Lung dendritic cells migrate to the spleen to prime long-lived TCF1^hi memory CD8⁺ T cell precursors after influenza infection

Abstract

CD8⁺ T cell responses to pulmonary challenges are primed by lung migratory dendritic cells (mDCs), which capture antigens in the lungs and migrate to the lung-draining mediastinal lymph node (med-LN) to activate T cells. Notably, the lungs and the spleen are not connected by the lymphatic vasculature. Thus, the current paradigm suggests that in response to respiratory virus infections that are restricted to the respiratory tract, priming of T cell responses by lung mDCs takes place entirely in the med-LN. Our results challenge this “LN-centric” paradigm by demonstrating that during influenza virus infection, lung mDCs egress the med-LN and traffic to the spleen, where they prime influenza-specific CD8⁺ T cells. Importantly, CD8⁺ T cells primed in the spleen are transcriptionally distinct and have enhanced ability to differentiate into long-lived memory cells compared to med-LN-primed counterparts. Thus, our data identify a lung mDC trafficking pathway that connects the lungs with the spleen.

ONE SENTENCE SUMMARY

Lung-derived dendritic cells carry influenza antigens to the spleen after egressing the lymph node by an S1P/S1PR-dependent mechanism.

INTRODUCTION

Priming of CD8⁺ T cell responses against viruses that are restricted to the respiratory tract takes place in the lung-draining mediastinal lymph node (med-LN) and requires the interaction of naïve T cells with antigen (Ag)-bearing, lung migratory dendritic cells (mDCs) (1–4). Mechanistically, lung mDCs capture viral antigens in the lung, upregulate CCR7, enter the lymphatic vessels, and subsequently migrate to the med-LN to cross-prime CD8⁺ T cells (5–9). Consistent with the essential role of lung mDCs, T cell priming to respiratory viruses is compromised in the absence of lung mDCs or when these cells are unable to migrate out of the lungs to the med-LNs (5–9). Importantly, it is generally believed that mDCs die within three days after reaching the med-LN (3, 10, 11). Thus, the current paradigm suggests interactions between lung mDCs and naive T cells take place entirely in the med-LN (1–4). However, despite the prevailing view that DCs do not leave the LNs, DCs can be found in the thoracic duct, which collects the lymph and cells egressing from the LNs (12–17). The origin and function of these DCs is unknown.

Intriguingly, although delayed, influenza-specific CD8⁺ T cell responses are observed in the absence of LNs (18, 19), suggesting the possibility of “non-LN” priming. In agreement with this view, mathematical modeling and computational approaches suggest that, in addition to the med-LN, the spleen might be major source of effector CD8⁺ T cells after influenza virus infection (20). Similarly, activated CD8⁺ T cells have been found in the spleen of influenza-infected mice as early as 24 h after antigen encounter (21), a time that is not sufficient for CD69 downregulation and subsequent LN egress (22–24). Altogether, these data suggest the possibility of local priming in the spleen after influenza virus infection. However, this idea is at odds with previous literature showing that respiratory antigens do not freely reach the circulation and instead have to be transported out of the lungs by mDCs (3, 25–28). Given that the lungs and the spleen are not connected by the lymphatic vasculature and that the current paradigm suggests that mDCs do not leave the med-LN, it is unclear how pulmonary-derived Ags could reach the spleen after infection. Additionally, the nature of the putative antigen-presenting cell population that might prime T cell responses in the spleen is unknown.

Using a mouse influenza virus infection model, we demonstrate that, together with the med-LN, the spleen is a primary site for the initiation of CD8⁺ T cell responses to influenza. We found that splenic influenza-specific CD8⁺ T cell responses were primed by lung mDCs that, after reaching the med-LN, egressed the med-LN and trafficked to the spleen. Importantly, priming in the spleen resulted in CD8⁺ T cell progeny with transcriptional characteristics of long-lived, self-renewing memory T cell precursors, while med-LN counterparts resembled terminal T effector (Teff) cells. Corresponding with these transcriptional differences, fate-mapping experiments demonstrated that spleen-primed CD8⁺ T cells were long-lived and contributed substantially to the pool of long-lived TCF1^hi memory cells compared to CD8⁺ T cells primed in the med-LN. Together, these results indicate that lung-derived mDCs can migrate to the spleen to prime long-lived memory CD8⁺ T cell precursors after influenza virus infection. Collectively, our data demonstrate a lung-DC trafficking pathway that connects the lungs with the spleen and suggest a mechanism by which CD8⁺ T cell functional diversity is generated after influenza virus infection.

RESULTS

Splenic migratory-like DCs (smDCs) prime influenza-specific CD8⁺ T cells in the spleen

To investigate whether the CD8⁺ T cell response to influenza could be primed outside the med-LN, we adoptively transferred CellTrace Violet (CTV)-labeled Ovalbumin (OVA)-specific TCR transgenic CD8⁺ T (OTI) cells into C57BL/6 (WT) mice. Recipient mice were then infected with A/WSN/33-OVA/I (WSN-OVA), an engineered influenza virus containing the K(b) restricted OVA_257–264 epitope in the neuraminidase stalk (29). As a negative control, mice were infected with influenza A/PR8/34 (PR8), which does not express the OVA_257–264 peptide. Three days later, we analyzed the phenotype of the donor cells in the med-LN, spleen, lung, non-draining popliteal LN (p-LN), and blood (Fig. 1a–d). The frequency of total donor OTI cells was similar in both groups in all the tissues (Fig. 1a). As expected, we found CD69⁺OTI cells (Fig. 1b, top row) with an activated CD25^hiCD62L^loCD127^loIRF4⁺ phenotype (Fig. 1c, left column) in the med-LN of WSN-OVA-infected but not PR8-infected mice. At this time point, the frequency of activated OTI cells remained at background levels in the lung, pLN, and blood (Fig. 1b). Despite the virtual absence of circulating activated OTI cells, nearly 40% of the OTI cells present in the spleen of the WSN-OVA-infected mice had an activated CD69⁺CD25^hiCD62L^loCD127^loIRF4⁺ phenotype (Fig. 1b and c). Thus, in agreement with a previous study (21), Ag-specific CD8⁺ T cells were present in the spleen of influenza-infected mice before the systemic dissemination of the CD8⁺ T cell response that was initiated in the med-LN. Importantly, activated OTI donor cells in the spleen remained unproliferated at the time of the analysis (Fig. 1d). Given that proliferation precedes LN egress (22–24), these data suggest the possibility that, rather than reflecting expanded cells that trafficked to the spleen, splenic activated OTI donor cells were locally primed in the spleen.

Figure 1. smDCs present influenza-derived Ag to CD8⁺ T cells in the spleen.

(a-c) OTI cells (CD45.2⁺) were adoptively transferred into B6 (CD45.1⁺) mice. One day later, recipient mice were i.n. infected with PR8 or WSN-OVA. Mice were analyzed on day 3 after infection. (a) Frequency of donor OTI cells in the PR8 and WSN-OVA infected mice. Representative plots gated on CD8⁺ T cells are shown. (b) Frequency of CD69⁺ cells within the donor OTI cells. Representative plots gated on CD45.2⁺CD8⁺ OTI cells are shown. (c) Expression of CD25, CD62L, CD127, and IRF4 in OTI cells from PR8-infected mice and CD69⁺OTI cells from WSN-OVA infected in the med-LN (left) and spleen (right). (d) CTV staining in donor OTI cells. Data are representative of three independent experiments. Data are shown as the mean ± SD (n=3–5 mice/group). **P < 0.01, ***P < 0.001. P values were determined using a two-tailed Studentś t-test. (e-f) Mice were infected i.n. with influenza and DC subsets in the med-LN (e) and spleen (f) were analyzed on day 3. Representative plots gated on B220⁻ Siglec-F⁻ 7AAD⁻ cells are shown. Data are representative of more than five independent experiments (n=3–5 mice). (g-h) (g) Expression of OVA_257–264-H-2Kbcomplexes in DC subsets from the spleen and med-LN of naïve and WSN-OVA infected mice. Representative plots are shown. (h) Frequency of OVA_257–264-H-2Kb expressing cells within the indicated DC subsets after WSN-OVA infection. Data are representative of two independent experiments. Data are shown as the mean ± SD (n=4–5 mice/time point). (i) B6 mice were infected with WSN-OVA or PR8 virus and splenic smDCs and rDCs and mDCs from the med-LN were sorted three days later. DCs were co-cultured with CTV-labeled OT-I cells without peptide or with OVA_257–264. T cell proliferation was assessed 72h later. Data are representative of three independent experiments.

Priming of influenza-specific CD8⁺ T cells responses requires cognate interactions with Ag-bearing DCs. Thus, we next assessed the capacity of splenic DCs to prime influenza-specific CD8⁺ T cells. It is well established that in the med-LN, influenza-specific CD8⁺ T cells are primed by lung migratory DCs (mDCs), which are phenotypically characterized as MHCII^hiCD11c^medCCR7^hi DCs (5–8). In contrast, LN-resident DCs (rDCs), which do not carry influenza-derived Ags, are MHCII^medCD11c^hiCCR7^lo (5–8) (Fig. 1e). Using a similar characterization strategy, we found that the majority of DCs in the spleen displayed a rDC-like phenotype (Fig. 1f and Supplementary Fig. 1a). However, we identified a population of splenic DCs that were MHCII^hiCD11^medCCR7^hi, phenotypically resembling mDCs from the med-LN (Fig. 1f and Supplementary Fig. 1a) that we subsequently refer to as splenic migratory-like DCs (smDCs). Staining with the 25-D1.16 monoclonal antibody revealed that, similar to mDCs from the med-LN (Fig. 1g–h), smDCs but not rDCs displayed membrane OVA257–264-H-2Kbcomplexes after WSN-OVA infection (Fig. 1g–h). The frequency (Fig. 1h) and number (Supplementary Fig. 1b) of smDCs expressing OVA257–264-H-2K^b complexes progressively increased after infection, peaking between days 3 and 4 post-infection (p.i.).

To test whether splenic smDCs were capable of priming CD8⁺ T cells, we sorted rDCs and smDCs from the spleen of WSN-OVA or PR8-infected mice, co-cultured them with naïve CTV-labeled OTI cells, and assessed T cell expansion 3 days later (Fig. 1i). As a reference, we also sorted mDCs from the med-LNs. Similar to mDCs from the med-LN, smDCs induced OTI cell expansion (Fig. 1i). In contrast, rDCs failed to activate OTI cells (Fig. 1i). Failure of splenic rDCs to prime OTI cells was not due to an intrinsic inability to stimulate naïve T cells as the inclusion of OVA_257–264 peptide rescued OTI proliferation (Fig. 1i). As a negative control, smDCs from PR8-infected mice failed to induce T cell activation (Fig. 1i). These results suggest that smDCs bear influenza-derived Ags and have the ability to prime influenza-specific CD8⁺ T cells in the spleen after influenza virus infection.

smDCs cross-present lung-derived antigens

Viral titers in the spleen of WNS-OVA infected were below detection limits (Supplementary Fig. 2a). This is not unexpected as, rather than direct Ag presentation by infected DCs, priming of influenza virus-specific CD8⁺ T cells requires cross-presentation of viral antigens that mDCs acquire from airway-infected cells before migrating into the med-LN (30–32). Thus, we examined the capacity of splenic smDCs to cross-present soluble proteins delivered through the intranasal route to T cells in vivo. We first adoptively transferred CTV-labeled OTI cells into PR8-infected mice. The next day, we administered soluble Alexa-647-labelled OVA (PR8+OVA) or PBS (PR8+PBS) intratracheally (i.t.) and enumerated activated OTI cells in the med-LN, spleen, lung, and non-draining p-LN 24h later (3 days p.i.) (Supplementary Fig. 2b and Fig. 2a–d). This approach allowed us to study CD8⁺ T cell activation following 24 hours of controlled Ag exposure, which, importantly, is a period of time that is sufficient for initial activation but not for CD69 downregulation and LN egress (22–24). The frequency of total donor OTI cells was similar in all organs in both groups (Supplementary Fig. 2b). Analogous to what we found with the WSN-OVA model, CD69⁺OTI cells with an activated phenotype were present in the med-LN and spleen but not in the lungs or the non-draining pLN of PR8+OVA infected mice (Fig. 2b, c). All donor cells remained unproliferated at this time (Fig. 2d). No activated donor cells were found in the PR8+PBS negative control infected mice (Fig. 2b and c) or in the spleen of mice that received OVA without PR8 infection (Fig. Supplementary Fig. 2c). However, we did find similar results under other lung inflammatory settings such as mice treated i.t. with OVA plus house dust mite (HDM+OVA) or LPS (LPS+OVA) (Fig. Supplementary Fig. 2c). Thus, following pulmonary challenges, cross-primed Ag-specific CD8⁺ T cells were present in the spleen before med-LN-primed cells could traffick out of the med-LN.

Figure 2. Cross-primed CD8⁺ T cells in the spleen have LN-independent origin.

(a-d) CTV-labeled OTI cells (CD45.2⁺) were transferred into PR8-infected B6 (CD45.1⁺) mice 1 day post-infection. On day 2, recipient mice were i.t. treated with PBS (PR8+PBS) or 60 μg of OVA (PR8+OVA). Mice were analyzed 24 h later (day 3 p.i.). (a) Experimental diagram. (b) Frequency of CD69⁺ cells within the donor OTI cells. Representative plots gated on CD45.2⁺CD8⁺ OTI cells are shown. (c) Expression of indicated markers in OTI cells from PR8+PBS and CD69⁺OTI cells from PR8+OVA infected mice in the med-LN (left) and spleen (right). (d) CTV staining in donor OTI cells. Data are shown as the mean ± SD (n=4–5 mice/group). Data are representative of three independent experiments. ***P < 0.001. P values were determined using a two-tailed Studentś t-test. (e-i) B6 mice were i.n. infected with PR8. On day 1, mice were treated or not with 250 μg of anti-CD62L Ab and CTV⁺OTI cells were transferred 30 min later. On day 2, mice were i.t. treated with OVA and analyzed 24 h later (day 3). Experimental diagram (e). Frequency and number of OTI cells in the med-LN (f and g) and spleen (h and i). Representative plots are shown. Data are shown as the mean ± SD (n=4–5 mice/group). Data are representative of three independent experiments. *P < 0.05. P values were determined using a two-tailed Studentś t-test. (j-k) OTI cells were transferred into day 1 PR8-infected B6 (WT) or lymphotoxin α-deficient (LTa^−/−) mice. On day 2, mice were i.t. treated with OVA, and the frequency (j) and number (k) of CD69⁺ OTI cells in the spleen were measured 24h later. Data are shown as the mean ± SD (n=3–5 mice/group). Data are representative of three independent experiments. P values were determined using a two-tailed Studentś t-test. n.s not statistically significant.

To rule out the possibility that cross-primed Ag-specific CD8⁺ T cell present in the spleen had a LN origin, we assessed whether blocking priming in the med-LN affected the presence of early-activated CD8⁺ T cells in the spleen. We treated PR8+OVA mice with anti-CD62L before OTI transfer and enumerated activated OTI cells in the med-LN and spleen 24h after OVA administration (Fig. 2e–i). As expected, anti-CD62L treatment prevented the entry of transferred OTI cells into the med-LN, thereby resulting in a virtual absence (~40-fold reduction) of activated OTI cells in this organ (Fig. 2f and g). However, despite the lack of med-LN priming, the frequencies (Fig. 2h) and numbers (Fig. 2i) of activated OTI cells in the spleen of control and anti-CD62L-treated mice were indistinguishable. These data suggest that early activation of CD8⁺ T cells in the spleen occurs independently of the med-LN priming. Furthermore, in lymphotoxin a-deficient mice (LTa^−/−), which despite having lymphatic vessels lack LNs, the absence of med-LNs did not preclude CD8⁺ T cell priming in the spleen (Fig. 2j and k). Altogether, our results suggest that rather than migrating from the med-LN, the early-activated CD8⁺ T cells found in the spleen are locally cross-primed in the spleen.

To continue testing the hypothesis that smDCs cross-presented lung-antigens in the spleen, we assessed the presence of OVA-bearing, Alexa-647⁺ DCs in the med-LN, spleen, and non-draining LNs of the PR8+OVA and PR8+PBS infected mice (Fig. 3a, b; Supplementary Fig. 3a). As predicted based on previous literature, mDCs but not rDCs were Alexa-647⁺ in the med-LN of PR8+OVA infected mice (Fig. 3a). No labeling was detected in the PBS-treated counterparts (Fig. 3a). In agreement with previous studies showing than i.n. delivered antigens do not freely reach the circulation (3, 25, 27, 33), we failed to detect OVA-bearing mDCs in non-draining LNs after PR8+OVA infection (Supplementary Fig. 3a). In contrast, correlating with the presence of activated OTI cells in the spleen, smDCs but not rDCs stained Alexa-647⁺ in the spleen of PR8+OVA infected mice (Fig. 3b and Supplementary Fig. 3b). This observation was not specific to the PR8+OVA model, as similar results were obtained in mice that were i.n. sensitized with Alexa Fluor 647-labeled HDM (Supplementary Fig. 3c–d).

Figure 3. smDCs cross-present lung-derived Ag.

(a-c) B6 mice were infected with PR8. On day 2, mice were treated i.t. with PBS or 60 μg of Alexa-647-labelled OVA (OVA-647). The frequency of OVA-647⁺ cells within the indicated DC subsets in the med-LN (a) and spleen (b) was assessed 24 h later (day 3). Representative plots are shown. Data are shown as the mean ± SD (n=5 mice/group). ***P < 0.001. P values were determined by one-way ANOVA with post-hoc Turkey’s multiple comparison test. n.s not statistically significant. (c) Expression of membrane OVA_257–264-H-2Kbcomplexes in the indicate subsets of DCs from the spleen and med-LN of OVA-647-treated mice. Dotted lines indicate FMO control in the OVA⁺ mDCs. Representative plots are shown. All data are representative of three independent experiments. (d-e) B6 mice were infected with PR8 and treated i.t. with 60 μg of OVA on day 2. 24h later, smDCs and rDCs were sorted from the spleen and co-cultured with CTV-labeled OT-I cells with or without OVA_257–264 peptide as indicated. (d) Proliferation of OTI cells after 72h. (e) Expression of CD25 and CD69 in OTI cells. Data are representative of four independent experiments. All values were obtained in triplicate. Data are shown as the mean ± SD. ***P < 0.001. P values were determined by one-way ANOVA with post-hoc Turkey’s multiple comparison test.

Correlating with these results, similar to mDCs from the med-LN (Fig. 3c, right panel) smDCs but not rDCs displayed membrane OVA_257–264-H-2Kbcomplexes (Fig. 3c, left panel), thus mimicking the data obtained with the WSN-OVA virus (Fig. 1g–i). Furthermore, smDCs but not rDCs from the spleen of PR8+OVA infected mice efficiently cross-primed naïve OTI cells (Fig. 3d and e). Altogether, these data demonstrate that smDCs bore and cross-presented lung-derived Ags to splenic CD8⁺ T cells after influenza virus infection.

smDCs are lung migratory DCs

We next investigated whether smDCs locally acquired pulmonary-derived Ags in the spleen or were lung mDCs that homed into the spleen after capturing Ags in the lung, similar to mDCs from the med-LN. First, we labeled lung DCs by i.t. instilling CFSE into the lungs of day 2 influenza-infected mice, which has been widely used to track lung mDCs after they migrate out of the lungs (7, 8, 34, 35). One day later, we analyzed the presence of lung-derived CFSE⁺ DCs in the med-LN, non-draining pLN, and spleen (Fig. 4a; Supplementary Fig. 4a, b). In agreement with previous studies (7, 8, 27), mDCs but not rDCs were CFSE⁺ in the med-LN of the CFSE-treated mice (Supplementary Fig. 4a). As a control, we failed to detect CFSE⁺ mDCs in non-draining LNs (Supplementary Fig. 4b), which was consistent with previous literature showing that CFSE-staining after i.t. administration is restricted to lungs and lung-derived cells (7, 8, 27).

Figure 4. smDCs migrate from the lung.

(a) PR8-infected B6 mice were treated i.t. with PBS or CSFE on day 2. The frequency of CFSE⁺ cells within the splenic smDCs and rDCs was evaluated 24h later. Data are shown as the mean ± SD (n=5 mice/group). ***P < 0.001. P values were determined by one-way ANOVA with post-hoc Turkey’s multiple comparison test. n.s not statistically significant. Data are representative of four independent experiments. (b) PR8-infected B6 mice received CSFE intravenously on day 2. Frequency of CFSE⁺ cells within mDCs and rDCs in the spleen 24 h later. Data are shown as the mean ± SD (n=5 mice). Data are representative of two independent experiments. (c) PR8-infected B6 mice were treated with CSFE on day 2. Mice also received 60 μg of OVA-647 i.t. or injected in the footpad (f.p.). Frequency of OVA-647⁺ cells within CFSE⁺ and CFSE⁻ smDCs in the spleen 24 h later. Data are shown as the mean ± SD (n=5 mice/group). ***P < 0.001. P values were determined by one-way ANOVA with post-hoc Turkey’s multiple comparison test. Data are representative of two independent experiments. (d) Concatenated plots from 5 independent samples showing the frequency of cDC1s and cDC2s within CFSE⁺ mDCs 24 h after CFSE administration. Data are representative of four independent experiments. (e-f) Frequency of CFSE⁺ smDCs (e) and rDCs (f) within the total cells in the spleen of PR8-infected B6 (WT) and Ccr7^−/− mice 24 h after CSFE administration at the indicated time points. Representative dot plots are shown. Data are shown as the mean ± SD (n=4–5 mice/group). *P < 0.05, **P < 0.01. P values were determined using a two-tailed Studentś t-test. Data are representative of two independent experiments. (g-k) 50:50 WT/Ccr7^−/− mixed BM chimeras were infected with PR8, treated with CFSE on day 2, and analyzed 24h later. The Frequency of WT (CD45.1⁺) and Ccr7^−/− (CD45.2⁺) cells within the DCs in the lungs (h), the CFSE⁺ mDCs in the med-LN (i), the CFSE⁺ smDCs in the spleen (j), or the splenic rDC (k) are shown. Data are shown as the mean ± SD (n=5 mice/group). Representative plots are shown. ***P < 0.001. P values were determined using a two-tailed Studentś t-test. Data are representative of three independent experiments.

Similar to our observations in the med-LN, smDCs but not rDCs were CFSE⁺ in the spleen of CFSE-treated mice (Fig. 4a). The kinetics of CFSE⁺ DC appearance in the spleen was, however, delayed compared to the med-LN (Supplementary Fig. 4c and d). Importantly, the lack of CFSE labeling in the rDC population was not due to the inability of these cells to acquire CFSE. In fact, rDCs more efficiently stained with CFSE than smDC when CFSE was administered intravenously (i.v) instead of i.t. (Fig. 4b). These results likely rule out the possibility that differences in CFSE labeling between rDCs and mDCS after i.t. administration were due to an enhanced capacity of smDCs to acquire residual “blood-derived” CFSE that could have leaked into the blood after i.t. treatment. Instead, these results suggest that, similar to mDCs in the med-LN, CFSE⁺ smDC were lung mDCs.

To further support this conclusion, we assessed whether smDCs bearing lung-delivered Ags in the spleen were lung-derived. To do this, we i.t. instilled CFSE into the lungs of influenza-infected mice, co-administered OVA-647 i.t., and analyzed the frequency of OVA-bearing cells within the CFSE⁺ and CFSE⁻ smDCs 24h later. We found that CFSE⁺ smDCs but not CFSE⁻ mDCs were OVA⁺ (Fig. 4c). As a control, we failed to detect Alexa-647⁺ cells within the CFSE⁺ smDCs when OVA-647 was administered in the footpad instead of i.t. (Fig. 4c). These data suggest that rather than capturing Ag that freely drained to the spleen, smDCs are lung mDCs that acquire influenza-derived Ags in the lungs and then traffick into the spleen to cross-prime influenza-specific CD8⁺ T cells. Further characterization of the smDCs revealed that, while CFSE⁺ mDCs in the lungs and the med-LN were a mix of cDC1 (XCR1^hiCD103⁺CD11b^loSIRP1α^loCCR7^h) and cDC2 cells (XCR1^loCD103⁻ CD11b^hiSIRP1α^hiCCR7^hi/med) (Supplementary Fig. 4e and Fig. 4d), CFSE⁺ smDCs contained mainly cDC2 cells (Fig. 4d). Together these results suggest that smDCs are lung migratory cDC2 cells.

Intriguingly, the lungs and the spleen are not connected by the lymphatic vasculature. Thus, we hypothesized that smDCs were lung mDCs that, after migrating into the med-LN, egressed through the efferent lymphatics to subsequently home into the spleen. To test this possibility, we assessed whether inhibiting the trafficking of mDCs from the lungs into the med-LN prevented the accumulation of smDCs in the spleen (Fig. 4e–k). Given that migration of mDCs from the lungs into the med-LN is CCR7-dependent (7, 36, 37), we first enumerated CFSE⁺ smDCs in WT and Ccr7^−/− mice. We found that CFSE⁺ smDCs were virtually absent in the spleen of influenza-infected Ccr7^−/− mice relative to WT counterparts (Fig. 4e). By contrast, no differences were detected within the rDC population (Fig. 4f). To confirm that the requirement for CCR7 was cell-intrinsic, we generated 50:50 WT/Ccr7^−/− mixed bone marrow (BM) chimeras (Fig. 4g–k), infected them with influenza, instilled CFSE i.t., and analyzed the presence of WT (CD45.1⁺) and Ccr7^−/− (CD45.2⁺) CFSE⁺ DCs in the med-LN and spleen 24h later. DCs in the lungs were a mixture of WT and Ccr7^−/− DCs (Fig. 4h). As predicted, because Ccr7^−/− DCs cannot migrate out of the lung, we failed to detect CD45.2⁺CFSE⁺ mDCs (Ccr7^−/−) in the med-LN (Fig. 4i). Similarly, all the CFSE⁺ smDCs in the spleen had a WT origin (Fig. 4j). In contrast, the rDC subset contained equivalent frequencies of WT and Ccr7^−/− cells (Fig. 4k). These results indicate that impaired migration of mDCs from the lungs to the med-LN prevented the accumulation of CFSE⁺ smDCs in the spleen. Collectively, these data suggest that, rather than passively acquiring Ags in the spleen, smDCs are lung-derived DCs that migrate to the spleen after egressing the med-LN.

smDCs egress from the mediastinal lymph node to prime splenic T cells

To determine whether smDCs were trafficking through the mLN, we assessed whether preventing med-LN egress affected the presence of smDCs in the spleen. Exit of lymphocytes from the LNs requires the expression of the sphingosine-1-phosphate (S1P) receptor 1 (S1PR1), which guides S1PR1-expressing cells into the efferent lymphatic vessels (38, 39). As such, treatment with the S1P synthetic analog FTY720 inhibits lymphocyte egress of S1PR1-expressing T and B cells from the LNs (38, 40). In agreement with previous studies (27, 41, 42), we found that mDCs in the med-LN expressed high levels of S1PR1 compared to non-migratory rDCs (Supplementary Fig. 5a). Corresponding with this observation, treatment with FTY20 resulted in the accumulation of lung-derived mDCs in the med-LN of influenza-infected mice (Supplementary Fig. 5b-c). Importantly, correlating with the retention of mDCs in the med-LN, the frequency (Fig. 5a) and number (Fig. 5b) of CFSE⁺ smDCs were significantly reduced in the spleen of FTY720-treated mice compared to control counterparts. The number of rDCs was, however, similar in both groups of mice (Fig. 5c). As a control, no differences were observed in the frequency or number of CFSE⁺DCs in the lungs of FTY720-treated compared to control mice (Supplementary Fig. 5d–f). These results indicate that blocking LN egress prevented the accumulation of lung-derived smDCs in the spleen, thereby suggesting that smDCs were lung mDCs that captured Ag in the lungs and migrated to the spleen via the med-LN.

Figure 5. Blockade of smDC egress from the mLN prevents T cell priming in the spleen.

(a-c) On day 2, PR8-infected B6 mice received CSFE i.t. and were treated with FTY720 or PBS (control) intraperitoneally (i.p.). Mice received a second injection 8h later. The frequency (a) and number (b) of CFSE⁺ mDCs and the number of rDCs (c) in the spleen on day 3 are shown. Representative plots gated on mDCs. Data are shown as the mean ± SD (n=4–5 mice/group). Data are representative of four independent experiments. **P < 0.01. P values were determined using a two-tailed Studentś t-test. (d-h) CTV-labeled OTI cells were transferred into day 1 PR8-infected B6 mice. On day 2, OVA (60 μg) was instilled i.t. and FTY720 or PBS (control) was administered i.p. Cells from the med-LN (e-f) and spleen (g-h) were analyzed 24h later. Representative plots are shown. Data are shown as the mean ± SD (n=4–5 mice/group). Data are representative of five independent experiments. ***P < 0.001. P values were determined using a two-tailed Studentś t-test. (i-l) On day 1, PR8-infceted B6 mice were treated or not with anti-CD62L Ab, and CTV-labeled OTI cells were adoptive transferred 30 min later. On day 2, the indicated groups received OVA or PBS i.t. and were treated i.p. with FTY720 or PBS. Cells from the spleen were analyzed on day 3. The schematic of the experiment is shown in (i). The number of total OTI donors (j), the frequency of CD69⁺ cells within the OTI donors (k), and the number of CD69⁺ OTI cells are shown (l). Data are shown as the mean ± SD (n=4–5 mice/group). Data are representative of three independent experiments. **P < 0.01, ***P < 0.001. P values were determined by one-way ANOVA with post-hoc Turkey’s multiple comparison test. n.s not statistically significant. (m-n). On day 2, PR8-infected B6 mice received 60 μg of OVA i.t. and were treated i.p. with FTY720 or PBS (control). 24h later, smDCs were sorted and co-cultured with CTV-labeled OT-I cells. The relative number of proliferating CTV^lo OTI cells (m) and the expression of CD25 (n) after 72h are shown. All values were obtained in triplicate. Representative histograms are shown. Data are shown as the mean ± SD. **P < 0.01. P values were determined by two-tailed Studentś t-test. Data are representative of three independent experiments.

To further support this conclusion, we assessed whether egress of lung-derived mDCs from the med-LN was required for priming CD8⁺ T cells in the spleen. To do this, we first transferred CTV-labeled OTI cells into PR8-infected mice. One day later, mice received OVA i.t. and were treated with FTY720 to prevent mDC egress from the med-LN (Fig. 5d). CD8⁺ T cell activation was evaluated in the med-LN and the spleen 24 h after OVA administration. The frequency of donor OTI cells in the med-LN and the spleen was similar in the FTY720 and control-treated groups (Fig. 5e, g). No differences in the frequency of activated OTI cells were detected in the med-LN (Fig. 5f). However, correlating with the lack of lung-derived smDCs (Fig. 5a), OTI cells failed to become activated in the spleen of mice treated with FTY720 relative to control mice (Fig. 5h). These results suggest that blockade of smDC egress from the med-LN prevented priming of CD8⁺ T cells in the spleen. One alternative explanation, however, is that FTY720 treatment inhibited the egress of an “early” wave of activated OTI cells from the med-LN to the spleen, thereby resulting in a diminished frequency of activated OTI in the spleen. While this remains a possibility, it is unlikely because 24h is not sufficient for proliferation and CD69 downregulation, which precedes LN egress (24). As such, all the activated OTI cells in the med-LN remained undivided and were CD69^hi at the time of the analysis (Supplementary Fig. 5g). Nonetheless, to further rule out this possibility, we inhibited med-LN priming with anti-CD62L and treated the mice with FTY720 (Fig. 5i). As expected, treatment with anti-CD62L efficiently prevented the entry of donor OTI cells in the med-LN in the anti-CD62L and anti-CD62L+ FTY720 treated mice (Supplementary Fig. 5h). In contrast, OTI donors similarly accumulated in the spleen in all groups (Fig. 5j). In agreement with our data in Fig. 2h, OTI cells were normally primed in the spleen of mice that received anti-CD62L alone (Fig. 5k and l). Co-treatment with FTY720, however, prevented splenic T cell priming in these mice (Fig. 5k and l). Given that OTI cells were absent from the med-LN after CD62L administration, these results ruled out the possibility that lack of activated CD8⁺ T cells in the spleen of anti-CD62L+FTY720 mice relative to anti-CD62L alone counterparts was due to the sequestration of activated T cells in the med-LN.

Finally, to more directly test whether the lack of splenic priming in the FTY720-treated mice was due an intrinsic defect in smDCs, we tested the capacity of smDCs from control and FTY720-treated mice to cross-prime CD8⁺ T cells in vitro. We administered OVA i.t. to day 2 influenza-infected mice, treated them or not with FTY720, and sorted smDCs from the spleen 24h later. Sorted smDCs were co-cultured with naïve OTI cells and T cell expansion and activation were assessed 72 h later. We found that smDCs from FTY720-treated mice poorly cross-primed OTI cells compared to smDCs from control counterparts (Fig. 5m and n). Collectively, these results suggest that FTY720 treatment impaired the migration of Ag-bearing, lung-derived mDCs from the med-LN to the spleen, thereby preventing the cross-priming of Ag-specific CD8⁺ T cells in the spleen after influenza infection.

Spleen-primed CD8⁺ T cells give rise to long-lived memory

Our data indicate that influenza-specific CD8⁺ T cells can be primed by lung-derived mDCs in both the med-LN and the spleen. Thus, we next investigated whether spleen-primed (SP-P) and med-LN-primed (LN-P) CD8⁺ T cells gave rise to progeny with different functional properties. Previous studies suggest that the transcriptional signature initiated at the time of the initial interaction with DCs (12–24h p.i.) fine-tunes the fate of the ensuing CD8⁺ T cell response (43–47). Thus, we first characterized the transcriptome profile of the SP-P and LN-P CD8⁺ T cells. We sorted SP-P and LN-P OTI cells from PR8+OVA infected mice 24 h after OVA administration and performed RNA-sequencing (RNA-seq) (Fig. 6a–f and Supplementary Fig. 6a–b). We also included naïve OTI cells in our studies. We found that SP-P and LN-P cells similarly up-regulated key Teff differentiation genes, such as Tbx21, Il2r, Batf, Irf4, and Gzmb compared to naïve T cells (Fig. 6a). Principal component analysis (PCA) segregated SP-P and LN-P cells into two separate clusters that, while distinct from naive T cells, were also transcriptionally different from each other (Fig. 6b). Approximately 1,900 genes were differentially expressed (FDR <0.05, FC ≥1.5) between SP-P and LN-P cells (Fig. 6c and Data file S1). Interestingly, a substantial number of transcripts that are characteristically expressed in long-lived, self-renewing memory T cells were highly expressed in SP-P relative to LN-P cells (Fig. 6c and d) (48–57). Conversely, SP-P cells had reduced mRNA for the inhibitory receptors Lag3, Pdcd (encoding PD1), and Ctla4 (Fig. 6c and d), which are upregulated in terminal Teff cells and exhausted CD8⁺ T cells (58).

Figure 6. Spleen- and med-LN-primed CD8⁺ T cells are transcriptionally distinct.

(a-f) OTI cells (CD45.2) were adoptively transferred into day 1 PR8-infected CD45.1 mice. The next day, recipient mice were i.t. treated with 60 μg of OVA and activated CD69⁺CD62L^lo OTI cells were sorted from the spleen (spleen-primed, SP-P) and med-LN (med-LN-primed, LN-P) 24 later. Naïve OTI cells were also purified from naïve OTI mice and RNA-seq was performed. (a) Heatmap showing the z-score of expression of key Teff genes in SP-P, LN-P, and naïve OTI cells. (b) Principal component analysis (PCA) of normalized gene expression (counts per million) based on all detected genes in naïve (open circles), SP-P (purple triangles), and LN-P (green squares) cells. Each symbol represents an individual sample (c). Volcano plot highlighting genes differentially expressed in SP-P over LN-P cells (FDR <0.05, ≥1.5 fold change). (d). Heatmap displays the z-score of expression of characteristic long-lived memory and terminal Teff genes that are differentially expressed in SP-P relative to LN-P cells. (e and f) Gene set enrichment analysis (GSEA) for the indicated gene signatures in SP-P vs. LN-P cells. Normalized enrichment score (NES). Three replicates for each cell type were obtained from three independent experiments. Genes for GSEA were ranked by -log10(p-value) times the sign of the fold change for the paired SP-P vs. LN-P comparison. (g-h) WT mice were infected with PR8 and the frequency (g) and number (h) of TCF1^lo and TCF1^hi NP-specific CD8⁺ T cells were determined at different times after infection in the med-LN. Data are shown as the mean ± SD (n=4–5 mice/group). Representative plots are shown. Data are representative of two independent experiments.

Among the top differentially expressed genes between SP-P and LN-P cells, Tcf7 (TCF1), a transcription factor that is critical for the differentiation of long-lived memory CD8⁺ T cells and their precursors (53, 54, 56, 57) was markedly upregulated in SP-P relative to LN-P cells. In agreement, Gene Set Enrichment Analysis (GSEA) revealed that the SP-P transcriptome was significantly enriched relative to the LN-P transcriptome in targets positively regulated by TCF1 (53, 57)(Fig. 6e) and previously identified gene signatures of TCF1^hi long-lived, self-renewing memory CD8⁺ T cells (51, 54, 57)Fig. 6f). Furthermore, Ingenuity Pathway Analysis (IPA) indicated that transcriptional networks promoting cell maintenance, survival, and hematopoietic progenitor cell development were induced in SP-P versus LN-P cells (Supplementary Fig. 6b).

Most of the studies on TCF1^hi memory CD8⁺ T cells have been performed using tumor or chronic infection models (53, 54, 56, 57). To characterize the kinetics of expression of TCF1 after acute influenza virus infection, we enumerated TCF1-expressing, influenza nucleoprotein (NP)-specific CD8⁺ T cells at different times after infection (Fig. 6g and h). Not surprisingly, the majority of NP-specific CD8⁺ T cells were TCF1^lo at the peak of the infection. However, the frequency and number of TCF1^loNP-specific CD8⁺ T cells sharply declined thereafter. As a result, virtually all the long-lived NP-specific memory CD8⁺ T cells identified 300 days after infection were TCF1^hi (Fig. 6g–h and Supplementary Fig. 6c). Similar data were obtained in the spleen and lungs (Supplementary Fig. 6c). Thus, similar to tumor or chronic infection models, TCF1 expression was a marker of long-lived memory CD8⁺ T cells after influenza infection. Collectively, these results suggested that SP-P T cells had transcriptional characteristics typically associated with long-lived TCF1^hi memory CD8⁺ T cells and their precursors. In contrast, LN-P T cells resembled terminal Teff cells.

Next, we examined the fate of SP-P and LN-P cells in vivo. To do this, we sorted congenically different SP-P and LN-P OTI cells from the spleen and med-LN of PR8+OVA infected mice, mixed together in a 1:1 ratio, and intravenously injected 4×10³ cells into WSN-OVA-infected secondary recipient mice (Fig. 7a). The relative frequencies and phenotype of SP-P and LN-P-derived cells were analyzed at different times after infection in the lung, med-LN, and spleen. Cells derived from each donor and from the host were distinguished based on the differential expression of CD45.1 and CD45.2 (Fig. 7b). We found similar frequencies of SP-P and LN-P-derived cells in the lungs on day 7 after infection (Fig. 7b). At this time, all donor cells were TCF1^lo regardless of their origin (Fig. 7c). Furthermore, SP-P and LN-P produced similar amounts of IFN-γ and Granzyme B, and both displayed a similar CD62L^loCD127^loT-bet^hi Teff phenotype (Fig. 7d–f). These results suggest that SP-P and LN-P cells similarly home into the lungs and differentiate into bona fide Teff cells at the peak of the infection.

Figure 7. Spleen-primed CD8⁺ T cells more efficiently generate long-lived memory cells than the med-LN-primed counterparts.

(a-k) CD45.2⁺CD45.1⁺ and CD45.2⁺CD45.1⁻ OTI cells were adoptively transferred into day 1 PR8-infected mice (CD45.2⁻CD45.1⁺). The next day, recipient mice were i.t. treated with 60 μg of OVA. Activated CD69⁺CD62L^lo OTI cells were sorted from the spleen (spleen-primed, SP-P) and med-LN (med-LN-primed, LN-P) 48 later. 4×10³ cells containing a 1:1 mixture of congenically different SP-P and LN-P cells was intravenously injected into secondary recipient mice (CD45.2⁻CD45.1⁺) that were infected with WSN-OVA 24 h later. (a) Experimental diagram. (b) Frequency of SP-P and LN-P cells within the OTI donors in the lungs 7 days after infection. (c) TCF1 expression in total donor OTI cells (left) and SP-P and LN-P donor OTI cells (right) from the lungs. (d-e) Production of IFN-γ (d) and Granzyme B (e) by SP-P and LN-P cells from the lungs. (f) Expression of CD127, CD62L, and T-bet in SP-P and LN-P cells from the lungs. (g) Frequency of TCF1^lo and TCF1^hi within the total OTI donors in the med-LN 7 days after infection. (h) Frequency of SP-P and LN-P cells within the TCF1^lo and TCF1^hi OTI donor cells in the med-LN 7 days after infection. (i) Frequency of TCF1^lo and TCF1^hi within the OTI donors in the med-LN 45 days after infection. (j) Frequency of SP-P and LN-P cells within the TCF1^lo and TCF1^hi OTI donor cells in the med-LN 45 days after infection. (k) Frequency of SP-P and LN-P cells within the total OTI donors in the med-LN 45 days after infection. Data are shown as the mean ± SD (n=6–7 mice). Representative plots are shown. Data are representative of four independent experiments. Data from the reverse transfer is shown in the Supplementary Figure. 6g. **P < 0.01, ***P < 0.001. P values were determined using a paired two-tailed Studentś t-test. n.s not statistically significant.

Whereas all donor cells in the lungs were TCF1^lo (Fig. 7c), close to 20% were TCF1^hi in the med-LN on day 7 after infection (Fig. 7g). Importantly, we found that SP-P donors expressed higher levels of TCF1 compared to the LN-P counterparts (Supplementary Fig. 7a). In contrast, T-bet expression was similar in both populations (Supplementary Fig. 7a), which is consistent with our RNA-seq data (Fig. 6a and d). As a result, whereas the TCF1^lo population contained an equivalent mix of LN-P and SP-P-derived cells, nearly 80% of the TCF1^hi cells were SP-P-derived cells on day 7 (Fig. 7h). Similar results were obtained in the spleen (Supplementary Fig. 7b–c).

As predicted, on day 45 after infection, the majority of the remaining OTI donors were TCF1^hi (Fig. 7i). Both, TCF1^hi and TCF1^lo cells were T-bet^loIRF4^lo (Supplementary Fig. 7d). In contrast, as previously described (53), TCF1^hi cells expressed high levels of Eomes compared TCF1^lo cells (Supplementary Fig. 7d). When donor cells were segregated into TCF1^loEomes^lo and TCF1^hiEomes^hi cells, we observed no significant differences in the relative frequency of LN-P and SP-P-derived cells within the TCF1^loEomes^lo population (Fig. 7j). In contrast, nearly 80% of the TCF1^hiEomes^hi donor cells were SP-P cells (Fig. 7j). As a result, SP-P-derived cells largely accumulated compared to LN-P counterparts on day 45 after infection (Fig. 7k). Similar results were obtained in the lungs, although the frequency of TCF1^hiEomes^hi cells in the lungs was relatively lower compared to the med-LN (Supplementary Fig. 7 e–f). As a control, similar results were obtained when SP-P cells derived from the CD45.2⁺CD45.1⁻ instead of CD45.2⁻CD45.1⁺ donors (Supplementary Fig. 7g). Altogether, these results indicate that SP-P cells differentiate into TCF1^hi long-lived memory CD8⁺ T cells more efficiently than LN-P cells after influenza virus infection.

Finally, we examined the capacity of LN-P and SP-P memory cells to expand and re-differentiate into Teff cells after re-challenge. We sorted CD8⁺ T cells from the secondary recipient mice on day 45 after primary infection and adoptively transferred a mix of LN-P and SP-P cells into naïve CD45.1⁺ tertiary recipient mice. One day later, we challenged the tertiary recipients with WSN-OVA and assessed the LN-P and SP-P responses in the lungs on day 7 after the challenge (Fig. 8a). Phenotypically, LN-P and SP-P cells were indistinguishable from each other since both downregulated TCF1 and acquired a similar CD127^loCD62L^loT-bet^hiIRF4^hiBlimp-1^hi Teff phenotype after re-challenge (Fig. 8b). Importantly, however, SP-P cells largely expanded compared to LN-P cells, as evidenced by the differences in fold-expansion (Fig. 8c). Correspondingly, the ratio of SP-P to LN-P cells dramatically increased after re-challenge (Fig. 8d). As a result, the majority of the TCF1^lo Teff donor cells found in the lungs derived from the SP-P donors (Fig. 8e). Similar results were obtained when we enumerated LN-P and SP-P cells within the Granzyme B/IFN-γ double-producing OTI cells (Fig. 7e) or when the SP-P and LN-P response was evaluated in the med-LN (Supplementary Fig. 8a–d) or the spleen (Supplementary Fig. 8e–g). These results demonstrated that SP-P-derived memory cells had a superior capacity to expand and repopulate the Teff compartment after re-challenge compared to LN-P cells.

Figure 8. Spleen-primed memory CD8⁺ T cells have a superior recall capacity compared to med-LN-primed memory cells.

(a-f) Congenically different spleen-primed (SP-P) and med-LN (LN-P) OTI cells were generated and transferred into WSN-OVA-infected secondary recipient mice as in Fig. 6a. On day 45 after primary infection, CD8⁺ T cells were sorted from the spleen, and a mixture containing 2×10³ LN-P and 2.8×10³ SP-P cells was intravenously injected into tertiary recipient mice (CD45.2⁻CD45.1⁺), which were infected with WSN-OVA 24 h later. (a) Experimental diagram. (b) Expression of TCF1, T-bet, Blimp-1, IRF4, CD127, and CD62L in SP-P and LN-P donors in the lungs of tertiary recipient mice 7 days after re-challenge. (c) The relative fold expansion of SP-P- and LN-P OTI cells in the lungs of tertiary recipient mice was calculated on day 7. (d) The ratio of SP-P to LN-P OTI cells in the mixture of transferred cells (d0) and the ratio of SP-P to LN-P OTI cells in the lungs of tertiary recipient mice on day 7 after re-challenge (d7) in the lung. (e) The frequency of SP-P and LN-P cells within the TCF1^lo OTI donors on day 7 after re-challenge. (f) The frequency of SP-P and LN-P cells within the Granzyme B/IFN-γ double-producing OTI cells on day 7 after re-challenge. Data are shown as the mean ± SD (n=12 mice). Representative plots are shown. Data are representative of three independent experiments. For simplicity, only data from the group in which SP-P cells are CD45.2⁺CD45.1⁺ are shown. ***P < 0.001. P values were determined using a paired two-tailed Studentś t-test.

Collectively, our data demonstrated that priming of CD8⁺ T cells in the spleen by lung migratory smDCs resulted in the generation of CD8⁺ T cell precursors with enhanced ability to differentiate into long-lived memory cells and superior re-challenge capacity compared to med-LN-primed counterparts.

DISCUSSION

The current paradigm suggests that lung mDCs prime CD8⁺ T cell responses to influenza in the med-LN and from there, CD8⁺ T cells disseminate systemically after 3–4 days of active proliferation (1–4). The results presented here challenge this paradigm by demonstrating that activated antigen-specific CD8⁺ T cells are observed in the spleen of influenza-infected mice before the systemic spreading of the LN-primed CD8⁺ T cells. Importantly, inhibition of LN priming did not prevent the presence of early-activated CD8⁺ T cells in the spleen, thus indicating that splenic-activated CD8⁺ T cells did not have a LN origin. Instead, we demonstrate that influenza-specific CD8⁺ T cells are locally primed in the spleen by lung-derived mDCs bearing influenza-derived antigens. Thus, our data show that, rather than exclusively acting as a receptacle for LN-primed CD8⁺ T cells, the spleen is a primary site for the initiation of the CD8⁺ T cell responses to influenza by lung-derived mDCs.

Our findings support a model in which lung mDCs capture exogenous antigens in the lung, process them through the MHC class I pathway, and migrate to the spleen via the med-LN to cross-prime CD8⁺ T cells. Supporting this model, when DCs were unable to migrate out of the lung, or when LN egress was inhibited, DCs bearing lung-derived antigens failed to accumulate in the spleen, thereby preventing splenic T cell priming. Altogether, our results demonstrate a new path of DC migration that connects the lungs with the spleen, hence providing an explanation for how splenic CD8⁺ T cell responses are initiated in response to respiratory challenges. These findings reconcile the conflicting reports in the literature showing that, while lung-derived antigens do not freely reach the circulation (3, 27, 33), activated T cells can be found in the spleen early after infection and T cells responses are initiated even in the absence of LNs (18, 19, 21).

Egress of mDCs from the med-LN was dependent on S1P/S1PR interactions, which regulate efferent lymphatic entry (38, 39). Neither the med-LN nor the lungs is directly connected to the spleen by the lymphatic vasculature. Thus, the only possible route for mDCs to access the spleen from the med-LN is via the thoracic duct, which channels the efferent lymph back into the blood circulation (38). Supporting this model, while technically challenging (14), previous studies demonstrate that DCs are present in the blood (16, 17, 59) and the thoracic duct lymph fluid of rats, mice, and sheep (12–17). Hence, altogether these data suggest that the same mechanism that regulates the exit of T and B cells from the LNs (38, 39) controls the egress of mDC from the med-LN. Interestingly, our results suggest that cDC2 but not cDC1 lung-derived mDCs leave the med-LN. These differences could be explained by the differences in CCR7 expression between these two subsets since CCR7 counteracts S1P-mediated exit from the LNs (60) and cDC1 cells characteristically express high levels of CCR7 relative to cDC2 cells (26, 61, 62).

The role of S1P/S1PR in lung mDC migration is not entirely surprising since previous studies show that mDCs express mRNA for all five known S1P receptors in the tissues (41, 42) and further upregulate S1PR1 expression after entering the corresponding draining LN (42). Interestingly, while the role of S1P/S1PR signaling in controlling mDC egress from the LNs was not evaluated, these studies indicate that blockade of S1P/S1PR after FTY720 treatment prevents the migration of DCs from the tissues to the draining LN (41, 42). Intriguingly, we found that lung mDCs trafficked normally into the med-LN in the absence of S1P/S1PR interactions. One potential explanation for this discrepancy is that in the highly inflammatory environment induced after influenza infection, sustained CCR7/CCL21-mediated chemotaxis is sufficient for guiding mDCs out of the lung. In this scenario, potent CCR7 up-regulation by mDCs after infection may compensate for the lack of S1P/S1PR interactions in the lungs of FTY720-treated mice. However, once in the med-LN, S1P/S1PR signaling appears required for the egress of mDCs, and its absence cannot be compensated by additional overlapping mechanisms.

Importantly, we found that priming of CD8⁺ T cells in the spleen and med-LN gives rise to a progeny of cells with different fates. Multiple mutually complementary mechanisms may account for this effect. For example, because cDC1s do not migrate out of the med-LN, all lung-derived Ags in the spleen are likely to be presented to CD8⁺ T cells by lung-derived cDC2s. Importantly, cDC1 produce large amounts of IL-12 (63–65) and express high levels of CD24 (66), signals that favor the generation of terminal Teff cells, but hinder the development of CD8⁺ T cell memory precursors and long-lived memory CD8⁺ T cells (66–69). Conversely, cDC2s express high levels of CD70 (8), which promotes memory CD8⁺ T cell differentiation (9, 70). Like cDC1s, monocyte-derived DCs also accumulate in the med-LN after influenza infection and produce large amounts of IL-12 (71). Therefore, it is reasonable to speculate that priming of naive CD8⁺ T cells in the presence of cDC1s and/or IL-12-producing monocyte-derived DCs in the med-LN skews T cell differentiation towards the terminal Teff differentiation pathway. On the contrary, the interaction of naïve T cells with cDC2 cells in the spleen gives rise to a progeny of relatively less differentiated and more plastic cells with the ability to become long-lived, self-renewing memory cells. Supporting this view, whereas T cells stimulated by cDC1 cells preferentially differentiate into Teff cells, activation by cDC2 cells give rise to CD8⁺ T cells with characteristics of progenitor cells poised to become memory CD8⁺ T cells (66).

In addition to the nature of the APC itself, the inflammatory and metabolic environment could be different in the spleen and the highly reactive med-LN. Similarly, due to the differences in size, the ratio of APCs to T cells, architecture, and the duration and/or frequency of DC:T cell contacts may be different in both organs. All these factors have been shown to direct the Teff/memory precursor differentiation balance by fine-tuning the early transcriptional program initiated during the initial encounter with DCs (43–47). While future studies are needed to understand key drivers, our results suggest a model in which priming of CD8⁺ T cells responding to the same antigen in different anatomical locations gives rise to cells with distinct functional capabilities, thereby providing a new mechanism for how T cell diversity is generated after infection.

In summary, our results identify a DC trafficking pathway that connects the lung with the blood circulation and identify the spleen as a primary site for the priming of long-lived memory T cell precursors. These data will be critical for the development of more efficient vaccination and therapeutic strategies to respiratory challenges.

MATERIALS AND METHODS

Study design

The goal of this study was to determine whether lung migratory DCs primed CD8⁺ T cell responses to influenza in the spleen and the mechanisms and functional consequences of splenic priming. Flow cytometry was used to identify lung-derived DCs in the spleen. The capacity of lung-derived DCs to prime influenza-specific CD8⁺ T cells responses in the spleen was confirmed in vivo and in vitro using various knock-out mice and treatments. The functional differences between med-LN and spleen-primed CD8⁺ T cells were determined using adoptive transfer experiments and RNA-seq studies.

Mice.

C57BL/6 (B6), B6.SJL-Ptprca Pepcb/BoyJ (B6.CD45.1), C57BL/6.OTI (OTI), B6.129P2(C)-Ccr7 ^tm1Rfor/J (Ccr7^−/−), and C57BL/6J-Lta^hlb382/J (Lta^−/−) mice were originally obtained from Jackson. All mice were bred in the University of Alabama at Birmingham (UAB) animal facility. All experimental procedures involving animals were approved by the UAB Institutional Animal Care and Use Committee and were performed according to guidelines outlined by the National Research Council. Control and experimental groups were age and sex-matched. Males and females were used in these experiments. All mice were eight week-old at the time of the experiment.

Infection, viral foci assay BM chimeras and in vivo treatments.

Infections were performed intranasally (i.n.) with 0.2 LD₅₀ of influenza A/PR8/34 (PR8) or A/WSN/33-OVA/I influenza virus diluted in 100 μl of PBS. In some experiments, mice were sensitized i.n. with 100μg of HDM extract labeled with the AF647 labeling kit (Invitrogen) prior to administration to mice. Viral foci assays were performed by homogenizing lung samples in zero–serum medium with 4 μg/ml trypsin. 5–fold serial dilutions were performed in the same media. MDCK cell monolayers were infected with100 μl of each sample. MDCK cells were fixed with 80% acetone and infected cells were detected using an antibody against NP (Clone A3. Millipore–Sigma) 16 h later. Spots were revealed using alkaline phosphatase–conjugated streptavidin (Life Technologies) and BCIP/NBT substrate (Moss Substrates Inc.).

BM chimeric mice were generated by irradiating the indicated recipient mice with 950 Rads from an X-ray source delivered in 2 equal doses administered 4–5h apart. Following irradiation, mice were intravenously injected with a 50:50 mix of BM cells (5 × 10₆ cells) obtained from the BM of naïve B6.CD45.1 and Ccr7^−/− mice. The irradiated mice were allowed to reconstitute for 8–10 weeks before influenza infection. Lung DCs were labeled in situ by intratracheally (i.t.) administering 30μl of 8 mM CFSE (Molecular Probes, ThermoFischer Scientific) in PBS 24h before sacrifice at the indicated times. In some experiments, mice instilled with 60 μg of Alexa-647-labelled OVA (OVA-647) (Molecular Probes, Thermo Fischer Scientific) i.t. or injected in the footpad 24h before sacrifice at the indicated times. Where indicated, mice were intraperitoneally (i.p.) injected with 100 μg of FTY720 (Cayman Chemicals) at the indicated times alone or in combination with 250 μg of anti-CD62L Ab (Mel-14, BioXCell)

Cell preparation and flow cytometry.

Lungs from infected mice were harvested, cut into small fragments and digested for 45 min at 37°C with 0.6 mg/ml collagenase A (Sigma) and 30 μg/ml DNAse I (Sigma) in RPMI-1640 medium (GIBCO). Digested lungs, LNs, and spleens were mechanically disrupted to obtained single cell suspensions. Red blood cells were lysed with 150 mM NH₄Cl, 10 mM KHCO₃ and 0.1 mM EDTA for 5 minutes. Cell suspensions were then filtered through a 70 μm nylon strainer. Cells were washed, resuspended, and incubated in PBS with 2% donor calf serum and 10 μg/ml FcBlock (2.4G2 -BioXCell) for 10 min on ice before staining with fluorochrome-conjugated antibodies. Fluorochrome-labeled anti-CD45.1 (clone A20, dilution 1/400), anti-CD45.2 (clone 104, dilution 1/400), anti-CD19 (clone 1D3, dilution 1/200), anti-CD25 (clone PC61, dilution 1/100), anti-CD62L (MEL-14, dilution 1/200), anti-CD69 (clone H1.2F3, dilution 1/200), anti-CD8α (clone 53–6.7, dilution 1/200), anti-CD40 (clone 3/23, dilution 1/400), anti-CCR7 (clone 4B12, dilution 1/200), anti-CD44 (clone IM7, dilution 1/500), anti-CD86 (clone GL1, dilution 1/400), anti-SiglecF (clone E50–2440, dilution 1/400), anti-CD80 (clone 16–10A1, dilution 1/400), anti-B220 (clone RA3–6B2, dilution 1/200), anti-MHC Class II (clone AF6–120.1, dilution 1/2500), and anti-CD11c (clone HL3, dilution 1/200) antibodies were obtained from BD Biosciences. Fluorochrome-labeled anti-CD127 (clone A7R34, dilution 1/200) antibody was purchased from eBioscience. Fluorochrome-labeled anti-T-bet (clone 4B10, dilution 1/200), anti-IRF4 (clone 3E4, dilution 1/2000), anti- mouse FcεRIα (clone MAR1, dilution 1/200), and anti-CD64 (clone X54–5/7.1, dilution 1/200), and anti-mouse H-2Kb bound to SIINFEKL (clone 25-D1.16, dilution 1/200) antibodies were purchased from BioLegend. Biotin-conjugated primary antibodies were detected with fluorochrome-labeled streptavidin from BD Biosciences Dead cell exclusion was performed using 7-AAD (Calbiochem). Intracellular staining for transcription factors was performed using the mouse regulatory T cell staining kit (eBioscience) following the manufacturer’s instructions. Flow cytometry was performed using an Attune NxT Flow Cytometer (ThermoFischer Scientific).

Cell purification, adoptive transfer, and sorting.

CD8⁺ OTI cells were purified from the spleens of naïve CD45.2⁺CD45.1⁻ B6.OTI or CD45.2⁺CD45.1⁺ B6.OTI mice by positive selection with the EasySep CD8⁺ positive selection kit (Stemcell Technologies). All T cell preparations were more than 95% pure as determined by flow cytometry. Where indicated, purified OTI cells were labeled with 1uL/mL CellTrace Violet (CTV) (Molecular Probes, ThermoFischer Scientific) for 20 minutes at 37°C. In the indicated experiments, purified OTI cells (6×10⁵) were transferred intravenously via the retro-orbital route into congenically different (B6.CD45.2⁻CD45.1⁺) recipient mice at the indicated time points. In the co-transfer experiments of LN-P and SP-P cells, secondary recipient mice received 4×10³ cells containing a 1:1 mixture of CD69⁺CD62^lo OTI cells sorted from the spleen (SP-P) and med-LN (LN-P) of previously PR8+OVA infected mice. DCs from pooled LNs and spleens were purified with LS columns using anti-CD11c MACs beads (Miltenyi). After staining with fluorochrome-conjugated antibodies, the indicated DC subsets were sorted. All sorting experiments were performed using a FACSAria (BD Biosciences) sorter in the University of Alabama at Birmingham Flow Cytometry core. All sorted DC and T cells populations were more than 95% pure as determined by flow cytometry.

In vitro cultures.

Sorted DC populations from the med-LN and spleen of the indicated animals were co-cultured with purified CTV-labelled OTI cells at a 1:10 ratio. Cells were cultured for 72 h at 37°C in 125 μl in round-bottomed 96-well plates in RPMI 1640 supplemented with sodium pyruvate, HEPES pH 7.4, non-essential amino acids, penicillin, streptomycin, 2-mercaptoethanol and 10% heat-inactivated FCS (all from GIBCO). In some experiments, OVA_257–264 peptide was added to the culture as a positive control.

Quantitative real-time PCR.

RNA was extracted from the indicated subsets of sorted DCs using the Norgen Single Cell RNA Purification Kit (Norgen Biotek) following the manufacturer’s instructions. Purified RNA was reverse transcribed with random hexamers and Superscript II (Invitrogen). Quantitative PCR was performed using Taqman master mix and primers and probes for S1pr1, S1pr2, S1pr4, and S1pr5 (Mm02619656_s1, ARGZGJC, Mm00468695_s1, Mm02620565_s1 Applied Biosystems). All reactions were run on a LightCycler 480 Real-time PCR System (Roche) located in the Heflin Center for Genomics Sciences at the University of Alabama at Birmingham. Transcript abundance was determined and then normalized to Gapdh mRNA. Data are reported as the fold-change in expression relative to rDCs.

RNA Sequencing

Primary Analysis

Library preparation and RNA sequencing was conducted through Genewiz. Libraries were sequenced using a 1×50bp single end rapid run on the HiSeq2500 platform. The quality of raw sequence fastq-formatted files was assessed using fastQC (http://www.bioinformatics.babraham.ac.uk/projects/fastqc). Sequences were trimmed using Trim Galore using phred33 scores (version 0.4.4, http://www.bioinformatics.babraham.ac.uk/projects/trim_galore), the paired setting and the nextera adapter option. Trimmed sequences were aligned with STAR aligner (72) (version 2.5.2a) using mouse mm10 (UCSC) genome (https://support.illumina.com/sequencing/sequencing_software/igenome.html) and ENCODE options (outputFilterMultimapNmax = 20, alignSJoverhangMin = 8, alignSJDBoverhangMin = 1, outFilterMismatchNmax = 999, alignIntronMin = 20, alignIntronMax = 1000000, alignMatesGapMax = 1000000). The STAR genome was constructed using the mm10 annotation file and sjdbOverhang = 100 as recommended in the documentation. Aligned reads were counted with HTseq-count (version 0.6.1p1) (73) set for un-stranded and using the mm10 genes.gtf annotation file.

Downstream Analysis

The R package edgeR (74) was used to assess differential expression between pairs of groups and to generate gene-by-sample matrices for both RPKM and counts per million (CPM). Genes were considered for further analysis if their CPM were above 1 for at least three samples. We used two models: first, we evaluated group differences between SA, LA and SN. Because SA and LA were derived from the same mouse per replicate, we then used a paired model just for the SA and LA comparison. Additional downstream analysis and visualization (principal components, clustered heat maps, volcano plots) was performed using custom Matlab (The Mathworks Inc., Natick MA, USA) scripts (https://github.com/afr-uab/Jenkins2021). Comparison of our data to published data sets GSE85367 (57), GSE20754 (53), GSE83978 (51), GSE84105 (54) was accomplished using Gene Set Enrichment Analysis (GSEA) (75). Identification of enriched functions based on differentially expressed genes was performed using Ingenuity Pathway Analysis (Ingenuity Pathway Analysis, Qiagen, Redwood City CA, USA)

Statistical analysis.

GraphPad Prism (Version 8) was used for data analysis. The statistical significance of differences in mean values was determined using two-tailed Student’s t test or one-way ANOVA with post-hoc Tukey’s multiple comparison test. P values of less than 0.05 were considered statistically significant. *P < 0.05, **P < 0.01, ***P < 0.001.

Supplementary Material

Supplemental Material

Figure S1. Gating strategy of smDCs in the spleen.

Figure S2. Cross-primed CD8⁺ T cells in the spleen after pulmonary challenges.

Figure S3. Splenic smDCs but not mDCs from non-draining LNs have lung-derived Ags.

Figure S4. smDCs are lung-derived cDC2 cells.

Figure S5. Effects of FTY720 in med-LN and lung mDCs.

Figure S6. Characterization of spleen-primed and med-LN primed CD8⁺ T cells.

Figure S7. Spleen-primed and med-LN primed CD8⁺ T cells differently populate the long-lived memory T cell compartment.

Figure S8. Spleen-primed and med-LN-primed re-challenge.

Data file S1. Genes were differentially expressed between SP-P and LN-P cells

Data file S2. Raw data file (Excel file).