**3.2 MEMS-based microdevice design and fabrication**

T-type mixing model and sandwich type mixing model were proposed. In the sandwich type model, line-type model and coil-type model of the mixing microchannels were designed. All the models (see Fig.2) were numerically simulated. The best model with the best mixing performance was used to construct the silicon substrate. The mixing condition of the best model was optimized by numerical simulation, and then it was verified by cell lysis experiments.

Fig. 1. Sequence for fabrication of the microdevice.

Triton-X 100, HNO3, KOH, ethanol, Tris, EDTA, and HCl were purchased from Beijing Chemical Reagents Company (Beijing China). SYBR Green I dye was purchased from Molecular Probes (Leiden, The Netherlands). Guanidine thiocyanate, λ-DNA, nucleic acid extraction kit, and PCR kit were purchased from Tianwei (Beijing China). Primers for the 203-bp' -gapd gene and primers for the 250-bp, -action gene were purchased from Sangon

Diluted buffer (0.9% NaCl in PBS) was used to dilute whole blood. TE buffer (10mM Tris, 1 mM EDTA, titrated to pH 8), load buffer (4M GuSCN in TE buffer with 1% Triton-X 100, titrated to pH 6.7), wash buffer (70% ethanol with 10mM NaCl) and eluted buffer (namely TE buffer) were used for the DNA purification procedure. All solutions were prepared in

T-type mixing model and sandwich type mixing model were proposed. In the sandwich type model, line-type model and coil-type model of the mixing microchannels were designed. All the models (see Fig.2) were numerically simulated. The best model with the best mixing performance was used to construct the silicon substrate. The mixing condition of the best model was optimized by numerical simulation, and then it was verified by cell

**3.2 MEMS-based microdevice design and fabrication** 

Fig. 1. Sequence for fabrication of the microdevice.

**3. Experimental** 

(Shanghai China).

distilled water.

lysis experiments.

**3.1 Reagents** 

Fig. 2. Numerical simulation results for the species concentration distributions at T-type mixing model (a), sandwich type mixing model with lined channel (b) and sandwich type mixing model with coiled channel (c) when νcell = 0.005m/s, νbuffer = 0.0025m/s. The color for "1.00e+00" and "0.00 e+00" in this figure refers to the concentration intensity of the pure cell sample and the pure lysis buffer, respectively. The software of FLUENT 6.2 was used to simulate.

The microdevice consisted of a silicon substrate and a glass cover. The fabrication process was shown in Fig. 1. A double-side polished n-type 0.01-0.1Ωcm silicon wafer of (100) crystal orientation was first deposited with 0.3μm thick of silicon nitride (Si3N4) as the mask for fabricating porous silicon by a chemical vapour deposition (CVD) method, and then with 0.3μm aluminum as the mask for deep reaction ion etching (DRIE) by a electron beam vapour deposition technology. The wafer was spin coated with a positive photoresist (AZ1500) and patterned. After the exposed photoresist was developed, the exposed aluminum was removed by phosphoric acid and then the exposed silicon nitride was removed by plasma etching. The wafer was etched in a deep reaction ion etcher (Adixen, AMS100) to produce rectangle channels of 100μm in depth. After the process of removing the remained aluminum, a porous silicon layer on the internal walls of channels was anodized in 10% HF electrolyte at 20mA/cm2 for 15min. After that, the silicon wafers were thoroughly rinsed in distilled water and oxidized at 1050°C for 1h. After oxidation process, the silicon nitride formerly patterned as the mask was removed by plasma etching for bonding. Finally glass covers (Corning Pyrex#7740), in which three holes had been drilled corresponding to the silicon substrate, were anodically bonded to the silicon wafers to form the closed channel by a bonder (Suss, SB6), thus microfluidic devices are fabricated.

#### **3.3 Online cell lysis and DNA extraction procedure**

The microdevice was mounted onto the stage of a microscope with a CCD camera and a video monitoring system. Blood was introduced through the cell inlet by a peristaltic pump, while

MEMS-Based Microdevice for Cell Lysis and DNA Extraction 29

where C is the species concentration profile across the width of the microchannel (h), c∞ is the completely mixed state (= 0.5) and C0 is the completely unmixed state (= 0 or 1). Note that the confluent streams are completely mixed if σ = 100%. In contrast, they are completely

Fig. 3. Numerical evaluations of mixing efficiency index (σ) at different mixing models when

Fig. 3 demonstrates the mixing efficiency of the proposed mixing models: T-type, sandwichtype with lined channel and sandwich-type with coiled channel. The use of coiled channel is effective in enhancing chaotic mixing, in which the mixing efficiency of 89.3% is obtained at the cross section located 4mm from inlet, by folding, stretching and reorienting fluid. Based on discussion above, the coiled channel of sandwich-type mixing model is chosen to

For cell lysis based on chemical disruption method, there are two steps which are mixing and chemical reaction respectively. In the microdevice, mixing is slow compared with the chemical reaction rate due to the fact that mixing in the microdevice is exclusively caused by diffusion because of laminar flow. The total lysis of cells requires that each cell should be fully exposed to lysis buffer, which is limited by the mixing speed. Thus the cell lysis efficiency is strongly depended on the mixing performance between the cell solution and the lysis buffer. From above results of simulation, a microdevice utilizing sandwich-type mixing model with coiled channel was designed and fabricated. As is shown in Fig. 4, the microdevice consists of a glass cover with two inlets and one outlet and a silicon substrate

The cell lysis performance is not only influenced by the design of the microchannels but also impacted by the velocity of the two solutions. With the increase of velocity, the time for cell lysis is reduced while the cell lysis efficiency decreases. In order to reduce the time for cell lysis and obtain high cell lysis efficiency, the velocity should be carefully examined. At first, the relationship between the mixing efficiency index and the inlet velocity was simulated

with an etched coiled channel of 200µm wide, 100µm deep and 20.16cm long.

using the coiled channel of sandwich-type mixing model.

unmixed if σ = 0%.

νcell = 0.005m/s, νbuffer = 0.0025m/s

construct the physical microdevice.

**4.2 Optimization of cell lysis process** 

lysis buffer (namely, load buffer) was pumped through the buffer inlet by another peristaltic pump. The value of the flow velocity was changed from 0.1μL/min to 25μL/min. When the lysis buffer and blood were blended in the channel during the continuous flow, cells were gradually lyzed online, and then genomic DNA was released and absorbed on the porous silicon matrix. DNA extraction procedure itself consisted of load, wash, and elution steps. In the load step, the released DNA was absorbed onto the porous silicon matrix in the presence of low pH (pH 6.7) and high concentration binding salt (4M GuSCN). And then proteins and possible PCR inhibitors were removed by passing 70% ethanol twice through the microdevice. Finally, DNA was eluted in TE buffer. 5 µL TE buffer was introduced in the microdevice for 10 min at 55°C and then 20 µL TE buffer was continuously passed through the microdevice.

#### **3.4 Fluorescence detection and PCR amplification**

DNA collected in the elution step was quantified by using SYBR Green I dye in a fluorometer using calibration curve of DNA which were generated using lambda DNA. Then DNA purified from whole blood was amplified by polymerase chain reaction (PCR). The 203-bp -gapd gene of rat's blood was amplified using the following primers: Forward Primer: 5′-AGAAGTACCTGCAACAGG- 3′, Reverse Primer: 5′- GACGGACACATTGGGGGT- 3′. PCR reactions consisted of 2.5 µL standard 10×PCR buffer, 100 μmol dATP, dGTP, dCTP, and dTTP, 2.5 units of Taq polymerase, 50nmol of each primer, 5µL of the initial collected fraction, in a total volume of 25µL. These reactions were cycled in under the following conditions: 95°C denaturation for 5min, 35 cycles of 94°C for 1min, 68°C for 1min, 72°C for 1min, followed by a 10min extension at 72°C. DNA amplification was confirmed by gel electrophoresis.

#### **4. Results and discussion**

#### **4.1 Design and simulation of mixing process**

In general, comprehensive computational fluid dynamics (CFD) simulations are conducted before physical models are built and tested since these simulations enable the system parameters to be varied over a wide range of values and permit the simultaneous and instantaneous data collection of various aspects of models. Numerical simulation plays a key role in optimizing designs of microdevices and enables a reliable interpretation of the experimental results. T-type mixing model, sandwich-type mixing model with lined channel and coiled channel were designed and numerical simulated using FLUENT 6.2 software and all the simulations were two-dimensional.

The numerical simulation results are shown in Fig. 2. The mixing performance in sandwichtype mixing model is better than that in T-type one because of the increase of the contact area, and the mixing performance in coiled channel is much better than that in lined channel. Those numerical simulation results mentioned above provide a very clear understanding of the physical phenomena taking place in the two-dimensional microfluidic channels. In order to get a more accurate evaluation of the degree of mixing in the microchannel, a mixing index (σ) is used as follows (Erickson & Li, 2002):

$$\sigma = \left( 1 - \frac{\int\_0^h |c - c\_\infty| dy}{\int\_0^h |c\_0 - c\_\infty| dy} \right) \times 100\,\% \tag{4}$$

lysis buffer (namely, load buffer) was pumped through the buffer inlet by another peristaltic pump. The value of the flow velocity was changed from 0.1μL/min to 25μL/min. When the lysis buffer and blood were blended in the channel during the continuous flow, cells were gradually lyzed online, and then genomic DNA was released and absorbed on the porous silicon matrix. DNA extraction procedure itself consisted of load, wash, and elution steps. In the load step, the released DNA was absorbed onto the porous silicon matrix in the presence of low pH (pH 6.7) and high concentration binding salt (4M GuSCN). And then proteins and possible PCR inhibitors were removed by passing 70% ethanol twice through the microdevice. Finally, DNA was eluted in TE buffer. 5 µL TE buffer was introduced in the microdevice for 10 min at 55°C and then 20 µL TE buffer was continuously passed through the microdevice.

DNA collected in the elution step was quantified by using SYBR Green I dye in a fluorometer using calibration curve of DNA which were generated using lambda DNA. Then DNA purified from whole blood was amplified by polymerase chain reaction (PCR). The 203-bp -gapd gene of rat's blood was amplified using the following primers: Forward Primer: 5′-AGAAGTACCTGCAACAGG- 3′, Reverse Primer: 5′- GACGGACACATTGGGGGT- 3′. PCR reactions consisted of 2.5 µL standard 10×PCR buffer, 100 μmol dATP, dGTP, dCTP, and dTTP, 2.5 units of Taq polymerase, 50nmol of each primer, 5µL of the initial collected fraction, in a total volume of 25µL. These reactions were cycled in under the following conditions: 95°C denaturation for 5min, 35 cycles of 94°C for 1min, 68°C for 1min, 72°C for 1min, followed by a 10min extension at 72°C. DNA

In general, comprehensive computational fluid dynamics (CFD) simulations are conducted before physical models are built and tested since these simulations enable the system parameters to be varied over a wide range of values and permit the simultaneous and instantaneous data collection of various aspects of models. Numerical simulation plays a key role in optimizing designs of microdevices and enables a reliable interpretation of the experimental results. T-type mixing model, sandwich-type mixing model with lined channel and coiled channel were designed and numerical simulated using FLUENT 6.2 software and

The numerical simulation results are shown in Fig. 2. The mixing performance in sandwichtype mixing model is better than that in T-type one because of the increase of the contact area, and the mixing performance in coiled channel is much better than that in lined channel. Those numerical simulation results mentioned above provide a very clear understanding of the physical phenomena taking place in the two-dimensional microfluidic channels. In order to get a more accurate evaluation of the degree of mixing in the

microchannel, a mixing index (σ) is used as follows (Erickson & Li, 2002):

*h*

*c c dy %*

1 100 (4)

*c c dy*

*h*

 0 <sup>0</sup> <sup>0</sup>

**3.4 Fluorescence detection and PCR amplification** 

amplification was confirmed by gel electrophoresis.

**4.1 Design and simulation of mixing process** 

all the simulations were two-dimensional.

**4. Results and discussion** 

where C is the species concentration profile across the width of the microchannel (h), c∞ is the completely mixed state (= 0.5) and C0 is the completely unmixed state (= 0 or 1). Note that the confluent streams are completely mixed if σ = 100%. In contrast, they are completely unmixed if σ = 0%.

Fig. 3. Numerical evaluations of mixing efficiency index (σ) at different mixing models when νcell = 0.005m/s, νbuffer = 0.0025m/s

Fig. 3 demonstrates the mixing efficiency of the proposed mixing models: T-type, sandwichtype with lined channel and sandwich-type with coiled channel. The use of coiled channel is effective in enhancing chaotic mixing, in which the mixing efficiency of 89.3% is obtained at the cross section located 4mm from inlet, by folding, stretching and reorienting fluid. Based on discussion above, the coiled channel of sandwich-type mixing model is chosen to construct the physical microdevice.

#### **4.2 Optimization of cell lysis process**

For cell lysis based on chemical disruption method, there are two steps which are mixing and chemical reaction respectively. In the microdevice, mixing is slow compared with the chemical reaction rate due to the fact that mixing in the microdevice is exclusively caused by diffusion because of laminar flow. The total lysis of cells requires that each cell should be fully exposed to lysis buffer, which is limited by the mixing speed. Thus the cell lysis efficiency is strongly depended on the mixing performance between the cell solution and the lysis buffer. From above results of simulation, a microdevice utilizing sandwich-type mixing model with coiled channel was designed and fabricated. As is shown in Fig. 4, the microdevice consists of a glass cover with two inlets and one outlet and a silicon substrate with an etched coiled channel of 200µm wide, 100µm deep and 20.16cm long.

The cell lysis performance is not only influenced by the design of the microchannels but also impacted by the velocity of the two solutions. With the increase of velocity, the time for cell lysis is reduced while the cell lysis efficiency decreases. In order to reduce the time for cell lysis and obtain high cell lysis efficiency, the velocity should be carefully examined. At first, the relationship between the mixing efficiency index and the inlet velocity was simulated using the coiled channel of sandwich-type mixing model.

MEMS-Based Microdevice for Cell Lysis and DNA Extraction 31

In order to test the above results of simulation, the experiments of cell lysis were conducted by using the microfluidic device (Fig.4c). The cell sample was whole blood, while the lysis buffer was load buffer, 4M GuSCN in TE buffer with 1% Triton-X 100, titrated to pH 6.7. Triton X-100 is generic surfactants, with amphoteric chemical properties, which can react with water and can also react with lipid. High concentration guanidine salt is typical bonding salt for solid phase extraction nucleic acid and is also a generic strong denaturant. Moreover guanidine salt can quickly dissolve protein and can destroy the structure of cells. So the mixture of the traditional surfactant and the typical bonding salt was used as the lysis

Fig. 6. Experimental evaluations of time for full cell lysis at variation of flow velocity ratio of

At first, the effect of flow velocity on the time taken for cell lysis was analyzed when the velocity ratio of cell/buffer was 1:1. For the flow velocity of 0.1, 0.2, 0.4 and 0.5μL/min, the time taken for complete lysis was tested to be 188s, 188s, 175s and 194s, at the location which was 3.92cm, 7.84cm, 14.56cm and 20.16cm from the inlet along the channel respectively. Note that the flow velocity of 1μL/min is equivalent to 0.833mm/s in the microchannel of 200μm wide and 100μm deep. The time taken for completely lysis at different flow rate with the same cell/buffer ratio was approximately the same; however cells could not be fully lyzed in the microdevice at more than 0.5μL/min. Then, the velocity ratio of cell/buffer was varied. As is shown in Fig. 6, the time required for full lysis is sharply reduced from approximate two minutes to several seconds with the decrease of the cell/buffer velocity ratio from 0.7 to 0.1. The photograph pictures of Fig. 7 show that the width of the cell stream is reduced with the decrease of the cell/buffer velocity ratio and is not affected by specific rate value. These experimental results are found to be consistent with the numerical results. Thus rapid cell lysis can be implemented under the condition of lower cell/buffer velocity ratio and higher specific rate value. In the end the cell velocity of 5μL/min and the lysis buffer velocity of

buffer for online cell lysis and DNA extraction.

cell/buffer when νbuffer = 1μL/min.

15~25μL/min were used for rapidly cell lysis.

Fig. 4. Schematic top view (a), cross view (b) and photograph (c) of the microdevice. (1) Cell inlet; (2) Buffer inlet; (3) Outlet.

Fig. 5. Numerical evaluations of mixing efficiency index (σ) of coiled channel of sandwichtype mixing model. (Left) With different inlet velocities: "1" νcell = 0.005m/s, νbuffer = 0.0025m/s; "2" νcell = 0.02m/s, νbuffer = 0.01m/s; "3" νcell = 0.2m/s, νbuffer = 0.1m/s. (Right) At the location near the inlet with different velocity ratio: "low velocity" νbuffer = 0.01m/s; "high velocity" νbuffer = 0.1m/s.

As is shown in Fig. 5A, the mixing efficiency is sharply decreased with the increase of the inlet velocity. This decrease of mixing efficiency can likely be attributed to the insufficient diffusion for the limited diffusion time. For the laminar flowing, the mixing time is proportional to the square of diffusion length so that decreasing diffusion length would reduce the mixing time sharply without deterioration of the mixing efficiency. The diffusion length can be changed by varying the cell and buffer velocity ratio. The lower cell/buffer velocity ratio leads to shorter diffusion length which can enhance mixing efficiency. Moreover at the location near the junction of three streams, the distribution of the laminar flow of the horizontal stream caused by the vertical streams enhances fluid mixing. And this enhancement is much stronger at the lower cell/ buffer velocity ratio. When the cell/buffer velocity ratio is 0.2, the mixing efficiency index is found to be more than 80% near the junction, which is insensitive to the value change of the flow velocity, shown in Fig. 5B.

Fig. 4. Schematic top view (a), cross view (b) and photograph (c) of the microdevice.

(A) (B) Fig. 5. Numerical evaluations of mixing efficiency index (σ) of coiled channel of sandwichtype mixing model. (Left) With different inlet velocities: "1" νcell = 0.005m/s, νbuffer = 0.0025m/s; "2" νcell = 0.02m/s, νbuffer = 0.01m/s; "3" νcell = 0.2m/s, νbuffer = 0.1m/s. (Right) At the location near the inlet with different velocity ratio: "low velocity" νbuffer = 0.01m/s;

As is shown in Fig. 5A, the mixing efficiency is sharply decreased with the increase of the inlet velocity. This decrease of mixing efficiency can likely be attributed to the insufficient diffusion for the limited diffusion time. For the laminar flowing, the mixing time is proportional to the square of diffusion length so that decreasing diffusion length would reduce the mixing time sharply without deterioration of the mixing efficiency. The diffusion length can be changed by varying the cell and buffer velocity ratio. The lower cell/buffer velocity ratio leads to shorter diffusion length which can enhance mixing efficiency. Moreover at the location near the junction of three streams, the distribution of the laminar flow of the horizontal stream caused by the vertical streams enhances fluid mixing. And this enhancement is much stronger at the lower cell/ buffer velocity ratio. When the cell/buffer velocity ratio is 0.2, the mixing efficiency index is found to be more than 80% near the junction, which is insensitive to the value change of the flow velocity, shown in Fig. 5B.

(1) Cell inlet; (2) Buffer inlet; (3) Outlet.

"high velocity" νbuffer = 0.1m/s.

In order to test the above results of simulation, the experiments of cell lysis were conducted by using the microfluidic device (Fig.4c). The cell sample was whole blood, while the lysis buffer was load buffer, 4M GuSCN in TE buffer with 1% Triton-X 100, titrated to pH 6.7. Triton X-100 is generic surfactants, with amphoteric chemical properties, which can react with water and can also react with lipid. High concentration guanidine salt is typical bonding salt for solid phase extraction nucleic acid and is also a generic strong denaturant. Moreover guanidine salt can quickly dissolve protein and can destroy the structure of cells. So the mixture of the traditional surfactant and the typical bonding salt was used as the lysis buffer for online cell lysis and DNA extraction.

Fig. 6. Experimental evaluations of time for full cell lysis at variation of flow velocity ratio of cell/buffer when νbuffer = 1μL/min.

At first, the effect of flow velocity on the time taken for cell lysis was analyzed when the velocity ratio of cell/buffer was 1:1. For the flow velocity of 0.1, 0.2, 0.4 and 0.5μL/min, the time taken for complete lysis was tested to be 188s, 188s, 175s and 194s, at the location which was 3.92cm, 7.84cm, 14.56cm and 20.16cm from the inlet along the channel respectively. Note that the flow velocity of 1μL/min is equivalent to 0.833mm/s in the microchannel of 200μm wide and 100μm deep. The time taken for completely lysis at different flow rate with the same cell/buffer ratio was approximately the same; however cells could not be fully lyzed in the microdevice at more than 0.5μL/min. Then, the velocity ratio of cell/buffer was varied. As is shown in Fig. 6, the time required for full lysis is sharply reduced from approximate two minutes to several seconds with the decrease of the cell/buffer velocity ratio from 0.7 to 0.1. The photograph pictures of Fig. 7 show that the width of the cell stream is reduced with the decrease of the cell/buffer velocity ratio and is not affected by specific rate value. These experimental results are found to be consistent with the numerical results. Thus rapid cell lysis can be implemented under the condition of lower cell/buffer velocity ratio and higher specific rate value. In the end the cell velocity of 5μL/min and the lysis buffer velocity of 15~25μL/min were used for rapidly cell lysis.

MEMS-Based Microdevice for Cell Lysis and DNA Extraction 33

(A) (B)

The microdevices with and without porous silicon were used to recover DNA from 200ng prepurified genomic DNA under the same experimental conditions, respectively. The performance of the microdevice with porous silicon is quite well, with an average of 83% (11.6%RSD) evaluated from five extractions, which is much higher than that of one without porous silicon with an average of 39.2% (8.7%RSD). It is important to note that 200ng DNA overloads the capacity of the microdevice without porous silicon due to the limitation of surface area. Therefore the binding capacity of the microdevice without porous silicon is approximately 75ng/cm2 which agrees with the results of Cady et al. (Cady et al., 2003) who found that the binding capacity of nonporous micropillars was approximate 82ng/cm2. The previous researches proved that the performance of DNA extraction microdevice was determined by the surface area of the matrix and the extracted DNA was found to increase linearly with the surface area (Fan et al., 1999; Kwakye et al., 2006). The binding capacity of porous silicon matrix would increase hundreds or thousands of times, while the extraction efficiency didn't improve so much. The reason could probably be that most of the internal pores and smaller pores might not be used to adsorb DNA, and the DNA adsorbed in these

Comparison with the microdevice without the porous silicon layer, the microdevice with the porous layer achieved higher extracting efficiency of DNA without problems of clogging and high backpressure. Moreover this SPE microdevice with porous silicon can be easily integrated with other microdevices, since the porous silicon can be directly generated on the

A microfluidic device capable of real time destroying cells and extracting DNA from the cell lysates during continuous flow was designed and fabricated. As is shown in Fig. 4, the microfluidic device with the size of 2cm×1.2cm consists of a silicon substrate within an etched coiled channel, and a glass cover with two inlets and one outlet according to the silicon substrate. The two inlets: cell inlet and buffer inlet are designed to introduce blood sample and lysis buffer simultaneously, leading to rapid lysis of blood cells near the location of cell inlet using the sandwich-type mixing model. Porous silicon layer fabricated on the

Fig. 8. SEM of porous silicon on the surface of the channel for DNA purification.

(A) ×40; (B) ×1000.

pores might not be easily eluted.

internal walls of channel in the microdevice.

**4.4 Cell lysis and purification of DNA on microdevice** 

Fig. 7. Experimental results for the cell and buffer concentration distributions at location near the inlet with different velocity: "a" νcell : νbuffer = 2:1, νbuffer = 1μL/min; "b" νcell : νbuffer = 1:1, νbuffer = 1μL/min; "c" νcell : νbuffer = 1:3, νbuffer = 1μL/min; "d" νcell : νbuffer = 1:5, νbuffer = 1μL/min; "e" νcell : νbuffer = 2:1, νbuffer = 10μL/min; "f" νcell : νbuffer = 1:1, νbuffer = 10μL/min; "g" νcell : νbuffer = 1:3, νbuffer = 10μL/min; "h" νcell : νbuffer = 1:5, νbuffer = 10μL/min.

#### **4.3 DNA purification using porous silicon matrix**

DNA purification on-microdevice is based on SPE method: firstly adsorption of DNA onto a solid-phase matrix surface and then washing the impurities such as protein and finally elution of the purified DNA. The mechanism of DNA adsorption on a solid-phase matrix surface elucidated by Melzak et al. (Melzak et al., 1996) was that the adsorption of highly charged duplex DNA to negatively charged silica was controlled by three competing effects: weak electrostatic repulsion forces, dehydration and hydrogen bond formation. The surface area of the matrix significantly affects the adsorption of DNA onto the solid-phase matrix surface.

Porous silicon gotten on the surface of the tortuous rectangle channel in the microdevice by electrochemical etching technology can enhance the available surface area (see Fig. 8). The details of the preparation and optimization process for porous silicon were reported in our previous research (Chen et al., 2006; Chen et al., 2007). By using BET nitrogen adsorption experiment, pore size of the porous silicon is in the range of 20nm to 30nm. And the surface area of porous silicon is approximate 400m2/g. Note that the volume of microfluidic device with a tortuous channel of 25cm long, 100µm deep and 200µm wide is 0.005cm3. Thus the surface area to volume of microdevice with porous silicon is approximate 300m2/cm3 which is thousands of times higher than that of one without porous silicon since it is known that the surface area to volume of the microdevice without porous silicon is approximate 200cm2/cm3.

Fig. 7. Experimental results for the cell and buffer concentration distributions at location near the inlet with different velocity: "a" νcell : νbuffer = 2:1, νbuffer = 1μL/min; "b" νcell : νbuffer = 1:1, νbuffer = 1μL/min; "c" νcell : νbuffer = 1:3, νbuffer = 1μL/min; "d" νcell : νbuffer = 1:5, νbuffer = 1μL/min; "e" νcell : νbuffer = 2:1, νbuffer = 10μL/min; "f" νcell : νbuffer = 1:1, νbuffer = 10μL/min;

DNA purification on-microdevice is based on SPE method: firstly adsorption of DNA onto a solid-phase matrix surface and then washing the impurities such as protein and finally elution of the purified DNA. The mechanism of DNA adsorption on a solid-phase matrix surface elucidated by Melzak et al. (Melzak et al., 1996) was that the adsorption of highly charged duplex DNA to negatively charged silica was controlled by three competing effects: weak electrostatic repulsion forces, dehydration and hydrogen bond formation. The surface area of the matrix significantly affects the adsorption of DNA onto the solid-phase matrix surface. Porous silicon gotten on the surface of the tortuous rectangle channel in the microdevice by electrochemical etching technology can enhance the available surface area (see Fig. 8). The details of the preparation and optimization process for porous silicon were reported in our previous research (Chen et al., 2006; Chen et al., 2007). By using BET nitrogen adsorption experiment, pore size of the porous silicon is in the range of 20nm to 30nm. And the surface area of porous silicon is approximate 400m2/g. Note that the volume of microfluidic device with a tortuous channel of 25cm long, 100µm deep and 200µm wide is 0.005cm3. Thus the surface area to volume of microdevice with porous silicon is approximate 300m2/cm3 which is thousands of times higher than that of one without porous silicon since it is known that the surface area to volume of the microdevice without porous silicon is approximate

"g" νcell : νbuffer = 1:3, νbuffer = 10μL/min; "h" νcell : νbuffer = 1:5, νbuffer = 10μL/min.

**4.3 DNA purification using porous silicon matrix** 

200cm2/cm3.

Fig. 8. SEM of porous silicon on the surface of the channel for DNA purification. (A) ×40; (B) ×1000.

The microdevices with and without porous silicon were used to recover DNA from 200ng prepurified genomic DNA under the same experimental conditions, respectively. The performance of the microdevice with porous silicon is quite well, with an average of 83% (11.6%RSD) evaluated from five extractions, which is much higher than that of one without porous silicon with an average of 39.2% (8.7%RSD). It is important to note that 200ng DNA overloads the capacity of the microdevice without porous silicon due to the limitation of surface area. Therefore the binding capacity of the microdevice without porous silicon is approximately 75ng/cm2 which agrees with the results of Cady et al. (Cady et al., 2003) who found that the binding capacity of nonporous micropillars was approximate 82ng/cm2. The previous researches proved that the performance of DNA extraction microdevice was determined by the surface area of the matrix and the extracted DNA was found to increase linearly with the surface area (Fan et al., 1999; Kwakye et al., 2006). The binding capacity of porous silicon matrix would increase hundreds or thousands of times, while the extraction efficiency didn't improve so much. The reason could probably be that most of the internal pores and smaller pores might not be used to adsorb DNA, and the DNA adsorbed in these pores might not be easily eluted.

Comparison with the microdevice without the porous silicon layer, the microdevice with the porous layer achieved higher extracting efficiency of DNA without problems of clogging and high backpressure. Moreover this SPE microdevice with porous silicon can be easily integrated with other microdevices, since the porous silicon can be directly generated on the internal walls of channel in the microdevice.

#### **4.4 Cell lysis and purification of DNA on microdevice**

A microfluidic device capable of real time destroying cells and extracting DNA from the cell lysates during continuous flow was designed and fabricated. As is shown in Fig. 4, the microfluidic device with the size of 2cm×1.2cm consists of a silicon substrate within an etched coiled channel, and a glass cover with two inlets and one outlet according to the silicon substrate. The two inlets: cell inlet and buffer inlet are designed to introduce blood sample and lysis buffer simultaneously, leading to rapid lysis of blood cells near the location of cell inlet using the sandwich-type mixing model. Porous silicon layer fabricated on the

MEMS-Based Microdevice for Cell Lysis and DNA Extraction 35

This chapter has demonstrated an integrated microdevice capable of performing online cell lysis and DNA extraction. Firstly micro total analytical systems (μTAS) base on MEMS technology was briefly introduced in Section 1. Then microdevices both for DNA extraction and cells lyzing were reviewed. The theory of flowing in microdevice was introduced in Section 2. In microdevice channels, the flow is laminar in nature. Then a novel MEMS-based microdevice capable of performing online cell lysis and DNA extraction was reported in Section 3 and 4. The fabrication procedure of the MEMS-based microdevice was presented in detail, while the experimental of cell lysis and DNA extraction were described respectively. Then the properties of this microdevice were studied by numerical simulation and experimental. According to the results of the simulation, the mixing performance in sandwich-type coiled channels was better than that in the T-type ones. According to the results of the experimental, a high performance of cell lysis was obtained in the sandwichtype microdevice at the optimized conditions of the cell/buffer velocity ratio <1/5. 83% DNA was recovered by the porous matrix, strongly contrary to the 39.2% DNA extracted by the non-porous one. And 39.7ng PCR-amplifiable genomic DNA was obtained from whole blood samples by using an integrated microdevice with sandwich-type coiled channels and porous silicon matrix. In general, the developed integrated microdevice providing a powerful tool for biological sample pre-treatment is shown to exhibit numerous advantages over its large-scale counterparts, including rapidness, much lower blood and reagent consumption, disposability, and portability and so on, which has the potential to integrate

The authors greatly acknowledge the financial support from the National Science Foundation of China under Grant number 60701019, 60427001 and 60501020. The authors are grateful to Mr. Feng Shen of Institute of Mechanics, Chinese Academy of Sciences for his

Fig. 9. Gel electrophoresis analysis of PCR products.

into μTAS for point-of -care medical diagnose.

**6. Acknowledgement** 

assistance.

**5. Conclusion** 

surface of the internal wall of the coiled channel is designed to absorb DNA, which can strongly improve the extraction efficiency with the huge surface area. Thus this microfluidic device makes it possible to lyze cells and to purify DNA orderly.

Microfluidic devices with porous silicon matrix were used for lyzing blood cells and purifying genomic DNA, including three steps: load step, washing step, and elution step. In load step, blood cells were damaged, and the genomic DNA was released and adsorbed on the porous silicon matrix. In washing step, impurity such as protein was removed while the genomic DNA was still held on the porous matrix. In elution step, the purified DNA was desorbed from the porous matrix into the elution buffer. The process of online cell lysis and DNA extraction was in details as follows. Firstly, 10µL whole blood was introduced into the channel from the cell inlet, while 50µL lysis buffer was introduced from the buffer inlet. The lysis buffer was load buffer, namely 4M GuSCN in TE buffer with 1% Triton-X 100, titrated to pH 6.7, which included lysis reagent and bonding salt for DNA adsorption. GuSCN and Triton-X 100 were both the chemical reagents for cell lysis, while GuSCN is the bonding salt for DNA extraction. The flow velocity of whole blood sample was 5µL/min, whereas the flow velocity of lysis buffer was 25µL/min. The whole blood sample and the lysis buffer were mixed effectively at the location near the inlet, and blood cells were rapidly lyzed. After cell lysis, genomic DNA in blood cells was released and adsorbed on the porous matrix at the presence of low pH (pH 6.7) and high concentration binding salt (4M GuSCN). Secondly, 50µL washing buffer was pumped in the microchannel at 25µL/min from the buffer inlet, while the cell inlet was closed during washing step. Finally, 10µL of TE buffer was introduced in the microdevices and incubated in the microdevices for 5min at 55°C and then another 50µL of TE buffer was passed through continuously. As a whole, online cell lysis and genomic DNA purification could be implemented in less than 20min by using this integrated microdevice, while its large-scale counterparts commonly require more than several hours to finish the job.

In order to value the performance of the integrated microdevice, the elution buffer flowing out from the outlet port, in which the purified DNA was desorbed and eluted, was collected and subjected to fluorescence detection using SYBR Green I. From fluorescence detection, average 39.7ng genomic DNA was extracted from 1µL whole blood with 10.5%RSD by using three microdevices, respectively, at the same conditions. The extracted efficiency of DNA by the integrated microdevices is higher than commercial kit with silica resin which can only extract about 20~30ng DNA per microlitre blood. Moreover, it is only 10μl blood sample needed for the integrated microdevice to extract genomic DNA, while traditional phenol extraction or commercial kit requires several milliliters blood because of centrifugal operation.

The extraction of genomic DNA from a crude biological sample must be PCR-amplifiable. The lyzed cells are a complex mixture of proteins, peptides, lipids, carbohydrates, and other low molecular weight compounds that are known to inhibit DNA amplification by PCR. The genomic DNA extracted from whole blood in two separate experiments, respectively, was submitted for PCR to ensure that no inhibitory compounds were present. Shown in Fig. 9, a 203-bp fragment of -gapd gene is successfully amplified, which is identified by gel electrophoresis separation. This illustrates that the eluted DNA using the microdevice has been purified effectively for subsequent enzymatic reactions. Also shown from Fig. 9, the repeatability of this system is good enough for the downstream analytical steps.

Fig. 9. Gel electrophoresis analysis of PCR products.
