Comparative genomics of strains of varying virulence of the gram-positive bacterial pathogen of maize, Clavibacter nebraskensis

Abstract

Background

Clavibacter nebraskensis (Cn) is a Gram-positive bacterial plant pathogen that infects the vascular tissue of maize, causing Goss’s Wilt and Leaf Blight. Detection of highly virulent strains since the mid-2000s suggests that genetic divergence has contributed to the re-emergence of the disease following its effective management and subsequent disappearance in the 1980s. However, few genomic resources exist for this pathogen, and little is known about genomic variation in the species. Our study aimed to characterize genomic diversity across a set of differentially virulent strains isolated from both historic epidemics and recent outbreaks.

Results

We sequenced 17 Cn strains of varying virulence isolated from epidemics spanning their earliest detection to subsequent outbreaks in the mid-2010s across various US states and Manitoba, Canada. Contrary to previous studies, we found no strong population structure associated with geographic origin or year of isolation based on haplotype analysis. We detected five heterogeneous plasmids in strains CDK032, CDK039, CDK044, CDK045, and CDK046, whereas to date, only one plasmid has been sequenced in a single Cn strain. Genomic islands were detected in all strains, and the putative virulence gene celA was found to be encoded in one such island. Secreted CAZymes, hypothesized to be involved in pathogenicity, are well conserved across all Cn genomes, indicating that allelic or expression diversity, and not gene copy number, might be implicated in virulence diversity. Importantly, we identified deletions in a secreted cellulase within the linker peptide region between the carbohydrate-binding domain and the cellulase domain that have not been reported in the literature previously.

Conclusion

Our analyses provide insight into genomic diversity within the species, revealing five novel plasmids and genomic islands harboring CAzymes, including the celA gene. Genes encoding cell wall degrading enzymes are well-conserved across strains of all virulence phenotypes. Deletions within a secreted cellulase not previously reported have been described for future functional analysis.

Supplementary Information

The online version contains supplementary material available at https://doi.org/10.1186/s12864-026-13129-4.

Background

Clavibacter, a genus within the class Actinomycetia (Gram-positive, high G + C content), is responsible for economically important diseases of various crops. Members of this genus can infect both dicotyledonous and monocotyledonous plants, but generally exhibit narrow host ranges as pathogens, causing disease on tomato (C. michiganensis), potato (C. sepodonicus), pepper (C. capsici), alfalfa (C. insidiosus), wheat (C. tessellarius), barley (C. zhangzhiyongii), bean (C. phaseoli), and maize (C. nebraskensis). Additionally, several strains within the genus have been isolated from asymptomatic tomatoes, existing primarily as endophytes, and new non-pathogenic species have been described as a result [1, 2]. Endophytic colonization of Clavibacter spp. has been reported frequently [3], and some of these endophytes cluster phylogenetically with Clavibacter pathogens of alfalfa, pepper, and wheat [4]. C. tessellarius, a wheat pathogen, has been isolated from tomato seeds [3]. Similarly, C. sepedonicus, known to infect potatoes, has been found to form an endophytic association with sugar beet and Solanum rostratum [5, 6]. Additionally, pathogenic strains of C. nebraskensis have been isolated from asymptomatic maize as epiphytes but later induced disease when artificially inoculated [7]. This suggests that the life cycle of Clavibacter strains can be complex, involving periods of nonpathogenic endophytic colonization of carrier plants and subsequent dissemination onto host plants, or periods of epiphytic colonization on host plants, followed by disease induction when environmental conditions are suitable.

Comparative genomics studies have shed light on the issues surrounding the taxonomic status of pathogens within this genus [4, 8]. While the genus Clavibacter was originally proposed by Davis et al. (1984) to encompass all plant pathogenic coryneform bacteria with cell walls containing diaminobutyrate, many species were eventually reclassified as members of Rathayibacter and Leifsonia, leaving Clavibacter as a monospecific genus containing only C. michiganensis subspp [9, 10]. In 2018, genetic analyses within the genus, including average nucleotide identity (ANI), DNA-DNA hybridization (DDH), and multi-locus sequence analysis (MLSA), supported the elevation of some C. michiganensis subspecies to the species level, including C. nebraskensis (Cn), the causative agent of Goss’s Wilt & Leaf Blight (GWLB) [8].

Cn was first identified infecting maize in Nebraska in 1969, but quickly spread to surrounding states [11–13]. Deployment of maize hybrids with quantitative resistance to GWLB helped quell the outbreak, and after the early 1980s, GWLB did not pose a major threat in the US Corn Belt [13, 14]. However, since the early 2000s, Cn has re-emerged as a major threat to maize-growing regions, and has spread further north through the Upper Midwest and into Canada, south through Louisiana and Mexico, and more recently in South Africa [15–21]. It is hypothesized that changes in agronomic practices, such as low- or no-tillage and continuous corn cropping, may have contributed to its re-emergence [22]. However, there is evidence that changes in the population dynamics and genetics of Cn may also have contributed to its re-emergence [23]. Genetic techniques, including PCR-AFLP and MLSA, revealed distinct differences in population structure between strains causing early epidemics in the 1960s and 1970s and those isolated from the mid-2000s onward [7, 23, 24]. However, these techniques did not identify gene variants underlying virulence in the pathogen, although putative virulence factors, including the cellulase gene celA, have been identified in the genome of Cn. Some members of the genus Clavibacter encode virulence factors on extrachromosomal plasmids, including C. michiganensis, causing bacterial canker of tomato, and C. sepedonicus, causing bacterial ring rot of potato [3, 25–27]. However, some of these plasmid-encoded genes have no obvious homologs in the chromosomal genome of Cn, necessitating genomic sequencing and comparative analyses to identify virulence factors in this species.

Whole-genome sequencing and comparative genomic analyses have been widely applied across the genus Clavibacter to compare genomic content of strains of various species and economic hosts [1, 26, 28]. However, whole-genome sequencing and comparative analysis have not been used to characterize genomic variation within the Cn species. Therefore, we generated complete genome assemblies for 17 representative strains from both historic and recent outbreaks, ranging from avirulent to highly virulent, and compared them to publicly available genome sequences for the highly virulent type strain NCPPB 2581 and an avirulent strain HF4 [7]. We found that all Cn strains have one chromosome that is highly structurally similar across strains, and we identified five novel plasmids in the strains CDK032, CDK039, CDK044, CDK045, and CDK046. However, plasmids do not appear to be correlated with pathogenicity in Cn as they are in the related species C. michiganensis and C. sepedonicus [3, 29], due to the discovery of heterogeneous plasmids in both highly virulent and lowly virulent strains and lack of conserved plasmid proteins in highly virulent strains in this study. Haplotype analysis did not reveal any population structure linked to year or location of isolation, contrary to previous studies that found population differentiation between strains from early and later outbreaks [23, 24]. CelA, a putative virulence factor, is encoded on a genomic island shared amongst all strains, while additional genomic islands are variable across strains. Investigation of secreted CAZymes across strains, also hypothesized to be involved in virulence, revealed that they are well-conserved within the species. Analysis of allelic variation within secreted CAZymes revealed variable deletions within a secreted cellulase that had not previously been described in the literature.

Methods

Virulence assays and strain selection

We identified 17 diverse Cn strains from those analyzed in Veregge et al. (2025), spanning multiple collection locations including five US states and the Canadian province Manitoba, and diverse collection dates including early strains from the 1970s and more recent strains from the mid-2000s onward (Table 1) [30]. Apart from maximizing spatiotemporal diversity of strains, we took care to select strains that spanned the genetic diversity of our collection by assessing the phylogeny generated from whole-genome SNP data in Veregge et al. (2025), and specifically selected strains that clustered closely on the phylogeny but had distinct virulence phenotypes [30]. We assayed these 17 strains for virulence on the susceptible maize line Inbred 34-1141 (PI 186218), identified from a genome-wide association study of resistance to GWLB conducted by Singh et al. [31]. Seeds were obtained from the USDA National Plant Germplasm System [32] and virulence phenotypes were measured according to the methods described previously [30]. Briefly, strains were grown on Nutrient Broth Yeast Extract (NBY) plates for 5–7 days at 25 °C. Then, cells were scraped and suspended in 1X Phosphate Buffered Saline (PBS) and adjusted to OD600 = 0.4, or an approximate CFU/mL of 1 × 10⁸. Maize seedlings at V2-V3 growth stage were inoculated using a stem-stab method wherein 100µL of inoculum was dispensed in a wound made approximately 1 inch above the soil line, and another 100µL was dispensed in a wound made perpendicularly to the original wound for a total of 200µL of inoculum per plant. Control plants were inoculated using the same method with sterile 1X PBS. After inoculation, plants were maintained in a growth chamber with a 16-hour photoperiod cycling between 25℃ and 23℃ with 25% relative humidity. Strains were inoculated in triplicate technical replicates, and phenotyping was repeated for a total of two biological replicates, totaling six scores per strain taken as a percentage of whole tissue necrosis at 7 days post-inoculation. Strains NCPPB 2581 and HF4 were also assayed for virulence using the same methods for inclusion in comparative analyses as high and low virulence check strains, respectively. Strain NCPPB 2581 is the type strain from 1971 and is highly virulent, and strain HF4 is an asymptomatic isolate recovered from maize in 2012 [7, 12].

Table 1.

Strains selected for PacBio HiFi sequencing. Locations are Nebraska (NE), North Dakota (ND), Manitoba, Canada (CAN), Colorado (CO), Minnesota (MN), and South Dakota (SD)

Strain	Year	Location	Virulence	Assembly size (bp)	GC%	Contigs	Sequencing Coverage	No. Genes	No. CDS	Plasmid Size (bp)
CDK032	2012	NE	78.3%	3,191,185	72.8%	2	220	3066	3010	61,970
CDK039	2014	ND	92.5%	3,147,131	72.9%	2	280	3034	2978	83,288
CDK044	2015	ND	38.3%	3,125,097	72.9%	2	280	3005	2949	26,737
CDK045	2016	ND	23.3%	3,146,143	72.9%	2	210	3030	2974	41,599
CDK046	2016	ND	22.5%	3,163,640	72.8%	2	340	3036	2980	50,233
CDK057	2015	CAN	84.2%	3,063,148	73.0%	1	270	2943	2887
CDK091	2014	CAN	94.2%	3,072,024	73.0%	1	310	2945	2889
CDK110	2017	CAN	59.2%	3,063,835	73.0%	1	220	2938	2882
CIC056	1990	CO	53.3%	3,065,253	73.0%	1	260	2934	2878
CIC251	2009	MN	72.5%	3,060,407	73.0%	1	210	2938	2882
CIC252	2009	MN	74.2%	3,065,153	73.0%	1	270	2939	2883
CIC296	2011	SD	88.3%	3,068,945	73.0%	1	290	2942	2886
CIC349	2008	NE	20.0%	3,073,701	73.0%	1	330	2951	2895
CIC359	2011	ND	26.7%	3,060,413	73.0%	1	270	2938	2882
CIC365	2012	ND	5.8%	3,064,755	73.0%	1	280	2938	2882
CIC369	1972	NE	43.3%	3,064,420	73.0%	1	280	2940	2884
CIC400	2014	MN	90.8%	3,068,926	73.0%	1	230	2944	2888

DNA extraction and sequencing

We streak-plated strains on NBY agar for 5–7 days at 25 °C, then transferred a single colony into 10 mL of sterile NBY broth and incubated the culture overnight at 25 °C with shaking at 200 rpm. To isolate high-molecular-weight DNA, we centrifuged overnight cultures at 10,000 x g for 1 min, removed the media, and extracted DNA using the High Molecular Weight DNA Extraction Kit from New England Biolabs (Cat. # T3050L) according to the manufacturer’s protocols.

DNA was then sent to the University of Minnesota Genomics Center (UMGC) for long-read sequencing. DNA size and quality were determined via the Agilent Femto Pulse, and samples were sequenced on a PacBio 8 M Sequel II cell and received 0.76–1.2 Gb of HiFi yield per isolate.

Genome assembly and annotation from PacBio long read sequencing data

Long-read PacBio sequencing reads were cleaned of potential adapter contamination using HiFiAdapterFilt [33]. Cleaned reads were then used to estimate the depth of coverage and assembled genome length for each isolate. K-mers of length 51 nucleotides were counted with Jellyfish 2.3.0 [34]. The k-mer frequency distribution was then used to estimate the genome size and depth of coverage with Genomescope 2.0, using a haploid genome model [35]. All isolates showed expected estimated genome sizes and depths of coverage and were advanced to genome assembly.

Assemblies were produced using the consensus approach implemented by Trycycler 0.5.4 [36]. Briefly, sequencing data from each isolate is split into partitions, independently assembled, clustered to screen for technical artefacts, then merged into a consensus assembly. Each isolate was assembled separately in the following scheme. First, four mutually exclusive partitions of the adapter-cleaned sequencing reads were subsampled. Then, each read partition was assembled separately with hifiasm 0.19.6-r595, Flye 2.9.2, and Canu 2.2 [37–39]. This resulted in 12 independent assemblies for each isolate. Hifiasm was run with the “-k 51 -l 0 -f 0 --hg-size 3100k --n-hap 1 -D 2” options. Flye was run with the “--pacbio-hifi --iterations 5” options. Canu was run with the “-p Assembly genomeSize=3100k useGrid=false -pacbio-hifi” options. The 12 independent assemblies from each isolate were then clustered for manual inspection using the “trycycler cluster” routine. Potential assembly errors and artefacts, such as very short linear contigs or contigs found in only one assembly, were discarded from the clusters. Assembled sequences were retained if they were circular and were supported by multiple assemblers in multiple read partitions. Plasmids were predicted by the presence of a second complete circular contig supported by multiple assemblers. This resulted in each isolate having a group of accepted sequences produced from independent reads and assemblers, with each sequence group representing the chromosome or a plasmid. These sequences that passed clustering were then processed with the “trycycler reconcile” routine to reorient them so they start at an accepted origin of replication and represent the same strand. Reconciled sequences were then aligned with the “trycycler msa” routine. A consensus for each isolate was produced with the “trycycler consensus” routine. Finally, we mapped the cleaned reads back to the consensus assemblies to assess assembly quality and to polish the sequences with racon 1.4.20 [40]. Polished assemblies were annotated with Prokka 1.14.6, using the annotated reference genome of NCPPB 2581 (GCF_000355695.1) as a resource for known genes [41]. The genome completeness was evaluated using BUSCO v5.8.2 and lineage microbacteriaceae_odb12 [42]. Assemblies were deposited under the BioProject number PRJNA1145386.

Comparative genomics

ANI calculation

Average nucleotide identity (ANI) was calculated using FastANI v1.1 with all genomes sequenced in this study, as well as reference genomes for all other Clavibacter species in the genus and the outgroup Rathayibacter tritici. NCBI accession numbers of reference genomes used are reported in Table S1. In the computation of ANI, all genomes were included as both references and queries to allow all-to-all comparisons. ANI heatmaps were created using the R package pheatmap v1.0.13.

Haplotypes

Genetic variability among the studied Cn strains was evaluated using concatenated sequences of six housekeeping genes: atpD, dnaK, gyrB, ppk, recA, and rpoB. These loci were selected because they represent single-copy, essential genes that accumulate mutations at a rate suitable for distinguishing closely related strains and have been widely employed in Clavibacter phylogenetic studies [43]. In particular, atpD and dnaK have been reported to reveal population structure across decades and geographic regions in Cn [23].

Gene sequences for each locus were retrieved from the assembled and rotated genomes through a BLAST-based extraction pipeline. Reference sequences from the type strain NCPPB2581 were compiled in a multi-FASTA file and used as queries for BLASTn (BLAST+ v2.13.0) searches against each genome. Each genome was indexed as a local BLAST database, and searches were performed under stringent parameters (E-value ≤ 1e–10 and ≥ 80% sequence identity) to ensure accurate identification of orthologs. The top-scoring hits were parsed, and the corresponding genomic coordinates were used to extract nucleotide sequences with Bedtools (v2.31.0) using the getfasta function, with strand orientation preserved as indicated by BLAST.

All extracted loci were individually aligned using MUSCLE implemented in MEGA 12 [44]. The alignments were manually inspected, trimmed, and concatenated into a single supermatrix comprising approximately 12,141 bp after curation. The concatenated alignment was exported in NEXUS format and subsequently imported into PopART (Population Analysis with Reticulate Trees) for haplotype analysis [45]. Metadata, including the year of isolation and geographic origin of each strain, were appended to the dataset, and haplotype networks were constructed using the median-joining algorithm implemented in PopART.

Genetic diversity indices, including the number of segregating sites, number of parsimony-informative sites, and nucleotide diversity (π), were computed using the “Statistics” tool within PopART. Tests for neutrality were performed by calculating Tajima’s D statistic and its associated p-value based on coalescent simulations under the standard neutral model. Population differentiation was evaluated through an Analysis of Molecular Variance (AMOVA) performed in PopART, using 1,000 permutations to assess the significance of the fixation index (Φ_ST).

Chromosome and plasmid comparisons

Each chromosome was rotated using the dnaA gene as a reference using the Python code from Dr. Ralf Koebnik (IRD, France). Then all chromosomes were compared using the program Mauve [46]. Whole genome alignments from Mauve output with NCPPB 2581 as the reference strain were generated in Geneious (v2025.1.3) to call variants and assess allelic variation across strains. Plasmids were compared to one another using the annotated gbk files in Clinker [47], and to PLSDB and NCBI databases [48]. For PLSDB, the fasta file from the plasmid was uploaded to their website and searched with Mash dist strategy. For NCBI, a blastn search was performed with the core nucleotide database.

Gene orthology and variation

Orthology between translated coding sequences from C. nebraskensis samples sequenced in this study and the reference proteomes for the strains listed in Table S1 were compared using the software OrthoFinder (v2.5.5). Presence-absence variation and copy number variation were assessed by comparing orthogroups between strains. A phylogeny was inferred using the core gene MAFFT alignment from Panaroo (3.13.0) and the IQ-TREE3 GTR + I+G model with 1000 boostraps for the 22 Cn isolates, 17 from this paper and five from NCBI, with additional eight Clavibacter species, and Rathayibacter tritici NCPPB 1953 as an outgroup [49].

Functional annotation

Secreted proteins were predicted using SignalP 6.0 with the “Other” setting to accommodate Gram-positive bacterial proteins, and proteins with predicted Sec and Tat secretion signals were further characterized [50]. Putative carbohydrate-active enzymes of secreted proteins were predicted using the CAZy database [51]. Genomic islands were predicted using IslandViewer4 [52]. Proteome BLAST searches were conducted against the UniRef_90 database and the non-redundant (nr) protein database on NCBI to further characterize genomic coding sequences [53, 54]. BLAST searches were conducted using peptide sequences (blastp) with a minimum query coverage of 20%, a maximum E-value of 1E-5, and up to 20 target sequences returned for analysis. Conserved domains and catalytic sites were predicted using InterProScan [55]. Structural analysis was performed with Foldseek [56]. Bacteriocin-associated gene clusters were predicted with the software BAGEL4 and then compared using Clinker v0.0.31 [47, 57].

Cellulase activity

The ability of Cn strains to degrade cellulose was evaluated using a carboxymethyl cellulose (CMC) plate assay adapted from Pfeilmeier et al. (2024) with modifications [58]. As basal medium, Rathaybacter Stunting Disease / Ratoon Stunting Disease (RSD) agar was prepared following the formulation modified from Davis and Vidaver (2001) in Laboratory Guide for Identification of Plant Pathogenic Bacteria [59], containing (per liter): yeast extract 1.0 g, (NH₄)₂HPO₄ 1.0 g, MgSO₄·7 H₂O 0.2 g, KCl 0.2 g, L-cysteine (free base) 1.0 g, bovine hemin chloride 15 mL (0.1% w/v in 0.5 N NaOH), bovine serum albumin fraction V 10 mL (20% aqueous solution), and agar 15 g. The medium was supplemented with 0.5% (w/v) sodium CMC (Sigma-Aldrich, C5678) as the sole carbon source.

The inoculum was prepared by culturing the strains overnight in liquid nutrient broth yeast extract (NBY) at 25 °C in a rotary shaker at 120 rpm. Cells were harvested by centrifugation at approximately 1,000 × g for 10 min, washed once, and resuspended in phosphate-buffered saline (PBS) to an optical density of 0.1 at 640 nm [60]. Each RSD-CMC plate was spotted in triplicate with 7 µL of bacterial suspension at equidistant positions and incubated at 25 °C for 10 days. The Cn type strain NCPPB 2581 was included on each plate as a positive control for CMC hydrolysis, and PBS alone was included as a negative control to verify that halo formation was attributable to bacterial activity rather than the spotting buffer or medium.

After incubation, cellulose degradation was visualized by flooding the plates with 0.1% (w/v) Congo red solution for 30 min, followed by repeated washing with 1 M NaCl until the background cleared and hydrolysis zones became visible. The presence of yellow halos surrounding the colonies was interpreted as indicative of carboxymethyl cellulase activity.

Halo size was quantified from plate images by measuring the halo diameter (mm) for each spot. Areas (cm²) were computed as 3.1416 × (d/2)² after conversion to centimeters. Statistical analyses were conducted in R version 4.4.2 [61]. For each strain, halo areas across technical replicates were treated as response values, normality and variance assumptions were assessed graphically, and a one-way ANOVA was applied with strain as the fixed factor. Multiple comparisons were performed using Tukey’s Honest Significant Difference (HSD) test (α = 0.05). Figures were generated with “ggplot2”; letter groupings from Tukey HSD were overlaid on boxplots to summarize pairwise differences [62].

Sequences at the locus “CMN_02390” were extracted from the Mauve whole genome alignment in Geneious version 2025.1.3 and aligned with MAFFT version 7.490 the alignment algorithm set to “auto”, the scoring matrix set to 200PAM / k = 2, a gap open penalty of 3, and an offset value of 0.123. The resulting alignment was exported as a fasta file and imported into R version 4.5.1 as a genclone object. A neighbor joining (NJ) tree was calculated from this genclone object using the “aboot” function in the “poppr” package (version 2.9.7) using Provesti’s distance and 200 bootstraps with a cutoff value of 50. The resulting dendrogram was overlaid with cellulase activity using the R package “ggtree” version 3.17.0.

AMOVA was performed with cellulase activity as a hierarchical factor explaining molecular variance at the “CMN_02390” locus. Strains were assigned to cellulase activity groups based on their statistical groupings from the previous ANOVA and Tukey HSD tests, and AMOVA was performed using the R ‘poppr’ package using the function ‘poppr.amova’ with the method set to ‘ade4’. Results from the AMOVA were then input into the ‘randtest’ function with 1000 repetitions to obtain a significance p-value.

Results

Virulence scoring

We assessed the ability of strains in this study to induce tissue necrosis at 7 days post-inoculation (dpi) on a susceptible maize cultivar through a stem-stab inoculation method. We found that the 19 Cn strains assayed, including the 17 strains sequenced in this study and the high- and low-virulence check strains NCPPB 2581 and HF4, showed a continuous spread of phenotypes. ANOVA followed by a post-hoc Tukey’s Honest Significance Difference (HSD) test grouped strains into eight statistical bins, with significant overlap amongst groups. Of these groupings, CIC365 and HF4 are significantly less virulent than all other strains, and strains CDK039, CDK091, and CIC400 are significantly more virulent (Fig. 1).

Fig. 1.

Phenotypic analysis of 17 Cn strains selected for PacBio sequencing, and two reference strains. A Inoculations were performed using a stem-stab inoculation method on susceptible maize line Inbred 34-1141. Virulence scores were assigned by measuring the percentage of necrotic tissue of the host plant at 7 dpi. Reference photos for tissue necrosis scores assigned 5%, 30%, 50%, and 100% are shown. B Boxplot of virulence data for each strain from six independent scores per strain. Reference strains NCPPB 2581 and HF4 are presented first, followed by strains sequenced in this study in ascending order of mean virulence

Genome sequencing and assembly

We sequenced and assembled the 17 Cn strains’ long reads into complete genomes, each with BUSCO completeness values above 99.3% (Table S2). Each assembly contained one complete chromosome represented by a single circular contig with a mean coverage of 270X, and strains CDK032, CDK039, CDK044, CDK045, and CDK046 each carried a plasmid with variable sizes, identified by the presence of a second complete circular contig (Table 1). Type strain NCPPB 2581 contains no plasmids and has a genome assembly size of 3.06 Mbp, and genome assembly sizes of Cn strains sequenced in this study ranged from 3.06 to 3.07 Mbp for the single chromosome. Average nucleotide identity (ANI) indicated that all strains belong to the species C. nebraskensis, with ANI values ranging from 99.6% to 100% within Cn strains and 91% to 95.2% with all other Clavibacter spp. (Fig. S1). Likewise, all Cn strains sequenced in this study cluster in the phylogenetic tree with the type strain of Cn (Fig. 2).

Fig. 2.

Phylogenetic tree of 17 sequenced Cn strains from a core genome alignment with other strains of Cn, other members of the genus Clavibacter, and the outgroup Rathayibacter tritici. All strains sequenced in this study are highlighted as bold fonts, and these strains cluster with publicly available genomes of Cn strains. Slashed lines on the phylogenetic tree indicate phylogenetic distance and were utilized to compare intragenic relationships between Clavibacter strains more closely. Scale bar indicates nucleotide substitutions per site, and branch labels indicate bootstrap values (1000 bootstraps)

Comparative genomics

Haplotype analysis reveals low genetic differentiation

A median-joining haplotype network was constructed from the concatenated alignment of six housekeeping genes (atpD, dnaK, gyrB, ppk, recA, and rpoB) to infer multilocus relationships among Cn strains collected from multiple U.S. states and Canada. A total of 23 variable sites were detected, including 20 parsimony-informative positions, which defined several distinct haplotypes.

Shared multilocus haplotypes were detected among several groups of isolates, including CDK091–CDK044, CIC365–CIC252–CIC056, CDK057–CDK045, CDK110–CDK039, and CIC400–CIC359–CIC296–CIC251–CDK046. These clusters suggest the presence of highly similar or identical genotypes distributed across multiple regions. When the haplotype network was color-coded by geographic origin (Fig. 3A), no clear spatial clustering was observed. Strains from different U.S. states and Canada were interspersed across the network, and several haplotypes contained isolates from multiple regions, indicating limited phylogeographic structure. When the same network was color-coded by year of isolation (Fig. 3B), haplotypes were again distributed across time points without evidence of temporal clustering. This pattern suggests that the same multilocus genotypes have persisted through different collection years.

Fig. 3.

Haplotype network analysis reveals low genetic differentiation. The haplotype network was built from six housekeeping genes (atpD, dnaK, gyrB, ppk, recA, and rpoB). A Colors represent the year of isolation. B Colors represent the location of isolation. IA = Iowa; NE = Nebraska; ND = North Dakota; CAN = Manitoba, Canada; CO = Colorado; MN = Minnesota; SD = South Dakota

The overall nucleotide diversity was low (π = 0.000734), and Tajima’s D was positive (D = 1.379, p = 0.095), showing no significant deviation from neutrality. An AMOVA indicated that most of the genetic variation occurred within populations (116.85%), while only a small, non-significant proportion was attributed to among-population differences (Φ_ST = − 0.168, p = 0.894). Together, these findings indicate minimal geographic or temporal differentiation among the studied Cn strains.

Chromosome structure and orthologous gene families

To examine chromosome structure and facilitate comparisons of gene families and alleles between strains, we aligned all Cn assemblies using Progressive Mauve [46], including the highly virulent reference strain NCPPB 2581 and the non-pathogenic strain HF4. Strains exhibit a conserved genome structure, with two collinear blocks identified (Fig. S2). Using NCPPB 2581 as the reference genome, we analyzed sequence variants produced by the Mauve genome alignment with a co-occurrence matrix of SNPs across strains to visualize conserved variants across strains (Fig. S3). Strain CIC349 contained the greatest number of SNPs from the reference strain (4,871) and strain CIC056 contained the least (688). Strains shared, at a minimum, 5.6% of their variants. CIC349 and CDK032 shared the least number of SNPs with other strains; CIC349 shared the greatest percentage of SNPs (67.9%) with strain CIC400, and CDK032 shared 79.6% of SNPs with strain CIC349. The remaining strains shared between 94.2% and 99.5% of SNPs with their most closely related neighbor. Similarly to the phylogenetic tree, proximity based on shared SNPs between strains did not indicate a shared virulence phenotype.

Individual Cn strains in this study encoded between 2,934 and 3,066 genes total, with an average of 2,967 genes per strain (Table 1). Including NCPPB 2581 and HF4, Cn strains share 2,771 gene families in their core genome (Fig. 4). The remaining gene families comprise 380 unique orthogroups with presence-absence variation (PAV) and copy-number variation (CNV) across Cn genomes (Fig. 4, Table S2). Of these 380 genes, 179 are encoded on five of the strains’ plasmids, and many of the remaining 201 genes are adjacent to recombinase or integrase family proteins, indicating putative horizontal gene transfer within genomic islands, discussed in later sections.

Fig. 4.

Gene family presence-absence variations across the 17 sequenced strains and two Cn reference strains. The reference strains included Cn strain NCPPB 2581 (GCF_000355695.1), and Cn strain HF4 (GCF_009739595.2). The horizontal bars on the left represent the number of predicted proteins. The vertical bars show the number of orthologous groups. For each bar, the panel of dots below indicates the presence of the group in each strain. Connected dots signify that the orthologous group is shared among those strains

Notably, we observed that the non-pathogenic strain HF4 is missing an orthologous gene family predicted to be a response regulator transcription factor (Table S3). After manual inspection, we found that in strain HF4, an insertion in the sequence at locus EGX36_RS05980 (CMN_01212 in strain NCPPB 2581) leads to a premature stop codon. This transcription factor is adjacent to a sugar ABC transporter substrate-binding protein, a sugar ABC transporter ATP-binding protein, and a sugar ABC transporter permease. It may therefore play a role in the transcriptional regulation of carbohydrate uptake.

Genomic islands

While many pathogenicity determinants are encoded on plasmids in phytobacterial pathogens, other regions associated with pathogenicity include genomic islands, regions within the genome that likely originated by horizontal gene transfer and are identified by the presence of mobility genes and differences in GC content from the remainder of the genome. Hence, genomic islands that harbor virulence genes are referred to as “pathogenicity islands” (PAIs). PAIs have been identified in C. michiganensis, including the chp/tom PAI, which encodes genes necessary for pathogenesis [63–66]. Apart from genes directly associated with virulence, genomic islands may also contain genes involved in toxin/antitoxin systems, phage protection, and metabolism, increasing the competitive fitness of bacterial isolates containing these regions [67]. We used IslandViewer4 and identified 9 genomic islands (GIs) in the 19 Cn genomes, with one GI associated with plasmids (Fig. S4, Table S3).

All Cn strains contained genomic island (GI) cluster 1, which spans approximately 16 kb and contains 22 proteins, including the CelA protein encoding an expansin EXLX1 family cellulose-binding protein (Fig. S4, Table S3). This protein contains a cellulose-binding domain CBD II in the N-terminal region, and an expansin EXLX1-like domain in the C-terminal region, but notably, it lacks the cellulase domain of its homolog in C. michiganensis. Also encoded within GI cluster 1 are proteins involved in signal transduction, antitoxin and bacteriocin resistance systems, response to environmental stress, and primary metabolism (Table S3).

Many of the remaining GIs were marked by the presence of prophage insertion elements. They did not appear to be correlated with strain virulence, as they were spread across strains with varying virulence phenotypes (Fig. S4). Some GIs were located in genomic regions harboring genes associated with transcriptional regulation, membrane transport, and carbohydrate metabolism (Table S3).

Plasmid structure and content

Strains CDK032, CDK039, CDK044, CDK045, and CDK046 each contain one circular plasmid, ranging from 26,737 to 83,288 bp in size (Fig. 5; Table 1). To date, there is only one sequenced strain, Cn strain 44, reported to harbor the plasmid pCML2 (GenBank CP086347). Of the strains sequenced in this study containing plasmids, CDK044, CDK045, and CDK046 are weakly virulent and CDK032 and CDK039 are highly virulent. Therefore, plasmid presence alone is not indicative of strain virulence. Plasmids are heterogeneous amongst these strains, and each contains between 29 and 89 proteins. Per plasmid, 27 to 84 of these proteins encode products that are annotated as hypothetical proteins, although many have homology to proteins in the UniRef database that suggest putative functions.

Fig. 5.

Plasmid alignment. Shared colors indicate similar annotations. Gray indicates less than 30% similarity

The CDK046 strain carries a conjugative MOBF plasmid that shares 99.3% identity and 100% coverage with the C. nebraskensis strain 44 plasmid pCM2L (GenBank CP086347) on NCBI. On PLSDB, the first hit was C. michiganensis subsp. michiganensis NCPPB 382. Alignment between the CDK046 plasmid and pCM2L revealed two locally collinear blocks: one 4 kbp and the other 70 kbp. Similarly, the conjugative MOBF plasmid in CDK032 showed C. michiganensis NCPPB 382 as the first hit on PLSDM and NCBI C. nebraskensis strain 44 plasmid pCM2L, with 92% and 60% identity, respectively. The CDK039 strain carries a plasmid lacking a clear relaxase type, with no hits on PLSDB. On NCBI, its closest match was C. sepedonicus strain K496 plasmid pCSL2, with 87% identity and 54% coverage. The CDK044 strain harbors a non-mobile plasmid highly similar to C. insidiosus and C. michiganensis NCPPB 382 on PLSDB. On NCBI, its main hit was C. tessellarius strain ATCC 33,566 plasmid pCT2, with 97% identity and 78% coverage. Finally, the CDK045 strain contains a plasmid with no hits on PLSDB. Its closest match on NCBI was C. insidiosus strain R1-3 plasmid pCI2, with 84% identity and 23% coverage. When comparing all plasmids, the CDK046 and CDK032 plasmids exhibited similarities, although CDK032 encoded more genes. CDK039 carried a higher number of coding genes but showed no similarity to the other four plasmids. CDK044 and CDK045 plasmids were more distantly related to the rest.

Across plasmids, various proteins involved in the Type II and Type IV secretion systems, conjugative DNA transfer, and DNA transcription and repair were encoded. Also, genes involved in toxin/antitoxin systems, important for plasmid stability and stress response were conserved. Some genes encoding proteins with CAZyme activities were identified, including alpha/beta fold hydrolases in CDK032 and CDK046 that have homology to a cutinase encoded by C. michiganensis (pCM2_0028). However, no conserved CAZymes were found across the plasmids of highly virulent strains CDK032 and CDK039, indicating that previously identified essential virulence factors are likely not encoded on these plasmids (Table S4). Most mobile elements were related to transfer, replication/recombination/repair, integration/excision, and stability/transfer/defense.

Bacteriocins

The prediction of bacteriocins using BAGEL4 identified a single sactipeptide biosynthetic gene cluster (BGC). Comparative genomic analysis of this BGC revealed that 90% of the coding genes within the cluster displayed 100% identity across all strains (Fig. S5). Its presence in both virulent and avirulent strains indicates it may be more related to bacteriocin production or competitive potential rather than directly linked to pathogenicity.

Secreted carbohydrate-active enzymes

Small secreted proteins are hallmarks of microbial effectors and are implicated in the virulence of numerous phytopathogens [68–70]. Within other Gram-positive phytopathogens, the general secretory pathway (Sec) and twin-arginine translocation pathway (Tat) have been implicated in the transport of virulence-related proteins across the cytoplasmic membrane [3, 71]. Within the Cn strains analyzed in this study, we did not detect any differences in cellular machinery associated with the Sec or Tat translocation pathways across genomes in this study, indicating that all strains can utilize these systems.

We further investigated proteins secreted through the Sec or Tat pathways with probable function in pathogenicity. Each strain secretes between 188 and 197 proteins predicted by SignalP based on conserved signal peptides, with 106 to 115 secreted through the Sec or Tat pathways. We narrowed our search to Carbohydrate-Active enZymes (CAZymes) secreted through the Sec or Tat pathways, which are crucial for the virulence of other pathogens within the genus Clavibacter [26, 65, 72, 73]. CAZymes have a multitude of functions in plant pathogenic microbes, including the metabolism of diverse carbohydrates in the plant cell wall to facilitate pathogen entry and colonization, as well as nutrient uptake and absorption. Overall, Cn strains in this study, including HF4 and NCPPB 2581, share 54 of their 59 total secreted CAzymes. We investigated the potential roles of proteins with predicted signal peptides and CAZyme activity in the polysaccharide lyase (PL), carbohydrate esterase (CE), glycoside hydrolase (GH), glycosyl transferase (GT) and carbohydrate-binding molecule (CBM) classes of CAZymes.

Conserved CAzymes

We observed no copy number variation of secreted proteins in orthogroups functioning as polysaccharide lyases (PL), carbohydrate esterases (CE), and glycosyltransferases (GT), indicating that they are likely essential to Cn (Fig. 6).

Fig. 6.

CAZyme families among C. nebraskensis strains. The CAZyme families were separated by CAzy class (CBM = Carbohydrate Binding Module, CE = Carbohydrate Esterase, GH = Glycoside Hydrolase, GT = GlycosylTransferase, PL = Polysaccharide Lyase). Variations in the number of orthologous gene families in CAZy classes CBM32, CBM13, GH19_1, GH184, and GH18 across strains are observed

CAzymes varying in copy number

We observed minor copy number variation in secreted proteins in orthologous gene families functioning as carbohydrate-binding modules (CBM) and glycoside hydrolases (GH), involved in the binding and breakdown of various carbohydrate structures, respectively. However, we did not identify any variation in orthogroups shared exclusively in either highly virulent or weakly virulent strains.

All Cn strains but three (CDK046, CDK057, and CIC359) secrete a putative lipase in the CBM13 family with domain similarity to the fermentation-respiration switch esterase FrsA, involved in respiratory response to sugar availability (Fig. 6). We identified seven sequence variants of this lipase, and a deletion in the coding sequence at position 49 truncated the start of the open reading frame, abolishing the secretion signal from those strains predicted to lack this protein. CDK032 secretes only one dienelactone hydrolase family protein protein in the CBM13 class while all other Cn strains secrete two (Fig. 6). These proteins are predicted to function in chlorocatechol degradation. Only 6 of 18 strains secrete a non-hemolytic phospholipase C (PlcN) in the CBM32 family (Fig. 6), while the strains not predicted to secrete this protein contain a single nucleotide deletion mid-gene that abolishes the signal peptide from this protein.

Strains CDK032 and CDK045 secrete one additional GH each, encoded on their plasmids, of classes GH19_1 and GH184, respectively (Fig. 6). The hypothetical protein of class GH19_1 in CDK032 contains an RHS repeat-associated core domain and likely functions as a toxin. In contrast, the hypothetical protein of class GH184 in CDK045 contains no conserved protein domains. CAZymes of class GH184 generally function as lysozymes, and therefore may play a role in intercellular competition. CIC400 was predicted to be missing a secreted protein of class GH18, but manual inspection of this region in CIC400 revealed that it shares 100% sequence identity with strains CIC296 and CIC349 that are predicted to encode this secreted glycosyl hydrolase.

Allelic variations in a secreted cellulase

While CAZyme copy number variation was relatively low amongst Cn strains, we investigated allelic variation within secreted glycoside hydrolases, which degrade complex sugars in plant cell walls and are likely important to the pathogenicity of Cn. Notably, we identified considerable allelic variation in a secreted endoglucanase corresponding to the locus “CMN_02390” in the reference strain NCPPB 2581. UniProt predicts the activity of this protein to be ‘endohydrolysis of 1,4-beta-D-glucosidic linkages in cellulose, lichenin and cereal beta-D-glucans’. The avirulent strain CIC365 contains a unique 60 bp deletion, resulting in a missing repeat region (PQPQPQPQPQPQPQPEPQP) corresponding to a linker polypeptide chain connecting the carbohydrate binding module and cellulose domains on either side (Fig. S6). Across all Cn strains analyzed in this study, there are 12 sequence types composed of variants from nucleotide positions 1,224 to 1,302 with varying effects on the total length of the repeat region (Fig. S6). CIC365 contains the largest deletion, followed by a sequence type shared by HF4 and CIC369. Notably, this protein is upstream of an extracellular solute-binding protein and a sugar ABC transporter.

We tested in vitro cellulase activity to explore correlations with sequence variation at this locus and the ability to degrade cellulose in vitro. Strains showed a range of cellulase activity, measured by the size of the halo formed on CMC media (Fig. 7A). When overlaying the in vitro cellulase activity to sequence types at the locus “CMN_02390”, we observed that strains visually clustered on the dendrogram by in vitro cellulase activity (Fig. 7B). However, AMOVA did not support a statistically significant relationship between cellulase activity and sequence variation at the “CMN_02390” locus (p = 0.39).

Fig. 7.

Cellulase activity and CMN_02390 sequence types in Cn strains. A In vitro cellulase activity of Cn strains, ordered by strain virulence as presented in Fig. 1. B Dendrogram of the sequence types of the secreted cellulase at the locus CMN_02390, overlaid with cellulase activity in vitro

Discussion

This study provides the first comprehensive analysis of intraspecific genomic variation in Cn using 17 PacBio HiFi genome assemblies from diverse isolates, compared with the virulent type strain NCPPB 2581 and the avirulent strain HF4. In our study, we sought to characterize the diversity of orthologous gene group families across the species. From genomic analyses of the 17 Cn genomes we assembled, queried against the genomes of Cn strains NCPPB 2581 and HF4, we found considerable copy number variation (CNV) in orthogroups amongst Cn strains, with 380 variable gene families. Many of these non-conserved orthologous gene families comprise plasmid-encoded genes or genes likely gained via horizontal gene transfer, as they are localized near recombinase or integrase family proteins. Genomic island analysis verified the likely horizontal acquisition of a number of these variable orthogroups, and revealed that the celA gene, encoding an expansin EXLX1 family cellulose-binding protein and widely hypothesized to be a key virulence factor of Cn, is encoded within a genomic island. In C. michiganensis, celA is essential for pathogenicity in tomato, suggesting that it may also be important for pathogenicity in Cn [25, 26, 73, 74]. Expansins are hypothesized to play important roles in the loosening of plant cell walls to allow cellulases and other CAZymes access to cell wall polymers [75]. It is worth noting that the celA gene in Cn contains carbohydrate-binding and expansin domains, but does not contain the cellulase domain present in the celA homolog in C. michiganensis. However, conflicting results of the role of the expansin domain in the activity of CelA in C. michiganensis have been reported, with Jahr et al. (2000) reporting its presence necessary for pathogenicity, and Hwang et al. (2019) reporting the ED domain as dispensable for such activities [25, 73]. Exploring the functional role of this expansin-like protein is the next step in determining its importance to the pathogenicity of Cn.

Additionally, we investigated the level of conservation of secreted CAzymes across strains. Although we hypothesized that variation in CAZyme content could be associated with strain virulence, we did not identify any CAZymes present only in highly virulent or weakly virulent strains, indicating that gene copy number variation in this class of proteins does not appear to underlie virulence in the strains studied. Within non-conserved CAZyme orthogroups, we identified two secreted lipases that vary in presence across strains. The first encodes a protein with a domain similar to FrsA (Fermentation-Respiration Switch Protein), which could modulate respiration in response to sugar availability in the host. The other lipase encodes phospholipase C and is secreted by only 6 strains. The role of phospholipases in microbial virulence in plant-microbe systems is relatively unexplored; however, the lipase lipA has been shown to function in virulence of plant pathogenic Xanthomonas species by operating as a cell wall-degrading enzyme [76–78]. Additionally, a phospholipase secreted by Verticillium dahliae was recently found to function as a virulence factor by hijacking host cellular transport to localize to the nucleus, subsequently suppressing pattern-triggered host immunity [79].

The remaining secreted CAzymes are highly conserved across strains, indicating their importance to the fitness of Cn. These gene families are likely to be involved in processes such as pectin degradation, cellulose binding and degradation, cell wall synthesis, and intercellular competition. Cell wall-degrading enzymes are likely a key contributor to the pathogenesis of Cn. Within the cell walls of grasses, mixed-linked glucans and xylans play important structural roles, and β-1,3/β-1,4-mixed linked glucans are unique components of the cell walls of members of Poales [80, 81]. The ability to bind and degrade these molecules, or to inhibit the host’s ability to modify and strengthen these components, is therefore likely important to pathogens of monocots [82–84]. Moreover, modification of these molecules has been shown to induce defense responses in plant host cells [85]. All Cn strains were found to secrete a hypothetical protein of CAZyme class CBM11, which is involved in the binding of β-1,4-glucan and β-1,3 − 1,4-mixed linked glucans. In addition, three endo-1,4-β-xylanases were found in the genomes of all strains sequenced in this study. Only one of these three genes was secreted via the TAT secretion pathway in all Cn strains, suggesting potential importance during infection, as xylan is a primary structural component of hemicellulose of monocots [80, 86]. Proteins involved in pectin degradation, including pectate lyases and pectinesterases, were identified in all pathogenic Cn isolates, although monocot cell walls contain very low levels of pectin compared to dicots [80]. However, the production of xylem-clogging compounds, including pectin, has been identified as a host plant method of controlling xylem-dwelling pathogens such as Cn. For example, the formation of tyloses or pectin-rich gels can halt pathogen proliferation and translocation in the host, effectively halting the spread of infection [87, 88]. In C. michiganensis, a pectate lyase pelA1 of the PL3 family in the chp/tom pathogenicity island of Cm greatly contributes to wilting of the host tomato plant [3].

While plasmids are often implicated in bacterial virulence, including in the closely related species C. michiganensis and C. sepedonicus, plasmids are not well-characterized in Cn. In early studies of the genus Clavibacter (previously within Corynebacterium), plasmids were found in all surveyed Cm strains and were highly diverse, while plasmids were found in only ~ 20% of surveyed Cn strains and were homogeneous [89]. Of the 31 publicly available genome sequences of Cn at this time, only Cn strain 44 contains a plasmid (RefSeq GCF_023279165). Of the seventeen strains sequenced in this study, five contained plasmids, accounting for 29% of strains sequenced. Notably, the five plasmids sequenced in this study are heterogeneous. Two have syntenic regions to the plasmid pCM2L from Cn strain 44, and the remaining three plasmids are distinct in their composition.

Interestingly, while we sequenced strains from both historic and current outbreaks, all plasmid-containing strains were isolated from 2012 to 2026 outbreaks, suggesting a recent diversification of plasmids in the species. Further genome sequencing of historic and recent isolates may confirm this hypothesis. Four of the five strains containing plasmids were isolated between 2014 and 2016 from North Dakota, although the two strains containing syntenic regions to pCM2L from Cn strain 44 were from different geographic locations, one from North Dakota and one from Nebraska. Cn strain 44 was originally isolated from Nebraska in 2011 and represents the earliest known occurrence of this plasmid. Nebraska is the site of the first outbreak of this pathogen in the United States, and therefore might indicate a potential point of diversification within the species. Similarly to early studies of Cn, we found no correlation between plasmid content and strain virulence. Therefore, it is likely that the functional roles of these plasmids confer a fitness advantage apart from virulence. For example, many toxin/antitoxin systems and MFS transporters were identified on Cn plasmids. These may function to regulate biofilm formation under environmental stress, or exclude antimicrobial compounds and heavy metals from the cell, fine-tuning bacterial survival in the host or under field conditions where competing microbes or agrochemical inputs may exert selective pressures on the pathogen population [90, 91]. Under intensified agricultural systems and increased inputs, plasmids conferring fitness advantages under these conditions may be selected for.

Since resistance to Goss’s Wilt is highly quantitative [31, 92, 93], it is likely that numerous sequence polymorphisms and gene copy number variations underlie virulence phenotypes. In other members of the genus Clavibacter, serine proteases and cell wall-degrading enzymes such as cellulases and pectinases have been identified as key virulence factors [25, 63, 66, 73]. However, in Cm, secreted CAZymes are generally highly conserved even amongst strains of varying virulence. It has been proposed that sequence variation and deletions in virulence genes in the Cm genome, such as celA, celB, xysA, xysB, and pat-1, may underlie differential virulence phenotypes [94, 95]. Gene homologs, such as celA/celB, which harbor considerable sequence variation and coding-sequence truncations, may be undergoing pseudogenization, a process that occurs during the adaptation of a pathogen to its host [96]. We observed a phenomenon in the secreted cellulases found in the genomes of Cn that supports this hypothesis. In Cn, sequence polymorphisms in celA (CMN_00144) have shown no correlation with the virulence phenotype [7, 23]. In our study, we identified variable deletions in a different secreted cellulase family glycosylhydrolase (CMN_02390) occurring in a repeat region located between the carbohydrate binding module and the cellulase domains. This region is a linker polypeptide chain connecting the active domains, and we hypothesize that shorter linker peptides may have downstream effects on the positioning of the active domains, leading to inefficient binding. Sequence variants at the locus “CMN_02390” have not previously been analyzed for correlation with virulence, but this gene was reported to be upregulated in the secretome of Cn in maize xylem sap, suggesting its importance in pathogenicity [97].

To further study the impact of sequence variation at the “CMN_02390” locus with cellulase activity in vitro, we conducted cellulase degradation assays on CMC media. We hypothesized that highly virulent strains would show a greater ability to degrade cellulose in vitro, as was shown in Cm [95] . However, our results showed that while most highly virulent strains exhibited higher cellulose degradation than weakly virulent strains on the CMC media, that did not always hold true. Strains CIC251 and CIC252, which were moderately virulent in our in planta phenotyping assays and grouped into the statistical bin ‘abcd’, showed very low cellulose degradation in vitro. Conversely, of the two avirulent strains, one exhibited very low cellulase activity (CIC365) while the other (HF4) exhibited cellulase activity similar to that of moderately virulent strains. While we observed some visual clustering on the dendrogram of sequence types of “CMN_02390” by cellulolytic ability, we did not find any statistically significant relationship between the two variables. Three factors may explain this observation. Firstly, genetic drift may have contributed to the proliferation of certain genotypes at the “CMN_02390” locus unrelated to cellulolytic phenotypes. Secondly, while “CMN_02390” functions as an endoglucanase, each strain secretes an additional two endoglucanases at the loci “CMN_02650” and “CMN_02651”. A previous study failed to link sequence variation at these two loci to virulence phenotype [7]. Therefore, these endoglucanases may act in concert, and low activity of one enzyme may be masked by the presence of a functionally redundant enzyme, supporting the hypothesis that pseudogenization at some virulence loci may be occurring in Cn, as in Cm. Finally, our study lacked the statistical power necessary to conduct association analyses that might link sequence variation at the locus “CMN_02390” with cellulase activity or virulence phenotype in planta due to the small number of genomes analyzed. Sequencing a larger number of genomes and performing genome-wide association analyses may shed more light on the importance of “CMN_02390” in cellulose degradation and pathogenicity of Cn.

Our study found that most genes hypothesized to be crucial to Cn pathogenicity, such as cell wall degrading enzymes, are largely conserved across strains of varying virulence. Future studies focusing on gene expression differences between weakly and highly virulent strains, either in planta or in xylem sap, may shed light on transcriptomic changes associated with high virulence. Resistance to GWLB is highly quantitative, and as such, it is possible that subtle differences in bacterial responses to oxidative stress or in host antimicrobial defense mechanisms, rather than genes directly involved in host cell wall degradation, might also influence overall strain virulence. Additionally, characterization of putative pathogenicity genes in Cn via functional analyses remains a clear point for further research. This study serves as an important source of genomic data for elucidating the diverse mechanisms underlying virulence in Cn, aiding in the development of resistant maize varieties and helping to inform future molecular pathotyping of Cn isolates.

Conclusions

This study utilized comparative genomic analyses to explore genomic variation between Cn strains from distinct outbreaks, both historic and recent, by generating complete genome sequences for 17 strains and including two publicly available genomes for which we generated virulence phenotype data. Haplotype analysis revealed no population structure underlying strains from various geographic locations and time points, contrary to previous findings. Between Cn strains, we identified hundreds of orthologous gene families with CNV located mainly within plasmids and genomic islands. We predicted five unique plasmids from strains ranging from weakly to highly virulent, representing the most diverse plasmid discovery in the species to date. Many other genes with CNV likely function in intercellular microbial competition and defense, and a smaller subset may function in carbohydrate metabolism and virulence. We found that the celA gene, predicted to be involved in Cn infection, is encoded within a genomic island, indicating horizontal transfer of this putative virulence factor. We found that secreted CAZymes are highly conserved in the species, with 54 of 59 identified CAZymes present in all Cn genomes in this study. We identified allelic variants of a secreted cellulase that have not been previously discussed in the literature at the locus “CMN_02390”. We tested cellulase activity in vitro and found that sequence types of “CMN_02390” loosely cluster based on cellulase degradation ability, although we were unable to detect statistically significant relationships between the two variables given a small sample size.

Abbreviations

Cn

Clavibacter nebraskensis

GWLB

Goss’s Wilt & Leaf Blight

PAV

Presence-absence variation

CNV

Copy number variation

GI

Genomic island

Authors’ contributions

M.V. performed DNA extractions, L.F.F. performed cellulase activity assays, M.V., L.F.F., T.K., and V.RR. performed data analyses, and M.V. drafted the manuscript. D.K. conceptualized the research objectives, designed the experiments, and edited the manuscript. All authors helped with data analyses, reviewing, and editing the manuscript.

Funding

This work was supported by the National Science Foundation under grant no. IOS 1856744 AM 002 and Research Computing grant at University of Minnesota to D.K., Bayer Crop Science/University of Minnesota Multifunctional Agriculture Initiative Graduate Student Fellowship to M.V., USDA National Institute of Food and Agriculture, and Federal Appropriations under project PEN04997 (accession 7007260) to V.RR.

Data availability

Genomes reported in this study are available on NCBI (GenBank Accession Numbers CP168417-CP168435).

Declarations

Ethics approval and consent to participate

This study did not involve human participants, vertebrate animals, or the use of personally identifiable data.

Competing interests

The authors declare no competing interests.

Consent for publication

Not applicable.