Both ADP-Ribosyl-Binding and Hydrolase Activities of the Alphavirus nsP3 Macrodomain Affect Neurovirulence in Mice.

Viral encephalomyelitis is an important cause of long-term disability, as well as acute fatal disease. Identifying viral determinants of outcome helps in assessing disease severity and developing new treatments. Mosquito-borne alphaviruses infect neurons and cause fatal disease in mice. The highly conserved macrodomain of nonstructural protein 3 binds and can remove ADP-ribose (ADPr) from ADP-ribosylated proteins. To determine the importance of these functions for virulence, recombinant mutant viruses were produced. If macrodomain mutations eliminated ADPr-binding or hydrolase activity, viruses did not grow. If the binding and hydrolase activities were impaired, the viruses grew less well than the wild-type virus, induced similar innate responses, and caused less severe disease, and most of the infected mice recovered. If binding was improved, but hydrolase activity was decreased, the virus replicated well and induced greater innate responses than did the WT, but clearance from the nervous system was impaired, and mice remained paralyzed. Therefore, macrodomain function determined the outcome of alphavirus encephalomyelitis.

To measure hydrolase activity, the catalytic domain of PARP10 (PARP10 CD ) was incubated with 32 P-NAD ϩ to generate 32 P-mono-ADP-ribosylated (MARylated) PARP10 CD as a substrate for incubation with buffer alone, WT, or mutant nsP3 MDs for 1 h at 37°C, followed by analysis by SDS-PAGE and autoradiography (Fig. 1A). Autoradiographs were scanned and the ability of the mutant nsP3 MDs to remove 32 P-MAR from MARylated PARP10 CD was quantified and compared to the WT set at 100% (Fig. 1B). N24A, G32E, and TM strains had no detectable activity, whereas the G32S mutant had 73.5%, the G32A mutant had 72%, and the Y114A mutant had 42% of WT activity.
MDs bind mono-and poly-ADP-ribosylated (PARylated) substrates through interactions with the terminal ADPr moiety (59)(60)(61). Therefore, we reasoned that a free PAR chain labeled at the 1Љ terminus with a fluorescent probe could serve as a substrate for determining the binding affinity of the SINV nsP3 MDs for ADP-ribosylated substrates using fluorescence-based biophysical methods such as microscale thermophoresis (MST) (62). To generate a defined PAR chain fluorescently labeled at the 1Љ terminus, we used enzymatically made PAR and cleaved it from modified proteins with alkaline hydrolysis that results in loss of the terminal phosphoribose moiety, generating a 5=-phosphate (63). After purifying PAR to defined length with high-performance liquid chromatography (HPLC), 1-ethyl-3-(N,N-dimethylamino)isopropyl carbodiimide (EDC) coupling was used to introduce an alkyne at the 1Љ terminus of (ADPr) 18 . Cy5-azide was then conjugated to alkyne-(ADPr) 18 via Cu(I)-catalyzed cycloaddition. Cy5-(ADPr) 18 was incubated with 2-fold serial dilutions of SINV WT and mutant MDs, followed by MST analyses. From these assays, we determined that the nsP3 MD WT binds to Cy5-(ADPr) 18 with a K D of 44.49 Ϯ 2.46 M (Fig. 1C), a value comparable to the affinity of the homologous CHIKV nsP3 MD for free ADPr (27,43). The affinities of SINV nsP3 MD mutants for Cy5-(ADPr) 18 , were higher for Y114A (34.81  Effects of nsP3 MD mutations on SINV replication in mouse neuronal cells. The nsP3 MD mutations were introduced into clones of the TE strain of SINV (5, 6), full-length viral RNA was transcribed and viruses recovered after transfection of RNA into BHK cells. All viruses could be rescued. To assess the effects of MD mutations on replication efficiency in mouse neuronal cells, NSC34 cells were infected (multiplicity of infection [MOI] of 10) with WT SINV and SINV with each of the nsP3 MD single mutations (N24A, G32A, G32E, G32S, and Y114A) or set of mutations (G32E/I113R/ Y114N; TM) ( Fig. 2). Peak virus production was at 24 h for all viruses ( Fig. 2A). At 12 h, the amount of infectious virus released was similar to WT for G32S, G32A, and N24A, whereas G32E, Y114A, and TM had produced less virus. Infectious virus production by TM was most impaired ( Fig. 2A), and sequencing of virus from 12 and 24 h showed mutation of G32E to G32S and reversion of position 113 from R to I and position 114 from N to Y. Cells infected with G32E showed less cell death compared to cells infected with other viruses at 24 and 48 h (Fig. 2B). Sequencing of viruses recovered at 12 and 24 h revealed that, in addition to the TM changes, G32E had reverted from E to WT G. Thus, altered MD function associated with mutation of G32 to E that eliminates hydrolase activity and ADPr-binding ( Fig. 1) resulted in selection for reversion to WT G or to the less impaired mutant S (Fig. 1) during growth in NSC34 cells. The other mutations (N24A, G32S, G32A, and Y114A) were retained during replication in neural cells.
To determine whether viruses reverted or developed compensatory mutations during in vivo CNS replication, the MDs of recovered viruses were sequenced. As observed during replication in NSC34 cells, G32E reverted to WT G and TM mutated to G32S and reverted to 113I and 114Y, while N24A retained this mutation. Thus, the two viruses with the G32E MD mutation were rapidly selected against during growth in neural cells both in vitro and in vivo. For G32E, reversion to G conferred WT virulence in mice, while selection of G32S in TM, along with reversion of position 113 to I and position 114 to Y led to virulence similar to single mutant G32S (Fig. 3A). Viruses with other MD mutations were less neurovirulent than WT but differed in whether survivors recovered neurologic function. Because G32E and TM reverted, N24A had been studied previously (55), and G32S and G32A had similar phenotypes, subsequent studies focused on comparisons between WT, G32S (decreased binding and hydrolase) and Y114A (increased binding and decreased hydrolase).

Effects of nsP3 MD mutations on CNS virus replication and clearance.
To determine the effect of nsP3 MD mutations on virus replication and clearance from the CNS, brain and spinal cord homogenates were assayed for SINV by plaque assay and for viral RNA by quantitative reverse transcription-PCR (qRT-PCR) (Fig. 4). In brain, all three viruses had similarly high levels of infectious virus 2 days after infection but differed in clearance (Fig. 4A). G32S was cleared more rapidly than WT (P Ͻ 0.05, day 4), while Y114A was cleared more slowly than WT (P Ͻ 0.05, day 6). In the spinal cord, the amount of virus at day 2 was lower for G32S than for the WT and Y114A (P Ͻ 0.0001), and the clearance of infectious virus occurred earlier (day 6). Y114A had titers similar to WT through day 4, but the amounts of infectious virus remained high at day 6 (P Ͻ 0.0001, WT versus Y114A) (Fig. 4B). SINV genomic and subgenomic RNAs were quantified in the brains ( Fig. 4C and E) and spinal cords ( Fig. 4D and F) as determined by qRT-PCR using primers specific for nsP2 (genomic; Fig. 4E and F) and E2 (subgenomic and genomic; Fig. 4C and D). The levels of viral RNA were generally lower for G32S than for the WT in both the brain and spinal cord at all times after infection, while the levels of Y114A were similar or higher than the WT. Therefore, replication of G32S in the CNS was impaired compared to the WT with rapid clearance, whereas the replication of Y114A was similar to the WT, but with impaired clearance.
Effects of nsP3 MD mutations on expression of Parp mRNAs in the CNS. Several PARPs are under diversifying evolutionary pressure, are induced by IFN, and/or have demonstrated antiviral activity, suggesting that MD function may be important for countering the antiviral effects of ADP-ribosylated proteins (46,50,51,64,65). In addition, although Parp mRNAs are not induced by infection of NSC34 cells, PARPs are activated for ADP-ribosylation and facilitate CHIKV replication (37). To determine whether Parp mRNA expression was induced by CNS infection, we measured changes in levels of selected Parp mRNAs in brain and spinal cord by qRT-PCR ( Fig. 5A and B). Expression of Parp1 mRNA changed little, but mRNAs for Parp9, -10, -12, -13, and -14 were increased by infection. At 2 days after infection, the brain levels of Parp9, -10, -12, -13, and -14 mRNAs and the spinal cord levels of Parp14 mRNA were higher in Y114A-infected mice than in either WT or G32S-infected mice. To determine whether PARP14 protein levels were increased after infection, brain and spinal cord homogenates were probed by immunoblotting ( Fig. 5C and D). PARP14 protein levels increased in the CNS after infection with all three viruses with higher levels at 4 days (P Ͻ 0.0001 versus WT/G32S) and lower levels at 6 days (P Ͻ 0.05 versus WT, P Ͻ 0.0001 versus G32S) in the brains of Y114A-infected mice. Therefore, a subset of Parp mRNAs was increased after infection with generally higher levels induced by Y114A than WT or G32S infection.
Effects of nsP3 MD mutations on IFN pathway activation in the CNS. Because Parp9, -10, -12, -13, and -14 are IFN inducible (46,(66)(67)(68)(69)(70), the MD-mediated regulation of innate responses was examined further (Fig. 6). The expression of mRNAs for Rig-I and Mda5, important cytoplasmic sensors of viral RNAs, was analyzed in the brains and spinal cords of infected mice by qRT-PCR (Fig. 6A). Rig-I and Mda5 were induced in all groups with higher levels of both mRNAs at day 2 in the brains of Y114A compared to WT-or G32S-infected mice (P Ͻ 0.0001), but with similar levels in the spinal cord. To compare levels of type I IFNs in the CNS after infection, the levels of alpha interferon (IFN-␣) and IFN-␤ protein in homogenates of the brain and spinal cord were assessed by enzyme immunoassay (EIA) (Fig. 6B). WT virus induced more IFN than either Y114A or G32S (P Ͻ 0.05 versus Y114A, P Ͻ 0.001 versus G32S IFN-␣ day 2 brain; P Ͻ 0.001 versus G32S IFN-␣ day 6 spinal cord; P Ͻ 0.0001 versus G32S IFN-␤ day 2 and P Ͻ 0.0001 versus G32S/Y114A IFN-␤ day 4 spinal cord). Therefore, infection with WT SINV FIG 4 Infectious virus and SINV RNA in brains and spinal cords of mice infected with WT and nsP3 MD mutants. Two-week-old CD-1 mice were inoculated intracranially with 1,000 PFU of SINV WT or nsP3 MD mutants G32S and Y114A. Brain (A) and spinal cord (B) homogenates from four mice from each group at each time point were assayed for infectious virus by plaque assay. RNA extracted from brain and spinal cord tissues was assayed for viral subgenomic and genomic (C and D) and genomic (E and F) RNA by qRT-PCR. Data pooled from two independent experiments are presented as means Ϯ the SD for eight mice for each time point per group. Significance was determined by two-way ANOVA with Tukey's multiple-comparison test. *, P Ͻ 0.05; ****, P Ͻ 0.0001 (WT versus G32S).^, P Ͻ 0.05;^^^^, P Ͻ 0.0001 (WT versus or Y114A). #, P Ͻ 0.05; ##, P Ͻ 0.01; ###, P Ͻ 0.001; ####, P Ͻ 0.0001 (G32S versus Y114A).
Alphavirus Neurovirulence in Mice ® induced more brain IFN, but the expression of Parps, Rig-I, and Mda5 IFN-stimulated genes (ISGs) was highest in response to Y114A infection.
To assess potential differences in IFN signaling, expression and phosphorylation of STAT1, an important cytoplasmic transcription factor activated by Jak kinase-mediated tyrosine phosphorylation in response to IFN (71,72), was examined by immunoblotting brain and spinal cord homogenates (Fig. 7A to D). CNS expression of STAT1 protein was increased in response to infection by all three viruses. Activation of STAT1 (pSTAT1) in FIG 5 Modulation of PARP mRNA and protein expression in the CNS of mice infected with WT and nsP3 MD mutants. Two-week-old CD-1 mice were inoculated intracranially with 1,000 PFU of SINV WT (TE) or nsP3 MD mutants G32S and Y114A. RNA was extracted from brain and spinal cord tissues and the expression of Parp1, Parp9, Parp10, and Parp12 (A) and of Parp13 and Parp14 (B) (upper panels, brain; lower panels, spinal cord) mRNAs were measured by qRT-PCR. C T values were normalized to Gapdh, and the fold change was calculated relative to samples from day 0 (ΔΔC T ). Data pooled from two independent experiments are presented as means Ϯ the SD for eight mice per group. Significance was determined by two-way ANOVA with Tukey's multiple-comparison test. *, P Ͻ 0.05; **, P Ͻ 0.01; ***, P Ͻ 0.001 (WT versus G32S).^, P Ͻ 0.05;^^, P Ͻ 0.01;^^^^, P Ͻ 0.0001 (WT versus Y114A). #, P Ͻ 0.05; ##, P Ͻ 0.01; ###, P Ͻ 0.001; ####, P Ͻ 0.0001 (G32S versus Y114A). (C) Immunoblots of brain and spinal cord homogenates (20 g of 10% [wt/vol]) probed for PARP14. Antibody against ␤-actin was used for loading controls. The levels of PARP14 (170-kDa band) relative to actin in the brain (upper panel) and spinal cord (lower panel) were determined using densitometry from five blots for brain and four blots for spinal cord and presented as a bar graph. Significance was determined by 2-way ANOVA with Tukey's multiple-comparison test.^, P Ͻ 0.05;^^^^, P Ͻ 0.0001 (WT versus Y114A). ####, P Ͻ 0.0001 (G32S versus Y114A). (D) Representative immunoblot images of brain (upper) and spinal cord (lower) homogenates probed for PARP14 and actin.  were probed for total and phosphorylated STAT1. Antibody against ␤-actin was used for loading controls. The levels of pSTAT1 (Y701) and STAT1 (91-ϩ 84-kDa band) relative to the actin in brain (A) and spinal cord (C) were quantitated using densitometry from five blots and are presented as a bar graph. Significance was determined by two-way ANOVA with Tukey's multiple-comparison test. *, P Ͻ 0.05 (WT versus G32S). ##, P Ͻ 0.01 (G32S versus Y114A). Representative immunoblot images of brain (B) and spinal cord (D) homogenates probed for pSTAT1, STAT1, and actin. (E) RNA extracted from brain and spinal cord tissues assayed for mRNA expression of Ifit1, Ifit2, and Isg15 by qRT-PCR. C T values were normalized to Gapdh, and the fold change was calculated relative to infected controls at day 0 (ΔΔC T ). Data pooled from two independent experiments are presented as means Ϯ the SD for eight mice for each time point per group. Significance was determined by two-way ANOVA with Tukey's multiple-comparison test. *, P Ͻ 0.05; **, P Ͻ 0.01; ***, P Ͻ 0.001; ****, P Ͻ 0.0001 (WT versus G32S).^, P Ͻ 0.05;^^^, P Ͻ 0.001;^^^^, P Ͻ 0.0001 (WT versus Y114A). #, P Ͻ 0.05; ##, P Ͻ 0.01; ###, P Ͻ 0.001; ####, P Ͻ 0.0001 (G32S versus Y114A). brain was evident by 2 days after infection, with lower levels for G32S-infected mice (day 2 P Ͻ 0.05 versus WT, day 6 P Ͻ 0.01 versus Y114A) (Fig. 7A to D), but no differences in the spinal cord were detected.
To further assess induction of ISGs, changes in levels of mRNAs for IFN-induced protein with tetratricopeptide repeat 1 (Ifit1), Ifit2, and Isg15 were measured (Fig. 7E). At 2 days after infection, the levels of Ifit1 mRNA were higher in the brains (P Ͻ 0.0001 versus WT/G32S) and spinal cords (P Ͻ 0.001 versus WT, P Ͻ 0.0001 versus G32S) of mice infected with Y114A. The level of Ifit2 mRNA was also higher in the brains of Y114A-infected mice at 2 days (P Ͻ 0.05 versus WT, P Ͻ 0.0001 versus G32S) and remained at this level at 6 days in both brains (P Ͻ 0.05 versus WT/G32S) and spinal cords (P Ͻ 0.0001 versus WT/G32S). Isg15 mRNA was higher in brains of WT-and Y114A-infected mice than G32S-infected mice throughout the infection and in the spinal cord at 4 days (P Ͻ 0.0001). Therefore, although IFN levels in the brain were highest with WT infection (Fig. 6B), Y114A generally induced a more vigorous and sustained ISG response than G32S or WT virus infection (Fig. 5, 6A, and 7), perhaps associated with the generally higher levels and slower clearance of viral RNA (Fig. 4).

Effects of nsP3 MD mutations on NF-B pathway activation in the CNS.
Because NF-B pathway activation is an important component of the innate response to viral infection that occurs in response to signaling through Toll-like receptors (TLRs), we assessed the changes in the expression of the mRNAs for endosomal TLRs 3, 7, 8, and 9 that were most relevant for responses to viral infection (Fig. 8A). All endosomal Tlr mRNAs were increased after infection. Tlr3 mRNA was most rapidly induced (day 2) and more highly expressed in the brains (P Ͻ 0.01 versus WT, P Ͻ 0.001 versus G32S) and spinal cords (P Ͻ 0.05) of mice infected with Y114A than in mice infected with WT or G32S. At 6 days after infection, the levels of Tlr7 and Tlr8 in both the brain (P Ͻ 0.0001) and the spinal cord (P Ͻ 0.01) were higher in WT-infected than in Y114A-or G32Sinfected mice.
The levels of mRNAs for several cytokines and chemokines dependent on NF-B pathway signaling were examined ( Fig. 8B and D). Il1␤, Tnf, Il6, Ccl2, Ccl5, and Cxcl10 mRNAs were all increased in the CNS after infection. In brain, levels of all innate response gene mRNAs were generally highest at 2 days in Y114A-infected mice and lowest in G32S-infected mice. In the spinal cord, WT-infected mice had higher levels of Il6, Tnf, Ccl2, and Ccl5 mRNAs at 2 or 4 days after infection, while Il1␤ was highest in Y114A at 2 and 6 days (P Ͻ 0.01 versus G32S, P Ͻ 0.0001 versus Y114A). The levels of mRNAs for chemokines Ccl2, Ccl5, and Cxcl10 remained higher at 6 days after infection in spinal cords of Y114A-infected than WT-infected mice (P Ͻ 0.0001) ( Fig. 8B and D). The levels of tumor necrosis factor alpha (TNF-␣) protein in the CNS also increased after infection and were higher in WT-infected mice at 2 days postinfection in the brain (P Ͻ 0.05 versus Y114A, P Ͻ 0.001 versus G32S) and remained high in the spinal cords of Y114A-infected mice at 8 days (P Ͻ 0.0001) (Fig. 8C). Therefore, Y114A-infected mice tended to have higher early levels of these cytokine and chemokine mRNAs in brain than WT mice and persistent expression in spinal cord, whereas the responses of G32S-infected mice were generally lower than WT-infected mice.

Effects of nsP3 MD mutations on adaptive immune responses in the CNS.
To determine the effects of MD mutations on induction of the adaptive immune response, we measured the levels of SINV-specific antibodies in the serum (Fig. 9A), brains (Fig. 9B) and spinal cords (Fig. 9C) of mice infected with WT, G32S, and Y114A viruses by EIA. Serum levels of SINV-specific IgM were lower for Y114A (P Ͻ 0.01 days 4 and 6) than G32S and WT (Fig. 9A), whereas levels of IgG were lower in mice infected with G32S (P Ͻ 0.01, days 4 and 6) (Fig. 9A). Because the antibody levels in the CNS reflect entry of antibody-secreting cells into the brain and spinal cord and are most important for virus clearance and recovery (73,74), we also measured SINV-specific antibodies in CNS tissue homogenates. In both the brain (Fig. 9B) and the spinal cord (Fig. 9C) the IgM and IgG levels were similar. These data suggest that MD mutations had little effect on induction of the antibody response to SINV.  Because CD4 and CD8 T cells also play an important role in adaptive immune responses, we assessed changes in the CNS levels of Cd4, Cd8a, and Cd8b mRNAs after infection as indicators of T cell infiltration and retention (Fig. 10A). Expression of Cd4 mRNA steadily increased, while Cd8a and Cd8b increased for 6 days and then decreased at 8 days in surviving G32S-and Y114A-infected mice. The levels of Cd4, Cd8a, and Cd8b mRNAs in the brain were similar in the three groups, but in the spinal cord at 6 days the mRNA levels of Cd4 (P Ͻ 0.0001), CD8a (P Ͻ 0.0001), and CD8b (P Ͻ 0.001) were lowest in WT-infected mice (Fig. 10A).
Because the primary T cell effector of SINV clearance is IFN-␥ with interleukin-10 (IL-10) as an important regulator (74)(75)(76), these cytokine mRNAs were measured as additional indicators of T cell responses in the CNS (Fig. 10B). Ifn␥ mRNA was highly induced in all animals and, at day 6, the level was lower in the brains and spinal cords of mice infected with WT virus than in G32S (P Ͻ 0.001, brain; P Ͻ 0.0001 spinal cord)or Y114A (P Ͻ 0.05, brain; P Ͻ 0.0001, spinal cord)-infected mice (Fig. 10B). The expression of IFN-␥-induced ISGs guanylate binding protein 1 (Gbp1) and Gbp2 mRNAs also increased in the brains and spinal cords of all mice. In the spinal cord the levels of Gbp1 and Gbp2 mRNAs at day 6 were lowest for WT-infected mice (Gbp1, P Ͻ 0.0001; Gbp2, P Ͻ 0.001). The mRNA for regulatory cytokine Il10 increased after infection, with the highest levels in the brains of WT-infected mice at 6 days (P Ͻ 0.0001) and in spinal cords at 4 days (P Ͻ 0.0001) (Fig. 10B). These data suggest that SINV with MD mutations induce a more vigorous IFN-␥-producing T cell response to infection than WT SINV, particularly in the spinal cord.

DISCUSSION
Viral nonstructural protein MDs possess both MAR hydrolase and ADPr-binding activities (43)(44)(45) and are important determinants of neurovirulence, but how pathogenesis of alphavirus encephalomyelitis is affected by alterations in the ADPr-binding and hydrolase functions of the nsP3 MD has not been evaluated. Previous studies of CHIKV nsP3 MD mutants showed that these activities are critical for different aspects of alphavirus replication in neuronal cells (37,43,55). If both functions are impaired, initiation of infection is inefficient, and little if any virus is produced. If there is better binding but diminished hydrolase activity, replication is initiated more quickly, but the amplification of replication complexes is impaired (37). To determine how altered MD functions affect the pathogenesis of encephalomyelitis, we introduced mutations similar to those studied in CHIKV into SINV, a better model system for analysis of CNS alphavirus infection in mice, and showed that the effects on ADPr-binding and hydrolase activities were similar to those observed for CHIKV.
In vitro replication of SINV in neural cells and in vivo replication in the CNS of mice were severely impaired by mutations that eliminate ADPr-binding and hydrolase activities (G32E) with reversion to WT (G) or selection of a less compromising change (S) during replication. SINVs with MDs deficient in both binding and hydrolase activities (G32S and G32A) or with hydrolase deficiency combined with better binding (Y114A) were less virulent than WT virus in mice but displayed different phenotypes. G32S replicated less well in both the brain and the spinal cord, induced similar innate from brain and spinal cord tissues and mRNA expression of Tlr3, Tlr7, Tlr8, and Tlr9 (A) and Il6, Il1␤, and Tnf (B) was measured by qRT-PCR (upper panels, brain; lower panels, spinal cord). The C T values were normalized to Gapdh, and the fold change was calculated relative to day 0 (ΔΔC T ). Data pooled from two independent experiments are presented as means Ϯ the SD for eight mice for each time point per group. Significance was determined by two-way ANOVA with Tukey's multiple-comparison test. (C) Brain (upper panel) and spinal cord (lower panel) homogenates were tested by EIA for TNF-␣. Graphs show the average concentration of TNF-␣ in pg/g tissue from four animals per group. The dotted line indicates the lowest assay range value in pg/ml. Significance was determined by two-way ANOVA with Tukey's multiple-comparison test. (D) Expression of Ccl2, Ccl5, and Cxcl10 mRNAs (upper panels, brain; lower panels, spinal cord) was measured by qRT-PCR. The C T values were normalized to Gapdh, and the fold change was calculated relative to day 0 (ΔΔC T ). Data pooled from two independent experiments are presented as means Ϯ the SD for eight mice per group for each time point. Significance was determined by two-way ANOVA with Tukey's multiple-comparison test. *, P Ͻ 0.05; **, P Ͻ 0.01; ***, P Ͻ 0.001; ****, P Ͻ 0.0001 (WT versus G32S).^, P Ͻ 0.05;^^, P Ͻ 0.01;^^^, P Ͻ 0.001;^^^^, P Ͻ 0.0001 (WT versus Y114A). #, P Ͻ 0.05; ##, P Ͻ 0.01; ###, P Ͻ 0.001; ####, P Ͻ 0.0001 (G32S versus Y114A).
Alphavirus Neurovirulence in Mice ® immune responses, and caused less severe disease, with 71% survival and full recovery of survivors. Y114A replicated better than WT and induced higher expression of IFN-stimulated and NF-B-induced genes with similar antibody responses but was cleared more slowly from the spinal cord, with 40% survival and persistent neurologic deficits in survivors. Therefore, MD function was important for neural cell replication both in vitro and in vivo and determined the outcome from alphavirus encephalomyelitis in mice.
Alanine substitution at residue 10 eliminates alphavirus MD ADPr-binding and is not tolerated by CHIKV (D10) or by SINV (N10), and D is present at this position in the encephalitic New World alphaviruses VEEV, EEEV, and WEEV (43,55). Mutation of this critical binding residue leads to a failure to synthesize viral RNA in neural cells, so that any virus recovered had either reverted to D, T, or N or acquired a compensatory mutation at position 31 (E to G) (37,55). The hydrolase loop formed by residues 24 to 33 also participates in ADPr-binding (necessary for enzyme function), but enzymatic activity per se is less critical for initiating alphavirus replication (37). The amino acid at position 24 (N) is conserved in all viral MDs examined (50,51), and mutation of N24 to FIG 9 Antibody responses of mice infected with WT and nsP3 MD mutants. Two-week-old CD-1 mice were inoculated intracranially with 1,000 PFU of SINV TE and nsP3 MD mutants G32S and Y114A. Serum (A), brain (B), and spinal cord (C) homogenates were tested for SINV-specific IgM (left panels) and IgG (right panels) by EIA. The graphs show the OD of serum (1:100 dilution) and 10% (wt/vol) brain and spinal cord homogenates (1:2 dilution) from four animals per group. Significance was determined by multiple Student t tests using the Holm-Sidak method. *, P Ͻ 0.05; **, P Ͻ 0.01 (WT versus G32S).^, P Ͻ 0.05;^^, P Ͻ 0.01 (WT versus G32S versus Y114A). ##, P Ͻ 0.01 (G32S versus Y114A).
A eliminated hydrolase activity (Fig. 1) and impairs shutoff host protein synthesis in NIH 3T3 cells (56) but did not affect replication in NSC34 cells ( Fig. 2A). However, mortality in 2-week-old mice was reduced from 100 to 84% in the present study (Fig. 3A) and to 40% in our previous study (55) and is associated with more rapid virus clearance. The reasons for differences in mortality between the two studies are not known but may be due to differences in genetic background or microbiome composition of outbred CD-1 mice. The phenotype was similar to that observed with mutation at the comparable position of the SARS-CoV MD (N41A/N1040A) that does not affect replication in Vero or FIG 10 Cellular immune response genes expressed by mice infected with WT and nsP3 MD mutants. Expression of Ifn␥, Il10, and IFN-␥-induced ISG mRNAs in the CNS of mice infected with WT and nsP3 MD mutants. Two-week-old CD-1 mice were inoculated intracranially with 1,000 PFU of SINV TE or nsP3 MD mutants G32S and Y114A. RNA was extracted from brain and spinal cord tissues, and the mRNAs for Cd4, Cd8a, and Cd8b (A) and Ifn␥, Il10, Gbp1, and Gbp2 (B) were measured by qRT-PCR (upper panels, brain; lower panels, spinal cord). The C T values were normalized to Gapdh, and the fold change was calculated relative to day 0 (ΔΔC T ). Data pooled from two independent experiments are presented as means Ϯ the SD for eight mice for each time point per group. Significance was determined by two-way ANOVA with Tukey's multiple-comparison test. *, P Ͻ 0.05; **, P Ͻ 0.01; ***, P Ͻ 0.001; ****, P Ͻ 0.0001 (WT versus G32S).^, P Ͻ 0.05;^^, P Ͻ 0.01;^^^, P Ͻ 0.001;^^^^, P Ͻ 0.0001 (WT versus Y114A). #, P Ͻ 0.05; ##, P Ͻ 0.01; ####, P Ͻ 0.0001 (G32S versus Y114A).
Alphavirus Neurovirulence in Mice ® Calu-3 2B4 epithelial cells but decreases lung virus replication and improves survival of mice after respiratory infection. These results were postulated to be linked to MD regulation of innate response genes with higher lung levels of IFN, ISG, IL-6 and TNF mRNAs at 24 h, but not 72 h, after infection (40). The specific mechanisms for these effects have yet to be identified.
The amino acid at position 32 is a critical determinant of binding, as well as enzymatic, activity, and G is highly conserved for all viral MDs, including the encephalitic alphaviruses VEEV, WEEV, and EEEV (51). E at this position eliminates both binding and hydrolase activities (Fig. 1) (43), and SINV G32E MD mutants reverted during replication in NSC34 cells and in the nervous systems of mice. Mutation of G32 to S or A partially preserved ADPr-binding, with hydrolase activity decreased 26.5% for G32S and 28% for G32A (Fig. 1). These viruses replicated less well in NSC34 cells than did the WT (Fig. 2) and induced less severe neurologic disease with reduced mortality in mice (Fig. 3). Replication of G32S in the CNS was lower than that of the WT, with the most striking difference in the spinal cord (Fig. 4). We postulate that decreased virus production is due to less efficient initiation and amplification of replication in neurons, as observed in vitro for CHIKV G32S (37), resulting in decreased spread of virus to the spinal cord. Therefore, efficient replication complex formation and amplification in neurons likely requires binding to one or more yet to be identified ADP-ribosylated proteins.
The affinity of MD binding to ADPr is determined in part by the amino acid at position 114 in the ADPr-binding pocket (50). The amino acid at 114 is more variable in alphaviruses than the other residues studied with a Y in EEEV and WEEV, as well as SINV and CHIKV, but F in VEEV. The Y114A substitution in SINV increases binding while decreasing hydrolase activity ( Fig. 1) (43) and thus provides an opportunity to independently examine the role of MD hydrolase activity in pathogenesis. As for CHIKV (37,43), SINV Y114A grew more slowly in NSC34 cells (Fig. 2) and was less virulent in mice (Fig. 3). However, Y114A replicated as well or better than WT SINV in the CNS, was cleared more slowly from both brain and spinal cord, and resulted in permanent neurologic damage in survivors (Fig. 4). Initial control of SINV replication in the CNS is dependent on the innate immune response, particularly local production of type I IFN (77)(78)(79). Because viral MDs can influence both the induction of and the response to innate immune effectors (40,65,80,81), we assessed multiple parameters of the immune response to SINV in the CNSs of infected mice. Mice infected with Y114A in general had similar CNS levels of IFN-␣ and IFN-␤ and greater induction of ISGs Parp9, -10, -12, -13, and -14; Ifit1 and -2; and Isg15 ( Fig. 5 and 7) and innate response genes Tlr3, Il1␤, Il6, Tnf, Ccl2, Ccl5, and Cxcl10 (Fig. 8) compared to WT-infected mice. Because these same parameters tended to be lower in G32S-infected mice where viral RNA loads were also lower than WT, we postulate that MD-determined levels of virus replication in the CNS have a more important role in regulating innate responses than direct MD regulation of these responses.
Surviving Y114A-infected animals had persistent hind limb paralysis associated with delayed virus clearance from the spinal cord ( Fig. 3B and 4D). SINV clearance is mediated by the adaptive immune response with the combined effects of antibody and IFN-␥ (74,75,82,83). Although production of serum IgM was decreased in Y114Ainfected animals compared to WT and G32S-infected mice, levels of SINV-specific IgM and IgG, as well as Ifn␥ mRNA, in the brains and spinal cords were similar or higher ( Fig. 9 and 10). Therefore, defects in clearance could not be ascribed to failure to induce an adaptive immune response. Because the mechanisms of antibody-mediated and IFN-␥-mediated clearance of virus from neurons are not known and MD hydrolase activity affects amplification of replication complexes, shutoff of host protein synthesis and translation of structural proteins (37), it is possible that MD interactions with intracellular mechanisms of virus clearance are affected by the mutations studied.
In summary, these studies have shown that MD function is an important determinant of alphavirus replication in the CNS, manifestations of encephalomyelitis, virus clearance, and recovery.

MATERIALS AND METHODS
developed using an Amersham ECL Plus Western blot developing kit (GE Healthcare), and densitometric analysis was carried out using ImageJ software.
Statistical analysis. Survival was assessed using Kaplan-Meier curves and a log-rank (Mantel-Cox) test. Time course studies were analyzed using two-way analysis of variance (ANOVA) with Tukey's multiple-comparison posttest to compare the groups infected with TE, G32S, and Y114A 0 to 6 days after infection. Two-way ANOVA with Bonferroni's posttest was used to compare surviving G32S and Y114A groups at 8 days. Differences in a single group were determined using one-way ANOVA with Dunnett's multiple-comparison test. Differences between groups at a single time point were determined using an unpaired two-tailed Student t test with a 95% confidence interval. The results are expressed as means Ϯ the SD. Statistical analyses were conducted using Prism 8 (GraphPad).