Respiratory Syncytial Virus Sequesters NF-κB Subunit p65 to Cytoplasmic Inclusion Bodies To Inhibit Innate Immune Signaling

Many viruses replicate almost entirely in the cytoplasm of infected cells; however, how these pathogens are able to compartmentalize their life cycle to provide favorable conditions for replication and to avoid the litany of antiviral detection mechanisms in the cytoplasm remains relatively uncharacterized. In this manuscript, we show that bovine respiratory syncytial virus (bRSV), which infects cattle, does this by generating inclusion bodies in the cytoplasm of infected cells. We confirm that both bRSV and human RSV viral RNA replication takes place in these inclusion bodies, likely meaning these organelles are a functionally conserved feature of this group of viruses (the orthopneumoviruses). Importantly, we also showed that these organelles are able to capture important innate immune transcription factors (in this case NF-KB), blocking the normal signaling processes that tell the nucleus the cell is infected, which may help us to understand how these viruses cause disease.

microscopy (CLEM) to colocalise RSV N protein and P65 within bRSV IBs; granular, 23 membraneless regions of cytoplasm with liquid organelle-like properties. Additional 24 characterisation of bRSV IBs indicated that although they are likely formed by liquid-25 liquid phase separation (LLPS), they have a differential sensitivity to hypotonic shock 26 proportional to their size. Together, these data identify a novel mechanism for viral 27 antagonism of innate immune signalling which relies on sequestration of the NF-κB 28 subunit p65 to a biomolecular condensatea mechanism conserved across the 29 Orthopneumovirus genus and not host-cell specific. More generally they provide Introduction 51 Bovine and human respiratory syncytial viruses (bRSV and hRSV, respectively), are 52 closely related viruses that cause acute respiratory illness in cattle and humans, 53 respectively. The viruses infect all ages, but severe illness associated with bronchiolitis 54 and pneumonia is more common in calves (for bRSV) and infants, the elderly and 55 immunocompromised (for hRSV) [1,2]. Although the process is poorly understood, 56 immune responses to RSV infections are incomplete leading to re-infection, even in 57 healthy adults [3]. In high-risk groups hRSV infection can be fatal; however, there is 58 no approved vaccine and only a single therapeutic option, monoclonal antibodies 59 against the F protein. Whilst there are available bRSV vaccines these are mildly 60 protective and there is evidence for an exacerbation of natural infection [4]. Both 61 viruses were recently taxonomically reclassified as species Bovine and Human 6 rightinset zoom). However, IF and intensity profile analysis revealed that, even in 162 the case of hTNFα stimulation, p65 nuclear translocation in bRSV infected cells was 163 absent and that most p65 remained in the observed perinuclear puncta (Fig 1A; bottom 164 leftinset zoom). bRSV can infect a broad range of host cells in vitrogrowing to 165 similar titres in both Vero and Madin-Darby bovine kidney (MDBK) cells 166 (supplementary Fig 1A and B). To examine the apparent innate immune antagonism 167 in bovine cells, equivalent infections were performed in MDBK cells. These 168 experiments confirmed the same p65 sequestration into perinuclear puncta following 169 bRSV infection, as well as the related insensitivity to TNFα stimulation (supplementary 170 Fig 1C) indicating a conserved mechanism of antagonism active in both primate and 171 ruminant cells. 172 To examine the effect of this sequestration on NF-κB signalling, we next employed a 173 luciferase reporter assay to assess NF-κB transactivation. HEK293T cells were 174 infected with bRSV at an MOI of 1, before being transfected with the NF-κB reporter 175 and subsequently treated with or without TNFα ( Fig 1B). Interestingly, infection without 176 TNFα treatment did not result in any significant activation of the reporter, despite 177 demonstrable viral protein production ( Fig 1B,

black bars and RSV F western blot), 178
indicating that even in the presence of active viral replication there is little to no 179 activation of the NF-κB signalling pathway in bRSV-infected cells. Indeed, activation 180 of the NF-κB reporter was only seen following addition of 20 ng/ml of exogenous 181 hTNFα; however, this activation was significantly less in infected cells, when 182 compared to the mock ( Fig 1B, grey bars). Separately, we also examined protein 183 levels of p65 (total and transiently phosphorylated) and IκBα, components of NF-κB 184 signal transduction, in infected Vero cells with and without TNFα stimulation. As 185 expected TNFα treatment of mock-infected cells resulted in an increase in p65 186 phosphorylation and a decrease in total IκBα (presumably the result of proteasomal 187 degradation following its own phosphorylation) (Fig 1C; mock -/+ TNFα) [9]. The 188 detected levels of phospho-NFκB p65 and total IκBα in infected cells (Fig 1C; infected 189 -/+ TNFα) confirmed the lack of activation during infection and also the modest  activation induced by bRSV infection with subsequent TNFα treatment observed in Fig  191   1B. Together, the data strongly suggests that NF-κB signalling is inhibited by bRSV 192 infection due to its sequestration into intracytoplasmic puncta. Importantly, these data 193 also indicate that the sequestered p65 is not in a transcriptionally active state, since 7 infection did not result in a marked increase in p65 phosphorylation nor evidence for 195 demonstrable IκBα degradation. 196 BRSV replication induces the recruitment of the NF-κB subunit p65 into intra-197 cytoplasmic bodies distinct from stress granules 198 NF-κB p65 puncta were only observed in bRSV infected cells showing detectable 199 levels of F protein, indicating a correlation between productive infection and 200 sequestration ( Fig 1A). To examine this correlation and define the kinetics of p65 201 sequestration over time, MDBK cells were infected at an MOI of 1 and fixed at different 202 times post infection (p.i), before being permeabilised and the distribution of p65 and 203 RSV F analysed by IF. Detectable NF-κB p65 puncta (>3 µm 2 ) were apparent in 204 infected cells by 16 h p.i. (Fig 2B), correlating with significant levels of F expression 205 (Fig 2A). Interestingly, two populations of F protein were present at this stage, a 206 perinuclear, presumably ER-or vesicle-associated population (Fig 2A; white arrow), 207 and a peripheral more filamentous-like population, possibly the site of virion 208 biogenesis (Fig 2A; beige arrow)neither of which appeared to colocalise in any 209 significant way with p65. By 24 h p.i., all infected cells contained at least one p65 210 puncta with none being observed in nearby uninfected cells. Using fluorophore line of 211 interest analysis, we were also able to assess the ratio of cytoplasmic-to puncta-212 localised p65 as well as the increasing diameter of these aggregates. As infection 213 proceeded the intensity of p65 in the puncta increased as the level of disperse p65 in 214 the cytoplasm decreased ( Fig 2C; 'p65 in puncta' vs. 'p65 outside puncta'), indicating 215 coalescence, and supporting our observations in Fig 1C that (Fig 2B). Smaller p65 puncta (<10 µm 2 ) were 219 also observed at 48 h p.i., most likely the result of nascent infections in nearby cells. 220 By this time F protein expression was markedly different, with less distinct populations 221 of protein; however, there was still no obvious co-localisation with the p65 puncta. A 222 similar pattern of results was also observed in Vero cells (supplementary Fig 2). 223 Our first line of inquiry following the identification of p65 puncta in bRSV infected cells 224 was based on their visual similarity to protein and mRNA aggregations that form in 225 cells in response to cellular stress and viral infections, so-called stress granules (SG). 226 A wide range of viruses have been shown to either induce or inhibit SG formation to 227 their advantage [33]; however, there are contradictory findings on SG induction by 228 RSV [25,[34][35][36]. To examine the potential relationship between these p65 puncta and 229 SG we induced SG formation in bRSV infected cells with sodium arsenite treatment 230 and performed co-immunostaining for p65 and G3BP1 (a SG marker) in fixed cells. 231 Although we were able to successfully stimulate the production of SGs in Vero cells 232 our analysis showed that the p65 puncta were entirely distinct from these granules 233 ( Fig 2D). Tangentially, this experiment also demonstrated that bRSV infection does 234 not significantly induce SG formation. 235 The NF-κB subunit p65 co-localises with viral inclusion bodies independently of 236

RSV-encoded immunomodulators 237
RSV has a relatively small genome, encoding 11 proteins from 10 genes ( Fig 3A). interest plots). The sub-IB localisation of bRSV N and P was similar to that previously 255 described for hRSV, with N and P being found on the periphery of the organelle [26]. 256 The significant intra-IB localisation of the M protein at 24 h p.i., as well as its partial 257 nuclear localisation, is consistent with previously reported IF in RSV-infected cells [37, 258 38]. However, the role of M in RNA virus IBs reflects an interesting point of divergence; 259 with some viral IBs being M positive (e.g. RSV) and others negative (e.g. rabies) [39].
Significantly, the larger N, P or M-positive IBs were, in the majority of cases, also p65 261 positive ( Fig 3B; red IF panels) identifying, for the first time, that this NF-κB component 262 was being recruited to RSV inclusion bodies in infected cells. To examine this in detail 263 we next characterised the number, size and p65 status of N-positive IBs in infected 264 cells, observing that they were numerous and mostly localised in the median section 265 of the cell. We therefore obtained images from multiple planes in this section to 266 assemble max intensity z-stacks to aid quantification. represents an entirely novel mechanism of viral inhibition of NF-κB signalling, since it 283 is the sequestration of signalling components to a viral organelle, rather than the 284 degradation commonly seen [12,22], which leads to the innate immune antagonism 285 We infected cells with wild type bRSV (wt), or recombinant bRSVs which do not 289 express these proteins (ΔNS1, ΔNS2, ΔNS1/2 (a double knockout) or ΔSH - [15,40]). 290 At 24 h p.i., infected cells were fixed and co-immunostained for p65 and the RSV F 291 protein. To confirm deletion of SH, immunostaining was performed using an anti-SH 292 antibody (supplementary Fig 4A). Since we did not have access to anti-NS antibodies, the genotype of these mutants was confirmed by RT-PCR, on RNA extracted from 294 infected cells targeting the region of NS deletion (supplementary Fig 4B). IF analysis 295 of these samples identified p65 puncta in all infected cells (Fig 3F), suggesting that 296 these bRSV encoded immunoantagonists do not play a significant role in either the 297 formation of IBs or the sequestration of p65 to these structures. 298 bRSV IBs are sites of RNA replication but p65 does not specifically co-localise 299 with M2-1 or nascent viral RNA in IB-associated granules (IBAGs) 300 hRSV inclusion bodies have previously been shown to be the sites of virus 301 transcription and replication [25,26,41]. To confirm bRSV IBs are also the site of viral 302 RNA replication, we carried out nascent RNA labelling using 5-ethynyl-uridine (5EU) 303 incorporation. Mock infected MDBK cells, incubated with 5EU for 1 h, revealed, as 304 expected, 5EU incorporation into cellular RNA in the nucleus ( Fig 4A; top row). When 305 cellular transcription was inhibited following pre-incubation of mock infected cells with 306 actinomycin D (Act D) for 1 hr this signal was lost. 5EU labelling performed on bRSV 307 infected cells without Act D treatment did not reveal significant evidence for viral 308 replication in IBs; perhaps due to over-representation of cellular RNA synthesis. 309 However, in the presence of Act D, labelled, newly synthesised RNA could only be 310 seen in the N-positive IBs, presumably the result of viral replication. This co-311 localisation of 5EU incorporation and N-protein within IBs provides strong evidence 312 that bRSV IBs are the sites of viral RNA replication. A more detailed look at the IBs 313 ( Fig 4A; inset zoom and line of interest plot -asterisks) revealed partial sub-IB 314 organisation to the RNA found within these structures. Interestingly, a recent study on 315 hRSV IBs identified similar functional compartments within IBs termed inclusion body-316 associated granules (IBAGs) [26]. These were shown to concentrate newly 317 synthesised viral mRNA and the viral M2-1 protein but not genomic RNA, or the N, P 318 and L proteins. To confirm the presence of IBAGs in bRSV IBs we immuno-stained 319 bRSV-infected cells for M2-1 following nascent viral RNA labelling, observing co-320 localisation of both these components ( Fig 4B). The intra-IB organisation of RNA 321 replication and M2-1 protein into IBAGs appears, therefore, to be a structurally 322 conserved aspect of orthopneumovirus IBs. We next examined the potential co-323 localisation of p65 with these sites of nascent vRNA localisation (IBAGs). Although we 324 observed partial sub-IB localisation signals for p65, this did not always co-localise with 325 vRNA (Fig 4B) or, in subsequent experiments, M2-1 ( Fig 4C). These findings suggest that there are multiple sub-compartments within bRSV IBs, in addition to IBAGs, which 327 potentially carry out a distinct range of functions. when compared to their larger (>3 µm in diameter), more pleomorphic counterparts 340 ( Fig 5A). As expected, these structures were not membrane-bound or directly 341 associated with sub-cellular organelles; however, rough endoplasmic reticulum (RER) 342 and mitochondria were frequently found in close proximity ( Fig 5A). These structures 343 are similar to those previously reported for other RNA viruses [28,39], supporting our 344 conclusion that bRSV also forms membraneless IBs in infected cells. 345 Various reports have also demonstrated that IBs can rapidly change their size due to 346 fusion or fission whilst remaining spherical in nature, a characteristic feature of these 347 liquid organelles [42]. Rabies virus inclusion bodies, termed negri bodies, have been 348 shown to rapidly dissolve and reform in response to hypotonic shock, demonstrating 349 the dynamic nature of these structures [27,28,46]. To assess the sensitivity of bRSV 350 IBs to hypotonic shock, Vero cells, infected with bRSV for 24 h, were incubated with 351 DMEM (diluted to 20% in H2O) for 20 mins. Cells were then fixed and immunostained 352 for N protein. Many smaller IBs showed evidence of dissolution following hypotonic 353 shock (Fig 5B; iv); however, unlike rabies virus negri bodies, the larger bRSV IBs 354 remained intact following this significant period of cellular osmotic shock (Fig 5B; iii). 355 Of note, incubation beyond 20 minutes was not possible because of the associated 356 cytotoxicity. In addition, a large percentage of the sequestered p65 in these larger IBs 357 remained tightly associated with the intact structure ( Fig 5C). Recently,Zhou et al.,358 demonstrated that larger measles IBs had slower rates of fluorescence recovery after photobleaching (FRAP), relative to their smaller counterparts, postulating that these 360 structures had acquired a more gel-like property. The acquisition of this gel-like status, 361 which are also less likely to exchange molecules with the surrounding cytoplasm, has 362 been linked to aging of phase separated organelles -a continuum which ends with the 363 formation of irreversible aggregates [47]. The insensitivity of large bRSV IBs to osmotic 364 shock, and the maintenance of p65 within the IB even under these harsh conditions, 365 is perhaps the result of them acquiring gel-like status, a property which may be linked 366 to the age and size of individual IBs within infected cells. 367 Finally, to examine the sub-IB localisation of RSV N and p65 in relation to our 368 ultrastructural analysis of IBs, we performed correlative light electron microscopy 369 (CLEM). Vero cells were infected at an MOI of 1 and analysed at 24 and 48 h p.i., 370 firstly by confocal microscopy using N and p65 antibodies to immunolabel these 371 proteins ( Fig 5D). The same cells, identified by grid reference, were then isolated, 372 embedded and sectioned with their ultrastructure subsequently analysed by TEM. 373 Importantly, these CLEM data confirmed that the electron dense granular structures 374 seen by TEM (Fig 5A) are synonymous with the N, P, M and p65 stained IBs seen in 375 IF microscopy ( Fig 3B). To our knowledge this is the first CLEM to be performed on 376 an RNA virus IB. An overlay of the two images confirmed that bRSV IBs had retained 377 the electron dense granular structure characteristic of liquid organelles, even with the 378 chemical permeabilization required for IF antibody labelling ( Fig 5C and  379 supplementary Fig 5). Our CLEM data also confirmed the p65 and N proteins localising 380 to the IB, with p65 present within the structure and N around the periphery. At 24 h 381 p.i., the p65-positive IB structures were mostly spherical, becoming larger and more 382 irregularly shaped by 48 h p.i., possibly as a result of transition into a more gel-like 383 status, as discussed above. A similar pattern of immunostaining and IB morphology 384 was also observed in bRSV-infected MDBK cells analysed by CLEM (Supplementary 385 Fig 5). 386

Co-expression of bRSV N and P proteins induces the formation of IB-like 387
structures which can sequester p65 388 In the absence of infection, ectopic co-expression of many Mononegavirales N and P 389 proteins has been shown to result in the formation of IB-like structures [24,26,28,30] 390 a finding which has been linked to their potential to induce LLPS independently of 391 viral infection. Although there has been broad discussion that this is related to the presence of intrinsically disordered regions within the N and P proteins, a definitive 393 functional mechanism for this viral-induced LLPS remains uncharacterised. In 394 addition, whether these infection-independent IB-like structures retain all the 395 properties of viral IBs is not entirely clear. For hRSV it was shown that IBAGs do not 396 form within these visually orthologous bodies [26]; however, the recruitment of MDA5 397 and MAVS to IB-like structures, following N and P overexpression, was maintained 398 [24]. To address similar questions for bRSV IBs, and to examine the related 399 sequestration of p65, Vero cells transiently transfected with plasmids expressing 400 bRSV N (pN) and bRSV P (pP) were fixed and stained at 24 h post transfection and 401 examined by IF. As has been reported previously, expression of N or P alone did not 402 lead to the formation of IB-like structures; however, co-expression did, resulting in the 403 formation of inclusions up to 6.9 µm 2 in area ( Fig 6A). Examination of the sub-cellular 404 localisation of p65 in this system also confirmed that the N-and P-induced inclusions 405 were proficient in sequestering p65, independent of viral replication, with a pattern of 406 expression mirroring that seen in infected cells (Fig 6A; inset zoom and fluorescent 407 line of interest analysis). 408 To examine the mechanism of p65 recruitment to, and sequestration within, the bRSV 409 IB we next investigated whether there was evidence for direct protein-protein 410 interactions between this protein and N or P. Endogenous p65 or p65 expressed from 411 a plasmid (pP65) were immunoprecipitated from bRSV-infected, or mock-infected, 412 293T cells (at 24 h p.i) using an anti-p65 antibody. When these immuno-precipitates 413 were analysed by western blot, both bRSV N and P were found to co-414 immunoprecipitate (co-IP) with endogenous or overexpressed p65 in infected cell 415 lysates, providing evidence of direct interactions being maintained post-lysis ( Fig 6B). 416 Experiments with beads alone did show a small amount of co-IP N protein; however, 417 this was markedly lower than in the p65 antibody experiment, background signal which 418 we believe may be the consequence of the high levels of N protein in infected cells at 419 24 h p.i. In summary, our results indicate that p65 recruitment into bRSV IBs is 420 maintained even in IB-like structures formed after N and P overexpression. 421 Furthermore, the recruitment of p65 to IBs is likely due to specific interactions with the 422 N and/or P proteins. Since RSV N and P are known to interact, yet the IB does not 423 form without both proteins being expressed together, more detailed characterisation 424 of this interaction is required to define the true binding partner, either N or P.

The sequestration of the NF-κB subunit p65 to cytoplasmic IBs is a conserved 426 mechanism of orthopneumovirus immunomodulation 427
Having established structural and functional similarity between bRSV and hRSV IBs, 428 we finally examined the regulation and sub-cellular localisation of the NF-κB subunit 429 p65 in hRSV infected cells. Beginning with the NF-κB luciferase reporter assay we 430 uncovered a pattern of signalling inhibition similar to bRSV. Infection with hRSV in the 431 presence of the NF-κB reporter did not lead to robust activation when compared to 432 mock infected cells, highlighting a lack of activation of this pathway in infected cells 433 ( Fig 7A, black bars). Again, similar to bRSV, infected 293T cells (24 h with hRSV) 434 which were stimulated for 6h with hTNFα induced significantly less NF-κB 435 transactivation, when compared to equivalently treated mock-infected cells (Fig 7A,  436 grey bars). This correlated well with an examination, by IF, of hRSV replication in Vero 437 cells, with and without hTNFα treatment, where again we did not observe significant 438 levels of p65 nuclear translocation ( Fig 7B). Indeed, as observed in bRSV infected 439 cells, p65 was recruited into intra-cytoplasmic puncta. These puncta were 440 subsequently shown to be synonymous with viral IBs (Fig 7C) in a set of experiments 441 which also confirmed that IB formation and the recruitment of p65 is host cell 442 independent. bRSV or hRSV infected MDBK (bovine) or Hep2 (human) cells 443 demonstrated the presence of p65-containing IBs in all scenarios, highlighting that the 444 mechanisms underpinning RSV IB formation, and the sequestration of p65 to these 445 bodies, are likely highly conserved (Fig 7C). We concluded this examination of host-446 range specificity with a more physiologically relevant model of the human bronchial 447 epithelium, BEAS-2B cells, which are derived from normal human tissues taken 448 following autopsy of non-cancerous individuals, identifying again the formation of IBs 449 and sequestration of p65, regardless of RSV species. Finally, we confirmed that IB-450 like structures formed by ectopic hRSV N and P co-expression recruited p65 to their 451 core ( Fig 7D). Taken together, these data indicate that the formation of IBs during viral 452 replication, together with the sequestration of the transcription factor NF-κB subunit 453 p65 to these bodies, is a common feature of orthopneumoviruses. 454

455
Recognition of viral pathogen-associated molecular patterns (PAMPs) by RIG-I or 456 MDA-5 can lead to activation of NF-κB transcription factors through the IKK complex 457 or IRFs through TBK-1/IKKε [9,11,48]. Activation of these innate responses is essential for inducing a robust adaptive response, firstly to clear viral infections and 459 secondly to elicit the establishment of a memory response [4,48]. However, in vivo 460 the various immune-evasion strategies employed by RSV combine to generate only a 461 short-lived response [4,19,20,23,48]. For instance, there is strong evidence that the 462 downregulation of key signalling molecules by the NS proteins suppresses IRF3 463 activation and type I IFN induction [17-20, 22, 23], although interestingly we did see 464 significant IRF3 nuclear translocation in our infected cells. As a key innate immune 465 pathway, NF-κB signalling is often a target for viral antagonism; however, to date RSV 466 modulation of its activation has remained less well defined. Although RSV lacking the 467 SH gene was shown to enhance NF-κB activation, the exact mechanisms employed 468 are unclear [14,15,49,50]. To address this, we monitored NF-κB p65 activation in 469 RSV-infected cells at multiple steps in the signalling pathway: IκBα degradation, p65 470 phosphorylation (at Ser536), p65 nuclear translocation, and more broadly NF-κB 471 transactivation. We present a novel mechanism of immune evasion wherein RSV 472 infection results in the sequestration of the NF-κB subunit p65 into viral inclusion 473 bodies (Fig 3B), a process which is independent of the known RSV immunomodulatory 474 proteins, NS1, NS2 and SH (Fig 3F). We also demonstrate that as a result, activation 475 of NF-κB p65 is suppressed in infected cells, even with exogenous TNFα stimulation 476 (Fig 1). Although small IBs were observed as early as 6 h p.i. (≤2.5 µm 2 ) these did not 477 colocalise with detectable levels of p65 (Fig 3E). This may reflect a technical limitation 478 of our IF, or alternatively that IBs need to grow in size before they can begin to 479 sequester p65. It remains to be determined if p65 is actively recruited to IBs by viral 480 proteins or if its sequestration is a result of the IB's position in the cell and that it 481 captures p65 by an indirect mechanism, perhaps involving trafficking. Interestingly, the 482 lack of p65 activation prior to IB formation and p65 aggregation, highlights that RSV 483 may employ additional mechanisms for NF-κB inhibition which remain 484 uncharacterised. From a wider perspective, this mechanism of immunomodulation 485 might be a common strategy utilised by RSV and other viruses that induce IB 486 formation. MAVS and MDA5 were similarly both found to be recruited into RSV IBs as 487 a mechanism of suppressing IFN signalling [24]. Similarly, p38 MAPK and OGT 488 sequestration into RSV IBs suppressed MAPK-activated protein kinase 2 signalling 489 and stress granule formation, respectively, enhancing virus replication [25]. Whether 490 viruses such as Ebola, Nipah or rabies adopt similar mechanisms of 491 immunomodulation remains to be determined.
From a mechanistic perspective our results also showed that the N and P proteins are 493 essential for the formation of bRSV IBs. As has been reported for rabies [28] and 494 measles [30] viruses, ectopic expression of these proteins resulted in the formation of 495 IB-like structures (Fig 6A and 7D). These were mostly spherical and at 24 h post 496 transfection, measured up to 6.9 µm 2 which is considerably less than the conventional 497 IBs observed in infected cells. We hypothesise that both pseudo-IBs and viral IBs form 498 by biomolecular condensation, but that their maturation into larger structures is 499 dependent on other factors present only in infected cells. That these pseudo-IBs could 500 also recruit p65 suggested a direct interaction between p65 and RSV N or P, which 501 we confirmed by co-IP (Fig 6B). Interestingly, our IF data was somewhat contradictory, 502 with the staining patterns and line intensity profiles showing p65 concentrated in the 503 middle of IBs with N and P at the periphery, separating the IB contents from the 504 cytoplasm. It is possible that exchange of biomolecules across the boundary, e.g. 505 during the sequestration of p65, may require transient N or P interactions. Intriguingly, 506 Lifland et al. also suggested MAVS and MDA-5 are recruited into IBs by interacting 507 with N and P in a macromolecular complex [24]. We propose that this recruitment may 508 involve low-affinity interactions with N and/or P and that maintenance within the IB is 509 enhanced by the same physicochemical properties of the IBs which enable them to 510 induce LLPS, namely macromolecular-macromolecular interactions. The RSV P 511 protein has been shown to bind and recruit M2-1 to IBs, potentially through intrinsically 512 disordered regions within P that allow it to form multiple interactions [51]. Although 513 further work is required to identify the exact mechanism of p65, MAVS, MDA5 etc. 514 recruitment into IBs, we postulate the physicochemical properties of these proteins 515 may also be an important factor. 516 Electron micrograph analysis of our RSV IBs showed greater electron density in the 517 IBs, when compared to the cytoplasm, a characteristic of biomolecular condensates 518 ( Fig 5A). These data also highlighted the structural complexity of the phase-separated 519 structure. Although we observed some association with the ER and RER, RSV IBs 520 were not membrane bound, unlike rabies virus negri-bodies which acquire a 521 membrane boundary later in infection, presumably derived from the ER [28,39]. 522 Interestingly, our CLEM analysis confirmed previous IF data from the field that the IB 523 boundary is surrounded by N protein (Fig 5D). A debate remains in the field as to 524 whether this is an artefact of disrupted antibody epitope accessibility to N, since GFP 525 tagged N proteins were shown to have a diffuse pattern throughout the IB [24]; 526 however, we would only note that we used an antibody developed in-house for this 527 staining. Nevertheless, the presence of viral RNA associated proteins, N, P and M2-528 1, in IBs (Fig 3B and 4C) strongly suggested the presence of RNA replication and 529 transcription within these structures. Building on previous work for hRSV and rabies 530 virus [26,39] , we used 5EU incorporation to confirm RNA synthesis in the IBs (Fig 4A  531 and B). Using fluorescence in situ hybridization (FISH) experiments, Rincheval et al. 532 showed that genomic RNA colocalised with the hRSV N and P proteins at the 533 periphery, whilst viral mRNA was found to concentrate in IBAGs, transient sites of 534 mRNA storage [26]. Our data showed the formation of similar structures, confirming 535 IBAGs are found in multiple orthopneumoviruses; however, there was no conclusive 536 colocalisation with p65. However, this sequestered cellular protein did localise to 537 distinct intra-IB bodies (Fig 4B and C), raising the intriguing possibility that multiple 538 microdomains exist within what is, by TEM, an apparently uniform granular 539 biomolecular condensate. 540 In summary our data shows that RSV IBs are highly ordered structures performing 541 multiple roles in the virus lifecycle including the compartmentalisation of virus 542 replication and transcription and the sequestration of cellular proteins involved in the 543 antiviral response. This mechanistic characterisation is potentially applicable to other 544 negative sense RNA viruses that have been shown to form IBs during replication. 545 Wild-type recombinant (r) bRSV and deletion mutant rbRSVs ∆SH, ∆NS1, ∆NS2, and 555 ∆NS1/2 were produced by reverse genetics from rbRSV strain A51908 variant 556 Atue51908 (GenBank accession no. AF092942) [18,40,52]. These were propagated in Vero cells and hRSV subtype A (A2 strain) grown in Hep-2 cells. All viruses were 558 further purified from total cell lysates using polyethylene glycol (molecular weight, 559 8,000) precipitation and discontinuous sucrose gradient centrifugation. 560

Materials and Methods
Plasmids and transfections. All viral gene sequences were derived from bRSV 561 A51908 (GenBank accession NC_038272) and hRSV A2 (GenBank accession 562 KT992094). Expression plasmids (pcDNA3.1) encoding codon-optimised N genes at 563 KpnI-BamHI sites referred to as pN were purchased from Bio Basic Inc. Full length P 564 genes were amplified by reverse transcriptase PCR using gene-specific primers and 565 Superscript II reverse transcriptase (Invitrogen). These were then cloned into 566 pcDNA3.1 at KpnI-BamHI sites and designated pP. The p65 open reading frame 567 (ORF) was amplified from pcDNA3.1-HA-p65 (kindly provided by Carlos Maluquer de 568 Motes, Uni. Of Surrey) and inserted at the HindIII-BamHI sites of pcDNA3.1; 569 designated pP65. All sequences were confirmed by conventional sanger sequencing. 570 Plasmids were transfected into cells using TransIT-X2 (Geneflow). 571 Antibodies and drugs. Mouse monoclonal antibodies raised against bRSV F 572 (mAb19), N (mAb89), P (mAb12), M (mAb105) and M2-1 (mAb91) were previously 573 described [53,54]. Rabbit polyclonal anti-bRSV SH antibody was purchased from 574 Ingenasa. Rabbit anti-NF-kB p65 (8242)  Fluorescence was imaged on a Leica TCS SP5 confocal microscope using 405nm, corresponding horseradish peroxidase-conjugated secondary antibodies (CST). 622 Protein bands were detected using Clarity Western ECL substrate (Bio-Rad) and 623 imaged with Bio-Rad ChemiDoc TM MP Imaging System. 624 5-ethynyl uridine (5EU) labelling. Infected cells growing on coverslips were 625 incubated with or without medium supplemented with 20 µg/ml actinomycin D (Act D) 626 to inhibit cellular transcription for 1 h. Cells were then incubated with medium 627 containing 1 mM 5EU and 20 µg/ml Act D for another hour. Medium was then washed 628 off and cells fixed in 4% PFA for 15 mins. Cells were then washed with PBS and 629 permeabilized with 0.2% Triton X-100 for 5 mins. These were both supplemented with 630 0.125 U/ml RNase inhibitor (Promega). Incorporated 5EU was labelled using the Click-631 IT RNA Imaging Kit (Invitrogen) following the manufacturer's protocol. Following that, 632 immunofluorescence staining was done as described above. 633 TEM. Cells seeded onto Thermanox coverslips (Thermo Scientific) were fixed at 24 h 634 and 48 h p.i in phosphate buffered 2% glutaraldehyde (Agar Scientific) for 1 hour 635 followed by 1 hour in aqueous 1% osmium tetroxide (Agar Scientific). Following 636 dehydration in an ethanol series; 70% for 30 min, 90% for 15 min and 100% three 637 times for 10 min, a transitional step of 10 min in propylene oxide (Agar Scientific) was 638 undertaken before infiltration with 50:50 mix of propylene oxide and epoxy resin (Agar 639 Scientific) for 1 hour. After a final infiltration of 100% epoxy resin for 1 hour, the 640 samples were embedded and polymerised overnight at 60°C. 80µm thin sections were 641 cut, collected onto copper grids (Agar Scientific) and grid stained using Leica EM AC20 642 before being imaged at 100kV in a FEI Tecnai 12 TEM with a TVIPS F214 digital 643 camera. 644 CLEM. Cells seeded onto gridded glass coverslips (MatTek) were fixed at 24 h and 645 48 h p.i in 4% PFA (Sigma) and labelled according to the described 646 immunofluorescence method. Selected grid squares were imaged on a Leica TCS 647 SP8 confocal using 405nm, 488nm and 568nm laser lines for the appropriate dyes. 648 The cells were then fixed in phosphate buffered 2% glutaraldehyde (Agar Scientific) 649 for 1 hour followed by 1 hour in aqueous 1% osmium tetroxide (Agar Scientific). 650 Following 15min in 3% uranyl acetate (Agar Scientific), the cells were dehydrated in 651 an ethanol series; 70% for 30 min, 90% for 15 min and 100% three times for 10 min. 652 After infiltration of 100% epoxy resin for 2 hours, the samples were embedded and polymerised overnight at 60°C. The glass coverslips were removed with liquid nitrogen 654 and the appropriate grid squares located. 80µm thin sections were cut, collected onto 655 copper grids (Agar Scientific) and grid stained using Leica EM AC20. The specific cells 656 imaged in the confocal were identified and imaged at 100kV in a FEI Tecnai 12 TEM 657 with a TVIPS F214 digital camera. 658 Co-Immunoprecipitation. 1 x10 5 293T cells cultured overnight in 12-well plates were 659 transfected with pcDNA3.1-empty vector (pEV) or pcDNA3.1-p65 (pP65) using 660 TransIT-X2 (Geneflow). 24 h later, cells were infected with bRSV at MOI 1 or mock 661 infected and incubated for another 24 h. Cells were then lysed on ice with RIPA lysis 662 buffer (EMB Millipore) and cell debris removed by centrifugation. Cell lysates pre-663 cleared with protein A coated magnetic beads (CST) were incubated with rabbit anti-664 p65 antibodies overnight at 4 o C. Lysates were then incubated with protein A coated 665 magnetic beads for 20 mins at room temperature with rotation. Following five washes 666 with PBS-T, immunoprecipitates were eluted with Laemmli sample buffer and 667 subjected to SDS-PAGE and western blot analysis as already described. were infected with bRSV or mock infected. At 24 h p.i., cells were treated with 500 µM 723 Sodium arsenite or mock treated for 1 hr. Cells were then fixed and immuno-stained 724 with anti-G3BP1 (green) and anti-NF-κB p65 (red) antibodies. Nuclei were stained with 725 DAPI (blue) and images obtained using a Leica TCS SP5 confocal microscope. 726  fixed. 5EU incorporated into newly synthesised RNA was detected using Alexa Fluor 754 488-azide (green) as described in the methods. Cells were then immuno-stained with 755  were mock infected or infected with hRSV at an MOI of 1. At 6 h p.i., cells were 800 transfected with 100 ng NF-κB FLuc reporter and 10 ng TK-renilla luciferase and 801