Analysis of the codon usage pattern of the RdRP gene of mycovirus infecting Aspergillus spp.

The AfuPmV-1 genome consists of four capped double-stranded RNAs (dsRNAs), of which the largest dsRNA1 segment encodes the RNA-dependent RNA polymerase (RdRP) [5, 6]. The data were downloaded from the NCBI GenBank database (https://​www.​ncbi.​nlm.​nih.​gov/​genbank) and the complete nucleotide sequences of the RdRP coding region of mycoviruses that infect Aspergillus species were used. Eight sets of data published between 2006 and 2015 were studied in the analysis. Other polymycoviruses and hypervirulence-related mycoviruses data were downloaded from NCBI for comparison (Table 2).

The programs CodonW (https://​sourceforge.​net/​projects/​codonw/) and CALcal (http://​genomes.​urv.​es/​CAIcal/) were used to conduct analyses of nucleotide composition, overall/local G + C content, relative synonymous codon usage (RSCU), effective number of codons (ENC), codon adaptation index (CAI), and correspondence analysis (COA) for each of the selected genetic data [9, 10]. Nucleotide composition analysis was conducted based on the overall frequency of nucleic acid occurrence (A%, C%, U%, and G%), total AU%/GC%, frequency of the third nucleic acid in synonymous codons (A3s, C3s, U3 s, and G3 s), and the GC3s values. The RSCU is a value represented as the ratio between predicted and observed usage rates of specific codons, assuming that all synonymous codons were used equally for an amino acid. An RSCU value of 1.0 indicates that all codons were used randomly or equally. The value would be greater (less) than 1.0 if certain codons were used more (less) frequently [11, 12]. In particular, codons with RSCU values ≥1.6 are considered over-represented codons, while those with RSCU values ≤0.6 are considered under-represented codons [13, 14]. The ENC is a simple and absolute method to measure codon usage bias in genes and genomes. The GC3s values are used in the calculation, and the ENC value ranges from 20 to 61. The ENC is inversely proportional to codon usage bias, and lower ENC values indicate stronger codon usage bias. A gene that does not show any codon usage bias would have an ENC value of 61. On the other hand, a gene that shows a high degree of codon usage bias would have an ENC value of 20, indicating that only one codon was used for an amino acid. Generally, ENC values < 35 are interpreted as showing strong codon usage bias [15, 16]. The CAI is the geometric mean of relative adaptation level and is a quantitative method to calculate the differences in codon usage bias against highly expressed known reference data. The CAI value ranges between 0 and 1, with values closer to 1 indicating a high degree of similarity between the reference data and the codon usage pattern and expression level, whereas smaller values indicate lower similarity in the codon usage pattern and expression level [17]. The reference data of codon usage levels in humans (Homo sapiens) and fungus (Aspergillus fumigatus) were acquired from the codon usage database (http://​www.​kazusa.​or.​jp/​codon/) for comparison [18]. Furthermore, the RSCU values were examined by the COA method with the XLSTAT 2016 program for visualization of the CodonW results. The COA method is the preferred method in the area of multivariate analysis and represents the data as vectors consisting of rows and columns [19]. Each individual data of the RdRP codon region was represented as a vector with 59 dimensions covering 59 codons except methionine (AUG) and tryptophan (UGG), which lack synonymous codons, as well as termination codons. A comparative analysis was performed including three groups: mycoviruses infecting Aspergillus (Group 1), polymycoviruses with a fungal host other than Aspergillus (Group 2), and newly reported mycoviruses that enhance host virulence (Group 3).

The phylogenetic relationships among the mycoviruses used for the analysis were inspected and phylogenetic analysis was conducted using the MEGA7 program (http://​www.​megasoftware.​net) to infer the influence of evolutionary processes on codon usage patterns. RdRP coding sequences were aligned with the MUSCLE algorithm, and the phylogenetic tree was constructed by applying the Maximum Likelihood method and the Kimura 2-parameter substitution model. The robustness of the tree was verified with the bootstrap value set to 1000 [20].

Basic nucleotide composition analysis was conducted for the RdRP gene of mycoviruses infecting Aspergillus spp. (Table 3, Fig. 1). The mean and standard deviation (SD) of A%, C%, U%, and G% were 23.16 ± 4.00, 25.09 ± 5.13, 25.13 ± 3.54, and 26.63 ± 4.04, respectively. The mean and SD of total AU% and GC% were 48.29 ± 6.76 and 51.71 ± 6.76, respectively. The mean and SD of A3s, C3s, U3 s, G3 s, and GC3s were 22.34 ± 8.66, 34.72 ± 12.09, 34.26 ± 10.20, 32.15 ± 7.47, and 53.48 ± 13.45, respectively. AfuPmV-1 showed the lowest frequencies of nucleotides A and U and the highest frequency of nucleotide C among the Aspergillus-infecting mycoviruses included in the analysis. Similarly, the frequencies of the four types of nucleotides in the third position of synonymous codons were the lowest with A3s and U3 s and highest with C3s. As nucleotide C occurred relatively abundantly in AfuPmV-1, the GC content and GC3s indices were highest in this virus with values of 62.74 and 78.30, respectively (Fig. 1). The GC content of AfuPmV-1 was reported to be approximately 63% in a previous study of the entire or partial genome sequences of Polymycoviridae viruses [6]. As a result of the comparison, Group 2 showed a similar pattern to that of AfuPmV-1. The GC content was between 59.97 and 61.98%, and the frequencies of nucleotides at the third position in the synonymous codons were higher in C3s and G3 s than in A3s and U3 s, respectively. Among these, BbPmV-1 is an experimentally reported virus that may be related to mild hypervirulence. The Group 3 results differed from those of Groups 1 and 2 (Table 3).

The RSCU values of AfuPmV-1, AfuCV, and AfuPV-1 of Group1 were compared to inspect the codon usage bias according to virus species (Table 4, Fig. 2). Of codons related to the entire 18 amino acids, 18 codons were preferred in AfuPmv-1 of which 11 showed RSCU values ≥1.6. AfuCV showed 17 preferred codons, three of which had RSCU values ≥1.6. AfuPV had 20 preferred codons, three of which showed RSCU values ≥1.6. AfuPmV-1 showed similarities to AfuCV in codons CAC (His), CAG (Gln) and to AfuPV-1 in codons CUC (Leu) and AUC (Ile). The codons preferred by each virus individually were as follows. The codons solely preferred in AfuPmV-1 were GUC (Val), UAC (Tyr), GAC (Asp), UCC (Ser), CCC (Pro), ACC (Thr), GCC (Ala), UGC (Cys), CGC (Arg), and GGC (Gly) for 10 of the amino acids. Remarkably, all of these were C-ended codons (Fig. 2). Codons solely preferred in AfuCV were CUG (Leu), AUA (Ile), CCG (Pro), GCU (Ala), AGC (Ser), and AGG (Arg) for six of the amino acids, and in AfuPV-1, CUU (Leu), CAU (His), CAA (Gln), UCA (Ser), CCA (Pro), GCA (Ala), GCG (Ala), UGU (Cys), and CGU (Arg) for eight of the amino acids. The codons UUC (Phe), AAC (Asn), AAG (Lys), and GAG (Glu) showed similar preferences in all three viruses. The RSCU values and end nucleotide composition indicated that, in the RdRP coding region, C-ended codons were strongly preferred in AfuPmV-1 (15 of 18), G-ended codons were preferred in AfuCV (7 of 17), and U-ended codons were preferred in AfuPV-1 (8 of 20). Interestingly, there were no A/U-ended codons among the preferred codons of AfuPmV-1, indicating that AfuPmV-1 has a codon bias toward C- and G-ended codons.

Group 2 showed a similar codon usage pattern to that of AfuPmV-1, which indicates a preference for the C/G-ended codon. Group 3 also preferred C- or G-ended codons, but the four nucleic acids were distributed more evenly.

The ENC value was calculated to quantitatively measure the magnitude of RdRP gene codon usage bias of the eight mycoviruses infecting Aspergillus spp. An ENC value < 35 indicates a strong codon usage bias. The results showed that the ENC value was lowest for AfuPmV-1 (40.67) and the other viruses showed ENC values > 50 (Table 3, Fig. 3). Taking into account the results of previous studies on RNA viruses in which the ENC values of Zaire Ebola virus, Chikungunya virus, Hepatitis C virus, and West Nile virus were 57.23, 55.56, 52.62, and 53.81, respectively [21], AfuPmV-1 appeared to have stronger codon usage bias relative to other viruses. The CAI was calculated to compare the adaptability of synonymous genetic codon usage in mutually different individuals, and the codon usage patterns were considered similar to the reference individual with CAI values closer to 1. This study referred to the CAI values of H. sapiens and A. fumigatus, which have ranges of 0.699–0.762 and 0.676–0.843, respectively. The mean and SD were 0.72 ± 0.02 and 0.74 ± 0.05 for H. sapiens and A. fumigatus, respectively. Remarkably, AfuPmV-1 showed the highest values for both H. sapiens and A. fumigatus with values of 0.762 and 0.843, respectively. These results indicated that AfuPmV-1 has the greatest similarity to the reference data in codon usage pattern and expression level, and that it has higher adaptability to human and fungal hosts compared with other mycoviruses. In Group 2, CAI and ENC values were similar to those of AfuPmV-1 (Table 3).

To inspect the trends related to codon usage patterns of the Aspergillus-infecting viruses AfuPmV-1, AfuCV, and AfuPV-1, COA was performed with the RSCU values. Axis1, Axis2, Axis3, and Axis4 of the COA-RSCU explained 44.31, 18.81, 17.33, and 9.78% of the total variation, respectively. The results were based on 59 codons, excluding the three termination codons and methionine (AUG)/tryptophan (UGG) that do not have synonymous codons. Although there were exceptions in the COA results according to the third nucleotide in the codon, G- and C-ended codons formed one group and A- and U-ended codons formed another group (Fig. 4(a)). Moreover, the COA results for over-represented codons with RSCU values ≥1.6 showed that the codons strongly preferred individually by the three viruses formed separate groups (Fig. 4(b)). The observations verified that there were differences in the codon pattern preferences among the Aspergillus-infecting viruses.

A phylogenetic tree was constructed to examine the phylogenetic relationships among the 14 mycoviruses, including AfuPmV-1. Polymycoviruses were grouped with AfuPmV-1. Polymycoviruses showed a relatively close relationship with alternaviruses (Fig. 5). To provide more information, the RdRP family for each sequence was examined from Pfam. AfuPmV-1 was assigned to RdRP_1 and all other polymycoviruses with similar codon patterns were assigned to the same RdRP_1 (pfam00680). Other mycoviruses infecting Aspergillus were classified as RdRP_1 or RdRP_4, although they were found in the same fungus (Table 5). This result suggests that the hypervirulent effects of AfuPmV-1 may be more affected by viral genome characteristics than by the effect of Aspergillus as a host.