In Silico Identification of Apple Genome-Encoded MicroRNA Target Binding Sites Potentially Targeting the ACLSV

1. Introduction

The cultivated Apple (Malus domestica Borkh) is an economically and culturally popular, most frequently produced fruit worldwide [1,2,3]. The first reference whole genome for domesticated apple (2n = 2X = 34) was released in 2010 [4]. Apple chlorotic leaf spot virus (ACLSV) is classified within the Trichovirus genus. ASLSV is an economically important, highly deleterious, graft-transmissible, latent pathogen that is distributed worldwide infecting cultivated, ornamental and wild fruit species plants including apple tree [5,6,7,8,9]. The genome of ACLSV (7474–7561 nucleotides) is a positive-sense single-stranded (+ssRNA) that comprises three ORFs. ORF1 encodes a large protein, replication-associated protein (Rep). Movement protein (MP) was coded by ORF2. ORF3 encodes a coat protein (CP) [10,11,12].

In plants, microRNAs (miRNAs) perform a crucial role in various biological processes including development, growth, response to environmental stresses and host-virus interaction by controlling gene expression and regulation. They are typically 20-24 nucleotides in length and can bind to complementary sequences in messenger RNAs (mRNAs), leading to their degradation or repression of translation [13,14]. Plant miRNAs are generated from miRNA precursors (termed pri-miRNAs). Dicer-like1 (GCL1), Ribonuclease (RNase) III enzyme is responsible for processing pri-miRNA transcripts into precursor miRNAs (pre-miRNAs). DCL cleaves pri-miRNA stem-loop, releasing a double-stranded RNA molecule that contains the miRNA sequence. Further, intermediate duplexes (miRNA/miRNA*) formation, stabilization, incorporation to the RNA induced silencing complex (RISC), which guide miRNA to target mRNA by base-pairing rule to lead for repression [15,16,17,18].

The widespread occurrence of microRNA-directed RNA silencing based on RNA interference (RNAi) is a conserved, stronger innate defense mechanism to control plant viruses. It involves gene regulation, host–virus interactions, and a strong inhibiter of virus replication [19,20,21]. The artificial microRNA (amiRNA)-based gene silencing impart resistance in plants against invading viruses. The amiRNA has been utilized successfully in economically important plants against rice stripe virus [22] and cucumber green mottle mosaic virus [23].

Apple tree has been investigated for potential diverse molecular mechanism to explore mature miRNAs which are natural source of immunity during biotic and abiotic stresses, and are important for growth and development [24,25,26,27,28]. The apple genome was mapped with experimentally verified 322 mature mdm-miRNAs that were available from miRBase [29]. The high-confidence mdm-miRNAs are hypothesized to have predicted binding sites in the ACLSV genome. The current study is based on predicting homologous amiRNAs for silencing of the ACLSV. The research work summarized key strategies for analyzing most effective target sites of apple locus-derived mdm-miRNAs and miRNA-mRNA target site interactions in the ACLSV genome. The study aims to elucidate the predicted apple locus-derived-mdm-miRNAs to induce resistance against ACLSV in the transgenic apple plants in future.

2. Materials and Methods

2.1. Apple Mature MicroRNAs and ACLSV Genomic Data Source

Recent studies indicated 322 mature apple (Malus domestica) miRNA sequences (mdm-miRNA156-mdm-miR11020) (Accession IDs: MIMAT0025867-MIMAT0043631) which were available in the miRBase database (version 22) (http://mirbase.org/) [29]. The mature sequences of apple mdm-miRNAs were acquired for analysis (Supplementary Table S1). The whole genome (7545 bases) of ACLSV (Isolate, SY03) (GenBank Accession number KU870525) was retrieved from the NCBI GenBank database [30].

2.2. Analysis of Multiple mdm-miRNA Target-Pairs in ACLSV Genome

A computational ‘four algorithm’ approach was implemented for effective binding prediction of mdm-miRNAs with the ACLSV genome using publicly available tools−miRanda, RNA22, TAPIR and psRNATarget which are most widely used in research (Table 1). Mature sequences of apple genome-encoded mdm-miRNAs and genomic transcript of the ACLSV (in FASTA format) were analyzed.

2.3. miRanda

The miRanda algorithm involves several features such as sequence complementarity, seed-based interaction, miRNA-mRNA duplex dimerization and was released in 2003 [31]. Cross-species target conservation is a key feature for prediction. It has been widely used standard scanning computational algorithm for prediction miRNA binding sites in the corresponding target region based on thermodynamically free energy of duplexes [32]. The miRanda algorithm has been written in C programming language. Standard parameters were set for miRNA-target prediction (Table 1).

2.4. RNA22

The RNA22 algorithm uses a pattern-recognition approach relied on non-seed based interactions of miRNA-mRNA pairs. A web-based server is used for prediction miRNA binding sites in the target sequence [33]. Highly sensitive and significant target patterns were predicted based on maximum folding energy (MFE) [34]. Default parameters were selected for prediction of multiple target sites (Table 1).

2.5. TAPIR

The TAPIR algorithm is a newly developed web server used to select highly specific miRNA-mRNA duplexes precisely. TAPIR is also named as Tapirhybrid. It has been widely used to identify seed-based miRNA-target binding sites based minimum free energy ratio in plants [35]. Target prediction was performed on standards parameters (Table 1).

2.6. psRNATarget

The psRNATarget algorithm is based on pattern of cleavage, high sensitivity, and diversity and uses complementarity scoring schema for prediction multiple binding sites of plant miRNAs in the target sequences using a web-based server [36,37]. The published 207 apple mdm-miRNAs were selected in the web server for analysis. Default standard parameters were selected for prediction of multiple target sites (Table 1).

2.7. Discovering Apple mdm-miRNA-Target Interactions

The apple mature mdm-miRNAs and target ORFs of ACLSV were plotted using CIRCOS algorithm [38].

2.8. RNAfold

The RNAfold algorithm is a web server implemented in ViennaRNA package [39]. It is used to construct the secondary structures of consensus mdm-miRNA precursors. Precursor sequences of apple mdm-MIRNAs were analyzed using default setting in the RNAfold web server.

2.9. RNAcofold

The RNAcofold algorithm is used to evaluate miRNA-mRNA interactions by estimating free energy (ΔG) of duplexes [40]. Consensus fasta sequences of mature apple mdm-miRNAs and target sequences of ACLSV were analyzed using default setting in the RNAcofold web server.

2.10. Statistical Analysis

The predicted miRNA datasets presented here were generated as graphical interpretations. The R-language (version 3.1.1) is widely used tool for biological data interpretation and visualization [41].

3. Results

3.1. Apple Genome-Encoded mdm-miRNAs Targeting ACLSV Genome

The in-silico approach to identify apple mdm-miRNAs with predicted potential for targeting the ACLSV +ssRNA-mRNA among the 322 apple genome-encoded mature mdm-miRNAs was investigated using a focused bioinformatic ‘four algorithms’ approach. The miRanda algorithm predicted binding of 37 mature apple mdm-miRNAs at 46 target sites in the ACLSV genome. RNA22 predicted 108 mdm-miRNAs targeting 161 genomic sites of ACLSV. The TAPIR algorithm identified 103 apple genome-encoded mdm-miRNA-target pairs. In contrast, highly significant ‘cleavable targets’ were identified by the psRNATarget algorithm: 109 mdm-miRNAs targeting at 166 genomic sites of ACLSV (Table S2-S3 and File S1).

3.2. Apple mdm-miRNAs Targeting ORF1 that Ecodes Replication-asscoiated Protein

The trichoviral ORF1 (140-5773) (5634 bases) encodes the replication associated polyprotein (Rep) containing RNA-dependent RNA polymerase (RdRp), which is essential for genome replication [42,43]. The miRanda algorithm predicted binding of thirty one apple mdm-miRNAs: mdm-miR159c (start site 4379), mdm-miR160 (a, b, c, d, e) (4197), mdm-miR169a (4217, 3355), mdm-miR169 (e, f) (1571), mdm-miR169 (g, h, i and j) (4217,3355), mdm-miR393 (g, h) (3091), mdm-miR395k (4691), mdm-miR3627d (2376), mdm-miR5225 (a, b) (182), mdm-miR5225c (4719), mdm-miR7121 (a, b, c, d, e, f, g and h) (149), mdm-miR10998 (3899) and mdm-miR11012 (a, b) (4585) (Figure 2A). RNA22 predicted eighty two apple mdm-miRNAs: mdm-miR160 (start site a, b, c, d, e,) (2133, 4424), mdm-miR166f (963, 4194), mdm-miR167a (975), mdm-miR167 (b, c, d, e ,f, g, h, i and j) (3231), mdm-miR168(a, b) (2504, 3030), mdm-miR169o (345), mdm-miR171f-3p (3230), mdm-miR171f-5p (5101), mdm-miR171o (836), mdm-miR171q (5101), mdm-miR172 (m, n) (3230), mdm-miR319 (a b-3p (5195), mdm-miR319c-5p (952, 1743, 3236), mdm-miR319h (952, 1743, 3236), mdm-miR393 (d, e, f, g, h) (3092), mdm-miR394 (a, b)(2587), mdm-miR395 (d-5p, g-5p, h, i-5p and j) (5181), mdm-miR395k (4691), mdm-miR395l (5592), mdm-miR408a (2376), mdm-miR477(a, b) (3659), mdm-miR482a-3p (2135), mdm-miR530 (a, b, c) (4683), mdm-miR535 (a, d) (1652), mdm-miR3627d (2376, 3541), mdm-miR5225 (a, b) (1364) mdm-miR5225c (1718), mdm-miR7121 (a, b, c) (153, 961, 2574), mdm-miR7121 (d, e, f, g, h) (2574, 4538), mdm-miR10978 (a, b)(1644), mdm-miR10979 (2150), mdm-miR10980 (a, b)(2630, 4198), mdm-miR10983 (5399), mdm-miR10984b-3p(2584), mdm-miR10993 (c, d, e, f) (972), mdm-miR10994-3p (2217), mdm-miR10995 (960, 4695), mdm-miR10996a (3100), mdm-miR11002 (a, b, c-3p) (558, 4139, 5331), mdm-miR11012(a, b) (4582) and mdm-miR11019 (964) (Figure 2B).

The TAPIR algorithm identified several mdm-miRNAs: mdm-miR167a (start site 976), mdm-miR168 (a, b) (2505), mdm-miR169 (k, l, m, n, o) (200), mdm-miR171 (m, n) (700), mdm-miR319 (b-5p, d, e, f) (5373), mdm-miR390 (a, b, c, d, e, f) (687), mdm-miR393 (d, e, f) (3091), mdm-miR394a (1426), mdm-miR394b (1970), mdm-miR395(a, b, c, d-3p, e, f, g-3p, h, and i-3p,) (1970), mdm-miR395k (4691), mdm-miR396 (a, c, d, e) (2702), mdm-miR397 (a, b) (3331), mdm-miR399 (e, f, g, h) (1443), mdm-miR403 (a, b) (2053), mdm-miR477a (2028), mdm-miR482a-3p (2136), mdm-miR482b (2207), mdm-miR530 (a, b, c) (4730), mdm-miR5225c (4490), mdm-miR7120 (a-3p, b-3p) (l2867), mdm-miR7121 (a, b, c, d, e, f, g and h) (1755), mdm-miR10981 (c, d) (5024), mdm-miR10983 (3324), mdm-miR10986 (5404), mdm-miR10989 (a, b, c, d, e) (3173) and mdm-miR10991(a, b, c, d, e) (4997) (Figure 2C). Several potential apple mature mdm-miRNAs were predicted by psRNATarget algorithm: mdm-miR156 (ad, ae) (2674), mdm-miR159 (a, b) (375), mdm-miR164 (b, c, d ,e, f) (2670), mdm-miR166(a, b, c, d, e, f, g, h and i) (781), mdm-miR169 (e, f) (2669), mdm-miR171o (581), mdm-miR172 (a,b ,c ,d ,e ,f ,g ,h ,I ,j ,k ,l ,m ,n, o) (5580), mdm-miR319 (a, b) (2232), mdm-miR394 (a, b) (1426, 4195), mdm-miR395 (a, b, c, d, e, f , g, h and i) (1970, 4121), mdm-miR396 (a, b, c, d, e, f, g) (2702, 3376), mdm-miR399 (e, f, g, h) (1443), mdm-miR408 (a) (2218, 1410), mdm-miR408 (b, c, d) (2668), mdm-miR482a-5p (4111), mdm-miR535 (a, d) (1652), mdm-miR858 (4465, 4296), mdm-miR2111 (a, b) (1021, 4893), mdm-miR5225 (a, b) (1340, 5233), mdm-miR5225c (266, 4490), mdm-miR7120 (a, b) (5128, 4556, 3088), mdm-miR7121(a, b, c, d, e, f, g and h) (1755), mdm-miR7123 (a, b) (1567), mdm-miR7125 (5489) and mdm-miR7126 (1910) (Figure 2D).

3.3. Apple mdm- miRNAs Targeting ORF2 that Encodes Movement Protein

The trichoviral ORF2 (5685-7067) (1382 nucleotides) encodes a multifunctional movement protein (MP) which is required for cell-to-cell movement for ACLSV [44,45,46,47]. The miRanda algorithm predicted several apple mdm-miRNAs to silence the movement protein by targeting ORF2: mdm-miR319d (start site 6744), mdm-miR828 (a, b) (6033), mdm-miR3627d (6736), mdm-miR5225 (a, b) (6226), mdm-miR10980 (a, b) (6561), mdm-miR11008 (6150) (Figure 2A).

RNA22 predicted binding of potential apple mdm-miRNAs to target ORF2: mdm-miR164 (a, b, c, d, e, f) (start site 6096), mdm-miR169 (e, f) (6750), mdm-miR171(a, b) (6071), mdm-miR171f-5p (6651), mdm-miR171 (j, k, l, p) (6071), mdm-miR171 (q) (6071, 5101), mdm-miR319d (6744), mdm-miR393 (d, e, f)(6423), mdm-miR394 (a,b) (6746), mdm-miR399 (a, b, c, d, i, j) (6083), mdm-miR482a-5p (6657), mdm-miR530 (a, b, c) (6312), mdm-miR3627d(6736), mdm-miR7121 (a, b, c, d, e, f, g, h) (6415), and mdm-miR10981 (a, b) (6085) (Figure 2B). The TAPIR algorithm predicted eight mdm-miRNAs: mdm-miR169b (6678), mdm-miR396 (e, f) (6447), mdm-miR3627d (6567), mdm-miR7125 (5806), mdm-miR10980 (a, b) (6561), mdm-miR10994-3p (5976) (Figure 2C). Several potential apple mdm-miRNAs were predicted by the psRNATarget algorithms: mdm-miR169 (b, c, d) (6678), mdm-miR171o (5985), mdm-miR390 (a, b, c, d, e, f) (6321), mdm-miR393 (d, e, f) (6432), mdm-miR395 (a, b, c, d, e, f, g ,h, i) (6286), mdm-miR396 (f, g) (6447), mdm-miR397 (a, b) (6152), mdm-miR482b (6432) and mdm-miR5225c (6200) (Figure 2D).

3.4. Apple mdm- miRNAs Targeting ORF3 that Encodes Coat Protein

The trichoviral ORF3 (6751-7332 bp) (581 nucleotides) encodes a capsid protein (CP) for encapsidation of trichoviral ssRNA [48,49]. Only one mdm-miRNA (mdm-miR5225c) was detected at nucleotide position 7297 by the miRanda algorithm (Figure 2A). Two unique apple mdm-miRNAs were identified by RNA22: mdm-miR168 (a, b) (7265) (Figure 2B). Several potential apple mdm-miRNAs were predicted to target the ORF3 for silencing the coat protein by TAPIR algorithm: mdm-miR156 (p, q, r, s, x, y, z) (6758), mdm-miR156 (aa, ab, ac, ad, ae) (6758, 7293), mdm-miR397b (6839), mdm-miR398 (b, c) (6839), mdm-miR11001 (6858), mdm-miR11016 (6858) (Figure 2C). In addition, psRNATarget predicted nineteen mdm-miRNAs: mdm-miR156 (p, q, r, s) (6758), mdm-miR156 (ab, ac, ad, ae) (6758, 7293), mdm-miR166 (a, b, c, d, e, f, g, h, i) (7335), mdm-miR398a (6890) and mdm-miR858 (6846) (Figure 2D and Figure 3).

3.5. Evaluation of Common Apple MicroRNAs

On the basis of predicted apple genome-encoded mdm-miRNAs, nine miRNAs (mdm-miR5225c and mdm-miR7121 (a, b, c, d, e, f, g and h) were detected by union of consensus between ‘all four algorithms’ (Figure 1 and Figure 3).

3.6. Evaluation and Identification of Consensual Apple MicroRNAs for ACLSV Silencing

Of the 322 targeting mature apple tree mdm-miRNAs, 58 apple mdm-miRNAs (mdm-miR156 (start site o, p, q, r, ab, ac) at nucleotide (nt) position 6758, mdm-miR156 (ad, ae) at nt position 7293, mdm-miR167a at nt position 976, mdm-miR168 (a, b) at nt position 2505, mdm-miR169b at nt position 6678, mdm-miR319d at nt position 6744, mdm-miR393 (d, e, f, g, h) at nt positions 3092 and mdm-miR393 (d, e, f) at nt positions, mdm-miR394a at nt position 1426, mdm-miR395 (a, b, c, d, e, f, g, h, i) at nt position 1970, mdm-miR395k at nt positions 4691, mdm-miR396 (a, c, d, e) at nt position 2702, mdm-miR396 (f, g) at nt positions 6447, mdm-miR399 (e, f, g, h) at nt position 1443, mdm-miR482a-3p at nt positions 2136, mdm-miR535(a, d) at nt position 1652, mdm-miR3627d at nt positions 2376 and 6736, mdm-miR5225c at nt positions 4490, mdm-miR7121 (a, b, c, d, e, f, g, h) at nt positions 153 and 1755, mdm-miR10980 (a, b) at nt positions 6561 and mdm-miR11012 (a, b) at nt positions 4585 were detected by consensus between two algorithms (Figure 4 and Table 2).

On the basis of identifying 58 consensus mdm-miRNAs, nine apple tree mdm-miRNAs, mdm-miR7121 (a, b, c, d, e, f, g, h) and mdm-miR395k were predicted top effective mdm-miRNAs for targeting the ACLSV genome (Figure 4).

3.7. Construction of Apple mdm-miRNAs Regulatory Network

Prediction and validation of host-virus interaction was visualized and created by a ‘Circos map’. The predicted Circos plot shows a comprehensive global purview of the integrated apple mdm-miRNAs and corresponding target genes of the ACLSV genome (Figure 5).

3.8. Secondary Structures of the Consensual RNA

In silico identification, prediction and validation of the consensus apple genome-encoded mdm-miRNAs (mdm-MIR5225c, mdm-MIR395k and mdm-MIR7121 (a, b, c, d, e, f, g, h) was achieved by selecting and predicting their secondary structures using pre-miRNA sequences (Table 3).

3.9. Assessment of Free Energy

Evaluation and validation of consensus apple genome-encoded mdm-miRNAs were determined by assessment of their free energies (ΔG) of duplex bindings (Table 4).

4. Discussion

The ACLSV is a deleterious pathogen of fruit trees belonging to genus trichovirus that reduces plant vigor and yield causing viral infection in fruit trees produced in many countries during the last three decades. The discovery that amiRNA may act as a best strategy to control plant viruses has initiated extensive research in the field plant biotechnology [50]. Recent studies demonstrated expression of numerous endogenous plant miRNAs that may directly target RNA or DNA viruses using an in-silico approach―employing several computational algorithms [51,52,53,54,55,56,57]. Recent studies demonstrated expression of amiRNA-based constructs in economically important transgenic crops to eliminate plant viral load against RNA and DNA viruses [22,23,58,59,60,61,62,63]. Although several studies have already been performed on ACLSV pathogenesis, no research work was reported to enunciate the role of apple genome-encoded miRNA-mediated regulation targeting ACLSV viral replication. We highlight pivotal studies first time to test computationally the mature apple genome-encoded miRNAs for predicting effective miRNA-binding sites and to understand complex host-trichovirus specific interactions with the ORF1, ORF2 and ORF3 of ACLSV genome.

In silico algorithms are widely used to predict miRNA-binding sites in the target region to screen host-virus interaction [64]. The miRanda, RNA22 and TAPIR algorithms identified a consensual binding site of mdm-miR395k at nucleotide position 4691. The apple common madm-miR5225c and mdm-miR7121 (a, b, c, d, e, f, g, h) were predicted as the most potent miRNAs by all four algorithms (Figure 4 and Table 2). The TAPIR and psRNATarget algorithms were utilized to predict binding strength of madm-miR5225c and mdm-miR7121 (a, b, c, d, e, f, g, h) at consensus genomic positions 4490 and 1755, respectively.

For predicting miRNA-binding sites of the target sequence based on MFE which is also interpreted for evolutionary inferences [65]. The stability of miRNA-mRNA duplex is associated with the binding energy. Functional miRNA–target recognition depends upon site accessibility which is a key feature of in silico algorithms to assess the false-positive miRNA–target interaction. Validation of the miRNA–target interaction also depends upon MFE [66]. High probability of miRNA–target interaction was set on lower value of MFE [67]. High stability of miRNA-mRNA duplex is directly dependent on the stronger ability of binding affinity of miRNA to target mRNA [68,69]. Prediction, evaluation and validation of miRNA targeting patterns were based on base-pairing probability of mdm-miRNA seed regions with complementary high affinity binding sites in the ACLSV genome. The MFE of mdm-miR395k was calculated as −20.85 kcal/mol (miRanda), −18.00 kcal/mol (RNA22) (Table 2), and −18.30 kcal/mol (RNAcofold) (Tables 4), supporting prediction results with high stability of miRNA-mRNA duplex―representing ‘true targets’ [82,83].

Taken together with assessment of free energy is a standard feature of miRNA-target binding sites prediction which identified ten consensual potent apple miRNAs (Table 2, 3, 4). These predicted apple mdm-miRNAs have robust potential for RNAi-based gene silencing. Future work is focused on amiRNA-based constructs of ACLSV genome for transformation in apple plants.

Computational analysis have been implicated the apple consensus mdm-miR395k targeting ORF1 of ACLSV. The apple precursor mdm-MIR395k (NCBI Accession ID: MI0035639) was located on apple chromosome MDC003846.250 (genome context coordinates 8270 to 8437)[70]. Whereas in apple, Md-miR395 was demonstrated to target transcription factor MdWRKY26 to regulate apple resistance to leaf spot disease [24]. The miR395 is involved in regulating carbohydrate accumulation gene (NADP-MDH) for flower development in tea-oil camellia [71]. Our studies show that apple genome-encoded miRNA pairs directly with specific locations on ACLSV +ssRNA-encoded mRNAs. In the current study, we demised a model to estimating the probability of prediction using different approaches to reduce false-positive results. These include at individual, union and intersection levels. Union approach is a highly sensitive for predicting candidate miRNAs based on combining more than one miRNA prediction tools. A lower level of prediction specificity was observed. While using intersection approach, which entirely based on prediction specificity suing more than two prediction tools with decrease in sensitivity of the predicted data [72]. Our reliable target prediction results demonstrated that in silico strategy concluded high efficiency of predicted data at individual, union and intersection levels with best target binding sites of apple mdm-miRNAs under study (Figure 1, 2, 3, 4 and Table 2).

While existing literature has mostly focused on full-length genome sequencing of ACLSV, here we identify apple genome-encoded miRNAs that bind directly to several consensuses genomic regions of ACLSV. In addition, there is no previous report that showed that apple tree mdm-miRNAs bind to ACLSV predicting diverse, functionally related, host-miRNA-virus-mRNA interactions. While interactions between apple genome-encoded mdm-miRNAs and ACLSV were evaluated in this study, the probability of the apple mdm-miRNAs to bind ACLSV in vitro is needed to explore. While more studies need to be performed to fully comprehend the mechanism of host-virus interaction on the ACLSV genome at experimental level. This study deepens our understanding of ACLSV biology and validation of predicted miRNAs to develop resistant apple varieties. Further experiments are needed to determine binding strength of predicted mdm-miRNAs in transgenic apple plants in future.