Human coronaviruses disassemble processing bodies

A dysregulated proinflammatory cytokine response is characteristic of severe coronavirus infections caused by SARS-CoV-2, yet our understanding of the underlying mechanism responsible for this imbalanced immune response remains incomplete. Processing bodies (PBs) are cytoplasmic membraneless ribonucleoprotein granules that control innate immune responses by mediating the constitutive decay or suppression of mRNA transcripts, including many that encode proinflammatory cytokines. PB formation promotes turnover or suppression of cytokine RNAs, whereas PB disassembly corresponds with the increased stability and/or translation of these cytokine RNAs. Many viruses cause PB disassembly, an event that can be viewed as a switch that rapidly relieves cytokine RNA repression and permits the infected cell to respond to viral infection. Prior to this report, no information was known about how human coronaviruses (hu CoVs) impacted PBs. Here, we show SARS-CoV-2 and the common cold hu CoVs, OC43 and 229E, induced PB loss. We screened a SARS-CoV-2 gene library and identified that expression of the viral nucleocapsid (N) protein from SARS-CoV-2 was sufficient to mediate PB disassembly. RNA fluorescent in situ hybridization revealed that N protein-mediated PB loss correlated with elevated RNA for PB-localized transcripts encoding TNF and IL-6. Ectopic expression of the N proteins from five other human coronaviruses (OC43, MERS, 229E, NL63 and SARS-CoV-1) did not cause significant PB disassembly, suggesting that this feature is unique to SARS-CoV-2 N protein. These data suggest that SARS-CoV-2-mediated PB disassembly contributes to enhanced proinflammatory cytokine production observed during severe SARS-CoV-2 infection.


142
Infection with human coronaviruses causes PB loss 143 Endothelial cells (ECs) have emerged as playing a significant role in severe COVID; as sentinel 144 immune cells they are important sources for many of the cytokines elevated in severe disease and 145 are infected by SARS-CoV-2 in vivo [56,[58][59][60]88]. However, others have shown that 146 commercial primary human umbilical vein endothelial cells (HUVECs) require ectopic 147 expression of the viral receptor, ACE2, to be susceptible to . We recapitulated 148 those findings and showed that after HUVECs were transduced with an ACE2-expressing 149 lentivirus (HUVEC ACE2 ), they were permissive for SARS-CoV-2 (Wuhan-like ancestral Toronto 150 isolate; TO-1) [90] (Fig S1A). To use HUVEC ACE2 for studies on PB dynamics, we confirmed 151 that ACE2 ectopic expression had no effect on PB number in HUVECs (Fig S1B). Confirming 152 this, we infected HUVEC ACE2 with SARS-CoV-2 (MOI=3) to determine if PBs were altered. 153 SARS-CoV-2 infected cells were identified by immunostaining for the viral nucleocapsid (N) 154 while PBs were identified by immunostaining for two different PB resident proteins, the RNA 155 helicase DDX6, and the decapping cofactor, Hedls. PBs, measured by staining for both markers, 156 were absent in most SARS-CoV-2 infected HUVECs ACE2 by 24 hours post infection (Fig 1A-D). 157 We quantified the loss of cytoplasmic puncta using a method described previously [91] and 158 showed that by 24 hours post infection, SARS-CoV-2 infected cells displayed a significant 159 reduction in PBs compared to mock-infected controls (Fig 1B, D). 160 161 To confirm that PBs were reduced by SARS-CoV-2 infection of naturally permissive cells 162 derived from respiratory epithelium, we infected Calu-3 cells with SARS-CoV-2. Infected cells 163 were identified by immunostaining for N protein 24 hours after infection and PBs were stained 164 for DDX6 and Hedls. We observed PB loss in most but not all infected cells (Fig 1E-H). We 165 quantified PB loss and showed that by 24 hours post infection, SARS-CoV-2 infected Calu-3 166 cells also displayed a significant reduction in PBs compared to mock-infected controls (Fig F,167 H). As the SARS-CoV-2 pandemic has progressed, new variants of concern (VOCs) have 168 continued to emerge [92,93]. To determine if VOCs also induced PB disassembly, we infected 169 HUVEC ACE2 with SARS-CoV-2 VOCs Alpha, Beta, Gamma and Delta (MOI=2) and compared 170 PB disassembly to an ancestral isolate of SARS-CoV-2 (Wuhan-like Toronto isolate; TO-1) (Fig  171   2). Infected cells were identified by immunostaining for N protein 24 hours after infection and 172 PBs were identified using Hedls. Significant PB disassembly was observed for all VOCs, 173 although we noted slightly less PB disassembly mediated by TO-1 compared to the experiments 174 in Fig 1, likely due to the decrease in MOI between these two experiments (Fig 2). 175

176
To determine if PBs were lost in response to infection with other human coronaviruses, we 177 established infection models for the Betacoronavirus, OC43, and the Alphacoronavirus, 229E. 178 We found HUVECs were permissive to both OC43 and 229E (Fig S2A-B). We then performed a 179 time-course experiment wherein OC43-infected HUVECs were fixed at various times post 180 infection and immunostained for the viral N protein and the PB-resident protein DDX6. We 181 observed that PBs were largely absent in OC43 N protein-positive cells but present in mock-182 infected control cells (Fig 3A-B). 229E-infected HUVECs were stained for DDX6 to measure 183 PBs and for dsRNA to denote infected cells due to a lack of commercially available antibodies 184 for 229E. CoV infected cells are known to form an abundance of dsRNA due to viral replication 185 and transcription from a positive-sense RNA genome making this a suitable marker for virally 186 infected cells [94] . In parallel, we confirmed 229E infection by performing RT-qPCR for viral 187 genomic and subgenomic RNA (Fig S2C-D). After 229E infection, we also found that PBs were 188 significantly reduced (Fig 3C-D). Because of antibody incompatibility, we were unable to co-189 stain infected cells for the PB protein Hedls and OC43 N protein or dsRNA. In lieu of this, we 190 performed additional OC43 and 229E infections and co-stained infected cells using antibodies 191 for Hedls and DDX6 (Fig S3). We then performed additional quantification of PB loss using the 192 Hedls marker to label PBs. These data also show robust PB loss in response to infection with 193 OC43 and 229E (Fig S3). 194 195 PBs will disassemble if key scaffolding proteins are lost; these include the RNA helicase DDX6,196 the translation suppressor 4E-T, the decapping cofactors Hedls/EDC4 and DCP1A, and the 197 scaffolding molecule Lsm14A [95] . To elucidate if CoV-infected cells displayed decreased 198 steady-state levels of PB resident proteins, we immunoblotted infected cell lysates for PB 199 proteins XRN1, DCP1A, or DDX6, and Hedls (Fig 4) and quantified protein levels by 200 densitometry (Fig S4). SARS-CoV-2 infection of HUVEC ACE2 cells did not alter steady-state 201 levels of these proteins compared to uninfected cells ( Fig 4A). OC43-infected HUVECs 202 displayed comparable levels of XRN1, DCP1A, and Hedls relative to uninfected cells; however, 203 OC43 infection decreased steady-state levels of DDX6 at both 12 hpi but not significantly at 24 204 hpi ( Fig 4B). 229E-infected HUVECs showed no detectible change in PB protein expression 205 after infection compared to controls (Fig 4C). 206 207 PBs are important sites for the post-transcriptional control of inflammatory cytokine transcripts 208 containing AU-rich elements, and PB loss correlates with enhanced levels of some of these 209 transcripts [7,[12][13][14][15]. To determine if ARE-mRNAs are elevated, and therefore subject to 210 expressors (nsp6, nsp11) from further studies and proceeded with validation of the top four hits 257 (ORF7b,N,ORF3b,nsp1). 258

259
The nucleocapsid protein of SARS-CoV-2 induces PB disassembly 260 We tested four top hits from our PB screen in more relevant endothelial cells as these cells can 261 be infected, express inflammatory cytokines, and stain robustly for PBs. HUVECs were 262 transduced with recombinant lentiviruses expressing N, nsp1, ORF3b, and ORF7b or empty 263 vector control lentiviruses. We also included recombinant lentiviruses expressing nsp14 in this 264 experiment because of its exoribonuclease activity and ability to diminish cellular translation and 265 interferon responses [84]. Transduced cells were selected for transgene expression and then fixed 266 and stained for the endogenous PB marker protein DDX6 and for the Strep-tag II on each of 267 SARS-CoV-2 constructs. We observed robust staining of the viral nucleocapsid (N) protein in 268 the transduced cell population (Fig 7A) but were unable to detect expression of nsp1, nsp14, 269 ORF3b or ORF7b by immunostaining ( Fig S6A). We quantified PB loss in the selected cells and 270 observed decreased PB numbers in cell populations expressing N, nsp14, ORF3b and ORF7b; 271 however, the most robust PB loss was induced in N-expressing cells, which displayed a 272 significant reduction in PB numbers as well as strong immunostaining ( Fig 7B, Fig S6). We were 273 concerned that we could not detect the other four transgenes by immunostaining; therefore, we 274 performed immunoblotting for the Strep-II tag on lysates from each transduced cell population 275 ( Fig S6C). Although we detected a strong band of ~50 kDa at the predicted molecular weight for 276 N, we were unable to detect bands for nsp14, ORF3b and ORF7b, while the most prominent 277 band for nsp1 did not migrate at the predicted molecular weight of ~20 kDa ( Fig S6)  We showed that human CoVs OC43 and 229E also cause PB loss during infection (Fig 3); 294 therefore, we were interested to determine if ectopic expression of nucleocapsid proteins from 295 these or other human CoVs were sufficient to mediate PB disassembly. To test this, we 296 transduced HUVECs with recombinant lentiviruses expressing the N protein from SARS-CoV-2 297 as well as N derived from Betacoronaviruses, MERS-CoV N-Flag and OC43 N. Expression of 298 MERS-CoV and OC43 N proteins did not lead to significant PB loss compared to SARS-CoV-2-299 N (Fig 8A-B). We also tested two N proteins from human Alphacoronaviruses, 229E N-Flag and 300 NL63 N-Flag. 229E N protein failed to induce significant PB loss compared to SARS-CoV-2 N, 301 while unexpectedly, it appeared the expression of NL63 N protein increased PB numbers . The significance of this increase is not yet clear. We also expressed the N protein from 303 the more closely related Betacoronavirus, SARS-CoV-1, and observed SARS-CoV-1 N protein 304 caused moderate PB loss, yet SARS-CoV-1 N protein-induced PB disassembly was not deemed 305 significant by statistical analysis (Fig 8E-F). We collected protein lysates in parallel and 306 performed immunoblotting to detect each CoV N protein (Fig 8G-I). We noted that CoV N 307 proteins are not expressed equally, thus, we cannot fully discount expression level as the reason 308 for the discrepancy in PB disassembly. This may be especially important for our analysis of 309 SARS-CoV-1 N protein, which showed an intermediate but non-significant PB disassembly 310 phenotype but was expressed at a lower level that SARS-CoV-2 N protein ( Fig 8I). 311 Immunoblotting of steady state levels of PB resident proteins after N protein overexpression 312 showed that most PB proteins tested remained unchanged in the context of ectopic expression 313 ( Fig 8G-I, Fig S7A). The exception to this was the decapping factor, Hedls/EDC4, which was 314 decreased after expression of SARS-CoV-2 N in one set of experiments but not in others, but 315 increased after expression of N proteins from MERS and OC43 (Fig 8G-I, Fig S7A). The 316 significance of this observation is also unclear. 317

318
To understand if PB disassembly correlates with changes to PB-regulated inflammatory cytokine 319 transcripts in our system, we performed immunofluorescence-RNA fluorescent in situ 320 hybridization (IF-FISH) to confirm the localization of inflammatory RNAs to PB foci. RNA 321 FISH was performed for two PB-regulated cytokine transcripts that contain AREs, those 322 encoding IL-6 and TNF, and for the GAPDH RNA, which does not contain AREs and is not 323 expected to localize to PBs [3]. To achieve a better signal-to-noise ratio for detection, we used 324 TNF to induce their transcription of IL-6 and TNF, which are lowly expressed in untreated cells. 325 First, we confirmed that TNF treatment alone did not significantly alter PB dynamics ( Fig S8). 326 We then stained PBs using our antibody to Hedls and co-stained with probes that bound 327 GAPDH, TNF and IL-6 RNA transcripts. We repeatedly observed co-localization of both IL-6 328 and TNF RNA with PBs; in contrast, we observed extremely limited GAPDH co-localization 329 with PBs ( Fig 9A). IL-6 and TNF transcripts were also present at much lower levels than 330 GAPDH RNA, consistent with the notion that ARE-containing transcripts are kept low by tight 331 transcriptional control and constitutive decay in PBs [14]. We then stained N protein-transduced 332 cells using our probes ( Fig 9B). As expected, N protein expression induced loss of Hedls-333 positive PB foci (not shown). In N-expressing cells, the signal for IL-6 and TNF RNA 334 redistributed from PB foci to the cytoplasm, and the probe intensity for both IL-6 and TNF was 335 markedly increased compared to control cells which lacked N protein ( Fig 9B). We quantified 336 the change in FISH probe signal intensity per cell ( Fig 9C). Compared to controls, N protein-337 expressing cells displayed strongly increased TNF probe signal intensity and increased IL-6 338 probe signal intensity, whereas GAPDH probe signal intensity was not increased (Fig 9C). These 339 data support our hypothesis that PBs directly regulate proinflammatory cytokine transcripts and 340 that these PB-regulated mRNAs increase upon PB disassembly. 341

342
We then asked if N protein from SARS-CoV-2 or other human CoVs would increase ARE-RNA 343 transcript levels by RT-qPCR. We analyzed IL-6 and TNF transcript level, as well as two other 344 ARE-containing RNAs, encoding CXCL8 and COX-2 which were elevated after infection with 345 SARS-CoV-2 and OC43 (Fig 5). In uninduced ECs, the transcription of these mRNAs is 346 minimal; for example, TNF mRNA could not be readily detected by RT-qPCR without 347 transcriptional activation. Therefore, we treated control and N-expressing HUVECs with TNF to 348 activate cytokine transcription and then assessed if N protein expression enhanced cytokine 349 mRNA level post-transcriptionally. In the absence of TNF, no change of mRNA abundance was 350 observed for any of the coronaviruses N proteins tested (Fig 10A-C, 0 hour no treatment). 351 Ectopic expression of SARS-CoV-2 N enhanced transcript levels of IL-6, CXCL8, and TNF 24 352 hours after transcription was induced; however, this enhancement was not deemed significant by 353 statistical analysis (Fig 10A-C). In contrast, we did not observe any increase in transcript levels 354 after expression of N protein from MERS-CoV, OC43, 229E, or NL63 (Fig 10A-C In this manuscript, we present the first evidence to show that human CoVs, including SARS-365 CoV-2, induce PB loss after infection. PBs are fundamental sites of post-transcriptional control 366 of gene expression and are particularly relevant to the regulation of cytokine production. Our 367 major findings are as follows: i) Three human coronaviruses, SARS-CoV-2, OC43, and 229E 368 induced PB loss. ii) The SARS-CoV-2 nucleocapsid (N) protein was sufficient to cause PB loss. identified eight candidate genes that reduced PB numbers (Fig 6). Validation of a subset of these 377 in HUVECs revealed that the most robust and consistent mediator of PB loss was the SARS-378 CoV-2 viral N protein (Fig 7). The N protein is the most abundant produced protein during CoV 379 replication [102]. The SARS-CoV-2 N protein is 419 amino acids long and has two globular and 380 three intrinsically disordered protein domains, including a central disordered serine-arginine 381 (SR-rich) linker region [102][103][104]. The N protein is a multifunctional RNA-binding protein 382 (RBP) essential for viral replication; it coats the viral genome and promotes viral particle 383 assembly [46,[104][105][106][107]. Several recent reports have shown that N protein undergoes liquid-liquid 384 phase separation with RNA, an event which may be regulated by phosphorylation of multiple 385 serine residues in the SR-region and is an important feature of viral assembly [107][108][109][110][111][112] . The N 386 protein is also an important modulator of antiviral responses [85,113,114]. A recent study 387 showed that low doses of N protein supressed the production of IFN and some PB-regulated 388 inflammatory cytokines, while high doses of N protein promoted their production [85] . These 389 observations are consistent with our phenotype of PB disassembly, which correlates with later 390 infection times, high expression of N protein and immunofluorescent staining throughout the 391 cytoplasm (Fig 1, Fig 7). 392

393
We subsequently screened five other coronavirus N proteins from OC43, MERS, 229E, NL63, 394 and SARS-CoV-1 discovered that the phenotype of N-mediated PB disassembly was not 395 conserved among N proteins but was unique to SARS-CoV-2-N, and perhaps SARS-CoV-1 N 396 protein which appeared to disassemble PBs in two of three independent experiments (Fig 8). 397 Despite conservation of structural motifs, the N proteins between most human CoVs possess low 398 sequence conservation at the amino acid level (~25-50%) and have been reported to exhibit 399 different properties [115] . In contrast, SARS-CoV and SARS-CoV-2 N proteins have ~94% 400 amino acid identity [116], making it difficult to reconcile our observation that PB disassembly 401 induced by SARS-CoV-2 N protein was significant, while PB disassembly induced by SARS-402 CoV-1 N protein was not (Fig 8). One difference that we observed by immunoblotting was the 403 presence of a lower molecular weight ~37 kDa band recognized by our anti-N antibody for 404 SARS-CoV-2. We did not observe the 37kDa N product after OC43 infection, transduction with 405 OC43 N protein or transduction with C-terminally Flag-tagged N proteins from 229E, NL63, 406 MERS-CoV or SARS-CoV-1 (Fig 3, Fig 8). Steady state levels of the lower molecular weight N 407 product increased over the course of SARS-CoV-2 infection, consistent with the timing of PB 408 disassembly. Other groups have noted that the SARS-CoV-2 variant of concern (VOC), Alpha, 409 produces an additional subgenomic mRNA from which a truncated version of N, termed N*, can 410 be produced [68,117,118]. Translation of the N* ORF is predicted to start at an internal in-frame 411 methionine residue (Met210) within the N protein [117,118] was not observed on our immunoblots for SARS-CoV N protein, leading us to speculate that the 417 downstream Met residue may not be used for translation initiation in this case. Viruses capitalize 418 on downstream methionine residues to translate truncated protein products with subcellular 419 localization or functions that differ from their full-length counterparts as a clever way to increase 420 coding capacity [119,120] . Our ongoing investigation of the precise nature of the SARS-CoV-2 421 N protein truncation product we observe during infection and overexpression may reveal that it 422 has a specific role in PB disassembly. 423

424
The PB protein MOV10, and other components of RNA processing machinery, were revealed as 425 potential interactors with the N protein [121] ; however, we do not observe colocalization of N 426 protein with PBs after immunofluorescent staining of SARS-CoV-2 infected cells or N-427 expressing cells. Based on our data, we consider two possible mechanisms of N protein mediated 428 PB disassembly. First, N protein may mediate PB disassembly by phase separation with a PB 429 protein(s). This is similar to what has already been shown for N-mediated disruption of 430 cytoplasmic stress granules, important cytoplasmic biomolecular condensates that correlate with 431 cellular translational shutdown [35]. N protein localizes to stress granules and binds the essential 432 protein, G3BP1, preventing its interaction with other stress granule proteins and blocking stress 433 granule formation [122][123][124]. Although the precise domain required for this effect has been 434 debated, more than one report suggests that the N-terminal intrinsically disordered region is 435 required for stress granule disruption [123,125] . Second, a possible reason for PB loss may be 436 the indiscriminate binding of RNA by N protein. N protein could be acting as sponge for RNA, 437 pulling it out of cytoplasm, thereby reducing the RNA-protein interactions required for phase 438 separation of PBs [24,122]. We are currently engaged in site-directed and truncation mutagenesis 439 studies to determine the precise region(s) of SARS-CoV-2-N that is essential for PB 440 disassembly. 441

442
Prior to this report, little was known about CoVs and PBs, and the information that was 443 published was contradictory. Infection with murine hepatitis virus (MHV) was reported to 444 increase PBs, whereas transmissible gastroenteritis coronavirus (TGEV) decreased PBs [86,87]. 445 Since the initiation of our study, one additional publication used ectopic expression of one of the 446 SARS-CoV-2 CoV proteases, nsp5, to test if it was capable of PB disassembly. Consistent with 447 the results of our screen (Fig 6), nsp5 did not mediate PB loss [32]. In this manuscript, we now 448 confirm that SARS-CoV-2, OC43, and 229E induce PB disassembly (Fig 1-3 One possibility is that PBs are antiviral because their proteins help the cell respond to signals that 466 activate innate immune pathways [28,30,129,130]. In support of this, TRAF6 was shown to 467 control Dcp1a localization to PBs using ubiquitylation, suggesting that antiviral signaling is 468 more complex than previously appreciated and integrates transcriptional responses with cytokine 469 mRNA suppression in PBs [129,131] . Moreover, the PB protein Lsm14A has been shown to 470 bind to viral RNA/DNA after infection-induced PB disassembly to promote IRF3 activation and 471 . Although it remains unclear if the higher order condensation of many 472 proteins into PBs is required for their proposed antiviral activity, what is clear is that the 473 outcome of PB disassembly is a reversal of the constitutive decay or translational suppression of 474 cytokine mRNA that would normally occur there [7,[12][13][14][15]132] . Our data support a role for PBs 475 as sensors of virus infection that release cytokine transcripts from repression when they 476 disassemble. Using IF-FISH, we observed that RNAs encoding IL-6 and TNF, two molecules 477 elevated in severe SARS-CoV-2 disease, localized to PBs. Upon SARS-CoV-2 N-induced PB 478 disassembly, these transcripts re-localized from PB foci to the cytoplasm, a redistribution that 479 was also accompanied by an increase in FISH probe signal intensity signifying increase RNA 480 abundance (Fig 9). Although we did not see a statistically significant increase in IL-6 and TNF 481 RNA level in SARS-CoV-2 N-expressing cells using a population-based assay like RT-qPCR 482 (Fig 10), our single-cell analysis by IF-FISH suggests that the biological relevance of our 483 observation derives from the combination of increased transcript level abundance and the 484 redistribution of cytokine transcripts from PB foci to the cytoplasm. We speculate that viral 485 infection causes PB loss and this event is viewed as a danger signal by the cell: it relieves 486 cytokine mRNA suppression and re-localizes these transcripts, returning them to the cytoplasmic 487 pool of RNA to permit translation of proinflammatory cytokines that then act as a call for 488 reinforcements. In this way, PB disassembly is connected to the innate immune response and is 489 one of many signals that notify the immune system that a cell is infected. In situations where 490 interferon responses are delayed or defective, as is emerging for SARS-CoV-2 and severe 491 , PB disassembly may occur to alert the immune system of an infection, and 492 may be a contributing factor to pathogenic cytokine responses. 493 494 In summary, our work adds to a growing body of literature which describes that many viruses 495 target PBs for disassembly, supporting the idea that PBs restrict viral infection. We showed that 496 the N protein of SARS-CoV-2 is sufficient for PB disassembly, and this phenotype correlated 497 with elevated levels of PB-localized cytokine transcripts encoding IL-6 and TNF, both which are 498 elevated during SARS-CoV2 infection [53]. Not only does this work describe a previously 499 uncharacterized cellular event that occurs during CoV infection, but we have identified a novel 500 mechanism which may contribute to the dysregulated cytokine response exhibited by severe 501 SinoBiological (cat #VG40068-CF) using BamHI and EcoRI (Table 1). Codon-optimized 529 pLJM1-SARS-CoV-N-FLAG was cloned SinoBiological (cat# VG40588-NF) using BamHI and 530

Transient Transfections 533
Transient transfections were performed using Fugene (Promega) according to manufacturer's 534 guidelines. Briefly, HeLa Flp-In TREx GFP-Dcp1a cells were seeded in 12-well plates at 535 150,000 cells/well in antibiotic-free DMEM. Cells were transfected with 1 μg of DNA and 3 μL 536 of Fugene for 48 hours before processing. 537

Production and use of Recombinant Lentiviruses 538
All recombinant lentiviruses were generated using a second-generation system. HEK293T cells 539 were transfected with psPAX2, MD2-G, and the lentiviral transfer plasmid containing a gene of 540 interest using polyethylenimine (PEI, Polysciences). 6 hours after transfection, serum-free media 541 was replaced with DMEM containing serum but no antibiotics. Viral supernatants were 542 harvested 48 hours post-transfection and frozen at -80°C until use. For transduction, lentiviruses 543 were thawed at 37°C and added to target cells in complete media containing 5 µg/mL polybrene 544 (Sigma). After 24 hours, the media was replaced with selection media containing 1 µg/mL 545 puromycin or 5 µg/mL blasticidin (ThermoFisher) and cells were selected for 48 h before 546 proceeding with experiments. 547 Immunofluorescence 548 Cells were seeded onto 18mm round, #1.5 coverslips (Electron Microscopy Sciences) for 549 immunofluorescence experiments. Following treatment, cells were fixed for 10 or 30 (if infected 550 with SARS-CoV-2) min in 4% (v/v) paraformaldehyde (Electron Microscopy Sciences). 551 Samples were permeabilized with 0.1% (v/v) Triton X-100 (Sigma-Aldrich) for 10 min at room 552 temperature and blocked in 1% human AB serum (Sigma-Aldrich) in 1X PBS 1 h at room 553 temperature. Primary and secondary antibodies (Table 2) were diluted in 1% human AB serum 554 and used at the concentrations in Table 2. Nuclei were stained with 1 µg/ml Hoechst 555 (Invitrogen). Samples were mounted with Prolong Gold AntiFade mounting media 556 (ThermoFisher). Images were captured using a Zeiss AxioObserver Z1 microscope with the 40X 557 oil-emersion objective unless otherwise stated in the respective figure caption. To account for 558 variability in staining, all experiments contained internal controls (negative control; mock or 559 EV). During image acquisition, exposure time was kept consistent within an independent 560 experiment. 561

IF-FISH was performed according to manufacturers protocol (Stellaris IF-FISH protocol). 563
Briefly, cells were fixed in 4% formaldehyde (Sigma-Aldrich) for 10 mins then permeabilized 564 with 0.1% Triton X-100 (Sigma-Aldrich) in PBS for 10 mins. Primary Hedls antibody was 565 diluted in PBS and incubated for 1 h at room temp, following which secondary antibody was 566 likewise diluted in PBS and allowed to incubate for 1 h at room temp. Cells were again fixed in 567 4% formaldehyde for 10 mins, washed with 1X PBS, and incubated in Wash Buffer A (LGC 568 Biosearch) for 5 minutes. Stellaris probes; GAPDH Quasar 670 (VSMF 2151-5), IL-6 Quasar 569 670 (LGC Biosearch, SMF 1065-5, custom order, probe sequences detailed in Table 4) and TNF 570 Quasar-670 (LGC Biosearch, SMF 1065-5, custom order, probe sequences detailed in Table 4). 571 were diluted in Hybridization Buffer (LGC Biocearch) to a concentration of 125 nM and cells 572 were incubated for hybridization at 37 °C overnight. The following day, cells were washed 573 sequentially with Wash Buffer A for 30 mins, Wash Buffer A containing 5 ng/mL DAPI 574 (Invitrogen) for 30 mins, and Wash Buffer B (LGC Biosearch) for 5 minutes. Samples were 575 mounted with Prolong Gold AntiFade mounting media (ThermoFisher). Images were captured 576 on a Zeiss LSM 880 confocal microscope with the 63X oil-emersion objective. At least three z-577 stacks were acquired per condition, maximal intensity projections (MIPs) are presented. 578 Immunoblotting 579 Cells were lysed in 2X Laemmli buffer and stored at -20°C until use. The DC Protein Assay 580 (Bio-Rad) was used to quantify protein concentration as per the manufacturer's instructions. 10-581 15 µg of protein lysate was resolved by SDS-PAGE on TGX Stain-Free acrylamide gels 582 (BioRad). Total protein images were acquired from the PVDF membranes after transfer on the 583 ChemiDoc Touch Imaging system (BioRad). Membranes were blocked in 5% BSA in TBS-T 584 (Tris-buffered saline 0.1% Tween-20). Primary and secondary antibodies were diluted in 2.5% 585 BSA, dilutions can be found in Table 2. Membranes were visualized using Clarity Western ECL 586 substrate and the ChemiDoc Touch Imaging system (BioRad). Densitometry was conducted 587 using ImageJ, area under the curve for each band was calculated, normalized to the respective 588 actin loading control band, and presented as fold change. 589

Quantitative PCR 590
RNA was collected using the RNeasy Plus Mini Kit (Qiagen) according to the manufacturer's 591 instructions and stored at -80°C until further use. RNA concentration was determined using 592 NanoDrop One C (ThermoFisher) and 500 ng of RNA was reverse transcribed using qScript XLT 593 cDNA SuperMix (QuantaBio) using a combination of random hexamer and oligo dT primers, 594 according to the manufacturer's instructions. Depending on starting concentration, cDNA was 595 diluted between 1:10 and 1:20 for qPCR experiments and SsoFast EvaGreen Mastermix (Biorad) 596 was used to amplify cDNA. The ∆∆quantitation cycle (Cq) method was used to determine the 597 fold change in expression of target transcripts using HPRT as a housekeeping control 598 gene. Variance in the mock or empty vector samples was calculated by dividing the ∆Ct value of 599 a single replicate by the average ∆Ct value of all replicates for that specific gene and condition. 600 qPCR primer sequences can be found in Table 3. (Table 3). 601

Virus Stocks and Virus Propagation 602
Experiments with SARS-CoV-2 and variants were conducted in a containment level-3 (CL3) 603 facility, and all standard operating procedures were approved by the CL3 Oversight Committee 604 and Biosafety Office at the University of Calgary. Stocks of SARS-CoV-2 Toronto-01 isolate 605 (SARS-CoV-2/SB3-TYAGNC) [90] , Alpha, Beta, Gamma and Delta were propagated in Vero 606 E6 cells. To produce viral stocks, Vero E6 cells were infected at an MOI of 0.01 for 1 hour in 607 serum-free DMEM at 37˚C. Following adsorption, DMEM supplemented with 2% heat 608 inactivated FBS and 100 U/mL penicillin/streptomycin/glutamine was added to the infected 609 wells. 24-60 hours post-infection (hpi), the supernatant was harvested and centrifuged at 500 x g 610 for 5 min to remove cellular debris. Virus stocks were aliquoted and stored at -80˚C for single 611 use. SARS-CoV-2 titres were enumerated using plaque assays on Vero E6 cells as previously 612 described [133] using equal parts 2.4% w/v semi-solid colloidal cellulose overlay (Sigma; 613 prepared in ddH2O) and 2X DMEM (Wisent) with 1% heat inactivated FBS and 1% PSQ. 614 Experiments with hCoV-OC43 (ATCC VR-1558) and hCoV-229E (ATCC VR-740) were 615 conducted in under containment level-2 conditions. hCoV-OC43 and hCoV-229E were 616 propagated in Vero E6 and MRC-5 cells, respectively. Cells were infected at an MOI of 0.01 for 617 1 hour in serum-free media at 33˚C. Following adsorption, the viral inoculum was removed and 618 replaced with fresh media supplemented with 2% heat inactivated FBS and 100 U/mL 619 penicillin/streptomycin/glutamine. After 5-6 days post infection (dpi), the supernatant was 620 harvested and cellular debris was cleared by centrifugation. Virus stocks were aliquoted and 621 stored at -80˚C. hCoV-OC43 and hCoV-229E titres were enumerated using Reed and Muench 622 tissue-culture infectious dose 50% (TCID50) in Vero E6 or MRC-5 cells, respectively. 623

Virus Infection 624
For experimental infections, cells were seeded into wells to achieve ~80% confluency at the time 625 of infection. The growth media was removed and replaced with 100 µL of viral inoculum diluted 626 in serum-free DMEM to reach the desired MOI and incubated at 37˚C for 1 hour, rocking the 627 plate every 10 min. Following incubation, the virus inoculum was removed and replaced with 1 628 mL of complete growth media. 629

Processing Body and FISH Probe Intensity Quantification 630
Processing bodies were quantified using an unbiased image analysis pipeline generated in the 631 freeware CellProfiler4.0.6 (cellprofiler.org) [134]. First, nuclear staining was used to identify 632 individual cells applying a binary threshold and executing primary object detection between 65 633 and 200 pixels for HUVECs/HUVEC ACE2 and 40 and 200 pixels for Calu3s. For each identified 634 object (nuclei), the peripheral boundary of each cell was defined using the "Propagation" 635 function. Propagation distance was customized depending on cell type to account for variance in 636 cell size (150 pixel radius from the nuclei for HUVECs/HUVEC ACE2 , 60 pixel radius from nuclei 637 for Calu3s). Using a subtractive function to remove the nuclei from the total cell area, the 638 cytoplasm of each cell was defined. The cytoplasm area mask was then applied to the matched 639 image stained for PB proteins (DDX6 or Hedls) to count only cytoplasmic puncta. Importantly, 640 multiple nuclei in close proximity (e.g., within the propagation distance) would be divided into 641 mutually exclusive cells, such that a single PB could not be counted more than once. Cells were 642 stratified into 'positive cells' (staining for viral proteins or dsRNA) or 'bystander cells'. In 643 control treatments (e.g., mock infected or EV) PBs in all cells were quantified. In treatment cells 644 only 'positive cells' were quantified unless otherwise indicated in corresponding figure captions. 645 Background staining was reduced using the "Enhance Speckles" function. Only DDX6 or Hedls-646 positive puncta with a defined size (3-13 pixels) and intensity range were quantified using 647 "global thresholding with robust background adjustments" function. All thresholds were 648 consistent within each replicate that used identical staining parameters. PBs were quantified per 649 cell, either control (EV or mock) or treatment (infected or viral protein expressing). For each 650 experiment, an equal number of control and treatment cells were analyzed. For Figure 5, puncta 651 counts were exported and RStudio was used for data analysis and PB numbers in treatment cells 652 were represented relative to control cells for ease of data interpretation, rather than raw PBs per 653 cell. For Figure 9, nuclei and cells were defined according to HUVEC boundaries as specified 654 above. Background of probe specific staining was reduced using a global thresholding strategy 655 with a minimum cross entropy method of execution. The MeasureImageIntensity function was 656 then used to identify the total intensity per image which was divided by the number of cells 657 identified within that image for a final readout of total intensity/cell. A minimum of 40 cells 658 were quantified per condition. 659

Statistics 660
All statistical analyses were performed using GraphPad Prism 9.0. 'Per cell' processing body 661 counts were plotted such that each independent biological replicate (including paired control and 662 treatment) could be visualized on a single graph. Given that per cell processing body counts are 663 naturally skewed and thus non-parametric, we elected to use rank-sum statistical methods 664 (Mann-Whitney U test and Kruskal Wallis test) when appropriate, as indicated in the 665 corresponding figure caption. In the case of Figure 7, a two-way ANOVA with a Tukey's post-666 hoc analysis was used to determine significance to (1) avoid the false discovery rate associated 667 with multiple T-tests and (2) because sodium arsenite-induced processing bodies appeared to be 668 parametrically distributed. Parametric distribution was assumed on all normalized data (Fig 5, 6      human ACE2 (HUVEC-ACE2), selected and infected with SARS-CoV-2 TO-1, alpha, beta, 737 gamma, or delta isolates (MOI=2), or a mock infection control. 24 hpi cells were fixed and 738 immunostained for SARS-CoV-2 N protein (green) and Hedls (white). Nuclei were stained with 739 Hoechst (blue). Representative images from one of two independent experiments are shown. B. 740 PBs were quantified as in Figure 1. These data represent two independent biological replicates 741 (n=2) with >80 cells measured per condition (mock and infected) per replicate. Each mock and 742 infected replicate pair plotted independently; mean. Statistics were performed using  Wallis H test with Dunn's correction (*, p < 0.0332; ****, p < 0.0001). Scale bar = 20 µM. 744  protein expressing lentiviruses were used as a positive control in each experiment. Cells were 826 selected, fixed and immunostained for DDX6 (PBs; white) and either authentic N protein or a 827 FLAG tag (green). Nuclei were stained with Hoechst (blue). Scale bar = 20 µm. DDX6 puncta in 828 EV or N-transduced cells were quantified using CellProfiler as in Figure 1. DDX6 puncta were 829 quantified as in Figure 1. Representative images from one independent experiment of three are 830 shown. These data represent three independent biological replicates (n=3) with >30 cells 831 measured per condition (EV and N) per replicate. Each EV and N replicate pair plotted 832 independently; mean. Statistics were performed using Kruskal-Wallis H test with Dunn's 833 correction (*, p < 0.0332; **, p < 0.0021; ****, p < 0.0001; ns, nonsignificant). G-I. HUVECs 834 were transduced as above, protein lysate was harvested and immunoblotting was performed 835 using XRN1, Hedls, DCP1A, DDX6, N protein or FLAG, and β-actin specific antibodies. One 836 representative experiment of three is shown. 837 Cells were fixed and immunostained for Hedls prior to hybridization with Stellaris probes 840 specific for GAPDH, IL-6 and TNF. Nuclei were stained with Dapi. Cells were imaged using a 841 Zeiss LSM 880 confocal microscope with Airyscan using the 63x objective. At least three Z-842 stacks were acquired per condition (probe) per replicate and a maximal intensity projection 843 (MIP) image is presented. One representative experiment of three is shown. Scale bar = 5 µm. B-844 C. HUVECs were transduced with recombinant lentiviruses ectopically expressing SARS-CoV-2 845 N or an EV control. Cells were selected, fixed and immunostained for Hedls prior to 846 hybridization with Stellaris probes specific for GAPDH, IL-6 and TNF. Nuclei were stained with 847 Dapi. Only probe-specific staining is displayed for simplicity. Scale bar = 20 μm. Representative 848 images from one of two independent experiments are shown. Total signal intensity per cell was 849 quantified using CellProfiler and is represented as fold-change relative to the intensity in the EV 850 control. 851  infected with OC43 (TCID50 = 2 x 10 4 ) or 229E (TCID50 = 2.4 x 10 3 ). 6-, 12-, or 24 hpi, virus-876 containing supernatant was harvested and titrated TCID50 assay in VeroE6 (A) or MRC5 (B) 877 cells, respectively. These data represent two independent experiments (n=2); mean ± SD. C. 878 HUVECs were infected with 229E as in B or mock infected. Total RNA was harvested 12 hpi 879 and RT-qPCR was performed using primers specific to 229E genomic RNA (gRNA) or 880 subgenomic RNA8 (sgRNA8), or HPRT (cellular house-keeping gene). Cycle-threshold (Ct) 881 values for each primer pair were plotted. These data represent three independent experiments 882 (n=3); mean ± SD. Ct values greater than 35 were not deemed biologically relevant and therefore 883 are scored as N/A. D. PCR products from C. were resolved using agarose gel electrophoresis. infected with OC43 (TCID50 = 2 x 10 4 ) or mock-infected. 12 or 24 hpi cells were fixed and 889 immunostained for DDX6 (green) and Hedls (white). Nuclei were stained with Hoechst (blue). 890 Representative images from one of three independent experiments shown. B. Hedls puncta in 891 OC43-infected or mock-infected cells were quantified per field of view using CellProfiler. C. 892 HUVECs were infected with 229E (TCID50 = 2.4 x 10 3 ) or mock-infected. 6 or 12 hpi cells were 893 fixed and immunostained for DDX6 (green) and Hedls (white). Nuclei were stained with 894 Hoechst (blue). Representative images from one of three independent experiments shown. D. 895 Hedls puncta in 229E-infected or mock-infected cells were quantified as in B. These data 896 represent three independent experiments (n=3) with ≥100 cells measured per condition per 897 replicate. Each mock and infected replicate pair plotted independently; mean. Statistics were 898 performed using a Mann-Whitney rank-sum test (****, p < 0.0001). Scale bar = 20 μm. 899 900 Figure S4. Processing body protein densitometry after human coronavirus infection. A. 901 HUVECs were transduced with human ACE2, selected, and infected with SARS-CoV-2 TO-1 902 isolate (MOI=3). Cells were lysed at 6 and 12 hpi and immunoblotting was performed using 903 XRN1, Hedls, DCP1A, DDX6, SARS-CoV-2 N, and β-actin specific antibodies. The results of 904 second independent experiment of two is shown. B-C. HUVECs were infected with OC43 905 (TCID50 = 2 x 10 4 ) or 229E (TCID50 = 2.4 x 10 3 ) or mock-infected. Cells were lysed at 12 and 24 906 hpi (B, OC43) or 6 and 12 hpi (C, 229E). Immunoblotting was performed using DDX6, Hedls, 907 XRN1, DCP1A, and actin. Protein densitometry was determined in ImageJ normalized to actin 908 and expressed as a fold-change relative to the 12 hpi (B; OC43) or 6 hpi (C; 229E) mock-909 infected control. These data represent three independent biological replicates (n=3). A one-way 910 ANOVA with a Dunnett's post-hoc analysis was performed; mean; bars represent SD (*, p < 911 0.05). 912 913 Figure S5. Ectopic expression of SARS-CoV-2 ORFs. HeLa cells expressing GFP-Dcp1a were 914 transfected with an empty vector (EV) control or 2xStrep-tagged SARS-CoV-2 ORFs for 48 915 hours then fixed and immunostained for Strep-Tag-II (viral ORF; green) or DDX6 (PBs; white). 916 Nuclei were stained with Hoechst (blue). A. SARS-CoV-2 ORFs that could be detected by  Tag-II staining (thresholded ORFs). B. SARS-CoV-2 ORFs that could not be detected by  Tag-II staining (unthresholded ORFs). Representative images from one of three or more 919 independent experiments shown. Scale is the same for all images. N=nucleocapsid protein, 920 E=envelope protein, M=membrane protein, S=spike protein 921 922 Figure S6. Validation of processing body loss induced by selected SARS-CoV-2 ORFs in 923 HUVECs. A. HUVECs were transduced with recombinant lentiviruses expressing 2xStrep-924 tagged SARS-CoV-2 nsp1, nsp14, ORF3b, and ORF7b constructs or control lentiviruses (EV) 925 and selected with puromycin for 48 hours. Cells were fixed and immunostained for Strep-Tag-II 926 (ORFs; green) and DDX6 (PBs; white). Nuclei were stained with Hoechst (blue). Representative 927 images from one of three independent experiments shown. Scale bar = 20 µm. B. DDX6 puncta 928 were quantified per field of view using CellProfiler. These data represent three independent 929 experiments (n=3) with >18 cells measured per condition per replicate. Each EV and ORF-930 expressing replicate pair plotted independently; mean. Statistics were performed using  Wallis H test with Dunn's correction (**, p < 0.0021; ****, p < 0.0001; ns, nonsignificant). C. 932 HUVECs were transduced as in A. Cells were lysed and immunoblotting was performed using 933 the Strep-Tag II antibody on a 4-15% gradient gel. 934 935 Figure S7. Processing body protein densitometry after CoV N protein overexpression. A-C. 936 937 HUVECs were transduced with recombinant lentiviruses ectopically expressing N protein from 938 the betacoronaviruses MERS-CoV and OC43 (A), N protein from alphacoronaviruses 229E and 939 NL63 (B) or SARS-CoV-1 N (C). A control lentiviral expressing an empty vector (EV) was used 940 as a negative control and SARS-CoV-2 N protein expressing lentiviruses were used as a positive 941 control in each experiment. Cells were selected and lysed and immunoblotting was performed 942 using XRN1, Hedls, DCP1A, DDX6, N protein or FLAG, and β-actin specific antibodies. 943 Protein densitometry was determined in ImageJ normalized to actin and expressed as a fold-944 change relative to the EV control. These data represent three independent biological replicates 945 (n=3). A one-way ANOVA with a Dunnett's post-hoc analysis was performed; mean; bars 946 represent SD (*, p < 0.05). 947 948 Figure S8. TNF treatment does not alter processing body number 949 HUVECs were treated with 0.01 ng/L soluble TNF for either 6 or 12 hours following which, 950 cells were fixed and immunostained for Hedls. Nuclei were stained with Hoechst. Hedls puncta 951 were quantified per field of view using CellProfiler. These data represent two independent 952 experiments (n=2) with >90 cells measured per condition per replicate. Each mock and infected 953 replicate pair plotted independently; mean. 954 955 956