Human blood samples, without any donor identifiers, were obtained after their informed consent, and the buffy coat was isolated by the South Texas Blood and Tissue Bank (STBT), San Antonio, TX, USA. Buffy coat was procured from STBT. All the methodologies and procedures were approved by Institutional Review Board (IRB), The University of Texas at San Antonio, San Antonio, TX, USA (protocol #12-227), and the experiments were conducted in accordance with these appropriate guidelines. PBMCs were isolated following the protocol detailed in Evani and Ramasubramanian (2016). Primary human umbilical vein endothelial cells (HUVECs) were obtained from Life Line Cell Technologies and cultured according the manufacturer’s instructions, stored at -80°C within three passages (eight doublings). The cells were thawed, used within one passage for the experiments.

Chlamydia pneumoniae TW183 EB was obtained from the University of Washington (Seattle, WA, USA), aliquoted and stored at -80°C. C. pneumoniae stock titer was determined following established protocols using TT401 antibody (Campbell and Kuo, 2009). For experiments to study the effect of chlamydial antigens, the bacteria were rendered ineffective by heat treatment at 95°C for 30 min.

Human monocytes were infected with C. pneumoniae EB at a Multiplicity of Infection (MOI) 1 as described in detail elsewhere (Evani and Ramasubramanian, 2016). A similar procedure was followed for infection with heat-killed bacteria. Monocytes that were treated with buffer (uninfected or mock) were used as controls.

Collagen gels of concentration 2 mg/ml with 3 million/ml monocytes/macrophages were prepared using rat tail collagen type l (Corning, USA) according to manufacturer’s protocol. Briefly, appropriate amounts of high concentration rat tail collagen solution, recommended quantity of neutralizing 1N NaOH and 100 μl of 10x dPBS were mixed on ice. Human monocytes from 4 h post infection were resuspended in complete media at a concentration of 3 × 10⁷ cells/ml (untreated, HK Chlamydia and Cpn infected). Hundred microliters of this cell suspension was added to the neutralized collagen solution. Sterile water was then added to the collagen solution to make up the volume of final solution to 1 ml and supplemented with gentamycin (1 μg/ml). Appropriate volumes of the collagen solution on ice was quickly loaded in culture dishes/well plates/glass slides and allowed to gel for 20 min at 37°C. Complete media was then added to gels to immerse and further incubated for up to 12 days.

Human umbilical vein endothelial cells were grown to confluence for a minimum of 3 days in the top well of a Transwell [Polyethylene terephthalate (PET) membranes, 6.5-mm diameter, 3.0 μm pore size; Corning, Falcon]. The confluency of the monolayer was verified by staining one of the inserts with Phalloidin-Alexa fluor 488 and DAPI and visualized by fluorescence microscope. Confluent HUVEC inserts were placed in companion plates that contain 200 μl collagen gels containing human monocytes (uninfected, heat killed, or infected with C. pneumoniae) supplemented with 400 μl of media (lower chamber). After 24 h, inserts were removed and placed in fresh 24-well companion plate containing 600 μl fresh lifeline vascular basal media (lower chamber) and 100 μl of 1 mg/ml Rhodamine B isothiocyanate–dextran in top chamber (insert). After 1 h incubation at 37°C, 100 μl aliquots were taken from the lower chamber and added to a black 96-well plate. Endothelial permeability was measured by estimating the fluorescence at ex/em of 528/600 nm using a fluorescence plate reader. TNF-α stimulation was used as a positive control and collagen gels without cells as negative control. For LDL permeability assays, dextran was replaced with 100 μl of BODIPY FL-LDL (Life Technologies), and the procedure described above was followed. LDL permeability was measured at ex/em of 485/528 nm using Synergy BioTek plate reader.

Human umbilical vein endothelial cells on PET membrane transwell inserts from the above experiment were fixed with 4% formaldehyde, and incubated with mouse antihuman vascular endothelial (VE)–cadherin antibody (1:100; Abcam), followed by incubation with secondary goat anti-mouse Alexa 488–conjugated antibody and Alexa 633–conjugated phalloidin (1:100) and DAPI for nucleus (1:100; Molecular Probes). After incubation with each antibody, the membranes were washed three times for 10 min with BD Permwash. Finally, membranes were dipped once in water and mounted with Fluorosave (Calbiochem) on microscopic glass slides. The membranes were then visualized under a Zeiss-510 confocal microscope at a magnification of 100×.

Five microliters of collagen gels were spotted in a black 96 well plate, supplemented with 50 μl media and incubated for up to 1 week. The matrices were then treated with 50 μl PBS containing Calcein-AM for 30 min at RT. The wells were gently/carefully washed with media, supplemented with 50 μl of PBS, and fluorescence readings of the spots (area scan) were taken at ex/em of 488/530 nm using BioTek plate reader. Fluorescence readings from fresh gels with various concentrations of monocytes were used to generate standard curve, from which the viability was determined. The gels were immediately fixed and mounted on a cover glass for 3D confocal imaging.

After incubation of up to 1 week, supernatant from collagen matrices were collected, and assayed using ELISA kits for detection of cytokines and chemokines IL-1β, IL-8, MCP-1, and TNF-α following manufacturer’s protocol (BD biosciences).

Five microliters of collagen gels were spotted in a black 96 well plate, supplemented with 50 μl media and incubated for up to 1 week. Intracellular ROS was stained using Image-IT green and mitochondrial SOD was stained with Mito-SOX red assay kits (Life technologies, USA), along with staining for nucleus using DAPI according to manufacturer’s instructions. Post staining, the gels were fixed with 4% formaldehyde and mounted on a cover glass for 3D confocal imaging. 3D confocal images were taken at ex/em set at 380/405, and 495/529 nm and 510/580 nm for gels stained with DAPI, Image-IT green and Mito-SOX red assay kit, respectively. Fluorescence intensity within the cells was measured from images using Image Pro analyzer, which was expressed in Fluorescence Intensity Unit (FIU).

Matrix metalloproteinases secreted by macrophages into the supernatant was measured by a colorimetric assay following the manufacturer’s instructions (Anaspec fluorimetric MMP assay kit). The MMPs analyzed by the kit were MMP-1, 2, 3, 7, 8, 9, 12, 13, and 14. Enzyme activity of the supernatants collected was measured at 37°C. Activity units in the plateau region of the enzyme activity were chosen as a representation of the MMP release/activity.

Five microliters of gels were placed in black 96 well plates, and incubated in media for up to 7 days. The matrices were then fixed with 2% formaldehyde, and stained with 10 μg/ml of anti C-terminal collagen antibody overnight (AdipoGen Life sciences). The gels were washed three times for 10 min each time using PBS-T and then stained with Alexa fluor 488 secondary anti rabbit-antibody (Abcam) for 2 h, along with Alexa fluor 633-phalloidin and DAPI. The gels were washed three times for 10 min each time using PBS-T and mounted on a cover glass with a drop of FluorSave. Z-stack images were recorded at 100× magnification at ex/em set at 380/405, and 488/520 nm and 510/580 nm for nucleus, collagen nicks and cytoskeletal actin, respectively, using confocal microscopy.

To quantify the cell spreading in 3D, gels with phalloidin/calcein AM stain were used to obtain at least five Z-stack images with a stack interval of 1.8 μm with a total Z-height of at least 60 μm. Using Image Pro analyzer, the 3D stack was compressed to a 2D image, and processed for estimating area occupied by single cells and by cluster of cells: for single cells normalized actin intensity per cell was used; and for cell clusters, calcein intensity was used. A group of cells was considered to be a cluster if >5 cells (or 500 μm²) were present as a continuum. To quantify the changes in the microstructure of gels doped with fluorescence collagen (Fl-collagen), at least five Z-stack images with a stack interval of 1.8 μm were obtained for in phase collagen fibers with a total Z-height of at least 60 μm. Using Image Pro analyzer/Image J, the 3D stack was compressed to a 2D image, and processed to obtain the distribution of fiber density, orientation, and alignment. To estimate fiber density, the 2D stacked image was converted to gray scale (0–255), and the intensity distribution was obtained. The lowest gray scale (<28) was expeditiously chosen to represent absence of any prominent fibers. Fiber orientation was estimated for a 2D stacked gray image using Orientation J plugin with a defined set of parameters in Image J software, and was defined as the fiber angle between 0° and 90°. The frequency distribution of the angles was used as a measure of skewness in fiber alignment. To estimate fiber length, Skeleton analysis plug in in Image J software was used. To estimate fiber thickness, the 2D stacked gray image was converted to gray scale, and the grayscale threshold was set to 75 units so that only bundled fibers (i.e., filaments) were identified. Filament properties were estimated by calculating filament thickness and area occupied by bundled filaments using Image J software.

The viscoelastic properties of collagen matrices (n = 5) were measured using Physica MCR 300 rheometer (Anton-Paar, Ashland, VA, USA). The samples were prepared with dimensions to the 15 mm circular geometry of the parallel plate spindle being used. Briefly, human monocytes (control, uninfected, HK-infected, and live-infected) were embedded in 200 μl (three spots/dish) collagen gels and placed in 35 mm petri dishes with 12 mm hydrophobic borders (drawn with a hydrophobic pen and kept under UV light over night before making gels). After gelation at 37°C, the gels were detached from the surface, supplemented with media and the dishes were placed in a 5% CO₂ incubator at 37°C and for up to 7 days. After 3 or 7 days of incubation, the gels were analyzed by rheometry under humidified conditions. To avoid any slippage during the experiment, a rough sand paper cut to the shape of the spindle was attached to the plates. Oscillatory amplitude testing over 0.1 to 100% strain of the gels was performed to determine the linear viscoelastic range of the storage modulus (G’). The elastic and loss moduli were measured from the frequency sweep ranging between 1 and 100 rad/s at 0.5% strain. The yield point was obtained from the amplitude sweep ranging between 0.1 and 1000% strain. In another experiment, the gels from 1 week post infection were placed in IBIDI sticky slide and 1x PBS was infused for 5 min at 50 dyn/cm² of wall shear stress. The samples were then fixed in neutral buffer formalin and processed for scanning electron microscopy analysis by dehydrating the samples with ethanol and coated with gold. The samples were then analyzed for collagen tear/rupture by SEM at a magnification of 5000×.

Statistical analysis was performed using Graph Pad Prism 6 (Groningen, The Netherlands). Each experiment was performed in triplicates and was conducted twice, and a Student’s t-test or a two way ANOVA (followed by Bonferroni correction/Tukey’s post hoc analysis) was used to estimate whether a significant differences between the treatments existed The differences were considered significant if P< 0.05 (t-test/ANOVA).

To understand the effect of matrix environment on infected macrophages, cells infected with PBS (Mock), heat-killed (HK Cpn) and live C. pneumoniae were encapsulated in 3D collagen matrices, and viability and infectivity were analyzed for up to 1 week. The viability of macrophages decreased gradually during the course of 7 days, although even at the end of 1 week more than 60% of the infected cells were viable suggesting that infected macrophages are capable of surviving in the matrix for substantially long periods of time (>40% cells were viable at the end of 12 days: data not shown). Monocytes infected with heat-killed chlamydial EBs show similar viability pattern to those of monocytes infected with live organisms with decrease in viability as early as 3 days (Figure 1A). Using confocal microscopy, we observed that at least 80% of the cells were infected and, consistent with our previous findings, the size of the inclusions ranged from very small to large inclusions (3–30 μm), and the cells contained a variable number of inclusions (Figure 1B) (Cheeniyil et al., 2015). Interestingly, C. pneumoniae can be seen outside cells in the matrices at the end of 1 week possibly due to the release of C. pneumoniae inclusions, which correlates with the decrease in viability of infected cells in 7 days. As expected, cells infected with HK C. pneumoniae did not form any inclusions.

Upon chlamydial infection, leukocytes in suspension cultures (Netea et al., 2000; Bas et al., 2008) secrete pro-inflammatory cytokines and chemokines, MMP, and reactive oxygen species (ROS), all of which contribute to development and progression of atherosclerosis. To examine the effect of C. pneumoniae infection in cells trapped in an extracellular matrix, we quantified the secretion of key pro-inflammatory cytokines and chemokines from infected macrophages encapsulated in 3D collagen gels. We observed that the infected cells secrete small quantities of TNF-α within the first day, which declined soon after (Figure 2A). TNF-α secretion was undetectable in uninfected and HK-infected controls. In contrast, cells infected with live C. pneumoniae secreted large quantities of IL-1β and IL-8 for 7 days, unlike cells infected with HK C. pneumoniae or uninfected cells. We also observed that cells infected with live C. pneumoniae produced significantly higher levels (threefold) of MCP-1 by 7 days compared to cells infected with HK C. pneumoniae or uninfected controls (Figure 2B). Cells infected with either live or HK organisms produced significantly more MMPs compared to uninfected cells. Interestingly, HK C. pneumoniae organisms triggered MMP production compared to uninfected controls (Figure 2C). C. pneumoniae infection triggers the secretion of, amongst others, MMP-1 and -8 which degrade collagen and play an important role in atherosclerosis (Mäntylä et al., 2004; Rupp et al., 2004; Kern et al., 2009). This data reveals both live C. pneumoniae and chlamydial antigen trigger a proinflammatory response in encapsulated macrophages, with live infection eliciting a higher and lasting response.

The breach of endothelial barrier is one of the early hallmarks of atherosclerosis. We hypothesized that the high levels of IL-1β and IL-8 secreted by infected macrophages will trigger endothelial dysfunction. To this end, we incubated an intact and confluent endothelial layer over the collagen matrix containing uninfected/infected cells for 24 h, and analyzed the barrier integrity based on the distribution of junctional protein VE-cadherin and permeability of dextran (Figure 3A). While the uninfected macrophages had a small effect (as indicated by arrows), macrophages infected with either HK or live chlamydial infection compromised the endothelial barrier significantly, with live infection disintegrating the endothelial junction substantially. We quantified the total gap between endothelial cells from the images stained with VE-cadherin. We observed that macrophages infected with live chlamydial infection in the sub-endothelium created threefold more gaps in endothelium than those with HK bacteria or mock infected macrophages. These leaky junctions disrupt the barrier permeability and might increase the transendothelial transport molecules, which otherwise cannot pass the junction via para-cellular route when endothelial junction is intact. We estimated the permeability of fluorescent dextran to correlate permeability with barrier disruption (Figure 3B). We observed that permeability of dextran increased 2.5-fold in the presence of active infection in sub-endothelium when compared to HK and mock infected controls. This disruption might facilitate para-cellular LDL movement into the arterial wall along with permeation of LDL by trans-cellular route by activated endothelium, as seen from a significant increase in LDL transport across the endothelial barrier due to infection (Figure 3C). Chlamydial antigens, though did not produce substantial IL-1 and IL-8, were enough to disrupt the endothelial barrier and allow transendothelial movement of LDL (Figures 3B,C). Of note, macrophages encapsulated in 3D matrices secreted threefold higher IL-8 compared to 2D cultures, which translated to a twofold increase in the transendothelial permeability of dextran (Supplementary Figures S1A,B). Although limited, our data indicates that the 3D cultures may elicit a higher inflammatory response than 2D cultures, thus underscoring the importance of relevant in vitro tissue models of infection. Further, basal activation of the endothelium is more effective in increasing the permeability when compared to apical activation, suggesting a major role of lodged macrophages in 3D on endothelial damage in atherosclerosis (Supplementary Figure S1C).

Low density lipoprotein is oxidized to oxLDL, which is taken up by macrophages to form foam cells, a crucial part of plaque. We found that cells infected with live C. pneumoniae in the matrix produced significantly higher levels of ROS and SOD when compared to cells infected with HK organisms, which in turn was higher than uninfected controls (Figure 4A). We hypothesized that the matrix containing infected cells rich in ROS and SOD will promote foam cell formation. To estimate uptake of LDL by macrophages, we exposed encapsulated macrophages to LDL, and analyzed the uptake 7 days after infection (Figure 4B). Analysis of accumulated LDL in macrophages shows that higher percentage of infected cells loaded with LDL when compared to HK and mock infected macrophages. Further, macrophages infected with live EBs were more effective in taking up LDL as evident from LDL load per cell (∼2.5-fold) in infected macrophages than uninfected macrophages or macrophages infected with chlamydial antigens. Again, infection of macrophages by chlamydial antigens still resulted higher LDL accumulation than uninfected controls, possibly due macrophage activation and hence oxidation of LDL (Figure 4B). These observations support the premise of accelerated atherosclerotic progression due to live infection or bacterial antigens in the intima.

Since, we observed substantially high levels of MCP-1 released by C. pneumoniae-infected macrophages (Figure 2B), we analyzed the migration of infected cells in the matrix by quantifying spreading and cytoskeletal changes in individual macrophages, and the density of macrophages within the collagen matrix. After 7 days, live-infected and HK-infected cells spread more when compared to uninfected cells, which retained their spherical shape and spread less (Figure 5A). In uninfected macrophages, actin predominantly concentrated around the nucleus but in infected macrophages, actin was visibly concentrated in the filopodial segments and was higher, which are characteristics of motile cells (Figure 5B). We also observed that infected macrophages form clusters, which are densely packed cells in certain areas of the 3D collagen matrix, thus supporting the cause of higher actin polymerization, spreading, and movement of infected cells (Figure 5C). This clustering may also be a result of aggregation due to the upregulation/activation of adhesion receptors, including integrin, and ICAM-1 in infected cells (Evani et al., 2011, 2013). Although macrophages infected with chlamydial antigen did not show clustering at the end of 1 week; at the end of 2 weeks, they also exhibited areas with significantly high cell density (data not shown). As shown in Figure 2C, C. pneumoniae infection elicits the production of MMPs, which is a key in the degradation of collagen matrix and subsequent remodeling. To understand the effect of MMP produced during C. pneumoniae infection on the microenvironment, we quantified the extent of degradation from the binding of fluorescently labeled antibody to c-terminal of collagen strands (Figure 5D). We observed higher collagen degradation around the infected cells suggesting a higher concentration of secreted MMPs in the neighborhood immediate to the cells after both 3 and 7 days.

To understand the role of infection on the matrix microstructure, we analyzed the morphological properties of fluorescently stained individual collagen fibers from CLSM images (Supplementary Figure S2). Visual observation suggested that infection resulted in less uniform and more regular alignment of the fibers compared to uninfected or no cell controls (Figure 6A). To quantify the uniformity of fibers, we quantified the intensity distribution in individual images, and a representative distribution is shown in Figure 6B, which shows that the infection results in an increase in areas that lack collagen fibers. The orientation of fibers was calculated from the angles that the fibers make to the X–Y axes in each of the quadrants. These angles vary between 0° and 90°, and reveal alignment of fibers (Figure 6C). Matrix containing uninfected cells showed uniform alignment, which becomes non-uniform following an infection. Matrix with infected cells showed a high percentage of fibers at 15° to 25° and 70° to 80° suggesting that infected cells impose fiber orientation. Fresh collagen production by activated macrophages and macrophage migration might distort and bundle up fibers as previously shown in studies on SMCs. Finally, we estimated the length of the fibers, thickness of individual clusters of fibers, and the area occupied by these clusters (Figure 6D). These estimations show that while the matrix containing uninfected cells are predominantly made of fibers ∼60 μm in diameter, infected cells bundle the collagen fibers laterally to form thicker filaments, and such thicker filaments dominate the distribution. Matrix containing cells infected with HK-C. pneumoniae showed lower but similar thickness throughout suggesting less/no bundling of collagen fibers (Figures 6D,E). Together, our analysis demonstrates a profound effect of infection on collagen architecture of the 3D gels.

Having established that infected macrophages has a profound effect on matrix architecture, we investigated the role of infection on the mechanical properties of the collagen matrix. To this end, we used dynamic mechanical analysis to estimate the viscoelastic properties of collagen gels. Prior to rheological analysis, we confirmed that uninfected/infected macrophages did not induce any morphological changes such as shrinkage or damage to the gels over 7 days (Figure 7A). To measure the elastic (G’) and loss moduli (G”), we subjected the gels to a steady strain rate of 0.5%, which falls in the linear viscoelastic limit of collagen gels. Consistent with previous reports, the elastic modulus or gel stiffness of collagen gels was estimated to be 50 Pa, which did not change due to macrophage encapsulation (Supplementary Figure S3). The concentration of naïve monocytes was adjusted so that the mechanical properties are not dependent on the cell concentration. After 3 days of infection, we did not observe any significant difference in gel stiffness due to infection. However, at 7 days post infection, the stiffness of gels containing macrophages infected with live C. pneumoniae EB increased to ∼80 Pa, which is 1.5-fold higher when compared to gels containing macrophages infected with chlamydial antigen or uninfected controls. After 7 days, the stiffness of the gels containing macrophages infected with chlamydial antigen gels increased compared to 3 days uninfected controls, whereas the stiffness of the gels with uninfected macrophages was unaltered (Figure 7B).

Next, we measured the yield strength of the gels as the point of irreversible deformation when the gels were subjected to increasing strain, i.e., during amplitude sweep. From the response, we obtained the yield point as the strain rate at which G’ starts to decrease, and crossover point as the strain rate at which a loss in gel integrity was observed (Figure 7C). Surprisingly, we observed that collagen gels containing live infection showed an early loss of mechanical strength as evident from yield point onset at ∼7% strain (Figure 7D) and cross-over point (Figure 7E) at lower strain percentage (∼26%) when compared to other conditions. To further confirm this observation, we exposed gels from 7 days post infection to high levels of wall shear stress (∼50 dyn/cm²), as may be encountered at plaque site in vivo, by infusing buffer over gels formed in a microfluidic channel. The samples were analyzed for collagen tear/rupture by SEM. Qualitative observation of the images revealed tear in collagen gels thus exposing live C. pneumoniae infected macrophages due to shear-induced rupture, and particularly in the immediate vicinity of the cells. This was not as pronounced in controls (Figure 7F) highlighting the impact of micro-environmental changes due to live chlamydial infection.