Genetic diversity of Hirsutella thompsonii isolates from Thailand based on AFLP analysis and partial-tubulin gene sequences

Amplified fragment length polymorphism (AFLP) was used to investigate polymorphism among 43 Hirsutella thompsonii isolates (33 from Thailand) obtained from various mite species. The outgroups were an unidentified Hirsutella isolate along with Hirsutella nodulosa and Hirsutella kirchneri. Phylogenetic analyses of the AFLP data showed significant variation among isolates and the existence of three H. thompsonii clades. We also investigated the isolates using PCR with specific primers for the Hirsutella exotoxin gene Hirsutelin A (HtA) and 18 of these isolates were used for sequencing of the partial -tubulin gene. Phylogenetic analyses of -tubulin sequences showed two distinct H. thompsonii clades, one of which included AFLP clades I and II. For both markers grouping of the H. thompsonii isolates was not related to either host mite species or geographical origin, although for the HtA gene one clade contained only isolates with no detectable HtA band. These results confirm the high intraspecific polymorphism of H. thompsonii, and maximum likelihood analysis showed no monophyletic group within this species. To refine the taxonomy of this genus other studies should be undertaken using additional molecular markers and several other Hirsutella isolates.


Introduction
The genus Hirsutella (Hypocreales) contains a variety of fungal pathogens isolated from numerous mite, nematode, and insect hosts (McCoy, 1996;Jaffee, 2000;Chandler et al., 2000;Van der Geest et al., 2000).For example, the Hirsutella species H. nodulosus, H. citriformis and H. gigantean, infect lepidopterans, hemipteran and dipterans producing structures composed of a compact group of erect conidiophores (synnemata) while H. rhossiliensis and H. minnesotensis infect plant-parasitic nematodes.The most wide-studied species from this genus is the mitespecific pathogen Hirsutella thompsonii (McCoy, 1981), a mononematous species that has been separated into three morphologically distinct groups consisting of the tropical variety synnematosa, the subtropical variety vinacea and the temperate variety thompsonii.This fungus has a worldwide distribution on different eriophyoid and tetranychid mite hosts and during hot, humid weather can cause spectacular natural epizootics among mite populations (e.g.citrus rust mites, blueberry, coconut and tomato mites, etc.) and is considered to be a major natural enemy of various mite pests (Chandler et al., 2000).
Isolates of H. thompsonii have displayed phenotypic plasticity in vitro and certain isolates produce multiple conidial states (McCoy, 1996).Isoenzyme analyses of 15 geographically diverse H. thompsonii isolates has shown a high degree of intra-specific variability even among morphologically similar Hirsutella isolates (Boucias et al., 1982).Maimala et al. (2002) reported that the major hosts of H. thompsonii are eriophyoid mites displaying specific host plant preferences and that many H. thompsonii isolated from a range of eriophyiids varied in their ability to produce exotoxins such as Hirsutellin A (HtA).More recently, (Maimala, 2004) reported extensive plasticity in the growth characteristics of H. thompsonii isolated from a range of eriophyiids and one may speculate that the variation exhibited by H. thompsonii strains may have co-evolved with the specific mite-plant interactions.
A major objective of the work described in this paper was to determine if the variation observed in H. thompsonii is linked to phenotypic plasticity or genotypic variation.Amplified fragment length polymorphism (AFLP) were used to examine H. thompsonii isolates exhibiting a spectrum of phenotypic differences and the results were complimented by an analysis of HtA and b-tubulin gene data.

Fungal strains
In this study we used 43 Hirsutella thompsonii Fisher isolates collected from various mite species, 33 isolates having been collected from various mite species inhabiting different plants in various regions of Thailand (Table 1, Maimala et al., 2002).The outgroups consisted of Hirsutella isolate 624 (unidentified to species level), Hirsutella nodulosa and Hirsutella kirchneri, from the United States Agricultural Research Service (USDA-ARS) Entomopathogenic Fungal Collection.

DNA extraction
Isolates were inoculated into 250 mL flasks containing 40 mL of malt extract broth (glucose, 1%; malt extract 0.3%; peptone 0.5%; yeast extract, 0.3%) and incubated at 25 °C at 250 revs min -1 for 4 days.Mycelial samples were collected on Whatman number 1 filter-paper using vacuum filtration, frozen at -70 °C, lyophilized and stored at -20 °C.About 25 mg of the lyophilized mycelia of each sample were frozen in liquid nitrogen, crushed in a mortar and the total genomic DNA extracted using a modification of the cetyltrimethylammonium bromide (CTAB) DNA extraction protocol of Boucias et al. (2000a).

AFLP analysis
For each sample, 0.3-2.0mg of genomic DNA was digested with EcoRI and ligated to EcoRI adaptors in a single overnight reaction at 37 °C (Suazo and Hall, 1999).The digestion-ligation reactions were diluted with TE buffer to a final volume of 200 mL and stored at -20 °C.A series of six 19-mer primers (synthesized by Gemini Biotech Houston, TX) were used, consisting of the EcoRI adapter sequence GACTGCGTACCAATTC plus three 3' selective nucleotides (AGG, AGT, GGC, ACC, CGA and GGG).The amplification reactions were conducted in a volume of 25 mL containing the anchor-annealed EcoRI fragments, primer and Taq DNA polymerase using the conditions established by Suazo and Hall (1999) and the PCR products separated by electrophoreses on 1.5% (w/v) agarose-synergel (0.7% (w/v) agarose and 0.4% (w/v) synergel; Diversified Biotech, Boston, MA), stained with ethidium bromide and photographed under UV light.

Analysis of the b-tubulin and Hirsutelin A genes
Partial b-tubulin sequences were PCR amplified using primers designed from consensus regions of selected fungal, algal and protozoan b-tubulin genes, the primers being btubF 5' TGGGCYAARGGYCACTACACYGA 3' and btubR 5' TCAGTGAACTCCATCTCRTCCAT 3' (Tartar et al., 2002).PCR reactions were performed in a final volume of 25 mL containing 1 mL of genomic DNA, 0.4 mM of primers, 200 mM of dNTPs, 1x Taq recommended buffer, and 1 unit of Taq DNA polymerase (Fisher Scientific, Pittsburg, PA).The cycling program was 1 min at 94 °C and then 30 cycles (1 min denaturation at 94 °C followed by 1 min annealing at 55 °C and 2 min extension at 68 °C) with a final 6 min elongation step at 68 °C.The PCR products were purified using a QIAquick PCR purification kit (QIAGEN Inc., Santa Clara, CA) and then sent to the Interdisciplinary Center for Biotechnology Research (ICBR) at the University of Florida for sequencing.
The HtA gene was PCR amplified from the different DNA samples using the primers DGB35 (5'-CATATGAA AGCCTTTACTGCCATTCTC-3') and DGB14 (3'-ATCT CAGAACCAACGAGCCTAGG-5') (Maimala et al., 2002).The PCR conditions were the same as for the b-tubulin gene amplification, except that the annealing temperature used for these primers was 60 °C.The PCR products were separated by electrophoreses on 1% (w/v) agarose gel and visualized using ethidium bromide staining.

Phylogenetic analyses
Bands from AFLP analysis were scored as present or absent using digitized photographs of the gels.Experiments were repeated at least once, and only DNA fragments consistently present were recorded and considered to be binary characters.DNA fingerprints of each isolate were converted into a binary matrix.The data were assumed unordered with no a priori weighting.
The partial b-tubulin gene sequences were initially aligned with GenBank data retrieved using the Clustal X program version 1.83 (Thompson et al., 1997) and the alignments refined visually using the MacClade program version 4.02 (Sinauer Associates, Inc., Sunderland, MA, USA).Three homologous GenBank b-tubulin gene sequences from the fungi Neurospora crassa (M13630) Colletotrichum gloeosporioides f. sp.aeschynomene (U14138) and Cordyceps bassiana (AY366063) were included in the alignment.Gene genealogies were reconstructed using the distance-based, neighbor joining (NJ; Saitou and Nei, 1987), and the discrete character-based, maximum parsimony (MP) and maximum likelihood (ML), algorithms included in the PAUP* program version 4.0b10 (Swofford, 2002).Maximum parsimony analysis was performed using a heuristic search including 1,000 bootstrap replicates of random addition sequences and the tree-bisection-reconnection branch swapping algorithm with the steepest descend option not in effect.The NJ tree was obtained using the mean-character difference for the AFLP dataset and the Log Determinant distance measure for the b-tubulin sequences, each using 1,000 bootstrap replicates.The ML analysis was applied only to the b-tubulin sequences dataset.The GTR+G model favored by ModelTest program version 3.06 (Posada and Crandall, 1998) was used for the ML searches which were conducted heuristically with 100 bootstrap replicates.

Results
The HtA gene PCR amplification with HtA genespecific primers confirmed the results previously obtained by Maimala et al. (2002).A consistent ~600 bp band (data not shown) was produced by 67% of the H. thompsonii isolates (HtA+), 17% of which produced weak bands (Table 1).The Hirsutella species other than H. thompsonii showed no detectable HtA bands.
The AFLP analysis of the 43 H. thompsonii isolates and Hirsutella isolate 624 (Table 1) with the five different primers produced identical AFLP profiles for isolates HTF75 and HTF87 and also for isolates 1707 and H4, so in these cases only the profiles of isolates HTF75 and 1707 were used in the final analyses.The amplification of the anchor-ligated EcoRI fragments with all the primers used produced an array of scoreable bands with abundant polymorphisms that were resolved on ethidium bromide-stained agarose synergels (Figure 1).The number of polymorphic AFLP bands selected varied from 25 to 39 per primer and their sizes ranged from ~250 to 3,000 bp.A total of 183 fragments were scored as AFLP markers.No single fragment was found to be amplified in all of the 42 isolates analyzed (i.e., no fragment was monomorphic) and 161 were parsimony-informative characters.The topologies of the NJ and MP trees were identical, so only the parsimony tree is shown in Figure 2. Three clades (I, II and III) were identified, with the Hirsutella isolate 624 and three H. thompsonii isolates (18 II, 2057 and 2110) being outliers which remained independent from the clades.In both MP and NJ analyses clades I and III were well defined with bootstrap support (bs) of 86%-100%, while clade II was poorly supported and included two groups with bs of 79-84%.Each of these groups contained subgroups that were supported by bootstrap values from NJ and MP analysis.Interestingly, the AFLP clades were not related to original host or geographical location (Table 1, Figure 2).For example, seven H. thompsonii isolates obtained from the mite Phyllocoptruta oleivorus were included among the nine isolates from the USA (Table 1) which all grouped in clade 1, while three isolates from the same host species were placed in clade II.Furthermore, the well-supported Clade III included isolates from different regions from Thailand and from various hosts (Table 1, Figure 2).In several instances, clade designation seemed to be correlated with the expression of the HtA gene (Table 1).For example, all of the members in clade III did not produce the HtA PCR product (HtA -) whereas clade I contained mainly HtA + isolates (17 out of 19) while clade II contained a mix of HtA + and HtA -isolates (Table 1, Figure 2).The b-tubulin sequences analyses showed that the btubF and btubR primers amplified a segment of the btubulin gene that includes 'intron F' from the Neurospora crassa b-tubulin gene sequence (GenBank M13630).The partial b-tubulin sequences obtained from the 20 Hirsutella isolates analyzed consisted of 750 (isolate 2109) to 805 (isolate 1730) bp.Identical sequences were obtained for isolates HTF75, 873 and 2121 so only the HTF75 sequence was included in subsequent analyses.The 57 bp related to the intron was not included in the alignment of the sequences.The final alignment contained 681 positions, of which 186 were variable and 124 were parsimonyinformative characters.The sequences have been deposited in GenBank.Phylogenetic trees generated by both distance and parsimony methods had similar topologies, and are represented by the maximum parsimony tree (MP) shown in Figure 3 which shows four clades (I to IV).Some of the isolates that were in clade I and II in the trees obtained with the AFLP dataset were shown to be in a strongly supported group (100% bs for both NJ and MP analyses) that included the well-supported clade I (98% bs) and clade II (85% bs).The maximum likelihood tree (ML) showed a slightly different topology, showing also clades I and II as a unique clade, but a close relationship (77% bs) between clade III and clade IV (Figure 4).

Discussion
Our research showed that, in general, H. thompsonii isolates obtained from various mite species showed high polymorphism for the molecular markers studied.Amplified fragment length polymorphism (AFLP) has proved to be a powerful DNA marker technique to evaluate polymorphism among isolates of H. thompsonii as well for other fungi (Boucias et al., 2000a, b).Based on AFLP analyses, the different H. thompsonii were grouped into three clades uncorrelated with either location or mite species.Different strains isolated from the same mite species were often found in two or more of the delineated clades (Table 1).Similarly, no correlation was found between phenotypic data (Maimala et al., 2002;Maimala, 2004) and clade designation, with, for example, isolates 873 (clade I) and 1690 (clade II) showing both polyblastic conidia and synnemata whereas other isolates (13, 1614, 1707, 1722, 1730, 1764 and 1813), distributed in all three clades, produce polyblastic conidia without synnemata (Maimala et al., 2002).Comparisons of other phenotypic data, such as plant host,  fungal growth rate, colony color and morphology, and phialide size (Maimala, 2004) failed to correlate any morphological phenotype(s) to the genotypic data resulting from our AFLP analysis.However, clades I and III showed a relationship with the presence of the HtA gene, with Clade I members (except for one isolate) being HtA+ and all clade III members being HtA-.
Because of the high AFLP marker polymorphism observed among the H. thompsonii isolates studied, we conducted partial sequencing of the b-tubulin gene on a select group of isolates.Analyses of sequence data is a widely used approach in the study of species delimitation and the analysis of DNA sequence information within the b-tubulin region has provided robust phylogenetic information for fungi (Jong et al., 2001;ODonnell et al., 1998), including the entomopathogenic fungi of the genus Paecilomyces and Nomuraea (Han et al., 2002, Luangsa-Ard et al., 2005).Based on b-tubulin analysis, we have observed that Hirsutella grouped with a branch that included C. bassiana, suggesting that this genus possesses affinities to the Hypocreales.Stensrud et al. (2005) reported that Cordyceps species are commonly the teleomorphic stage of anamorph genera of many entomopathogenic fungi.The connection of Hirsutella and Cordyceps was first observed for the ant pathogen Cordyceps unilateralis whose anamorph is Hirsutella formiacarum (Petch, 1924) and several other Cordyceps species have been reported to be the teleomorphs of various Hirsutella species (Kendrick and Carmichael, 1973;Samson et al., 1988).
Our b-tubulin gene analysis allowed the detection of intraspecies variability.In both b-tubulin trees (MP and ML), clades I and II from AFLP data were placed in a unique well-supported clade, and clade III included two isolates which were in separate clades in the AFLP analyses.We also found that b-tubulin clade IV included Hirsutella isolate 624, H. nodulosa and H. kirchneri (the non-thompsonii isolates ), indicating that Hirsutella isolate 624 obtained in Thailand was closely related to the mite pathogen H. nodulosa.
The ML tree indicates that H. thompsonii is not monophyletic because clade III isolates were more closely related to H. nodulosa and H. kirchneri than to the rest of the H. thompsonii isolates studied.Clade III isolates were all HtA -and were isolated from Thailand, albeit from different locations and hosts.The fact that H. thompsonii isolates are difficult to identify is due to the pleomorphic character of this fungus and the high polymorphism detected among different isolates in regard to several biological characteristics and molecular markers (Boucias et al., 1982, Mozes-Koch et al., 1995, Van der Geest et al., 2000).Our results have confirmed that the taxonomy of Hirsutella is complex (Samson et al., 1988) and that genomic characterization, through the analysis of other DNA sequences, should be developed in order to refine the taxonomy of H. thompsonii.

Figure 1 -
Figure 1 -Example of ethidium bromide-stained AFLP reaction products electrophoresed on 1.5% agarose-synergel.The first and last lanes represent the 250 bp DNA Ladder (Invitrogen).

Diversity of Hirsutella thompsonii 719 Figure 3 -
Figure 3 -Maximum parsimony (MP) tree based on b-tubulin sequences.This tree also represents the topology obtained with the neighbor joining (NJ) analysis.Numbers next to the branches are bootstrap percentage values, based on 1,000 replicates and obtained with MP (above) and NJ (below) analyses.The outgroups are Neurospora crassa and Colletotrichum gloeosporioide.

Figure 2 -
Figure 2 -Maximum parsimony (MP) tree based on AFLP characters.This tree also represents the topology obtained with the neighbor joining (NJ) analysis.Numbers next to the branches are bootstrap percentage values, based on 1,000 replicates and obtained with MP (above) and NJ (below) analyses.The outgroup is Hirsutella isolate 624.

Figure 4 -
Figure 4 -Maximum likelihood tree based on b-tubulin sequences.Numbers next to the branches are bootstrap percentage values, based on 1,000 replicates.The outgroups are Neurospora crassa and Colletotrichum gloeosporioide.

Table 1 -
Data on the Hirsutella isolates examined in this study.The HtA gene was detected using the PCR reaction (-no band; + weak band; ++,normal band) and the AFLP and b-tubulin clades (I, II, III and IV) were constructed by maximum parsimony tree analysis of AFLP and partial b-tubulin gene sequence data (NA = not analyzed and an asterisk (*) represents the isolates standed separated from the clades).