Environmental variables associated with Nothophaeocryptopus gaeumannii population structure and Swiss needle cast severity in Western Oregon and Washington

Abstract The environment has a strong influence on the abundance and distribution of plant pathogenic organisms and plays a major role in plant disease. Climatological factors may also alter the dynamics of the interactions between plant pathogens and their hosts. Nothophaeocryptopus (=Phaeocryptopus) gaeumannii, the causal agent of Swiss needle cast (SNC) of Douglas‐fir, is endemic to western North America where it exists as two sympatric, reproductively isolated lineages. The abundance of this fungus and the severity of SNC are strongly influenced by climate. We used statistical and population genetic analyses to examine relationships between environment, pathogen population structure, and SNC severity. Although N. gaeumannii Lineage 2 in western Oregon and Washington was most abundant where SNC symptoms were most severe, we did not detect a significant relationship between Lineage 2 and disease severity. Warmer winter temperatures were inversely correlated with foliage retention (AFR) and positively correlated with the relative abundance of Lineage 2 (PL2). However when distance inland, which was strongly correlated with both AFR and PL2, was included in the model, there was no significant relationship between Lineage 2 and AFR. Spring/early summer dew point temperatures also were positively associated with total N. gaeumannii abundance (colonization index (CI)) and inversely correlated with AFR. Warmer summer mean temperatures were associated with lower CI and higher AFR. Our results suggest that the two lineages have overlapping environmental optima, but slightly different tolerance ranges. Lineage 2 was absent from more inland sites where winters are colder and summers are warm and dry, while Lineage 1 occurred at most sites across an environmental gradient suggesting broader environmental tolerance. These relationships suggest that climate influences the abundance and distribution of this ecologically important plant pathogen and may have played a role in the evolutionary divergence of these two cryptic fungal lineages.


| INTRODUC TI ON
Climate plays a major role in influencing the geographic distributions of plant pathogens and their hosts and may lead to changes in host-pathogen dynamics (Sturrock et al., 2011). Climate change is predicted to directly affect pathogens most strongly, and fungi causing foliage diseases in particular are most likely to be influenced by environmental variables (Harvell et al., 2002). Outbreaks of damaging forest pathogens such as Dothistroma septosporum, which causes foliage loss, growth reductions, and mortality of native and introduced Pinus species, have been linked to regional weather patterns (Kliejunas et al., 2009;Welsh, Lewis, & Woods, 2014). The recent emergence of this major forest disease and others around the globe may be attributable, at least in part, to climate change (Harvell et al., 2002;Kliejunas et al., 2009;Welsh et al., 2014;Woods, Coates, & Hamann, 2005;Woods et al., 2016). Microbial pathogens and insects, because of their shorter generation times, respond more rapidly to environmental changes than forest communities and may amplify the effects of climate change on forest health (Kliejunas et al., 2009). Changes to forest disturbance regimes due to the influences of climate on the interactions between disease, fire, and insects are expected to impact the future ecological trajectories of forested landscapes (Agne et al., 2018;Sturrock et al., 2011).
Although the disease was first described from a Douglas-fir plantation in Switzerland in 1925, Nothophaeocryptopus gaeumannii is presumed to be native to the Pacific Northwest (Hansen et al., 2000) where the population is subdivided into two reproductively isolated lineages (Bennett & Stone, 2016;Winton, Hansen, & Stone, 2006).
Lineage 1 occurs throughout the natural range of Douglas-fir and worldwide where Douglas-fir is grown as an exotic (Winton et al., 2006). Lineage 2 has a more restricted distribution (Bennett & Stone, 2016;Winton et al., 2006) and is most abundant along the western Coast Ranges in Oregon and Washington within a few kilometers of the coast. The abundance of Lineage 2 decreases further inland, where it is often supplanted entirely by Lineage 1 (Bennett & Stone, 2016;Winton et al., 2006). The factors influencing the spatial distributions of these lineages are not currently understood. Previous studies have identified strong genetic differentiation between these two lineages (Bennett & Stone, 2016;Winton et al., 2006). Despite the fact that there is circumstantial evidence for reproductive isolation between the two lineages, there is currently insufficient evidence to describe them as distinct biological or phylogenetic species.
Prior to the late 1970s, SNC had not caused any serious damage in Pacific Northwest forests (Hansen et al., 2000). However, symptoms of SNC have been intensifying in this region since 1990s, and a 2016 aerial survey identified approximately 200,000 hectares of affected land in Oregon (Hansen et al., 2000;Ritóková et al., 2016). Its proliferation and subsequent emergence as a threat to Douglas-fir forest health and productivity in western North America is thought to have been perpetuated by the widespread planting of Douglas-fir where the climate is particularly conducive to disease in a narrow band of low-elevation coastal forests along the western slopes of the Coast Ranges in Oregon and Washington that has historically been considered the Picea sitchensis vegetation zone (Franklin & Dyrness, 1973;Hansen et al., 2000).
Because the distribution of Lineage 2 in the western Coast Ranges in Oregon and Washington corresponds to the region where the most severe SNC symptoms (foliage loss and growth reduction) have been documented, a causal relationship between Lineage 2 and the recent intensification and expansion of SNC in this region has been suggested (Winton et al., 2006). Compared to Lineage 1, recovery of Lineage 2 isolates from diseased foliage was twice as likely in severely diseased stands, and only half as likely in healthier stands (Winton et al., 2006). In coastal Oregon, Douglas-fir stands with higher relative proportions of Lineage 2 had more sparse canopies, suggesting less foliage retention due to SNC (Winton et al., 2006). Visual estimates of foliage discoloration as an indicator of disease severity also suggested a causal relationship with the relative abundance of Lineage 2 in the stand (Winton et al., 2006).
Isolates of N. gaeumannii collected from sites with severe disease also seemed to cause more severe SNC symptoms in an inoculation study, suggesting that more virulent or aggressive genotypes may be more prevalent in the low-elevation coastal forests in Oregon and Washington where SNC is most severe (Winton, 2001). These observations informed our hypotheses about the potential relationships between the distribution of N. gaeumannii Lineage 2 and SNC severity.

The regional climate in the western Coast Ranges in Oregon and
Washington has a strong influence on the distribution and severity of SNC, and variation in N. gaeumannii abundance and SNC symptom severity in relation to site-specific climate factors has been well documented (Coop & Stone, 2010;Hansen et al., 2000;Lee, Beedlow, Waschmann, Burdick, & Shaw, 2013;Lee et al., 2017;Manter, Reeser, & Stone, 2005;Rosso & Hansen, 2003;Stone, Coop, et al., 2008;Stone, Hood, Watt, & Kerrigan, 2007;Watt, Stone, Hood, & Palmer, 2010;Zhao, Maguire, Mainwaring, & Kanaskie, 2012;Zhao, Mainwaring, Maguire, & Kanaskie, 2011). Winter temperature consistently has been identified as being strongly correlated with N. gaeumannii abundance, and thus SNC severity, given that needle colonization and the development of pseudothecia continue throughout winter following the spring/summer infection period (Capitano, 1999;Manter et al., 2005;Stone, Capitano, & Kerrigan, 2008). Colder winter temperatures limit the colonization of needles by N. gaeumannii (Manter et al., 2005). Leaf wetness and free moisture (as precipitation, fog, or dew) during the spring and early summer are necessary for spore dispersal, adhesion, and germination on the needle surface (Capitano, 1999;Manter et al., 2005), and thus are highly influential in predictive spatial models of SNC severity when severity estimates are based on the abundance of N. gaeumannii. Temperatures above 30°C inhibit the growth of N. gaeumannii, and thus, warmer average summer temperatures also are associated with less severe SNC symptoms (Capitano, 1999;Lee et al., 2017;Rosso & Hansen, 2003;Zhao et al., 2011). Predictive models based on combinations of these factors explain much of the variability (R 2 = 57%-78%) in SNC severity in western Oregon and Washington forests (Lee et al., 2017;Manter et al., 2005;Rosso & Hansen, 2003;Stone, Coop, et al., 2008). A trend of increasing mean winter temperatures and spring precipitation in the Pacific Northwest in recent decades has resulted in conditions conducive to the intensification and expansion of SNC (Abatzoglou, Rupp, & Mote, 2014;Lee et al., 2017;Stone, Coop, et al., 2008).
Our objectives for this study were to (a) assess the spatial distributions of the two N. gaeumannii lineages in relation to SNC severity in the Oregon (OR) and Washington (WA) Coast Ranges, (b) determine whether any relationship exists between SNC severity and the relative abundance of N. gaeumannii Lineage 2, and (c) examine the relationships between the genetic structure of N. gaeumannii populations, SNC severity, and key environmental variables with a multivariate statistical ordination. Given previous observations of the relationship between the spatial distribution of N. gaeumannii Lineage 2 and SNC severity, we aimed to test the hypothesis that the relative abundance of Lineage 2 within sites is positively correlated with disease severity. The analyses performed here were also designed to test the hypothesis that SNC severity and the spatial distributions of the two lineages (and thus the spatial genetic structure of N. gaeumannii populations) are correlated with environmental variables that have been previously identified as being critical for the development of N. gaeumannii.

| Foliage sampling
Douglas-fir foliage was collected from 23 sites in the western Oregon Coast Range in a plot network maintained by the Oregon State University Swiss Needle Cast Cooperative (SNCC) (Ritóková et al., 2016;Shaw, Filip, Kanaskie, Maguire, & Littke, 2011). For these sites, the sampling was concurrent with annual disease assessments Beach, OR, the sampling sites were generally arranged in transects stratified by distance inland, with sites ranging from the shoreline to 56 km inland (Figure 1). At each site, foliage was collected from second-and third-year internodes on secondary branches in the upper crowns of five randomly selected 10-to 30-year-old Douglasfir trees. From one of the five trees sampled at each of the SNCC sites, foliage samples were also collected from the lower, middle, and upper crowns to assess within-tree diversity. The foliage was stored on ice and promptly returned to the campus of Oregon State University for storage in a cold room for no longer than 7 days prior to processing.

| Isolation of N. gaeumannii from Douglasfir Foliage
The protocol for isolating N. gaeumannii from infected Douglas-fir needles is described in detail in Bennett and Stone (2016). Briefly, needles with pseudothecia of N. gaeumannii were attached to the lids of Petri dishes with double-sided adhesive tape, placed over water agar, and incubated for 48-72 hr. Individual ascospores were removed from the agar with sterilized forceps and transferred onto 2% malt agar (MA) (Difco Laboratories). Cultures were F I G U R E 1 Genetic distance dendrogram (UPGMA, unweighted pair-group method with arithmetic means) from bootstrap analysis of Nei's genetic distance showing divergence between genotypes corresponding to two Nothophaeocryptopus gaeumannii lineages from Oregon and Washington (clone-censored, N = 663). Node labels represent bootstrap statistics (≥ 70%) from 10,000 replicate trees incubated at 18°C for a minimum of 2 months, and often up to 6 months.

| Molecular techniques
The protocols for DNA extraction, PCR amplification, and genotyping of simple sequence repeats (SSRs) are described in Bennett and Stone (2016). The DNeasy Plant Kit (Qiagen) was used to extract genomic DNA from vegetative mycelium. The manufacturer's protocol was modified with the addition of a maceration procedure (Bennett & Stone, 2016). For each isolate, ten SSR loci (Winton, Stone, & Hansen, 2007) were amplified in three multiplexed PCR reactions (Bennett & Stone, 2016). The Qiagen Type-It Microsatellite PCR kit (Qiagen, Hilden, Germany) and its associated protocols were used, but with reaction volumes of 12.5 µl. The amplification was performed with the thermal cycling protocol described in Bennett and Stone (2016 Corporation) and were also visually examined for each locus and isolate to confirm accuracy. A positive control isolate was included with each PCR and genotyping run to ensure that the results were consistent and reproducible. One of the loci, Pgdi5, did not amplify reliably for all isolates and therefore was omitted from the dataset prior to analysis. The actual number of trees per site represented by the isolates in the SSR dataset was often less than five due to a number of steps that may have resulted in sample loss, such as variation in the presence of pseudothecia, contamination during isolation and culturing, and inconsistent amplification during PCR. The genotypes used for this study included those analyzed in Bennett and Stone (2016) (i.e., those sampled in 2014) along with additional isolates that were collected in 2015 and 2016.

| Data analysis
All statistical and population genetic analyses were performed with the R statistical computing software version 3.4.1 (R Core Team, 2017). The multilocus genotypes (MLGs) consisting of alleles for the 9 SSR loci were formatted in Microsoft Excel 2016 with GenAlEx 6.503 (Peakall & Smouse, 2012. The MLGs were then imported into R version 3.4.1 (R Core Team, 2017) with the R package poppr version 2.5.0. The MLGs were organized in a stratified population hierarchy that included levels for lineages, states within lineages, sites within states, and trees within sites.
Estimates of genetic diversity, including genotypic diversity (Shannon-Weiner index, H) and Nei's unbiased gene diversity (expected heterozygosity, H e ) (Nei, 1978), were calculated with the R package poppr 2.5.0 (Kamvar et al., 2015(Kamvar et al., , 2014. The Shannon-Weiner diversity index, H, was estimated from 1,000 iterations of a bootstrap analysis with rarefaction. Genotypic richness was estimated as the number of expected multilocus genotypes (eMLG) with a rarefaction sample size ≥ 10 (Kamvar et al., 2014).
The membership of each isolate in either Lineage 1 or Lineage 2 was visualized as a UPGMA dendrogram constructed from 10,000 bootstrap replicates of Nei's genetic distance (Nei, 1978). This analysis was performed with the R packages poppr 2.5.0 (Kamvar et al., 2015(Kamvar et al., , 2014 and visualized with ggtree version 1.12.0 (Yu, Smith, Zhu, Guan, & Lam, 2017). Genetic differentiation between the two N. gaeumannii lineages was estimated by calculating a standardized differentiation metric, G′ ST (Hedrick, 2005), with the R package mmod (Winter, 2012). This measure is scaled such that the value for two populations that do not share any alleles is one, and the value for population pairs that share all alleles is equal to zero (Hedrick, 2005). This metric is also corrected for the observed population heterozygosity and the number of subpopulations being compared (Hedrick, 2005;Winter, 2012). If we assume that the rate of gene flow between populations is higher than the mutation rate within populations, indices such as G′ ST can provide a reasonable estimate of migration (Balloux & Lugon-Moulin, 2002;Hedrick, 2005).
Aerial disease survey data were obtained from the Oregon Department of Forestry (https ://www.oregon.gov/ODF/Fores tBene fits/Pages/ Fores tHeal th.aspx) (Ritóková et al., 2016), and from the Washington Department of Natural Resources (Ramsey, Omdal, Dozic, Kohler, & Boderck, 2015). Researchers with the Oregon State University Swiss Needle Cast Cooperative (SNCC, http://sncc.fores try.orego nstate.edu/) and the Washington Department of Natural Resources provided estimates of both average foliage retention (AFR) and colonization index (CI) for sites in OR and WA (Ramsey et al., 2015;Ritóková et al., 2016). These disease severity metrics were measured using methods similar to those described in Manter et al. (2005) and Watt et al. (2010). For this study, AFR was expressed as the average percentage of foliage remaining across four needle age classes from secondary branches collected in the midcanopies of 10 trees from each site. An estimate of the average percentage of stomata occluded by pseudothecia, CI, was calculated as the product of incidence (the proportion of needles bearing pseudothecia, N = 50 needles) and the percentage of stomata occluded by pseudothecia (N = 10 needles). The percentage of stomata occluded was calculated by averaging the numbers of pseudothecia from 100 stomata in each of three sections per needle (base, middle, and tip) for each sample of ten 2-year-old needles from each of three canopy sections per tree (lower, middle, and upper) for ten trees from each site.
Pearson correlation coefficients were calculated for all pairwise correlations between geographic and disease variables (for the 34 sites for which disease severity data were available) with the R package Hmisc version 4.1-1 (Harrell, 2017). Scatter plots associated with these statistical analyses were constructed with ggplot2 (Wickham, 2016). To account for the possibility that Douglas-fir foliage retention, SNC severity, and the relative proportion of N. gaeumannii Lineage 2 may vary independently along the west-east sampling gradient (due to climatic and other spatial geographic factors), a linear mixed model was utilized to investigate the influence of the proportion of Lineage 2 on AFR with the distance inland from the coast held at its mean value (26 km). Distance inland is related to continentality and serves as a proxy for a complex combination of environmental variables (i.e., average temperature, precipitation, RH, dew point deficit, fog, wind speed, and other unknown factors) that are expected to influence Swiss needle cast severity and foliage retention (Zhao et al., 2011). The following model was designed to test the null hypothesis that the proportion of Lineage 2 had no effect on AFR after accounting for distance inland (H 0 : where Y i is the average foliage retention of the i th site. The intercept β 0 is the mean average foliage retention when the relative proportion of Lineage 2 recovered at the site is 0 and the distance inland is 0 km, β 1 is the coefficient for the effect of the relative proportion of Lineage 2 on mean average foliage retention, X 1i is the relative proportion of Lineage 2 recovered from the i th site, β 2 is the coefficient for the effect of the distance inland (km) on mean average foliage retention, X 2i is the distance inland (km) of the i th site, and ε i is the random effect of the i th site on mean average Nonmetric multidimensional scaling (NMDS) was used to visualize genetic differentiation between each of the sample sites.
Roger's euclidean genetic distance (Rogers, 1972) was calculated pairwise between each of the sampling sites with the R package adegenet (Jombart, 2008). The NMDS ordination based on this genetic distance matrix was performed with the function metaMDS in the R package vegan version 2.4-5 (Oksanen et al., 2017). This method ranks sample units according to their dissimilarity and then attempts to minimize the stress in the relationship between ordination distances and genetic distances (McCune, Grace, & Urban, 2002). Correlations between the ordination, SNC severity, and the environmental/geographic variables associated with each site were calculated with the function envfit from the R package vegan version 2.4-5 (Oksanen et al., 2017).
The environmental and geographic overlays were displayed as a series of radiating vectors, with the direction of the vector corresponding to its relationship to the ordination axes, and the length of the vector proportional to the strength of the correlation between the variable and the ordination. Included as vectors on the joint plot were SNC severity (AFR and CI), the relative proportion of Lineage 2, latitude, longitude, elevation, and several environmental variables related to SNC severity and N. gaeumannii abundance (Lee et al., 2017;Manter et al., 2005;Rosso & Hansen, 2003;Stone, Coop, et al., 2008;Stone et al., 2007;Watt et al., 2010;Zhao et al., 2011). Interpolated spatial climatic data were obtained for each of the sample sites from the Parameter-ele-

| Genetic diversity and population structure
This study included a total of 663 unique multilocus genotypes Total gene diversity (H e ) for the 1,061 isolates was 0.82 (Tables   1 and 2). There were 403 distinct Lineage 1 MLGs and 260 distinct Lineage 2 MLGs (Table 1). The number of private alleles was 88 for Lineage 1 and 55 for Lineage 2. Overall, Lineage 1 had greater genotypic diversity, genotypic richness, and gene diversity than Lineage 2, even after correcting for the difference in sample sizes by rarefaction (Table 1). The two N. geaumannii lineages were strongly differentiated (G′ ST = 0.941) ( Table 1). In the UPGMA dendrogram constructed from a bootstrap analysis of Nei's distance, isolates from the two lineages clustered into two distinct groups ( Figure 1). The branch lengths and topologies reflected strong genetic differentiation between the lineages and high genetic diversity within each lineage. Of the 657 isolates collected in Oregon, 384 had distinct MLGs, and of the 404 isolates  e Total MLG not equal to sum of population totals due to shared MLGs. The genotypes of isolates collected in 2014 were analyzed previously in Bennett and Stone (2016). *Estimated genotypic diversity from 1,000 bootstrap replicates with rarefaction (n ≥ 10).
TA B L E 2 (Continued) collected in Washington, 282 had distinct MLGs (Table 2). There were 39 private alleles in the Oregon population and 24 in the Washington population.

| Spatial distributions of N. gaeumannii lineages in relation to SNC severity
There was a strong association between the geographic distributions of N. gaeumannii Lineage 2 and SNC symptom severity assessed by aerial surveys (Figure 2). With few exceptions, sites nearest the coast had the highest proportions of Lineage 2 and occurred in areas where moderate-to-severe SNC symptoms were observed in the aerial surveys. Sites further inland, where Lineage 2 was generally rare or absent, had less severe SNC, or symptoms were not visible ( Figure 2). However, aerial surveys did not detect symptoms of SNC in southwestern Oregon, where only Lineage 2 was isolated from foliage samples and Lineage 1 was not recovered ( Figure 2).

| Variation in geography, environment, and disease severity across sites
The 35 sampling sites in Oregon and Washington covered a latitudinal range from 42°N to 48°N, and a longitudinal range from -124.6°W to -123.3°W (Table 3). Sampling sites ranged from 16 to 595 m elevation and were located from 1.8 to 57.6 km inland (  CI was at site T5-3 (158 m elevation, 16.7 km inland; Table 2). The sites with the least severe SNC symptoms were at higher elevations and further inland. The highest AFR was at site N25-5 (251 m elevation, 51 km inland), and the lowest CI was at T25-2 (591 m elevation, 49.5 km inland; Table 2).

| Correlations between environment, disease, and the genetic structure of N. Gaeumannii populations
Pearson's coefficient (r) was calculated pairwise between the environmental, geographic, and disease variables in our dataset. PL2 was negatively correlated with distance inland (Figure 3) and elevation (Table 4), and was positively correlated with mean winter temperature (Table 4). AFR was positively correlated with elevation and distance inland (Figure 3, Table 4). CI was negatively correlated with elevation and distance inland (Figure 3, Table 4) and was positively correlated with MADT (Table 4). AFR was negatively correlated with the relative proportion of N. gaeumannii Lineage 2 isolates recovered from the site (PL2) (Figure 3), CI ( Figure 5), and MADT (Table 4).
The linear model described the site-level AFR as a function of the relative proportion of N. gaeumannii Lineage 2 (PL2) recovered from the site, after accounting for distance inland (a variable that was strongly correlated with both PL2 and SNC severity; Figure 4, Table 4). Although it initially appeared that there was a significant correlation between AFR and PL2 (Figure 3d), this association was not significant when distance inland was held at its mean (25.95 km) (t 31 = −0.01, p = .99) (Figure 4). There was a significant negative association between AFR and CI, but there also was considerable variation in AFR for a given value of CI (Figure 5), though foliage retention may be affected by a combination of environmental variables in addition to colonization by N. gaeumannii.
In initial runs of NMDS that included all of the 34 sites for which disease severity data were available, site CB5-2 appeared as an outlier in the periphery of the plot and had a disproportionate influence on the ordination (see Section 4). For this reason, CB5-2 was removed from subsequent analyses. The NMDS ordination of the 33 remaining sites for which disease severity data were available (Table 2) had a final stress of 0.095 ( Figure 6). PL2 had the strongest relationship with the NMDS ordination (R 2 = .947, p = .001) ( Figure 6, Table 5). Therefore sites were aligned along NMDS axis 1 according to the relative proportion of the isolates recovered from the site that were identified as N. gaeumannii Lineage 2.
In addition to the genetic differentiation across sites related to the divergence between lineages, there also appeared to be genetic variation between sites that was correlated with environment and SNC severity ( Figure 6). Disease severity varied along NMDS axis 2 (Figure 6), and AFR was most strongly correlated with the ordination of genetic distance between the sites (R 2 = .569, p = .001) ( Figure 6, Table 5). CI was inversely related to AFR, but was not correlated as strongly with the ordination (R 2 = .228, p = .016) ( Figure 6, Table 5). Winter temperature (TmWinter) was the environmental variable most strongly correlated with the genetic differentiation between sites (R 2 = .513, p = .001), and with PL2 ( Figure 6, Table 5). Mean summer temperature (TmSummer) (R 2 = .284, p = .012) and average spring/early summer dew point temperature (MADT) (R 2 = .247, p = .015) had much weaker, but nonetheless significant, correlations with the ordination, while average precipitation in the spring and early summer was not significantly correlated with the ordination (Figure 6, Table 5). Latitude was the geographic variable with the strongest correlation with the ordination (R 2 = .342, p = .004), followed by elevation (R 2 = .283, p = .009), and longitude (R 2 = .253, p = .017) ( Table 5). AFR and TmSummer were positively correlated with elevation and longitude, while CI was positively correlated with latitude and MADT ( Figure 6, Table 5).

| D ISCUSS I ON
The population of N. gaeumannii in Oregon and Washington is diverse with a genetic structure that reflects the presence of two strongly differentiated noninterbreeding lineages (Figure 1) (Bennett & Stone, 2016;Winton et al., 2006). These characteristics are consistent with the presumed endemism of N. gaeumannii in northwestern North America (Hansen et al., 2000;Winton et al., 2006). The presence of repeated multilocus genotypes (i.e., clones) within the two lineages (Tables 1 and 2) likely reflects reproduction via homothallism, as N. gaeumannii is not known to reproduce asexually (Winton et al., 2006;Winton, 2001). This reproductive mode likely contributes to the genetic structure of N. gaeumannii populations, but the level of genetic variation observed suggests that outcrossing also occurs.
Although the two N. gaeumannii lineages have overlapping distributions, it appears that each is adapted to slightly different environmental conditions and thus exhibit different habitat distributions that are determined, directly or indirectly, by climate.
The spatial distribution of N. gaeumannii Lineage 2 corresponded to the regions where SNC symptoms were most severe suggesting that some variation in aggressiveness between the two lineages might exist, as suggested by previous authors (Winton et al., 2006).
However, we found no evidence to suggest that N. gaeumannii Lineage 2 is more aggressive than Lineage 1. After accounting for distance inland, a confounding variable in the relationship between PL2 and AFR, there was no significant correlation between Lineage 2 and SNC severity. Thus, the association between Lineage 2 and greater defoliation can be attributed to overall abundance of N. gaeumannii rather than dominance of Lineage 2 genotypes.
The separation of sites along axis 2 in the NMDS analysis ( Figure 6) reflects genetic differentiation between the isolates collected at sites with severe disease (higher CI and lower AFR) and F I G U R E 3 The relationships between distance inland (km) and (a) the relative proportion of isolates corresponding to Nothophaeocryptopus gaeumannii Lineage 2, (b) average foliage retention (AFR) (%), and (c) the colonization index (i.e., the average proportion of stomata occluded by N. gaeumannii pseudothecia). (d) The relationship between the relative proportion of Lineage 2 and average foliage retention when the distance inland (a major confounding variable) is not taken into account. Shaded regions represent 95% confidence intervals for the fitted lines those collected at sites with lower disease severity (lower CI and higher AFR). This genetic differentiation was not related to the relative abundances of Lineages 1 and 2. This suggests that some variation in aggressiveness among isolates exists that is not related to the genetic differentiation between the two lineages. These observations also suggest that adaptation to local climate or natural selection for advantageous genotypes has occurred in the geographic regions where SNC is most severe. Whether N. gaeumannii populations in the coastal SNC epidemic zone are in fact more aggressive or have increased fitness (and thus cause more severe symptoms) is still unclear and should be the focus of future studies.
The relative abundances of the lineages varied along a west-east gradient, with Lineage 2 more abundant relative to Lineage 1 in sites near the coast and decreasing in relative abundance further inland.
At sites approximately 40-56 km inland, Lineage 2 was supplanted entirely by Lineage 1 in some sites. The correspondence between the geographic distributions of N. gaeumannii Lineage 2 and severe SNC symptoms in western Oregon and Washington ( Figure 2) TA B L E 4 Pearson's correlation coefficients (r) for the relationships between SNC severity and each of the environmental and geographic variables used for this study F I G U R E 5 The relationship between average foliage retention (AFR) (%) and the colonization index (CI), an estimate of the average percentage of stomata occluded by pseudothecia of Nothophaeocryptopus gaeumannii. AFR and CI were measured for ten Douglas-fir trees from each site. The data here reflect only the 34 sites for which disease severity data were available F I G U R E 4 The relationship between average foliage retention (AFR) (%) and the relative proportion of Nothophaeocryptopus gaeumannii Lineage 2, after accounting for distance inland. Each point corresponds to one of the 34 sites for which disease severity data were available. The line corresponds to predicted values from the model when the distance inland is fixed at its mean (26 km), with the shaded region representing the 95% confidence intervals for the predicted values initially suggested a causal relationship between the relative abundance of this lineage and disease severity. At the landscape level, the SNC symptoms documented by aerial surveys were most prevalent along the western slopes of the Coast Ranges, where Lineage 2 was generally more abundant, or where the two lineages coexisted within ~30 km of the coast (Figure 2). This trend was observed in the low-elevation forests along the western slopes of the Coast Ranges in Oregon and Washington from Coos Bay, OR to the northern Olympic Peninsula (Figure 2, Table 2). Further inland, Lineage 2 was generally less abundant, or absent, and SNC symptoms were generally less severe or not detected ( Figure 2).
The observation that the regions where distributions of the two N. gaeumannii lineages overlap correspond to the regions with severe SNC is in agreement with previous studies (Winton et al., 2006).
The apparent association between PL2 and AFR when distance inland was not included in the model (Figure 3, Table 4) also supports the interpretation by Winton et al. (2006) that SNC symptoms were more severe in Douglas-fir stands with higher proportions of Lineage F I G U R E 6 Nonmetric multidimensional scaling (NMDS) ordination based on the genetic distances between 33 sample sites calculated from multilocus SSR genotypes of Nothophaeocryptopus gaeumannii isolates. Joint plot vectors show correlations between environmental variables and the ordination. Only environmental variables with statistically significant correlations (p < .05) are shown. Vector labels correspond to the variables listed in Table 4. Final stress = 0.095. Locations in legend correspond to the sampling blocks. One Coos Bay site was removed from the ordination because it was an outlier, and one S Olympic Peninsula site was not included because disease severity estimates were not available   (Manter et al., 2005(Manter et al., , 2003. Given the current understanding of the mechanisms of disease in this pathosystem, where premature foliage loss associated with SNC results from cumulative occlusion of the stomata by the pseudothecia of N. gaeumannii (Manter et al., 2000(Manter et al., , 2005(Manter et al., , 2003Stone, Capitano, et al., 2008), the most plausible mechanism by which Lineage 2 could cause more severe defoliation than Lineage 1 would be by colonizing the host needle more rapidly and subsequently producing more abundant pseudothecia relative to Lineage 1. Our analyses do not support such a mechanism.
Lineage 1 was not isolated from the Douglas-fir foliage samples collected at the two southernmost sites in Oregon near Gold Beach  (Lavender & Hermann, 2014) and N. gaeumannii is relatively common. The fact that we only recovered isolates of N. gaeumannii Lineage 2 from those sites, yet symptoms of SNC are not observed there, provides further evidence that Lineage 2 is not in fact associated with increased SNC severity.
The NMDS ordination of sites in relation to genetic distance revealed strong spatial genetic differentiation between inland and coastal sample sites ( Figure 6). This approach allowed for a visualization of relationships between genetic variation and environmental, geographic, and disease variables. The spatial distribution of genetic variation was strongly correlated with Tmwinter, Tmsummer, and MADT in the year prior to sampling, but not PPT ( Figure 6, Table 5).
Precipitation in the western Coast Ranges is generally abundant during the period of sporulation and infection by N. gauemannii, so spatial variation in spring/summer precipitation has not been a useful variable for modeling N. gaeumannii abundance in this region (Manter et al., 2005). The distribution of N. gaeumannii Lineage 2 was most strongly correlated with Tmwinter ( Figure 6, Table 5). Foliage retention varied most strongly with Tmsummer, elevation, and longitude, and was negatively correlated with MADT ( Figure 6, Table 5).
Sites with lower AFR values and higher CI values generally occurred at lower elevations nearest the coast where summers were cooler and spring/early summer dew point temperatures were warmer ( Figure 6). The N. gaeumannii populations occurring in these sites appeared to be genetically differentiated from the higher elevation inland sites where the winters were colder, summers were hotter, and AFR was greater ( Figure 6).
The environmental variable most strongly correlated with the spatial distribution of Lineage 2 (Tmwinter) is also strongly correlated with MADT, the variable most strongly associated with needle colonization by both lineages (Table 4). This suggests that the environment most conducive to needle colonization is also optimal for N. gaeumannii Lineage 2, but that Lineage 1 has a broader range of environmental tolerance than Lineage 2. These apparent differences in environmental tolerance may be related to the underlying causes of reproductive isolation and evolutionary divergence between the two lineages. These results also suggest that the distribution of genetic variability within and between N. gaeumannii populations is influenced by environmental factors, possibly due to the influences of natural selection and local adaptation. Although the SSR markers used here are presumed to be selectively neutral, and thus are not directly influenced by the environment, they provided a tool with which to detect genetic differentiation between populations that may be associated with adaptation to local climate.
Previous studies have identified significant associations between environmental variables, colonization by N. gaeumannii, and SNC symptoms (Coop & Stone, 2010;Hansen et al., 2000;Lee et al., 2013Lee et al., , 2017Manter et al., 2005;Rosso & Hansen, 2003;Shaw, Woolley, & Kanaskie, 2014;Stone, Coop, et al., 2008;Stone et al., 2007;Watt et al., 2010;Zhao et al., 2012Zhao et al., , 2011 Given that the occlusion of stomata causes the foliage loss associated with SNC, we expected the correlation between AFR and CI to be stronger, as reported in previous studies (Manter et al., 2003;Watt et al., 2010). In our dataset, there was considerable variation in AFR for a given value of CI, suggesting genetic variation in host tolerance (i.e., some host genotypes can tolerate higher levels of infection before needle abscission occurs) and/or that AFR may be affected by environmental factors in combination with SNC severity.
This variation in AFR in relation to N. gaeumannii abundance has been investigated for SNC in coastal Douglas-fir (Temel, Johnson, & Stone, 2004). In that study, the authors concluded that some genetic variation in host tolerance reflected the historical SNC pressure in the environment where the host genotype evolved (Temel et al., 2004).
Generally, Douglas-fir provenances from regions where rainfall and humidity are low are less tolerant of SNC because natural selection has not favored tolerant individuals where disease pressure is historically low (Hood, 1982;Mcdermott & Robinson, 1989;Temel et al., 2004). The variation in AFR observed in our study may also reflect some direct interaction between environment and disease; SNC is one of several factors that may affect AFR. It seems reasonable to assume that some unfavorable environmental conditions may lead to greater foliage loss at lower levels of infection and that favorable conditions may allow hosts to maintain a healthy level of foliage with higher levels of infection.
Distance inland and elevation are two geographic variables that had significant associations with both PL2 and disease severity in our data.
While these geographic variables may not directly affect biological processes that influence N. gaeumannii abundance or SNC severity, their effects on the biological system are a reflection of interacting climatic and environmental factors along a gradient from the low-elevation forests along the western slopes of the Coast Range to the higher elevation forests further inland along the Coast Range (Hansen et al., 2000;Shaw et al., 2014;Zhao et al., 2011). In fact, distance inland and elevation were collinear with several environmental variables examined in this study. We therefore used distance inland as a proxy for continentality, a complex combination of multicollinear environmental factors that vary along this geographic gradient (Zhao et al., 2011) (Lee et al., 2013(Lee et al., , 2017(Lee et al., , 2016Manter et al., 2005;Stone, Coop, et al., 2008;Stone et al., 2007;Watt et al., 2010). Our observations are also in agreement with studies that suggested that high summer temperatures inhibit the growth of N. gaeumannii (Rosso & Hansen, 2003), resulting in lower SNC severity (Lee et al., 2017;Manter et al., 2005;Stone, Coop, et al., 2008;Stone et al., 2007;Watt, Stone, Hood, & Manning, 2011;Watt et al., 2010;Zhao et al., 2012). Climate change (i.e., increasingly warmer winter temperatures) will likely exacerbate SNC severity in the western Coast Range leading to an intensification of symptoms in areas already affected by SNC, as well as an expansion of the area affected by SNC (Coop & Stone, 2010;Lee et al., 2017;Stone, Coop, et al., 2008;Watt et al., 2011Watt et al., , 2010. One site near Coos Bay, Oregon (CB5-2) did not fit with the overall trends observed for the geographic distributions of the two lineages in relation to coastal proximity. This site was within 25 km of the coast, but all of the isolates recovered from two separate trees were Lineage 1 (Table 2, Figure 2). This was unexpected, considering that the isolates sampled from a site just a few kilometers to the west were all Lineage 2, and both lineages were recovered from the closest sites to the northeast ( Figure 2, Table 2). This suggests that little gene flow is occurring between these sites, even though they are only a few kilometers apart. There were only eight isolates sampled from this site, and thus the estimated PL2 (and the estimated relationships between PL2 and other variables) may have been affected by sampling bias. The sample of eight isolates from this site had MLGs that were very dissimilar from those collected from nearby sites. Because all isolates collected from CB5-2 were Lineage 1, but Lineage 2 should have been relatively abundant given the site location, this site disproportionately influenced the fitting of the geographic variables in the NMDS, especially latitude. This suggests that the isolates at CB5-2 were more similar to sites further north, even though this site is near the southern coast of Oregon.
The results of our analyses suggest that climate influences the population structure of this ecologically important Douglas-fir pathogen. Not only does the environment influence SNC severity through direct effects on growth rate and reproduction of N. gaeumannii, but environmental variables were also correlated with spatial genetic differentiation in the N. gaeumannii population. This suggests that climate may play a role in the evolutionary divergence of these cryptic fungal lineages. Climate change has the potential to be a driver of further genetic change in N. gaeumannii populations.
The strong genetic differentiation between the two N. geaumannii lineages observed in this study is in agreement with the results of previous analyses that suggested that the lineages were reproductively isolated (Bennett & Stone, 2016;Winton et al., 2006).
Although the reproductive incompatibility of these two lineages cannot be definitively demonstrated, as this fungus does not produce ascomata in culture, the divergence observed suggests that the two lineages constitute separate ecological species. However, designation of Lineage 1 and Lineage 2 as distinct species will require phylogenetic or phylogenomic analyses to evaluate the degree of evolutionary divergence between the lineages.
The scope of this study was limited due to the use of highly variable neutral markers, which are useful for identifying popula-