Unique Mutations in the Murine Hepatitis Virus Macrodomain Differentially Attenuate Virus Replication, Indicating Multiple Roles for the Macrodomain in Coronavirus Replication

ABSTRACT

All coronaviruses (CoVs) contain a macrodomain, also termed Mac1, in nonstructural protein 3 (nsp3) that binds and hydrolyzes mono-ADP-ribose (MAR) covalently attached to proteins. Despite several reports demonstrating that Mac1 is a prominent virulence factor, there is still a limited understanding of its cellular roles during infection. Currently, most of the information regarding the role of CoV Mac1 during infection is based on a single point mutation of a highly conserved asparagine residue, which makes contact with the distal ribose of ADP-ribose. To determine if additional Mac1 activities contribute to CoV replication, we compared the replication of murine hepatitis virus (MHV) Mac1 mutants, D1329A and N1465A, to the previously mentioned asparagine mutant, N1347A. These residues contact the adenine and proximal ribose in ADP-ribose, respectively. N1465A had no effect on MHV replication or pathogenesis, while D1329A and N1347A both replicated poorly in bone marrow-derived macrophages (BMDMs), were inhibited by PARP enzymes, and were highly attenuated in vivo. Interestingly, D1329A was also significantly more attenuated than N1347A in all cell lines tested. Conversely, D1329A retained some ability to block beta interferon (IFN-β) transcript accumulation compared to N1347A, indicating that these mutations have different effects on Mac1 functions. Combining these two mutations resulted in a virus that was unrecoverable, suggesting that the combined activities of Mac1 are essential for MHV replication. We conclude that Mac1 has multiple functions that promote the replication of MHV, and that these results provide further evidence that Mac1 is a prominent target for anti-CoV therapeutics.

IMPORTANCE In the wake of the COVID-19 epidemic, there has been a surge to better understand how CoVs replicate and to identify potential therapeutic targets that could mitigate disease caused by SARS-CoV-2 and other prominent CoVs. The highly conserved macrodomain, also termed Mac1, is a small domain within nonstructural protein 3. It has received significant attention as a potential drug target, as previous studies demonstrated that it is essential for CoV pathogenesis in multiple animal models of infection. However, the functions of Mac1 during infection remain largely unknown. Here, using targeted mutations in different regions of Mac1, we found that Mac1 has multiple functions that promote the replication of MHV, a model CoV, and, therefore, is more important for MHV replication than previously appreciated. These results will help guide the discovery of these novel functions of Mac1 and the development of inhibitory compounds targeting this domain.

INTRODUCTION

Coronaviruses (CoVs) are of the family Coronaviridae in the Nidovirales order and are responsible for a variety of diseases of both clinical and veterinary significance. These diseases range from potentially lethal human respiratory diseases, such as severe acute respiratory syndrome (SARS)-CoV, SARS-CoV-2, and Middle East respiratory syndrome (MERS)-CoV to mammalian gastrointestinal diseases such as porcine epidemic diarrhea virus (PEDV) and avian respiratory diseases such as infectious bronchitis virus (IBV) (1). As SARS-CoV-2 continues to be a significant health threat, and as there will likely be further CoV outbreaks in the future, there is an urgent need for a better understanding of the mechanisms used by CoVs to promote their replication and cause severe disease.

Like other members of the Nidovirales order, CoV genetic information is stored as nonsegmented, positive-sense RNA ranging in size from 26 to 32 kb. These genomes can be broken down into the region encoding structural and accessory proteins, comprising approximately 10 kb at the 3′ end of the genome, and the region encoding nonstructural proteins (nsps), consisting of approximately 20 kb at the 5′ end of the genome. Conserved structural proteins include the spike, envelope, membrane, and nucleocapsid proteins. Accessory proteins have important roles in viral pathogenesis, such as antagonism of the type 1 IFN antiviral response, but are not essential for in vitro viral replication. The nsps are translated into two long polyproteins, named polyprotein 1a (pp1a) and polyprotein 1ab (pp1ab), which are cleaved into individual proteins by viral proteases. The nsps perform a variety of functions and include the polymerase, helicase, 2 proteases, and many others. However, the functions of many nsps are still being fully determined or remain completely unknown. The largest nonstructural protein, nsp3, contains several modular domains, including a ubiquitin-like domain, an acidic domain, one or two papain-like protease (PLP) domains, multiple transmembrane domains, and one or more macrodomains (1, 2).

Macrodomains are globular protein domains present in many different life forms, including humans, yeast, bacteria, and several families of positive-sense RNA viruses. They have a highly conserved “sandwich” structure that includes several central β-sheets surrounded by 3 α-helices on each side (3, 4). The primary biochemical functions of macrodomains are to bind and hydrolyze mono-ADP-ribose (MAR) from proteins. There are additional macrodomains (Mac2/Mac3) in some CoVs, including SARS-CoV-2, that do not bind MAR and instead bind to nucleic acids or cellular proteins to promote virus replication (5–11).

ADP ribosylation is the posttranslational covalent addition of ADP-ribose to proteins by ADP-ribosyltransferases that utilize NAD⁺ as a substrate. ADP-ribose can be added to proteins as single subunits (MAR) or as chains of multiple subunits (poly-ADP-ribosylation, or PAR). This process is performed intracellularly by poly-ADP-ribose polymerases (PARPs), also known as diphtheria toxin-like ADP-ribosyltransferases (ARTDs). There are 17 human PARPs, with PARP-1 being the most well studied, as it mediates most of the PARylation that occurs in the cell. Much less is known about the MARylating PARPs, although many of them are interferon (IFN)-stimulated genes (ISGs) and some have demonstrated antiviral activities (12). For example, PARP12-mediated ADP ribosylation impedes the replication of Zika virus by ADP-ribosylating NS1 and NS3, leading to their proteasomal degradation (13). PARP activity is countered by several dePARylating or deMARylating enzymes, including PARGs (polyADP-ribose glycohydrolase), ARHs (ADP-ribosylhydrolases), and macrodomains (14).

Recombinant viruses mutated at a highly conserved asparagine residue in the primary CoV macrodomain (here referred to as Mac1) have been engineered for multiple CoVs to understand the role of Mac1 during infection. This residue was targeted because its mutation to alanine had eliminated Mac1 phosphatase activity, was later shown to eliminate ADP-ribosylhydrolase activity, and is completely conserved among all enzymatically active macrodomains (15–18). Structurally, it is positioned to provide critical hydrogen bonds with the terminal ribose, positioning the ADP-ribose for hydrolysis (15, 19, 20). Recombinant viruses with this mutation generally replicate normally in tissue culture cells but are highly attenuated in vivo, and they have also been shown to replicate poorly in primary bone marrow-derived macrophages (BMDMs) (16, 21–23). These reports have established Mac1 as a prominent virulence factor and potential therapeutic target (4). For instance, the SARS-CoV Mac1 mutant virus, N1040A, replicates poorly in mice and induces an increased IFN and proinflammatory cytokine response in mice and Calu-3 cells, a bronchial epithelial cell line. (16). These results demonstrated that Mac1 is required for the ability of SARS-CoV to fully inhibit IFN and cytokine induction. Similar results were observed following infection of BMDMs with the JHM strain of MHV (JHMV) containing the same asparagine-to-alanine mutation (N1347A) in Mac1 (19, 23). We also found that treatment with PARP inhibitors or short interfering RNAs targeting PARP12 or PARP14 enhanced the replication of N1347A in BMDMs. These results demonstrated that the CoV nsp3 Mac1 domain is required to prevent PARP-mediated inhibition of virus replication (23).

While these studies have provided significant insight into the role of the CoV Mac1 domain, they are largely based on this single asparagine-to-alanine mutation. As this mutation may more significantly impact enzyme activity than MAR binding, it remains unclear whether other activities of Mac1 have additional roles in CoV replication. MAR can be covalently bound to a number of different amino acids, including acidic, basic, serine, and cysteine residues, while macrodomains have only been shown to hydrolyze MAR attached to acidic residues. Previous studies on the alphavirus macrodomains showed that mutations separating the ADP-ribose binding and hydrolase activities result in distinct phenotypes during virus infection (24, 25). These results indicate that macrodomain binding to MARylated proteins with ADP-ribose attached at non-cleavable residues, such as serine, may have functions distinct from its ADP-ribosylhydrolase activity.

Here, we compared the replication of recombinant JHMV with point mutations at three distinct residues (N1347, D1329, and N1465) that are located at the opposite ends of the ADP-ribose binding pocket. Biochemical data from other macrodomain proteins have shown that mutation of the aspartic acid, which is located in close proximity to the adenine base of ADP-ribose, dramatically reduces MAR binding but retains some enzyme activity. In contrast, the mutation of the first asparagine, which interacts with the distal ribose, dramatically reduces enzyme activity with only modest effects on MAR binding (16, 25–27). The final asparagine, N1465, appears to contact the proximal ribose but has not been previously studied biochemically. We found that mutation of D1329 to alanine (D1329A) had more severe replication defects in cell culture than N1347A but retained some ability to block IFN production compared to N1347A. In contrast, N1465A acted like wild-type (WT) virus in all assays tested. PARP inhibitors enhanced the replication of D1329A, and NAD-enhancing compounds further decreased its replication, indicating that the defect of this virus was due to PARP activity. Finally, we failed to recover a recombinant virus containing both the D1329A and N1347A mutants, or a separate mutant, G1439V, predicted to have diminished MAR binding and hydrolysis activity, suggesting that the combined activities of Mac1 are essential for JHMV replication.

RESULTS

Structure of the JHMV Mac1 ADP-ribose binding pocket.

As there is no published structure for the MHV Mac1, we used computer modeling to predict its overall structure and the structure of its ADP-ribose binding pocket (Fig. 1A and B). Not surprisingly, the structure of the MHV Mac1 protein is similar to that of other CoV Mac1 proteins (Fig. 1C and D). One residue predicted to impact ADP-ribose binding is a highly conserved aspartic acid residue located in a loop region between β2 and α1 of MHV Mac1 (D16) (Fig. 1A, B, and E to F). This residue is present in most macrodomains across all species, including bacteria, yeast, and humans, and is known to either make a critical hydrogen bond with the N6 nitrogen of the adenine ring or mediate water contacts with this molecule (Fig. 1B) (15, 20, 28, 29). Its position in the ADP-ribose binding pocket is well conserved and largely superimposes with that of the SARS-CoV-2 and MERS-CoV Mac1 proteins (Fig. 1E and F). Mutation of this aspartic acid in multiple macrodomains virtually eliminates MAR binding but maintains some ADP-ribosylhydrolase activity (16, 26, 27). Therefore, determining the role of this residue in virus replication and pathogenesis may provide unique insight into the functions of Mac1. Another residue of interest is an asparagine located in a loop between β7 and α6 of MHV Mac1 (N152) (Fig. 1A, B, E, and F). This residue is mostly conserved among β-CoVs, and in our MHV-1 modeled structure, it appears to provide hydrogen bonds to the proximal ribose. It has also been proposed to provide similar interactions in the SARS-CoV and MERS-CoV Mac1 proteins (20, 28). This residue is often found as a hydrophobic amino acid in other CoV and viral macrodomains, including phenylalanine in SARS-CoV-2 Mac1 (30). These side chains appear to be in close contact with the adenine base and may stack against the adenine ring to create water-mediated hydrogen bonds with the proximal ribose (31). Based on these observations, we hypothesized that alanine mutations at these residues are detrimental to MHV replication or pathogenesis.

FIG 1.

Rosetta-derived structure of the MHV Mac1 protein. (A) Cartoon representation of MHV Mac-1 with ADP-ribose as determined by Rosetta. (B) Hydrogen bond interactions (dashed lines) between ADP-ribose and amino acids with modeled water molecules. (C and D) Superposition of MHV Mac1 (green) (PDB entry 6WOJ) with other CoV Mac1 structures. (C) SARS-CoV-2 Mac1 with ADP-ribose (magenta) (PDB entry 6WOJ). (D) MERS-CoV Mac1 with ADP-ribose (cyan) (PDB entry 5DUS). (E and F) Superposition of MHV Mac1 (green) with other CoV Mac1 structures highlighting the ADP-ribose binding site. (E) SARS-CoV-2 (magenta). (F) MERS-CoV (cyan). The ADP-ribose molecules are colored gray for MHV (A to F) and are rendered as orange cylinders for SARS-CoV-2 Mac1 (C and E) and blue cylinders for MERS-CoV Mac1 (D and F). Conserved waters are shown as red spheres. Residue numbering is based on MHV sequence.

The MHV Mac1 D1329A virus is highly attenuated in all cell types tested, while N1465A replicates like WT virus.

To test the role of the D16 (D1329 in pp1a) and N152 (N1465 in pp1a) residues in the context of MHV replication, we first engineered recombinant JHMV bacterial artificial chromosomes (BACs) containing the D1329A and N1465A mutations using a two-step Red recombination with the endonuclease I-SceI as previously described (22, 32). These recombinant viruses, termed D1329A and N1465A, were easily recovered and replicated in cell culture. Following reconstitution of virus, we sequenced the Mac1 region and confirmed that these mutations had been retained following passaging.

We first compared the replication of D1329A and N1465A with that of WT and N1347A viruses in BMDMs. We previously showed that N1347A had significantly decreased replication in BMDMs and initially hypothesized that D1329A and N1465A would have similar defects. However, N1465A replicated like WT virus in BMDMs, indicating that this residue does not dramatically affect Mac1 functions. In contrast, N1347A and D1329A were highly attenuated and replicated at similar levels, showing a greater than 1-log replication defect throughout the infection (Fig. 2A). We next tested the replication of D1329A and N1465A in 17Cl-1 fibroblasts, a common cell line that is highly permissive for MHV replication. As expected, N1347A replicated like WT virus in these cells, as did N1465A, but, surprisingly, D1329A replicated poorly in these cells, with a replication defect of ∼1 log (Fig. 2B and C), and produced reduced levels of viral nucleocapsid (N) and spike (S) protein (Fig. 2D). To confirm that this defect was not due to a second-site mutation in the BAC, we repaired this mutation to create the BAC clone, repD1329. The virus recovered from this BAC clone replicated well and produced viral proteins at WT virus levels in 17Cl-1 cells, confirming that the replication defect was due to mutation of D1329 (Fig. 2C and D).

FIG 2.

D1329A, but not N1465A, is highly attenuated in both primary cells and in cell lines. BMDM (A), 17Cl-1 (B to D), L929 (E and F), and DBT (G and H) cells were infected with WT, N1347A, D1329A, N1465A, and repD1329 viruses as described in Materials and Methods. Cell-associated and cell-free virus was collected at indicated time points, and virus titers were determined by plaque assay. In addition, cell lysates were collected and viral protein levels were determined by immunoblotting. The data show one experiment representative of at least two independent experiments (A to C, E, and G) with n = 4 (A) or n = 3 (B, C, E, and G) in each experiment. The data in panels D, F, and H show one experiment representative of two independent experiments.

Next, we tested whether D1329A was defective in additional MHV-susceptible cells. In both L929 fibroblasts and DBTs (delayed brain tumors), an astrocytoma cell line, D1329A was highly attenuated, with replication defects of approximately 2 logs, and very little N and S protein accumulated in both cell types compared to WT virus (Fig. 2E to H). Interestingly, we found that N1347A also replicated poorly and produced reduced N and S protein in DBT cells, although it replicated better than D1329A (Fig. 2G and H). These results indicate that the D1329A mutation disrupts a particular function of Mac1 that is not disrupted by N1347A.

The MHV Mac1 D1329A virus mostly retains the ability to repress IFN production in BMDMs.

We have previously demonstrated that the N1347A infection resulted in a >10-fold increase in IFN-β transcript levels in BMDMs compared to WT virus (23). To determine if D1329A or N1465A also induced IFN production above WT levels, we measured the level of IFN-β mRNA at 6 and 12 h postinfection (hpi) in BMDMs infected with WT, N1347A, D1329A, and N1465A. Here, infection with N1347A resulted in a >19-fold increase in IFN-β mRNA levels at 12 hpi and a smaller difference at 6 hpi (Fig. 3A), consistent with our previous report. N1465A-infected cells had IFN-β transcript levels similar to those of cells infected with WT virus, further indicating that this mutation does not affect Mac1 function. Finally, while IFN-β mRNA levels were increased 5.5-fold in D1329A-infected cells compared to WT virus, these levels were reduced 3.5-fold compared to N1347A, despite having roughly the same level of viral genomic RNA (gRNA) in the cells in this specific experiment (Fig. 3A and B). These results indicate that D1329A generally replicates worse than N1347A but retains some ability to block IFN-β transcription, providing strong evidence that D1329A differentially impacts the functions of Mac1 compared to N1347A. This result, in combination with the results in Fig. 2, suggests that Mac1 utilizes multiple mechanisms to promote virus replication and block innate immune responses in cell culture.

FIG 3.

D1329A infection results in lower IFN-β mRNA levels than N1347A infection in BMDMs. (A and B) BMDMs were infected with WT, N1347A, D1329A, or N1465A recombinant virus. Cells were collected at the indicated times postinfection and RNA was purified. IFN-β mRNA (A) and gRNA (B) levels were determined by RT-qPCR using primers listed in Table 2 and normalized to HPRT mRNA levels. The data in panels A and B show one experiment representative of four independent experiments with n = 4 for each experiment.

PARP inhibitors significantly enhance the replication of MHV D1329A, while boosting the PARP substrate NAD⁺ with nicotinamide riboside further inhibits D1329A replication.

We have recently shown that the N1347A mutant is susceptible to PARP-mediated inhibition of virus replication (23). We hypothesized that D1329A also enhances the susceptibility of MHV to PARP-mediated inhibition; however, it is also conceivable that this mutation could affect other functions of Mac1, such as papain-like protease (PL-Pro) or nucleic acid binding (33, 34). To test this hypothesis, we treated cells with inhibitors that target host PARP enzymes. If the ADP-ribosyltransferase activity of PARP enzymes are restricting the replication of D1329A, we would expect to see at least partial restoration of virus replication in the presence of PARP inhibitors, as we previously showed with N1347A (23). First, we performed 3-(4,5-dimethyl-2-thiazolyl)-2,5-diphenyl-2H-tetrazolium bromide (MTT) assays to confirm that the PARP inhibitors XAV-939 and olaparib (2281) did not affect the metabolic capacity of BMDMs or 17Cl-1 cells. BMDM and 17Cl-1 cells were treated with XAV-939 and 2281 for 24 h, and MTT levels were measured. Neither compound resulted in notable metabolic changes at a working concentration of 10 μM (Fig. 4A and B). In virus replication experiments, both compounds significantly enhanced the replication of D1329A in BMDMs and in 17Cl-1 cells, by approximately 5 and 3-fold, respectively, but had no effect on WT virus (Fig. 4C and D). In addition, the enhancement of D1329A replication in BMDMs was similar to that of N1347A. These results indicate that PARP-mediated ADP-ribosylation inhibits the replication of D1329A.

FIG 4.

D1329A replication is significantly increased by the addition of PARP inhibitors. (A and B) BMDMs (A) and 17Cl-1 cells (B) were treated with the indicated compounds, and, at 24 h, cell viability was measured using an MTT assay as described in Materials and Methods. (C and D) WT BMDMs (C) and 17Cl-1 cells (D) were infected with WT, N1347A, or D1329A and then treated with 0.25% dimethyl sulfoxide (DMSO), 10 μM olaparib (2281), or 10 μM XAV-939 as described in Materials and Methods. Progeny virus was collected at the indicated time points, and virus titers were determined by plaque assay. The data in panels A to D show one experiment representative of two independent experiments, with n = 4 (A and B) or n = 3 (C and D) for each experiment.

To provide further evidence for PARP-mediated inhibition of D1329A, we hypothesized that increasing PARP activity using nicotinamide riboside chloride (NR) would lead to further reduction in D1329A replication. NR enhances PARP activity by increasing the intracellular levels of the PARP substrate, NAD⁺ (35, 36). We recently showed that NAD levels are depleted following MHV infection, and that restoring NAD levels with NR and other NAD-boosting compounds both increased PARP activity and decreased the replication of N1347A (37). We first confirmed that NR did not significantly decrease the metabolic activity of BMDM or 17Cl-1 cells at and above the working concentration of 100 μM (Fig. 5A and B). Rather than decreasing metabolic activity, 17Cl-1 cells treated with NR seemed to have slightly increased metabolic activity, although this was not statistically significant (Fig. 5B). For the infection, we pretreated these cells with NR for 4 h, infected them with WT and D1329A, added fresh NR after the infection, and then collected both cell-free and cell-associated virus at 20 hpi. Similar to our results with N1347A, we found that NR significantly reduced the replication of D1329A virus in both cell types by >5-fold but had no impact on WT virus (Fig. 5C and D). These results indicate that NR enhanced PARP activity that further reduced D1329A virus replication but was countered by Mac1 in WT virus-infected cells.

FIG 5.

Addition of NR, a precursor of the PARP substrate NAD⁺, further decreases the replication of D1329A. (A and B) BMDMs (A) and 17Cl-1 cells (B) were treated with NR as described in the text, and at 24 h cell viability was measured using an MTT assay as described in Materials and Methods. (C and D) WT BMDMs (C) and 17Cl-1 cells (D) were either mock treated (H₂O) or treated with NR as described in Materials and Methods. Cell-associated and cell-free virus was collected at the indicated time points, and virus titers were determined by plaque assay. The data in panels A to D show one experiment representative of two independent experiments, with n = 4 (A and B) or n = 3 (C and D) for each experiment.

Infection of mice with Mac1 ADP-ribose binding mutants.

As results in cell culture may not mimic conditions in vivo, we tested the effect of these mutations in mice, where WT JHMV causes lethal encephalitis (22, 38). We first infected C57BL/6 mice intranasally with recombinant WT, N1347A, and N1465A JHMV at 3 × 10⁴ PFU and monitored weight loss as a clinical sign of disease progression (Fig. 6A and B). N1347A was included as an attenuated control virus, as it was previously shown to cause minimal disease in B6 mice (22). We hypothesized that N1465A would cause disease similar to that of WT virus, as all cell culture experiments indicated that this mutation had not affected Mac1 function. Indeed, N1465A-infected mice lost weight and were euthanized due to severe disease at nearly the same rate as WT-infected mice, while N1347A did not cause any weight loss and all mice survived its infection (Fig. 6A and B). These results provide further evidence that mutation of N1465 does not significantly affect the function of Mac1.

FIG 6.

D1329A, but not N1465A, is highly attenuated in vivo. (A and B) WT male and female B6 mice were infected with 3 × 10⁴ PFU of WT, N1347A, and N1465A intranasally and monitored for survival and weight loss daily for 12 days. The data show the combined results from two independent experiments. WT, n = 5 for male and female mice; N1347A, n = 5 for male and female mice; N1465A, n = 4 for male and female mice. (C and D) WT male and female B6 mice were infected with 3 × 10³ PFU of WT and D1329A as described for panels A and B. The data show the combined results of two independent experiments. WT, n = 5 for male and female mice; D1329A, n = 7 for male and n = 6 for female mice. (E) D1329A virus readily reverts in vivo. The brains of 3 mice (2 male [M]; 1 female [F]) that succumbed to infection with D1329A were harvested, and their viral RNA was amplified by RT-PCR with Mac1-specific primers. The PCR product was sequenced by Sanger sequencing and analyzed using DNA Star software. Mouse number 1 and number 3 reverted to wild-type virus sequence, while mouse number 2 evolved 2 new mutations, L1313V and K1327E. (F) Rosetta predicted structures of WT MHV Mac1 (green) around the adenine base showing the positions of D1329, L1313, and K1327.

Next, we infected mice intranasally with 3 × 10³ PFU of WT and D1329A virus. This low dose was used because titers of the D1329A virus stocks were significantly lower than those of other viruses. Regardless, WT virus still caused significant weight loss, and the infected mice succumbed to this infection at roughly the same rate as those receiving the higher dose of virus. Interestingly, the D1329A virus did lead to weight loss and lethality in 5 out of 7 male mice and 1 out of 6 female mice (Fig. 6C and D). Since this result was unexpected, we collected the brains of 1 female and 2 male mice that succumbed to infection with D1329A and sequenced the Mac1 region of the MHV genome. We found that in 2 of the 3 mice, the alanine reverted back to aspartic acid, while in the third mouse there were two second-site mutations in the N terminus of the macrodomain, L1313V and K1327E (Fig. 6E), with both residues located just outside D1329 (Fig. 6F). As all mice sequenced had reverted virus, it appears that D1329A replicates poorly in vivo and is especially prone to reversion. We conclude that D1329A is highly attenuated in vivo.

Recombinant JHMV with both D1329A and N1347A mutations is not recoverable.

To further test the hypothesis that Mac1 has multiple roles in virus replication, we created a D1329A/N1347A double mutant BAC clone. We hypothesized that if the D1329 residue confers a unique role in virus replication compared to N1347, then a double mutant virus may be even more attenuated than either single mutant. Consistent with this hypothesis, we were unable to recover the D1329A/N1347A virus following 12 transfection attempts from 2 separate BAC clones. Each of these experiments included successful transfections of other BACs as positive controls (Table 1).

TABLE 1.

MHV BAC recovery rates in BHK-MVR cells

BAC	No. of viruses recovered/no. of attempts	Recovery (%)
N1347A	9/9	100
D1329A	9/9	100
N1347A/D1329A	0/12	0
G1439V	4/11	36.37
A1438T/G1439V	6/6	100

Introducing multiple mutations into the macrodomain could disrupt its structure; thus, the inability to recover this virus could be due to impairment of additional nsp3 functions outside Mac1. To provide further evidence that the loss of Mac1 function is lethal for JHMV, we engineered an additional mutant, G1439V, into our JHMV BAC clone. This mutation has previously been introduced into SARS-CoV and SARS-CoV-2 recombinant Mac1 proteins, and both proteins had minimal, if any, ADP-ribosylhydrolase activity (16, 39). Furthermore, the insertion of a large valine residue may prevent ADP-ribose from entering the binding pocket. Indeed, computational modeling of the SARS-CoV and MERS-CoV Mac1 proteins found that ADP-ribose binding was highly unfavorable when introducing the G-V mutation, with a ddG of binding of ∼9 and 10 Rosetta energy units (REUs), respectively. After transfection, we only recovered this virus 4/11 times for a 37% recovery rate (Table 1). Upon sequencing the recovered viruses, we found that most had reverted to WT virus. However, one clone instead had evolved a second-site mutation of alanine to threonine in the residue immediately upstream of V1439. To determine if this mutation was responsible for the ability of this virus to replicate, we created an A1438T/G1439V mutant BAC clone. We were able to easily recover this virus, indicating that this mutation was the reverting mutation that allowed the outgrowth of one of the G1439V viruses (Table 1). A1438T/G1439V replicated better than D1329A on L929 cells but had a 3-fold replication defect and produced less viral protein on L929 cells than WT virus (Fig. 7A and B). It also did not cause severe disease in mice, indicating this mutation only partially recovered Mac1 functions (Fig. 7C and D). In total, we have created two separate Mac1 recombinant JHMV clones (D1329A/N1347V and G1439V) that were not recoverable without second-site or reverting mutations. These results suggest that the combined activities of Mac1 are essential for JHMV replication.

FIG 7.

A1438T/G1439V virus, a revertant of G1439V, is recoverable but replicates at slightly reduced levels compared to WT virus and is attenuated in vivo. (A and B) L929 cells were infected with WT, D1329A, and AG1438/1439TV viruses as described in Materials and Methods. Cell-associated and cell-free virus was collected at 24 hpi, and virus titers were determined by plaque assay. In addition, cell lysates were collected at 24 hpi, and viral protein levels were determined by immunoblotting. The data in panel A show one experiment representative of two independent experiments, with n = 3 for each experiment. The data in panel B show one experiment representative of two independent experiments. (C and D) WT male B6 mice were infected with 3 × 10³ PFU of WT and A1438T/G1439V intranasally and monitored for survival (C) and weight loss (D) daily for 12 days. WT, n = 3; A1438T/G1439V, n = 6. (D and E) Rosetta predicted structures of WT MERS-CoV Mac1 around the proximal ribose compared to G128V (E) and A127T/G128V (F) proteins. G128V (orange cylinders) is predicted to cause a disruption of water molecules (E) (WT, red spheres; G128V or A127T/G128V, black spheres). (F) The A127T mutation (salmon cylinders) is predicted to restore this water network back to their original location, likely restoring critical hydrogen bonds.

DISCUSSION

Based on prior reports demonstrating the importance of Mac1 for CoV pathogenesis, there has been particular interest in developing therapeutic strategies targeting the coronavirus Mac1 domain (16, 31, 39–44). Here, we show that the JHMV Mac1 protein domain uses multiple functions to promote virus replication and block innate immune signaling. These results further demonstrate the importance of Mac1 for CoV replication, will help identify novel mechanisms by which Mac1 promotes virus replication, and could have implications in the design of compounds targeting the ADP-ribose binding domain of Mac1.

The two most common functions for viral macrodomains are to bind MARylated proteins and hydrolyze MAR from proteins. Viral macrodomain proteins are in the MacroD family of macrodomains and have only been demonstrated to remove MAR from acidic residues (4). Several PARPs can MARylate acidic residues (27, 45), but recent proteomic analyses have shown that under certain conditions, MARylation of acidic residues is rare or even absent (46). Thus, it is likely that viral macrodomains would also target proteins modified at nonacidic residues, such as serine, cysteine, or lysine, even if they cannot hydrolyze these modifications. Previous studies of the alphavirus macrodomain have provided unique insight into their separate MAR binding and hydrolase activities (24, 25, 27, 47). Chikungunya (CHIKV) and Sindbis virus mutations that impact these activities are highly attenuated in cell culture and in vivo. In fact, CHIKV macrodomain mutations that reduced hydrolysis activity by as little as 25% are attenuated, and any mutation that eliminated binding activity was unrecoverable (27). In addition, two mutants were further analyzed for their impact on the virus replication cycle, G32S (reduced hydrolysis and binding) and Y114A (decreased hydrolysis but increased binding). While both mutants had reduced virus replication, as measured by plaque assay, the G32S mutant had greatly reduced levels of viral RNA and protein production early in the infection, while Y114A had normal levels of viral RNA and protein until later stages of infection. It was concluded that a CHIKV with MAR binding defects is unable to properly initiate replication, while a hydrolase deficiency negatively affects the later stages of virus replication.

Most studies of the CoV Mac1 domain have focused on mutation of a highly conserved asparagine residue. Biochemical data from other macrodomains have shown that mutation of this residue to alanine largely ablates hydrolase activity while only reducing MAR binding 2- to 3-fold (15, 16, 18, 25, 27). Studies of CoVs containing this mutation have demonstrated that these viruses often replicate normally in cell culture but induce increased IFN and cytokine expression during infection and are highly attenuated in vivo (16, 21–23). In addition, we have shown that the JHMV N1347A mutant virus replicates poorly in BMDMs and DBT cells but not in other cell lines (Fig. 2). It is unclear why N1347A is only attenuated in certain cells. We hypothesize that this could be due to the presence of a specific PARP or cellular target(s) in these cells.

To determine if there are additional or altered phenotypes associated with the MAR binding ability of Mac1, we chose to mutate residues in the adenine binding pocket of the JHMV Mac1 domain. First, we targeted N1465, located in a loop between β7 and α6, which appears to provide a hydrogen bond with the proximal ribose (Fig. 1B), although its role in ADP-ribose binding has not been studied biochemically. Despite its apparent interaction with the proximal ribose (20, 28), we found that mutation of this residue to alanine had no impact on JHMV replication or IFN induction in any cell type we tested (Fig. 2 and 3) and also did not affect the ability of JHMV to cause disease in mice (Fig. 6A and B). From these results, we conclude that N1465 does not contribute significantly to the ability of the JHMV Mac1 to promote replication or cause disease.

Next, we targeted a conserved aspartic acid that makes a critical hydrogen bond with the N7 nitrogen of the adenine base, D1329. Biochemical data from other macrodomains has demonstrated that this aspartic acid mutation leads to a dramatic loss (>10-fold) of MAR binding (25–27). While D1329A replicated similarly to N1347A in BMDMs, it had lower levels of replication than N1347A in all other cell lines tested, including 17Cl-1, L929, and DBT cells. These results suggest that D1329 contributes to virus replication in a unique manner compared to N1347. We hypothesize that MAR binding is critical for JHMV replication in cell culture, and that the D1329A mutation has a greater effect on MAR binding than does the N1347A mutation, leading to more severe replication defects. In contrast, the D1329A virus retained some ability to block IFN production compared to N1347A (Fig. 3). This result suggests that IFN repression depends more on hydrolysis activity than binding activity, as the asparagine mutation has been shown in other macrodomain proteins to have a more severe effect on hydrolysis than the aspartic acid mutation (16, 27). It will be important to test these hypotheses using recombinant MHV Mac1 proteins to directly measure their binding and hydrolysis activities and identify mutations with different levels of activity so that phenotypes from virus infections can be correlated with biochemical activities.

While ADP-ribose binding and hydrolysis are the two major activities of the viral macrodomains, it is possible that other functions contribute to the phenotypes seen here. Other potential Mac1 functions include nucleotide binding, PL-Pro binding, ADP-ribose-1″-phosphatase activity, and PAR binding or hydrolysis (15, 33, 34). We have shown that the defects of D1329A and N1347A are largely due to PARP-dependent ADP-ribosylation (Fig. 4 and 5) (23, 37). Thus, it is unlikely that loss of PL-Pro binding, nucleotide binding, or ADP-ribose-1″-phosphatase activity significantly contributes to the phenotypes of either D1329A or N1347A. However, we have not ruled out the possibility that PAR binding or hydrolysis impacts the replication of JHMV. Some studies have reported structural models describing how Mac1 might bind to PAR (15, 34, 39), but biochemical data have indicated that PAR binding and hydrolysis activities of CoV Mac1 domains are often weak and significantly reduced compared to other viral macrodomains (30, 34, 48). Regardless, in vitro assays may lack precise cellular conditions or factors that could affect these activities; thus, the possibility that these domains could bind or hydrolyze PAR cannot be excluded.

In addition to its robust attenuation in cell culture, we found that D1329A often reverted in mice. While in some cases it reverted to WT, in one case two novel mutations near D1329, L1313V and K1327E, appeared. Both residues are found just behind the aspartic acid residue, though not in direct contact with the substrate (Fig. 6F). It is possible that the loss of the hydrogen bond with adenine in the N1329A mutant leads to destabilization at the N terminus of the protein, allowing for new contacts to be made. We hypothesize that without this important hydrogen bond, the lysine at position 1327 moves into a position where it can interact with the adenine base. Mutation of this residue to glutamic acid then would allow this residue to provide a hydrogen bond with the adenine base, restoring ADP-ribose binding. It will be interesting to test this hypothesis by determining if this lysine-to-glutamic acid mutation can help restore replication or MAR binding activity of D1329A.

Surprisingly, we were unable to recover a double mutant virus, D1329A/N1347A, or a separate mutant designed to negatively impact both binding and hydrolase activities, G1439V. These results indicate that the combined activities of Mac1 are essential for JHMV replication. Intriguingly, we identified a second site mutation for G1439V in an immediately adjacent residue, A1438T, that allowed it to replicate. Of note, all alphaviruses have a threonine in this position. Based on computer modeling of the MERS-CoV Mac1 protein, we predict that the G1439V (G128V by MERS-CoV protein numbering) mutation displaces a network of water molecules in the vicinity of these residues, and that the A1438T (A127T by MERS-CoV protein numbering) mutation restores this water network to its original location (Fig. 7E and F). After recreating an isogenic BAC clone with these two mutations, A1438T/G1439V, we found that this clone was easily recoverable and replicated in cell culture, although it had a mild replication defect compared to WT virus in L929 cells and was attenuated in vivo (Fig. 7A to D). Regardless, these results confirmed that this mutation allowed us to recover the G1439V mutant virus and provided more evidence that Mac1 may be essential for JHMV replication.

The primary challenge remaining is to identify targets of Mac1 that contribute to the inhibition of Mac1 mutant viruses. While Mac1 is likely to target both cellular and viral proteins during infection, the severe and universal defect of D1329A indicates that this mutation impacts the ability of Mac1 to target an ADP-ribosylated CoV protein. We recently found that the CoV nucleocapsid (N) protein is ADP ribosylated, and this modification was unchanged following infection with N1347A, indicating it is not cleaved by Mac1 but could be a target for Mac1 binding (49). Regardless, further investigation into how Mac1 impacts the function of viral and cellular proteins is likely to uncover unique insights into the replication of CoVs.

MATERIALS AND METHODS

Molecular modeling of the MHV Mac1 protein.

A model of MHV Mac1 protein bound to ADP-ribose was generated using RosettaCM (50), using the ADP-ribose-bound MERS-CoV Mac1 structure, PDB entry 5DUS, as a template (20). Following the threading step, the pose of ADP-ribose bound to the MERS-CoV macrodomain was included in the refinement process to create a model of the MHV Mac1 protein bound to ADP-ribose. The top-scoring model was aligned to the MERS Mac1 protein, and three structural waters that participate in hydrogen bond networks bridging the protein to ADP-ribose (water residues 312, 365, and 384) were added to the MHV model, followed by hydrogen bond optimization and minimization within Maestro by Schrodinger.

Rosetta Cartesian ddG calculations.

The structures of the MERS-CoV and SARS-CoV Mac1 proteins bound to ADP-ribose, PDB entries 5DUS (20) and 2FAV (15), respectively, were stripped of all waters but the three structural waters bridging the macrodomain to ADP-ribose (20). These structures, as well as the model of the MHV bound to ADP-ribose and the three waters, were prepared for ddG prediction by relaxing into Cartesian space using coordinate restraints via Rosetta (51). The lowest-energy structure was selected for ddG prediction of the G128V mutant using Rosetta’s Cartesian ddG protocol (52). By default, the Cartesian ddG protocol runs for three iterations for the wild-type and G1439V proteins, and the predicted ddG was taken as the difference between the wild type and the mutant for the average of the three simulations. Commands are available upon request.

Cell culture.

17Cl-1, delayed brain tumor (DBT), 17Cl-1, L929, HeLa cells expressing the MHV receptor carcinoembryonic antigen-related cell adhesion molecule 1 (CEACAM1) (HeLa-MHVR), and baby hamster kidney cells expressing CEACAM1 (BHK-MVR) (all cell lines gifts provided by Stanley Perlman, University of Iowa) were grown in Dulbecco’s modified Eagle medium (DMEM) supplemented with 10% fetal bovine serum (FBS), 100 U/ml penicillin, 100 mg/ml streptomycin, HEPES, sodium pyruvate, nonessential amino acids, and l-glutamine. Bone marrow-derived macrophages (BMDMs) sourced from WT mice were differentiated by incubating cells in Roswell Park Memorial Institute (RPMI) medium supplemented with 10% L929 cell supernatants, 10% FBS, sodium pyruvate, 100 U/ml penicillin, 100 mg/ml streptomycin, and l-glutamine for 7 days. Cells were washed and placed in fresh medium every day after the 4th day.

Cell viability assay.

BMDMs and 17Cl-1 cells were treated with the indicated compounds for 24 h. Cellular metabolic activity was assessed using a CyQUANT MTT cell proliferation assay (Thermo Fisher Scientific) by following the manufacturer’s instructions.

Mice.

Pathogen-free C57BL/6 (B6) mice were originally purchased from Jackson Laboratories. Mice were bred and maintained in the animal care facility at the University of Kansas, as approved by the University of Kansas Institutional Animal Care and Use Committee (IACUC) and following guidelines set forth in the Guide for the Care and Use of Laboratory Animals (53).

Generation of recombinant pBAC-JHMV constructs.

All recombinant pBAC-JHMV constructs were created using Red recombination (primers are listed in Table 2). Recombinant WT (rJ^IA-GFPrevN1347) and N1347A (rJ^IA-GFP-N1347A) MHV were previously described. Recombinant BACs with the D1329A, G1439V, A1438T, and N1465A point mutations in the nsp3 macrodomain were engineered using the Kan^r-I-SceI marker cassette for dual positive and negative selection as previously described, using the primers listed in Table 2 (32). BAC DNA from Cml^r Kan^s colonies was analyzed by restriction enzyme digest, PCR, and direct sequencing for isolation of correct clones. PCR and sequencing were done using the following primers located just outside the Mac1 gene sequence: forward, 5′-GGCTGTTGTGGATGGCAAGCA-3′, and reverse, 5′-GCTTTGGTACCAGCAACGGAG-3′. Wild-type repaired BACs were engineered by reintroducing the wild-type sequence into the BAC clones containing the D1329A mutation using the same procedure described above. The resulting BAC clones were termed pBAC-JHMV^IA-GFP-D1329A, pBAC-JHMV^IA-GFP-N1465A, pBAC-JHMV^IA-GFP-D1329A/N1347A, pBAC-JHMV^IA-GFP-G1439V, pBAC-JHMV^IA-GFP-G1439V/A1438T, and pBAC-JHMV^IA-GFP-repD1329.

TABLE 2.

Primers used to create recombinant BACs

Gene	Forward 5′→3′a	Reverse 5′→3′a
D1329A	ggttcatgtataacaccaaatgtttgttttgttaaaggagatgttataaaggttttgcgcAGGATGACGACGATAAGTAGGG	atgacttcagcaccaactctgcgcaaaacctttataacatctcctttaacaaaacaaacaGCCAGTGTTACAACCAATTAACC
repD1329A	ggttcatgtataacaccaaatgtttgttttgttaaaggagctgttataaaggttttgcgcAGGATGACGACGATAAGTAGGG	atgacttcagcaccaactctgcgcaaaacctttataacagctcctttaacaaaacaaacaGCCAGTGTTACAACCAATTAACC
N1465A	tctacttggtgtagtgacgaagaatgtcattcttgtcagtaataaccaggatgattttgaAGGATGACGACGATAAGTAGG	cctgacacttctctatcacatcaaaatcatcctggttattactgacaagaatgacattctGCCAGTGTTACAACCAATTAACC
G1439V	aataagtgtgacaatgttgtcaccactttaatttcggctgttatatttagtgtgcctactAGGATGACGACGATAAGTAGGG	agataagttaaggaaacatcagtaggcacactaaatataacagccgaaattaaagtggtgGCCAGTGTTACAACCAATTAACC
A1438T/G1439V	tattaataagtgtgacaatgttgtcaccactttaatttcgactgttatatttagtgtgccAGGATGACGACGATAAGTAGGG	aagttaaggaaacatcagtaggcacactaaattaacagtcgaaattaaagtggtgacaaGCCAGTGTTACAACCAATTAACC

^aViral sequences indicated in lowercase; marker sequence indicated in lowercase.

Reconstitution of recombinant pBAC-JHMV-derived virus.

Approximately 1 × 10⁶ BHK-MVR cells were transfected with approximately 0.5 to 1 μg of pBAC-JHMV DNA and 1 μg of pcDNA-MHV-N plasmid using PolyJet (SignaGen) as a transfection reagent. New recombinant viruses used in this study were termed D1329A (rJ^IA-GFP-D1329A), N1465A (rJ^IA-GFP-N1465), G1439V (rJ^IA-GFP-G1439V), A1438T/G1439V (rJ^IA-GFP-A1438T/G1439V), and repD1329 (rJ^IA−GFP-repD1329). Virus stocks were created by infecting ∼1.5 × 10⁷ 17Cl-1 cells at a multiplicity of infection (MOI) of 0.1 PFU/cell and collecting both the cells and supernatant at 16 to 20 hpi. The cells were freeze-thawed, and debris was removed prior to collecting virus stocks. Virus stocks were quantified by plaque assay on HeLa-MHVR cells and sequenced by collecting infected 17Cl-1 or L929 cells using TRIzol. RNA was isolated and cDNA was prepared using MMLV-reverse transcriptase per the manufacturer’s instructions (Thermo Fisher Scientific). The Mac1 gene sequence was amplified by PCR using the same primers as described above for sequencing BACs, and then resulting PCR products were sequenced by Sanger sequencing using the forward primer. Sequence was analyzed using DNA Star software.

Virus infection.

Cells were infected with recombinant virus at an MOI of 0.05 to 0.1 PFU/cell with a 45- to 60-min adsorption phase, unless otherwise stated. Olaparib (number A4154; APExBIO) and XAV-939 (number A1877; APExBIO) were added to cells following the adsorption phase at 10 μM. Nicotinomide riboside chloride (NR) (Chromadex) was added to cells at 100 μM 4 h prior to infection and then readded immediately following the adsorption phase of the infection. Male and female mice, 5 to 8 weeks old (unless otherwise indicated), were anesthetized with ketamine-xylazine and inoculated intranasally with either 3 × 10³ or 3 × 10⁴ PFU of recombinant virus in a total volume of 12 μl DMEM. To obtain viral RNA from infected animals to sequence the virus following infection, mice were sacrificed and brain tissue was collected and homogenized in TRIzol (Invitrogen). RNA was isolated and cDNA was prepared using MMLV-reverse transcriptase per the manufacturer’s instructions (Thermo Fisher Scientific). The macrodomain gene sequence was amplified by PCR and sequenced as described above. Sequence was analyzed using DNA Star software.

Real-time qPCR analysis.

RNA was isolated from BMDMs using TRIzol (Invitrogen), and cDNA was prepared as described above. Quantitative real-time PCR (qRT-PCR) was performed on a QuantStudio3 real-time PCR system using PowerUp SYBR green master mix (Thermo Fisher Scientific). Primers used for quantitative PCR (qPCR) are listed in Table 3. Cycle threshold (C_T) values were normalized to the housekeeping gene hypoxanthine phosphoribosyltransferase (HPRT) by the following equation: C_T = C_{T(gene of interest)} − C_T(HPRT). Results are shown as a ratio to HPRT calculated as 2^−ΔCT.

TABLE 3.

Quantitative real-time PCR primers

Gene	Forward, 5′→3′	Reverse, 5′→3′
gRNA	AGGGAGTTTGACCTTGTTCAG	ATAATGCACCTGTCATCCTCG
IFN-β	TCAGAATGAGTGGTGGTTGC	GACCTTTCAAATGCAGTAGATTCA
HPRT	GCGTCGTGATTAGCGATGATG	CTCGAGCAAGTCTTTCAGTCC

Immunoblotting.

Total cell extracts were lysed in sample buffer containing SDS, protease and phosphatase inhibitors (Roche), β-mercaptoethanol, and a universal nuclease (Fisher Scientific). Proteins were resolved on an SDS polyacrylamide gel, transferred to a polyvinylidene difluoride (PVDF) membrane, hybridized with a primary antibody, reacted with an infrared (IR) dye-conjugated secondary anti-body, visualized using a Li-COR Odyssey Imager (Li-COR), and analyzed using Image Studio software. Primary antibodies used for immunoblotting included anti-MHV N and S monoclonal antibodies (54) and anti-actin monoclonal antibody (clone AC15; Abcam, Inc.). Secondary IR antibodies were purchased from Li-COR.

Statistical analysis.

All statistical analyses were done using an unpaired two-tailed Student's t test to assess differences in mean values between groups, and graphs are expressed as geometric means ± geometric standard deviations (SD) (virus titers) or ± standard errors of the means (SEM) (qPCR). The n value represents the number of biologic replicates for each figure. All data were analyzed using GraphPad Prism software. Significant P values are denoted with asterisks: *, P ≤ 0.05; **, P ≤ 0.01; ***, P ≤ 0.001.