The genome of influenza A viruses consists of eight RNA segments. Reassortment of the gene segments plays a prominent role in the biology and evolution of influenza viruses. In fact, the high frequency of reassortment of genetic characters [2] was the first indication of the segmented nature of influenza virus genome. For a long time, there was a controversy concerning the random or non-random packaging of influenza virus genome segments [4, 5]. At present, the problem seems to have been resolved by the discovery of specific packaging signals in the terminal coding sequences of the genomic segments [6]. However, the specific packaging does not necessarily imply the absence of partial heterozygosis in the reassortant progeny. The general possibility of partial heterozygosis was demonstrated by the incorporation of nine RNA segments in an influenza virion [5]. Further, irrespective of the packaging mechanism, the frequency of reassortment for any given segment is usually regarded as the frequency expected at independent segregation [14, 20]. Several deviations from this rule have been described, mostly resulting from a functional cooperation between virus genes. Such cooperation was described for the HA and M genes of human and avian influenza viruses [18]. In our earlier studies on human-avian influenza virus reassortment, we observed a functional mismatch of HA and NA genes in the reassortants having HA originating from an avian parent and NA from a human H1N1 strain [9, 10]. In at least one case, a non-random association of parent genes was detected under non-selective conditions [12].

The initial goal of the present studies was to verify whether the necessity of a functional match leads to genetic linkage during the reassortment between two live influenza viruses under non-selective conditions, and whether the formation of partial heterozygotes (that is, virus particles containing at least one gene represented by two copies originating from different parents) is a frequent event in the course of mixed infection. However, the analysis of the genetic content of the reassortants provided unexpected results, suggesting that factors other than functional linkage may be involved in a strong deviation of the frequency of reassortment for a specific gene from the frequency expected for the independent segregation.

Influenza viruses A/WSN/33 (H1N1) and A/Duck/Czechoslovakia/56 (H4N6) were obtained from the virus collection of the D. I. Ivanovsky Institute of Virology, Moscow. The viruses were propagated in 10-day-old embryonated chicken eggs, and the virus-containing allantoic fluid was stored at 4°C. The infectious titer of the virus was assessed by end-point titration in the embryonated chicken eggs and by plaque titration in MDCK cells.

Goat polyclonal serum against HA of /Duck/Czechoslovakia/56 (H4N6) virus was a kind gift of Dr. R. G. Webster (St. Jude Children’s Research Hospital, Memphis, Tennessee). Rabbit serum against A/WSN/33 (H1N1) virus was kindly given by Dr. H.-D. Klenk (Institut fuer Virologie, Philipps Universitaet, Marburg, Germany). The hemagglutination-inhibition test was performed by a conventional technique [16].

For double infection, the embryonated chicken eggs were inoculated in the allantoic cavity with 0.2 ml of the mixture of parent viruses. The mixed yields were used for plaque titration and clone isolation in MDCK cells. The cells were infected by serial virus dilutions with 0.5 log10 steps, and wells containing no more than ten plaques were used for picking plaque isolates. Each isolate was propagated in chicken eggs (five eggs per plaque, 48-h incubation at 37°C).

The conditions of double infection used in our experiments were close to the ones used in the standard procedure for influenza A virus reassortment [19]. The inoculum contained 1 × 108 to 4 × 108 egg infectious 50% doses (EID50) of each virus. Groups of five eggs were inoculated with each mixture. The eggs were incubated for 13 h at 37°C, cooled, and the virus-containing allantoic fluid was collected. The total amount of infectious virus used for the infection of embryonated chicken eggs with the mixtures of A/Duck/Czechoslovakia/56 (H4N6) and A/WSN/33 (H1N1) viruses was 3 × 108 to 5 × 108 EID50 per egg; that is, the majority of the cells of the chorioallantoic membrane had to be infected by the virus used as inoculum. The ratio of the infectious viruses in the inoculum (A/Duck/Czechoslovakia/56:A/WSN/33) ranged from 2:1 to 1:4. The virus yields produced by the parent viruses and by the mixed inocula were collected and analyzed in the HI test with anti-H1 and anti-H4 immune sera. The yields produced by the ratios 2:1, 1:1 and 1:2 exhibited a clear-cut double neutralization pattern (Table 1), indicating that a very large fraction of the virus yields was represented by phenotypically mixed virus particles [8, 21]. This result suggested that a significant proportion of the chorioallantoic cells were infected by both viruses and could be expected to produce reassortants.

Table 1 Hemagglutination-inhibition test with mixed yields
Full size table

To obtain data on the distribution of virus genes in the reassortants, the progeny of individual infectious virus particles had to be analyzed. For this goal, we performed the analysis of plaque isolates. As we were interested in registering partial heterozygosis, the plaque isolates were not subjected to plaque-to-plaque passaging. The complete genetic content of the isolates was determined on the basis of the strain-specific differences in mobility of NP and NS1 proteins in polyacrylamide gel electrophoresis (Fig. 1), partial sequencing of the M gene, and the analysis of PCR products using strain-specific primers for the other genes. For the analysis of virus-specific proteins, MDCK cells were infected at an m.o.i of 5–10 plaque-forming units (PFU) per cell, labeled with 14C-amino acids from 5 to 6 h postinfection (p.i.) and chased 6–7 h p.i. with an excess of unlabelled amino acids. Polyacrylamide gel electrophoresis was performed as described in our earlier publication [22]. The sequencing was performed using a DNA ABI Prism 3130 sequencer (Applied Biosystems) and a BigDye Terminator v3.1 kit; DNA sequences were completed and edited by using the DNASTAR sequence analysis software package (DNASTAR Inc.). For genotyping based on PCR analysis, two pairs of strain-specific primers were used for each gene. The primers were chosen to be complementary to a specific gene of one parent virus and non-functional as a primer for the same gene of the other (primer sequences are available upon request). The parent origin of each gene was tested by the presence or absence of the corresponding band in the agarose gel (Fig. 2).

Fig. 1
figure 1

The mobility of NP and NS1 proteins of A/Duck/Czechoslovakia/56 (H4N6) (lane 1) and A/WSN/33 (H1N1) (lane 2) viruses in polyacrylamide gel electrophoresis. MDCK cells were infected and labeled as described in the text

Full size image
Fig. 2
figure 2

Distribution of PA (a) and PB2 (b) genes in plaque isolates as determined by PCR analysis. Electrophoresis of PCR products obtained with A/WSN/33-specific (upper part) and A/Czechoslovakia/56-specific (lower part) oligonucleotide primers. Electrophoresis was performed in 1% agarose gel. a Plaques CW1-Pla2, CW1-Pla4, CW1-Pla5, CW1-Pla22, A/WSN/33, and A/Czechoslovakia/56 (lanes 1, 2, 3, 4, 5 and 6, respectively). b Plaques CW2-Pla2, CW2-Pla3, CW2-Pla18, CW2-Pla26, A/WSN/33, and A/Czechoslovakia/56 (lanes 1, 2, 3, 4, 5 and 6, respectively)

Full size image

The mixed yields produced in two independent experiments by a mixture of A/Duck/Czechoslovakia/56 and A/WSN/33 viruses at the m.o.i ratio 1:1 were used for the infection of MDCK cells, and 37 plaque isolates were genotyped. The genotyping demonstrated that ten clones were non-reassortants. Most of them (eight isolates) obtained all their genes from A/WSN/33 virus, and two isolates were A/Duck/Czechoslovakia/56 virus. The other 27 isolates were reassortants (Table 2).

Table 2 The genetic content of reassortants between A/Duck/Czechoslovakia/56 and A/WSN/33 viruses generated at a multiplicity ratio 1:1
Full size table

The overall genetic content of the reassortants was represented by the genes of both parents to nearly the same extent. The sum of the A/WSN/33 genes in 27 reasortants was 116, and the sum of A/Duck/Czechoslovakia/56 genes was 103. However, for some individual genes the ratio of parent genes in the reassortant genomes deviated strongly from the overall representation. Genotyping revealed that the HA gene of A/WSN/33 virus and the NP gene of A/Duck/Czechoslovakia/56 virus were preponderant in the reassortants. The parent origin of the M gene often coincided with that of HA. No clear-cut predominance of PB1 and NS genes of a specific parent origin was revealed, whereas the preponderance of A/WSN/33 genes for PB2, NA and M was not as evident as for HA and M.

We have no means to determine whether the representation of parent genes in the reassortants characteristic for any specific gene reflects a random incorporation pattern. However, if one presumes that the nearly equal representation of NS genes of each parent reflects a random incorporation of the parent genes into reassortant virions, the deviation from the random pattern for HA and NP genes as calculated using the χ2 criterion [12] is statistically significant at the level of probability P < 0.05 and P < 0.01, respectively; that is, at the level considered reliable in biological experiments. If the ratio revealed for the PB1 gene reflects a random representation of parent genes, the deviation of the ratio for HA, NP and M genes is highly significant (at the level of probability <0.005, <0.005 and <0.025, respectively). If one presumes that the genes of A/WSN/33 virus prevail, so that the ratio observed for HA gene reflects the random incorporation of segments, the deviation from the random pattern is significant for the PB1, PA, NP and NS genes (Table 2). If the ratio observed for any other gene beside NS, PB1 and HA is presumed to result from random incorporation, the calculation of P values for the difference between this ratio and the ratios observed for the other genes reveals that in each case the representation of at least one gene deviates from the random pattern (not shown).

Three reassortants appeared to be heterozygotic with respect to PB2 (CW2-Pla18), PA (CW1-Pla2), or M (CW1-Pla4). These isolates may be regarded as derived either from confluent plaques produced by two reassortant virus particles, or, alternatively, from real heterozygotes, that is, virus particles containing more than eight segments [5]. Importantly, the majority of isolates contained one copy of each gene, in accordance with electron microscopy data demonstrating eight segments in each particle [15].

The dominance of HA gene of A/WSN/33 virus in the mixed yields is in contrast with the equal representation of HA of both parents as revealed by HI test (Table 1). However, the ratios of virus proteins in phenotypically mixed particles may differ from the genetic content [20].

It is generally believed that the genetic content of a specific reassortant virus is determined by selective conditions. Specific combinations of virus genes in natural isolates, such as the preponderant HA-NA combinations in the avian influenza viruses, are regarded as resulting from natural selection [24]. When reassortants are produced by double infection under experimental conditions, special selection factors are applied, such as antibodies [23] or selective temperature [3, 13]. The selection pressure results in the isolation of reassortants with specific genetic content from a wide array of reassortants produced by the doubly infected cells. On the other hand, if the reassortants are produced under non-selective conditions, the distribution of gene segments is expected to be random, rare exceptions being observed for functionally linked genes [17]. However, the concept of the random distribution of virus genes in the virus progeny produced in the course of double infection is based on limited evidence obtained mostly with the use of closely related viruses, such as two human strains [20], or two variants of FPV [14]. A limited number of genetic traits were analyzed, and they were registered as phenotypic features without the application of molecular techniques. In the only case where two closely related viruses were analyzed for the origin of all eight segments [12], a slight deviation from the random pattern was observed. For more distantly related influenza A viruses, that is, a human parent virus and an avian parent virus, to our knowledge, the distribution of the parent virus genes in the reassortants produced in the course of a productive mixed infection under non-selective conditions has never been studied in detail.

The data of the present study describe the genetic content of reassortants produced by double infection with influenza A strains in embryonated chicken eggs. The results demonstrate a strong deviation from a random pattern of distribution of the genomic segments in the reassortants. The results described in the present report suggest that a kind of competition between the genes of the parent viruses takes place in the course of double infection, and this competition is differential. A specific gene of one parent (such as HA of A/WSN/33 in our experiments) may be dominant in the mixed progeny, whereas the other gene originates predominantly from the other parent (NP in our experiments). Since the genes preponderant in the reassortants originated from different parents, the observed effect cannot be explained by functional intergenic linkage.

At least three explanations for the observed phenomenon are possible: First, although we applied no selection factors, one has to keep in mind that absolutely non-selective conditions do not exist. Specifically, egg-adapted viruses were used in our study, and the generation of the reassortants proceeded in the embryonated chicken eggs. The growth properties of the chosen strains were roughly equivalent, the virus titres in the allantoic fluid ranging for A/WSN/33 from 108.9 to 109.3 EID50, and for A/Czechoslovakia/56 from 108.7 to 109.1 EID50. However, the cloning was performed in MDCK cells, and a better replication of a reassortant in MDCK cells could be a selection factor. However, the reassortants containing the NP gene of A/WSN/33 virus and the HA gene of A/Duck/Czecoslovakia/56 virus, though less frequent, were good plaque-formers, and we did not observe any differences in plaque size or the level of accumulation in MDCK culture fluid correlating with a specific genetic content of a reassortant (not shown). Therefore, the selection at cloning does not seem to be a proper explanation for the low frequency of such reassortants. Another explanation for the observed deviation from the random pattern may be differential competition at the level of the RNA segments’ replication in the doubly infected cell. We did not observe any indications of such competition when virus proteins synthesized in MDCK cells infected with virus mixtures (the same as used for the generation of reassortants) were analyzed in polyacrylamide gel electrophoresis (not shown). However, even if there is no differential accumulation of segments in MDCK cells, we cannot exclude the existence of such competition during the replication in the doubly infected cells of the chorioallantoic membrane.

The third possible mechanism is the preferential packaging of specific segments. The specificity of packaging of influenza virus genomic segments was a controversial problem for a long time, with some data supporting the model of nonspecific recruitment of the segments into the virion [1, 5], and other data indicating that a specific packaging mechanism should exist [4]. Recently, the specific recognition at the level of packaging operated by the terminal parts of the coding sequences of the segments was demonstrated [6, 11]. Moreover, data showing drastic effects of nucleotide substitutions in the packaging signals have been reported, and the existence of strain-specific differences in the packaging signals have been predicted [7]. Our data are not sufficient to prove that such differences exist, but they indicate that a search in this direction may prove to be fruitful.

In any case, whatever the subtle mechanism of the observed phenomenon, it seems obvious that the intrinsic properties of the influenza genes, at least in some cases, exert a strong effect on the distribution of genomic segments in the reassortant progeny, and this may affect the genetic content of the reassortants, leading to a non-random reassortment pattern when no specific selective pressure is applied.