Sterol Biosynthesis and Azole Tolerance Is Governed by the Opposing Actions of SrbA and the CCAAT Binding Complex

Azole drugs selectively target fungal sterol biosynthesis and are critical to our antifungal therapeutic arsenal. However, resistance to this class of drugs, particularly in the major human mould pathogen Aspergillus fumigatus, is emerging and reaching levels that have prompted some to suggest that there is a realistic probability that they will be lost for clinical use. The dominating class of pan-azole resistant isolates is characterized by the presence of a tandem repeat of at least 34 bases (TR34) within the promoter of cyp51A, the gene encoding the azole drug target sterol C14-demethylase. Here we demonstrate that the repeat sequence in TR34 is bound by both the sterol regulatory element binding protein (SREBP) SrbA, and the CCAAT binding complex (CBC). We show that the CBC acts complementary to SrbA as a negative regulator of ergosterol biosynthesis and show that lack of CBC activity results in increased sterol levels via transcriptional derepression of multiple ergosterol biosynthetic genes including those coding for HMG-CoA-synthase, HMG-CoA-reductase and sterol C14-demethylase. In agreement with these findings, inactivation of the CBC increased tolerance to different classes of drugs targeting ergosterol biosynthesis including the azoles, allylamines (terbinafine) and statins (simvastatin). We reveal that a clinically relevant mutation in HapE (P88L) significantly impairs the binding affinity of the CBC to its target site. We identify that the mechanism underpinning TR34 driven overexpression of cyp51A results from duplication of SrbA but not CBC binding sites and show that deletion of the 34 mer results in lack of cyp51A expression and increased azole susceptibility similar to a cyp51A null mutant. Finally we show that strains lacking a functional CBC are severely attenuated for pathogenicity in a pulmonary and systemic model of aspergillosis.


Introduction
Sterols are components of most eukaryotic cell membranes playing key roles in sustaining membrane integrity and fluidity. The de novo synthesis of sterols by fungi is essential for their viability and numerous antifungal drugs have been developed that exploit the differences between enzymes in the sterol biosynthetic pathway of fungal pathogens and their hosts. The most notable sterol biosynthetic inhibitors are the azoles which are extensively used in crop protection and have been the cornerstone of systemic antifungal therapy in man for the last 30 years [1].
Triazoles such as voriconazole (VORI), itraconazole (ITRA) or posaconazole (POSA) represent the main antifungal drug class employed to treat disease caused by Aspergillus spp. Although in general Aspergillus fumigatus is susceptible to these drugs, resistance is emerging and reaching levels that have prompted some health centres to move away from the use of azoles as a sole first line therapeutic, opting instead for high cost combination therapies and/or less effective agents [2,3,4,5]. Azole drugs act by inhibiting the function of the sterol C14-demethylase Cyp51, leading to ergosterol depletion and simultaneous accumulation of toxic sterol compounds [6]. A principal cause of azole resistance in clinical strains of A. fumigatus is modification of the cyp51A gene, one of two genes that encode isoforms of sterol C14-demethylase in this pathogen. A particular family of pan-azole resistant isolates dominates. Typified by the TR34/L98H modification, but also including TR46/Y121F/T289A and TR53, they harbor a tandem repeat in the promoter of cyp51A along with a non-synonymous mutation resulting in one or more amino acid changes in the Cyp51A protein. In the case of TR34/L98H this is manifest as a duplication of a 34 mer within the 5' non-translated (5' NTR) region of cyp51A, combined with a lysine to histidine substitution at position 98 in the protein.
Understanding the mechanisms by which the TR34/L98H family and non-cyp51A type mutations lead to resistance is critical to formulating strategies to both detect and treat resistant infections.
What drives resistance in the TR34/L98H family is only partially understood. Introduction of the L98H modification into a hitherto wild-type (wt) isolate of A. fumigatus results in a modest but significant increase in triazole tolerance. Introduction of the TR34 variant into a conventional cyp51A promoter results in an approximate doubling of cyp51A gene expression and an associated increase in tolerance to azoles [7]. Only when these modifications are combined can isolates reach tolerance levels that exceed clinically relevant breakpoints as defined by EUCAST (European Committee on Antimicrobial Susceptibility Testing) [8]. Although the mechanistic basis of L98H driven increase in azole tolerance has been linked to modification of the tertiary structure of Cyp51A [7], the cause of increased expression resulting from the TR34 promoter variant has not been elucidated.
In mammalian systems, sterol production is transcriptionally regulated by sterol regulatory element binding proteins (SREBPs). SREBPs belong to the basic helix-loop-helix (bHLH) transcription factors. In their inactive form they are membrane-bound, once activated the regulator is released and accumulates in the nucleus where it binds to sterol regulatory elements (SREs) in the promoters of target genes, including sterol C14-demethylase and activates expression [9,10]. At sterol excess SREBPs remain inactive causing decreased transcript levels of their targets [10]. The activating role of the SREBPs is facilitated by the action of the CCAAT-binding complex (CBC) NF-Y which is made up of three subunits NF-YA, NF-YB and NF-YC. The SREBPs and NF-Y synergistically activate expression of essentially all genes involved in sterol metabolism [11].
Orthologues to mammalian SREBPs have been found in several fungi including A. fumigatus. The A. fumigatus SREBP homologue, termed SrbA, is a non-redundant transcription factor which, like its mammalian counterparts, plays a key role in the regulation of sterol biosynthetic genes, including cyp51A. In keeping with this role, A. fumigatus strains lacking SrbA have reduced tolerance to the azoles [12,13,14].
Similarly, A. fumigatus has an orthologue of the CBC. As with the mammalian regulator, the A. fumigatus CBC is a multimeric transcription factor complex comprising three subunits (HapB/HapC/HapE) and is highly conserved from yeast (Hap2p/Hap3p/Hap5p) to man (NF-YA/NF-YB/NF-YC) [11,15,16,17,18,19]. A mutation in the HapE subunit (P88L) was recently identified as the causative modification that led to azole resistance in an A. fumigatus strain isolated from a patient in the Netherlands [20].
As both SrbA and the CBC have been directly implicated in modified azole tolerance in A. fumigatus, and their orthologues have positively acting sterol regulatory functions, we sought to understand the role of the CBC in sterol biosynthesis and both regulators in clinical azole resistance. Particularly we were interested in uncovering any role of these regulators in TR34 mediated resistance.

Results
The CBC recognises a CGAAT motif within the azole resistance associated 34 mer in the promoter of cyp51A, a key gene in the sterol biosynthetic pathway The A. fumigatus CBC functions as a heterotrimer comprising HapB, HapC and HapE and is known to interact with the highly abundant consensus motifs CCAAT and CGAAT [21,22]. A number of potential binding sites are evident within the cyp51A promoter, therefore we sought to identify if the CBC interacted directly with the promoter of cyp51A using chromatin-immunoprecipitation followed by next-generation sequencing (ChIP-seq). To this end, we replaced the native hapC gene in isolate A1160P+ with a chimeric gfp-construct. The respective strain expresses C-terminally GFP-tagged HapC protein (hapC GFP , Fig 1A). We sequenced anti-GFP bound DNA fractions from cell extracts and identified a single region within the promoter (within 1kb upstream of the translational start site) of cyp51A to which the GFP-CBC was bound (S1 Table). The sequencing peak identified by ChIP-seq was at position -297 relative to the translation start placing it at the 3' end of the 34 mer. This result was validated by ChIP-qPCR by comparing enrichment of the region identified in the ChIP-seq analysis with region not bound by the CBC in the promoter of actA (Fig 1B). By examining the promoter region around the ChIP-seq peak, we found the CBC binding motif CGAAT at position -293 to -289 ( Fig 2A). To further analyse the interaction of the CBC to this motif, binding kinetics and affinity were monitored by surface plasmon resonance spectroscopy (SPR) protein-DNA interaction analysis. Kinetic CBC binding responses to a 37-bp cyp51A promoter duplex revealed a dissociation constant (K D ) of 74.4 nM, demonstrating that the CBC effectively recognised the CGAAT -293 motif within the original 34 mer in vitro (Fig 2B, upper panel 2). Importantly, although the CGAAT sequence is duplicated in TR34 (CGAAT -327 ), the binding affinity of the CBC at the duplicated sequence is about 8-fold decreased and consequently not effectively bound (Fig 2C panel 2).
HapX, a transcriptional regulator associated with adaptation to environmental iron cues, binds to the promoter of cyp51A Cyp51A is known to use iron as a catalytic co-factor and the CBC is a known interaction partner of HapX, an iron responsive basic region leucine zipper (bZIP) transcription factor. We therefore screened the nucleotide sequence downstream the CGAAT -293 motif for potential HapX binding sites [19,21,24]. We identified a pseudo-palindromic TTACTAA sequence at The hapC GFP strain is phenotypically indistinct from the wt host and HapC GFP localises to the nucleus. For confocal laser microscopy 10 4 spores of the respective isolates were grown in liquid AMM at 37°C for 24 h. DAPI was used to co-stain the nuclei. (B) In vivo binding of the CBC was confirmed by comparing % recovery of DNA (ChIP-qPCR) from promoters of ergosterol biosynthetic genes (erg7B, erg13B, cyp51A and cyp51B) to an unbound region of the actA promoter (for more details, see Material and Methods position -275 to -269 in the cyp51A promoter that corresponds exactly to the yeast AP-1 (YAP1) bZIP consensus binding site [25]. Employing SPR co-injection analysis we confirmed that this region of the cyp51A promoter is synergistically bound by the CBC/HapX complex (Fig 2B panel 4 and 5). Consistent with the fact that the HapX-binding site is not duplicated in the 34 base tandem repeat, combinatorial CBC/HapX recognition was not detectable on a DNA duplex that contained the duplicated CGAAT motif (Fig 2C, panel 4

and 5).
Together these data demonstrate that only the wt 34 mer of the cyp51A promoter is a direct target of the CBC and its interaction partner HapX.

The CBC represses cyp51A expression, this repression is liberated in a HapE P88L mutant
To determine if the CBC is functioning as a repressor, or an activator of cyp51A gene expression and to assess the role of the HapE P88L mutation we generated CBC gene deletion mutants of each subunit of the CBC namely, HapB (ΔhapB), HapC (ΔhapC) and HapE (ΔhapE), a strain expressing HapE P88L (hapE P88L ), and a deletion mutant of HapX (ΔhapX) (S2 Table). Disruption of any of the CBC subunits, HapX or mutation of HapE resulted in increased tolerance to ITRA, VORI and POSA (Table 1). Additionally, similar to HapE P88L [20] cyp51A expression  was increased by about 2-fold in the ΔhapC strain ( Fig 3A). This result indicates that the CBC and HapX act as repressors of cyp51A and that the HapE P88L mutation abolishes this repressing function. The loss of repressing activity of HapB/HapC/HapE P88L seems to be linked to a reduction in binding affinity of the mutated CBC to the CGAAT -293 motif as indicated by SPR analysis (Fig 2B; compare panel 1 to panel 2).
Combined, these results suggest that azole resistance caused by HapE P88L is linked to impaired CBC DNA-binding function.

The CBC is a repressor of several genes in the ergosterol biosynthetic pathway
The discrepancy between the increase in expression of cyp51A (c. 2-fold) and azole resistance (>32-fold) suggested that cyp51A-independent factors may be influencing azole resistance in CBC deficient strains. This is supported by the fact that a ΔhapCΔcyp51A is more resistant to the azoles than a single cyp51A null mutant (Table 1). We therefore hypothesised a possible function of the CBC in the repression of other genes in the ergosterol biosynthetic pathway.
Interrogation of our ChIP-seq data allowed us to identify CBC binding sites in the promoter regions (< 1.0 kb 5' NTR) of around half of the ergosterol biosynthetic genes (14 out of 27, S1 Table). We validated the results of the ChIP-seq analysis by ChIP-qPCR for four of the genes erg13B, erg7B, cyp51A and cyp51B (Fig 1B). To assess if binding of the CBC was affected by azole treatment, which would mimic cellular sterol depletion, we performed ChIP-qPCR using A. fumigatus cultures grown in the presence or absence of ITRA. We found enrichment of the binding regions in both sterol replete (-ITRA) and sterol deplete (+ITRA) conditions indicating that the CBC is constitutively bound to all four promoters under these conditions ( Fig 1B).
To investigate if binding of the CBC was linked to regulation of gene expression, we monitored expression levels of several genes associated with the ergosterol biosynthetic pathway in our CBC mutant (ΔhapC). Specifically we monitored genes coding for enzymes of the first and committed steps in ergosterol biosynthesis, HMG-CoA synthase (paralogs erg13A and erg13B) and HMG-CoA reductase (paralogs hmg1 and hmg2), the genes encoding the azole drug target sterol C14-demethylase (paralogs cyp51A and cyp51B) as well as erg7B, that encodes lanosterol synthase. For four of these genes, erg13A, erg13B, hmg1 and erg7B, expression was significantly increased by at least 3-fold in ΔhapC when compared to the isogenic host strain (Fig 3A). For Table 1. Azole sensitivity of strains generated in this study. MIC testing has been carried out using the EUCAST broth microdilution reference method [26]. cyp51B we detected a modest increase in transcript levels (>1.5 fold). We were unable to detect expression of hmg2 in either strain due to the very low abundance of the transcript. It is noteworthy that our ChIP-seq results did not indicate binding of the CBC to the 1.5 kb region upstream of the erg13A translation start site, suggesting an indirect regulation of this gene by the CBC. Based on the idea that increased expression of several ergosterol biosynthetic genes in CBC mutants might lead to increased activity of the enzymes, we determined MIC levels of ΔhapC for simvastatin and terbinafine ( Fig 3B). These drugs, belonging to the families of statins and allylamines, are ergosterol biosynthesis inhibitors that inhibit the enzymes HMG-CoA reductase (Hmg) and squalene epoxidase (Erg1) respectively. Both of these enzymes perform functions upstream of the C14-demethylation carried out by the Cyp51 enzymes ( Fig 3C). In keeping with our hypothesis, ΔhapC displayed a significant increase in tolerance to both of these drugs.
Taken together, these data illustrate that loss of the CBC leads to upregulation of genes involved in several steps of ergosterol biosynthesis, and increases the activity of several enzymatic steps in this pathway. This results in resistance to ergosterol biosynthesis inhibitors belonging to the azoles, statins and allylamine drug classes.
These findings suggest that transcriptional upregulation of ergosterol biosynthetic genes in ΔhapC leads to increased sterol levels, particularly ergosterol, which is likely to contribute to resistance to ergosterol biosynthesis inhibitors in CBC mutants.
CBC mutation results in severe growth reduction and attenuated virulence in a murine model of aspergillosis The high abundance of CCAAT and CGAAT motifs in eukaryotic promoters suggests that the expression of a large proportion of protein coding genes might be affected by CBC mutation. Hence, it seemed likely that CBC defective mutants would have significant phenotypic defects. We assessed growth of CBC mutants (ΔhapB, ΔhapC, ΔhapE, hapE P88L ) on Aspergillus minimal medium (AMM), RPMI as well as Sabouraud dextrose agar (SAB, rich-nutrient medium). Lack of either subunit resulted in significant growth defects on all media with individual deletion mutants showing identical phenotypes (Fig 4A). The hapE P88L isolate revealed a less dramatic phenotype compared to hap-deletion mutants. Growth of this isolate, carrying a single amino acid mutation, was impaired particularly on RPMI.
To assess the requirement of the CBC in infection, we compared the virulence of ΔhapC and hapC REC with that of the isogenic control strain in systemic and pulmonary murine models of aspergillosis. The pathogenicity of the isogenic control and hapC REC strains were indistinguishable, causing 100% mortality (n = 10) between 9 and 10 days after systemic infection and between 12 and 14 days after pulmonary infection. Systemic infection by ΔhapC caused no mortality while in the pulmonary model only one animal succumbed to the infection (Fig 4B). Tissue burden studies showed a reduced fungal load in animals infected i.v. (n = 10) and i.n. (n = 8) with ΔhapC in comparison to those challenged with wt and hapC REC (S1 Fig). Pulmonary infection resulted in high fungal burden in all animals challenged with wt and hapC REC while ΔhapC was cleared in all mice except two. Kidneys from animals inoculated with wt or hapC REC showed low viable fungal elements, however no colonies were detected in ΔhapCinfected mice. With exception of the brain, the systemic infection resulted in high fungal loads in all tissues after wt and hapC REC inoculation. Animals challenged with ΔhapC showed significantly lower CFU/g in all organs except for the liver and brain when compared to those challenged with wt or hapC REC .
Taken together, these results demonstrate the crucial role of the CBC for A. fumigatus growth and show that transcriptional circuits mediated by this regulator are critical during infection of an immunocompromised host.
The positive regulator of transcription SrbA binds to the 34 mer in the cyp51A promoter Our analysis of the CBC suggests that it is associated with the 34 mer but not directly responsible for the transcriptional enhancement of cyp51A observed in strains with the TR34 mutation. This led us to assess if the other known transcriptional regulator of sterol biosynthetic genes, SrbA, was involved.
In A. fumigatus the sterol regulatory element binding protein SrbA has been described as direct positive regulator of sterol biosynthesis in general and cyp51A in particular [13,27]. A previous study identified a motif similar to the reported sterol regulatory element (Sre1) binding motif of S. pombe upstream of the TR34 site [13]. In an attempt to corroborate the location of this site, we performed a retrospective analysis of a recent genome-wide ChIP-Seq evaluation of SrbA binding [27] and identified a sequencing peak summit at position -302 from the cyp51A coding sequence start site, which in contrast to the previous predicted location of the binding site, is within the duplicated 34 mer. By comparing the 34 bases of TR34 with the SrbA consensus binding motif [27] we identified two putative SREs (SRE1 and SRE2, Fig 5A). These SREs are also evident in the overlapping TR46 and TR53 found in other resistant clinical isolates ( Fig 5A) [28,29,30].
To test if SrbA directly interacts with the predicted SREs, we assessed binding of recombinant SrbA to DNA-duplexes containing either SRE1, SRE2 or both sites by SPR (Fig 5B). SrbA showed interaction with both of the predicted binding sites within the 34 mer, however affinity for SRE1 (K D = 22.8 nM) was 12-fold higher than that for SRE2 (K D = 272.6 nM) (Fig 5B; panels 1 and 2 respectively). Employing a DNA duplex containing both SREs we measured a 2-fold increase in the saturating SrbA response (R max value of 197.6) combined with an apparent K D of 22.0 nM, which does not represent the simple average of the SrbA affinities measured for the single SREs. In conclusion, we propose cooperative binding of two SrbA homodimers to the 34 mer (Fig 5B; panel 3). Notably, the affinities of SrbA and the CBC to their partially overlapping binding motifs are at a similar level (74.4 vs. 22.0 nM), indicating that both regulators bind competitively to the 34 mer within the wt cyp51A promoter. By contrast, the 8-fold lower affinity for CBC at the CGAAT motif duplicated in TR34 will favor cooperative SrbA binding to both of the SREs in the 5´-TR34 region.
This finding suggests that the increased expression of cyp51A in strains harboring TR34, TR46 or TR53 could be the result of increased SrbA activity as a result of the duplication of its consensus binding sites and a lack of CBC and accordingly CBC/HapX repression.

The 34 mer is crucial for activation of cyp51A gene expression and azole resistance
Based on the hypothesis that the 34 mer is required for SrbA mediated cyp51A activation and consequent azole resistance, we generated a cyp51A promoter variant lacking the 34 mer (cyp51A Δ34 , S1 Fig). In order to confirm the previous results described by Snelders in 2011 [7] and to draw a direct comparison to the cyp51A Δ34 isolate in an isogenic background, we also generated a strain carrying the tandem repeat TR34 (cyp51A TR34 ). Deletion of the 34 mer (cyp51A Δ34 ) led to a >90% reduction in cyp51A expression similar to the srbA deletion mutant (ΔsrbA) (Fig 6), which is consistent with our hypothesis. A similar result has been observed in a previous study using a cyp51A promoter based luciferase reporter approach [31]. In addition, the cyp51A Δ34 isolate displayed a 4-fold reduction in MIC (from 0.50 mg/L to 0.13 mg/L, Fig  6), which mimics the azole susceptibility phenotype of the isogenic cyp51A deletion mutant (Δcyp51A). MIC levels for the azoles were even lower for ΔsrbA, which is likely a result of lack of activation of several ergosterol biosynthetic genes in this mutant [12,27].
Taken together, our data demonstrates that cyp51A activation relies on the presence of the 34 mer. This region is required for SrbA mediated activation of cyp51A gene expression, hence, for azole resistance in A. fumigatus.

Discussion
In this study, we report a novel mechanism for the regulation of ergosterol metabolism in A. fumigatus and highlight the interplay between the transcriptional regulators SrbA and CBC. We also describe how perturbation of this regulatory mechanism through numerous routes can induce resistance to the primary class of agent used to treat fungal disease, the triazoles, which act by inhibiting ergosterol biosynthesis through depleting sterol C14-demethylase activity.
In A. fumigatus, the sterol regulatory element binding protein SrbA has been shown to be a positive acting transcriptional regulator of sterol biosynthesis and, aligned with this, has a crucial role in azole tolerance [14]. A recent study revealed that several ergosterol biosynthetic genes, including those encoding the two paralogs of Cyp51 (Cyp51A and Cyp51B) and Erg25 (Erg25A and Erg25B), are under direct control of SrbA [27], and strains lacking the regulator have significant reduction in the expression of the aforementioned genes. The role of SREBPs in activating expression of sterol biosynthesis is highly conserved.
In mammals, two genes encode three distinct isoforms SREBP-1a, SREBP-1c and SREBP-2, which are involved in positively regulating various aspects of lipid and sterol metabolism [10]. The activating role of the SREBPs in mammalian cells is facilitated by the action of the CCAAT-binding complex NF-Y.
The discovery that an amino acid alteration P88L within the HapE subunit of the CBC in A. fumigatus leads to resistance to all clinically used azole antifungals provided a potential link between CBC function and ergosterol biosynthesis and suggested the role of the CBC may also be conserved. Indeed in keeping with a role of the CBC in regulation of ergosterol biosynthesis we have demonstrated that the CBC binds the promoters of several genes in the pathway including cyp51A and cyp51B, erg13B and erg7B. Intriguingly, we identified that the CBC binds within the 34 mer in the promoter of cyp51A commonly duplicated in strains that are panazole resistant (TR34 family strains).
If the CBC were a positive regulator of ergosterol biosynthesis this may explain the mechanistic link between regulation of cyp51A and TR34 associated resistance. Our findings however are contrary to this hypothesis and suggest that the CBC of A. fumigatus is playing the opposite role to that of the mammalian orthologue and acting as a negative regulator of sterol biosynthesis. Lack of CBC activity (ΔhapC) leads to derepression of genes in the ergosterol biosynthetic pathway, including genes coding for enzymes of the initial and committed steps of sterol biosynthesis, erg13A, erg13B and hmg1, as well as the azole drug target cyp51A and a concomitant increase in sterol production and tolerance to drugs inhibiting the ergosterol biosynthetic pathway (azoles, allylamines and statins). Additionally, we have shown that the CBC binds efficiently only to the original CGAAT motif (Fig 2, CGAAT -293 ) but not to the duplicated version within TR34 (Fig 2, CGAAT -327 ). The reason for the differential binding affinities at these two sites were not investigated further however several lines of evidence point towards the importance of the bases outside this core motif as being critical. First, the structure of the Aspergillus nidulans CBC in complex with DNA demonstrated that hydrogen bonds between the phosphate groups of the DNA backbone outside the core motif, and the protein main or side chain atoms of HapC and HapE, significantly stabilise the protein:DNA complex and induce DNA bending. Second, alignment of the 23 bp cycA promoter sequence from which this structure was derived with the native and repeat cyp51A promoter sequences revealed different nucleotides at positions at which HapE amino acid residues mediate DNA sequence-independent interactions with phosphate oxygen atoms (S2 Fig). Finally our SPR data clearly indicate that an 8-fold lower CBC affinity to the duplicated TR34 CBC site is due to the presence of eight different nucleotides at the 3´end of the motif. Therefore, our working hypothesis is that altered DNA-bending might be the major cause for the decreased CBC affinity.
We have determined that the azole resistance exhibited by isolates with the HapE P88L modification is linked to an inability of the modified CBC to bind effectively to its recognition site in the cyp51A promoter, leading to increased expression. Interestingly, the growth phenotype exhibited by a strain with the HapE P88L mutation is less severe than that of the HapE null, suggesting only a partial loss of function. One could hypothesise two potential mechanisms for this partial phenotype, 1) the modified CBC retains binding capability for all targets but with reduced efficacy or 2) the modified CBC loses the ability to bind a subset of its targets. These scenarios are not mutually exclusive and more likely than not, a combination of both of these factors are responsible. Interestingly, structural analysis of the closely related A. nidulans CBC indicates that HapE stabilises the protein:DNA complex by binding through residues immediately adjacent to P88, namely L89 to K94 [23]. These residues are thought to interact in a DNA sequence-independent manner. Therefore it is likely that the modified CBC exhibits a general reduction in binding capability across all of its targets, with other factors such as the strength of binding of the CBC to the remaining sections of its binding site, or the presence of stabilising protein interaction partners dictating the strength of binding of the modified CBC.
Our genome-wide binding data revealed that the CBC interacts with the cyp51A promoter during sterol depletion as well as repletion. We found this to be true for almost all the ergosterol biosynthetic genes tested, which indicates that the complex might constitutively bind to these targets. This could suggest a simplistic binary regulatory mechanism at the promoters of sterol regulatory genes with the CBC carrying out a repressing role in the absence of an activator, e.g. SrbA. The regulation of sterol metabolism is however far more complex. We have demonstrated that the regulatory element HapX also has a role in the regulation of cyp51A and azole tolerance. As HapX is known to govern the regulation of genes involved in iron acquisition as well as iron consumption its association with cyp51A, a heme-iron containing enzyme is logical. We propose that HapX enhances or stabilises binding of the CBC at the cyp51A promoter in iron limiting conditions and by doing so provides a more nuanced way of regulating ergosterol biosynthesis in environments that are unfavorable for sterol production. It is possible therefore that further as yet undefined factors will have a role in the regulation of ergosterol biosynthesis.
Our data suggests a role for the CBC in TR34 family mediated azole resistance as its repressing function is not duplicated effectively in the repeat. This alone however, could not explain an increase in cyp51A expression observed in TR strains. Our hypothesis was that a positive regulator must be binding here. We therefore re-evaluated the sequence of the tandem repeat and identified two sequences that matched the consensus of the positive regulator SrbA and which were close to a putative binding site highlighted in a recent ChIP-seq study [27]. We confirmed SrbA interaction to both of these sequences in the 34 mer and cooperative binding at a region on the TR34. This leads to the conclusion that TR34 family driven upregulation of cyp51A is a result of effective duplication of SrbA binding sites in combination with ineffective duplication of the CBC site. It is interesting to note that tandem repeats of transcriptional enhancer elements are known mechanisms of azole resistance in several phytopathogenic moulds [32]. For example Hamamoto et al. identified a 126-bp tandem repeat in the cyp51 5' NTR of the plant pathogen Penicillium digitatum, which caused resistance to Cyp51 inhibitors [33]. Whether resistance in this case is linked to duplication of SREs and/or ineffective duplication of repressor elements is unclear, however we have identified sequences within this repeat that resemble the SrbA binding consensus defined in A. fumigatus [27].
In contrast to A. fumigatus and most other eukaryotes, yeasts such as Saccharomyces cerevisiae and Candida albicans SREBPs have been replaced by Upc2p, a zinc-finger transcription factor which positively regulates sterol synthesis in these species [34]. In C. albicans antifungal azole resistance caused by overexpression of target genes has been extensively studied identifying Upc2p gain-of-function mutations. These types of mutations result in upregulation of the Cyp51 encoding gene and appear to be frequent in azole resistant C. albicans clinical isolates [35]. In A. fumigatus similar mutations in SrbA have not been identified so far. However, our study demonstrates an alternative route to enhance the activity of the positive regulator, SrbA, through increasing its DNA-binding ratio by duplicating its binding site.
The cause of the growth deficiency in the CBC null mutants is difficult to pinpoint exactly. The CBC is a global regulator, its consensus site is present in >30% eukaryotic promoters [22]. In A. nidulans and A. fumigatus several iron metabolic genes are known to be under control of the CBC in complex with HapX [19,24,36]. Additionally the CBC has also been shown to play a prominent role in the regulation of secondary metabolism, development and oxidative stress response [37,38]. These roles and probably the involvement of the CBC in many other biological processes are likely to be the cause for reduced fitness in the mutants. Although the same could be said for the role of the CBC in virulence, its interconnection with the transcription factor HapX is probably more important in this context. HapX mediated transcriptional adaptation to low iron environments has been demonstrated to be crucial for virulence in A.
fumigatus and other pathogenic fungi such as C. albicans, Cryptococcus neoformans and Fusarium oxysporum [36,39,40,41]. Hence, we consider one major reason of the reduced pathogenicity to be a direct result of inefficient adaptation to low iron in the CBC mutant. It is interesting to note that despite a clear attenuation in virulence, an azole resistant CBC loss of function mutant, namely HapE P88L , was the likely cause of death in a chronic granulomatous disease patient [42].
This work provides novel insight into the molecular basis of sterol regulation and TR34 mediated azole drug resistance in A. fumigatus and suggests that if the binding of SrbA could be countered one could reverse drug resistance. The direct targeting of transcriptional binding is fraught with difficulties. Unlike many enzymes transcription factors have large surface areas for protein-protein and protein-DNA interactions which are difficult to disrupt with a small molecule. However modulation of these interactions may be possible using oligonucleotide or peptide therapeutics if these molecules can be effectively delivered into fungal cells [43]. Some, albeit limited, data suggests that this may be possible, at least in the case of oligonucleotides [44] and as significant differences exist between human SREBP and SrbA (less than 40% sequence identity at the DNA binding domain) selective targeting could be achieved. Alternatively, it should be possible to disrupt upstream activators of SrbA, for example the protease RbdA which was recently linked to SrbA cleavage and activation [45]. It is encouraging that a number of small molecules that inhibit Upc2-dependent transcriptional signaling in vivo have been identified in yeast, providing evidence that a similar, unbiased approach may be successful in A. fumigatus [46].

Ethics statement
All animal care procedures were supervised and approved by the Universitat Rovira i Virgili Animal Welfare and Ethics Committee.

Strains, oligonucleotides and growth conditions
Strains and oligonucleotides used in this study are listed in S2 and S3 Tables. Conidia were grown on SAB agar plates. For phenotypic analysis, A. fumigatus liquid cultures were grown at 37°C in AMM according to Pontecorvo et al. [47]. AMM included 1% glucose as carbon source (w/v) and 20 mM ammonium tartrate as nitrogen source. In vitro susceptibility testing of mutant strains was carried out using the EUCAST broth microdilution reference method [26]. Strains are defined in this manuscript as resistant where their MIC exceeds the relevant breakpoint as defined by EUCAST [8].
Generation of hapB, hapC, hapE, srbA, cyp51A deletion mutants and a strain expressing hapE P88L in the CEA10 derivative A1160 Δku80 pyrG+ (A1160P+) Coding sequence of hapB, hapC, hapE, srbA and cyp51A was disrupted in A1160P+. The deletion fragments for homologous recombination were generated using the FusionPCR approach previously described [48]. Around 1 kb of 5' NTR and 1kb 3' NTR were PCR amplified and subsequently linked to an antibiotic resistance cassette via PCR Fusion using primers listed in S3 Table. The P88L causing mutation of HapE was carried out using a similar FusionPCR strategy. hapE coding sequence and 1kb 3' NTR were amplified from wt genomic DNA. The respective amplicons were linked to a hygromycin resistance conferring cassette via FusionPCR using a mutation generating forward primer.

Reconstitution of Δcyp51A, ΔhapC and gfp-tagging of hapC
To complement Δcyp51A and generate strains harboring modified cyp51A promoters, the basic plasmid pcyp51A REC , comprising a pyrithiamine resistance cassette, was generated. The plasmid for the reconstitution of ΔhapC, phapC REC , was generated following the same principle described for pcyp51A REC . The backbone of the hygromycin resistance cassette carrying plasmid pAN7-1 was amplified using primers pAN7-1-hapC-f/pAN7-1-hapC-r. hapC coding sequence plus 1.3 5' NTR and 0.7 3' NTR were amplified with primers hapCrec-f/hapCrec-r. Both fragments were gel-purified and linked via Gibson assembly as described above.
To tag hapC coding sequence at its 3' end with gfp, phapC REC was PCR amplified using primers phapC-GFP-FW/phapC-GFP-RV. The gfp encoding gene was amplified employing primers GFPhapC-FW/GFPhapC-RV using the recently described plasmid pgfpcccA as template [49]. Gel-purified fragments were combined using Gibson assembly yielding phapC GFP .

Generation of cyp51A promoter mutants
Mutation of specific DNA sections was carried out according to the PCR based Q5 site-directed mutagenesis protocol (NEB). pcyp51A REC was used as template DNA. TR34 was introduced into the promoter region employing primers TR34-FW/TR34-RV. Δ34-FW/Δ34-RV were used to delete the 34 mer. After amplification of each construct yielding linearized PCR amplicons, PCR fragments were gel-purified and circularised using the Quick Ligation Kit (NEB). Plasmids carrying cyp51A promoter mutations were designated pcyp51A TR34 and pcyp51A Δ34 .

Fungal transformation
Generally 2 μg of the deletion constructs or 2 μg of the respective plasmids were used for transformation. Prior to transformation pcyp51A REC and promoter mutated plasmids were linearised using BsrGI, phapC REC and phapC GFP were digested with PmlI (S3 Fig). For transformation 1M sucrose was supplemented to AMM or SAB. Depending on the resistance cassette transformed, 0.1 mg/L pyrithiamine (ptrA), 0.2 g/L hygromycin B (hph) or 0.15 g/L zeocin (ble) were used for selection. AMM was used for pyrithiamine based transformation, SAB was used for hygromycin B and zeocin selection at pH 6 and pH 8, respectively.

Sterol analysis: Cultivation and extraction of the cells
Liquid cultures for sterol measurements were grown in AMM for 24 h at 37°C, 200 rpm. Mycelia were harvested through filtration, shock frozen using liquid nitrogen and freeze-dried. The lyophilisate was ground and dissolved in 2M NaOH to obtain a suspension of 3.0 mg/mL. The work-up procedure can be taken from Müller et al. [50]. The residue was dissolved in 800 μL MtBE, 100 μL cholesterol solution (calibration standard, 10 μg/L), and 100 μL of silylation reagent MSTFA/TSIM (9:1) was added. The sample was gently shaken and stored for complete silylation at room temperature for at least 30 min, before being subjected to GC-IT-MS analysis [51,52].
Each sample was prepared in triplicate and measured in duplicate.  Fig). Therefore, an Äkta Explorer system (GE Healthcare) was connected to a miniDawn TREOS static light scattering (SLS) detector equipped with an internal quasi-elastic light scattering (QELS) system in series with an OPTILab T-rEX differential refractometer (Wyatt). Absolute molar mass and hydrodynamic radius (R h ) were determined using the ASTRA 6 software (Wyatt). The CBC and CBC P88L were chromatographed in 20 mM Tris/HCl, 400 mM NaCl, 1 mM DTT pH 7.5 using a Superdex 200 Increase 10/300 GL column (GE Healthcare). The basic region/helix-loop-helix/leucine zipper (bHLHZ) region of A. fumigatus SrbA (amino acids 161-267) was purified following the procedure described in Linde et al. [54]. Real-time SPR protein-DNA interaction measurements were performed by using protocols published previously [19,54].

Fluorescence microscopy
Microscopy images were taken on the Leica TCS SP8 inverted confocal microscope (Leica Microsystems CMS, Mannheim, Germany) using a 63×/1.2 NA objective. Images were captured with LAS AF V3.3. Images were processed using ImageJ and Adobe Photoshop CS6.

Chromatin immunoprecipitation and ChIP-qPCR analysis
For ChIP-qPCR analysis 1x10 6 spores/ml of hapC GFP were grown in 50 mL of AMM for 18 h at 37°C and 200 rpm.
ChIP was performed as previously described by Chung et al. [27] in this work using an Anti-GFP antibody (A-11122, Life technologies) or an anti-IgG control (ab46540, abcam) on Dynabeads Protein A magnetic beads (Thermo Fischer Scientific). Immunoprecipitated DNA was reverse cross-linked, treated with RNase A (Sigma), and then purified using a PCR purification kit (QIAGEN). ChIP'd DNA was eluted with 50 μL of elution buffer and 1 μL of the elution was used for ChIP-qPCR. All ChIP experiments were performed in biological duplicates.
ChIP-qPCR was carried out in an Applied Biosystems 7500 Fast Real-Time PCR System (Thermo Fischer Scientific). Amplification reactions were performed in a final volume of 20 μL using the iTaq Universal SYBR Green Supermix (Bio-Rad) with 0.4 μM forward primer, 0.4 μM reverse primer, and 1 μL of immunoprecipitated DNA. Oligonucleotides used in ChIP-qPCR are listed in S3 Table. Percent input method was applied to analyse the enrichment of the promoter region for each gene and the values were calculated according to the Thermo Fischer Scientific web site (https://www.thermofisher.com/uk/en/home/life-science/ epigenetics-noncoding-rna-research/chromatin-remodeling/chromatin-immunoprecipitationchip/chip-analysis.html). ChIP-qPCR experiments were run in triplicates and the results are presented together with the background signal and standard error.

Virulence study in a murine model of systemic and pulmonary infection
Virulence of wt, ΔhapC and hapC REC was compared in a model of systemic and pulmonary infection, both developed in four-week old OF-1 male mice (Charles River; Criffa SA, Barcelona). For the pulmonary model, groups of 20 animals (10 for survival and 10 for fungal burden studies) were immunosuppressed with 125 mg/kg of cortisone acetate, given intraperitoneally (i.p.), administered four days prior infection and then 3 days per week. Animals were anaesthetized by inhalatory sevofluorane and challenged by nasal instillation (i.n) with conidial suspensions of each strain containing 1x10 5 CFU/animal in a volume of 25 μl. Systemic infection was performed in groups of 16 animals (8 for survival and 8 for tissue burden studies) by intravenous (i.v.) inoculation into the lateral tail vein of 3x10 4 CFU/animal of each strain. Five days after i.v. or i.n. infection, animals included in the tissue burden study were euthanatised by CO 2 anoxia. Liver, lungs, kidneys, spleen and brain from animals challenged i.v. were aseptically removed for CFU determination, while lungs and kidneys were used in the pulmonary model. Approximately, one half of each organ was weighted and mechanically homogenized in 1 to 1.5 mL of PBS. Homogenates were serially 10-fold diluted and placed onto potato dextrose agar plates for CFU/g determination.