A number of species have been identified in the ascomycete Metarhizium genus that are insect pathogenic fungi (Kepler et al., 2014). Among these, Metarhizium anisopliae and M. acridum have been developed as biocontrol agents to mitigate insect pest populations (de Faria and Wraight, 2007; St Leger and Wang, 2010). Of particular interest is the divergence of Metarhizium species to adapt to different varieties of invertebrate hosts. For example, in contrast to the invasion of versatile hosts by the generalist species such as M. anisopliae and M. robertsii, M. acridum, and M. album are more specific toward orthopteran and hemipteran insect hosts, respectively (Wang et al., 2012). Pan-genomic analyses of Metarhizium species revealed the expansion/contraction of protein families associated with fungal host adaptation (Hu et al., 2014). Transcriptional and developmental differences have been also observed when Metarhizium species were challenged with host or non-host cuticles. These differences include the up-regulation of divergent G-protein coupled-receptors, differential expression of down-stream singling pathway genes, and finally, production of the infection structure appressoria to determine fungal host recognition and infection (Wang and St Leger, 2005; Gao et al., 2011; Chen et al., 2016). Despite these advances, metabolomic and physiological differences among the Metarhizium species are largely unclear when the fungi are grown under the same or different conditions.

Different bioactive metabolites have been isolated from Metarhizium species, including the well-known non-ribosomal peptide destruxins (Dtxs) (Molnar et al., 2010; Gibson et al., 2014), which are non-selectively insecticidal. Interestingly, the toxin-producing gene cluster was not present in the genomes of the specialist species M. acridum and M. album (Wang et al., 2012). Presence or absence of gene clusters involved in secondary metabolisms (SMs) has been frequently observed in the genomes of closely related or unrelated fungal species. Horizontal transfer of gene clusters between species has been suggested (Khaldi et al., 2008; Slot and Rokas, 2011). Divergent evolution also contributed to the diversity of fungal SMs. For example, a highly conserved polyketide synthase (PKS) involved in the biosynthesis of dihydroxynaphthalene (DHN) melanin is present in the genomes of Metarhizium species. However, DHN-type melanin could not be produced by Metarhizium due to the divergent evolution of the tailoring enzyme laccase (Chen et al., 2015). On the other hand, differential regulation (including epigenetic control) of secondary metabolic gene expression largely determines chemical diversity in fungi (Yu and Keller, 2005; Keller, 2015). Relative to the number of identified gene clusters in each species (Hu et al., 2014), chemical diversity and the associated genetics in different Metarhizium species are still poorly understood (Donzelli and Krasnoff, 2016).

High-resolution technique oriented metabolomics can systematically identify and quantify metabolites from a biological sample, and has emerged as a powerful tool for microbiological studies (Oldiges et al., 2007; Xu et al., 2014). For example, metabolomic analyses of the insect pathogens M. brunneum and Beauveria bassiana cultured in different conditions revealed that both fungi produce significantly different arrays of secondary metabolites during infectious and saprophytic growths (de Bekker et al., 2013). Similarly, the chemical diversity of B. bassiana varied in association with the host and non-host growth cultures, and oxidative stress conditions (Luo et al., 2015; Xu et al., 2015; Zhang et al., 2016). The exploration of metabolomic technique could also help identify known metabolites that have not been reported in a given sample or organism. For example, metabolomic analysis of B. bassiana mycelial samples identified the macrocyclic lactone, doramectin and a magnesium-containing antibiotic, magnesidin that were originally identified in bacteria (Luo et al., 2015). Our previous analysis of metabolic dynamics in silkworm (Bombyx mori) larvae identified a mulberry leaf component maclurin in insect hemolymph that reduced the feeding tendency of insects infected with B. bassiana, thus indicating an anti-feeding effect on insects infected with fungal pathogen (Xu et al., 2015). Metabolomic analysis has been also used to investigate and compare the intracellular metabolic responses among different yeast species (Christen and Sauer, 2011).

In this study, we performed a comparative analysis of gene clusters involved in SMs in seven Metarhizium species. Metabolomic analyses of seven species grown in the same medium were also conducted by using gas chromatography (GC) and liquid chromatography-mass spectrometry (LC-MS). Consistent with the conserved and divergent evolution of SM gene clusters, our metabolome data indicate that fungal metabolic profiles are linked to fungal host specificity. In addition, the high resolution of the MS technique facilitated the identification of a few compounds that have not been reported previously in Metarhizium species.

Genomic analyses indicated that entomopathogenic fungi belonging to the Metarhizium genus have the potential to synthesize a plethora of secondary metabolites (Hu et al., 2014). However, apart from the handful of small molecules previously identified in Metarhizium, the capacity of secondary metabolite production in Metarhizium species has not been fully investigated (Donzelli and Krasnoff, 2016). In this study, re-analysis of the genome data revealed the occurrence of both conserved and divergent evolution of SM gene clusters among the seven Metarhizium species analyzed. In particular, the gene cluster involved in the production of insecticidal destruxins is found to have evolved divergently in the different species, including the incident of lineage/species-specific mutagenesis in the toxin-biosynthesis gene that likely disabled the production of toxins. By employing integrated metabolomic approaches, we found that the obtained data allowed the grouping of Metarhizium species into clusters correlating with fungal host ranges. In addition, we found that the pattern of small molecule production is generally consistent with the conserved and divergent evolution of SM gene clusters.

Secondary metabolic conservation and diversification have been frequently observed in fungi. It includes two different scenarios: closely related fungi containing divergently evolved SM gene clusters that produce structurally different small molecules, whereas divergently evolved fungi producing the same metabolites (Keller, 2015). Consistent with this notion, the analysis of seven Metarhizium species in this study demonstrated clearly the conserved and divergent evolution of SM genes. Among > 30 SM gene clusters encoded by different species, less than half (12 clusters) are conserved in all examined Metarhizium species while the remaining clusters are either lineage- or species-specific. In particular, the gene cluster responsible for Dtx production is not present in the host-specific fungi M. album and M. acridum. Considering that destruxins are non-selectively insecticidal (Pedras et al., 2002; Gibson et al., 2014), the absence of this gene-cluster in host-specific fungi but its presence in non-specific Metarhizium species shows a clear association with the evolution of fungal host specificity. On the other hand, this observation also raises the question whether the gene cluster was lost in specialists or acquired in non-specialist species. Metarhizium species evolved after the divergence of the ascomycete plant pathogenic fungi (Shang et al., 2016; Wang et al., 2016), and the specialist species evolved first (Hu et al., 2014). To adapt to diverse insect hosts, it is possible that the non-specialists acquired the toxin-producing gene cluster during their speciation processes. However, considering that Dtx B can be produced by plant pathogens (Wang et al., 2012), it cannot be ruled out that the ancestor of Metarhizium contained the gene cluster but the specialist species lost it. Gain or loss event could also be applied to explain the presence/absence of other lineage/species-specific clusters. For example, comparative analysis of the putative genes involved in helvolic acid biosynthesis indicated the absence of the cluster in the genomes of the specialists M. acridum and M. album (Donzelli and Krasnoff, 2016). Besides the incidence of the horizontal transfer of gene clusters (Khaldi et al., 2008), the driving force(s) leading to the gain/loss evolution of SM genes among those closely related fungal species remains elusive.

Based on our metabolomic data, PCA analysis divided the examined Metarhizium species into separate groups in association with fungal host ranges (Figure 3). In particular, the small molecule profiles clearly distinguished between the generalist and non-generalist species (Figure 3B). These findings are consistent with previous studies that demonstrated the contribution of secondary metabolites to fungal virulence; e.g., destruxins in Metarhizium (Wang et al., 2012), and beauvericin and oosporein in B. bassiana (Xu et al., 2008; Feng et al., 2015). For destruxins, about 40 analogs have been identified (Pedras et al., 2002). Based on chromatography, however, only a few common destruxin derivatives such as Dtx A, Dtx B, and Dtx E could be detected in any given Metarhizium species or strain (Amiri-Besheli et al., 2000; Wang et al., 2004). In this study, the metabolomics analysis based on the highly sensitive LC-MS technique indicated that additional derivatives could be produced by fungi grown in an artificial medium (Figure 5A). This supports previous results that the adenylation of NRPS DtxS1 domain 3 can select either proline (for Dtx A-F series) or pipecolic acid (for Dtx A₁-F₁ series) as substrate, and the DtxS1 domain 4 can use either isoleucine (for Dtx A-F series) or valine (for Dtx A₂-F₂ series) (Donzelli et al., 2012; Wang et al., 2012). It is unexpected to find that the specialist M. album, with the fewest number of PKSs, produced higher amounts and varieties of polyketide molecules compared to other species (Figure 6A). Future studies are required to determine the mechanism(s) involved in the differential control of SMs in Metarhizium species.

Our metabolomics analysis also identified small molecules that were not previously reported in any Metarhizium species (Figure 4; Supplementary Table S2). Interestingly, these metabolites were originally identified from other insect pathogenic fungi except the Streptomyces-origin of polyketones naphthgeranines (Wessels et al., 1991). For example, torrubiellutins were identified from the insect pathogen T. luteorostrata (Pittayakhajonwut et al., 2009), hirsutellones from Hirsutella nivea (Isaka et al., 2005), and trichothecanes tenuipesine (Kikuchi et al., 2004a), spirotenuipesines (Kikuchi et al., 2004b), and paecilomycines (Kikuchi et al., 2004c) from the insect pathogen Paecilomyces tenuipes (now re-classified as Isaria tenuipes). The identification of similarly conserved SM gene clusters among the Metarhizium and Torrubiella species (Figure 1C) supports, at least in part, the production of same metabolites by different fungal species that has been frequently observed (Gibson et al., 2014). This finding would suggest that these metabolites may contribute to fungal entomopathogenicity. Likewise, by using a high resolution MS technique, the immunosuppressive compounds α-pyrone diterpenoids subglutinols were identified in M. robertsii (Kato et al., 2016). In contrast to the compounds mentioned above, these diterpenoids were originally identified from the plant pathogen Fusarium subglutinans (Lee et al., 1995). Future studies are required not only to verify the production of these molecules in Metarhizium species but also to elucidate their biosynthetic mechanisms and biological functions.

There are also the cases that the known products identified from Metarhizium species have not been detected in our metabolome analysis. These compounds include the non-ribosomal cyclopeptides ferricrocin, serinocyclins, and metachelins; the polyketides aurovertins; the polyketide/peptide hybrids metacridamides, NG-391 and NG-393; the terpenoids viridoxins, metarhizins and ovalicins; and the indolizidine alkaloids swainsonine and fungerins (Gibson et al., 2014; Donzelli and Krasnoff, 2016). First, it could be the reason of species-specific production of certain metabolites, which can be supported by the presence of species/lineage-specific gene clusters in different Metarhizium species. For example, the diterpenes viridoxins and metarhizins were identified from M. flavoviride (Gupta et al., 1993; Kikuchi et al., 2009). These diterpenoids have not yet been reported in other Metarhizium species. Second, the growth media may determine fungal chemical diversity. For example, aurovertins F-H were identified from M. anisopliae grown in a culture containing starch and glycerol (Azumi et al., 2008). In this study, to normalize the growth condition for comparative analysis, the fungi were grown in a nutrient-rich SDB medium. It is thus possible that certain compound(s) will not be produced by the fungi grown in this condition. Chemical diversity analysis is still required by growing the fungi in different conditions, especially that the insect-host-oriented stimulations may facilitate the identification of novel compounds from these fungi.

As indicated above, the putative gene cluster for helvolic acid biosynthesis is not present in the specialist fungi (Donzelli and Krasnoff, 2016). However, our metabolomic analysis detected the compound in M. acridum and M. album (Figure 5A; Supplementary Table S2). The biosynthetic mechanism of helvolic acid has not been fully understood. A loss-of-function study revealed that a geranylgeranyl diphosphate synthase (BAJ05823) is involved in the production of helvolic acid in M. anisopliae (Singkaravanit et al., 2010). Our genome survey indicated that the homolog of this protein is present not only in the non-specialists (e.g., MAA_03020 of M. robertsii, 82% identity) but also in M. acridum (MAC_02619, 80%) and M. album (MAM_05601, 75%). A previous study in Aspergillus fumigatus indicated that an oxidosqualene cyclase (Afu4g14770) is required for helvolic acid biosynthesis (Mitsuguchi et al., 2009). Likewise, the homolog of this cyclase is present in M. robertsii (MAA_06578, 63%), M. acridum (MAC_09947, 38%), and M. album (MAM_07210, 39%). However, the homologs of acyltransferases (Afu4g14820 and Afu4g14820) are not present in the genomes of the specialist fungi. Thus, the biosynthetic nature of helvolic acid in Metarhizium species remains to be determined in the future.

We performed comprehensive genomics and metabolomics analyses of seven insect pathogenic Metarhizium species by focusing on their gene contents and abilities in producing bioactive metabolites. The gene clusters involved in the biosynthesis of secondary metabolites are diverse among the examined species. The presence of non-selective insecticidal toxin producing gene cluster in non-specialist species is generally associated with fungal adaptation to diverse insect hosts. Consistent with the biosynthetic gene diversification, the MS-based metabolomics data also revealed the chemical diversification among the seven Metarhizium species, which could separate the fungal species into groups correlating with fungal host ranges. The results obtained in this study will benefit future investigations and elucidations of secondary metabolic mechanisms and the function of small molecules involved in insect-fungal interactions.

Y-JX, and FL performed experiments. Y-JX, BL, YS, and CW analyzed the data. Y-JX and CW wrote the manuscript. CW designed the experiments.

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.