Lab on a Chip

Micromagnetic–microfluidic blood cleansing device

Chong Wing Yung ^a, Jason Fiering ^b, Andrew J. Mueller ^b and Donald E. Ingber *^ac
aVascular Biology Program, Department of Surgery and Pathology, Children's Hospital and Harvard Medical School, 300 Longwood Ave., KFRL 11.127, Boston, MA 02115-5737, USA. E-mail: donald.ingber@childrens.harvard.edu; Fax: +617-730-0230; Tel: +617-919-2223
^bCharles Stark Draper Laboratory, 555 Technology Square, Cambridge, MA 02138, USA
^cWyss Institute for Biologically Inspired Engineering, School of Engineering & Applied Sciences, Harvard University, Cambridge, MA 02139, USA

Sepsis is a lethal disease caused by a systemic microbial infection that spreads via the bloodstream to overwhelm the body's defenses. Current therapeutic approaches are often suboptimal, in part, because they do not fully eliminate the pathogen, and hence the source of deadly toxins. Here we describe an extracorporeal blood cleansing device to selectively remove pathogens from contaminated blood and thereby enhance the patient's response to antibiotic therapy. Immunomagnetic microbeads were modified to create magnetic opsonins that were used to cleanse flowing human whole blood of Candida albicans fungi, a leading cause of sepsis-related deaths. The micromagnetic–microfluidic blood cleansing device generates magnetic field gradients across vertically stacked channels to enable continuous and high throughput separation of fungi from flowing whole blood. A multiplexed version of the device containing four parallel channels achieved over 80% clearance of fungi from contaminated blood at a flow rate of 20 mL/h in a single pass, a rate 1000 times faster than a previously described prototype micromagnetic–microfluidic cell separation system. These results provide the first proof-of-principle that a multiplexed micromagnetic–microfluidic separation system can be used to cleanse pathogens from flowing human blood at a rate and separation efficiency that is relevant for clinical applications.

Sepsis, a body's response to a systemic microbial infection, is the leading cause of death of immunocompromised patients, and is responsible for over 200,000 deaths per year in the United States.^1,2 The onset of sepsis occurs when rapidly growing infectious agents saturate the blood and overcome the body's immunological clearance mechanisms.³ Most existing therapies are ineffective, and patients die because of clot formation, hypoperfusion, shock, and multiple organ failure.⁴

Current treatments for sepsis and septic shock include use of recombinant drugs, membrane filtration devices, antibiotics and blood transfusion with packed red blood cells, with the major focus being removal of inflammatory mediators that contribute to multi-organ system failure.^5–9 One form of recombinant therapy for sepsis is a human recombinant activated protein C (Xigris) used to restore dysfunctional anticoagulation, prevent microvascular thrombosis, and properly modulate the systemic inflammatory response.^10,11 However, this treatment can cause severe bleeding, deplete pro-inflammatory mediators necessary for proper immune function, and it has questionable efficacy.^11–15 Hemofiltration and hemadsorption devices, which are used to non-specifically remove blood proteins, can also remove many inflammatory cytokines that are required to fight infectious agents.^8,9 Interestingly, antibiotic therapies that simply act by reducing the pathogen load can be effective in some sepsis patients.^5,16 There are also clinical cases where the lives of septic infants were saved by performing whole body transfusions which removed pathogens from the blood.¹⁷ Generalized use of these antibiotic therapies are unfortunately limited by dose toxicity and blood transfusions can deplete the patient of endogenous immune components and cells needed to fight the infection.¹⁸ However, the success of these approaches indicate that a biomedical device that can selectively cleanse pathogens from a patient's blood without altering critical blood components can be an effective method for treating sepsis when used in conjunction with conventional therapies, such as fungicides.

Microfluidics technologies, such as the H-filter, offer a novel approach for selective separation of particulates from flowing liquids without the need for a filter membrane and they provide large interfacial surfaces to increase separation efficiency.^19–22 For example, microfluidic systems that utilize strong rare-earth magnets positioned adjacent to microfluidic channels have been developed to separate cells, magnetically-labeled cells and particles from a single sample stream.^23–26 But these devices have limited clearance capacity because captured species tend to accumulate near the magnet and obstruct flow over time. To address this limitation, we developed a prototype micromagnetic–microfluidic device that uses a collection fluid in a second laminar flow stream in direct contact with the flowing sample to continually carry away cells that are magnetically pulled across the flow stream boundary.²⁷ For targeted removal, magnetic opsonins (i.e. magnetic nano- or micro-beads coated with pathogen-specific antibodies) were used to selectively bind bacterial pathogens within a red blood cell solution at a rate of 20 µL/h.²⁷ Here we have developed a more clinically relevant microfluidic device that employs a multiplexed design to achieve high separation efficiency of fungal pathogens from human whole blood (hWB) at a rate 1000 times faster than our initial prototype.

The micromagnetic–microfluidic blood cleansing device (MMBCD) is composed of three major components: a polymeric microfluidic flow cell, a magnetic field concentrator, and a tunable external electromagnet (Fig. 1).²⁸ The microfluidic flow cell was created with standard microfabrication techniques and was assembled through plasma-bonding of four layers of molded polydimethylsiloxane (PDMS) cast from SU-8 masters (Fig. 2). The two inner layers contained wide channel features of (0.5 × 0.2 × 20 mm; w × h × l) that allowed direct contact between blood and collection flow streams, necessary for the magnetic separation. The two outer layers contained bifurcated channel manifolds that distributed fluid into the four inner separation channels. Thin inner PDMS layers (250 µm) were cast in order to position the channels in close proximity to the field concentrator and thereby maximize the magnetic field gradient within the separation chambers. Tubing connected from the side of the device allows for device stacking.

	Fig. 1 Micromagnetic–microfluidic blood cleansing device (MMBCD) is assembled from three major components. The polymeric microfluidic flow cell (1) contains an array of channels where blood (bottom channel) and saline collection (top channel) fluids flow side-by-side under laminar conditions. A solenoid electromagnetic (2) generates magnetic fields that are conducted vertically across the microfluidic channels by a C-shaped steel core. (Inset) Strong magnetic field gradients directed vertically upwards (gray lines) are created by the ridged geometry of the magnetic field concentrator (3) and pull opsonized pathogens (green dots) from the blood into the collection fluid. Multiplexing was achieved by increasing the number of channels within each device, and in the future could be achieved by stacking up multiple concentrators and devices (dashed outlines).

	Fig. 2 Microfabrication approach used to create the microfluidic device. The polymeric microfluidic flow cell was fabricated by plasma-bonding four individually molded polydimethylsiloxane (PDMS) layers. The two inner layers were molded on a silicon wafer master (thick black line), which contains two levels of SU8 features (gray lines). The short features form channels for fluid flow, and the tall features form through-holes that connect to the outer layers and also act as a stop for the piston. The two outer layers are molded by standard polymer micromolding techniques and form channel features for fluid inlet, distribution and outlet.

Our design has several advantages when compared to previous magnetic separation flow cells. It employs a separate metal field gradient concentrator layer with surface ridges that run directly above the entire length of each channel (inset of Fig. 1). Since this magnetic field concentrator is not placed within the PDMS layers as in past model systems,²⁷ multiple channels can be densely arrayed within a single polymeric device to increase throughput. Further multiplexing can also be achieved by stacking multiple devices vertically, interposed with multiple metal gradient concentrators that are placed between each PDMS layer inside a single electromagnet housing (Fig. 1). The electromagnet was constructed from a 1500 turn, 47 solenoid and a C-shaped steel core. The magnetic field concentrator, also machined from high magnetic permeability steel, has four individual ridges (1 × 1 × 20 mm; w × h × l), spaced 3 mm apart, and is attached to the top side of the C-shaped core. The total air gap between the top surface of the ridges and the opposing face of the magnet is 5.7 mm. The electromagnetic field strength of the concentrator was measured using a Teslameter (F.W. Bell 5080) and field gradient was quantified by measuring the change in the field strength at a distance of 0.25 mm normal to the surface of a ridge.

Candida albicans used for experiments were cultured in suspension in YPD media (BactoYeast 10 g/L, BactoPeptone 20 g/L, Dextrose 20 g/L) at 37 °C, shaking at 250 rpm overnight. The yeast form was harvested through centrifugation (1000g) and resuspended in phospate buffer saline (PBS) before use.

Super-paramagnetic beads (1 µm diameter, tosyl-coated; MyOne Dynabeads®, Invitrogen) were conjugated to biotinylated polyclonal antibodies (Abcam) against C. albicans fungi surface antigens. Beads were incubated with antibody (0.08 to 1.60 µg/µL) dissolved in carbonate buffer (Na₂CO₃ 1.6 g/L, NaHCO₃ 0.93 g/L, pH = 9.4) overnight at 37 °C on a rotator, and then collected magnetically and washed twice in PBS (plus 0.5% Human Serum Albumin (HSA), 0.05% Tween20, 0.02% NaN₃) before rotating overnight at 37 °C. The beads were then washed again with PBS containing 0.1% HSA, 0.05% Tween20, and 0.02% NaN₃ and stored at 4 °C.

C. albicans cells were labeled using CellTracker Blue CMAC (353/466; 7-amino-4-chloromethylcoumarin; Molecular Probes) for bead binding assays. Cells were incubated with magnetic opsonins for 1 h at room temperature before bead-bound-fungi were removed with a permanent (Nd-Fe-B) magnet. The bead binding efficiency was indirectly quantified by measuring the remaining fluorescence after magnetic removal of the fungal cells (Victor3; PerkinElmer).

For cell binding assays performed in whole blood, a multi-fluorescence labeling strategy was used to distinguish beads from unbound and bead-bound-fungi (Fig. 3a). Beads conjugated with biotinylated antibodies were labeled with streptavidin-Alexa647-R-PE (496/668; Molecular Probes), and C. albicans cells were labeled with CellTracker Green CMFDA (491/517; 5-chloromethylfluorescein diacetate; Molecular Probes). Human whole blood samples that contained beads and fungi were incubated in a red blood cell (RBC) lysis buffer (FACS Lysing Solution, BD Biosciences) for 15 min at room temperature and assayed without washing in a four-color flow cytometer (FACS Calibur, BD Biosciences). Fluorescence-based gates were drawn around individual populations within dot plots to identify lone beads (only red fluorescence), lone fungi cells (only green fluorescence) and bead-bound-fungi (both red and green fluorescence) (Fig. 3b). CountBright counting beads (UV-635/385-800; Invitrogen) were used to simultaneously quantitate the absolute concentrations of all populations within one assay, thereby allowing accurate comparison among samples.

	Fig. 3 Multi-fluorescence labeling and quantitation of magnetic opsonins and C. albicans. (a) Magnetic beads were conjugated with antibodies and Alexa647-R-PE (red) to create magnetic opsonins that selectively bound live C. albicans fungal pathogens stained with CMFDA CellTracker (Green) in human whole blood. (b) Magnetic opsonins (red) can be distinguished from bound (orange) and unbound pathogens (green) in RBC-lysed whole blood samples using a 4-color flow cytometer. Counting beads were added to blood sample to accurately quantitate the concentration of individual populations.

Binding kinetics of fungi (C. albicans) and beads were first examined under batch conditions to optimize bead binding. C. albicans cells were suspended in PBS or hWB (from Children's Hospital Boston blood bank) at a concentration of 10⁶ fungi/mL. Citrate–phosphate–dextrose–adenine (cpda-1) was used as a preservative and heparin (0.05 U/mL) was added as an anti-coagulant to maintain clot free flow (in microfluidic experiments). Bead binding saturation experiments were performed by varying the ratio of beads to fungi (R_b/p = 20 to 600) and samples were incubated in an end-over-end rotator (16 rpm) at 37 °C for 1 h. Kinetic binding experiments were performed over 4 h under similar incubation conditions. Flow cytometry was used to quantify the concentrations of bound fungi (positive for both CMFD and Alexa647-R-PE fluorescence) and unbound fungi (positive for only CMFD) in order to determine the fraction of cells bound by beads (binding efficiency). Flow cytometry was also used to quantify the number of beads bound per fungi based on the increase of Alexa647-R-PE fluorescence between bead bound fungi and lone beads.

To assess the cell separation efficiency of the multiplexed MMBCD, beads and fungi were pre-incubated in PBS to allow maximum binding before resuspending in heparinized hWB. The effects of varying electromagnet current on fungal separation efficiency were analyzed using bead-bound-fungi suspended in PBS. Two individual syringe pumps (New Era Pumps) were used to inject PBS or hWB samples into the bottom channels and PBS collection fluid into the top channels. To test how the viscosity of the collection fluid affected its hydrodynamic interaction with blood, medical grade dextran (40 kDa, Sigma) was used to vary the viscosity. Dextran was dissolved in PBS at 5, 10 and 20% to produce solutions with viscosities of 2, 3, 11 centipoise at room temperature, and all experiments were performed with a sample flow rate of 5 mL/h/channel. The 4-channel MMBCD, which was used to separate C. albicans fungal pathogens from heparinized hWB, had a total flow rate of 20 mL/h. The PBS collection fluid flow rate was 80 mL/h. Magnetic separation in hWB samples was achieved at a 10 V setting, which corresponds to a magnetic gradient of approximately 50 T/m. Fluid samples were collected from bottom-inlet, top-outlet, and bottom-outlet channels and analyzed by flow cytometry to assess the separation efficiency of beads and bead-bound fungi (Efficiency = 1 − X _bottom-out/X _bottom-in). Blood loss was quantified by measuring the OD₆₀₀ of red blood cells (Loss = OD _top-out/OD _bottom-out).

All measured values are reported as an average of at least triplicate samples ± standard deviation (SD, as indicated by error bars in all graphs). Non-linear regressions of saturation and kinetic binding data were performed using Mathematica 6.0. The ANOVA t-test was used to determine p-values; p < 0.05 was considered statistically significant.

The optimal antibody concentration required for bead conjugation was assessed using a pathogen binding assay. The pathogen binding efficiency of the beads followed a first-order binding relationship, increasing with the concentration of conjugated antibodies and saturating at approximately 96 ± 1% at an antibody concentration of 1 µg/µL (Fig. 4). In contrast, only a low level of unspecific binding (p < 8 × 10⁻⁷) was detected when beads conjugated with anti-E. coli antibodies were mixed with C. albicans cells. Functionalized beads were also found to not bind blood cells based on visual examination under bright-field and fluorescence microscopy, as shown in Fig. 3.

	Fig. 4 Magnetic opsonins have high binding specificity and efficiency. Magnetic opsonins made from 1 µm magnetic beads conjugated with anti-C. albicans antibodies showed a high level of binding affinity for the fungal pathogens in PBS (96%). The high specificity of the magnetic opsonins was demonstrated by a low level of background binding when similar beads were conjugated with anti-E. coli antibodies.

The optimal amount of beads needed to bind fungal cells was determined by varying the bead : pathogen ratio (R_b/p) under batch conditions. The binding efficiency followed a Langmuir adsorption model with fungal cell binding increasing with bead concentration. A maximum binding efficiency of 80% was achieved at an R_b/p of 120 in hWB ( see ESI Fig. S1). A cost-benefit analysis, based on minimizing the change in slope of the regression curve, indicated that an R_b/p of 120 (beginning of the plateau) was the optimal concentration for binding fungal cells in hWB. The optimal incubation period required for magnetic microbead binding to fungi in blood was determined using a 4 h kinetic assay. Although maximum binding (85 to 90%) occurred between 30 to 60 minutes in whole blood, approximately half (48 ± 4%) of all pathogens were bound within the first 5 minutes of incubation (see ESI Fig. S2). Thus, future devices may utilize the shorter incubation period but compensate by further multiplexing to maintain the same overall system efficiency to meet target clearance levels and process rates. Alternatively, a semi-batch process can be used where several small aliquots of blood can be drawn from a patient in stages that are staggered in time and premixed for continuous operation. Optimal binding conditions (1 h incubation at 37 °C and R_b/p of 120) were used for all subsequent fungal cell binding experiments in whole blood. Under these optimized conditions, an average of 28 ± 15 beads was bound per fungal cell.

Our design for the MMBCD contained channels that merged with a transverse velocity component (Fig. 5a); under these conditions the interface between the laminar streams of saline collection fluid and hWB rotated by 90 degrees when the streams joined, resulting in the two streams flowing partially side-by-side along the length of the channel rather than being stacked vertically (Fig. 5b). This rotation of the laminar flow interface was predicted for a water–water interface in computational fluid dynamic modeling;²⁸ however, the model also predicted that the laminar streams would reflect when the channels diverge at the outlets, so that separation of the blood and collection streams would be preserved. However, the rotated interface was not similarly reflected at the outlets in experiments that utilized hWB and PBS solutions, and this resulted in significant blood loss (50–60%) to the collection outlet, and comparable blood dilution at the blood outlet.

	Fig. 5 Transverse merging of microfluidic channels in MMBCD caused rotation of inlet fluid streams. (a) The separation channel within the prototype MMBCD has inlets that merge from top and bottom layers of PDMS. As laminar streams of fluids from the inlets merged, the transverse angle of the inlets caused the fluids to rotate about the longitudinal axis of the separation channel. The rotation of a water–water interface at the inlets and reflection at the outlets was predicted by computational fluid dynamics modeling. (b) However, the predicted reflected rotation at the outlets did not occur in experiments that used human whole blood (hWB) and phosphate buffer saline (PBS) and this resulted in significant blood loss (50–60%). Replacement of PBS with a 6% Dextran solution, which has a similar viscosity as hWB, did little to reduce blood loss. Blood loss was significantly reduced to 13% (p < 0.01) when the PBS to hWB flow ratio was increased by 4-fold (80 to 20 mL/h) in order to hydrodynamically confine the blood stream.

Attempts at matching the viscosity of the collection fluid and the apparent viscosity of blood, by replacing PBS with 6% Dextran solution (4 cp), did little to prevent blood loss. This is most likely because blood is a complex non-Newtonian fluid that exhibits both shear-thinning and viscoelastic properties and thus, its hydrodynamic behavior cannot be completely matched by a simple Newtonian fluid.^20,29,30 But we have found that it is possible to hydrodynamically confine blood to the bottom channel by increasing the flow rate and thus pressure of the collection fluid, as previously observed by others.³¹ Utilizing a PBS : blood flow ratio of 4, blood loss was significantly reduced to 13% (p < 0.01) with a similar level of dilution of the blood stream. Although the hematocrit level of diluted blood can be easily readjusted with conventional hemaconcentrators used in hemafiltration or renal dialysis machines, the increased saline flow may exert a momentum transfer into the blood phase, which can potentially impede the motion of magnetic particles and reduce separation efficiency.

The magnetic field strength (B) that is directly normal to the surface of the concentrator increased linearly from 80 to 272 mT as the voltage supplied to the electromagnet was increased from 10 to 40 V (Fig. 6). The field strength began to saturate with settings beyond 40 V and reached a maximum of 295 mT at 50 V. The corresponding electromagnetic gradient (B) varied from 50 T/m at 10 V up to a maximum of 200 T/m at 40 V. The efficiency of fungi separation utilizing a single-channel MMBCD was found to be tunable based on the voltage supplied to the electromagnet. Utilizing beads and fungi mixed in PBS, the separation efficiency increased linearly with the voltage up to 5 V, and began to plateau thereafter (see ESI Fig. S3). A 90% separation efficiency was achieved at the maximum setting of 15 V; however, magnetic particles and microaggregates (which are normally present in blood-bank blood prior to filtration) were found to accumulate within the microfluidic channels when separations were performed in hWB. This was because beads and bead-bound-fungi were too strongly magnetized and were pulled tightly against the channel walls, which served as nucleation sites for further accumulation; but decreasing the voltage to 10 V prevented this accumulation.

	Fig. 6 Tunable electromagnetic cell separation was initially tested with pathogens in PBS. Initial studies to test the strength and tenability of electromagnetic separation in microfluidic experiments were performed with C. albicans in PBS. Fungal cell separation was found to be broadly tunable from 0–90% based on a voltage setting of 0–15 V. This demonstrated that our electromagnet had sufficient strength and sensitivity necessary for the further optimization of cell separation in whole blood samples.

The magnetic force () exerted on a point-like magnetic dipole () is proportional to the gradient () of an external magnetic field ():^32,33

(1)

This equation can be simplified to calculate the total force on a number of beads (n) in one dimension (z) as:³⁴

(2)

Based on eqn (2), a 10 V setting that generates a magnetic field gradient of 50 T/m would be able to exert 50 pN of force on each 1 µm bead, which has a saturated magnetic moment (m_sat) of 10⁻¹³ A m².²⁷ Because a minimum of 10 magnetic beads were bound to each opsonized fungi (average = 28 ± 15 beads/cell), at least 500 pN can be exerted on each cell. With this force, opsonized pathogens (5–10 µm diameter) would require 0.3–0.6 seconds to traverse the 100 µm height of the blood channel. Because blood was flowing at a rate of 5 mL/h and has a residence time of 0.7 seconds in the MMBCD, a 10 V setting would provide sufficient time to pull magnetic particles from the blood stream. Future systems used in the clinical setting, however, should integrate inline low-capacity, but ultra-high efficiency, magnetic traps to pull out any remaining magnetic particles in the blood before it is reintroduced into a human patient.

It is important to note that there is minimal undesired diffusion of blood proteins, such as albumins from blood into the saline using the flow conditions described here. Given that HSA has an approximate diffusion coefficient (D) of 6 × 10⁻⁷ cm²/s,³⁵ the diffusion time scale (t = L²/2D) across the height of the blood channel (L = 0.01 cm) is approximately 84 s, which is much longer than the time the blood and saline fluids were in direct contact (i.e., residence time)..

Utilizing the 4-channel multiplexed MMBCD, it was possible to clear 80% of bead-bound-fungi from a 10 mL volume of hWB within 30 min (Fig. 7). Excess beads not bound to fungal cells were also removed with relatively high efficiency (82%). Accumulation of magnetic particles within the channels was not found under these conditions. These results demonstrate that a majority of fungi can be cleared upon a single pass through one 4-channel microfluidic system. The potential clinical value of this device is that multiple microfluidic cartridges can be positioned in parallel and in series, and it enables continual recirculation of contaminated blood, such that a patient's entire pathogen load can be further reduced over time through repeated passage through the device.

	Fig. 7 Multiplexed MMBCD achieved high throughput separation of pathogens from hWB. The multiplexed 4-channel MMBCD was capable of cleansing 80% of all bead-bound-fungi from a 10 mL volume of hWB at a rate of 20 mL/h. Excess beads were also removed at a relatively high efficiency (82%).

A simplified mathematical model, based on Monod kinetics,³⁶ to describe the systemic clearance of fungi in a recirculated system is defined by the differential equation:

(3)

Where C_p is the concentration of pathogen in the patient, t is time, µ is the specific growth rate of the pathogen (0.7 h⁻¹ under optimal conditions), n is the total number of microfluidic channels used in the multiplexed device, F is the blood flow rate per channel, is the separation efficiency per single pass, and V is the entire blood volume of the patient (500 mL in neonates). The systemic rate of pathogen clearance is thus determined by the rate of pathogen growth (1st term) minus the rate of magnetic separation (2nd term). Utilizing our current results, a multiplexed system that contains 100 channels will theoretically be able to reduce the total pathogen load by 90% in 3.6 h or 99% in 7.3 h. The respective clearance times can be further reduced to only 1.2 h and 2.3 h by doubling the number of channels.

Microfluidic devices have been used to separate molecules, particles and cells from small sample volumes at low flow rates (µL/h) for various analytical applications.^37–39 However, this technology is rarely exploited for therapeutic applications or for high throughput processes (mL/h) that utilize two fluid streams due to the challenges in multiplexing multiple pairs of channels while incorporating other separation features, such as magnetic field concentrators. We have demonstrated that this can be achieved by developing an array of vertically aligned channels that utilizes a separate magnetic field concentrator placed external to the microdevice. The advantage of this design is that it allows channels to be densely arrayed within each device. Additionally, in principle, multiple devices can be stacked with interposed magnetic field gradient concentrators, which ensure application of similar magnetic pulling forces across multiple interposed microfluidic systems, to achieve high throughput processing required for therapeutic applications. We successfully demonstrated proof-of-principle for this multiplexing concept by using a microfluidic device to cleanse 80% of living fungal pathogens from hWB flowing at a rate of 20 mL/h, which is 1,000 times faster than previously achieved in an earlier prototype microfluidic-micromagnetic separation device.²⁷ These results clearly demonstrate that the improved, multiplexed microfluidic–micromagnetic cell separation design provides much higher volume throughput while maintaining cell separation efficiencies, and thus, confirm its potential value for future clinical applications where it would provide a potentially exciting adjuvant to conventional antibiotic therapies.

Future design challenges include development of fully multiplexed arrays that can handle practical patient blood volumes, semi-batch mixing processes that will allow longer bead–pathogen incubation periods while maintaining continuous blood flow, as well as integration into conventional continuous veno–venous hemafiltration units, which have proven hemaconcentrators, blood warmers and oxygenation technologies. Furthermore, additional safety features such as ultra-high-efficiency magnetic traps must also be developed to capture excess beads before such a prototype can be used clinically.

Footnote

Electronic supplementary information (ESI) available: Fig. S1–S3. See DOI: 10.1039/b816986a