Grapes were collected from Jeongeup (35°424.176' and 126°547.308'), and 5g grape was surface-sterilized with 1.05% sodium hypochlorite for 10min, followed by rinsing five times with sterile distilled water. After grinding in 20ml sterile water, 100μl of each sample was spread onto tryptic soy agar (TSA) medium and incubated at 25°C for 4days. Single colonies were obtained and subsequently streaked on fresh TSA medium. One grayish white, circular, smooth, opaque, and flat isolate, designated as GR-TSA-9^T, was selected for further taxonomic study and preserved in sterile skimmed milk (10%, w/v) at −80°C. The strain was deposited at the Korean Collection for Type Cultures (KCTC) and the China General Microbiological Culture Collection (CGMCC) as KCTC 82386^T and CGMCC 1.18820^T, respectively. Unless otherwise stated, the cells were grown on TSA medium at 25°C for 5days for subsequent tests.

Cell morphology was observed using scanning electron microscopy (Quanta 250 FEG) at the KRIBB Microscopy Core Facility. A Gram staining kit (Difco) was used according to the manufacturer’s instructions. Cells were grown on soft King B (peptone 20g/L, MgSO₄7H₂O 1.5g/L, and K₂HPO₄ 1.5g/L) medium containing agar (0.3, 0.5, and 1.5% w/v) for 5days to test swimming, swarming, and twitching motilities (Belgini et al., 2014). To determine catalase activity, bubble production was observed after adding 3% (v/v) hydrogen peroxide solution to fresh cells (Lee and Jeon, 2017). An oxidase reagent kit (bioMérieux) was used to determine oxidase activity. Growth of strain GR-TSA-9^T was observed using different media, namely nutrient agar (NA, beef extract 3g/L, peptone 5g/L, and agar 15g/L), potato dextrose agar (PDA), Luria–Bertani agar (LB), TSA, marine agar 2216 (MA), reasoner’s 2A agar (R2A), and reinforced clostridial agar (RC). Various temperatures of 4, 10, 15, 20, 25, 30, 37, 40, 45, and 60°C were assessed for optimal growth. Tryptic soy broth (TSB) with different pH values ranging from 3.0 to 12.0 was prepared using 10mM Tris–HCl (Hwang et al., 2018) and adjusted with 1N HCl and NaOH, while salt concentrations ranging from 0 to 15% (w/v; 1% concentration increments; Lee et al., 2019) were prepared to determine the optimal pH and salt concentration, respectively. Optical density at 600nm (OD₆₀₀) was measured using an Optizen POP UV/VIS spectrophotometer (Optizen) to monitor the pH range and NaCl tolerance. The anaerobic growth was tested by incubating strain GR-TSA-9^T on TSA medium in an anaerobic chamber (Coy Scientific) filled with 86% N₂, 7% CO₂, and 7% H₂ at 25°C for 7days. Other biochemical features, such as enzymes, were tested using API 20NE (bioMérieux) or API ZYM with NaCl 0.85% medium (bioMérieux) according to the manufacturer’s protocol.

The almost-complete 16S rRNA gene sequence of strain GR-TSA-9^T was amplified by polymerase chain reaction using the universal primers 27F and 1492R (Lane, 1991), and sequencing was performed using primers 27F/1492R and 518F (5'-CCAGCAGCCGCGGTAATACG-3') and 800R (5'-TACCAGGGTATCTAATCC-3'). The nearly full-length 16S rRNA gene was 1394 nucleotides in length and was compared to the corresponding sequences of the type strains obtained from the EzBioCloud server (Yoon et al., 2017).² BioEdit (V.7.0.5) was used to perform multiple sequence alignments using all validly published members of the family Caulobacteraceae. Phylogenetic trees were constructed using the neighbor-joining (NJ), minimum evolution (ME), and maximum likelihood (ML) algorithms with 1,000 bootstrap iterations in the Molecular Evolutionary Genetics Analysis version 7.0 program (Kumar et al., 2016). Evolutionary distances for the NJ, ME, and ML analyses were calculated using Kimura’s two-parameter model.

Genomic DNA of strain GR-TSA-9^T was isolated using a genomic DNA purification kit (MGmed, Republic of Korea). The quantity and quality of the extracted genomic DNA were measured using PicoGreen and Nanodrop (ratio A260/A280). Whole-genome sequencing was performed at the Macrogen facility (Macrogen, Korea) on the PacBio RSII (Pacific Biosciences, Inc.) and the Illumina sequencing platform and assembled using SMRT Portal (version 2.3) de novo assembler. Potential contamination of the genome sequence was verified using ContEst16S (Lee et al., 2017).³ The assembled sequences were annotated using the National Center for Biotechnology Information Prokaryotic Genome Annotation Pipeline (PGAP) and the Rapid Annotation Subsystem Technology (RAST) server with the SEED database (Tatusova et al., 2016). Metabolic pathways were reconstructed using BlastKOALA based on the Kyoto Encyclopedia of Genes and Genomes pathway database (Kanehisa et al., 2016). The digital DNA–DNA hybridization (dDDH) and average nucleotide identity (ANI) values between strain GR-TSA-9^T and several closely related strains were calculated using the Genome-to-Genome Distance Calculation web server⁴ using the BLAST method and recommended formula 2 (Meier-Kolthoff et al., 2013) as well as an ANI calculator.⁵ The orthoANI values among the closely related strains were calculated using the standalone Orthologous Average Nucleotide Identity (OAT) software (Lee et al., 2016). Whole-genome sequences of closely related strains publicly available at NCBI GenBank were collected, and a whole-genome-based phylogenetic tree was constructed using the up-to-date bacterial core gene set and pipeline as described by Na et al. (2018). Briefly, the 92 core genes were extracted from genomes using Prodigal v2.6.3 (Hyatt et al., 2010) and hmmsearch v3.1b2 (Eddy, 2011). Predicted coding sequences (CDS) of 92 core genes were aligned by using Multiple Alignment using Fast Fourier Transform (MAFFT) for alignments (ver. 7.310; Katoh and Standley, 2013). Then, the phylogenomic tree was inferred by using the FastTree (Price et al., 2010) and viewed using MEGA v7.0 (Kumar et al., 2016). The branch support inference was based on 100 nonparametric bootstrap replicates, and the branch supports for the phylogenomic tree were evaluated using gene support index (GSI).

For fatty acid methyl ester analysis, cells grown for 3days on TSA were extracted according to the instructions of the standard MIDI (Sherlock Microbial Identification System version 6.0), and cellular fatty acid content was analyzed by gas chromatography (Model 6890 N; Agilent) using the Microbial Identification software package (Sasser, 2006). Freeze-dried cells (100mg) were used to extract isoprenoid quinones using a chloroform-to-methanol (2:1, v/v) mixture, followed by analysis using thin-layer chromatography as described by Collins et al. (1980). Subsequent analysis was performed using reverse-phase high-performance liquid chromatography with ultraviolet (UV) absorbance detection at 270nm. Polar lipids were extracted from 100mg of freeze-dried cells with a chloroform/methanol mixture (1:2, v/v) and then identified by two-dimensional thin-layer chromatography on Kieselgel 60F254 plates (silica gel, 10cm×10cm; Merck). In addition, 0.2% ninhydrin (Sigma-Aldrich), α-naphthol, molybdenum blue (Sigma-Aldrich), 4% phosphomolybdic acid reagent, and Dragendorff’s solution were sprayed onto the plates to detect amino group-containing lipids, sugar-containing lipids, phosphorus-containing lipids, total lipids, and quaternary nitrogen-containing lipids, respectively.

The melanin-producing ability of strain GR-TSA-9^T was confirmed by growth on TSA medium supplemented with 0 and 10mg/ml of l-tyrosine followed by incubation at 25°C. Production was performed in TSB consisting of 0, 0.2, 0.4, 0.6, 0.8, and 1.0mg/ml l-tyrosine, and after inoculation with 1ml of the cell suspension (OD₆₀₀=1.0) in 100ml TSB, shaking at 25°C and 150rpm for 1–7days, and melanin production in the broth was measured at OD₄₀₀ using the standard synthetic melanin (Sigma-Aldrich, M8631) calibration curve method (Singh et al., 2018). Melanin was purified as previously described (El-Naggar and El-Ewasy, 2017). Briefly, the supernatant was centrifuged at 8,000rpm to remove the cells, followed by adjustment of the pH of the supernatant to 2.0 using 6M HCl; the samples were then allowed to stand for 4h and centrifuged at 8,000rpm to collect the precipitate. The melanin pellets were washed with distilled water three times and then centrifuged at 8,000rpm for 10min to obtain melanin. The purified melanin was freeze-dried for further use. In vitro-synthesized pyomelanin was produced by auto-oxidation of 10mM HGA (Sigma-Aldrich, H0751) solution at pH 10 with constant stirring for 3days as described by Schmaler-Ripcke et al. (2009).

To analyze the pigment content of GR-TSA-9^T, purified melanin was dissolved in 0.5M NaOH solution, and the absorption value was recorded in the wavelength range of 200–1000nm with a UV–visible spectrophotometer (Thermo Fisher Scientific, USA) by comparing it to a synthetic melanin standard. FT-IR spectroscopy was performed at the Center for Instrumental Analysis, Korea Basic Science Institute, Busan, Republic of Korea. The samples were pressed into disks under vacuum using a KBrpress. The FT-IR spectra of the KBr discs were recorded on an FT-IR Vertex 80v spectrophotometer (Bruker, USA). The spectra were read in the wavenumber region of 400–4000cm⁻¹.

Mass spectrometry analysis was performed using an API 32000 mass spectrometer (AB SCIEX, USA). Samples were dissolved in 1M NaOH solution and then diluted with 50:50 methanol/water. The optimized mass spectrometry parameters were as follows: curtain gas, 10; spray voltage, 4200; ion source gas, 20psi; and flow rate, 10μl/ml. Full spectra were collected in the m/z range of 50–1800 in the positive ion mode.

GR-TSA-9^T cells were cultured on TSA (non-melanized cells) and TSA with 10mg/ml l-tyrosine (melanized cells) at 25°C for 48h. The cells were washed twice with 1× phosphate-buffered saline. Next, 5×10⁷ cells were irradiated with UVC (254nm) at 450mJ/cm, and 10μl of aliquots from tenfold serial dilutions was spotted on TSA plates. All plates were incubated at 25°C for 3days. The experiments were repeated at least twice, with similar results.

Strain GR-TSA-9^T grew well on LB, TSA, MA, NA, and R2A (optimum TSA and LB) but not on RC and PDA. Growth occurred at 10–37°C (optimal 25–30°C), in 0–1% (w/v) of NaCl (optimum 0%), and at pH 7.0–8.0. Cells were rod shaped (0.2–0.3μm in width and 0.9–2.6μm in length) with flagella (Supplementary Figure 1), gram negative, facultatively anaerobic, motile by swimming and twitching, catalase weakly positive, and oxidase positive. Other characteristics of strain GR-TSA-9^T, which is different from closely related strains, are shown in Table 1.

The almost-complete 16S rRNA gene amplicon of strain GR-TSA-9^T contained 1394 nucleotides and has been submitted to GenBank (accession number: MW442968.1). Based on 16S rRNA gene sequence analysis, strain GR-TSA-9^T appeared to belong to the genus Brevundimonas and was closely related to B. lenta DS-18^T (97.9%), B. kwangchunensis KSL-102^T (97.8%), and B. aurantiaca DSM 4731^T (97.5%), whereas it shared less than 97.5% similarity with other related species. Based on the novel species recognition threshold of 98.6% (Kim et al., 2014), strain GR-TSA-9^T was regarded as a novel species in the genus Brevundimonas. A phylogenetic tree was reconstructed using the NJ, ME, and ML methods. The results of these analyses suggest that strain GR-TSA-9^T belongs to the genus Brevundimonas. The genera Asticcacaulis, Caulobacter, and Phenylobacterium formed independent and monophyletic branches (Figure 1). Moreover, B. fluminis LA-55^T, B. viscosa CGMCC 1.10683^T, B. kwangchunensis KSL-102^T, B. lenta DS-18^T, and B. diminuta AJ 2068^T, the type species of genus Brevundimonas, were used to compare phenotypic properties and for chemotaxonomic analyses based on 16S rRNA similarity and phylogenetic tree analysis.

The genome of strain GR-TSA-9^T was assembled using the SMRT Portal (version 2.3) de novo assembler, which is a single circular chromosome of 2,976,716bp in size (the N50 value was 13,858bp), with a coverage of 389X. By comparing two copies of the 16S rRNA gene fragment from the whole-genome sequence, a contaminating DNA sequence was not present in the genome assembly. The genomic GC content of GR-TSA-9^T was 66.4%, which was similar to that of other type strains in the genus Brevundimonas. According to the PGAP annotation, there were 2940 protein-coding genes and 56 RNA genes, including two 5S rRNA genes, two 16S rRNA genes, two 23S rRNA genes, four ncRNAs, and 46 tRNA genes. Cluster orthologous group (COG) annotation results showed that the functional categories of most coding sequences were classified as unknown (28.6% of total assigned COGs), general function prediction only (9.1% of the total assigned COGs), and amino acid transport and metabolism (5.8% of the total assigned COGs; Supplementary Figure 2).

The dDDH value between strain GR-TSA-9^T and the closest species B. lenta DS-18^T was 20.9%, while the ANI value was 76.7%. OrthoANI values based on the entire genome (Supplementary Figure 3) were 72.8–77.6% for the most closely related strains. Thus, strain GR-TSA-9^T was considered a distinct species of the genus Brevundimonas, considering the values obtained were significantly lower than the proposed dDDH (<70%) and ANI cutoff (95–96%) values for bacterial species delineation (Chun et al., 2018). The 92 core gene set-based phylogenetic tree also supported strain GR-TSA-9^T forming a phylogenetic lineage within the genus Brevundimonas, consistent with the 16S rRNA-based phylogenetic tree (Figure 2).

Genome annotation was performed using the RAST server and BlastKOALA to reconstruct the metabolic pathway, and the following sections were predicted from genome sequences.

The cellular fatty acid (>1%) composition of strain GR-TSA-9^T and five closely related strains is listed in Table 2; the most abundant fatty acids of strain GR-TSA-9^T were C_16:0 (14.2%) and summed featured 8 (65.6%). These results are consistent with those of closely related species within the genus Brevundimonas. However, the amount of some components, such as summed featured 3, C_17:1 w8c, and C_18:1 w7c 11-methyl, differs from the reference strains. The respiratory quinone detected in strain GR-TSA-9^T was Q-10, which was consistent with the genus Brevundimonas. The polar lipid profile of strain GR-TSA-9^T consisted of phosphoglycolipids, phosphatidylglycerol, 1,2-di-O-acyl-3-O-[d-glucopyranosyl-(1→4)-α-d-glucopyranuronosyl]glycerol, and unidentified lipids (L1, L2, and L4; Supplementary Figure 4), and lipid spots different from other closely related strains suggested that strain GR-TSA-9^T represents a novel species in the genus Brevundimonas.

Strain GR-TSA-9^T was considered a potential melanin producer on TSA medium and can produce a higher amount of melanin in l-tyrosine-containing medium than without l-tyrosine (Figures 3A,B). On TSA with l-tyrosine, the synthesis of brown pigment was initiated at 3 d and the pigment increased to brown, dark brown, and black-brown at 4, 5, and 6days, respectively. Melanin production in TSB was observed after 3days of shaking, and at 6days, the highest yield of melanin (0.19g/L) was obtained (Figure 3C).

Strain GR-TSA-9^T was found to produce dark brown melanin-like pigment in TSA medium containing l-tyrosine (Figure 3A), suggesting that l-tyrosine is a precursor for melanin biosynthesis in strain GR-TSA-9^T. In addition, it has been reported that Brevundimonas sp. SGJ produces L-DOPA melanin (Surwase et al., 2012a,b, 2013). However, we failed to identify annotated tyrosinases (EC 1.14.18) and laccases (EC 1.10.3.2) with the PGAP and RAST annotation pipelines. These data indicate that melanin synthesis in strain GR-TSA-9^T may not occur via the L-DOPA melanin biosynthetic pathway. Some interesting features of strain GR-TSA-9^T were observed HGA-melanin biosynthesis pathway in the whole genome (Figure 4). According to the HGA-melanin synthesis pathway described previously (Hunter and Newman, 2010; Drewnowska et al., 2015; Wang et al., 2015), all genes responsible for the HGA melanin synthesis pathway were predicted from the whole genome as follows: araT encoding amino acid aminotransferase (E.E.2.6.1.57) converts tyrosine to 4-hydroxyphenylpyruvic acid (HPP), and hppD encoding 4-hydroxyphenylpyuvate dioxygenase (EC 1.13.11.27) catalyzes conversion of 4-HPP to HGA. Then, the produced HGA is converted to pyomelanin through auto-oxidation and spontaneous polymerization. Homogentisic acid is catabolized by a central metabolic pathway that involves three enzymes, homogentisate dioxygenase (hmgA; EC 1.13.11.5), fumarylacetoacetate hydrolase (hmgB/fahA), and maleylacetoacetate isomerase (maiA/hmgC), and then, final products are fumarate (EC 4.2.1.2) and acetoacetate (Arias-Barrau et al., 2004). The pyomelanin production results from a defect in the catabolism pathway (Fonseca et al., 2020). When hmgA levels are low (e.g., gene mutation, deletions, or overexpression of hmgR, HGA accumulates and secretes out of cells, and it lead to pyomelanin though HGA auto-oxidation and self-polymerization. (Arias-Barrau et al., 2004; Rodríguez-Rojas et al., 2009; Wang et al., 2011; Fonseca et al., 2020). In the whole genome of strain GR-TSA-9^T, TetR family member hmgR (PA2010), which was previously shown to bind directly to the hmgA promoter and repress hmgA expression, blocks the central pathway and finally accumulates homogentisic acid and produces pyomelanin (Figure 4A). In strain GR-TSA-9^T, the gene hppD (EC 1.13.11.27) encodes the 4-hydroxyphenylpyruvate dioxygenase, araT (EC 2.6.1.57), and fumarate hydratase (EC 4.2.1.2), which share 62.5, 48.6, and 64% identity with the corresponding protein of Pseudomonas aeruginosa PAO1 (AE004091.2), which was found to produce pyomelanin through the HGA pathway (Bolognese et al., 2019). It is possible that strain GR-TSA-9^T produces pyomelanin through HGA melanin, which encodes araT, phhR encoding σ⁵⁴-dependent transcriptional activator, and hppD, constituting a linear pathway for converting phenylalanine to HGA, which is widely conserved in the genus Pseudomonas (Arai et al., 1980; Arias-Barrau et al., 2004; Rodríguez-Rojas et al., 2009; Orlandi et al., 2015; Hocquet et al., 2016; Zeng et al., 2017; Kurian and Bhat, 2018; Bolognese et al., 2019). This suggests that encoding the most critical enzymes 4-hydroxyphenylpyruvate dioxygenase (EC 1.13.11.27), araT (EC 2.6.1.57), and other enzymes such as hmgR annotated in the whole genome demonstrated that the HGA pathway is the melanogenic pathway in strain GR-TSA-9^T (Figure 4B). In addition, we found brown or black pigments in some of the most closely related strains such as B. Kwangchunensis KCTC 12380^T, B. viscosa JCM 17426^T, and B. lenta KCTC 12871^T strains, but not in B. fluminis KCTC 72717^T and B. diminuta AJ2067^T (Table 1). To determine whether the HGA pathway is specific to strain GR-TSA-9^T or common to the genus Brevundimonas, we investigated the important enzyme-coding genes, such as hppD, araT, and hmgA, involved in the HGA melanin synthesis pathway. Our bioinformatic analysis results showed that the key genes related to the HGA-pathway are widely distributed in the genus Brevundimonas, including Brevundimonas naejangsanensis BRV3, Brevundimonas subvibrioides ATCC 15264^T, Brevundimonas abyssalis TAR-001^T, and Brevundimonas viscosa CGMCC 1.10683^T. However, the genes phhR and araT could not identify from whole genome of B. fluminis KCTC 72717^T. Although the presence of HGA-melanin-related genes is necessary but insufficient for pyomelanin production (if it cannot block and convert into a central pathway), some Brevundimonas species cannot produce pyomelanin because of the incomplete pathway. Based on these analyses, we suggest that the brown-to-black pigment production in many species in the genus Brevundimonas is because of the production of pyomelanin via the HGA pathway rather than the L-DOPA pathway.

The wavelength of maximum absorbance was scanned in the range of 200–1000nm. The wavelength of maximum absorbance of the extracted pigment from the culture supernatant and synthetic melanin standard was observed at 210nm (Supplementary Figure 5), which is a typical feature of melanin (ranges between 196 and 300nm; Surwase et al., 2013; Zeng et al., 2017). The FT-IR spectra were analyzed to confirm that the extracted bacterial pigment was pyomelanin type of melanin. The FT-IR spectra of the extracted bacterial pigment and synthetic pyomelanin standard are shown in Figure 5A. The spectra were recorded at 4000–400cm⁻¹ using an FT-IR spectrophotometer. Both spectra revealed a broad absorption peak at 3650–3100cm⁻¹, corresponding to the ▬OH group and ▬NH group. The peaks at 3000–2900 and 1250–1200cm⁻¹ were attributed to C▬H groups. The strong absorption peak observed at 1450–1770cm⁻¹ was ascribed to the C〓C and C〓O groups. The peak between 1250 and 1200cm⁻¹ corresponds to the N▬H group and C▬N (secondary amine). The high degree of resemblance in the main absorption peaks indicated that the pigment isolated from strain GR-TSA-9^T was pyomelanin. Synthetic pyomelanin and purified bacterial melanin were analyzed by ESI-MS. In addition, nine prominent peaks were observed in the mass spectrum of synthetic pyomelanin with m/z values of 96.5, 121.4, 135.5, 141.4, 167.5, 181.5, 223.6, 239.6, and 263.4 (Figure 5B). These nine peaks were also detected in the mass spectrum of purified bacterial melanin (Figure 5C). In particular, m/z 141.4 and 263.4 peaks were also detected in a previous report (Singh et al., 2018). These results suggest that the purified bacterial melanin is similar to synthetic pyomelanin.

To test the photoprotective property of bacterial melanin, cells were irradiated with short-wavelength UVC, which is a damaging type of UV radiation. UVC is absorbed by DNA, resulting in the formation of pyrimidine adducts and strand breaks (Rastogi et al., 2010). The cultured cells on TSA (non-melanized cells) and TSA with 10mg/ml l-tyrosine (melanized cells) were irradiated with UVC. The melanized cells grew better than the non-melanized cells after UVC irradiation (Figure 6). This finding suggests that cells grown in melanin-containing media showed significantly higher resistance to UV radiation than those grown on TSA medium alone.

Brevundimonas vitisensis (vi’ tis.ne.sis. N.L. fem. Vitis, the generic name of grapevine, refers to the source from which the bacteria were isolated).

Colonies are grayish white, circular, smooth, and opaque (2–4mm in diameter) after growth on TSA medium at 25°C for 3days. Brown pigment is found to be a common attribute of melanin production, with optimal growth occurring at 25–30°C and pH 7.0 and in 0% (w/v) NaCl. Additionally, strain growth occurs on NA, MA, LB, TSA, and R2A (optimal TSA and LB) but not on PD and RC media. Cells are facultatively anaerobic, gram negative, rod shaped, (0.2–0.3μm×0.9–2.6μm), motile, catalase weak positive, and oxidase positive. In the assay with the API 20NE strips, reactions for esculin hydrolysis and β-galactosidase are positive, whereas for other substrates are negative. The production of acid is only from salicin. In the API ZYM strips, alkaline phosphatase, esterase (C4), esterase lipase (C8), leucine arylamidase, trypsin, β-glucosidase, and α-glucosidase are positive and valine arylamidase, cystine arylamidase, α-chymotrypsin, and naphthol-AS-BI-phosphohydrolase are weak positive. The major fatty acids are C_16:0 (14.2%) and summed feature 8 (65.6%). The respiratory quinone detected in strain GR-TSA-9^T is Q-10, while the polar lipid profile of strain GR-TSA-9^T consisted of phosphoglycolipids, phosphatidylglycerol, 1,2-di-O-acyl-3-O-[d-glucopyranosyl-(1→4)-α-d-glucopyranuronosyl]glycerol, and unidentified lipids (L1, L2, and L4).

The type strain is B. vitisensis GR-TSA-9^T (=KCTC 82386^T=CGMCC 1.18820^T), isolated from surface-sterilized grapes from Jeongeup, South Korea.

The accession numbers for 16S rRNA and the whole genome of strain GR-TSA-9^T are MW442968.1 and CP067977.