|Home | About | Journals | Submit | Contact Us | Français|
The critical role of angiogenesis for solid tumor growth and metastatic spread has been well established. In contrast, even though increased vascularity has been commonly observed in bone marrows of patients with hematological malignancies (liquid tumors), the pathophysiology of leukemia induced angiogenesis in the bone marrow remains elusive. In this paper, we demonstrated the usage of a microengineered 3D biomimetic model to study leukemic cell induced bone marrow angiogenesis. Rational design of the 3D angiogenesis chip incorporating endothelial cells (ECs), leukemic cells, and bone marrow stromal fibroblasts provided an efficient biomimetic means to promote and visualize early angiogenic processes. Morphological features of angiogenesis induced by three different leukemic cell lines (U937, HL60, and K562) were investigated and compared. Quantitative measurements of angiogenic factors secreted from monocultures and cocultures of leukemic cells with bone marrow stromal fibroblasts suggested a synergistic relationship between ECs, leukemic cells, and bone marrow stromal fibroblasts for angiogenic induction, and also confirmed the necessity of conducting functional angiogenic assays in proper 3D biomimetic cell culture systems like the one developed in this work.
Angiogenesis describes the dynamic cellular process of formation of neo blood vessels from existing vasculatures. Angiogenesis is a highly regulated process orchestrated by a variety of factors either stimulating or inhibiting the proliferation, migration, and lumen formation of endothelial cells (ECs) [1,2]. Angiogenesis is involved in various physiological processes, such as development, growth, and wound healing [3–5]. In addition to its role in physiological processes, angiogenesis also plays a critical role in progression and metastatic spread of solid tumors [5–8]. As solid tumors grow beyond a certain size, additional blood supply is needed for oxygen and other nutrient transportation to support continuous growth of tumors. Neovessels generated from the angiogenesis process also facilitate spread of cancer cells through the circulation during metastasis [4,9–11]. Thus, for certain solid tumors, increased levels of angiogenic factors (e.g., VEGF) in the circulation have been associated with poor prognosis, and some anti-angiogenic drugs are currently used in the clinic to suppress solid tumor growth with varied levels of success [9,10,12–15].
In contrast to solid tumors, fewer studies so far have been conducted to examine specifically the role of angiogenesis in hematological malignancies (liquid tumors, including leukemia and lymphoma) [5,11–13,16]. In fact, most hematological malignancies, such as acute and chronic leukemia, lymphoma, myelodysplastic syndromes, myeloproliferative neoplasms, and multiple myeloma, are associated with angiogenesis in the bone marrow or lymphatic organs [2,5,6,17,18]. Histologically, angiogenesis is considered irrelevant to the pathologies of hematologic malignancies, based on the fact that these malignances do not develop compact tumor mass as solid tumors [11,19]. Recently, however, there is emerging clinical evidence of increased bone marrow microvessel density and increased levels of plasma proangiogenic cytokines associated with tumor progression in hematological malignancies. It has also been observed that after chemotherapy with cytosine-1-D-arabinofuranoside, residual leukemic cells are mainly localized in the perivascular endothelium and in contact with the trabecular endosteum, implying that the vascular endothelium may release antiapoptotics to help leukemic cells evade chemotherapy treatments [20–22]. This clinical correlation raises the interesting possibility for the functional role of angiogenesis in the bone marrow to promote proliferation and survival of leukemic cells [22,23].
The pathophysiology of leukemia induced angiogenesis in the bone marrow remains elusive. Most existing studies on angiogenesis in bone marrow are focused on measurements of secretion levels of selected angiogenic factors (such as VEGF and bFGF) from leukemic and stromal cells and studying effects of these angiogenic factors on endothelial cell behaviors in vitro using conventional two-dimentional cell cultures [6,24–26]. However, angiogenesis is a very complex cellular process involving endothelial invasion and proliferation in a three-dimentional (3D) environment regulated by dynamic cell-cell and cell-matrix interactions. Furthermore, bone marrow contains both cellular (e.g., hematopoietic cells, fat cells, blood vessel cells, bone cells, and fibroblasts) and non-cellular (e.g., hypoxia) components that can all have important effects on angiogenesis. As such, simply measuring a few angiogenic factors secreted from leukemic and stromal cells can only provide a partial picture that may not be sufficient to delineate the complex dynamic angiogenic processes.
Conventionally, angiogenesis under coculture conditions are studied using transwells, where ECs are seeded in the bottom chamber coated with Matrigel . Such 2D monolayer-based cell culture systems lack the complexity of in vivo 3D environments or key features of native angiogenesis, such as directional endothelial invasion into a 3D matrix and proper apical-basal polarity of lumen formation [28–30]. It is now well appreciated that conventional 2D cell culture systems are incapable of recapitulating dynamic and highly complex tissue architectures, leading to productions of results different from in vivo 3D environments [31–33].
Recently, different microengineered 3D biomimetic culture systems have been successfully demonstrated to model in vivo environments for different biological and biomedical studies. Compared to conventional 2D cell culture methods, microengineered 3D biomimetic models have been shown to resemble better the physiological environment while simultaneously allowing high-resolution imaging and direct quantification of dynamic cellular processes [34,35]. Importantly, microengineered platforms integrating 3D extracellular matrix (ECM) confined by surface tension have been utilized by several groups for studying cell-cell communication, cell migration, as well as vasculogenesis and angiogenesis [29,35–40]. Adapting the same approach, herein we demonstrated the usage of a microengineered 3D culture system (“the microfluidic 3D angiogenesis chip”) to quantitatively study leukemic cell induced bone marrow angiogenesis in vitro for the first time. Rational design of the microfluidic 3D angiogenesis chip provided an efficient means to promote and visualize early angiogenic processes induced by leukemic cells and bone marrow stromal cells. Morphological features of angiogenesis including endothelial invasion distance and area, tip cell number, and lumen structure were further investigated and compared for three different leukemic cell lines. Furthermore, we examined the effect of coculture of leukemic cells with bone marrow stromal cells on angiogenic sprouting and quantified 10 common angiogenic factors secreted from monocultures and cocultures of leukemic cells and bone marrow stromal cells. Together, the results demonstrated the utility of the microfluidic 3D angiogenesis chip as an in vitro 3D biomimetic model to study leukemic cell induced bone marrow angiogenesis and highlighted the potential applications of the chip to elucidate complex cell-cell interactions and their roles in coordinating bone marrow angiogenesis
To study the effect of leukemic cells on angiogenic invasion, sprouting, and lumen formation from ECs, we designed and fabricated a 3D biomimetic angiogenesis chip using PDMS by conventional soft lithography to facilitate controlled cell-cell communications while allowing direct characterization of angiogenic sprouting morphogenesis in situ (Fig. 1A). The 3D biomimetic angiogenesis chip, with its design comparable to previous studies [40,41], consisted of three parallel microchannels (100 μm in height and 1,000 μm in width) partitioned by trapezoid-shaped supporting posts spaced 100 μm apart. Each channel had two loading reservoirs at its both ends for sample loading and culture medium exchange. Rat tail collagen I gel matrix (2.5 mg mL−1) injected into the central channel (“the gel channel”) was locally confined in the channel owing to surface tension. The collagen matrix served as a paracrine interaction medium separating two side channels that would be loaded with leukemic (“the leukemic channel”) and endothelial cells (“the endothelial channel”). Upon gelation, ECs were injected into the endothelial channel and allowed to adhere onto the collagen gel interface within the endothelial channel (Fig. 1A, Materials and Methods). Importantly, collagen matrix in the gel channel provided a 3D space for angiogenic sprouting and neovessel formation. In addition, the vast majority of the organic matrix in the bone marrow is collagen, and collagen matrix has been extensively used as a two-dimensional substrate to study the effect of bone marrow matrix environment on related cell behaviors. For angiogenic assays with only leukemic cells, leukemic cell suspensions at a concentration of 2 × 106 cells mL−1 were loaded into the leukemic channel. For studies with leukemic cells cocultured with stromal cells, 0.1 × 106 HS5 stromal cells suspended in 50 μL collagen gel (5 mg mL−1) was first injected into both the loading reservoirs of the leukemic channel and allowed for gelation before leukemic cell suspensions were pipetted into the leukemic channel. The 3D microfluidic angiogenesis device was maintained at 37 °C and 5% CO2, and continuously monitored for 3 days to examine dynamic in situ angiogenic sprouting and neovessel formation of ECs under angiogenic induction by leukemic cells (Fig. 1B and Video S1). Figure 1B shows representative phase-contrast microscopic images taken daily to demonstrate angiogenic sprouting morphogenesis of ECs in the collagen gel towards the leukemic channel where an acute promyelocytic leukemia cell line (HL60) was cultured. Directional migration and invasion of ECs as either single, solitary cells or in the form of collective cell migration with long, multicellular sprouts were clearly visible in the collagen gel as early as Day 1 (Fig. 1B). Progressively more elongated and mature multicellular sprouts into the collagen matrix were observed over the 3 day assay period (Fig. 1B). To quantify dynamic angiogenic sprouting, phase-contrast microscopic images were further analyzed to determine invasion distance, invasion area, and the number of tip cells during angiogenic sprouting morphogenesis (see Materials and Methods, Fig. 1C). Together, observations in Fig. 1B&C supported that culturing leukemic cells in the leukemic channel could establish gradients of angiogenic factors across the collagen matrix to the ECs in the endothelial channel. The 3D microfluidic angiogenesis chip was able to provide a means to promote and visualize endothelial sprouting that could emulate early angiogenic processes.
Three different leukemic cell lines, U937, HL60, and K562, were selected as different leukemia disease models to study their angiogenic induction using the 3D microfluidic angiogenesis chip. U937, HL60, and K562 were derived from patients with histiocytic lymphoma, acute promyelocytic leukemia, and chronic myelogenous leukemia, respectively. Importantly, all three leukemic cell lines can secrete VEGF and other proangiogenic factors. In addition, clinical biopsy data also report higher bone marrow vascularity in acute promyelocytic leukemia and chronic myelogenous leukemia patients [17,42,43].
When the leukemic channel was loaded only with culture medium (Fig. 2A, Fig. S1), 3-day angiogenic assays using the 3D microfluidic angiogenesis chip revealed minimal invasion of ECs into the collagen matrix. By day 3, only a small number of ECs invaded into the collagen matrix as single, solitary cells with limited invasion distance or invasion area (Fig. 2B-D). No multicellular sprout was observed in the collagen matrix without leukemic cells loaded into the leukemic channel, consistent with previous studies showing critical requirements of angiogenic factors (either endogenous or exogenous) to promote endothelial sprouting and neovessel formation [30,40]. When loaded into the leukemic channel, each of the three leukemic cells alone induced directional migration and invasion of ECs into the collagen matrix towards the leukemic channel, supporting potent angiogenic induction by the three leukemic cell lines (Fig. 2A). Interestingly, different leukemic cell lines induced remarkably different angiogenic morphologies (Fig. 2A). Specifically, among all three leukemic cells, angiogenic induction by K562 cells drove ECs to invade furthest from the collagen gel interface and resulted in the highest number of tip cells. Endothelial invasion distance under angiogenic induction by K562 cells increased over time during the 3 days of culture (Fig. 2B); however, endothelial invasion area remained constant (Fig. 2C), since ECs invading into the collagen matrix were primarily single, isolated cells with very few multicellular stalks (Fig. 2A, D). HL60 cells induced shorter invasion distance and less tip cells when compared to K562; however, most tip cells were followed by multicellular stalks forming mature lumen structures in the collagen (Fig. 2A). For HL60 cells, endothelial invasion distance and invasion area both increased over time during the 3 day culture. Among all three leukemic cells, U937 cells demonstrated the weakest angiogenic induction potential as evidenced by the lowest EC invasion distance and area. For all the three leukemic cells, the number of tip cells and the percentage of isolated tip cells did not change significantly during the 3 day assay period (no significant difference, two group t-test) (Fig. 2D).
We further investigated the effect of leukemic cell concentration on angiogenic induction by increasing leukemic cell concentration from 1 × 106 cells mL−1 to 2 × 106 cells mL−1 (Fig. S2). Increasing U937 cell concentration resulted in heightened endothelial invasion distance and invasion area by day 3. Increasing HL60 cell concentration resulted in a significant increase of endothelial invasion area by day 3; however, endothelial invasion distance by day 3 was not significantly affected by HL60 cell concentration increase. In contrast, cell concentration increase of K562 cells led to a greater endothelial invasion distance by day 3, with endothelial invasion area not significantly affected by K562 cell concentration. For all the three leukemic cells, the total number of tip cells (including both isolated tip cells and tip cells with multicellular stalks) was not significantly affected by leukemic cell concentration increase by day 3.
Alongside the hematopoietic cells in the bone marrow are stromal cells such as fibroblasts, which secrete signals to mediate the bone marrow microenvironment [44,45]. In vivo studies of solid tumors have suggested a critical role of stromal fibroblasts in pathological angiogenesis and tumor progression [46,47]. In vitro studies have also demonstrated the important role of fibroblasts in promoting endothelial angiogenic sprouting and vascularization [29,48]. To examine specifically the effects of bone marrow fibroblasts on endothelial sprouting and angiogenic induction by leukemic cells, a commercially available bone marrow stromal cell line, HS5, was incorporated into the 3D microfluidic angiogenesis chip to mimic the bone marrow microenvironment. Herein, 50 μL collagen gel dispersed with 0.1 × 106 HS5 cells was first injected into both loading reservoirs of the leukemic channel and allowed gelation prior to loading leukemic cells into the leukemic channel. By day 3, HS5 stromal cells alone induced significant angiogenic sprouting and neovessel formation (Fig. 3A and Fig. S3). Compared to angiogenesis assays with leukemic cells only (Fig. 2A), most invading tip cells induced by HS5 cells alone were followed by multicellular stalks forming mature lumen structures in the collagen gel (Fig. 3A). Very few isolated, single tip cells were resulted from endothelial invasion induced by HS5 cells alone, presenting a good agreement with in vivo vasculature in healthy bone marrow samples [6,17,18]. Interestingly, at day 2 and day 3, HS5 cells cocultured with HL60 and K562 leukemic cells induced significantly greater endothelial invasion distance and invasion area and much more isolated tip cells compared to HS5 cells alone (Fig. 3 and Fig. S3). These observations agreed well with clinical observations of bone marrows of patients with acute myeloid leukemia, where endothelial sprouts and small microvessels without visible lumens are prevailing [18,49]. For HS5 cells cocultured with K562 cells, some tip cells in fact successfully invaded through the entire width of the collagen matrix (1,000 μm) and reached the leukemic channel by day 3. In contrast, HS5 cells cocultured with U973 cells resulted in a reduced invasion area compared to HS5 cells alone at both day 2 and day 3, even though endothelial invasion distance and isolated tip cell number were both significantly increased at both day 2 and day 3 (Fig. 3C).
We further compared endothelial invasion at day 3 induced by leukemic cells alone vs. leukemic cells cocultured with HS5 cells (Fig. S4). Compared to U937 cells alone, coculture of U937 cells with HS5 cells increased significantly endothelial invasion distance but decreased invasion area by day 3. No significant difference was observed for the total tip cell number or percentage of isolated tip cells between U937 cells alone and U937 cells cocultured with HS5 cells. For HL60 cells, there was no significant effect of HS5 coculture on endothelial invasion distance or invasion area, even though coculture with HS5 cells increased significantly the number of isolated tip cells. Coculture of K562 cells with HS5 cells increased significantly both endothelial invasion distance and invasion area but decreased the tip cell number and percentage of isolated tip cells by day 3.
In vivo, angiogenesis is mediated at least partially by combinational effects of different angiogenic factors, such as VEGF, bFGF, and HGF. Here, we applied microELISA array to quantify secretions of 10 proangiogenic factors (VEGF, bFGF, HGF, HB-EGF, ANG, ANG-2, PIGS, Leption, PDGF-BB and EGF) by leukemic (U937, HL60, and K562) and stromal (HS5) cells (see Materials and Methods). To this end, monoculture and coculture of leukemic cells (U937, HL60, and K562) and stromal cell HS5 were conducted in 24-well plates (Fig. 4A and Materials and Methods). For assays with HS5 cells, HS5 cells were first suspended in collagen gel before the gel solution was added to the bottom of the 24-well plate and allowed to gelation. For coculture conditions, leukemic cells were then seeded on top of the collagen gel. In monoculture conditions with leukemic cells alone, U937, HL60, and K562 cells were directly seeded into 24-well plates. After 24 hr, supernatants of cell cultures and leukemic cells were collected for measurements of angiogenic factor secretions and mRNA levels, respectively. The use of collagen gel to suspend HS5 cells in 24-well plates was to mimic culture conditions of HS5 cells in the 3D microfluidic angiogenesis chip. In the 24-well plate culture, the total leukemic cell and stromal cell numbers, cell concentration, and collagen gel concentration were controlled to be exactly the same as in the 3D microfluidic angiogenesis chip.
PIGS, Leption, PDGF-BB, or EGF was not detected by the microELISA array under either monoculture or coculture conditions for any leukemic or stromal cell (data not shown). Notable amounts of VEGF, bFGF, HB-EGF, ANG, and ANG-2 were detected from monocultures of HS5 cells (Fig. 4B-G), consistent with their potent angiogenic induction property demonstrated through in situ angiogenic assays conducted using the 3D microfluidic angiogenesis chip. Under monoculture conditions, VEGF, HB-EDF, ANG, and ANG-2 were all detected for U937, HL60, and K562 cells; both HL60 and K562 secreted detectable amounts of bFGF, but not U937, in agreement with previous studies [17,25]. Furthermore, HGF secretion was detected only from HL60 cells (either monoculture or coculture with HS5 cells), but not from any other leukemic or stromal cells.
VEGF, bFGF, and HGF are among the best studied angiogenic factors. It has also been shown that to efficiently induce angiogenic lumen structures, VEGF, bFGF, or HGF needs to be presented to ECs in combination [50–54]. Compared to HL60 and K562 monocultures, monoculture of U937 cells secreted the lowest levels of bFGF and HGF (Fig. 4B-D), consistent with their weakest angiogenic induction potential demonstrated using the 3D microfluidic angiogenesis chip. In contrast, monoculture and coculture of K562 cells led to the highest secretion levels of VEGF and bFGF among the three leukemic cells (Fig. 4B&C), correlating well with their potent angiogenic induction potential observed in the 3D microfluidic angiogenesis chip. Among the leukemic and stromal cells examined in this work, only HL60 cells would secrete a detectable amount of HGF (Fig. 4D). It should be noted that among all angiogenic factors measured, the level of ANG-2 secretion detected by microELISA under cocultures of leukemic cells with HS5 cells was significantly lower than HS5 monoculture condition, suggesting an inhibitory effect of leukemic-stromal cell-cell interactions on ANG-2 secretion.
mRNA expression of VEGF and bFGF by leukemic cells under both monoculture and coculture conditions were quantified using real-time PCR (Fig. 5). When cocultured with HS5 cells, HL60 cells showed 3-fold and 10-fold increases of VEGF and bFGF mRNA expression, respectively, from monoculture conditions. For U937 and K562 cells, there was no significant difference for mRNA expression of VEGF or bFGF between monoculture and coculture conditions.
The microELISA and mRNA data for HL60 cells consistently demonstrated that coculture of HL60 cells with HS5 cells could significantly increase secretion levels of VEGF and bFGF in cell culture. However, angiogenic assays conducted using the 3D microfluidic angiogenesis chip did not show any significant effect of coculture of HL60 cells with HS5 cells on endothelial invasion distance or invasion area when compared to monoculture of HL60 cells; only the number of isolated tip cells was increased significantly by coculture with HS5 cells (Fig. S4). This discrepancy underscored the limitation of using a few selected angiogenic factors measured through conventional cell cultures to gauge angiogenic induction potential of cells. Conducting functional angiogenic assays in a 3D biomimetic environment incorporating proper spatiotemporal cellular and cell-matrix interactions should provide more accurate determination of angiogenic properties of cells.
To further demonstrate the utility of the 3D microfluidic angiogenesis chip to examine angiogenic induction by specific angiogenic factors secreted from leukemic cells, additional angiogenesis assays under monoculture conditions of HS5 cells were performed using the 3D microfluidic angiogenesis chip with HGF (20 ng mL−1) supplemented into the leukemic channel. HGF concentration was selected based on microELISA measurements in Fig. 4D. As shown in Fig. S5, adding HGF into the leukemic channel induced greater endothelial invasion distance and invasion area and more isolated, single tip cells compared to monoculture of HS5 cells alone, confirming the angiogenic induction potential of HGF. It is of interest to note that monoculture of HS5 cells with HGF supplementation still showed different angiogenic invasion profiles from coculture of HL60 cells with HS5 cells, as evidenced by different endothelial invasion area and total tip cell number by day 3, underscoring again angiogenic invasion as an integrative cellular process involving combinational effects of different angiogenic factors.
Increased vascularity has been commonly observed in bone marrows of patients with hematological malignancies. However, the role of new blood vessel formation (or angiogenesis) in bone marrows in regulating the pathogenesis and progression of hematological malignancies is still elusive. The difficulty of in vivo study of bone marrow angiogenesis for hematological malignancies necessitates the development of in vitro functional angiogenesis assays that are capable of mimicking key features of 3D in vivo bone marrow environment. In this work, we demonstrated the usage of a microengineered 3D microenvironment for in situ characterization of angiogenic sprouting morphogenesis induced by leukemic cells. Through rational designs of different functional microfluidic channels, gradients of angiogenic factors secreted by leukemic cells were established across a collagen matrix. Angiogenic induction induced directional endothelial migration and invasion, and neovessel formation in the collagen matrix, emulating early angiogenic processes. Using the 3D microfluidic angiogenesis chip, we were able to discover unique morphogenic signatures of angiogenesis induced by different types of leukemic cells with or without cocultures with bone marrow stromal cells, which have not been demonstrated using any existing 2D cell culture approaches.
Importantly, monoculture of bone marrow stromal cells in the 3D microfluidic angiogenesis chip demonstrated such stromal cells’ potent angiogenic induction property in promoting angiogenic invasion and sprouting, and mature lumen formation with a limited number of isolated, solitary tip cells. Compared to stromal cell monoculture, coculturing bone marrow stromal cells with leukemic cells further promoted angiogenesis, featuring greater endothelial invasion distance and invasion area, and a larger number of isolated tip cells without multicellular stalks, in good agreement with bone marrow biopsy samples of leukemia patients, where leaky endothelial sprouts and small microvessels without visible lumens are common.
In addition to stromal cells and blood vessel cells, in vivo bone marrow microenvironment contains other cellular (e.g., bone cells, mesenchymal stem cells, and macrophages) and non-cellular (e.g., hypoxia) compartments that may have important effects on angiogenesis. In the future, these additional cellular and non-cellular components can be incorporated using existing, compatible microfluidic technologies in a straightforward manner into the 3D microfluidic angiogenesis chip to further establish the biomimetic bone marrow niche for in situ characterization of angiogenic sprouting morphogenesis. We also envision the exciting possibility to leverage the 3D microfluidic angiogenesis chip in the future to examine the specific roles of stromal cells and blood vessel cells in regulating leukemic cell behaviors in a bone marrow-like environment.
Most existing studies on angiogenesis have focused on a few potent angiogenic factors (such as VEGF) and studying their effects on endothelial cell behaviors. However, angiogenesis in the bone marrow is a dynamic, integrative cellular process that critically depends on spatiotemporal cell-cell and cell-matrix interactions that are abundant in the bone marrow microenvironment. It has now been appreciated as a limitation to use only a few growth factors as prognostic markers of angiogenesis, further corroborating the critical need of developing functional angiogenic assays using proper 3D biomimetic cell culture systems (such as the 3D microfluidic angiogenesis chip developed in this work). The experimental results presented in this work further provide important information for understanding leukemic cell induced bone marrow angiogenesis. In addition to be used as a general 3D in vitro culture platform to study angiogenic interactions between different bone marrow cellular and non-cellular components, by incorporating leukocytes isolated directly from peripheral blood or bone marrow of leukemia patients, the 3D microfluidic angiogenesis chip holds the promise for evaluating the effect of anti-angiogenic (anti-cancer) drug treatments on disease progression of individual leukemia patients [55,56].
The 3D microfluidic angiogenesis chip was fabricated using soft lithography by bonding a PDMS (Sylgard 184; Dow Corning) layer with designed features to a coverslip, as described previously [40,41]. Briefly, the PDMS layer was obtained by mixing PDMS base and curing agent (10:1 w/w) and then cast from a positive relief pattern of SU-8 (MicroChem) generated using standard lithography. Four loading reservoirs were then punched on the PDMS layer with an 8-mm biopsy puncher. After cleaning, both the PDMS layer and the coverslip were treated with oxygen plasma for 80 sec and gently pressed together to achieve irreversible bonding. The microfluidic angiogenesis device was then flushed with 1 mg mL−1 poly-D-lysine (PDL) solution (Sigma), incubated at 37 °C for 24 hr, and washed three times with DI water. To restore hydrophobicity of the PDMS device surface, the device was incubated at 80 °C for 2 hr before sterilized by exposing to UV light for 25 min. 20 μL type I collagen solution (2.5 mg mL−1; BD Biosciences) was prepared per the manufacturer's protocol and pipetted into the middle gel channel and allowed to gelation at 37 °C for 30 min.
Human umbilical vein endothelial cells (HUVECs; Lonza) were cultured in endothelial cell growth medium-2 (EGM; Lonza), grown till 80% confluency before subculture, and used at passages 3 - 6 in all experiments. Human leukemia cell lines U937, HL60, and K562 (ACTT) were cultured in RPMI-1640, IMDM, and DMEM (ATCC), respectively, and passaged per the manufacturer's protocol. RPMI-1640 and IMDM were supplemented with 10% FBS (Invitrogen) and 0.5% penicillin-streptomycin (Sigma). DMEM was supplemented with 20% FBS and 0.5% penicillin-streptomycin. Human bone marrow stromal cell line HS5 (ACTT) was cultured in DMEM supplemented with 10% FBS and 0.5% penicillin-streptomycin.
For cell seeding, HUVECs were first suspended at a concentration of 5 × 106 cells mL−1 during subculture. 20 μL HUVEC suspension and 20 μL EGM medium were then introduced into one of the loading reservoirs of the endothelial channel, before sucked into the device by applying gentle vacuum. HUVECs were allowed to adhere to collagen gel by tilting the device by 90 ° and incubated at 37 °C for 10 min. 300 μL EGM was then immediately added to the loading reservoirs of the endothelial channel. For experiments with only leukemic cells in the leukemic channel, leukemic cells were first suspended at a concentration of 1 × 106 cells mL−1 or 2 × 106 cells mL−1 before 300 μL cell suspensions were pipetted into the reservoirs of the leukemic channel. For assays with leukemic cells cocultured with stromal cells, HS5 cells were first suspended in DMEM at a concentration of 20 × 106 cells mL−1 before mixed with 5 mg mL−1 collagen solution to achieve a final cell concentration of 2 × 106 cells mL−1. After seeding HUVECs in the endothelial channel, 50 μL collagen gel containing HS5 cells was injected into the two loading reservoirs of the leukemic channel without blocking the channel. The device was slightly tilted and incubated at 37 °C for 30 min to allow gelation of collagen gel. 250 μL leukemic cell suspension was then loaded into the leukemic channel by applying gentle vacuum.
After cell seeding, the microfluidic angiogenesis device was incubated at 37 °C and 5% CO2 and continuously monitored for 3 d. During the 3 d assays, one third of culture medium in the reservoirs of the endothelial and leukemic channels was replaced daily by fresh EGM. During the 3 d assays, all leukemic cells remained suspended in the leukemic channel and did not migrate into the collagen gel. Therefore, there was no direct contact between endothelial cells and leukemic cells inside the 3D microfluidic angiogenesis chip. For angiogenic assays supplemented with HGF, 250 μL EGM supplemented with 20 ng mL−1 HGF was used as culture medium and exchanged every 12 hr.
The microfluidic angiogenesis device was washed twice with PBS before cells were fixed in 4% (w / v) paraformaldehyde (Sigma) in PBS for 10 min and then permeabilized using a 0.1% Triton X-100 (Sigma) in PBS solution for 5 min. Cell nuclei and F-actin were then labeled with DAPI for 15 min and phalloidin for 1 hr, respectively. Confocal fluorescent images were acquired using either a 10 × air objective or a 40 × oil immersion lens on an Olympus-IX81 microscope equipped with spinning disk confocal scanner unit (CSU-X1) and EMCCD camera (iXon X3, Andor) with the slice thickness of 1 μm. 3D image reconstruction and projection from z-resolved confocal stacks were rendered using Imaris 7.1.1 software (Bitplane).
Phase-contrast images were captured daily for 3 d with a Zeiss Observer.Z1 microscope (Carl Zeiss MicroImaging) attached with a thermoelectrically-cooled monochrome CCD camera (AxioCam HRM, Carl Zeiss MicroImaging). The entire gel channel of the microfluidic angiogenesis device was examined by stitching mosaic images of multiple adjacent areas. For each mosaic area, 7 z-stack images were taken at different z planes with a separation distance of 13 μm between each stack. Number of leading tip cells (either trailed with a multicellular stalk or as solitary single cells) was counted manually by two independent observers using the phase-contrast images [30,39,57]. Vertical distance between leading tip cells and the collagen gel interface within the endothelial channel was quantified as invasion distance. Three furthest tip cells from each measurement area (the collagen gel space between two neighboring supporting posts; there were 10 measurement areas in each device) were selected before the average invasion distance was calculated for all the tip cells as the endothelial invasion distance. A custom MATLAB algorithm was developed to measure projected area of angiogenic invasion of ECs as multicellular sprouts forming neovessels in the collagen gel using the 7 z-stack images (see Video S2). Briefly, the standard deviation of the pixel values at the same coordinate within the 7 z-stack images was calculated as a deviation matrix. The average of the numerical gradients along x and y directions of the deviation matrix was used to construct new images through dilation and erosion steps. Small contaminants with area smaller than 20 pixels were removed, and coordinates with pixel values lower than the background was also regarded as occupied by cells. The image was then binarized, and the projected area was measured by selecting specific area of interest (Fig. S6).
IMDM medium supplemented with 10% FBS was used for culturing cells in 24-well plates. For monoculture of leukemic cells (U937, HL60, and K562), 600 μL cell suspension solutions containing 1.2 × 106 cells were added into 24-well plates. For assays with HS5 cells, HS5 cells were first suspended in 5 mg mL−1 collagen gel at a concentration of 2 × 106 cells mL−1 before 100 μL gel solution was added to the bottom of the 24-well plate and allowed to gelation. After gelation, 500 μL culture medium was added into the 24-well plate for HS5 monoculture conditions. For coculture conditions, 500 μL leukemic cell suspensions containing 1.0 × 106 cells were added on top of the collagen gel in the 24-well plate. (In the 24-well plate culture, the total leukemic cell and stromal cell numbers, cell concentration, and collagen gel concentration were controlled to be exactly the same as in the 3D microfluidic angiogenesis chip.) Cell cultures were incubated at 37 °C for 24 hr before cell medium (for angiogenic factor measurements) and leukemic cells (for mRNA quantification) were collected. An angiogenesis antibody array kit (Quantibody Human Angiogenesis Array 1; Raybiotech) was used to quantify human angiogenic factors (VEGF, bFGF, HGF, HB-EGF, ANG, and ANG-2) per manufacturer's instructions. IMDM medium supplemented with 10% FBS was used as negative controls. Total mRNA of leukemic cells was extracted using RNeasy kit (Qiagen) following manufacturer's instructions. cDNA was synthesized using iScript cDNA synthesis kit (Bio-rad). RT-PCR was conducted using the CFX connect real-time PCR detection system (Bio-rad) using the following primers: bFGF (forward, 5′-AGAGCGACCCTCACATCAAG-3′; and reverse, 5′-TCGTTTCAGTGCCACATACC-3′), and VEGF (forward, 5′-CTACCTCCACCATGCCAAGT-3′; and reverse, 5′-GCAGTAGCTGCGCTGATAGA-3′).
Measurement results were compared using independent, two-tailed Student t-test in Excel (Microsoft). P < 0.05 was considered as statistically significant.
We acknowledge financial support from the National Science Foundation (CBET 1149401 and CMMI 1536087), the American Heart Association (12SDG12180025), and the UM Comprehensive Cancer Center Prostate SPORE Pilot Project (NIH/NCI P50 CA069568). The Lurie Nanofabrication Facility at the University of Michigan, a member of the National Nanotechnology Infrastructure Network (NNIN) funded by the National Science Foundation, is acknowledged for support in microfabrication.