Repeated Antigen Exposure Extends the Durability of Influenza-Specific Lung-Resident Memory CD8 + T Cells and Heterosubtypic Immunity

Lung-resident primary memory CD8 + T cell populations (T rm ) induced by a single influenza infection decline within months, rendering the host susceptible to new heterosubtypic influenza infections. Here, we demonstrate that, relative to single virus exposure, repeated antigen exposure dramatically alters the dynamics of influenza-specific lung T rm populations. Lung T rm derived from repeatedly stimulated circulating memory CD8 + T cells exhibit extended durability and protective heterosubtypic immunity relative to primary lung T rm . Parabiosis studies reveal that the enhanced durability of lung T rm after multiple antigen encounters resulted from the generation of long-lasting circulating effector memory (T em ) populations, which maintained the ability to be recruited to the lung parenchyma and converted to T rm , in combination with enhanced survival of these cells in the lung. Thus, generating a long-lasting T rm precursor pool through repeated intranasal immunizations might be a promising strategy to establish long-lasting lung T rm mediated heterosubtypic immunity against influenza. repeated antigen exposure by influenza virus infection in the lung enhances the durability of lung CD8 + T resident memory populations and extends the duration of heterosubtypic immunity against influenza


INTRODUCTION
Based on health and socioeconomic impact, influenza virus infections are a major global health burden (Kondrich and Rosenthal, 2017;Nicholson et al., 2003). This public health burden remains despite the approval of the first influenza vaccine almost 7 decades ago (Barberis et al., 2016). Current vaccine formulations aim for induction of neutralizing antibodies specific for the main surface antigen (hemagglutinin [HA]) of the influenza virus particle (Barberis et al., 2016). However, the HA protein undergoes high rates of mutation (antigenic drift) (Doherty et al., 2006) that enables successful escape from the immunological pressure of vaccination-induced antibodies and dramatically limits vaccine efficacy (Boni, 2008;de Jong et al., 2000). Additionally, reassortment of the segmented influenza virus genome in animal reservoirs can result in new HA sequences (antigenic shift) (Kim et al., 2018) that have not previously circulated in humans and have the potential for pandemic infections (Kim et al., 2018). It has been well documented in humans and rodent models that influenza-specific CD8 + T cells targeting conserved internal proteins of the virus can control virus titers and limit disease development in the absence of neutralizing antibodies (Altenburg et al., 2015;Kreijtz et al., 2007;McMichael et al., 1983;Sridhar et al., 2013). Recent research suggests that the population of lung-resident CD8 + T cells (T rm ) induced by primary influenza infection plays a critical role in such heterosubtypic immunity (HI) (Hogan et al., 2001;Slütter et al., 2017;Wu et al., 2014). Thus, inducing a potent and long-lasting influenza-specific T rm population should be considered as a potentially useful vaccination target.
Waning of protection is one of the main limitations of T cellmediated heterosubtypic immunity after primary influenza infection (Liang et al., 1994;Slütter et al., 2017;Wu et al., 2014). We and others have shown that the gradual loss of protection closely correlates with the decrease in size of the influenza-specific lung T rm population (Slütter et al., 2017;Wu et al., 2014). Mechanistically, lung T rm cells undergo increased apoptosis that, in combination with time-dependent decreases in recruitment and conversion of circulating T rm precursors, limit the longevity of influenza-specific T rm in the lung (Slütter et al., 2017€ ). Of note, the vast majority of published studies addressing the formation and maintenance of influenzaspecific lung T rm , are based on a "single-exposure" model (Slütter et al., 2017;Takamura et al.,€ 2016;Wu et al., 2014). This represents an important limitation given the repetitive, seasonal nature of influenza infections and the current approaches of yearly vaccine applications. Thus, it is pivotal to understand how repeated influenza antigen encounters impact the dynamics of lung T rm and, consequently, the longevity of heterosubtypic immunity.

Experimental Model
To study the impact of repeated influenza infections on lung T rm , we initially infected C57BL/6 mice with an antibody escape variant of the mouse adapted H1N1 A/PR/8, designated SEQ12 (Das et al., 2013;Van Braeckel-Budimir et al., 2017), followed at 60-day intervals by infection with X31 (H3N2), and then a high dose of the parental A/PR/8 (PR8) virus. Despite the careful selection of viruses, sequential infection with SEQ12 and X31 induced antibodies that neutralized PR8 and no virus was recovered after the tertiary challenge (not shown). Furthermore, repeated infections may lead to generation of memory populations with different antigen exposure history, by incomplete recruitment of existing memory CD8 + T cells together with de novo recruitment of naive CD8 + T cells (Masopust et al., 2006;Wirth et al., 2010). Thus, we sought to develop a model that would allow us to generate distinct populations of T rm with defined numbers of antigen encounters after lung infection.
To address these issues, we took advantage of a previously published model based on adoptive transfer (AT) of TCR transgenic CD8 + T cells (Wirth et al., 2010). Naive Thy1.2 C57BL/6 mice were seeded with low numbers (10 4 ) of naive spleenderived transgenic Thy1.1 P14 cells and subsequently intranasally (IN) infected with recombinant PR8 expressing the GP33 epitope of lymphocytic choriomeningitis virus (LCMV) recognized by P14 cells (PR8-GP33). Ninety days after infection, 10 5 spleen-derived primary memory (1°M) P14 CD8 + T cells were transferred into new naive congenic recipients followed by IN infection with PR8-GP33. Simultaneously, another group of naive recipients was seeded with naive P14 cells and infected IN with PR8-GP33. Adoptive transfer and viral infections were repeated until we simultaneously obtained 1°M and quaternary memory (4°M) P14 cells ( Figure 1A). Of note, based on published data indicating that 3°M cells proliferate less than naive CD8 T cells in response to systemic infection (Wirth et al., 2010) and in order to achieve similar frequencies of circulating 1°M and 4°M P14 cells, mice received 10 5 3°M P14 cells. This model not only permitted isolation and transfer of P14 cells with a known antigen exposure history but also incorporated the recently established notion that circulating memory T cells can be recruited to tissues and generate T rm after repeated infections (Beura et al., 2018;Park et al., 2018). Figure S1A depicts selected phenotypic characteristics of naive and 3M spleenderived P14 which were the precursors of 1M and 4M lung T rm . Of note, the circulating precursor populations used for adoptive transfer lacked cells expressing both CD69 and CD103, which represent the canonical markers of lung T rm (Wakim et al., 2013;Wu et al., 2014). As expected, naive and 3°M populations exhibited differential expression of a number of cell surface markers, including CD44, CD11a, CD62L, and KLRG1. To verify that the different numbers of transferred cells did not impact the overall course of influenza infection and to confirm that both 1°M and 4°M cells developed in similar conditions, we assessed the lung PR8-GP33 titers at different days after infection. As depicted in Figure S1B, titers measured 4 and 6 days after infection were similar in recipients of naive or 3°M P14 and in the same range as titers measured in mice receiving no adoptive transfer prior to infection. This finding suggested that the input of circulating naive or 3°M P14 did not alter the course of infection and that 1°M and 4°M cells developed under similar infection conditions. For clarity, although all of the recipient mice were infected with influenza only once, we will refer to mice bearing 1°M P14 cells as 1°M mice, and those bearing 4°M P14 cells as 4M mice.

Multiple Antigen Encounters Enhance the Durability of Influenza-Specific Lung T rm
To determine whether multiple antigen encounters altered the durability of lung T rm , lungs of 1°M and 4°M mice were analyzed at different time points after the PR8-GP33 infection. To discriminate circulating cells from those within the tissue parenchyma, we performed intravascular (IV) exclusion based on IV injection of a fluorescently labeled pan-lymphocyte marker (CD45.2) 3 min prior to euthanasia as described (Galkina et al., 2005;Teijaro et al., 2011). T rm were defined as CD8 + IV CD69 + CD103 + based on our previous results (Slütter et al., 2017). As described earlier (Slütter et al., 2017;Wu et al. 2014), 1°M lung T rm populations gradually waned over time, with only a few hundred cells detectable in the lung parenchyma 5 months after influenza exposure ( Figure 1B). In sharp contrast, after an initial ~3-fold decline between day 30 and day 50 postinfection (p.i.), 4°M T rm numbers stabilized and were maintained at >10,000 cells/lung out to day 150 p.i. As a result, the 4°M T rm population was >50-fold larger than the 1°M T rm population at day 150 after infection ( Figure 1B). Strikingly, 4°M T rm could be readily detected as late as day 450 p.i., a time point when 1°M T rm have essentially disappeared ( Figure 1C). Thus, multiple antigen encounters after lung infection extended the durability of influenza-specific lung T rm .

Antigen Exposure History Defines the Tissue-Specific Distribution of Influenza-Specific CD8 + T Cells without Altering the Basic T rm Phenotype
Previous studies have emphasized the impact of repeated systemic infection on fundamental biological characteristics of memory CD8 + T cells and their distribution in tissues (Jabbari and Harty, 2006;Masopust et al., 2006;Rai et al., 2014;Vezys et al., 2009;Wirth et al., 2010). Thus, we assessed whether and how the history of influenza antigen encounter impacted the distribution of memory CD8 + T cells. To be able to study memory cell distribution while allowing for competition for distinct niches, we generated 1M and 4M cells in the same host ( Figure S2A). For this purpose, naive Thy1.2/1.2 C57BL/6 recipients were seeded with mixture of 10 4 Thy1.1/1.2 naive P14 and 10 5 Thy1.1/1.1 3°M P14 cells and subsequently IN infected with PR8-GP33. Mice were analyzed at day 90 p.i., and the distribution of 1°M and 4°M cells was assessed in the spleen, blood, and lungs ( Figure S2B). While the distribution of 1°M and 4°M cells was similar in the blood and spleen, 4°M cells were the dominant population in the lung, with slightly stronger preference for the parenchymal (IV − ) niche compared to the vasculature (IV + ) niche ( Figure S2B).
Furthermore, we investigated whether repeated influenza antigen exposures altered some of the important phenotypic properties of lung T rm . For this purpose, we evaluated expression of CD103, CD69 (classical T rm markers) (Schenkel and Masopust, 2014;Wu et al., 2014), Eomes (a key transcription factor downregulated in T rm ) (Mackay et al., 2015), CX3CR1 (expressed by T rm precursors) (Gerlach et al., 2016), and CXCR3 (facilitates homing to the lung/airway) (Slütter et al., 2013 ) on 1°M and 4°M P14 generated in the same host ( Figures  S2A and S2C). Overall, we observed no substantial difference in selected phenotypic characteristics between 1°M and 4°M lung T rm . Most 1°M and 4°M IVCD103 + cells expressed CD69, had low expression of Eomes and CX3CR1, and were CXCR3 + . These data demonstrated that repeated influenza antigen exposure enhanced 4°M entry into the lung parenchyma, where some of the cells developed T rm phenotypic characteristics similar to those of their 1°M T rm counterparts. Thus, major changes in canonical T rm properties did not appear to account for the persistence of 4°M T rm in the lung parenchyma.

Repeated Influenza Antigen Exposures Extend the HalfLife of T rm in the Lung by Increasing Cell Survival
As shown previously, 1°M T rm have a relatively short half-life in the tissue parenchyma and exhibit higher frequencies of cells with active caspase 3/7 and decreased Bcl-2 expression compared to 1°M T rm in skin (Slutter et al., 2017 ). In the light of these findings, we hypothesized that the 4°M T rm population could exhibit prolonged survival in the lung relative to 1°M T rm , leading to their longer persistence. To probe this notion, we labeled lung T rm 90 days p.i. by IN inoculation of carboxyfluorescein diacetate succinimidyl ester (CFSE) (Legge and Braciale, 2003;Slütter et al., 2017 ). This approach labels all of the cells in the lung, with negligible labeling of cells in the other tissue compartments, and loss of CFSE signal with time relates to loss of labeled cells, not dilution due to proliferation (Slütter et al., 2017). To determine the initial level of labeling, we evaluated CFSE intensity in the lung T rm at 1 hr post-inoculation ( Figure 2A). At this time point, 100% of lung T rm were CFSE + ( Figure 2A). When mice were analyzed 120 hr after labeling, only 20% of 1°M T rm were CFSE + , confirming the relatively short survival of this population in the lung. A significantly (p = 0.0127) larger fraction (>30%) of 4°M T rm were CFSE + at the same time post-labeling ( Figure 2A), suggesting prolonged survival of these cells in the lung. Additionally, 4°M T rm exhibited higher expression of Bcl-2, relative to 1°M T rm ( Figure  2B), consistent with increased survival of this population in the tissue parenchyma. Finally, we observed a modest, but significant (p < 0.05) increase in the proportion of 1°M relative to 4°M T rm cells staining positive with Flica reagent, suggesting increased proapoptotic caspase 3/7 activity ( Figure 2C). Together, these results suggested that enhanced durability of 4°M T rm in the lung parenchyma was due, at least in part, to increased survival. However, the differences in each "survival" assay were modest and are unlikely sufficient to entirely explain the enhanced durability of 4°M T rm .

Repeated Influenza Antigen Exposures Stabilize the Circulating T rm Precursor Pool
We previously showed that, despite their short half-life after CFSE labeling, 1°M T rm are sustained for some time by recruitment to the lung of circulating effector memory (T em ) precursors, some of which undergo phenotypic and functional conversion to T rm (Slütter et al., 2017 ). As a consequence of differentiation of circulating 1°M memory populations from T em to primarily central memory (T cm ) with time, replenishment of lung T rm eventually ceases and the population declines. Thus, we hypothesized that, due to their increased antigen exposure history, circulating 4°M may exhibit a sustained T em population allowing for prolonged recruitment and conversion to lung T rm . To probe this hypothesis, we took advantage of the mouse parabiosis model. 1°M mice, bearing Thy1.1/1.1 circulating memory and lung T rm cells, and 4°M mice, bearing Thy1.1/1.2 circulating memory and lung T rm cells, were surgically joined 90 days post-PR8GP33 infection ( Figure 3A). Three weeks later, parabiotic pairs were analyzed for the distribution of 1°M and 4°M P14 cells in blood, spleens, and lungs ( Figure 3A). Of note, similar frequencies of circulating P14 were found in the blood and spleen of both groups ( Figures 3B and 3C), indicating that equilibrium had been reached. However, this equilibrium did not reflect a 1:1 distribution of 1°M and 4°M circulating cells. Instead, both groups of mice exhibited 75% 1°M P14 and 25% 4°M P14 ratio in the blood and spleen ( Figures 3B and 3C). In contrast, a higher frequency of memory P14 was recovered from the total lung and IV − fraction of the lung in the 4°M parabiont ( Figure 3D). As expected from prior work showing that late 1°M loses the ability to home to the lung parenchyma and convert to T rm (Slütter et al., 2017 ), and despite the dominance of 1°M in the circulation, T rm in the lungs of the 4°M parabiotic partner were exclusively 4°M T rm . In contrast, despite their reduced presence in the circulation, a clear population of 4°M T rm could be found within the T rm population of the 1°M parabiotic partner ( Figure 3D). These data suggested that circulating 4M exhibited a prolonged capability to home to the lung parenchyma and convert to T rm .
The contribution of 4°M T rm P14 to the T rm compartment in 1M mice at 3 weeks of parabiosis was low (5%). Thus, we investigated whether representation of these cells would increase with additional time of parabiosis. For this purpose, 1°M and 4°M parabiotic partners were euthanized 13 weeks after joining, and distribution of P14 cells in lung tissue was assessed. Extended time of parabiosis did not change the contributions of 1°M and 4°M cells to blood and spleen compartments of parabiotic partners (not shown). Despite the added time of parabiosis, we still detected no 1 T rm in the lungs of the 4°M parabiotic partners. However, the representation of 4°M T rm P14 increased from 5% to 30% in the lungs of the 1°M mice with the additional parabiosis time. These data suggested sustained accumulation of 4°M-derived T rm from the circulating 4°M pool with time.
Consistent with this notion, splenic 1M progressed from 50% T em (CD62L lo ) at day 50 p.i. to <10% T em at day 150 p.i., and this resulted in a decrease in total numbers of splenic T em ( Figure S3A). In sharp contrast, the splenic 4°M population remained CD62L lo and the T em population was numerically stable in the same time frame ( Figure S3A). This trend was extended well beyond day 150, as even 7 months p.i. circulating 4°M continued to express a T em phenotype (CD62L lo ) and a permissive transcription factor profile (Eomes lo ), whereas the 1°M population was exclusively T cm (CD62L hi Eomes hi ) at this late time point ( S3B). Although Tbet was modestly upregulated in 4°M, the Tbet regulated homing receptor CXCR3 was not differently expressed between splenic 1°M and 4°M ( Figure S3C). These data were consistent with the notion that general impairment of lung-homing capacity was unlikely to underlie the failure to maintain 1°M T rm (Slutter et al., 2017€ ). In summary, circulating 4°M CD8 + T cells, due to their prolonged display of T em properties were recruited to the lung and converted into T rm for an extended period of time, compared to circulating 1°M. Given the modest evidence for increased survival, sustained replenishment of lung T rm from the circulatory memory pool likely accounted for the enhanced durability of 4°M lung T rm .

Repeated Influenza Antigen Exposures Extend the Duration of the Heterosubtypic Protection
Waning of 1°M T rm correlates strongly with loss of heterosubtypic immunity (Slütter et al., 2017;Wu et al., 2014 ). Our results show that multiple exposures of circulating CD8 + T cell populations to pulmonary influenza infection extended the durability of lung T rm far beyond that observed with 1°M lung T rm . Thus, our next step was to investigate whether extending the durability of lung T rm through multiple antigen exposures resulted in prolonged heterosubtypic immunity. For this purpose, 1°M mice and 4°M mice, from the cohorts analyzed in Figure 1C, were challenged with a high dose of the heterologous X31-GP33 virus 120 days post-PR8-GP33 exposure ( Figure 4A). Heterosubtypic immunity was evaluated by measuring X31-GP33 titers in the lung of challenged mice ( Figure 4A).
It is important to note that both groups of mice were exposed to PR8-GP33 only once, both contained similar percentages of endogenous 1°M CD8 + T cells recognizing the NP366 influenza epitopes in their circulation, and challenge took place at a late time point when protection by endogenous lung T rm would have waned ( Figure 4B). Thus, potential differences in protection would be due to differences in lung P14 T rm populations, which are lost in 1°M mice and sustained in 4°M mice ( Figure 1C). While the magnitude of the endogenous influenza-specific 1°M responses were similar, in line with previous data (Figure 3) 1M mice had higher circulating P14 frequencies than 4M mice ( Figure 4B) (Wirth et al., 2010). As expected at this late time point, 1°M memory mice showed only a modest reduction of X31-GP33 virus titers compared to naive infected mice, which did not reach statistical discernibility. In contrast, 4°M mice had 99% less virus than the naive infected group ( Figure 4C). Thus, the extended durability of 4°M T rm in the lung enabled superior heterosubtypic immunity at time points after infection when protection mediated by 1°M T rm had waned.

DISCUSSION
Heterosubtypic immunity against influenza virus has been extensively studied in a singleinfection mouse model of the disease (Budimir et al., 2012;Christensen et al., 2000;Furuya et al., 2010;Liang et al., 1994;Nguyen et al., 1999;Slutter et al., 2017;€ Wu et al., 2014). Recent studies show that heterosubtypic immunity mediated by 1°M CD8 + T cells wanes over time, as a consequence of gradual loss of lung T rm cells (Slutter et al.,€ 2017;Wu et al., 2014). In the context of annual influenza outbreaks and the potential for yearly immunizations, it has remained unknown how multiple antigen exposures influence the formation and maintenance of lung T rm and subsequent T rm -mediated heterosubtypic protection. To address these important knowledge gaps, we modified a multiple antigen exposure strategy previously used to assess responses to systemic infections (Wirth et al., 2010), by stimulating circulating memory CD8 + T cell populations with localized lung influenza infections. This approach allowed us to overcome limitations of repeated infections with influenza virus and to study homogeneous populations of memory CD8 + T cells with well-defined influenza exposure history (Masopust et al., 2006;Wirth et al., 2010). Our results revealed that 4°M T rm cells exhibit substantially increased durability in the tissue parenchyma compared to 1°M T rm . Mechanistically, the enhanced durability of 4M cells resulted from improved survival of this population in the lung and a dramatically extended capacity for the circulating 4M cells to enter the lung parenchyma and convert to T rm . These findings strongly suggest that, by inducing a long-lived T em population, multiple influenza exposures license a circulating T rm precursor pool for sustained seeding and conversion in the lung, thus extending the duration of heterosubtypic immunity.
Our prior studies in systemic infection models (Jabbari and Harty, 2006;Nolz and Harty, 2011) showed that the history of antigen exposures defines fundamental biological properties of circulating memory CD8 + T cells, which jointly determine their main function -protection against subsequent infection. For example, 1°M and 2°M CD8 + T cells show differential capacity to protect against acute and chronic infection, with 2°M cells establishing better control of acute LCMV Armstrong or Listeria monocytogenes infection, but failing to protect against chronic LCMV clone 13 infection (Jabbari and Harty, 2006;Nolz and Harty, 2011). Importantly, this difference in protective capacity was driven by differential tissue preferences of 1°M and 2°M CD8 + T cells, determined by antigen exposure history. Furthermore, the previously described dramatic changes in gene expression signature between memory T cells with different antigen exposure histories, strongly suggests that a complex network of fundamental properties, defined by the number of antigen encounters, determines the protective capacity of memory CD8 + T cells. Here, we show that, similar to systemic infections, repeated exposures of circulating memory CD8 + T cells to influenza infection in the lungs determines their ability to persist as T rm and provide heterosubtypic protection against lung infection. Although further work will be needed to address possible differences in the per-cell quality of protection between 1°M and 4°M lung T rm , it is clear that 4°M T rm , on a population level, provide superior protection, as they can successfully control virus replication even very late after initial infection, when protection mediated by 1°M cells is lost ( Figure 4C).
We have previously shown that maintenance of lung 1°M T rm is a dynamic process, defined by the short half-life and high apoptotic rate of these cells on one side and replenishment from circulating memory populations on the other (Slütter et al., 2017). Importantly, not all circulating memory populations exhibit the same capacity to home to the lung and convert to T rm (Slütter et al., 2017€ ). We have identified the population of circulating T em cells (CD62L lo ) as precursors for lung T rm cells, as their circulating T cm counterparts were not able to successfully seed the tissue and convert to T rm . As 1°M CD8 + T cells differentiate toward a uniform T cm phenotype with time (Martin et al., 2015), the number of potential T rm precursors decreases, resulting in a gradual loss of 1°M lung T rm . Of note, we previously showed that multiple encounters with systemic pathogens generate memory populations with sustained characteristics of T em (Wirth et al., 2010). We now show that CD8 + T cells that undergo multiple encounters with influenza-derived antigens after lung infections exhibit similarly sustained T em characteristics ( Figure S3G), allowing for extended recruitment of these cells to the influenza-experienced lung and their conversion to T rm (Figures 3A-3F). Furthermore, we were able to find well-defined, albeit reduced populations of 4°M lung T rm as late as 450 days after infection ( Figure 1C), suggesting that their lung persistence, although dramatically extended, may still be limited. Given that circulating 4M populations express exclusively a T em phenotype as late as 210 days after influenza exposure, it is likely that a slow transition to T cm is not the sole explanation for the eventual decline of 4°M lung T rm . At this point, we cannot rule out that possible time-dependent changes in the architecture of the lung tissue and immunological landscape after influenza infection have an important impact on long-lasting maintenance of the lung T rm population. These important knowledge gaps merit further in-depth studies in the realm of lung physiology in the context of influenza virus infection.
It is important to acknowledge that our study did not compare lung T rm cells that encountered influenza once or multiple times due to repeated infection of the same host. Instead, we build on recent data describing the capacity of circulating memory CD8 + T cells to be recruited to the site of infection and convert to T rm (Beura et al., 2018;Park et al., 2018). Thus, we describe fundamental differences between T rm populations that were established by recruitment from the pool of circulating CD8 + T cells that were exposed to influenza lung infections once or multiple times, in an otherwise naive host. Understanding whether and how multiple influenza exposures shape the already existing lung T rm population remains an important knowledge gap and a major challenge, given the limitations of the available animal models and potential ethical and sampling concerns with multiple experimental challenges of humans.
Current influenza vaccines are highly purified subunit or splitvirion formulations (Soema et al., 2015;Wong and Webby, 2013), composed mainly of surface viral antigens (HA and neuraminidase [NA]), while conserved internal antigens (e.g., nucleoprotein [NP]) represent only a fraction of a split-virus vaccine (Soema et al., 2015). These vaccines aim at inducing neutralizing antibodies against surface antigens, such as HA. High mutation rate and antigenic plasticity of the main antibody targets can severely compromise the efficacy of seasonal influenza vaccines and preclude their efficacy against pandemic influenza (Boni, 2008;Kim et al., 2018). Thus, much effort is currently directed toward the concept of universal influenza vaccines. Although CD8 + T cells targeting conserved viral epitopes cannot provide sterilizing immunity to influenza infection, cumulative evidence points toward induction of lung-residing memory CD8 + T cells as a promising strategy to provide heterosubtypic immunity that could broadly limit influenza disease in humans (Kumar et al., 2017;Pizzolla et al., 2018). Our findings suggest that repeated lung immunizations with formulations containing conserved CD8 + T cell target epitopes, might generate durable lung T rm populations capable of sustained heterosubtypic immunity.

CONTACT FOR REAGENT AND RESOURCE SHARING
Further information and requests for resources and reagents should be directed to and will be fulfilled by the Lead Contact, John T. Harty (john-harty@uiowa.edu). (PBS), and mice were inoculated with the specific viral dose after induction of anesthesia by ketamine + xylazine (80-100 mg/kg + 10-12.5 mg/kg) injection.

Tissue Preparation, T Cell Analysis, Flow
Cytometry-For quantification of antigen-specific memory CD8 + T cells, influenza immune mice were euthanized at the indicated time points p.i., 3 min after IV administration of 2 mg fluorescently labeled anti-CD45.2 antibody (IV exclusion) (Slutter et al., 2017€ ). Spleen and PBL were harvested and processed into a single-cell suspension. Lungs were cut into small pieces and incubated in the presence of collagenase (125 U/ml) and deoxyribonuclease (0.1 mg/ml) for 1h at 37C. After incubation a single-cell suspension was obtained by physical homogenization through a 70-mm cell strainer. Leukocytes were enriched by centrifugation in 35% Percoll (GE Healthcare) diluted in HBSS and red blood cells were lysed using Vitalize (BioE Under ketamine anesthesia hair was removed from one side of the first mouse by shaving, starting approximately 1 cm above the elbow to 1 cm below the knee. Hair was then removed from the opposite side of the second mouse.
On the day of the surgery mice were anesthetized using ketamine/xylazine and injected subcutaneously with Meloxicam to induce analgesia. Once anesthesia was induced, mice were transferred to the surgery area, and the shaved skin was aseptically cleaned with betadine and alcohol. Mice were placed back to back, on their sides, with adjacent shaved areas facing up. To avoid contamination of the surgical area, mice were covered with a sterile surgical drape exposing only the surgical area. To prevent hypothermia, mice were positioned on the heat pad for the duration of the surgery. Longitudinal skin incisions were performed on the shaved side of each animal, starting at 0.5cm above the elbow to 0.5cm below the knee joint. Skin was gently detached from the subcutaneous tissue and the separation was performed along the entire incision. After attachment of knee and elbow joints with non-absorbable 3-0 suture the skin of the two animals was connected with a continuous absorbable 5-0 suture starting ventrally from the elbow toward the knee. Once ventral skin attachment was completed with a double surgical knot, the suture was continued dorsally ending with a double surgical knot. Bupivacaine was applied locally to both sutures and each mouse was subcutaneously injected with sterile saline to prevent dehydrataion.
To prevent bacterial infections, animals were fed an antibiotic-containing diet starting from day 2 prior to surgery until day 10 post surgery. Mouse recovery was followed daily for two weeks after the procedure.

QUANTIFICATION AND STATISTICAL ANALYSIS
Statistical differences between two study groups were evaluated using an unpaired, twotailed t test. Statistical differences between more than two study groups (single factor) were evaluated using one-way ANOVA with Tukey's multiple comparison post hoc test. Two-way ANOVA with Sidak's multiple comparison post hoc test was used to assess comparison between more than two groups based on more than one parameter (multiple factors). Statistical significance was assigned as *p < 0.05, **p < 0.01, ***p < 0.001 and ****p < 0.0001. Statistical analyses were performed using Prism 7 software (GraphPad).

Supplementary Material
Refer to Web version on PubMed Central for supplementary material.