Decision letter: Dynamic persistence of UPEC intracellular bacterial communities in a human bladder-chip model of urinary tract infection

Abstract

Article Figures and data Abstract eLife digest Introduction Results Discussion Materials and methods Data availability References Decision letter Author response Article and author information Metrics Abstract Uropathogenic Escherichia coli (UPEC) proliferate within superficial bladder umbrella cells to form intracellular bacterial communities (IBCs) during early stages of urinary tract infections. However, the dynamic responses of IBCs to host stresses and antibiotic therapy are difficult to assess in situ. We develop a human bladder-chip model wherein umbrella cells and bladder microvascular endothelial cells are co-cultured under flow in urine and nutritive media respectively, and bladder filling and voiding mimicked mechanically by application and release of linear strain. Using time-lapse microscopy, we show that rapid recruitment of neutrophils from the vascular channel to sites of infection leads to swarm and neutrophil extracellular trap formation but does not prevent IBC formation. Subsequently, we tracked bacterial growth dynamics in individual IBCs through two cycles of antibiotic administration interspersed with recovery periods which revealed that the elimination of bacteria within IBCs by the antibiotic was delayed, and in some instances, did not occur at all. During the recovery period, rapid proliferation in a significant fraction of IBCs reseeded new foci of infection through bacterial shedding and host cell exfoliation. These insights reinforce a dynamic role for IBCs as harbors of bacterial persistence, with significant consequences for non-compliance with antibiotic regimens. eLife digest Urinary tract infections are one of the most common reasons people need antibiotics. These bacterial infections are typically caused by uropathogenic Escherichia coli (also known as UPEC), which either float freely in the urine and wash away when the bladder empties, or form communities inside cells that the bladder struggles to clear. It is possible that the bacteria living within cells are also more protected from the immune system and antibiotics. But this is hard to study in animal models. To overcome this, Sharma et al. built a ‘bladder-chip’ which mimics the interface between the blood vessels and the tissue layers of the human bladder. Similar chip devices have also been made for other organs. However, until now, no such model had been developed for the bladder. On the chip created by Sharma et al. is a layer of bladder cells which sit at the bottom of a channel filled with diluted human urine. These cells were infected with UPEC, and then imaged over time to see how the bacteria moved, interacted with the bladder cells, and aggregated together. Immune cells from human blood were then added to a vascular channel underneath the bladder tissue, which is coated with endothelial cells that normally line blood vessels. The immune cells rapidly crossed the endothelial barrier and entered the bladder tissue, and swarmed around sites of infection. In some instances, they released the contents of their cells to form net-like traps to catch the bacteria. But these traps failed to remove the bacteria living inside bladder cells. Antibiotics were then added to the urine flowing over the bladder cells as well as the vascular channel, similar to how drugs would be delivered in live human tissue. Sharma et al. discovered that the antibiotics killed bacteria residing in bladder cells slower than bacteria floating freely in the urine. Furthermore, they found that bacteria living in tightly packed communities within bladder cells were more likely to survive treatment and go on to re-infect other parts of the tissue. Antibiotic resistance is a pressing global challenge, and recurrent urinary tract infections are a significant contributor. The bladder-chip presented here could further our understanding of how these bacterial infections develop in vivo and how good antibiotics are at removing them. This could help researchers identify the best dosing and treatment strategies, as well as provide a platform for rapidly testing new antibiotic drugs and other therapies. Introduction Urinary tract infections (UTIs), the second most-common cause for the prescription of antibiotics (Foxman, 2010) are characterized by the high frequency of recurrence, defined as a reappearance of infection within 12 months despite the apparently successful completion of antibiotic therapy. Recurrence occurs in about 25% of all UTIs (Foxman et al., 2000) and strongly impacts the cost of healthcare and reduces the quality of life, particularly since more than 60% of women are diagnosed with a UTI at least once in their lifetime (Klein and Hultgren, 2020). Uropathogenic Escherichia coli (UPEC), the causative agent for the majority of UTIs, exhibits a complex lifestyle in the bladder, with planktonic sub-populations within the urine co-existing with intracellular bacteria. UPEC invasion of the urinary bladder generates substantial changes to bladder morphology and a robust immune response. Much of our current understanding of early stages of UTI and the intracellular lifestyle is derived from studies in the mouse model (Anderson, 2003; Duraiswamy et al., 2018; Hannan et al., 2012; Hung et al., 2009; Justice et al., 2004; Schwartz et al., 2011; Yang et al., 2019). Examinations of mouse bladder explants via microscopy have revealed the formation of intracellular bacterial communities (IBCs) composed of thousands of bacteria within individual superficial bladder cells (Justice et al., 2004). IBCs also play a prominent role in clinical infection and have also been harvested from the urine of cystitis patients (Robino et al., 2013; Rosen et al., 2007). These structures are considered to have a linear progression from an early stage of colonization by single bacteria to an intermediate stage of biofilm-like communities culminating in the release of bacteria at the late stage (Anderson, 2003; Justice et al., 2004). However, long-term imaging of infected animals or of explant-tissue is technically challenging and therefore it has been difficult to capture the dynamic changes that underlie the formation of these structures and their impact on infection and clearance by subsequent antibiotic treatment. In addition, the bladder is an extremely complex organ; it has a stratified architecture with well-differentiated cell types; a lumen filled with urine whose composition and chemical properties can vary depending on the physiological state of the individual and is subjected to periodic and large changes in organ volume and surface area (Korkmaz and Rogg, 2007). These features may play important roles in infection, but they have been hitherto hard to capture outside of a whole animal model, where they are present in their entirety and cannot be dissected in a modular manner. Organotypic models are well-suited to address these outstanding questions. A key strength of advanced organotypic models such as organs-on-chip (Benam et al., 2016; Huh et al., 2010; Jang et al., 2013; Jang et al., 2019; Kim et al., 2016; Novak et al., 2020; Zhou et al., 2016) or organoids (Cakir et al., 2019; Clevers, 2016; Rossi et al., 2018; Sato et al., 2009) is that they increasingly recapitulate the complexity of host physiology. Organ-on-chip models have been developed for several organs either in isolation or in combination (Novak et al., 2020; Ronaldson-Bouchard and Vunjak-Novakovic, 2018). These systems are increasingly being used to model infectious diseases including viral and bacterial infections of the respiratory tract (Nawroth et al., 2020; Si et al., 2021; Thacker et al., 2020; Thacker et al., 2021), gut (Jalili-Firoozinezhad et al., 2019; Kim et al., 2016; Shah et al., 2016; Tovaglieri et al., 2019), kidney (Wang et al., 2019), and liver (Kang et al., 2017). Recently, bladder organoids that mimic the stratified architecture of the urothelium have been used to study infections (Horsley et al., 2018; Smith et al., 2006; Sharma et al., 2021). However, the organoid model suffers from certain shortcomings inherent to the 3-D architecture such as the lack of vasculature, the inability to manipulate the cells mechanically, and the constrained volume of the lumen. In contrast, organ-on-chip models offer a complementary approach that does not suffer from these limitations. However, to our knowledge, there have been no reports of a bladder-on-chip model. Although there are studies that have developed in vitro bladder models that recreate the stratified architecture of the bladder epithelium (Horsley et al., 2018; Smith et al., 2006; Suzuki et al., 2019), they have not been used to recapitulate the multiple stages of IBC formation. Similarly, while a few studies have visualized different stages of IBC development by culturing human bladder cells under urine flow, these models are restricted to monoculture experiments (Andersen et al., 2012; Iosifidis and Duggin, 2020). Furthermore, the lack of vasculature in these models restricts the extent to which immune cell components or drugs can be introduced in a physiologically relevant manner and none of the systems reported to date offer the possibility of mimicking the mechanics of filling and voiding in a functioning bladder (Andersen et al., 2012; Horsley et al., 2018; Iosifidis and Duggin, 2020; Smith et al., 2006). Here, we report the development and characterization of a bladder-chip model that mimics the bladder architecture by co-culture of a well-characterized human bladder epithelial cell line with bladder microvascular endothelial cells in a device geometry that allows the two cell types to be exposed to urine and nutritive cell culture media, respectively. The flow rates in the apical and vascular channels can be controlled independently, and multiple rounds of micturition are recreated via the application of a linear strain to the PDMS membrane that serves as the substrate for co-culture. Using this model, we show that diapedesis of neutrophils to sites of infection on the epithelial side can lead to the formation of neutrophil swarms and neutrophil extracellular traps (NETs), and that IBCs offer substantial protection to bacteria from antibiotic clearance. Our observations suggest that the role of IBCs in reseeding bladder infections upon cessation of antibiotic treatment or failure to complete a course of antibiotics might be more important than previously assumed and therefore strategies aimed toward eradication of IBCs are very crucial for treatment efficacy. Results Reconstitution of bladder uroepithelium and bladder vasculature We established a bladder-chip infection model by co-culturing HMVEC-Bd primary human bladder microvascular endothelial cells with the 5637 human bladder epithelial cell line (HTB-9, epithelial cells) in a novel bladder-chip approach (Figure 1A). In a small subset of experiments, primary human bladder epithelial cells (primary epithelial cells) were used in place of 5637 cells. Immunostaining verified that the epithelial and endothelial cells formed confluent monolayers (cell densities in Table 1) with high expression of junction markers such as epithelial cell adhesion molecule (EpCAM) in the epithelial cell layer, and platelet endothelial cell adhesion molecule-1 (PECAM-1) and VE-cadherin in the endothelial cell layer (Figure 1B,C, Figure 1—figure supplement 1A–H). Expression of these and other junction markers such as E-cadherin and zonula occludens-1 (ZO-1) was consistent across monocultures (Figure 1—figure supplement 2A–G). Figure 1 with 3 supplements see all Download asset Open asset Human Bladder-chip model of UTI recapitulates the physiology of bladder filling and voiding. (A) Schematic of the human bladder-chip with co-culture of the 5637 human bladder epithelial cell line (epithelium, top) and primary human bladder microvascular endothelial cells (endothelial, bottom) on either side of the stretchable and porous membrane. Pooled human urine diluted in PBS and endothelial cell medium were perfused in the apical and vascular channels respectively to mimic bladder physiology. A negative pressure in the ‘vacuum’ channels (magenta) on either side of the main channel was applied to stretch the porous membrane to mimic stretching of the bladder. (B, C) Immunofluorescence staining of confluent epithelial and endothelial cell monolayers (anti-EpCAM (magenta) and anti-CK7 (yellow) for the epithelial cells and anti-PECAM-1 (green) for the endothelial cells) in an uninfected control chip. Some endothelial cells also stained positive for CK7. Cell nuclei were labeled with DAPI (azure). (D) Schematic of the reconstitution of the bladder filling and voiding cycle via stretching of the membrane with a duty cycle of 6 hr. The cycle consisted of a linear increase in strain through stretching of the membrane (filling bladder, 0–2 hr), maintenance of the membrane under stretch (filled bladder, 2–4 hr), a quick relaxation of applied strain over 2 min (voiding bladder, 4:02 hr) and maintenance without applied strain (voided bladder, 4:02 hr to 6 hr). (E) An overview of the timeline of the experimental protocol including infection, addition of neutrophils via the vascular channel, and two cycles of antibiotic treatment interspersed by two bacterial growth cycles. The consecutive bladder duty cycles are indicated. Table 1 Characterization of epithelial and endothelial cell densities from a total of n=18 fields of view in both the epithelial and endothelial layers from n=two bladder-chips. Cell typeCell density/104 μm2Epithelial31.7 ± 5.1Endothelial4.0 ± 0.6 Similarly, a majority of the epithelial cells expressed cytokeratin 7 (CK7), a general uroepithelial marker and cytokeratin 8 (CK8), a differentiated uroepithelial marker both in monoculture (Figure 1—figure supplement 2I,J) and on-chip (Figure 1B, Figure 1—figure supplement 1B, Figure 1—figure supplement 1F). Some endothelial cells were also CK7+ (Figure 1C, Figure 1—figure supplement 2H). Co-culture of bladder epithelial and bladder endothelial cells in the human bladder-chip therefore did not alter the cellular expression patterns of these markers as compared to monocultures. We further characterized the 5637 cells in monoculture for markers for uroepithelial differentiation. A total of 5637 cells showed high expression of uroplakin-3a (UP3a) which has been shown to be essential for UPEC infection (Figure 1—figure supplement 2K; Martinez et al., 2000; Mulvey et al., 1998) but only a small proportion of epithelial cells were positive for the basal cell marker cytokeratin 1 (CK1), in agreement with the umbrella cell nature of the 5637 cells (Figure 1—figure supplement 2L,M; Duncan et al., 2004; Smith et al., 2006). Overall, these observations confirmed that the bladder-chip is populated with cells that mimic the physiology of the bladder vasculature and the superficial uroepithelial cell layer, although the relatively small size of the epithelial cells is likely an artefact inherent in the 5637 cells. Further, the chip design enables the flow of media with different compositions through the epithelial and vascular channels. We used this feature to mimic bladder physiology by perfusion of diluted pooled human urine on the epithelial side and endothelial cell medium on the vascular side. Modeling bladder filling and bladder voiding in the bladder-chip The human bladder experiences vast changes in volume and surface area on a periodic basis. The first phase of this cycle, (‘filling bladder state’), occurs through a slow and gradual increase in bladder volume due to addition of urine from kidney via the ureters, resulting in a large increase in bladder volume (Figure 1D). At the level of the uroepithelium, these changes manifest as a two-dimensional biaxial stretch of the bladder tissue. Ex vivo studies in rat bladders report that a maximum strain of 90% and 130% can be applied to bladder explants in the circumferential and longitudinal directions (Gloeckner et al., 2002; Parekh et al., 2010), although micturition in vivo is likely triggered before this maximum value is reached. Subsequently, volume growth slows, and the bladder attains a relatively constant volume, with the bladder epithelium in a corresponding state of maximal stretch, (‘filled bladder state’). Voiding of the bladder through urination rapidly reduces bladder volume, (‘voiding bladder state’) and removes the strain experienced by the bladder cells. The bladder epithelium, as a viscoelastic material, responds to the removal of strain and relaxes over the subsequent period (De Pascalis et al., 2018) where the bladder volume remains low (‘voided bladder state’) (Figure 1D). The architecture of the bladder-chip device allows for a linear strain to be applied to the porous membrane via negative pressure in the channels adjacent to the main channel of the device (‘the vacuum channels’) (Grassart et al., 2019; Huh et al., 2010; Figure 1A). We characterized the linear strain experienced by epithelial cells due to application of negative pressure (Figure 1—figure supplement 3). A plot of applied pressure vs. strain was linear, and we achieved a dynamic range of linear strain (0% to 19%) by application of negative pressure (0 to −900 mbar). High levels of applied strain were incompatible with long-term time-lapse microscopy as it generated a significant drift in the axial position of the membrane. We therefore limited the linear strain applied to a maximum of 10%, which is of the same order of magnitude, albeit a small fraction of the typical strain experienced by the bladder tissue in vivo. We then modeled the bladder filling and voiding over a 6 hr duty cycle with different states: filling bladder (0 to 2 hr, 0% to 10% strain), filled bladder (2 to 4 hr, 10% strain), voiding bladder over a period of 2 min (4 hr to 4:02 hr, 10% to 0% strain), and voided bladder (4:02 hr to 6 hr, 0% strain) (Figure 1D). This 6 hr duty cycle was repeated for the remaining duration of each experiment. Overall, the bladder-chip model enables co-culture of two cell-types in nutritionally different microenvironments with an applied strain that partly mimics the physiology of bladder filling and voiding cycles. UPEC infection of the epithelial layer under flow in the bladder-chip model UPEC attachment and invasion of bladder epithelial cells has been shown to be sensitive to fluidic shear stress (Andersen et al., 2012; Zalewska-Piątek et al., 2020). Although UPEC infection in the bladder can occur in the presence or absence of shear stress, we infected the epithelial layer on the bladder-chip under a flow rate of 1.2 ml/hr (corresponding to a shear stress of 0.02 dyne cm−2) for a period of 1.5–2 hr. Low shear stress conditions have been recently reported to enhance bacterial adhesion to epithelial cells (Zalewska-Piątek et al., 2020) and likely reflects the in vivo environment except during urination when the shear stress increases dramatically to 3–5 dyne cm−2 (Sokurenko et al., 2008). This choice was also motivated by the small volumes of the microfluidic chip, as rapid bacterial growth in the diluted human urine could lead to acidification of the medium. Under these conditions, an overwhelming majority of bacteria did not attach to the epithelial cells. Time-lapse microscopy showed that bacterial attachment to the epithelial cells increased steadily over this period, but that the typical infectious dose at the end of this period was low (<one bacterium per epithelial cell, Figure 2—figure supplement 1). Accordingly, we established a 28 hr live-cell imaging experimental protocol as shown in the schematic in Figure 1E and described in greater detail in the subsequent sections and in the Materials and methods. Briefly, this consisted of an initial period of acclimatization, followed by UPEC infection, introduction of neutrophils, and two consecutive cycles of antibiotic treatment and recovery to monitor IBC dynamics at the single-cell level. This experimental protocol therefore mimicked key aspects of the host-pathogen interactions in early stages of UTIs as well as the response to antibiotic treatment. Diapedesis of neutrophils across the epithelial-endothelial barrier in response to UPEC infection in the bladder-chip The bladder-chip platform enables continuous imaging while maintaining flow in both epithelial and vascular channels and mechanically stretching the membrane as part of the bladder voiding cycle. At the start of each live-cell imaging experiment, bladder epithelial cells in an initial relaxed state were perfused with sterile pooled human urine diluted in PBS on the epithelial side and endothelial cell medium on the vascular side respectively (Figure 2A1, B1, Figure 2—figure supplement 2A1). The epithelial side was inoculated with a low dose of UPEC in diluted urine and infection was performed under flow (Figure 2A2, B2, Figure 2—figure supplement 2A2) for ca. 1.5–2 hr. The first bladder duty cycle was also initiated at this timepoint. We maintained the optical focus of the microscope on the epithelial layer over the subsequent course of infection. UPEC attachment was evident during this infection phase by visual inspection (Figure 2B2, Figure 2—figure supplement 2A2). Figure 2 with 10 supplements see all Download asset Open asset Neutrophil diapedesis and NET formation on-chip. (A1–A5) Schematic of the UPEC infection, introduction of neutrophils and diapedesis of neutrophils across the epithelial-endothelial barrier to sites of infection. Flow in the epithelial and endothelial channels is indicated by arrows; the flow rate was increased upon introduction of neutrophils in the vascular channel to increase attachment. Increased flow rate in the epithelial and endothelial channel is indicated by black arrows. (B1–B5) Snapshots from time-lapse imaging highlighting each stage in the infection cycle shown in (A1–A5). Bladder epithelial cells (magenta) and neutrophils (amber) were identified with membrane (Cell Mask Orange) and cytoplasmic (Cell Tracker Deep Red) dyes, respectively. UPEC identified via constitutive expression of YFP are shown in green. Neutrophil swarms could either control bacterial growth (yellow dashed square, compare B4 vs. B5) or did not manage to restrict bacterial growth (white dashed square, compare B4 vs. B5). (C) Bar charts for relative frequency of neutrophil diapedesis (black) in n=three uninfected control bladder-chips and n=four infected bladder-chips. Data obtained from n=95 and n=116 fields of view, each of which was 206 x 206 μm2. (D) Quantification of the number of neutrophils detected on the epithelial layer, in control and infected bladder-chips. The red bar represents the median value, p<1E-15. In many instances in the uninfected control bladder-chips, no neutrophil diapedesis is detected. Data obtained from n=95 and n=130 fields of view, each of which was 206 x 206 μm2. (E) A plot of the maximum neutrophil swarm volume on the epithelial layer normalized to the total volume for n=67 and n=118 fields of view on n=three uninfected control and n=three infected bladder-chips. The red bar represents the median value. p<1E-15. (F) A plot of neutrophil swarm volume on the epithelial layer over time for n=51 and n=40 fields of view for n=two technical replicates indicated by squares and circles. For each time profile, the volume is normalized to the maximum volume attained over the timeseries and t=0 refers to the timepoint at which neutrophils are introduced into the vascular channel (G) Plot of the time to reach the maximum swarm volume in n=154 fields of view across n=four infected bladder-chips. (H1–H5, I1–I5) NET formation by neutrophils on the epithelial layer. The neutrophils are identified by a cytoplasmic dye (CellTracker Deep Red, false colored in amber) (H2, I2) and immunostaining with an anti-myeloperoxidase antibody (H3, zooms in H5) or an anti-neutrophil elastase antibody (I3, zooms in I5). Merged images in each case are shown in H4 and I4. UPEC identified via YFP expression is shown in spring green. Nuclear labelling with DAPI is shown in azure. (H5) NETs, identified via anti-myeloperoxidase staining or anti-elastase staining are indicated with white arrows in (H5) and (I5), respectively. (J1–J4) Scanning electron micrographs of the epithelial layer of an infected bladder-chip 2 hr after the introduction of neutrophils in the endothelial channel. (J1) Neutrophils (amber arrowheads) are visible above a layer of epithelial cells. Thick bundles consisting of many thinner NET structures between neutrophils are indicated by white arrowheads, and examples of individual UPEC bacteria within the NETs are indicated by green arrowheads. A heavily infected epithelial cell with UPEC visible on the epithelial cell (purple arrowhead), and appendages between epithelial cells (cyan arrowheads) are also visible. (J2) Micrographs showing thick bundles consisting of many thinner NET structures (white arrowheads) that extend between cells and trap many individual bacteria. Thinner NET fibers (yellow arrowheads) are also visible. (J3) Zooms of the regions in J2 identified by a yellow dashed square. Two bacteria held by a thick bundle composed of many thinner NET fibers are shown. (J4) Micrograph highlighting multiple bacteria trapped between NET bundles. p-Values in D and E were calculated using a Mann-Whitney test. Scale bar = 50 μm in B1–B5, H1-H5, I1-I5. After this period of infection, unattached bacteria in the epithelial channel were removed by perfusion of diluted urine. As our aim was to study the interaction of UPEC with host cells, this perfusion was maintained throughout the experiment. Continuous perfusion reduced the accumulation of bacteria in the urine and enabled imaging of intracellular bacteria without large amounts of background noise in the fluorescence channels from planktonic bacteria. We subsequently performed a series of interventions that mimic the course of UTI infections and used time-lapse imaging to monitor the simultaneous changes in host-pathogen interaction dynamics. Immune cells were introduced into the vascular channel to mimic the host response to infection. Among innate immune cells, neutrophils are the first line of defense in UTIs (Haraoka et al., 1999). Neutrophils were isolated from human blood with a high degree of purity (Son et al., 2017) verified by immunostaining for the neutrophil-specific marker CD15 (Zahler et al., 1997; Figure 2—figure supplement 3). Neutrophils were pre-labeled with the cytoplasmic CellTracker Deep Red dye to enable identification and tracking during time-lapse imaging and subsequently introduced in the bladder-chip devices through the vascular channel at a cell concentration of ca. 2 million cells/ml similar to the neutrophil concentration in human blood (Hsieh et al., 2007). This mimics the natural route of immune cell migration into the bladder (schematic in Figure 2A3–A5, snapshots in Figure 2B3–B5, Figure 2—figure supplement 2A3-A5) During this period, the flow rate through the vascular channel was increased to 3 ml/hr (shear stress η = 1.0 dyne/cm2), which aids neutrophil attachment to endothelial cells (Alon et al., 1995). In uninfected control chips, neutrophil attachment to the endothelial layer was minimal (Figure 2—figure supplement 4A,B) and diapedesis of neutrophils to the epithelial layer was rare (Figure 2C, Figure 2—figure supplement 5A). In stark contrast, infection of the epithelial layer with UPEC elicited a robust attachment of neutrophils to endothelial cells (Figure 2—figure supplement 4C,D), along with rapid diapedesis across to the epithelial side (Figure 2B2–B4, Figure 2—figure supplement 2A2-A4, Figure 2—figure supplement 5C). We confirmed that neutrophil diapedesis was also stimulated in the uninfected chips upon exposure to a gradient across the epithelial-endothelial barrier of pro-inflammatory cytokines known to be upregulated during UTIs - IL-1α, IL-1β, IL-6, and IL-8, each at 100 ng/ml (Agace et al., 1993; Hedges et al., 1992; Nagamatsu et al., 2015; Song et al., 2007; Figure 2—figure supplement 5B). However, infection elicited a more robust neutrophil diapedesis response compared to the cytokine gradient, evident in the higher number of neutrophils observed across multiple fields of view on the epithelial side (Figure 2—figure supplement 5B–D). These results suggest that infection on-chip generates a strongly pro-inflammatory environment locally and that neutrophil diapedesis is further stimulated by the presence of UPEC (Agace et al., 1995; de Oliveira et al., 2016). Using time-lapse imaging, we quantified the kinetics of neutrophil migration with a temporal resolution of up to 7.5 min. Neutrophil diapedesis to the epithelial sites of infection was detected as early as ca. 15–17 min post-introduction into the vascular channel (Figure 2B3, Figure 2—figure supplement 2A3, Figure 2—video 1, Figure 2—video 2). Neutrophils that migrated to the epithelial side aggregated on the epithelial cells (Figure 2B4–B5, Figure 2—figure supplement 2A4-A5, Figure 2—video 1, Figure 2—video 2) In some cases, these neutrophils were able to control bacterial growth (Figure 2B4–B5, solid yellow boxes), whereas in others, bacterial growth was uncontrolled (Figure 2B4–B5, dashed white boxes). Diapedesis was observed in every field of view examined across infected chips (n=4), whereas diapedesis occurred infrequently in uninfected controls (Figure 2C). Furthermore, the size of these aggregates was significantly larger in infected bladder-chips vs. uninfected controls, with the large aggregates in infected-chips exhibiting the characteristic features of neutrophil swarms (Kienle and Lämmermann, 2016; Kreisel et al., 2010; Lämmermann et al., 2013). We quantified the size of the swarm-like aggregates b