Direct antiviral activity of interferon stimulated genes is responsible for resistance to paramyxoviruses in ISG15-deficient cells

Interferons (IFNs), produced during viral infections, induce the expression of hundreds of IFN- stimulated genes (ISGs). Some ISGs have specific antiviral activity while others regulate the cellular response. Besides functioning as an antiviral effector, IFN-stimulated gene 15 (ISG15) is a negative regulator of IFN signalling and inherited ISG15-deficiency leads to autoinflammatory interferonopathies where individuals exhibit elevated ISG expression in the absence of pathogenic infection. We have recapitulated these effects in cultured human A549-ISG15-/- cells and (using A549-UBA7-/- cells) confirmed that posttranslational modification by ISG15 (ISGylation) is not required for regulation of the type-I IFN response. ISG15-deficient cells pre-treated with IFN-α were resistant to paramyxovirus infection. We also showed that IFN-α treatment of ISG15-deficient cells led to significant inhibition of global protein synthesis leading us to ask whether resistance was due to the direct antiviral activity of ISGs or whether cells were non-permissive due to translation defects. We took advantage of the knowledge that IFN-induced protein with tetratricopeptide repeats 1 (IFIT1) is the principal antiviral ISG for parainfluenza virus 5 (PIV5). Knockdown of IFIT1 restored PIV5 infection in IFN-α-pre-treated ISG15-deficient cells, confirming that resistance was due to the direct antiviral activity of the IFN response. However, resistance could be induced if cells were pre-treated with IFN-α for longer times, presumably due to inhibition of protein synthesis. These data show that the cause of virus resistance is two-fold; ISG15-deficiency leads to the ‘early’ over-expression of specific antiviral ISGs, but the later response is dominated by an unanticipated, ISG15- dependent, loss of translational control. Key points Cell culture model of ISG15-deficiency replicate findings in ISG15-/- patient cells Cause of resistance in ISG15-/- cells differs depending on duration of IFN treatment ISG15-/- patients without serious viral disease don’t prove ISGylation is unimportant


Introduction 44
The innate immune response against pathogens is underpinned by the evolutionary conserved 45 interferon (IFN) system. All cells express pathogen recognition receptors (PRRs) that sense the 46 products of infection and establish a signalling cascade leading to the production of cytokines, 47 including type I IFN (IFN-α/β) (1, 2). IFN is secreted from cells and binds to cell surface receptors 48 expressed on both infected and non-infected cells, initiating a Janus kinase/signal transducer and 49 activator of transcription (JAK/STAT) signalling cascade, culminating in the expression of hundreds of 50 interferon stimulated genes (ISGs) (3). The biological effects of ISGs are extensive and their principle 51 role is to generate an unfavourable environment for the replication of viruses. Many ISGs have 52 broad antiviral activity, such as double-stranded RNA dependent protein kinase (PKR) that, upon 53 recognition of viral dsRNA, dampens general protein synthesis and prevents the translation of viral 54 mRNAs (4). Other antiviral ISGs, such as IFN-induced protein with tetratricopeptide repeats (IFIT) 55 proteins, inhibit specific viruses, but for many, they are inconsequential (5). Additionally, multiple 56 ISGs are generally required to limit infection because the majority of ISGs result in low to moderate 57 Viruses used were human parainfluenza virus 2 (HPIV2) strain Colindale (HPIV2-Co), HPIV3 strain 118 Washington/47885/57 (HPIV3-Wash) (20), PIV5 strain W3 (PIV5-W3) (22) and PIV5 strain CPI-(PIV5-119 CPI-) (23). Virus stocks were prepared by inoculating Vero cells at a multiplicity of infection (MOI) of 120 0.001 with continual rocking at 37°C. Supernatants were harvested at 2 d p.i., clarified by 121 centrifugation at 3,000 xg for 15 min, aliquoted and snap frozen. Titres were estimated by standard 122 plaque assay on Vero cells in 6-well plates. 123 For infection studies, cell monolayers were infected in 6-well plates with virus diluted in medium to 124 achieve a MOI of 10, unless stated otherwise. Virus adsorption was for 1 h, after which the viral 125 inoculum was removed and replaced with media supplemented with 2% (v/v) FBS and incubated in 126 5% (v/v) CO2 at 37°C until harvested. When cells were treated with IFN-α prior to infection (pre-127 treated) this was done with 1000 IU/ml IFN-α2b (referred to as IFN-α from here on; IntronA, Merck 128 Sharp & Dohme Ltd.) 18 h prior to infection, unless otherwise stated. IFN-α remained on cells for the 129 duration of experiments. Cells were either processed for immunoblot analysis or (if infecting with 130 rPIV5-mCherry, kind gift of Dr He, University of Georgia, USA) imaged using an IncuCyte Zoom 131 imaging system (Sartorius). 132 For plaque assays 30-40 PFU PIV5-CPI-in 1 ml DMEM, 2% FBS were adsorbed for 1 h onto confluent 133 monolayers of cells in 6-well plates while rocking at 37°C. Following adsorption, 2 ml overlay 134 (DMEM, 2% FBS, Avicel) was added to wells and incubated for 6 d. Cells were fixed with 5% 135 formaldehyde (10 min), washed in PBS and either stained for 10 min with 1 mg/ml toluidine blue O 136 (Sigma) followed by rinses with water or permeabilised for 10 min (PBS, 1% Triton X-100, 3% FBS) 137 washed again and incubated for 1 h with a pool of PIV5-specific antibodies (24) or mouse 138 monoclonal anti-HPIV3 NP (25) diluted in PBS, 3% FBS (1:1000). Following PBS washes, cells were 139 incubated for 1 h with goat anti-mouse IgG antibodies conjugated to alkaline phosphatase (Abcam 140 Cat# ab97020) diluted 1:1000 in PBS, 3%FBS. Cells were washed in PBS and signals were detected 141 using SIGMAFAST BCIP/NBT (Sigma). 142

ISG15-knockout A549 cells recapitulate ISG15-deficient patient cells 191
Among the several immune modulatory roles of ISG15 (8) B6 and C4). We also selected a clone that had gone 199 through the CRISPR/Cas9 process but retained ISG15 expression (C4+) (Fig. 1a). In addition to control 200 A549 cells, all clones were treated with IFN-α for 24 h, 48 h or left untreated. Immunoblot analysis 201 showed that, compared to control cells, expression of the ISGs MxA and IFIT1 were higher in A549-202 ISG15 -/cells (Fig. 1a). It was previously reported that increased ISG expression in ISG15-deficient 203 cells was due to enhanced signalling resulting from the destabilisation of the type I IFN negative 204 regulator USP18. To determine if IFN-α treatment led to enhanced signalling in A549-ISG15 -/cells we 205 selected clone B8 for further analyses. Cells were treated with IFN-α for 30 min, extensively washed 206 and media without IFN-α was replaced. Immunoblot analysis of cell lysates taken after 30 min 207 treatment (and following washes; 0') and 30 min later (30') showed that IFN-α treatment led to the 208 phosphorylation of STAT1, an indicator of IFN signalling, in both A549 and A549-ISG15 -/cells (Fig.  209 1b). Following 24 h treatment, there was clear evidence of ISG expression as shown by the 210 expression of MxA and ISG15 (in A549 cells) and enhanced expression of STAT1 (Fig. 1b). However, 211 while phospho-STAT1 levels had abated in both cell lines 24 h post-IFN-α treatment, levels were 212 clearly higher in A549-ISG15 -/cells indicating that in these cells there was a higher degree of 213 signalling. We also tested the impact of ISG15-deficiency on the expression of various ISG mRNAs. 214 A549-ISG15 -/cells were, in addition to control A549 cells, treated with IFN-α, or left untreated, for 215 24 h and the expression of various ISGs were examined by RT-qPCR. Whilst IFN-α treatment 216 enhanced the expression of all ISGs tested, this increase was larger in ISG15-deficient cells compared 217 to control A549 cells (between 5-and 10-fold, depending on the ISG) (Fig. 1c). Importantly, the 218 expression of ISGs in non-stimulated cells was equivalent to control cells suggesting that ISG15-219 dependent regulation is specific to the IFN response and not required for the regulation of basal 220 gene expression. Further experiments showed that lack of ISG15 prolonged the longevity of ISG 221 protein expression, which presumably has an impact on patients with autoinflammatory diseases 222 associated with ISG15 loss-of-function. Here, control A549 and knockout cells were treated with IFN-223 α for 24 h. The cells were washed and media (without IFN-α) was then added. Cells were harvested 224 every 24 h for 72 h and MxA expression was assessed by immunoblotting ( Fig. 1d). In control A549 225 cells MxA expression peaked at 24 h (the point at which IFN was removed) and had returned to basal 226 levels between 48 and 72 h. In knockout cells MxA expression was clearly higher than in control cells, 227 corroborating our mRNA analyses. Furthermore, while MxA expression in A549-ISG15 -/did recede 228 between 48 and 72 h, high protein levels remained at 72 h (Fig. 1d). A dysregulated IFN response in 229 ISG15-deficient cells is thought to be due to destabilisation of USP18, a known negative regulator of 230 JAK/STAT signalling (10). To determine if USP18 is similarly affected in our cell lines, A549-ISG15 -/-231 cells were treated with IFN-α for 24 or 48 h (or left untreated) and whole cell lysates were probed 232 for USP18 by immunoblotting. USP18 was robustly induced in A549 cells following IFN-α treatment; 233 however, levels of USP18 were much lower in IFN-α-treated ISG15-deficient cells (Fig. 1e). USP18 234 mRNA levels were approximately 10-fold higher in IFN-treated ISG15-deficient cells compared to 235 control A549s, demonstrating that reduced USP18 in A549-ISG15 -/cells was not due to reduced 236 transcription (Fig. 1c). Together, these data show that ISG15 is critical for the regulated expression of 237 ISGs. Moreover, they demonstrate that the effects of IFN treatment on our ISG15 knockout A549 cell 238 lines recapitulate the findings in cells derived from ISG15-deficent, patient cells. 239

ISG15-deficiency leads to translational repression following IFN treatment 241
During our studies we observed that IFN-α-treatment of ISG15-knockout cells led to a reduction in 242 protein synthesis and reasoned that this was a likely contributor to the reported virus resistance in 243 ISG15-deficient cells (14) analysis that measures the accumulation of viral protein over time; here, the levels of viral protein 257 appeared as high, if not higher, at 48 h p.i. than 24 h p.i. (Fig. 2c). In contrast, the levels of viral 258 protein synthesis following IFN-α treatment was higher at 48 h p.i. than at 24 h p.i. because IFN-α 259 treatment delayed PIV5 infection (Fig. 2b). This was also indicated by immunoblot analysis where the 260 accumulation of NP was higher at 48 h p.i. that 24 h p.i. (Fig. 2c). When A549-ISG15 -/cells were 261 infected, there was clear evidence of NP protein synthesis (Fig. 2b) and accumulation (Fig. 2c); 262 however, when these cells were pre-treated with IFN-α and infected, there was very little evidence 263 of viral protein synthesis (Fig. 2b) or accumulation (indicating that viral protein synthesis was barely 264 initiated) (Fig. 2c) (14), and this seems to extend to 273 PIV5 with our in vitro system (Fig. 2b-c). To investigate this in A549-ISG15 -/cells, control A549 cells 274 and the ISG15 knockout clones described above were either untreated or treated with 1000 IU/ml 275 IFN-α2b (the same concentration and IFN-α type used in (14)) for 18 h. Cells were then infected with 276 PIV5 (strain W3) (22) for 24 and 48 h and analysed by immunoblotting. In all cell lines, the levels of 277 PIV5 nucleoprotein (NP) expression was equivalent at 24 and 48 h in unstimulated cells (Fig. 3a). In 278 IFN-α pre-treated control cells, including C4+ that retained ISG15 expression, the level of NP 279 expression was markedly reduced at 24 h. By 48 h, the level of NP increased showing that infection 280 had progressed even in the presence of IFN-α (Fig. 3a). This is because the PIV5-V protein targets 281 STAT1 for proteasomal degradation, and once sufficient V is expressed, the IFN response is 282 dismantled allowing the virus to replicate (23). Indeed, there was no detectable STAT1, and as a 283 result, markedly reduced levels of ISGs MxA and IFIT1 in PIV5-infected, ISG15-expressing cells (Fig.  284   3a). However, all A549-ISG15 -/cell lines that had been pre-treated with IFN-α were resistant to PIV5 285 infection as shown by dramatically reduced, or even absent, NP expression at both time points (Fig.  286 3a). Moreover, these cells displayed STAT1 expression and the expression of associated MxA and 287 IFIT1 (indicating that PIV5 infection was inhibited) (Fig. 3a). 288 Previous reports have shown that the ISG15 regulation of IFN signalling is independent of its ability 289 to covalently modify proteins by ISGylation (10). To confirm this, we again applied CRISPR/Cas9 290 genome engineering technology and knocked out expression of UBA7, the E1 enzyme required for 291 ISGylation. For this we took a different approach compared to generating our ISG15 knockout cells 292 (19). Here, we introduced constitutive expression of Cas9 by lentiviral transduction of A549 cells and 293 transduced A549-Cas9 cells with lentiGuide-Puro lentivirus carrying a guide RNA specific for UBA7, 294 followed by single-cell cloning. We confirmed that all clones were UBA7-deficient by immunoblot 295 analysis, which demonstrated that they retained expression of ISG15 but had lost the ability to 296 ISGylate proteins (Fig. 3b). Additionally, following the scheme used in Fig. 3a, these cells were 297 infected with PIV5-W3. These data showed that, compared to ISG15 knockout cells that were 298 resistant to infection, all IFN-α-pre-treated UBA7-knockout cells were infected as efficiently as 299 control cells (Fig. 3b), confirming reports that ISG15-dependent regulation of type I IFN signalling 300 does not require ISGylation (10). 301 302

The direct antiviral activity of ISGs is responsible for virus resistance 303
Virus resistance can be induced following 8 h IFN-α treatment (shorter times were not tested), well 304 before any obvious effect on global protein synthesis (Fig. 2). Therefore, shutdown of translation is 305 unlikely to be the sole contributor to virus resistance at early time points and so we wished to 306 determine whether the direct antiviral activity of ISGs was responsible. Addressing this question is 307 complex since, for most viruses, the specific ISG(s) responsible for blocking replication is not known. 308 However, for PIV5, it has been established that IFIT1 is the principle ISG responsible for most of the 309 IFN-dependent antiviral activity (17,27). We therefore hypothesised that if virus resistance was 310 caused by the direct antiviral activity of ISGs, knockdown of IFIT1 in ISG15-deficient cells would 311 permit PIV5 replication during an antiviral response. We reduced IFIT1 (according to (17) were pre-treated, or left untreated, with IFN-α and then infected with PIV5-W3 (MOI 10) for 24 and 314 48 h. Expression of PIV5 NP, analysed by semi-quantitative immunoblotting, was used to measure 315 virus infection (Fig. 4a). IFIT1 levels and expression of ISG15 were likewise tested. Typically, pre-316 treatment of naïve cells with IFN-α reduced infection, as shown by a reduction in NP levels, 317 compared to non-treated cells (Fig. 2b-c & 3a-b); nevertheless, because PIV5 expresses the IFN 318 antagonist V protein, NP levels reach similar levels to untreated cells by 48 h p.i. However, this IFN-319 dependent reduction in virus infection is diminished when IFIT1 is knocked down, confirming earlier 320 reports of IFIT1's antiviral activity against PIV5 (17, 27). While IFN- pre-treatment of A549-ISG15 -/-321 cells renders them resistant to infection, when IFIT1 was also knocked down, PIV5 infection was 322 restored (Fig. 4a). Because we performed semi-quantitative immunoblotting of NP and β-Actin, we 323 were able to quantify NP levels, allowing us to analyse these changes statistically (Fig. 4b). These 324 data show that in IFN-α-pre-treated cells, knocking IFIT1 down restored NP to similar levels to those 325 seen in untreated cells, regardless of ISG15 status. While IFN-α pre-treatment of A549 cells 326 significantly reduced NP levels when we compared 24 h and 48 h p.i. samples, there was no 327 difference at these time points when IFIT1 was knocked down (Fig. 4b). Importantly, while NP levels 328 were virtually absent in IFN-α-pre-treated ISG15-deficient cells, when IFIT1 was knocked down in 329 these cells NP levels were equivalent to A549-shIFIT1 cells (Fig. 4b). 330 Rather than solely relying on viral protein expression as a surrogate for virus infection, we also 331 tested virus replication using biologically relevant plaque assays. Because paramyxoviruses (like 332 most wild type viruses) are poor inducers of the IFN response (28,29), are able to efficiently and 333 rapidly counteract it if it were induced, and our data showed that basal ISG expression was not 334 effected in ISG15-deficient cells (Fig. 1c), we predicted that infection of naïve A549-ISG15 -/cells 335 would be equivalent to naïve A549 cells. To determine if this was the case, plaque assays were 336 performed with various paramyxoviruses. These data show that each virus formed plaques that 337 were analogous on both A549 and A549-ISG15 -/cells (Supplemental Fig. 1). There were subtle 338 differences in plaque phenotype; for instance, infection of ISG15-deficient cells, particularly with 339 HPIV2 but also evident following PIV5 infection, resulted in plaques with poorer defined edges (hazy 340 plaques) (Supplemental Fig. 1). The reason for this is currently not clear but may indicate an antiviral role for ISG15 against HPIV2 and PIV5. Nevertheless, this, and data in figures 2 and 3, supports the 342 notion that naïve cells were not resistant to wild type viral infection. However, viruses unable to 343 counteract the IFN response should be restricted and therefore provide a means of assessing the 344 role of ISG15 and virus resistance. 345 To do this cells were infected with approximately 30-40 PFU of PIV5 strain CPI-(PIV5-CPI-) (30), a 346 strain unable to block IFN signalling due to a mutation in its V protein. Infected cells were fixed 6 d 347 p.i. and stained for viral antigen (Fig. 4c). As previously demonstrated (17), PIV5-CPI-was unable to 348 efficiently form plaques in IFN-competent A549 cells. However, PIV5-CPI-did replicate when cells 349 were unable to produce IFN, such as in A549-Npro cells that constitutively express bovine viral 350 diarrhea virus (BVDV) Npro that cleaves IRF3 (a transcription factor critical for IFN induction (21)). 351 Furthermore, when IFIT1 was knocked down, PIV5-CPI-was able to replicate (albeit less efficiently), 352 further highlighting the major role of IFIT1 as an anti-PIV5 protein. As expected, and like A549 cells, 353 there was very little virus replication in A549-ISG15 -/cells; however, when IFIT1 was knocked down, 354 cells were able to support virus replication. It must be noted however that virus replication in A549-355 ISG15 -/-/shIFIT1 cells did not recover to the same degree as A549-shIFIT1 cells. We propose that the 356 reason for this will be complex and may include the likelihood that additional, yet to be identified, 357 anti-PIV5 ISGs exist which are expressed at higher levels in ISG15-deficient cells. Another possible 358 explanation is the inhibition of protein synthesis, including that of viral proteins, in ISG15-deficent 359 cells; cells were infected for 6 days prior to performing the plaque assays, a time point beyond that 360 required to observe a significant effect on protein synthesis (Fig. 2a). Therefore, the plaques 361 observed in A549-ISG15 -/-/shIFIT1 cells likely result from virus that replicated prior to the inhibition 362 of global protein synthesis. 363 IFIT1 restricts viral infection post-transcriptionally by blocking the translation of viral mRNA (17, 27); 364 therefore, we predicted that IFN-α-pre-treated A549-ISG15 -/cells would remain susceptible to 365 infection, but that high levels of IFIT1 would mean these cells would not be permissive to PIV5 366 infection. Furthermore, investigating this could highlight additional restrictions to viral infection, 367 such as entry. A549 and A549-ISG15 -/cells were pre-treated for 8 h with IFN-α and then infected 368 with PIV5-W3 (MOI 10) (Fig. 4d). Analysis of PIV5 NP transcription showed that ISG15-deficent cells 369 were infected and that viral transcription increased over time; however, this was muted compared 370 to A549 control cells. Importantly however, the levels of NP transcription at 1 h p.i. was equivalent in 371 both cell lines, a time point that likely represents primary transcription ( Fig. 4d; see inset graph). 372 These data suggest that both cell lines were susceptible to infection and that high levels of pre-373 existing IFIT1 strongly restricted further viral transcription by preventing the translation of the virally 374 encoded mRNAs. To investigate if IFIT1 restriction was responsible for reduced viral transcription in 375 ISG15-deficient cells, we repeated the experiment in A549-shIFIT1 and A549-ISG15 -/-/shIFIT1 cells 376 (Fig. 4e). These data show that in IFN-α-treated cells, viral transcription was markedly increased 377 compared to cells with intact IFIT1 expression. Furthermore, in A549-shIFIT1 cells, transcription 378 peaked between 12 and 18 h p.i. and then receded. We have recently described the transcription 379 and replication of various paramyxoviruses, including PIV5-W3, using un-biased high throughput, 380 RNA-seq approach (26); this report shows that this pattern of transcription is typical of PIV5- W3 and 381 likely results from the phosphoprotein (P)-dependent repression of viral transcription and replication 382 (31). This repression also occurred in A549-ISG15 -/-/shIFIT1 cells, but this occurred later (Fig. 4e), 383 suggesting that ISG15 may be an additional antiviral factor that curtail PIV5 transcription. 384 Nevertheless, these data showed that when IFIT1 levels were knocked down, the transcriptional 385 repression identified in IFN-α-pre-treated ISG15-deficient cells was relieved, demonstrating that 386 virus resistance was due to the post-transcriptional activity of IFN-inducible IFIT1. We also 387 investigated infection of these cell lines with other paramyxoviruses whose sensitivity to IFIT1 has 388 been previously reported. Cells were treated with IFN-α and then infected with HPIV2 strain 389 Colindale (MOI 10; family Paramyxoviridae, sub-family Orthorubulavirinae), which is reported to be 390 moderately sensitive to IFIT1-restriction (27), for 24 and 48 h (untreated cells were not analysed 391 because of high cytopathic effect in the absence of IFN). To investigate infection, we detected expression of HPIV2 phosphoprotein (P) by semi-quantitative immunoblotting (Fig. 5a), which 393 showed that IFN-α-pre-treated A549-ISG15 -/cells were largely resistant to infection, although by 48 394 h p.i. there was some, albeit low level, evidence of viral protein accumulation. Nevertheless, 395 infection of A549-ISG15 -/-/shIFIT1 did allow significantly more viral protein expression. Semi-396 quantitative analyses demonstrated that viral protein accumulation in A549-ISG15 -/-/shIFIT1 cells 397 was significantly higher than in A549-ISG15 -/cells, but this was not as high as in A549 control cells, 398 which agrees with the reported partial sensitivity of HPIV2 to IFIT1 restriction indicating that 399 additional ISGs target HPIV2 (Fig. 5b). We performed a similar analysis with HPIV3 strain Washington 400 (20)  Our data have so far suggested that early virus resistance is mediated by the direct antiviral activity 412 of the IFN response. However, protein synthesis is reduced at later times post-IFN treatment and 413 this is likely to cause resistance; therefore, we investigated whether PIV5 resistance could be 414 induced independently of the direct antiviral activity of IFIT1. To do this we pre-treated the four cell 415 lines (A549, A549-shIFIT1, A549-ISG15 -/and A549-ISG15 -/--shIFIT1) with IFN-α for different periods of 416 time, infected with a recombinant PIV5 that expresses the fluorescent protein mCherry (rPIV5-417 mCherry) for 48 h (MOI 10) and measured fluorescence as a marker of virus replication (Fig. 6a). 418

Virus replication in A549 cells was equivalent regardless of the time cells had been pre-treated with 419
IFN-α and, as expected, A549-ISG15 -/cells were resistant to infection at any time post IFN-α 420 treatment (Fig. 6b). Any advantage to PIV5 replication as a result of IFIT1 knockdown in A549-shIFIT1 421 cells was lost when cells had been pre-treated for 16 h or more, as longer periods of pre-treatment 422 resulted in replication equivalent to IFN-pre-treated A549 cells. Similarly, PIV5 replication in A549-423 ISG15 -/--shIFIT1 cells was higher than A549 control cells, and equivalent to A549-shIFIT1 cells, 424 following 8 and 16 h pre-treatment; however, when cells were pre-treated for 24 h, replication was 425 lower than in A549 and A549-shIFIT1 cells. Interestingly, as the time of pre-treatment of A549-ISG15 -426 /--shIFIT1 cells extended, virus replication reduced further until cells became resistant (e.g. at 60 h 427 and 72 h pre-treatment, Fig. 6b), which was not observed in A549 or A549-shIFIT1 cells. These data 428 suggest that cell permissiveness progressively reduced with longer times of IFN-α pre-treatment, 429 which correlated with the effects of IFN-α treatment on protein synthesis in ISG15-deficient cells 430 (Fig. 2). 431 A previous report demonstrated that ISG15-dependent stabilisation of USP18 was required to bring 432 about regulation of the type I IFN response and this was sufficient for these cells to once again be 433 infected (14). However, what aspects of the antiviral response was responsible for resistance was 434 not investigated. Taken together, these data strongly suggest that virus resistance in early IFN-435 treated ISG15-deficient cells was caused by the direct antiviral activity of ISGs and not due to a lack 436 of permissiveness as a result of IFN-dependent inhibition of protein synthesis. Nevertheless, because 437 of the reduced protein synthesis in IFN-α-treated ISG15-deficient cells, cells later become non-438 permissive to infection, even when key ISGs are eliminated. Previous work had shown that virus resistance was observed in cells that had been treated with IFN-443 α and then left to rest for 36 h prior to challenge (14). We had observed that IFN-α treatment of 444 A549-ISG15 -/cells led to dramatic decreases in protein synthesis, particularly between 24 and 48 h; 445 therefore, it was not clear whether the initially reported virus resistance was due to defects in 446 translation (including of viral mRNAs) at the timepoint used in (14) or due to the direct antiviral 447 activity of the IFN response. For most viruses, the specific ISG(s) with antiviral activity for a given 448 virus is not known, making the latter difficult to discern; however, for PIV5, it is well established that 449 IFIT1 is responsible for the majority of the antiviral response (17). To study this we generated A549-450 ISG15 -/cells and showed these cells recapitulated the effects observed in ISG15-deficient patient 451 cells following treatment with IFN-α which included dysregulated ISG expression and reduced USP18 452 protein levels following IFN-α treatment (Fig. 1). Additionally, by knocking-out UBA7, the first 453 enzyme in the ISGylation cascade, we showed that ISGylation is not required for a regulated 454 response (Fig. 3b), confirming previous reports that 'free' ISG15 is required for regulation (10). 455 Using these cell lines in combination with a PIV5 infection model, we showed that infection of IFN-α-456 pre-treated ISG15-deficient cells in which IFIT1 had been knocked down restored infection, thus 457 confirming that at early times post infection, resistance was indeed due to the direct antiviral activity 458 of the IFN response. Furthermore, because IFIT1 blocks the translation of viral transcripts, our data 459 show that IFN-treated A549-ISG15 -/cells were still susceptible to infection, allowing viral 460 transcription to take place prior to IFIT1 restriction, and that ISG15 was unlikely to significantly 461 regulate processes involved in entry (Fig. 4d-e). Nevertheless, if ISG15-deficient cells were treated 462 for longer periods with IFN-α prior to infection they did become resistant, even when IFIT1 was 463 knocked down, suggesting that at later times the inhibition of protein synthesis was the principal 464 cause of resistance (Fig. 6). These data suggest that the virus resistance reported by Speer et al. (14) 465 was due to a lack of permissiveness and not a result of the direct antiviral activity of the IFN 466 response, although different cells were used in that study. 467 The data here demonstrate that the mechanism of resistance is likely two-fold, depending on the 468 duration that cells are exposed to IFN-α. It is not currently possible to know which mechanism is 469 dominant in ISG15-deicient patients, but it is likely to be a combination of both. Nevertheless, virus 470 resistance results from a lack of IFN signalling control -as a consequence of ISG15-loss-of-function -471 which would explain why ISG15-deficient patients were not more susceptible to severe infection. 472 This observation, therefore, cannot be used to support the notion that human ISG15 does not 473 possess direct antiviral activity, as proposed (14,16). It is likely that many viruses will not be 474 sensitive to ISG15-dependent antiviral activity; however, this is true of many antiviral effectors. For 475 example, and as confirmed in this study, IFIT1 strongly restricts PIV5 infection, yet it has reduced 476 activity against HPIV2 and likely no activity against HPIV3 or human respiratory syncytial virus (27). It 477 is also true that several ISGs are often required to limit infection (6); therefore, if one antiviral 478 effector mechanism is absent (such as ISGylation), there is sufficient redundancy to avoid severe 479 effects of infection (redundancy that can complicate the investigation of specific antiviral 480 mechanisms in in vitro studies). Nevertheless, several human viruses have been shown to be 481 sensitive to ISGylation and many have evolved specific mechanisms to counteract antiviral 482 ISGylation, adding further weight to the argument that human ISG15 does have antiviral activity 483 (reviewed in (8)). Indeed, other than the handful of patients that have been found to lack ISG15 484 expression (10, 32), individuals will possess an intact IFN response where the antiviral activity of 485 ISG15 (and other effectors) will function, if the infecting virus is sensitive to it. 486 It was surprising that protein synthesis was so affected in ISG15-deficient cells following IFN 487 treatment. It is well established that inhibition of general protein translation is a key feature of the 488 antiviral response and this is through the actions of proteins such as PKR or PERK (PKR-like ER kinase) 489 (4). However, for PKR to be activated it must recognise dsRNA, which was absent in IFN-α-treated 490 cells. Similarly, PERK is activated upon endoplasmic reticulum stress which might be expected during 491 a viral infection, but not following treatment with IFN alone. Previous reports have shown that 492 carcinoembryonic antigen-related cell adhesion molecule 1 (CEACAM1) has antiviral activity against 493 human cytomegalovirus, influenza virus and metapneumovirus by supressing mTOR-mediated 494 protein synthesis (33, 34). The membrane protein CEACAM1 is induced by innate sensors such as 495 TLR-4 (35) and IFI16 (34) and delivers inhibitory signals via SHP1 (haematopoietic cells) or SHP2 496 (epithelial and endothelial cells) phosphatase activity through CEACAM1 immunoreceptor tyrosine-497 based inhibitory motifs (ITIMs) (36). CEACAM1 expression is rapidly induced following activation of 498 NF-κB and IRF1, but whether IFN-α alone (as used here) can induce it expression is not clear. The 499 IRF1 promoter possesses a single GAS element, but no ISRE, and so its expression is induced by 500 STAT1 homodimers (37). Type I IFN signalling predominantly leads to the formation of STAT1-STAT2 501 heterodimers that associate with IRF9 (to form the ISGF3 transcription factor) to drive expression of 502 ISGs that possess ISRE elements in their promoters; however, STAT1 homodimers are formed after 503 type I IFN treatment, but these are at lower concentrations. It is possible that 'late' inhibition of 504 protein synthesis in ISG15-deficient cells (compared to the swifter antiviral activity of ISRE-505 containing genes such as IFIT1) may relate to the kinetics of CEACAM1 expression as the 506 accumulation of STAT1 homodimers is required to drive the expression of IRF1, that itself needs to 507 be translated before it induces CEACAM1. Of course, the accumulation of STAT1 homodimers may 508 be higher in ISG15-deficient cells because of a dysregulated type I IFN response. Nevertheless, it is 509 plausible that the overamplified type I IFN response in ISG15-deficient cells led to high levels of 510 CEACAM1 (compared to control cells) resulting in inhibition of protein synthesis. Moreover, ISG15 511 may have yet-to-be characterised functions in regulating the cellular response to stressors that lead 512 to inhibition of protein synthesis. 513 It has been reported that ISG15 has a role in regulating the cell cycle through its interactions with 514 SKP2 and USP18, although experiments in that study were not performed in IFN-treated cells, nor 515 were ISG15 knockout cells tested (15). While rates of protein synthesis differ during different stages of the cell cycle, translation is thought to be lowest during mitosis (38). Perturbation of the ISG15-517 SKP2-USP18 axis following ablation of USP18 led to a delayed progression from G1 to S phase which 518 is not generally thought to be associated with translational repression (39). Of note, we have not 519 observed any obvious differences in cell growth in non-treated A549-ISG15 -/cells. Further work is 520 required to dissect the mechanism responsible for ISG15's effects on general protein translation 521 during an antiviral response. 522 ISG15 has emerged as a central regulator of immunity. It is a pleotropic protein that is strongly 523 expressed following activation of innate immune sensors and connects innate and adaptive 524 immunity. In this study, we have shown that a lack of ISG15 leads to virus resistance by two 525 kinetically distinct mechanisms; the rapid induction of antiviral ISGs and the unexpected effects on 526 protein synthesis. Our newly developed cell lines and infection model will pave the way for further 527 studies investigating the regulatory mechanisms of ISG15 during the antiviral response. 528 529 Acknowledgments 530 We are grateful to ERASMUS+ for supporting DH. The authors also acknowledge technical support 531 provided by Christopher Simmons-Riach and Miroslav Botev. 532 was used to knockout ISG15 expression in A549 cells followed by single-cell cloning (following 657 previously reported procedures (18)). Four independent clones were treated with 1000 IU/ml IFN-α 658 for 24 and 48 h, or left untreated, and protein