Population BiologyFree Access icon

Novel Introductions and Epidemic Dynamics of the Sudden Oak Death Pathogen Phytophthora ramorum in Oregon Forests

    Affiliations
    Authors and Affiliations
    • Nicholas C. Carleson1
    • Hazel A. Daniels1
    • Paul W. Reeser1
    • Alan Kanaskie2
    • Sarah M. Navarro2
    • Jared M. LeBoldus1 3
    • Niklaus J. Grünwald4
    1. 1Department of Botany and Plant Pathology, Oregon State University, Corvallis, OR
    2. 2Oregon Department of Forestry, Salem, OR
    3. 3Forest Engineering, Resources and Management Department, Oregon State University, Corvallis, OR
    4. 4Horticultural Crops Research Laboratory, United States Department of Agriculture-Agricultural Research Service, Corvallis, OR

    Abstract

    Sudden oak death caused by Phytophthora ramorum has been actively managed in Oregon since the early 2000s. To date, this epidemic has been driven mostly by the NA1 clonal lineage of P. ramorum, but an outbreak of the EU1 lineage has recently emerged. Here, we contrast the population dynamics of the NA1 outbreak first reported in 2001 to the outbreak of the EU1 lineage first detected in 2015. We performed tests to determine whether any of the lineages were introduced more than once. Infested regions of the forest were sampled between 2013 and 2018 (n = 903), and strains were genotyped at 15 microsatellite loci. Most genotypes observed were transient, with 272 of 358 unique genotypes emerging during one year and disappearing the next year. The diversity of EU1 was very low and isolates were spatially clustered (less than 8 km apart), suggesting a single EU1 introduction. Some forest isolates are genetically similar to isolates collected from a local nursery in 2012, suggesting the introduction of EU1 from this nursery or simultaneous introduction to both the nursery and latently into the forest. In contrast, the older NA1 populations were more polymorphic and spread more than 30 km2. A principal component analysis supported two to four independent NA1 introductions. The NA1 and EU1 epidemics infest the same area but show disparate demographics because of the initial introductions of the lineages spaced 10 years apart. Comparing these epidemics provides novel insight regarding patterns of emergence of clonal pathogens in forest ecosystems.

    Sudden oak death (SOD) is a disease caused by the oomycete pathogen Phytophthora ramorum Werres, de Cock, and Man int’l Veld (Werres et al. 2001) that was first observed along the northern coast of California in the mid 1990s (Rizzo et al. 2005). Tanoak [Notholithocarpus densiflorus (Hook. & Arn.) Manos, C.H. Cannon & S. Oh], coast live oak (Quercus agrifolia Née), and, later, California black oak (Q. kellogii Newb.) showed symptoms, including bleeding bole cankers, wilting shoots, rapid foliage browning, and mortality. P. ramorum is now known to infect more than 130 plant species in 75 genera, including forest species that dominate ecotypes in California and Oregon, where the disease has reached epidemic status (Elliott and Yuzon 2018; Grünwald et al. 2012; Rizzo and Garbelotto 2003).

    In 2001, P. ramorum was discovered in Curry County, Oregon; however, its first introduction is thought to be as early as 1998 (Goheen et al. 2017). The pathogen has likely been independently introduced to forested sites in Oregon at least three times (Grünwald et al. 2016, 2019; Kamvar et al. 2015b). Since its first discovery in Oregon, an interagency team consisting of state and federal groups has worked to manage the epidemic through quarantine regulations and destruction of hosts by cutting and burning (Goheen et al. 2017; Rizzo et al. 2005). After a sharp increase in disease between 2009 and 2012, regulatory agencies established a generally infested area (GIA) in which host eradication is no longer required.

    P. ramorum is genetically split into four known clonal lineages named for the continent and the order in which they were discovered: NA1, NA2, EU1, and EU2 (Grünwald et al. 2009, 2012, 2019; Van Poucke et al. 2012). Definitions of these lineages are based on a consensus of multiple differentiating nuclear genes, mitochondrial genes, and microsatellite loci (Grünwald et al. 2009). The initial founder populations of Oregon’s forest infestation are of the NA1 lineage, and the EU1 lineage was discovered within the Curry County quarantine in 2015 (Grünwald et al. 2016, 2019; Kamvar et al. 2015b; LeBoldus et al. 2018). The origin of this recent EU1 introduction is unknown. Both NA1 and EU1 have been found in nurseries in California, Oregon, and Washington in the United States and in British Columbia in Canada. NA2 has been found in these provinces as well, but it has not yet been detected in Oregon forests (Frankel 2008; Goss et al. 2011; Grünwald et al. 2008a; Grünwald et al. 2019); however, it was recently discovered in the Midwest (Press et al. 2020). To date, EU2 has not been observed outside of Europe (Van Poucke et al. 2012).

    P. ramorum primarily reproduces clonally in all epidemics documented to date. Sexual reproduction is theoretically possible but has not been observed under field conditions. Because P. ramorum is heterothallic, its two mating types (A1 and A2) must interact to form oospores (Werres et al. 2001). Thick-walled Phytophthora oospores can persist in the environment longer than other propagules, posing a high risk for latent infection (Erwin and Ribeiro 1996). All NA1 isolates to date are of the A2 mating type, whereas most EU1 are of the A1 mating type (Grünwald et al. 2008b, 2016; Hansen et al. 2003). Co-occurrence of opposite mating types in the same region could lead to sexual reproduction and novel genotypes through recombination (Brasier and Kirk 2004). However, it has been shown that P. ramorum oospores are aberrant under laboratory settings with poor germination and unexpected segregation of alleles (Boutet et al. 2010). Furthermore, it is unclear whether recombinant P. ramorum genotypes are more or less pathogenic or aggressive than either parental strain (Boutet et al. 2010; Vercauteren et al. 2011).

    Understanding the spatiotemporal and genetic distribution of the NA1 and EU1 P. ramorum lineages is important to inform effective management practices. Identification of high-risk areas, such as those populated by abundant genotypes, genotypes prone to long-distance dispersal, or regions with co-occurring opposite mating types, may allow for more effective deployment of the limited resources for management. In the long-term, high-risk areas could be targeted for ground surveys and host sampling to characterize host susceptibility and patterns of pathogenicity, informing epidemic models that more accurately predict disease transmission (Cobb et al. 2019).

    Population genetic analyses also aid in the evaluation of previous management strategies. If eradication efforts are effective, then the same genotype(s) would not be observed consistently across multiple years. In the forests we sampled, all diseased trees and surrounding areas were destroyed quickly after the diagnosis of the EU1 infection, and trees infected with NA1 were destroyed only if found outside of the GIA (Goheen et al. 2017). Therefore, the genotypes that dominate across several consecutive years are likely to be derived from sporulation prior to eradication or secondary inoculum sources that were not sufficiently removed from the environment, such as twigs and branches (Maloney et al. 2005). This is complicated by heterogeneous eradication treatments, which have changed over time, and are dependent on specific risk factors at each site and available program funding. Generally, after positive identification of the pathogen, infected trees at the site are killed in addition to healthy but susceptible trees in a surrounding buffer of 100 to 200 m. Among infected and buffer trees killed, tanoaks were always cut, and other potential SOD hosts, such as Oregon myrtle and Rhododendron shrubs, were sometimes destroyed.

    Our overall objective was to determine how the SOD epidemic in the coastal forests of Curry County in southwest Oregon has changed from that described in previous reports that summarized data gathered from 2001 to 2012 (Kamvar et al. 2015b). Similar to previous work, we used microsatellite markers to identify the spatiotemporal dynamics (Kamvar et al. 2015b; Prospero et al. 2007). The first objective was to determine how many times the EU1 lineage had been introduced since it was first recognized. Our second objective was to determine if any additional NA1 introductions occurred independent of the two reported previously. Third, we compared population dynamics of the NA1 and EU1 epidemics and searched for evidence of sexual recombination. Finally, we related genetic diversity and geographic data using the emergence and disappearance of dominant and rare genotypes to gain insight regarding the flow of P. ramorum in a coastal forest ecosystem.

    MATERIALS AND METHODS

    Location and sampling strategy.

    Aerial surveys in southwest Oregon are performed quarterly each year to identify potential SOD-diseased trees. Surveys center on the SOD quarantine and surrounding area and focus on tanoak mortality. Tanoak, the most commonly infected host tree in the region, inhabits riparian and upland zones with common understory species, including salal (Gaultheria shallon Pursh), Pacific rhododendron (Rhododendron macrophyllum D. Don ex G. Don), and Oregon grape (Mahonia aquifolium (Pursh) Nutt.). The sampled region is a mixed evergreen forest with upper canopies mostly dominated by Douglas fir (Pseudotsuga menziesii (Mirb.) Franco). Human disturbance has contributed to a mosaic of vegetation types in the landscape containing variable components of tanoak and other vegetation types, including redwood (Sequoia sempervirens (Lamb. ex D. Don) Endl.), Douglas fir, and other conifers (Franklin and Dyrness 1973; Hansen et al. 2019). Several of these species are hosts to P. ramorum, but all samples analyzed in this study were isolated from tanoak hosts. Ground surveys located and sampled trees identified by the aerial survey, as well as symptomatic and dead tanoak trees not visible by aerial observation. Samples consisted of canker margin tissue from the inner bark of tree boles and, occasionally, foliage or twig samples (Goheen et al. 2017; Parke et al. 2007). Geographic coordinates of sampled trees were collected with handheld GPS devices.

    Samples were placed on Phytophthora-selective media (corn meal agar amended with 10 ppm natamycin, 200 ppm Na–ampicillin, and 10 ppm rifamycin SV), from which individual P. ramorum isolates were collected and isolated according to the methods of Hansen et al. (2019). Bark samples were direct-plated in the field to minimize sample desiccation and expedite sample processing, enabling laboratory results to be obtained sooner for eradication prioritization. Twig, foliage, and a second bark sample were transported to the laboratory. Coordinates of each sampled tree were collected and imported into ArcGIS 10.6 software (Esri, Redlands, CA) for the generation of geospatial distribution maps. Additional base layers were provided by Esri and the U.S. Geological Survey (USGS) National Hydrography Dataset.

    Drainage areas were delineated for the purposes of this study by using hydrologic unit code (HUC) 10 or 12 delineations when appropriate (U.S. Geological Survey and National Resources Conservation Service 2013). We refer to HUC 12 delineations (Myers Creek, Thomas Creek, and Tuttle Creek) utilized along the Pacific coast as subwatersheds, and the inland HUC 10 delineations (Windchuck River, Chetco River, Hunter Creek, and Pistol River) as watersheds. To describe results with mixed HUC 12 and 10 delineations, we refer to the drainage areas generically as hydrologic regions. Although P. ramorum is isolated from rivers in the area, only data from isolates of infected tanoak tissue are presented.

    Genetic analysis.

    An assay using the cellulose-binding elicitor lectin (CBEL) locus was performed to rapidly identify the lineage of each isolate (Gagnon et al. 2014). Genotyping was subsequently conducted using cultures of P. ramorum at 15 simple sequence repeat (SSR) microsatellite loci by the Center for Genome Research and Biocomputing (CGRB) at Oregon State University on a 3730 capillary sequencer (Thermo Fisher Scientific, Waltham, MA). The loci were selected based on previous work showing polymorphisms within and between the two clonal lineages currently described in Curry County (Goss et al. 2009b; Goss et al. 2011; Ivors et al. 2006; Kamvar et al. 2015b; Prospero et al. 2007; Vercauteren et al. 2010). A panel of reference strains and the corresponding reference database for previously observed genotypes of P. ramorum were also included (Grünwald et al. 2010). Genetic data were analyzed for isolates collected from 2013 to 2018. During the analysis, we included genotypes previously found in Rhododendron sampled in a nursery near the first tanoak that tested positive for infection by an EU1 strain (Grünwald et al. 2016).

    SSR data were imported into R 3.5.1 considering each codominant locus as diploid in all isolates and analyzed using various R packages (R Core Team 2018). Diversity statistics and minimum spanning networks were calculated using adegenet 2.1.0 and poppr 2.8.1 (Jombart 2008; Kamvar et al. 2014). For evidence of sexual recombination, the standardized index of association r¯d was calculated in poppr, including permutations to test the null hypothesis of linkage equilibrium that would be maintained by random mating and recombination (Agapow and Burt 2001; Brown et al. 1980; Smith et al. 1993). Poppr was also used to detect if the observed diversity in NA1 or EU1 lineages would appreciably increase with additional markers (Arnaud‐Haond et al. 2007; Kamvar et al. 2015a). Alleles at microsatellite loci were assumed to follow a stepwise mutation when calculating Bruvo’s genetic distance. Missing alleles, observed at low frequencies (Supplementary Table S1), were replaced according to the genome addition model in which missing alleles were replaced with all possible combinations of all observed alleles to calculate distance (Bruvo et al. 2004). Previously published repeat lengths were used to calculate Bruvo’s distance (Supplementary Table S2).

    A principal component analysis (PCA) was performed using ade4 1.7-13 and ape 5.4 (Cori et al. 2018; Dray and Dufour 2007; Jombart 2008; Kamvar et al. 2014; Paradis et al. 2004). A discriminant analysis of principal components (DAPC) was conducted using the adegenet package (Jombart et al. 2010) with a K-means clustering approach. To maximize the power of DAPC clustering while avoiding overfitting the model, we performed a discriminant analysis of the first 35 principal components; the total variance captured was saturated. Cohesion of clusters was evaluated with the Bayesian information criterion (BIC) and silhouette scores (SS), and the number of K groups that resulted in the smallest BIC and maximum SS was identified. SS were calculated using both Bruvo’s distances and Euclidean distances between the first eight principal components for each isolate (Rousseeuw 1987; Schwarz 1978). The minimum BIC is often unclear and cannot be used for selecting the optimal K clusters in populations (see helpfile provided with adegenet); therefore, we also calculated the difference in BIC between successive K-means groupings. Then, we maximized both this ΔBIC and the SS together. Visualizations were created using ggplot2 (Wickham 2016). All scripts and custom functions are available at GitHub (https://github.com/grunwaldlab/Ph_ramorum_ORforests).

    RESULTS

    Low diversity within the EU1 lineage.

    The first EU1 site was found in Curry County, Oregon, in February 2015. The host tanoak was cut and burned, but the stump was resampled 3 months later to confirm the EU1 identification. The first isolate shares complete genotype identity with three of the isolates obtained from the resampling, with all isolates clustering closely within a network (Fig. 1). None of the MLGs originally isolated was observed in 2018. Three isolates sampled in 2012 from nurseries near the infection site (Grünwald et al. 2016) cluster within a network, and the nursery genotypes were observed in the forest in 2016, within the Myers Creek subwatershed (Fig. 1). Most isolates in 2015 to 2018 were from the Myers Creek subwatershed, with low abundance in the Pistol River watershed (Table 1).

    Fig. 1.

    Fig. 1. Minimum spanning network of Phytophthora ramorum samples of the EU1 clonal lineage collected from tanoaks in the forests in Curry County, Oregon, between 2015 and 2018. Each circle represents one multilocus genotype; it is colored according to the proportion of isolates that were collected during a certain year. Collections in 2012 were isolated from Rhododendron hosts in a nearby nursery.

    Download as PowerPoint

    TABLE 1. Phytophthora ramorum isolates and multilocus genotypes (MLGs) collected during 2013 to 2018 from Curry County, OR, and divided among hydrologic regions

    Reflecting their limited geographic range, the genetic distance within a clonal lineage is expected to be small for a recently introduced, novel lineage. EU1 isolates were highly related across 30 unique multilocus genotypes (MLGs). Across the 15 microsatellite loci sequenced, only five were polymorphic within the EU1 lineage. The genotype accumulation curve showed modest power to resolve genotypes, which could be explained by the recent introduction of EU1 (Supplementary Fig. S1). Isolates were spread randomly in the PCA space (Supplementary Fig. S2). Axis one and axis two captured only 34.5 and 24.9% of the total variation, respectively, reflecting the low diversity given the recent introduction. Furthermore, no reasonable value of K selected in K-means clustering resulted in group membership was accurately reflected in the PCA. The low genetic diversity and lack of clustering indicated that there was likely a single introduction of EU1 into the area, and that this clonal population is gradually starting to accumulate mutations.

    To determine the role of the nursery industry in the EU1 invasion, we evaluated whether the nursery isolates of EU1 shared alleles with all MLGs found in the forest. All alleles found in the three nursery isolates were also found in the forest populations sampled. In the forest populations, no alleles private to the Pistol River watershed or private to isolates collected in 2015 were detected. There was a single allele at the locus KIPrMS18 and two alleles at PrMS145c exclusively observed during any single year or any single hydrologic region. There were two (by year) or four (by hydrologic region) such alleles at locus ILVOPrMS131.

    Independent clustering of EU1 and NA1.

    EU1 was introduced independently of the existing NA1 epidemic, but the two lineages are found in close proximity. In general, EU1 and NA1 did not share the same geographic range until recently (Fig. 2). NA1 was present in the Myers Creek subwatershed, where EU1 was primarily confined; however, there was only a small area in which both lineages were found in close proximity (≈0.1 km2) (Table 1; Fig. 2).

    Fig. 2.

    Fig. 2. Geospatial distribution of multilocus genotypes (MLGs) of Phytophthora ramorum isolates collected between 2013 and 2018, in order of greatest abundance by lineage. (a), The area of interest (yellow) is located between the cities of Gold Beach and Brookings within Curry County, the southwestern-most county in Oregon. (b), A histogram shows the distribution and count of MLGs identified since 2012 and grouped by lineage (EU1 = purple; NA1 = orange) for those MLGs with more than one occurrence (116 MLGs). (c) to (q), Each panel shows one combination of clonal MLGs by lineage rank-ordered by dominance across panels. ChRv = Chetco River watershed; HuCr = Hunter Creek watershed; MyCr = Myers Creek subwatershed; PiRv = Pistol River watershed; ThCr = Thomas Creek subwatershed; TuCr = Tuttle Creek subwatershed; and WiRv = Winchuck River watershed.

    Download as PowerPoint

    Genotyping can potentially improve the accuracy of dispersal distance estimates, which are typically calculated by simply measuring the longest distance between a newly infested site and its nearest neighboring infested site. This method assumes that the further site is not an independent introduction; however, when that assumption is true, it could still underestimate the true distance if the inoculum came from a site further away. The longest dispersal distance of EU1, defined as observation of the same MLG in consecutive years but separated in space, was less than 6 km. NA1 was dispersed into six hydrologic regions, with the most dominant MLGs occurring in up to four different regions and dispersed as far as 35 km apart between 2017 and 2018.

    Genetic distance within a clonal lineage is expected to be higher for older and more widely dispersed lineages because mutations accumulate in a lineage presumably at a constant rate. Similarly, genetic distances between two clonal lineages are expected to be larger than those within a clonal lineage. All genetic distances calculated pairwise between MLGs within each lineage were much lower than pairwise comparisons performed between MLGs from opposite lineages (Fig. 3). Of 388 unique P. ramorum MLGs, none was shared between isolates from NA1 and EU1 lineages (Fig. 4). The first axis in a PCA separated the two lineages, capturing 64% of variation (Supplementary Fig. S3). NA1 isolates (n = 700) were placed in 358 unique MLGs and EU1 isolates (n = 203) collapsed into 30 MLGs (Table 1). Most MLGs were transient; 272 MLGs were observed a single time between 2013 and 2018. Only the most dominant MLGs were observed consistently and in high abundance across multiple years (Fig. 4).

    Fig. 3.

    Fig. 3. Histogram of pairwise Bruvo’s genetic distances between all Phytophthora ramorum multi-locus genotypes within EU1 (yellow; n = 30) and NA1 (purple; n = 357) and across EU1 and NA1 lineages (red) clonal lineages.

    Download as PowerPoint
    Fig. 4.

    Fig. 4. Rank-size distribution of Phytophthora ramorum multilocus genotypes (MLGs) found between 2013 and 2018 in Curry County, Oregon. All EU1 lineage MLGs (orange triangles; n = 30) and the most abundant NA1 lineage MLGs (purple circles; n = 15) are sorted by abundance, with the highest abundance at the top of the figure. The number of isolates that are described by each MLG during each year are presented inside shapes. Lines connecting shapes represent the discovery of that same MLG during any subsequent year.

    Download as PowerPoint

    Because EU1 and NA1 are opposite mating types, sexual recombination is theoretically possible within their shared geographic range in Curry County. We found no genetic evidence of sexual recombination. We rejected the null hypothesis of random mating (P < 0.005) at all levels of population structure tested using data before clone correction, including comparisons within lineages, hydrologic regions, years, lineages × hydrologic regions, and overall. After clone correction, the index of association was significant for EU1 at all levels tested. For NA1 at the level of lineages × hydrologic regions, we could not reject the null hypothesis in the Thomas Creek subwatershed after clone-correcting the data (n = 47; r¯d = 0.01; P = 0.15) (Supplementary Table S3).

    NA1 diversity.

    The epidemic of the NA1 clonal lineage was spatially and genetically diverse. Of 15 SSRs sequenced, 14 were polymorphic across both lineages and 13 were polymorphic in NA1 (Supplementary Figs. S4 and S5). The genotype accumulation curve for NA1 was saturated, indicating the isolates captured most of the diversity in the area. The most genetic diversity was in the Chetco River watershed. The second most diverse population was in the neighboring Winchuck River watershed. In most watersheds, the region-specific MLGs were mostly singletons; however, in these two watersheds, at least two isolates shared most MLGs (Table 1).

    We used an assumption-free cluster analysis to analyze population dynamics, including the number of independent NA1 and EU1 introductions in the region. All isolates from each lineage clustered independently in the PCA, although EU1 isolates were more tightly clustered than NA1 (Supplementary Figure S3). Although there was no clearly minimal BIC in the cluster analysis, there was a large decline in the improvement of fit after K=3 groups, suggesting three genetic clusters of NA1 isolates (Supplementary Figure S6). Cluster 2 (cyan ellipse), which contained many isolates from all hydrologic regions, and cluster 3 (yellow), which was found mostly in the Chetco River watershed, were more closely related to each other than cluster 1. Cluster 1 (purple), which was more distant from the other two clusters, was mostly isolated to the Winchuck River watershed (Fig. 5). The median genetic distance for all within-cluster pairwise comparisons was lower than all distances calculated between MLGs from different clusters, thus supporting PCA-based K-means clustering (Supplementary Figure S7). Despite uneven sampling, at least one isolate from each hydrologic region did not cluster within groups containing 95% of the isolates from a given cluster, indicating cryptic diversity.

    Fig. 5.

    Fig. 5. Distribution of Phytophthora ramorum isolates collected between 2013 and 2018 belonging to the NA1 (n = 701) clonal lineages sampled in Curry County, Oregon. A, A principal component analysis of the first two axes captured 37.5% of total variation. Each point represents a single isolate. The shape of each point (square, circle, triangle) represents which cluster an isolate was assigned to during the discriminant analysis of principal components (DAPC), with likelihoods of all isolates plotted B, by year and C, by watershed.

    Download as PowerPoint

    Cryptic diversity in the NA1 population confounded clustering approaches. SS, which compare distances within and between clusters, were highest for either the K=2 group or the K=4 group (Supplementary Figure S8). Maximizing both the ΔBIC and SS, the best number of groups was K=3 (Supplementary Figure S9). Despite discordance between results, metrics of the clustering performance we measured suggested either K=3 or K=4 clusters. Importantly, the differing numbers of clusters revealed no major differences in population structure. With K=4 groups, the diverse isolates between cluster 1 and cluster 2 and cluster 3 were grouped into a fourth cluster (Supplementary Figure S10). Therefore, we observed evidence of at least three separate NA1 introductions.

    Because oomycetes travel in wind-splashed water, we expect that the spread of P. ramorum is more likely to occur within rather than across hydrologic regions. In the Pistol River, Thomas Creek, and Myers Creek hydrologic regions, cluster 2 was dominant. During the 5-year study period, the Myers Creek subwatershed consisted of almost entirely isolates belonging to cluster 2, with a single isolate collected in 2016 that was assigned to cluster 1 but with an assignment probability <98%, unlike the other isolates, which had nearly 100% posterior membership probability of being assigned to their cluster (Fig. 5B and C). The Chetco River watershed appeared to have the highest membership to cluster 2, mostly because of its dominance in 2013. In subsequent years, isolates assigned to cluster 2 and cluster 3 were collected in similar proportions. A low but stable presence of cluster 1 was observed in Chetco River within the study period. Winchuck River was dominated by cluster 1, which grew substantially in 2015 and again in 2018 (Fig. 5C). Although cluster 1 was dominant in 2015 and 2018, these were the two least diverse years during the study period (Supplementary Table S4).

    DISCUSSION

    Our first objective was to assess how many introductions of EU1 occurred and whether they are linked to nursery epidemics. Initial reports implicitly linked local nursery activity in Curry County to the outbreak of the P. ramorum clonal lineage EU1 in the nearby forest (Grünwald et al. 2016). Even though nurseries have been implicated in introducing founder EU1 populations, no genotypes of isolates collected in 2015, which were all recovered from the same site initially infected, shared complete identity with a genotype from a nursery. However, isolates collected 1 year after the initial report did match the exact genotype of nursery isolates. This could be the result of landscape plants that originated in the infested nursery and subsequently sporulated, or the result of latent infection from spores that were still in the environment or released by the nursery or other nearby unsampled nurseries. More likely, our results could be an effect of sample bias. Unsampled standing variation in EU1 nursery populations not captured by our three analyzed isolates could explain the apparent 4-year delay between these infections (Grünwald et al. 2016). There were no alleles private to the nursery population at any locus. Highly variable loci that separated EU1 populations in the forest from the nursery, primarily locus PrMS145a, converged over the study period. A comprehensive sampling of multiple nurseries in 2012 likely would have found exact matches with the founder EU1 population in the forest. Our results are consistent with the northward spread of the nursery infestation in Pistol River (Hansen et al. 2019) and support a single introduction of the EU1 clonal lineage.

    Another objective of the study was to determine how many independent NA1 introductions have occurred. There was prior support for at least two independent NA1 introductions, as reported by Kamvar et al. (2015b), with the second introduction occurring in the Hunter Creek watershed. Since then, the regional HUC boundaries were redefined, and these isolates were assigned to the Myers Creek subwatershed in the present study. Because this region contains isolates almost exclusively assigned to cluster 2, we suggest that this cluster contains the second outbreak. Despite treatment of the infested site (Goheen et al. 2017), P. ramorum isolates assigned to this cluster were found in high abundance in all hydrologic regions. Alternatively, it is possible that cluster 2 was a separate introduction from the Hunter Creek event, and that our study did not capture the previous introduction in the area after successful eradication. However, it is more parsimonious that infections linked to the independent Hunter Creek outbreak persisted and spread through Curry County.

    Without improved resolution in the data, it is challenging to determine how many NA1 introductions occurred in Curry County. The clusters we report may be independent introductions, or they could be the result of clonal divergence from a single or unknown number of earlier introductions. In EU1, genetic drift appears to affect the population evenly, introducing random variation with no apparent subpopulations evolving. It is possible that in NA1, subpopulations in the process of evolving are not clearly captured either by the marker system used or by the assumption that hydrologic regions serve as borders to P. ramorum migration. For example, NA1 cluster 2 and cluster 3 could be from one introduction, and we are now observing divergence into two subpopulations. However, eradication could have selected for certain genotypes at the beginning of the epidemic. Future work should more thoroughly investigate the flow of P. ramorum between hydrologic regions defined at various levels by the U.S. Geological Surveys to supplement the investigations of inoculum dispersal (Eyre and Garbelotto 2015). Furthermore, using whole-genome sequence data will resolve the apparent homoplasy observed in the microsatellite data.

    Given that the two clonal lineages present are of opposite mating type, the final objective was to determine the presence of sexual reproduction. Both mating types must be present for P. ramorum to produce oospores that may result in the emergence of increased virulence through sexual recombination (Erwin and Ribeiro 1996; Li et al. 2012; Werres et al. 2001). We report a small area in which these two lineages of opposite mating types (EU1: A1; NA1: A2) are co-occurring. Recombination is most likely to occur in the Myers Creek subwatershed that contains most of the EU1 genetic diversity and multiple independent introductions of NA1 and EU1 clones. Although the overlapping area between NA1 and EU1 outbreaks in the Myers Creek subwatershed presents a risk of sexual recombination, we found no genetic evidence of recombination between 2013 and 2018. Furthermore, we have not observed oospores in soil or tissue samples collected from the forest in any region. Because both mating types are within dispersal range of each other, and because the area infected by NA1 is expanding north toward the EU1 outbreak while EU1 is expanding south toward NA1 infestations, recombination is an ongoing risk and should continue to be monitored.

    The viability and likelihood of recombination and outcrossing are dependent on the organism and biotic and abiotic environmental factors. In a powdery mildew fungus, it was recently shown that co-infection of the same host increased the probability of overwintering and emergence of novel genotypes after outcrossing (Laine et al. 2019). The powdery mildew strains most likely to show evidence of recombination were isolated from co-infected hosts, which is a concept that has not been explored in P. ramorum. Inside or outside of co-infected hosts, the possibility of outcrossing should be the subject of future studies to more deeply investigate the importance of recombination, as has been observed in Phytophthora infestans at its center of origin in Mexico (Fernández-Pavía et al. 2002; Goss et al. 2014; Wang et al. 2017). However, in vitro oospore production of P. ramorum is challenging. In crosses between P. ramorum strains from lineages of opposite mating types or with other Phytophthora spp. strains of known mating type, oospore production rates are generally low in culture, their progeny are often not viable, and there is little evidence that recombinants are more pathogenic than their parental strains (Boutet et al. 2010; Vercauteren et al. 2011). The lack of genetic or physical evidence for outcrossing, despite the close proximity of mating types and unsampled co-infection, supports the hypothesis that P. ramorum may have effectively lost the ability to sexually reproduce after ancient isolation of clonal lineages (Goss et al. 2009a). As the P. ramorum metapopulation in Curry County expands and the number of co-infected hosts increases, local environmental changes could reveal the reproductive ability of the pathogen.

    It is expected that older epidemics will show greater genetic diversity than newer epidemics because older populations have had more time to evolve. Clonal lineages in longer-lasting epidemics have had more time to accumulate mutations over larger populations and experience evolutionary forces acting on populations with entirely linked genomes. By all measures of genetic diversity, the NA1 population is more diverse than the EU1 population. However, these results are subject to an ascertainment bias based on the genotyping method used. Most loci sequenced were developed to detect polymorphisms within the NA1 lineage. The nine additional polymorphic loci in NA1 resulted in substantially more power and accuracy understanding the population genetics and dynamics of NA1 rather than EU1. Despite this, if only five polymorphic loci were used with the NA1 lineage, then there would still likely be more than double the number of MLGs, as observed in the EU1 lineage. Additionally, the observed genetic diversity exceeds what was previously reported during the epidemic. Even assuming only five polymorphic loci, as in the study by Kamvar et al. (2015b), we likely would have observed more MLGs now than were observed in data leading up until 2013.

    The present study found that a large proportion of all unique genotypes was recorded only one or two times across the entire landscape. Future work could compare NA1 populations outside and within the GIA of the quarantine zone, encompassed by the Thomas Creek subwatershed, and the Chetco River and Pistol River watersheds. The purpose of establishing the GIA was to reduce management costs within it and allocate funds toward eradicating infestations near the leading edges of the outbreak that would increase the total infested area (Kanaskie et al. 2013). As a result, NA1 populations within the GIA have expanded for several years, unimpeded by eradication efforts. Typical species accumulation curves indicate that the rarest species occur only a few times (Fisher et al. 1943; Magurran and Henderson 2003; Preston 1948). This ecological property also applies to the accumulation of genotypes within a species, where we expect the rarest genotypes to be observed only a few times (Grünwald et al. 2003). If eradication efforts are successful, then rare genotypes would be even less frequently observed because of genetic bottlenecking of the population. However, in the context of the Curry County epidemic, only the populations outside the GIA for NA1 would be experiencing a bottleneck. Although the distribution of abundance per MLG may prove effective for evaluating management practices, we would need to sample both within and outside the GIA to understand how the resource tradeoff of GIA establishment affects eradication efforts. Because all hosts infected by EU1 are being destroyed, we would expect to see a bottleneck effect as we collect more data, which would indicate that eradication efforts are mitigating the impact, even if we are unable to completely eradicate this lineage from the area.

    Data collection and subsequent analysis are limited by sampling bias. Because the EU1 lineage was discovered in the area more recently than NA1, management priority has been given to data collection and host eradication concentrated in these condensed areas. There is also a recognized delay between initial infection and symptom generation, which impacts sampling and temporal analysis of both lineages (Goheen et al. 2017). Future research to improve rapid detection of SOD symptoms and generate lineage-neutral markers will reduce the impact of these known limitations. Therefore, we are exploring whole-genome single-nucleotide polymorphisms as a better marker system.

    ACKNOWLEDGMENTS

    We thank Wendy Sutton, Meg Larsen, Karan Fairchild, and Caroline Press for maintenance of the cultures and excellent technical support; the CGRB at Oregon State University for access to their high-performance computing facility and sequencing; Brian Knaus for bioinformatic support; and Casara Nichols, Randy Weise, Jon Laine, and Vimal Golding for their tireless efforts to find SOD-positive trees.

    The author(s) declare no conflict of interest.

    LITERATURE CITED

    N. C. Carleson and H. A. Daniels contributed equally to this work and are considered co-first authors.

    The author(s) declare no conflict of interest.

    Funding: This project was supported in part by funds from the U.S. Department of Agriculture-Agricultural Research Service (CRIS Projects 2072-22000-041-00-D and 2072-22000-043-00-D), the USDA-ARS Floriculture Nursery Initiative (to N. J. Grünwald), and the Phytophthora Diagnostics grant from the U.S. Forest Service (to J. M. LeBoldus).

    Current address for S. M. Navarro: United States Forest Service, Portland, OR.