A Novel C2H2 Transcription Factor that Regulates gliA Expression Interdependently with GliZ in Aspergillus fumigatus

Secondary metabolites are produced by numerous organisms and can either be beneficial, benign, or harmful to humans. Genes involved in the synthesis and transport of these secondary metabolites are frequently found in gene clusters, which are often coordinately regulated, being almost exclusively dependent on transcription factors that are located within the clusters themselves. Gliotoxin, which is produced by a variety of Aspergillus species, Trichoderma species, and Penicillium species, exhibits immunosuppressive properties and has therefore been the subject of research for many laboratories. There have been a few proteins shown to regulate the gliotoxin cluster, most notably GliZ, a Zn2Cys6 binuclear finger transcription factor that lies within the cluster, and LaeA, a putative methyltransferase that globally regulates secondary metabolism clusters within numerous fungal species. Using a high-copy inducer screen in A. fumigatus, our lab has identified a novel C2H2 transcription factor, which plays an important role in regulating the gliotoxin biosynthetic cluster. This transcription factor, named GipA, induces gliotoxin production when present in extra copies. Furthermore, loss of gipA reduces gliotoxin production significantly. Through protein binding microarray and mutagenesis, we have identified a DNA binding site recognized by GipA that is in extremely close proximity to a potential GliZ DNA binding site in the 5′ untranslated region of gliA, which encodes an efflux pump within the gliotoxin cluster. Not surprisingly, GliZ and GipA appear to work in an interdependent fashion to positively control gliA expression.


Introduction
Secondary metabolites are small, low-molecular mass molecules made by numerous organisms that are not essential for normal growth, but can play important roles in defense or signaling [1][2][3][4]. They can be benign in nature, such as pigments or molecules used in interspecies communication, but they can also be malignant, exhibiting antimicrobial or toxic activities to eliminate competing organisms [5,6]. Some of these compounds have been exploited by scientists because of their potential benefit to humans. For example, penicillin, produced by P. chrysogenum, is used as an antibiotic and lovastatin, produced by A. terreus, reduces cholesterol [7]. While some of these secondary metabolites benefit humans, others cause harm. Aflatoxin, produced by A. flavus, is carcinogenic, and gliotoxin, produced by A. fumigatus, exhibits immunosuppressive properties [4,[8][9][10]. Genes involved in secondary metabolite production are frequently found as clusters, which are often located in subtelomeric positions [2][3][4][5]11]. Furthermore, these clusters are often coordinately regulated, being almost completely dependent on induction by transcription factors located within the clusters themselves [3][4][5]12,13].
Zinc binuclear (Zn(II) 2 Cys 6 ) transcription factors are uniquely found in fungi and represent the most common type of regulators located within these clusters [3,4,14,15]. AflR, a Zn 2 Cys 6 transcription factor, is located within the aflatoxin/sterigmatocystin gene cluster and is required for production of both metabolites [3,4,10]. Zn 2 Cys 6 transcription factors, like aflR, generally recognize and bind as homodimers to palindromic sequence motifs, such as CGG(N x )CCG [14][15][16][17]. Interestingly, although the palindromic sequences of these binding motifs can be similar or identical for multiple Zn 2 Cys 6 transcription factors, the length and base composition of the linker sequence is highly variable. Therefore, this linker sequence greatly contributes to the specificity of binding for each individual transcription factor [15,17,18]. Although gene clusters are often coordinately regulated by the cluster-specific transcription factor, some members can be independently regulated. For example, gliT encodes an oxidoreductase of the gliotoxin biosynthetic cluster, which is required for self-protection against gliotoxin. Even though gliT expression is decreased when the Zn 2 Cys 6 transcription factor, gliZ, is deleted, exogenous gliotoxin induces the expression of gliT, even in a DgliZ background [4,19].
Aside from pathway-specific transcription factors that reside within the cluster, there are numerous other regulatory elements that affect the expression of secondary metabolite clusters. Nutritional and environmental factors, as well as developmental processes, have been shown to affect secondary metabolite production in multiple fungal species [3][4][5]. For instance, penicillin production in A. nidulans is repressed in the presence of glucose, a phenomenon termed carbon catabolite repression [8,20]. Secondary metabolite repression also occurs in response to nitrogen source, which involves AreA, the global positive regulator of nitrogen metabolite repression. Indeed, in A. nidulans, sterigmatocystin production is repressed in the presence of ammonium and induced when nitrate is the sole nitrogen source [2,6,21]. This type of regulation is not seen in all fungi, as carbon and nitrogen source have the opposite effect in A. parasiticus, where growth in the presence of glucose or ammonium results in higher levels of aflatoxin production [2,6,21].
A. fumigatus, the leading cause of mold infections worldwide, is an opportunistic pathogen that causes severe problems in immunecompromised populations [9,22]. These populations include: AIDS patients, cancer patients receiving chemotherapy, solid organ transplant/skin graft patients and victims of chronic granulomatous disease [12,[23][24][25]. One of the most studied secondary metabolites produced by A. fumigatus is gliotoxin, which is also produced by several other Aspergillus species, Trichoderma species, and Penicillium species [13,[26][27][28]. Gliotoxin is a member of the epidithiodioxopiperazine (ETP) class of toxins, which are characterized by a disulfide bridge across a piperazine ring [13,[23][24][25][26][27][29][30][31][32][33]. The oxidized form of gliotoxin travels into host immune cells where it is able to affect cellular functions essential to the immune response. These include impediment of phagocytosis and NF-kB activation, as well as induction of apoptosis [23,25,26,29,32,33]. As with other secondary metabolites, most of the genes responsible for the production and transport of gliotoxin exist within a gene cluster. The gliotoxin biosynthesis cluster was first identified based on its homology to the sirodesmin PL biosynthesis gene cluster in the ascomycete Leptosphaeria maculans [13,27,34,35].
Within this cluster lies a Zn 2 Cys 6 binuclear finger transcription factor, GliZ, thought to be responsible for general gliotoxin induction and regulation. Indeed, over-expression of gliZ leads to an increase in gliotoxin production and deletion of gliZ results in a loss in gliotoxin production [12,26,28]. A DNA binding site has been proposed for GliZ (TCGGN 3 CCGA), but has not been experimentally proven. This site is present within the promoter region of every gene within the gliotoxin cluster, except gliZ and gliA, which encodes an efflux pump within the cluster [14]. Gliotoxin itself positively regulates expression of the genes within the gliotoxin cluster, as deletion of gliP, the non-ribosomal peptide synthetase (NRPS) required for the first step in the synthesis of gliotoxin, virtually eliminates expression of the other genes in the cluster. This loss in gene expression can be reversed by the addition of exogenous gliotoxin to culture medium [19,26,30]. Interestingly, gliT, encoding an oxidoreductase required for resistance of the fungus to gliotoxin, is induced by exogenous gliotoxin even in the absence of gliZ [19]. This demonstrates regulation of genes within the cluster that are independent of the coordinate regulation by GliZ. LaeA, a global regulator of secondary metabolism, has also been shown to regulate the gliotoxin cluster, as gliotoxin is among the secondary metabolites that are lost with deletion of laeA [36][37][38]. Furthermore, loss of vel1 in T. virens (homologous to VeA in Aspergilli) results in a loss in gliotoxin production [39]. This is not surprising, since VeA, VelB, and LaeA form a heterotrimeric complex that regulates secondary metabolism in several fungal species [2,25,36,37,40]. RsmA, a bZIP transcription factor, positively regulates the gliotoxin cluster through LaeA and GliZ, as loss of either protein abolishes the inducing effects of RsmA over-expression [41]. Map kinase signaling is another element that is important for gliotoxin production, as a strain lacking mpkA, the map kinase in the cell wall integrity pathway, is significantly reduced in gliotoxin production [42]. Nutrient starvation, murine infection, and exposure of germlings to neutrophils have also been shown to up-regulate the gliotoxin biosynthesis cluster through microarray analysis [11,26].
At this time, information is still lacking as to what proteins and transcriptional regulators are responsible for differential expression of the gliotoxin cluster. We have identified a novel C 2 H 2 transcription factor, gipA, which plays a role in regulating the gliotoxin biosynthesis gene cluster. This protein was identified using a high-copy inducer screen and has not been discussed in detail before now. High-copy expression of gipA results in an increase in transcription of multiple genes within the gliotoxin cluster. Conversely, loss of gipA negatively affects expression of multiple genes within the gliotoxin cluster. In addition, this C 2 H 2 transcription factor specifically regulates gliA expression through a binding site we have identified. Here, we propose a model for gliA expression that involves both GliZ and GipA. Our aim was to identify a novel protein that is involved in gliotoxin production, which we achieved through the discovery of GipA.

High-Copy Inducer Screen Identifies a Novel Protein That Induces Expression of a gliA P -lacZ Cassette
To identify genes that regulate the gliotoxin biosynthesis cluster, we performed a high-copy inducer screen. Our goal was to identify genes that, when present in extra copies, induce the gliotoxin

Author Summary
Fungal infections cause severe problems for immunecompromised individuals. Current antifungal treatment is problematic, as some therapies are toxic to humans and others are not highly effective. These fungal infections also burden hospitals, as costs to treat and prevent such disease runs high. Aspergillus fumigatus is the most common cause of fungal infections worldwide. A. fumigatus produces a variety of toxins that aid the fungus in survival both in the environment and within host systems. Genes involved in producing such toxins are often found in clusters within the genome, being almost exclusively dependent on transcription factors located within the clusters. Gliotoxin, one such toxin, is known to negatively affect immune cell function. Although gliotoxin has been studied extensively, information is still lacking with regards to regulation of gliotoxin biosynthesis. Our lab has discovered a novel C 2 H 2 transcription factor, GipA, which plays an important role in gliotoxin production. Not only does GipA enhance gliotoxin production when overexpressed, but loss of GipA causes a significant reduction in gliotoxin production. As this gene is not located within the gliotoxin cluster, understanding its mode of action and upstream partners could shed light on toxin production in general and lead to better, more effective antifungal therapies.
biosynthesis cluster in repressing conditions. We used a LacZ reporter system, under the control of the gliA promoter (identified as GL in strain names). GliA encodes an efflux pump within the gliotoxin cluster that is thought to be involved in the transport of gliotoxin out of the fungal cell [13,27]. Although GliA has not been shown to be essential to gliotoxin transport in A. fumigatus, expression of gliA in a mutant strain of L. maculans provided protection against gliotoxin, but not sirodesmin [27]. We chose gliA for several reasons. First, it has been shown that expression of gliA peaks when the amount of gliotoxin in surrounding medium is maximal [13,27]. Second, experiments in our lab revealed that gliA is induced within 30 min of A. fumigatus germlings being exposed to human neutrophils, while nutritional induction of the gliotoxin cluster takes 24-48 hrs (data not shown). These data led us to conclude that gliA expression would be a good indicator of gliotoxin cluster expression, as well as gliotoxin production.
The first round of the high-copy inducer screen involved transforming an AMA1-NotI A. fumigatus genomic library [43] into Af293.1-GL (Fig. 1a). AMA1 stands for autonomous maintenance in Aspergillus and was first discovered and isolated from A. nidulans [44,45]. Due to the presence of this element, the A. fumigatus library plasmids replicate autonomously upon transformation. This feature allows for easy recovery of the plasmids from the fungal genome by transformation of the genomic DNA into bacteria [46,47]. Furthermore, this screen is a high-copy inducer screen, because there are on average 10 to 30 copies of any given AMA1containing plasmid per genome [44,45]. We grew transformants on gliotoxin-repressing medium with X-gal and screened for colonies that produced a blue pigment. This indicated that the AMA1-NotI plasmid within the genome was inducing gliotoxin cluster expression in repressing conditions. For the second round of the high-copy inducer screen, we transformed individual plasmids, isolated in the first round of the genetic screen, into Af293.1-GL (Fig. 1b). We transformed individual vectors to be sure that the effect we observed from the first round was the result of only one plasmid and not multiple plasmids. To measure LacZ levels in a more quantitative manner, we isolated total protein from transformants and measured b-galactosidase activity. We transformed a pDONR AMA empty vector (AMA.GL) and pDONR AMA-gliZ (AMA-gliZ.GL) as a negative control and positive control, respectively. The third round of the high-copy inducer screen entailed isolating the individual genes from each AMA1-NotI library plasmid and transforming them into Af293.1-GL (Fig. 1c). For this round, we only measured b-galactosidase activity.
The gene that induced lacZ the most was a C 2 H 2 transcription factor (AFUA_6G01910), which we have designated gipA for gliotoxin inducing protein. High-copy expression of gipA in the Af293.1-GL background (AMA-gipA.GL) induces a 400-fold increase in b-galactosidase activity, compared to the empty vector control (AMA.GL) (Fig. 2). The level of b-galactosidase activity in our positive control, AMA-gliZ.GL, was almost 30-fold higher than the empty vector control, AMA.GL. These data indicate that GipA positively regulates gliA expression, similar to GliZ.
gipA Is a C 2 H 2 Transcription Factor with an Unusually Long 59 UTR The predicted sequence of gipA contains one intron and two C 2 H 2 regions at the 39 end (Fig. 3a). The two C 2 H 2 regions are canonical, the first being X 2 -C-X 2 -C-X 12 -H-X 3 -H and the second being X 2 -C-X 2 -C-X 12 -H-X 5 -C, which is a natural variant [48,49]. The 59 untranslated region (UTR) of gipA appears to be unusually long (877 bp) and contains three mORFs, which indicates that gipA may be under post-transcriptional control due to regulation of translational efficiency and mRNA stability. Northern analysis was consistent with cDNA predictions (Fig. 3b). Sequence similarity searches indicated that proteins homologous to GipA are present in only a limited number of species, the closest being a C 2 H 2 transcription factor in N. fischeri, a close relative to A. fumigatus (Fig. 3c). The rest of the homologous proteins were from other  Aspergillus species and a few Penicillium species, which suggests that GipA is not highly conserved at the primary sequence. For proteins, primary sequences evolve and change much more rapidly than do the tertiary structures. There have been numerous examples where two proteins have low sequence similarity, but once crystallized, exhibit almost identical folding patterns and subsequently share similar functions [50]. For example, Gcn4 of Saccharomyces cerevisiae and CpcA of A. niger share a 35% identity, yet they both function in amino acid biosynthesis. Furthermore, CpcA is able to complement a Dgcn4 mutant in S. cerevisiae [51]. When solely comparing the putative DNA binding domain of cpcA, the identity between CpcA and Gcn4 increases to 70% [51]. Therefore, the lack of homologous counterparts to GipA in other organisms does not necessarily mean that there is not a protein present in other fungi that functions similarly to GipA.

High-Copy Expression of gipA Induces Gliotoxin Production
In a previous section, we showed that extra copies of gipA induce expression of lacZ, under the control of the gliA promoter, which suggests that GipA is inducing gliA. Since the gliotoxin biosynthesis cluster is coordinately regulated, extra copies of gipA should also induce other genes within the cluster. As expected, AMA-gipA.GL had a higher amount of gliA mRNA, than AMA.GL. Transcript levels of gliA in AMA-gipA.GL were 4.5-fold higher, compared to AMA.GL (Fig. 4a). The mRNA levels of the other gliotoxinspecific genes tested were also significantly higher in AMA-gipA.GL, compared to AMA.GL, as gliZ was induced 12-fold, gliP was induced 5-fold, and gliT was induced 2-fold (Fig. 4a).
Gliotoxin production reflected what was seen with mRNA levels, as AMA-gipA.GL produced gliotoxin at a higher amount than AMA.GL (7-fold higher) (Fig. 4b). AMA-gliZ.GL was the positive control and showed the same pattern as AMA-gipA.GL, with respect to induction of the gliotoxin biosynthesis cluster.

High-Copy Expression of gipA Possibly Affects Multiple Secondary Metabolite Clusters
To expand our view of GipA regulation, we performed a microarray analysis of AMA-gipA.GL vs. AMA.GL, grown in repressing conditions for 24 hrs. Of the 9,436 total genes analyzed, 443 genes were up-regulated .2-fold and 75 genes were downregulated .2-fold in the AMA-gipA.GL strain (Fig. S1). There were several genes common to secondary metabolism clusters (e.g. transporters, oxidoreductases, methyltransferases, nonribosomal peptide synthetases and polyketide synthases) up-regulated (Table 1). Approximately 31 secondary metabolism clusters have been proposed using genomic mapping and microarray techniques [52,53]. Of these 31 potential secondary metabolism clusters, 18 contained at least one gene that was up-regulated .2-fold in AMA-gipA.GL, compared to AMA.GL (Table 2 & Fig. S2). Based  . High-copy expression of gipA induces gliotoxin production. All cultures were grown in repressing conditions. Total RNA was isolated and quantified by dot blot analysis in triplicate. Gliotoxin levels were quantified by RP-HPLC using a standard curve and calculated as mg gliotoxin/g dry mycelial mass. All data sets are normalized to AMA.GL. (a) mRNA transcript levels of several gliotoxin cluster genes after 48 hrs of growth relative to AMA.GL. The results of one representative experiment of three independent experiments are shown as mean 6 SD. (b) Gliotoxin levels in growth medium relative to AMA.GL. The data presented is an average of three biological replicates, shown as mean 6 SD. The asterisk (*) indicates a statistically significant difference (p-value,0.05), compared to AMA.GL, calculated by one-way ANOVA and Tukey comparison test. doi:10.1371/journal.pgen.1004336.g004 on microarray data done previously [52], loss of laeA, a global regulator of secondary metabolism, affected 13 of 22 identified secondary metabolite clusters. This suggests that GipA potentially induces numerous other secondary metabolism gene clusters in A. fumigatus, similar to LaeA. One caveat of this microarray, however, is that we measured the effects of high-copy expression of gipA, while the microarray done with LaeA compared a deletion strain to a wild-type and complemented strain. Therefore, based on these results, we cannot conclude that loss of gipA will have a significant effect on multiple secondary metabolite clusters, as is observed with loss of laeA.

Deletion of gipA Significantly Reduces Gliotoxin Production
Since GipA can induce gliotoxin production, we sought to discover if loss of gipA has any effect on the gliotoxin cluster. We replaced the coding region of gipA with pyrG and designated this strain as DgipA. We created a complemented strain, gipA(R), using a hygromycin resistance cassette (hygroR) as the selective marker. We also created a gliZ deletion strain as a control, since previous studies have shown that loss of gliZ results in a significant decrease in mRNA levels of gliotoxin-specific genes [12]. Loss of gipA caused a significant decrease in mRNA levels of gliA, gliZ, gliP, and gliT in non-repressing conditions, as most genes exhibited close to a 50% reduction (Fig. 5a). The gliotoxin biosynthesis cluster is not completely dependent on gipA, as there was still mRNA being made for the genes we tested. The gipA deletion mutant also produced significantly less gliotoxin than the 1160G control strain (50% reduction) (Fig. 5b). Gliotoxin-specific gene expression and gliotoxin production of gipA(R) were restored beyond wild-type levels, which demonstrates that the effect we observed with the gipA deletion was due to the absence of gipA. As expected, loss of gliZ caused almost a complete loss in gene expression for gliA, gliP, and gliT and abolished gliotoxin production. A gliZ/gipA double deletion mutant revealed a pattern of gene expression and gliotoxin production similar to the gliZ single deletion mutant (data not shown).

Putative GipA Binding Sites Are Present Throughout the Gliotoxin Cluster
Since GipA is a C 2 H 2 transcription factor, it is likely that GipA is directly binding to DNA. We sought to identify a consensus sequence and to discover if this sequence was present within the gliotoxin biosynthesis cluster. A protein-binding microarray analysis identified a consensus DNA binding sequence for GipA (59-TNNVMGCCNC-39) (Fig. 6a). This putative sequence is 10 nucleotides, which coincides with one complete turn of the DNA double helix. Interestingly, there are 7 high-content positions (59-VMGCCNC-39), which would correlate with two C 2 H 2 zinc finger DNA binding domains [54], and then a ''T'' residue two base pairs upstream of the 7 high content core positions. The protein-binding microarray verified direct DNA binding of GipA, as the purified DNA-binding domain, and not whole cell extract, was analyzed in the microarray. We analyzed the genomic sequence of the gliotoxin biosynthesis cluster to locate any potential GipA binding sites. Indeed, we found variations of this consensus sequence scattered throughout the gliotoxin biosynthesis cluster. In fact, we identified a possible GipA binding site within the intergenic region of gliA (59-TTGCCGCCAC-39 315 bp upstream of the start site), as well as all other gliotoxin-specific genes, except gliM (Fig. 6b). Not all putative GipA DNA binding sites throughout the intergenic regions in the gliotoxin biosynthetic cluster contain the ''T'' residue in the 1 position, which means that either these sites are possibly only weakly recognized by GipA, if at all, or the ''T'' residue is not part of the actual GipA DNA recognition element.

Mutational Analysis of the gliA Promoter Verifies the Presence of an Active GipA Binding Site
We created two mutant backgrounds: BSM1 (TTGCCGC-CACCTGCCGCCAC) and BSM2 (TTGCCGCCACRTTG-GGTGAGC), by mutating the putative GipA DNA binding site on the gliA-lacZ construct we used in the original screen. We measured lacZ expression with b-galactosidase assays in the presence of different high-copy plasmids (strains used in this study are listed in Table S2). When normalized to AMA.GL, AMA-gipA.GL displayed increased LacZ levels (13-fold), as was to be expected from previous results (Fig. 6c). High-copy expression of gipA in the BSM1 background also induced lacZ significantly (53fold), relative to AMA.BSM1 (Fig. 6c). The magnitude of induction was enhanced with the BSM1 mutation when compared to the wild-type binding site (13-fold vs. 53-fold, respectively). This enhanced expression was not due to higher overall lacZ levels, as shown when comparing empty vector controls in all three backgrounds (Fig. 6d). AMA-gipA.BSM2 only weakly induced lacZ (4-fold), compared to AMA.BSM2 (Fig. 6c). This suggests that mutation of a core sequence in the putative GipA DNA binding site significantly reduces the ability of GipA to induce lacZ. Mutation of the 59 ''T'' residue did not reduce GipA-specific gliA induction, but rather increased the response of gliA to GipA highcopy expression.
Interestingly, expression of lacZ in the pDONR AMA-gliZ strains followed a similar pattern to that of pDONR AMA-gipA strains (Fig. 6c). Therefore, lacZ was induced in AMA-gliZ.GL (19fold) and AMA-gliZ.BSM1 (93-fold), compared to AMA.GL and AMA.BSM1, respectively. The level of induction was enhanced by the BSM1 binding site, compared to the wild-type binding site (19fold vs. 93-fold, respectively). Furthermore, lacZ was only weakly induced in AMA-gliZ.BSM2, relative to AMA.BSM2 (4-fold). Therefore, mutation of the putative GipA DNA binding site is affecting the ability of GliZ to induce gliA.

GipA Requires GliZ for Induction of Both gliA and gliP
To determine if GipA is dependent on GliZ for expression of gliA, we created a series of mutants and measured mRNA levels of gliA and gliP. High-copy expression plasmids used in previous sections were transformed into a wild-type background (AMA.G, AMA-gliZ.G, and AMA-gipA.G) and a gliZ deletion background (AMA.Z, AMA-gliZ.Z, and AMA-gipA.Z) (strains used in this study are listed in Table S2). All strains were grown in non-repressing conditions to induce expression of the gliotoxin biosynthesis cluster. In the pyrG+ background, AMA-gliZ.G and AMA-gipA.G had increased mRNA levels for gliP, while only AMA-gipA.G had increased mRNA levels for gliA, compared to AMA.G (Fig. 7). These changes were not significant because growth in nonrepressing conditions already induced these genes to high levels, so having extra copies of GliZ or GipA did not greatly contribute to gene expression. In AMA.Z, mRNA for both gliA and gliP was almost completely undetectable, as to be expected from previous experiments. High-copy expression of gliZ brought transcript levels of both genes back to AMA.G levels (Fig. 7). AMA-gipA.Z displayed a reduction in mRNA levels similar to AMA.Z. For gliA, the level of mRNA present in AMA-gipA.Z was slightly higher (close to 5-fold) than what was observed for AMA.Z. However, the level of gliP mRNA did not exceed that of the AMA.Z empty vector control. Therefore, GipA was not able to induce gliA or gliP in the absence of GliZ. This suggests that GliZ is required for GipA to induce both gliP and gliA.
GliZ Requires GipA for Induction of gliA, but not gliP To determine if GliZ is dependent on GipA for expression of gliA, we created a series of mutants and measured mRNA levels of gliA and gliP. High-copy expression plasmids used in previous sections were transformed into a wild-type background (AMA.G, AMA-gliZ.G, and AMA-gipA.G) and a gipA deletion background (AMA.A, AMA-gliZ.A, and AMA-gipA.A) (strains used in this study are listed in Table S2). All strains were grown in non-repressing conditions to induce expression of the gliotoxin biosynthesis cluster. As expected from previous experiments, mRNA levels in AMA.A of both gliA and gliP were reduced significantly, 50% and 80%, respectively. Levels of gliP mRNA in AMA-gliZ.A were comparable to those of AMA.G; however mRNA levels of gliA were not significantly higher than background levels (AMA.A) (Fig. 7). This indicates that GliZ is not dependent on GipA for induction of gliP, however induction of gliA by GliZ does appear to be dependent on GipA. The first 18 secondary metabolism clusters are the ones we predict to be positively regulated by GipA, while the last 13 did not have any genes up-regulated .2-fold from the microarray data. *Cluster ranges for unknown products are defined in [52] or were predicted using SMURF at http://jcvi.org/smurf/index.php. 1 Although several genes in this cluster have been shown to be involved in production of fumiquinazolines, PesL (Afu6g12050) was recently revealed to be necessary for fumigaclavine C production. doi:10.1371/journal.pgen.1004336.t002

Discussion
Although recent studies have revealed gliotoxin intermediates, which have led to a better understanding of the biosynthesis of gliotoxin, information on regulation of the genes involved in the biosynthesis pathway is lacking [28]. With the use of a high-copy inducer screen, our lab has uncovered a novel protein, GipA, which appears to be involved in the regulation of the gliotoxin cluster. GipA is a C 2 H 2 transcription factor, which harbors an unusually long 59 UTR. Furthermore, there are three mORFs within the 59 UTR, which suggests that gipA may be under posttranscriptional control. There are two canonical C 2 H 2 zinc finger DNA binding regions at the 39 end of gipA, the first being X 2 -C-X 2 -C-X 12 -H-X 3 -H and the second being X 2 -C-X 2 -C-X 12 -H-X 5 -C, which is a natural variant [48,49]. Two C 2 H 2 zinc finger binding domains is consistent with a 6-8 base pair DNA recognition element [54]. The putative DNA binding site that we identified for GipA contains 7 high content positions , in addition to a 59 ''T'' residue two base pairs upstream of the core residues. Based on mutagenesis of the putative GipA DNA binding site in the gliA promoter, the 7 high content positions are likely part of an active GipA DNA binding site. While the ''T'' residue is not necessary for GipA-specific gliA induction, mutation of the ''T'' residue does enhance GipAspecific gliA expression. We have devised a possible model for the specific regulation of gliA that involves GliZ and GipA (Fig. 8). We propose that GliZ and GipA work together at the same binding site or in close proximity to induce gliA. In this model, GipA and GliZ are dependent on each other for inducing gene expression of gliA.
GipA appears to play an important role in gliotoxin biosynthesis, as high-copy expression of gipA causes increased gliotoxin production and loss of gipA causes a significant reduction in gliotoxin levels. Furthermore, GipA is possibly regulating other secondary metabolite clusters within A. fumigatus, similarly to LaeA, a global regulator of secondary metabolism [52]. In contrast to LaeA, though, GipA is a C 2 H 2 transcription factor, which likely directly binds to DNA recognition sites. Other C 2 H 2 transcription factors, such as PacC, NsdC, NsdD and CreA, have also been identified as regulators of secondary metabolism. PacC regulates a wide range of genes in response to ambient pH, including those involved in secondary metabolism. Penicillin production in A. nidulans is induced in alkaline conditions, while sterigmatocystin production is repressed [4,55,56]. NsdC and NsdD are both involved in sexual development in A. nidulans. However, recent work by Jeffrey Cary et al. [57] supported additional roles for NsdC and NsdD in asexual development and secondary metabolism in A. flavus. Although AflR expression was not affected, loss of either nsdD or nsdC resulted in decreased expression of other genes in the aflatoxin biosynthesis gene cluster [57]. Finally, CreA is a global regulator of carbon catabolite repression in fungi. In addition to carbon metabolism, CreA also affects secondary metabolism, as cephalosporin production is reduced in Acremonium chrysogenum in response to high glucose [56,58]. It is possible that GipA serves to enhance expression of the gliotoxin cluster in certain environmental conditions. Both high-copy expression and loss of gipA affect gliZ expression, so the effects we see for the entire cluster could be the direct consequence of GipA binding to each promoter region, or an indirect consequence of GipA partially regulating gliZ, which in turn regulates other genes within the cluster. Loss of both gliZ and gipA does not give conclusive evidence for either possibility, as loss of gliZ alone completely abolishes gliotoxin production. None of the mutants tested displayed abnormal growth rates or attenuated virulence in a Drosophila melanogaster model system, indicating that high-copy expression of gipA or loss of gipA does not affect overall fitness of A. fumigatus (data not shown).
Our data show a dependency between GipA and GliZ with respect to gliA expression. Firstly, there is a putative GliZ DNA binding site embedded within a GipA DNA binding site in the gliA 59 UTR. Although a GliZ binding site has not yet been experimentally determined, one has been predicted (TCGGN 3 CCGA). This sequence is present in the intergenic region of every gene within the gliotoxin cluster, except gliZ and gliA [14]. Recognition of these sequences by Zn 2 Cys 6 binuclear finger transcription factors is often very specific and even a slight change to the length or base composition of the linker sequence can result in reduced binding in vivo [15,17,18]. Within the gliA promoter region, a sequence overlapping the GipA binding site has the CGG-CCG inverted repeats common to Zn 2 Cys 6 binuclear finger transcription factor binding sites, but the linker sequence is longer (8 bps) than that of the predicted GliZ DNA binding site (3 bps). Mutation of the CCG repeat almost completely abolishes gliZ-mediated induction of gliA. This suggests that either the prediction for the GliZ DNA binding site Figure 5. Loss of gipA negatively affects gliotoxin production. All cultures were grown for 48 hrs in non-repressing conditions. Total RNA was isolated and quantified by dot blot analysis in triplicate. Gliotoxin levels were quantified by RP-HPLC using a standard curve and calculated as mg gliotoxin/g dry mycelial mass. All data sets are normalized to 1160G. (a) mRNA transcript levels of several gliotoxin cluster genes relative to 1160G. The results of one representative experiment of three independent experiments are shown as mean 6 SD. (b) Gliotoxin levels in growth medium relative to 1160G. The data presented is an average of three biological replicates, shown as mean 6 SD. The asterisk (*) indicates a statistically significant difference (pvalue,0.05), compared to 1160G, calculated by one-way ANOVA and Tukey comparison test. doi:10.1371/journal.pgen.1004336.g005 is incorrect and is actually CGGN 8 CCG or that the GliZ DNA binding site within the gliA UTR is unique. Due to the position of the putative GliZ DNA binding site, mutation of this CCG also changes the core sequence of the putative GipA DNA binding site. Accordingly, gipA-mediated induction of gliA is almost completely abolished.
Dependent dual regulation of two transcription factors has been uncovered in other organisms. For example, in A. nidulans, FlbB and FlbD both bind in close proximity to the brlA promoter to regulate asexual development through brlA activation [59,60]. Furthermore, FlbB does not bind to the brlA promoter in the absence of FlbD, indicating that these two transcription factors are dependent on each other for DNA binding and activation of brlA [60]. Although the mutational analysis of the gliA promoter raises the possibility of a GliZ-GipA dependency, it does not give conclusive evidence. One could argue that disruption of the GliZ DNA binding site causes loss of gliA induction that indirectly abolishes gipA-mediated induction of gliA. However, our hypothesis of a dependency between GipA and GliZ with respect to gliA expression is further supported by the fact that GliZ cannot induce gliA in the absence of GipA. As expected, loss of gliZ severely decreases mRNA levels of gliA, which cannot be rescued by high-copy expression of gipA. Interestingly, loss of gipA also significantly decreases mRNA levels of gliA, which cannot be rescued by high-copy expression of gliZ. Clearly, these two proteins work together to induce gliA beyond basal levels in non-repressing conditions. This pattern of dependency is unique to gliA expression, as gliP is induced by high-copy expression of gliZ, even in the absence of gipA.
Together, these data support a model in which GliZ and GipA are working together to regulate gliA (Fig. 8). There appears to be a dependency with regards to gliA expression that we demonstrated in our experiments. This model does not apply to every gene within the gliotoxin cluster, as no other genes have a GipA binding site embedded within a possible GliZ binding site, except gliZ, although these sequences are located farther upstream of the gliZ start site (Fig. 9). Possibly, GipA serves to aid GliZ in binding to the gliA promoter region, as this binding site is different from the others present in the gliotoxin cluster. Although there are possible GipA binding sites in all gliotoxin gene promoters, except gliM, we cannot say with certainty whether GipA is directly binding to these other promoter regions or if GipA is simply binding in the gliA promoter in conjunction with GliZ. This adds to mounting evidence that genes within a gene cluster are typically coordinately regulated, but can also be individually expressed in response to certain stimuli. It is possible that gliT and gliA are both independently regulated to protect the fungus from exogenous gliotoxin, although gliA was not induced in a DgliZ mutant in the presence of exogenous gliotoxin as gliT was [19]. Another possibility is that gliA is independently regulated to aid in the transport and/or expression of other secondary metabolism clusters in A. fumigatus. There has been evidence in other fungal species that crosstalk between these gene clusters exists [61].

Materials and Methods
All primers used in this study are listed in Table S1.

Strains and Growth Conditions
All strains used in this study and genotypes are listed in Table  S2. We maintained Af293.1, Af293.1-GL, 1160, Af293.1-BSM1, and Af293.1-BSM2 on YAG medium supplemented with uridine and uracil (0.5% yeast extract, 1% glucose, trace elements and vitamin mix as modified [62], 10 mM MgCl 2, 1.5% agar, 5 mM uridine, and 10 mM uracil,). We grew AMA.G, AMA.Z, AMA.A, AMA-gliZ.G, AMA-gliZ.Z, AMA-gliZ.A, AMA-gipA.G, AMA-gipA.Z, and AMA-gipA.A on YAG medium with 400 mg/ml hygromycin. We maintained all other strains on YAG medium. Unless otherwise noted, we grew all strains at 37uC for 48 hrs. For phenotypic growth assays of high-copy and deletion strains, we inoculated approximately 1000 spores of each strain onto MMVAT ( [62], and 1.25% agar), MMVAT with 10 mg/ml gliotoxin, and YAG. MMVAT plates were incubated for 72 hrs. We repeated plate growth assays twice for a total of three independent tests. We measured radial growth of each colony and scanned plates on the final day of growth.

High-Copy Inducer Screen
We used fusion PCR (f-PCR) to create a gliA P -lacZ-gliA T construct. For the first reaction, we created three cassettes: gliA 59 P , lacZ, and gliA 39 T , using primer pairs GliA F1 and GliA 59 R, lacZ F and lacZ R, and GliA 39 F and GliA R, respectively. We    We cloned the fusion product into pDONR HPH A [63] using a BP recombination reaction (Invitrogen, Grand Island, NY). We transformed the reaction mix into TOP10 cells (Invitrogen, Grand Island, NY) by electroporation, as recommended by the manufacturer. We grew the transformation mix on LB (1% tryptone, 0.5% yeast extract, 16 SOB salts [10 mM NaCl, 2.5 mM KCl], 1.5% agar) +50 mg/ml kanamycin at 37uC overnight. We picked colonies and transferred to 2 ml of LB liquid +50 mg/ml kanamycin to grow overnight in a 37uC shaking incubator. We isolated plasmid DNA from each culture using a miniprep kit (Qiagen, Hilden, Germany). We digested plasmid DNA with specific enzymes to verify the correct insertion. We designated this vector as pDHGL.
We grew Af293.1 in MAG medium supplemented with uridine and uracil (2% malt extract, 0.2% peptone, 1% glucose, trace elements and vitamin mix as modified [62], 2% agar, 5 mM uridine, 10 mM uracil). We performed the transformation as previously described [64], keeping pDHGL as a circular vector. We grew transformants on MMVAT supplemented with 5 mM uridine, 10 mM uracil, 0.2 M sucrose, and 400 mg/ml hygromycin at 37uC for 3-5 days. We identified the presence of pDHGL by Southern hybridization [65] of the lacZ coding region (Fig. S3). We also streaked transformants onto MMVSN supplemented with uridine and uracil (as described above for MMVAT, except 16MM salts contain 20 mM sodium nitrate instead of ammonium tartrate) and 40 mg/ml X-gal to grow at 37uC for 2 days. We screened for transformants that grew in the presence of hygromycin and developed a blue pigment on X-gal, signaling that lacZ expression was working properly. We designated this strain as Af293.1-GL.
For the first round of the high-copy inducer screen, we grew Af293.1-GL in MAG supplemented with uridine and uracil. We transformed the AMA1-Not1 A. fumigatus genomic library [43] into Af293.1-GL as described previously [64], but with changes. We combined each transformation mix with 50 mls of CM top agar (MMVAT, as described above, 0.1% yeast extract, 0.2% peptone, 0.1% tryptone, 1% CM supplement [27 mM adenine HCl, 33.5 mM methionine, 173 mM arginine, and 1.3 mM riboflavin], and 1% agar) supplemented with 1 M sucrose and spread the mixture over 10 plates (5 ml/plate). We grew transformants on CM supplemented with 0.2 M sucrose and 40 mg/ml X-gal at 37uC for 3-5 days. We screened for transformants that were both prototrophic for uridine and uracil and producing a blue pigment. We prepared genomic DNA from transformants [66] and transformed 1 ml of genomic DNA into TOP10 cells (Invitrogen, Grand Island, NY) by electroporation, as recommended by the manufacturer. We grew the transformation mix on LB +100 mg/ ml ampicillin at 37uC overnight. We picked colonies and transferred to 2 ml of LB liquid +100 mg/ml ampicillin to grow overnight in a 37uC shaking incubator. We isolated plasmid DNA from each culture using a miniprep kit (Qiagen, Hilden, Germany) and digested it with KpnI to identify individual plasmids.
For the second round of the high-copy inducer screen, we grew Af293.1-GL in MAG supplemented with uridine and uracil. We transformed each individual plasmid isolated from the first round of the genetic screen as described above, but with changes. We plated two amounts of protoplasts (20 ml and 100 ml) each in 4 mls of MMVAT top agar (as described above but with 1% agar) supplemented with 1 M sucrose. We grew transformants on MMVAT supplemented with 0.2 M sucrose and 40 mg/ml X-gal at 37uC for 3-4 days and put plates in the 4uC refrigerator to facilitate blue pigment production. Plasmids causing at least 80% of colonies to turn blue were sequenced using primers AMA-NotI F and AMA-NotI R. We also grew transformants in 10 mls CM at 37uC stationary overnight. We collected mycelia, froze in liquid nitrogen, and lyophilized overnight. We collected total protein and performed b-galactosidase assays to measure LacZ levels quantitatively (detailed below).
For the third round of the high-copy inducer screen, we PCR amplified individual genes from genomic library plasmids, flanked by native 59 and 39 non-coding regions. For gipA (Afu6g01910), we used the primer pair 6g01910 F and 6g01910 R. We amplified the fragments from Af293 genomic DNA using e2TAK DNA polymerase (Takara Bio Inc., Otsu, Shiga, Japan) following manufacturer recommendations. We cloned the PCR fragments into pDONR AMA [63] with a BP recombination reaction (Invitrogen, Grand Island, NY). We transformed the reaction mix into TOP10 cells (Invitrogen, Grand Island, NY) by electroporation, as recommended by the manufacturer. We grew the transformation mix on LB +100 mg/ml ampicillin at 37uC overnight. We picked colonies and transferred to 2 ml of LB liquid +100 mg/ml ampicillin to grow overnight in a 37uC shaking incubator. We isolated plasmid DNA from each culture using a miniprep kit (Qiagen, Hilden, Germany). We digested plasmid DNA with specific enzymes to verify the correct insertion. We designated the gipA-containing vector as pDONR AMA-gipA. We created a control vector, pDONR AMA-gliZ, which contained the gliZ coding region flanked by promoter and terminator regions. We generated this vector as described above for pDONR AMA-gipA using primers GliZ attB 1 and GliZ attB 2.
We grew Af293.1GL in MAG medium, supplemented with uridine and uracil. We performed the transformation as previously described [64], with changes, using pDONR AMA and pDONR AMA-gliZ as controls. After the 3 hour incubation of protoplasts, we carried out all reactions at half the specified volume. We used 500 ng-750 ng of circular vector DNA for each reaction. We plated two amounts of protoplasts (20 ml and 50 ml) each in 4 mls of MMVAT top agar supplemented with 1 M sucrose. We grew transformants on MMVAT medium supplemented with 0.2 M sucrose at 37uC for 2-3 days. To measure LacZ levels quantitatively, we grew transformants in 10 mls CM at 37uC stationary overnight. We collected mycelia, froze in liquid nitrogen, and lyophilized overnight. We collected total protein and performed bgalactosidase assays (detailed below). We chose three strains to use for future analysis: AMA.GL, AMA-gliZ.GL, and AMA-gipA.GL.

l Phage Library Screen
We isolated gipA cDNA clones from a l phage library constructed with the UniZAP vector and poly (A) + mRNA, as described by the manufacturer (Stratagene, La Jolla, California).
We performed a primary screen of the l phage library as recommended by manufacturer (Stratagene, La Jolla, California). We used the gipA coding region as a probe [65]. We excised plugs containing positive plaques and placed them in SM (.1 M NaCl, 10 mM MgSO 4 ?7H 2 O, 50 mM Tris?HCl [pH 7.5], and .01% gelatin) at 4uC overnight. After determining pfu concentration of plaques, we performed a secondary screen, as performed for the primary screen. We PCR amplified cDNA from excised phagemids using e2TAK DNA polymerase (Takara Bio Inc., Otsu, Shiga, Japan) and following manufacturer recommendations. We used M13 F and M13 R as primers for the reaction. Based on the PCR band sizes, we picked colonies from selected plates and transferred to 2 ml of LB liquid +100 mg/ml ampicillin to grow overnight in a 37uC shaking incubator. We isolated plasmid DNA from each culture using a miniprep kit (Qiagen, Hilden, Germany) and sequenced the cDNA insert using M13 F and M13 R primers. We used DNASTAR software (DNASTAR Inc.) to assemble and analyze the sequences obtained from the library screen.

RNA Dot Blot Analysis
For all Dot Blot assays, we grew stationary cultures in 25 mls of CM (repressing) (described above) or CD (non-repressing) (87.6 mM sucrose, 35.3 mM sodium nitrate, 5.8 mM K 2 HPO 4 , 2 mM MgSO 4 ?7H 2 O, 6.7 mM KCl, and 0.025 mM ferrous ammonium sulfate) at a concentration of 5610 6 spores/ml at 37uC for 48 hrs. We included 400 mg/ml hygromycin in dot blot assays involving AMA.G, AMA.Z, AMA.A, AMA-gliZ.G, AMA-gliZ.Z, AMA-gliZ.A, AMA-gipA.G, AMA-gipA.Z, and AMA-gipA.A. We prepared total RNA from freeze-dried mycelia using the TRIzol method [62]. We incubated total RNA with denaturing solution (50% formamide, 16%formaldehyde, 16 borate buffer [206 borate buffer: 0.4 M Boric Acid, 4 mM EDTA, pH 8.3 with NaOH], 0.025% bromophenol blue) for 10 min at 65uC. We quenched samples on ice for 10 min, then added equal volume 206 SSC (3 M NaCl, 0.3 M sodium citrate, pH 7.0). We placed a nylon membrane that had been equilibrated in 106 SSC for 10 min into a 96 well dot blot apparatus attached to a vacuum manifold. We collected samples, each containing 3 mg of RNA unless otherwise noted, in 100 ml volumes by aspiration. We aspirated 50 ml of 106 SSC through the membrane in duplicate immediately before and after samples were collected. Once all samples were aspirated, we air dried nylon membranes overnight and then baked them in an oven for 2 hrs. For prehybridization and hybridization, we sealed nylon membranes in a bag using a heated sealer. We prehybridized membranes for 4-6 hrs at 42uC and hybridized membranes overnight at 42uC. For DNA probes, we only used the coding region of each gene of interest (gliA, gliP, gliZ, gliT, gipA, and actin). After hybridization, we washed membranes as we would for Southern hybridization [65] and exposed them to a Typhoon 8600 PhosphorImager (GE Healthcare Life Sciences, Pittsburgh, PA) overnight. We quantified the intensity of hybridization using ImageQuant5.1 software (GE Healthcare Life Sciences, Pittsburgh, PA).

Gliotoxin Extraction and HPLC Analysis
We extracted spent culture medium from RNA dot blot assays in 15 ml chloroform for 30 min at room temperature on an orbital shaker set to 250 rpm. We transferred the chloroform phase to a 50 ml conical tube (BD Biosciences, San Jose, California) and repeated the extraction twice for a total of 45 mls of chloroform. We dried open tubes under a hood until the chloroform was completely evaporated. We added 15 mls of chloroform to each tube and mixed, to concentrate the extracted material at the bottom. We dried open tubes under a hood until chloroform was completely evaporated. We added 1 ml of methanol to each tube and mixed to dissolve all methanol-soluble substances and transferred extracts to microcentrifuge tubes (Fisher Scientific, Pittsburgh, PA). We dried open tubes under a hood until all methanol was completely evaporated.
We dissolved extracts in 50 ml of dimethyl sulfoxide (DMSO), spun tubes to pellet insoluble debris, and transferred DMSO to fresh microcentrifuge tubes. We quantified gliotoxin levels by running samples through a reverse-phase high performance liquid chromatography (RP-HPLC) system with a Waters 996 photodiode array detector (Waters, Milford, MA). We ran samples through a Sonoma 2.16250 mm C18 column (100 Å pore size) packed with 5 mM particles (VWR, Radnor, PA). The mobile phase consisted of H 2 O, 0.1% TFA (solution A) and 100% Acetonitrile, 0.1% TFA (solution B): 10% B up to 80% B over 30 min. The injection volume was 10 ml and flow rate was set at 0.4 ml/min. Gliotoxin eluted from the column at 14.7 min and absorbance was read at 268 nm wavelength. We determined gliotoxin concentrations by interpolation from a 9 point standard curve (39 ng to 10 mg) prepared using purified gliotoxin (Sigma-Aldrich Corp., St. Louis, MO).

Microarray
The DNA amplicon microarray for Af293 was created previously [67]. We grew AMA.GL and AMA-gipA.GL in 25 mls CM (described above) at 37uC for 24 hrs in stationary cultures. The spore concentration was 5610 6 spores/ml. We used two independent biological replicates for AMA.GL and AMA-gipA.GL growth assays. We prepared total RNA from freeze-dried mycelia using the TRIzol method [62]. We carried out RNA labeling reactions and hybridizations as described in the J. Craig Venter Institute Microarray Protocols (http://pfgrc.jcvi.org/ index.php/microarray/protocols.html). We repeated all the hybridizations in dye-swap sets. We scanned and analyzed hybridized slides as described previously [67]. We averaged all replicates for the official data used in our analysis.

Virulence Assays with Toll-Deficient D. melanogaster
We injected the dorsal side of the thorax of CO 2 -anesthetized, adult, Toll-deficient D. melanogaster flies with a sterile 0.25 mm needle that had been dipped in a solution containing 10 7 spores/ ml of A. fumigatus conidia. We infected 20-25 flies per strain for each virulence assay, which was repeated at least twice. Flies were kept in a 29uC incubator to maximize susceptibility to microbial challenge and monitored for 7 days. Flies that died within 3 hrs of the injection were not included in the survival graph, as these flies most likely died as a result of the puncture wound.

Deletion and Complementation of gipA in Af1160
We amplified the 59 flanking region (FR) and the 39 FR from gipA using primers 01910 59 F, 01910 59 R, 01910 39 F, and 01910 39 R. We engineered a unique NotI site into 01910 59 F to linearize the final deletion construct. We amplified the fragments from Af293 genomic DNA using e2TAK DNA polymerase (Takara Bio Inc., Otsu, Shiga, Japan) and following manufacturer recommendations. We cloned the gipA 59 FR into pDONR P4-P1R and we cloned the gipA 39 FR into pDONR P2R-P3, using BP recombination reactions (Invitrogen, Grand Island, NY). We transformed BP reaction mixes into TOP10 cells (Invitrogen, Grand Island, NY) by electroporation, as recommended by the manufacturer. We grew transformed cells on LB (described above) +50 mg/ml kanamycin at 37uC overnight. We picked colonies and transferred to 2 ml LB liquid +50 mg/ml kanamycin to grow overnight in a 37uC shaking incubator. We isolated plasmid DNA from each culture using a miniprep kit (Qiagen, Hilden, Germany). We digested plasmid DNA with specific enzymes to verify the correct fragment orientation.
To create the deletion constructs, we combined pDONR P4-P1R-gipA 59 FR, pDONR P2R-P3-gipA 39 FR, and pDONR 221-AnpyrG [63] in an LR recombination reaction with pDEST R4-R3 as the destination vector (Invitrogen, Grand Island, NY) [68]. We transformed the LR reaction mix into TOP10 cells (Invitrogen, Grand Island, NY) by electroporation, as recommended by the manufacturer. We grew transformed cells on LB +100 mg/ml ampicillin at 37uC overnight. We picked colonies and transferred to 2 ml LB liquid +100 mg/ml ampicillin to grow overnight in a 37uC shaking incubator. We isolated plasmid DNA from each culture using a miniprep kit (Qiagen, Hilden, Germany). We digested plasmid DNA with specific enzymes to verify the correct fragment orientation. We grew bacterial cultures containing the correct plasmid in 250 ml LB liquid +100 mg/ml ampicillin overnight in a 37uC shaking incubator. We purified plasmid DNA by banding on cesium chloride ethidium bromide gradients [65]. We designated the plasmid pDEST R4-R3-gipA 59 FR-AnpyrG-gipA 39 FR.
We grew A. fumigatus 1160 (obtained from FGSC) in MAG supplemented with uridine and uracil. We performed the transformation as previously described [64], linearizing the deletion construct with NotI to facilitate homologous recombination. We grew transformants on MMV supplemented with 0.2 M sucrose at 37uC for 3-5 days. We screened for mutants that were prototrophic for uridine and uracil. We prepared genomic DNA from transformant strains [66], and we identified deletion mutants by Southern blot analysis (Fig. S4) [65]. We made a DNA probe using the gipA 59 and 39 FRs to verify DgipA.
To create a vector for complementation, we PCR amplified the gipA coding region, flanked by a 3 kilobase promoter region and 500 base pair terminator region, using primers gipA 3 kb F and 6g01910 R. We amplified the fragment from Af293 genomic DNA using e2TAK DNA polymerase (Takara Bio Inc., Otsu, Shiga, Japan) and following manufacturer recommendations. We cloned the PCR product into pDONR HPH B [63], as described previously for pDHGL and designated this vector pDONR HPH-gipA. We grew DgipA in MAG medium and performed a transformation as previously described for pDHGL. For Southern hybridization, we made a DNA probe using the coding region of gipA (Fig. S5). We designated this strain as gipA(R).

Construction of DgliZ/DgipA
We grew DgipA on YAG supplemented with uridine and uracil +1 mg/ml 5-Fluororotic acid (5-FOA) at 37uC. Colonies that reverted to a pyrGphenotype grew as outgrowths from the original streak. We prepared genomic DNA from these pyrGoutgrowths [66] and tested them for the presence of gipA by Southern hybridization (Fig. S6) [65]. We used the gipA coding region as a DNA probe. We designated this mutant as DgipA.0. We transformed pDEST R4-R3-gliZ 59 FR-AnpyrG-gliZ 39 FR into DgipA.0, as described above for DgipA. We used gliZ 59 and 39 flanking regions as a DNA probe for Southern hybridization (Fig. S7).

Protein Binding Microarray
We amplified the C 2 H 2 DNA binding region from gipA using primers gipA C2H2 F and gipA C2H2 R. We used Af293 as template and AccuPrime Pfx DNA polymerase (Life Technologies, Grand Island, NY) and per manufacturer's recommendations. We cloned the PCR product into pDONR 221 using a BP recombination reaction (Invitrogen, Grand Island, NY). We transformed the reaction mix into TOP10 cells (Invitrogen, Grand Island, NY) by electroporation, as recommended by the manufacturer. We grew the transformation mix on LB +50 mg/ml kanamycin at 37uC overnight. We picked colonies and transferred to 2 ml of LB liquid +50 mg/ml kanamycin to grow overnight in a 37uC shaking incubator. We isolated plasmid DNA from each culture using a miniprep kit (Qiagen, Hilden, Germany). We digested plasmid DNA with specific enzymes to verify the correct insertion. We recombined the gipA C 2 H 2 region into pDEST 15 using an LR recombination reaction (Invitrogen, Grand Island, NY). We transformed the reaction mix into TOP10 cells (Invitrogen, Grand Island, NY) by electroporation, as recommended by the manufacturer. We grew the transformation mix on LB +100 mg/ml ampicillin at 37uC overnight. We picked colonies and transferred to 2 ml of LB liquid +100 mg/ml ampicillin to grow overnight in a 37uC shaking incubator. We isolated plasmid DNA from each culture using a miniprep kit (Qiagen, Hilden, Germany). We digested plasmid DNA with specific enzymes to verify the correct insertion. This vector, pDEST 15-gipA C 2 H 2 , was used in a protein binding microarray analysis as previously described [69].

b-Galactosidase Assays
We ground 50 ml lyophilized mycelia to a fine powder with acid-washed glass beads (400-650 mm) (Sigma-Aldrich Corp., St. Louis, MO). We suspended the ground powder in 200 ml protein extraction buffer (PEB) ( , pH to 7.0). For samples grown in repressing conditions (CM), the total protein added was 1 mg. For samples grown in non-repressing conditions (CD), the total protein added was 0.1 mg. To begin the b-galactosidase assay, we added 20 ml of 2-Nitrophenyl-b-D-galactopyranoside (ONPG) (Sigma-Aldrich Corp., St. Louis, MO), diluted to 4 mg/ml in Z Buffer, and placed the 96-well plate in a 37uC incubator. We timed reactions and stopped samples with 50 ml 1 M Na 2 CO 3 . We measured absorbance at OD420 and calculated Units of bgalactosidase activity/mg protein with the following equation: (OD4206TV)/(0.00456T6V6C), where TV is total volume of the reaction in ml, T is time in min, V is volume of protein added in ml, and C is concentration of protein used in mg/ml.

Creation of Strains for Bypass Suppression Experiments
We grew Af293.1 in MAG supplemented with uridine and uracil (described above). We transformed pDEST R4-R3-gliZ 59 FR-AnpyrG-gliZ 39 FR and pDEST-gipA 59 FR-AnpyrG-gipA 39 FR into Af293.1, as previously described for DgipA, except we plated transformants on YAG supplemented with 0.2 M sucrose. We used gliZ 59 and 39 FRs and gipA 59 and 39 FRs for Southern hybridization and designated these strains DgliZ.1 and DgipA.1 (Fig. S9). We chose one strain from the DgliZ transformation that did not show homologous recombination at the gliZ locus, but was prototrophic for uridine and uracil. We designated this strain pyrG+. We collected total RNA and performed dot blot analysis on pyrG+, DgliZ.1, and DgipA.1 to verify loss of gliZ and gipA, respectively (described in RNA dot blot Analysis) (Fig. S10). We cloned gliZ and gipA into pDONR AMA/HPH [63], as described above for pDONR AMA-gliZ and pDONR AMA-gipA, respectively, except we used primers gipA 3 kb F and 6g01910 R for the gipA cassette. We designated these plasmids pDONR AMA/HPH-gliZ and pDONR/HPH-gipA. We transformed these plasmids, along with pDONR AMA/HPH empty vector, into pyrG+, DgliZ.1, and DgipA.1, as described above for AMA-gipA.GL, except we grew transformants on YAG supplemented with uridine and uracil, 0.2 M sucrose, and 400 mg/ml hygromycin. We designated these strains as AMA.G, AMA-gliZ.G, AMA-gipA.G, AMA.Z, AMA-gliZ.Z, AMA-gipA.Z, AMA.A, AMA-gliZ.A, and AMA-gipA.A. Figure S1 Complete microarray data. All genes tested are listed by locus. The average Log2 value from four independent experiments is shown for each locus listed. (XLSM) Figure S2 Secondary metabolism cluster graphs. Graphs depict Log2 values based on microarray data. Data are presented as mean 6 SD. Clusters 1-18 (graphs 1-5) all contain at least one gene that is induced .2-fold in the presence of high-copy gipA expression. Clusters 19-31 (graphs 6-8) do not contain any genes that are induced .2-fold in the presence of high-copy gipA expression. (TIF) Figure S3 Southern hybridization of Af293.1-GL transformants. Genomic DNA was digested with EcoRI overnight at 37uC then run on a 0.8% agarose gel. The lacZ coding region was used as the probe. WT is wild-type (Af293.1) and is expected to have no band. Proper transformants should only have one band of unknown size. Lanes with asterisks are transformants that produced blue pigment when grown on MMVSN U/U (non-repressing) medium with X-gal. (TIF) Figure S4 Southern hybridizations of gipA and gliZ deletion mutants. Samples were run on a 0.8% agarose gel. WT is wildtype (Af1160).  Figure S6 Southern hybridization of DgipA mutants treated with 5-FOA. Samples were run on a 0.8% agarose gel. WT is wild-type (Af1160) and DA is the untreated deletion strain. DgipA.0 mutants were digested with EcoRV and probed with the gipA coding region. Only WT should display a band, which is 7616 bp. (TIF) Figure S7 Southern hybridization of DgliZ/DgipA double mutants. Samples were run on a 0.8% agarose gel. WT is wildtype (Af1160). DgliZ/DgipA transformants were digested with SacI and probed with the gliZ and gipA coding regions. WT should display two bands, which are 7474 bps (gipA), and 4566 bps (gliZ). The DgipA control (DA) should display one band, which is 4566 bps (gliZ). Correct transformants should display not display the 7474 bp band or the 4566 bp band, signifying that the gipA and gliZ coding regions are not present, respectively. There is a 11712 bp band present in all lanes as a result of a third probe, but is not being mentioned here because it is not included in this paper. (TIF) Figure S8 Southern hybridizations of promoter mutagenesis transformants. Samples were digested with EcoRI and run on a 0.8% agarose gel. WT is wild-type (Af293.1) and GL is Af293.1-GL, which contains the gliA-lacZ construct. The lacZ coding region was used as the probe. WT should not display a band and GL should have a single band ,6500 bps. Correct transformants should display a single band, any size.