Iron homeostasis in the absence of ferricrocin and its consequences in fungal development and insect virulence in Beauveria bassiana

The putative ferricrocin synthetase gene ferS in the fungal entomopathogen Beauveria bassiana BCC 2660 was identified and characterized. The 14,445-bp ferS encodes a multimodular nonribosomal siderophore synthetase tightly clustered with Fusarium graminearum ferricrocin synthetase. Functional analysis of this gene was performed by disruption with the bar cassette. ΔferS mutants were verified by Southern and PCR analyses. HPLC and TLC analyses of crude extracts indicated that biosynthesis of ferricrocin was abolished in ΔferS. Insect bioassays surprisingly indicated that ΔferS killed the Spodoptera exigua larvae faster (LT50 59 h) than wild type (66 h). Growth and developmental assays of the mutant and wild type demonstrated that ΔferS had a significant increase in germination under iron depletion and radial growth and a decrease in conidiation. Mitotracker staining showed that the mitochondrial activity was enriched in ΔferS under both iron excess and iron depletion. Comparative transcriptomes between wild type and ΔferS indicated that the mutant was increased in the expression of eight cytochrome P450 genes and those in iron homeostasis, ferroptosis, oxidative stress response, ergosterol biosynthesis, and TCA cycle, compared to wild type. Our data suggested that ΔferS sensed the iron excess and the oxidative stress and, in turn, was up-regulated in the antioxidant-related genes and those in ergosterol biosynthesis and TCA cycle. These increased biological pathways help ΔferS grow and germinate faster than the wild type and caused higher insect mortality than the wild type in the early phase of infection.

Iron homeostasis is essential in nearly all living forms, from mammals to microbes. It maintains a balance between the iron supply and the prevention of cellular toxicity due to iron overload 1,2 . In fungi and bacteria, small iron-chelating compounds called 'siderophores' and transporters mediate iron-associated mechanisms, similar to those found in humans and mammals [3][4][5] . Insect-pathogenic fungi are microbes used for the biological control of agricultural insect pests. Beauveria bassiana is a widely-regarded biological control agent used commercially in several countries, including Thailand. As an entomopathogen, iron is indispensable for the initiation and establishment of B. bassiana in insect hosts 6 . Fungal cells primarily utilized small molecules "siderophores" to sequestor iron from the environment, thus being an extracellular siderophore, and to store and distribute iron in the cell, being an intracellular siderophore. Ferricrocin is a common intracellular siderophores in several fungi, including Aspergillus fumigatus 7 , A. nidulans 8 , and Metarhizium robertsii 9 .
A major group of enzymes that synthesize siderophores belong to the family of nonribosomal peptide synthetases (NRPSs). Several siderophores are synthesized by these multimodular NRPSs. Each NRPS module has an adenylation domain (A), thiolation domain (T), and condensation domain (C), thus activating and incorporating www.nature.com/scientificreports/ one amino acid. The phylogeny of ferrichrome synthetases is based on their modular evolution, and their domain architecture divides these enzymes into two lineages, NPS1/SidC lineage, and NPS2 lineage. Most ferrichromes consist of three N 5 -acyl-N 5 -hydroxy-L-ornithine (AHO) and three amino acids. One amino acid is always a glycine, and the remaining two can be a combination of alanine, serine, or glycine. For example, ferrichrome A consists of three AHOs, one glycine, and two serines. Ferricrocin consists of three AHOs, with two glycines and one serine 10 . Although several fungal NRPSs associated with intracellular siderophore biosynthesis have been studied, there are distinct roles for the intracellular siderophores of different fungi, particularly among fungal pathogens. For example, the ferricrocin synthesis gene ssm1 is involved in intracellular siderophore production in the phytopathogenic fungus Magnaporthe grisea. It contributes to the plant infection process, including the formation of a penetration peg. The ssm1 mutation affected fungal pathogenicity in rice 11 . In contrast, the disruption of ferrichrome synthetase gene sid1 (Δsid1) in plant pathogenic fungus Ustilago maydis did not affect its phytopathogenicity 12 . Previously, sidC1 that encodes a monomodular nonribosomal peptide synthetase has been knocked down by RNA silencing in B. bassiana BCC 2660 13 . In this study, we completely knocked out the ferricrocin synthetase gene ferS by targeted disruption. We performed comprehensive studies of ΔferS compared with B. bassiana wild type. The biosynthesis of ferricrocin has been abolished in ΔferS, which unexpectedly led to gains of functions in conidial germination and virulence against insects. Comparative transcriptomes between the wild type and ΔferS suggest several potential genes associated with ferroptosis, oxidative stress response, ergosterol biosynthesis, TCA cycle, and mitochondrial expansion. These processes might serve as acquired oxidative stress responses, which promote oxidative stress resistance of ΔferS during B. bassiana infection.

Results and discussion
The multimodular ferricrocin synthetase gene in B. bassiana BCC 2660. Before the complete genome of B. bassiana BCC 2660 was obtained and analyzed, the function of a sidC-like gene was determined by RNA silencing. The sidC1-silenced mutants showed deficiency in production of des-ferricrocin and ferricrocin, and had an increase in tenellin and iron-tenellin complex in iron-replete conditions 13 . However, the B. bassiana BCC 2660 genome sequence 14 revealed that the fungus has four sidC-like genes, which are three monomodular NRPSs, sidC1 (accession No. MZ086759; encoding a 1525-aa protein), sidC2 (MZ086760; a 1417-aa protein) and sidC3 (MZ086761; a 1380-aa protein), and a multimodular NRPS 'ferS' (MZ031022) that encodes a 4818-aa protein. The domain organization of each putative SidC-like protein is shown in Fig. 1A. All the three SidC-like NRPSs comprise only one set of A, T and C domains. By contrast, FerS consists of three complete modules of A-T-C, an extra set of T-C domains interrupted between the second and third modules, and a double set of the T-C domains at the C terminus. The monomodular SidC1 alone might not confer the ferricrocin biosynthesis based on its domain composition. Since there was a sequence similarity (33%) between sidC1 and the first adenylation domain of ferS, the off-target effect of RNA silencing might account for the reduction in ferricrocin production in our previous study 13 . Therefore, in this study, the function of the putative ferricrocin synthetase gene ferS in B. bassiana BCC 2660 was verified by insertional mutagenesis.
We have assessed the evolutionary conservation of B. bassiana BCC 2660 ferricrocin synthetase and their homologs. The domain architecture of FerS is remarkably similar to the modular architecture of ferrichrome synthetases (type IV NRPSs) such as NPS2 from F. graminearum and SSM1 from M. grisea 10 (Fig. 2A). We performed multiple alignment of the adenylation domains from B. bassiana BCC 2660 FerS and the three monomodular SidCs and other known fungal ferrichrome and ferricrocin synthetases, and constructed a phylogenetic tree (Fig. 2B) using the neighbor-joining method in CLUSTAL-X 15 . The NRPS signature sequences for substrate specificity were also predicted by NRPS-PKS, which is a knowledge-based resource for analyzing nonribosomal peptide synthetases and polyketide synthases 16 . Amino acid residues at the signature sequences of adenylation domains from the four B. bassiana BCC 2660, including FerS, were compared to other known ferrichrome and ferricrocin synthetases (Fig. 2B). The phylogeny indicated that B. bassiana BCC 2660 FerS and three SidC-like NRPSs could be placed in two lineages, NPS1/SidC and NPS2, according to the previous classification 10 . The monomodular SidC-like NRPSs were clustered with the first adenylation domains of A. nidulans and A. fumigatus SidCs, which have substrate specificity to serine ( Fig. 2A,B). Nevertheless, the signature sequences of the three monomodular SidCs do not match the signature sequence of the adenylation domains that are specific for serine, and neither do the signature sequences of adenylation domain in other ferrichrome and ferricrocin synthetases. On the other hand, FerS was clustered with ferricrocin synthetases in the NPS2 lineages. The signature sequences of all FerS adenylation domains were identical with the adenylation domains of F. graminearum ferricrocin synthetase NPS2 (FgNPS2); the first adenylation domain is specific for glycine, the second domain for serine, and the third domain for N 5 -acyl-N 5 hydroxy-L-ornithines (AHO). Thus, our sequence analysis suggested that FerS is a complete ferricrocin synthetase, most likely crucial for ferricrocin biosynthesis in B. bassiana BCC 2660. The three SidC-like monomodular NRPSs could result from evolutionary events that include deletion of the second and third adenylation domains and a following triplication of the first adenylation domain.
The ferS-null mutants abolished the ferricrocin production. Transformation of B. bassiana BCC 2660 with the ferS-disruption plasmid pCXFB4.4 generated 28 glufosinate-resistant transformants. Southern analysis indicated that two out of 28 transformants had an integration of the bar cassette at the targeted ferS locus, demonstrated by an increase of the 4-kb ferS fragment by the 1-kb size of bar (Fig. 1B). The Southern result also confirmed the presence of bar in the transformant but not in the wild type (Fig. 1B). Furthermore, our PCR analysis verified the similar bar integration in the same locus of ferS and the 5′ and 3′ border regions of the bar integration site (Fig. 1C).

AG T C AR
T C T C T C T C AHO T C www.nature.com/scientificreports/ Then, our metabolite analysis using HPLC indicated the lack of desferricrocin and ferricrocin production in ΔferS (Fig. 3A). The metabolite profile of mycelial extracts from wild type showed desferricrocin and ferricrocin production at the retention time (Rt) of 10.408 and 10.887 min, respectively. Under the iron-replete conditions, the amount of ferricrocin has increased, while the amount of desferricrocin drastically decreased in the wild-type extract. The spectrum absorption of desferricrocin and ferricrocin are shown in Fig. 3B. In contrast, both the desferricrocin and ferricrocin peaks were undetected in the metabolite profile from ΔferS (Fig. 3A). Notably, the ΔferS metabolite profile had an unknown compound (c) peak at Rt of 10.867 min with www.nature.com/scientificreports/ the distinct spectrum absorption from those of ferricrocin and desferricrocin (Fig. 3B). We have analyzed the mycelial extracts of both wild type and ΔferS using TLC, and verified that the mutant ΔferS had abolished the ferricrocin production (Fig. 3C).
The ferS disruption affected radial growth, germination and conidiation. The mutant ΔferS surprisingly had some particular advantages in growth and development over the wild type. For the radial growth, as a mean of vegetative, hyphal growth, ΔferS grew larger than the wild type on the same day of incubation under all the culture conditions supplemented by 10-400 µM Fe (Fig. 4A,B). At the low (10 µM) iron condition, the mutant radial growth increased by 13% over the wild type. When the iron concentrations were increased to 100 and 200 µM, the growth increases were more pronounced by 31-35% in ΔferS. At the highest Fe concentration tested, the mutant grew bigger than the wild type by 400%, which was clearly observed by visual colony inspection (Fig. 4A,B). Under the iron depletion (MM + bathophenanthrolinedisulfonic acid (BPS); conducted in separate independent experiments), the mutant radial growth increased by 11% over the wild type. The sidC1-silenced mutants also increased radial growth compared to wild type under minimal medium agar supplemented by 10 µM Fe 13 . Conidial germination was also enhanced in ΔferS. Our microscopic observation data indicated that ΔferS conidia germinated at a significantly (p < 0.05) higher percentage than the wild-type conidia under the iron depletion ( Fig. 4C), remarkably similar to the increase in the vegetative (hyphal) growth described above. However, under the iron-replete conditions, both the strains germinated similarly. Together, iron appears not necessary for the hyphal growth (shown by the data of radial growth and conidial germination) in B. bassiana BCC 2660, and indeed appears to have an inhibitory effect on vegetative growth. In contrast, asexual reproduction, as a measurement of conidiation, was reduced in ΔferS, consistent with a decreasing trend in conidiation found in sidC1-silenced mutants (Supplemental File S1). On potato dextrose agar (PDA) cultivation, the mutant produced a smaller number of conidia than the wild type (p < 0.05) per area of PDA culture (Fig. 4D). There was a clear distinction in aerial hyphae formation and conidiation between the wild type and 'the ferricrocin-deficient/ferricrocin-free mutants' . The wild-type colony had a lawn of aerial mycelia and numerous, dense clusters of conidia; however, the mutants' colonies appeared to have sparse growth with fewer conidial clusters (Supplemental File S1). In A. fumigatus, ferricrocin is responsible for iron transport and distribution, especially iron transport from substrate hypha to the aerial parts. Therefore, the ferricrocin deficiency results in a reduction of conidial production 7 . Similarly, the reduction of both aerial hyphae and conidiation results suggested that the reduction or the abolishment in ferricrocin production impaired the development of aerial hyphae, conidiophores, and conidia in B. bassiana BCC 2660 mutants. Thus, the role of ferricrocin in the iron supply used for asexual development has been demonstrated in this study.

The ferricrocin-free mutants had increased insect virulence. The mutant ΔferS lacks ferricrocin,
an important iron-storage molecule. As iron is essential for the pathogenicity of several pathogens in the hosts, the lack of ferricrocin in the mutant would have been assumed to lead to a deficiency in the virulence against the insect. However, our insect bioassay data from three independent experiments showed that ΔferS was not deterred in the virulence against insect, compared to the wild type (Fig. 5). Indeed, the mutant was significantly increased in the ability to kill the insects, compared to wild type, on day 2 after inoculation (Fig. 5). The LT 50 of ΔferS was 2.46 days, 7 h shorter than wild type (LT 50 of 2.75 days). This is interesting because we would not have anticipated a gain of function from a gene deletion unless the gene serves as a repressor or negatively relates to the phenotype.

Comparative transcriptomes indicated differential gene expression patterns in response to iron depletion and iron excess between the mutant ΔferS and wild type.
We investigated what mechanisms that can lead to the increases in radial growth, germination, and insect virulence in ΔferS as we observed. RNA Seq was conducted to compare the gene expression of wild type and ΔferS under iron-depleted conditions (WT-and ΔferS-BPS) and under iron-replete conditions (WT-and ΔferS-Fe). These conditions were used to mimic the host-pathogen interaction process. The pathogen B. bassiana encounters the iron-limited environment at an early stage of infection, and the oxidative burst from the host defense response in the insect hemocoel. Our transcriptomic analysis with Cufflinks showed a total expression of 9879 genes and 10,066 isoforms in all eight replicates (each of the four treatments having two replicates). The pairwise comparison results identified 308 differentially-expressed genes (DEGs) (p < 0.01). Wild-type responses to iron-replete conditions were represented by the expression of 58 up-regulated DEGs and 41 down-regulated DEGs, of which 93 and 90% have putative known functions (Table 1). In ΔferS, 41 DEGs were up-regulated, and 46 were down-regulated, of which 88 and 76% have putative functions under the iron-replete conditions ( Table 1). The enriched functions of up-regulated DEGs in ΔferS included cytochrome P450 and ABC transporter genes. In contrast, the enriched functions of down-regulated DEGs included those of coagulation factor, ricin b, and TauD.
Furthermore, the enriched DEGs were classified into 11 clusters based on gene expression patterns among four treatments using K-means clustering (k = 11) (Supplemental File S2). The overview of the expression profile of the clusters is shown in the graph. The bold black line is the medoid line that demonstrates the trend of expression profile in each DEG cluster. The complete list of clustering results is provided in Supplemental File S3. The expression profile of DEG clusters was evaluated in relation to gene functions and the pathway in which they involve. The KofamKOALA 17 and BlastKOALA 18 web tools were used to map DEGs into KEGG gene accession and pathways. There were 123 DEGs that can be mapped to the KEGG database. Among these DEGs, 46 genes were specific for, i.e., have the highest specificity scores in, the iron-depleted conditions (30 genes in WT-BPS and 16 genes in ΔferS-BPS). By contrast, 77 genes were specific for the iron-replete conditions (  www.nature.com/scientificreports/ DEG clusters 2 and 5 included genes that had increased expression in ΔferS over wild type under iron-depleted and replete conditions (Supplemental File S2). In cluster 2, the long-chain acyl-CoA synthetase gene acsl4 is involved in ferroptosis, a regulated form of cell death and characterized by a production of reactive oxygen species (ROS) from accumulated iron and lipid peroxidation. ACSL4 regulated ROS via phospholipid-hydroperoxide glutathione peroxidase. We also found the low-affinity ferrous iron transporter gene fet4 in cluster 2. Cluster 5 contained six DEGs, four of which have not been characterized for functions in the literature. It is fascinating that under the iron-depleted condition, cluster 5′s DEGs were not expressed at all in wild type but strongly expressed in ΔferS. These could be the genes that were apparently induced in ΔferS in response to iron depletion. Understanding the functions of these genes would lead to finding de novo proteins that have essential roles in response to no or low iron conditions in fungi.
On the other hand, cluster 11′s DEGs had the opposite expression pattern to those in clusters 2 and 5. These genes' expression was utterly missing in ΔferS, but was high in the wild type under the iron-replete conditions. One of these genes was the ferric reductase required for the high-affinity iron uptake 19 , suggesting that ΔferS might be impaired in the reductive iron uptake. A likely hypothesis for this phenomenon might be to limit or reduce the level of labile Fe 2+ in the ΔferS cells, which often causes iron toxicity. Moreover, as reported above ΔferS exhibited the increased virulence against the insect host. This is strikingly similar to the hypervirulence phenotype found in the mutant Δfet1 knocked-out in the ferroxidase gene, a core component of the reductive iron assimilation system in the phytopathogen Botrytis cinera 20 .
Cluster 9 was particularly intriguing that the mutant ΔferS was considerably increased in expression of fusarinine C synthase, cytochrome P450 52A10, cytochrome P450 CYP56C1, C-14 sterol reductase, ergosterol biosynthesis ERG4/ERG24 family protein, autophagy-related protein, oxaloacetate acetylhydrolase, L-lactate dehydrogenase and two major facilitator superfamily transporters, compared with wild type (Fig. 6). The data of the other clusters are provided in Fig. 6 and Supplemental Files. S2 and S3.
Increase in certain parts of siderophore biosynthesis and other iron homeostasis mechanisms in ΔferS. The wild type and ΔferS had a notably similar pattern of gene expression in three siderophore biosynthetic genes, sidA, sidD, and sidL, under the iron-depleted condition. On the other hand, when the fungal cells were exposed to the high-iron condition, sidA, sidD, and sidL were markedly enhanced in the expression in the mutant ΔferS (Fig. 6). SidD is a nonribosomal siderophore synthetase required for biosynthesis of the extracellular siderophore, fusarinine C. Its production is usually induced upon a low-iron environment, and suppressed  www.nature.com/scientificreports/ upon a high-iron condition 21 . Here, in ΔferS the expression of sidD was significantly more pronounced than in wild type in the iron-replete conditions (Fig. 6). The wild-type sidD expression level under the iron-replete conditions decreased to half of that in iron depletion. However, there was no clear statistical significance in the difference (p = 0.058). On the other hand, ΔferS had significantly enhanced expression of sidD to a threefold under the iron-replete conditions, compared to that under iron depletion (p = 0.006). The siderophore-mediated iron homeostasis regulation could sense the unavailability of ferricrocin and, in turn, produce and secrete fusarinine  www.nature.com/scientificreports/ C for sequestering iron. Such a higher level of fusarinine C could promote the infection of ΔferS in the host, as we observed the higher insect virulence of the mutant than the wild type. Inside the cell, SidL is N 5 -hydroxyornithine-acetylase required for biosynthesis N 5 -acetyl-N 5 -hydroxyornithine, an essential intermediate of ferricrocin biosynthesis. The expression of sidL was drastically increased to 26.9-fold in ΔferS (p < 5E −05 ), but to only 5.0-fold in the wild type (p < 5E −05 ) when the expression in iron-replete conditions was compared to that in iron deplete (Fig. 6). The drastic increase of sidL expression could be due to the similar regulatory mechanism that senses no ferricrocin in the cell. Lastly, SidA is L-ornithine N 5 -monooxygenase essential for biosynthesis of N 5 -hydroxy-L-ornithine, the building block of all siderophores in fungi. Similarly to the sidL expression pattern with a less extent, the expression of sidA was increased to 5.2-fold in ΔferS (p < 5E −05 ), but to only 3.4-fold in the wild type (p < 5E −05 ) when expression in iron-replete conditions was compared to that in iron depletion (Fig. 6).
In addition to those in siderophore biosynthesis, the iron homeostasis genes had differential gene expression patterns under the iron-replete conditions. The vacuolar iron transporter (vit) gene was up-regulated in response to the high iron condition by an increase of 58.5-fold in ΔferS (p < 5E −05 ), but 31.3-fold in the wild type (p < 5E −05 ). In contrast, reductive iron assimilation-related genes such as iron transport multicopper oxidase (fet3) and highaffinity iron transporter (ftr) genes were down-regulated under high iron conditions. Nevertheless, for fet3, the mutant ΔferS had a two-fold expression level over that of wild type under low and high iron conditions (Fig. 6).
ΔferS was increased in ferroptosis, oxidative stress response, ergosterol biosynthesis, TCA cycle, and mitochondrial expansion. Interestingly, ΔferS showed remarkable up-regulation of genes for cytochrome P450 and those in TCA cycle, ergosterol biosynthesis, alternative iron homeostasis, autophagy, and ferroptosis under iron depletion iron-replete conditions, compared to the wild type.
Ferroptosis, oxidative stress response and ergosterol biosynthesis. The oxaloacetate acetylhydrolase and cellobiose dehydrogenase (CDH) genes were up-regulated in ΔferS, especially in the high iron environment. Oxaloacetate acetylhydrolase is involved in oxalate production. The gene was up-regulated in ΔferS, especially in iron-replete conditions. In the meantime, oxalate decarboxylase gene, required for decomposition of oxalate to formate and carbon dioxide 22 , was down-regulated in ΔferS. Oxalate can reduce the toxicity of metals by forming metal-oxalate complexes, therefore being able to act as an iron chelator. The formation of iron oxalates has been reported in B. bassiana 23 . The CDH is a heme-containing oxidoreductase that can transfer electrons to electron acceptors such as cytochrome c and ferric-oxalate 24 . CDH has an essential role in wood decomposition 25,26 . This oxidoreductase can generate hydrogen peroxide by oxygen reduction and helps degrade cellulose, xylan, and lignin in the presence of hydrogen peroxide and ferrous ions 24,27 . Therefore, the up-regulation of oxaloacetate acetylhydrolase and CDH in ΔferS is consistent with the process that leads to the ferroptosis pathway via the Fenton reaction and lipid peroxidation. Oxalate binds to Fe 3+ to form iron-oxalate complex. CDH acts as a hydrogen peroxide (H 2 O 2 ) generator and iron-reducing agent, which reduces Fe (III)-oxalate complex to ferrous ions (Fe 2+ ). The accumulation of Fe 2+ in the cytoplasm induced the expression of vacuolar iron transporter (VIT). The mutant ΔferS had a significant (p < 5E −05 ) increase of vit expression compared to wild type (Fig. 6). The coincidence of Fe 2+ and H 2 O 2 could lead to hydroxyl radical generation via the Fenton reaction.
The generation of such free radicals can damage the cell membrane by the process of membrane lipid peroxidation. However, our transcriptomic data indicated that ergosterol biosynthesis genes and oxidative stress response gene were up-regulated in ΔferS, compared with wild type (Fig. 6). These ergosterol biosynthesis genes included genes for ergosterol biosynthesis proteins ERG4/ERG24 and C-14 sterol reductase. The oxidative stress response genes included catalase peroxidase (katG), glutathione transporter, autophagy-related protein (ATG22), and Zn(II)2Cys6 type transcription factor. Catalase peroxidase is an antioxidant enzyme that is active in response to H 2 O 2 accumulation in fungal cell 28 . ATG22 is a vacuolar efflux of amino acids, which helps maintain protein synthesis and viability under nitrogen starvation during the autophagy-associated processes 29 . Nitrogen starvation is related to oxidative stress and membrane peroxidation 30 . Interestingly, the ATG22 homolog of B. bassiana has been reported to be involved in fungal pathogenicity 31,32 .
Bbpc1 and BbThm1 encode Zn(II)2Cys6 type transcription factors in B. bassiana. Bbpc1 plays a role in oxidative stress response, virulence, and conidial and blastospore production 33 . BbThm1 has been reported as a regulator of membrane homeostasis and heat and sodium/lithium dodecyl sulfate (S/LDS) stress 34 . In A. fumigatus, Zn(II)2Cys6 type transcription factor AtrR has been reported to be involved in ergosterol biosynthesis, adaptation in hypoxia condition, and virulence. The cytochrome P450 14-alpha sterol demethylase, Cyp51A is an iron-dependent enzyme and a target of Zn2-Cys6 Transcription Factor (AtrR) in ergosterol biosynthesis 35 . Ergosterol can protect lipid against peroxidation, and the increasing ergosterol level in the cell membrane can inhibit the membrane damage and sustain membrane permeability 36,37 .
Furthermore, a positive correlation between ergosterol biosynthesis and the ability of oxidative stress protection has been demonstrated in Saccharomyces cerevisiae 38 . Therefore, the notably increased expression of stress response genes and ergosterol biosynthesis genes in ΔferS in both iron-replete and iron-depleted conditions might result from the cell acclimation processes. This cell acclimation occurred during oxidative stress conditions, generated from the Fenton reaction in the iron excess and oxidative stress induced by iron starvation. In iron starvation, some iron-dependent mechanisms such as oxidative phosphorylation can be affected and lead to ROS generation 39 . TCA cycle and mitochondrial expansion. In the viewpoint of primary metabolism, under iron-replete and iron-depleted conditions, ΔferS showed higher expression levels of genes involved in TCA cycle and the central carbon metabolism such as citrate synthase (gltA), L-lactate dehydrogenase (ldh) isocitrate lyase (Icl1), and choline/carnitine O-acyltransferase, compared to the wild type (Fig. 6). These results might be consequences www.nature.com/scientificreports/ of mitochondrial expansion and increased iron pool in mitochondria, promoting TCA cycle activity. In this study, the expansion of mitochondria in ΔferS was clearly detected using fluorescence staining, compared to the wild type. The mitochondrial expansion was found under both iron-depleted and replete conditions, suggesting a constitutive pattern (Fig. 7). In contrast, wild-type mitochondria were expanded only under iron depletion (Fig. 7). The wild-type occurrence was consistent with the phenomenon in Saccharomyces cerevisiae, in which the yeast cells can expand the mitochondrial compartments during iron starvation due to diauxic shift condition 40 . On the other hand, the ΔferS mitochondrial expansion occurred regardless of iron availability. The expansion in mitochondrial volume leads to an increase of iron pool in mitochondria, which induces the expression of high-affinity iron transporter such as Fet3 and Ftr1 under iron starvation, as reported in S. cerevisiae 41 . The expansion of the mitochondrial compartment, as well as mitochondrial iron pool, was consistent with the increase in heme and Fe-S cluster-dependent proteins in TCA cycle and respiratory complexes in Ascomycetes 40 .
In conclusion, ΔferS that lacks intracellular siderophore ferricrocin responds to iron-depleted and ironreplete conditions using specific processes. Both iron starvation and iron excess can result in ROS generation. The ferricrocin-free mutant produced oxalate (predicted by transcriptomic data) as an iron chelator. However, the induced expression of CDH could generate H 2 O 2 and promote ROS production (via the Fenton reaction), lipid peroxidation, and ferroptosis. Therefore, the mutant ΔferS might sense the iron excess and the oxidative stress. In turn, the antioxidant-related genes, ergosterol biosynthesis and TCA cycle was up-regulated under both iron-depleted, and iron-replete condition. These responses are potentially analogous to the priming, in which the ΔferS cells are trained for adaptation to severe stresses. Therefore, these increased biological pathways empower the mutant ΔferS during the host infection and lead to higher insect mortality than the wild type in the early phase of infection. XbaI restriction sites are included in the two primers for facilitating the cloning. The ferS fragment was cloned into the vector pCAMBIA1300 at the XbaI site to generate plasmid pCXF3.4. Next, the bar cassette was amplified from the plasmid pCB1534 using the primers Bar-F and Bar-R (Supplemental File S4). The underlined bases indicate the BglII restriction site. The pCXF3.4 was digested with BglII and then ligated with the BglII-restricted bar cassette. Therefore, we obtained the ferS-disruption plasmid pCXFB4.4, of which ferS is interrupted by the bar cassette (Fig. 1). The disruption vector pCXFB4.4 was transformed into Agrobacterium tumefaciens strain EHA 105 using the protocol described previously 42 with some critical modifications 43 .

ΔferS
To determine the integration of the bar cassette in ferS transformants, the genomic DNA was analyzed by Southern and PCR analyses in glufosinate-resistant transformants, compared with the wild type. For Southern analysis, 10 ug of completely BamHI-digested genomic DNA from wild type and ferS transformants were loaded onto 1% agarose gel electrophoresis, and the DNA was transferred and cross-linked to a nylon membrane (Hybond N + ; GE healthcare Bio-sciences, U.S.A.). The 415 bp of ferS fragment was non-radioactively labeled using an alkaline phosphatase-based system (CDP-Star; GE Healthcare Bio-Sciences). The hybridization was performed with the CDP-Star-labelled ferS fragment probe at 55 °C overnight. After high stringency wash, the membrane was incubated with CDP-Star detection solution and exposed to X-ray film (Hyperfilm_ECL; GE Healthcare Bio-Sciences). PCR analysis was performed by three primer pairs. The first pair was used to amplify a ferS region covering the bar integration site and includes Upstart_Fp and FerS4880_Rp (Supplemental File S4). The second and third primer pairs were used to amplify the border regions between the bar cassette and the ferS locus at the bar's 5′ and 3′ ends, respectively. The second pair included Upstart_Fp and Bar-360R. The third pair had Bar-100F and FerS4880_Rp (Supplemental File S4).
HPLC and TLC analysis. Determination of ferricocin in wild type and ΔferS were performed by HPLC analysis, as previously described 13 with some modifications. B. bassiana wild-type or ΔferS was grown on a cellophane sheet laid on top of MM or MM + 10 µM FeSO 4 . The culture was incubated at 25 °C for 20 days. The harvested mycelia were air-dried and extracted with 50 ml of methanol for 2 days. After discarding the mycelia, the methanol fraction was concentrated under reduced pressure to obtain a crude extract. HPLC analysis was conducted using a reverse-phase column (VertiSep HPLC Column; Vertical Chromatography, Thailand) and diode array detector (996 Photodiode Array Detector, Waters). The crude extract was dissolved in methanol to a final concentration of 10 mg ml −1 . Metabolite separation was performed on a VertiSep HPLC Column. Analysis was performed at a flow rate of 0.8 ml min − 1 at 210 nm with a water-acetonitrile step gradient as follows: 0 min/2% acetonitrile, 14 min/60% acetonitrile, 16 min/60% acetonitrile, 19 min/100% acetonitrile, 50 min/100% acetonitrile, 51 min/50% acetonitrile, 60 min/50% acetonitrile and 64 min/2% acetonitrile.
For TLC analysis, the crude mycelial extracts were spotted on a TLC plate (TLC silica gel 60 F254 25 aluminum sheets 20 × 20 cm, Merck, Germany), and developed by a freshly prepared solvent chloroform/methanol/ water (70:24:4) system, as previously reported 44 . Insect bioassay. We have compared the virulence against insects of B. bassiana wild type and ΔferS using beet armyworm (Spodoptera exigua). We performed intrahaemocoelic injection of beet armyworms by using 3 µl of conidial suspension at the density of 1 × 10 7 conidia mL −1 as previously described 14 . Control larvae were injected with saline (0.85% NaCl). The inoculated insect larvae were then placed and fed with the armyworm medium 14 in a plastic container, kept in a large carton at 25 °C. The relative humidity inside the carton was maintained above 80% by using a fine-nozzle spray. There were ten beet armyworm larvae for each treatment, and the experiment was repeated four times. Insect mortality was determined at 24, 48, 72, 96, and 120 h postinoculation (PI).
Comparative analysis of radial growth, conidiation and conidial germination between ΔferS and wild type. For radial growth determination, ferricrocin-deficient mutant ΔferS and the wild type were grown under the iron-depleted and iron-replete conditions, 10 μl of 1 × 10 5 conidia mL −1 were inoculated at the center of MM, MM + BPS, MM + 100Fe and MM + 200Fe. The colony diameter was measured at 3, 5, 7, 9, and 12 days after inoculation. To determine conidiation, the number of conidia produced in a 1 × 1 cm 2 area of culture was determined by using a hemocytometer 14 days after inoculation. www.nature.com/scientificreports/ We conducted the germination assay in slide culture. For each strain, conidia were incubated in 200 µL of 5% PDB (v/v) containing 100 µM BPS (PDB + BPS) or 100 µM FeSO 4 (PDB + 100Fe) broth for a final concentration of 1 × 10 6 conidia mL −1 at 25 °C for 16 h. Conidial germination was determined by counting the number of germinated conidia relative to the total number of conidia in a hemocytometer. There were three replicates for each treatment, and the experiment was repeated three times.
Comparative transcriptomic analysis under iron-depleted and iron-replete conditions. The wild type and ΔferS strains of B. bassiana were cultured in MM + BPS and MM + 200Fe as described above for 4 h. The mycelia were harvested by filtration through cheesecloth and ground to the fine powder in liquid nitrogen, and total RNA was extracted using Ambion TM TRIzol Reagent (Invitrogen, USA). For the four treatments (WT-BPS, WT-Fe, ΔferS-BPS, and ΔferS-Fe), there were two replicates (two sets of total RNAs) for each treatment. Total RNA quality and quantity were measured by NanoDrop One Microvolume UV-Vis spectrophotometer. Poly (A) mRNA was isolated from 75 µg of total RNA using Dynabeads™ mRNA Purification Kit (Thermo Fisher Scientific). Fragment Analyzer (Advanced Analytical Technologues) was used to quantify the concentration and quality of isolated mRNA by DNF-472M33 kit (HS mRNA 15nt). The mRNAs were used to construct RNA libraries using Ion Total RNA-Seq kit v2 protocol (Life Technologies). cDNA was synthesized using SuperScript® III Enzyme Mix, purified by magnetic bead cleanup module, and eluted in 6 µl of pre-heated nuclease-free water. Sequencing adapters and barcode adapters were ligated and amplified using Platinum® PCR SuperMix High Fidelity, Ion Express™ RNA 3′ Barcode primer, and Ion Express™ RNA-Seq Barcode BC primer. RNA libraries were sequenced using on 540™ Kit-OT2 on Ion S5™ XL.
The transcriptomic read data were mapped to the annotated genome of B. bassiana BCC 2660 using Cufflinks version 2.2.1 45 . The genome annotation was conducted using the MAKER annotation pipeline version 2.31.10 46 . The transcriptomic expression profile of each replicate was quantified into Fragments Per Kilobase Million (FPKM). The FPKM values were log-transformed and normalized using geometric normalization. The normalized data were imported to R version 4.0 and analyzed using cummeRbund package version 2.30.0 47 . The pairwise comparison was employed to determine the significant differentially expressed genes (DEGs) for each pair of experiment conditions (p < 0.01). In order to assess to which condition each DEG was specific, the specificity scores of DEGs in 4 treatment conditions (WT-BPS, ΔferS-BPS, WT-Fe, and ΔferS-Fe) were calculated using csSpecificity method in cummeRbund package. For functional assessment, the DEGs between wild type and ΔferS in different conditions were classified into up-regulated and down-regulated groups. The functional enrichment analysis was then conducted using STRING v11 with a false discovery rate < 0.05 48 .