For this study, the Ac group II strain, KACC17005, whose genome information has been fully sequenced (Park et al., 2017), was used as the wild-type strain. Ac strains were grown in TSB (tryptic soy broth, 30 g/l) with appropriate antibiotics at 28°C. For generating the plasmid constructs, Escherichia coli strain DH5α was used, and for cloning Tn5-inserted DNA fragments, the E. coli strain EC100D was used (Epicentre, Madison, WI, USA). E. coli strains were grown in LB (Luria-Bertani; 1% tryptone, 0.5% yeast extract, and 1% NaCl) with appropriate antibiotics. The final antibiotic concentrations in the media were as follows: kanamycin, 50 μg/ml; rifampicin, 50 μg/ml; gentamicin, 10 μg/ml; and ampicillin, 100 μg/ml.

To generate the knockout of glpdAc, the EZ-Tn5 <R6Kγori/KAN-2> Insertion Kit (Lucigen, Middleton, WI, USA) was used. Following the kit procedure, Tn5 was randomly inserted on one site of the Ac genome, and this insertion was verified using Tn5-specific primers 5′-GGATCGCGATGGATATGTTCT-3′ and 5′-AACAGGAATCGAATGCAACC-3′. Genomic DNA was extracted, and the Tn5-disrupted DNA fragment was cloned following the manufacturer's protocol (Lucigen). After cloning the DNA fragment, its DNA sequence was verified by Sanger sequencing with the following primers: forward (KAN-2F), 5′-ACCTACAACAAAGCTCTCATCAACC-3′ and reverse (R6KAN-2R), 5′-CTACCCTGTTGGAACACCTACATCT-3′ (Lucigen). The confirmed strain was named AcΔglpdAc. To create the construct for the complemented strain, the open reading frame of glpdAc was amplified using the following gene-specific primers: 5′-CTCGAGATGAAGATCATCGT-3′ and 5′-AAGCTTCAGTGGTGGTGGTGGTGGTGGCAGGAGCGCAA-3′. The amplicon was cloned into the pGem-T Easy vector (Promega, Madison, WI, USA), generating pGem-GlpdAc, and the cloned DNA was confirmed by Sanger sequencing. The construct was digested with restriction enzymes XhoI and HindIII, and the digested DNA fragment was subcloned into the pBBR1-MCS5 vector, which is a broad host range vector containing the lacZ promoter (Kovach et al., 1995), generating pBBR1-GlpdAc. The generated plasmid was introduced into AcΔglpdAc and selected on tryptic soy agar (TSA) with appropriate antibiotics. The selected transformant was confirmed using PCR with GlpdAc-specific primers and called AcΔglpdAc(GlpdAc). To control for vector side-effects, an empty pBBR1-MCS5 vector was also introduced into Ac and AcΔglpdAc, generating Ac(EV) and AcΔglpdAc(EV) strains, respectively. The transformants were selected on TSA with appropriate antibiotics and then confirmed by PCR with the following pBBR1-MCS5-specific primers: 5′-CAGGGTTTTCCCAGTCACGA-3′ and 5′-ATGCTTCCGGCTCGTATGTT-3′. The generated plasmids and strains are described in Supplementary Table 1.

To predict the putative 3D structure of GlpdAc, the deduced amino acid sequence of GlpdAc was evaluated through the I-TASSER server (Roy et al., 2010), and the PDB file was obtained. This file was imported into the PyMOL program (Molecular Graphics System, San Carlos, CA, USA), generating the putative 3D structure. For comparison, the 3D structure of GpsA, a G3P dehydrogenase in Coxiella burnetii, was also generated using the same method.

Watermelon (Citrullus lanatus var. vulgaris) line SBA, which was kindly provided by Partner Seed Company (Gimje, Korea), was used for virulence testing. A germinated-seed inoculation was used according to a previously established method with slight modifications (Bahar et al., 2009). For effective inoculation, a pre-germination step was added. To prepare the germinated seeds for inoculation, they were placed on sterilized filter papers with 4 ml sterilized water and incubated at 28°C for 2 days; germination was confirmed by visual observation. The Ac strains were grown on TSA with antibiotics at 28°C. The bacteria were collected and suspended in 10 mM MgCl2 to an optical density at 600 nm (OD₆₀₀) of 0.3 (approximately 108 colony forming units [cfu]/ml) and then diluted to 106 cfu/ml with 10 mM MgCl2. Ten pre-germinated seeds were incubated in 20 ml of bacterial suspension and gently agitated for 1 h. After incubation, the inoculated seeds were sown in a 50-pot tray with sterilized soil and sealed with a tray cover for 2 days. The planted seeds were then grown in a growth chamber and checked for disease severity for 7 days. Disease severity was evaluated using a 0-2 scale (0, no symptoms; 1, water-soaked region; 2, seedling wilt) as described previously (Kim et al., 2020; Song et al., 2020). A disease index was then calculated as follows:

The infiltration of watermelon leaves was performed according to a previous protocol (Bahar et al., 2011). The pre-germinated watermelon seeds were sown and grown for 2 weeks, at which time the seedlings had acquired four true leaves. The Ac strains were suspended in 10 mM MgCl2 to an OD₆₀₀ of 0.3 and diluted to 105 cfu/ml with 10 mM MgCl2. The undersides of first and second leaves were inoculated with bacterial suspension using a 3 ml needleless syringe. For counting CFUs, the inoculated leaves were punched using cork-borers (0.4 cm diameter), and two punches were ground in 200 μl sterilized water using a tissue grinder. The leaf extracts were serially diluted (10-fold). The diluted suspensions were spotted onto TSA with antibiotics and incubated at 28°C for 2 days. Cfus were determined 0, 2, 4 6, and 8 days after inoculation. Data from three biological replicates were collected.

The proteomics protocols and conditions, including protein extraction, peptide generation, liquid chromatography-tandem mass spectrometry (LCMS/MS), peptide quantification/identification, comparison of identified proteins, and COG classification, were as reported previously (Park et al., 2020). Briefly, three biological replicates of the Ac strains (a total of six samples) were grown in TSB to an OD₆₀₀ of 0.5 and collected by centrifugation. Total soluble proteins were then extracted from the bacteria, and peptides were generated. A split-free nano-LC system (EASY-nLC II, Thermo Fisher Scientific, Waltham, MA, USA) combined with an LTQ Velos Pro instrument (Thermo Fisher Scientific) was used for the proteomics analysis. Mass spectra from the LC-MS/MS analysis were identified using the Ac strain KACC17005 database from the National Center for Biotechnology Information, and the identified peptides were quantified using Thermo Proteome Discoverer 1.3 (ver. 1.3.0.399) and the SEQUEST program. After peptide quantification, protein levels in Ac and AcΔglpdAc were compared. Finally, differentially expressed proteins were categorized by COG analysis. Student's t-tests (P < 0.05) were used for statistical comparisons.

For bacterial growth assay in TSB medium, Ac(EV), AcΔglpdAc(EV), and AcΔglpdAc(GlpdAc) were harvested, washed twice with sterilized water, and suspended in TSB to an OD₆₀₀ of 0.3 (108 cfu/ml). The bacterial suspensions were then diluted to 105 cfu/ml with TSB and measured for 96 h at intervals of 12 h using a spectrophotometer. To determine the effect of carbon sources on the Ac strains, bacterial growth was assayed in M9 minimal medium containing glucose or glycerol (Hames et al., 2009). For the assay, M9 minimal medium (5× M9 salts, 200 ml; 1 M MgSO4, 2 ml; 1 M CaCl2, 0.1 ml in 1 l) with 20 mL of 20% glycerol or 20% glucose was used. The bacteria were grown on TSA, harvested, washed twice, and suspended in 10 mM MgCl2 to an OD₆₀₀ of 0.05. The bacterial growth was measured for 10 days at 24 h intervals using a spectrophotometer. Three biological replicates were used in this assay.

The assay for cell-cell aggregation was carried out as previously described (Bahar et al., 2009). Ac strains were grown in TSA, washed, and suspended in TSB. The bacterial suspension was adjusted to an OD₆₀₀ of 0.1 (4 ml in 15 ml tube), incubated for 12 h at 28°C, and kept without shaking at 22°C for 2 h. After the bacterial cells had settled at the bottom of the tube, the absorbance at 600 nm (OD_s) of the upper 1 ml of liquid was measured with a spectrophotometer. The residual bacteria were completely dispersed by vortexing and the absorbance at 600 nm was measured (OD_t). The cell-cell aggregation rate was calculated as [(OD_t ‒ OD_s) × 100]/OD_t. For this assay, six biological replicates were collected.

To estimate bacterial tolerance to stress conditions, treatments with 2.5% NaCl for osmotic stress, and 0.1 μg/ml ciprofloxacin for antibiotic stress were used. For osmotic stress, the bacterial strains were grown, harvested, and adjusted to an OD₆₀₀ of 0.3. The bacteria were then treated with 2.5% NaCl for 10 min. For antibiotic stress, the bacteria strains were harvested, suspended at an OD₆₀₀ of 0.1, and exposed to 0.1 μg/ml ciprofloxacin for 2 h. The treated and untreated bacterial suspensions were serially diluted and plated on TSA with appropriate antibiotics. After incubation for 2 days at 28°C, the numbers of colonies were counted. The ratio of stress-treated bacterial cell count to untreated bacterial cell count was calculated for each strain and stress treatment. For this study, at least three biological replicates were used for each strain.

By screening a Tn5 insertion library, we identified one mutant with reduced virulence in watermelon and confirmed that one gene (accession no. ATG93866), identified as G3P dehydrogenase, was disrupted by Tn5. The predicted 3D structure of ATG93866 is similar to that of G3P dehydrogenase (GpsA) from C. burnetii, which is the causal agent of Q fever in humans (Fig. 1A and B). As shown in Fig. 1C, the deduced amino acid sequence of ATG93866 has high similarity with the NAD(P)H-dependent GlpdAc strain AAC00-1, G3P dehydrogenase in Nocardioides sp. SLBN-35, and G3P dehydrogenase in Burkholderiales bacterium 64-34. Therefore, ATG93866 was named GlpdAc (glycerol-3-phosphate dehydrogenase in Ac).

To investigate the role of GlpdAc in virulence, germinatedseed inoculation and leaf infiltration were performed. Ac(EV) (wild-type strain carrying an empty vector), AcΔglpdAc(EV) (glpdAc knockout mutant carrying an emphajeoty vector), and the complemented strain, AcΔglpdAc(GlpdAc) (AcΔglpdAc carrying glpdAc on pBBR1MCS5) were tested. In the germinated-seed inoculation, the disease index of Ac(EV) continuously increased and reached a value of 2 at 7 days after inoculation (DAI). In contrast, AcΔglpdAc(EV) did not cause disease, with a disease index was 0 at 7 DAI (Fig. 2A), demonstrating that GlpdAc is indispensable for virulence in Ac. The virulence of AcΔglpdAc(GlpdAc) was similar to that of Ac(EV), indicating that there were no positional effects by the insertion of Tn5 in GlpdAc. As shown in Fig. 2B, watermelon infected by AcΔglpdAc(EV) did not show disease symptoms. The results of the leaf-infiltration experiments were similar to those of the seed-germinated inoculation. Although the AcΔglpdAc(EV) population slightly increased during the observation period, it was significantly lower than that of Ac(EV) (Fig. 2C). However, the growth of the complemented strain was not different from that of Ac(EV). Similar to the seed-germinated inoculation, AcΔglpdAc(EV) did not induce symptoms on the leaves of watermelon (Fig. 2D). However, leaves infiltrated with Ac(EV) and AcΔglpdAc(GlpdAc) showed necrosis near the site of inoculation.

It is clear that GlpdAc is involved in virulence. In addition, it has been shown that one protein related to biochemical pathways can produce pleiotropic effects in bacteria, including the alteration of virulence and other physiological mechanisms (Felgner et al., 2016; Kim et al., 2020). Therefore, we carried out comparative proteomics to postulate biological and cellular mechanisms related to GlpdAc using the wild-type and AcΔglpdAc strains. In three biological replicates of the Ac strain, 751, 757, and 752 proteins were found from 59,405, 58,831, and 59,569 peptide spectral matches (PSMs), respectively; 726, 695, and 722 proteins from 59,367, 58960, and 59,261 PSMs, respectively, were identified from the AcΔglpdAc strain (Supplementary Table 2). Among them, 696 and 658 proteins were commonly detected in the three biological replicates of Ac and AcΔglpdAc, respectively, and these proteins were used for the comparative analysis. It was found that 51 and 27 proteins were more abundant (>2-fold) in Ac and AcΔglpdAc, respectively (Fig. 3A). Seventy-eight proteins were subjected to COG classification. The COG analysis revealed that proteins classified in groups F (nucleotide metabolism and transport), G (carbohydrate metabolism and transport), N (cell motility), and U (intracellular trafficking and secretion) were uniquely found in the Ac strain, and proteins categorized in groups I (lipid metabolism), L (replication and repair), and Q (secondary structure) were detected only in AcΔglpdAc (Fig. 3B, Supplementary Tables 3 and 4). Proteins belonging to other categories were mostly abundant in the Ac strain. Interestingly, fructose 1/6-bisphosphatase, C4-dicarboxylate ABC transporter substrate-binding protein, twitching motility protein PilT, and universal stress protein were only detected in Ac. These results indicate that GlpdAc is involved in diverse cellular processes in Ac, including carbohydrate metabolism and cell wall/membrane biogenesis.

Sequence comparison and proteomics analysis of GlpdAc reveal that the protein may be associated with glucose/glycerol metabolism. Therefore, bacterial growth in the presence of glucose or glycerol as a sole carbon source was investigated. First, we examined the growth of Ac(EV), AcΔglpdAc(EV), and AcΔglpdAc(GlpdAc) in a rich medium to test the effect of GlpdAc on bacterial multiplication (Fig. 4A). During the measurement period, the OD₆₀₀ values of the three strains were not significantly different, indicating that GlpdAc is not associated with bacterial reproduction. As shown in Fig. 4B, the growth of AcΔglpdAc(EV) barely increased in M9 medium with glucose, but the complemented strain showed similar growth to that of Ac(EV). Notably, the growth of AcΔglpdAc(EV) in M9 medium supplemented with glycerol was not different from that of Ac(EV) and AcΔglpdAc(GlpdAc) (Fig. 4C). These data suggest that the function of GlpdAc is related to glycolysis.

In the comparative proteomics analysis, the abundance of diverse proteins related to cell wall/membrane/envelope, including PilT, which is associated with cell aggregation (Bahar et al., 2009), was altered (Supplementary Tables 3 and 4). To examine the effect of GlpdAc on cell aggregation, a bacterial cell sedimentation assay was conducted using fully grown Ac(EV), AcΔglpdAc(EV), and AcΔglpdAc(GlpdAc). As shown in Fig. 5, the bacteria that remained in suspension after allowing the medium to stand for 2 h and those that had pelleted were quantified, and the cell-cell aggregation ratio was calculated. Ac(EV) showed a cell-cell aggregation ratio of 7.76%. The ratio for AcΔglpdAc(EV) was 16%, which was approximately 2-fold higher than that for Ac(EV). The cell-cell aggregation in AcΔglpdAc(GlpdAc) was restored to the level in Ac(EV).

In the proteomics analysis, a universal stress protein (ATG96448) was found to be more abundant in Ac than in AcΔglpdAc. Proteins associated with cell membrane/wall/envelope biogenesis were also abundantly detected in the comparative proteomics analysis. Thus, the tolerance to osmotic stress and antibiotics in Ac(EV), AcΔglpdAc(EV), and AcΔGglpdAc(GlpdAc) was investigated. After treatment with 2.5% NaCl, 6.61% of AcΔglpdAc(EV) had survived, compared to 16.7% of Ac(EV) (Fig. 6A), indicating that the mutant was less tolerant to osmotic stress. Survival was restored to wild-type levels in the complemented strain. Notably, with 0.1 μg/ml ciprofloxacin treatment, AcΔglpdAc(EV) displayed higher survival (7.79%) than both Ac(EV) (5.93%) and AcΔglpdAc(GlpdAc) (4.22%) (Fig. 6B).