The ascomycete Geosmithia morbida and the walnut twig beetle Pityophthorus juglandis are associated with thousand cankers disease of Juglans (walnut) and Pterocarya (wingnut). The disease was first reported in the western United States (USA) on several Juglans species, but has been found more recently in the eastern USA in the native range of the highly susceptible Juglans nigra. We performed a comprehensive population genetic study of 209 G. morbida isolates collected from Juglans and Pterocarya from 17 geographic regions distributed across 12 U.S. states. The study was based on sequence typing of 27 single nucleotide polymorphisms from three genomic regions and genotyping with ten microsatellite primer pairs. Using multilocus sequence-typing data, 197 G. morbida isolates were placed into one of 57 haplotypes. In some instances, multiple haplotypes were recovered from isolates collected on the same tree. Twenty-four of the haplotypes (42%) were recovered from more than one isolate; the two most frequently occurring haplotypes (H02 and H03) represented 36% of all isolates. These two haplotypes were abundant in California, but were not recovered from Arizona or New Mexico. G. morbida population structure was best explained by four genetically distinct groups that clustered into three geographic regions. Most of the haplotypes isolated from the native range of J. major (Arizona and New Mexico) were found in those states only or present in distinct genetic clusters. There was no evidence of sexual reproduction or genetic recombination in any population. The scattered distribution of the genetic clusters indicated that G. morbida was likely disseminated to different regions at several times and from several sources. The large number of haplotypes observed and the genetic complexity of G. morbida indicate that it evolved in association with at least one Juglans spp. and the walnut twig beetle long before the first reports of the disease.
Citation: Zerillo MM, Ibarra Caballero J, Woeste K, Graves AD, Hartel C, Pscheidt JW, et al. (2014) Population Structure of Geosmithia morbida, the Causal Agent of Thousand Cankers Disease of Walnut Trees in the United States. PLoS ONE 9(11): e112847. https://doi.org/10.1371/journal.pone.0112847
Editor: Ned B. Klopfenstein, USDA Forest Service - RMRS, United States of America
Received: January 19, 2014; Accepted: October 16, 2014; Published: November 13, 2014
This is an open-access article, free of all copyright, and may be freely reproduced, distributed, transmitted, modified, built upon, or otherwise used by anyone for any lawful purpose. The work is made available under the Creative Commons CC0 public domain dedication.
Funding: This research was supported by funds provided by the USDA NIFA Western Region IPM Center and Critical Issues-Emerging and New Plant and Animal Pests and Diseases grant programs, the USDA Forest Service Forest Health Monitoring Program (Detection Monitoring Grant No. INT-DM-09-01 and Evaluation Monitoring Grant No. INT-EM-B-11-03), and USDA Forest Service Special Technology Development Grant R4-2011-01 (administered by A.S. Munson). The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript.
Competing interests: The authors have declared that no competing interests exist.
Juglans nigra L. (Juglandaceae), commonly referred to as black walnut or eastern black walnut, is a native tree species of eastern North America (Figure 1A). Its wood is highly prized for use in cabinetry, gunstocks, veneer, and other finished wood products, and the nuts are an important nutritional source for wildlife –. This species was widely planted in the western United States (USA) as an ornamental and nut-bearing tree ,  during European colonization and the subsequent development of rural and urban landscapes. However, black walnut does not constitute a major proportion of trees in the modern urban landscape in this region.
Regions/counties were color-coded according to the tree species with green = J. major; red = J. californica; yellow = J. regia; blue = J. hindsii; black = J. nigra; pink = J. cinerea; light gray = unidentified Juglans spp. or hybrids. (Figure adapted from U.S. Census Bureau and U.S. Geological Survey at https://www.census.gov/).
In the early 1990 s, widespread decline and death of J. nigra from an undetermined cause was observed in Oregon (OR), USA , . Similar mortality in Utah (UT) in the late 1990 s  and in New Mexico (NM) in 2001  was attributed to drought and attack by the walnut twig beetle (WTB), Pityophthorus juglandis Blackman (Coleoptera: Scolytidae) , . Beginning in 2001, decline and mortality of trees was also noted in several cities in Colorado (CO) . Symptoms included yellowing of foliage and thinning of the upper crown, followed by twig and branch dieback and eventually tree death . In 2008 an undescribed species of the ascomycete Geosmithia associated with the WTB was shown to cause bark cankers around WTB galleries in dying J. nigra in CO , and the fungus was subsequently described as Geosmithia morbida M. Kolařík, E. Freeland, C. Utley, & N. Tisserat sp. nov. (Ascomycota: Hypocreales) . Many other Geosmithia species are symbionts of bark beetles, but G. morbida was the first phytopathogenic species reported in this genus . The disease in J. nigra was given the common name thousand cankers disease (TCD) because of the enormous number of coalescing cankers that are formed around WTB entrance holes and galleries when the beetle aggregates in the inner bark of severely affected trees . The WTB and TCD have now been documented as causing J. nigra mortality throughout much of the western USA and more recently, in many locations within the native range of this species , –. In 2013, TCD was also confirmed in J. nigra planted in northeastern Italy .
The source of the current TCD epidemic remains unclear. There were no recorded collections of the WTB or G. morbida from J. nigra in the western USA prior to 1959 in southern California (CA); none prior to 1996 in UT ; and none in the native range of J. nigra until 2010 , . WTB was first collected in 1896 from J. major (Torr.) A. Heller (Arizona walnut) in NM , , and it has since been collected widely throughout the northern native range of this species in Arizona (AZ), NM, and once in Chihuahua in Mexico (Figure 1A) , . Geosmithia morbida has also been isolated consistently from small, superficial cankers surrounding WTB galleries in native stands of J. major in AZ and NM, but the fungus does not cause extensive branch dieback or mortality in this species , . This has led to speculation that the origin of the WTB and G. morbida is from J. major, and that there was a host shift in the recent past by the beetle and its fungal symbiont to the more susceptible J. nigra. Another possible source of the WTB and G. morbida in the current epidemic are populations from Los Angeles County where WTB was collected in 1959 from J. nigra and from J. californica S. Watson (southern California black walnut) , the latter is a native tree in this region . Geosmithia morbida has been recovered subsequently from declining and also relatively healthy appearing J. californica throughout its range in southern CA (Figure 1A)  and from many other species of Juglans and the closely related Pterocarya growing in arboreta and germplasm collections in CA (SJS, unpublished data).
The population structure of G. morbida in the USA has been partially characterized. Previous studies have indicated that the fungus is genetically complex , . Freeland and collaborators  identified 12 haplotypes from 145 G. morbida isolates collected in the western USA based on rDNA ITS sequences, and 10 haplotypes from 56 isolates based on partial β-tubulin (BT) sequence comparisons. There were no correlations among haplotypes and the hosts or geographic regions from which the isolates were collected . Representative ITS haplotypes recovered from different geographic regions and hosts were all pathogenic, and although there were slight differences in aggressiveness among isolates, canker sizes were not significantly different . Freeland and collaborators  also showed that of four G. morbida isolates recovered from different cankers on the same tree, all had different di-locus haplotypes based on rDNA ITS and BT sequences. Hadziabdic and collaborators  used 15 polymorphic microsatellite loci to reveal high haploid genetic diversity in G. morbida isolates collected from the eastern USA and OR. They identified two genetic clusters that corresponded to isolates collected in: 1) OR and North Carolina (NC), and 2) Pennsylvania (PA) and Tennessee (TN). A sexual state for G. morbida has not been detected that could account for the observed genetic variability, but clonal organisms may have limited recombination events that can change population structure , .
The main objective of our study was to determine the diversity and spatial pattern of G. morbida haplotypes collected widely and intensively from both native and introduced Juglans and Pterocarya hosts in various locations in the USA. Pterocarya, a non-native member of the Juglandaceae in the USA, was included because trees in this genus were recently reported with TCD symptoms . We used multilocus sequence typing (MLST) and a microsatellite (or SSR, simple sequence repeats) analysis comprised of three genomic regions and ten markers, respectively, to observe the population structure of 209 G. morbida isolates collected from 17 different geographic regions (Figure 1B). Specifically we were interested in: i) comparing the population structure of G. morbida from disjunct geographic locations and ii) deducing the potential source of the TCD outbreak and whether multiple introductions of the pathogen have occurred.
Materials and Methods
Fungal collection and isolation
Isolates of G. morbida were collected from 54 counties in 12 states in the USA (Table 1). Samples were collected, transported, and shipped under the terms of USDA-APHIS permit P526P-11-03416. No specific permissions were required for collection of isolates from any of the locations and the field studies did not involve endangered or protected species. It is important to clarify that “population” in the manuscript means a group of geographically co-located isolates. Because G. morbida is primarily asexual, possibly these isolates do not represent genetically interacting entities. For the purpose of analyses, isolates were grouped into one of 17 geographic regions separated by relatively large distances: (1) NM_AZ (New Mexico and Arizona), (2) central AZ, (3) southwestern CA (California), (4) central CA, (5) northern CA, (6) NV (Nevada), (7) southwestern OR (Oregon), (8) central OR, (9) OR_WA (northern OR and southern Washington), (10) ID (Idaho), (11) UT (Utah), (12) northern CO (Colorado), (13) western CO, (14) southern CO, (15) TN (Tennessee), (16) VA (Virginia), and (17) PA (Pennsylvania) (Figure 1B and Table 1). Isolates were cultured from the margins of stem or branch cankers surrounding P. juglandis galleries located in the phloem or bark of symptomatic J. californica, J. cinerea L. , J. hindsii (Jeps.) Jeps. ex R.E. Sm., J. major, J. nigra, and J. regia L., from hybrids of these species, or from undetermined species of Juglans, and from Pterocarya stenoptera C. DC. (Chinese wingnut) and P. fraxinifolia (Lam.) Spach (Caucasian wingnut) (Figure 1B and Table 1). Isolates from J. californica, J. hindsii, J. nigra, and J. regia were collected in areas where TCD epidemics were in progress and where infected trees were readily apparent. Many J. nigra were large trees located in urban areas and often on private property, so sampling was restricted to trees with cankered branches that were easily accessible. Branch samples from J. californica and J. hindsii were collected from trees located in their native ranges (parks, National Forest lands, roadside riparian areas), or from germplasm collections. Samples from J. hindsii were also collected from roadside plantings in the Central Valley (CA) or in OR and WA. Samples from J. regia were collected primarily in Central Valley walnut orchards, with a few exceptions, Cache County, UT and Jefferson County, CO (Table 1). Samples from J. major were collected from trees located in their native range (Figure 1A), which were widely scattered, sparse and did not exhibit severe symptoms of TCD. Thus, for J. major, samples were collected arbitrarily from declining or asymptomatic trees that were identified during road surveys. Isolation from cankers was performed as described previously .
In order to obtain a haploid individual of G. morbida, all isolates were either single spored by serial dilution plating of conidia or by inducing the yeast phase by rapid shaking (250 rpm) of liquid culture, followed by serial dilution plating. A total of 209 haploid isolates were maintained on one-half strength potato dextrose agar (½ PDA, Difco Corp., Sparks, MD, USA) . Of these isolates, 197 or 107 were assessed by MLST or SSR analysis, respectively, whereas 95 were tested by both methods (Table 1). In some cases, multiple G. morbida isolates were collected from different cankers on the same tree, or from more than one Juglans species from the same location (Table 1) to ascertain whether individual trees or adjacent trees contained more than one G. morbida haplotype.
Single-spore isolates were grown on half-strength potato dextrose broth (Difco) for 3 days at 25°C with shaking (120 rpm). Mycelium was collected and lyophilized and DNA extraction was performed by using the Easy DNA Kit (Invitrogen, Carlsbad, CA, USA) or DNeasy Blood & Tissue Kit (Qiagen, Valencia, CA), according to the manufacturer’s recommendations. DNA concentrations and purities were estimated with a spectrophotometer (Thermo Scientific NanoDrop 1000, Loughborough, UK).
Isolation and identification of SSR sequences
DNA was extracted from two single-spore-derived cultures of isolates 1217 (CBS 124663) and 1218 (CBS 124664) (Table 1), known a priori to be characterized by different ITS sequences. DNA from both isolates was pooled and sent to the Cornell University Genomics Facility; enzymatically digested with HincII; ligated; and enriched for microsatellites by hybridization to probes containing random repeats. Enriched fragments were processed into a mate-pair library and sequenced with 454 GS FLX (Roche, Indianapolis, IN, USA) sequencing technology. Sequences were assembled by using SeqMan Pro (Lasergene version 8.1.1; DNASTAR, Madison, WI, USA) yielding a total of 13,392 contigs and 21,737 singlets that were screened for the presence of microsatellite repeats by using EagleView software . Primers were designed from a subset of selected contigs based on: presence of microsatellite repeats; putative presence of polymorphism within a contig; location of repeats within the sequence; and read quality. The software Primer 3  was used to identify primer pairs from contigs. Primer pairs that produced a strong and consistent signal and produced polymorphic amplicons were selected and used to characterize DNA from designated isolates (Table 1, underlined isolates). Sequence data from loci used in this study were deposited in GenBank (http://www.ncbi.nlm.nih.gov/genbank/) (Table S1). For all loci, polymerase chain reaction (PCR) protocol was 4 min denaturation at 94°C, followed by 25 cycles of 45 sec at 94°C, 45 sec at 56°C and 45 sec at 72°C, followed by 8 cycles of 10 sec at 94°C, 45 sec at 51°C and 45 sec at 72°C. At the end of the amplification, the samples were held at 72°C for 11 min and then at 4°C until they could be removed and stored at −20°C for dilution and mixing with ROX standards. Reactions included dNTP (0.3 mM each), bovine serum albumin (10 µg/µl) (New England Biolabs, Ipswich, MA, USA), in a final volume of 15 µl and all other reagents as described previously . Microsatellite haplotypes were produced by using the three primer method , and alleles were binned by using Gene Mapper software (Applied Biosystems, South San Francisco, CA, USA). Quality control of microsatellite markers was guaranteed by genotyping the loci of all samples at least twice and independently. All genotyping plates contained DNA from three isolates (1217, 1218, and 1234), which were used as control (allelic standards). Locus/isolate combinations that produced inconsistent allele sizes or were null were considered as missing data. The ten amplicon sizes (corresponding to each SSR) of each isolate were entered into an Excel spreadsheet and compared visually (manually). Isolates sharing the same amplicon sizes were considered to be clonal.
Isolation and identification of MLST sequences
We attempted to amplify DNA from 16 genomic regions by using the following primer pairs: for housekeeping genes identified in the G. morbida genome; for those developed for Fusarium solani MLST analysis ; for the rRNA internal transcribed spacer (ITS) region ; and for the β-tubulin (BT) gene , to obtain a MLST-based analysis of G. morbida (Table 2). Details on primer design for MLST sequences based on the G. morbida genome are provided in File S1.
Amplification of these 16 regions by PCR consisted of 10X Standard Taq Reaction Buffer (New England Biolabs), 0.2 U of Taq DNA Polymerase, 2 mM MgCl2, 0.2 mM of each dNTP, 0.25 µM of each primer, 20 ng of DNA, and sterile deionized water added for a final volume of 20 µl. Annealing temperatures were initially set 5°C below the lowest Tm of the primer pairs. Because some of the primers were not based on the G. morbida genome, we also performed gradient PCR with annealing temperatures ±5°C of those set initially. Parameters were: 95°C for 5 min; followed by 35 cycles of 95°C for 30 sec, varying annealing temperatures based on the primer pair for 45 sec, 72°C for 1.5 min; and completed with 72°C for 5 min. The presence of amplicons was verified by using a Sub-Cell GT Cell (BioRad, Hercules, CA, USA) electrophoresis system. Amplified sequences that exhibited a single well-defined band were purified by using PureLink PCR Purification Kit (Invitrogen). Ten nanograms of DNA per 100 base pairs (bp) were combined with 10 pmol of primer and sequenced with the BigDye version 3.1 ready reaction kit (Applied Biosystems) on an ABI 3730 automated sequencer at the Proteomics and Metabolomics Facility at Colorado State University. Chromatograms were visualized by using Sequence Scanner Software 1.0 from Applied Biosystems (http://www.appliedbiosystems.com/) and bases were only accepted if the Phred quality score was equal to or greater than 20 . Low quality reads and those containing overlapping sequences were either re-sequenced or sequenced with the primer flanking the opposite side. Sequences were edited and aligned by using BioEdit . Only three of the 16 primer sets amplified genomic regions with SNPs. These resulted in 7, 13, and 7 polymorphisms in the ITS, BT, and methionine aminopeptidase (MAP) sequences (Figure S1) of 197 G. morbida isolates, respectively (Table 2). PCR amplicons of the ITS, BT, and MAP region/genes were 565, 438, and 468 bp in length and the sequences were trimmed to 516, 363, and 365 bp, respectively, for MLST analysis. Sequences were deposited in GenBank with the following accession numbers: ITS, KJ148225 to KJ148419; BT, KJ148030 to KJ148224; and MAP, KJ148420 to KJ148614.
Genetic structure and host specificity.
Haplotypes (isolates in the case of SSR data) were placed into genetic clusters based on the posterior probability of their allele frequencies by using STRUCTURE software  and based on multivariate methods by using Discriminant Analysis of Principal Components (DAPC)  in the Adegenet software package in R . Haplotypes were assigned to the clusters based on independent analyses of the SSR and MLST data. SSR inputs for the analysis were the amplicon sizes of the ten microsatellite markers (Table S1) from each isolate, whereas MLST inputs were each one of the polymorphic sites (SNPs and insertions/deletions) of the three genomic regions (Figure S1). Even though some polymorphic sites were consecutive and likely linked, each one of the 27 (as shown in results) was treated as a single locus (Figure S1). The three MLST genomic regions were aligned by using BioEdit 7.2.4 ; concatenated (total length of 1,244 bp), and then combined into unique haplotypes (Table S1A). Isolates containing identical haplotypes were considered as one, i.e., the dataset was clone-corrected.
A provisional genetic cluster assignment independent of geographical location was determined by using the Bayesian clustering algorithm implemented in the software STRUCTURE . We used the admixture, correlated frequency model, and tested K (number of genetic clusters) = 1 to 10. Parameters were estimated under the null hypothesis of panmixia where loci are at Hardy-Weinberg (H-W) equilibrium. Twenty independent runs were performed with a burn-in period and run length of 50,000 and 500,000 iterations, respectively. The optimal number of populations (K) was estimated by using Structure Harvester  according to the ad hoc statistic ΔK, which is based on the rate of change in the log probability of the data between succeeding K values . The assignment of each isolate/haplotype to a cluster was based on the quality threshold of q, which denotes the admixture proportion for each individual belonging to a cluster. In a situation where cluster x had the highest q-value for a determined individual, this individual was assigned to that cluster only if cluster x had a q-value of ≥0.75, or if the q-value was ≥0.5 and twice the q-value of another cluster. Clones whose q-values did not fit these parameters were not assigned to any specific cluster.
Violations of the assumptions in STRUCTURE (panmixia, H–W equilibrium, linkage equilibrium in clonal subgroups) can produce incorrect assignments; therefore, the results from STRUCTURE were compared to a second genetic cluster assignment by using DAPC. DAPC optimizes variation among clusters to the detriment of variation within clusters and, contrary to the results from STRUCTURE, is neutral to any a priori genetic hypothesis. The differentiation of individuals within and among populations was calculated by using Analysis of Molecular Variance (AMOVA) , which is based on a fixation index (FST), using the Arlequin software package . The resulting FST values were entered into a table and used to determine the most reliable genetic cluster model assignment for G. morbida.
Host specificity was also analyzed independently with both SSR and MLST data by using AMOVA. The analysis was only conducted for host species from which more than five G. morbida-isolates had been collected. FST P-values (α = 0.05) were used to determine host specificity.
Phylogenetic analyses, sexual reproduction/recombination, neutrality, and linkage disequilibrium tests.
Due to the higher robustness of genetic clustering with the MLST data (see Results) and because more isolates were genotyped by using this approach, we performed most of the population genetic analyses with a focus on the MLST data.
Phylogenetic analysis showing the relatedness of G. morbida haplotypes and other Geosmithia species was conducted by using two concatenated genomic regions (ITS/BT). MAP was not included because of the absence of sequence data for other species in public databases. Fifty-five sequences were compared, six of which corresponded to other species of Geosmithia downloaded from GenBank (accessions are shown in Figure 2); the others corresponded to G. morbida haplotypes. Because MAP was not considered, some of the haplotypes sharing the same ITS and BT sequences were combined in a single leaf (e.g., H32 and H33, Figure 2). Sequences were aligned by using ClustalW  and trimmed in MEGA 5.05 , with lengths of 520 bp and 406 bp for ITS and BT, respectively. We performed two analyses: a Bayesian inference of phylogeny with a variant of Markov chain Monte Carlo  in MrBayes v3.1.2  by using a general time-reversible model with inverse-gamma rates of evolution for 1,500,000 generations and a burn-in of 0.25; and a maximum-likelihood analysis in PhyML 3.0  (http://www.atgc-montpellier.fr/phyml/) with default values except for bootstrap of 1,000 replicates. The unrooted phylogenetic tree obtained in MrBayes was formatted in MEGA5  and branches with bootstrap values equal or greater than 50% were shown. Bootstrap values greater than 500 obtained in maximum-likelihood analysis were also indicated in that tree.
A Bayesian analysis was performed for 1,500,000 generations by using a GTR-gamma distributed model of evolution (invariant sites). Bayesian percentages (≥50%) are depicted above each branch, and maximum likelihood bootstrap values (≥500) obtained by using PhyML (default parameters) are shown below most branches. Geosmithia morbida haplotypes are color coded according to their genetic cluster assignment (four-cluster-MLST-DAPC model, as in Figure 3) and haplotypes sharing the same ITS and BT sequences are co-located. Leaves pertaining to the same branch were arranged together according to their cluster assignment. GenBank accession numbers of other Geosmithia spp. are identified within parenthesis.
The program GenClone 2.0  was used to calculate pgen, i.e., the likelihood that MLST haplotypes resulted from sexual reproduction. Neutrality of the polymorphic loci was tested based on coalescent simulations with the program DnaSP 5.10 ; 1,000 replicates were used to test Tajima’s D; Fu and Li’s D, D*, F and F*; and Fu’s F neutrality statistics. MultiLocus 1.2  was used to test linkage disequilibrium of alleles at different loci, and to test the null hypothesis of recombination. Indexes of association IA and r-bar d were obtained with their corresponding probabilities by using 10,000 randomizations.
We used the most robust cluster assignment (four-cluster-MLST-DAPC model, as explained under “Genetic Structure”) to investigate private alleles in SSR analysis, i.e., alleles found in only one cluster or one group of interest. This analysis was performed manually.
Genotype (or haplotype) diversity of each geographic region was also calculated by using DnaSP 5.10 , measuring the genetic variance of haplotypes located in the same region. Significance of the proposed geographic structures was subsequently tested with Hudson’s Permtest, which computes the mean number of pairwise differences within (K*ii or K*jj) and between groups (K*ij) along with the probability that K*ij exceeds K*ii by chance  (http://wwwabi.snv.jussieu.fr/public/mpweb/). Robustness of geographic clusters was calculated by using AMOVA in the Arlequin software package , .
Isolation and identification of SSR sequences
Ten microsatellite (SSR) loci revealed a total of 59 alleles in G. morbida (Table S1). Amplicon sizes were compared, and two isolates (1348 and 1352) were identified as clonal because all but one amplicon from these isolates had the same size. Loci 2514 and 3416 failed to amplify in an unusually high percentage of samples (75% and 63%, respectively), from which other loci amplified consistently, which may indicate the presence of null alleles at those loci (Table S2).
Isolation and identification of MLST sequences
From the set of 16 pairs of primers tested with G. morbida DNA as template (Table 2) all eight based on the Fusarium solani genome  failed to amplify a single well-defined band. The six pairs of primers designed based on the G. morbida genome produced a single amplicon (Table 2), and one pair corresponded to a polymorphic region (MAP) when the sequences of eight experimental isolates were compared (Table 1, isolates with two asterisks). ITS and BT produced polymorphic sequences , as reported in other species of Geosmithia , and were therefore used in this population genetic study. BT was originally amplified by using a pair of primers based on Fusarium spp. genome , but after sequencing the corresponding fragment from G. morbida one primer was refined accordingly (Table 2). We did not detect sequences with consistent noise or base call overlap, which would be an indication of different variants (alleles) of multi-copy genes (or genomic regions) within the same individual.
Using MLST data, 197 G. morbida isolates were placed into one of 57 haplotypes based on 11, 16, and 4 variants of ITS, BT, and MAP, respectively (Table S2 and Table 1), according to the combination of 7, 13, and 7 polymorphic loci, respectively (Figure S1). Multiple haplotypes were identified from isolates collected from different cankers on the same tree in cases from AZ, CA, CO, NM, TN, and VA (Table 1). Twenty-four of the haplotypes (42%) were represented by more than one isolate (Table 1). Overall, haplotypes H02 and H03 occurred the most frequently in the survey, representing 36% of all G. morbida isolates examined (Figure 3). They were detected in all geographic regions where more than 4 isolates were collected, except neither haplotype occurred in the native range of J. major (central AZ and NM_AZ), and haplotype H03 did not occur in southern CO (Figure 3). Haplotype H02 was especially abundant in the three geographic regions in CA (southwestern, central, and northern). Haplotype H03 was also abundant in northern CA and central CA but less so in the native range of J. californica in southwestern CA. Most of the geographic regions shared haplotypes with other regions, with the exception of central AZ and NM_AZ. In fact, most of the haplotypes present in those two regions were either exclusive or only shared between them, and all isolates from those locations were collected from J. major (Figure 3). The exceptions were haplotypes H38 and H40, which were also collected from a single county (Jefferson County) in northern CO, but from J. nigra. Isolate 1503 (Table 1) was not grouped with the other NM_AZ isolates because it was collected from J. nigra in an urban planting in Albuquerque (NM) and carried a haplotype that was not found in the native range of J. major.
The size of wedges in each pie chart is proportional to the number of isolates. Haplotype colors relate to genetic clusters identified in the four-cluster-MLST-DAPC model where cluster 1 = shades of blue, cluster 2 = shades of red/brown, cluster 3 = shades of yellow, and cluster 4 = shades of green. Callouts are color-coded according to the three-region geographic-Hudson’s Permtest model, where: 1) blue = NM_AZ, central CA, northern CA, northern CO, and TN; 2) green = central AZ; and 3) red = southwestern CA, OR_WA, and southern CO. Callouts in white indicate regions not assessed by using Hudson’s Permtest. Counties are color-coded as in Figure 1B. (U.S. map adapted from US Census Bureau at https://www.census.gov/).
The SSR data was not organized into haplotypes because no repeated haplotypes were identified when the allele sizes of the ten loci were analyzed, with the exception of isolates 1348 and 1352, as described above.
According to analysis with STRUCTURE, 107 G. morbida isolates were best organized into two clusters based on SSR allele frequency with a smaller ΔK-peak at 4 (Figure S2). Similarly, the clone-corrected MLST data from 57 haplotypes resulted in a ΔK peak at 4 following analysis with STRUCTURE (Figure S2).
Assignment of the SSR data to genetic clusters by using DAPC resulted in an optimal theoretical K = 6 (Figure 4A). One cluster containing only three isolates, 1234, 1303, and 1307 (in green in Figure 4A) segregated from a group of five clusters. When the three distant isolates were removed and data re-analyzed, the remaining isolates were re-organized into three clusters (Figure 4B). Therefore, in further SSR-DAPC analysis we considered a K = 4, comprising the one cluster containing the three distant isolates (in green in Figure 4A) as well as the three additional clusters in Figure 4B. The fixation index of K = 4 was similar to the six-clusters-SSR-DAPC model obtained originally (Table 3).
The MLST data analyzed with DAPC was first assigned to two clusters containing 55 and 2 haplotypes each (Figure 5A). When the distant cluster containing haplotypes H32 and H33 was excluded (in green in Figure 5A), the remaining 55 haplotypes segregated into three distinct clusters (Figure 5B). Fixation indices were similar for both two and four MLST-DAPC-clusters (Table 3). When the cluster assignment of the haplotypes on the four-cluster-MLST-STRUCTURE model was compared to the four-cluster-MLST-DAPC, they were mostly correlated, despite some discrepancies (Figure 5A, green cluster; and Figure 5B, all clusters). The four-cluster-MLST-DAPC model had a high (>0.25) fixation index and it was the most informative when compared to other models of genetic and geographic (as described below) clustering of G. morbida isolates. Therefore, the four-cluster-MLST-DAPC model was used to assign haplotypes to clusters in further analyses (Table 1 and Figure 3).
The most distant cluster (cluster 4 in green) comprised of haplotypes H32 and H33 is identified (A), as well as the coordinates of all remaining haplotypes when H32 and H33 were excluded (B). A comparison between the assignments of the MLST-DAPC and MLST-STRUCTURE models are shown in detail. Pie charts give the probability of assignment of haplotypes to the four genetic clusters obtained in the four-clusters-MLST-STRUCTURE model. They are represented by colors, cluster 1 = blue, cluster 2 = red, cluster 3 = yellow and cluster 4 = green. Haplotypes in the box (in B) were amplified for better resolution.
Examining the haplotype assignment, based on the four-cluster-MLST-DAPC model, cluster 1 (blue color in pie charts of Figure 3) contained 21 haplotypes (63 isolates) including the abundant and widely distributed haplotype H03 (Table 1 and Figure 3). Cluster 1 was especially abundant in northern CO, northern CA, central CA and TN. Cluster 2 (red in pie charts, Figure 3) contained 25 haplotypes (103 isolates), including the abundant and widely distributed haplotype H02. Cluster 2 was especially abundant in the three regions in CA, OR_WA, and in southern CO, where it accounted for more than half of the individuals (Table 1 and Figure 3). Cluster 3 (yellow/amber in pie charts, Figure 3) was represented by 9 haplotypes (24 isolates) and was abundant in central AZ and NM_AZ, but not present in most of the other regions, except for two haplotypes in northern CO and one in southwestern CA and TN (Table 1 and Figure 3). Cluster 4 (green in pies charts, Figure 3) contained 2 haplotypes (7 isolates) that were only detected from J. major in central AZ (Table 1 and Figure 3).
We found correspondence in cluster assignment between MLST and SSR data by using DAPC (Figure 4A, 4B, 5A, and 5B). The outliers (isolates) 1234, 1303, and 1307 in SSR analysis (Figure 4A) corresponded to H33 (Figure 5A), which is also one of the outliers in MLST analysis. The single isolate (1299) pertaining to H32 (Figure 5A) was not assessed by the SSR method. Most of the isolates (75%) in SSR-cluster 1 corresponded to cluster 2 in MLST analysis, and most of the isolates (47%) in SSR-cluster 3 corresponded to cluster 3 in MLST analysis. The SSR-cluster 2 was comprised of 67% of MLST-cluster 2 and 33% of MLST-cluster 1 isolates.
Based on pairwise FST analysis, the SSR and MLST data indicated that J. major isolates were different from J. californica, J. hindsii, J. nigra and J. regia isolates collectively, whereas J. nigra isolates were different from J. californica isolates (Table S3).
In a phylogenetic tree based on Bayesian inference with ITS and BT sequences (Figure 2), all 57 haplotypes of G. morbida (Table S2) grouped in the same clade with strong bootstrap support, apart from several other Geosmithia species. Within the G. morbida clade, isolates in the haplotypes H32 and H33 formed their own sub-clade with strong bootstrap support and the remaining haplotypes clustered together into a second sub-clade.
Sexual reproduction/recombination, neutrality, and linkage disequilibrium tests.
Independent analyses of both the whole set of 57 haplotypes and haplotypes assigned to MLST-DAPC clusters showed that the likelihood of the observed haplotypes arising from sexual reproduction was extremely low (pgen <0.05). In both the complete dataset and clone-corrected data, the rd test rejected the null hypothesis of recombination in MLST loci (P<0.001) and we obtained similar results by using SSR data (P<0.001). The polymorphic MLST-loci were all selectively neutral, whereas polymorphisms in SSR loci were presumed neutral as they were extra-genic. Overall, the three genomic regions used in the MLST analysis were in linkage disequilibrium (index of association = 0.8005, P<0.001) and pairwise tests indicated the same (index of association = 0.8005, P = 0.029).
Private alleles may indicate restrictions in dispersal or evolutionary isolation. The four-cluster-MLST-DAPC model was used to investigate private alleles in SSR analysis (Table S2). In this regard, isolates in cluster 4, showed a high number of private alleles and included private alleles at seven of ten loci examined. Moreover, the most common allele at locus 194 (270 bp), locus 2849 (310 bp and 316), and locus 4045 (184 bp) were completely absent from isolates of cluster 4, reinforcing their genetic distinctiveness. Isolates belonging to cluster 3 showed private alleles at SSR loci 223, 194, 1851, and 6823; cluster 2 had private alleles at 2849, 3416 and 7713, and cluster 1 at 4045, 6823 and 7713 (Table S2).
Geosmithia morbida isolates were placed initially into one of 17 geographic regions based on the proximity of the counties where the isolates were collected. These regions, while arbitrarily determined, were for the most part spatially isolated from one another (Figure 1B). Even though the fixation index (FST = 0.167) (Table 3) was not low when the 17 regions were tested, the population structure in some regions was not different from others according to Hudson’s Permtest . Because of that, a series of Hudson’s test-based pairwise comparisons of 8 and then 6 regions (regions that we compared are described in Table 3) that included more than four isolates indicated that the populations could be combined into three different (P<0.001) macro regions encompassing: 1) NM_AZ, central CA, northern CA, northern CO, and TN; 2) central AZ and 3) southwestern CA, OR_WA, and southern CO (depicted in callouts in Figure 3). Therefore the three-region geographic model was the most robust when compared to other geographic models. This conclusion is based on fixation indices (FST) that identified the lowest within and highest among population variation (Table 3), and all pairwise FST values were significant (Table 4). The haplotype diversity was high across all three macro regions [0.70 (SD±0.02)]. The values obtained for the three regions individually were 0.67 (SD±0.03), 0.77 (SD±0.08), and 0.54 (SD±0.08) in: 1) NM_AZ/Ce CA/No CA/No CO/TN; 2) Ce AZ; and 3) Sw CA/OR_WA/So CO, respectively.
Isolates and genetic diversity
Geosmithia morbida was isolated consistently from cankers surrounding WTB galleries from all Juglans and Pterocarya showing TCD symptoms and in all regions that we surveyed (Fig. 1A), with the exception of J. major in Texas where the WTB was not detected . Considering the substrate specificity of bark beetles and Geosmithia associates to their respective plant hosts , and the highly diverse and complex genetic structure in G. morbida that we and others , ,  have observed, it is unlikely that the current TCD epidemic was a result of a new association between the WTB and G. morbida.
The number of polymorphic MLST loci described here and elsewhere  was not low, as we had first expected. However, it is unlikely that haplotypes are admixed (sexual recombinants) because no teleomorph has been observed and according to tests for sexual reproduction, neutrality, and disequilibrium, sexual recombination in G. morbida is nonexistent or infrequent. Considerable intra-species variation of the ITS rDNA region has been reported in other Geosmithia species , , and this may be a common characteristic of this genus. Thus, this variability does not necessarily support the hypothesis that G. morbida is a species complex.
All isolates were identified initially as G. morbida based on the similarity of their morphological characteristics and their growth patterns and color in culture. However, the placement of haplotypes H32 and H33 in a group distinct from the other G. morbida haplotypes according to the ITS/BT-based phylogenetic tree and based on analyses of genetic and geographic cluster assignments could be viewed as evidence for a hypothetical cryptic species (Figure 2, 4A, and 5A). All isolates characterized by H32 and H33 were derived from J. major, whereas the other haplotypes were collected from many Juglans and Pterocarya species. However, in previous work, our team has found that isolate 1234 (characterized by H33 and first documented from J. major in Kolařík et al. 2011 ) is indistinguishable from other G. morbida isolates in terms of its morphology and pathogenicity. A more thorough multi-gene phylogenetic analysis in the future including other species of Geosmithia and more isolates of G. morbida from J. major in the southwestern USA and Mexico might be necessary to investigate the potential for cryptic species related to G. morbida.
Genetic, geographic, and host groupings
Our results indicated that the G. morbida genetic structure in the USA was best explained by four distinct genetic groups. Cluster assignment correspondence between MLST and SSR by using the DAPC method was moderate, probably due to some missing data in SSR (Table S2). The most robust analysis was achieved by DAPC analysis with MLST data (Figure 5 A, B and Table 3), with support of STRUCTURE. The four-cluster-MLST-DAPC model differed from the two clusters identified by Hadziabdic and collaborators , probably because we sampled more extensively and in areas not sampled in their study.
Haplotypes in genetic clusters 1 and 2 (blue and red) dominated in locations outside of AZ and NM (i.e., outside the native range of J. major); vice versa for clusters 3 and 4 (yellow and green) (Figure 3). The high proportion of widely-dispersed, closely-related G. morbida haplotypes supports the hypothesis of a recent invasion by at least three genetic groups into regions other than central AZ and NM_AZ (Figure 3).
The haplotypes were further grouped into three diverse geographic groups (callouts in Figure 3), possibly indicating that regions within the same group exchanged or had a common source of haplotypes. A comparable genetic and geographic structure was observed for Ophiognomonia clavigignenti-juglandacearum, an exotic canker pathogen of J. cinerea in North America , which consisted of three genetic clusters of haplotypes that were geographically disjunct. In the past, many epidemics caused by invasive plants, pathogens, or insects were assumed to be initiated by a single introduction event followed by a radiation from the introduction point. However, several recent studies have found that this is likely not the case and that, in most invasion events, multiple introductions have occurred followed by migration and admixture of populations –. This may be particularly true for forest pathogens as damage caused by these organisms may go unnoticed and undetected for decades in more remote areas. This allows greater migration and genetic drift to occur between and among populations. Based on the amount of genetic diversity observed in the ITS, BT, and MAP regions and the fact that G. morbida is present but does not cause mortality in some Juglans species, it is likely that this fungus was present in at least one location in North America for decades, being moved virtually unnoticed amongst Juglans populations by the WTB. However, once mortality on J. nigra was observed out of its natural range and the fungus was detected and described for the first time, the full extent of the distribution of G. morbida was finally realized , , .
Although we found some evidence of host specialization, these data were strongly influenced by the uneven geographic distribution of the Juglans species, i.e., host and region were highly confounded. Geosmithia morbida isolates from J. californica and J. major were derived almost exclusively from trees in or near their native ranges, whereas isolates from other hosts were collected in multiple regions and from planted trees. Isolates from the Eurasian native, J. regia, were derived primarily from adventive orchards in CA and two outlying samples from CO and UT. Moreover, the number of isolates collected from the hosts was extremely unbalanced. Further studies that extend the work of Utley et al.  by evaluating virulence of selected isolates on various hosts are needed to confirm any evidence of host specialization.
Introduction of multiple G. morbida haplotypes to geographically isolated areas
We recovered multiple G. morbida haplotypes representing different genetic clusters from different cankers on the same tree; their introduction could have occurred in different events or all at once from a single WTB-infested log. Even small infested logs harbor large numbers of beetles . The WTB is not considered a strong flier  and even though G. morbida produces dry conidia, they are unlikely to be dispersed by the wind because the spores are only formed in beetle galleries and feeding sites . Therefore, the most probable means of G. morbida introduction into geographically isolated areas is by importation of contaminated logs or wood products (with bark attached) . The anthropogenic movement of wood containing many haplotypes would explain the complex mixture of G. morbida haplotypes observed at most locations.
The high proportions of haplotypes H02 and H03 in most of the regions may represent their relative abundance in the original or secondary invasive population; the frequency of haplotype reintroduction during subsequent dispersal events; or it could indicate some type of competitive advantage , .
Possible origin of G. morbida isolates in the TCD epidemic: J. major native range as primary source
We hypothesized that G. morbida was dispersed from the native range of J. major, where the WTB vector was first collected , . There is circumstantial evidence of an isolated introduction of G. morbida (H38 and H40) from central AZ into northern CO (yellow in pie charts in Figure 3), but these haplotypes represent a small fraction of the total number. Thus, the two haplotypes were probably part of an independent and secondary introduction of G. morbida into northern CO.
In general, we did not detect haplotypes collected from J. major in AZ and NM in other areas in the USA. Although we did not find shared haplotypes, the Hudson’s Permtest  analysis indicated that the genetic makeup of haplotypes from NM_AZ was not significantly different from those found in central CA, northern CA, northern CO, and TN (blue callouts in Figure 3). The high number of exclusive G. morbida haplotypes collected from J. major in central AZ and NM_AZ (Figure 3) and the large number of private alleles, combined with the absence in this region of the two most frequently occurring haplotypes (H02 and H03), make it unlikely that these two regions are the direct geographic sources of G. morbida isolates associated with TCD outbreaks in the majority of locations. Juglans major has an extensive but non-contiguous range that extends south into Mexico , and our sampling was restricted to the extreme northern portion of the full range (Figure 1A) of this species. If G. morbida is present throughout the range of J. major, its population structure may be variable because of genetic isolation. Even within central AZ and NM_AZ, we noted what appeared to be geographically isolated genetic clusters of G. morbida (Table 1). A more intensive survey in unsampled regions of J. major is warranted.
Possible origin of G. morbida isolates in the TCD epidemic: J. californica native range as primary source
Another possible source of G. morbida haplotypes causing the TCD epidemic could be southwestern CA. All G. morbida isolates from this region were collected from J. californica, a species that may be a native host of the WTB and its symbiont G. morbida . Moreover, a very high proportion (89%) of the haplotypes identified in southwestern CA, in particular haplotypes H02 and H03, was also found in at least one other region (Figure 3). Furthermore, the result of the geographic structure analysis indicated that southwestern CA was similar to OR_WA and southern CO (Figure 3).
Geographic structure analysis also indicated genetic similarities among G. morbida haplotypes in northern CO, TN, central CA and northern CA (Figure 3), with the latter region representing the natural range of J. hindsii (Figure 1A). Central CA and northern CA shared 88% and 83% of their haplotypes with other regions (Figure 3). It is possible that the TCD epidemic was triggered when infested raw walnut wood products were moved from northern CA to northern CO, or vice versa.
Geosmithia morbida dispersal
Regardless of its origin, TCD was first noted after its appearance on the highly susceptible J. nigra , , probably as a consequence of transportation of infested logs that allowed movement of WTB and G. morbida into regions that they would not have reached naturally. Introduction to new sites could also have been facilitated by tourists/campers carrying even small amounts of wood. For example, there were more than 13 million visitors to Arizona’s State and National Parks from 2001–2012 (www.azot.gov). Visitor attendance during this period to California’s State Parks alone was even higher, ranging from 85.5 million (2001–2002) to 67.9 million (2011–2012) (http://www.parks.ca.gov/?page_id=23308). The genetic makeup of the G. morbida populations in regions where J. nigra has been affected suggests that the outbreaks in most of the USA are related to those in CA and from a fungus population that was already established in at least one species of Juglans.
This study helped to explain the genetic structure of G. morbida and its multi-factorial and disjunct distribution in the USA. Even though the observation of TCD in Juglans and Pterocarya species has only been recent, the observed haplotype diversity and the genetic complexity of G. morbida indicate it has been in association with at least one Juglans spp. and the WTB for a long period of time. Some haplotypes and genetic clusters were found in specific regions and in association only with certain Juglans species. The scattered geographic distribution of genetic clusters indicates that G. morbida was disseminated several times and from several sources and most likely by transportation in beetle-infested wood.
Twenty-seven nucleotide polymorphic loci detected in three genomic regions in G. morbida. Sequencing ITS, BT and MAP of G. morbida isolates resulted in 7, 13 and 7 polymorphisms, respectively. Those were combined in 57 haplotypes (Table S2). Polymorphisms are depicted in colors and loci are identified.
Results of ΔK computation for Geosmithia morbida isolates. Scenario where K = 1-10 (20 iterations) by using STRUCTURE . SSR data supports clusters of 2 and 4, whereas MLST data supports a cluster of 4.
Microsatellite (SSR) loci used to characterize Geosmithia morbida isolates.
Collection Information of 209 Geosmithia morbida-isolates and their MLST and SSR profile.
Host specificity analysis. FST values calculated using both MLST (above diagonal) and SSR (below diagonal) data, grouping isolates by host. Significant (P<0.05) FST values are denoted by (*).
We thank R.M. Bostock, T.W. Coleman, P.L. Dallara, G. Durham, E.J. Fichtner, S. Fraedrich, K.J. Greby, B. Hammon, J.E. Henrich, S.M. Hishinuma, D. Leatherman, C.A. Leslie, A. Liu, J. McKenna, L.M. Ohara, M. Putnam, S. Schlarbaum, C. Utley, and D.L. Wood for assistance in collecting samples. We also thank E. Freeland, E. Luna, N. Thiry, R. Billings, L. Holder, R. Feild, M. Shenk, M. Nelson, R. Sitz and M. Curto for providing technical support. Mention of a trademark, proprietary product, or vendor does not constitute a guarantee or warranty of the product by the U.S. Dept. of Agriculture and does not imply its approval to the exclusion of other products or vendors that also may be suitable.
Conceived and designed the experiments: KW WC SJS NT. Performed the experiments: MMZ CH JT. Analyzed the data: MMZ JIC KW KB. Contributed reagents/materials/analysis tools: ADG JP WC SJS NT. Wrote the paper: MMZ JIC KW ADG JP KB WC SJS NT. Performed statistical analyses: JIC.
- 1. Williams RD (1990) Juglans nigra L., black walnut. In: Burns RM, Honkala, BH, tech coords. 2. Hardwoods. Agriculture Handbook 654. U.S. Department of Agriculture, Forest Service, Washington, DC. vol. 2: pp. 391–410.
- 2. Hardin JW, Leopold DJ, White FM (2000) Harlow & Harrar’s Textbook of Dendrology: New York: McGraw-Hill. 544 p.
- 3. Nicodemus MA, Salifu KF, Jacobs DF (2008) Nitrate reductase activity and nitrogen compounds in xylem exudate of Juglans nigra seedlings: relation to nitrogen source and supply. Trees 22: 685–695.
- 4. Tisserat N, Cranshaw W, Putnam ML, Pscheidt J, Leslie CA, et al.. (2011) Thousand cankers disease is widespread in black walnut in the western United States. Online. Plant Health Progress. doi:10.1094/PHP-2011-0630-01-BR. Accessed 10 July 2014.
- 5. Walnut (Juglans spp.)-Thousand cankers disease (black walnut decline). In: Pscheidt JW, Ocamb CM, eds. (2014) Pacific Northwest plant disease management handbook [online]. Corvallis, OR: Oregon State University. http://pnwhandbooks.org/plantdisease/walnut-juglans-spp-thousand-cankers-disease-black-walnut-decline. Accessed 10 July 2014.
- 6. USDA-FS (2001) Forest Insect and Disease Conditions of the Southwestern Region. USDA-FS Southwestern Region, Forestry and Forest Health Publication R3-02-01. 21 p.
- 7. Bright DE Jr (1981) Taxonomic monograph of the genus Pityophthorus Eichhoff in North and Central America (Coleoptera: Scolytidae). Memoirs of the Entomological Society of Canada 113: 1–378.
- 8. Bright DE Jr (2014) A catalog of Scolytidae and Platypodidae (Coleoptera), Supplement 3 (2000–2010), with notes on subfamily and tribal reclassifications. Insecta Mundi 356: 1–336.
- 9. Tisserat N, Cranshaw W, Leatherman D, Utley C, Alexander K (2009) Black walnut mortality in Colorado caused by the walnut twig beetle and thousand cankers disease. Online. Plant Health Progress. doi:10.1094/PHP-2009-0811-01-RS. Accessed 10 July 2014.
- 10. Kolařík M, Freeland E, Utley C, Tisserat N (2011) Geosmithia morbida sp. nov., a new phytopathogenic species living in symbiosis with the walnut twig beetle (Pityophthorus juglandis) on Juglans in USA. Mycologia 103: 325–332.
- 11. Seybold SJ, Haugen D, O’Brien J, Graves AD (2013) Thousand cankers disease. USDA Forest Service, Northeastern Area, State and Private Forestry Pest Alert NA-PR-02-10. http://www.na.fs.fed.us/pubs/detail.cfm?id=5225. Accessed 10 July 2014.
- 12. Grant JF, Windham MT, Haun WG, Wiggins GJ, Lambdin PL (2011) Initial assessment of thousand cankers disease on black walnut, Juglans nigra, in eastern Tennessee. Forests 2: 741–748
- 13. Haun WG, Powell S, Strohmeier C, Kirksey J (2010) State of Tennessee thousand cankers disease action plan: Tennessee Department of Agriculture, 34 p.
- 14. Hanson MA, Bush EA, Day E, Griffin G, Dart NL (2011) Walnut Thousand Cankers Disease Alert. Virginia Cooperative Extension, 4 p.
- 15. Hadziabdic D, Windham MT, Baird R, Vito L, Cheng Q, et al. (2014) First report of Geosmithia morbida in North Carolina: The pathogen involved in thousand cankers disease of black walnut. Plant Disease 98: 992 http://dx.doi.org/10.1094/PDIS-06-13-0630-PD.
- 16. Services. VDoAaC (2011) VDACS - Plant and pest services - Thousand cankers disease.
- 17. Force PFPT (2012) Commonwealth of Pennsylvania thousand cankers disease action plan.
- 18. Fisher JR, McCann DP, Taylor NJ (2013) Geosmithia morbida, thousand cankers disease of black walnut pathogen, was found for the first time in southwestern Ohio. Online. Plant Health Progress. doi:10.1094/PHP-2013-1201-01-BR. Accessed 10 July 2014.
- 19. Montecchio L, Faccoli M (2014) First record of thousand cankers disease Geosmithia morbida and walnut twig beetle Pityophthorus juglandis on Juglans nigra in Europe. Plant Disease. 98: 696 http://dx.doi.org/10.1094/PDIS-10-13-1027-PDN.
- 20. Cranshaw W (2011) Recently recognized range extensions of the walnut twig beetle, Pityophthorus juglandis Blackman (Coleoptera: Curculionidae: Scolytinae), in the western United States. Coleop Bull 65: 48–49.
- 21. Blackman MW (1928) The genus Pityophthorus Eichh. in North America: a revisional study of the Pityophthori, with descriptions of two new genera and seventy-one species. Bull New York State College of Forestry 1(3–6): 183.
- 22. Seybold SJ, Coleman TW, Dallara PL, Dart NL, Graves AD, et al. (2012) Recent collecting reveals new state records and geographic extremes in the distribution of the walnut twig beetle, Pityophthorus juglandis Blackman (Coleoptera: Scolytidae), in the United States. Pan-Pac Entomol 88: 277–280.
- 23. Graves AD, Coleman TW, Seybold SJ (2011) Monitoring walnut health and decline in response to thousand cankers disease and infestation by the walnut twig beetle, Pityophthorus juglandis, in southern California and New Mexico. Year one Progress Report for USDA Forest Service Forest Health Monitoring Project INT-EM-B-11-03. http://fhm.fs.fed.us/em/funded/12/INT-EM-B-11-03.pdf. Accessed 10 July 2014.
- 24. Griffin JR, Critchfield WB (1972) The distribution of forest trees in California. USDA Forest Service Research Paper PSW-82, Pacific Southwest Forest and Range Experiment Station. 114 p.
- 25. Newton LP, Fowler G, Neeley AD, Schall RA, Takeuchi Y (2009) Pathway Assessment: Geosmithia sp. and Pityophthorus juglandis Blackman movement from the western into the eastern United States. United States Department of Agriculture Animal and Plant Health Inspection Service Rev. 1: 10.19.2009. http://tn.gov/agriculture/publications/regulatory/tc_pathwayanalysis.pdf Accessed 10 July 2014.
- 26. Freeland E, Cranshaw W, Tisserat N (2012) Effect of Geosmithia morbida isolate and temperature on canker development in black walnut. Online. Plant Health Progress. doi:10.1094/PHP-2012-0618-01-RS. Accessed 10 July 2014.
- 27. Hadziabdic D, Vito LM, Windham MT, Pscheidt JW, Trigiano RN, et al. (2013) Genetic differentiation and spatial structure of Geosmithia morbida, the causal agent of thousand cankers disease in black walnut (Juglans nigra). Curr Genet 60: 75–87.
- 28. Halkett F, Simon JC, Balloux F (2005) Tackling the population genetics of clonal and partially clonal organisms. Trends Ecol Evol 20: 194–201 DOI:https://doi.org/10.1016/j.tree.2005.01.001.
- 29. Fournier E, Giraud T (2008) Sympatric genetic differentiation of a generalist pathogenic fungus, Botrytis cinerea, on two different host plants, grapevine and bramble. J Evol Biol 21: 122–132.
- 30. Serdani M, Vlach JJ, Wallis KL, Zerillo MM, McCleary T, et al.. (2013) First report of Geosmithia morbida and Pityophthorus juglandis causing thousand cankers disease in butternut. Plant Health Progress. doi:10.1094/PHP-2013-1018-01-BR. Accessed 10 July 2014.
- 31. Huang W, Marth G (2008) EagleView: a genome assembly viewer for next-generation sequencing technologies. Genome Res 18: 1538–1543.
- 32. Rozen S, Skaletsky H (1999) Primer3 on the WWW for general users and for biologist programmers. Bioinformatics methods and protocols: Springer. 365–386.
- 33. Schuelke M (2000) An economic method for the fluorescent labeling of PCR fragments. Nature Biotechnology 18: 233–234.
- 34. Debourgogne A, Gueidan C, Hennequin C, Contet-Audonneau N, de Hoog S, et al. (2010) Development of a new MLST scheme for differentiation of Fusarium solani species complex (FSSC) isolates. J Microbiol Methods 82: 319–323.
- 35. Gardes M, Bruns T (1993) ITS primers with enhanced specificity for basidiomycetes-application to the identification of mycorrhizae and rusts. Mol Ecol 2: 113–118.
- 36. O’Donnell K, Cigelnik E (1997) Two divergent intragenomic rDNA ITS2 Types within a monophyletic lineage of the fungus Fusarium are nonorthologous. Mol Phylogenet Evol 7: 103–116.
- 37. Ewing B, Green P (1998) Base-calling of automated sequencer traces using phred. II. Error probabilities. Genome Res 8: 186–194.
- 38. Hall TA (1999) BioEdit: a user-friendly biological sequence alignment editor and analysis program for Windows 95/98/NT. Nucleic Acids Symposium Series 41: 95–98.
- 39. Pritchard JK, Stephens M, Donnelly P (2000) Inference of population structure using multilocus genotype data. Genetics 155: 945–959.
- 40. Jombart T, Devillard S, Balloux F (2010) Discriminant analysis of principal components: a new method for the analysis of genetically structured populations. BMC Genetics 11: 94.
- 41. Jombart T, Ahmed I (2011) Adegenet 1.3–1: new tools for the analysis of genome-wide SNP data. Bioinformatics 27: 3070–3071.
- 42. Earl DA, vonHoldt BM (2011) Structure Harvester: a website and program for visualizing Structure output and implementing the Evanno method. Conserv Gen Res 4: 359–361.
- 43. Evanno G, Regnaut S, Goudet J (2005) Detecting the number of clusters of individuals using the software Structure: a simulation study. Mol Ecol 14: 2611–2620.
- 44. Excoffier L (1995) AMOVA: Analysis of molecular variance (version 1.55). Geneva, Switzerland: University of Geneva. 47 p.
- 45. Excoffier L, Laval G, Schneider S (2005) Arlequin (version 3.0): an integrated software package for population genetics data analysis. Evolutionary Bioinformatics Online 1: 47.
- 46. Larkin MA, Blackshields G, Brown NP, Chenna R, McGettigan PA, et al. (2007) Clustal W and Clustal X version 2.0. Bioinformatics 23: 2947–2948.
- 47. Tamura K, Peterson D, Peterson N, Stecher G, Nei M, et al. (2011) MEGA5: molecular evolutionary genetics analysis using maximum likelihood, evolutionary distance, and maximum parsimony methods. Mol Biol Evol 28: 2731–2739.
- 48. Green PJ (1995) Reversible jump Markov chain Monte Carlo computation and Bayesian model determination. Biometrika 82: 711–732.
- 49. Huelsenbeck JP, Ronquist F (2001) MrBayes: Bayesian inference of phylogenetic trees. Bioinformatics 17: 754–755.
- 50. Guindon S, Dufayard JF, Lefort V, Anisimova M, Hordijk W, et al. (2010) New algorithms and methods to estimate maximum-likelihood phylogenies: assessing the performance of PhyML 3.0. System Biol 59: 307–321.
- 51. Arnaud-Haond S, Belkhir K (2007) GenClone: a computer program to analyse genotypic data, test for clonality and describe spatial clonal organization. Mol Ecol Notes 7: 15–17.
- 52. Rozas J, Sánchez-DelBarrio JC, Messeguer X, Rozas R (2003) DnaSP, DNA polymorphism analyses by the coalescent and other methods. Bioinformatics 19: 2496–2497.
- 53. Agapow PM, Burt A (2001) Indices of multilocus linkage disequilibrium. Mol Ecol Notes 1: 101–102.
- 54. Hudson RR, Boos DD, Kaplan NL (1992) A statistical test for detecting geographic subdivision. Mol Biol Evol 9: 138–151.
- 55. Freeland E (2012) Intraspecific variability of Geosmithia morbida the causal agent of thousand cankers disease, and effects of temperature, isolate and host family (Juglans nigra) on canker development. Fort Collins, CO: Colorado State University. 91 p.
- 56. Kolařík M, Jankowiak R (2013) Vector affinity and diversity of Geosmithia fungi living on subcortical insects inhabiting Pinaceae species in central and northeastern Europe. Microb Ecol 66: 682–700.
- 57. Broders KD, Boraks A, Sanchez AM, Boland GJ (2012) Population structure of the butternut canker fungus, Ophiognomonia clavigignenti-juglandacearum, in North American forests. Ecol Evol 2: 2114–2127.
- 58. Goss EM, Larsen M, Vercauteren A, Werres S, Heungens K, et al. (2011) Phytophthora ramorum in Canada: Evidence for migration within North America and from Europe. Phytopathology 101: 166–171.
- 59. Keller SR, Gilbert KJ, Fields PD, Taylor DR (2012) Bayesian inference of a complex invasion history revealed by nuclear and chloroplast genetic diversity in the colonizing plant, Silene latifolia. Molecular Ecology 21: 4721–4734.
- 60. Ciosi M, Miller NJ, Kim KS, Giordano R, Estoup A, et al. (2008) Invasion of Europe by the western corn rootworm, Diabrotica virgifera: multiple transatlantic introductions with various reductions of genetic diversity. Molecular Ecology 17: 3614–3627.
- 61. Delmotte F, Giresse X, Richard-Cervera S, M’Baya J, Vear F, et al. (2008) Single nucleotide polymorphisms reveal multiple introductions into France of Plasmopara halstedii, the plant pathogen causing sunflower downy mildew. Infection Genetics and Evolution 8: 534–540.
- 62. Utley C, Nguyen T, Roubtsova T, Coggeshall M, Ford TM, et al. (2013) Susceptibility of walnut and hickory species to Geosmithia morbida. Plant Disease 97: 601–607.
- 63. Sitz R (2013) Management options for the walnut twig beetle, Pityophthorus juglandis Blackman, vector of the fungal canker pathogen Geosmithia morbida: Colorado State University. 68 p.
- 64. Little EL, Jr. (1979) Checklist of United States trees (native and naturalized). US Department of Agriculture Forest Service, Agriculture Handbook No. 541. 375 p.