Human severe sepsis cytokine mixture increases β2-integrin-dependent polymorphonuclear leukocyte adhesion to cerebral microvascular endothelial cells in vitro

Introduction Sepsis-associated encephalopathy (SAE) is a state of acute brain dysfunction in response to a systemic infection. We propose that systemic inflammation during sepsis causes increased adhesion of leukocytes to the brain microvasculature, resulting in blood-brain barrier dysfunction. Thus, our objectives were to measure inflammatory analytes in plasma of severe sepsis patients to create an experimental cytokine mixture (CM), and to use this CM to investigate the activation and interactions of polymorphonuclear leukocytes (PMN) and human cerebrovascular endothelial cells (hCMEC/D3) in vitro. Methods The concentrations of 41 inflammatory analytes were quantified in plasma obtained from 20 severe sepsis patients and 20 age- and sex-matched healthy controls employing an antibody microarray. Two CMs were prepared to mimic severe sepsis (SSCM) and control (CCM), and these CMs were then used for PMN and hCMEC/D3 stimulation in vitro. PMN adhesion to hCMEC/D3 was assessed under conditions of flow (shear stress 0.7 dyn/cm2). Results Eight inflammatory analytes elevated in plasma obtained from severe sepsis patients were used to prepare SSCM and CCM. Stimulation of PMN with SSCM led to a marked increase in PMN adhesion to hCMEC/D3, as compared to CCM. PMN adhesion was abolished with neutralizing antibodies to either β2 (CD18), αL/β2 (CD11α/CD18; LFA-1) or αM/β2 (CD11β/CD18; Mac-1) integrins. In addition, immune-neutralization of the endothelial (hCMEC/D3) cell adhesion molecule, ICAM-1 (CD54) also suppressed PMN adhesion. Conclusions Human SSCM up-regulates PMN pro-adhesive phenotype and promotes PMN adhesion to cerebrovascular endothelial cells through a β2-integrin-ICAM-1-dependent mechanism. PMN adhesion to the brain microvasculature may contribute to SAE.


Introduction
Sepsis, a systemic inflammatory response to a known or suspected infection, is a leading cause of admission and mortality in intensive care units (ICUs) [1]. Severe sepsis refers to sepsis accompanied by acute organ dysfunction [2]. Up to 70% of patients with severe sepsis experience an acute neurological dysfunction known as sepsis-associated encephalopathy (SAE) [3]. Acute brain dysfunction in severe sepsis patients is associated with increased rates of mortality and long-term cognitive impairment in survivors [4,5].
The mechanisms of SAE are unknown, but possible instigating factors may be the early activation of the innate immune response and microcirculatory dysfunction that are common manifestations of severe sepsis [6,7]. Activated leukocytes and vascular endothelial cells contribute to systemically increased concentrations of cytokines and chemokines in patients with sepsis [8].
Circulating inflammatory mediator(s) activate polymorphonuclear leukocytes (PMN), leading to up-regulation of the pro-adhesive phenotype and recruitment of PMN to the inflamed tissues and/or organs [9,10]. While PMN are critical for the host defense against pathogens, the systemic activation of PMN also correlates with organ dysfunction in septic patients [11]. Animal models of sepsis demonstrate that overwhelming PMN recruitment is associated with dysfunction of the affected organs, including the brain [12][13][14].
Endothelial cells of the blood-brain barrier (BBB) are likely disrupted during SAE, as vasogenic edema is consistently found on head magnetic resonance imaging in sepsis [16]. Lesions in the subcortical white matter, also consistent with vasogenic edema, are particularly prominent in cases of fatal septic shock [17]. Increased signal associated with contrast agent administration during SAE indicates BBB disruption [6]. The mechanism(s) of BBB disruption in sepsis, however, remain poorly understood.
In this study, we hypothesized that the inflammation associated with severe sepsis would result in greater PMN interactions with the brain microvasculature. Thus, we first aimed to measure circulating inflammatory mediators in blood plasma from patients with severe sepsis, and then investigate the effects of these inflammatory analytes on the pro-adhesive phenotype of human PMNs and human cerebrovascular endothelial cells (hCMEC/D3) in vitro.

Materials and methods
The institutional review boards of Hamilton Health Sciences (Hamilton, ON, Canada) and the University of Western Ontario (London, ON, Canada) approved this study. Consent was obtained from the ICU patients or their legal guardians. Consistency of patient blood sampling in the participating centers was monitored by a study investigator (PCL).

Severe sepsis patients and blood plasma
Blood plasma collection was described previously [18]. Briefly, blood was collected from ICU patients within the first 24 hours of severe sepsis diagnosis according to the American College of Chest Physicians/Society of Critical Care Medicine (ACCP/SCCM) guidelines [2,19]. Venous blood (4.5 mL) was drawn via indwelling catheters (severe sepsis patients) or fresh venipuncture (from age-and sex-matched healthy controls), and immediately transferred into 15 mL polypropylene tubes (VWR, Mississauga, ON, Canada) containing 0.5 mL of 0.105 mol/L buffered trisodium citrate (pH 5.4) (Sigma-Aldrich, St. Louis, MO, USA) with 100 μL of 1 mol/L benzamidine HCl (approximately 20 mmol/L benzamidine), a serine protease inhibitor (Sigma-Aldrich, St. Louis, MO, USA). The blood was immediately centrifuged at 1,500 g for 10 minutes at 20°C, and the plasma was stored in aliquots at −80°C.
Care was taken to ensure that freeze/thaw cycles were avoided. Baseline characteristics of severe sepsis patients were recorded at ICU admission (Table 1). Baseline
Briefly, 100 μL of multiplexed standards (1,000-12.3 pg/mL) or diluted plasma was applied to each well and incubated at room temperature on a shaker at 200 rpm for two hours. After sample incubation, the microarrays were rinsed three times with wash buffer. Prior to reincubation on a shaker for one hour, 75 μL pre-titered biotinylated detector antibodies were added to each well.
Microarrays were washed three times and 100 μL Streptavidin DyLight 649 (part of Thermo Scientific ExcelArray™kit; Thermo Fisher Scientific, Rockford, IL, USA) was applied to each well, preceding a final incubation of 30 minutes on a shaker. Microarrays were washed five times and dipped briefly in a final rinse solution. Slides were then centrifuged at approximately 200 g until dry. Methodologies for microarray production and antibody cross reactivity assays have been described previously [20]. Cross reactivity on the microarray antibody pairs is ≤5%, as reported by the manufacturer. Microarrays were imaged using an Alphascan™ Microarray Imager (Alpha Innotech, San Leandro, CA, USA) and spot densitometry was performed by Thermo Fisher (Rockford, IL, USA) using ArrayVision™ Software (GE Healthcare, Piscataway, NJ, USA).

Human cytokine mixtures
The most up-regulated (P <0.001) human analytes detected in the plasma of severe sepsis patients (IL-8, G-CSF, IP-10, GRO-α, HGF, MIP-1β, and MCP-1; Table 2) were selected to make the severe sepsis cytokine mixture (SSCM) and control cytokine mixture (CCM). The sepsis-relevant cytokine IL-6 [21,22] was also significantly up-regulated in the plasma of severe sepsis patients (P = 0.017) and was added to each cytokine mixture. Recombinant proteins were suspended in PBS containing 1% BSA at 100X concentration measured in severe sepsis blood plasma or healthy control blood plasma (Table 2). Each cytokine mixture was stored at −20°C until thawed for experiments. To treat cells, each cytokine mixture was diluted at 100X concentration in either PBS or VascuLife® EnGS-Mv (LifeLine cell technology, Walkersville, MD, USA) cell culture medium to the corresponding analyte concentrations measured in control blood Plasma was obtained from severe sepsis patients and compared to plasma from age-and sex-matched healthy controls. Analytes were measured with antibody microarray and presented as mean ± SE. Values are in ng/mL (n = 20 per group). MCP-1-monocyte chemotactic protein-1; GCSF-granulocyte colony-stimulating factor; IL-8-interleukin-8; HGF-hepatocyte growth factor; MIP-1β-macrophage inflammatory protein-1 beta; IP-10-interferon-inducible protein 10 kDa; GRO-α-growth-regulated oncogene alpha; IL-6-interleukin-6.
plasma or severe sepsis blood plasma, respectively. All human recombinant analytes were obtained from either Ebioscience (San Diego, CA, USA) or Invitrogen (Carlsbad, CA, USA).

Human polymorphonuclear leukocyte isolation
Human PMN were isolated from venous blood of healthy adults immediately before experiments as previously described by us [24]. Briefly, heparinized whole blood (10 mL; 10 U/mL; Pharmaceutical Partners of Canada, Richmond Hill, ON, Canada) was gently mixed by inversion (10 times) with 5 mL of 3% dextran in PBS (pH 7.4). Sedimentation of red blood cells was performed via incubation of the mixture for 20 minutes at 23°C. Blood plasma was removed and gently added onto 5 mL of Histopaque® lymphocyte separation medium (density of 1.077 g/mL; Sigma-Aldrich, St Louis, MO, USA). Samples were centrifuged for 30 minutes at 400 × g at 23°C. Following centrifugation, the pellet (PMN and remaining red blood cells) was re-suspended in 15 mL cold red blood cell lysis buffer, containing NH4Cl (8.3 g/L), KHCO3 (1.0 g/L), and EDTA (0.14 mM) (all purchased from Sigma-Aldrich, St. Louis, MO, USA) in water, adjusted to a pH of 7.2. After five minutes incubation at 4°C, mixtures were centrifuged at 300 × g for five minutes. PMN were re-suspended in 1 mL PBS (4°C), stored on ice until use, and counted using a hemocytometer (VWR, Mississauga, ON, Canada). This procedure yields a PMN population that is 95 to 98% viable (as measured by trypan blue exclusion) and 98% pure (as measured by acetic acid-crystal violet staining) (both reagents were purchased from Sigma-Aldrich, St. Louis, MO, USA).

Polymorphonuclear leukocyte adhesion to hCMEC/D3
hCMEC/D3 were grown to confluence in μ-Slide VI 0.4 parallel flow channels (ibidi; MOFA Global, Ingersoll, ON, Canada) and interacted with PMN as previously described by us [24,25]. Briefly, hCMEC/D3 were seeded at a density of 11,000 cells per channel and grown for 5 days, followed by 5 hours of stimulation with either CCM or SSCM diluted in serum free-VascuLife EnGS-Mv cell culture medium. The CCM and SSCM treatments did not affect hCMEC/D3 viability as assessed by ViaCount Flex (Cedarlane, Burlington, ON, Canada) assay and flow-cytometry approach (data not shown). Each channel was then perfused (that is, washed) in the presence of laminar shear stress of 0.7 dyn/cm 2 for one minute with VascuLife EnGS-Mv cell culture medium, using a Harvard Apparatus 22-syringe pump (Instech, Plymouth Meeting, PA, USA). Subsequently and continuously, hCMEC/D3 were perfused with freshly isolated PMN (1 × 10 6 PMN/mL; suspended in serum free-VascuLife EnGS-Mv cell culture medium) that have been stimulated either with CCM or SSCM for 10 minutes. The experiment was visualized and video recorded using a Nikon DIAPHOT-300 inverted microscope (Nikon, Chiyoda-ku, Tokyo, Japan) equipped with a temperature-controlled chamber (37°C) and Panasonic WJ-810 video-camera (Panasonic, Kadoma, Osaka, Japan) connected to an SV-video recorder (Panasonic). For each experiment, 5 random fields of view (0.1 mm 2 each) were recorded for 30 seconds, starting at 4 minutes after initiation of the PMN perfusion. The recorded PMN-hCMEC/ D3 adhesive interactions were video-analyzed to count the number of firmly adherent PMN (stationary for at least five seconds). The average number of adherent PMN was expressed as PMN/0.1 mm 2 .
RNA isolation, cDNA synthesis and quantitative PCR hCMEC/D3 were seeded (500,000 per well) on six-well plates and grown for two days, preceding four hours of stimulation with either CCM or SSCM. Following stimulation, hCMEC/D3 were washed three times with PBS and RNA was extracted with 1 mL TRIzol® reagent (Life Technologies, Carlsbad, CA, USA), as per the manufacturer's protocol except the RNA pellet was washed three times with 75% ethanol. RNA was resuspended in 20 μL of 10 mM Tris, pH 8. The purity and concentration of RNA was determined using a Biophotometer (Eppendorf, Hamburg, Germany). One microgram of RNA per sample was reverse transcribed using iScript™ Reverse Transcription Supermix (Bio-Rad, Hercules CA, USA) according to the manufacturer's protocol.
Quantitative PCR (qPCR) was performed using Sso Fast™ Probes Supermix (Bio-Rad, Hercules, CA, USA), according to the manufacturer's protocol, using cDNA at concentrations previously determined to be within 90 to 110% efficiency. Primers were predesigned (Applied Biosystems, Foster City, CA, USA): 18S (Hs99999901_s1), glyceraldehyde-3-phosphate dehydrogenase (GAPDH; Hs99999905_m1), ICAM-1 (Hs00164932_m1), and VCAM-1 (Hs01003372_m1). Real-time PCR was run using a C1000™ Touch Thermal Cycler with CFX96™ Real-Time System (Bio-Rad, Hercules CA, USA). Cycling conditions included one-time enzyme activation at 95°C for 30 seconds, followed by 50 cycles of denaturation at 95°C for 5 seconds, and annealing and extension at 60°C for 15 seconds. Target gene expression was normalized to GAPDH and 18S and was expressed relative to gene expression by hCMEC/D3 treated with serum-free VascuLife EnGS-Mv cell culture media.

Statistical analysis
Data were analyzed using SigmaStat 3.5 (Systat Software Inc., San Jose, CA, USA). Continuous variables are reported with mean ± SE. Groups were pre-screened for normality and compared with either a Student's t test or Mann-Whitney U test. Multiple groups were analyzed using one-way analysis of variance (ANOVA) with Tukey's post-hoc test. A P <0.05 was considered statistically significant.

Polymorphonuclear leukocyte adhesion to hCMEC/D3
PMN and/or hCMEC/D3 were stimulated with either SSCM or CCM and assessed for PMN adhesion to hCMEC/D3 in the presence of flow (laminar shear stress 0.7 dyn/cm 2 ). Stimulation of hCMEC/D3 with SSCM did not significantly increase the adhesion of naïve (unstimulated) PMN (Figure 1). On the contrary, stimulation of PMN with SSCM significantly increased PMN adhesion to hCMEC/D3 (P <0.05; n = 7; Figure 1), while simultaneous treatment of PMN and hCMEC/D3 with SSCM (co-stimulation) had no additive effects.

Discussion
The present study identified an inflammatory profile in severe sepsis patients at ICU day one that up-regulated a pro-adhesive phenotype of PMN, resulting in increased adhesion to hCMEC/D3 in vitro. These effects may be important for the pathophysiology of SAE, as adhered and/or activated PMN could result in dysfunction of the BBB.
The extracellular environment of the brain is highly regulated by the BBB. Dysfunction of the BBB has been demonstrated with brain imaging in patients with SAE [6,16,17], and BBB dysfunction is also observed in lipopolysaccharide-challenged animals [26]. Injury to the cellular components of the BBB, including cerebrovascular endothelial cells [27], pericytes [28], and astrocytes [29], was observed in experimental animal models of sepsis. Cerebrovascular endothelial cell dysfunction may lead to dysregulation of the brain extracellular environment and subsequent neuronal dysfunction, thereby causing SAE.
In the current study, we collected blood plasma samples from ICU patients within the first 24 hours of severe sepsis diagnosis. Encephalopathy occurs early in sepsis, often developing on the first day of ICU admission [30]. Thus, blood was sampled at a critical time point for SAE development.
A total of 13 inflammatory analytes were up-regulated in blood plasma obtained from severe sepsis patients compared to blood plasma from age-and sex-matched healthy controls. Most of the up-regulated analytes are  pro-inflammatory cytokines (such as IL-6) and CXC-and CC-family chemokines (such as IL-8, GRO-α, MCP-1, and MIP-1β), with well-documented roles in cellular activation and leukocyte recruitment to vascular endothelium [31,32]. In addition, a marked increase in cell growthmodulating factors (such as HGF, FGFβ, HB-EGF, and KGF) in the circulation of severe sepsis patients was also found. The role of these growth factors in tissue remodelling has been shown previously, but the exact role of Figure 4 Effects of anti-ICAM-1 antibody on severe sepsis cytokine mixture (SSCM)-induced PMN adhesion to hCMEC/D3. hCMEC/D3 were stimulated with CCM or SSCM for 5 hours, treated with anti-ICAM-1 function neutralizing antibody for 30 minutes, and interacted with CCM-or SSCM-stimulated (10 minutes) PMN to assess PMN adhesion in the presence of flow (laminar shear stress 0.7 dyn/cm 2 ). n = 6; *P <0.05 as compared to corresponding control cytokine mixture (CCM); # P <0.05 as compared to isotype IgG treatment. CCM-control cytokine mixture; SSCM-severe sepsis cytokine mixture; hCMEC/D3-human cerebral microvascular endothelial cells/D3; PMN-polymorphonuclear leukocytes; Ab-antibody; ICAM-1intercellular adhesion molecule-1.  these molecules in the pathophysiology of severe sepsis is unknown.
Acute expression of individual cytokines and chemokines (such as IL-6, IL-8, MCP-1, and G-CSF) has been reported to correlate with the presence and severity of organ dysfunction in patients with severe sepsis and septic shock [22]. Interestingly, cytokines that are most commonly employed in experimental models of inflammation and/or sepsis (TNF-α, IL-1β, and IFN-γ) were not significantly increased in severe sepsis plasma obtained within 24 hours of ICU admission. One possible explanation for the latter phenomenon could be that the cytokines mentioned above may be expressed more prominently at different stages (for example, earlier or later than 24 hours post-admission time) and severities of sepsis, or in subgroups of severe sepsis patients [22,33]. Thus, it would be plausible to assume that the severe sepsis analyte profile within the first 24 hours of ICU admission represents not an 'early' , but rather a 'later' cytokine response, that may intensify and perpetuate the inflammatory response.
One unique aspect of this study was the use of SSCM that consisted of eight cytokines and chemokines at the actual concentrations detected in the plasma of severe sepsis patients. In contrast, previous sepsis studies in the field have used a single cytokine and chemokine-based stimulation of the cells at often 'un-physiologically' high concentrations. In the current study we employed a severe sepsis analyte profile obtained at a time-point of disease that is clinically relevant to SAE development [3]. Our results indicate that PMN, but not hCMEC/D3, are more responsive to SSCM, with regards to up-regulation of the pro-adhesive phenotype. Stimulation of PMN, or PMN and hCMEC/D3, with SSCM resulted in an increased adhesion of PMN to hCMEC/D3 under in vitro conditions of flow, whereas stimulation of hCMEC/D3 only had no effect on PMN adhesion. This latter phenomenon may be explained by the inability of SSCM to up-regulate inducible adhesion molecule (such as ICAM-1 or VCAM-1) expression in hCMEC/D3 (as evidenced by qPCR analysis; Figure 5). On the other hand, the levels of constitutively expressed ICAM-1 on the surface of endothelial cells are sufficient to promote adhesion of activated PMN. Whether cerebrovascular endothelial cells (such as hCMEC/D3) are less sensitive to the pro-inflammatory cytokine and chemokine stimulation at physiologic concentrations as compared to the non-brain endothelial cells (such as pulmonary microvascular endothelial cells or umbilical vein endothelial cells) remains to be determined.
Our data indicate that β 2 -integrins (such as LFA-1 and Mac-1) are required for SSCM-induced PMN adhesion to hCMEC/D3. Interestingly, immune-neutralization of ICAM-1 (a receptor for β 2 -integrins) in hCMEC/D3 was also very effective in reducing PMN adhesion, despite the fact that SSCM failed to up-regulate ICAM-1 expression in hCMEC/D3. The latter may be explained by reduced availability of the constitutively expressed ICAM-1 following treatment of hCMEC/D3 with the immunoneutralizing anti-ICAM-1 antibody.
Up-regulated surface expression of β 2 -integrin has been reported on freshly isolated PMN stimulated with blood plasma obtained from severe sepsis patients 24 hours after diagnosis [34]. Treatment of PMN with individual chemokines, such as IL-8, may also induce LFA-1-or Mac-1-dependent PMN adhesion [35,36]. The up-regulation of β 2 -integrins (such as LFA-1 and Mac-1) surface expression is not always a prerequisite for LFA-1-and Mac-1-dependent adhesion. Rather, β 2 -integrins that are constitutively expressed on the leukocyte cell membrane undergo conformational activation to a pro-adherent state that promotes leukocyte adhesion to molecules such as ICAM-1 [37,38]. LFA-1 and Mac-1 are known to bind ICAM-1 on vascular endothelial cells to facilitate PMN adhesion [39,40]. Ameliorated PMN recruitment to brain tissue has been demonstrated during endotoxemia in ICAM-1-deficient mice, similar to our findings in which an anti-ICAM-1 antibody reduced PMN adhesion to hCMEC/D3, suggesting an important role for β 2 -integrin-ICAM-1 interactions in PMN adhesion to CMEC [41].
There is currently no data published regarding the use of antibody treatment directed against LFA-1, Mac-1, or β 2 -integrin in humans for treatment of severe sepsis or SAE. However, LFA-1-and Mac-1-dependent recruitment of PMN is not unique to the brain, and studies that employed murine and rabbit models of sepsis have demonstrated that antibodies directed against β 2 -integrin, LFA-1, and Mac-1 can reduce PMN recruitment and ameliorate subsequent dysfunction of lung and liver [42][43][44].
Whether PMN adhesion to the brain microvasculature in severe sepsis causes brain dysfunction is unclear. However, PMN are armed with many enzymes involved in host defense against pathogens, including NADPH oxidase, myeloperoxidase (MPO), matrix metalloproteinases (MMPs), elastase, and cathepsins [45]. Overwhelming local recruitment of activated PMN to brain microvasculature could expose CMEC to the deleterious effects of these enzymes. Both NADPH oxidase and MPO produce reactive oxygen species (ROS), which directly or indirectly contribute to the disruption of vascular endothelial cell intercellular junctions [27,46,47]. ROS production by PMN isolated from severe sepsis patients, and elevated MPO concentrations in blood plasma isolated from septic patients have both been reported [11,48]. MMPs and elastase can also damage endothelial cell adherens and tight junctions [49,50]. Disruption of the brain microvasculature may contribute to pathological activation of glial cells or neurons and, subsequently, to the development of SAE [27,28].
At present, there is little pathophysiological data on SAE, and no studies using human tissues. Animal models of sepsis are a predominant source of data regarding potential mechanisms of brain dysfunction in sepsis. Excessive PMN recruitment, associated with neurological impairment in the brains of septic mice, has been reported [13,51]. PMN recruitment was ameliorated by treatment with antibodies directed against ICAM-1, β 2 -integrin, and P-selectin adhesion molecules. Anti-P-selectin antibody treatment was associated with decreased sickness behaviour by septic mice, suggesting a possible role for PMNendothelial interactions in brain dysfunction during sepsis.
Our data have several limitations. First, our in vitro culture model may not fully represent the cerebral microvasculature in severe sepsis. Brain endothelial cells in vivo are in close contact with astrocytes, and co-culture of hCMEC/D3 with astrocytes has been demonstrated to decrease basal protein expression of ICAM-1 [52]. Therefore, hCMEC/D3 may express ICAM-1 at a higher basal rate compared to brain endothelial cells in vivo, which may reduce the window for up-regulation of ICAM-1 in response to cytokine stimulation. Second, our SSCM does not include all potential mediators that may play a role in leukocyte and/or endothelial activation in inflammation. The use of SSCM, however, provides valuable information regarding the acute inflammatory state in severe sepsis patients, at a time point when brain dysfunction most often occurs [3,53].

Conclusions
Our data elucidate the late inflammatory changes in plasma during the first 24 hours of severe sepsis diagnosis in ICU patients, and demonstrate inflammation-induced PMN adhesion to hCMEC/D3, mediated by β 2 -integrin-ICAM-1 interaction.

Key messages
The pathophysiology of sepsis-associated encephalopathy (SAE) is unknown, but systemic inflammation may cause dysfunction of the brain microvasculature. We measured inflammatory analytes in plasma during severe sepsis, and created a severe sepsis cytokine mixture (SSCM). SSCM induced neutrophilic leukocyte (PMN) adhesion to cultured cerebrovascular endothelial cells in vitro. β2-integrins on PMNs, and ICAM-1 on cerebrovascular endothelial cells, mediated increased cellular adhesions. PMN adhesion to the brain microvasculature during severe sepsis may contribute to SAE.