Co-culture Microdevice with Oxygen Gradient for Tumor Microenvironment Model and Metastasis Imaging

Tumor hypoxia is a major therapeutic problem since it decreases radiation effects and leads to metastasis. Oxygen is delivered to tumor tissue via abnormal and dysfunctional microvessels, which forms heterogeneity of tissue oxygenation in the tumor. Mimicking the oxygen gradient fo r cellu lar experiments in vitro is important to clarify the mechanis ms involved in tumor bio logy, but the only method to produce hypoxic conditions at a constant level is using gas-controlled incubators, because there is currently no technique for creating an oxygen gradient using culture dishes. We designed a polydimethylsiloxane (PDMS) microflu idic device integrated with microchannels for cell cultures that enables visualizat ion of cellu lar distribution under a microscope and co-culture to determine interactions between cancer and other cells. Phosphorescence-based partial oxygen measurements quantified the oxygen grad ient, which can be controlled by the gas pressure between the inlet and outlet of the device. A monoculture of end othelial cells with an oxygen gradient in the device showed an increase in cell death in the hypoxic area. In addition, Lewis lung carcinoma cells co -cultured with endothelial cells showed gradient-dependent migration through a membrane pore filter, indicating that the interaction between tumor and endothelial cells under hypoxia is crucial in metastasis. The results suggest that the developed microdevice can be used to study the mechanisms of tumor metastasis under hypoxic conditions.


Introduction
Tumor cells consume o xygen during active division, and abnormal and dysfunctional microvessels form severe hypoxic conditions in the tu mors [1,2]. Tissue hypoxia causes cells to accumulate hypoxia -inducible factor-1 alpha (HIF-1α), which is an o xygen-sensing transcriptional factor that transactivates genes encoding proteins contributing to homeostatic responses to hypoxia [3] and is implicated in tumor metastasis [4]. In addition, hypoxic conditions in tumors decrease rad iation effects [5]. To clarify the mechani sms of tumor gro wth and metastasis and the effects of radiation under hypo xia, cellu lar experiments in an environment with a controlled o xygen concentration are required.
However, d ifficu lty in rep roducing tumo r internal en v iro nment is a barrier to tumor research in vitro. Tumor cells are constant ly incub ated at 37℃ und er 20% O 2 as a controlor at 37℃ under 1%-5% O 2 as hypoxia, wh ich does not reproduce the conditions inside tumors. Furthermore, interactions between tumor and other types of cells, such as endothelial cells, pericytes, stromal cells, and immunocompetent cells, are important in tu mor metastasis. Co-culture experiments in vitro with mu ltip le types of cells are therefore necessary in hypoxia research. Cell cu lture inserts with a pore membrane are availab le to culture different cells in wells to examine cellular migration and cell-cell interactions [6,7]. However, only stepwise changes between normo xia and hypoxia, which do not mimic the physiological o xygen gradient, can be applied to cells.
Recent microfluid ic devices fo r controlling themicroenvi ronment have been used in bio medical research. Microfluid ic devices for creating a concentration gradient of a reagent and protein [8] and o xygen gradient [9,10] have been developed, and various microdevices for cell culture to facilitate microscopic observation of cellular behavior [11,12] and drug research [13,14] have been reported; some devices are available fo r cell cultures with an o xygen gradient [15,16]. However, the absence of a technique for co-cultures in microdevices to study interactions between cells of different types, has delayed the understanding of mechanis ms that relate to tumor metastasis under hypoxic conditions.
In this study, we aimed to develop a co -culture microdevice with an o xygen gradient. To validate our technique, we quantified the o xygen gradient with laser-assisted phosphorimetry and imaged endothelial cells M odel and M etastasis Imaging in the device for viability assay. Finally, we examined the effect of the oxygen gradient on tumor mig ration by performing co-culture experiments with endothelial cells. Figure 2. An optical setup for evaluating hypoxic conditions in the microdevice with phosphorimetry. A Nd:YAG laser through an objective lens was irradiated to the PDMS surface excit ing Pd-TCPP on it. The emitted phosphorescence depending on oxygen tension was detected with a photomultiplier via a long-pass filter and analyzed to calculate pO2

Fabricating the co-culture Microdevice
The co-culture microdevice, as shown in the cross section in Fig. 1(a), consisted of polydimethylsilo xane (PDMS) with microchannels for gas flo w and a membrane pore filter for trans migrat ion of tu mor cells, which were located 1 mm above the PDMS coated with collagen for endothelial cell culture. SU-8 (epo xy-based negative photoresist) was spin-coated uniformly on a substrate. A mold with microchannels was designed by photolithography, i.e., by irradiating UV light on the SU-8. Hardener-mixed PDMS was poured into the mo ld and fixed at 65℃ for 1 h. After PDMS was separated from the SU-8 mold, o xygen plas ma bonding with a glass substrate was performed. Fig. 1(b) shows the pattern and size of the microchannels, with two microchannels set in parallel and supplying o xygen, such that the central area between the channels became hypoxic. Stainless rings with pore memb rane[red area in Fig. 1(b)] were placed on the PDMS substrate for co-culture of tumor cells. The co-culture device was set in a gas-tight chamber made of acry lic p lates containing a mixture of 5% CO 2 and 95% N 2 , keeping the partial o xygen tension (pO 2 ) inside the device under 1 mmHg, as shown in Fig. 1(c). The pressure difference between the inlet and outlet fo r pure o xygen flow was measured by pressure sensors and controlled to change the oxygen gradient in the device. Figure 1. Schematic diagram of the co-culture microdevice with an oxygen gradient (a). Microfabricated PDMS with microchannels for oxygen supply was attached to a glass substrate through plasma bonding. T umor and endothelial cells were cultured on PDMS and a membrane filter with 3 -µ m pore size, respectively. Oxygen gas was supplied from the inlet to the outlet through microchannels, and membrane filters for observation of migrated cells were placed on PDMS (b). The microdevice was installed in a gas-tight chamber filled with a mixture of 5% CO2 and 95% N2 (c)

Measuring oxygen in the cell culture area
Oxygen supplied fro m microchannels was diffused to the cell cu lture surface through PDMS. To evaluate the oxygen gradient formed on PDMS, we used laser-assisted phosphorimetry installed on a fluorescence microscope (Fig.  2). pO 2 was measured by oxygen-dependent quenching of phosphorescence, as described previously [17]. In brief, Pd-meso-tetra-(4-carbo xyphenyl)-porphyrin (Pd-TCPP, Frontier Scientific, Inc.) was dissolved in phosphate buffer with 5% bovine serum albu min, and the mixture was spread on PDMS. The second harmonic of a Q-switched Nd :YA G pulse laser (532-n m wavelength, 6-ns pulse width at half maximu m, 1-Hz pulse recurrence frequency, 200-nJ/pulse irradiation energy) was irrad iated at the measurement point through the objective lens of a microscope. Lifet ime was obtained by least-squares fitting of the remain ing data to a single exponential curve. pO 2 was calculated using the Stern-Vo lmer equation, where I 0 and τ 0 are phosphorescence intensity and lifet ime in absence of o xygen, I and τ are phosphorescence intensity and lifetime at a given o xygen tension, and k q is the rate constant of oxygen quenching. The k q and τ 0 values for our system were 206 Torr −1 · s −1 and 0.53 ms at 37°C with a pH of 7.4.
The chamber was filled with nitrogen gas, and the inside pO 2 was maintained under 5 mmHg, creating a hypoxic condition in the device. Pure o xygen gas was supplied fro m the inlet of the device with pressure differences of 10, 25, 50, 100 mmHg between the inlet and outlet. In addition, pO 2 was measured at points every 1 mm on PDM S for 30 s at a pressure difference of 50 mmHg . A pseudocolor showing pO 2 was imaged using Mathematica 7 (Wolfram Research).

Cell Culture
Mono and co-cultures can be used in our device. Lewis lung carcino mas (LLC) cells, wh ich have a high metastatic ability, were used to visualize and quantify the d istribution of tumor cells during migration, which is dependent on the oxygen gradient. LLCs were grown in RPM I media with 10% FBS, penicillin, and streptomycin. Hu man umb ilical vein endothelial cells (HUVECs) were used for co-culture with LLCs, and were gro wn in EBM-2 med ia (Lonza Group Ltd.) with 10% FBS, penicillin, and streptomycin.
To quantify the dependence of cell viability on the oxygen gradient, HUVECs were cultured in a collagen gel, rather than directly on the PDM S surface, so that the cells would adhere to the device surface while alive and could be isolated fro m the device surface on death. HUVECs were suspended at a cell density of 8 × 10 5 cells/ ml in Cellmatrix (Kurabo Industries Ltd) on ice. Next, 500 µl of the suspension was spread on PDMS with a cell scraper, resulting in a final cell density of 4.4 × 10 4 cells/cm 2 . The device was incubated for 15 min for gelificat ion, and then, 2 ml of EBM-2 was added. The device was further incubated for 12 h under the o xygen gradient.

Cancer cell migration under the oxygen gradient
Tumor cell migrat ion through the pore membrane under the oxygen gradient, which imp lied metastasis, was quantified. A part of the pore membrane with the size of 3 µm pore was cut from a 24-well Fluoro Blo k (BD Falcon) and attached to a circular stainless ring (outside diameter: 16 mm, inside diameter: 6 mm). In the co-culture experiments, HUVECs were cultured on the PDMS surface coated with collagen at a density of 4.4 × 10 4 cells/cm 2 and LLC cells were seeded at a density of 33 × 10 4 cells/cm 2 , and then incubated separately for 6 h under normo xic conditions. The ring with the membrane was installed on PDMS, and the device was incubated for 12 h under an oxygen gradient formed by the oxygen flo w fro m microchannels. The culture med iu m for co-culture was a mixture of RPM I and EBM -2 in a ratio of 1:1.

Fl uorescence imaging and processing
Cells were fixed in 4% parafo rmaldehyde after staining with calcein -AM (Ex: 488 n m; Em: 515 n m) for living cells and propidiu m iodide (PI, Ex: 530 n m; Em: 615 n m) for dead cells. Fluorescent images were taken using a confocal microscope (TE 2000-U, Nikon), and the number of living and dead cells were counted offline using ImageJ software.
In the experiments on migration, cells that migrated through the pore membrane were fixed and stained with PI, and the average number o f cells was calculated fro m 10 images because the distribution of migrated cells on the membrane was non-uniform.

Statistical Analysis
Data shown are means ± SE. The statistical significance of the results was determined using Student's t-test. P < 0.05 was considered significant. Figure 3. The oxygen gradient in the microdevice, shown by the red line, was quantified when the pressure differences of the pure oxygen gas were 10, 50, 100 mmHg between the inlet and outlet (a). A two-dimensional pO2 distribution on the PDMS surface was imaged in pseudocolor, indicating oxygen diffusion from the microchannels supplying PDMS with oxygen for cell culture (b)

Oxygen gradient in the co-culture microdevice
The only oxygen supply to the co-culture microdevice was via the microchannels in PDMS, since the device was set in a chamber filled with nitrogen gas. Oxygen diffused in PDMS, forming an o xygen gradient with a maximu m pO 2 above the microchannels. The o xygen gradient could be controlled by changing the pressure difference of o xygen gas between the inlet and outlet. Figure 3(a) shows the measured o xygen gradient fro m a microchannel to the central part of the device at pressure differences of 10, 50, and 100 mmHg . pO 2 peaked immed iately above the microchannel and decreased exponentially 5 mm away fro m the channel, and then gently thereafter. The larger the difference in the pressure of the o xygen supply at the inlet and the outlet, the higher the inner pressure of the microchannel, resulting in a higher level of o xygen M odel and M etastasis Imaging diffusion at the inlet than at the outlet; this in turn resulted in a steeper oxygen gradient in the device. Two-dimensional pO 2 distribution on the PDMS surface at a pressure of 50 mmHg is shown in Fig. 3(b). Oxygen was diffused fro m the microchannels; the lowest pO 2 was at the central part of the device, while the upper part of the microchannels showed no difference between the inlet and outlet.
In the liv ing body, oxygen is delivered main ly by red blood cells and diffused in microcirculat ion, resulting in an oxygen gradient fro m microvessels to tissue. In addition, the oxygen gradient depends to a large extent on vessel structure, which is specific to each organ. A major advantage of the microdevice using micro fabrication is that channels can be designed to match an organ-specific vessel structure. For instance, the lobule structure typical of the liver can be designed, producing an organ-specific o xygen gradient.
Different types of theoretical models, for examp le, Krogh's cylindrical model, have been used to estimate oxygen delivery to tissues. However, theoretical calculations are no match for working with actual tissue; it is extremely difficult to take into account, in vivo, the consistency of the simulat ion and experiments. Cellular experiments in vitro play an intermediate role between animal and theoretical experiments. However, most experiments in vitro have been performed in standard incubators, indicating that even tumor cells are cultured in about 20% O 2 . Most experiments in vitro may thus be unrealistic. Most recently, there has been an increase in research on regenerative medicine with ES and iPS cells, and the general consensus is that stem cells should be cultured under hypoxic conditions because a hypoxic micro -environ ment is crit ical in stem cell b iology. Fo r this reason, a microdevice with an oxygen gradient close to that of physiological t issue holds great potential as a useful tool in various biological experiments.

Cellular vi ability under Oxygen Gradient
To validate the oxygen gradient formed in the co-culture microdevice, the viability of the cultured cells was quantified. A mosaic of fluorescent images of cultured HUVECs fro m the microchannel to 7-mm inside at 50 mmHg of pressure is shown in Fig. 4(a). The green fluorescence of calcein-AM fro m liv ing cells was dominant on the left side, while the red fluorescence of PI fro m the nuclei of dead cells increased with d istance fro m the channel, indicating that viability decreased under severe hypoxia. Magnified images located within 1 and 6 mm of the channel in Fig. 4(b) showed that green fluorescence fro m HUVECs near the microchannels indicated normal cellu lar shapes, but in the hypoxic area, red fluorescence increased, and even green fluorescence-emitting cells got rounder, indicating a possible decrease in activity. In Fig.  4(c), plotting the viability of HUVECs with distance from the microchannel, a dramatic decrease in viability was indicated at a distance of 5 mm, with a further significant decrease at 6 or 7 mm. The result of Fig. 3(b) indicates that superficial pO 2 on PDMS at 6 mm fro m the microchannel was about 30 mmHg. Ho wever, HUVECs were inside the collagen gel, and therefore, actual pO 2 around the cells was assumed to be lower than 30 mmHg. These experiments demonstrated that the oxygen gradient in the microdevice affects cellular activity and that severe hypoxia cause death of endothelial cells. Live and dead cells were stained with calcein-AM (green) and PI (red), respectively. A mosaic image from the microchannel to the center of the microdevice showed that viability decreased with distance and oxygen gradient (a, scale bar = 500 µ m). In a magnified image near the microchannel, the presence of live cells forming normal shapes was major. However, at 6 mm away from the channel, dead cells increased, and live cells with green fluorescence showed rounded shapes (b, scale bar = 100 µ m). Cell viability decreased in accordance with the oxygen gradient in the microdevice (c)

Metastasis Imaging of Cancer Cells Under An
Oxygen Gradient Figure 5. Tumor cell migration mimicking metastasis was observed in the microdevice. LLCs were cultured alone or with HUVECs under an oxygen gradient, and LLCs that migrated through the membrane pore filter were stained with PI (a). Cell migration was quantified using the distance from the microchannel; the hypoxic area with co-culture conditions accelerated the migration of tumor cells (b) We evaluated cell migration in a culture with endothelial cells under an o xygen gradient to simu late the interior of a tumor. Nuclear staining of LLCs that had migrated to the bottom o f the pore membrane was imaged with a confocal microscope, as shown in Fig. 5(a), and cell nu mber/unit area with distance fro m the microchannel was counted[ Fig. 5(b)]. The migrated cells in the monoculture showed minimal increase in the hypoxic area, and in contrast, the co-cultured LLCs with HUVECs increased their migrat ion with d istance, i.e., the degree of hypoxia, and fro m a point at a distance of 4 mm fro m the microchannel particularly, migrated cells increased notably.
A defin ing characteristic of the microdevice is that it facilitates the co-culture of tumor cells with other cells under an oxygen gradient. The interaction between tumor cells and endothelial cells, pericytes, or stromal cells is very important in vivo, and co-culture experiments have revealed various cellu lar mechanis ms. Fig. 5(b) clearly ind icates that not only hypoxic conditions but also the co-culture of tumor cells with HUVECs increased their migration. Th is confirmed the importance of the microdevice in research applications requiring co-culture under an o xygen gradient.
The device developed in this study has a role in the identification of gene exp ression in cells and cytokines associated with hypoxia by analyzing cultured cells and culture mediu m. For instance, HIF-1α is a type of "master switch" of the transcriptional response to hypoxia and is known to be associated with tumor metastasis. Previous studies have provided evidence that Snail and Twist are downstream targets of HIF-1α, and they repress E-cadherin expression in tumor cells that is a major co mponent of adherens junctions [18,19]. This leads to epithelial-mesenc hymal transition in tu mor cells, causing metastasis. In endothelial cells, cytokines activating cellu lar mot ility are produced in the hypoxic area. MMPs, CXCR4, HGF, and AMF are factors associated with tumor metastasis and invasion. Hypo xia stimulates CXCR4 expression in endothelial and tu mor cells [20], and CXCR4 overexpression increases tumor invasion [21]. Our results, as shown in Fig.  5(b), lead to the presumption that co-culture under an o xygen gradient produced a synergistic effect on tumor cell migrat ion through the expression of transcriptional factors and cytokines in LLCs and HUVECs. The addition of an oxygen gradient to co-culture experiments on tumor and endothelial cells has the potential to clarify the details of mo lecular mechanisms involved in tu mor metastasis. Moreover, changing the combination of tu mor cells with various other cells enables further discussion in tumor biology, and we can use the device for the evaluation of radiation effects under hypoxia by setting the chamber, including the microdevice, in a rad iotherapy unit.

Conclusions
We developed a co-culture microdevice with an oxygen gradient to visualize cell d istribution and quantify cell viability, and used the device to determine the effect of hypoxia on tumor metastasis. LLC cells co-cultured with endothelial cells showed gradient-dependent migration through a membrane pore filter, indicating that the interaction between tumor and endothelial cells under hypoxia is crucial in metastasis. The results suggest that the developed microdevice can be used to study the mechanisms of tumor metastasis under hypoxic conditions.