An RNA Pseudoknot Is Required for Production of Yellow Fever Virus Subgenomic RNA by the Host Nuclease XRN1^▿

Cell culture.

The origin and culture conditions of the BHK-21J, Vero E6, and SW13 cells used in this study have been described before (7, 52). C6/36 cells (23) were obtained from the ATCC and grown in Eagle's minimal essential medium (EMEM) supplemented with 8% fetal calf serum (Bodinco, Netherlands) and 5% nonessential amino acids.

Recombinant DNA techniques and plasmid construction.

Unless explained in more detail, standard nucleic acid methodologies were used (3, 48). Chemically competent Escherichia coli DH5α cells (24) were used for cloning. YFV-17D nucleotide numbering was according to Rice et al. (47) (GenBank accession no. X03700).

Plasmid pBlsrcptSK-YFV_9845-10861 (R. Molenkamp et al., unpublished data), which contains the complete 3′ UTR of YFV-17D, was used as a template for site-directed mutagenesis using the QuikChange strategy (Stratagene) to introduce mutations into the sequences of stem-loop E (SL-E) or in the nucleotides predicted to be involved in the formation of RNA pseudoknot 3 (PSK3) (41). After sequencing to verify the presence of the introduced mutations and to exclude unintended nucleotide changes, the mutant pBlsrcptSK-YFV_9845-10861 derivatives were digested with SfiI and XbaI, and the 492-bp DNA fragments containing the mutated SL-E or PSK3 sequences were cloned into pACNR-FLYF17Da (37).

A 194-bp HindIII-XbaI fragment encompassing the region between nucleotides (nt) 10,520 and 10,714 in the 3′ UTR of YFV was isolated from pHYF5′3′IVΔRS (7) and cloned into pBluescript SK(−). The resulting plasmid, pBlscrptSK-YFV_10520-10714, was used as a template to generate a minus-strand YFV RNA fragment for use as a probe in RNase protection assays.

Plasmid pSinrep5-YFV_10531-10611 was constructed by inserting a linker encompassing nt 10,531 to 10,611 of YFV flanked by an MluI and SphI adapter into MluI- and SphI-digested pSinrep5eGFP. The pSINrep5-YFV_10521-10662 recombinant was constructed by PCR. Plasmid pBlsrcptSK-YFV_9845-10861 was used as a template with oligonucleotides that contained either an MluI site (forward primer) or an SphI site (reverse primer). The resulting PCR product was cloned into MluI- and SphI-digested pSinrep5-eGFP (6).

Plasmid DNAs of the pACNR-FLYF17Da mutants and recombinant pSinrep5 were linearized with AflII and XhoI, respectively, and used for in vitro RNA transcription (52).

XRN1 RNA silencing.

Stocks of lentivirus particles, each expressing a short hairpin RNA (shRNA) (TRCN-049675, TRCN-049676, or TRCN-049677) against the human XRN1 gene (GenBank accession no. NM_019001.3), and a lentivirus expressing the scrambled shRNA SHC-002 were prepared from the MISSIONTRC-Hs1.0 library (Sigma) according to the manufacturer's recommendations. The particle titers of the lentivirus stocks were determined using a p24 enzyme-linked immunosorbent assay (ELISA) (Zeptometrix).

To analyze the effect of XRN1 silencing on YFV sfRNA production, 5 × 10⁵ SW13 cells were transduced with a combination of two different lentiviruses expressing one of the shRNAs specified above at a multiplicity of infection (MOI) of 5 (each) or with the SHC-002-expressing virus at an MOI of 10. At 72 h posttransduction, the cells were infected with YFV-17D at an MOI of 5, and 30 h later, the cells were lysed (16) and analyzed for the expression of the host XRN1 protein and actin by Western blotting (28) after PAGE on 5 and 10% gels, respectively. Antibodies directed against human XRN1 (Bethyl Laboratories) and actin (Santa Cruz) were used at dilutions of 1:5,000 and 1:2,000, respectively. Viral sfRNA production was analyzed by Northern blotting as described below.

In vitro XRN1 assay.

Plasmid pYF-R.luc2A-RP (25) was linearized with XhoI and used as a template for in vitro transcription of YFV replicon RNA in the absence of a cap analogue. To remove the 5′-triphosphate, 7 μg of transcript was incubated with tobacco acid pyrophosphatase (TAP; Epicentre) as specified by the manufacturer. The TAP-treated RNA was purified by phenol-chloroform extraction and ethanol precipitation. The pellet was dissolved in 16 μl H₂O and used for digestion with XRN1 (Terminator 5′-phosphate-dependent exonuclease; Epicentre) under the conditions described by the manufacturer. The units of XRN1 are indicated in the relevant figure.

RNA transfection and analysis of viral RNA synthesis.

BHK-21J cells were transfected with 5 μg of full-length YFV-17D or Sinrep5 transcript as described previously (52). In general, 2.5 ml (approximately 1.5 × 10⁶ cells) of the transfected BHK-21J cell suspension was seeded in a 35-mm plate. Total RNA was isolated from the transfected cells at 8 h postelectroporation (p.e.) for the recombinant Sinrep5-transfected cells and at 24 h p.e. for the YFV-transfected cells. Analysis of RNA synthesis by [³H]uridine labeling was performed as described previously (7). Trizol (Invitrogen) was used for cell lysis and subsequent RNA purification. [³H]uridine-labeled RNAs were denatured with glyoxal and analyzed in 0.8% agarose gels (48).

For Northern blotting, samples containing 10 μg of total RNA from electroporated cells or in vitro-transcribed XRN1-treated RNA mixed with 5 μg total RNA from BHK cells were denatured using formaldehyde, separated in a formaldehyde-containing 1.5% agarose gel, and blotted onto a Hybond-N⁺ membrane (GE Healthcare) (48). The blots were hybridized with ³²P-labeled oligonucleotides as described previously (34, 57), except for the analysis of sfRNA production in the XRN1 silencing experiments, in which case randomly primed, [α-³²P]dATP-labeled cDNA fragments directed against the YFV 3′ UTR (nt 10,555 to 10,862) and the human glyceraldehyde-3-phosphate dehydrogenase (GAPDH) gene (GenBank accession no. NM_00246; nt 321 to 724) were used as probes in 5× SSC (1× SSC is 0.15 M NaCl plus 0.015 M sodium citrate) and 50% formamide at 42°C (34).

Virus stocks, infections, and plaque assays.

Medium was harvested from transfected cells to obtain virus stocks when a complete cytopathic effect (CPE) was observed. YFV-17D infections and plaque assays were performed essentially as described before (52), except for the agarose in the overlay, which was replaced by 1.2% Avicel (33).

Primer extension assay.

Primer extension analysis was performed as described by Sambrook et al. (48), with minor modifications. Total RNA (5 to 7 μg) from YFV or recombinant Sinrep5 RNA-transfected cells or XRN1-treated pYF-R.luc2A-RP transcript was annealed to ³²P-labeled oligonucleotide 1632 or 1648. Oligonucleotide 1632 is complementary to YFV nt 10,690 to 10,708, and oligonucleotide 1648 is complementary to YFV nt 10,580 to 10,598. After hybridization and subsequent ethanol precipitations, primer extension reactions were performed using 200 U of RevertAid H Minus Moloney murine leukemia virus reverse transcriptase (Fermentas) as described by the manufacturer. After 1 h of incubation at 42°C, the samples were treated with RNase A (10 μg/μl; Qiagen) and purified by phenol-chloroform extraction and ethanol precipitation. The primer extension products were analyzed in a denaturing 5% polyacrylamide-8 M urea sequencing gel. A ³³P-labeled Cycle Reader (Fermentas) sequence reaction mix using oligonucleotide 1632- or 1648-primed pBlsrcptSK-YFV_9845-10861 as a template was run in the same gel as the primer extension products and served as a size and sequence marker.

RNase protection assay.

pBlscrptSK-YFV_10520-10714 was linearized using HindIII, purified by phenol-chloroform extraction, and used as a template for T7 RNA polymerase-mediated in vitro transcription to yield a 229-nt probe that contains 34 nt derived from the vector and 195 nt that are complementary to YFV nt 10,520 to 10,714. The RNA probe was purified from a 6% polyacrylamide gel and hybridized overnight at 42°C to 10 μg of total RNA from mock- or YFV-infected BHK cells, using the solutions and protocols supplied with an RPAIII RNase protection assay kit (Ambion). As a positive control, 10 ng of a positive-strand T7 RNA polymerase transcript containing the 3′ 1 kb of the YFV genome was mixed with 10 μg of total BHK cell RNA and treated similarly to the samples containing the YFV- or mock-infected total BHK cell RNA. After hybridization, the samples were treated with RNase A/T1, ethanol precipitated, and subsequently analyzed in a 6% denaturing sequencing gel. A sequence reaction mix that served as a size marker was run in parallel with the samples in the RNase protection assay.

RNA structure determination by selective 2′-hydroxyl acylation and primer extension (SHAPE) probing.

A DNA template containing a T7 RNA polymerase promoter fused to the 3′-terminal 342 nt of YFV was generated by PCR, using Pfu DNA polymerase (Fermentas) as described by the manufacturer. The template was digested with EcoRI and HindIII and cloned into pUC9. After linearization with XbaI, the plasmid was used for in vitro RNA transcription (T7 MEGAscript kit; Ambion) to obtain a YFV RNA fragment of 188 nt (YFV nt 10,520 to 10,708) for probing of the RNA structure in the region encompassing the predicted RNA pseudoknot (41). RNA was purified as described by the manufacturer (Ambion), and the yield was determined by spectrophotometry.

Probing was performed essentially as previously described (36). Briefly, RNA (20 pmol) in 6 μl H₂O was heated at 95°C for 3 min, cooled down on ice, and subsequently incubated with 3 μl folding buffer at 37°C. After 25 min of incubation, 1 μl of 65 mM N-methylisatoic anhydride (NMIA; Sigma) in anhydrous dimethyl sulfoxide (DMSO) was added to the RNA and allowed to react for 60 min at 35°C. The control reaction mixture contained 1 μl of anhydrous DMSO and no NMIA. To determine the NMIA-induced modifications in the 188-bp YFV transcript, 4 pmol (2 μl) of the NMIA-treated RNA was annealed to 10 pM ³²P-labeled oligonucleotide 1632 and used for primer extension without any further purification.

Small virus-specific RNAs can be detected in YFV-infected mammalian and insect cells.

Recent work (44) demonstrated that cells infected with arthropod-borne flaviviruses produce a small, virus-specific RNA derived from the 3′ UTR. To analyze whether the YFV sfRNA is also produced in infected cells other than BHK-21 cells (44), the mammalian cell lines Vero E6 and SW13 and the mosquito cell line C6/36 were infected with YFV-17D. At 30 h postinfection (p.i.) (mammalian cell lines) or 36 h p.i. (insect cell line), total RNAs were isolated and analyzed by Northern blotting for YFV sfRNA production, using ³²P-labeled oligonucleotide 1632, complementary to nt 10,690 to 10,708 in the YFV 3′ UTR, as a probe. Apart from the YFV genomic RNA, three smaller YFV-specific RNAs were detected in the infected mammalian cells (Fig. 1). Using in vitro RNA transcripts of various lengths as size markers (data not shown), these RNAs were estimated to be 630 nt (RNA A), 330 nt (RNA B), and 235 nt (RNA C) in length. Based on their estimated sizes and assuming that the RNAs were colinear with the 3′ end of the genome, the 5′ end of RNA A would be located within the carboxy-terminal coding region of NS5, whereas the 5′ ends of RNAs B and C would be in the 3′ UTR. RNA B was the most abundant of these RNAs, and the concentrations of RNA A and RNA C varied depending on the cell line studied. Hardly any RNA A was detected in the infected BHK cells, in which RNA C was relatively abundant, whereas the reverse was true for SW13 cells (Fig. 1, lanes 1 and 3). Surprisingly, RNA B was the only small YFV-related RNA detected in the mosquito cells (C6/36 cells) (Fig. 1). Of the three RNAs detected in the analyzed mammalian cell lines, RNA B appeared to be similar to the recently described YFV sfRNA (44). To keep in line with the nomenclature used in previously published work, RNA B was named sfRNA1. The slightly smaller RNA C was named sfRNA2, which does not imply that this sfRNA is similar to WNV sfRNA2, which can be detected only when the XRN1 stalling site for WNV sfRNA1 is deleted (44).

YFV sfRNA1 and sfRNA2 have identical 5′ ends.

Oligonucleotide 1632 (Fig. 2A), which was expected to bind to both sfRNA1 and sfRNA2, was used for primer extension analysis of total RNA isolated from YFV-infected or mock-infected BHK cells. As an additional control, a reaction was performed with a sample containing total RNA of uninfected cells mixed with 1 μg of full-length YFV genome transcript to detect products that resulted from strong stops of the reverse transcriptase on the YFV genome. As shown in Fig. 2B, lane 1, primer extension of oligonucleotide 1632 on total RNA of YFV-infected cells yielded a product that was not present in the RNA of uninfected cells (Fig. 2B, lane 2) or in the sample containing the full-length YFV transcript (Fig. 2B, lane 3). Two primer extension products were observed that differed in size by only one nucleotide. The 5′ ends of these primer extension products were mapped to the A residues at positions 10,532 and 10,533 by use of an oligonucleotide 1632-primed sequence reaction with pBluescript-YFV_9845-10861 as a size marker. Based on these results, the length of the YFV sfRNA is 329 or 330 nt, which is in agreement with the size estimate for sfRNA1 based on the Northern blots. The result of the primer extension analysis was verified using an RNase protection assay with a gel-purified, ³²P-labeled probe that is complementary to 34 nt of the vector and to YFV nt 10,520 to 10,714. This probe has extra YFV nucleotides at the 3′ end relative to the predicted 5′ end of the sfRNA to discriminate between YFV genome and sfRNA protected fragments. As shown in Fig. 2C, some large YFV genome-derived bands were detected in both total RNA of YFV-infected cells and the control sample (Fig. 2C, lanes 2 and 4, respectively). In addition, the probe protected a unique 181- or 182-nt fragment in the total RNA of YFV-infected BHK cells (Fig. 2C, lane 2). The size of the protected RNA fragment agreed with that predicted based on the results of the primer extension analysis and confirmed that the 5′ end of sfRNA1 is at position 10,532 or 10,533.

The most surprising result of these assays was that only these primer extension products and these RNase-resistant RNA fragments were detected, suggesting that sfRNA1 and sfRNA2 have a common 5′ end and that the observed difference in size between the two RNAs was due to a truncation at the 3′ end of sfRNA2. To verify this hypothesis, total RNAs were isolated from YFV-infected BHK-21, Vero E6, SW13, and C6/36 cells and analyzed by Northern blotting, using oligonucleotides 1648 and 1296 as probes. Oligonucleotide 1648 was complementary to nt 10,580 to 10,597 of the YFV genome (Fig. 3A), near the predicted 5′ end of the sfRNAs. Oligonucleotide 1296 should bind to the 3′-terminal 32 nt of the YFV genome (Fig. 3A). As shown in Fig. 3B, oligonucleotide 1648 hybridized to both sfRNA1 and sfRNA2, whereas oligonucleotide 1296 detected only sfRNA1 (Fig. 3C). These results demonstrated that the two YFV sfRNAs have a common 5′ end and that sfRNA2 is indeed truncated at the 3′ end.

YFV sfRNA is generated by the host enzyme XRN1.

As recently shown, the host exoribonuclease XRN1 hydrolyzes the WNV genome in a 5′-3′ direction until it is stalled by a currently unknown signal yielding the WNV sfRNA (44). To determine whether XRN1 was also involved in the production of the YFV sfRNAs, the effects of silencing of XRN1 expression on sfRNA production in YFV-infected cells and the in vitro production of sfRNA by incubation of YFV transcripts with purified XRN1 were studied. The human-derived SW13 cell line was used in the XRN1 silencing experiments because the nucleotide sequence of the human XRN1 gene is the only one known among the cell lines that were analyzed for YFV sfRNA expression (Fig. 1). SW13 cells were transduced with lentiviruses expressing shRNA direct against the XRN1 transcript or a control shRNA, infected with YFV, and analyzed for XRN1 expression and YFV sfRNA production by Western and Northern blotting. Expression of each combination of two of the three shRNAs, TRCN-049675, TRCN-049676, and TRCN-049677, in SW13 cells resulted in significantly less production of XRN1 (Fig. 4A, lanes 2, 3, and 4) than that in cells transduced with a similar lentiviral vector expressing a scrambled shRNA (Fig. 4A, lane 1). The silencing of XRN1 expression by the indicated shRNAs was likely specific, since no effect on the expression of actin was observed in these cell lysates (Fig. 4A, bottom panel). As shown by the Northern blot analysis, silencing of XRN1 decreased the production of YFV sfRNA by approximately 90% for all three combinations of shRNAs tested compared to that in cells transduced with a lentivirus expressing a scrambled shRNA (Fig. 4B). The signals for the host cell GAPDH gene were similar in all these lysates, indicating that the observed differences in sfRNA production were not due to significant experimental error. Visual inspection of the cells expressing the shRNAs directed against XRN1 indicated that silencing affected the homeostasis of the cells. This might explain why the hybridization signal for the YFV genome was also somewhat weaker than that in cells transfected with the scrambled shRNA (Fig. 4B).

To determine whether purified XRN1 was able to produce YFV sfRNA in vitro, linearized pYF-R.luc2A-RP (25) was used as a template for the production of an uncapped YFV transcript. After treatment with TAP to remove the XRN1-blocking 5′-triphosphate, this RNA was incubated with 0.1 or 1 unit of XRN1 and analyzed by Northern blot analysis for sfRNA production. Incubation of YF-R.luc2A-RP RNA with XRN1 resulted in the production of a small YFV-specific RNA that comigrated with the sfRNA1 produced in YFV-infected BHK-21J cells (Fig. 5A). Primer extension analysis was performed to provide additional evidence that this small RNA produced in vitro truly reflected the YFV sfRNA. As shown in Fig. 5B, primer extension using oligonucleotide 1632 resulted in similarly sized products for both XRN1-treated YF-R.luc2A-RP RNA transcript and total RNA from YFV-infected BHK cells. The fact that the smaller sfRNA2 was readily detectable in YFV-infected BHK-21J cells and not in the in vitro XRN1 assays is in line with our previous findings that sfRNA1 and sfRNA2 have a common 5′ end and suggests that additional RNA processing is required to yield sfRNA2.

From the results obtained with the RNA silencing experiments and the in vitro YFV sfRNA production by purified XRN1, it was concluded that similar to the case for WNV, the host RNase XRN1 is responsible for the production of YFV sfRNAs.

An RNA pseudoknot in the YFV 3′ UTR is required for production of sfRNAs.

Compared to the predicted RNA structure of the YFV 3′ UTR (41), the 5′ end of sfRNA1 and sfRNA2 is located just a few nucleotides upstream of SL-E (Fig. 2A), which is part of an RNA pseudoknot. To determine whether the predicted RNA structures that form this pseudoknot (PSK3) are required for YFV sfRNA production, site-directed mutagenesis was used to construct pACNR-FLYF-17DΔSL-E, which lacked nt 10,537 to 10,596 (which form SL-E) and was therefore unable to form PSK3. In vitro-transcribed RNA of AflII-linearized pACNR-FLYF-17DΔSL-E was electroporated into BHK-21J cells that were subsequently analyzed for viral RNA replication by [³H]uridine labeling and for sfRNA synthesis by Northern blotting. Despite the relatively minor effects on viral replication (Fig. 6A), no YFV sfRNA could be detected in the cells transfected with YFV-ΔSL-E RNA (Fig. 6B). To demonstrate that the lack of sfRNA production in the cells transfected with YFV-ΔSL-E RNA was due to disruption of the XRN1 stalling site, transcripts of this mutant were incubated in vitro with purified XRN1. As shown in Fig. 6C, XRN1 was not stalled when SL-E was deleted and therefore no sfRNA was produced. These in vitro and in vivo results together demonstrate that SL-E encompasses RNA structures and/or sequences that are essential for stalling XRN1 and therefore for YFV sfRNA production. The YFV-ΔSL-E mutant virus was delayed in inducing CPE in BHK cells. It also showed a slight delay in initial virus production but reached a similar maximum titer to that of the parental virus (Fig. 6D).

Two sets of additional mutants were constructed to dissect the role of SL-E in YFV sfRNA synthesis in more detail. The first set of mutants focused on the role of the top stem structure of SL-E (referred to as e2) (41). In mutant YFV-e2AA, the right arm of the stem (5′-GCAGU-3′; YFV nt 10,583 to 10,587) was replaced by the sequence of the left part of the stem (5′-ACUGC-3′) (Fig. 7A). Due to these mutations, the formation of SL-E e2 was expected to be disrupted. YFV-e2BB was the opposite of the YFV-e2AA mutant. In this mutant, the left part of the e2 stem (5′-ACUGC-3′; YFV nt 10,567 to 10,571) was replaced by the nucleotide sequence of the right arm of SL-E e2 (5′-GCAGU-3′) (Fig. 7A). In the YFV-e2BA mutant, the possibility to form stem e2 was restored by switching the parental YFV-17D original sequences of the left and right arms of the stem. BHK-21J cells were electroporated with in vitro-transcribed RNAs of the parental YFV-17D strain and the YFV-e2 stem mutants. Cells were analyzed for genome RNA synthesis by [³H]uridine labeling and for sfRNA production by Northern blotting with oligonucleotide 1632. As shown in Fig. 7B, all three YFV-e2 mutants were able to synthesize viral genomic RNA efficiently. Strikingly, neither YFV-e2AA nor YFV-e2BB was able to produce detectable amounts of sfRNA (Fig. 7C). Restoring the possibility to form stem e2, as in YFV-e2BA, also restored the production of YFV sfRNA. Although in comparison to YFV-17D YFV-e2BA produced less sfRNA, both viruses induced CPE in the cells about 48 h after electroporation, whereas YFV-e2AA and YFV-e2BB were clearly delayed in CPE induction compared to the parental virus. The kinetics of virus production in BHK-21 cells were similar to that shown for YFV-ΔSL-E (e.g., YFV-e2AA) or closer to that for the parental virus (data not shown). The fact that the YFV sfRNAs were produced in cells infected with YFV-e2BA indicated that the SL-E structure was more important than the primary sequence for producing the sfRNA.

SL-E is predicted to be part of a more complex RNA structure in which the sequence 5′-GCUGU-3′ (pk3′; YFV nt 10,575 to 10,579) in the top loop of e2 is predicted to base pair with the sequence 3′-CGACA-5′ (pk3; YFV nt 10,598 to 10,602) immediately downstream of SL-E, forming an RNA pseudoknot (PSK3) (Fig. 7A) (41). A second set of YFV mutants was constructed to determine whether the pk3′-pk3 interaction is important for YFV sfRNA production. In YFV-pk3′pk3′ and YFV-pk3pk3, the downstream pk3 sequence or the upstream pk3′ sequence was replaced by the complementary sequence. For both of these mutants, PSK3 formation was expected to be disrupted. In YFV-pk3pk3′, the possibility to form PSK3 was restored, albeit with the positions of the pk3′ and pk3 sequences reversed in comparison to those in YFV-17D. [³H]uridine labeling of BHK cells transfected with in vitro-transcribed full-length RNAs of these mutants demonstrated that they replicated quite efficiently (Fig. 7D). More importantly, no sfRNA could be detected in cells transfected with YFV-pk3′pk3′ or YFV-pk3pk3, whereas YFV sfRNA could be detected in cells electroporated with the YFV-pk3pk3′ mutant, in which the ability to form PSK3 was restored (Fig. 7E). However, the amount of sfRNA produced by YFV-pk3pk3′ was far less than that observed for the parental YFV-17D strain and required contrast enhancement of the phosphorimager data, which changed only the view of the data, not the data themselves. All three pk3-pk3′ mutant viruses showed a significant delay in the onset of CPE in BHK cells compared to YFV-17D. The kinetics of virus production in BHK-21 cells were similar to that for YFV-17D, except for YFV-pk3′pk3′, whose kinetics was like that of YFV-ΔSL-E (data not shown).

The combined results of the mutagenesis of the SL-E e2 stem and the pk3′-pk3 sequences strongly suggest that the predicted H-type RNA pseudoknot PSK3 is required to stall XRN1, resulting in the production of YFV sfRNA.

Insertion of the PSK3 sequence into a heterologous RNA results in production of an sfRNA-like RNA.

The fact that the base pairings in SL-E stem e2 and between pk3′ and pk3 are important determinants of YFV sfRNA synthesis did not exclude the possibility that this region is part of a more complex RNA structure, including additional elements downstream of PSK3. To address this possibility, a synthetic MluI-SphI DNA mimicking YFV nt 10,521 to 10,662 and YFV nt 10,531 to 10,611 was inserted into a Sinrep5 vector (6) that contained an enhanced green fluorescent protein (GFP) gene just upstream of the insertion site (Fig. 8A). This resulted in two constructs: pSinrep5-eGFP-YFV_{10,521-10,662}, encoding the SL-E to SL-C fragment of the YFV 3′ UTR; and pSinrep5-eGFP-YFV_{10,531-10,611}, encoding the YFV SL-E/PSK3 region. The GFP gene allowed for easy determination of the transfection efficiency and for discrimination between the Sindbis virus subgenomic RNA and a potentially produced sfRNA-like RNA. BHK-21J cells were transfected with in vitro-transcribed RNAs of the Sinrep5-eGFP and Sinrep5-eGFP-YFV mutants. RNAs were isolated at 8 h p.e. and analyzed for sfRNA-like RNA production by Northern blotting and hybridization, with two oligonucleotides as probes. Probe 1674 was complementary to nt 7,601 to 7,625 of Sindbis virus and was expected to hybridize to both the Sinrep5 genomic and subgenomic mRNAs, whereas probe 1648, which was complementary to YFV SL-E (nt 10,580 to 10,597), was expected to hybridize not only to Sinrep5 mRNAs but also to any sfRNA-like RNA produced in the transfected cells. As shown in Fig. 8B, hybridization of total RNA of the transfected cells with oligonucleotide 1674 demonstrated that Sinrep5-eGFP and both Sinrep5-eGFP-YFV_{10,521-10,622} and pSinrep5-eGFP-YFV_{10,531-10,611} replicated efficiently and produced a subgenomic mRNA of the expected size in the transfected BHK cells. In addition to the recombinant Sinrep5 genomic and subgenomic mRNAs, hybridization with oligonucleotide 1648 revealed the production of an additional small RNA that was unique for the cells transfected with Sinrep5-eGFP-YFV_{10,521-10,662} and pSinrep5-eGFP-YFV_{10,531-10,611} (Fig. 8B, lanes 6 and 7). The size of this RNA was in agreement with what was expected for the production of an sfRNA-like RNA in pSinrep5-eGFP upon insertion of the YFV sequences. Primer extension of total RNAs from cells transfected with Sinrep5-eGFP-YFV_{10,521-10,662} and pSinrep5-eGFP-YFV_{10.531-10.611} (Fig. 8C) showed that the 5′ end of the sfRNA-like RNA produced in these transfected cells was similar to that of the sfRNA produced in YFV-infected cells. Taken together, these results clearly demonstrate that YFV nt 10,531 to 10,611, containing the region that allows the formation of PSK3, are sufficient to stall XRN1. Furthermore, these results also indicate that RNA structures downstream of PSK3 are not required for the production of sfRNA.

Chemical RNA probing provides evidence of the formation of PSK3.

To obtain additional evidence to support the formation of PSK3, in vitro structure probing by SHAPE with NMIA (36) was performed on an in vitro-synthesized RNA template containing YFV 3′ UTR nt 10,520 to 10,708, encompassing PSK3. As shown in Fig. 9B, chemical probing of this region of the wild-type YFV-17D genome essentially confirmed the previously predicted RNA structure model (41). Apart from the nucleotides predicted to form the top loop of the side stem SL-E e3 (Fig. 9A), most of the other nucleotides did not react with NMIA, indicating that they were in a double-stranded conformation. This observation also includes the nucleotides of the pk3′ and pk3 sequences (Fig. 9B). The lack of reactivity to NMIA of these particular sequences was in agreement with the prediction that these complementary sequences could base pair to form the PSK3 pseudoknot.

The mutants that should have disrupted (YFV-pk3′pk3′ and YFV-pk3pk3) or restored (YFV-pk3pk3′) PSK3 formation were also subjected to RNA structure probing. The NMIA reactivity patterns of these mutant RNAs showed that the overall SL-E structure was maintained but that especially the pk3′ sequence, and to a somewhat lesser extent the pk3 sequence, were now susceptible to modification by NMIA (Fig. 9B), indicating that these sequences were no longer base pairing to form PSK3. Surprisingly, a similar result was also obtained for the putatively restored YFV-pk3pk3′ mutant RNA, indicating that despite the nucleotide complementarity between the pk3 and pk3′ sequences, the formation of PSK3 was not restored in this mutant.

Nonetheless, the structural probing data obtained with the RNA fragment encompassing YFV-17D nt 10,520 to 10,708 provided strong support for the previously proposed structural model (41).