The rfbA Gene as a High-Resolution Marker for Pathovar-Level Classification of Xanthomonas citri and Pseudomonas syringae

Article information

Plant Pathol J. 2026;42(2):222-227

Publication date (electronic) : 2026 April 1

doi : https://doi.org/10.5423/PPJ.NT.10.2025.0141

Seung Yeup Lee, Suin Yang, Minju Yun, Eunjung Roh

Crop Protection Division, National Institute of Agricultural Sciences, Rural Development Administration, Wanju 55365, Korea

^*Corresponding author. Phone) +82-63-238-3284, FAX) +82-63-238-3838 E-mail) rosalia51@korea.kr

Handling Editor: Hee Jin Park

Received 2025 October 2; Revised 2026 February 6; Accepted 2026 February 6.

After Antonie van Leeuwenhoek first identified bacteria, Carl Linnaeus first suggested the scientific classification of living organisms in 1735, which propelled the evolution of bacterial taxonomy (Linnaeus, 1758). Bacteria are classified within the domain Bacteria into multiple phyla (and lower taxa) based on phenotypic, genetic, and phylogenetic analyses (Cavalier-Smith, 1998; Ruggiero et al., 2015). Classification through 16S rRNA-sequence-based phylogenetic analysis, proposed by Carl Woese and George Fox in 1977 is used as the formal classification method (Woese et al., 1990). However, to precisely classify bacteria exhibiting indistinct resolution at the species level based on 16S rRNA sequences, fingerprinting techniques such as REP−, ERIC−, and BOX-A1R-based repetitive extragenic palindromic–PCR, or methodologies such as multilocus sequence typing (MLST) are employed. The capability to determine taxonomy without the requirement of complex analytical procedures, relying instead on a single gene, would confer significant advantages in terms of efficiency and resource allocation. In this context, we identified a novel marker gene (rfbA) that facilitates a more accurate classification at both the species and subspecies (pathovar) levels than previously utilized marker genes.

Lipopolysaccharide (LPS) is a biomolecule composed of a lipid and a polysaccharide and it is predominantly localized in the outer membrane of gram-negative bacteria. LPS plays an integral role in non-pathogenic activities, including the promotion of surface adhesion and modulation of bacteriophage sensitivity (Saini and Wood, 2020; Yu et al., 2024). However, the significance of LPS extends to pathogenic functions in which it acts as a barrier to protect bacteria from the hostile environment of the host immune system (Bertani and Ruiz, 2018; Ranf, 2016). Additionally, as a pathogen-associated molecular pattern (PAMP), LPS is detected in plants, triggering an immune response (Lien et al., 2000). This phenomenon suggests a strategy for evading host innate immunity through structural modifications, as observed in various Gram-negative bacteria (Matsuura, 2013).

The LPS of pathogenic bacteria and the immune system of host plants are closely associated with disease development. Therefore, we hypothesized that bacteria classified as different pathovars within the same species have evolved different sequences of genes associated with LPS biosynthesis. While investigating the sequence variations of genes related to LPS biosynthesis from the genomic data of Xanthomonas, it was confirmed that the rfbA gene (encoding glucose-1-phosphate thymidylyltransferase) had many variations within the species and had an appropriate gene size (approximately 850 bp) for Sanger sequencing. Several attempts have been made to classify bacterial strains within a serotype using the rfb locus, which is involved in LPS biosynthesis (Fitzgerald et al., 2003; Manning et al., 1995). However, these studies were limited to animal pathogens, and it was predictable that the bacterial serotypes classified by LPS structure would be distinctly grouped by the rfb locus due to the strong correlation between LPS structure and rfb locus. Therefore, in this study, we compared classification using the rfbA gene sequence with existing standard marker genes by constructing a phylogenetic tree of plant pathogens, Xanthomonas species, and Pseudomonas species. These two genera were specifically selected as model systems because they encompass numerous pathovars and subspecies that remain taxonomically ambiguous and are often difficult to differentiate using conventional single-marker genes or even MLST. By focusing on these challenging groups, we sought to demonstrate the superior discriminatory power of the rfbA gene in resolving fine-scale phylogenetic relationships where standard methods reach their resolution limits.

Genomic sequences (16S rRNA sequence, gyrB, rfbA genes) and MLST (atpD, dnaK, efp, and gyrB genes for Xanthomonas species; acn, cts, gapA, and gyrB genes for P. syringae strains) sequences were collected from genome sequences that were completely assembled and registered in the National Center for Biotechnology Information (NCBI) (Słomnicka et al., 2015). Genomic sequences were retrieved using basic local alignment search tool (BLAST) with searches restricted to Xanthomonadaceae (taxid: 32033) or Pseudomonas syringae (taxid: 136849) group, and the corresponding genome accession numbers are listed in Supplementary Tables 1 and 2. Multiple sequence alignments were conducted using MAFFT v7.490 with the “--auto” option, which automatically selects the appropriate alignment algorithm based on the size and characteristics of the dataset (Katoh and Standley, 2013). Maximum likelihood phylogenetic trees were constructed using IQ-TREE v2 (Minh et al., 2020). The best-fit substitution model was determined by ModelFinder with the option “-m MFP+I+G4”. Branch supports were assessed using 1,000 ultrafast bootstrap replicates (-bb 1000) and 1,000 SH-aLRT replicates (-alrt 1000). Visualization of the phylogenetic tree was conducted in Interactive Tree Of Life (iTOL; Letunic and Bork, 2021) with an unrooted option. To assess selective pressure and recombination, Branch-Site Unconstrained Statistical Test for Episodic Diversification (BUSTED) analysis in HyPhy package and the Pairwise Homoplasy Index (Phi) test were performed to calculate the dN/dS ratio and recombination P-values (Bruen et al., 2006; Murrell et al., 2015).

Our first objective was to determine whether rfbA could be used for taxonomic identification of gram-negative bacteria at the species or subspecies level. The phylogenetic tree of Xanthomonas species generated using the 16S rRNA gene sequence classified most strains at the species level but failed to resolve X. citri or X. campestris at the subspecies level (sequence similarity: 100%; Supplementary Fig. 1A). Furthermore, X. arboricola and X. hortorum were poorly differentiated, with some X. arboricola strains clustering with X. campestris. In contrast, the gyrB-based tree demonstrated significantly enhanced taxonomic resolution, providing clear differentiation between species and distinguishing strains effectively even at the subspecies level (Supplementary Fig. 1B). The phylogenetic tree generated by MLST (atpD, dnaK, efp, and gyrB; approximately 2,234 bp) classified the strains similarly to the gyrB-based tree, but the X. campestris strains were divided into more detailed groups by MLST (Fig. 1A, Supplementary Table 3). The phylogenetic tree based on the rfbA gene also classified strains of Xanthomonas species and subspecies with a high level of resolution, similar to the trees based on gyrB and MLST. However, a critical difference was observed in the rfbA-based tree within the pathovar group of X. citri (Fig. 1B, Supplementary Table 4). In the phylogenetic trees, except for the rfbA-based tree, the subspecies strains belonging to X. citri species were clustered together; however, in the rfbA-based tree, citri group (X. citri pv. citri and X. citri pv. punicae; black curly brace in Fig. 1) was clearly separated from glycines group (X. citri pv. glycines, X. citri pv. malvacearum, X. citri pv. phaseoli var. fuscans, and X. citri pv. fuscans; red curly brace in Fig. 1) (sequence similarity: approximately 80%). Based on these characteristics of the rfbA gene, we propose a new classification system for glycines group of X. citri using rfbA as the marker gene.

Fig. 1

Phylogenetic trees of Xanthomonas species based on (A) MLST (comprising partial sequences of atpD, dnaK, efp, gyrB genes) and (B) the rfbA gene. Maximum Likelihood based phylogenetic trees were generated, and they were visualized as unrooted tree in interactive tree of life (iTOL). A total of 489 strains were analyzed. Bootstrap support values (70–100%) are displayed at the nodes. The tree is unrooted, and topological relationships should not be interpreted as directional. The color of each node’s name represents taxon information, which is shown in the legend. The curly brace indicates pathovars belonging to X. citri, and the color of the brace corresponds to their host plants (displayed on the right side of the panel).

Xanthomonas has been reclassified several times, from the time when bacteria were classified based on their various phenotypic or pathological characteristics to the time when a classification system was established based on the identification of genetic characteristics. Changes in the classification of Xanthomonas species often occur with the development of new classification methods (Bansal et al., 2023; Rademaker et al., 2005; Rodriguez-R et al., 2012; Vauterin et al., 1995; Young et al., 1978). For example, X. citri pv. glycines, X. citri pv. malvacearum, and X. citri pv. fuscans were initially reported to belong to X. campestris but are now classified as a subspecies (pathovar) of X. citri (Schaad et al., 2005; Vauterin et al., 1995). After a classification system using the sequences of several marker genes (including 16S rRNAsequence and gyrB) was standardized, X. citri pv. glycines, X. citri pv. malvacearum, and X. citri pv. fuscans were clustered into X. citri because these strains were the closest to X. citri, as shown in Fig. 1A. However, in the present study, analysis of the rfbA gene sequence showed that X. citri pv. glycines, malvacearum, and fuscans were located further away from X. citri pv. citri. This is thought to be because, unlike other marker genes, the rfbA gene is subject to pressure from the immune system of the host plant. The rfbA gene is associated with LPS biosynthesis, and it is predicted that this difference in the phylogenetic tree is because of mutations in LPS-related genes, such as the rfbA gene, enabling the adaptation of each strain to the specific immune system of the host plant. For Xanthomonas, the Phi test indicated significant recombination (P < 0.01), and BUSTED analysis revealed significant episodic positive selection (P = 0.0002). This significant evidence of recombination and episodic positive selection in rfbA suggests that this gene is not merely a neutral phylogenetic marker but a functional locus undergoing adaptive evolution. Specifically, the high recombination rate in Xanthomonas and the selective pressure observed in both genera indicate that rfbA sequences are likely shaped by host-pathogen interactions and immune evasion. The rfbA gene encodes a key enzyme in the biosynthesis of dTDP-L-rhamnose, a crucial precursor for LPS O-antigens. Our evolutionary analysis supports the functional significance of this locus; the significant positive selection and recombination observed in our BUSTED and Phi test results indicate that rfbA undergoes rapid adaptive evolution. This evolutionary dynamism explains why rfbA provides superior resolution at the pathovar level compared to conserved housekeeping genes or 16S rRNA, which primarily reflect vertical descent and lack the sensitivity to capture recent ecological niche adaptations.

Phylogenetic trees were also constructed for the subspecies (pathovar) of Pseudomonas syringae using 16S rRNA sequence, gyrB, MLST (acn, cts, gapA, and gyrB; approximately 2,957 bp), and rfbA sequences (Fig. 2 and Supplementary Fig. 2). The phylogenetic tree constructed using the 16S rRNA sequences showed that many bacteria were mixed and could not be properly clustered; therefore, the 16S rRNA sequence was not suitable for classifying the pathovars of Pseudomonas syringae (Supplementary Fig. 2). In contrast, phylogenetic trees constructed based on gyrB and MLST analyses yielded similar results, effectively clustering most bacteria by pathovar. These markers provided a substantially higher degree of resolution than the 16S rRNA sequence, which was found unsuitable for classifying the pathovars of P. syringae due to poor clustering (Supplementary Fig. 2). The phylogenetic tree based on rfbA clustered most of the pathovars, similar to gyrB- and MLST-based trees, but showed distinct differences between them at some points. The rfbA-based tree was classified into an upper cluster (Group 1; Herbaceous plants) and a lower cluster (Group 2; Woody plants) (Fig. 2). When the host plants infected by the strains in each group were investigated, it was confirmed that the strains in Group 1 infected various herbaceous, whereas the hosts of the strains in Group 2 were more concentrated in woody plants, such as horse chestnut and kiwi tree. Although we did not find evidence that the immune systems of the host plants differed according to groups in their responses to Pseudomonas syringae, it will be of great significance to further study the relationship between LPS and the immune system of plants, because the separation of groups according to pathovars was clear.

Fig. 2

Comparison of phylogenetic clustering of Pseudomonas syringae strains based on (A) MLST (consisted of partial sequences of acn, cts, gapA, gyrB gene) and (B) rfbA gene. Maximum likelihood based phylogenetic trees were generated, and they were visualized as unrooted tree in interactive tree of life (iTOL). The color of each node’s name represents taxon information, which is shown in the legend. A total of 88 strains were analyzed. Strains associated with the same host are consistently labeled to demonstrate the transition from dispersal in MLST to distinct clustering in rfbA. Bootstrap support values (70–100%) are displayed at the nodes. The tree is unrooted, and topological relationships should not be interpreted as directional.

PAMPs are conserved molecular structures found in microorganisms that allow the host plant’s innate immune system to recognize and respond to infections. These patterns are detected by pattern recognition receptors in plants, and their detection triggers immune responses. Examples of PAMPs include bacterial LPS, flagellins, and viral nucleic acids. The rfbA gene is involved in synthesizing the O-antigen component of LPS, and it has been reported that the O-antigen can regulate the effectiveness of immune evasion strategies depending on changes in its structure, thereby affecting the pathogenicity of various bacteria (Lerouge and Vanderleyden, 2002). However, limited research has been conducted on the structural modification of LPS in response to environmental stress (pressure on the innate immunity of host plants) during the interaction between phytopathogens and host plants (Caroff and Novikov, 2019; Lerouge and Vanderleyden, 2002). Through taxonomic classification using the rfbA gene, we confirmed that pressure from the host immune system might have been exerted on the rfbA gene of phytopathogens during adaptation to their hosts. This finding suggests the function of the rfbA gene as a new marker gene that can ensure higher resolution and accuracy in classifying bacteria and as a cornerstone for the study of the co-evolution between pathogenic bacteria and host plants.