Macrophage-B Cell Interactions in the Inverted Porcine Lymph Node and Their Response to Porcine Reproductive and Respiratory Syndrome Virus

Swine lymph nodes (LN) present an inverted structure compared to mouse and human, with the afferent lymph diffusing from the center to the periphery. This structure, also observed in close and distant species such as dolphins, hippopotamus, rhinoceros, and elephants, is poorly described, nor are the LN macrophage populations and their relationship with B cell follicles. B cell maturation occurs mainly in LN B cell follicles with the help of LN macrophage populations endowed with different antigen delivery capacities. We identified three macrophage populations that we localized in the inverted LN spatial organization. This allowed us to ascribe porcine LN MΦ to their murine counterparts: subcapsular sinus MΦ, medullary cord MΦ and medullary sinus MΦ. We identified the different intra and extrafollicular stages of LN B cells maturation and explored the interaction of MΦ, drained antigen and follicular B cells. The porcine reproductive and respiratory syndrome virus (PRRSV) is a major porcine pathogen that infects tissue macrophages (MΦ). PRRSV is persistent in the secondary lymphoid tissues and induces a delay in neutralizing antibodies appearance. We observed PRRSV interaction with two LN MΦ populations, of which one interacts closely with centroblasts. We observed BCL6 up-regulation in centroblast upon PRRSV infection, leading to new hypothesis on PRRSV inhibition of B cell maturation. This seminal study of porcine LN will permit fruitful comparison with murine and human LN for a better understanding of normal and inverted LN development and functioning.

Swine lymph nodes (LN) present an inverted structure compared to mouse and human, with the afferent lymph diffusing from the center to the periphery. This structure, also observed in close and distant species such as dolphins, hippopotamus, rhinoceros, and elephants, is poorly described, nor are the LN macrophage populations and their relationship with B cell follicles. B cell maturation occurs mainly in LN B cell follicles with the help of LN macrophage populations endowed with different antigen delivery capacities. We identified three macrophage populations that we localized in the inverted LN spatial organization. This allowed us to ascribe porcine LN M to their murine counterparts: subcapsular sinus M , medullary cord M and medullary sinus M . We identified the different intra and extrafollicular stages of LN B cells maturation and explored the interaction of M , drained antigen and follicular B cells. The porcine reproductive and respiratory syndrome virus (PRRSV) is a major porcine pathogen that infects tissue macrophages (M ). PRRSV is persistent in the secondary lymphoid tissues and induces a delay in neutralizing antibodies appearance. We observed PRRSV interaction with two LN M populations, of which one interacts closely with centroblasts. We observed BCL6 up-regulation in centroblast upon PRRSV infection, leading to new hypothesis on PRRSV inhibition of B cell maturation. This seminal study of porcine LN will permit fruitful comparison with murine and human LN for a better understanding of normal and inverted LN development and functioning.

INTRODUCTION
In most species, free antigens and DC migrating from the peripheral tissues enter the lymph node (LN) through the afferent lymphatic vessels into the external capsular sinus. Naïve lymphocytes enter the LN from the blood through the high endothelial venules (HEV) and then patrol the B and T cell areas. Upon antigen encounter in the context of the adequate antigen presenting cells, T and B lymphocytes get activated, mature and exit the LN through the medullary sinus and the efferent lymphatic vessel. Thus, most mammals, including mice and humans, possess LN with a centripetal lymphatic motion, i.e., afferent lymph enters the LN through the peripheral capsule and finds its way out toward the LN central hilum and the efferent lymph.
Conversely, in pigs (1), like in some species belonging to the Laurasiatheria superorder such as dolphins, hippopotamus (2), and rhinoceros (3), as well as in elephant (4), lymph presents a centrifuged motion. The porcine afferent lymphatic vessels enter the capsule at one site and penetrate deep into the area occupied by the B follicles and the T cells. Then they join the trabecular sinuses and filters into the subcapsular sinus from which efferent vessels originate (5). Naïve lymphocytes entered the LN through HEV as in other mammalian species, however, after having scanned the B and T cell areas, they exit directly in the blood through the same HEV (6).
In mouse, five populations of LN M have been identified [for review (7,8)]. The subcapsular sinus M (SCS M ) (CD169 pos /F4/80 neg ) transfer the antigens from the subcapsular space into the B cell follicle. SCS M have been demonstrated as mandatory for mounting efficient cytotoxic (9) and humoral immune (10) responses. In the follicle, tangible body M (TBM) scavenge the dead B lymphocytes whereas T cell zone M (TZM) might do the same for T lymphocytes. The medullary cord M (MCM) have a role in the plasma cells terminal maturation (11) and medullary sinus M (MSM), situated at the exit of the LN would be involved in the final clearance of lymph borne particles.
Porcine reproductive and respiratory syndrome (PRRS) is a disease induced by the PRRS virus (PRRSV), a positive single stranded RNA virus from the Arterivirade family within the Nidovirales order (12). After oronasal transmission, PRRSV colonizes the respiratory tract and could play an immunomodulatory role delaying and weakening host responses, finally leading to virus persistence. Although anti-PRRSV antibodies are detected in the serum as early as one-week postinfection, the antibody serum titers to several viral proteins decline over time despite the continuous presence of the virus (13). Moreover, the emergence of neutralizing antibodies is strongly delayed, up to several months. Such delay has been proposed to be the main reason for PRRSV escape to the immune response [for review see (14)].
PRRSV strongly impacts the swine industry due to reproductive failures, reduced weight gain and predisposition to super-infections (15). The two main PRRSV cellular receptors are CD169/Sialoadhesin that allows the binding of the virus and CD163 which is essential for the release of the viral genome in the cytosol [for review see (16)]. PRRSV cellular targets are cells from the monocytic lineage, among them so far, only alveolar macrophages (M ) (17)(18)(19), pulmonary intravascular M (20, 21) and CD163-positive tonsil macrophages (22) have been shown to be actually infected in vivo. We and others recently showed that in vivo, primary dendritic cells from lung (23) and tonsil (22) were not infected by the virus.
During the persistence phase, the virus is no longer detected in lung but could be isolated from LN up to 5 months post infection (24). Although virus are detected in secondary lymphoid tissues such as tonsils and lymph nodes, and at a lesser level spleen (25,26), its target(s) in these tissues have not been precisely identified. One team studied the porcine LN M for the sake of vaccine targeting to CD163 or CD169 (27, 28) but without making a distinction between different M subpopulations. A study on the porcine respiratory tract draining LN upon PRRSV infection has been recently published (26). Using microdissection of follicular and interfollicular areas, the authors detected similar viral loads in both regions. The PRRSV ability to replicate in the respiratory draining LN for more than a month (25,26) and the parallel delay in the appearance of neutralizing antibodies suggest that LN-M infection by PRRSV might directly influence the B cell maturation process.
Herein, using whole LN imaging, flow cytometry analysis, cell sorting and RT-qPCR, we first described the localization and phenotype of three different M subpopulations of the pig respiratory LN and assign each of them to their likely mouse counterpart. We described the different B cell maturation stages according to their expression of key transcription factors of B cell differentiation and their follicular localization. Subsequently, in vivo PRRSV infections were performed in order to study the susceptibility to infection of previously identified cells and to tentatively get information on how PRRSV infection may impacts the B cell maturation process.

In vivo Infections
Two different strains of the European originated PRRSV1 species were used: the PRSSV1.1 emergent Flanders13 (Fl13) strain (25) and the PRRSV1.3 highly pathogenic Lena strain (29). For in vivo experiments, PRRSV infections were performed at INRA PFIE (Nouzilly, France) for FL13 and ANSES (Ploufragan, France) for Lena infections. The animal experiments were authorized by the French Ministry for Research (authorization no.2015051418327338 and no.2015060113297443, respectively) and approved by the national ethics committee (authorizations no.09/07/13-1 and no.07/07/15-3). Ten-week-old Large White piglets were tested PRRSV free and inoculated intranasally at 5.10 5 TCID50/animal or mock inoculated. For FL13, 3 pigs were used per group and euthanized 5 days post infection (dpi). For Lena, 4 pigs were used per group and euthanized at 10 dpi. Tracheobronchial LN were collected and processed as described above. slaughterhouse (Houdan, France) and from the controlled UE-PAO-INRA (Nouzilly, France) herd. Tissues were minced and incubated with RPMI 1640 supplemented with 100 IU/ml penicillin, 100 mg/ml streptomycin, 250 ng/ml Fungizone R (Antibiotic-Antimycotic 100X ThermoFisher Scientific, Illkirch, France), 2 mM L-glutamine and 10% inactivated fetal calf serum (FCS, Invitrogen, Paisley, UK). Digestion were performed for 30 min at 37 • C with 2 mg/ml collagenase D (Roche, Meylan, France), 1 mg/ml dispase (Invitrogen) and 0.1 mg/ml DNase I (Roche). Filtration on 40 µm cell strainers were performed and red blood cells were lysed using erythrolysis buffer (10 mM NaHCO3, 155 mM NH4Cl, and 10 mM EDTA). Cells were washed in PBS/EDTA and further processed fresh for flow cytometry staining and sorting as much as possible. Alternatively, LN cells were frozen in FCS + 10% dimethyl sulfoxide.

Immunohistochemical Staining
Tracheobronchial LN were frozen in Tissue Teck (Sakura, Paris, France) and conserved at −80 • C. Sections of 14 µm were obtained using a cryostat (Leica CM3050S, Nanterre, France) and deposed on Superfrost R glass slides (ThermoFisher scientific). Cryosections were fixed in methanol/acetone (1:1) at −20 • C for 20 min. Fixed slides were saturated using PBS supplemented with 5% horse serum (HS) and 5% swine serum (SS) for 30 min at room temperature (RT). When biotinylated antibodies were used, a specific step of endogenous biotins saturation using Avidin/Biotin Blocking Kit (Invitrogen) was added. Primary and secondary antibodies ( Table 1) were added at 4 • C overnight or 30 min, respectively.
For beads draining experiments, tracheobronchial explants were injected at multiple sites in proximity of the target LN with 0.1 µm red fluorescent beads (FluoSpheres TM Carboxylate-Modified Microspheres, 0.1 µm, red fluorescent (580/605), 2% solids, ThermoFisher Scientific) diluted 1/4 in physiological serum. Explants were incubated 30 min à 37 • C to allow the drainage of beads into the LN. The targeted LN was then sampled and frozen in Tissue Teck and processed as above for immunohistochemical staining, except for the fixation, in 4% PFA, 15 min at room temperature, in order to avoid bleaching

Whole LN Staining and Clearing
Tracheobronchial LN was fixed by overnight incubation in paraformaldehyde 4% at 4 • C. Immunohistochemical staining and clearing were performed following the iDISCO+ protocol (30). The sample was acquired on a light-sheet ultramicroscope (LaVision Biotec, Bielefeld, Germany) with a 2x objective. Whole LN and isolated follicles were acquired using 0.63X and 4X zoom factor, respectively. LN was reconstructed in 3D using Imaris 9.1 (Bitplane). The number of follicles was manually quantified.

Flow Cytometry Analysis
Cells were stained in blocking solution, composed of PBS-EDTA (2 mM) supplemented with 5% HS and 5% SS. Staining were made in 4 steps, including PBS/EDTA with 2% FCS washing between each step: Uncoupled primary anti-CD169, anti-IgM and anti-CD172a antibodies were added to the blocking solution for 30 min on ice and then washed. Fluorescent, secondary, mouse isotype specific antibodies were then added, respectively anti-IgG2a-PE-Cy7, anti-IgM-Alexa647 and anti-IgG2b-APC-Cy7 for 20 min on ice and then washed. Then, to saturate the potential unbound IgG1-directed secondary antibodies sites, we incubated cells 30 min with isotype-control IgG1 (10 µg/ml) in blocking solution. Third the fluorochromcoupled or biotinylated primary IgG1: anti-CD21 coupled to BV510, anti-CD163 coupled to PE and biotinylated anti-CD2 antibodies were added for 30 min on ice and then washed. Finally, streptavidin-coupled to Alexa700 was added for 20 min on ice and washed before resuspension in DAPI containing buffer for sorting ( Table 1).

Cell Sorting
LN cells were stained as for flow cytometry analysis and the different populations were sorted. Since LN cells are fragile the sorting was carried out on fresh cells to maximize cell yield and viability. Dead cells were excluded by DAPI staining (Sigma-Aldrich). Cells were sorted using a Moflo Astrios sorter (Beckman-Coulter, Paris, France) driven by summit 6.2. Puraflow 1X was used as sheath and run at a constant pressure of 25 PSI. FlowJo software (version X.1.0, Tree Star, Ashland, OR, USA) was used for analysis.

May-Grünwald-Giemsa Staining
Sorted cells were fixed and stained with May-Grünwald-Giemsa methods as previously described (31). Images were acquired and analyzed with a Pannoramic Scan II (3DHISTECH Ltd., Budapest, Hungary).

RNA Extraction, Reverse Transcription, and Real-Time qPCR
Total RNAs from sorted cell populations were extracted using the Arcturus PicoPure RNA Isolation kit (ThermoFisher Scientific, St Aubin, France) according to the manufacturer's instructions. The Qiagen RNase-Free DNase Set (Courtaboeuf, France) was used to remove contaminating genomic DNA. Reverse transcription (RT) was made using Multiscribe reverse transcriptase (ThermoFisher Scientific) according to the manufacturer's instructions and qPCR carried out using the iTaq Universal SYBR Green Supermix (Biorad, Hercules, CA). Ribosomal protein S24 (RPS24) was chosen as reference gene as previously described in pig lung (31,32). Primers used in this publication are reported in Supplementary Table 1. Using the porcine new born pig trachea epithelial cell line NPTR (33) we validated Topoisomerase IIA and Cyclin B2 as transcriptomic markers of cell proliferation, since their expression by RT-qPCR were respectively 30 and 17 times higher in exponential proliferation cultures compared with confluent NPTR cells. Conversely Ki67, which is a good proliferation marker at protein level, was invalidated for transcriptomic studies since its expression only varied by a factor of 2 between proliferating and confluent cells (Supplementary Figure 1a).

Statistical Analysis
All data were analyzed using GraphPad Prism. The unpaired, non-parametric Mann-Whitney Statistical test was used.

Macrophages Identification and Localization in the Tracheobronchial LN by CD169, CD163, and CD21 Microscopic Staining
We first identified swine LN M according to their LN localizations with regard to B and T cell areas. The anti-CD21b antibody B-Ly4 mainly stains immature pre-classswitched B cells (34) and was used to localize B cell follicles. Anti-CD8α expressed on CD8 T cells and memory/activated CD4 T cells in swine (35) was used to identify the T cell areas (Figures 1a,c). As previously observed in swine (26), T cell areas are diffused, whereas B cell follicles are precisely delineated. CD169 and CD163 markers previously used to identify swine LN M (27) and mouse LN M were used [for review (8)]. As expected LN from conventionally reared animals (Houdan slaughterhouse) frequently presented several well-developed B cell follicles, whereas control-reared animals from INRA heard presented fewer and smaller follicles (data not shown). We observed CD169 expression at the periphery of the LN (Figure 1a) in an extended subcapsular location (Figure 1b). These CD169 subcapsular staining colocalized with CD163 staining (Figures 1a,c,d). Because of the reverse structure of the pig LN, these peripheral cells are localized next to the efferent vessels. Interestingly, CD169 was also expressed in close association with the B cell follicles (Figure 1a). CD169 pos /CD163 neg /CD21 neg cells were situated on a thin band at the periphery of the CD21-positive follicles (Figure 1a, green perifollicular crescent) whereas CD169 pos /CD163 neg /CD21 pos cells were identified in a more diffuse "crescent shape" intrafollicular area (Figure 1a, yellow intra-follicular staining). Finally, CD169 neg /CD163 pos cells were present in the medulla, tightly associated with the lymphatic cords, both in the LN parenchyma as well as in the sinus (Figure 1d, red staining).

Three Different M Cell Types Can Be Identified in the Tracheobronchial LN
A flow cytometry analysis of enzyme-digested LN was performed to isolate the 4 populations identified using microscopy. The 4 populations were distinguished using a CD163/CD169/CD21 and FSC gating and sorted as described in Figure 2A, followed by MGG staining (Figure 2B) or by RT-qPCR ( Figure 2C). CD163 pos /CD169 pos cells presented a rounded or slightly indented nucleus with coarse chromatin and abundant clear vacuolated cytoplasm, in agreement with a macrophagic phenotype ( Figure 2B). Moreover, RT-qPCR analyses revealed that these cells expressed the macrophagic CSF1R, MAFB, and MerTK genes ( Figure 2C). Because of the reverse structure of the pig LN, these cells located at the periphery are next to the efferent vessels. Thus, we called them efferent M (effM ). CD163 pos /CD169 neg cells displayed abundant clear vacuolated cytoplasm and a large deeply indented nucleus with lacy chromatin similar to blood monocytes and to the MCM from murine LN (11). They expressed similar levels of CSF1R and MerTK than effM but higher levels of MAFB ( Figure 2C). According to their association with the cord vessels, these cells were called cord M (cordM ).
CD163 neg /CD169 pos /CD21 neg cells displayed a blue-gray cytoplasm with a grainy texture surrounding a large vesicular or slightly indented nucleus with coarsely clumped chromatin ( Figure 2B). These cells expressed CSF1R, MAFB, and MerTK genes ( Figure 2C). According to their morphology and their localization at the periphery of the follicle, these cells were called perifollicular M (PFM ). CD163 neg /CD169 pos /CD21 pos cells displayed a high nucleo-cytoplasmic ratio with a large round nucleus with coarse and dense chromatin and a nucleolus surrounded by a perinuclear halo and a rim of basophilic cytoplasm, looking very similar to porcine bone marrow preB cells (36), in strong support with an immature B cell identity. Moreover, their localization in the periphery of the follicle (Figures 1a,d), reminds the dark zone localization of centroblasts in mouse and human [for review see (37)]. In agreement with their immature B cells microscopic profile, these cells expressed CD19, a component of the B cell receptor complex, and PAX5, a regulatory gene involved in the somatic hypermutation process (38) (Figure 2D) but none of the macrophagic markers tested (Figure 2C). The cells were subsequently named CD169 pos B cells. Although CD169 protein expression was detected by FACS (Figure 2A) and microscopy (Figures 1a,d), CD169 pos B cells did not express detectable levels of CD169 mRNA ( Figure 2D). Conversely, CD169 mRNA quantification in effM (CD169 pos ), cordM (CD169 neg ) and PFM (CD169 pos ), was in agreement with cytometry and microscopy data (Figure 2D).

PFM Cells Cap the B Cell Follicles and Interact With Intrafollicular CD169 pos B Cells
To investigate the relationship between the PFM and the B cell follicles we proceeded to whole LN transparisation Frontiers in Immunology | www.frontiersin.org  (Figure 3a), combined with CD169/CD21 immunostaining and imaging using fluorescent light sheet microscope. This allowed to reconstitute the whole LN 3D structure (Figure 3b and Supplementary Movie 1). This tracheobronchial LN, sampled from a conventionally reared animal, contained 539 B cell follicles (Supplementary Movie 2), each of them being individually associated with a PFM area (Supplementary Movie 1). The afferent central entry appeared composed of a collection of smaller afferent vessels which separated from each other while entering deeper in the LN parenchyma (Supplementary Movie 1). One of this afferent vessel and its continuous sinus was tracked (in red, Figure 3c and Supplementary Movie 2). The sinus appeared interconnected with other sinuses along its parenchymal trail to finally pour in the efferent subcapsular sinus.
The 3D rendering of follicles acquired at higher resolution delineated a PFM area forming a semi-spherical structure that cap the B cell follicle on one of its sides (Supplementary Movie 3).
To identify the M taking in charge particulate antigens drained through afferent lymphatics, we injected ex vivo red fluorescent 0.1 µm beads in the tissue surrounding a tracheobronchial LN. Injected beads were allowed to drain for 30 min at 37 • C. The draining LN was then sampled and immunostained for CD169 and CD21 expressions. Beads were present in effM and PFM but not in cordM (Figure 3d). Zooming on B cell follicles, beads were mostly associated with PFM although some signal was observed deeper in the follicle associated with CD169 pos B cells (Figure 3e). By referring in DAPI staining, PFM appeared clearly situated in the space between the follicle and the LN parenchyma, reminiscent of mouse LN subcapsular sinus space (Figures 3f,g). To precisely set the limit between the CD21 pos cells and the PFM , a view of the interface between intrafollicular B cells and PFM is depicted in Figure 3h with saturated CD21 staining. Once this limit affixed on the CD169/CD21 co-staining (Figure 3i), several beads appeared situated at the contact between PFM protrusions and intrafollicular CD169 pos B cells (Figure 3i). Thus, PFM and CD169 pos B cells closely interact with antigen drained from the peripheral tissue, at the frontier of the B cell follicle.

Identification of Five LN B Cell Differentiation/Activation Stages
The CD21, IgM, and CD2 markers, previously proposed by Sinkora et al. (39) for porcine bone marrow B cell development analysis, were used to better integrate CD169 pos B cells across the LN B cell differentiation steps. The CD169 pos /CD21 pos follicular B cells (Figure 4A, purple cells) did not express IgM and were mixed with CD21 pos /CD169 neg /IgM neg B cells (red cells) in the intrafollicular area contiguous with PFM (blue cells). IgM pos /CD21 pos cells (yellow cells) were localized in the center of the follicle. Finally, IgM pos /CD21 neg cells (green cells) were rarely present in the center of the follicles, but well represented in extrafollicular area, as well as in the LN periphery, in direct contact with effM ( Figure 4A). Thus, this first overview of LN B cells phenotypic localization is in agreement with CD21 pos /IgM neg cells (CD169 positive and negative) being dark zone-localized centroblasts, IgM pos /CD21 pos cells being light zone centrocytes and plasmablasts and IgM pos /CD21 neg cells being mostly extrafollicular mature plasma cells.
We completed this microscopic study by FACS analysis using the panel utilized for M identifications, which was complemented with the myeloid marker Sirpa/CD172a and the B cell markers CD21, IgM and CD2 (Figure 4B). The different cell-types were sorted as described in Figure 4B and stained using MGG coloration (Figure 4C). Like CD169 pos B cells, centroblasts displayed a large, round nucleus with multiple peripheral nucleoli surrounded by a rim of basophilic cytoplasm. Centrocytes displayed a small, round nucleus with clumped chromatin and with scant cytoplasm. Plasmablasts displayed some heterogeneity; their size was intermediate to large with scant to moderate slightly basophilic cytoplasm and a vesicular nucleus with reticular chromatin. Plasma cells demonstrated an intermediate-sized nucleus with reticular chromatin and a rim or just a crescent of clear cytoplasm with a prominent juxtanuclear archoplasm.
Using Ki67 staining we identified proliferating B cells that appeared mostly localized in the follicle area contiguous to CD169 pos PFM , in agreement with the phenotype and dark zone-localization of CD169 pos B cells and centroblasts ( Figure 4D).
The sorted B cells were characterized by analyzing B cell-transcription factors genes expression by RT-qPCR (40) [for review see (37)]. CD169 pos B cells and centroblasts expressed significantly more BCL-6 ( Figure 5) than all the other B cells. BCL-6 is a master transcription factor expressed specifically in centroblasts (40) and involved in the inhibition Frontiers in Immunology | www.frontiersin.org of their differentiation into plasma and memory B cells (40,41). Centrocytes expressed median levels of BCL-6, PAX5, and IRF4 as expected for these cells at an intermediate differentiation stage (Figure 5). As expected, plasmablasts and plasma cells expressed the lowest levels of BCL-6 and PAX5, whereas plasma cells expressed the highest levels of IRF4 (42), XBP1 (43,44), and Blimp1 (45,46) which are known markers of terminally differentiated B cells (Figure 5). CD169 pos B cells and centroblasts (CD21 pos /IgM neg ) expressed the proliferation markers Topoisomerase IIA ( Figure 5) and Cyclin B2 (Supplementary Figure 1b) at significantly higher levels than all the other B cells, which is consistent with the expression of Ki67 in the follicular area contiguous with PFM .
In conclusion, the phenotype, the localization, the expression of commonly used transcription factors and the proliferation status consistently defined five B cell maturation stages. To note, we confirmed here that the CD169 pos B cells were strongly related to centroblasts, and might be an early step of this differentiation stage.

In vivo PRRSV Interact With EffM and PFM
In vivo infections were performed to explore the mechanisms used by PRRSV to persist in the porcine LN, and its impact on the follicular B cell maturation process.
In a first experiment, 4 animals were infected with Lena PRRSV and tracheobronchial LN were collected 10 dpi. At the time of this preliminary experiment we did not yet distinguish PFM from CD169 pos B cells, we thus sorted one single CD169 pos /CD163 neg population. We also sorted LN cDC2 as a myeloid negative control of infection (23) as described in Supplementary Figure 2a. The sorting strategy was validated by RT-qPCR of key specific genes (Supplementary Figure 2b). As previously validated (23), we measured cell-associated PRRSV by RT-qPCR on viral RNA normalized using RPS24 reference gene (2 − Ct ). PRRSV RNA was detected in effM (10 dpi) (Supplementary Figure 2c) in agreement with effM population decrease proportion among total LN live cells (Supplementary Figure 2d), whereas a weaker expression was detected in the pooled PFM /CD169 pos B cells (Supplementary Figure 2c), with no significant decrease of this last composite population (Supplementary Figure 2d).
Realizing that the CD163 neg /CD169 pos population was composed of mixed B and macrophagic cells we then performed a second in vivo infection in which we discriminated PFM from CD169 pos B cells. Moreover, to ascertain that our first results could be extended to different strains and times post-infection, we then used Flanders 13 strain and sacrificed the animals 5 dpi. Three infected animals were sacrificed, tracheobronchial LN cells were isolated and analyzed using the same gating as in Figure 4B. Upon infection, a proportional decrease of effM and

DISCUSSION
In mammalians, the lymph node is the control center of the adaptive immune response, that's why efforts have been developed recently in the mouse model to better describe its structure/function relationships, leading to a better understanding of LN cells complex relationships, in steady states and infectious situations (47)(48)(49)(50). Although several swine pathogens, among them PRRSV, are known both to persist in the LN and to strongly delay the onset of an effective immune response, the structure of the porcine lymph node and its striking inverted peculiarity had never been precisely investigated.
Herein, we first explored the inverted structure of the pig LN according to LN M populations and B cell maturation stages. The effM population had a subcapsular localization at the exit of the LN, whereas human and mice SCS M are positioned at the entrance of the LN. Because of this distinction, pig subcapsular M are expected to be endowed with different functions compared to mouse and human SCS M . Due to their localization before the efferent lymphatic, we considered them as the likely functional equivalent of the mouse MSM, localized in the medullary sinus, just before the efferent lymphatic vessel. Indeed their phenotype (CD169 pos /CD163 pos /CD21 pos ) corresponds to the CD169 pos /CD163 pos phenotype known for MSM (8) rather than to the phenotype of murine SCS (CD169 pos /CD163 neg ) (51) ( Table 2). In mouse, these MSM have been proposed to play a role in the clearance of the lymphatic fluid before its exit to the main blood circulation (8). CordM are clearly positioned along the medullary cords with a phenotype similar to the mouse MCM (CD163 pos / CD169 neg ). Finally, PFM appeared to be the equivalent of murine SCS M ( Table 2). They presented a CD169 pos /CD163 neg /CD21 neg phenotype compatible with mouse SCS M and with the porcine CD169 pos spleen M (27), which are the likely counterpart of metallophilic marginal zone M , i.e., the spleen equivalent of SCS (52). We brought here evidences that PFM are positioned at the interface between afferent lymph and B cell follicles and

Main function
Lymph clearance before exit / that they are endowed with the capacity to bind and transfer antigens from the sinus to the intrafollicular compartment as described for murine SCS M (10, 53-55). We did not find evidence neither of tangible body M , nor of T cell zone M , probably because of their absence of expression of CD169 and CD163, the two markers we used here for porcine LN M identification. We observed a CD169 pos B cell population positioned inside the B cell follicles at the contact with PFM . Except their exhibition of CD169, these cells are CD21 pos /IgM neg /CD2 neg . They are proliferating and their MGG phenotype as well as transcription factors expression (BCL-6 and PAX5) are characteristic of the centroblastic stage. Interestingly these cells exhibit CD169 proteins at their surface without detectable mRNA. This discrepancy between mRNA and protein expression associated with their observed close interaction with PFM supports the acquisition of CD169 molecules by intrafollicular CD21 pos centroblasts through trogocytosis upon intimate contact with CD169-expressing PFM . This membrane material exchange gives credit to the possibility of antigen transfer from PFM to CD169 pos B cells. Indeed, the B cell interaction with antigen-bearing PFM in the porcine lymph node appears similar to the previously described murine SCS M /B cells interaction in their relay on antigen transport from the capsule to the follicle (10,53,54). Interestingly, in pig, these antigentransporting B cells belong to the centroblast stage, whereas in mouse, naïve B cells are supposed to fulfill this task (55). Centroblasts are the cells defining the dark zone, this means that, differently from mouse SCS M which are in contact with the centrocyte-occupied light zone, porcine PFM are in contact with the centroblast-occupied dark zone. It remains to be explored how this striking difference may affect the follicular functions of the porcine LN. After profiling the steady state of LN cells, we analyzed the different M and B cell populations upon PRRSV infection. According to their PRRSV RNA level associated with their depletion upon infection, effM are likely productively infected by both FL13 and Lena viruses. Because of effM localization, and since their depletion might preclude them to play their role of efferent lymph scavenger, this infection might increase the dumping of viral particles from the LN to the blood circulation, thus contributing to the viremia. Compared to effM , PCM and cDC2 did not present consistent PRRSV RNA levels. Interestingly, PFM presented comparable viral RNA load as effM , whereas PFM did not express the major PRRSV entry receptor CD163. For several viruses (53), it has been described that SCS (the bona fide PFM murine counterpart) are endowed with the capacity to capture lymph-borne viruses for presentation to follicular B cells, being infected or not. Unfortunately, the scarcity and the fragility of PFM did not allow us to unambiguously test their capacity to replicate PRRSV in vitro. However, the absence of CD163 expression associated to the fact that PFM did not experience depletion upon in vivo infection are in agreement with the occurrence of a viral interaction without productive infection.
Concerning the modulation by PRRSV of cytokine expressions related to B cell homeostasis, BAFF was found to be upregulated in LN M upon infection whereas levels of IL21 and IL10 were unchanged. Although limited, this information is in agreement with an adequate survival environment for activated B cells. In the B cell compartment, only the CD169 pos B cells, in close contact with PFM , presented increase of BCL-6 expression and proliferation. Although it is difficult to interpret these data without a comparison with another respiratory virus triggering a normal neutralizing antibody response such as influenza, the strict limitation of BCL-6 upregulation and proliferation in CD169 pos B cells led us to hypothesize a blocking of early centroblasts at this immature low affinity state. Interestingly, in germ free piglets PRRSV infection triggers a non-antigen selective expansion of B cells in all immunoglobulin classes, suggestive of a PRRSV-induced defect in the B cells maturation process (56). This blocking might contribute to the PRRSV-induced defect in the B cell maturation process observed by Sinkora et al. (56), linked with the production of hydrophobic binding sites autoimmune-like antibodies (57). These observations might be related to similar B cell maturation defects (58) as well as low-affinity autoimmune antibodies appearance (59) observed in HIV infections, another single strand positive RNA virus, in which the gp120 virus envelope protein act directly on B cell membrane proteins to trigger a polyclonal, non-protective B cell response (60). In conclusion, we observed in porcine tracheobronchial LN a strong PRRSV signal associated with depletion of the effM , in agreement with a productive infection that can participate to the viremia. Conversely, interaction of PRRSV with PFM might favor a direct or indirect action of the virus on CD169 pos B cells, leading to their blocking in a BCL6 high state.
The fine description of the inverted porcine LN M and B cell differentiation steps will open the possibility to visualize the action of PRRSV and other porcine viruses, such as the African swine fever virus, the classical swine fever virus and the porcine circovirus type 2 in this main place of the adaptive immune response initiation. Finally, an in depth understanding of inverted porcine LN structure may also allow to look from a new point of view on the "normal" structure of human and mice lymph nodes and may help veterinarians to better understand the immunology of endangered species such as rhinoceros and dolphin.

ETHICS STATEMENT
The animal experiments were authorized by the French Ministry for Research (authorization no.2015051418327338 and no.2015060113297443 respectively) and approved by the national ethics committee (authorizations no.09/07/13-1 and no.07/07/15-3).

AUTHOR CONTRIBUTIONS
ElB, EC, EdB, and NB processed the samples and performed the in vitro and ex vivo experiments. MF and CL performed the transparisation, the immunological staining, and the image analysis of the whole LN. SR, ElB, and NB performed, acquired and analyzed the regular LN immunostaining. JP, AP, PR, OBour, and OBoul performed the in vivo experiments. ElB, NB, and MB performed the cell-sorting. TL provided strong support for MGG and immunofluorescence images interpretation. OBour and OBoul supervised the in vivo experiments. ElB, EC, EG, and IS-C provided thorough discussions and critical manuscript reading. EG, IS-C, and NB provided financial supports. NB designed experiments and wrote the manuscript.

FUNDING
This work was supported by ERANET KILLeuPRRSV. ElB was supported by H2020 SAPHIR. EC was supported by ERANET KILLeuPRRSV and the AgreenSkills+ fellowship programme which has received funding from the EU's Seventh Framework Programme under grant agreement N • FP7-609398 (AgreenSkills+ contract). This project has received funding from the European Union's Horizon 2020 Program for research, technological development and demonstration under the Grant Agreement n • 633184. This publication reflects the views only of the author, and not the European Commission (EC). The EC is not liable for any use that may be made of the information contained herein.