The enzymatic activity of the nsp14 exoribonuclease is critical for replication of Middle East respiratory syndrome-coronavirus

Coronaviruses (CoVs) stand out for their large RNA genome and complex RNA-synthesizing machinery comprising 16 nonstructural proteins (nsps). The bifunctional nsp14 contains an N-terminal 3’-to-5’ exoribonuclease (ExoN) and a C-terminal N7-methyltransferase (N7-MTase) domain. While the latter presumably operates during viral mRNA capping, ExoN is thought to mediate proofreading during genome replication. In line with such a role, ExoN-knockout mutants of mouse hepatitis virus (MHV) and severe acute respiratory syndrome coronavirus (SARS-CoV) were previously found to have a crippled but viable hypermutation phenotype. Remarkably, using an identical reverse genetics approach, an extensive mutagenesis study revealed the corresponding ExoN-knockout mutants of another betacoronavirus, Middle East respiratory syndrome coronavirus (MERS-CoV), to be non-viable. This is in agreement with observations previously made for alpha- and gammacoronaviruses. Only a single MERS-CoV ExoN active site mutant could be recovered, likely because the introduced D191E substitution is highly conservative in nature. For 11 other MERS-CoV ExoN active site mutants, not a trace of RNA synthesis could be detected, unless – in some cases – reversion had first occurred. Subsequently, we expressed and purified recombinant MERS-CoV nsp14 and established in vitro assays for both its ExoN and N7-MTase activities. All ExoN knockout mutations that were lethal when tested via reverse genetics were found to severely decrease ExoN activity, while not affecting N7-MTase activity. Our study thus reveals an additional function for MERS-CoV nsp14 ExoN, which apparently is critical for primary viral RNA synthesis, thus differentiating it from the proofreading activity thought to boost long-term replication fidelity in MHV and SARS-CoV. Importance The bifunctional nsp14 subunit of the coronavirus replicase contains 3’-to-5’ exoribonuclease (ExoN) and N7-methyltransferase (N7-MTase) domains. For the betacoronaviruses MHV and SARS-CoV, the ExoN domain was reported to promote the fidelity of genome replication, presumably by mediating some form of proofreading. For these viruses, ExoN knockout mutants are alive while displaying an increased mutation frequency. Strikingly, we now established that the equivalent knockout mutants of MERS-CoV ExoN are non-viable and completely deficient in RNA synthesis, thus revealing an additional and more critical function of ExoN in coronavirus replication. Both enzymatic activities of (recombinant) MERS-CoV nsp14 were evaluated using newly developed in vitro assays that can be used to characterize these key replicative enzymes in more detail and explore their potential as target for antiviral drug development.


57
The bifunctional nsp14 subunit of the coronavirus replicase contains 3'-to-5' 58 exoribonuclease (ExoN) and N7-methyltransferase (N7-MTase) domains. For the 59 betacoronaviruses MHV and SARS-CoV, the ExoN domain was reported to promote 60 the fidelity of genome replication, presumably by mediating some form of proofreading.

61
For these viruses, ExoN knockout mutants are alive while displaying an increased 62 mutation frequency. Strikingly, we now established that the equivalent knockout 63 mutants of MERS-CoV ExoN are non-viable and completely deficient in RNA 64 synthesis, thus revealing an additional and more critical function of ExoN in coronavirus

INTRODUCTION
Interestingly, a quite different phenotype was described for the corresponding ExoN-122 knockout mutants of two betacoronaviruses, mouse hepatitis virus (MHV) and  CoV. While ExoN inactivation decreased replication fidelity in these viruses, conferring 124 a 'mutator phenotype', the mutants were viable, both in cell culture (22,23) and in 125 animal models (24). These findings suggested that ExoN may indeed be part of an 126 error correction mechanism. Subsequently, the ability of ExoN to excise 3'-terminal 127 mismatched nucleotides from a double-stranded (ds) RNA substrate was 128 demonstrated in vitro using recombinant SARS-CoV nsp14 (25). Furthermore, this 129 activity was shown to be strongly enhanced (up to 35-fold) by the presence of nsp10, 130 a small upstream subunit of the CoV replicase (26). The two subunits were proposed 131 to operate, together with the nsp12-RdRp, in repairing mismatches that may be 132 introduced during CoV RNA synthesis (21,27). In cell culture, MHV and SARS-CoV 133 mutants lacking ExoN activity exhibit increased sensitivity to mutagenic agents like 5-134 fluoracil (5-FU), compounds to which the wild-type virus is relatively resistant (28,29). 135 Recently, ExoN activity was also implicated in CoV RNA recombination, as an MHV 136 ExoN knockout mutant exhibited altered recombination patterns, possibly reflecting its 137 involvement in other activities than error correction during CoV replication and 138 subgenomic mRNA synthesis (30). Outside the order Nidovirales, arenaviruses are the 139 only other RNA viruses known to employ an ExoN domain, which is part of the 140 arenavirus nucleoprotein and has been implicated in fidelity control (31) and/or immune 141 evasion, the latter by degrading viral dsRNA (32,33). Based on results obtained with 142 TGEV and MHV ExoN knockout mutants, also the CoV ExoN activity was suggested 143 to counteract innate responses (34,35). 144 In the meantime, CoV nsp14 was proven to be a bifunctional protein by the discovery 145 of an (N7-guanine)-methyltransferase (N7-MTase) activity in its C-terminal domain 146 (36) (Fig. 1). This enzymatic activity was further corroborated in vitro, using biochemical 147 assays with purified recombinant SARS-CoV nsp14. The enzyme was found capable 148 of methylating cap analogues or GTP substrates, in the presence of S-adenosyl 149 methionine (SAM) as methyl donor (36,37). The N7-MTase was postulated to be a 150 key factor for equipping CoV mRNAs with a functional 5'-terminal cap structure, as N7-151 methylation is essential for cap recognition by the cellular translation machinery (25). 152 Although, the characterization of the nsp14 N7-MTase active site and reaction 153 mechanism was not completed, alanine scanning mutagenesis and in vitro assays with 154 nsp14 highlighted several key residues ( Fig. 1) (36,38,39). Moreover, crystal 155 structures of SARS-CoV nsp14 in complex with its nsp10 co-factor (PDB entries 5C8U 156 and 5NFY) revealed several unique structural and functional features (Ma et al., 2015; 157 Ferron et al., 2018). These combined structural and biochemical studies confirmed that 158 the two enzymatic domains of nsp14 are functionally distinct (36) and physically 159 independent (20, 21). Still, the two activities are structurally intertwined, as it seems 160 that the N7-MTase activity depends on the integrity of the N-terminal ExoN domain, 161 whereas the flexibility of the protein is modulated by a hinge region connecting the two 162 domains (21). 163 Coronaviruses are abundantly present in mammalian reservoir species, including 164 bats, and pose a continuous zoonotic threat (40)(41)(42)(43). To date, seven CoVs that can 165 infect humans have been identified, and among these the severe acute respiratory continues to circulate and cause serious human disease, primarily in the Arabian Peninsula (45). Occasional spread to other countries has also occurred, including an 172 outbreak with 186 confirmed cases in South Korea in 2015 (46-48 viable, while displaying a 15-to 20-fold increased mutation rate (Eckerle, Lu et al. 2007, 203 Eckerle, Becker et al. 2010). An alignment of CoV nsp14 amino acid sequences is 204 presented in Fig. 1, including SARS-CoV-2, which emerged during the course of this 205 project. It highlights the key motifs/residues of the two enzymatic domains of nsp14, 206 as well as other structural elements, like the nsp10 binding site, the hinge region 207 connecting the ExoN and N7-MTase domains, and three previously identified zinc 208 finger domains (20,21). The alignment also illustrates the generally high degree of 209 nsp14 sequence conservation across different CoV (sub)genera.

210
In the present study, we targeted all five predicted active site residues of the MERS-

211
CoV ExoN domain (D90, E92, E191, D273, and H268) by replacing them with alanine 212 as well as more conservative substitutions (D to E or Q; E to D or Q). This yielded a 213 total of 14 ExoN active site mutants ( Fig. 2A) Transcripts were electroporated into BHK-21 cells, which lack the DPP4 receptor 220 required for natural MERS-CoV infection (Chan, Chan et al. 2013, Raj, Mou et al. 2013 but are commonly used to launch engineered CoV mutants because of their excellent 222 survival of the electroporation procedure (19,22,23,34,51,54). As BHK-21 cells have 223 a severely compromised innate immune response (55), they would seem an 224 appropriate cell line to launch ExoN knockout mutants also in case the enzyme would 225 be needed to counter innate immunity (Becares, Pascual-Iglesias et al. 2016, Case, Li 226 et al. 2018. To amplify any progeny virus released, transfected BHK-21 cells were 227 mixed with either innate immune-deficient (Vero) or -competent (HuH7) cells, which 228 both are naturally susceptible to MERS-CoV infection.

229
In stark contrast to what was previously described for MHV and SARS-CoV, 230 mutagenesis of ExoN active site residues was found to fully abrogate MERS-CoV 231 replication. When transfected cell cultures were analyzed using immunofluorescence 232 microscopy at 2 days post transfection (d p.t.), abundant signal was always observed 233 for wild-type MERS-CoV, but no sign of virus replication was observed for 13 out of 14 234 mutants tested (Fig. 2), regardless whether Vero or HUH7 cells were used for 235 propagation of recombinant virus. Furthermore, infectious progeny was not detected 236 when transfected cell culture supernatants were analyzed in plaque assays ( Fig. 2 and   237 data not shown). The single exception was the mutant carrying the conservative E191D 238 replacement in ExoN motif II (Fig. 1), which was alive but somewhat crippled, as will 239 be discussed in more detail below. These results were consistent across a large  The lack of MERS-CoV-specific RNA synthesis was further analyzed using RT-PCR 259 assays specifically detecting genomic RNA or subgenomic mRNA3. RNA   specifying an A1235D substitution in the betacoronavirus-specific marker (βSM) 287 domain of nsp3, which has been predicted to be a non-enzymatic domain (58) and is 288 absent in alpha-and delta-coronaviruses (59, 60). Thus, we assumed that any 289 changes in viral replication were likely caused by the E191D mutation in nsp14 ExoN.

290
The same virus stock was used to assess growth kinetics in HuH7 cells (Fig. 4B) and 291 Vero cells (Fig. 4A), which were found to be very similar for wt and mutant virus. Still, 292 the E191D mutant was found to be somewhat crippled, yielding smaller plaque sizes 293 and somewhat lower progeny titers in HuH7 cells (Fig. 4 B-C), but not in Vero cells 294 (Fig. 4A). 295 We next examined the sensitivity of E191D and wt virus to the mutagenic agent 5-296 FU, which intracellularly is converted into a nucleoside analogue that is incorporated 297 into viral RNA (61, 62). Previously, MHV and SARS-CoV ExoN knockout mutants were 298 found to exhibit increased sensitivity to 5-FU treatment, in particular in multi-cycle 299 experiments, which was attributed to a higher mutation frequency in the absence of 300 ExoN-driven error correction (28). We employed this same assay to assess the 301 phenotype of the E191D mutant in more detail, by performing plaque assays in HuH7  Previously, the ExoN activity of SARS-CoV nsp14 was found to be dramatically 366 stimulated by the addition of nsp10 as co-factor (26). Consequently, we also expressed 367 and purified MERS-CoV nsp10 and optimized the ExoN assay by testing different 368 molar ratios between nsp14 and nsp10 (Fig. 6A, left-hand side), different nsp14 369 concentrations (Fig. 6B, left-hand side), and by different incubation times (Fig. 7, left-370 hand side). MERS-CoV nsp14 ExoN activity was found to be stimulated by nsp10 in a 371 dose-dependent manner (Fig. 6A), while nsp10 did not exhibit any nuclease activity by 372 itself (Fig. 6B, nsp10 lane). The full-length substrate is more completely degraded 373 when a fourfold (or higher) excess of nsp10 over nsp14 was used compared to the 374 effect of merely increasing the nsp14 concentration in the assay (Fig. 6B). Similar  (Bouvet, Imbert et al. 2012, Ma, Wu et al. 2015.

381
Using a 4:1 ratio of nsp10 versus nsp14, MERS-CoV ExoN activity was analyzed in 382 a time-course experiment, (Fig. 7). Over time, the full-length substrate was 383 progressively converted to a set of degradation products in the size range of 6-18 nt. 384 We anticipated that the structure of the H4 RNA substrate would change from a

408
When the two proteins were combined in the same reaction, a strong increase of ExoN 409 activity was observed for both nsp14-nsp10 pairs, with the SARS-CoV pair appearing 410 to be somewhat more processive than the MERS-CoV pair (Fig. 8, lanes 2 and 9).  (Fig. 2). 434 We also evaluated the impact of the H229C ZF1 mutation, which -despite its 435 conservative nature -yielded a crippled mutant virus (Fig. 2) and of two N7-MTase 436 mutations (discussed below). The N7-MTase mutants displayed wt nsp14-like ExoN 437 activities (Fig. 9), suggesting that -as in SARS-CoV nsp14 -ExoN and N7-MTase 438 activities are functionally separated (36). Analyzing the substrate degradation pattern 439 of the H229C mutant (Fig. 7), the enzyme seems to be somewhat crippled when 440 compared to wt nsp14. This suggests that this mutation alters ExoN activity in vitro, 441 potentially by affecting the structure of the ExoN domain, as ZF1 is in close proximity of the nsp10 interaction surface (20). However, a similar reduction of ExoN activity was 443 observed for the E191D mutant (Fig. 7), which was much more viable than the H229C  Recombinant MERS-CoV nsp14 was found to methylate GpppA, but not m7GpppA 464 (Fig. 10A), which yielded a signal that was similar to the background signal in assays 465 lacking nsp14 or substrate (data not shown). Methylation increased with time until 466 reaching a plateau after 120 min (Fig. 10B). The N7-MTase activity of the various 467 nsp14 mutants was compared with that of wt nsp14 after reaction times of 30 and 120 468 min (Fig. 10C). While the R310A and D331A control mutations fully inactivated the N7- ExoN catalytic residues abolished all detectable viral RNA synthesis (Fig. 3) and the 484 release of viral progeny (Fig. 2). The only exception was the conservative E191D 485 mutant, which was found to exhibit near-wild-type levels of ExoN activity ( Fig. 7 and   486 9). Based on nsp14 conservation (Fig. 1) and the viable phenotype of SARS-CoV and 487 MHV ExoN-knockout mutants, MERS-CoV was expected to tolerate ExoN inactivation, 488 in particular since the enzyme was proposed to improve the fidelity of CoV replication 489 without being essential for RNA synthesis per se (10, 14, 21-23, 26, 29). This notion is

544
In the only viable MERS-CoV ExoN active site mutant obtained, E191D, the catalytic 545 motif was changed from DEEDh to the DEDDh that is characteristic of all members of 546 the exonuclease family that ExoN belongs to (14,15,75). The phenotype of the E191D 547 virus mutant was comparable to that of wt virus (Fig. 4). Biochemical assays revealed 548 that E191D-ExoN enzyme is able to hydrolyze a dsRNA substrate with an activity level 549 approaching that of the wt protein (Fig. 9). Additionally, the E191D mutant behaved  In this study, we developed an in vitro assay to evaluate MERS-CoV ExoN activity 554 using a largely double-stranded RNA substrate ( Fig. 6 and 7). As previously observed is interchangeable between CoV subgenera in its role as co-factor for the nsp16 2′-O-565 methyltransferase, which was attributed to the high level of conservation of the nsp10-566 nsp16 interaction surface (77). As nsp14 and nsp16 share a common interaction 567 surface on nsp10 (21, 26, 66), we explored whether a similar co-factor exchange was 568 possible in the context of nsp14's ExoN activity, which was indeed found to be the case 569 (Fig. 8). Structurally, nsp14 interacts with nsp10 figuratively similar to a "hand (nsp14) 570 over fist (nsp10)" conformation (21). In the formation of this complex, nsp10 induces 571 conformational changes in the N-terminal region of ExoN that adjusts the distance 572 between the catalytic residues in the back of the nsp14 palm and, consequently, impact 573 ExoN activity (21). The exchange of the nsp10 co-factor between the two beta-CoVs 9), indicating that each of these residues is important for catalysis.

588
Our study suggests that, in addition to the active site residues, also other motifs in 589 MERS-CoV ExoN are important for virus viability, specifically the two ZF motifs that 590 were probed using two point mutations each (Fig. 2A). In previous ZF1 studies, a 591 mutation equivalent to H229A created solubility issues during expression of recombinant SARS-CoV nsp14 (20) and resulted in a partially active ExoN in the case 593 of white bream virus, a torovirus that also belongs to the nidovirus order (79). It was 594 suggested that ZF1 contributes to the structural stability of ExoN, as it is close to the 595 surface that interacts with nsp10 (20). Here, we demonstrate that the more 596 conservative H229C replacement, which converts ZF1 from a non-classical CCCH 597 type ZF motif into a classical CCCC type (63), was tolerated during recombinant protein 598 expression and yielded an ExoN that is quite active in vitro (Fig. 7). This likely 599 contributed to the fact that the H229C virus mutant retained a low level of viability ( Fig.   600 2), although its overall crippled phenotype and the non-viable phenotype of mutant 601 C201H clearly highlight the general importance of ZF1 for virus replication. In contrast, 602 the corresponding TGEV mutant (ZF-C) was not strongly affected and could be stably