Penicillium roqueforti strain CECT 2905 (ATCC 10110) was used in this work. This strain and all the transformants obtained in this work were kept on Potato dextrose agar (Merck, Germany), excepting when andrastin A production was required. In these cases, the strains were grown on YES agar [Bacto Yeast Extract (Difco, USA) 20 g/L, sucrose (Merck, Germany, biochemical grade) 150 g/L and Bacto Agar (Difco, USA) 20 g/L].

DNA from P. roqueforti CECT 2905 was obtained as described before (Gil-Durán et al., 2015). This DNA was used to perform the sequencing of the P. roqueforti CECT 2905 genome, by using Illumina technology. This genome was assembled in several contigs and currently is under annotation (unpublished data). We take advantage of this genome sequence to identify the adr cluster from P. roqueforti CECT 2905 as follows: the adrD gene from P. chrysogenum (7,930 nucleotides) encoding the putative PKS, was used to scan all the contigs from P. roqueforti CECT 2905 by BlastN. One contig containing a gene with very high similarity to adrD (coverage 99%, identity 84%, and E-value closer to zero) was obtained. Then, the vicinities of this gene were compared to the rest of the adr cluster from P. chrysogenum by BlastX, BlastP, and BlastN, revealing the presence of the rest of the adr genes, excepting adrB (see Results). All the genes were manually analyzed, delimited, and annotated using a combination of the Blast tools, multiple alignments by Clustal Omega, and the Translate Tool from Expasy web interface¹.

The nucleotide sequence of the adr gene cluster from P. roqueforti CECT 2905 described in this work has been deposited in the GenBank database under accession number KY349137.

To silence the ten genes of the adr cluster of P. roqueforti, RNA-mediated gene silencing technology was used. The strategy was essentially the same described by Del-Cid et al. (2016). Briefly, a small sequence of each gene was amplified by PCR using suitable primers (Supplementary Table S1). Each amplicon was digested with NcoI and then, ligated into plasmid pJL43-RNAi (Ullán et al., 2008) previously digested with the same restriction enzyme, thus giving rise to ten RNAi-silencing constructs (pJL-RNAi-adrA to pJL-RNAi-adrK, Supplementary Table S1 and Figure S1). These constructs were then used to transform P. roqueforti, exactly as was described before (Gil-Durán et al., 2015).

For RT-qPCR experiments, total RNA was purified as described previously (Gil-Durán et al., 2015) and quantified in a MultiSkan GO quantification system (Thermo Scientific, Germany). Two μg of RNA were used to synthesize cDNA using RevertAid Reverse Transcriptase (Thermo Scientific, Germany). For each RT-qPCR reaction (20 μl) the following conditions were set: 10 μl of KAPA SYBR Fast qRT-PCR Master Mix 2x (Kapa Biosystems, USA), 0.4 μl of each primer (at a concentration of 10 μM each), 0.4 μl de 50x ROX High/Low, 6.8 μl of water and 2 μl of the cDNA previously synthesized. RT-qPCR reactions were carried out in the StepOne Real-Time PCR System (Applied Biosystems, USA). Amplification conditions were 20 s at 95°C and 40 cycles of 3 s at 95°C and 30 s at 50°C. Three replicates were performed for each analysis and suitable negative controls were included. Relative gene expression values were determined by the comparative Ct (ΔΔCt) method using β-tubulin gene expression as a normalization control. The sequences of the primer sets used in RT-qPCR experiments are described in the Supplementary Table S2.

The extraction of andrastin A was done using the same method described by Del-Cid et al. (2016). Briefly, the fungal strains were grown on YES agar for 7 days at 28°C. The selected sample (mycelium or triturated agar) was extracted overnight with 50 mL of an ethyl acetate: dichloromethane: methanol (3:2:1) mixture containing formic acid (1%). After that, the mixtures were sonicated during 30 min and filtered through a 0.45 μm Millex-HV hydrophilic PVDF syringe filter (Merck Millipore). The filtrated was evaporated to dryness in a rotary evaporator, resuspended in 500 μl of methanol (HPLC grade) and submitted to HPLC analysis. The HPLC equipment used consists in a Waters 1525 HPLC system (Waters, Ireland) equipped with a Waters 1525 Binary HPLC pump, a Waters 2996 Photodiode Array (PDA) Detector and a 4.6 × 250 mm (5 μm) SunFire C18 column. HPLC runs were performed as was described by Del-Cid et al. (2016): samples (20 μL) were injected into the HPLC using water (solvent A) and acetonitrile (solvent B), both acidified with 0.02% trifluoroacetic acid. The elution gradient was as follows: 15% solvent B to 68% solvent B linear over 25 min, 68% solvent B to 100% solvent B linear over 2 min, isocratic for 5 min and 100% solvent B to 15% solvent B linear over 2 min. The flow used was 1.2 mL/min and the column was held at 35°C.

Under the HPLC conditions described above, andrastin A was identified by the “spiking” technique. Briefly, known amounts of pure andrastin A (Santa Cruz Biotechnology, Dallas, TX, USA) were added to the samples (spiked samples). As control, non-spiked samples were used. In all the spiked samples, a single peak (with retention time of 24.72 min) increased its size. The increase of the peak size was proportional to the amount of pure andrastin A added. No increase of the peak size was observed in non-spiked samples and no new chromatographic peaks or “shoulder peaks” were seen in the spiked samples. Moreover, the identity of the peak was confirmed by UV-Vis absorption spectrum (200–600 nm). Both the UV-Vis spectrum and the retention time of the peak were identical to those observed for the pure andrastin A.

Finally, the quantity of andrastin A of each sample was obtained from a calibration curve constructed with the pure compound as standard and the UV detector set at 254 nm. The quantity of andrastin A obtained was normalized to the dry weight of the fungal mycelia. For this purpose, the mycelium of each P. roqueforti strain was dried as described before (García-Rico et al., 2009).

To find the adr gene cluster in P. roqueforti CECT 2905, we sequenced and scanned its genome as was described in Material and Methods. As a result, a genomic region of approximately 29.4 kbp containing 10 genes was identified (Figure 1). This region has high similarity (84% overall identity) to the adr gene cluster from P. chrysogenum (Matsuda et al., 2013). Thus, according to the nomenclature of their orthologs in P. chrysogenum, the genes found in the P. roqueforti adr cluster were named adrA, adrC, adrD, adrE, adrF, adrG, adrH, adrI, adrJ, and adrK (Figure 1). At this point, it should be noted that recently, the genome of another P. roqueforti strain (named FM164) was sequenced and annotated (Cheeseman et al., 2014). We also analyzed this genome and as expected, we found the adr cluster. The adr clusters from both P. roqueforti strains have identical size, contain the same 10 genes with identical organization, and are almost identical at nucleotide level (only 75 nucleotide differences in 29,321 total nucleotides, representing 99% overall identity, 100% coverage and no gaps). In Table 1, the correspondence between each adr gene from strain CECT 2905 and the respective ORF from the genome of strain FM164 is described.

Despite several efforts, we did not found any ORF with similarity to the adrB gene from P. chrysogenum in the P. roqueforti genome. Specifically, we used the adrB gene from P. chrysogenum and its deduced protein to exhaustively scan the whole genome of P. roqueforti CECT 2905 using several bioinformatics analysis (BlastN, BlastX, tBlastN) with no positive results. It should be noted that this gene was not found either in the genome of P. roqueforti strain FM164 (Cheeseman et al., 2014), suggesting that its absence is a common fact in P. roqueforti strains.

Recently, it has been suggested that some fungal gene clusters can suffer reorganization processes leading to the total or partial loss of some genes (Martín and Liras, 2016). In accordance with this suggestion, it can be hypothesized that the adrB gene originally present in P. chrysogenum was entirely or partially lost in the adr cluster from P. roqueforti. To address this hypothesis, we take the region comprising the intergenic zone between adrA and adrC in P. roqueforti and we compared it with the syntenic region from the P. chrysogenum genome. A first interesting observation was that the intergenic zone between adrA and adrC in P. roqueforti is 1,444 nucleotides shorter that the syntenic region from P. chrysogenum (1,218 bp vs. 2,662 bp; Figure 2A), supporting the possibility that the adrB gene was total or partially lost in P. roqueforti. Using these regions, we performed several alignments using nucleotide and protein sequences, and we found that the adrB gene in P. roqueforti was partially lost and is found as a residual pseudogene (Figure 2B). In this pseudogene, the first 10 aminoacids almost exactly match to the first 10 aminoacids of the AdrB protein from P. chrysogenum (Figure 2B). However, from this point, a deletion of 49 nucleotides (and other minor insertion/deletion events) changes the sequence and produces in-frame stop codons in all the frame shifts (Figure 2B). Finally, and due to additional four consecutive deletion events in the pseudogene, the similarity between these sequences is entirely lost.

Finally, regarding the deduced proteins encoded by the adr genes from P. roqueforti, Table 1 summarizes the sizes and putative functions of these proteins. As can be observed, all the deduced proteins have high similarity to their respective orthologs from P. chrysogenum (between 83 and 94% identities), suggesting that in P. roqueforti, these proteins should perform the same function assigned in P. chrysogenum.

To test the participation of each gene of the adr cluster from P. roqueforti CECT 2905 in andrastin A biosynthesis, we employed RNA-mediated gene-silencing technology. This technology has been widely used to demonstrate the functionality of several biosynthetic gene clusters in P. roqueforti (Kosalková et al., 2015; Del-Cid et al., 2016; Hidalgo et al., 2016). For this purpose, ten suitable plasmids were constructed and used to transform P. roqueforti CECT 2905 (see Material and Methods for details).

In each transformation event, around 40–50 phleomycin-resistant transformants were obtained. From each transformation event, fifteen transformants were chosen randomly and submitted to preliminary RT-PCR analysis (data not shown). Those two transformants showing the most significant decrease in the mRNA level were selected for further quantification of the down-regulation using RT-qPCR (Figure 2). The results indicate that depending on the gene, the transformants selected exhibited between 1.4- and 10- fold of decrease in mRNA levels compared with the wild-type strain of P. roqueforti (Figure 3), confirming the successful knock-down of all the genes of the adr cluster. Additionally, in all cases the presence of the full silencing cassette was confirmed (Supplementary Figure S1). The transformants selected were used for further analysis.

Previously, the participation of adrI, adrF, adrE, adrJ and adrA genes in the biosynthesis of andrastin A in the fungus P. chrysogenum was experimentally demonstrated (Matsuda et al., 2013). Therefore, as first approach, we tested if their orthologs in P. roqueforti are effectively involved in the andrastin A biosynthesis. For this purpose, RNA-mediated silenced transformants were obtained (see above) and they were used to evaluate the production of andrastin A by HPLC (Figure 4). As control, P. roqueforti CECT 2905 (wild-type strain) and a strain harboring empty plasmid pJL43-RNAi were used.

Silencing of the five genes aforementioned drastically reduced the production of andrastin A by P. roqueforti CECT 2905 (Figure 4). Specifically, transformants with attenuated levels of adrI, adrF, adrE, adrJ and adrA transcripts produced between 14 and 57.7% of the andrastin A produced by the wild-type strain, depending on the gene and the transformant analyzed (Figure 4). These results experimentally confirm the participation of these genes in the production of andrastin A by P. roqueforti.

It has been hypothesized that adrD, adrG, adrK and adrH encode for four enzymes that would act consecutively in the four first putative steps of the andrastin A biosynthesis (Matsuda et al., 2013). However, to the best of our knowledge, the role of these genes has not been experimentally tested yet. Therefore, it was of great interest to address the participation of these genes in the biosynthesis of andrastin A in P. roqueforti.

Silencing of the adrD, adrG, adrK, and adrH genes drastically reduced the production of andrastin A by P. roqueforti CECT 2905 (Figure 4). Specifically, transformants with attenuated levels of adrD, adrG, adrK, and adrH transcripts produced between 8.6 and 56.7 % of the andrastin A produced by the wild-type strain, depending on the gene and the transformant analyzed (Figure 4). These results provide the first experimental support for the participation of adrD, adrG, adrK and adrH in the in vivo production of andrastin A by fungi.

The protein encoded by adrC has high similarity with MFS transporters proteins (Table 1). MSF proteins are ubiquitous membrane proteins that are responsible for the movement of a wide range of substrates across biological membranes (Quistgaard et al., 2016). In the case of genes encoding for MSF proteins located into fungal biosynthetic gene clusters, it has been suggested that these proteins could be involved in the secretion of the secondary metabolites produced (Martín et al., 2005; Martín and Liras, 2016). Therefore, we tested if adrC could influence andrastin A production and/or secretion. Interestingly, the silencing of adrC drastically reduced the production of andrastin A by P. roqueforti CECT 2905 (Figure 3). The transformants with attenuated levels of adrC transcripts produced between 15.2 and 18.7 % of the andrastin A produced by the wild-type strain (Figure 4), suggesting that adrC is necessary for the production of andrastin A.

Regarding a putative role of adrC in andrastin A secretion, we compared the quantity (μg) of the compound found in mycelium and agar medium (Figure 5A). The result, expressed as the percentage of the total andrastin A produced by the strain, indicates that the quantity of andrastin A found in agar and mycelium is similar in the wild type strain and the transformants with attenuated levels of adrC transcripts. The same result was obtained when all the rest of strains with adr genes attenuated were analyzed (Figure 5A). In addition, in order to assess a putative effect of the volume of the sample used (agar or mycelium) we also compared the concentrations (μg/mL) of andrastin A in these samples. Our results indicate that the concentrations of andrastin A in the cells of the transformants with attenuated levels of adrC transcripts are around three times higher than those found in agar (Figure 5B). However, this effect was not specific to adrC and was observed in all the strains, including the wild-type strain (Figure 5B). Taken together, these results suggest that adrC (or any other gene from the adr cluster) has no specific role in andrastin A secretion.