Virus-derived circular RNAs populate hepatitis C virus–infected cells

Significance

The cytoplasmic hepatitis C viral RNA genome is shown to be processed into hundreds of circular RNAs. Some of these circular RNAs contain the viral internal ribosome entry site and are translated to yield viral peptides which display proviral functions. Other nontranslated viral circular RNAs also enhance viral RNA abundances, suggesting that they can be targeted in antiviral approaches.

Abstract

It is known that pre-mRNAs in eukaryotic cells can be processed to circular RNAs by a backsplicing mechanism. Circular RNAs have great stability and can sequester proteins or small RNAs to exert functions on cellular pathways. Because viruses often exploit host pathways, we explored whether the RNA genome of the cytoplasmic hepatitis C virus is processed to yield virus-derived circRNAs (vcircRNAs). Computational analyses of RNA-seq experiments predicted that the viral RNA genome is fragmented to generate hundreds of vcircRNAs. More than a dozen of them were experimentally verified by rolling-circle amplification. VcircRNAs that contained the viral internal ribosome entry site were found to be translated into proteins that displayed proviral functions. Furthermore, two highly abundant, nontranslated vcircRNAs were shown to enhance viral RNA abundance. These findings argue that novel vcircRNA molecules modulate viral amplification in cells infected by a cytoplasmic RNA virus.

It is well established that pre-mRNAs in eukaryotic cells can be spliced into circular forms (circRNAs) by a nuclear backsplicing mechanism (1, 2). It has been estimated that approximately 25,000 circRNAs exist per cell at steady state (1, 3). These circRNAs are generated from an estimated 20% of transcribed genes (4), with hundreds of unique circRNAs being expressed in a tissue-specific manner. CircRNAs are more stable than their linearly spliced cousins due to their resistance to exoribonucleases (4). Very little is known about the biological functions for circRNAs. This is mainly because the abundance of most specific circRNAs ranges only between 1 and 10 copies per cell (4). One of the most studied cytoplasmic circRNAs is CDR1as (or ciRS7). CDR1as contains multiple binding sites for microRNA 7 and functions as a microRNA “sponge” to control midbrain development (5, 6). There is also accumulating evidence that circRNAs function in cell proliferation and immune regulation (7–9). Not surprisingly, certain DNA viruses, that use cellular pathways for transcription and RNA splicing, generate pre-mRNAs that are spliced into virus-derived circRNAs (vcircRNAs) (10). For example, Epstein–Barr virus encodes vcircRNAs that sequester microRNAs, resulting in the upregulation of genes that modulate viral malignancies (11, 12); Kaposi’s sarcoma virus generates vcircRNAs that modulate the lytic infectious cycle of the virus (13, 14); human papillomavirus also generates a vcircRNA, circE7, which is translated to encode the E7 oncoprotein (15). Here, we demonstrate that the cytoplasmic hepatitis C virus (HCV) generates vcircRNAs which display proviral functions.

Results

Prediction of vcircRNAs Generated from the HCV RNA Genome.

We conducted a study to identify any host circRNAs whose abundance was altered during HCV infection of hepatoma cells (16). Several of identified host circRNAs displayed proviral functions (16). Upon further interrogation of the RNase R-treated circRNA libraries used to identify these cellular circRNAs, we also identified, astonishingly, hundreds of putative vcircRNA species derived from the 10,000-nucleotide HCV genome (SI Appendix, Table S1 and Dataset S1). A putative vcircRNA was defined by the presence of a discontinuous reading with two breakpoints in the junction region, and additional two nucleotides located at the start and end of junction-spanning reads (Fig. 1A and SI Appendix, Fig. S1A). Although most of total viral reads were mapped to continuous regions on the viral genome, consistent with their templating from linear viral sequences, approximately 1% of the reads contained such novel junction sequences (SI Appendix, Table S1). Most of these junction reads (97.7%) were derived from the positive strand of the RNA genome (SI Appendix, Table S1 and Dataset S1). Although the 5′ and 3′ breakpoints in each junction were distributed across the entire HCV genome, their locations were not random (Fig. 1B). One cluster of reads (cluster I) was observed at the 5′ end of HCV and contained both an intact viral internal ribosome entry site (IRES) and the sequences that encode the N terminus of the viral core protein. The most abundant cluster (cluster II) contained 5′ breakpoints between nucleotide positions 830 to 890 in the viral core coding sequence and 3′ breakpoints between 1150 to 1154 in the envelope E1 coding sequence (Fig. 1B). In addition, another abundant cluster (cluster III) was found within the NS5B region (Fig. 1B). The most abundant vcircRNA junctions are shown in Fig. 1C. The sizes of predicted vcircRNAs range from 130 nts to larger than 2,000 nts, with an average size of 200 to 300 nts (Fig. 1D). Separately, analyses of the raw RNA-seq data using CIRI2 (https://sourceforge.net/projects/ciri/) also predicted vcircRNAs with similar junctions, although the number of junction reads were considerably fewer (Dataset S2).

Fig. 1.

Landscape of HCV-derived circRNA (vcircRNA) junctions predicted by RNA-seq analysis. (A) Identification of putative vcircRNA junctions derived from the linear HCV RNA genome. The curved arrow represents a junction-spanning read, which consists of two discontinuous sequences colored in blue and orange and joined by their 5′ and 3′ breakpoints. Aligning with the viral genome also generates the start and the end of read positions. To ensure that the junction originates from vcircRNAs, it is essential to validate the exact order of nucleotide positions, specifically the 5′ breakpoint, end of read, start of read, and finally the 3′ breakpoint. (B) Distribution of predicted vcircRNA junctions on the entire HCV genome. Each junction read is indicated by an arch that connects its 5′ and 3′ breakpoints. The three clusters examined in this study are indicated on the Top. Blue arches on the Bottom demonstrate the distribution of all the other predicted vcircRNA junctions. (C) List of top 20 predicted vcircRNA junctions. (D) Length distribution of the vcircRNAs predicted from junction-spanning reads.

Experimental Verification of vcircRNAs.

Although the presence of discontinuous junction sequences suggested their derivation from vcircRNAs, this needed to be experimentally verified. Reverse transcriptase-mediated rolling circle amplification was employed (17) to verify the putative IRES-containing cluster I vcircRNAs (~10/~405), as well as the most abundant cluster II (~870/~1151) and cluster III (a single member 7952/8189). Briefly, vcircRNAs were first reverse transcribed into cDNA products. These products, predicted to contain multiple iterated sequences of their template circular RNAs, were amplified by PCR using primers that flanked the novel junctions and, thus, diverged on linear viral counterpart sequences (Fig. 2A). If template RNAs are truly circular, multiple rounds of transcription will occur, resulting in PCR products of various lengths that contain multiple junction sites, denoted as a, b, c, and d (Fig. 2A). In contrast, linear RNAs should yield no products in the PCR reaction. Indeed, PCR fragments predictive of harboring multiple junction sequences were identified in infected, but not mock-infected cells (Fig. 2B and SI Appendix Fig. S1B). No specific PCR products containing junction sequences were amplified from full-length control viral RNAs that were in vitro transcribed by T7 RNA polymerase (SI Appendix, Fig. S1C), strengthening the argument that the vcircRNAs were generated in infected cells and did not represent PCR template switching or other artifacts. Sanger sequencing of the cloned PCR products revealed complete, iterated sequences of vcircRNAs in all the three clusters of vcircRNAs (Fig. 2C). Each iterative sequence was flanked by the predictive identical junction sites. So far, 14 predicted vcircRNAs have been verified by rolling circle RT-PCR and Sanger sequencing (Dataset S3).

Fig. 2.

Identification of HCV vcircRNAs in HCV-infected cultured hepatoma cells. (A) Schematic views of vcircRNAs subjected to reverse transcription and subsequent PCR using primers that flank the junction sites of predicted vcircRNAs but diverge on linear RNAs. RT-PCR products of single and multiple repeats of the vcircRNA sequences are denoted as b, c, and d while the smallest amplicon “a” only contains one junction sequence (Right). (B) Agarose gel electrophoresis of RT-PCR products. PCR products representing multiple copies of vcircRNAs from clusters I, II, and III are shown. Primers flanking the variable junctions of cluster I or cluster II vcircRNAs were used. For cluster III 7952/8189, junction-spanning primers were used (listed in Dataset S4). Nested PCR was performed to amplify cluster I vcircRNAs due to their low abundance and a representative image of the second round PCR for cluster I vcircRNAs is shown here in the Upper panel. (C) Representative Sanger sequencing results of RT-PCR products inserted into a TOPO vector. Each sequence trace displays two junction sites denoted as triangles with one full copy of the indicated vcircRNA sequence. (D) Northern blot detection of cluster II vcircRNA 873/1151 from HCV JFH-1 infected cells. Total RNA was extracted from uninfected or infected Huh7 cells at 3 d.p.i. Lane “pc873/1151” denotes RNA isolated from plasmid-based expressed circRNA873/1151 (see also SI Appendix, Fig. S2 A–C). Ribosomal RNA 28S and 18S are indicated by methylene blue staining (Top) of the membrane. A 5′ UTR probe (84 to 374 nt) was hybridized to detect HCV genomic RNA (Middle). An oligonucleotide probe specific for the junction site was hybridized to detect vcircRNA 873/1151 (Bottom). * and ** suggests the presence of RNAs with different topologies or of broken RNAs. (E) RT-PCR amplification of cluster II vcircRNAs with or without RNase R treatment. (F) Strand-specificity and RNase R resistance of cluster III vcircRNAs. After RNase R treatment (30 min at 37 °C), RNA was reverse transcribed with a specific reverse primer or forward primer spanning the junction (listed in Dataset S4) to differentiate between plus-stranded or negative-stranded vcirc7952/8189 followed by subsequent PCR. (G) Copy numbers of cluster II vcircRNAs quantified by droplet digital PCR. Divergent primers that flank the junctions of ~870/~1151 were used to amplify the vcircRNA (open circles), while convergent primers targeting the HCV Core region (squares) were used to amplify linear viral RNA. “Control IVT RNA” is in vitro transcribed full-length HCV RNA. Error bars indicate means ± SEM. Statistical significance was determined by ordinary one-way ANOVA. ns, not significant; ***P < 0.001.

To further validate the presence of viral circRNAs, Northern blot analysis using an oligonucleotide probe with sequence complementarity to the junction site of vcircRNA 873/1151 was performed. Extensive degradation of host ribosomal RNA (Fig. 2 D, Upper panel) and HCV genomic RNA (Fig. 2 D, Middle panel) was observed in RNA samples treated with RNase R, which selectively digests linear RNA. In contrast, vcircRNA ~873/~1151 was detected in nuclease-treated samples obtained from vcircRNA-overexpressing cells (pc873/1151; SI Appendix, Fig. S2 A–C) and infected cells (Fig. 2 D, Bottom panel). Of note, HCV genomic RNA seems less sensitive to RNase R digestion compared to rRNA, possibly due to its highly structured untranslated regions. Finally, rolling circle RT-PCR amplification also revealed the presence of cluster II (Fig. 2E) and positive-stranded cluster III (Fig. 2F) vcircRNAs in the RNase R-treated samples. Notably, the banding patterns (Fig. 2 E and F) in RNaseR− lanes were less prominent than in RNaseR+ lanes, suggesting that viral circRNAs were enriched after total RNA was digested by RNaseR. Thus, the junction-specific RT-PCR, Northern blot analysis, and RNase R results argue that bona fide vcircRNAs populate HCV-infected cells. Droplet digital PCR (ddPCR) showed that the abundance of vcircRNA~870/~1151 is approximately 1% of HCV genomic RNA in infected cells (Fig. 2G). In control experiments using in vitro transcribed full-length viral RNA, linear HCV RNAs can be easily detected by ddPCR, while only background levels of vcircRNAs could be observed (Fig. 2G). These findings strongly argue that this vcircRNA is a genuine product of viral infection.

Translation of Cluster I vcircRNAs.

In 1995, we showed that synthetic circRNAs that contained the IRES of encephalomyocarditis virus (EMCV) could be translated (18). Thus, we hypothesized that the HCV IRES-containing vcircRNAs represent a class of naturally derived vcircRNAs that can be translated into novel proteins. We determined the abundance of cluster I vcircRNAs to be approximately 0.25% of the abundance of linear viral RNA (SI Appendix, Fig. S2D) and verified the circularity of several members of the cluster I vcircRNAs (vcircRNA10/405, vcircRNA 26/369, vcircRNA30/372 and vcircRNA14/362) by rolling circle RT-PCR and Sanger sequencing (Dataset S3). To determine whether these HCV IRES-containing circRNAs could mediate translation, they were expressed from a split-GFP plasmid, in which the C-terminal coding sequence of GFP is placed before its N-terminal sequence, separated by the individual vcircRNA sequence that contains the HCV IRES (SI Appendix, Fig. S2E). This cassette was flanked by two inverted Drosophila lacasse2 intronic repeats, which mediate backsplicing events and facilitate the circularization of the linear transcript, thus allowing IRES-mediated production of full-length circular GFP in transfected cells (19). As a positive translation control, EMCV IRES-containing circGFP was expressed efficiently in transfected cells (Fig. 3A). In contrast, the 873/1151 HCV sequence, which does not contain an IRES, did not promote the synthesis of GFP (Fig. 3A). All the four tested cluster I vcircRNAs (10/405, 14/362, 26/369, 30/372) exhibited similar translation efficiency of circular GFP when compared to EMCV IRES-containing circGFP, demonstrating that the HCV IRES is active when placed into a circular RNA. Of note, the HCV 5′ UTR expressed half as much GFP from circGFP, indicating that a fully functional HCV IRES requires extra codons after AUG (20).

Fig. 3.

Translation of cluster I IRES-containing vcircRNAs. (A) Translation of the HCV IRES within a circular GFP RNA expressed from a polymerase II promoter-containing split-GFP plasmid that promotes backsplicing to create circular RNAs (see also SI Appendix, Fig. S2E). Plasmids containing the IRES from EMCV, or HCV 5′ UTR, or individual cluster I vcircRNA sequence were transfected into 293FT cells for 48 h. GFP fluorescence in each panel was then analyzed by flow cytometry. (B) Illustrations of two infectious clones of HCV that contain a HiBiT tag of 11 amino acids (21) inserted after the microRNA 122 (miR-122) binding sites in the 5′ noncoding region of HCV, so as not to interfere with miR-122 binding (22). The HCV-c10/405-Hibit mutant has an additional U to G mutation (the G is highlighted in red) to eliminate the upstream UGA codon and allow the expression of the HiBiT tag at the C terminus of vcirc10/405 ORF. In the HCV-c25/1277-Hibit mutant, HiBiT is inserted in-frame with the vcirc25/1277 ORF. (C) HiBiT expression from infectious viral genomes described in Fig. 3B. HiBiT expression from wild-type HCV or mutants were measured as HiBiT NanoLuc activities at indicated time points. (D) Effects of junction-specific siRNAs on HiBiT expression from IRES-containing vcircRNAs in virus-infected cells. Huh7 cells were transfected with scrambled control siRNA (siScramble), or a pool of siRNAs targeting the junction sites of vcirc10/405, vcirc26/369, vcirc30/372 and vcirc25/1277 at a final concentration of 100 nM. At 1 d posttransfection, cells were transfected with full-length viral RNA of HCV-c10/405-Hibit or HCV-c25/1277-Hibit. HiBiT expression was analyzed at 24 h after viral genome transfection. At least three replicas were used for each group. Error bars indicate means ± SEM. Statistical significance was determined by ordinary two-way ANOVA. **P < 0.01, ***P < 0.001, and ****P < 0.0001.

Although the IRES-containing vcircRNAs are of various lengths, all of them contain nucleotides 42 to 356 and are predicted to use the AUG at nucleotide 341 that is normally used for translation initiation of the HCV core protein. The proteins encoded by these vcircRNAs, depending on the location of their junction sequences, are predicted to contain a partial sequence of HCV core protein, followed by a novel sequence encoded from their 5′ breakpoints in the junction site until a stop codon is met. To determine whether the IRES-containing vcircRNAs were translated, we chose two different cluster I vcircRNAs, 10/405 and 25/1277, to monitor the production of 29 and 336 amino acid proteins, respectively (SI Appendix, Fig. S3). To enable their detection without appropriate antibodies specific to their predicted products, a sequence encoding a small peptide tag of 11 amino acids, called HiBiT (21), was inserted downstream of the 5′ junction site of vcircRNA 10/405 and 25/1277 in the noncoding region of the HCV genome (Fig. 3B). Care was taken not to disrupt the two microRNA122 binding sites which are essential for viral gene amplification (22). The HiBiT sequence has been engineered to be a required component of the luminescent NanoBiT enzyme, allowing HiBiT-containing proteins to be quantified as bioluminescence (21, 23). Since it is placed in the 5′ UTR of the HCV genome, HiBiT is exclusively expressed from IRES-containing vcircRNAs by using the natural AUG codon at HCV 341nt. Two variant genomes were constructed to test the translatability of the IRES-containing vcircRNAs (Fig. 3B): HCV-c10/405-Hibit, which is predicted to generate a core-HiBiT fusion peptide, and HCV-c25/1277-Hibit, which is predicted to generate a HiBiT fusion protein from a different reading frame (SI Appendix, Fig. S3). Both wild-type and variant RNA genomes were expressed in transfected cells, with two HiBiT mutants exhibiting attenuated phenotypes (SI Appendix, Fig. S4). Next, translation of the IRES-containing vcircRNAs was monitored. Unlike the HiBiT-lacking wild-type HCV, transfection of either the HCV-c10/405-Hibit or HCV-c25/1277-Hibit RNA genomes gave rise to significant luminescence (Fig. 3C). This result indicates that both vcirc10/405 and vcirc25/1277 can be translated when expressed from full-length viral RNA during viral infection. In further support that the observed luminescence resulted from translation of vcirc10/405-Hibit and vcirc25/1277-Hibit, short interfering RNAs (siRNAs) directed to the junction sequences of cluster I circRNAs were tested and shown to diminish HiBiT accumulation (Fig. 3D). To exclude the possibility that HiBiT was translated from linear RNA in an IRES-independent manner, linear control RNA fragments from the 5′ end of HCV-c10/405-Hibit were generated, and their translation was examined after transfection. Compared to the full-length HCV-c10/405-Hibit RNA, all HiBiT-containing linear control RNAs yielded only background luminescence (Fig. 4 A and B), strongly arguing that HiBiT expression is originated exclusively from vcircRNAs. These experiments support the conclusion that IRES-containing vcircRNAs can be translated into novel products in HCV-infected cells.

Fig. 4.

Translation of the HiBiT-containing linear control IVT RNAs. (A) Schematic view of designing short linear control RNAs of various lengths derived from HCV-c10/405-Hibit. Linear RNA fragments containing HiBiT were generated by in vitro transcription of indicated PCR products amplified from 5′ end of the mutant plasmid of HCV-c10/405-Hibit. (B) Huh7 cells in 12-well plates were transfected with 1 μg of each linear control RNA or in vitro transcribed full-length HCV-WT (lacking HiBiT) or HCV-c10/405-Hibit RNAs. At 12 and 24 h after transfection, HiBiT expression was quantified in the HiBiT Lytic Assay (Promega). At least four replicas were used for each group. Error bars indicate mean ± SEM. Statistical significance was determined by ordinary two-way ANOVA. **P < 0.01 and ***P < 0.001.

Functions of vcircRNAs during Viral Infection.

Do the identified vcircRNAs have pro- or antiviral functions in HCV-infected cells? To test this, selected vcircRNAs were depleted by siRNAs against their specific junctions in HCV infected cells. As shown in Fig. 5 A and B, depletion of either cluster I, or the most abundant circ873/1151 in cluster II, or the single member circ7952/8189 in cluster III resulted in a significant decrease in viral RNA abundance in infected cell lysates. More importantly, the cluster II and cluster III depletion led to reduced infectious virus production (Fig. 5C), further demonstrating that these two abundant viral circRNAs promote the HCV infectious cycle during infection. To further test potential off-target effects of siRNAs on linear HCV RNAs, which might reduce their abundance, a replication-deficient HCV JFH-1 genome carrying a nanoluciferase gene (HCV-Nluc-GND) was generated (SI Appendix, Fig. S5 A–C). While the positive control siRNA with sequence complementarity to HCV linear RNA decreased luciferase expression from the reporter viral genome, transfection of junction-specific siRNAs or their siRNA mutants did not alter the nano-luciferase expression produced from the replication-deficient HCV (Fig. 5D and SI Appendix, Fig. S5D), arguing that the vcircRNA junction siRNAs do not display off-target effects on linear viral RNA sequences.

Fig. 5.

Viral circRNAs and circRNA translational product have proviral effects on HCV infection. (A) The orange and blue arms show the corresponding 5′ and 3′ sequences forming the junction of the viral circRNA. JCT siRNAs (siJCT) target vc873/1151 (the most abundant member in cluster II) and vc7952/8189 (cluster III) at junction sequences. Mismatched sequences in the corresponding JCT mutant control siRNA (siJCT mutant) are shown in dashed lines. siHCV was designed to only target HCV linear RNA due to its complete complementarity to the linear genomic sequences. Due to low abundance of individual cluster I circRNAs, a pool of siRNAs targeting the junctions of vc10/405, vc26/369, vc30/372 and vc25/1277 were used. (B and C) Effects of cluster I, II, and III vcircRNA depletion on HCV infection. Huh7.5.1 cells were transfected with each indicated siRNA at a final concentration of 100 nM overnight and were further infected with HCV JFH-1 strain at a multiplicity of infection of 0.5. Three days later, HCV RNA abundance at HCV Core and NS5B regions in cell lysates (B) were quantified by qRT-PCR. Infectious viral titers in cell culture supernatant (C) were measured by focus forming unit (FFU) assay. At least three biological replicates were used for each group. (D) Effects of siRNAs on translation of a replication-defective HCV RNA genome containing a luciferase reporter (HCV-Nluc-GND). Huh7.5.1 cells were transfected with siRNAs overnight and then transfected with in vitro transcribed HCV-Nluc-GND RNA. HCV Nluc activity was further measured by NanoGlo assay 16 h later. (E) Effects of vcirc10/405 ORF overexpression on HCV reporter virus (HCV-Nluc-WT). Huh7.5.1 cells were transfected in vitro transcribed HCV-Nluc-WT RNA for 24 h and then transfected with each of control EGFP, control Stop codon-ORF, N-terminal, or C-terminal tagged ORF plasmids. After 24 h, HCV Nluc activities were measured in NanoGlo assays. At least four replicas were used for each group. Error bars indicate means ± SEM. Statistical significance was determined by ordinary two-way ANOVA. ns, not significant, *P < 0.05, **P < 0.01, ***P < 0.001, and ****P < 0.0001.

Finally, it was examined whether the protein product encoded by the IRES-containing vcircRNA10/405 enhanced viral gene expression. Overexpression of N- or C-terminal HiBiT-tagged vcircRNA10/405 open reading frames, but not a control reading frame interrupted by two stop codons (pSTOP-N-Hibit-10/405ORF), enhanced viral gene expression (Fig. 5E and SI Appendix, Fig. S6 A and B), arguing that the 29-amino acid peptide itself has a proviral function. Collectively, these results indicate that viral circRNAs themselves and peptides generated by IRES-containing vcircRNAs are beneficial to the HCV life cycle, most likely via distinct mechanisms.

Discussion

This study shows that vcircRNAs with proviral functions are generated from distinct parts of the HCV RNA genome, thus presenting a novel class of viral RNA species that populate infected cells. Some HCV-derived vcircRNAs contain the HCV IRES, and their unique open reading frames are translated. In addition to the translation potential, miR122-binding potential of some vcircRNAs that harbor miR-122 binding sites would expand the possible roles that they have in regulating the HCV life cycle. The 5′ end of the HCV genome can bind one, simultaneously two miR122 molecules or being devoid of miR122 (24). Cluster I vcircRNAs were found to be poorly expressed during infection, which may indicate that cluster I vcircRNAs are being made from HCV RNAs which are devoid of miR122. Abundant vcircRNAs could sequester proteins or small RNAs that modulate viral infection, or they could function in bulk after translocating to uninfected bystander cells (25). It is quite unlikely that the non-IRES-containing viral circRNAs all have distinct functions, considering that hundreds of them were predicted in our RNA-Seq study. Perhaps vcircRNAs inhibit initial innate immune responses, for example by inhibiting the activation of protein kinase R in uninfected bystander cells (26). A limitation of this study is that it was performed in cultured liver cells that were infected with the HCV JFH1 genotype 2 which replicates efficiently in Huh7 cells. While animal (27) and organoid models (doi.org/10.1101/2021.10.26.465357) for HCV are available, they are not very robust to study viral gene expression. However, putative vcircRNAs have been detected in the silkworm midguts infected with Bombyx mori cypovirus, a member of the Reoviridae (28). Thus, it is likely that HCV vcircRNAs are produced in animal tissues as well.

At present, it is not clear by what mechanism these vcircRNAs are being generated from HCV whose infectious cycle resides exclusively in the cytoplasm. There is no evidence that the cellular splicing machinery is relocated to the cytoplasm in infected cells, and HCV viral RNA cannot be detected in the nucleus (29). Furthermore, sequence alignment in the proximity of the junction sites did not yield any conserved motifs commonly seen in host pre-mRNA splicing, suggesting that these RNA virus-derived circular RNAs are probably generated by pathway(s) that are distinct from those utilized by cellular circRNAs involving spliceosomes (30). We speculate that a genome-wide knockout screen study could provide valuable insights into host factors responsible for the biogenesis of HCV circRNAs.

Another important question to ask is whether viral circRNA production is dependent on ongoing HCV RNA replication? According to Fig. 2G and SI Appendix, Fig. S1, runoff in vitro transcription of full-length RNA samples does not generate vcircRNA junctions, indicating the viral circRNAs are produced within virally infected cells. Furthermore, preliminary data (not shown) suggested that cluster III vcircRNAs were generated in both WT and replication-deficient GND mutant RNA transfected Huh7 cells by performing rolling circle amplification. Surprisingly, even in the presence of DAA inhibitor sofosbuvir, specific PCR banding patterns for cluster III circRNAs were still detected. Further experiments are in progress to study the mechanism.

There is evidence that other RNA viruses, such as SARS coronaviruses (31–33) and the Bombyx mori reovirus (28, 34) accumulate RNA species that contain noncontiguous viral sequences. In those studies, selected junction sequences were demonstrated by RT-PCR. However, it cannot be ruled out that some of these RNAs represent linear recombinant RNAs. Here, we validated complete sequences of vcircRNAs by rolling circle amplification and demonstrated their presence during infection by Northern blot analyses with resistance to RNaseR. Future investigations of novel vcircRNAs across a diverse range of RNA viruses will reveal the prevalence of virus-derived circRNAs and their potential significance in viral growth and pathogenesis. Such studies will ultimately contribute to the development of effective therapeutic strategies with which these novel virus-derived RNAs can be targeted.

Methods

Cell Culture, HCV Infection, and RNA Extraction.

The human hepatocarcinoma cell lines Huh7 and Huh7.5.1 were maintained in Dulbecco's modified Eagle's medium (DMEM, Invitrogen) supplemented with 10% heat-inactivated fetal bovine serum (FBS), pen/strep, and nonessential amino acids. 293FT cell line was purchased from ThermoFisher (Catalog#R70007) and maintained in DMEM supplemented with 10% heat-inactivated FBS, pen/strep, nonessential amino acids, and 500 µg/mL Geneticin^TM. Cells were grown in an incubator with 5% CO₂ at 37 °C. For infection, six-well plates were seeded with 2.5 × 10⁵ Huh7 cells per well. On the next day, supernatants were removed, and cells were infected with HCV JFH-1 viral stock at MOI of 0.1 for 3 h. The inoculum was removed and replaced with 2 mL of DMEM. At 48 or 72 h postinfection, total RNA from infected and noninfected cells was extracted using TRIzol reagent (Invitrogen), following the manufacturer’s instructions. Genomic DNA was removed in-column by Turbo DNase I treatment (ThermoFisher) using the RNA concentrate and clean kit (ZYMO lnc.).

Sample Preparation for RNA-Seq.

RNA samples with an RNA integrity ≥ 9 were first rRNA depleted and then treated with RNase R (Lucigen). After phenol/chloroform/isopropanol extraction, samples were used to construct RNA-Seq libraries, using the True-Seq Stranded Total RNA Library Kit (Illumina Inc.). These libraries were then sequenced on the Illumina sequencing platform (NovaSeq 6000 platform), and 150-bp/125-bp paired-end reads were generated. Illumina sequencing and QC of the raw data were performed at Stanford Genomics.

In Silico Identification of Viral Circular RNAs (vcircRNA).

The Galaxy tool (https://usegalaxy.org/) was used to perform Q30 and adaptor trimming on the raw data, and the reads were aligned with the HCV JFH-1 reference genome (GenBank No. AB047639). An expectation value cutoff of 10^-5 was used in the blastn (35). On the blast-hit table, reads that mapped to two discontinuous regions on the viral genome were further subjected to an analysis pipeline. This pipeline specifically sought circRNA junction-specific reads, which produced four nucleotide positions (P1 to P4) on the subject HCV reference genome in a precise order of 5′ breakpoint (P1) < end of read (P2) < start of read (P3) < the 3′ breakpoint (P4), or vice versa. To ensure that, the detailed step-by-step calculation filters are performed as follows: 1) [(query 1st split end) – (query 2nd split start)] greater or equals −1; 2) (P4 − P3)*(P2 − P1) greater than 0; 3) (P4 − P1)*(P3 − P2) greater than 0; and 4) [(P4 − P1) − (P3 − P2)] ranging from 1 to 151 are plus stranded junctions while [(P4 − P1) − (P3 − P2)] ranging between −151 and −1 are minus stranded.

Validation of HCV vcircRNAs using Rolling Circle Amplification and PCR.

Reverse transcription of 1 μg of RNA was performed using the Superscript III kit (Invitrogen), following the manufacturer’s instructions. A reverse primer specific to the predicted circRNA was added to the reaction to increase the yield of cDNA from circRNAs (Dataset S4). Subsequently, PCR (PrimeSTAR MAX DNA polymerase, Takara Bio USA, Inc.) was performed using divergent primers flanking or spanning the junction site of the specific circRNA. The oligonucleotides used in this study are listed in Dataset S4. PCR cycling parameters were as follows: 98 °C 30 s, 34 cycles of 98 °C 10 s, 59 °C 5 s, 72 °C 30 s, and 72 °C 3 min. PCR products were then separated in 1.2% agarose gel running with ladder (1 Kb Plus and 100 bp Plus DNA Ladders, Invitrogen). The DNA band at expected size of the circle was extracted by the Qiagen gel purification kit and then cloned into the pCR4 Blunt-TOPO vector (Invitrogen) for sequencing. Plasmids were sequenced and blasted to the JFH-1 sequence. A few vcircRNAs were then validated, containing one or iterated vcircRNA sequences with two identical repeated junction sites flanked at both ends (Dataset S3).

RNase R Treatment and Northern Blot Analysis.

RNase R treatment was performed by incubating 15 µg of total RNA with 2.5 μL RNaseR (10 U/μL, Abcam) at 37 °C for 30 min followed by cleanup using RNA concentrate and clean kit (ZYMO Inc.). For Northern blot analysis of HCV genomic RNA and vcircRNA, 10 μg of undigested total RNA samples and the above-mentioned RNase R-treated RNA were added with RNA loading dye (NEB) and denatured for 10 min at 70 °C. Subsequently, RNA samples were separated in 1% agarose gel containing 1X MOPS-EDTA-Sodium acetate (MESA, Sigma) and 6% formaldehyde, and then transferred onto Zeta-probe membranes (Bio-Rad) by upload capillary blotting overnight. Detection of HCV RNA (HCV nucleotides 84 to 374) and vcircRNA873/1151 (oligonucleotide probe 5′-CAGGACAACAGGGCCAGCAAGGTGGCGGACATCACAACCA-3′ against the junction) were performed using γ-³²P-dATP-end labeled (PNK kinase, NEB) probes. Autoradiographs of phosphor screen were quantitated using ImageQuant (GE Healthcare).

HCV vcircRNA and Viral mRNA Quantification by qPCR and Droplet Digital PCR (ddPCR).

One µg of total RNA was used to prepare cDNA using the High-Capacity RNA-to-cDNA Kit (Applied Biosystems). qPCR was performed using the PowerUP SYBR green PCR mix in a 96-well format in a CFX96 qPCR machine according to the manufacturer’s instructions (Bio-Rad). HCV RNA abundance at Core, Envelope 1, NS5B regions, and vcircRNA~870/~1151 was quantified by using primers listed in Dataset S4. Droplet Digital PCR was performed using the Bio-Rad’s QX200 Droplet Digital PCR system (Stanford Genomics). Briefly, 1 μg RNA was reverse transcribed using the Superscript III kit and a reverse primer for cluster II vcircRNA (Dataset S4). Subsequently, cDNA was mixed with 10 μL of QX200ddPCR EvaGreen supermix and 125 nM primers. The reaction was then dispensed into sample wells in the DG8 cartridge so that droplets were generated according to the manufacturer’s instructions. The droplets were then transferred to a 96-well plate and PCR was performed with the following conditions: 95 °C for 5 min, 42 cycles at 98 °C for 30 s and 59 °C for 1 min, 4 °C for 5 min and 90 °C for 5 min. Subsequently, the plate was transferred in to a QX200 droplet reader and analyzed by QuantaSoft Analysis Pro software (Bio-Rad). Both HCV RNA in the Core region and cluster II vcircRNA were quantitated separately for each sample, with at least three biological repeats.

Plasmid Overexpression of vcircRNA873/1151.

The plasmid placcase2-splitGFP was a generous gift from Dr. Jeremy E. Wilusz (Baylor College). To insert the sequence (5′ to 3′) of HCV circRNA873/1151 in between the two laccase2 inverted repeats, the plasmid was amplified with the following primer pair to exclude the entire split-GFP-EMCV-IRES region (SI Appendix, Fig. S2A) forward 5′-GTAAGTATTCAAAATTCCAAAAT-3′, reverse 5′-CTGCataaaataaaaaaaCTT-3′. The sequence (from 5′ breakpoint to 3′ breakpoint) of vcircRNA873/1151 was amplified from HCV JFH-1 infected cells by using the following primer pair: forward 5′- TTTTATTTTATGCAGTTGCTGGCCCTGTTGTCCT-3′, reverse 5′-ATTTTGAATACTTACGGTGGCGGACATCACAACC-3′. These two PCR products were subjected to InFusion Cloning (Takara Bio.) to generate the overexpression plasmid pc873/1151. For transfection, 1.7 × 10⁶ 293FT cells were seeded in a 6-cm dish and transfected with 3 µg pc873/1151. Total RNA was extracted at 2 d posttransfection for Northern blot analysis.

Construction of Placasse2-Split-GFP Plasmids Containing IRES Sequences from Various vcircRNAs.

To replace the EMCV-IRES in placcase2-splitGFP with the sequence (5′ to 3′) of HCV circRNA10/405, two fragments were amplified from placcase2-splitGFP with primer pair 1 and 2. Primer sequences were as follows: pair 1 forward: 5′-CTTGGTACCGAGCTCGGATCCACTAGTCCAGTGTGGTGGAAT-3′, pair 1 reverse: 5′ TTATTAATTAAGTCGACTGCAGAATTCAGATCC-3′; pair 2 forward: 5′-GT CCTGCAGG-GT ATGGTGAGCAAGGGCGAGGAGCT-3′, pair 2 reverse: 5′-AACGGGCCCTCTAGACTCGAGCGGCCGCCTGAGGTGCCAC-3′. A third fragment was amplified from HCV JFH-1 infected cells from 10 to 405 nt. The following primers were used to also include two restriction enzyme sites PacI and SbfI at its 5′ or 3′ end, respectively: forward 5′-gcagtcgac TTAATTAATAATAGGGGCGACACTCCGCC, reverse 5′-ctcaccatACCCTGCAGGACGTCTTCTGGGCGACGGTTG-3′. The plasmid placcase2-splitGFP was then digested with BamHI and XhoI and subjected to InFusion Cloning with three fragments mentioned above. The resulting plasmid placcase-splitGFP-10/405 was later digested with PacI and SbfI and then used in the InFusion Cloning to generate plasmids inserted with the individual sequences (5′ to 3′) of HCV circRNA30/372, circRNA14/362, circRNA26/369, and circ873/1151 (primers are listed in Dataset S4).

Flow Cytometry.

For transfection, 4 × 10⁵ 293FT cells per well were seeded in 12-well plates. On the following day, 1 μg of each placasse2-split-GFP plasmid was transfected using Lipofectamine 3000 according to the manufacturer’s protocol. At 2 d posttransfection, cells were collected by trypsinization, fixed with 3.2% paraformaldehyde, and then washed and suspended in FACS buffer (PBS containing 2% FBS). GFP fluorescence in each transfection was assessed with a Sony MA900 Multi-Application Cell Sorter (Stanford).

Generation of Mutant HCV JFH-1 DNA Infectious Clones Containing a HiBiT Tag Inserted in the 5′ Noncoding Region.

pHCV-JFH-1 genotype 2a DNA infectious clone was a generous gift from Dr. Charles Rice (Rockefeller University). To insert a HiBiT tag (VSGWRLFKKIS) and U32G mutation to eliminate a stop codon, an upstream fragment of 125 bp was amplified from the wild-type pJFH-1 plasmid by nested PCR with the forward primer 5′-AAAAGCAGGCTACTCGAATTCTAATACGACT-3′ and the first reverse primer 5′-AACAGCCGCCAGCCGCTCACGGGAGTGATTCCTGGCGGAGT-3′, and then the second reverse primer 5′-agTTAGCTAATCTTCTTGAACAGCCGCCAGCCGCT-3′. A downstream fragment of 2951 bp was amplified with the forward primer 5′- TTCAAGAAGATTAGCTAActgtgaggaactactgtcttc-3′ and the reverse primer 5′-CACGCGATGCCATCGCGGCC-3′. These two fragments were then included in the InFusion Cloning, together with the pJFH-1 plasmid digested with NotI and EcoRI. The resulting infectious clone was named pHCV-c10/405-Hibit. To generate pHCV-c25/1277-Hibit, an upstream fragment of 139 bp was synthesized (​AAA​AGC​AGG​CTA​CTC​GAA​TTC​TAA​TAC​GAC​TCA​CTA​TAG​ACC​TGC​CCC​TAA​TAG​GGG​CGA​CAC​TCC​GCC​ATG​AAT​CAC​TCC​CCG​TGA​GCG​GCT​GGC​GGC​TGT​TCAAGAAGATTAGCTGTGAGGAACTACTGTCTTC) by IDT. The downstream fragment was amplified from the wild-type pJFH-1 plasmid by using forward primer 5′-TGTGAGGAACTACTGTCTTCACGCAG-3′ and the reverse primer 5′-CACGCGATGCCATCGCGGCC-3′. Both fragments were fused to pJFH-1 plasmid in a similar way as mentioned above.

Generation of Nanoluciferase-Containing HCV DNA Infectious Clones with GDD (Wild-Type) or GND Mutation in the NS5B Gene.

A gBlock fragment (19aaCore-Nluc-P2AUbi, Dataset S4) was synthesized to construct pHCV-Nluc-WT. This fragment comprises a Nanoluciferase gene fused with the initial 19 amino acids of the HCV JFH-1 Core protein at its N terminus. Additionally, it incorporates a 19-residue porcine teschovirus-1 2A self-cleaving peptide and a Ubi sequence at its C terminus. This fragment was inserted immediately after HCV 5′ UTR, followed by the complete HCV polyprotein. GDD to GND mutation was further introduced in HCV-Nluc-GND by PCR primers (Dataset S4).

In Vitro Synthesis of RNA and Transfection.

Plasmids pHCV-WT (wild-type strain JFH-1), pHCV-c10/405-Hibit, pHCV-c25/1277-Hibit, pHCV-Nluc-WT, and pHCV-Nluc-GND were linearized with XbaI and transcribed using the T7 MEGAscript kit (Ambion), using the manufacturer’s instructions. PCR products of different lengths (1-226nt-Hibit, 1-444nt-Hibit, and 1-820nt-Hibit) were amplified from pHCV-c10/405-Hibit (primers listed in Dataset S4) and subsequently used as DNA templates for in vitro transcribed control RNAs. To detect HiBiT expression from virus mutants, Huh7 cells in 6-well plates were transfected with 3 μg of each in vitro transcribed RNA using the TransIT mRNA transfection kit (Mirus Bio) as previously described (36). After 6, 12, or 24 h of incubation, cells were harvested for the HiBiT lytic assay (Promega).

Small Interfering RNA Experiments.

The following siRNAs against junction sequence of certain IRES-containing circRNAs were synthesized: for depletion of vcircRNA10/405, sense siRNA-a, 5′-GAAGACGUUAAUAGGGGCGACdTdT-3′ and sense siRNA-b, 5′-UCGCCCAGAAGACGUUAAUAGdTdT-3′; for vcircRNA26/369, sense siRNA-a, 5′-AAACCUCAAAGAAACCGCCAUdTdT-3′ and sense siRA-b, 5′-ACCUCAAAGAAACCGCCAUGAdTdT-3′; for vcircRNA30/372, sense siRNA-a, 5′- AAAGAAAAACCAUGAAUCACUdTdT-3′, sense siRNA-b, 5′-GAAAAACCAUGAAUCACUCCC dTdT-3′ and sense siRNA-c, 5′-AGAAAAACCAUGAAUCACUCCdTdT-3′; for vcircRNA25/1277, sense siRNA 5′ CCCUGGCACCAUCCGCCAUGAdTdT-3′, antisense siRNA 5′-UCAUGGCGGAUGGUGCCAGGGdTdT-3′. The siRNA duplexes were formed by combining sense and their corresponding antisense strands in 1X siRNA Buffer (Dharmacon) as described previously (16). As a negative control siRNA, the following oligonucleotides were used: sense 5′-GAUCAUACGUGCGAUCAGAdTdT-3′ and antisense 5′-UCUGAUCGCACGUAUGAUCdTdT-3′. To knock down HiBiT-tagged vcircRNAs, 3 × 10⁵ Huh7 cells were seeded overnight in 6-well plates and then transfected with pooled siRNAs mentioned above targeting different IRES-containing vcircRNAs at a final concentration of 100 nM. After incubation for 24 h, cells were transfected with 3 μg IVT RNAs of HCV-c10/405-Hibit and HCV-c25/1277-HiBiT. Further after 24 h, HiBiT expression was measured by the HiBiT lytic assay (Promega). To test off-target effects of siRNAs, Huh7.5.1 cells were similarly transfected with 100 nM of control siRNA, or pooled siRNAs against cluster I vcircRNAs or cluster II vcirc 873/1151 or cluster III vcirc7952/8189 in 12-well plates. After 24 h, 1 μg in vitro transcribed HCV-Nluc-GND RNA was introduced into cells. After additional 16 h, HCV Nluc was measured by NanoGlo assay (Promega). To determine the effects of vcircRNAs on HCV replication, 100 nM of siRNA duplexes were transfected into Huh7.5.1 cells in a 12-well plate using Dharmafect I (Dharmacon) and incubated overnight. Cells were then infected with JFH-1 at a MOI of 0.5. Three days later, viral titers in cell supernatants were analyzed by focus-forming unit assay. Total RNA was extracted from cell lysates and HCV viral RNA levels were assessed by real-time qPCR (Primers are listed in Dataset S4).

Fluorescent Focus-Forming Assay.

Infectious titers were determined by a fluorescent focus forming units (FFU) assay. Huh7.5.1 cells (6 × 10⁴) were seeded in a 48-well plate and incubated overnight. Serial dilutions of virus stock were added to cells. After 6 h of incubation, the diluted virus supernatant was removed and replaced with fresh medium (2% FBS). At day 3 postinfection, cells were washed once with PBS and fixed with 3.2% paraformaldehyde. Intracellular HCV core protein was stained by using a mouse monoclonal antibody directed against HCV core (C7-50, Abcam) at 1:1,000 dilution in 1% fish gelatin/PBS (37 °C/1 h) and an AlexFluor 488–conjugated goat anti-mouse antibody (Invitrogen) at 1 μg/mL (37 °C/1 h). Whole-well fluorescent focus forming units were counted at appropriate dilutions using an inverted microscope Keyence BZ-X700.

Overexpression of the Polypeptide Translated from vcRNA10/405 ORF.

To construct overexpression plasmids for the translational product of vc10/405, two gBlock fragments were synthesized by IDT, which contain a HiBiT tag and a linker (GSSG) either N-terminally or C-terminally fused to c10/405 ORF. These two fragments were fused with pcDNA3.1(+) vector by InFusion cloning to generate pC-hibit-10/405ORF and pN-hibit-10/405ORF. In addition, a third plasmid pSTOP-N-hibit-10/405ORF was constructed as a negative control in which two stop codons were introduced immediately after AUG. For transfection, 8 × 10⁴ Huh7.5.1 cells were seeded in 24-well plate. On the next day, 200 ng of IVT RNA of HCV-Nluc-WT was transfected using the TransIT mRNA transfection kit (Mirus Bio). After 24-h incubation at 37 °C, cells were transfected with 400 ng of each plasmid DNA using Lipofectamine 3000 and P3000 Reagent (Invitrogen) according to the manufacturer’s instructions. After additional 24 h, Nluc activity from HCV was measured by the NanoGlo assay.

Statistical Analysis.

Ordinary one-way ANOVA or two-way ANOVA was used to evaluate the data for statistical differences (*P < 0.05, **P < 0.01, ***P < 0.001, and ****P < 0.0001) in GraphPad Prism (version 9.0).

Supplementary Material

Appendix 01 (PDF)

Dataset S01 (XLSX)

Dataset S02 (XLSX)

Dataset S03 (XLSX)

Dataset S04 (XLSX)

Data, Materials, and Software Availability

The raw RNA-sequencing (RNA-seq) data were deposited in the National Center for Biotechnology Information (NCBI) Sequence Read Archive under the accession number PRJNA1068101 at http://www.ncbi.nlm.nih.gov/bioproject/1068101 (37). Other data in this study are included in the article and/or SI Appendix.