Extensive Differences in Antifungal Immune Response in Two Drosophila Species Revealed by Comparative Transcriptome Analysis

The innate immune system of Drosophila is activated by ingestion of microorganisms. D. melanogaster breeds on fruits fermented by Saccharomyces cerevisiae, whereas D. virilis breeds on slime flux and decaying bark of tree housing a variety of bacteria, yeasts, and molds. In this study, it is shown that D. virilis has a higher resistance to oral infection of a species of filamentous fungi belonging to the genus Penicillium compared to D. melanogaster. In response to the fungal infection, a transcriptome profile of immune-related genes was considerably different between D. melanogaster and D. virilis: the genes encoding antifungal peptides, Drosomycin and Metchnikowin, were highly expressed in D. melanogaster whereas, the genes encoding Diptericin and Defensin were highly expressed in D. virilis. On the other hand, the immune-induced molecule (IM) genes showed contrary expression patterns between the two species: they were induced by the fungal infection in D. melanogaster but tended to be suppressed in D. virilis. Our transcriptome analysis also showed newly predicted immune-related genes in D. virilis. These results suggest that the innate immune system has been extensively differentiated during the evolution of these Drosophila species.


Introduction
In natural environments, Drosophila species feed and breed on fermenting fruits, slime fluxes on decaying parts of tree, and so forth, where biochemical processes of bacteria and fungi are extremely active [1][2][3]. Therefore, Drosophila species are exposed to a huge number of microorganisms throughout their developmental stages. Feeding on decaying or fermented materials results in the ingestion of a wide variety of microorganisms in their digestive organs. Recent studies on larval immune response of D. melanogaster to oral infection of bacteria and fungi showed that the fat body mediated systemic immune response including antimicrobial peptide (AMP) production was triggered by infections of Gram-negative bacterial species such as Pseudomonas entomophila and Erwinia carotovora carotovora 15 (Ecc15) and of a dimorphic fungal species, Candida albicans [4][5][6].
AMPs are cationic small secretory peptides that exhibit a wide range of activities against bacteria, fungi, and/or viruses, playing an essential role in the innate immune system of Drosophila [7]. To date, seven AMP families, that is, Attacin, Cecropin, Defensin, Diptericin, Drosocin, Drosomycin, and Metchnikowin, have been identified in Drosophila melanogaster [7]. According to Sackton et al. (2007), it was indicated by their sequence analysis of the 12 Drosophila genomes that only the species belonging to the melanogaster species group of the subgenus sophophora had Drosomycin genes [8]. Drosomycin is known to be a major antifungal peptide [9][10][11]. This suggests that antifungal immune response varies among different Drosophila species and attacks from different bacteria and/or fungi might have produced different immune responses in Drosophila. Therefore, it is hypothesized that the differences in the environmental factors caused the difference in the immune system. For instance, D. virilis feeds and breeds on slime flux and decaying bark of trees, which are infected by various bacteria, yeasts and molds. Indeed, many yeasts, other than Saccharomyces cerevisiae and filamentous fungi, such as Xanthophyllomyces dendrorhous, Cryptococcus spp., and Fusarium spp., have been isolated from slime flux and decaying wood [14,15], whereas S. cerevisiae solely ferments various fruits, which D. melanogaster thrives on [1][2][3]. From this difference in the microbial community in host materials of D. virilis and D. melanogaster, it is conceivable that D. virilis is exposed to a wider variety of fungi and therefore D. virilis has a higher resistance to fungi compared to D. melanogaster. To test this hypothesis, we examined the immune response of D. virilis and D. melanogaster to a fungus species belonging to the genus Penicillium. Since Penicillium species are commonly found in both slime flux and rotting fruits [16,17], both D. virilis and D. melanogaster likely have high risk of Penicillium infection during all their developmental stages. To measure resistance of D. virilis and D. melanogaster to the fungal infection, adult flies of these species were reared on the culture medium that Penicillium fungi grew. The results showed that D. virilis adult flies survived more than two times longer than D. melanogaster flies (Figure 1), suggesting that D. virilis has a higher resistance to Penicillium infection. This higher antifungal activity without having Drosomycin motivated us to investigate the immune system of D. virilis.
In this study, to clarify the immune mechanism responsible for the higher antifungal resistance of D. virilis, larval immune response to the fungal infection between D. virilis and D. melanogaster were compared by means of comparative transcriptome analyses. Using a Roche 454 GS Junior sequencer, we examined the transcriptome of fat body and salivary gland of 3rd-inster larvae with and without infection of a Penicillium species. Genes showing different expression patterns in response to the fungal infection between D. virilis and D. melanogaster were extracted and compared. These genes included the genes encoding AMPs and "immune-induced molecule (IM). " Extensive differences were observed in the expression pattern of already known AMP and IM genes between D. virilis and D. melanogaster. Additionally, two potential AMP genes were newly identified from function-unidentified genes. Furthermore, three novel putative immune-related genes were identified: the products of them had a homology to an IM, Ras-like GTP-binding protein Rho1 involved in many signaling pathways and Ficolin-2 binding to a cell wall component of bacteria and fungi, respectively.

Measurement of Antifungal
Resistance. Twenty to twentyfive adult flies 1 day after eclosion were reared at 25 ∘ C on a cornmeal-malt medium (50 g cornmeal, 50 g malt powder, 40 g dried brewer's yeast, 50 g sucrose, 5 mL propionic acid and 5 g agar in 1 liter water) with and without Penicillium fungi. The medium containing Penicillium fungi was prepared by inoculating a small amount of spores of a Penicillium species (identified by its nucleotide sequence of 18S RNA  gene) onto the cornmeal-malt medium and incubated at 20 ∘ C for a week or more until the surface was completely covered by the growing fungi. After the flies were transferred onto the medium with or without fungi, the number of flies alive was counted every day. To measure the resistance to the infection of the Penicillium species, the 50% lethal time (LT50) was estimated by the generalized linear method implemented in R version 2.15.2 software [18]. These processes were independently replicated three times.

Induction of Gene Expression by Fungal Infection.
A small amount of Penicillium spores were inoculated and cultured on a Sabouraud dextrose agar (SDA) medium (10 g peptone, 40 g dextrose, and 15 g agar in 1 liter water) at 20 ∘ C for several days until the fungi grew on to cover the surface of the medium. To prepare the fungus-infected larvae, twenty 3rdinstar larvae of D. virilis or D. melanogaster were reared on the fungus-covered SDA medium for 12 hours at 20 ∘ C. The induction of AMP genes is usually detected in three hours after the infection and continued at least 24 hours at 25 ∘ C [4,6]. However, we reared the larvae at 20 ∘ C to postpone their pupation. The responses to the fungal infection was confirmed by the raised expression level of the Metchnikowin gene (known antifungal AMP gene) measured by RT-PCR and only the induction confirmed samples were used for the transcriptome sequencing described in the next section. As the control, the naïve larvae were prepared by rearing with the same condition on fungus-free SDA medium.

Transcriptome Sequencing.
We analyzed transcriptome of larval fat body and salivary grand. This is because all AMPs were shown to be expressed in fat body and a major antifungal International Journal of Genomics 3 AMP, Drosomycin, was highly expressed in larval salivary gland in D. melanogaster [19]. Larval fat bodies and salivary glands dissected from twenty fungus-infected or naïve 3rdinstar larvae were pooled and the total RNA was extracted from these fat bodies and salivary glands by acid-guanidium phenol-chloroform (AGPC) method [20]. Then, mRNA was isolated by using Dynabeads mRNA purification kit (Invitrogen) according to the supplier's instruction. The complementary DNA (cDNA) library was constructed according to the Roche GS Junior cDNA rapid library preparation protocol with a modification to keep short molecules expected for AMP genes. The double-stranded cDNA was synthesized by using cDNA synthesis system (Roche Diagnostics) with random hexamer primers. The resultant cDNA was purified by using AMPure XP kit (Agencourt) and the end-polished cDNA fragments were ligated with the FAM-labeled RL adaptor included in Lib-L GS FLX Titanium Rapid Library Preparation kit (Roche Diagnostics). The adaptor-ligated cDNA was then purified by using Agencourt AMPure XP system and finally eluted in 50 L TE buffer. The cDNA solution was then concentrated by extracting with the equal volume of 2-butanol twice and subsequently with diethyl ether to remove the residual 2-butanol. Instead of the sizing procedure described in the standard protocol, we conducted 2% agarose-gel electrophoresis, excised the gel section containing 200 bp to 1 kb DNA fragments, and extracted the cDNA using High Pure PCR Clean-up kit (Roche diagnostics). The quality and quantity of the cDNA were evaluated by using QuantiFluor-P Handheld Fluorometer (Promega) and Agilent 2100 Bioanalyzer High Sensitivity DNA kit (Agilent Technologies). The pyrosequencing was conducted by using a 454 GS Junior sequencer after the emulsion PCR according to the manufacturer's instructions (Roche diagnostics). software, we first queried against the CDS database and the reads that did not hit were subsequently queried against gene and transcript databases (Figure 2(a)). Finally, the reads that did not hit any target were used for further analyses to search for novel immune-related genes as explained later in Section 2.6. For the genes identified in the D. virilis genome, most of them have different names from their orthologues in the D. melanogaster genome. In this study, however, we used the gene names of D. melanogaster for both species for the ease of comparison between species. The correspondence of gene ID between the two species was according to the 12 Drosophila genome analyses (ftp://ftp.flybase.net/genomes/12 species analysis/clark eisen/homology/) [23]. For genes that have multiple IDs corresponding to multiple copies in either or both species, one-to-one correspondence of homologue between the two species was determined by TBLASTN search with the translated protein sequence of D. virilis gene as the query against the D. melanogaster CDS database. Whether a gene is immune-related or not was determined by referring to the list of Drosophila immune-related genes [8].

Gene Prediction for
The D. virilis genes of unknown function, which did not have homologue in the D. melanogaster genome, were further BLAST-searched for their homologues in other organisms' genomes (http://blast.ncbi.nlm.nih.gov/) [21]. In this homology search, only the genes, for which the number of reads was significantly different between fungus infected and naïve larvae, were used. For the genes that did not hit any homologue in any organism (D. virilis-specific genes), their functions were predicted by using domain and motif search programs available in NCBI Conserved Domain Database (http://www.ncbi.nlm.nih.gov/Structure/cdd/cdd.shtml) and Pfam (http://pfam.sanger.ac.uk/) (Figure 2(b)). When any conserved domain or motif was not predicted, the presence of signal peptide was predicted by using SignalP (v4.0) [24] and ProP (v1.0) [25] programs as a criterion to consider the possibility of antimicrobial peptide. For the candidates with putative signal peptide, the molecular weight, net charge, and structural features were computed by using JEMBOSS (v1.5) program [26]. Finally, from the amino acid sequence of putative mature peptide after removal of the putative signal peptide, the possibility of antimicrobial peptide was examined by AMP prediction web programs, AntiBP2 [27], CAMP [28], and AMPA [29].

Estimation of Gene Expression Level.
To estimate the expression level of each gene, the total number of reads to hit the gene in the BLAST search was counted (Figure 2(b)). To calibrate the difference in transcript length among different genes, the number of reads counted was then standardized to be the number of reads per site per million reads (RPSM) as follows: RPSM = ( number of reads/total number of reads transcript length ) × 1,000,000. (1) We further normalized RPSM to take the difference in total gene expression level between the samples into account and computed trimmed Mean of values (TMM) [30], using TCC package implemented in R version 2.15.2 software [18,31]. For each gene, the TMM for the fungus infected larvae was compared to that for the control naïve larvae to quantify To test the statistical significance of the induction, the difference in the number of actual reads was compared between the fungus infected and naïve larvae. In this test, RpL32 and GAPDH genes were used as endogenous control genes. Although actin was also a well-known endogenous control gene, actin was reported to play an important role in phagocytosis against fungi in Drosophila S2 cell [32] and that the expression of an actin gene (Act42A) of D. melanogaster 3rd-instar larvae was induced by Saccharomyces cerevisiae contained in the culture medium [33]. Indeed, the expression of D. melanogaster Act42A gene was not detected in the control naïve larvae but in the fungus infected larvae (the number of reads was 6 and TMM = 0.0619). Therefore, only RpL32 and GAPDH genes were used as the endogenous control genes in this study. Since the homogeneity of the International Journal of Genomics 5 numbers of reads for the two genes between the fungus infected and the naïve larvae was statistically supported ( = 0.14 in D. virilis and = 0.51 in D. melanogaster by Fisher's exact test, Supplementary Table 1 available online at http://dx.doi.org/10.1155/2013/542139), the total number of reads derived from the two genes was used as the number of reads for the endogenous control genes. Finally, the difference in the number of reads between the fungus infected larvae and the naïve larvae was tested on the 2 × 2 contingency table with the numbers for the endogenous control genes by Pearson's chi-square test or Fisher's exact test dependent on whether the minimum number of reads was five or more or not.

Prediction of New Immune-Related Genes in D. virilis.
The pyrosequencing reads which were derived from the fungus infected D. virilis but not mapped to any known gene were subject to predicting a new gene (Figure 2(c)). These pyrosequencing reads were mapped to the D. virilis genome sequence by Newbler GS reference mapper software (Roche Diagnostics) with the default parameter setting designated for CDS sequences to obtain continuous transcript sequences. Since the median length (192 bp) of the obtained contigs was similar to that (230 bp) of 3 -UTR of D. melanogaster [34], many contigs might not include protein coding region at all. Therefore, for each contig, the corresponding genome sequence plus 250 bp each of its upstream and downstream flanking regions were extracted to build a query sequence to search for new gene. All the query sequences obtained were subjected to BLASTX search against Swissprot protein database downloaded from the Uniprot web site (http://www.uniprot.org/downloads) with the condition of e-value ≤ 1 − 05. For the identified putative genes, the difference in the number of reads was statistically tested between the fungus infected and the naïve larvae in the same way as that for the known genes described above and if the number of reads was significantly different, then the gene ontology was analyzed by STRAP software (v1.1.0.0) [35].

Difference in Antifungal Resistance between D. virilis and D. melanogaster.
To compare antifungal resistance between D. virilis and D. melanogaster, adult flies of these species were reared on a culture medium harboring Penicillium fungi and their survival time was measured. The results showed that the D. virilis flies survived more than two times longer than the D. melanogaster flies did ( Figure 1); the average 50% lethal times (LT50) of D. virilis and D. melanogaster flies were 6.04 days and 1.75 days, respectively, whereas their survival time on the normal culture medium without fungi was much longer (LT50 ≫ 10). This suggests that D. virilis has a higher resistance to the infection of the Penicillium species than D. melanogaster at the adult stage.

Transcriptome Analysis Summary.
Many AMP genes encode relatively short peptides less than 100 amino acids long. Therefore, to avoid the loss of sequences derived from such  Table 1).
After removing the reads derived from mitochondrial genes and rRNA genes, the total numbers of the remaining reads were 77,558 and 90,836 for the fungus infected and naïve D. virilis larvae, respectively, and 65,670 and 48,474 for the fungus infected and naïve D. melanogaster larvae, respectively. They were thought to be derived from mRNA transcribed from nuclear protein-coding genes. For 55,358 and 62,110 out of the 77,558 and 90,836 reads, respectively, we found BLAST hits for 5,155 and 4,709 genes, respectively, in D. virilis, whereas for 63,555 and 46,536 out of the 65,670 and 48,474 reads, respectively, we found BLAST hits for 4,735 and 4,275 genes, respectively, in D. melanogaster. It is noteworthy that the numbers of the remaining reads for D. virilis were 22,200 (fungus infected) and 28,726 (naïve), which were more than ten times as many as the corresponding 2,115 (fungus infected) and 1,938 (naïve) for D. melanogaster (Table 1).

Expression Pattern of Immune-Related Genes.
According to Sackton et al. (2007) [8], innate immune system is categorized into three functional classes: "recognition, " "signaling, " and "effector. " In the D. virilis transcriptome analysis, 128 immune-related genes were detected, in which 23, 68, and 37 were assigned to recognition, signaling, and effector classes, respectively ( Table 2, Supplementary Table 2 and Supplementary Figure 1). In the case of the D. melanogaster transcriptome, 129 immune-related genes were detected, in which 28, 62, and 39 genes were assigned to recognition, signaling and effector classes, respectively (Table 3, Supplementary Table 3 (Table 3, Supplementary  Table 3), whereas the expressions of their homologs in D. virilis were not induced by the fungal infection (Table 2,  Supplementary Table 2). We also detected the genes for negative regulators of systematic immune response, such as PGRP-SC1a, PGRP-SC2, and PGRP-LB [36][37][38][39], as well as the genes for activators. Consistent with the expression of these recognition class genes, the expressions of signaling class genes, for example, Myd88, Rel, STAT92E, hep, and so forth, involved in Toll, Imd, JNK, and JAK/STAT pathways, were also detected in both species (see Tables 2 and 3 and  Supplementary Tables 2 and 3 for details).

Between-Species Differences in the Expression Pattern of
Effector Class Genes. Since the effectors directly function against infected microbes, in this study, we focus on the response of the effector class genes to the Penicillium infection to elucidate the differences in the antifungal resistance between D. melanogaster and D. virilis. In contrast to the shared expression pattern between the species observed in the recognition and signaling class genes, substantial differences in the expression pattern were observed in the effector class genes.
AMPs are known to be a major effector that has a critical role in the innate immune system of Drosophila [11]. In D. melanogaster, 20 AMP genes belonging to seven AMP gene families have been found, whereas 15 AMP genes belonging to five AMP gene families have been identified in D. virilis (Drosocin and Drosomycin in D. melanogaster are missing in D. virilis) [8]. In both D. virilis and D. melanogaster, many AMP genes (11 of 15 in D. virilis and 14 of 20 in D. melanogaster) were expressed in the fungus infected larvae (Tables 2 and 3, Supplementary Tables 2 and 3). In D. virilis, genes encoding Diptericin (GJ19916, TMM = 3.812), Defensin (GJ22479, TMM = 2.445), and Cecropin (Cec2B, TMM = 1.604 and Cec3, TMM = 1.475) showed high TMM values and Diptericin (GJ19916) was most highly expressed in the fungus infected larvae (Table 2). In contrast, the expression level of Metchnikowin, which was only the known antifungal peptide in D. virilis, was not so high (TMM = 0.660; Table 2). In contrast, Drosomycin (Drs) and Metchnikowin (Mtk), which were known as antifungal peptide genes, were most strongly expressed in the fungus infected D. melanogaster larvae (TMM = 23.817 and 23.719, resp.),    Table 3). Among the Drosomycin gene family, only Dro5 responded to the fungal infection, suggesting that D. melanogaster uses the specific Drosomycin gene copy against the Penicillium species; However, the expression level of Dro5 was 100-fold lower than that of Drs (TMM = 0.276) ( Table 3). These observations indicate substantial differences in the AMP usage between the species, that is, against the fungal infection, Diptericin, Defensin, and Cecropin were the three major AMPs in D. virilis, whereas Drosomycin and Metchnikowin were the two major AMPs in D. melanogaster ( Figure 3). Among other effector class genes, the immune-induced molecule (IM) genes showed distinct expression pattern between the species. The IM genes are known as the genes induced by bacterial or fungal infection in D. melanogaster. However, their functions mostly have not been characterized. In this study, 10 IM genes were identified to be expressed in the fungus infected D. melanogaster larvae and five of them, IM1, IM4, IM10, IM14, and IM18, were significantly upregulated by 2-fold or more (Table 3 and Supplementary  Table 3). For most of the D. melanogaster IMs, their expressions tended to be induced by the fungal infection. On the other hand, five IM genes, IM1 (GJ19885), IM4 (GJ18607), IM10 (GJ21308, GJ21309), and IM23 (GJ22454), were identified to be expressed in D. virilis, but their expression tended to be downregulated by the fungal infection (Table 2,  Supplementary Table 2). Particularly, the expressions of IM1 (GJ19885), IM4 (GJ18607), and IM10 (GJ21308) were significantly reduced by the fungal infection by half or less (Table 2). These differences in the expression pattern may indicate that IMs play separate roles in the immune response to fungal infection in D. melanogaster and D. virilis.

Novel AMP Genes in the Annotated D. virilis Genes.
Using the BLAST search against all the known D. melanogaster genes, we could not find the homologues for three D. virilis annotated genes significantly upregulated by the fungal infection. They were GJ10737 (IC = 2.503, = 0.0037), GJ11722 (IC = 3.198, = 0.032), and GJ18291 (IC = 3.909, = 0.047). Additional queries to orthologue database (orthoDB: http://cegg.unige.ch/orthodb6) [40] and the nonredundant gene database in the NCBI BLAST web server failed to find any known gene, suggesting that they were D. virilisspecific genes. Although we further searched for annotated domains and motifs in the expected products of these genes using the domain and motif search programs on NCBI Conserved Domain Database and Pfam, no conserved domain or motif was predicted. However, using SignalP (v4.0) [24], ProP (v1.0) [25], and JEMBOSS (v1.5) [26] programs, the expected products of GJ10737 and GJ18291 were predicted to be secretory peptides having propeptide sequences and positively charged mature peptide (Table 4). These features are commonly found in AMPs. Indeed, AMP prediction web programs, CAMP [28] and AMPA [29], predicted them to be AMPs, although another program, AntiBP2 [27], did not (Table 4). These results suggested the possibility that D. virilis possesses unknown AMP genes functioning in its innate immune system.

Novel
Immune-Related Genes in D. virilis. In our BLAST analysis described above, 22,200 and 28,726 pyrosequencing reads, respectively, from the fungal infected and naïve D. virilis larvae did not hit any known gene, whereas such reads were only 2,115 (infected) and 1,938 (naïve) in D. melanogaster (Table 1). We hypothesized that this is because there were many unidentified genes in D. virilis. To examine whether or not these reads were derived from unidentified immunerelated genes, we assembled these reads by mapping each read onto the D. virilis genome sequence to make contigs.   2 . 0 2 0 PG02420 2 * 10 0.208 ---- * , * * Significant difference from the number of reads for naïve larvae ( * < 0.05, * * < 0.01).
3,269 contigs of the average length 237 bp in total ( Figure 4). This indicates that these reads were actually derived from transcripts of the D. virilis genome rather than possible contaminants and that there are unidentified transcription units potentially encoding polypeptide. Since most of the contigs were shorter than the median length of 3 -UTR of D. melanogaster genes, we extended each contig with 250 bp each of upstream and downstream genome sequences to make a query sequence subject to the BLAST search against Swissprot protein database. As a result, we identified 620 putative genes in the 3,269 contigs. Among them, 27 putative genes showed a statistically significant difference in the number of reads between the fungus infected and naïve larvae. Three out of the 27 putative genes, PG00034, PG01778, and PG02420, were assigned to potential immune-related genes for subsequent GO analysis (Supplementary Table 4). PG00034 was homologous to IM14 of D. melanogaster. Although the expression of IM14 was significantly upregulated in D. melanogaster (Tables 3 and 5), the expression of PG00034 was significantly downregulated by the fungal infection in D. virilis. PG01778 was homologous to a Ras-like GTP-binding protein, Rho1, of D. melanogaster. This gene is known to play a role in regulating actin genes involved in phagocytosis [41][42][43][44]. The expression was observed only in the infected larvae in D. virilis and induced by the fungal infection (IC = 2.020) in the D. melanogaster larvae, indicating that this gene was up-regulated by the fungal infection in both species. PG02420 was homologous to Ficolin-2 that binds to the cell wall component of bacteria and fungi [45,46], and the expression of PG02420 was significantly downregulated in the infected D. virilis (IC = 0.208) ( Table 5).

Discussion
In this study, we first clarified that the antifungal resistance against Penicillium fungal infection is higher in D. virilis than in D. melanogaster. In general, adult flies of most Drosophila species are attracted to, feed, and breed upon a variety of fermenting substances such as fallen fruit and flowers, slime fluxes of forest trees, decaying bark of trees, and mushrooms [1]. However, there are interspecies variations of the fermenting substances utilized by Drosophila species for feeding and breeding. For instance, D. virilis is known to feed on slime flux and decaying bark of tree harboring many yeasts and filamentous fungi, such as Xanthophyllomyces dendrorhous, Cryptococcus spp., and Fusarium spp. [14,15], whereas D. melanogaster feeds on fermented fruits, which mainly harbor Baker's yeast, Saccharomyces cerevisiae [1][2][3].
The Penicillium species is ubiquitously and abundantly found in natural environment, where Drosophila species live, and grows on both decaying woods and fruits [16,17]. Therefore, both D. virilis and D. melanogaster are likely to be infected by them in nature during their life time. According to the theory of evolutionary adaptation, the higher antifungal resistance of D. virilis observed in this study (Figure 1) is expected to reflect the result of higher risk of the infection in their living environments over the evolutionary time compared to D. melanogaster. This raises the question of the immune mechanism attributed to the higher antifungal resistance of D. virilis, and it is thought to be a key factor for understanding the adaptive evolution of D. virilis to its habitat in moldy environment. To answer this question, we compared the immune responses to the fungal infection between D. virilis and D. melanogaster by analyzing their transcriptome extracted from larval salivary gland and fat body. Although the antifungal resistance was compared at the adult stage, we focused on the transcriptome at the larval stage. Since the larvae live and feed on fermented substances in their habitat environment and cannot escape from the surrounding microbes as the adults fly away, the larvae are consistently infected by microbes. Therefore, we assume that the resistance at the larval stage is more important for their adaptation to the environment. Unfortunately, it was difficult to measure the antifungal resistance at the larval stage since the larvae became pupae within several days and some larvae avoided immediate infection of fungi by digging the medium deeply. Accordingly, our interpretation in the following is on the basis of the assumption that the resistance at the adult stage correlates with the resistance at the larval stage.
Our comparative transcriptome analysis revealed that the genes involved in all major signaling pathways for immune response, that is, Toll, Imd, JAK/STAT, and JNK, were triggered by the infection of the Penicillium species in both D. virilis and D. melanogaster (Tables 2 and 3, Supplementary Tables 2 and 3). These pathways regulate humoral and cellular immune responses, such as AMP production and phagocytosis [7,47,48]. Among the signaling pathways, the Toll pathway plays an essential role against fungal infection in D. melanogaster [10,49]. The Toll pathway regulates expressions of two antifungal peptides, Drosomycin and Metchnikowin [50]. Consistent with this fact, the expression levels of Drosomycin and Metchnikowin genes were highest in the fungus infected D. melanogaster larvae ( Table 3). The response of these AMP genes to the infection of an entomopathogenic fungus, Beauveria bassiana, was highest in adult D. melanogaster as well [51,52]. Interestingly, seven genes encoding Drosomycin have been found in D. melanogaster genome (Drs, Drsl, Dro2, Dro3, Dro4, Dro5, and Dro6) [8]. Nevertheless, we found that only Drs and Dro5 were induced by the fungal infection in the D. melanogaster larvae (Table 3). This specificity of the expression pattern was consistent with the result of the microarray analysis by De Gregorio et al. (2001) [51], suggesting that the specific genes, Drs and Dro5, are used against the fungal infection at both larval and adult stages. In contrast, any Drosomycin gene is absent in the D. virilis genome and the expression of the Metchnikowin gene (Mtk) was not high (TMM = 0.660) compared to that of other AMP genes in the fungus infected D. virilis larvae (  Figure 3). This result was rather unexpected since Metchnikowin was the only known antifungal peptide in D. virilis, suggesting that Metchnikowin of D. virilis does not compensate for the lack of Drosomycin. Since the comparison of D. melanogaster and D. virilis genomes revealed that Mtk is present as a single copy gene in both species [8], it is implausible that D. virilis has an additional copy of Mtk responsible for the observed higher antifungal resistance.
On the other hand, the genes encoding Diptericin (GJ19916), Defensin (GJ22479), and Cecropin (Cec2B and Cec3) were highly expressed (TMM = 3.812, TMM = 2.445, TMM = 1.604 and TMM = 1.475, resp.) in the fungus infected D. virilis larvae compared to other AMP genes ( Table 2), suggesting a substantial difference in the AMP usage in response to the fungal infection between the two species and a possibility that Diptericin, Defensin, and Cecropin have an antifungal function in D. virilis. The antifungal activity of Diptericin and Defensin against an ascomycete fungus, Fusarium oxysporum, has been reported, although they are not effective against other fungi (Neurospora crassa, Beauveria bassiana, and Aspergillus fumigatus) in D. melanogaster [11]. Comparing the Diptericin protein sequence of D. virilis to its orthologue in D. melanogaster, we found substantial amino acid differences (50-70%). This may indicate the possibility that Diptericin of D. virilis has a different activity spectrum against fungi from that of D. melanogaster, although the main activity of the latter is not antifungal but antibacterial [53]. In contrast, amino acid sequences of mature peptide of Cec2B and Cec3 of D. virilis are almost identical (92.5-100%) to those of Cecropin of D. melanogaster, and the few amino acid substitutions observed are all conservative to maintain physicochemical properties of the peptide. Therefore, it is likely that the functions of Cecropin are conserved in the two species. A notable difference was observed in the Defensin gene. Defensin is known to be an AMP of main specificity to Gram-positive bacteria in D. melanogaster [54]. However, the Drosophila Defensin is classified into Defensin 2 superfamily (Pfam: PF01097), which has antifungal activity in mosquito (Anopheles gambiae) and sand fly (Phlebotomus duboscqi) [55,56]. D. virilis has two Defensin genes (GJ21126 and GJ22479). The mature peptide sequence of GJ21126 is closely related to the D. melanogaster Defensin gene as expected from their phylogenetic relationship of species, whereas the mature peptide sequence of GJ22479 is more similar to those of Anopheles gambiae (AgaDef) and Phlebotomus duboscqi (PduDef), which have antifungal activity ( Figure 5). In our transcriptome analysis, we detected the expression of GJ22479 but not GJ21126 in response to the Penicillium infection. A possible speculation based on these observations is that Defensin functions differently as an antifungal peptide in D. virilis from that in D. melanogaster. Since the expression of these three AMPs is under the regulation of the Imd pathway rather than the Toll pathway [50,57], this result suggests that the Imd pathway plays an important role in the response to the fungal infection in D. virilis, in contrast to the fact that the Toll pathway is more important to regulate the Drosomycin genes as the antifungal response in D. melanogaster. Alternatively, the Diptericin, Defensin, and Cecropin genes may be under the Toll pathway regulation in D. virilis. To examine this possibility, we analyzed the upstream region of these genes to see differences in DIF (Toll pathway) and Relish (Imd pathway) binding sites [58] between D. virilis and D. melanogaster. However, there was International Journal of Genomics no clear difference in the number, position, and direction of these binding sites, suggesting that the alternative possibility is not likely. A striking difference in the expression pattern was observed in the immune-induced molecule (IM) genes. The IM genes of D. melanogaster showed a similar expression pattern to that observed in the previous study conducted by De Gregorio et al. (2001) [51]. In this study, ten IM genes were expressed in the fungus infected D. melanogaster larvae and five of them, IM1, IM4, IM10, IM14, and IM18, were significantly up-regulated by 2-fold or more and downregulated gene was not observed (Table 3, Supplementary  Table 3). Similar inductions of IM genes were observed in adult flies by the infection of B. bassiana [51]. This suggests that the IM genes play a similar role in antifungal immunity in larvae and adults of D. melanogaster and against Penicillium and Beauvaria fungi, although the function of the IM genes has not been characterized. However, the IM genes showed contrary expression pattern in D. virilis: the expressions of five IM genes, IM1 (GJ19885), IM4 (GJ18607), IM10 (GJ21308, GJ21309), and IM23 (GJ22454), detected in D. virilis, were rather down-regulated by the fungal infection ( Figure 3). Indeed, three of them, IM1 (GJ19885), IM4 (GJ18607), and IM10 (GJ21308), showed statistically significant reductions ( Table 2, Supplementary Table 2). This result suggests differences in the functions of IMs between D. virilis and D. melanogaster. In other words, the definition of immuneinduced molecule (IM) holds true in D. melanogaster but not necessarily so in other Drosophila species. It can be speculated that D. virilis may have other immune-related genes that have the functions of IMs in D. melanogaster. Based on the comparative transcriptome analysis using bacterialinfected D. melanogaster and D. virilis flies, Sackton and Clark (2009) suggested that new components were recruited into the immune system of D. virilis [34]. Therefore, our results as well as their observation motivated us to search for novel immune-related genes in D. virilis.
In our transcriptome analysis, we found that three D. virilis-specific genes were induced by the fungal infection and two of them, GJ10737, and GJ18291, were predicted to encode novel AMPs (Table 4). This suggests that D. virilis has acquired lineage-specific AMPs against fungal infection through its evolution. Since no orthologous sequences of these genes were found in other Drosophila genomes either, these genes seemed to be recruited to the D. virilis genome de novo. In addition to the fraction of these genes of unknown function, we also predicted new D. virilis genes from the pyrosequencing reads that did not show any BLAST hit.
In our BLAST analyses of the pyrosequencing reads, approximately 30% of the reads from D. virilis did not hit any gene, whereas only 3-4% of the reads from D. melanogaster fell in the same situation (Table 1). This may suggest the possibility that many genes in the D. virilis genome have not been identified yet. Actually, we found 620 putative genes in 3,469 contigs and three of them, PG00034, PG01778 and PG02420, were predicted to be immune-related genes with expression level significantly changed by the fungal infection. PG00034 is homologous to IM14 and PG01778 is homologous to a Ras-like GTP-binding protein, Rho1, which regulates actin cytoskeletal organization [41,42] and is involved in phagocytosis [43,44] in D. melanogaster (Table 5). PG02420 is homologous to Ficolin-2 of Bos taurus. Ficolin binds to a cell wall component of bacteria and fungi and is involved in phagocytosis [45,46]. Although the expression of the IM14 gene was significantly up-regulated by the fungal infection in the D. melanogaster larvae, the expression of PG00034 was significantly down-regulated as in the case of other homologues of IM genes in the D. virilis larvae. Similarly, the expression of PG02420 was significantly downregulated in the infected D. virilis larvae. On the other hand, the expression of PG01778 was significantly up-regulated by the fungal infection in D. virilis. For the remaining 2,649 contigs, we could not find any homologue in Swissprot protein database. This seems partly to be because many of them are too short to find a homology to a known gene, domain, or motif in the homology search ( Figure 4). Further experimental determination of their full length sequence is necessary for a better prediction of novel protein-coding genes. Therefore, there is a possibility that some of these putative genes constitute novel components in the immune systems of D. virilis and contribute to the higher resistance against the fungal infection.
Our comparative transcriptome analysis revealed extensive differences in the immune response to the infection of Penicillium species between D. virilis and D. melanogaster at the transcriptome level. These results provide an important insight into the different role of immune system between ecologically diverged species. It is quite natural to consider that the observed differences resulted from evolutionary adaptation to their different habitats. This presumption should be further experimentally examined by the investigation of antimicrobial activities of AMPs, for example, Diptericin and Defensin, to identify the component responsible for the higher antifungal resistance of D. virilis.

Conclusion
In general, Drosophila species feed and breed on fermenting fruits, slime fluxes on decaying parts of tree, and so on, in which a variety, of microbes are extremely active [1][2][3]. Therefore, antimicrobes immune system is an essential trait for Drosophila species to survive. The evolution of the immune system is likely responsible for the diversity of Drosophila species adapting to a variety of microbial environments. In this study, a substantial difference in antifungal activity against a Penicillium species between two Drosophila species, D. virilis and D. melanogaster, living in different environments, was demonstrated.
Our comparative transcriptome analysis showed extensive differences in the expression pattern of immune-related genes, that is, antimicrobial peptide (AMP) and the immuneinduced molecule (IM) genes, in response to the Penicillium infection between D. virilis and D. melanogaster. Furthermore, we predicted novel immune-related genes responding to the fungal infection in D. virilis. These results indicate that the innate immune system has been substantially differentiated during the evolution of these Drosophila species. The extensive differences in the immune system may have been evolved as an adaptive response to microbial environments, which remains open to further investigations.