Glu333 in rabies virus glycoprotein is involved in virus attenuation through astrocyte infection and interferon responses

Summary The amino acid residue at position 333 of the rabies virus (RABV) glycoprotein (G333) is a major determinant of RABV pathogenicity. Virulent RABV strains possess Arg333, whereas the attenuated strain HEP-Flury (HEP) possesses Glu333. To investigate the potential attenuation mechanism dependent on a single amino acid at G333, comparative analysis was performed between HEP and HEP333R mutant with Arg333. We examined their respective tropism for astrocytes and the subsequent immune responses in astrocytes. Virus replication and subsequent interferon (IFN) responses in astrocytes infected with HEP were increased compared with HEP333R both in vitro and in vivo. Furthermore, involvement of IFN in the avirulency of HEP was demonstrated in IFN-receptor knockout mice. These results indicate that Glu333 contributes to RABV attenuation by determining the ability of the virus to infect astrocytes and stimulate subsequent IFN responses.


INTRODUCTION
Rabies virus (RABV) is the causative agent of rabies, a fatal neurological disease in mammals causing at least 59,000 human deaths annually, particularly in Asian and African countries. Because of its global impact, rabies is notifiable to the World Health Organization (WHO) and International Epizootic Office (OIE) (Hampson et al., 2015;Singh et al., 2017). The RABV virion carries a negative-sense single-stranded RNA genome possessing five open reading frames encoding the nucleoprotein (N), phosphoprotein (P), matrix protein (M), glycoprotein (G), and large protein (L) (Finke and Conzelmann, 2005).
Highly pathogenic strains of RABV often exhibit characteristics related to their strict neurotropism (Ito et al., 2001;Morimoto et al., 1999Morimoto et al., , 2000. In contrast, attenuated strains often exhibit a broader cell tropism not specific to neuronal cells in vitro and show limited ability to spread to the central nervous system in vivo (Takayama- Ito et al., 2006;Tao et al., 2010). Among the viral proteins, the G protein exists as a trimer on the surface of the RABV virion. Because G protein is responsible for host cell receptor recognition (Sasaki et al., 2018;Thoulouze et al., 1998;Tuffereau et al., 1998;Wang et al., 2018) and membrane fusion (Gaudin et al., 1991), mutations in the G protein often alter viral pathogenicity. In particular, Arg at position 333 in the G protein (G333) contributes to virulence in some fixed strains of RABV in adult mice (Dietzschold et al., 1985;Ito et al., 2021;Morimoto et al., 2001;Nakagawa et al., 2012;Shuai et al., 2015). To understand this property, several studies have attempted to characterize the role of the amino acid at G333. An amino acid substitution for Glu 333 enhanced viral-induced apoptosis in infected cells, leading to a loss of pathogenicity in mice (Tao et al., 2010). Infection with RABV encoding dual G proteins with Glu 333 also resulted in enhanced apoptosis in cells (Faber et al., 2002). One study has reported that amino acid at G333 influences the binding affinity of G protein to one of the receptors for RABV, p75NTR, suggesting that amino acid change at G333 may affect the cell tropism of RABV (Tuffereau et al., 1998). Collectively, it has been clearly demonstrated that replacement of Arg 333 or Lys 333 to other amino acids causes a pathogenic shift of RABVs to an avirulent phenotype, whereas the mechanisms regulated by the amino acid residue at G333 remain unclear and controversial.
However, involvement of the amino acid residue at G333 in astrocyte infection is yet to be established. Recently, astrocytes abortively infected with diverse neurotropic viruses, including RABV, Theiler's murine encephalomyelitis virus, and vesicular stomatitis virus, have been reported to be the main source of interferon (IFN)-b production in the brain conferring antiviral protection (Pfefferkorn et al., 2016). In addition, other studies showed that type-I IFN signaling in astrocytes is important to build an antiviral state in a virus-infected brain (Detje et al., 2015;Hwang and Bergmann, 2018;Kallfass et al., 2012;Pfefferkorn et al., 2016;Tian et al., 2018). RABVs are sensitive to IFN, and the importance of IFN in controlling RABVs has long been proposed (Postic and Fenje, 1971;Weinmann et al., 1979).
To further understand the attenuation mechanism dependent on the amino acid residue at G333 on aspects of viral infection and IFN responses in astrocytes, we investigated the tropism for astrocytes using a recombinant HEP-Flury strain (rHEP; Glu 333 ) and a single amino acid mutant HEP 333 R strain (Arg 333 ) in vitro. Infection of astrocytes and IFN responses were also examined in vivo. Finally, we examined the pathogenicity of rHEP in mice deficient in IFN signaling pathways, which might play a significant role in G333-dependent attenuation. Understanding the mechanism of RABV attenuation is essential in considering the use of live attenuated vaccines to control rabies in wild animals. Therefore, our present study has yielded new insights into the pathogenicity of RABV associated with G333.

RESULTS
Impact of amino acid substitution at position 333 of the G protein on rRABV growth in neuron-and astrocyte-derived cell lines To investigate the role of the amino acid at position 333 of RABV G protein (G333) on RABV infection, we generated recombinant RABV clones (rRABV) that included rHEP carrying Glu 333 and rHEP 333 R carrying Arg 333 in G protein by reverse genetics methods. First, we evaluated the replication efficacy between rHEP and rHEP 333 R in a neuron-derived cell line, NA cells, and an astrocyte-derived cell line, SVG-A cells. rHEP and rHEP 333 R showed similar growth kinetics in NA cells ( Figure 1A). By contrast, in SVG-A cells, rHEP A B Figure 1. Infectivity of rabies virus (RABV) in neuron-derived NA cells and astrocyte-derived SVG-A cells Monolayers of NA or SVG-A cells were inoculated with rHEP or rHEP 333 R at a multiplicity of infection (MOI) of 1. (A) Viral growth curve. Supernatants were collected at the indicated time points, and virus titers were measured by a focus forming assay. Means G standard deviations of triplicate data from a representative experiment are shown in the graph. A multiple t test was performed by the Holm-Sidak method for statistical analysis. *p < 0.05, **p < 0.01. (B) Images of RABV-infected cells. Cells were fixed at 24 h postinfection (hpi) and stained with FITC-conjugated anti-RABV N antibody for RABV N (green) and Hoechst 33342 for the nucleus (blue). The representative images were captured by fluorescent microscopic analysis. Scale bar, 50 mm.  Figure 1A). These results were reflected in the population of infected cells stained by FITC-labeled anti-RABV N antibody. In short, rHEP and rHEP 333 R exhibited similar infectivity to neuron-derived NA cells, whereas rHEP 333 R showed limited growth in astrocyte-derived SVG-A cells ( Figure 1B). These results demonstrated that the amino acid residue at G333 influences the cell tropism of RABV.

IFN production in primary astrocytes infected with rRABVs
Astrocytes act as IFN producers in brain tissue infected with neurotropic viruses (Detje et al., 2015;Hwang and Bergmann, 2018;Kallfass et al., 2012;Pfefferkorn et al., 2016;Tian et al., 2018), and thus we assessed the relationship between RABV infection and IFN responses in astrocytes. Primary astrocytes were used in these experiments because the astrocyte-derived cell line SVG-A lacks an IFN response against rRABV infection ( Figure S1). In line with the results of rRABV growth in SVG-A cells (Figure 1), mouse-derived primary astrocytes were susceptible to rHEP infection but less susceptible to rHEP 333 R infection ( Figure 2A). To confirm whether rRABV infection triggers IFN production in astrocytes, the mRNA levels of IFN-b in rRABV-infected primary astrocytes were quantified by qRT-PCR. The IFN-b gene expression levels in primary astrocytes were significantly higher in rHEP-infected cells than in rHEP 333 R-infected cells ( Figure 2B). These results suggested that rHEP can infect astrocytes more efficiently and induce subsequent IFN production as compared with rHEP 333 R.

rRABV tropism for astrocytes and the IFN responses in vivo
To further characterize the role of the amino acid at G333 with respect to astrocyte infection and IFN induction, we isolated astrocytes from the brains of immunocompetent S129 mice at 5 days after intracranial inoculation of rHEP or rHEP 333 R. To estimate the incidence of rRABV infection in astrocytes in the whole brain, we quantified viral RNA levels in astrocytes sorted from the rRABV-infected brain, and the data were described relative to viral replication in the whole brain. In rHEP-infected mice, viral RNA levels in astrocytes relative to the whole brain were significantly higher than these in rHEP 333 R-infected mice (Figure 3A), suggesting preferential infection of astrocytes by rHEP in vivo. Similarly, and consistent with the viral RNA levels, the gene expression levels of IFN-b were higher in astrocytes derived from rHEP-infected mice compared with those from rHEP 333 R-infected mice ( Figure 3B). These results indicated that rHEP has tropism for astrocytes and induces high levels of IFN production in astrocytes in vivo.
Comparison of infectivity and pathogenicity between rHEP and rHEP 333 R in IFN-receptor knockout (KO) mice rHEP is an attenuated RABV strain that causes transient infection in immunocompetent mice (Takayama-Ito et al., 2006). To evaluate the protective efficacy of IFN in mice infected with rHEP, we next examined the pathogenicity of rRABV in AG129 mice: type-I/II IFN receptor-knockout (KO) mice in an S129 background, which are A B Figure 2. Infectivity of RABV in mouse-derived primary astrocytes (A and B) Astrocyte-derived primary cells were cultured on 24-well plates to 80% confluency and infected with rHEP or rHEP 333 R at an MOI of 1. Supernatants and RNAs were collected at 48 hpi and subjected to virus titration by a focus forming assay (A) and qRT-PCR for the measurement of IFN-b gene expression (B), respectively. Expression levels of the IFN-b gene were normalized to the b-actin gene and presented as fold changes relative to the mock controls using the DDCt method. The values in the graph show the means G standard deviations of a representative experiment. Statistical analysis was performed by the Student's t test (*p < 0.05, **p < 0.01).

OPEN ACCESS
iScience 25, 104122, April 15, 2022 3 iScience Article highly susceptible to infection by various viruses (Aliota et al., 2016;Milligan et al., 2017;Tan et al., 2010). AG129 and S129 mice were infected with rRABVs intracranially to evaluate virus replication and host IFN responses in the central nervous system. Consistent with a previous report (Takayama -Ito et al., 2006), rHEP caused a transient decrease in body weight without any neurological symptoms in immunocompetent S129 mice ( Figure 4A); however, rHEP 333 R showed body weight loss and neuropathogenicity at 6 days postinfection (dpi), resulting in 100% mortality by 8 dpi ( Figure 4B). In contrast, in AG129 mice, rHEP as well as rHEP 333 R caused significant body weight loss and fatal neurological manifestations at 4-5 dpi, resulting in a 100% fatality rate at 6 dpi (Figures 4A and 4B). Furthermore, virus titration of the whole brain tissue at 5 dpi showed a high virus titer in S129 mice infected with rHEP 333 R compared with rHEP ( Figure 4C). Conversely, there was no remarkable difference in virus titer in AG129 mice infected with rHEP or rHEP 333 R ( Figure 4C). The virus titers in AG129 mice infected with rHEP and rHEP 333 R were approximately 10 4 or 10 2 times higher than in S129 mice infected with rHEP and rHEP 333 R, respectively ( Figure 4C). These results indicated that attenuated rHEP exhibits similar growth and virulence to the authentic pathogenic RABV in AG129 mice, intimating that the IFN responses may be related to the attenuation of rHEP in immunocompetent mice. Consistent with our observations on virus titration ( Figure 4C), no significant difference in IFN gene expression levels was found between the two viruses in AG129 mice (Figure 4D). Interestingly, no significant difference was observed in expression levels of IFN-b gene or IFN-stimulated genes (ISGs) in S129 mice infected with rHEP or rHEP 333 R ( Figures 4D-4F), despite the virus titer being markedly higher in rHEP 333 R-infected S129 mice ( Figure 4C). Considering the relatively low virus titer and high IFN gene expression level of rHEP in S129 mice, rHEP may have induced IFN-b gene or ISGs expression more strongly than rHEP 333 R in S129 mice.
Immunohistochemistry of brain sections from rRABV-infected mice showed extensive viral infection in the neurons of AG129 mice infected with either rHEP or rHEP 333 R ( Figures 4G and S3A). Notably, rHEP effectively infected astrocytes as well as neurons in AG129 mice compared with rHEP 333 R ( Figures 4G, 4I, and S3B). In S129 mice infected with rHEP, a limited number of neurons and astrocytes were found to be positive for RABV N proteins ( Figures 4H, S3C, and S3D). Even though rHEP 333 R showed RABV signals distributed across the whole brain in S129 mice, RABV-infected astrocytes were rarely observed in S129 mice ( Figures 4H, S3C, and S3D). These results strengthen the hypothesis that astrocytes are highly susceptible to infection with rHEP but not rHEP 333 R in vivo.

IFN induction and the sensitivity of rRABVs in neuronal cells
Finally, to further investigate the relationship between the IFN responses and pathogenicity of the rRABVs, we compared IFN inducibility and the sensitivity of rRABVs under an IFN-mediated antiviral state using a human neuroblastoma cell line SYM-I, in which IFN signaling pathways are validated (Honda et al., 1984). SYM-I cells were appropriate to evaluate the ability of IFN induction and the sensitivity of rRABVs because the growth kinetics of rRABVs were comparable in this cell line ( Figure 5A). We then analyzed A B Figure 3. Infectivity of RABV and IFN expression in astrocytes in the brains of mice infected with rRABVs S129 mice were inoculated intracranially with 10 4 focus forming units (ffu) of rHEP or rHEP 333 R. qRT-PCR was performed using homogenate of the brain tissue and isolated astrocytes at 5 days postinfection (dpi). (A) Relative RABV N mRNA level in astrocytes. The data were normalized to the b-actin gene and presented as fold changes relative to the whole brain using the DDCt method.
(B) Relative IFN-b mRNA expression in astrocytes. The data were normalized to the b-actin gene and presented as fold changes relative to the whole brain using the DDCt method. All values in the bar graph show the means G standard deviations of three mice from a representative experiment. Statistical analysis was performed by the Student's t test (*p < 0.05, **p < 0.01). iScience Article IFN-b responses against rRABV infection and found that IFN-b production was consistent between the two rRABVs both at the mRNA ( Figure 5B) and the protein level ( Figure 5C) in culture supernatants. These results indicated that there is no difference between rHEP and rHEP 333 R in terms of IFN-b inducibility in response to virus infection.
Next, the sensitivity of rRABVs to the antiviral state induced by IFN was examined. An antiviral state was induced in SYM-I cells by pretreatment with recombinant IFN-b before rRABV infection. Under these conditions, the proliferations of both rHEP and rHEP 333 R were inhibited in SYM-I cells ( Figure 5D). By contrast, when the IFN signaling pathway was blocked by treatment of the cells with anti-IFN receptor neutralizing antibody, the progeny virus titer was increased at the same rate for rHEP and rHEP 333 R relative to the control infection ( Figure 5E). These results suggested that rHEP and rHEP 333 R have similar potential to induce IFN and similar sensitivity to the subsequent antiviral state induced by IFN.

DISCUSSION
G protein plays major roles in RABV pathogenicity in terms of cell attachment, apoptosis induction, and stimulation of the host immune responses (Faber et al., 2002;Morimoto et al., 1999;Pré haud et al., 2003;Sarmento et al., 2005;Sasaki et al., 2018;Seif et al., 1985). In particular, an amino acid residue at position 333 in the G protein (G333) has been shown to be a major determinant of RABV pathogenicity and attenuation (Dietzschold et al., 1985;Tuffereau et al., 1989). In this study, to further understand the mechanism underlying the attenuation of RABV dependent on the amino acid residue at G333, we specifically focused on RABV tropism for astrocytes using an RABV nonpathogenic strain (rHEP) carrying Glu 333 and a pathogenic mutant (rHEP 333 R) carrying Arg 333 in G protein. We herein demonstrated that astrocytes are susceptible to infection with rHEP but not rHEP 333 R both in vitro and in vivo. These results indicated that the infectivity of the attenuated strain HEP in astrocytes is dependent on a single amino acid at G333 and that Glu 333 has a higher affinity for astrocytes than Arg 333 .
Previous studies have reported that avirulent RABV strains including rHEP carrying Glu 333 in G protein exhibit a broader cell tropism not specific to neuronal cells (Takayama- Ito et al., 2006;Tao et al., 2010) compared with pathogenic strains carrying Arg 333 or Lys 333 . Our findings indicated that replacement of Arg 333 to Glu 333 increased astrocyte tropism of HEP strain. Even though the position G333 is not located at any of reported receptor binding sites on the G protein (Langevin et al., 2002;Lentz et al., 1982Lentz et al., , 1984, replacement of Arg 333 with Glu 333 is known to greatly reduce the binding affinity of G protein to one of the RABV receptors, p75NTR (Tuffereau et al., 1998). Accordingly, the amino acid residue at G333 might influence the conformation of the receptor binding sites of G protein, possibly altering the RABV tropism for astrocytes.
In cells, virus infection induces IFN-b production immediately, and mass production of IFN-a/b is triggered by the earlier IFN-b response (Marié et al., 1998;Sato et al., 1998). Previously, abortive virus infection in astrocytes was found to be the main source of IFN production in the central nervous system (Pfefferkorn et al., 2016). Also, it has been reported that RABV infection of astrocytes is abortive (Pfefferkorn et al., 2016;Tian et al., 2018). In this study, viral RNA replication was detected in isolated astrocytes from RABV-infected S129 mice (Figures 3A and S3E); however, RABV N protein signals were rarely observed in astrocytes of S129 mice (Figures 4H and S3D) in contrast to strong signals in AG129 mice ( Figures 4G and S3B). Based on these observations, infection of astrocytes in immunocompetent mice may be abortive with activated Figure 4. Evaluation of RABV pathogenicity in IFN-receptor knockout mice (A and B) Twelve-week-old AG129 or S129 mice were intracranially inoculated with 10 4 ffu of rHEP or rHEP 333 R. Virus-infected mice were monitored for (A) body weight changes and (B) survival every day until 13 dpi. The values in the graph are shown as the means G standard deviations (mock group; n = 4, virus challenge group; n = 8).
(C) Virus titer in the 10% homogenate of the whole brain at 5 dpi was determined by a focus forming assay. The bar graphs show the means G standard deviations (n = 8).
(D-F) mRNA expression of IFN-b, OAS, or IFIT2 in the mouse brains at 5 dpi was determined by qRT-PCR. The results were normalized to the b-actin gene and presented as fold changes relative to the mock controls using the DDCt method. The bar graphs show the means G standard deviations (n = 8). Statistical analysis was performed by the Student's t test (*p < 0.05, ****p < 0.0001).
(G and H) Brain sections of (G) AG129 and (H) S129 mice at 5 dpi were stained for NeuN, GFAP, and RABV iScience Article IFN responses as previously reported (Pfefferkorn et al., 2016;Tian et al., 2018). Besides, compared with rHEP, rHEP 333 R showed lower infectivity to astrocytes (Figures 1, 2, 3, and 4I) and consequently induced lower IFN-b production in astrocytes ( Figures 2B, 3B, and 4D), which was independent of IFN inducibility or susceptibility to IFN ( Figure 5). Yet, in the absence of an IFN response, rHEP showed an equivalent level of virus replication and pathogenicity to rHEP 333 R both in vitro ( Figure 5E) and in vivo ( Figures 4A-4C). These results suggested that IFN production in astrocytes is associated with attenuation of rHEP.
Previous studies have proposed several possible mechanisms of G333-dependent RABV attenuation. Even though HEP strain comprising Glu 333 in G protein induces higher levels of apoptosis in infected neurons than pathogenic RABVs carrying Arg 333 or Lys 333 (Faber et al., 2002;Tao et al., 2010), another attenuated ERA strain carrying the Arg 333 was reported to cause apoptosis (Morimoto et al., 1999). Moreover, it has been demonstrated that the poor neuronal transmissibility of HEP strain in vivo, which is one of the attenuation properties, was gained by the substitution of Glu 333 to Arg 333 in G protein. Conversely, a high pathogenic CVS strain lost its virulence through Arg 333 mutation, still, it retained its distribution pattern in brain (Yan et al., 2002). Nonetheless, neuroinvasiveness was lost by an additional mutation at Lys 330 in G protein (Coulon et al., 1998). (E) Neutralization of IFN signaling. Cells were treated with anti-human IFN receptor neutralizing antibody at 800 ng/mL in the culture medium for 16 h before RABV infection at an MOI of 0.1. Virus titer in the supernatant at 24 hpi was measured by a focus forming assay. All values in the graphs show the means G standard deviations of triplicate data from a representative experiment. Statistical analysis was performed by multiple t tests using the Holm-Sidak method for (A), Student's t test for (B), (D), and (E), and standard one-way ANOVA and Tukey's multiple comparisons test for (C) (****p < 0.0001, ns = not significantly different).

OPEN ACCESS
iScience 25, 104122, April 15, 2022 7 iScience Article Although previous studies suggest that the amino acid at G333 contributes to the attenuation of RABV in multiple aspects in a strain-dependent manner indicated earlier, our novel proposed mechanism of Glu 333 -dependent astrocyte infection and IFN responses could partially account for the attenuation of HEP strain. Evidently, ERA strain comprising Arg 333 in the G protein showed diminished infection of astrocytes (Potratz et al., 2020), which further supports the G333-dependent astrocyte infection. Controversially, recent studies have shown that other field RABV strains or even some bat-associated lyssaviruses could cause infection in astrocytes regardless of the amino acid at G333 or pathogenicity (Potratz et al., 2020;Klein et al., 2022). Based on these previous findings, we note that the amino acid at position 333 alone is not sufficient to determine astrocyte tropism of all RABV strains.
In conclusion, we have demonstrated here that rHEP infected astrocytes more efficiently than rHEP 333 R and consequently, evoked higher IFN responses in astrocytes in vitro and in vivo. In addition, we revealed that the IFN response is indispensable for both virulence and attenuation of rHEP using AG129 mice deficient in IFN responses. Taken together, our findings confirm that Glu 333 in G protein is involved in the attenuation of rHEP, at least partially, by excessive infection of astrocytes triggering the induction of IFN. Our study provides new insights into the mechanism of RABV virulence attenuation and highlights the importance of astrocytes as potential drivers of the host immune response against lethal infection with RABV. In future studies, a profound understanding of RABV infection of astrocytes and the evoked immune responses should help to unravel the mechanism of RABV attenuation.

Limitations of the study
Our results suggested that the IFN-mediated antiviral activity in astrocytes could, at least in part, lead to the attenuation of RABV pathogenicity. However, we cannot exclude the possibility that neuronal cells, as well as astrocytes, also induce an IFN-mediated antiviral responses and are involved in the attenuation of RABV pathogenicity; this is because inhibition of the IFN pathway generally improves the growth of various viruses. Another limitation is the choice of the RABV strains. In our study, we used a laboratory fixed strain HEP-Flury, but infectivity of astrocytes is variable among RABV strains (Potratz et al., 2020). Thus, comparison using different RABV strains including RABV street strains will provide a better understanding of inter-strain variations. In addition, further investigations are needed to elucidate the specific mechanism of clearance of RABVs dependent on the amino acid at G333 that results in reduced pathogenicity.

STAR+METHODS
Detailed methods are provided in the online version of this paper and include the following:
Takayama-Ito, M., Inoue, K.I., Shoji, Y., Inoue, S., Iijima, T., Sakai, T., Kurane, I., and Morimoto, K.   iScience Article provided enzymes. C Tubes were attached to the gentleMACS Dissociator (Miltenyi Biotec), and the following gentleMACS programs were performed: m_brain_01_01, m_brain_02_01, m_brain_03_01. Each program was repeated twice with 5-min intervals on a tube rotator at 37 C. The dissociated brain was applied onto a 70 mm cell strainer and debris and red cell removal steps were performed in accordance with the manufacturer's protocol.

QUANTIFICATION AND STATISTICAL ANALYSIS
All statistical analyses were performed using GraphPad Prism software. For analyses between two groups, a two-tailed unpaired Student's t-test was used. For the comparison of two groups at multiple time points, a multiple t-test by the Holm-Sidak method was performed. For comparisons among more than two groups, one-way ANOVA with Tukey's multiple comparisons test was used (*p < 0.05, **p < 0.01, ****p<0.0001, ns=not significantly different). Data were presented as the mean G standard deviation (SD) in graphs.