Influenza Virus Infects and Depletes Activated Adaptive Immune Responders

Abstract Influenza infections cause several million cases of severe respiratory illness, hospitalizations, and hundreds of thousands of deaths globally. Secondary infections are a leading cause of influenza's high morbidity and mortality, and significantly factored into the severity of the 1918, 1968, and 2009 pandemics. Furthermore, there is an increased incidence of other respiratory infections even in vaccinated individuals during influenza season. Putative mechanisms responsible for vaccine failures against influenza as well as other respiratory infections during influenza season are investigated. Peripheral blood mononuclear cells (PBMCs) are used from influenza vaccinated individuals to assess antigen‐specific responses to influenza, measles, and varicella. The observations made in humans to a mouse model to unravel the mechanism is confirmed and extended. Infection with influenza virus suppresses an ongoing adaptive response to vaccination against influenza as well as other respiratory pathogens, i.e., Adenovirus and Streptococcus pneumoniae by preferentially infecting and killing activated lymphocytes which express elevated levels of sialic acid receptors. These findings propose a new mechanism for the high incidence of secondary respiratory infections due to bacteria and other viruses as well as vaccine failures to influenza and other respiratory pathogens even in immune individuals due to influenza viral infections.


Influenza Virus Infection Decreases Vaccine Response in Human PBMCs
Cryo-preserved human peripheral blood mononuclear cells (PBMCs) collected from healthy adult vaccinees were treated as follows following polyclonal stimulation to increase the number of activated responders. PBMCs were infected with 5 multiplicity of infection (MOI) of CA09-RFP (A/California/07/09 backbone; H1N1) or NS1-GFP (A/Puerto Rico/8/34 backbone; H1N1) -which fluoresce during viral replication and the number of vaccine-specific ASC were measured by ELISPOT (Figure 1A-D). The PR8 strain was used to represent the seasonally circulating H1N1, often predominating the influenza season. The infection of PBMC by H1N1 was confirmed by flow cytometry 24 h post-infection (p.i.), allowing time for virus to fully replicate and trigger cell death ( Figure 1E-H). A reduction (p < 0.01) in total IgG ASCs as well as those specific to H1N1 The number of measles-specific ASCs were already low in uninfected controls, and the modest decrease did not reach statistical significance. Progressive decline in circulating measles-specific antibodies post-vaccination has been shown. [32][33][34][35][36] Furthermore, the subjects from whom we obtained PBMC were all in the age group of 30-65. People born in 1960s through 80s used to receive one dose of measles vaccine and may have had poor memory responses to measles. Poor vaccine-specific immunity due to one dose of vaccine and waning may have contributed to the low levels of measles-specific ASCs observed in this study. Unlike influenza A viruses, influenza B viruses have been shown to mutate 2-3 times slower. [37] Hence, B components in influenza vaccine tend to remain constant for several seasons potentially leading to reduced immunogenicity due to repeat vaccination as antigenic distance between B components between seasons is negligible. For example, B/Brisbane/60/2008like virus has been the B component of influenza vaccines since 2009-10 through 2017-18 influenza seasons. Thus, the low responses observed for influenza B virus component could be due to reduced responses owing to repeat vaccination effect. All antigen-specificities saw a reduction with infection, with subtle differences that may be due to differing initial ASC. ) and infected with NS1-GFP, with A) total numbers of IgG-expressing ASCs and B) antigen-specific IgG ASCs in uninfected controls and infected samples are shown, matched pairs connected in red. C) Percentage loss of total IgG and D) antigen-specific IgG for each individual, comparing paired infected and uninfected samples from each individual. The median and spread of up to 8 subjects are shown (where cells and titers were available), and (*), (**) indicating Student's paired t-test P < 0.05, P < 0.01, respectively, calculated from total cell comparison ( Figure 1A,B). E) Representative B cell subset flow plot highlighting percent NS1-GFP infected for both control and day 1 p.i. PBCS, and F) bar graph showing averages across samples. G) Representative T cell subset flow plot highlighting percent NS1-GFP infected for both control and day 1 p.i. lung cells, and H) bar graph showing averages across PBMC samples. The mean (±SD) of 3 subjects are shown, comparing Activated B cells and plasmablasts with naïve B cells, Activated CD4 with naïve CD4, and Activated CD8 and NKT cells with naïve CD8, and (*), (**) indicating Student's paired t-test (where appropriate to single comparisons) or One-way ANOVA (Dunnett's multiple comparison) of P < 0.05, P < 0.01, respectively. Total cells were excluded from statistical analysis. populations tended to approximate or outnumber activated populations even after stimulation, and yet active responders were disproportionately impacted.
To confirm this targeting of activated lymphocytes also occurred in the lungs, BALB/c mice were immunized intranasally with 10 µg of inactivated H1N1 monovalent vaccine and boosted 7 days later, allowing for the development of adaptive responders. Lymphocytes were isolated from lungs on day 14 and infected in vitro with 5MOI of NS1-GFP. NS1-GFP fluorescence was measured by flowcytometry at 24 h p.i. in infected or uninfected lung lymphocytes, both from pre-immunized mice, to ensure equivalent numbers of influenza-specific immune responders in both samples. Significantly (p < 0.01), activated B cells (CD69+) and plasmablasts (B220+CD138+), (Figure 2A,B), activated CD4 and CD8 (CD62L-CD44hi), and NKTs (CD3+DX-5+) ( Figure 2C,D), were preferentially infected by NS1-GFP in pre-immunized mice (gating strategy, Figure S2: Supporting Information).
To further investigate viral targeting specificity, BALB/c mouse cohorts were immunized and boosted with 10 9 plaque-forming units (pfu) of non-replicating human adenovirus (HAd) intranasally before NS1-GFP infection. Just as in lymphocytes activated by influenza vaccine, those HAd-activated B and T cells and NKTs were similarly infected at higher levels (p < 0.05) than naïve lung cells ( Figure 2E-H), showing that this targeting of immune responders is independent of the activating antigen. These results were further confirmed with in vivo infection of HAdimmunized mice with 5 MOI NS1-GFP virus, where the same pattern is observed ( Figure 2I-L). As in human PBMCs, naïve B and T cells tended to approximate or outnumber active responders (particularly plasmablasts), but these populations are disproportionately infected. The observed impact appears to be lower in mice than in humans, even at the same MOI, likely due to the use of the same human NS1-GFP virus across our experiments rather than a mouse-adapted influenza strain.

Influenza Virus Infects and Lyses Activated Immune Cells
To assess the outcome of virus infection of immune responders, B-and T-cell subsets were isolated from pooled murine lungs at day 7 post-HAd immunization and infected in vitro with 5 MOI of PR8 virus for 24 h. The levels of viral replication (determined by NS1 expression) within each subset of cells were measured by PCR, confirming viral replication in activated B and T cells and NKTs ( Figure 2M,N). This was also confirmed in sorted B and T subsets isolated from lungs of mice immunized with HAd and infected at day 7 with NS1-GFP in vivo ( Figure 2O). Further, these cells were shown to increase cell apoptosis (compared to uninfected cells), measured by 7AAD and Annexin V costaining ( Figure 2P,Q), demonstrating that viral targeting directly reduces the numbers of adaptive immune responders. This also indicated that no other external factors (such as cell-cell contact or cytokines) are necessary for cell death in infected lymphocytes.

High Levels of 2,3 Sialic Acids Linked to Viral Entry in Mice and 2,6 Sialic Acids in Human PBMC
In an earlier study of adaptive cell surface labeling, it was observed that H1N1 hemagglutinin and neuraminidase proteins bound increasingly to targeted cells [unpublished data], leading us to hypothesize that increased binding and/or viral entry might be responsible for the observed cell targeting seen here. It is wellestablished that the influenza virus enters the cell via sialic acid receptors on the cell's surface. This occurs through the 2,6 sialic receptor in humans and the 2,3 in mice. [27] To probe the potential mechanism of this observed viral targeting, levels of 2,3 and 2,6 sialic acids were measured at day 7 on the relevant B and T subsets following infection with 10MID 50 of PR8 or 10 9 pfu of HAd. Expression was measured via binding of lectins, Maackia amurensis lectin I (MAA-1) or II (MAA-2) to 2,3 sialic acids, or Sambucus nigra lectin (SNA) to 2,6 sialic acids. Activated B and T, as well as NKTs, revealed higher levels of 2,3 sialic acids compared to naïve lymphocytes in response to either PR8 ( Figure 3A-D) or HAd ( Figure 3E-H) infection, as well as in response to immunization with Keyhole limpet hemocyanin (data not shown). Smaller differences were observed in SNA expression, along similar lines. Increases in the sialic acid expression on activated cells were observed at all-time points and varying antigen levels (data not shown). Influenza viruses typically target 2,6 sialic acids in humans, [38] so SNA-binding was used to measure an increase in 2,6 sialic acids in activated human B and T and NKTs

Lectin Binding of Sialic Acids Blocks Viral Adhesion
To further demonstrate that the expression of sialic acids is the viral entryway, increasing amounts of lectins were added to splenocytes in vitro for 30 min before infection to show that they could block viral binding. A dose-dependent effect was observed as increasing amounts of MAA-1, MAA-2, or SNA (1:200, 1:100, and 1:20 dilutions in media) were incubated with murine cells before infection with 5MOI PR8-FITC ( Figure 3L). Prior incubation of cells with 1:20 MAA-1 lectin reduced viral adhesion to half compared to control cells. A similar, though a lesser effect, was observed with MAA-2 and SNA.

In Vivo Influenza Challenge Reduces Vaccine Responses in Mice
To demonstrate the in vivo effects of viral targeting of adaptive immune responders, cohorts of mice were immunized and boosted with 10 µg of influenza vaccine intranasally or with PBS control before challenge with 10MID 50 of PR8 at day 7 post-boost. The influenza-specific adaptive response was allowed to develop before the influenza challenge to ensure that only direct effects on adaptive responders were observed. On day 2 post-challenge, lung cells from challenged and control cohorts were isolated, and antigen-specific and total ASCs were measured by ELISPOT. A decrease in the percentage of H1N1-specific IgG ASCs (p < 0.05) was found after the PR8 challenge ( Figure 4A) as well as in total IgG, and IgM ASCs in H1N1 vaccine immunized mice (Figure 4E,F). A decrease in H3N2-specific IgG ASCs (p < 0.05) was also observed ( Figure 4A). There is no significant increase in PR8responding ASCs to account for this diminishment. Next, a non-replicating HAd was used to assess the effect of influenza on HAd-specific responders to mimic the impact of influenza on recent vaccination and model the impact on respiratory co-infection. [3,5] Cohorts of mice were immunized and boosted intranasally with 10 9 pfu of HAd before challenge with 10MID 50 of PR8 at day 7 post-boost, as before. A reduction (p < 0.05) in the percentage of HAd-specific IgG ASCs was found after the PR8 challenge ( Figure 4G), as well as a decrease (p < 0.05) in total IgG ASCs ( Figure 4K). While declines were trending in both HAd-specific IgM ASCs ( Figure 4H) and total IgM ASCs ( Figure 4L), they were not significant. No significant numbers of PR8-specific ASCs were measured in challenged or control cohorts ( Figure 4I,J), so the recruitment of PR8-specific ASCs does not explain the decrease in HAd-specific ASCs.
In a separate set of experiments, mice were immunized and boosted on day 14 post-priming intramuscularly with PREVNAR-13 or intranasally with PspA, and then challenged with 10MID 50 of PR8 or PBS control on day 28. On day 34 (day 6 post-PR8 challenge), all mice were challenged with the S. pneumoniae strain A66.1 (serotype 3) to simulate human infection with pneumococcus following influenza infection. Morbidity was assessed over the following 48 h, and the mice were sacrificed to determine A66.1 lung titers ( Figure 4S). Unvaccinated animals experienced pulmonary bacterial growth whether they had been challenged with PR8 or not, but both PREVNAR-13and PspA-vaccinated groups conferred protection following bacterial challenge with significant bacterial clearance. However, this protection was ablated in the PR8-challenged group resulting in bacterial replication. Assessment of body weight ( Figure 4T) and clinical score ( Figure 4U) changes confirmed the vaccine protection patterns displayed in the A66.1 lung burden data ( Figure 4S). These findings suggest that influenza virus infection compromises pneumococcal vaccine effectiveness in a co-infection model by infecting and depleting pneumococcal antigen-specific adaptive immune responders.

Discussion
Our results demonstrate that in addition to the influenza virus's suppressive effects on the innate immune system, [25,29] the virus can also directly suppress adaptive immune responders by infecting and killing activated lymphocytes. This targeting likely results through increased expression of sialic acid on the surface of activated lymphocytes. ASCs were reduced by as much as 50% in infected mouse cohorts or human PBMC samples, which could negatively impact vaccine efficacy and leave patients susceptible to a secondary infection. Though our confirmatory findings are in a mouse model, they are suggestive of a mechanism by which many clinical observations of the immunosuppressive effects of influenza vaccination may be explained and warrant further studies in human challenge models.
Notably, our findings suggest a possible mechanism for previous observations demonstrating a decrease of B and T helper numbers in the lymph nodes during influenza infection, [31] as well as evidence of fewer peripheral lymphocytes in humans. [40][41][42] Studies have shown diminished adaptive immune responses during influenza to a cell-free purified protein derivative (PPD) obtained from a human strain of Mycobacterium tuberculosis, candida, mumps virus, and trichophytin were attributed to T cell suppression, [43] weakened Ig production, NK functionality [44] and/or decreased lymphocyte responsiveness/proliferation. [45][46][47] A decrease in circulating lymphocytes, or lymphopenia, is not uncommon in influenza patients. [40][41][42]48,49] However, the reasons for this decline were previously unknown, though low levels of influenza viral RNA   have been detected in the PBMCs of influenza patients. [50] More recently, it has been demonstrated that innate cells and activated B cells were preferentially infected over naïve lymphocytes, although the mechanism is unknown. [51] Susceptibility of NK cells as well as B cells to Influenza A virus was also demonstrated via whole lung imaging using NS1-GFP. [52] Similarly, enhanced susceptibility of activated B cells to H5N1 avian influenza was also observed. These investigators reasoned that increased expression of DC-SIGN on the cell surface might contribute to this and theorized that sialic acid expression might play a role along with the presence of monocytes, though they did not investigate this or any other cell subsets. [53,54] Here, we have shown that the immune-suppressive effect of influenza may occur directly through viral infection and killing of responding immune cells, even in the absence of monocytes -through these and other immune mediators may contribute to additional adaptive cell death in vivo -and that this targeting occurs through increased sialic acid expression. This effect is also observed in innate cells, which may explain previous findings of depleted dendritic cells, macrophages, and NK cells during influenza infection (Figure S3, Supporting Information). [19,[21][22][23][24][25][26][27][28][29][30] Similar observations were made with measles -which has been shown to infect naïve and memory cells, leading to immunological amnesia lasting up to two years. [6,55] In the case of influenza, the impact of this immunosuppressive effect is likely limited to the lungs and respiratory tract, based on the localization of the virus itself, although we are yet to test the impact on other compartments -the spleen, bone marrow, and distant lymph nodes. The observed increases in the sialic acid expression on activated cells are part of a normal immune response to infection and do not appear to be specific to influenza infection/vaccination. The increase in cell surface sialic acids is necessary for cell motility and the prevention of cellular adherence. Previous studies have demonstrated sialic acid loss causes clustering and lymphocytosis in humans, [56] while upregulation increases cell motility and cancer metastasis. [57,58] However, it should be noted that other factors, including other cell surface receptors, may play a role in this targeting as well. While HA has been known to bind sialic acid receptors for cellular entry, NA has been shown to cleave sialic acid from the budding viruses to facilitate their release; however, NA of certain influenza viruses such as N9 have been shown to bind to sialic acids close to enzymatic cleavage site, thus suggesting a potential role along with HA in viral entry. [59,60] Sialic acids are merely one of the earliest factors in viral binding and entry, and are targets of both Influenza Hemagglutinin and Neuraminidase, and thus appear to play a significant, early role in this immunosuppressive effect.
Taken together, our findings provide a new mechanism for the reduced efficacy of not only influenza vaccines but also pneu-mococcal vaccines and the higher incidence of secondary bacterial/viral infections during the influenza season, even in vaccinated or immune individuals. Similarly, co-circulation of SARS-CoV-2 [61] during the upcoming influenza season may increase disease severity of COVID-19 even if a vaccine is available. [62] Hence, it may be preferable that any vaccine against respiratory pathogens be given well before the influenza season, ensuring that the recipients have high serological and cellular protective responses. Finally, individuals who lack seroprotective immune responses may benefit from passive immunization during influenza seasons.

Experimental Section
Mice: BALB/c females were purchased from Jackson Laboratories (Bar Harbor, ME) and used beginning at 5 weeks old and were 8 weeks Sinai School of Medicine, New York City, NY47. Viruses were propagated, purified, and stored as described previously. [20] A non-replicating human adenovirus with deletions in E1 and E3 regions (HAd) was provided by Dr. Suresh Mittal, Purdue University, and aliquots stored at −80°C until use.
In Vitro PBMC Infection: Two million PBMCs were stimulated with 1 µg mL −1 of TLR7/8 agonist R848 in the presence of 10 ng mL −1 of IL-2 for 3 days. At 2 days post-stimulation, PBMCs were infected with 5 MOI of NS1-GFP or NS1-RFP for 1 h at 37°C in 200 µL of RPMI or treated similarly without adding virus. The cells were cultured overnight at 37°C in complete RPMI media containing 10% fetal bovine serum (FBS), 5% Penicillin/Streptomycin, 2 × 10 −3 m L-Glutamine. On day 3 post-stimulation, 1 day post-infection, NS1-GFP or NS1-RFP infection was measured via or intramuscular prime/boost with Prevnar, half of the mouse cohorts were challenged with 10MID 50 of PR8 virus at day 14 post-immunization and measured by ELISPOT at day 2 post-challenge. The mean (±SD) of 5 mice per cohort are shown, comparing control and challenge cohorts. In a separate experiment, five-week-old mice were vaccinated by priming (day 0) with intranasal PspA-CoPoP/PHAD (PspA) or PBS control, or intramuscular Prevnar-13 and similarly boosted on day 14. Mice were challenged with 10MID 50 of PR8 virus or control (PBS) at day 28 post-immunization. At day 6 post-viral challenge, all mice received 7.25-10.0 × 10 5 colony-forming units (cfu) of S. pneumoniae (serotype 3 strain, A66.1) in 50 µL PBS, intranasally. Two days post-bacterial instillation (PBI), mice were sacrificed and S) lungs homogenized and A66.1 burden determined by titration on blood agar. T) Bodyweight change from the time of PBI (%), and U) clinical score at the time of harvest were also assessed. Data represent mean ± SD. of n = 6 for each group (performed in 2 experiments of n = 3) except the PBS vaccination/PR8 challenge group experienced 1 death during the 2nd PBI day (n = 5). Data from the lung A66.1 burden data were log transformed for normalization before statistical analysis. With (*), (**) indicating Student's paired t-test P < 0.05, P < 0.01, respectively.