David
Barat†
a,
Daniel
Spencer†
a,
Giuseppe
Benazzi
a,
Matthew Charles
Mowlem
b and
Hywel
Morgan
*a
aSchool of Electronics and Computer Science, University of Southampton, Highfield, Southampton, SO17 1BJ, UK. E-mail: hm@ecs.soton.ac.uk
bNational Oceanography Centre, University of Southampton Waterfront Campus, European Way, Southampton, S014 3ZH, UK
First published on 3rd November 2011
We describe a microfluidic cytometer that performs simultaneous optical and electrical characterisation of particles. The microfluidic chip measures side scattered light, signal extinction and fluorescence using integrated optical fibres coupled to photomultiplier tubes. The channel is 80 μm high and 200 μm wide, and made from SU-8 patterned and sandwiched between glass substrates. Particles were focused into the analysis region using 1-D hydrodynamic focusing and typical particle velocities were 0.1 ms−1. Excitation light is coupled into the detection channel with an optical fibre and focused into the channel using an integrated compound air lens. The electrical impedance of particles is measured at 1 MHz using micro-electrodes fabricated on the channel top and bottom. This data is used to accurately size the particles. The system is characterised using a range of different sized polystyrene beads (fluorescent and non-fluorescent). Single and mixed populations of beads were measured and the data compared with a conventional flow cytometer.
Many commercial machines use a combination of both optical and electrical techniques to measure cells and other particles. These systems are complex and are capable of extremely high throughputs; up to 105cells per second.1 However, they are expensive and bulky, and are unsuitable for applications such as point of care diagnosis or in situ analysis. In an attempt to miniaturise the functionality of a flow cytometer, miniaturised Lab-on-Chip (LoC) systems have been developed; for recent reviews see.3–6 Chip-based devices have a number of advantages - they can process extremely small sample volumes (tens of μLs) containing low numbers of particles while maintaining high statistical accuracy. They offer the possibility of integrating sophisticated sample handling and detection modalities, within a single chip, allowing the development of novel assays and measurement systems. Miniature flow cytometers were first demonstrated by Kamentsky in the mid-1960s.7 Over the last decade a number of different microfabricated flow cytometers have been described, mostly optical. Early devices demonstrated the principle of miniaturisation, but the sensitivity and particle throughput was very poor. Miniaturisation and integration of optical components into LoC devices is an area of considerable activity,8 and important to the development of miniature flow cytometer systems. Efficient coupling and focusing of light into a microfluidic channel is a challenge and generally an optical fibre or waveguide is used; many different designs have been published.9–17
One of the first attempts at integration was by Kruger et al.,9 who implemented a leaky waveguide and an avalanche photodiode into a microfluidic channel. A holographic diffraction grating was used to collect light; they demonstrated fluorescence detection of beads. Wang et al.10 fabricated a device with integrated waveguides and a lens that focused the light into the centre of a channel. They used scattered light to count and discriminate beads, at speeds of up to 25 beads per second. Sheath flow was used to centre the sample stream in the optical detection region. This was one of the first papers to demonstrate integrated optical elements in a micro-flow cytometer. Tung et al.11 used solid-state lasers and PIN photodiodes with lock-in amplification. Using microgrooves for fibre alignment, they were able to guide incident light and collect fluorescence from a detection region. They demonstrated two-colour fluorescence detection (440 nm and 635 nm) and multi-angle detection (45°, 135° and 180°), measuring fluorescently labelled yeast cells at a rate of up to 500 particles per second.
Simple waveguides are generally lossy and a number of techniques have been developed to improve them. Using an SU-8 core and PDMS cladding, Mogensen et al.12 produced waveguides with a propagation loss of about 4.5 dB cm−1 and 1.4 dB cm−1 at 532 nm and 633 nm respectively. Bliss et al.13 launched light into a device using liquid-core/liquid-cladding waveguide (L2 waveguide). The incident light was launched through a fibre-to-waveguide coupler propagating into an L2 waveguide. Although these waveguides have low propagation losses, the overall losses are significant due to the number of interfaces. An improvement uses integrated wavelength-selective optical waveguides.14 Using a long length liquid-core waveguide (PDMS and glycerol) doped with dye molecules they demonstrated a detection waveguide which selectively absorbs the excitation light (532 nm), with propagation losses of 120 dB cm−1 at this wavelength, but only 4.4 dB cm−1 at 633nm. However this system is limited by the auto-fluorescence of the dyes, adding noise to the detection. Another technique is an antiresonant reflecting optical waveguide, ARROW. Bernini et al.15 demonstrated a micro flow cytometer using silicon integrated hollow core ARROW. The excitation light and sample fluid travel co-linearly through the detection region. Fluorescence is collected using two orthogonal fibres and by adding an L2 waveguide,16 they focused the incident light vertically and horizontally. However, the fluidic part was complex and no data on cell/particle analysis presented. Lee et al.17 described a micro flow cytometer with buried SU-8/SOG optical waveguides with propagation losses of 4 dB cm−1 at 633 nm. They used an etched single-mode optical fibre to couple the incident light into the waveguide. Integration of micro-lenses can improve focusing of incident light into the detection region. Compound micro-lenses collect and focus light without optical aberration but these lenses must be precisely fabricated and aligned in order to minimise optical aberration and achieve precise focusing along the optical axis. Multimode fibres have also been used to collect light in miniature cytometers. These fibres have a larger diameter core than single mode fibres so that they collect more light; device fabrication is somewhat simpler too. Examples of the use of such fibres for miniature micro-machined cytometers can be found in the work of Ligler et al.,18–20
Various impedance cytometers have also been developed and these show promise for high-speed analysis of particles based upon size and dielectric properties - for a review see.21 With single frequency excitation, micro-cytometers can count and size particles, but multi frequency measurements have been used for discrimination of cells.22,23 In conventional cytometry, particle diameter is measured from the FSC signal, but this parameter is difficult to implement in a miniaturised system. Impedance analysis offers an attractive alternative for particle volume determination. The technology is somewhat simpler to miniaturise than optical systems since the device only requires pairs of aligned microelectrodes fabricated within a microfluidic channel.
To maximise the functionality of a miniature particle analysis system, we have developed a micro-cytometer that can simultaneously measure the optical and electrical properties of single particles. Incident light is launched through an integrated compound lens.24 The system measures side scattered light using optical fibres inserted in grooves, and fluorescence light with fibres placed orthogonal to the excitation light. Micro-electrodes simultaneously measure the low-frequency electrical impedance (volume) of the particles.22,25 The complete system is characterised by measuring a range of different sized beads (fluorescent and non-fluorescent) and the data compared with a conventional flow cytometer.
Fig. 1 Overview of the cytometer chip and details of the optical and electronic setup. (a): Schematic diagram showing 1-D hydrodynamic focusing, micro-electrodes for impedance spectroscopy and the different grooves for holding the fibres (incident light, scattered light and fluorescence). (b): Photograph of a chip fabricated from glass and SU-8. (c): Cross section of the detection area showing the microelectrodes and the setup for impedance spectroscopy. (d): View of the electrodes as seen along the microfluidic channel demonstrating how 1-D hydrodynamic focusing centres the particles in the electric field. (e): Details of the optical design, showing the compound air lens, the light collection fibres and the optical paths of the detection region, (the refractive index of SU-8 and water are at 532 nm). (f): Photograph of the final microfabricated structure. The two black lines crossing the channels are the microelectrodes which are slightly misaligned. |
The microfluidic chips were fabricated using photolithography and full wafer thermal bonding; the fabrication process has been described elsewhere.26,27 Metal electrodes consisting of 200 nm thick platinum with a 20 nm Ti seed layer were fabricated on 100 mm diameter glass wafers by photolithography. Electrodes were 20 μm wide with 40 μm gap. SU-8 resist was used to define the fluidic channel, grooves for the fibres and the air lens. The microfluidic channel is 80 μm high and 200 μm wide in the detection region. SU-8 negative photoresist (Microchem) was spin coated onto the surface of Pyrex wafers over the patterned electrodes, followed by a soft bake. The two SU-8 layers (one on each wafer) were each 40 μm thick, giving a total channel depth of 80 μm, which enabled insertion of the optical fibres. Pairs of wafers were aligned and bonded using a thermo-compression technique to form sealed microfluidic structures. The wafers were designed with alignment marks on each chip to aid accurate alignment, so that the electrodes and channels could be aligned to within 5 μm. Individual chips were released from the bonded wafer pair by dicing, and holes for fluids made using a mechanical drill. Fig. 1(b) shows a photograph of a chip, which is approximately 25 mm wide.
Details of the optics are shown in Fig. 1(e). A groove in the SU-8 holds a fibre which launches incident light perpendicular to the channel. This light is focused into a sheet across the width of the channel using an air compound lens (r1 = −102 μm, r2 = 83 μm) - for further details see.24 Fluorescence emission is collected with fibres placed in two grooves on the same side as the incident light (at 135°). A 7° fibre was used to measure the optical extinction (EX) signal - light loss due to absorption or scatter out of the field of view of the detector when a particle passes through an incident beam.2 The chip was designed for two more collection fibres to be placed at 22° and 45°, to measure side scattered light (SSC). The data from these two collection fibres was qualitatively similar (for the beads) and therefore only light from the 45° fibre was collected and analysed. Fig. 1(f) is a photograph of the detection region showing the lens, fibre grooves and the two pairs of electrodes (with slight offset due to fabrication).
Fig. 2 Diagram showing focusing of the incident light across the channel. (a): Ray tracing simulation of the optical system designed in ZEMAX with 100,000 analysis rays and an incident power of 1 W @ 532 nm. (b) Profile of the measured light intensity across the channel width at the midpoint (see image in (c)). Also shown are the simulated results which give a FWHM of 70 μm compared with a measured FWHM of 140 μm. The two sharp dips in measured intensity occur because the electrodes obscure the light in these regions. The detector viewer for the simulation is 400 × 400 μm2 with 150 pixels for X and Y axis. (c) Photographs of the light in the channel imaged using fluorescein solution. The lines indicate the boundaries of the channel and the dotted line is the centre of the channel. The dark areas are the electrodes. |
Bead diameter (μm) | Events measured | Sample flow rate (μL min−1) | Manufacture's CV (%) | CV size (%) | CV 45° SSC (%) | CV Fluorescence (%) | ||||
---|---|---|---|---|---|---|---|---|---|---|
Micro-cytometer | FACS | Micro-cytometer | Micro-cytometer (|Z|) | FACS (FSC-A) | Micro-cytometer | FACS (SSC-A) | Micro-cytometer | FACS (PE-A) | ||
7 | 1036 | 1419 | 14.2 | 7.2 | 23.4 | 6.8 | 38.4 | 5.9 | ||
10 | 1622 | 1539 | 43.4 | 4.1 | 24.4 | 3.9 | 44.8 | 12.3 | ||
15 | 804 | 1415 | 29.4 | 6.5 | 10.8 | 10.3 | 12.0 | 13.4 | 17.1 | 12.7 |
20 | 1114 | 1652 | 32.8 | 2.4 | 9.4 | 13.3 | 26.7 | 13.5 | ||
25 | 3111 | 1195 | 30.6 | 5.9 | 4.4 | 1.7 | 18.8 | 4.9 | ||
31 | 822 | 994 | 24.5 | 11.0 | 11.9 | 7.7 | 20.1 | 13.0 | 23.8 | 15.8 |
Fig. 3 Scatter plots for fluorescent 15 μm beads measured using the micro-cytometer and the BD FACSAria. (a) 580 nm fluorescence vs. impedance (at 1 MHz), inset is a histogram of impedance. (c) 45° Side scatter signal vs. impedance at 1 MHz; inset shows histogram of the 45°SSC. Both data sets are from the micro-cytometer. (b) 575 nm fluorescence signal vs. forward scattered light, signal area (FSC-A) for the FACS, inset shows FSC histogram. (d) 45° side scattered signal vs. FSC-A from the FACS; inset shows SSC histogram. |
Fig. 4 Summary of data obtained with the micro-cytometer for a wide size range of beads from 7 μm to 31 μm diameter. (a) Histograms of impedance, (b) Histograms of the 45° side scatter and (c) extinction at 7°. For particles of 15 μm and above, two distinct populations are observed in the impedance and SSC signals, owing to the onset of inertial focusing. |
At the flow rate used in these experiments, two populations were observed in impedance and SSC for beads greater than 15 μm diameter (Fig. 4). Di Carlo et al.29,30 has shown that when particles flow through microfluidic channels at intermediate Reynolds numbers, they experience a combination of counteracting inertial lift forces (wall lift) and shear gradient lift which focuses them into equilibrium positions near the channel walls. Our microfluidic chip has a rectangular cross section, which leads to focusing into two equilibrium positions.25,31,32
The impedance signal from our chip, which has two pairs of overlapping parallel electrodes, depends on the vertical position of the particle in the channel, but not on the horizontal position.25 Particles moving along the bottom of the channel close to the measurement electrodes (see Fig. 1(c)) have a higher impedance signal than those flowing in the top half of the channel (connected to the source). Particles flowing in the centre of the channel have the lowest impedance signal. The particle transit time can be obtained from the impedance signals and since the flow profile across the channel is parabolic, this transit time reflects the distance of the particle from the centre of the channel in the vertical or z-axis (Fig. 1). A density plot of transit time vs. impedance magnitude for 25 μm diameter beads is shown in Fig. 5(a). The data shows that there are two populations of particles, one moving in the top half of the channel (lobe L1) with a small distribution in impedance magnitude, and another in the bottom half of the channel with a much wider distribution (L2). As particles move off centre (along the vertical axis) the impedance signal increases in magnitude but by a different amount in each of the upper and lower halves of the channel. This leads to the tick-shaped distribution shown in Fig. 5(a,b). As particle move off centre, the shape of the impedance-time signal also becomes distorted from the simple double anti-symmetric Gaussian that occurs in the centre. The transit time is calculated using simple peak detection, but as the signal becomes more distorted this method introduces errors and the transit time becomes biased, see ref. 25 for further details. Therefore, although the particles flow with the same speed in the top and bottom halves (lobes L1 and L2), experimentally calculated transit times are skewed. The actual particle transit time is the mean of the two values shown in L1 and L2. The flow rate used in this experiment was 50 μL min−1, equivalent to a mean fluid velocity of 50 mm s−1 and a peak velocity of 120 mm s−1. This equates to a minimum particle transit time of 0.5 ms (60 μm electrode centre to centre spacing), equal to the minimum experimental value determined from the impedance signals - Fig. 5(a).
Fig. 5 (left) Density plot of transit time vs. impedance magnitude for 25 μm diameter beads. (a) Experimental results, (b) Simulated data. Two populations of particles are observed in the experimental results, one in the top half of the channel (L1) and a second in the bottom half of the channel (L2). Right panel: (c) Extinction, and (d) SSC vs. impedance for the same data as in (a) Each event is colour coded according to its position within one of the two lobes in (a); red top, green bottom of the channel. The extinction data does not depend on position within the channel. The SSC data demonstrates clear top/bottom differentiation. |
The distribution in transit times and impedance seen in this figure shows the onset of inertial focusing where particles begin to move from the fastest flow region (channel centre) into the upper and lower halves. Observation of the 1-D hydrodynamically focused particle stream with a microscope indicated that the sample width was approximately 40 μm wide, and flowed in the centre of the channel. Quantitative analysis of the number density of particles in each of the two lobes in Fig. 5(a) shows that the distribution top to bottom is 50/50 with greatest numbers of particles in the centre of the channel, as observed for 1-D hydrodynamic focusing in a rectangular channel.
This data can be compared with simulation – Fig. 5(b). The electric field and impedance signal was simulated using COMSOL, the particle velocity determined from the analytical solution to the Navier Stokes equation33 and the impedance calculated as described previously.25 The simulation was performed assuming viscous flow and assuming a random distribution of particle position in the channel. The combination of a parabolic flow profile together with the random particle distribution means that most of the particles flow through the channel centre (with τ = 0.6 ms). Comparing simulations with data (Fig. 5(a)), demonstrates that at the intermediate Reynolds numbers used in these experiments (≈10) inertial lift forces are beginning to have an effect and the particles are focused away from the centre into two vertical equilibrium positions.
Fig. 5(c, d) shows SSC and extinction data, against impedance for the same beads (25 μm diameter). Each event has been colour coded according to its position within one of the two lobes in Fig. 5(a). The extinction data is independent of position within the channel; particles simply obscure the light path as they move, regardless of z-position. However, the SSC data demonstrates clear top/bottom differentiation. Particles that move along the bottom half of the channel and have a higher impedance also have a higher SSC signal than those with the lower impedance that move in the top half. There are two populations as seen in the histogram in Fig. 4(b), demonstrating that variations in the z-position of particles leads to large changes in the reflected optical signal strength. By contrast the extinction signal has a single wide distribution, and is less dependent on particle position.
The simultaneously measured electrical (impedance) and optical (SSC and fluorescence) properties of the mixed population of beads after w-filtering is shown in Fig. 6. Approximately 2500 events are plotted; both SSC and fluorescence against impedance (trigger channel). Histograms for each parameter are also shown. The figure shows that the 4 different bead sizes can be distinguished from optical side scatter, but the impedance signal provide much better discrimination between the populations. The fluorescence signal from the 10 μm and 15 μm beads provides easy discrimination in this channel. For comparison, the same mixed population analysed with FACS (fluorescence vs. forward scatter) is also shown.
Fig. 6 SSC, fluorescence and impedance data from a mixture of different beads (10, 15, 20 and 25 μm diameter) after w-filtering. Also shown is FACS fluorescence data for the same mixture. Particles can be discriminated from either SSC or impedance, as well as fluorescence intensity. |
Footnote |
† These authors contributed equally to the work. |
This journal is © The Royal Society of Chemistry 2012 |