Influenza A Virus Superinfection Potential Is Regulated by Viral Genomic Heterogeneity

Superinfection, the sequential infection of a single cell by two or more virions, plays an important role in determining the replicative and evolutionary potential of influenza A virus (IAV) populations. The specific mechanisms that regulate superinfection during natural infection remain poorly understood. Here, we show that superinfection susceptibility is determined by the total number of viral genes expressed within a cell and is independent of their specific identity. Virions that express a complete set of viral genes potently inhibit superinfection, while the semi-infectious particles (SIPs) that make up the bulk of IAV populations and express incomplete subsets of viral genes do not. As a result, viral populations with more SIPs undergo more-frequent superinfection. These findings identify both the primary determinant of IAV superinfection potential and a prominent role for SIPs in promoting coinfection.

I nfluenza A viruses (IAV) are estimated to cause hundreds of thousands of deaths across the world every year during seasonal epidemics, despite widespread preexposure and vaccination (1). In addition to the yearly burden of seasonal influenza viruses, novel zoonotic IAV strains periodically emerge into humans from swine or birds, triggering unpredictable pandemics that can dramatically increase infection and mortality rates (2). Defining the specific factors that influence the evolution of influenza viruses is critical for designing more-effective vaccines, therapeutics, and surveillance strategies.
The prevalence of coinfection can play an enormous role in determining the replicative and evolutionary potential of IAV populations. This is a function of both the segmented nature of the viral genome and the enormous amount of genomic heterogeneity present within IAV populations (3,4). Coinfection allows reassortment, i.e., the production of novel viral genotypes through the intermixing of the individual IAV genome segments (5,6). Reassortment events have contributed to the emergence of every major influenza pandemic of the past century (7). Coinfection also facilitates the complementation and productive replication of the semi-infectious particles (SIPs) that make up the majority of IAV populations (8)(9)(10)(11)(12). Finally, increasing the frequency of coinfection can accelerate viral replication kinetics and virus output by increasing the average multiplicity of infection (MOI) (13)(14)(15). Thus, to better understand how IAV populations transmit and evolve, we must identify the specific host and viral factors that govern coinfection. One of the primary means by which coinfection can occur is superinfection, the sequential infection of one cell by multiple viral particles. For some viruses, superinfection appears to occur freely (16,17). In contrast, a diverse range of viruses actively inhibit superinfection through a variety of mechanisms, a phenomenon known as superinfection exclusion (SIE) (18)(19)(20)(21)(22)(23)(24)(25)(26). The only in-depth study of IAV superinfection performed to date concluded that the viral neuraminidase (NA) protein acts to potently and rapidly inhibit IAV superinfection by depleting infected cells of the sialic acid receptors required for viral entry (27). More recently, Dou et al. reported a narrow time window during which IAV superinfection was possible (13). These findings are consistent with recent studies that have argued that reassortment is rare during human infection (28,29). However, the existence of a potent mechanism of IAV SIE is at odds with both the frequent coinfection observed in a variety of experimental settings and the widespread occurrence of reassortment at the global scale (30)(31)(32)(33)(34)(35). Marshall et al. showed that superinfection occurring at up to 8 h after a primary infection leads to robust coinfection and reassortment in cell culture (36). Extensive coinfection and complementation have also been observed in the respiratory tracts of IAV-infected mice and guinea pigs (9,37). Collectively, these results suggest that IAV superinfection can be restricted, but to what extent and through which specific mechanisms remain crucial open issues.
Here, we reveal that IAV superinfection potential is regulated by the extent of genomic heterogeneity within the viral population. We observed that superinfection susceptibility is influenced in a dose-dependent fashion by the number of viral genes expressed by the initially infecting virion. Further, we show that superinfection occurs more frequently in IAV populations with more SIPs than in those with fewer. Finally, we demonstrate that SIE is mediated by the presence of active viral replication complexes and is completely independent of the gene coding sequence. Taken together, our results reveal how genomic heterogeneity influences IAV superinfection potential and demonstrate how SIPs can modulate collective interactions within viral populations.

Influenza virus SIE occurs in multiple cell types and is independent of type I interferon (IFN) secretion.
A previous study of IAV SIE concluded that NA expression completely blocks susceptibility to superinfection by 6 h postinfection (hpi) (27). To explore the potential mechanisms of IAV SIE in greater detail, we developed a flow cytometry-based assay that allows us to precisely measure the effects of previous infection on superinfection efficiency. To clearly identify cells infected by the first virus or by the superinfecting virus or by both, we used two recombinant viruses that express antigenically distinct hemagglutinin (HA), NA, and NS1 proteins that we could distinguish using specific monoclonal antibodies (MAbs) that we had on hand (see Fig. S1 in the supplemental material). For the primary infection, we used a recombinant version of H1N1 strain A/Puerto Rico/8/34 (rPR8). For the secondary infection, we used a recombinant virus (rH3N2) that contained the HA and NA gene segments from H3N2 strain A/Udorn/72, the NS gene segment from A/California/04/09, and the remaining 5 segments from PR8.
We first asked whether prior infection with rPR8 affected cellular susceptibility to superinfection with rH3N2. We infected Madin-Darby canine kidney (MDCK) cells with rPR8 at an MOI of Ͻ0.3 50% tissue culture infective doses (TCID 50 )/cell, and at 3 hpi (all times postinfection are expressed relative to the time at which the first virus was added) we added the PR8-HA-specific neutralizing MAb H17-L2 to block secondary spread of rPR8 within the culture. At 6 hpi, we infected with rH3N2 at an MOI of Ͻ0.3 TCID 50 /cell (6 hr). To prevent spread of both rPR8 and rH3N2, we added 20 mM NH 4 Cl at 9 hpi (38,39). In parallel, we performed simultaneous coinfections (0 hr) with rPR8 and rH3N2 to measure coinfection frequencies under conditions SIE should not be possible. At 19 hpi, we harvested cells and examined primary and secondary virus infections by flow cytometry, using H1 expression and H3 expression as markers of rPR8 infection and rH3N2 infection, respectively. We observed that the H3-positive (H3 ϩ ) frequency within H1 ϩ cells was significantly reduced when rPR8 infection preceded rH3N2 by 6 h compared with when rPR8 and rH3N2 were added simultaneously (Fig. 1A). This indicated that rPR8 infection significantly reduces the susceptibility of cells to superinfection by 6 hpi.
We next asked whether the SIE effect was cell type specific and whether it depended on activation of the type I interferon (IFN) system. We performed the experiment described above in MDCK cells, A549 cells, human embryonic kidney HEK293T (293T), and Vero cells (which are incapable of type I IFN secretion) (40,41). We observed that the levels of SIE were comparable among all cell lines tested, suggesting that SIE occurs in multiple distinct cell types and does not depend upon IFN secretion ( SIE does not depend upon viral neuraminidase activity. In an attempt to confirm the previously reported role for NA activity in SIE, we directly measured the effect of NA expression on SIE in our system (27). We took advantage of our previous observation that IAV populations consist primarily of SIPs that fail to express one or more viral genes (8). When carrying out the primary infection at a low MOI, we generate populations of infected cells that are either positive or negative for expression of a given viral gene. We can then assess the effects of specific viral proteins on superinfection susceptibility by comparing superinfection frequencies between infected cells that do or do not express the protein in question ( Fig. 2A).
We performed the same superinfection experiment as described above in MDCK cells; however, we used slightly different viruses to ensure that the NA specificity of the primary virus was well matched to the HA specificity of the secondary virus. The primary virus used here encoded the HA gene from A/Udorn/72 and the NA gene from PR8 (rH3N1), while the secondary virus encoded the HA gene from PR8 and the NA gene from A/Udorn/72 (rH1N2). The remaining 6 segments for both viruses came from PR8. Influenza A Virus Superinfection Regulation ® At 19 hpi, we harvested and stained with MAbs against H1, N1, and H3. To compare rH3N1 infected cells that did or did not express NA, we individually gated cells into H3 ϩ N1 ϩ and H3 ϩ N1 Ϫ subpopulations ( Fig. 2A). Comparison of H1 ϩ frequencies between H3 ϩ N1 ϩ and H3 ϩ N1 Ϫ cells revealed that NA expression in infected cells was clearly associated with decreased susceptibility to superinfection (Fig. 2B). This finding was consistent with the previously reported role for NA in IAV superinfection exclusion (27). Importantly, while SIE was more pronounced in the H3 ϩ N1 ϩ cells, we also observed a significant decrease in superinfection susceptibility within the H3 ϩ N1 Ϫ cell population by 6 hpi, suggesting that viral factors other than NA also act to restrict superinfection.
To directly test whether NA enzymatic activity was required for the increased potency of SIE that we observed in NA ϩ cells, we asked whether treatment with the NA inhibitor (NAI) zanamivir could diminish this effect. We observed that the addition of zanamivir at 1 M (a concentration that completely blocked rH3N1 NA enzymatic activity) did not decrease the strength of SIE in NA ϩ cells ( Fig. 2B; see also Fig. S3). We also examined the effects of two substitutions (NP:F346S and NA:K239R) that decrease cellular NA expression relative to wild-type PR8 on superinfection efficiency (9, 42) ( Fig. S4A). In accordance with our zanamivir results, these mutants did not exhibit higher superinfection frequencies than wild-type PR8 (Fig. S4B). Taken together, these results reveal that superinfection is more effectively inhibited by virions that express NA than by those that do not but that NA enzymatic activity is dispensable for this effect. Superinfection susceptibility is determined by the number of viral genes expressed in the cell. On the basis of our observation that superinfection was also inhibited within N1 Ϫ cells (Fig. 2B), we hypothesized that expression of other viral gene products might also inhibit superinfection. We examined the effects of HA and NS1 expression on superinfection susceptibility, using rPR8-specific MAbs. Surprisingly, we found that both HA expression and NS1 expression within rPR8-infected cells were associated with significant decreases in superinfection by rH3N2 and that the results were comparable to the effect associated with NA expression (Fig. 3A and B). To further dissect how viral gene expression patterns influence SIE, we individually gated all seven possible combinations of HA, NA, and NS1 expression by rPR8 (HA ϩ NA ϩ NS1 ϩ , HA ϩ NA ϩ , HA ϩ NS1 ϩ , NA ϩ NS1 ϩ , HA ϩ , NA ϩ , NS1 ϩ ) and directly compared their rH3N2 infection frequencies (the gating scheme is shown in Fig. S5A). We observed that the fraction of cells superinfected with rH3N2 was inversely correlated with the number of rPR8 genes expressed (among the three we examined), regardless of their specific identities ( Fig. 3C and D). Thus, susceptibility to IAV superinfection is determined by the number of viral genes expressed in the host cell by the initially infecting virion. Superinfection is more prevalent in IAV populations with more SIPs. If the number of viral genes expressed in a cell determines superinfection susceptibility, then decreasing the average number of functional viral genes successfully delivered by individual virions should increase the overall incidence of superinfection. We tested this by artificially decreasing the functional gene segment content of rPR8 through exposure to UV irradiation (43). Exposure to low-dose UV irradiation generates SIPs that carry gene-lethal UV-induced lesions at frequencies proportional to genome segment length. On the basis of our previous findings, we hypothesized that superinfection frequencies would increase with longer exposure of rPR8 to UV.
We used UV (302-nm wavelength) to irradiate rPR8 for either 30 s or 60 s and confirmed that the TCID 50 concentration was reduced and the SIP concentration was increased as a function of treatment duration (Fig. 4A to C). We then performed superinfection assays as described above, comparing rH3N2 superinfection frequencies following infection by untreated or UV-irradiated rPR8 in MDCK cells. To fairly compare superinfection frequencies between viral populations with differing particle-toinfectivity ratios, we normalized our rPR8 infections based on equivalent numbers of particles capable of expressing NA (NA-expressing units [NAEU]) (9).

FIG 3
Superinfection is more frequent in cells that express fewer viral genes. MDCK cells were infected with rPR8 and were simultaneously (0hr) or sequentially (6hr) infected with rH3N2; all infections were performed at MOI ϭ Ͻ0.3 TCID 50 /cell. (A and B) Representative FACS plots comparing H3 ϩ frequencies between (A) N1 ϩ H1 Ϫ and N1 ϩ H1 ϩ cells and (B) H1 ϩ NS Ϫ and H1 ϩ NS ϩ cells. (C) Superinfection assay performed as described for panels A and B; however, cells are gated into all 8 possible combinations of HA, NA, and NS1 expression by rPR8, and H3 ϩ percentages are compared between these subpopulations. Data represent the values of 6hr groups normalized to the means of 0hr controls. Each data point represents the H3 ϩ frequency for the indicated cell population within a single cell culture well, shown as a percentage of the mean H3 ϩ frequency of the same cell subpopulation in the 0hr controls. (D) Data from panel C grouped by total numbers of the three examined viral gene products (HA, NA, and NS1) expressed by rPR8 rather than by their specific identities. *, P Ͻ 0.05; **, P Ͻ 0.01; ***, P Ͻ 0.001; ****, P Ͻ 0.0001; N.S., not significant (t test).

®
We first examined the effect of UV treatment on superinfection when rPR8 and rH3N2 were added to cells simultaneously (at h 0). This was a critical control because UV treatment can increase the measured incidence of coinfection, independently of SIE effects, purely by creating a larger pool of SIPs that show up in our assays only when complemented by secondary infection (43). Consistent with this, we observed a small increase in coinfection frequency with UV treatment when the two viruses were added simultaneously ( Fig. 4D and E). When rH3N2 was added 6 h after rPR8, however, we observed a much more pronounced increase in superinfection frequency with UV treatment, consistent with our hypothesis that superinfection can be regulated by the proportion of SIPs present within the viral population ( Fig. 4D to F). SIE is mediated by active IAV replication complexes and is independent of the gene coding sequence. Our data reveal that IAV superinfection potential is determined by the number of viral genes expressed within a cell, independent of their specific identity. This suggests that the viral gene products themselves are dispensable for SIE. We thus hypothesized that active replication and/or transcription of viral RNAs by the viral replicase complex might be responsible for decreasing cellular susceptibility to subsequent infection. To test this, we cotransfected 293T cells with pDZ vectors encoding the individual viral replicase proteins (PB2, PB1, PA, and NP) together with a pHH21 vector encoding either the HA vRNA gene segment (HA vRNA ) or a vRNA-derived reporter gene segment in which the enhanced green fluorescent protein (eGFP) open reading frame (ORF) is flanked by the 5= and 3= untranslated-region (UTR) sequences from the NA segment (eGFP vRNA ). These UTR sequences are required for replication and transcription of the reporter RNA by the viral replicase. At 24 h posttransfection, we infected cells with rH3N2 at an MOI of 0.2 TCID 50 /cell and measured infectivity at 8 hpi using an M2-specific MAb.
Infection frequencies were decreased ϳ50% in cells expressing the replicase components plus the eGFP vRNA construct compared with control cells transfected with the replicase-expressing constructs plus an empty pHH21 vector (Fig. 5A). In analyses of differences in the levels of rH3N2 infectivity between cotransfected cells (eGFP ϩ , HA ϩ ) and uncotransfected cells (eGFP Ϫ , HA Ϫ ) within the same culture wells, the inhibitory effects mediated by eGFP vRNA or HA vRNA expression were found to be comparable ( Fig. 5B; see also Fig. S6). Importantly, this effect was not seen when we left out the plasmid encoding PA (RNP PAϪ ) or used an eGFP reporter RNA that lacked the viral UTR sequences (eGFP ORF ) (Fig. 5A). Taken together, these data indicate that inhibition of infection requires both an intact replicase complex and an RNA template containing the viral UTR sequences but not the viral coding sequence.
Our data demonstrate that IAV SIE is mediated by the specific activity of viral replication complexes. One potential explanation is that large amounts of recently synthesized negative-sense vRNA within the cell might outcompete incoming genome segments for replication and expression. To test this, we transfected 293T cells with a pHH21 vector that overexpresses the eGFP vRNA segment and measured susceptibility to rH3N2 infection 24 h later using an NP-specific MAb. Compared to the empty vector control, we observed no effect of eGFP vRNA vRNA overexpression on cellular susceptibility to infection (Fig. 5C). Similarly, we observed no effect when we overexpressed the cRNA and vRNA forms of the HA gene segment, either individually or together (Fig. 5C). It must be noted that bulk levels of viral RNA were roughly 5-fold lower in these cells than in cells that express the viral replicase, so we cannot rule out a role for the intracellular abundance of viral RNA as a determinant of susceptibility to subsequent infection (Fig. S7).
Another potential explanation is that viral mRNA or protein overexpression might inhibit subsequent infection. To test this, we transfected 293T cells with pCI vectors that overexpress mRNA and protein of eGFP and HA and measured susceptibility to rH3N2 infection 24 h posttransfection using the M2-specific MAb. Compared to the empty vector control, mRNA/protein overexpression of eGFP or HA had no effect on infection susceptibility ( Fig. 5D; see also Fig. S8). Collectively, these data demonstrate that IAV superinfection exclusion is triggered by the activity of viral replication complexes.

DISCUSSION
Superinfection plays an enormous role in influencing the outcome of IAV infection, both by promoting reassortment and by facilitating the multiplicity reactivation of SIPs and defective interferring particles (4). Despite this importance, the specific factors that govern the occurrence of superinfection have remained obscure. Here, we reveal that IAV superinfection susceptibility is regulated by the number of viral genes successfully expressed by a virion. We further demonstrate that the presence of SIPs within viral populations significantly increases the frequency of superinfection. This represents a completely novel mechanism of viral superinfection exclusion and identifies a clear functional consequence of the enormous genomic heterogeneity within IAV populations.
The only other published study that examined IAV SIE in detail concluded that NA expression mediates SIE by depleting the pool of available sialic acid receptors on the Influenza A Virus Superinfection Regulation ® cell surface (27). In this study, we directly quantified the contribution of NA expression to SIE during IAV infection and found that the SIE effect of NA expression is actually comparable to that of other viral genes. We also show that treatment with NAIs has no appreciable effect on superinfection susceptibility. The conclusions of the study by Huang et al. were based primarily on two observations: (i) overexpression of NA within cells rendered them refractory to infection by an HA-pseudotyped virus, and (ii) IAV superinfection occured only when cells were treated with NAIs (27). While we cannot conclusively explain the discrepancies between the two studies, we can offer a couple of plausible explanations. First, the cellular overexpression studies in Huang et al. likely Sun and Brooke ® involved levels of cellular NA expression that are far beyond those seen during IAV infection. Second, the observation that NAI treatment dramatically increases superinfection frequencies may be explained by the effects of cell death. In their experiments, Huang et al. infected cells at a relatively high MOI, did not block secondary spread of the virus within cultures, and assessed superinfection frequency at 20 hpi or later. Under those conditions, many of the initially infected cells may have been dead or dying and thus lost from the analysis. That may be especially true of superinfected cells, which tend to be infected at a higher than the average effective MOI. Even under low-MOI conditions, we had to limit the time frame of our experiments and block secondary spread of virus to prevent cell death from skewing our results. NAI treatment may act to help preserve coinfected cells so that they are detected at the endpoint of the experiment, thus increasing the measured superinfection rate.
Our results reveal that SIE is triggered in a dose-dependent fashion by the number of functional viral gene segments delivered by the initially infecting virion. The surprising irrelevance of the specific IAV gene segments involved is explained by our finding that the viral coding sequence of a gene segment can be replaced with that of eGFP without any loss of inhibitory effect. This suggests a direct role for viral replicase activity itself in triggering SIE rather than any effect of the viral gene segments themselves. We hypothesize that the overall levels of replicase activity in the cell, particularly during the early stages of infection, are determined by the number of functional RNPs delivered to the cell. Thus, cells infected by SIPs that deliver fewer functional RNPs exhibit lower overall levels of replicase activity, resulting in less-potent SIE.
The specific mechanism by which the activity of viral replicase complexes may inhibit subsequent infection remains unclear; however, one potential explanation is that the activity of viral replication complexes triggers a cell-intrinisic antiviral response. As each incoming gene segment brings its own replicase complex, this could potentially explain the gene dose-dependent suppression of superinfection that we observed. While our experiments in Vero cells demonstrated that the secretion of type I IFN is not required for SIE, they do not preclude the involvement of type I IFNindependent mechanisms. These could include either the type III IFN-mediated induction of antiviral effectors or the engagement of completely IFN-independent antiviral mechanisms (44)(45)(46). Future studies will be aimed at delineating the role of the host in the regulation of IAV superinfection.
Our results demonstrate that SIPs can directly influence the prevalence of superinfection and thus, potentially, the frequency of reassortment. Fonville et al. used a UV irradiation-based method similar to that shown here to demonstrate that increasing the frequency of SIPs within a viral popuation increases the overall reassortment rate (43). The explanation given for this effect was that increasing the abundance of SIPs increases the proportion of the viral population that depends upon coinfection to replicate. As a result, within a certain MOI range, a greater share of productively infected cells are coinfected and subject to reassortment. In our study, we confirmed this effect by observing a slight increase in coinfection frequency with increasing UV dose when rPR8 and rH3N2 were added simultaneously (Fig. 4E). When we controlled for this, however, we still observed a significant increase in superinfection frequency as we increased the proportion of SIPs through UV treatment (Fig. 4D to F). Thus, the relationship between SIPs and SIE that we describe here is completely independent of the increased multiplicity reactivation observed by Fonville et al. and likely represents the effects of decreasing the strength of SIE. Between these two studies, it is clear that SIPs can modulate the frequency of IAV coinfection and reassortment through two distinct mechanisms.
IAV strains can differ significantly with respect to the relative production and gene expression patterns of SIPs (8,9). This raises the possibility that strains with distinct SIP production phenotypes may differ in their reassortment potential, given the influence of SIPs on coinfection and reassortment frequencies. If this is the case, it would suggest a significant role for production of SIPs in governing the evolutionary potential of IAV populations.
The relationship between viral gene expression patterns and superinfection exclusion that we report here demonstrates that viral genomic heterogenity has distinct functional consequences during infection. A crucial implication is that all infected cells cannot be thought of as equal but may in fact exhibit distinct phenotypes based on the number and identity of viral genome segments that they harbor (47). The relationship between viral genomic heterogeneity and the phenotypic diversity of infected cells likely extends to cellular features beyond superinfection susceptibility.
It remains to be seen whether the relationship between viral gene dose and superinfection susceptibility that we describe here exists for other segmented viruses besides IAV. Beyond the segmented viruses, it has become increasingly clear that collective interactions mediated by cellular coinfection significantly influence the replicative and evolutionary dynamics of nonsegmented viruses as well (48). More work is needed to better understand the factors that govern coinfection for different virus families, both in vitro and in vivo.
In summary, our work reveals a unique mechanism of IAV superinfection regulation that is governed by viral genomic heterogeneity. Critically, we show that the abundance of SIPs within a viral population directly influences the prevalence of superinfection; suggesting that IAV strains may differ in their superinfection potential and thus in their potential for reassortment. This finding has significant consequences for understanding how SIP production can influence the evolutionary potential of IAV populations. More broadly, we demonstrate how genomic diversity within viral populations can have clear functional consequences during infection.
Cells. Madin-Darby canine kidney (MDCK) and human embryonic kidney HEK293T (293T) cells were maintained in Gibco's minimal essential medium (MEM) with GlutaMax (Life Technologies). Vero cells were maintained in Dulbecco's modified Eagle medium (Life Technologies). Human lung epithelial A549 cells were maintained in Gibco's F-12 medium (Life Technologies). MDCK, Vero, and A549 cells were obtained from Jonathan Yewdell; 293T cells were obtained from Joanna Shisler. All media were supplemented with 8.3% fetal bovine serum (Seradigm). Cells were grown at 37°C and 5% CO 2 .
Viruses. Recombinant A/Puerto Rico/8/1934 (rPR8) and rH3N2 viruses were generated using 8-plasmid rescue systems. The rH3N2 virus is a reassortant with the HA and NA segments from A/Udorn/72 (H3N2), the NS segment from A/California/04/09 (H1N1), and the other 5 segments from PR8. The rPR8 clones differ from the published sequence (GenBank accession no. AF389115 to AF389122) at two positions: PB1 A549C (K175N) and HA A651C (I207L) (numbering from initiating Met). Molecular clone-derived mutants (rPR8 NP:F346S and rPR8 NA:K239R) were generated using standard site-directed PCR mutagenesis. All viruses were rescued by transfecting subconfluent 293T cells with 500 ng of each of the appropriate reverse genetics plasmids using JetPRIME (Polyplus) according to the manufacturer's instructions. Plaque isolates derived from rescue supernatants were amplified into seed stocks in MDCK cells. Working stocks were generated by infecting MDCK cells at an MOI of 0.0001 TCID 50 /cell with seed stock and collecting and clarifying supernatants at 48 hpi. All viral growth was carried out in MEM with 1 g/ml trypsin treated with L-(tosylamido-2-phenyl) ethyl chloromethyl ketone (TPCK-treated trypsin; Worthington), 1 mM HEPES, and 100 g/ml gentamicin. The titers of the virus stocks were determined via standard 50% tissue culture infectious dose (TCID 50 ) assay.
Superinfection assay. For the 6-h sequential-infection group, confluent mammalian cells (MDCK, Vero, A549, or 293T) in six-well plates were infected with rPR8 at an MOI of Ͻ0.3 TCID 50 /cell for 1 h. At 1 h postadsorption, monolayers were washed with phosphate-buffered saline (PBS) and incubated in serum-containing medium. At 3 hpi, neutralizing anti-PR8-HA mouse MAb H17-L2 (5 g/ml) was added to cultures to prevent spread of rPR8. At 6 hpi, monolayers were superinfected with rH3N2 at an MOI of Ͻ0.3 TCID 50 /cell in the presence of H17-L2 (which does not interfere with rH3N2 infection) (see Fig. S2 in the supplemental material). At 1 h postadsorption, monolayers were washed with PBS and incubated in serum-containing medium with H17-L2. At 9 hpi of rPR8 (3 hpi of rH3N2), the medium was changed to MEM with 50 mM HEPES and 20 mM NH 4 Cl to block spread of both viruses. At 19 hpi of rPR8 (13 hpi of rH3N2), monolayers were trypsinized into single-cell suspensions.

Sun and Brooke
® For the 0-h simultaneous infection group, cells were infected with a mixture of rPR8 and rH3N2 at the same MOIs as in 6-h superinfection group. The NH 4 Cl medium was added to block viral spread at 3 hpi, and cells were harvested at 19 hpi.
UV treatment and analysis. rPR8 stocks were placed in six-well plates (500 g/well) on ice. Plates were placed 5 cm underneath a 302-nm-wavelength UVP-57 handheld UV lamp (UVP) and irradiated for 30 s or 60 s. TCID 50 titers and single virion expression patterns of untreated and UV-treated virus were determined on MDCK cells, and the superinfection assays described above were performed using these viruses and rH3N2.
Statistical analysis. Unpaired, two-sided Student's t tests were applied to the data shown in Fig. 1B, 2B, 3A, 3B, and 4F. An unpaired, two-sided Welch's t test was applied to the data shown in Fig. 3D. All statistical analyses were performed with GraphPad Prism 7.0a.