Protection from septic peritonitis by rapid neutrophil recruitment through omental high endothelial venules

Acute peritonitis is a frequent medical condition that can trigger severe sepsis as a life-threatening complication. Neutrophils are first-responders in infection but recruitment mechanisms to the abdominal cavity remain poorly defined. Here, we demonstrate that high endothelial venules (HEVs) of the greater omentum constitute a main entry pathway in TNFα-, Escherichia coli (E. coli)- and caecal ligation and puncture-induced models of inflammation. Neutrophil transmigration across HEVs is faster than across conventional postcapillary venules and requires a unique set of adhesion receptors including peripheral node addressin, E-, L-selectin and Mac-1 but not P-selectin or LFA-1. Omental milky spots readily concentrate intra-abdominal E. coli where macrophages and recruited neutrophils collaborate in phagocytosis and killing. Inhibition of the omental neutrophil response exacerbates septic progression of peritonitis. This data identifies HEVs as a clinically relevant vascular recruitment site for neutrophils in acute peritonitis that is indispensable for host defence against early systemic bacterial spread and sepsis.

M ultiple clinical conditions such as trauma, surgery, liver cirrhosis, peritoneal dialysis catheters or disruption of the gastrointestinal tract can cause peritonitis. Subsequent systemic spread of intra-abdominal pathogens results in sepsis with high mortality and medical treatment is challenging 1 .
Leucocyte influx into the peritoneal cavity is a hallmark of abdominal inflammation but the routes of neutrophil emigration remain enigmatic. Here we studied the greater omentum, an intraperitoneal organ deriving from the greater curvature of the stomach and the spleen. It consists of a double layer of mesothelial cells that encapsulates a highly vascularized parenchyma of fatty tissue and multiple immune cell aggregates called milky spots. These are considered unconventional secondary lymphoid organs 2 and consist mostly of B1-cells, macrophages, T cells and mast cells, and lack follicular dendritic cells and germinal centres 3,4 . Milky spots dynamically increase in number and size in chronic inflammation or irritation 5,6 and a role in natural antibody production and adaptive peritoneal immunity is well established 2,7,8 . A key feature is the presence of a specialized type of postcapillary venule called high endothelial venule (HEV) that is essential for lymphocyte trafficking 9 . Similar innate B1-cell-rich clusters in adipose tissues have been observed in the mouse mesentery, the pleural cavity and the pericardium, and are termed fat-associated lymphoid clusters 10,11 .
To date, little is known about the regulatory functions of the greater omentum in innate immunity. Therefore, we studied its impact on leucocyte recruitment to and bacterial spread from the abdominal cavity in murine models of acute peritonitis. We can show that the first wave of neutrophils recruited to the inflamed peritoneum uses milky spot HEVs as exit site with unique molecular properties and that specific inhibition of this pathway leads to systemic bacterial spread and clinical exacerbation of septic peritonitis.

Results
Omental HEVs as neutrophil exit route in peritonitis. Milky spots are located predominantly at the lateral border of the greater omentum and peripheral node addressin (PNAd) as key marker for HEVs can be detected using the MECA-79 monoclonal antibody ( Fig. 1a; Supplementary Fig. 1a) 9 . At steady state, there are few neutrophils present in the omentum or the abdominal cavity, but neutrophil influx can be provoked by intraperitoneal application of TNFa (Fig. 1b) 12 . Omentectomized mice showed a strong reduction in peritoneal neutrophil influx at 1 h after TNFa application indicating a functional role of the greater omentum in acute peritonitis (Fig. 1b). Using LysM-eGFP reporter mice in which neutrophils are brightly fluorescent 13 with 1 h intraperitoneal (i.p.) TNFa treatment, we observe that neutrophil invasion is first triggered in milky spots in close vicinity to HEVs (Fig. 1c). Electron microscopy revealed polymorphonuclear cells in close contact to lymphocytes at luminal, subendothelial and perivascular sites in inflamed milky spots (Fig. 1d) suggesting a similar exit site for both leucocyte subsets.
Next, we developed a surgical technique to perform intravital microscopy on the omental microcirculation ( Supplementary  Fig. 1b) showing early neutrophil-endothelium interactions in milky spots and parenchymal neutrophil migration adjacent to HEVs in LysM-eGFP mice (Supplementary Movies 1 and 2; Supplementary Fig. 1c). Intravital 4D spinning disc confocal microcopy confirmed HEVs as predilection site for active neutrophil transmigration in acute TNFa-induced inflammation (Fig. 1e, Supplementary Movie 3). A similar phenotype was observed in a neutrophil-specific Ly6G-tdTomato reporter mouse 14 (Supplementary Movie 4).
We noticed that at early (1 h) but not later stages (43 h) of TNFa-inflammation neutrophils accumulated intraluminally in PNAd À /PECAM-1 þ postcapillary venules (PCV; Fig. 1c; Supplementary Fig. 2a). Tracking of neutrophil motion 60 min after TNFa revealed distinct HEV exit ramps as described for lymphocytes 15 , whereas PCVs predominantly showed rolling neutrophils (Fig. 1f). Time-lapse video recordings indicate a different temporospatial response of neutrophil recruitment in HEVs compared with PCVs with the former being dominant in the first 60-80 min of TNFa-inflammation ( Fig. 1g; Supplementary Fig. 2b; Supplementary Movie 5). The first wave of neutrophil egress out of the omental parenchyma into the surrounding area is initiated in close vicinity to milky spot HEVs as shown by intravital microscopy (Fig. 1h). Together, this data indicates that HEVs constitute a neutrophil exit platform in the early phase of peritoneal inflammation allowing rapid neutrophil entry into omental milky spots and subsequent release into the abdominal cavity.
Intravital imaging indicated that intraluminal neutrophils exhibited rolling, crawling and arrest in inflamed HEVs (Fig. 2c). Blocking concentrations of MECA-79 mAb abolished most neutrophil-HEV interactions ( Fig. 2c; Supplementary Movie 6). E-selectin can be detected in TNFa-exposed HEVs using immunofluorescence microscopy ( Supplementary Fig. 3d) and its inhibition induced a prominent crawling phenotype ( Inhibition of HEV-mediated neutrophil exit aggravates sepsis. Our data suggested that the greater omentum may play a role in the early innate immune response of bacterial infections. We used a sub-lethal caecal ligation and puncture (CLP) model to induce polymicrobial peritonitis and found that substantial neutrophil recruitment in omental milky spots is triggered at 4 h after surgery, whereas sham-operated controls showed only little signs of inflammation ( Fig. 3a and Supplementary Fig. 4a). To investigate the spread of abdominal bacteria, fluorescently labelled E. coli were injected i.p. Confocal imaging of the omentum after 2 h incubation revealed their presence predominantly in milky spots (Fig. 3b). Recruited neutrophils could be observed in close vicinity to milky spot HEVs and largely colocalize with E. coli (Fig. 3b, Supplementary Movie 9). Flow cytometry showed that CD11b þ F4/80 þ -macrophages but also CD11b þ Ly6G þneutrophils can phagocytose bacteria (Fig. 3c). Blocking L-selectin or PNAd abrogates neutrophil phagocytosis whereas macrophage function, assessed at 2 h after E. coli injection, is unaffected in both peritoneal lavage (Fig. 3c) and in omentum extracts ( Supplementary Fig. 4c). Next, we injected a sub-lethal dose of E. coli i.p. and assessed sepsis using a murine sepsis score 17 . While control mice developed mild clinical signs over a 2-h time course, blocking PNAd or L-selectin disabled the protective neutrophil response ( Supplementary Fig. 4d) which was associated with pronounced sepsis starting B45 min after E. coli injection (Fig. 3d). The severe clinical phenotype was associated with greatly increased bacterial loads in the omentum, peritoneal lavage and blood (Fig. 3e). Bacteremia was increased 50 to À 100fold in L-selectin blocked mice. The clinical sepsis score was positively correlated with bacterial burden of the omentum (Fig. 3f). This data suggests neutrophil recruitment via omental milky spot HEVs to be critical for abdominal pathogen clearance and preventing early onset of sepsis.

Discussion
The outstanding capability of the greater omentum to resolve inflammation and fight bacteria has been clinically appreciated for many decades but the underlying pathomechanisms remain mostly unknown 18 . We report here that a hyperacute phase of mono-and polymicrobial peritoneal inflammation induces rapid neutrophil influx into the omentum by means of milky spot HEVs.
The spatial and temporal molecular pattern of leucocyte-and endothelial-expressed adhesion molecules, chemokines and other regulatory proteins defines the quality and quantity of local cell recruitment. Organ-specific variations of the adhesion cascade ARTICLE such as in liver, brain and lung have been described 19,20 . Several lines of evidence argue for a distinct mechanism of neutrophil recruitment in HEVs. We detected a shorter response time compared with PCV after TNFa stimulation. In fact, the mode of leucocyte diapedesis in HEVs seems to be unique as shown by a mouse model with locked interendothelial junctions 21 . Electron microscopy studies highlighted a discontinuous basement membrane and loose lining of mesothelial cells at milky spots that might facilitate fast egress into the peritoneal cavity 22,23 . Furthermore, the panel of required adhesion receptors for neutrophil extravasation is exceptional. Constitutive lymphocyte homing to secondary lymphatic tissue requires L-selectin/PNAd and LFA-1/ICAM-1 interactions in HEVs for rolling and arrest, respectively 9,24 . It is known that neutrophils are also capable of rolling in HEVs at steady state in an L-selectin/PNAd-dependent manner 24 and we show that this is a prerequisite in inflammatory conditions. These findings confirm previous studies on the impact of L-selectin 25 and its ligand epitope 6-sulfo sialyl Lewis X (ref. 26) in abdominal inflammation and now identify the molecular and cellular mechanisms. In contrast to lymphocytes, Mac-1 and E-selectin but not LFA-1 critically contribute to neutrophil extravasation. Mac-1 is important for luminal crawling 27 and blockade of E-selectin or its counterreceptor PSGL-1 fully abrogates the downstream adhesion cascade. Interestingly, intravital microscopy reveals that the transition from crawling to arrest requires E-selectin engagement. This finding corroborates a recent report on constitutive E-selectin transcription in HEVs (ref. 28) suggesting modulatory functions in inflammation. While inside-out signalling to LFA-1 is well documented 29,30 , a role of E-selectin in LFA-1-independent arrest remains to be investigated.
A rapid phagocyte response is decisive for successful resolution of infection 31,32 . Our study attributes the greater omentum a critical task of particularly rapid neutrophil deployment in the peritoneal cavity via HEV-specific mechanisms. Inhibition of this pathway increases bacterial load in the peritoneum and exacerbates septic peritonitis. The critical time window is also known from clinical settings and referred to as 'golden hours of sepsis' 33 . Some data hint at an association between omentectomy and higher sepsis rates and mortality in patients 34,35 . However, murine peritonitis models do not seem to fully recapitulate human pathophysiology 36 and clinical studies are needed to definitively address the role of the greater omentum in inflammation and sepsis in humans.

Methods
Peritonitis models. Acute peritonitis was induced by i.p. injection of TNFa (500 ng) in 500 ml PBS. In some experiments, blocking antibodies (40 mg in 100 ml PBS) were administered intravenous (i.v.) 15 min before. After 30, 60 or 120 min animals were euthanized by CO 2 and the abdominal cavity was flushed with 5 ml cold PBS. Subsequently, the omentum was extracted for further analysis. Alternatively, LPS (25 mg, Sigma) was used with an incubation time of 12 h.
The CLP model was performed according to a published protocol with a sub-lethal mortality 40 . We confirmed the low-mortality time course of this protocol ( Supplementary Fig. 4e). Briefly, anaesthesia was induced using isoflurane and buprenorphine was administered subcutaneously for post-operative analgesia.
After a 1-cm midline incision of the abdomen the caecum was exteriorized. At 1 cm from the distal end, the caecum was ligated using 2-0 suture and perforated using a 27 gauge needle (both sides of the caecum). For short time points (o4 h), the animal was kept in anaesthesia after the operation until euthanasia and subsequent analysis. Sham operation as control did not include ligation and perforation.
For live E. coli experiments, non-pathogenic DH5a were cultured in lysogeny broth medium overnight. Before each experiment, E. coli were incubated for 2 h in fresh medium to induce log phase. After i.v. administration of blocking antibody or vehicle, 1 Â 10 8 E. coli were injected i.p. in 300 ml PBS and clinical signs of sepsis were assessed using a murine sepsis score with a maximum of 28 points 17 . After 2 h the peritoneal lavage, the omentum and blood (in EDTA) was obtained and plated on lysogeny broth agar plates in serial dilutions. Following 24 h of incubation at 37°C, the number of colonies was determined and expressed as mean of the diluted series.
For imaging studies, heat-inactivated fluorescently labelled E. coli (Bioparticles) were purchased from Molecular Probes. Particles (6 Â 10 7 ) were diluted in 500 ml PBS and injected i.p. After 2 h, peritoneal lavage and omentum were harvested for confocal microscopy or flow cytometry analysis.
Intravital microscopy. Mice were injected with fluorescently labelled antibodies i.v. and anaesthetized by i.p. injection of ketamine (125 mg kg À 1 ) and xylazine (12.5 mg kg À 1 ). In some experiments a carotid catheter was inserted to readminister anaesthesia, inject antibodies or take blood samples. All antibodies used in vivo were azide-free. After removal of abdominal fur, a 1-1.5-cm median incision was done at the linea alba of the upper abdomen to exteriorize the greater omentum. It extends from the greater curvature of the stomach and the spleen and can be distinguished from mesenterial fatty tissue by its lighter appearance. Gentle handling with cotton-wool tips is essential to avoid tissue damage and bleeding was electrocautered if necessary. A custom-made microscopic stage allowed fixation of the omentum between two cover slips. Throughout the experiment the animals were kept warm on a 37°C heating pad or incubation chamber and the exteriorized specimen was kept moist using heated bicarbonate-buffered saline. Omental inflammation was induced by topically applying TNFa (2.0 mg ml À 1 ) in PBS. To visualize HEVs only low doses of MECA-79 antibody (2-3 mg per mouse) were injected. 200 kDa FITC-or rhodamine-dextran was used as blood tracer. Videos were recorded using either a Axio Scope.A1 microscope (Zeiss) at Â 10 (numerical aperture 0.25) or Â 20 (numerical aperture 0.45) equipped with stroboscopic epifluorescent illumination (Colibri 2, Zeiss) or an upright spinning disc microscope (Zeiss Imager Axio Examiner D1). After imaging, mice were euthanized in anaesthesia and the omentum was extracted and fixed in 2% paraformaldehyde for further analysis.
Data analysis and cell tracking was done using Fiji 41 . Where indicated, movies were denoised via total variation denoising (according to ref. 42 with b ¼ 0). All video frames were rigidly aligned based on minimizing the so-called L1-distance between them. Colour channels with strong photobleaching but no other temporal change were replaced by the first frame to improve their visibility.
Omentectomy. Mice were anaesthetized as described above. A 1-cm incision was made in the upper linea alba to expose the greater curvature of the stomach. The omentum was clamped on both most lateral ends using micro clips (0.8 Â 5 mm), ligated and subsequently removed. Great care was taken to avoid bleeding. Next, 500 ng TNFa in 500 ml PBS were applied into the peritoneal cavity and the abdominal wall was closed up. After 1 h, the abdominal cavity was lavaged with 2 ml cold PBS. Sham operation consisted of abdominal incision and gentle manipulation of the omentum.
Confocal microscopy. Omenta were isolated, fixed in 2% PFA for 1 h, blocked with 5% foetal calf serum for 30 min and incubated with primary antibodies in PBS at 4°C. Specimens were subsequently stained with the secondary antibodies for 4 h at 4°C. Whole mount tissues were examined using a Zeiss LSM 700 confocal microscope equipped with a Hamamatsu ORCA-ER CCD camera. Images were analysed using Volocity software (ImproVision).
Electron microscopy. Milky spots were manually dissected from the omentum, fixed, dehydrated and embedded in epon resin using standard protocols. Sections (60 nm thin) were cut on a microtome and stained with uranyl acetate and lead citrate. Images were taken using a Philips CM10 electron microscope. Milky spots were identified by the high density of parenchymal leukoytes. HEVs showed a higher cell morphology and were luminally populated with mononuclear cells.