Diversity in the Bacterial Genus Dickeya Grouping Plant Pathogens and Waterways Isolates

Jacques Pédron ^*, Frédérique Van Gijsegem ^*

Sorbonne Université, INRA, Institute of Ecology and Environmental Sciences-Paris (iEES-Paris), 75252 Paris cedex, France

* Correspondences: Jacques Pédron, Frédérique Van Gijsegem

Academic Editors: Ben Krause-Kyora, Sergey Pisarenko, Dmitry Kovalev and Alexander Kulichenko

Special Issue: Bacterial Genomes

Received: July 25, 2019 | Accepted: November 07, 2019 | Published: November 26, 2019

OBM Genetics 2019, Volume 3, Issue 4, doi:10.21926/obm.genet.1904098

Recommended citation: Pédron J, Van Gijsegem F. Diversity in the Bacterial Genus Dickeya Grouping Plant Pathogens and Waterways Isolates. OBM Genetics 2019; 3(4): 098; doi:10.21926/obm.genet.1904098.

© 2019 by the authors. This is an open access article distributed under the conditions of the Creative Commons by Attribution License, which permits unrestricted use, distribution, and reproduction in any medium or format, provided the original work is correctly cited.

1. Introduction

Soft rot Pectobacteriaceae and Dickeya – are Enterobacterales responsible for considerable economic losses in several important crops and ornamental plants and were listed as one of the top ten phytopathogenic bacteria [1]. Recent severe outbreaks in potato, one of the primary food crops worldwide, resulted from the action of a cohort of bacteria belonging to different Pectobacterium and Dickeya species in a complex population dynamics history [2]. These bacteria often exhibit a very broad host range and are able to infect both monocot and dicot plants [2,3]. Their hallmark is the ability to macerate plant tissues by disintegrating plant cell walls leading to cell lysis and liberation of the cell content. They do so by producing and secreting a battery of plant cell wall degrading enzymes (PCWDEs) that cause maceration of the plant tissue. Their virulence relies however on several other factors that allow them to adapt to environmental changes encountered in planta and to face the stresses produced by plant defense responses [4].

The genus Dickeya was formed in 2005 by the reclassification of former Erwinia chrysanthemi into six species as Dickeya chrysanthemi, D. dadantii, D. diffenbachiae, D. dianthicola, D. zeae and D. paradisiaca [5]. Subsequently, D. dieffenbachiae was reclassified as a subspecies of D. dadantii based on DNA-DNA hybridization analysis and MLSA [6]. Each of these species comprises strains isolated from various plant hosts including both dicots and monocots and does not possess real host specificity [5]. Three new Dickeya species have also been described: D. aquatica isolated from freshwaters in Europe [7], D. solani associated with severe outbreaks in the 2000s in potato in Europe and Israel [8,9] and D. fangzhongdai isolated from diseased pear trees but also from several monocots, including orchids, in Asia and Europe and from freshwaters [3,10,11]. Comparative analyses of these Dickeya proteomes allowed Duprey et al. [12] to infer the complex evolutionary history of the genus Dickeya.A recent systematic search for Dickeya in waterways led to the characterization of two new Dickeya species, D. lacustris [13] and D. undicola [14] that were isolated exclusively from freshwaters or from the plant rhizosphere at pond edges so far.

Thus, the genus Dickeya thus currently comprises ten characterized species that group bacterial isolates from various environments including monocot versus dicot plants, diseased plants versus freshwaters. The purpose of this study was to analyze the inter- and intra-specific diversity in genus Dickeya with a special emphasis on the two new species D. undicola and D. lacustris, isolated from waterways.

2. Materials and Methods

2.1 Genome Collection and Annotation

We downloaded the complete and draft genomes of all Dickeyas available as of March 2019 from the NCBI (https://www.ncbi.nlm.nih.gov/), except D. aquatica 174/2, whose complete genome was downloaded from the ASAP database (http://asap.ahabs.wisc.edu/asap/home.php). Coding sequences were predicted using the Rapid Annotation Subsystem Technology (RAST) server [15] using default parameters.

2.2 Species Delimitation

Average nucleotide identity (ANI) was computed using the pyani python module [16] with the BLASTp algorithm. In silico DNA-DNA hybridization (DDH) was carried out and analyzed according to Meier-Kolthoff et al. [17], using a dedicated pipeline (http://ggdc.dsmz.de/). Species threshold was set to 96% and 70% for ANI and DDH, respectively. Genome synteny was visualized by a dot plot using the D-Genies web site ([18], http://dgenies.toulouse.inra.fr/).

2.3 Phylogenetic Analysis

In silico multilocus sequence analysis (MLSA) was performed on 1191 concatenated amino acid orthologous sequences (82093 variable sites) of the 59 genomes. Pectobacterium atrosepticum 21A was used as an outgroup. The clustering of orthologous sequences into homologous families was conducted using the SiLix software package [19], using an identity threshold of 70% and a minimal overlap of 80%. Orthologous sequences were aligned using the Muscle software [20] and then concatenated. SeaView (version 4.6.5), a multi-platform program for molecular phylogeny, was used to implement specified tools and options [21]. Alignments were curated using Gblocks [22] and phylogenic analysis was performed using the BioNJ algorithm [23] with 100 bootstraps value. Mega7 tool [24] was used to visualize the phylogenetic tree.

For the pangenome clustering, hierarchical clustering was performed for the pan-genome as described by Meric et al. [25]. Briefly, a presence/absence matrix (0/1) for the pangenome was constructed. Manhattan distances were calculated and used for hierarchical clustering to generate the tree. The plotted distance between two genomes revealed the proportion of protein families where their present/absent status differs. Thus, in the pan-genome hierarchical clustering analyses, protein families that were not conserved vary in their presence or absence between genomes. Protein families occurring only in a single genome (singletons) were not included in the analysis. Mega7 tool [24] was used to visualize the phylogenetic tree with the BioNJ algorithm.

2.4 Assessment of Core Genome, Pan Genome, and Species-Specific Genes

Genome-to-genome comparisons were performed by bi-directional protein-protein BLAST sequence comparison of translated open reading frames (ORFs) with a 10⁻⁵ e-value threshold. Genes were defined using the RAST server [15] applying default parameters and without a minimum length cut-off. Orthologous sequences were clustered into homologous families using the SiLix software package [19], with a constraint of 70% amino acid identity and 80% overlap.

3. Results and Discussion

3.1 Contrasting Diversities among Dickeya Species

To investigate the diversity of the genus Dickeya, we analyzed 59 Dickeya genomes from databases, including the 49 genomes analyzed by Duprey et al. [12], two additional genomes of recently described D. undicola species [14] and the genome of the new D. lacustris species [13]. Our panel comprised 20 fully sequenced genomes and 39 draft genomes consisting of one to nearly 200 scaffolds (Table 1). A significant number of the analyzed strains were isolated from potato (Solanum tuberosum) diseased tissues (19 isolates) reflecting the important economic implications of the recent outbreaks that occurred in this crop in Europe and in the USA [2]. The other strains were collected from monocots or other dicots plants, from the rhizosphere of healthy potato plants (2 isolates) or from natural freshwaters (14 isolates). Geographically, the analyzed bacteria originated from all around the world but the distribution was not uniform with half of them isolated from Europe and only one isolated from Comores islands, Africa (Table 1).

As already described [26,12], a whole-genome MLSA analysis of the 59 studied genomes confirmed that all strains fit in one of the ten characterized Dickeya species except Dickeya sp. NCPPB569 that was proposed to be a new species [12] (Figure 1).

Based on MLSA analysis, Dickeya species could be further organized in different branches, clade I grouping D. chrysanthemi, D. zeae and Dickeya sp. NCPPB569, clade II grouping the new D. undicola species with D. solani, D. dadantii, D. fangzhongdai and D. dianthicola, a third clade grouping D. lacustris/D. aquatica clade, and the D. paradisiaca branch (Figure 1).

We analyzed the relatedness between strains of similar species by calculating the two-by-two average nucleotide identities (ANI) for each Dickeya genome (Figure 1) and highlighted contrasting situations. Bacteria belonging to D. aquatica, D. dianthicola and D. undicola group shared around 99% or more ANI among them (Figure 1), revealing very high sequence conservation. However, only three genomes of D. aquatica and D. undicola are available until now and they may not reflect the complete diversity of these two species. The high conservation observed in these three species does not correlate to similar geographical origins or similar hosts. Indeed, if all D. dianthicola strains (except one) were isolated from diseased potato, they originate from locations all over the world (Table 1). Similarly, while the three D. undicola strains were isolated from freshwaters, they were from either France or Malaysia.

Table 1 General features of the different Dickeya strains/genomes analyzed in this study.

Figure 1 Phylogeny of the Dickeya genus. A. Phylogenic tree built up from the concatenated sequences of 1191 homologous protein sequences (82093 variable sites). One hundred bootstrap replicates were performed to assess the statistical support of each node. Only one node (labelled ◇) was below 100% (bootstrap value 49%). Pectobacterium atrosepticum 21A was used as an outgroup. The numbers of protein families common to all genomes of the different clades/nodes are written in the left part. ANI values ranges for each species are given. B. Pangenome tree: distance was calculated from a presence/absence matrix of the pangenome (see Materials and Methods). Origins: M: monocot; D: dicot; R: potato rhizosphere; : potato; : water. The stars indicate differences between the phylogenic and pangenome trees.

Similarly, no link was observed between plant host origin and genome variability. For instance, as already reported [27,28,29], most D. solani strains exhibiting very high ANI values above 99.9%, were isolated either from potato or from ornamentals, pointing to a clonal origin and possible contamination of potato from nearby fields planted with ornamentals [30]. However, one D. solani genome, RNS05.1.2 was found to be more divergent with an ANI value of about 98% with the other D. solani genomes [28,29]. This divergence is not related to a different host or geographical origins because, like other D. solani strains of our panel, this strain was isolated from potato in France [28].

Strains grouped in the D. fangzhongdai, D. zeae, D. dadantii and D. chrysanthemi species were more diverse and, in some cases, fell below the limits of species definition, now generally recognized as 95%–96% for average identity value (ANI) and 70% for digitally derived DNA-DNA hybridization (dDDH) [31,32]. While all D. dadantii and D. chrysanthemi strains harbored more than 96/70% ANI/dDDH values between genomes of the same species, these values were at the limits of species definition for some D. fangzhongdai or even fell below these limits for D. zeae genomes. In both cases, a further subcladal organization was observed in these species and the distribution of the analyzed genomes could be assigned to three branches, two subclades grouping genomes that share ANI values ranging from 97% to 99%, and one divergent strain (NCPPB3274 in D. fangzhongdai and 586 in D. zeae) (Figure 1) [33]. The ANI/dDDH values of genomes belonging to the two D. zeae subclades were as low as 94/58.2% (Table S1), indicating the separation of these two lineages into two distinct subspecies or even separate species. Further phenotypic characterizations and analysis of more genomes are needed to solve this ambiguity.

3.2 Dickeya Species Share a Large Panel of Genes

In order to further address the relationships between the different Dickeya species, we carried out a pan-genome analysis of this genus and compared the protein-coding sequences of the 59 genomes using the SiLix gene family clustering tool. Proteins were classified as homologous to others in a given family if the amino acid identities were above 70% with 80% minimal overlap. Bacterialpan-genomes could be divided into three gene classes as i) the core genome that groups protein families shared by all members of the analyzed genus/species, ii) the specific genome gathering protein families unique to one genome or a genome group, iii) the accessory genome including protein families present in more than one but not all genomes of a group. The genus Dickeya core genome comprises 1800 genes representing nearly 40% (37.5% to 43%) of a single genome content (Figure 1). These common genes encompass most of the genes shown to be involved in virulence (for a review, [4]).The hallmark of virulence of the soft rot causing Dickeya is the production and secretion of a battery of plant cell wall degrading enzymes, such as pectinases, cellulase, and proteases. The out type 2 secretion system (T2SS) drives the secretion of nine pectate (PelABCDEILNZ), one pectin (Pnl), one or two rhamnogalacturonate (RhiE, RhiEbis) lyases and various enzymes that strip off the pectin decorations (pectin methylesterases PemA and PemB, acetyl esterase PaeY) as well as the cellulase CelZ. As recently described [12],the corresponding encoding genes are largely conserved in the Dickeya genus (Table 2). Like all members of clade II, D. undicola possesses all PCWDE-encoding genes except the gene encoding the second rhamnogalacturonate lyase. A notable exception is PelA whose gene is truncated in all D. dianthicola genomes. Likewise, D. lacustris harbors almost the same PCWDE battery as its closest relative D. aquatica, except that it has only one copy of the pelD/E genes and possesses the Pel10 encoding gene otherwise present only in D. fangzhongdai and some D. chrysanthemi strains. D. paradisiaca and Dickeya sp. 569 were found to be less equipped, and they both lacked PelA, PelI, Rhi, and PemB. In addition, PelZ, and PnlG were absent in Dickeya sp. 569 while D. paradisiaca harbored only one copy of the PelD/E isoenzymes and lacked PaeY acetyl esterase.

Table 2 Conservation of genes encoding pectinolytic enzymes.

All Dickeya species except D. paradisiaca were found to possess a Prt T1SS that allows the secretion of up to four metalloproteases (PrtABCG). The genes encoding the three PrtABC proteases are also largely shared while the prtG gene is absent in D. dianthicola, D. undicola, D. chrysanthemi and Dickeya sp. NCPPB569 (Table 3).

Four other protein secretion systems (T3SS to T6SS) might be present in gram-negative bacteria where they are involved in interactions with eukaryotic cells or in bacterial intra- and inter-competitions [34]. All four TSS were present in the D. dadantii 3937 model strain but T5SS comprised very long proteins truncated in the draft genomes and difficult to tackle. So, we excluded T5SS from our study. The different Dickeya species possess variable sets of T3SS, T4SS, and T6SS; D. paradisiaca and Dickeya. sp. NCPPB569 do not have any; D. lacustris and D. aquatica possess only T3SS, and for the other species, the presence and number of these TSS are variable (Table 3).

Upon plant colonization, bacteria encounter various stressful environmental conditions like the low pH in apoplast, as well as osmotic and oxidative stresses. The D. dadantii 3937 model strain is well equipped to cope with these stresses [35].Like many other bacteria, it responds to oxidative stress by producing antioxidant enzymes like catalases (KatE, KatG), superoxide dismutases (SodA, SodC) and alkylhydroperoxide reductase (AhpCF) as well as repair systems like the peptide methionine sulfoxide reductase (MsrA), an enzyme repairing oxidative damage caused to proteins [36],or the Suf and Isc systems, involved in the repair of damaged Fe/S clusters containing proteins [37,38].It also produces the ROS scavenging pigment indigoïdine, It can also resist antibacterial peptides produced by plants through the Sap system and detoxify nitric oxide produced by plants during infection with the nitric oxide dioxygenase HmpX [39,40]. Most of the encoding genes are well conserved in different Dickeya species (Table 3). However, the suf gene cluster is present only in clade II as well as in D. chrysanthemi, HmpX is truncated in D. undicola and absent in Dickeya sp. 569 and D. paradisiaca. The indigoïdine encoding genes are absent in D. paradisiaca and D. lacustris. (Table 3), although strikingly, D. lacustris was more resistant to oxidative stress than D. dadantii despite the lack of indigoïdine production [13].

Table 3 Conservation of virulence-related factors.

At the species level, core genomes are very large comprising nearly 3000 to 4000 protein families, i.e., around two-thirds of the genetic content in a strain (Figure 2).

This high genetic conservation persists in clades that group different species. Indeed, the D. solani/D. dadantii clade, D. undicola/D. fangzhongdai clade and D. lacustris/D. aquatica clade also share around 3000 protein families (Figure 1). The D. zeae/D. chrysanthemi/Dickeya sp.NCPPB569 clade I is less related because it shares only around 2500 protein families, similar to the number of protein families shared by all five clade II Dickeya species (Figure 1).

Therefore, even at the genus level, Dickeya members share a very large pool of genes as compared to other Enterobacterales species. For comparison, using less stringent conditions (50% identity on 50% of the length of the proteins) the E. coli core genome includes only around 1500 orthologous genes [41]. Furthermore, most virulence determinants are shared by most Dickeya species with the notable exceptions of D. paradisiaca and the Dickeya sp. 569 strain.

Figure 2 Conserved and strain-specific gene content in the different Dickeya species. The protein families present in the core, accessory (in brackets) and strain-specific (in petals) are depicted.

3.3 Accessory Genome and Strain-Specific Genes

The sizes of the accessory genome of species, which groups protein families present in more than one but not all genomes of a given species, ranged from about 400 to more than 2100 protein families (Figure 2). To analyze them further, we performed a pan-genome clustering analysis that builds a hierarchical clustering based on the proportion of the presence/absence status of each gene family in each genome (Figure 1B). The allocation in species determined by the whole genome MLSA analysis is conserved reflecting the high number of protein families present in the core genome of each species. Strikingly, however, the relatedness between Dickeya species at the pan-genome level is quite different.

For instance, D. chrysanthemi did not cluster with D. zeae anymore but was now closer to clade II species, D. undicola fell outside clade II and D. paradisiaca was closer to P. atrosepticum 21A than to any Dickeya species (Figure 1B). At the species level, the distribution of D. zeae strains into two branches as per the MLSA and ANI analyses was conserved in the pan-genome distribution, further substantiating the separation into two subspecies/species. No correlation was observed between this pan-genome clustering and habitat or geographical origin (Figure 1B and Table 1) except in D. undicola (Figure 1B). Indeed, in D. undicola, the two strains isolated from France were closer to each other than to the 2B12 strain in the pan-genome clustering, while in the MLSA core genome analysis, D. undicola FVG10 was closer to 2B12 than to the other strain from France. Although the current analysis was based on only three genomes, it points to a possible closeness of the accessory genomes in relation to their geographical origin.

We also determined the number of protein families specific to each analyzed genome, and absent from all other Dickeya genomes considered in this study (Figure 2). We found that D. solani strains harbored very few strain-specific protein families (except 05.1.1.2), reflecting the already reported clonal origin of these strains [27,28,29,30]. Three D. dianthicola strains (DE440, WV516, and ME23) also possess only a few dozen strain-specific genes coupled with very high ANI values (> 99.9%) among them. Interestingly, these three strains were isolated from potato in the USA, pointing to the same clonal origin as reported for the recent blackleg outbreak devastating potato crops in this country [2]. For the other strains, the number of strain-specific genes ranged from a few dozen to several hundreds. Further, we did not observe any link between the proximity of the core genomes revealed by ANI values in a given species and the extent of the strain-specific genome. For example, the members of the D. dianthicola and D. undicola species exhibited ANI values above 99% within a given species but harbored up to several hundred strain-specific genes, in the same range as D. dadantii members that shared only 96.7% ANI among them.

3.4 Features Specific to Each Dickeya Species

To tackle the functions that might be specific to each Dickeya species, we identified the species-specific genes present in all members of one species and missing in all other Dickeya genomes analyzed in this study (corresponding proteins harboring less than 70% identity). The number of species-specific genes ranged from 20 to 560 gene families (Table 4), although most of these species-specific genes encode hypothetical proteins or proteins with undefined function.

This may be due to the fact that the RAST CDS definition program considers very short proteins (less than 30 aa) for which no annotations are available. Expression profiling analyses of D. solani and D. dianthicola strains,however, revealed that most of the small CDS are expressed and may be differentially expressed in macerated tubers versus their growth in rich medium [42]. Annotation of the species-specific genes is listed in Table S2. Out of the five D. undicola genes with a known or predicted function, two genes encode regulatory proteins, and interestingly, a few (1 to 7) species-specific regulatory genes are present in all Dickeya species except D. dadantii. Several species-specific genes clustered in genomic regions, grouping at least five genes. D. aquatica harbors seven such genomic regions that group 5 to 33 genes. These regions mainly contain genes encoding hypothetical proteins. Further, GR1 encodes an ABC-transport system related to nickel transporters and GR2 harbors genes related to secondary metabolism (Table S2). Most D. chrysanthemi species-specific genes with a known function are grouped in a genomic region of 26 genes that contain the Flp/Tad operon and its associated two-component regulatory system (Table S2). The genes in this operon encode the synthesis of a type IVb pilus shown to be involved in biofilm formation and virulence in several bacteria including Pectobacteria [43].

Table 4 Dickeya species and D. aquatica/D. lacustris clade specificities.

In animal pathogens, diversity in virulence at the species level is often related to the absence instead of the presence of genes, suggesting that high virulence is related to specialization as it was described for Yersinia pestis [44]. We thus examined if species specificity may reside in the absence of specific protein families that are present in all other Dickeya species. Very few such families are present in Dickeya species, less than ten in all, except in D. paradisiaca that lacks 124 such genes (Table S2). As already described above, those families absent from D. paradisiaca include several proteins associated with virulence like those encoding the Prt type 1 secretion system of proteases, the pectin acetylesterase PaeY, a catalase, part of the achromobactine siderophore biosynthesis pathway as well as several genes related to metabolism and regulation (Table S2). Strikingly, D. paradisiaca strains possess all genes encoding proteins involved in flagellum biosynthesis, although, these genes are more related to orthologous genes in Lonsdalea and Brenneria than to their Dickeya counterparts.

3.5 The D. aquatica/ D. lacustris Clade

The D. aquatica/D. lacustris clade is the only clade grouping strains isolated only from water, hence, we analyzed the genomic specificities of this clade. Members of this clade share 239 genes that are absent in all other Dickeya species (or exhibit less than 70% identity) (Table 4). Besides genes encoding hypothetical proteins or proteins with undefined function (124 genes), these clade-specific genes could be divided mainly into categories related to metabolism (45 genes), transport (14 genes), regulation (6 genes), and resistance to stress (9 genes) (Table S3). Several of these genes grouped into nine genomic regions (GR) defined as regions clustering at least five genes (Table 5). Five of these genomic regions include genes involved in interesting pathways/features.

Table 5 Genomic regions present only in the D. aquatica/D. lacustris clade.

GR1 encodes genes involved in the transport and catabolism of N-acetyl-glucosamine, the monomer of chitin present in the cuticles and shells of insects and crustaceans and in fungal cell walls. The genes of GR2 and GR3 are predicted to be involved in capsule biosynthesis and LPS export indicating specificities in the bacterial cell envelope. GR5 genes are similar to the fim genes involved in the biosynthesis of type 1 fimbriae that confer adhesion to a variety of eukaryotic cells. In uropathogenic E. coli, these fimbriae play a critical role both in the colonization of the lower urinary tract and in the formation of biofilms by intracellular bacteria after the invasion of epithelial cells [45]. GR6 encompasses genes homologous to the Ter stress response gene cluster involved in tellurium resistance and that are activated during pathogenesis in Yersinia [46].

Likewise, 54 genes absent in the strains of the D. aquatica/D. lacustris clade were present in all other Dickeya species (Tables 4 and Table S3). Among them, only five were annotated as hypothetical proteins and several of these genes encode proteins involved in transport systems or metabolism. In particular, D. aquatica/D. lacustris lacks the gene cluster encoding proteins involved in the metabolism of xylose available in plant xyloglucans, and thus, these bacteria are not able to use xylose as a carbon source [12,13].

For comparison, we performed a similar analysis with the members of the D. zeae/D. chrysanthemi/Dickeya sp. 569 clade I. This clade shares only nine genes that are absent in all other Dickeya species and conversely, no genes are absent in its members while present in all other Dickeya.

Therefore, although D. aquatica and D. lacustris were isolated from water until now, they are well equipped for invading plants as they possess most of the virulence-related genes (see 3.2) and, contrary to D. paradisiaca, none of the genes absent in these species are known to be involved in interactions with plants, except the genes involved in xyloglucan catabolism. Accordingly, D. aquatica was shown to be capable of macerating acidic fruits like tomato or cucumber [12].

3.6 Does D. paradisiaca belong to the Dickeya Genus?

Contrary to the definition of bacterial species, the delimitation of genus adjuncts is still a matter of debate. One of the proposed limits is an average nucleotide identity above 80% [32]. In the case of D. paradisiaca, the ANI values ranged from 78 to 80% with the other Dickeya species (at or below this limit), D. aquatica/D. lacustris beingthe more distant clade. Furthermore, D. paradisiaca lacks several enzymes involved in pectinolysis as well as virulence determinants (see 3.2). It was however considered as the earlier branching lineage in the Dickeya genus in evolution studies that analyzed the emergence of the different virulence-related genes [47,12].Finally, in our pan-genome analysis (Figure 1B), D. paradisiaca clusters with P. atrosepticum, outside the Dickeya genus.

To further analyze the relatedness of D. paradisiaca to the Dickeya genus, we performed synteny analyses taking advantage of the availability of complete genomes in our panel (Figure 3).

Figure 3 Synteny between D. paradisiaca and the other Dickeya clades.

The synteny of clade II members of Dickeya species is very high as already described using the MAUVE program [27,42].This high synteny was conserved between the clade I and clade II clusters that show collinearity with only a few inversions of genomic regions, as exemplified by the D. zeae EC1/D. solani IPO2222 synteny (Figure 3A). In D. aquatica 174/2, although the inversed regions were larger compared to these clusters the syntenic regions were still dense (Figure 3B). In contrast, the syntenic regions between D. paradisiaca NCPPB2511 and members of clades I and II were more scattered (Figure 3C). Comparison between D. paradisiaca and D. aquatica gave similar results (Figure 3D). For comparison, we also analyzed the synteny between Dickeya and the Pectobacterium related genus exemplified by a P. atrosepticum strain (Figure 3E and Figure 3F). The homologous regions were even more scattered despite a quite similar overall organization of D. zeae and P. atrosepticum. This analysis revealed a gradient of synteny between the members of the genus Dickeya with a very high closeness between the genomes of species belonging to clade I and clade II whileless homologous regions were observed in D. paradisiaca, and an intermediate synteny was revealed in D. aquatica. Although D. aquatica wasproposed as the second deepest branching lineage in the genus Dickeya, it was not more syntenic to D. paradisiaca than to the other Dickeya species. Nevertheless, our analyses only comprise two D. paradisiaca genomes (the only ones available) and further phenotypic, as well as genomic analyses of a larger number of strains, should be carried out before proposing a possible categorization of this bacterial group into different genera.

4. Conclusions

One hallmark of the genus Dickeya is the high genetic closeness among its members. Indeed, despite using high stringency parameters (70% identity with minimal overlap of 80%), all genomes analyzed shared around as much as 1800 protein families and this was also true for D. paradisiaca that is at the limit of affiliation to the genus Dickeya. At the species level, very few gene specificities exist and even fewer specific gene losses frequently found in specialized human pathogens. This, accompanied by high conservation of the characterized virulence-related genes in most species, may account for the broad host range of the different Dickeya species that can infect both monocot and dicot plants. Nevertheless, such a high relatedness in genetic content does not necessarily imply similarity in strategies to adapt to the plant hosts. In fact, transcriptomic studies during maceration of potato tubers revealed differences in gene expression even in very close Dickeya species like D. solani and D. dianthicola [42]. Likewise, pangenome analysis does not highlight strain clustering according to host or environment of their origin, further substantiating the lack of specialization to a given environment. Among the 239 protein families present only in the quite distant D. aquatica/D. lacustris clade that groups the strains isolated only from water, we did not find any genetic trait that might contribute to habitat specificity. Instead, the members of this clade shared two gene clusters that play a role in the interactions of animal pathogens with their hosts and the other one facilitates interaction with insects and fungi. No protein family was shared by the three Dickeya species found only in water so far, corroborating the absence of “water-specific” genes already reported in D. fangzhongdai water isolates [33]. None of the protein families shown to play a role in interactions with plants were found missing in the D. aquatica/D. lacustris clade. Hence, their spread via waterways, specifically irrigation water might be a potential threat for economically important crops.

Earlier studies, by compiling hundreds of strains previously identified as Erwinia chrysanthemi, defined seven Dickeya species [3,26,48]. In addition, the strain NCPPB569 (isolated from sugar cane plantations in Australia) is significantly distinct to belong to an additional new species [3,26,48]. Recent studies in aquatic environments allowed the identification of three new Dickeya species that were found only in water so far. It would be interesting to extend the search for Dickeya isolates to new environments, for instance, insects that are recognized as hosts for some Dickeya [49] to explore if this will further broaden the diversity of genus Dickeya.

Acknowledgments

We thank Nicole Hugouvieux-Cotte-Pattat for sharing the D. lacustris genomic sequence before publication and for fruitful discussions.

Additional Materials

The following additional materials are uploaded at the page of this paper.

1. Table S1: Relatedness of the Dickeya species. in silico DNA-DNA hybridization (DDH, upper triangle) and Average Nucleotide Identity (ANI, lower triangle) values are presented. The specific threshold value is 70% for DDH and 96% for ANI. Strains fulfilling these thresholds are highlighted in tangerine.

2. Table S2: Dickeya species-specific and species specifically absent gene families.

3. Table S3: Gene families specific to and specifically absent from the D. aquatica/D. lacustris clade.

Author Contributions

JP performed the bioinformatics analyses. Both authors analysed the data and wrote the manuscript.

Funding

This research was supported by the COMBICONTROL ((ANR-17-CE32-0004-04) and SPREE (ANR-17-CE32-0004-04) projects financed by the French National Agency for Research.

Competing Interests

The authors have declared that no competing interests exist.