The velvet complex governs mycotoxin production and virulence of Fusarium oxysporum on plant and mammalian hosts

Authors

  • Manuel S. López-Berges,

    1. Departamento de Genética, Facultad de Ciencias, Universidad de Córdoba, Córdoba, Spain
    2. Campus de Excelencia Internacional Agroalimentario ceiA3, Córdoba, Spain
    Current affiliation:
    1. Department of Molecular and Cellular Medicine, Centro de Investigaciones Biológicas CSIC, Madrid, Spain
    Search for more papers by this author
  • Concepción Hera,

    1. Departamento de Genética, Facultad de Ciencias, Universidad de Córdoba, Córdoba, Spain
    2. Campus de Excelencia Internacional Agroalimentario ceiA3, Córdoba, Spain
    Search for more papers by this author
  • Michael Sulyok,

    1. Christian Doppler Laboratory for Mycotoxin Research, University of Natural Resources and Applied Life Sciences Vienna, Tulln, Austria
    Search for more papers by this author
  • Katja Schäfer,

    1. Departamento de Genética, Facultad de Ciencias, Universidad de Córdoba, Córdoba, Spain
    2. Campus de Excelencia Internacional Agroalimentario ceiA3, Córdoba, Spain
    Search for more papers by this author
  • Javier Capilla,

    1. Mycology Unit, Medical School, Universitat Rovira i Virgili, IISPV, Reus, Spain
    Search for more papers by this author
  • Josep Guarro,

    1. Mycology Unit, Medical School, Universitat Rovira i Virgili, IISPV, Reus, Spain
    Search for more papers by this author
  • Antonio Di Pietro

    Corresponding author
    1. Campus de Excelencia Internacional Agroalimentario ceiA3, Córdoba, Spain
    • Departamento de Genética, Facultad de Ciencias, Universidad de Córdoba, Córdoba, Spain
    Search for more papers by this author

For correspondence. E-mail ge2dipia@uco.es; Tel. (+34) 957 218981; Fax (+34) 957 212072.

Summary

Fungal pathogens provoke devastating losses in agricultural production, contaminate food with mycotoxins and give rise to life-threatening infections in humans. The soil-borne ascomycete Fusarium oxysporum attacks over 100 different crops and can cause systemic fusariosis in immunocompromised individuals. Here we functionally characterized VeA, VelB, VelC and LaeA, four components of the velvet protein complex which regulates fungal development and secondary metabolism. Deletion of veA, velB and to a minor extent velC caused a derepression of conidiation as well as alterations in the shape and size of microconidia. VeA and LaeA were required for full virulence of F. oxysporum on tomato plants and on immunodepressed mice. A critical contribution of velvet consists in promoting chromatin accessibility and expression of the biosynthetic gene cluster for beauvericin, a depsipeptide mycotoxin that functions as a virulence determinant. These results reveal a conserved role of the velvet complex during fungal infection on plants and mammals.

Introduction

Fungi are a highly versatile group of organisms comprising mostly saprophytes that thrive on dead organic material. A number of species have evolved a pathogenic lifestyle, enabling them to cause disease on living organisms. Fungal plant pathogens produce more than 70% of all major crop diseases and destroy 15% of global agricultural production through yield losses and mycotoxin contamination, making them by far the most harmful class of plant pathogens (Strange and Scott, 2005; Fisher et al., 2012). In clinical settings, filamentous fungi (moulds) provoke life-threatening systemic infections, particularly on immunocompromised patients (Fridkin, 2005).

To sustain the infectious lifestyle, fungal pathogens have evolved virulence mechanisms that allow them to enter the host and overcome its innate defences. How this multifaceted infection programme is orchestrated by the fungal cell remains largely unknown (Sexton and Howlett, 2006). A second unsolved question concerns the genetic bases of host range. While certain fungi cause disease only on a single host species, others infect a wide range of hosts from evolutionary distant groups or even different kingdoms, such as plants and animals. The soil-borne ascomycete Fusarium oxysporum produces vascular wilt disease in over 100 different plant species, and was ranked fifth in a recent survey of the Top 10 fungal plant pathogens, based on scientific and economic importance (Dean et al., 2012). F. oxysporum also causes infections in humans that range from superficial dermato- or keratomycosis to fatal disseminated fusariosis, and its clinical impact has increased over the last decades (Nucci and Anaissie, 2007). Fusarium spp. now represent the second most frequent mould causing invasive fungal infections after Aspergillus, with F. oxysporum ranking second after Fusarium solani in the incidence of invasive fusariosis (Guarro and Gene, 1995; Nucci and Anaissie, 2007). Both human and plant pathogenic isolates of F. oxysporum have polyphyletic origins and respond poorly to available antifungal agents (O'Donnell et al., 1998; 2004; Azor et al., 2009).

Previous work established that a single isolate of F. oxysporum f. sp. lycopersici causes disease both on tomato plants and on immunodepressed mice (Ortoneda et al., 2004). The genome sequence of this isolate is available (Ma et al., 2010), making it an ideal model for studying the genetic and metabolic repertoire of a trans-kingdom pathogen (Ortoneda et al., 2004). In a forward genetic screen we identified a mutant with attenuated virulence on tomato plants that carries a transposon insertion in an orthologue of the Aspergillus nidulans VelB (velvet-like B) gene (Lopez-Berges et al., 2009). VelB belongs to the velvet protein family, whose founding member is A. nidulans velvet A (VeA) (Kaefer, 1965). Additional members of the velvet protein family include VelB, VelC and VosA. In the absence of light, VeA and VelB interact and enter the nucleus (Bayram et al., 2008a), where they assemble with the non-velvet protein LaeA, a global regulator of secondary metabolism (Bok and Keller, 2004; Bok et al., 2006). The heterotrimeric velvet complex co-ordinates fungal development and biosynthesis of secondary metabolites by modulating chromatin accessibility and gene expression (Bayram et al., 2008a; Reyes-Dominguez et al., 2010).

In this study we genetically characterized the members of the velvet protein complex in F. oxysporum, and show that VeA, VelB and LaeA govern hyphal growth and conidiation, as well as virulence on tomato plants and on immunodepressed mice. We provide evidence that velvet promotes chromatin accessibility and transcription of gene clusters encoding biosynthesis of the siderophore ferricrocin as well as the mycotoxin beauvericin, a virulence factor of F. oxysporum.

Results

Velvet proteins and LaeA control hyphal growth and development of F. oxysporum

The previous identification of the VelB orthologue FOXG_00016 as a virulence factor on tomato plants (Lopez-Berges et al., 2009) led us to characterize the entire velvet protein family from F. oxysporum. A search in the genome database identified two additional velvet-like genes, FOXG_11273 and FOXG_02050, which encode predicted orthologues of A. nidulans VeA and VelC, respectively, but no VosA orthologue (Fig. 1). We also identified FOXG_00975 encoding a putative orthologue of A. nidulans LaeA which contains a conserved protein methyltransferase SAM-binding domain previously reported in A. nidulans LaeA (Fig. S1). Yeast two-hybrid experiments with cDNA clones of the putative velvet complex genes from F. oxysporum detected the following protein–protein interactions: VeA–VelB, VeA–VelC, VeA–LaeA, VelB–VelB and VelB–VelC (Fig. S2).

Figure 1.

The velvet protein family in Fusarium oxysporum. Cladogram of fungal velvet-like proteins. Bootstrap values obtained from 1000 replicates are indicated at the nodes. clustalw was used for protein alignment. Necha, Nectria haematococca (Fusarium solani); FOXG, Fusarium oxysporum; FVEG, Fusarium verticillioides; FGSG, Fusarium graminearum; MGG, Magnaporthe oryzae; UM, Ustilago maydis; AN, Aspergillus nidulans. Velvet proteins from F. oxysporum are boxed, while those from A. nidulans are shaded.

Null mutants in veA, velC and laeA [in addition to the previously generated ΔvelB mutant (Lopez-Berges et al., 2009)], as well as a ΔveAΔvelB double mutant, were obtained by homologous gene replacement (Figs S3–S6). F. oxysporum ΔveA mutants displayed a characteristic flat colony phenotype with dramatically reduced aerial mycelium, leading to a significant decrease in surface hydrophobicity (Fig. S7). The colony phenotype of the ΔvelB mutant was similar, albeit less severe than that of ΔveA. The ΔveAΔvelB double mutant recapitulated the ΔveA phenotype, whereas the ΔvelC mutant closely resembled the wild-type strain.

Light significantly affected colony morphology and pigmentation of the wild-type and ΔvelC strains on solid medium (Fig. 2A). The ΔveA, ΔvelB and ΔveAΔvelB mutants showed an attenuated light response and a strong increase in hyphal branching and differentiation of phialides during submerged growth in the dark, resulting in a dramatic (up to 30-fold) increase in microconidia production relative to the wild type (Fig. 2B and C). Two independent ΔvelC mutants also showed a modest, but significant surge in conidiation. Increased conidiation in the velvet mutants was reversed by high concentrations of the osmoticant sorbitol. Microconidia of the ΔveA mutant differed in shape from those of the wild-type strain, being more elongated (Fig. 2D). The elongated conidial phenotype was exacerbated in the ΔvelB mutant, while the ΔveAΔvelB double mutation recapitulated the ΔveA phenotype. Strikingly, the ΔvelC mutants produced significantly smaller microconidia than the wild type.

Figure 2.

VeA, VelB and LaeA have overlapping and distinct roles in hyphal development and conidiation.

A. Colony phenotype of the indicated strains grown on PDA plates for 6 days at 28°C under conditions of continuous dark or light. wt, wild type. Bar = 1 cm.

B. Hyphal morphology of strains grown in static liquid cultures on YPG with or without 1 M sorbitol in continuous dark for 24 h at 28°C. Images were taken in a Leica DMR microscope using the Nomarsky technique. Bar = 20 μm.

C. Numbers of microconidia produced by the indicated strains grown in static liquid cultures on YPG with or without 1 M sorbitol in continuous dark for 48 h at 28°C, represented relative to the wild-type strain grown on YPG. Bars represent standard errors from three independent experiments with three technical replicates each.

D. Morphology of microconidia produced by the indicated strains grown in liquid cultures on PDB for 3 days at 28°C. Images were taken using the Nomarsky technique. Bar = 10 μm.

The ΔlaeA mutant shared several phenotypes with ΔveA, including the flat colony morphology, lack of aerial mycelium or decrease in colony hydrophobicity, but also displayed differential phenotypes such as an increase rather than a decrease in hyphal growth rate or a stronger pigmentation in response to light (Fig. 2). Moreover, the ΔlaeA strains produced less microconidia in submerged culture than the wild type, in contrast to the velvet mutants. All the mutant phenotypes were restored in the complemented strains, confirming that they are caused by loss of the respective gene (Fig. 2). Collectively these results indicate that VeA, VelB, LaeA and, to a minor extent VelC, have both overlapping and distinct functions in hyphal development, conidiation and light response of F. oxysporum.

The velvet complex regulates chromatin structure and transcription of siderophore biosynthetic genes

In Aspergillus, LaeA (Bok and Keller, 2004; Bok et al., 2006; Perrin et al., 2007) and the velvet complex (Bayram et al., 2008a) promote transcription of gene clusters containing non-ribosomal peptide synthetase (NRPS) genes. Some of these NRPSs function in biosynthesis of secondary metabolites such as penicillin or gliotoxin. We searched the F. oxysporum genome database for putative orthologues of Aspergillus fumigatus NRPSs regulated by LaeA (Perrin et al., 2007). Our survey identified 11 NRPS genes, which are mostly located in putative gene clusters with a predicted function in biosynthesis of secondary metabolites. RT-PCR analysis of RNA obtained from mycelia grown in the dark (35 cycles) detected transcripts for three of the identified NRPS genes, FOXG_06448, FOXG_09785 and FOXG_11847 (Table S1).

FOXG_06448 and FOXG_09785 encode predicted orthologues of SidC and SidD, two NRPSs from A. nidulans that function in biosynthesis of the siderophores ferricrocin and triacetylfusarinine C respectively (Eisendle et al., 2003; Oide et al., 2006). We noted that the gene located next to F. oxysporum sidC, FOXG_06447 (Fig. 3A) encodes a predicted orthologue of the l-ornithine N(5)-monooxygenase SidA, which catalyses the initial step in both the intra- and extracellular siderophore biosynthetic pathways in A. nidulans (Eisendle et al., 2003). A survey of sidA and sidC orthologues in different ascomycetes revealed that these two genes are also clustered in other Fusaria, Neurospora crassa and Magnaporthe oryzae, but not in Aspergilli.

Figure 3.

The velvet complex and light regulate chromatin structure and transcription of genes required for biosynthesis of the siderophores ferricrocin and triacetylfusarinine C.

A. Physical map of the promoter region of the sidC gene located in the ferricrocin siderophore gene cluster. The amplified promoter fragment is indicated by a box.

B. Real-time qPCR performed on gDNA from the indicated strains germinated 14 h in potato dextrose broth (PDB) in the dark, transferred to fresh PDB and maintained for 10 h either in dark or in light. Relative chromatin accessibility was calculated as the ratio of amplification levels with gDNA obtained from untreated mycelia versus gDNA from MNase-treated mycelia, and represented relative to that of the wild-type (wt) strain grown in the dark.

C and D. Real-time RT-PCR of cDNA from the indicated strains grown as in (B). Transcript levels are represented relative to those of the wt strain grown in the dark. Bars represent standard errors from three independent experiments with three technical replicates each.

We next examined the role of the velvet complex in chromatin remodelling and transcriptional regulation of siderophore biosynthetic genes. To analyse the chromatin structure of the F. oxysporum sidC promoter, mycelia were treated with micrococcal nuclease (MNase), and the extracted genomic DNA (Fig. S8) was submitted to real-time qPCR with promoter-specific primers. Relative chromatin accessibility in the wild type, calculated as the ratio of amplification from untreated versus MNase-treated mycelia, was more than twofold higher during growth of F. oxysporum in the dark relative to the light (Fig. 3B). The increase of chromatin accessibility in the dark was abolished in the ΔveA, ΔvelB, ΔveAΔvelB and ΔlaeA mutants, and strongly reduced in ΔvelC. Consistent with these differences in chromatin structure, we detected a sharp increase of sidC and sidA transcript levels during growth in the dark in the wild-type strain, but not in the ΔveA, ΔvelB, ΔveAΔvelB and ΔlaeA mutants (Fig. 3C). An analogous expression pattern was observed for the sidD gene (Fig. 3D). We conclude that chromatin accessibility and transcription of siderophore biosynthetic genes are increased in F. oxysporum in the absence of light, and that this induction depends on the velvet complex.

Velvet regulates biosynthesis of the cyclooligomer depsipeptide beauvericin

The third of the transcribed NRPS genes identified in our initial survey, FOXG_11847 (named beas hereafter), encodes a predicted orthologue of enniatin and beauvericin synthetases from Fusarium avenaceum and Beauveria bassiana respectively (Herrmann et al., 1996a; Xu et al., 2008). Inspection of the adjacent chromosomal region identified two contiguous genes that are transcribed divergently from beas (Fig. 4A). FOXG_11846 (named kivr hereafter) encodes a putative orthologue of B. bassiana 2-ketoisovalerate reductase (KIVR), which converts 2-ketoisovalerate into d-hydroxyisovalerate, the precursor for biosynthesis of the cyclooligomer depsipeptide beauvericin (BEA) (Xu et al., 2008). FOXG_11845 (named abc3 hereafter) encodes a predicted ABC multidrug transporter, whose closest characterized orthologue is ABC3 from M. oryzae (Sun et al., 2006).

Figure 4.

The velvet complex and light affect chromatin structure and transcription of the beauvericin (BEA) gene cluster and BEA production.

A. Physical map of the promoter regions of the beas (BEA non-ribosomal peptide synthetase), kivr (ketoisovalerate reductase) and abc3 (ABC multidrug transporter) genes located in the BEA gene cluster. The promoter fragment amplified in (B) is indicated by a box.

B. Real-time qPCR performed on gDNA from the indicated strains grown as detailed in Fig. 3. Relative chromatin accessibility was calculated as described in Fig. 3.

C. Real-time qPCR of cDNA obtained from the indicated strains grown as in (B). Transcript levels are represented relative to those of the wild type grown in the dark. Bars represent standard errors from three independent experiments with three technical replicates each.

D. Quantification of BEA in cultures of the indicated strains grown for 3 days on PDA in the dark or the light, by liquid chromatography/tandem mass spectrometry. BEA levels are expressed in μg cm−2 of fungal culture. Bars represent standard errors from three independent cultures.

Relative chromatin accessibility at the beas/kivr promoter was significantly higher in wild-type mycelia grown in the dark compared with those grown in the light (Fig. 4B), resulting in higher transcript levels of the beas, kivr and abc3 genes (Fig. 4C). In contrast, the ΔveA, ΔvelB, ΔveAΔvelB and ΔlaeA mutants failed to increase chromatin accessibility in the dark, resulting in a 16- and 50-fold reduction of transcript levels of the beas and kivr genes, respectively, in the ΔveA mutant relative to the wild type (Fig. 4B and C).

To determine whether the velvet complex governs mycotoxin production in F. oxysporum, extracts from cultures of the different strains on Potato Dextrose Agar (PDA) or Czapek Dox Agar plates were analysed by liquid chromatography/electrospray ionization tandem mass spectrometry (HPLC/ESI-MS/MS). This approach allows reliable and sensitive quantification of more than 100 fungal analytes, including almost all mycotoxins for which standards are commercially available (Sulyok et al., 2007). BEA was the major mycotoxin analyte detected in F. oxysporum cultures on PDA, while fusaric acid was predominantly produced on Czapek Dox medium. Approximately 60% of total BEA in liquid culture was found in the mycelium while 40% was detected in the supernatant, indicating that part of the toxin is exported to the extracellular medium. In agreement with the gene expression data, the wild type produced significantly more BEA in the dark than in the light (Fig. 4D), while the ΔveA, ΔvelB, ΔveAΔvelB and ΔlaeA mutants showed a significant decrease of BEA levels in the dark (132-, 26-, 129- and 55-fold, respectively, compared with the wild type). Production of fusaric acid was slightly increased in the dark and dramatically reduced in the ΔlaeA mutant (Fig. S9). In the complemented strains, chromatin accessibility, gene expression and BEA production were restored to wild-type levels (Fig. 4C and D; Fig. S9).

To confirm that beas encodes the NRPS responsible for BEA biosynthesis in F. oxysporum, the gene was deleted by homologous replacement with the hygromycin resistance cassette (Fig. S10). The Δbeas mutants did not display detectable changes in colony morphology, hyphal growth or conidiation. However, HPLC/ESI-MS/MS analysis failed to detect BEA in cultures from two independent Δbeas strains (Fig. 4D). We conclude that the velvet complex regulates chromatin structure and transcription of the BEA biosynthesis gene cluster, as well as BEA production by F. oxysporum.

The velvet complex is essential for virulence on immunodepressed mice

Fusarium oxysporum f. sp. lycopersici strain 4287 can infect and kill immunodepressed mice (Ortoneda et al., 2004). Inoculation of mice with conidia of the wild-type strain or the ΔvelC mutant resulted in 70–80% killing, while animals infected with the ΔveA, ΔveAΔvelB and ΔlaeA mutants showed significantly lower mortality (P < 0.002) (Fig. 5A and C). Fungal burden in kidneys and lungs of surviving mice sacrificed 7 days after inoculation with ΔveA, ΔveAΔvelB and ΔlaeA mutants was reduced by two to three orders of magnitude with respect to those inoculated with the wild type (Fig. 5B and D). Mortality and tissue colonization were fully restored in the complemented strains. We conclude that VeA and LaeA are required for virulence of F. oxysporum on a mammalian host.

Figure 5.

VeA and LaeA are essential for production of BEA in blood and for virulence on mice.

A, C and G. Groups of 10 immunosuppressed OF-1 male mice were inoculated with 2 × 107 microconidia of the indicated fungal strains by lateral tail vein injection. Per cent survival was plotted for 15 days. Mortality rates of mice infected with the ΔveA, ΔveAΔvelB, ΔlaeA and Δbeas#10 mutants were significantly lower (P < 0.05) than those of mice infected with the wild type, ΔvelC or complemented strains. Data shown are from one representative experiment. All experiments were performed at least three times with similar results.

B, D and H. Five surviving mice infected with the indicated fungal strains were sacrificed on day 7 post infection and homogenates obtained from the indicated organs were quantitatively cultured on PDA medium. Bars represent standard errors from three technical replicates. All experiments were performed twice with similar results. Columns with the same letter within the same organ are not significantly different (Mann–Whitney, P ≤ 0.05).

E and F. Production of BEA (E) and fusaric acid (F) was quantified by liquid chromatography/tandem mass spectrometry in extracts from cultures of the indicated fungal strains grown 3 days in whole human blood at 37°C. Mycotoxin levels are expressed in ng ml−1 blood culture. Bars represent standard errors from three independent experiments.

The velvet complex is required for infection of tomato plants

Tomato plants, whose roots were inoculated with conidia of the wild-type strain or the ΔvelC mutant, showed progressive wilt symptoms and usually died before day 20 post inoculation (Fig. 6A). In contrast, plants inoculated with the ΔveA, ΔvelB and ΔveAΔvelB mutants displayed significantly lower mortality rates (P = 0.001, 0.003 and 0.001 respectively). The ΔlaeA strain was even more attenuated in virulence (P = 0.0005), and most of the plants inoculated with this mutant survived the assay or developed only mild disease symptoms (Fig. 6B). Complementation with the wild-type allele rescued the virulence defect of the different mutants.

Figure 6.

Velvet proteins and LaeA contribute to production of BEA in planta and to virulence of F. oxysporum on tomato plants.

A–C. Groups of 10 plants (cultivar Monica) were inoculated by immersing roots into a suspension of 5 × 106 freshly obtained microconidia ml−1 of the indicated fungal strains and planted in minipots. Per cent survival was recorded for 35 days. Mortality rates of plants infected with the ΔveA, ΔvelB, ΔveAΔvelB, ΔlaeA, Δbeas#4 and Δbeas#10 mutants were significantly lower (P < 0.05) than those of plants infected with the wild type, ΔvelC or complemented strains. Data shown are from one representative experiment. All experiments were performed at least three times with similar results.

D. Real-time qPCR performed on gDNA extracted from roots or stems of tomato plants at different times after inoculation with the indicated fungal strains. Amplification levels are expressed relative to those obtained from plants inoculated with the wild-type strain. Bars represent standard errors from three independent experiments with three technical replicates each.

E. Production of BEA was quantified by liquid chromatography/tandem mass spectrometry in extracts from tomato roots infected with the indicated fungal strains, at 7 days post inoculation. BEA levels are expressed in μg g−1 plant tissue. Bars represent standard errors from three independent experiments.

Fusarium oxysporum penetrates the roots of the host plant through wounds or natural openings (Perez-Nadales and Di Pietro, 2011) and grows inter- and intracellularly through the cortex before entering the vascular bundles (Pareja-Jaime et al., 2010). The fungus then uses the xylem vessels as avenues to grow upwards into the stem and colonize the plant, provoking the characteristic wilt symptoms. To determine the role of LaeA in host colonization, we measured the amount of fungal DNA in tomato plants infected with the different strains, using real-time qPCR with Fusarium-specific primers. No significant differences were detected in the root colonization ability of the wild-type strain and the ΔlaeA mutant at 7 or 14 days post inoculation, but colonization of stems at 14 days post inoculation was markedly reduced in the ΔlaeA mutant (Fig. 6D). These results demonstrate that VeA and VelB contribute to virulence of F. oxysporum on tomato plants, while LaeA plays a key role during late infection stages and is essential for successful colonization of the xylem vessels and for development of vascular wilt symptoms.

The mycotoxin BEA acts as a virulence factor during infection of F. oxysporum on mammals and plants

The velvet complex controls multiple developmental and metabolic processes, some of which may account for the observed virulence attenuation in the mutants. We found that F. oxysporum produces significant levels of BEA (3.5 ng ml−1) during growth in human blood (Fig. 5E). Besides BEA, fusaric acid, was also detected in blood cultures of the wild-type strain (Fig. 5F). The ΔveA and ΔlaeA mutants showed a sharp reduction in BEA and fusaric acid levels, while the Δbeas mutant failed to produce BEA, but still produced wild-type levels of fusaric acid.

The Δbeas#10 mutant caused significantly lower mortality (P = 0.01) on immunodepressed mice than the wild type (Fig. 5G). Mortality was also decreased in a second independent knockout strain (Δbeas#4), although the difference with the wild type was not statistically significant (P = 0.1). Kidneys and lungs of mice infected with both Δbeas mutants contained significantly less fungal burden than those of animals inoculated with the wild-type strain (Fig. 5H).

During infection of tomato roots, the F. oxysporum wild-type strain produced detectable levels of BEA (1.15 μg g−1 of root tissue), but not of fusaric acid. In plants inoculated with the ΔveA, ΔvelB or ΔlaeA mutants, BEA levels were reduced 58-, 3- or 3-fold, respectively, and no BEA was detected in plants inoculated with the ΔveAΔvelB, Δbeas#4 or Δbeas#10 strains (Fig. 6E). Mortality of tomato plants inoculated with two independent Δbeas strains was attenuated (P = 0.007) in relation to plants inoculated with the wild type (Fig. 6C). We conclude that BEA is produced by F. oxysporum during infection of mice and tomato roots and contributes to virulence on mammalian and plant hosts.

We also generated targeted deletion mutants in the abc3 gene (Fig. S11), which is located next to the beas and kiv genes, and whose expression is reduced in the velvet complex mutants (see Fig. 4C). Two independent Δabc3 mutants displayed wild-type virulence levels on immunodepressed mice, but caused significantly lower mortality on tomato plants (P = 0.008) (Fig. S12). Thus, the ABC3 multidrug transporter is required for full virulence of F. oxysporum on a plant host.

Discussion

Developmental roles of the F. oxysporum velvet proteins

An inventory of the velvet protein family in F. oxysporum detected three members, VeA, VelB and VelC. Deletion of veA and velB had profound effects on hyphal growth and development. Major phenotypes of the mutants include a flat colony morphology, premature asexual development and increased conidiation in submerged culture, which are reminiscent of veA mutants in other fungal species (Calvo, 2008), such as A. nidulans, A. fumigatus, N. crassa, Penicillium chrysogenum and Fusarium graminearum (Kim et al., 2002; Kato et al., 2003; Bayram et al., 2008b; Hoff et al., 2010; Jiang et al., 2011; Park et al., 2012). In contrast, the ΔvelC mutant showed only minor developmental phenotypes. This suggests that VelC has an auxiliary role in F. oxysporum, as previously reported in A. nidulans and A. fumigatus (Sarikaya Bayram et al., 2010; Park et al., 2012).

Most of the velvet phenotypes are more severe in ΔveA than in ΔvelB mutants. This suggests that VeA has additional functions that are independent of VelB, as previously reported in Aspergillus (Bayram et al., 2008a; Bayram and Braus, 2012). We speculate that some of these additional VeA functions could be carried out in association with VelC. This idea is supported by genetic evidence showing a significant contribution of VelC in light-dependent functions such as repression of submerged conidiation or chromatin remodelling at secondary metabolite gene clusters (Figs 2C, 3B and 4B). Our yeast two-hybrid results suggest that VeA can interact both with VelB and with VelC, consistent with a hypothetical model in which VeA forms at least two distinct complexes (Fig. 7). In vivo protein interaction experiments are required to corroborate this hypothesis.

Figure 7.

Roles of the F. oxysporum velvet complex in regulation of hyphal growth and development. Proposed model for the role of different velvet complex proteins in hyphal growth and conidiation. VeA forms complexes with VelB or VelC that have both overlapping and contrasting functions in different developmental processes. Interactions depicted in the model are based on yeast two-hybrid analysis and on genetic evidence. For simplicity, putative interactions VelB–VelB and VelB–VelC are not shown. Model partially based on that proposed by Bayram and Braus (2012).

The increase in length and size of microconidia was more severe in ΔvelB than in ΔveA, which is in contrast to most other mutant phenotypes. Unexpectedly, deletion of veA in the ΔvelB background recapitulated the less severe ΔveA phenotype. Together with the finding that ΔvelC displays the opposite phenotype, namely smaller and shorter conidia, this result suggests a hypothetical model in which VeA–VelB and VeA–VelC complexes have opposite roles in determination of conidia size: positive for VeA–VelC and negative for the VeA–VelB (Fig. 7). Deletion of velB would result in loss of the negatively regulating complex and increased conidia size, whereas deletion of velC creates the opposite effect. Knockout of veA abolishes both complexes, leading to a moderate increase in conidia size. Such exquisite fine-tuning of conidial development is consistent with the ability of Fusarium to differentiate different types of asexual spores termed microconidia, macroconidia and chlamydospores. Even though F. oxysporum generally does not produce macroconidia in submerged culture, macroconidia-like spores were frequently observed in cultures of the ΔvelB mutant. Likewise, knockout of veA in Fusarium verticillioides led to a marked increase in the ratio of macroconidia to microconidia (Li et al., 2006).

Yeast two-hybrid experiments also detected a self-interaction VelB–VelB, which was recently described in A. nidulans (Sarikaya Bayram et al., 2010), as well as a previously unreported interaction VelB–VelC whose biological role is currently unknown. Unexpectedly, we failed to detect a F. oxysporum orthologue of VosA, the fourth member of the velvet protein family in Aspergillus (Ni and Yu, 2007). Lack of VosA appears to be specific for the genus Fusarium, since F. graminearum, F. verticillioides and F. solani also lack VosA orthologues while other ascomycetes such as N. crassa, Chaetomium globosum or M. oryzae all have predicted VosA proteins (Fig. 1). VosA is primarily involved in regulation of sporogenesis and trehalose biogenesis of A. nidulans (Ni and Yu, 2007) and forms a light-regulated complex with VelB (Sarikaya Bayram et al., 2010). It is currently unknown whether the VosA-specific functions in Fusarium are taken over by the other velvet complex proteins.

The conidiation phenotype of the F. oxysporum ΔlaeA mutants was opposite to that of ΔveA, decreased versus increased respectively. P. chrysogenum ΔlaeA mutants also displayed reduced conidiation, in contrast to ΔveA mutants (Hoff et al., 2010). This indicates that, as previously suggested in A. nidulans (Sarikaya Bayram et al., 2010), F. oxysporum LaeA inhibits certain developmental functions of velvet.

In summary, our data suggest that F. oxysporum velvet proteins form different complexes, some of which could have competing functions. VeA has a key role in complex formation, since it is the only velvet-like protein to interact with LaeA in the yeast two-hybrid assay. This finding explains the generally more severe phenotype of the ΔveA mutant relative to ΔvelB and ΔvelC mutants. However, our interaction data are solely based on yeast two-hybrid experiments and thus should be viewed with caution. A recent study confirmed in vivo interaction of VeA and LaeA in the nucleus of Fusarium fujikuroi, using an alternative approach based on bimolecular fluorescence complementation (Wiemann et al., 2010).

Velvet participates in chromatin remodelling and transcriptional activation of secondary metabolite gene clusters

In this study, we establish a light-dependent role for the velvet complex in biosynthesis of three secondary metabolites, ferricrocin, triacetylfusarinine C and BEA. In A. nidulans, VeA and LaeA govern biosynthesis of the secondary metabolites sterigmatocystin, penicillin and lovastatin (Bok and Keller, 2004; Calvo, 2008; Bayram et al., 2008a). VeA was also shown to activate expression of genes involved in biosynthesis of cephalosporin C in Acremonium chrysogenum (Dreyer et al., 2007), or production of the deleterious mycotoxins fumonisin, fusarin C, trichotecene or deoxynivalenol in the plant pathogens F. verticillioides, F. fujikuroi and F. graminearum (Myung et al., 2009; Wiemann et al., 2010; Jiang et al., 2011; Merhej et al., 2012). Transcriptional profiling in the human pathogen A. fumigatus revealed that LaeA controls expression of 13 of 22 secondary metabolite gene clusters, including those involved in biosynthesis of siderophores and mycotoxins (Perrin et al., 2007). Likewise, LaeA regulates expression of multiple secondary metabolism gene clusters in the plant parasite Aspergillus flavus (Georgianna et al., 2010). LaeA deletion caused downregulation of gene clusters encoding biosynthesis of bikaverin, fumonisin, fusaric acid, fusarin and two unknown secondary metabolite clusters (Butchko et al., 2012).

Importantly, our work correlates reduced transcript levels in the F. oxysporum velvet complex mutants with a significant decrease in chromatin accessibility at the ferricrocin and BEA gene clusters (Figs 3 and 4). This result fits the generally accepted view that tightly positioned nucleosomes repress gene activity whereas loss of nucleosome positioning leads to transcriptional activation (Felsenfeld and Groudine, 2003). Moreover, these findings strongly suggest that the main regulatory function of the velvet complex in F. oxysporum resides in the modification of chromatin structure at the target loci. The exact mechanism of chromatin remodelling by velvet remains to be elucidated but is likely to depend on LaeA function. Like other fungal orthologues, F. oxysporum LaeA contains a conserved domain of methyltransferase SAM binding residues (Bayram and Braus, 2012). Site-directed mutation of the s-adenosyl methionine binding site in A. nidulans LaeA resulted in a loss-of-function phenotype, suggesting that LaeA acts as a methyltransferase (Bok et al., 2006). Chromatin modification by LaeA was recently shown to involve reversal of histone H3 lysine 9 trimethylation and the concomitant removal of heterochromatic marks (Reyes-Dominguez et al., 2010).

Velvet controls fungal virulence factors such as the mycotoxin BEA

Our study establishes a conserved role of the velvet complex during infection of F. oxysporum on both plant and mammalian hosts. LaeA had a more severe effect on virulence than VeA, indicating that regulation of secondary metabolism is a key contribution of the velvet complex during infection. In support of this idea, production of BEA and fusaric acid, two known mycotoxins, was impaired in the velvet complex mutants during growth in human blood and on tomato roots. BEA and its close structural relative enniatin are cyclic hexadepsipeptides consisting of alternating d-α-hydroxy-isovaleryl-(2-hydroxy-3-methylbutanoic acid) and amino acid-units (Jestoi, 2008). BEA was first isolated from the culture of the entomopathogenic fungus B. bassiana (Hamill et al., 1969), and its production has been reported in several species of the genus Fusarium including different formae speciales of F. oxysporum (Logrieco et al., 1998; Moretti et al., 2002; Song et al., 2008).

We confirmed that BEA is a virulence factor of F. oxysporum by deleting the NRPS responsible for BEA biosynthesis. The Δbeas mutants were attenuated in virulence on mice and on tomato plants. Targeted knockout of the beas gene in the entomopathogen B. bassiana revealed a significant role of this mycotoxin in virulence on insect hosts (Xu et al., 2008). Mutants of the cereal pathogen F. avenaceum lacking enniatin synthetase exhibited significantly reduced virulence in an infection assay on potato tubers (Herrmann et al., 1996a). These compounds display a wide array of biological activities including antiviral, antibacterial, nematicidal, insecticidal and cytotoxic (Jestoi, 2008). So far, however, their exact function during infection of plant and animal hosts remains unknown. BEA and enniatins act as potent and specific inhibitors of ABC-type multidrug efflux pumps such as Saccharomyces cerevisiae Pdr5p (Hiraga et al., 2005) and human ABCB1 and ABCG2 (Dornetshuber et al., 2009). Moreover, BEAS induces membrane scrambling and apoptosis in human erythrocytes (Qadri et al., 2011), while enniatin causes necrosis of potato tuber tissue, suggesting that cyclic hexadepsipeptides can trigger cell death both in mammalian and in plant cells (Herrmann et al., 1996b).

Components of the velvet complex have previously been associated with virulence in fungal human and plant pathogens. A role of LaeA in virulence of A. fumigatus was partly attributed to regulation of biosynthesis of gliotoxin (Sugui et al., 2007), an epidithiodioxopiperazine that induces apoptosis in different human cell types (Scharf et al., 2012). Interestingly, reduced virulence of ΔlaeA mutants of A. fumigatus was accompanied by decreased levels of pulmonary gliotoxin and changes in susceptibility to host phagocytes ex vivo (Bok and Keller, 2004; Bok et al., 2005). In the dimorphic human pathogen Histoplasma capsulatum, the orthologues of VosA, VelB and VeA are required for correct switching from hyphal to yeast stage and for full virulence (Webster and Sil, 2008; Laskowski-Peak et al., 2012). Null mutants in veA or laeA of the mycotoxigenic fungus A. flavus failed to metabolize host cell lipid reserves, resulting in inhibition by oleic acid and reduced seed colonization (Amaike and Keller, 2009). Strains of F. graminearum lacking VeA or VelB showed reduced in vitro expression of trichothecene and deoxynivalenol biosynthesis genes and attenuated virulence on wheat heads (Jiang et al., 2011; Lee et al., 2012; Merhej et al., 2012), while F. fujikuroi ΔveA, ΔvelB and ΔlaeA mutants caused less bakanae etiolation symptoms on rice seedlings, likely due to impaired biosynthesis of gibberelic acid (Wiemann et al., 2010), and F. verticillioides ΔveA strains displayed reduced fumonisin and fusarin production and infection on maize seedlings (Myung et al., 2009; 2012). In a recent study, production of the host selective polyketide T-toxin in the corn pathogen Cochliobolus heterostrophus was reduced in ΔveA or ΔlaeA mutants and increased in veA or laeA overexpression strains (Wu et al., 2012). In contrast, veA was dispensable for virulence in two other plant pathogens, Mycosphaerella graminicola (Choi and Goodwin, 2011) and Dothistroma septosporum (Chettri et al., 2012).

The reduction in virulence of the Δbeas mutants was less severe than that of the ΔveA and ΔlaeA mutants, suggesting that the velvet complex controls additional pathogenicity functions. We found that velvet is also required for production of fusaric acid (5-n-butyl-2-pyridine carboxylic acid) during growth of F. oxysporum in human blood. Fusaric acid was identified more than 50 years ago as a phytotoxin produced by F. oxysporum on tomato plants, and has been implicated in the development of vascular wilt symptoms by altering water permeability of the plant plasma membrane and causing solute leakage (Gäumann, 1957; 1958). Although we were unable to detect fusaric acid in infected tomato plants, possibly due to a low level of production or to instability of the compound in the plant tissue, it cannot be excluded that F. oxysporum produces fusaric acid during specific stages of infection. The recent identification of the fusaric acid biosynthetic gene cluster in F. verticillioides (Butchko et al., 2012) opens the way for future studies on the role of this mycotoxin in fungal virulence. Moreover, mutants in a third target gene of the velvet complex encoding an ABC3 multidrug transporter were also attenuated in virulence on tomato plants. Interestingly, the orthologous abc3 gene in M. oryzae functions as a virulence factor on rice plants (Sun et al., 2006). Additional putative virulence genes controlled by velvet include those encoding two NRPSs that participate in biosynthesis of the siderophores ferricrocin and triacetylfusarinine C. While the role of siderophores during infection of F. oxysporum has not been determined, a recent study found that the regulator of iron homeostasis HapX is required for virulence on mice and on tomato plants (López-Berges et al., 2012). Finally, important developmental functions of LaeA may account for the reduced virulence of the ΔlaeA mutant during late stages of infection. Reduced conidiation of the mutant could lead to a delay in colonization of the vascular bundles and attenuated wilt symptoms. Collectively, these results demonstrate that the velvet complex controls expression of multiple virulence-related genes in F. oxysporum, some of which are specific for either plant or mammalian infection while others, such as beas, contribute to virulence on both types of hosts.

Experimental procedures

Fungal isolates and culture conditions

Fusarium oxysporum f. sp lycopersici race 2 wild-type isolate 4287 (FGSC 9935) was used in all experiments. Fungal strains were stored as microconidial suspensions at −80°C with 30% glycerol. For extraction of genomic DNA and microconidia production, cultures were grown in potato dextrose broth (PDB) at 28°C (Di Pietro and Roncero, 1998). For analysis of gene expression and relative chromatin accessibility, freshly obtained microconidia were germinated in the dark for 14 h in PDB. Mycelia were harvested by filtration, washed three times in sterile water, transferred to fresh PDB and maintained for 10 h under continuous light or dark. For microscopic analysis, fungal strains were grown 3 days in PDB or 24 h in Yeast Peptone Glucose (YPG) medium (0.3% yeast extract, 1% peptone and 2% glucose), respectively, and observed in a Leica DMR microscope using the Nomarsky technique. Images were recorded with a Leica DFC 300 FX digital camera. Conidiation is was quantified in static cultures grown in continuous dark as described (Li et al., 2006). For determination of colony growth, 2 × 104 microconidia were spotted onto PDA or YPGA (0.3% yeast extract, 1% peptone, 2% glucose and 1.5% agar). Plates were incubated at 28°C for the indicated time periods. All experiments included two replicate plates and were performed at least three times with similar results.

Nucleic acid manipulations and quantitative real-time RT-PCR analysis

Total RNA and genomic DNA was extracted from F. oxysporum mycelia following previously reported protocols (Raeder and Broda, 1985; Chomczynski and Sacchi, 1987). Quality and quantity of extracted nucleic acids were determined by running aliquots in ethidium bromide-stained agarose gels and by spectrophotometric analysis in a NanoDrop ND-1000 spectrophotometer (NanoDrop Technologies) respectively. Routine nucleic acid manipulations were performed according to standard protocols (Sambrook and Russell, 2001). DNA and protein sequence databases were searched using the blast algorithm (Altschul et al., 1990).

RT-qPCR was performed as described (Lopez-Berges et al., 2010) using iQ SYBR Green Supermix in an iCycler iQ real-time PCR System (both from Bio-Rad). Gene-specific primers (Table S2) were designed to flank an intron, if possible. Transcript levels were calculated by comparative ΔCt and normalized to act1. Expression values are presented as values relative to the expression in the wild-type strain.

Analysis of chromatin structure

Mycelia of F. oxysporum strains grown under the indicated conditions were harvested by filtration, ground to a fine powder under liquid nitrogen and lyophilized. Nuclease digestion was performed as described (Gonzalez and Scazzocchio, 1997; Basheer et al., 2009). Briefly, 20 mg of lyophilized mycelium was suspended in 1 ml of MNase buffer (250 mM sucrose, 60 mM KCl, 15 mM NaCl, 0.5 mM CaCl2, 3 mM MgCl2), and 300 μl of the suspension was treated for 3 min with 0.5 U of micrococcal nuclease (MNase, Sigma) at 37°C. The reaction was terminated, by adding stop buffer (2% SDS, 40 mM EDTA). DNA was obtained by phenol/chloroform extraction, precipitated, washed with 70% ethanol and dissolved in water (see Fig. S9). Quantitative real-time PCR was performed as described above using promoter-specific primer pairs. Chromatin accessibility was expressed by comparative ΔCt as the ratio between amplification levels from untreated gDNA relative to those obtained from MNase digested gDNA. Values were represented relative to those of the wild-type strain.

Targeted gene knockout

Targeted gene replacement was performed as detailed in Figs S3–S6, using the split-marker method. A 2.8 kb fragment from the plasmid pAN7-1 containing the hygromycin B resistance gene (hygr) under control of the A. nidulans gpdA promoter (PgpdA) and trpC terminator (TtrpC) (Punt et al., 1987) was used. All PCR reactions were routinely performed with the High Fidelity Template PCR system (Roche Diagnostics) using a MJ Mini Bio-Rad personal thermal cycler. Fungal transformation and purification of the transformants by monoconidial isolation were done as described (Di Pietro and Roncero, 1998). Transformants showing homologous insertion of the construct were detected by PCR of genomic DNA and by Southern blot analysis. Complementation of deletion mutants with a PCR fragment encompassing the wild-type allele was done by co-transformation with the phleomycin B resistance (phleor) gene under control of the A. nidulans gpdA promoter and trpC terminator, obtained from plasmid pAN8-1 (Mattern et al., 1988). Complemented transformants were identified by PCR with gene-specific primers.

Plant infection assays

Tomato root inoculation assays were performed as described (Di Pietro and Roncero, 1998), using 2-week-old tomato seedlings (cultivar Monika). Severity of disease symptoms and plant survival was recorded daily for 30 days. Ten plants were used for each treatment. Virulence experiments were performed at least three times with similar results. Plant survival was calculated by the Kaplan-Meier method and compared among groups using the log-rank test. Data were analysed with the software GraphPad Prism 4. Quantification of fungal biomass in planta was performed as described (Pareja-Jaime et al., 2010), using total genomic DNA extracted from tomato roots or stems infected with F. oxysporum strains at 7 or 14 days post inoculation. Relative amounts of fungal gDNA were calculated by comparative ΔCt of the Fusarium-specific six1 gene normalized to the tomato-specific gadph gene.

Animal infection assays

Mice were cared for in accordance with the principles outlined by the European Convention for the Protection of Vertebrate Animals Used for Experimental and Other Scientific Purposes (European Treaty Series, No. 123; http://conventions.coe.int/Treaty/en/Treaties/Html/123.htm). Experimental conditions were approved by the Animal Welfare Committee of the Faculty of Medicine, Universitat Rovira I Virgili. Infection assays with immunodepressed mice were performed as described (Ortoneda et al., 2004). Survival was recorded daily for 15 days. Infection experiments with each individual strain were performed at least three times with similar results. Survival was calculated by the Kaplan-Meier method and compared among groups using the log-rank test. Determination of fungal tissue burden was performed as described (Ortoneda et al., 2004). Fungal colony counts were converted to log10 and compared using the anova test. Data were analysed with the software GraphPad Prism 4.

Sequence alignments and phylogenetic analysis

Velvet family proteins from different fungal genomes were initially identified by blastp search on the Broad Institute (http://www.broadinstitute.org/) or the NCBI (http://www.ncbi.nlm.nih.gov/blast/Blast.cgi) web sites, using the sequence of the A. nidulans VeA protein. Full-length protein sequences were aligned with clustal w (Thompson et al., 1994) and inspected manually. A maximum likelihood tree was built from the alignment by PhyML version 4.0 using both parsimony and distance analysis (neighbour joining; NJ) with 1000 bootstrap replicates and represented as a cladogram with Dendroscope v1.2.3 (Guindon and Gascuel, 2003).

Mycotoxin quantification

Samples were obtained from fungal colonies grown 3 days at 28°C on PDA or Czapek Dox Agar (Difco) plates, or 3 days at 37°C in heparinized human whole blood (Dunn Labortechnik GmbH, Asbach, Germany; Cat. No. IPLAWB) that was drawn from healthy donors of ages 18–65 at FDA licensed facilities; or from roots of tomato plants infected with F. oxysporum as described above, at 7 days post inoculation. Samples were homogenized in acetonirile/water/glacial acetic acid (79:20:1; v : v : v) with a workcentre T10 basic (IKA®) for 1 min at a rate of 4 ml solvent per gram of sample. The mix was re-homogenized after 2 min of repose, filtered, centrifuged 10 min at 12 000 g, and the supernatant was lyophilized. Dry crude extracts were reconstituted in the solvent, and mycotoxin detection and quantification was performed with a QTrap 4000 LC-MS/MS System (Applied Biosystems, Foster City, CA) equipped with a TurboIonSpray electrospray ionization (ESI) source and an 1100 Series HPLC System (Agilent, Waldbronn, Germany), as described (Vishwanath et al., 2009).

Accession numbers

Sequence data from this article can be found in the GenBank/EMBL database or in the Fusarium Genome database under the following accession numbers: F. oxysporum VeA, FOXG_11273; VelB, FOXG_00016; VelC, FOXG_02050; Beas, FOXG_11847; Kivr, FOXG_11846; Abc3, FOXG_11845; SidC, FOXG_06448; SidD, FOXG_09785; SidA, FOXG_06447; Act1, FOXG_01569; Six1, FOXG_16418; Gapdh, M64114; F. fujikuroi LaeA, CBE54370; A. nidulans LaeA, CBF88745; N. crassa LaeA, NCU00646; M. oryzae LaeA, MGG_07964; pAN7-1 (PgpdA-hygr-TtrpC), Z32698; and pAN8-1 (PgpdA-phleor-TtrpC), Z32751.

Acknowledgements

We are grateful to Esther Martínez Aguilera (Universidad de Córdoba) for technical assistance and to Gerhard Adam (University of Natural Resources and Applied Life Sciences Vienna) for initiating the collaboration between the Córdoba and Vienna labs. This research was supported by the following grants: BIO2010-15505 from Ministerio de Ciencia e Innovación (MICINN); ERA-NET/PathoGenoMics project TRANSPAT (BIO2008-04479-E from MICINN); EUI2009-03942 from MICINN/Plant KBBE; BIO-3847 from Junta de Andalucia; Marie Curie ITN ARIADNE (FP7-PEOPLE-ITN-237936). M.S.L.-B. was recipient of a PhD fellowship from MICINN.

Ancillary