Abundant Genetic Diversity and Extensive Differentiation among Geographic Populations of the Citrus Pathogen Diaporthe citri in Southern China

The fungal pathogen Diaporthe citri is a major cause of diseases in citrus. One common disease is melanose, responsible for large economic losses to the citrus fruit industry. However, very little is known about the epidemiology and genetic structure of D. citri. In this study, we analyzed 339 isolates from leaves and fruits with melanose symptoms from five provinces in southern China at 14 polymorphic simple sequence repeat (SSR) loci and the mating type idiomorphs. The genetic variations were analyzed at three levels with separate samples: among provinces, among orchards within one county, and among trees within one orchard. The five provincial populations from Fujian, Zhejiang, Jiangxi, Hunan, and Guizhou were significantly differentiated, while limited differences were found among orchards from the same county or among trees from the same orchard. STRUCTURE analysis detected two genetic clusters in the total sample, with different provincial subpopulations showing different frequencies of isolates in these two clusters. Mantel analysis showed significant positive correlation between genetic and geographic distances, consistent with geographic separation as a significant barrier to gene flow in D. citri in China. High levels of genetic diversity were found within individual subpopulations at all three spatial scales of analyses. Interestingly, most subpopulations at all three spatial scales had the two mating types in similar frequencies and with alleles at the 14 SSR loci not significantly different from linkage equilibrium. Indeed, strains with different mating types and different multilocus genotypes were frequently isolated from the same leaves and fruits. The results indicate that sexual reproduction plays an important role in natural populations of D. citri in southern China and that its ascospores likely represent an important contributor to citrus disease.


Introduction
Diaporthe citri (anamorph Phomopis citri) is one of the most important pathogenic fungi on citrus. It induces melanose on the leaves, twigs, and fruits; stem-end rot of postharvest fruits; and gummosis and blight of shoots, perennial branches, and trunks of citrus. Among these diseases, melanose brings perhaps the most significant economic loss since it significantly reduces the marketability of citrus fruits [1][2][3][4][5]. This fungal pathogen is widespread in humid citrus-producing areas, except for California in the US and Europe [5,6]. All citrus species are susceptible to D. citri infection, with grapefruit and lemons being the most severely affected [5]. the same tree and the same orchard. Furthermore, based on the observed broad host range of D. citri [10][11][12], we hypothesize that there will be limited host tree-based and host organ-based population structure of this fungal pathogen. To test the third hypothesis, melanose-diseased samples from two host tree species from within the same county were analyzed. Similarly, melanose-diseased leaves and fruits from the same trees within a single orchard were sampled and analyzed. Together, these hypotheses were tested using a variety of samples and statistical and population genetic methods.

Isolate Collection
In total, 339 single-conidial isolates of D. citri were collected between June and September in 2019 from Citrus reticulata (Ponkan) and C. sinensis (navel orange) in 21 orchards in five provinces of southern China (Table 1). These samples were collected to address strain relationships and population genetic structures at three spatial levels, as described below. The first level was the finest scale where we sampled extensively among five trees (the trees were at least 10 m from each other) in one orchard in Luxi county, Hunan Province. For each tree, five leaves and five fruits with melanose symptoms were sampled and 10 melanose spots per leaf/fruit were individually treated for isolation of D. citri. All host trees in this orchard were C. reticulata and a total of 95 isolates were obtained for study.
The second spatial level analyzed D. citri samples from four orchards within Luxi county in Hunan Province. Two of the four orchards had C. reticulata while the other two were planted with C. sinensis. Here, four leaves or fruits with melanose symptoms were randomly collected from each of 10 random trees (at least 10 m apart) in each orchard. A total of 102 isolates were obtained from these four orchards. At this level, we analyzed the potential effects of intermediate spatial separation and host tree species on the total genetic variations.
The third level is the largest spatial scale at the provincial level, where 142 isolates were collected from 18 orchards of C. reticulata located in five provinces. These included 24 isolates from four orchards in Fujian (FJ) Province, 30 isolates from three orchards in Guizhou (GZ) Province, 30 isolates from four orchards in Jiangxi (JX) Province, 28 isolates from five orchards in Zhejiang (ZJ) Province, and 30 isolates from two orchards in Hunan (HN) Province. Similar to the two smaller scales, each isolate was derived from an independent melanose spot on either leaves or fruits. Here, the isolates were derived from at least five trees in each orchard. For each tree, 1-4 leaves or fruits with melanose symptoms were randomly sampled for isolation of D. citri. However, to reduce the potential overrepresentation of isolates from certain trees, only one isolate from each tree was randomly chosen for genotyping and subsequent data analyses.
To minimize confounding factors, the three levels of analyses used separate samples. Strains of D. citri were isolated following the method described previously [2]. Cultures were stored at 4 • C. For long-term storage, the isolates were kept in 30% glycerin in a −80 • C freezer.

DNA Extraction
For each isolate, its genomic DNA was extracted from mycelia scraped off with a sterile blade from a colony growing on PDA at 25 • C for about 7 days. The scraped mycelia were lyophilized, and then the DNA was extracted using a SIMGEN Fungi Kit (Simgen, Hangzhou, China) according to the manufacturer's protocol. The extracted genomic DNA samples were quantified using a Nanodrop spectrophotometer (Nanodrop Products, Wilmington, DE, USA) and stored in sterile water in a −20 • C freezer.

Identification of D. citri-like Isolates
The obtained D. citri-like isolates were confirmed based on their colony morphology, microscopic conidia features, and molecular characteristics [2]. Specifically, we used D. citrispecific primer pair Dc-F/Dc-R (5 -CCCTCGAGGCATCATTAC-3 /5 -ATGTTGCAGATG GTCAAATGG-3 ) designed based on sequence variations in the Tub gene among 21 Diaporthe species from citrus [2]. Each PCR reaction was conducted in a final volume of 20 µL, containing 1 µL of template genomic DNA (~20 ng), 0.8 µL forward primer, 0.8 µL reverse primer, 10 µL of 2 × Taq Plus Master Mix (Vazyme Biotechnology Co. Hangzhou), and 7.4 µL sterile water. PCR cycling condition was performed as follows: 94 • C for 3 min, 31 cycles of 94 • C for 30 s, 60 • C for 30 s, 72 • C for 1 min, with a final extension of 3 min at 72 • C. The PCR products were analyzed by electrophoresis at 120 V for 30 min in a 1% (w/v) agarose gel and visualized under UV light after ethidium bromide staining. Genomic DNAs from two strains (ZJUD2 and ZJUD14), which were identified as D. citri [2], were used as positive controls. The genomic DNAs of Diaporthe citriasiana strain ZJUD31 and D. citrichinensis strain ZJUD40 [2] were used as negative controls. These four reference isolates are stored at the Institute of Biotechnology at Zhejiang University (Hangzhou, Zhejiang Province, China).

SSR Marker Development and Screening
The genome sequence of D. citri strain ZJUD2 [22] was screened using the online tool Batchprimer 3 (https://wheat.pw.usda.gov/demos/BatchPrimer3/, accessed on 5 November 2020) [23] to identify potential SSR markers. SSRs containing 2 to 54 tandem repeats of di-to hexa-nucleotides as well as their flanking regions were retrieved. Primers were designed using Batchprimer 3 following the default setting to obtain amplification products in the range of 100 bp to 250 bp to allow optimum genotyping and scoring [24]. The selected primers were screened for levels of polymorphism and amplification success using a subset of 12 isolates of D. citri (ZJUD2, ZJUD6, ZJUD14, ZJUD26, ZJUD13, ZJUD24, ZJUD22, ZJUD28, ZJUD19, ZJUD21, ZJUD15, and ZJUD8) [2], which were kept at the Institute of Biotechnology at Zhejiang University. The PCR cycling conditions used the following touch-down protocol: 94 • C for 5 min, 10 cycles of 94 • C for 30 s, 60 • C for 30 s, with a decrease in annealing temperature of 1 • C in each cycle, and 72 • C for 30 s, followed by 25 cycles of 94 • C for 30 s, 52 • C for 30 s, and 72 • C for 30 s, with a final extension of 10 min at 72 • C. The PCR products were separated on polyacrylamide gels and stained with ethidium bromide before visualization under UV light. Loci were considered polymorphic if two or more alleles were observed among the 14 isolates evaluated.
Based on amplification success and the level of polymorphisms, we selected 14 markers to genotype the subset of 14 isolates from our recent collection. The optimum annealing temperature of the selected 14 pairs of primers was derived by gradient PCR using the genome DNA of ZJUD2. The PCR reaction system was the same as described above. The PCR cycling conditions were: 94 • C for 5 min, 31 cycles of 94 • C for 30 s, 50~62 • C for 30 s, 72 • C for 30 s, with a final extension of 5 min at 72 • C. The PCR products were separated and visualized as described above for confirmation of amplification success and polymorphisms among the 14 representative strains. All 14 primer pairs successfully amplified a single product from each of the 14 representative strains. These 14 primer pairs were then used to genotype all 339 isolates following protocols described below.

SSR Loci Genotyping
For each of the 14 SSR markers, the forward primer was labeled with either FAM, HEX, or TAMRA fluorophore. The PCR reaction system was the same as described above. PCR cycling was performed as follows: 94 • C for 5 min, 35 cycles of 94 • C for 30 s, 53, 55, or 56 • C (depending on the primer pair) for 30 s, 72 • C for 30 s, with a final extension of 10 min at 72 • C. The amplicons were separated using capillary electrophoresis. Capillary electrophoresis was performed in multiplex reactions of three amplicons with three different fluorescent dyes per reaction. The PCR products were analyzed using an ABI 3730xl Genetic Analyzer (Applied Biosystems, Foster City, CA, USA) with a GS500LIZ Size Standard (Applied Biosystems). The output files were analyzed using GeneMapper v4.1 software (Applied Biosystems) to identify the lengths of amplified fragments at each locus.

Diversity Characteristics of the 14 SSR Markers
For most diversity estimates, two sets of samples were analyzed: non-clone-corrected (i.e., the raw sample) and clone-corrected. For clone-corrected datasets, only a single isolate of each multilocus genotype (MLG) was retained from each sub-population. Isolates with the same alleles at all loci were considered as belonging to the same clone. Clone-corrected datasets were used to minimize potentially distorted estimates of allelic diversity due to differences in sampling efforts and in D. citri isolation successes [25].
To determine gene diversities of the subpopulations and populations at different spatial scales, the following diversity estimates were calculated using the program GenAlEx v. 6.5.2 [26]: number of alleles (Na), number of private alleles, effective number of alleles, Shannon's polymorphism information index [27], Nei's gene diversity [28], and unbiased gene diversity [29].
To determine the genotypic diversities in subpopulations and populations at each of the three spatial scales, the number of MLGs was determined in R v. 3.6.3 (R Core Team 2013) [30] using the analysis package poppr [31]. Genetic and genotypic diversities were not calculated for samples from individual trees and orchards due to their small numbers of isolates. To assess the power of our SSR markers for discriminating between strains and multilocus genotypes in the dataset, a rarefaction curve was generated using poppr [31].

Genetic Structure Analysis of D. citri Populations
Population genetic structure was analyzed using several approaches. In the first, we analyzed the sources of genetic variation at different spatial scales using GenAlEx version 6.501 [26]. The relationships among geographic subpopulations were investigated based on clone-corrected data using hierarchical analysis of molecular variance (AMOVA) at different scales, using 999 permutations. The PhiPT values, representing a 0 to 1 scaled estimator of population differentiation [32], were calculated with clone-corrected data for all pairwise comparisons for subpopulations from different provinces using 999 permutations. Principal Component Analysis (PCA) was also performed to highlight the genetic similarities and differences among the provincial subpopulations. In the second analysis, we assessed the relationships among the MLGs in the subpopulations from Fujian, Guizhou, Jiangxi, Zhejiang, and Hunan provinces, using a minimum spanning network approach to analyze the non-clone-corrected data based on Bruvo's distance [33], using the R package poppr.
In the third approach, we used the program STRUCTURE v. 2.3.4 [34] to infer the number of genetic clusters in the total sample, as well as to estimate the extent to which genetic admixture might have occurred among the subpopulations from the five provinces. The STRUCTURE analyses used the admixture model, with 10 replicated runs of K = 1 to 10 after an initial burn-in of 100,000 generations followed by a run length of 1,000,000 generations. The optimal number of genetic clusters (K) was identified by following the method described in [35], using StructureHarvester (http://taylor0.biology.ucla. edu/structureHarvester, accessed on 1 June 2021).
At the two smaller spatial scales, we also used GenAlEx to identify the potential influences of two host species (C. reticulata vs. C. sinensis among geographically closely located orchards within Luxi county in Hunan) and two host organs (leaves vs. fruits of the same five trees within one orchard) on the observed genetic variations, respectively.

Mating Type Ratios
The ratios of the MAT1-1-1 and MAT1-2-1 mating types among the isolates at each subpopulation and in the total population were calculated based on both the non-clonecorrected and clone-corrected datasets. A χ 2 test with a critical value of p < 0.05 was performed to determine whether the mating type frequencies deviated significantly from a 1:1 ratio [37]. The null hypothesis was that the ratio of the two mating types was 1:1 (in equilibrium). Due to issues with using a χ 2 test when analyzing small sample sizes, an exact binomial goodness-of-fit test (two-tailed) was also performed to assess whether the observed mating type ratios for analyzed populations deviated from 1:1. The χ 2 test and the exact binomial test (two-tailed) were performed using IBM SPSS Statistics 20 (International Business Machines Corporation, Arkmonk, NY, USA). Occurrence of the two mating types in equilibrium would be an indication of sexual reproduction and recombination.

Analysis of Linkage Disequilibrium and Recombination
Linkage disequilibrium was analyzed by calculating the index of association (I A ) [38] and the standardized index of association (r d ) [39] after 999 permutations in package poppr [31]. The null hypothesis in this test was that frequent recombination occurred and would be seen as random associations among alleles among the 14 different loci. Significant deviations from random associations would result in rejection of the null hypothesis and support the alternative hypothesis of linkage, indicating infrequent or no recombination in nature. In addition, we also tested whether the null hypothesis of no recombination would be supported. Here, the proportion of phylogenetically compatible pairs of loci (Prp) was calculated using Multilocus 1.3b with 1000 permutations. Two loci are phylogenetically compatible (Prp = 1) if they arise from the same phylogenetic position in the absence of homoplasy or recombination. Under the assumption that parallels, reversals, or convergences are rare, phylogenetic incompatibility (Prp < 1) provides the evidence of sexual recombination and exchange of genetic material between genomes [40,41]. The hypothesis of random mating would be rejected if fewer incompatible pairs of loci were observed than expected in a randomized recombining population (p < 0.05) [39].

SSR Marker Development and Screening
The fourteen primer pairs we screened showed polymorphisms in the subset of 14 isolates (data not shown). The BLAST analyses of the 14 SSRs using the assembled D. citri genome showed that these 14 SSR loci were all located on different contigs and not closely linked. The details about these 14 loci are presented in Table 2.

Diversity Characteristics of the 14 SSR Markers in the D. citri Population
At the largest spatial scale that includes the five provincial populations, all 14 loci were polymorphic, and the number of alleles per locus among the 142 isolates ranged from 4 to 20, Nei's gene diversity ranged from 0.40 to 0.89, Simpson index ranged from 0.39 to 0.89, and Evenness ranged from 0.42 to 0.86. At the next spatial scale among the four orchards used for analyzing potential local population structuring and host tree effects on pathogen gene diversity, 13 of the 14 loci were polymorphic, while the DC1 locus was monomorphic among the 102 isolates. The number of amplified alleles, Simpson index, Nei's gene diversity, and Evenness per locus among the 102 isolates ranged from 2 to 15, 0.04 to 0.78, 0.04 to 0.79, and 0.35 to 0.88 in the orchard population, respectively (Table 3). Similarly, 13 of the 14 loci were polymorphic among the 95 isolates from the five trees in the same orchard, with locus DC1 being monomorphic. The number of amplified alleles, Simpson index, Nei's gene diversity, and Evenness per locus ranged from 2 to 13, 0.11 to 0.80, 0.11 to 0.81, and 0.39 to 0.89 in the tree population, respectively ( Table 3). The genotype rarefaction curve showed increases in the number of genotypes as the number of the loci increased at all three examined spatial scales and reached a plateau at 13 loci at the provincial level, and at 12 loci at the orchard and tree levels ( Figure 1). Together, these results indicate that the 14 SSR markers were sufficient to discriminate among strains at the three spatial scales.   The above results on gene diversities described the variations among the 14 loci at different spatial scales. When all loci are considered, individual populations at all three spatial scales also showed high diversities. For example, the numbers of different and effective alleles per locus ranged from 2.957 (tree level) to 4.104 (provincial level) and 1.920 (orchard level) to 2.466 (provincial level); the Shannon's information indices were 0.908, 0.684, and 0.681; and the unbiased Nei's diversities were 0.472, 0.348, and 0.363 for populations at provincial, orchard, and tree scales, respectively. The values of unbiased allelic diversities were overall similar to Nei's diversity (Table 4).
Among the five provincial subpopulations, that from Fujian Province showed the highest gene diversity, with Nei's gene diversity and Shannon's information index being 0.638 and 1.252, respectively, followed by the Zhejiang subpopulation with Nei's gene diversity and Shannon's information index at 0.569 and 1.132, respectively. The levels of gene diversity of the Hunan, Jiangxi, and Guizhou subpopulations were similar, with Nei's gene diversity and Shannon's information index ranging from 0.403 to 0.424 and 0.839 to 0.874, respectively (Table 4).
Similar to gene diversity analyses, results from the genotype diversity analysis also revealed high genotypic diversities within all examined populations, with 139 (98%), 99 (97%), and 87 (92%) MLGs identified among 142, 102, and 95 fungal isolates analyzed The above results on gene diversities described the variations among the 14 loci at different spatial scales. When all loci are considered, individual populations at all three spatial scales also showed high diversities. For example, the numbers of different and effective alleles per locus ranged from 2.957 (tree level) to 4.104 (provincial level) and 1.920 (orchard level) to 2.466 (provincial level); the Shannon's information indices were 0.908, 0.684, and 0.681; and the unbiased Nei's diversities were 0.472, 0.348, and 0.363 for populations at provincial, orchard, and tree scales, respectively. The values of unbiased allelic diversities were overall similar to Nei's diversity (Table 4). Among the five provincial subpopulations, that from Fujian Province showed the highest gene diversity, with Nei's gene diversity and Shannon's information index being 0.638 and 1.252, respectively, followed by the Zhejiang subpopulation with Nei's gene diversity and Shannon's information index at 0.569 and 1.132, respectively. The levels of gene diversity of the Hunan, Jiangxi, and Guizhou subpopulations were similar, with Nei's gene diversity and Shannon's information index ranging from 0.403 to 0.424 and 0.839 to 0.874, respectively (Table 4).
Similar to gene diversity analyses, results from the genotype diversity analysis also revealed high genotypic diversities within all examined populations, with 139 (98%), 99 (97%), and 87 (92%) MLGs identified among 142, 102, and 95 fungal isolates analyzed from the provincial, orchard, and tree levels, respectively (Table 4). At the provincial level, three of the 139 MLGs were shared, with each shared MLG found between two isolates from different trees in the same orchard. One shared MLG each was found in Guizhou, Hunan, and Zhejiang provinces (Table 4). Similarly, at the orchard level, three of the 99 MLGs were shared, with each shared MLG found between two isolates from different trees in the same orchard. Two of the shared MLGs were in orchard 1 and the third shared MLG was found in orchard 2 (O2 in Table 4). Interestingly, among the five trees within the same orchard, there was no shared MLG in our sample-all the shared MLGs were found among isolates either from different leaves/fruits on the same tree or from different melanose spots on the same leaf and fruit (Figure 2).

Genetic Structure among Populations of D. citri
AMOVA analysis showed that 31% of the genetic variation was distributed among the subpopulations from the five provinces, and 69% was distributed within the subpopulations. At the orchard level, 10% of the total genetic variations were distributed among the four orchards, while 90% were found within the orchards, with less than 2% coming from between the two citrus species (p > 0.05), consistent with limited host tree specialization in D. citri genotypes. At the tree level within the orchard in Luxi, 1% of the total genetic variations were found among the subpopulations from the five trees, while 99% were found within the individual tree subpopulations (Figure 3). Furthermore, there was no evidence of genetic differentiation between D. citri populations from the two different organ types (leaf vs. fruit) (Figure 4).

Genetic Structure among Populations of D. citri
AMOVA analysis showed that 31% of the genetic variation was distributed among the subpopulations from the five provinces, and 69% was distributed within the subpopulations. At the orchard level, 10% of the total genetic variations were distributed among the four orchards, while 90% were found within the orchards, with less than 2% coming from between the two citrus species (p > 0.05), consistent with limited host tree specialization in D. citri genotypes. At the tree level within the orchard in Luxi, 1% of the total genetic variations were found among the subpopulations from the five trees, while 99% were found within the individual tree subpopulations (Figure 3). Furthermore, there was no evidence of genetic differentiation between D. citri populations from the two different organ types (leaf vs. fruit) (Figure 4).

Genetic Structure among Populations of D. citri
AMOVA analysis showed that 31% of the genetic variation was distributed among the subpopulations from the five provinces, and 69% was distributed within the subpopulations. At the orchard level, 10% of the total genetic variations were distributed among the four orchards, while 90% were found within the orchards, with less than 2% coming from between the two citrus species (p > 0.05), consistent with limited host tree specialization in D. citri genotypes. At the tree level within the orchard in Luxi, 1% of the total genetic variations were found among the subpopulations from the five trees, while 99% were found within the individual tree subpopulations (Figure 3). Furthermore, there was no evidence of genetic differentiation between D. citri populations from the two different organ types (leaf vs. fruit) (Figure 4).

Genetic Structure among Populations of D. citri
AMOVA analysis showed that 31% of the genetic variation was distributed among the subpopulations from the five provinces, and 69% was distributed within the subpopulations. At the orchard level, 10% of the total genetic variations were distributed among the four orchards, while 90% were found within the orchards, with less than 2% coming from between the two citrus species (p > 0.05), consistent with limited host tree specialization in D. citri genotypes. At the tree level within the orchard in Luxi, 1% of the total genetic variations were found among the subpopulations from the five trees, while 99% were found within the individual tree subpopulations (Figure 3). Furthermore, there was no evidence of genetic differentiation between D. citri populations from the two different organ types (leaf vs. fruit) (Figure 4).
Pairwise PhiPT values showed significant genetic differentiation between all provincial subpopulations (p = 0.001), with the PhiPT values ranging from 0.104 (HN vs. GZ) to 0.421 (FJ vs. JX) ( Table 5). PCA analysis revealed that the subpopulations from five provinces were divided into two genetic groups. One group included subpopulations from Hunan, Guizhou, and Jiangxi, while the other group included subpopulations from Fujian and Zhejiang ( Figure 5). The minimum spanning network revealed that the MLGs from Fujian and Zhejiang are distinct from the Hunan, Guizhou, and Jiangxi subpopulations ( Figure 6). among the four orchards, while 90% were found within the orchards, with less than 2% coming from between the two citrus species (p > 0.05), consistent with limited host tree specialization in D. citri genotypes. At the tree level within the orchard in Luxi, 1% of the total genetic variations were found among the subpopulations from the five trees, while 99% were found within the individual tree subpopulations (Figure 3). Furthermore, there was no evidence of genetic differentiation between D. citri populations from the two different organ types (leaf vs. fruit) (Figure 4).   among the four orchards, while 90% were found within the orchards, with less than 2% coming from between the two citrus species (p > 0.05), consistent with limited host tree specialization in D. citri genotypes. At the tree level within the orchard in Luxi, 1% of the total genetic variations were found among the subpopulations from the five trees, while 99% were found within the individual tree subpopulations (Figure 3). Furthermore, there was no evidence of genetic differentiation between D. citri populations from the two different organ types (leaf vs. fruit) (Figure 4).    Pairwise PhiPT values showed significant genetic differentiation between all provincial subpopulations (p = 0.001), with the PhiPT values ranging from 0.104 (HN vs. GZ) to 0.421 (FJ vs. JX) ( Table 5). PCA analysis revealed that the subpopulations from five provinces were divided into two genetic groups. One group included subpopulations from Hunan, Guizhou, and Jiangxi, while the other group included subpopulations from Fujian and Zhejiang ( Figure 5). The minimum spanning network revealed that the MLGs from Fujian and Zhejiang are distinct from the Hunan, Guizhou, and Jiangxi subpopulations ( Figure 6).    The STRUCTURE analysis also identified two genetic clusters (K = 2). However, isolates in the two genetic clusters showed different frequency distributions among the five provincial subpopulations. Isolates of one genetic cluster were distributed mostly in the Fujian subpopulations, and the other genetic cluster was mainly distributed in subpopulations from Hunan, Jiangxi, and Guizhou provinces. The Zhejiang subpopulation had similar representations of both genetic clusters (Figure 7). The results of the STRUCTURE analysis are consistent with the results of the pairwise PhiPT comparisons, the PCA, and minimum spanning network (MSN), which indicated low-level gene flow among five provinces' subpopulations. Interestingly, isolates with genetic elements of both clusters were found in all five provinces, consistent with the presence of hybridization among these geographic populations. The STRUCTURE analysis also identified two genetic clusters (K = 2). However, isolates in the two genetic clusters showed different frequency distributions among the five provincial subpopulations. Isolates of one genetic cluster were distributed mostly in the Fujian subpopulations, and the other genetic cluster was mainly distributed in subpopulations from Hunan, Jiangxi, and Guizhou provinces. The Zhejiang subpopulation had similar representations of both genetic clusters (Figure 7). The results of the STRUCTURE analysis are consistent with the results of the pairwise PhiPT comparisons, the PCA, and minimum spanning network (MSN), which indicated low-level gene flow among five provinces' subpopulations. Interestingly, isolates with genetic elements of both clusters were found in all five provinces, consistent with the presence of hybridization among these geographic populations. Mantel analysis showed significant correlation between genetic and geographic distances (r = 0.7228, p = 0.001), consistent with geographic separation playing a significant role in the genetic differentiation among these five provincial populations.

Mating Type Identification and Ratios
A positive PCR amplification was obtained for all isolates with either primer pair DcM1F/DcM1R, which is specific to isolates of mating type MAT1-1, or primer pair DcM2F/DcM2R, which is specific to isolates of mating type MAT1-2. In populations at all three spatial scales of analyses, the mating type frequencies did not deviate significantly from a 1:1 ratio based on the χ2 test (non-clone-corrected data, 0.166 ≤ p ≤ 0.918; clone-corrected data, 0.132 ≤ p ≤ 0.932) and the exact binomial analysis Mantel analysis showed significant correlation between genetic and geographic distances (r = 0.7228, p = 0.001), consistent with geographic separation playing a significant role in the genetic differentiation among these five provincial populations.

Mating Type Identification and Ratios
A positive PCR amplification was obtained for all isolates with either primer pair DcM1F/DcM1R, which is specific to isolates of mating type MAT1-1, or primer pair DcM2F/DcM2R, which is specific to isolates of mating type MAT1-2. In populations at all three spatial scales of analyses, the mating type frequencies did not deviate significantly from a 1:1 ratio based on the χ 2 test (non-clone-corrected data, 0.166 ≤ p ≤ 0.918; clonecorrected data, 0.132 ≤ p ≤ 0.932) and the exact binomial analysis (non-clone-corrected data, 0.198 ≤ p ≤ 1; clone-corrected data, 0.159 ≤ p ≤ 1) (Table 6). Indeed, within each subpopulation from a province, orchard, and tree, the mating types were also in similar frequencies (Table 6). Furthermore, the two mating types were found to co-exist on 13 of the 19 (68%) fruits and 6 of the 9 (67%) leaves from which at least two isolates were obtained for analyses (Figure 2).

Genetic Diversity of the Two MAT Idiomorph Populations
Allele diversity analyses, including the number of different alleles, the number of effective alleles, Shannon's information indices, Nei's gene diversities, and unbiased allelic diversities between the MAT1-1 and MAT1-2 subpopulations, showed high and similar levels of genetic diversity at all three spatial levels ( Table 8). For both the non-clone-corrected and clone-corrected dataset, AMOVA showed that less than 1% of the genetic variation could be explained by differences between the MAT1-1 and MAT1-2 subpopulations at all three spatial scales ( Figure 8). Furthermore, the minimum spanning network revealed that the MLGs of two mating types were intermixed at all three of the studied scales ( Figure 9). Together, these results are consistent with frequent mating and recombination in natural populations of D. citri in citrus orchards.

Discussion
Melanose caused by D. citri is an important disease of citrus in southern China and has significantly influenced the marketability of citrus fruits [3]. In this study, we analyzed the patterns of genetic variation of D. citri in orchards of C. reticulata and C. sinensis to understand the mode of reproduction of D. citri in nature and to reveal the relationships among populations of D. citri from different trees, orchards, and provinces in southern China. Consistent with what we hypothesized, geographic separation was found to have played a significant role in structuring the genetic diversity of D. citri in southern China. In addition, as was hypothesized, limited differences were found between host tree species-based subpopulations and between plant organ-based subpopulations. However, different from what we expected, both linkage disequilibrium analysis and phylogenetic incompatibility tests revealed abundant evidence for recombination within individual subpopulations. Furthermore, the ratios of the two mating types MAT1-1-1 and MAT1-2-1 in both the total sample and in most subpopulations of D. citri analyzed here did not deviate significantly from 1:1. Together, the genetic signatures of the D. citri populations in southern China are consistent with this pathogen being highly endemic in these regions [19,41].
Our analyses revealed that geographical isolation played a significant role in population differentiations among the five provincial populations of D. citri. The influences of geography on population structure have been reported in many other plant fungal pathogens such as the wheat pathogen Mycosphaerella graminicola [42], the chili pepper anthracnose pathogen Colletotrichum truncatum [43], the grapevine dieback pathogen Diaporthe spp. [15], the yam wilt pathogen Fusarium oxysporum species complex [44], and the tea-oil tree anthracnose pathogen Colletotrichum fructicola [45]. Indeed, in southern China, a diversity of genetic relationships among regional populations has been observed for plant fungal pathogens, e.g., [43][44][45]. For example, multilocus sequence typing revealed significant genetic differentiations among regional populations of C. fructicola from tea-oil tree plantations in Hunan and Jiangxi provinces [45]. However, long-distance dispersals were also found, likely related to tea-oil tree seedling trade among regions by humans [45]. Long-distance sharing of multilocus genotypes has also been reported for other fungi in southern China, including the human fungal pathogens Cryptococcus neoformans [46] and Candida tropicalis [47], likely due to anthropogenic influences on these fungal pathogens. However, different from most previous studies, no shared multilocus genotype was found in this study among isolates of D. citri from the five provinces in southern China. The result suggested one of two possibilities. In the first, long-distance dispersal of asexual spores of D. citri was uncommon. In the second, long-distance dispersal was common but after dispersal, sexual mating and recombination were frequent enough to break the allelic associations in these asexual propagules to generate new genotypes. Based on the observed allelic distributions among geographic populations at most loci, it seemed that the second possibility was more likely than the first. However, with increased trade and human travel among these provinces, we will likely observe long-distance dispersals of D. citri among geographic populations.
We found that the two mating types (MAT1-1-1 and MAT1-2-1) were distributed at similar frequencies in most subpopulations of D. citri at all three spatial scales. In combination with results from linkage equilibrium and Prp analyses, this result is consistent with the frequent mating and recombination of D. citri in nature. Similar results have been reported in several plant fungal pathogens such as Phyllosticta citricarpa [48,49], Zymoseptoria tritici [50], Dothistroma septosporum [51], and C. truncatum [43]. High genetic diversity is typical of sexually reproducing species [18,20,41]. Carstens et al. [49] studied the global genetic diversity of P. citricarpa, the causal agent of citrus black spot, and found that the populations from China and Australia were sexually reproducing populations with very high genetic diversity. In contrast, the population from United States was consistent with asexual reproduction, with only one multilocus genotype (MLG), and all strains had the same mating type, MAT1-2-1. Our research showed that D. citri populations have high levels of genetic diversity at all three studied spatial scales in China, likely due to frequent sexual recombination in orchards. Similar results have been reported another citrus fungal pathogen P. citriasiana [52].
Interestingly, we observed very few shared MLGs among the 339 isolates. Previous studies have suggested that the large number of conidia of D. citri produced on dead twigs are likely the infectious propagules that are transmitted to healthy fruits and leaves through rain or irrigation water to cause new infections [53]. If this hypothesis were true, we should observe frequent MLG sharing throughout the same tree and the same orchard. However, our study found very few shared MLGs of D. citri, even in a very small spatial area. Therefore, we speculate that although D. citri can produce a large number of conidia in the orchard, the infection rate on citrus by conidia is likely low, with most of the lesions not caused by conidia but likely by sexual spores (ascospores).
A previous study suggested that D. citri likely has multiple modes of nutrient acquisition, as a biotroph that can cause diseases, a saprophyte that grows on a variety of dead branches and twigs, and as endophyte living under the epidermis in the intercellular space of trees [2,7,8]. As most citrus trees are perennial plants, the D. citri populations could have existed in these orchards for an extended time and represented populations in relative equilibrium. In such equilibrium populations, sexual reproduction and local adaptation are likely common for populations of D. citri, potentially making it difficult for immigrants to establish in the endemic niches. However, the observation of significant genetic differentiations among the five provincial populations does not mean that there was no migration. As shown in the minimum spanning tree, closely related strains were found from several geographically distinct areas. In addition, STRUCTURE analyses identified evidence for genetic admixture between the two highly differentiated genetic clusters in most geographic populations (but most prevalent in Zhejiang Province). Together, these results suggest that the D. citri populations are likely changing. With increasing human influences, both directly through the exchange of citrus seedlings and indirectly by impacting climate change, the changes in D. citri populations will likely accelerate.
Though the sample size was relatively large, our study only analyzed samples of D. citri from two Citrus tree species in 21 orchards in five provinces in southern China. Many issues about the ecology, evolution, and population genetics remain unanswered. For example, at present, the exact host range of D. citri outside citrus trees remains undefined. Further sampling and analyses of D. citri populations from host plants outside of Citrus species, from other Citrus species (i.e., species other than C. reticulata and C. sinensis that we analyzed here), and from decaying twigs on the ground over time are needed in order to provide a more robust understanding on the origins and epidemiology of the D. citri populations causing citrus melanose.