A Conserved Homeobox Transcription Factor Htf1 Is Required for Phialide Development and Conidiogenesis in Fusarium Species

Conidia are primary means of asexual reproduction and dispersal in a variety of pathogenic fungi, and it is widely recognized that they play a critical role in animal and plant disease epidemics. However, genetic mechanisms associated with conidiogenesis are complex and remain largely undefined in numerous pathogenic fungi. We previously showed that Htf1, a homeobox transcription factor, is required for conidiogenesis in the rice pathogen Magnaporthe oryzae. In this study, our aim was to characterize how Htf1 homolog regulates common and also distinctive conidiogenesis in three key Fusarium pathogens: F. graminearm, F. verticillioides, and F. oxysporum. When compared to wild-type progenitors, the gene-deletion mutants in Fusarium species failed to form conventional phialides. Rather, they formed clusters of aberrant phialides that resembled elongated hyphae segments, and it is conceivable that this led to the obstruction of conidiation in phialides. We also observed that mutants, as well as wild-type Fusaria, can initiate alternative macroconidia production directly from hyphae through budding-like mechanism albeit at low frequencies. Microscopic observations led us to conclude that proper basal cell division and subsequent foot cell development of macroconidia were negatively impacted in the mutants. In F. verticillioides and F. oxysporum, mutants exhibited a 2- to 5- microconidia complex at the apex of monophialides resulting in a floral petal-like shape. Also, prototypical microconidia chains were absent in F. verticillioides mutants. F. graminearum and F. verticillioides mutants were complemented by introducing its native HTF1 gene or homologs from other Fusarium species. These results suggest that Fusarium Htf1 is functionally conserved homeobox transcription factor that regulates phialide development and conidiogenesis via distinct signaling pathways yet to be characterized in fungi.


Introduction
Asexual sporulation is the preferred mode of reproduction in most pathogenic fungi [1,2]. More importantly, these asexual spores, commonly known as conidia, are used as a primary dissemination tool as well as for initiating infection [3][4][5][6][7]. Under favorable conditions, fungal pathogens can rapidly propagate and spread to cause diseases in economically important crops as well as humans and animals. Significantly, recent studies have shown that fungal pathogens responsible for plant diseases can also cause opportunistic mycosis in humans [8,9]. However, the mechanisms of asexual sporulation are diverse and complex and remain largely undefined in numerous pathogenic fungi. In model organisms, namely Neurospora crassa and Aspergillus nidulans, signaling pathways that regulate conidiation have been extensively studied, and excellent reviews are available [2,10,11]. Briefly, there are key transcriptional regulators known to be involved in this process. One important gene that plays a critical role in the transition from conidiophore to condia formation is brlA in A. nidulans [12]. Further genetic and biochemical studies led to the discovery of abaA and wetA [13,14]. These three genes (brlA-abaA-wetA) have been proposed to constitute a central regulatory pathway that acts in concert with other genes to control conidiation in Aspergillus [15,16]. However, we also need to recognize that different fungal species may have developed different regulatory mechanisms for producing various types of conidia.
The genus Fusarium is considered the most important and diverse genera of plant pathogenic fungi and causes a wide range of diseases in every economically important crop species [17,18]. Several species within the genus are also associated with the production of mycotoxins which poses a significant threat to food safety and human health [19][20][21][22] Moreover, once considered a relatively uncommon cause of ocular disease, Fusarium species have emerged as one of the leading causes of human keratomycosis outbreaks, along with Aspergillus and Candida species [23,24] The genomes of closely related Fusarium species, F. graminearum, F. verticillioides, and F. oxysporum, have been sequenced mainly due to their economic and scientific importance [18,25,26]. In addition, these Fusarium species offer a unique opportunity to investigate numerous biological features, including distinct asexual sporulation modes. In recent years, a number of conidia-related genes in Fusarium species have been identified by insertional mutagenesis or targeted gene deletion approaches. Several genes are important transcriptional regulators, such as FgSTUA, FoSTUA and REN1, which are conserved in filamentous fungi and essential for conidiogenesis [27][28][29]. Genes such as FgVEA, FvVE1 and FgTEP1 are involved in multiple signaling pathways, regulating virulence, secondary metabolism and conidiation [30][31][32]. Some important signal transduction related genes, e.g., GPMK1, GzSNF1 and FAC1, which encode various protein kinases are also required for conidiation in Fusarium species [33][34][35][36][37]. However, molecular mechanisms underlying conidiogenesis in Fusarium species is complex and does not seem to adhere to the regulatory pathway established in A. nidulans and N. crassa [2,10,11,38]. Therefore, it is currently difficult to unambiguously define genetic mechanisms or signaling pathways required for this important biological process in Fusaria.
Although F. graminearum, F. verticillioides and F. oxysporum are defined into the same genus, they exhibit distinct features in asexual sporulation. For instance, F. graminearum only produces macroconidia on solitary phialides or on multiple phialides borne on conidiophores [27]. In F. verticillioides, the fungus grows as haploid mycelia and propagates vegetatively via hyphal elongation and produces two types of asexual spores, macroconidia and microconidia [39]. Macroconidia emerge from macroconidiophores, which are branched and unbranched monophialides [40,41]. Similarly, unicellular and uninucleate microconidia also arise from branched and unbranched monophialides, frequently forming long conidial chains and false heads. When compared to the other two, F. oxysporum is unique in the fact that it can only reproduce asexually, but through three different types of conidia: microconidia, macroconidia, and chlamydospores [40][41][42]. Microconidia are ellipsoidal and have no or one septum, macroconidia are falcate and have three or four septa, and chlamydospores are globose-like with thick walls [29,40].
Despite the structural difference between macroconidia and microconidia, conidiation pattern is very similar, and the use of enteroblastic mechanisms from the phialide is common in all three species [43]. In Fusarium species, phialides are cylindrical, solitary or produced as a component of a complex branching system. Microconidia are formed from phialides with false heads or from long chains by basipetal division, from the apex toward the base [44,45]. Macroconidia with pronounced foot cells are generally produced from phialides on conidiophores also by basipetal division [46]. However, in F. oxysporum the chlamydospores are produced acrogenously from hyphae or by the modification of hyphal cells and conidial cells through the condensation of their contents [47].
Transcriptional regulation plays a critical role in altering the expression of specific subsets of genes associated with development and differentiation in cells [48,49]. In our previous study, we identified and characterized Htf1, an important homeobox transcription factor (TF) required for conidiogenesis, in Magnaporthe oryzae [7,50]. We reasoned that Htf1 may also play a critical role in Fusarium conidiation. However, considering that different Fusarium species have different mode of conidiogenesis, we also hypothesized that Htf1 can perform unique function in different species. In this study, our aim was to functionally characterize Htf1 orthologs in F. graminearum, F. verticillioides and F. oxysporum, and determine its role in conidiogenesis. Results showed that Htf1 ortholologs in three Fusarium species have significant similarity in specifically regulating phialidegenesis and subsequent macroconidiation. Moreover, Htf1 ortholologs in F. verticillioides and F. oxysporum are required for morphogenesis of microconidial chains and false heads.

Comparative analysis of Htf1 orthologs in Fusarium species
Htf1 was reported previously as a key regulator of conidiogenesis in M. oryzae [7,50]. To investigate functional conservation, we first isolated genes encoding Htf1 homolog in three Fusarium species as well as other filamentous fungi via BLAST analysis. Study of databases (Broad Institute Fungal Genome Initiative [www.broadinstitute.org] and Fungal Transcription Factor Database [ftfd.snu.ac.kr]) revealed that M. oryzae Htf1 homolog is present in these fungal species and that these homologs contain a conserved homeodomain motif, predominantly located in the Nterminus region (data not shown). As anticipated, the homeodomain motif was highly conserved in Fusarium species and M. oryzae ( Figure 1A), but a high level of variability was observed in the Cterminus region between M. oryzae and Fusarium species. When we compared the predicted protein sequence of F. graminearum FGSG_07097 gene (designated FgHTF1), F. verticillioides FVEG_08072 gene (designated FvHTF1), F. oxysporum FOXG_01706 gene (designated FoHTF1), and M. oryzae MGG_00184 gene (previously HTF1 [50], designated MoHTF1 in this manuscript), identity was greater than 95% at the protein level within the homeodomain amongst Fusarium species, and MoHtf1 homeodomain shared 60% identity with Fusarium counterparts ( Figure 1B). These genes also share a highly conserved exon/intron structure ( Figure S1A). Outside the homeobox domain region, protein similarity drops significantly when compared across fungal species (data not shown), and no known functional motifs or biologically significant domains exist ( Figure S1B). In Arabidopsis thaliana Athb-12, the C-terminus region has been shown to serve as the activation domain [51], but functional role of Htf1 C-terminus region in fungi have not been characterized to date.
FgHTF1 is dispensable for vegetative growth and fertility but essential for conidiation In order to study the function of Htf1 ortholog in Fusarium conidiogenesis, we first deleted FgHTF1 in F. graminearum using gene replacement approach (Table S1, Figure S2A). Transformants were selected on hygromycin-amended medium, and gene deletion was confirmed by polymerase chain reaction (PCR) and Southern blot analyses ( Figure S2B and S2C). FgHTF1 deletion mutant (DFghtf1) showed no discernible difference in vegetative growth and sexual reproduction when compared to the wild-type strain PH-1 on complete medium (CM) and wheat kernels medium, respectively (Table 1, Figure S3A and S3B). However, DFghtf1 mutant showed significantly reduced macroconidia production in liquid carboxymethylcellulose (CMC) medium [52]. After three days, only 1.0060.87610 4 macroconidia were observed in DFghtf1, whereas 51.4466.64610 4 macroconidia were in PH-1 (Table 1). Even after fifteen days of incubation, DFghtf1 mutant did not recover the production of macroconidia compared to PH-1, indicating that this reduction in conidiation was not related to the duration of incubation ( Table 1). The defect in conidia production was fully recovered to the wild-type level in the complemented strain DFghtf1-Com, where the native promoter-driven FgHTF1 gene cassette was re-introduced into DFghtf1 ( Figure S2B and S2C, Table 1). These results indicate that FgHTF1 is critical for macroconidia production in F. graminearum.

FgHTF1 specifically regulates phialidegenesis and subsequent conidiation
In order to investigate the reason for significantly reduced conidiation in DFghtf1, we microscopically observed fungal tissues grown in CMC, a medium that promotes fungal conidiation. Under the same culture condition, PH-1 and DFghtf1-Com produced typical conidiogenous cells, i.e., phialides, which divide to produce incipient macroconidia. The morphology of phialides in PH-1 and DFghtf1-Com assumes a bottle-like shape ( Figure 2A). The mutants, however, did not produce these structures on its conidiophores, but rather formed clusters consisting of hyphal segments ( Figure 2A). Fluorescence staining of nuclei with 496diamidino-2-phenylindole (DAPI) showed that a phialide in the wild type was uninucleate and harbors a macroconidium ( Figure 2B). However, in DFghtf1 it appeared that multiple phialide-like structures were disorderly formed on a conidiophore with no macrocondia development ( Figure 2C). Presumably, the mutation in FgHTF1 led to abnormal conidiogenous cells with uncontrolled proliferation and the loss of conidiation capacity. These results suggest that FgHtf1 governs proper differentiation of phialides and is continuously required for maintenance of conidiogenesis in F. graminearum.

FgHTF1 regulates macroconidia basal cell division and foot cell development
While DFghtf1 deletion mutant lost its ability to produce macroconidia from phialides, which is the main conidiogenesis structure in F. graminearum, we still observed some conidia produced in CMC medium, suggesting that alternative conidiation mechanisms exist. After further microscopic observation, we detected DFghtf1 and PH-1 producing spores directly from hyphae, similar to budding observed in Saccharomyces cerevisiae, albeit at low frequencies ( Figure 3A). However there was also a significant difference in conidiogenesis between DFghtf1 and PH-1. In the early stages of culturing PH-1 (within 48 h), incipient conidium broke off from intercalary or terminal hyphae without distinct septation, whereas in DFghtf1 no conidium was observed at the early stages of culturing in CMC medium. Only after five days of incubation, DFghtf1 produced some matured conidia with evident septation at the tip of hyphae ( Figure 3A). In addition, we noticed that there was no recognizable narrow region that allows conidium to detach easily from hyphae, and hence, these conidia lack the typical enteroblastic phenomenon associated with PH-1 ( Figure 3A). In PH-1, the narrow region serves as the site for the production of macroconidium by cell division. Therefore, we inferred that the dissociative spores may be ruptured away from DFghtf1 hyphae by mechanical force during shaking incubation ( Figure 3B), and this may explain why macroconidia produced by DFghtf1 were morphologically aberrant ( Figure 3C). The wild-type macroconidia were moderately curved on the dorsal side and straight on the ventral surface with papillate apical cells and distinct foot-shaped basal cells ( Figure 3C). However, DFghtf1 conidia were grotesque without proper foot-shaped basal cells ( Figure 3C). These observations indicate that FgHTF1 is important for proper basal cell division and subsequent foot cell development in macroconidia produced directly from hyphae.
Aberrant macroconidia of DFghtf1 mutant still can germinate properly and be pathogenic on hosts While DFghtf1 produced a limited number of macroconidia by budding and have a defect in the foot cell, these spores were still able to germinate like wild type at 25uC in liquid CM with gentle agitation. After 1 h incubation, approximately 70% of macroconidia in DFghtf1 and PH-1 looked swollen ( Figure 4A). After 2 h, over 95% of macroconidia had at least one germ tube from terminal cells, intercalary cells, or both in the mutant ( Figure 4A, Table 1). To determine whether FgHTF1 has a role in pathogenicity, we inoculated wheat heads and wheat coleoptiles with conidia from PH-1 and DFghtf1. At 14 days post inoculation (dpi), typical ear rot symptoms were observed on wheat head inoculated with PH-1 and the DFghtf1 ( Figure 4B, Table 1). Similar brown lesions on coleoptiles and corn stalks infected by PH-1 and DFghtf1 were observed ( Figure 4C and 4D). These results showed that the aberrant DFghtf1 macroconidia can germinate properly and be pathogenic on hosts. Expression of FgHTF1 correlates with conidiophore development in F. graminearum In order to investigate the temporal and spatial pattern of FgHTF1 expression during conidiogenesis, FgHTF1 gene with its native promoter was fused in-frame to the green fluorescent protein (GFP)-encoding gene. The construct was then transformed into DFghtf1 protoplasts. Subsequently, we isolated three transformants expressing GFP in hyphae, and the presence of the FgHtf1-GFP construct was confirmed by PCR (data not shown). All positive transformants (DFghtf1-GFP) produced a similar number of conidia when compared to the wild-type progenitor. To investigate the expression patterns of FgHtf1 during conidia germination in F. graminearum, we followed GFP expression by fluorescence microscopy at different time points (24 h, 36 h, 48 h, 60 h and 72 h) after inoculating DFghtf1-GFP mycelia into CMC medium, which is conducive to spore production. GFP signals were not detectable or extremely weak from 24 h to 36 h when DFghtf1-GFP strain was incubated in CMC (data not shown). However, at 48 h GFP signal spiked, and the localization of FgHtf1 to nucleus was verified by GFP and ethidium bromide (EB) stain ( Figure 5A). To study expression patterns of FgHTF1 in PH-1, we extracted total RNA from PH-1 cultured in CMC medium at 24   stage ( Figure 5B), but leveled off at 60 h and 72 h. We have observed previously that conidiophores typically proliferate after 48 h time point in CMC culture (data not shown). Thus, we can conclude that FgHtf1 transcription factor is activated and localized to the nucleus prior to conidiogenesis and perhaps regulates expression of genes associated with conidiophore and phialide development in a temporal and spatial specific manner.
FvHTF1 regulates microconidia chain and false head formation in F. verticillioides F. graminearum only produces macroconidia as asexual reproduction, and to further investigate whether Htf1 plays a role in microconidiation in Fusarium species, we generated FvHTF1 null mutation (DFvhtf1) in F. verticillioides, which produces both macroconidia and microconidia. Homologous recombination that resulted in FvHTF1 deletion was identified by PCR and was further confirmed by Southern blot (Figure S4A and S4B). DFvhtf1 mutant did not show a significant difference in vegetative growth and microconidia morphology when compared to the wild-type strain A149 grown on various media, including complete medium (CM) and mung bean agar medium ( Figure S5A and S5B).
When examined under optical and scanning electron microscopes, A149 produced long chains of microconidia and false heads of microconidia aggregates ( Figure 6A and 6B), which are customary taxonomic features of F. verticillioides. In contrast, we did not find microconidia chains in DFvhtf1 even with repeated efforts on various cultures conditions, even though there was no distinct difference in conidiophore morphology. In addition, microconidia of DFvhtf1 did not stay attached to each other to form false heads as typically observed in the wild-type progenitor ( Figure 6A and 6B). In DFvhtf1, the shape of conidia false head produced from the conidiophore varied, some assumed a trifolium pretense-like shape, others looked like petal, and some also possess dichotomization shape ( Figure 6A and 6B). Although the shape of microconidia cluster produced by DFvhtf1 varied in their appearances, they displayed a common conidiogenesis pattern, which suggests that all microconidia were verticillate branches at the apex of monophialides ( Figure 6A and 6B). To further confirm the microconidial chain and false head shape defect in DFvhtf1, the mutant strain was transformed with the corresponding wildtype gene. Complementation of FvHTF1, where the native promoter-driven FvHTF1 gene cassette was re-introduced into DFvhtf1, restored microconidia chains and false head pattern ( Figures S4B and 6A), demonstrating that FvHTF1 is needed for the formation of microconidial chains and false head pattern in F. verticillioides.
When we assayed the production of microconidia on 7-day mung bean agar and synthetic low-nutrient agar (SNA) cultures, the wild-type strain produced about twice as much microconidia than DFvhtf1 ( Figure 6C and 6D), suggesting that a significant reduction in conidiation could be due to a defect in formation of microconidia chains. It is conceivable that altered microconidiation pattern observed in DFvhtf1 prohibits the fungus from developing long chains of microconidia through basipetal division typically observed in the wild-type F. verticillioides [40,41,44]. Microconidia production was fully restored to the wild-type level in the complemented strain DFvhtf1-Com ( Figure 6C and 6D).
The Htf1-regulated macroconidiation is conserved in F. verticillioides and F. graminearum While microconidia are the predominant form of asexual spores in F. verticillioides, it also produces macroconidia in nature and in certain laboratory conditions [36,53]. When we monitored the development process of macroconidiation in F. verticillioides, we observed the mechanism similar to F. graminearum, in which macroconidia are produced on solitary phialides or on multiple phialides borne on conidiophores ( Figure 7A). In DFvhtf1, we found that the macroconidiogenesis from phialide was impaired ( Figure 7A), however, the mutant used budding pattern to produce foot cell-defective macroconidia, which is identical to what we observed in F. graminearum DFghtf1 mutant ( Figure 7A).
In addition, we assayed for the amount of macroconidia in mung bean liquid medium under continuous dark and UV light conditions. In continuous dark condition, the wild-type and complemented strains produced a similar level of macroconidia, which typically accounts for approximately 5% of the total conidia harvested from F. verticillioides cultures after 7 days. Under the same conditions, however, only 2% of the conidia were macroconidia in the DFvhtf1 mutant ( Figure 7B). These data suggested that FvHTF1 plays an important role in macroconidia development. UV light is known to stimulate macroconidia production in Fusarium species [36,53], and under UV light approximately 17% of the total conidia harvested after 7 days of incubation were macroconidia in wild-type and complemented strains. This is a significant increase when compared to the continuous dark condition. However, in DFvhtf1 mutant the percentage of macroconidia (2%) was consistent with that produced under continuous dark condition ( Figure 7C). These results suggested that FvHTF1 is important for macroconidia production and that it may plays a role in cellular responses to UV light stimulus.
FoHTF1 is also required for the development of microconidia and macroconidia, but not for chlamydospores The results we obtained from F. graminearum and F. verticillioides studies led us to further explored F. oxysporum conidiogenesis. This fungus produces three types of asexual spores: microconidia, macroconidia, and chlamydospores. To determine whether the function of Htf1 is conserved in F. oxysporum conidiogenesis, we generated a gene-replacement mutant of FoHTF1 ( Figure S6). The mutant (DFohtf1) was normal in vegetative growth and microconidia morphology ( Figure S7), but when assayed for conidiation on SNA medium we found that DFohtf1 had a significant reduction in macroconidia and a slight reduction in microconidia when compared to the wild type ( Figure 8A and 8B). The DFohtf1 and the wild-type strain were examined under an optical microscope, and we noticed that the mutants lacked normal false head microconidia and formed windmill-shaped structure, which was congruent with microconidiogensis in DFvhtf1 mutant ( Figure 8C). However, DFohtf1 produced normal chlamydospores acrogenously from hyphae or by the modification of hyphal cells, as the wild type ( Figure 8D). Phialides of macroconidia also redundantly proliferated and developed constant extension very similar to DFghtf1 and DFvhtf1 mutants ( Figure 8E). These results indicate that FoHTF1 is important for conidiophore and phialide development, and ultimately microconidia and macroconidia production. However we concluded that FoHTF1 is not involved in hyphal differentiation that leads to chlamydospores in F. oxysporum.
The function of Htf1 is conserved in three Fusarium species Htf1 in three Fusarium species showed highly conserved functions in macroconidiogenesis and microconidiogenesis. F. graminearum FgHtf1 homeodomain and whole protein sequence  share greater than 95% and 85% amino acid identity, respectively, when compared to FvHtf1 and FoHtf1 ( Figure 1B). To test whether Htf1 homologs from three Fusarium species are functional orthologs, we transformed FvHTF1 and FoHTF1 genes with their respective promoter regions into DFghtf1 mutant. Positive transformants were identified by PCR with respective specific primers (Table S2), and these showed rescued conidiogenesis in DFghtf1 mutant when incubated in CMC medium ( Figure 9A, Table 1). In addition, we also transformed FgHTF1 gene into the DFvhtf1 mutant. Significantly, the complemented strain DFvhtf1-Fg produced abundant microconidia in chain and false head shape although F. graminearum species does not produce microconidia ( Figure 9B). These results suggest that Htf1 is conserved in three Fusaria and that FgHtf1 can transcriptionally regulate F. verticillioides proteins that are involved in microconidiogenesis.

Discussion
Conidiation is an important characteristic in fungi that requires spatial and temporal regulation of gene expression that leads to specialized cellular differentiation and intercellular communications [1,2,10]. In our previous study, we found that HTF1 is essential for conidiation in M. oryzae. Further observation revealed that DMohtf1 mutant produces greater amounts of conidiophores, which showed curvature slightly near the tip but could not develop into sterigmata-like structures ( Figure 10) [50]. This led us to conclude that MoHTF1 is an essential positive regulator responsible for switching from conidiophore maturation to the initiation of conidia development in M. oryzae. Concurrently, we also proposed that MoHTF1 functions as a negative regulator of conidiophore development. In other filamentous fungi, homeodomain transcription factors have been linked to the shaping of fruiting body structure, sexual reproduction, and mycelial branch formation [49,[54][55][56]. In this study, we hypothesized that, while there are similarities and conservation in HTF1 gene function between M. oryzae and Fusarium species, there are divergent biological features exhibited by Htf1 in Fusarium species. In three Fusarium species, we found that the deletion of HTF1 also abolished macroconidia development from conidophores. In   addition, Dhtf1 mutants in these Fusarium species failed to form morphologically discernible phialide, but rather forming ''clusters'' consisting of hyphal segments ( Figure 10). This phenotype was more profound in F. graminearum than F. verticillioides and F. oxysporum, however it was consistently observed in three Fusaria. In addition, the deletion of HTF1 in Fusarium species caused excessive elongation of conidiogenous cell, suggesting that Htf1 is a negative regulator of conidiogenous cell development similar to M. oryzae.
But we also discovered a major difference in conidiogenesis between M. oryzae and Fusarium species. In M. oryzae, the conidiogenous cell usually was deemed as conidiophore, while in Fusarium species it often represent phialides, and therefore it would be reasonable to presume that the function of HTF1 in Fusarium species underwent a further specialization for phialidegenesis. Some genes associated with phialide development have been reported in Fusarium species, such as FgStuA in F. graminearum, FoStuA and REN1 in F. oxysporum. The sequence of FgStuA protein showed a very high level of homology (72%) with FoStuA [27,29]. Not surprisingly, the deletion mutants of DFgStuA and DFoStuA lacked conidiophores and uninucleate phialides, suggesting a conserved function in two orthologs. In A. nidulans, where phialidic conidiation has been extensively studied, stuA mutants produced significantly stunted conidiophores and lacked normal metulae and phialides [57,58]. It is also recognized that stuA affects conidiation through the spatial and temporal modifier of brlA and abaA expression [59,60]. However, FgStuA regulates sexual development and pathogenicity in addition to conidiogenesis, suggesting that this transcription factor may have a broader and diverse impact on F. graminearum lifestyle [27]. Significantly, in contrast to FgStuA all gene-deletion mutants we studied (DMohtf1, DFghtf1, DFvhtf1, DFohtf1) showed phenotypic deformity only limited to conidiogenesis. Our SAGE data (unpublished) and previously published microarray study in PH-1 and DFgStuA [27] showed no reciprocal influence in FgHTF1 and FgSTUA, and it is reasonable to hypothesize that FgStuA and FgHtf1 regulate phialide development through different cellular networks.
REN1 encodes a protein analogous to A. nidulans MedA and M. oryzae Acr1, and all of these are involved in conidiogenesis [28]. The ren1 mutant strains lacked normal conidiophores and phialides and formed rod-shaped, conidium-like cells directly from hyphae by acropetal division [28], but maintained pathogenicity on host. These results showed that Ren1 specifically regulate conidiogenesis. In our study, we concluded that FgHTF1 does not directly play a role in conidial germination and pathogenicity, although there were many significant defects in phialidiogenesis and macroconidiogenesis in DFghtf1 mutant. This similarity in cellular function led us to further analyze the expression of REN1 orthologous gene FGSG_02471 in DFghtf1 mutant. However, the result showed no significant change in expression (data not shown), suggesting that these two transcription factors are not epistatic. However, it remains to be tested whether these two genes regulate signaling pathways that converge downstream and impact conidiogenesis in F. graminearum.
In Fusarium species, macroconidia have distinct basal foot cell and pointed distal ends. In this study, we discovered that macroconidia in Dhtf1 mutants, those produced through hyphal budding, exhibit significant defect in the foot cell ( Figure 10). When we monitored conidiogenesis in microscopic detail, we recognized that the first initial conidium of the wild-type strain is formed within the apical extension of the phialide or hyphae at the early stage of development. Before the macroconidium is released, the characteristic foot cell at the base of macroconidium is formed.  Macroconidia formation directly from hyphae were also affected, namely in foot cell development. (C) In F. verticillioides DFvhtf1 mutants, prototypical microconidia chain and false heads were not observed but rather a 2-to 5-microconidia complex at the apex of monophialides resulting in a floral petal-like shape. This suggests that FvHtf1 is important for basipetal division typically observed in the wild-type F. verticillioides. (D) In F. oxsporum DFohtf1 mutants, we observed floral petal-shaped microconidia false head similar to F. verticillioides DFvhtf1 mutants suggesting FoHtf1 is important for phialide development and subsequent microconidiogenesis. However, it is not involved in chlamydospore development. c, conidia; cp, conidiophore; ma, macroconidia; tp, terminal phialide; ip, intercalary phialide; fc, foot cell; ch, chain; mp, monophialide; fh, false head; pe, petal; cs, chlamydospore.
Nevertheless, at late stages of development, the expanding conidium ruptures the conidiophore wall and is released by an abscissional splitting of the basal septum. Internal septation of the conidium normally occurs after it is released. In the Dhtf1 mutant, it is unlikely that a conidium is released by abscissional splitting of the basal septum ( Figure 10). Perhaps this is why we found mature conidium with distinct septum born on hyphae in the Dhtf1 mutant. This implies that Htf1 play a key role in regulating foot cell division during macroconidiogenesis.
Production of both microconidia and macrocondia is a common phenomenon in most Fusarium species, and both conidia are formed from phialides in false heads by basipetal division, the developmental mode from the apex toward the base without catenation of cells [28]. In our study, we learned that Htf1 not only regulated macroconidiogenesis, but also for microconidiogenesis. In DFvhtf1 mutant, long chains of microconidia were completely absent in contrast to the wild-type strain ( Figure 10). In F. verticillioides, the cAMP signaling pathway gene FAC1 and two hydrophobin genes HYD1 and HYD2 have been reported to be important for the production of microconidial chains [39]. In addition, Choi and Xu [36] further showed that FAC1 positively regulates microconidia production and the expression of two hydrophobin genes, HYD1 and HYD2 [36]. Notably, all these reported genes had no discernable effect on false-head pattern of microconidia, suggesting that these gene-deletion mutants still can produce microconidia through basipetal division. Intriguingly, our study revealed that the deletion of HTF1 in F. verticillioides and F. oxysporum led to the formation of petal-shaped pattern by sharing the apical branches (Figures 6 and 10) instead of producing typical false heads with characteristic ball-shaped assemblage of microconidia held together apparently by mucilage. As we described earlier, this aberrant microconidiogenesis from monophialide may have interfered with typical microconidia chain development that occurs through basipetal division and ultimately led to significant reduction in microconidia in F. verticilioides and F. oxysporum. Therefore, our results collectively provide evidences that Htf1 regulates microconidial formation from chains and false heads by basipetal division in Fusarium species.

Strains, media and growth condition
All wild-type and mutant strains used in this study are listed in Table S1. In F. graminearum, growth and morphology were evaluated by culturing strains on complete medium (CM: 0.6% yeast extract [w/v], 0.6% casein hydrolysate [w/v], and 1% sucrose [w/v]) at 28uC for 4 days. Formation of perithecia was assayed on wheat kernels medium as described previously [61]. To assay conidiation, an agar block (3 mm in diameter) carrying mycelia was introduced into 50 ml of liquid CMC medium [52]. The suspension was shaken at 180 rpm for 3-15 days, and the concentration of conidia was determined with a hemacytometer. For spore germination assays, fresh macroconidia were suspended in CM for 4 h with gentle agitation [62]. Macroconidia of PH-1 and mutants were observed using an Olympus BX51 Microscope. Infection assays on flowering wheat heads, wheat coleoptiles and corn stalks were conducted as previously described [63][64][65].
The F. oxysporum wild-type and mutant strains were cultured on CM agar to observe morphology and growth. To induce conidiation in F. oxysporum strains, SNA and mung bean agar were used as described above. All tests were repeated three times.

Microscopy and histological visualization
To observe conidiogenesis in F. graminearum, an agar block carrying mycelium was inoculated into CMC as described above and then were imaged at different culture stages with Olympus BX51 Research Microscope. Nuclear visualization in phialides was observed by DAPI staining. Mycelia were collected by centrifugation, washed with PBS buffer (pH 7.2) and then resuspended in PBS containing 0.1% Triton X-100. Cells were then fixed with PBS paraformaldehyde (3.7%, w/v) and stained with 10 mg/ml DAPI (Sigma). The cell nuclei were observed with Olympus BX51 Research Microscope at UV excitation wavelength. For spore germination studies, fresh PH-1 macroconidia were suspended in CM for 4 h with gentle agitation. Cell walls and septa of germinating conidia were visualized by staining with Calcofluor White (10 mg/ml, Sigma).
To directly visualize F. verticillioides and F. oxysporum conidial chains and false heads without immersion in water or buffer, agar squares were removed from actively growing colonies and placed in a slide glass with the fungal colony surface oriented perpendicular to the cover slip. Images were acquired from Olympus BX51 Research Microscope.
For scanning electron microscopy (SEM), blocks of 5-day-old mung bean agar cultures (5 mm 2 ) were fixed in 4% glutaraldehyde at 4uC for 16 h. The samples were then dehydrated in a graded ethanol series and dried in a critical point dryer as described [67]. Samples were coated with a thin gold layer and observed with JSM-6360LV (Jeol Ltd., Tokyo) scanning electron microscope.

Fungal transformation and generation of gene-deletion mutants
The F. graminearum, F. verticillioide and F. oxysporum protoplast preparation and fungal transformation were performed following standard protocols [29,53,63]. Hygromycin-or neomycin-resistant transformants were selected on media supplemented with 250 g/ mL hygromycin B (Roche Applied Science) or 200 g/mL G418 (Invitrogen).
To generate the DFghtf1 mutant, a 1,291-bp fragment upstream from FgHTF1 was amplified with primers FG07097AF and FG07097AR, and this amplicon was subsequently cloned into the HindIII and EcoRI sites upstream of the hph cassette on pCX63 [68]. Then, 1,033-bp fragment downstream from FgHTF1 was amplified with primers FG07097BF and FG07097BR, and cloned into the BamHI and SacI sites downstream of hph cassette, and this plasmid was transformed into protoplasts of the wild-type PH-1 strain as described [63]. Hygromycin-resistant transformants were screened by PCR with primers FG07097UA and H853 and primers FG07097OF and FG07097OR (Table S2). An isolate that tested positive with PCR was further verified by Southern blot analysis performed with the digoxigenin high prime DNA labeling and detection starter Kit I (Roche, Mannheim, Germany).
We generated HTF1 gene-replacement constructs in F. verticillioide and F. oxysporum using the split-marker approach [69,70]. Upstream and downstream fragments were amplified with specific primer pairs that are listed in Table S2. Partial fragments of the hygromycin phosphotransferase (hph) gene were amplified with primers HYG/F, HY/R, YG/F, and HYG/R as described [71]. After transformation, hygromycin-resistant transformants were screened by PCR with designated primers (Table  S2) and further characterized by Southern blot analysis.
For complementation of DFghtf1 and DFvhtf1, FgHTF1 and FvHTF1 (with upstream promoter and downstream terminator) was amplified with primer sets FG07097CF4/FG07097CR4 and FV08072CF/FV08072CR, respectively (Table S2). The resulting constructs were co-transformed into protoplasts of the target mutant along with a vector harboring geneticin-resistance marker (pKNTG). Transformants exhibiting resistance to both geneticin and hygromycin were selected, screened by PCR for the presence of the complementation construct, and further validated by Southern blot analyses. Inter-species complemented strategies were similar to generation of DFghtf1-Com strain. FgHTF1, FvHTF1 or FoHTF1 gene was amplified with a set of primers (Table S2), and subsequently co-transformed into the target fungal protoplasts with pKNTG vector. The selected isolates were further analyzed by PCR, using primers (Table S2) to determine the presence of FgHTF1, FvHTF1 or FoHTF1 gene.

Construction of FgHTF1-GFP vector and complementation
The FgHTF1-GFP fusion vector, pGM-FgHTF1-GFP, was constructed by amplification of 2,886-bp fragment including 1,459-bp FgHTF1 coding sequence and a 1,427-bp promoter region using primers FG07097CF3-GFP and FG07097CR3-GFP (Table S2). The 2,886-bp PCR product was then cloned into pGEM-T easy vector to generate pGM-FgHTF1. The 1.5-kb GFP allele [72] carrying the A. nidulans trpC terminator was amplified using primers HindIII-GFPF and HindIII-GFPR (Table S2), then cloned into pGEM-T easy vector. It was subsequently digested with HindIII to release the GFP allele with HindIII sticky ends, which was inserted into HindIII site of pGM-FgHTF1 to create pGM-FgHTF1-GFP. We verified the orientation of GFP insertion and in-frame fusion by sequencing the pGM-FgHTF1-GFP vector. To generate FgHTF1-GFP strain, pGM-FgHTF1-GFP vector and pKNTG vector were cotransformed into DFghtf1 mutant. Transformants carrying a single insertion were selected and phenotypic restoration in DFghtf1 mutants was sought. GFP fluorescence was observed using a Leica TCS SP5 inverted confocal laser scanning microscope (Leica, Germany)

Quantitative RT-PCR
Wild-type conidia were harvested at growth stages (24 h, 36 h, 48 h, 60 h and 72 h incubated on CMC medium). RNA was isolated with TRIzol reagent (Invitrogen) and purified with the DNA-free kit (Ambion). First-strand cDNA was synthesized with the M-MLV reverse transcriptase (Invitrogen), and qRT-PCR was performed with the ABI 7500 sequence detection system (Applied Biosystem) using QuantiTect SYBRgreen PCR Master Mix (Qiagen). Primers used to amplify selected genes in qRT-PCR reactions are listed in supplemental Table S2. TUB2 (FGSG_06610.3) was used as the endogenous reference gene. The relative quantification of each transcript was calculated by the 2 2DDC T method [73]. All qRT-PCR reactions were conducted in triplicates for each sample and the experiment was repeated three times.    Figure S2A). PH-1, wild-type strain; DFghtf1-Com, complementation strain; DFghtf1-7 and DFghtf1-8, null mutants. (C) Total RNA samples isolated from mycelia of PH-1, DFghtf1-8 and DFghtf1-Com were subjected to RT-PCR using FgHtf1 gene-specific primers FG07097OF and FG07097OR (Table S3). As predicted, the RT-PCR amplicon (1,178 bp) was observed in PH-1 and DFghtf1-Com, but was absent the deletion mutant DFghtf1-8.