The A. oryzae mcrA homolog was identified by searching for any A. oryzae protein sequences on the NCBI non-redundant protein database using the A. nidulans McrA protein sequence as the query [accession number AN8694 (Oakley et al., 2017)]. Three homologous putative proteins were identified from three separate A. oryzae strains; A. oryzae 3.042 (Zhao et al., 2012), A. oryzae BCC7051 (Thammarongtham et al., 2018), and A. oryzae RIB40 (Machida et al., 2005) (accession numbers: EIT82532.1, OOO05597.1 and XP_023092425.1, respectively). These three predicted proteins were all different lengths (366, 318, and 234 amino acids, respectively). The latter two appear to be truncated at the N-terminus and have an additional 22 amino acid sequence internally (Supplementary Figure 1). Interrogating the genome sequences of each strain and re-annotating each gene manually determined that these apparent differences were due to incorrect annotation. The full protein sequence previously annotated in the genome of strain 3.042 (Zhao et al., 2012) is present and conserved with 100% identity in every strain. This protein sequence shares 62% identity with the A. nidulans McrA with 98% query coverage (Supplementary Figure 2). This is equivalent to that found and reported for other Aspergillus strains (Oakley et al., 2017), suggesting that this protein is also conserved in A. oryzae. A GAL4-like Zn₂Cys₆ binuclear cluster DNA-binding domain (accession number cd00067) was identified from residues 118–157, supporting the previous assertion that McrA is likely to function as a transcription factor, or have a similar DNA binding function.

To disrupt Ao_mcrA, a bipartite strategy (Nielsen et al., 2006) was used. A. oryzae NSAR1 was transformed with the relevant bi-partite knock-out fragments, containing homology to the mcrA flanking regions and the argB cassette for selection, and after multiple rounds of subculturing on selective media, nine of the resulting transformants were analyzed by PCR to evaluate integration of the targeting cassette and subsequent gene disruption (Supplementary Figure 3). Transformant AoΔmcrA-7 passed all diagnostic tests, with PCRs confirming correct integration of both bipartite fragments, and with the transformant being shown to be genetically pure by the absence of any wild-type mcrA (Supplementary Figure 3).

Metabolite production in strain AoΔmcrA-7 was assessed on a range of media; namely CMP, MEB, PDP, GN, and static rice, and compared to the parental strain; A. oryzae NSAR1, to identify any changes in the metabolic profile as a result of the gene disruption. On all media, the metabolic profiles were found to be similar, with no novel metabolites observed. CMP cultures were then used for further analysis, including purification and quantification, as this is the medium routinely employed for heterologous production of compounds in A. oryzae, and supports high yields of crude extract. Both NSAR1 and AoΔmcrA-7 were fermented in CMP media for 1 week in triplicate. Quantitative analysis by LCMS again demonstrated that there were no dramatic changes in the metabolic profile of AoΔmcrA-7 (Figure 2), with the only evident change being an increase in the production of kojic acid from 1.23 to 2.52 g/L (Supplementary Table 1).

Fermentation of ΔmcrA-7 was scaled up to purify minor compounds for NMR analysis and subsequent structural elucidation (Supplementary Figures 4–42). In addition to kojic acid 1, a kojic acid dimer 2 was also shown to be present in the extracts. Five diketopiperazines, 3–7, were also isolated and characterized as the known cyclo-[(_D)Pro-(_L)Tyr] (3), cyclo-[(_D)Pro-(_L)Val] (4), cyclo-[(_D)Pro-(_L)Ile] (5), cyclo-[(_D)Pro-(_L)Leu] (6), and cyclo-[(_D)Pro-(_L)Phe] (7) (Figure 3). Some of these compounds have previously been reported as being isolated from Aspergillus species (Dong et al., 2014; Shaaban et al., 2014; Uchoa et al., 2017). However, a comparison to extractions from CMP medium containing no A. oryzae demonstrated that in this case, these compounds were likely derived from the growth medium and not A. oryzae (Supplementary Figure 43).

To disrupt kojic acid production in strain NSAR1 kojA was mutated using CRISPR-Cas9 technology. KojA encodes an FAD-dependent oxidoreductase shown previously to be necessary for kojic acid production (Terabayashi et al., 2010). A plasmid named pAYargCas9-kojA2 was constructed, containing Cas9 under the control of the inducible P_amyB promoter and a guide RNA cassette consisting of the Cas9-binding scaffold sequence and target-specific guide RNA, placed between the hammerhead (HH) and hepatitis delta virus (HDV) ribozymes. This design was based on that of Nødvig et al. (2015) and the use of ribozymes removes the requirement for an RNA polymerase III promoter. The targeting guide RNA (sequence GTCGCTCGAACAAAGAAACTCGG; complementary to nucleotides 90–109 of kojA) was designed using the online tool WU-CRISPR (Wong et al., 2015), and chosen based on its high predicted potency and 5′ location. An Interpro analysis (Mitchell et al., 2019) of KojA identified an FAD dependent oxidoreductase domain (IPR006076) located at residues 9–319, placing the target sequence toward the N-terminus of this domain. The approach used here is known to generate double-strand breaks (DSBs) at the target locus, with repair via the NHEJ pathway likely to introduce small indel (insertion-deletion) mutations (Nødvig et al., 2015). It was reasoned that any resulting indel mutations at this position, and particularly any out-of-frame mutations, would likely lead to a loss of function. pAYargCas9-kojA2 was also designed to contain the AMA1 autonomously replicating sequence, to allow for the subsequent loss of the plasmid and recycling of the selectable marker.

Plasmid pAYargCas9-kojA2 was used to transform A. oryzae strain NSAR1, generating 42 transformants, which were transferred onto secondary plates lacking arginine. After two rounds of subculturing under such selection, transformants were transferred to KA-Fe medium. This medium contains iron trichloride (FeCl₃), which reacts with kojic acid to produce a distinct red color, thus acting as an initial screen of kojic acid production. One transformant, Ao-ko18, failed to trigger such a color change and was taken forward for further analysis. A ~1 kb region covering the targeted kojA site was amplified from both the original NSAR1 strain and Ao-ko18 and sequenced using primer pair kojA-F/kojA-R. An alignment of the sequences identified an 8 bp deletion in the mutant strain precisely at the target site (Figure 4). This out-of-frame deletion is sufficient for a loss of function for kojA.

To fully confirm a loss of kojic acid production a chemical extraction was performed for Ao-ko18 fermentation cultures and the metabolic profile of the resulting crude extract was analyzed by both TLC and HPLC analysis (Supplementary Figures 44, 45). This confirmed Ao-ko18 as an essentially “empty” strain, producing no major compounds under our laboratory production conditions.

Ao-ko18 was then subcultured multiple times on non-selective medium (MEA), to allow for the loss of the autonomously replicating plasmid pAYargCas9-kojA2. This loss was confirmed by restoration of arginine auxotrophy, and the strain renamed A. oryzae NSARΔK.

NSARΔK, and the original NSAR1 strain were both transformed with the pretenellin A 8 producing plasmid pTYGS-arg-TenC+TenS (Yang et al., 2019), to demonstrate the potential of using NSARΔK as a platform for the heterologous production of secondary metabolites. Two resulting strains, ΔK-Ten120 and N-Ten402, were fermented in production medium to induce the expression of the tenellin genes TenS and TenC. Both strains were confirmed as producing pretenellin A 8 in similar yields. In the case of ΔK-Ten120, the absence of kojic acid meant that the crude extract was almost pure pretenellin A 8 (Figure 5). The presence of kojic acid in large quantities can hinder the purification and downstream use of heterologously produced metabolites, making NSARΔK a valuable platform for the heterologous production of secondary metabolites.

Escherichia coli strain TOP10 (Invitrogen) was used as a host for plasmids and was maintained according to standard procedures (Sambrook and Russell, 2001). Saccharomyces cerevisiae strain YPH499 (Stratagene) was used for plasmid assembly by homologous recombination and was maintained on YPAD medium (1% (w/v) yeast extract, 2% (w/v) peptone, 2% (w/v) glucose, 0.004% (w/v) adenine sulfate, 1.5% (w/v) agar) at 28°C. A. oryzae NSAR1 was obtained as a gift from the Kitamoto group and was maintained on MEA medium (Sigma) at 28°C (Jin et al., 2004).

Transformation of A. oryzae was performed according to the protocol outlined by Williams et al. (2016).

A. oryzae strains were cultured in 10 ml GN medium [2% (w/v) D(+)-glucose monohydrate, 1 % (w/v) Nutrient broth No. 2 (Oxoid, UK)] at 28°C with shaking at 200 rpm, then pelleted, lyophilised and ground under liquid nitrogen. Genomic DNA was prepared using the GenElute Plant Genomic DNA Miniprep kit (Sigma).

Ao_mcrA was deleted from A. oryzae strain NSAR1 using the bipartite method (Nielsen et al., 2006). This method relies on the splitting of the selection marker (with a ~500 bp overlap), which leads to the requirement for homologous recombination between the two selection marker halves for selection to occur. This requirement for homologous recombination leads to higher gene disruption levels.

A plasmid was constructed using homologous recombination in S. cerevisiae to provide a template for the bi-partite fragment amplification (Figure 6). Flanking regions either side of Ao_mcrA were amplified using the proof-reading polymerase KAPA HiFi HotStart (Roche), with the primer pairs AO-mcrAKO-P1F/AO-mcrAKO-P2R and AO-mcrAKO-P3F/AO-mcrAKOP4R. The argB cassette which provides selection in A. oryzae NSAR1 was amplified using primer pair PargB-P5F/TargB-P8R using the plasmid pTYarg as the template (Lazarus et al., 2014). All three fragments were recombined into the E. coli/yeast shuttle plasmid pE-YA (Pahirulzaman et al., 2012), which had been linearised with NotI. The resulting recombinant plasmid, pE-YA-mcrAKO, was used as a template for PCR amplification of the bi-partite fragments using primers AO-mcrAKO-P1F with argB-P6R, and argB-P7F with AO-mcrAKOP4R. Primers used are shown in Table 1.

A kojA mutant strain of NSAR1 was created using CRISPR-Cas9 technology, through the construction and application of plasmid pAYargCas9-kojA2.

Plasmid pAYargCas9-kojA2 was built from the A. oryzae expression plasmid pTYGS-arg (Lazarus et al., 2014) in multiple steps (Figure 7). Firstly, cas9 was recombined into pTYGS-arg by yeast recombination. Cas9 was amplified from plasmid pKWGF2::hCas9 (Nekrasov et al., 2013) in two overlapping fragments, to allow the removal of an internal AscI site, using the primer pairs PamyB-Cas9-F/Cas9-Asc-R and Cas9-Asc-F/Cas9-TamyB-R. These fragments were combined with pTYGS-arg digested with NotI, placing cas9 within the amyB cassette and creating plasmid pTYargCas9. pTYargCas9 was then adapted to contain the AMA1 autonomously replicating sequence to allow for extrachromosomal replication rather than integration, and the subsequent loss of the plasmid to recycle the selectable marker. This was achieved by digesting pTYargCas9 with NheI and inserting AMA1 into this break in two overlapping fragments which were amplified from pFC330 (Nødvig et al., 2015) using the primer pairs AMA1-1F/AMA1-2R and AMA1-2F/AMA1-1R. Again, this was done via yeast recombination, yielding the plasmid pAYargCas9. Finally, the single chimeric guide RNA (sgRNA) was constructed by recombining two PCR products into pAYargCas9 to place them between the gpdA promoter and the enolase terminator. This utilized the design of Nødvig and co-workers (Nødvig et al., 2015) where the sgRNA is embedded within a larger transcript from which it is released by the action of two ribozyme sequences; the 5′-end hammerhead (HH) and 3′-end hepatitis delta virus (HDV), which flank the sgRNA. This allows for the use of standard RNA polymerase II promoters, removing the requirement for an RNA polymerase III promoter. Portions of the sgRNA cassette were amplified from plasmid pFC334 (Nødvig et al., 2015) using the primer pairs Pgpd-kojA2-HH-F/Guide-kojA2-R and HH-kojA2-Guide-F/Teno-HDV-R. These primers introduced the necessary kojA-specific sequences including the protospacer, the sgRNA backbone and the variable HH sequence. The two resulting PCR products were combined with pAYargCas9 that had been digested with AscI as well as a short PCR product (amplified using primer pair ExAdhSeq.F/ExSeqAdh.R) to repair an AscI break site in another cassette and yeast recombination was conducted (Figure 7). This produced the final plasmid; pAYargCas9-kojA2, which was used to transform A. oryzae NSAR1. Primers used in construction of this plasmid are shown in Table 1.

Putative ΔkojA transformants were screened by culturing on KA-Fe solid medium (5% (w/v) glucose, 0.25% (w/v) yeast extract, 0.1% (w/v) K₂HPO₄, 0.05% (w/v) MgSO4.7H₂O, 1 mM FeCl₃, 2% (w/v) agar, pH 5) (Terabayashi et al., 2010). Kojic acid reacts with the ferrous chloride in the medium and causes a clear color change to red. An absence of color change therefore indicates a loss of kojic acid production.

gDNA from nine putative ΔmcrA A. oryzae strains was analyzed using 4 sets of primers (Table 1). The first pair tested for integration of the “left hand” flanking region (primer pair AO-mcrA-test-F/pArgB-Bg-R). These primers should only amplify a product if integration has occurred correctly. The second pair (TargB-Nd-F/AO-mcrA-test-R) tested for correct integration of the “right-hand” flanking region. The third pair (AO-mcrA-F/AO-mcrA-R) tested for the presence of the wild-type Ao_mcrA gene. If the transformants are genetically pure, then no product should be amplified. As final confirmation, primers AO-mcrA-test-F/AO-mcrA-test-R were used to amplify over the whole disrupted region. Primers used are shown in Table 1. Analytical PCR was performed using Biomix Red (Bioline).

gDNA from A. oryzae NSAR1 and the putative ΔkojA transformant was analyzed using the primer pair kojA-F/kojA-R (Table 1) which amplified a ~1 kb product. This was subsequently sequenced to identify the 8 bp deletion in the ΔkojA transformant. Amplification of fragments for sequencing was performed using the high-fidelity KAPA HiFi HotStart (Roche).

Sample solutions (20 μl) were analyzed using a Waters LCMS system comprising of a Phenomenex KINETEX column (2.6 μ, C18, 100 Å, 4.6 × 100 mm) equipped with a Phenomenex Security Guard precolumn (Luna C5 300 Å) eluted at 1 ml/min. Detection was by Waters 2998 Diode Array detector between 200 and 600 nm; Waters 2424 ELSD and Waters SQD-2 mass detector operating simultaneously in ES⁺ and ES⁻ modes between 100 and 800 m/z. Solvent A was HPLC grade water containing 0.05% formic acid, and solvent B was HPLC grade acetonitrile containing 0.05% formic acid. Gradient program was as follows: 5–95% of B over 15 min. 0 min, 5% B; 1 min, 5% B; 2 min, 40% B; 15 min, 95% B; 17 min, 95% B; 18 min, 5% B; 20 min, 5% B; flow rate 1 ml/min.

Instruments used; Varian 400-MR (400 MHz), Varian VNMRS500 (500 MHz), Bruker 500 Cryo (500 MHz), or Varian VNMRS600 Cryo (600 MHz). Chemical shifts (δ) quoted in parts per million (ppm) and coupling constants (J) in Hertz (Hz), rounded to 0.5 Hz intervals. Two-dimensional NMR techniques (HSQC, COZY, HMBC) were used routinely for the assignment of structures.

Strains ΔK-Ten120 and N-Ten402 were cultured on solid CMP medium (3.5% (w/v) Czapek Dox broth, 2% (w/v) maltose, 1% (w/v) peptone, 2% (w/v) agar) for 5 days at 28°C. Subsequently, 1 cm² plugs of agar from the rim and the center of growing colonies were extracted in 2 ml of a solution of ethyl acetate, dichloromethane and methanol (3:2:1) with 1% acetic acid. After 1 h in an ultrasonic bath, the liquid was removed and dried. The crude extract was then dissolved in 200 μl acetonitrile and analyzed by HPLC-MS/UV/ELSD.