A first assessment of Fraxinus excelsior (common ash) susceptibility to Hymenoscyphus fraxineus (ash dieback) throughout the British Isles

Ash dieback (ADB), caused by Hymenoscyphus fraxineus, has severely damaged a large proportion of ash trees (Fraxinus excelsior) in continental Europe. We have little damage data for the British Isles where the disease was found only five years ago in the Southeast, and is still spreading. A large-scale screening trial to evaluate ADB damage to provenances of F. excelsior sourced from throughout the British Isles was planted in 2013 in the southeast of England. In 2016, we scored trees by their level of ADB damage observed in field at the two worst affected (based on assessments in 2015) of the 14 sites. Significant differences were found in average ADB damage among planting sites and seed source provenances. Trees from certain provenances in Scotland were the least damaged by ADB, whereas trees from Wales and Southeast England were the most badly damaged in both trial sites. Thus the levels of ADB damage currently seen in ash populations in Southeast England may not be an accurate predictor of the damage expected in future throughout the British Isles. Given all provenances contained some healthy trees, a breeding programme to produce genetically variable native ash tree populations with lower ADB susceptibility may be feasible.

In the British Isles, H. fraxineus inoculum levels and damage are currently worst in the Southeast of England. The fungus is still spreading though the North and West of Great Britain, and in Ireland. For policy makers to manage the effects of the epidemic, we need accurate predictions of the severity of future damage. In the British Isles this is difficult as ash populations in the North and West appear to be derived from a Pleistocene refugium in Iberia, whereas populations in the south and east are more closely related to trees in France, Germany and the Low Countries [27][28][29][30] . Sutherland et al. found two chloroplast haplotypes that were widespread in Great Britain, and three that were found only in Scotland, but not the Northwest of Scotland, which had a more common British haplotype 31 . This means the future impact of ADB may vary for different locations in the British Isles. Predictions using transcriptome markers 32 have suggested that ash trees in the North and West of Great Britain may be less damaged by ash dieback than Danish populations 30 .
Here, we provide the first in-field assessments of the damage caused by ash dieback to genotypes of ash sourced from locations across the British Isles. Our results are based on screening trials of ash saplings from British, Irish and continental provenances set up in 2013 in the Southeast of England. We report detailed assessments of ADB damage from the autumn of 2016, after three and a half year's exposure to natural H. fraxineus inoculum.

Methods
Trial design. This study is based on a Forest Research screening experiment planted in spring 2013, comprising 48 hectares of trials on 14 sites in Southeast England. Saplings were sourced from five nurseries in the UK (Supporting Information Table S3) and each site was planted with trees grown from seed sourced from up to 15 different provenances. These were 10 British native seed zones ( Table S1). Each provenance had two to six complete replications per site, in 16 m × 16 m blocks of 256 trees, randomly distributed (Supporting Information Table S2). All saplings were inspected, and in some cases tested, to ensure they were disease free at the time of planting. In September 2014, establishment of the saplings in all sites was surveyed by Forest Research. In autumn 2015, all sites were surveyed by Forest Research with each tree scored as 1 -dead or missing, 2 -alive and infected and 3 -alive and healthy (Forest Research, unpublished data).
Detailed Phenotyping. In autumn 2016 we scored the ash dieback symptoms of 28,160 trees at the two trial sites that had showed highest damage in 2015, which were located near Norwich in Norfolk (Site 16) and near Royal Tunbridge Wells in East Sussex (Site 35). We considered that these sites had the highest inoculum pressure, and would be the most reliable sites for gaining a preliminary understanding of the susceptibility of different provenances to ADB, as fewer trees would be likely to have simply avoided infection by chance. We used the methodology of Pliura et al. to score the damage phenotypes 17 of the trees in these two sites. This method uses a scale of 1-7 (Supporting Information Fig. 1S, Supporting Information Fig. 2S) as follows: 1 -missing tree; 2 -dead, dry tree without significant development; 3 -dry tree that produced shoots before dying; 4 -heavily damaged living tree with dead main stem and resprouted stems with highly visible leaf and stem ADB lesions; 5 -moderate damage with dry leading shoot and once or repeatedly resprouted leading shoot with highly visible leaf and stem ADB lesions; 6 -limited damage with multiple brown dry or wilted leaves or peripheral shoot or/ and brown lesions on stem or branches; 7 -healthy tree with minimal signs of damage only on leaves (brown, dry or wilted leaves). We later introduced a category of zero for trees that were missing both in autumn 2016 and autumn 2014: these trees were subtracted from score one.
We analysed the damage scores using a general linear model in a factorial ANOVA with score as the dependent variable and site and seed source provenances as categorical predictors (factors). For this analysis, seed source provenances that did not have an equal amount of replication (NSZ 403 has only three blocks at Site 16) and seed source provenances that are not present at both sites (DEU, FRA and CLARE only present at Site 16) were excluded. Score zero and one (missing trees) were excluded from our initial analyses since trees in both categories most likely died for reasons other than ADB. In case score one trees has in fact died from ash dieback, we repeated the analyses including score one trees and only excluding score zero trees. We then analysed each site separately, including all seed source provenances present at that site and conducting a post-hoc pairwise comparison     (Tukey's test) to find significant differences between provenances; this was done twice, once with score one and zero trees excluded, and then with only score zero trees excluded.
Data availability. The dataset generated and analysed during the current study are available as Supplementary Data.

Results
The percentages of trees in each damage class, site and from each provenance are shown in Fig. 1a,b. Site 35 had a higher percentage (20.3% Score zero and 3.3% Score one) of missing trees than Site 16 (1.1% Score zero and 1.7% Score one) for all provenances (Fig. 1b, Supporting Information Fig. S3). We are convinced that this was due to poor establishment at Site 35 and not caused by ash dieback for three reasons: (1) The majority of missing trees had disappeared by autumn 2014, (2) because the trees that did establish well at Site 35 were less badly affected by ash dieback in 2016 than the trees at Site 16 (see below), suggesting higher inoculum pressures at Site 16, and (3) the missing trees at Site 35 were spatially clustered in a wide strip suggestive of a soil condition problem, such as poor drainage, rather than the spread of wind-borne spores. We therefore excluded Score zero trees from our analyses of ash dieback damage at both sites. For Score one, where trees were present in 2014 but missing by 2016, we could not be sure if they had died from ADB or not, so we analysed the data twice, one including and one excluding them. The two sites showed significant (p ≤ 0.001) differences between them in level of ADB damage ( Table 1). The average score (excluding Score zero and one trees), for Site 35 was 5.05 (S.D. = 1.28) and for Site 16 was 3.76 (S.D. = 1.12). When only Score zero trees were excluded, in case Score one trees had in fact died from ADB, the same pattern was found with the average score for Site 35 being 4.89 (S.D. = 1.49) and for Site 16 being 3.72 (S.D. = 1.17). This pattern suggests that site 35 has experienced a lower inoculum pressure than site 16, perhaps partly due to a lower density of established trees, and is therefore at an earlier stage of infection. Average mortality that can be definitely attributed to ash dieback (Scores two and three) at Site 16 was 27% and at Site 35 was 7%, with an average across both sites of 18%. Average percentage of healthy trees (Score seven) at Site 16 was 3.1% and at Site 35 was 2.9%.
Provenance was a significant factor for ash dieback damage when differences between sites were accounted for (Table 1). This was true whether or not Score one trees were included in the analysis. Pairwise comparisons of provenance scores are shown in Table 2, where both Score zero and score one trees are excluded. The Scottish provenance NSZ 106 had the highest average score at both sites, being 3.97 at Site 16 (Table 2a & Fig. 2a,c) and 5.39 at Site 35 (Table 2b and Fig. 2b,d). Mortality figures for NSZ 106 are amongst the lowest for all provenances (Table 2a,b and Table 3a,b). Overall, NSZ106 is significantly different (p ≤ 0.05) to eight of the fifteen provenances at Site 16 (Table 2a) and eight of the twelve provenances at Site 35 (Table 2b). The British southeastern provenance NSZ 405 had the lowest average score at both sites (Table 2a,b & Fig. 2a,b) with the highest mortality (Table 2a,b) for all provenances. The BSO FTT provenance, which is comprised of superior trees for commercial forestry, was found to have low (Site 35) to intermediate (Site 16) average scores in this study (Table 2a,b). Performance analysis for FRA, DEU and CLARE can only be made at Site 16, as they were not planted at Site 35. Interestingly, CLARE has very few trees at the upper end of the scoring scheme (0% Score six and 1% Score seven) (Fig. 1a), a large percentage of highly damaged trees (73% Score four) but also lower mortality (Scores two and three) at 21% (Table 2a). The German provenance (DEU) has the second highest percentage of Score six (limited damage) trees and performed well overall, with an average score of 3.83 at Site 16 (Table 2a).
When Score one trees were included in the analysis, in case they had died from ADB, the overall pattern of results was similar for Site 16. At Site 35, two provenances, NSZ 204 and NSZ 302 decreased in their average score (18% and 7% respectively), altering their ranking with respect to other provenances at the site (Table 3 and Fig. 2d).

Discussion
We found significant differences among provenances of ash trees in their susceptibility to ash dieback in the British Isles. Trees from the middle of Scotland appeared least susceptible and interestingly, these areas were also found, by a previous study, to contain rare chloroplast haplotypes 31 . Sutherland et al. suggested that these were sourced from a separate glacial refugium to other ash populations in Great Britain 31 .
The most southeasterly British ash provenance in our experiment was consistently amongst the most damaged. Trees from this provenance sampled in Sutherland et al. had the same chloroplast haplotype as most of Great Britain 31 , but the study of Heuertz et al. showed this provenance to also contain haplotypes present in France, Switzerland and Italy 27 . Thus this provenance may be more closely related to the French ash trees that have also shown high mortality in this study. Our results suggest that current high damage due to ADB in the Southeast of England may be greater than what we might expect to see throughout the rest of the British Isles in the coming years as ADB continues to spread. In particular, parts of Scotland may be less badly affected.
Our results suggest that a few ash trees with low susceptibility to ADB are present in most provenances from across the British Isles. Similar results have been found in Europe 15,19,22,23,33 . It should be remembered that this study represents just two of the 14 sites planted in 2013 and the trees are only four years old, but these early findings suggest that the prospects for selective breeding for lower susceptibility to ADB in a variety of native provenances are good. The suite of Forest Research trial sites will provide valuable starting materials for such breeding programmes in the near future.