification for different cell types. We also observed anti-LysAc labeling of the nuclei, visualized with 4#,6-diaminophenylindole (DAPI) nuclear stain (Fig. 3, E-H).
Highly condensed chromatin regions, or chromocenters, are visualized as the bright DAPI-positive domains, and often contain the pericentric chromosome regions. In contrast, the gene-rich portions of chromosomes are less condensed and stain more weakly with DAPI (Tessadori et al., 2007). Importantly, histones associated with chromocenters are typically methylated whereas chromatin in other regions of the nucleus tends to be associated with acetylated histones (Soppe et al., 2002;Jackson et al., 2004). Consistent with this notion, we found weak colocalization of DAPI staining with protein LysAc inside the chromocenters, suggesting that proteins in these regions may be hypoacetylated. Finally, the specificity of the immunolabeling was demonstrated in Figure 3, I to L, where no primary anti-LysAc antibodies were added and as a result, no signals were observed in the antibody channel, while the chlorophyll autofluorescence showed the presence of the cells. To summarize, the immunolocalization data reinforced the LC-MS identification that LysAc proteins in Arabidopsis are diverse and localized in many cellular compartments.

Differential LysAc of Peripheral LHCb versus Tightly PSII-Bound LHCb
LHCb proteins are encoded in the nucleus, and processed into the mature form in the chloroplast. N-acetylation of LHCb has been reported (Michel et al., 1991), but «-acetylation of these proteins is not known. In the immunoblotting results (Fig. 1C), LHCb family proteins were one of the major LysAc proteins in Arabidopsis. The mature LHCb1 polypeptide begins with Arg-36 (Michel et al., 1991), and our LC/ MS analysis identified Lys-37 to be a site of acetylation in the LHCb1 Glu-C proteolytic peptide: H2N R(ac)KTVAKPKGPSGSPWYGSD. The LC/MS spectrum also contained the N-acetylated form peptide: ACETYL RKTVAKPKGPSGSPWYGSD, which lacked the «-acetylation on Lys-37. This mixed identification of acetylated Lys-37 and N-acetyl-Arg-36 peptides also occurred for the phosphorylated forms of each peptide (phosphoThr-38; Supplemental Fig. S5). However, no peptides with double acetylation (N-Ac-R36 and acK37) or completely lacking acetylation (no N-Ac-R36 or acK37) were identified in the same digests. Thus, it is possible that the «-acetylation of Lys-37 and N-acetylation of Arg-36 are mutually exclusive.
It is known that LHCb occurs in several complexes in the thylakoid membranes and to study the possible relation of LysAc with LHCb protein association, we performed a 2-D native green gel analysis of thylakoid membranes (Fig. 4). In the first native (i.e. nondissociating) dimension (Fig. 4A), three chlorophyllcontaining bands were found to be acetylated: band 1, corresponding to photosystem supercomplexes; band 2, PSII:LHCb complex; and band 3, peripheral LHCb (14). Peripheral LHCb includes the LHCb that is loosely bound to PSII and the mobile pool of LHCb. Each of the three bands was cut out individually and the component proteins in each band were resolved by SDS-PAGE in the second dimension, which resolved two major LysAc bands at 50 and 25 kD. The highly LysAc 50-kD protein (both in band 1 and band 2) was identified as RbcL by LC-MS. It is established that RbcL can interact with the thylakoid membrane (Smith et al., 1997), but the surprising observation of RbcL comigrating together with PSII and LHC in different complexes has not been reported. The 25-kD LysAc proteins were identified as LHCb1 and LHCb2. Furthermore, it turned out that the peripheral LHCb (band 3) had substantially higher levels of LysAc than the tightly PSII-bound LHCb (band 2), despite the similar amount of LHCb protein (Fig. 4B). These quantitative differences are summarized in Figure 4C. For both light and dark samples, the peripheral LHCb antennae (band 3) had 2-to 3-fold higher LysAc/ LHCb ratio (i.e. acetylation normalized for protein) compared to the LHCb that was more tightly bound to PSII (band 2). The results suggest that LysAc may play a role in LHCb associations in the thylakoid membrane and thus could be of regulatory significance.
We monitored the modification of LHCb by LysAc and Thr phosphorylation during light-to-dark transitions (Fig. 4D). As expected, the phosphorylation of LHCb on Thr residues decreased rapidly in the dark, while the LysAc remained constant. This result is consistent with the observation from the green gel study. Hence, unlike phosphorylation, which is a big part of LHCb regulation during state transitions (16), the LysAc of LHCb did not respond directly to shortterm light/dark changes.
Interestingly, LysAc was detected not only in the major antennae (LHCb1 and LHCb2), but was also in the minor antennae (LHCb3-5). We identified alterna-tive N-terminal forms of LHCb5 starting with Leu-38, Phe-39, or Ser-40 (Supplemental Fig. S6). However, only the peptide starting with Leu-38 contained the acetylated modification at Lys-41. Thus, the possible relation between LysAc and alternative transit peptide processing in LHCb5 emerges as an interesting topic to explore in the future. In summary, we have shown that LysAc is an important modification of LHCb proteins and may play a role in LHCb protein interactions and N-terminal processing.
LysAc of eEF-1A Lys-306 in Relation to Calmodulin Binding eEF-1A was another abundant LysAc protein identified in Arabidopsis by immunoblotting (Fig. 1B)  sites have been reported to be trimethylated in eEF-1A in maize (Zea mays; Lopez-Valenzuela et al., 2003). In addition, two Lys residues (Lys-44 and Lys-55) were identified as monomethylated or dimethylated in this study (Table II). To confirm the LysAc of eEF-1A, we produced sequence-and modification-specific anibodies to recognize acetylated Lys-306 (antieEF-1A-acK306) and acetylated Lys-227 (antieEF-1A-acK227) and confirmed that both residues were acetylated in vivo in the shoot and root (Fig. 5, A and B). The sequence and modification specificity of these custom antibodies was documented with dot-blot peptide assays (Supplemental Fig. S7). As expected, the generic anti-LysAc antibodies could recognize the acetylated peptides corresponding to Lys-306 (eEF-1A-acK306) and Lys-227 (eEF-1A-acK227).
Carrot (Daucus carota) eEF-1A was reported to interact with calmodulin (CaM; Durso and Cyr, 1994), but the specific sequences involved in the binding are not known. We used a predictive algorithm (Hoeflich and Ikura, 2002) for CaM-binding sequences and found the sequence from Leu-291 to Ala-314 (containing Lys-306) had the highest score for the Arabidopsis eEF-1A protein. Using synthetic peptides, we confirmed that recombinant CaM could interact with the eEF-1A Lys-306 peptide. More interestingly, the acetylated form of the Lys-306 peptide had strongly reduced binding of Ca 2+ /CaM compared to the nonacetylated peptide (Fig. 5C). These data suggest that LysAc of eEF-1A Lys-306 may play a role in regulating the interaction of eEF-1A and Ca 2+ /CaM.

DISCUSSION
To our knowledge, our report along with the companion manuscript (Finkemeier et al., 2011) are the first systematic studies of LysAc of nonhistone proteins in plants. Thus, Arabidopsis is shown to contain a Lys acetylome, as documented recently for eubacteria (E. coli [Zhang et al., 2009] and Salmonella [Wang et al., 2010]) and eukaryotes (various human cell lines: HeLa cells [Kim et al., 2006], cancer cells [Choudhary et al., 2009], and liver cells ). In our study, we identified 64 Lys modification sites on 57 proteins, which are involved in a wide variety of pathways and processes and located in various cellular compartments. One characteristic group was the photosynthesisrelated proteins. From light harvesting and energy metabolism to carboxylation, proteins representative of the full photosynthetic pathway harbor LysAc modifications, as it is also reported by Finkemeier et al. (2011). Understanding the impact of the modifications on these sites holds considerable promise for a deeper understanding of the regulation of the component processes of photosynthesis.
In our LC-MS identifications, we could not distinguish LysAc from trimethylation, because these two modifications only differ by 0.0363 D. The discriminative immunion ions for LysAc or Lys trimethylation peptides reported by Zhang et al. (2004) were not detected in most cases, perhaps due to the limited abundance of in vivo peptides. We took best advantage of ion error maps to manually verify each putative acetylation site (Carroll et al., 2008). Nonetheless, some false positives of Lys trimethyl peptides may still remain. However, our immunoblot results (Supplemental Fig. S3) indicated that LysAc was a much more common modification in Arabidopsis compared to Lys trimethylation, suggesting that peptides containing a lysyl residue with a mass gain of approximately 42 units are most likely to be sites of acetylation. In the case of eEF-1A, the one protein that was clearly acetylated and trimethylated on Lys residues, we confirmed two in vivo acetylation sites with specific antibodies.
We studied two abundant LysAc proteins in Arabidopsis-LHCb and eEF-1A. For LHCb proteins, LHCb1 and LHCb5 were found to contain Lys modification sites close to the N termini of the proteins (acetylated Lys-37 for LHCb1 and Lys-41 for LHCb5). These sites were isoform specific (Jansson, 1999). In the structural view of LHCb proteins, Standfuss et al. (2005) reported that N-terminal-positive residues of LHCb proteins are important in the association of LHCb trimers and cohesion of thylakoid grana. The LysAc on the LHCb N-terminal Lys residues would remove the positive charges of the Lys side chain and consequently may affect the LHCb trimer interactions and thylakoid membrane structure. In our experiments, we found the overall LysAc of the peripheral LHCb was much higher than the tightly PSII-bound LHCb, although the LHCb LysAc did not differ significantly between light and dark samples. This suggests that LysAc may influence the building of LHCb complexes, but may not be directly involved in shuffling LHCb complexes during state transition as is Thr-38 phosphorylation (Vener et al., 2001).
In eEF-1A, we identified five potential LysAc sites and two Lys methylation sites. We showed that one of these sites, Lys-306, may be important in regulating interaction with Ca 2+ /CaM, and thus may regulate eEF-1A function indirectly by altering protein:protein interactions. Whether calcium signaling regulates eEF-1A function in vivo is unclear at present and remains as an important topic for future studies. Acetylation of the Lys-306 site may also directly regulate eEF-1A function. In yeast (Saccharomyces cerevisiae) EF-1a, the directed mutant E317K (equivalent to D307K in Arabidopsis eEF-1A) caused a substantial reduction in protein translation efficiency and fidelity (Sandbaken and Culbertson, 1988). As acetylation of Lys-306 would also affect charge in a similar domain of the eEF-1A protein it is reasonable to speculate that acetylation of Lys-306 might have the same effect as the E317K directed mutant of the yeast enzyme and lower protein translation fidelity. Alternatively, other functions of eEF-1A may be impacted such as actin association (Gross and Kinzy, 2005).
In addition to the characterization of LHCb and eEF-1A, we summarized other LysAc candidate sites that have potential functional implications (Supplemental Table S1). One example is acetylation of Lys-252 in RbcL. In the crystal structure of the spinach (Spinacia oleracea) Rubisco quaternary complex (L 4 S 4 ), Lys-252 is bonded via ionic interactions with Asp-286 at the dimer interface (Knight et al., 1990). This ionic interaction is thought to be involved in the stabilization of the Rubisco hexadecameric complex (Knight et al., 1990). Acetylation of Lys-252 would neutralize the positive charge of the side chain and preclude the salt bridge with Asp-286, likely leading to less stability of the Rubisco holoenzyme. Therefore, the activity of Rubisco holoenzyme would be negatively affected (Gutteridge and Gatenby, 1995), and might explain the inhibition of Rubisco activity by acetylation that was recently observed (Finkemeier et al., 2011). Another interesting example is acetylation of Lys-178 of chloroplast ATP synthase b-subunit. This residue is located in the ATP-binding motif ( 172 GGAGVGKT 179 ) that is conserved across eukaryotic ATP synthase b-subunits (Omote et al., 1992). The equivalent residue of the E. coli F1 ATPase b-subunit ( 149 GGAGVGKT 156 ) when substituted with Ala or Ser (to generate the K155A or K155S directed mutants, respectively) resulted in a10-fold decrease in ATP binding and 100fold decrease in the rate of ATP catalysis (Ida et al., 1991;Omote et al., 1992). It is possible that acetylation of Lys-178 (equivalent to Lys-155 in the E. coli protein) would mimic the charge depletion effect of the Ala and Ser substitutions, and thereby cause down-regulation of ATP synthase catalytic activity. Along with the companion article (Finkemeier et al., 2011), our findings indicate that LysAc may be an important regulatory mechanism across various pathways and processes in Arabidopsis. The findings of LysAc on nonhistone proteins will likely provide new insights and tools for plant biologists to engineer plants in future.

Plant Material
Arabidopsis (Arabidopsis thaliana) ecotype Ws-2 was used in the study and grown in soil as described previously (Oh et al., 2009) and described further in the Supplemental Text S1.

Protein Extraction
Frozen tissue powder was mixed with extraction buffer containing 100 mM Tris, pH 8.0, 1.5 M 2-mercaptoethanol, 4% SDS, 15% glycerol, 5 mM NaF, 1 mM Na 3 VO 4 , 1 mM AEBSF, 2 mM EDTA, and 0.005% bromphenol blue (1 g/3 mL) and vortexed for 30 s. The mixture was heated at 95°C for 5 min, and vortexed another 30 s. Supernatants were collected after centrifugation at 16,000g for 15 min at room temperature and extracts were analyzed by 1-DE and 2-DE followed by immunoblotting as described in the Supplemental Text S1.

Green Native Gels
Shoots were homogenized in 50 mM Tris, pH 7.4, 10 mM MgCl 2 , and 5 mM KCl, and normalized for chlorophyll content before centrifugation at 6,000g for 10 min. The thylakoid pellets were solubilized and loaded for 1-D native green gel analysis according to Allen and Staehelin (1991). Individual bands were excised from green gels and resolved by SDS-PAGE before transfer to polyvinylidene difluoride for immunoblotting.

LC-MS Analysis
Denatured and alkylated proteins were digested with Trypsin or Glu-C (Roche) as described by Shevchenko et al. (2006) and reconstituted for immunoenrichment of acetylated peptides as described by Choudhary et al. (2009). In addition, selected spots of 1-or 2-DE gels were plugged and digested. LC-MS/MS analysis was performed on a Waters Q-Tof API-US Quad-ToF mass spectrometer with a nanoAcquity UPLC system, with a linear gradient of 1% to 60% acetonitrile in 0.1% formic acid over 60 min. MS/MS data were collected using the data directed analysis method in MassLynx to fragment the top four ions in each survey scan. ProteinLynx (Waters) was used to process the mass spectral data into peak list files (PKL) for analysis by Mascot (Matrix Science) and/or Phenyx (Genebio).
The mass spectra were searched against the National Center for Biotechnology Information nr database using the Mascot search engine, version 2.3. The search was limited with a taxonomy filter of Arabidopsis and mass tolerance was set to 0.5 D for peptide and MS/MS masses. The significance threshold was set to the default P value of 0.05. The reverse decoy database setting within Mascot was used to determine the false discovery rate for the searches and only those with values less than 5% are presented. Relevant spectra were visually inspected to confirm the proper assignment of peaks for acetylated residues, and selected spectra annotated to identify the characteristic acetyl-ion peaks shown in Supplemental Figure S9.

Immunocytochemistry
Fully expanded leaves were fixed with 4% paraformaldehyde (Electron Microscopy Sciences), dehydrated in a graded ethanol series, and embedded in paraffin using an automatic tissue processor (ASP 300, Leica). Sections were blocked by Image-iT FX signal enhancer (Invitrogen), incubated in a cocktail of anti-LysAc mouse monoclonal antibodies (Cell Signaling) overnight, and probed with Alexa 488 goat anti-mouse secondary antibody (Invitrogen) and counter stained with either a membrane dye (10 mg/mL) FM143 (Invitrogen) or a nuclear dye (100 mg/mL in water) Hoechst 33342 (Invitrogen). Samples were imaged using a Zeiss LSM 710 laser-scanning confocal microscope (Carl Zeiss) as described in the Supplemental Text S1.

Protein Interaction Analysis
Real-time interactions between recombinant CaM and immobilized biotinylated eEF-1A peptides were monitored with an Octet QK (Forte-Bio) that is based on biolayer biointerferometry (Abdiche et al., 2008).

Supplemental Data
The following materials are available in the online version of this article.
Supplemental Figure S4. Gene ontology for cellular compartments.
Supplemental Figure S9. Mass spectra for LysAc pepdies identified in this study.
Supplemental Table S1. Functional examples of LysAc sites.
Supplemental Table S2. Summary of LysAc sites identified by LC/MS in this study.
Supplemental Text S1. Additional experimental Materials and Methods.