Pseudomonas viridiflava: An internal outsider of the Pseudomonas syringae species complex

Abstract Pseudomonas viridiflava is a gram‐negative pseudomonad that is phylogenetically placed within the Pseudomonas syringae species complex. P. viridiflava has a wide host range and causes a variety of symptoms in different plant parts, including stems, leaves, and blossoms. Outside of its role as a pathogen, P. viridiflava also exists as an endophyte, epiphyte, and saprophyte. Increased reports of P. viridiflava causing disease on new hosts in recent years coincide with increased research on its genetic variability, virulence, phylogenetics, and phenotypes. There is high variation in its core genome, virulence factors, and phenotypic characteristics. The main virulence factors of this pathogen include the enzyme pectate lyase and virulence genes encoded within one or two pathogenicity islands. The delineation of P. viridiflava in the P. syringae complex has been investigated using several molecular approaches. P. viridiflava comprises its own species, within the complex. While seemingly an outsider to the complex as a whole due to differences in the core genome and virulence genes, low average nucleotide identity to other of P. syringae complex members, and some phenotypic traits, it remains as part of the complex. Defining phylogenetic, phenotypic, and genomic characteristics of P. viridiflava in comparison to other P. syringae members is important to understanding this pathogen and for the development of disease resistance and management practices. Taxonomy Kingdom Bacteria; Phylum Proteobacteria; Class Gammaproteobacteria; Family Pseudomonadaceae; Genus Pseudomonas; Species Pseudomonas syringae species complex, Genomospecies 6, Phylogroup 7 and 8. Microbiological properties Gram‐negative, fluorescent, aerobic, motile, rod‐shaped, oxidase negative, arginine dihydrolase negative, levan production negative (or positive), potato rot positive (or negative), tobacco hypersensitivity positive. Genome There are two complete genomes, five chromosome‐level genomes, and 1,540 genomes composed of multiple scaffolds of P. viridiflava available in the National Center for Biotechnology Information Genome database. The median total length of these assemblies is 5,975,050 bp, the median number of protein coding genes is 5,208, and the median G + C content is 59.3%. Disease symptoms P. viridiflava causes a variety of disease symptoms, including spots, streaks, necrosis, rots, and more in above‐ and below‐ground plant parts on at least 50 hosts. Epidemiology There have been several significant disease outbreaks on field and horticultural crops caused by P. viridiflava since the turn of the century. P. viridiflava has been reported as a pathogen, epiphyte, endophyte, and saprophyte. This species has been isolated from a variety of environmental sources, including asymptomatic wild plants, snow, epilithic biofilms, and icepacks.


| INTRODUC TI ON
Pseudomonas viridiflava is a species in the Pseudomonas syringae species complex, an amalgam of closely related pseudomonads that altogether comprises nine genomospecies, 13 phylogroups, and 60 pathovars with a vast host range Gardan et al., 1999). Originally isolated in Switzerland from dwarf bean (Phaseolus vulgaris) with reddish-brown lesions on the pods in 1930 (Billing, 1970), P. viridiflava has been shown to infect a range of both monocot and dicot hosts; it has been reported to cause disease in over 50 hosts (Table 1). Since 2000, there have been at least 13 outbreaks of diseases on annual plants caused by P. viridiflava; this makes up 18% of all outbreaks caused by the entire P. syringae species complex worldwide (Lamicchane et al., 2015). Diseases caused by P. viridiflava have been reported in countries such as Saudi Arabia, New Zealand, Italy, Hungary, Spain, and Greece, among others (Gonzáles et al., 2003;Morretti et al., 2005;Sarris et al., 2012;Taylor et al., 2011;Végh et al., 2012). In addition to its role as a crop pathogen, P. viridiflava also acts as an endophyte, epiphyte, and saprophyte in both agricultural and natural environments. Genomic and phenotypic characteristics of P. viridiflava make it an "internal outsider" within the P. syringae species complex. For example, while P. viridiflava is lumped within the P. syringae complex, it has distinct characteristics, including pectate lyase as a virulence factor, atypical pathogenicity islands, and phenotypic phase variation (Araki et al., 2006;Bartoli et al., 2014;Liao et al., 1988). Delineation of P. viridiflava, among other species within the P. syringae species complex, has been a major research focus in the past decade (Baltrus et al., 2017;Berge et al., 2014;Bull & Koike, 2015;Dillon, Thakur, et al., 2019).
Genetic relationships have been investigated through techniques such as DNA-DNA hybridization, comparisons of 16S rRNA and housekeeping gene sequences, molecular fingerprinting, and more recently comparisons of whole-genome sequences (Anzai et al., 2000;Berge et al., 2014;Dillon, Thakur, et al., 2019;Gardan et al., 1999). Another recent focus of P. viridiflava research is understanding the genetic and phenotypic variability within the species (Bartoli et al., , 2015. Intraspecies phylogeny and pathogenicity island variability have been investigated (Araki et al., 2006(Araki et al., , 2007Bartoli et al., 2014). For phenotypic variability, research has focused on soft rotting potential and the two main colony phenotypes, transparent and mucoid, which exhibit varied pathogenic and antibiotic-resistant phenotypes (Bartoli et al., , 2015. This pathogen profile will highlight the research regarding current knowledge and advances in (1) delineating P. viridiflava in the P. syringae species complex, (2) lifestyle and epidemiology, (3) host range, (4) virulence, and (5) suggestions for future research on P. viridiflava.

| Taxonomy
A goal of recent P. syringae species research is to clarify the taxonomic delineation of associated species and pathovars, including P. viridiflava. Currently, P. syringae is referred to as a species complex, defined as a cluster of related monophyletic groups based on historical trends in bacterial classification initially based on phenotypes and progressively based on genotypes including DNA-DNA hybridization and phylogenetic analysis of housekeeping genes sequences . In the past, the pseudomonads were differentiated using the LOPAT profile test (levan production, oxidase production, pectinolytic activity, arginine dihydrolase production, and tobacco hypersensitivity) and DNA-DNA hybridization TA B L E 1 All plant hosts in which Pseudomonas viridiflava has been reported to cause disease in natural infections or in experimental conditions

| Intraspecies phylogeny
Currently, the P. syringae species complex comprises nine genomospecies, 13 phylogroups, and 64 pathovars. A genomospecies within the P. syringae complex is defined by DNA-DNA hybridization ability; through this method, P. viridiflava was determined to be a distinct species from other members of P. syringae sensu lato (Gardan et al., 1999).
P. viridiflava strains do not hybridize with other species in the P. syringae complex (Gardan et al., 1999). In a study by Gomila et al. (2017) including type strain ICMP 2848, P. viridiflava had an average nucleotide identity of less than 97% compared to other members of the species complex, which is below the accepted threshold for species separation (Goss et al., 2005). Phylogroup delineations were determined by Berge et al. (2014) using the four-gene MLSA classification schema developed by Hwang et al. (2005) Gardan et al., 1999). The majority of named "P. viridiflava" strains group within PG7a and are typically isolated from a variety of environmental sources .
Some key features of PG7 are the soft-rotting capability on potato tubers, phenotypic phase variation, and the presence of a noncanonical type III secretion system (T3SS) . PG8 shares the key features described for PG7, but it is differentiated by the production of a toxin in bioassays with Geotricum candidum . P. viridiflava exhibits high intraspecific genetic variation. In the paramount study by Bartoli et al. (2014) assessing the intraspecific variability of P. viridiflava, the genetic diversity of strains was characterized per the structure and sequences of the pathogenicity islands (PAIs), critical portions of the genome for virulence. Not only did strains contain divergent types of PAIs, both single-partite (S-PAI) and tripartite (T-PAI), but several strains of the S-PAI type contained an exchangeable effector locus that was not present within other S-PAI types, but present in all T-PAI types . P. viridiflava strains contain higher diversity in PAI structure compared to P. syringae sensu stricto strains, which typically only contain one type of canonical T-PAI (Dillon, Thakur, et al., 2019). Genetic diversity was also previously demonstrated in a population of P. viridiflava strains isolated from Arabidopsis thaliana from various geographic locations (Goss et al., 2005). There was substantial variation in the five genomic fragments examined, with an average of 33.4% synonymous site nucleotide divergence between two clades defined within the population, and 9.3% synonymous site nucleotide divergence within a single clade. This variation was not correlated with differences in geographic location of isolates. In another study, the hypothesis that intraspecific genetic variation of P. viridiflava is not due to host-specific adaptation was supported, given by divergent clade groupings in which strains
High rates of HGT in PG7 strains, as well as strains from other phylogroups, may contribute to its intraspecific diversity. The evolutionary potential of P. viridiflava due to high rates of HGT and occupation of varied niches may contribute to the intraspecific diversity.
Although MLSA and molecular fingerprinting techniques, including rep-PCR, have been useful tools for exploring intraspecies diversity in the past, the increase in affordability and accessibility of whole-genome sequencing will soon be the gold standard for delin- While most strains (PG1, PG2, PG3, PG4, PG5, PG6, PG10) group into the primary group, P. viridiflava strains of PG7 grouped within the secondary group. PG8 isolates of P. viridiflava were not included in this analysis, but probably can be considered members of the secondary phylogroup due to genetic and phenotypic similarities to PG7 strains.
[Correction added on 23 September 2021, after first online publication: The type strain in the cited study by Gomila et al. (2017) in Section 2.2 has been corrected in this version.]

| EPIDEMI OLOGY AND LIFE S T YLE
P. viridiflava is widely distributed and plays various roles in the environment. This species has been reported as an epiphyte, endophyte, saprophyte, and pathogen on a variety of agricultural and wild plant hosts Bordjiba & Pruner, 1989;Samad et al., 2017).
P. viridiflava has also been commonly isolated from nonplant environmental reservoirs, such as snowpack, rain, epilithic biofilms, and lake water Pietsch et al., 2017). Like P. syringae, some P. viridiflava strains have ice nucleation capabilities, which supports their potential relevance in nonagricultural environments such as the water cycle. Studies show P. viridiflava isolates having ice nucleation capabilities in roughly 33%-45% of strains tested Pietsch et al., 2017). Although widespread, its broad distribution does not seem to follow a particular geographical pattern or structure . Because members of the P. syringae species complex, including P. viridiflava, are seemingly ubiquitous throughout various environments yet diverse in population structure, it is conceivable to place P. viridiflava into the ecotype model by Cohan (2002), which emphasizes the role of recurrent selective sweeps in defining the niche of distinct populations of bacteria (Baltrus et al., 2017). Therefore, the existence of epiphytic, endophytic, and pathogenic states could be considered ecotypes of P. viridiflava shaped by certain environmental conditions and selection pressures. Goss et al. (2005) hypothesize that because P. viridiflava is often deemed a weak or opportunistic pathogen, it could experience selection pressure in its epiphytic phase that the pathogenic ecotypes do not. P. viridiflava has demonstrated various roles in the microbial community as an epiphyte and endophyte. Some P. viridiflava isolates have been shown to produce a family of antimycotics, called ecomycins, that have significant bioactivity against both human and plant-pathogenic fungi (Miller et al., 1998). Presumably, this capability could be important to establishing as a plant epiphyte or pathogen when encountering fungal competitors. P. viridiflava has also been identified as an endophyte of weeds with the capability of herbicidal activity (Samad et al., 2017). The P. viridiflava strain CDRTc14, originally isolated as an endophyte in a vineyard in Australia, significantly inhibited seed germination and root growth of the weed Lepidium draba in greenhouse conditions (Samad et al., 2016). The CDRTc14 genome contained abiotic stress tolerance genes, such as genes for heavy metal and herbicide resistance, but it did not contain a complete pathogenicity island or pathogenicity phenotype typical of pathogenic P. viridiflava strains. The ability of P. viridiflava to act as an endophyte, saprophyte, and pathogen supports the idea that P. viridiflava, like many other members of the complex, is a generalist rather than a specialist. Its ability to infect a wide range of hosts corroborates its validity as a generalist pathogen (Goss et al., 2005;.
P. viridiflava is responsible for 13 economically relevant disease outbreaks on annual plants since 2000  and has been reported to cause disease on over 50 hosts since its discov-  (Jakob et al., 2002;Lamicchane et al., 2015). Ice nucleating properties of P. viridiflava may be beneficial to create frost wounds on the host, which can serve as an entry point for the bacteria (Lindow et al., 1982;Varvaro & Fabi, 1992).
An extensive study was performed on the epidemiology of P. viridiflava causing kiwifruit blossom blight over two decades ago.
Interestingly, data from this study showed that weather variables (air temperature, surface wetness, rainfall, and relative humidity) did not seem to affect development of disease while timing of the infection before the budding phase was critical for disease (Everett & Henshall, 1994 (Canaday et al., 1991). Most recently, P. viridiflava was reported to cause bacterial stem blight disease of alfalfa along with P. syringae PG2 strains (Lipps et al., 2019). The mechanism of synergy is currently unknown in this system. Overall, the synergistic potential of P. viridiflava with other microbes may be a factor in its ability to cause disease in certain situations.

| HOS T R ANG E
P. viridiflava has a wide natural and experimental host range. This species has also been isolated as an endophyte and epiphyte of wild plants as well as from various environmental sources. Here, the currently known host range of P. viridiflava as a pathogen is summarized, including data from natural hosts and explicit host range studies (Table 1).  (Liao, 1991;Liao et al., 1988). Mutant strains with a defective pel gene resulted in no leaf maceration after infection on Arabidopsis (Jakob et al., 2007). Pectate lyase activity has been shown to differ based on the type of PAI of the bacterial strain; single-PAI isolates exhibited twofold higher enzyme activity than tripartite-PAI isolates on Arabidopsis, even though the pel gene is encoded outside of the PAI (Jakob et al., 2007). The production of pectate lyase may be considered a significant biological difference between P. viridiflava and other members of the P. syringae species complex. Although the soft rotting phenotype is unique to P. viridiflava within the P. syringae complex, a phenotypic study showed that 8% of sampled P. viridiflava strains were not able to induce soft rot on potato tubers, therefore soft rot may be used as a general descriptor for the species, but not a diagnostic trait .

| Phase variation and mutability
An important discovery regarding P. viridiflava is the phenotypic plasticity of pathogenicity-related traits. Historically, "levan-production negative" was a characteristic of the typical LOPAT profile of P. viridiflava. However, yellow mucoid, levan-positive bacterial colonies originally isolated from bean, kiwifruit, and lettuce were identified as an atypical form of P. viridiflava (Gonzáles et al., 2003). More recent discoveries of levan-production positive P. viridiflava isolates, which also display yellow, mucoid growth on King's B medium, are evidence for phase variation within the species; in fact, 56% of P. viridiflava strains tested in a study by Bartoli et al. (2014) were levanproduction positive. Thus, the current knowledge of P. viridiflava phenotypic variability is that there are two phase variants of isolates: a yellow, mucoid, levan-positive variant and a transparent, flat, levan-negative variant. Interestingly, isolates can switch between the variant phenotypes, and the variants correlate with pathogenic potential. In a pathogenicity study of 11 mucoid strains and 11 transparent strains stably cloned from the same original 11 isolates, the mucoid variant could induce soft rot on potato tubers, while the nonmucoid variant could not . Also, wild-type (defined as whichever of the two variants naturally occurred in original isolate) and mucoid variant isolates were able to induce disease on bean stems (Figure 1), while the transparent variant could not (Bartoli et al., 2015). Although phase variation in P. viridiflava could be linked to pathogenic potential, there may be other advantages to possessing this type of plasticity. The presence of exopolysaccharide could increase bacterial tolerance to plant defences, or the pectinolytic capability of the mucoid strains could be important to bacterial colonization via release of sugars . More recently, it was reported that the transparent variant has a mutator phenotype and general antibiotic resistance in additional to low pathogenic potential on bean (Bartoli et al., 2015). Conversely, the mucoid variant did not show mutability or antibiotic resistance potential but did effectively cause disease in bean. Though P. viridiflava strains are probably plastic in their mucoid and nonmucoid phenotypes, the genetics underlying this phase switch are currently unknown and may be of interest for future research.

| Pathogenicity islands, associated virulence genes, and effectors
In the P. syringae complex and commonly in gram-negative bacteria, virulence factors such as the type III secretion system (T3SS) and associated effectors are arranged in a cluster known as a pathogenicity island (PAI). In the early 2000s, Araki et al. (2006Araki et al. ( , 2007 contributed significantly to the understanding of the genetic basis  (Araki et al., 2006). The two forms, a single pathogenicity island (S-PAI) and a tripartite pathogenicity island (T-PAI), differ in structure and phenotype. The T-PAI contains three components: the hrp/hrc gene cluster, the 5′ effector locus or the exchangeable effector locus (EEL), and the 3′ effector locus or the conserved effector locus (CEL); the T-PAI variant region is typically c.45 kb (Araki et al., 2006(Araki et al., , 2007. The S-PAI differs in that it only contains one of the components of the T-PAI, the hrp/hrc cluster, as well as a 10 kb insertion; the S-PAI variant region is typically c.30 kb (Araki et al., 2006(Araki et al., , 2007. In Araki et al. (2007), S-PAI-associated virulence genes include avrE (avirulence gene), avrF (putative avrE chaperone), and hrpA, hrpZ, and hrpW (type III secreted proteins). In the same study, T-PAI associated genes included those of S-PAI as well as hopPsyA (avirulence gene) and shcA (putative hopPsyA chaperone) (Araki et al., 2007; Figure 2). An association between PAI type and hostspecific virulence was also noted; S-PAI variant isolates were found to cause disease more rapidly on Arabidopsis, while the T-PAI variant isolates were faster in causing a hypersensitive response (HR) in tobacco (Araki et al., 2006). In a study of 286 P. viridiflava isolates from around the world, 10% contained a T-PAI and the other 90% contained an S-PAI; in both cases, each isolate contained a single type of PAI (Araki et al., 2006). Thus, the majority of P. viridiflava isolates examined harboured a S-PAI.
Since the work of Araki et al. (2006Araki et al. ( , 2007, the previous understanding that S-PAI and T-PAI do not share a common EEL region has shifted. In a study of environmental P. viridiflava isolates, a genomic region resembling an EEL was detected in S-PAI strains, probably from a recombination between the two types of PAI . This amends the previous understanding that only two distinctly different PAIs, one containing an EEL and CEL and one without either, exist in P. viridiflava. Additionally, although it was previously accepted that the two different PAIs were associated with varied virulence phenotypes, recent research showed that the two PAI configurations are not correlated with pathogenicity or soft rotting capability; instead, it was found that the only gene linked with pathogenicity was the presence or absence of the avrE effector on the PAI . As sampling and sequencing of P. viridiflava increases, it is possible that there will be more isolates with variable PAIs due to recombination than have previously been discovered. To infer evolutionary history, Bartoli et al. (2014) (Figure 3). This corroborates the findings in Araki et al. (2007) that the PAI types have a deep and divergent evolutionary history. It was also found that regardless of PAI type, there were two genes, coding for lipoprotein and monooxygenase, that were present in nearly all strains that were analysed by Bartoli et al. (2014). These genes are present in the EEL of the T-PAI and in a region resembling an EEL, yet lacking effectors, in S-PAI types. Phylogenetic analysis of these two genes showed that they grouped in accordance with their PAI type . Finally, in Bartoli et al. (2014) PG7 strains contained either an S-PAI or T-PAI type, while PG8 strains contained only the T-PAI type.
This finding led to the hypothesis that the T-PAI in PG7 strains was probably acquired later in its evolutionary history, which was supported by the placement PG8 strains at the root of the phylogenetic tree of PG7 and PG8 strains constructed with four-gene MLSA. In the context of all PGs within the P. syringae complex based on fourgene MLSA , PG7 and PG8 group more proximally to PG11 (P. cichorii), which is closest to the root of the tree, than most other phylogroups. A distinct feature of P. cichorii is its oxidasepositive phenotype. P. viridiflava does not have an oxidase-positive phenotype, but the cytochrome c oxidase operon, responsible for this phenotype in P. cichorii, was found in two strains in PG7, but not in any other phylogroup . The shared S-PAI and cytochrome c oxidase operon between P. cichorii and some PG7 strains is corroborative of their evolutionary history.
Recently, the phylogenetic distribution of the T3SS in P. viridiflava was studied. Four types of T3SS were detected in P. viridiflava strains: canonical T-PAI, alternate T-PAI (which acts as a replacement for the canonical T-PAI), S-PAI, and a Rhizobium-like PAI, or R-PAI, which differs from other PAIs by the splitting of the hrcC gene (Dillon, Thakur, et al., 2019;Gazi et al., 2012). Interestingly, all P. viridiflava isolates containing a S-PAI T3SS also contained a R-PAI

| Other potential virulence factors
While pathogenicity islands, phase variants, and soft rot capability are the main contributors to virulence, there are a few other potential virulence factors in the arsenal of P. viridiflava. The P. syringae group in general is known for its use of toxins, particularly coronatine, syringomycin, syringopeptin, tabtoxin, and phaseolotoxin, in induction of plant disease (Bender et al., 1999). Although P. viridiflava isolates in PG7 have not been shown to produce toxins, isolates in PG8 all produced a toxin inhibiting the fungus Geotricum candidum in an in vitro bioassay . Because P. viridiflava probably does not produce syringomycin, it is possible that the toxicity could be a product of an antimycotic, ecomycin, that was previously identified as a toxin produced by P. viridiflava Miller et al., 1998). The production of ecomycin may serve as a virulence factor by means of eliminating fungal competitors.
Another potential virulence factor is ice nucleation activity (INA).
This is a general characteristic that spans across the phylogroups of the P. syringae complex at varied intensities. Currently, only isolates F I G U R E 3 Phylogenetic tree of Pseudomonas viridiflava isolates constructed with pathogenicity island (PAI)-associated gene hrcC. This figure from Bartoli et al. (2014)  To prevent and control plant disease, it is necessary to be able to detect the pathogen. Fortunately, there have been advances in detection methods of P. viridiflava in recent years. PCR primers for lipoprotein and monooxygenase genes, which are present in the majority of P. viridiflava strains regardless of PAI type, were created for species-specific detection . Primers for the lipoprotein and monooxygenase genes in P. viridiflava  have been used in multiplex PCR with primers for the lipodepsipeptide toxin gene (Sorensen et al., 1998) present most commonly in P. syringae sensu strico for the detection and diagnosis of pathogens causing bacterial stem blight of alfalfa (Lipps et al., 2019). Currently, there are no highly effective methods for management of P. viridiflava diseases. Generally, elimination or reduction of pathogen inoculum is recommended for diseases caused by the P. syringae species complex . For P. viridiflava, the recent discovery of irrigation water, streams/rivers, snowpack, and epilithic biofilms serving as inoculum sources should shape the practices for eliminating or reducing pathogen inoculum. There have been some successes using Bacillus as a biocontrol in vitro, as well as some promise in using copper compounds to control epiphytic populations (Balestra & Bovo, 2003;Orel, 2020). Additionally, further exploration of the mechanisms behind the nonpathogenic transparent phase variants of P. viridiflava could pave the way for developing control strategies based on increasing the occurrence of these variants (Bartoli et al., 2015).
As far as achieving disease resistance, the use of translational taxonomy and the application of basic taxonomic research to advance knowledge regarding disease control will be crucial in the case of P. viridiflava due to its muddled relationship to the P. syringae complex. Classifying, naming, and identifying isolates of P. viridiflava based on relevant characteristics will enhance the ability of researchers to develop resistant plants. For example, current knowledge of pectolytic capability, PAI type diversity, and effector and (a) virulence gene repertoires specific to P. viridiflava will help accelerate research on P. viridiflava-specific avenues of disease resistance.
At present, there are no cases of plants bred or engineered specifically for resistance of diseases caused by P. viridiflava.

ACK N OWLED G EM ENTS
This paper is a joint contribution from the Plant Science Research