Host-to-Pathogen Gene Transfer Facilitated Infection of Insects by a Pathogenic Fungus

Metarhizium robertsii is a plant root colonizing fungus that is also an insect pathogen. Its entomopathogenicity is a characteristic that was acquired during evolution from a plant endophyte ancestor. This transition provides a novel perspective on how new functional mechanisms important for host switching and virulence have evolved. From a random T-DNA insertion library, we obtained a pathogenicity defective mutant that resulted from the disruption of a sterol carrier gene (Mr-npc2a). Phylogenetic analysis revealed that Metarhizium acquired Mr-npc2a from an insect by horizontal gene transfer (HGT). Mr-NPC2a binds to cholesterol, an animal sterol, rather than the fungal sterol ergosterol, indicating it retains the specificity of insect NPC2 proteins. Mr-NPC2a is an intracellular protein and is exclusively expressed in the hemolymph of living insects. The disruption of Mr-npc2a reduced the amount of sterol in cell membranes of the yeast-like hyphal bodies that facilitate dispersal in the host body. These were consequently more susceptible to insect immune responses than the wild type. Transgenic expression of Mr-NPC2a increased the virulence of Beauveria bassiana, an endophytic insect-pathogenic fungus that lacks a Mr-NPC2a homolog.


Introduction
New infectious diseases are constantly appearing, and their origins are diverse. Human behavior and practices are important factors, for example, catastrophic declines in amphibian and bat populations have been attributed to pathogenic fungi spread by humans [1,2]. Another important origin is the ability of infectious agents themselves to evolve different host ranges, which would certainly contribute to the impact of invasive species. Such host switching probably accounts for the wide variety of fungal associations with animals, plants and other fungi [3]. There must be mechanisms for such host shifts, although these remain largely unknown [4,5,6].
Horizontal gene transfer (HGT) between distantly related bacteria contributes significantly to the emergence of new pathogens but HGT is usually thought to play a minor role in eukaryotes. However, data from multiple genomic sequences suggests that HGT has also occurred between eukaryotes [7], and may bestow a clear selective advantage to fungi [8]. A gene encoding a critical virulence factor was transferred from one species of fungal pathogen to another, leading to the emergence of a new damaging disease of wheat [9]. HGT from bacteria to fungi is relatively common [10], and a few examples are known where HGT has occurred between plants and either pathogenic fungi or parasitic plants [11,12,13,14]. Microsporidian intracellular parasites Enephalitozoon may have acquired a purine nucleotide phosphorylase from an arthropod that is not a host [15]. There is also a longstanding controversy as to whether the malaria causing pathogen Plasmodium vivax has horizontally acquired human genetic material that might facilitate its long stay in the body [16]. Otherwise, HGT between eukaryotic pathogens and animal hosts has been neither predicted nor characterized.
Fungi are the commonest pathogens of insects and crucial regulators of insect populations [17,18]. Metarhizium robertsii (formerly known as Metarhizium anisopliae) is the best studied entomopathogenic fungus [19] as well as being an important plant-growth-promoting endophyte [20,21]. Phylogenomic analysis suggests that the ancestor of Metarhizium was an endophyte, with entomopathogenicity being an acquired characteristic [22]. The evolutionary transition of a fungus from an endophyte to an insect pathogen provides a novel perspective on how new functional mechanisms important for host switching and virulence are acquired. In this study, we identified a new virulence factor (Mr-NPC2a) of M. robertsii by screening an Agrobacterium tumefaciensmediated T-DNA insertion library. We provide evidence that Mr-npc2a was horizontally acquired from an insect and allowed Metarhizium to compete with insect hosts for the sterols necessary to maintain cell membrane integrity.

Results
Characterization of the sterol carrier gene Mr-npc2a as a pathogenicity mutant Using Agrobacterium tumefaciens-mediated fungal transformation, we generated a set of 20, 328 T-DNA-tagged M. robertsii strain ARSEF2575 mutants [23] which we screened for virulence mutants against wax worm larvae (Galleria mellonella) (Pet Solutions, OH USA). A transformant (M298) that showed a significantly slower killing speed than the wild type strain was selected for further study (Fig. 1). The open reading frame (ORF) of the gene MAA_03817 (Genbank accession no.: EFZ01221) was found to be disrupted in this mutant. MAA_03817 showed significant similarity to insect Niemann-Pick type C2 (NPC2) proteins, and thus was designated as Mr-NPC2a. We deleted the ORF of Mr-npc2a in wild type M. robertsii, and the resulting disruptant (DMr-npc2a) showed the same loss of virulence as M298 (Fig. 1). The details of the T-DNA insertion and gene deletion alleles are shown in Fig.  S1. DMr-npc2a was complemented by a genomic clone of Mr-npc2a. The details of deletion of Mr-npc2a and the complementation of DMr-npc2a are described in Fig. S2.
The phenotype of DMr-npc2a on PDA (potato dextrose agar) plates was indistinguishable from the wild type strain, M298 and the complemented DMr-npc2a. Formation of infection structures (appressoria) by DMr-npc2a on the hindwings of Locusta migratoria manilensis (kindly provided by Dr. Wangpeng Shi at China Agricultural University) was not significantly different from the wild type strain, M298 and the complemented DMr-npc2a. The hyphal bodies [unicellular hyphal bodies (blastospores) and multicellular hyphal bodies] of DMr-npc2a, M298, the wild type strain and the complemented DMr-npc2a all began to appear in the hemolymph 36-hours after topical infection of G. mellonella, but the number of DMr-npc2a hyphal bodies (2066.8 hyphal bodies/mL) and M298 (2263.2 hyphal bodies/mL) was significantly less than the wild type (36 hyphal bodies 69.3/mL) and the complemented DMr-npc2a (38 hyphal bodies 69.5/mL). From 36 to 60 h postinoculation, wild type hyphal bodies and the complemented DMr-npc2a hyphal bodies significantly outnumbered DMr-npc2a and M298 in the hemolymph. The number of the complemented DMr-npc2a hyphal bodies was not significantly different from the wild type strain at all time points (P.0.05). Likewise, the number of M298 hyphal bodies was not significantly different from DMr-npc2a at all time points (P.0.05) (Table S1). This set of data suggests that disruption of Mr-npc2a only impairs pathogenesis after M. robertsii enters the haemocoel.
All Galleria infected with the wild type strain or the complemented DMr-npc2a died as larvae, whereas nearly 40% of insects infected by DMr-npc2a or M298 survived. Similar numbers of conidia were produced on the cadavers of insects killed by the wild type strain, by the complemented DMr-npc2a, by DMr-npc2a and by M298 (P.0.05), so in terms of fecundity Mr-NPC2a enhanced fitness by 40%. Based on above data, M298 and DMr-npc2a are functionally indistinguishable, so only DMr-npc2a was used for following studies.
Metarhizium acquired the sterol carrier Mr-NPC2a from insect hosts by horizontal gene transfer Genome analysis showed that there are 3 NPC2-like proteins in M. robertsii: MAA_03817 (Mr-NPC2a), MAA_10401 (Mr-NPC2b, Genbank accession no.: EFY94139) and MAA_03340 (Mr-NPC2c, Genbank accession no.: EFZ00744). All three Metarhizium NPC2-like proteins have conserved putative cholesterol/lipid binding sites, and six cysteine residues that form a hydrophobic core for sterol binding typical of Niemann-Pick type C2 (NPC2) sterol carriers. Mr-NPC2b shows 70% similarity to Mr-NPC2c (6e 253 ), but except for the aforementioned cholesterol/lipid binding sites and hydrophobic core, they show no significant similarity to Mr-NPC2a. We screened for homologs of Mr-NPC2a, Mr-NPC2b and Mr-NPC2c proteins using a protein based local alignment search tool (BLASTP, e-value cutoff 1e 205 ) against databases provided by NCBI (Genbank), JGI (Joint Genome Institute) and Broad Institute. No Mr-NPC2a homologs were identified when Mr-NPC2b or Mr-NPC2c were used as the query sequences; likewise no Mr-NPC2b and Mr-NPC2c homologs were identified when Mr-NPC2a was used as the query. Thus, BLASTP clearly differentiates between the evolutionary relationships of Mr-NPC2a and the related Mr-NPC2b and Mr-NPC2c. Homologs of Mr-NPC2b and Mr-NPC2c are widespread in the fungal kingdom being found in Ascomycota, Basidomycota, Zygomycota and Chytridiomycota. A phylogenetic analysis conducted with Mr-NPC2b, Mr-NPC2c and 30 other NPC2s showed that their phylogenetic relationship is consistent with

Author Summary
The ability of infectious agents to evolve different host ranges contributes to the emergence of new diseases, and this host switching could also account for the wide variety of fungal associations with animals, plants and other fungi. There must be mechanisms for such host shifts, but these remain largely unknown. In this study, we phylogenetically predict that the endophytic fungus Metarhizium robertsii acquired a sterol carrier gene from insects through horizontal gene transfer (HGT). This sterol carrier is involved in maintaining cell membrane sterols, and thus membrane integrity, when M. robertsii proliferates in the haemocoel of living insects. Therefore, the acquisition of genetic material from a host has contributed to the development of fungal entomopathogenicity. In order to simulate this evolutionary event, the sterol carrier gene was transformed into an endophytic insect-pathogenic fungus (Beauveria bassiana) that lacks an endogenous Mr-NPC2a homolog. The virulence of B. bassiana was increased by expression of Mr-NPC2a.
Mr-NPC2a homologs were identified in only two fungi (BLASTP, e-value cutoff 1e 205 )-the ergot fungus Claviceps purpurea and Metarhizium acridum. All other Mr-NPC2a homologs were from insects and other animals including vertebrates. No homologs were found in plant, bacteria, archaea and viruses. A TBLASTn search against NCBI genomes (e-value cutoff 1e 205 ) only identified the same homologs as BLASTP. The origin of NPC2a was assessed using a variety of models and methods for phylogenetic reconstruction. The phylogenies consistently showed that Metarhizium and C. purpurea sequences exclusively clustered with insect sequences with high support (Fig. 2, Table S2). We then compared the topology of the obtained tree with those of alternative trees using eight tests including SH-test and AU-test provided by the program CONSEL [24]. The cluster of fungal NPC2as with vertebrate NPC2s was not statistically rejected, but these constrained trees were significantly less supported than the obtained tree. Furthermore, vertebrates lack close contact with Metarhizium and C. purpurea, whereas insects regularly encounter both fungi [17,25]. Since sympatry is almost a prerequisite for HGT [26], the fungi are much more likely to have acquired NPC2a from insects than vertebrates. Details about the tree topology comparisons are presented in Table S3.
Although it is the most likely that Metarhizium and C. purpurea acquired NPC2a from insects, the fungal NPC2as show low overall similarity to insect sequences (the lowest E value was 6e 215 ). However, NPC2 paralogs are remarkably divergent even in the same insect species, consistent with rapid diversification. For example, the fruitfly Drosophila melanogaster has 8 NPC2 proteins designated as NPC2a-NPC2h (Genbank accession numbers are NP_608637, NP_650331, NP_649976, NP_649975, NP_731439, NP_651219, NP_651823 and NP_651824, respectively), and the similarity of D. melanogaster NPC2a to other NPC2 proteins ranges from 3e 216 (NPC2b, 50% similarity) to as low as 3e 203 (NPC2g, 37% similarity). It is likely therefore that the insect ancestor of Mr-NPC2a could also have had low similarity to currently known insect NPC2 proteins.
We then looked at the genomic context of the npc2a genes in Metarhizium and C. purpurea to investigate how these fungi may have acquired npc2a gene from insects. The microsyntenies are almost the same in ,40 kb M. robertsii and M. acridum DNA fragments containing the npc2a genes, indicating that the npc2a gene was acquired before the diversification of M. robertsii and M. acridum ,33-43 MYA [22], and no significant genomic rearrangements have happened since. However, little microsynteny is retained between Metarhizium and other fungi including C. purpurea; only one M. robertsii gene (MAA_03818, Genbank accession no.: EFZ01222) adjacent to Mr-npc2a (MAA_03817) has a homolog (CPUR_07480, Genbank accession no.: CCE3355) adjacent to the C. purpurea npc2a gene (CPUR_07479, Genbank accession no.: CCE33554). CPUR_07480 and MAA_03818 have homologs in other fungi but not insects, showing that this gene is vertically inherited. CPUR_07480 and MAA_03818 have LITAF-like zinc ribbon domain, but their functions have not been characterized. Conversely, the microsynteny of the sequence around the C. purpurea npc2a gene (CPUR_07479) and CPUR_07480 is conserved among many fungi including Metarhizium (Fig. 3).

The lipid binding activity of Mr-NPC2a
Mr-NPC2a was expressed in Escherichia coli using the pET-15 system. The recombinant protein was purified to homogeneity and subjected to a lipid binding assay. Mr-NPC2a binds to cholesterol and bacterial lipid A that are absent in fungi, but did not bind to fungal ergosterol (Fig. 4), indicating that its binding specificity was similar to D. melanogaster NPC2 proteins that binds to cholesterol and bacterial lipid A [27], and distinct from yeast yNPC2p that binds to ergosterol and is involved in ergosterol homeostasis [28].

Expression pattern and cellular localization of Mr-NPC2a
We then investigated the biological function of Mr-NPC2a in M. robertsii. RT-PCR (reverse transcription PCR) did not detect Mr-npc2a transcripts in aerial hyphae, conidiating mycelium and conidia collected from PDA plates, hyphae and blastospores collected from SDB (Sabouraud dextrose broth), hyphae and blastospores grown in the in vitro prepared insect hemolymph, and appressoria on locust hindwings. However, Mr-npc2a was transcribed in the hyphal bodies collected from the hemolymph of living insects (Fig. 5A). The expression pattern of Mr-NPC2a was also investigated by observing GFP signal in three randomly selected transformants containing PMr-NPC2a:GFP. No differences in GFP signal at different developmental stages were observed between these transformants, so only data about one of the transformants is shown. GFP fluorescence was only observed in both unicellular (blastospores) and multicellular hyphal bodies after the fungus had penetrated the insect host (Manduca sexta larva) (Carolina, Burlington, NC) cuticle (Fig. 5B). Therefore, both RT-PCR analysis and GFP observation showed that Mr-NPC2a expresses exclusively in the hemolymph of living insects irrespective of whether the hyphal bodies were unicellular (blastospores) or multicellular. This pattern differs from other hemolymph specific promoters such as Mcl1 [29], as these are also active in in vitro prepared hemolymph.
We tried to identify the hemolymph inducers of Mr-npc2a expression, but found sterols, an insect steroid hormone (20-Hydroxyecdysone), oxidative stress, hypoxia stress, osmotic stress did not trigger the expression of Mr-npc2a (Table S4).
In order to investigate the cellular distribution of Mr-NPC2a we placed the fusion protein Mr-NPC2a:GFP under control of the constitutive glyceraldehyde-3-phosphate dehydrogenase (gpd) promoter from Aspergillus nidulans [30]. No GFP was observed in the hyphae of the transformants, but Western blotting analysis detected Mr-NPC2a:GFP in cell lysates though not in liquid culture, indicating that Mr-NPC2a is an intracellular protein as reported for other NPC2s (Fig. 5C) [28]. There are other reports of failure in detecting GFP signal in fungal transformants expressing a hybrid gene, and the reason for the failure is not clear [31].

Mr-NPC2a is essential for membrane integrity of hyphal bodies
M. robertsii rapidly proliferates as hyphal bodies in host hemolymph. Having confirmed that Mr-NPC2a is a cholesterol carrier and specifically expressed in hemolymph, we investigated its role in hyphal body formation. Flipin staining confirmed a large reduction in sterol content of DMr-npc2a hyphal bodies relative to those of wild type and the complemented DMr-npc2a (Fig. 6A), indicating a role in maintaining sterols as the fungus proliferates. When cultured in a complex nutrient medium (PDA), DMr-npc2a, the wild-type and the complemented DMr-npc2a cells had similar levels of cell membrane sterols (Fig. S4).
The nucleic acid stain Sytox is unable to cross intact cell membranes and therefore only stains cells with poor membrane integrity [32]. We found that 762.7% of wild type hyphal bodies, 863.1% of the complemented DMr-npc2a and 5566.5% of DMr-npc2a hyphal bodies were permeable to Sytox (Fig. 6B), confirming that the provision of sterols by Mr-NPC2a has an important role in maintaining the membrane integrity of hyphal bodies. This result  Table S2. doi:10.1371/journal.ppat.1004009.g002 is consistent with DMr-npc2a producing fewer hyphal bodies in hemolymph than the wild type strain.

Expression of Mr-NPC2a increased the virulence of the entomopathogenic fungus Beauveria bassiana
The genome of the well-studied insect-pathogenic fungus Beauveria bassiana (Genbank accession number: ADAH00000000) was searched for Mr-NPC2a homologs using BLASTP and TBLASTN (e-value cutoff 1e 205 ), but no significant hits were  Top row: the expression of Mr-npc2a in 1: aerial hyphae collected from a PDA plate; 2: conidiating mycelium collected from a PDA plate which has hyphae, conidiphores and conidia; 3: conidia collected from a PDA plate; 4: mycelium collected from SDB broth which comprises of hyphae and blastospores; 5: appressoria formed on locust hindwings; 6: hyphae and blastospores grown in the in vitro prepared hemolymph; 7: Unicellular hyphal bodies (blastospores) and multicellular hyphal bodies collected from the hemolymph of living insects which also contains insect hemocytes; 8: insect hemocytes of healthy insects. Bottom row: 1 to 7 represents the expression of the gene gpd in the same samples described in the top row; 8 is the expression of ribosomal protein S3 gene (Genbank accession no. U12708) in the hemocytes of M. sexta. (B) Analysis of expression pattern of Mr-npc2a by following GFP signal in a transformant where GFP expression was driven by the promoter region of Mr-npc2a. GFP fluorescence only occurred in hyphal bodies in the hemolymph of living M. sexta larvae. PMr-npc2a:GFP: a transformant with egfp driven by the promoter region of Mr-npc2a; wild type: the wild type strain of M. robertsii; DIC: differential interference contrast; GFP: GFP fluorescence. (C) Localization of Mr-NPC2a. M. robertsii transformants expressing the fusion protein Mr-NPC2a: GFP driven by the constitutive Pgpd promoter were constructed to investigate the localization of Mr-NPC2a. A representative transformant was grown in SDB for 36 h, and total proteins in mycelium and supernatant were separately prepared for Western blot analysis using rabbit anti-GFP antibody. Total protein was prepared from mycelium of (1) wild type, (2)   identified suggesting that this fungus lacks a Mr-NPC2a homolog. So, we decided to test whether transfer of Mr-NPC2a into B. bassiana can enhance its pathogenicity. Expression of Mr-NPC2a did not alter the growth, conidiation and germination of B. bassiana. Compared to the wild type strain, the LT 50 (time taken to kill 50% of insects) of a transformant expressing Mr-NPC2a (Bb-Mr-NPC2a) was reduced by 22% (Fig. 7). Filipin staining did not detect obvious differences in sterol in the hyphal bodies of B. bassiana in hemolymph, but Sytox staining showed that the transformant expressing Mr-NPC2a had significantly fewer hyphal bodies (561.2%) with impaired cell membranes than did the wild type strain (1061.8%).

Discussion
The virulence of most fungal pathogens of plants and animals is a quantitative attribute, and thus many different genes and strategies are used simultaneously to cause disease [33]. Through evolution, fungi have acquired new virulence mechanisms, and increases in pathogen virulence have likely therefore evolved in a stepwise manner [34]. The important role of HGT in the evolution of fungal virulence is being illuminated by comparative genomics, but in only a few cases has the importance of the transferred genes been functionally demonstrated with gene knockouts, and no reports describe functional characterization of genes transferred from an animal to a eukaryotic pathogen. Lack of functional characterization is a major bottleneck in understanding the contribution of HGT to pathogen virulence. In this study, we describe how the insect pathogenic fungus Metarhizium may have acquired a sterol carrier gene Mr-npc2a from its insect hosts by HGT. This sterol carrier is involved in maintaining cell membrane sterols and thus membrane integrity of hyphal bodies in insect hemolymph. It thus helps Metarhizium adapt to the hemolymph and facilitates colonization of the host. A feature of Mr-npc2a is its specificity for the hemolymph of living insects. The promoter will thus be useful for driving expression of insecticidal proteins, and for manipulating gene expression in hemolymph, via antisense or overexpression constructs, to test whether candidate proteins play critical roles.
Ergosterol is an important constituent of fungal membrane lipids, similar to animal cholesterol, and modulates the fluidity, permeability and thickness of the membrane [35]. Under aerobic conditions, fungi synthesize ergosterol de novo and this process is oxygen dependent [36]. Under anaerobic conditions, ergosterol biosynthesis is suppressed and fungi need to take up sterols from the environment [37]. Metarhizium infects insects by direct penetration of the cuticle and entry into the haemocoel where the fungus kills insects with a combination of toxins and invasive growth by hyphal bodies [18]. However, the insect haemocoel is a hypoxic or even oxygen free environment, necessitating M. robertsii takes up sterols for the rapid proliferation of hyphal bodies. The insect hemolymph is the only environment where the fungal hyphal bodies can acquire sterols, and cholesterol is most likely to be exploited because it is the dominant insect sterol [38]. The bulk membrane function of the fungal ergosterol can be provided by structurally related sterols, including cholesterol [39]. Fungi can convert cholesterol into ergosterol or incorporate cholesterol directly into plasma membranes [28,40].
Cholesterol is insoluble and insects use low density lipoproteins (LDL) to transport cholesterol through hemolymph to target tissues where they bind to LDL receptors. Metarhizium has a variety of genes with predicted functions that mimic those in insects including an LDL receptor (MAA_08841, Genbank accession no.:  [41]. The mechanisms of intracellular cholesterol transport are largely unknown, however, in mammals, cholesterol is exported to organelles by an NPC team that comprises an NPC1 (a lysosomal membrane protein) and NPC2 (a soluble protein in the lysosomal lumen). These mechanisms seem to be conserved in yeast and other fungi [28,42]. M. robertsii has 1 NPC1-like protein (MAA_02001, Genbank accession no.: EFZ02419) as well as the 3 NPC2 proteins described in this study. Mr-NPC2b and Mr-NPC2c are typical fungal NPC2-like proteins, and their homolog in yeast (yNPC2p) is involved in ergosterol homeostasis [28]. The insect-like Mr-NPC2a is specifically expressed in insect hemolymph and is involved in maintaining cell membrane integrity only after the fungus enters the haemocoel. Although the disruption of Mr-npc2a greatly reduced the number of hyphal bodies in the hemolymph of living insects, i.e. suppressed growth, DMr-npc2a was still able to kill insects, suggesting that unidentified components can partially substitute for Mr-NPC2a when M. robertsii is in the hemolymph. The competition for cholesterol from Metarhizium might reduce the amount of sterols available to insects, which could alter the development of insects because insects cannot synthesize sterols and thus have dietary requirement for sterols. However, we did not see significant differences in development (pupation) between G. mellonella larva infected by DMr-npc2a and the wild type strain.
The existence of insect NPC2a homologs both in Metarhizium and the ergot fungus C. purpurea raises the question: when and how did these fungi acquire NPC2a from insects? There are four possible explanations. First, ancestral fungi had NPC2a proteins that were homologs of insect NPC2 proteins, but during evolution they were only retained by Metarhizium spp. and C. purpurea. This is unlikely as it would require independent loss of npc2a genes in all other fungal lineages. Secondly, Metarhizium and C. purpurea belong to the family Clavicipitaceae, and it is possible HGT occurred into the common ancestor of Metarhizium and C. purpurea. But microsynteny of sequence around C. purpurea npc2a is conserved with many fungi including Metarhizium. spp, whereas the microsynteny of the sequence around Metarhizium npc2a (Mr-npc2a) showed no conservation with other fungi including C. purpurea. So, if the ancestor of Metarhizium and C. purpurea acquired npc2a gene from an insect, this npc2a gene had moved from its ancient position to its current area during speciation of Metarhizium. spp. There is no evidence for Metarhizium npc2a (Mr-npc2a) having ever been within a transposon so genome rearrangement rather than transposition would be required for this move. Around Mr-npc2a (MAA_03817), only the gene MAA_03815 (Genbank accession no.: EFZ01219) is possibly involved in sterol metabolism which encodes C-4 methylsterol oxidase-like protein (Table S5), but M. acridum's C-4 methylsterol oxidase-like gene (MAC_09281, Genbank accession no.: EFY84691) does not cluster with its npc2a gene (MAC_00943 Genbank accession no.: EFY93160). Therefore, no obvious evolutionary force is seen that selected the genome rearrangement in Metarhizium genus. Thirdly, ancestors of C. purpurea and Metarhizium spp. may have independently acquired npc2a gene from ancestral insects. It is conceptually easy to understand an insect pathogenic fungus acquiring a gene from its host by HGT.
In nature, C. purpurea also has physical contact with insects, as it produces ergot honeydew that attracts insects and these spread its conidia [25]. However, the gene adjacent to M. robertsii npc2a (Mr-npc2a) and that to C. purpurea npc2a are homologs with related genes in other fungi but not in insects. It is improbable that insect npc2a genes independently inserted beside the same gene in M. robertsii and C. purpurea. The final possibility is that an ancestor of C. purpurea or Metarhizium spp. acquired npc2a which was inserted beside the gene MAA_03818 in Metarhizium to become MAA_03817 (M. robertsii npc2a) or beside the gene CPUR_07480 in C. purpurea to become CPUR_07479 (C. purpurea npc2a). C. purpurea then obtained the two genes (the genes MAA_03818 and MAA_03817) from Metarhizium or vice versa. Ecological niche overlap allows HGT of gene clusters between fungi [8], making this the most likely possibility. The genetic distances between the generalist M. robertsii (can kill many kinds of insects) and the specialist M. acridum (specific for grasshoppers) is greater than that between most other species in the Metarhizium genus [43]. Since the ancestor of Metarhizium may have acquired npc2a gene from an insect before speciation and both genetically distant M. robertsii and M. acridum retain this gene, it is expected that most of other Metarhizium species still have this gene. Therefore, based on currently available information, more Metarhizium than Claviceps species have npc2a genes, so it is more likely that C. purpurea was the recipient. However, more Claviceps genomes will be needed to confirm the direction of HGT.

Fungal and bacterial strains
M. robertsii ARSEF2575 and B. bassiana ARSEF252 were cultured as previously described [18,44]. Escherichia coli DH5a was used for plasmid construction. A. tumefaciens AGL1 was used for fungal transformation as described [45].

Gene cloning and disruption
The flanking sequences of T-DNA in mutants generated by the insertion of T-DNA into the genome of M. robertsii were cloned by Y-shaped adaptor-dependent extension as described [46]. The primers employed in this study and their usages are given in Table  S6. PCR products were cloned into pGEM-T Easy (Promega, USA) for sequencing.
To construct the Mr-npc2a disruption plasmid, the 59-end and 39-end of Mr-npc2a were cloned by PCR and inserted into the XbaI and SpeI sites, respectively, of the plasmid pFBARGFP [18] to form the disruption plasmid pFBARGFP-dMr-npc2a. The disruption mutant (DMr-npc2a) was obtained utilizing A. tumefaciens-mediated fungal transformation. To complement DMr-npc2a, the genomic sequence of Mr-npc2a was cloned, inserted into the XbaI site of pSURGFP [43], and transformed into DMr-npc2a as described [47].

Phylogenetic analyses and tree topology tests
M. robertsii Mr-NPC2a protein sequence was used as the query for BLASTP searches against the NCBI non-redundant database (e-value cutoff 1e 205 ) to identify and retrieve its homologs. The selected homologous proteins of representative taxa were aligned using MUSCLE 3.7 [48], with a maximum number of iterations of 16. Poorly aligned positions and divergent regions of the alignment were removed using GBlocks0.91b [49], with half the gapped positions allowed, the minimum number of sequences for a conserved and a flank position set to 50% of the number of taxa plus one, the maximum of contiguous nonconserved positions set to 16, and the minimum length of a block set to 4. Phylogenetic analyses were performed with three approaches: maximum likelihood (ML), bayesian inference (BI) and distance-based neighbor-joining (NJ). ML phylogenies were constructed by PhyML 3.1 [50] using the best-fit evolutionary model as suggested by ProtTest 3.2 [51], in which case, a discrete gamma-distribution model with four rate-categories plus invariant positions was used. The gamma parameter and proportion of invariant sites were estimated from the data. Branch support values were obtained by 100 bootstrap pseudo-replicates. Bayesian reconstructions were performed with MrBayes 3.1.2 [52] under the WAG+C4+I model of amino acid substitution. The Markov chain Monte Carlo searches were run for 1,000,000 generations, sampling the Markov chains every 1000 generations; the first 250 trees were discarded as 'burn-in'. In addition, we also employed distance-based neighborjoining (NJ) trees, which were constructed by using the program Neighbor in MEGA5 [53]. Bootstrap support values were obtained by generating 1,000 pseudo-replicates.
We statistically tested the significance of the topology of the obtained tree (Fig. 2) by phylogenetic analysis in comparison with alternative trees in which fungal NPC2a proteins were constrained to the vertebrate clade to form monophyletic groups using the program CONSEL [24]. The value of site-wise likelihood of every constrained topology was calculated by PhyML [50].

Sterol binding assay of Mr-NPC2a
The DNA fragment encoding Mr-NPC2a (signal peptide excluded) was cloned with RT-PCR using total RNA extracted from hyphal bodies collected from the hemolymph of living insects as described below. The sequences of the primers are presented in Table S6. PCR product was purified and cloned into pGEM-T vector (Promega, Madison, WI, USA) for sequence verification. The DNA fragment was then subcloned into EcoRI/NotI restriction sites in the expression vector pET32a (EMD Millipore Corporation, Billerica, MA, USA) to form the pET32a-MrNPC2a plasmid that was then transformed into E. coli strain BL21 (DE3) (EMD Millipore Corporation, Billerica, MA, USA). Isopropyl-1thio-b-D-galactopyranosid (1 mM) was used to induce recombinant protein expression. Recombinant Mr-NPC2a was purified with HisTALON Superflow Cartridges (Clontech, Mountain View, CA, USA), concentrated and dialyzed with Tris buffer (50 mM Tris-HCl, 50 mM NaCl, pH 8.0) at 4uC using Amicon Ultra centrifugal filter unit (EMD Millipore Corporation, Billerica, MA, USA). Control protein was expressed and purified from BL21 (DE3) cells transformed with pET32a empty vector with the same methods described above.
The ligand-binding assay was performed as described with some modifications [27]. Since the D. melangaster NPC2a can bind to bacterial Lipid A and cholesterol [27], both of them were used as ligands to compare binding specificity of Mr-NPC2a with that of the insect NPC2a. Ergosterol was also tested to look at the binding ability of Mr-NPC2a to the fungal sterol. Lipid A monophosphoryl from E. coli F583 (Rd mutant) and cholesterol were purchased from Sigma Aldrich (MO, USA), and ergosterol was from Santa Cruz Biotechnology (Texas, USA). Wells of flat-bottom 96-well plates (BD falcon, NJ, USA) were coated with 2 or 10 mg/well of each ligand. Control wells with no ligand were coated with the same amount of chloroform/water. The plates were then air dried overnight at room temperature, heated at 60uC for 30 min and then blocked with 200 mL/well of 1 mg/mL BSA in Tris buffer (50 mM Tris-HCl, 50 mM NaCl, pH 8.0) at 37uC for 2 h. Wells were washed with the Tris buffer (four times, each for 5 min) and 500 nM of purified Mr-NPC2a fusion protein or vector protein diluted in binding buffer (the Tris buffer with 0.1 mg/mL BSA) were added to each well of the coated plates (50 mL/well) and incubated at room temperature for 3 h. Plates were then washed and mouse monoclonal anti-polyHistidine antibody was added (100 mL/well) and incubated at 37uC for 2 h. Plates were washed again and alkaline phosphatase-conjugated goat anti-mouse IgG (Promega, 1:2000 dilution in binding buffer) was added (100 mL/ well) and incubated at 37uC for 2 h. Alkaline phosphatase substrate liquid (Sigma Aldrich MO, USA) was then added to each wells and incubated for 15 min at room temperature. Plates were shaken for 15 s and the absorbance at 405 nm of each well was determined using a plate reader (FilterMax F3, Molecular Devices, CA, USA). The absorbance from control wells was subtracted and specific binding was calculated by subtracting absorbance from vector protein from the total binding of NPC2 protein. Binding assays were repeated 3 times with 3 repeated wells.

Expression pattern of Mr-npc2a
The expression pattern of Mr-npc2a was investigated by RT-PCR and by following GFP fluorescence in transformants expressing GFP driven by the promoter region (2,203 bp) of Mr-npc2a. The promoter region of the Mr-npc2a and the ORF of egfp were both cloned by PCR using primers described in Table S6. The resultant egfp product was digested with EcoRV and XhoI, and inserted into the corresponding sites of pBARGPE1 [30] to form pGFP. The Mr-npc2a promoter was digested with EcoRI and SmaI, and inserted into EcoRI and EcoRV of pGFP to form PMr-NPC2a:GFP. The egfp cassette was then mobilized into Ppk2-bar [21] to form pPMr-NPC2a:GFP. This Ti plasmid was then transformed into the wild type M. robertsii mediated by A. tumefaciens [45]. For RT-PCR analysis, the gpd gene was used as a reference gene [54]. Primers for Mr-npc2a and gpd are presented in Table S6. Total RNA was isolated using the Plant RNeasy Kit (Qiagen USA). First-strand cDNA was synthesized using Revert Aid First Strand cDNA Synthesis Kit (Fermentas, USA). PCR was conducted using DreamTaq DNA polymerase (Thermo Scientific, USA).
Expression of Mr-npc2a was examined during saprophytic growth, infection and in several possible inducing conditions (Table S4). For saprophytic growth, the fungus was grown on PDA, aerial hyphae, conidiating mycelium and conidia were sampled as previously described [23] for GFP observation or RNA preparation. Blastospores were obtained by growing the fungus in nutrition rich medium SDB for 10 d and in the in vitro prepared hemolymph for 2 d. GFP fluorescence in blastospores and hyphae was examined directly under microscopy. Blastospores and hyphae were combined for RNA preparation.
The expression of Mr-npc2a during infection was examined in appressoria on insect cuticles (the hindwings of L. migratoria locusts) and hyphal bodies in the hemolymph of living insects. Appressoria were induced on locust hindwings as previously described [55]. To obtain hyphal bodies in insect hemolymph, 1610 5 conidia were injected into a 5 th instar M. sexta larvae. After 20 h at 27uC, the insect was bled and GFP signal in the fungus in the hemolymph was checked by microscopy. The hemolymph with insect hemocytes and fungal hyphal bodies was subjected to RNA preparation. RNA of the hemolymph of healthy insects was prepared as a control for RT-PCR analysis.
In order to test the effect of sterols on the Mr-npc2a expression, the wild type strain (for RT-PCR analysis) or the transformant with PMr-npc2a:GFP (for GFP fluorescence observation) were grown in SDB for 24 h. The mycelium were then filtered, washed 3 times with sterile water and cultured in the minimal medium M100 [21] supplemented with the insect steroid hormone 20-Hydroxyecdysone (Sigma, USA), cholesterol (Sigma, USA) or ergosterol (Sigma, USA) (Table S4). To detect the effect of stresses on Mr-npc2a expression, the mycelium from the 24 h culture in SDB were cultured in the minimal medium M100 supplemented with different agents to generate stresses described in Table S4. For all treatments, GFP fluorescence in mycelium was examined and RNA was also prepared for RT-PCR.

Investigating localization of Mr-NPC2a
To investigate the localization of Mr-NPC2a in M. robertsii, we constructed the fusion protein Mr-NPC2a: GFP. The coding sequence of Mr-NPC2a (stop codon was excluded) was cloned by PCR, and inserted into the EcoRV site of the plasmid pGFP described above, resulting in pMr-NPC2a:GFP. Since Mr-NPC2a is only expressed in hyphal bodies produced in the hemolymph of living insects, it is hard to determine the localization of this protein when its native promoter is used to drive the expression of NPC2a:GFP. Therefore, the expression of the fusion gene was driven by the constitutive promoter Pgpd from A. nidulans in pGFP [30]. The Mr-NPC2a:GFP cassette was removed from pMr-NPC2a:GFP and inserted into Ppk2-bar to produce Ppk2-gpd:Mr-NPC2a:GFP that was then transformed into wild type M. robertsii mediated by A. tumefaciens to produce transformant Mr-NPC2a:GFP.
GFP observation and Western blot analysis were used to test the localization of Mr-NPC2a:GFP. The transformant Mr-NPC2a:GFP was inoculated in SDB broth (10 6 spores/mL) and grown for 36 h at 27uC with constant shaking (200 rpm). Mycelium was collected by filtration and ground into powder for protein preparation. The protein in the filtrant was precipitated with 80% (NH 4 ) 2 SO 4 . The salt was then removed by centrifugation with Microcon filter with a 3 kDa cutoff (Millipore, USA). Total protein (15 mg) was subjected for Western blot analysis. Rabbit anti-GFP antibody (Proteintech Group Inc. USA) was used for Western blot analysis.

Expression of Mr-NPC2a in B. bassiana
The ORF of Mr-npc2a was cloned by PCR and inserted into BamHI and EcoRV sites downstream of the Pgpd promoter in the plasmid pBARGPE1 [30]. The Mr-npc2a expression cassette was then moved into pFBARGFP to form the expression plasmid pMr-NPC2a that was transformed into B. bassiana using A. tumefaciens.

Sytox staining of hyphal bodies collected from the hemolymph of living insects
Sytox Green nucleic acid staining is widely used to assess the integrity of eukaryotic and prokaryotic cell membranes [32]. Sytox Green can only enter cells if their cell membrane integrity is compromised, in which case it stains nucleic acids, resulting in a green fluorescence emission (absorption and emission maxima at 502 and 523 nm, respectively). To prepare hyphal bodies for Sytox staining, 50 mL of conidial suspension (10 7 conidia/mL) was injected into a fifth instar larva of M. sexta. After 20 h incubation at 27uC, the insect was bled and the hemolymph was collected into a 1.5 ml centrifuge tube that was then placed on ice. The hemolymph (100 mL) was then quickly mixed with 5 mL of sytox staining solution (10 mM) by pipetting up and down. The mixture was then immediately loaded on a glass slide and covered with a cover slip. This set was subsequently placed on wet paper towel in a small container and incubated at room temperature for 10 min. Stained nuclei were observed under microscopy.

Flipin staining
Hyphal bodies were prepared from the hemolymph of M. sexta as described above and stained with 5 mM Filipin (Sigma Aldrich, MO, USA) in ACES buffer as described [28].

Quantification of hyphal bodies in insect hemolymph
A fungal spore suspension (1610 7 spores/mL) was topically applied on the wax worm larvae (G. mellonella). At 12 h intervals after inoculation, ten larvae per treatment were surface sterilized with 1% bleach and individually bled as described [56]. Five microliter aliquots of blood from each larva were mixed with 95 mL of sterile water and spread onto Metarhizium selective medium for colony counts [56]. The number of colony forming units resulting from unicellular hyphal bodies (blastospores) or multicellular hyphal bodies is used to show the number of hyphal bodies.

Bioassay
Fungal virulence was assayed against wax worm (G.mellonella) as previously described [56]. The worms were maintained as manufacture's instruction (Pet Solutions, OH USA). The SPSS program was used to calculate the LT 50 for each strain. There were 40 insects for each strain and the experiment was repeated 3 times. Mr-npc2a ORF is deleted without T-DNA inserted into other parts of the genome. Middle: transformants are resistant to the herbicide with GFP signal. In these transformants, T-DNA is randomly inserted in the genome and Mr-npc2a t is not disrupted. Bottom: transformants are resistant to the herbicide with GFP signal. In these transformants, Mr-npc2a ORF is deleted, but T-DNA inserts into other parts in the genome which could disrupts other genes. Transformants with GFP signal (middle and bottom panels) will be discarded. (B) Further confirmation of the deletion of Mr-npc2a ORF by PCR in the mutants with herbicides resistance and without GFP signal which were obtained from above screening. 1 and 2 are two mutants, and C is the wild type strain. Top panel: PCR conducted with the primers Bar-UP and CF2; Bottom panel: PCR conducted using primers CF1 and CF2. The positions of the primers are shown in the left panel of (A). The PCR data and the screening data [right panel in (A)] selected out the mutants where only Mr-npc2a ORF is deleted and no other genes are disrupted. (C) PCR confirmation of the complementation of DMr-npc2a. A genomic DNA fragment of Mr-npc2a including the promoter region, ORF and termination region was cloned by PCR using primers Mr-npc2a-5 and Mr-npc2a-3 (Table S6) and inserted into pPK2-SUR-gfp to form Ppk2-SUR-GFP-Mr-npc2a (Top panel) that was then transferred in to DMr-npc2a. Bottom panel: Confirmation of the complementation of DMr-npc2a by PCR using the primers F (Mr-npc2aORF-5) and R (Mr-npc2aORF-3) that were used to amplify the deleted region (ORF). 1 to 6: six different transformants with DMr-npc2a complemented; C: wild type strain; M: DMr-npc2a. The primers are described in Table S6. (TIF) Figure S3 Phylogeny of Mr-NPC2b and Mr-NPC2c and their homologs. The Bayesian inference tree is shown unrooted. Numbers at nodes represent Bayesian posterior probabilities. The scale bar corresponds to the estimated number of amino acid substitutions per site. This tree shows that the phylogenetic relationship between these proteins is consistent with previously established species phylogenies, demonstrating vertical inheritance. (TIF) Figure S4 Filipin staining of M. robertsii conidia collected from a PDA plate (Potato dextrose agar). Left panels: differential interference contrast images of conidia; Right panels: fluorescence due to Filipin staining of ergosterol in the cell membrane of the same conidia shown in the left panels. DMr-npc2a: A M. robertsii strain with Mr-npc2a deleted; Wild type: the wild type M. robertsii strain; Comp: the complemented DMr-npc2a.

(TIF)
Table S1 The number of hyphal bodies (hyphal bodies/mL) in the hemolymph of living wax worms. (DOCX)

Table S2
GenBank accession numbers used for phylogenetic reconstruction in Figure 2. (DOCX)

Table S3
Statistics of comparison of topologies of constrained trees with the tree obtained by phylogenetic analyses (nonconstrained tree) (Fig. 2). (DOCX)