Modeling Group B Streptococcus and Blood-Brain Barrier Interaction by Using Induced Pluripotent Stem Cell-Derived Brain Endothelial Cells

Here for the first time, human iPSC-derived BMECs were used to model bacterial interaction with the BBB. Unlike models previously used to study these interactions, iPSC-derived BMECs possess robust BBB properties, such as the expression of complex tight junctions that are key components for the investigation of bacterial effects on the BBB. Here, we demonstrated that GBS interacts with the iPSC-derived BMECs and specifically disrupts these tight junctions. Thus, using this BBB model may allow researchers to uncover novel mechanisms of BBB disruption during meningitis that are inaccessible to immortalized or primary cell models that lack substantial tight junctions.

pathogen (1)(2)(3). To cause meningitis, bacteria must gain access to the bloodstream and replicate to a high level, causing bacteremia (1). Following bacteremia, bacteria must then interact with and penetrate the blood-brain barrier (BBB) to gain access to the central nervous system (CNS). The specialized brain microvascular endothelial cells (BMECs) that comprise the BBB respond to these bacterial interactions with a cellular immune response and contribute to the disease progression (1)(2)(3).
Streptococcus agalactiae (group B Streptococcus [GBS]) is a Gram-positive, non-sporeforming bacterium that is the leading cause of neonatal meningitis and is an emerging pathogen in specific adult populations (1,4,5). Although advancements have been made in diagnosis and therapy, death still occurs in up to 10% of cases, with 25 to 50% of surviving infants exhibiting permanent neurological sequelae (1,4,5). More recently, much work has been conducted to determine bacterial virulence factors that contribute to interaction with the BBB and allow bacterial access to the CNS. Bacterial surfaceexpressed factors, such as lipoteichoic acid (LTA) (6), pilus components (PilA, PilB, and PilC) (7,8), serine-rich repeat proteins (Srr) (9)(10)(11), streptococcal fibronectin binding factor (SfbA) (12), fibrinogen-binding protein (FbsA) (13), and hypervirulent GBS adhesin (HvgA) (14), have all been demonstrated to promote direct association with the BBB. Additionally, regulatory two-component signal transduction systems, such as CovR/S (15), and CiaR/H (16), have been implicated in the ability to regulate virulence factors that contribute to the pathogenesis of GBS meningitis. Studies have also been conducted to determine the BBB response to GBS. During infection, immortalized BMECs have been shown to downregulate tight junctions through the induction of host transcription factor Snail1, a known repressor of tight junction components. This response resulted in a significant loss of barrier function during infection, which was dependent on Snail1 expression (17). Previous studies have also demonstrated that GBS infection of immortalized BMECs upregulates proinflammatory chemokines and cytokines that act to orchestrate the recruitment and activation of neutrophils and enhance their survival (1,7,18). The recruitment of neutrophils has been linked to further BBB destruction during infection, and bacterial determinants that include CovR/S, PilA, and ␤-hemolysin/cytolysin contribute to this process (7,15,18). Together, these factors promote GBS penetration of the BBB, allowing access to the CNS and the development of meningitis.
The BBB is comprised of highly specialized BMECs that serve to separate the brain from the circulation and, along with cells of the neurovascular unit (NVU), maintain CNS homeostasis (19)(20)(21). BMECs express a spectrum of nutrient transporters and multidrug efflux transporters while also displaying intercellular tight junctions and low endocytosis rates (19)(20)(21). To study the BBB, researchers have relied on complementary in vivo and in vitro techniques. Animal models have been utilized to examine bacterial interactions with the BBB in the context of the full CNS microenvironment (6,(22)(23)(24)(25)(26)(27)(28); however, these are inherently nonhuman models and subject to interspecies differences (29). Primary cell culture of animal and human BMECs has been employed (30)(31)(32)(33); however, after removing BMECs from the brain microenvironment, they routinely lose BBB characteristics (29,34,35). Immortalized human BMECs offer a facile human-based model (29,(34)(35)(36), yet many of these cell lines lack critical BBB properties, such as high transendothelial electrical resistance (TEER) and complex tight junctions (29,(34)(35)(36). Recently, induced pluripotent stem cells (iPSCs) have offered the prospect of renewable BBB models with superior barrier properties (29,34,(37)(38)(39)(40). The model has been used to examine drug delivery, genetic human disease, and ischemic stroke, but it has not yet been evaluated for its applicability to infectious disease (41)(42)(43)(44). Here, we demonstrate that an iPSC-derived BMEC model can be used to examine hostpathogen interactions using the meningeal pathogen GBS. These results motivate the application of this model in infectious disease research.

RESULTS
Wild-type GBS interacts with iPSC-derived BMECs. Previous studies have shown that GBS has the ability to interact with and invade immortalized human brain endothelial cells (hBMECs) (1,3,4,14,31). Thus, we sought to determine if iPSC-derived BMECs could be used to model these interactions. The iPSC-derived BMECs were differentiated as previously described and were shown to express the expected markers and respond to astrocyte cues ( Fig. S1A to G) (29,40). In addition, the iPSC-derived BMECs express ␤1 integrin, a receptor for GBS virulence factors (Fig. S1H) (7,12). Wild-type serotype III, hypervirulent, multilocus sequence type 17 (MLST-17) GBS strain COH1 was used to examine GBS interactions with iPSC-derived BMECs. Confluent iPSC-derived BMEC monolayers were infected with GBS to assess bacterial attachment and intracellular invasion. The numbers of total cell-associated and intracellular bacteria recovered increased proportionally with the multiplicity of infection (MOI) of the original inoculum ( Fig. 1A and C). However, as the MOI increased, the percentage of adherent or intracellular GBS recovered relative to the original inoculum decreased in a stepwise fashion ( Fig. 1B and D), indicating that bacterial attachment and uptake mechanisms are saturable. These results were specific to GBS, as the cell-associated and intracellular populations of the nonpathogenic bacterium Lactococcus lactis were significantly less than those observed for GBS (Fig. S2). To determine if GBS is able to survive intracellular uptake into iPSC-derived BMECs, we performed a modified invasion assay in which the numbers of intracellular GBS bacteria were quantified at different time points after the addition of antibiotics to eliminate extracellular bacteria. As shown by the results in Fig. 1E, GBS persisted in iPSC-derived BMECs for up to 6 h and then exhibited a decrease in intracellular survival over time. Overall, these data demonstrate that GBS can specifically interact with iPSC-derived BMECs and are consistent with results obtained in immortalized hBMECs and other cell types (16,31,45,46). GBS virulence factors contribute to interaction with iPSC-derived BMECs. Recently, much work has been conducted to identify and characterize virulence factors that promote GBS interaction with the BBB (6-8, 11-14, 47). To determine whether several well characterized GBS virulence factors affected interactions with iPSC-derived BMECs, a cohort of mutants were examined for their ability to promote attachment and invasion of the iPSC-derived BMECs. We selected GBS factors that have been shown to be critical for BBB interaction and the pathogenesis of GBS meningitis. Specifically, we chose surface-expressed factors, including PilA (7,8), SfbA (47), and Srr2 (9, 10), all of which contribute to GBS interaction with extracellular matrix (ECM) components acting to bridge the bacteria to host ECM receptors (1). Additionally, we analyzed the contribution of an invasion-associated gene (iagA) whose product acts to anchor lipoteichoic acid (LTA) to the cell surface and promote bacterial uptake into the brain endothelium (1,(6)(7)(8)(9)(10)(11)(12). We observed that infection of iPSC-derived BMECs with these previously generated and described GBS mutant strains resulted in significant decreases in the numbers of adherent and/or intracellular bacteria recovered compared to the results for the wild-type parental GBS strains recovered (Fig. 2). Thus, these data demonstrate that these mutant GBS strains exhibited the expected attenuated attachment and invasion phenotypes using iPSC-derived BMECs.
iPSC-derived BMECs are activated in response to GBS infection. Previous reports have shown that the major proinflammatory BBB response to GBS infection is the upregulation of chemokines and cytokines that promote neutrophilic influx (7, 18).

FIG 2 Contribution of GBS virulence factors to interaction with iPSC-derived BMECs. Adherence (A) and
invasion (B) of wild-type GBS strains compared to those of GBS mutants lacking the virulence factors pilA, iagA, sfbA, and srr2. Data are presented as mean values from three independent iPSC-derived BMEC differentiations conducted in triplicate. Error bars represent SEM. Student's t test was used to determine the significance of the difference between WT NCTC10/84 and the ⌬pilA mutant. ANOVA was used to determine the significance of the difference between WT COH1 and the ⌬iagA, ⌬sfbA, and ⌬srr2 mutants. **, P Ͻ 0.01; ***, P Ͻ 0.001.
Thus, the transcript expression of key chemokines in iPSC-derived BMECs was evaluated following GBS infection. Wild-type GBS at an MOI of 10 was used to infect iPSC-derived BMECs for 5 h, and no loss in cell viability was observed under these conditions (Fig. S3). We observed that CXCL8 (encoding interleukin-8 [IL-8]), CXCL1, and CXCL2, as well as CCL20, encoding neutrophil chemoattractants, were significantly upregulated in response to GBS infection compared to their levels in the noninfected control ( Fig. 3A to D). In contrast, the expression of the gene encoding the global proinflammatory cytokine IL-6 was unchanged in response to GBS infection (Fig. 3E). Together, these data show that key proinflammatory chemokines are induced in iPSC-derived BMECs during GBS infection.

GBS infection disrupts tight junctions of iPSC-derived BMECs.
It is known that GBS infection can affect tight junction components and BBB barrier function (17). To determine whether iPSC-derived BMECs were affected by GBS interactions, we used TEER to measure barrier integrity. After 3 h post-GBS infection, we observed a dramatic decrease in the barrier function of iPSC-derived BMECs (Fig. 4A). Tight junction dysfunction during GBS infection was recently reported to be linked to the expression of the tight junction transcriptional repressor Snail1 (SNAI1) (17). Consistent with this finding, upregulation of SNAI1 transcript expression was observed in iPSC-derived BMECs following GBS infection (Fig. 4B). Transcripts for genes encoding tight junction proteins occludin (OCLN), claudin-5 (CLDN5), and ZO-1 (TJP1) were all also decreased during GBS exposure (Fig. 4C to E). The transcript downregulation corresponded to a loss of tight junction complexes, as immunostaining revealed that occludin, claudin-5, and ZO-1 proteins were noticeably discontinuous or absent from cell-cell junctions following infection (Fig. 5A to C). Quantitation using the area fraction index indicated that this decrease was significant compared to the results for the uninfected control ( Fig. 5D to F). In contrast, an immortalized hBMEC cell line exhibited a lack of tight junction continuity and did not express claudin-5 ( Fig. S4A to E), which is consistent with previous studies using this cell line to examine GBS infection (17). Taken together, these results demonstrate that GBS infection results in the disruption of tight junctions in iPSC-derived BMECs.

DISCUSSION
Here, we demonstrate for the first time the use of iPSC-derived BMECs to model the critical steps of bacterial-brain endothelial interactions that lead to BBB penetration and the development of bacterial meningitis. Until now, in vitro studies of bacterium-BBB interactions have largely relied on immortalized human BMECs (hBMECs) (6-8, 12, 15, 17, 18, 24, 30, 31, 45, 47-52). While the immortalized hBMECs offered the ability to begin understanding the molecular interactions leading to BBB penetration, the model itself lacks important BBB properties, such as the proper expression and localization of tight junction proteins (Fig. S4) (17). More recently, other immortalized human BBB models, such as the human cerebral microvascular endothelial cell (hCMEC) line hCMEC/D3, have been developed. In contrast to the immortalized hBMECs described here, hCMEC/D3s do express junctional claudin-5; however, occludin expression is discontinuous, leading to modest TEERs (~40 ⍀ ϫ cm 2 ) that are not substantially improved upon coculture with astrocytes (35,36,(53)(54)(55)(56). Similar to immortalized human BMEC lines, the iPSC-derived BMEC model offers a reliable and scalable method of generating BMECs that express BBB markers. iPSC-derived BMECs offer the additional advantages of continuous tight junctions, barrier formation, and elevated TEER in response to astrocyte cues (Fig. S1) (29,38,40,57). Furthermore, the addition of retinoic acid to the iPSC-derived BMEC differentiation can greatly elevate TEER values, to physiological levels (38,40). Our results suggest that the iPSC-derived BMEC model can be utilized to study bacterial attachment, invasion, immune activation, and tight junction disruption using GBS as a model meningeal pathogen. In addition, our data compare well with published data regarding attachment and invasion percentages using the immortalized BMEC models. Previous work has demonstrated a range of GBS attachment of between 15 and 28%, with intracellular CFU in the range of 2.5 to 4% (11,12,47); in the present study, we observed comparable percentages. This new model may also be helpful for future studies examining other bacterial meningeal pathogens, such as Streptococcus pneumoniae (pneumococcus), Neisseria meningitidis (meningococcus), and Escherichia coli strain K1.
The physical interaction between GBS and brain endothelial cells has been characterized previously using immortalized human BMECs (1). GBS possesses the ability to adhere to and invade brain endothelial cells through the use of a variety of virulence factors, such as the PilA, SfbA, Srr2, and IagA proteins (6)(7)(8)(9)(10)(11)(12). Our results demonstrate that these factors also contribute to bacterial interaction with the iPSC-derived BMECs. Many of the GBS virulence factors we examined are proteins that promote bacterial interaction with ECM components, including collagen (for PilA), fibronectin (for SfbA), and fibrinogen (for Srr2). These interactions are thought to allow GBS to bridge to host cell receptors directly, promoting bacterial attachment and invasion of multiple host cell types, including brain endothelial cells (7,9,12). Here, we show that these adhesins also promote GBS interaction with iPSC-derived BMECs, suggesting that the iPSCderived BMECs maintain similar cellular receptors necessary for GBS attachment. As it has been previously thought that attachment would precede invasion, it is then unsurprising that adhesion deficiencies, such as in the case of PilA, may also result in a decrease in invasion. Studies have also described a role for IagA, a glycosyltransferase that acts to generate the glycolipid anchor for LTA, in the pathogenesis of GBS using immortalized BMECs in vitro and a mouse model of GBS infection. Our results support these previous findings, as the ΔiagA mutant was less invasive with the iPSC-derived BMECs, although we observed that the ΔiagA mutant also had reduced adherent properties. It is unknown at this point whether the deletion of iagA only impacts anchored LTA or whether other surface factors are disrupted in the ΔiagA mutant that may impact GBS-BBB interactions. The work characterizing GBS SfbA in immortalized hBMECs has demonstrated its contribution to bacterial entry (12), while we observed a role for both attachment and invasion of iPSC-derived BMECs. Interestingly, SfbA is a homolog of the pneumococcal adhesin PavA, previously shown to contribute to pneumococcal adherence to BMECs (58,59), and therefore, it is possible that SfbA could be contributing to GBS adherence, which is better detected in the iPSC-derived model. Additionally SfbA has been reported to contribute to GBS adherence to fibronectin (12), and the binding of extracellular matrix as a means to adhere to host BBB has been discussed previously (7,13,49). Regardless, our results demonstrate that known GBS mutants are attenuated in their interaction with iPSC-derived BMECs, suggesting that in general, similar mechanisms for bacterial-host cell interactions are preserved in the iPSC-derived-BMEC model.
Bacterial activation of the brain endothelium has been hypothesized to contribute to disease progression through the upregulation of cytokines and chemokines that attract circulating leukocytes, specifically neutrophils (1)(2)(3)(4)(5). Previous work has demonstrated that the recruitment of neutrophils contributes to further BBB destruction in murine models of GBS meningitis (7). As expected, in response to GBS infection, iPSC-derived BMECs upregulate CXCL8 (IL-8), CXCL1, and CXCL2, encoding potent neutrophil chemoattractants. Our findings agree with published results on the immortalized BMEC models, where the expression levels of CXCL8, CXCL1, and CXCL2 were upregulated between 6-and 15-fold (7). Interestingly, however, we did not observe upregulation of IL6, encoding the global proinflammatory cytokine IL-6. It is possible that IL-6 may be generated from other sources besides the brain endothelium in vivo, given that it is detected at high levels in the sera of infected animals (7). Further investigation is needed to determine the overall transcriptional profile of iPSC-derived BMECs during GBS infection. However, GBS infection of iPSC-derived BMECs induced a set of chemokines that act to orchestrate neutrophil recruitment and activation.
We recently demonstrated that GBS is able to induce the expression of the transcriptional repressor of tight junctions, Snail1, which contributes to tight junction disruption in immortalized hBMECs during GBS infection (17). In agreement with these findings, GBS infection of iPSC-derived BMECs upregulated Snail1 and substantially decreased the tight junctional continuity and resultant barrier properties. Furthermore, our previous work examining tight junction disruption was limited by the use of immortalized human BMEC cell lines that did not express claudin-5 and lacked continuous occludin staining and, thus, lacked an optimal barrier phenotype ( Fig. S4) (17). The iPSC-derived BMECs express claudin-5 and have properly localized occludin and, thus, allow for the much more relevant analysis of tight junction function during GBS infection.
In vivo, BMECs are supported by a number of other CNS cell types, such as the astrocytes, neurons, and pericytes that make up the NVU. Currently, little is known about the contribution of astrocytes, neurons, and pericytes to BBB function during meningitis. Models combining iPSC-derived neurons and astrocytes along with iPSCderived BMECs have been reported by us and others (29,34,35,37,38,57,60). Combined with this report establishing iPSC-derived BMECs as a model to study bacterial meningitis, these tools will likely support future studies examining the role of other NVU cell types and their contributions during infection.  ⌬srr2 (63, 64), and ⌬sfbA (12) and NCTC10/84 ⌬pilA (8) mutants have been described previously. GBS strains were all grown in Todd-Hewitt broth (THB) at 37°C. Lactococcus lactis was grown in M17 medium at 30°C (8). iPSC line DF19-9-11 (WiCell) was chosen for this study as this iPSC line was not generated through the use of viral integration vectors, eliminating the potential for inherent antiviral or anti-inflammatory responses. DF19-9-11 cells were grown in mTeSR1 medium (WiCell) that was changed daily and maintained on Matrigel (WiCell)-coated plates (Corning) consistent with previously published methods (29,(38)(39)(40)57). Immortalized hBMECs (a gift of Kwang Sik Kim and Monique Stins, Johns Hopkins University, Baltimore, MD) were cultured in RPMI 1640 containing 10% fetal bovine serum (FBS), 10% NuSerum, and 1% nonessential amino acids as described previously (6,7,12,17,31).

MATERIALS AND METHODS
Generation of BMECs and astrocytes from iPSCs. DF19-9-11 iPSCs were differentiated into BMECs according to methods published previously (29,(38)(39)(40)57). Briefly, a single-cell suspension of iPSCs was seeded at a density of 10,000/cm 2 onto Matrigel (WiCell) cell culture plates or flasks (Corning) and grown for 3 days. Differentiation was initiated by changing to UM medium (Dulbecco modified Eagle medium [DMEM]-F-12 medium plus 20% knockout serum replacement [KOSR], 1% minimal essential medium [MEM], 0.5% Glutamax, and 0.07% beta-mercaptoethanol) for 6 days, refreshing the medium daily. The medium was then changed to EC medium (human endothelial cell serum-free medium plus 1% platelet-poor plasma-derived serum [Fisher] and 500 ng/ml basic fibroblast growth factor [bFGF]) for 2 days. Finally, BMECs were purified on fibronectin-and collagen-coated plates or Transwell inserts (Corning, product number 3460). BMECs were analyzed for TEER using an EVOM II instrument (World Precision) and for the expression of endothelial markers as described previously (29,40). The addition of retinoic acid during BMEC differentiation has been shown to elevate BBB properties, including TEER and vascular endothelial (VE)-cadherin expression (38,39). However, since retinoic acid has been shown to have anti-inflammatory properties and could interfere with cellular responses to bacteria (65)(66)(67)(68)(69), it was not used. Astrocytes were generated as described previously (57).
Infection assays. iPSC-derived BMECs purified onto collagen-fibronectin-coated 24-well plates (Corning) were grown to a confluent monolayer. A multiplicity of infection (MOI) of 10 was used for all experiments unless noted specifically. Overnight cultures of GBS or L. lactis were grown in THB or M17, respectively, subcultured the following day, and grown to an optical density at 600 nm (OD 600 ) of 0.4. Bacteria were spun down and washed in phosphate-buffered saline (PBS) prior to infecting BMECs. Enumeration of total cell-associated and intracellular CFU was performed as described previously (6,7,31). Briefly, for adherence, bacteria were incubated with BMECs for 30 min at 37°C and 5% CO 2 , followed by 5 washes in PBS. Mammalian cells were lysed in 0.025% Triton X-100 and plated in a dilution series onto THB plates (M17 for L. lactis) to determine bacterial loads. For invasion, bacteria were incubated at 37°C and 5% CO 2 for 2 h, followed by 3 washes in PBS and incubation with antibiotic medium for an additional 2 h at 37°C and 5% CO 2 . After the second 2-h incubation, BMECs were lysed in 0.025% Triton X-100 and plated in a dilution series onto THB plates (M17 for L. lactis) to determine bacterial loads. Intracellular survival assays were run exactly like the intracellular assessment assays; however, cultures were incubated in antibiotic medium for extended periods of time as described previously (45). The following day, bacterial loads were determined by counting the colonies of countable dilutions and back calculating. Data are presented as total CFU recovered or as the percentage of the initial inoculum.
Immunofluorescence and area fraction index calculations. Immortalized hBMECs and iPSCderived BMECs were either infected with GBS for 5 h at 37°C and 5% CO 2 or left uninfected as controls. After infection, cells were fixed and stained exactly as reported previously (40), with the single exception that anti-ZO-1 antibody (catalog number 339100; Thermo Scientific) was used in place of the previously reported anti-ZO-1 antibody. Briefly, BMECs were fixed in 100% ice-cold methanol and stained overnight at 4°C. Secondary antibodies were utilized at a 1:200 dilution, and samples were visualized with an Olympus IX70 inverted fluorescence microscope using Nikon NIS image acquisition software. Fiji ImageJ was used to create merged images. Determination of the area fraction index of tight junction staining and continuity was performed as described previously (57). Using Fiji ImageJ, images were corrected for inconsistent fluorescence illumination using the Background Correction plugin. Next, the gray scale intensity profile was examined to determine a threshold value that minimizes background while maintaining the staining profile. Images were then converted to a binary image and processed using the outline filter to determine the perimeter of tight junction staining in pixels. Total pixels were normalized to the square root of the number of cells in the image.
Flow cytometry. To assay the expression of the ␤1 integrin, iPSC-derived BMECs were differentiated, fixed in 1% paraformaldehyde, and stained using a 1:1,000 dilution of anti-␤1 integrin antibody (catalog number NB100-63255; Abcam, Inc.) in 0.5% bovine serum albumin (BSA) in PBS overnight at 4°C. The following day, fixed cells were washed in 0.5% BSA in PBS twice and stained with anti-mouse IgG Alexa Fluor 488 antibody (catalog number A-11001; Life Technologies, Inc.) for 1 h at room temperature. Cells were then washed twice in 0.5% BSA in PBS and run on a BD Accuri C6 flow cytometry instrument for fluorescence-activated cell sorting (FACS) analysis.
Western blot analysis. Immortalized hBMECs and iPSC-derived BMECs were infected as described above for 5 h at an MOI of 10. Then, cells were washed three times in sterile PBS, and cell lysates were taken using radioimmunoprecipitation assay (RIPA) buffer plus protease inhibitor cocktail (catalog number 78443; Thermo Scientific). Proteins were quantified using a bicinchoninic acid (BCA) assay (catalog number 23227; Thermo Scientific), and equal amounts were loaded onto protein gels and transferred to nitrocellulose membranes. Anti-COX IV antibody was used as a protein loading control (catalog number 4850T; Cell Signaling Technologies), and anti-claudin-5 antibody (catalog number 35-2500; Thermo Scientific) to visualize claudin-5. Horseradish peroxidase (HRP)-conjugated secondary antibodies (Jackson Laboratory) and a BioRad ChemiDoc XRSϩ instrument were used to image the blots.
Statistics. GraphPad Prism version 5.0 (GraphPad Software, Inc.) was used for all statistical analysis. For pairwise comparisons, the 2-tailed Student t test was used where appropriate. For multiple comparisons, analysis of variance (ANOVA) was used to determine statistics. Data are represented as mean values Ϯ standard errors of the means (SEM) where triplicate mean values are presented and as mean values Ϯ standard deviations (SD) where raw values are presented. Statistical significance was accepted at a P value of less than 0.05.