Viral mimicry of p65/RelA transactivation domain to inhibit NF-κB activation

Sensing of virus infection activates NF-κB to induce the expression of interferons, cytokines and chemokines to initiate the antiviral response. Viruses antagonise these antiviral defences by interfering with immune sensing and blocking the actions of antiviral and inflammatory molecules. Here, we show that a viral protein mimics the transactivation domain of the p65 subunit of NF-κB. The C terminus of vaccinia virus (VACV) protein F14 (residues 51-73) activates transcription when fused to a DNA-binding domain-containing protein and F14 associates with NF-κB co-activator CBP, disrupting p65-CBP interaction. Consequently, F14 diminishes CBP-mediated acetylation of p65 and the downstream recruitment of the transcriptional regulator BRD4 to the promoter of the NF-κB-responsive genes CXCL10 and CCL2, hence inhibiting their expression. Conversely, the recruitment of BRD4 to the promoters of NFKBIA, which encodes the inhibitor of NF-κB (IκBα), and CXCL8 remains unaffected in the presence of either F14 or JQ1, a competitive inhibitor of BRD4 bromodomains, indicating its recruitment is acetylation-independent. Therefore, unlike other viral NF-κB antagonists, F14 is a selective inhibitor of NF-κB-dependent gene expression. A VACV strain lacking F14 showed that it contributes to virulence in an intradermal model of infection. Our results uncover a mechanism by which viruses disarm the antiviral defences through molecular mimicry of a conserved host protein and provide insight into the regulation of NF-κB-dependent gene expression by BRD4.


Sensing of virus infection activates NF-κB to induce the expression of interferons, cytokines 21
and chemokines to initiate the antiviral response. Viruses antagonise these antiviral defences 22 by interfering with immune sensing and blocking the actions of antiviral and inflammatory 23 molecules. Here, we show that a viral protein mimics the transactivation domain of the p65 24 subunit of NF-κB. The C terminus of vaccinia virus (VACV) protein F14 (residues 51-73) 25 activates transcription when fused to a DNA-binding domain-containing protein and F14 26 associates with NF-κB co-activator CBP, disrupting p65-CBP interaction. Consequently, F14 27 diminishes CBP-mediated acetylation of p65 and the downstream recruitment of the 28 transcriptional regulator BRD4 to the promoter of the NF-κB-responsive genes CXCL10 and 29 CCL2, hence inhibiting their expression. Conversely, the recruitment of BRD4 to the promoters 30 of NFKBIA, which encodes the inhibitor of NF-κB (IκBα), and CXCL8 remains unaffected in 31 the presence of either F14 or JQ1, a competitive inhibitor of BRD4 bromodomains, indicating 32 its recruitment is acetylation-independent. Therefore, unlike other viral NF-κB antagonists, F14 33 is a selective inhibitor of NF-κB-dependent gene expression. A VACV strain lacking F14 34 showed that it contributes to virulence in an intradermal model of infection. Our results uncover 35 a mechanism by which viruses disarm the antiviral defences through molecular mimicry of a 36 conserved host protein and provide insight into the regulation of NF-κB-dependent gene 37 expression by BRD4. 38

INTRODUCTION 40
Viruses provide constant selective pressure shaping the evolution of the immune systems of 41 multicellular organisms [1][2][3]. At the cellular level, an array of receptors detects virus-derived 42 molecules, or more broadly pathogen-associated molecular patterns (PAMPs), allowing the 43 recognition of invading viruses and the activation of a gene expression programme that 44 initiates the antiviral response [reviewed by [4,5]]. The induced gene products, which include 45 interferons, cytokines and chemokines, are secreted and function as signals to activate more 46 specialised immune cells and attract them to the site of infection, thereby generating 47 inflammation [reviewed by [6][7][8]]. This coordinated inflammatory response evolved to achieve 48 the control and (or) elimination of the infection, and the establishment of an immunological 49 memory against future infection [reviewed by [4,9]]. 50 Engagement of pattern recognition receptors (PRRs) by their cognate PAMPs activates 51 multiple transcription factors, including nuclear factor kappa light-chain enhancer of activated 52 B cells (NF-κB) [reviewed by [10][11][12]]. NF-κB is a homo-or heterodimer of Rel proteins, with 53 the heterodimer of p50 (also known as NF-κB1 or NFKB1) and p65 (also known as RelA or 54 RELA) being the prototypical form of NF-κB [13]. Through an interface formed by the Rel 55 homology domains of the two Rel subunits, NF-kB recognises and binds to a consensus DNA 56 sequence in the promoter elements and enhancers of target genes [reviewed by [10], [14]].

57
NF-κB-responsive gene products include inflammatory mediators, such as cytokines, 58 chemokines and cell adhesion molecules, as well as proteins involved in other immune 59 processes, like MHC molecules, growth factors and regulators of apoptosis [15,16]. Moreover, 60 cytokines, such as interleukin (IL)-1 and tumour necrosis factor (TNF)-α, also trigger NF-κB 61 activation upon engagement of their receptors on the cell surface [reviewed by [10,11]]. 62 In resting conditions, NF-κB remains latent in the cytoplasm bound to the inhibitor of κB (IκB) 63 α (also known as NFKBIA) [17,18]. Upon activation, the IκB kinase (IKK) complex 64 phosphorylates IκBα, triggering its ubiquitylation and subsequent proteasomal degradation, 65 thus releasing NF-κB to accumulate in the nucleus [18] [reviewed by [10,11]]. In the nucleus, 66 NF-κB interacts with chromatin remodelling factors, coactivators and general transcription 67 factors to activate the transcription of antiviral and inflammatory genes by RNA polymerase 68 (RNAP) II [14,[19][20][21][22][23][24]. The specificity and kinetics of NF-κB-dependent gene expression is 69 determined by several factors including dimer composition [25], cooperation with other 70 transcription factors [24], duration of the stimulus [26,27], cell type [28,29] and chromatin 71 context on the promoters of target genes [14,30,31]. In addition, NF-κB undergoes multiple 72 posttranslational modifications, in the cytoplasm or nucleus, that control its transcriptional 73 activity through interactions with coactivators and basal transcription machinery [reviewed by 74 [13,32,33]]. 75 Following the stimulation with PAMPs or inflammatory cytokines (e.g., TNF-α), two conserved 76 residues in p65 are phosphorylated: S276, mainly by protein kinase A (PKA), and S536 within 77 the transactivation domain (TAD), by the IKK complex [34,35]. Phosphorylation of either site 78 enhances NF-κB transcriptional activity by promoting the interaction with the coactivators 79 CREB-binding protein (CBP) or its paralogue p300 (also known as CREBBP and EP300, 80 respectively). These coactivators acetylate both p65 at K310 and histones on the target gene 81 promoters to allow transcription initiation and elongation to proceed [35][36][37][38]. The bromodomain 82 and extraterminal domain (BET) protein BRD4 docks onto acetylated p65-K310, via its two 83 bromodomains, and subsequently recruits positive transcription elongation factor b (P-TEFb) 84 to drive transcription of inflammatory genes by RNAP II [39]. This latter study highlighted the 85 complexity of the gene expression programme downstream of NF-κB, with subsets of genes 86 differentially expressed depending on the transcriptional regulatory events following NF-κB 87 recruitment to DNA [[39-41]; reviewed by [42]]. Targeting of the nuclear activity of NF-κB and 88 its coactivator CBP by viral proteins has been described as a strategy to antagonise antiviral 89 responses (e.g. high-risk human papillomavirus (HPV) E6 protein [43] and herpes simplex 90 virus (HSV) type 1 protein VP16 [44]). However, these studies do not elucidate in detail how 91 viral interference with NF-κB in the nucleus affects induction of the inflammatory genes by this 92 transcription factor. Furthermore, there are contradictory reports regarding the interaction 93 between VP16 and CBP [21,44]. 94 The confrontations between viruses and hosts leave genetic signatures over their evolutionary 95 histories [3,45]. On one hand, host innate immune factors display strong signs of positive 96 selection to adapt to the pressure posed by viruses [reviewed by [46]]. On the other hand, 97 viruses acquire multiple mechanisms to antagonise host innate immunity, such as mimicking 98 host factors to disrupt their functions in the antiviral response or to subvert them for immune 99 evasion [reviewed by [46,47]]. Poxviruses have been a paradigm in the study of virus-host 100 interactions [reviewed by [48]]. Their large DNA genomes encode a plethora of proteins that 101 antagonise the host antiviral response. Some poxvirus proteins show structural similarity to 102 host proteins and modulate innate immune signalling during infection [reviewed by [49,50]]. 103 For instance, vaccinia virus (VACV), the smallpox vaccine and the prototypical poxvirus, 104 encodes a family of proteins sharing structural similarity to cellular Bcl-2 proteins despite very 105 limited sequence similarity. Viral Bcl-2-like proteins have evolved to perform a wide range of 106 functions, such as inhibition of NF-κB activation [reviewed by [51]]. VACV protein A49, 107 notwithstanding its Bcl-2 fold, also mimics the IκBα phosphodegron that is recognised by the 108 E3 ubiquitin ligase β-TrCP, thereby blocking IκBα ubiquitylation [52,53]. Upon NF-κB 109 activation, the IKK complex phosphorylates A49 to create the complete phosphodegron mimic 110 that then engages β-TrCP to prevent IκBα ubiquitylation [54]. 111 Despite the existence of multiple inhibitors of NF-κB from VACV, virus strains lacking individual 112 inhibitors have reduced virulence in mouse models, arguing against their functional 113 redundancy [reviewed by [55,56]]. Therefore, the detailed study of the mechanisms 114 underpinning the antagonism of NF-κB by VACV and other poxviruses offers an opportunity 115 to dissect the signalling pathways leading to NF-κB activation and their relative contributions 116 to antiviral immunity. Previous work from our laboratory predicted that VACV encodes 117 additional inhibitors of NF-κB because a mutant VACV strain (vv811ΔA49)  IL-1β-stimulated NF-κB activity in HEK 293T cells in a dose-dependent manner ( Figure 1A, B 150 and Figure S1). This inhibitory activity was specific for the NF-κB pathway, because F14 did 151 not affect IFN-α-stimulated IFN-α/β receptor (IFNAR)/signal transducer and activator of 152 transcription (STAT) or phorbol 12-myristate 13-acetate (PMA)-stimulated mitogen-activated 153 protein kinase (MAPK)/AP-1 pathways ( Figure 1C, D). The inhibitory activity was exerted 154 despite the lower levels of F14 when compared to protein B14 ( Figure 1E) [64]. Conversely, 155 VACV protein C6 suppressed type I IFN signalling and B14 upregulated AP-1 activity, as 156 observed previously ( Figure 1C, D) [65,66]. 157 The virulence of VACV strains lacking specific genes has been tested mostly in intranasal or 158 intradermal murine models [reviewed by [55,56]]. Deletion of genes encoding VACV 159 immunomodulatory proteins may give a phenotype in either, neither or both models. To 160 evaluate if loss of F14 expression affected virulence, a recombinant VACV lacking F14 was 161 generated, termed vΔF14. Intradermal injection of the ear pinnae of mice with vΔF14 produced 162 smaller lesions, and reduced virus titres at 7 and 10 d post-infection (p.i.) ( Figure 1F, G) 163 compared to wildtype virus (vF14) and a revertant strain (vF14-Rev), generated by reinserting 164 F14 into vΔF14 at its natural locus ( Figure 1F, G). Attenuation of vΔF14 in the intradermal 165 mouse model correlated with reduced viral titres in the infected ears 7 and 10 d p.i., but not 3 166 d p.i. ( Figure 1G). In contrast, in an intranasal mouse model, vΔF14 caused the same extent 167 of body mass loss as wildtype and revertant controls ( Figure S2). In cell culture, vF14, vΔF14 168 and vF14-Rev displayed no differences in replication and plaque size ( Figure S3) F14 was tagged with a C-terminal TAP tag. The vF14-TAP strain replicated normally ( Figure  176 S3) and immunoblotting showed F14 protein expression was detected from 4 h p.i. and peaked 177 by 8 h p.i., matching the accumulation of the early VACV protein C6 ( Figure 1H) [70]. F14 178 levels were notably low either when expressed ectopically ( Figure 1E) or during infection 179 ( Figure 1H). This might explain why F14 was not detected in our recent quantitative proteomic 180 analysis of VACV infection, which detected about 80% of the predicted VACV proteins [71].

181
Pharmacological inhibition of virus DNA replication with cytosine arabinoside (AraC) did not 182 affect F14 protein levels, consistent with early expression, whereas late protein D8 was 183 inhibited ( Figure 1H). 184 The existence of multiple VACV inhibitors of NF-κB that each contribute to virulence indicates 185 they are not redundant. To test if F14 affects NF-κB activation during infection, we deleted 186 F14 from the vv811ΔA49 strain that lacks other known inhibitors of NF-κB [57] and then 187 infected an NF-κB reporter cell line [57]. As shown previously, vv811ΔA49 inhibited NF-κB to 188 a reduced extent when compared to the parental vv811 strain ( Figure 1I) [57] and deletion of 189 F14L from vv811ΔA49 reduced NF-κB inhibition further ( Figure 1I). Immunoblotting confirmed 190 equal infection with these viruses ( Figure 1J). Notably, vv811ΔA49ΔF14 still suppressed  κB activation considerably, which might be explained by: (i) the existence of additional virally-192 encoded inhibitors that cooperate to inhibit NF-κB in the nucleus, or (ii) the actions of D9 and 193 D10 decapping enzymes to reduce host mRNA [72,73].  To dissect how F14 functions, its impact on three hallmarks of NF-κB signalling were studied: 197 namely, degradation of IκBα, phosphorylation of p65 at S536 and p65 nuclear translocation.

198
A cell line that expresses F14 inducibly upon addition of doxycycline was used to study the 199 degradation of IκBα, and phosphorylation and nuclear translocation of p65 following 200 stimulation with TNF-α. IκBα degradation was evident 15 min after stimulation and its re-201 synthesis had started by 30 min, but neither process was influenced by F14 ( Figure 2A). F14 202 also did not affect phosphorylation of p65 at S536 (Figure 2A) or p65 translocation into the 203 nucleus as measured by immunofluorescence ( Figure 2B and Figure S4). In contrast, VACV 204 protein B14 inhibited translocation efficiently as reported [64], and VACV protein C6, an IFN 205 antagonist [65, 70], did not ( Figure 2B). 206 Next, the NF-κB inhibitory activity of F14 was tested by reporter gene assay following pathway 207 activation by p65 overexpression. In contrast to B14, F14 inhibited p65-mediated activation in 208 a dose-dependent manner without affecting p65 levels ( Figure 2C). Altogether, these results 209 showed that F14 blocks NF-κB in the nucleus at or downstream of p65. F14 thus fits the criteria 210 described previously for the unknown inhibitor of NF-κB encoded by VACV and expressed by 211 vv811ΔA49 [ κB activation [34,38], is occupied by the negatively charged residue D59 ( Figure 3A). The 241 negative charge of F14-D59 closely resembles the negative charge conferred by 242 phosphorylation of p65-S536 during NF-κB activation [34]. 243 These observations and the key role of CBP in NF-kB-dependent gene activation [20] 244 prompted investigation of whether F14 could interact with CBP. Immunoprecipitation (IP) of 245 HA-tagged F14 co-precipitated CBP-FLAG from HEK 293T cells ( Figure 3B). Reciprocal IP 246 experiments showed that ectopic CBP co-precipitated F14-HA, but not GFP-HA, with or 247 without prior TNF-α stimulation ( Figure 3C). These interactions were also seen at endogenous 248 levels in both HEK 293T and HeLa cells infected with vF14-TAP. F14, but not C6, co-249 precipitated endogenous CBP ( Figure 3D). 250 To test whether the C terminus of F14 mediated transactivation via its binding to CBP, F14 aa 251 51 to 73 were fused to the C terminus of a p65 mutant lacking the TA1 subdomain of the TAD 252 (ΔTA1) and the fusion protein was tested in a NF-κB reporter gene assay. Compared to 253 wildtype p65, the p65ΔTA1 mutant was impaired in its transactivating activity, which was 254 restored to wildtype levels upon fusion to F1451-73 ( Figure 3E). This result argues strongly that 255 the C terminus of F14 mimics the TA1 of p65 and this mimicry might explain how F14 inhibits 256 NF-κB activation. 257

F14 outcompetes NF-κB for binding to CBP 259
The similarity between the C termini of F14 and p65 led us to investigate if conserved aa 260 residues contributed to the NF-κB inhibitory activity of F14. Based on the structure of CBP KIX 261 domain in complex with p65 TA1 [79], residues of the F14 TAD-like domain corresponding to 262 residues of p65 important for its transcriptional activity and binding to CBP were mutated. 263 Three sites were altered by site-directed mutagenesis: the dipeptide D62/63, and the following 264 L65 and L68 of the ΦXXΦΦ motif. F14 L65A or L68A still inhibited NF-κB ( Figure 4A), although 265 the L65A mutant was slightly impaired. In contrast, mutation of D62/63 to either alanine 266 (D62/63A) or lysine (D62/63K) abolished the inhibitory activity ( Figure 4A). Protein levels were 267 comparable across the different F14 mutants ( Figure 4A). The loss of NF-κB inhibitory activity 268 of D62/63A and D62/63K mutants correlated with their reduced capacity to co-precipitate CBP, 269 whereas L65A and L68A mutants co-precipitated CBP to the same extent as wildtype F14 270 ( Figure 4B). The mutation of the negatively charged D62/63 to positively charged lysine 271 residues was more efficient in disrupting the interaction between F14 and CBP than only 272 abolishing the charge ( Figure 4B). Collectively, these results highlight the importance of the 273 negatively charged dipeptide D62/63 within the TAD-like domain for NF-κB inhibition by F14. 274 Next, we tested if F14 could disrupt the interaction of p65 with its coactivator CBP [20]. HEK 275 293T cells were transfected with vectors expressing p65 and CBP or RIG-I (negative control), 276 and VACV proteins F14 or C6. The amount of p65-HA immunoprecipitated by ectopic CBP 277 was reduced by increasing amounts of F14 but not C6 ( Figure 5A, B). Quantitative analysis 278 showed equivalent ectopic CBP immunoprecipitation with or without F14 ( Figure 5C).

279
Furthermore, the mutation D62/63K diminished the capacity of F14 to disrupt the interaction 280 of CBP and p65 ( Figure 5D). This observation correlated well with the reduced capacity of the 281 D62/63K mutant to co-precipitate CBP ( Figure 4B). 282

F14 suppresses expression of a subset of NF-κB-responsive genes 284
To address the impact of F14 on the induction of endogenous NF-κB-responsive genes by 285 TNF-α, the cell line inducibly expressing F14 was utilised. NF-κB-responsive genes display 286 different temporal kinetics upon activation, with "early" gene transcripts peaking between 30 -287 60 min after stimulation before declining, whilst "late" gene transcripts accumulate slowly and 288 progressively, peaking 3 h post stimulation [15,16]. When F14 expression was induced, 289 mRNAs of NFKBIA and CXCL8 "early" genes had equivalent induction kinetics compared to 290 uninduced cells ( Figure 6A, B, I). The lack of inhibition of F14 on the expression of NFKBIA 291 mRNA is in agreement with the previous finding that the re-synthesis of IκBα (NFKBIA protein 292 product) is unaffected by F14 after its proteasomal degradation induced by TNF-α ( Figure 2A). 293 Conversely, F14 induction inhibited the accumulation of the mRNAs of CCL2 and CXCL10 294 "late" genes in response to TNF-α ( Figure 6D, E). Similar results were observed when the 295 F14-expressing cell line was compared to the cell line inducibly expressing C6 ( Figure S5). 296 CXCL8 and CXCL10 encode chemokines CXCL8 and CXCL10 (also known as IL-8 and IP-297 10, respectively). Following induction of VACV protein expression, the levels of these secreted 298 chemokines were measured by ELISA and showed that levels of CXCL10, but not CXCL8, 299 was inhibited by F14. In contrast, the secretion of both chemokines was inhibited, or 300 unaffected, by VACV proteins B14 or C6, respectively, as expected ( Figure 6C, F, J, I and 301 Figure S6). Thus, unlike other VACV NF-κB inhibitors, F14 is selective and inhibits only a 302 subset of NF-κB-responsive genes. 303 Differential regulation of transcription activation downstream of NF-κB has been ascribed to 304 the recruitment of BRD4 to some NF-κB-dependent inflammatory genes [39]. Via its 305 bromodomains 1 and 2, BRD4 docks onto acetylated histones and non-histone proteins and 306 recruits transcriptional regulatory complexes to chromatin [reviewed by [83,84]]. The specific 307 recognition of acetyl-lysine residues by BRD4 is competitively inhibited by small-molecule BET 308 bromodomain inhibitors, such as JQ1 [85]. Therefore, to gain more insight into the mechanism 309 underpinning the selective inhibition of inflammatory genes by F14, the effect of JQ1 on the 310 inducible expression of CXCL8 and CXCL10 was investigated. Following TNF-α stimulation, 311 JQ1 inhibited the secretion of CXCL10, but not CXCL8, phenocopying the selective inhibition 312 of inflammatory protein expression by F14 ( Figure 6G, H). 313 314

Acetylation of p65 and recruitment of BRD4 are inhibited by F14 315
Posttranslational modifications of p65 accompany NF-κB translocation to the nucleus and 316 some, such as acetylation by acetyltransferases CBP and p300, are associated with increased 317 transcriptional activity [[35, 36, 38]; reviewed by [13,32,33]]. F14 did not interfere with the 318 phosphorylation of p65 at S536 (Figure 2A), so the acetylation of p65-K310 was investigated. 319 Cell lines that express F14 inducibly or contain the empty vector (EV) control were transfected 320 with plasmids expressing p65 and CBP in the presence of the inducer, doxycycline. Although 321 both cell lines expressed equivalent amounts of ectopic p65 and CBP, the amount of p65 322 acetylated at K310 was greatly diminished by F14 ( Figure 7A). Quantitative analysis showed 323 acetylated p65 was reduced 90% by F14 ( Figure 7B). This result, together with data in Figure  324 5, indicated that the reduced acetylation of p65 was due to disruption of the interaction 325 between p65 and CBP by F14. 326 Acetylated K310 on p65 serves as a docking site for the bromodomains 1 and 2 of BRD4, 327 which then recruits P-TEFb to promote RNAP II elongation during transcription of some NF-328 κB-responsive genes [39]. The differential sensitivity of TNF-α-stimulated genes to the 329 inhibition of NF-κB by F14 might reflect the differential requirement of p65 acetylated at K310, 330 and the subsequent recruitment of BRD4, to activate the expression from NF-κB-responsive 331 promoters [39]. This hypothesis was tested by chromatin immunoprecipitation with an anti-332 BRD4 antibody followed by quantitative PCR of the promoters of four representative genes: 333 NFKBIA and CXCL8, resistant to F14 inhibition, and CCL2 and CXCL10, sensitive to inhibition.

334
BRD4 was recruited to these promoters after TNF-α stimulation, with BRD4 present on 335 NFKBIA and CXCL8 promoters at 1 and 5 h post-stimulation, whereas BRD4 was more 336 enriched on CCL2 and CXCL10 promoters only at 5 h post-stimulation, mirroring the kinetics 337 of mRNA accumulation (Figure 7B-F; see Figure 6A, B, D, E). In the presence of F14, the 338 inducible recruitment of BRD4 to the NFKBIA and CXCL8 promoters remained unaffected, 339 whilst its recruitment to CCL2 and CXCL10 was blocked ( Figure 7C). This strongly suggests 340 that inhibition of acetylation of p65 at K310 by F14 is relayed downstream to the recruitment 341 of BRD4 to the "F14-sensitive" promoters, but not to the "F14-resistant" promoters. 342 The BRD4 recruitment to NFKBIA and CXCL8 promoters despite inhibition of p65-K310 343 acetylation prompted investigation of whether other acetyl-lysine residues are recognised. The 344 bromodomain-mediated docking onto acetylated lysine residues is generally accepted as 345 responsible for the recruitment of BRD4 to the chromatin [reviewed by [83,84]]. For instance, 346 histone 4 acetylated on K5, K8 and K12 (H4K5/K8/K12ac) is responsible for BRD4 recruitment 347 to NF-κB-responsive genes upon lipopolysaccharide stimulation [86]. The recruitment of 348 BRD4 to the NFKBIA, CXCL8, CCL2, and CXCL10 promoters was tested in the presence of 349 the bromodomain inhibitor JQ1, by chromatin immunoprecipitation and quantitative PCR.

350
BRD4 was still recruited to NFKBIA and CXCL8 promoters after TNF-α stimulation in the 351 presence of JQ1, whilst inducible recruitment to CCL2 and CXCL10 promoters was abolished 352 by JQ1 ( Figure 7H-K). As a control for JQ1 pharmacological activity, BRD4 recruitment to the 353 CCND1 gene promoter was diminished by this small-molecule inhibitor ( Figure S7). CCND1 354 is a BRD4 target gene that encodes the cell cycle regulator cyclin D1 and was used as a 355 positive control [87]. Altogether, these results suggest that the inducible recruitment of BRD4 356 to some promoters is independent of the bromodomains. 357

F14 is unique among known viral antagonists of NF-κB 359
The TAD domain of p65 belongs to the class of acidic activation domains, characterised by a 360 preponderance of aspartic acid or glutamic acid residues surrounding hydrophobic motifs [22].

361
VP16 is a transcriptional activator from HSV-1 that bears a prototypical acidic TAD ( Figure  362 8A) and inhibits the expression of virus-induced IFN-β by association with p65 and IRF3 [44].

363
Although the VP16-mediated inhibition of the IFN-β promoter was independent of its TAD, we 364 revisited this observation to investigate the effect of VP16 more specifically on NF-κB-365 dependent gene activation. VP16 inhibited NF-κB reporter gene expression in a dose-366 dependent manner and deletion of the TAD reduced NF-κB inhibitory activity of VP16 about 367 2-fold, but some activity remained ( Figure 8A). 368 A search for other viral proteins that contain motifs resembling the ΦXXΦΦ motif present in 369 acidic transactivation activation domains detected a divergent ΦXXΦΦ motif in protein E7 (aa 370 79-83) from HPV16, with acidic residues upstream (D75) or within (E80 and D81) the motif 371 ( Figure 8B). E7 has been reported to inhibit NF-κB activation, in addition to its role in promoting 372 cell cycle progression [43,[88][89][90]. We confirmed that HPV16 protein E7 inhibits NF-κB-373 dependent gene expression ( Figure 8B). Furthermore, E7 mutants harbouring aa substitutions 374 that inverted the charge of D75 (D75K) or added a positive charge to the otherwise 375 hydrophobic L83 (L83R) were impaired in their capacity to inhibit NF-κB ( Figure 8B). 376 Lastly, the ability of VP16 and E7 to associate with CBP was assessed after ectopic 377 expression in HEK 293T cells. Neither VP16 nor E7, like VACV protein C6 used as negative 378 control, co-precipitated CBP under conditions in which F14 did ( Figure 8C). These findings 379 indicate that the mimicry of p65 TAD by F14 is a strategy unique among human pathogenic 380 viruses to suppress the activation of NF-κB. 381

DISCUSSION 383
The inducible transcription of NF-κB-dependent genes is a critical response to virus infection. 384 After binding to κB sites in the genome, NF-κB promotes the recruitment of chromatin 385 remodelling factors, histone-modifying enzymes, and components of the transcription 386 machinery. Thereby, NF-κB couples the sensing of viral and inflammatory signals to the 387 selective activation of the target genes. In response, viruses have evolved multiple immune 388 evasion strategies, including interference with NF-κB activation. VACV is a paradigm in viral 389 evasion mechanisms, inasmuch as this poxvirus encodes 15 proteins known to intercept NF-390 κB activation downstream of PRRs and cytokine receptors [reviewed by [55,56], [91,92]]. 391 Nonetheless, a VACV strain lacking all these inhibitors still prevented NF-κB activation after 392 p65 translocation into the nucleus [57], indicating the existence of other inhibitor(s). 393 Here, VACV protein F14, which is conserved in all orthopoxviruses, including ancient variola 394 viruses, is shown to inhibit NF-κB activation within the nucleus and its mechanism of action is 395 elucidated. First, ectopic expression of F14 reduces NF-κB-dependent gene expression 396 stimulated by TNF-α or IL-1β ( Figure 1A, B). Second, F14 is expressed early during VACV 397 infection, and is small enough (8 kDa) to diffuse passively into the nucleus ( Figure 1E, H). 398 Third, a VACV strain lacking both A49 and F14 (vv811ΔA49ΔF14) is less able to suppress 399 cytokine-stimulated NF-κB-dependent gene expression than vv811ΔA49 ( Figure 1I). Fourth, 400 following TNF-α stimulation, IκBα degradation, IKK-mediated phosphorylation of p65 at S536 401 and p65 accumulation in the nucleus remained unaffected in the presence of F14 (Figure 2A, 402 B). Lastly, F14 blocked NF-κB-dependent gene expression stimulated by p65 overexpression, 403 indicating that it acts at or downstream of p65 ( Figure 2C). Mechanistically, F14 inhibits NF-404 κB via a C-terminal 23 aa motif that resembles the acidic activation domain of p65. F14 405 disrupts the binding of p65 to its coactivator CBP (Figures 4 and 5) and reduces acetylation of 406 p65 K310. Subsequently, F14 inhibits the inducible recruitment of BRD4 to CCL2 and CXCL10 407 promoters, but not to NFKBIA and CXCL8 promoters (Figure 7). These findings correlated 408 with F14 suppressing CCL2 and CXCL10, but not NFKBIA and CXCL8, mRNA expression 409 ( Figure 6A, B, D, E). The selective inhibition of a subset of NF-κB-dependent genes by F14, 410 despite the interference with molecular events deemed important for p65-mediated 411 transactivation, underscores the complexity of the nuclear actions of NF-κB. Initial 412 understanding of NF-κB-mediated gene activation was derived mostly using artificial reporter 413 plasmids, but subsequent genome-wide, high-throughput studies uncovered diverse 414 mechanisms of gene activation [14,16,24,28,29,86,93]. Because multiple promoters 415 containing κB sites are preloaded with CBP/p300, RNAP II and general transcription factors, 416 the activation of transcription by NF-κB relies on the recruitment of BRD4 [86,93]. 417

Recruitment of BRD4 to promoters and enhancers occurs via bromodomain-mediated docking 418
onto acetyl-lysine residues on either histones or non-histone proteins and promotes chromatin 419 remodelling and transcription [reviewed by [83,84]]. For NF-κB-bound promoters, BRD4 420 recognises p65 acetylated at K310 [39]. This explains how F14 reduced inducible enrichment 421 of BRD4 on the CCL2 and CXCL10 promoters following TNF-α stimulation: namely, reduced 422 acetylation of p65 by CBP ( Figure 7A, B, E, F). Nonetheless, BRD4 enrichment on the NFKBIA 423 and CXCL8 promoters remained unaffected in the presence of F14 ( Figure 7C, D). Genes 424 whose expression is resistant to F14 inhibition might be activated independently of the p65 425 TA1 domain, as is the case for some NF-κB-responsive genes in mouse fibroblasts stimulated 426 with TNF-α, including Nfkbia. For those genes, p65 occupancy on the promoter elements 427 suffices for gene activation, via recruitment of secondary transcription factors [24]. However, 428 BRD4 enrichment on NFKBIA and CXCL8 promoters also remained unaffected in the 429 presence of the bromodomain inhibitor JQ1 ( Figure 7H, I), indicating alternative mechanism(s) 430 of BRD4 recruitment to some promoters. Downstream of p65, alternative recruitment via 431 protein-protein interactions through the C-terminal domains of BRD4 might mediate BRD4 432 recruitment to some NF-κB-bound promoters independently of the recognition of acetyl-lysine 433 residues by the N-terminal bromodomains, which is recognised as the main mechanism of 434 BRD4 recruitment to chromatin [reviewed by [83,84]]. For instance, BRD4 interacts directly 435 with multiple transcription factors and chromatin remodellers independently of acetylation [94]. 436 Further investigation of acetylation-independent recruitment of BRD4 to inducible promoters 437 observed here and elsewhere [95] is warranted. 438 In the nucleus, p65 engages with multiple binding partners via its transactivation domains, 439 including the direct interactions between TA1 and TA2 and the KIX and transcriptional adaptor 440 zinc finger (TAZ) 1 domains of CBP, respectively. These interactions are mediated by 441 hydrophobic contacts of the ΦXXΦΦ motifs and complemented by electrostatic contacts by 442 the acidic residues in the vicinity of the hydrophobic motifs [79,93]. Sequence analysis 443 suggested that F14 mimics the p65 TA1 domain ( Figure 3A). Indeed, fusion of the TAD-like 444 domain of F14 to a p65 mutant lacking the TA1 domain restored its transactivation activity to 445 wildtype levels ( Figure 3E). This explains the observation from a yeast two-hybrid screen of 446 VACV protein-protein interactions, in which F14 could not be tested because it was found to 447 be a strong activator when fused to the GAL4 DNA-binding domain [96]. Site-directed 448 mutagenesis of F14 revealed that the dipeptide D62/63, but not L65 or L68 of the ΦXXΦΦ 449 motif, is required for inhibition of NF-κB ( Figure 4A), for interaction with CBP ( Figure 4B) and 450 for the efficient disruption of p65 binding to CBP ( Figure 5D). This contrasts with the molecular 451 determinants of p65 TA1 function, i.e., both hydrophobic (F542) and acidic (including D539 452 and D541) residues contribute to p65 TA1 transactivation activity [80,81]. Although the p65 453 TA1 binding to the KIX domain of CBP was shown to depend on F542, the importance of the 454 electrostatic interactions by D539 and D541 is yet to be tested [79]. Of note, a recent high-455 throughput mutagenesis analysis of a model acidic activation domain provided useful insight 456 into the relative contributions of hydrophobic and acidic residues for transcriptional activity.

457
This analysis supports a model in which key hydrophobic residues require the acidic residues 458 to keep them exposed to solvent where they can interact with coactivators [97]. We cannot 459 rule out that F14 function depends on other C-terminal hydrophobic residues, but our 460 observation that F14-D62/63K (F14-D62 aligns with p65-D539) mutant is impaired in 461 disrupting p65-KIX interaction in cells is in line with the hypothesis of how acidic activation 462 domains work. Future elucidation of the structure of F14-KIX complex and its comparison with 463 p65 TA1-KIX co-structure will be necessary to address this apparent discrepancy. The 464 "imperfect" nature of F14 mimicry is not without precedent in poxviruses. VACV protein A49, 465 the mimic of IκBα phosphodegron, contains an extra aa residue between the two 466 phosphorylatable serine residues of the degron and requires the phosphorylation of just one 467 of the two serines to interact with the E3 ligase β-TrCP and thus to prevent IκBα degradation 468 [54]. 469 The diminished acetylation of p65 K310 is a direct consequence of the disruption of CBP p65 TA1 to bind to CBP and prevent its interaction with p65. HPV16 E6 also disrupts the 486 interaction of CBP with p65 but, unlike F14, E6 lacks a ΦXXΦΦ motif surrounded by acidic 487 residues and inhibits the expression of CXCL8 and therefore is mechanistically distinct [43]. 488 After searching for additional viral proteins that might mimic p65 TAD, we focused on HPV16 489 E7 and HSV-1 VP16. The latter protein has a prototypical acidic TAD ( Figure 8A), the former 490 bears a motif resembling the ΦXXΦΦ motif ( Figure 8B), and both proteins inhibit NF-κB 491 activation [43,44,[88][89][90]. Data presented here confirm that VP16 and E7 each inhibit NF-κB-492 dependent gene expression ( Figure 8A, B). However, neither co-precipitated CBP under 493 conditions in which F14 did ( Figure 8C) Overall, our search for additional inhibitors of NF-κB activation encoded by VACV unveiled a 501 viral strategy to inhibit this transcription factor that is unique among known viral antagonists of 502 NF-κB. By mimicking the TA1 domain of p65, F14 disrupts the interaction between p65 and its 503 coactivator CBP, thus inhibiting the downstream molecular events that trigger the activation of 504 a subset of inflammatory genes in response to cytokine stimulation. Among these events, the 505 recruitment of RNAP II processivity factor BRD4 is important for induction of the inflammatory 506 response. This study also showed BRD4 is recruited to some inducible NF-κB-dependent 507 promoters independently of the recognition of acetylated chromatin (i.e., acetyl-lysine 508 residues), via an unknown mechanism that warrants further investigation. Two lines of 509 evidence illustrated the biological importance of F14. First, F14-D62/63 site is conserved in 510 F14 orthologues from different orthopoxviruses, including human pathogens cowpox and 511 monkeypox viruses, and ancient (10 th century CE) and modern variola virus strains ( Figure  512 3A). Second, a VACV strain lacking F14 is attenuated in an intradermal model of infection 513 ( Figure 1F), despite the presence of several other VACV-encoded NF-κB inhibitors [reviewed 514 by [55,56]]. The attenuation of vΔF14 also shows the function of F14 is not redundant with 515 these other inhibitors of NF-κB, despite the selective inhibition imparted by F14 ( Figure 6A-F). 516 From the viral perspective, the selective inhibition of only a subset of NF-κB-responsive genes 517 by F14 might represent an adaptation to counteract the host immune response more 518 efficiently. If an NF-κB-activating signal reached the nucleus of an infected cell, maintaining 519 expression of some NF-κB-dependent genes, particularly NFKBIA, might promote the signal 520 termination by IκBα. Newly synthesised IκBα not only tethers cytoplasmic NF-κB, but can also 521 remove NF-κB from the DNA and cause its export from the nucleus [17,18,109]. We anticipate 522 that other viruses might also use the selective inhibition of NF-κB to exploit the pro-viral 523 functions of active NF-κB whilst dampening its pro-inflammatory and antiviral activities.  and Gerd Blobel (University of Pennsylvania, Philadelphia, USA) for providing us with 532 reagents. We are also grateful to Tony Kouzarides for helpful advice and to Callum Talbot-533 Cooper for critical reading of the manuscript.  Copenhagen, if sequences diverged between strains) were codon-optimised for expression in 586 human cells and synthesised by GeneArt (Thermo Fisher Scientific), with an optimal 5' Kozak 587 sequence and fused to an N-terminal FLAG epitope. For ease of subsequent subcloning, 5' 588 BamHI and 3' XbaI restriction sites were included as well as a NotI site + 1G between the 589 epitope tag and the ORF. The NotI site + 1G generates an (Ala)3 linker between the epitope 590 tag and the protein of interest. For mammalian expression, nucleotide sequences encoding 591 N-terminal FLAG-tagged VACV proteins were subcloned between the BamHI and XbaI 592 restriction sites of a pcDNA4/TO vector (Invitrogen). Alternatively, codon-optimised F14 was 593 PCR-amplified to include a 3' HA tag or a 3' FLAG or no epitope tag, and 5' BamHI and 3' 594 XbaI sites to clone into pcDNA4/TO plasmid. In addition, codon-optimised F14 sequence was 595 PCR-amplified to include 5' BamHI and 3' NotI sites to facilitate cloning into a pcDNA4/ respectively. 621 The oligonucleotide primers used for cloning and site-directed mutagenesis are listed in Table  622 S2. Nucleotide sequences of the inserts in all the plasmids were verified by Sanger DNA 623 sequencing. supplemented with zeocin (100 µg/mL, Gibco). 637 The absence of mycoplasma contamination in the cell cultures was tested routinely with 638 MycoAlert detection kit (Lonza), following the manufacturer's recommendations. 639 640

Construction of recombinant viruses 641
A VACV Western Reserve (WR) strain lacking F14 (vΔF14)  The oligonucleotide primers used to generate the recombinant VACV strains are listed in Table  681 S2. To verify that all the final recombinant viruses harboured the correct sequences, PCR 682 fragments spanning the F14L locus were sequenced.

Virus growth and spread assays 698
To analyse virus growth properties in cell culture, single-step growth curve experiments were 699 performed in HeLa cells. Cells were grown to about 90% confluence in T-25 flasks and then 700 infected at 5 p.f.u./cell in growth medium supplemented with 2% FBS. Virus adsorption was at 701 37⁰C for 1 h. Then the inoculum was removed, and the cells were replenished with growth 702 medium supplemented with 2% FBS. At 1, 8, and 24 h p.i., infected-cell supernatants and 703 monolayers were collected for determination of extracellular and cell-associated infectious 704 virus titres, respectively, by plaque assay on BS-C-1 cells. Supernatants were clarified by 705 centrifugation to remove cellular debris and detached cells, whereas cell monolayers were 706 scraped and disrupted by three cycles of freezing/thawing followed by sonication, to release 707 intracellular virus particles. 708 The virus spread in cell culture was assessed by plaque formation. Confluent monolayers of reporter plasmid (NF-κB, ISRE, or AP-1), TK-Renilla luciferase reporter plasmid (as an internal 727 control) and the desired expression vectors or empty vector (EV) using (v/v) Tween-20, membranes were probed with fluorophore-conjugated secondary antibodies 797 (LI-COR Biosciences) diluted in 5% (w/v) non-fat milk at room temperature for 1 h. After 798 washing, membranes were imaged using the Odyssey CLx imaging system (LI-COR 799 Biosciences), according to the manufacturer's instructions. For quantitative analysis of protein 800 levels, the band intensities on the immunoblots were quantified using the Image Studio 801 software (LI-COR Biosciences). The antibodies used for immunoblotting are listed in Table  802 S3. 803 804

Co-immunoprecipitation and pulldown assays 805
HEK 293T or HeLa cells in 10-cm dishes were infected at 5 p.f.u./cell for 8 h or transfected 806 overnight with the specified epitope-tagged plasmids using polyethylenimine (PEI, 807 Polysciences, 2 μl of 1 mg/ml stock per μg of plasmid DNA). For the competition assays, cells 808 were starved of FBS for 3 h and stimulated with TNF-α (40 ng/ml, PeproTech) in FBS-free 809 DMEM for 15 min before harvesting. Cells were washed with ice-cold PSB, scraped in 810 immunoprecipitation (IP) buffer [50 mM Tris-HCl pH 7.4, 150 mM NaCl, 0.5% (v/v) NP-40, 0.1 811 mM EDTA], supplemented with protease (cOmplete Mini, Roche) and phosphatase 812 (PhosSTOP, Roche) inhibitors, on ice, transferred to 1.5-ml microcentrifuge tubes and rotated 813 for 30 min at 4⁰C. Cell lysates were centrifuged at 17,000 × g for 15 min at 4⁰C and the soluble 814 fractions were incubated with 20 μl of one of the following affinity resins equilibrated in IP 815 buffer: (i) anti-FLAG M2 agarose (Sigma-Aldrich, Cat# A2220) for IP of FLAG-or TAP-tagged 816 proteins; (ii) anti-HA agarose (Sigma-Aldrich, Cat# A2095) for IP of HA-tagged proteins; or (iii) 817 Strep-Tactin Superflow agarose (IBA, Cat# 2-1206-025) for pulldown of TAP-tagged protein 818 via Strep-tag II epitope. After 2 h of rotation at 4⁰C, the protein-bound resins were washed 819 three times with ice-cold IP buffer. The bound proteins were eluted by incubation with 2× SDS-820 gel loading buffer and boiled at 100⁰C for 5 min before analysis by SDS-polyacrylamide gel 821 electrophoresis and immunoblotting, along with 10% input samples collected after clarification 822 of cell lysates. The antibodies used for immunoprecipitation are listed in Table S3. 823 824

Reverse transcription and quantitative PCR 825
To analyse mRNA expression of NF-κB-responsive genes, T-REx-293-F14 in 12-well plates 826 were left uninduced or induced overnight with 100 ng/ml doxycycline (Melford, UK) to induce 827 the expression of F14. Alternatively, T-REx-293-F14 and C6 in 12-well plates were induced 828 overnight with 100 ng/ml doxycycline (Melford, UK) to induce the expression of the VACV 829 proteins. The next day, cells were starved for 3 h by removal of serum from the medium and 830 then stimulated in duplicate with TNF-α (40 ng/ml, PeproTech) in FBS-free DMEM for 0, 1 or 831 6 h. RNA was extracted using RNeasy Mini Kit (Qiagen) and complementary DNA (cDNA) 832 was synthesised using SuperScript III reverse transcriptase (Invitrogen) and oligo-dT primers 833 (Thermo Scientific), according to the instructions of the respective manufacturers. The mRNA 834 levels of CCL2, CXCL8, CXCL10, GAPDH and NFKBIA were quantified by quantitative PCR 835 using gene-specific primer sets, fast SYBR green master mix (Applied Biosystems) and the 836 ViiA 7 real-time PCR system (Life Technologies). The oligonucleotide primers used for the 837 qPCR analysis of gene expression are listed in Table S2. Fold-induction of the NF-κB-838 responsive genes was calculated by the 2 -ΔΔCt method using non-induced and non-stimulated 839 T-REx-293-F14 cells, or induced and non-stimulated T-REx-293-C6 cells, as the reference 840 sample, and GAPDH as the housekeeping control gene. washed twice with ice-cold PBS and fixed in 4% (v/v) paraformaldehyde for 10 min. After 861 quenching of free formaldehyde with 150 mM ammonium chloride for 5 min, the fixed cells 862 were permeabilised with 0.1% (v/v) Triton X-100 in PBS for 5 min and blocked with 10% (v/v) 863 FBS in PBS for 30 min. Staining was carried out with primary antibodies for 1 h, followed by 864 incubation with the appropriate AlexaFluor fluorophore-conjugated secondary antibodies 865 (Invitrogen Molecular Probes) for 30 min and mounting onto glass slides with Mowiol 4-88 866 (Calbiochem) containing 0.5 μg/ml DAPI (4',6-diamidino-2-phenylindole, Biotium). Images 867 were acquired on an LSM 700 confocal microscope (Zeiss) using ZEN system software 868 (Zeiss). Quantification of nuclear localisation of p65 was done manually on the ZEN lite 869 software (blue edition, Zeiss). The details about the antibodies used for immunofluorescence 870 are listed in Table S3. Cells were crosslinked with 1% (v/v) formaldehyde added directly to the growth medium. After 892 10 min at room temperature, crosslinking was stopped by the addition of 0.125 M glycine. Cells 893 were then lysed in 0.2% NP-40, 10 mM Tris-HCl pH 8.0, 10 mM NaCl, supplemented with 894 protease (cOmplete Mini, Roche), phosphatase (PhosSTOP, Roche) and histone deacetylase 895 (10 mM sodium butyrate, Sigma-Aldrich) inhibitors, and nuclei were recovered by 896 centrifugation at 600 × g for 5 min at 4⁰C. To prepare the chromatin, nuclei were lysed in 1% 897 (w/v) SDS, 50 mM Tris-HCl pH 8.0, 10 mM EDTA, plus protease/phosphatase/histone 898 deacetylase inhibitors, and lysates were sonicated in a Bioruptor Pico (Diagenode) to achieve 899 DNA fragments of about 500 bp. After sonication, samples were centrifuged at 3,500 × g for 900 10 min at 4⁰C and supernatants were diluted four-fold in IP dilution buffer [20 mM Tris-HCl pH 901 8.0, 150 mM NaCl, 2 mM EDTA, 1% (v/v) Triton X-100, 0.01% (w/v) SDS] supplemented with 902 protease/phosphatase/histone deacetylase inhibitors. 903 Protein G-conjugated agarose beads (GE Healthcare, Cat# 17-0618-02) equilibrated in IP 904 dilution buffer were used to preclear the chromatin for 1 h at 4⁰C with rotation. Before the 905 immunoprecipitation, 20% of the precleared chromatin was kept as input control. 906 Immunoprecipitation was performed with 8 µg of anti-BRD4 antibody (Cell Signalling 907 Technology, #13440) or anti-GFP (Abcam, #ab290), used as negative IgG control, overnight 908 at 4⁰C with rotation. Protein-DNA immunocomplexes were retrieved by incubation with 60 µl 909 of equilibrated protein G-conjugated agarose beads (GE Healthcare), for 2 h at 4⁰C, followed 910 by centrifugation at 5,000 × g for 2 min at 4⁰C. were reversed by incubation overnight at 67⁰C in presence of 1 µg of RNase A and 300 mM 917 NaCl, followed by proteinase K digestion for 2 h at 45⁰C. Co-immunoprecipitated DNA 918 fragments were purified using the QIAquick PCR purification kit (Qiagen) and analysed by 919 quantitative PCR targeting the promoter elements of NFKBIA, CXCL8, CCL2, and CXCL10 920 genes. The oligonucleotide primers used for the qPCR analysis of ChIP are listed in Table S2. 921 The primers target regions containing consensus κB sites (5'-GGGRNYYYCC-3′, in which R 922 is a purine, Y is a pyrimidine, and N is any nucleotide), prioritising amplicons overlapping areas 923 with histone modification often observed near active regulatory elements (H3K27ac) according 924 to ENCODE Histone Modification database on UCSC Genome Browser 925 (https://genome.ucsc.edu/index.html). Some primers have been described previously [15, 926 119, 120]. 927 The ChIP-qPCR data were analysed by the fold enrichment method. Briefly, the signals 928 obtained from the ChIP with each antibody were first normalised to the signals obtained from 929 the corresponding input sample (ΔCt = CtIP -CtInput). Next, the input-normalised signals (ΔCt) 930 were normalised to the corresponding 0 time-point control (i.e. ΔΔCt = ΔCt -ΔCt0  proteins are labelled with the protein name followed the epitope tag antiboby in parentheses.

1049
When multiple tagged proteins are shown in the same immunoblot, each protein is indicated 1050 by a red arrowhead. Statistical significance was determined by the Student's t-test.  sequence. If present, sequences coding the tag epitopes are highlighted in bold, whilst the 1148 Kozak sequence are shown in italics. Plasmids marked with an asterisk (*) were constructed 1149 by site-directed mutagenesis, with the mutated codons underlined. 1150