Involvement of the Rabies Virus Phosphoprotein Gene in Neuroinvasiveness

Abstract

Rabies virus (RABV), which is transmitted via a bite wound caused by a rabid animal, infects peripheral nerves and then spreads to the central nervous system (CNS) before causing severe neurological symptoms and death in the infected individual. Despite the importance of this ability of the virus to spread from a peripheral site to the CNS (neuroinvasiveness) in the pathogenesis of rabies, little is known about the mechanism underlying the neuroinvasiveness of RABV. In this study, to obtain insights into the mechanism, we conducted comparative analysis of two fixed RABV strains, Nishigahara and the derivative strain Ni-CE, which cause lethal and asymptomatic infections, respectively, in mice after intramuscular inoculation. Examination of a series of chimeric viruses harboring the respective genes from Nishigahara in the genetic background of Ni-CE revealed that the Nishigahara phosphoprotein (P) gene plays a major role in the neuroinvasiveness by mediating infection of peripheral nerves. The results obtained from both in vivo and in vitro experiments strongly suggested that the Nishigahara P gene, but not the Ni-CE P gene, is important for stable viral replication in muscle cells. Further investigation based on the previous finding that RABV phosphoprotein counteracts the host interferon (IFN) system demonstrated that the Nishigahara P gene, but not the Ni-CE P gene, functions to suppress expression of the beta interferon (IFN-β) gene (Ifn-β) and IFN-stimulated genes in muscle cells. In conclusion, we provide the first data strongly suggesting that RABV phosphoprotein assists viral replication in muscle cells by counteracting the host IFN system and, consequently, enhances infection of peripheral nerves.

INTRODUCTION

Rabies virus (RABV), a member of the genus Lyssavirus of the family Rhabdoviridae, infects almost all kinds of mammals, including humans, and causes a severe neurological disease with a high mortality rate of about 100% after a long and inconstant incubation period (usually 20 to 90 days in humans) (reviewed in reference 1). It is estimated that more than 55,000 people die of rabies every year, mainly in Asia and Africa (2), due to the absence of an effective cure and also the complexity and expensiveness of current postexposure prophylaxis, which requires medical treatment (i.e., rabies vaccination) five times over a period of 28 days. In order to develop both therapeutic and novel prophylaxis approaches for rabies, it is necessary to fully understand the pathogenesis of rabies.

The pathogenesis of rabies essentially relies on viral spread to and in the nervous system of the infected individual (reviewed in reference 1). RABV secreted into saliva of a rabid animal is generally transmitted via a bite wound caused by the infected animal. After transmission, RABV infects peripheral nerves and then spreads to the central nervous system (CNS) via retrograde axonal transport, followed by active viral replication and spread in the CNS, culminating in severe neurological symptoms and lethal outcome. To date, studies on pathogenesis of rabies have mainly focused on ability of the virus to spread in the CNS and to cause disease (neurovirulence), and results of those studies have revealed that the cell-to-cell infection and viral evasion of neuronal apoptosis and innate immunity contribute to the neurovirulence of RABV (3–6).

In contrast, less is known about ability of the virus to spread from a peripheral site to the CNS (neuroinvasiveness). For instance, it is not clear whether RABV replication in nonneural peripheral tissue is an important key to infection of peripheral nerves. Previous histopathological studies using animals inoculated with street RABV strains (field isolates) or fixed RABV strains (laboratory and vaccine strains) via the intramuscular (i.m.) route have demonstrated the presence of the viral antigen in muscle cells prior to detection of the antigen in peripheral nerves (7–10). These observations indicate the possibility that RABV replication in muscle cells contributes to efficient infection of peripheral nerves. However, this possibility has been questioned by results of other previous studies using both street and fixed RABV strains, strongly suggesting that RABV directly infects peripheral nerves without replication in muscle cells after i.m. inoculation (11–13). Therefore, the contribution of RABV replication in muscle cells to infection of peripheral nerves is still controversial. This is mainly due to a lack of experimental data showing a strong relationship between viral replication in muscle cells and infection of peripheral nerves.

The RABV particle contains an unsegmented negative-sense genomic RNA encoding five structural proteins: nucleoprotein (N), phosphoprotein (P), matrix (M) protein, glycoprotein (G), and large (L) protein (reviewed in reference 14). The N, P, and L proteins and viral genomic RNA constitute a ribonucleoprotein (RNP) complex. The N protein enwraps the genomic RNA, whereas the P protein, in combination with the L protein, known as an RNA-dependent RNA polymerase, is involved in viral RNA synthesis. The P protein also functions as an interferon (IFN) antagonist by inhibiting both IFN induction and response (4, 15–21). The M protein, which interacts with both the RNP and G protein, participates in recruiting RNP to the host cell membrane and budding of enveloped virus particles. The G protein forms spikes that project from the viral envelop and plays an indispensable role in interaction with receptors on host cells.

Among these viral proteins, G protein is known to play an important role in the neuroinvasiveness of RABV. Prehaud et al. (22) demonstrated by i.m. inoculation of mice with a series of neutralizing escape mutants that Lys-to-Gln mutation at position 147 in G protein significantly decreases the lethality of an RABV strain. Notably, it was shown that lentiviral vectors pseudotyped with RABV G protein invade peripheral nerves and reach the CNS after i.m. inoculation (23, 24), clearly indicating that G protein is responsible for both internalization to peripheral nerves and retrograde axonal transport of RABV. Other previous studies also indicated or strongly suggested the importance of G protein in the neuroinvasiveness of RABV (25–29).

Meanwhile, there is also accumulating evidence that viral proteins other than G protein are also involved in the neuroinvasiveness of RABV. Pulmanausahakul et al. (27) showed by comparative analysis of the highly neuroinvasive SB strain and attenuated SN strain that, after i.m. inoculation, the chimeric virus with the G gene from the SB strain in the genetic background of the SN strain causes lower morbidity and mortality in mice than does the SB strain. This indicates that viral proteins other than G protein also contribute to the ability of the SB strain to infect peripheral nerves, to invade the CNS, and to cause lethal neurological symptoms. Indeed, further analysis and a previous study revealed that the M and L genes play supplementary roles in these abilities of SB strain (25, 27). However, the roles of RABV proteins other than G protein in the neuroinvasiveness still remain to be elucidated.

We previously reported that the fixed RABV strain Nishigahara (Ni) causes lethal neurological symptoms in mice after intracerebral (i.c.) inoculation, whereas the derivative Ni-CE strain causes mild symptoms, such as transient body weight loss in mice (30). Although these strains differ in lethality in mice, this indicates that both strains have the ability to infect neurons in the brain and to cause symptomatic infection in mice. Interestingly, we found that, after i.m. inoculation, the Ni-CE strain causes asymptomatic infection in mice, in contrast to the Ni strain, which causes lethal symptomatic infection (unpublished data). Therefore, the Ni and Ni-CE strains differ in neuroinvasiveness in mice.

In the present study, we examined the mechanisms underlying the different levels of neuroinvasiveness of the Ni and Ni-CE strains by analyzing a series of chimeric viruses with the respective genes from Ni strain in the genetic background of the Ni-CE strain. The results indicated that the P gene is mainly responsible for the different levels of neuroinvasiveness of the Ni and Ni-CE strains and mediates their abilities to infect peripheral nerves. We also demonstrated that the Ni P gene, but not the Ni-CE P gene, functions to assist viral replication in muscle cells. The findings obtained from further experiments based on the IFN-antagonistic function of RABV P protein (4, 15–21) revealed that the Ni P gene, but not the Ni-CE P gene, functions to evade induction of the IFN-β (Ifn-β) gene and IFN-stimulated genes (ISGs), such as the myxovirus resistance A (Mx1) and 2′-5′-oligoadenylate synthetase 1 (Oas1) genes. This study provides the first data strongly suggesting that replication of RABV in muscle cells, which is supported by the P protein's IFN-antagonistic function, enhances infection of peripheral nerves.

MATERIALS AND METHODS

Cells.

Mouse neuroblastoma NA cells were maintained in Eagle's minimal essential medium (EMEM) supplemented with 10% fetal calf serum (FCS). Mouse muscle myoblast G-8 cells (American Type Culture Collection [ATCC] no. CRL-1456) were grown in high-glucose Dulbecco's modified Eagle's medium (DMEM) supplemented with 10% FCS and 10% horse serum (HS). Mouse muscle myoblast C2C12 cells (ATCC no. CRL-1772) and human rhabdomyosarcoma A-673 cells (ATCC no. CRL-1598) were grown in DMEM supplemented with 10% FCS. Differentiation of G-8 and C2C12 cells was carried out according to the instructions of ATCC. Briefly, G-8 and C2C12 cells were differentiated by reducing the FCS and HS to 2% each and by supplying 10% HS instead of FCS to the growth medium, respectively.

Viruses.

As Ni and Ni-CE strains, we used recombinant viruses of the respective strains that had been recovered from cloned cDNA (30, 31). Chimeric CE(NiN), CE(NiP), CE(NiM), CE(NiG), and CE(NiL) strains with respective genes of the Ni strain in the background of the Ni-CE genome were also previously generated by a reverse-genetics approach (30) (Fig. 1).

Fig 1.

Schematic diagrams of genome organizations of the RABV strains used in this study. The black and white bars represent the viral genes that originate from Ni and Ni-CE strains, respectively. The striped bars and dotted bars represent the GFP gene and the firefly luciferase gene, respectively.

To generate a recombinant Ni-CE strain expressing green fluorescent protein (GFP) (Ni-CE-GFP) (Fig. 1), we amplified a cDNA fragment containing the GFP gene downstream of transcriptional stop and start signals of RABV by PCR using the pEGFP-N1 vector (Clontech, Mountain View, CA) as a template and then inserted the fragment into the G-L intergenic region (upstream of the transcriptional stop signal of the G gene) of the full-length genome plasmid of the Ni-CE strain. We rescued the Ni-CE-GFP strain from the resulting plasmid, as previously reported (32). In the same manner, we generated a recombinant chimeric CE(NiP) strain expressing GFP [CE(NiP)-GFP] from the full-length genome plasmid. Recombinant Ni, Ni-CE, and chimeric CE(NiP) strains expressing firefly luciferase [Ni-Luc, Ni-CE-Luc, and CE(NiP)-Luc, respectively] (Fig. 1) were generated by the same procedure as that described above using pGL3-Control vector (Promega, Madison, WI) as a template for PCR. Details of the construction of these plasmids and sequences of the primers are available from the authors on request. All virus stocks were prepared in NA cells and stored at −80°C. Titers of viruses were determined by focus assays in NA cells using monoclonal antibody 13-27 against RABV N protein (33) and fluorescein isothiocyanate (FITC)-labeled goat IgG anti-mouse IgG.

Pathogenicity of each virus in mice.

Five 4-week-old female ddY mice (Japan SLC, Inc., Shizuoka, Japan) per group were inoculated intracerebrally with 0.03 ml of 10⁶ focus-forming units (FFU) of each strain or intramuscularly with 0.1 ml of 10⁶ FFU of each strain into the left thigh muscle. In another experiment, 20 mice per group were intramuscularly inoculated with 10⁶ FFU of the Ni-CE or CE(NiP) strain. The mice were observed daily for 14 days. The symptoms in mice were classified into 5 grades: (i) normal, (ii) body weight loss (5% reduction from maximum body weight), (iii) mild neurological symptoms (such as stagger or gait abnormality of a unilateral hind limb), (iv) severe neurological symptoms (such as gait abnormality of bilateral hind limbs), and (v) death. We euthanized mice when they showed lack of righting reflex (mice unable to right themselves within 10 s after being placed on their side). All animal experiments in this study were conducted in accordance with the Regulations for Animal Experiments in Gifu University; the protocols were approved by the Committee for Animal Research and Welfare of Gifu University (approval no. 10086).

Biodistribution of each strain in mice.

Three 4-week-old ddY mice (Japan SLC, Inc.) per group were intramuscularly inoculated with 10⁶ FFU of the Ni or Ni-CE strain, and five mice per group were inoculated with the CE(NiP) strain. Mice were euthanized at 5 days postinoculation (dpi), and their brains, spinal cords at cervical spine 1 (C1) to the last lumbar vertebra, sciatic nerves, and thigh muscles were quickly collected. In another experiment, 10 mice per group were intramuscularly inoculated with 10⁶ FFU of the Ni-CE or CE(NiP) strain, and their sciatic nerves and thigh muscles were collected at 8 dpi. Tissues were frozen in liquid nitrogen and stored at −80°C until being used for RNA extraction and reverse transcription (RT)-nested PCR. After the stored tissues were thawed and homogenized, total RNAs were extracted from each tissue using TRIzol reagent (Invitrogen, Carlsbad, CA) and were purified using a PureLink RNA minikit (Invitrogen) according to the manufacturer's instructions. The extracted RNAs were reverse transcribed into cDNAs by using SuperScript III reverse transcriptase (Invitrogen) with reverse transcriptase primer RHN19(+) specific to the RABV genome (Table 1). First and 2nd PCRs were performed with TaKaRa HS Ex Taq (TaKaRa Bio, Inc., Shiga, Japan) using primer pairs N501(+)-RHN18(−) and N540(+)-N1000(−), respectively, which are specific to the N gene regions of the RABV genome (Table 1).

Table 1.

Sequences of the primers used in this study

Analysis	Primer (sense)	Sequence (5′ to 3′)	Annealing site positionsa
RT	RHN19 (+)	AAAATGTAACACCTCTACAATG	52–73
PCR
1st	N501 (+)	TCCTGAGTCTGTATAGGTTGAGCAA	501–525
	RHN18 (−)	CCATGTAGCATCCAACAAAGT	1008–1028
2nd	N540 (+)	ACACCGGTAACTATAAAACAAACAT	540–564
	N1000 (−)	ATTGAACACATGACCAACGGCATTC	976–1000

^aSite from the 5′-terminal region of the full-genome sequence of the RABV Ni strain (accession no. AB044824.1).

In vitro examination of the ability of the virus to infect neurons via axon terminals.

Microfluidic culture platforms with two compartments connected by 450-μm-long microgrooves, which allow cell bodies of primary cultured neurons and their axon terminals to grow separately, were fabricated in polydimethylsiloxane (PDMS) based on previous studies (34), and protocols for assembly of the culture platform and dish were modified from previous reports (34, 35). In brief, a culture platform was placed on a lumox dish (Sarstadt, Numbrecht, Germany), and the culture platform and dish were coated with 1,500 μg/ml of poly-dl-ornithine (Sigma-Aldrich, Saint Louis, MO) for 2 days at 4°C. On the day of cell plating, after being washed with phosphate-buffered saline (PBS), the culture platform and dish were coated with 3 μg/ml laminin (Sigma) for 5 h at 37°C.

Primary cultured motor neurons were prepared as described previously (36). Briefly, spinal cords were dissected from embryos of ICR mice (Japan SLC, Inc.) at embryonic day 13 and incubated at 37°C in 0.025% trypsin (Invitrogen) for 10 min. Tissue was resuspended in neurobasal medium (Invitrogen) containing bovine serum albumin (Sigma) and DNase I (Sigma). After dissociation of tissue by gently pipetting, the supernatant containing cells was collected and centrifuged at 340 × g for 5 min. The cell pellet was resuspended in the enriched medium (37), and cells were plated in the cell body side of the coated culture platform at densities of approximately 4 × 10⁶ cells/ml. The culture medium was replaced at 3 days postplating, and cultures were kept in a CO₂-regulated 37°C incubator.

At 5 days postplating, after axons had extended across the microgrooves, cells were infected with the Ni-CE-GFP or CE(NiP)-GFP strain from the axon terminal side of the culture platform. Briefly, the wells of the axon terminal side were completely depleted of medium, and they were inoculated with 10⁶ FFU of each virus. After a 1-h incubation for virus adsorption, the virus inoculum in wells of the axon terminal side was removed and replaced with enriched medium. The medium in wells of the axon terminal side was maintained at a lower volume than that of medium in wells of the cell body side. This hydrostatic pressure completely prevented diffusion of virus through the microgrooves to the cell body side of the culture platform. At 12, 24, 36, and 48 h postinoculation (hpi), cell bodies were observed by using a Biozero fluorescence microscope (BZ-8000 series; Keyence, Osaka, Japan), and then the percentages of GFP-positive cell bodies were calculated. Images were analyzed by using IMARIS (Carl Zeiss Microscopy Co., Ltd., Oberkochen, Germany).

In vivo examination of viral replication in muscle.

Three 4-week-old ddY mice (Japan SLC, Inc.) per group were intramuscularly inoculated with 10⁶ FFU of the Ni-Luc, Ni-CE-Luc, or CE(NiP)-Luc strain. Thigh muscles of infected mice were collected at 0, 12, 24, and 72 hpi, and then they were homogenized and lysed by using 1 ml of passive lysis buffer (Promega), and the samples were stored at −80°C. After being thawed and vortexed, the samples were centrifuged at 20,600 × g for 5 min, and then 10 μl of each of the supernatants was used to measure luciferase activities (calculated as relative light units [RLU]/s/g muscle weight) with the Promega luciferase assay systems.

In vitro examination of viral replication in muscle cell lines.

Differentiated G-8, C2C12, and A-673 cells grown in a 24-well tissue culture plate were infected with the Ni-Luc, Ni-CE-Luc, or CE(NiP)-Luc strain at a multiplicity of infection (MOI) of 1. The cells were lysed at 7 dpi by using 100 μl of Promega cell culture lysis reagent and stored at −80°C until use. After being thawed and vortexed, the samples were centrifuged at 13,200 × g for 2 min, and then 10 μl of each of the supernatants was used for the luciferase assays as described above.

Differentiated G-8 cells grown in a 6-well tissue culture plate were infected with the Ni, Ni-CE, or CE(NiP) strain at an MOI of 1. At 1, 3, 5, and 7 dpi, viruses in the culture supernatants were harvested and titrated in NA cells by focus assays, as described above. All assays were carried out in triplicate, and the results are expressed as means ± standard errors of the means.

Real-time RT-PCR.

Differentiated G-8 cells grown in a 24-well tissue culture plate were infected with the Ni, Ni-CE, or CE(NiP) strain at an MOI of 1. At 24 hpi, total RNAs were extracted using an RNeasy Plus minikit (Qiagen, Hilden, Germany) and reverse transcribed into cDNAs with a random primer [hexadeoxyribonucleotide mixture; pd (N)₆] (TaKaRa Bio). The expression of the Ifn-β, Mx1, Oas1, and Gapdh (glyceraldehyde-3-phosphate dehydrogenase) genes in the infected cells was analyzed by using a TaqMan gene expression master mix (Applied Biosystems, Carlsbad, CA) with specific TaqMan probes (Ifn-β [Mm00439546], Mx1 [Mm00487796], Oas1 [Mm00449297], and GAPDH [4352339E]) in an ABI 7300 real-time PCR system (Applied Biosystems). The expression levels of Ifn-β, Mx1, and Oas1 genes are indicated as the number of copies of specific mRNA per copy of mouse Gapdh mRNA. All assays were carried out in triplicate, and the values in the graph are shown as means ± standard errors of the means.

In another experiment, three 4-week-old ddY mice (Japan SLC, Inc.) per group were intramuscularly inoculated with 10⁶ FFU of the Ni, Ni-CE, or CE(NiP) strain. Thigh muscles of infected mice were collected at 12 hpi, and then total RNAs were extracted from each tissue using TRIzol reagent (Invitrogen) and a PureLink RNA minikit (Invitrogen). The extracted RNAs were reverse transcribed into cDNAs, and the expression of the Ifn-β and Gapdh genes in the thigh muscles of infected mice was analyzed by the same method described above. The values in the graph are shown as means ± standard errors of the means.

Statistical analysis.

Student's t test was used to determine statistical significance. P values of <0.05 were considered statistically significant.

RESULTS

Examination of pathogenicities of Ni and Ni-CE strains in mice by i.c. and i.m. inoculations.

To confirm the different levels of neuroinvasiveness of the Ni and Ni-CE strains, we examined their pathogenicities in 4-week-old mice by both i.c. and i.m. inoculations (Fig. 2). After i.c. inoculation, all of the mice infected with the Ni and Ni-CE strains developed symptoms, although the severities of infection were clearly different: all of the Ni-infected mice died within 6 dpi after showing obvious neurological symptoms, whereas all of the Ni-CE-infected mice survived following transient body weight loss and, occasionally, mild neurological symptoms (Fig. 2A and B). These results strongly suggest that both the Ni and Ni-CE strains have the ability to cause symptomatic infection in mice once they have reached the brain. After i.m. inoculation, all of the Ni-infected mice developed severe neurological symptoms, similar to those after i.c. inoculation, and died within 7 dpi. In contrast, all of the Ni-CE-infected mice continuously gained body weight without showing any symptoms (Fig. 2C and D). These findings confirm that the ability of Ni strain to cause symptomatic infection in mice after i.m. inoculation is significantly higher than that of the Ni-CE strain.

Fig 2.

Progression of symptoms and changes in body weight of mice inoculated (A and B) intracerebrally or (C and D) intramuscularly with 10⁶ FFU of the Ni or Ni-CE strain. The mice were observed daily for 14 days. The symptoms in mice were classified into 5 grades: (i) normal, (ii) body weight loss, (iii) mild neurological symptoms, (iv) severe neurological symptoms, and (v) death. The changes in body weight are shown as ratios, considering the body weights of mice at day 0 as 1. The values in the graph are averages and standard errors of the ratio. ∗, all of the mice died.

Identification of the viral gene related to neuroinvasiveness.

To identify the viral gene or genes related to the difference in levels of neuroinvasiveness of the Ni and Ni-CE strains, we examined pathogenicities of chimeric CE(NiN), CE(NiP), CE(NiM), CE(NiG), and CE(NiL) strains, which have N, P, M, G, and L genes from the Ni strain in the genetic background of the Ni-CE strain, respectively (Fig. 1), by i.m. inoculation into 4-week-old mice (Fig. 3). The CE(NiM) and CE(NiL) strains caused asymptomatic infection in mice, as did the Ni-CE strain (Fig. 2C). In contrast, the CE(NiN), CE(NiP), and CE(NiG) strains caused symptomatic infection in some of the mice. We found that 80% of the CE(NiP)-infected mice developed neurological symptoms and died within 13 dpi and that 40% of the CE(NiN)- and CE(NiG)-infected mice showed symptoms and died, except for one mouse that recovered from CE(NiG) infection. These results indicate that multiple viral genes, especially the P gene, are related to the difference in levels of neuroinvasiveness of the Ni and Ni-CE strains. The importance of the Ni P gene in neuroinvasiveness was confirmed by data obtained from two independent experiments showing 75% to 80% morbidity rates of CE(NiP)-infected mice (Table 2). Based on the results, it is concluded that the P gene plays a major role in the ability of the Ni strain to cause symptomatic infection in mice after i.m. inoculation.

Fig 3.

Progression of symptoms in mice inoculated intramuscularly with 10⁶ FFU of chimeric strains with the respective genes of the Ni strain in the background of the Ni-CE genome. The mice were observed daily for 14 days.

Table 2.

Morbidity rates of mice inoculated i.m. with each strain

Strain	% morbidity (no. sick/inoculated) in:
	Expt 1a	Expt 2a	Expt 3b
Ni	100 (5/5)	100 (5/5)	NTc
Ni-CE	0 (0/5)	0 (0/5)	0 (0/20)
CE(NiN)	40 (2/5)	20 (1/5)	NT
CE(NiP)	80 (4/5)	80 (4/5)	75 (15/20)
CE(NiM)	0 (0/5)	0 (0/5)	NT
CE(NiG)	40 (2/5)	0 (0/5)	NT
CE(NiL)	0 (0/5)	0 (0/5)	NT

^aGroups of five mice were i.m. inoculated with each strain.

^bGroups of 20 mice were i.m. inoculated with the Ni-CE or CE(NiP) strain.

^cNT, not tested.

Biodistribution of the Ni, Ni-CE, and CE(NiP) strains in mice after i.m. inoculation.

To obtain an insight into the mechanism by which the Ni and Ni-CE P genes are differently involved in neuroinvasiveness, we investigated the biodistribution of the Ni, Ni-CE, and CE(NiP) strains in mice after i.m. inoculation. For this purpose, we collected brains, spinal cords, sciatic nerves, and thigh muscles from mice infected with the respective strains at 5 dpi, when 100% (3/3 mice), 0% (0/3 mice), and 20% (1/5 mice) of Ni-, Ni-CE-, and CE(NiP)-infected mice, respectively, showed symptoms (Fig. 4A). Next, we examined the presence of viral genomic RNAs in these tissues by RT-nested PCR, of which the detection limit is equivalent to 10 FFU of infectious viruses (data not shown). As this RT-nested PCR system is designed to target viral genomic RNAs, the system detects, in addition to viral RNAs of replicating viruses, those of nonreplicating viruses, such as viruses traveling along axons of the sciatic nerve. In all of the infected mice, viral RNAs were detected in thigh muscles, which were the sites of viral inoculation (Fig. 4A). Considering the design and high sensitivity of this RT-nested PCR, it is highly possible that not only genomic RNAs of replicating viruses but also those of nonreplicating viruses contained in the inoculum were detected in muscles. We found that in all of the Ni-infected mice, viral RNAs were present in the brains, spinal cords, and sciatic nerves, whereas in all of the Ni-CE-infected mice, viral RNAs were detected in none of these nerve tissues. In 40% (2/5 mice) of the CE(NiP)-infected mice, viral RNAs were found in all of the examined nerve tissues. These results indicate that the abilities of the Ni and CE(NiP) strains to infect peripheral nerves are higher than that of Ni-CE strain.

Fig 4.

Viral biodistribution in mice inoculated with each strain. Mice were intramuscularly inoculated with 10⁶ FFU of the Ni, Ni-CE, or CE(NiP) strain. Tissues were collected (A) at 5 dpi or (B) at 8 dpi and analyzed for viral genomic RNAs by RT-nested PCR targeting the N gene region (461 bp). M, marker; #1 to #10, identification numbers of mice; PC, positive control (mouse brain intracerebrally inoculated with challenge virus standard [CVS] strain); NC, negative control (each tissue i.m. inoculated with medium). Symptoms: +, mouse with symptoms such as body weight loss and neurological symptoms on the day of tissue collection; −, mouse without any symptoms on the day of tissue collection.

To confirm the difference between the abilities of Ni-CE and CE(NiP) strains to infect peripheral nerves, we intramuscularly inoculated 10 mice per group with the Ni-CE or CE(NiP) strain and examined the presence of viral RNA in sciatic nerves and thigh muscles by using the same method at 8 dpi, when 0% (0/10 mice) and 40% (4/10 mice) of Ni-CE- and CE(NiP)-infected mice, respectively, showed symptoms (Fig. 4B). We detected viral RNAs in the thigh muscles of all of the infected mice. Importantly, in 60% (6/10 mice) of the CE(NiP)-infected mice, viral RNAs were found in sciatic nerves, whereas viral RNAs were not detected in sciatic nerves of any of the Ni-CE-infected mice, These results indicate that the Ni P gene, but not the Ni-CE P gene, mediates infection of peripheral nerves.

Abilities of Ni-CE and CE(NiP) strains to infect peripheral nerves.

Next, using an in vitro system, we examined whether the Ni-CE strain, but not the CE(NiP) strain, has a defect in the ability to infect peripheral nerves via axon terminals. For this purpose, we utilized primary cultured neurons grown by using a microfluidic culture platform, which enables maintenance of their cell body and axon terminal compartments fluidically isolated from each other (Fig. 5A and B). We inoculated the GFP-expressing Ni-CE or CE(NiP) strain [Ni-CE-GFP or CE(NiP)-GFP] (Fig. 1) into the axon terminal side and then checked for the presence of GFP signals in the cell bodies. At 48 hpi, GFP signals were observed in the cell bodies of neurons infected with the Ni-CE-GFP strain and those infected with the CE(NiP)-GFP strain (Fig. 5C). After GFP signals were first detected at 24 hpi, the percentages of GFP-positive cell bodies in both neurons infected with Ni-CE-GFP and CE(NiP)-GFP strains similarly increased over time (Fig. 5D). We confirmed the reproducibility of these results by a repeated experiment (data not shown). These results suggest that both strains invade neurons via axon terminals, travel retrogradely in axons, and replicate in cell bodies with similar efficiencies. Therefore, we concluded that the Ni-CE strain has the ability to infect peripheral nerves via axon terminals.

Fig 5.

Ability of virus to infect neurons via axon terminals. (A) Diagram of microfluidic culture platform. (B) Image of Ni-CE-GFP-infected neurons at 48 hpi. (C) Cell bodies of Ni-CE-GFP- or CE(NiP)-GFP-infected neurons at 48 hpi. (D) Percentages of GFP-positive cell bodies in neurons infected with the Ni-CE-GFP or CE(NiP)-GFP strain. The scale bars correspond to 50 μm.

Contribution of the P gene to viral replication in muscle cells.

The above results obtained from in vitro experiments are inconsistent with the finding that the Ni-CE strain does not infect peripheral nerves in vivo (Fig. 4), leading to the possibility that the Ni and Ni-CE P genes are differently involved in viral replication in muscle. To examine this possibility, we inoculated the firefly luciferase-expressing Ni, Ni-CE, or CE(NiP) strain [Ni-Luc, Ni-CE-Luc, or CE(NiP)-Luc] (Fig. 1) into thigh muscles of mice and compared the activities of luciferase expressed in the inoculated thigh muscles after sequential collections of muscle samples (Fig. 6A). In Ni-Luc- and CE(NiP)-Luc-infected mice, but not in Ni-CE-Luc-infected mice, the luciferase activities in muscle gradually increased over time. At 72 hpi, the luciferase activity in muscle from Ni-Luc-infected mice was significantly higher than that in muscle from Ni-CE-Luc-infected mice (P < 0.05). The luciferase activity in muscle from CE(NiP)-Luc-infected mice also tended to be higher than that in muscle from Ni-CE-Luc-infected mice. Viral genomic RNAs were detected in the inoculated thigh muscles of all of the infected mice (Fig. 6B), confirming that each strain had been inoculated into thigh muscle. These results suggest that the Ni P gene functions to assist viral replication in muscle in vivo.

Fig 6.

Viral replication in mouse thigh muscle. Mice were intramuscularly inoculated with 10⁶ FFU of the Ni-Luc, Ni-CE-Luc, or CE(NiP)-Luc strain. (A) At 0, 12, 24, and 72 hpi, mouse thigh muscles were collected and used for analyses of luciferase activities (calculated as relative light units [RLU])/s/g muscle weight) by the luciferase assay system. The values in the graph are shown as means ± standard errors of the means. ∗, significant difference (P < 0.05). (B) Muscle tissue debris in the lysate prepared at 72 hpi for the above luciferase assay was examined for viral genomic RNAs by RT-nested PCR targeting the N gene region (461 bp). M, marker; #1 to #3, identification numbers of mice; PC, positive control (mouse brain i.c. inoculated with CVS strain).

To investigate the importance of the Ni P gene in viral replication in muscle cells in vitro, we inoculated the Ni-Luc, Ni-CE-Luc, or CE(NiP)-Luc strain into three cell lines derived from mouse or human muscle cells and compared the luciferase activities expressed in the infected muscle cells (Fig. 7). We found that the activities in mouse muscle myoblast G-8 cells infected with the Ni-Luc and CE(NiP)-Luc strains were significantly higher than that in cells infected with Ni-CE-Luc strain (P < 0.05) (Fig. 7A). Similar results were obtained with infection of mouse muscle myoblast C2C12 cells and human rhabdomyosarcoma A-673 cells with these luciferase-expressing strains (Fig. 7B and C). These results indicate that the efficiencies of replication of the Ni and CE(NiP) strains in these muscle cells are higher than that of the Ni-CE strain.

Fig 7.

Viral replication in cultured muscle cell lines. (A) Mouse muscle myoblast G-8 cells, (B) C2C12 cells, and (C) human rhabdomyosarcoma A-673 cells were infected with the Ni-Luc, Ni-CE-Luc, or CE(NiP)-Luc strain at an MOI of 1. At 7 dpi, the cells were used for analyses of luciferase activities (calculated as RLU/s) by the luciferase assay system. All assays were carried out in triplicate, and the values in the graph are shown as means ± standard errors of the means. ∗, significant difference (P < 0.05).

Next, we compared the levels of production of infectious viruses in G-8 cells infected with the Ni, Ni-CE, and CE(NiP) strains (Fig. 8). The virus titers of the Ni and CE(NiP) strains in culture supernatants slightly increased over time, whereas the titer of the Ni-CE strain drastically decreased: while the titers at 7 dpi of the Ni and CE(NiP) strains reached about 10⁵ FFU/ml, the titer of the Ni-CE strain decreased to <10² FFU/ml (P < 0.05). We confirmed that <10² FFU/ml of the respective viruses was carried over from the inoculum into the culture supernatants at 0 dpi (data not shown). The results presented above indicate that infectious viruses were produced more efficiently in G-8 cells infected with the Ni and CE(NiP) strains than in the cells infected with the Ni-CE strain.

Fig 8.

Growth curves of each strain in muscle G-8 cells. G-8 cells were infected with the Ni, Ni-CE, or CE(NiP) strain at an MOI of 1. The viruses in the supernatants were harvested at 1, 3, 5, and 7 dpi and titrated in NA cells by focus assay. All assays were carried out in triplicate, and the values in the graph are shown as means ± standard errors of the means. ∗, significant difference (P < 0.05).

Based on the above findings, it is concluded that the Ni P gene, but not the Ni-CE P gene, mediates stable viral replication in muscle cells, supporting the possibility that viral replication in muscle cells enhances infection of peripheral nerves.

IFN induction and responses in muscle cells infected with the respective strains.

According to the results of previous studies showing that RABV P protein functions to antagonize the host IFN system (4, 15–21), we checked whether the CE(NiP) and Ni-CE strains differ in the ability to suppress IFN induction in muscle cells. Specifically, we compared the expression levels of the Ifn-β gene and also the Mx1 and Oas1 genes, known as ISGs, in G-8 cells infected with the Ni, Ni-CE, and CE(NiP) strains (Fig. 9). To minimize the influence of different replication efficiencies of those strains in G-8 cells, we chose 1 dpi, at which time the titers of the respective strains in the culture supernatants were comparable (Fig. 8), as the time for RNA collection. We found that the expression levels of the Ifn-β gene in Ni- and CE(NiP)-infected cells were significantly lower than the level in Ni-CE-infected cells (P < 0.05) (Fig. 9A). Similar results were obtained with an equivalent assay using mouse muscle C2C12 cells (data not shown). Consistent with these results, the expression levels of the Mx1 and Oas1 genes were suppressed in Ni- and CE(NiP)-infected G-8 cells more efficiently than that in Ni-CE-infected cells (P < 0.05) (Fig. 9B and C). These results indicate that the Ni and CE(NiP) strains suppress expression of the Ifn-β gene and ISGs in muscle cells more efficiently than does the Ni-CE strain.

Fig 9.

Relative expression levels of IFN-related genes in virus-infected muscle G-8 cells. G-8 cells were infected with the Ni, Ni-CE, or CE(NiP) strain at an MOI of 1 and then lysed at 24 hpi for RNA extraction. The expression levels of the (A) Ifn-β, (B) Mx1, and (C) Oas1 genes are indicated as the number of copies of specific mRNA per copy of mouse Gapdh mRNA. All assays were carried out in triplicate, and the values in the graph are shown as means ± standard errors of the means. ∗, significant difference (P < 0.05).

To investigate the abilities of the Ni, Ni-CE, and CE(NiP) strains to suppress IFN induction in muscle in vivo, we compared the expression levels of the Ifn-β gene in inoculated thigh muscles of Ni-, Ni-CE-, and CE(NiP)-infected mice (Fig. 10). We found that expression levels of the Ifn-β gene in the muscles from Ni- and CE(NiP)-infected mice tended to be lower than that in the muscles from Ni-CE-infected mice.

Fig 10.

Relative expression levels of the Ifn-β gene in thigh muscles of infected mice. Mice were intramuscularly inoculated with 10⁶ FFU of the Ni, Ni-CE, or CE(NiP) strain. Thigh muscles were collected at 12 hpi, and total RNAs were extracted from each tissue. The expression level of the Ifn-β gene is indicated as the number of copies of specific mRNA per copy of mouse Gapdh mRNA. The values in the graph are shown as means ± standard errors of the means.

These results indicate that the Ni P gene, but not the Ni-CE P gene, functions to suppress IFN induction in muscle cells and also strongly suggest that the function of P protein to antagonize the host IFN system affects the efficiency of viral replication in infected muscle cells.

DISCUSSION

Previous studies have demonstrated that G protein plays an important role in neuroinvasiveness of RABV (22–29). On the other hand, there is accumulating evidence that G protein is not the only viral component determining the ability of RABV to cause lethal neurological symptoms after peripheral inoculation (25–29). Although minor contributions of M and L proteins to this viral ability have been reported (25, 27), the importance of viral proteins other than G protein in neuroinvasiveness of RABV remains to be elucidated. In this study, we conducted comparative analysis of the two RABV strains Ni and Ni-CE, which cause lethal and asymptomatic infections, respectively, in mice after i.m. inoculation (Fig. 2). Using chimeric viruses with the respective genes from the Ni strain in the genetic background of the Ni-CE strain, we revealed that the P gene is mainly related to the difference in levels of neuroinvasiveness of the Ni and Ni-CE strains (Fig. 3 and Table 2). To our knowledge, this is the first report that the P gene plays a major role in neuroinvasiveness of RABV.

Since the Ni-CE and CE(NiP) strains are genetically identical except for five amino acid substitutions in the P protein (at positions 56, 58, 66, 81, and 226) (30), it is obvious that one or a combination of these amino acid substitutions determine the levels of neuroinvasiveness of the Ni and Ni-CE strains. Our preliminary experiment revealed that neither of the Ni-CE mutant strains, CE(NiP56-81) and CE(NiP226), in which amino acids at positions 56, 58, 66, and 81 and the amino acid at position 226 in P protein were replaced with those the Ni strain, respectively, caused symptomatic infection in any of the mice after i.m. inoculation (data not shown). This indicates that another combination of amino acid substitutions in P protein is related to the difference in levels of neuroinvasiveness of the Ni and Ni-CE strains. Further experiments with additional Ni-CE mutants will be required to identify the key amino acid residues in P protein.

We revealed by i.m. inoculation of mice with 10⁶ FFU of each strain that, in contrast to the Ni and CE(NiP) strains, the Ni-CE strain does not infect peripheral nerves (Fig. 4). This raises the possibility that Ni-CE P protein, but not Ni P protein, has a functional defect that disrupts the ability of the virus to be internalized into and/or to be retrogradely transported in peripheral nerves. However, previous findings strongly suggest that this is not the case. More specifically, it was reported that retrograde axonal transport of RABV takes place in the form of a viral particle, not a viral RNP complex (38). On the other hand, P protein is a component of the viral RNP complex and is theoretically not exposed to the outside of the viral particle through the processes of viral internalization and retrograde axonal transport. Therefore, it is expected that P protein is not able to directly interact with host cellular machineries, which play important roles in these processes.

To experimentally check the above theory, we grew primary cultured neurons by using a microfluidic culture platform, which allows us to fluidically separate their cell body and axon terminal compartments. This culture system has been found to be a useful tool in neuroscience studies, such as studies on transport of neurotrophic factors along axons (39, 40), regeneration of axons (34, 41), and interaction of axons with glial cells (34, 42). This culture system is also thought to be useful in studies on neurotropic viruses, although a limited number of virology studies have utilized this system. In fact, it has been successfully used for the study of axonal transport of herpesvirus (43, 44). Using this system, we inoculated axon terminals of primary cultured neurons with the Ni-CE-GFP or CE(NiP)-GFP strain and checked the presence of GFP signals in the cell bodies. The fact that GFP signals were detected in both cell bodies of neurons infected with Ni-CE-GFP and CE(NiP)-GFP strains (Fig. 5C) clearly indicated that both strains have the ability to infect neurons via axon terminals. Notably, at 24 hpi, when GFP signals were first detected in the cell bodies of neurons infected with the Ni-CE-GFP and CE(NiP)-GFP strains, their percentages of GFP-positive cell bodies were comparable (Fig. 5D), suggesting that both strains were transported from axon terminals to cell bodies of neurons with almost identical efficiencies. These data support the theory that Ni-CE P protein does not negatively affect the ability of the virus to be internalized into and/or transported retrogradely in the peripheral nerves.

Previous studies using both street and fixed RABV strains have demonstrated that muscle cells are the infection sites in peripheral tissue (7–10, 45). In this study, we compared in vivo replication efficiencies of the Ni-Luc, Ni-CE-Luc, and CE(NiP)-Luc strains in muscle and obtained data strongly suggesting that the Ni P gene, but not the Ni-CE P gene, mediates stable viral replication in muscle (Fig. 6). This finding is supported by the results of in vitro experiments showing that the Ni and CE(NiP) strains replicate in cultured muscle cells more efficiently than does the Ni-CE strain (Fig. 7 and 8). Notably, we previously reported that the Ni-CE and CE(NiP) strains grew similarly in mouse and human neuroblastoma cells (4, 30). This indicates that the low replication efficiency of the Ni-CE strain in muscle cells is not due to a defect in the fundamental replication ability.

Previous histopathological studies showed that RABV replicates in muscle cells before infecting peripheral nerves, implying that viral replication in muscle cells enhances the infection of peripheral nerves (7–10). However, this possibility has not been supported by experimental data showing a correlation between the abilities of the virus to replicate in muscle cells and to infect peripheral nerves. In this study, we demonstrated that the Ni and CE(NiP) strains, which have the ability to infect peripheral nerves in vivo, are capable of stably replicating in muscle cells, whereas the Ni-CE strain, with a defect in this ability, is not (Fig. 6, 7, and 8). Hence, we present here the first data indicating that efficiency of RABV replication in muscle cells correlates with the ability of the virus to infect peripheral nerves. This correlation leads to the possibility that replication of RABV in muscle cells is important to present a certain number of infectious viruses to axon terminals of peripheral nerves, consequently enhancing infection of peripheral nerves.

While the results above highlight the contribution of viral replication in muscle cells to the efficient infection of peripheral nerves, several studies using both street and fixed viruses have demonstrated that RABV has the ability to directly infect peripheral nerves without viral replication in muscle cells (11–13, 46, 47). Previous findings that lentiviral vectors pseudotyped with RABV G protein reach the CNS after i.m. inoculation (23, 24) also indicated the lack of necessity of viral replication in muscle cells for infection of peripheral nerves. Notably, our preliminary data demonstrated that when mice were inoculated intramuscularly with a higher dose (i.e., 5 × 10⁷ FFU) of the Ni-CE strain, which has a defect in the ability to replicate in muscle cells, some (∼60%) of them developed symptoms such as transient body weight loss (data not shown), strongly suggesting direct infection of peripheral nerves with the Ni-CE strain. We believe that various and complex factors, including the dose and biological characteristics (e.g., viral tropisms to neuron and muscle cells) of the RABV strain used for inoculation affect the efficiency of direct infection of peripheral nerves. In the previous studies described above, this efficiency might have increased under certain conditions, masking the contribution of RABV replication in muscle cells to efficient infection of peripheral nerves.

In response to viral infection, host cells produce type I IFN, including the IFN-α family and IFN-β, which, on binding to IFN receptors on the cell surface, leads cells to an antiviral status by inducing the expression of ISGs encoding antiviral proteins (reviewed in reference 48). It was previously demonstrated that RABV P protein functions to antagonize the host IFN system by inhibiting both cellular signaling pathways for induction of the IFN gene and ISGs (4, 15–21). Based on those findings, we hypothesized that this IFN-antagonistic function of P protein is involved in the different replication efficiencies of CE(NiP) and Ni-CE strains in muscle cells. Consistent with this hypothesis, in muscle G-8 cells, the expression levels of the Ifn-β gene and also the Mx1 and Oas1 genes, known as ISGs, in CE(NiP)-infected cells were significantly lower than the levels in Ni-CE-infected cells (Fig. 9). In addition, the expression levels of the Ifn-β gene in inoculated thigh muscles of CE(NiP)-infected mice tended to be lower than that in the muscles of Ni-CE-infected mice (Fig. 10). Importantly, we found that in Vero cells, which are known to be IFN deficient (49), the Ni-CE-Luc and CE(NiP)-Luc strains expressed almost identical levels of luciferase (data not shown), indicating that the replication efficiencies of the Ni-CE-Luc and CE(NiP)-Luc strains are comparable in the absence of an intact IFN system. These findings strongly suggest that the host IFN system suppresses replication of the Ni-CE strain in muscle cells more efficiently than that of the CE(NiP) strain.

We showed that Ni-CE infection induces the Ifn-β gene more efficiently than does CE(NiP) infection in muscle cells (Fig. 9), but the molecular mechanism remains to be elucidated. Brzózka et al. (16) reported that RABV P protein inhibits phosphorylation of interferon regulatory factor 3 (IRF-3), which is an important transcription factor for IFN induction. Since there are five amino acid substitutions in P protein between the Ni and Ni-CE strains (30), these amino acid mutations in Ni-CE P protein might disrupt its function to inhibit IRF-3 phosphorylation in muscle cells. Interestingly, in neuroblastoma SYM-I cells, such a difference between expression levels of the Ifn-β gene in Ni-CE- and CE(NiP)-infected cells was not observed (data not shown). Consistent with this, we previously reported that in SYM-I cells, Ni-CE and Ni P proteins expressed from a plasmid equally inhibited activation of the Ifn-β promoter induced by infection with Newcastle disease virus (5). These findings suggest that the inhibitory function of Ni-CE P protein in IFN induction is deficient in muscle cells but not in neural cells. Further studies will be required to elucidate the mechanism underlying the cell-type dependency of IFN-antagonistic function of Ni-CE P protein.

Our data strongly suggested that viral evasion of the host IFN system in peripheral tissue is important for neuroinvasiveness of RABV. Importantly, this finding is supported by results of previous studies: it was shown that administration of IFN inducers, such as polyriboinosinic-polyribocytidylic acid containing poly-l-lysine and carboxymethylcellulose [poly(ICLC)] following i.m. challenge with virulent RABV enhances the protective effect of postexposure rabies vaccine in monkeys (50, 51). According to these findings, we believe that the host IFN system in peripheral tissue is a potential target for development of novel prophylactic approaches for rabies.

While we demonstrated the importance of the P gene in neuroinvasiveness of RABV, our data also indicated that the P gene is not the only viral gene determining the levels of neuroinvasiveness of the Ni and Ni-CE strains: the morbidity rate of mice inoculated with the CE(NiP) strain (75 to 80%) was lower than that of mice inoculated with the Ni strain (100%) (Table 2). Notably, the CE(NiN) and CE(NiG) strains caused symptomatic infection in 20 to 40% and 0 to 40% of mice, respectively, suggesting a minor contribution of N and G genes to the neuroinvasiveness. Consistent with the results of previous studies (25–29), the findings of this study indicate that multiple genes are involved in the ability of RABV to cause symptomatic infection after i.m. inoculation, strongly suggesting that the mechanism underlying neuroinvasiveness of RABV is complex.

In conclusion, the present study has demonstrated the importance of the RABV P gene in neuroinvasiveness. Also, we have provided findings strongly suggesting that RABV P protein assists stable viral replication in muscle cells by its IFN-antagonistic function and consequently enhances infection of peripheral nerves. We believe that our findings provide basic information for development of novel prophylactic approaches for rabies and also for establishment of a live rabies vaccine strain with a high level of safety.