Lipocalin 2 modulates dendritic cell activity and shapes immunity to influenza in a microbiome dependent manner

Lipocalin 2 (LCN2) is a secreted glycoprotein with roles in multiple biological processes. It contributes to host defense by interference with bacterial iron uptake and exerts immunomodulatory functions in various diseases. Here, we aimed to characterize the function of LCN2 in lung macrophages and dendritic cells (DCs) using Lcn2-/- mice. Transcriptome analysis revealed strong LCN2-related effects in CD103+ DCs during homeostasis, with differential regulation of antigen processing and presentation and antiviral immunity pathways. We next validated the relevance of LCN2 in a mouse model of influenza infection, wherein LCN2 protected from excessive weight loss and improved survival. LCN2-deficiency was associated with enlarged mediastinal lymph nodes and increased lung T cell numbers, indicating a dysregulated immune response to influenza infection. Depletion of CD8+ T cells equalized weight loss between WT and Lcn2-/- mice, proving that LCN2 protects from excessive disease morbidity by dampening CD8+ T cell responses. In vivo T cell chimerism and in vitro T cell proliferation assays indicated that improved antigen processing by CD103+ DCs, rather than T cell intrinsic effects of LCN2, contribute to the exacerbated T cell response. Considering the antibacterial potential of LCN2 and that commensal microbes can modulate antiviral immune responses, we speculated that LCN2 might cause the observed influenza phenotype via the microbiome. Comparing the lung and gut microbiome of WT and Lcn2-/- mice by 16S rRNA gene sequencing, we observed profound effects of LCN2 on gut microbial composition. Interestingly, antibiotic treatment or co-housing of WT and Lcn2-/- mice prior to influenza infection equalized lung CD8+ T cell counts, suggesting that the LCN2-related effects are mediated by the microbiome. In summary, our results highlight a novel regulatory function of LCN2 in the modulation of antiviral immunity.

Introduction Lipocalin 2 (LCN2) is a 25-kDa secreted glycoprotein involved in a variety of biological processes, including immune responses, iron homeostasis and metabolism [1][2][3]. Acting as scavenger of bacterial siderophores, LCN2 is known as host defense molecule with potent antibacterial activity against siderophore-dependent bacteria, such as Escherichia coli and Klebsiella pneumonia [4,5]. In line with this function, LCN2 is highly expressed in barrier tissues exposed to microorganisms including the lungs [6] and further induced upon Toll-like receptor (TLR) stimulation [4].
Multiple studies have uncovered immunoregulatory effects of LCN2 beyond scavenging bacterial siderophores, and LCN2 has been found to protect from excessive inflammationrelated morbidity in sterile endotoxemia [7], neuroinflammation [8] and non-alcoholic steatohepatitis (NASH) [9]. Intriguingly, all these observations are linked to effects of LCN2 on myeloid cells, such as macrophages or, in the context of NASH, macrophage-neutrophil interplay. Our group could previously show that LCN2 deactivates lung macrophages and worsens disease outcome from pneumonia caused by the siderophore-independent pathogen Streptococcus pneumoniae [10]. These studies highlight effects of LCN2 on myeloid cell function and plasticity, with important consequences in various infectious and non-infectious diseases. Given the expression pattern at mucosal surfaces including the intestine and LCN2's potential to alter the availability of microbial nutrients, it is tempting to speculate that LCN2 might shape the composition of the microbiome. Since the microbiome can strongly influence many aspects of the host immune system [11], this could be a major relay mediating LCN2-related immunomodulatory effects.
While immune responses need to be of sufficient intensity for pathogen clearance, inflammatory responses can cause substantial damage to the host. Indeed, certain pulmonary viral infections, including influenza and SARS-CoV-2 [12,13], are associated with particularly severe degrees of inflammation, which, considering the delicate architecture of lung tissue, require complex and efficient regulatory mechanisms [14]. Immunopathology during viral infections can be caused by both innate and adaptive immune activity, for instance by dysregulated production of proinflammatory cytokines, or exuberant cytolytic activity of CD8 + T cells [15]. Pulmonary dendritic cells (DCs) represent the interface between the innate and adaptive immune systems in the lungs by uptake, processing and presentation of antigens leading to initiation of specific T cell responses. In particular, CD103 + DCs contribute substantially to the cytotoxic T lymphocyte response during influenza infection [16]. Potential effects of LCN2 on DCs could consequently influence adaptive anti-viral immune responses. Here, we aimed to examine the role of LCN2 as modulator of lung immunity in the context of homeostasis and viral infections. Since previous studies [7][8][9][10] suggested an immunomodulatory effect of LCN2 on myeloid cells, we focused our studies on alveolar macrophages and CD103 + dendritic cells.

Lipocalin 2 shapes the transcriptome of pulmonary myeloid cells during homeostasis
To assess if LCN2 exerts any effects on the homeostatic transcriptome of lung-resident myeloid immune cells that could impact acute immune responses [17], we performed RNA sequencing of pulmonary CD103 + DCs and alveolar macrophages (AMs) isolated from wild type (WT) and Lcn2 -/mice ( Fig 1A). Using principal component analysis on all expressed genes, we revealed a clear separation of CD103 + DCs according to genotype (Fig 1B), and a tendency for AM samples (Fig 1C). Differential gene expression analysis identified 56 Table for differentially impacted KEGG pathways in AMs, with antigen processing and presentation (based on 11 differentially expressed genes) affected the most. Remarkably, three pathways related to infections (Legionellosis [9 genes], Measles [11 genes] and Influenza A [11 genes]), were among the 10 most significantly impacted ( Fig 1D). Focusing on antigen processing and presentation and Influenza A, we identified three heat-shock proteins (Hspa8, Hspa1b, Hspa1a), the M alpha chain of the MHC-II complex (H2-DMa) and Tnf (Fig 1E) to be differentially regulated. The genes related to antigen processing and presentation further included the nonclassical MHC class I molecule H2-M3, highlighting that LCN2 affected expression of both, MHC class I and II associated genes. Interestingly, the Influenza A pathway was related to six additional differentially regulated genes, including Il1b and Nlrp3. These results reveal that LCN2 shapes pulmonary CD103 + DC gene expression profiles in homeostasis and suggest potential effects of LCN2 on antigen processing and presentation by DCs in the context of antiviral immune responses.

Lipocalin 2-deficient animals show increased disease morbidity and lung T cell numbers upon influenza infection
Considering the importance of CD103 + DC in cross presenting antigen during influenza infection [16], and the obvious impact LCN2 has on homeostatic CD103 + DC signatures, we tested the functional implications of these findings. Using a mouse model of influenza infection based on intranasal inoculation with PR/8 (a mouse-adapted H1N1 influenza strain [18]) we determined the potential functions of LCN2 in antiviral immunity. LCN2 deficiency was associated with increased body weight loss over nine days after infection (Fig 1F), which led to higher mortality in Lcn2 -/mice ( Fig 1G). Interestingly, this impaired disease phenotype was not associated with increased lung viral loads in Lcn2 -/mice (Figs 1H and S1E). In contrast, LCN2 deficiency was associated with increased bronchoalveolar lavage (BAL) levels of proinflammatory cytokines (i.e. IL12, interferon [IFN]γ, IP-10, IL6, MCP-1, BAFF and IFNα) three days after influenza infection (S1F Fig). Furthermore, Lcn2 -/animals showed higher BAL protein levels at the peak of inflammation (nine days after influenza infection; Fig 1I), indicating increased lung vascular permeability and inflammation [19,20]. We conclude from these results that LCN2 attenuated influenza infection-related inflammation and disease morbidity independent of viral clearance, suggesting a role for LCN2 in limiting virus-related immunopathology.
To characterize the regulatory function of LCN2 on immunological processes during influenza infection, we analyzed the influenza-induced lung T cell response by flow cytometry (S2A Fig). We observed more CD4 + T helper cells (Fig 2A) and CD8 + cytotoxic T cells (Fig 2B) in lungs of Lcn2 -/mice seven, nine and 16 days after infection. Furthermore, LCN2 deficiency was associated with elevated proportions of CD69 + activated pulmonary CD4 + (S2B Fig) and CD8 + T cells ( Fig 2C) and higher numbers of IFNγ-expressing ( Fig 2D) and of antigen-specific CD8 + T cells (Fig 2E). This enhanced adaptive immune response was further reflected by increased influenza-specific serum IgG antibody levels in Lcn2 -/mice ( Fig 2F). Apart from an increase in lung neutrophils (day nine) and B cells (day 16 after infection) in Lcn2 -/mice, we observed no significant differences in the composition of lung immune cells (S2C- S2E Fig). However, the enlarged mediastinal lymph nodes (mLNs), which are considered the primary sites of early T cell proliferation after influenza infection [21], and increased mLN cell numbers nine and 16 days after influenza infection underline the increased adaptive immune response in Lcn2 -/animals ( Fig 2G). We next depleted CD8 + T cells to assess their contribution to the increased morbidity of Lcn2 -/-. This depletion protected Lcn2 -/mice from exaggerated weight loss between day 4 and 7 post infection (Fig 3C-3E), indicating that CD8 + T cells mediate the increased weight loss during this time period. Taken together, these results suggested that LCN2 influenced disease morbidity during influenza infection by regulation of T cell-related adaptive immunity.

LCN2 alters the antigen presentation efficiency of CD103 + DCs, but not intrinsic T cell proliferation
We next aimed to determine whether the increased lung T cell numbers in Lcn2 -/mice were related to an increased intrinsic T cell proliferation potential. To do so, we applied a competitive T cell chimera tracing strategy based on transfer of equivalent numbers of CD45.2 + splenic T cells isolated from green fluorescent protein (GFP) expressing WT mice and GFP -Lcn2 -/-

Fig 1. LCN2 shapes myeloid cell transcriptome during homeostasis and limits disease morbidity during influenza infection. (A)
Experimental layout to assess the influence of LCN2 on the transcriptome of pulmonary myeloid immune cells. CD103 + dendritic cells (DCs) and alveolar macrophages (AMs) were isolated by FACS and processed for RNA sequencing. (B-C) Principal component analysis of CD103 + DCs (B) and AMs (C) derived from lungs of WT and Lcn2 -/mice. (D) Top ten KEGG pathways with lowest SPIA (Signaling Pathway Impact Analysis) p values in CD103 + DCs. Circle sizes indicate the number of differentially expressed genes (DEGs) associated with the respective pathway. (E) Heatmap of DEGs in antigen processing and presentation or Influenza A KEGG pathways. Raw counts are rlog transformed followed by z-score scaling. (F) Comparison of relative body weight (normalized to baseline) between WT and Lcn2 -/mice after infection with 10 PFU PR/8, shown as group means +/-SEM. Pooled data from three independent experiments are shown, n = 17 mice per genotype. (G) Survival of WT and Lcn2 -/mice after infection with 15 PFU PR/8. n = 7 (WT) and 6 (Lcn2 -/-). Data are representative of two independent experiments. (H) Viral load, assessed by plaque assay on MDCK cells, in lung homogenates of WT or Lcn2 -/mice after infection with 10 PFU PR/8. n = 4-10 mice per genotype and timepoint. (I) Total protein concentration in bronchoalveolar lavage (BAL) 9 days after infection with 10 PFU PR/8. n = 4 (WT) and 5 (Lcn2 -/-). Data are representative of two independent experiments. (E) Columns represent samples collected from individual mice. Bar diagrams (H, I) show group means +/-SEM. Significance was assessed using Student's t test for (F), (H) and (I), or asymptotic two-sample logrank test for (G). � p < 0.05, �� p < 0.01, ��� p < 0.001.
Another reason for the increased T cell numbers in Lcn2 -/mice could be an altered antigen-presentation process. The proliferation of CD8 + T cells during influenza infection is driven by cross-presentation of viral antigens by lung-resident DCs, which migrate from the infected lung to the mLNs [16]. To compare the potential of WT and Lcn2 -/-DCs to induce CD8 + T cell proliferation, we used an antigen presentation assay utilizing a genetically modified PR/8 strain expressing an ovalbumin peptide (PR/8-Ova) [22] in combination with OT-I CD8 + T cells (derived from mice which possess exclusively ovalbumin-specific CD8 + T cells [23]). Sixty hours after infection with PR/8-Ova, CD103 + DCs and CD8 + DCs were sorted from mediastinal lymph nodes and then incubated with OT-I T cells (Fig 4A). Using this approach, we observed increased CD8 + T cell proliferation upon incubation with Lcn2 -/-CD103 + DCs (Fig 4B and 4C). This LCN2-related effect was limited to CD103 + DCs, which are considered the dominant antigen cross-presenting DCs in the lung [16,24,25], and was not observed for CD8a + DCs (S3A and S3B Fig). To further dissect the effect of LCN2 on DCs in antiviral immunity, we compared the transcriptome of WT and Lcn2 -/-CD103 + DCs collected from mLNs 60 hours after influenza infection and observed separation of samples according to genotype, suggesting robust LCN2 effects on the CD103 + DC transcriptome during infection ( Fig 4D). Among 106 upregulated and 63 downregulated genes in Lcn2 -/-CD103 + DCs (Fig 4E), only very few (one and four genes, respectively) showed similar regulation in cells isolated from homeostatic lungs (S3D Fig), indicating specific differentially regulated gene sets during homeostasis and infection. The most significantly impacted KEGG pathway in CD103 + DCs isolated from infected Lcn2 -/mice was intestinal immune network for IgA production ( Fig  4F). Differentially expressed genes mapping to this pathway included Ccr9, Icosl and Cd40, the latter two of which are involved in interactions between DCs and T cells [26,27], and H2−Eb1 and H2−Aa which encode subunits of the MHC II complex ( Fig 4G). These two genes were also associated with the KEGG pathway antigen processing and presentation, which was also significantly affected (Fig 4F and 4G). Further genes linked to this pathway were H2-M2, which encodes an MHC class Ib antigen, and Ciita, a transcriptional coactivator that regulates MHC class I and II genes [28]. Taken together, these results show that LCN2 negatively regulates antigen presentation of CD103 + DCs to CD8 + T cells during influenza infection, possibly by influencing the expression of specific functionally important gene sets.

The microbiome is shaped by LCN2 and influences the immune response to influenza infection
Having shown that LCN2 protects from influenza-associated disease severity by modulating DC-driven T cell activation during infection, we wanted to investigate potential mechanisms that could link those observations. To this end, we generated DCs from bone marrow (BMDCs) of WT and Lcn2 -/mice and investigated their antigen presentation potential in vitro. Interestingly, ovalbumin-pulsed WT and LCN2-deficient BMDCs were similarly proficient inducers of OT-I CD8 + T cell proliferation (S3D Fig). Additionally, pre-incubation of BMDC with recombinant LCN2 during antigen pulsing did not alter the antigen presentation capacity (S3E Fig). In line with these results, we considered direct effects of LCN2 on DC antigen presentation rather unlikely, and decided to explore potential indirect mechanisms. Since it is known that LCN2 modulates the availability of microbial nutrients [4] and that the microbiome can influence antiviral immunity [29,30], we tested if LCN2 availability might alter the microbiome. We therefore compared commensal intestinal and pulmonary bacterial communities in WT and Lcn2 -/mice. While microbiome analysis of F2 generation offspring of heterozygous breeding pairs by 16S rRNA gene sequencing ( Fig 5A) showed that the BAL microbial profile of WT and Lcn2 -/animals was similar, we observed highly different LCN2-dependent compositions of ileum luminal and mucosal, cecal and stool microbial communities (Fig 5B-5F). Shannon diversity was comparable between genotypes for all sampled sites, but we observed trends towards decreased amplicon sequencing variant (ASV) richness in cecal and stool samples (S4A and S4B Fig Table). As such, an ASV identified as segmented filamentous bacteria (SFB), which accounted for up to 90% of all bacterial sequencing reads in the ileal mucosa of WT mice, was absent in Lcn2 -/mice (S4D Fig). In summary, our data show that LCN2 shapes the intestinal microbiome in a site-specific manner.
We next aimed to determine whether the LCN2-related microbiome differences play a role in the dysregulated antiviral immune response in Lcn2 -/mice. To deplete the intestinal microbiome, we treated WT and Lcn2 -/mice with antibiotics (vancomycin, metronidazole, ampicillin and gentamicin) in glucose-supplemented drinking water for 4 weeks prior to influenza infection ( Fig 5G). While mock-treated Lcn2 -/mice (receiving glucose-supplemented water) exhibited increased pulmonary T cell and mediastinal lymph node cell counts as previously observed, microbiome depletion equalized these differences to WT mice (Fig 5H-5J). In addition, antibiotic treatment abolished differences in the antigen presentation potential of Lcn2 -/and WT CD103 + DCs (S5A and S5B Fig). To further study the role of the microbiome, we cohoused WT and Lcn2 -/mice for 4 weeks prior to PR/8 infection ( Fig 5K) to partially hybridize microbial profiles by passive transfer of microbiota [31,32]. Confirming the effectivity of this approach, co-housing led to similar levels of the bacterial phylum Bacteroidetes in WT and Lcn2 -/animals, while significantly different in separately housed mice (S4C Fig). Finally, cohousing decreased lung CD8 + T cell differences between influenza-infected WT and Lcn2 -/mice ( Fig 5L and 5M), while mLN size was unaffected ( Fig 5N). Taken together, these results indicate that LCN2 modulated the antiviral immune response and prevented exaggerated CD8 + T cell immunity during influenza infection in a microbiome-dependent mechanism.

Discussion
Previous studies provided evidence for the immunomodulatory potential of LCN2 [7][8][9][10]. We here extend and substantiate these observations by showing that LCN2 skewed the transcriptome of lung myeloid cells during homeostasis and dampened T cell responses upon influenza infection, resulting in diminished disease-related morbidity and mortality independent of viral clearance. We succeeded in linking the effects of LCN2 to the magnitude of CD8 + T cell responses via altered CD103 + DC functionalities. Finally, the LCN2-driven

PLOS PATHOGENS
immunomodulation did not occur directly, but relied on LCN2's impact on the microbiome composition, which we verified by experiments using antibiotics or cohousing approaches.
Influenza infection-related pathology in humans, non-human primates and mice is predominantly caused by exaggerated antiviral and hyper-inflammatory responses, rather than direct cytopathic effects of the virus [12,[33][34][35]. While critical for viral clearance and immunological memory, pro-inflammatory cytokine production and cytolytic activity by CD4 + and CD8 + T cells, respectively, importantly contribute to immunopathology [36,37]. Our data indicate that excessive T cell proliferation, resulting from more potent antigen presentation by CD103 + DCs to CD8 + T cells, mediate increased morbidity in Lcn2 -/mice. Our study identifies LCN2 as an important modulator of antiviral immunity and suggests that LCN2 fine-tunes DC activity during homeostasis and to prevent excessive T cell expansion and immunopathology during viral infections. Of note, tissue damage due to aberrant immune activation is not restricted to influenza infections, but has been suggested to play an important role in COVID-19 mortality [13]. While we here focused on influenza infection, LCN2-regulated inflammation and immunopathology might also importantly influence the course of other viral infections [15].
Recent research highlights the critical effect of microbiome compositions on maturation and functionality of the immune system [11]. As such, the gut microbiome was reported to influence DC migration to draining lymph nodes and T cell priming during influenza infection by providing signals leading to pro-IL1β and pro-IL18 expression at steady state and inflammasome activation upon influenza infection [29]. While antibiotic treatment interfered with these signals, intranasal or intrarectal injection of TLR ligands was sufficient to restore the microbiome dependent effects. Furthermore, sensing of host microbiota via TLR5 was found to facilitate antibody responses to influenza vaccines by promoting plasma cell differentiation [38]. Additionally, microbial metabolites such as desaminotyrosine and short-chain fatty acids were discovered to regulate antiviral immunity and to protect from influenza associated disease morbidity [39,40]. We here observed that LCN2-deficiency was associated with an altered site-specific intestinal microbial composition and that microbiome disruption by antibiotics abrogated LCN2-related differences in anti-influenza immunity. While we did not identify which specific features of the microbiota mediated the observed effects on host immunity, we identified several candidates. Among the multiple ASVs with differential abundance in ileal, cecal or stool samples, SFB (which were among the most abundant ASVs in the ileal mucosa of WT mice) were completely absent in Lcn2 -/mice. This common bacterial strain is involved in host-microbiome crosstalk as it can induce Th17 cells in the gut [41], and its abundance is affected by influenza infection [42].
Two previous studies have set out to assess the effect of LCN2 on stool microbial composition. Singh et al. reported distinct bacterial communities in Lcn2-deficient mice, which were linked to exacerbated colitis or neutralization of IL10 [43]. In agreement with our findings, this study found increased Bacteroidetes and Proteobacteria phyla and a decreased Tenericutes phylum in stool samples of Lcn2 -/mice. In contrast, Moschen et al. reported that LCN2 exclusively affects the intestinal microbiome in mice lacking IL-10 leading to colitis and spontaneous emergence of right sided colonic tumors [44]. These discrepancies could be explained by altered baseline microbiota due to differences in animal housing or the lower number of experimental animals with differences in statistical power in the study by Moschen et al. (4 vs 13-16 mice per genotype in the present study).
Several cell surface receptors for LCN2 have been identified: both SLC22A17 (also named 24p3R) and megalin bind LCN2 and mediate its cellular uptake [45,46]. Furthermore, LCN2 can bind to and signal through MC4R in the hypothalamus [3]. We conclude from our results that the effects of LCN2 on influenza-associated T cell responses are primarily dependent on the microbiome. While LCN2 exposure of DCs did not alter their antigen presentation potential in vitro, we cannot completely rule out that direct effects of LCN2 on immune cells contribute to the immune-regulatory properties of LCN2 in vivo. Further research is needed to fully elucidate these mechanisms. Considering its drastic immunomodulatory effects, variable LCN2 levels in human subjects [47] could have important clinical consequences. For instance, decreased LCN2 levels might identify patients at risk of exaggerated responses to inflammatory triggers such as viral infections, and, on the other hand, supraphysiological LCN2 levels might prevent proper function of the adaptive immune system. Recent evidence suggests a beneficial effect of chronically increased LCN2 levels in patients with obesity or type 2 diabetes [48,49], as it assists in counteracting obesity-induced glucose intolerance by reducing appetite and driving beta-cell proliferation [50]. Interestingly, diabetes and obesity are also associated with increased susceptibility to severe influenza infection [51,52], but the potential correlation between LCN2 levels and disease outcome in humans is not known. Considering the dampening effects of LCN2 on adaptive immunity and influenza-specific antibody levels, elevated LCN2 levels could be a reason for the poor vaccination responses observed in obese individuals [53]. While these observations can be interpreted as potential therapeutic scenarios, further research is required to test the applicability and translational feasibility of such LCN2-related treatment strategies.

Mouse model of influenza infection
For influenza infection, female mice were intranasally infected with 10-12.5 PFU PR/8 in 50 μl sterile injection-grade (sterile and endotoxin-free) 0.9% NaCl solution. For survival experiments, male mice received 15 PFU PR/8 in 50 μl. Mice were monitored daily and graded according to an internally developed scoring sheet, which assessed appearance, posture, change in body weight and body temperature, natural behavior and clinical signs. Animals with a combined score of >5 or individual scores (in one of the categories) of >3, which could result from hunched posture, more than 30% body weight loss, >5 degrees of body temperature drop, squeaking, self-mutilation or inactive behavior and visible signs of unrelated infections or bleeding, were euthanized. In some planned end-point experiments, bronchoalveolar lavage was performed by inserting a tracheal cannula (Venflon, BD Bioscience) and flushing the lungs with 1 ml NaCl. Otherwise, lungs were flushed by injecting 5 ml of endotoxin-free PBS (Gibco) into the left ventricle, removed under sterile conditions and processed. For plaque assay or RT-PCR, lung tissue was homogenized using a Precellys 24 homogenizer (Peqlab). For plaque assays, homogenates where directly stored at -80˚C, whereas for RT-PCR, aliquots of lung homogenates where frozen (-20˚C) in RA1 buffer (Macherey-Nagel) with 10% betamercaptoethanol (Calbiochem).

Determination of viral load by qPCR
RNA was isolated using the NucleoSpin1 RNA II kit (Macherey-Nagel), reverse transcription was performed using 700ng of isolated RNA and the iScript cDNA Synthesis Kit (Biorad), according to manufacturer's protocol. Real-time PCR was performed with SYBR Green Master Mix reagents (Applied Biosystems) on a StepOnePlus Real-Time PCR System (Applied Biosystems). Primer sequences targeting the influenza M gene [58] were CATGGAATGGCTAAAGA-CAAGACC (forward) or CCATTAAGGGCATTTTGGACA (reverse). The reference gene HPRT was quantified using the sequences GTTAAGCAGTACAGCCCCAAAATG (forward) and AAATCCAACAAAGTCTGGCCTGTA (reverse).

Multiplex immunoassay for measurement of BAL cytokines
BAL cytokines were measured using a custom multiplex immunoassay (eBioscience), according to the manufacturer's protocol. Briefly, magnetic beads where incubated with undiluted BAL (performed with 1ml sterile and endotoxin-free 0.9% NaCl solution) samples in a 96-well plate for two hours, followed by washing and incubation with detection antibody (30 minutes). After washing, incubation with streptavidin-PE and another washing step, the plate was read on a Luminex 200 instrument (R&D Systems).

Cell suspension preparation and flow cytometry
For flow cytometry analysis of lung tissue samples, representative pieces of each pulmonary lobe were weighed and homogenized using the lung dissociation kit (Miltenyi Biotec), according to the manufacturer's protocol. Tissue was digested for 30 minutes at 37˚C using DNAse I (12 U/ml; Sigma) and Collagenase I (160 U/ml; Gibco). Next, suspensions were filtered using 70 μM cell strainers (BD Biosciences), red blood cells were lysed using ACK lysis buffer (150 mM NH 4 Cl, 10 mM KHCO 3 , 0.1 mM Na 2 EDTA, pH 7.2-7.4; all chemicals from Sigma) and suspensions were again filtered through 40 μM strainers. PBS with 0.5% bovine serum albumin (BSA; BDA, cat. no. 8076.3) was used to stop red blood cell lysis and as buffer for all subsequent steps. In experiments assessing T cell antigen specificity, single cell suspensions were incubated with fluorescently labelled recombinant PR8 pentamer (ProImmune, dilution 1:5) for 10 minutes at room temperature. Cells were then incubated with anti-mouse CD16/32 antibody (TruStain fcX; BioLegend, dilution 1:500) to block unspecific binding to Fcγ receptors and viability dye (Fixable Viability Dye eFluor 780; eBioscience, dilution 1:1000) for 20 minutes at 4˚C, followed by staining with fluorescence-labelled antibodies (see S4 Table, 40 minutes, 4˚C). Mediastinal lymph nodes were mashed through 70 μM cell strainers, followed by Fcγ receptor blocking and staining. Cells were fixed using Fix and Perm (Nordic MUbio) reagents, according to the manufacturer's protocol. Analysis was performed using an LSR Fortessa (BD Biosciences) and FlowJo software (FlowJo LLC). Absolute cell numbers for lungs and lymph nodes were determined using counting beads (123 count eBeads, Thermo Fisher Scientific) or a hematocytometer (for spleens).

CD8+ T cell depletion
To deplete CD8 + T cells, WT and Lcn2 -/mice where intraperitoneally injected with 200 μg of anti-CD8a (clone YTS 169.4) in 200 μl NaCl on the day of infection with 12.5 PFU PR/8, and on day 3 and day 6 post infection [59]. Control mice received sterile PBS (carrier) injections on the same timepoints. Depletion of CD8 + T cells was assessed by FACS analysis of blood samples collected on days 2 and 5 post infection from the retro-orbital vein.

Bone marrow derived dendritic cells (BMDCs)
Bone marrow was isolated by flushing sterilized mouse femurs with 10 ml of RPMI medium. Bone marrow cells were cultured in tissue-culture-treated 6-well plates (Corning), in 4ml RPMI medium containing 10% FCS, 1% penicillin/streptomycin and 25 ng/ml murine GM-CSF (PeproTech), at a density of 10 6 cells per ml. On day 2, half of the medium was replaced with fresh medium containing 25 ng/ml murine GM-CSF. On day 3, the complete medium (containing non-adherent cells) was replaced with fresh medium containing 25 ng/ ml murine GM-CSF. Non-adherent cells were harvested on day 6 by gentle washing with sterile endotoxin-free PBS (Sigma) and incubated with biotinylated anti-mouse CD11c (clone N418; BioLegend). CD11c + cells (BMDCs) were purified by magnetic separation using streptavidin-coated magnetic beads (IMag Cell Separation System; BD). Isolated BMDCs were pulsed with ovalbumin (Grade V, Sigma, 100 μg/ml) for 16 hours. For some experiments, recombinant LCN2 (BioLegend) was added during the pulsing step at indicated concentrations. Ovalbumin-pulsed BMDCs were used for antigen presentation assays as described below.

RNA sequencing data analysis
Sequencing reads were adapter-trimmed using Trimmomatic [61] and aligned to the mm10 reference genome (STAR aligner [62]). Reads mapping to genes were counted using the sum-marizeOverlaps function (Bioconductor R package GenomicAlignments [63]). Differential gene expression was assessed using DESeq2 [64], whereby separate models per cell type and condition (lung-homeostasis, lymph node-infection) were formulated for all pairwise comparisons between genotypes. Genes were filtered using independent hypothesis weighting (ihw R package [65]). Impact of differential expression on KEGG pathways was assessed using Signaling Pathway Impact Analysis (SPIA, SPIA R package [66]). Genes with an FDR-adjusted p value of < 0.1 were considered differentially expressed.

Bacterial DNA collection, extraction, library preparation and sequencing
Sterile surgical tools were used for collection of microbial samples from mice. For lung samples, BAL was performed 10 times with 1 ml NaCl (aliquots were subsequently pooled) as described above. Next, the abdomen of the animal was sterilized using 70% EtOH and the peritoneal cavity was opened. Ileal samples were collected by cutting a 2.5 cm long piece of the terminal ileum 2 cm distal of the ileo-cecal valve. Cecal samples were collected by dissecting the whole cecum. From ileum and cecum, luminal samples were collected in a sterile microcentrifuge tube (Eppendorf Biopur) by gently squeezing out content from the distal end of the dissected intestinal segment. For collection of mucosal samples, the ileal segments where flushed with 10 ml NaCl to remove all luminal content, cut open longitudinally and scraping biopsies where collected using sterile glass slides. Bacterial DNA was isolated using the QIAamp DNA Microbiome Kit (Qiagen) according to the manufacturer's protocol and isolated DNA was eluted in 50 μl of DEPC-treated water (Roth). To control for contaminations introduced during sample collection, two harvest controls (NaCl used to flush cannulas inserted into mouse tracheas, without performing BAL), one isolation control and one PCR control were prepared identically to BAL samples and sequenced. Similarly, another PCR control sample was prepared for ileal mucosa samples. Isolated bacterial DNA from samples and controls was amplified using barcoded, Illumina adaptor-linked PCR primers that target the V1-V2 hypervariable region of the bacterial 16S rRNA gene [67]. Each sample was amplified using the Accuprime Taq DNA Polymerase High Fidelity kit (Invitrogen) with the following cycling parameters: Initial denaturation for 2 min at 94˚C, followed by amplification cycles starting with 30 sec denaturation at 94˚C, 30 sec annealing at 56˚C, and 60 sec elongation at 68˚C, with a final extension at 68˚C for 7 min. An initial enrichment PCR reaction (10 amplification cycles) was performed, followed by the amplification PCR. The number of cycles performed for amplification PCR varied between sampling sites: Stool: 20 cycles; ileum lumen: 25 cycles; cecum lumen: 35 cycles; ileum mucosa: 40 cycles. For BAL, two amplification PCRs were performed, with 40 (first reaction) and 35 (second reaction) cycles. The ideal number of amplification cycles for each sampling site was determined by increasing the number of cycles by five (starting from 20), until samples were sufficiently amplified (i.e. presence of a visible band by automated electrophoresis using an Agilent 4200 TapeStation system). All PCR reactions contained 2.5 μl of 10x AccuPrime buffer II, 0.5 μl of each 10 mM forward and reverse primers and 0.1 μl of Accu-Prime Taq DNA Polymerase, to which 1 μl of template DNA and 20.4 μl of ultrapure water (for the enrichment PCR reaction) or 2 μl of the enrichment PCR reaction and 19.4 μl of ultrapure water (for the amplification PCR reaction) were added. PCR products were screened for sufficient amplification and quantified using automated electrophoresis. Finally, libraries containing pooled equimolar PCR products and spiked with 40% phiX (Illumina) were sequenced using Illumina MiSeq technology in the 2x 350bp configuration (MiSeq Reagent kit v3) at the Biomedical Sequencing Facility of CeMM and the Medical University of Vienna.

16S rRNA gene sequencing data processing and analysis
Raw sequences were demultiplexed using qiime [68] commands split_libraries.py with options-r 999 -n 999 -q 0 -p 0.0001 (to prevent quality filtering at this stage) and split_sequen-ce_file_on_sample_ids.py. ASVs (amplicon sequence variants) were selected using dada2 [69]. Forward and reverse reads were trimmed to a length of 225 bp and filtered using parameters maxN = 0, maxEE = c (2,2) and truncQ = 2. Potential contaminants within unique ASVs were identified with the decontam R package [69] using prevalence of ASVs in control samples (for ileal mucosa and BAL samples) and correlation of ASV frequency with DNA concentration after PCR amplification as identification methods. Following contaminant removal, taxonomy was assigned against the SILVA 16s rRNA database [70] and ASVs which could not be assigned to kingdom Bacteria were removed from the dataset. One ASV was identified as belonging to genus Candidatus_Arthromitus. Since the misclassification of mammalian segmented filamentous bacteria (SFB) as Candidatus_Arthromitus (which would not be expected to be found in mammalian guts) is well described [71], we refer to this ASV as SFB. Next, samples with a total ASV count of less than 10000 were excluded from further analysis. The remaining samples were rarefied to the smallest library size above 10000 ASVs for the sampling site (between 10033 for stool samples and 27693 for cecal samples). Shannon diversity and Chao 1 ASV richness were calculated using the vegan R package [72]. Differences between genotypes (controlling for cage effects) were assessed using the nested.npmanova function (bio-diversityR package) on ASV counts normalized to sample library size (relative abundances). Statistical comparisons of Chao1 richness and Shannon diversity were carried out with linear mixed models using the lme4 and lmerTest R packages [73,74], while cage association was entered as a random factor. Linear mixed models were also used to determine which ASVs were differentially abundant according to genotype, after transforming ASV counts with a centralized log ratio (clr) transformation.

Antibiotic treatment and cohousing
For experiments involving microbiome perturbation by antibiotics, female WT or Lcn2 -/mice were treated with a mix of metronidazole ('Kabi', Fresenius), vancomycin (Vancocin, Baxter), ampicillin (Standacilin, Sandoz) and gentamicin (Braun, concentration of each 500 mg/l) in drinking water supplemented with 0.4% D-glucose (Sigma) starting at 5-6 weeks of age. Control mice received drinking water supplemented with 0.4% D-glucose. After 4 weeks of antibiotic treatment, mice were infected with 10 PFU PR/8. Antibiotic treatment was continued until the end of the experiment. For co-housing experiments, 4 weeks old WT or Lcn2 -/mice were housed together at equal ratios for 4 weeks prior to influenza infection. Control WT or Lcn2 -/mice were housed separately. As with antibiotic treatments, housing conditions were maintained throughout the infection period.

Quantification of bacterial classes using qPCR
Bacterial DNA from stool of separately housed and co-housed WT and Lcn2 -/animals was isolated using the QIAamp DNA Mini Kit (Qiagen) according to the manufacturer's recommendations. Perfecta SYBR Green SuperMix (Quanta) on a StepOnePlus Real-Time PCR System (Applied Biosystems) were used for real-time qPCR. Primer sequences for Bacteroidetes were CRAACAGGATTAGATACCCT (forward) or GGTAAGGTTCCTCGCGTAT (reverse) [75]. The conserved bacterial (pan-bacteria) 16S-gene was quantified using the sequences TCC TACGGGAGGCAGCAGT (forward) and GGACTACCAGGGTATCTAATCTT (reverse) and used for calculation of log 2 fold changes.

Statistical analysis
Comparisons between two groups were performed using Student's t test, except where noted otherwise. In experiments involving two grouping factors (e.g. genotype and antibiotic treatment), 2-way ANOVA was conducted, and group means were compared using the emmeans R package. Survival curves were compared using the asymptotic two-sample logrank test, as implemented in the coin R package [76]. In antigen presentation assays, technical replicates of biological replicates were analyzed. To avoid pseudo-replication, nested ANOVA was conducted in these experiments. Where appropriate, p values were corrected for multiple testing using the false-discovery rate approach [77]. Single p values below 0.05 or FDR-corrected p values below 0.1 (RNA-Seq and 16S rRNA gene sequencing) were considered statistically significant. WT and Lcn2 -/mice were treated with broad-spectrum antibiotics in drinking water. After four weeks treatment, mice were infected with PR/8-OVA and CD103 + DCs were sorted (60 hours post infection) from mediastinal lymph nodes (mLNs), followed by co-culture with purified splenic OT-I T cells. (B) Numbers of proliferated (proliferation dye low ) CD8 + T cells after 96h co-culture (1:100 DC: T cell ratio) with WT or Lcn2 -/-CD103 + DCs derived from antibiotics-treated mice. Replicates of 3 pools per genotype (each consisting of 6-7 mice) are shown. (C) Experimental setup for co-housing experiments. WT and Lcn2 -/mice were separately-(sep) or cohoused (co) for four weeks prior to stool sample collection. (D) Log 2 fold change between the phylum Bacteroidetes and total bacteria in stool samples, measured by qPCR. Statistical significance for (B) was assessed using nested ANOVA, and for (C) using Student's t test. n.s. not significant, �� p < 0.01. (TIF) S1