Microfluidic devices for size-dependent separation of liver cells
- 1.2k Downloads
Liver is composed of various kinds of cells, including hepatic parenchymal cells (hepatocytes) and nonparenchymal cells, and separation of these cells is essential for cellular therapies and pharmacological and metabolic studies. Here, we present microfluidic devices for purely hydrodynamic and size-dependent separation of liver cells, which utilize hydrodynamic filtration. By continuously introducing cell suspension into a microchannel with multiple side-branch channels, cells smaller than a specific size are removed from the mainstream, while large cells are focused onto a sidewall in the microchannel and then separated into two or three groups. Two types of PDMS-glass hybrid microdevices were fabricated, and rat liver cells were successfully separated. Also, cell size, morphology, viability and several cell functions were analyzed, and the separation performances of the microfluidic devices were compared to that of a conventional centrifugal technique. The results showed that the presented microfluidic devices are low-cost and suitable for clinical use, and capable of highly functional separation with relatively high-speed processing.
KeywordsLiver cell Hepatocyte Hydrodynamic filtration Microfluidic device Separation
Liver is a highly differentiated organ, and is composed of hepatic parenchymal cells, which are known as hepatocytes, and nonparenchymal cells, including sinusoidal endothelial cells, Kupffer cells (resident liver macrophages), hepatic stellate cells (myofibroblast-like cells), and pit cells (Morin et al. 1988). These cells play important roles in metabolism and various physiological functions, such as production of serum proteins, storage of nutrients, secretion of cytokines, and blood filtration and detoxification. Separation of liver cells, or selection of specific cells from liver, is therefore essential for the study of intact liver functions. For instance, isolated hepatocytes are often used for pharmacological and toxicological assessment (Kane et al. 2006) and cellular therapies such as cell transplantation (Ohashi et al. 2001). Also, separation of highly-proliferating small hepatocytes, which will be useful for liver reconstruction, has been reported by several groups (Mitaka et al. 1992; Tateno and Yoshizato 1996; Katayama et al. 2001).
For the purpose of separation of specific liver cells, various techniques have been hitherto employed. Low-speed centrifugation is a frequently used method to separate hepatocytes and nonparenchymal cells, utilizing the difference in sizes of these cells (Seglen 1976; Nilsson and Berg 1977; Tateno and Yoshizato 1996). A variety of centrifugal techniques have been also employed to prepare specific liver-cell fractions from the complex mixture. Density gradient centrifugation and iso-density Percoll centrifugation were employed (Singh et al. 1983; Knook et al. 1982; Kreamer et al. 1986), and centrifugal elutriation, a kind of continuous counter-stream centrifugation, has been adopted for liver cell separation based on cell size and/or density (Knook and Sleyster 1976; Asahina et al. 2006). Although fluorescence-activated cell sorting (FACS) systems are highly accurate and therefore widely used to select rare cells that express specific marker proteins (Sigal et al. 1995; Katayama et al. 2001), typical FACS systems are expensive and unsuitable for large-scale treatment. Furthermore, its architecture often prevents the disposable use of flow channels.
Recently, with the advance of microfabrication technologies, various chemical and biological systems have been realized in miniaturized devices (Auroux et al. 2002; Andersson and van den Berg 2003). In finely fabricated microfluidic channels, stable laminar flow profiles can be easily maintained, which is useful for highly effective and accurate cell manipulation (El-Ali et al. 2006). Among the various applications of microfluidic devices, cell separation therefore seems promising in biochemical and biomedical fields. Several researchers have reported on microfluidic systems for passive and continuous separation of biological particles, such as blood cells (Shevkoplyas et al. 2005; Takagi et al. 2005; Yamada and Seki 2005; Sethu et al. 2006), bacterial chromosome (Huang et al. 2004), plant cells (Yamada et al. 2004), and other kinds of animal cells (Nam et al. 2005; Murthy et al. 2006).
We have reported on microfluidic systems of hydrodynamic filtration for continuous and size-dependent particle separation (Yamada and Seki 2005, 2006). By introducing particle suspension into a microchannel with multiple side-branch channels, micrometer-size particles could be accurately and passively separated according to size, without the help of outer controls. Compared to other separation schemes, hydrodynamic filtration is advantageous in the high-speed processing without any need for outer controls. Depending on the microchannel deign, cells or particles could be easily separated into two or more groups. These properties of hydrodynamic filtration prompted us to separate complex liver cell mixture according to cell size.
In this study, we fabricated two types of microfluidic devices for continuous and size-dependent liver cell separation by hydrodynamic filtration. To investigate their abilities for cell separation, separated cells were analyzed by image processing and immunological staining. Finally, the hepatocyte separation efficiency of the microfluidic devices was compared with that of the conventional low-speed centrifugation method.
2 Materials and methods
To completely remove the small particles, and to focus the positions of large particles onto one sidewall, the introduction of the liquid flow that does not contain particles is essential. For this purpose, we previously proposed flow splitting in the upstream and recombining in the downstream, using multiple loop channels (Yamada and Seki 2006). This scheme is advantageous in the sense that multiple pumps are not required, although the microchannel structure becomes complex. Introduction of liquid flow without particles using another pump is an alternative, although the throughput is decreased compared to the flow splitting. In this study, these two schemes were employed.
2.2 Microdevice fabrication and design
PDMS-glass microfluidic devices were fabricated as described previously (Duffy et al. 1998). Briefly, PDMS prepolymer was poured onto an SU-8 master on a silicon wafer, and a PDMS plate having micro-grooves was obtained. Then inlet and outlet holes were punched, and after O2 plasma treatment, the PDMS plate was bonded with a flat glass slide to form the microchannel structure. Finally, silicone tubes with inner and outer diameters of 1 and 2 mm, respectively, were inserted into the inlet holes, and glued, to form the inlet connection.
In Microdevice 2 (Fig. 2(b)), on the other hand, split-flows from the main stream are re-injected through the 40 loop channels (20 μm in width), which act as the flow without cells in the case of Microdevice 1. Approximately 50% of the introduced liquid flow will be split into the 40 loop channels when the channel depth is 30 μm, and re-injected into the main channel. Then ∼80% of the flow will be removed through the 80 side channels, so it was expected that all the particles will be perfectly separated, unless the particles are small enough to go through the loop channels. Particles with diameter of 4∼6, 6∼10, and 10 μm∼ will be collected from Outlets 3, 2, and 1, respectively. In contrast, particles smaller than ∼4 μm were expected to enter into the loop channels, so these particles would be randomly dispersed and distributed through all the outlets.
2.3 Evaluation of the microfluidic devices
To examine whether the fabricated microdevices work as we designed, two experiments were conducted. First, flow rates distributed to each outlet were estimated, by introducing pure water and by measuring the output volumes. Second, standard particles were introduced. Green fluorescent polymer particles with diameter of 9.9 μm (G1000, Duke Scientific Corp., Palo Alto, CA, USA) were suspended in 0.5% (w/v) Tween 20 aqueous solution with a concentration of 4 × 105 particles/ml, and introduced into the microchannel using a syringe pump (KSD210, KD Scientific Inc., Holliston, MA, USA), and their behaviors were observed under a fluorescence microscope (ECLIPSE TE2000-U, Nikon Corp., Tokyo, Japan).
2.4 Cell isolation
Liver cells were isolated from male rats (Wistar or Lewis rat, 4 week age, 90–95 g weight) by using the two-step collagenase perfusion method (Seglen 1976), with minor modifications. Typically ∼108 cells were harvested from the liver with 80–95% viability, verified by trypan blue exclusion test.
2.5 Cell separation using microfluidic devices
Before using microfluidic devices for cell separation, they were placed in a vacuumed atmosphere for degassing in order to accelerate the removal of the inner air through PDMS substrate when liquid flow is introduced (Hosokawa et al. 2004). Liver cells were suspended at a concentration of 2∼3 × 106 cells/ml in a cell culture medium (EGM-2 BulletKit, Cambrex Bio Science Walkersville Inc., Walkersville, MD, USA). The cell suspension was filtered twice through 40 μm mesh cell strainer (BD Biosciences, Bedford, MA, USA), to get rid of large cell aggregates and other particulates. In Microdevice 1, the medium with and without cells were respectively introduced into the microchannel from Inlets 1 and 2, respectively, while cell suspension was simply introduced in Microdevice 2, using the syringe pump and plastic syringes with a volume of 1 ml (Terumo Corp., Tokyo, Japan), through Teflon tubing with inner and outer diameters of 0.5 and 1.5 mm. The flow rates for these suspensions were 50 μl/min for both microdevices. Silicone tubes were inserted into each outlet, and they were used for collection of separated cells. Experiments were repeated at least three times using different microfluidic devices, and obtained data were averaged.
2.6 Characterization of the separated cells
After cell separation, the same volumes of the recovered suspension and 1 mg/ml trypan blue solution in Dulbecco’s phosphate buffered saline (PBS) were mixed. Photographs of the stained cells from each inlet were taken with an inverted microscope (ECLIPSE TE300, Nikon Corp.) and a digital camera (FinePix S1Pro, Fujifilm Corp., Tokyo, Japan). Cell concentration and viability were estimated from the photographs, and the size distributions of the cells were analyzed using an image-processing software (Scion Image Beta 4.03, Scion Corp., Frederick, MD, USA), assuming that the cell shape is spherical. When two or more cells form an aggregate, these cells were individually analyzed. Also, the ratios of the single cells were obtained from the photographs.
In addition, the ratio of the albumin-producing cells, which are parenchymal cells (hepatocytes), was obtained by immunofluorescence staining (Tsuda et al. 2006). Cell nuclei were co-stained with Hoechst 33258 (Molecular Probes Inc., Eugene, OR, USA). The recovered cells from each outlet were plated onto a type I collagen-coated polystyrene Petri dish with diameter of 35 mm (Asahi Techno Glass Corp., Chiba, Japan). After 4 h of incubation at 37°C in a 5% CO2 incubator, cells were fixed with 4% (w/v) paraformaldehyde in PBS. After washing, cells were treated with 0.5% (w/v) Triton X in PBS for 2 min, followed by blocking using 0.1% (w/v) bovine serum albumin in PBS. Rabbit anti-rat albumin polyclonal antibody (Cappel, ICN Pharmaceuticals Inc., Aurora, OH, USA) was then added, and the dish was incubated at 4°C overnight. Finally, secondary antibody (FITC-conjugated goat IgG fraction to rabbit immunoglobulins, Cappel, ICN Pharmaceuticals Inc.) and Hoechst 33258 were added, and then washed with PBS. Fluorescent micrographs were captured with the fluorescence microscope.
To compare the separation efficiency of the presented microfluidic devices to that of the conventional method, low speed centrifugation was also performed to isolate hepatocytes. Cell suspension was loaded into a 15 ml plastic centrifuge tube (Corning Inc., Corning, NY, USA), and centrifuged for 1 min at 50 G. The supernatant and pellet were respectively collected, and the cells were examined.
3 Results and discussion
Comparison of the theoretical and measured flow rates distributed into each outlet
Microdevice 1 (Theoretical)
Microdevice 1 (Measured)
Microdevice 2 (Theoretical)
Microdevice 2 (Measured)
3.2 Separation of liver cells
The ideal cell/particle separation would not be achieved before the flow profile reaches a steady state. Therefore, when the cell suspension was introduced into a syringe connected with Teflon tubing, 0.8 ml of cell suspension was first introduced by aspiration, and then, 0.2 ml of the culture medium was introduced. This operation could prevent the cell introduction before the inner air was completely purged from the microchannel.
The photographs showing liver cell separation near the outlets are shown in Fig. 3(c) and (d). As can be seen from the figures, cells were continuously flowing though the microchannel. Since the flow speed was decreased near the outlets due to the broadened channel width, and since the lower surface of the microchannel was glass, cells were tend to attach to the lower surface near each outlet. However, the ratio of the attached cells was less than 0.01% of the introduced cells, thus the influence of the channel surface, or the interaction between cells and channel surface, could be neglected.
In the experiment, we introduced the cell suspensions into the microdevice with a flow rate of 50 μl/min. It took ∼20 min to process 1 ml of sample containing 2∼3 × 106 cells, and consequently 1.3∼2.0 × 103 cells were processed per second in both microdevices. The throughput of the presented devices is comparable to that of typical FACS systems, showing the ability of relatively high-speed separation. By optimizing the operating conditions, such as cell concentration and flow rate, the throughput will be increased up to ten-fold compared to the presented experiments, using the same devices. Also, other several strategies can be adopted, such as modification of the microchannel design, fabrication of higher-aspect-ratio structures, and parallel arrangement of multiple channels by stacking multiple PDMS layers.
3.3 Characterization of the separated cells
Characteristics of cells, separated using Microdevice 1 and 2, without/before separation (control), and separated using low-speed centrifugation
Cell numbers ± SE (%)
3.3 ± 0.8
26.0 ± 8.7
34.3 ± 9.0
36.4 ± 13.1
54.6 ± 15.5
9.0 ± 4.2
36.4 ± 13.1
44.5 ± 2.4
55.5 ± 2.4
Cell diameter ± SD (μm)a
20.2 ± 3.9
18.7 ± 4.7
7.7 ± 2.0
18.8 ± 5.3
9.2 ± 2.5
7.2 ± 1.8
14.8 ± 6.7
9.3 ± 4.5
18.7 ± 5.3
SE of cell diameters (μm)a
Cell viability ± SEb
107.0 ± 5.8
96.5 ± 4.5
96.6 ± 5.1
109.6 ± 4.5
93.0 ± 2.4
99.8 ± 1.5
109.8 ± 5.6
96.2 ± 8.4
94.9 ± 4.9
Single cells ± SE (%)c
6.9 ± 1.4
18.7 ± 2.2
80.6 ± 3.4
38.8 ± 3.2
51.1 ± 15.7
36.0 ± 2.2
Albumin-positive cells ± SE (%)
94.4 ± 1.8
96.0 ± 2.4
87.4 ± 2.5
4.0 ± 1.9
88.4 ± 2.1
5.1 ± 0.7
5.3 ± 2.9
64.3 ± 1.7
22.3 ± 6.4
83.1 ± 2.1
Binuclear cells ± SE (%)c
46.5 ± 4.0
55.5 ± 8.0
34.3 ± 5.1
47.5 ± 4.7
44.2 ± 1.7
43.7 ± 4.6
By applying the liver cell suspension into Microdevice 1, hepatocytes and nonparenchymal cells could be almost perfectly separated as shown in Figs. 4(a) and 5(b); hepatocytes were mainly recovered from Outlets 1, 2, and 3, while smaller cells were recovered from Outlet 4. A small number of cells smaller than 10 μm in diameter were recovered from Outlets 1, 2, and 3, although model particles with diameter of 9.9 μm did not go through these outlets. This was caused by the adhesion of nonparenchymal cells to hepatocytes, or by the formation of relatively large aggregates of nonparenchymal cells. Also, hepatocytes from Outlet 1 formed large aggregates, although their population was much smaller than that of cells from other outlets. The difference in the average cell diameters obtained from Outlets 2 and 3 was ∼1.5 μm, and this difference was significant by the statistical analysis using one-tailed Student’s t-test (p < 0.01). However, the difference was smaller than we expected, while the microchannel was designed so that spherical particles with diameters of 18∼30 and 10∼18 μm would be respectively recovered from Outlets 2 and 3. Non-spherical cells or particles would be separated based on their minor axis (Takagi et al. 2005), so the indefinite shapes of the cells would be one of the reasons for this discrepancy. Also, as can be seen from Fig. 4(a) and Table 2, it was observed that a large portion of the hepatocytes from Outlet 3 was single, while cells from Outlets 1 and 2 formed aggregates. When low-speed centrifugation was adopted to separate liver cells, cells could be separated only into two groups (hepatocytes and nonparenchymal cells), and single and aggregated hepatocytes could not be separated (Fig. 5(d) and Table 2).
On the other hand, Microdevice 2 was designed so that hepatocytes and nonparenchymal cells can be separated, and nonparenchymal cells can be further separated into two groups, without introducing the liquid flow without particles. From Figs. 4(b) and 5(c), it was confirmed that the hepatocytes and nonparenchymal cells could be separated with high accuracy, and these cells were respectively recovered from Outlet 1, and Outlets 2 and 3. However, the ratio of nonparenchymal cells in the cell fraction from Outlet 1 was higher than that in the cells recovered from Outlets 2 and 3 in Microdevice 1. This increase in small cell population might be attributed to the cells that flowed through the loop channels, despite the theoretical cell size that can enter the loop channels was ∼4 μm. There was a difference in average sizes of the cells recovered from Outlets 2 and 3; the average cell diameters were 9.2 and 7.2 μm, respectively, although the two peaks partially overlapped.
In both microdevices, the cell viabilities before and after separation were on nearly the same level. The slight increases in viabilities of cells from Outlet 4 in Microdevice 1, and Outlet 3 in Microdevice 2, and the slight decrease in that of cells from Outlets 2 and 3 in Microdevice 1, and Outlet 1 in Microdevice 2, would be due to the relatively high viability of small cells (nonparenchymal cells) compared to large cells (hepatocytes). So the damages to cells by the shear stress using these microfluidic devices were negligible.
It is known that mononuclear and binuclear hepatocytes exist in liver, and the ratio of the binuclear cells increases with age, which is related to the growth potential of hepatocytes (Sigal et al. 1995; Mossin et al. 1994). So the ratio of the binuclear cells among hepatocytes was evaluated from the captured micrographs. As a result, there was a slight difference in the ratios of the binuclear hepatocytes, only when Microdevice 1 was used for cell separation. The ratio of binuclear hepatocytes from Outlet 2 was higher than that of hepatocytes from Outlet 3 (Table 2), which would be due to the tendency that mononuclear hepatocytes are smaller in size. Since the separation scheme of the presented microdevices is simply based on cell size, complete separation of multinuclear and mononuclear cells would not be achieved. However, the potential for the highly-functional separation of liver cells was successfully shown.
In the current study, we demonstrated size-dependent liver cell separation using microfluidic devices, utilizing the principle of hydrodynamic filtration. The results showed that the finely-designed microdevices are suitable for the separation of liver cells, since the simple introduction of cell suspension enables accurate and relatively high-speed cell separation. It is expected that the presented microfluidic systems can be used for clinical settings and biochemical research fields.
This research was supported in part by the Core Research for Evolution Science and Technology from the Japan Science and Technology Agency, and by Grant-in-aid for JSPS Fellowship. We are grateful to Ms. Chinatsu Kohno for her technical assistance.
- S. Katayama, C. Tateno, T. Asahara, K. Yoshizato, Am. J. Pathol. 158, 97 (2001)Google Scholar
- O. Morin, F. Goulet, C. Normand, in Cell Biology Reviews (Springer International, New York, 1988), pp. 1–69Google Scholar
- M. Nilsson, T. Berg, Biochim. Biophys. Acta 497, 171 (1977)Google Scholar
- C. Tateno, K. Yoshizato, Am. J. Pathol. 149, 1593 (1996)Google Scholar