Dynamic and functional changes in vascular endothelial cell monolayers under combined hypoxic and inflammatory stimuli Authors: Kazuki Sone, Ai Funatsu, So Sampei, Kenichi Funamoto Published: 2025 ABSTRACT AbstractIn sepsis, dysregulated inflammatory responses cause microthrombi formation, leading to organ hypoxia. While endothelial cell (EC) responses to inflammation are well-studied, the combined effects of hypoxia and inflammation remain unclear. We investigated EC dynamics and function under simultaneous hypoxia and lipopolysaccharide (LPS) exposure using microfluidic devices with controllability of oxygen concentration. Collective cell migration was measured by particle image velocimetry, while monolayer permeability was assessed by diffusion of fluorescently labeled dextran. Additionally, we examined adhesion and signaling protein expression. Migration speed decreased under approximately 1% O2, but LPS exposure delayed this decrease, attenuating hypoxic responses. Both hypoxia and LPS increased permeability, though this effect was partially suppressed by LPS preconditioning at a specific concentration. LPS exposure increased ICAM-1 and VCAM-1 expression on endothelial surfaces, except for VCAM-1 under hypoxia, suggesting complex interactions between HIF-1α and NF-κB pathways activated by hypoxia and LPS, respectively. NF-κB inhibition suppressed both collective migration and permeability, confirming its role in these dynamic and functional changes. These findings provide novel insights into EC phenotypic alterations under combined hypoxic and inflammatory conditions in an in vitro model that effectively recapitulates key aspects of sepsis pathophysiology.Supplementary InformationThe online version contains supplementary material available at 10.1038/s41598-025-31172-1.Keywords: Hypoxia, Inflammatory response, Vascular endothelial cells, Microfluidic device, Collective cell migration, Vascular permeability, Lipopolysaccharide, SepsisSubject terms: Cell biology, Diseases, Medical research, Physiology Introduction Sepsis is a condition in which the body’s response to infection (inflammatory response) becomes dysregulated, leading to serious organ dysfunction. When pathogens (bacteria, viruses) enter the body, the innate immune system is activated through Toll-like receptors (TLRs), resulting in massive production of inflammatory cytokines (such as TNF-α and IL-6) and chemokines1–4. These biochemical substances contribute to physiological responses such as capillary dilation and increase vascular permeability. However, when uncontrolled inflammatory responses occur due to excessive production of cytokines (cytokine storms), they can lead to systemic inflammatory response syndrome (SIRS) and multiple organ failure5,6. As an experimental model to simulate the pathophysiology of sepsis, the method of administering lipopolysaccharide (LPS) to experimental animals has been widely used7,8. LPS administration has been reported to rapidly increase the production of inflammatory cytokines and induce various diseases including acute organ injury and Parkinson’s disease9,10. Polyinosinic:polycytidylic acid (Poly I:C), a synthetic analog of double-stranded RNA, is also used as a model to mimic viral infection11. While LPS triggers inflammatory responses mainly against bacterial infections via TLR4, Poly I:C induces antiviral responses primarily consisting of type I interferons via TLR311,12. Experimental systems using these different pathogen-associated molecular patterns have provided important insights into understanding the differences in immune responses depending on the type of pathogen13. Hypoxic conditions within the body are also important factors in sepsis. The internal environment of the body is hypoxic compared to the atmosphere (21% O2 (160 mmHg)), with oxygen levels decreasing to 13.2% O2 (~ 100 mmHg) in oxygen-rich arterial blood and to 5.3% O2 (~ 40 mmHg) in venous blood14. When cells sense surrounding hypoxia, the intracellular transcription factor, namely hypoxia-inducible factor (HIF)-1α, accumulates within the cell and translocates to the nucleus, ultimately promoting the expression of genes involved in processes such as angiogenesis15–17. In sepsis, significant tissue hypoxia occurs due to a combination of factors: microthrombus formation caused by inflammatory cytokines, tissue edema and perfusion disorders resulting from increased vascular permeability, and increased oxygen consumption by activated immune cells18–20. Furthermore, HIF-1α interacts with nuclear factor-kappa B (NF-κB), which is involved in inflammatory responses. It has been reported that HIF-1α can indirectly enhance inflammatory responses by promoting the production of inflammatory cytokines, while conversely, under certain conditions, it can also suppress excessive inflammatory responses by inhibiting the activation of NF-κB21–23. Thus, while both hypoxic and inflammatory responses play important roles in the pathological progression of sepsis, many aspects of their interactions remain unclear. Vascular endothelial cells (ECs) covering the blood vessel lumen are constantly exposed not only to mechanical stimuli such as shear stress and pressure from blood flow, but also to chemical stimuli from blood components (nutrients, inflammatory cytokines, etc.). Inflammatory cytokines activate the intracellular transcription factor NF-κB, promoting the expression of cell adhesion molecules such as intercellular adhesion molecule (ICAM)-1 and vascular cell adhesion molecule (VCAM)-1 on the vascular endothelial surface24–26. These molecules are involved in the adhesion of immune cells to the vascular endothelium and play important roles in the induction and regulation of immune responses. Furthermore, ECs are also exposed to fluctuations in blood oxygen concentration. Vascular endothelial cells forming a monolayer maintain tissue integrity by collectively migrating in response to changes in environmental factors while influencing adjacent cells27–29. This collective migration is important not only for maintaining vascular homeostasis but also in various physiological and pathological processes such as angiogenesis, tissue regeneration, and tumor formation30–33. Understanding these processes has significant physiological and clinical implications. Previous studies using wound-healing assays have revealed that both inflammatory cytokines and hypoxic stress independently increase the migration of ECs34,35. However, these studies have primarily focused on tissue regeneration and wound healing, and the dynamics of normal ECs maintaining a monolayer have not been sufficiently examined. In addition, in previous in vitro experiments on inflammation, physiological oxygen concentrations in the body have often not been considered for the sake of experimental simplification. Therefore, the combined effects of oxygen concentration and chemical stimuli due to inflammation on EC dynamics, as well as their underlying mechanisms, remain unclear. Understanding how the collective migration of ECs in monolayers is disrupted under simultaneous hypoxic and inflammatory stimuli provides novel insights into mechanisms governing vascular homeostasis and identifies potential therapeutic targets aimed at preserving vascular integrity and function during sepsis. In recent years, cell experiments using microfluidic devices have become increasingly popular as in vitro research. By creating micro-sized channels using polydimethylsiloxane (PDMS), which is transparent and highly permeable to gases, and culturing cells within these channels, it is possible to observe cell dynamics in real-time while strictly controlling environmental factors around the cells. We have developed unique microfluidic devices that precisely and instantaneously control the oxygen concentration around cells36,37. In addition to the media channel for cell culture, the devices include gas channels for supplying gas mixtures with adjusted oxygen concentrations, achieving control of oxygen levels through gas exchange via PDMS. Using the developed devices, we have elucidated changes in the dynamics and functions of ECs in hypoxic environments38–44. Specifically, we found that the collective migration of ECs varies depending on the surrounding oxygen concentration, reaching its maximum at around 2% O2, while decreasing in extremely low oxygen environments below this level39,40,44. We also investigated the combined effects of chemical stimulation by glucose and oxygen concentration on cell migration, demonstrating the involvement of mitochondrial aerobic energy production in oxygen-dependent changes in migration39,40. Moreover, by quantifying the diffusion of fluorescently labeled dextran passing through a monolayer of ECs formed on the surface of collagen gel, we evaluated the permeability of the EC monolayers38. The results suggested that permeability increases in hypoxic environments, indicating changes in the barrier function of vascular endothelium. These experimental techniques using microfluidic devices are valuable for evaluating dynamics when exogenous inflammatory stimuli, such as LPS during sepsis, and hypoxic stress simultaneously affect ECs. The purpose of this study is to elucidate the changes in dynamics and functions of ECs and their mechanisms in response to simultaneous exposure to hypoxic stress and inflammatory stimulation by LPS. An EC monolayer was formed in a microfluidic device capable of controlling oxygen concentration, and LPS at various adjusted concentrations was added to the cell culture medium to provide inflammatory stimulation to the cells (Fig. 1). Gas mixtures with adjusted oxygen concentrations were supplied to the gas channels to control the oxygen concentration around the cells by facilitating gas exchange between the channels where cell culture was performed. Time-lapse observations of cell dynamics under simultaneous control of oxygen and LPS concentrations were conducted using phase-contrast microscopy. The time-series microscopy images were then analyzed using particle image velocimetry (PIV) to measure the collective migration velocity of the EC monolayer. Moreover, fluorescently labeled dextran solution was injected into the right media channel, and its diffusion through the EC monolayer formed on the gel interface was quantified, thereby evaluating permeability. After these cell experiments, the cells were fixed and subjected to immunofluorescence staining to examine several key proteins: the cell adhesion molecules ICAM-1 and VCAM-1; vascular endothelial (VE)-cadherin, which mediates cell–cell adhesion; HIF-1α, the master regulator of hypoxic response; and NF-κB, a transcription factor that orchestrates inflammatory responses. To further elucidate the underlying mechanisms of the observed changes, additional experiments were conducted using an NF-κB inhibitor. Schematics of microfluidic device45 and cell experiments. (a) Perspective view generated using 3D modeling software (Rhinoceros ver. 5 SR14, Robert McNeel & Associates, USA, https://www.rhino3d.com/), (b) enlarged cross-sectional view in the zx-plane (y = 0 mm), and (c) schematic of migration and permeability measurements with the regions of interest. Results Variations in the collective cell migration by hypoxic and inflammatory stimuli Migration velocity was measured by performing PIV analysis on time-series phase-contrast microscopy images of the EC monolayers formed on the bottom surface of the media channel in the microfluidic device (Fig. 1c). Under normoxic condition N (21% O2), the spatial average migration speed in the absence of LPS (control condition) showed a nearly constant value of ~ 5 µm/h throughout the experimental period (Fig. 2a). Following acute inflammatory stimulation with LPS, the migration speed remained comparable to the control condition for approximately the first 2 h. Subsequently, cells treated with 0.1 μg/ml LPS exhibited a slight increasing trend in migration speed, whereas those exposed to 1 μg/ml LPS demonstrated a modest decreasing trend. Under hypoxic condition H1 (1.3% O2), the spatial average migration speed in the absence of LPS decreased over time, reaching approximately 3 µm/h (Fig. 2b). Throughout the experimental period, the migration speed remained consistently higher with 1 μg/ml LPS compared to other LPS conditions. With 0.1 μg/ml LPS, migration speed gradually decreased after 2 h and eventually fell below values observed in the LPS-free conditions. Here, we quantitatively assessed cell proliferation during the experimental period by calculating the cell density ratio between the start and end points of our observations (Fig. S1a). This ratio was approximately 1.0 across all experimental conditions, demonstrating that cell populations remained stable throughout the experiment. These data conclusively indicate that the observed differences in migration patterns cannot be attributed to changes in cell density, thus validating that our migration measurements reflect genuine cellular responses to the experimental conditions rather than artifacts of varying confluency. Under moderate hypoxic condition H5 (5.3% O2), the migration speed tended to be high in the absence of LPS, likely due to the hypoxic stress response (Fig. S2a). Stimulation with 0.1 μg/ml LPS decreased migration speed, whereas 1 μg/ml LPS had minimal effect. After 1 h of the experiment under normoxia, almost the same values were observed for all LPS conditions (Fig. 2c). In contrast, under hypoxic condition H1, a general trend toward reduced migration was observed, except at 1 μg/ml LPS where this reduction was notably absent. After 6 h of the experiment, the migration speed under normoxia exhibited a non-monotonic response to LPS concentration, peaking at 0.1 μg/ml and subsequently decreasing with higher concentration (Fig. 2d). Under severe hypoxic condition H1, the migration speed decreased at LPS concentrations of 0.1 μg/ml, but increased at an LPS concentration of 1 μg/ml. By normalizing the migration speeds by the value obtained immediately after the start of the experiment under the control condition, the time-dependent changes in migration speed according to oxygen concentration and LPS concentration become clear (Fig. 2e,f). A decrease in migration speed by hypoxic exposure at an extremely low oxygen condition occurred rapidly (Fig. 2e). The distribution of migration speed changed after 6 h, resulting in the oxygen concentration dependence of migration speed becoming pronounced under exposure to low concentration LPS (0.1 μg/ml) due to the decrease in migration caused by hypoxic stress (Fig. 2f). In contrast, under exposure to 1 μg/ml LPS, the oxygen concentration dependence of migration speed became less distinct. The total migration distance during the 6-h experimental period was further evaluated by integrating the migration speed over time (multiplying the instantaneous speed by each time interval and summing these values) (Fig. S3a). Analysis revealed that variations in total migration distance under different hypoxic and inflammatory conditions followed similar trends to those observed for average migration speed. Specifically, the total migration distance significantly decreased under hypoxic conditions compared to normoxic conditions, with the exception of inflammatory stimulation by 1 μg/ml LPS under hypoxia, which showed a relatively large total migration distance. Collective migration of HUVECs under acute exposure to hypoxia and LPS at different concentrations of 0, 0.1, or 1 µg/ml. Average migration speed, V, over time for 6 h under (a) normoxia (N: 21% O2) and (b) hypoxia (H1: 1.3% O2), and the average migration speed at (c) 1 h and (d) 6 h under different oxygen and LPS conditions. Error bars show the standard deviation. Significant differences in migration speed between the oxygen and LPS conditions were assessed by two-way ANOVA followed by post hoc Tukey’s test for multiple comparisons. *P < 0.05; ***P < 0.001. Contour map of relative migration speed over that under the control condition (normoxia and no LPS stimulus) with respect to the oxygen concentration and LPS concentration at (e) 1 h and (f) 6 h. Preconditioning by exposing cells to LPS for 24 h to simulate a long inflammatory state resulted in different patterns of time-dependent changes in migration speed under several conditions (Fig. 3). Under normoxic condition N, migration decreased in the presence of LPS (Fig. 3a). Unlike during acute exposure to LPS, the migration speed was slowest at an LPS concentration of 1 μg/ml. The migration speed under hypoxic condition H1 showed similar time-dependent patterns across all LPS conditions (Fig. 3b). In all cases, migration speeds remained relatively stable for the first 2 h, after which they decreased to approximately 3 µm/h. Under hypoxic condition H5, the migration speed also tended to decrease at all LPS concentrations, with a particularly pronounced decrease in the case of 0.1 μg/ml LPS (Fig. S2b). Preconditioning with exposure to LPS under normoxia tended to decrease migration (Fig. 3c). After 6 h, migration decreased by LPS stimulation, and the specific increase in migration speed induced by acute inflammatory stimulus of 1 μg/ml LPS under hypoxia was not observed (Figs. 2d and 3c). The cell increase rate during the experimental period was approximately 1.0 across all conditions, confirming that cell density remained stable and therefore did not influence the observed migration patterns (Fig. S1b). The normalized migration speeds after 6 h of the experiment with preconditioning tended to decrease overall (Fig. 3d). The oxygen concentration dependence became apparent in migration speed with relatively smaller migration speeds under hypoxia. The total migration distance remained consistently low under hypoxic conditions regardless of LPS concentration, while it exhibited a dose-dependent decrease with increasing LPS concentration under normoxic conditions (Fig. S3b). Collective migration of HUVECs under different oxygen conditions and long exposure to LPS (preconditioning (PreC)) at different concentrations of 0, 0.1, or 1 µg/ml. Average migration speed, V, over time for 6 h under (a) normoxia (N: 21% O2) and (b) hypoxia (H1: 1.3% O2), and (c) the average migration speed at 6 h under different oxygen and LPS conditions. Error bars show the standard deviation. Significant differences in migration speed between the oxygen and LPS conditions were assessed by two-way ANOVA followed by post hoc Tukey’s test for multiple comparisons. †P < 0.1; *P < 0.05; ***P < 0.001. (d) Contour map of relative migration speed over that under the control condition (normoxia and no LPS stimulus) with respect to the oxygen concentration and LPS concentration at 6 h. Changes in the permeability of the vascular endothelial cell monolayer The permeability of the EC monolayers formed on the interface between the gel and media channels was quantitatively evaluated using fluorescently labeled dextran solution. When 40 kDa fluorescently labeled dextran solution was injected into the media channel, diffusion through the EC monolayers into the gel was observed (Fig. 4a). A rectangular region 2,048 × 200 pixels (1,272 × 124 μm) including the gel interface with the EC monolayer was set in the fluorescence microscopy image, and the distribution of the average intensity in the y-direction at each x-coordinate was calculated and normalized by the intensity in the media channel (Fig. 4b). Due to the presence of the EC monolayer, there was a gap in intensity, and the intensity decreased linearly in the -x direction due to diffusion into the gel. Under the control condition, the fluorescence intensity in the gel was low, suggesting that permeability was maintained at a low level. In contrast, under other conditions, the intensity was higher, indicating relatively high permeability. The permeability under normoxia tended to increase with increasing LPS concentration (Fig. 4c). On the other hand, under hypoxia, permeability was higher than the control condition. With preconditioning by LPS supplementation, similar permeability increase was observed under normoxia (Fig. 4d). Under hypoxia, in the cases of 0.1 μg/ml LPS, permeability was slightly lower compared to the condition without LPS stimulation but tended to be higher than the control condition. In the case of 1 μg/ml LPS, permeability similar to the condition without LPS was observed. Similar permeability variations were observed when using dextrans of different molecular sizes (10 or 70 kDa) (Fig. S4). Under control conditions, a size-selective barrier function was evident; however, this selectivity became less pronounced when hypoxic or LPS stimulation was applied. Evaluation of permeability of HUVEC monolayers formed at the gel interface by observing diffusion of 40 kDa fluorescently labeled dextran. (a) Representative phase-contrast fluorescent microscopy images and (b) the x-directional profiles of the normalized fluorescence intensity after 10 min of the infusion of the dextran solution into the right media channel after 6-h exposure to different oxygen and LPS concentrations. In each image, the gel and media channels are on the left and right, respectively. Scale bar represents 200 μm. The fluorescence intensity was normalized by dividing it by the average intensity in the media channel. Box-whisker plots of permeability values of HUVEC monolayers under (c) acute or (d) long inflammatory stimulus (preconditioning (PreC)) with LPS and different oxygen conditions measured at four arbitrary locations in three microfluidic devices (N = 3, 12 interfaces in total). Significant differences in permeability between the oxygen and LPS conditions were assessed by two-way ANOVA followed by post hoc Tukey’s test for multiple comparisons. †P < 0.1; **P < 0.01; ***P < 0.001. Expression and localization of adhesion and signaling proteins Intracellular signaling during exposure to hypoxia and LPS was investigated by immunofluorescence staining on proteins in ECs on the bottom surface of the media channel, followed by microscopic observation. ICAM-1 and VCAM-1 were expressed on the cell surface (Fig. 5a). The intensity of ICAM-1 in fluorescence microscopy images significantly increased with LPS stimulation, particularly showing a notable increase in the case of 0.1 μg/ml LPS after preconditioning (Figs. 5b,c). Moreover, with preconditioning, there was little difference between oxygen conditions. The intensity of VCAM-1, similar to ICAM-1, tended to increase with the LPS stimulation under normoxia (Figs. 5d,e). With preconditioning, the expression of VCAM-1 under normoxia greatly increased at an LPS concentration of 0.1 μg/ml. In contrast, during simultaneous exposure to hypoxic and LPS stimuli, the increase of VCAM-1 expression did not occur with the LPS stimulation but rather decreased compared to under normoxia. Particularly, under exposure to 0.1 μg/ml LPS, the expression of VCAM-1 significantly decreased. Under hypoxic conditions, there was little difference in VCAM-1 expression with or without preconditioning. Here, VE-cadherin, responsible for cell–cell adhesion, was expressed at the cell periphery (Fig. S5a). Microscopy images showed that the HUVECs maintained a typical cobblestone morphology with no specific polarity under the static or non-directional flow conditions. The relative area of VE-cadherin to the total cell area, A*cad, showed no significant differences at any oxygen and LPS concentrations, and no clear differences were observed (Fig. S5b,c). Expression of ICAM-1 and VCAM-1 in HUVECs after 6 h of exposure to hypoxia and LPS with/without preconditioning (PreC) by LPS. (a) Representative images of maximum intensity projections of confocal microscopy images of HUVECs to the xy-plane. ICAM-1 and VCAM-1 were stained with red and green, respectively. The fluorescence intensity of ICAM-1 integrated over the whole image and normalized by that under the control condition (normoxia and no LPS stimulus), I*ICAM-1, under (b) acute or (c) long inflammatory stimulus with LPS and different oxygen conditions. The fluorescence intensity of VCAM-1 integrated over the whole image and normalized by that under the control condition, I*VCAM-1, under (d) acute or (e) long inflammatory stimulus with LPS and different oxygen conditions. Error bars show the standard deviation. Significant differences in expression of ICAM-1 and VCAM-1 between the oxygen and LPS conditions were assessed by two-way ANOVA followed by post hoc Tukey’s test for multiple comparisons. †P < 0.1; *P < 0.05; **P < 0.01; ***P < 0.001. HIF-1α translocated into the cell nucleus under hypoxic stress (Fig. 6a). The nuclear translocation ratio of HIF-1α, αHIF-1α, increased by hypoxic stress and 0.1 μg/ml LPS stimulation (Fig. 6b). Under hypoxic conditions, 1 μg/ml LPS stimulation relatively suppressed the hypoxia-induced increase in nuclear translocation of HIF-1α. With preconditioning, the increase in nuclear translocation of HIF-1α by 0.1 μg/ml LPS was attenuated under normoxic conditions (Fig. 6c). Under hypoxic conditions, nuclear translocation rates were comparable regardless of preconditioning. Expression and nuclear translocation of HIF-1α in HUVECs after 6 h of exposure to hypoxia and LPS with/without preconditioning (PreC) by LPS. (a) Representative images of maximum intensity projections of confocal microscopy images of HUVECs to the xy-plane. HIF-1α and nuclei were stained with green and blue, respectively. Variation of the average intensity of HIF-1α in nucleus relative to that in the whole image, αHIF-1α, under (b) acute or (c) long inflammatory stimulus with LPS and different oxygen conditions. Error bars show the standard deviation. Significant differences in nuclear translocation of HIF-1α between the oxygen and LPS conditions were assessed by two-way ANOVA followed by post hoc Tukey’s test for multiple comparisons. *P < 0.05; **P < 0.01; ***P < 0.001. Modulation of the cell migration and permeability by PFI-1 treatment To examine the effect of the inflammatory response on cell dynamics and function, NF-κB was stained after 1 h of the experiment. The transcription factor NF-κB was highly expressed in the cell nucleus (Fig. 7a). The integrated intensity of NF-κB in the nucleus significantly increased at 0.1 μg/ml LPS, showing its maximum under each oxygen condition (Fig. 7b). Under both oxygen conditions, the integrated intensity at 1 μg/ml LPS tended to be higher than that under the control condition, though no significant difference was observed. The nuclear translocation ratio of NF-κB, αNF-κB , tended to increase with LPS concentration under normoxia, but slightly decreased in the case of 0.1 μg/ml LPS under hypoxia (Fig. 7c). Expression and nuclear translocation of NF-κB in HUVECs after 1 h of exposure to hypoxia and LPS. (a) Representative images of maximum intensity projections of confocal microscopy images of HUVECs to the xy-plane. NF-κB and nuclei were stained with green and blue, respectively. (b) The fluorescence intensity of NF-κB in nuclei, normalized by that under the control condition (normoxia and no LPS stimulus), In*NF-κB, and (c) the variation of the average intensity of NF-κB in nucleus relative to that in the whole image, αNF-κB, under different oxygen and LPS conditions. Error bars show the standard deviation. Significant differences in nuclear translocation of NF-κB between the oxygen and LPS conditions were assessed by two-way ANOVA followed by post hoc Tukey’s test for multiple comparisons. †P < 0.1; *P < 0.05. The collective migration and permeability of ECs were evaluated when hypoxic and LPS stimuli were applied with PFI-1, which inhibits the transcriptional activity of NF-κB, supplemented to the cell culture medium. The PFI-1 supplementation decreased migration under both oxygen conditions, with little difference due to LPS concentration (Fig. 8a,b). The migration speed remained nearly constant throughout the experimental period under normoxia. Whereas, under hypoxia, it maintained high migration for approximately the first 2 h of the experiment before decreasing (Fig. 8b), which was similar to the case when 0.1 μg/ml LPS was added without PFI-1 supplementation (Fig. 2b). Comparing migration speeds after 6 h of the experiment, migration decreased in all conditions compared to the control condition, particularly under hypoxia where the migration speed was ~ 2.5 μm/h, about half the value of the control condition (Fig. 8c). Moreover, permeability decreased to levels comparable to the control condition under all conditions (Fig. 8d). Effects of PFI-1 treatment on HUVEC monolayers. Average migration speed, V, over time for 6 h under (a) normoxia (N: 21% O2) and (b) hypoxia (H1: 1.3% O2), and (c) the average migration speed at 6 h under different oxygen, LPS, and PFI-1 conditions. Error bars show the standard deviation. (d) Box-whisker plots of permeability values Pm of HUVEC monolayers under different oxygen, LPS, and PFI-1 conditions measured at four arbitrary locations in three microfluidic devices (N = 3, 12 interfaces in total). Significant differences of the results between the conditions were assessed by one-way ANOVA followed by post hoc Tukey’s test for multiple comparisons. ***P < 0.001 versus control condition without PFI-1. Discussion Hypoxic and inflammatory stimuli complexly alter the dynamics and function of ECs. The migration ability of ECs is oxygen concentration-dependent at low LPS concentrations, but at high LPS concentrations, the influence of LPS becomes dominant. The permeability of the EC monolayers is enhanced by hypoxia and LPS stimulation. Additionally, during simultaneous exposure to hypoxic and LPS stimuli, there exist conditions where migration ability specifically increases or permeability decreases compared to when only hypoxic stress is applied. Furthermore, cell adhesion molecules (ICAM-1 and VCAM-1) that are expressed on ECs during inflammatory response and promote the adhesion of immune cells from the blood exhibit different expression patterns depending on oxygen condition, and intracellular signaling via HIF-1α and NF-κB is involved in these changes in cell dynamics and function. The migration speed of ECs decreased under hypoxia at 1.3% O2 compared to normoxia in the absence of LPS (Fig. 2). This was consistent with results from previous studies39,40,44. The significant decrease in cell migration ability under extremely low oxygen environments is due to the reduction in aerobic energy production within ECs40. Under normoxia, no clear changes occurred in the migration speed of ECs even when acute inflammatory stimulation was applied by adding LPS (Fig. 2a). In contrast, when long inflammatory stimulation was applied through preconditioning by exposing cells to each concentration of LPS for 24 h, the migration speed decreased with increasing LPS concentration (Fig. 3c). These results suggest that under oxygen-rich conditions, the effect of relatively low concentrations of LPS is limited, but migration is reduced when the stimulation persists over a long period. Without LPS stimulation, migration speed decreased after the hypoxic exposure (Figs. 2b and 3b). In contrast, when hypoxic and LPS stimuli were simultaneously applied, migration speed maintained its initial values for approximately 2 h before subsequently decreasing. This indicates the possibility that inflammatory stimulation by LPS temporarily attenuates the hypoxic response. Inflammatory stimulation by 0.1 μg/ml LPS acted to maintain cell migration in the initial stages, overcoming the effect of insufficient energy production in a hypoxic environment. However, after 2 h, the influence of hypoxic stress became dominant over time, and the migration speed decreased. In the case of an LPS concentration of 1 μg/ml, similar to the results under normoxia, the migration speed was equivalent to the control condition throughout the experimental period (Fig. 2a,b). Hence, the inflammatory stimulation by this concentration of LPS had a greater impact on cell migration ability than hypoxia, and the effect of maintaining migration throughout the experimental period overcame the effect of insufficient energy production due to hypoxia. When preconditioning was performed by adding LPS, the migration speed after hypoxic exposure showed almost constant values for about 2 h regardless of LPS concentration, and then gradually decreased (Fig. 3b). This change was similar to the condition where hypoxia and LPS at a concentration of 0.1 μg/ml were acutely applied (Fig. 2b). These findings indicate that the migration-enhancing effect of inflammatory stimulation by LPS is well manifested in the initial stages, and the influence of hypoxic stress becomes dominant in long inflammatory states. The permeability of the EC monolayers increased in accordance with increasing LPS concentration when acute inflammatory stimulation was applied under normoxia (Fig. 4c). This increase in permeability is consistent with results from numerous previous studies46–48. The size selectivity observed under conditions without LPS stimulation, where the permeability decreased with larger molecular sizes of fluorescently labeled dextran, became less distinct (Figs. 4 and S4). When preconditioning was performed by exposing cells to LPS stimulation for 24 h, the permeability also increased with LPS concentration, though the differences between LPS concentration conditions became smaller (Fig. 4d). Compared to acute inflammatory conditions, the permeability increased under long inflammatory conditions. It is evident that inflammatory stimulation by LPS enhances permeability with increasing concentration and exposure time. When hypoxic stress was applied, the permeability of EC monolayers was enhanced at all LPS concentrations (Figs. 4c, S4a,b), indicating that the influence of hypoxia is more dominant than acute inflammatory stimulation in enhancing permeability. Notably, when preconditioning was performed using 0.1 μg/ml LPS, the permeability decreased compared to other LPS concentration conditions (Figs. 4d, S4c,d). This suggests that certain specific mild inflammatory stimulation acting over a long period may inhibit the hypoxia-induced increase in permeability of EC monolayers. Microscopic observation by immunofluorescence staining of proteins provided insights into the mechanisms of changes in migration ability and permeability of ECs associated with hypoxic and inflammatory stimuli. It is known that NF-κB is rapidly expressed in cells in response to LPS stimulation, with its activation peaking 15–30 min after the exposure, and nuclear translocation of NF-κB decreases when cells acquire LPS tolerance49,50. Consistent with many previous studies49–51, the fluorescence intensity of NF-κB that translocated into the nucleus tended to increase at 1 h after LPS stimulation (Fig. 7). The expression of adhesion molecules ICAM-1 and VCAM-1, which have been reported to be regulated by NF-κB activated through LPS stimulation24–26, also increased under normoxia with inflammatory stimulation by LPS (Fig. 5). Interestingly, the trend in changes in ICAM-1 expression after preconditioning coincided with NF-κB activation, suggesting a relationship between the two (Figs. 5c and 7b). On the other hand, when hypoxic stress was simultaneously applied, the increase in VCAM-1 expression was suppressed (Figs. 5d and e). Under hypoxic conditions, HIF-1α was stabilized and translocated into the nucleus due to hypoxic stress at all LPS concentrations, indicating that cells sensed hypoxia regardless of inflammatory state (Fig. 6). VCAM-1 expression is reportedly reduced in a hypoxic environment independently of NF-κB nuclear translocation52, which is consistent with our present results. These findings indicate that the complex response of ECs to hypoxic and inflammatory stimuli alters the expression pattern of adhesion molecules. There are reports that activation of HIF-1α is promoted by inflammatory stimulation by LPS in immune cells such as macrophages53. Increased nuclear translocation of HIF-1α in ECs due to LPS stimulation was observed with 0.1 μg/ml LPS even under normoxia and became less pronounced after preconditioning (Fig. 6), showing dependence on LPS concentration and exposure time. Here, the change in the nuclear translocation rate of HIF-1α under normoxic conditions was almost consistent with the change in the intensity of NF-κB in the nucleus (Figs. 6b and 7b). Previous studies have reported that NF-κB promotes the activation of HIF-1α and the same relationship was confirmed in the present study54,55. Additionally, under conditions of simultaneous exposure to hypoxic and inflammatory stimuli with LPS concentration of 1 μg/ml, where collective cell migration specifically increased, the nuclear translocation ratio of HIF-1α was relatively decreased compared to conditions without LPS stimulation (Figs. 2b and 6b). In previous study44, when nuclear translocation of HIF-1α in ECs was significant (≤ 1% O2), migration was strongly suppressed, and when it was moderate (~ 3% O2), migration became active. The results of this study similarly suggest a correlation between HIF-1α nuclear translocation and cell migration ability. In another previous study under hypoxic conditions of ~ 3% O238, the permeability was enhanced, with loosening of adherens junctions due to internalization of VE-cadherin responsible for cell–cell adhesion. However, no clear changes in the expression of VE-cadherin were observed under extremely low oxygen conditions of 1.3% O2 and inflammatory stimulation by LPS (Fig. S5), and therefore its contribution was considered limited. These results suggest that changes in migration ability and permeability of ECs occur as result of the interrelationship between NF-κB-mediated responses to acute inflammatory stimulation and responses mediated by HIF-1α stabilization to hypoxic stress. Inhibition of NF-κB transcriptional activity by PFI-1 reduced the migration ability of ECs (Fig. 8a,b,c). In particular, the specific increase in migration speed observed during simultaneous exposure to 1.3% O2 hypoxic and 1 μg/ml LPS stimuli disappeared (Fig. 8b). Additionally, under hypoxic conditions, there was little LPS concentration dependence of migration with NF-κB inhibition, and the inhibitory effect of hypoxic stress on migration became more dominant, further supporting the involvement of NF-κB in changes in dynamics due to inflammatory stimulation. Moreover, under all oxygen and LPS conditions, permeability maintained the same value as the control (Fig. 8d). Here, the final dilution ratio of dimethyl sulfoxide (DMSO) used as a solvent for PFI-1 was 4,000-fold, but it is considered that there is little effect if the dilution is more than 1,000-fold44. There are reports that NF-κB inhibition restored the expression of ZO-1 protein responsible for tight junctions that decreased under TNF-α stimulation56, and the decrease in migration ability and permeability observed in this study may also be due to the enhancement of cell–cell integrity such as tight junctions and focal adhesion responsible for adhesion to the substrate, in addition to changes in adherens junctions (Fig. S5). Detailed understanding including these biochemical mechanisms is expected to contribute to elucidating the mechanisms of vascular disruption and immune response abnormalities during sepsis. This study elucidated changes in the dynamics and function of EC monolayers during simultaneous exposure to hypoxic and inflammatory stimuli using microfluidic devices. However, the evaluation of protein expression and localization was limited to microscopic observation and image analysis by immunofluorescence staining, and quantitative evaluation could not be performed. Additionally, since NF-κB and HIF-1α react quickly to stimuli, there was a possibility that their active states might change during oxygen concentration control or cell fixation procedures. Although NF-κB was inhibited using PFI-1 to consider its effects on migration ability and permeability, there remain unclear points regarding mechanisms that directly affect these, such as changes in cell–cell integrity. Further investigation is needed regarding changes in gene expression and signal transduction between proteins. As a physiological defense response, immunothrombosis could occur, where microthrombi are intentionally formed to reduce blood flow to prevent an invasion of microorganisms. In this phenomenon, a hypoxic state is first generated, and then ECs are exposed to inflammatory stimuli such as LPS once the thrombi are removed. Cell experiments with an inflammatory stimulus after hypoxic preconditioning would also give important insights for sepsis. The use of HUVECs as our primary EC model represents a potential limitation of this study. Being derived from large fetal veins, HUVECs may not fully recapitulate the physiological characteristics and molecular responses of ECs found in small capillaries and arterioles that are predominantly affected during sepsis. While HUVECs provide a well-established and reproducible model system, the microvascular endothelial heterogeneity across different organs cannot be captured in this approach. Future studies employing organ-specific microvascular ECs would provide more physiologically relevant insights into the complex vascular dysfunction observed in sepsis and potentially reveal tissue-specific responses to inflammatory and hypoxic challenges. Moreover, incorporating directional fluid shear stress would be valuable to investigate how flow-induced polarity might interact with hypoxic and inflammatory responses in ECs. Conclusion The collective cell migration varied depending on the oxygen and supplemented LPS concentrations, and the inflammatory stimulation attenuated the cellular response to hypoxia. The permeability of EC monolayers increased in response to both hypoxic and inflammatory stimuli, whereas under long inflammation, the increase of permeability was somewhat suppressed at a certain LPS concentration. Furthermore, the expressions of ICAM-1 and VCAM-1 suggested complex effects of hypoxia and inflammation on vascular endothelial monolayer function through nuclear translocations of NF-κB and HIF-1α. These results enhance our understanding of EC behavior in an in vitro model that effectively recapitulates key pathophysiological features of sepsis, thereby providing a robust foundation for further mechanistic studies on vascular dysfunction and adaptive responses under sepsis. Methods Microfluidic device The double-layer microfluidic device (35 mm in diameter, 4 mm in thickness) developed in our previous study45 was used (Fig. 1). It was fabricated by transferring channel patterns to transparent and gas-permeable polydimethylsiloxane (PDMS; SILPOT 184, Dow Corning Toray, Japan) using soft lithography and bonding with a glass coverslip (35 mm in diameter, Matsunami Glass Ind., Japan) after plasma treatment (PCD-001, Harrick Plasma, USA). On the bottom of the device, a gel channel (1,300 μm in width, 150 μm in height) was formed at the center for placing a hydrogel that reproduced an extracellular matrix (ECM), and media channels (500 μm in width, 150 μm in height) were positioned on both sides of the gel channel for culturing cells. In addition, gas channels (1,000 μm in width, 150 μm in height) were located 500 μm from the bottom for supplying a gas mixture at predefined oxygen concentrations. Gas exchange between the gas channels and the gel and media channels via PDMS allowed for control of the oxygen concentration around the cultured cells. A polycarbonate film (32 mm in diameter, 0.5 mm in thickness) was embedded in the device 1 mm from the bottom to reduce oxygen infusion from the surrounding atmosphere around the device. The fabrication process of the device is described in our previous papers45,57. Briefly, the lower PDMS layer with gel and media channels and the upper PDMS layer with gas channels containing the polycarbonate film were separately fabricated. After punching the inlets and outlets for the gas channels, the two layers were bonded together using plasma treatment. Subsequently, the inlets and outlets for the gel and media channels were punched in the PDMS mold. After sterilization by autoclaving, the PDMS mold was bonded to a glass coverslip using plasma treatment. For hydrogel stability and cell adhesion, the gel and media channels were coated with 1 mg/ml poly-D-lysine (PDL) solution (P7886, Sigma-Aldrich, USA). A 2.5 mg/ml type I collagen solution (354236, Corning, USA) adjusted to pH 7.3–7.5 was injected into the gel channel and polymerized for 40 min in the incubator to mimic ECM. Subsequently, the cell culture medium (EGM-2; CC-3162, Lonza, Switzerland) was injected into the media channels, and the device was left in an incubator overnight to stabilize the collagen gel. To further enhance cell adhesion, the media channels were coated with a 2.0 mg/ml Matrigel solution (354230, Corning) on ice before cell seeding38,42,44. Cell culture Pooled human umbilical vein endothelial cells (HUVECs; C2519A, Lonza), which contain cells isolated from multiple donors to minimize donor-to-donor variation, were used for cell experiments. HUVECs between the 5th and 7th passages were cultured on cell culture dishes using EGM-2 in an incubator (37°C, 5% CO2) and harvested before reaching confluence. The cell suspension of HUVECs was adjusted to 5.0 × 106 cells/ml and 25 μl was introduced into each of the left and right media channels in the microfluidic device. Here, to promote cell adhesion to the gel interface, the device was tilted more than 90° and maintained for 15 min after the cells were injected into the right media channel, followed by the same procedure for the left media channel. The device was subsequently returned to its initial position. After 30 min incubation in the incubator, the cell culture medium was replaced. The cells were cultured for several days with daily medium change until they formed a monolayer on the media channels. Inflammatory stimuli for the cells were induced by supplementing the cell culture medium with LPS (L2880, Sigma-Aldrich) at concentrations of 0.1 or 1 μg/ml. For acute inflammatory stimulus, the cell culture medium in the media channels was replaced with LPS-supplemented medium immediately before the experiment. For long inflammatory stimulus, the cells were preconditioned by replacing the cell culture medium in the media channels with LPS-supplemented medium 24 h before the experiment. Here, our previous study revealed that high LPS concentration at 10 μg/ml decreased the cell migration speed immediately after LPS stimulation57. This decrease in migration ability was thought to be attributed to cellular damage caused by inflammatory stimulation from very high concentrations of LPS47. When ECs were cultured for 24 h with LPS added in a 96-well plate, at an LPS concentration of 10 µg/ml, cell viability hardly changed, while the glycocalyx on the cell membrane tended to be damaged57. Considering the cytotoxicity of high concentrations of LPS, LPS concentrations in the present study were set at ≤ 1 μg/ml. To confirm the involvement of NF-κB in responses to hypoxic and inflammatory stimuli, experiments were performed by supplementing PFI-1 (PF-6405761, Selleck, USA) to the cell culture medium at 250 nM. A 1 mM stock solution was prepared in DMSO (049–07213, FUJIFILM Wako Pure Chemical Corporation, Japan) and further diluted to 250 nM in cell culture medium. PFI-1 blocks BRD4 activity, thereby suppressing the transcriptional activity of NF-κB. The cell culture medium was changed to the one supplemented with PFI-1 one day after cell seeding in the device. Measurement of collective cell migration The device with a monolayer of HUVECs formed in the media channels was placed in a stage incubator (STXG-WSKMX, Tokai Hit, Japan) (37 °C, 5% CO2) mounted on a phase-contrast microscope (EVOS M7000, Thermo Fisher Scientific, USA). Subsequently, a gas mixture containing 21% O2 and 5% CO2 balanced with N2 was injected into both gas channels using a gas blender (BR-3C, KOFLOC, Japan) for 1 h to generate a steady normoxic state and to equilibrate the temperature and pH in the device. After that, the oxygen concentration in the supplied gas mixture was switched to 21%, 5% or 1% O2, while maintaining 5% CO2 balanced with N2, to control the uniform oxygen environment around the cells. This resulted in uniform oxygen concentrations of 21%, 5.3% or 1.3% O2, respectively, in the media channels45. Time-series phase-contrast microscopy images were then acquired every 10 min for 6 h. Cell migration was measured using time-lapse microscopy images captured at 20-min intervals with particle image velocimetry (PIV) software JPIV58. First, the effects of device misalignment were eliminated using the StackReg plugin in the image analysis software ImageJ (National Institutes of Health, USA). Next, PIV analysis was performed to measure cell migration velocity in a region of interest (ROI) of 640 × 1,280 pixels (397 × 794 μm). Permeability measurement Permeability of a HUVEC monolayer formed at the interface between gel and media channels was measured by quantifying the diffusion of fluorescently labeled dextran into the gel. The device was placed in a stage incubator and subjected to oxygen concentration control for 6 h, as in the migration measurement experiments. Fluorescently labeled dextrans of 10, 40, and 70 kDa, conjugated with Cascade Blue (D1976, Invitrogen, USA), Alexa Fluor 488 (FD40S, Sigma-Aldrich), and Texas Red (D1864, Invitrogen), respectively, were supplemented to EGM-2 at 10 μg/ml each. A 10 μl of the dextran solution was injected into the right media channel. Fluorescence images were acquired at 5 and 10 min after the right media channel filled with dextrans reached a steady state. The sequential fluorescence images were analyzed using ImageJ, setting an ROI of 2,048 × 200 pixels (1,272 × 124 μm) covering the gel and media channels. The average fluorescence intensity in the media channel, Im, and the average fluorescence intensity at the gel interface, Ig, were quantified from the profile of average fluorescence intensity in the y-direction. The average fluorescence intensity within the gel channel, Īg, after 5 and 10 min was also obtained in a ROI of 2,048 × 700 pixels (1,272 × 435 μm) at the same x-coordinate. The permeability Pm was then calculated using the following equation59. where V and A are the volume of the gel region and the area of the gel interface, respectively, and t is time. Immunofluorescence staining and image analysis The expression and localization of proteins in HUVECs were visualized by immunofluorescence staining following protocols described in our previous works39–42,44. Briefly, after 6-h exposure to controlled oxygen concentrations, cells were fixed with 4% paraformaldehyde phosphate buffer (163–20145, Wako Pure Chemical Industries, Japan) for 10 min. For NF-κB localization studies, fixation was performed after 1 h of exposure instead. Triton X-100 (Pharmacia Biotech, Sweden) and 1% Block Ace (DS Pharma Biomedical, Japan) diluted in phosphate-buffered saline (PBS; P5119, Sigma-Aldrich) were used for permeabilization and blocking, respectively. ICAM-1, VCAM-1, VE-cadherin, HIF-1α, and NF-κB were labeled for 1 h with each antibody diluted in PBS at appropriate concentrations: ICAM-1 antibody (62133, Cell Signaling, USA) at 1:200; VCAM-1 antibody (39036, Cell Signaling) at 1:200; VE-cadherin antibody (sc-9989, Santa Cruz Biotechnology, USA) at 1:200; HIF-1α antibody (ab51608, Abcam, USA) at 1:100; and NF-κB antibody (8242, Cell Signaling) at 1:200. After that, the primary antibodies were labeled with a secondary antibodies for 1 h. Alexa Fluor 488 goat anti-rabbit antibody (A11008, Invitrogen) diluted at 1:100 with PBS was used for VCAM-1, HIF-1α, or NF-κB, and Alexa Fluor 594 goat anti-mouse antibody (A11032, Invitrogen) diluted at 1:100 with PBS was used for ICAM-1 or VE-cadherin. Furthermore, nuclei were stained with 5 μg/ml DAPI (D21490, Thermo Fisher Scientific) for 10 min. The immunofluorescence staining was conducted at room temperature, and the cells were washed twice with PBS between each step. The cells were observed using a confocal laser scanning microscope (LSM800, Carl Zeiss Microscopy, Germany). Twenty cross-sectional images were acquired at 0.60 μm intervals in the vertical direction to capture the entire cell volume on the bottom of the media channels. Maximum intensity projection was then performed to generate composite images on the horizontal plane. The expressions of ICAM-1 and VCAM-1 were evaluated by integrating the fluorescence intensity over the whole image and normalizing to those under the control condition, denoted as I*ICAM-1 and I*VCAM-1, respectively. The cell–cell integrity was investigated by calculating the relative VE-cadherin area as A*cad = Acad/Aout, where Acad and Aout are the VE-cadherin area and total cell area for one cell, respectively38. The image analysis was performed using ImageJ. The nuclear translocation of HIF-1α or NF-κB was quantified by calculating the ratio of mean fluorescence intensity in nuclei, Īnucleus, to that in the whole image, Īwhole, denoted as αHIF-1α or αNF-κB, respectively42. The nuclear translocation of NF-κB was also evaluated by integrating the fluorescence intensity in the nuclei and normalizing with that under the control condition, denoted as In*NF-κB. Statistical analysis For migration measurement, four devices were used for each experimental condition, with two ROIs set in each device (N = 4, 8 samples in total), except for the experiments with PFI-1 supplementation where three devices were used (N = 3, 6 samples in total). For permeability measurement, three devices were used for each experimental condition, with four locations at gel interfaces evaluated in each device (N = 3, 12 interfaces in total). For immunofluorescence staining observation, three devices were used for each condition, and confocal microscopy images captured at four arbitrary locations were quantified for each device (N = 3, 12 images in total). For VE-cadherin analysis, three cells were randomly selected per image and examined to calculate the relative VE-cadherin area (N = 3, 36 cells in total). Significant differences in metrics were assessed by the one-way or two-way analysis of variance (ANOVA) followed by post hoc Tukey’s test, with statistical significance inferred at P < 0.05. Supplementary Information Below is the link to the electronic supplementary material. Supplementary Material 1 Acknowledgments We gratefully acknowledge Ms. Yukie Kobayashi and Mr. Masamitsu Nakajima for their valuable technical assistance in fabricating the microfluidic devices. Author contributions S.S. and K.F. conceived the idea for the study. K.S., A.F., and K.F. performed time-lapse imaging of collective cell migration using microfluidic devices and analyzed the data. K.S. conducted permeability measurement of vascular endothelial monolayers. K.S and A.F performed confocal microscopy imaging with immunofluorescence staining and image analysis. K.S., A.F., S.S., and K.F. interpreted and validated the data. K.S. S.S., and K.F. obtained funding support. K.F. supervised the study. K.S. and K.F. drafted the manuscript, and K.S., A.F., S.S., and K.F. edited the manuscript. Funding This work was partially supported by the JST SPRING (JPMJSP2114) to K.S., the JSPS KAKENHI (23K15623) to S.S., and the JST PRESTO (JPMJPR22O8) to K.F. Data availability The data presented in this study are available from the corresponding author upon reasonable request. Declarations Competing interests The authors declare no competing interests. Footnotes Publisher’s note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations. References Associated Data Supplementary Materials Supplementary Material 1 Data Availability Statement The data presented in this study are available from the corresponding author upon reasonable request.