Main

A fundamental tenet of immunology is that primary adaptive immune responses are initiated in secondary lymphoid organs, such as lymph nodes, Peyer's patches or spleen1,2,3. These lymphoid organs are organized to recruit naive lymphocytes from the blood and to promote their interaction with activated antigen-presenting cells from surrounding tissues1. Once lymphocytes have been activated and clonally expanded in centralized lymphoid organs, the resulting effector cells localize to the infected or inflamed tissues and perform their effector functions4. For example, B and T cells responding to influenza virus are first detected in the lymph nodes that drain the respiratory tract, and are only later found in the lung5.

Consistent with an essential role for lymphoid organs in primary immune responses, splenectomized lymphotoxin-α-null (Lta−/−) mice6,7 or alymphoplastic (aly/aly, also known as Map3k14−/−) mice8,9 — which lack spleen, lymph nodes and Peyer's patches — are unable to generate primary immune responses to a variety of pathogens10, antigens11 and allografts12. However, Lta−/− mice infected with respiratory viruses are capable of generating antigen-specific B and T cells13,14, albeit with delayed kinetics, suggesting that lymphocytes can also be primed in non-lymphoid tissues. Consistent with this idea, the bone marrow seems to function as an alternative site of lymphocyte priming to systemic antigens when lymph nodes are unavailable15,16. Another tissue that could facilitate primary immune responses to respiratory infections is the BALT17. Although the presence of BALT in mouse and human lungs is controversial, there are reports that infection or inflammation triggers the organization of lymphoid structures in the lungs of both species18,19,20,21. These structures do not fit the classical definition of BALT, as they are not formed independently of antigen22,23. Because inducible BALT (iBALT) appears in the lung only after infection or inflammation, it is generally assumed that iBALT is simply an accumulation of effector cells that were initially primed in conventional lymphoid organs; however, it is also possible that inflammatory responses directly trigger the neo-formation of iBALT, which promotes the recruitment, priming and expansion of antigen-specific lymphocytes in situ. Thus, iBALT may functionally replace conventional lymphoid organs in respiratory immune responses.

Results

Respiratory immunity in the absence of lymphoid organs

To determine whether CD8+ T cells could be primed in the absence of conventional lymphoid organs, we generated bone marrow chimeric mice lacking lymphoid organs or LTα (Supplementary Fig. 1 online). These chimeras include wild-type (WT) mice, which possess all lymphoid organs and are populated with Lta+/+ cells; lymphotoxin chimeric (LTCh) mice, which contain all lymphoid organs and are populated with Lta−/− cells; spleen-, lymph node– and Peyer's patch–deficient (SLP) mice, which lack lymphoid organs and are populated with Lta+/+ cells; lymphotoxin knockout (LTKO) mice, which lack lymph nodes and Peyer's patches and are populated with Lta−/− cells; and, finally, splenectomized LTKO (LTKO-S) mice, which lack lymphoid organs and are populated with Lta−/− cells. After reconstitution, we infected the chimeric mice with influenza and followed influenza-specific CD8+ T cells in the lungs. Although influenza nucleoprotein (NP)-specific CD8+ cells were observed in the lungs of WT, LTCh and SLP chimeric mice by day 7 (Fig. 1a), the appearance of NP-specific CD8+ cells was delayed until after day 9 in LTKO mice, and was even further delayed in LTKO-S mice. Despite the delayed appearance of NP-specific CD8+ cells in LTKO and LTKO-S mice, the total number of these cells in the lungs eventually reached almost wild-type levels (Fig. 1b).

Figure 1: Influenza-specific CD8+ T cells are generated in mice lacking peripheral lymphoid organs.
figure 1

(a) Chimeric mice were infected with influenza and cells from the lung were analyzed on the indicated days. Representative plots show CD62L expression and H-2DbNP366–374 tetramer binding on CD8+ cells. Numbers in each plot refer to the mean percentage and standard deviation of tetramer-binding cells in the CD8+ population. (b) The combined number of CD8+CD62LloH-2DbNP366–374 and CD8+CD62LloH-2DbPA224–233 tetramer-binding T cells was determined by flow cytometry on days 7, 9, 12, 15 and 20 after infection in the lungs, mediastinal lymph nodes, spleens and NALT. Data were obtained from 4–5 mice per group per time point. Cells from NALT were pooled before analysis. The data shown are representative of four independent experiments.

Full size image

The rapid appearance of CD8+ cells in the lungs of WT and LTCh mice could easily be explained by migration of effector CD8+ cells primed in the draining lymph nodes, because the number of influenza-specific CD8+ cells had already peaked in the lymph nodes by day 7, and subsequently expanded in the spleen (Fig. 1b). In contrast, the robust CD8+ response in the lungs of SLP mice was more difficult to explain because these mice lacked lymph nodes and spleen. Even in nasal-associated lymphoid tissue (NALT), the only lymphoid tissue constitutively present in SLP mice24, the number of NP-specific CD8+ cells was at background levels on day 7 (about 50 cells per mouse) (Fig. 1b). To account for the number of influenza-specific CD8+ cells observed in the lungs of SLP mice at this time, the influenza-specific cells in NALT would have to perform roughly ten cell divisions, making the NALT an unlikely source of the cells in the lung. It also seemed unlikely that the CD8+ cells in the lungs of SLP mice were primed in the bone marrow, as no influenza-specific T cells were detected at this site in any group until day 15 (data not shown). Rapid influenza-specific CD8+ T cell responses were also observed in the lungs of splenectomized Rorc−/− mice (Supplementary Fig 2 online), which lack lymph nodes owing to a developmental defect unrelated to LTα expression25. Thus, conventional lymphoid organs are not required to generate relatively rapid and robust CD8+ T-cell responses to influenza. In addition, although LTα is not required for CD8+ T-cell responses when lymph nodes are present (LTCh mice), LTα is very important for CD8+ T-cell responses when lymphoid organs are absent (LTKO and LTKO-S mice).

Next, we tested the ability of CD8+ cells from the various chimeric mice to produce IFN-γ and TNFα and to kill peptide-loaded targets. IFN-γ- and TNFα-producing cells were observed on day 10 in the lungs of WT, LTCh and SLP mice (Fig. 2a, c). These cells were also able to kill peptide-pulsed target cells (Fig. 2d). In contrast, the number of influenza-specific, cytokine-producing effector CD8+ T cells in the day-10 infected LTKO mice was about 2% of WT levels, and the cells had no detectable CTL activity. By day 14, however, substantial numbers of cytokine-producing CD8+ effector T cells were found in the lungs of all chimeric mice (Fig. 2b, c). Similarly, CTL activity towards both NP- and PA-pulsed targets was observed in cells isolated from the lungs of all groups of chimeric mice (Fig. 2d). These data show that neither LTα nor peripheral lymphoid organs are required for the generation of influenza-specific CD8+ effector T cells.

Figure 2: Influenza-specific CD8+ effector T cells accumulate in the lungs of mice lacking peripheral lymphoid organs.
figure 2

(a,b) Chimeric mice were infected with influenza and cells were pooled from the lungs of five mice per group on days 10 (a) and 14 (b). Cells were stimulated with NP366–374 (top row), PA224–233 (middle row) or no peptide (bottom row) for 5 h and intracellular cytokine expression analyzed by flow cytometry. (c) Total numbers of NP-specific and PA-specific, TNFα–IFN-γ-producing CD8+ T cells in the lungs on day 10 or day 14 after infection were determined by intracellular cytokine staining. (d) Chimeric mice were infected with influenza, and cells from the lungs were pooled from groups of 5–8 mice at the indicated times after infection. Cells were cultured with 51Cr-labeled, peptide-pulsed EL4 targets for 5 h at the indicated effector-to-target ratios, and the specific lysis calculated. The data shown are representative of two independent experiments.

Full size image

Next, we determined the kinetics of influenza-specific IgM and IgG production in chimeric animals. The titers of influenza-specific IgM rapidly increased in WT mice, peaked between days 9 and 15 and then declined (Fig. 3a). The IgM response was delayed in the remaining chimeras, and only reached WT levels in those mice with spleens (Fig. 3a). In addition, the titers of influenza-specific IgG climbed rapidly in WT, LTCh and SLP mice starting on day 9 (Fig. 3a). In contrast, the appearance of IgG was delayed until day 12 in LTKO mice, and was further delayed until day 15 in LTKO-S mice (Fig. 3a). Influenza-specific IgM and IgG production with similar kinetics was also observed in splenectomized Rorc−/− mice that had been infected with influenza (Supplementary Fig. 3 online). Thus, isotype-switched, influenza-specific antibodies could be generated in mice that lacked conventional lymphoid organs.

Figure 3: Influenza-specific IgG and germinal-center B cells are generated in mice lacking peripheral lymphoid organs.
figure 3

(a) Mice were infected with influenza and serum obtained at days 7, 9, 12, 15 and 20 after infection. The titers of influenza-specific IgM and IgG in serum of individual mice were determined by influenza-specific ELISA. Data are representative of four independent experiments. (b) Chimeric mice were infected with influenza and cells from the spleens, NALT, MLNs and lungs of chimeric mice were stained with anti-CD19, anti-FAS and anti-GL7. The plots shown were gated on CD19+ cells, and the FAShiGL7hi germinal-center B cells are circled. Numbers in each plot refer to the mean total number and standard deviation (in thousands) of germinal-center B cells per organ.

Full size image

To determine where B-cell responses were occurring, we looked for FAS+GL7+ germinal-center B cells (GCBs). GCBs were observed in WT spleens, but not in LTCh and LTKO spleens (Fig. 3b), consistent with previous data26. Similarly, GCBs were present in the NALT of WT and SLP mice, but were absent from the NALT of LTCh, LTKO and LTKO-S mice (Fig. 3b). In contrast, GCBs were found in the draining lymph nodes of both WT and LTCh mice, consistent with the ability of lymph nodes to support germinal-center formation in the absence of LTα27. Interestingly, GCBs were also found in the lungs of all groups of influenza-infected chimeric mice. Similar results were observed using FAS and PNA as markers for GCBs (Supplementary Fig. 4 online). These data suggest that germinal centers can be formed in the lungs of animals lacking conventional lymphoid organs and also that germinal centers are formed in the lungs independently of LTα after influenza infection.

Formation of iBALT after influenza infection

Because germinal centers are typically found in secondary lymphoid tissues, we reasoned that the lungs of SLP mice may contain iBALT. In fact, upon examination of serial sections, we found that the lungs of influenza infected SLP mice contained iBALT, which primarily consisted of B-cell follicles centered around CD21-expressing follicular dendritic cells (FDCs; Fig. 4a). Some of these B cells were organized in well-defined germinal centers (Fig. 4b). The interfollicular regions contained CD8+ (Fig. 4c) and CD4+ (Fig. 4d) T cells as well as CD11c+ dendritic cells (DCs; data not shown). Although CD8+ T cells were primarily confined to the interfollicular regions (Fig. 4c), CD4+ T cells were found in the interfollicular regions and throughout the B-cell follicle and germinal center (Fig. 4d). To test whether iBALT was competent to support T and B-cell proliferation in situ, we administered bromodeoxyuridine (BrdU) 1 h before the mice were killed, and probed sections with anti-BrdU. Proliferating CD8+ T cells were found in the interfollicular T-cell areas (Fig. 4e), whereas proliferating CD4+ T cells were observed in the interfollicular area (Fig. 4f) and in the germinal center in close proximity to BrdU+ B cells (Fig. 4g). Thus, iBALT contains organized B- and T-cell areas and supports lymphocyte proliferation.

Figure 4: iBALT is formed in influenza-infected SLP mice and supports T- and B-cell proliferation.
figure 4

Influenza-infected SLP mice were pulsed with BrdU 1 h before being killed on day 7. Frozen serial sections of the lung were stained with various antibodies and analyzed by fluorescence microscopy. (a) B cells and FDCs were identified with anti-B220 (green) and anti-CD21 (red). (b) Germinal centers were identified with anti-GL7 (green) and DAPI (blue). The germinal center is outlined with a dotted circle that appears in all the serial sections. (c) Proliferating CD8+ T cells and B cells were identified with anti-CD8 (green), anti-B220 (blue) and anti-BrdU (red). (d) Proliferating CD4+ T cells and B cells were identified with anti-CD4 (green), anti-B220 (blue) and anti-BrdU (red). (e) A higher magnification of the boxed area shown in c. Yellow arrows indicate CD8+ cells that incorporated BrdU within 1 h. (f) A higher magnification of the upper boxed area shown in d. Yellow arrows indicate CD4+ cells that incorporated BrdU. (g) A higher magnification of the germinal center shown in d. Yellow arrows indicate CD4+ cells that incorporated BrdU. The data are representative of sections from at least five independent mice. Images a–d were originally obtained at 20× magnification and those in e–g at 40× original magnification.

Full size image

Next, we investigated whether mRNAs for CXCL13 and CCL21 were expressed in the lungs of influenza-infected mice. As expected28, CXCL13 and CCL21 were expressed constitutively in the spleens of C57BL/6 (B6) mice, but were much less abundant in the spleens of Lta−/−, Tnf-Lta−/− and μMT mice (Fig. 5a). In contrast, CXCL13 was undetectable and the amount of CCL21 was low in the lungs of uninfected B6 mice (Fig. 5b). However, after influenza infection, CXCL13 was induced in the lungs of B6 mice by 4 d after infection and peaked between 6 and 10 d (Fig. 5b). In contrast, CCL21 was already expressed in the lungs of B6 mice before infection, and was induced to higher levels at day 10 after infection and beyond (Fig. 5b). Similar patterns of CXCL13 and CCL21 expression were observed in the lungs of Lta−/−, Tnf-Lta−/− and μMT mice (Fig. 5b). These data were confirmed using real-time PCR (Supplementary Fig. 5 online). We also examined whether the CCL21-Ser locus, which is predominantly active in secondary lymphoid organs, or the CCL21-Leu locus, which is predominantly active in peripheral tissues29,30, was responsible for CCL21 expression in the lung. Both forms of CCL21 were observed in the lungs of B6 and Lta−/− mice at 0, 10 and 30 d after infection; however, CCL21-Ser was predominant at later times (Supplementary Fig. 5 online). Thus, it seems that CXCL13 and CCL21 are induced in the lungs of influenza-infected mice independently of LTα and TNFα and that the CCL21-Ser form is preferentially induced.

Figure 5: CXCL13 and CCL21 are induced in the lungs of influenza-infected mice and are found in areas of iBALT.
figure 5

(a,b) RNA was prepared from the spleens (a) and lungs (b) of influenza-infected B6, Lta−/−, Tnf-Lta−/− and μMT mice at various times after infection and northern blot analysis performed with cDNAs probes for CXCL13, CCL21 and CHO-B (loading control). Each time point includes RNA from two individual mice. (c,d) Sections from the lungs of SLP mice (c) and LTKO mice (d) were probed with anti-CXCL13, anti-CCL21, anti-PNAd, anti-B220 or anti-CD3 as indicated. No staining was observed using secondary antibodies alone or in combination with nonspecific goat IgG. Images in c were originally obtained at 10× magnification and those in d at 20× magnification. (e,f) Paraffin-embedded sections from the lungs of naive or day-10 influenza-infected B6 (e) and Lta−/− (f) mice were stained with H&E (top and lower left) or with anti-B220 (lower right).

Full size image

We also determined whether CXCL13 and CCL21 co-localize with iBALT in SLP mice. The majority of CXCL13 was localized in follicular areas that contained B cells and FDCs (Fig. 5c). In contrast, CCL21 was observed on vascular endothelia, as well as on reticular cells around the edge of the follicles and in the interfollicular areas (Fig. 5c). In addition, PNAd was expressed on vascular endothelium in the lymphoid areas of the lung in SLP mice (Fig. 5c). CXCL13 and CCL21 were also observed in the lymphoid areas of Lta−/− mice after infection with influenza. CXCL13 was expressed in a reticular pattern in areas containing B cells (Fig. 5d), and CCL21 expression was observed on vascular endothelial cells in this same region (Fig. 5d). However, iBALT in mice lacking LTα was not as well organized as it is in mice expressing LTα, and consisted primarily of B cells with few T cells (Fig. 5d). Furthermore, the B cell areas in the lungs of Lta−/− mice were devoid of FDCs (data not shown) and the T cells were intermixed rather than being separated into interfollicular areas (Fig. 5d). Finally, PNAd expression was never observed in the absence of LTα (data not shown). Thus, CXCL13 and CCL21 are observed in iBALT independently of LTα expression; however, iBALT is not structurally equivalent in wild-type and Lta−/− mice.

We also examined the lungs of uninfected B6 and Lta−/− mice for iBALT. As expected, no iBALT was observed in naive B6 mice (Fig. 5e). Similarly, and contrary to previous reports6,21, the lungs of naive Lta−/− mice did not contain significant areas of iBALT (Fig. 5f); however, iBALT was easily detected in the lungs of both B6 and Lta−/− mice within 10 d after infection (Fig. 5e,f). Furthermore, the lymphoid areas in the infected lungs of both groups of mice were organized into structures that contained prominent B-cell areas (Fig. 5e,f). Areas of iBALT peaked in size between 1 and 2 weeks after infection in B6 mice and diminished thereafter. In contrast, areas of iBALT in Lta−/− mice were larger and were maintained for longer periods (data not shown), despite the lack of FDCs and PNAd expression. Regardless of the kinetics, iBALT seems to be formed as a result of viral infection in both wild-type and Lta−/− mice.

Viral clearance and survival in the absence of lymphoid organs

Next, we assayed viral clearance and survival in the chimeric mice. We found that the viral burden in the lung was similarly high in all groups of mice on day 6, but by day 9 the viral burden in WT, LTCh and SLP mice had declined significantly, whereas the viral burden in the lungs of LTKO mice remained high (Fig. 6a). Similarly, on day 10, virus was nearly eliminated from WT and LTCh mice, was substantially reduced in SLP mice and remained high in LTKO mice. However, by day 14, virus was cleared from the lungs of all groups of mice, including LTKO mice (Fig. 6a). Thus, viral clearance was consistent with the kinetics of T- and B-cell responses (Figs. 1,2,3).

Figure 6: Viral clearance is slightly delayed in SLP mice, but survival is enhanced.
figure 6

(a) Chimeric mice were infected with influenza, and viral titers in the lungs were determined at the indicated times. The dashed line indicates the limit of detection of the assay. The data shown are representative of three independent experiments. (b) Groups of ten wild-type (WT), LTCh, SLP and LTKO mice were infected with influenza and their survival and weight were monitored daily. The asterisk indicates that two of the ten LTKO mice were humanely killed at that point. The data shown are representative of two independent experiments. (c,d) Groups of 6–8 WT or SLP mice were infected with the indicated doses of virus and their survival (c) and weight (d) were monitored daily. The data are representative of two independent experiments. (e) WT, LTKO and LTKO-S mice were infected with 100 EIU of influenza and their weight monitored daily. (f) B6, LTKO and Tnf-Lta−/− mice were infected with 100 EIU influenza and their weight monitored daily.

Full size image

Next, we followed the weight loss and survival of the chimeric mice after infection. WT mice rapidly lost weight between days 4 and 7 after infection, and recovered between days 9 and 13 (Fig. 6b). The LTCh mice showed weight loss with kinetics similar to those of WT mice. As observed previously13, the onset of weight loss was delayed in LTKO mice, but was ultimately more severe (Fig. 6b). In fact, two of the ten mice in the LTKO group succumbed to disease on day 12 (Fig. 6b, asterisk). In marked contrast, SLP mice survived high doses of influenza (6,000 and 1,200 egg infectious units, EIU) significantly better than WT mice (Fig. 6c). Furthermore, when infected with lower doses of virus (240 and 50 EIU), the SLP mice showed substantially less morbidity throughout the experiment (Fig. 6b and d). Thus, the lack of peripheral lymphoid organs does not increase morbidity and mortality after influenza infection but, instead, substantially increases the ability of these mice to survive infection with normally lethal doses of virus.

We reasoned that the reduced weight loss in the SLP mice might be the result of a smaller immune response and less systemic cytokine production. To test this, we first determined whether the smaller immune response in LTKO-S mice relative to that in LTKO mice (Fig. 1) resulted in reduced weight loss. The weight loss of LTKO-S mice was less than that in LTKO mice (Fig. 6e). Thus, the spleen of LTKO mice seems to facilitate the expansion of the immune response and to contribute to the pathology. As TNFα is known to contribute to systemic inflammation and weight loss in a number of infections, including influenza31, we also compared the weight loss of WT, Lta−/− and Tnf-Lta−/− mice. As predicted, Tnf-Lta−/− mice lost less weight than Lta−/− mice (Fig. 6f), even though both groups of mice lack lymph nodes and have disorganized spleens. Thus, TNFα is one inflammatory cytokine that is involved in weight loss after influenza infection.

Discussion

In typical respiratory immune responses, antigen-specific lymphocytes first appear in the lymph nodes that drain the respiratory tract and subsequently appear in the lungs5, suggesting that lymphocytes are primed in the draining lymph node and then migrate as effector cells to the infected lungs. However, our data demonstrate that peripheral lymphoid organs are not necessary for primary immune responses to influenza. Instead, we find that influenza infection rapidly induces the formation of iBALT, which is capable of priming influenza-specific T and B cells that function to clear virus and facilitate host survival without any help from conventional lymphoid organs. These data are consistent with published reports showing that T cells can be primed directly in the lung32. Therefore, it is somewhat surprising that iBALT is often overlooked as an inductive site for immune responses to respiratory infections5, particularly as GCB cells and iBALT are easily detected in the lungs of normal mice (Figs. 3 and 4). The most probable explanation for this oversight is that immune responses initiated in iBALT are delayed relative to the immune response in the draining lymph nodes, owing to the time it takes to form iBALT. Therefore, immune responses initiated in iBALT of normal mice are obscured by the rapid and robust responses initiated in conventional lymphoid organs. Notably, other studies have shown that iBALT is particularly prominent after multiple respiratory infections33. Thus, we predict that iBALT induced by one infection will alter the kinetics and location of subsequent immune responses to respiratory antigens. For example, the ability of previous influenza infection to prevent tolerance induction by the intranasal administration of ovalbumin may be the result, in part, of the presence of iBALT34. In addition, the formation of iBALT in chronic lung diseases, such as tuberculosis or COPD, may be a component of the pathology35.

Although the iBALT that develops after influenza infection resembles constitutive BALT, it is distinct in that it is not induced by an antigen-independent developmental pathway18,22 but by inflammatory insult. In addition, the size of iBALT varies widely, from small clusters of B cells, T cells and DCs to the well-developed follicular areas shown in Figures 4 and 5. Furthermore, unlike constitutive BALT, which is found in the upper bronchi of some species, iBALT is found in perivascular, peribronchial and even interstitial areas in the lower airways of the lung, and does not always occur under a dome epithelium. However, iBALT does contain B-cell follicles centered around networks of FDCs, separated by interfollicular regions containing DCs and T cells. High endothelial venules (HEVs) expressing PNAd are also present in iBALT, allowing for the recruitment of naive T cells36. Thus, iBALT shares functional and structural characteristics with classically defined BALT18,22. However, whereas the development of secondary lymphoid organs — such as lymph nodes and Peyer's patches — and the expression of CXCL13 and CCL21 in the spleen are typically dependent on LTα7,28,37, the formation of iBALT and the expression of CXCL13 and CCL21 in the lung occur independently of LTα. Similarly, germinal-center formation in the lung, but not the spleen or NALT, is LTα independent. Thus, the development and organization of iBALT is governed by mechanisms that are distinct from those that regulate the development and organization of conventional secondary lymphoid organs.

Notably, iBALT is often observed in the lungs of Lta−/− mice that have not been intentionally infected38. A number of possibilities have been proposed to explain this phenomenon. For example, the lack of IgE in Lta−/− mice may result in respiratory inflammation and lymphocyte accumulation in the lung39. In addition, the inability of Lta−/− mice to express AIRE in the thymus may cause autoimmune inflammation in the lung and other tissues21. Alternatively, the constitutive expression of CCL21 in the lungs may result in the enhanced recruitment of lymphocytes owing to the loss of CCL21 in conventional lymphoid organs40. Finally, it is possible that respiratory infection with opportunistic pathogens may result in increased lymphocyte accumulation in the lungs of Lta−/− mice. To prevent this, we re-derived our Lta−/− mice by embryo transfer and maintained them on antibiotics to prevent infection with Pneumocystis carinii and other respiratory pathogens. Possibly because of these measures, we did not observe significant infiltration in the lungs of Lta−/− mice under the age of 4–5 months (Fig. 5). However, as these mice aged, we did begin to see accumulations of lymphocytes in the lungs (particularly B cells). This may be the result of delayed colonization of the lung by opportunistic pathogens or to other defects in Lta−/− mice. Importantly, we eliminated the possibility that the relatively normal respiratory immune responses observed in SLP mice resulted from pre-existing iBALT in Lta−/− recipient mice by confirming our results with splenectomized Rorc−/− mice (Supplementary Figs. 2 and 3 online). As Rorc−/− mice lack lymph nodes and Peyer's patches25, but express LTα normally and do not spontaneously develop iBALT, they provided an independent method to evaluate the requirement for spleen, lymph nodes and Peyer's patches in respiratory immune responses. These experiments showed that T- and B-cell responses were generated with near-normal to normal kinetics in splenectomized Rorc−/− mice, and suggest that any pre-existing lymphoid accumulations in the lungs of Lta−/− recipient mice did not measurably influence our results. Thus, the formation of iBALT occurs as a normal component of the immune response to respiratory infections and may also occur as a consequence of pathological inflammation in the lung.

One surprising outcome of these experiments was that iBALT initiated immune responses that were not only protective, but resulted in less morbidity and mortality than immune responses initiated in conventional lymphoid tissues. We suspect that the reduced morbidity and mortality in influenza-infected SLP mice is simply a function of the reduced size of the immune response at early times after infection in SLP mice. Although the influenza-specific CD8+ T-cell response in SLP mice was about 20% of WT levels on days 9 and 10, this was sufficient to substantially lower the viral load in SLP mice (Fig. 6a). By day 14, when large numbers of T cells were present in the lungs of SLP mice, virus was already eliminated, which limited additional T-cell activation. Thus, the T-cell response in SLP mice is limited at both early and late time points of infection. Because activated T cells produce large quantities of TNFα (formerly known as cachexin), a reduced T-cell response in SLP mice should result in less weight loss and morbidity. Therefore, the increased survival and reduced morbidity of SLP mice is probably the result of 'just enough' immunity, which begins to clear virus without the systemic production of pathologic levels of cytokines.

Together, these data alter our perception of how respiratory immune responses occur in normal animals, and overturn the immunological dogma that primary immune responses are generated only in constitutively present secondary lymphoid organs. We have shown that secondary lymphoid organs, such as spleen and lymph nodes, accelerate and amplify immune responses to antigens or pathogens in peripheral tissues. Yet these lymphoid organs are not absolutely required for immunity and may, under certain circumstances, facilitate the development of overly robust and perhaps damaging immune responses. However, smaller and possibly less pathologic immune responses can be initiated locally within the lung, but only after a delay, during which lymphocytes are recruited and organized at the site of infection. Although our data demonstrate that peripheral lymphoid organs are not absolutely required for the initiation of immune responses, these data do not contradict the idea that lymphoid tissues facilitate immune responses and provide a significant benefit to the host, particularly in response to highly virulent pathogens. Instead, our data suggest that the lung, and perhaps other non-lymphoid organs, have the ability to organize local immune responses; furthermore, they suggest that the immune response generated in these local sites can be protective, yet functionally different, to those initiated in conventional lymphoid organs. If so, then this argues that vaccines that target immune responses to local tissues rather than to conventional lymphoid organs may provide protection with reduced potential for immune-mediated damage.

Methods

Mice.

C57BL/6, μMT and C57BL/6.129Ltatm1Dch (Lta−/− mice) were obtained from Jackson Laboratory. Rorc−/− mice were obtained from D. Littman and Tnf-Lta−/− mice were obtained from J. Sedgewick. All mice were bred and maintained at the Trudeau Institute. Recipient mice were irradiated (950 rad), reconstituted with 1 × 107 bone marrow cells and allowed to recover for 6 weeks before infection. Mice were maintained on 1.25 mg/ml sulfamethoxazole and 0.25 mg/ml trimethoprim. All procedures were approved by the Trudeau Institute Institutional Animal Care and Use Committee and were conducted according to the principles outlined by the National Research Council.

Influenza virus.

Mice were infected intranasally with 100 EIU of influenza A/PR8/34 in 100 μl. Viral titers were quantified in embryonated chicken eggs and the viral end-point titer was defined as the highest dilution in which two or more eggs scored positive in the hemagglutination assay.

Flow cytometry.

Tissues were disrupted by passage through wire mesh, and live cells were obtained by density gradient centrifugation using Lympholyte-Poly (Cedarlane). Cells were incubated with 10 μg/ml FcBlock and then fluorochrome-conjugated antibodies or MHC class I tetramers were added. The H-2Db tetramers containing NP366–374 or PA224–233 were generated by the Trudeau Institute Molecular Biology Core Facility

Intracellular cytokine staining.

Cells were cultured for 5 h at 5 × 106/ml in media containing 10 μg/ml brefeldin A, 10 U/ml IL-2 and 1 μg/ml NP366–374 or PA224–233. Simulated cells were then washed, blocked with FcBlock and probed with anti-CD8 before fixation in 4% paraformaldehyde. Fixed cells were washed with 0.1% Triton X-100, 3% fetal calf serum in PBS and probed with fluorochrome-labeled anti–IFN-γ and anti-TNFα in the same buffer.

CTL assays.

EL4 cells were cultured with no peptide, influenza NP366–374 peptide or influenza PA224–233 peptide at 1 μg/ml and were labeled overnight with 51Cr. After labeling, the EL4 targets were washed and cultured for 5 h at 104 cells per well with effector cells at effector-to-target ratios of between 0.1:1 and 100:1 (performed in triplicate). Radioactivity in supernatants was counted in a gamma counter. Spontaneous 51Cr release was determined from labeled EL4 cells in the absence of effectors, and the maximum possible release was determined from EL4 cells cultured in 1% Triton X-100.

ELISA.

Plates were coated with purified influenza proteins at 1 μg/ml. Serum samples were diluted in threefold serial dilutions in PBS with 10 μg/ml BSA and 0.1% Tween-20 before incubation on coated plates. Bound antibody was detected with horseradish-peroxidase-conjugated goat anti–mouse IgM or goat anti–mouse IgG (Southern Biotechnology Associates).

Northern blots, PCR and sequencing.

RNA was extracted from the spleens and lungs of influenza-infected mice using the RNeasy kit (Qiagen). Total RNA (15 μg per lane) was loaded into the wells of a denaturing 1.2% agarose gel and resolved by electrophoresis. The RNA was blotted onto a nylon membrane, cross-linked to the membrane by ultraviolet irradiation and hybridized with 32P-labeled probes of the coding sequences of CXCL13, CCL21 and CHO-B. Sequencing was performed using the Prism Dye Terminator Cycle Sequencing Ready Reaction Kit from ABI.

Taqman PCR.

cDNA was generated from DNase-digested RNA by reverse transcription using Oligo dT and Superscript II. CDNA (50 ng) was amplified using primers and probes from Applied Biosystems (CXCL13) and primers and probes described previously for CCL21 (ref. 41). Amplifications were performed using Taqman Universal PCR master mix according to the Applied Biosystems protocol.

Immunohistochemistry.

Sections of paraffin-embedded lungs were processed from xylene through graded alcohol to PBS. Endogenous peroxidase was blocked with 0.1 M NaN3/1% H2O2, incubated with biotinylated anti-B220 (Pharmingen) and developed with ABC and DAB kits from Vector. Slides not processed for immunohistochemistry were stained with Gill's H&E (hematoxylin and eosin; Fisher).

Immunofluorescence and BrdU administration.

BrdU (1 mg) in PBS was administered intravenously 1 h before mice were killed. Biotinylated anti-B220, anti-CD11c, anti-CD21/CD35, anti-CD4 and anti-CD8 were purchased from PharMingen. Biotinylated peanut agglutinin (PNA) was obtained from Vector Laboratories. Biotinylated antibodies were visualized with streptavidin conjugated to Alexa-594, Alexa-350 or Alexa-488 (Molecular Probes). Unlabeled anti-GL7 (PharMingen) was visualized with anti-IgM–FITC (PharMingen). Unlabeled anti-PNAd (PharMingen) was visualized through biotinylated goat anti–rat IgM (Jackson ImmunoResearch) followed by streptavidin–Alexa-594. Anti-CXCL13 and anti-CCL21 antibodies (R&D Systems) were visualized with donkey anti–goat IgG conjugated to Alexa 594. Anti-BrdU conjugated to Alexa-594 was purchased from Molecular Probes. To expose BrdU epitopes, slides were incubated in 4 M HCl for 30 min, followed by 0.1 M Na2B4O7 at pH 8.5 for 10 min. Sections were mounted either with Polymount (Polysciences) or with SlowFade Light antifade kit containing DAPI (Molecular Probes). Slides were viewed with a Zeiss Axioplan 2 microscope using a 480/30 bandpass filter set (Alexa 488 and FITC), a 560/40 bandpass filter set (Alexa 594) and a 330/80 filter set (Alexa-350 and DAPI). Images were recorded with a Zeiss AxioCam digital camera using the Zeiss proprietary software, Axiovision 3.0.6.0. Images were manipulated in Adobe Photoshop 7.0.

Note: Supplementary information is available on the Nature Medicine website.