Nsp14 of SARS-CoV-2 inhibits mRNA processing and nuclear export by targeting the nuclear cap-binding complex

Abstract To facilitate selfish replication, viruses halt host gene expression in various ways. The nuclear export of mRNA is one such process targeted by many viruses. SARS-CoV-2, the etiological agent of severe acute respiratory syndrome, also prevents mRNA nuclear export. In this study, Nsp14, a bifunctional viral replicase subunit, was identified as a novel inhibitor of mRNA nuclear export. Nsp14 induces poly(A)+ RNA nuclear accumulation and the dissolution/coalescence of nuclear speckles. Genome-wide gene expression analysis revealed the global dysregulation of splicing and 3′-end processing defects of replication-dependent histone mRNAs by Nsp14. These abnormalities were also observed in SARS-CoV-2-infected cells. A mutation introduced at the guanine-N7-methyltransferase active site of Nsp14 diminished these inhibitory activities. Targeted capillary electrophoresis-mass spectrometry analysis (CE-MS) unveiled the production of N7-methyl-GTP in Nsp14-expressing cells. Association of the nuclear cap-binding complex (NCBC) with the mRNA cap and subsequent recruitment of U1 snRNP and the stem-loop binding protein (SLBP) were impaired by Nsp14. These data suggest that the defects in mRNA processing and export arise from the compromise of NCBC function by N7-methyl-GTP, thus exemplifying a novel viral strategy to block host gene expression.


INTRODUCTION
Since eukaryotic cells are compartmentalized to the nucleus and the cytoplasm, newly synthesized mRNAs must be transported to the cytoplasm for decoding genetic information by the translation machinery ( 1 , 2 ). Packaging of a mature mRNA into a messenger ribonucleoprotein (mRNP) particle, which is a pr er equisite for transport through the nuclear pore complex (NPC), occurs cotranscriptionally in the nucleus (3)(4)(5). The carboxy-terminal domain (CTD) of the largest subunit of RN A pol ymerase II (RN APII) initiall y plays essential roles in this process. The CTD code, which consists of various combinations of posttranslational modifications of the heptapeptide repeats (YSPTSPS), such as the phosphorylation of specific serine residues, leads to the pertinent recruitment of differ ent pr e-mRNA processing / ma tura tion factors to acti v e genes (6)(7)(8). During transcription initiation phase, S 5 of the CTD is phosphorylated and attracts the capping enzyme (mRN A guanyl yltr ansfer ase; RNGTT), which adds an inverted guanosine moiety to the 5 -end of a nascent transcript by the triphosphatase and guanylyltr ansfer ase activities ( 9 , 10 ). The cap methyltr ansfer ase (RNA guanine N-7 methyltr ansfer ase; RNMT), in turn, is then recruited to the transcription initiation site and adds a methyl-group to the N7-position of the cap guanosine base, although in this process the interaction with the phosphorylated CTD is likely to be indirect ( 11 ).
Capped nascent pre-mRNA is then bound by the nuclear cap binding complex (NCBC) for transcriptional elongation and further processing. NCBC is a multifunctional heterodimer consisting of CBP20 and CBP80, of which the former directly binds to the N7-methylated cap structure (12)(13)(14). NCBC, along with the CTD, influences subsequent steps in mRNA metabolism through interactions with numerous factors ( 15 ). The downstream processes in which NCBC and specific interaction partners have been implicated are highly di v ergent and include the following: promoter clearance and producti v e transcription elongation with positi v e transcription elongation factor b (P-TEFb) ( 16 ), splicing with U1, U2 and U4 / U6 ·U5 snRNPs ( 17 , 18 ), cleavage and polyadenylation with the 3 -end processing factors ( 19 ), stem −loop binding protein (SLBP)-mediated replication-dependent (RD) histone mRNA ma tura tion with nega ti v e elongation factor (NELF) ( 20 ), nuclear export with the TRanscription-EXport complex (TREX) ( 21 ), and translation with eIF-4G ( 22 ) and CBC-dependent translation initiation factor (CTIF) ( 23 ). In addition to these broad functions on mRNA metabolism, NCBC determines the fates of the other RNAPII-transcribed RNAs: the intracellular transport of U snRNAs as well as a small subset of capped snoRNAs is regulated by NCBC-PHAX ( 24 , 25 ), whereas the degradation of short-li v ed RNAs such as promotor upstream transcripts (PROMPTs) is accelerated by NCBC-ZC3H18 ( 26 ).
Although detailed molecular mechanisms remain under acti v e inv estigation ( 15 ), recent studies hav e suggested that NCBP3, w hich was originall y identified as an alternati v e for CBP20 ( 27 ), participates in the nascent mRNP through interaction with NCBC and recruits TREX, thus playing an important role in licensing mRNPs for nuclear export ( 28 , 29 ). Ev entually, the conserv ed mRNA nuclear export receptor Tap-p15 (also called NXF1-NXT1) recognizes the fully ma tura ted mRNP as a tr ansport cargo via inter action with ALYREF, an mRNA binding adaptor component of TREX, allowing translocation of the mRNP through the NPC ( 5 , 30 ).
Viruses often exploit and / or inhibit the host nuclear RNA export processes to evade host defense systems and to promote their own gene expression for selfish replication. Pathogenic lentiviruses, such as human immunodeficiency virus (HIV), use the Rev protein to seize the CRM1mediated U snRNA export pathwa y f or the expression of their unspliced mRNAs ( 31 , 32 ). Simple retroviruses, such as simian retrovirus type 1, hav e e volv ed the constituti v e transport element (CTE) in their RNA genomes to hijack the Ta p-p15-mediated mRN A export pathway (33)(34)(35). Herpes simplex virus ICP27 facilitates the export of viral intronless mRNAs by using the ALYREF / Tap-p15-mediated mRNA export pathway ( 36 ). The vesicular stomatitis virus (VSV) matrix (M) protein shuts down host mRNA export by inhibiting the nucleoporins Nup98-Rae1 at the NPC ( 37 , 38 ). ORF10 of Kaposi's sarcoma-associated herpesvirus also targets Nup98-Rae1 and inhibits the nuclear export of select cellular mRNAs ( 39 ). Poliovirus 2A protease blocks the nuclear export of mRN A, rRN A and U snRN As by cleaving Nup98 ( 40 ).
Inhibition of cellular mRNA export by SARS-CoV-2, the causati v e agent of the present pandemic of se v ere acute respiratory syndrome, has also been noted as a mechanism to contend for the host defense system ( 41 ) (see Supplementary Figure S1). Recent studies have indica ted tha t both the ORF6 and NSP1 proteins of SARS-CoV-2 target the mRNA export process by inhibiting Nup98-Rae1 and Tap-p15, respecti v ely (42)(43)(44), although the latter mechanism could also be explained differently ( 41 , 45 , 46 ). In this study, we assessed 24 different SARS-CoV-2 proteins for their mRNA export inhibitory activities and added Nsp14 to the short list of viral mRNA export inhibitors. The data presented in this report indicate that increased 7-methyl-GTP (m7GTP), which is produced by Nsp14 and can act as a cap mimic, perturbs the functions of NCBC in the processing and nuclear export of mRNAs, thus demonstrating a novel vir al str ategy to b lock host cell gene e xpression.

Antibodies
Antibodies against Tap and the human THO / TREX components have been previously described ( 47 , 48 ). Rabbit polyclonal antibodies against GFP, mouse IgG, CBP20, SLBP, and SARS-CoV-2 spike (S) protein, mouse monoclonal antibodies against ALYREF, SC35, ␤-actin, FLAGpeptide ta g, Rho-1D4 ta g, Strep-ta g II and a rat monoclonal antibody against U1C wer e commer cially acquir ed. Antibodies against SARS-CoV-2 Nsp14 and puromycin were obtained from MRC PPU Reagents and Services, Uni v ersity of Dundee and DSHB, Uni v ersity of Iowa, respecti v ely. Horseradish peroxidase (HRP)-conjugated secondary antibodies were purchased from Bio-Rad, whereas Alexa-and DyLight-conjugated secondary antibodies were purchased from Thermo Fisher Scientific. HRP-conjugated anti-rat IgG was purchased from Abcam. The monoclonal antibody 38A1 against CBP80 ( 49 ) was a gift from Drs Hito Ohno and Ichiro Taniguchi of Kyoto Uni v ersity. Details of the primary antibodies used in this study are listed in Supplementary Table S1.

Plasmid construction
Mammalian e xpression v ectors of the SARS-CoV-2 proteins ( 50 ), a DOX-inducible mammalian expression vector pInducer20 ( 51 ), and a human ACE2 expression vector pLENTI hACE2 HygR ( 52 ) were obtained from Addgene (listed in Supplementary Table S2). The mammalian e xpression v ector pCMV-FLAG3 was constructed by replacing the GFP ORF of the pEGFP-C1 vector (Clontech) with synthetic double-stranded oligo DNA encoding a 1xFLAG peptide tag (MDYKDDDDK). Expression vectors for GFP-Nsp14 fusion proteins were constructed as follows. Complementary DNAs (cDNAs) encoding GFP and Nsp14 together with a bovine growth hormone polyadenylation signal (BGH pA) cassette from pcDNA3.1 (Invitrogen) were amplified by polymerase chain reaction (PCR) and inserted into the pENTR4 entry vector (Invitrogen) by using the In-Fusion HD cloning kit (Clontech). The resulting GFP-Nsp14-BGH pA cassette was transferred to the pInducer20 vector by the Gateway system (Invitrogen). The D 90 VE → AVA and D 331 → A mutations were introduced by the Quick Change kit (Agilent). For transient expression experiments, cDNAs encoding wild-type and mutant Nsp14 proteins were inserted into the BspEI-SalI site of the pEGFP-C1 vector. To remove the GFP-tag, the pEGFP -Nsp14 vector was doubly digested with NheI-BspEI, and a synthetic double-stranded oligonucleotide encoding the initiation methionine was inserted. The protein expressed from the resulting vector is Nsp14 appended with only four amino acids (MSGA) at the amino-terminus. For the construction of DcpS expression vectors, a cDNA fragment encoding mouse DcpS (GenBank accession NM 027030) was amplified by PCR from a mouse 7-da y embry o cDNA library (Clontech) and inserted into the pCMV-FLAG3 and pENTR4-GFP-BGHpA vectors by the In-Fusion HD cloning kit. Transfer of the GFP-DcpS-BGH pA cassette was carried out as described above, and the resulting plasmid was named pInducer20-GFP-DcpS. All the sequences were verified by Sanger sequencing. Detailed plasmid maps are available upon request.

Cell cultur e, tr ansfection, and establishment of a stable cell line
Human 293F (Invitrogen) and HeLa (ATCC) cells were cultured in Dulbecco's modified Eagle's medium supplemented with 10% fetal calf serum (DMEM-10% FCS) under a 5% CO 2 atmosphere. Vero E6 cells were cultured in minimal essential medium supplemented with 10% fetal calf serum (MEM-10% FCS). Transfection of plasmids was carried out by using Effectene transfection reagent (Qiagen) according to the manufacturer's protocol. To establish DOXinducible cell lines, pInducer20 vectors harboring GFP fusions of wild-type and D 331 → A Nsp14 were linearized by SfiI and transfected into 293F cells. At 48 h after transfection, the cells were reseeded into 10 cm dishes containing DMEM-10% FCS supplemented with 1.2 mg / ml G418 (Nacalai Tesque). After 10 days, cell colonies were manually picked, and the individual clones wer e scr eened for DOXinducib le e xpr ession of the fusion proteins under a fluor escence microscope. Clones named 293F Nsp14 wt2 (wildtype) and 293F Nsp14 AIG44 (D 331 → A) were used for fur-ther analysis. A 293F cell line stab ly e xpressing hACE2 was established essentially as described abov e, e xcept that 400 g / ml hygromycin was used instead of G418 for selection. The clones were screened for their hACE2 expression le v els by Western b lotting and IFA using an anti-Rho-1D4 tag and SARS-CoV-2 susceptibility (Supplementary Figure S9). A clone named 293F hACE2 21 was used for further analysis. To establish GFP-DcpS-inducible cell lines, SfiI-digested pInducer20-GFP-DcpS was transfected into 293F hACE2 21 cells. Cells harboring the hACE2 and GFP-DcpS e xpression v ectors were selected by 1.2 mg / ml G418 and 400 g / ml hygromycin and a cell line designated 293F hACE2 DcpS 29 was used for further analysis. DOX-inducib le e xpression of the fusion protein in each cell clone was confirmed as described above (Supplementary Figure S9).

Viral infection
The SARS-CoV-2 strain JPN / TY / WK / 521 was provided by the National Institute of Infectious Diseases, Tok yo , Japan. Viruses were propagated in a monolayer of Vero E6 / TMPRSS2 cells cultured in DMEM-2% FCS. Monolayers of Vero E6, 293F hACE2 21 or 293F hACE2 DcpS 29 cells in DMEM-2% FCS were infected with viruses at the indicated multiplicity of infection. After 24 hr of incubation, the cells were processed for further analysis.

Oligo-dT in situ hybridization
Fluorescent in situ hybridization (FISH) using a Cy3labeled oligo-dT 50 probe was carried out as previously described ( 53 ). For double staining with antibodies, fixed cells were first subjected to immunofluorescence analysis using the indicated antibodies, postfixed with 4% paraformaldehyde, and then probed with a Cy3-labeled oligo-dT 50 probe.

RNA-seq and data analysis
Total RNA was isolated from uninduced or induced 293F Nsp14 wt2 cells by TRIzol reagent (Ambion). Library construction using the TruSeq stranded mRNA kit and sequencing by NovaSeq 6000 (Illumina) were carried out by Macrogen J apan Corp. P oly(A) + RNA-seq data SRR13952621 (SARS-CoV-2 infected) and SRR13952627 (uninfected) ( 54 ), which were downloaded by fasterq-dump from the National Center for Biotechnology Information sequence read archi v e (NCBI SRA), were used for gene expression analysis of SARS-CoV-2-infected A549 cells. For the analysis of inhibition of telescripting, SRR9864939 (control morpholino-treated cells) and SRR9864940 (U1antisense morpholino-treated cells) ( 55 ) were used. Adapter trimming and quality filtering of the 101 base paired-end r eads wer e carried out by fastp ( 56 ). The passed r eads wer e mapped to the human genome 19 (hg19) assembly by using HISAT2 ( 57 ) or STAR ( 58 ). Differ entially expr essed genes were mined by featureCounts ( 59 ). Mapping to different gene features, i.e. coding sequences (CDSs), introns, intergenic regions, and untranslated regions (UTRs), were carried out by gatk4 collectRnaSeqMetrics ( 60 ) using an hg19 refFlat file downloaded from the Uni v ersity of California Santa Cruz (UCSC) site. Intron retention ratios (IR ratio) at different introns were calculated by IRFinder ( 61 ) using the bam files generated by STAR aligner. Of 249325 introns anal yzed automaticall y by the software, those with the 'Low-Cover' warning under both uninduced and induced conditions were removed. Of the remaining 79724 introns, the IR ratios of 93% (74262 introns) were in the range of 0 to 0.3. Thus, only these introns were included in the heatmap and the box plot to avoid low confidence data ( 61 ). IR ratio analysis of SARS-CoV-2-infected cells was carried out in the same manner. Box plots were made by BoxPlotR ( 62 ). Conversion of the bam files to CPM-normalized bigWig files was performed using Deeptools bamCoverage ( 63 ). Metagene analysis of 69 human replication-dependent histone genes was carried out by Deeptools computeMatrix r efer ence-point using a custom GTF file generated from data downloaded from the UCSC table browser. Heatmaps were drawn by Deeptools plotHeatmap.

Quantitati ve PCR (qPCR) and quantitati ve RT −PCR (qRT −PCR)
qPCR analysis was carried out by using Luna Uni v ersal qPCR Master Mix. Reaction conditions were as per the manufacturer's protocol.
For qRT −PCR anal ysis, RN A isolation was carried out as described above. The total RNA samples were pretreated with a Turbo DNA-free kit (Thermo Fischer Scientific) to remove contaminated genomic DNA. Re v erse transcription r eactions wer e carried out using LunaScript RT SuperMix with oligo-dT and random hexamers. qPCR was carried out as described abov e. Relati v e amounts of the targets were determined by the standar d curv e method. The nucleotide sequences of the qPCR primers used are listed in Supplementary Table S3. Data are presented as averages of three biological replicates with standar d de viation. Statistical significance was verified by two-tailed Welch's t test.

RNA immunoprecipitation assay (RIP)
Cells with or without DOX induction (1 × 6 cm dish / IP) were harvested and resuspended in 600 l of RIPA buffer (10 mM Tris-HCl (pH 8.0) / 100 mM NaCl / 1 mM EDT A / 0.5 mM EGT A / 1% Triton X-100 / 0.1% deoxycholate / 0.05% SDS) and disrupted by brief sonica tion. The insoluble ma terials wer e r emoved by centrifugation. An aliquot (100 l) of the supernatant was removed, and total RNA was isolated by TRIzol reagent (input fraction). The remaining portions were mixed with Dynabeads ProteinG that had been preincubated with each antibody. After 7 h of incubation at 4 • C, the beads were e xtensi v ely washed with RIPA buffer, and bound materials were extracted by TRIzol reagent (IP fraction). Contaminated genomic DNA was removed by a Turbo DN A-Free Kit. qRT −PCR anal ysis was carried out as described above. RIP efficiency was calculated by dividing the amount of each RNA in the IP fraction by that in the corresponding input fraction. Data are presented as averages of at least three technical replicates with standard deviation. The nucleotide sequences of the qPCR primers used are listed in Supplementary Table S3.

Capillary electrophoresis coupled with mass spectrometry (CE-MS)
CE-MS and m ultitarget anal ysis were carried out as previously described ( 64 ). The m7GTP standard was purchased from Sigma.
By comparing the area of the CE peak of the known amount of the standard with those of the test samples, m7GTP produced in 4.3 × 10 6 Nsp14-expressing cells was estimated to be 954 ± 117 pmol. Assuming that a 293F cell is a sphere with a diameter of ∼15 m ( 65 ), the cellular concentration of m7GTP was calculated to be ∼119 M.

UV cr osslinking and cr osslinking-immunoprecipitation (CLIP)
UV crosslinking was performed essentially as described previously ( 47 ). 293F Nsp14 cells (3 × 10 cm dishes / condition) with or without DOX induction were used. The UV dose was 300 mJ / cm 2 . The cells were lysed in RIPA buffer, and the soluble fractions were adjusted to final concentrations of 1% SDS and 0.5 M LiCl, then hea t dena tured and subjected to oligo-dT cellulose (Wako pure chemicals) chromato gra phy. After e xtensi v e washing with RIPA containing 1% SDS and 0.5 M LiCl, bound materials were eluted with TE buffer (10 mM Tris-HCl (pH 8.0) / 1 mM EDTA). The samples were concentrated by ethanol pr ecipitation, tr eated with 100 g / ml RNase A, and analyzed by SDS −PAGE followed by Western blotting using the antibodies specified in the figure.
CLIP was carried out as follows. UV-irradiated cells were harvested and resuspended in RIPA buffer. The cells were disrupted by brief sonication, and insoluble materials were removed by centrifugation. An aliquot (100 l) of the supernatant was removed and treated with proteinase K (250 g / ml), and total RNA was isolated by TRIzol reagent (input fraction). The remaining proteins were mixed with Dyna beads ProteinG , which had been pr etr eated with each antibody. After 8 h of incubation at 4 • C, the beads were extensi v ely washed with RIPA buffer. The bound materials were eluted with 50 mM Tris-HCl (pH 8.0) / 5 mM EDTA / 1% SDS at 65 • C and diluted 2 times with H 2 O. After proteinase K treatment, RNA was extracted by TRIzol reagent (IP fraction). Contaminated genomic DNA was removed by a Turbo DNA-Free Kit. qRT −PCR analysis was carried out as described above. CLIP efficiency was calculated by dividing the amount of each RNA in the IP fraction by that in the corresponding input fraction. Data are presented as averages of at least three technical replicates with standard deviation. Reproducibility was confirmed by two to three independent experiments. The nucleotide sequences of the qPCR primers used are listed in Supplementary Table S3.
Global protein synthesis was monitored by analyzing incorporation of puromycin into the nascent peptide chains according to a method described in ( 68 ). 293F Nsp14 wt2 cells without or with DOX induction (48 hr) were pulse labeled with puromycin at the concentrations specified in the figure for 10 min and chased for 60 min. Whole cell extracts were subjected to SDS-PAGE followed by Coomassie brilliant blue staining or Western blotting. The monoclonal anti-puromycin antibody was used at the concentration of 0.5 g / ml.

Identification of Nsp14 as a novel viral mRN A e xport inhibitor
Nuclear export of host cell mRNA was blocked upon SARS-CoV-2 infection (Supplementary Figure S1: SARS-CoV-2 infected Vero cells). To identify which viral proteins are responsible for export inhibition, we screened individual SARS-CoV-2 proteins for their nuclear export inhibitory activities by oligo-dT FISH analysis. Expression of the epitope-tagged viral proteins was readily detectable by imm unofluorescence anal ysis (IFA) using an anti-tag antibody ( Figure 1 and Supplementary Figure S2). Among the 24 different viral proteins tested, only three showed apparent export inhibitory activities ( Figure 1 , also see Supplementary Figure S2 for factors not affecting export). As previousl y reported, nuclear accum ulation of pol y(A) + RN A was observed in cells expressing either ORF6 or Nsp1 proteins ( 42 , 43 ). In some instances, the FISH signals accumula ted a t the nuclear rim in ORF6-expr essing cells (Figur e 1 , inset), which might reflect the disturbance of NPC function by the viral protein ( 42 , 50 ). In addition to these known factors, we found that Nsp14 strongly inhibits the nuclear export of bulk pol y(A) + RN As (Figure 1 ).
Nsp14 is a bifunctional viral replicase subunit consisting of the N-terminal exoribonuclease (ExoN) and the Cterminal guanine N7-methyltr ansfer ase (N7-MTase) domains (Figure 2 A). Since the Nsp14 pr oteins fr om SARS-CoV and SARS-CoV-2 are almost identical (overall sequence identity is 95%), the previous data of SARS-CoV Nsp14 ( 69 , 70 ) could be extrapolated to SARS-CoV-2 Nsp14. Indeed, recent biochemical and structural studies re v ealed tha t the residues critical for the enzyma tic activities are well conserved (71)(72)(73). Ther efor e, we selected two well-characterized mutants D 90 VE → AVA (ExoN mutant) and D 331 → A (N7-MTase mutant) to examine which enzymatic activity is r equir ed for pol y(A) + RN A export inhibition. The wild-type SARS-CoV-2 Nsp14 as well as the mutants were expressed as GFP fusion proteins. As shown by Western blotting, these mutant proteins were expressed as efficiently as the wild-type protein (Figure 2 B). In transfected cells, they were localized in both the nucleus and the cytoplasm, which was indistinguishable from the localization of the wild-type protein. FISH analysis re v ealed that the D 90 VE → AVA mutant blocked poly(A) + RNA export as efficiently as the wild-type protein, whereas the D 331 → A mutant showed almost complete loss of inhibitory activity (Figure 2 C). These data indicate that the N7-MTase activity of Nsp14 is indispensable for the inhibition of the nuclear export of mRNAs. In Nsp14-expressing cells, pol y(A) + RN As accum ulated in nuclear speckle domains, and dissolution and / or coalescence of the speckles were observed (Figure 2 D). Removal of most part of the GFP tag sequence did not affect the pol y(A) + RN A export inhibitory activity (Supplementary Figure S3), excluding a possibility that the addition of the tag at the N-terminus grossly altered the function of Nsp14.

Nsp14 dysr egulates pr e-mRNA splicing and the 3 -end formation of replication dependent histone mRNAs
To analyze the genome-wide gene expression changes induced by Nsp14, doxy cy cline (DOX)-inducib le cell lines were established (Figure 3 A). As expected, the expression of the wild-type and D 331 → A mutant Nsp14 proteins was tightly regulated, and the GFP-fusion proteins were detectable only after DOX induction (Figure 3  Total RNA pr epar ed befor e (uninduced) and 48 hr after the induction (induced) of wild-type Nsp14 expression was subjected to pol y(A) + RN A-seq anal ysis. Of a pproximatel y 50 million paired reads each prepared from the samples, most (92.09% of uninduced and 93.24% of induced samples, Supplementary Figure S4) were uniquely mapped to the human r efer ence genome hg19. The mapping rate of the reads to different gene features was further analyzed, and the intron reads were considerably increased (8.6 to 16.9%) upon Nsp14 expression at the expense of CDS reads (48.3 to 40.7%) (Figure 4 A). To examine whether splicing defects actually occur, the intron-retention (IR) ratio was calculated ( 75 ). As shown in Figure 4 B and C, the IR ratio was significantly increased by Nsp14 expression. Closer inspection of individual genes revealed that intron coverage was actually increased a t dif ferent genes ( Figure 4 D and E). The increase in intron coverage at the ACTB and GAPDH loci was confirmed by qRT-PCR analysis, and wild-type Nsp14, but not the D 331 → A mutant, caused these changes ( Figure 4 F and  G). Interestingly, in the IR analysis, nearly 25% of the retained introns were appended with the 'nonUniformIntron-Cover' warning. In fact, at long first introns of a subset of genes, pr ematur e polyadenylation was detected, which was similar to that observed when U1 snRNA was inhibited by antisense morpholino (Supplementary Figure S5A, C, D, E). The data indicate that a process called 'telescripting' ( 55 ) was affected. Since the relati v e amount of U1 snRNA was incr eased (Figur e 4 H) and uncleaved pre-U1 snRNA was virtually undetectable under Nsp14-expressing conditions, the downregulation and / or malfunction of U1 snRNP per se was not the direct cause of this defect.
Of a pproximatel y 50,000 significantl y expressed transcripts, most wer e downr egulated, and only 25% were upregulated upon Nsp14 expression. Of note, the pol y(A) + RN A-seq read coverage on a subset of the  Table S4). Normally, transcription of the RD histone genes is terminated just downstream of the stem −loop (SL) sequence by SLBP-dependent cleavage, and ther efor e, the RD histone mRNAs ar e nonadenylated ( 76 , 77 ). Nsp14 inhibited SLBP-dependent RD histone termination, and the cleavage sites shifted toward the downstream canonical polyadenylation sites (Figure 4 I and J). As the result, the amount of polyadenylated RD histone mRNAs was increased significantly. An increase in the RNA le v el downstream of the SL sequence of the H4C5 gene was confirmed by qRT −PCR analysis (Figure 4 K). Termination by the canonical cleavage / polyadenylation factors at other genes with alternati v e polyadenylation sites, such as the TIMP2 and RPL22 genes, was not grossly altered (see Supplementary Figure S6), indicating that the downstream shift of the polyadenylation site is specific for the RD histone genes.

Nsp14 increases the cellular concentration of m7GTP
SARS-CoV Nsp14 reportedly methylates not only the cap guanosine, but also the guanosine base of 'free' mononucleotide GTP in in vitro methylation assays ( 69 ). Thus, we addressed whether the closely related SARS-CoV-2 Nsp14 actually increases the cellular m7GTP le v el. Cell e xtracts pr epar ed befor e and after 48 h of DOX induction were subjected to capillary electrophoresis coupled with mass spectrometry (CE-MS). Strikingly, CE peaks with migration times of a pproximatel y 15 min were detected only in the extracts pr epar ed from wt Nsp14 expr essing cells (Figure 5 A). When the extracts pr epar ed either from uninduced cells or from cells expressing the D331 → A mutant were analyzed, nothing was detected in the same migration time range (Figure 5 A). The average monoisotopic m / z of the CE peaks was 267.4987, which was almost identical to the 7608 Nucleic Acids Research, 2023, Vol. 51, No. 14  theoretical m / z of m7GTP bivalent cation (267.4959), indica ting tha t Nsp14 actually methyla tes GTP. Moreover, the results indicate that the D331 → A mutant is enzymatically inacti v e. The slight difference in the migration times between the samples and synthetic m7GTP was most likely due to the presence of residual ingredients originating from the culture medium or cell extracts, because the peak was indistinguishable from the spike-in m7GTP standard (Supplementary Figure S7).
The scavenger mRN A deca pping enzyme DcpS ( 78 ) eliminates cellular m7GTP by converting it to N7-methylated GMP ( 79 ). When DcpS was ov ere xpressed in Nsp14expressing cells, the pol y(A) + RN A export block was considerably mitigated and the size and the shape of the nuclear speckles returned to normal (Figure 5 B). In addi-tion, the incr eased r eadthrough H4C5 RNA le v el by Nsp14 was significantly reduced by the DOX-induced expression of DcpS, although DcpS expression alone did not significantly alter it (Figure 5 C and D, see also Supplementary Figure S9D for establishment of the cell line). From these da ta, we concluded tha t SARS-CoV-2 Nsp14 produces m7GTP in vivo and that the accumulated m7GTP inhibited mRNA nuclear export and RD histone mRNA 3 -end formation.

Nsp14 inhibits the NCBC-cap interaction
The data thus far indicate that m7GTP produced by Nsp14 caused the observed defects in gene expr ession; i.e. pr e-mRN A splicing, histone mRN A 3 -end formation, and (right panels) cells were cultured for 48 h in the presence of 2.5 g / ml DOX. The cells were fixed and subjected to IFA using the indicated antibodies. The cell nuclei were stained with Hoechst 33342. The cells were observed by confocal microscopy. In the merged pictures, the fluorescent signals of THOC2 and SC35 were pseudocolored red and green, respecti v ely. mRNA nuclear export. We conceived that these seemingly di v ergent outcomes could converge to functional defects of NCBC. As the mRN A ca p and NCBC interaction can be competiti v ely inhibited by m7GTP ( 13 ), we first e xamined whether the interaction between the capped RNAs and the components of NCBC is affected by Nsp14. To this end, RN A imm unoprecipitation (RIP) using anti-CBP80 and -CBP20 antibodies was carried out. As determined by Western blotting, the expression of CBP80 and CBP20 was not changed significantly upon Nsp14 expression (Figure 6 A, also see Supplementary Figure S8 for the expression and localization of CBP20 in individual cells). The amounts of capped RNAs coprecipitated with either CBP80 or CBP20 were se v erely decreased after Nsp14 induction (Figure 6 C, see Figure 6 B for primer positions). In contrast, the background precipitation of uncapped 5S rRNA with the NCBC components was unaffected or slightly increased upon expression of Nsp14 (Figure 6 C). These data indica te tha t ca p reco gnition b y NCBC is abated b y Nsp14 expression.

Nsp14 impairs both U1 snRNP and SLBP recruitment to pre-mRNAs
Gi v en the absence of functional NCBC on nascent pre-mRNAs, we hypothesized that the subsequent recruitment of pre-mRNA processing factors might also be affected by Nsp14. To examine whether Nsp14 alters the recruitment of U1 snRNP and SLBP to pre-mRNAs, CLIP followed by qRT −PCR was carried out by using antibodies specific for U1C and SLBP. Western blotting using these antibodies indica ted tha t the expression of U1C was mostly unaffected ( ∼80%) and SLBP was decreased ( ∼40%) at 48 h after Nsp14 induction (Figure 6 A). Coprecipitation of GAPDH pre-mRNA with anti-U1C antibody was substantially decreased upon Nsp14 expression (Figure 6 D, 55F / 55R and F1 / R3). As the anchoring of U1 snRNP to long first introns via NCBC with Ars2 has been implicated in telescripting ( 55 ), coprecipitation of the ACTN4 pre-mRNA with anti-U1C was also se v er ely impair ed by Nsp14 (Supplementary Figure S5B). Coprecipitation of 5S rRNA was almost undetectable (Figure 6 D) The data are presented as the means ± SDs of three biological replicates. * means P value < 0.05. experimental settings. In addition, the interaction of SLBP with RD histone H4C5 pre-mRNA was also decreased to background le v els (Figure 6 E). The e xtent of the reduction in RNA coprecipitation with either U1C or SLBP far exceeded that at the protein le v el. Ther efor e, from these data, we concluded that Nsp14 disturbs the recruitment of U1 snRNP and SLBP to pre-mRNAs.

Nsp14 inhibits the formation of export-competent mRNP
To elucidate the molecular mechanism of the defect in mRNA nuclear export caused by Nsp14, association of transport factors with pol y(A) + RN As was examined by UV-crosslinking. As shown in Figure 6 F, only upon UVirradiation were both Tap and ALYREF r ecover ed in the pol y(A) + RN A-containing fraction (Figure 6 F, lanes 1-4).
The crosslinking of Tap and ALYREF to poly(A) + RNAs was se v er ely impair ed under Nsp14-expr essing conditions (Figure 6 F, lanes 5-8). The association of these mRNA export factors with GAPDH mRNA was further examined by CLIP followed by qRT-PCR. The amount of GAPDH mRNA associated with ALYREF and Ta p was significantl y r educed (Figur e 6 G). These da ta indica te tha t Nsp14 im-pairs host mRNA export by inhibiting the recruitment of export factors.

Gene expression changes in SARS-CoV-2-infected cells
Finally, to assess the gene expression changes in SARS-CoV-2-infected cells, we anal yzed publicl y available pol y(A) + RN Aseq data ( 54 ) (see Supplementary Figure  S4 for mapping details). Intronic reads were increased significantly by SARS-CoV-2 infection, whereas CDS r eads wer e decr eased (Figur e 7 A). As shown in Figure 7 B, the IR ratio was also increased genome-wide in infected cells. In addition, increases in both the readthrough transcript and pol yadenylated mRN A le v els of RD histone genes were observed, indicating that SLBP-dependent 3 -end formation was perturbed (Figure 7 C-F). In this analysis, we noted that intergenic reads were also increased (Figure 7 A), which was not obvious in Nsp14-expressing cells (Figure 4 A). This might be due to the presence of other viral factors. Although less pronounced than that in Nsp14-expressing cells, we confirmed the increase in the readthrough transcript le v el of the H4C5 locus in SARS-CoV-2-infected 293F hACE2 cells (Supplementary RNAs copurified with anti-CBP80 and anti-CBP20 antibodies as well as control antibodies were subjected to qRT −PCR with the indicated PCR primers. RIP efficiency was calculated by dividing the amounts of RNAs immunopurified with each antibody by that of the corresponding input. The data are presented as the means ± SDs of three technical replicates. ** means P value < 0.01. ( D ) Whole-cell extracts pr epar ed from UV-crosslinked 293F Nsp14 wt2 cells induced for the indicated periods were immunoprecipitated with an anti-snRNP U1C antibody. The amounts of GAPDH pre-mRNA and 5S rRNA in the immune-pellets were divided by that in the corresponding input to calculate CLIP efficiency. Shown is a r epr esentati v e of three independent experiments. The data are presented as the means ± SDs of three technical replicates. * and ** indicate P values < 0.05 and < 0.01, respecti v ely. ( E ) Whole-cell extracts pr epar ed fr om UV-cr osslinked 293F Nsp14 wt2 cells induced for the indicated periods wer e immunopr ecipita ted with an anti-SLBP antibod y. The amounts of H4C5 pre-mRNA in the immune pellets were divided by the amounts in the corresponding input to calculate CLIP efficiency. Shown is a r epr esentati v e of three independent e xperiments. The data ar e pr esented as the means ± SDs of thr ee technical r eplicates. ** means P value < 0.01. Right : The amount of SARS-CoV-2 N mRNA normalized to that of GAPDH mRNA was also quantitated. The data are presented as the means ± SDs of three biological replicates. *** means P value < 0.001. ND means none detected. Figure S9) by qRT −PCR analysis (Figure 7 G). Importantly, the defect in RD histone mRNA 3 -end processing by SARS-CoV-2 infection was alleviated by DcpS expression (Figure 7 H, left). DcpS expression itself seemed not to decrease SARS-CoV-2 infectivity, as judged by viral N gene expr ession (Figur e 7 H, right). Thus, these da ta indica te that gene expression changes similar to those observed in Nsp14 expressing cells actually occur in SARS-CoV-2 infected cells.

DISCUSSION
SARS-CoV-2 disrupts the host gene expression system in various means to escape from antiviral defenses. The disturbance of gene expression also serves as a basis for diverse pathologies associated with SARS-CoV-2 infection. Although SARS-CoV-2 completes its life cycle exclusively in the cytoplasm, different steps in the host gene expression process in both the nucleus and the cytoplasm are influenced by different viral factors ( 41 , 80 ). In this study, we have successfully identified Nsp14 as a novel viral mRNA export inhibitor. On analysis, the inhibitory effects were found not to be limited to mRNA export but extended to the nuclear mRNP ma tura tion processes, all of which are accountable for NCBC dysfunction.

m7GTP as a mediator of host gene expression shutoff by Nsp14
Among the different SARS-CoV-2 factors involved in host gene expression shutdown mechanisms, Nsp14 has been linked to the inhibition of transcription, splicing and translation ( 81 , 82 ). Howe v er, elucidation of the detailed molecular basis of the inhibition by this viral factor has been hampered mostly due to the absence of experimental evidence of m7GTP production in cells. Despite the distincti v e structural features ( 83 ), the N7-MTase activity of Nsp14 of SARS-CoV, a closely related species of SARS-CoV-2, is essentially the same as that of the cellular counterpart ( 84 ). Indeed, SARS-CoV Nsp14 is able to complement the growth of the otherwise lethal yeast abd1 mutant ( 85 ). One notable difference between the human and viral N7-MTases is their substrate specificity, and SARS-CoV Nsp14 has been shown to methylate GTP in in vitro assays ( 69 , 86 ). In this study, we extended the previous in vitro data and unequivocally identified m7GTP produced in cells expressing SARS-CoV-2 Nsp14. As has been predicted from circumstantial evidence, our data have established that the inhibition of mRN A ca p function by m7GTP does indeed underlie the virulence of this viral protein. We have also identified additional and previously unappreciated abnormalities brought about by this viral factor: defects in bulk pol y(A) + RN A export, the suppression of prema ture transcription termina tion (telescripting) and the 3end formation of RD histone mRNAs, all of which are also attributable to the inhibition of mRN A ca p function ( 15 ). Intriguingly, most of these changes were also detectable in SARS-CoV-2-infected cells. The inhibition of host gene expression via the production of the small molecule m7GTP described in this report is an unprecedented strategy among other viral factors.

The mechanism of the inhibition
The cap structur e m7GpppN (wher e N is any nucleotide) is a characteristic feature of all eukaryotic mRNAs. The cap is important not only for protecting mRNAs from exonucleolytic degradation but also for providing a binding site to NCBC and eIF4E, which play pivotal roles in nuclear mRNP ma tura tion and cytoplasmic translation initiation, respecti v ely ( 15 , 87 ). Thus, e xcess cap analogs impede eukaryotic gene expression a t dif ferent le v els by limiting the available cap binding proteins from mRNAs. Our data clearly indica te tha t Nsp14, via the production of m7GTP, disturbs cap recognition by NCBC and the subsequent recruitment of mRNA processing and export factors.
Each cap binding protein exhibits dif ferent af finities to different cap analogs. For example, eIF4E prefers capped short ribonucleotides (m7GpppG + N 20 : K D ∼3 nM) to either m7GTP or m7GpppG ( K D ∼260 nM). In contrast, NCBC binds m7GTP and m7GpppN with much higher affinities ( K D ∼10 nM and ∼100 pM, respecti v ely) ( 87 ). The rough estimate of m7GTP concentration in the Nsp14expressing cells deduced from our CE-MS data was > 100 M, which far exceeds the K D values of the cap binding proteins to the methylated mononucleotide. In addition, cells express similar numbers of CBP20 (2.59 × 10 4 / cell) and eIF4E (1.8 × 10 4 / cell) molecules ( 88 ). Hence, it is concei vab le that under such extremely high m7GTP condition, the direct and simultaneous inhibition of both NCBC and eIF4E would be possible. In an in vitro reticulocyte lysate system, 100 M of m7GTP is already high enough to efficiently inhibit translation ( 89 ). We also observed reduced incorporation of puromycin to nascent polypeptides in Nsp14-expressing cells (Supplementary Figure S10), indica ting tha t transla tion is actually inhibited in our experimental condition. Howe v er, in a situation w here onl y a limited amount of m7GTP is produced, NCBP would be preferable and a principal target. Consequently, nuclear phenotypes might be dominant over translational disturbance.
The scavenger decapping enzyme DcpS hydrolyzes both m7GpppN cap and m7GTP, which are intermediates of the 3 → 5 and 5 → 3 mRNA deca y pathwa ys, respecti v ely, into metabolically 'inert' m7GMP ( 79 ). The hydrolyzing activity is ther efor e important for cir cumventing the competition between the residual cap and mRNAs as well as for pre v enting the misincorporation of methylated ribonucleotide into RN As, both of w hich are potentiall y harmful for cells. In fact, the depletion of DcpS by shRNAs resulted in splicing defects ( 90 ), which coincide well with our observations in Nsp14-expressing cells and in SARS-CoV-2-infected cells. We also found that ov ere xpression of DcpS counteracted Nsp14 and alleviated the defect in RD histone mRNA 3end formation as well as the block in bulk pol y(A) + RN A export, the regulation of which was proposed as a possible DcpS function more than a decade ago ( 91 ). Thus, another persuasi v e and non-m utuall y e xclusi v e inhibitory mechanism is that m7GTP indirectly impedes mRNA cap function via suppression of the cap degradation process. Mammalian DcpS hydrolyzes m7GTP at a slower rate than m7GpppG ( 92 ). By slowing DcpS-mediated cap catabolism, Nsp14 could induce toxic accumulation of very short mRNA fragments with the ca p structure, w hich perturbs the cap binding proteins more efficiently.
Surprisingl y, our CE-MS anal ysis also re v ealed the downregulation of cellular ribonucleotides, especially their monophosphate forms, by Nsp14 (Supplementary Table  S5). Interestingly in this context, Nsp14 appeared to upregulate SMN2 gene expression (see Supplementary Table S4), which also occurred when DcpS was inhibited by small-molecule inhibitors, although the molecular mechanisms and biological significance of this phenomenon have yet to be addressed ( 93 ). That RNA catabolism is the major source of cellular CMP and UMP pools ( 94 ) and that ther e ar e functional links between the RNA degradation process and either NCBC or DcpS ( 78 , 95 ) suggest that these might be additional evidence that Nsp14 targets the cap catabolism.

Implications for viral pathogenesis
Although se v eral factors hav e alread y been a ttributed ( 41-44 , 46 , 80 ), the splicing and nuclear export defects caused by Nsp14 could contribute to those observed in SARS-CoV-2 infected cells. As discussed elsewhere ( 41-44 , 46 , 80 ), blocking gene expression at these le v els might disturb invocation of the host defense system.
We also found that the recruitment of SLBP to RD histone mRNA via NCBC was se v er ely impair ed by Nsp14, resulting in the increased expression of polyadenylated RD histone mRNAs. SLBP is r equir ed not only for 3 -end formation but also for efficient nuclear export, cytoplasmic transla tion and stabiliza tion of RD histone mRNAs ( 96 ). As a consequence, it is anticipated that the expression of the RD histone genes is distorted by Nsp14. It has also been reported that SLBP depletion causes S-phase arrest ( 97 , 98 ). During the course of this study, we noted that the expression of Nsp14 considerably delayed cell growth. We found that SLBP, which is synthesized immediately prior to entry into S-phase and degr aded r a pidl y as cells exit from S-phase ( 77 ), was downregulated upon the expression of Nsp14. Moreov er, the e xpression of RD histone and a subset of cell cycle regulators such as cyclin A, geminin, and Emi1, all of which increase as S-phase proceeds ( 99 , 100 ), was found to be reduced in Nsp14-expressing cells (Supplementary Figure S11A and B), suggesting that Nsp14 impairs cell cycle progression during S-phase. Thus, one possibility is that inefficient supply of RD histones by Nsp14 bolsters the recently reported cell cycle arrest during SARS-CoV-2 infection ( 101 , 102 ).
Neuronal impairment by direct SARS-CoV-2 infection of brain tissues is still under debate, and direct brain infection appears to be regarded as a rather uncommon mechanism ( 103 ). Howe v er, r ecent r eports hav e estab lished that SARS-CoV-2 is able to infect both astrocytes and neurons (104)(105)(106). NCBC has been linked to mRNP localization and local translation at neuronal processes, which are important for neuronal de v elopment and function ( 49 ). Intriguingl y, loss-of-function m utations of DCPS cause inherited neuronal diseases, implicating DcpS, or more generally cap catabolism, in neuronal de v elopment and functions (107)(108)(109). Thus, another plausible scenario is that Nsp14, via its inhibition of NCBC functions, contributes to viral patho-genesis as a possible underpinning of the persistent cogniti v e symptoms associated with SARS-CoV-2 infection.
In summary, our present study uncovered the previously unr ecognized r elevance of m7GTP produced by SARS-CoV-2 Nsp14 to host gene expression shutdown mechanisms and di v erse viral pathologies. Nsp14 would undoubtedly be a valuable in vivo tool to study the physiological and pathological functions of the mRNA cap and cap-binding proteins in more detail. In addition, our data suggest that the process of cap catabolism might be a novel target worth in vestigating f or the de v elopment of treatments for the after-effects of SARS-CoV-2 infection, or long COVID.

DA T A A V AILABILITY
RNA-seq data are available from the sequence read archi v e SRA under the accession numbers DRR440984 (uninduced sample) and DRR440985 (induced sample).