On 13th October 2019, four corms of taro plants showing wilting leaves were collected from Laojun Village, Ruyuan County, Shaoguan City, Guangdong Province, China. Microbial separation from the symptomatic taro was carried out using the method previously described (Hu et al., 2021b). Firstly, the corms were washed with tap water and then surface-disinfected with 70% ethanol solution. Secondly, the corms were cut open using a sterile knife in the biosafety cabinet. Then, rotten tissue was picked out from the center part with a sterile inoculation loop and streaked onto a Luria-Bertani (LB, containing 10 g/L typtone, 5 g/L yeast extract and 10 g/L NaCl) agar (1.5% w/v) plate for incubation at 28°C for 24 h. Finally, single colonies were streaked onto fresh medium plates for purification and grown in LB medium with shaking at 200 rpm overnight for preservation.

Bacteria were grown in LB medium until OD₆₀₀ of 1.0, and genomic DNAs were extracted using the EasyPure Bacteria Genomic DNA Kit (TransGen Biotech, Beijing, China). The 16S rDNA gene sequences were amplified using the primers 27f and 1492r (Coenye et al., 1999). The products were examined using 1% agarose gel electrophoresis and sent to Sangon Biotech Company in Shanghai, China, for sequencing. SeqMan V.5.00 was used to assemble sequences generated from forward and reverse primers.

To test whether LJ1 and LJ2 could cause soft rot in the field, we carried out potted experiments with wound and non-wound inoculation. Bacterial strains were first cultured to the logarithmic phase till OD₆₀₀ of 1.2 in LB medium. For wound inoculation, 100 μL of LJ1, LJ2, and LJ1 + LJ2 mixture were, respectively, injected into the taro seedling tuberous roots. LB medium was used as a negative control. For non-wound inoculation, 10 mL of LJ1, LJ2, and LJ1 + LJ2 mixture were, respectively, centrifuged, re-suspended with 50 mL of PBS (Phosphate Buffer Solution), and then poured into the potted plants. PBS buffer was used as the negative control. All the pots were kept in a greenhouse (28 ± 2°C, 75 ± 15% relative humidity, and 12 h alternating light and dark cycles) and watered once a day with sterile water until disease symptoms were observed. The experiment was repeated three times in triplicate. To fulfill Koch’s postulates, the bacteria were re-isolated from the diseased slices compared with the corresponding sequences of the inoculated ones for their 16S rDNA gene sequences.

To test the pathogenicity of LJ1 and LJ2 on different plants, bacterial strains were grown in LB medium till OD₆₀₀ of 1.2, and different plant organs were selected using different inoculation methods. P. carotovorum BC9 was used as a control. For potato (Solanum tuberosm), Chinese cabbage (Brassica pekinensis), celtuce (Lactuca sativa var. angustata), and onion (Allium cepa), tubers/stems were washed with tap water and dried with a paper towel. Potato and celtuce were subsequently surface-sterilized with 70% ethanol and then sliced evenly about 5 mm in thickness. Plant tissue was placed in a tray with moistened filter paper, and 2 μL of bacterial culture was applied to the center of the tissue pierced with pipette tips. For Alocasia macrorrhiza, 100 μL of bacterial culture was injected into the basal stems of the seedlings, and 5 μL onto the leaves. All trays were kept in a growth chamber (Shanghai YiHeng Scientific Instruments Co., Ltd., Shanghai, China) with controlled conditions as 28 ± 2°C, and 75 ± 15% relative humidity till symptoms appeared. Same volume of LB medium was inoculated as the negative control. Each assay was repeated three times with triplicate. The areas of lesions were measured using Image J 1.52s software (Schneider et al., 2012).

Biochemical characteristics of strains LJ1 and LJ2 were determined using Biolog GEN III microplates (bioMérieux) following the manufacturer’s protocol. For chemotaxonomic analysis of fatty acids, biomasses of exponentially growing cells of LJ1 and LJ2 cultured in MB medium (TOPBIO, Shandong Tuopu Bioi-engineering Co., Ltd., Zhaoyuan, China) at 25°C for 3 days were harvested and lyophilized. Cellular fatty acids were saponified, methylated, and extracted according to the protocol of the Sherlock Microbial Identification System (MIDI) and identified by gas chromatography (model 7890B; Hewlett Packard, Santa Clara, United States) using the Microbial Identification software package with the MIDI Sherlock system and the Sherlock Aerobic Bacterial Database (RTSBA6 6.21).

To determine whether there is an antagonistic interaction between LJ1 and LJ2, we performed an antibacterial assay using the method previously described (Hu et al., 2018). In brief, single colonies of isolates LJ1, LJ2, and Escherichia coli DH5α on LB agar plates were transferred into LB medium for growth with shaking at 200 r/m at 28°C (or 37°C for E. coli DH5α) till OD₆₀₀ of 1.2. To test the antagonism of LJ1 against LJ2, 1 mL of LJ2 bacterial culture was added into 20 mL of 1% melted agarose (cool to 50∼60°C), mixed thoroughly, and then laid onto a LB agar plate (10 cm × 10 cm). After solidifying, wells of 5 mm in diameter were punched and applied with 20 μL of LJ1 bacterial culture. Antagonistic assay of LJ2 against LJ1 was performed in the same way. LB medium was used as the negative control, and P. carotovorum BC9 reserved in our lab, which produces antibacterial substances, was used as the positive control. The plates were incubated at 28°C for 24 h. The experiment was repeated three times in triplicate.

To determine whether strains LJ1 and LJ2 promote or inhibit each other in planta, we measured the bacterial cell numbers by counting their corresponding colony-forming units (CFU) according to the method previously described (Hu et al., 2021a; Xue et al., 2021). Firstly, we visualized these two strains by, respectively, importing plasmids pBBP_gdh (pmCherry) (Liao et al., 2018) and pLAFR3-GFP (pGFP) (Shi et al., 2019) into strains LJ1 and LJ2 by triparental mating with the help of E. coli HB101(RK2013). Secondly, 10 μL of bacterial cultures (grown in LB medium till OD₆₀₀ of 1.5) of LJ1(pBBP_gdh), LJ2(pLAFR3-GFP), and LJ1(pBBP_gdh) + LJ2(pLAFR3-GFP) mixture were, respectively, spotted onto taro slices. LB medium was spotted as the negative control. The slices were placed on Petri dishes, sealed and kept in a biological incubator under the conditions described above. After 24 and 48 h, the diseased tissue was cut off and weighed, and then ground into homogenate with 10 mL of PBS buffer. The homogenate was diluted in 10× series; 100 μL of each gradient dilution was spread onto fresh LB agar plates and incubated at 28°C overnight. Colonies on plates were counted (plates with less than 30 or more than 300 colonies were considered outliers and discarded). The experiment was repeated three times in triplicate.

Genomic DNAs were extracted from LJ1 to LJ2 in LB medium culture using EasyPure^® Bacteria Genomic DNA Kit (Transgen, Beijing, China). The harvested DNAs were detected by the agarose gel electrophoresis and quantified by Qubit^® 2.0 Fluorometer (Thermo Scientific, Waltham, MA, United States). The complete genomes were sequenced by Grandomics Biosciences (Beijing, China) using the Nanopore PromethION platform and Illumina MGISEQ T7 platform. The genome assemblies were generated using Flye v2.7 with long-read data alone, which were then polished with both long- and short-read data by medaka v1.2.5 and pilon v1.2.3 (Wick et al., 2017), respectively. The sequencing data and genome assemblies have been deposited in the NCBI database under the accessions CP084032.1 (PRJNA765110) and CP084023.1 (PRJNA765112), respectively.

The circular visualization of LJ1 and LJ2 genome characteristics was created by Circos (Krzywinski and Schein, 2009). PGAP (2020-09-24.build4894) (Tatusova et al., 2016) was used for genome annotation. Virulence factors were predicted by PathoFact v1.0 with default settings (de Nies et al., 2021), and only the predictions with high confidence levels (i.e., “1: Secreted Virulence factor” and “2: Non-secreted Virulence factor”) were considered.

Pairwise average nucleotide identity (ANI) values between LJ1, LJ2, and all 327 Pectobacterium genomes available in the NCBI RefSeq database (Supplementary Table 1) were calculated using fastANI v1.3. Digital DNA-DNA hybridization (dDDH) values between LJ1, LJ2, and Petobacterium genomes that share the highest ANI values with LJ1 or LJ2 were then determined by using the GGDC server.² Orthologous gene groups were constructed using OrthoFinder v2.5.4 from annotated proteins in the genomes of LJ1, LJ2, and representative genomes of the 20 Pectobacterium species. The OrthoFinder analysis was conducted with default settings except that the sensitivity of diamond search was set to “ultrasensitive.” The orthologous groups defined in the “Orthogroups.tsv” result file were used in downstream analyses such as the identification of unique genes or orthologs of previously reported SRP genes encoding secretion systems and CWDEs (Bell et al., 2004; Coulthurst and Palmer, 2008; Rodríguez-Sanz et al., 2010; Pérez-Mendoza et al., 2011; Zhou et al., 2015; Gorshkov et al., 2018; Zhang et al., 2018). The genome annotations and orthologous groups generated in this study have been deposited in the FigShare repository and available at https://figshare.com/projects/Isolation_and_genome_analysis_of_Pectobacterium_colocasium_sp_nov_and_P_aroidearum_two_new_pathogens_of_taro/134858. A phylogenetic analysis of LJ1, LJ2, and the representative Pectobacterium genomes was inferred based on a set of 1,621 single-copy orthologous genes present in most analyzed genomes. Single-gene alignments were generated using MAFFT v7.490 (with the “E-INS-i” iterative refinement method), trimmed for gappy columns using trimAl v1.4 (with the “gappyout” option), and then concatenated into one supermatrix. The phylogenetic inference was performed using IQ-TREE v2.1.2 with an automatic selection of the best-fitting evolutionary model for each gene, and the reliabilities of internal branches were estimated by ultrafast bootstrap analysis with 1,000 replicates.

The activities of cell wall degrading enzymes (CWDEs) were measured using the assay medium recipe described previously (Zhou et al., 2016). Briefly, 30 mL of each assay medium containing enzyme reaction substrate was poured into a 13 × 13 cm square plate. After solidification, wells (5 mm in diameter) were punched, and 20 μL of bacterial cells (OD₆₀₀ of 1.5) were added into the wells. All the assay plates were incubated at 28°C. After 11 h, the pectate lyase (Pel) assay plate was covered with 4 M HCl; after 14 h, the cellulase (Cel) assay plate was stained with 0.1% (w/v) Congo Red for 15 min and discolored with 1 M NaCl twice. Protease (Prt) activity was indicated by the transparent halos surrounding the wells after 24 h without any treatment. The experiment was repeated three times in duplicate. GraphPad Prism 8.4.1 was used to perform the unpaired two-tailed t-test.

This disease happens in all growing stages. In the early seedling stage (1–5 leaves, 0–3 months old), the plant shows green wilting with drooped down leaves (Figure 1A); inside the basal pseudostem, brown rot with smelly bacterial pus can be seen in vascular bundles, the edge of the lesion is dry, and the middle is wet and rotten (Figure 1B). In the middle stage (6–8 leaves, 4–6 months old), leaves of the infected plant curl up, some lodge (Figure 1C), and bean curd-like rotten tissue can be observed in the corm (Figure 1D). In the harvest stage (corm growing, 7–8 months old), the infected plant wilts, leaves remain green but become frizzy (Figure 1E). The corm has a strong odor, and the internal tissue decays completely after splitting (Figure 1F). The petiole and leaf blade remain green throughout the disease course, indicating that pathogens preferentially colonize underground corms.

On 13th October 2019, four corms of taro plants showing wilting leaves were collected from Laojun Village, Ruyuan County, Shaoguan City, Guangdong Province, China. Rotten tissues of the soft rot corms were used for pathogen separation. As a result, 18 isolates in total (names as LJ1–LJ18) were isolated from the tissue, which were putatively classified as Pectobacterium aroidearum (eight isolates), Bacillus sp. (four isolates), Klebsiella sp. (three isolates), P. carotovorum (two isolates), and Weissella cibaria (one isolate) based on 16S rDNA sequence analysis (Supplementary Table 1).

Given that Pectobacterium spp. can cause soft rot disease in many kinds of crops and vegetables, we selected two of the Pectobacterium isolates for pathogenicity tests, namely LJ1 and LJ2, which were closest to P. aroidearum SCRI 109 and P. aroidearum L6, respectively. Inoculums of pure LJ1, pure LJ2, and LJ1 + LJ2 mixture were each applied to taro seedlings with both non-wound (irrigation) and wound (injection) inoculation methods. At 20 days post-inoculation (dpi), wilting symptoms were observed in taro plants irrigated with LJ1, LJ2, or their mixture. Similar symptoms were visible at 7 dpi in injected plants (Figure 2A). Brown rotten tissue could be observed after longitudinal incision of the corms of inoculated plants, whereas all control taro plants remained symptom-free (Figure 2A). The pathogens were further re-isolated from the diseased tissue and confirmed as the same original strains by 16S rDNA sequencing, thus fulfilling the Koch’s postulates.

To explore the host range of LJ1 and LJ2, we also performed pathogenicity tests on potato, Chinese cabbage, celtuce, onion, and Alocasia macrorrhizos. Results showed that both strains can also infect these plants except onion; LJ2 was more virulent than LJ1 on Chinese cabbage and celtuce, while both caused only slight disease symptoms on the leaves of A. macrorrhizos (Supplementary Figure 1).

Since LJ1 and LJ2 were isolated from the same diseased corm, we further examined whether there is any interaction between them. Firstly, antagonistic activity between the two isolates was measured, and results showed no growth inhibitory halo around the wells applied with LJ1 (or LJ2) culture on the bacterial lawn plate of LJ2 (or LJ1) (Supplementary Figure 2). Furthermore, we inoculated both taro corm slices and plants with mCherry-labeled LJ1, GFP-labeled LJ2, or the mixture of both to assess the interaction of the two strains within host by evaluating disease severity and cell propagation. Twenty-four hours post-inoculation, both individual inoculations of either strain and the co-inoculation of LJ1 + LJ2 resulted in macerated taro corm slices with similar lesion areas (Figure 2B). Forty-eight hours post-inoculation, LJ2 alone and the LJ1 + LJ2 mixture resulted in comparable lesions significantly larger than the lesion caused by LJ1 (Figure 2B). In concordance with this data, LJ1 and LJ2 exhibited comparable numbers of CFU in decayed tissues in both pure inoculation and co-inoculation at 24 dpi, whereas LJ2 became considerably more abundant than LJ1 in both conditions at 48 dpi (Figure 2C). Notably, LJ1 was only slightly more abundant (1.47-fold) in pure inoculation than in co-inoculation at 24-hpi, and exhibited comparable CFU numbers in both conditions at 48-hpi (Figure 2C). A similar trend was observed for LJ2, which had similar CFU numbers at 24-hpi, and was 2.07-fold more abundant in pure inoculation than in co-inoculation at 48-hpi (Figure 2C). Overall, the results of co-infection assays suggest that LJ1 and LJ2 do not interact with each other.

To better understand the molecular basis of their pathogenicity, the genomes of LJ1 and LJ2 were sequenced using both Nanopore PromethION platform and Illumina NovaSeq platform. The genomes of LJ1 and LJ2 were both assembled into single, circular chromosomes of 4,912,018 bp and 4,857,800 bp in size, and 51.65 and 51.87% in GC-content, respectively (Supplementary Figure 3 and Table 1). A total of 4,316 and 4,255 protein-coding genes were predicted in the genomes of LJ1 and LJ2, respectively; the two strains both have 108 RNA genes, all of which are conserved except one unique ncRNA gene in LJ1 and one unique tRNA gene in LJ2. Interestingly, LJ1 contains substantially more transposases (71:21) and pseudogenes (122:70) than LJ2 (Table 1).

To determine the taxonomic classifications of LJ1 and LJ2, we compared their genomes with the 327 Pectobacterium genomes available in the NCBI RefSeq database and a custom assembly of P. cacticida ICMP 11136 (see Section “Materials and Methods”), which altogether represent all the 20 known Pectobacterium species. We first calculated the average nucleotide identity (ANI) values between all genomes. The results showed that LJ1 is most closely related to Pectobacterium sp. CFBP8739 (GCF_013449375.1, isolated from irrigation river water in France) with an ANI value of 93.68% (Supplementary Table 2) and a dDDH value of 52.00%, both below the commonly accepted thresholds (ANI ≥ 95%, and dDDH ≥ 70%) for species delineation. For LJ2, the closest relative is P. aroidearum L6 (GCF_015689195.1) with an ANI value of 98.33% (Supplementary Table 2) and a dDDH value of 84.50%, indicating that LJ2 belongs to P. aroidearum.

We then performed a phylogenomic analysis to investigate further the relationships between LJ1, LJ2, and representative strains of the 20 known Pectobacterium species. The phylogeny showed that LJ2 and P. aroidearum L6 are sister to each other and very closely related, corroborating the assignment of LJ2 to P. aroidearum. These two P. aroidearum strains are then grouped with LJ1; however, they are separated by a longer evolutionary distance in the tree than that between LJ2 and J6, consistent with the relatively low ANI values between LJ1 and LJ2 (91.35%) or L6 (91.40%) (Supplementary Table 2 and Figure 3). All other Pectobacterium species are more distantly related to LJ1/LJ2.

Since our genomic analyses indicate that LJ1 might represent a novel species, we decided to perform biochemical and fatty acid analyses of LJ1 and LJ2. As shown in Table 2, the two strains exhibited several differences in their utilization of carbon sources, including that: (1) LJ1 could not use citric acid, while LJ2 could use; (2) LJ1 was weakly responsive to D-cellobiose, D-raffinose, D-melibiose, and Tween 40, while LJ2 was negative to them; and (3) LJ2 had a weak response to 3-methyl glucose, inosine, D-arabitol, and L-alanine, while LJ1 did not. Our chemotaxonomic analysis showed that the major (>5%) cellular fatty acids of strains LJ1 and LJ2 both included summed feature 3 (28.96 and 29.44%, respectively, comprising C_16:1ω6c and/or C_16:1ω7c), C_16:0 (28.75 and 25.55%, respectively), summed feature 8 (18.06 and 19.47%, respectively, comprising C_18:1ω6c and/or C_18:1ω7c), summed feature 2 (10.39 and 9.29%, respectively, comprising C_14:0 3-OH, iso-C_16:1 I and/or C_12:0 aldehyde), and C_12:0 (5.71 and 6.69%, respectively) (Table 2). In addition, C_17:0 cyclo and C_15:1 ω8c were detected in LJ2 but not LJ1 (Table 2). Both strains were most closely related to P. carotovorum in their chemotaxonomic profiles, but the similarity index of LJ1 (0.542) was considerably lower than that of LJ2 (0.673). The above phenotypic and genomic data all indicate that strain LJ1 should represent a novel species of the genus Pectobacterium, for which the name “Pectobacterium colocasium sp. nov.” is proposed after the species name of its natural host.

Since P. aroidearum LJ2 and “P. colocasium sp. nov.” LJ1 are close relatives but differ in their virulence against multiple hosts, we carried out pan-genome analysis to identify genes unique to either species (i.e., lack orthologs in the other species). As a result, 570 genes were found to be unique to LJ1, and 612 genes were unique to LJ2 (Supplementary Table 3). Among these, 107 LJ1 genes and 165 LJ2 genes were predicted to encode virulence factors, accounting for 18.8 and 27.0% of all unique genes in the two strains, respectively (Supplementary Table 3). Furthermore, we examined the presence of genes encoding major types of secretion systems and CWDEs in Pectobacterium, and the results were as follows.