The microfluidic devices were fabricated as previously described (51). In brief, SU-8–100 (Y131273, Kayaku Advanced Materials, Westborough, MA) was patterned onto silicon wafers (857, University Wafer Inc., Boston, MA) using soft lithography to create a top and bottom master for the devices. Polydimethylsiloxane (PDMS, 24236–10, Electron Microscopy Sciences, Hatfield, PA) was then polymerized over the masters for 4 hours at 60 °C. Then, the PDMS layers were aligned, and a PDMS rod (0.337 mm inner diameter) was inserted between the two layers before the devices were bonded to a glass-bottom dish (P50G-1.5–30-F, MatTek Corporation, Ashland, MA) with oxygen plasma from a PE25-JW Plasma Etcher (Plasma Etch, Carson City, NV).

Pooled human umbilical vein endothelial cells (HUVEC, 50–305-964, Promocell GmbH C12203, Heidelerg, Germany) were cultured in Endothelial Growth Medium 2 (EGM-2, NC9525043, Lonza Walkersville CC3162, Basel, Switzerland). Media was changed every 2 days. HUVECs were seeded on cell culture-treated flasks (658175, Greiner Bio-one, Monroe, NC), passed at 80% confluency, and used through passage 7.

LB agar (22700025, Thermo Fisher Scientific, Waltham, MA) was used according to the manufacturer’s instructions to create LB plates. Pseudomonas aeruginosa strain K (PAK) was streaked onto LB plates and incubated at 37 °C for 16 hours. The plates were then placed into a refrigerator at 4 °C until they were used. A single colony was grown overnight in 5 mL of LB broth in a bacterial shaker at 37 °C. The next day the 1 mL of the cultured solution was diluted in 4 mL of fresh LB broth and grown in a bacteria shaker at 37 °C for 1.5 hours. 1 mL of bacterial culture was pelleted by centrifugation (17,000xg for 1 minute) and resuspended in 100 μL of EGM-2. The optical density (OD) was measured, and the bacterial solution was diluted in EGM-2 to an OD of 5 (measured at 600 nm, 1.25 x 10⁶ CFU/mL). The EGM-2 contains gentamicin, which prevents replication of the bacteria.

The microfluidic devices were UV sterilized for 15 minutes in a biosafety cabinet. Subsequently, the central chambers were incubated with a 2% polyethylenimine solution in deionized water (03880, Sigma Aldrich, St. Louis, MO) for 10 minutes at room temperature, followed by incubation with a 0.4% glutaraldehyde solution in deionized water (G6257, Sigma Aldrich, St. Louis, MO) for 30 minutes at room temperature. High concentration type-I collagen (354249, Corning, Corning, NY) was neutralized to pH 7.2 using a combination of 0.5 N NaOH (02153495.5, MP Biomedicals Inc., Pittsburgh, PA), 1X phosphate buffered saline (PBS, 10010049 Gibco, Waltham, MA), and 10X PBS (BP2944100, Fisher BioReagants, Pittsburgh, PA) with a final concentration of 2, 4, or 6 mg/mL. The collagen was then loaded into the device chambers around the PDMS rod. After allowing the collagen to polymerize overnight at 37 °C, tweezers were used to remove the PDMS rods, leaving a cylindrical void in the collagen network. HUVECs were seeded in these lumens at a concentration of 20,000 cells/μL. The devices were flipped every 5 minutes for 30 minutes before being placed on a rotator overnight at 37 °C with 5% CO₂. The devices were cultured for two days, and media was changed twice daily. For experiments without an endothelium, the HUVEC seeding and rotating steps were skipped.

All characterization was performed using a Discovery HR20 rheometer (TA Instruments, New Castle, DE) fitted with a 20 mm diameter cone geometry, 1° cone angle, and a 23 μm gap. Collagen solutions were prepared as described above. Oscillatory time sweeps (10 rad/s, 1% strain) were performed with an initial temperature of 10 °C, which was increased to 37 °C within the first 70 seconds of each test. Stress relaxation experiments (15% strain) were conducted at 37 °C.

Collagen pores were visualized via confocal reflectance as described previously (52). In brief, images were obtained with a Nikon A1R HD25 Laser Scanning Confocal Microscope along with a 405 nm laser and a Nikon 20x/0.95 (NA) water immersion objective operated by Nikon Elements software.

Images were loaded into Matlab (MathWorks, 2023b) and preprocessed to remove background noise using a Wiener filter (0.5 x 0.5 μm neighborhood) and a TopHat filter (1.0 μm strel disk diameter). Then, the images were converted to binaries, such that fibers had pixel values of 0 and pores had pixel values of 1. Binaries were eroded with progressively larger disk sizes until a threshold of 50% erosion was crossed. Clusters of pixels with the same values were grouped as objects and the regionprops function in Matlab was used to get the areas and minor axis lengths of each pore.

After 2 days of culture in the microfluidic devices, HUVEC lumens were incubated with prewarmed 4% paraformaldehyde (PFA, AAJ19943K2, Thermo Scientific, Waltham, MA) in phosphate buffered saline (PBS, B2944–100, Thermo Scientific, Waltham, MA) for 30 minutes at 37 °C. The PFA was removed with 3 washes of PBS. Next, PBST (PBS with 0.1% Tween 20) was added to the lumens and left to incubate at room temperature for 10 minutes. Lumens were then incubated with 0.005 mg/mL Hoescht (23491–45-4, Sigma Aldrich, St. Louis, MO), phalloidin (1:200, ab176757, abcam, Cambridge, United Kingdom), and anti-VE-cadherin (1:120, 130–123-688, Miltenyi Biotec, Bergisch Gladbach, Germany) in PBST overnight at 4 °C. The next morning the lumens were washed 3 times with PBS to remove any unconjugated staining molecules. Single timepoint Z-stacks were taken with 2 μm steps along the Z-axis with a Nikon A1R HD25 Laser Scanning Confocal Microscope along a Nikon 20x/0.95 (NA) water immersion objective operated by Nikon Elements software.

For visualizing ICAM-1 expression the above protocol was followed except 3 μL of PAK in EGM-2 was added to the top port of each device and allowed to incubate for 2 hours before fixing with 4% PFA. The devices were stained with 0.005 mg/mL Hoescht (23491–45-4, Sigma Aldrich, St. Louis, MO), phalloidin (1:200, ab176757, abcam, Cambridge, United Kingdom), and anti-ICAM-1 (1:200, BBA20, R&D systems, Minneapolis, Minnesota).

HUVECs in the microfluidic devices were stained and imaged as described above with Hoechst, anti-VE-cadherin, and phalloidin. A maximum intensity projection along the Z-axis was created for the bottom half of each lumen. The number of nuclei in a 450 x 250 μm region of interest placed at the center of each lumen were counted.

HUVECs in the microfluidic devices were stained and imaged as described above with Hoechst, anti-ICAM-1, and phalloidin. A maximum intensity projection along the Z-axis was created for the bottom half of each lumen. The mean pixel intensity of a 625 x 250 μm region of interest centered on the lumen was calculated and subtracted by the mean pixel intensity of a 400 x 100 μm segment of the background of each lumen.

After 2 days of culture in the microfluidic devices, 3 μL of either EGM-2 or PAK in EGM-2 was added to the top port of each device. The devices were incubated at 37 °C with 5% CO₂ for 2 hours. Then, the media in the lumens was replaced with about 4 μL of EGM-2 containing 12.5 μM of 10 kDa fluorescein isothiocyanate (FITC) labeled dextran (FD10S-100MG, Sigma, St. Louis, MO) and imaged immediately. Images were taken in the center Z-plane of each device every 5 seconds for 15 minutes. The videos were transferred to FIJI (ImageJ) for analysis. A vertical line three times the width of the lumen was drawn perpendicular to the lumen and placed at its center. The intensity profile was measured for every pixel on the line at t=0, 5, 10, and 15 minutes. Intensity was normalized to the highest pixel intensity on the t=0 profile for each lumen. Bins of 10 pixels were averaged to smooth the curves.

All blood samples were obtained according to the institutional review board-approved protocols per the Declaration of Helsinki. Peripheral blood neutrophils were isolated from healthy donors, using a MACSxpress Neutrophil Isolation Kit (130–104-434, Miltenyi Biotec, Bergisch Gladbach, Germany) and a MACSxpress Erythrocyte Depletion Kit (130–098-196, Miltenyi Biotec, Bergisch Gladbach, Germany), according to the manufacturer’s instructions. Informed consent was obtained from donors at the time of the blood draw according to our institutional review board.

Isolated neutrophils were resuspended and stained in a 10 nM solution of calcein AM (C3100MP, Thermo Scientific, Waltham, MA) before being resuspended at 7.5 million cells/mL in EGM-2. About 5 μL were loaded into each lumen; then 3 μL of PAK was added to the top port of each device. The devices were imaged immediately following bacterial loading every 10 minutes for 8 hours over a Z-stack of 350 μm with 10 μm steps.

Neutrophil and bacterial loading were carried out as described for extravasation experiments; however, only 20% of the neutrophils were stained to make it easier to distinguish individual cells during data analysis. Images were taken every 30 seconds for 20 minutes at 2-hour intervals over a Z-stack of 80 μm with 10 μm steps.

Intracellular ROS production by extravasated neutrophils was measured in the microfluidic devices as reported previously (46). In brief, dihydrorhodamine-123 (DHR-123, D23806, Thermo Fisher Scientific, Waltham, MA) was used to label intracellular ROS. The lumens were preincubated with 10 mM DHR-123 in EGM-2 for 30 minutes at 37 °C. Isolated neutrophils were stained with DiI (V22885, Thermo Fisher Scientific, Waltham, MA) diluted 1:200 in serum-free RPMI (11875119, Thermo Fisher Scientific, Waltham, MA) for 20 minutes at 37 °C. Neutrophils were resuspended at 7.5 million cells/mL in EGM-2 with 10 mM DHR-123. About 5 μL were loaded into each lumen; then 3 μL of PAK was added to the top port of each device. The devices were imaged immediately following bacterial loading every 10 minutes for 8 hours over a Z-stack of 350 μm with 10 μm steps.

All imaging was performed using a Nikon A1R HD25 Laser Scanning Confocal Microscope built on the Nikon TI2-E Inverted Microscope System, a fully automated stage, and Nikon Elements acquisition software. Stained lumens and confocal reflectance images were captured with a Nikon 20x/0.95 (NA) water immersion objective. Extravasation and ROS experiments were imaged with a Nikon 10x/0.45 (NA) objective. Migration experiments were imaged with a Nikon 20x/0.75 (NA) objective. Extravasation, ROS, and migration experiments were performed at 37 °C and 5% CO₂ with a cage incubator (H201-T-Unit-BL, Oko Labs, Sewickley, PA).

Max intensity projections in the Z-axis were created for each extravasation and migration video in Nikon Elements software. Those videos were transferred to FIJI (ImageJ) for data analysis. To eliminate edge effects, neutrophils were analyzed in the center of the device. To analyze extravasated neutrophils, a 690 μm x 345 μm rectangular region of interest defined by the top edge of the lumen in the center of the device was created. The number of neutrophils within this region were counted at each time point. The number of extravasated neutrophils at each time point was normalized to the initial number of neutrophils in the lumen to account for any variability in cell loading. To determine the initial number of neutrophils, a 345 μm x 122.5 μm rectangular region was drawn within the center of the lumen at the first timepoint. The neutrophils within this luminal region were counted.

For migration experiments, extravasated neutrophils that remained in frame through the duration of the 20-minute time lapse and were clearly distinguishable from other cells were tracked using the MtrackJ plugin in FIJI.

A multiplexed ELISA was employed to measure secretion levels of inflammatory cytokines in lumen-conditioned media from six different conditions (each collagen concentration with either 3 μL of 5 OD PAK in EGM-2 or EGM-2 alone added to the top port). A MagPix Luminex Xmap system (Thermo Fisher Scientific, Waltham, MA) with a ProcartaPlex Human Inflammation Panel 20Plex (EPX200–12185-901, Invitrogen, Waltham, MA) was used following the manufacturer’s protocol. Lumens were incubated for 8 hours. Media was collected from 6 lumens per replicate, pooled, and centrifuged to remove cellular debris. Data are from 4 replicates. Samples were frozen at -20 °C from the time of collection until the assay was run. Secreted protein levels for each replicate with PAK were normalized from each replicate of the same collagen concentration with EGM-2.

For all extravasation, migration, reactive oxygen species, and multiplexed ELISA experiments, data were pooled from three or more independent replicates. Statistical analysis was performed using the estimated marginal means (emmeans) package in R. The statistical analysis accounts for variability within and between replicates and considers independent replicates as individual entities. All collagen concentrations were compared to one another at each time point and analyzed with analysis of variance (ANOVA). For each condition, estimated marginal means and standard errors of the mean (SEM) were calculated and pairwise comparisons were made with Tukey’s adjustment. P-values are labeled as *p<0.05, **p<0.01, ***p<0.001, and ****p<0.0001, unless otherwise noted.

For stained image, rheology, and confocal reflectance microscopy experiments, data were pooled from three or more independent replicates. A one-way ANOVA followed by Tukey’s multiple comparisons test were performed between conditions with an alpha value of 0.05 for the stained image and rheology results. Welch’s ANOVA followed by Games-Howell pairwise comparisons were performed on the confocal reflectance data with an alpha value of 0.05. The mean and SEM were calculated for each condition and P-values are labeled as *p<0.05, **p<0.01, ***p<0.001, and ****p<0.0001, unless otherwise noted.

Neutrophil extravasation and migration are regulated by the structural and mechanical properties of the surrounding ECM (19, 23, 27, 47, 53, 54); therefore, it was important to characterize collagen gels over the range of concentrations used in this study. We created gels with 2, 4, and 6 mg/mL collagen, as this range represents physiologically relevant concentrations throughout various tissues in the body (55). To determine bulk mechanical properties of the collagen gels, we performed oscillatory rheology and found the moduli of the three concentrations were significantly different from each other ( Figure 1A , Supplementary Figure S1 ). After a 30-minute polymerization period, the Storage Moduli for the 2, 4, and 6 mg/mL collagen gels were 6.4 ± 0.2 Pa, 27.4 ± 0.6 Pa, and 56.7 ± 6.2 Pa, respectively ( Supplementary Tables S1 , S2 ). Following polymerization, stress relaxation tests were conducted ( Figure 1B ). All three collagen gels relaxed to 50% of the original strained modulus 75 seconds after being strained ( Supplementary Tables S1 , S2 ).

We next sought to characterize the microscale organization of the different collagen concentration gels. We used confocal reflectance microscopy to visualize the pore structures of the collagen gels ( Figure 1C ). Pore area and minor axis length (M.A.L.) were determined using Matlab’s erosion function, as has been described previously (52) ( Figures 1D, E ). The pore area of the 2 mg/mL collagen (8.89 ± 2.14 μm²) was significantly larger than that of the 4 mg/mL collagen (4.68 ± 0.79 μm²), which was significantly larger than the pore area of the 6 mg/mL collagen (0.92 ± 0.05 μm²). The same trends were seen for pore M.A.L. ( Supplementary Table S2). Together, these data show that increasing collagen concentration increases the modulus of the gels, while also decreasing pore size. However, changing the collagen concentration over the selected range did not affect the viscoelastic nature of the gels.

We previously developed a microphysiological system with relevant components of a tissue microenvironment, including a model blood vessel surrounded by extracellular matrix (ECM) proteins to quantify the human neutrophil response to infection (45). Here, we used that model to investigate the effect of collagen concentration on the neutrophil response. To simulate an infection, neutrophils were loaded into the lumen of the vessel, and bacteria were added to the surrounding matrix via a port at the top of the device ( Figure 2A ). For this study, we used type I collagen gels as our model ECM, as this is the most abundant protein in tissue matrices (17).

The barrier function of endothelial cells, which has been linked to neutrophil activation, has been shown to increase with increasing stiffness of the underlying substrate on which they are seeded (40, 56). Therefore, we sought to determine if collagen concentration affected the formation or barrier function of endothelial lumens. To determine if collagen concentration affects cell density in the lumens, confluent monolayers were formed with human umbilical vein endothelial cells (HUVECs) in each collagen concentration, and endothelial nuclei (Hoechst), actin (phalloidin), and tight junctions (anti-VE-cadherin) were stained ( Figure 2B ). Counting the number of nuclei in a fixed region of interest revealed no significant differences in the number of cells per area across the three conditions ( Figure 2C , Supplementary Table S2 ). Collectively, these results suggest that there was no significant difference in endothelial cell loading or morphology in the different collagen concentration substrates.

During homeostasis endothelial cells form tight junctions to prevent vascular leakage; however, upon infection the cells permit the passage of inflammatory markers between the ECM and the lumen to aid in neutrophil activation and extravasation (32). Therefore, we tested the barrier function of the lumens in the three collagen concentrations using a FITC-dextran to visualize vascular permeability. P. aeruginosa was added to the top ports of devices to simulate an infection. FITC-dextran in medium was added to the lumens 2 hours after adding either medium alone or P. aeruginosa in medium to the top bacterial ports of each device, and diffusion into the surrounding ECM was visualized for 15 minutes. Representative graphs show the percentage of the maximum fluorescence intensity, relative to t=0, at t=0, 5, 10, and 15 minutes across a region three lumen diameters in length and centered around each lumen ( Figure 2D , Supplementary Figure S2A, B ). In each collagen concentration, the endothelial cells retained the FITC in the absence of P. aeruginosa, indicating significant barrier function. However, in the presence of P. aeruginosa the endothelial lumens became leaky, permitting dextran to diffuse into the ECM.

Upon inflammatory challenge luminal endothelial cells upregulate the expression of surface proteins, such as ICAM-1, to facilitate neutrophil extravasation (32, 33, 57). To determine if endothelial cells produce ICAM-1 in our system, we added P. aeruginosa to the top bacterial port of lumen devices and allowed them to incubate for 2 hours before fixing and staining the endothelia for nuclei (Hoechst), actin (phalloidin), and ICAM-1 (anti-ICAM-1) ( Supplementary Figure S2C ). No significant differences in ICAM-1 were observed by endothelial lumens in each of the 2, 4, and 6 mg/mL collagen conditions ( Supplementary Figure S2D , Supplementary Table S2 ).

After characterizing the collagen networks and confirming no changes in lumen structure or barrier function in the various collagen concentrations, we sought to study the effect of collagen concentration on neutrophil extravasation, as this is the first step in the neutrophil response to infection. To quantify extravasation, primary human neutrophils were isolated from whole blood samples, stained with a fluorescent marker (calcein AM), and loaded into the lumens. Extravasation was then induced by loading P. aeruginosa into the top bacterial port of each device. Confocal microscopy was used to image neutrophil extravasation for 8 hours ( Figure 3A ). Extravasated neutrophils were quantified as the number of extravasated neutrophils in a region of interest (ROI) outside of the lumen divided by the number of neutrophils in a second ROI inside the lumen at t=0 to normalize for any discrepancies in neutrophil loading ( Figure 3B ). Significantly more neutrophils extravasated into the 4 mg/mL collagen than in the 6 mg/mL collagen 5–8 hours after P. aeruginosa stimulation. Neutrophil extravasation in 2 mg/mL collagen was greater than in 6 mg/mL collagen but less than in 4 mg/mL collagen; however, the differences did not reach statistical significance ( Figure 3C ). This extravasation is dependent on the presence of an inflammatory stimulus as no extravasation occurs in 2, 4, or 6 mg/mL collagen in response to EGM-2 alone ( Supplementary Figure S3 ).

Given the mechanosensitive nature of the endothelial cells and their role in the activation of neutrophils, we wanted to determine if the differences in neutrophil extravasation were due to the collagen concentration affecting the neutrophils directly or indirectly through interactions with the endothelial cells. Therefore, we repeated the extravasation experiments in the absence of an endothelium ( Supplementary Figure S4 ). We found no significant differences in the number of extravasated neutrophils across the 2, 4, and 6 mg/mL collagen conditions. Together these data show that there is greater neutrophil extravasation into the 4 mg/mL collagen compared to the 6 mg/mL collagen and imply that this difference is due to an endothelial cell-dependent mechanism.

Following extravasation, neutrophils must navigate through the ECM to respond to infection. To determine the effect of collagen concentration on this next step of the neutrophil response, we quantified the migration properties of extravasated neutrophils in the different collagen matrices. Extravasated neutrophils were imaged every 30 seconds for 20-minutes 2, 4, 6, and 8 hours after P. aeruginosa stimulation ( Figure 4A ). We calculated the net displacement as the distance between the starting and ending positions for each 20-minute interval ( Figure 4B ). Migration path length was measured for each neutrophil and was used with displacement to determine the straightness of migration: the path length of a neutrophil divided by its net displacement ( Supplementary Figure S5 ). Lastly, speed was computed as the total path length travelled by a neutrophil divided by the time elapsed during the track ( Figure 4C ).

Generally, migration speed, displacement, and path length significantly decreased with increasing collagen concentration ( Figures 4B, C , Supplementary Figure S5B ). This trend was observed for straightness at the 2-hour time point; however, it did not persist at later timepoints ( Supplementary Figure S5A ).

To determine if these differences were due to direct neutrophil-collagen interactions or indirect interactions involving the endothelium, migration experiments were repeated in the absence of an endothelium ( Supplementary Figure S6 ). Neutrophil migration was not reported beyond the 2-hour timepoint for conditions without an endothelium as most neutrophils do not survive past 2 hours without the pro-survival cytokine granulocyte macrophage colony stimulating factor (GM-CSF) (45). Without an endothelium, neutrophils in the 4 mg/mL collagen migrated faster and farther than neutrophils in the 2 mg/mL collagen ( Supplementary Figures S6B, D ). However, straightness was not significantly different ( Supplementary Figure S6E ). Together these results suggest that the observed differential migration is, at least in part, regulated by an endothelial cell-dependent mechanism.

During an infection, endothelial cells secrete inflammatory markers that recruit and activate neutrophils circulating in the blood (58). To determine if the differential neutrophil extravasation and migration were due to changes in endothelial secretion profiles of inflammatory markers, a multiplexed ELISA was performed. Endothelial lumens were formed in 2, 4, and 6 mg/mL collagen. Devices were treated with media or P. aeruginosa and incubated for 8 hours. Conditioned media was then collected from inside the lumens, and a ProcartaPlex Human Inflammation Panel was used to determine the net mean fluorescence intensity (MFI) of 20 select pro- and anti-inflammatory signaling molecules ( Figure 5 , Supplementary Figure S7 ). Secretion of key inflammatory markers (MIP-1α, TNFα) increased in the 2 and 4 mg/mL conditions when stimulated with P. aeruginosa compared to controls ( Figures 5A, B ). Furthermore, IL-6 and GM-CSF, which have previously been reported to increase neutrophil extravasation and lifetime, respectively (45), were increased in the 2 mg/mL compared to the 6 mg/mL collagen in the presence of P. aeruginosa ( Figures 5C, D ). These results suggest that altered endothelial secretion of inflammatory markers, such as IL-6 and GM-CSF may contribute to the observed differences in neutrophil migration. However, none of the secretion profiles match the trends seen in extravasation, suggesting that the differences in extravasation are not due to secreted factors by the endothelium ( Figures 3 , 5 , Supplementary Figure S7 ).

Once neutrophils reach the site of infection, they employ a variety of antimicrobial functions to eliminate pathogens, and one of the primary functions is the production of reactive oxygen species (ROS). It has been shown previously that neutrophil binding to collagen, via β1 integrins on the neutrophil surface, facilitates ROS production (29–31). Thus, to determine if ROS production was altered by the surrounding collagen concentration, a ROS indicator (DHR-123) was used to measure the percentage of ROS-producing neutrophils following extravasation ( Figure 6A ). The percentage of ROS-producing neutrophils was calculated by dividing the number of rhodamine positive neutrophils in a ROI outside the lumen by the total number of neutrophils in that ROI ( Figure 6B ). No significant differences in percentage of ROS-producing neutrophils were seen across the three collagen concentrations over the 8-hour observation period ( Figure 6C ).