In vitro endothelial hyperpermeability occurs early following traumatic hemorrhagic shock

2.1 Study protocol

The Microshock study was approved in the UK under Research Ethics Committees reference: 14/YH/0078 (Clinical trial registration: NCT02111109) [8]. The current study includes a subgroup analysis of 16 adult patients from a single site (Birmingham, UK) [8]. These patients had evidence of hemorrhagic shock as represented by a mechanism of injury consistent with blood loss. Eligible patients needed to be intubated and ventilated, have a serum lactate concentration > 2 mmol.l^-1 recorded at any stage prior to admission to the intensive care unit (ICU), and received any blood products (e.g., packed red blood cells (PRBC), fresh frozen plasma (FFP), cryoprecipitate) during the initial period of resuscitation, prior to admission to the ICU. A process of deferred consent was approved by the research ethics committee [8]. Characteristics and serum analyses of syndecan-1 and soluble thrombomodulin [9, 10] and microcirculatory perfusion measurements of these patients [4] were published previously.

2.2 Blood sampling

Blood was collected in both citrate and z-serum clotting activator tubes (for plasma and serum samples respectively) at three different time points, which include arrival at the emergency department (ED), admission at the ICU (ICU) and 24 hours after admission at the ICU (ICU+24 h) (Fig. 1A). To reach n = 8 per time point as derived by our power calculation, plasma was withdrawn from 16 patients in total as it was difficult to include plasma at three time points from the same patient. Eight patients had available blood samples at two time points, the remaining patients had an available blood sample at one time point. Plasma was also collected from eight age- and sex-matched controls. For serum samples, blood from z-serum clotting activator tubes was centrifuged at 1,620 G for 10 min at 4°C. For plasma samples, blood from citrate tubes was centrifuged at 2,000 G for 20 min at 4°C, and then the supernatant was additionally centrifuged at 13,000 G for 2 min at 4°C to achieve platelet-free plasma. The platelet-free plasma and serum samples were stored at –80°C.

OPEN IN VIEWER

Fig.1

Study protocol and in vitro experimental set-up. (A) Blood was collected (for plasma and serum samples respectively) and sublingual non-invasive incident dark field videomicroscopy was performed at three different time points: after arrival at the emergency department (ED; n = 8), after admission at the intensive care unit (ICU; n = 8) and 24 hours later (ICU+24 h; n = 8). (B) Human endothelial cells were seeded and culture until a monolayer was formed. After starvation in 1% human serum albumin (HSA), endothelial cells were exposed to plasma from traumatic hemorrhagic shock patients collected at ED admission, ICU admission or 24 h later. Measurements were performed using electric cell-substrate impedance sensing (ECIS; experiment I.) or via immunofluorescence imaging (IF imaging; experiment II.).

For cell culture experiments, the following materials were used: bare medium (bM199) consisting of Medium 199 supplemented with penicillin 100 U.ml^-1 and streptomycin 100 mg.ml^-1 (all from Biowhittaker, Verviers, Belgium); complete medium (cM199) consisting of bM199 supplemented with 10% heat inactivated new-born calf serum (Gibco, Grand Island, NY, USA), 10% heat inactivated human serum (pooled serum of 10–20 healthy donors, stored at 4°C, Sanquin CLB, Amsterdam, the Netherlands), 2 mmol.l^-1 glutamine (Biowhittaker, Verviers, Belgium), 5 U.ml^-1 heparin (Leo Pharmaceutical Products, Weesp, The Netherlands), 150 μg.ml^-1 crude endothelial cell growth factor prepared from bovine hypothalamus; 1% HSA solution (dilution of human serum albumin in bM199; Sanquin, Amsterdam, the Netherlands) and Trypsin (Gibco, Grand Island, NY, USA).

Human umbilical vein endothelial cells (HUVECs) were isolated and cultured as described before [11, 12]. Briefly, endothelial cells were isolated from human umbilical cords obtained from healthy donors from Amstelland Hospital (Amstelveen, the Netherlands) and subsequently cultured on gelatin-coated well plates in cM199 medium at 37°C, in an atmosphere of 95% air and 5% CO₂.

2.4 Electric cell–substrate impedance sensing

Electric cell–substrate impedance sensing (ECIS; Applied BioPhysics, Troy, NY, USA) was used to measure impedance of endothelial cells as previously described [13]. Confluent, passage one HUVECs pooled from three donors were transferred to gelatin-coated 96-wells ECIS culture plates (Applied BioPhysics, Troy, NY, USA) pre-treated with 10 mM L-cysteine (Merck, Darmstadt, Germany). After 72 h of culturing in cM199 medium, the ECIS device was used for continuous, multi frequency scanning to confirm a confluent monolayer. Confluent monolayers were washed with and exposed to 1% human serum albumin (HSA) in bare medium for 60 min followed by addition of 10% platelet-free plasma collected from traumatic hemorrhagic shock patients or age- and sex-matched controls. Impedance was recorded for 2 hours until steady state was reached at multiple frequencies ranging from 250–64000 Hz using ECIS software (v1.2.210.0 PC; Applied Bio-Physics). A schematic overview of the experimental set-up is shown in Fig. 1B. Plasma concentration and exposure time were based on previously performed experiments [12] and confirmed in pilot experiments.

Resistance is one of the parameters that is derived from impedance measurements and represents quality and function of the cell barrier [13]. Impedance was also used to calculate cell–cell integrity (Rb) and cell–matrix integrity (α). All measurements were performed in triplicate and normalized to baseline.

2.5 Immunofluorescence imaging

Immunofluorescence was used to visualize cell structures of interest. HUVECs were seeded on 5 μM fibronectin-coated coverslips. After reaching a confluent monolayer, cells were washed with and exposed to 1% HSA for 60 min followed by 2 hour stimulation with 10% platelet-free plasma from either traumatic hemorrhagic shock patients collected upon arrival at the ED (n = 5) or age- and sex-matched controls (n = 8) (Fig. 1B). Cells were washed with bM199 and fixed with 4% paraformaldehyde at room temperature. Upon fixation, cells were permeabilized with 0.2% Triton X-100 in PBS for 3 min and unspecific staining was blocked by incubation with 1% HSA. Cells were incubated with a primary antibody against VE cadherin (C-19, 1:400, Santa Cruz, Dallas, TX, USA) and a secondary antibody against donkey anti-goat Alexa Fluor 488 (Invitrogen, Carlsbad, CA, USA). DAPI (Thermo Fisher Scientific, Waltham, MA, USA) and acti-stain phalloidin 670 (Cytoskeleton, Denver, CO, USA) were used to stain nuclei and F-actin, respectively. Coverslips were mounted in Mowiol mounting solution (Sigma Aldrich, Zwijndrecht, The Netherlands). Imaging was performed with a Zeiss Axiovert 200M Marianas^TM digital imaging inverted microscope system using a 63x Zeiss oil objective (Zeiss, Breda, the Netherlands).

2.6 Microcirculatory perfusion measurements

Microcirculatory perfusion has previously been reported for 58 patients included in the Microshock study [4]. Briefly, sublingual microcirculatory perfusion was measured using non-invasive incident dark field (IDF) videomicroscopy (Cytocam, Braedius Medical, Huizen, the Netherlands). In vivo microcirculatory perfusion measurements were performed at time points corresponding to blood sampling (Fig. 1A). At least four video clips of 5 seconds (100 frames) duration were recorded and stored de-identified for blinded analysis. Microcirculatory videos were analyzed by a single operator using automated vascular analysis (AVA) v.3.02 (Microvision Medical, Amsterdam, The Netherlands) by manually tracing all vessels by hand. These video analyses give values for total vessel density (TVD), perfused vessel density (PVD) and proportion of perfused vessels (PPV) according to the current best practice guidelines for reporting microcirculatory variables [4]. Here, in vivo microcirculatory perfusion measurements of the 16 patients that were included in the present study were used for additional analysis.

2.7 Serum analysis

Serum concentrations of angiopoietin-1, angiopoietin-2, soluble Tie2 (R&D Systems, Minneapolis, MN, USA) and von Willebrand factor (Abcam, Cambridge, MA) were analyzed using commercially available enzyme-linked immunosorbent assays. Serum levels of both syndecan-1 and soluble thrombomodulin have previously been reported of the patients in this cohort [9, 10], and were included in additional analyses in the current study.

2.8 Statistical analysis

Data were analyzed using Graphpad Prism 7.0 (Graphpad Software, La Jolla, CA, USA).

At least a 20% reduction (Δ= 230 Ohm) in in vitro endothelial resistance with a standard deviation of 140 Ohm following exposure to plasma of traumatic hemorrhagic shock patients was expected [12]. With a significance level (α) of 0.05 and beta of 0.9 group sizes of n = 8 were calculated.

Continuous data are expressed as mean±standard deviation for normally distributed data, or median followed by the interquartile range (Q1–Q3) for non-normal data. Normality of distribution was tested with the Shapiro-Wilk test. Changes in endothelial resistance over time were evaluated using two-way ANOVA with Bonferroni post-hoc analyses. An one-way ANOVA with Bonferroni post-hoc analysis or Kruskal Wallis test was used to evaluate differences between groups for normally and non-normally distributed data, respectively. Associations between in vitro endothelial resistance with patient characteristics and microcirculatory perfusion parameters were analyzed using Pearson’s correlation coefficient or Spearman’s rank correlation coefficient for normally and non-normally distributed data, respectively. P values < 0.05 were considered statistically significant.

3.1 Patient characteristics

Sixteen traumatic hemorrhagic shock patients enrolled in the Microshock study with available blood samples were included. Patient characteristics and clinical parameters are listed in Table 1. Patients had a median age of 38 (23 – 53) years, were mostly male (94%) and had an injury severity score (ISS) of 27 (15 – 34), lowest systolic blood pressure of 94 (62 – 109) mmHg and highest lactate level of 5.9 (3.5 – 13.1) mmol.l^-1. Median sequential organ failure assessment (SOFA) score on admission was 10 (6 – 12). Median time between ED admission and ICU admission was 4 (2 – 8) h, median hospital stay was 18 (6 – 42) days and mortality rate was 25%.

Plasma obtained directly at ED admission decreased in vitro endothelial resistance with 19% compared to plasma from controls (p < 0.001, Fig. 2A and B). This reduction in in vitro endothelial resistance was absent following exposure to plasma obtained at ICU admission and 24 hours later compared to controls (both p > 0.99). Compared to ED admission, plasma collected at ICU admission and 24 hours later increased endothelial resistance (p < 0.05, p < 0.01, Fig. 2A and B).

OPEN IN VIEWER

Fig.2

Loss of in vitro endothelial barrier function. Human endothelial cells were exposed to plasma from traumatic hemorrhagic shock patients collected at admission at the emergency department (ED), intensive care unit (ICU), 24 hours (ICU+24 h) after admission at the ICU and from controls (all n = 8). Endothelial resistance after plasma exposure over time (A) and after 2 hours of plasma exposure (B). A: Data represent mean and were tested with a two-way ANOVA with Bonferroni post-hoc analysis. B: Data represent mean±SD and were tested with a one-way ANOVA with Bonferroni post-hoc analysis. ^*p < 0.05, ^**p < 0.01, ^***p < 0.001 compared to ED admission.

Patients following traumatic hemorrhagic shock showed a decrease in hematocrit over time (Table 1). No association was found between hematocrit and in vitro endothelial resistance (r = –0.33, p = 0.11).

Two parameters can be modelled from endothelial resistance measurements, which distinguish between cell-cell (Rb) and cell-matrix (alpha) integrity (Fig. 3A). Plasma collected at ED admission decreased cell-cell integrity by 35% compared to plasma from controls (p < 0.01), but no differences were measured between the remaining time points and controls (Fig. 3B). No differences in cell-matrix integrity were found between all time points (Fig. 3C).

OPEN IN VIEWER

Fig.3

Impaired in vitro endothelial cell-cell integrity. Rb and alpha can be modelled from endothelial resistance measurements, which distinguish between cell-cell (Rb) and cell-matrix (alpha) integrity (A). Human endothelial cells were exposed to plasma from traumatic hemorrhagic shock patients collected at admission at the emergency department (ED), intensive care unit (ICU), 24 hours (ICU+24 h) after admission at the ICU and from controls (all n = 8). Cell-cell integrity (Rb; B), cell-matrix integrity (alpha; C) and representative images of endothelial cells stained for VE cadherin (adherents junctions; green), F-actin (stress fibers; white) and DAPI (nuclei; blue) (D) after 2 hours of plasma exposure. Data represent mean±SD and were tested with a one-way ANOVA with Bonferroni post-hoc analysis, ^**p < 0.01 compared to ED admission.

Increased intercellular gap formation was observed using immunofluorescence staining of human endothelial cells exposed to plasma from traumatic hemorrhagic shock patients collected at ED admission compared to plasma from controls, which was paralleled by a reorganization of actin filaments from cortical actin distribution to fibers that stretch throughout the endothelial cell body, indicating increased stress-fiber formation (Fig. 3D).

In vitro endothelial resistance was positively associated with in vivo microcirculatory perfusion, as represented by the proportion of perfused vessels and perfused vessel density (PPV: r = 0.56, p < 0.01, Fig. 4 and PVD: r = 0.38, p = 0.07), but not total vessel density (TVD: r = 0.08, p = 0.73), measured at matching time points.

OPEN IN VIEWER

Fig.4

Microcirculatory perfusion associated with in vitro endothelial barrier function. Association between proportion of perfused vessels and in vitro endothelial resistance determined after exposure to plasma collected at corresponding time points. Data are presented with a linear regression with 95% CI and tested with a Pearson’s correlation test.

Patients showed an increase over time in circulating angiopoietin-2 (p < 0.01), whereas circulating angiopoietin-1, soluble Tie2 and von Willebrand factor levels remained unchanged (Table 2). Circulating levels of syndecan-1 and soluble thrombomodulin have previously been reported [9, 10]. Similar results were found if circulating markers were corrected for hematocrit (data not shown). No association was found between in vitro endothelial resistance and angiopoietin-1 (r = –0.57, p = 0.15), soluble Tie2 (r = –0.33, p = 0.43), syndecan-1 (r = –0.14, p = 0.78), soluble thrombomodulin (r = 0.21, p = 0.66) and von Willebrand Factor (r = –0.40, p = 0.33) at ED admission, whereas a negative correlation was found between in vitro endothelial resistance and circulating angiopoietin-2 levels (r = –0.78, p < 0.05).

4.1 Limitations

There were several limitations to the present study. First, a relatively small number of patients was included in the present sub-study. Although powered for in vitro endothelial barrier measurements, additional analyses, such as determining a biomarker of interest, were limited and no hard conclusions could be drawn. Interesting is however, that although sample size was small and samples came from a heterogenic population, significant differences were found in endothelial permeability.

Secondly, the majority of the patients included in this sub-study were male, making it difficult to translate the findings to the female population. However, the population of patients with traumatic hemorrhagic shock is known to be predominantly male. Also, animal studies showed that the female gender exhibits microvascular protection [29], suggesting that the effects of plasma from female patients on in vitro endothelial barrier function should be studied separately.

Finally, the effect of plasma was studied on “healthy” endothelial cells in an in vitro setting, which limits translation to the in vivo situation. However, previous studies confirmed the representability of these cells as a model for endothelial barrier function [12, 30].