Identifying resistance in wild and ornamental cherry towards bacterial canker caused by Pseudomonas syringae

Abstract Bacterial canker is a major disease of stone fruits and is a critical limiting factor to sweet cherry (Prunus avium) production worldwide. One important strategy for disease control is the development of resistant varieties. Partial varietal resistance in sweet cherry is discernible using shoot or whole tree inoculations; however, these quantitative differences in resistance are not evident in detached leaf assays. To identify novel sources of resistance to canker, we used a rapid leaf pathogenicity test to screen a range of wild cherry, ornamental Prunus species and sweet cherry × ornamental cherry hybrids with the canker pathogens, Pseudomonas syringae pvs syringae, morsprunorum races 1 and 2, and avii. Several Prunus accessions exhibited limited symptom development following inoculation with each of the pathogens, and this resistance extended to 16 P. syringae strains pathogenic on sweet cherry and plum. Resistance was associated with reduced bacterial multiplication after inoculation, a phenotype similar to that of commercial sweet cherry towards nonhost strains of P. syringae. Progeny resulting from a cross of a resistant ornamental species Prunus incisa with susceptible sweet cherry (P. avium) exhibited resistance indicating it is an inherited trait. Identification of accessions with resistance to the major bacterial canker pathogens is the first step towards characterizing the underlying genetic mechanisms of resistance and introducing these traits into commercial germplasm.

avii (Psa), and P. cerasi (Kałużna et al., 2016;Ménard et al., 2003;Parisi et al., 2019). Psm R1 and Psm R2 are genetically distinct, belonging to different phylogroups within the species complex and can be alternatively referred to as within the species P. amygdali and P. avellanae, respectively (Gomila et al., 2017). Bacterial strains differ in host range and aggressiveness towards particular species in the genus (reviewed in Bultreys & Kałużna, 2010). A recent study identified a range of factors that contributed to bacterial virulence, but also found that knockout of genes encoding possible avirulence proteins, including the effector HopAU1, led to hypervirulent bacterial phenotypes, suggesting a quantitative level of resistance exists even in susceptible cultivars (Neale et al., 2021).
Control measures available for bacterial canker are limited. The genotypically diverse P. syringae clades causing the disease may vary in sensitivity to control measures and can rapidly evolve and transfer genes conferring resistance to chemicals such as copper-based biocides and antibiotics (Sundin et al., 2016). The genetic diversity of bacterial canker pathogens poses a challenge in the generation of novel controls because responses to all potential pathogens must be tested.
Progress is being made in the development of specific biological controls such as the use of bacteriophages (Rabiey et al., 2020) that could be used in combinations effective against all clades. A complementary approach is to breed for resistance, a strategy particularly important in forestry, where spraying control is impractical (Vicente et al., 2004).
Ideally, resistance against multiple clades would be most beneficial or an alternative strategy would be to stack resistance-associated loci effective against the different clades into new varieties.
The molecular mechanisms involved in plant resistance towards bacterial pathogens, such as P. syringae, have been extensively characterized in model plant species such as Arabidopsis thaliana.
Resistance involves heightened immunity that occurs at the plant cell surface through receptor detection of pathogen-associated molecular patterns, as well as the intracellular detection of pathogen virulence proteins (effectors) injected into plant cells. These two components of resistance are now known to be intrinsically linked (Ngou et al., 2021).
There is limited knowledge of resistance in Prunus towards bacterial canker pathogens. Cherry and apricot varieties with partial resistance to one or more of the pathogens have been identified using methods such as laboratory-based shoot inoculations and field tree inoculations (Farhadfar et al., 2016;Hulin, Mansfield, et al., 2018;Omrani et al., 2019;Santi et al., 2004). In our previous study, we found that the partial resistance seen in woody tissue of certain cherry cultivars was not detected using detached leaf syringe-infiltration assays (Hulin, Mansfield, et al., 2018). This partial resistance seen in woody tissues is probably quantitative, involving multiple alleles having small effects, with the most resistant varieties still succumbing to disease under favourable conditions. Although only partial, such resistance could be highly useful for Prunus breeding as it could reduce overall pathogen load in orchards as part of an integrated disease management approach (Sundin et al., 2016).
In addition, it is arguably more durable than single resistance (R) gene-based immunity, which is, theoretically, more frequently overcome during pathogen evolution (Pilet-Nayel et al., 2017). Progress towards understanding the genetic factors involved in bacterial canker resistance has been made by Omrani et al. (2019) who identified quantitative trait loci (QTLs) involved in partial resistance in apricot.
These loci contain genes involved in phytohormone signalling, a process known to play a pivotal role during the plant immune response.
Studies reporting the screening of Prunus for canker resistance have focused on established commercial varieties. However, wild relatives can provide robust sources of disease resistance not found in crop genotypes and may be introduced during crop breeding.
Nonhost resistance is defined as the ability of all genotypes of a plant species to resist all genotypes of a pathogen (Heath, 2000). Such resistance traits can be transferred into crops. For example, relatives of apple such as Malus × robusta 5 and Malus floribunda have been used extensively to introduce complete resistance towards the fireblight pathogen Erwinia amylovora, both through breeding and transgenic strategies (Campa et al., 2019). In addition, wild accessions of kiwifruit have been identified with resistance towards the canker pathogen P. syringae pv. actinidiae using large-scale in vitro assays (Wang et al., 2020). Prunus is a diverse genus that includes five subgenera: Amygdalus, Cerasus, Prunus, Laurocerasus and Padus (Chin et al., 2014), with many natural and artificial interspecific hybrids.
The subgenus Cerasus includes P. avium (sweet and wild cherry), P. cerasus (sour cherry) and P. mahaleb. Wild cherry is native to Europe, Africa and western Asia (Miljković et al., 2019) and exhibits greater genetic diversity than sweet cherry (Avramidou et al., 2010), potentially including diversity in genes conferring resistance to pathogens.
Studies have already shown wild Prunus species to be important sources of resistance to pathogens such as plum pox virus (Decroocq et al., 2005). Therefore, in this study, we aimed to identify resistance in accessions of wild cherry. Sweet cherry cultivars are known to vary in their resistance towards bacterial canker disease under field conditions (Farhadfar et al., 2016;Mgbechi-Ezeri et al., 2017), but no complete resistance has been reported. We screened a wide variety of wild cherry accessions and Prunus species related to cherry for resistance to the bacterial canker pathogens. We also screened several hybrids of susceptible sweet cherry crossed with ornamental species. Our results have identified potential sources of resistance to members of each of the pathogenic clades of P. syringae.

| Plant material
The Prunus germplasm used in this study (Table 1)

| Bacterial strains
Strains of P. syringae used and the experiments they were included in are listed in Table S2. The most used strains were Psm R1-C (R1-5244), originally isolated from sweet cherry; Psm R1-P (R1-5300), isolated from plum with low virulence on sweet cherry; Psm R2 (PsmR2-leaf, renamed MH001), isolated from sweet cherry; and Pss (Pss 9644), also isolated from sweet cherry. For the wild cherry screening, the pathogen P. syringae pv. avii (avii5271) was also included. Screening was later extended to a diverse range of strains on selected Prunus accessions. The pathogenicity of the strains was extensively characterized in Hulin, Mansfield, et al. (2018). Culturing and inoculum preparation were as described in this previous work (Hulin, Mansfield, et al., 2018). Briefly, strains obtained from longterm 20% glycerol stocks held at −80°C were grown on King's B agar (King et al., 1954)  b Experiments in which each accession is included: a: sweet cherry cut-shoot assay of susceptibility ( Figure 1, Figure (Figure 9). c Accessions tested further after initial screening.
d Accessions showing significantly reduced symptom development compared to susceptible sweet cherry controls.
e Prunus subgenus Cerasus interspecific hybrids from crosses with P. avium.

TA B L E 1 (Continued)
F I G U R E 1 Susceptibility of sweet cherry cultivars to Pseudomonas syringae infection. Boxplots show length of necrotic lesions in cut-shoots inoculated with P. syringae pv. morsprunorum race 1-C (Psm R1-C 5244), P. syringae pv. morsprunorum race 2 (Psm R2 MH001), or P. syringae pv. syringae (Pss 9644), 6 weeks after inoculation. The boxplots are ordered by estimated marginal means derived from the linear model to visualize the range of responses, although the graphs are of raw data. Individual data points are included and coloured for each separate experiment and the arithmetic mean is shown with a black diamond. This experiment was repeated up to five times per cultivar × strain combination. This figure shows the results for the three main pathogens, while the full data including results using Psm R1-Plum and mock-inoculated controls (neither of which caused significant symptoms) are presented in Figure S1. REML analysis indicated a significant difference between cultivars (p < 0.01, df = 15), strains (p < 0.01, df = 4) and an interaction between them (p < 0.01, df = 60). Tukey HSD (p = 0.05, confidence level: 0.95) significance groups obtained from the estimated marginal means (emmeans) are presented separately for each bacterial strain as letters under the graph [Colour figure can be viewed at wileyonlinelibrary.com] for 10 min before resuspending in 10 mM MgCl 2 to an optical density at 600 nm (OD 600 ) of 0.2, which corresponds to approximately 2 × 10 8 cfu/ml. This inoculum was then diluted to generate the different inoculum concentrations required for each experiment.

| Pathogenicity assays
Shoots were collected from mature trees and inoculated using the dip inoculation method described in Hulin, Mansfield, et al. (2018).
Briefly, 1-year-old shoots, 12 cm in length, were collected from fieldgrown trees when dormant (December-February). Before inoculation, shoots were surface sterilized with 70% ethanol and allowed to air dry. The apical end was cut with secateurs (removing 1 cm) and dipped in bacterial inoculum of 2 × 10 7 cfu/ml for 5 min. Shoots were blotted dry on paper towel and sealed with Parafilm. The basal end of the shoot was then cut (removing 1 cm) and the shoot was kept in water for 1 week at 16°C with 16:8-h light:dark cycles. Shoots were then placed in Oasis foam in a fully randomized design and kept at Leaf pathogenicity assays and population counts were conducted as in Hulin, Mansfield, et al. (2018). Leaves were infiltrated using a blunt-ended syringe with inoculum usually at a concentration of 2 × 10 6 cfu/ml (100-fold dilution of a 0.2 OD 600 suspension). After incubation for 10 days at 22°C, this concentration of inoculum allowed clear differentiation between responses to strains pathogenic to cherry and to other hosts (Hulin, Mansfield, et al., 2018). Each leaf received a mock inoculation as a control and, where appropriate, different strains were compared on the same leaves (up to six inoculation sites) to reduce plant variability. Symptoms were scored on a scale of 0-5 (0, none; 1, limited browning; 2, browning <50% inoculated area; 3, browning >50% inoculated area; 4, complete browning; 5, complete browning with spread away from initial lesion).
Experiments were repeated at least three times. Population counts of bacteria within disease lesions were conducted as previously de-

| Statistical analysis
All statistical analyses and graph generation were performed using R software (R Core Team, 2012), and the packages ggplot2, lmerTest, lme4, emmeans, ordinal and multcomp (Bates et al.,

F I G U R E 2
Bacterial population counts for three cherry pathogens, Pseudomonas syringae pv. morsprunorum race 1-C (Psm R1-C), P. syringae pv. morsprunorum race 2 (Psm R2) and P. syringae pv. syringae (Pss), after their inoculation at different concentrations (2 × 10 4 , 2 × 10 5 , 2 × 10 6 cfu/ml) into leaves of three sweet cherry cultivars. Boxplots show the day 10 population counts for cultivars that showed differential responses in the cut-shoot assay ( Figure  1). Individual data points are included and the arithmetic mean is shown with a black diamond. This experiment was performed once. There were significant differences between strains (p < 0.01, df = 2), concentrations (p < 0.01, df = 15) and an interaction between them (p < 0.01, df = 4), while cultivars were not significantly different in this analysis (p = 0.055, df = 2). Tukey HSD (p = 0.05, confidence level: 0.95) significance groups for the different strains at particular concentrations are shown as letters above the boxes [Colour figure can be viewed at wileyonlinelibrary.com] FIGURE 3 Use of leaf inoculation to screen wild cherry accessions for susceptibility to canker pathogens. (a) Boxplots of symptom scores from 52 wild cherry accessions and four sweet cherry cultivars (shaded in red) 10 days after inoculation with five Pseudomonas syringae strains. The strains P. syringae pv. avii (Psa; black), P. syringae pv. morsprunorum race 1-C (Psm R1-C; white), P. syringae pv. morsprunorum race 1-P (Psm R1-P; light grey), P. syringae pv. morsprunorum race 2 (Psm R2; dark grey) and P. syringae pv.

F I G U R E 4
Leaf inoculation-based screen of a range of Prunus species and hybrids (see Table 1 for full descriptions) for susceptibility to cherry canker pathogens. The boxplots show symptom scores 10 days after inoculation with four Pseudomonas syringae strains. The strains P. syringae pv. morsprunorum race 1-C (Psm R1-C), P. syringae pv. morsprunorum race 1-P (Psm R1-P),  (b) were also screened, the plot is shaded to show this (e.g., P. incisa E621 in (a) is the parent of three hybrids coloured in blue). P. avium 'Napoleon' (highlighted in red) was a parent of most hybrids (see Table 1 for more details and abbreviations). Ordinal regression analysis indicated that there were significant differences between cultivars (p < 0.01, df = 34) and strains (p < 0.01, df = 3). The accessions that showed significantly reduced symptoms across the strains compared to cherry cultivar Napoleon (P. av Nap) are marked with an asterisk. Symptoms were scored as 0, no symptoms; 1, limited browning; 2, browning To analyse the symptom score data from pathogenicity assays, the ordinal package was used, specifically the function clmm, which is optimized for ordinal data.

F I G U R E 6
Bacterial population counts of cherry pathogens Pseudomonas syringae pv. morsprunorum race 1-C (Psm R1-C), P. syringae pv. morsprunorum race 2 (Psm R2) and P. syringae pv. syringae (Pss) inoculated into leaves of sweet cherry (Penny, Sweetheart), wild cherry (Groton A and Groton B) and ornamental cherry (Prunus incisa). (a) Boxplots show the day 10 population counts for each strain on each cultivar after inoculation with 2 × 10 6 cfu/ml of each strain. Individual data points are included and the arithmetic mean is shown with a black diamond. This experiment was performed once. Analysis of variance revealed there were significant differences between strains (p < 0.01, df = 2) and cultivars (p < 0.01, df = 4) as well as an interaction between them (p < 0.01, df = 8). Tukey HSD (p = 0.05, confidence level: 0.95) significance groups for the whole data set comparison are labelled. (b) Representative pictures of disease symptoms for each strain × cultivar combination (images taken during initial screens documented in Figures 3 and 4); infiltration sites were within the four black pen marks. Note the lack of macroscopic lesions in P. incisa [Colour figure can be viewed at wileyonlinelibrary.com] F I G U R E 5 Screening of accessions of wild cherry (Groton A and B), ornamental cherry (Prunus incisa) and sweet cherry (Penny, Sweetheart) with 16 strains of the cherry canker pathogen Pseudomonas syringae. The boxplots show symptom scores 10 days after inoculation. The strains are coloured in shades of grey by clade: P. syringae pv. avii (Psa), P. syringae pv. morsprunorum race 1-C (Psm R1-C), P. syringae pv. morsprunorum race 1-P (Psm R1-P), P. syringae pv. morsprunorum race 2 (Psm R2) and P. syringae pv. syringae (Pss). Individual data points are included and the experiment was performed only once. Ordinal analysis confirmed differences between cultivars (p < 0.01, df = 4). Symptoms were scored as 0, no symptoms; 1, limited browning; 2, browning <50% of inoculated site; 3, browning >50% of inoculated site; 4, complete browning; 5, spread from site of inoculation [Colour figure can be viewed at wileyonlinelibrary.com]

| Partial resistance is seen in woody tissue but not leaf tissue of sweet cherry cultivars
Varietal resistance has been reported in sweet cherry under field conditions (Hulin, Mansfield, et al., 2018). To extend the range of sweet cherry cultivars screened for differences in resistance, detached shoot assays were conducted using representative strains from the three major canker-causing clades Psm R1, Psm R2 and Pss, as shown in Figure 1. The strain Psm R1-P, recognized as virulent on plum but not cherry (Hulin, Mansfield, et al., 2018), was also included (see full data Figure S1). Statistical analysis revealed significant differences in length of necrosis between cultivars (p < 0.01, df = 15), between strains (p < 0.01, df = 4) and an interaction between them (p < 0.01, df = 60).
Overall, cultivars showed a large degree of variability in the length of necrotic lesion produced, which meant that apparent differences in susceptibility of many cultivars were deemed not significantly different. However, cultivars such as Merton Glory and Colney showed partial resistance to all three of the major canker pathogens, with necrosis lengths significantly lower than in the most susceptible varieties such as Van and Roundel. We previously reported that the cultivar Merton Glory exhibited partial resistance to bacterial canker (Hulin, Mansfield, et al., 2018). All cultivars showed very limited susceptibility to Psm R1-P, the strain virulent on plum but less virulent on cherry ( Figure S1).
In an earlier study, detached leaf syringe-infiltration assays did not reproduce the quantitative differences seen in woody tissues of cherry varieties (Hulin, Mansfield, et al., 2018). To examine further the use of leaf inoculation to differentiate varietal resistance within sweet cherry, leaves of three cultivars that had F I G U R E 7 Bacterial population counts of cherry pathogens, Pseudomonas syringae pv. morsprunorum race 1-C (Psm R1-C), P. syringae pv. morsprunorum race 2 (Psm R2) and P. syringae pv. syringae (Pss), inoculated into leaves of wild cherry (Groton B), sweet cherry (Penny) and ornamental cherry (Prunus incisa) at different inoculum concentrations. (a) Boxplots show the day 0 population counts for cultivars. Individual data points are included and coloured for each separate experiment and the arithmetic mean is shown with a black diamond. This experiment was repeated up to four times per cultivar × strain combination. Analysis of variance (ANOVA) revealed a significant difference between strains (p < 0.01, df = 2), concentrations (p < 0.01, df = 2) and an interaction between them (p < 0.01, df = 4). There was no significant difference in bacterial populations between cultivars (p = 0.32, df = 2). Tukey HSD (p = 0.05, confidence level: 0.95) significance groups for the different strains at particular concentrations are presented. (b) Boxplots show the day 10 population counts for cultivars. The layout is the same as in (a). ANOVA revealed a significant difference between strains (p < 0.01, df = 2), cultivars (p < 0.01, df = 2), concentrations (p < 0.01, df = 2) and a cultivar × strain interaction (p < 0.01, df = 4), cultivar × concentration interaction (p < 0.01, df = 4) and strain × concentration interaction (p = 0.03, df = 4). (c) Symptom scores at day 10, scored as 0, no symptoms; 1, limited browning; 2, browning <50% of inoculated site; 3, browning >50% of inoculated site; 4, complete browning; 5, spread from site of inoculation. Data are presented as in (a) and (b). Ordinal analysis revealed a significant difference between strains (p < 0.01, df = 2), concentrations (p < 0.01, df = 2), cultivars (p < 0.01, df = 2) and a cultivar × concentration interaction (p < 0.01, df = 4) [Colour figure can be viewed at wileyonlinelibrary.com] varied in their response in the shoot assays (Figure 1), ranging from partially resistant to susceptible and highly susceptible, (Colney, Sweetheart and Van, respectively), were inoculated with progressively lower bacterial concentrations than the 10 6 cfu/ml used in earlier work. Bacterial population counts were determined after 10 days (Figure 2). There were significant differences between strains (p < 0.01, df = 2), concentrations (p < 0.01, df = 15) and an interaction between them (p < 0.01, df = 4). However, even from the lowest inoculum level, the different cultivars did not vary significantly in final bacterial populations 10 days postinoculation (p = 0.055, df = 2). The cultivar Colney, which had exhibited reduced susceptibility in the shoot assay, did not show any reduction in bacterial populations compared to Sweetheart and Van at any of the concentrations, although at the lowest, Psm R1 and Psm R2 grew to higher levels in Van than in the other cultivars. These experiments confirmed that, in these sweet cherry cultivars, leaf infiltration inoculations did not reproduce the differential susceptibility to canker scored using cut shoots.

| Wild cherry and other Prunus species exhibit leaf-based resistance to P. syringae
Although leaf inoculation assays did not reproduce the differential susceptibility observed in cut shoots of sweet cherry cultivars, in previous work the more tractable leaf tests did clearly demonstrate nonhost resistance to strains of P. syringae pathogenic on other plants (Hulin, Armitage, et al., 2018;Hulin, Mansfield, et al., 2018).
Therefore, we examined whether any leaf-based resistance could be found in the wider germplasm that would give levels of resistance to the cherry pathogens comparable to nonhost resistance.
Fifty-two wild cherry accessions, and four susceptible sweet cherry accessions for comparison, were screened using young leaves from mature trees (Figure 3). In initial experiments, Psm R1, Psm R2 from cherry and plum, and Pss were used for inoculation at 2 × 10 6 cfu/ml, and in the final screen, P. syringae pv. avii (Psa) was also included as this has been reported to be a pathogen of wild cherry (Ménard et al., 2003). The wild cherries exhibited a wide range of responses to the bacterial canker pathogens, from no, or very limited symptoms to complete necrosis of the inoculated region (see representative images of scores in Figure 3b). Results are presented in Figure 3a in order of the increasing severity of symptoms observed (mean overall symptom score per cultivar). Several accessions produced limited or no symptoms during this screening. In particular, the wild cherries P.a. Groton B, P. a. FD1-57-4/122, P. a. Deadmans Wood and P. a. Thruxton Vallets (numbered 23, 19, 16 and 48, respectively, in Figure 3a) were scored as highly resistant.
Ordinal statistical analysis confirmed that there were significant differences between accessions (p < 0.01, df = 55) and between strains (p < 0.01, df = 4). However, an interaction model could not be fitted due to complete separation of the response factor preventing model convergence (e.g., where in selected cases all scores were the same for a particular strain × cultivar combination) as discussed in Allison (2008). Nevertheless, in some genotypes there were clear differential reactions to the pathogenic strains (listed in Table S3).

For example, genotypes 15 (Coed-y-Stig) and 25 (Howley Wood)
showed resistance to Psa and Psm R1-P, respectively, but were susceptible to other strains. Sweet cherry cultivars were resistant to the plum strain Psm R1-P (graphs shaded in red in Figure 3a), but several wild cherries were susceptible, for example, 31 (Marlow Common 1902) and 21 (Frydd Wood 1908), the latter recording very little symptom development with the other strains. Another pattern to emerge was lesion formation following inoculation with Psa and Psm R1-C from cherry, but resistance to other strains as recorded F I G U R E 8 Bacterial populations following inoculation at 2 × 10 8 cfu/ml into leaves of Prunus incisa and sweet cherry cv. Sweetheart of a cherry pathogen Pseudomonas syringae pv. morsprunorum race 1-C (Psm R1-C) and two strains of P. syringae originating from different plants (from plum, P. syringae pv. morsprunorum race 1-P [Psm R1-P] and from Aquilegia vulgaris [RMA1]) that are nonpathogenic to cherry. Boxplots show the day 4 population counts for cultivars. Individual data points are included and the arithmetic mean is shown with a black diamond. This experiment was performed once. Analysis of variance revealed a significant difference between strains (p < 0.01, df = 2), cultivars (p < 0.01, df = 1) and an interaction between them (p < 0.01, df = 2). Tukey HSD ( incisa), as well as susceptible sweet (Penny and Sweetheart) and wild (Groton A) cherry cultivars for comparison, were screened with 16 previously characterized P. syringae strains pathogenic on cherry and plum ( Figure 5). The wild cherry Groton B generally recorded low levels of symptom development, but a tree from the same woodland, Groton A, was highly susceptible and comparable to the sweet cherry varieties ( Figure 3). This test with further strains confirmed that Groton B exhibited resistance, although some strains of Pss were able to cause lesions. Inoculation with each of the 16 strains tested failed to cause symptoms in the ornamental species P. incisa. Statistical analysis confirmed differences between cultivars (p < 0.01, df = 4), with Groton B and P. incisa recording significantly lower symptom scores with all pathogenic strains.

F I G U R E 9
Susceptibility of sweet (Colney, Merton Glory and Penny) and wild cherry (Groton B) cultivars to Pseudomonas syringae infection using cut shoots. Boxplots show length of necrotic lesions in cut shoots 6 weeks after inoculation with the cherry pathogens Pseudomonas syringae pv. morsprunorum race 1-C (Psm R1-C), P. syringae pv. morsprunorum race 2 (Psm R2), P. syringae pv. syringae (Pss) or the plum pathogen P. syringae pv. morsprunorum race 1-P (Psm R1-P). Individual data points are included. This experiment was performed once.

| Relationship between resistance response and bacterial inoculum dose
To see if the observed resistance in certain accessions was robust to increasing concentrations of bacterial inoculum, Groton B, P. incisa and the susceptible cultivar Penny were inoculated using increasing doses ranging from 2 × 10 6 cfu/ml to 2 × 10 8 cfu/ml (Figure 7). At day 0 (Figure 7a), there was no significant difference between bacterial numbers in accessions (p = 0.32, df = 2). At day 10, the wild cherry Groton B supported high bacterial populations of all pathogens when inoculated at 2 × 10 7 cfu/ml and 2 × 10 8 cfu/ml, with resistance only apparent at the lower inoculum concentration (Figure 7b).
By contrast, the ornamental species P. incisa recorded significantly reduced bacterial populations even when inoculated at 2 × 10 8 cfu/ ml with Psm R1 and Psm R2, although Pss appeared to overcome any resistance using the highest inoculum concentration. Symptom scoring in these experiments revealed that at the lower concentration (2 × 10 6 cfu/ml) Groton B and P. incisa recorded very limited symptom formation after 10 days (Figure 7c), confirming the results presented in Figures 3 and 4. By contrast, at the higher inoculum concentrations, symptoms were more apparent and similar to those observed in sweet cherry cv. Penny, particularly for the more virulent Pss.
The restriction of bacterial populations in P. incisa, particularly towards Psm R1 and Psm R2 at higher inoculum concentrations, was similar to a nonhost resistance response as seen previously in cherry towards plum and Aquilegia pathogens (Hulin, Mansfield, et al., 2018).
To examine if the multiplication of the sweet cherry pathogen Psm R1-C was similar to nonpathogens of cherry in P. incisa, several strains were inoculated at the highest inoculum concentration (2 × 10 8 cfu/ml) on P.
incisa and compared with a susceptible cherry, 4 days after infiltration ( Figure 8). The nonpathogens Psm R1-P from plum and RMA1 (a pathogen of Aquilegia) reached levels of 1 × 10 5 to 1 × 10 6 cfu/leaf disk in cherry cv. Sweetheart, while the pathogenic strain Psm R1-C grew 10 times higher. Psm R1-C did not grow as well in P. incisa where it reached levels of 1 × 10 5 to 1 × 10 6 cfu/leaf disk. However, the nonpathogens of cherry multiplied even less in P. incisa than they did in the sweet cherry. These results indicated that Psm R1-C may be more adapted to P. incisa than strains originating from unrelated plant hosts even though the ornamental cherry species still appears to have significant resistance.
Finally, to confirm if the resistance response of Groton B and P. incisa seen in leaves was reflected in woody tissue, a cut-shoot assay was performed ( Figure 9). Unfortunately, the P. incisa shoots were not amenable to this assay and dried out, probably due to their thinness. However, the assay confirmed that Groton B, like the more resistant sweet cherry cultivars Merton Glory and Colney, showed much reduced necrosis compared to the susceptible sweet cherry Penny.  (Crosse, 1966) and resistance in these tissues is of use for breeding programmes.

| DISCUSS ION
Although, the responses of sweet cherry cultivars tested could not be differentiated on leaves, we reasoned that relatives of sweet cherry might exhibit resistance in nonwoody tissues as seen in previous work (Vicente & Roberts, 2003 (Garrett, 1979). A sibling of FD1-57-4/122, FD1-57-4/166, was also found to be more resistant in plantlet assays (Vicente & Roberts, 2003).
Differential symptom development in some accessions also suggests the existence of a pattern of resistance and susceptibility, as ob- served in examples of race-and cultivar-specific resistance in other plant-bacterium interactions, for example in bean halo blight disease (Arnold et al., 2011). Differential reactions observed are listed in Table   S3, but no simple model based on the presence of R genes matching each clade could be fitted to the data. The reactions observed to the plum strain Psm R1-P are of particular interest. Resistance to Psm R1-P in sweet cherry could be due to resistance triggered by the intracellular detection of pathogen effectors such as HopAB1 by the plant immune system. Genomic analysis revealed the hopAB1 effector gene is present in this strain but not its relatives that are pathogenic on cherry (Hulin, Armitage, et al., 2018;Hulin, Mansfield, et al., 2018). Several wild P.
avium accessions were susceptible to infection by the plum strain, developing distinct lesions, and presumably these accessions could lack a receptor recognizing HopAB1, such as Pto in tomato species (Chien et al., 2013). The role of HopAB1 as an inducer of effector-triggered immunity and/or a virulence determinant should be tested by genetic dissection through deletion of hopAB1 from Psm R1-P.
The study was extended to other Prunus species and sweet cherry hybrids. In particular, some Prunus species also displayed resistance to the major pathogen strains, and the Fuji cherry accession P. incisa proved resistant to all 16 canker pathogens tested. The resistance suggested by lack of symptom development in wild cherry and related Prunus spp. was confirmed through analysis of bacterial multiplication in leaves. Bacterial populations reached in P. incisa were lower than those recorded in the selected wild cherry accession Groton B. The dynamics of population growth in P. incisa were similar to those recorded for nonpathogens in sweet cherry. The similarly reduced multiplication of the nonhost plum and Aquilegia pathogens in P. incisa compared with sweet cherry indicates that there may be a more rapid deployment of resistance, perhaps mediated through an enhanced level of cell surface-based immunity and/or effector-triggered intracellular responses. Whatever the biochemical nature of resistance, the lack of symptoms found in the hybrids between P. incisa and the sweet cherry cv. Napoleon after challenge with the major pathogens suggests that the resistance from P. incisa is probably inherited as a dominant trait.
The resistant wild cherry and Prunus accessions selected, Groton B and P. incisa, respectively, have now been incorporated into breeding programmes to introgress resistance into commercial sweet cherry genotypes and generate more resistant varieties for growers. Progeny of Groton B have also been selected for wild cherrybreeding programmes to improve canker resistance in the forestry industry (K. Russell, unpublished data). Such work can take up to 15 years. The routine testing of progeny performance against the main canker pathogens during these projects and future genetic research will provide further insights into the genetic controls underlying the outcome of the Prunus-P. syringae interaction.

ACK N OWLED G EM ENTS
We thank Ken Tobutt and Marzena Lipska for generation of hybrids and propagation and maintenance of these tree resources, which are important genetic material for future use. We are also grateful to the East Malling Farm and Glass team for propagation and maintenance of plant material throughout this work. We thank Joana Vicente and Steven Roberts for the provision of bacterial strains used in this study. We are grateful to Greg Deakin for assistance with statistical analysis for this manuscript. Finally, we thank the University of Sassari Erasmus + Programme for funding F.C.'s traineeship.

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
The data that support the findings of this study are available from the corresponding author upon reasonable request.