Functional roles for viral RNA ribosylation in mammalian cells

Abstract

In mammalian cells, the addition of ADP-ribose to proteins and DNA plays well established roles in regulating cell function. Recently, RNA ribosylation was also found in mammalian cells under conditions of cell stress, though the functional consequences remain unclear. Here we find that infection with chikungunya virus, a positive strand RNA virus that causes frequent widespread epidemics, increases overall levels of RNA ribosylation in human fibroblasts. During infection, viral RNA is ribosylated by the PARP12 ribosyltransferase, which is counteracted by the viral nsP3 ribosylhydrolase. Increased viral RNA ribosylation resulted in decreased translation in cell-free systems and infected fibroblasts, and more rapid viral RNA decay. Further, ribosylated RNA potently induced the expression of antiviral host response genes. Together these data show the first functional consequences of RNA ribosylation in mammalian cells by showing that RNA ribosylation inhibits translation, decreases RNA stability, and creates a novel pathogen-associated molecular pattern that activates the host innate immune response. As macrodomains are present in multiple unrelated viruses, our data suggest RNA ribosylation is a novel component of cellular antiviral sensing pathway. These results define a framework for understanding how RNA ribosylation regulates RNA metabolism during viral infection.

Graphical Abstract

Graphical Abstract.

Introduction

RNA modifications play critical roles in regulating RNA activity and metabolism at multiple levels. For example, the addition of the 5′ 7-methylguanosine cap and its subsequent methylation modulate RNA translation, while other modifications such as methylation and acetylation impact RNA stability, localization, and degradation [1–4]. RNA modifications are especially important under conditions of cellular stress, such as viral infection, where they help integrate environmental signals and determine RNA fate [5, 6]. Understanding the functional consequences of RNA modifications is therefore critical for uncovering how cells adapt to stress and dynamically reprogram gene expression to promote cell survival.

ADP ribosylation, the covalent attachment of adenosine diphosphate ribose (ADPr) to substrates, has long been recognized as a conserved mechanism for responding to stress across diverse biological systems. Both proteins and nucleic acids can be ADP-ribosylated, and modification of both regulates cellular stress responses [7]. For example, in bacteria and plants, ribosylation of exposed DNA ends is recognized as a damage signal and helps recruit DNA repair proteins. Protein ribosylation of recruited factors further facilitates chromatin remodeling and promotes error-free DNA repair in bacterial, plant, and mammalian cells [7]. During viral infection, protein ADP ribosylation also plays key roles by attaching ADPr to essential viral replication proteins to inhibit their function [8, 9]. Additionally, ribosylation of host innate immune effectors can enhance their antiviral activity [10]. These examples underscore the evolutionary conservation and functional importance of ADP ribosylation in regulating cellular responses to diverse stressors.

In addition to DNA and protein substrates, RNA has recently been identified as a target of ADP ribosylation, though the function of RNA ribosylation in mammalian cells is unknown. In prokaryotes, RNA ribosylation serves as an antiviral defense mechanism against bacteriophage infection. For example, bacterial cells upregulate the ADP-ribosyltransferase CmdT upon bacteriophage infection, which modifies phage RNA to block its translation and thereby inhibit viral replication [11]. In mammalian systems, cellular stressors that inhibit translation such as nutrient deprivation or the accumulation of unfolded proteins also induce RNA ribosylation [12, 13]. In cell-free systems RNA modified with ADPr on the 5′ terminus in place of the 7-methylguanosine messenger RNA (mRNA) cap reduces translational efficiency and protects from degradation by the XRN-1 exoribonuclease [12]. These observations, along with its known antiviral role in bacteria, suggest that RNA ribosylation may similarly contribute to stress adaptation in mammalian cells. However, to date, no physiological role for RNA ribosylation has been reported in any mammalian system.

RNA ribosylation is catalyzed by RNA ADP-ribosyltransferases, which utilize NAD⁺ as a substrate to covalently add ADPr to specific sites on RNA molecules [12, 14]. ADPr can be attached to various locations on RNA, including 5′ and 3′ terminal phosphates [13], the exocyclic amine of guanosine bases [15], and the 2′ hydroxyl group of ribose [16]. ADP ribosyltransferases are conserved across all domains of life, supporting the notion that RNA ribosylation is a fundamental and ancient regulatory mechanism [17–19]. In cell-free systems, several mammalian ADPr ribosyltransferases, including multiple members of the poly-ADPr polymerase (PARP) family such as PARP10, 11, 12, 15, and TRPT1, can ribosylate RNA [12]. In mammalian cells these ADPr ribosyltransferases are required for RNA ribosylation in response to cellular stress. Although the physiological consequences of this modification remain unknown in mammals, its conservation and induction under stress suggest important regulatory roles, possibly analogous to its functions in prokaryotic systems.

Specialized ribosylhydrolases remove ADPr moieties from RNA by catalyzing the hydrolysis of the glycosidic or diphosphate bond. The NADAR family found in bacteria, archaea, viruses, and certain eukaryotes specifically removes ADPr from guanosine bases [16], however NADAR orthologs are absent in mammals [20]. Both ARH and macrodomain ribosylhydrolases are expressed in mammals and can remove ADPr from 2′ hydroxyl groups and terminal phosphates on RNA in cell-free systems [16, 21]. Several viruses, including coronaviruses (e.g. SARS-CoV-2) and alphaviruses [e.g. chikungunya virus (CHIKV)], encode macrodomains that efficiently remove ADPr from RNA in cell-free systems [16, 21], suggesting that viral ADPr ribosylhydrolases may suppress RNA ribosylation during infection.

Here we explore functional roles for RNA ribosylation in mammalian cells, using CHIKV infection of human fibroblasts as a model system. We demonstrate that RNA ribosylation functions as an antiviral defense mechanism by coordinately reducing viral RNA translation, enhancing RNA decay, and promoting innate immune sensing. These findings establish RNA ribosylation as a previously unrecognized regulator of viral RNA fate during infection. Our data provide new insights into how RNA ribosylation integrates with known cellular signaling networks such as the type I interferon pathway, suggesting a central role for RNA ribosylation in the response to infection. More broadly, these results define a framework for understanding how RNA ribosylation regulates RNA metabolism during viral infection and suggest that similar mechanisms may operate in other cellular stress contexts, which will require further investigation.

Materials and methods

Cell culture and virus production

Human MRC-5 hTert fibroblasts and Vero-81 cells were grown in Dulbecco’s modified Eagle medium (Gibco) supplemented with 10% fetal bovine serum (FBS, Gibco) and 1% L-glutamine (Gibco). BHK-21 cells were grown in MEM-alpha media (Gibco) supplemented with 10% FBS, 1% L-glutamine, and 10% tryptose phosphate broth. All cells were cultured 37°C in 5% CO₂. Viruses were produced by electroporating BHK-21 cells with in vitro transcribed (IVT), capped RNA from an infectious clone of CHIKV (181/25). Briefly, an infectious clone of 181/25 was linearized with NotI-HF (NEB), purified, and used as a template for in vitro transcription using mMessage Machine Kit (Invitrogen) according to the manufacturer’s directions. IVT RNA was precipitated with LiCl, and then 10 µg of RNA was electroporated into 4 × 10⁷ BHK-21 cells. Virus was purified from cell supernatants 24 h after electroporation by centrifugation through a 20% sucrose cushion. The resulting pellet was resuspended in phosphate buffered saline supplemented with 1% FBS and stored at −80°C until use.

Western blotting

Protein was extracted from cells using radioimmunoprecipitation assay (RIPA) buffer [50 mM Tris–HCl (pH 7.4), 150 mM NaCl, 1 mM ethylenediaminetetraacetic acid (EDTA), 1% NP-40, 1% sodium deoxycholate] supplemented with protease inhibitor cocktail (Roche cOmplete EDTA-free protease inhibitor cocktail) as previously described [22]. Protein concentration was determined using a Bradford assay (Bio-Rad). Then, 30 µg of protein was resolved on 10% sodium dodecyl sulphate–polyacrylamide gel electrophoresis gels and then transferred to a polyvinylidene difluoride (PVDF) membrane. Membranes were blocked with 5% nonfat dairy milk in TBST [20 mM Tris–HCl (pH 6.8), 140 mM NaCl, 0.1% Tween 20] for 1 h prior to incubation with primary antibody overnight at 4°C in tris-buffered saline with Tween 20 (TBST) supplemented with5% bovine serum albumin (BSA). Blots were washed with TBST prior to incubation with the appropriate secondary antibody conjugated to horseradish peroxidase (Advansta) in TBTS + 5% BSA and visualized by chemiluminescence. Antibodies used in this studies include: Poly/Mono ADPr (Cell Signaling Technology; #E6F6//Cell Signaling Technology; # D9P7Z), CHIKV nsP2 (Invitrogen; #HL1432), Tubulin (Cell Signaling Technology; #2144), PARP10 (Bethyl; #A300-665A), PARP11 (Proteintech; #16692-1-AP), PARP12 (Invitrogen; #PA5-89749), PARP15 (Proteintech; #18126-1-AP), CHIKV Capsid (Invitrogen; #PA5-143450), GAPDH (Cell Signaling Technology; #14C10), Anti-Rabbit horseradish peroxidase (HRP) IgG (Seracare Life Sciences #54 500 010), and Anti-Mouse HRP IgG (Seracare Life Sciences #54 500 011).

Dot blot assay

Ribosylated RNA was detected following previously described protocols [23]. In brief, RNA was extracted using RNAeasy Kits (Qiagen) according to the manufacturer’s directions. RNA was dotted onto nylon membrane (Invitrogen) and allowed to dry for 15 min prior to ultraviolet cross-linking using the auto-cross link function of the Stratalinker (Stratagene). Membranes were blocked with 5% milk in PBST prior to incubation overnight with Poly/Mono-ADPr antibody in phosphate-buffered saline with Tween 20 (PBST) supplemented with 5% BSA. Membranes were washed with PBST and probed with secondary antibody in 3% BSA prior to visualization via chemiluminescence. For experiments to evaluate ADPr antibody specificity, RNA was extracted as above from human fibroblasts at 18 h after infection with CHIKV (181/25). Then, 1 µg of RNA was treated with either 1 unit Proteinase K (NEB), 1 µg RNAse A (Invitrogen), or 2 units of DNAse (Invitrogen) for 2 h at 37°C, and then re-purified. An equal volume of each sample was dotted onto nylon membrane to detect ribosylation as above.

Infectious virus quantification

The specific infectivity of viral RNA was measured as previously described [24]. Briefly, BHK-21 cells were electroporated with IVT RNA as described above. Serial 10-fold dilutions of electroporated cells were plated onto monolayers of Vero81 cells and then overlayed with 2× α-MEM (Gibco) containing 2.5% carboxmethylcellulose. Seventy-two hours after plating, cells were fixed with 2% paraformaldehyde and stained with crystal violet to visualize plaques. Plaque assays were used to quantify levels of cell-free infectious virus, as previously described [24]. Briefly, serial 10-fold dilutions of virus stocks were plated on Vero81 cells. After adsorption for 1 h at 37°C, the inoculum was removed and replaced with 2× α-MEM (Gibco) containing 2.5% carboxmethylcellulose, incubated for 72 h and were fixed and stained as above. The presence of reversion mutations in nsP3 mutant stocks was assessed by sequencing viral RNA extracted from virus stocks using previously described methods [25].

Quantification of viral genomes in virus stocks

Viral genomes were quantified by quantitative reverse transcriptase PCR (RT-qPCR) as described previously [26]. Briefly, RNA was extracted from 100 μL of concentrated virus stocks using Trizol. Viral RNA was reverse transcribed and quantified by RT-qPCR using primers specific for nsP2 (F 5′-TGCGTGAGACACACGTAGCCT-3′, R 5′-GGGCCTTCAAAAAGGCGCTGTCA-3′). The number of viral genomes was determined by the absolute quantification method by comparing the results to a standard curve generated using a range of viral genomes from 10⁸ to 10¹ copies.

RNA immunoprecipitation

RNA was isolated from human MRC-5 fibroblasts at 18 h after infection using an RNAeasy kit (Qiagen) according to the manufacturer’s directions. Mono/Poly-ADPr antibody (Cell Signaling Technologies; #E6F6) or an IgG control was conjugated to protein A/G PLUS agarose beads (Santa Cruz). Then, 5 µg of RNA was incubated with the beads for 1 h, after which the beads were washed with a series of buffers of increasing stringency to minimize nonspecific binding, as before [27]. RNA was extracted from the beads with Trizol, and the number of viral genomes quantified by RT-qPCR using the absolute quantification method and nsP2-specific primers noted above.

RT-qPCR quantification of RNA expression

For RNA decay and interferon stimulated gene (ISG) expression experiments RNA was extracted from MRC-5 fibroblasts using the RNAeasy kit (Qiagen) as described above. 1 µg of total RNA was reverse transcribed using the High Capacity cDNA Reverse Transcription Kit (Applied Biosystems #4 368 813) according to the manufacturer’s directions. Viral RNA was quantified via absolute quantification method using nsP2-specific primers as described above. Host genes were quantified using the ΔΔCT method normalized to actin. Primers sequences are listed below: nsP2 (F 5′-TGCGTGAGACACACGTAGCCT-3′, R 5′-GGGCCTTCAAAAAGGCGCTGTCA-3′), Capsid (F 5′-GACCGATCTTCGACAACAAG-3′, R 5′-CGATGTCTTTGTTCCAGGTC-3′), ISG15 (F 5′-TGGCGGGCAACGAATT-3′,R 5′-GGGTGATCTGCGCCTTCA-3′), IFIT1 (F 5′-GCCTTGCTGAAGTGTGGAGGAA-3′,R 5′-ATCCAGGCGATAGGCAGAGATC-3′), RIG-I (F 5′-CACCTCAGTTGCTGATGAAGGC-3′,R 5′-GTCAGAAGGAAGCATTGCTACC-3′), OAS1 (F 5′-AGGTGGTAAAGGGTGGCTCC-3′, R 5′-ACAACCAGGTCAGCGTCAGAT-3′).

Rabbit reticulocyte lysate translation assay

50 ng of RNA was incubated at 65°C for 5 min prior to incubation on ice. The RNA was then added to nuclease treated rabbit reticulocyte lysates (Promega #L4960) supplemented with 50 mM KCl, 20 µM unlabelled amino acids, and 25 units RNAse inhibitor (Promega #N2111). Reactions were incubated at 30°C for 90 min and 5 µl was removed at 0, 5, 15, 30, 45, 60, and 90 min and immediately treated with Nano-Glo reagent (Promega #N1110) and luciferase activity was quantified using a Glomax plate reader (Promega).

Analysis of translation, decay, and immune response

hTert MRC-5 fibroblasts were infected with an equal number of genomes of WT or N24A CHIKV virus from control or siPARP12 knockdown cells equivalent to a multiplicity of infection (MOI) of 3 for WT in the presence of >10 µM of SR-42718, and/or RA-2034, or cycloheximide (CHX). Wells were washed three times with media following 1 h infection and translation of incoming RNA was measured by adding Nano-Glo reagent (Promega #N1110) directly to wells and quantifying luciferase activity using Glomax plate reader (Promega) at indicated timepoints. RNA decay was measured by RT-qPCR for CHIKV genomes as described above and decay rate was measured by comparing change in CHIKV genome copy number to the starting amount immediately following infection at indicated timepoints. Expression of ISGs was measured at 12 h post infection by RT-qPCR as described above to measure innate immune response.

XRN-1 degradation assay

Purified virion RNA or control mouse thymus RNA was treated with either 10U RppH (NEB, cat # M0356) or with buffer as negative control. The reaction was incubated at 37°C for 1 h and then re-purified as described above. 2U XRN-1 (NEB; #M0338) or buffer was added to each sample and incubated at 37°C for 1 h. Equal volumes of each reaction were analyzed by reverse transcriptase-polymerase chain reaction (RT-PCR) using the nsP2-specific primers described above and quantified by the absolute quantification method. The percentage of XRN-1 resistant RNA was calculated by dividing the number of viral genomes in XRN-1 treated samples by the number of viral genomes in buffer-treated samples and multiplying by 100.

PARP siRNA knockdown

MRC-5 hTert human fibroblasts were reverse transfected with SMARTPool ON-TARGETplus siRNAs (Dharmacon) specific for the indicated genes, nontargeting siRNAs, or a combination of all PARP siRNAs together as previously described [27]. Cells were infected at 48 h after transfection and culture supernatants were harvested at 24 h after infection. Total RNA was extracted using the RNeasy Kit, and RT-qPCR was performed using nsP2 specific primers as above to measure viral RNA abundance. Knockdown efficiency and impact on viral replication were measured by western blot and RT-qPCR.

Results

Virus infection increases RNA ribosylation

Recent studies have found that RNA can be modified with ADP ribose (ADPr RNA) in mammalian cells and that RNA ribosylation is induced under conditions of cell stress such as nutrient deprivation and heat shock [12]. As viral infection also activates a cell stress response, we tested if infection with CHIKV, a positive sense RNA virus in the Togaviridae family that causes widespread, explosive outbreaks and epidemics of severe acute and chronic polyarthralgia, induced RNA ribosylation in infected cells. Total RNA and protein were extracted from CHIKV infected cells or uninfected control cells and protein and RNA ADPr levels were measured by western blot and dot blot [23], respectively. Infection led to a consistent increase in ribosylated protein and RNA levels compared to controls (Fig. 1A and B). To confirm that the signal from the dot blot was specific for ribosylated RNA rather than contamination from ADPr modified DNA or protein, we treated total RNA extracted from infected cells with RNAse A, DNase, or Proteinase K, and then re-purified the RNA and again measured ribosylated RNA levels. Only samples treated with RNAse A displayed a significant decrease in the ADPr RNA signal, demonstrating that the assay specifically measures ribosylated RNA rather than DNA or protein (Fig. 1C). As a further control we also confirmed the absence of protein contamination in RNA preparations by western blot using antibodies specific to the cellular protein tubulin or the viral capsid protein (Fig. 1D). Together these data show that like other cell stressors, viral infection induces RNA ribosylation in mammalian cells.

Figure 1.

CHIKV infection increases RNA ribosylation. (A) Human fibroblasts were infected with CHIKV strain181/25 at an MOI = 3 and harvested at the indicated times after infection. Protein ribosylation was determined by western blot of infected cell extracts. (B) Total RNA was extracted from cells infected as above and harvested at 18 h after infection. The levels of ribosylated RNA were determined by dot blot probed with ADPr specific antibody and total RNA loaded was measured with methylene blue staining. (C) RNA from panel (B) was treated with dimethyl sulfoxide (DMSO), 1 µg RNAse A, 2 units DNAse, or 1 unit proteinase K for 2 h. ADP-ribose levels were measured by dot blot as above and quantified relative to DMSO treated samples with equivalent loading confirmed by methylene blue staining. (D) Equal numbers of mock or CHIKV-infected MRC-5 fibroblasts were lysed in RIPA buffer or extracted with an RNAeasy kit. Equal volumes of each extraction were analyzed by western blot using antibodies specific for the cellular tubulin protein or the CHIKV capsid protein. (N = 3; ***P <.001).

The CHIKV nsP3 macrodomain reverses RNA ribosylation in mammalian cells

The CHIKV nsP3 protein contains a macrodomain, an evolutionarily conserved ADPr ribosylhydrolase domain essential for efficient viral replication, that removes inactivating ADPr modifications from viral proteins [28–30]. Together with the increase in ADPr RNA observed after infection, the presence of a viral macrodomain suggests that the CHIKV nsP3 macrodomain may limit ribosylated RNA accumulation during infection. To determine if analogous mutations in CHIKV similarly impacted virus replication, we measured the replication of a set of mutations to the CHIKV nsP3 macrodomain that we predicted to decrease nsP3 ribosylhydrolase activity based on previous studies [28, 29, 31], their conservation across the alphavirus family, and their location relative to the nsP3 macrodomain active site [32, 33]. As mutations that fully ablate nsP3 macrodomain activity are lethal for virus replication, we expected that decreased nsP3 activity would decrease viral specific infectivity and replication. Of the mutants tested both the G32E and N24A mutation displayed a significant decrease in specific infectivity and replication. However the G32E mutation was unstable, and rapidly reverted to the wild-type (WT) sequence as determined by sequencing the viral genome after passage, consistent with previous results (Supplemental Fig. 1A–E) [29]. Based on the impact of N24A on viral infectivity and replication and its ability to be stably maintained across the viral replication cycle, we used the N24A mutant in all subsequent studies. We first measured the impact of N24A on replication using a specific infectivity assay, which quantifies the number of productively infected cells after electroporation and provides a measure of whether a given mutation adversely affects the ability of the viral RNA to initiate infection (Fig. 2A and Supplemental Fig. 1C) and the yield of infectious virus 24 h following electroporation (Fig. 2B and Supplemental Fig. 1D). As with previous studies utilizing Sindbis virus (SINV) [29], the N24A mutation significantly decreased RNA specific infectivity and cell-free virus yield at 37°C. However, this defect could be rescued when transfected or infected cells were incubated at 30°C, demonstrating that the defect was temperature dependent (Fig. 2A and B).

Figure 2.

CHIKV nsP3 macrodomain is required for efficient virus replication. (A) BHK21 cells were electroporated with WT or N24A RNA and the specific infectivity of WT and N24A was determined by plating electroporated BHK21 cells on a monolayer of Vero81 cells. (B) Supernatants were collected 24 h after electroporation and titered on Vero81 cells at 37°C. (C) The ratio of viral genomes to plaque forming units (PFU) was determined by measuring genome copies from viral supernatants by RT-qPCR and comparing it to viral titers. (D) Human fibroblasts were infected with an equivalent number of viral genomes of WT or N24A virus, which was equal to a multiplicity of infection of 3 PFU/cell for WT virus. Viral RNA levels in infected cells was quantified by RT-qPCR at 1 h after infection. (E) Human fibroblasts were infected at 37°C with the equivalent number of genomes of WT and N24A as in panel (D), and amount of infectious virus at the indicated times after infection was measured by plaque assay. (F) The levels of viral protein expression (nsP2) and overall levels of protein ribosylation (Poly/Mono ADPr) were measured by western blot in lysates from fibroblasts infected with equal genomes of WT or N24A, as described above. (G) Cells were infected as in panel (F), and levels of RNA ribosylation in total cellular RNA was measured via dot blot. Loading consistency was confirmed by methylene blue staining, and relative ADPr levels were quantified by densitometry (N = 3; *P <.05; ***P <.001).

To determine if the different replication rates of N24A as compared to WT virus were caused by an accumulation of defective genomes, we assessed the ratio of viral RNA to the number of PFUs in N24A or WT virus stocks as measured by plaque assay. We observed a significant disconnect between the number of viral genomes in virus stocks when compared to the number of infectious particles. The N24A and WT virus stocks contained a similar number of viral genomes, but the N24A virus stocks had a significant defect in the number of infectious particles leading to a significantly higher genome to PFU ratio indicating the presence of noninfectious genomes (Fig. 2C). While this difference could be due to a defect in viral binding and entry, we found that when cells were infected with an equal number of N24A and WT viral genomes, equivalent numbers of viral genomes were found inside the cell at 1 h after infection, demonstrating equivalent binding and entry for both viruses (Fig. 2D). As a further measure of virus replication, we infected cells with an equal number of N24A and WT viral genomes and analyzed viral replication kinetics by measuring the accumulation of infectious virus across a time course of the viral replicative cycle. When MRC-5 fibroblasts were infected with an equal number of genomes, the CHIKV N24A mutant replicated less efficiently than WT virus (Fig. 2E). The lower number of PFUs observed at the time of infection reflects the defect in genome to PFUs, consistent with the results of Fig. 2C. A similar reduction in N24A replication was observed following infection of fibroblasts with an equal number of PFUs at both high and low multiplicity of infection (Supplementary Fig. 2A–C). We conclude that, similar to SINV, nsP3 macrodomain activity is required for efficient CHIKV replication, and attenuating CHIKV nsP3 macrodomain activity results in inhibition of virus replication at a step after virus binding and entry.

The SINV nsP3 macrodomain removes inactivating ADPr from viral and cellular proteins to enhance virus replication, and mutating nsP3 at N24A significantly reduces its ribosylhydrolase activity [28]. To confirm that the N24A mutation in CHIKV similarly decreased nsP3 macrodomain activity, we measured the accumulation of ribosylated proteins in human fibroblasts infected with WT CHIKV or the N24A CHIKV mutant. As expected, increased levels of protein ribosylation were observed in N24A infected cells as compared to cells infected with WT virus, confirming that the N24A mutation decreased ribosylhydrolase activity of the CHIKV nsP3 macrodomain (Fig. 2F).

Previous studies found that the purified CHIKV nsP3 macrodomain can remove ADPr from model RNA templates in cell-free systems [12]. To determine if CHIKV nsP3 macrodomain also impacted RNA ribosylation in infected mammalian cells, human fibroblasts were infected with WT CHIKV or the N24A mutant, and the ribosylation of total cellular RNA was measured by dot blot. Human fibroblasts were used in these studies as they are a physiologically relevant model for alphavirus infection and have fully intact innate immune signaling. Similar to protein ribosylation, the level of ADPr RNA was significantly increased in N24A infected cells as compared to WT CHIKV infection in all tested infection conditions (Fig. 2G and Supplementary Fig. 2D–F), with the largest difference occurring at later times after infection when viral RNA is the predominant newly synthesized RNA species in the cell [34]. Similar results were observed in two additional infected cell lines, Vero and BHK-21 cells (Supplementary Fig. 3). These results show that CHIKV nsP3 macrodomain activity limits the accumulation of ribosylated RNA during viral infection of mammalian cells.

Viral RNA is ribosylated in mammalian cells and in virions

To directly test whether viral RNA is ribosylated, we purified WT or N24A virions by ultracentrifugation through a sucrose density cushion, extracted RNA from the purified virus stocks, and measured ribosylated RNA levels by dot blot. N24A virion RNA was significantly more ribosylated than WT (Fig. 3A and B), showing that ribosylated RNA can be packaged into CHIKV virions.

Figure 3.

CHIKV genomes are ribosylated. (A) Virion RNA was extracted from purified CHIKV virus stocks, and levels of ribosylation were measured by dot blot. Equivalent loading was assessed by dot blot using a probe specific for the CHIKV 3′UTR. A representative is shown. (B) The results from three independent experiments described in A were quantified via densitometry. (C) Purified virion RNA from WT or N24A virus (N24A) was incubated with a control antibody (IgG) or an antibody specific to poly/mono ADP ribose (ADPr). Antibody:RNA complexes were captured by immune precipitation, and the amount of co-purifying viral RNA was quantified by RT-qPCR. The fold enrichment of viral RNA in ADPr-specific immune complexes versus control immune complexes is shown (N = 3; *P <.05).

To confirm the above result, we incubated purified viral RNA from WT and N24A virions with an antibody specific for ADPr or an IgG control and quantified the number of viral genomes co-purifying with each antibody by RT-qPCR. Both WT and N24A genomes were recovered at levels above the control samples, indicating that both are ribosylated to some level. However, viral genomes were significantly enriched in N24A samples incubated with the ADPr-specific antibody as compared to WT virus, demonstrating that CHIKV RNA is ribosylated during infection and that the CHIKV macrodomain suppresses viral RNA ribosylation during infection (Fig. 3C).

PARP12 ribosylates viral RNA

To identify the cellular enzyme(s) that ribosylate viral RNA, we used siRNA to deplete select PARPs previously shown to regulate RNA ribosylation in mammalian cells [12], and then measured the effect on the ribosylation of purified virion RNA and viral protein expression. Reducing the expression of PARPs 10, 11, 15, and TRPT1 had no effect on the expression of viral nonstructural (nsP2) or structural (capsid) protein expression (Fig. 4A). In contrast, depletion of PARP12 increased nsP2 and capsid RNA and protein levels compared to a nontargeting siRNA control (Fig. 4B and C, respectively). Analysis of virion RNA from virus stocks grown on cells depleted of each PARP alone or all in combination revealed a striking loss of virion RNA ribosylation after PARP12 depletion or the combined depletion of all PARPs together (Fig. 4D and E). These data support the conclusion that PARP12 is necessary for efficient ribosylation of viral RNA in mammalian cells. These findings also suggest that RNA ribosylation primarily regulates translation efficiency per RNA molecule rather than total RNA abundance. While PARP12 depletion increases viral RNA levels, the corresponding increase in protein expression is more modest, consistent with ribosylation directly limiting the translational competence of viral RNA.

Figure 4.

PARP12 ribosylates viral RNA in mammalian cells. (A) Human fibroblasts were transfected with a nontargeting siRNA (Ctrl), siRNA specific to the indicated PARPs, or a pooled library of all siRNAs (Mixed). Cells were infected with N24A virus at an MOI of 3 and collected at 24 h post infection. The expression of the indicated proteins was measured by western blot. (B) The abundance of viral RNA in siRNA treated cells was measured by RT-qPCR using primers specific for the CHIKV nsP2 nonstructural protein or (C) the capsid structural gene. (D) The ribosylation levels of virion RNA from siRNA treated cells was measured by dot blot and (E) quantified by densitometry (N = 3, ^****P <.0001).

Ribosylation decreases the translation efficiency of viral RNA

In cell-free systems, RNA can be ribosylated on its 5′ end in place of the 7-methylguanosine mRNA cap, inhibiting RNA translation [12, 35]. To measure the effect of viral RNA ribosylation on viral protein synthesis, we isolated RNA from WT or N24A virions and measured the ability of each to be translated in rabbit reticulocyte lysates. In this experiment we used a CHIKV reporter virus encoding a nanoluciferase (nLuc) reporter gene as part of the nonstructural open reading frame [36]. nLuc activity in this assay thus serves as a direct measure of viral RNA translation. As controls we also included IVT, capped full length WT and N24A genomes. WT virion RNA and IVT WT and N24A full length genomes, which are not ribosylated (not shown), were translated at similar rates, demonstrating that the presence of the N24A mutation alone did not affect protein synthesis. However the translation of N24A virion RNA was significantly reduced (Fig. 5A), demonstrating that the defect in the translation of N24A virion RNA was due to the presence of ADPr rather than the result of a nonspecific defect. To determine if ribosylation also inhibited CHIKV RNA translation in the context of infection, human fibroblasts were infected with N24A or WT CHIKV expressing nLuc in the presence of SR-42718 [37], a potent and specific small molecule inhibitor of the CHIKV RNA-dependent RNA polymerase (RdRP) or CHX, an inhibitor of global translation, and nLuc expression was measured over the first 5 h after infection. In the presence of SR-42718, nLuc expression is derived solely from incoming viral genomes as de novo viral RNA synthesis is inhibited. SR-42718 decreased nLuc expression after WT infection, consistent with inhibition of the RdRP. In control cells, nLuc expression was significantly lower in N24A infected fibroblasts as compared to cells infected with WT virus, and the same was true in the presence of SR-42718 (Fig. 5B), indicating that the input N24A genomic RNA is translated less efficiently than WT genomic RNA.

Figure 5.

Ribosylation decreases the translation efficiency of CHIKV RNA. (A) Capped, IVT genomic CHIKV RNA or purified virion RNA (Virion) from WT or N24A mutants (N24A) virus stocks expressing a nanoluciferase reporter gene were incubated with nuclease free rabbit reticulocyte lysates for the indicated times, at which nanoluciferase activity was quantified to measure the translation of each viral genome (n = 3). (B) Human fibroblasts were infected with an equivalent number of viral genomes of WT or N24A nanoluciferase reporter viruses as in panel (A) and treated with CHX (100 µg/ml) or SRI-42718 (50 µM) where indicated (n = 3). (C) RNA extracted from purified WT and N24A virions expressing a nanoluciferase reporter grown in MRC-5 fibroblasts treated with a nonspecific siRNA (siSCR) or siRNA targeting PARP12 (siPARP12). Purified virion RNAs were incubated with nuclease free rabbit reticulocyte lysates for the indicated times at which nanoluciferase activity was quantified to measure translation of the viral genomes (n = 3). (D) Purified virion RNA from WT and N24A virus stocks was treated with buffer or RppH (10U) for 1 h. Ribosylated RNA levels were measured by dot blot. (E) Mouse thymus RNA (Ms thymus) or WT or N24A virion RNA was treated with buffer or RppH (10U) for 1 h followed by 2U of XRN1 for 1 h. Viral genome abundance was quantified by RT-qPCR. The graph shows the percent of input RNA remaining after XRN1 digestion. The results of a representative experiment are shown in panels (C) and (D) (*P <.05, **P <.01, ***P <.001).

To directly test the impact of PARP12-dependent ribosylation on the functional competence of virion RNA, we generated WT and N24A viruses expressing a nLuc reporter in control or PARP12 knockdown cells. Virion RNA was extracted from each virus stock, and the translation efficiency was measured in cell-free reticulocytes lysates. Virion RNA derived from PARP12 knockdown cells, which exhibits reduced ribosylation, was translated more efficiently than RNA derived from control cells for both WT and N24A virus (Fig. 5C). These results demonstrate that PARP12-dependent ribosylation reduces the translational capacity of incoming viral genomes. Because translation efficiency is a key determinant of specific infectivity, these findings indicate that RNA ribosylation reduces the functional infectivity of viral particles.

Previous cell-free experiments found that RNA can be ribosylated on a free 5′ monophosphate in place of the 5′ 7-methylguanosine cap, and 5′ end ribosylation significantly decreases translation and protects RNA from degradation by 5′ to 3′ exoribonucleases such as XRN1 [12]. To determine if viral RNA was ribosylated on the 5′ end, we treated virion RNA with the decapping enzyme RppH followed by digestion with XRN1 and measured the amount of viral RNA remaining after each step by RT-qPCR. Viral RNA was ribosylated at similar levels in RppH and buffer treatment, demonstrating that RppH does not have ribosylhydrolase activity (Fig. 5D). As expected, the control, capped RNA was minimally susceptible to degradation by XRN1 in the absence of prior RppH treatment. However the vast majority of the control RNA was degraded by XRN1 after prior incubation with RppH, indicating the control RNA has a 5′ m7G cap and removing the cap with RppH treatment makes the RNA susceptible to degradation by XRN1. We found that ∼80%–90% of WT virion RNA was degraded by XRN1 in the absence of RppH pre-treatment, consistent with prior studies showing that the majority of WT virion RNA is uncapped [38]. Prior incubation of WT virion RNA with RppH led to a further reduction in the levels of WT RNA remaining, consistent with the XRN1 resistant RNA having a 5′ m⁷G cap. In contrast, significantly more N24A virion RNA was resistant to XRN1 degradation, whether or not the RNA was previously incubated with RppH (Fig. 5E). These data show N24A viral RNA is protected from XRN1 degradation in a cap-independent manner and suggest that ∼5% of WT and 40% of N24A virion RNA is ribosylated. The differential sensitivity of WT and N24A RNA to XRN1 digestion alongside the increase in viral RNA ribosylation with the N24A mutant and corresponding decrease in translation N24A translation efficiency support the conclusion that viral RNA is ribosylated on its 5′ end. However these results cannot rule out the presence of other modifications that might protect from degradation or ADPr in segments of viral RNA other than the 5′ end. Together these data show that ribosylation inhibits the translation of viral RNA in cell-free systems and in infected human fibroblasts and that the defect in translation is consistent with ADP-ribosylation at the 5′ end of the RNA, although additional modification sites cannot be excluded.

RNA ribosylation decreases RNA stability during infection

As RNA modifications can impact RNA stability, we next determined if RNA ribosylation impacts RNA degradation in mammalian cells by measuring the decay of CHIKV RNA after infection with WT virus or the N24A mutant. Virus was allowed to adsorb and enter infected cells for 1 h, and the protein synthesis inhibitor CHX or the viral RdRP inhibitor SR-42718 was then added to the cultures to prevent de novo RNA synthesis. The abundance of viral RNA as a function of time was then measured by RT-qPCR. Consistent with previous studies [38], there was an initial rapid decay of WT viral RNA over the first hour after infection, which continued, albeit more slowly, as time progressed (Fig. 6A), which was unaffected by treatment with CHX, a global inhibitor of protein synthesis. Viral RNA levels decreased significantly more rapidly after infection with the N24A virus as compared to WT in SR-42718 treated cells (Fig. 6A and B). These data suggest that ribosylation enhances viral RNA decay independent of its effects on translation or viral genome replication.

Figure 6.

Viral RNA ribosylation enhances RNA decay. (A, B) Human fibroblasts were infected with an equal number of genomes of WT or the N24A mutant virus. Cells were treated with CHX (100 µg/ml) or SR-42718 (50 µM) to prevent de novo viral RNA synthesis. Viral genome abundance was quantified by RT-PCR at the indicated times after infection. The amount of input RNA remaining at each as a percent of viral genome abundance at 1 h after infection is shown (N = 3, **P <.01).

RNA ribosylation increases antiviral gene expression in response to infection

Beyond direct effects on RNA stability and translation efficiency, RNA modifications can also be recognized by cells as danger-associated molecular patterns or pathogen-associated molecular patterns (PAMPs), which trigger the expression of genes that function to resolve cell stress. In viral infections, PAMP recognition leads to increased expression of antiviral effector genes such as ISGs that control pathogen replication. To determine if ribosylated RNA acts as a PAMP, we measured the expression of exemplar ISGs after infection of human fibroblasts with N24A or WT CHIKV. Two CHIKV replication inhibitors, SR-42718 [37] or the alphavirus nsP2 protease inhibitor RA-2034 [39], were included in some samples to allow us to specifically measure changes in ISG induction triggered by incoming, rather than newly replicated, viral genomes. N24A viral genomes accumulated to lower levels than WT viral genomes in the absence of the inhibitors, consistent with the defect in virus replication for the N24A virus (Fig. 2). Both replication inhibitors prevented new viral genome accumulation in cells infected with either virus, as expected (Fig. 7A). Significantly higher ISG levels were observed in cells infected with N24A virus as compared to WT virus, and similar results were observed in the presence of the CHIKV replication inhibitors (Fig. 7B–E). These results indicate that CHIKV macrodomain activity is required to suppress induction of the antiviral response and suggest that ribosylated viral RNA is a novel PAMP sensed by infected cells to initiate an antiviral response.

Figure 7.

Ribosylated viral RNA triggers an enhanced cellular antiviral response. (A) Human fibroblasts were infected with an equal number of WT or N24A mutant (N24A) viral genomes as in Fig. 2. Infected cells were treated with a viral replicase inhibitor (SR-42718; 50 µM) in combination with an inhibitor of the viral protease (RA-2034; 50 µM), or vehicle alone (DMSO). Cells were collected at 12 h after infection, and the amount of intracellular viral RNA was quantified by RT-qPCR using the absolute quantification method. (B–E) The fold increase in the expression of the indicated genes was measured relative to the mock sample using the ΔΔCt method normalized to actin expression (N = 3, *P <.05, **P <.01, ***P <.001, ^****P <.0001).

Discussion

Previous studies show that cell stress induces RNA ribosylation in mammalian cells, and cell-free studies have identified potential impacts of ribosylation in regulating RNA metabolism and translation. Here we demonstrate, to our knowledge, the first functional roles for RNA ribosylation in mammalian cells. We find that RNA ribosylation serves as a novel cellular defense during viral infection by inhibiting viral RNA translation, inducing viral RNA decay, and stimulating the innate antiviral response to infection. Taken together, our data support a temporal model in which ribosylated viral RNA undergoes accelerated decay early after entry, exhibits reduced translation efficiency during initial rounds of protein synthesis, and subsequently triggers enhanced innate immune signaling. These coordinated effects impair viral replication at multiple stages of the early infection cycle. Our data provide new insights into how RNA ribosylation integrates into known cellular antiviral pathways suggesting a central role in the response to infection with multiple viruses. More broadly, our results define a framework for understanding how RNA ribosylation regulates RNA metabolism during viral infection and suggest that similar mechanisms may operate under other cellular stress conditions, which will require further investigation.

One key finding from our studies is that ribosylated viral RNA is poorly translated, both in cell-free systems and in human fibroblasts. Importantly, independently manipulating RNA ribosylation levels by altering nsP3 macrodomain activity and PARP12 depletion produces reciprocal effects on translation, suggesting antagonistic roles for nsP3 and PARP12 in regulating the functionality of viral RNA during infection. Our data suggest that ADPr is added to the 5′ terminal end of viral genomes in place of the 5′ mRNA cap. We found that ∼80%–90% of WT virion RNA is uncapped and instead has a 5′ monophosphate, making it sensitive to XRN1 degradation. In contrast, N24A virion RNA contains similar levels of uncapped RNA, but a significant fraction of this uncapped ribosylated RNA is XRN1 resistant, which is consistent with results from cell-free systems where the predominant ADPr modification of RNA by human PARPs occurs on 5′ terminal monophosphates of uncapped RNAs. Together with the known requirement for the 5′ mRNA cap in the translation of incoming CHIKV genomes [40], our data suggest a model in which the presence of ADPr in place of the 5′ mRNA cap reduces the recruitment of translation initiation factors needed for viral protein synthesis. We hypothesize that ribosylated viral RNA fails to recruit the cellular cap binding protein eIF4E, which is needed for the formation of translation initiation complexes that recruit ribosomes to viral RNAs. In the absence of translation initiation complex recruitment, ribosylated viral RNAs are poorly translated, limiting the expression of viral replicase proteins and decreasing virus replication. However, further studies to define the impact of viral RNA ribosylation on translation factor recruitment will be needed to more directly test this hypothesis. While our data support the conclusion that ADPr modification occurs on the 5′ end of the viral genomes, we cannot rule out a potential role for ADPr modification at additional nucleotides throughout the viral genome that could potentially limit additional steps in translation such as ribosome processivity. Biophysical studies to comprehensively identify sites of ribosylation across the viral genomes and their role in viral RNA translation will be needed to determine if RNA ribosylation may play additional roles in inhibiting protein synthesis.

Our data show that PARP12 plays a critical role in ribosylating viral RNA, which is reversed by the nsP3 macrodomain. Based on the accumulation of ribosylated RNA as infection progresses, we hypothesize that PARP12 adds ADP ribose to nascent viral RNA chains. Our results are consistent with a model where PARP12 adds ADP ribose to uncapped viral RNAs that normally accumulate during replication and are packaged into nascent virions, where they play a role in enhancing viral infectivity upon infection of new cells [38]. In this model, PARP12-mediated ribosylation not only decreases the translation efficiency and stability of the viral genomic RNA but also converts uncapped genomes from proviral factors to antiviral PAMPs that inhibit virus replication in newly infected cells by stimulating the expression of antiviral effectors.

Our data show that RNA ribosylation decreases RNA stability in mammalian cells, as ribosylated N24A virion RNA is more rapidly degraded than WT viral RNA. While decreased translation efficiency often correlates with increased RNA decay, our results suggest the two phenotypes are separable, as the decay of the more ribosylated N24A was accelerated compared to nonribosylated WT virion RNA in the presence of the translation inhibitor CHX. The mechanism underlying the accelerated decay of ribosylated N24A is currently unclear. While speculative, one potential explanation is that ribosylated RNA is specifically recognized by a cellular factor that recruits the RNA degradation machinery. Such a protein would presumably have both RNA and ADPr binding activities and interact with cellular nucleases that degrade RNA. Based on our results, future efforts to define proteins that specifically interact with ribosylated RNA and regulate RNA decay in the context of infection will likely reveal key factors in the host antiviral response.

Our data also indicate that ribosylated viral RNA acts as a PAMP, as ISG induction is significantly enhanced after infection with the N24A macrodomain mutant as compared to WT CHIKV. As with RNA decay, the enhanced ISG response was independent of de novo viral RNA synthesis, suggesting that the incoming ribosylated RNA is the trigger for enhanced ISG expression. As above, while the identity of a ribosylated RNA sensor is unknown, we speculate that this sensor would specifically recognize and bind ribosylated RNA. When bound to ligand, PAMP sensors most often activate a common set of signaling pathways (e.g. TBK1, IKKβ kinases, etc.), which in turn activate transcription factors that stimulate ISG expression. As the ISGs preferentially induced by ribosylated N24A virion RNA are also induced by other PAMP sensors [41, 42], we hypothesize that known RNA sensors such as RIG-I or MDA-5 may activate similar cell signaling pathways, though further studies will be needed to test this hypothesis. As before, a complete understanding of how ribosylated RNA stimulates an enhanced ISG response awaits the discovery of cellular factors that recognize ribosylated RNA.

Beyond alphavirus infection, we expect our results to have implications for additional human disease states. Multiple human viral pathogens encode functional macrodomains, including multiple coronaviruses, togaviruses [43], hepatitis E virus [44], and rubella [45]. Our results suggest RNA ribosylation could play similar roles in limiting the replication of these viruses as well. Interestingly, the SARS-CoV-2 nsP3 protein contains a macrodomain and RNA binding domain, similar to the alphavirus nsP3 protein. In cell-free systems the SARS-CoV nsP3 macrodomain also hydrolyze ADPr from RNA [13] and SARS-CoV-2 nsP3 mutants induce elevated antiviral gene expression [13]. Perhaps RNA ribosylation generally triggers an antiviral response that limits virus replication in mammalian cells, similar to prokaryotes where RNA ribosylation plays a key role defending against bacteriophage infection [11, 46].

Outside the context of viral infection, RNA ribosylation is induced in response to a variety of cell stressors in mammalian cells, suggesting it may play similar roles in other disease states. ADP ribosylation is implicated in a wide range of human diseases including cardiovascular, inflammatory, and autoimmune disease, neurodegenerative disorders, and cancer. While to date the ribosylation of proteins and DNA has been viewed as the driving event in these disease states, our results suggest dysregulated RNA ribosylation may also contribute to disease, potentially providing additional avenues for therapeutic intervention.

Supplementary data

Supplementary data is available at NAR online.

Conflict of interest

None.

Funding

This work was supported by the READDI-AC AViDD grant from the National Institute of Health U19 AI17129201, R21 AI196787, and an RNA Discover Center (RDC) Collaborative Team Science Fellowship to N.J.M. and M.T.H. J.D.S. and R.D. received support from T32 AI007419. Funding to pay the Open Access publication charges for this article was provided by Lineberger Comprehensive Cancer Center, University of North Carolina at Chapel Hill, RNA Discover Center (RDC) Collaborative Team Science Award.

Data availability

The data underlying this article are available in the article and in its online supplementary material.