Duplication and Functional Divergence of Branched-Chain Amino Acid Biosynthesis Genes in Aspergillus nidulans

The penultimate step in leucine biosynthesis is catalyzed by β-IPM dehydrogenase (Fig. 1). A single gene in yeast, LEU2, encodes β-IPM dehydrogenase (35, 36), whereas in A. niger, two enzymes, Leu2A and Leu2B, encoded by separate genes, carry out this role (18). Two A. nidulans β-IPM dehydrogenase enzymes, encoded by AN0912 and AN2793, were identified in BLASTp searches with S. cerevisiae Leu2p as the query. AN0912 and AN2793 showed high levels of similarity and identity with Leu2p, Leu2A, and Leu2B, with AN0912 most similar to Leu2A and AN2793 most similar to Leu2B (Table 1). AN0912 and AN2793 showed 50.5% identity and 67.3% similarity with each other. Alignment of the five proteins revealed strong conservation throughout the protein, including in the substrate binding loop and NAD binding motif (see Fig. S1 in the supplemental material). We designated AN0912 leuD and AN2793 leuE. leuD is found on chromosome VIII in a region of highly conserved gene colinearity in all 27 Aspergillus species genomes available at FungiDB (Fig. S2A). In contrast, leuE is located on chromosome VI and lacks colinearity with its 24 orthologs in the 27 Aspergillus species (Fig. S2B).

We investigated the relationships of the two A. nidulans β-IPM dehydrogenase genes through construction of a phylogenetic tree (Fig. 2). LeuD and LeuE formed distinct clades with their respective Aspergillus orthologs. The LeuD clade is consistent with the position of A. nidulans in the fungal evolutionary tree (37), whereas the LeuE clade lies between the Ascomycota and the Basidiomycota clades.

To determine whether leuD and leuE are functional genes, we generated deletion mutants by gene replacement (Fig. S3A; see Materials and Methods). Deletion of genes required for leucine biosynthesis results in leucine auxotrophy (19, 27), yet neither leuDΔ nor leuEΔ strain conferred strict leucine auxotrophy (Fig. 3A). However, while the leuEΔ strain grew similarly to the wild type in the absence of leucine, the leuDΔ mutant showed reduced growth compared with the wild type unless supplemented with exogenous leucine. Transformation of the leuD gene into the leuDΔ mutant restored leucine prototrophy (Fig. S4A). To determine whether the leaky nature of the leuDΔ leucine auxotrophy resulted from LeuE activity, we constructed a leuDΔ leuEΔ double mutant by meiotic crossing and found that the double mutant was a strict auxotroph, showing growth only when supplemented with exogenous leucine (Fig. 3A). Leucine supplementation of a C. neoformans auxotroph lacking α-IPM isomerase is possible when glutamine or asparagine, but not ammonium, is the nitrogen source (6). In contrast, the leuCΔ mutant lacking α-IPM isomerase can be supplemented on ammonium (27). Likewise, the leuDΔ leuEΔ leucine auxotrophy could be supplemented on the preferred nitrogen sources ammonium and glutamine and on the alternative nitrogen source nitrate (Fig. 3A). Therefore, regulation of leucine uptake in A. nidulans is not regulated by nitrogen metabolite repression.

To complement the tight leucine auxotrophy of the leuDΔ leuEΔ double mutant, we introduced a plasmid carrying the wild-type leuE gene and directly selected transformants in the absence of leucine (Fig. S4B to D). Single-copy integration conferred partial leucine auxotrophy that resembled the leuDΔ single mutant, whereas multicopy transformants showed stronger growth, indicating that additional copies of the leuE gene partially suppress the leuDΔ phenotype. We next considered whether levels of expression were the source of the different degrees of effect of leuDΔ and leuEΔ. We found, using reverse transcription-quantitative PCR (RT-qPCR), that leuD had ∼64-fold higher expression than leuE after 16 h of growth in 10 mM ammonium-minimal medium. In transcriptome sequencing (RNA-seq) data from wild-type mycelia, leuD showed higher expression than leuE when grown on ammonium (35-fold), alanine (12-fold), and glutamine (13-fold) (Fig. 3B). As leucine production is regulated by feedback inhibition, we examined the effect of the leuDΔ mutation on expression of leuE and two other leucine biosynthesis genes, luA and leuC, by RT-qPCR, and gdhA, which is coregulated with leucine biosynthesis, using enzyme activity of LacZ expressed from the gdhA-lacZ translational fusion reporter gene (19, 27). For all three leucine biosynthesis genes, and for gdhA-lacZ, we found that leuDΔ resulted in increased expression over wild-type levels (Fig. 3C and D). Therefore, reduced leucine production as a result of leuDΔ results in compensation by upregulation of leuE and the other leucine biosynthesis genes as well as gdhA.

As leuEΔ had no effect on growth and leuE upregulation in the leuDΔ deletion mutant is expected to be LeuB dependent, we constructed a leuBΔ leuDΔ double mutant (Fig. 3A). In contrast to the leuBΔ and leuDΔ single mutants, which are leaky leucine auxotrophs, the leuBΔ leuDΔ double mutant is a strict leucine auxotroph, suggesting that LeuB regulation of leuE is required for leucine biosynthesis in the absence of leuD. We assayed gdhA-lacZ reporter gene expression in the double mutant (Fig. 3D). Unlike the single leuDΔ mutant, there was no increase in expression above leuBΔ levels in the double mutant, consistent with the leuDΔ-induced upregulation of leucine biosynthesis genes occurring through LeuB.

The final step in leucine biosynthesis, catalyzed by the BCAA aminotransferase (BAT), is common to isoleucine and valine biosynthesis (Fig. 1). In S. cerevisiae, BAT enzymes are encoded by two genes, BAT1 and BAT2 (38, 39). Six BAT enzymes predicted to catalyze this step have been previously identified in A. nidulans (20). We confirmed the identity of these six BATs, and their coding genes, using BLASTP analysis and designated them BatA (AN4323), BatB (AN5957), BatC (AN7878), BatD (AN7876), BatE (AN0385), and BatF (AN8511). Pairwise protein sequence comparisons with Bat1p and Bat2p revealed >21% identity and >31% similarity to both proteins (Table 2). Alignment of these eight proteins showed strong conservation of NAD cofactor binding residues and absolute conservation of the catalytic lysine residue (Fig. S5). The two S. cerevisiae BATs function in different subcellular compartments. Bat1p is primarily targeted to mitochondria, whereas Bat2p is cytoplasmic (39). To predict the subcellular location of the six A. nidulans BAT enzymes, we used DeepLoc-1.0, TargetP v1.1, and Predotar targeting signal predictions (40–43). For all three algorithms, BatA and BatC, like Bat1p, were predicted to be predominantly mitochondrial, and the remaining BAT enzymes were predicted by DeepLoc-1.0 to localize in the cytoplasm (Data Set S1). The BAT protein alignment revealed that Bat1p, BatA, and BatC have extended N termini containing a predicted mitochondrial targeting signal (Fig. S5).

We examined the colinearity of genes surrounding each of the six A. nidulans BAT-encoding genes to identify orthologous genes (Fig. S6). batA and batB orthologs are conserved in regions of high colinearity in all 27 species. batE orthologs are found in a region of moderate colinearity in 13 species. In contrast, batC, batD, and batF were located in regions lacking colinearity. batD only had orthologs in A. niger and A. oryzae, whereas batC and batF have no predicted ortholog. Interestingly, two of the BAT-encoding genes, batC and batD, are separated by just 2 kbp within the aspercryptins secondary metabolite gene cluster (44–47). The tight physical linkage of these two genes suggests that they arose from gene duplication by unequal crossover and, therefore, would show high sequence homology. However, the proteins encoded by these genes are highly diverged, showing only 28.9% protein sequence identity.

To determine the relationship of the six A. nidulans BATs, we performed phylogenetic analysis (Fig. 4). The BATs formed two distinct groups within the fungi. Group I, the larger group containing 37 out of 52 of the fungal BATs, included BatA, BatB, BatC, and S. cerevisiae Bat1p and Bat2p, as well as at least one protein from every other fungus examined. Group II was a smaller group, with only 15 of the 52 proteins, and was almost entirely composed of BAT enzymes from Pezizomycotina genera (Aspergillus, Penicillium, Fusarium, Neurospora, Magnaporthe) and lacked any Saccharomycotina genera (Saccharomyces, Candida). Notably, BatC is in group I and BatD is in group II, consistent with separate recruitment to the aspercryptins cluster.

The expansion of the number of BAT-encoding genes in A. nidulans indicates specialization for the production of isoleucine, leucine, or valine by specific BATs or the evolution of completely new roles. To determine which BAT-encoding genes were required for BCAA biosynthesis, we constructed individual knockout mutants of each of the six BATs (Fig. S3B; see Materials and Methods). Growth tests of the six individual bat knockout mutants showed none were BCAA auxotrophs (Fig. 5A). Therefore, each of the six BATs is dispensable for BCAA biosynthesis. During this study, the two BAT genes found in the aspercryptins gene cluster batC (AN7878) and batD (AN7876) were published by others as atnH and atnJ, respectively, and are thought to be involved in biosynthesis of 2-aminocaprylic acid, 2-aminododecanoic acid, and 2-aminodecanoic acid, three unusual BCAAs that are components of aspercryptins (46, 47).

Analysis of RNA-seq expression data from wild-type mycelia grown on ammonium, alanine, or glutamine (Fig. 6A) showed that batA has the highest expression under all three conditions. batB was the next most highly expressed and showed increased expression on alanine and glutamine compared to ammonium. batC, batD, and batE all showed intermediate expression levels, whereas batF was not expressed under these conditions. As batC and batD are involved in biosynthesis of unusual BCAAs (46, 47), we focused on the other four BAT genes. We measured expression of batA, batB, batE, and batF using RT-qPCR of RNA prepared from samples grown on ammonium, alanine, or nitrate. batA, batB, and batE expression did not substantially change under these conditions (Fig. 6B). batF was not expressed under these conditions, consistent with it being undetectable by RNA-seq. We constructed double, triple, and quadruple mutants combining batAΔ, batBΔ, batEΔ, and batFΔ by meiotic crossing. The batAΔ batBΔ double mutant, which combined deletions of the two most related and highly expressed genes, was a strict BCAA auxotroph and could only grow if supplemented with all three BCAAs (Fig. 5B). Therefore, BatA and BatB are the major BAT enzymes for isoleucine, leucine, and valine (ILV) biosynthesis. The batAΔ batBΔ batEΔ and batAΔ batBΔ batFΔ triple mutants and the batAΔ batBΔ batEΔ batFΔ quadruple mutant showed BCAA auxotrophy identical to that of the batAΔ batBΔ double mutant. In contrast, all of the other double and triple mutants constructed, which contained a wild-type copy of either batA or batB, were BCAA prototrophs. We confirmed that introduction of either the batA or batB gene into the batAΔ batBΔ mutant restored BCAA prototrophy (Fig. S4E). We investigated whether loss of either batA or batB would cause a compensatory increase in expression of batB or batA, respectively. However, on ammonium, batA expression was not upregulated in the batBΔ mutant and batB expression was not upregulated in the batAΔ mutant (Fig. 6C). This indicates that the expression levels of either one of the major bat genes for BCAA biosynthesis is sufficient for prototrophy. We constructed leuBΔ batAΔ and leuBΔ batBΔ double mutants. These two double mutants showed leaky leucine auxotrophy similar to that of the leuBΔ single mutant, indicating that leuBΔ is epistatic to batAΔ and batBΔ (Fig. 6D).

In addition to their role in BCAA biosynthesis, BATs also form the first step in ILV catabolism (28). We examined expression of batA, batB, batE, and batF with ILV as the sole nitrogen source to determine their expression pattern during catabolic conditions (Fig. 6B). For both batA and batE, expression levels were similar under anabolic and catabolic conditions. However, batB levels were elevated substantially during ILV catabolism compared with biosynthetic growth conditions, suggesting that BatB is the predominant catabolic enzyme. batF expression was undetectable. During BCAA catabolic growth, neither batA nor batB expression showed compensatory upregulation in the batBΔ or batAΔ strain, respectively (Fig. 7A). We assessed whether mutants carrying single or multiple BAT gene deletions could utilize each BCAA as the predominant nitrogen source in the presence of lower levels of the other two BCAAs to supplement the auxotrophy (Fig. 7B). All six single BAT mutants could utilize the three BCAAs. Mutants lacking batB but not batA showed slightly reduced colony morphology compared with batB⁺ strains. Notably, mutants lacking both batA and batB showed severely reduced growth on each of the BCAAs as a predominant nitrogen source, and the reduction in growth was greater on isoleucine and valine than on leucine. We also examined growth of the batAΔ and batBΔ single and double mutants on increasing concentrations of equimolar ILV and found that batBΔ shows reduced colony morphology compared with both wild-type and batAΔ strains but stronger growth than the batAΔ batBΔ double mutant (Fig. 7C). Therefore, BatA and BatB are the major BAT enzymes in A. nidulans for both BCAA biosynthesis and utilization. We did not observe a phenotype for batEΔ or batFΔ mutant in BCAA catabolism. Transformation analysis of the batA or batB gene into the batAΔ batBΔ recipient repaired BCAA utilization to the wild-type phenotype (Fig. S4F).

The transcription factor LeuB is thought to regulate leucine biosynthesis genes because the leuBΔ mutant is a leaky leucine auxotroph (19). To determine whether LeuB regulates these genes in response to leucine levels, we performed RT-qPCR on RNA isolated from mycelia grown with exogenous leucine, which represses LeuB activation (27), and in a leuBΔ strain (Fig. 8). leuB expression was not altered in response to leucine. The six genes we demonstrated to function in leucine biosynthesis, leuC, luA, leuD, leuE, batA, and batB, as well as batE, showed decreased expression in response to exogenous leucine and/or in the leuBΔ mutant compared to the wild type.

A. nidulans strains and genotypes are listed in Table 3 using conventional nomenclature (98). A. nidulans growth conditions and media were as described previously (63, 64). Aspergillus nitrogen-free minimal medium (ANM), pH 6.5, containing 1% (wt/vol) glucose as the sole carbon source, was supplemented for auxotrophies and nitrogen sources (10 mM final concentration), unless otherwise stated. A. nidulans growth testing and genetic analysis were as described previously (64).

Escherichia coli NM522 cells [F′ proA⁺B⁺ lacI^q Δ(lacZ)M15/Δ(lac-proAB) glnV thi-1 Δ(hsdS-mcrB)5] (65) were employed for molecular cloning (66). Plasmid DNA was isolated using the Wizard Plus SV miniprep DNA purification kit (Promega). A. nidulans genomic DNA was isolated according to reference 67. PCR products and DNA fragments isolated from agarose gels were cleaned with the Wizard SV gel and PCR clean-up system (Promega). Restriction enzyme digestions (Promega, New England Biolabs), dephosphorylation with Arctic shrimp alkaline phosphatase (Promega), and ligations using T4 DNA ligase (Promega) followed the manufacturers’ instructions. DNA was separated on 1 to 2% agarose gels by electrophoresis in 1× Tris-acetate-EDTA (TAE) buffer. PCRs used Ex Taq (TaKaRa), Phusion (Finnzymes), or AccuStart II Geltrack PCR supermix (Quanta Biosciences) DNA polymerases according to the manufacturers’ instructions, with 1 ng plasmid or 100 ng A. nidulans genomic DNA templates. All reactions followed recommended denaturing and annealing conditions with 33 to 36 amplification cycles. Oligonucleotide PCR primers (Integrated DNA Technologies) are described in Table S1 in the supplemental material. DNA sequencing to confirm correct amplifications and cloning was performed at the Kansas State University DNA Sequencing and Genotyping Facility. Southern hybridizations used either Hybond N+ or Hybond XL membranes (GE Healthcare) and the DIG (digoxigenin) high prime DNA labeling and detection starter kit II (Roche) by following the manufacturer’s instructions.

A. nidulans transformation was performed as described previously (27) using the nkuAΔ nonhomologous integration-defective mutant for targeted integration (68). Knockout constructs, generated by the A. nidulans whole-genome gene deletion constructs program (69), were sourced from the Fungal Genetics Stock Center, Manhattan, KS (70), and were transformed into MH11068 (pyrG89 nkuAΔ::Bar) and selected for uracil and uridine prototrophy to generate leuDΔ (AN0912Δ; RT411, Δ−7 to +1,431 bp), leuEΔ (AN2793Δ; RT413, Δ−6 to +1,233 bp), batAΔ (AN4323; RT415, Δ+65 to +1,722 bp), batBΔ (AN5957Δ; RT440, Δ+25 to +1,395 bp), batCΔ (AN7878Δ; RT475, Δ−10 to +1,222 bp), batDΔ (AN7876Δ; RT419, Δ−7 to +1,297 bp), batEΔ (AN0385Δ; RT417, Δ+27 bp to 1,302 bp), and batFΔ (AN8511; RT441, Δ−9 to +1,230 bp) strains. Selection media for leuDΔ and leuEΔ transformants were supplemented with 2 mM leucine, and BAT gene deletion transformants were supplemented with ILV (2 mM each). The Aspergillus fumigatus pyrG (AfpyrG) marker showed position effect in the batCΔ mutant, incompletely complementing the pyrimidine auxotrophy of the pyrG89 mutation. Full complementation of pyrG89 by AfpyrG was observed in the other deletion mutants generated in this study. Pyrimidine supplementation was used in all growth tests. All deletion mutants were confirmed by Southern blotting as a single homologous double-crossover integration at the correct locus by probing with the 982-bp KpnI-SspI fragment of AfpyrG⁺ (data not shown). Meiotic crossing was used to generate double, triple, and quadruple mutants. The presence of each deletion in the progeny of crosses was confirmed by diagnostic Southern blotting or diagnostic PCR. The leuDΔ mutant was repaired by introduction of a wild-type leuD PCR product (−960 to +3216) amplified from MH1, with direct selection for simultaneous resistance to 1 mg ml⁻¹ 5-fluoroorotic acid (5-FOA) in the absence of exogenous leucine. The leuDΔ leuEΔ mutant was complemented with the plasmid pJS249, which carries leuE (−913 to +2877) PCR amplified from MH1 and cloned into pGEMTeasy by transformation with direct selection for leucine prototrophy. The batAΔ batBΔ double mutant was complemented with the wild-type batA (−717 to +2558) or batB (−725 to +2187) gene using plasmids (pJS244 and pJS255, respectively) containing PCR-amplified DNA from MH1 cloned into pGEMTeasy. Transformants were directly selected for growth in the absence of exogenous ILV.

β-Galactosidase assays were performed as described previously (71) using soluble protein extracts. β-Galactosidase specific activity is defined as A₄₂₀ × 10³ min⁻¹mg⁻¹ of soluble protein. Protein concentrations were determined using Bio-Rad assay reagent (Bio-Rad).

Total RNA was isolated by grinding mycelia under liquid nitrogen and subsequent addition to RNA extraction buffer (7.0 M urea, 100 mM Tris-HCl, pH 8.0, 10 mM EDTA, 1.0% sodium dodecyl sulfate) followed by two phenol-chloroform-isoamyl alcohol extractions and one chloroform extraction (66). RNA was precipitated in 3 M ammonium acetate and 50% isopropanol, resuspended in diethyl pyrocarbonate-H₂O, and reprecipitated overnight in 4 M lithium chloride at −20°C. RNA quality was determined by visualization after electrophoretic separation in a 1.2% agarose gel containing 1.1% formaldehyde run in 1× morpholinepropanesulfonic acid (MOPS) buffer (20 mM MOPS, pH 7.0, 5 mM sodium acetate, 1 mM EDTA). RQ1 DNase (Promega) treatment of RNA followed the manufacturer’s instructions.

For reverse transcriptase-quantitative PCR (RT-qPCR), cDNA was produced using the reverse transcriptase system (Promega) or qScript cDNA supermix (Quanta Biosciences). RT-qPCR used a MyiQ thermocycler (Bio-Rad) with iTAQ universal SYBR green supermix (Bio-Rad), and results were analyzed with iQ5 v2.1 (Bio-Rad). Fold change was calculated using the ΔΔC_T method with β-tubulin-encoding benA as the reference gene (72–74). Primers (IDT) were designed to specifically amplify cDNA by overlapping a splice junction. Primer sequences used for RT-qPCR, target regions, and efficiencies are listed in Table S2.

PolyA⁺ RNA, isolated from three independent biological replicates of wild-type (MH1) mycelia grown for 16 h in supplemented liquid ANM with 10 mM ammonium, glutamine, or alanine, was fragmented to 180 bp and indexed using the TruSeq stranded total RNA sample preparation kit (Illumina). Multiplexed libraries were sequenced using 50-bp single-end reads on the Illumina Hi-Seq 2500 system (Kansas University Medical Center Genome Sequencing Facility, Kansas City, KS). RNA-seq analysis was conducted using Galaxy (www.galaxyproject.org) (75–77). Reads were processed with FASTQ Groomer (78) and FastQC and aligned to the A. nidulans FGSC_A4 genome (79, 80) using TopHat (v2.0.6) (81) default settings, with exceptions (minimum intron length, 10; maximum intron length, 4,000; maximum alignments, 40; minimum read length, 20). Strand-specific reads were separated using SAMtools view (v1.1) (82). Strand-specific transcripts were identified using AspGD annotations (s10_m03_r15) and Cufflinks (v2.1.1.7) (83, 84) default settings, with exceptions (max intron length, 4,000; bias correction, yes; multiread correction, yes). Identified transcripts from all growth conditions were combined into a single annotation using Cuffmerge guided by the reference annotation. Differential expression was determined using CuffDiff (84) and cummeRbund (v2.8.2) (85).

DNA and protein sequences were downloaded from the Aspergillus Genome Database, AspGD (www.aspgd.org [83]), the Saccharomyces Genome Database, SGD (www.yeastgenome.org [86]), the Broad Institute genomes database (www.broadinstitute.org), the NCBI protein database (www.ncbi.nlm.nih.gov/protein/), and the EMBL-EBI Pfam database (http://pfam.xfam.org [87]). Protein sequence database searches used BLAST (https://blast.ncbi.nlm.nih.gov/Blast.cgi). Protein conserved domains were identified using the NCBI Conserved Domain Database (88). Pairwise protein sequence comparisons, and percent identity and similarity were calculated using EMBOSS Needle (EMBL-EBI) with default parameters. Sequences were analyzed in Geneious version 5.3.5, created by Biomatters (www.geneious.com). Multiple sequence alignments were made using ClustalW2 (89) or Clustal Omega (90) on the EMBL-EBI server (http://www.ebi.ac.uk/Tools/msa) and shaded using online BoxShade 3.2 (K. Hofmann and M. D. Baron) at ExPASy (https://embnet.vital-it.ch/software/BOX_form.html). Predicted subcellular localization of proteins was determined using Predotar 1.03 (https://urgi.versailles.inra.fr/predotar/ [42]), TargetP v1.1 (http://www.cbs.dtu.dk/services/TargetP/ [40, 41]), and DeepLoc-1.0 (https://services.healthtech.dtu.dk/service.php?DeepLoc-1.0 [43]). Colinearity of syntenic regions was illustrated using the GBrowse genome browser of FungiDB with genomes clustered based on whole-genome phylogenies (91–93, 99).

The Pfam database (http://pfam.xfam.org/ [87]) was used to identify orthologs in the isocitrate/isopropylmalate dehydrogenase family (PF00180) and the aminotransferase class IV family (PF01063). Protein sequences were aligned and phylogenies were constructed using MAFFT (94). The neighbor-joining method with 100 bootstraps was used to generate consensus unrooted phylogenetic trees of β-IPM dehydrogenases and BATs. Tree visualization and label editing used Interactive Tree Of Life (iTOL) (95).

RNA-seq fastq files and bigwigs have been deposited in NCBI's Gene Expression Omnibus (96) and are accessible through GEO Series accession number (GSE145035).