Comparative genomic analysis of Herbaspirillum strains associated with banana sheath rot reveals potential virulence factors

Abstract

Background

Banana (Musa spp.) is a critical global fruit crop whose production is severely threatened by sheath rot disease. Klebsiella variicola has been identified as one of the main causative agents of this disease, while the pathogenicity of Herbaspirillum spp. strains isolated from sheath rot-infected banana tissues remains to be clearly defined.

Results

We isolated five Herbaspirillum strains from infected banana sheaths, namely Herbaspirillum huttiense Musa1 (hereafter abbreviated as Her1), Herbaspirillum huttiense Musa2 (hereafter Her2), Herbaspirillum sp. Musa3 (hereafter Her3), Herbaspirillum huttiense Musa4 (hereafter Her4), and Herbaspirillum huttiense Musa5 (hereafter Her5), and performed whole-genome sequencing on each isolate. Comparative genomics revealed genomes of similar size and stable GC content across strains. Functional annotation uncovered a core set of carbohydrate-active enzymes (CAZymes) and antibiotic resistance genes (ARGs), indicating conserved metabolic capabilities. Notably, two pectate lyase genes, potentially involved in plant cell wall degradation, were uniquely identified in strain Her3. We observed marked variation in the repertoire of putative virulence determinants across the isolates: strains Her1 and Her4 each encoded 17 predicted type III secretion system (T3SS) effector proteins, a notably higher number than that of the other strains. All strains also harbored a substantial number of predicted type VI secretion system (T6SS) effectors.

Conclusion

This study provides the first genomic resource for Herbaspirillum species associated with banana sheath rot. Our comparative analysis highlights key genomic differences, particularly in secreted effector profiles, that likely underlie strain-specific pathogenic mechanisms. These results enhance our fundamental understanding of the phytopathogenicity of Herbaspirillum in plants.

Supplementary Information

The online version contains supplementary material available at https://doi.org/10.1186/s12864-026-13097-9.

Background

Banana (Musa spp.) is an important economic crop in tropical and subtropical regions of China [1], primarily distributed in Guangdong, Hainan, Guangxi, Yunnan, Fujian, and Taiwan provinces. Among them, Guangdong, Hainan and Guangxi, which fall within the official geographical division of South China, constitute one of the core banana-producing areas in the country [2]. In recent years, a new disease has emerged in banana plantations in this area, characterized by initial yellowing of the petiole, followed by rot, swelling with exudate, and ultimately leading to petiole breakage and leaf wilting. Notably, this disease occurs year-round without distinct seasonal patterns and can infect all cultivated varieties, including plantain, horn plantain, banana, Formosana banana, Brazil banana, and Fenza, causing severe bacterial sheath rot. It has emerged as a major threat to banana production in South China, second only to Fusarium wilt of banana. Yunhao Sun et al. confirmed that Klebsiella variicola is one of the main causative agents of banana sheath rot in South China [3]. However, the pathogenicity of Herbaspirillum spp. strains isolated from sheath rot-infected banana tissues has not been clearly defined, and the genomic basis underlying their potential interaction with banana hosts remains largely uncharacterized.

Herbaspirillum spp. are Gram-negative, rod-shaped bacteria belonging to the family Oxalobacteraceae within the class Betaproteobacteria. Since its first description by Baldani et al. in 1986 [4], the genus has expanded to include 13 recognized species. Most Herbaspirillum species are endophytic colonizers of plants, widely distributed in grasses such as maize, wheat, and sugarcane, as well as in certain dicotyledons including banana and pineapple. Some strains have also been isolated from contaminated soils, well water, and groundwater, and have exhibited plant growth-promoting traits such as nitrogen fixation [5, 6]. In recent years, this bacterial genus has attracted increasing attention due to its multifunctional roles in plant growth promotion, heavy metal remediation, and potential pathogenicity. Notably, several Herbaspirillum species have been identified as plant pathogens. Among them, H. rubrisubalbicans and H. frisingense have been relatively well-studied and are known to act both as endophytes and as causative agents of plant diseases [7, 8]. Studies have shown that the type III secretion system (T3SS) of H. rubrisubalbicans is essential for causing mottled stripe disease in sugarcane and for endophytic colonization in rice [9]. In China, the first report of sugarcane stripe disease caused by this pathogen was documented in 2010 [10]. Another significant species, H. huttiense, was originally isolated from well water in 2004 [11]. More recently, strains closely related to H. huttiense were isolated from Boston fern in Florida, USA, in 2021, where they induced tan leaf spots and mild blight symptoms, thereby expanding the known host range and pathogenic potential of this species [12]. Furthermore, Herbaspirillum species are recognized as opportunistic human pathogens, capable of causing infections in immunocompromised individuals such as those with cystic fibrosis, cancer, or recipients of stem cell transplants [13, 14]. A recent case report highlighted their clinical significance, documenting Herbaspirillum-associated bacteremia in a neonate receiving high-dose corticosteroid therapy [15].

Based on the above findings, we isolated five Herbaspirillum strains (Her1–Her5) from infected banana plants and performed whole-genome sequencing. Using comparative genomics and evolutionary analysis, we focused on investigating the genetic characteristics and evolutionary relationships of host-specific candidate type III effectors (T3Es).

Methods

Strain isolation and pathogenicity tests

In this study, diseased tissues infected with sheath rot were collected from banana plantations in seven cities of Guangdong Province for pathogen isolation. Leaf sheath tissues from the junction between diseased and healthy areas were surface-sterilized and inoculated onto NA medium. After incubation at 30 °C for 24 h, purification was achieved through three successive streak plates. Based on 16 S rDNA sequencing, five strains of Herbaspirillum spp. (Her1–Her5) were isolated from Enping and Yingde cities, among other locations. Single colonies were inoculated into liquid LB medium and cultured at 30 °C with shaking at 180 rpm for 18 h. Bacterial cells were collected by centrifugation and resuspended to 10⁷ CFU/mL. Pathogenicity assays were conducted using 60-day-old Brazilian banana plants (Musa nana Lour). Inoculation was performed via leaf puncture (three sites per leaf, 100 µL per site) and sheath needle injection (2 mm depth, 100 µL per site), with sterile water used as a negative control. Symptoms were recorded and images were captured on the 7th day after inoculation. The experiment included three biological replicates per treatment in a randomized layout. Data were analyzed by one-way ANOVA with Duncan’s multiple range test (p < 0.05) using SPSS Statistics 27.0.1.

DNA extraction, genome sequencing and assembly

Genomic DNA was extracted from five Herbaspirillum strains using the HiPure Bacterial DNA Kit (Magen Bio, Guangzhou, China). DNA concentration and purity were assessed using a Qubit fluorometer and a NanoDrop spectrophotometer, respectively. DNA integrity was verified via 1% agarose gel electrophoresis to confirm all samples met the requirements for library construction. We employed a hybrid sequencing strategy combining Oxford Nanopore Technologies (ONT) long-read and Illumina short-read sequencing for this study. ONT data were used for de novo genome assembly, while Illumina data were applied for single-base level polishing to generate complete, high-accuracy genome sequences. For ONT sequencing, genomic DNA was fragmented to an average size of ~ 10 kb using a g-Tube (Covaris, Woburn, MA, USA). Sequencing libraries were constructed using the SQK-LSK109 Ligation Sequencing Kit (Oxford Nanopore Technologies, Oxford, UK) following the manufacturer’s standard protocol. After library quality validation, sequencing was performed on the ONT PromethION platform using R9.4.1 flow cells. Basecalling of raw electrical signal data was performed using Dorado v0.5.3 [16] (Oxford Nanopore Technologies, Oxford, UK) with the super accuracy (sup) model (model version: dna_r9.4.1_e8_sup@v3.6). The resulting raw reads were subjected to a three-step quality control (QC) process, including removal of reads with an average Phred quality score (Q-score) < 7, retention of valid long fragments between 200 bp and 50 kb for downstream assembly, and trimming of sequencing adapters using Porechop v0.2.4 [17]. High-quality clean reads obtained after these steps were used for subsequent genome assembly.

For Illumina sequencing, paired-end libraries with an insert size of 300 ~ 400 bp were constructed from randomly sheared genomic DNA. Validated libraries were sequenced on the Illumina platform to generate raw paired-end reads. Adapter trimming and strict QC of raw reads were performed using Fastp v0.23.2 [18], with defined filtering parameters: reads shorter than 50 bp after adapter trimming were discarded, reads consisting entirely of adenine (A) bases were removed, reads with an ambiguous base (N) proportion > 10% were filtered out, and low-quality reads where over 50% of bases had a Phred Q-score ≤ 20 were discarded. The final adapter-trimmed ONT reads were subjected to de novo assembly using Flye v2.9 [19] with parameters --asm-coverage 50, --genome-size 5.6 M, --iterations 3. Genome circularization was automatically determined by the built-in algorithm of Flye. The draft assemblies were subsequently polished using Pilon v1.23 [20] with high-quality Illumina clean reads following a detailed workflow. First, Illumina clean reads were mapped to the draft genome assemblies using BWA-MEM v0.7.17 [21] with default parameters. Next, the resulting sequence alignment map (SAM) files were converted to sorted binary alignment map (BAM) files, followed by duplicate removal and index construction using Samtools v1.15 [22]. Finally, Pilon v1.23 [20] was run with default parameters using the sorted BAM files and draft genome assemblies as input to complete single-base error correction, insertion and deletion (InDel) fixing, and misassembly repair. Ultimately, complete, circularized chromosome sequences for each strain were successfully obtained via this hybrid sequencing approach, and all genome sequences have been deposited in the GenBank database under the following accession numbers: JBSXAD000000000 (Her1), JBSWZC000000000 (Her2), JBSXAE000000000 (Her3), JBSPTD000000000 (Her4), and JBSXAF000000000 (Her5).

Functional annotation and systems analysis

We predicted open reading frames (ORFs) using both the NCBI Prokaryotic Genome Annotation Pipeline and Prokka v1.11 [23, 24]. Non-coding RNAs were identified using tRNAscan-SE v1.3.1 [25], RNAmmer v1.2 [26], and cmscan v1.1.2 (small non-coding RNAs) [27]. Screening for mobile genetic elements and repetitive sequences employed CRISPRfinder v4.2.17 for CRISPR arrays [28], IslandPath-DIMOB v1.0.0 for genomic islands [29], Phage_Finder v2.0 for prophages [30], TransposonPSI 20,100,822 for transposons [31], and Tandem Repeats Finder (TRF) v4.09.1 [32] with RepeatMasker v4.0.5 [33]for tandem and dispersed repeats. Functional annotation was performed by querying protein sequences against the NR [34], Swiss-Pro^t[35], GO [36], COG [37], and KEGG databases [38], with redundancy reduction of GO terms performed using rrvgo v1.14.0 [39], and additional context provided by Pfam [40], PHI-base [41], CARD [42], and CAZy [43]. Protein feature analysis included prediction of signal peptides by SignalP v4.1 [44], transmembrane helices by TMHMM v2.0 [45], and putative type III secreted effectors by EffectiveT3 v1.0.1 [46]. Secondary metabolite biosynthetic gene clusters were identified using antiSMASH v4.1.0 [47]. Genomic islands were visualized with IslandViewer 4 [48]. All analyses were conducted with default parameters unless otherwise specified.

Identification of orthologous genes

Whole-genome alignment of the five sequenced Herbaspirillum strains in this study with 16 qualified genus reference genomes (retrieved from NCBI GenBank as of November 11, 2024; inclusion criteria: complete genome assembly, clear species-level taxonomic assignment, ≥ 95% completeness and ≤ 5% contamination verified by CheckM [49]) was performed using MUMmer v3.3.3 [50]. Subsequently, DIAMOND v2.0.11 [51] was used for the alignment of all reference strains’ protein sequences (E-value < 1e⁻⁵, query coverage > 30%). OrthoMCL v2.0 [52] was then employed to identify orthologous gene clusters based on the alignment results. For phylogenetic analysis, genomes closely related to the target strains were prioritized for gene presence-absence analysis. Whole-genome average nucleotide identity (ANI) was calculated using pyani v0.2.0 [53] in ANIb mode to evaluate genomic similarity. Pairwise ANI values were summarized and compared using CompareM v0.1.2 [54].

Core genome phylogenetic analysis

Whole-genome-based core genome phylogenetic analysis was conducted using the 5 sequenced Herbaspirillum strains and 16 qualified Herbaspirillum reference genomes (retrieved from NCBI GenBank as of November 11, 2024; inclusion criteria: complete genome assembly, clear species-level taxonomic assignment, ≥ 95% genome completeness and ≤ 5% contamination verified by CheckM [49]). Core single-copy orthologs were identified with Roary v3.13.0 [55], aligned with MUSCLE [56, 57], trimmed with Gblocks [58], and used to construct a maximum likelihood phylogenetic tree with 1000 bootstrap replicates via IQ-TREE v2.2.0 [59].

Results

Herbaspirillum spp. are the pathogen of banana sheath rot

In November 2023, an epidemiological survey of banana sheath rot was conducted in major banana-growing regions of Guangdong Province, China, including Enping (22°15′27″N, 112°24′41″E) and Yingde (24°08′39″N, 112°54′52″E). Five bacterial isolates exhibiting pale yellow, translucent colonies were obtained from diseased leaf sheath tissues of the banana cultivars ‘Brazilian’ and ‘Fenza No. 1’ (Fig. 1A). Among them, strains Her1, Her4, and Her5 were isolated from ‘Fenza No. 1’, while Her2 and Her3 were isolated from ‘Brazilian’. We first performed 16 S rRNA-based phylogenetic analysis for preliminary genus identification of the five isolates, which confirmed all strains belonged to the genus Herbaspirillum; the corresponding phylogenetic tree is provided in Figure S1. To establish a more robust evolutionary framework, we further conducted phylogenetic analysis of the five Herbaspirillum isolates and 16 qualified reference strains of the genus based on whole-genome core single-copy genes. The resulting core genome phylogenetic tree, with significantly improved bootstrap support for all branch nodes, is presented in Fig. 1B.

Fig. 1.

Phylogenetic tree of Herbaspirillum strains constructed based on whole-genome sequences. A Plate colony morphology of five Herbaspirillum isolates. Single colonies were streaked onto LB medium and incubated at 30 °C for 36 h.; B Phylogenetic tree constructed based on the whole genomes of 16 representative Herbaspirillum strains and the five target strains

Pathogenicity tests confirmed that the strains induced typical sheath rot symptoms on the ‘Fenza No. 1’ cultivar (Fig. 2), with quantitative lesion area measurements provided in Fig. S2. Bacterial isolates reisolated from the lesion margins of inoculated plants were confirmed to be identical to the original strains via colony morphology observation and 16 S rRNA gene sequencing alignment (Fig. S3-S7), fulfilling Koch’s postulates.

Fig. 2.

Symptoms observed on ‘Fenza No. 1’ banana plants 7 days post-inoculation. Panels A to F represent disease symptoms 7 days after inoculation with sterile water (CK) and strains Her1 to Her5, respectively. Bacterial suspensions were cultured for 18 h and applied via needle puncture inoculation (200 µL per plant). Bar = 10.0 cm.Panels G to L show disease symptoms 7 days after inoculation with sterile water (CK) and strains Her1 to Her5, respectively, using a leaf punch inoculation method (200 µL of bacterial suspension cultured for 18 h). Bar = 1.0 cm

Genome sequencing, assembly, and functional annotation

To characterize the genomic features of the five Herbaspirillum strains, we performed hybrid whole-genome sequencing using Oxford Nanopore Technologies (ONT) long-read and Illumina short-read platforms. We obtained high-quality reads after quality filtering. Clean data were assembled into a single scaffold using Flye, indicating that each Herbaspirillum strain possesses one large chromosome. The chromosome sizes of Her1 to Her5 were approximately 5.82 Mb, 5.60 Mb, 5.54 Mb, 5.70 Mb, and 5.70 Mb, respectively (Fig. 3). The chromosomes contained 5,182, 4,931, 4,909, 4,981, and 4,928 predicted genes, respectively, with GC contents of 62.31%, 62.39%, 62.56%, 62.22%, and 62.39%. Various prediction strategies were employed to identify clustered regularly interspaced short palindromic repeats (CRISPR), genomic islands (GIs), prophages, interspersed repeats, and tandem repeats (Table 1). Species distribution statistics based on the Non-Redundant Protein Database (NR) showed that over 91% of the genes in all five strains belonged to the genus Herbaspirillum, indicating high quality of the assembled genomes (Fig. S8).

Fig. 3.

Whole-genome circular maps of the five Herbaspirillum strains. A Circular map of the chromosomes from Herbaspirillum strain Her1. The complete contig length is 5,824,446 bp, with a GC content of 62.31% and 5,182 predicted genes. B Circular map of the chromosomes from Herbaspirillum strain Her2. The complete contig length is 5,607,265 bp, with a GC content of 62.39% and 4,931 predicted genes. C Circular map of the chromosomes from Herbaspirillum strain Her3. The complete contig length is 5,544,868 bp, with a GC content of 62.56% and 4,909 predicted genes. D Circular map of the chromosomes from Herbaspirillum strain Her4. The complete contig length is 5,702,389 bp, with a GC content of 62.22% and 4,981 predicted genes. E Circular map of the chromosomes from Herbaspirillum strain Her5. The complete contig length is 5,607,230 bp, with a GC content of 62.39% and 4,928 predicted genes. The outermost ring of each genomic map indicates the sequence coordinates. From outer to inner, the rings represent: genes on the forward strand, genes on the reverse strand, ncRNAs (tRNAs shown in black, rRNAs in red), GC content (red indicates values above the mean, blue below the mean), and GC skew (used to measure the relative abundance of G and C, often applied in circular chromosomes to mark origin and terminus; GC skew = (G - C) / (G + C); purple indicates values greater than 0, orange less than 0)

Table 1.

General genomic features of Herbaspirillum strains Her1–Her5

Feature	Megachromosome
	Her1	Her2	Her3	Her4	Her5
Size (bp)	5,824,446	5,607,265	5,544,868	5,702,389	5,607,230
G + C content (%)	62.31	62.39	62.56	62.22	62.39
Coding genes	5182	4931	4909	4981	4928
tRNA	60	60	58	60	60
23S_rRNA	4	4	3	4	4
16S_rRNA	4	4	3	4	4
5S_rRNA	0	0	0	0	0
sRNA	6	7	6	6	7
CRISPR number	0	0	0	0	0
Genomic islands	11	14	14	9	12
Prophage	4	3	1	1	3
Interspersed repeats	57	54	70	54	54
Tandem repeats	202	188	176	195	187
ARGS	405	383	403	400	384

To functionally characterize the strain-specific genes, Gene Ontology (GO) and Kyoto Encyclopedia of Genes and Genomes (KEGG) enrichment analyses were conducted, revealing distinct functional profiles across the five Herbaspirillum strains. After redundancy reduction using rrvgo, genes unique to strain Her1 were enriched in GO terms including redox state regulation and biotic interaction/defense processes, while KEGG analysis emphasized their roles in Metabolic pathways and ABC transporters. In contrast, Her2-specific genes were primarily associated with core metabolic and biosynthetic processes like “gene expression” and “heterocycle biosynthetic process,” though no KEGG pathways were significantly enriched. For strain Her3, specific genes were implicated in “transposition” and “type VI secretion system-mediated protein secretion” according to GO analysis, and KEGG analysis further underscored their roles in “Two-component system” and “Xylene degradation”. Strain Her4 exhibited enrichment in biological processes including biotic interaction/defense-like processes and endomembrane system organization, alongside KEGG pathways for “Nicotinate and nicotinamide metabolism” and “Bacterial chemotaxis”. Finally, while Her5-specific genes were linked to DNA integration and cellular component organization, no significant KEGG pathway enrichment was detected (Fig. 4).

Fig. 4.

Top 20 non-redundant GO biological process enrichment terms and KEGG pathways of strain-specific genes in the five Herbaspirillum strains. The q-value was determined by multiple-test correction of the p-value. A and B Top 20 non-redundant GO terms and KEGG pathways of strain-specific genes in Her1, respectively; (C) Top 20 non-redundant GO terms of strain-specific genes in Her2; (D): Top 20 non-redundant GO terms of strain-specific genes in Her5; E and F): Top 20 non-redundant GO terms and KEGG pathways of strain-specific genes in Her3, respectively; G and H: Top 20 non-redundant GO terms and KEGG pathways of strain-specific genes in Her4, respectively. Non-redundant GO terms were reduced using the rrvgo package (version 1.14.2) with the method="Rel” and threshold = 0.7 based on semantic similarity measures

Genetic relationships between five Herbaspirillum strains and other pathogenic strains

In this study, we explored the unique and shared genes among these strains. We found that strain Her1 has 273 unique genes, Her2 has 7, Her3 has 233, Her4 has 64, and Her5 has 5 (Fig. 5). Using MUMmer software, we aligned the target genomes with reference genomes and identified syntenic blocks between each of the five Herbaspirillum strains and the sixteen Herbaspirillum strains of different species. The proportion of each syntenic block relative to the total gene length is shown in Table 2. Whole-genome synteny comparisons were performed, and pairwise average nucleotide identity (ANI) values between the sequenced strains and closely related reference genomes were calculated using pyani. Average amino acid identity (AAI) was also determined as the mean identity across all pairwise orthologous proteins shared between genomes (Table S1). Among the Herbaspirillum strains examined, ANI values between Her1 and Her4, as well as between Her2 and Her5, exceeded 99%. All four strains also showed ANI values above 97.5% relative to the type strains H. huttiense ZXN111 and NFYY, well above the widely accepted threshold of 95% for species circumscription [60]. Core genome phylogenetic analysis based on single-copy orthologous genes further confirmed that these four strains formed a strongly supported monophyletic clade within H. huttiense.

Fig. 5.

Genetic relationships between the five Herbaspirillum strains and other pathogenic Herbaspirillum strains. Note: Venn diagram showing shared and unique gene families among the five target strains and the 16 representative Herbaspirillum strains

Table 2.

Percentage of syntenic blocks relative to the total gene length

Strain	Collinear (%)
	Her1	Her2	Her3	Her4	Her5
Herbaspirillum rubrisubalbicans_M1	75.65	73.71	74.28	75.95	73.71
Herbaspirillum rubrisubalbicans _DSM_11543	77.13	73.89	73.51	75.78	73.89
Herbaspirillum rubrisubalbicans_Os34	70.89	69.58	70.40	71.15	69.58
Herbaspirillum huttiense_ZXN111	86.04	83.77	76.94	87.06	83.77
Herbaspirillum huttiense_NFYY	86.59	85.37	76.95	86.50	85.37
Herbaspirillum frisingense_IAC152	73.60	73.46	72.14	73.00	73.47
Herbaspirillum frisingense_AU14559	74.81	75.24	73.82	74.90	75.24
Herbaspirillum robiniae_AA6	61.40	61.40	60.64	61.89	61.72
Herbaspirillum seropedicae_Z67	72.12	72.07	72.19	72.07	72.07
Herbaspirillum seropedicae_AU14040	72.89	72.53	72.11	72.79	72.53
Herbaspirillum seropedicae_AU13965	73.71	72.81	72.14	73.24	72.81
Herbaspirillum seropedicae_SmR1	72.15	72.09	72.18	72.09	72.09
Herbaspirillum camelliae_WT00C	65.49	64.77	63.36	64.98	64.77

All four strains recovered from banana sheath rot in this study were taxonomically assigned to the species H. huttiense based on standard genomic criteria. In contrast, strain Her3 exhibited a maximum ANI of only 88% relative to the 16 reference genomes of Herbaspirillum, markedly below the 95% threshold for species definition, indicating that Her3 likely represents a distinct novel species within the genus. Accordingly, these four strains were taxonomically assigned to Herbaspirillum huttiense, with their formal designations as follows: Herbaspirillum huttiense Musa1 (abbreviated as Her1), Herbaspirillum huttiense Musa2 (Her2), Herbaspirillum huttiense Musa4 (Her4), and Herbaspirillum huttiense Musa5 (Her5) (Fig. 6).

Fig. 6.

ANI heatmap of the five Herbaspirillum strains and other pathogenic Herbaspirillum strains

Identification and comparative analysis of T3SS effector proteins in five Herbaspirillum strains

Herbaspirillum bacteria deliver various effector proteins through their T3SS, which play a critical role in determining host range and pathogenicity [62]. To characterize the effector profiles of the five Herbaspirillum strains (Her1–Her5), we first identified genes associated with type I–VII secretion systems (T1SS–T7SS) based on functional annotations against the KEGG and NR protein databases, and predicted T3SS effector proteins in these Gram-negative bacteria using EffectiveT3. The analysis revealed marked differences in the number of effector proteins among the five strains: 51 effector proteins were predicted in both Her1 and Her4, 30 each in Her2 and Her5, and 36 in Her3. Focusing specifically on T3Es, the predicted counts were 17 (Her1), 2 (Her2), 3 (Her3), 17 (Her4), and 2 (Her5). Among them, Her1 and Her4 each contain two Hrp gene clusters. Notably, T6Es accounted for a significant proportion of the effector repertoire in these strains, with counts of 25 (Her1), 22 (Her2), 26 (Her3), 25 (Her4), and 22 (Her5). These results show that Her1 and Her4 share highly similar effector protein profiles, Her2 and Her5 also show comparable patterns, while Her3 displays a distinct effector profile (Fig. 7). The five Herbaspirillum strains harbor both T3SS and abundant T6SS effectors. As coordinated functions of T3SS and T6SS in host interaction have been demonstrated in plant-associated bacteria [63, 64], we suggest that these systems may act together to mediate interactions with host plants.

Fig. 7.

Identification and Comparison of Type III Effector Genes in Five Herbaspirillum Strains. Rectangles of different colors indicate the presence status of effector protein genes in the corresponding strains. The meaning of each color is shown in the legend at the top right

Genomic islands and prophage elements

Genomic islands (GIs) are DNA segments transferred between bacterial genomes through horizontal gene transfer (HGT). In this study, a systematic screening of genomic islands in the chromosome sequences of five Herbaspirillum strains (Her1–Her5) was performed using the IslandViewer 4 software platform. As shown in Fig. 8, prediction results based on the IslandPath-DIMOB algorithm identified 11, 14, 14, 9, and 12 genomic islands in the chromosomes of Her1, Her2, Her3, Her4, and Her5, respectively. The characteristics of the genomic islands in each strain are as follows: Her1: GI length ranged from 5,732 to 43,159 bp, containing 247 coding genes (Fig. 8A, a); Her2: 3,527–59,142 bp (255 genes) (Fig. 8B, b);Her3: 7,754–65,835 bp (321 genes) (Fig. 8C, c); Her4: 10,519–66,426 bp (229 genes) (Fig. 8D, d); Her5: 10,003–60,073 bp (254 genes) (Fig. 8E, e). These genomic islands primarily encode functions related to genetic information processing (DNA replication and repair, regulation of gene expression); protein biosynthesis and modification; metabolic pathways (amino acid, vitamin, and ion metabolism); cell structure maintenance and substance transport systems. Notably, several GIs encode transposases (IS3 and IS256 families) and recombinases (e.g., tyrosine recombinase XerC), suggesting their potential involvement in genomic dynamics and evolution. Additionally, various functionally characteristic proteins were identified, including DNA primase, transcriptional regulator DecR, protease HtpX, putative signal proteins, threonine efflux proteins, and chemotaxis protein CheY. The functions of some proteins, such as the 50 S ribosomal protein L21 and several hypothetical proteins, require further validation.

Fig. 8.

Genomic Island Circular and Linear Diagrams of Five Herbaspirillum Strains. A Circular map of genomic island distribution in the chromosomes of Herbaspirillum strain Her1; a: shows the linear map; (B) Circular map of genomic island distribution in the chromosomes of Herbaspirillum strain Her2; b: shows the linear map; (C) Circular map of genomic island distribution in the chromosomes of Herbaspirillum strain Her3; c: shows the linear map; (D) Circular map of genomic island distribution in the chromosomes of Herbaspirillum strain Her4; d: shows the linear map; (E) Circular map of genomic island distribution in the chromosomes of Herbaspirillum strain Her5; e: shows the linear map. Prediction results from different tools are represented by colored bars: orange indicates genomic islands predicted by SIGI-HMM, green represents results from IslandPick, and blue denotes analysis results from IslandPath-DIMOB. Genomic island regions confirmed through integrated analysis are marked with red boxes

Analysis of carbohydrate-active enzymes and antibiotic resistance

CAZymes are a group of multifunctional enzymes that catalyze the formation, breakdown, and modification of glycosidic bonds, thereby participating in carbohydrate metabolism [65]. To investigate the CAZymes involved in carbohydrate synthesis and degradation in the five Herbaspirillum strains and to elucidate their carbon source utilization preferences and ecological functions (such as plant interactions and biocontrol potential), we performed genomic annotation. The results revealed a highly conserved distribution of CAZyme gene families: Her1 contained 795 genes; Her2 and Her5 each contained 798; Her3 and Her4 contained 797 and 792, respectively. Her3 uniquely harbored two Polysaccharide Lyase (PL) genes. Glycoside Hydrolases (GHs) and Glycosyl Transferases (GTs) were the most abundant families, followed by Carbohydrate-Binding Modules (CBMs), Carbohydrate Esterases (CEs), and Auxiliary Activities (AAs) (Table 3).

Table 3.

Numbers of carbohydrate-active enzymes (CAZymes) identified in Herbaspirillum strains Her1–Her5

CAZymes	Her1	Her2	Her3	Her4		Her5
Auxiliary Activities	12	11	12	10	11
Carbohydrate-Binding-Module	73	73	80	74	73
Carbohydrate Esterases	43	46	42	44	46
Glycoside Hydrolases	260	255	243	256	255
Glycosyl Transferases	407	413	418	408	413
Polysaccharide Lases	0	0	2	0	0
Total	795	798	797	792	798

Antibiotic resistance genes (ARGs) were annotated using the Resistance Gene Identifier (RGI) by aligning predicted gene sequences against the Comprehensive Antibiotic Resistance Database (CARD), which uses the Antibiotic Resistance Ontology (ARO) for classification [66]. Analysis identified 383–405 ARGs in the five strains, associated with resistance to a wide range of compounds, including rifamycins, tetracyclines, beta-lactams, fluoroquinolones, glycopeptides, and macrolides, among others (Table 1).

Discussion

In this study, we isolated five Herbaspirillum strains (Her1–Her5) from banana leaf sheaths affected by sheath rot in major banana-growing regions of Guangdong Province. Pathogenicity tests confirmed that all five strains could induce typical sheath rot symptoms on banana plants, and fulfilled Koch’s postulates, confirming that Herbaspirillum spp. are causative agents of banana sheath rot. These findings expand the known pathogenic spectrum of banana sheath rot, and provide new etiological insights into this destructive disease.

In 2011, the complete genome sequence of Herbaspirillum seropedicae strain SmR1 was first reported [67]. We performed high-quality complete genome sequencing of the five pathogenic Herbaspirillum strains using a hybrid ONT and Illumina sequencing strategy, generating circularized chromosome sequences with high completeness and accuracy. The five strains showed similar genome size (5.54–5.82 Mb) and stable GC content (62.22%–62.56%), with over 91% of genes annotated to the genus Herbaspirillum, providing a high-quality foundational genomic resource for subsequent comparative analysis.

GO and KEGG enrichment analyses of strain-specific genes suggested distinct functional profiles across the five Herbaspirillum strains. Genes unique to strain Her1 were enriched in redox state regulation and biotic interaction/defense-related GO terms, with KEGG analysis emphasizing their roles in metabolic pathways and ABC transporters, suggestive of broad metabolic potential and environmental adaptability. Strain Her3 showed features in transposition and type VI secretion system-mediated protein secretion (GO terms), as well as two-component system and xylene degradation (KEGG pathways), suggesting that it may possess enhanced adaptive potential to complex plant-associated niches. Relative to the other four strains in this study, strain Her4 exhibited distinct functional signatures associated with host interaction and environmental sensing, supported by enrichment in biotic interaction/defense-like processes, endomembrane system organization, and KEGG pathways of nicotinate and nicotinamide metabolism and bacterial chemotaxis. No significantly enriched KEGG pathways were detected for strains Her2 and Her5, with their strain-specific genes primarily associated with basal metabolic and biosynthetic processes.

Comparative genomic analysis was conducted between the five isolated strains and sixteen reference Herbaspirillum strains retrieved from NCBI. To clarify the taxonomic status of the five sequenced strains, we performed core genome phylogenetic analysis combined with ANI and AAI calculation. Four of the strains shared ANI values > 97.5% with the type strains of H. huttiense (ZXN111 and NFYY), well above the widely accepted ≥ 95% threshold for prokaryotic species circumscription [60], and formed a well-supported monophyletic clade with H. huttiense reference strains in the core genome phylogenetic tree. These findings confirm the taxonomic assignment of the four strains to H. huttiense. Notably, these four banana-derived pathogenic strains formed a distinct subclade within the H. huttiense lineage, with unique genomic and virulence-related features that differentiated them from non-plant-pathogenic H. huttiense reference strains. This finding is consistent with the recent report by Poudel et al. (2024) [61], which identified plant-pathogenic lineages within H. huttiense for the first time, and our results further expand the known plant-pathogenic diversity of H. huttiense in banana hosts.

Notably, over the past two decades, research on Herbaspirillum has increasingly focused on its role as a plant pathogen. Using the T3E database [68], we analyzed the T3Es in the five strains. Her1 and Her4, as well as Her2 and Her5, shared similar T3E compositions, while Her3 exhibited a distinct effector profile. Although some strains showed comparable T3E patterns, whole-genome ANI and phylogenetic analyses confirmed that all five strains (Her1–Her5) were genetically distinct independent isolates without redundancy. Notably, multiple T3Es—such as hrpW, hrpB3, lcrD, sctC, sctN, among others—were predicted in Her1 and Her4. Among these, HrpW contains a C-terminal pectate lyase domain belonging to the polysaccharide lyase family 3 (PL3). This enzyme is Ca²⁺-dependent and degrades highly esterified pectin via a β-elimination mechanism, thereby disrupting plant cell wall integrity and facilitating bacterial invasion [69].

GIs are key mobile genetic elements associated with HGT in bacteria, which drive genomic evolution and adaptive divergence. In this study, we identified 9–14 genomic islands across the five strains, with Her3 harboring the highest number of GI-associated coding genes. These GIs mainly encoded functions related to genetic information processing, metabolism, substance transport and genomic mobility, including transposases, recombinases and chemotaxis-related proteins. These genomic features suggest that GI-associated HGT events may have contributed to the genomic differentiation and functional diversification among these banana-derived Herbaspirillum strains.

In addition, CAZyme family analysis showed that their distribution was highly conserved across the five strains. Notably, two genes in Her3 were annotated as polysaccharide lyases, suggesting a potential capacity for pectin degradation. Pectinases are known key virulence factors in many plant pathogens. For example, Fusarium oxysporum f. sp. cubense secretes exo-polygalacturonases (e.g., PGC3) that degrade pectin, directly leading to tissue maceration and cell necrosis in banana [70]. Therefore, further experimental validation is required to confirm the role of these two PL genes in the pathogenicity of Her3. In addition, we identified 383–405 antibiotic resistance genes (ARGs) in the five strains, associated with resistance to multiple classes of antibiotics including rifamycins, tetracyclines and beta-lactams. These ARGs may contribute to the survival of these strains in complex agricultural environments.

In previous studies, comparative analysis of microbial interaction networks between sheath tissues of sheath rot-infected and healthy banana plants revealed a potential correlation between Herbaspirillum spp. and the colonization of Klebsiella variicola [3]. However, the present study did not conduct targeted functional validation of the predicted virulence factors. Further systematic studies are needed to clarify the colonization dynamics, infection process and pathogenic contribution of these Herbaspirillum strains in banana sheath rot, which will help to fully reveal the pathogenic mechanism of this disease.

This study lays a foundation for subsequent functional investigations of Herbaspirillum. The whole-genome sequences of the five banana-derived pathogenic Herbaspirillum strains generated expand the genomic resources for pathogenic lineages within the genus Herbaspirillum, and provide data support for the taxonomic classification and dissection of pathogenicity evolution of this genus.

Conclusion

In this study, we isolated five Herbaspirillum strains from banana sheath rot lesions, confirmed their pathogenicity via Koch’s postulates, and completed whole-genome sequencing. Taxonomic analysis assigned four strains (Her1, Her2, Her4, Her5) to H. huttiense, and suggested that Her3 represents a candidate novel species within the Herbaspirillum genus. Comparative genomics revealed distinct genetic and functional divergence across the five strains, and we further characterized key pathogenicity-related genetic factors including type III effectors, carbohydrate-active enzymes and antibiotic resistance genes. These findings confirm that Herbaspirillum species are causal agents of banana sheath rot in South China, expand genomic resources for pathogenic Herbaspirillum lineages, and provide a foundation for follow-up studies.

Abbreviations

CAZymes

Carbohydrate-Active zymes

GHs

Glycoside Hydrolases

GTs

Glycosyltransferases

PLs

Polysaccharide Lyases

CEs

Carbohydrate Esterases

CBMs

Carbohydrate-Binding Modules

AAs

Auxiliary Activities

ARGs

Antibiotic Resistance Genes

CARD

Comprehensive Antibiotic Resistance Database

ARO

Antibiotic Resistance Ontology

RGI

Resistance Gene Identifier

T3SS

Type III Secretion System

T6SS

Type VI Secretion System

T3Es

Type III Secretion Effectors

T6Es

Type VI Secretion Effectors

HGT

Horizontal Gene Transfer

ORFs

Open Reading Frames

ONT

Oxford Nanopore Technologies

GIs

Genomic Islands

GO

Gene Ontology

KEGG

Kyoto Encyclopedia of Genes and Genomes

ANI

Average Nucleotide Identity

TRF

Tandem Repeat Finder

NR

Non-Redundant

Authors’ contributions

Nianjia Li: Conceptualization, Data curation, Writing – Original Draft, Investigation, Supervision, Writing – Review & Editing. Shuyin Luo, Shijie You: Conceptualization, Supervision, Writing – Review & Editing. Yanhui Wu, Sijia Luo: Methodology, Data curation, Writing – Review & Editing. Yulan Ruan: Conceptualization, Writing – Review & Editing. Yuefeng Li, Huiwen Yang, Yutao Huang, Shurui Hu: Writing – Review & Editing. Guohui Yu, Yunhao Sun: Supervision, Writing – Review & Editing.

Funding

This work was supported by the Approved Project of the Special Fund for Scientific and Technological Innovation Strategy of Guangdong Province (for the Cultivation of College Students’ Scientific and Technological Innovation, pdjh2024a201), the grants from the Second-round Key Deployment of Rural Technology Commissioners under the “Hundred, Thousand, Ten-Thousand” Project (No. KTP20240299 and 2025A03011001), Integrated Key Technology for Intensive Vegetable Seedling Cultivation and Demonstration Base (No. 2024A0304009), Research and Demonstration of Key Technologies for Agricultural Microbial Technology to Improve the Production Quality and Efficiency of Organic Vegetables and Fruits in South China (No. 2023B0202010012), Research, Application and Demonstration of Key Simplified Green Prevention and Control Technologies for Banana Fusarium Wilt Based on Rhizosphere Immunity Regulation (No. 202206010083).

Data availability

The complete genome sequences of the five *Herbaspirillum* strains have been deposited in the DDBJ/ENA/GenBank databases under the accession numbers JBSXAD000000000 (Her1), JBSWZC000000000 (Her2), JBSXAE000000000 (Her3), JBSPTD000000000 (Her4), and JBSXAF000000000 (Her5).

Declarations

Ethics approval and consent to participate

Not applicable.

Consent for publication

Not applicable.

Competing interests

The authors declare no competing interests.