Characterization of the sterol 14 a-demethylases of Fusarium graminearum identifies a novel genus-specific CYP 51 function

CYP51 encodes the cytochrome P450 sterol 14α-demethylase, an enzyme essential for sterol biosynthesis and the target of azole fungicides. In Fusarium species, including pathogens of humans and plants, three CYP51 paralogues have been identified with one unique to the genus. Currently, the functions of these three genes and the rationale for their conservation within the genus Fusarium are unknown. Three Fusarium graminearum CYP51s (FgCYP51s) were heterologously expressed in Saccharomyces cerevisiae. Single and double FgCYP51 deletion mutants were generated and the functions of the FgCYP51s were characterized in vitro and in planta. FgCYP51A and FgCYP51B can complement yeast CYP51 function, whereas FgCYP51C cannot. FgCYP51A deletion increases the sensitivity of F. graminearum to the tested azoles. In ΔFgCYP51B and ΔFgCYP51BC mutants, ascospore formation is blocked, and eburicol and two additional 14-methylated sterols accumulate. FgCYP51C deletion reduces virulence on host wheat ears. FgCYP51B encodes the enzyme primarily responsible for sterol 14α-demethylation, and plays an essential role in ascospore formation. FgCYP51A encodes an additional sterol 14α-demethylase, induced on ergosterol depletion and responsible for the intrinsic variation in azole sensitivity. FgCYP51C does not encode a sterol 14α-demethylase, but is required for full virulence on host wheat ears. This is the first example of the functional diversification of a fungal CYP51.


Introduction
The cytochrome P450 sterol 14a-demethylase (CYP51, syn. ERG11) is an essential enzyme in the biosynthesis of sterols, critical components of cell membranes of all eukaryotic organisms required for the regulation of membrane fluidity and permeability (Parks et al., 1995;Lepesheva & Waterman, 2007). Of the many sterols identified in fungi, ergosterol is the most common and is required for fungal growth (Rodriguez et al., 1985;Weete et al., 2010). Consequently, CYP51 is a widely exploited target for the control of fungal pathogens of humans and plants, with the azole (imidazole and triazole) fungicides being the leading class of antifungals for over three decades (Sheehan et al., 1999). Surprisingly, despite their widespread long-term use and single-site mode of action, incidences of resistance to azole fungicides are rare. More commonly, reductions in sensitivity are reported with cross-resistance within the azole class often incomplete. To date, three mechanisms of resistance to azoles predominate in filamentous fungi: mutations in the CYP51 gene encoding amino acid alterations, resulting in decreased affinity of the protein for inhibitors (Loffler et al., 1997;Sanglard et al., 1998;Wyand & Brown, 2005;Cools et al., 2010); over-expression of the target CYP51 gene, most frequently caused by insertions in the predicted promoter regions (Mellado et al., 2007;Cools et al., 2012); and over-expression of genes encoding efflux pumps (Sanglard et al., 1995;Kretschmer et al., 2009). These mechanisms can combine, and therefore resistance levels in fungal strains are often determined by combinations of CYP51 amino acid alterations, CYP51 gene overexpression and/or enhanced efflux (Perea et al., 2001;Mellado et al., 2007;Cools et al., 2012).
Until recently, CYP51 was thought to exist as a single gene in all phyla. Mammalian genomes, for example, contain one CYP51, with some nonfunctional pseudogenes identified (Rozman et al., 1996). However, increasing genome sequence information has shown that this is not the case in all kingdoms. To date, multiple CYP51 genes have been found in plants, including rice (12), oats (two), tobacco (two) and Arabidopsis thaliana (two) (Lepesheva & Waterman, 2007). Filamentous fungi, particularly Ascomycetes, often possess two or more CYP51 paralogues, for example in Penicillium digitatum (two), Aspergillus fumigatus (two), A. nidulans (two), A. flavus (three), Magnaporthe oryzae (two) and species of Fusarium, including F. verticillioides, F. oxysporum f. sp. lycopersici and F. graminearum (three). Molecular phylogenetic analysis has shown that the CYP51 genes of fungi within the subphylum Pezizomycotina fall into three clades, designated A, B and C (Becher et al., 2011). All species possess a CYP51 in clade B (CYP51B). Species with multiple paralogues carry an additional CYP51 in clade A (CYP51A), with duplications of CYP51A or CYP51B generating the third paralogue in some species, for example A. flavus and A. terreus, respectively. Uniquely, the third CYP51 paralogue in Fusarium species forms a distinct clade, CYP51C. The CYP51C gene is found exclusively in Fusarium species, and is ubiquitous across the genus, as demonstrated by its use as a reliable phylogenetic marker (Fern andez-Ortuño et al., 2010).
Fungi with multiple CYP51s are intrinsically resistant to some azoles, although some remain effective. For example, A. fumigatus is well controlled by itraconazole and voriconazole, whereas fluconazole is ineffective. Species of Fusarium, for example F. solani, are resistant to commonly used medical azoles, including fluconazole, voriconazole and the recently introduced posaconazole (Nucci & Anaissie, 2007). Deletion of CYP51A increases the intrinsic sensitivity to some azoles in M. oryzae (e.g. tebuconazole and prochloraz), A. fumigatus (e.g. fluconazole and ketoconazole) and F. graminearum (e.g. tebuconazole and prochloraz) (Mellado et al., 2005;Liu et al., 2011;Yan et al., 2011). Furthermore, resistance to effective azoles in fungi with multiple CYP51s is most frequently mediated by changes in the CYP51A paralogue. For example, over-expression of CYP51A has been reported for resistant isolates of P. digitatum (Hamamoto et al., 2000;Ghosoph et al., 2007), and mutation of AfCYP51A is the most common mechanism of resistance in A. fumigatus isolates (Diaz-Guerra et al., 2003;Mellado et al., 2005), which, when combined with AfCYP51A over-expression, confers a multiazole-resistant phenotype (Mellado et al., 2007;Snelders et al., 2008).
The control of F. graminearum, the most important pathogen causing Fusarium head blight (FHB) or head scab disease on wheat and barley, is primarily dependent on effective azole fungicides (e.g. tebuconazole, metconazole and prothioconazole). Effective application of azoles reduces the content of the harmful trichothecene mycotoxin deoxynivalenol (DON) in wheat grains (Beyer et al., 2006;Paul et al., 2008). However, an increased level of DON has been detected after treatment with sublethal doses of prothioconazole both in vitro and in planta (Audenaert et al., 2010), and trichothecenes accumulate in grain samples after treatment with tebuconazole (Kulik et al., 2012). In addition, a study using the enhanced green fluorescent protein gene (egfp) as a reporter demonstrated activation of the F. graminearum trichothecene biosynthetic TRI5 gene by sublethal concentrations of tebuconazole (Ochiai et al., 2007); both TRI4 and TRI5 transcript levels were higher after tebuconazole treatment in culture (Kulik et al., 2012). Currently, although stress responses and enhanced secondary metabolism have been proposed, the mechanism(s) responsible for altered mycotoxin production after azole treatment is unknown.
Functional analysis of multiple CYP51s has identified roles additional to primary sterol biosynthesis. In oats, a CYP51 homologue, AsCyp51H10, is dispensable for sterol biosynthesis, but is required for the synthesis of avenacins, antimicrobial compounds unique to the genus Avena (Qi et al., 2006). Heterologous expression in yeast identified AsCyp51H10 as the first CYP51 not classified as a sterol demethylase (Kunii et al., 2012). Recent analysis of the 12 CYP51s of rice has differentiated some genes as CYP51H, a group likely to have functions beyond sterol biosynthesis (Inagaki et al., 2011). Previous work on the CYP51 genes of F. graminearum (FgCYP51) has demonstrated that the deletion of individual FgCYP51 genes can reduce conidiation, but otherwise causes no changes in in vitro morphology, mycelial growth rate or ergosterol content (Liu et al., 2011).
In this study, we have determined the competence of the three paralogous FgCYP51 genes to act as sterol 14a-demethylases by heterologous expression in S. cerevisiae strain YUG37:erg11, which carries a regulatable promoter controlling native CYP51 expression, and found that FgCYP51C cannot complement the CYP51 function of the yeast gene. We generated single (DFgCYP51A, DFgCYP51B and DFgCYP51C) and combined (DFgCYP51AC and DFgCYP51BC) CYP51 deletion mutants and characterized their function in vitro and in planta. We report distinct roles of the three FgCYP51 genes, with the FgCYP51C gene specifically required for full virulence on host wheat ears, but not on Arabidopsis floral tissue or the fruits of apple and tomato.

Materials and Methods
Fungal strains, growth assays and sporulation tests Fusarium graminearum isolates (Table 1) used in this study were routinely cultured on synthetic nutrient-poor agar (SNA) plates, as described previously (Urban et al., 2002). Conidiation assays were performed on SNA plates, and the spores were germinated in YPS medium (0.3% (w/v) yeast extract, 0.3% (w/v) bactopeptone and 20% (w/v) sucrose) (Cuzick et al., 2008). Colony morphology was assayed on potato dextrose agar (PDA) plates in the dark at 22°C. Growth rate tests were conducted in potato dextrose broth (PDB) and Czapek Dox liquid medium (CZD) using an amended 96-well microtitre plate method. Absorbance measurements were made every 8 h for 96 h at 22°C (Fraaije et al., 2007). Carrot agar plates were used for perithecia formation and ascospore production (Ehrenshaft et al., 1995). Approximately 50 asci per strain/mutant were scraped from the plates and physically broken to determine ascospore formation.
Heterologous expression of FgCYP51 genes in S. cerevisiae YUG37::erg11 Heterologous expression of FgCYP51 genes in S. cerevisiae YUG37::erg11 followed the method described by Cools et al. (2010). FgCYP51 genes were amplified from cDNA of F. graminearum isolate Fg1955 (Nicolaisen et al., 2009) with the primers shown in Supporting Information Table S1 designed and cloned into the yeast expression vector pYES3/CT (Invitrogen). Plasmids were transformed into S. cerevisiae strain YUG37:: erg11 (Revankar et al., 2004) using an S.c. EasyComp transformation kit (Invitrogen) with the pYES3/CT vector as a negative control. The complementation efficiency of two independent transformants per construct was screened on Synthetic Drop-out (SD) galactose (GAL) + raffinose (RAF) medium containing 3 lg ml À1 doxycycline (Sigma-Aldrich) by droplet inoculation with 5 ll of cell suspensions (six five-fold dilutions of a starting concentration of 1 9 10 6 cells ml À1 ) for each transformant and incubation for 6 d at 30°C. The growth rate of YUG37::erg11 transformants was determined by inoculation of liquid SD GAL + RAF medium containing 6 lg ml À1 doxycycline with 100 ll of a cell suspension of 1 9 10 6 cells ml À1 (grown for 24 h at 30°C). Absorbance was measured every 4 h for 96 h at 30°C and the growth rate was calculated at the linear growth stage.

Generation of gene deletion mutants
FgCYP51 gene deletion mutants were generated in F. graminearum strain PH-1 (NRRL 31084) (Cuomo et al., 2007), the genome of which has been sequenced, using polyethyleneglycol (PEG)-mediated protoplast transformation of splitmarker PCR fragments (Catlett et al., 2003). The primers used to amplify the flanking sequences for each gene are listed in Table S1. PCR products were transformed into protoplasts of wild-type PH-1 to delete single FgCYP51 genes. Transformants were selected with 75 lg ml À1 hygromycin B or geneticin. The split hygromycin markers of the FgCYP51A and FgCYP51B genes were transformed into the FgCYP51C gene deletion mutants carrying a geneticin selectable marker to generate the double deletion mutants, DFgCYP51AC and DFgCYP51BC. Candidate transformants were screened using 5′ flank and 3′ flank PCRs to confirm the targeted deletion of the genes (Fig. S1). An additional PCR to amplify the 5′ flank and 3′ flank of the nontargeted FgCYP51 gene confirmed that the remaining FgCYP51 genes were not disrupted. Double deletion of FgCYP51A and FgCYP51B was considered to be lethal.

Fungicide sensitivity testing
Fungicide sensitivities were tested using 96-well microtitre plate sensitivity assays (Fraaije et al., 2007). Briefly, spore suspensions of Fusarium isolates with 2 9 10 4 spores ml À1 were added to PDB amended with serial diluted concentrations of fungicides (Table S2). For S. cerevisiae transformants, SD GAL + RAF medium containing 6 lg ml À1 doxycycline (Sigma-Aldrich) and amended with the serial diluted concentrations of fungicides was inoculated with 1 9 10 6 cells ml À1 (Table S2) (Cools et al., 2010). The absorbance was measured at 630 nm using a FLUOstar OPTIMA microplate reader (BMG Labtech GmbH, Offenburg, Germany) after 96 h of incubation at room temperature for F. graminearum and at 30°C for S. cerevisiae transformants. Fungicide sensitivities were determined as the 50% effective concentration (EC 50 ) using a dose-response relationship according to the BMG Labtech Optima Software. Three independent replicates were performed for statistical analysis.

Gene expression analysis
Total RNA was extracted from freeze-dried samples with TRIzol reagent (Invitrogen), followed by a 4 M lithium chloride purification and DNase I treatment. For in vitro FgCYP51 expression analysis, RNA was extracted from mycelia of isolate Fg1955 grown in PDB for 24, 48, 72 and 96 h in the absence of azole, or after treatment with tebuconazole, epoxiconazole or prochloraz at the calculated EC 50 at 24 h of incubation. In vitro gene expression of DFgCYP51 mutants used RNA extracted after 48 h of growth in PDB, compared with expression in wild-type parent strain PH-1. For in planta expression analysis, RNA was extracted from inoculated wheat ears harvested at 7 d post-inoculation (dpi). First-strand cDNA was prepared using the SuperScript III first-strand synthesis system with oligo(dT) (Invitrogen). Real-time PCR amplifications were performed on an ABI 7500 sequence detection system (Applied Biosystems, Foster City, CA, USA) using SYBR Green I fluorescent dye detection. The relative quantities (RQs) were calculated using the 2 ÀDDC t method with F. graminearum actin and b-tubulin as reference genes and the statistical analysis of three biological replicates. The primers used for quantitative PCR are listed in Table S1.

Sterol extraction and analysis
Fusarium graminearum strains PH-1 and DFgCYP51 mutants were collected after growth on PDB medium at 23°C for 48 h. Nonsaponifiable lipids were extracted as described previously (Kelly et al., 1995). Briefly, pellets were resuspended in 2.5 ml of methanol, followed by the addition of 1.5 ml of 60% (w/v) KOH and 1 ml of 5 mg ml À1 pyrogallol (in methanol). Samples were refluxed at 90°C for 2 h, and then extracted twice with hexane and dried in a rotary evaporator (Heto, Allerod, Denmark). Samples were derivatized by the addition of 100 ll of 90% bis(trimethylsilyl)-trifluoroacetamide (BSTFA) -10% trimethylsilyl (TMS) (Sigma-Aldrich) and 50 ll of anhydrous pyridine (Sigma-Aldrich) and heating for 2 h at 80°C. Gas chromatography-mass spectrometry was performed using a VG12-250 mass spectrometer (VG Biotech, Mumbai, India) with splitless injection and chemical ionization. Data were analysed using an MSD Enhanced ChemStation (Agilent Technologies), and TMS esters of individual sterols were identified by reference to relative retention times, mass ions and fragmentation patterns. Data present the relative composition in total sterol in each sample. Three biological replicates were performed.

Plant disease assays
Wheat ear infection, gene expression and DON concentration measurement Wheat (Triticum aestivum L., cv Bobwhite) plants were grown and infected by point inoculation as described previously (Urban et al., 2003). The 11th and 12th flowering spikelets counted from the stem were inoculated with 5 ll of 5 9 10 4 spores per spikelet or mock inoculated with sterile water. The number of infected spikelets was recorded at 3, 7, 11, 16 and 21 dpi. For measurements of DON concentration and in planta gene expression assays, 14 flowering spikelets were inoculated per ear. The infected spikelets were harvested at 7 dpi for gene expression analyses and at 10 dpi for DON measurements. Samples were lyophilized and infected spikelets were ground to a fine powder in liquid nitrogen using a mortar and pestle. The powder was resuspended in water as described previously (Urban et al., 2003). After incubation at 30°C for 30 min with shaking at 80 rpm, the solutions were centrifuged and the supernatants were analysed quantitatively for DON content using an EZ-Quant Vomitoxin (DON) plate kit (Diagnostix, Ontario, Canada), according to the manufacturer's instructions. Three wheat ears of different wheat plants were inoculated for each strain/mutant in each replicate experiment. Three independent replicate experiments (total of nine ears per strain/mutant inoculation) were carried out and analysed statistically.
Arabidopsis floral inoculations Arabidopsis plants (ecotype Landsberg) were grown and inoculated as described previously (Cuzick et al., 2008). Five-week-old flowering plants were spray inoculated with c. 0.5 ml of conidial suspensions of 1 9 10 5 spores ml À1 of wild-type PH-1 or DFgCYP51 strains, and incubated in a 100% relative humidity box. Each isolate was sprayed onto three plants per experiment, with two experimental replicates performed (total of six plants per strain/mutant inoculation), with sterile water as a negative control. Disease symptoms on apical flowers and siliques were assessed using the Fusarium-Arabidopsis disease (FAD) scoring system (Urban et al., 2002), with the modification of combining old and new silique scores. Floral and silique scores were added to give the total FAD score shown in Table S3.
Infection assays on apple fruit sections The ability of DFgCYP51 mutants to colonize apple fruit sections relative to strain PH-1 was tested on cv Braeburn. After surface sterilization with 70% ethanol, apples were sliced into 8-mm-thick sections and inoculated in three discrete areas with a 5-ll droplet of a spore suspension containing 5 9 10 5 ml À1 conidia of wild-type isolate PH-1, sterile water control or a DFgCYP51 gene mutant. In each experiment, three apple slices were inoculated for each mutant.
Three independent experiments were carried out. Apple slices were sealed in a 9-cm Petri dish and incubated at 28°C for 6 d.
Infection assays on tomato fruits To assay the colonization of ripe tomato fruits (cultivar Encore) by DFgCYP51 mutants, three wounds were made with a hyperdermic needle (BD microlance 23G 10.6 9 25) on 70% ethanol surface-sterilized tomatoes. Fruits were inoculated with 3 9 5 ll of 5 9 10 5 ml À1 conidia suspensions and incubated at 28°C in a sealed plastic container at 100% humidity for 6 d (Urban et al., 2003). For each strain/ mutant, three tomatoes were inoculated in each of three independent experiments. Mock inoculations were with sterile distilled water.

Statistical analysis
Data were analysed from two or three independent replicates using ANOVA. Following a significant F-test result (P < 0.0001 for plant infection, P < 0.05 for fungicide sensitivity tests and in vitro/in planta gene expression), means of independent replicates were compared using least-significant differences (LSD). No transformation of the data was required. P < 0.05 was considered to be significant. Greek letters were used to mark statistically significant differences (P < 0.05).

FgCYP51A expression is induced by azole fungicides in vitro
In the absence of fungicide, transcript levels of FgCYP51A and FgCYP51B were highest at 72 h of incubation ( Fig. 1).
FgCYP51C expression did not change significantly across the time course. Twenty-four hours after treatment (48 h after inoculation) with subinhibitory concentrations of azoles, the expression of FgCYP51A was induced up to 30-fold (prochloraz and tebuconazole) or 100-fold (epoxiconazole) relative to the untreated control. FgCYP51B and FgCYP51C transcript levels did not change significantly after azole treatment.

New Phytologist
Heterologous expression of FgCYP51 genes in S. cerevisiae strain YUG37::erg11 FgCYP51A, FgCYP51B and FgCYP51C were expressed in S. cerevisiae strain YUG37::erg11, in which native CYP51 gene (ScCYP51) expression is under the control of a doxycyclinerepressible promoter (Revankar et al., 2004). FgCYP51A and FgCYP51B genes were able to complement ScCYP51 function; however, no growth was observed with transformants expressing FgCYP51C on doxycycline amended medium, a phenotype similar to the empty vector control (Fig. 2). Yeast transformants expressing FgCYP51A grew more slowly than those expressing FgCYP51B in the presence of doxycycline (Table 2). Yeast transformants expressing FgCYP51A were less sensitive to epoxiconazole than to tebuconazole and prochloraz. However, the sensitivity of FgCYP51B transformants was similar to all three azoles (Table 3). There was no difference in cycloheximide sensitivity, a non-azole fungicide, of yeast transformants expressing FgCYP51A relative to FgCYP51B.
In vitro growth characteristics of FgCYP51 deletion strains There were no differences in growth rate between mutants DFgCYP51A, DFgCYP51B, DFgCYP51C, DFgCYP51AC and DFgCYP51BC (Table 4) and the wild-type parental strain PH-1 in PDB or CZD (Table 4). Nor were there morphological differences on SNA plates under UV light (data not shown), or for single FgCYP51 deletion strains on PDA incubated in the dark (Fig. 3a), confirming the findings of Liu et al. (2011). However, the morphology of double gene deletion mutants uniquely generated in this study (DFgCYP51AC and DFgCYP51BC) on PDA incubated in the dark (Fig. 3a) was different. Growth was slower and the colony pigment was yellow rather than the pink of wild-type and single gene deletion strains. Although DFgCYP51AC> and DFgCYP51BC strains produced around half the number of conidia of the PH-1 strain, conidial morphologies (data not shown) and germination rates were normal (Table 4).
FgCYP51B is essential for ascospore production Perithecia were produced by all strains with no morphological differences relative to PH-1. However, 30 d after inoculation, no ascospores oozed from perithecia of DFgCYP51B and DFgCYP51BC and, when perithecia were physically broken, no ascospores were found (Fig. 3b). Ascospores produced by DFgCYP51A, DFgCYP51C and DFgCYP51AC showed wild-type morphology and germinated normally (data not shown).   According to the sensitivity of DFgCYP51A and DFgCYP51B mutants, the seven azole fungicides tested can be divided into two groups (Table 5). Sensitivity to group 1, which includes tebuconazole, epoxiconazole, propiconazole and imazalil, is only affected significantly by FgCYP51A deletion. For example, DFgCYP51A strains are > 30-fold more sensitive to epoxiconazole than are wild-type and DFgCYP51B strains. Group 2 includes metconazole, prochloraz and difenoconazole. Deletions of both FgCYP51A and FgCYP51B increased the sensitivity to these compounds. There was no significant difference in the sensitivity of single FgCYP51C gene deletion mutants to all the azoles tested relative to the wild-type PH-1, in contrast with previous reports (Liu et al., 2011). In addition, there were no significant differences in azole sensitivity between DFgCYP51A and DFgCYP51AC, confirming that FgCYP51C has no effect on azole sensitivity.   Conidia produced from 7 d of incubation on synthetic nutrient-poor agar (SNA) plate with 10 4 conidia inoculation. c Percentage of germinated conidia was measured after 16 h of incubation in YPS medium at room temperature. *Data were analysed using ANOVA (P < 0.05, F test). Different Greek letters are used to mark statistically significant differences between strains/mutants (P < 0.05, least-significant difference (LSD)).

New Phytologist
Deletion of FgCYP51B and FgCYP51C decreases sensitivity to fenhexamid, an erg27 inhibitor Testing the sensitivity to other ergosterol biosynthesis inhibitors (EBIs , Table 6) revealed no differences in amphotericin B sensitivity, an inhibitor that binds to ergosterol (Gray et al., 2012), between PH-1 and the FgCYP51 deletion mutants, suggesting no differences in ergosterol content between strains. DFgCYP51B and DFgCYP51C mutants were significantly less sensitive to fenhexamid, an erg27 inhibitor, relative to the wildtype strain. The DFgCYP51BC double mutants were least sensitive. There was no effect of FgCYP51A deletion on fenhexamid sensitivity. For all single FgCYP51 deletion mutants, there was no difference in sensitivity to the erg1 inhibitor terbinafine relative to the wild-type strain PH-1. However, the double gene deletion mutants, DFgCYP51AC and DFgCYP51BC, were significantly less sensitive to terbinafine. There were no differences in sensitivity between mutant strains and the wild-type to the other EBIs tested, or to the b-tubulin assembly inhibitor carbendazim.

Ó 2013 The Authors
New Phytologist Ó 2013 New Phytologist Trust New Phytologist (2013) 198: 821-835 www.newphytologist.com 5,7-dienol. Although the composition of sterol intermediates differed between the DFgCYP51 mutants and PH-1, there were no substantial differences in ergosterol content, a finding consistent with the similar in vitro growth rates of the various DFgCYP51 mutants and PH-1.

Expression of genes in the ergosterol biosynthesis pathway
The relative expression of genes involved in the ergosterol biosynthesis pathway was measured after 48 h of incubation in PDB (Fig. 4a). FgCYP51A gene expression was almost 10-fold ) are means of EC 50 of two transformants from three independent replicates AE SD. Mean resistance factor (RF), representing the sensitivity change of each isolate, was calculated as the fold change in EC 50 compared with wild-type PH-1. *Data were analysed using ANOVA (P < 0.05, F test). Different Greek letters are used to mark statistically significant differences between strains/mutants (P < 0.05, least-significant difference (LSD)).

Research
New Phytologist higher in DFgCYP51B and DFgCYP51BC mutants relative to PH-1. When present, the expression of FgCYP51B and FgCYP51C was unchanged in the DFgCYP51 mutants relative to PH-1. Of the other genes involved in ergosterol biosynthesis, there were no significant differences in the expression of ERG2, ERG3A or ERG4 in any mutants relative to PH-1 (Fig. S2). ERG7 and ERG27 expression was significantly higher in the DFgCYP51A and DFgCYP51BC mutants. The expression of ERG1, encoding squalene epoxidase, the first enzyme in ergosterol biosynthesis, was significantly lower in the single DFgCYP51A and DFgCYP51C, and double DFgCYP51AC and DFgCYP51BC mutants, which are less sensitive to terbinafine, an erg1 inhibitor (Table 6). ERG3B expression, encoding a C-5 (6) desaturase, was lower in DCYP51AC and DCYP51BC mutants. There were no significant differences in actin or btubulin gene expression (Fig. S3).

FgCYP51C is required for full virulence on wheat ears
In wheat ear infection assays, four of the FgCYP51 gene deletion mutants showed reduced virulence relative to wild-type PH-1 (Fig. 5). There were no significant differences (P < 0.05, LSD) in DFgCYP51A infection over the time course relative to the wild-type. This suggests that FgCYP51A is not essential for wheat ear infection. The percentage ear infection caused by DFgCYP51B was significantly less than that of the wild-type at 16 dpi (P < 0.05, LSD), but infection levels were similar at 21 dpi, indicating that FgCYP51B gene deletion delays infection. Deletion of FgCYP51C decreased significantly ear infection at both 16 dpi and 21 dpi (P < 0.05, LSD), suggesting that FgCYP51C is required for full virulence on wheat ears. The virulence of DFgCYP51AC was similar to DFgCYP51C, supporting the assertion that FgCYP51A has no effect on infection. The strongest reduction in wheat ear infection was measured for DFgCYP51BC mutants, demonstrating a strong combined effect of the absence of both FgCYP51C and FgCYP51B on virulence. In addition, wheat spikelets inoculated with DFgCYP51C, DFgCYP51AC and DFgCYP51BC mutants contained five-fold less DON at 10 dpi than those inoculated with PH-1 (Fig. 7). There was no significant difference in the amount of DON in wheat ears inoculated with DFgCYP51A and DFgCYP51B (Fig. 6). The grains in wild-type PH-1, DFgCYP51A and DFgCYP51B infected wheat ears were shrivelled and bleached (Fig. 7c). Not all the grains of ears inoculated with DFgCYP51C, DFgCYP51AC and DFgCYP51BC were infected, although some spikelets were bleached (Fig. 7a). Rachises infected with PH-1, DFgCYP51A and DFgCYP51B were all bleached, whereas those infected with DFgCYP51C, DFgCYP51AC and DFgCYP51BC were a darkbrown colour (Fig. 7b). In addition, grains in DFgCYP51C, DFgCYP51AC and DFgCYP51BC infected ears had a pink pigment. This was absent from ears infected with PH-1 and the DFgCYP51A and DFgCYP51B mutants (Fig. 7c).

Gene expression in infected wheat ears
Similar to the in vitro assay, FgCYP51A expression was highest in wheat ears inoculated with DFgCYP51B and DFgCYP51BC (Fig. 4b). Single FgCYP51A and FgCYP51C gene deletions did not affect the expression of FgCYP51A and FgCYP51B in planta (Fig. 4b). The expression of TRI genes, involved in DON production, was measured in inoculated wheat ears (Fig. S4). Although the RQs of TRI4, TRI5 and TRI15 transcripts varied between DFgCYP51 mutants, these differences were not significant. There were no significant differences in ERG1, ERG7 and

Virulence on dicotyledonous hosts: Arabidopsis, tomato fruit and apple fruit sections
In contrast with wheat ear infection, there were no significant differences in disease values between the wild-type and DFgCYP51A, DFgCYP51B and DFgCYP51C on Arabidopsis plants, a host which does not require DON mycotoxin for colonization (Cuzick et al., 2008). Both DFgCYP51AC and DFgCYP51BC were significantly less virulent than wild-type PH-1 on Arabidopsis (Fig. 8), consistent with the virulence of these mutants on wheat ears. In addition, 8 d after inoculation, less aerial mycelia were visible and more green siliques were present on Arabidopsis plants infected with DFgCYP51AC and DFgCYP51BC in comparison with PH-1 and DFgCYP51A, DFgCYP51B and DFgCYP51C (Fig. S5). All mutants infected tomato fruits and apple fruit sections, with phenotypes similar to that of wild-type PH-1 (Figs S6, S7).

Discussion
The recent increase in genome sequence information has revealed that many fungi, particularly ascomycetes, carry more than one copy of the azole fungicide target encoding gene, CYP51. In the Pezizomycotina, a subphylum of the ascomycetes that includes important pathogens of animals and plants, such as Aspergillus spp., M. oryzae, Mycosphaerella graminicola and Fusarium spp., CYP51 paralogues have been classified into three phylogenetic clades, designated A, B and C (Becher et al., 2011), with the CYP51C paralogue only found in Fusarium species. The CYP51C clade is ubiquitous across the genus Fusarium and, consequently, can be used as a reliable phylogenetic marker for the identification of different species (Fern andez-Ortuño et al., 2010). A previous study has reported that the deletion of individual F. graminearum CYP51 genes (FgCYP51A, FgCYP51B or FgCYP51C) has no effect on colony morphology, vegetative growth rate or ergosterol content, although conidiation is reduced in all mutants and deletion of the FgCYP51A and FgCYP51C genes increases azole sensitivity (Liu et al., 2011). In this study, by heterologous expression in yeast and systematic characterization of the impact of individual (FgCYP51A, FgCYP51B, FgCYP51C) and double (FgCYP51AC and FgCYP51BC) gene deletions on in vitro growth, fungicide sensitivity, total sterol composition and virulence on wheat ears and other plants, we describe distinct roles for the FgCYP51 paralogues of F. graminearum.
FgCYP51B encodes the primary sterol 14a-demethylase and is essential for ascospore production The three FgCYP51 genes of F. graminearum isolate Fg1955 were heterologously expressed in S. cerevisiae strain YUG37::erg11, which has been used previously to assess the impact of CYP51 mutations on Mycosphaerella graminicola azole sensitivity and enzyme function (Cools et al., 2010(Cools et al., , 2011 and to analyse the role of AfCYP51A and AfCYP51B genes in A. fumigatus (Martel et al., 2010b). FgCYP51A and FgCYP51B were able to substitute ScCYP51 function, whereas FgCYP51C could not (Fig. 2). In addition, transformants expressing FgCYP51B grew faster than those expressing FgCYP51A (Table 2). This suggests that the FgCYP51B protein is a more effective sterol 14a-demethylase than FgCYP51A in yeast.
FgCYP51 gene deletion did not impact on ergosterol content, in accordance with Liu et al. (2011). However, the abundance of intermediate sterols was different in all mutants relative to the wild-type. The specific accumulation of eburicol in DFgCYP51B and DFgCYP51BC suggests that the overall eburicol demethylation activity is perturbed in mutants lacking FgCYP51B. This is in contrast with those deficient in FgCYP51A or FgCYP51C activity, although the product of CYP51, 4,4-dimethyl ergosta-8,14,24(28)-trienol, was significantly less abundant in all DFgCYP51 mutants relative to the wild-type. These data are consistent with studies of A. fumigatus. Deletions of either AfCYP51A or AfCYP51B blocked C14-demethylation, but far more eburicol accumulated in the AfCYP51B mutant than the AfCYP51A mutant (Alcazar-Fuoli et al., 2008). Furthermore, substrate binding studies of AfCYP51 proteins expressed in Escherichia coli detected strong binding with purified AfCYP51B using eburicol and lanosterol, in contrast with AfCYP51A (Warrilow et al., 2010). As a consequence of accumulated eburicol, two additional novel 14-methylated sterol intermediates (4,4,14-trimethyl ergosta-trienol and 4,4,14-trimethyl ergostadienol) were detected in DFgCYP51B and DFgCYP51BC (Fig. S8). These data suggest that Pezizomycotina CYP51B, including FgCYP51B, is central to effective sterol C14-demethylation.
Ascospores forcibly ejected from mature perithecia, formed by the overwintering fungus on field debris, are the primary source of inoculum for F. graminearum epidemics (Parry et al., 1995;Trail et al., 2005). In this study, no ascospores were formed in DFgCYP51B and DFgCYP51BC mutants (Fig. 3b), although all the FgCYP51 gene deletion mutants produced superficially normal perithecia. This finding demonstrates that FgCYP51B is specifically required in the development of the sexual stage, a role that cannot be fulfilled by the up-regulation of FgCYP51A. Similarly, the delayed colonization of wheat ears by DFgCYP51B suggests that FgCYP51A cannot fully complement FgCYP51B function during wheat infection.
FgCYP51A encodes an inducible sterol 14a-demethylase that determines azole sensitivity In the absence of fungicide treatment, the relative transcript quantities of FgCYP51A and FgCYP51B were highest at 72 h of incubation, decreasing at 96 h during growth in rich medium (PDB). This pattern of expression is coincident with rapid fungal growth, which is linear between 24 and 60 h of incubation, and into stationary phase after 72 h of incubation (data not shown). The relative transcript quantities of FgCYP51A changed most over this time course. Previous studies have shown an increase in CYP51A gene expression after azole treatment and CYP51B deletion in vitro in F. graminearum (Liu et al., 2010;Becher et al., 2011) and M. oryzae (Yan et al., 2011). We report similar results, with FgCYP51A expression induced over 100-fold by azoles in vitro, and c. 10-fold by FgCYP51B deletion both in vitro and in planta. The enhanced transcription of FgCYP51A on exposure to azoles suggests that this gene is not only responsive to chemical or genetic perturbation of FgCYP51B activity, but also other stresses induced by fungicide treatment. For example, in S. cerevisiae, ScCYP51 expression is higher during growth on glucose, in the presence of haem, under oxygen-limiting growth conditions and during exposure to anaerobic conditions (Turi & Loper, 1992).
Amino acid substitutions and CYP51 over-expression are the most common mechanisms of resistance to azoles in filamentous ascomycetes. In fungi with multiple CYP51s, it is the CYP51A paralogue that is most commonly altered, for example in A. fumigatus and P. digitatum (Ghosoph et al., 2007;Mellado et al., 2007). In addition, single gene deletion has confirmed that AfCYP51A is involved in intrinsic azole resistance, for example to fluconazole in A. fumigatus (Mellado et al., 2005), and azole affinity studies have shown that the AfCYP51A protein has lower affinity than AfCYP51B for a wide range of azoles (Warrilow et al., 2010). In F. graminearum, it has been suggested previously that different azole fungicides target different FgCYP51s (Liu et al., 2011). In this study, deletion of FgCYP51C, either alone or with FgCYP51A (DFgCYP51AC), had no effect on azole sensitivity. Deletion of FgCYP51B caused an increase in sensitivity to some azoles, with sensitivities to metconazole and prochloraz particularly affected. These data conflict with those of Liu et al. (2011), who reported increased sensitivity to tebuconazole and prochloraz of DFgCYP51C mutants, with FgCYP51B deletion having no impact on azole sensitivity. The reason for this discrepancy is unclear, although the different methodologies used for fungicide sensitivity testing and the different origins of the strains used in gene deletion studies may have contributed. However, clearly, in this study, an interaction of azole fungicides with FgCYP51B is consistent with the assertion that this paralogue is the primary sterol 14a-demethylase in F. graminearum. In both studies, however, the deletion of FgCYP51A increased the sensitivity to all azoles tested and, particularly, epoxiconazole. Considering that F. graminearum isolate Fg1955 is least sensitive to epoxiconazole, FgCYP51A expression is inducible on azole exposure and yeast transformants expressing FgCYP51A are least sensitive to epoxiconazole (Table 3), it can be concluded that the intrinsically lower sensitivity to some azoles in F. graminearum is primarily determined by FgCYP51A.
FgCYP51C is a novel genus-specific CYP51 gene required for full virulence on wheat ears In the S. cerevisiae heterologous expression system, FgCYP51A and FgCYP51B were able to substitute for ScCYP51 function. By contrast, FgCYP51C could not complement ScCYP51. This suggests that FgCYP51C cannot function as a sterol 14a-demethylase. This loss or diversification of function is probably caused by substitutions in conserved putative substrate recognition sites (SRSs) of FgCYP51C. The predicted FgCYP51 amino acid sequences are sufficiently identical (over 40%) to be considered as members of the same P450 family (Liu et al., 2011). Analysis of residues conserved in eukaryotic CYP51s (Fig. S9, Lepesheva & Waterman, 2011) identified two residues (N304 and T305) unique to FgCYP51C. Although the importance of these residues in the function of FgCYP51C is unknown, substitutions T315N or S316T of rat CYP51, equivalent to N304 and T305, caused significant reductions in lanosterol demethylase activity (Nitahara et al., 2001). Deletion of FgCYP51C had no impact on in vitro fungal morphology, growth rate, conidiation and spore germination at almost all vegetative stages, perithecia production, ascospore formation or azole sensitivity. In addition, there was no difference in eburicol or ergosterol content in DFgCYP51C mutants. However, DFgCYP51C mutants had less 4,4-dimethyl ergosta-8,14,24(28)-trienol, the product of CYP51, and accumulated the sterol intermediates episterol, ergosta-5,7,24(28)-trienol and ergosta-5,7-dienol, products of ERG2, ERG3 and ERG4 activity, respectively (Fig. S8). This suggests that FgCYP51C can impact indirectly on sterol 14a-demethylation, ERG2, ERG3 and ERG4 activity. There was no difference in ERG2, ERG3A, ERG3B and ERG4 gene expression in single DFgCYP51C mutants relative to wild-type PH-1 in vitro. However, the ERG3B gene was expressed less in both DFgCYP51AC and DFgCYP51BC mutants, which grew more slowly and produced less aerial mycelia on rich medium in the dark and when inoculated on wheat ears and Arabidopsis floral tissues. The CYP51 gene is required for aerobic viability in S. cerevisiae, C. albicans and C. glabrata. In a CYP51-deficient mutant, aerobic growth can be restored by null mutation or deletion of ERG3 (Bard et al., 1993;Kelly et al., 1993Kelly et al., , 1995Geber et al., 1995). However, to date, there is no evidence for ERG3-mediated azole resistance in filamentous fungi.
After inoculation with the DFgCYP51C, DFgCYP51AC and DFgCYP51BC mutants, the number of infected spikelets per wheat ear was reduced significantly relative to inoculations with wild-type PH-1, DFgCYP51A and DFgCYP51B. In addition, not all the grain in bleached spikelets had a rough, shrivelled appearance, although infected grains were pink in wheat ears inoculated with DFgCYP51C, DFgCYP51AC and DFgCYP51BC, rather than the grey in plants inoculated with PH-1, DFgCYP51A and DFgCYP51B (Fig. 7c). The rachises of wheat heads infected with DFgCYP51C, DFgCYP51AC and DFgCYP51BC were dark brown, in contrast with the bleached rachises of plants inoculated with PH-1, DFgCYP51A and DFgCYP51B. The blocking of fungal growth from inoculated spikelets to adjacent spikelets is correlated with an unidentified brown substance deposited in the rachis node in the additional wheat line CS-7EL, which carries resistance to FHB on the long arm of chromosome 7E(7EL) (Miller et al., 2011). These data suggest that hyphal development of DFgCYP51C, DFgCYP51AC and DFgCYP51BC during wheat ear infection is impaired.
In contrast with wheat ear infection, there were no differences in virulence between wild-type PH-1 and DFgCYP51C on Arabidopsis, in which the trichothecene mycotoxin DON is not required for fungal infection (Cuzick et al., 2008), although colonization by the double DFgCYP51AC and DFgCYP51BC mutants was impaired. This suggests that altered DON production is New Phytologist (2013) 198: 821-835 Ó 2013 The Authors New Phytologist Ó 2013 New Phytologist Trust www.newphytologist.com responsible for the decreased virulence of the DFgCYP51C mutant on wheat ears. However, unlike mutants unable to produce DON, for example TRI5 gene mutants, which cause only discrete eye-shaped lesions on spikelets and fail to infect the rachis (Cuzick et al., 2008), DFgCYP51C, DFgCYP51AC and DFgCYP51BC mutants are able to infect beyond the inoculated spikelet. In addition, TRI4 and TRI5 gene expression is not altered significantly in wheat ears inoculated with DFgCYP51C, DFgCYP51AC and DFgCYP51BC relative to the wild-type. Therefore, rather than a biosynthetic requirement for DON biosynthesis, FgCYP51C is likely to have an indirect effect. A relationship between the sterol and trichothecene biosynthesis pathways has been reported previously. They share a common precursor, farnesol pyrophosphate, and the global regulator TRI6, located in the core TRI gene cluster. For example ERG9 (FGSG_09381), encoding squalene synthase, the first step of sterol biosynthesis, was down-regulated in the DTRI6 strain under nitrogen-deprived conditions, and ERG25 (FGSG_10666) was up-regulated, although there was no impact on FgCYP51C expression (Nasmith et al., 2011).

Conclusion
We have identified distinct functions of the three CYP51 paralogues of F. graminearum. FgCYP51B, as the most conserved CYP51 gene in all fungi, encodes the enzyme primarily responsible for sterol 14a-demethylation, a role essential for ascospore formation. FgCYP51A, found in many human and agricultural pathogens, is induced by azoles and environmental stress, encodes a sterol 14a-demethylase with the capacity to compensate for disruption of FgCYP51B function, and is responsible for intrinsic variation in sensitivity to different azoles. FgCYP51C, a Fusarium-specific CYP51 gene, no longer functions as a sterol 14a-demethylase, but rather is specifically required for full virulence on host wheat ears. This is the first example of functional diversification of a fungal CYP51 gene.

Supporting Information
Additional supporting information may be found in the online version of this article.        The proposed sterol biosynthesis pathway in Fusarium graminearum. The sterol pathway based on compounds detected in the wild-type PH-1 and FgCYP51 gene deletion mutants incubated in PDB media for 48 h. The proposed pathway in wild-type PH-1 is shown in solid arrows. The pathway after FgCYP51B deletion is shown with dashed arrows in the dashed box. The two additional sterols, 4,4,14-trimethyl ergosta-trienol and 4,4,14-trimethyl ergosta-dienol, in the dashed box were identified as TMS esters with m/z 510 and m/z 512.   Table S3 The Fusarium-Arabidopsis disease scoring system Please note: Wiley-Blackwell are not responsible for the content or functionality of any supporting information supplied by the authors. Any queries (other than missing material) should be directed to the New Phytologist Central Office.
New Phytologist is an electronic (online-only) journal owned by the New Phytologist Trust, a not-for-profit organization dedicated to the promotion of plant science, facilitating projects from symposia to free access for our Tansley reviews. Regular papers, Letters, Research reviews, Rapid reports and both Modelling/Theory and Methods papers are encouraged. We are committed to rapid processing, from online submission through to publication 'as ready' via Early View -our average time to decision is <25 days. There are no page or colour charges and a PDF version will be provided for each article.
The journal is available online at Wiley Online Library. Visit www.newphytologist.com to search the articles and register for table of contents email alerts.