All subjects involved in the study were between the age of 18 and 62, healthy, did not suffer from any illness, and were not pregnant (all signed a questionnaire). Male and female donors were equally distributed across the study.

Blood from healthy donors was taken in a heparin flask. Red blood cell (RBC) lysis-buffer (BD Bioscience, Heidelberg, Germany) was added and incubated for 15 min at room temperature. After centrifugation (1,300 rpm, 10 min), the supernatant was discarded to remove lysed RBCs, and the pellet was resuspended in an endothelial cell growth medium (PromoCell, Heidelberg, Germany).

Human umbilical vein cells (PromoCell, Heidelberg, Germany) were seeded (3 × 10⁴ cells per well) in endothelial cell growth medium (PromoCell, Heidelberg, Germany) on the filter of collagen-coated transwells (Corning, Kaiserslautern, Germany) for 8 days (37°C and 5% CO₂) until a confluent monolayer was formed. Complete formation of the monolayer was confirmed by transendothelial electrical resistance (TEER) measurement. Then, medium from the upper well was removed, and erythrocytes-depleted-whole-blood (edWB) was added to mimic the physiological condition of a blood vessel and naturally occurring contact to blood cells.

The HUVECs/edWB coculture was stimulated with 5 μg/ml sDLL1 (Pe-Protech, Hamburg, Germany) or 100 ng/ml LPS (Invivogen, Toulouse, France). For inhibition of Notch, signaling cells were pretreated for 1 h with 2.5 μM DAPT (Sigma-Aldrich, Taufkirchen, Germany). For blocking DLL1 binding, 5 μg/ml anti-DLL1 antibody (AbD3593, Bio-Rad, Puchheim, Germany) or 5 μg/ml anti-Notch1 antibody (BioLegend GmbH, Koblenz, Germany) was added 1 h before stimulation.

Escherichia coli isolate was obtained from a rectal swab of a patient of University Hospital Heidelberg during routine diagnostics. E. coli was cultured overnight on Columbia blood sheep agar at 37°C at 5% CO₂ in a humidified atmosphere. The following day one colony of culture was transferred into Tryptic Soy Broth media and cultured at constant shaking at 200 rpm/37°C until mid-log phase. Then, bacterial suspension was adjusted using absorption measurement. Coculture was infected with multiplicity of infection (MOI) 50. After 2 h of infection, bacteria were killed with gentamicin (Fisher Scientific, Schwerte, Germany).

The HUVECs/edWB coculture was stimulated with 100 ng/ml LPS or an E. coli strain (MOI 50) overnight (Figure 1) or with sDLL1 (5 μg/ml), LPS (100 ng/ml) ± DAPT (1 μM, pretreatment for 1 h), or LPS + anti-DLL1 antibody (5 μg/ml) overnight (Figure 2). After removing edWB and rinsing with phosphate-buffered saline (PBS), HUVECs were analyzed for surface expression of DLL1 (anti-DLL1-APC, Miltenyi Biotec, Bergisch Gladbach, Germany), Notch1 receptor (anti-Notch1-APC, BioLegend GmbH, Koblenz, Germany), ICAM1 (anti-CD54-APC), and E-selectin (anti-CD62E-PE) (both BD Bioscience, Heidelberg, Germany) with antibody staining and flow cytometry. Fluorescence was recorded using the FACS DIVA V 4.12 software on a FACS Canto (BD Biosciences, Heidelberg, Germany).

The HUVECs with or without edWB were stimulated with 100 ng/ml LPS or an E. coli strain (MOI 50) overnight. Blood cells were removed from the coculture by repeated washing with PBS. Total RNA was extracted from HUVECs using the High Pure RNA Isolation Kit (Roche, Mannheim, Germany) according to the protocol of the manufacturer, and cDNA was produced using ReverseAid First Strand cDNA Synthesis Kit (Thermo Fischer Scientific, Karlsruhe, Germany). The cDNA obtained was used for quantitative PCR utilizing the “SYBR Green ROX Mix” (Thermo Fischer Scientific, Karlsruhe, Germany) and the following sequence-specific primers: actin forward 5′-AGA GCT ACG AGC TGC CTG AC-3′, actin reverse 5′-AGC ACT GTG TTG GCG TAC AG-3′, HES1 forward 5′-CTGAAGAAAGATAGCTCGCG-3′, HES1 reverse 5′-ACTTCCCCAGCACACTT-3′, HEY1 forward 5′-AGCCGAGATCCTGCAAGATGA-3′, and HEY1 reverse 5′-GCCGTATGCAGCATTTTCAG-3′.

The HUVECs with or without edWB were stimulated with 100 ng/ml LPS or 5 μg sDLL1 or LPS + anti-DLL1 antibody (5 μg/ml), LPS + anti-Notch1 antibody (5 μg/ml), and LPS + DAPT (1 μM pretreatment for 1 h) for 24 h. The TEER was measured using a volt-ohm meter EVOM² and an STX1 electrode (WPI, Friedberg, Germany) following the protocol of the manufacturers after 0, 0.5, 1, 3, and 24 h. Data were calculated by subtracting resistance values of a transwell without cells and multiplied with the surface area (1.12 cm²) of the transwell. To confirm that cell numbers are equal in each condition during the experiment, we harvested cells of all conditions by trypsinization after 24 h and counted trypan blue stained cells.

The HUVECs, cultivated with edWB, were stimulated using 100 ng/ml LPS or 5 μg/ml sDLL1 or LPS + anti-DLL1 antibody (5 μg/ml), LPS + anti-Notch1 antibody (5 μg/ml), and LPS + DAPT (1 μM pretreatment for 1 h). Post stimulation, 15 μg/ml (0.1 nmol/ml) fluorescein isothiocyanate (FITC)-dextran (Sigma Aldrich, Taufkirchen, Germany) was added to the upper chamber of each well. After 0, 0.5, 1, 3, and 24 h, 50 μl from the lower chamber of each well was transferred into a Black 96 Well Microplate (Greiner Bio-One, Frickenhausen, Germany) and Black 96-Well Plate Reader (LUMIstar OPTIMA, BMG LABTECH, Ortenberg, Germany) and measured at 488 nm excitation and 555 nm emission. In addition, a serial dilution of FITC-dextran was added to the plate in each experiment as a standard for the calculation of dextran concentration.

Coculture was stimulated with sDLL1 (5 μg/ml) for 3 h or left unstimulated. After removing edWB, the HUVECs were fixed (2% PFA in PBS, 30 min at RT) on the transwell membrane and permeabilized with 0.1% Triton-X in PBS (TX-PBS) for 15 min at RT. Non-specific binding sites were blocked with TX-PBS 5.5% FCS for 1 h at RT. Staining was performed with primary antibody (45 min) against VE-Cadherin and secondary antibody (1 h) anti-mouse-Cy5 or anti-ZO-1 (all three Thermo Fischer Scientific, Karlsruhe, Germany) and secondary antibody anti-rabbit-DyLight488 (Abcam, Cambridge, United Kingdom). All antibodies were diluted in PBS to a final concentration of 5 μg/ml. Actin was visualized later with phalloidin-tetramethylrhodamine B isothiocyanat, 1:2,000 in PBS for 45 min (Sigma Aldrich, Taufkirchen, Germany). DNA was stained (20 min) with Hoechst-33342, 1:10,000 in PBS (Thermo Fischer Scientific, Karlsruhe, Germany). Membranes with stained cells were transferred to glass slides and mounted on a microscope (ROTI^®Mount Aqua, Carl Roth GmbH + Co., KG, Karlsruhe, Germany). Images were taken using a Leica SP8 confocal microscope and LAS X software (Leica Microsystems GmbH, Wetzlar, Germany). Pictures were analyzed, processed, and quantified in Fiji/ImageJ (Wayne Rasband). For quantification of Ve-Cadherin and ZO-1, microscopic images were converted into black and white, and integrated density was measured from total cells and from the cytoplasm of cells. Then, corrected total fluorescence (CTCF) was calculated using the following formula: CTCF = Integrated density – (Area of selected cell × Mean fluorescence of background readings).

Statistical analysis was performed using Graphpad Prism version 9 (San Diego, California, United States). Values were presented as mean with standard deviation. Comparison of quantitative variables between groups was calculated using the non-parametric Mann-Whitney U-test. The level of significance was set at a p-value ≤ 0.05.

This study was carried out in accordance with the recommendations of the ethics committee of the Medizinische Fakultät, Heidelberg (S-157/2006). All subjects gave written informed consent in accordance with the Declaration of Helsinki.

First, we assessed the ability of HUVECs to respond to bacterial stimulation with upregulation of DLL1 and DLL1-receptor Notch1. The HUVECs/edWB coculture was stimulated with LPS or infected with an E. coli strain (MOI 50) and analyzed using antibody staining and flow cytometry for the induction of DLL1 and Notch1 receptor surface expression. LPS stimulation and in vitro infection with E. coli triggered the expression of ligand and receptor significantly (Figure 1A). Then, we confirmed that bacterial stimulation induced Notch signaling by performing quantitative real-time (rt)-PCRs for Notch target genes Hes1 and Hey1. LPS as well as E. coli stimulation induced a pronounced gene induction of both target genes (Figure 1B). Stimulation of HUVECs without edWB induced only a minor induction of target genes implicating the need for activated blood cells (Figure 1C). Having shown that whole E. coli activates Notch signaling in HUVECs, the following experiments were performed exclusively with LPS as one major stimulatory component of gram-negative bacteria such as E. coli.

To verify the involvement of Notch signaling in LPS-stimulated vEC activation, HUVECs/edWB coculture was stimulated with LPS and pretreated with Notch inhibitor DAPT. Flow cytometry analysis revealed that LPS stimulation significantly induced the endothelial activation marker ICAM-1 and E-selectin, as expected. Inhibition of the Notch cascade through DAPT diminished the expression significantly, implicating the involvement of Notch signaling in LPS-induced endothelial cell activation. Importantly, pretreatment with an anti-DLL1 blocking antibody also dampened the LPS-induced activation, implicating a role for DLL1-induced Notch signaling (Figures 2A,B). To further investigate the impact of DLL1, the coculture was stimulated with recombinant soluble DLL1 (sDLL1), which significantly upregulated ICAM-1 as well as E-selectin on HUVECs (Figures 2A,B).

To evaluate the impact of Notch signaling on the integrity of vascular endothelial monolayer, the barrier function of HUVECs was analyzed. First, the integrity of HUVECs was determined by TEER measurement across the cellular monolayer. Post stimulation, TEER was measured at different time points. LPS stimulation of HUVECs without edWB induced a minor, although significant, decrease in resistance after 1 and 3 h [Δ mean 1 h (3 h) in Ω: 2.93 (2.30); both p = 0.05*]. Similar finding was observed for stimulation with sDLL1 [Δ mean 1 h (3 h) in Ω: 0.9 (1.7); p = 0.05* (p = 0.412)] (Figure 3A). However, in the HUVECs/edWB coculture, LPS stimulation induced a pronounced decrease in resistance of HUVEC monolayer after 1 h (Δ mean in Ω: 6.4: p = 0.05*) that stayed stable up to 3 h (Δ mean in Ω: 7.1; p = 0.05*), confirming once again that blood cells are important in terms of loosening endothelial cell structure. After 24 h, the decrease in resistance faded (Δ mean in Ω: 2.6; p = 0.06), and the integrity began to stabilize again. sDLL1 mediated a distinct decrease in TEER after 1 h of stimulation (Δ mean in Ω: 4.07; p = 0.05*). This effect was stable for 24 h and slightly increased further during that time (Δ mean in Ω: 6.80; p = 0.05*) (Figure 3B). Importantly, blocking DLL1/receptor binding via an anti-DLL1 antibody and an anti-Notch1 antibody diminished the LPS-induced decrease in resistance significantly after 1 and 3 h [anti-DLL1: Δ mean 1 h (3 h) in Ω: 3.2 (2.93); anti-Notch1: Δ mean 1 h (3 h) in Ω: 3.1 (3.6); all p = 0.05*]. The same was true for inhibiting Notch signaling via DAPT (Δ mean 1 h [3 h] in Ω: 2.9 (4.2); both p = 0.05*] (Figure 3C).

Next, fluorescence-based transwell assays were performed. FITC-dextran was added on top of the transwells. Over time, transmigration of dextran in the lower well was analyzed. Comparable to the TEER measurement, transmigration of dextran started after 1 h of LPS and DLL1 stimulation. sDLL1 increased transmigration around 3-fold after 3 h of stimulation and almost 5-fold after 24 h. LPS-mediated transmigration of dextran increased significantly more than 6-fold after 3 h and stabilized to around 4-fold after 24 h (Figure 3D). Importantly, anti-DLL1 antibody, anti-Notch1 antibody, and DAPT could dampen the LPS-associated transmigration of dextran almost by half (Figure 3E). According to these results, DLL1-mediated Notch signaling mediates an increase in the permeability of HUVEC monolayers.

We further characterized the effect of sDLL1 and Notch signaling on endothelial cell integrity by antibody staining and confocal microscopy imaging techniques. Cells were analyzed for morphological changes and expression and localization of VE-cadherin, the major component of adherens junctions. Furthermore, ZO-1, as an important regulator of adherens junctions and barrier stability, was analyzed. Stimulation with sDLL1 for 3 h clearly modulated the structure of the HUVEC monolayer. The tight structure that can be observed in the image of the unstimulated sample started to loosen and gaps began to form between cells (Figure 4). In the unstimulated HUVECs, VE-cadherin (CD144) (Figure 4A) and ZO-1 (Figure 4B) locate mostly in the plasma membrane, showing the tight cohesion between single cells. However, stimulation with sDLL1 seems to mediate the translocation of the proteins into the cytoplasm. The increased appearance in the cytoplasm is significant for ZO-1 and not significant but replicable for VE-cadherin (Figures 4A–C).