Dynamics of Small RNA Profiles of Virus and Host Origin in Wheat Cultivars Synergistically Infected by Wheat Streak Mosaic Virus and Triticum Mosaic Virus: Virus Infection Caused a Drastic Shift in the Endogenous Small RNA Profile

Co-infection of wheat (Triticum aestivum L.) by Wheat streak mosaic virus (WSMV, a Tritimovirus) and Triticum mosaic virus (TriMV, a Poacevirus) of the family Potyviridae causes synergistic interaction. In this study, the effects of the synergistic interaction between WSMV and TriMV on endogenous and virus-derived small interfering RNAs (vsiRNAs) were examined in susceptible (‘Arapahoe’) and temperature-sensitive resistant (‘Mace’) wheat cultivars at 18°C and 27°C. Single and double infections in wheat caused a shift in the profile of endogenous small RNAs from 24 nt being the most predominant in healthy plants to 21 nt in infected wheat. Massive amounts of 21 and 22 nt vsiRNAs accumulated in singly and doubly infected Arapahoe at both temperatures and in Mace at 27°C but not 18°C. The plus- and minus-sense vsiRNAs were distributed throughout the genomic RNAs in Arapahoe at both temperature regimens and in Mace at 27°C, although some regions served as hot-spots, spawning an excessive number of vsiRNAs. The vsiRNA peaks were conserved among cultivars, suggesting that the Dicer-like enzymes in susceptible and resistant cultivars similarly accessed the genomic RNAs of WSMV or TriMV. Accumulation of large amounts of vsiRNAs in doubly infected plants suggests that the silencing suppressor proteins encoded by TriMV and WSMV do not prevent the formation of vsiRNAs; thus, the synergistic effect observed is independent from RNA-silencing mediated vsiRNA biogenesis. The high-resolution map of endogenous and vsiRNAs from WSMV- and/or TriMV-infected wheat cultivars may form a foundation for understanding the virus-host interactions, the effect of synergistic interactions on host defense, and virus resistance mechanisms in wheat.


Introduction
Co-infection of plants with two or more viruses is common in nature because plants are susceptible to multiple viruses. Additionally, many hemipteran vectors transmit more than one virus, thus causing mixed infections by related or unrelated viruses [1], leading to antagonistic or synergistic interactions [2]. Antagonistic interactions result from infection of plants by a virus, followed by different strains of the same virus at different times, where only one virus strain is the beneficiary, and its preexistence prevents superinfection by a second virus [3,4]. In contrast, synergistic interactions among unrelated viruses in mixed infections display a facilitative effect on one or both of the viral partners, with more severe symptoms than those induced by individual viruses [5][6][7][8][9][10][11]. Synergistic interactions among some other virus combinations have been reported as host-dependent [8,[11][12][13], suggesting that in addition to interactions among viral partners, successful interaction between viral and host proteins might play an important role in synergistic interactions. However, several aspects of synergistic interactions are not clearly understood. For example, how does co-infection by two unrelated viruses cause more severe symptoms than infection by individual viruses? What are the effects of virus-encoded RNA silencing suppressor proteins on host defense mechanisms in co-infected plants compared with plants infected by individual viruses? Plant viruses are considered to be strong inducers as well as targets of host RNA silencing. RNA silencing is one of the most important antiviral defense mechanisms used by plants against viral infections [14,15]. In addition to antiviral defense, RNA silencing regulates several plant developmental processes, transposon control, DNA methylation, and chromatin modification [16,17]. Viral double-stranded (ds) RNAs, generated by viral RNA-dependent RNA polymerases as replication-intermediate forms or as highly structured portions of single-stranded viral genomic and sub-genomic RNAs, trigger the antiviral RNA silencing defense mechanism in plants [18][19][20][21][22]. These dsRNAs are recognized by Dicer-like (DCL) ribonucleases and are processed into 21 to 24 nucleotide (nt) short dsRNAs known as virus-derived small interfering RNAs (vsiRNAs) [23][24][25][26][27]. Two classes of vsiRNAs have been identified in virus-infected plants, namely, primary vsiRNAs derived from the initial cleavage of virus derived dsRNA-like structures by DCLs, and secondary vsiRNAs formed by amplification of primary vsiRNAs by host RNA-dependent RNA polymerases (RDRs) [24,28]. These vsiRNAs are recruited by Argonaute (AGO) proteins to form a multiprotein effector complex, an RNA-induced silencing complex (RISC), and guide the cleavage of homologous RNAs in a sequence-specific manner, or induce translational repression or chromatin modification [29][30][31][32].
Both monocotyledonous and dicotyledonous plants contain DCL ribonucleases, AGOs and RDRs, the core components of host RNA silencing [30]. In Arabidopsis thaliana (L.) Heynh, 4 DCLs, 10 AGOs, and 6 RDRs have been identified [33]. Among monocots, the RNA silencing phenomenon and its components have extensively been studied only in rice (Oryza sativa L.). So far, in rice, 8 DCLs,19 AGOs and 5 RDRs have been identified [34]. The presence of different numbers of RNA silencing core components in rice and A. thaliana suggests that different RNA silencing pathways might be utilized by these two angiosperms. The mechanisms of RNA silencing were extensively studied in A. thaliana using mutant plants that defined the roles of DCLs, AGOs and RDRs in RNA silencing [35]. In addition to the processing of endogenous precursor RNAs to sRNAs, all four DCLs found in A. thaliana have been implicated in the formation of vsiRNAs [36,37]. However, their action is hierarchical, with DCL4 and DCL2 being the main contributors [23]. Loading vsiRNAs on to the AGO complex in plants is preferentially, but not exclusively, dictated by the 59-terminal nucleotide of vsiRNAs [38].
Wheat (Triticum aestivum L.) is the world's most widely grown crop, covering 200 million hectares with an annual production of 681 million tons in 2011, slightly behind maize (Zea mays L.) and rice, providing the primary source of carbohydrates, proteins, vitamins, and minerals for the world's population [39]. The 17gigabase-pair hexaploid wheat genome is complex, about three times larger than the human genome [40,41]. Among viral pathogens infecting wheat in the Great Plains region of the United States, the main U.S. wheat-growing region, Wheat streak mosaic virus (WSMV) is the most economically important virus causing significant yield losses [42,43]. Recently, Triticum mosaic virus (TriMV), a new virus, was reported from several Great Plains states [44][45][46]. WSMV and TriMV are the type members of the Tritimovirus and Poacevirus genera, respectively, in the family Potyviridae [47,48]. Both of these viruses are vectored by Aceria tosichella Keifer, the wheat curl mite [49][50][51], resulting in mixed infections in growers' fields [44,52]. Wheat cultivar (cv.) Mace is resistant to WSMV at temperatures below 19uC [53]. Mace's resistance to WSMV is conditioned by the Wsm-1 gene, which is also found to provide similar levels of resistance to TriMV [8,52].
WSMV and TriMV possess limited sequence homology but interact synergistically in co-infected susceptible wheat cvs. Arapahoe and Tomahawk at 19uC and 20 to 26uC, resulting in increased accumulation of both viruses and enhanced disease severity [8,52]. However, in Mace, double infections induced moderate disease synergism at 20 to 26uC but not at 19uC. The P1 proteins of WSMV and TriMV have been reported to be involved in the suppression of local and systemic RNA silencing in N. benthamiana 16c plants [55,56]. However, the response of wheat defense mechanisms against single and double infections in susceptible and resistant wheat cultivars with WSMV and TriMV has not been examined, and it is not known how viral infection(s) trigger cascades of host defense mechanisms that cause production of vsiRNAs, an important component of host RNA silencing.
In this study, the dynamics of endogenous sRNAs and vsiRNAs were examined in WSMV and/or TriMV-infected susceptible and temperature-sensitive resistant wheat cultivars as a first step toward understanding virus-host interactions in synergistic infections.

Virus source
In vitro transcripts from an infectious cDNA clone of WSMV isolate Sidney 81 were inoculated onto wheat (cv. Tomahawk) seedlings at the single leaf stage [57]. A TriMV Nebraska isolate was activated from dried wheat leaves that were stored at 4uC in a 125-ml Erlenmeyer flask over anhydrous calcium sulfate [48] by inoculating wheat seedlings at the single-leaf stage. WSMV, TriMV and WSMV plus TriMV inocula were prepared from 12 dpi wheat cv. Arapahoe infected with WSMV or TriMV at 1:20 dilution in 20 mM sodium phosphate buffer, pH 7.0, and mechanically inoculated to wheat cvs. Arapahoe and Mace at the single leaf stage, two pots per inoculum with 12 to 15 seedlings per pot. Inoculated wheat plants were incubated in growth chambers (Percival, Perry, IA) at 18uC and 27uC with a 16 h photoperiod. Fully expanded upper leaves of infected wheat at 16 dpi were harvested for total RNA extraction.

Isolation of total RNA
One gram of fully expanded upper leaves of wheat infected with WSMV, TriMV, WSMV plus TriMV, or healthy plants was ground in liquid nitrogen into a fine powder, followed by the addition of 5 ml of TriPure isolation reagent (Roche, Indianapolis, IN). The tissue was ground thoroughly and the extract was transferred to an Oak ridge tube. The mortar was washed with an additional 2 ml of TriPure isolation reagent. The extract was vortexed for 30 s, followed by the addition of 1.4 ml of chloroform and vortexed for 30 s. This mixture was incubated at room temperature for 10 min, followed by clarification at 12,0006g for 15 min at 4uC. Total RNA was precipitated from the 3 ml aqueous phase by adding an equal volume of isopropanol and incubated at room temperature for 10 min. The RNA was pelleted at 12,0006g for 10 min at 4uC, and the pellet was washed with 5 ml of 70% ethanol. The air-dried RNA pellet was suspended in 500 ml of sterile water and stored frozen at -70uC. The amount of total RNA was quantified using a NanoPhotometer (Implene Inc., Westlake Village, CA).

Northern blot hybridization
Two mg of total RNA was separated on 15% polyacrylamideurea gels (Bio-Rad, Hercules, CA), followed by electro-transfer to Nylon membranes (Roche). Nylon membranes were probed with DIG-labeled minus-strand RNA riboprobes specific to nt 6303 to 9384 of WSMV or nt 6401 to 10,245 of TriMV. These probes were further hydrolyzed into ,50 nt long RNA pieces by treatment with sodium carbonate as described in Dalmay et al. [58]. Prehybridization and hybridization were carried out essentially as described in Tatineni et al. [55].

Isolation of small RNAs
Thirty mg of total RNA was denatured in the presence of 47.5% formamide, 0.1X TBE buffer plus 0.015% of each bromophenol blue and xylene cyanol FF at 95uC for 4 min, followed by incubation on ice for 5 min. Denatured total RNA was separated on a 15% polyacrylamide-urea-TBE gel (Bio-Rad) at 120 V until the bromophenol blue reached the bottom of the gel. Gel slices from just below the center of the xylene cyanol band and extending to three-quarters of the way toward the bromophenol blue band were collected into a 30 ml corex tube containing 5 ml of 0.3 M sodium chloride. The gel slices were crushed into small pieces with a 10 ml syringe piston and incubated in a shaker at 4uC overnight. The supernatant was collected after centrifugation at 10,000 g for 15 min at 4uC. Two ml of 0.3 M sodium chloride was added to acrylamide pieces and incubated in a shaker for 1 h at room temperature. The small RNAs were precipitated from the pooled supernatants by addition of 3 volumes of ice-cold ethanol in the presence of 0.3 M sodium acetate, pH 5.2 and 2 ml of glycogen (Roche), and incubated at 270uC for 1.5 h. The small RNAs were suspended in 350 ml of 0.3 M sodium chloride and loaded onto a spin-x-cellulose acetate filter (Sigma-Aldrich, St. Louis, MO). The spin column was centrifuged at 10,000 g for 2 min at room temperature. The small RNAs were precipitated from the elutant by adding 3 volumes of ethanol in the presence of 0.3 M sodium acetate and 1 ml glycogen for 30 min at 270uC, followed by centrifugation at 12,000 g for 15 min at 4uC. The RNA pellet was washed with 70% ethanol, and the air dried small RNA pellet was suspended in 10 ml sterile water and stored at 2 70uC.

Deep sequencing
Small RNA libraries were prepared from gel-eluted small RNA fractions of total RNA using Illumina TruSeq small RNA Sample Prep Kit (Illumina Inc, San Diego, CA), essentially as described in the manufacturer's instructions at the University of Nebraska-Lincoln Genomics Core Research Facility. Illumina TruSeq SR Cluster Generation Kit v5 was used to generate the cluster at one sample per lane. Amplified library was sequenced on Illumina Genome Analyzer IIx using a 36-cycle sequencing kit to read 36 nts of sequence from a single end of each insert by standard v8 protocol. The raw sequence reads for all 16 samples were deposited in the GEO Omnibus under accession GSE54026.

Bioinformatics analysis of small RNAs
Residual Illumina TruSeq adapters were removed from the small RNA samples via the cut-adapt tool (version 0.9.5, [59]) using the following parameters: a maximum error rate of 15% (2e 15), base trimming when the Phred-like quality score was lower than 45 (2q 45), and iterative removal of adapters up to three times (2n 3). Thus, the cleaned 16 samples contained a grand total of 14 billion bases spanning approximately 570 million reads (Table S1). The number of reads found in healthy and infected wheat cultivars was normalized against the reads of healthy Arapahoe at 18uC (Table S2). The initial size distribution of the small RNAs were determined for each sample via a custom inhouse Perl script.
The small RNAs were mapped against WSMV and TriMV genome sequences and small RNAs that did not align to these viral sequences were considered to be of host origin. Similarly, crosscontamination of healthy samples with virus was detected and removed. A custom in-house Perl script was developed to target the small RNAs of interest and to generate information on the final length distribution and the nucleotide frequency at the 59 end of the small RNAs, as well as coverage information on the virus genomes. As extra filters for the final output, in the case of the healthy samples, any read with a non-called base (N) was removed; for both the healthy and virus-infected samples, small RNAs were size-selected between 18 to 26 nucleotides (inclusive) and reads containing alignment mismatches, or that do not map uniquely, were removed. For the samples with the double virus infection, any RNAs that cross match between the two viruses were removed as well.
MATLAB (2012b, The Mathworks, Inc.) was used to generate coverage graphs of the small RNAs on the virus genomes.
Chi-square analysis was used to determine whether the treatments induced statistically significant changes in the distribution or composition of small RNAs. For each cultivar and temperature treatment, the observed distribution (as percent) of sRNA from 20 to 25 nt in healthy tissue was used as the ''expected'' class in chi-square calculations.

Co-infection of wheat cultivars with WSMV and TriMV caused an increase in the accumulation of vsiRNAs
The effect of the synergistic interaction between WSMV and TriMV on the accumulation of vsiRNAs in wheat cultivars Arapahoe and Mace at 18uC and 27uC was examined by Northern blot hybridization of total RNA using strand-specific digoxigenin (DIG)-labeled riboprobes. The riboprobes hybridized neither with total RNA from healthy wheat cultivars nor another co-infecting virus, suggesting that these riboprobes are highly specific to their respective viruses ( Fig. 1). In singly and doubly infected wheat cv. Arapahoe, increased levels of WSMV-and TriMV-specific vsiRNAs accumulated in plants incubated at 27uC compared to 18uC (Fig. 1). In Mace, vsiRNAs were readily detected in singly and doubly infected plants at 27uC but not 18uC, further confirming Mace's resistance to WSMV and TriMV at 18uC. In wheat cv. Arapahoe at 18uC and 27uC, and cv. Mace at 27uC, increased levels of vsiRNAs accumulated in doubly infected plants relative to levels observed in singly infected plants ( Fig. 1), suggesting that synergistic interaction between WSMV and TriMV is not due to the presence of two RNA silencing suppressor proteins.

High-throughput sequencing of small RNAs from WSMVand/or TriMV-infected wheat cultivars
To obtain insight into the mechanisms of synergistic interactions among unrelated viruses and the effects of these interactions on host defense mechanisms, the profile of endogenous small RNAs and vsiRNAs was examined from singly (WSMV or TriMV) and doubly (WSMV plus TriMV) infected wheat cultivars Arapahoe (susceptible) and Mace (temperature sensitive-resistant) at 18uC and 27uC.
Composition of small RNAs obtained through Illumina Solexa high-throughput sequencing was examined by computational alignment of sequences identical or complementary to the genomes of WSMV and TriMV. Similarly, small RNAs not corresponding to the viral genomes from doubly infected wheat cv. Arapahoe at 27uC were also BLAST searched for sequences identical to the wheat genome and most of the sRNAs were aligned with the wheat genome. Since there are few gaps in the wheat genome sequence [40,41], we considered sRNA sequences that do not align with the genomes of WSMV and TriMV as wheat genomic sequences. Moreover, we did not find significant matches to other pathogen sequences present in the database, suggesting that sequences not aligning with WSMV or TriMV most likely represent the wheat genome. Small RNAs between 18 and 26 nt in size were selected for further analyses.
A total of 29.8 to 37.6 million reads were obtained from WSMV-and/or TriMV-infected Arapahoe and Mace at 18uC and 27uC, and 34.2 to 39.2 million reads were obtained from healthy samples (Table 1). In infected samples, 6.8 to 14.3 million size-selected (18 to 26 nt) reads were specific to the wheat genome, in contrast to 10.4 to 13.7 million size-selected reads from healthy samples ( Table 1). In healthy samples, only 25 to 1321 virus-specific size selected reads were found, and the authenticity of these reads will be known only after the availability of the complete wheat genome sequence without gaps. In Arapahoe at 18uC and 27uC and Mace at 27uC, 6.5 to 17.1 million reads were aligned with the WSMV or TriMV genome, and 94 to 98% (6.3 to 16.5 million) of these reads were in 18 to 26 nt size class (Table 1). Among size selected reads, 20 to 25 nt sRNAs accumulated at significantly higher levels compared with 18, 19 and 26 nt sRNAs in healthy and virus-infected samples. Hence, further analyses of small RNAs from healthy and virus-infected samples were performed for 20 to 25 nt RNAs.

Virus infection caused a shift in the profile of endogenous sRNAs accumulation
In healthy wheat cultivars Arapahoe and Mace at 18uC and 27uC, 24 nt endogenous sRNAs were the most predominant (33 to 41%), followed by 21 nt (22 to 24%), 22 nt (13 to 15%), 23 nt (9 to 10%), and 20 nt (7 to 10%) ( Table 2). Among endogenous sRNAs, the predominant 24 nt sRNA reads accumulated at 4.5 to 4.8 and 2.6 to 3.3 million in healthy Arapahoe and Mace, respectively ( Table 2; Fig. 2). However, the profile of endogenous sRNAs changed as 21 nt became the most predominant in all virusinfected wheat samples, except in temperature-sensitive resistant Mace at 18uC ( Table 2; Fig. 2). In single infections, 29 to 39% of   (Table 2). In single infections of Arapahoe with WSMV or TriMV at both temperature treatments and in Mace with WSMV at 27uC, the distribution of sRNAs was significantly different from healthy treatments by chi-square analysis, with the largest increases observed in the 21 nt class (Table 2). Similarly, the shift in the profile of endogenous sRNAs was significantly different from healthy in double infections of Arapahoe at both temperature treatments and in Mace at 27uC, with 21 nt sRNAs as the most abundant (37 to 41%), followed by 22 nt (20 to 22%) and 24 nt (13 to 17%) sRNAs (Table 2). Coinfection caused an increase in the accumulation of 21 and 22 nt sRNAs and a decrease in the accumulation of 24 nt sRNAs. In double infections, 24 nt endogenous sRNAs accumulated at 13 to 17% compared with 33 to 41% and 16 to 25% in healthy and single infections, respectively ( Table 2; Fig. 2), suggesting that coinfection with WSMV and TriMV caused a drastic reduction in the accumulation of 24 nt endogenous sRNAs in wheat cultivars. Co-infection of Arapahoe at both temperatures and Mace at 27uC caused a statistically significant shift in sRNA accumulation compared to healthy wheat cultivars ( Table 2).
In contrast, the profile of endogenous sRNAs in singly and doubly infected Mace at 18uC did not differ significantly from healthy ( Table 2; Fig. 2), most likely due to Mace's resistance to WSMV and TriMV at 18uC. Accordingly, the lack of Mace's infection at 18uC did not alter the profile of endogenous sRNAs, further supporting that virus infection caused a shift in the profile of endogenous small RNAs.

Profile of vsiRNAs in WSMV and/or TriMV-infected wheat cultivars
The most prevalent vsiRNAs from WSMV-and/or TriMVinfected Arapahoe at both temperatures and Mace at 27uC were 21 nt (4.5 to 10.2 million reads; 58 to 73%), followed by 22 nt (1.6 to 5.1 million reads; 23 to 39%) ( Table 3; Fig. 2). The 20, 23 and 24 nt vsiRNAs accumulated at much reduced levels with only 0.3 to 2.8% (Table 3; Fig. 2). Further examination of vsiRNAs revealed differential accumulation of 21 and 22 nt vsiRNAs in singly and doubly infected Arapahoe (     Table S2). The small RNA pool was analyzed independently for WSMV and TriMV-derived vsiRNAs, and the small RNAs that did not align with the plus-and minus-sense genomic strands of WSMV and TriMV were considered as endogenous sRNAs of wheat. c Probability (p) of observing a numerically larger X 2 using healthy sRNA distribution as ''expected'' and treatment sRNA distribution as ''observed''. Arapahoe at 18uC and 27uC, 22 nt vsiRNAs accumulated similarly at 34 to 35% (Table 3). In singly and doubly infected Mace at 18uC, vsiRNAs accumulated to lower levels than in Arapahoe and Mace at 27uC ( Fig. 2; Table 3). In singly infected Mace at 18uC, for 21 and 22 nt vsiRNAs, 12,424 and 4968 reads were detected in WSMVinfected, and 36,088 and 9690 reads were detected in TriMVinfected samples, respectively (Table 3). However, in double infections, 21 and 22 nt vsiRNAs accumulated at higher levels at 90,624 and 37,193 of WSMV-derived and 53,255 and 19,380 of TriMV-derived reads, respectively (Table 3). These numbers correspond to a 7-fold increment for WSMV reads and to a 1.5to 2-fold increment for TriMV reads in doubly infected samples. Reduced levels of vsiRNAs accumulation in Mace at 18uC are consistent with its tolerance to WSMV and TriMV. Hence, reduced levels of viral RNAs were available as a substrate for DCLs to process, which led to basal levels of vsiRNAs.

Effect of the synergistic interaction between WSMV and TriMV on the biogenesis of vsiRNAs compared with endogenous sRNAs
Accumulation of vsiRNAs and endogenous sRNAs was compared within a sample, but not between the samples because high-throughput sequencing of small RNAs may not reflect accurate quantitative differences between the samples due to variation in the quality and quantity of starting small RNA preparations. Hence, the pool of small RNA reads of deep sequencing most accurately represents differences within a sample but not between the samples.
The effect of the synergistic interaction between WSMV and TriMV on vsiRNA biogenesis was examined by comparing vsiRNAs from singly and doubly infected wheat with those of endogenous sRNAs in Arapahoe and Mace. Accumulation of 20 to 25 nt endogenous sRNAs and vsiRNAs in singly and doubly infected Arapahoe and Mace is presented in Table 4. In singly infected Arapahoe at 18uC and 27uC, WSMV-and TriMVderived vsiRNAs accumulated to 91 to 95% and 106 to 120% of corresponding endogenous sRNAs, respectively. In contrast, double infections in Arapahoe caused increased accumulation of WSMV-derived vsiRNAs but not those of TriMV when compared with endogenous sRNAs (Table 4). In Mace at 27uC, WSMV-and TriMV-derived vsiRNAs accumulated at 176% and 100%, respectively, of endogenous sRNAs in single infections compared with 215% and 138% in double infections (Table 4). These data suggest that double infections in Arapahoe at 18uC and 27uC caused a significant increase in the accumulation of WSMVderived vsiRNAs, but little or no increase in TriMV-derived vsiRNAs compared with endogenous sRNAs. In contrast, double infections in Mace at 27uC caused a substantial increase in accumulation of both WSMV-and TriMV-derived vsiRNAs compared with endogenous sRNAs. However, in singly and doubly infected Mace at 18uC, vsiRNAs accumulated at much reduced levels compared with Arapahoe (at 18uC and 27uC) and Mace (at 27uC) ( Table 4). A slightly increased level of vsiRNAs accumulation in doubly infected Mace at 18uC compared with those of single infections could be due to marginally increased virus replication in co-infected plants [8].

The polarity of vsiRNAs
The polarity of the most abundant 21 and 22 nt redundant vsiRNAs was analyzed and presented in Figure 3. The plus-and minus-sense polarity of vsiRNAs of all the samples accumulated at 52 to 59% and 41 to 48%, respectively, except for an equimolar accumulation of plus-and minus-sense 22 nt vsiRNAs of WSMV in doubly infected Arapahoe at 27uC (Fig. 3). More plus-sense vsiRNAs accumulated in TriMV-infected samples (56 to 59%) compared to WSMV (50 to 56%) (Fig. 3), suggesting that slightly more vsiRNAs originated from the genomic RNA of TriMV compared to that of WSMV.
Analysis of strand polarity of nonredundant vsiRNAs revealed a different picture from those of redundant vsiRNAs (Table 5). Nonredundant vsiRNAs of WSMV origin accumulated at an equimolar ratio of plus-and minus-sense polarity in single and double infections in Arapahoe at 18uC and 27uC and Mace at 27uC (Table 5). However, nonredundant vsiRNAs of TriMV origin accumulated with a moderate excess of plus-sense polarity (54 to 55%) compared with those of minus-sense polarity (46 to 45%). In Mace at 18uC, fewer WSMV-and TriMV-derived nonredundant vsiRNAs accumulated in singly and doubly infected plants, but vsiRNAs of plus-sense polarity accumulated at 53 to 56% compared with 44 to 47% of minus-sense polarity ( Table 5).  (Fig. 4). The second most predominant 59terminal nucleotide in the vsiRNAs of TriMV in single and double infections was U in 24 to 28% of the population while nucleotide C was the least predominant 59-terminal nucleotide in 18 to 20% of vsiRNAs (Fig. 4). In WSMV-specific vsiRNAs, nucleotide C or G was found as the second most predominant 59-terminal nucleotide in 24 to 27% of vsiRNAs in singly and doubly infected plants (Fig. 4). These data suggest that a modest number of vsiRNAs contain a 59-terminal nucleotide as an A, and nucleotide C, G, or U was present as a 59-terminal nucleotide in the remainder of vsiRNAs. However, there was no strong bias toward a particular nucleotide for the 59-terminal nucleotide.
We also analyzed nonredundant vsiRNAs for the 59-terminal nucleotide from all the samples. In WSMV-and/or TriMVinfected samples, the predominant 59-terminal nucleotide of nonredundant 21 and 22 nt vsiRNAs was A (30 to 33%), followed by G (23 to 27%), U (20 to 28%), and C (18 to 22%) (Fig. 4). These data suggest that nucleotide A was moderately predominant and nucleotide C was the least predominant 59-terminal nucleotide in nonredundant vsiRNAs (Fig. 4).

Distribution of vsiRNAs in the genomes of WSMV and TriMV
Because a large number of WSMV-and TriMV-derived redundant vsiRNAs accumulated in infected wheat, the distribution of 21 to 24 nt vsiRNAs along the corresponding genomic RNAs of WSMV or TriMV was examined to reveal any hot-and cold-spot regions for vsiRNAs. A region of 5 nt or more without any 21 to 24 nt vsiRNA reads was considered to be a 'cold-spot', and a peak with more than 50,000 reads was considered to be a hot-spot for vsiRNAs.
The genomes of WSMV and TriMV were covered completely with both plus and minus-sense polarity vsiRNAs of 21 to 24 nt in all the samples, except in Mace at 18uC. In Mace at 18uC, fewer vsiRNAs accumulated, especially in single infections with a few cold-spots without vsiRNAs in the genomes of WSMV and TriMV. WSMV-derived vsiRNAs in Mace at 18uC covered most of the genomic RNA except nt 2719, 3503 to 3505, 3843 to 3844, and 5475 to 5478. However, in doubly infected Mace at 18uC, Table 3. Percent 20 to 25 nt vsiRNA reads in WSMV and/or TriMV-infected wheat cv. Arapahoe and Mace at 27uC and 18uC*. The number of vsiRNA reads were normalized against Arapahoe 18uC (see Table S2 (Fig. S1). Similar cold-spot regions were found in the TriMV genome with respect to redundant and nonredundant vsiRNAs.

Hot-spots for vsiRNAs in the genomes of WSMV and TriMV
In WSMV and/or TriMV-infected Arapahoe at 18uC and 27uC and Mace at 27uC, 6.3 to 16.2 million reads of 20 to 25 nt plus-and minus-sense vsiRNAs accumulated in each virus infected sample (Table 3). Several peaks representing multiple vsiRNA reads were found along the plus and minus-sense genomic RNAs of WSMV and TriMV (Fig. 5). The location of peaks with respect to position in the WSMV or TriMV genome was conserved in singly and doubly infected wheat cultivars at 18uC and 27uC. However, the number of vsiRNA reads varied among different treatments of WSMV or TriMV (Fig. 5).
A clear difference in size and composition of vsiRNA peaks was observed between WSMV-infected Arapahoe 18uC and 27uC samples (Fig. 5). In singly and doubly infected Arapahoe at 27uC, two large vsiRNA peaks were clustered predominantly with 22 nt vsiRNAs toward the 39 end of the genomic RNA between nt 9100 to 9131 and 9231 to 9291 (Fig. 5). However, much smaller peaks were observed at similar locations with 21 and 22 nt vsiRNAs in Arapahoe at 18uC. Similar hot-spots, as observed in Arapahoe at 27uC, were also found in singly and doubly infected Mace at 27uC (Fig. 5).
Several vsiRNA hot-spots in the TriMV genome were also found in singly and doubly infected Arapahoe at 18uC and 27uC and Mace at 27uC (Fig. 5). The hot-spots were spread throughout the TriMV genome in singly infected Arapahoe at 18uC and 27uC. However, internal hot-spots in the TriMV genome disappeared in doubly infected Arapahoe at 27uC and were reduced to 3 in Arapahoe at 18uC (Fig. 5). In Mace at 27uC, two internal hot-spots were detected in singly infected TriMV, but the number of hot-spots increased toward the termini of genomic RNA in the doubly infected sample (Fig. 5).
In singly and doubly infected Mace at 18uC, fewer WSMV-and TriMV-derived vsiRNAs accumulated due to its temperature sensitive-resistance, and peaks with fewer vsiRNAs were found at the same position as in Mace at 27uC (Fig. 5).

Discussion
Synergistic interactions among unrelated viruses in plants have been well-documented [2], but the mechanisms behind these interactions have not been deciphered. It has been postulated that the combination of silencing suppressor proteins of two unrelated viruses additively counteract host RNA silencing defense mechanisms, leading to an effective co-invasion of plants by unrelated viruses [2,5,6]. In this study, the endogenous sRNA and vsiRNA profiles in singly and doubly infected wheat cultivars were examined in an effort to understand the role of RNA silencing on synergistic interaction between WSMV and TriMV in wheat.
Even though WSMV and TriMV are distantly related type members of two distinct genera in the family Potyviridae [47,48], co-infection of these two viruses interacted synergistically in wheat cultivars Arapahoe and Tomahawk and in Mace at 27uC [8,52]. We found a large number of WSMV and TriMV-derived vsiRNAs accumulated in singly and doubly infected Arapahoe at both temperatures (18uC and 27uC) and in Mace only at 27uC, indicating that virus infection triggered antiviral RNA silencing responses in these hosts. Additionally, accumulation of abundant vsiRNAs in doubly infected wheat cultivars shows that the presence of two silencing suppressors did not inhibit the biogenesis of virus-derived small RNAs. Our previous work demonstrated that silencing suppressor proteins of WSMV and TriMV repressed host RNA silencing by inhibiting production of vsiRNAs [55,56]. Taken together, these data suggest that synergism results from events that occur downstream of vsiRNA biogenesis; hence, synergistic interaction and RNA silencing are likely unrelated.
A. thaliana and rice encode four and eight DCL proteins, respectively [33,34]; however, there is no information on the numbers and roles of DCL proteins in wheat. The roles of four A. thaliana DCL proteins in the biogenesis of 21 to 24 nt vsiRNAs have been thoroughly characterized [61]: DCL2 and DCL4 produces 22 nt and 21 nt vsiRNAs, respectively, while DCL3 is involved in the biogenesis of 24 nt vsiRNAs in the absence of DCL4 and DCL2 [23,36]. In healthy Arapahoe and Mace, 24 nt endogenous sRNAs are the most predominant, followed by 21, 22, and 23 nt. However, virus infections caused a dramatic shift in the accumulation of endogenous sRNAs and 21 nt became the most prevalent size class in all hosts, except in Mace at 18uC, which is resistant to infection and accumulates the lowest levels of viral RNA [8,53]. Virus infections also caused an increased accumulation of 22 nt endogenous sRNAs in singly and doubly infected plants compared to healthy wheat cultivars, except in Mace at 18uC. Our data indicate that WSMV and/or TriMV infections up-regulated production of 21 and 22 nt sRNAs and downregulated biogenesis of 24 nt endogenous sRNAs. The shift in the profile of endogenous sRNAs is greater in doubly infected than in singly infected plants. These data suggest that in the absence of virus infection, DCL3 is involved in the biogenesis of 24 nt endogenous sRNAs, but it appears that virus infection changes DICER preference to DCL4 and DCL2 for the biogenesis of 21 and 22 nt endogenous sRNAs, respectively. It is conceivable that in healthy plants, DCL1 and DCL3 are active in regulating growth and development, but upon virus infection, plants shift to DCL2 and DCL4 to protect from invading viral pathogens.
Plants have evolved to tolerate biotic and abiotic stress by reprogramming gene expression via the biogenesis of endogenous small RNAs [62]. The up-regulation of endogenous 21 and 22 nt sRNAs and down-regulation of 24 nt sRNAs in virus-infected wheat cultivars suggests that biogenesis of endogenous sRNAs is reprogrammed in virus-infected wheat. This hypothesis is further supported as increased differential accumulation of 21, 22, and 24 nt sRNAs was observed in a doubly infected susceptible wheat cultivar with essentially no change in singly and doubly infected temperature-sensitive resistant wheat cv. Mace at 18uC. It has been reported that stress-induced sRNAs might down-regulate their target genes [63], which may encode negative regulators of stress responses. Conversely, down-regulation of small RNAs in response to stress might cause increased accumulation of their target mRNAs, leading to a positive contribution to stress adaptation. Identification of the target genes for 21, 22, and 24 nt endogenous sRNAs in wheat is beyond the scope of this work. Future research would decipher the target genes of these endogenous sRNAs, enabling better understanding of host-virus interactions. Alternatively, down-regulation of endogenous 24 nt sRNAs could be directly induced by virus infection or could result indirectly from changes in gene expression in response to virus infection.
In RNA silencing, 21 and 22 nt vsiRNAs play a crucial role in combating viral infections [30]. In WSMV-and TriMV-infected plants, 93 to 98% of vsiRNAs are composed of 21 and 22 nt. Double infections in Arapahoe at 18uC and 27uC and Mace at 27uC induced severe symptoms [8] and caused an increased accumulation of vsiRNAs compared to single infections (Fig. 1). In singly and doubly infected Arapahoe, 20 to 25 nt vsiRNAs accumulated at 91 to 120% and 87 to 142% of endogenous sRNAs, respectively, indicating efficient biogenesis of vsiRNAs in doubly infected Arapahoe. In contrast, vsiRNAs in singly and doubly infected Mace at 27uC accumulated at substantially elevated levels of 100 to 176% and 138 to 215%, respectively, of endogenous sRNAs. These data suggest that compared to Arapahoe, the host defense system in Mace at 27uC effectively counteracted invading viruses by processing viral RNAs into vsiRNAs, thus causing a surge in the production of vsiRNAs. These data, together with increased accumulation of vsiRNAs in doubly infected wheat cultivars in Northern blots (Fig. 1) further suggest that synergism and vsiRNA accumulation are most likely not related.
Polarity analysis of redundant vsiRNAs of WSMV and TriMV from singly and doubly infected plants revealed that the plus-sense vsiRNAs were predominant than the minus-sense, suggesting that the majority of vsiRNAs were generated from plus-sense genomic RNAs. In contrast, the plus-and minus-sense nonredundant vsiRNAs exhibited stark differences between WSMV and TriMV infection with symmetrical and asymmetrical accumulation, respectively. The polarity of nonredundant vsiRNAs might give some cues on the vsiRNA biogenesis: WSMV nonredundant vsiRNAs originate from viral dsRNAs and subsequent biogenesis of redundant vsiRNAs might be a combination of primary vsiRNAs from dsRNAs and secondary vsiRNAs by host RDRs. It is possible that TriMV-specific vsiRNAs were generated using the latter method. Several studies indicate that cloned vsiRNAs are predominantly plus-sense polarity in nature compared to those of minus-sense polarity as exemplified in Turnip mosaic virus [64], Potato virus X [65], and Citrus tristeza virus [66]. The polarity of vsiRNAs of Rice stripe virus are host dependent [67] and some other viruses produced an approximately equal ratio of plus-and minus-sense polarity [for eg. [68][69][70][71].
The 59-terminal nucleotides of vsiRNAs determine the vsiRNA loading preference into specific AGO complexes: vsiRNAs with a 59-terminal U nucleotide preferentially load into the AGO1 complex, nucleotide A into AGO2 and AGO4, and C into AGO5 [72][73][74][75]. The 59 nucleotide in 32 to 37% of vsiRNAs in singly and doubly infected plants of two wheat cultivars is an A. The 59nucleotide of vsiRNAs of WSMV and TriMV approximately followed the nucleotide composition of their genomic RNAs with A(32%).U(24%).C(20%).G(24%) in WSMV and A(30%). U(28%).C(18%).G(24%) in TriMV. Hence, there is no preferential accumulation of vsiRNAs based on 59 nt. In TriMV and WSMV, vsiRNAs with 59-nucleotide A and U comprise about 56 to 58%; these vsiRNAs might preferentially load into wheat homologs of AtAGO1, AtAGO2 and AtAGO4. vsiRNAs were generated from most of the genomic RNAs of WSMV and TriMV in wheat cultivars, suggesting that DCL enzymes target most of genomic RNAs. However, some regions of the WSMV and TriMV genomes were found to be hot-spots for preferential excessive vsiRNA accumulation of more than 50,000 reads. More hot-spot regions in the genomes of TriMV and WSMV were found in singly and doubly infected wheat at 18uC than 27uC, suggesting that DCL enzymes at 18uC preferentially cleave more regions than at 27uC. In addition to hot-spot regions, we also found several cold-spot regions of 5 to 21 nt long in the genomic RNA of TriMV with no vsiRNA reads. In the WSMV genome, though no cold-spot regions were found, some regions had fewer reads than others. The genomic regions with hot-and cold-spots for vsiRNAs might depend on the accessibility of RNA to DCL enzymes, or the visRNAs may preferentially accumulate by associating with AGOs or with the silencing suppressor. The hot-spot regions of genomic RNA might have extensive secondary structures that are favorable for easy access to DCL enzymes. The WSMV and TriMV genomic regions with hot-spot regions were folded into stem-and-loop structures in the mFOLD program (data not shown). However, the in vivo nature of these secondary structures needs to be verified by probing with nuclease treatment.