Requirement for Host RNA-Silencing Components and the Virus-Silencing Suppressor when Second-Site Mutations Compensate for Structural Defects in the 3′ Untranslated Region

ABSTRACT

Turnip crinkle virus (TCV) contains a structured 3′ region with hairpins and pseudoknots that form a complex network of noncanonical RNA:RNA interactions supporting higher-order structure critical for translation and replication. We investigated several second-site mutations in the p38 coat protein open reading frame (ORF) that arose in response to a mutation in the asymmetric loop of a critical 3′ untranslated region (UTR) hairpin that disrupts local higher-order structure. All tested second-site mutations improved accumulation of TCV in conjunction with a partial reversion of the primary mutation (TCV-rev1) but had neutral or a negative effect on wild-type (wt) TCV or TCV with the primary mutation. SHAPE (selective 2′-hydroxyl acylation analyzed by primer extension) structure probing indicated that these second-site mutations reside in an RNA domain that includes most of p38 (domain 2), and evidence for RNA:RNA interactions between domain 2 and 3′UTR-containing domain 1 was found. However, second-site mutations were not compensatory in the absence of p38, which is also the TCV silencing suppressor, or in dcl-2/dcl4 or ago1/ago2 backgrounds. One second-site mutation reduced silencing suppressor activity of p38 by altering one of two GW motifs that are required for p38 binding to double-stranded RNAs (dsRNAs) and interaction with RNA-induced silencing complex (RISC)-associated AGO1/AGO2. Another second-site mutation substantially reduced accumulation of TCV-rev1 in the absence of p38 or DCL2/DCL4. We suggest that the second-site mutations in the p38 ORF exert positive effects through a similar downstream mechanism, either by enhancing accumulation of beneficial DCL-produced viral small RNAs that positively regulate the accumulation of TCV-rev1 or by affecting the susceptibility of TCV-rev1 to RISC loaded with viral small RNAs.

IMPORTANCE Genomes of positive-strand RNA viruses fold into high-order RNA structures. Viruses with mutations in regions critical for translation and replication often acquire second-site mutations that exert a positive compensatory effect through reestablishment of canonical base pairing with the altered region. In this study, two distal second-site mutations that individually arose in response to a primary mutation in a critical 3′ UTR hairpin in the genomic RNA of turnip crinkle virus did not directly interact with the primary mutation. Although different second-site changes had different attributes, compensation was dependent on the production of the viral p38 silencing suppressor and on the presence of silencing-required DCL and AGO proteins. Our results provide an unexpected connection between a 3′ UTR primary-site mutation proposed to disrupt higher-order structure and the RNA-silencing machinery.

INTRODUCTION

The structure of the genomic RNA (gRNA) of positive-strand RNA viruses participates in many basic functions, including replication, translation, and regulation of the competing processes of translation and transcription (1–3). Studies examining the secondary structure and tertiary interactions of full-length gRNAs of human immunodeficiency virus, satellite tobacco mosaic virus (STMV), and tomato bushy stunt virus (TBSV) suggest that viral gRNAs fold into distinct structural domains of different shapes and sizes that protrude from a central backbone (4–6). For TBSV, specific inter- and intradomain long-distance RNA:RNA interactions regulate replication, subgenomic RNA (sgRNA) synthesis, cap-independent translation, and ribosome recoding (6–9). In addition, gRNA structure can both promote and provide an escape from host defense responses. For example, extensive RNA secondary structures throughout a gRNA correlate with virus persistence (10), and the degree of RNA folding can be inversely correlative with production of antiviral responses in host cells (11). Highly structured viral genomes can also provide structural barriers for target site accessibility of the host RNA-silencing machinery (12–14).

Turnip crinkle virus (TCV), a small positive-strand RNA virus in the genus Carmovirus (family Tombusviridae), lacks a 5′ cap and 3′ poly(A) tail. The 4,054-nucleotide (nt) TCV gRNA has a 63-nt 5′ untranslated region (UTR) followed by 5 overlapping open reading frames (ORFs) (Fig. 1A). The 5′-proximal ORF encodes the replication-required protein p28 and ribosome readthrough product p88, which is the viral RNA-dependent RNA polymerase (RdRp). p8 and p9 movement proteins are synthesized from the larger of the two sgRNAs, and the p38 coat protein is translated from the smaller sgRNA (15–17). In addition to its role in capsid formation, p38 functions as a silencing suppressor (18, 19), likely through a combination of activities that include binding to dsRNAs of different sizes (20, 21) and by interaction with Argonaute1 (AGO1) and AGO2, which confer slicer activity on the RNA-induced silencing complex (RISC) and are major effectors of antiviral defense (22–24). Interaction of p38 with AGO1 is proposed to inhibit loading of host microRNAs (miRNAs) and viral small interfering RNAs (siRNAs) into the RISC, which affects homeostasis of DCL proteins in Arabidopsis thaliana, including downregulation of DCL3 and DCL4 (25, 26).

FIG 1.

Complex interactions within the 3′ region of TCV. (A) Genome organization of TCV. The genomic RNA and two subgenomic RNAs are depicted. (B) Secondary and tertiary interactions near the 3′ end of TCV. The presence of hairpins and pseudoknots (arrowheads with dotted black lines) has been confirmed by genetic mutational analyses (30, 31). The termination codon for the p38 ORF is shown in red. Interactions defined by examining structural changes in response to specific mutations are designated by blue lines. Interactions defined by the location of second-site mutations that arose in response to primary mutations are designated by green lines. The location of the TSS 3′cap-independent translation enhancer is shown. Hairpin names are boxed. (C) Primary mutations (m21) in the asymmetric loop of H4 (H4AL) that are the subject of the current study are shown. H4TL, H4 terminal loop.

Biochemical structure probing combined with genetic, phylogenetic, and computational analyses has identified several key structural elements in the TCV gRNA 3′ UTR and 3′-proximal portion of the p38 ORF that function in replication, translation, and ribosome recoding (Fig. 1B). The 3′-terminal hairpin (Pr), which is structurally conserved in all carmoviruses, serves as the core promoter for minus-strand synthesis of a TCV-derived satellite RNA (satC) and is also important for TCV replication (27). The Pr terminal loop additionally engages in a long-distance RNA:RNA interaction with a bulge loop in a hairpin just downstream from the ribosome readthrough site at the termination of the p28 ORF (7). 3′ UTR hairpins H4a, H4b, and H5 are important for replication of satC (28, 29) and, along with pseudoknots Ψ₂ and Ψ₃, form a T-shaped structure (TSS) in the gRNA that functions as a translational enhancer through interaction with the 60S ribosomal subunit and/or 80S ribosome (30, 31). Upstream of the TSS is H4, which is critical for both replication and translation (31–33).

Widespread structural changes that occur throughout the 3′ UTR when specific 3′ hairpins/pseudoknots were mutated led to the proposal that these 3′-terminal elements participate in a complex network of noncanonical RNA:RNA interactions that support an integrated higher-order structure in the region (32) (Fig. 1B). Support for this proposition also came from finding that addition of purified TCV RdRp causes a widespread conformational shift affecting nearly all elements in the 3′ UTR and extending into the p38 ORF (33). This extensive conformational change disrupts the TSS, suggesting that RdRp binding mediates the transition between gRNA translation and replication. H4 occupies a central position in this network of structural interactions, and the flexibility of residues in its terminal loop (H4TL) and internal asymmetric loop (H4AL) was substantially altered following RdRp binding to the 3′ UTR (32–34). Specific mutations in H4TL and H4AL as well as in the A-rich sequence just upstream of the hairpin affected virus accumulation by restricting translation enhancement by the TSS, reducing cell-free transcription by the TCV RdRp and altering the RNA structure of the Pr terminal loop (32–34). Second-site changes that reversed the observed structural changes and that significantly restored virus accumulation were located mainly in several clusters throughout the 3′ UTR as well as in nearby upstream sequences, suggesting that the primary mutations do not disrupt a specific RNA:RNA interaction but rather affect regional RNA conformation that is critical for virus viability (32).

Second-site mutations in response to several different 3′ UTR primary-site mutations were also located throughout a 400-nt region within the p38 ORF, but whether these distal alterations were compensatory was not examined. In this study, second-site mutations that arose in the p38 ORF in response to a primary mutation in H4AL were investigated. Two independent second-site changes and one set of 3 second-site changes partially restored virus levels in conjunction with identical additional alterations in the vicinity of the primary-site mutation. Although one distal second-site change altered the RNA structure in the vicinity of H4, the compensatory attributes of the two single second-site changes were dependent on the production of p38 and on the presence of DCL and AGO proteins. These results provide an unexpected connection between TCV 3′ UTR primary-site mutations and the RNA-silencing machinery.

MATERIALS AND METHODS

Generation of constructs.

Oligonucleotide-mediated site-directed mutagenesis was used to generate mutations in pTCV66, which contains full-length TCV cDNA downstream from a T7 RNA polymerase promoter. PCR was performed using Phusion high-fidelity (HF) PCR master mix with HF buffer (NEB) according to the manufacturer's protocol. PCR products were digested with DpnI for 1 h at 37°C, followed by transformation of DH5α cells. All mutations were subjected to regional sequencing to confirm the desired alteration. For agrobacterium infiltration, binary construct PZP-TCVp38, containing mutation G3561A or U3329C in the p38 ORF within vector pRTL2, was used (18). PZP-TCVp38 contains a 35S promoter, TEV translational enhancer, p38 ORF insert, and 35S terminator. Oligonucleotide-mediated site-directed mutagenesis was used to generate mutations. The expression cassette was excised from the vector using PstI digestion and ligated into binary vector pPZP212 to produce constructs p38_G3561A and p38_U3329C.

Isolation of second-site mutations.

TCV-m21, containing a 3-nt alteration in H4AL (3897UUA to ACU), was used to generate the second-site mutations. Plants (Turnip cv Just Right) at the two leaf stage were mechanically inoculated with in vitro-transcribed TCV-m21 RNA (2 μg for each of two leaves), as described previously (35). Total RNA was extracted at 21 days postinoculation (dpi) and used to reinoculate seedlings (5 μg for each of the two leaves). The process was repeated for a total of three times. Total RNA isolated from the third passage was used for reverse transcription-PCR (RT-PCR) amplification of fragments corresponding to the 5′ 1,564 nt or 3′ 900 nt, which were cloned and subjected to sequencing.

Protoplast preparation, inoculation, and RNA gel blots.

TCV gRNA constructs were linearized with SmaI and used as templates for in vitro transcription using T7 RNA polymerase. Protoplasts (Arabidopsis thaliana ecotype Col-0) were prepared from seed callus cultures as previously described (36). To assay for accumulation of TCV gRNA, 20 μg of uncapped, in vitro-transcribed wt TCV or mutant gRNA was inoculated onto protoplasts using 50% polyethylene glycol as previously described (36). Total RNA was extracted at 40 h postinoculation (hpi), subjected to electrophoresis, and transferred to a nitrocellulose membrane, and the gRNA was detected using three [γ-³²P]ATP-labeled oligonucleotides that were complementary to positions 3931 to 3951, 3869 to 3883, and 4035 to 4054 in the 3′ UTR. A two-tailed Student t test for two independent samples in replicates was used to test the null hypothesis that the mean relative accumulations in the protoplasts of TCV-rev1 and those of TCV-rev1 plus a second-site mutation are the same versus the alternative hypothesis that they are not the same.

SHAPE structure probing.

Six picomoles of full-length TCV transcript, synthesized using T7 RNA polymerase, was resuspended in water, then heated at 65°C for 5 min, and snap-cooled on ice for 2 min. SHAPE (selective 2′-hydroxyl acylation analyzed by primer extension) folding buffer [final concentrations of 80 mM Tris-Cl (pH 8), 11 mM Mg(CH₃COO)₂, and 160 mM NH₄Cl] was added, followed by incubation at 37°C for 20 min. Three picomoles of the RNA was combined with N-methylisatoic anhydride (NMIA) to a final concentration of 15 mM, and an equal amount of RNA was combined with the same volume of dimethyl sulfoxide (DMSO) solvent for the negative control. Reaction mixtures were incubated at 37°C for 35 min (5 half-lives of the NMIA). RNA was collected by ethanol precipitation and resuspended in 8 μl of 0.5× Tris-EDTA (TE) buffer. Primer extension reactions were carried out as previously described (37) with Superscript III reverse transcriptase (Invitrogen) and ³²P-labeled oligonucleotides. For domain 1 structural analysis, oligonucleotides were complementary to positions 3871 to 3894, 4005 to 4022, and 4039 to 4054. For extreme 3′-end structure probing of TCV, the 3′ end of the gRNA was extended by inclusion of 106 residues of plasmid sequence (by introduction of a SacII site into the vector) and an oligonucleotide (5′-GCTGTTTCCTGTGTGAAA-3′) corresponding to a region that was 40 bases downstream from the SmaI site in pUC19 was used. For domain 2, the oligonucleotides were complementary to positions 3130 to 3157, 3327 to 3348, 3395 to 3414, 3496 to 3518, 3628 to 3648, and 3707 to 3730. Radioactively labeled products of reverse transcription were resolved on 8% denaturing polyacrylamide gels and visualized using a phosphorimager. Band density was quantitated using Multi Gauge v3.1 software (Fuji Photo Film Co.).

Agrobacterium infiltration.

Agrobacterium infiltration was performed as previously described (18). Individual agrobacterium cultures carrying various constructs were suspended in buffer (10 mM morpholinepropanesulfonic acid [pH 5.5], 10 mM MgCl₂, and 100 μM acetosyringone) to an optical density of 1.0 at 600 nm. Suspensions were incubated at 23°C for 2 h, and equal amounts were mixed for coinfiltrations. Plants (Nicotiana benthamiana) constitutively expressing green fluorescent protein (GFP) (line 16c) were grown at 22°C with a 12-h/12-h day/night cycle. Infiltration was performed at the 4- to 5-week stage using 1-ml needleless syringes. GFP fluorescence was observed under long wavelength using a UV lamp (Black Ray model B 100AP) and photographed using a Nikon D90 digital camera. For detection of GFP, total RNA from the infiltrated leaves was isolated using TRIzol reagent (Invitrogen) according to the manufacturer's protocol and RNA was transferred to a nitrocellulose membrane following electrophoresis. The membrane was hybridized with [α-³²P]dCTP-labeled DNA probes corresponding to the full-length GFP ORF.

Protein gels and immunoblotting.

Total protein was extracted from infiltration sites and subjected to electrophoresis through 12% SDS-PAGE gels. Proteins were transferred onto Immunobilon-P transfer membranes (Millipore) at 10 V for 30 min using a TransferBlot semidry transfer cell (Bio-Rad). p38 was probed using a polyclonal antiserum. A Super Signal West Pico chemiluminescent substrate kit (Pierce) was used for chemiluminescent staining according to the manufacturer's protocol.

RESULTS

Second-site mutations in the p38 ORF along with partial reversion of the 3′ UTR m21 primary-site mutation can partially compensate for reduced accumulation of TCV-m21.

Alterations in H4TL and H4AL affect both translation and replication and cause a number of structural changes throughout the 3′ UTR and nearby upstream region in TCV gRNA (31, 34). One 3-nt alteration in H4AL (UUA to ACU), known as m21, reduced TCV gRNA accumulation in protoplasts by ∼80% (34) and reduced in vivo translation of a luciferase reporter construct containing the TCV 5′ UTR and 400 3′ residues by 77% (31). To determine if H4AL participates in any direct RNA:RNA interactions, a second-site mutation approach was initiated whereby TCV gRNA containing m21 (TCV-m21) was inoculated onto turnip seedlings and both the 5′-terminal 1,564 nt or the 3′-terminal 900 nt of progeny accumulating after three passages were cloned and examined for second-site mutations. As previously described (34), no second-site mutations were recovered in clones containing the 5′ 1,564 nt. Second-site mutations that were recovered in the 3′ UTR included single residue changes in the stem of hairpin H4b, a single change in the loop of H4TL and three bases altered in the Pr loop (UCG→AA). None of these locations contained sequence capable of canonical base pairing with H4AL (for a complete list of TCV-m21 second-site mutations, see Table 1 in reference 34).

A number of TCV-m21 second-site mutations were located within the p38 ORF and were coupled (as were all 3′ UTR second-site changes) with partial primary-site reversions designated “rev1” or “rev2” (UUA [wt]→ACU [m21]→UUU [rev1] or ACU [m21]→UCU [rev2]) (34). In this study, a selection of p38 ORF second-site mutations were examined, all of which were associated with the rev1 partial reversion, which differs from wt TCV at a single position (A3899U) (Fig. 2A). Two second-site mutations examined were single missense mutations recovered in separate plants (G3561A [G→E] and U3329C [F→L]), and one was a series of 3 second-site alterations that were present together in the same clone (A3475G/A3709G/U3741C), which also produced altered p38 (U3741C [V→A]).

FIG 2.

Second-site mutations located in the p38 ORF are compensatory for virus accumulation. (A) Selected second-site mutations, their effects on the sequence of p38, and association with the primary-site partial reversion (rev1) are shown. (B) Effect of second-site mutations on accumulation of wt TCV, TCV-m21, and TCV-rev1 are shown. (Left) G3561A; (middle) U3329C; (right) A3475G, A3709G, and U3741C. Arabidopsis protoplasts (Col-0) were inoculated with full-length transcripts and levels of gRNA determined by RNA gel blots at 40 hpi. Values from independent experiments conducted in triplicate and standard deviations are shown. *, P < 0.05 (Student's t test).

To determine whether any of these second-site alterations were compensatory in conjunction with the initial m21 alteration or the recovered rev1 variant, each second-site mutation was introduced into three full-length gRNA backbones and assayed for accumulation in Arabidopsis thaliana protoplasts (Fig. 2B). These backbones were (i) wt TCV, to determine if the second-site mutations by themselves affected TCV accumulation; (ii) TCV-m21, to assess if the second-site changes improved the accumulation defect of m21 and thus may have arisen prior to the m21 partial reversion to rev1; and (iii) TCV-rev1. Note that TCV-rev1 accumulations were quite variable in these experiments (performed by two individuals using the same starting material), although results were consistent for the individuals performing the experiments. This is highly unusual for TCV mutations and cannot currently be explained. Thus, for each second-site mutation, all constructs, including wt TCV, TCV-m21, and TCV-rev1, with and without second-site mutations, were assayed at the same time by the same individual. At least three independent experiments were performed for each alteration before final averaged values were obtained.

TCV gRNA that includes second-site mutation G3561A (TCV-G3561A) accumulated to 38% of wt TCV (Fig. 2B, left graph), suggesting that the mutation negatively impacts either p38 function or gRNA structure. TCV-m21 accumulated to 17% of wt TCV, and addition of G3561A did not appreciably improve accumulation. TCV-rev1 accumulated to 75% of wt TCV, indicating that rev1 partially reverses the defect caused by the m21 mutations (Fig. 2B, left graph). TCV gRNA containing both rev1 and G3561A (TCV-rev1+G3561A) improved accumulation of TCV-G3561A over 2-fold to 88% of wt TCV levels. These results suggest that G3561A did not arise prior to the m21 partial reversion to rev1 and that the combination of rev1 (U3899A) and G3561A is synergistic for improved accumulation of TCV gRNA.

Second-site mutation U3329C had only a slight effect on TCV gRNA levels, with TCV-U3329C accumulating to 88% of wt TCV (Fig. 2B, middle graph). As with G3561A, U3329C had no beneficial effect on accumulation of TCV-m21. TCV-rev1, in this set of experiments, accumulated to an average of 42% of wt TCV, and addition of U3329C enhanced accumulation 1.6-fold. These results suggest that, as with G3561A, U3329C likely originated in the TCV-rev1 background and was able to partially compensate for the rev1 (U3899A) alteration.

The second-site mutation set A3475G/A3709G/U3741C was evaluated in wt TCV for A3475C alone, A3709G and U3741C together, and all three alterations combined. TCV with A3475C accumulated slightly better than wt TCV (125%), whereas TCV containing both A3709G and U3741C accumulated to a level similar to that of wt TCV. gRNA containing all 3 second-site alterations accumulated to only 30% of wt TCV (Fig. 2B, right graph), suggesting that all together, these mutations are detrimental for gRNA accumulation. A similar negative effect of the three mutations was found when combined with TCV-m21. TCV-rev1 accumulated to an average of 65% of wt TCV in these experiments, and addition of all 3 second-site mutations enhanced accumulation to wt TCV levels. This suggests that one or more of the alterations in this second-site mutation set is compensatory with rev1.

Second-site mutations are located within a discrete RNA domain.

We previously found that second-site alterations in the Pr loop that were isolated together with rev1 affected the structure of the 3′ UTR (34). In addition, several second-site mutations in the 3′ UTR that compensated for primary mutations in H4TL were compensatory for virus accumulation and eliminated RNA structural changes caused by the primary-site mutations (32, 34). To determine if G3561A and U3329C affect local (p38 ORF) RNA structure and/or structure within the 3′ UTR, the secondary structure of most of the p38 ORF was mapped using SHAPE (selective 2′-hydroxyl acylation analyzed by primer extension) structure probing combined with mFold computational predictions (38). SHAPE takes advantage of the electrophile N-methylisatoic anhydride (NMIA), which reacts with flexible 2′-OH groups and impedes the progression of reverse transcriptase during primer extension (39, 40). SHAPE therefore uses a single reagent to interrogate the flexibility of all four nucleotides and can be applied to full-length gRNA.

The location of all flexible residues in the secondary structure of the wt TCV p38 ORF region is presented in Fig. 3A. Whereas most of the flexible residues mapped to single-stranded terminal loops, bulged loops, and base pairs next to loops, a notable exception was the stem flanking the 3′ base of hairpin H2-2 (stem 1). Stem 1 had a number of reactive bases mainly on one side of the stem that likely denote instability of a poorly stable structure (7/10 G-U or A-U pairs and a central A-A pair) and/or participation of that region in a higher-order structure that cannot be currently be ascertained. G3561A is located near the base of H2-2, U3329C is located in one of the bulge loops of H2-4 near two other second-site changes, A3475G is within H2-3, and the two companion mutations are located downstream of this region (data not shown). Eight of 11 second-site changes in progeny of TCV-m21 were predicted to weaken hairpin stems, and none are in sequences that might canonically pair with H4AL. Thus, the compensatory nature of these second-site mutations likely involves a mechanism that does not involve direct canonical base pairing with the rev1 H4AL sequence.

FIG 3.

Locations of second-site mutations in the structure of domain 2. (A) SHAPE-derived structure that includes most of the p38 ORF (2744 to 3799). Residues corresponding to high and low reactivities to NMIA are in red and green, respectively. Names of hairpins and locations of second-site mutations are indicated. Identical symbols denote association in the same clone. Second-site mutations investigated in this study are in blue. (B) mFold-generated secondary structure of full-length TCV gRNA. Structural predictions used constraints according to SHAPE-derived flexibility information and included placing stem 1 at the base of hairpin H2-2. Locations of domain 1 and domain 2 are shown.

Secondary-structure predictions of full-length TCV gRNA by mFold, which take into account the SHAPE data generated for the p38 ORF region, suggest that the genome is organized into multiple domains emanating from a central backbone, similar to the SHAPE-determined genome structure of TBSV and STMV (4, 6). This structure prediction places most of the p38 ORF into a domain (domain 2) that is separate from the domain (domain 1) containing the 3′ region of TCV gRNA and known 3′ hairpins, including H4, the site of the primary mutations. Thus, any direct RNA linkages between H4 and the second-site mutations would need to involve long-distance tertiary interactions.

Effect of second-site mutations on the RNA structure of domain 2.

The effect of second-site mutations G3561A and U3329C on the structure of domain 2 was examined by SHAPE analysis of full-length TCV-G3561A and TCV-U3329C. G3561A, located near the base of hairpin H2-2, enhanced the flexibility of residues on both sides of the lower stem (Fig. 4A and B, compare lanes 2 and 4), supporting the existence of the lower H2-2 stem (Fig. 4D, right). However, G3561A also enhanced flexibility of residues on only the 5′ side of the upper H2-2 stem, suggesting that the upper portion of the hairpin adopts a higher-order structure different from the structure shown in Fig. 4D. G3561A also affected residue flexibility in the adjacent poorly organized stem 1 sequence, with 3564GGG showing reduced flexibility and A3567 becoming more flexible (Fig. 4B). These changes in residue flexibility produced a profile that was more consistent with the predicted secondary structure of the stem, suggesting that G3561A also disrupts higher-order structure in the region. No discernible differences were found within H2-3 and H2-4 (data not shown). However, G3561A enhanced the flexibility of residues in the loop of hairpin H2-5, which is predicted to be spatially proximal to H2-2 within domain 2 (Fig. 4C, compare lanes 2 and 4). Enhanced flexibility in the base and upper 5′ side of H2-2, adjacent stem 1, and upstream H2-5 due to a single nucleotide change near the base of H2-2 suggests that tertiary interactions may connect H2-2, H2-5, and stem 1. Domain 2 SHAPE structure profiles were also determined for TCV-m21, TCV-rev1, TCV-m21+G3561A, and TCV-rev1+G3561A (Fig. 4). No changes in residue flexibility compared with wtTCV were discernible for TCV-m21 and TCV-rev1, and profiles were very similar for all gRNAs containing G3561A (i.e., TCV-G3561A, TCV-m21+G3561A, and TCV-rev1+G3561A).

FIG 4.

Effect of G3561A on the structure of resident hairpin H2-2 and other proximal regions. (A) SHAPE autoradiogram showing the effect of domain 1 and domain 2 mutations on the structure of the RNA in the vicinity of H2-2. RNAs subjected to SHAPE are designated above the lanes. A, G, C, and U, ladder lanes; N, NMIA; D, DMSO. The arrow shows the position of G3561. Orange circles and yellow circles indicate residues with increased and reduced flexibilities, respectively, compared with that of wt TCV. No other changes in the flexibility of residues were found within domain 2 (data not shown). (B) Shorter run of the samples shown in panel A. (C) Region of hairpin H2-5 showing additional effects of G3561A. (D) Mapping the flexibility of residues in wt TCV and TCV-G3561A. Strong and weak reactivities to NIMA are indicated by red and green colors, respectively. Only changes that were consistent in repeated experiments are indicated by orange (enhanced flexibility) and yellow (reduced flexibility) circles. G3561 and G3561A are boxed.

The domain 2 structure profile was also determined for second-site mutation U3329C. Although U3329 is predicted to reside in a small internal loop, U3329 is not flexible as assayed by SHAPE in wt TCV gRNA. U3329C caused flexibility changes in three nearby residues on the 5′ side of the H2-4 lower stem (G3323, C3325, and U3332) (Fig. 5A and B) and two spatially proximal residues on the opposite side of the stem (A3420, A3421) (Fig. 5A, lower blot, and Fig. 5B), supporting the existence of this stem. This suggests that U3329 is involved in local interactions in the mid-portion of the H2-4 stem and/or that U3329C causes new interactions in the region. Residue flexibilities were very similar for all gRNAs containing U3329C (TCV-U3329C, TCV-m21+U3329C, and TCV-rev1+U229C), and the structure of domain 2 was not detectably altered in the gRNAs of TCV-m21 and TCV-rev1 (Fig. 5A and data not shown). Together, these results suggest that the compensatory effects of G3561A and U3329C in combination with rev1 cannot be explained by any discernible common structural changes in domain 2.

FIG 5.

Effect of U3329C on the structure of resident hairpin H2-4. (A) SHAPE autoradiogram showing structural changes in the vicinity of H2-4 in response to U3329C with and without m21 or rev1. The arrow indicates the location of U3329C. RNAs subjected to SHAPE are designated above the lanes. See the legend to Fig. 4. (B) Comparison of residue flexibility in H2-4 between wt TCV and TCV-U3329C. Strong and weak reactivities to NIMA are indicated by red and green colors, respectively. Only changes that were consistent in repeated experiments are indicated by orange (enhanced flexibility) and yellow (reduced flexibility) circles. No other structural changes were found within domain 2.

G3561A affects the structure of the lower stem of hairpin H4.

The structures of fragments containing portions of domain 1 were previously mapped by in-line probing, a technique that evaluates the flexibility of individual residues within short fragments (30, 32). In-line probing of the 3′ region of domain 1 produced nearly identical results when using fragments that either began just upstream of H4, included the entire 3′ UTR, or began just upstream of hairpin H3 within the p38 ORF (32, 33, 41). For the current study, SHAPE was used to determine the structure of the 3′-end region of domain 1 within the full-length gRNA. The majority of domain 1 residues were similarly flexible as determined by either SHAPE or in-line probing, suggesting that sequences upstream of domain 1 do not interact substantially with most sequences in the domain (Fig. 6A). Major differences with the in-line probing data were found, however, for two pyrimidine-rich segments within the large unstructured region (USR) that links H3 and H4 (positions 3818 to 3827 and 3837 to 3846, boxed in Fig. 6A). These sequences lacked flexibility only when full-length gRNA was probed, suggesting that they interact with sequences absent in 3′-terminal fragments. In addition, several residues in H4TL and three of the five residues of H4AL exhibited different flexibility from the profile previously determined using 3′ UTR fragments (compare H4 in Fig. 6A and B), suggesting that sequences upstream of the 3′ UTR affect the structure of the H4 loops and/or that the different techniques have variable results for residues likely involved in complex higher-order structure (42).

FIG 6.

G3561A enhances the flexibility of the H4 lower stem. (A) Structure of domain 1 determined using SHAPE and full-length TCV gRNA. Residues corresponding to high and low reactivity to NMIA are indicated by red and green colors, respectively. Residues in dashed boxes differed significantly in their stability when assayed using SHAPE of full-length gRNA versus in-line probing of 3′ UTR fragments (32). Residues whose flexibility changes in gRNA containing G3561A are boxed in blue (G3901 and G3902). (B) In-line probing results for H4 using a 3′-terminal fragment beginning at position 3859 (33). Red residues are highly flexible, and green residues are moderately flexible. (C) SHAPE autoradiogram showing the effect of G3561A with and without m21 or rev1 on the structure of the RNA in the vicinity of H4 in domain 1. The locations of hairpins H4 and H4a and the H4 asymmetric loop (H4AL) are shown. Orange and yellow circles indicate residues with enhanced flexibility and reduced flexibility, respectively.

Susceptibility of domain 1 residues to NMIA was determined for the gRNAs shown in Fig. 4 and 5. TCV-m21 had structural changes in H4AL (the local region containing the mutations), with reduced flexibility of G3894, U3898, and U3899 (Fig. 6C, compare lanes 2 and 6). Although these residues are predicted to be unpaired, this result supports our earlier findings that H4AL participates in widespread higher-order interactions in the 3′ region. Rev1 partially restored the wt pattern in H4AL, with the exception that U3898 and A3899 (the mutated residue in rev1) retained their reduced flexibilities (Fig. 6C, compare lanes 2 and 10). Interestingly, domain 2 alteration G3561A generated a reproducible difference within the 3′ UTR of TCV, increasing the flexibility of the H4 lower stem guanylates G3901 and G3902 (Fig. 6C, compare lanes 2 and 4). This enhanced flexibility remained when G3561A was combined with m21 and rev1 (Fig. 6C, compare lanes 4, 8, and 12). In all three repetitions of this experiment, the control (DMSO) lane also contained reverse transcriptase stops at residues G3901 and G3902 in the G3561A and G3561A+rev1 gRNAs (and in the G3561A+m21 gRNA in one experiment), which were absent from other lanes. Second-site mutation U3329C did not produce any discernible structural changes in the H4 region or elsewhere in domain 1 (Fig. 7 and data not shown), and m21 and rev1 produced similar local structural changes in H4AL with or without U3329C (data not shown). These results suggest that while a long-range structural connection appears to exist between domain 1 and domain 2, there are no discernible common RNA structural changes specifically associated with the synergism between rev1 and both G3561A and U3329C.

FIG 7.

Second-site alteration U3329C did not affect the structure of the H4 lower stem. The SHAPE autoradiogram shows no discernible flexibility changes due to U3329C on the structure of the RNA in the vicinity of H4 in domain 1.

Compensatory effects require a functional p38.

p38 is a multifunctional protein that assembles into virions, serves as an effector of virus resistance (43), and mediates suppression of virus accumulation by the virulent satellite RNA satC (44). Importantly, p38 also functions as the TCV silencing suppressor, a critical protein required to suppress the innate RNA-silencing defense system of the host plant (18, 19). Both G3561A and U3329C, as well as U3741C from the triple mutation set A3475G/A3709G/U3741C, are missense mutations and thus may alter one or more of the functional properties of p38. To determine if the beneficial effects of G3561A and U3329C on TCV-rev1 accumulation are connected with an altered p38, G3561A and U3329C were introduced into a TCV backbone (TCVCP-) with an engineered stop codon at the 6th codon position in the p38 ORF that eliminates detectable p38 synthesis (44). In the absence of p38, TCVCP- accumulates to only ∼10% of wt TCV in protoplasts with a functional RNA-silencing system, indicating that p38 is required for efficient virus accumulation due at least in part to its role in suppressing silencing.

Whereas G3561A had a negative effect on wt TCV levels (Fig. 2B, right graph), it had an insignificant effect on TCVCP- accumulation (Fig. 8A, left graph), suggesting that the negative effect of this alteration on wt TCV accumulation is due to a defect in p38 and not in the structure of the RNA. Rev1 reduced accumulation of TCVCP- to an extent similar to that with wt TCV. G3561A did not enhance accumulation of TCVCP-/rev1, suggesting that the beneficial effect of G3561A on TCV-rev1 depends on a functional p38.

FIG 8.

G3561A affects the silencing suppressor activity of p38. (A) Accumulation of viral gRNA incapable of producing p38 (TCVp38-) in Arabidopsis protoplasts (Col-0). (Left graph) Accumulation of TCV-G3561A, TCV-rev1, and TCV-rev1+G3561A in the absence of p38 (right graph) accumulation of TCV-U3329C, TCV-rev1, and TCV-rev1+U3329C in the absence of p38. **, P < 0.01 (Student's t test). (B) (Top) structural domains of p38. R, RNA binding domain; A, Arm; S, surface domain; H, hinge; P, protruding domain. G3561A is located in the P domain, and U3329C is located in the S domain. (Bottom) Pymol-generated 3D structure of p38 showing the location of the GW motifs and the second-site mutations. The N-terminal GW motif is located in an unresolved region in p38. G3561A alters the C-terminal GW motif (red). U3329C modifies a surface nonpolar residue (blue). (C) Effect of second-site mutations on silencing suppressor activity of p38. (Left) expression of a reporter GFP transgene with and without mutant and wt p38 in N. benthamiana leaves. Photographs were taken at 5 days postinfiltration under UV light to show the green fluorescence of GFP and the red fluorescence of chlorophyll. (Right) accumulation of GFP RNA in leaves assayed by RNA blotting using a [α-³²P]dCTP-labeled DNA probe corresponding to full-length GFP ORF. To detect levels of p38, equal amounts of total protein were separated by 12% SDS-PAGE and subjected to immunoblotting using a polyclonal anti-p38 antibody.

As with wt TCV, U3329C did not negatively impact accumulation of TCVCP- (Fig. 8A, right graph). TCVCP-/rev1 accumulated to 82% of TCVCP- levels in this set of experiments, and unexpectedly, addition of U3329C reduced TCVCP-/rev1 levels 4-fold. This significant reduction in TCVCP-/rev1 levels by U3329C suggests that in the absence of p38, U3329C negatively affects gRNA structure when combined with A3899U. Since these two mutations are in separate RNA domains, this further supports possible structural connections between domain 1 and domain 2.

G3561A alters one of two GW motifs in the p38 AGO binding platform.

Since the positive effects of G3561A and U3329C on rev1 were dependent on production of p38, one possibility was that the mutations alter the silencing suppressor activity of p38. p38 contains two GW motifs located proximal to the N and C termini of the protein (25), both of which are necessary for efficient binding to AGO1 and AGO2 (24, 25) and for binding to synthetic 19-nt duplex RNAs (21). The N-terminal GW motif is conserved in nearly all carmoviruses, and the C-terminal motif is unique to TCV. The C-proximal GW motif at position 273 is located on the surface of the protrusion (P) domain of p38, while the N-proximal GW motif at position 25 is in the RNA-binding (R) domain that is structurally unsolved due to disorganization of the protein in this region (Fig. 8B). In addition to the GW motifs, R74, E122, R130, and R137 have been reported to be implicated in p38 silencing suppressor activity (45, 46). However, R74 disrupts several p38 functions, including virion formation, indicating that (at least) this defect causes a general disruption of p38 structure (44). In contrast, R130 did not affect virion formation in dcl2-dcl3-dcl4 triple mutant plants (45).

Figure 8B shows the location of G3561A and U3329C in the three-dimensional (3D) structure of p38. Strikingly, G3561A replaces a nonpolar glycine R group in the C-proximal GW motif with the negatively charged glutamic acid R group (Fig. 8B), thus likely disrupting the interaction of this GW motif with AGO1/AGO2 and double-stranded RNAs (dsRNAs). U3329C, which causes a conservative phenylalanine-to-leucine alteration at position 196, maps to the surface of p38 in a region of the S domain that is distal to the GW motifs. There are currently no specific activities of p38 associated with this region (S. C. Harrison, personal communication).

To determine if G3561A or U3329C affect silencing suppressor activity of p38, cotransient expression assays were conducted by infiltrating Nicotiana benthamiana line 16c with agrobacteria carrying a reporter GFP transgene or the transgene together with wt p38, p38_G3561A, or p38_U3329C. Leaves infiltrated with the GFP transgene alone showed a marked decreased in GFP expression after 5 dpi due to rapid induction of host RNA silencing (Fig. 8C). In contrast, coinfiltration with wt p38 suppressed RNA silencing, allowing for GFP expression and detection. When leaves were infiltrated with GFP along with p38_G3561A, a 40% reduction in GFP mRNA accumulation was observed, supporting previous reports on the importance of at least the C-proximal GW motif for efficient RNA silencing (25). Since G3561A negatively impacts TCV accumulation only when p38 is synthesized, this suggests that one effect of G3561A is to reduce silencing suppressor activity of p38. In contrast, U3329C had no detectable effect on silencing suppressor activity of p38 (Fig. 8C). Therefore, these results suggest that G3561A and U3329C enhance accumulation of TCV-rev1 using different mechanisms; however, both mechanisms require p38.

The compensatory effects of G3561A and U3329C require DCL proteins.

Of the four DCL proteins found in Arabidopsis, DCL4 is the key enzyme in the plant's antiviral defensive arsenal, targeting structured, imperfect duplexes in viral gRNA or double-stranded replicative intermediates (47). In dcl4 mutant backgrounds, DCL2 becomes the principal antiviral factor and may also have an additional role in production of secondary viral small interfering RNA (siRNAs) that are generated following host RdR synthesis of long viral dsRNAs using DCL-generated primers (48). In the absence of p38, TCV accumulates to wt levels in dcl2-dcl4 plants but not in plants with defects at single loci (25). In addition, TCV gRNA in the absence of p38 can accumulate to wt levels in dcl2-dcl4 protoplasts (44), demonstrating that RNA silencing is active in callus-derived single cells.

If G3561A enhancement of TCV-rev1 levels is due to reduced silencing suppressor activity of p38_G3561A, then G3561A should have no positive effect if cells cannot synthesize DCLs that target viral RNAs. To test this prediction, the effects of G3561A and U3329C on TCV-rev1 accumulation were assessed in Arabidopsis dcl2-dcl4 protoplasts. TCV-G3561A accumulated in dcl2-dcl4 cells to 85% of wt TCV, supporting the suggestion that the negative effect of G3561A on wt protoplasts is primarily due to reduced silencing suppressor activity of p38_G3561A (Fig. 9A, left graph). Similar to results obtained using wt protoplasts, G3561A had no effect on TCV-m21 accumulation. As predicted, G3561A also did not enhance levels of TCV-rev1 in dcl2-dcl4 protoplasts, suggesting that DCL2 and/or DCL4is necessary for G3561/rev1 synergy (Fig. 9A, left graph).

FIG 9.

Compensatory effect of the second-site mutations requires DCL2/DCL4 and AGO1/AGO2. (A) Effect of second-site mutations on accumulation of wt TCV, TCV-m21, and TCV-rev1 in dcl2-dcl4 protoplasts. (Left) G3561A; (right) U3329C. (B) Effect of second-site mutations on accumulation of wt TCV, TCV-m21, and TCV-rev1 in ago1-ago2 protoplasts. (Left) G3561A; (right) U3329C. Viral RNA levels were assayed at 40 hpi by RNA gel blots. Values are from independent experiments conducted in triplicate, and standard deviations are shown. *, P < 0.05; **, P < 0.01 (Student's t test).

TCV-U3329C accumulated to near wtTCV levels in dcl2-dcl4 cells (Fig. 9A, right graph), and U3329C had no effect on TCV-m21 levels. Addition of U3329C to TCV-rev1 reduced accumulation by 55% in dcl2-dcl4 cells, similar to the results obtained when U3329C was included with TCVCP-/rev1 in wt protoplasts. These results suggest that, as with G3561A, DCL2/DCL4 are necessary for U3329C to have a positive effect on TCV-rev1, and, in the absence of the DCLs, the combination of U3329C and A3899U (the rev1 alteration) is negative for TCV accumulation.

The compensatory effects of G3561A and U3329C require AGO1/AGO2.

Active DCL proteins do not by themselves limit virus infection (22), suggesting that additional RISC cleavage events are necessary to clear viruses from plants. AGO slicer enzymes are the active components within RISC, and AGO proteins are critical for antiviral defense and are popular targets for virus silencing suppressors, including p38 (47). AGO1 and AGO2, which are two of the 10 AGO proteins in Arabidopsis, act synergistically to control virus infections, and both are targeted by p38 (24, 25).

To determine if AGO1/AGO2 are needed for TCV-rev1 compensation by G3561A and U3329C, TCV containing the second-site changes in combination with the rev1 alteration were assessed in ago1/ago2 protoplasts. TCV-G3561A and TCV-U3229C accumulated in ago1/ago2 protoplasts to 68% and 65% of wt TCV, respectively, and neither second-site mutation significantly affected levels of TCV-m21 or TCV-rev1 (Fig. 9B). The reduction in TCV-G3561A levels compared with wt TCV in ago1/ago2 protoplasts suggests that there is an effect of this alteration in addition to reduced AGO1/AGO2 binding. Together, these results suggest that neither G3561A nor U3329C is compensatory with rev1 in the absence of p38 or when cells lack important components of RNA silencing.

DISCUSSION

The goals of this study were to investigate whether second-site changes that arose in the p38 ORF compensated for primary mutations in the 3′ UTR and, if so, what mechanism(s) might be responsible for such compensation. Primary-site mutations were constructed in H4AL because the internal (and terminal) loops of H4 participate in a network of higher-order structural interactions that connect 3′ elements (32). In-line probing of the 3′ UTR and extended 3′ UTR fragments previously showed that (i) mutations generated in H4AL and H4TL disrupt RNA structure in several downstream regions (32–34), (ii) mutations in the 3′-terminal Pr hairpin and in the pseudoknot that connects 3′-terminal residues to hairpin H5(Ψ₁) alter the flexibility of residues in H4AL (34), and (iii) the flexibilities of H4TL and H4AL residues were altered upon RdRp binding to 3′ UTR fragments as part of a wide-spread conformational shift (33). Higher-order structure in the 3′ region is likely required for important 3′-end-associated activities such as translation mediated by the TSS 3′cap-independent translation enhancer (31), the long-distance interaction involved in ribosome recoding that extends the p28 ORF to synthesize p88 (7), and RdRp recognition of the promoter for complementary-strand synthesis (33). Attempts to detect structural changes caused by H4AL alterations m21 and rev1 outside the H4AL region using SHAPE and full-length TCV gRNA were not successful (data not shown). Likewise, SHAPE analysis of the TBSV genome could account for only two of six known long-distance inter- and intradomain interactions (6). Therefore, SHAPE analysis using NMIA may not be effective for ascertaining disturbances in subtle tertiary interactions, such as those associated with the TCV 3′ UTR, as has been previously reported (42).

m21 primary-site alterations were detrimental for virus accumulation in plants and protoplasts, and all second-site changes that arose in response to m21 were associated with one of two partial reversions, including rev1, which leaves only a single base alteration in H4AL (U3899A) (34). wt TCV was not recovered, even though this would have required only a single alteration from TCV-rev1 to the wt. One possibility is that the selection for second-site mutations was conducted with plants, and assays for accumulation used single-cell protoplasts. It is possible that these second-site changes, in combination with TCV-rev1, provided additional benefits in a multicellular host.

Surprisingly, TCV-rev1 accumulation was quite variable, ranging from 40% to 65% of wt TCV in Col-0 protoplasts in independent experiments (Fig. 2B) and 58% to 75% of wt TCV in dcl2-dcl4 and ago1-ago2 backgrounds (Fig. 9). This highly unusual variation in results following inoculation of a TCV mutant transcript does not have a definitive explanation. One possibility is that the structure of the transcript when inoculated into protoplasts differs in different preparations and experiments (conducted by two individuals), affecting early translation/replication events.

Second-site mutations accumulating in progeny of TCV-m21 transcripts were found in several locations proximal to H4 (34), as well as throughout a 400-nt region within the upstream p38 ORF (this study). When combined with the rev1 alteration, all second-site mutations tested enhanced levels of gRNA accumulation compared with those of TCV-rev1 and/or wt TCV containing the second-site mutations assayed in the same experiment (Fig. 2). None of the mutations compensated for m21 (and many were negative for accumulation), suggesting that they arose subsequent to the conversion of m21 to rev1. U3329C was located very near two other second-site changes present on both sides of domain 2 hairpin H2-4 (Fig. 3A), suggesting a possible role for regional RNA structure disruption in the U3329C compensatory mechanism. Primary mutations in the 3′-terminal Pr hairpin loop, a region that also affects the higher-order structure of the 3′ UTR (34), gave rise to second-site mutations that cluster in the same vicinity on both sides of the H2-4 stem as well as in domain 1 locations very similar to second-site mutations of m21 (X. Yuan and A. E. Simon, unpublished data). We are currently testing if other clustered second-site mutations in domain 1 and domain 2 are compensatory in the presence of rev1 or Pr terminal loop mutations.

When translation of p38 was blocked, U3329C caused a 4-fold decrease in levels of gRNA with the A3899U rev1 alteration, supporting the hypothesis that structural alterations caused by U3329C on domain 2 impact the structure of domain 1. Although these proposed structural alterations could not be detected by SHAPE, G3561A consistently altered the flexibility of two guanylates in the H4 lower stem, supporting a structural connection between domain 1 and domain 2 (Fig. 6A and C). Since this particular structural change was G3561A specific, it did not correlate with compensation in general by alterations in domain 2. The phenomenon of RNA:RNA interactions connecting domains protruding from a central backbone is consistent with observations of viral RNA genomes by cryo-electron microscopy, which show condensation of domains in the presence of magnesium, suggesting that tertiary interactions connect secondary-structure elements within and between RNA domains (49).

Since G3561A reduced wt TCV accumulation by 62% but had no detrimental effect in the absence of p38, the G3561A G→E amino acid alteration negatively impacts p38. Several pieces of evidence suggest that the G→E alteration reduces silencing suppressor activity of p38 and that this reduction is at least partially responsible for the compensatory property of the mutation. These include the facts that (i) p38_G3561A has 40% less silencing suppressor activity than wt p38 (Fig. 8C); (ii) the G→A alteration changes the C-terminal GW motif, known to be important for binding to AGOs and dsRNA; (iii) G3561A had no compensatory effect in the absence of p38; (iv) G3561A had no compensatory effect if cells lacked one target of the p38 silencing suppressor (ago1/ago2); and (v) G3561A had no compensatory effect if cells lacked RNA silencing (dcl2/dcl4).

A significant question is how reducing the silencing suppressor activity of p38 compensates for rev1-associated structural changes in the 3′ UTR of TCV. Any mechanism must also account for compensation not occurring in the absence of key RNA-silencing components. One possibility is that weakening silencing suppressor activity of p38 allows for enhanced production of one or more viral siRNAs. Such a viral siRNA could benefit the virus by correcting structural defects through complementary base pairing. This hypothesis would account for the lack of compensation in the absence of DCL2/DCL4 and AGO1/AGO2, which are responsible for the production of primary and, presumptively, secondary viral siRNAs (22, 46, 50). An alternative explanation for the lack of compensation in the ago1/ago2 background is that a lack of AGO1 affects homeostasis of DCL proteins, including reduced accumulation of DCL4, which should affect primary viral siRNAs (25). Several studies have shown that virus- or host-derived small RNAs can regulate virus accumulation (51–53). For example, efficient accumulation of HCV in Huh7 cells requires an endogenous microRNA (mi122) (54), and West Nile virus (WNV) encodes a miRNA-like viral siRNA (KUN-miR-1) that facilitates viral replication in mosquito cells (55). Similar to the case with WNV, a microRNA-like viral siRNA autoregulates dengue virus 2 replication in mosquito cells (56). Enhanced production of a viral siRNA would not, however, account for the lack of compensation in the absence of p38, which would be expected to produce more viral siRNAs. However, the absence of the silencing suppressor severely decreases virus levels in Col-0 protoplasts (Fig. 8A), which may negate any benefit from enhanced production of a beneficial viral siRNA.

The conservative F→L amino acid alteration in TCV-U3329C did not affect TCV accumulation (Fig. 2B) and did not reduce silencing suppressor activity of p38 (Fig. 8C). These results suggest that the compensatory mechanism for U3329C differs from that of G3561A. Despite this difference, the compensatory effect of U3329C on rev1 was also negated in the absence of p38, or in dcl2/dcl4 or ago1/ago2 protoplasts. These findings suggest that as with G3561A, the compensatory mechanism requires active RNA silencing. Since the negative effect of combining rev1 with U3329C is mitigated in cells with functional DCL2/DCL4 (i.e., with functional DCL proteins, the combination of the two mutations is compensatory and not strongly inhibitory), a DCL-produced viral siRNA may also play a role in the compensatory effect of U3329C. One possibility is if combined structural changes by rev1 and U3329C create enhanced opportunity for interaction with a beneficial viral siRNA. In the absence of the viral siRNA (in dcl2/dcl4 protoplasts) or when there is no silencing suppressor, these combined structural changes must be detrimental, since they result in a 2-fold or 4-fold decrease in virus levels, respectively.

In conclusion, this is the first example, to our knowledge, of a requirement for RNA-silencing components for second-site changes to be compensatory for mutations in an untranslated region of a virus. Future experiments will be necessary to determine if other clusters of second-site changes in domains 1 and 2 are also only compensatory in cells with active RNA silencing.