PACT- and RIG-I-Dependent Activation of Type I Interferon Production by a Defective Interfering RNA Derived from Measles Virus Vaccine

ABSTRACT

The live attenuated measles virus vaccine is highly immunostimulatory. Identification and characterization of its components that activate the innate immune response might provide new strategies and agents for the rational design and development of chemically defined adjuvants. In this study, we report on the activation of type I interferon (IFN) production by a defective interfering (DI) RNA isolated from the Hu-191 vaccine strain of measles virus. We found that the Hu-191 virus induced IFN-β much more potently than the Edmonston strain. In the search for IFN-inducing species in Hu-191, we identified a DI RNA specifically expressed by this strain. This DI RNA, which was of the copy-back type, was predicted to fold into a hairpin structure with a long double-stranded stem region of 206 bp, and it potently induced the expression of IFN-β. Its IFN-β-inducing activity was further enhanced when both cytoplasmic RNA sensor RIG-I and its partner, PACT, were overexpressed. On the contrary, this activity was abrogated in cells deficient in PACT or RIG-I. The DI RNA was found to be associated with PACT in infected cells. In addition, both the 5′-di/triphosphate end and the double-stranded stem region on the DI RNA were essential for its activation of PACT and RIG-I. Taken together, our findings support a model in which a viral DI RNA is sensed by PACT and RIG-I to initiate an innate antiviral response. Our work might also provide a foundation for identifying physiological PACT ligands and developing novel adjuvants or antivirals.

IMPORTANCE The live attenuated measles virus vaccine is one of the most successful human vaccines and has largely contained the devastating impact of a highly contagious virus. Identifying the components in this vaccine that stimulate the host immune response and understanding their mechanism of action might help to design and develop better adjuvants, vaccines, antivirals, and immunotherapeutic agents. We identified and characterized a defective interfering RNA from the Hu-191 vaccine strain of measles virus which has safely been used in millions of people for many years. We further demonstrated that this RNA potently induces an antiviral immune response through cellular sensors of viral RNA known as PACT and RIG-I. Similar types of viral RNA that bind with and activate PACT and RIG-I might retain the immunostimulatory property of measles virus vaccines but would not induce adaptive immunity. They are potentially useful as chemically defined vaccine adjuvants, antivirals, and immunostimulatory agents.

INTRODUCTION

The innate immune response is the front line of host defense against pathogens. Such a response is mediated through the recognition of pathogen-associated molecular patterns (PAMPs) by host pattern recognition receptors (PRRs) (1–3). One major cytoplasmic sensor of viral RNA is RIG-I, which recognizes both the 5′-di/triphosphate end and the double-stranded region of viral RNA to induce an innate immune response (4–7). Optimal RIG-I-dependent production of type I interferons (IFNs) requires the double-stranded RNA-binding protein PACT, which facilitates RIG-I activation by intermediate-length poly(I·C) (8). However, the identity and nature of the viral RNA sensed by PACT and RIG-I remain to be elucidated. In searching for viral ligands of PACT and RIG-I, we focused on the vaccine strains of measles virus (MV), which are known to be potent inducers of type I IFNs (9–11).

MV belongs to the family Paramyxoviridae and contains a nonsegmented negative-stranded genome. MV is highly contagious and spreads through respiratory droplets. Patients with the disease commonly present with mild symptoms, such as fever, coryza, and rash, but serious complications, including pneumonia and damage to the central nervous system, may also occur (12). The devastating impact of MV has largely been contained since the development of a live attenuated vaccine in 1963. The MV vaccine is one of the most successful vaccines in human history and has nearly completely eliminated this once inevitable disease in developed countries (12, 13). However, MV still causes substantial morbidity and mortality in unvaccinated children (14).

The MV vaccine is safe and effective (13). A trivalent measles, mumps, and rubella virus vaccine is also safe, well tolerated, and highly efficacious (15). In contrast to MV infection, which results in immunosuppression, infection with a live attenuated MV vaccine induces strong protective immunity (12). Defining the molecular basis of the difference between wild-type and vaccine strains of MV might not only shed light on the mechanism of attenuation but also reveal new strategies for the rational design of adjuvants and antivirals. In addition to the mutations found in vaccine strains that inactivate immunosuppressive proteins (16–18), the robust production of defective interfering (DI) RNA from vaccine but not wild-type strains is also thought to be critical to this difference between wild-type and vaccine strains (10, 11, 18, 19). As a subgenomic RNA naturally produced during the course of viral infection, DI RNA may not form an infectious virion but can interfere with viral replication partly through the induction of IFNs (20). MV vaccines were developed by adaptation in chick embryos and chick embryonic fibroblasts (MEFs). During this adaptation, strains that produced large amounts of DI RNA were selected (21). In stark contrast, DI RNA is undetectable in clinical samples from measles patients (22).

Among the different types of DI RNA produced by MV, the copy-back type generated by the error-prone viral RNA-dependent RNA polymerase is particularly noteworthy. This DI RNA is abundantly found in cells infected with vaccine strains of MV (18, 22). Although its biological function largely remains unknown, it contains a long double-stranded region and is known to be a potent inducer of type I IFNs (18, 23, 24). Thus, the overproduction of copy-back DI RNA by the vaccine strains leads to activation of the innate immune response. In other words, copy-back DI RNA functions as a built-in adjuvant of the live attenuated MV vaccines. In support of this notion, a Sendai virus-derived copy-back DI RNA has been shown to be an agonist of RIG-I that potently induces type I IFNs and exhibits adjuvant activity in mice immunized with an influenza vaccine (25–27). However, exactly how copy-back DI RNA of MV activates RIG-I remains incompletely understood. Since it bears some similarity to the intermediate-length poly(I·C) that activates PACT (8), it will be of interest to see whether copy-back DI RNA of MV is an agonist of PACT.

In this study, we characterized a copy-back DI RNA of the Hu-191 vaccine strain of MV. We found that Hu-191 potently induces type I IFNs and that this DI RNA is an activator of PACT and RIG-I. Our findings support a model in which this DI RNA is recognized by PACT and RIG-I to induce type I IFN and cytokine expression. Our work has implications for the rational design and development of novel immunomodulatory agents, antivirals, and vaccine adjuvants.

MATERIALS AND METHODS

Cell culture and transfection.

Cells were grown in Dulbecco's modified Eagle's medium with 10% fetal bovine serum. HEK293-p125 and HEK293-PACT were established by transfecting the p125-Luc reporter and FLAG-tagged PACT into HEK293 cells, respectively, and were selected with 600 μg/ml of G418. Vero cells stably expressing the measles virus receptor, human signaling lymphocytic activation molecule (hSLAM) (28, 29), were a kind gift from Wenbo Xu (China CDC, Beijing, China) and were maintained in 400 μg/ml of G418. RIG-I^−/−, MDA5^−/−, and PACT^−/− MEFs have previously been described (30–32).

A549 cells were transfected using the Lipofectamine 3000 reagent (Life Technologies). HEK293-p125 and HEK293-PACT were transfected using the GeneJuice transfection reagent (Novagen). MEFs were transfected using the Lipofectamine 2000 reagent (Life Technologies). Cells were transfected with 0.25 μg/ml of DI RNA.

Plasmids, antibodies, and viruses.

The p125-Luc reporter, in which luciferase (Luc) expression is driven by the IFN-β promoter (4), was a generous gift from Takashi Fujita (Kyoto University, Kyoto, Japan). The expression plasmid for adenosine deaminase acting on RNA 1 (ADAR1) has been described previously (33). cDNA fragments for DI RNA and the DI RNA loop (DI-lp) were cloned into the HindIII and NheI sites of the in vitro transcription vector p2RZ using the following primer pairs: for DI RNA, 5′-GAATTCAAGC TTTAATACGA CTCACTATAG GGACCAGACA AAGCTGGGAA-3′ and 5′-ACCATGGCTA GCGGGACCAG ACAAAGCTGG GAA-3′, and for DI-lp, 5′-GAATTCAAGC TTTAATACGA CTCACTATAG GGATCCATTC AGATATAGAG-3′ and 5′-ACCATGGCTA GCGTCATAAT AATCTGTTTC-3′.

Mouse anti-FLAG (M2) and anti-α-tubulin antibodies were purchased from Sigma-Aldrich. Rabbit anti-PACT antibodies were bought from Abcam.

MV Edmonston strain VR-24 was purchased from the American Type Culture Collection. MV vaccine strain Hu-191, also known as Shanghai-191 (34), was kindly provided by Wenbo Xu (China CDC, Beijing, China). Viruses were propagated in Vero cells stably expressing hSLAM. The viral titer used to calculate the multiplicity of infection was determined by plaque assay.

Quantitative real-time RT-PCR.

Total RNA extraction was performed using RNAiso Plus reagents (TaKaRa). cDNA synthesis was achieved with a PrimeScript reverse transcription (RT) master mix kit (TaKaRa). Real-time PCR was performed with SYBR Premix Ex Taq reagents (TaKaRa) in a StepOne real-time PCR system (Applied Biosystems). The normalized value in each sample was derived from the relative quantity of target mRNA divided by the relative quantity of GAPDH (glyceraldehyde-3-phosphate dehydrogenase) mRNA. The relative mRNA expression level was derived from 2^−ΔΔCT by use of the comparative threshold cycle (C_T) method. Two-tailed Student's t test was performed to statistically assess the differences between selected groups.

The primers used in quantitative real-time RT-PCR were as follows: for IFN-β, 5′-TTGAATGGGA GGCTTGAATA-3′ and 5′-GCCAGGAGGT TCTCAACAAT AG-3′; for IFN-α1, 5′-CCTCGCCCTT TGCTTTACT-3′ and 5′-GCATCAAGGT CCTCCTGTTA TC-3′; for IFN-γ, 5′-TTCAGCTCTG CATCGTTTTG-3′ and 5′-CATGTATTGC TTTGCGTTGG-3′; for CCL5, 5′-GCATCTGCCT CCCCATATT-3′ and 5′-AGCACTTGCC ACTGGTGTAG-3′; for CXCL10, 5′-CCATTCTGAT TTGCTGCCTT AT-3′ and 5′-TTTCCTTGCT AACTGCTTTC AGTA-3′; for the L gene, 5′-CAAAGATACT ATAGAGAAGC TA-3′ and 5′-ACCAGACAAA GCTGGGAATA GAAACTTCG-3′; and for GAPDH, 5′-AGAAGGCTGG GGCTCATTTG-3′ and 5′-CTGTGGTCAT GAGTCCTTC-3′.

cDNA cloning and folding analysis of DI RNA.

cDNA of copy-back DI RNA was cloned by RT-PCR using a single primer. A series of nonoverlapping primers was tested. The sequences of all these primers perfectly matched the sequences of the genomes of both the Edmonston and Hu-191 strains. The sequence of the primer that amplified the longest copy-back DI RNA was 5′-ACCAGACAAA GCTGGGAATA GAAA-3′.

RNA folding analysis was performed with RNAfold software, supported by the Institute for Theoretical Chemistry, University of Vienna (35). The folding prediction was optimized by calculating the partition function and base-pairing probability matrix, in addition to the minimum free energy structure.

ELISA.

Reagents for human IFN-β enzyme-linked immunosorbent assay (ELISA) were purchased from PBL Assay Science. The amounts of IFN-β in the culture media were determined using the manufacturer's protocol. The range of the detection limit was 1.2 to 150 pg/ml.

Luciferase reporter assay.

Cells were harvested at 16 h after transfection. Cells were lysed using passive lysis buffer, and a dual-luciferase reporter assay was carried out as described previously (36–38). The pRL-TK Renilla luciferase reporter (Promega) was used as an internal control to normalize for transfection efficiency. Standard deviations (SDs) were calculated on the basis of the results of three independent experiments. Intermediate-length poly(I·C) with an average size of 0.2 to 1 kb, also known as low-molecular-weight poly(I·C), was purchased from InvivoGen.

In vitro transcription.

In vitro transcription was carried out using a MEGAscript T7 transcription kit (Ambion). One microgram of linear DNA was used as the template, and the reaction was carried out at 37°C for 4 h. RNA products were purified by phenol-chloroform extraction.

RNAi.

RNA interference (RNAi) experiments were performed as described previously (8, 39). A549 cells were transfected with 50 nM small interfering RNA (siRNA) using Lipofectamine 3000. siRNA sequences were as follows: for siRNA against green fluorescent protein (siGFP), 5′-GCAAGCUGAC CCUGAAGUU-3′; for siRNA against DI RNA (siDI), 5′-CAAACCCCCC AUCCAUUCAG-3′; for one independent siRNA against PACT (siPACT#1), 5′-CACCGAUUCA GGUAUUACA-3′; and for a second independent siRNA against PACT (siPACT#2), 5′-GAGAGAAUAU ACUACAAUU-3′.

RNA immunoprecipitation.

RNA immunoprecipitation was performed essentially as described previously (36, 39). Cells were cross-linked by incubation with formaldehyde (from a stock of 37% HCHO and 10% methanol) to a final concentration of 1% (vol/vol, 0.36 M) at room temperature for 10 min with gentle mixing. The cross-linking reaction was stopped by the addition of glycine, pH 7.0, to a final concentration of 0.25 M and incubation at room temperature for 5 min. Cells were harvested using radioimmunoprecipitation assay buffer (150 mM NaCl, 50 mM Tris, pH 7.4, 1 mM EDTA, 1% NP-40, 0.5% sodium deoxycholate, 0.05% SDS) with a 5× protease inhibitor cocktail (Roche). Cells were then lysed with 20 rounds of sonication for 3 s with pauses of 5 s between each round. Insoluble materials were removed by centrifugation at top speed for 10 min at 4°C. Lysates were incubated with protein G agarose beads (Invitrogen) coated with specific antibody at 4°C for 4 h. Immunoprecipitates were resuspended in 100 μl of 50 mM Tris, pH 7.0, 5 mM EDTA, 1% SDS, and 10 mM dithiothreitol. The cross-link was reversed by incubation at 70°C for 45 min. RNA was then extracted from the immunoprecipitates using RNAiso Plus reagents (TaKaRa).

Nucleotide sequence accession numbers.

The sequence of the copy-back DI RNA of Hu-191 has been deposited in the GenBank nucleotide sequence database under accession number KU057377.

RESULTS

Induction of IFN-β and cytokines by MV vaccine strain.

Many vaccine strains of MV potently activate the production of type I IFNs, which contributes to the subsequent induction of protective immunity (10, 11). To verify this, we first examined the induction of IFNs and cytokines by a vaccine strain of MV named Hu-191. The Hu-191 vaccine was developed in Shanghai, China, using a locally isolated strain and has been the most extensively used MV vaccine in the Chinese vaccination program (40, 41). In fact, it has been used for measles immunization in all provinces in China since 2003. It is therefore of interest to compare the abilities of Hu-191 and the more virulent Edmonston strain to induce IFNs and cytokines. The Edmonston strain was originally isolated from the blood of a patient in the acute phase of typical measles, and several vaccine strains were derived from it (34). We infected human alveolar basal epithelial A549 cells expressing the MV receptor SLAM with the Hu-191 and Edmonston viruses and determined the expression levels of IFNs and cytokines by quantitative real-time RT-PCR. Hu-191 potently induced the IFN-β transcript and protein in infected cells, but the Edmonston strain had only a mild effect (Fig. 1A and B). Since the levels of IFN-β mRNA and protein correlated well in Hu-191-infected cells, only mRNA was analyzed in our subsequent experiments. Similar patterns were also observed for the cytokines CCL5/RANTES and CXCL10/IFN-inducible protein 10 (IP10) (Fig. 1C and D), both of which are known to be induced by IFNs (42). In contrast, neither the Hu-191 nor the Edmonston virus dramatically induced IFN-γ mRNA (Fig. 1E), although both viruses replicated and transcribed well in these cells, as shown by the levels of viral RNA (Fig. 1F). Thus, the expression of type II IFN was not significantly activated by MV in A549 cells in our experimental setting. Nevertheless, our results demonstrate that Hu-191 is indeed a potent inducer of IFN-β and cytokines.

FIG 1.

Expression profiling of IFN and IFN-stimulated genes in A549 cells challenged with MV. SLAM-expressing A549 cells were infected with either the Hu-191 (191) or Edmonston (Ed) strain of MV at a multiplicity of infection of 0.5 for 18 h. Quantitative real-time RT-PCR analysis was performed to analyze IFN-β (A), CCL5 (C), CXCL10 (D), IFN-γ (E), L gene (F), and GAPDH mRNA transcripts. The level of target gene RNA expression relative to the level of GAPDH mRNA expression was calculated by the comparative C_T method. (B) The levels of IFN-β protein recovered from the culture medium were also determined by ELISA. All relative expression values were normalized to the value for the mock-infected sample. Results represent the means from three independent experiments, and error bars indicate SDs. The differences in the levels of IFN-β mRNA (A) and protein (B), CCL5 mRNA (C), as well as CXCL10 mRNA (D) expression between the Hu-191 and Edmonston strains were statistically significant by two-tailed Student's t test, with P values being equal to 0.00034 (highlighted by ***), 0.0032 (highlighted by **), 4.8 × 10⁻⁵ (highlighted by ***), and 0.0012 (highlighted by **), respectively.

Isolation of IFN-inducing DI RNA from Hu-191.

To look for IFN-inducing components in Hu-191, we focused on viral RNA, particularly DI RNA of the copy-back type, which is known to be a potent IFN inducer (10, 11). A recent analysis of RIG-I- and MDA5-associated RNA from MV-infected cells using deep sequencing also indicated an exclusive enrichment of RNA derived from the 5′ end trailer region of the viral genome (43). This is compatible with the notion that copy-back DI RNA might be RIG-I and MDA5 agonists. Thus, we set out to verify the production of copy-back DI RNA in Hu-191. Vero cells expressing hSLAM were infected with the Hu-191 and Edmonston viruses. Total RNA was extracted, and copy-back DI RNA was specifically isolated using a single-primer RT-PCR method. Since it contains complementary 5′ and 3′ ends, copy-back DI RNA but not other species can be amplified with a single primer serving as both forward and reverse primers (Fig. 2A and B). Several nonoverlapping primers perfectly complementary to the 5′ end trailer region of the genomes of both Hu-191 and Edmonston strains were tested. One longest DI RNA of the copy-back type was successfully amplified from Hu-191-infected cells by one primer (Fig. 2C). This DI RNA unlikely contains additional nucleotide residues, since RT-PCRs with any one of the nonoverlapping primers complementary to adjacent sequences outside the region were nonproductive. In this setting, no copy-back DI RNA was found in cells infected with the Edmonston strain. The DNA band was excised from the gel and sequenced to reveal the identity of the DI RNA. Folding analysis predicted that this copy-back DI RNA of Hu-191 would form a double-stranded stem region of 206 bp. The loop region of 823 nucleotides could also be highly structured (Fig. 2B). The GC content of this DI RNA was 40.8%.

FIG 2.

Isolation of an IFN-inducing DI RNA from the MV Hu-191 strain. (A) Rationale of the single-primer RT-PCR. Copy-back DI RNA has a characteristic complementary pair of 5′ and 3′ ends derived from the genome trailer sequence, represented by regions Tr and Tr′, respectively. The noncomplementary region is shown as a loop. A single primer can amplify copy-back DI RNA, as region Tr′ is complementary to region Tr. (B) Predicted folding pattern of DI RNA. The DI RNA sequence was subjected to folding analysis using RNAfold software. The DI RNA is predicted to form a 206-bp double-stranded stem region and an 823-nucleotide (nt) loop region. (C) DI RNA detection. Vero cells stably expressing hSLAM were infected with either the Hu-191 (191) or the Edmonston (Ed) virus at a multiplicity of infection of 0.5 for 72 h. Total RNA was extracted, and RT-PCR was performed to detect the DI RNA of the copy-back type. The MV L gene was included as a control (ctrl). Lane M, molecular mass markers. (D) DI RNA activation of the IFN-β promoter. HEK293 cells stably expressing p125-Luc were transfected with in vitro-transcribed MV DI RNA for the indicated duration. A dual-luciferase assay was performed, and relative luciferase activity (in arbitrary units) is shown. Error bars indicate SDs from three independent experiments.

In order to have a controllable amount of this DI RNA in experiments, the in vitro-transcribed product was used in subsequent studies. The IFN-inducing property of this DI RNA was tested with HEK293 cells stably expressing p125-Luc driven by the IFN-β promoter. Cells transfected with in vitro-transcribed DI RNA were harvested at different time points, and the luciferase activity recovered showed that DI RNA alone was sufficient to activate the IFN-β promoter potently as early as 8 h after transfection (Fig. 2D). This suggested that this DI RNA might be a major factor contributing to the ability of Hu-191 to induce IFN-β production.

To further verify the role of this DI RNA in Hu-191-induced IFN-β production, we used an siRNA directed against the junction sequence present in the DI RNA but not in the viral genome. Preexpression of this siRNA, named siDI, mitigated the activation of IFN-β mRNA expression by Hu-191 (Fig. 3A; compare bar 3 to bar 2). This result provides crucial support for the role of DI RNA in IFN-β induction by Hu-191. In light of the recent report that ADAR1, which catalyzes A-to-I RNA editing, can specifically edit and destabilize measles virus DI RNA (44), we asked how expression of ADAR1 might affect Hu-191-induced activation of IFN-β production. An inhibitory effect of ADAR1 on IFN-β induction by Hu-191 was observed (Fig. 3B; compare bar 4 to bar 3), and it was even more pronounced than that of siDI (Fig. 3A). Thus, compromising DI RNA expression prevents the induction of IFN-β production by Hu-191.

FIG 3.

Suppression of DI RNA expression counteracts IFN induction. (A) siRNA knockdown of DI RNA. SLAM-expressing A549 cells were transfected with either a DI RNA-targeting siRNA (siDI) or an irrelevant siRNA against GFP (siGFP) for 48 h. Cells were then infected with the Hu-191 virus at a multiplicity of infection of 0.5 for 16 h. Total RNA was extracted, and quantitative real-time RT-PCR was performed to detect IFN-β mRNA expression. Results represent the means from three independent experiments, with error bars indicating SDs. The difference between the samples whose results are shown as bars 2 and 3 was statistically significant, as determined by a two-tailed Student's t test with a P value of 0.012 (highlighted by *). (B) Suppression of Hu-191-induced IFN production by ADAR1. SLAM-expressing A549 cells were transfected with an ADAR1 expression plasmid for 48 h. Cells were then infected with the Hu-191 virus at a multiplicity of infection of 0.5 for 16 h. Results are representative of those from three independent experiments, and error bars denote SDs. The difference between the samples whose results are shown as bars 3 and 4 was statistically significant, as determined by a two-tailed Student's t test with a P value of 0.0093 (highlighted by **).

Potentiation of the IFN-inducing activity of MV DI RNA by PACT and RIG-I.

The long double-stranded stem region of the DI RNA derived from the Hu-191 vaccine strain might serve as a PAMP recognized by PRRs. In particular, the 206-bp stem constitutes a perfect RIG-I ligand with both an intermediate-length double-stranded region and 5′-triphosphates (5, 6). To assess the impact of RIG-I and MDA5 on the induction of IFN-β production by this DI RNA, RIG-I and MDA5 were overexpressed in HEK293 cells. After 24 h, DI RNA was further transfected. Quantitative real-time RT-PCR results showed that RIG-I augmented DI RNA-induced IFN-β expression, indicating the requirement for RIG-I in host recognition of DI RNA (Fig. 4A; compare bar 3 to bar 2). In contrast, MDA5, which is mainly involved in the recognition of long double-stranded RNA (30, 45), did not show any stimulatory effect (Fig. 4A; compare bar 4 to bar 2).

FIG 4.

PACT and RIG-I augment IFN-β induction by MV DI RNA and the Hu-191 virus. (A) Plasmids expressing the indicated combinations of RIG-I, MDA5, and PACT were transfected into HEK293 cells. After 48 h, in vitro-transcribed MV DI RNA was further transfected. The levels of IFN-β and GAPDH mRNA were measured by quantitative real-time RT-PCR after 6 h. The level of IFN-β mRNA expression relative to the level of expression of the GAPDH mRNA transcript in a mock-transfected control sample was taken as 1. Results are the means from three independent experiments, and error bars indicate SDs. The difference between the samples whose results are shown as bars 3 and 6 was statistically significant, as determined by a two-tailed Student's t test with a P value equal to 0.0089 (highlighted by **). (B) The same transfection described in the legend to panel A was performed in SLAM-expressing A549 cells. After 48 h, cells were infected with the Hu-191 virus at a multiplicity of infection of 0.5 for 18 h. Error bars denote SDs from three independent experiments. The difference between the samples whose results are shown as bars 3 and 6 was statistically significant, as determined by two-tailed Student's t test with a P value equal to 0.0054 (highlighted by **).

We previously demonstrated that the double-stranded RNA-binding protein PACT is required for optimal activation of RIG-I by intermediate-length poly(I·C) with an average size of 0.2 to 1 kb but not by short 5′-triphosphate RNA (8). This DI RNA of MV has both a base-paired region of 206 bp and a 5′-triphosphate end, so it was of great interest to see whether PACT might also be involved in its recognition in the activation of IFN production. To test this idea, we overexpressed PACT together with RIG-I in HEK293 cells and challenged them with DI RNA. Indeed, the expression of PACT further enhanced IFN-β induction by DI RNA in the presence of RIG-I (Fig. 4A; compare bar 6 to bar 3). These results were compatible with a model in which PACT facilitates the activation of RIG-I by this DI RNA of MV. Thus, PACT and RIG-I cooperate to induce IFN-β expression in the innate sensing of MV DI RNA.

We next extended our analysis to the role of RIG-I, MDA5, and PACT in the viral induction of IFN-β in Hu-191-infected A549 cells. Similar to the pattern described above for DI-RNA, IFN-β expression in infected cells was the most robust when both PACT and RIG-I were overexpressed (Fig. 4B; compare bar 6 to bar 3). This is generally in keeping with the notion that PACT cooperates with RIG-I to sense MV DI RNA derived from the Hu-191 vaccine strain. Consistent with previous findings for other MV strains (17, 46), MDA5 was also capable of augmenting IFN-β induction by Hu-191, and this effect was further potentiated by PACT (Fig. 4B; compare bar 7 to bar 4). Plausibly, DI RNA-independent but MDA5-dependent activation of IFN-β production mediated by long double-stranded RNA, such as the replicative form of viral RNA, might also occur in Hu-191-infected cells.

Loss of RIG-I or PACT abolishes DI RNA-induced IFN-β expression.

A loss-of-function study is required to fully establish the importance of RIG-I and PACT in the recognition of MV DI RNA. To this end, we first asked how DI RNA-induced IFN-β production might be affected upon depletion of PACT by siRNA. Two independent siRNAs against PACT were designed, and their effectiveness in achieving a high degree of knockdown was verified by Western blotting (Fig. 5A). When PACT was depleted by the two siRNAs, the activation of IFN-β expression by DI RNA in A549 cells was diminished (Fig. 5B; compare bars 4 and 6 to bar 2). Thus, PACT is required for the DI RNA-induced activation of IFN-β expression.

FIG 5.

PACT and RIG-I are required for IFN induction by MV DI RNA. (A) Effectiveness of siPACT. A549 cells were transfected with two independent siRNAs against PACT (siPACT#1 and siPACT#2) and one against GFP (siGFP). Cells were harvested after 48 h, and Western blotting was performed. PACT and α-tubulin were visualized using anti-PACT and anti-α-tubulin (α-tub) antibodies, respectively. (B) Effect of PACT knockdown. In vitro-transcribed MV DI RNA was transfected 48 h after siRNA transfection. The levels of the IFN-β transcript relative to those of GAPDH mRNA were measured by quantitative real-time RT-PCR and were normalized to the readouts of the mock-transfected siGFP control. Results are the means from three independent experiments, and error bars indicate SDs. The differences between the samples whose results are shown as bars 2 and 4 as well as bars 2 and 6 were statistically significant by a two-tailed Student's t test with P values equal to 0.022 (highlighted by *) and 0.026 (highlighted by *), respectively. (C) Effect of RIG-I or PACT knockout. Wild-type (WT), RIG-I^−/−, MDA5^−/−, and PACT^−/− MEFs were transfected with poly(I·C) (IC) and in vitro-transcribed MV DI RNA (DI). The levels of mRNA for the IFN-β gene were measured by quantitative real-time RT-PCR after 6 h. The mRNA levels were normalized to the readouts of GAPDH mRNA, and the fold induction compared with the level of induction for the mock-transfected controls is shown. Error bars denote SDs from three independent experiments. The differences between the samples whose results are shown as bars 3 and 6 as well as bars 3 and 12 were statistically significant according to a two-tailed Student's t test with P values equal to 4.2 × 10⁻⁵ (highlighted by ***) and 2.6 × 10⁻⁵ (highlighted by ***), respectively.

Our overexpression experiments indicated a role for both PACT and RIG-I but not MDA5 in the innate sensing of MV DI RNA (Fig. 4). In order to define the requirement for RIG-I, MDA5, and PACT in DI RNA-induced activation of IFN-β expression, we employed MEFs derived from RIG-I^−/−, MDA5^−/−, or PACT^−/− mice. The abilities of MV DI RNA and intermediate-length poly(I·C) to activate the expression of the IFN-β transcript in the MEFs of the different genotypes were then compared. Consistent with our results in the overexpression study (Fig. 4), the IFN-β-inducing activity of intermediate-length poly(I·C) and DI RNA was affected only moderately or minimally in the absence of MDA5 (Fig. 5C; compare bar 8 to bar 2, and compare bar 9 to bar 3). In sharp contrast, neither of them was capable of inducing the IFN-β transcript in RIG-I^−/− cells (Fig. 5C; compare bar 5 to bar 2, and compare bar 6 to bar 3). It is noteworthy that DI RNA could not activate the expression of IFN-β mRNA in PACT^−/− cells (Fig. 5C; compare bar 12 to bar 3). In the control group, the loss of PACT also impaired the induction of the IFN-β transcript by intermediate-length poly(I·C) (Fig. 5C; compare bar 11 to bar 2). Hence, RIG-I and PACT are required for the DI RNA-induced activation of IFN-β production. Results from our gain-of-function and loss-of-function experiments consistently support a model in which MV DI RNA is recognized by PACT and RIG-I.

Association of MV DI RNA with PACT.

The critical role of PACT in the recognition of DI RNA (Fig. 3 and 4) suggested that PACT and DI RNA might interact with each other. To explore this possibility, we performed RNA immunoprecipitation in HEK293 cells that stably expressed FLAG-tagged PACT and that were transfected with DI RNA. The presence of both DI RNA and PACT in the anti-FLAG precipitate indicated the association of the two entities (Fig. 6A; compare lane 4 to lane 3). This association was specific, since it was not seen in the absence of PACT (Fig. 6A, lane 2) and PACT was not bound to the L gene of Hu-191 (Fig. 6A, lane 1). The association of DI RNA and PACT was also observed in Hu-191-infected A549 cells (Fig. 6B; compare lane 2 to lane 1). Results from the titration of the amount of PACT protein in the immunoprecipitate combined with those of quantitative real-time RT-PCR analysis of the amount of DI RNA in the immunoprecipitate suggested the association of 268 copies of DI RNA with 1 ng of PACT protein. Thus, PACT might have a role in selecting and concentrating DI RNA. Our results are generally compatible with a model in which PACT selectively interacts with MV DI RNA in infected cells to transmit an activation signal to RIG-I in the induction of IFN-β expression.

FIG 6.

Association of MV DI RNA with PACT. (A) HEK293 cells stably expressing FLAG-PACT were transfected with MV DI RNA. After 6 h and before harvest, cells were treated with formaldehyde. Cell lysates were immunoprecipitated with anti-FLAG antibody, and total RNA was extracted from the precipitates. PACT and α-tubulin in the input lysate (5%) were visualized by Western blotting with anti-FLAG and anti-α-tubulin antibodies, respectively. DI RNA was detected by agarose gel electrophoresis after RT-PCR with DI RNA-specific primers. (B) SLAM-expressing A549 cells were infected with the Hu-191 virus at a multiplicity of infection of 0.5. After 18 h, cells were treated with formaldehyde and cell lysates were immunoprecipitated (IP) with anti-PACT antibodies. The input lysate (5%) was also probed for PACT and α-tubulin. The experiments were repeated twice, and similar results were obtained each time.

Both 5′-di/triphosphates and the double-stranded stem region are essential for DI RNA recognition.

To determine the crucial features on MV DI RNA that are important in the activation of an innate IFN response, we used two different approaches. First, in vitro-transcribed DI RNA was treated with calf intestinal phosphatase (CIP) to remove 5′-di/triphosphates. Second, a mutant construct was made to generate a DI RNA without the 206-bp stem region by in vitro transcription. The CIP-treated DI RNA (DI-CIP) and the DI RNA with the loop alone (DI-lp), as shown in Fig. 7A, were then used to transfect A549 cells. Notably, their abilities to induce the expression of IFN-β and CCL5 transcripts were severely impaired (Fig. 7B and C; compare bars 4 and 5 to bar 3), indicating the importance of 5′-di/triphosphates and the stem region in the sensing of DI RNA by PACT and RIG-I. In the control experiments, all three types of MV DI RNA were incapable of inducing IFN-α1 or IFN-γ (Fig. 7D and E). Collectively, our results demonstrate the requirement for both 5′-di/triphosphates and the double-stranded stem region in the activation of IFN production by MV DI RNA.

FIG 7.

Both 5′-di/triphosphates and the stem region in MV DI RNA are important for IFN induction. (A) DI RNA samples used in this experiment. (Left) Diagrams of MV DI RNA (DI), DI RNA treated with calf intestinal phosphatase (DI-CIP), and the DI RNA mutant containing the loop only (DI-lp). (Right) In vitro-transcribed RNA samples were loaded into a formaldehyde agarose gel and visualized under UV. (B to E) Requirement for 5′-di/triphosphates and the stem region. Poly(I·C) (IC), DI RNA, DI-CIP, and DI-lp were transfected into A549 cells. Quantitative real-time RT-PCR analysis was performed after 6 h to measure the levels of the IFN-β (B), CCL5 (C), IFN-α1 (D), IFN-γ (E), and GAPDH mRNA transcripts. The levels of target gene mRNA expression relative to the level of GAPDH mRNA expression were calculated by the comparative C_T method. All expression values were normalized to those for the mock-transfected control sample. Results are the means from three independent experiments, and error bars indicate SDs. The differences in the levels of IFN-β mRNA between DI RNA and DI-CIP as well as between DI RNA and DI-lp were statistically significant, as judged by a two-tailed Student's t test with P values equal to 3.2 × 10⁻⁵ (highlighted by ***) and 0.00028 (highlighted by ***), respectively (B), and the differences in the levels of CCL5 mRNA between DI RNA and DI-CIP as well as between DI and DI-lp were statistically significant, as judged by a two-tailed Student's t test with P values equal to 0.00074 (highlighted by ***) and 0.00055 (highlighted by ***), respectively (C).

DISCUSSION

In this study, we identified and characterized an IFN-inducing DI RNA of the copy-back type from MV vaccine strain Hu-191. With 50 nucleotide substitutions, the Hu-191 virus is more divergent from the Edmonston strain than most other vaccine strains (41). It has not previously been documented whether the Hu-191 virus produces DI RNA like other vaccine strains, such as strains Zagreb, AIK-C, and CAM-70 (21, 34). In this regard, our work provides the first evidence for the production of DI RNA in Hu-191-infected cells. We demonstrate the IFN-inducing property of both the Hu-191 virus and its copy-back DI RNA with a long double-stranded stem region. Through both overexpression and loss-of-function experiments, we further established the requirement for PACT and RIG-I in DI RNA-induced activation of IFN-β production. Our results support a model in which the copy-back DI RNA of Hu-191 associates with PACT and its recognition by PACT and RIG-I leads to the activation of type I IFN production. This DI RNA is the first viral RNA shown to activate PACT and RIG-I. Our analysis in PACT^−/− cells provides key evidence in support of the essentiality of PACT in innate sensing of this DI RNA. We further dissected the molecular patterns in this DI RNA that are required for this sensing. We found that both the 5′-di/triphosphate end and the double-stranded region are indispensable for IFN induction. Our work sheds light on the interplay between viral RNA and PACT and thereby provides a glimpse of the model for PACT- and RIG-I-dependent innate RNA sensing. Our findings also provide the foundation for further analysis of physiological ligands of PACT and RIG-I.

Our results suggest that the Hu-191 vaccine strain of MV is a potent activator of type I IFNs and IFN-stimulated genes. Type I IFN production and signaling are important in the innate immune response, and they also facilitate the subsequent induction of adaptive immunity (2, 3). In this sense, potent activation of the type I IFN response by Hu-191 plausibly contributes to successful attenuation and the high immunogenicity of the MV vaccine. Other vaccine strains of MV and their DI RNAs have been shown to induce type I and III IFNs in dendritic cells (11). Although the same group recently reported that the production of type I and III IFNs in vivo was limited in macaques infected with wild-type and vaccine strains of MV free of a common species of DI RNA (47), this and other types of DI RNA are commonly found in MV vaccines (10, 19, 23, 44, 48). Although DI RNA was not detected in the Edmonston strain in our study, the cell type-dependent production of DI RNA by the Edmonston strain and Edmonston-derived vaccine strains has been shown previously (19). We cannot exclude the possibility that the Edmonston strain used in that study had already been adapted to Vero cells and selected for DI RNA production during this adaptation.

HEK293 cells are known to express low levels of RIG-I and PACT (8). Overexpression of RIG-I and PACT in HEK293 cells resulted in the further induction of IFN-β expression by copy-back DI RNA, implicating that endogenous expression of RIG-I and PACT might be suboptimal. Thus, it will be of interest to clarify whether the DI RNA-induced activation of IFN-β production in HEK293 cells seen in our study might be supported by basal levels of RIG-I and PACT, although the possibility of the involvement of other RNA-binding proteins with a complementary function cannot be totally excluded. We demonstrated the requirement for PACT and RIG-I in the innate sensing of copy-back DI RNA of MV using gene-knockout MEFs. Our results are generally consistent with our previous finding that PACT facilitates RIG-I activation by intermediate-length poly(I·C) of 0.2 to 1 kb but not by short 5′-triphosphate RNA (8). On the basis of the model for the length-dependent recognition of viral RNA by RIG-I and MDA5 (30), the DI RNA of Hu-191 with a 206-bp double-stranded region should be an ideal ligand for RIG-I but not MDA5. So it is not too surprising that either expression or the loss of expression of MDA5 had a minimal effect on the IFN-inducing activity of this DI RNA. However, MDA5 has previously been shown to be influential in the activation of IFN production by Sendai virus DI RNA (49). Further investigations are required to clarify whether the Sendai virus DI RNA examined in that study might be exceedingly long or the requirement for MDA5 might be cell type and virus specific. Although intermediate-length poly(I·C) without 5′-triphosphates sufficiently activates PACT and RIG-I (8, 30), our work suggested that both 5′-di/triphosphates and the double-stranded region are required for the recognition of MV DI RNA by PACT and RIG-I. This is generally consistent with the known molecular patterns recognized by RIG-I (6, 50). Plausibly, other DI RNAs and viral RNAs with both 5′-di/triphosphates and a double-stranded region are agonists of PACT and RIG-I.

In MV-infected cells, DI RNA is encapsidated, and the nucleocapsid (N) protein is thought to contact 6 nucleotides; thus, the total number of nucleotides in DI RNA should be a multiple of six (51). The DI RNA that we identified from Hu-191 vaccine strain contains 1,235 nucleotides, which is 1 nucleotide short of a multiple of six. Although many MV DI RNAs contain nucleotides that are indeed a multiple of six (10, 23, 44), exceptions to this rule of six have also been reported (10, 44). In particular, the number of nucleotides in one DI RNA found in the ICV laboratory-adapted strain of MV was also 1 nucleotide short of a multiple of six (10). In addition, we found the association of DI RNA with PACT in Hu-191-infected cells. It will therefore be of great interest to determine whether PACT, N, and DI RNA are in the same protein-RNA complex that mediates innate RNA sensing.

Our findings on the requirement for PACT for optimal activation of RIG-I by copy-back DI RNA support the critical involvement of PACT in innate sensing of RNA viruses. This is consistent with the recent observations that viral induction of type I IFNs is severely reduced in the absence of PACT in cells infected with Sendai virus, Ebola virus, herpes simplex virus 1, or influenza A virus (32, 52, 53). PACT is also targeted by different viral IFN-antagonizing proteins (32, 36, 52–54). Although deletion of PACT in mice was reported to be embryonically lethal by one group (55), two other independently constructed PACT mutant mice did not show this phenotype but displayed similar developmental defects (31, 56, 57). Whereas PACT knockdown in mice had a significant impact on cytokine induction and viral infection (58), the loss of PACT was found by another group to have no effect on Sendai virus-induced activation of STAT1 and IFIT1 production (31). It was not known whether the induction of STAT1 and IFIT1 had already plateaued in that study. In stark contrast, we and others documented the severely impaired induction of IFN-β in PACT^−/− MEFs derived from the same mice (32, 52). Nevertheless, more detailed analyses of IFN production, the IFN response, and viral replication in PACT^−/− cells and mice infected with different viruses would clarify these discrepancies.

Several lines of additional work are required to derive further mechanistic insight into the sensing of copy-back DI RNA by PACT and RIG-I. First, the association between RIG-I and DI RNA and the impact of PACT on this association should be further analyzed. If PACT indeed acts to select and concentrate RNA ligands for RIG-I, it should bind with these ligands with a higher affinity. Second, whether PACT and RIG-I bind simultaneously or sequentially with DI RNA should be determined. Third, structural analysis should be performed to elucidate whether and how DI RNA might induce conformational changes in PACT and RIG-I. As a natural ligand of PACT and RIG-I, this copy-back DI RNA might be used to shed light on how it triggers their activation. Finally, ADAR1 is known to edit and destabilize measles virus DI RNA (44). ADAR1 also suppresses RIG-I recognition of double-stranded RNA by A-to-I editing, RNA binding, and other mechanisms (44, 59–61). In addition, ADAR1 inhibits IFN induction, protein kinase R activation, stress granule formation, and MV replication in MV-infected cells and mice (62–66). We showed that expression of ADAR1 dampened the Hu-191-induced activation of IFN-β production. It will be of interest to clarify whether ADAR1 might also affect DI RNA recognition by PACT and RIG-I.

Other than host factors, viral proteins might also affect DI RNA sensing as a means to evade innate immunity. In particular, V proteins from different paramyxoviruses are known to be suppressors of IFN signaling (67–69). Remarkably, it was shown that the MV V protein inhibits poly(I·C)-induced IFN-β promoter activation (17). Such inhibition might also have an impact on DI RNA-induced IFN-β production. Indeed, the V protein of Sendai virus has already been shown to be capable of blocking DI RNA-induced activation of the IFN-β promoter (25). There are several nucleotide variations in the V genes between the Hu-191 and Edmonston strains (34, 41). Whether these variations might have any impact on the activation of IFN production by MV copy-back DI RNA warrants further investigations.

The molecular determinants of DI RNA production are poorly understood. DI RNAs exist in both the positive sense and the negative sense (20). However, in our experimental setting, only one major DI RNA of negative sense was detected. This DI RNA was probably generated from the transcription of the positive-sense antigenome of MV. It remains to be determined whether there might be a bias in strand selection during DI RNA synthesis or most of the positive-sense DI RNAs are selectively degraded. A recent study reveals that in MV the loss of the C protein, which is a known suppressor of IFN-β expression (70), boosts the production of copy-back DI RNA through an unknown mechanism (18, 44). On the other hand, the cysteine-rich carboxy terminus of the V protein of MV can also bind with RNA and modulate the transcription of the MV genome (70, 71). It will therefore be intriguing to see whether the C and V proteins as well as the sequence variations in these proteins in Hu-191 might affect the production of DI RNA.

MV vaccines have been extensively tested as immunotherapeutic agents for cancer and other diseases (72). However, the activation of specific humoral and cellular immunity against MV prevents readministration. Copy-back DI RNAs in MV vaccines serve as built-in adjuvants to provoke innate immunity. They might retain the immunostimulatory property of MV vaccines but might not elicit adaptive immunity against MV. Just like a similar DI RNA from Sendai virus (27), they have the potential to be further developed as chemically defined vaccine adjuvants, antivirals, and immunostimulatory agents. In this regard, the copy-back DI RNA from Hu-191 merits further testing for its adjuvant activity in animal models.