Microfluidic cell sorter with integrated piezoelectric actuator
- 1.6k Downloads
We demonstrate a low-power (<0.1 mW), low-voltage (<10 Vp-p) on-chip piezoelectrically actuated micro-sorter that can deflect single particles and cells at high-speed. With rhodamine in the stream, switching of flow between channels can be visualized at high actuation frequency (~1.7 kHz). The magnitude of the cell deflection can be precisely controlled by the magnitude and waveform of input voltage. Both simulation and experimental results indicate that the drag force imposed on the suspended particle/cell by the instantaneous fluid displacement can alter the trajectory of the particle/cell of any size, shape, and density of interest in a controlled manner. The open-loop E. Coli cell deflection experiment demonstrates that the sorting mechanism can produce a throughput of at least 330 cells/s, with a promise of a significantly higher throughput for an optimized design. To achieve close-loop sorting operation, fluorescence detection, real-time signal processing, and field-programmable-gate-array (FPGA) implementation of the control algorithms were developed to perform automated sorting of fluorescent beads. The preliminary results show error-free sorting at a sorting efficiency of ~70%. Since the piezoelectric actuator has an intrinsic response time of 0.1–1 ms and the sorting can be performed under high flowrate (particle speed of ~1–10 cm/s), the system can achieve a throughput of >1,000 particles/s with high purity.
KeywordsHydrodynamics Flow cytometer Microfluidics Sorter Piezoelectric actuation
Compared to conventional fluorescence-activated cell sorter (FACS), µFACS has the advantages of lower cost, reduced reagent usage, portability and rapid analysis time, which make them attractive for various biotechnological applications (e.g. bioanalysis and drug discovery) (Dittrich 2003). Due to the growing interest in the study of single living cells (e.g. effect of drugs and gene expression studies in cancer research (Sims and Allbritton 2007; Andersson and Berg 2003; Meredith et al. 2000), along with advances in microfabrication technology, significant progress has been made toward developing µFACS that can sort single cells with high efficiency and throughput.
Numerous types of µFACS based on the principle of continuous flow separation (i.e. continuous injection, monitoring detection, and sorting based on spatial separation (Pamme 2007) have been developed to sort various types of cells including yeast, bacterial, and blood cells. Among those published micro sorting techniques are electroosmotic (Fu et al. 2004; Fu et al. 1999), dielectrophoretic (DEP) (Lapizco-Encinas et al. 2004; Fiedler et al. 1998; Holmes et al. 2005; Doh and Cho 2005; Braschler et al. 2007; Hu et al. 2005), magnetic (Pamme 2006; Pamme and Wilhelm 2006), and hydrodynamic (Bang et al. 2006; Krüger et al. 2002; Wolff et al. 2003) sorting. Electroosmotic sorters can achieve precise flow switching, but such devices operate under high DC voltage (e.g. hundreds of volts), require frequent change of voltage settings due to ion depletion (Fu 2002), and suffer from low sorting rate (e.g. tens of particles per sec). On the other hand, DEP-based sorting exhibits precise manipulation down to single cell level, but the fabrication is highly complex and the cell differentiation capability is rather limited. For DEP-based sorters, particles and cells with similar dielectric properties cannot be easily separated from one another since they experience similar DEP forces. Magnetic sorting demonstrates high selectivity but requires labeling of magnetic particles, which cause irreversible cell damages. The aforementioned drawbacks may be resolved with hydrodynamic flow-switching methods such as sheath flow steering (Blankenstein and Lasen 1998) and integration of membrane valve to alter fluid flow (Fu et al. 2002). However, these techniques have long response time due to the inherent properties of the mechanical components. As a result, the sorting rate of these devices is low in general. Due to the advance in check valve technology, several groups (Bang et al. 2006, Wolff et al. 2003) have implemented high-speed off-chip flow-switching check valve (i.e. response time of 2.5 ms) to enhance sorting throughput (~12,000 cells/sec). However, due to the lack of precise electronics control and large electronics-induced timing jitter, both the recovery rate and purity of sorting can be low. Also, the use of bulky external actuators (syringe pumps, external check valves, etc) in this sorting scheme hinders miniaturization.
In this paper, we describe a micro-sorter with a piezoelectric/metal bi-morphous actuator. The piezoelectric film is made of lead zirconate titanate (PZT), and the metal may be copper or stainless steel. The piezoelectric actuator has a diameter of 20 mm and an intrinsic resonant frequency of around 5 kHz. This PZT actuator is integrated with the microfluidic channels using UV ozone assisted bonding process to be discussed later. Compared to hybrid systems, the integrated approach offers several advantages: 1) low voltage and low power consumption (<10 Vp-p and <0.1 mW), 2) precise control of the magnitude of transverse cell/particle deflection to enable single particle/cell sorting, 3) intrinsically much faster response (~0.1–1 ms) than conventional check or membrane valves. In addition, to suppress noise in fluorescence detection, we have designed a spatial filter and signal processing algorithm for signal-to-noise ratio enhancement and sorting verification (i.e. sorting error detection). To carry out automated sorting, characterize sorting efficiency and accuracy, and to reduce timing jitter to <10 µsec, the signal processing algorithms as well as the actuator control algorithms have been implemented in field-programmable-gate-array (FPGA).
2 Working principle
2.1 Sorting mechanism
2.2 Dynamic simulation
PA is the actuation pressure, t is time, f is the actuation frequency, and Po is the hydrodynamic pressure established by the flow. To see the effect of PZT actuator-induced flow response, 5-µm particles are released into the center of the flow stream at a velocity of 5 cm/s and the movements of the particles are modeled using Khan and Richardson’s force (Coulson et al. 1982). Time-dependent boundary parameters, PA, f, and Po, are set at 1.5 kPa, 250 Hz, and 0.63 kPa respectively.
3.1 Device fabrication
3.2 Rhodamine and bead deflection
The device is mounted on a microscope stage with a high-speed camera (Photron Fastcam, Photron Inc.) attached for visualization, and PZT actuation is driven by a function generator (Tektronix Inc.). 5 mM of Rhodamine 6G is introduced to the sample channel and hydrodynamically focused (1:10 sample to sheath flow). Once the fluid flow is stabilized, the PZT actuator is electrically modulated and the behaviors of stream under various voltage and frequency are analyzed. Similarly, 5-µm polystyrene beads (conc. ~3.2 × 105 beads/ml, from Bangs Lab) are introduced to the microfluidic channel under the same flow condition with PZT actuation operated at 250 Hz and 9 Vp-p. For both experiments, videos are taken at the sorting junction and recorded at 3,000 fps.
3.3 Deflection of E. Coli cells
An E. Coli concentration of ~7 × 105 cells/ml is prepared and suspended in 1X PBS solution. For low speed sorting, the total volumetric flow rate is maintained at 2 µl/min, and the PZT actuator is operated under 3 Vp-p at 20 Hz frequency. To increase throughput, the flowrate is increased to 18 μl/min, and the applied voltage and frequency are increased to 6 Vp-p and 200 Hz. Again, videos are recorded at 6,000 fps, and the sorted E. Coli are enumerated using frame-by-frame images to characterize the cell deflection capability of the device.
3.4 Automated sorting
4 Results and discussion
4.1 Flow switching capability
4.2 Bead trajectory
4.3 E. Coli deflection experiment
4.4 Noise suppression by spatial filter and signal processing
4.5 Sorting efficiency and sorting accuracy
The current experiment uses filtered mercury lamp as the fluorescence excitation source. To increase sensitivity for sorting of fluorescently-tagged biological agents, 488 nm laser (~40 mW) will be used as the excitation source. These developments, in conjunction with integration of optical devices with microfluidic channels, could lead to high-throughput, high efficiency, and high accuracy microfabricated cell sorter at the single cell level.
We have demonstrated, for the first time, a microfluidic cell sorter with integrated piezoelectric actuator. The device is easy to fabricate and operates at less than 10 Vp-p. In the experiment of instantaneous flow switching, we have shown that the flow stream responds to the piezoelectric actuator at high frequency (~1.7 kHz) and the amount of deflection of cells/particles in the flow can be precisely controlled. Both simulation and experiment show that particles of any size, shape, and density of interest can be individually sorted in a controlled manner. In the experiment of E. Coli deflection, a sinusoidal voltage deflects cells at a rate of 330 cells/s and shows a highly repeatable operation consistent with the theory. Using a specially design spatial filter and a real-time signal processing algorithm implemented in FPGA, a closed-loop sorting system is demonstrated with zero error rate and a sorting efficiency of around 70%. Compared with other µFACS, our sorting system has advantages in three areas. First, the spatial filter design and the real-time signal processing algorithm enhance the signal-to-noise ratio by 18 dB and allow verification of sorting. Second, the PZT-actuated sorting module is easy to fabricate, consumes minimum power (0.1 mW), operates at low voltage (<10 Vpp), and has a much faster response (0.1–1 ms) than off-chip mechanical actuators such as check-valves and syringe pumps. Third, the FPGA-based electronics control enables real-time signal amplification, user-defined delay time, programmable output waveform, and low timing jitter (<10 µsec). These features contribute significantly to a low-cost sorter that can perform high-throughput cell sorting at a single-cell level.
The research is supported by the NIH grants 1R21RR024453-01 and 1R01HG004876. We would like to thank the Nano3 staff for their technical support.
This article is distributed under the terms of the Creative Commons Attribution Noncommercial License which permits any noncommercial use, distribution, and reproduction in any medium, provided the original author(s) and source are credited.
- H. Andersson and A.v.d. Berg, Sensors and Actuators B 92, 315 (2003)Google Scholar
- J.M. Coulson, J.F. Richardson, J.R. Backhurst, J.H. Harker, Chemical engineering, 2nd edn. (Butterworth-Heinemann Ltd, Oxford, 1982), p. 970Google Scholar
- A. Y. Fu, California Institute of Technology, Pasadena (thesis). 115 (2002)Google Scholar
- D. Holmes, M. E. Sandison, N. G. Green, H. Morgan, Nanobiotechnol. (2005) doi: 10.1049/ip-nbt:20050008
- G. D. Meredith, C. E. Sims, J. S. Soughayer, N. L. Allbritton, Nat Biotech 18, 309 (2000)Google Scholar