Identification and characterization of a pathogenicity-related gene VdCYP1 from Verticillium dahliae

Verticillium dahliae is a phytopathogenic fungus that causes vascular wilt disease in a wide variety of crop plants, thereby causing extensive economic loss. In present study, one V. dahliae T-DNA mutant M01C06 showed the pathogenicity loss on cotton, and the expression of a flanking gene encoding cytochrome P450 monooxygenase (P450, VdCYP1) was strongly repressed. P450s of fungi could affect the fungal pathogenicity by involving in the synthesis of secondary metabolites. However, there was no report about the pathogenic function of P450s in V. dahliae. VdCYP1 gene deletion and complementation experiments confirmed that VdCYP1 was the pathogenicity-related gene in V. dahliae. A comparison of culture supernatants of the VdCYP1 deletion mutants and wild-type strains indicates that at least 14 kinds of secondary metabolites syntheses were affected due to VdCYP1 gene deletion. One of these compounds, sulfacetamide, had the ability to induce the necrosis and wilting symptoms in cotton. Above results indicate that VdCYP1 could participate in pathogenesis by involving the secondary metabolism in V. dahliae, such as the compound sulfacetamide. In conclusion, VdCYP1 acts as an important pathogenicity-related factor to involve in secondary metabolism that likely contributes to the pathogenic process in V. dahliae.

To identify pathogenicity-related genes in V. dahliae, the virulence of 1,344 transformants was assayed on cotton plants, and 69 transformants were identified to display a significant reduction or loss of pathogenicity. One mutant strain, M01C06, the T-DNA insertion affected the expression of a candidate pathogenicity-related gene encoding a cytochrome P450 monooxygenase (named VdCYP1; VDAG_05890 in VdLs.17 genome), resulting in a near-complete loss of pathogenicity on cotton plants. P450s are heme-thiolate proteins distributed across the biological kingdoms, and involved in the metabolism of a diverse array of endogenous and xenobiotic compounds [23][24][25] . Especially, P450s extensively participate in a wide variety of physiological reactions in fungi that contribute to the fitness and fecundity of fungi in various ecological niches 26 . Analysis of fungal genomes has revealed the presence of a surprisingly large number of P450s which often involve in oxygenation steps in the biosynthesis of fungal secondary metabolites 23,27,28 . Moreover, the P450 family shows significantly expansion in several plant pathogens by the comparative genomics analysis 23,25 , and numerous P450s are significantly regulated during host-pathogen interactions to involve in fungal pathogenic processes 29,30 , such as in Heterobasidion annosum sensu lato 31 , Moniliophthora perniciosa 32 , Botrytis cinerea 33 , and Fusarium spp. 34 . Taken together, the previous reports suggested that the P450s probably involve in the secondary metabolism and plays an important role in the pathogenesis. However, the involvement of P450s in V. dahliae pathogenicity has not been reported.
In this study, we mainly focused on determining whether VdCYP1 is a pathogenicity-related gene in V. dahliae. The expression of VdCYP1 was disturbed by the T-DNA insertion in the M01C06 mutant, which showed a loss of virulence on the cotton plant. Further gene deletion and complementary experiments confirmed that gene VdCYP1 contribute to virulence during infection. And the secondary metabolites analysis indicates that VdCYP1 may be involved in the secondary metabolism, by which to affect the pathogenesis on cotton.

Results
Identification of a non-pathogenic mutant of V. dahliae. Using the root-dipping method with susceptible cotton (Gossypium hirsutum cv. Junmian 1), mutants showing changes in virulence were isolated from a small T-DNA random insertion library that contained 1,344 mutants. The library was generated from a highly aggressive defoliating strain of V. dahliae, Vd991, by using ATMT. In total, 69 mutants exhibited reduced virulence relative to the wild-type strain, and 16 of them showed significantly less pathogenicity to cotton (see Supplementary Fig. S1). Further assessment of their pathogenicity on cotton plants confirmed that one of the mutants, termed M01C06, showed near-complete loss pathogenicity to cotton, while the wild-type Vd991 strain caused severe necrosis and wilting of the leaves (Fig. 1a). Investigation of vertical shoot sections showed that the M01C06 mutant almost lost all capacity for vascular discoloration when compared to the symptoms associated with wild-type infection (Fig. 1b). The disease index (DI) of cotton inoculated with the M01C06 mutant was significantly lower than that of the wild-type strain. The M01C06 DI was 8.78 ± 3.07, similar to the value for mock-inoculation (which is caused by cotyledon senescence under normal conditions), while the wild-type Vd991 strain showed a DI of 68.05 ± 3.62 after inoculation (Fig. 1c).
The colony-growth phenotype of the M01C06 mutant was unchanged relative to wild-type strains. The development rates and colony diameters were similar on media containing several different carbon sources, including sucrose, pectin, cellulose, and starch (Fig. 1d). Together, these results revealed that the M01C06 mutant showed a significantly reduced virulence on cotton but did not present alterations in development, implying that the gene associated with the integration of T-DNA in the genome was specifically involved in the pathogenicity of V. dahliae.
Analysis the T-DNA tagged genes in the mutant. To identify the targeted gene that was involved in the pathogenicity of M01C06 mutant, the sequences flanking the T-DNA insertion were isolated from the mutant using hiTAIL-PCR as previously described 35 . Fragments of approximately 200-1000 bp were amplified from the genomic DNA of the mutant using 5 degenerate primers (see Supplementary Fig. S2). Among the amplified fragments, most were impaired PCR amplicons, with the exception of a 795-bp amplicon that completely matched the T-DNA and its flanking sequence; the sequence indicated that the T-DNA integrated into an intergenic region at nucleotide 793,894 of supercontig 1.11 (chromosome 4) between the VDAG_05889 and VDAG_05890 genes, according to the reference genome for VdLs.17 36 . The physical distances of VdIF5 and VdCYP1 from the T-DNA insertion were 1,471 bp and 286 bp, respectively (Fig. 2a). VDAG_05889 (termed VdIF5) encodes a protein that is homologous to eukaryotic translation initiation factor 5 ( Fig. 2b), which plays a central role in protein synthesis by organizing the formation of the pre-initiation complex 37 . VDAG_05890 (termed VdCYP1) is predicted to encode a cytochrome P450 monooxygenase belonging to the P450s superfamily (Fig. 2c), the proteins of which are involved in many essential cellular processes and play diverse functional in fungi 28 . A transcript-level comparison of VdIF5 and VdCYP1 in the mutant showed that the relative expression of VdCYP1 but not VdIF5 was markedly reduced relative to that in the wild-type strain (to only 38% of the level in the wild-type strain; Fig. 2d), indicating that the integration of the T-DNA into the intergenic region likely affects the function of VdCYP1 in the M01C06 mutant.
Identification of the cytochrome P450 monooxygenase-encoding VdCYP1 gene. Based on information predicting that the VDAG_05890 gene contains 4 introns in VdLs.17 genome 36 , the VdCYP1  transcript was obtained using RT-PCR (see Supplementary Fig. S3a). The sequence of the cDNA verified that the full-length open reading frame was 1,458 bp in length and encoded 485 amino acids (Fig. 3a). After comparison, the genomic sequence of VdCYP1 was confirmed to contain four introns (Fig. 3a). Therefore, the VdCYP1 gene had the identical sequence and gene structure as the annotation in VdLs.17 genome 36 . Moreover, a 506-bp 3′ untranslated region (3′ UTR) was cloned using 3′ rapid amplification of cDNA ends; this region included the integration site of the T-DNA (nucleotide 793,894 of supercontig 1.11 in VdLs.17) (see Supplementary Fig. S3b,c). These results confirmed that the T-DNA was integrated into the 3′ UTR of VdCYP1, providing an explanation for the suppressed expression of the VdCYP1 gene that ultimately affected the pathogenicity of the M01C06 mutant.
Bioinformatics analysis of the VdCYP1 protein confirmed that VdCYP1 contains a P450 domain (IPR002401). Further comparison to the Fungal Cytochrome P450 Database 23 revealed that VdCYP1 had a greater similarity to members of group I of the E-class P450 proteins (CYP family 548), which includes numerous orthologous genes in fungi. Although there is high variation in the amino acid sequences of VdCYP1 and its orthologs, group I of the E-class P450 family contains six conserved signature motifs (E-I), four of which are conserved in VdCYP1 (see Supplementary Fig. S4). In particular, P450s contain the highly conserved cysteine heme-iron ligand signature at the C-terminal end of the protein 24 . Correspondingly, VdCYP1 encodes the signature sequence of ten amino acids (FAFANGGRGC) in the E-I-5 motif, including a conserved cysteine residue (see Supplementary Fig. S4).
Several P450s have been implicated in mechanisms of plant pathogenicity 28 . To determine whether VdCYP1 is related to V. dahliae pathogenicity, the expression level of VdCYP1 was detected during the cotton infection process using quantitative real-time PCR. As expected, the expression of VdCYP1 was continuously up-regulated within five days after infection of cotton; the transcript level was significantly elevated by over 20-fold one day after inoculation (Fig. 3b). These results suggest that VdCYP1 may be involved in V. dahliae pathogenicity and that it plays a curial role in the extreme environment of the cotton cell wall during the early infection process.
VdCYP1 is crucial for the virulence of V. dahliae. To determine whether VdCYP1 was the disrupted gene that caused the loss of pathogenicity in M01C06 mutant, knockout transformants of VdCYP1 and VdIF5 were generated in the Vd991 wild-type strain by using homologous recombination with ATMT (see Supplementary  Fig. S5). As expected, the VdCYP1 gene was necessary for the virulence of the wild-type strain. The pathogenicity of three random knockout strains (Δ VdCYP1-1-3) was significantly reduced on cotton, and these strains caused no aberrant stunting, chlorosis or wilting symptoms in the cotton plants, similar to the phenotype observed after inoculation with the T-DNA M01C06 mutant (Fig. 4a). Moreover, deletion of VdCYP1 attenuated vascular discoloration in cotton in a way that was observably different from the wild-type strain (Fig. 4b). The degree of cotton leaf wilting that was observed with the knockout strains was also markedly reduced compared to that of the wild-type strain, and the knockout strains showed DI values of only 16.22 ± 4.08 after 17 days inoculation, similar to the value of the T-DNA mutant M01C06 (Fig. 4c). In addition, the phenotypes of the VdCYP1 deletion strains and their radial growth on media with different carbon sources were similar to those of the wild-type strain during the development process over 11 days, with the exception of a minor difference in hyphal compaction ( Fig. 4d; see Supplementary Fig. S6). As expected, the homologous recombination of VdIF5 was unsuccessful after ATMT in five experiments with six replicates (see Supplementary Fig. S5c), which probably due to the function of VdIF5 should be necessary for basic biological processes involving the eukaryotic translation initiation factor, such as protein translation. These results, together with our observation in mutant M01C06, show that the integration of T-DNA in the intergenic area significantly affected the function of VdCYP1 but not VdIF5 and caused a loss of pathogenicity on cotton, strongly indicate that gene VdCYP1 is crucial for the virulence of V. dahliae.

Re-introduction of VdCYP1 into the T-DNA mutant and knockout strains restores the pathogenicity on cotton.
To ensure that the pathogenicity loss in the VdCYP1 deletion strain was caused by the T-DNA insertion or the gene replacement event, the full-length of gene VdCYP1 was re-introduced into the T-DNA mutant and knockout strains using ATMT with geneticin selection. The transformants in which whether the VdCYP1 gene has been integrated were verified by PCR. Three complementation transformants, C M 1, C M 2, and C M 3, in which the VdCYP1 gene was integrated in the T-DNA M01C06 mutant background, and three other transformants, C K 1, C K 2, and C K 3, in which the VdCYP1 gene was integrated in the gene knockout strains, were selected for phenotypic analysis (see Supplementary Fig. S7). As expected, the virulence of above six transformants were restored on cotton after re-introducing the VdCYP1 gene into the gene deletion and T-DNA strains, showing the virulence comparable to that of the wild-type Vd991 strain (Fig. 5a). Above results confirm that VdCYP1 acts as a pathogenic gene in V. dahliae. An assay of the degree of Verticillium wilt symptoms showed that the DI values of all the transformants were between 57.95 ± 1.67 and 65.56 ± 7.52, which were close to that of the wild-type strain (approximately 68.0). These results reconfirmed that VdCYP1 acts as a pathogenic factor and is crucial for the virulence of V. dahliae. VdCYP1 involved in the secondary metabolism in V. dahliae. Previous research on fungal genomes revealed that the genome encodes a large number of P450 genes, which are typically involved in fungal pathogenic processes by participating in the biosynthesis of fungal secondary metabolites 23,27,28 . To determine whether VdCYP1 is involved in V. dahliae secondary metabolism, extracellular metabolites were extracted from the culture supernatants of two independent knockout mutant strains (Δ VdCYP1-1 and Δ VdCYP1-2) and the Vd991 wild-type strain. The metabolites were detected using an ultra-performance liquid chromatography-quadrupole time-of-flight mass spectrometry (UPLC-Q-TOF-MS) method and many compounds showed significant difference between wild-type strain and VdCYP1 deletion mutants (see Supplementary Table S1). Unexpectedly, the qualitative analysis found that the number of the different compounds presented significant difference in the two independent knockout mutants compared with the wild-type strain, identified that 206 and 699 compounds in the Δ VdCYP1-1 and Δ VdCYP1-2 supernatants, respectively, were less than in the wild type (see Supplementary  Table S1). These results suggest that the secondary metabolism was extremely sensitive to the independent recombination events of VdCYP1 in V. dahliae. Further combination analysis found that the metabolism of 14 compounds (Sulfacetamide, 1-Methoxy-1H-indole-3-acetonitrile, Sinalexin, etc.) was simultaneously affected both in the Δ VdCYP1-1 and Δ VdCYP1-2 mutants compared to the wild-type strain (Table 1). This result provided the solid evidence for the opinion that VdCYP1 could be involved in the secondary metabolism of V. dahliae.
According to the KEGG annotation of the reference genome of VdLs.17 36 , at least 148 genes are involved in aminobenzoate degradation pathway, and among them 41 genes encode proteins containing the P450 domain (see Supplementary Table S2). In particular, the pathway annotation of VdCYP1 was matched to the reaction of parathion to paraoxon by oxygenation (see Supplementary Fig. S8), indicating that VdCYP1 may be involved in the aminobenzoate degradation pathway. However, we did not find the variation of parathion or paraoxon in Δ VdCYP1-1 and Δ VdCYP1-2 strains, which indicate that parathion or paraoxon was not the final products of VdCYP1 in the aminobenzoate degradation pathway. Interestingly, the result of UPLC-Q-TOF-MS showed that the amount of one typical compound, sulfacetamide (SFA), was significantly higher in the wild-type strain than in the two independent knockout mutants (Table 1). Coincidentally, the structure of sulfacetamide is similar to sulfanilic acid, which is one metabolite of the aminobenzoate degradation pathway. Therefore, we infer that gene VdCYP1 probably affect the metabolism of the sulfacetamide by involving aminobenzoate degradation pathway. In addition, previous reports indicated that sulfacetamide derivatives are toxic compounds that can inhibit the biosynthetic pathway of folate, which is an essential molecule that is required by numerous organisms, especially plants [38][39][40] . Correspondingly, the in vitro assay showed that the cotton and tobacco leaves were sensitive to sulfacetamide, and the cotton and tobacco leaves all showed the obvious necrosis and wilting symptoms after infiltration with the sulfacetamide solution (Fig. 6b,c). Moreover, in vivo detection also showed that sulfacetamide could  cause significant wilting symptom in cotton leaves that was similar to the typical disease symptoms (darkening and wilting of the leaves) of Verticillium wilt (Fig. 6d). Therefore, above results indicate that VdCYP1 play important function in the pathogenesis of V. dahliae by involving in the secondary metabolism, such as the aminobenzoate degradation pathway, to contribute to virulence during infection.

Discussion
Recently, several studies of V. dahliae have revealed that this pathogen could employ various and complex mechanisms for penetration, colonization, host-pathogen interaction, and expansion in plants; and the complex mechanisms include the utilization of carbon/nitrogen metabolism to obtain nutrition, overcoming the osmotic stress and reactive oxidative species produced by the host plant in response to infection, and modulating host physiology or biochemistry with effectors 41 . Correspondingly, numerous genes have been identified as pathogenicity-related factors in V. dahliae and facilitate the pathogenic progression, including the cell wall-degrading enzymes 9,18 , various effectors 7,10,11,42 , protein kinases 13 , and transcriptional regulators 11,12 . With the completion of the genome sequence of V. dahliae 36,42 , ATMT technology has been more widely used for the study of V. dahliae genes that are associated with diverse phenotypes [18][19][20] , especially the genes associated with pathogenicity. At present, several genes associated with pathogenicity have been identified from the mutants with reduced or lost pathogenicity 18,19,21,22 , confirming that ATMT is an effective method to identify the genes associated with V. dahliae pathogenicity. In China, Verticillium wilt caused by V. dahliae is the most destructive disease of cotton, as more than 50% of the cotton acreage has been affected by Verticillium wilt in recent years, significantly reducing the fiber quality and resulting in yield losses (National Cotton Council of America Disease Database). Based on the premise that the resistant genes are lacking in cotton, the investigations of pathogenicity-related genes and revealing the pathogenic mechanism of V. dahliae could be beneficial for the development of Verticillium wilt control strategies. In our previous research, the T-DNA random insertional mutagenesis of Vd991 (the typical highly virulent strain of the cotton host origin) was obtained using ATMT, which prompted us to investigate the pathogenicity of V. dahliae 20 . In present research, at least 16 mutants with significant pathogenicity reduce or loss were screened (see Supplementary Fig. S1). And the locations of the T-DNA insertion sites have already revealed several genes associated with pathogenicity phenotypes, such as VdCYP1 (encoding the cytochrome P450 monooxygenase), VdPT1 (encoding the palmitoyltransferase), VdVSP1 (encoding the vacuolar protein sorting-associated protein), and VdOR1 (encoding the oxidoreductase) (unpublished data). These results confirmed that the ATMT methods can be effectively used in the study of pathogenicity-related genes in V. dahliae.
In present research, one T-DNA mutant that lost the pathogenicity on cotton was selected for pathogenesis analysis (Fig. 1). The identification of the flanking sequence indicate that the T-DNA was integrated into the intergenic region between gene VdCYP1 and VdIF5 (Fig. 2); and the further targeted gene deletion and complement confirmed that VdCYP1 was the pathogenic related gene (Figs 4 and 5). The protein encoded by VdCYP1 is highly similar to P450s of other fungi, and it contains the conserved region with the sequence FXXGXXXCXG (see Supplementary Fig. S4), which is the heme-binding domain that contains the axial Cys ligand for the heme 24 . P450s as a superfamily of heme-containing monooxygenases, distribute widely in fungi, and play diverse and pivotal roles in metabolic versatility and fungal adaptation to specific ecological niches 23,25 . Associations between P450s and pathogenicity have been identified in many fungi 30,33,43,44 , and several studies showed that mycotoxin synthesis in several fungi requires the participation of P450s in the pathogenic process 45 . Moreover, comparative genomics revealed that P450s family in several plant pathogens has been significantly expanded to promote the pathogenicity and adaptation 23,25 . Numerous P450s have been found to be significantly up-regulated during host-pathogen interactions 29,30 ; such regulation also occurred in V. dahliae, as VdCYP1 was observably up-regulated during cotton infection (Fig. 3). Together, our research strongly indicates that VdCYP1, which encodes a P450, is a pathogenicity-related gene to involve in the pathogenicity in V. dahliae.
Secondary metabolites can be the pathogenic or virulent factors and cause significant disease in agricultural crops 46 , as demonstrated by several studies of Fusarium spp 34 . For example, the trichothecene mycotoxin that are produced by certain Fusarium species can promote the virulence towards wheat and maize 47,48 , by inhibiting protein expression and elicitor-inducing activities that could stimulate the plant defenses and promote plant cell death 49 . The increasing evidences suggest that P450s, as one huge and incompletely discovered family, could play important and various roles in the pathogenic processes of fungi infecting host plants by participating in the secondary metabolites synthesis 44 . Recent fungal transcriptome data analysis found that the expression of many fungi P450 genes have been obviously induced in infection process, which maybe influence the fungal secondary metabolites synthesis and participate in the pathogenic process. For example, one putative benzoate 4-monooxygenase cytochrome P450 gene of Fusarium graminearum was obviously induced during wheat coleoptiles infection, and was speculate to involve in the pathogenic process by participating in the secondary metabolism 29 . The idea that P450s could involve in secondary metabolites synthesis and influence fungi pathogenicity was also confirmed in our investigation, as the qualitative analysis of some secondary metabolites showed observably decrease in Δ VdCYP1 mutants compared to the wild-type strain (see Supplementary Table S1). In particular, 14 compounds were thought to be affected by the VdCYP1 deletion in the two independent Δ VdCYP1 mutants (Table 1). But until now, there were no other report about the regulatory relationship between P450 and these compounds, and the possible pathogenic function of these compounds were also unclear.
The KEGG annotation found that VdCPY1 might be involved in the aminobenzoate degradation pathway (see Supplementary Fig. S8 and Table S2). Coincidentally, one of the 14 kinds of compounds, sulfacetamide, which is the analogous compound of sulfanilic acid that involves in the aminobenzoate degradation pathway, could cause the necrosis and wilting symptom on cotton showed by the in vivo and vitro assay (Fig. 6). These result indicate that VdCPY1 may be involved in the aminobenzoate degradation pathway of V. dahliae and indirectly influence the synthesis of sulfacetamide, which could act as one virulent factor during infecting cotton. This study is the first report of the characterization and pathogenic function analysis of one P450 gene VdCYP1 in V. dahliae, and this gene could affect the pathogenicity of V. dahliae by influencing the synthesis of secondary metabolites synthesis, such as the metabolism of sulfacetamide, which probably one of the typical case in its' complex secondary metabolic regulatory functions.
Above results suggest that compound sulfacetamide probably played as one virulent factor in V. dahliae pathogenic process. Still, the pathogenic mechanism of the compound sulfacetamide in cotton was also unclear. Previous reports indicate that sulfacetamide can impair folate synthesis in Arabidopsis thaliana 50 . And folate as essential water-soluble B-vitamins is required by numerous organisms, especially on the plants 51 . During folate synthesis, dihydropteroate synthase (DHPS) acts as one of the crucial enzymes that catalyzes the condensation of dihydropteridine diphosphate and para-aminobenzoate to yield dihydropteroate 52,53 . Sulfonamide is an analogue of para-aminobenzoate and can act as a competitive inhibitor of DHPS 53 that can inhibit DHPS activity, and finally disturb the folate synthesis in Arabidopsis 51 . Therefore, VdCYP1 may involve in the secondary metabolism by the aminobenzoate degradation pathway to influence the compounds sulfacetamide that could impair folate synthesis and ultimately induce the wilting symptom in host plants.
In conclusion, P450 VdCYP1 is one necessary pathogenic factor during V. dahliae infecting cotton. VdCYP1 could involve in the synthesis of secondary metabolites in V. dahliae, such as the pathogenic compound sulfacetamide synthesis. The characterization of VdCYP1 expression and pathogenic function during V. dahliae infecting cotton in this study would provide insight into the process of V. dahliae infecting host plant, promote the identification of P450s' roles in V. dahliae secondary metabolism and accelerate the understanding of other functions of P450s in fungal pathogenic process.

Materials and Methods
Identification of mutants with a loss of pathogenicity. The T-DNA library, consisting of 1,344 mutants, was cultured on potato dextrose agar (PDA) medium at 25 °C for 10 days before the conidia were washed with sterile water and the concentration was adjusted to 5 × 10 6 conidia/mL. The susceptible cotton (G. hisutum cv. Junmian 1) was planted in sterilized soil at 28 °C with a 16/8 light-dark photoperiod for 2 weeks. In the preliminary screening, 12 cotton seedlings were immersed into 30 mL of prepared conidial suspension for 2 minutes and then transferred into a new nutrition pot before being cultured in the greenhouse at 28 °C with a 16/8 light-dark photoperiod for 3 weeks. The pathogenic loss mutants were further selected using a pathogenicity assay with an expanded population of 30 cotton seedlings with 3 replicates. Symptom development was measured from the first day to the sixth week and categorized according to the disease index (DI), which uses four grades: grade 1: the cotyledon beginning to yellow; grade 2: the cotyledon and one euphylla wilting; grade 3: all leaves wilting and chlorosis; and grade 4: plant death. The formula of DI value = [∑ (the seedling of every grade × relative grade)/ (total seedlings × 4)] × 100 54 . The symptoms of Verticillium wilt were photographed after three weeks inoculation. The pathogenicity loss mutant M01C06 was further used for functional analysis.
T-DNA tagged gene analysis in the mutant. The M01C06 mutant was cultured on liquid complete medium (CM) at 25 °C for 5 days, and the mycelium was then harvested for genomic DNA isolation using the DNAsecure Plant Kit (TianGen Biotech., CO., Ltd., Beijing, China). The flanking sequences of T-DNA were identified using hiTAIL-PCR methods with degenerate primers, as previously described 35 . All amplicons were cloned, sequenced, and then aligned to the reference genome of VdLs.17 using BLASTN (http://blast.ncbi.nlm.nih.gov/ Blast.cgi) to ascertain the physical location of the T-DNA. According to the alignment results, the 2 sequences flanking the T-DNA were reconfirmed using PCR with specific primers. Quantitative reverse transcription-PCR (qRT-PCR) was performed to identify the expression levels of genes that were affected by the T-DNA insertion using the FastFire qPCR premix (SYBR Green, TianGen, Beijing, China), and the relative gene expression level was calculated using the 2 −△△CT method 55 with the β -tubulin gene (VDAG_10074, VdLs.17) as an internal control. The primers are listed in Supplementary Table S3. Real-time PCR conditions consisted of an initial 94 °C denaturation step for 10 min that was followed by 40 cycles of 94 °C for 15 s and 60 °C for 1 min.
Gene cloning and bioinformatic analysis. Total RNA was extracted using the Total RNA Miniprep Kit (Axygen, MA, USA), and first-strand cDNA was synthesized using the FastQuant cDNA Reverse Transcriptase Kit (TianGen, Beijing, China). The flanking genes were cloned using RT-PCR with cDNA and PCR with genomic DNA, and then the protein-coding genes were identified using Clustal X software with the transcript and genomic sequence information (http://www.clustal.org/clustal2/). The rapid amplification of cDNA ends (RACE) system (Invitrogen, Carlsbad, California, USA) was used for 3′ end amplification using the standard protocol. Conserved domain and functional homology searches were performed using the SMART programs (http://smart. embl-heidelberg.de/smart). All primers are listed in Supplementary Table S3.
Fungal transformation. The homologous recombination method was used to generate deletion mutants of the two flanking genes, VdIF5 and VdCYP1, as previously described 9 . The flanking sequences were amplified from the genomic DNA, integrated into a hygromycin cassette using fusion PCR, and then assembled into the pGKO2-gateway vector 56 in a BP recombinant reaction (Invitrogen). To generate the complementation vector, 3 fragments of the trpC promoter, open reading frame of VdCYP1 and Nos terminator were fused using fusion PCR to obtain a complementation fragment with KpnI/XbaI enzyme sites. The fragment was then ligated into the pCT-HN vector, which possesses hygromycin and geneticin resistance selection markers 9 .
The vectors were transferred into the Agrobacterium tumefaciens AGL-1 strain for fungal transformation. A. tumefaciens-mediated transformation of V. dahliae was conducted for the homologous recombination and complementation transformants as described previously 57 , and transformants were selected on PDA medium Scientific RepoRts | 6:27979 | DOI: 10.1038/srep27979 supplemented with hygromycin or geneticin at 50 μ g/mL. Homologous recombination was verified by PCR. The primer sequences are shown in Supplementary Table S3.

Gene expression analysis.
To determine the expression level of VdCYP1 during cotton infection, two-week-old cotton seedlings were root-inoculated with 5 × 10 6 conidia/mL of a V. dahliae suspension; the roots were then harvested at 1, 2, 3, 4, and 5 days post-inoculation and flash frozen in liquid nitrogen for RNA extraction. V. dahliae cultured on PDA medium without being used to inoculate cotton plants was set as the control. After grinding, 100 mg of ground material was used for total RNA extraction and cDNA synthesis. The transcript level of VdCYP1 was determined using qPCR, as described above.

Phenotypic characterization.
To determine the colony morphology on different carbon sources, conidial suspensions of V. dahliae strains were prepared at 2 × 10 6 conidia/mL and placed in the center of basic Czapek-Dox medium agar plates that included either sucrose (30.0 g/L), pectin (10.0 g/L), cellulose (10.0 g/L) or starch (15.0 g/L). The growth phenotypes were investigated after 10 days of incubation at 25 °C, and the colony diameter was measured. Pathogenicity assays of the deletion strains and the complemented transformants were conducted with 30 cotton seedlings and in triplicate. The severity of Verticillium wilt was calculated using the DI method, as used to quantify the pathogenicity of T-DNA mutants. Phenotypic investigations of Verticillium wilt in whole plants and in vertical sections of the cotton seedling shoots were also conducted at three weeks post-inoculation.
Secondary metabolite determination. The secondary metabolites in the culture supernatants of the VdCYP1 deletion strains were identified using the UPLC-Q-TOF method. Two independent VdCYP1 deletion strains were cultured in basic Czapek-Dox medium without antibiotics at 25 °C in a shaking incubator at 150 rpm for five days. The cultured solution was immediately centrifuged at 15,000 g at 4 °C for 20 min to collect the supernatants. The supernatants were filtered through a cellulose acetate filter (0.45 μ m) and then cleaned up by liquid-liquid extraction (LLE) using 250 mL of chloroform:acetonitrile (1:1 v/v) twice. Next, the lower layer (organic solvent) was collected in a round-bottom flask and concentrated to near-dryness using a rotary evaporator at 35 °C. The residue was reconstituted in 2.0 mL of acetonitrile:water (50:50 v/v, containing 0.1% formic acid) and was then briefly centrifuged. Subsequently, 1.0 mL of the upper layer (acetonitrile) was filtered through a 0.22-μ m syringe filter for UPLC-Q-TOF analysis. The wild-type strain was used as the control.
Samples were analyzed on an UPLC-Q-TOF mass spectrometer (Waters, Manchester, U.K.). Metabolites were separated using a Waters BEH Shield RP C18 column (50 mm × 2.1 mm, 1.7 μ m) at 30 °C. The mobile phases were composed of phase A (0.1% formic acid in water) and phase B (0.1% formic acid in acetonitrile). The gradient conditions were optimized as follows: 0-1 min, 2-5% B; 1-8 min, 5-90% B; 8-10 min, 90-100% B; 10-12 min, 100-2% B. The flow rate was 0.3 mL/min. The MS system was operated in ESI+ mode. The mass range was set at m/z 100-1000 Da in the full-scan mode. The optimized ESI parameters were as follows: capillary voltage, 3.0 kV; cone voltage, 35 V; source temperature, 100 °C; desolvation temperature, 350 °C; desolvation gas flow, 600 L/h. For accurate mass measurement, leucine enkephalin was used as the lock spray standard at a concentration of 100 ng/mL with a flow rate of 50 μ L/min. Data were processed using MarkerLynx software (Waters). Peak detection was performed across the mass range of m/z 100-1000 with a retention time (RT) between 2 and 12 min, peak widths automatically detected, an intensity threshold of 100, a mass window of 0.02, a RT window of 0.1, and a noise elimination value of 6.0. Multivariate statistical analysis was performed using EZinfo software (Waters) for unsupervised principal component analysis (PCA) to obtain a general overview of the variance of metabolic phenotypes. Next, orthogonal projection to latent structures-discriminant analysis (OPLS-DA) was used to maximize the differences in the metabolic profiles between the wild-type and mutant strains. The metabolite features were identified based on retention behavior, mass assignment, and online database query. Metabolic pathway analysis was performed in the HMDB database (http://www.hmdb.ca/) to investigate the disturbed metabolic pathways and to facilitate biological interpretation.
Virulence assays of sulfacetamide on cotton. The cotton cotyledon was used for evaluating the virulence of SFA by infiltration. The SFA was dissolved in 2% dimethyl sulfoxide (DMSO) at concentration of 25 mM. The solutions were infiltrated into the cotyledons of 10-day-old cotton seedlings, and the wilting phenotype was investigated after 24 hours; the solvent was used as the control. Virulence assays were also performed on cotton seedlings using the root-dipping method. The bottom of the nutrition pot was wiped out to expose the root of a two-week-old seedling, and it was transferred into a new pot containing only 20.0 mL of at 25 mM SFA solution; the wilting symptoms of the cotton seedling were investigated after 36 hours post inoculation.