ICAMs are dispensable for influenza clearance and anti-viral humoral and cellular immunity

αLβ2 (LFA-1) mediated interactions with ICAM-1 and ICAM-2 predominate leukocyte-vascular interactions, but their functions in extravascular cell-cell communications is still debated. The roles of these two ligands in leukocyte trafficking, lymphocyte differentiation, and immunity to influenza infections were dissected in the present study. Surprisingly, double ICAM-1 and ICAM-2 knock out mice (herein ICAM-1/2-/- mice) infected with a lab adapted H1N1 influenza A virus fully recovered from infection, elicited potent humoral immunity, and generated normal long lasting anti-viral CD8+ T cell memory. Furthermore, lung capillary ICAMs were dispensable for both NK and neutrophil entry to virus infected lungs. Mediastinal lymph nodes (MedLNs) of ICAM-1/2-/- mice poorly recruited naïve T cells and B lymphocytes but elicited normal humoral immunity critical for viral clearance and effective CD8+ differentiation into IFN-γ producing T cells. Furthermore, whereas reduced numbers of virus specific effector CD8+ T cells accumulated inside infected ICAM-1/2-/- lungs, normal virus-specific TRM CD8+ cells were generated inside these lungs and fully protected ICAM-1/2-/- mice from secondary heterosubtypic infections. B lymphocyte entry to the MedLNs and differentiation into extrafollicular plasmablasts, producing high affinity anti-influenza IgG2a antibodies, were also ICAM-1 and ICAM-2 independent. A potent antiviral humoral response was associated with accumulation of hyper-stimulated cDC2s in ICAM null MedLNs and higher numbers of virus-specific T follicular helper (Tfh) cells generated following lung infection. Mice selectively depleted of cDC ICAM-1 expression supported, however, normal CTL and Tfh differentiation following influenza infection, ruling out essential co-stimulatory functions of DC ICAM-1 in CD8+ and CD4+ T cell differentiation. Collectively our findings suggest that lung ICAMs are dispensable for innate leukocyte trafficking to influenza infected lungs, for the generation of peri-epithelial TRM CD8+ cells, and long term anti-viral cellular immunity. In lung draining LNs, although ICAMs promote lymphocyte homing, these key integrin ligands are not required for influenza-specific humoral immunity or generation of IFN-γ effector CD8+ T cells. In conclusion, our findings suggest unexpected compensatory mechanisms that orchestrate protective anti-influenza immunity in the absence of vascular and extravascular ICAMs.


Introduction
The respiratory tract is comprised of distinct epithelial layers and vascular beds including the alveolar gas-exchange surfaces, which contain the largest vascular bed in the body (1). The bronchial and alveolar epithelial cells are key targets for viral infections (2). Infected respiratory epithelial cells triggers multiple sequential immune responses critical for both virus clearance and for the establishment of anti-viral immune memory (3). It is commonly accepted that leukocyte adherence and emigration across pulmonary capillaries is molecularly distinct from the classical multistep cascade of leukocyte adhesion to and emigration across inflamed postcapillary venules (4,5). When infected with highly pathogenic pneumotropic influenza strains, different innate and adaptive immune cells must be selectively recruited to various compartments of virus-infected upper and lower airways and lung alveoli using distinct trafficking cues. In addition, different virus-specific naïve T and B lymphocytes must enter the main lung draining lymph nodes in which the presentation of virus antigens is executed by professional antigen presenting dendritic cells (6). Comprehensive knowledge of the trafficking signals and specifically the cell adhesion molecules functioning on these distinct vascular beds during influenza infections is still missing. Three of such adhesion molecules ICAM-1 (CD54), ICAM-2 (CD102) and VCAM-1 (CD106) are single chain Ig superfamily transmembrane glycoproteins (1,7). ICAM-1 and ICAM-2 are ligands of the key leukocyte b2-integrin LFA-1 (8), whereas VCAM-1 is a major vascular ligand for the integrin a4b1 (VLA-4). ICAM-1 and VCAM-1 expression is generally upregulated by inflammatory cytokines, while ICAM-2 (CD102) is constitutively expressed (9,10).
These three ligands are also constitutively expressed on high endothelial venules (HEVs), specialized postcapillary venules that are the main portal of lymphocyte entry to all lymph nodes (LNs) (11). ICAM-1 and ICAM-2 are also expressed on numerous immune cells, in particular antigen presenting cells (APCs) like DCs, B cells and macrophages (12), and are thought to stabilize functional immune synapses between these cells and other immune cells (13). Nevertheless, the roles of these b2integrin ligands in the complex communications between distinct lung APCs and T cells necessary for the establishment of early and long-lasting antiviral cellular and humoral immunity have never been investigated in the context of influenza infections.
Multiple studies have suggested the usage of lung ICAMs or their main CD18 (b2) integrin receptors in leukocyte recruitment to the lungs in distinct infections (4). To systematically dissect the roles of the two ICAMs in innate and adaptive immunity to influenza infections, we followed the course of H1N1 influenza infection in WT and in ICAM-1/2 -/mice. Additionally, we addressed the role of a4 integrin interactions in leukocyte entry to distinct lung compartments and lung draining lymph nodes. Strikingly, ICAM-1/2 -/mice normally recovered from infection, elicited potent humoral immunity, and generated long lasting cellular immunity to both homosubtypic and heterosubtypic variants of influenza A strains. Intravital staining of lung vessels following influenza infection suggested a unique compartmentalization of VCAM-1 and ICAM-1, respectively, between virus infected peribronchial vessels and lung capillaries. Nevertheless, vascular ICAMs were not obligatory for neutrophil or NK cell recruitment to virus infected lungs. On the other hand, vascular ICAMs promoted T and B lymphocyte entry to lung draining lymph nodes both prior to and after infection. Strikingly, however, viral Ag dependent proliferation and differentiation of naïve CD4 + T cells into effector Tfh cells took place normally and coincided with the appearance of hyperactive cDC2 subsets which accumulated in ICAM-1/2 -/-MedLNs of Organ preparation for flow cytometry Bone marrow was extracted from the femurs and tibias of mice as described (15). Spleens were harvested, placed on a 40 mm cell strainer in a 100x15 mm petri dish with 5 mL 1X PBS and crushed with a plunger to dissociate it into the dish. The suspension was pipetted into a 15 mL tube and centrifuged at 394 x g for 5 minutes at 4°C. RBCs were subsequently lysed with an RBC lysis buffer (Sigma Aldrich, Cat. R7757) and the cells resuspended in ice-cold flow cytometry buffer (PBS, 1% BSA, 0.1% Na azide and 5 mM EDTA). Blood samples were collected from the retro-orbital sinus of the mice. Single cell suspensions were prepared using the BD Pharm Lyse (BD Bioscience, Cat. 555899) lysing buffer according to the manufacturer's instruction (BD Bioscience). Samples were centrifuged at 394 x g for 5 minutes at 4°C the blood samples were re-suspended in ice-cold flow cytometry buffer.
Isolation of BAL fluid and preparation of total lung cell suspension for flow cytometry Mice were euthanized, lungs were transcardially perfused (10 mL 1X PBS) and BAL fluid (BALF) was collected twice with 5 mM EDTA/PBS aliquots introduced through a tracheal catheter. The BAL fluid was centrifuged at 394 x g for 5 minutes at 4°C. For preparation of lung cell suspensions, lungs were extracted, minced and incubated in RPMI-1640 containing collagenase type 4 (1.5 mg/mL) and DNase I (20 mg/mL) at 37°C for 45 minutes. The lung cell suspensions were transferred through a 100 mm cell strainer and centrifuged at 394 x g for 5 minutes at 4°C. RBCs were subsequently lysed with an RBC lysis buffer (Sigma Aldrich, Cat. R7757). The cells from both BALF and lungs were re-suspended in ice-cold buffer and analyzed by flow cytometry. For analysis of protein surface expression, cells were labeled with fluorescent conjugated antibodies (10 mg/mL), incubated on ice for 20-30 minutes, centrifuged at 394 x g for 5 minutes at 4°C and resuspended in ice-cold flow cytometry buffer (PBS, 1% BSA, 0.1% Na azide and 5 mM EDTA). Cells were analyzed on either an LSR-II Flow Cytometer (BD Biosciences) or a CytoFLEX flow cytometer (Beckman Coulter). Intracellular staining for BCL6 and Foxp3 was performed using the True-Nuclear ™ Transcription Factor Buffer Set (Biolegend, 424401) according to the manufacturer's recommendations. Data were analyzed using FlowJo software (BD Biosciences).

Pulmonary intravital microscopy
Imaging and analysis of lung neutrophils was performed as previously described (16). For neutrophil phenotypic quantification the following was defined (17): tethering was defined as a discreet neutrophil interaction with the vascular wall, which arrests its circulatory movement for less than 30 seconds. Crawling was defined as continuous interaction of a neutrophil with the vascular wall for more than 30 seconds, which involves a polarized cell that does not remain stationary. Adhesion was defined as a stationary neutrophil that was not mobile and remained static for at least 30 seconds.

Pathogens
The PR8 strains used in this study were kindly provided by Prof. Natalio Garbi (University Hospital Bonn). The X31 strain was kindly provided by the Shulman lab (WIS). Pseudomonas aeruginosa (strain PAO1) was grown on agar plates overnight at 37°C. Multiple single colonies were picked from plates washed 3 times in sterile PBS and quantified by OD600.

In vivo influenza infections
Anesthetized mice were held in supine and the tongue was gently pulled out using forceps to gain visualization of the larynx. Various PR8 influenza doses (sublethal: 15-155 PFU) and lethal: 3x10 3 PFU) were delivered. For I.T. administration: PR8 doses were suspended in a total volume of 50 ml of sterile 1X PBS and placed on the base of the tongue while the nose was closed, thus forcing the mice to breathe through the mouth. The nose and tongue were released after at least two breaths had been completed. For I.N. delivery: anesthetized mice were placed in the dorsal recumbency. A pipet tip containing 30 mL/mouse of sublethal doses of PR8, PR8-SIINFEKL or PR8-OVAII (OVA 323-339 peptide) influenza inoculum in sterile PBS was placed above the mouse nostril and the solution was slowly injected. The mice were validated to have inhaled the entire inoculum drop. For recall infections: infected mice were challenged 40 days after the primary infection. Weight loss was tracked throughout the course of infection.
Quantification of pulmonary PR8 viral load by real-time quantitative PCR WT and ICAM-1/2 -/mice were euthanized on day 4 or 12 post I.N. infection with 30 PFU of PR8 influenza. Lungs were snap-frozen and kept at -70°C. Lungs were slowly thawed on ice. Viral genomic PR8-RNA was extracted from the infected lungs and transcribed into cDNA by RT-PCR using NA specific PR8 influenza virus primers (NA Fw: TTAATGAGCTGCCC TGTCGG, NA Rv: CACTTGCTGACCAAGCAACC). The cDNA was used as a template for the RT-qPCR reaction and results are represented as mean PFU equivalent per microgram RNA (mean PFUe/mg RNA). PR8-cDNA samples containing virus-complementary sequences served as quantification standards for the RT-qPCR reaction.

MedLN preparation for flow cytometry
Isolated LNs were placed on a 40 mm cell strainer, crushed with a plunger and dissociated into a 10x60 mm petri dish. The suspension was transferred into a 96 v-bottom plate (approximately 200ul/well) and the plates were centrifuged at 394 x g for 4 minutes at 4°C. Pellets were collected with two rounds of washes with 100 ml of 1X PBS. Plates were then centrifuged at 394 x g for 5 minutes at 4°C and the cells were resuspended in ice-cold flow cytometry buffer (PBS, 1% BSA, 0.1% Na azide and 5 mM EDTA). For DC isolation, each LN was placed in a well of a 12-well plate and suspended in 0.5 mL of PBS +/+ supplemented with 1 mg/mL Collagenase D (Roche, COLLD-RO). The treated LN was torn with two syringes, incubated for 30 minutes at 37°C and subsequently placed on a 40 mm cell strainer in a 10x60 mm petri dish in the presence of 2 mL PBS.

Blocking of a4 integrin on innate leukocytes and T cells
Innate leukocyte blocking: On day 0, 60 mgs of the anti-a4 antibody (BioXcell clone: PS/2) or isotype control (BioXcell clone: 2A3) were administered I.V. in a total volume of 150 ml (in 1X PBS) via the retro-orbital sinus. 2 hours later, mice were I.N. infected with 30 PFU of the PR8 influenza virus. For the next three days (day 1, 2 and 3), mice were I.V. administered with 30 mg of the same blocking regimen (PS/2 or 2A3). On day 4 mice were euthanized and BALFs of WT and ICAM-1/2 -/mice were analyzed by flow cytometry for neutrophils and NK cells. For T cell blocking: Spleen-derived 10 7 WT CD45.1 OT-I T cells (2x10 6 naïve T cells) were incubated for 20 minutes with 30 mg of PS/2 or 2A3 in 100 ml and subsequently administered I.V. via the retro-orbital sinus. After 4 hours, an additional 30 mg dose of PS/2 or 2A3 was administered I.V. and 14 hours later the accumulation of the I.V. injected CD45.1 OT-I CD8 + T cells in the MedLNs was determined by flow cytometry. In an additional experiment, at this time point ICAM-1/2 -/mice were I.N. administered either with a sublethal dose of PR8-SIINFEKL or PBS and 4 hours later were administered with 30 mg of PS/2 or 2A3 I.V. One or four days later, ICAM-1/2 -/were euthanized and the total number of CD45.1 OT-I CD8 + T cells recovered from the MedLNs of the differently treated mice groups was determined by flow cytometry.

MedLN immunohistochemistry
Naïve WT and ICAM-1/2 -/-MedLNs were harvested, fixed for 2 hours at RT in 1% PFA/PBS solution and then placed in 30% sucrose solution O.N. at 4°C. The next day, samples were transferred to a fresh solution of 30% sucrose for an additional 24 hours at 4°C. On the third day, each MedLN was placed in a small cryomold (Fisher scientific, Cat. NC9511236), embedded in OCT compound (Sakura, Cat. 4583), snap frozen on dry ice nuggets and stored at -80°C. Tissue sections (10mm thick) were cut using a cryostat and slides were placed in chilled 70% alcohol for 7 minutes. Prior to staining slides were rehydrated twice with PBS, treated for 5 minutes in 1% SDS/PBS, washed 3 times and blocked with 3% BSA in PBS-T (PBS/0.05% Tween-20). Sections were incubated with primary antibodies diluted in 1% BSA in PBS-T O.N. at 4°C. The following antibodies were used for staining: mouse monoclonal anti-VCAM-1 IgG (clone 429, BD Bioscience, Cat. 550547) and mouse monoclonal anti-PNAd IgM (MECA-79, Biolegend, Cat. 120802). After washing 3 times for 5 minutes in PBS-T at room temperature, the sections were incubated at 4°C O.N. with the following secondary antibodies: anti-rat IgG conjugated to Dylight 549 (Jackson, Cat. 712-505-153) and anti-rat IgM conjugated to Alexa Fluor 647 (Jackson, Cat. 716-605-020). The next day slides were washed and stained with DAPI (1:3000) (Biolegend, Cat. 422801).

OT-II CD4 + Tfh effectors and GC B cell generation in MedLNs following PR8-OVAII influenza infection
Spleen-derived 0.1x10 6 CD45.1 OT-II T cells were injected I.V. into WT, ICAM-1/2 -/-, DC DICAM-1 or WT (ICAM-1 fl/fl ) mice. After 18 hours, mice were I.N. infected with a sublethal dose of PR8-OVAII influenza. Mice were euthanized on days 4 or 12 and the totals of the CD45.1 OT-II CD4 + Tfh cells were determined by flow cytometry in the MedLNs. The number of GC B cells was determined on day 12 post infection.

ELISA detection of serum IgG and IgG2a antibodies against PR8 influenza virus
WT and ICAM-1/2 -/mice were infected with PR8 influenza and sera samples were collected on days 12, 20 or 40 post infection. The levels of PR8 influenza virus-specific serum IgGs were determined by enzyme-linked immunosorbent assay (ELISA). PR8 virus was purified on sucrose gradient (22-50%). Nunc MaxiSorp plates (Thermo Fisher Scientific, USA) were coated with the purified PR8 virus (5x10 5 PFU/well) in Na 2 CO 3 / NaHCO 3 buffer (Sigma Aldrich, Cat. C3041) at 4°C overnight. Plates were blocked with ELISA buffer (50 mM Tris, pH 7.6 + 140 mM NaCl + 0.05% Tween 20 + 2% BSA) for 60 minutes at 37°C. Following blocking and washing with PBST buffer (PBS + 0.05% Tween 20), the plates were incubated with serum samples diluted 1:100 to 1:3x10 4 in ELISA buffer for 1 hour at 37°C. Following washing, alkaline phosphatase-conjugated anti-mouse IgG (Sigma Aldrich cat. A5153) was used (diluted 1:1000) as a secondary antibody. After washing, P-nitrophenyl phosphate substrate (Sigma, Israel, cat. N2770) was added, and after 60 minutes of substrate incubation at room temperature, the optical density was measured by using a spectrophotometer (SpectraMax 190 microplate reader, Molecular Devices, Sunnyvale, CA), at 405 nm. Each data point represents the mean OD values from the ELISA analysis of three replicates.

OT-I CD8 + T effectors generated in MedLNs and lungs following PR8-SIINFEKL influenza infection
Spleen-derived 2x10 6 CD45.1 OT-I T cells were labeled with CFSE (5 mM) (Molecular Probes, Cat. 11524217) as described (19)  In vivo T cell proliferation and intracytoplasmic IFN-g staining CFSE labeled OT-I CD8 + T cells were recovered from the MedLNs of WT and ICAM-1/2 -/mice 4 or 7 days after sublethal PR8-SIINFEKL infection and lymphocyte proliferation was assessed by CFSE dilution. "Percent Divided" refers to the percentage of the original T cell population that underwent at least one division. "Division Index" is defined as the average number of cell divisions per cell in the original population. For intracellular staining of IFN-g, cells were stimulated with 50 ng/ mL PMA (Sigma Aldrich, Cat. 16561-29-8) and 500 ng/mL Ionomycin (Sigma Aldrich, Cat. 13909) for 1.5 hours at 37°C and then treated with Brefeldin A (Biolegend, Cat. 420601) and Monensin (Biolegend, Cat. 420701) as described in Intracytoplasmic cytokine staining of IL-6. Cells were fixed, permeabilized, and subjected to intracellular IFN-g staining with the Cytofix/Cytoperm kit (BD Biosciences, Cat. BD 554714).

Statistical analysis
Data were statistically analyzed using GraphPad Prism software. Student's two-tailed unpaired t-test was used to determine the significance of the difference between the means of two groups. One-way ANOVA tests were used to compare the means among three or more independent groups/categories. Data are shown as mean ± SEM. Asterisks indicate significant differences. Significance was set to p < 0.05 (*p < 0.05, **p < 0.01 and ***p < 0.001, ****p < 0.0001).

Results
ICAMs are largely dispensable for leukocyte generation and margination inside the lung microvasculature We initially characterized the cellular composition of all major types of leukocytes in the lungs of ICAM-1/2 -/mice. The total number of the major sentinel leukocytes, alveolar macrophages as well as lung DCs, were normal in the ICAM-1/2 -/mice (Supplementary Figures 1A-D). At steady state, the pulmonary capillary bed contains a large pool of marginated leukocytes which comprise 20-60 fold more neutrophils as well as NK cells and monocytes (20) than in the systemic circulation (21). In spite of a similar number of neutrophils produced in the bone marrow of ICAM-1/2 -/mice ( Figure 1A), we found higher amounts of blood circulating neutrophils, as well as much more abundant spleen neutrophils in these mice ( Figure 1A). The higher numbers of these neutrophils could be explained by inability of the blood circulating neutrophils to get retained by ICAM-1 and ICAM-2 expressing capillaries and microvessels outside the lungs (22). Nevertheless, the number of neutrophils recovered from WT and ICAM-1/2 -/lungs were similar ( Figure 1A). Notably, the majority of these lung neutrophils were intravascular, whereas most spleen and bone marrow neutrophils were extravascular in both WT and ICAM-1/2 -/mice ( Figure 1A). The loss of both ICAMs also did not affect the numbers of Ly6C high inflammatory monocytes marginating through the lung vasculature and their partition between intravascular and extravascular compartments at steady state conditions ( Figure 1B). In contrast, NK cells and patrolling monocytes (Ly6C low ) (23), were significantly reduced in the resting lungs ( Figures 1C, D) highlighting a key role of the two vascular ICAMs in the retention of these leukocytes inside the lung vasculature. Likewise, the numbers of intravascular Tregs recovered from ICAM-1/2 -/lung vessels were significantly reduced relative to WT lung vessels ( Figure 1E), in spite of similar totals of conventional lung CD4 + and CD8 + T cells (Supplementary Figure 2). Similar to patrolling monocytes, Tregs express constitutively adhesive LFA-1 (24) and may use constitutively expressed MHC-II/self-peptide complexes within subsets of alveolar capillaries for survival (25). In contrast to these leukocytes, iNKT cells appeared normal in spite of their dramatic loss in the ICAM-1/2 -/liver (Supplementary Figure 3). Furthermore, a subset of HSCs populating the lungs was also normally retained in ICAM-1/2 -/lungs consistent with their lack    Leukocyte counts, their vascular partition, and neutrophil margination in the lung vasculature are not affected by loss of ICAM-1 and ICAM-2.
The relative numbers of neutrophils (CD45 + CD11b + Ly6G + ) (A), classical (Ly6C hi ) monocytes (CD45 + CD11b + Ly6G -Ly6C hi ) (B), NK cells (CD45 + CD11b + NKp46 + ) (C), patrolling (Ly6C lo ) monocytes (CD45 + CD11b + Ly6G -Ly6C low ) (D) and Tregs (CD45 + CD4 + CD25 + Foxp3 + ) (E) in the indicated organs. The partition of each leukocyte type in intravascular (red) and extravascular (white) organ compartments is shown in the insets next to each graph. n= 5 for all bone marrow determinations; n= 5-12 for spleen; n= 5-13 for lung determinations of either WT or ICAM-1/2 -/mice and n=12-14 for blood samples analyzed for each leukocyte type. Data were combined from at least 2 independent experiments. (F) Pulmonary intravital microscopy of neutrophils marginating in either WT or ICAM-1/2 -/lungs. The experimental design is shown in the left panel. Neutrophils and blood vessels were stained by I.V. injection with the indicated fluorescently labeled mAbs. The middle panel depicts representative still images from two-photon video microscopy movies (scale bars represent 128 mm), and the bar graph shows the total neutrophils per field of view (FOV) as assessed by imaging. The total numbers of circulating and adherent neutrophils and their dynamic phenotypes grouped in three categories (tethering, crawling and arrested (stationary)) are depicted in the right panels. n= 8 for each group. Statistical significance was determined by two-tailed, unpaired Student's t tests. *P < 0.05, ****P < 0.0001, ns, not significant. The error bars indicate the SEM of each measurement.
of b2 integrin expression (data not shown). Interestingly, this subset also lacked a4 integrin expression, in contrast to BM HSCs (Supplementary Figure 4).
To assess whether marginating neutrophils use the capillary expressed ICAM-1 and ICAM-2 for arrest or crawling, we next assessed by intravital microscopy the ability of individual neutrophils to tether (i.e., transient adhesion), continuously crawl or firmly arrest (i.e., stationery) along either WT or ICAM-1/2 -/pulmonary capillaries. Notably, about 50% of neutrophils entering WT capillaries crawled significant distances on these vessels, but this crawling behavior was only mildly reduced in ICAM-1/2 -/lung capillaries ( Figure 1F, middle bar graph). In contrast, the number of neutrophils firmly arrested (i.e., stationary) inside the capillaries was significantly reduced ( Figure 1F, right bar graph). Thus, the combined fraction of crawling and firmly arrested neutrophils visualized was slightly but significantly reduced in ICAM-1/2 -/mice. In contrast, deficiency of the CD11b integrin on similar neutrophils totally abrogated their crawling behavior (17). Furthermore, the presence of ICAMs inside all lung vessels as well as on alveolar epithelial cells was dispensable for neutrophil recruitment, extravasation and entry into the bronchoalveolar space during an acute infection with Pseudomonas aeruginosa, as well as for bacterial clearance (Supplementary Figure 5). Taken together our results suggest that vascular and epithelial ICAMs are dispensable for lung homeostasis and the survival of all major sentinel leukocytes, as well as for the margination of major blood circulating leukocytes except for patrolling monocytes, NK cells and Tregs passing through the pulmonary vasculature.
ICAMs in the lungs and lung draining lymph nodes are dispensable for influenza clearance, mice recovery and protection from homosubtypic influenza re-infection A classical model of viral lung infections is influenza, a leading cause of respiratory tract diseases worldwide (26). To test the involvement of lung ICAMs in leukocyte recruitment and clearance of this pathogen, we used the lab adapted influenza A virus (A/Puerto Rico/8/1934 (H1N1) (herein: PR8 influenza) (27). We compared two routes of lung infections: the physiological intranasal route (I.N.) and the widely used experimental route of intratracheal infection (I.T.). WT and ICAM-1/2 -/mice were infected with different doses of the PR8 influenza virus and both their weight loss and survival rates were compared. As expected, this virus spread primarily along the bronchial tree of the lungs ( Figure 2A and Movie 1). Strikingly, the mortality rates caused by either semi-lethal or lethal doses of PR8 influenza were indistinguishable between WT and ICAM-1/2 -/mice ( Figure 2B). Furthermore, similar rates of weight loss were observed in both mice groups in either administration route (I.T. and I.N.) ( Figures 2C, D). In both cases, by day 12, all mice started to regain weight and by day 40 they returned to or surpassed their original body weight ( Figure 2D). Consistently with these clinical observations, both PR8 expansion and clearance rates were comparable for both WT and ICAM-1/2 -/mice groups ( Figure 2E). Since PR8 expands mainly in epithelial cells (2), these results indicate that loss of epithelial ICAMs doesn't impair influenza infection, contrasting a previous report on ICAM-1 dependence of this infection in airway derived epithelial cells as assessed in vitro (28).
The hallmark of normal adaptive response to viral infection is the ability of the infected animals to develop full protection against a secondary challenge with a homosubtypic virus (immunity to the same serotype virus) (29, 30). Both WT and ICAM-1/2 -/mice groups were therefore pre-infected I.N. or I.T. with low doses of PR8 and 40 days later, when full humoral and cellular immunity are established (31), all mice were I.T. challenged with a lethal dose of 3x10 3 PFU of PR8 ( Figure 2C). Strikingly, both WT and ICAM-1/2 -/mice similarly survived this secondary lethal virus challenge ( Figure 2D) suggesting that the ICAM-1/2 -/mice were fully protected from the viral challenge. Thus, ICAM-1/2 -/mice not only normally cleared the PR8 influenza introduced via different routes, but also mounted long lasting memory responses against a secondary infection with a lethal dose of a homosubtypic influenza virus challenge.

ICAMs are dispensable for neutrophil and NK cell entry to influenza infected lungs
Initial clearance of influenza A virus infection involves neutrophils and NK cell activities (32,33). Neutrophils are among the first cells recruited to the lungs in response to this infection, where they phagocytose the virus and initiate various antiviral programs (34,35). NK cells co-recruited to the infected lungs facilitate elimination of influenza by binding viral HA via their NKp44 and NKp46 receptors (36). We therefore next addressed the roles of ICAMs in the recruitment of these leukocytes to PR8 infected lungs during this time frame. Since PR8 infection was restricted to the bronchial tree during the first 4 days of infection ( Figure 2A and Movie 1), we first assessed the luminal distribution of these ligands in different lung vessels around the bronchial tree and in pulmonary capillaries, the two major vascular platforms for leukocyte extravasation into infected lungs (2). To that end, we used LSM microscopy for whole lung imaging of both peribronchial vessels and pulmonary capillaries following intravital luminal staining with fluorescently labeled anti-ICAM-1 and VCAM-1 specific mAbs ( Figure 3A). Notably, while alveolar capillaries in influenza-infected lungs lacked VCAM-1 expression, high surface ICAM-1 expression was detected on all capillaries (Figures 3B, C and Movies 2, 3). In contrast, high and uniform luminal VCAM-1 expression was detected on both large and small bronchi and bronchioles ( Figure 3D top panel, Movie 4). Surprisingly, PR8 infection did not significantly elevate the luminal expression of either ICAM-1 or VCAM-1 in either the peribronchial or capillary lung vasculature ( Figures 3B, C). Furthermore, VCAM-1 expression on infected peribronchial vessels was not elevated in ICAM-1/2 -/lungs compared to WT lungs, ruling out a compensatory contribution of this ligand to leukocyte entry into the infected ICAM-1/2 -/lungs ( Figure 3D bottom panel and Movie 5). Thus, during infection, the main vascular ligand expressed by the vast majority of peribronchial vessels is VCAM-1, while the main vascular ligand expressed by all pulmonary capillaries is ICAM-1 with constitutive expression of the low affinity LFA-1 ligand ICAM-2 shared by both capillary and peribronchial blood vessels (22).
Assessing the contribution of these ligands to the entry and accumulation of either neutrophils or NK cells in lungs infected with different sublethal viral doses, we found that neither leukocyte used ICAM-1 or ICAM-2 for recruitment into the infected lungs ( Figure 3E). In order to follow the potential role of VCAM-1 in neutrophil and NK cell recruitment to PR8 infected lungs, we first verified that both leukocytes comparably expressed the main VCAM-1 binding integrin VLA-4 (a4b1) on their surface during active infection in both WT and ICAM-1/2 -/mice ( Figure 3F). Notably, the majority of the a4 integrin subunit expressed by these leukocytes was associated with b1 rather than with b7 subunits ( Figure 3F). Interestingly, the introduction of a4 blocking mAb prior and during early phases of infection ( Figure 3G) did not inhibit neutrophil recruitment to infected lungs of either WT or ICAM-1/2 -/mice in a statistically significant manner ( Figure 3H). Surprisingly, similar introduction of a4 integrin blocking mAb resulted in significant inhibition of the recruitment of NK cells to PR8 infected WT lungs, consistent with NK cell usage of VCAM-1 for their recruitment ( Figure 3I). NK cell recruitment to   ICAM-1/2 -/lungs, on the other hand, was not inhibited by similar a4 integrin blocking ( Figure 3I). Importantly, this mAb was not reported to exert cytopathic effects in various chronic inflammatory settings (37). Taken together, neither neutrophils nor NK cells required ICAMs to get recruited into PR8 infected lungs. Since infected lungs do not express the alternative a4 integrin ligand MadCAM-1 (38) and NK cells express negligible a4b7 ( Figure 3F), our results highlight a potential role of VLA-4-mediated NK cell interactions with VCAM-1 expressed by the peribronchial lung vessels during their entry into influenza infected WT lungs. Neutrophils, on the other hand, enter PR8 infected lungs independently of expression of either capillary expressed ICAMs or peribronchial vessel-expressed VCAM-1 ( Figure 3H).
ICAMs are dispensable for CD4 + T cell differentiation, Tfh generation, T-dependent B lymphocyte differentiation, and virus specific IgG production following PR8 infection The MedLNs are the key lung-draining LNs of the lower respiratory tract (39). The early anti-influenza virus clearance is tightly dependent on generation of virus specific IgG antibodies (40) produced by plasma cells in B cell follicles of the MedLNs (41,42). To further explore the cellular basis for the normal humoral response of ICAM-1/2 -/mice, reflected by normal recovery from sublethal PR8 infection ( Figure 2B), normal viral clearance ( Figure 2E), and full protection from a lethal homosubtypic viral challenge 40 days post primary infection ( Figure 2D), we next dissected the accumulation of individual B and T cell subsets involved in antiviral humoral immunity in MedLNs that drain influenza infected lungs. At steady state, the hematopoietic lineage cells (CD45 + ) populating these LNs including CD19 + and B220 + B cells as well as CD4 + T cells poorly accumulated in the resting MedLNs of ICAM-1/2 -/mice ( Figure 4A), in spite of normal HEVs in these mice ( Figure 4B). Strikingly, however, ICAMs were not required for endogenous B-2 cell accumulation in the MedLNs following viral infection ( Figure 4C). Furthermore, the CD5 + B-1a cell subset, which produces natural broad specificity anti-influenza IgMs that contributes to early protection (43), also normally accumulated in ICAM-1/2 -/-MedLNs of PR8 infected mice ( Figure 4D). Similarly, extrafollicular IgG producing plasma cells which provide an initial burst of T-dependent antiinfluenza antibodies essential for early control of infection (44), were also recovered at similar numbers in the MedLNs of both WT and ICAM-1/2 -/mice following infection ( Figure 4E and Supplementary Figure 6).
Since B-2 B cells undergo isotype switching and affinity maturation depending on helper viral antigen specific CD4 + T cells, we next followed how influenza antigen specific CD4 + T cells accumulate inside the lung draining MedLNs of virus infected ICAM-1/2 -/mice. To that end, we infected either WT or ICAM-1/2 -/mice with a PR8 strain encoding the OVAII sequence (OVA 323-339 peptide) recognized by OT-II CD4 + transgenic T cells. As expected, prior to and shortly after infection the homing of transferred virus specific OT-II CD4 + T cells into ICAM-1/2 -/-MedLNs was markedly impaired ( Figure 4F). Consistent with their poor homing, the OVA specific CD4 + T cells also gave rise to significantly lower numbers of monoclonal OT-II effector CD4 + T cells (i.e., CD44 hi CD62 lo ) by day 4 post infection ( Figures 4G, H). Strikingly, however, similar numbers of early OT-II T follicular helper (Tfh) cells were recovered inside the MedLNs of both influenza infected WT and ICAM-1/2 -/mice at later time points post infection ( Figure 4I). Nevertheless, these early Tfh cells expressed lower content of the major Tfh transcription factor, BCL6 ( Figure 4J), suggesting that their early maturation was partially ICAM-1 and ICAM-2 dependent, possibly due to loss of B cell ICAMs implicated in T-B cell synapses (45).
The differentiation of viral specific naïve CD4 + T cells into early and late B cell helper T cells and Tfh effectors, in the MedLNs, depends on proper education by migratory cDC2 subsets carrying influenza antigens into these LNs (46)(47)(48). Importantly, at steady state, the cellularity of all cDCs populating the MedLNs remained normal in ICAM-1/2 -/mice (Supplementary Figure 7). Nevertheless, the cDC2 subset was recruited at higher numbers to ICAM-1/2 -/-LNs following PR8 infection ( Figures 4K, L), and were hyperactivated, as evident by their significantly elevated levels of the major co-stimulatory molecules CD80 and CD86, as well as elevated IL-6 production ( Figure 4M). Thus, the proliferation and differentiation of CD4 + T cells into effective Tfh lymphocytes was probably driven by DC overexpression of these critical Tfh differentiation cytokines and co-stimulatory molecules. Importantly, the cDC2 hyperactivation and their enriched IL-6 content in the MedLN of infected ICAM-1/2 -/mice did not appear to be the result of the reduced Treg content in the lungs and MedLNs of these mice ( Figure 1E and Supplementary Figure 8A). This conclusion arose from the finding that forced depletion of endogenous Tregs from WT mice with anti-CD25 mAb (Supplementary Figure 8B) did not increase cDC2 accumulation or stimulation following PR8 infection (Supplementary Figures 8C-E).
We next followed the fate of monoclonal Tfh cells on day 12 post infection, when germinal centers (GCs) are generated inside the MedLNs ( Figure 5A). Strikingly, significantly elevated numbers of monoclonal OVA specific Tfh cells were generated from adoptively transferred naïve CD4 + OT-II T cells inside MedLNs of influenza infected ICAM-1/2 -/mice ( Figure 5B). In contrast, the total numbers of endogenous polyclonal GC B cells (Fas + , CD38 low ) recovered 12 days post infection were similar in WT and ICAM-1/2 -/-MedLNs ( Figure 5C). To dissect the potential contribution of DC ICAMs to this multistep humoral response separately from the roles of endothelial and stromal ICAMs, we prepared a new mouse model devoid of DC ICAM-1 by first  Defective homing of early CD4 + T cells post influenza infection is counterbalanced by augmented cDC2 accumulation and activation inside ICAM-1/2 -/mediastinal LNs. (A) The total numbers of CD45 + cells, CD4 + T cells, CD19 + and B220 + B cells in resting MedLNs of either WT or ICAM-1/2 -/mice determined by flow cytometry. The subset totals were recovered from the same mice, except for the B220 + totals, which were analyzed from a separate experiment.  Figures 9A, B). These mice lost ICAM-1 expression on all major cDC subsets and moDCs in the MedLNs, but retained normal ICAM-1 expression by pDCs, accessory IFN-a producing DCs involved in Th1 and Tc1 differentiation processes (49), as well as by subcapsular sinus macrophages and B cells (Supplementary Figure 9B). Importantly, the negligible ICAM-2 levels expressed by cDCs were not elevated in our DC DICAM-1 mice (Supplementary Figure 9C). Notably, the ability of adoptively transferred naïve OT-II CD4 + T cells to accumulate and differentiate into GC Tfh cells was normal in the DC DICAM-1 mice ( Figure 5D) as opposed to their enhanced differentiation in the total ICAM knockout (ICAM-1/2 -/-) MedLNs ( Figure 5B). Collectively, these results suggest that in the absence of stromal ICAM expression (endothelial, fibroblastic and B cell), CD4 + effector T cells gave rise to elevated numbers of long lasting highly effective Tfh lymphocytes. These T cells provided critical help for the viral Ag-specific ICAM-1/2 -/-B-2 cells that entered the ICAM deficient MedLNs in response to viral infection. Consequently, the ICAM-1/2 -/influenza specific plasmablasts, produced normal serum levels of anti-PR8 IgGs after both primary infection and secondary challenge (Supplementary Figure 10). The levels of the main influenza clearing IgG subtype IgG2a (50), and the PR8 binding affinities of these polyclonal Abs recovered from the blood of ICAM-1/2 -/mice determined at high sera dilutions, were also normal ( Figure 5E). These surprising results suggest that Tdependent B cell immunity against the PR8 influenza virus does not require DC or B cell ICAM-1 and ICAM-2 expression, but is fine-tuned by stromal ICAMs, possibly expressed by fibroblastic reticular cells (FRCs) and follicular dendritic cells (FDCs). Thus, all major components of anti-viral humoral immunity are normally elicited in lung draining LNs deficient in ICAM-1 and ICAM-2 during a primary influenza infection, resulting in normal IgG dependent viral clearance.

CD8 + lymphocytes use ICAMs to enter MedLNs during influenza infection but undergo robust virus dependent differentiation into effectors in the absence of stromal and DC ICAMs
Virus specific CD8 + T cells contribute to efficient viral clearance by killing of virus infected cells and later on, generate tissue resident memory (T RM ) cells that can rapidly respond to a secondary heterosubtypic viral infection (51,52). In light of the enhanced differentiation of CD4 + T cells into Tfh cells taking place in PR8 infected ICAM-1/2 -/-MedLNs ( Figure 5B), we next asked whether virus specific naïve CD8 + T cells also undergo enhanced differentiation into virus specific CD8 + T effectors and T RM CD8 + cells following infection with PR8-SIINFEKL (OVA 257-264 peptide). As expected, the number of endogenous naïve CD8 + T cells harvested from ICAM-1/2 -/-MedLNs was significantly reduced compared with WT MedLNs ( Figure 6A). Furthermore, and reminiscent of the defective OT-II CD4 + T cell homing into MedLNs of ICAM-1/2 -/mice, the entry of OT-I CD8 + T cells into resting MedLNs of ICAM-1/2 -/mice was also markedly reduced compared to WT MedLNs ( Figures 6B, C). Although OT-I CD8 + T cells express comparable levels of the a4 integrins VLA-4 and a4b7 (Supplementary Figure 11) and HEVs of the MedLNs express both MadCAM-1 (53) and VCAM-1 (data not shown), under resting conditions, a4 integrin blocking of these T cells had no effect on their homing to either WT or to ICAM-1/2 -/-MedLNs ( Figure 6B). Nevertheless, following influenza infection, the homing of OT-I CD8 + T cells into ICAM-1/2 -/-MedLNs was only partially reduced ( Figure 6C; 3 rd vs. 4 th column), and the residual CD8 + OT-I T cell homing observed could be completely blocked by anti a4 integrin mAb treatment ( Figure 6C; 4 th vs. 5 th column). These results indicated that during infection, naïve CD8 + T cells can use both HEV expressed ICAMs and a4 integrin ligands like VCAM-1 and MadCAM-1 to enter MedLNs.
We next examined if viral Ag presenting ICAM deficient DCs can drive the proliferation and differentiation of the few OT-I CD8 + T cells accumulated in the ICAM-1/2 -/-MedLNs on day 4 p.i ( Figure 6D). In spite of the lack of ICAMs on all APCs and stromal cells of the MedLNs, the numbers of proliferating OT-I CD8 + T cells as well as their division rates were comparable in both WT and ICAM-1/2 -/-MedLNs ( Figures 6E, F). As expected, when OT-I CD8 + T cells were blocked from entering the infected ICAM-1/2 -/-MedLNs in the presence of a4 blocking mAb ( Figure 6D), no accumulation and generation of effector OT-I CD8 + T cells could be detected on day 4 p.i. (Figure 6E; 3 rd column). Nevertheless, once entering the ICAM-1/2 -/-MedLNs following infection and without a4 blocking, the accumulating OT-I CD8 + T cells underwent enhanced differentiation into effector IFN-g producing CD8 + T cells (Figures 6G-J) similar to the enhanced CD4 + T cell differentiation into Tfh cells observed in infected ICAM-1/2 -/mice ( Figure 5B). This IFN-g producing subset could evolve by homotypic interactions of the daughter ICAM-1 expressing OT-I CD8 + effector T cells which secrete IFN-g to each other (54).
In contrast to their altered homing and differentiation patterns in the total ICAM-1/2 -/mice, the number of OT-I CD8 + T cells entering and proliferating in the MedLNs of the DC DICAM-1 mice (Figures 6K, L) were similar to those found in the MedLNs of the WT mice ( Figure 6E). Furthermore, the differentiation of the adoptively transferred naïve OT-I CD8 + T cells into IFNg + effector T cells was not affected by ICAM-1 deficiency on all MedLN cDCs in the DC DICAM-1 mice ( Figure 6M). Collectively, these results indicated that DC ICAM-1 is not required for normal CD8 + T cell priming and differentiation triggered by influenza expressed antigens ( Figures 6L, M). Since ICAM-1 deficient DCs also lacked ICAM-2 expression (Supplementary 9C) these results further suggest that the presence of ICAMs on these essential APCs is dispensable for normal naïve CD8 + T cell priming and differentiation in the course of primary influenza lung infection.

Effector virus specific CTLs accumulate at reduced numbers in infected ICAM-1 and ICAM-2 deficient lungs but give rise to normal T RM CD8 + cells that protect from homo-and heterosubtypic infections
Effector CD8 + T cell lymphocytes migrate to the inflamed lung and clear virus infected airways (1) by killing infected epithelial cells (26) and by producing proinflammatory cytokines (e.g., IFN-g, TNF-a, IL-17 and CCR5 chemokines) that contribute to the recruitment and in situ activation of additional inflammatory DCs and pDCs at sites of viral infection (55). To follow the fate of OT-I CD8 + T cell effectors generated in the ICAM-1/2 -/lung draining LNs in response to PR8-SIINFEKL infection, we next determined the accumulation of effector OT-I CD8 + T cells in the infected lung ( Figure 7A). Notably, the accumulation of virus Ag specific OT-I CD8 + T cells in the ICAM-1/2 -/lungs was significantly reduced compared with WT lungs ( Figure 7B). The accumulation of polyclonal influenza specific CD8 + T cells (i.e., expressing the TCR Vb8.3 gene segment enriched in CD8 + TCRs specific for the immunodominant NP366-374/Db influenza epitope) (56) was also reduced ( Figure 7C).
Heterosubtypic immunity is defined as cross-protection to infection with an influenza A virus serotype other than the one used for primary infection (29). Effector CD8 + T cells receive local instructive signals that promote their subsequent differentiation into T RM cells (57), which are critical for first line protection from a heterosubtypic distinct influenza A virus challenge (58, 59). Successful heterosubtypic responses rely on Virus specific CTLs poorly accumulate in infected ICAM-1-and ICAM-2-deficient lungs but give rise to normal T RM CD8 + cells that protect from homo-and hetero-subtypic infections. the presence of these T RM cells which are specific for conserved antigenic moieties shared by non-identical influenza strains (60).
In the face of a heterosubtypic influenza A viral infection, the neutralizing antibodies generated by viral-specific B cells, can recognize only the specific influenza A viral surface proteins from a previous infection, while viral-specific CTLs target viral peptides derived from internal proteins of the virus conserved across different subtypes (61). We therefore next followed the fate of polyclonal effector CD8 + T cells residing in the ICAM-1/2 -/lungs 40 days post primary infection with the influenza H3N2 subtype A/HK/X31 strain. Strikingly, the number of polyclonal T RM cells generated following X31 influenza infection was similar in ICAM-1/2 -/and WT mice ( Figure 7D). These results confirmed that long lasting protective T RM CD8 + cells are generated independently of ICAM-1 and ICAM-2 (59). Consequently, when ICAM-1/2 -/mice were pre-infected with non-lethal doses of X31 influenza and then subjected to a secondary infection with a high dose of the heterosubstypic PR8 strain 40 days later, the ICAM-1/2 -/mice were fully protected with comparable efficacy to that of WT mice from the heterosubtypic influenza infection ( Figure 7E). These results indicated that long lasting protective T RM CD8 + cells are generated independently of vascular, epithelial or APC ICAMs. These results are also consistent with the lack of ICAM-1 expression by bronchial epithelia (22) and the downregulation of LFA-1 by interstitial T RM CD8 + cells at their early stages of differentiation from virus specific effector T cells accumulated early on in the airway mucosa (62). Finally, we asked whether polyclonal anti-viral Vb8.3 + /CD8 + T cells can be generated from memory CD8 + T cells also following a secondary homosubtypic PR8 influenza challenge of ICAM-1/2 -/mice ( Figure 7F). Notably, the numbers of Vb8.3 + /CD8 + effectors recovered in total lung suspensions of I.N. infected lungs subjected to a secondary I.T. homosubtypic PR8 influenza challenge were reduced in ICAM-1/2 -/lungs compared with WT lungs ( Figure 7G). Nevertheless, the numbers of Vb8.3 + /CD8 + effectors recovered in total lung suspensions of I.T. infected lungs, subjected to a secondary I.T. homosubtypic PR8 influenza challenge, were higher than in the primary I.N. infection and similar for both WT and ICAM-1/2 -/mice ( Figure 7H). These results suggest that the primary I.T. and I.N. infection routes induce either ICAM-1 independent or ICAM-1 dependent generation of Vb8.3 + /CD8 + effectors, respectively, following the recall I.T. infection. Furthermore, the generation and retention of T RM CD8 + cells following I.N. infection ( Figure 7I) was also normal in ICAM-1/2 -/lungs ( Figure 7J). In sharp contrast, the number of T RM CD4 + cells generated in similarly infected lungs was reduced in ICAM-1/2 -/lungs compared with infected WT lungs ( Figure 7K). Since this subset was reported to retain its LFA-1 expression (63), it could depend on lung ICAM expression for survival, possibly because these T cells use this integrin to reside in ICAM-1 rich areas like the alveoli. Taken together, our results suggest that although generated at lower rate inside ICAM-1/2 -/-MedLN following influenza infection, and poor accumulation of this subset in the infected lungs on day 7 post infection, effector CD8 + T cells successfully accumulate in the infected airway mucosa and give rise to T RM CD8 + cells ( Figure 5J), the major resident memory T cells that provide protection from secondary influenza challenges, in particular of heterosubtypic nature.

Discussion
Influenza virus (flu) is the most common respiratory infectious pathogen (64). Active influenza viruses uses their hemagglutinin (HA) envelope proteins to bind to specific sialic acid residues on epithelial cells in the upper and lower respiratory tract and subsequently gain entry into these cells and replicate (65,66). We have used the well-studied murine H1N1 influenza A virus to follow the role of ICAMs in this classical model of respiratory viral lung infection. The epithelial cells of the respiratory tract are the main source of productive influenza infection (67). Multiple redundant effector mechanisms of innate and adaptive cells contribute to clearance of this virus and are crucial for the establishment of long lasting immunity against recall infections with identical and related strains (68).
Leukocyte emigration from the blood to the influenzainfected respiratory tract occurs through three major types of blood vessels (1): the postcapillary venules along the bronchial tree (2), the alveolar capillaries in the lung parenchyma and (3) the HEVs of lung draining LNs and tertiary lymphoid tissues that are composed of specialized cuboidal postcapillary venules and serve as the main portal for the entry of circulating naïve and memory T and B lymphocytes into these organs (2). Influenza virus infection is initiated in the trachea and bronchial tree but can spread to alveolar lung regions and cause pneumonia (69). The endothelial lining of the alveolar capillaries wrapped around the numerous alveoli that comprise the lung parenchyma is structurally distinct from that of postcapillary venules (70). The capillary endothelia are elongated and flattened to improve gas exchange (2), demonstrate a lower permeability to solutes and fluids compared to the larger pulmonary blood vessels (70) and have smaller diameters than leukocytes, resulting in slow transit time. Consequently, leukocyte adhesion to and crossing of these blood vessels is drastically different from leukocyte adhesion and emigration from postcapillary venules. Indeed, the classic recruitment paradigm for these lung capillary vessels does not involve selectin-mediated tethering and rolling, and the role of leukocyte integrins is probably far more restricted than in post capillary lung venules (2). These capillaries contain a large number of marginated neutrophils, monocytes, as well as naïve, memory and effector T cells released from infected lung draining LNs (1) The role of integrin ligands, in general and of ICAMs, in particular in this leukocyte margination and extravasation across the lung capillaries has remained vague (4,17). Our results rule out that influenza infection elevates ICAM-1 expression on alveolar capillaries and suggests that ICAM-1 and ICAM-2 do not serve as the primary ligands for neutrophil margination at steady state conditions. Furthermore, during viral infection, the presence of ICAM-1 or of ICAM-2 on all lung capillaries is also not required for neutrophil or NK cell emigration into the alveolar space. ICAM-1 expression on alveolar epithelial cells is also not critical for leukocyte entry into the bronchoalveolar spaces during this viral infection. Thus, neutrophils localized inside the lung capillaries do not seem to use their LFA-1 integrin to bind endothelial ICAMs. Instead, they may use their Mac-1 integrin for margination through the capillaries (17). Inflammatory monocytes also seem to accumulate in the lung capillaries independently of endothelial ICAMs. In contrast, NK cells, patrolling monocytes and Tregs passing through the lung vasculature seem to use capillary ICAMs, consistent with expression of highly adhesive LFA-1 on the last two (23). It is therefore tempting to speculate that these ICAMs contribute to endothelial repair by patrolling monocytes and Tregs following influenza related injury. In addition, the expression of capillary ICAMs might be critical for the ability of effector T cells to interact with these blood vessels post infection where they encounter critical chemotactic exit signals into infected alveoli (22, 71).
Tracheal and peribronchial postcapillary venules and other blood vessels support massive leukocyte recruitment into the various airways following viral infection (72). In a previous study we found that at steady state peribronchical vessels express high levels of the key VLA-4 integrin ligand VCAM-1 as well as the low affinity LFA-1 integrin ligand ICAM-2 (22). In the present study, using intravital staining with distinct anti VCAM-1 and ICAM-1 mAbs, we have confirmed these results in resting lungs and extended them to influenza infected lungs, where we found a striking compartmentalization of VCAM-1 and ICAM-1 expression by peribronchial vs. alveolar capillaries, respectively. Canonical inflammatory cytokines like TNF-a, IL-1b and interferons, are secreted by virus-infected epithelial and stromal cells during various stages of influenza infection (2). Surprisingly, however, we did not find evidence for de novo expression of ICAM-1 on the surface of endothelial cells comprising the capillaries or the peribronchial vessels in influenza infected lungs. These staining results suggest that the entry of virus specific effector T cell egressing the MedLNs on their way to the bronchial mucosa during viral infection, likely involves VCAM-1 rather than ICAM-1 and ICAM-2 mediated adhesions. This subset of effector T cells that accumulate nearby the bronchioles and bronchi, once entering the bronchial mucosa, downregulates its LFA-1 within hours (62,73) and progressively differentiates into T RM cells by elevating CD103 (the integrin receptor for E-cadherin enriched in epithelial cell junctions), CD49a (a collagen type IV receptor) (58) and CD69 that negatively regulates the egress receptor S1P1 (73). Furthermore, in spite of considerable expression of VLA-4 on neutrophils and NK cells the VLA-4-VCAM-1 axis was not mandatory for neutrophil and NK leukocyte entry into similar bronchoalveolar compartments of influenza infected lungs in the absence of ICAMs. Thus, innate leukocyte trafficking through peribronchial blood vessels triggered by viral infection may involve alternative integrin-ligand pairs rather than ICAMs or VCAM-1. In addition, neutrophils and NK cells may emigrate through the pulmonary capillaries of infected alveoli also via ICAM-1, ICAM-2 and VCAM-1 independent interactions. One such interaction could involve the b2 integrin Mac-1 shared by neutrophils and NK cells. Mac-1 (CD11b) binds at least 30 alternative ligands in addition to ICAM-1 (74), including endothelial heparan sulfates, JAM-C and ECM ligands that can coat pulmonary vessels like fibrinogen and fibronectin (75). Thus, the direct involvement of Mac-1 in neutrophil and NK cell recruitment to influenza infected lungs remains to be investigated in future studies (76).
HEVs are the third class of specialized blood vessels (77) that drain lymphocytes into the lung draining LNs and into specialized tertiary lymphoid organs along the bronchial tree (78). These vessels serve as the main portal entry for naïve T and B lymphocytes, central memory lymphocytes and subsets of NK and innate lymphoid cells (ILCs) (79). In our influenza infection model, the major posterior MedLNs drain most of the migratory DCs, including the main viral-antigen cross presenting CD103 + cDC1 subset (80). In addition, cervical LNs drain parts of the upper airways and are also implicated in virus-specific T cell priming by influenza antigens (81). The LFA-1 integrin ligands ICAM-1 and ICAM-2 are critical for lymphocyte arrest on peripheral LN HEVs (82). Our study confirms the role of these two ligands in T and B cell entry to resting MedLNs but is the first to suggest that during the course of acute influenza infection both types of lymphocytes can use their a4 integrins to enter these LNs in the absence of HEV expressed ICAMs. Indeed, MedLN HEVs constitutively express the VLA-4 ligand VCAM-1 as well as MadCAM-1. The contribution of these ligands to ICAM independent lymphocyte entry is higher following infection, possibly due to upregulation of these ligands. In addition, whereas naïve T and B lymphocytes require the chemokine receptor CCR7 to activate their LFA-1, arrest on HEVs and cross these vessels, following infection, additional signals such as CXCR3 chemokines may provide critical a4 integrin activating signals that allow these lymphocytes to overcome the absence of ICAMs on MedLNs and use their a4 integrins to arrest on HEVs expressed VCAM-1 and MadCAM-1.
Both the cellular and humoral responses critical for influenza clearance depend on the proper presentation of viral antigens by respiratory viral antigen loaded DCs entering the MedLNs (83, 84). The migratory lung CD103 + cDC1 subset is the dominant lung DC population to transport influenza virus antigen to the MedLNs from primarily dying respiratory epithelial cells (85). In primary infections, the CD103 + cDC1 subset is crucial for efficient processing and presentation of viral antigens to both naïve CD8 + and CD4 + T cells in these LNs (84, 86). Another migratory DC subset, CD11b + cDC2, is involved mainly in CD4 + T cell priming into Th1 and Tfh effector lymphocytes (87). This subset is also capable of transferring viral antigens to resident cDC1 subsets which can in turn present these antigens to CD8 + T cells in the MedLNs (88). An asymmetrical distribution of these LN cDC1 and cDC2 subsets create regionalized activation of CD4 + and CD8 + T cells in the outer cortex and in the paracortex of the LNs, respectively (88). In addition, an influx of CD11b + cDC2 subset can also promote CD8 + T cell priming and differentiation (89). The interactions of lymphocyte LFA-1 with ICAM-1 expressed by these DCs, elevated upon infection, maturation and entry into MedLNs, were thought to be critical for functional immune synapses between both CD4 + and CD8 + T cells and antigen presenting DCs. Nevertheless, deleting ICAM-1 from these DCs in our DC DICAM-1 mice did not impair monoclonal OVA specific CD8 + or CD4 + T cell proliferation or differentiation to PR8 encoded OVA peptides in MedLNs following influenza virus infection. Since cDCs express negligible levels of the low affinity LFA-1 ligand, ICAM-2, our results suggest that CD4 + and CD8 + T cells do not use their LFA-1 for priming and differentiation by migratory and resident DCs that present viral antigens to these T cells. Interestingly, on day 4 post infection with PR8 encoding the OT-I cognate ligand, SIINFEKL, OT-I CD8 + T cells in the MedLNs of ICAM-1/2 -/mice proliferated and differentiated vigorously, whereas by day 7, their numbers declined in the ICAM-1/2 -/-MedLNs as well as in the ICAM-1/2 -/lungs. Since a similar reduction of CD8 + effector T cells was not observed in the MedLNs of our conditional DC DICAM-1 mice (data not shown) these results suggest that newly generated virus specific CD8 + effector T cells use pools of stromal ICAM-1 and ICAM-2 (e.g., FRCs) for survival (90), a possibility that will need to be studied in future studies using new ICAM deficient mice models.
The MedLNs are major sites of B-cell responses following influenza virus infection (44). Both innate-like B-1 cells and conventional B-2 cells (44, 91) provide critical humoral protection against primary influenza virus infections. The polyclonal IgM produced by B-1 cells can initially bind a variety of different influenza strains (92), reduce viral load, present viral particles to incoming B-2 cells and help initiate the stronger T-dependent B-2 cell dependent anti-viral responses in extrafollicular regions prior to GC generation necessary for affinity maturation and generation of long lasting B cell memory (44).
In response to type I IFN produced in the infected respiratory tract, innate-like B-1 cells enter the MedLNs via HEVs (92). We found that this critical entry step was not inhibited in the ICAM-1/2 -/-MedLNs, arguing for usage of VLA-4 and possibly of Mac-1 expressed by these lymphocytes. The accumulation and generation of extrafollicular B-2 cells were also normal in ICAM-1/2 -/-MedLNs, reflecting a potential usage of a4 integrins also by these lymphocytes in the entry into ICAM-1 and ICAM-2 deficient MedLNs. Furthermore, B-2 lymphocytes entering ICAM-1/2 -/-LNs combated the primary influenza infection as efficiently as in WT MedLNs, suggesting normal interaction of these lymphocytes with FDCs, the key antigen presenting cells that comprise the B follicles. During early stages post infection, FDCs are involved in virus Ag uptake, early B-2 activation and subsequent B cell migration towards the T zone border of the follicles (93). Our findings suggest that in the ICAM-1/2 -/-MedLNs of influenza infected mice, B-2 cells collected normal activation and differentiation signals from FDCs, possibly via their VCAM-1 (94), which is likely sufficient to overcome the absence of ICAMs in the follicles. Later on, GCs develop from these follicles within 10-12 days post influenza infection (44) and persist up to several months after influenza infection clearance in both the MedLNs and in tertiary lymphoid organs such as nasal associated lymphoid tissues (NALTs) (95) and bronchial associated lymphoid tissues (BALTs) (96). The role of ICAMs in the persistence and functions of these distinct GCs remains to be explored.
Our unexpected results indicate that B-2 cells acquired normal T cell help in the ICAM-1/2 -/-MedLNs. Indeed upon encounter and activation by viral antigen-presenting DCs, viralspecific CD4 + T cells differentiated into early Tfh CD4 + cells which stably express CXCR5 + , ICOS + , PD-1 + , and are capable of migrating to the T/B border (97). The expression of the key Tfh transcription factor, BCL6, by these early Tfh cells was reduced in the absence of DC and FDC ICAMs, suggesting that their early maturation is partially ICAM-1 and ICAM-2 dependent. Nevertheless, these specialized B helper T effector cells gave rise to normal GC reactions on day 12 post infection. The strength of the TCR dependent immune synapses generated between monoclonal Ag presenting B cells and cognate Tfh cells was recently shown to be augmented by B cell expressed ICAM-1 and ICAM-2 (45). Our findings suggest, however, that in the course of influenza virus infection, ICAM-1 and ICAM-2 deficient polyclonal B cells can still efficiently present antigenic peptides to Tfh cells. Furthermore, high affinity antibodies (50), were normally generated in MedLNs deficient in both DC and B cell ICAMs. In addition, both virus specific memory CD4 + T cells and memory B cells, likely persisted in ICAM-1/2 -/deficient MedLNs, since ICAM-1/2 -/mice were fully protected from secondary infections with lethal doses of homosubtypic influenza and generated similar levels of virus specific IgGs during these recall responses.
In conclusion, our study suggests an exceptional flexibility in the usage of vascular, stromal, and epithelial integrin ligands by innate and adaptive leukocytes recruited to influenza infected lungs. Our study also highlights multiple compensatory roles of alternative integrin ligands expressed by lung vascular cells and by APCs operating in lung draining lymph nodes in the priming and differentiation of virus specific T and B lymphocytes elicited during primary influenza infection. Although initial homing of these lymphocytes into lung draining lymph nodes is defective, specific cDC subsets in ICAM-1/2 -/mice appear hyperactivated and thereby compensate for the initially reduced accumulation of T and B cells in the lymph nodes of influenza-infected mice. The identification of these hyperactivated DC subsets in ICAM deficient lung draining lymph nodes eludes to DC suppressory mechanisms that depend on lung stromal ICAMs and affect cDC maturation and migration into lung draining lymph nodes during respiratory viral infections. New genetic models with multiple cell type specific deletions of ICAMs and other integrin ligands (e.g., VCAM-1) expressed by distinct stromal cells and cDC subsets could shed new light on where and how ICAMs and VCAM-1 regulate cDC maturation, lymphocyte differentiation, and lymphocyte memory generation in distinct viral infections. These new genetic models can also aid to new understanding of T cell education by ICAMs in the context of distinct lung tumors. These future studies may help in the design of improved vaccines against different pathogens and malignancies.

Data availability statement
The original contributions presented in the study are included in the article/Supplementary Materials, further inquiries can be directed to the corresponding author.

Ethics statement
The animal study was reviewed and approved by The Animal Care and Use Committee of the Weizmann Institute of Science.

Author contributions
SK designed and performed a major part of the in vivo experiments, analyzed the data, and contributed to the writing of the manuscript. OR, AS, MK and EP-K assisted with in vivo experiments. HA and NE produced purified PR8 for ELISA determinations, quantified the pulmonary PR8 viral load via qPRC, conducted the ELISA for IgG antibody detection against PR8 and assisted with the detection of IgG2a. JE performed IF staining of MedLNs. OR and MK assisted with LSM imaging and data analysis. YA assisted with data acquisition of LSM. FC imaged neutrophils inside lung capillaries by pulmonary intravital imaging. PK supervised the pulmonary intravital imaging experiments. NG provided the various PR8 variants and advised on several experimental directions. RA supervised all the experiments and wrote the manuscript. All authors contributed to the article and approved the submitted version.