Neurovirulence of Avian Influenza Virus Is Dependent on the Interaction of Viral NP Protein with FMRP in the Murine Brain

Infection of the CNS is a serious complication of human cases of AIV infection. The viral and host factors associated with neurovirulence of AIV infection are not well understood.

ABSTRACT

Avian influenza viruses (AIVs) are zoonotic viruses that exhibit a range of infectivity and severity in the human host. Severe human cases of AIV infection are often accompanied by neurological symptoms; however, the factors involved in the infection of the central nervous system (CNS) are not well known. In this study, we discovered that the avian-like sialic acid (SA)–α2,3-galactose (α2,3-Gal) receptor is highly presented in mammalian (human and mouse) brains. In the generation of a mouse-adapted neurotropic H9N2 AIV (SD16-MA virus) in BALB/c mice, we identified key adaptive mutations in its hemagglutinin (HA) and polymerase basic protein 2 (PB2) genes that conferred viral replication ability in the mouse brain. The SD16-MA virus showed binding affinity for the avian-like SA–α2,3-Gal receptor, enhanced viral RNP polymerase activity, and increased viral protein production and transport that culminated in elevated progeny virus production and severe pathogenicity. We further established that host fragile X mental retardation protein (FMRP), a highly expressed protein in the brain that is physically associated with viral nucleocapsid protein (NP) to facilitate RNP assembly and export, was an essential host factor for the neuronal replication of neurotropic AIVs (H9N2, H5N1, and H10N7 viruses). Our study identified a mechanistic process for AIVs to acquire neurovirulence in mice.

IMPORTANCE Infection of the CNS is a serious complication of human cases of AIV infections. The viral and host factors associated with the neurovirulence of AIV infections are not well understood. We identified and functionally characterized specific changes in the viral HA and PB2 genes of a mouse-adapted neurotropic avian H9N2 virus responsible for enhanced virus replication in neuronal cells and pathogenicity in mice. Importantly, we showed that host FMRP was a crucial host factor that was necessary for neurotropic AIVs (H9N2, H5N1, and H10N7 viruses) to replicate in neuronal cells. Our findings provide insights into the pathogenesis of the neurovirulence of AIV infections.

INTRODUCTION

Avian influenza viruses (AIVs), particularly the three most prevalent subtypes of H5, H7, and H9 viruses found in poultry, pose a constant threat to public health by repeatedly breaking through species barrier and infecting humans (1–4). Extended epizootics and panzootics of H5N1 viruses have led to the emergence of novel H5NX reassortants, including H5N2, H5N6, and H5N8 viruses (6–8). Until now, H5NX viruses have caused 880 human infection cases (861 cases with H5N1 and 24 cases with H5N6) at around 60% mortality (https://www.who.int/influenza/human_animal_interface/2020_01_20_tableH5N1.pdf?ua=1). H7N9 AIVs have caused five influenza epidemic waves in China since the first human infections were reported in 2013 (3, 9, 10). To date, the total number of confirmed human cases was 1,568, including 616 deaths, which translated to an approximately 39% case fatality rate (https://www.who.int/influenza/human_animal_interface/influenza_h7n9/Risk_Assessment/en). H9N2 viruses circulate globally in wild birds and are endemic in domestic poultry in some areas (11, 12). Recent studies indicate that H9N2 influenza viruses have acquired higher binding affinity for the human-like sialic acid (SA)–α2,6-Gal-linked receptor and increased virulence and transmissibility in mammals (13). Collectively, current AIVs in circulation are a growing public health threat.

Although respiratory disease is a hallmark of human influenza virus infection, AIV infections can often cause neurological complications with fatal outcomes. A review of clinical features of human infected with avian H5N1 virus found that 31.36% (74/236) of cases reported fussiness and irritability, and 25% (59/236) of cases showed consciousness disorder (14). Previously, research had demonstrated that avian H5N1 viral RNAs and antigens were detected in neuronal cells in a human infection case (15). Besides, H7N9 virus has also been reported to cause encephalitis in patients (16). These observations indicate that AIVs could cause significant damage to the central nervous system (CNS). Little is known about the host factors involved in the pathogenesis of neurological complication of AIV infection.

The attachment of viral hemagglutinin (HA) spikes to SA-containing receptors on the host cell surface initiates viral infection (17). Yen et al. (18) found that changes in the HA receptor-binding domain alter the ability of the H5N1 virus to spread systemically in mice and are important for viral neurotropism. Schrauwen et al. (19) and Suguitan et al. (20) identified a multibasic cleavage site in the HA protein as a virulent factor in the systemic spread of H5N1 virus in ferrets. Residue HA 328Y in H1N1 virus (A/WSN/33) and HA 325S in H5N1 virus were found to contribute to the ease of cleavage of HA protein into HA1 and HA2, which permits fusion of the viral envelope with the secondary endosome (21, 22). Additionally, mutations of PB2 E627K and NA R146N in mouse-adapted AIVs were shown through increasing polymerase activity and binding to fibrinolytic proteasomes to be neurovirulence factors (23, 24). Gaining insight into viral and host factors critical in neuronal infection and understanding their roles in the pathogenesis could develop new targeting treatment strategies to prevent fatal outcomes or lasting damage from neurological effects of AIV infection in humans.

In the present study, we found that mutations in the HA and PB2 genes of a mouse-adapted neurotropic avian H9N2 virus conferred binding affinity for the avian-like SA–α2,3-Gal receptor type and enhanced virus polymerase activity leading to elevated pathogenicity and replication ability in mice. We further demonstrated that the neurovirulence of AIVs was dependent on the interaction between viral NP protein and host fragile X mental retardation protein (FMRP).

RESULTS

The avian-like α-2,3-linked SA receptor is dominant in human and murine brains.

We compared the distribution of SA receptors in the brain between human and mouse using lectin histochemistry. Avian and human influenza viruses preferentially bind to α-2,3-linked and α-2,6-linked SAs, respectively (25). We found that both avian-like SA–α2,3-Gal and human-like SA–α2,6-Gal receptors were expressed in human and murine cerebrum tissues (Fig. 1). The avian-like SA–α2,3-Gal receptor appeared to be dominant in the brain of both species (Fig. 1), unlike in the human upper respiratory tract, where the SA–α2,6-Gal receptor was more abundant (26). This distribution could facilitate preferential binding of avian virus to neuronal cells of humans and mice.

FIG 1.

The avian-like SA–α2,3-Gal receptor appears to be dominant over the human-like SA–α2,6-Gal receptor in murine and human brains. The mouse and human brain tissues were stained with fluorescein isothiocyanate (FITC)-labeled MAL I (SA–α2,3-Gal receptor, green) or biotinylated SNA (SA–α2,6-Gal receptor, red); nuclei were stained with 4′,6-diamidino-2-phenylindole dihydrochloride (DAPI) (blue).

Mouse-adapted H9N2 influenza virus acquired neurovirulence.

To study neuronal adaptation of AIVs, we passaged an H9N2 virus (A/chicken/Shandong/16/05; SD16) in BALB/c mice via the intranasal (i.n.) route at 10⁶ egg infectious dose (EID₅₀) of virus per mouse. By passage 13 (P13), the mouse-adapted (MA)-H9N2 virus-infected mice exhibited clear signs of depression and tremor. Histopathology examination (by hematoxylin and eosin [H.E.] stain) showed that SD16 virus-infected brain appeared normal (Fig. 2B and E), but SD16-MA virus-infected brains showed typical symptoms of encephalitis, characterized by infiltrating inflammatory cells surrounding the blood vessels and neuronophagia (Fig. 2C and F). By immunohistochemistry (IHC), viral antigens were detected in neurons and glial cells of SD16-MA virus-infected brain (Fig. 2I and L), but not in brain tissue infected with the parental (P0) virus (Fig. 2H and K), indicating that the SD16 virus has gained neurotropism. P13 MA virus recovered from brain of infected mice was plaque purified three times in Madin-Darby canine kidney (MDCK) cells, and one clone, designated SD16-MA, was used for further studies.

FIG 2.

Mouse-adapted avian H9N2 influenza virus gained neurovirulence in mice. Representative Histological (hematoxylin and eosin [H.E.] staining; A to F) and immunohistochemistry (IHC) (G to L) of brain sections at 5 dpi are shown. Mouse-adapted H9N2 virus (passage 13 [P13])-infected brain showed typical encephalitis; white arrows indicate infiltrating inflammatory cells surrounding the blood vessels. Viral NP proteins (brown) were detected in the cerebral tissues of mice infected with mouse-adapted H9N2 virus (P13). Open arrows indicate virus presence in infected brain tissue. Bar, 200 μm.

The neurotropic SD16-MA virus showed increased pathogenicity in mice.

Influenza patients with CNS manifestations are more likely to experience severe illness with unfavorable outcomes (27). To assess the neurotropic pathogenicity of the MA-H9N2 virus, each mouse in two groups of eight BALB/c mice was i.n. inoculated with SD16 or SD16-MA virus at a dose of 10⁶ EID₅₀/mouse. Clinical signs, mortality, and body weight loss were monitored over 14 days. Mice infected with SD16-MA exhibited clear signs of depression (huddling, tremor, decreased activity, wheezing, and ruffled fur) and all mice died by 6 days postinfection (dpi) (Fig. 3A and B). In contrast, mice infected with SD16 showed only moderate weight loss at 18.7%, and all mice regained weight from 8 dpi (Fig. 3A and B). SD16-MA virus titers in the lungs at 5 dpi were at least 10-fold higher than those derived from parental SD16 virus infection (Fig. 3C). Crucially, SD16-MA virus could be isolated from the brains of infected mice (mean titers of 3.2 log₁₀ EID₅₀/ml at 5 dpi) but no virus was detected from brains of SD16 virus-infected mice (Fig. 3C). These results indicated that SD16-MA H9N2 virus produced higher viral titers and was neurotropic in mice.

FIG 3.

Mouse-adapted avian H9N2 influenza virus gained enhanced pathogenicity in mice. (A) Body weight change (percentage) and (B) survival (percentage) of mice (n = 5 per group) infected separately with SD16 virus and SD16-MA virus at 10⁶ EID₅₀/mouse. Body weight changes are presented as means ± standard deviation (SD) of five mice. Mice that lost >25% of their baseline body weight were euthanized. (C) SD16 virus and SD16-MA virus titers were determined in the lungs and brains of infected mice (n = 3 per group) at 5 dpi. The dashed black line indicates the lower limit of detection (10^0.75 EID₅₀/ml). Data are presented as means ± SD of three mice. **, P < 0.01; ***, P < 0.001 (determined by analysis of variance [ANOVA]).

Mouse-adapted mutations in PB2 and HA played important roles in H9N2 virus infection in the mouse brain.

To identify adaptive viral genetic changes in SD16-MA associated with invasion into the mouse brain, 30 plaque clones of SD16-MA were randomly picked, sequenced, and aligned with the parental SD16 parental virus. Eight consensus mutations were identified in five viral proteins, namely, PB2 (M147L, V250G, and E627K), PB1 (Y657H), HA (R211K and L226Q; position numbers are based on the H3 gene sequence), M1 (R210K), and NS1 (L214F) (Table 1). To determine the functional effect of the SD16-MA mutations, the following five recombinant viruses were constructed, each with a single SD16-MA segment, in the parental SD16 background: rSD16-MA/PB2, rSD16-MA/PB1, rSD16-MA/HA, rSD16-MA/M1, and rSD16-MA/NS1. rSD16 and rSD16-MA were also generated by reverse genetics for inclusion as controls. Groups of eight BALB/c mice were i.n. inoculated with each recombinant virus at the dose of 10⁶ EID₅₀. Clinical signs, mortality, and body weight loss of five mice were monitored over 14 days. Brains and lungs were collected from 3 BALB/c mice per group at 5 dpi for virus titration. Mice infected with rSD16-MA, rSD16-MA/PB2, and rSD16-MA/HA virus showed 25% to 28% weight loss and 100% mortality by 8 dpi (Fig. 4A and B). On the contrary, similar to rSD16 virus infection, all mice infected with rSD16-MA/PB1, rSD16-MA/M1 and rSD16-MA/NS1 viruses survived with a maximum 18% weight loss (Fig. 4A and B). Like the P13 SD16-MA virus, rSD16-MA virus was detected in brains of all 3 mice (100% isolation rate), with an average titer of 3.1 ± 0.3 log₁₀ EID₅₀/ml, while rSD16 was not detected in any mouse brain (Fig. 4C). Of the 5 recombinants, only rSD16-MA/PB2 and rSD16-MA/HA viruses were detected in 2/3 mice, with average virus titers of 2.25 to 2.75 log₁₀ EID₅₀/ml (Fig. 4C). Virus titer in lungs showed that rSD16-MA-infected group had the highest virus titer, of about 6.9 log₁₀ EID₅₀/ml. When infected with rSD16-MA/PB2 and rSD16-MA/HA, the average virus titers in lungs were 6.4 and 6.3 log₁₀ EID₅₀/ml, respectively, significantly higher than that in the rSD16 group (P < 0.05). There was no significant difference between the virus titers of lungs in rSD16-MA/PB1-, rSD16-MA/M1-, and rSD16-MA/NS1-infected groups compared with that in the rSD16-infected group (5.1 to 5.4 log₁₀ EID₅₀/ml; Fig. 4D). Thus, the mutations sited in PB2 and HA are closely involved in replication in the mouse brain and in pathogenicity of the MA H9N2 virus (SD16-MA).

TABLE 1.

Amino acid substitutions of mutants isolated from mice

Virus protein	Amino acid position	Presence of amino acid in virus:		Mutation frequency (% [no./total])
		SD16	SD16-MA
PB2	147	M	L	100 (30/30)
	250	V	G	100 (30/30)
	627	E	K	100 (30/30)
PB1	657	Y	H	100 (30/30)
	642	N	K	3 (1/30)
PA	129	I	T	3 (1/30)
	151	T	S	3 (1/30)
	588	S	P	3 (1/30)
HA	211	R	K	100 (30/30)
	226	L	Q	100 (30/30)
NA	204	S	T	20 (6/30)
M1	210	R	K	100 (30/30)
NS1	214	L	F	100 (30/30)

FIG 4.

Mutations in PB2 and HA of mouse-adapted H9N2 virus (rSD16-MA virus) contributed to neurovirulence and enhanced pathogenicity in mice. Mice were infected separately with 10⁶ EID₅₀ of rSD16, rSD16-MA, rSD16-MA/PB2, rSD16-MA/PB1, rSD16-MA/HA, rSD16-MA/M1, and rSD16-MA/NS1 viruses. (A) Body weight change (percentage), presented as means ± SD of 5 mice, and (B) survival (percentage) of mice (n = 5 per group) were monitored over 14 days. Mice that lost >25% of their baseline body weight were euthanized. Virus titers in the brain (C) and lung (D) were determined at 5 dpi. Each colored bar represents the virus titer from an individual animal. The dashed black line indicates the lower limit of detection (10^0.75 EID₅₀/ml). *, P < 0.05; **, P < 0.01 (determined by ANOVA).

Mouse-adapted H9N2 virus reverted to SA-α2,3 receptor binding preference.

The binding specificity of HA to the host receptor is a critical determinant for influenza virus attachment and entry (17). AIVs typically show binding affinity for the SA–α2,3-Gal receptor; however, many naturally occurring avian H9N2 viruses have acquired the ability to preferentially bind to the SA–α2,6-Gal receptor (13). HA 226L confers an SA–α2,6-Gal receptor-binding preference, whereas HA 226Q virus prefers SA–α2,3-Gal receptor binding (28). The two HA mutations identified in the SD16-MA virus, one being the avian-like HA 226Q, are sited at the receptor-binding pocket. Direct binding assays with SA–α2,3-Gal and SA–α2,6-Gal sialylglycopolymers showed that the parental rSD16 virus and the rSD16-MA/HA virus preferentially bound the SA–α2,6-Gal receptor and SA–α2,3-Gal receptors, respectively (Fig. 5A). The reverse adaptation of HA in the rSD16-MA virus facilitated viral binding to avian-type SA–α2,3-Gal receptors, thus promoting the virus to infect avian-type receptor-enriched neuronal cells. To confirm this hypothesis, mouse neuroblastoma (N2a) cells infected with rSD16 or rSD16-MA/HA viruses (at a multiplicity of infection [MOI] of 0.1) for 16 h showed viral NP protein presence in 34% and more than 95% of cells, respectively (P < 0.05) (Fig. 5B). Thus, the rSD16-MA/HA virus appeared better adapted than rSD16 to replicate in neuronal cells.

FIG 5.

HA in mouse-adapted H9N2 virus reverted back to avian-like SA-α2,3 receptor-binding preference. (A) Binding affinity of inactivated viruses to SA-α2,3-linked and SA-α2,6-linked polymers. Mouse-adapted HA from rSD16-MA virus in parental SD16 virus backbone (rSD16-MA/HA virus) displayed binding affinity for avian-like SA-α2,3-linked polymers. Control A/Beijing/7/2009 (H1N1) and A/Anhui/1/2005 (H5N1) viruses selectively bound SA-α2,6 and SA-α2,3 polymers, respectively. Each data point is the mean ± SD of three independent experiments. (B) N2a cells were infected with rSD16 and rSD16-MA/HA viruses at a multiplicity of infection (MOI) of 0.1. NP protein (red) was detected by immunofluorescence at 16 hpi. Nuclei were stained with DAPI (blue). Bar, 50 μm.

Mouse-adapted PB2 promoted its nuclear import and increased viral RNP polymerase activity.

PB2 protein, synthesized in the cytoplasm, is imported into the nucleus to form part of a multiprotein RNP polymerase complex with PB1, PA, and NP (29). At 4 h postinfection (hpi), detection of nuclear import of PB2 protein from SD16-MA/PB2 virus infection of N2a cells was earlier than that of SD16 virus at 6 hpi (Fig. 6A). At 8 and 10 hpi, nuclear detection of PB2 protein in cells infected with rSD16-MA/PB2 virus was 68% and 87%, respectively; corresponding nuclear detection of PB2 from rSD16 virus infection was lower at 26% and 52%, respectively (P < 0.05) (Fig. 6B). Furthermore, Western blotting of nuclear protein extracts from time course infection of N2a cells clearly showed earlier and greater PB2 protein accumulation in the nucleus with rSD16-MA/PB2 virus infection than that with parental rSD16 virus infection (P < 0.05) (Fig. 6C). Thus, the PB2 gene of rSD16-MA/PB2 virus conferred earlier and higher nuclear accumulation of its PB2 protein relative to that of the parental PB2 gene from the rSD16 virus.

FIG 6.

PB2 from mouse-adapted H9N2 virus in rSD16 backbone (rSD16-MA/PB2 virus) conferred increased PB2 protein production and its nuclear import. (A) Ten-hour infection time course nuclear localization of PB2 protein (red) in N2a cells infected separately with parental rSD16 virus and rSD16-MA/PB2 virus, each at 2.0 MOI. Nuclei were stained with DAPI (blue). Bar, 20 μm. (B) Relative quantification of PB2 protein nuclear localization. Cells (n = 100; blue nuclei) were randomly selected from multiple microscopic fields to determine the presence of intranuclear PB2 (red). Data are presented as means ± SD of three independent experiments (*, P < 0.05; determined by ANOVA). (C) Western blotting of nuclear extracts from correspondingly infected N2a cells to detect nuclear PB2 protein. Proliferating cell nuclear antigen (PCNA) immunodetection demonstrated normalized protein loading of each sample.

Viral RNP assays were performed in 293T cells to compare the PB2-associated polymerase function of rSD16 virus and rSD16-MA/PB2 virus. Polymerase activity derived from rSD16-MA/PB2 virus was 37-fold higher than that from rSD16 virus (Fig. 7A). Western blotting based on corresponding cell lysates did not detect any quantitative difference between the two PB2 proteins (Fig. 7B). Thus, the higher polymerase activity associated with MA PB2 was due to increased enzymatic activity. Taken together, MA PB2 increased its nuclear import (temporal and spatial) efficiency and RNP polymerase activity, which could facilitate infection of neuronal cells.

FIG 7.

PB2 from SD16-MA (rSD16-MA/PB2 virus) increased the viral polymerase activity of parental SD16 virus in 293T cells. (A) Polymerase activity of rSD16 and rSD16-MA/PB2 was determined by minigenome assays. Four protein expression plasmids (PB2, PB1, PA, and NP) for RNP constitution of the respective viruses were transfected into 293T cells along with luciferase reporter plasmid pYH-Luci and internal control Renilla plasmid. Results presented are means ± SD of three independent experiments and are referenced to rSD16 activity set at 100%. (B) Western blotting of cell lysates from corresponding transfections harvested at 24 hpi showed no quantitative difference in PB2 protein between sSD16 and rSD16-MA/PB2 viruses (***, P < 0.001; determined by ANOVA).

PB2 and HA from mouse-adapted H9N2 virus individually promoted virus replication and NP protein production in neuronal cells.

The replication profiles of three neurovirulent viruses (rSD16-MA, rSD16-MA/PB2, and rSD16-MA/HA) in N2a cells were compared with the control rSD16 virus over a time course of 72 h. Throughout the monitoring period at 2, 12, 24, 36, 48, 60, and 72 hpi, all three neurovirulent viruses produced much greater progeny virus (close to 1,000-fold greater) (P < 0.001) than that of rSD16 virus (Fig. 8A). Expression of NP protein, a major component of the RNP complex (29, 30), of each neurovirulent virus at 24 hpi was also higher than that of the corresponding rSD16 control in N2a cells (Fig. 8B). Thus, rSD16-MA/HA and rSD16-MA/PB2 viruses were more replication proficient than parental rSD16 virus in N2a cells.

FIG 8.

PB2 and HA from SD16-MA virus increased expression of NP protein in N2a cells. (A) Cells were infected with rSD16, rSD16-MA, rSD16-MA/PB2, and rSD16-MA/HA viruses, each at 0.1 MOI. Supernatants of the infected cells were collected at the indicated time points for virus titration on MDCK cells; virus titers are means ± SD of three independent experiments. Indicated significance was relative to corresponding rSD16 virus infection. (B) Infected N2a cells were harvested at 24 hpi to detect NP protein by Western blotting, and β-actin served as the loading control.

FMRP is a critical host factor for mouse-adapted H9N2 virus infection of murine brain.

FMRP is an RNA-binding protein that interacts with NP protein to promote viral RNP assembly in an RNA-dependent manner (30). FMRP is abundant in “fragile X granules” in neuronal axons and presynaptic terminals, where it seems to regulate recurrent neuronal activity (31). To assess the involvement of FMRP in influenza virus replication in neuronal cells, groups of FMRP knockout (FMRP^−/−) or wild-type (WT) mice (FVB strain background), 11 mice per group, were i.n. infected with rSD16-MA virus (at 10⁶ EID₅₀/mouse). Clinical signs, mortality, and body weight loss of 5 mice from each group were monitored over 14 days. Brains and lungs were collected from 3 mice per group at 3 and 5 dpi for virus titration. FMRP^−/− and WT mice infected with rSD16-MA virus exhibited severe clinical signs, including huddling, decreased activity, and wheezing; all of the mice showed more than 25% weight loss and 100% mortality by 6 dpi (Fig. 9A). The virus replicated to comparable titers in the lungs of both genotypes (all above 4 log₁₀ EID₅₀/ml). However, in the brains of WT mice, virus titer at 5 dpi (Fig. 9B) was significantly higher (at 3.1 ± 0.3 log₁₀ EID₅₀/ml) than those in FMRP^−/− murine brains; FMRP RNA expression was correspondingly more abundant (19.6-fold higher) in WT murine brain than in lung (GAPDH was used to normalize the input samples by the comparative threshold cycle [2^−ΔΔCT] method, Fig. 9C). Thus, FMRP appeared to facilitate rSD16-MA virus replication in the murine brain but not in the lung.

FIG 9.

FMRP promoted rSD16-MA virus replication in murine brain and primary neuronal cells. FMRP^−/− and WT mice (FVB strain background) were intranasally infected with 10⁶ EID₅₀ of rSD16-MA. (A) Body weight change (percentage) presented as means ± SD of 5 mice. Mice that lost >25% of their baseline body weight were euthanized. (B) Virus titers from brains and lungs of WT and FMRP^−/− mice were determined at 3 and 5 dpi. The dashed black line indicates the lower limit of detection (10^0.75 TCID₅₀/ml). (C) Relative expression of FMRP mRNA in WT murine brain and lung. Total RNA was extracted from brain and lung tissues of a group of 3 BALB/c mice, and FMRP mRNA was quantitated by quantitative reverse transcription-PCR (qRT-PCR). GAPDH was used to normalize the input samples by the comparative threshold cycle (2^−ΔΔCT) method. Data are presented as the mean ± standard deviation of three independent experiments. (D) FMRP^+/+ and FMRP^−/− primary murine cortical neuronal cells were infected with 0.1 MOI of rSD16-MA virus. Virus titers of supernatants, collected at indicated time points, were determined and are presented as means ± SD of three independent experiments. (E) FMRP^+/+ and FMRP^−/− primary cells were infected with rSD16-MA for 6 h for fluorescence immunodetection of NP (green) and nuclear staining (DAPI, blue). Graph shows detection rate of NP-positive cells of the two genotypes. ***, P < 0.001 (determined by ANOVA). Bar, 100 μm.

Primary neuronal cortical cells derived from FMRP^+/+ and FMRP^−/− mice were infected with 0.1 MOI of rSD16-MA for 72 h. Viral titers monitored from 12 to 72 hpi consistently showed that FMRP^+/+ cells produced more progeny virus (up to 10^2.75-fold more) than FMRP^−/− cells (Fig. 9D). At 6 hpi, FMRP^+/+ cells infected with rSD16-MA (at 1.0 MOI) exhibited more abundant NP-positive cells (38.7%) than corresponding FMRP^−/− cells (3.7%), as determined by viral NP immunodetection (Fig. 9E). Taken together, FMRP promotes rSD16-MA virus replication in primary murine neurons.

FMRP and NP protein association promoted NP export from the nucleus in rSD16-MA virus-infected neuronal cells.

As NP protein was more highly expressed in N2a cells infected with rSD16-MA virus than in those infected with rSD16 virus (Fig. 8B), FMRP-NP interaction was examined by coimmunoprecipitation with an anti-NP antibody for cell lysates from both virus infection. Interaction between NP protein and FMRP could be detected with rSD16-MA virus infection but not with rSD16 virus (Fig. 10). The detectable interaction between FMRP and NP from rSD16-MA virus was associated with higher NP protein expression (Fig. 10).

FIG 10.

Physical interaction of viral NP and host FMRP in N2a cells. Cells were infected separately with rSD16 and rSD16-MA virus, each at 0.1 MOI. At 24 hpi, cell lysates were harvested for coimmunoprecipitation and Western blotting, as indicated. rSD16-MA virus exhibited stronger expression of NP protein than did SD16 virus, and its NP coimmunoprecipitated with FMRP.

Since the interaction of FMRP and NP facilitates viral RNP (vRNP) export from the nucleus (30), we assessed this functional association using a neuroblastoma cell line (SH-SY5Y) in which FMRP was stably knocked down by short hairpin RNAs (shRNAs) (Fig. 11A). WT and FMRP knockdown SH-SY5Y cells infected with rSD16-MA (at 2.0 MOI) were examined for nuclear presence of NP over 12 hpi by confocal microscopy. At 12 hpi, significantly more FMRP knockdown cells than WT cells contained NP protein in the nucleus (Fig. 11B and C). Thus, FMRP and NP association are involved in the nuclear export of NP protein in SH-SY5Y cells. Taken together, the stronger expression of NP protein from rSD16-MA virus infection promotes FMRP-NP interaction, which in turn facilitates NP export from the nuclei of infected neuronal cells.

FIG 11.

FMRP knockdown increased viral NP retention in the nuclei of SH-SY5Y cells infected with rSD16-MA virus. (A) Protein levels of FMRP in wild-type SH-SY5Y cells (WT), negative-control SH-SY5Y cells (NC) and FMRP-stable knockdown SH-SY5Y cells (shRNA). (B) Relative nuclear localization of NP protein. At least 100 cells (blue nuclei) from randomly selected microscopic fields of infected SH-SY5Y cells (short hairpin RNA [shRNA] and WT) were scored for the presence of intranuclear NP (red) at 8, 10, and 12 hpi. Data are presented as means ± SD deviations of three independent experiments. *, P < 0.05 (determined by ANOVA). (C) NC and shRNA cells were infected with rSD16-MA virus for 12 h, followed by immunofluorescence detection of NP protein. All infections were performed at 2.0 MOI. Bar, 15 μm.

FMRP is required for replication of neurotropic H5N1 and H10N7 viruses in the murine brain.

To determine whether FMRP supports other subtypes of influenza viruses in neuronal replication, FMRP^−/− and WT mice were infected separately with two neurotropic viruses, an H5N1 virus (22) and a mouse-adapted H10N7 virus (H10N7-MA) (32) at 10⁶ EID₅₀/mouse. Brains and lungs of 3 mice from each group were collected at 3 and 5 dpi for virus isolation. H5N1 and H10N7-MA viruses replicated efficiently in WT mouse brains, with average titers of 3.1 log₁₀ EID₅₀/ml and 2.5 log₁₀ EID₅₀/ml, respectively, at 5 dpi. In FMRP^−/− mice, neither H5N1 virus nor H10N7-MA virus could be detected at 5 dpi (Fig. 12A). However, these viruses could replicate efficiently in mouse lungs, and no significant differences in virus titers (all above 5 log₁₀ EID₅₀/ml) were found between WT and FMRP^−/− groups (P > 0.05), indicating that the FMRP was not indispensable for AIV replication in mouse lungs (Fig. 12B). These results demonstrated that the FMRP is a necessary host factor for the replication of neurotropic AIVs in the murine brain but not in the lungs.

FIG 12.

FMRP is necessary for the replication of neurotropic H5N1 and H10N7 AIVs in murine brain but not in lungs. WT and FMRP^−/− mice, in groups of 6, were infected separately with neurotropic H5N1 and H10N7 virus at 10⁶ EID₅₀ per mouse. Virus isolation from brains (A) and lungs (B) of 3 mice from both groups was performed at 3 and 5 dpi.

DISCUSSION

AIVs are zoonotic viruses that exhibit a certain degree of infectivity and severity in the human host. Severe human cases of AIV infection are sometimes accompanied by neurological symptoms. In this study, we discovered that the avian-like SA–α2,3-Gal receptor is highly represented in mammalian (human and mouse) brains, and, through generation of a mouse-adapted neurotropic H9N2 AIVs, identified key adaptive mutations in HA and PB2 genes that conferred viral infection ability in mouse brain. We further established that host FMRP protein, which is highly expressed in the brain tissue and physically associated with viral NP protein in the assembly and export of RNP complex, was a necessary host factor for neurotropic AIVs (H9N2, H5N1, and H10N7 viruses) to undertake neuronal replication.

Influenza virus neurovirulence, characterized by the ability to gain entry and subsequent replication in the CNS, could be found in some influenza virus-infected cases with severe illness (33–35). However, little is known about the strategy of AIV infection in mouse brain. Here, we generated a neurovirulent H9N2 influenza (SD16-MA) virus through repeated passaging in mice. SD16-MA virus showed binding affinity for the SA–α2,3-Gal receptor, a reversal from the parental virus with a SA–α2,6-Gal receptor-binding preference, and was more replicative than parental SD16 virus in neuronal N2a cells. The two HA mutations (R211K and L226Q) of SD16-MA virus are located around the HA receptor-binding pocket, and 226Q is a known critical determinant for avian-like SA–α2,3-Gal receptor binding (26). We also showed that another reassortant virus, rSD16-MA/PB2, without the binding affinity for the SA–α2,3-Gal receptor, replicated effectively in the murine brain. Thus, SA–α2,3-Gal receptor-binding specificity appears to facilitate viral entry but is not indispensable during SD16-MA virus infection in the murine brain.

The three mutations (M147L, V250G, and E627K) identified in PB2 of the SD16-MA virus are associated with increased RNP activity and promotion of PB2 protein production and its nuclear import, culminating in increased progeny virus production from infected neuronal cells. PB2-E627K is a well-characterized PB2 mutation that mediates increased polymerase activity, replication, and pathogenicity in mammals (36–38). E627K is frequently found in avian H5N1 and H7N9 virus strains isolated from humans (36–38) and in H9N2 viruses from infected mice (13). A role of PB2 627K in infection of the mouse brain was also demonstrated in H5N1 influenza virus (24). As residues at position 147 and 250 are in the PB2-NP-binding and cap-binding domains, they could functionally affect polymerase function and vRNP assembly. In summary, HA and PB2 genes from mouse-adapted H9N2 (SD16-MA) virus, when separately introduced into a parental SD16 virus backbone, can facilitate viral replication in the mouse brain.

In addition to the identification of HA and PB2 mutations that are responsible for the replication ability in the mouse brain of mouse-adapted avian H9N2 (SD16-MA) virus, we identified FMRP as an essential host factor that mediates the replication in mouse brain of neurotropic H9N2, H5N1, and H10N7 influenza viruses. FMRP forms part of a large RNP complex that is involved in the transport and translation of mRNA in neurons (39). It was previously shown that FMRP stimulated replication of human influenza A/PR/8/34 (PR8) virus in the upper respiratory tract of mice through RNA-mediated interaction with NP protein (30). However, FMRP is not necessary for supporting SD16-MA virus replication in the mouse lung. The absence or specific mutation(s) of FMRP leads to fragile X syndrome and causes inherited mental retardation and autism (40). Clinically, the frequency and severity of influenza virus infection in individuals with FMR1 mutations should be paid more attention.

In summary, the adaptive mutations of HA and PB2 that affect host receptor affinity, enhance viral polymerase activity, and facilitate nucleocytoplasmic shuttling of viral proteins are the key changes needed by AIVs to achieve replication ability in the mouse brain, which in turn relies on the interaction between host FMRP and viral NP protein to affect viral replication in neuronal cells.

MATERIALS AND METHODS

Ethics statement.

All animal work was approved by the Beijing Association for Science and Technology (approval identifier SYXK [Beijing] 2007-0023) and conducted in accordance with the Beijing Laboratory Animal Welfare and Ethics guidelines, as issued by the Beijing Administration Committee of Laboratory Animals, and in accordance with the China Agricultural University (CAU) Institutional Animal Care and Use Committee guidelines (identifier SKLAB-B-2010-003).

Cells and viruses.

Human embryonic kidney (293T) cells, mouse N2a cells, and human neuroblastoma (SH-SY5Y) cells were maintained in Dulbecco’s modified Eagle’s medium (DMEM; Life Technologies) supplemented with 10% fetal bovine serum (FBS; Life Technologies), 100 U/ml of penicillin, and 100 μg/ml of streptomycin.

The influenza viruses of WT (parental) H9N2 [A/chicken/Shandong/16/05 (SD16)], H5N1 (A/chicken/Sheny/0606/2008), and H10N7-MA (mouse adapted A/mallard/Beijing/27/2011) were previously described (22, 32, 41). Virus titers were measured by 50% tissue culture infectious dose (TCID₅₀) assay in MDCK cells or by EID₅₀ assay in eggs (42). All experiments with H5 subtype viruses were performed in biosafety level 3 containment.

Isolation and cultural of murine primary neuron cortical cells.

Whole cerebral cortices were removed from FVB neonatal mice (1 to 2 days old), taking care to discard the hippocampal formation, basal ganglia, and most of the meninges and vessels. The tissue was minced, incubated in 2 mg/ml papain supplement (Sigma) with 0.05 mg/ml DNase (Sigma) for 45 to 60 min at 37°C, dissociated by trituration, and plated as a single-cell suspension on a lysine (Sigma)-treated 6-well plate (3 × 10⁶ cells/well) in a plating medium of DMEM-nutrient mixture F-12 (DMEM/F-12; Gibco) supplemented with 10% FBS, 0.5% penicillin-streptomycin solution (Thermo Fisher Scientific), and 1% B27 supplement (Gibco). The plates were maintained at 37°C in a humidified 5% CO₂ atmosphere. After 24 h of in vitro incubation, the plating medium was replaced with neurobasal-A cell culture medium supplemented (Thermo Fisher Scientific) with 0.5 mmol/liter l-glutamine (Thermo Fisher Scientific), 0.5% penicillin-streptomycin solution, and 1% B27 supplement. After 48 h, the old cell culture medium was replaced with a cell culture medium containing 10 μmol/liter cytosine arabinoside (MedChemExpress). Subsequent medium replacement was carried out every 48 h. At 6 to 7 days after isolation, cells could be used for infection assays.

Adaptation of H9N2 virus in mice.

Groups of three BALB/c mice (6-week-old female BALB/c; Vital River Laboratory) were lightly anesthetized with Zoletil 50 (tiletamine-zolazepam, 20 mg/g; Virbac S.A.) and inoculated i.n. with 10⁶ EID₅₀ of virus in 50 μl phosphate-buffered saline (PBS; Gibco). At 3 dpi, three inoculated mice were euthanized, the lungs were harvested and homogenized in 2 ml of sterile, cold PBS, and 50 μl of supernatant from the centrifuged homogenate was used as inoculum for the next passage. After 13 passages, the virus could be isolated in the brain of infected mice. The virus isolated from the brain was cloned three times by plaque purification in MDCK cells, and the cloned virus was passaged once in the allantoic cavities of 10-day-old embryonated chicken eggs at 37°C for 72 h to generate virus stock.

Sequence analysis.

The virus present in the brain from the passage 13 virus-infected mice was plaque purified three times in MDCK cells. Thirty clones were chosen randomly for sequencing. Viral RNA was extracted from the allantoic fluid of the 30 clone-inoculated eggs, and the eight viral genes of each clone were amplified by reverse transcription-PCR (RT-PCR). The segments were sequenced, and adaptive mutations were identified by comparing the consensus sequences of the 30 clones to the sequences of WT SD16 virus.

Plasmid construction and virus rescue.

All eight gene segments were amplified by reverse transcription-PCR from SD16 and SD16-MA viruses and cloned into the dual-promoter plasmid pHW2000. All the constructs were sequenced to confirm the mutations. rSD16, rSD16-MA/PB2, rSD16-MA/HA, rSD16-MA/PB2-HA, rSD16-MA/M1, rSD16/MA-NS1, and rSD16-MA were generated by reverse genetics. Briefly, 0.5 μg of plasmid for each gene segment was mixed and incubated with 8 μl of TransIT-LT1 reagent (Mirus Bio, USA) at 25°C for 30 min. The TransIT-LT1-DNA mixture was transfected into 70% confluent 293T cultured monolayers and incubated at 37°C with 5% CO₂. At 6 hpi, the supernatants were replaced with 2 ml of Opti-MEM containing 2 μg/ml tosylsulfonyl phenylalanyl chloromethyl ketone (TPCK)-treated trypsin (Sigma-Aldrich). At 48 hpi, the cell suspensions were harvested and inoculated into 10-day-old specific-pathogen-free (SPF) eggs and incubated at 37°C for 72 h to prepare a virus stock. Viral RNA was extracted and amplified by RT-PCR, and each viral segment was sequenced to confirm the sequence identity.

Virus titration and replication kinetics.

TCID₅₀ was determined in MDCK cells with 10-fold serially diluted viruses inoculated at 37°C for 72 h. The TCID₅₀ value was calculated by the Reed-Muench method. Multistep replication kinetics were determined by inoculating N2a and murine primary cortical neuronal cells with viruses at an MOI of 0.1. After 1 h of incubation at 37°C, the cells were washed twice and further incubated in serum-free DMEM containing 1 μg/ml TPCK-trypsin (Sigma). Supernatants were collected at 2, 12, 24, 36, 48, 60, and 72 hpi and titers were determined on MDCK cells. Three independent experiments were performed.

Mouse experiments.

Six-week-old male WT and FMR1 gene knockout (FMRP^−/−) mice in the FVB.129P2 (B6)-Fmr1tm1Cgr/J strain background were bought from the Chinese Academy of Medical Sciences. Mice were genotyped, and the absence or presence of the Fmr1 gene was confirmed by PCR by the Chinese Academy of Medical Sciences.

Groups of 6-week-old female BALB/c mice were anesthetized with Zoletil 50 (Virbac S.A) and inoculated i.n. with 10⁶ EID₅₀ of virus in 50 μl PBS. Three mice in each group were euthanized at 3 and 5 dpi. Lungs and brains were collected for virus titration and histology assay. The remaining five mice in each group were monitored for weight loss and mortality for 14 days. Mice that lost more than 25% of their body weight were humanely euthanized.

Western blotting.

Total cellular proteins were extracted from transfected 293T cells or infected N2a cells with RIPA lysis buffer, and total protein concentration was determined by a bicinchoninic acid (BCA) protein assay kit (Beyotime). Cellular proteins were separated by 12% sodium dodecyl sulfate-polyacrylamide gel electrophoresis (SDS-PAGE) and transferred to a polyvinylidene difluoride (PVDF) membrane (Bio-Rad). Each PVDF membrane was blocked with 0.1% Tween 20 and 5% nonfat dry milk in PBS and subsequently incubated with a primary antibody. Primary antibodies were specific for influenza A virus PB2 (diluted 1:1,000; GenScript, China), NP (diluted 1:1,000; Abcam), and host protein FMRP (diluted 1:1,000; Abcam). The secondary antibodies used was horseradish peroxidase (HRP)-conjugated anti-rabbit or anti-mouse antibody (diluted 1:10,000; Beyotime). HRP presence was detected using a Western Lightning chemiluminescence kit (Amersham Pharmacia), following the manufacturer’s protocols.

Viral ribonucleoprotein polymerase assay.

RNP polymerase (minigenome luciferase) assays were based on the cotransfection of four pcDNA3.1 expression plasmids housing PB1, PA, and NP with PB2 or PB2-MA into human 293T cells (125 ng of each plasmid), together with a pYH-NS1-Luci plasmid expressing a reporter firefly luciferase gene under the control of the human RNA polymerase I promoter (10 ng), and an internal control plasmid expressing Renilla (2.5 ng). Cultures were incubated at 37°C. After 24 h of transfection, cell lysates were prepared with the Dual-Luciferase Reporter assay system (Promega), and luciferase activity was determined in a GloMax 96 microplate luminometer (Promega). LARII (100 μl) was dispensed to each well to measure firefly luciferase activity, and then100 μl of Stop & Glo reagent was dispensed to measure Renilla luciferase activity. The final polymerase activity was determined as firefly luciferase activity divided by Renilla luciferase activity.

RNA isolation and quantitative RT-PCR.

Groups of 6-week-old female BALB/c mice were anesthetized with Zoletil 50 (Virbac S.A), and brains and lungs from 3 mice were collected. Total RNA was isolated from each collected tissue with RNA isolation reagent (Thermo Fisher) following the instruction of the manufacturer. The RNAs were reverse transcribed into cDNA using the TransScript RT reagent kit (TransGen). Oligo(dT) primers were used for detecting the FMRP gene. The obtained cDNA was amplified by a fast two-step amplification using FastStart Universal SYBR green mastermix (Roche, China). GAPDH was used to normalize the input samples by the 2^−ΔΔCT method. For detection of FMRP and GAPDH, 5′-GAGATCGTGGACAAGTCAGGAG-3′ (FMRP forward), 5′-CTTCAGAGGAGTTAGGTCCAACC-3′ (FMRP reverse), 5′-ACAACTTTGGCATTGTGGAA-3′ (GAPDH forward), and 5′-GATGCAGGGATGATGTTCTG-3′ (GAPDH reverse) primers were used in this study.

Coimmunoprecipitation assay.

N2a cells were infected separately with 0.1 MOI of rSD16 virus and rSD16-MA virus. At 24 h later, cells were washed with cold PBS and lysed in RIPA buffer (Beyotime). The lysates were incubated with anti-NP (diluted 1:250, Abcam) antibody at 4°C for 16 h, and protein G Plus-agarose (Santa Cruz) was then added and rotated at 4°C for 6 h. The beads were washed 6 or 7 times with lysis buffer, and the bound proteins were separated by SDS-PAGE. followed by Western blotting with the indicated antibody.

Lectin histochemistry.

Each mouse was perfused transcardially with 10 to 20 ml of PBS followed by 20 ml of freshly prepared 4% paraformaldehyde. The brains were removed and postfixed in 4% paraformaldehyde at room temperature for more than 24 h and less than 48 h. Both sagittal and transverse sections of the brains were prepared. For detection of host influenza receptors in the tissues, the organs were sectioned at 4-μm thickness. The human brain tissue sections were provided by Beijing Longmaidasi Technology Development. Detection details of host influenza receptors are found in Kuchipudi et al. (43). Briefly, sections were presoaked in Tris-buffered saline (TBS) and blocked using a biotin-streptavidin blocking kit (Vector Laboratories) according to the manufacturer’s instructions, followed by overnight incubation at 4°C with biotinylated SNA (Vector Laboratories) or fluorescein isothiocyanate (FITC)-labeled MAL I (Vector Laboratories), each at a concentration of 10 μg/ml. After three washes with TBS, the sections were incubated with streptavidin-Alexa Fluor 594 conjugate (Invitrogen) for 2 h at room temperature. The sections were washed and mounted with ProLong Gold antifade reagent with 4′,6-diamidino-2-phenylindole (DAPI; Invitrogen).

Virus detection by immunofluorescence.

N2a cells were grown on the carry sheet glass in 24-well plates and infected with the indicated viruses. At the specified time points postinfection, the cells were fixed with 4% paraformaldehyde in PBS for 30 min and permeabilized with 0.5% Triton X-100 in PBS for 30 min. After blocking with 5% bovine serum albumin (BSA) in PBS, the cells were incubated with antiserum against PB2 (diluted 1:500; GeneTex) or NP (diluted 1:500; Abcam) at 4°C for 12 h. The cells were then washed three times with PBS and incubated with goat anti-rabbit (FITC, diluted 1:500; Abcam) or goat anti-rabbit IgG (Alexa Fluor 594, diluted 1:500; Abcam) secondary antibodies for 1 h at 37°C. The cells were subsequently washed three times with PBS and incubated with DAPI for 10 min. Cells were imaged with a laser scanning confocal microscope (Leica). The frequency of nuclear localization of the PB2 protein was determined by cell counting (n = 100).

RNA interference.

Cells were transfected with small interfering RNAs (siRNAs) at 50 nM for indicated times. The following sequences were targeted for FMR1 (5′ to 3′): no. 1, 5′-CCAAAGAGGCGGCACAUAA-3′; no. 2, 5′-AAAGCUAUGUGACUGAUGA-3′; and no. 3, 5′-CAGCUUGCCUCGAGAUUUC-3′. Lentivirus expressing FMRP-specific short hairpin RNA was generated by the GenePharma Company (Shanghai). Briefly, two complementary oligonucleotides with BamHI and EcoRI endonuclease sites at each end were synthesized, annealed, and cloned into an HIV-based lentiviral expression vector (LV3-pGLV-H1-GFP/PURO; GenePharma, Shanghai) to express a hairpin transcript (5′-GCAGCTTGCCTCGAGATATCT-CAAGAGGATATCTCGAGGCAAGCTGC TT-3′). The lentiviral particles were then produced by cotransfecting the short hairpin RNA expression plasmids with packaging plasmids into 293 packaging cells. After 72 h, viruses were collected and titers were determined. To generate FMRP-stable knockdown or control cell lines, SH-SY5Ycells were infected with the lentiviral particles and selected with puromycin (1 mg/ml) for 3 weeks.

Statistical analyses.

All statistical analyses were performed using Prism software version 5.00 (GraphPad Software, Inc.). Statistically significant differences between experimental groups were determined using Dunnett’s test following one-way analysis of variance (ANOVA). Differences were considered statistically significant at P < 0.05.

Accession number(s).

The nucleotide sequences of the eight gene segments of influenza viruses of WT H9N2 SD16 are available from GenBank under accession numbers MW397163 to MW397170. The nucleotide sequences of the eight gene segments of mouse-adapted H9N2 influenza virus SD16-MA are available from GenBank under accession numbers MW397155 to MW397162.