Diversity and Genetic Variation among Brevipalpus Populations from Brazil and Mexico

Brevipalpus phoenicis s.l. is an economically important vector of the Citrus leprosis virus-C (CiLV-C), one of the most severe diseases attacking citrus orchards worldwide. Effective control strategies for this mite should be designed based on basic information including its population structure, and particularly the factors that influence its dynamics. We sampled sweet orange orchards extensively in eight locations in Brazil and 12 in Mexico. Population genetic structure and genetic variation between both countries, among locations and among sampling sites within locations were evaluated by analysing nucleotide sequence data from fragments of the mitochondrial cytochrome oxidase subunit I (COI). In both countries, B. yothersi was the most common species and was found in almost all locations. Individuals from B. papayensis were found in two locations in Brazil. Brevipalpus yothersi populations collected in Brazil were more genetically diverse (14 haplotypes) than Mexican populations (four haplotypes). Although geographical origin had a low but significant effect (ca. 25%) on the population structure, the greatest effect was from the within location comparison (37.02 %). Potential factors driving our results were discussed.


Introduction
Citrus crops are important worldwide and sweet orange, Citrus sinensis (L.) Osbeck (Sapindales: Rutaceae), is the most economically important species produced [1]. Currently, Brazil is the world's largest producer of oranges followed by the United States, China, India and Mexico [1]. Currently, one of the most important viral diseases affecting citrus production in Brazil is leprosis, caused by Citrus leprosis virus C (CiLV-C) [2]. Transmission of this virus has been related to mite species in the genus Brevipalpus, specifically B. phoenicis (Geijskes) [3][4][5]. However, the existence of a species complex within B. phoenicis [6] makes it more difficult to assess the true role of each species in the transmission of CiLV-C. Previously, this species complex has been referred to as B. phoenicis group species A-G [6], but recently the species status of B. phoenicis has been revised and the putative species referred to as species groups within B. phoenicis have been elevated to separate species, specifically: B. azores Beard [7]. All the information reported so far regarding B. phoenicis has been done without considering the existence of a species complex. Therefore, we will consider this species as B. phoenicis sensu lato, and wherever possible we will refer to the new species description made by Beard et al. [7].
Brevipalpus phoenicis s.l. is a tropical and subtropical species that feeds on at least 486 host plants including agricultural, ornamental and weed species [8][9][10]. Reproduction of Brevipalpus mite species is by thelytokous parthenogenesis with females producing females that are genetically similar [11]. Females are haploid with two chromosomes and males are rarely found [12]. Interestingly, asexuality in this species complex is due to the presence of bacteria from the genus Cardinium [13]. When these bacteria were removed from mite populations, some males were produced [14], although they appeared to be unable to reproduce [15]. Although some males can be produced, perhaps by an inefficient transmission of the bacteria, as suggested by Groot et al. [13], the lack of functional males induces parthenogenesis as the main mode of reproduction in this species [14]. The fact that this species is polyphagous contradicts the hypothesis that asexual species are unsuccessful colonizers of different environments [16]. However, it has also been reported that successful colonization of new host plants depends on the original host plant. For example, B. phoencisis s.l originating from acerola (Malpighia glabra L. Malpighiales: Malpighiaceae) did not adapt to two new host plants testedsweet orange and hibiscus (Hibiscus rosa-sinensis, L. Malvales: Malvaceae). However, populations originating from sweet orange adapted well to the other two host plant species [17]. It is not known whether these results may be explained by the recent description of cryptic species within B. phoenicis s.l. [7] where each new species might be associated with different host ranges.
In Brazil, only B. phoenicis has been reported causing damage on citrus, mainly due to the transmission of CiLV-C [2,18]. However, in Mexico B. phoenicis s.l. is part of a community of mites on citrus that also include Brevipalpus obovatus (Donnadieu) and Brevipalpus californicus (Banks) [19][20][21][22][23]. In both countries, the presence of B. phoenicis s.l. is consistent with damage, suggesting the effect of CiLV-C in Mexico may become as severe as in Brazil if no new control strategies are designed and implemented. Currently, the most common strategy to control CiLV-C is by controlling mite populations [24][25]. In Brazil, the citrus industry spends $US 62 million per year on the control of B. phoenicis [26]. Effective control strategies for this mite should be based on an understanding of its population structure, and particularly the factors that influence its dynamics [27]. In Mexico, to our knowledge, there have only been two studies on the population dynamics and distribution of B. phoenicis on different citrus species and in different regions of Mexico [23][24][25][26][27][28].
Brevipalpus yothersi (formerly B. phoenicis morphotype B) has been reported in both Brazil and Mexico, but B. papayensis (formerly B. phoenicis morphotype C) has only been reported from Brazil [29]. Despite studies determining genetic population structure and genetic variation [30], and the recent publication of Beard et al. [7], studies of different species within the B. phoenicis s.l. species complex in Brazilian and Mexican populations are practically non-existent and therefore urgently needed. There are several molecular markers used to resolve taxonomic relationships and quantify genetic variation within the same population of a particular mite species [31,32]. DNA sequence information of the mitochondrial cytochrome oxidase subunit I (COI) has also been used previously to quantify genetic variation and population structure of other mite pests on citrus [33] and to aid taxonomic identification of mites within Tetranychidae and Tenuipalpidae [34]. Improved analyses of genetic diversity and population structure over both large and local geographic scales are important to understand the factors affecting population dynamics and to design effective control strategies [27,35].
With this aim in mind, we took extensive samples of mites from different locations in Mexico and Brazil and compared the species diversity, genetic diversity and population structure of the B. phoenicis s.l. species complex using COI sequence information to infer relationships between haplotypes and evaluate genetic differentiation among different populations.

Sampling of mites
Brevipalpus mites were collected from sweet orange orchards. In total, we sampled 35 orchards that were distributed as follows: one orchard at each of 11 locations in Brazil ( Table 1) from five states, and one orchard at each of 24 locations in Mexico (Table 1) from three states. Collections were made from September to November 2012 in Brazil, and from February to July 2013 in Mexico. At each location in Mexico, samples of leaves, branches and fruits were taken from each of five trees within the orchard (each tree represented one sample) and returned to the laboratory. The trees were selected based on their position in the orchard, one from each of the four corners and one from the centre. In the laboratory the plant material from each tree was searched for mites using a binocular stereomicroscope. Adult mites were collected using a fine brush and deposited into microtubes containing absolute ethanol. Where possible up to 100 individuals were collected from each tree. When only very few mites were found, they were allowed to reproduce on sweet orange fruits under laboratory conditions, following the methods proposed by Chiavegato-Gonzaga [36], to acquire sufficient numbers for identification and analysis. Although five samples were always taken from each location, some had no mites at all and therefore, the final number of trees sampled differed among locations. In Brazil, mite samples were collected and processed in a similar way except from the localities of Gurupi, Sao José Castanhal and Capitao Poco, where mites were stored in 70% ethanol. Before processing, these mites were lyophilized for 20 min to remove the alcohol. The study was conducted in private orchards with the permission of the landowners. The field studies did not involve endangered or protected species. In total 59 samples of mites were collected (Table 1).

Morphological identification
Only adult females were used for morphological identification. From each sample, 30 mites were separated and processed for microscopy. Mites were deposited in Hoyer mounting liquid on a glass slide and covered with a coverslip [37]. The glass slides were maintained at 45°C for 15 days. All slides were cleaned using 70% ethanol and cotton swabs. The coverslip was sealed using transparent nail polish. Species identifications were made according to dichotomous keys [19,20,22]. Species determination within the B. phoenicis s.l. species complex was done using the descriptions of Beard et al [7]. All specimens were examined by phase and differential interference contrast (DIC) microscopy.

Analysis of genetic variation among populations of identified Brevipalpus species
DNA was extracted from ten adults per sample using the DNA extraction kit DNeasy Blood & Tissue (Qiagen, Germantown, MD, USA) following the manufacturer's instructions. Partial sequences of the COI gene were obtained using the primers DNF-5' TGA TTT TTT GGT

Data analysis
Sequence traces were assembled using BioEdit [38]. For this, clear and unambiguous peaks representing each base were located in each raw sequence trace. Data beyond this point at both ends of the sequence were discarded. The same was done for forward and reverse sequences for each sample and, by combining both sequences (forward and reverse) we were able to generate only one sequence for each sample. All sequences were truncated to the same length (352 bp) to eliminate missing data. Multiple sequence alignments were made using Clustal W [39]. Maximum parsimony, maximum likelihood and neighbour joining analyses were done using Molecular Evolutionary Genetic Analysis (MEGA) software ver. 5.0 for Windows [40], using the Close-Neighbour-Interchange algorithm. The robustness of branches was estimated by bootstrap analysis with 1000 repeated samplings of the data [41]. Tree reconstructions were made excluding non-synonymous substitutions, without any effect on tree topology. We show the tree including all sites. Sequences from related species within the genus Brevipalpus were retrieved from GenBank and used for comparison (Table 1). A Cenopalpus pulcher (Canestrini and Fanzago) (Acari: Tenuipalpidae) sequence was used as the outgroup for this analysis. In addition, the Nei-Gojobori method [42], as implemented in the Z test in the program MEGA 5.0 [40], was used to compute the synonymous and nonsynonymous distances at a 5% significance level. Genetic differences among haplotypes were represented in a maximum parsimony network [43] using TCS 1.21 [44] with 95% confidence in the connection limit (limits of parsimony) and where gaps were treated as a 5th state. Haplotype and nucleotide diversity [45] were calculated using DnaSP v5 [46]. Finally, the partition of genetic variation between countries, among populations (locations) within each country and amongst all populations regardless of country or location of origin was estimated only for B. yothersi populations using analysis of molecular variance (AMOVA) and by computing F-statistics using Arlequin v. 3.5 [47] with 10000 permutations. No AMOVA analysis was attempted for B. papayensis as there were only sufficient samples from Brazil for molecular study.

Results
All specimens were identified morphologically from Brazilian and Mexican samples, and the majority were from Brevipalpus yothersi and B. papayensis. In only three samples (O2, O4 and O5, Table 1), B. californicus was also found. When two species were found, single adults were placed on oranges (one mite per orange and maintained separately to avoid any cross-contamination), and allowed to reproduce parthenogenetically. When a significant number of mites were produced (more than 100), ca. five mites per orange were mounted, identified, and only the sample oranges containing B. yothersi or B. papayensis were used in the genetic analysis. Those sample oranges with B. californicus were discarded because they were in such low numbers that any meaningful molecular analysis of B. californicus was not possible. In both countries, B. yothersi (Fig 1) was the most common species and was found in almost all locations. Individuals from B. papayensis (Fig 2) were found in two locations in Brazil (Pouse Alegre and Lavras, Minas Gerais, Table 1), and in only one location in Mexico (O1: Ocozocoautla de Espinosa) (Table 1), where B. yothersi was also found. As very low numbers of mites (less than ten mites) from both B. yothersi and B. papayensis were found in location O1, increasing their population by rearing them on sweet oranges was attempted in the laboratory. However, successful reproduction was only achieved for B. yothersi meaning there were insufficient Mexican B. papayensis for genetic analysis. The morphological characteristics used to separate B. yothersi and B. papayensis, are listed in Table 2, and illustrated in Figs 1 and 2.   Table 1. Phylogenetic analyses showed a clear separation among species within the genus Brevipalpus with all bootstrap values above 90% (Fig 3). All samples morphologically identified as B. yothersi were grouped together with the B. phoenicis type 2 [32] (now considered as B. yothersi) sequences retrieved from GenBank that were used as a reference. Samples morphologically identified as B. papayensis were grouped together with the B. phoenicis type 1 [32] (now considered as B. papayensis) sequences (Fig 3). All samples of B. papayensis evaluated were from Brazil only, while the samples of B. yothersi evaluated were collected from both Brazil and Mexico. The B. yothersi samples could be further separated into two distinct groups, G1 and G2 (Fig 3). Group G1 contained 48 samples, with samples SJRPB-1 and SJRPP-3 forming a distinct group separated from the other samples with a bootstrap value above 90%. Group G2 contained five samples, all from Brazil, with samples SJRPL-2 and SJRPL-3 forming a distinct group, followed by sample SJRPL-1 with bootstrap values above 90% (Fig 3).
Haplotype Network analysis showed the existence of 20 haplotypes from 35 sampled trees at 20 orchard localities in Mexico and Brazil (Table 1; Fig 4). There were three discreet networks, network N1 contained only samples identified as B. papayensis, and networks N2 and N3 contained only samples identified as B. yothersi. Network N1 contained three haplotypes: H01, H02 and H03. H02 and H03 were each found in only one sample but H01 was found in four samples; all six samples were collected in Brazil. The second network (N2) contained four haplotypes, each found in only one sample and all from Brazil. The third and largest network (N3) contained 11 haplotypes that were found in 47 samples. The most common haplotype was H08, which was found in 36 samples, all from Mexico except PAP4-1 and SJRPB-2, which were from Brazil. Finally, there were two independent haplotypes from Brazil that did not belong to any network (H07 and H17) and were each found in only one sample (Fig 4). There was greater haplotype (0.966±0.028) and nucleotide (0.06248±0.00710) diversity in B. yothersi populations from Brazil, where 17 haplotypes were recorded from 22 samples analysed, compared with the haplotype (0.158±0.080) and nucleotide (0.000092±0.00152) diversity found in Mexico with only four haplotypes from 37 samples analysed.
AMOVA analysis (Table 3) revealed that the greatest amount of variation among B. yothersi populations could be accounted for by differences between individuals within each population Dorsal palp femur seta Setiform and barbed (Fig 1A) Broadly setiform and barbed (Fig 2A) Sublateral region of propodosoma Posterior region forming large cells, anterior region minus reticulate (Fig 1B) Reticulations like large cells only in the posterior end ( Fig 2B) Opisthosoma Reticulation between setae e1 and h1 with "V" shaped folds (Fig 1C) Reticulation between setae e1 and h1 starting with transverse folds abruptly becoming longitudinal folds towards h1 (Fig 2C) Ventral region posterior to setae 4a Rounded reticulations (Fig 1D) Elongate reticulations forming transverse bands (Fig 2D) Spermatheca With a long narrow duct, which merges to an oval vesicle with small distal stipe ( Fig 1E) With a long moderately thick duct, which ends in a spherical vesicle with a crown of small projections (Fig 2E) doi:10.1371/journal.pone.0133861.t002 (81.15%), followed by differences between Brazil and Mexico (16.33%) and finally by differences among populations within each country (2.52%). Although the figure for the difference between Brazil and Mexico was not the greatest, it showed a highly significant P value (P<0.0001) suggesting a geographically structured population (Table 3).

Discussion
Combining morphological and genetic analyses is a powerful way to obtain the maximum information on taxonomic and genetic variation in any species [48]. Considerable taxonomic and genetic information is available for B. phoenicis s.l. [29,30,[49][50][51], but there remain many  gaps in understanding genetic variation within and between populations and species. For example, the recent report raising many morphotypes within the B. phoenicis species complex to species level [7], and the fact that specific relationships between virus and host have been reported [52], suggests the necessity for assessing the ability of each of these new species to transmit CiLV-C. Currently, only the relationship between CiLV-C and B. yothersi has been reported [53], so the relationship between CiLV-C and B. papayensis remains to be investigated. In addition, the factors that drive genetic variation within these new species, and how this might vary between their geographical origins are poorly understood. Morphological identification showed the existence of two main species, B. yothersi and B. papayensis. Mites from B. papayensis were only found in Lavras and Pouso Alegre (Minas Gerais), Brazil and Ocozocoautla de Espinosa (Chiapas), Mexico, the latter representing the first report of this species from Mexico. Unfortunately, there were insufficient numbers of mites for DNA extraction to evaluate genetic variation as was done with the samples from Brazil. Brevipalpus papayensis specimens in this study were all collected from citrus orchards located in coffee growing regions, which was consistent with previous studies in Brazil [30]. More sampling near coffee plantations must be done to confirm this association.
The use of COI sequencing has been used previously to study B. phoenicis s.l. populations from Brazil, Chile, The Netherlands and USA [52]. Using the same marker, Groot and Breeuwer [13] found conflicts between genetic analysis and morphological identification. Navia et al. [30] suggested that these conflicts were due to the existence of cryptic species within the genus Brevipalpus, which was later confirmed by Beard et al. [7]. Our study confirms the existence of a species complex within B. phoenicis s.l. (Figs 1 and 2), which corresponds to the species B. yothersi and B. papayensis reported by Beard et al. [7]. Our data showed that populations from both species were indeed genetically different (Figs 3 and 4), which was clearly demonstrated in the Brazilian populations. Unfortunately, insufficient individuals from B. papayensis were collected in Mexico to allow us to confirm this genetic separation in Mexican populations. Although Beard et al. [7] have raised all of these morphotypes to species level, we still suggest that their study should be complemented with molecular data from more than one gene [54].
The haplotype network (Fig 4) showed the existence of three discrete networks where network 1 (Fig 4) contained only samples from B. papayensis. The existence of the other networks (2 and 3) including haplotypes 7 and 17 can be considered as a consequence of the genetic variation within B. yothersi. The majority of the specimens collected in both countries, corresponded to B. yothersi, and within this species, greater genetic variation was observed in populations from Brazil than in populations from Mexico (Fig 4). In fact, the majority of the Mexican populations were clustered in one group following phylogenetic analyses (Fig 3) and there were only four haplotypes (Fig 4). Although, thelytokous parthenogenesis may be responsible, it is also likely that selection pressures on B. yothersi populations, such as the number and type of acaricide applications made, are greater in Brazil compared to Mexico, resulting in genetically more diverse populations in Brazil, This mechanism has also been suggested to account for variability in genetic diversity in Panonychus citri (Acari: Tetranychidae), another important mite pest in citrus orchards worldwide. ITS1 sequence analysis revealed greatest genetic diversity among P. citri populations from different locations in China, where the control of this mite relied most heavily on acaricides [55].
Additionally, it is possible that the host plant may also be playing a significant role in the greater genetic diversity found in the Brazilian samples of B. yothersi. Although all Brazilian mites were collected from sweet orange (C. sinensis) orchards, samples were collected from four different varieties, Pêra, Lima, Bahia and a non-commercial variety. With the exception of one sample (O5) all samples from Mexico were from the same variety, Valencia (Table 1). This may contribute to a lack of genetic diversity within B. yothersi populations in Mexico. Hostassociation differentiation (HAD) has been described for mites previously. For example the existence of a large number of genetically distinct lineages of the mite Aceria tosichella Keifer (Prostigmata: Eriophyidae) were associated with specific plant hosts, regardless of geographic origin [56]. Based on this, we propose that sampling different varieties of C. sinensis may have contributed to the existence of more haplotypes within B. yothersi populations in Brazil compared with Mexico where only one orange variety is normally grown by producers. Although they reproduce clonally, genetically distinct lineages or haplotypes of B. yothersi may be related to the orange variety on which they were collected suggesting some level of host-plant specialisation, a process described as the Frozen Niche Variation model (FNV) as previously reported for B. phoenicis s.l. by Groot et al. [17]. Currently, we are performing more studies to assess the role of the host plant in shaping patterns of genetic variation in B. yothersi and B. papayensis populations.
In summary, morphological and genetic analysis has demonstrated the existence of B. yothersi and B. papayensis populations collected in Brazil and Mexico, although genetic variation between these two species was only confirmed for Brazilian populations. In both countries, B. yothersi was the most abundant species and was more genetically diverse in Brazil than in Mexico. The existence of these two species in Mexico and Brazil requires research, including behavioural and ecological studies, as it is likely that one species may be more efficient in transmitting CiLV-V. Such information should be included in the design of monitoring and control programs, especially for B. yothersi, which was the most abundant species compared to B. papayensis.