Mutations in the M-Gene Segment Can Substantially Increase Replication Efficiency of NS1 Deletion Influenza A Virus in MDCK Cells

Abstract

Influenza viruses unable to express NS1 protein (delNS1) replicate poorly and induce large amounts of interferon (IFN). They are therefore considered candidate viruses for live-attenuated influenza vaccines. Their attenuated replication is generally assumed to result from the inability to counter the antiviral host response, as delNS1 viruses replicate efficiently in Vero cells, which lack IFN expression. In this study, delNS1 virus was parallel passaged on IFN-competent MDCK cells, which resulted in two strains that were able to replicate to high virus titers in MDCK cells due to adaptive mutations especially in the M-gene segment but also in the NP and NS gene segments. Most notable were clustered U-to-C mutations in the M segment of both strains and clustered A-to-G mutations in the NS segment of one strain, which presumably resulted from host cell-mediated RNA editing. The M segment mutations in both strains changed the ratio of M1 to M2 expression, probably by affecting splicing efficiency. In one virus, 2 amino acid substitutions in M1 additionally enhanced virus replication, possibly through changes in the M1 distribution between the nucleus and the cytoplasm. Both adapted viruses induced levels of IFN equal to that of the original delNS1 virus. These results show that the increased replication of the adapted viruses is not primarily due to altered IFN induction but rather is related to changes in M1 expression or localization. The mutations identified in this paper may be used to enhance delNS1 virus replication for vaccine production.

INTRODUCTION

The nonstructural (NS1) protein of influenza A virus is an antagonist of the cellular antiviral response. Infection with virus either not encoding NS1 protein (delNS1) or encoding a truncated NS1 protein results in high levels of type I interferons (IFN) such as IFN-α or IFN-β. Replication of such viruses is attenuated in IFN-competent cell lines, indicating that the NS1 protein is not essential for replication in such hosts (14). In vivo, viruses lacking a fully functional NS1 protein induce IFN in the absence of detectable virus replication (12), which are favorable conditions for use as live attenuated vaccines. The local release of IFN and other cytokines and chemokines appears to be an excellent adjuvant that enhances production of immunoglobulins and contributes to the activation of dendritic cells required for antigen presentation (22, 31). DelNS1 candidate vaccines against influenza A and B viruses have been developed (35, 51), and initial trials in humans showed successful induction of antibody responses (48). Apart from the use in vaccines, delNS1 viruses also show potential as oncolytic agents (27) and viral expression vectors (50).

NS1 is expressed at high levels directly after infection and facilitates virus replication in many different ways (reviewed in reference 15). Its antiviral properties are focused on reducing the IFN-mediated innate immune response and act at several levels. Cytoplasmic double-stranded RNA (dsRNA) and 5′-triphosphate-containing RNA are produced during influenza infection and recognized as pathogenic patterns by antiviral proteins like retinoic-acid inducible gene I (RIG-I), dsRNA-dependent protein kinase R (PKR), and 2′-5′-oligoadenylate synthetase (OAS). NS1 binds both dsRNA and RIG-I and blocks the activation of PKR and OAS, thereby limiting the onset of several pathways that lead to IFN induction (31). Other functions of NS1 are inhibition of cellular pre-mRNA processing (including IFN pre-mRNA) and mRNA nuclear export (15). Furthermore, NS1 regulates both viral genome replication and translation (49), splicing of M segment mRNA (32), nuclear export of viral mRNA (13) and viral ribonucleoprotein (vRNP) (49), and viral protein synthesis (7, 10). Recently it was found that NS1 binds the human PAF1 transcription elongation complex (hPAF1C) by a histone-mimicking sequence, thereby inhibiting the role of hPAF1C in the antiviral response (26). NS1 mRNA is transcribed from the eighth vRNA segment. It is partially spliced to generate mRNA that encodes the nuclear export protein (NEP) (15). In the nuclei of infected cells, NEP facilitates the export of the vRNP complexes containing the viral genome segments to the cytoplasm, where assembly of the viral components is completed before virus budding takes place (1). Independently from vRNP export, NEP also regulates viral genome transcription and replication (33). During development of delNS1 strains it is therefore essential to retain the NEP protein.

In cells and animals with a low or absent IFN response, such as Vero cells or STAT1 or PKR knockout mice, delNS1 virus replicates to high titers (8, 14, 20), whereas replication is attenuated in MDCK cells and other IFN-competent hosts. However, when Vero cells are externally stimulated with IFN-α before infection, delNS1 virus replication is also attenuated (12). Moreover, delNS1 replicates efficiently only in embryonated chicken eggs younger than 8 days, when the host immune response is not yet fully developed (43). It is therefore generally assumed that the inability of delNS1 virus to counter the cellular innate immune response is the major cause for its attenuated phenotype (31). In addition to the unimpaired IFN response, the absence of NS1 during influenza virus infection results in enhanced apoptosis induction (41, 57). Activation of caspases, a group of cysteine proteases that play an important role in apoptosis, results in cleavage of viral NP protein and thereby limits the amount of viral protein available for assembly of viral particles (56). The antiviral effect of apoptosis is therefore believed to contribute to the attenuated replication of delNS1 virus. Inhibition of the apoptotic response is attributed to the activation by NS1 of the phosphatidylinositol 3-kinase (PI3K)-Akt signaling pathway, which is known to result in an anti-apoptotic response (9, 55), as well as the inhibition of IFN. IFN sensitizes cells for apoptosis (57) through its transcriptional induction of PKR (44) and activation of the FADD/caspase-8 death signaling pathway (4). The role of apoptosis in influenza infection is, however, still uncertain, as several influenza proteins, including NS1, also exhibit proapoptotic functions. Furthermore, influenza virus replication is impaired in the presence of caspase inhibitors (52), which appears to be caused by retention of vRNA complexes in the nucleus, preventing formation of progeny virus particles. A possible explanation for this double role of NS1 in apoptosis regulation could be prevention of cell death by inhibition of apoptosis early in the infection, followed by induction at a later stage (55). Ludwig et al. (24) suggested that caspases enhance vRNP export from the nucleus later in the infection by widening of the nuclear pores, thereby allowing diffusion of vRNP out of the nucleus.

Previously we showed that delNS1 virus can efficiently be propagated on a MDCK cell line showing inducible expression of NS1 from a trans-complementing genomic gene (46). A 500-fold increase in infectious virus titer was observed, even though the NS1 level was 1,000-fold lower than that in cells infected with wild-type (WT) virus. Furthermore, apoptosis was reduced to levels similar to those found in WT virus-infected cells, whereas the induction of IFN by delNS1 virus was not significantly reduced in these cells. Because of the limited effect on IFN induction, we then hypothesized that the low yield of delNS1 virus on normal MDCK cells could be caused by loss of another NS1 regulatory function rather than the inability of the virus to interfere with the host cells' antiviral response. In this work, we increased the replication efficiency of delNS1 virus by adaptation to IFN-competent MDCK cells during serial passage. Next, we determined if the observed increase in virus yield was related to decreased IFN and apoptosis induction. Furthermore, we identified the mutations and partly characterized the mechanism that allowed the virus to efficiently replicate in the absence of the NS1 protein.

MATERIALS AND METHODS

Cell culture and virus strains.

MDCK-SFS (serum-free suspension) cells (47) were grown in suspension in SFM4BHK21 medium (HyClone, Waltham, MA), supplemented with 8 mM glutamine, 5 mg/liter phenol red, and 1.5 g/liter sodium bicarbonate, or were grown adherent in serum-free UltraMDCK medium (Lonza Biowhittaker, Basel, Switzerland) supplemented with 4 mM glutamine. Adherent NS1Bon2 MDCK cells (46) were also grown in UltraMDCK medium, additionally supplemented with 200 μg/ml G418 (Promega, Fitchburg, WI) and 100 μg/ml hygromycin B (Clontech, Mountain View, CA). G418 and hygromycin were not used during virus infections. Vero and human embryo kidney (293T) cells were cultured in Glutamax medium (Invitrogen, Carlsbad, CA) supplemented with 10% fetal bovine serum (FBS). All culture media were provided with 100 units/ml penicillin and 100 μg/ml streptomycin (Gibco). Cells were grown at 37°C and 5% CO₂. Suspension cells were grown in shaker flasks at 100 rpm. Cell density and viability were determined with a Countess automated cell counter (Invitrogen).

All virus strains described in this paper are based on the A/PR/8/34 (H1N1) (PR8) strain in which the HA and NA genes are replaced by those from A/turkey/Turkey/1/05 H5N1 (46). The multibasic cleavage site of H5 was replaced by that of a low pathogenic H6 subtype (46) and we refer to this HA gene segment as H5(6). This H5(6)N1 virus (46) strain containing the complete NS segment was passaged once in 9-day-old embryonated eggs after virus rescue and is referred to as WT virus. The influenza delNS1 virus strain (46) used for the adaptation to MDCK-SFS cells is isogenic to WT virus except for the NS gene segment. It was previously passaged 10 times in 7-day-old embryonated chicken eggs and is referred to as delNS1^EA. Note that in the comparison of reassortant virus replication, protein expression, apoptosis, and IFN induction, a nonadapted delNS1 virus strain was used, to which we refer as delNS1. The infectious influenza virus titer was measured by determining the tissue culture infective dose required to infect 50% (TCID₅₀) of MDCK cells, as previously described (47). All virus strains were propagated on the NS1-expressing NS1Bon2 MDCK cell line to generate virus seed stocks with high infectious virus titers (>7 log¹⁰ TCID₅₀/ml) prior to further viral characterization. For this purpose, NS1 expression was induced in this cell line 24 h before infection by the addition of 1 μg/ml doxycycline (Clontech) to the culture medium, followed by infection at multiplicity of infection (MOI) of 0.01 and harvesting at 3 days postinfection (dpi).

Virus adaptation.

Two independent adaptation experiments were performed. In the first experiment, MDCK-SFS cells in suspension were infected with delNS1^EA at an MOI of 0.1. After 2 to 3 days the supernatant was collected and a 500-fold dilution was used for subsequent infection of fresh MDCK-SFS cells (unknown MOI). The virus was serially passaged 10 times in this manner. In the second adaptation experiment, 5 serial passages were performed, starting with infection of adherent MDCK-SFS cells with delNS1^EA at an MOI of 0.01. The infectious virus titer was determined daily and the supernatant with the highest titer was used in subsequent infection of fresh MDCK-SFS cells at an MOI of 0.01. The two adapted virus strains were cloned 3 times by limiting dilution on MDCK-SFS cells. Of each adapted strain, eight clones were screened for virus replication on MDCK-SFS cells and one clone of each strain with high titer was selected and amplified on NS1Bon2 cells. The resulting strains are referred to as delNS1^CA1 and delNS1^CA2, respectively.

Plasmids.

Ten plasmids containing single or multiple mutations found in delNS1^CA1 and delNS1^CA2 (Table 1) were made. Viral RNA of delNS1^CA1 and delNS1^CA2 was isolated from seed virus, and HA and M gene segments were amplified by PCR, using primers with BsmBI restriction sites (18). The resulting cDNAs were inserted in plasmid pHW2000 to create pROM33-pROM36. Plasmids containing the mutated PB1 segments of delNS1^CA1 and delNS1^CA2 could not be made. pROM16 was made by cloning a synthetic 675-bp BsrGI-NgoMIV fragment (GenScript Corporation, Piscataway, NJ) with mutation A1381G in pHW195 containing the PR8 NP gene (17). Synthetic fragments with suitable BsmBI restriction sites comprising the complete M segment with one or more of the CA2 mutations or the complete delNS1 NS segment with all six CA1 mutations were inserted in pHW2000 to construct pROM13 and pROM51-pROM54. All plasmid inserts were sequenced to ensure the absence of additional nucleotide substitutions.

Table 1.

pHW2000-derived plasmids encoding mutant gene segments from cell-adapted delNS1 viruses

Plasmid	Segment	Nucleotide substitution(s)	Amino acid substitution(s)a
pROM34	HACA1	U796C	F257L
pROM33	HACA2	C1326A	None
pROM16	NPCA1	A1381G	R446G
pROM36	MCA1	U640C, U643C, U652C, U688C	None
pROM35	MCA2	U277C, U298C, U315C, U316C, U323C, U325C	V97A, Y100H
pROM51	MCA2.1	U315C	V97A
pROM52	MCA2.2	U323C	Y100H
pROM53	MCA2.3	U315C, U323C	V97A, Y100H
pROM54	MCA2.4	U277C, U298C, U316C, U325C	None
pROM13	NSCA1	A148G, A173G, A179G, A180G, A248G, A252G	Y41C, M52V, I76V

^aAmino acid positions are relative to methionine in the open reading frame.

Rescue of recombinant influenza virus.

To generate recombinant influenza virus, a mixture of 1.5 × 10⁶ 293T and 5 × 10⁵ MDCK-SFS cells was transfected with equal amounts of the eight plasmids containing the different gene segments, using Fugene HD (Roche Applied Science, Penzberg, Germany). DelNS1 viruses were made using plasmids pHW191 (GenBank accession no. of vRNA coding region, AB671295), pHW192 (EF467819), pHW193 (CY058518), pHW195 (EF467822), pHW197 (EF190985) (17), pPolsaprib H5, pPolsaprib N1, and pHW NEP (46), whereas WT virus was made by replacing pHW NEP with pHW198, which contains the full-length NS segment. All other recombinant viruses were made by replacing one or more plasmids with those in Table 1. At 24 h posttransfection (hpt) the transfection mixture was replaced by Glutamax medium supplemented with 0.3% bovine serum albumin (Chemie Brunschwig AG, Basel, Switzerland) and 1 μg/ml trypsin-TPCK (tosylsulfonyl phenylalanyl chloromethyl ketone) (Sigma-Aldrich, St. Louis, MO). Virus was harvested at 96 hpt. Virus stocks were generated using NS1Bon2 cells as described above, and virus identity was confirmed by sequence analysis. Reassortant delNS1 virus strains are referred to as delNS1, followed by the mutated gene segment(s) it contains (Table 1) and the adapted virus from which they originate (e.g., delNS1:[NP M]^CA1, which contains the mutated NP and M gene segments from delNS1^CA1).

Virus genome sequencing.

Viral RNA genomes were isolated using a High Pure viral RNA isolation kit (Roche Applied Science). Universal influenza genome primer uni12 (18) was used for reverse transcriptase reactions with the Superscript III first strand synthesis system (Invitrogen), followed by segment-specific PCRs with an Expanded high-fidelity PCR system (Roche Applied Science). DNA sequencing was performed at Baseclear (Leiden, Netherlands), and sequence analysis was done with Lasergene (DNASTAR Inc., WI).

Comparison of virus replication.

To study the infection kinetics of the different virus strains, 10⁶ MDCK-SFS or Vero cells per well were incubated in 6-well plates, in 5 ml UltraMDCK medium containing 2 μg/ml trypsin-TPCK. Cells were infected in triplicate at an MOI of 0.01. Because of their ability to inactivate trypsin (19), Vero cells were supplied with additional trypsin-TPCK (1 μg/ml) at 24 and 48 h postinfection (hpi). Supernatant was sampled at the indicated intervals and stored at −80°C before determining the infectious virus titer.

The main effect and interactions of the mutated gene segments on the infectious virus titer were analyzed by a repeated-measures analysis of variance (ANOVA) with data from two infection experiments, both performed in triplicate, using the R statistical software package (R Foundation for Statistical Computing, Vienna, Austria). Analysis was performed separately for delNS1^CA1 and delNS1^CA2 reassortant data sets, with the mutated gene segments as explanatory variables, the two infection experiments as random effects, and delNS1 virus as the baseline. Nonsignificant explanatory variables were excluded from the model.

IFN reporter assay.

MDCK-SFS cells were allowed to attach to the surface of eight 96-well plates for 1 h (4.5 × 10⁴ cells/well) and then transiently cotransfected with a reporter plasmid carrying a firefly luciferase gene under the control of the IFN-β promoter (p125Luc, kindly provided by Takashi Fujita, Kyoto University, Japan) (54) and a Renilla luciferase control plasmid pGL4.73 (Promega), using Fugene HD. The next day, supernatant was removed and cells were infected in triplicate (MOI, 5) with WT, delNS1, delNS1^CA1, delNS1^CA2, delNS1:M^CA1, delNS1:M^CA2, or delNS1:M^CA2.3 or were mock infected. One hour later, supernatant was replaced with fresh medium. At 9, 12, 16, 20, and 24 hpi, one plate was stored at −20°C without supernatant. The firefly luciferase activity of all plates was measured with a GloMax-Multi luminometer (Promega) using the dual luciferase reporter assay system (Promega) and normalized to the Renilla luciferase activity.

Apoptosis assay.

MDCK-SFS cells were allowed to attach to the surface of six 96-well plates for 1 h (10⁴ cells/well) and infected in sextuple with WT, delNS1, delNS1^CA1, or delNS1^CA2 virus (MOI of 5) or mock infected. One hour later, supernatant was replaced with fresh medium. At 10, 14, 16, 19, 22, and 26 hpi, one plate was stored at −20°C without supernatant. Apoptosis was determined by the activity of caspase-3 and caspase-7 using the Caspase-Glo 3/7 assay (Promega). Caspase-Glo reagent, diluted 1:1 with phosphate-buffered saline (PBS), was added to the frozen cells (50 μl/well). After 1.5 h of incubation at room temperature, the luminescence was measured with a GloMax-Multi luminometer.

Western blot analysis of M1 and M2 expression.

MDCK-SFS cells were allowed to attach to the surface of 24-well plates for 1 h (3.3 × 10⁵ cells/well) and infected (MOI of 5) or mock infected in triplicate. HEK293T cells (10⁶ cells/well in 6-well plates) were transiently transfected in triplicate, using Fugene HD with 2 μg/well either pHW195, pROM35, pROM36, pROM51, pROM52, pROM53, pROM54 (Table 1), or mock transfected. At 10 hpi or 48 hpt, cells were lysed with reducing NuPAGE sample buffer containing, in addition, complete EDTA-free protease inhibitor cocktail (Roche). Samples were sheared with a 21G needle, incubated for 10 min at 75°C, and loaded onto NuPAGE Novex 12% Bis-Tris precast gels (Invitrogen). Polypeptides were transferred to polyvinylidene difluoride membranes and detected by immunoblotting using monoclonal mouse antibodies (Santa Cruz Biotechnology, Santa Cruz, CA) against M1 (SC-57881, 0.2 μg/ml) or M2 (SC-32238, 0.4 μg/ml). After subsequent incubation with peroxidase-conjugated rabbit anti-mouse immunoglobulins (0.13 μg/ml; Dako, Glostrup, Denmark), proteins were visualized with ECL plus (GE Healthcare, Buckinghamshire, United Kingdom) and quantified with a Storm840 imaging system and ImageQuant software (Molecular Dynamics, Sunnyvale, CA). The M1/M2 band intensity ratio of each sample was calculated before determining the mean ratio of the triplicate infections or transfections.

Subcellular M1 localization.

MDCK-SFS cells were cultured in suspension (6.6 × 10⁵ cells/ml) and infected with WT, delNS1, or delNS1:M^CA2.3 in triplicate (MOI, 5). At 6 and 10 hpi, 10⁶ cells were harvested and cytoplasmic and nuclear extracts were prepared with the NE-PER nuclear and cytoplasmic extraction kit (Thermo Fisher Scientific, Rockford, IL), according to the manufacturer's description. However, complete EDTA-free protease inhibitor cocktail was added to the extraction reagents. Cytoplasmic and nuclear extractions were subjected to reducing SDS-PAGE and Western blot analysis as described above. Polypeptides were detected by immunoblotting as described in the previous section with monoclonal mouse antibodies (Santa Cruz Biotechnology) against lamin A/C (SC-7292, 0.2 μg/ml), tubulin (SC-5286, 0.2 μg/ml) or M1 (SC-57881, 0.2 μg/ml). Cytoplasmic and nuclear specific proteins tubulin and lamin A/C were used to assess the purity of nuclear and cytoplasmic extracts. The amount of M1 protein was corrected for the extraction efficiency by either tubulin (cytoplasmic extracts) or lamin A/C (nuclear extracts). The ratio of corrected nuclear M1 to corrected cytoplasmic M1 was calculated before determining the mean ratio of the triplicate infections.

RESULTS

delNS1 virus adaptation.

The egg-adapted influenza virus delNS1^EA replicated poorly in MDCK cells, reaching a maximum infectious virus titer that was 10⁴-fold lower than that of WT virus (Fig. 1B). To investigate whether the virus was able to overcome the negative effect of the NS1 deletion by acquiring compensating mutations, we serially passaged delNS1^EA virus on MDCK-SFS cells in two independent adaptation experiments (see Materials and Methods). In the first cell adaptation experiment (CA1), virus was blindly passaged 10 times. In the second experiment (CA2), virus taken from the time point where the titer was maximal was used to infect cells in the next passage at a controlled MOI of 0.01. Both adaptation experiments on MDCK-SFS cells resulted in virus populations with increased replication rate and increased maximum titers compared to the parent strain. The maximum virus titer of strain delNS1^CA2 increased during the first 3 to 4 passages but did not increase further during the fifth passage (Fig. 1A). To obtain clonal virus, the adapted strains were further cultured during 3 limiting dilution steps on MDCK-SFS cells. The infectious virus titers of the resulting cloned cell-adapted virus strains delNS1^CA1 and delNS1^CA2 were about 250-fold higher than the parental delNS1^EA virus titer, but they remained 25-fold lower than the WT virus titer (Fig. 1B). In Vero cells, delNS1^EA was only weakly attenuated in comparison to the WT virus (Fig. 1C). Adaptation of delNS1 virus to MDCK-SFS cells caused only a small increase in delNS1^CA2 virus replication in Vero cells, whereas delNS1^CA1 virus replication was comparable to that of the parental delNS1 virus (Fig. 1C). Thus, the increase in virus replication due to virus adaptation is much higher during propagation on MDCK-SFS cells than during that on Vero cells.

Fig 1.

Replication of delNS1 virus adapted to growth on MDCK-SFS cells. (A) Maximum infectious virus titer during each passage step of the second delNS1 adaptation experiment. (B and C) Replication kinetics of the two adapted delNS1^CA1 and delNS1^CA2 virus strains in comparison to WT and parental delNS1^EA virus after infection of MDCK-SFS (B) or Vero (C) cells at an MOI of 0.01. Geometric mean titers and 95% confidence intervals of the means of virus infections performed in triplicate are presented.

Sequence analysis of adapted virus.

All eight gene segments of the delNS1^CA1 and delNS1^CA2 virus strains and their parental delNS1^EA strain were sequenced (Table 2). No mutations were found in the PB2, PA, and NA segments. The delNS1^EA virus contained two mutations in PB1 and one in HA with all mutations present in approximately 50% of the virus population. These mutations were silent and must have arisen during the passaging in eggs. Both PB1 mutations were present in both cell-adapted viruses, whereas the HA mutation was present only in delNS1^CA2 virus. Due to the presence of mutations in the delNS1^EA virus, a nonadapted delNS1 virus was used in the comparison of reassortant virus replication, protein expression, apoptosis, and IFN-induction experiments. We refer to this virus as delNS1.

Table 2.

Mutations in egg-adapted delNS1^EA and cell-adapted delNS1^CA1 and delNS1^CA2 virus strains

Gene segment	Nucleotide positiona	Nucleotide substitutionb			Amino acid substitutionb
		delNS1EA	delNS1CA1	delNS1CA2
PB1 (S2)	798	—	G to U	—	—
	1953	A to U 50%c	A to U	A to U	—
	2133	U to C 50%c	U to C	U to C	—
HA (S4)	796	—	U to C	—	F257L
	1326 (1338)	C to A 50%c	—	C to A	—
NP (S5)	1381	—	A to G	—	R446G
M (S7)	277	—	—	U to C	—
	298	—	—	U to C	—
	315	—	—	U to C	V97A
	316	—	—	U to C	—
	323	—	—	U to C	Y100H
	325	—	—	U to C	—
	640	—	U to C	—	—
	643	—	U to C	—	—
	652	—	U to C	—	—
	688	—	U to C	—	—
NS (S8)	148 (620)	—	A to G	—	Y41C
	173 (645)	—	A to G	—	—
	179 (651)	—	A to G	—	—
	180 (652)	—	A to G	—	M52V
	248 (720)	—	A to G	—	—
	252 (724)	—	A to G	—	I76V

^aNumbering refers to nucleotide positions in the vRNA template (GenBank accession numbers: EF467819 [PB1], DQ407519 [HA], EF467822 [NP], EF190985 [M], AF389122 [NS]). Note that the numbering of the HA and NS mutations refers to the recombinant gene segments (46). The numbering in line with the GenBank sequences is placed in parentheses.

^bSubstitutions as compared to sequence of plasmid used for initial delNS1 virus generation. —, no substitution.

^cPresence of a second nucleotide sequence within the seed virus.

Compared to delNS1^EA virus, delNS1^CA1 virus had 13 additional mutations, including a silent mutation in PB1, F257L in HA, R446G in NP, four silent U-to-C mutations in M, and six A-to-G mutations in the delNS1 segment. Three of the NS segment mutations resulted in the amino acid substitutions Y41C, M52V, and I76V in the NEP protein. The first 4 A-to-G mutations in the NS^CA1 segment are located on the part that normally encodes NS1, resulting in two amino acid substitutions at the NS1 C terminus. These mutations may be disadvantageous to the WT virus, and removal of the NS1 open reading frame (ORF) therefore increased the freedom of this segment to acquire mutations. DelNS1^CA2 virus had 6 additional mutations that were all located in the M segment. These mutations started to appear simultaneously at passage 4 and had increased at passage 5 (Fig. 2A). Again, these 6 mutations were all U-to-C mutations, two of which resulted in amino acid changes V97A and Y100H in M1. Notably, all mutations on the M and NS segments are clustered in regions of 50 to 100 nucleotides. Both M segment mutation clusters are located on a region of the M1 mRNA that is removed by splicing to generate the M2 mRNA and thus do not affect M2 mRNA structure (Fig. 2B). Apart from the two silent mutations in PB1 which were already present in the initial delNS1^EA virus, the two parallel adaptations did not lead to identical mutations.

Fig 2.

Nucleotide substitutions in M-gene segments of adapted delNS1 viruses. (A) Sequence analysis of delNS1^CA2 showing the simultaneous accumulation of six U-to-C substitutions (indicated by arrows) in the M segment vRNA between positions 277 and 325 from passages 3 to 5 on MDCK-SFS cells. Double peaks are visible at passages 4 and 5. Note that in the electropherograms U is shown as T. (B) Schematic overview of the M segment mRNAs, with the locations of the adaptive M^CA1 and M^CA2 mutations and splicing products M2 mRNA and mRNA3. Open reading frames are indicated by thick bars. (C) Nucleotide sequences of the M^CA1 and M^CA2 regions shown in panel B, including the amino acid sequence of M^CA2, where dots indicate sequence identity of M1^CA2 to M1^WT. The square box indicates the location of the NLS in M1, with positively charged (+) amino acids (53). The arrow indicates the location of the adaptive mutation in influenza B M1 protein found earlier (51). The amino acid sequence of M1^CA1 is not shown since it is identical to that of M1^WT (i.e., all mutations were silent).

Infection with delNS1 viruses containing mutated gene segments.

The effect of each mutated gene segment on the infectious virus titer and possible interactions between the gene segments were determined by employing a full-factorial analytical approach. MDCK-SFS cells were infected with reassortant delNS1 viruses consisting of all 15 possible combinations of delNS1^CA1 virus HA, NP, M, and NS gene segments or the 3 possible combinations of delNS1^CA2 virus HA and M gene segments (Fig. 3A). All reassortant viruses contained the WT PB1 segment, as we were unable to generate plasmids containing the mutated PB1 segments. Infectious virus titers were examined using a repeated-measures ANOVA to determine the relative importance of the gene segments, and possible interactions between gene segments, on virus replication. Segments with significant effects on the delNS1^CA1 virus titer were NS^CA1, M^CA1, and NP^CA1, with coefficients of, respectively, 0.68, 0.76, and 0.50 (all P < 0.001). These coefficients specify the average increase of the virus titer in log¹⁰ TCID₅₀/ml, when the mutated segment was included in the delNS1 reassortant strain. There was also an interaction effect between the mutated M and NP segments of −0.36 (P < 0.05), which indicated an average decrease in virus titer when the mutated M and NP segments were combined in the delNS1 virus. Thus, the enhanced delNS1^CA1 virus replication was the effect of these three mutated gene segments together. The enhanced replication of delNS1^CA2 was determined by the mutated M segment alone (Fig. 3A). This observation was confirmed by statistical analysis, which appointed a coefficient of 2.0 (P < 0.001) to M^CA2, indicating that the 100-fold increase in virus titer was solely determined by the M segment mutations. Reassortant viruses containing all mutated gene segments (delNS1:[HA NP M NS]^CA1 and delNS1:[HA M]^CA2) replicated as well as the viruses from which their segments originated, delNS1^CA1 and delNS1^CA2, indicating that the mutated PB1 gene segments did not contribute to the enhanced virus replication.

Fig 3.

Comparison of infectious virus titers 3 days after infection of MDCK-SFS cells with the cell-adapted or the various reassortant virus strains (MOI, 0.01). (A) Titers of delNS1 reassortant viruses made with original delNS1 plasmids (empty cells) or plasmids containing single or multiple mutations originating from the adapted virus strains delNS1^CA1 (cells marked 1) and delNS1^CA2 (cells marked 2). (B) Titers of delNS1 reassortant viruses containing one or more of the delNS1^CA2 M segment mutations (cells marked C) and of WT reassortant virus containing the M^CA2 mutations. In both panels, geometric mean titers and 95% confidence intervals of the means of triplicate measurements are presented.

Because the M segment plays a major role in increased replication of both adapted viruses, we further focused on the mechanism by which mutations in this segment could overcome the decreased replication in the absence of NS1. To determine which individual M^CA2 mutation was responsible for virus titer increase, four additional mutant virus strains were made containing either the V97A or Y100H mutation, the combination of V97A and Y100H, or the remaining four silent mutations (Fig. 3B). Compared to delNS1, the two strains with single amino acid substitutions did not replicate more efficiently. However, when V97A and Y100H were combined in delNS1:M^CA2.3, a 50-fold increase in virus titer was observed. Furthermore, the four silent mutations increased the virus yield approximately 10-fold, as indicated by the comparison of delNS1 to delNS1:M^CA2.4 and delNS1:M^CA2.3 to delNS1^CA2. Interestingly, when introduced into the WT virus, the M^CA2 segment decreased replication (Fig. 3B).

IFN-β and apoptosis induction by cell-adapted delNS1 virus.

To assess if viral adaptation affected IFN-β expression, cells were transfected with a firefly luciferase reporter gene under the control of an IFN-β promoter and subsequently infected with WT or different delNS1 virus strains at a high MOI. The low luciferase activity of cells infected with WT virus compared to cells infected with delNS1 virus indicates inhibition of IFN-β induction by NS1 (Fig. 4A). The two cell-adapted viruses (Fig. 4A) as well as delNS1 reassortant virus containing either M^CA1, M^CA2, or the M segment containing the two CA2 amino acid mutations (Fig. 4B) did not show lower IFN induction than parental delNS1 virus (Fig. 4A). This indicates that the increase in virus replication due to virus adaptation was not caused by a lower IFN induction.

Fig 4.

Effect of delNS1 virus adaptation on IFN-β and apoptosis induction. (A and B) IFN induction in MDCK-SFS cells infected with either WT, delNS1, delNS1^CA1, or delNS1^CA2 virus or mock infected (A) and with delNS1:M^CA1, delNS1:M^CA2, or delNS1:M^CA2.3 virus (B) was measured with an IFN-β dependent luciferase reporter construct and corrected for transfection efficiency with Renilla luciferase. (C) Induction of apoptosis in MDCK-SFS cells infected with either WT, delNS1, delNS1^CA1, or delNS1^CA2 virus or mock infected. Apoptosis induction was assessed by measuring the activity of caspase-3 and caspase-7. Geometric mean activities and 95% confidence intervals of the means of experiments performed in triplicate are presented.

The induction of apoptosis was determined by measuring the activity of caspase-3 and caspase-7, two proteases that are induced late in the apoptosis pathway. Again, the inhibiting effect of NS1 was visible, as little caspase activity was seen in cells infected with the WT virus in comparison to cells infected with delNS1 viruses. Both cell-adapted delNS1^CA1 and delNS1^CA2 viruses similarly induced caspase activity to a level that is far higher than observed with WT virus (Fig. 4C).

Effect of M segment mutations on M1 and M2 protein expression.

To determine if the M segment mutations affected splicing of the M1 mRNA (Fig. 2B) we measured the ratio of M1 and M2 protein expression in cells at 10 hpi by Western blot analysis (Fig. 5A). The M1 protein of all Y100H mutant viruses (delNS1:M^CA2, delNS1:M^CA2.2, and delNS1:M^CA2.3) migrated slightly more slowly in SDS-PAGE than that of the other virus strains (Fig. 5A, lanes 4, 6, and 7, respectively). Slight changes in mobility in SDS-PAGE due to amino acid changes that affect protein charge, such as Y100H, have been observed before (30).

Fig 5.

Western blot analysis of M1 and M2 protein expression 10 h after infection of MDCK-SFS cells (A) or 48 h after transient transfection of HEK293T cells (C) with plasmid (Table 1) containing either the WT M segment (pHW197), the mutated M^CA1 or M^CA2 segments (pROM36 and pROM35), or the M segment containing one or more delNS1^CA2 mutations (pROM51-54). M1 and M2 protein levels of three independent experiments were quantified using phosphorimager densitometry. The M1/M2 ratio for each virus (B) or plasmid (D) is shown as the mean and 95% confidence interval of the mean. Lanes 1 to 4 and lanes 5 to 8 in panel A originated from two different blots.

Cells infected with delNS1 virus (Fig. 5A, lane 2) appeared to express more M2 protein than those infected with WT virus-infected cells (Fig. 5A, lane 1). This difference was consistently observed in several experiments, even though the difference in M1/M2 ratio was not statistically significant from that of the WT virus (Fig. 5B). Infection with virus containing the mutated M segments, delNS1:M^CA1 (Fig. 5A, lane 3) and M^CA2 (Fig. 5A, lane 4), resulted in M1/M2 ratios that were 2- and 3-fold higher, respectively, than that of delNS1 (Fig. 5B). Virus containing only the single or double M1 amino acid mutations (delNS1:M^CA2.1, M^CA2.2, and M^CA2.3) (Fig. 5A, lanes 5, 6, and 7) showed an M1/M2 ratio similar to delNS1, whereas delNS1:M^CA2.4 virus (containing the 4 silent M^CA2 mutations) showed a 2-fold increase (Fig. 5A, lane 8, and Fig. 5B). The effect of the mutations on M1 and M2 expression was confirmed by transfection of HEK293T cells with plasmids encoding the different M segments (Fig. 5C). The M1/M2 ratio of cells transfected with WT M segment (0.35; Fig. 5D), is comparable to that of delNS1-infected cells (0.42; Fig. 5B). The M1/M2 expression ratio was higher with segments containing the original mutations acquired during the adaptation (M^CA1 and M^CA2) and with the M^CA2 segment containing the four silent mutations (M^CA2.4) (Fig. 5C, lanes 3, 4, and 8) but not with the M^CA2 segments containing one or both of the nonsilent mutations (Fig. 5C, lanes 5, 6, and 7). Taken together, these results show that both cell-adapted viruses acquired mutations that increased the M1/M2 protein ratio in infected cells.

Effect of M1 amino acid substitutions on subcellular localization.

The M1 mutations V97A and Y100H present in delNS1^CA2 are located close to the NLS at position 101 to 105 (Fig. 2C). To determine whether they affected the subcellular localization of M1 protein, cells were infected with WT, delNS1, or delNS1:M^CA2.3 virus and the nucleoplasm and cytoplasm were isolated early (6 hpi) and late (10 hpi) in the infection process. The level of M1 present in both compartments was then determined by Western blot analysis and quantified by phosphorimager densitometry (Fig. 6A and B).

Fig 6.

Subcellular localization of M1 protein. MDCK-SFS cells were infected with WT, delNS1, and delNS1:M^CA2.3 virus at a high MOI in triplicate. Then, cytoplasm and nucleoplasm fractions were prepared at 6 hpi and 10 hpi and M1 protein was quantified by Western blot analysis. (A) Representative Western blot. Tubulin (Tub.) and lamin A/C (Lam.) were used as cytoplasm- and nucleoplasm-specific controls, respectively. The isolations were successful as no tubulin was observed in the nucleoplasm isolates (lanes 1 to 6) and no lamin A/C was observed in the cytoplasm isolates (lanes 8 to 13). The cytoplasmic isolate of WT virus at 6 hpi was used as a source of tubulin in lane 7, whereas the nucleoplasmic isolate of WT virus at 6 hpi was used as a source of lamin A/C in lane 14. Nuclear isolates show aspecific staining by the tubulin antibody (lanes 1 to 6), which is not present in the cytoplasmic control (lane 7). (B) M1 protein levels were quantified by phosphorimager densitometry using tubulin and lamin A/C as controls to compensate for the isolation efficiency. The ratio of corrected nuclear M1 (M1n) to corrected cytoplasmic M1 (M1c) was used to determine changes in the distribution of M1 and is shown as the mean and 95% confidence interval of the mean at 6 hpi (empty bars) and 10 hpi (filled bars).

At 6 hpi, the ratio of nuclear to cytoplasmic M1 was comparable between WT and delNS1 virus-infected cells (Fig. 6B). Four hours later, this ratio remained constant in WT virus-infected cells whereas it was 4-fold lower in delNS1 virus-infected cells, mainly due to a decrease in nuclear M1. Thus, NS1 appears to affect the concentration of M1 in the nucleus late but not early in the infection. The mutant virus containing only M1 amino acid substitutions V97A and Y100H showed a higher ratio at early stages, although this did not significantly differ from the ratio of the other two viruses. At 10 hpi, nuclear M1 was not detectable anymore, thereby reducing the localization ratio below that of delNS1-infected cells.

DISCUSSION

The low yield of delNS1 virus on MDCK cells is assumed to be caused by the inability of this virus to inhibit the antiviral host response (14). In this study, we obtained two delNS1 influenza virus variants that replicated to 250-fold-higher infectious virus titers (TCID₅₀) after two parallel, serial passages on MDCK cells. However, the increase in virus replication did not correlate with a lower induction of IFN or apoptosis, which is linked to IFN induction (4, 44, 57). These results thus indicate that IFN induction may not be the only cause of limited delNS1 virus replication in MDCK cells. Several recent studies already suggest that IFN has a minor effect on influenza virus replication in MDCK cells because canine myxovirus resistance proteins lack anti-influenza activity (38) and secreted IFN is proteolytically degraded by trypsin, which is normally present during influenza virus production on MDCK cells (39). Furthermore, delNS1 virus titers could be increased by recombinant NS1 expression, without lowering IFN induction (46), indicating that IFN induction in MDCK may play a less important role than generally assumed.

The two adapted viruses contained a high frequency of either A-to-G or U-to-C substitutions (19 out of 22) that occur mostly (16 out of 22) in three clusters in M^CA1, M^CA2, and NS^CA1 gene segments. Furthermore, all mutations in M^CA2 appeared to be acquired at the same time. Taken together, these results suggest that these substitutions result from hyperediting by adenosine deaminases acting on RNA (ADAR), which cause A-to-G substitutions in RNA (37). Hyperediting of the influenza virus genome by ADAR has previously been reported (42). The occurrence of both A-to-G and U-to-C substitutions can be explained by hyperediting of the positive-sense cRNA as well as the negative-sense vRNA, respectively. Interestingly, ADAR1 is induced by IFN and is believed to have an antiviral role during influenza virus infection (37). Furthermore, NS1 interacts with ADAR1, suggesting that it inhibits ADAR1 function (28). Possibly, the high level of IFN induction and the absence of NS1 in delNS1-infected cells resulted in more RNA hyperediting.

By sequence analysis of the two adapted viruses and subsequent analysis of reassortant viruses generated by reverse genetics, we showed that six substitutions in the M segment were responsible for the increase in delNS1^CA2 virus titers. The increase in delNS1^CA1 viral titers was caused by substitutions in the NP, M, and NS segments, where the M segment was most important. Previously, adaptation to Vero cells yielded an influenza B delNS1 virus with increased titers due to M1 amino acid substitution M86V (51). This is in striking contrast to results of the many studies of the adaptation of WT virus originating from eggs or clinical specimens to propagation in mammalian hosts or cell lines, which showed that adaptive mutations predominantly accumulated in the HA segment or segments encoding the RNA polymerase (PB1, PB2, and PA); this effect is assumed to be caused by adaptation to the different host species (16, 29, 34, 45). This suggests that the preferential isolation of M-segment mutations upon delNS1 virus adaptation compensates for the absence of the NS1 protein rather than the replication in a different host species. This conclusion is further supported by our observation that the introduction of the M^CA2 gene segment into a WT virus (that produces NS1) does not enhance, but even reduces, viral replication.

Therefore, we focused on the mechanism by which the M segment mutations could improve replication in the absence of NS1. The M segment encodes the M1 matrix protein from unspliced M1 mRNA, whereas the M2 ion channel protein is transcribed from a spliced mRNA (Fig. 2B). A second splice product, mRNA₃, can arise from an alternative 5′ splice site and encodes a hypothetical and as-yet-undiscovered 9-amino-acid peptide (21). It was previously shown that NS1 expression limits splicing, including that of the M segment-derived mRNA (23, 32), resulting in a higher ratio of M1 to M2 mRNA. Furthermore, M1 expression is reduced in MDCK cells infected with influenza virus expressing truncated NS1 (8, 10). We therefore measured M1 and M2 expression in infected cells and calculated the M1/M2 expression ratio, assuming that this would be dependent on the efficiency of M1 mRNA splicing. Indeed, cells infected with WT virus showed a higher M1/M2 expression ratio than delNS1-infected cells. The M^CA1 and M^CA2 segments showed significantly increased M1/M2 expression ratios compared to the WT M segment, both when expressed using a delNS1 virus backbone and after transfection of cells with M gene-encoding plasmids. The M^CA1 segment contains four silent mutations whereas the M^CA2 segment contains four silent and two nonsilent mutations. By generating novel reassortant viruses we could show that the altered M1/M2 expression ratio of delNS1^CA2 was due to the 4 silent mutations. Furthermore, these silent M^CA2 mutations caused an increase in delNS1 viral titers although not to the same extent as a segment that also contains the two nonsilent M segment mutations. The silent M^CA1 and M^CA2 mutations lay in a region that contains the major determinants for M segment splicing (3). Thus, it is likely that these two sets of mutations lower M1 mRNA splicing efficiency in a similar manner by restoring a balance that was disturbed due to the absence of NS1. Such a mechanism, aimed at restoration of M1 splicing efficiency, may also explain why introduction of the M^CA2 segment into a backbone of virus that produces NS1 (WT virus) reduces replication efficiency. Surprisingly, NS1 does not affect the M1/M2 expression ratio in Vero cells (36). Furthermore, absence of NS1 causes reduced M1 expression in MDCK but not in Vero cells (8). Taken together with our results, this suggests that the improved replication of delNS1 virus in Vero cells compared to MDCK cells is determined not only by the lack of an IFN response but also by the ability of Vero cells to retain efficient M segment splicing and M1 expression in the absence of NS1.

The major part of the increase in delNS1^CA2 virus titer resulted from the combination of M1 amino acid substitutions V97A and Y100H. As single substitutions these mutations did not affect replication efficiency. Mutation V97A was previously introduced into the A/WSN/33 (H1N1) strain (which is able to express NS1) and resulted in a 100-fold lower virus yield (6). Residues 97 and 100 are located on the helix 6 (H6) domain, a positively charged surface region between amino acids 91 and 105 of M1 (40). The influenza B M1 M86V mutation that enhanced delNS1 virus replication (51) is located near this region. The exact mechanism by which this mutation affected viral replication was not further investigated. The H6 domain has multiple functional motifs, including a nuclear localization signal (NLS) between amino acids 101 and 105 (Fig. 2C) that binds to cellular importin-α (5). Inside the nucleus, M1 binds to the vRNP complex, after which NEP can bind to the NLS of M1 (1). The vRNP-M1-NEP complex can then be exported to the cytoplasm, where virus particles are assembled at the cell membrane (1). The localization of V97A and Y100H within the H6 domain suggests that they could affect M1 binding to NEP. Furthermore, Y100H is located immediately next to the NLS and may also affect importin-α binding, as described earlier for a mutation next to an NLS in PB2 (29). In this manner, these mutations could affect M1 (and vRNP) subcellular distribution. The absence of NS1 during infection resulted in decreased levels of M1 in the nucleus at late stages of the infection. This may result from increased apoptosis induction by delNS1 virus as widening of the nuclear pores (11) allows diffusion of vRNPs out of the nucleus (24). Amino acid changes V97A and Y100H resulted in full depletion of M1 in the nucleus at 10 hpi; thus, these mutations do not restore the M1 localization balance in delNS1 toward infection in the presence of NS1. It is therefore difficult to speculate how these two mutations could cause a 50-fold increase in virus titer.

Interestingly, the increased viral titers of delNS1^CA1 virus were in part due to six nucleotide substitutions causing three amino acid substitutions in the NEP protein. In particular, substitution I76V, which is located within the domain that binds to M1 (1), could—similarly to the M1 mutations described above—affect vRNP nuclear export. Similar effects of M1 and NEP mutations that affect their interaction were observed earlier in WT virus. Mutations of NEP glutamate residues 67, 74, and 75, which bind M1, decreased the vRNP content of viral particles and caused morphological virion changes similar to those that occur in virus particles with mutated positively charged M1 residues 95, 98, 101, and 102, which bind NEP (2, 6).

In this paper, we showed that mutations in the M segment can enhance the replication of delNS1 virus due to both nonsilent and silent mutations, the latter presumably affecting the M1 mRNA splicing efficiency. These findings contradict the previous suggestion that the restricted replication of this virus in MDCK cells is primarily due to the inability to inhibit the IFN response. The mutations described may have direct applications, as they, for example, allow the development of delNS1-based viruses with improved replication efficiency, thereby making it possible to produce such a virus in cell lines other than Vero cells (35) or NS1-expressing MDCK cells (46). Moreover, these mutations may be combined with the G3A and C8U mutations in HA vRNA, which increased the HA expression level of a live, attenuated, NS1-truncated influenza vaccine strain (25). However, it will be necessary to determine the effect of these mutations on vaccine safety and efficacy.