Cherry picking by pseudomonads: After a century of research on canker, genomics provides insights into the evolution of pathogenicity towards stone fruits

Abstract Bacterial canker disease is a major limiting factor in the growing of cherry and other Prunus species worldwide. At least five distinct clades within the bacterial species complex Pseudomonas syringae are known to be causal agents of the disease. The different pathogens commonly coexist in the field. Reducing canker is a challenging prospect as the efficacy of chemical controls and host resistance may vary against each of the diverse clades involved. Genomic analysis has revealed that the pathogens use a variable repertoire of virulence factors to cause the disease. Significantly, strains of P. syringae pv. syringae possess more genes for toxin biosynthesis and fewer encoding type III effector proteins. There is also a shared pool of key effector genes present on mobile elements such as plasmids and prophages that may have roles in virulence. By contrast, there is evidence that absence or truncation of certain effector genes, such as hopAB, is characteristic of cherry pathogens. Here we highlight how recent research, underpinned by the earlier epidemiological studies, is allowing significant progress in our understanding of the canker pathogens. This fundamental knowledge, combined with emerging insights into host genetics, provides the groundwork for development of precise control measures and informed approaches to breed for disease resistance.

Since it was first reported in the early 1900s (Brzezinski, 1902), bacterial canker has contributed to major production losses worldwide (Kennelly et al., 2007). It is a serious problem for young plantations and nurseries, and cherry tree losses of up to 75% have been reported in Oregon due to the disease (Spotts et al., 2010). Spraying with copper-based biocides is the major control method; however, this is being phased out in Europe and strains with copper resistance are frequently reported (Sundin and Bender, 1993;Ghorbani and Wilcockson, 2007). No fully resistant host varieties have been identified. Resistance breeding is also complicated by the remarkable diversity of bacteria causing the disease (Spotts et al., 2010;Farhadfar et al., 2016).
The genus Prunus contains economically important species such as cherry, plum, almond, apricot, and peach, all of which are vulnerable to Pseudomonas diseases. Bacteria infect all aerial plant organs throughout the season, causing fruit spot, shoot necrosis, and blossom blight, as well as causing leaf spot symptoms where the tissue can drop out to leave shot-holes (Crosse, 1966). The disease becomes most serious when bacteria infect woody tissues, causing black necrotic cankers and dieback that may lead to tree death. During the growing period, a diverse population of epiphytic bacteria survive on the leaf surface. This population, that may not induce symptoms provides the inoculum for leaf scar and wound infections in the autumn (Crosse, 1959). The pathogens exist in several different niches on and within the host at various stages of the disease cycle ( Figure 1). This review highlights how recent genome-based research has brought this and other woody plant-infecting bacterial pathogens into focus, and could stimulate development of novel diagnostics, precise control methods, and accelerated resistance breeding in woody perennials.

| IDENTIF YING THE C AUSAL AG ENTS OF BAC TERIAL C ANK ER
Research on the disease in the UK began at East Malling in the 1920s, with the pioneering work of Dr Harry Wormald, who noticed a dieback disease in plum was caused by at least two species of bacteria that entered through wounds (Wormald, 1930). He named one of the species Pseudomonas morsprunorum, and the other Pseudomonas prunicola (Wormald, 1932). It was later determined that P. prunicola is Pss and that P. morsprunorum is a pathovar within P. syringae (Crosse, 1966). Following this original work, strains of Psm were divided into two races (Psm R1 and Psm R2) based on biochemical markers and differential pathogenicity on host genotypes (Freigoun and Crosse, 1975). Genetic data has subsequently shown that the two races are distantly related and should not be considered as races of a single pathovar (Bultreys and Kałużna, 2010). Other pseudomonads closely related to Pss, Psm R1, or Psm R2 have recently been associated with cankers of cherry, namely P. cerasi and P. syringae pv. avii (Ménard et al., 2003;Kałużna et al., 2016b). P. viridiflava strains have F I G U R E 1 Overview of the canker disease cycle on Prunus based primarily on Crosse (1966), and Crosse and Garrett (1966) also been isolated from cankerous shoot tissues of various Prunus species (Harzallah et al., 2004;Parisi et al., 2019). Cherry and relatives in the Prunus subgenus Cerasus are additionally susceptible to bacterial gall disease caused by P. syringae pv. cerasicola (Kamiunten et al., 2000). Aside from cherry, cankers are also associated with the clades P. syringae pv. persicae on peach and P. amygdali on almond (Psallidas and Panagopoulos, 1975;Young, 2010).
The epidemiology of this complex disease has previously been reviewed in detail (Crosse, 1966;Kennelly et al., 2007;Bultreys and Kałużna, 2010;Konavko et al., 2014). Unravelling the pathogen's disease cycle (Figure 1) has greatly informed control strategies. Figure 2 shows symptoms of natural and artificial infection. P. syringae clades that cause bacterial canker in Prunus may differ in lifestyle and aggressiveness on different host tissues, reflecting the multiple niches that perennial crops provide. For example, leaf scars may be the dominant route of infection of cherry by Psm R1 but not by Pss, which preferentially invades through wounds (Cameron, 1962;Crosse and Garrett, 1966). Additional information also suggested that Pss cannot survive in cankers as long as Psm R1 (Crosse and Garrett, 1966); however, it belongs to a phylogroup with strains characterized as better epiphytes (Feil et al., 2005;Helmann et al., 2019), so may be more suited for colonization of the plant surface. Environmental factors such as frost occurrence can also significantly increase the incidence of infection (Kennelly et al., 2007), presumably due to wound creation by frost damage. Inoculation of immature cherry fruits has also revealed differences in induced symptoms, with Pss strains causing large black necrotic lesions whilst Psm races 1 and 2 cause small water-soaked lesions (Bultreys and Kałużna, 2010). Within each pathovar, strains may also exhibit differences in aggressiveness towards a particular host species within Prunus (Gilbert et al., 2009;Hulin et al., 2018b), indicating there is significant variation even between closely related strains equivalent to differentiation into physiological races.
ulmi, a pathogen of elm. P. syringae pv. avii is within the same phylogroup as Psm R2, but is more closely related to strains belonging to P. syringae pv. persicae. P. cerasi is within phylogroup 2a (although it is proposed as a new phylogenomic species; Gomila et al., 2017) and is closely related to the apple blister spot pathogen P. syringae pv. papulans (Hanvantari, 1977). By contrast, strains classified as Pss are highly diverse, originating in different subphylogroups within phylogroup 2 (2b and 2d) and interspersed with Pss pathogens of various plant species and also strains present in the wider environment such as water sources (Ahmadi et al., 2017). P. viridiflava falls within phylogroup 7, with strains isolated from various plants and the environment .
The designation of strains as pathovar syringae (i.e., Pss) within phylogroup 2 provides no information about their host range or symptoms produced, and ignores genetic diversity. In terms of host specificity, closely related strains of Pss have been isolated from different Prunus species, such as apricot, cherry, and plum and many can cross-infect (Gilbert et al., 2008;Rezaei and Taghavi, 2014;Hulin et al., 2018b). Members of the other Prunus-infecting clades may be more host-restricted. In particular, Psm R1 strains can be differentiated into at least two host-specific lineages that show differential virulence on Prunus species (Crosse, 1953;Crosse and Garrett, 1970;Hulin et al., 2018a). Phylogenomics revealed two closely related sister groups within this clade that are >99.7% identical at the genome level (M. T. Hulin, unpublished observation), using the program PYANI, for genomic average nucleotide identity (ANI) analysis (Pritchard et al., 2016). One group is virulent on both cherry and plum and the other only virulent on plum (Hulin et al., 2018a(Hulin et al., , 2018b. Similarly, P. syringae pv. avii and P. syringae pv. persicae are very closely related but may have host range differences within Prunus, with the former being restricted to wild cherry (Ménard et al., 2003). Further rigorous pathogenicity testing will be required to determine the level of host specificity within the other Prunusinfecting clades.
The diversity of bacteria causing canker raises issues of taxonomy and nomenclature, not only for the scientific community, but also practically for development of control measures, plant health regulations, and diagnostics. Whole genome data confirm that the various canker pathogens do indeed fall into separate phylogenomic species based on the established ANI cut-off of 94%-95% (Gomila et al., 2017). This divergence in their core genome does not necessarily mean that they use distinct mechanisms of pathogenicity (see discussion on shared flexible effector genes below). In terms of classification, the naming of Psm R1 and R2 as P. syringae pv. morsprunorum (Psm) wrongly suggests genetic similarity and thus the use of separate phylogenomic species names P. amygdali (Psm R1) and P.
avellanae (Psm R2) has been suggested to reflect their phylogenetic positions (Bultreys and Kałużna, 2010; Marcelletti and Scortichini, F I G U R E 3 Core genome phylogenetic tree of Pseudomonas syringae species complex highlighting strains isolated from Prunus. The tree was generated using IQ-TREE (Nguyen et al., 2015). Phylogenomic clades based on whole genome average nucleotide identity (ANI) of ≥95% are coloured and phylogroups numbered. Circles going outwards: (1) strains isolated from cherry, other Prunus and Rosaceae are coloured according to the scale; (2) presence of the canonical type III secretion system (T3SS) in grey (see Text S1 and Table S3  This would allow more focus on and dissection of pathogenicity mechanisms encoded by genes that may be shared requirements for pathogenicity, but are outside the different core genomes. The addition of phylogroup information after this could then add a level of genomic information, for example, P. syringae pv. cancriprunorum phylogroup 1.

| WHAT MAK E S A CHERRY PATHOG EN?
How do the different clades of P. syringae cause bacterial canker? P. syringae is a model organism in the study of plant-microbe interactions (Mansfield et al., 2012). Pathogenicity factors have been well studied in model pathosystems using tomato, Arabidopsis, and bean (Preston, 2000;Arnold et al., 2011). Essential requirements for virulence have been identified, including the production of effector proteins injected into plant cells through the type III secretion system which is induced by the detection of effector proteins (Jones and Dangl, 2006). Effectors are essential for virulence as they suppress both PTI and ETI, but conversely can activate ETI when resistance

| Type III effectors (T3Es)
Despite their common host, cherry-infecting strains show remarkable variation in their number and composition of T3E repertoires (Hulin et al., 2018a), reflecting their phylogenetic divergence and perhaps the functional redundancy among T3Es (Lindeberg et al., 2012). Extensive research of the model pathogen P. syringae pv.

| T3Es promoting virulence?
Using phylogenetically aware association statistics on the cherry pathogens with a canonical T3SS, a set of T3Es was identified that are over-represented in cherry pathogenic clades, and therefore are putatively of common importance in this disease. This approach identified several candidate virulence-associated T3E genes: hopAR1, hopBB1, hopBF1, and hopH1, which are present in many cherry pathogen genomes across the phylogeny and have probably been gained in these clades.

HopAR1
(also known as AvrPphB) is a C58 YopT cysteine protease that recognizes and cleaves receptor-like cytoplasmic protein kinases (RLCKs) such as BIK1, RIPK, PBS1, and various PBS-like kinases in Arabidopsis crucial for both the PTI and ETI response of the plant (Zhang et al., 2010;Russell et al., 2015). HopAR1 has also been shown to cleave homologues of PBS1 in barley (Carter et al., 2019), indicating its targets may be conserved across plant families.
We cannot assume it has the same role in cherry without validation by identification of interacting partners, so its function in this host is still to be determined. Members of Psm R1, Psm R2, P. syringae pv. avii, P. cerasi, and Pss within the subphylogroup 2d possess the hopAR1 gene. P. syringae also possesses homologous YopT cysteine proteases, such as HopAY1 and HopAW1, for which the host targets are yet to be characterized and are potentially different from those of HopAR1 (Dowen et al., 2009). Interestingly, many cherry pathogens also possess multiple copies of genes encoding these effector families (Hulin et al., 2018a).
HopBB1 is part of the HopF family containing multidomain proteins with homologous N-termini and myristoylation sites for plant cell membrane localization (Lo et al., 2016). HopBB1 is a chimeric effector as its C-terminus is not homologous to the other family members, which may indicate different functionality. In Arabidopsis, HopBB1 specifically interferes with hormonal signalling, promoting jasmonic acid-activated pathways by the removal of two repressors, TCP14 and JAZ3. This leads to the down-regulation of salicylic acid signalling, important for defence against P. syringae (Yang et al., 2017). The manipulation of jasmonic acid signalling to suppress immunity is a role shared by other effectors such as HopX1 and HopZ1a as well as the toxin coronatine (Jiang et al., 2013;Gimenez-Ibanez et al., 2014). By contrast, other HopF members, such as HopF2, have been shown to interfere with immune signalling through interactions with RIN4, MKK4, and BAK1 (Lo et al., 2016). It is unclear if HopBB1 shares any functional similarities with other HopF members and if its role in manipulation of hormone signalling is conserved in cherry. However, an interesting note is that members of Psm R1, P. syringae pv. cerasicola, and Psm R2 each possess two different alleles of the hopF gene as well as the hopBB1 gene (Hulin et al., 2018a). Such duplications indicate that this effector family may be enriched in cherry pathogens and is therefore a target for further study.
The HopBF1 effector is a protein kinase that can inactivate the chaperone Hsp90, preventing its role in activating key immune receptors during plant immunity (Lopez et al., 2019). Its representation across the cherry pathogenic clades is intriguing as this effector is relatively rare across the whole P. syringae phylogeny (18% of all strains compared to 45% of cherry isolates).

| T3Es triggering resistance
In contrast to the positive link between effector presence and viru-  (Jackson et al., 1999;Helmann et al., 2019). As this function is important for disease on these hosts, perhaps other T3Es take on this role in suppression of the cherry immune system? The alternative possibility, that some functions are not required or less important for infection of cherry, should also be considered.
To determine if AvrPto and HopAB trigger ETI in cherry, they were ectopically expressed in cherry-pathogenic strains in the laboratory (Hulin et al., 2018a). Expression of several alleles of the hopAB gene triggered a hypersensitive reaction (HR) in cherry (faster symptom onset at high inoculum concentration) and restricted bacterial multiplication from lower doses in planta. The avrPto1 and hopAB1 effector genes were found in the clade of Psm R1 plum-infecting strains that exhibit reduced aggressiveness on cherry. One hypothesis is that selection pressures due to the activation of ETI could have driven the loss of these genes to enable greater virulence on cherry. The truncation of hopAB3 in Psm R2 and indel in this allele in P. syringae pv. avii could also have been selected during the evolution of these clades to maximize disease on cherry by avoiding recognition of sites in the HopAB effector protein (Hulin et al., 2018a). As HopAB3 is a multidomain protein, the truncated version in Psm R2 may still function and play a role during virulence on cherry. Similar truncated forms of HopAB lacking the E3 ubiquitin ligase domain or with diminished enzymatic activity of this domain have also been identified in other pathovars (Lin et al., 2006;Chien et al., 2013).
Other studies have examined similarly divergent lineages of P.

| Toxins
Other virulence factors used to manipulate the host include toxins that are not secreted via the T3SS and have been historically identified based on their contributions to symptom development (Bender et al., 1999). Strains of Pss in phylogroup 2 possess several gene clusters for toxin biosynthesis, and also generally have much smaller effector repertoires than other phylogroups; the toxins may compensate for the reduced set of effector proteins (Figure 3). Gene clusters for up to four toxins, syringomycin, syringolin A, syringopeptin, and mangotoxin, are present in Pss strains.
Syringomycin and syringopeptin are produced after detection of plant signal molecules in a similar manner to the T3SS (Wang et al., 2006). Single and double knockouts of genes involved in these pathways revealed that both toxins contribute to symptom development in immature cherry fruits (Scholz-Schroeder et al., 2001). A study of Pss infecting bean showed that syringomycin may also be required for competitive fitness of bacteria in the plant apoplast (Helmann et al., 2019). The toxins are antagonistic to other bacteria and also fungi, suggesting that they also function in microbial competition (Lavermicocca et al., 1997). By contrast, syringolin A, a nonribosomal cyclic peptide fused with a polyketide synthase, acts within plant cells to inhibit the proteasome interfering with the immune response (Schellenberg et al., 2010). It has also been shown to promote the spread of bacteria by suppressing immune responses in neighbouring plant tissues and increasing bacterial motility (Misas-Villamil et al., 2013).
Mangotoxin is an antimetabolite toxin present in phylogroup 2a and 2b, which includes some cherry and plum strains. Originally identified in Pss pathogenic to mango, it inhibits the biosynthesis of ornithine and arginine and has been implemented as a virulence factor as it increases disease symptoms and may also improve epiphytic fitness (Arrebola et al., 2009).
Some Psm R1 strains possess the gene clusters for the biosynthesis of coronatine, a chlorosis-inducing polyketide toxin that interferes with hormone signalling in plant cells to induce stomatal opening and interrupt the salicylic acid-associated immune response (Grant and Jones, 2009). Coronatine expression is co-regulated with the T3SS as it is induced in a hrpL-dependent manner in P. syringae pv. tomato DC3000 (Fouts et al., 2002). The possession of coronatine genes is one of several discriminating factors that separate strains of Psm R1 with high and low aggressiveness on cherry (Gilbert et al., 2009;Hulin et al., 2018a).

| Hormones
Hormone production is a common trait of plant-associated bacteria, which can lead to hormonal imbalance and therefore modification of plant growth and development (Aragón et al., 2014). Auxin and cytokinin production have been extensively studied in the gall-forming olive knot pathogen P. savastanoi pv. savastanoi, in which their production is required for full expression of knot symptoms (see reviews by Ramos et al., 2012 andCaballo-Ponce et al., 2017). Two key genes involved in auxin (IAA) biosynthesis, iaaH and iaaM, are present in many pathovars of P. syringae (Kunkel and Harper, 2018), including Prunus pathogens (Hulin et al., 2018a). By contrast, the gene encoding an isopentenyl transferase (ptz) required for cytokinin production is not widely distributed within the P. syringae species complex. Although it is present in the gall-forming P. syringae pv. cerasicola in phylogroup 3, it is absent from the other pathogens of Prunus that principally cause cankers (Ruinelli et al., 2019). Disruption of the phytohormone balance would be expected to lead to the overgrowth associated with canker symptoms but the precise role of IAA and cytokinin production by these pathogens is unclear and remains to be explored.

| Diverse weaponry leads to pathogenicity
The differences in the complements of virulence factors between the Prunus-infecting P. syringae clades may reflect their subtle differences in pathogenicity. In particular, toxin production is probably the major factor that leads to different symptoms on fruits seen in Figure 2 (Scholz-Schroeder et al., 2001). Cherry leaf inoculations revealed that Pss induces disease symptoms more rapidly than the Psm races, indicating that it may have a more necrotrophic lifestyle (Hulin et al., 2018b). With only a core set of T3Es lacking redundancy, Pss may be less able to suppress the immune system and therefore be unable to remain biotrophic for as long as members of other clades.
This idea requires support from further experimental evidence, such as looking at the timing of toxin expression and when the switch to necrotrophy occurs. Continued proliferation of the bacterial population after plant cell death would also provide evidence for necrotrophy alongside symptomology.
The reduced repertoire of T3Es in Pss and other members of phylogroup 2 may also expand the host range of this clade, as they hypothetically possess fewer ETI-activating effectors. However, a preliminary assessment of different Pss strains isolated from cherry, plum, bean, pea, and lilac inoculated onto detached cherry leaves revealed that the strains from Prunus did grow to higher population levels in planta than those from other plants, indicating that some host adaptation does exist within phylogroup 2 (Hulin et al., 2018a).
A host specificity study on cucurbits showed remarkable variation in the virulence of phylogroup 2 strains on two different host species (Newberry et al., 2019), which was postulated to be due to differences in effector repertoires. Another study by Rezaei and Taghavi (2014)

What are the origins of factors important for virulence on Prunus?
An examination of horizontal gene transfer of T3E genes found that transfers were predicted between Psm R1, Psm R2, and P. syringae pv. avii (Hulin et al., 2018a). With the use of long-read sequencing technologies, recent studies have been able to resolve the complete genomes of cherry pathogens into chromosomal and plasmid sequences (Hulin et al., 2018a;Ruinelli et al., 2019). Complete genomes can greatly improve the accuracy of downstream analyses by resolving highly repetitive regions and providing structural context to gene locations (Baltrus and Clark, 2019;Smits, 2019).
Careful re-examination of the presence of T3E genes in the whole genomes of cherry pathogens characterized by long-read sequencing technologies has revealed some effector genes are present in multiple copies (authors' unpublished observation). The hopAY1 gene is present in two copies in certain complete Psm R1, Psm R2, and P. syringae pv. avii genomes. In addition, two copies each of hopF3, hopBL2, and hopAO2 are present in Psm R1 (Table S2) that were unresolved in draft genomes. The significance of these duplications is unclear but they may allow enhanced expression of valuable effectors.
Extending the effector analysis to the recently identified cherry pathogen P. cerasi shows that it may also share effectors with the other cherry-infecting clades (Ruinelli et al., 2019;M. T. Hulin, unpublished observation). Many of the shared T3Es are located on plasmids (Figure 4), suggestive of horizontal gene transfer between the cherry pathogens in different phylogroups. Strains within the clades Psm R1, Psm R2, P. cerasi, and P. syringae pv. avii possess a highly variable number of plasmids (none to seven), with most possessing at least one, whilst Pss strains rarely possess plasmids, ranging from none to two (Liang et al., 1994;Hulin et al., 2018a;Ruinelli et al., 2019). In Psm R1, coronatine has been found to be plasmid-encoded, suggesting that it has been gained in this clade via horizontal gene transfer (Hulin et al., 2018a). However, the striking lack of plasmids F I G U R E 4 Role of horizontal gene transfer in evolution of cherry pathogens. (a) Circular plot of the complete Pseudomonas syringae pv. morsprunorum (Psm) R1-5244 genome created using Circos (Krzywinski et al., 2009). Circles going outwards: predicted prophages (green) and genomic islands (blue), effectors (red), and toxins (purple). Prophage regions and genomic islands identified using PHASTER (Arndt et al., 2016) and Examination of other mobile elements (Hulin et al., 2018a) (Ohnishi et al., 2001) and Vibrio cholerae (Waldor and Mekalanos, 1996); however, this was the first finding of a P. syringae T3E within a putatively active prophage.
The selective retention of some common T3Es among cherry-pathogenic clades points to a role in virulence. In addition, some T3Es may be transferred together or hitchhike with functionally important genes due to linkage on plasmids or other mobile genetic elements and therefore could be less important than hypothesized.
For example, the hopH1 gene is within 10 kb of other effector genes such as hopF4 in Psm R2 (on a putative integrative conjugative element, ICE) and a disrupted copy of the hopAW1 gene in Pss (Hulin et al., 2018a). Interestingly, Newberry et al. (2019) also found the hopH1 gene within ICEs containing a variety of other effectors such as hopC1, hopAR1, hopAW1, avrRpt2, hopZ1, and hopZ5 in curcurbit-infecting Pss. This gene therefore appears to be moving in close association with other effectors. Dissecting the individual contributions of these co-gained effectors may be challenging, particularly in the background of effector-rich strains such as Psm R1 and R2.

| EFFEC TOR S AND LIFE OUTS IDE THE HOS T PL ANT
An important caveat in population genomic analyses is that there is an inherent bias towards sampling pathogenic strains of P. syringae from important crops (discussed in Monteil et al., 2016). This may mean that we are missing the interactions occurring between pathogens and the P. syringae metapopulation in the wider environment such as stages of the water cycle and during interactions with wild plant species (reviewed in Morris et al., 2013). Even within fruit orchards, sampling has been focused on diseased tissues, giving us a narrow view of the lineages occupying this environment. More unbiased sampling, including leaf washing and sampling symptomless tissues, has revealed other P. syringae strains within Prunus and kiwifruit orchards that belong to unknown clades and often exhibit low virulence (Vicente et al., 2004;Gilbert et al., 2008Gilbert et al., , 2009Straub et al., 2018;Visnovsky et al., 2019).

| HOW TO INFEC T WOODY TISSUE S
The ability to colonize and cause disease in trees has evolved multiple times within the P. syringae complex (Nowell et al., 2016).
Woody tissues are so heterogeneous, comprising living cambium and dead lignified vessels, it is difficult to define precisely what special factors, if any, may be required for their colonization.
However, once inside a cherry branch for example, bacteria are within an environment rich in available nutrients and protected from adverse conditions. The symptomology associated with P. syringae diseases in wood varies considerably between host plants and P. syringae clades and reflects dramatic host responses as well as pathogen-induced necroses (Lamichhane et al., 2014). The term canker is defined as a necrotic, usually sunken, lesion on the woody tissue of a plant (Agrios, 1936) and is a symptom of various bacterial and fungal diseases. It is usually associated with gummosis and dieback. Canker symptoms are common in many tree hosts infected with P. syringae other than Prunus, including horse chestnut, kiwifruit, and hazelnut (Schmidt et al., 2008;Ferrante and Scortichini, 2010;Scortichini, 2010). Infection of actively growing shoot tips is also associated with necrosis, referred to as shoot dieback or when rapid, shoot blight/blast (Wimalajeewa, 1987). Blister bark is a symptom seen during Pss infection of apple, whereby the tissue becomes raised, flaky, and necrotic (Scortichini and Morone, 1997). Finally, certain clades within phylogroup It is also present in Psm R2 and its close relatives and is partially present in some Prunus-infecting Pss strains (Hulin et al., 2018a). It may therefore have a role in bacterial fitness within Prunus woody tissues. However, this gene cluster is absent from other clades that cause canker on Prunus, such as P. syringae pv. avii, P. syringae pv.
This absence indicates that it is either not essential for virulence or that other unidentified genes are fulfilling the same role in other clades. Studies of the horse chestnut and kiwifruit pathogens Cunty et al., 2015) have revealed that strains vary in ability to invade leaf or woody tissues and similar experiments are needed to assess the significance of wood colonization in Prunus. Nowell et al. (2016) also found that several effector genes such as hopAY1 and hopAO1 were enriched in clades of woody plant-infecting pathovars. The role of T3Es during woody tissue infection has been shown in olive where the T3SS is essential for growth of P.
Canker development in Prunus species occurs when the tissue is dormant or just coming out of dormancy and there is extensive lignification that the bacteria must overcome or avoid (Caballo-Ponce et al., 2017). The bacteria colonize various tissues such as the cortex, cambium, phloem, and xylem (Crosse, 1956;Lamichhane et al., 2014).
In the growing season, the other stages of the disease such as shoot, fruit, flower, and leaf necrosis, occur in actively growing nonlignified tissue (Crosse, 1966). In-depth studies of bacterial gene expression in-different tissues could provide some insights into whether colonization requires the induction of distinct mechanisms of virulence (Yu et al., 2013). This could be coupled with the identification of mutants deficient in growth and/or symptoms in the different tissues (Somlyai et al., 1986). A recent high-throughput approach called transposon insertion sequencing (TnSeq) involves inoculating pools of transposon mutants and using sequencing to determine mutant frequencies within this pool before and after exposure to a selective environment, such as within a plant (van Opijnen et al., 2009). This approach has proved powerful in identifying different genes important for bacterial fitness in the apoplast versus surface of leaves in the Pss-bean interaction (Helmann et al., 2019). It could be applied to the Pseudomonas-Prunus system to identify candidate virulence genes through an unbiased genetic dissection (Phan et al., 2013).

| TOWARDS VARIE TAL RE S IS TAN CE
There are no examples of cherry cultivars with complete immunity or single gene-based resistance to canker. Broad-acting partial resistance to all clades exists in certain cultivars, notably Merton Glory (Hulin et al., 2018b). Perennial plants provide challenging study systems as differences in tissue-specific resistance, plant age, seasonal environment conditions, and nutrient status impact on susceptibility (Kus et al., 2002;Mur et al., 2017;Velásquez et al., 2018;Tahir et al., 2019).
Historical studies have provided clues to responses involved in canker resistance. Timing of leaf-drop may have a major effect on field resistance as the leaf scar provides a route to infection, particularly in Psm infections. In addition, the timing of break from dormancy and the generation of active phellogen, the meristematic cell layer that generates periderm, which is thought to provide summer immunity, may vary between different cultivars. Varieties that bloom earlier also show reduced canker spread (Wilson, 1939;Crosse, 1966).
Since these early studies, little work has been done to identify the immune responses that occur in woody tissues. It is likely that immunity involves the creation of antimicrobial barriers in the cambium, and phellogen activity to seal off and prevent the spread of bacterial infections (Rioux, 1996;Abe et al., 2007). Quantitative differences in these responses between host genotypes may contribute to varietal resistance. The age of the tree and rootstock choice may also have effects on resistance levels due to differences in vigour (Garrett, 1979(Garrett, , 1981Santi et al., 2004). Analysis of leaf populations revealed that different cultivars may support different levels of epiphytic bacteria.
As these epiphytic populations are the major inoculum source, any genetic differences that reduce the growth of aggressive pathogens on the plant surface may also be important in the field (Crosse, 1966).
Although a range of resistance assays in the field and laboratory ( Figure 2) distinguish high and low virulence strains, the pathogens often show a large degree of variability in their aggressiveness, particularly after field inoculations. This meant that any clade-specific trends in resistance were not statistically significant in the experiments described in Hulin et al. (2018b). For field inoculations, multiyear, multisite experiments using a variety of P. syringae strains may be required to fully assess canker resistance. Genomics and functional studies have revealed that cherry mounts a nonhost ETI response towards particular effectors (Hulin et al., 2018a). Small additive ETI-regulated responses may contribute to varietal resistance and could be clade-specific.
The use of plant genomic association techniques to identify genes for resistance is increasing. Recently, an association study (Omrani et al., 2019) mapped partial resistance towards Pss in apricot to several quantitative trait loci (QTLs), with two loci explaining 41% and 26% of phenotypic variation, respectively. The candidate loci contained genes annotated as producing products involved in phytohormone signalling, specifically abscisic acid, and may mediate crosstalk between different signalling pathways during immunity. Similarly, another association study of kiwifruit canker resistance identified candidate genes within QTLs involved in PTI.
The candidate genes identified are involved in cell wall metabolism and signalling (Tahir et al., 2019). Interestingly, this study used multiple pathogenicity assays to identify resistance QTLs that were tissue-specific and associated with particular disease symptomology.
An approach like this would be highly relevant to Prunus canker as it is also a multiple tissue disease. to identify the causes of epidemics in hospitals and determine antibiotic resistance profiles to inform decision-making (Jeukens et al., 2019). Such precision diagnostics should be applied in the horticultural industry as it would inform the ideal timing and composition of spray routines. One fast-growing field of research and development is the use of strain-specific bacteriophages to control diseases (reviewed in Svircev et al., 2018). The combination of using advanced genomics-based diagnostics with strain-specific bacteriophages would be a precise alternative to current control measures for canker, which involve the blanket spraying of copper-based products.

| LINKING G ENOMI C S TO THE CONTROL OF C ANK ER
Given the close association between the presence of certain effectors and virulence (although their importance is still to be examined in planta), a simplistic outcome of the genomic analysis is to propose that expression of the matching R genes in cherry could confer ETI (Kennelly et al., 2007). For example, the R proteins RPS5 and ZAR1 are known to detect members of the HopAR1 and HopF effector families, respectively (Laflamme et al., 2020) that are common in cherry-infecting strains. The cherry genome sequence (Shirasawa et al., 2017) could be searched for homologues of these R genes and other genes required for effector recognition that have been extensively characterized in model plants. A transgenic approach to stack multiple known R genes against the common cherry pathogen effector set would provide a proof of concept for such approaches. Although the idea is promising it would probably be challenging to transfer genes across plant genera and families and maintain the correct functionality whilst not triggering autoimmunity (Zhang and Coaker, 2017). It has been shown to work experimentally in some cases; for example, the transfer of the RPS4 and RRS1 genes, involved in effector recognition, from Arabidopsis into tomato conferred resistance to several pathogens including P. syringae pv. tomato (Narusaka et al., 2013).
An additional molecular approach to accelerate the search for R gene-mediated resistance would be to utilise effectoromics, which involves screening single or combinations of effectors for their ability to elicit immunity in planta. In this targeted approach the wider Prunus germplasm, which is potentially amenable to conventional plant breeding into cherry, could be screened for responses to core effectors. Effectoromics has been utilized in wild relatives of potato to identify novel resistance sources to the oomycete pathogen Phytophthora infestans (Vleeshouwers et al., 2008). If multiple distinct sources of resistance to one or more clades of P. syringae could be identified, the underlying resistance genes could be pyramided to provide durable resistance to the canker pathogens. In addition, genomics provides a route to understanding the redundancy and mobility of virulence factors and therefore could be used to predict and monitor virulence factor diversity (Thilliez et al., 2019) and predict the likelihood of resistance failing, due to the loss, modification, or pseudogenization of targeted effectors.

| FUTURE PROS PEC TS
Bacterial canker of Prunus species is a highly complex disease.
Perennial plants provide multiple niches for bacterial occupation including different tissues and host life stages. Progress towards the development of resistant cultivars is impeded by the diversity of bacteria that cause the disease. It is clear that further research into the fundamental mechanisms of resistance and underlying genetics will be beneficial to inform breeding programmes. The diversity of the bacteria associated with the canker syndrome presents a challenge that breeders must appreciate, and strain selection for phenotyping will be key. There has been valuable progress in the development of tools to diagnose the cause of the disease and this will be further enhanced by novel sequencing-based technologies that could allow rapid in-field diagnostics and monitoring of disease potential and host range. With a greater knowledge of pathogen lifestyle and genetics, control measures can be made more precise and durable by targeting the particular pathogens that occupy particular tissue types in each orchard during the annual cycle of tree production.

| Targets of future research
• A revision of the taxonomy of canker pathogens of Prunus would be beneficial.
• Any newly identified pathogens should be examined thoroughly for virulence and host range to assess their possible impact on their species of origin and wider Prunus.
• Genetic dissection of the role and regulation of virulence factors such as T3Es and toxins in the colonization of woody tissues will provide insight into niche specialization.
• Further investigation of epiphytic populations on hosts and nonhosts that may act as reservoirs for the pathogens and facilitate gene exchange.
• The identification of the genetic basis of resistance to canker in related Prunus.

ACK N OWLED G EM ENTS
The authors acknowledge BBSRC for funding this work (BB/ P006272/1). They also thank Penny Greeves for generating the disease cycle graphic in Figure 2. This review builds upon the extensive work that has been conducted by various research groups across the world on bacterial canker in the last century. Without prolonged study, diseases of perennial crops cannot be well characterized. The dedication of many researchers over the years to studying the epidemiology of these diseases provides the groundwork for genomic analyses today.

CO N FLI C T O F I NTE R E S T
The authors declare no conflict of interest.

DATA AVA I L A B I L I T Y S TAT E M E N T
Data sharing is not applicable to this article as no new data were created or analysed in this study.