Genetic Analyses of the Laurel Wilt Pathogen , Raffaelea lauricola , in Asia Provide Clues on the Source of the Clone that is Responsible for the Current USA Epidemic

Laurel wilt is caused by the fungus Raffaelea lauricola T.C. Harr., Fraedrich and Aghayeva, a nutritional symbiont of its vector the redbay ambrosia beetle, Xyleborus glabratus Eichhoff. Both are native to Asia but appeared in Georgia in the early 2000s. Laurel wilt has since spread to much of the southeastern United States killing >300 million host trees in the Lauraceae plant family. The aims of this research were to elucidate the genetic structure of populations of R. lauricola, to examine its reproductive strategy, and determine how often the pathogen had been introduced to the USA. A panel of 12 simple sequence repeat (SSR) markers identified 15 multilocus genotypes (MLGs) in a collection of 59 isolates from the USA (34 isolates), Myanmar (18), Taiwan (6) and Japan (1). Limited diversity in the USA isolates and the presence of one MAT idiotype (mating type locus) indicated that R. lauricola was probably introduced into the country a single time. MLG diversity was far greater in Asia than the USA. Only three closely related MLGs were detected in the USA, the most prevalent of which (30 of 34 isolates) was also found in Taiwan. Although more work is needed, the present results suggest that a Taiwanese origin is possible for the population of R. lauricola in the USA. Isolates of R. lauricola from Myanmar were distinct from those from Japan, Taiwan and the USA. Although both MAT idiotypes were present in Myanmar and Taiwan, only the population from Taiwan had the genetic structure of a sexually reproducing population.


Introduction
Laurel wilt is a systemic vascular wilt of trees in the Lauraceae plant family.It is caused by the fungus Raffaelea lauricola T.C. Harr., Fraedrich and Aghayeva (Ophiostomatales), which is cultivated and transmitted by its symbiont, the redbay ambrosia beetle, Xyleborus glabratus Eichhoff (Curculionidae: Scolytinae) [1,2].Xyleborus glabratus was first detected in the United States at Port Wentworth, GA in 2002 [3], and soon after laurel wilt began to kill Persea borbonia (L.) Spreng., a native tree in the Lauraceae [1].To date, 14 species in the family are known hosts, most of which are American endemics [4].
Laurel wilt has spread throughout much of the southeastern USA and devastated forest and agricultural landscapes.It has killed an estimated 330 million redbay trees [5], and now affects sassafras (Sassafras albidum (Nutt.)Nees) outside of the range of redbay [6,7].The potential for laurel wilt to continue to spread where sassafras is the predominant host is under investigation [8].As it spreads, laurel wilt will have an unclear effect on the large reservoir of native Lauraceae that exists in tropical America.In addition, the disease affects commercial avocado (Persea americana Miller) production in Florida, and threatens larger producers of this important crop in California and Mexico [9,10].
Xyleborus glabratus has been reported in China, India, Bangladesh, Japan, Taiwan and Myanmar, where the beetle is associated with plants in the Lauraceae [3,11].As the beetle's primary symbiont [1,2], R. lauricola is probably found wherever X. glabratus occurs, presumably throughout southern Asia.Previously, Harrington et al. [12] regularly isolated the fungus from individuals of X. glabratus from Japan and Taiwan.
Although the pathogen and its vector occur in Asia, significant disease is not known on native members of the Lauraceae in the region [10].Hulcr et al. [13] recovered the beetle and fungus from injured Machilus zuihoensis Hayata trees, but symptoms were localized to small stems near areas of X. glabratus attack and did not spread throughout the crown.It is hypothesized that the susceptibility of American hosts is due to a lack of coevolved resistance, which apparently occurs in Asian members of the family [10,14].For example, the Asian camphortree (Cinnamomum camphora (L.)J. Presl.)tolerates R. lauricola infection in the USA [15], as do other Asian taxa that have been tested experimentally (Smith et al., unpublished data).Outside the USA, laurel wilt has only been reported on cultivated avocado in the Tuanggyi and Ywangan Districts of Myanmar [16].Since X. glabratus is native to Myanmar [3] and R. lauricola is closely associated with its beetle symbiont, we assume that R. lauricola is also native to Myanmar.
Shortly after laurel wilt was described in the USA, it was suggested that X. glabratus and R. lauricola were each introduced into the country a single time [12].Recently, this hypothesis was investigated with genetic markers in R. lauricola.Wuest et al. [17] used two simple sequence repeat (SSR), a single nucleotide polymorphism (SNP) and mating type markers with isolates from the USA, Japan and Taiwan, whereas Hughes et al. [5] used amplified fragment length polymorphism (AFLP) and SSR markers to examine isolates recovered early in the USA epidemic.Although the studies used a limited number of markers or isolates, Hughes et al. [5] and Wuest et al. [17] suggested that a single introduction of the pathogen had occurred in the USA.
The present study had several objectives.To robustly test the single introduction hypothesis, we developed a multiplexed panel of SSR markers based on those developed by Dreaden et al. [18] and used by others [5,17,19].Isolates of R. lauricola from the USA and the only three countries in Asia from which isolates of the pathogen were available (Myanmar, Japan and Taiwan) were then assayed with SSR and mating type data (Table 1).In total, isolates from seven species of host tree and eight species of ambrosia beetles, seven of which acquired the fungus via lateral transfer [10,19], were represented.These data were used to assess clonality in the USA population of R. lauricola and whether a source of the putative clonal population in the USA was evident in Asia.In addition, we examined the hypothesis that R. lauricola reproduces sexually.

Isolates of Raffaelea lauricola
Isolates of R. lauricola were recovered by the authors or obtained from collaborators and culture collections (Table S1).A total of 59 isolates were examined from seven species of host tree, eight ambrosia beetle species, four countries (Myanmar, Japan, Taiwan and the USA) and five states in the USA.
The ability to consistently genotype isolates was calculated with seven randomly selected samples.Original DNA extractions were re-amplified and analyzed with four biological replicates of subcultures of the original isolates.Technical and biological replicates along with all samples that were analyzed multiple times were analyzed to calculate genotyping error rate, which is reported as the number of incorrect alleles per number of repeated reactions (Table 1).All samples/loci that produced null alleles were amplified in singleplex PCRs and reanalyzed to confirm the presence of a null allele.Primer sequences, repeat motifs and amplicon sizes in isolate PL716 (GQ996063) are given in supplemental Table S2.

Statistical Analyses
Poppr [23,24] was used to analyze the microsatellite data, including null alleles.A genotype accumulation curve was constructed using 1000 random samples of n loci.Analyses of Molecular Variance (AMOVA) were determined among populations, and calculated with and without clone correction using Euclidean distances [25].Significance of the variance component was tested with 1000 nonparametric permutations.
Discriminant Analysis of Principal Components (DAPC) was conducted with cross-validation to determine the appropriate number of principal components, as described by Jombart and Collins [26].A Minimum Spanning Network (MSN) was constructed using Bruvo's distance [27], and K-means hierarchical clustering was conducted, per Grunwald et al. [28].An index of association (I A ) was calculated for populations from Taiwan and Myanmar using 999 permutations [29].

Mating-Type Assay
All isolates produced an amplicon for either the MAT1-1-3 or MAT1-2-1 genes, which were then classified as MAT1 or MAT2, respectively.The MAT2 idiomorph was found in 15 of 18 (83%) of the isolates from Myanmar, three of six (50%) from Taiwan, and all 34 isolates from the USA, whereas MAT1 was found in three isolates each from Mynamar and Taiwan and the single Japanese isolate (Figure 1, Table S1).

Microsatellite Markers
Of 18 SSR loci that were examined, six were not analyzed due either to difficulty in scoring (X21, 9V8) or high genotyping error rates (OCT, 46Z, QI5, REH) (Table 1).Null alleles that were found at four of the 12 remaining loci (0DS, ZWC, F81, KTR) were confirmed in singleplex PCRs and used in the population analyses.At each locus, a mean of four alleles (range = two-six) were detected (Table 1).A genotype accumulation curve indicated that the SSR panel was sufficient to distinguish clonal lineages in the examined populations (Figure 2).

163
The diversity of multilocus SSR genotypes (MLGs) was far greater in Asia than in the USA 164 (Figure 1).A total of 15 MLGs were detected in the collection we studied, only three of which were 165 found in the USA; 30 of the 34 USA isolates had MLG5, which was also detected in Taiwan (Figure 1, 166 Table S1).Three USA isolates had MLG3, and another had MLG4 (Figure 1, Table S1); each of these 167 less common MLGs shared an IFW allele that was not found in MLG5 or the Asian isolates.MLG4 168 also had a unique MNT allele that was not found in other USA isolates, but was present in four of six 169 MLGs from Myanmar.With the exception of two SSR loci that had two alleles each, all other loci in 170 the USA collection were monomorphic.2).Thus, there was evidence for differentiation, and the null hypothesis of random 174 mating between populations could be rejected.The diversity of multilocus SSR genotypes (MLGs) was far greater in Asia than in the USA (Figure 1).A total of 15 MLGs were detected in the collection we studied, only three of which were found in the USA; 30 of the 34 USA isolates had MLG5, which was also detected in Taiwan (Figure 1, Table S1).Three USA isolates had MLG3, and another had MLG4 (Figure 1, Table S1); each of these less common MLGs shared an IFW allele that was not found in MLG5 or the Asian isolates.MLG4 also had a unique MNT allele that was not found in other USA isolates, but was present in four of six MLGs from Myanmar.With the exception of two SSR loci that had two alleles each, all other loci in the USA collection were monomorphic.

Statistical Analyses
In the AMOVA, most variance occurred among populations (p = 0.001), with and without clone correction (Table 2).Thus, there was evidence for differentiation, and the null hypothesis of random mating between populations could be rejected.Discriminant Analysis of Principal Components utilized four Principal Component Analysis axes and three Discriminant Analyses.Although the Bayesian information criteria (BIC) supported three genetic groups, the scatter plot clearly separated the samples into two main groups, the first of which contained the isolates from Myanmar and the second of which included all other isolates (Figure 3).Minimum Spanning Network (MSN) separated the isolates from Myanmar from the remaining isolates, which in turn were separated into two groups (Figure 1).K-means hierarchical clustering also detected the three groups identified by the MSN (Figure 1).remaining isolates, which in turn were separated into two groups (Figure 1).K-means hierarchical 183 clustering also detected the three groups identified by the MSN (Figure 1).The index of association for the Myanmar population was p = 0.001 for rd = 0.752 without clone correction and p = 0.001 for rd = 0.652 with correction.Thus, the null hypothesis of no linkage among markers could be rejected indicating that this population is mostly clonal.In contrast, the index of association for the Taiwan population was p = 0.566 for rd = 0.008 without clone correction and p = 0.532 for rd = 0.008 with correction.Thus, the null hypothesis could not be rejected, which indicated a sexually reproducing population.

Discussion
The AMOVA found evidence of differentiation among the Asian and USA populations.DAPC, K-means hierarchical clustering, and MSN were mostly congruent, and indicated that the population from Myanmar differed from those from Japan, Taiwan and the USA.Despite the high level of diversity, including the recovery of three MLGs from a single tree and the presence of both mating types, the index of association indicated that the pathogen in Myanmar had the structure of a clonally reproducing population.We recovered MLG14 (Table S1) from the three locations that were sampled in Myanmar, which provided further support for clonal reproduction in this population.In contrast, the index of association and presence of both mating types in Taiwan provided indirect evidence for sexual reproduction of the pathogen in that country.Wuest et al. [17] had suggested that R. lauricola could reproduce sexually in parts of its native range.Nevertheless, direct evidence for sexual reproductive in R. lauricola is lacking.
We detected limited genetic diversity and a single mating type in a collection of 34 isolates of R. lauricola from diverse geographic sources and ambrosia beetles in the USA.Thus, as suggested by Wuest et al. [17] and Hughes et al. [5], we conclude that the population of R. lauricola in the USA has gone through a genetic bottleneck.Unlike Wuest et al. [17], who found no geographic structure in their collection, we detected geographic clustering, mainly among isolates from Myanmar (Figures 1 and 3).Although Wuest et al. [17] did not examine isolates from Myanmar, it is also possible that the markers they used played a role in this discrepancy.
Since isolates of R. lauricola from ambrosia beetles other than X. glabratus were all of MLG5 it did not appear that beetle species influenced MLG in the USA.Likewise, as found previously [5,17] isolates from different host trees in the USA were also genetically uniform.Although scant MLG diversity did not enable a good test of this relationship in the USA.We note that three MLGs were detected in a single tree of avocado in Myanmar (Table S1) and five MLGs were detected in a single tree of Cinnamomum osmophloeum Kaneh. in Taiwan [17].This MLG diversity from a single tree is surprising as laurel wilt is generally considered a systemic disease in highly susceptible host like avocado [10].
Wuest et al. [17] and Hughes et al. [5] identified polymorphic isolates of R. lauricola from southeast Georgia and northeast Florida.Wuest et al. [17] identified isolates with an alternate IFW allele that was not found elsewhere.Three polymorphic isolates identified by Hughes et al. [5] with AFLPs were included in the present study, all of which were members of MLG5, the common SSR MLG in the USA.Although we did not have isolates used by Wuest et al. [17], the polymorphic isolates in the present study were recovered at the same times and from similar areas that they recovered the rare IFW alleles.
Minor differences were evident among the three MLGs that were detected in the USA.For example, isolates in MLG3 shared all its SSR alleles with MLG5, with the exception of one uncommon IFW allele.If a mutation at the IFW locus in MLG5 resulted in MLG3, a second mutation at the MNT locus in MLG3 could have resulted in MLG4.Two mutations in the predominant MLG5 would be consistent with the single introduction hypothesis for R. lauricola in the USA [5,17].Since the prevalent MLG5 in the USA was also detected in Taiwan and no other MLG in the USA was detected in Asia, it is possible that the relatively homogeneous population of R. lauricola in the USA originated from a single introduction from Taiwan.Additional isolates from Asia would be needed to test this hypothesis.
Our ability to determine the origins and relatedness of populations of R. lauricola is limited by the small number of isolates that are available from the native range of the pathogen and its beetle symbiont, X. glabratus.As more isolates from more diverse geographic areas become available the multiplex microsatellite panel we developed could help elucidate R. lauricola's genetic structure and possibly the source of the USA introduction.Clearly, better understandings of the pathogen's global population structure could clarify unknown aspects of its biology, pathology and epidemiology.

Conclusions
An invasive fungus from Asia, R. lauricola, causes Laurel wilt, a lethal vascular disease of plants in the Lauraceae plant family.The hypothesis that the fungus was introduced into the USA a single time was corroborated with 12 simple sequence repeat (SSR) markers and a diverse collection of the pathogen from Asia (Myanmar, 18 isolates, Taiwan, 6 isolates, and Japan, 1 isolate) and the USA (34 isolates).The results indicate that a Taiwanese origin is possible for the population of R. lauricola in the USA.Sexual reproduction in the pathogen, which was statistically supported only in the population in Taiwan, needs to be confirmed empirically.

Figure 2 .
Figure 2. Genotype accumulation curve for isolates of R. lauricola from the USA and Asia.The y axis

178
axes and three Discriminant Analyses.Although the Bayesian information criteria (BIC) supported 179 three genetic groups, the scatter plot clearly separated the samples into two main groups, the first of 180 which contained the isolates from Myanmar and the second of which included all other isolates 181 (Figure3).Minimum Spanning Network (MSN) separated the isolates from Myanmar from the

Figure 2 .
Figure 2.Genotype accumulation curve for isolates of R. lauricola from the USA and Asia.The y axis gives the number of multilocus genotypes (MLGs), the x axis the number of SSR loci that were assayed (12 total), and the dashed red line represents 100% MLG resolution.

Figure 3 .
Figure 3. Discriminant Analysis of Principal Components (DAPC) of R. lauricola sampled from United

195 4 .Figure 3 .
Figure 3. Discriminant Analysis of Principal Components (DAPC) of R. lauricola sampled from United States and Asian populations using four Principal Component Analysis axes and three Discriminant Analysis separates isolates from Myanmar from the rest of the isolates.Upper left inset, the Bayesian information criteria (BIC) supported three genetic groups.

Table 1 .
Genotyping statistics for microsatellite/simple sequence repeat (SSR) loci in isolates of Raffaelea lauricola from Asia and the USA.

Multiplex PCR SSR Loci x Error Rate (Incorrect Calls/Total Calls) Alleles
Minimum spanning network for SSR genotypes of Raffaelea lauricola from Asia and the USA.Minimum spanning network for SSR genotypes of Raffaelea lauricola from Asia and the USA.Node color indicates source country of a genotype, the relative node size indicates the number of clones that are represented by a given genotype, and the thickness of lines between nodes represents Bruvo's (27) genetic distance, wherein thicker lines denote smaller genetic distances.Isolate numbers in blue are MAT1 and black are MAT2.Three groups detected by K-means hierarchical clustering are in separate boxes, and multilocus genotypes (MLG) in the USA are indicated.
147 Node color indicates source country of a genotype, the relative node size indicates the number of 148 clones that are represented by a given genotype, and the thickness of lines between nodes represents 149 Bruvo's (27) genetic distance, wherein thicker lines denote smaller genetic distances.Isolate numbers 150 in blue are MAT1 and black are MAT2.Three groups detected by K-means hierarchical clustering are 151 in separate boxes, and multilocus genotypes (MLG) in the USA are indicated.thepopulation analyses.At each locus, a mean of four alleles (range = two-six) were detected (TableFigure 1.

Table 2 .
Analysis of molecular variance (AMOVA) using twelve microsatellite loci from Raffaelea 176 lauricola from Myanmar, Japan, Taiwan, and the United States.
Discriminant Analysis of Principal Components utilized four Principal Component Analysis

Table 2 .
Analysis of molecular variance (AMOVA) using twelve microsatellite loci from Raffaelea lauricola from Myanmar, Japan, Taiwan, and the United States.