The respiratory environment diverts the development of anti-viral memory CD8 T cells

Abstract

Our understanding of memory CD8⁺ T cells has been largely derived from acute, systemic infection models. However, memory CD8⁺ T cells generated from mucosal infection exhibit unique properties and, following respiratory infection, are not maintained in the lung long-term. To better understand how infection route modifies memory differentiation, we compared murine CD8⁺ T cell responses to a vesicular stomatitis virus (VSV) challenge generated intranasally (IN) or intravenously (IV). IN infection resulted in greater peak expansion of VSV-specific CD8⁺ T cells. However, this numerical advantage was rapidly lost during the contraction phase of the immune response, resulting in memory CD8⁺ T cell numerical deficiencies when compared to IV infection. Interestingly, the anti-viral CD8⁺ T cells generated in response to IN VSV exhibited a biased and sustained proportion of early effector cells (CD127^loKLRG1^lo) akin to the developmental program favored after IN influenza infection, suggesting that respiratory infection broadly favors an incomplete memory differentiation program. Correspondingly, IN VSV infection resulted in lower CD122 expression and EOMES levels by VSV-specific CD8⁺ T cells, further indicative of an inferior transition to bona-fide memory. These results may be due to distinct (CD103⁺CD11b⁺) dendritic cell subsets in the IN vs. IV T cell priming environments, which express molecules that regulate T cell signaling and the balance between tolerance and immunity. Therefore, we propose that distinct immunization routes modulate both the quality and quantity of anti-viral effector and memory CD8⁺ T cells in response to an identical pathogen and should be considered in CD8⁺ T cell-based vaccine design.

Introduction

Respiratory infections, including influenza, continue to be a major cause of morbidity and mortality globally (1). Current influenza vaccines target protective strain-specific antibody responses which prevent viral entry into the host cell (2). However, due to mutation and evolution of the targeted influenza hemagglutinin antigens these vaccines lose efficacy and do not provide long-term protection against infection. Evidence in mouse models and human studies not only implicate CD8⁺ T cells as requisite for viral clearance but also as protective against heterologous challenge with novel influenza strains (3). To date, no approved vaccine has been developed to specifically generate CD8⁺ memory T cells (T_mem) despite the fact that the intranasal influenza vaccine FluMist® elicits a larger effector CD8⁺ T cell (T_eff) response than inactivated or subunit vaccines in mice and humans (4). While the contribution of CD8⁺ T_mem to vaccine efficacy has not been tested comprehensively or longitudinally in the latter population, studies in mice have shown that CD8⁺ T cell responses are generated from a single intranasal dose of FluMist®. However, these cells are lost within 30 days and are not protective against lethal heterogeneous challenge (5). Correspondingly, in June of 2016, the Centers for Disease Control and Prevention’s Advisory Committee on Immunization Practices (ACIP) voted that FluMist® should not be using during the 2016-2017 influenza season due to a lack of a protective benefit compared to the inactivated influenza vaccine (3 vs 63%, respectively) (6). Together, these studies suggest that CD8⁺ T_eff generated following respiratory infection or delivery of live-attenuated influenza vaccines either do not form or retain T_mem at protective levels.

Over the last few decades, several laboratories have delineated pathways important in CD8⁺ T_mem development and defined the attributes and molecules which support robust T cell memory long-term. This gold standard for CD8⁺ T cell memory has been defined in murine models of acute viral infection whereby the pathogen of interest was delivered via the intravenous (IV) route (5). However, it is becoming increasingly clear that the formation of T_mem is a dynamic process, with memory potential influenced by a variety of factors including cytokines (6), the type of antigen presenting cells involved (7), and the strength/duration of antigen exposure (8), all of which can be differentially affected by inherent properties of both the pathogen and the exposure site. Indeed, our laboratory and others have shown that mucosally-derived anti-viral CD8⁺ T cells acquire properties incongruent with memory formation as defined from the systemic infection models (9, 10). For example, by simply altering the route of viral acquisition, from IV to intranasal (IN), CD8⁺ T_mem are not only less abundant overall, but those that do develop are maintained independent of the cytokine IL-15 (10). In contrast, IL-15 deficiency results in T_mem decay after systemic infection (10–12). Thus, as CD8⁺ T_mem generation does not appear to be a “one model fits all” scenario, it is important to understand how and why the respiratory T_mem program is offset from the benchmark T_mem derived in systemic model systems to improve vaccine formulation.

To date, the sole contribution of a respiratory infection on CD8⁺ T_mem potential has not been determined since many respiratory pathogens, like influenza, do not generate a productive infection outside of this environment. In this study, we used differential routes of inoculation of the same virus, vesicular stomatitis virus (VSV), to determine how the properties of the respiratory environment influence CD8⁺ T_mem development. VSV is highly suitable to our studies as the CD8⁺ T cell response after systemic IV infection is well characterized and generates long-lived T_mem. In addition, unlike influenza virus, VSV is transmitted via multiple routes in both mice and its natural bovine host, including the respiratory tract (13). Importantly, we have shown that T_mem derived from IN VSV infection numerically and phenotypically resemble anti-influenza CD8⁺ T_mem (14), validating the use of this model for evaluating how the respiratory environment affects T_mem formation. Here, we report that although pulmonary VSV infection resulted in higher CD8⁺ T_eff responses early after infection, this numerical advantage was lost rapidly though the contraction phase of the CD8⁺ T cell response, resulting in a quantitatively reduced T_mem pool. This loss is facilitated by altered programming whereby IN infection favors the development of T_eff with delayed CD127 expression and a subsequent reduced ability of these cells to express Eomes and CD122 as emerging T_mem. Coincidently, IN infection promotes the accumulation of a distinct population of CD11c⁺CD103⁺CD11b⁺CCR2⁺ monocyte/dendritic cells (moDCs) in the respiratory tract draining lymph nodes (LNs) during CD8 T cell priming which may regulate CD8⁺ T_mem differentiation. In summary, our data suggest that the environments encountered in distinct immunization routes are sufficient to modulate both the quality and quantity of anti-viral effector and memory CD8⁺ T cells in response to an identical pathogen and should be considered in CD8⁺ T cell-based vaccine design.

Materials and Methods

Mice and Viruses

C57BL/6 mice were purchased from Charles River (Wilmington, MA) through the National Cancer Institute program and bred in house. For VSV infections, age and sex matched mice were infected with 10⁴ pfu (plaque forming units) of VSV-Indiana serotype or VSV-Indiana-ova (originally obtained from Dr. Leo Lefrancois) either IN in 50 μl of PBS or IV through the tail vein in 200 μl of PBS. VSV stocks were maintained and isolated by growth in BHK cells while viral titers were determined by plaque assay. The influenza virus A/HK-x31(x31, H3N2) was generously provided by Dr. S. Mark Tompkins (University of Georgia, Athens, GA). For influenza experiments, age and sex matched animals were infected IN with 10³ pfu x31 in 50 μl PBS. For respiratory syncytial virus (RSV) infections, mice were intranasally infected with 2×10³ pfu of the RSV A2 strain (generously provided by Dr. Biao He) delivered in 50ul PBS. All animal experiments were approved by the Institutional Animal Care and Use Committee of the University of Georgia.

Tissue Preparation

Single cell suspensions from tissues were obtained as previously described (10). Briefly, cells were isolated from the lung parenchyma after first perfusing the lungs with ~10 mL PBS/heparin. Lungs were excised, minced and incubated with 1.25 mM EDTA at 37°C for 30 minutes followed by a 1 h incubation with 150 units/mL collagenase (Life Technologies, Grand Island, NY). After passage through 40 μM cell strainers, lymphocytes were resuspended in 44% Percoll, underlaid with 67% Percoll, centrifuged and the cellular interface collected. Lymph nodes and splenic tissues were mechanically disrupted then passed through a cell strainer. Erythrocytes were depleted from the spleen samples using Tris-buffered ammonium chloride (TAC). Blood samples were obtained either by retro-orbital eye bleeding or by cardiac puncture at time of sacrifice. Erythrocytes were depleted from blood samples by two serial treatments, 10 min each at 37° C, with TAC. Cell numbers were determined using a Z2 Coulter Particle Counter (Beckman Coulter, Fullerton, CA). For lymphocyte numbers obtained from blood samples 250 μl of blood was counted and expressed as lymphocytes/mL.

Flow cytometry

The VSV nucleoprotein (N) MHC class I [H-2K^b/RGYVYQGL] tetramer and the influenza nuclear protein (NP) MHC class I [H-2D^b/ASNENMETM] tetramer (conjugated to APC) were obtained from the NIH Tetramer Core Facility (Emory University, Atlanta, GA). Staining was carried out at RT for 1 h in conjunction with other surface staining. Antibodies (clones) used for staining include: PE-conjugated αCD127 (A7R34), αEomes (Dan11mag), or αCD103 (2E7); FITC-conjugated αCD11b (M1/70), or αCD44 (IM7); PerCP-Cy5.5-conjugated αCD8a (53-6.7), αCD19 (1D3), αNK1.1 (PK136), αCD3e (145-2C11), or αT-bet (eBio4B10); PE-Cy7-conjugated αCD44 (IM7), αCD11c (N418), αKLRG-1 (2F1), αPD-L1 (MIH5), or αCD86 (GLI); violetFluor 450-conjugated αCD8a (2.43), or αCD11c (N418); APC-conjugated αCD80 (16-10A1), or αSIINFEKL-H-2K^b (25-D1.16); APC-eFluor 780-conjugated αCD62L (MEL-14), αMHC II (M5/114.15.2) and APC-eFluor 647-conjugated phospho-s6 ribosomal protein (Ser235/236) (D57.2.2E) (Cell Signaling Technology Danvers, MA). All other antibodies were purchased from eBioscience, Tonbo Bioscience (both San Diego, CA) or BD Biosciences (San Jose, CA). When tetramer was not used, cells were surface stained for 20 min at 4°C. Tissue resident memory cells were identified by intravascular staining where mice were injected with (3ug of FITC-conjugated αCD45.2 (104) in 200ul PBS 3 minutes prior to sacrifice, cells were isolated, and subsequently stained ex vivo. For analysis of intracellular proteins (T-bet and Eomes) cells were fixed, permeabilized using eBioscience Fix Perm and intracellularly stained according to the manufacturer’s instructions (eBioscience, San Diego, CA). Following staining, cells were fixed in 2% paraformaldehyde and flow cytometric analysis was performed using a BD LSR II and data were acquired with FacsDiva software (BD Biosciences, San Jose, CA). For sorting experiments, fluorescently-activated cell sorting (FACS) was performed on a Beckman Coulter MoFlo (XDP) (Indianapolis, IN). For analysis of p-s6, lymph nodes and spleens were removed from mice and immediately disassociated as previously described in ice-cold methanol. Cells were then washed 2× in BD Perm/Wash (San Jose, CA) and stained in BD Perm/Wash for 45 min at RT. After washing, cells were resuspended in BD Perm/Wash and flow cytometric analysis was immediately performed. Data were analyzed using FlowJo software version 9.6.2 from Tree Star Inc. In all analyses, cells were first gated on single cells and where indicated followed by a lymphocyte gate as determined by forward (FSC-A) and side scatter (SSC-A). Subsequent gating strategies are noted in the figure legends.

Statistics

Statistical analysis was carried out using GraphPad Prism, version 5 or 6 using the specific analysis indicated in the individual figure legends. Significance was determined when the p-value was <0.05.

Results

IN infection generates fewer CD8⁺ T_mem than IV infection despite being derived from a larger effector cell pool

Respiratory infections produce CD8⁺ T_mem responses which are limited, both in number and lifespan, compared to systemic infections (10, 14). This is believed to be due, at least in part, to the heightened level of immune regulation at these surfaces which limit inadvertent immunopathology. However, the mechanisms linked to the development of substandard T_mem have been poorly understood, mainly due to inequitable comparisons made using two different pathogens with distinct tissue tropisms and inflammatory signals. Therefore, we sought to modify route of infection alone to address how the respiratory environment affects the development CD8⁺ T_mem.

To first validate that the respiratory route of infection modifies the development of CD8⁺ T_mem, mice were infected either IN or IV with a sub-lethal dose of VSV. This virus produces a replicating viral infection via multiple routes (15) and thus provides a model to compare the emergence of CD8⁺ T_mem arising from systemic and respiratory infection. Antigen-specific CD8⁺ T cell responses were assessed at 35 days post infection (dpi) using MHC-I tetramers against the immunodominant epitope of the VSV-nucleoprotein (N-tet⁺ cells). As previously reported (10), IN delivery of VSV resulted in a lower frequency of VSV N-tet⁺ CD8⁺ T_mem when compared to systemic infection using the same viral dose (data not shown). Moreover, IN infection also resulted in numerically deficient CD8⁺ T_mem in sites proximal (lung) and distal (spleen) to the respiratory tract (Fig. 1A). Thus, in a system where an identical pathogen is used, intranasal infection resulted in a quantitatively reduced T_mem.

FIGURE 1. Respiratory infection results in quantitatively deficient memory CD8 T cell pools, despite higher effector responses.

(A) Number of VSV N-tet⁺ CD8 T cells assessed at 35 days post infection (dpi) in the lung and spleen after infection with 10⁴ pfu VSV via the IV (black bar) or IN (open bar) route. (B) Mice were infected IN or IV with 10⁴ pfu VSV and CD8⁺ T cell responses were monitored in the blood 5-9 dpi. Data is displayed as frequency of N-tet⁺ CD8 T cells; n=3 mice/group and data represents 2 independent experiments. (C,D) Mice were infected with 10⁴ pfu VSV by either route and lymphocytes assessed for N-tet reactivity at 8 dpi in the indicated tissues and displayed as total frequency of CD8⁺ T cells (C) or quantified (D); n=10 mice/group and data is representative of 3 independent experiments, Significance between groups was assessed using a two-tailed student t-test (*p=<0.05, **=p<0.01, ***=p<0.001).

The size of the CD8⁺ T_mem pool is often correlated with the overall size of the corresponding effector population (16). Given that the lung has multiple innate barriers and may promote immune tolerance (17), we speculated that the CD8⁺ T_eff response may also be suppressed following respiratory infection, resulting in the observed T_mem reduction. To test this hypothesis, we assessed the CD8⁺ T_eff responses to IV and IN VSV infection. Surprisingly, the quantitative deficiency of IN derived N-tet⁺ CD8⁺ T_mem was in direct contrast to the peak antigen-specific CD8⁺ T cell response, where the IN derived CD8⁺ T_eff were more prominent. IN VSV infection produced greater CD8⁺ T cell activation in the blood, as assessed by CD44 expression (data not shown), and increased frequencies of circulating VSV-specific CD8⁺ T cells at the peak of the response (~7 dpi; Fig. 1B). Moreover, a higher antigen-specific CD8⁺ T_eff response was also observed in the spleen, lung and lung draining mediastinal LN (MdLN) 8 dpi (Fig. 1C & D). Together, these data demonstrate that the reduced CD8 T_mem responses observed following IN infection versus IV infection are not merely the result of numerically reduced T_eff responses.

Intranasally derived VSV-specific CD8⁺ T cells display reduced conversion to memory during the contraction phase of the immune response

Intranasal VSV infection results in heightened CD8⁺ T_eff responses yet quantitatively decreased memory pools (Fig. 1). This indicates that respiratory derived CD8⁺ T cells, compared to their systemically derived counterparts, are either unable to transition to T_mem or are not maintained. To address the former of these scenarios we monitored antigen-specific CD8⁺ T cell responses throughout the contraction phase of immune response. Comparison of the kinetics of CD8⁺ T cell contraction between IN and IV VSV infected mice indicated that antigen-specific CD8⁺ T cells that develop following IN infection are lost more rapidly during the contraction phase (8-15 dpi) in all analyzed tissues (Fig. 2). The average fold loss of T_eff during this period was 1.7× (in the MdLN), 3× (in the spleen), and 3.5× (in the lung) greater in IN vs IV infected animals (Fig. 2C). While it has been known for some time that respiratory-derived CD8⁺ T cells are not maintained in the lung long-term, this loss is thought to occur months following infection (14). Our data suggest that the instability of respiratory T_mem may not be solely be due to simple attrition but is more complex, involving a programmatic modification of respondent CD8⁺ T cells much earlier in the T_eff response.

FIGURE 2. Respiratory-derived CD8⁺ T cell responses contract more rapidly, and to a greater extent, than those derived from systemic infection.

A) Representative flow plots of splenic lymphocytes isolated from mice infected with 10⁴ VSV IN or IV, and sacrificed at 8, 12, and 15 dpi. Plots show populations previously gated on CD8⁺ lymphocytes. Frequencies (B) or total cell number (C) of VSV-N-tet⁺ CD8⁺ T cells isolated from the indicated issues 8, 12 and 15 dpi. N=5-10 mice/group and data are representative of three independent experiments. Significance between groups was assessed using a two-tailed student t-test (*p=<0.05, **=p<0.01, ***=p<0.001).

The phenotypic characteristics of the developing anti-viral CD8⁺ T cells derived from respiratory infection diverge from those derived from systemic infection

One of the key transcription factors identified in promoting memory cell development is eomesodermin (Eomes) (18). Eomes-deficient CD8⁺ T cells undergo primary clonal expansion but are defective in long-term survival (19). As respiratory derived N-tet⁺CD8⁺ T cells expanded normally, yet developed substandard memory responses, we tested the hypothesis that N-tet⁺ CD8⁺ T_eff derived from a respiratory infection fail to initiate Eomes expression and the subsequent memory cell program. As early as 6 dpi IN-derived N-tet⁺CD8⁺ T cells isolated from the lung and spleen expressed less Eomes than those Ag-specific cells derived from systemic infection (Fig. 3B). By 35 dpi this deficiency in Eomes expression was exacerbated and reduced to 25-50% of T_mem derived following systemic infection (Fig. 3A and C). These data indicate that a known initiator of memory cell programming was considerably less in T_mem cells which are also numerically reduced after respiratory infection.

FIGURE 3. Respiratory infection results in N-tet⁺ CD8⁺ T cells that have reduced Eomes and CD122 expression.

A) Representative flow of Eomes and CD122 expression of CD8 T cells isolated from the spleen 35 days following IV or IN VSV infection. Frequency of Eomes⁺ N-tet⁺ CD8⁺ T cells following IV (black bar) or IN (white bar) VSV infection in the lung or spleen was assessed by intracellular staining at 6 (B) and 35 (C) dpi (n=3mice/group; data are representative of 3 independent experiments). Frequency (D) and number (E) of CD122^hi of N-tet⁺CD8⁺ T cells following IV (black bar) or IN (white bar) VSV infection in the lung or spleen was assessed at 35 dpi (n=3mice/group; data are representative of 3 independent experiments). F) Average proportion of T_CM (CD62L^hi, black), T_EM (CD62L^lo, grey), and T_RM (CD62L^lo, intravascular-CD45⁻, white) populations in the lung and of T_CM (CD62L^hi, black) and T_EM (CD62L^lo, grey) in the spleen 35 days post IV or IN VSV infection is depicted. Significance between groups was assessed using a two-tailed student t-test (*p=<0.05, **=p<0.01, ***=p<0.001).

CD122 is a downstream target of Eomes (20), expressed at high levels on T_mem (12), and confers IL-15 reactivity which is necessary for the maintenance of T_mem following systemic infection (11). In contrast, IL-15 is dispensable for the development of memory following respiratory infections (10). Since Eomes expression was higher on N-tet⁺ CD8⁺ T cells following IV infection compared to IN infection, we assessed whether the IV memory cell program is responsible for the differences in IL-15 dependency between infection routes. Indeed, the proportion of N-tet⁺ CD8⁺ T_mem cells that express CD122 following IN infection was reduced in both the spleen and the lung, compared to IV VSV infection (Fig. 3D). This loss in CD122^hi cells is even more apparent when overall numbers of CD122^hi cells were quantified (Fig. 3E). Thus the reduction in Eomes and CD122 expression, along with the more significant contraction of CD8⁺ T_eff, indicate that early signals following IN infection may suppress the transition of CD8⁺ T_eff to bona fide long-lived, IL-15 dependent T_mem.

We next wished to determine whether the reduction of Eomes and CD122 expression in antigen-specific T_mem was simply due to differences in the development of specific T_mem subsets after infection by the two routes. Therefore, we assessed the contribution of central memory (T_CM; CD62L^hi) and effector memory (T_EM; CD62L^lo) cells (21) to the overall T_mem pool in the lung and spleen 35dpi with VSV delivered IN and IV. We additionally examined the proportion of tissue resident memory cells (T_RM; CD62L^lo and negative for intravascular staining (22)) in the lung. At 35 days post IV VSV infection, T_EM were found at a 28% greater number in the lung compared to IN infected animals whereas T_CM and T_RM were represented more in the lung following IN (vs IV) VSV infection (Fig. 3F). Since T_EM express the highest level of Eomes compared to the other T_mem subsets (23), more Eomes expression in the lung following IV VSV (Fig. 3) infection could be due to increased representation of this population within the tissue. However, the T_mem pools generated in the spleen were identical irrespective of infection route (Fig. 3F) while Eomes and CD122 expression was globally reduced in splenic T_mem after IN infection (Fig. 3). Together these data indicate that the early memory cell programming and subsequent development of the T_mem pools are distinct following respiratory VSV infection, the former favoring a global reduction in T_mem potential.

Respiratory derived anti-viral CD8⁺ T_eff do not fully transition to memory precursor cells

At the proliferative peak of the anti-viral CD8⁺ T cell response, T_mem can be identified within the effector cell pool using the IL-7 receptor alpha chain (CD127) and killer-cell lectin like receptor G1 (KLRG1) (24–26). These markers have been used extensively in CD8⁺ T cell memory studies and can predict which cells will survive the contraction phase, largely in the context of systemic viral infections (25, 26). Memory precursor cells or MPECs are CD127^hiKLRG1^lo and will dominate the Ag-specific CD8 T cell pool over time based on the enhanced survival conferred by IL-7 (24, 27). These MPECs differentiate from ancestral clones, referred to as early effector cells (EEC, CD127^loKLRG-1^lo). EECs have the greatest developmental plasticity, with the potential to develop into any of the other phenotypes (28), but are generally thought not to persist into memory due to their lack of CD127 expression. Short-lived effector cells or SLECs (CD127^loKLRG1^hi) constitute the majority of the early anti-viral CD8⁺ T cell responses during systemic infection (26) yet are terminally differentiated and lost during CD8⁺ T cell contraction (24). Since respiratory-derived Ag-specific CD8⁺ T cells become activated, but do not appear to phenotypically or quantitatively match their systemically derived counterparts (Fig. 1 and Fig. 3), we sought to determine whether memory cell differentiation was stalled following IN infection, aborting the development of MPECs and affecting T_mem development.

To this end, we monitored the emergence and persistence of the aforementioned effector CD8⁺ T cell subsets in the blood following IN and IV VSV infection using the CD127 and KLRG1 markers. Early after IV infection (Fig. 4A), both SLEC and EEC N-tet⁺ cells predominated until ~11 dpi when MPECs surpassed these subsets as the dominant phenotype. The survival advantage of these MPECs is very apparent by 50 dpi where this subset prevails. In contrast, N-tet⁺ T_eff derived from an IN infection (Fig. 4B) harbor predominately EECs with sustained persistence compared to IV infection. Moreover, MPECs do not emerge as the dominant subset until ~15 dpi. Direct comparison of the composition of the effector pools between the two routes of infection at 12 dpi support the conclusion that the prolonged frequency of EECs observed after IN infection is largely at the expense of the generation of SLECs and MPECs (Fig. 4C). Furthermore, the pattern of effector CD8⁺ T cell distribution following respiratory VSV infection is similar to that observed following influenza infection, where EECs are observed for a sustained period of time in the blood (Fig. 4D).

FIGURE 4. Respiratory infection results in skewed populations of effector cells based on CD127 and KLRG1 expression.

Antigen-specific CD8⁺ T cell responses were measured in the blood over time, populations were assessed for their expression of CD127 and KLRG1 and identified as previously defined effector/memory cell subsets using these markers. Phenotypes expressed after IV (A) or IN (B) VSV infection over time and comparing between the two routes specifically at 12 dpi (C). D) The frequency of NP-specific memory CD8 T cell phenotypes in the blood over time following infection with10³ pfu of of influenza A/HKx31; n≥3 mice/group/time point and repeated 2×.

We next wanted to confirm that the anti-viral T_eff/developing T_mem in the tissues were equally impacted by route of infection. Thus, we monitored the kinetics of the appearance and contribution of the various T_eff subsets to the overall CD8⁺ T cell pool in the lung, spleen and MdLN at 8, 12 and 15 dpi (Fig. 5). At 8 dpi, SLECs dominate the IV derived N-tet⁺ CD8⁺ T cell response, totaling up to 50% of the overall antigen-specific T_eff pool (Fig. 5A&B). In contrast, after IN infection, ~50% or more of the VSV-specific CD8 T cells are EECs in all tissues at 8 and 12 dpi and numerically total twice as many compared to IV infection (Fig. 5C). While EECs are lost during contraction in both IV and IN VSV infected animals (Fig. 5C), the fold loss is greater after IN infection (Fig. 5D). These data, combined with the earlier emergence of MPECs after IV infection, has consequences for T_mem development as MPECs have a survival advantage due to expression of CD127 (24). The inability of the enhanced numbers of EECs to transition to MPECs as efficiently after IN infection could mechanistically explain the steep and persistent decline of CD8⁺ T cells during contraction resulting in numerically reduced CD8⁺ T_mem.

FIGURE 5. Respiratory infection results in increased and sustained proportions of EEC which impact MPEC numbers.

(A) Representative flow plots of splenic lymphocytes following IV or IN infection with 10⁴ pfu VSV at 8, 12, and 15 dpi. Data was previously gated on CD8⁺ lymphocytes, then VSV-N-tet⁺ cells. (B) N-tet⁺ responses in the indicated tissues following IV (top) or IN (bottom) infection. Size of bar indicates the total frequency of VSV-N-specific CD8⁺ T cells, while the different shading within indicates the proportion of these cells of the indicated phenotype. Statistical signficance is noted for an increased frequency of EECs (*), SLECs (†) and MPECs (#) between the given route and tissue as indicated. (C) Total numbers of N-tet⁺ cells with an EEC (top) or MPEC (bottom) phenotype in IV (black bar) or IN (open bar) infected animals at 8, 12 and 15 dpi. (D) Fold change in N-tet⁺ EEC (top) and MPEC (bottom) numbers was calculated from 8-15dpi in IV (black bar) and IN (open bar) infected animals. Data is shown as log₂ and n= 5-10 mice/group. Data is representative of 3 independent experiments. Significance between groups was assessed using a two-tailed student t-test (*p=<0.05, **=p<0.01, ***=p<0.001)

CD8⁺ T_mem population size is ultimately governed by the rate of cell death and conversion between phenotypes. We would therefore expect that as EECs (and SLECs) numerically decrease through contraction, MPECs would remain numerically stable or increase due to EEC → MPEC conversion. Any accelerated loss of EECs or stalled conversion to MPEC would impact the resultant MPEC/T_mem pool. Interestingly, while the overall numbers of MPECs remain relatively stable in the spleen or even increase in the lung through contraction after IV VSV infection, there is a numerical attrition of MPECs in all tissues monitored after IN VSV infection, with an enhanced fold loss compared to those MPECs generated after IV infection (Fig. 5C&D). Importantly, the number of MPECs recovered from the lung 15 dpi is significantly reduced after IN infection. Together, these data suggest that inherent defects in EEC programming leads to stalled CD127 expression which is requisite for survival and adequate seeding of the respiratory MPEC pool.

A diverse population of monocytes and DCs accumulate in the draining lymph nodes after respiratory infection

Our data thus far indicate that early events following respiratory (vs. systemic) infection likely result in a different developmental program, ultimately resulting in numerically deficient CD8⁺ T_mem development. By using an identical virus (VSV) in our studies, we have eliminated pathogen-specific pattern recognition receptor (PRR) bias and the events underlying disparate PRR signaling pathways. Furthermore, differences in antigen availability was not observed between either route of infection at a time coordinate with T_mem programming (Supplemental Figure 1) suggesting that other extrinsic environmental factors are likely responsible for modified T_mem programming.

Two populations of migratory CD11c⁺ DCs exist in the lung: CD103⁺ (CD11b⁻) and CD11b⁺ (CD103⁻) respiratory DCs (RDC). Upon activation, both CD103⁺ and CD11b⁺ RDCs migrate to the lung draining lymph nodes (dLN) (7, 29). CD103⁺ RDC and LN resident DCs (CD8α⁺) are believed to primarily participate in CD8 T cell priming after influenza infection both indirectly and via cross-presentation, respectively whereas the role of CD11b⁺ RDC in this process is controversial (30–33). Nonetheless, CD11b⁺ RDC are a source of inflammatory chemokines and thus are likely to impact T cell activation and programming (34, 35). CD103⁺ RDCs can drive effector function in activated CD8⁺ T cells, increasing T_eff transcriptional profiles and migratory capabilities, while diverting cells from a central memory fate (7). Therefore, we hypothesized that the preferential activation by respiratory CD103⁺ RDCs following IN infection may inherently favor the development of CD8 T_eff, while resulting in a delayed and substandard transition to memory.

To first assess if CD103⁺ RDCs were activated and migrating to LNs following respiratory VSV infection, DC populations were identified in LNs at 3 dpi. DCs were defined as lineage-negative (Lin⁻) cells (CD3⁻, CD19⁻, NK.1.1⁻) expressing CD11c and MHC II (Fig. 6A). As previously reported following influenza infection (30), VSV IN infected mice harbored high numbers of CD11c⁺ DCs in their lung dLNs (MdLN and cervical (cLN)), with a significant proportion of these cells expressing CD103 (Fig. 6A and B). Interestingly, three quarters of the CD103⁺ DC co-expressed CD11b, and a reduced expression of MHC II compared to the other DC populations (Table 1.). These cells were not found either in analogous LNs after IV infection or in the lung after VSV infection by either route (Fig. 6C, 6G and data not shown). Importantly, these cells were also identified following respiratory infection with influenza (Fig. 6C) and respiratory syncytial virus (Fig. 6C). The morphological (cytoplasm:nucleus ratio, kidney shaped nucleus;) (Fig. 6D) and phenotypical (CCR2⁺GR1⁺MHC class II^lo) attributes of these cells (Table 1), as well as their transient appearance in the LNs after IN VSV infection (Fig. 6G) suggests these cells are derived from circulating monocyte precursors (36). From this point further, we will refer to these cells as monocyte/DCs (moDCs).

FIGURE 6. Repiratory infection results in the accumulation of distinct DC populations at sites of T cell priming.

A) Representative flow samples showing the initial gating of lineage- (CD3⁻,CD19⁻,NK1.1⁻ cells) DCs based on CD11c and MHC II (left) followed by CD103 (right). B) Total numbers of DCs (height of bar) and CD103⁺ DCs (gray portion of bar) isolated from indicated tissues. The color of the asterixes indicate signigicant differences between the corresponding populations within IV and IN VSV infected animals. C) Representative contribution of classical CD103⁺CD11b⁻ and CD103⁻CD11b⁺ RDC and CD103⁺CD11b⁺ moDC in the cLN after infection with the indicated pathogen. D) The cytospin image of giemsa-stained CD11c⁺MHC II⁺ cells FACS sorted based on CD11b and CD103 expression is shown. E) Representative OVA-H2Kb staining of indicated DC subsets isolated from the cLN 3 days post IN VSV-OVA (white histogram) or IN VSV (grey histogram) infection compared to the FMO control for each subset (black histogram). F) Average adjusted mean fluorescence intensity (MFI) of OVA-H2Kb by indicated DC subsets. Adjusted MFI calculated as the average OVA-H2Kb expression (VSV-OVA) with average background OVA-H2Kb expression (VSV) subtracted out; data pooled from 3 independent experiments. G) The lung tissue and pooled cLN and mLN were isolated on the indicated days post infection with IN VSV and the frequency and total number of CD103⁺CD11b⁺ moDCs determined by flow cytometry. H) cLN, iLNs, and spleens were isolated at the indicated hour post infection with IV (black bar) or IN (open bar) VSV and the levels of S6 phosphorylation at serine 235 and 236 in CD44^hi CD8⁺ T cells was determined by flow cytometry. Significance was determined using a two-tailed student t-test to test between IV and IN infected mice (*p=<0.05, **=p<0.01, ***=p<0.001), N=3 mice/group, representative of two independent experiments.

Table I.

Phenotypic Characterization of Respiratory DC Subsets¹

	CD11b+CD103−	CD11b−CD103+	CD11b+CD103+
CCR2	++	−	+++
Gr1	+++	−	++
MHC II	++	+++	+
MHC I	++	++	+++
B7-1	++	++	+++
B7-2	++	+++	+
PD-L1	++	++	+++
Ag Presentation	+	+++	+

¹Phenotypic analysis of DC populations in the cLN 3 days following IN VSV infection. Isolated Lin-⁻ (CD3e, CD19, NK1.1), CD11c⁺MHCII⁺ cells were gated into the three DC lineages based on CDllb and CD103 expression. Antigen presentation was determined via reactivity to a fluorescently labeled SIINFEKL-H-2K^b antibody. The phenotype of each of the subsets was determined based on mean fluorescence intensity (MFI) of individual cell surface markers relative to naïve control animals (CD11c⁺MHCII⁺ cells). Expression was determined as: ⁻ =negative, ⁺= low, ⁺⁺= intermediate, ⁺⁺⁺=high. N=4 mice/group and data is representative of 2 separate experiments.

The moDCs were present in the dLNs only after respiratory infection, coordinate with T cell priming and the skewed development T_effs. To begin to decipher how these moDCs could be influencing CD8⁺ T_mem development, we sought to determine whether these cells compromise T_eff-T_mem programming via modified antigen presentation or co-stimulation/inhibition. Thus, we infected animals IN with VSV or VSV-ova and measured the surface expression of the ova epitope SIINFEKL presented by H-2K^b. While CD103⁺CD11b⁻ cells expressed the highest level of antigen/MHC class I complexes (Fig. 6E and 6F) and the co-stimulatory molecule B7-2 (Table 1), the moDCs expressed low levels of antigen (Fig. 6E), low levels of B7-2, high levels of B7-1, and high levels of the inhibitory ligand programmed death ligand 1 (PD-L1) (Table 1). Activated CD8⁺ T cells express PD-1 and CTLA-4 (37, 38). Engagement of PD-1 by PD-L1 or preferential signaling of CTLA-4 in the context of B7-1/2 expression can affect signaling via the PI3K/Akt pathway (39), which is known to regulate T_mem programming via mTOR (40). Early in priming (48 h), we observed an increase in the phosphorylation of ribosomal protein S6 (p-S6) which would occur via either the MAPK (41) or mTOR-dependent pathway (42, 43) specifically in the respiratory dLN after IN infection (Fig. 6H) suggesting hyper-activation of T_eff after respiratory infection.

Overall the data presented in this paper shows that there is a common developmental pathway for CD8⁺ T_mem generated following respiratory infection resulting in the prolonged survival of EEC with concomitant reduced conversion to MPEC. The overall consequence of this is the reduced size and durability of the memory pool (Supplemental Figure 2). To our knowledge, our data demonstrate for the first time that IN infection preferentially recruits moDCs to the dLN, which expressed PD-L1 and increased levels of B7-1 over traditional RDC (Table 1). This inhibitory phenotype, coupled with rapid pS6 suppression within the dLN suggests an intrinsic modulation of respiratory CD8⁺ T cell responses. In conclusion, these results offer emerging insight as to why memory CD8⁺ T cell development following intranasal vaccines and natural respiratory infections is neither as robust nor long lasting as those generated in systemic models of infection.

Discussion

Systemic infection models have dominated the field of CD8⁺ T cell memory development since its infancy (44). This is likely due to the ability to consistently produce large, traceable pools of T_mem using acute, viral infections delivered by this route (45). However, neither vaccines nor most naturally transmitted infections are acquired through an IV route. Therefore, the possibility arises that the systemic models we use to study CD8⁺ T_mem development may not fully encapsulate the properties of CD8⁺ T_mem derived from physiologically relevant infection routes. Indeed, Mueller et al. showed that infection via a mucosal route (IN influenza infection) results in qualitatively deficient memory CD8⁺ T cells with reduced protective capacity compared to those acquired via a systemic route (IV LCMV) (9). How the respiratory environment regulates this response was unclear in these studies as distinct inflammatory and cytokine profiles elicited by the divergent priming viruses could not be eliminated as a confounding variable. By using VSV infection in our studies, we directly tested the impact of the respiratory environment on the developmental pathways responsible for anti-viral CD8⁺ T cell development.

Respiratory VSV infection recapitulates many features of anti-viral CD8⁺ T cells derived from a native influenza infection, including the generation of a robust effector cell pool yet reduced frequencies of T_mem (Fig. 1). The rate of attrition was most pronounced immediately after the peak number of CD8⁺ T_eff were detected in all tissues (Fig. 2) which led us to speculate that respiratory infection may differentially program anti-viral CD8⁺ T cells in a way that favors short over long-term protection. Indeed, EECs which maintain the plasticity to differentiate into either MPECs or SLECs (28) were enriched and selectively maintained within the Ag-specific CD8⁺ T cell pool after respiratory VSV infection, where they constituted ~ 1/3 of the antigen-specific response out to 15 dpi (Figs. 4 & 5). The sustained EEC phenotype is also observed following influenza infection (Fig. 4D), providing evidence that infection via the respiratory route may uniformly contribute to a developmental stall, preventing full transition to memory. Indeed, the sustained population of EECs generated following IN infection came at the expense of generating MPECs (Figs. 4C & 4D) and correlated with the timing of the greatest loss of the IN VSV-specific CD8⁺ T cells (Fig. 2). Prior to our study it was unclear whether EEC maintained beyond contraction could convert to MPECs; our data suggest that they do not (Fig. 5D), however this remains a possibility. Nonetheless, the delayed appearance of CD127⁺ (MPECs) after respiratory infection likely accounted for the greater loss of anti-viral CD8⁺ T cells during contraction.

Limitations on T_mem in the respiratory tract makes teleological sense given the plethora of respiratory assaults an individual encounters over its lifetime and the limited space to harbor accumulating T_mem without compromising tissue function. However, reductions in T_mem were also observed in the spleen (Figs. 1A &B) suggesting T_mem development was not selectively suppressed in the respiratory tract when exposure was by the respiratory route. Taking this into account, it is quite possible that respiratory derived restrictions may directly or indirectly lead to the selective development of certain subsets of T_mem. Our data suggest this is true, at least in part, as the lung parenchyma of IN infected mice harbor more T_RM and T_CM but less T_EM than IV infected animals, while pools are equivalent in the spleen after VSV infection by either route (Fig. 3F). However, respiratory VSV infection resulted in lower expression of Eomes, as well as the Eomes-regulated IL-15 receptor, CD122 by both splenic and lung T_mem (Fig. 3) and MPECs (data not shown), suggesting that differences in T_mem Eomes expression between the two infection routes is likely cell intrinsic and not necessarily T_mem subset specific. Eomes expression is key factor relevant to maintaining systemically derived CD8⁺ T_mem cells, partially due to its ability to up-regulate CD122 expression (20). Therefore, respiratory viral infections may favor memory cells which are IL-15 independent. Intriguingly, tissue resident memory cells (T_RM), defending in mucosal sites such as the lung express less CD122 than other T_mem subsets (46) and are 2× more abundant after IN VSV infection. Additionally, a subset of T_RM isolated from the LNs of mice was found to develop independent of IL-15 signaling as well (47). While only a subset of T_mem have reduced CD122 expression, we have shown that equivalent numbers of T_mem are isolated from the lung in IL-15 deficient and replete mice (10). Therefore, CD8⁺ T cell programming after respiratory infection may favor the development of specific subsets of T_mem, many of which will provide protection at the site of infection (perhaps with reduced longevity), over large pools of “classical” memory cells which are considered dependent on IL-15. Since the role of IL-15 in CD8⁺ T_mem generation and maintenance has not been well studied after oral or intra-vaginal infection, it is possible that these results are not exclusive to respiratory environment; other mucosal immunizations may provoke similar changes in T_mem formation.

The difference in T_mem programming after respiratory infection is likely associated with events occurring very early after infection as VSV is a promiscuous virus (48), eventually resulting in a brief systemic infection, even after IN delivery (49). Respiratory-resident CD103⁺ DCs are particularly important in the priming of naïve CD8⁺ T cells after influenza infection (30, 35). Given that these DCs can influence effector cell differentiation and migration (7), we hypothesized that these migratory CD103⁺ DCs were also responsible for the altered developmental phenotypes observed following intranasal infection. On first glance, our data seemed to confirm this hypothesis, as CD103⁺ T cells were specifically enriched in the respiratory tract draining LNs following IN infection (Fig. 6). However, upon further analysis we found that only ~10% of the total dLN DCs were classic CD103⁺ RDCs, while a striking proportion (~30%) of the DCs co-expressed CD103 and CD11b and had low expression of MHC II, a phenotype more indicative of a monocyte-derived CD103⁺ population of DCs (50). This DC phenotype is not observed after IV infection. Since many classical tissue-resident CD103⁺ will die in the lung-draining LN after priming CD8⁺ T cells (36), perhaps this newly recruited CD103⁺CD11b⁺ pool develops to replace the tissue-resident CD103⁺ DCs as a result of emergency myelopoiesis. Unlike lymphocytes which can undergo rapid proliferation, innate immune cells like DCs must be replenished through expansion of hematopoietic stem cell populations in the bone marrow (51) which is crucial for controlling infection in many infection models (52, 53). It is possible that IN infection induces either selectively or at a greater level certain inflammatory signals which trigger myelopoiesis and the development of moDCs. Inflammatory cytokines IFNα/β and IFNɣ can stimulate proliferation of hematopoietic stem cell populations in the bone marrow (54), both of which are produced in abundance following viral infections (55–57). It remains unclear why the moDCs are preferentially recruited to the lung draining LN (and not the lung) after IN infection. One possibility is that CCR2 ligands are projected to the dLN, selectively recruiting the CCR2⁺ moDC (58). Nonetheless, once there, moDCs have the potential to regulate the CD8 T_mem program by delivering inhibitory signals via PD-1 or CTLA-4 engagement due to their high expression of PD-L1 and B7-1 (Table 1). Both PD-1 and CTLA-4 engagement can result in T cell inhibition by directly blocking TCR signaling (59, 60), as well as via the mTOR signaling pathway (39). Indeed, ribosomal S6 protein is phosphorylated following both IV and IN VSV infection yet reduced only in IN infected animals 72 h in the lung dLN and coordinate moDC arrival. Therefore, it is possible that the developmental stall in memory formation observed following respiratory infection could be due to suppressive signals delivered directly to respondent T cells by moDCs during T cell priming. Future experiments will examine the roll of myelopoiesis and specific immunosuppressive pathways in regulating T_mem development.

By comparing VSV infection delivered by the IN or IV route, we showed that the respiratory environment results in T_mem that is skewed from the archetypical memory developmental programs defined in systemic models of infection, resulting in numerically deficient memory (Supplemental Fig. 2). The implications of this work suggest that the induction of protective memory CD8⁺ T cells should be studied in the context of appropriate infection route, as the developmental pathways and requirements for memory vary between routes of priming. As there continues to be a growing interest in developing CD8⁺ T cell based vaccines (particularly those which will induce respiratory specific responses) (61), it is imperative that we continue to improve our understanding regarding the mechanism of how the respiratory environment modifies T_mem. Fine tuning of the local respiratory environment via targeting specific DC pools or perhaps induction of responses via other routes mucosal infection may be necessary to secure the desired T_mem outcome.