The Penicillium brasilianum Histone Deacetylase Clr3 Regulates Secondary Metabolite Production and Tolerance to Oxidative Stress

Most of the biosynthetic gene clusters (BGCs) found in microbes are silent under standard laboratory cultivation conditions due to the lack of expression triggering stimuli, representing a considerable drawback in drug discovery. To access the full biosynthetic potential, studies towards the activation of cryptic BGCs are essential. Histone acetylation status is an important regulator of chromatin structure, which impacts cell physiology and the expression of BGCs. In this study, clr3, a gene encoding a histone deacetylase in Penicillium brasilianum LaBioMMi 136, is deleted and associated phenotypic and metabolic changes are evaluated. The results indicate reduced growth under oxidative stress conditions in the ∆clr3 strain, higher intracellular reactive oxygen species (ROS) levels, and a different transcriptional profile of 13 ROS-related genes of both strains under basal and ROS-induced conditions. Moreover, the production of 14 secondary metabolites, including austin-related meroterpenoids, brasiliamides, verruculogen, penicillic acid, and cyclodepsipeptides was evaluated in the ∆clr3 strain, most of them being reduced. Accordingly, the addition of epigenetic modulators responsible for HDAC inhibition into P. brasilianum’s growth media also culminated in the reduction in secondary metabolite production. The results suggest that Clr3 plays an essential role in secondary metabolite biosynthesis in P. brasilianum, thus offering new strategies for the regulation of natural product synthesis by assessing chromatin modification.


Introduction
Filamentous fungi can produce a vast array of low molecular-weight molecules resulting from their secondary metabolism, aiding the fungus' adaptation to environmental conditions, resisting and fighting back predators and competing microbes in their environmental niches [1]. These natural products possess a range of bioactive activities, ranging from the treatment of infectious diseases to potent toxic and carcinogenic properties [2]. Due to their important pharmaceutical potentialities, many efforts have been devoted in recent decades to identifying genes involved in the biosynthesis and regulation of these compounds [3].

Fungal Strains and Culture Conditions
The fungus isolation procedure from the root bark of Melia azedarach was previously described [9]. P. brasilianum (LaBioMMi 136) [13] was cultivated on commercial potatodextrose-agar (PDA) (Acumedia, San Bernardino, CA, USA) and potato-dextrose broth (PD) (Acumedia). Plates were inoculated with fresh conidia and grown at 30 • C for 7 days in darkness. Conidia were harvested by washing the agar surface with sterile distilled water and diluted to a final concentration of 10 5 or 10 6 conidia.mL −1 .

Genomic DNA Extraction
The extraction of fungal genomic DNA was performed according to the method described by Malavazi and Goldman (2012) [19]. Briefly, conidia were incubated in 50 mL of PD at 25 • C, 150 rpm for 72 h. Mycelia were harvested, ground in liquid nitrogen, and suspended in 500 µL of Lysis buffer (200 mM Tris-HCl, 250 mM NaCl, 25 mM EDTA, 0.5% [w/v] SDS, pH 8.0). Genomic DNA purification was performed by phenol/chloroform extraction, followed by isopropanol precipitation. Purified DNA was air-dried and resuspended in ddH 2 O.

Construction of ∆clr3 Mutant
Construction of the ∆clr3 mutant was performed following the protocol described by Malavazi and Goldman, 2012 [19]. The clr3 deletion cassette used in this study was constructed by in vivo recombination in S. cerevisiae [19]. For the cassette construction, fragments of the 5 and 3 UTR regions that flanked the clr3 gene were amplified from the genomic DNA of the wild-type strain. The primer sequences used in this study are listed in Table S1. Flanking regions contained a small sequence homologous to cloning sites of the pRS426 plasmid. The hph gene, which confers resistance to hygromycin, was PCR-amplified from the pAN7-1 plasmid and used as a selection marker in the deletion cassette. The three independent fragments, and BamHI-EcoRI-cut pRS426, were transformed into the S. cerevisiae FGSC 9721 strain [19]. The plasmids containing the clr3 deletion cassette were isolated using QIAprep Spin Miniprep Kit (QIAGEN, Hilden, NW, DEU) and used as templates to amplify the cassette using the outermost primers (5F and 3R). All PCR amplifications were performed using Phusion Flash High-Fidelity DNA Polymerase (Thermo Scientific, Waltham, MA, USA). The clr3 deletion cassette was transformed into the P. brasilianum wild-type strain LaBioMMi 136 (Table S2) via protoplast transformation, as previously described [19]. Transformants were analyzed by diagnostic PCR and Southern Blot to confirm a single insertion at the clr3 locus.

Southern Blot Analysis
Genomic DNA of the parental P. brasilianum and ∆clr3 strains were isolated, as described above, and PstI-restricted. Chromosomal DNA fragments were separated on 1% agarose gel and blotted onto Hybond N + nylon membranes (GE Healthcare, Chicago, IL, USA), following standard techniques [20,21]. Probe labeling for detection was performed using AlkPhos Direct Labeling and Detection System (GE Healthcare), according to the manufacturer's description. A ChemiDoc™ MP imager (Bio-Rad, Hercules, CA, USA) was used for gel/blot documentation.

Phenotypic Assays for Oxidative Stress Sensibility
To monitor the growth of the ∆clr3 and wild-type strains under oxidative stressing conditions, 1 × 10 4 conidia of each strain were grown in 200 µL of PD in 96-well plates supplemented with varying concentrations of paraquat, menadione, H 2 O 2 , and the crop fungicides fludioxonil and iprodione ( Figure 2) [22]. Plates were incubated for 72 h at 30 • C and photographed.

RNA Extraction and RT-qPCR Analysis
Samples subjected to oxidative stress caused by H 2 O 2 were disrupted, and the total RNA obtained was extracted and processed for cDNA synthesis, as previously described [23,24]. The primers for the individual genes (Table S3) are listed in Table S4. At least three independent biological replicates were used for each condition and the relative fold change for each gene was calculated using comparative cycle threshold (C T ), i.e., ∆∆C T , analysis [25]. All the values were normalized to the expression of the P. brasilianum tub1 gene, an ortholog of the Aspergillus fumigatus tubA gene [21]. Statistical analysis was performed by using one-way analysis of variance (ANOVA) with a Tukey's post hoc test to assess differences in the mutant strain compared to the same condition in the wild-type strain.

ROS Quantification
Reactive oxygen species levels were quantified using the probe CM-H 2 DCFDA (Molecular probes, Invitrogen, Eugene, OR, USA). Both fungal strains were cultivated in minimum media for 24 h in a 96-well plate. Each test condition (different time of exposure to H 2 O 2 ) was repeated in octuplicate. H 2 O 2 was added to growth media to a final concentration of 5 mM before incubation with a 0.25 mM CM-H 2 DCFDA probe for 30 min at 30 • C, and then washed twice with Phosphate-Buffered Saline pH 7.4 (PBS). Fluorescence (485/535 nm) was measured immediately using Cytation TM 5 Imaging Multi-mode Reader (BioTek, Winooski, VT, USA).

Secondary Metabolite Extraction
Cultivation was performed for both wild-type and ∆clr3 strains, as previously described in Section 2.1. After incubation, the Petri dish content, including solid media and the fungal colony, was cut into small pieces and transferred into an Erlenmeyer flask. Extraction was performed using a solvent mixture consisting of methanol, ethyl acetate, and dichloromethane (1:2:3) [26]. Flasks were sonicated for 30 min in an ultrasonic bath and vacuum filtered. The extraction process was repeated twice. The solvent was removed under reduced pressure and the final extract was stored at −20 • C.

UPLC-DAD-MS Analyses
The chromatographic system was an ACQUITY TM UPLC system (Waters, Milford, MA, USA) equipped with a diode array detection system. Waters Acquity UPLC BEH C18 analytical column (50 mm × 2.1 mm, 1.7 µm) was used as the stationary phase. The mobile phase was composed of 0.1% formic acid (A) and acetonitrile (B). Eluent profile (A/B %): 95/5 up to 2/98 within 8 min, maintaining 2/98 for 5 min and down to 95/5 within 1.2 min. Total run time was 18 min for each run and flow rate was 0.2 mL.min −1 . The injection volume was 5 µL. Mass spectrometry detection was performed on a Xevo TQD mass spectrometer (Waters Corp., Milford, MA, USA) with an electrospray ionization (ESI) source. Analyses were performed in positive ion mode with m/z range of 100-1000; capillary voltage at 1.54 kV; and source temperature at 149 • C. MassLynx software (MassLynx, version 4.1, Waters, Milford, MA, USA) was used for data acquisition, equipment control, and spectra analysis.

High-Resolution Mass Spectrometry Analyses
Samples were diluted in methanol. High-resolution mass spectrometry analyses (HPLC-HRMS/MS) were performed in a Thermo Scientific QExactive© Hybrid Quadrupole-Orbitrap Mass Spectrometer. Analyses were performed in positive mode with m/z range of 133.4-2000; capillary voltage at 3.5 kV; source temperature at 250 • C; and S-lens 100 V. The stationary phase was a Thermo Scientific column Accucore C18 2.6 µm (2.1 mm × 100 mm × 1.7 µm). The mobile phase was 0.1% formic acid (A) and acetonitrile (B). Eluent profile (A/B %): 95/5 up to 2/98 within 10 min, maintaining 2/98 for 5 min, down to 95/5 within 1.2 min and maintaining for 8.8 min. Total run time was 25 min for each run and flow rate was 0.2 mL.min −1 . Injection volume was 3 µL. MS/MS was performed by collisioninduced dissociation (CID) with m/z range of 100-800 and the collision energy ranged from 10 to 50 V. MS and MS/MS data were processed with Xcalibur software (version 3.0.63) developed by Thermo Fisher Scientific, Waltham, MA, USA.

Chemical Epigenetic Modulation Experiments
Epigenetic modulation experiments were achieved by using suberoylanilide hydroxamic acid (SAHA) and nicotinamide (NAA) treatments alone, and combined using 48-well microplates. In each well, 1 × 10 5 conidia were inoculated in 1 mL of PD medium containing 100 and 200 µM of NAA or SAHA, respectively. Cultures were incubated for 7 days at 30 • C (70 rpm). Extraction was conducted by liquid-liquid partition after transferring the content of each well to separation funnels using ethyl acetate (3 times of 2 mL). The organic phase was dried under reduced pressure, and the extracts were analyzed by HPLC-HRMS/MS.

Mass Spectrometry Imaging (MSI)
MSI analyses were performed directly on the agar surface using a Prosolia DESI source Modelo Omni Spray 2D ® -3201 coupled to a Thermo Scientific QExactive ® Hybrid Quadrupole-Orbitrap Mass Spectrometer. DESI configuration used was the same set by Angolini et al. [27]. The methanol flow rate was set at 10.0 mL min −1 . MS data were processed with Xcalibur software (version 3.0.63) developed by Thermo Fisher Scientific. IMS data was acquired using a mass resolving power of 70,000 at m/z 200. DESI-MSI data was converted into image files using Firefly data conversion software with a bin width of ∆m/z ± 0.03 (version 2.1.05) and viewed using BioMap software (version 3.8.0.4) developed by Novartis Institutes, Cambridge, MA, USA for Bio Medical Research. Color scaling was adjusted to a fixed value for the comparison between the samples.

∆clr3 Strain Construction and Phenotypic Analysis
To better access the P. brasilianum cryptic natural products, activation of silent BGCs is one of the approaches to dissect the identity of natural products of this organism. Modification of chromatin landscape has been a widely used strategy to achieve metabolic diversity in fungi [28]. Histone deacetylase activity inhibition, either through gene deletion or epigenetic modulation, has presented relevant results in altering fungal secondary metabolism [29,30] and, in some phytopathogenic species, such as Fusarium fujikuroi, virulence properties have been altered [31]. Similarly, the deletion of the hdaA gene in P. chrysogenum resulted in significant expression changes of genes related to pigment production and the upregulation of a sorbicillinoids BGC [32], indicating that HDAC inhibition is a feasible strategy in the Penicillium genera.
Based on the genome annotation in the NCBI database, four HDACs were identified in the P. brasilianum genome. To further identify and analyze each HDAC, Blastp searches and phylogenetic analyses with the amino acid sequences of known HDACs from other fungal species were performed. A phylogenetic tree was constructed using MEGA6 software based on the alignment of amino acid sequences ( Figure S1). Altogether, the following HDAC homologue genes were identified: hosB (PEBR_24088); sir2 (PEBR_32801); clr3 (PEBR_10023); and rpdA/rpd3 (PEBR_38155).
Classical HDACs can be grouped into four classes (I, II, III, and IV) based on the homology of its catalytic domain to yeast HDACs [33]. The gene P. brasilianum PEBR_10023, termed here as clr3, putatively encodes an ortholog of the class II had1 histone deacetylase gene in S. cerevisiae revealing 40% identity and 59% similarity (3.10 −170 e-value), as shown in Figure S1.
Further sequence analyses also indicated that the clr3 gene presented significant sequence identity and similarity with other known HDACs previously characterized in other filamentous fungi, such as the hdaA in A. fumigatus (68% identity, 60% similarity, 0 e-value); A. nidulans (65% identity, 78% similarity, 0 e-value); and P. chrysogenum (69% identity, 81% similarity, 0 e-value) [32,34,35]. The hdaA knockout strains for these species presented different fungal development and metabolic profiles. In P. chrysogenum, hdaA deletion resulted in the transcriptional activation of sorbicillinoids biosynthesis [32]. The A. nidulans ∆hdaA strain presented higher sensitivity to oxidative stressing conditions [34], while the null mutant in A. fumigatus displayed a decreased germination rate and reduced vegetative growth. In addition, A. fumigatus hdaA negatively regulates the expression of four NRPS clusters [35].
To evaluate the impact of HDAC activity in the biology of P. brasilianum and in the secondary metabolism profile, a clr3 deletion strain was constructed. Gene deletion was achieved through homologous recombination using a deletion cassette constructed in vivo in S. cerevisiae harboring the hph gene (hygromycin B phosphotransferase), as a selection marker. Gene deletion strategy and diagnostic PCR are shown in Figure 1A,B. Southern Blot analysis confirmed the clr3 deletion via a single copy integration of the deletion cassette at the clr3 locus ( Figure 1). One of the mutant strains was chosen and further used for phenotypic and the analyses of secondary metabolites. achieved through homologous recombination using a deletion cassette constructed in vivo in S. cerevisiae harboring the hph gene (hygromycin B phosphotransferase), as a selection marker. Gene deletion strategy and diagnostic PCR are shown in Figure 1A,B. Southern Blot analysis confirmed the clr3 deletion via a single copy integration of the deletion cassette at the clr3 locus ( Figure 1). One of the mutant strains was chosen and further used for phenotypic and the analyses of secondary metabolites. Figure 1. Construction of clr3 deletion mutant. Gene replacement strategy for clr3 deletion, in which the hph gene was used as a selection marker. The primer names and annealing regions are indicated by arrows (primer sequences are described in Table S1). Deletion cassettes were constructed by in vivo recombination in S. cerevisiae (A). Diagnostic PCR was performed to evaluate clr3 loci after gene replacement using primers located 500 bp upstream of the deletion cassette, shown in blue letters and arrows. No amplification was observed in the wild-type (WT) strain (B). Southern blot analysis indicating the hybridization of the probe, which recognizes the region indicated by orange letters and lines (C). Phenotype of the Δclr3 and wild-type (WT) strains grown in PDA at 30 °C for 7 days (D), top (left) and bottom (right) of the colony.
After clr3 knockout confirmation, phenotypic assays were performed to evaluate possible changes in fungal physiology in the mutant strain. Our phenotypic analyses Figure 1. Construction of clr3 deletion mutant. Gene replacement strategy for clr3 deletion, in which the hph gene was used as a selection marker. The primer names and annealing regions are indicated by arrows (primer sequences are described in Table S1). Deletion cassettes were constructed by in vivo recombination in S. cerevisiae (A). Diagnostic PCR was performed to evaluate clr3 loci after gene replacement using primers located 500 bp upstream of the deletion cassette, shown in blue letters and arrows. No amplification was observed in the wild-type (WT) strain (B). Southern blot analysis indicating the hybridization of the probe, which recognizes the region indicated by orange letters and lines (C). Phenotype of the ∆clr3 and wild-type (WT) strains grown in PDA at 30 • C for 7 days (D), top (left) and bottom (right) of the colony.
After clr3 knockout confirmation, phenotypic assays were performed to evaluate possible changes in fungal physiology in the mutant strain. Our phenotypic analyses revealed that the ∆clr3 mutant is more sensitive to the oxidative damaging compounds tested, such as H 2 O 2 , paraquat, and menadione ( Figure 2). Interestingly, higher levels of susceptibility were observed in the mutant strain exposed to H 2 O 2 .
revealed that the Δclr3 mutant is more sensitive to the oxidative damaging comp tested, such as H2O2, paraquat, and menadione ( Figure 2). Interestingly, higher le susceptibility were observed in the mutant strain exposed to H2O2. Oxidative stress results from an imbalance between pro-oxidant species a levels of antioxidant defenses, resulting from the generation of reactive oxygen (ROS). In contrast to Aspergillus ssp., minimal data are available on the antioxidat fense system of P. brasilianum. For instance, in A. nidulans, expression analysis re that catalase catB is upregulated upon ROS increase in the wild-type strain, but no ΔhdaA mutant [34], suggesting that chromatin modification is part of the reg mechanism against oxidative stress. As CatB is one of the known enzymes respons detoxifying hydroperoxides in A. nidulans hyphae, the authors hypothesize that tive failure in the positive expression of CatB is one of the main reasons for the sen of ΔhdaΑ strains against ROS in this fungus [36]. Our data suggest that a similar s occurs upon the deletion of clr3 in P. brasilianum, since peroxidases may be the m fected ROS detoxifying enzymes in the Δclr3 mutant, considering the results obtai the H2O2 challenge. In summary, our phenotypic analysis suggested that clr3 wa vant in both conditions where the highly membrane-diffusible H2O2 molecule w ogenously added to the cells and when cells were challenged with paraquat and dione, two compounds that continuously generated superoxide and H2O2 by the cycling in the mitochondria [37].
In both Trichoderma atroviride and A. nidulans, the deletion of histone deace caused reduced growth under oxidative stressing conditions compared to their tive parental strains [32,35,38]. The mechanism underlying oxidative stress resp particularly relevant in phytopathogenic fungi, since most hosts' responses to fun fection are based on ROS production by the plant and counteracting responses fr fungus [38]. Different P. brasilianum strains have already been reported as onion Oxidative stress results from an imbalance between pro-oxidant species and the levels of antioxidant defenses, resulting from the generation of reactive oxygen species (ROS). In contrast to Aspergillus ssp., minimal data are available on the antioxidative defense system of P. brasilianum. For instance, in A. nidulans, expression analysis revealed that catalase catB is upregulated upon ROS increase in the wild-type strain, but not in the ∆hdaA mutant [34], suggesting that chromatin modification is part of the regulatory mechanism against oxidative stress. As CatB is one of the known enzymes responsible for detoxifying hydroperoxides in A. nidulans hyphae, the authors hypothesize that a positive failure in the positive expression of CatB is one of the main reasons for the sensitivity of ∆hdaA strains against ROS in this fungus [36]. Our data suggest that a similar scenario occurs upon the deletion of clr3 in P. brasilianum, since peroxidases may be the most affected ROS detoxifying enzymes in the ∆clr3 mutant, considering the results obtained for the H 2 O 2 challenge. In summary, our phenotypic analysis suggested that clr3 was relevant in both conditions where the highly membrane-diffusible H 2 O 2 molecule was exogenously added to the cells and when cells were challenged with paraquat and menadione, two compounds that continuously generated superoxide and H 2 O 2 by the redox cycling in the mitochondria [37].
In both Trichoderma atroviride and A. nidulans, the deletion of histone deacetylases caused reduced growth under oxidative stressing conditions compared to their respective parental strains [32,35,38]. The mechanism underlying oxidative stress response is particularly relevant in phytopathogenic fungi, since most hosts' responses to fungal infection are based on ROS production by the plant and counteracting responses from the fungus [38]. Different P. brasilianum strains have already been reported as onion (Allium cepa L.) pathogens [39], but little is known about this host-pathogen interaction and the virulence attributes of this fungus, which may involve the function of clr3.

The Transcriptional Basis for P. brasilianum ∆clr3 Sensitivity to Hydrogen Peroxide
The P. brasilianum ∆clr3 strain was shown to be more sensitive to H 2 O 2 ( Figure 2A). Therefore, an RT-qPCR approach was used to characterize the transcriptional basis of the oxidative stress sensitivity of the ∆clr3 mutant ( Figure 3). virulence attributes of this fungus, which may involve the function of clr3.

The Transcriptional Basis for P. brasilianum Δclr3 Sensitivity to Hydrogen Peroxide
The P. brasilianum Δclr3 strain was shown to be more sensitive to H2O2 (Figure 2A). Therefore, an RT-qPCR approach was used to characterize the transcriptional basis of the oxidative stress sensitivity of the Δclr3 mutant ( Figure 3). About 30-fold increase in clr3 expression was observed in the wild-type strain after 60 min of exposure to 5 mM H2O2, indicating that clr3 plays an important role in positively regulating the antioxidant defense system of P. brasilianum against H2O2. Similar results were observed in T. atroviride, in which the expression of the histone deacetylase-encoding gene hda-2 was also increased under the influence of H2O2 and menadione [40].
Several homolog genes identified in A. fumigatus as being involved in oxidative stress tolerance were identified by Blastp searches in the P. brasilianum genome (Table  S4), and sequences were used to design primers for the detection of mRNA abundance in response to H2O2 (5 mM) in the wild-type and Δclr3 mutant.
Peroxiredoxins (Prxs) are cysteine-based peroxidases containing one (1-Cys) or two (2-Cys) catalytic cysteine residues that contribute for maintenance of intracellular peroxide levels and hydroperoxide removal under various subcellular locations [41]. The P. brasilianum genome possesses two 1-Cys Prxs, PEBR_28450 and PEBR_20685; three genes encoding putative enzymes from the PRX5-subfamily, PEBR_37093, PEBR_01141, and PEBR_39037; as well as a glutathione peroxidase: PEBR_14961. We named these genes prx1, prx2, prxA, prxB, prxC, and hyr1, respectively. Protein products for the orthologs of prx1 and prxA in A. fumigatus (Table S4) have been reported as being differently synthesized after fungal incubation with 2 mM of H2O2, indicating an associated role in oxidative stress response [42]. About 30-fold increase in clr3 expression was observed in the wild-type strain after 60 min of exposure to 5 mM H 2 O 2 , indicating that clr3 plays an important role in positively regulating the antioxidant defense system of P. brasilianum against H 2 O 2 . Similar results were observed in T. atroviride, in which the expression of the histone deacetylase-encoding gene hda-2 was also increased under the influence of H 2 O 2 and menadione [40].
Several homolog genes identified in A. fumigatus as being involved in oxidative stress tolerance were identified by Blastp searches in the P. brasilianum genome (Table S4), and sequences were used to design primers for the detection of mRNA abundance in response to H 2 O 2 (5 mM) in the wild-type and ∆clr3 mutant.
Peroxiredoxins (Prxs) are cysteine-based peroxidases containing one (1-Cys) or two (2-Cys) catalytic cysteine residues that contribute for maintenance of intracellular peroxide levels and hydroperoxide removal under various subcellular locations [41]. The P. brasilianum genome possesses two 1-Cys Prxs, PEBR_28450 and PEBR_20685; three genes encoding putative enzymes from the PRX5-subfamily, PEBR_37093, PEBR_01141, and PEBR_39037; as well as a glutathione peroxidase: PEBR_14961. We named these genes prx1, prx2, prxA, prxB, prxC, and hyr1, respectively. Protein products for the orthologs of prx1 and prxA in A. fumigatus (Table S4) have been reported as being differently synthesized after fungal incubation with 2 mM of H 2 O 2 , indicating an associated role in oxidative stress response [42].
Superoxide dismutase (Sods) play an essential role in the detoxification of highly reactive superoxide anions [43]. Sods representatives in the P. brasilianum's genome, including one copper-zinc (PEBR_25708), one manganese (PEBR_04850), and one iron (PEBR_29521) Sod, were identified and named sod1, sod2, and sod3, respectively. The deletion of orthologs for these genes in A. fumigatus led to high sensibility to superoxide ions produced by menadione [43]. Lastly, our searches for catalases identified four genes, PEBR_36641; PEBR_24619; PEBR_24196; and PEBR_36037, named as cta1, cta2, catA, and catB, respectively. Deletion mutants for the respective orthologs of these genes in A. fumigatus also presented higher sensitivity to H 2 O 2 [44].
The expression analyses of the above-selected oxidative stress response genes indicate a broad range of results in response to H 2 O 2 treatment (Figure 4). Both up and downregu-lated genes were identified across different enzyme families, through different time points. To better address our findings, in silico predictors of protein subcellular localization were utilized to look for patterns between enzyme localization and mRNA abundance (Table S3).
The expression analyses of the above-selected oxidative stress response genes indicate a broad range of results in response to H2O2 treatment (Figure 4). Both up and downregulated genes were identified across different enzyme families, through different time points. To better address our findings, in silico predictors of protein subcellular localization were utilized to look for patterns between enzyme localization and mRNA abundance (Table S3).  Figure S2). In silico prediction of subcellular location for each enzyme is also shown on the left of the heat map.
In silico predictions indicated that the ROS detoxifying enzymes analyzed here reside on cytoplasm, mitochondrion, peroxisome, and one of them secreted to the extracellular environment. However, the hierarchical clustering did not return a clear correlation between the mRNA levels of each enzyme and their subcellular localization, thus indicating a complex antioxidant regulation system and a potential regulatory role of clr3 upon different classes of antioxidant enzymes. Both enzymes, encoded by catB and cta1, respectively, were downregulated in the Δclr3 strain across all time points, as well as prxB and sod2. Based on our findings, we hypothesize that clr3 is an important regulator for these genes. Whether the control exerted by clr3 on these genes is direct or indirect remains to be determined.
Interestingly, clr3 negatively controls the expression of some other antioxidants, such as the cytoplasmic hyr1, catA, prxC, and mitochondrial sod1, which presented increased mRNA abundance in the deletion strain, suggesting that the lack of Clr3 activity triggers a compensatory effect to maintain ROS homeostasis, especially in cytosol. In the mitochondria, cta2 and prx2 activity was upregulated during the first 15 min of cell ex-  Figure S2). In silico prediction of subcellular location for each enzyme is also shown on the left of the heat map.
In silico predictions indicated that the ROS detoxifying enzymes analyzed here reside on cytoplasm, mitochondrion, peroxisome, and one of them secreted to the extracellular environment. However, the hierarchical clustering did not return a clear correlation between the mRNA levels of each enzyme and their subcellular localization, thus indicating a complex antioxidant regulation system and a potential regulatory role of clr3 upon different classes of antioxidant enzymes. Both enzymes, encoded by catB and cta1, respectively, were downregulated in the ∆clr3 strain across all time points, as well as prxB and sod2. Based on our findings, we hypothesize that clr3 is an important regulator for these genes. Whether the control exerted by clr3 on these genes is direct or indirect remains to be determined.
Interestingly, clr3 negatively controls the expression of some other antioxidants, such as the cytoplasmic hyr1, catA, prxC, and mitochondrial sod1, which presented increased mRNA abundance in the deletion strain, suggesting that the lack of Clr3 activity triggers a compensatory effect to maintain ROS homeostasis, especially in cytosol. In the mitochondria, cta2 and prx2 activity was upregulated during the first 15 min of cell exposure to H 2 O 2 and downregulated in later stages of exposure, implying their important role in the early stages of ROS accumulation. For the later time points, cytoplasmic glutathione peroxidase hyr1 and catA mRNA levels were increased in the null mutant after 30 min of H 2 O 2 incubation.
Interestingly, prxC, sod1, cta2, prx2, cta1, catB, prx1, prxA, sod3, and sod2 presented different expression levels in the mutant strain, even under non stressing conditions. Tribus and colleagues addressed a similar finding in their evaluation on the effect of HdaA deletion in A. nidulans growth under oxidative stress [34]. In their study, ∆hdaA strain grown in the absence of ROS-generating drugs presented higher levels of protein carbonylation, which is a measure of the overall oxidative damage within the cells [45].
To evaluate the intracellular levels of ROS in the P. brasilianum ∆clr3 strain, we fluorometrically monitored the ROS formation in the wild-type and mutant strains through the oxidative conversion of non-fluorescent CM-H 2 DCFDA into fluorescent CM-DCF ( Figure 5). different expression levels in the mutant strain, even under non stressing conditions. Tribus and colleagues addressed a similar finding in their evaluation on the effect of HdaA deletion in A. nidulans growth under oxidative stress [34]. In their study, ΔhdaA strain grown in the absence of ROS-generating drugs presented higher levels of protein carbonylation, which is a measure of the overall oxidative damage within the cells [45].
To evaluate the intracellular levels of ROS in the P. brasilianum Δclr3 strain, we fluorometrically monitored the ROS formation in the wild-type and mutant strains through the oxidative conversion of non-fluorescent CM-H2DCFDA into fluorescent CM-DCF ( Figure 5). As predicted in our qPCR analyses, intracellular ROS levels in the absence of H2O2 were higher in the null mutant than in the wild-type strain. Consistently, the expressions of sod2, cta1, and catB were significantly downregulated under basal conditions. The overall increase in the transcription of hyr1, catA, prxC, and sod1 after 30 min (Figure 4) was not sufficient to lower the ROS levels in the Δclr3 strain, again suggesting multiple regulatory mechanisms, which are possibly deficient in the mutant.

Natural Product Diversity in Δclr3 Strain
To further probe the contribution of clr3 in P. brasilianum physiology, we subsequently used the Δclr3 mutant to investigate the secondary metabolite production as an initial approach to natural product discovery in this species. For metabolic profile comparison, wild-type and Δclr3 strains were grown in identical cultivation conditions and As predicted in our qPCR analyses, intracellular ROS levels in the absence of H 2 O 2 were higher in the null mutant than in the wild-type strain. Consistently, the expressions of sod2, cta1, and catB were significantly downregulated under basal conditions. The overall increase in the transcription of hyr1, catA, prxC, and sod1 after 30 min (Figure 4) was not sufficient to lower the ROS levels in the ∆clr3 strain, again suggesting multiple regulatory mechanisms, which are possibly deficient in the mutant.

Natural Product Diversity in ∆clr3 Strain
To further probe the contribution of clr3 in P. brasilianum physiology, we subsequently used the ∆clr3 mutant to investigate the secondary metabolite production as an initial approach to natural product discovery in this species. For metabolic profile comparison, wild-type and ∆clr3 strains were grown in identical cultivation conditions and crude extracts were analyzed by UPLC-DAD. The resulting chromatograms ( Figure S3) were plotted to the same scale for a better comparison.
We observed no alteration in the chromatogram profile between the two strains, indicating that the deletion of clr3 did not induce the production of new metabolites or repression of those constitutively formed in our culture conditions. On the other hand, peak areas were significantly different for both strains, indicating that clr3 has an important regulatory role in secondary metabolite production.
To identify the molecules with different production levels in both strains, crude extracts were analyzed further by HPLC-HRMS/MS, and natural product dereplication was performed by manually searching on Natural Products databases. Furthermore, the obtained data were compared to HRMS data from previous related studies on P. brasilianum secondary metabolism [6][7][8][9][10][11][12]32,[46][47][48][49]. Interestingly, a total of 15 differentially produced compounds already known to be produced by P. brasilianum were putatively identified. HRESI-MS data for all metabolites described can be found in Table 1, Figure 6, and Figures S4-S17. Austin-related meroterpenoids numbered (Table 1) have previously been related to both Aspergillus and Penicillium genera, presenting various biological activities, such as convulsive and insecticidal [8,9,36,46]. Brasiliamides are rare examples of fungal phenylpropanoids produced by P. brasilianum, possessing convulsive and a weak antibacterial activity [7,8]. Verruculogen is a common tremorgenic mycotoxin also produced by both the Aspergillus and Penicillium genera [9,47]. Furthermore, verruculogen derivatives, such as verruculogen TR-2, have also had their production induced in P. brasilianum [12]. Penicillic acid is a polyketide produced by many strains of Penicillium fungi, being an important mycotoxin with antibacterial activity [50]. Cyclodepsipeptides JBIR 114 and JBIR 115 present a unique structure with three neighboring cyclic amino acids, one proline and two pipecolinic acids, indicating a great NRPS-like enzymatic potential in P. brasilianum [51].
The lack of novel-induced secondary metabolites due to clr3 deletion was unexpected based on its transcription role to suppress gene expression and may be highly related to our cultivation condition, again suggesting the complexity of secondary metabolite production. Nonetheless, there are several reports in the literature demonstrating that the perturbation of HDACs activity has led to both the up-and downregulation of a number of BGCs, the suppression of a metabolite, or the induction of new molecules, thus indicating a complex response mechanism to chromatin landscape modification in natural product biosynthesis [50][51][52].
Since the deletion of clr3 did not induce the production of any new metabolites, different approaches to modulate chromatin structure were sought. In addition to gene deletion, HDAC inhibition can be achieved through chemical epigenetic modulation [28,53]. In Aspergillus niger, growth in media supplemented by suberoylanilide hydroxamic acid (SAHA), a HDAC inhibitor, was able to alter its secondary metabolism and induce a new pyridine [54]. In the Penicillium genera, the same strategy was applied in Penicillium mallochii, resulting in two new natural sclerotioramine derivatives [55].
In this study, two chemical epigenetic modulators were used, NAA and SAHA, as well as a mixture of both. Extracts from the fungus grown in the presence of these epigenetic modulators were analyzed through HPLC-HRMS/MS. Previously dereplicated compounds were also detected in these experiments, although production levels varied through the different modulators. Austin-related meroterpenoids numbered (Table 1) have previously been related to both Aspergillus and Penicillium genera, presenting various biological activities, such as convulsive and insecticidal [8,9,36,46]. Brasiliamides are rare examples of fungal phenylpropanoids produced by P. brasilianum, possessing convulsive and a weak antibacterial activity [7,8]. Verruculogen is a common tremorgenic mycotoxin also produced by both the Aspergillus and Penicillium genera [9,47]. Furthermore, verruculogen derivatives, such as verruculogen TR-2, have also had their production induced in P. brasilianum [12]. Penicillic acid is a polyketide produced by many strains of Penicillium fungi, being an important mycotoxin with antibacterial activity [50]. Cyclodepsipeptides JBIR 114 and JBIR 115 present a unique structure with three neighboring cyclic amino acids, one proline and two pipecolinic acids, indicating a great NRPS-like enzymatic potential in P. brasilianum [51].
The lack of novel-induced secondary metabolites due to clr3 deletion was unexpected based on its transcription role to suppress gene expression and may be highly re-

The Regulatory Role of HDACs in Penicillium brasilianum Secondary Metabolism
To better understand the magnitude of HDAC regulation of the secondary metabolism in P. brasilianum, and to compare HDAC inhibition through clr3 deletion and chemical epigenetic modulation, metabolite quantification was performed. Relative quantification of the metabolites was conducted by comparing integrated peak areas of the identified metabolites in the HPLC-HRMS/MS analyses of the crude extracts from both strains (Figure 7), as well as the ones obtained from the chemical epigenetic modulation experiment (Figure 8). lism in P. brasilianum, and to compare HDAC inhibition through clr3 deletion and chemical epigenetic modulation, metabolite quantification was performed. Relative quantification of the metabolites was conducted by comparing integrated peak areas of the identified metabolites in the HPLC-HRMS/MS analyses of the crude extracts from both strains (Figure 7), as well as the ones obtained from the chemical epigenetic modulation experiment ( Figure 8).  By comparing the secondary metabolism of both wild-type and ∆clr3 strains, significant production level differences can be noted. Notably, brasiliamide A was the most produced amide in both extracts; however, its biological role has not yet been fully unveiled. This is the first report in which brasiliamide production has been evaluated upon HDAC-inhibition conditions.
Similarly, the deletion of SntB, a global histone deacetylase inhibitor in A. flavus, resulted in the downregulation of the flavotoxin BGC [56]. In our study, HDAC inhibition also caused the downregulation of mycotoxins, such as verruculogen and penicillic acid. On the other hand, isoaustinone, brasiliamide D, verruculogen TR-2, JBIR 114, and JBIR 115 presented remarkably similar production levels, indicating a low correlation between Clr3 activity and the biosynthesis of these natural products.
In A. fumigatus, wild-type, ∆hdaA mutants and the over-expression of ∆hdaAcomplement strains exhibited significant differences in their secondary metabolism profiles, which also resulted in altered virulence properties [35]. Ethyl acetate extracts from each strain were added to macrophages and the hdaA over-expression strain induced cell death at equivalent biomass [35]. Albeit the metabolism of the P. brasilianum ∆clr3 strain was not evaluated in vivo, the metabolic changes observed here might alter its endophytic and pathogenic properties. Further studies are necessary to evaluate this proposition.
Regarding the chemical HDAC modulators, the most significant differences were observed in the verruculogen TR-2 concentration, in which the mixture of modulators caused a total suppression of its production. Penicillic acid and brasiliamide A also presented significant differences, especially under the combination of SAHA and NAA. By comparing the secondary metabolism of both wild-type and Δclr3 strains, significant production level differences can be noted. Notably, brasiliamide A was the most produced amide in both extracts; however, its biological role has not yet been fully unveiled. This is the first report in which brasiliamide production has been evaluated upon HDAC-inhibition conditions.
Similarly, the deletion of SntB, a global histone deacetylase inhibitor in A. flavus, resulted in the downregulation of the flavotoxin BGC [56]. In our study, HDAC inhibition also caused the downregulation of mycotoxins, such as verruculogen and penicillic acid. On the other hand, isoaustinone, brasiliamide D, verruculogen TR-2, JBIR 114, and JBIR 115 presented remarkably similar production levels, indicating a low correlation between Clr3 activity and the biosynthesis of these natural products.
In A. fumigatus, wild-type, ΔhdaA mutants and the over-expression of ΔhdaA-complement strains exhibited significant differences in their secondary metabolism profiles, which also resulted in altered virulence properties [35]. Ethyl acetate extracts from each strain were added to macrophages and the hdaA over-expression strain As an alternative approach to validate these findings, Mass Spectrometry Imaging (MSI) was used to confirm the metabolic production differences of brasiliamide A and verruculogen ( Figures S18 and S19), the two most produced metabolites in P. brasilianum in this study (Figure 9), as well as to monitor the spatial distribution of both molecules in the fungal colony.
DESI-MSI analyses of wild-type and ∆clr3 strains indicated a lower accumulation of brasiliamide A and verruculogen in the colony surface of ∆clr3 than in the wild type. Additionally, the DESI-MSI images indicate that the production of these secondary metabolites occurs in both the conidia and hyphal tissue since the detection occurred throughout the colony, including areas of lower conidiation.
Based on the metabolomic approaches, it was possible to verify a close relation between Clr3 activity and secondary metabolite production in P. brasilianum. Both the deletion of clr3 and chemical epigenetic manipulation led to a downregulation of secondary metabolite production, indicating that histone deacetylases play an important role in regulating the P. brasilianum secondary metabolism regulation. sented significant differences, especially under the combination of SAHA and NAA.
As an alternative approach to validate these findings, Mass Spectrometry Imaging (MSI) was used to confirm the metabolic production differences of brasiliamide A and verruculogen ( Figures S18 and S19), the two most produced metabolites in P. brasilianum in this study (Figure 9), as well as to monitor the spatial distribution of both molecules in the fungal colony. Images are plotted on the same color scale from 0 (black) to 2 × 10 4 (red); ion concentration cannot be compared across images due to ionization differences between molecules.
DESI-MSI analyses of wild-type and Δclr3 strains indicated a lower accumulation of brasiliamide A and verruculogen in the colony surface of Δclr3 than in the wild type. Additionally, the DESI-MSI images indicate that the production of these secondary metabolites occurs in both the conidia and hyphal tissue since the detection occurred throughout the colony, including areas of lower conidiation.
Based on the metabolomic approaches, it was possible to verify a close relation between Clr3 activity and secondary metabolite production in P. brasilianum. Both the deletion of clr3 and chemical epigenetic manipulation led to a downregulation of secondary metabolite production, indicating that histone deacetylases play an important role in regulating the P. brasilianum secondary metabolism regulation.

Conclusions
Understanding filamentous fungi secondary metabolism and its regulation by chromatin structure is an important step towards natural product discovery. Here, we demonstrated for the first time the effect of HDAC inhibition on P. brasilianum's development and secondary metabolite production. In terms of fungal development, the Δclr3 Images are plotted on the same color scale from 0 (black) to 2 × 10 4 (red); ion concentration cannot be compared across images due to ionization differences between molecules.

Conclusions
Understanding filamentous fungi secondary metabolism and its regulation by chromatin structure is an important step towards natural product discovery. Here, we demonstrated for the first time the effect of HDAC inhibition on P. brasilianum's development and secondary metabolite production. In terms of fungal development, the ∆clr3 strain exhibited sensitivity in growth under oxidative stress conditions, higher ROS levels in basal and ROS-induced conditions, presenting different transcript levels for clr3 and crucial ROS-related genes compared to the parental strain. Few genes were also upregulated in the null mutant, likely compensating for the lack of clr3 activity. Based on the metabolic approaches, both the deletion of clr3 and chemical inhibition of HDACs caused the reduction in the production of secondary metabolites, such as austin-related meroterpenoids, brasiliamides, verruculogen, penicillic acid, and cyclodepsipeptides. Finally, the paradoxical relationship between higher histone acetylation status and decrease in natural product biosynthesis might be related to the fungal requirements to direct cell energy sources to ROS detoxification, rather than eliciting secondary metabolite pathways, at least in in vitro conditions [57].
Supplementary Materials: The following supporting information can be downloaded at: https: //www.mdpi.com/article/10.3390/jof8050514/s1, Figure S1: Neighbor-joining phylogenetic tree of HDACs; Figure S2: Relative expression of gene-encoding regulators of oxidative stress are different in the ∆clr3 mutant when compared to wild type; Figure S3: UPLC-DAD chromatograms obtained for the crude extracts from (A) ∆clr3 and (B) wild-type strains of P. brasilianum; Figure S4: HRESI-MS data for isoaustinone (1); Figure S5 Table S1: Primers used in this study for construction of ∆clr3 strain; Table S2: Penicillium brasilianum strains used in this study; Table S3: Individual genes selected for RT-qPCR analyses based on the orthology with oxidative stress regulation genes in A. fumigatus; Table S4: Primers used in this study for RT-qPCR analyses.