Structural and Functional Insights into a Lysine Deacylase in the Cyanobacterium Synechococcus sp. PCC 7002 1

Lys deacylases are essential regulators of cell biology in many contexts. Here, we have identi ﬁ ed CddA (cyanobacterial deacetylase/depropionylase), a Lys deacylase enzyme expressed in the cyanobacterium Synechococcus sp. PCC 7002 that has both deacetylase and depropionylase activity. Loss of the gene cddA led to slower growth and impaired linear and cyclic photosynthetic electron transfer. We determined the crystal structure of this depropionylase/deacetylase at 2.1 Å resolution and established that it has a unique and characteristically folded a / b structure. We detected an acyl binding site within CddA via site-directed mutagenesis and demonstrated that this site is essential for the deproprionylase activity of this enzyme. Through a proteomic approach, we identi ﬁ ed a total of 598 Lys residues across 382 proteins that were capable of undergoing propionylation. These propionylated proteins were highly enriched for photosynthetic and metabolic functionality. We additionally demonstrated that CddA was capable of catalyzing in vivo and in vitro Lys depropionylation and deacetylation of Fru-1,6-bisphosphatase, thereby regulating its enzymatic activity. Our identi ﬁ cation of a Lys deacylase provides insight into the mechanisms globally regulating photosynthesis and carbon metabolism in cyanobacteria and potentially in other photosynthetic organisms as well.

Cyanobacteria are the most ancient prokaryotes capable of conducting oxygenic photosynthesis (Schirrmeister et al., 2015). Cyanobacteria-mediated photosynthesis is a major component of total primary production on Earth, playing a vital role in global carbon and nitrogen cycles (Zwirglmaier et al., 2008;Flombaum et al., 2013). Cyanobacteria are likely the ancestors of the chloroplasts found in plant cells through endosymbiosis (Chellamuthu et al., 2013), with many cyanobacterial genes having undergone transfer to eukaryotic cell nuclei during the endosymbiotic process (Gagat and Mackiewicz, 2017). These properties make cyanobacteria ideal model organisms for the study of photosynthesis and many aspects of algal and plant metabolism.
Multiple complex regulatory mechanisms govern the photosynthetic and metabolic pathways within cyanobacteria (Padmasree et al., 2002;Noctor et al., 2004;Singh et al., 2008;Wegener et al., 2010;Tan et al., 2011). While these mechanisms remain incompletely understood, there is increasing evidence indicating that a wide range of posttranslational modifications (PTMs) in both cyanobacteria (Xiong et al., 2016) and plants (Canut et al., 2016) can regulate these processes, making the study of the enzymes regulating these processes of great interest. Of the over 300 known PTMs, protein acylation, including Lys acetylation, propionylation, butyrylation, malonylation, succinylation, and crotonylation are known to be central regulators of a broad range of cellular processes (Lee, 2013). When systematically reviewing PTMs identified in the cyanobacterium Synechococcus sp. PCC 7002 (Synechococcus 7002), we detected many different PTMs, including two major forms of Lys acylation-acetylation and propionylation (Yang et al., 2014). Interestingly, both Lys acetylation and propionylation were found as functional PTMs with the potential to regulate photosynthesis and carbon metabolism in Synechocystis sp. PCC 6803 (Synechocystis 6803; Mo et al., 2015;Yang et al., 2019).
There are two enzyme families responsible for regulating Lys acylation: Lys acyltransferases and deacylases (Olsen, 2012). The CobB enzyme identified in Escherichia coli is a Sir2 (silent information regulator)like enzyme with well-characterized Lys deacetylase functionality (Zhao et al., 2004). It was initially shown to activate acetyl-CoA synthetase (Teo et al., 2014) and to regulate bacterial energy metabolism in E. coli (Castaño-Cerezo et al., 2014). CobB was later found to exhibit Lys deacetylase (Zhao et al., 2004), desuccinylase (Colak et al., 2013), and depropionylase (Sun et al., 2016) activity in E. coli. We have previously shown that CobB also exhibits dual Lys deacetylase and desuccinylase activities in Mycobacterium tuberculosis . At present, however, CobB in E. coli is the only known depropionylase (Sun et al., 2016), and it is not present in cyanobacteria.
In this study, we tested all predicted deacylases in Synechococcus 7002 and demonstrated that A2791, which we have named CddA (cyanobacterial deacetylase/depropionylase), can catalyze Lys depropionylation and deacetylation both in vivo and in vitro. We determined the crystal structure of CddA at a 2.1 Å resolution, revealing the site that is essential for its deproprionylase activity. Based on the results of our global Lys propionylation analysis, we further demonstrated that CddA can catalyze Lys depropionylation and deacetylation of Fru-1,6-bisphosphatase (F/SBPase) both in vivo and in vitro. The functional significance of this enzyme in the regulation of Lys propionylation in photosynthetic organisms is discussed.

Identification of a Novel Enzyme with Deacetylase and Depropionylase Activities in Synechococcus 7002
In order to identify potential deacylases in cyanobacteria, we surveyed the genome of Synechococcus 7002 and identified a total of five enzymes with potential deacetylase activity: acetylpolyamine aminohydrolase (A2791), histone deacetylase/AcuC/AphA family protein (A1628), polysaccharide deacetylase domain-containing Figure 1. CddA (A2791) exhibits deacetylase and depropionylase activity in Synechococcus 7002. A, Coomassie Brilliant Bluestained gel of the purified A2827, A0319, A1994, A1628, CddA, and CobB proteins from E. coli. B, Fluorometric detection of the in vitro deacetylase and depropionylase activities of A2827, A0319, A1994, A1628, CddA, and CobB. Data are presented as means 6 SD and represent results from three independent experiments. C, Fluorometric detection of the in vitro demalonylase, desuccinylase, and deglutarylase activities of A2827, A0319, A1994, A1628, CddA, and CobB. Data are presented as means 6 SD and represent results from three independent experiments. D to G, HPLC assessment of the Lys deacetylase and depropionylase activities of CddA-and CobB-based on the presence of modified and unmodified peptides. Peptides used were as follows: SATEINK(ac)ILSECLNR (D), DLIK(ac)EIRATGARVRL (E), DLIK(pr)EIRATGA RVRL (F), and LVIIVMDRPRHK(pr)DLI (G). Data represent results from three independent experiments. protein (A1994), GlcNAc-6-phosphate deacetylase (A2827), and UDP-3-0-acyl GlcNAc deacetylase (A0319). We found through sequence alignments that both A2791 and A1628 featured conserved domains similar to those of eukaryotic histone deacetylases (Supplemental Fig. S1). To test the potential Lys deacylation activity of these five enzymes, we expressed these genes in E. coli and purified the resultant proteins for use in a fluorescence-based assay (Fig. 1A). In this assay, Boc-Lys(acyl)-7-AMINO-4-METHYLCOUMARIN (AMC) containing different acyl groups with low fluorescence were synthesized as substrates and incubated with each of the five potential deacylases. Any protein with Lys deacylase activity would remove the acyl group and generate a trypsin digestion site. Subsequent tryptic digestion will release the fluorescent molecule AMC and emit an intense fluorescence signal. Based on this assay, we found that A2791 exhibited both strong Lys deacetylation activity and high depropionylation activity similar to that of CobB in E. coli (Fig. 1B). Both A1628 and A1994 were able to deacetylate their target substrates; however, they lacked any depropionylation activity (Fig. 1B). No Lys deacylase activity was detected for proteins A0319 or A2827. Unlike CobB, none of these proteins exhibited any activities of Lys demalonylation, desuccinylation, or deglutarylation (Fig. 1C).
In order to confirm that A2791 was capable of Lys deacylation/depropionylation activity, we used synthetic peptides bearing acetylated lysines ([SATEIN-K(ac)ILSECLNR] and [DLIK(ac)EIRATGARVRL]) and propionylated lysines ([LVIIVMDRPRHK(pr)DLI] and [DLIK(pr)EIRATGARVRL]) as substrates, combining these peptides with recombinant A2791. The resultant peptides were then characterized by high-performance liquid chromatography (HPLC), which revealed that following incubation with A2791 or CobB there were additional detectable HPLC peaks associated with peptides lacking acetyl or propionyl groups ( Fig. 1, D-G). These results demonstrated that A2791 is an enzyme with deacetylase/depropionylase activity, and we have chosen to hereafter refer the gene A2791 as cddA (cyanobacterial deacetylase/depropionylase).

The Structure of CddA
In order to explore the catalytic mechanism of CddA, we next determined its crystal structure at a resolution of 2.1 Å. Pertinent data collection and refinement statistics are summarized in Table 1. This crystal structure revealed that CddA is composed of a single globular domain containing 10-stranded parallel b-sheets and 12 a-helices (including two 3 10 helices labeled h1 and h2; Fig. 2A; Supplemental Fig. S2). Notably, parallel strands 1, 2, and 5 to 10 comprise a slightly twisted plain, with the antiparallel helices 1, 7, 9, and 10 on one side of the plain and the helices 2, 3, 4, 5, and 6 on the other side. This unique a/b structure is consistent with a fold first detected in the binuclear manganese metalloenzyme arginase (Kanyo et al., 1996). The CddA structure has substantial similarity to that of the Aquifex aeolicus HDLP (histone deacetylase-like protein; Finnin et al., 1999), with an overall root mean square deviation of ;1.6 Å. Given this similarity, we predict that CddA is also likely to exhibit a similar catalytic mechanism.
Consistent with the essential role played by zinc atoms in the activity of the majority of HDAC (histone deacetylase) enzymes (Supplemental Fig. S1A), we were able to determine by inductively coupled plasma mass spectrometry (ICP-MS) that CddA binds to Zn 21 with a stoichiometry of 1.0539 per CddA molecule (Table 1). We were further able to detect a Zn 21 binding site in the CddA crystal structure ( Fig. 2A). This binding site is adjacent to a tube-like pocket and is stabilized by His-166, Asp-164, and Asp-248 with coordinated bonds. This tube-like pocket is formed by loops 1, 2, 3, and 4 and likely plays a role in substrate binding (Fig. 2B), given its structural similarities to the known bacterial deacetylase APAH complexed with its substrate acetylpolyamine (PDB: 3Q9F; Lombardi et al., 2011). Residues within this pocket, such as , may be essential for recognizing acetylated or propionylated Lys residues. Together these findings strongly support a role for CddA as a zinc-dependent deacylase.

Mutational Analysis of CddA
We found CddA to exhibit substantial sequence homology to APAH (Lombardi et al., 2011), with conservation of the His-126,  residues that mediate Zn 21 and substrate binding (Supplemental Fig. S2), suggesting substantial structural homology between CddA and APAH. While both CddA and APAH are known to exhibit deacetylase activity, APAH lacks any known depropionylase activity. When we compared the structures of these enzymes, we found that while the overall root mean square deviation of these two structures was only 1.6 Å, there was an additional b-sheet (TELYGEEL) located between helix a1 and a2 in APAH but not present in CddA (Fig. 2C). As this b-sheet was localized directly above the putative substrate binding pocket, we therefore hypothesized that it may act as a lid, partially covering the pocket and thereby preventing entry by larger propionylated Lys groups.
To test this hypothesis, we generated site-directed mutations (H126A, H127A, F136Y, Y192F, L255I, and Y287F) and an insertion mutant (inserting TELYGEEL into CddA) in order to assess the importance of these residues and structures for substrate and Zn 21 binding. The H126A, H127A, and Y287F CddA mutations were associated with markedly decreased deacetylase and depropionylase activity, whereas the F136Y and insertion mutations were associated with a substantial and specific decrease in depropionylase activity (Fig. 2D), supporting our "lid" hypothesis. The Y192F and L255I mutations to residues on top of this pocket region did not affect depropionylase activity.

Effects of cddA Deletion on Synechococcus 7002
In order to explore the importance of CddA in a physiological context, we disrupted cddA via insertion of the chloramphenicol resistance (Cm r ) gene into the coding sequence of the cddA gene at a unique Eco47 III site (Supplemental Fig. S3), and we confirmed successful gene disruption in the DCddA mutant strain via PCR and immunoblotting (Fig. 3A). The expected band at 2 kb from the inactivated gene was detected, while no immunoblotted signal of CddA was observed at 35 kD in the DCddA mutant strain (Fig. 3A). To confirm that no undesirable alterations were introduced, Figure 2. Crystal structure and catalytic activities of CddA. A, Left, overall structure of CddA, with colored secondary structural elements as follows: light green (helices), yellow (strands), and blue (loops). Middle, the N and C terminus of the protein are shown, with a magenta sphere indicating a zinc ion in the active site sphere. Loops 1 to 5 around zinc ion are highlighted in red, orange, pink, yellow, and blue, respectively. Right, a close-up view of the zinc ion binding pocket, with residues coordinated with zinc shown as sticks. B, the CddA tube-like pocket. Left, the potential CddA substrate binding pocket, with the zinc ion (purple) positioned near the bottom of the pocket, surrounded by loops and residues. Right, a closeup view of the pocket presented by surface presentation and key residues. C, Comparative structural analysis of CddA (cyan) and cetylpolyamine amidohydrolase (APAH; blue). The TELY-GEEL sequence in APAH is highlighted in yellow. D, Measurement of the in vitro deacylase activity of CddA and APAH. A fluorometric assay was used to measure the catalytic activities of wildtype CddA and APAH, along with mutant isoforms (H126A, H127A, Y287F, F136Y, L255I, and TELYGEEL). CobB from E. coli was used as a positive control. Data are presented as means 6 SD and represent results from three independent experiments.
we sequenced the genome around cddA (from 2,911,064 to 2,917,665) of the DCddA strain and compared it with the sequence of the wild type (Supplemental Fig. S4A). As expected, no mutation was observed in this genomic region, which contained six Synechococcus 7002 genes (A2789, A2790, cddA, A2792, A2793, and A2794; Supplemental Fig. S4B). The sequencing data files were uploaded onto the PeptideAtlas database with the identifier PASS00818 under the file name "DNA sequencing." However, there is a potential for secondary effects due to the insertion of the Cm r cassette. Under continuous normal light (250 mmol photons m 22 s 21 ) and low light (50 mmol photons m 22 s 21 ) conditions, DCddA grew much more slowly than did the wild type ( Fig. 3, B and C). Under high light conditions (HL; 2000 mmol photons m 22 s 21 ), DCddA growth was not observed (Fig. 3D). Under photoheterotrophic growth conditions, when linear photosynthetic electron transfer was blocked using the PSII inhibitor dichlorophenyldimethylurea (DCMU) and in the presence of glycerol (10 mM), the wild type exhibited a 12.96-h doubling time, whereas DCddA strain exhibited a 28.05-h doubling time (Fig. 3E). When we assessed maximal PSII efficiency (F v /F m ), we found that DCddA mutants and the wild type exhibited comparable efficiency under normal light conditions, whereas under HL conditions, the efficiency of the DCddA strain (0.138 6 0.001) was markedly decreased relative to the wild type (0.162 6 0.003; Fig. 3F).
We next measured photosystem I activity of the mutant and the wild type by monitoring P700 1 reduction. Under an illumination condition, electron flow to P700 1 through cytochrome b 6 f complex (Cytb 6 f) is mostly from PSII (linear electron transport; Yu et al., 1993;Huang et al., 2003). It was reported that the electron flow to P700 1 is likely from reduced ferredoxin through the NDH complex (cyclic electron transport) if PSII activity is inhibited (Huang et al., 2003;Xu et al., 2005;He and Mi, 2016). As shown in Figure 4, when no electron transfer inhibitor was present, linear electron transport is dominant and postillumination P700 1 reduction in wild-type cells is very rapid, with a half-life (t 1/2 ) of 8.4 ms, whereas for DCddA cells, this increased to 13.5 ms (Fig. 4A). The reduction curve in wild-type cells was consistent with an exponential decay profile, whereas in DCddA cells there was an initial bump after the light source was removed. Curve fittings for P700 1 reduction in these cells revealed that the reduction in wild-type cells was consistent with a two-component reduction curve (Fig. 4B). This curve was composed of a rapid component with a t 1/2 of 7.6 ms accounting for 23.2% of overall decay, and a slightly slower component with a t 1/2 of 8.7 ms accounting for 76.8% of overall decay. In DCddA cells, the majority of the decay curve was composed of two phases, although the initial curve was more complex than in wild-type cells (Fig. 4C). In the mutant cells, the fast component with a t 1/2 of 8.6 ms accounted for about 19.2%, and the slow component with a t 1/2 of 14.4 ms accounted for 80.8% of total reduction, respectively. When cells were grown with 10 mM DCMU, thereby blocking electron transfer from PSII, electron flow to the Cytb 6 f complex stemmed from cyclic electron flow around PSI, resulting in an overall decrease in the rate of postillumination reduction of P700 1 (Fig. 4D). In wild-type and DCddA cells, the P700 1 reductions had a t 1/2 of 348 ms and t 1/2 of 568 ms, respectively. Under these conditions, a curve fitting analysis yielded two-component P700 1 reduction curves for both cell types. A fast component with a t 1/2 of 200 to 300 ms was observed in both strains (Fig. 4, E and F). Substantial differences were then evident in the following slow component, which had a t 1/2 of 706 ms in wild-type cells accounting for 31.1% of overall decay, while in DCddA cells it had a t 1/2 of 1129 ms accounting for 46.3% of overall decay, suggesting that a change in this slow component was responsible for the majority of the observed change in P700 1 reduction between wild type and DCddA strains (348 versus 568 ms).

Identification of Propionylated Proteins and Propionylation Sites
The phenotypes observed in the DCddA strain suggested that disrupting deacetylation and/or depropionylation activity in cells had a substantial impact on physiological aspects of Synechococcus 7002. While previous studies have conducted global surveys of propionylation in Synechocystis 6803 (Yang et al., 2019), there have been no published propionylation profiles to date for Synechococcus 7002. We first performed immunoblotting analyses with an antibody specific for propionyl-Lys residues, using whole-cell lysates prepared from Synechocystis 6803, Synechococcus 7002, Synechococcus elongatus PCC 7942, Chlamydomonas reinhardtii, Chlorella sp. NJ-18, Arabidopsis (Arabidopsis thaliana), Spinacia oleracea, and Oryza sativa to broadly characterize propionylation. We detected strong Lys propionylation in all tested organisms, indicating that this PTM is widely used across photosynthetic species (Fig. 5A). To assess how global propionylation shifts in response to environmental changes, we next assessed propionylation in Synechococcus 7002 exposed to different stress conditions. As shown in Figure 5B, under stress conditions including HL and nitrogen depletion, propionylation profiles in these cells changed. A different protein profile was observed under nitrogen-deficient conditions, suggesting that propionylation may be linked to cellular mechanisms involved in responding to nitrogen deficiency and/or nitrogen-carbon balance in Synechococcus 7002. . The effects of cddA deletion on P700 reduction in Synechococcus 7002. A, P700 reduction kinetics for wild type (WT) and DCddA in the absence of DCMU. A.U., Arbitrary units. B, P700 1 reduction curve fitting for the wild-type strain in the absence of DCMU. C, P700 1 reduction curve fitting for the DCddA strain in the absence of DCMU. D, P700 1 reduction kinetics for wild type and DCddA in the presence of 10 mM DCMU. E, P700 1 reduction curve fitting for the wild-type strain in the presence of DCMU. F, P700 1 reduction curve fitting for the DCddA strain in the presence of DCMU. The half times and relative contribution to overall P700 1 reduction for each of the two components are inset. Residuals (Res) are the differences between experiments and fit curves. All depicted spectra represent averages of three biological replicates with five individual readings of each replicate.
Using a combination of immune-mediated enrichment and mass spectrometry, we next characterized in vivo protein propionylation in Synechococcus 7002 (Fig. 5C). We identified 613 unique propionylated peptides encompassing 598 propionylation sites (class I), representing a total of 382 proteins based on a false discovery rate , 1% for modified peptides in this study. We manually inspected all spectra containing propionylation modifications as previously described (Chen et al., 2005;Macek et al., 2008), with detailed information regarding these peptides compiled in Supplemental  Table S1. All tandem mass spectrometry (MS/MS) raw data were uploaded onto the PeptideAtlas database with the identifier PASS00818 under the file name "Propionyl-proteome." We achieved an overall absolute peptide mass accuracy of 0.04439 ppm (SD, 0.8091 ppm) and an average peptide score of 93.02, suggesting that the identification of peptides in this experiment was highly accurate and reliable (Fig. 5D). Of the identified proteins, approximately 33.8% were found to possess multiple propionylation sites (Fig. 5E). We next assessed position-specific amino acid frequencies for the residues flanking identified propionylation sites by comparing the 12 residues surrounding each site to those residues surrounding nonpropionylated lysines in the Synechococcus 7002 proteome (Supplemental Fig.  S5A). Through this analysis, we identified a substantial preference for polar basic residues (Arg) at the 11, 12, and 13 positions, whereas Lys was frequently present at the 26 and 23 positions. In addition, nonpolar hydrophobic residues (Phe) were also enriched at the 21 position. These results suggest that there may be specific propionylation motifs that exist in Synechococcus 7002.
To explore the structural aspects of the identified propionylation sites, we next used the NetSurfP software to assess local secondary structural elements. Of the identified propionylation sites, we found .50% (310) to be located in coil regions, while the fewest (79) were found in b-strand regions. We also found that the average relative side chain solvent accessibility of these propionylated lysines (89.73) was lower than that of all lysines (86.13), unlike what has been found for other Lys PTMs (Ogawa et al., 2015;Yang et al., 2015), suggesting that Lys propionylation may induce unique functional changes in modified proteins (Supplemental Fig. S5B).

Comparative Analysis of Lys Propionylation
We conducted a comparative analysis of the Synechococcus 7002 propionylome with our published Synechocystis PCC 6803 propionylome (Yang et al., 2019). It was found that 45 propionylated proteins were identified in both strains, while 337 and 24 propionylated proteins were identified exclusively in Synechococcus 7002 and Synechocystis 6803, respectively (Supplemental Fig. S6A; Supplemental Table S2). The enrichment analysis showed that 45 shared propionylated proteins were significantly enriched in the photosynthesis (P 5 2.71E-06), light-harvesting complex (P 5 7.22E208), phycobilisome (P 5 2.69E207), and carbon fixation in photosynthetic organisms (P 5 0.00235; Supplemental Fig. S6B; Supplemental Table  S3), suggesting a conserved role of propionylation in the regulation of photosynthesis in cyanobacteria. Among 337 propionylated proteins identified exclusively in Synechococcus 7002, 221, and 40 were found to belong to cellular metabolic process and response to stress, respectively (Supplemental Fig. S6C; Supplemental Table S4), implying the involvement of propionylation in metabolism and adaptive responses in Synechococcus 7002. Moreover, molecular function classification showed that 86 and 67 propionylated proteins were involved in protein binding and hydrolase activity, respectively (Supplemental Fig. S6D; Supplemental Table S4), supporting the idea that propionylation may be involved in regulating enzyme activities in Synechococcus 7002. These results also suggest that propionylation may have different functions in Synechococcus 7002 and Synechocystis 6803.
Our propionylome results were further compared with the published Synechococcus 7002 acetylome (Chen et al., 2017). A total of 1653 acetylation sites in 802 proteins were compared with our identified 598 propionylation sites in 382 proteins (Supplemental Table  S5). We determined that 288 propionylated proteins (75.4% of all propionylated proteins) were also acetylated in Synechococcus 7002 (Supplemental Fig. S7A). Those proteins with overlap between these PTMs were primarily associated with photosynthesis and carbon metabolism, which may indicate that both acetylation and propionylation play important roles in regulating these processes (Supplemental Fig. S7B; Supplemental Table S6A). We additionally conducted enrichment analyses of the nonoverlapping proteins (514 acetylated proteins and 94 propionylated proteins), finding 19 specific biological processes to be enriched for 70 acetylated proteins, and one biological process (translation; P 5 3.43E204) to be enriched for 13 propionylated proteins (Supplemental Table S6B). Further comparison of the specifically modified residues in identified proteins revealed that 227 sites in 160 proteins could be either acetylated or propionylated. These shared sites represented 37.9% (227 of 598) of all detected propionylation sites and 13.7% (227of 1653) of all detected acetylation sites (Supplemental Fig. S7C). There was a marked difference in the frequencies of specific amino acid residues near Lys residues that underwent propionylation or acetylation (Supplemental Fig. S7D), suggesting that these modifications may play different regulatory roles in vivo.

Propionylation Events Linked to Photosynthesis and Metabolism
To explore the functional roles of propionylated proteins in Synechococcus 7002, we next conducted Kyoto Encyclopedia of Genes and Genomes (KEGG) pathway, Gene Ontology (GO), and protein domain enrichment analyses for these proteins (Supplemental Table S7). The GO enrichment analysis of biological processes revealed that the propionylated proteins were mostly enriched in translation (P 5 2.48E215) and photosynthesis-related processes (P 5 0.00317; Supplemental Fig. S8). The Pfam domain and KEGG pathways analyses also revealed that the propionylated proteins were enriched in ribosome, glycolysis, and photosynthesis (Supplemental Fig. S8). In total, 23 proteins involved in light harvesting and photosynthetic electron transfer were propionylated based on mapped KEGG pathways (Supplemental Fig. S9A; Supplemental Table S8). Consistent with previous studies (Sun et al., 2016;Okanishi et al., 2017;Yang et al., 2019), we also found that many enzymes that play key roles in metabolic pathways were propionylated (Supplemental Fig. S9B; Supplemental Table S8). We further found that the large subunit of Rubisco (RbcL) and key regulatory enzymes including glyceraldehyde-3-phosphate dehydrogenase (Gap), phosphoribulokinase (Prk), phosphoglycerate kinase (Pgk), and F/SBPase were propionylated.
To identify the potential substrates catalyzed by CddA, we then compared the global Lys acetylation and propionylation levels between the wild type and DCddA using pan anti-propionylated or acetylated Lys antibodies. In comparison to the wild-type strain, the acetylation and propionylation levels of a few bands in the DCddA strain were increased, although the amount of total protein was almost the same (Supplemental Fig. S10A). Then, the gel bands with the apparent increases in acetylation and propionylation levels were cut from three replicate gels and trypsin digested for further MS analysis. Some proteins with important cellular functions were identified from these gels, including F/SBPase (Supplemental Fig.  S10B; Supplemental Table S9). Among these proteins, the propionylated or acetylated peptides of F/SBPase were further analyzed (Supplemental Table S10). Consistent with the immunoblotting analysis, we observed unique propionylated or acetylated peptides identified in F/SBPase in the DCddA strain, indicating that CddA could deacetylate and depropionylate of F/SBPase in vivo (Supplemental Fig. S10, C-F). The MS raw data of in-gel digestion were uploaded onto the Pepti-deAtlas public database with the identifier PASS00818 under the file name "In-gel proteome". On the basis of the observations, we speculated that F/SBPase is one of CddA's substrates in Synechococcus 7002. F/SBPase is an enzyme capable of dephosphorylating Fru-1,6-bisphosphate (FBP) and sedoheptulose-1,7bisphosphate (SBP) as part of the Calvin cycle and during gluconeogenesis in oxygenic photosynthetic organisms (Fig. 6, A and B). To further confirm F/SBPase catalyzed by CddA, we used it as a substrate for CddA treatment followed by immunoblot-mediated detection of acetyl and propionyl modifications (Fig. 6C). We found that CddA treatment reduced the acetylation and propionylation levels of F/SBPase (Fig. 6, D and E). Moreover, two mutated CddA enzymes (H126A and F136Y) Figure 6. CddA can catalyze F/SBPase Lys deacetylation and depropionylation in vitro and in vivo. A and B, Schematic illustration of the Calvin cycle (A) and gluconeogenesis (B) catalyzed by F/SBPase. C to E, Purified F/SBPase was combined with CddA or heatinactivated CddA in vitro at 37°C for 2 h. CobB from E. coli served as a positive control. The proteins were separated by SDS-PAGE and were visualized through Coomassie Brilliant Blue staining (C) or by transferring to polyvinylidene difluoride membranes for immunoblotting. Densitometric quantification of the F/SBPase acetylation (D) or propionylation (E) levels using anti-acetyl-(Anti-Kac) or anti-propionyl-(Anti-Kpr) antibodies. Each sample was standardized relative to its corresponding Coomassie Blue-stained gel. F, MS/MS analysis of acetylated/propionylated F/SBPase from samples treated or untreated with CddA. G, Enzymatic activity of F/SBPase after incubation with CddA. 1, only CddA; 2, only F/SBPase; 3, F/SBPase with inactivated CddA; 4, F/SBPase with CddA; 5, F/SBPase with CobB. H, Densitometric quantification of the in vivo F/SBPase acetylation and propionylation levels in wild type and DCddA as determined via immunoprecipitation and immunoblotting. I, The specific F/SBPase enzymatic activity in whole-cell extracts of wild type and DCddA. F/SBPase immunoblotting was used to ensure equal loading. Data are presented as means 6 SD from three independent experiments. Asterisks indicate statistical significance was determined by two-sample Student's t test (*P , 0.05, **P , 0.01).
were also purified for the enzymatic activity analysis (Supplemental Fig. S11A). Consistent with the fluorescence-based assay, a modest reduction on deacetylase and depropionylase activity was observed for H126A mutation when compared to the wild-type CddA. The F136Y mutation led to the decreased depropionylase activity of CddA and had no effect on its deacetylase activity (Supplemental Fig. S11, B-D). In order to characterize the sites of acetylation and propionylation sites affected by CddA treatment, we next conducted a MS analysis aimed at assessing the deacetylation and depropionylation of F/SBPase by CddA. Precursor ion intensities were manually quantified using the MaxQuant program (Tyanova et al., 2016). As above, we found that F/SBPase acetylation and propionylation levels were decreased after CddA treatment (Fig. 6F). Specifically, Lys-184 acetylation and Lys-266 propionylation modifications were removed by CddA. Lys acetylation and propionylation remained intact in the absence of the CddA, indicating that CddA mediates Lys-184 deacetylation and Lys-266 depropionylation (Supplemental Fig. S12). Importantly, the enzymatic activity of F/SBPase decreased after incubation with either CddA or CobB (Fig. 6G).
We finally analyzed the Lys acetylation and propionylation levels of F/SBPase in wild-type and DCddA cells via immunoprecipitation and subsequent immunoblotting. The acetylation and propionylation levels of F/SBPase were significantly elevated in the DCddA strain (Fig. 6H), and the enzymatic activity of F/SBPase was also higher in this strain relative to wild-type controls (Fig. 6I).

DISCUSSION
In this study, we tested all predicted deacylases in Synechococcus 7002 and identified CddA as a Lys deacylase in cyanobacteria (Fig. 1). We were able to clearly demonstrate its ability to remove acetyl and propionyl groups from an enzyme (Fig. 6). As CddA lacked a Sir2 domain and exhibited no desuccinylase activity, we concluded that it was not closely related to CobB. Notably, CddA is well conserved across plants and algae (Supplemental Fig. S13). Thus, discoveries regarding how CddA affects Synechococcus 7002 physiology are possibly applicable to plants and algae. It would be interesting and important to investigate the biological functions of CddA homologs in these photosynthetic organisms.
Our reported CddA structure offers insight into the physical basis for its depropionylase activity (Fig. 2). CddA shares a high degree of structural homology with the APAH bacterial deacetylase with a major difference in that the APAH exhibits an additional b-sheet (TELYGEEL; Lombardi et al., 2011). Through sitedirected mutagenesis, we found that the conserved His-126, His-127, Phe-136, and Tyr-287 residues are important for stabilizing the enzyme activities of CddA.
We additionally found that the extra b-sheets (TELY-GEEL) present in APAH act as a lid, potentially influencing the recognition and fit of large substrates in this binding pocket region (Lombardi et al., 2011). Overall, our structural analysis strongly suggests that these conserved residues are important for mediating both the conformation and catalytic activity of CddA.
As we found that CddA in Synechococcus 7002 exhibited dual functionality, we next chose to conduct a proteomic survey of Lys propionylation in this cyanobacteria, as the Lys acetylomes of Synechocystis 6803  and Synechococcus 7002 (Chen et al., 2017) have been published. These reports suggested that Lys acetylation is important for the global regulation of key cellular activities pertaining to photosynthesis and carbon metabolism. Recently, we reported the profiling of Lys propionylation in Synechocystis 6803 (Yang et al., 2019). The propionylated proteins identified in Synechocystis 6803 and Synechococcus 7002 were also enriched in metabolic and photosynthetic pathways (Supplemental Fig. S6). Given these findings, it was not surprising that a strain lacking CddA (DCddA) exhibited impaired growth and electron transfer activity. In this study, the DCddA strain was generated via insertion of the Cm r cassette (Supplemental Fig. S3). While the sequencing detected no mutations in genes close to the site of Cm r insertion (Supplemental Fig. S4), it is possible that the Cm r insertion could alter expression of surrounding genes. For example, A2792 is predicted as a potential membrane protein, and A2793 encodes a homolog of FtsY, which is involved in insertion of proteins in membranes, notably in chloroplast thylakoids. Thus, the insertion of the Cm r cassette may influence the results regarding growth and photosynthesis in the DCddA strain. In the HL conditions, PSII activity (as indicated by F v /F m ) was significantly decreased for the DCddA strain relative to wild-type controls, indicating that without the deacylation activity of CddA, there is additional stress on the electron transfer of PSII. We additionally found that in wildtype cells, P700 1 reduction was rapid, exhibiting exponential decay (Fig. 4A), whereas in mutant cells, this decay was substantially slower (;60%) and substantially deviated from an exponential curve during the initial phase (Fig. 4C). We deduced that both the fast and slow phases of decay were impaired by this mutation. It is possible that multiple pathways may have been affected and that PSII electron donation to the plastoquinone pool and/or Cytb 6 f complex was slowed in DCddA cells (Fig. 4A).
In the presence of DCMU, we found that P700 1 reduction in DCddA cells was roughly 60% slower than in wild-type cells (Fig. 4). P700 1 did however remain exponential in both strains. The slow pace of P700 1 reduction in the DCddA cells treated with DCMU suggests that this mutation may also affect cyclic electron transfer around PSI and/or respiratory electron transfer, thus explaining the marked impairment in photoheterotrophic growth observed in DCddA cells (Fig. 3E). F/SBPase, which can dephosphorylate both FBP and SBP in cyanobacteria (Feng et al., 2014), was identified as a CddA substrate. Our results revealed that CddA was able to deacylate F/SBPase both in vitro and in vivo. The enzymatic activity of F/SBPase was evidently regulated by its acylation status, such that CddA treatment decreased the in vitro activity of this enzyme (Fig. 6). F/SBPase activity has been reported to be mediated at the structural level (Feng et al., 2014). Given that acetylation and propionylation add functional groups to proteins, they may potentially induce structural and functional changes in this enzyme. Lys acetylation and particularly Lys propionylation can also neutralize positive charges and increase surface hydrophobicity (Es-Haghi et al., 2012), increasing the activity of enzymes as hydrophobicity increases (Maiti et al., 2012;Jamil et al., 2015). Cyanobacterial F/SBPase is known to enhance CO 2 fixation and photosynthetic efficiency in transgenic plants (Tamoi et al., 1996(Tamoi et al., , 1998(Tamoi et al., , 2006Miyagawa et al., 2001;Ogawa et al., 2015). Thus, CddA could play a role in metabolism by regulating F/SBPase activity. It is important to note that the overall effects of inactivating CddA in vivo are likely complex and that additional studies are required to understand the regulation of CddA itself and to identify other substrates of CddA.
In conclusion, we identified the first deacylase enzyme that has both deacetylase and depropionylase activities in cyanobacteria. This deacylase has a unique and characteristically folded a/b structure, and its depropionylase activity depends upon an acyl binding site. It can catalyze both in vivo and in vitro Lys depropionylation and deacetylation of a metabolic enzyme, thereby regulating its enzymatic activity. Our results provide novel insight into the mechanisms globally regulating photosynthesis and carbon metabolism in cyanobacteria and potentially in other photosynthetic organisms as well. Future studies are needed to understand how this type of global regulation forms a regulatory network with other regulatory mechanisms, such as the regulation of a rate-limiting step and feedback regulation of metabolism.

Protein Expression and Purification
The Sir-2-like enzyme CobB was cloned from Escherichia coli DH5a (Takara) then expressed and purified as previously described . The fulllength cddA (A2791), A1628, A0319, A1994, A2827, and A1301 genes were amplified from Synechococcus 7002 genome DNA by PCR with the primers shown in Supplemental Table S11. The APAH gene from Mycoplana ramosa was synthesized by using the primers shown in Supplemental Table S11. All genes were then cloned to the expression vector pET21b system (Novagen) to overexpress proteins in E. coli following the procedures described by the manufacturers. The TELYGGEL, H126A, H127A, F136Y, Y192F, L255I, and Y287F mutants were prepared from the pET-21b/A2791 plasmid using standard protocols outlined in the QuikChange site-directed mutagenesis kit (Agilent Technologies), according to the manufacturer's recommendations. Seven pairs of mutagenic primers are shown in Supplemental Table S11. The nucleotide changes are displayed in bold. All proteins were purified by using an affinity Ni 21 column (Qiagen) as described . Purified proteins were concentrated with a centrifugal filter (Amicon) and protein concentration was measured by a bicinchoninic acid protein assay (Beyotime). The purified proteins were checked by SDS-PAGE analysis.

In Vitro Deacetylase or Depropionylase Assay Using F/SBPase
About 3 mg of the purified CddA and its mutants (H126A, F136Y) were mixed with reaction buffer containing 50 mM Tris-Cl, 20 mM KCl, 100 mM NaCl, 25 mM ZnCl 2 , 1 mM b-NAD hydrate (NAD 1 ), and 1 mM DTT, pH 8.0. Then, 3 mg F/SBPase was mixed with the reaction buffer and the mixture was incubated at 37°C for 2 h and subsequently quenched with 1 mL of 10% (v/v) trifluoroacetic acid. The CobB protein from E. coli was used as the positive control. Samples were finally centrifuged at 12,000g for 10 min, followed by western blot.

Crystallization and Structure Determination
The protein was mixed with equal volume of crystallization solution containing 1.26 M sodium phosphate monobasic monohydrate, 0.14 M potassium phosphate dibasic at temperature 291 K. Crystallization was done in hanging drop plates. Crystals were rapidly soaked in the reservoir solution supplemented with 20% (v/v) glycerol as cryo-protectant, mounted on loops, and flash-cooled at 100 K in a nitrogen gas cryo-stream. Crystal diffraction data were collected from a single crystal at Shanghai Synchrotron Radiation Facility BL18U beamline, with a wavelength of 0.9793 Å at 100 K. The diffraction data were processed and scaled with HKL-3000 (Otwinowski and Minor, 1997). The structure was solved by the molecular replacement method and Pseudomonas aeruginosa histone deacetylase-like amidohydrolase (PDB entry 5G12) was used as the starting model (Krämer et al., 2016). The initial model was built using PHENIX autobuild (Adams et al., 2002). Manual adjustment of the model was carried out using the program COOT (Emsley and Cowtan, 2004), and the models were refined by PHENIX refinement (Murshudov et al., 1997;Adams et al., 2002) and Refmac5 (Murshudov et al., 1997). Stereochemical quality of the structures was checked by using PROCHECK (Laskowski et al., 1993). All of residues were located in the favored and allowed region and none in the disallowed region. Refinement resulted in a model with excellent refinement statistics and geometry (Table 1).

Inductively Coupled Plasma Mass Spectrometry Analysis
Purified CddA was digested in 100 mL nitric acid at 65°C for 24 h. Deionized water was added to a final volume of 4 mL, and the sample was then filtered using a nylon membrane (0.45 mm). Zn 21 levels were assessed by PerkinElmer NexION300X ICP-MS equipment (PerkinElmer) by three biological replicates, according to the manufacturers' specifications. Different concentrations (2, 15, 50, and 100 mg L 21 ) of the ICP multielement standard solution IV (Zn; Merck) were used to calibrate the MS unit and deionized water was served as blank solution. Standard curves were plotted, and Zn 21 levels were calculated as micrograms per liter.

Mutant Construction
The cddA gene was amplified from Synechococcus 7002 genome DNA by PCR using the following primers: 59-TGCCCAAGAGGACAAGACCA-39 and 59-CGTAGCCCGCACTGACAATT-39. The PCR product was then cloned into the TA cloning vector pMD19-T (Takara, Japan). Then, the plasmid containing the cddA gene was interrupted at a unique Eco47 III site within the cddA gene by a Cm r cassette (derived from the pRL271 plasmid). Finally, the plasmid containing the interruption was transformed into Synechococcus 7002 as previously described (Eaton-Rye, 2011), and the transformants were selected on chloramphenicol. The cddA gene was then inactivated by a homologous recombination strategy and the resulting mutant was further confirmed by PCR using the mutagenic primers (A2791-F/A2791-R) in Supplemental Table S11. To confirm that no undesirable alterations were introduced, the DNA sequences around cddA from 2,911,064 to 2,917,665 were amplified from Synechococcus 7002 genome DNA by PCR. The amplification products were then purified and sequenced by an ABI 3730XL DNA sequencer (Applied Biosystems). The sequence data were analyzed using CExpress tool in Vector NTI software (Lu and Moriyama, 2004). The primer sequences used for PCR amplification and sequencing were provided in Supplemental Table S11.

Measurements of Fluorescence and P700
Chlorophyll fluorescence and P700 were measured using pulse-amplitude fluorimeter model Dual-PAM-100 (Heinz Walz) as described (Zhao et al., 1998;Shen et al., 2002b;Shen et al., 2002a;Huang et al., 2003). Cells were adjusted to a concentration of OD 730 nm 2.0 in A 1 medium and fluorescence induction was induced by switching on red actinic light with an intensity of 95 mmol photons m 22 s 21 . Saturation pulses (7000 mmol photons m 22 s 21 ) at 30 s intervals were applied to probe the maximal fluorescence yields F m and F m 9, respectively. Minimal (F o ) and variable (F v ) fluorescence values were obtained during the first flash. The F v /F m was calculated with the steady-state and maximal chlorophyll fluorescence by Dual-PAM software. For P700 oxidation reduction kinetics, cells in exponential growth were adjusted to a cell suspension of 30 mg chlorophyll mL 21 with A 1 medium in the absence or presence of 10 mM DCMU. After dark adaptation for 15 min, the redox state of P700 was determined by measuring the absorbance difference signal of 820 2 860 nm, and the P700 1 reduction kinetics were fitted and calculated by GraphPad software as previously described (Zhao et al., 1998;Shen et al., 2002aShen et al., , 2002bHuang et al., 2003).

Cyanobacterial Strains, Culture Conditions, and Extraction of Protein
The Synechococcus 7002 cells were grown in A 1 liquid medium bubbled with filtered air under continuous illumination (250 mmol photons m 22 s 21 ) at 38°C (Ludwig and Bryant, 2011). To determine growth rates, the OD 730 was measured every 8 h for 3 d with a spectrophotometer (MAPADA). Then cells were grown under different treatments and harvested at different growth stages. For nitrogen-deficient conditions, the cells were cultured to exponential phase (OD 730 5 0.7 to 0.8) and resuspended in A 1 medium without nitrogen (Schlebusch and Forchhammer, 2010). For HL treatment, cultures in the exponential growth phase (OD 730 0.7 to 0.8) were harvested and adjusted to an OD 730 of ;0.3 and further exposed to HL (2000 mmol photons m 22 s 21 ) for designated times. For high salt stress and heterotrophic conditions, cells grown to exponential phase were moved to fresh medium containing 2.5 M NaCl or 10 mM glycerol and 10 mM DCMU. Cells at different treatments were harvested and then disrupted by sonication. The cell lysates were stored at 280°C in aliquots for later use.

Production of Specific Antibodies against Synechococcus 7002 Proteins
The generations of polyclonal antibodies against Synechococcus 7002 proteins were carried out by Proteingene. In brief, to produce the polyclonal antibodies against CddA and F/SBPase, full-length cDNA of these genes were amplified and cloned into the pGEX-4T expression vector (Pharmacia). The resulting plasmids were transformed into E. coli strain BL21 (DE3) for overexpression. Cells growing logarithmically were treated with 1 mM isopropyl-b-D-thiogalactopyranoside for 4 h at 30°C. The fusion proteins were then purified by performing His-tag affinity chromatography. Following purification of these antigens, immunization and sampling of the antisera from rabbit were performed by Proteingene, according to standard operating procedures. The specificity of the generated antibodies was determined by the manufacturer using ELISA and western blotting.

Trypsin Digestion, Immunoaffinity Enrichment, and Mass Spectrometric Analysis
For the identification of Lys propionylation, cell lysates were precipitated, washed, and digested by trypsin as previously described (Yang et al., 2013(Yang et al., , 2014. The propionylated peptides were enriched using anti-propionyl Lys antibody (PTM Biolabs) as previous described . After desalting by self-packed C 18 STAGE tips, the enriched propionylated peptides were resuspended in 0.1% (v/v) formic acid and analyzed using an easy nLC-1000 system (Thermo Fisher Scientific) connected to a Q-Exactive (Thermo Fisher Scientific) mass spectrometer with similar parameters as those previously described .
For the identification of modified proteins from Coomassie Blue-stained gels, the specific bands were cut from the gel and digested in-gel by trypsin (Yang et al., 2013;Yang et al., 2014). The peptides were analyzed on an easy nLC-1200 system (Thermo Fisher Scientific) coupled to Q Exactive HF-X (Thermo Fisher Scientific) mass spectrometer. MS and MS/MS data acquisition was performed using Xcalibur 3.0 in the data-dependent acquisition mode. Full MS survey with an mass-to-charge ratio range of 350 to 1800 were acquired in the Orbitrap with resolution r 5 60,000, and the 20 most intense precursor ions with charge state 2 to 6 were sequentially fragmented in each scan cycle by higher energy C-trap dissociation with normalized collision energy of 28%. The exclusion duration for the data-dependent scan was 30 s, and the isolation window was 1.6 mass-to-charge ratio.

Database Search
The MS/MS spectra were searched against the Cyanobase database (http:// genome.annotation.jp/cyanobase/SYNPCC7002, released 2012, 3186 protein sequences) concatenated with a reverse decoy database and common contaminants by using MaxQuant software (version 1.3.0.5; Cox and Mann, 2008). Two maximum missed cleavage sites were permitted for trypsin. The mass error of precursor ions and fragment ions was set to 10 ppm and 0.02 Da, respectively. Carbamidomethylation of Cys was set as a fixed modification and protein N termini acetylation, deamidation of Gln/Asn, oxidation of Met, acetylation (Lys), and propionylation (Lys) were included as variable modifications. Minimum peptide length was six. The maximum false discovery rate for modification site, peptide, and protein were 1%. All MS/MS spectra for acetylated or propionylated peptides were manually checked using a method as described previously (Chen et al., 2005). The identified peptides with C-terminal acetylation and propionylation were discarded to improve the data quality.

Bioinformatics Analysis
The identified propionylated proteins were classified into biological process and molecular function class based on the GO terms using Blast2GO software (Conesa et al., 2005). The enrichment analyses of KEGG pathways, GO terms, and Pfam domains were carried out using the DAVID bioinformatic resource (Huang et al., 2009a(Huang et al., , 2009b. The pathway analysis of propionylated protein was performed using the KEGG pathways database (http://www.genome.jp/ kegg) and previously published data (Zhang and Bryant, 2011 Structure and Function of Cyanobacterial CddA analysis was performed with multisequence alignments using BLASTP (Altschul et al., 1990). Sequence alignment was also performed using ClustalX2 (Larkin et al., 2007) and visualized by CLC sequence viewer (http://www. qiagenbioinformatics.com/product-downloads/). Amino acid sequence motifs (six amino acids upstream and downstream of the propionylated Lys) were analyzed using pLogo (O'Shea et al., 2013). The heat map was also performed by plotting the log10 of the ratio of frequencies.

Immunoprecipitation and Western Blotting
For immunoprecipitations, the experiments were performed according to the instructions of Dynabeads protein G (Invitrogen) with the following steps. First, the beads were incubated with 10 mg of antibody (F/SBPase) for 4 to 6 h at 4°C with gentle rocking. Second, after the supernatant buffer was removed, the conjugated antibodies were incubated with the whole-cell lysates (containing 1 mg of protein in each sample) overnight at 4°C on a rotator. Finally, the conjugate was washed twice with PBST buffer (phosphate-buffered saline with 0.1% [v/v] Tween 20) and then eluted with 30 mL loading buffer. Western blot analysis was carried out according to standard methods. The extracted proteins were boiled in SDS loading buffer for 5 min and then subjected to 12% SDS-PAGE and transferred to a polyvinylidene difluoride membrane for western blotting analysis. The hybridization signals were developed using enhanced chemiluminescence immune-blotting kit (Advansta) and visualized with a fluorescence scanner (ImageQuant). The band intensities of the western blots were quantified using Image J software (NIH).

Enzyme Activity Assay
The activities of F/SBPase were determined as described previously (Sun et al., 2012). In brief, the reaction was started by adding 5 mL 5 mM FBP to 45 mL of assay buffer (50 mM Tris-HCl, pH 8.0, 15 mM MgCl 2 , 10 mM DTT, 0.1 mg purified enzyme, or 5 mg whole-cell lysates) and incubated at 28°C for 5 min. The reaction was stopped by adding 50 mL of 1 M perchloric acid and centrifuged for 5 min at 14,000g, 4°C. About 10 mL sample or standards (Na 2 H 2 PO 4 ) were incubated with 100 mL malachite green solution (0.035% [w/v] malachite green, 0.35% [v/v] polyvinyl alcohol) for 30 min at 25°C, and the A620 was measured using a microplate reader. (Bio Rad). The control FBP was purchased from Sigma. One unit (1U) of FBPase activity was defined as the amount of enzyme that hydrolyzes the substrate to release 1 mmol of inorganic phosphate per minute.

Statistical Analysis
All data represent results from at least three independent experiments. Statistical analysis was performed to assess differences between different experiments. Student's t test was used to analyze the statistical significance. P , 0.05 was considered to be statistically significant. One asterisk and two asterisks indicate P , 0.05 and P , 0.01, respectively.

Accession Numbers
All of the raw MS data files, sequencing data files, and the annotated spectra of all propionylated peptides were submitted to the public PeptideAtlas repository with the identifier PASS00818 (http://www.peptideatlas.org/PASS/ PASS00818) and can be accessed as Supplemental Dataset S1. The coordinates of the CddA have been deposited with the Rutgers Collaborative Structural Bioinformatics database (Protein Data Bank) under accession number 5ZMP.

Supplemental Data
The following supplemental materials are available.
Supplemental Figure S1. Identification of deacetylase and depropionylase in Synechococcus 7002.
Supplemental Figure S3. Schematic showing the construction of DCddA mutant strain by homologous recombination.
Supplemental Figure S4. Comparison of the genomic context around cddA gene.
Supplemental Figure S5. Bioinformatics analysis of propionylation sites.
Supplemental Figure S6. Comparative analysis of propionylated proteins between Synechococcus 7002 and Synechocystis 6803.
Supplemental Figure S7. Comparative analysis of Lys propionylation with acetylation in Synechococcus 7002.
Supplemental Figure S8. Enrichment analysis of propionylated proteins.
Supplemental Figure S9. Overview of the Lys propionylation events involved in photosynthesis and carbon metabolism in Synechococcus 7002.
Supplemental Figure S10. Identification of potential substrates catalyzed by CddA using in-gel digestion and MS identification.
Supplemental Figure S11. The effects of two mutations on the deacylase activities of CddA.
Supplemental Figure S12. LC-MS/MS analysis to determine the residues of F/SBPase deacetylated and depropionylated by CddA.
Supplemental Figure S13. Multiple sequence alignment of CddA from different species (cyanobacteria and plants).
Supplemental Table S1. Detailed information on identified propionylated peptides in Synechococcus 7002.
Supplemental Table S3. Enrichment analysis of 45 shared propionylated proteins identified in Synechococcus 7002 and Synechocystis 6803.
Supplemental Table S4. Go functional classification of propionylated proteins in Synechococcus 7002 and Synechocystis 6803.
Supplemental Table S5. List of overlapping of acetylated and propionylated proteins and sites identified in Synechococcus 7002.
Supplemental Table S6. Enrichment analysis of acetylated and propionylated proteins in KEGG pathways and cellular component in Synechococcus 7002.
Supplemental Table S7. Complete list of GO terms, KEGG pathways, and PFAM domains enriched in Synechococcus 7002 propionylome.
Supplemental Table S8. Complete list of KEGG pathways in the identified propionylated proteins.
Supplemental Table S9. Detailed list of all identified peptides and proteins from the wild type and ΔCddA samples.
Supplemental Table S10. Detailed list of identified peptides of F/SBPase from the wild type and ΔCddA samples.
Supplemental Table S11. List of primers used in this study.
Supplemental Dataset S1. All of the raw MS data files, sequencing data files, and the annotated spectra of all propionylated peptides.