The 3′-Terminal Hexamer Sequence of Classical swine fever virus RNA Plays a Role in Negatively Regulating the IRES-Mediated Translation

The 3′ untranslated region (UTR) is usually involved in the switch of the translation and replication for a positive-sense RNA virus. To understand the 3′ UTR involved in an internal ribosome entry site (IRES)-mediated translation in Classical swine fever virus (CSFV), we first confirmed the predicted secondary structure (designated as SLI, SLII, SLIII, and SLIV) by enzymatic probing. Using a reporter assay in which the luciferase expression is under the control of CSFV 5′ and 3′ UTRs, we found that the 3′ UTR harbors the positive and negative regulatory elements for translational control. Unlike other stem loops, SLI acts as a repressor for expression of the reporter gene. The negative cis-acting element in SLI is further mapped to the very 3′-end hexamer CGGCCC sequence. Further, the CSFV IRES-mediated translation can be enhanced by the heterologous 3′-ends such as the poly(A) or the 3′ UTR of Hepatitis C virus (HCV). Interestingly, such an enhancement was repressed by flanking this hexamer to the end of poly(A) or HCV 3′ UTR. After sequence comparison and alignment, we have found that this hexamer sequence could hypothetically base pair with the sequence in the IRES IIId1, the 40 S ribosomal subunit binding site for the translational initiation, located at the 5′ UTR. In conclusion, we have found that the 3′-end terminal sequence can play a role in regulating the translation of CSFV.


Introduction
Classical swine fever virus (CSFV), a member of the genus Pestivirus of the family Flaviviridae [1], is the causative agent of classical swine fever (CSF), previously known as hog cholera. CSF is highly contagious and often causes fatal hemorrhagic disease in pigs, resulting in highly significant economic losses worldwide [2,3]. The genome of CSFV is a single-stranded, positive-sense RNA of approximately 12.3 kb in length, neither capped at its 59-end nor polyadenylated at its 39-end. The genomic RNA comprises a single open reading frame (ORF), flanked by the 59 and the 39 untranslated regions (UTRs), which serves as mRNA for the synthesis of a single large polyprotein of 3,898 amino acids [4]. The polyprotein precursor is further processed into mature structural and nonstructural proteins by cleavage of viral and cellular proteases. From the N-to the C-terminus, the viral proteins are arranged in the following order: Npro, C, Erns, E1, E2, P7, NS2, NS3, NS4A, NS4B, NS5A and NS5B [5].
Viral protein synthesis of CSFV is initiated in a cap-independent manner by a highly conserved structure known as internal ribosome entry site (IRES), located in the 59 UTR. The structure and function of CSFV IRES is similar to that of the well-characterized Hepatitis C virus (HCV), comprising two main structural domains marked as II and III [6]. The IRES of both viruses can capture the 40 S ribosomal subunit directly without any known initiation factor [7].
The domain III of HCV IRES is the major anchoring site for the 40 S ribosomal subunit and recruits translation initiation factor eIF3 for efficient 80 S complex formation [8]. The domain II is essential for the conformational change of 40 S that allows the mRNA to go into the ribosomal binding channel [9], and promotes GTP hydrolysis to release eIF2 [10].
A number of cellular RNA-binding proteins have been identified as IRES-transacting factors (ITAFs), which play various roles in a number of biological situations [11,12]. These ITAFs such as polypyrimidine tract-binding proteins [13], poly(C)binding protein [14], NS1-associated protein1, and La autoantigen [15] could bind to the viral IRES and modulate the gene expression as non-canonical translation factors. For the IREScontaining viral RNA genome, the 39 UTR is usually in a highly ordered structure and harbors the signals for replication and translation in the viral life cycle [16]. During translation, the viral RNAs as well as cellular messenger RNAs can form a ''closedloop'' structure through the 59-39 RNA-RNA or RNA-protein interaction [17]. This concept of end-to-end communication has been reported to contribute to viral translation stimulation [17,18,19], replication [20,21,22], and the switch between protein synthesis and RNA replication [23]. However, translational control by which the 39 UTR or its binding proteins regulate IRES-mediated translation has not been discussed extensively.
Here we investigate the 39 UTR of CSFV that participates in its translation regulation. First, we determined the secondary structure of the 39 UTR of CSFV by enzymatic probing. Second, we constructed the reporter system by transfecting the RNA into cells with the UTRs of CSFV flanking the luciferase gene. We found that the 39 UTR of CSFV could positively and negatively regulate the IRES-dependent translation. Furthermore, our results also suggest that the hexamer sequence of the 39-terminus CGGCCC can shut down the translation, possibly by a long distance RNA-RNA interaction with the apical loop of domain IIId1 in IRES.

Plasmid constructs
The cDNA clones of CSFV containing the 39 UTRs of ALD or LPC strain [3] were used as the templates for polymerase chain reaction (PCR) to generate the fragments containing a T7 RNA promoter for structural probing. The PCR primers used in the reaction are: CSFV/59T7+12,058 (59TAATACGACTCAC-TATAGGGTATGAGCGCGGGTAACCCGGGATCTGGA39), ALD3912,328 (59GGGCCGTTAGGAAATTACCTTAGTC39) and LPC3912,269 (59GGGCCGTTAGAAATTACCTTA39). The T7 promoter sequence is underlined. The PCR products were cloned into pUC18 and verified by sequencing.
Constructs for translation studies were all derived from plasmid pGEM3Z/ALD containing the full-length cDNA of CSFV. At first, pGEM3Z/ALD was digested by Kpn I and then self-ligated to remove the 7-kb fragment of CSFV coding region. The resulting clone was then digested with ApaI, and followed by a self-ligation to obtain the clone pALD/DAK/A that contains UTRs. For the in vivo translation assay, pALD/L/A were derived from pALD/ DAK/A. At first, p2Luc containing the firefly luciferase was used as the template for PCR to generate the fragments containing Apa I and Spe I with the 59 primer Apa L (59GAAGGGCCC-CGAAGACGCCAAAAACATAAAGAA39) and the 39primer Lspe (59GAAACTAGTTTACAATTTGGACTTTCCGCCCTT39). The restriction enzyme site is in Italic. The relative DNA fragment in pALD/DAK/A was replaced with the PCR-amplified fragments containing luciferase gene by digestion with Apa I and Spe I. The resulting plasmid pALD/L/A containing a T7 promoter, the ALD 59 UTR (nucleotides 1 to 374, numbering is from CSFV genomic RNA), partial Npro protein coding sequence, firefly luciferase gene (L), and ALD 39 UTR. The replacement of the 39 UTR of CSFV with poly (A) tail (30A) or the 39 UTR of HCV was done by PCR with Spe I containing forward and Hind III containing reverse primers ( Table 1). The PCR products were cloned into T-easy vector (Promega, Madison, WI, USA). The sequences were verified and sub-cloned into pALD/L/A with Spe I and Hind III sites. The resulting plasmids were designated as pALD/L/(A) n and pALD/L/H, respectively. The rest of constructs, pALD/L/SLI, -/SLII, -/SLIII, -/AupSLI, and -/AlowSLI, with the mutations in the 39 UTR of pALD/L/A were constructed using the same approach. To generate the plasmids pALD/L/AupSLI and pALD/L/AlowSLI, the 39 reverse primers were the megaprimers obtained by first PCR with the primers (primer 1 and primer 2) indicated in the Table 1.

In vitro transcription
The templates used for in vitro transcription were mostly derived from PCR fragments except ALD/L/(A) n which was cleaved with Hind III for run-off transcription. The gel-purified PCR products for transcription were generated using the common 59 primer (59GACGTCTAAGAAACCATTATTATC39) and the specific 39 primer (Table 2).
In vitro transcription was carried out at 37uC for 2 h in a 50 ml reaction containing 150 U of T7 RNA polymerase, 40 mM Tris-HCl pH 8.0, 8 mM MgCl 2 , 2 mM spermidine-(HCl) 3 , 10 mM dithiothreitol, 1 mM ATP/UTP/CTP/GTP. DNA templates was removed by digestion with RNase-free DNase I, and followed by a phenol/chloroform extraction, ethanol precipitation, and salt precipitation. RNAs were resolved through a 1% agarose gel to analyze the RNA quality and quantify the yield by a densitometry.

RNA labeling
To label the 59-ends of the LPC39UTR, LPC39SLII/I, and LPC39SLI transcripts, 1.5 mg of gel-purified transcripts were dephosphorylated first and then treated with T4 polynucleotides kinase [24]. Labeled transcripts were purified by electrophoresis on a 10% sequencing gel and eluted.

Structure prediction and mapping of the CSFV 39UTR
As predicted by mfold [25], 39 UTR of ALD and LPC strains form four stem-loops. Labeled LPC39UTR, LPCSLII/I, and LPCSLI were cleaved with specific RNases. Labeled RNAs were partially digested with the alkaline buffer (55.5 mM Na/carbonate pH 9.0 and 1.1 mM EDTA) and used as markers. Besides, the labeled RNAs were digested with RNase T1 (10 to 15 units) at 55uC or RNase A (3 ng) on ice for 10 minutes, and subsequently denatured in boiling water for 90 and 30 sec, respectively, to serve as the sequencing ladder markers.

Cell culture, transfection and analysis of luciferase activity
Porcine kidney-15 (PK-15; ATCC CCL-33) cells were maintained in Dulbecco's modified Eagle medium (DMEM) supplemented with 10% fetal bovine serum (FBS), 1% penicillinstreptomycin at 37uC, and supplying with 5% CO 2 . About 4610 5 cells were seeded into 12-well plates and grown to 90% confluence. The 4 ml of Lipofectamine (Invitrogen, Carlsbad, CA, USA) and 0.8 mg of the transcribed RNA were each diluted with 100 ml of Opti-MEM and incubated at room temperature for 5 min. Both of the diluted components were mixed and incubated at room temperature for 25 min, and then added to the culture.
For the harvest, cells were washed with phosphate-buffered saline (PBS), and about 150 ml of passive lysis buffer (Promega) was added to each well, and the cells were scraped, collected, and with a freeze-thaw process. After a centrifugation at 13,0006g for 5 min, 20 ml of lysate was mixed with 100 ml of Luciferase Assay reagent, and the firefly luciferase activity was measured by a luminometer (FLUOstar OPTIMA, BMG Labtech, Ortenberg, Germany).

Structural prediction and enzymatic probing of the 39 UTR of CSFV
The 39 UTR sequences of CSFV strains ALD and LPC [3] were aligned using the CLUSTAL W program [27]. The major difference between these two strains is the presence of a thirteennucleotide sequence (CUUUUUUCUUUUU) in the 39 UTR of LPC (Fig. 1A), which is similar to other lapinized vaccine strains such as C-and HCLV strain [3,28,29,30]. The 39 UTRs of these two strains, predicted with the mfold program [25], appeared to fold similarly into four consecutive independent stem loops, designated as SLI, SLII, SLIII, and SLIV from the 39 end (Fig. 1B), in agreement with previous predictions [31]. The extra thirteen-nucleotide region of LPC was found to lie in the loop structure of SLIII.
To confirm the predicted structure of the 39 UTR in solution, four RNases were used to probe the structure of the 39 UTR of CSFV. A summary of the enzymatic structural probing results is shown in figure 1B. A representative result of the probing experiments is shown in figure 1C. Due to the difficulties encountered using the 59 end-labeled LPC39UTR transcripts to clearly and completely map the structures of SLII and SLI, shorter transcripts LPCSLII/I (Fig. 1C) and LPCSLI (Fig. S1) were generated.  Using LPC39UTR as the substrate, appearance of RNase Aspecific bands corresponding to the C104 (bulge), C94 (apical loop), C84, and U81 (internal loop) positions strongly substantiate their predicted locations within single-stranded regions. The production of a strong signal at G64 by RNase T1 also suggests that G64 is located in a single-stranded region, in agreement with the prediction. Efficient cleavage of nucleotides U109 to A105, A88 to A85, and A78 to U71 by RNase V1 digestion, as indicated by the prominent cleavage product bands, strongly supports the premise that they are located in the predicted stem regions. RNase T2 digestion resulted in the production of the products at A98 to A93 and U70 to A57, corresponding to the predicted apical loop region, and the singlestranded junction between SLII and SLI, respectively. Overall, the results of RNases T2 and V1 digestion, specific for single-strand RNA and double-strand RNA, respectively, did not overlap with each other (Fig. 1C). Similar cleavage patterns with the single-strand specific RNases A, T1, and T2, and with the double-strand specific RNase V1 on SLI, SLII, SLIII, and SLIV were matched to that of the predicted structure (Fig. S1, S2, and S3). Overall, these enzymatic structural probing results correlate considerably well with the existence of four independent stem-loops in the 39 UTR of CSFV as predicted by computer analysis. The 39 UTR of CSFV can stimulate its IRES-dependent translation in cells Previously, we tested the effect of the 39 UTR derived from ALD and LPC strains, and found that the structural difference of the 39 UTR could be involved in translation efficiency in an in vitro translation system (unpublished data). Therefore, we systematically evaluated the roles of the 39 UTR participating in the regulation of translation in a cell-based system. The first, the reporter plasmid pALD/L/A, was generated by replacing nearly the entire coding region of CSFV with firefly luciferase gene ( Fig. 2A). This plasmid turned out to contain a T7 promoter, 59 UTR of ALD, N-terminal region of N pro fused to the firefly luciferase gene, and the 39 UTR of ALD ( Fig. 2A). We then optimized the conditions for the transfection by evaluating the translation efficiency. However, we have observed that the untranslated regions could cause the cap-dependent translation shutdown (unpublished data). Therefore, the capped reporter control used for normalization [32,33] showed varied expression when co-transfected with the RNAs containing different structure of the CSFV 39 UTR. To avoid the varied expression caused by cap-dependent translation shutdown, most of the transfection experiments with the cap-independent CSFV RNAs were repeated at least three times. The translation efficiency in PK-15 cells is in a 39 UTR-and dose-dependent manner (Fig. 2B). The translation efficiency, measured as the activity of firefly luciferase, reached maximum at six hours after transfection (Fig. 2C). Results indicate that the translation of the RNA containing the 39 UTR of ALD was about 10 to 54-fold higher than that of the RNA without 39 UTR (ALD/L/D) (Fig. 2B), indicating that the 39 UTR of CSFV can strongly stimulate its IRES-dependent translation in cells. The poly(A) tail and the 39 UTR of HCV enhance the CSFV IRES-mediated translation Although the IRESes in the 59 UTR of the related viruses from the Flaviviridae are structurally conserved, the 39 UTRs of these viruses are distinctive from each other. The 39 UTR of HCV was shown to facilitate the assembly of the translation initiation complex and enhance the IRES-mediated translation activity [34,35,36]. Since the IRES of HCV is structurally similar to that of CSFV, we tested whether the 39 UTR of HCV can also regulate the IRES of CSFV in translation. A chimera RNA containing the 59 UTR of CSFV but with the 39 UTR of HCV was generated ( Fig. 2A). Poly(A) sequence has been reported to have a general stimulatory effect not only on cap-dependent translation but also on cap-independent translation [34,36,37]. We then generated the RNA with the poly(A) tail to replace the 39 UTR of CSFV, to test the translation (Fig. 2A). Results indicate that the firefly luciferase activity of the reporter system with poly(A) tail and the 39 UTR of HCV increased about 2.5 and 3.8-fold of that with the 39 UTR of CSFV, respectively (Fig. 2D).

SLI in the 39 UTR of CSFV has a negative effect on translation
We wondered why the 39 UTR of CSFV was 2 to 4-fold less efficient in assisting its own IRES-mediated translation than those of heterologous 39 UTR or poly(A) tail. Since the 39 UTR of CSFV is composed of four stem-loops ( Fig. 1 and 3A), we tested whether these stem-loops (SLs) may have a negative role in regulating translation. We generated a series of deletion mutants to determine the contribution of these four SLs in the 39 UTR of CSFV in translation (Fig. 3B). The translation activity increased 2.7-or 4.1-fold when SLI alone (ALD/L/ADSLI) or SLI and SLII (ALD/L/ADSLISLII) is deleted, respectively. The resulting activity of these mutants similar to that of the constructs with the poly(A) tail or the 39 UTR of HCV (Fig. 2D). These results imply that each of the stem-loops in the 39 UTR of CSFV may have different roles in regulating the translation. A series of reporter constructs each with one of the individual stem-loops of 39 UTR was used to evaluate their contribution to translation activity. Results indicate that both the SLII and SLIII had a positive effect, but SLI and SLIV had a negative effect on the translation compared to that of the entire 39 UTR (Fig. 3B). Considering that luciferase activity increased in the absence of SL1,i.e. ALD/L/ADSLI (2.7-fold increase), and ALD/L/ADSLI-SLII (4.1-fold increase), but in the presence of SL1 (ALD/L/SLI), it reduced to less than 50% of that with entire 39 UTR, suggests that SLI has a negative regulation in CSFV IRES-mediated translation. Although SLI or SLIV in the 39 UTR of CSFV plays a negative role in translation, they somehow contribute to the translation about 5 to 6 folds compared to that of ALD/L/D (Fig. 3B).
To pin down the region of SLI involved in the translational regulation, we constructed mutants (ALD/L/AupSLI and ALD/ L/AlowSLI) containing either the upper or the lower region of SLI (Fig. 3A). The deletions on SLI showed no effect on the overall structure of the 39 UTR predicted with the mfold program. Results showed that the 39 UTR containing the lower portion of SLI (ALD/L/AlowSLI) had an effect similar to that of wild-type containing the entire SLI (Fig. 3C). The other mutant, with the 39 UTR containing the upper portion of SLI, (ALD/L/AupSLI) had an activity similar to that of the SLI deletion (ALD/L/ADSLI) (Fig. 3C).

The terminal hexamer CGGCCC is the negative regulater in IRES-mediated translation of CSFV
Since the lower portion of SLI contributes to the translation repression, we then generated the mutants containing the left strand sequence (ALD/L/AlowSLI-L) or the right strand sequence (ALD/L/AlowSLI-R) of the lowSLI mutant (Fig. 3A). The RNA transcript containing the left strand of lowSLI (22 nts) resulted in similar translation activity to that of the SLI deletion (Fig. 3D). Although the RNA transcript containing the right strand sequence of lowSLI (17 nts) showed different results, the activity was only 50% that of wild type (Fig. 3D). These results indicate that the primary sequence rather than secondary structure at the very 39end of CSFV genome can negatively regulate the translation (Fig. 3D).
We further dissected this sequence into two parts UCCUAA and CGGCCC to examine their potential effects on translation repression. In the absence of SLI, the RNA harboring CGGCCC hexamer (ALD/L/ADSLI+CGGCCC) but not UCCUAA (ALD/ L/ADSLI+UCCUAA) responded to the repression (Fig. 4A). We have also tested whether the hexamer CGGCCC should be located in the very 39-end of the molecture to response in the repression. Results indicate that the hexamer CGGCCC reponding in translation repression is required to be located at the very 39-end (Fig. 4B). These results also suggested that the single-strand form of UCCUAACGGCCC is more efficient to block the translation than that in the stem-loop form (as in SLI) statistically significant. We have wondered why the very 39-end sequence influenced the IRES-mediated translation in cells. After a sequence alignment and comparison with the help of the mfold program, we noticed that the very 39-end sequence 59UC-CUAACGGCCC39 could have the potential base-pairing to the IIId1 subdomain in the IRES (Fig. 5A). Therefore, we hypothesised that the interaction of the 39-end sequence to the subdomain IIId1 of IRES in the 59 UTR downregulated the IRES-mediated translation. The apical loop of sub-domain IIId1 contains the phylogenetically conserved GGG triplet, which is the primary determinant of the 40 S ribosomal subunit binding site in the IRES [38]. Therefore, the 59-39 interaction could possibly block the accessibility of the 40 S ribosomal subunit and repress the translation.
We then investigated whether addition of the hexamer to the terminus of the poly(A) tail or the 39 UTR of HCV could block the CSFV IRES-mediated translation. Our results indicated that the activity of ALD/L/(A)n+CGGCCC and ALD/L/H+CGGCCC was reduced to about 21% and 16% of that of ALD/L/(A)n and ALD/L/H, respectively ( Fig. 4C and 4D). When the 39-ends of ALD/L/(A)n and ALD/L/H were added to the longer sequence UCCUAACGGCCC for base-pairing, the reduction in translation activity was enhanced, down to 1% and 9%, respectively ( Fig. 4C  and 4D). These results suggest that the 39-terminal sequence CGGCCC represses the IRES-mediated translation through the 59-39 interaction by base-pairing.

Discussion
In this study, we have demonstrated that the 39 UTR of CSFV plays a regulatory role in IRES-mediated translation. However, the regulation in CSFV is somewhat different from those of HCV [35,36,37,39] and Dengue virus [40], the other members of the same family. The secondary structure in the 39 UTR of CSFV can have both up-(SLII and SLIII) and down-regulaing effects (SLI) in translation (Fig. 3). The switch in helping or repressing the translaion may rely on the availability of the 39-end hexamer sequence CGGCCC. One of the possibilities to shut down the translation is to go through the interaction of the CGGCCC with the IIId1 domain in IRES which has been reported to be the 40 S ribosomal subunit binding site [38]. Mutants ALD/L/ (A)n+CGGCCC and ALD/L/H+CGGCCC contain the only sequence (the hexamer) derived from CSFV in their 39 UTR which could repress the translation by about 80% of that without the terminal hexamer (Fig. 4). We have also observed that the RNA transcripts containing the extra non-viral nucleotides at the very 39-end (abutting to the CGGCCC) derived from restriction enzyme-linearized plasmids could vary the repression effects when transfected into the cells. The variation effect was overcome when the RNA transcripts with precise ends were all derived from the PCR fragments containing the T7 promoter. The results suggest that the terminal non-viral sequence may interfere with the availability of the CGGCCC sequence. These observations were confirmed by switching the terminal hexamer CGGCCC with its upstream sequence UCCUAA (from 59-UCCUAACGGCCC39 to 59-CGGCCCUCCUAA39) and losing the repression activity (Fig. 4).
The long-range RNA-RNA interaction of the 5BSL3.2 domain near the 39-end and the IRES IIId domain located at the 59 UTR of HCV (Fig. 5C) is involved in modulating the switch from translation to replication [21,41]. The apical loop of IIId1 in CSFV and BVDV and IIId in HCV containing the phylogenetically conserved GGG triplet (Fig. 5), the primary determinant of the 40 S ribosomal subunits binding site [38], could be blocked by antisense oligonucleotides or RNA aptamer and inhibit the IRESmediated translation [42,43]. These results suggested that the apical loop of IRES IIId1 in CSFV and BVDV or IIId in HCV appears to be a control element in regulating the viral gene expression.
Due to the limited size of the viral genome, the viruses have evolved various strategies to modulate their gene expression. It is not surprising that the viral genome harbors the cis-regulatory element at the 39-end of the genome to control the translation apart from the 59 UTR. When the positive-sense RNA virus infects the target cells, the 39 UTR can strongly enhance the translation via the 59-39 cross-talk, as with those of cellular mRNAs. The cellular proteins such as polypyrimidine tractbinding protein and insulin-like growth factor 2 mRNA-binding protein 1 bind to the 59-and the 39-end of HCV RNA that cause the viral RNA cyclization, and therefore facilitate translation [18,19]. Further, common cellular proteins such as NF45 and RNA helicase A, identified by UV crosslinking assays or RNApulldown assays, were found to associate with the 39 UTRs of CSFV, BVDV, and HCV [18,20,44,45]. These 39 UTR-binding proteins would be expected to be involved in translation/ replication regulation. Since the template for the translation and replication of a positive-sense RNA virus is the same, the switch between these two processes should be in a tight regulation [46]. The availability of the 39-terminus CGGCCC for the interaction with IRES IIId1 of CSFV in this study could be facilitated by resolving the stem-loop structure in SLI using a cellular protein such as RNA helicase which was reported to interact with the 39 UTR of CSFV [45]. Finally, it is possible that through the longrange 59-39 interaction of CSFV either by base-pairing or with the help of viral [32,33] or host proteins [45], the translation could be well-regulated.