8-oxoguanine DNA glycosylase1 recognizes oxidatively-generated epitranscriptomic marks on nascent mRNAs to promote RSV replication


 Respiratory syncytial virus (RSV) infection induces an oxidizing environment linked to increased viral load, expression of pro-inflammatory genes, and excessive lung inflammation. The mechanisms of how reactive oxygen species (ROS) promotes viral gene expression have remained largely elusive. Here we show that nascent (n)RNAs of RSV acquire 8-oxo-7,8-dihydroguanine (8-oxo(r)Gua) -a covalently modified guanine base in their 5’-UTR peritranscriptionally, while paired with the 3’-terminus of viral gene(s). 8-oxo(r)Gua is bound by 8-oxoguanine DNA glycosylase1 (OGG1), a complex that physically interacts with and recruits the anti-terminator protein M2-1 to increase viral gene transcription. Knockdown of OGG1 (but not other DNA glycosylases) or inhibition of its binding, significantly decreased RSV mRNA, protein levels and yield of progeny in cultured cells and airways. Collectively, these data suggest that Gua oxidation in vRNA, serves as an epitranscriptomic mark that repurposes OGG1 to increase lytic viral replication. Pharmacological inhibition of OGG1 binding to the epitranscriptomic mark could have clinical utility to decrease manifestations of RSV infection.


Introduction 70
Respiratory syncytial virus (RSV) is one of the most frequent causes of severe lower 71 respiratory tract infections 1 . Worldwide, RSV infection is responsible for upper/lower clinical 72 manifestations in nearly all children under age two, resulting in over 3 million hospitalizations 73 and nearly 200,000 in-hospital deaths 2 . In the U.S., more than 100,000 children are hospitalized 74 yearly due to severe consequences of RSV infection, making it the most frequent disease 75 among hospitalized infants 3 . Re-infections with RSV may occur throughout life and are mostly 76 limited to the upper airways. However, in immunocompromised adults and the elderly, RSV 77 infection spreads into the lower airways, producing severe bronchiolitis and pneumonia, 78 respiratory morbidity and mortality 4,5 . RSV infections produce clinical impact similar to that seen 79 with influenza, rhino-, and coronaviruses 2,6 . No effective vaccine or drug is currently available, 80 and patient's treatment is limited to supportive medical care 7 . 81 RSV infects airway epithelial cells by binding cellular receptors, producing envelope 82 fusion with the plasma membrane, delivering nucleocapsids into the cytoplasm 8,9 . Once 83 internalized, the negative-sense viral genome encodes for ten mRNAs regulated by 3' and 5' 84 genome, guiding mRNA synthesis 13,14 , which is modulated by the anti-terminator matrix protein 90 2 (M2-1) also called the transcription elongation factor 15 . The primary cellular sites of RSV 91 transcription and replication are cytoplasmic inclusion bodies (IBs) 16 . 92 Efficacy of vRNA transcription and replication are also dependent on cellular factors 17 . 93 For example, host cytoskeletal, membrane, heat-shock proteins or those involved in protein 94 trafficking have all been associated with RSV replication 18-20 . Additionally, cellular enzymes, 95 including methyl-and acetyl-transferases, deposit/induce covalent base modifications to vRNAs 96 (epitranscriptomic marks; e.g., N6-methyladenosine (m6A), 5-methylcytidine (m5C), or N4-97 acetylcytidine) [21][22][23][24] . Other types of covalent modifications include 5-ribosyl uracil (pseudo-98 uridine) and inosine also generated enzymatically; however, there are no "reader" or "eraser" 99 enzymes identified to date 24 . Epitranscriptomic marks facilitate various steps in viral replication, 100 viral mRNA stability, translation, and/or shield viral RNAs from recognition by host RNA-specific 101 innate receptors 21,22,25 . Similar epitranscriptomic modifications control viral latency and lytic 102 replication of DNA viruses 26 . 103 RNAs are 10-25-times more susceptible to oxidation than DNA in vivo due to their 104 single-stranded nature, their lack of or partial association with protecting proteins, and their 105 close proximity to sites of ROS generators [27][28][29] . In an oxidizing cellular milieu, ROS primarily 106 attack guanine (Gua) heterocycles due to their having the lowest oxidation potential among 107 nucleic acid bases 30 . Gua oxidation results in 8-oxo-7,8-dihydroriboguanine (8-oxo(r)Gua) or 8-108 oxo-7,8-dihydrodesoxyguanine (8-oxo(d)Gua) in RNA and DNA respectively 31 . The mutagenic 109 8-oxo(d)Gua lesions are removed from DNA by 8-oxoguanine DNA glycosylase1 32 , however, in 110 RNA it may affect codon-anticodon pairing during translation, or when in excess, induce 111 apoptosis through cellular RNA surveillance mechanisms 28,33 . In vRNA, oxidation at Gua is 112 unexplored; however, the addition of antioxidants or inhibition of ROS production significantly 113 lowers RSV as well as human metapneumo-virus (HMPV) replication 34-37 , raising the potential 114 for 8-oxo(r)Gua playing roles in vRNAs. 115 The aim of the present study was to examine the role(s) of oxidative base modification(s) 116 particularly to Gua in the RSV transcriptome and understand its consequences. Our results 117 show that RSV infection-induced ROS co-transcriptionally generated 8-oxo(r)Gua in the 5'-118 UTRs of nascent vRNAs while paired with the 3'-end of genes. 8-oxoguanine DNA 119 glycosylase1 (OGG1), an oxidized Gua base specific DNA repair protein 32,38 , physically 120 interacted with 8-oxo(r)Gua in vRNAs and recruited the RSV-encoded transcription elongation 121 factor M2-1 to promote mRNA synthesis. Knockdown of OGG1 (but not other repair proteins) or 122 inhibition of OGG1 binding to the oxidatively modified Gua(s) decreased viral mRNA, protein 123 levels as well as yield of RSV progeny (100 to 1000-fold) in both airways and cultured cells. 124 These data imply that RSV has adopted cellular biosynthetic machinery via Gua oxidation in 125 vRNA as an epitranscriptomic marker to recruit the cellular reader, OGG1 for efficient virus 126

Oxidatively modified guanine base lesions in viral mRNAs 132
Given the profound effect of oxidative stress on RSV replication, pathogenesis, as well 133 as vulnerability of Gua(s) to ROS 28,36 , we examined oxidative modification to vRNA by 134 assessing Gua base modification(s). The immortalized human small airway epithelial cell 135 cultures (hSAEC -a cell type target of RSV infection involved in disease pathogenesis 39,40 ) were 136 infected with RSV and harvested to isolate RNAs using buffers containing the iron chelator 137 desferioxamine (DFO) to prevent RNA oxidation 41 . Purified RNAs were incubated with 8-138 oxo(r)Gua-specific antibody (Ab) 41 , cross-linked and samples were immunoprecipitated (IP-ed, 139 RNA-IP). In parallel, (FLAG)OGG1-expressing hSAECs were infected as above and vRNA 140 associated with proteins was cross-linked and chromatin-IP-ed (CLIP-ed) with Ab to the FLAG 141 epitope tag. RNA-protein crosslinking was performed using formalin, to avoid ultraviolet (UV) 142 irradiation-induced generation of ROS and oxidative modifications to the RNA 42 . In controls, 143 transcriptional processivity and anti-termination factor (M2-1) and nucleoprotein (N) proteins 144 associated vRNAs were CLIP-ed using the corresponding Abs. IP-ed and CLIP-ed RNAs were 145 isolated and converted into cDNA using oligo-dT primers to assess abundance of transcripts for 146 nonstructural protein-1(NS1), N, attachment glycoprotein G (G), fusion protein (F) and L using 147 quantitative real time (qRT)-PCR (primer sequences: Table 1; primer validation and amount of 148 RNA used in qRT-PCR were determined in preliminary studies (Supplementary Fig. 1a,b). 149 Results show that Abs to 8-oxo(r)Gua or FLAG(OGG1) enriched all RSV mRNAs tested (Fig.  150   1a). The level of IP-ed mRNA was the highest for NS1 (encoded by the 3'-end proximal gene) 151 ( Fig. 1a, most left panel), and mRNA levels from subsequent genes were decreased, -the most 152 promoter-distal L gene product was the least (Fig. 1a, most right panel). These data are in line 153 with those showing that RdRp re-engages with less efficiency the more distant genes from the 154 3'-end of the genome 14,43 . In controls, Ab to M2-1 IP-ed viral mRNAs, similar to that of CLIP-ed, 155 IP-ed by Ab to FLAG(OGG1) and Ab to 8-oxoGua, while levels of N protein-Ab IP-ed viral 156 mRNAs were similar to that of IgG (Fig. 1a). These results are in agreement with M2-1' RNA 157 binding core domain and its roles in de novo mRNA synthesis 44 and those showing that the N 158 protein binds genome and anti-genome as they have encapsidation signals, which are absent in 159 mRNAs 10 . There was no interaction observed between viral mRNA the repair protein MTH1 (a 160 human homolog of E. coli Mutator 1), an enzyme that specifically recognizes 8-oxo(r)Gua 161 generated in the cytoplasmic ribonucleotide pool 45 . 162 To obtain information on the cellular site of vRNA oxidation, cells were infected as 163 above, and dual-color microscopic imaging was performed 24 h post-infection. IBs were 164 visualized using Ab to N protein (green, Alexa Fluor 488) and then co-stained with Ab to 8-165 oxo(r)Gua (red, Alexa Fluor 594). These data show that oxidatively modified RNAs at Gua(s) 166 primarily co-localized to IBs, although they are also found in the cytoplasm (Fig. 1b). In support 167 of specificity, our analysis showed that oxidatively modified proteins (carbonylated and 168 nitrosylated) were seen at the periphery and/or adjacent to IBs (Fig 1c,d). Because IBs are the 169 sites of vRNA synthesis, these results also imply that oxidative modification to Gua in the RSV 170 transcriptome possibly occurs co-transcriptionally, either by direct oxidation or by incorporation 171 of oxidatively modified Gua nucleoside triphosphate (8-oxo(r)GuaNTP). The latter may be due 172 to the lack of an innate capacity of RdRp to differentiate between GNTP and 8-oxo(r)GuaNTP, 173 similar to mammalian RNA polymerases 46 . 174

OGG1 converts oxidatively modified Gua into enhanced RNA expression 176
Based on binding of 8-oxo(r)Gua-Ab to and association of OGG1 with vRNAs ( Fig. 1a,  177 b, c), we investigated whether the oxidative Gua modification plays a role in RSV RNA 178 transcription. To test our hypothesis, 8-oxo(r)Gua in vRNA was decreased by normalizing RSV-179 induced ROS through treatment with AOs, EUK-8 (100 or 500 µM) or N-acetyl-L-cysteine [NAC, 180 a GSH precursor) at 10 or 20 µM)] 34 . Total RNAs were isolated, IP-ed using Ab to 8-oxoGua 181 and converted into cDNA for qRT-PCR. Results showed that AO treatment significantly lowered 182 ROS levels (Supplementary Fig. 2d), and abundance of 8-oxo(r)Gua-containing mRNA coding 183 for the attachment protein G, compared to that isolated from infected mock-treated cells (Fig.  184   2a). Similarly, there were lower mRNA levels of NS1, N, G, F and L (Fig. 2b) and genomic 185 vRNA (Fig. 2c), as well as viral progeny (Fig. 2d). This phenomenon was not restricted to 186 hSAECs, as we found that a decrease in 8-oxo(r)Gua levels correlated well with RSV replication 187 (mRNA, genome and progeny) in established airway epithelial cells, A549, which is a model of 188 infection (Supplementary Fig. 2a,b,c). In controls, ribavirin, a guanosine analog, an RdRp 189 inhibitor producing co-transcriptional inhibition 47 was used, which showed an inhibitory effect 190 irrespective of cell type (Fig. 2b,c,d; Supplementary Fig. 2a,b,c). 191 Results showing that decrease in ROS levels led to a lower abundance of 8-oxo(r)  containing vRNA and physical association of OGG1 with vRNAs (Fig. 1a,c) suggested to us that 193 oxidatively modified Gua along with OGG1 has a role in viral replication. To test OGG1 was 194 siRNA-silenced or permanently knocked out by CRISPR-Cas9 genomic editing in hSAECs 195 (hSAEC OGG1-/-), control, silenced or hSAEC OGG1-/cells were infected with RSV. Total RNAs were 196 isolated, converted into cDNA, and vRNA levels were assessed by qRT-PCR. Results showed 197 significantly lower mRNA (NS1, N, G, F and L) levels in OGG1 silenced hSAECs (Fig. 2e) and 198 hSAEC OGG1-/cells (Fig. 2f). Additionally, there were decreased levels of M2-1 and N proteins in 199 knockout cells compared to those expressing OGG1 as determined by Western immunoblotting 200 (Wb) (Fig. 2g, left and right panels). Similar results obtained using OGG1-silenced A549 cells 201 suggested that the observed effects are not restricted to hSAECs ( Supplementary. Fig. 3a). 202 Absence of OGG1 significantly decreased levels of genomic RNAs in hSAECs and A549 cells at 203 12h, 24h and 36h post-infection (RNA levels at 2h post-exposure served as input, Fig. 2h,i and 204 Supplementary Fig. 3d). Lack of OGG1 led to a more than 10-fold decrease in viral yield in 205 hSAEC (Fig. 2j, k) and A549 cells (Supplementary Fig. 3b). Transgenic expression of OGG1 206 in hSAEC OGG1-/restored RSV output to a level similar to that of OGG1 expressing cells (Fig.  207   2k). Silencing other DNA glycosylases, endonuclease VIII-like 1 (NEIL1, recognizes and repairs 208 oxidatively modified ring-fragmented purines, pyrimidines, 5-hydroxyuracil 48 ), or the 8-209 oxo(r)Gua-specific MTH1 45 had no effect on RSV replication (Supplementary Fig. 3a). These 210 results suggest that OGG1 bound to 8-oxo(r)Gua in RNA is actively utilized to maximize output 211

of RSV infection. 212
To examine whether changes in cellular physiology due to OGG1 depletion (selection, 213 puromycin) affected RSV replication, we inhibited OGG1 binding to 8-oxo(r)Gua in vRNAs by 214 using the selective binding-site inhibitor, TH5487 49 . hSAECs were RSV-infected and treated 215 with TH5487 (10 µM or the equivalent volume of solvent). RNAs isolated were subjected to 216 qRT-PCR analysis. Results show that TH5487 treatment of infected cells significantly 217 decreased mRNA levels coded by NS1, N, G, F, and L genes (Fig. 3a). Consistent with these 218 data, we observed significant decreases in the levels of N and M2-1 proteins (Fig. 3b, left and 219 right panels) and genomic RNA (24 and 36 hpi; Fig. 3c) levels. The Inhibitory effects of TH5487 220 were not cell type specific because it also inhibited RSV output in A549 cells (Supplementary 221 Fig. 3b,c,f,h). Inhibition of OGG1's binding to oxidatively modified Gua in vRNA decreased 222 RSV-yield in a concentration dependent (5, 10 µM) manner in hSAECs (Fig. 3d) and in A549 223 cells (Supplementary Fig. 3b). TH2840, an inactive analog of TH5487 or O8, which inhibits 224 only OGG1's enzymatic activity 49,50 , had no effect on RSV titers (Fig. 3d, Supplementary Fig.  225 3b). In line with this, TH5487 inhibited RSV-induced cytopathology while TH2840 and O8 had 226 no effect (Supplementary Fig. 3c). Treatment with TH5487, TH2840 or O8 showed no toxicity 227 in hSAECs or A549 cells (Supplementary Fig. 3i, left and right panels). 228 To validate the data derived from cultured cells, mice were challenged with purified RSV 229 (10 6 PFU/lung) and mock treated or treated with TH5487 at -2h, +1h and 12h intervals over 96 h 230 (30 mg/kg, intraperitoneal (i.p) 49 . At 2h, 48h and 96h post-challenge, the lungs were harvested 231 and total RNAs isolated. The RNAs were converted into cDNA by using oligo-dT and random 232 primers to assess RSV mRNA and genomic RNA levels, respectively. Quantitative RT-PCR 233 results showed that the inhibition of OGG1's substrate binding significantly lowered levels of N, 234 G, and F mRNAs, and genomic RNA compared to mock-treated RSV-infected animals (Fig. 3f, 235 g). There was no sign of TH5487 toxicity in infected animals as shown previously 49 . Data from 236 cell culture and animal experiments strongly suggest that OGG1 binding to oxidatively modified 237 Gua in viral RNAs is essential for efficient RSV replication. Future studies will determine the 238 clinical potential of small molecule inhibitor(s) of OGG1. complementary to the G mRNA) (Fig. 4a, Table 3). G:C contents of these probes were 45% 247 and 47%, respectively. The sequence of the third probe is homologous to the region located 248 between GE of small hydrophobic protein (SH) and GS of G gene (4506-4546 nt; intergenic 249 sequences complementary to antigenome). The RNA isolated from RSV-infected cells was 250 incubated with individual probes, denatured at 95 ⁰C for 5 min, and hybridized. Un-hybridized 251 probes and RNAs were digested with mung bean nuclease 48 , then OGG1 was added, cross-252 linked and mixtures were subjected to EMSA. Results showed that OGG1 extensively bound to 253 the probe-RNA hybrid formed at the 5'-end of viral mRNA, while a weak hybridization signal was 254 observed at the 3'-end of mRNAs (Fig. 4b, lane 6). The probe that was complementary to 255 antigenome gave a weak signal, suggesting that the hybridization observed at the 3'-end of 256 mRNA may be considered as background signal (Fig. 4b, lane 6). RNase-H completely 257 eliminated EMSA signals (Fig. 4b, lane 7, 8 and 9), indicating that OGG1 specifically bound to 258 the 5'-end of the mRNA region that acquired the oxidative modification of Gua, 8-oxo(r)Gua. 259 Accordingly, OGG1 bound to the synthetically made DNA-RNA hybrid containing 8-oxo(r)Gua in 260 the RNA strand in a concentration dependent manner (Fig. 4c). The active-site inhibitor, 261 TH5487, prevented OGG1 binding to the DNA-RNA hybrid containing 8-oxo(r)Gua, lending 262 additional support for binding specificity (Fig. 4d). These results suggest that oxidative 263 modification of Gua occurs at the 5'-UTR of mRNAs. Results from EMSAs show that OGG1 binds 8-oxo(r)Gua containing dsRNA and not ssRNA 273 ( Fig. 4e, lanes 6,7,8), an interaction that could be prevented by TH5487 (Fig. 4f,g). OGG1 had 274 no enzymatic activity on dsRNA (Fig. 4h, lanes 2,3,4), while it efficiently excised damaged 275 bases from dsDNA 32 (Fig. 4h, lanes 6,7,8; Supplementary Fig. 4a). Therefore, removal by 276 OGG1 of 8-oxo(r)Gua (cleavage of the N-glycosylic bond in RNA) and subsequent degradation 277 of RNA is an unlikely scenario in cellulo. These results also imply that OGG1 interacts with 8-278 oxo(r)Gua formed co-transcriptionally at the transcription active site. 279

OGG1 physically interacts with the transcriptional elongation factor M2-1 281
In eukaryotic cells OGG1 binds to 8-oxo(d)Gua, an epigenetic-like mark in gene 282 promoters, and physically interacts with transcription factors, including NF-κB, to facilitate its 283 DNA occupancy 53 . By analogy, we propose that OGG1 binds to 8-oxo(r)Gua at the 5'-end of 284 RNA and interacts with viral protein(s), potentially aiding efficiency of transcriptional machinery. 285 To test this possibility, FLAG-tagged OGG1, (FLAG)OGG1, expressing hSAECs were infected 286 with RSV, lysed and IP-ed using Ab to FLAG. Complexes that were IP-ed were subjected to 287 Wb analysis using a polyclonal Ab to RSV proteins. The results showed that (FLAG)OGG1 co-288 IP-ed with M2-1 (~25 kDa subunit of the tetrameric M2-1 is shown after SDS-PAGE; (Fig. 5a). 289 To confirm this, M2-1 was identified using a monoclonal Ab to M2-1 (Fig. 5b). The lower panel 290 of Fig. 5a and 5b shows intrinsic and transgenic expression of OGG1 using Ab to OGG1. We 291 also noted that there was another ~40 kDa OGG1-interacting protein immuno-precipitate ( Fig.  292 5a, * right panel); however, its identity remains unknown. To test whether OGG1 directly interacts 293 with M2-1, we purified it to homogeneity 54 (Supplementary Fig. 4b,c), mixed it with OGG1, and 294 performed pull-down assays. Results showed that OGG1 pulled down M2-1 protein, and Ab to 295 M2-1 IP-ed with OGG1 ( Fig. 5c,d), suggesting a direct interaction between M2-1 and OGG1. 296 There was no interaction between M2-1 and MTH1, which was used as control protein (Fig. 5e). 297 OGG1 typically localizes to the nucleus and mitochondria 32 , while M2-1 is found in the 298 cytoplasm, primarily in virus-induced IBs 15,55 . To obtain information on the site of the OGG1-M2-299 1 interaction in cellulo, microscopic imaging (immunohistochemistry; IHC) and in situ proximity 300 ligation assays 56 were performed using RSV infected hSAECs. Time course studies showed an 301 accumulation of OGG1 in the cytoplasm of RSV-infected cells from 6 hpi, as shown by IHC and 302 Wb analysis ( Fig. 5f, g, Supplementary Fig. 4d), while its level did not change in the nuclear 303 compartment (Fig. 5g, Supplementary Fig. 4e, upper panels), suggesting that de novo 304 synthesized OGG1 remained in the cytoplasm. After overlaying IHC images, strong co-305 localization of OGG1 and M2-1 was observed, especially in IBs. This observation was confirmed 306 by calculating the Pearson fluorophore-moment correlation coefficient 57 . E.g., in contrast to 6 307 hpi, correlation coefficient (R) at 12, 24 and 42 hpi were R = 0.93, R = 0.93 and R = 0.89, 308 respectively ( Fig. 5f, most right-hand panels). The positive proximity ligation assays strongly 309 suggest there was a physical interaction between OGG1 and M2-1 within IBs (Fig. 5h,i). The physical interaction observed between OGG1 and M2-1 in vitro and in cellulo ( Fig.  313 5a,b,c,d,f) and the fact that OGG1 bound to 8-oxo(r)Gua at the 5'-end of nRNA (Fig. 4b, c, d), 314 suggest that OGG1 may promote interaction of M2-1 with the transcriptionally active site at GS. 315 To model this scenario, we utilized ss and ds RNA probes containing ±8-oxo(r)Gua. The 316 presence or absence of 8-oxo(r)Gua in ss or ds RNAs had no effect on M2-1 binding (Fig. 6a). 317 M2-1 bound to ssRNA and dsRNA as a tetramer (arrow heads) and also formed higher 318 molecular size complexes (Fig. 6a) similar to those described for glutaraldehyde cross-linked 319 forms of RNA-M2-1 complexes 58 . Compared to ssRNA (Fig. 6a, lane 1,2) there was a 25-fold 320 decrease in M2-1's association with dsRNA ±8-oxo(r)Gua (Fig. 6a, lane 3, 4). To test whether 321 M2-1 binding to dsRNA was specific, we show that Ab to M2-1 super-shifted all M2-1-dsRNA 322 complexes (Fig. 6b, lane 3). Ab to OGG1 inhibited its binding to 8-oxo(r)Gua containing dsRNA 323 Fig. 6b, lane 2). IgG was used as control (Fig. 6b, lane 1). 324 To examine if OGG1 promotes binding of M2-1 to RNA, OGG1 was pre-incubated with a 325 dsRNA probe ±8-oxo(r)Gua and M2-1 was added. OGG1 increased M2-1 occupancy on 8-326 oxo(r)Gua-containing dsRNA in a concentration dependent manner (Fig. 6c, lanes 6 to 10 and 327 Fig. 6c, right panel), while without 8-oxo(r)Gua in the dsRNA, OGG1 had no effect (Fig. 6c, 328 lanes 1 to 5 and Fig. 6c). Time course experiments showed that OGG1 accelerated the 329 interactions between 8-oxo(r)Gua-containing dsRNA and M2-1 (Fig. 6d, lanes 7 to 11 and Fig.  330 6d right panel). For instance, the level of dsRNA-associated with M2-1 was similar at t = 1 min 331 in the presence of OGG1 to that of t = 6 min without OGG1 (Fig. 6d, lanes 2 and 7,  332   Supplementary Fig 4g). Interestingly, there was no M2-1-OGG1-dsRNA complex observed; 333 however, OGG1 formed a separate complex with dsRNA (Fig. 6d, arrow). Active site inhibitor 334 (TH5487) decreased the effect of OGG1 on the interaction between M2-1 and dsRNA in a 335 concentration dependent manner (Fig. 6e, right panel). Without 8-oxo(r)Gua OGG1 had no 336 effect (Fig. 6c, lanes 1 to 5 and right panel), and in the absence of OGG1, TH5487 had no 337 effect on M2-1 binding (Fig. 6f). These data imply that prior OGG1 binding to dsRNA is needed 338 for efficient association of M2-1 with dsRNA. In control experiments, MTH1 had no effect on M2-339 1 binding to dsRNA (8-oxo(r)Gua:C) (Supplementary Fig. 4h). From these results, we 340 speculate that OGG1 binds to dsRNAa hybrid between nRNA and the 3'-end of viral gene(s) 341 via co-transcriptionally generated 8-oxo(r)Gua, where it interacts with and recruits M2-1 to the 342 transcriptionally active site(s). OGG1 had no effect on the interaction of M2-1 with ssRNA 343 containing 8-oxo(r)Gua (Supplementary Fig. 4f). 344 We next explored the fate of OGG1 once the full-length mRNA disengages from 345 genome. To do so, mRNAs isolated form RSV-infected cells were mixed with His-tagged OGG1 346 in binding buffer, and IP assays were performed. Levels of mRNA encoding for G protein in IPs 347 was determined using qRT-PCR. Compared to IgG, Ab to His-(OGG1) had enriched viral RNAs 348 coding for G protein in IPs (Supplementary Fig. 4j). Heat denaturation of RNA (65 o C for 5 min) 349 prevented OGG1 binding. In controls, M2-1 extensively IP-ed with G-mRNA, while N protein 350 lacked interaction with mRNA (Supplementary Fig. 3j). Taking that the 5'-end of nRNAs is the 351 preferential site of OGG1 binding (Figs 4b,c,d), we performed a series of experiments that 352 examined OGG1 binding to a Cy5-tagged 40 nt base-long ssRNA (identical to 5-end of G-353 mRNA). Results showed that OGG1 bound to the non-denatured RNA probe only when it 354 contained 8-oxo(r)Gua (Fig. 6g, lanes: 5 to 8) similar to dsRNA (Fig. 6g, lanes: 9 to 12). Heat 355 denaturation (65 o C for 5 min) of RNA abolished OGG1 binding (Fig. 6g, lanes 1 to 4) implying 356 that OGG1 binds to the substrate via temperature-sensitive secondary RNA structure(s) at the 357 5-end of G-mRNA. The predicted secondary structure of the probe is shown in Fig. 6g (right 358 panel). A secondary structure of G-mRNA is shown in Supplementary, Fig. 4i as predicted by 359 "RNAstructure" software 59 . OGG1 lacks enzymatic activity on dsRNAs (Fig. 4h, Supplementary  360   Fig. 4a), therefore we speculate that OGG1 remains in complex with full-length viral mRNAs 361 (Fig. 6h). The post-transcriptional roles of OGG1 is a matter for future studies. 362

Discussion 363
RSV is one of the leading causes of life-threatening respiratory system illnesses globally 364 in infants, preschool children, immunocompromised adults and the elderly; yet, treatment is 365 mostly limited to supportive medical care 1,4,7 . RSV infection-induced generation of ROS is linked 366 to expression of inflammatory genes and severity of clinical disease; however, the mechanism 367 by which the oxidizing environment modulates virus replication remains unknown. Here we 368 report that vRNAs acquire ROS-generated 8-oxo(r)Gua as an epitranscriptomic mark that is 369 "read" by the repurposed DNA repair protein OGG1. Inhibition of 8-oxo(r)Gua generation 370 through AOs, OGG1 knockdown or inhibition of its binding to epitranscriptomic mark significantly 371 decreased the level of viral mRNAs, proteins, and viral progeny in cultured cells and airways. 372 Reconstitution experiments show that OGG1 binds to the 5'-UTR of nRNAs containing 8-373 oxo(r)Gua while paired with the 3'-end of viral gene(s) interacts with and recruits the 374 transcriptional elongation factor, M2-1 to transcriptionally active sites. Taken together, these 375 data suggest that ROS generated in RSV-infected cells is utilized to oxidize Gua in vRNA, which 376 serves as an epitranscriptomic mark that repurposes OGG1 to enhance RSV replication. These 377 studies also illustrate how the virus usurps cellular proteins to enhance its replication. 378 RNAs are highly vulnerable to oxidative modifications, due to their single stranded 379 nature, partial protection by RNA-binding proteins, and unprotected Watson-Crick edges of 380 nucleotides 27-29 . The nucleobase of guanosine is especially vulnerable to modifications as it has 381 the lowest oxidation potential 63 . Therefore, 8-oxoGua is often generated in RNAs and in 382 nucleotide triphosphate pool by ROS 28, in an oxidizing intra-cellular environment, such as in 383 RSV-infected cells 36 . Therefore, association of lesion-specific OGG1 with or IP by 8-oxoGua Ab 384 of mRNAs of NS1, N, G, F and L genes in RSV-infected cells is not surprising. Importantly, the 385 level of OGG1-associated vRNA was similar to that of RNA that was IP-ed using Ab to M2-1. 386 M2-1 interacts with RNA via its zinc-binding and core domains 44,60 . OGG1 lacks a zinc-finger 387 motif 61 , so its interaction with vRNA likely occurs through its recognition domain, which is 388 specific for 8-oxoGua in DNA and RNA. 389 Although oxidative modification to Gua can occur at any Gua islet (GGG, GGGG) within 390 genomic RNA and mRNAs, hybridization-coupled EMSA clearly indicated that 8-oxo(r)Gua was 391 The exact position of 8-oxo(r)Gua in these islets is yet to be determined; however, based on the 395 previously characterized nature of charge migration, it is most likely located at the end of G 396 stretches 51 . It is known that during viral replication, the dsRNA hybrid may reach 30 nucleotides 397 or more. For instance, in the case of VSV, mRNA is separated from the genome and then 398 capped after reaching 31 nucleotides 52 . Thus, to reconstitute an in cellulo scenario, 8-oxo(r)Gua 399 was synthetically placed at the end of 3'-end of G series in a 40-nucleotide long RNA that is 400 complementary to the 3'-end of genomic RNA (probe), which mimics dsRNA (probe) similar to 401 the polymerization active site at GS 52 . Binding of OGG1 to dsRNA with 8-oxo(r)Gua was similar 402 to DNA containing 8-oxo(d)Gua. Although M2-1 is an RNA binding protein, it showed poor 403 interaction with dsRNA ±8-oxo(r)Gua compared to ssRNA ±8-oxo(r)Gua. However, OGG1 404 significantly increased M2-1 binding (over 10-fold) only to dsRNA containing 8-oxo(r)Gua. The 405 exact molecular mechanism by which OGG1 enhances M2-1 binding is yet to be elucidated; 406 however, it is possible that OGG1-induces structural changes in dsRNA that are creating a 407 specific binding interface in the hybrid between the genome and nRNA that facilitates M2-1 408 interaction with the polymerization complex (Fig. 6h). This scenario could be similar to that seen . In contrast to epitranscriptomic moieties that are deposited and removed by cellular 424 enzymes called "writers" and "erasers", respectively 22 , 8-oxo(r)Gua is directly generated by the 425 interaction of Gua with one of several reactive species (primarily hydroxyl radical 63 ). Thus ROS 426 is considered to be a covalent modifier ("writer") of Gua by oxidizing --subtracting an electron 427 generating the epitranscriptomic mark 8-oxo(r)Gua of which "reader" is the repurposed base 428 specific DNA repair protein OGG1. As a "reader" OGG1 facilitates M2-1 binding to 5'-UTR, 429 thereby promoting mRNA synthesis. Of note, OGG1 has an innate ability to migrate/diffuse 430 along nucleic acid strands without energy requirement 38  Interestingly, OGG1 level in the nucleus was unchanged suggesting that the newly made OGG1 449 remained in the cytoplasm. Because cytoplasmic and nuclear extracts consistently showed the 450 ~39 kDa OGG1, we exclude the possibilities that new variants were made by alternative splicing 451 or that the mitochondrial OGG1 (42.5 kDa) is involved. The molecular mechanism by which 452 OGG1 is targeted to IBs remains unclear, but we note that site-specific posttranslational 453 modifications (acetylation, phosphorylation, O-GlcNAcylation and others) of OGG1 could be the 454 culprit. Importantly, the Pearson fluorophore-moment correlation coefficient 57 clearly showed 455 that OGG1 co-localized with M2-1 within IBs and physically interact as shown by in situ 456

proximity ligation assays and their co-IP from RSV-infected cells. 457
The fate of OGG1 on viral mRNA remains unclear; however, we note that OGG1 has an 458 innate ability to migrate/diffuse along nucleic acid strands without energy requirement 38 . Thus, 459 OGG1 may migrate along with M2-1 on RNA to facilitate transcription elongation. Another 460 possibility is that OGG1 remains engaged with the 8-oxo(r)Gua-containing UTR of mRNA via 461 temperature sensitive secondary stem-loops of RNA (Fig. 6h). providing the OGG1 inhibitor (TH5487) during initial phases of these studies. 499 We thank Dr. Sherry Haller Editor (Department of Microbiology and Immunology, University of 500 Texas Medical Branch at Galveston) for English editing of the manuscript. Healthcare, Piscataway, NJ) using a linear gradient of 50-1000 mM NaCl that further removed 602 nucleic acids. The purified His-M2-1 recombinant protein was exchanged into the low salt buffer 603 (20 mM Tris-HCl pH 7.4, 150 mM NaCl) and its purity was confirmed by spectrophotometry (OD 604 260nm/280nm ratio) and Western blot analysis. 605 RNA extraction and qRT-PCR. Total RNAs were extracted using RNeasy Mini kit (Qiagen) 606 according to the manufacturer's instructions. Crude RNAs were DNaseI-treated and loaded onto 607 an RNeasy column and subjected to washes with RW1 and RPE buffers. RNAs were eluted 608 with the RNase-free water included in the kit. In specific experiments, mRNAs were isolated 609 using a magnetic mRNA isolation kit (NEB, S1550S). The RNA concentration was determined 610 spectrophotometrically on an Epoch Take-3™ system (BioTek, Winooski, VT) using Gen5 v2.01 611 software. The quality of the RNAs were confirmed via the 260/280 nm ratio (varied between 1.9 612 -2.0). 1000 ng total RNA was used to generate cDNA with either oligo-dT or random hexamer 613 primers (Takara, Item #: RR037A). qPCR was performed using specific primers for RSV NS1,

614
N, G, F and L mRNA. The level of RSV genome was determined using primer pairs amplifying 615 inter-genome (upstream from the attachment glycoprotein G) and the 3'-end of G gene (genome 616 coordinates 3' c). As internal control, cellular GAPDH was used (sequence of primers are listed 617 in Table 1). Changes in mRNA and genomic RNA levels were expressed as fold increases over 618 levels at 2h (after adsorption) in infected cells using the 2 -∆∆Ct method. 619  Table 1. qPCR data were normalized using the fold 687 enrichment method. In brief, this method is based on the assumption that the IP using a specific 688 Ab contains both specific signal and background signal, whereas the IP of a negative control-689 IgG represents the background only. The background was subtracted from the signal, and the 690 remaining value corresponded to the net pull-down of a specific RNA by a given specific Ab 691 normalized to both total RNA and nonspecific (negative control) IP. 692 Electrophoretic mobility shift assay (EMSA). To evaluate OGG1 and/or M2-1 binding ssRNA 693 or dsRNA with or without oxidative modification to Gua, EMSA was performed. RNA oligos 694 (Table 3)  HCl (pH 7.5), 10 mM of NaCl, 1 mM of EDTA, 1 mg/mL BSA, and 1 mM of DTT). After 718 incubation for 10 min at room temperature, the reaction was halted by adding 10 µL loading 719 buffer (containing 8 µL of formamide, 2µL of 10 mM of NaOH) and heated for 5 minutes at 95°C. 720 The cleaved product was separated from the intact probe in a 15% polyacrylamide gel 721 containing 8 M urea in Tris-borate-EDTA buffer (pH 8.4). The separated bands were visualized 722 by using Amersham TM Imager 680.    Table 1).  1,2,3).