Cutting Edge: Influenza-Induced CD11a^lo Airway CD103⁺ Tissue Resident Memory T Cells Exhibit Compromised IFN-γ Production after In Vivo TCR Stimulation

Memory CD8⁺ T cells generated on challenge with invading pathogens are broadly characterized into three main subsets. These include central memory T cells that localize to secondary lymphoid organs and effector memory T cells that recirculate through nonlymphoid tissues. The third population consists of noncirculating tissue resident memory T cells (T_RM) that are widely distributed in nonlymphoid organs to protect the host against pathogen re-encounter (1, 2). The archetypical T_RM populations are defined by the markers CD103 and CD69 and downregulate the expression of CCR7, CD62L, and sphingosine-1-phosphate receptor 1 to enhance their residence capacity (2). Initially identified using parabiosis experiments, T_RM are distinct from circulating T cells in their unique transcriptional program of tissue residency and, with some exceptions such as the liver, are spared from short-term intravascular (IV) Ab labeling (3–5).

CD8⁺ T_RM occupy the airway and parenchyma of the lung, with cells in the latter location expressing higher levels of the adhesion molecule CD11a for lung entry (6–8). Recently arrived airway CD8⁺ T_RM are CD11a^hi, but they rapidly lose expression of this integrin with time (8). Airway T_RM fail to proliferate on secondary challenge in vivo and are relatively short-lived at a half-life of 14 d (7). Parenchymal CD8⁺ T_RM are retained for longer periods, yet both populations decline relatively faster than T_RM at other mucosal tissues (7, 8). The decline in CD8⁺ T_RM coincides with loss of heterosubtypic immunity to influenza challenge, despite the maintenance of influenza-specific effector memory T and central memory T cells in circulation (9–15). Yet, the mechanistic contribution of these cells in each lung compartment to clearance of secondary respiratory infections has yet to be elucidated.

As a first line of defense, CD8⁺ T_RM at mucosal sites, namely, the skin, were shown to rapidly mobilize and exert effector functions via cytotoxicity and cytokine production. Cytokines that promote a so-called sense and alarm function include IFN-γ and TNF-α that induce the activation of dendritic cells (DCs) and recruitment of circulating T cells, B cells, and NK cells (16–20). In one human study, a large proportion of influenza-specific CD8⁺ T cells in the total lung were CD103⁺ and capable of producing IFN-γ to a far greater extent than their CD103⁻ counterparts (21).

To date, few experimental murine studies have implicated the role of CD8⁺ T cell–derived IFN-γ in mediating protection against respiratory reinfections (7, 8, 22). In the context of CD8⁺ T_RM-specific protection, only one study to date demonstrated that airway CD8⁺ T_RM harboring a CD11a^lo phenotype effectively controlled viral infection on heterologous challenge (7). Airway CD8⁺ T_RM from influenza-experienced wild type (WT) or IFN-γ–deficient mice were intratracheally transferred into naive WT mice with titers measured via quantitative PCR 3 d postchallenge with influenza (7). Naive mice that received airway influenza-specific CD8⁺ T_RM deficient in IFN-γ production were modestly less effective in controlling viral load with significant differences depicted on a linear scale (7). In addition, the number of viral copies did not match that of mice receiving PBS control, implicating other IFN-γ–independent mechanisms for antiviral protection (7). In the same study, parenchymal CD8⁺ T_RM were shown to possess enhanced cytolytic capacity with increased levels of granzyme B, yet the direct contribution of IFN-γ from this cell type compared with that of airway CD8⁺ T_RM for protective responses is unknown (7).

The majority of studies evaluate effector function of CD8⁺ memory T cells with in vitro TCR stimulation. In addition, the assessment of airway CD8⁺ T_RM was undertaken using bronchoalveolar lavage harvests that may be an underrepresentation of the total number in the airway or result in incomplete representation of airway T_RM populations (23). Hence we employed an in vivo labeling technique for the simultaneous detection of both airway and interstitial CD8⁺ T_RM to determine the in vivo compartment-specific production of IFN-γ using a model of influenza infection.

Specific pathogen–free C57BL/6 (Thy1.2) mice were purchased from Charles Rivers and used at 8–12 wk old. Thy1.1 B6 P14 mice with TCR specific for lymphocytic choriomeningitis virus (LCMV) gp33–41 were originally obtained from M. Bevan (University of Washington) and maintained in-house. All experimental and animal procedures were approved by the University of Iowa Animal Care and Use Committee under U.S. Public Health Service assurance, Office of Laboratory Animal Welfare guidelines.

P14 T cells from naive donor blood were transferred into naive B6 hosts at 2–2.5 × 10⁴ cells/mouse before influenza virus infection. Mature DCs for priming were generated in mice injected with B16-melanoma cells expressing FLT3L as described previously (24). Splenocytes from donor were subsequently coated with gp33–41 peptide for 2 h, washed, and column purified to enrich CD11c⁺ cells per manufacturer’s protocol (Miltenyi). A total of 0.5 × 10⁶ peptide-coated DCs were injected i.v.

Recombinant influenza A/PR/8/34 (H1N1) expressing gp33–41 (PR8-gp33) and H3N2 X31-gp33 as previously described were both kind gifts from S. Varga (University of Iowa) (25). Anesthetized (ketamine/xylazine) mice were intranasally (IN) inoculated with 25 μl of recombinant PR8-gp33 (10 PFUs) or X31-gp33 (1 × 10³ PFUs) diluted in PBS. For rechallenge experiments, PR8-gp33 at 10⁴ PFUs was administered IN. For in vivo peptide stimulation, mice were administered IN with 1 μg or 50 μg gp33–41 peptide or PBS with brefeldin A (BioLegend) at a concentration of 5 μg/ml 6 h before lung harvest.

Mice were treated before lung harvest as previously described (23). In brief, mice were anesthetized and inoculated i.n. with 100 μl of 0.25 μg of CD45 Ab (BioLegend) to label cells in the airway. Two minutes later, mice were injected i.v. with 2 μg of anti-CD45.2 to label circulating cells. Lungs were subsequently removed 3 min later on euthanasia, cut into small pieces, and digested with collagenase (125 U/ml) and DNase (0.1 mg/ml). Subsequently, the lungs were passed through a 70-μm cell strainer and centrifuged in 35% Percoll (GE Healthcare) in HBSS to enrich for lymphocytes. RBCs were lysed with Vitalise (CMDG).

Lung cells were stained with viability dye eF780 (Invitrogen) in PBS incubated at 4°C for 20 min to distinguish live from dead cells. The following Abs were used for surface staining of lung cells: CD90.1 allophycocyanin, CD90.1 A700, CD90.1 FITC, CD103 PE, CD11a BV510, CD69 PE-cyanine 7, and CD25 PerCPCy5.5 (BioLegend; Tonbo). For detection of IFN-γ after in vitro stimulation, autologous splenocytes (1–2 × 10⁶) were prelabeled with anti–Thy1.1-PerCPCy5.5 and cocultured with lung cells for 5 h in the presence of LCMV gp33–41 and brefeldin A. For intracellular staining, cells were fixed, permeabilized (BD), and stained for either IFN-γ allophycocyanin or IFN-γ BV421. LSR Fortessa (Becton Dickinson) was used for flow cytometry data acquisition in FACSDiva, and data were analyzed using FlowJo 10 (BD Biosciences, Ashland, OR). For in vivo stimulations, harvested lung cells were immediately surface stained, fixed, permeabilized, and stained for intracellular IFN-γ.

Statistical analysis was carried out in GraphPad Prism. Statistically discernable differences between two different populations were determined using a paired two-tailed t test. For comparison of varying concentrations of peptide, a one-way ANOVA with Tukey’s multiple comparison test was used. In all cases, p < 0.05 was set at statistical significance.

To verify the production of IFN-γ as a functional characteristic of lung CD8⁺ T_RM, we used PR8-gp33 as a mouse-adapted strain for influenza. One day before infection, we adoptively transferred Thy 1.1 P14 transgenic T cells with a fixed TCR specific for the LCMV epitope gp33–41 (gp33) in naive Thy1.2 B6 mice. This approach allows for a trackable T cell population independent from influenza-specific TCRs with variable affinities. The mice were subsequently primed with gp33-loaded DCs to ensure for a sufficient number of CD8⁺ T_RM after boosting in both the airway and the parenchyma (24). Seven days after priming with DCs, mice were inoculated with 10 PFUs of PR8-gp33, and lungs were harvested 37–43 d postinfection (Fig. 1A). In vitro restimulation of airway and parenchymal T_RM for assessment of IFN-γ was achieved by mixing the lung cells with autologous splenocytes and a saturating concentration of gp33 peptide (10 nM). CD8⁺ T_RM were distinguished based on expression of CD103 and exclusion of intravascular-positive circulating cells. The mice were also inoculated with Ab IN to distinguish IN⁺ airway cells from IN⁻ parenchymal cells (Fig. 1B) (23). Within both populations, the percentage of IFN-γ–producing cells was analyzed (Fig. 1B, 1C). Although there was a clear population of IFN-γ⁺ cells irrespective of location, the IN⁻ compartment (representing cells in the parenchyma) had a significantly greater percentage of IFN-γ⁺CD103⁺CD8⁺ T_RM compared with the IN⁺ compartment (Fig. 1C). Hence airway CD8⁺ T_RM are compromised in their in vitro production of IFN-γ.

A distinctive feature of CD4⁺ or CD8⁺ memory T cells in the lung airways compared with the parenchyma and vasculature is the relatively rapid loss of LFA-1 (CD11a/CD18) expression after entry to the airways (8). These CD11a^lo airway CD8⁺ T_RM, which exhibit less cytolytic capacity than parenchymal T_RM, have been suggested to be protective against rechallenge through IFN-γ production (7). To test whether CD11a expression in CD8⁺ T_RM impacted the production of IFN-γ, we used the in vitro restimulation method as described in Fig. 1. After gating on CD103⁺ P14 cells, we assessed the percentage of IFN-γ production in the total CD11a^lo and CD11a^hi subset (Fig. 2A). As predicted, based on the role of LFA-1 in T cell–APC interactions (26), the percentage of IFN-γ production was significantly higher in the CD11a^hi T_RM compared with CD11a^lo T_RM (Fig. 2B).

Higher Ag concentrations could overcome the physical loss of LFA-1 expression. Thus, we tested higher concentrations of 200 nM and 1 μM gp33 peptide. In addition, we determined whether bypassing the T cell membrane receptor complex for T cell activation using PMA/ionomycin would further enhance IFN-γ production in airway CD103⁺CD8⁺ T_RM. Increasing concentrations of gp33 peptide did not further increase the percentage of IFN-γ, maintaining around 20%–30% of all IN⁺CD11a^loCD103⁺ T_RM (Fig. 2C). No cytokine-producing cells were observed after stimulation with the OVA peptide control, hence discounting potential bystander responses during the incubation (Fig. 2C). In addition, bypassing TCR stimulation with PMA and ionomycin did not further increase these responses with an average of 25%–30% with all stimuli (Fig. 2C). To rule out the possibility that TCR stimulation results in upregulation of CD11a, we assessed the percentage and expression level within IN⁺CD103⁺ T_RM (Supplemental Fig. 1A, 1B). As depicted, there were no significant differences in CD11a expression with peptide or PMA iono stimulation. Together, these data suggest a functional compromise of IFN-γ–producing capacity in CD11a^lo airway T_RM even under maximal in vitro stimulation conditions.

Cytokine stimulation under optimal conditions in vitro may not mimic in vivo T cell responses. Mice received P14 cells and were inoculated with a sublethal dose of PR8-gp33 1 d later with lung harvest 30–50 d after the primary infection to address the capacity of CD11a^lo and CD11a^hi airway and parenchymal T_RM to produce IFN-γ in vivo (Fig. 3A). To stimulate in vivo IFN-γ production, we administered gp33 peptide or irrelevant OVA257 peptide at various doses IN together with brefeldin A 6 h before harvest. Lung CD8⁺ T_RM were identified based on CD103 as a canonical marker of residence and because CD69 can be upregulated by TCR stimulation. Three main subpopulations were analyzed based on the IN label to identify CD103⁺CD8⁺ T_RM in the parenchyma and airway and the expression level of CD11a. These were IN⁺CD11a^lo, IN⁺CD11a^hi, and IN⁻CD11a^hi CD103⁺CD8⁺ T_RM (Fig. 3B). Notably, a small IN⁻CD11a^lo was not further analyzed because the number of events was not high enough to give an accurate percentage of gp33-stimulated IFN-γ above the OVA257-peptide control. A clear population of IFN-γ⁺ T_RM was detected in the CD11a^hi P14 cells after in vivo gp33 stimulation in both the airway and the parenchymal compartments (Fig. 3B, 3C). Strikingly, the fraction of IFN-γ⁺ T_RM stimulated by gp33 in the IN⁺CD11a^lo subset was barely detectable above the irrelevant peptide control, despite instilling as much as 50 μg of gp33 IN (Fig. 3C). The increase in IFN-γ⁺ in CD11a⁺ cells was not a result of increasing CD11a expression within IN⁺CD103⁺ T_RM (Supplemental Fig. 1C, 1D). Together, these data suggest that the level of IFN-γ is even further compromised in CD11a^lo airway T_RM in vivo compared with the standard in vitro method of restimulation.

Because we observed compromised IFN-γ production in CD11a^lo airway T_RM in vivo on peptide stimulation, we next addressed whether these cells could respond within the natural context of a secondary influenza infection. We adoptively transferred P14 cells into a naive host that was challenged with a sublethal dose of X31-gp33 (H3N2). More than 30 d after the primary infection, mice were rechallenged with a lethal dose of PR8 expressing either the cognate Ag gp33 or OVA257–264 as an irrelevant Ag (Fig. 4A). Two days after rechallenge, the lungs were harvested to assess for IFN-γ production from reactivated airway and parenchymal CD8⁺ T_RM. To monitor TCR activation specificity, we used CD25 (IL-2Rα) expression as a surrogate marker. Two days after rechallenge with PR8-gp33, both IFN-γ and CD25 were expressed by CD11a^hi P14 T_RM in both airway and parenchyma (Fig. 4B). In contrast, minimal expressions of IFN-γ and CD25 were observed in CD11a^lo P14 T_RM from the airways (Fig. 4B). Notably, the fraction of CD11a^hi T_RM increases in the IN⁺ population at 2 d after PR8 challenge (data not shown), and these cells, which are likely recently arrived in the airways, do express both IFN-γ and CD25 (Fig. 4B). Consistent with our results with IFN-γ production after in vivo peptide instillation, the expressions of CD25 and IFN-γ were significantly increased on infection with PR8-gp33 compared with PR8-OVA (Fig. 4C). However, this increase remained specific to CD11a^hi T_RM residing in both the airway and the parenchyma. Together with the peptide instillation experiment (Fig. 3), these data show that influenza-induced CD11a^lo airway T_RM are compromised in IFN-γ production after in vivo Ag exposure.

Airway T_RM generated by respiratory virus infection are compromised in cytolytic activity and recall proliferative responses. Our results are consistent with a study by Kohlmeier and colleagues (7) that some CD11a^lo airway T_RM induced by IAV can produce IFN-γ after optimal in vitro peptide stimulation. This study also used intratracheal adoptive transfer of WT and IFN-γ–deficient airway T_RM to suggest a contribution of IFN-γ to control airway virus titers. Our finding that CD11a^loCD103⁺CD8⁺ T_RM in the airway produce minimal IFN-γ on TCR stimulation by peptide or IAV infection in vivo is in seeming contrast with a major role for IFN-γ production by these cells in control of influenza challenge. More work will be required to fully resolve this issue. However, it appears clear that CD11a^lo airway T_RM, which have prolonged residence in the airways, are not fully competent in manifesting potent CD8 T cell effector mechanisms, such as cytolysis or IFN-γ production.

CD11a is a component of the LFA-1 integrin that interacts with ICAM-1 to enhance T cell–APC interactions (26, 27). The lack of LFA-1 could be a major driver in the dysfunction of long-term airway T_RM given its involvement in CTL activation. However, we and others observed clear IFN-γ production from these cells with peptide stimulation in vitro (7, 28). Hence the loss of CD11a may not be the only lesion compromising effector functions in vivo of lung T_RM with prolonged airway residence. Several studies have implicated the metabolic control of CD8 T cell effector mechanisms (29, 30). Consistent with this notion, a recent study suggests that the harsh environment of the airway may compromise CD8 T cell longevity through amino acid starvation, leading to transcriptional and epigenetic changes and induction of apoptosis in airway CD8⁺ T_RM (30). In this study, apoptosis was reversed on transfer of airway T_RM to an environment replete with nutrients (30), which may account for their ability to also produce IFN-γ after in vitro peptide stimulation. These data suggest the possibility that airway environment and metabolic constraints could contribute to loss of effector functions by CD11a^lo T_RM with prolonged airway residence. Another possibility for the suppressed IFN-γ production by airway T_RM in vivo is the interaction with neighboring cells that control immunopathology, which include alveolar macrophages or inflammatory macrophages that negatively regulate DC function (31–33).

For future vaccine strategies targeting influenza or other respiratory viruses, it is imperative to elicit durable responses, as well as those proximal to the mucosal site. Given the surprisingly limited research so far, this study highlights the value of probing the biology of mucosal CD8⁺ T_RM and their in vivo regulation for effective antiviral functions.

The authors have no financial conflicts of interest.

We thank members of the Harty and Badovinac laboratories for helpful discussion and Lecia Epping and Lisa Hancox for technical assistance.