A.
Link
and
T.
Franke
*
Division of Biomedical Engineering, School of Engineering, University of Glasgow, Oakfield Avenue, G12 8LT, Glasgow, UK. E-mail: thomas.franke@glasgow.ac.uk
First published on 21st April 2020
We demonstrate an acoustic device to mechanically probe a population of red blood cells at the single cell level. The device operates by exciting a surface acoustic wave in a microfluidic channel creating a stationary acoustic wave field of nodes and antinodes. Erythrocytes are attracted to the nodes and are deformed. Using a stepwise increasing and periodically oscillating acoustic field we study the static and dynamic deformation of individual red blood cells one by one. We quantify the deformation by the Taylor deformation index D and relaxation times τ1 and τ2 that reveal both the viscous and elastic properties of the cells. The precision of the measurement allows us to distinguish between individual cells in the suspension and provides a quantitative viscoelastic fingerprint of the blood sample at single cell resolution. The method overcomes limitations of other techniques that provide averaged values and has the potential for high-throughput.
Single cells can be mechanically probed using a variety of forces18 and some techniques have been used to assay the mechanical properties of erythrocytes. Erythrocyte mechanics has been examined dielectrically,19 mechanically, magnetically,18 optically,13,20 in shear or extensional flow21,22 and acoustically.23
Micropipette aspiration has been used to measure the viscoelastic properties of single RBCs.24–26 However, precise alignment of glass capillaries to each cell before the measurement limits this technique to small cell numbers. Similarly, atomic force microscopy, though providing high spatial resolution, suffers from low throughput and a time-consuming cell selection procedure.27,28 Dielectric forces excited by inhomogeneous electric fields have been applied to characterize immobilized erythrocytes. In these experiments cells are attached to an electrode and subjected to high frequency and high voltage electric fields (several 100 V cm−1) to measure the shear elastic modulus and viscosity of the plasma membrane.19,29,30 Magnetic and optical tweezers use microbeads attached to erythrocyte membranes to transfer magnetic and electric forces to the cell. Magnetic tweezers and magnetic twisting cytometry use ferromagnetic beads attached to the cell membrane. Optical tweezers have also been used without attached beads in multiple beam mode in more complex optical setups31 that need precise optical alignment of lasers. However, exposure to the high intensity of the electric field of the laser beam causes local heating, which affects the measurement and may cause optical damage to the cell. Other optical methods to assess RBC properties are ektacytometry,32,33 which combines a bulk rheometer setup with diffraction analysis of cells. Diffraction patterns have also been used in diffraction phase microscopy34 – a label-free method that analyses shape fluctuations of the cell contour. Bulk acoustic waves have been used to deform osmotically swollen erythrocytes in glass capillaries by acoustic radiation forces. However, only elastic effects were considered, and the cell number was low (n = 8).23 Many of these techniques suffer from complex experimental setups or low cell numbers.
Microfluidic channels enable simple and controlled dispensing of cells in a single file, allowing for continuous measurement, and have the potential for high-throughput. Several methods have been integrated into microfluidic channels in the context of erythrocyte measurements including deformation in flow,35–37 dielectrophoresis,30 optical stretching,38 and flow methods analysing cell transit through microchannels and pores.39–41 However, such methods probe cells at constant deformation and only yield the elastic parameters.3
The combination of surface acoustic waves with planar microfluidic design has been proven to be a powerful tool for cell manipulation.42 Recently, traveling surface acoustic waves (T-SAWs) in microfluidics have been used to create time averaged, stationary wave fields that have been used to custom-made pattern acoustic fields to trap particles.43–45
In this paper, we introduce an acoustic method to probe for the first time both the viscous and elastic mechanics of single RBCs, in a microfluidic device. We use a travelling surface acoustic wave (T-SAW) to generate a tuneable, standing acoustic wave field to capture and deform RBCs. Individual cells are dispensed in a single file to the nodes of the standing acoustic wave field through a microchannel and are repeatedly deformed by changing the amplitude of the acoustic field. We determine the shape deformation and the shape relaxation by comparing two independent methods of analysis. From the spatio-temporal observation of deformation of a RBC we can derive each cell's individual viscoelastic properties. We screen a blood sample of healthy RBCs and demonstrate its cell heterogeneity by quantifying the viscous and elastic properties of individual erythrocytes in this population. The method allows us to differentiate between individual members of the cell population and provides characterization of the blood sample based on its viscoelastic fingerprint.
Fig. 2 Schematic sketch of the acoustic erythrocytometer device showing the probing section of the device. Single RBCs enter the inlet flow into the measurement region and are aligned to the centreline of the microchannel which is the position of the acoustic nodal line (dashed horizontal line). The other two horizontal nodal lines at the side walls of the channel very rarely attract cells since most cells enter the proximity of the centre. Focused cells then flow along the centreline and are captured as soon as they reach the first intersection with the vertical λnf1. Further vertical nodes exist at different positions on the substrate given by the periodic spacing condition as shown where λl is the wavelength in the liquid and cl and cs are the speed of sound in the liquid and on the substrate, respectively. In the experiments we only use the first intersection to probe cells. The vertical nodes are caused by the interference of the traveling SAW excited by the IDT and its diffracted cylindrical wave excited at the edge of the PDMS wall at the fluidic channel. The superposition of these fields results in a grid-like pattern of intersecting and perpendicular nodal lines (dotted and dashed lines).43–45 |
(1) |
(2) |
This force attracts an RBC to the node in order to maximise the cell volume close to the node. The acoustic force is opposed by elastic forces that occur as soon as the RBC gets deformed. Eventually, the balance of acoustic and elastic forces determines the overall stationary deformed shape of the RBC. However, when the acoustic field rapidly changes with time as compared to the typical response time of the cell deformation, the viscous properties of the cell and its environment have to be considered. These additional viscous forces slow down the transient process of relaxation in the final stationary shape.
(3) |
Fig. 3 Micrographs and analysis of a RBC in the probing section. (a) Without application of an acoustic field the fitted contour (green line) is nearly circular and becomes progressively elliptical by increasing the acoustic amplitudes from 1 mW to 64 mW (0 dBm to 18 dBm) in doubling steps. The green contour around the cell was automatically analysed using image analysis software ImageJ and the built-in function “Analyse Particles”. The erythrocyte is labelled with a number, here cell #1. The black scale bar is 10 μm. (b) Plot of the Taylor deformation D as a function of applied power. With no applied SAW the deformation index of the RBC is D = 0.019 and for the power sweep D = 0.073, D = 0.095, D = 0.116, D = 0.144, D = 0.176, D = 0.191 and D = 0.225, accordingly. The error bars for each measure point are derived from the scattering from a single elongation around the saturation plateau for each power step. Data points are indicated by red circles and error bars by horizontal lines. Refer to ESI† Movie 1 which shows the deformation of the red blood cell with a stepwise increasing acoustic field. |
The curve progression of the Taylor deformation D follows a typical stress–strain relation of a relaxation curve with an increasing flank, a saturation at high deformation Di, a decreasing flank and again a saturation at minimal deformation. In total, we analysed 65 different RBCs in separate measurements, as shown in the scatter and box plots of Fig. 5a. The box plot displays a rather broad and slightly asymmetric distribution centred at around an average value of Dsample = 0.0729 ± 0.01896, where is the average value over all individual cells and cycles. Hence, the values for D scattering between different cells can be expressed by its relative error ΔDsample/Dsample = 0.01896/0.0729 = 26.0%. To demonstrate the significance of the distribution of the population, we plot Dcell for each cell as shown in Fig. 5a. The standard deviation for a single cell experiment is calculated from the scattering of the saturation value Di over all elongations. The plot in Fig. 5a shows that the absolute value of Dcell is much larger than the error of the repeated measurement on a single cell. For each cell we calculate the relative error ΔDcell/Dcell. On average the relative errors over all cells yield 2.6%, which is significantly smaller than the relative width of the population (26.0%). Therefore, our measurement can distinguish between single RBCs in the cell suspension and hence, has the potential to identify subpopulations of RBCs within a heterogeneous sample of RBCs. To effectively decouple the effect of cell deformation and size, we present the results in a scatter plot of D against an effective cell radius , with the projected area A as obtained at no deformation. The result as shown in Fig. 5b again underpins that the method allows individual erythrocytes to be resolved in the scatter plot and provides an elastic fingerprint of the sample.
Fig. 5 (a) Summarized results of measurements for the deformation index as obtained from single cell experiments in Fig. 4. Cells were probed at a constant acoustic amplitude of 8 mW (9 dBm) for five elongations for the Taylor deformation index Dcell and represented in scatter and box plots. The heterogeneity of the RBC population for the deformation has a mean value Dsample = 0.0728 ± 0.01896, where is the average value over all individual cells and cycles i. The standard deviation of the deformation index Dcell is much smaller than the differences of values among different cells. (b) Cell deformation and size results in a scatter plot of D against the effective cell radius . The projected area A is obtained from the area enclosed by the contour without deformation. |
Fig. 6 Summarised results of measurements of the relaxation times as obtained from single cell experiments and simple exponential fits in Fig. 4. Cells were probed at a constant acoustic amplitude of 8 mW (9 dBm) for five elongations for the relaxation times τ1 and τ2 and represented in scatter and box plots (a) for the rising deformation followed by an increase to 8 mW and (b) for the falling deformation back to the trapping amplitude of 0.8 mW. The heterogeneity of the RBC population for the relaxation time τ1 has a mean value of τ1,sample = 0.06 ± 0.05 s for the rising deformation and that for the relaxation time τ2 has a mean value of τ2,sample = 0.07 ± 0.03 s for the falling deformation, where is the average value over all individual cells and cycles i. The values for cell #18 and cell #61 exceed the range of the plot. The corresponding values are τ1,#18 = 0.165 ± 0.329 s and τ1,#61 = 0.385 ± 0.794 s. |
Using a square wave function for the acoustic amplitude of 8 mW (9 dBm), we have probed a sample of human erythrocytes. The oscillation period was chosen as T = 6 s to allow the cell deformation to relax and approach its saturation value. We analysed the cell response to the square wave excitation by fitting the data points between the square wave plateaus with a single exponential function as shown in Fig. 4. This fit assumes a simple model with the time constant, representing the ratio of viscous to elastic forces. From the analysis of the increasing and decreasing power jumps, we obtain average response times of τ1 = 0.06 ± 0.05 s and τ2 = 0.07 ± 0.03 s, respectively. These values are on the low side of previously measured time constants using different methods. However, the obtained mean values are in agreement with values from the literature using different methods, e.g., micropipette aspiration55 (∼0.3 s), microfluidic device21 (0.11 s < τ1 < 0.52 s), rheometer56 (light reflection 0.119 ± 0.017 s, ektacytometry 0.097 ± 0.015 s) and optical tweezers.57 It is apparent that the error bars for the sample average in our experiments are much higher than the ones given in the literature. This may be an indication that the scattering of viscoelastic values of a heterogeneous cell population is much larger than that previously assumed in studies where only a few cells have been probed and only represent a subpopulation. In this context, it is important to note that the error bars in our measurement mainly represent the variation of τ values between cells of the probed heterogeneous population and are not measurement errors.
The heterogeneity of response times may be an indication of the heterogeneity of the cell age. The average lifetime of erythrocytes is about 120 days and hence, a blood sample contains cells of very different cell ages. It is well known that erythrocytes are subjected to several changes upon aging, including morphological changes that are caused by the loss of the membrane material, and this has been argued to cause increased cell stiffness. Moreover, it has been reported that the viscosity of the erythrocyte cytosol increases with age.58 Density gradient isolated ‘young’ cells have been found to exhibit shorter (0.162 s) relaxation times as compared to ‘old’ cells (0.353 s) or the average of the whole cell population (0.271 s).59 In addition, measurements of response times depend on the surrounding fluid, since the shape changes are coupled to the viscous environment the cell is embedded in. For measurements in blood plasma, the response times are longer (∼0.3 s)59 than in phosphate-buffered saline solution in the literature (0.114 s),59 as well as in our experiments presented in Fig. 6.
It still remains challenging to link the cell response and the temporal progression of the shape deformation to the simple biomaterial properties of the cell, as given by its constituents. The RBC membrane is a complex composite envelope composed of a plasma membrane and macromolecular network, the cytoskeleton, attached to the cytoplasmic face of the plasma membrane.60 The mechanics of these two coupled shells contribute to the viscoelastic response of the erythrocyte. There are two restoring elastic moduli, the elastic bending stiffness κ and the shear elasticity μ, as given by the properties of the plasma membrane and the cytoskeleton, respectively. During the relaxation process, this elastic energy is opposed by dissipation in the membrane and the bulk fluid. These two contributions depend on the membrane shear viscosity ηmem and effective bulk viscosity of the fluid, which represent both the cytosol and the surrounding fluid. From these moduli, characteristic times for bending and shear can be estimated by τb and τs.21,24
(4) |
(5) |
The transient erythrocyte deformation in our experiment contains both shear and bending deformations and it is difficult to estimate the relative contribution of each without having analysed the exact 3D shape of the cell. Moreover, even if a full 3D-contour is known, simulations are still required to model the temporal rearrangement of the cytoskeleton network that dominates the shear effects.62 Therefore, in our simple model using a one exponential fit for the relaxation, we cannot resolve all the complex mechanical aspects of the cell. A more refined model using multiple exponentials based on Kelvin–Voigt-like models has been used in the past.63 However, a more thorough theoretical analysis in three dimensions, which includes both the erythrocyte morphology and the three-dimensional acoustic field of our device, will be necessary, and that can only be provided by simulations.64,65
The scatter plot in Fig. 5a shows the time-independent deformation of the erythrocyte sample and only depends on the elastic parameters. This comparatively simpler situation disregarding viscous effects has been analysed for cells using iso-elasticity lines.3 Yet, for erythrocytes, this still remains challenging since multiple elastic moduli, such as bending and shear elasticity, are involved. In contrast to spherical cells, the analysis is further complicated by the non-spherical discoid shape of RBCs. The erythrocytes in our experiments align with their axis of symmetry perpendicular to the substrate so that the non-deformed contour is circular and simple to analyse. However, in the deformed state, also components of deformation out of this projection may occur which have not been taken into account. Again, a full 3D-analysis would be necessary to provide a more detailed analysis and provide lines of constant elastic moduli.
Here, we have demonstrated that a simple phenomenological model is useful to analyse the experiments of our acoustofluidic device and to determine elastic and viscous experimental parameters. The results indicate that the error of a repeated measurement of a single cell is significantly smaller than the scattering of experimental values between different cells in the population. Hence, the resolution of our device allows the analysis of an erythrocyte sample at the single cell level.
In recent publications,23,66 it was shown that ultrasonic manipulation provides precision and speed and can potentially work with thousands of cells whereas micropipette aspiration and optical tweezers are limited to a small number of cells. There are techniques available to probe the mechanics of cells at a high-throughput of a few hundred cells per second, for example using deformation in flow.3,67 However, these techniques exploit the static deformation such as the ellipticity of cells allowing only to derive the elastic parameters. Determination of viscous parameters requires an extended observation of temporal shape changes. Generally, the throughput of our device is therefore inherently limited by the response time of the cells. The cell sample rate has to be larger than the inverse of the response time to allow observation of the transient states of relaxation. For typical response times of 0.1 s this limits the throughput to 10 cells per second. However, exploiting multiple nodes of the acoustic wave field is a way to increase the sampling rate. This can be done for example by using the nodes in the channel shown in Fig. 2. In conclusion, combining microfluidics and acoustics allows the non-invasive, repeatable, electrically tuneable manipulation of living cells.
Footnote |
† Electronic supplementary information (ESI) available: Movie showing the surface acoustic wave actuated deformation of a red blood cell for stepwise increasing power amplitudes. See DOI: 10.1039/c9lc00999j |
This journal is © The Royal Society of Chemistry 2020 |