Tissue-engineered 3D microvessel model of the blood-tumor barrier
Following the formation of metastatic lesions in the brain, the blood–brain barrier (BBB) displays locally altered structural and functional properties and is termed the blood-tumor barrier (BTB) [3]. To better understand the mechanisms of BTB formation and the interactions between cancer cells and brain microvascular endothelial cells (BMECs), we created a tissue-engineered model with a 150 μm diameter microvessel surrounded by brain metastatic cancer cells within a collagen I and Matrigel matrix [11, 18] (Fig. 1a, b). Both single cells and spheroids were formed using the human JIMT-1-BR cell line which displays brain tropism and is lethal in mice within 3–4 weeks following intracardiac injection due to extensive formation of brain metastases [15]. Spheroids embedded in the model mimicked sizes observed in rodent models of HER2+ brain metastases [30]. Microvessels were formed from iPSC-derived BMEC-like cells (iBMECs), which assemble into a confluent endothelium. For direct comparison, we also formed microvessels in the absence of cancer cells mimicking our previously reported model of the BBB [18]. All cell types were fluorescently-labeled to enable live-cell imaging: cancer cells (JIMT-BR-1 s) display stable GFP expression and iBMECs with RFP-labeled plasma membrane (Fig. 1c). To characterize the phenotype of the BTB over the course of 6 days, we utilized phase contrast and fluorescence imaging, functional assays, and analysis of gene and protein expression (Fig. 1d).
A tissue-engineered model of the blood-tumor barrier during metastatic breast cancer. a Schematic of model fabrication. A hydrogel matrix containing tumor cells or spheroids was formed around a template rod. After removal, the channel was seeded with iPSC-derived BMEC-like cells (iBMECs) to form a confluent monolayer. Microvessels were perfused by gravity flow at a shear stress of ~ 2 dyne cm−2. b Representative phase contrast images of metastatic breast cancer spheroids (JIMT-1-BR) and iBMECs on tissue-cultured treated plates prior to seeding into the BTB model. c Representative fluorescence images of the BBB microvessel model (formed by RFP-labeled iBMECs) and BTB model (formed by co-culture with GFP-labeled JIMT-1-BRs) at day 2 after iBMEC seeding. d Schematic of model cross-section highlighting key processes occurring within the BTB niche, including cancer cell proliferation, tumor-vessel interactions, vascular proliferation (angiogenesis) or degeneration, regulation of drug delivery, and immune cell interactions
Microenvironmental regulation of cancer cell growth
Prior to exploring tumor-vascular interactions, we sought to understand the dynamics of cancer cell growth within our model (Fig. 2). We first formed models with single cells or spheroids homogeneously distributed in the hydrogel matrix surrounding the microvessel. Tumor cell fate was determined from analysis of the fluorescence intensity of the cells over 6 days. In devices seeded with singularized cancer cells (seeded at a similar total cell density as spheroids), the fluorescence intensity in the matrix decreased slightly over time, where only a subset of single cells survived and proliferated (Fig. 2a). These cells were randomly distributed throughout the matrix with no preference for a specific location or proximity to the microvessel. In contrast, cancer cell spheroids showed robust growth over 6 days of culture (~ twofold increase in fluorescence), significantly higher growth compared to single cells (p = 0.011, unpaired t-test) (Fig. 2b). Cancer spheroids maintained well-defined borders with no evidence of shedding of single cells or clusters of cells. Thus, growth of cancer spheroids could be tracked by either fluorescence intensity or by spheroid area, which were strongly correlated (r2 = 0.677). Analyses of brain metastases from autopsy specimens previously characterized three patterns of cancer cell invasion: (1) well-demarcated growth (51%), (2) diffuse (single cell) infiltration (32%), and (3) vascular co-option (18%), where the percentages represent relative frequencies [31]. In our model, cancer spheroids matched patterns of well-demarcated growth but also displayed vascular co-option and single cell infiltration at later time points (discussed further below).
Dynamics of cancer growth within the BTB model. a Representative time course images of single cells, spheroids, and spheroids under static conditions in the BTB model. Day 0 images shows phase contrast, while other timepoints show JIMT-1-BRs (green). Tumor cells were seeded at an average density of 150,000 cell mL−1. White arrows show similarly sized spheroids at day 0 that grow only under flow conditions. b Spheroids showed a twofold increase in fluorescence over six days. Single cells displayed a small decrease in overall fluorescence over six days with both cell loss and proliferation. Data collected across n = 3 devices formed using single cells and n = 6 devices formed using spheroids. c Microvessel perfusion increased growth of cancer spheroids. Spheroid size is reported as the projected area of all spheroids in maximum intensity projections. Data collected across n = 85 spheroids under flow (11 independent devices) and n = 18 spheroids cultured under static condition (5 independent devices). d–e Relationship of spheroid growth rate to distance from microvessel and initial spheroid size. The growth rate was determined from the difference in projected area of all spheroids between days 2 and 6. Data collected across n = 85 spheroids from 11 independent devices. Initial spheroid area is plotted on a log scale. Data are presented as mean ± SD. *p < 0.05
We also tested the contribution of microvessel perfusion on spheroid growth rate. The growth rate was inferred from the fold change (FC) in projected area of each spheroid between days 2 and 6 (FCarea). In the absence of flow, spheroids displayed reduced growth compared to spheroids cultured under perfusion (p = 0.027, unpaired t-test) (Fig. 2c). Next, we assessed the influence of proximity to the microvessel and initial spheroid size. Spheroid growth rate was independent of vascular proximity, suggesting that nutrient transport does not lead to a gradient of spheroid growth in our model (Fig. 2d). However, we did observe a significant dependence on initial spheroid size, with the growth rate increasing exponentially with decreasing spheroid area (Fig. 2e). Cancer cells within larger tumors have less access to nutrients, suggesting that spheroids recapitulated nutrient gradients observed in patient tumors [32].
Vascular degeneration and vessel co-option within the blood-tumor barrier
Preclinical studies have shown that the BTB displays heterogeneous leakiness and hence therapeutic doses of anti-cancer agents are achieved inconsistently in most metastatic lesions [30, 33]. To better understand the barrier properties of the BTB, we conducted time-course imaging of BBB and BTB microvessels over 1 week. We first describe vascular degeneration and, in the next section, tumor—vessel interactions. The presence of cancer spheroids resulted in more rapid degeneration compared to control microvessels (Fig. 3a). We observed two modes of degeneration: (1) vascular collapse, and (2) vascular defects. Vascular collapse was most common and occurred as the endothelium physically detached from the surrounding matrix, while defect formation occurred as small holes (~ 1–5 cells) in the endothelium without changes in lumen diameter (Fig. 3a). Upon the appearance of defects or endothelium collapse, these effects quickly worsened over time as evident by widespread distribution of holes and large absence of endothelial cells at day six (Fig. 3a). We quantified the lifespan of BBB and BTB microvessels as the day on which vascular degeneration was first observed: BTB microvessels displayed a lifespan of 3.50 ± 0.23 days, while BBB microvessels displayed a higher lifespan of 5.00 ± 0.31 days (significant lower survival in BTB, p < 0.001, Gehan-Breslow-Wilcoxon test) (Fig. 3b).
Vascular degeneration and co-option by cancer cells within in vitro metastatic blood-tumor barrier. a Time course images of BBB and BTB models (day 2 to day 6). Two modes of vascular degeneration (collapse and defects) were observed in the presence of metastatic spheroids. White arrows identify the locations of initial signs of vascular collapse or defect formation. iBMECs (magenta) and JIMT-1-BR (green). b Lifespan across n = 17 BBB microvessels and n = 32 BTB microvessels. ***p < 0.001. c–d Schematic of tumor-vessel interactions and their cumulative frequency across n = 32 devices. e Higher magnification images of tumor-vessel interactions. The inset shows the xz projection of the confocal image, while other the images are epifluorescence. Images have brightness and contrast enhanced to enable visualization of the interactions. f Time course imaging of perivascular tumor growth and mosaic vessel formation. See also Additional file 1: Fig. S1
Mechanisms of tumor – vessel interactions in the BTB
Various mechanisms of tumor – vessel interactions have been observed in different settings [2, 3, 11, 34, 35]. During metastasis, extravasated cancer cells can remain in physical contact with the abluminal vessel wall and can proliferate along the basement membrane, a process termed vascular co-option. Extravasated cancer cells thrive in this perivascular niche due to proximity to nutrients and autocrine factors secreted by endothelial cells, while also being resistant to anti-angiogenic therapies which are effective in primary tumors, in part, due to the leaky vasculature (known as the enhanced permeability and retention effect). In mosaic vessel formation, cancer cells physically displace endothelial cells in the vessel wall, a process that can mediate intravasation of single cancer cells or cancer cell clusters. The growth of tumor spheroids can also locally compress the lumen of microvessels, resulting in vessel constriction and formation of dead-ends or string vessels. Finally, release of growth factors can attract endothelial cells to a nearby spheroid, a process termed vessel pull.
In our BTB model we observed only two tumor – vessel interactions: mosaic vessel formation and vascular co-option; we further subdivided vascular co-option based on whether cell migration occurred before or after BTB degeneration (Fig. 3c). Mosaic vessels formed in 53% of devices and roughly linearly increased in frequency over the initial 4 days of culture. Vascular co-option at early timepoints was directly associated with mosaic vessel formation (prior to BTB degeneration), while vascular co-option at late timepoints was directly associated with BTB degeneration (Fig. 3d). Using confocal microscopy, we confirmed that mosaic vessels were formed as evident by vascular cross-sections with cancer cells replacing endothelial cells in the vessel lumen (Fig. 3e). Migration along microvessels before and after BTB degeneration, suggests that vascular co-option can occur throughout progression of metastatic growth (Fig. 3e). Critically, directed proliferation and growth along blood vessels has been observed in autopsy specimens of brain metastases from solid cancers [31], but the spatiotemporal dynamics of this process are not well characterized. While multiphoton microscopy of brain metastasis in mice has visualized the early stages of tumor-vessel interactions when cancer cells are in close contact with blood vessels and perivascular growth occurs by vascular co-option [35], our studies suggest that cancer growth can also promote vascular degeneration. Spheroids in close proximity to microvessels displayed sustained perivascular growth and mosaic vessel formation through progressive displacement of endothelial cells (Fig. 3f). For a subset of devices, we continued perfusion for 2 weeks to determine terminal tumor-vessel interactions. We observed that mosaic vessels were formed in 100% of devices (n = 4) and that cancer cells, in some cases, completely replaced the endothelium (Additional file 1: Fig. S1).
Time course of BTB barrier function
Brain metastases are typically identified using gadolinium-enhanced MRI; however, despite gadolinium permeation into these tumors, evidence suggests that the metastatic BTB is heterogeneously permeable to drugs and other compounds [3]. In pre-clinical animal studies, the distribution of many chemotherapeutics is similar to fluorescently-labeled dextrans, indicating a paracellular pathway for drug transport across the BTB [3, 30, 33]. To quantify how barrier properties change in the presence of metastatic cancer cells, we measured permeability of BBB and BTB microvessels by co-perfusion with non-human-specific immunoglobulin G (IgG; Cascade blue-conjugated) and anti-HER2 IgG (Alex Flour-647-conjugated). Anti-HER2 IgG is a research grade biosimilar of Trastuzumab (Herceptin®), which is widely used for treatment of human epidermal growth factor receptor 2 (HER2)–positive metastatic breast cancer [36].
From time-lapse fluorescence images, we found that both BBB microvessels and BTB microvessels displayed negligible permeability to antibody at early time points (day 2) (Fig. 4a). However, after 4 days of culture, BTB microvessels were uniquely leaky to antibodies, matching observations of physical degeneration and formation of defects in the endothelium. While limited sites of leakage were observed at day 2 for both model types and day 4 for BBB microvessels, ~ 10 leakage sites per cm were observed in BTB microvessels, representing a significant increase compared to controls (p = 0.027, unpaired t-test) (Fig. 4b, Additional file 1: Fig. S2a).
Antibody permeability and accumulation dynamics within an in vitro metastatic blood-tumor barrier model. a Representative images of antibody permeability with and without cancer cells. Images are at 30 min after perfusion with non-specific or anti-HER2 IgG in BBB and BTB microvessels. iBMECs (red), non-specific IgG (blue), anti-HER2 IgG (magenta), JIMT-1-BR (green). b Quantification of focal leaks between BBB and BTB microvessels over time (n = 4–5 independent microvessels per condition). c Representative images of antibody accumulation within the BBB and BTB at day 4. Images are normalized to day 2 fluorescence. d Quantification of antibody accumulation across BBB endothelium, BTB endothelium, and cancer spheroids (n = 4 independent microvessels per condition). Data are presented as mean ± SD. *p < 0.05. See also Additional file 1: Fig. S2
To probe pathways for transport, we measured the accumulation of the two compounds in the endothelium and cancer spheroids (Fig. 4c, d). Following 30-min perfusion with IgG or anti-HER2 IgG, microvessels were perfused in the absence of antibody for 2 days prior to reassessing permeability. This washout period enabled us to determine how antibodies accumulated across conditions and locations. Critically, these experiments were conducted prior to degeneration of BTB or BBB microvessels, so that accumulation in cancer cells or spheroids, or endothelial cells, resulted from transport at day 2 of initial exposure to the IgGs. We found that non-specific IgG did not accumulate in endothelial cells or cancer cells (p > 0.05 for all compartments as tested by a one sample t-test to a hypothetical value of 1.0); however, anti-HER2 IgG was significantly accumulated in cancer spheroids (p = 0.029), without significant accumulation in the endothelium of BBB or BTB microvessels (p = 0.081 and 0.099) (Fig. 4c, d). HER2 antibody–drug conjugates (ADCs, 198 kDa) are able to cross the BTB in vivo but not 3 kDa dextran [37]. In vitro studies have suggested that uptake is limited to a subset of endothelial cells that support an endocytic transcellular pathway for transport [37]. Our results corroborate these findings, suggesting that Herceptin but not dextran can accumulate in the endothelium of the BTB and subsequently accumulate within perivascular metastatic tumors, even prior to paracellular barrier breakdown which is highly heterogeneous within metastatic lesions in vivo. Similar experiments were conducted using fluorescently-labeled 3 kDa dextran and fluorescently-labeled bovine serum albumin (BSA) (data not shown); paracellular focal leaks were also increased in BTB models at day 4 for these two solutes. Albumin accumulated in both cell types (endothelial and cancer), without significant differences between BBB and BTB microvessels, while 3 kDa dextran displayed no intracellular accumulation.
Metastatic BTB displays unique gene expression
The full repertoire of tumor-vessel interactions at the BTB is not well understood as most studies focus on functional assays such as permeability. To characterize changes in the BTB with genome-wide resolution, we performed transcriptomic profiling of iBMECs within BBB and BTB microvessels using bulk RNA sequencing. A day two timepoint was chosen to identify changes in BTB phenotype prior to BTB degeneration, mosaic vessel formation, and vascular co-option. This approach minimized cancer cell contamination as we lysed endothelial cells in microvessels that did not display mosaic vessels.
We observed distinct clustering of BBB and BTB endothelium using principal component analysis (PCA) (Additional file 1: Fig. S3a) and identified 988 BTB-enriched and 480 BBB-enriched transcripts (Fig. 5a). To validate these findings, we conducted semi-quantitative immunofluorescence for three proteins: serine protease 3 (PRSS3), intracellular adhesion molecule 1 (ICAM1), and k-ras (KRAS). BTB enrichment was confirmed at the protein level for serine protease 3 (p = 0.022, unpaired one-tailed t-test) and intracellular adhesion molecule 1 (p = 0.048), while a non-differentially expressed transcript (KRAS) maintained similar protein expression between both models (p = 0.187) (Fig. 5b, Additional file 1: Fig. S2c). To predict phenotypic differences between the BBB and BTB, we conducted gene-set enrichment analysis (GSEA) on Hallmark gene sets (Fig. 5c, Additional file 1: Fig. S4). BTB-associated Hallmark gene sets and transcripts driving their enrichment included: interferon alpha/gamma responses (IRF1, IRF5, CCL2, CCRL2, ICAM1), apoptosis (CASP8, MMP2, TNFRSF21), IL-6 signaling (IL6, IL18R1, TNFRSF1A, CXCL10), coagulation (TIMP3, PECAM1, MMP9), complement (PRSS3, PLAT, PLAUR, SERPINE1, ADAM9), hypoxia (VEGFA, FOS, FOSL2), among many others. BBB associated Hallmark gene sets included those associated with canonical BBB functions including wnt-beta catenin signaling and NOTCH signaling (WNT5A, HEY1).
Blood-tumor barrier phenotype within in vitro metastatic lesions. a Volcano plots depicting significantly (adjusted p < 0.05) upregulated genes (blue) and downregulated genes (red) between BBB and BTB microvessels. Bulk RNA was collected from control microvessels (n = 3) and microvessels surrounded by JIMT-1-BR spheroids (n = 3), two days after seeding of iBMECs. b Semi-quantitative validation of PRSS3, ICAM-1, and k-ras protein levels (n = 4). Representative immunofluorescence images of BBB and BTB microvessels at day 2 are shown with DAPI-labeled nuclei in blue. c Lollipop plot of select Hallmark gene sets enriched and depleted in BTB microvessels. d–e Representative images and quantification of THP1 (monocyte-like) immune cell and JIMT-1-BR cancer cell adhesion to BBB and BTB microvessels (n = 4). Arrows denote adherent cells. iBMECs (red), JIMT-1-BRs (green), THP-1 s (magenta). f–g Representative images and quantification of cell turnover events (proliferation and cell loss) between BBB and BTB microvessels (n = 7–8). Asterisks denote proliferation and cell loss events. Turnover is calculated as the difference between rates of cell proliferation and cell loss. Data are presented as mean ± SD. Statistical analysis was performed using a student’s unpaired t-test (two-tailed with unequal variance); *p < 0.05. See also Additional file 1: Figs. S3, S4 and Additional file 2: Data S1
Canonical endothelial transcripts (i.e. KDR, CDH5, VWF) were not uniformly altered within the BTB endothelium, suggesting that endothelial identity was not perturbed by cancer co-culture (Additional file 1: Fig. S3b). Given recent findings on iBMEC expression of epithelial transcripts [38], we sought to determine if cancer co-culture mediates vascular changes by augmenting epithelial identity of iBMECs. However, we found that most epithelial transcripts (i.e. CDH1, EPCAM, CLDN6) were not differentially expressed between the two models. Additionally, breast cancer marker genes were not broadly upregulated in iBMECs in the BTB model, suggesting minimal contamination of non-endothelial cells (Additional file 1: Fig. S3b); cancer associated transcripts that were upregulated including ESR1 and PLAU are known to be expressed by endothelial cells [39, 40]. Lastly, we benchmarked our findings to recent transcriptomic profiling of endothelial cells isolated from normal brain tissue and from patients with lung adenocarcinoma metastases [41]. 10% of BTB-enriched genes found here were also upregulated in endothelial cells isolated from lung adenocarcinoma metastases, including transcripts involved in epithelial-to-mesenchymal transition (SERPINE1), angiogenesis (VEGFA, CNN1) and extracellular matrix organization (ITGB6, ITGA5, COL6A1, COL6A2) (Additional file 2: Data S1). Since endothelial isolations from human tissue also contain contaminating mural and glia cells, a more precise benchmarking of our findings is not possible.
Metastatic BTB displays elevated immune cell adhesion and endothelial turnover
Gene set enrichment analysis was suggestive of diverse functional differences within the BTB, as well as loss of canonical BBB functions. We explored various functional responses between BTB and BBB microvessels, including adhesion of cancer/immune cells and endothelial turnover. Post-capillary venules are the preferential site of cancer cell and immune cell extravasation due to low shear stress and unique protein/gene expression [42]. To probe differences in cell adhesion to the endothelium, we perfused BTB and BBB microvessels with fluorescently labeled cancer cells and monocytes (Fig. 5d). The BTB displayed significantly increased adhesion of monocytic cells to the endothelium (~ 55 cm−1) compared to BBB microvessels (~ 5 cm−1) (p = 0.017, unpaired t-test) (Fig. 5e). As monocyte-derived macrophages accumulate in brain metastases [43], our model recapitulates the early stages of the transmigration cascade. This same effect was not observed following perfusion with single cancer cells (p = 0.421), suggesting differences in the mechanism of adhesion. Although adhesion of cancer cells to the endothelium were rare events in both models, these cells remained adherent for multiple days validating this model for studies of early stages in the metastatic cascade. In mouse models of brain metastases, the BTB endothelium expresses tumor necrosis factor (TNF) receptors conferring selective vulnerability to TNF-induced permeabilization compared to the BBB [44]. Indeed, multiple TNF receptors were upregulated in BTB microvessels, including TNFRSF1A, TNFRSF4, TNFRSF9, TNFRSF18, TNFRSF21. Furthermore, monocyte-derived macrophages (MDMs) are enriched in human brain metastases [43, 45], consistent with the BTB being conducive of monocytic infiltration. Immune cell adhesion is likely mediated by an upregulation of both endothelial surface adhesion molecules (i.e. ICAM1) and TNF receptor superfamily members.
To determine the influence of tumor spheroids on endothelial cell dynamics, we quantified iBMEC proliferation and cell loss prior to vascular degeneration (at day 2). Cell proliferation and cell loss events were manually counted from time-lapse phase contrast microscopy focused on the microvessel midplane and are reported as the number of events per hour (% h−1) (Fig. 5f). In control BBB microvessels, the rates of cell proliferation, cell loss, and overall turnover were similar to previous measurements [18]. However, BTB microvessels displayed slightly lower rates of cell proliferation (p = 0.458, unpaired t-test) and slightly higher rates of cell loss (p = 0.260), leading to significantly lower cell turnover (p = 0.047) (Fig. 5g). The negative turnover shows that cell loss dominates in the presence of metastatic spheroids, matching observations of vascular degeneration at late time-points. Cancer cells can mediate endothelial damage and angiogenesis by the secretion of soluble factors [46, 47] and by direct tumor-vessel interactions [48]. Indeed, BTB-enriched transcripts included mediators of apoptosis (CASP8) that are induced by cancer secreted factors and death receptor 6 (TNFRSF18) a master regulator of tumor cell-induced endothelial necroptosis [48]. iBMECs show angiogenic activity in the presence of growth factors [49] and VEGF signaling was upregulated at the transcriptional level; however, we did not observe angiogenic sprouting within the BTB model, which was also not observed in our prior model of metastatic breast cancer [11]. These observations are likely model dependent and cancer cell type dependent, as observed in multiphoton imaging studies of the metastatic cascade in vivo [35].
Mechanisms of BTB degeneration and dysfunction
BTB phenotype is derived from both physical and chemical interactions between cancer cells and the brain endothelium. As elevated immune cell adhesion and reduced endothelial cell turnover were observed at early time points prior to vascular degeneration and direct cancer-endothelial contact, chemical interactions likely represent a key mediator of BTB phenotype. To explore chemical interactions further, we conducted experiments in 2D Transwell models using: (1) conditioned microvessel media, and (2) co-culture with cancer spheroids (Additional file 1: Fig. S5a). These experiments remove contributions from physical interactions by depleting conditioned media of cells (by centrifugation) and by using cancer spheroids in the basolateral chamber enabling only chemical crosstalk. Interestingly, exposing 2D iBMEC monolayers to BBB or BTB-conditioned media did not elicit changes in barrier function over 1 week (p = 0.910, one-way ANOVA) (Additional file 1: Fig. S5b). Similarly, direct co-culture of spheroids in the basolateral chamber of 2D Transwells did not induce barrier loss but instead resulted in a small increase in TEER values (p = 0.039, paired t-test) (Additional file 1: Fig. S5c). These findings suggest that 2D and 3D microenvironments may possess differences in cancer-derived factors and/or iBMEC responses to cancer-derived factors. Indeed, in 3D models, cancer spheroid proximity to endothelial cells is greatly reduced compared to Transwells (~ 2 mm in Transwells) and cancer growth is significantly elevated (~ 1.15-fold growth in Transwells to ~ twofold growth in 3D) (p = 0.031, unpaired t-test) (Additional file 1: Fig. S5d).
To probe chemical factors that may mediate BTB phenotype, we performed ELISA for six analytes in perfusate from BBB and BTB microvessels collected at day 2 (Additional file 1: Fig. S5e–f). We note that analyte concentrations were highly variable in BTB microvessels compared to BBB microvessels (concentration standard deviation was 13-fold higher), suggesting that differences in spheroid density, size, and proximity may alter analyte concentrations and, in turn, BTB phenotype. TNFα receptor (TNFR1), TNFα, and VEGF were not significantly altered between the two perfusates (p > 0.05, unpaired t-test). Only interleukin-8 (IL-8, CXCL8) was significantly elevated in BTB microvessel perfusate (fold change = 11; p = 0.013), while IL-6 displayed small but non-significant increases (fold change = 5; p = 0.079) and IL-1b displayed small but non-significant depletion (fold change = 2; p = 0.094).
Macrophages augment BTB phenotype
Human brain metastases originating from primary breast cancer are comprised of ~ 30% immune cells, including resident microglia, monocyte-derived macrophages (MDMs), neutrophils, and T cells [45]. MDMs are particularly enriched in brain metastases and are localized to the perivascular region [43], suggesting that significant immune cell extravasation occurs during tumor progression. However, the contribution of macrophages to BTB phenotype remains unknown [3]. Tissue-engineered models are uniquely suitable for exploring immune cell contributions to tumor progression by avoiding issues of species-to-species differences in animal models [5].
To explore the contributions of monocyte-derived macrophages, we differentiated a monocytic cell line into macrophages using phorbol 12-myristate 13-acetate (PMA) and seeded these cells into the hydrogel matrix of the BTB model (Fig. 6a). Interestingly, tumor growth was slightly, but not statistically significantly, reduced by macrophage co-culture (p = 0.097, unpaired t-test) (Fig. 6b). Macrophage co-culture did not affect rates of microvessel degeneration (p = 0.913, Gehan-Breslow-Wilcoxon test) (Fig. 6c). To determine the effects of macrophage co-culture in an unbiased manner, bulk RNA sequencing of iBMECs was conducted at day 2. The magnitude of gene expression differences was much lower than comparison of BTB and BBB microvessels (Additional file 1: Fig. S3a). However, we identified 142 genes upregulated and 52 genes downregulated in the presence of macrophages (Fig. 6d). While macrophage marker genes were not broadly upregulated, increased expression of some markers suggests possible low levels of contamination (i.e. CD68, Additional file 1: Fig. S3b). Determining the degree of contamination is challenging as many gene families enriched in immune cells can also be expressed in endothelial cells and are responsive to inflammatory conditions. For example, toll-like receptors (TLRs) are widely expressed by endothelial cells and expression is increased in response to TLR ligands [50]; BTB microvessels with co-cultured macrophages displayed increased TLR1 and TLR2 expression. Transcripts depleted from iBMECs in BTB microvessels with macrophages included those involved in extracellular matrix organization (COL17A1, COL3A1, COL5A2). Hallmark gene sets associated with macrophages included interferon gamma responses (CASP1, IRF5, CCL5), among many other pathways, suggesting that macrophages can further augment BTB phenotype (Fig. 6d, e, Additional file 1: Fig. S4). To understand possible inflammatory cues that may mediate these gene expression changes, we compared analyte concentrations in microvessel perfusate with and without macrophage co-culture. IL-8, IL-1b, and TNFα were significantly elevated in BTB models with macrophage co-culture (fold changes = 11, 3.5, and 1.7, respectively) (p = 0.024, 0.004, and 0.045, respectively; unpaired t-test) representing possible factors secreted by macrophages or produced by cancer/endothelial cells in response to macrophage co-culture. The IL-8 concentration was ~ 11-fold higher with macrophage co-culture, representing the largest fold-change across analytes (Fig. 6f). Previous in vitro studies identified the effect of IL-8 on endothelial cells including tight junction downregulation in a dose- and time-dependent manner [51]; however, these findings are at concentrations orders of magnitude higher than levels measured here. Further studies are needed to determine functional differences induced by macrophage co-culture, but they appear to be more nuanced then directly augmenting cancer growth or microvessel degeneration.
Blood-tumor barrier phenotype in the presence of macrophages. a Representative image of BTB model with CellTracker-labeled macrophage co-culture. iBMECs (red), JIMT-1-BR (green), macrophages (magenta). b–c Quantification of BTB spheroid growth and microvessel lifespan with and without macrophage co-culture. End points determined when iBMECs are > 50% detached or collapsed from the hydrogel. Quantification across n = 10 BTB devices and n = 11 BTB + macrophage devices. d Volcano plots depicting significantly (adjusted p < 0.05) upregulated genes (blue) and downregulated genes (red) between BBB and BTB microvessels (n = 3 replicates each). Bulk RNA was collected from control microvessels (n = 3) and microvessels surrounded by JIMT-1-BR spheroids (n = 3), 2 days after seeding of iBMECs. e Lollipop plot highlighting selected Hallmark gene sets enriched and depleted in BTB microvessels. f Comparison of IL-8 concentration between BTB and BTB + macrophage microvessels (n = 5 microvessels per condition). Data are presented as mean ± SD. *p < 0.05. See also Additional file 1: Figs. S3, S4 and Additional file 2: Data S1
Add Comment