Simplifying Microfluidic Separation Devices towards Field-Detection of Blood Parasites

The device consists of three sections, each with parameters optimized to carry out that section’s specific task. Section 1 is deep and has a large critical size in order to sort out WBCs. Section 2 has a lower critical size and is shallow, which maximizes the effective size of the particles making it easy to laterally displace them along the channel walls. Section 3 has the same array parameters as section 2 but the height has been decreased. The height has been optimized to take advantage of the shape difference between erythrocytes and the longer and slender parasites. At the specific height of section 3 (9.0 μm) the parasites will continue to be laterally displaced, this time however towards the centre. The erythrocytes, on the other hand, will appear with a smaller effective size and switch from displacement mode to zig-zag.


Introduction
For many tropical diseases, identifying parasites or infected blood cells among a huge background of normal blood cells is the gold standard for diagnosis. 1Particle separation techniques such as acoustophoresis, 2 margination 3 and inertial focusing 4,5 have been developed that show great promise as candidates for the eld diagnosis of tropical diseases, however the gap between how microuidics devices are run in labs and how they need to be run in the eld is still considerable.In well-equipped labs there are pressure pumps, expensive pressure control units or syringe pumps plus advanced microscopes to monitor device function, giving information that is in turn used to make changes to ow rates and pressures.In the eld however the lack of equipment, and in many cases maybe even access to a power grid, means that devices designed to work well in the lab, are doomed to failure.One approach would be to develop cheaper, less power consuming ow-control and optical systems.Another approach is to develop systems that require as little control as possible, which can perform complex operations with minimum requirements on external control.Paperbased microuidics 6 is the epitome of the second approach, where all functions are designed into the device and the only external control required is the application of a droplet of sample followed by a simple visual readout.While these systems are promising for many applications, they are not wellsuited to high-resolution particle separations for which device structures oen need to be much more well-dened on the microscopic scale than the random brous structure of paper.
Here we describe a device that is designed to perform complex, high-resolution size and shape-based particle separation without expensive uidic control equipment.The separation mechanism in our device is deterministic lateral displacement (DLD). 7The basic concept is to use simple and cheap pressure generation (a disposable syringe where air is compressed to give an over-pressure), simple sample input (one inlet only for the entire device), and sequential particle sorting units based on DLD to decomplex a mixture of particles.While applicable to any system containing particles that differ in size and/or shape we focus here on the separation of parasites from the genus Trypanosoma from blood cells.We have previously shown how trypanosomes and erythrocytes exhibit different behaviours in DLD devices as the device depth is changed, a fact that we previously used to perform separations. 8In the present work we use multiple array parameters, including depth, to realize complex functions in our device while simultaneously simplifying uid control and use the decomplexing of a mixture of human blood and trypanosomes as a test system.
Endemic to sub-Saharan Africa, human African trypanosomiasis (HAT) is a fatal vector-borne parasitic disease caused by the protozoan parasite Trypanosoma brucei, transmitted by tsetse ies. 9Its diagnosis relies to a great extent on tedious and laborious microscopic detection in blood smears or lymph node aspirates. 10The key bottleneck with conventional microscopic diagnosis of HAT is the relatively low number of parasites against the overwhelming background of blood cells.Consequently the detection threshold for normal microscopic diagnosis is relatively high ($10 000 ml À1 ) 11 in comparison to the parasitaemia (parasite number density in blood).This varies periodically during the course of the disease but can at times be as low as 100 ml À1 , and consequently, there could be more than 50 million blood cells (predominantly erythrocytes) per parasite. 11Today, expensive and power-consuming technologies, such as ion exchange chromatography and centrifugation, are available at most modern facilities, and can be used for relatively efficient parasite detection (via enrichment relative to blood cell concentration).However, it is challenging to employ these tools in the geographical areas where the disease is prevalent, due to the lack of reliable power and storage facilities.Our aim is to use a microuidic sorting scheme to extract the parasites for visual inspection, thus addressing the limitations of current technologies.

Deterministic lateral displacement
Our device, Fig. 1, can be fabricated in cheap materials, has no moving parts or power requirements and being no larger than a microscope slide, is highly portable.It is based on Deterministic Lateral Displacement (DLD), a method rst shown in 2004 by Huang et al. that in its simplest conguration separates particles based on size. 7In brief, DLD uses an array of obstacles in a uidic channel.Particles smaller than a critical size follow the overall ow direction, while larger particles are deected at an angle relative to the ow, dened by the array.][14][15][16] The method has been shown capable of size-fractionating a wide range of samples including blood components, 17,18 cancer cells 19 and synthetic particles such as droplets in two-phase ow. 20A recent review 21 by Bridle et al. gives an expansive overview of this particle separation technique.
By controlling and exploiting shear forces our group recently added deformability 22 to the list of parameters by which particles can be separated using DLD devices.Also, by controlling the orientation of particles as they pass through the device, we have been able to separate particles based on shape.As a proof of principle we showed that channel depth could be used to control the orientation of particles, and used this effect to accentuate the differences in shape of parasites and human blood cells 8 making otherwise very difficult separations possible.
The work presented herein builds on these ndings and involves a complete redesign of the device.The outcome is a more robust, easier to use device, with a signicantly increased throughput.Our method provides a dramatically simpler and cheaper alternative to existing methods with the potential of enabling a fast and cheap point-of-care device that will have signicant impact on eld diagnosis.

Device design
In designing our device we prioritized simplicity and ease of use.We require the device to handle blood with low concentrations of parasites, with as little dilution as possible, and output close to all parasites in a sample stream free of blood cells.We designed a device with one inlet only, so that ow can be driven simply using a disposable syringe.To avoid pressure control systems that are both expensive and demanding with regards to user input, all functionality must be designed into the device: (1) Removal of leukocytes in order to avoid clogging in subsequent steps, (2) creation of cell-free plasma, and (3) transfer of parasites into the cell-free plasma.The functionality in each section of the device comes from a combination of the array spacing parameters and the depth of the channel (height of the posts).The nal device consists of three DLD arrays in series, each with its own depth, optimized to carry out these three tasks, Fig. 3A-B.
Section (1) is designed to function like a traditional sizebased DLD with a depth larger than the diameter of all blood cells.As we have previously shown, 8 in deep devices particles rotate due to shear forces and are consequently sorted according to their smallest dimension.Here, the spherical leukocytes, being in the range 8-15 mm, are displaced and removed from the sample via a side channel.Erythrocytes and trypanosomes, on the other hand, have a smallest dimension of $2.5 mm which is below the critical size in this section.As a consequence, they are not laterally displaced and mainly follow the ow to the subsequent section.
Section (2) is shallow (3.5 mm) to minimize the rotational freedom of the cells and consequently maximize their effective size.The array is designed so that all cells are laterally displaced towards the channel walls creating a cell-free stream in the centre.A mirrored design 23 is employed to minimize the migration distance of cells to the nearest wall which results in a higher throughput.
Section (3) has the same post size, post gap and row shi as section (2) but is deeper, (9 mm from previous optimization 8 ).Here, the fluid is actuated by compressing air with a simple syringe.When the desired pressure is reached, the valve is closed allowing for a stable pressure throughout the measurement.The sample is injected into the single inlet reservoir prior to attaching the pressure system.Following an analysis, the sorted particles (erythrocytes, leucocytes and T. cyclops parasites) can be visualized in, or collected from the outlets for further processing.
The row shi direction is also mirrored such that displaced particles are focused into the centre of the channel.The key idea is that the depth of this section is sufficient for erythrocytes to rotate and behave like 2.5 mm particles, as decided by their thickness, and they consequently remain near the walls.At the same time the longer parasites are still partially conned and behave as effectively larger particles being displaced back into the cell-free plasma and into a separate outlet where a dense array of posts (8 mm in diameter with a separation of 2 mm while keeping the array wide, 4 mm, in order to minimize the effect of clogged particles) serves as a sieving structure to capture and concentrate the cells of interest once they have been sorted.The parasites can thereaer be detected (future work will focus on detection schemes for the enriched parasites).

Simple pump system
The single inlet design not only simplies the sample handling and introduction to the device, but it also allows for a device which does not require advanced and expensive pressure controllers that would otherwise be necessary to ne-tune and balance the pressures applied to the different inlets.Instead, here a simple approach is employed where the air in a syringe is compressed in a controlled way, thereby resulting in a well-dened overpressure (see Fig. 4).The relatively large volume of air in the system compared to the small volume of uid handled in the device, together with well-sealed connections ensures that the overpressure is maintained at a stable level over the course of the experiment.The sample was applied to the inlet reservoir, which was subsequently connected to the syringe to drive the sample through the device.

Method
To maximize the throughput, the device has been kept as wide as possible.This means however that, together with the relatively small angle of displacement, the device needs to be long.In fact, the nal device is around 14 cm.To t this onto a standard glass slide the device is split up into its three different sections with multiple channels connecting the end of each array with the start of the proceeding array, see Fig. 5A-B.To ensure that the particles maintain their relative lateral position while moving between the different sections, careful calculations and CFD simulations (COMSOL Multiphysics 4.3, Comsol AB, Stockholm, Sweden) have been conducted, the key being to keep a constant resistance across the width of the device such that the ow and particle proles are preserved around the 180 bends.More info on this optimization can be found in ESI section (3).†

Experimental setup
All experiments were performed in compliance with the relevant laws and institutional guidelines.Blood was extracted via nger pricking (Haemolance, MedCore AB, Kista, Sweden) of healthy volunteers from which informed consent was obtained.Prior to each experiment, the devices were ushed with sterile ltered Cunningham's medium 26 with 20% fetal calf serum (Cat.No. F2442, Sigma-Aldrich, St. Louis, MO, USA), the medium used to culture the parasites, and 2 mM EDTA (Cat.No. E6758, Sigma-Aldrich) to inhibit coagulation of the blood.It is important to note that Cunningham's medium is only necessary to sustain our model system.
In order to obtain good statistics on the device performance, measurements were oen conducted with the same device for several hours.Consequently, to counteract particle sedimentation and ensure a homogenous sample, a small magnetic stir bar (length 5 mm, diameter 2 mm) was placed in the inlet reservoir and controlled by a magnet connected to a small electric motor.With a stir rate of $1 Hz and the dimensions of the stir bar and inlet reservoir as given above, the shear rate is estimated to be on the order of 10 s À1 to 100 s À1 , which is less than the physiological shear rates of up to $10 4 s À1 . 27,28Further, microscopic examination conrmed that the stir bar did not induce any morphological changes in the sample.For actual eld use of the nal device, the short time of analysis would eliminate the need for any stirring.

Measurements and analysis
The lateral distribution of cells was characterized at the end of each section in the device through an inverted Nikon Eclipse TE2000-U microscope (Nikon Corporation, Tokyo, Japan) using an EMCCD camera (iXon EM+ DU-897, Andor Technology Ltd, Belfast, UK) and various objective lenses (Nikon, 10/20x Plan Fluor and 60x Apo TIRF).Differential Interference Contrast (DIC) was used to ensure adequate contrast of blood cells and trypanosomes.
Particle tracking soware, based on available MATLAB code, was written in MATLAB R2014b (The MathWorks Inc, Natick, MA, USA) and optimized for accurate and efficient particle recognition.Detailed information together with the code can be found in ESI section (4).† Blood and trypanosomes were analysed at concentrations, which allowed for automated particle tracking while the addition of anticoagulants opened up for measurements over several hours per device without any blood clotting affecting the ow behaviour.
For all measurements, the morphologically similar but much less dangerous (and less restricted) parasite T. cyclops was used instead of the human parasites T. brucei rhodesiense or T. brucei gambiense. 8lood and parasites were initially analysed separately in order to better characterize how the output of the device varied with ow rate.Subsequently, with the optimum ow rate determined, spiked samples were separated.For the initial samples where blood was run separately, it was diluted 10Â in autoMACS Running Buffer (Miltenyi Biotec GmbH, Bergisch Gladbach, Germany).This is an isotonic PBS solution containing 2 mM EDTA, 0.5% BSA and 0.09% azide.In addition to preventing blood clotting, a lower concentration decreases the particle-particle interaction.For eld-diagnostics we envision the collection of blood in pre-lled vials (e.g.BD Vacutainer® 366450, BD Biosciences, Franklin Lakes, NJ, USA).For the spiked blood samples, the blood was washed twice in Cunningham's buffer to remove the lytic factors causing natural lysis of T. cyclops, which is not adapted to human physiology (this step would not be necessary for the T. brucei).The 10Â diluted blood sample was then spiked with trypanosomes at a concentration of approximately 1000 ml À1 .This is in most cases a higher concentration than what occurs in normal rare cell detection.However, it allows us to accurately determine the device performance in terms of sorting efficiency.Detailed protocols are available in the ESI section (7).†

Multi-layered device characterization
During UV-lithographic fabrication of the master, the different sections of varying depths could be precisely aligned to each other with an estimated positioning error less than 3 mm.This is small enough to have a negligible effect on the ow, and consequently, the relative lateral position of the particles in the device.Fig. 5A-E shows SEM micrographs of a nalized continuous device in PDMS consisting of multiple depths.Here, three of the main features of the device are shown.First, the channels of equal resistance connecting the subsequent arrays can be seen in Fig. 5A-B.Due to the complexity of the device with multiple arrays in series it will be longer than a normal DLD device.Consequently, to t the entire device on a normal glass slide the device design is folded, forming a serpentine.In order to maintain the relative particle positions between sections, multiple channels with identical uidic resistances are used to connect subsequent sections.To accomplish this, the shorter channels are designed to be narrower than the longer ones.Between section (1) and ( 2), there are in total 16 channels while 28 channels connect section (2) and (3).A detailed description of these channels can be found in ESI section (3).† In Fig. 5C-D the sieving structure in the outlet is shown.This feature has been implemented to capture sorted cells of interest and to focus them to a smaller area allowing for easier detection.In Fig. 5E-G the two step changes in height of the device are shown.The rst of these two SEM micrographs Fig. 4 (A-B) Schematic overview of the device connected with a simple syringe that was used in all measurements to induce fluid flow.A manometer was initially used to relate the syringe compression with the resulting overpressure, ensuring that the correct pressure was achieved.Thereafter a valve is closed resulting in a consistent pressure over longer times.shows how the deep channels in section (1) transition to the shallow channels in section (2).The other micrograph shows the end of the same section but here the transition is from the shallow section (2) to the intermediate section (3).Fig. 5G further shows the transition from deep to shallow channel depth, here visualized aer being lled with 1 mg ml À1 rhodamine B (#83689, Sigma) in H 2 O and imaged using confocal microscopy.

Using a simple syringe to control ow rate
Compressing the syringe leads to a pressure drop across the device acting to drive the sample, which prior to this has been loaded into the sample inlet.Having a stable pressure is required as ow rate affects the sorting of deformable cells, as shown previously, allowing for sensitive measurements to take place over a longer time.Here we succeeded in achieving high pressure stability, Fig. 4C, by utilizing a large dead volume ($12 ml) together with sealed connections.At an initially applied overpressure of 1 bar across the device the resulting pressure decreased less than 3% over 90 minutes.The volumetric throughput is $390 ml during this time, which according to Boyle's law explains the pressure decrease, indicating little to no leakages in the system.
The results from the calibration of the ow speed in relation to the syringe compression are given in Fig. 4D.Here, it can be seen that the hydrodynamic resistance of the device decreases with increasing pressure, which is expected due to the elastic PDMS device becoming deformed, and inated, at high pressures.Consequently, the ow rate is not linearly proportional to the applied driving pressure, instead it varies between 2.5 and 3.8 ml (min bar) À1 .The volumetric ow rate here is given as the sample input rate.From this, approximately 20% of the total ow is separated out to outlet 1, enriched with leukocytes, while the sorting between erythrocytes and trypanosomes takes place in the remaining volume.

Erythrocytes and leukocytes at various ow rates
The behaviour of so, deformable cells in a DLD device depends to a great extent on the ow rates.Increasing the ow rates leads to a larger force exerted on the cells by the uid acting to deform them.Consequently their size and, as a result, the sorting in the device might change.Further, cells in channels deeper than their size can rotate freely under the inuence of the shear eld of the uid, which as shown previously, acts to make the cells present their smallest dimension to the posts.An erythrocyte would due to this effect have an effective size the same as its thickness, $2.5 mm instead of its diameter, $7.5 mm.However, at low ow rates the shear stress exerted by the uid becomes smaller, which acts to decrease the rotational effect.Another effect of higher ow rates is the increased pressure drop, which acts to inate the deformable PDMS devices, something which in turn leads to an increase of the critical size.The erythrocytes and parasites were initially measured separately at various ow rates to allow for automated particle tracking of a large number of cells.Thereaer, spiked blood samples were analysed at the most promising ow rates.
The results from the separate measurements of erythrocytes and T. cyclops are presented in Fig. 6, with the sorting efficiency further analysed in Fig. 7.
In Fig. 7 the output of each section is given for both erythrocytes and T. cyclops at several different ow rates as a fraction of the total number of that cell type in that specic section.It can be seen that in section (1), Fig. 6B, the fraction of cells separated out are larger at low ow rates but converges to $20% with increasing ow rates (and consequently $80% of the cells continue to section ( 2)).Section ( 1) is deeper than the largest dimensions of both cell types, the length of the T. cyclops and the diameter of the erythrocytes.As a result the cells can rotate freely as discussed previously, resulting in a decreased effective size.With the section designed with a critical size of 7.1 mm, the cells are not expected to be laterally displaced.However, at low ow rates the shear stress exerted on the particles by the uid is limited which leads to a decrease in both the orientation of the cells and also the cell deformation.As a consequence, the fraction of cells sorted out into outlet 1 can be higher than the expected 20%.
In section (2), Fig. 6C, the cells are steered towards the sides of the channel to open up a cell-free stream in the centre.Due to the low depth of this section, the cells are restricted from rotating which leads them to display a larger effective size and consequently they are more easily laterally displaced into the side streams along the channel walls.Here, the side streams are dened as the outermost 25% on both sides of the channel while the centre stream constitutes the remaining 50%.The size of these streams is dened by how the end of section ( 3) is divided into outlet 2 and 3, meaning that any cells residing in the centre stream at the end of section ( 2) are likely to also be sorted into outlet 3.In the same way, any cells residing in the side streams will end up in outlet 2 unless they are laterally displaced in section (3).
At lower ow rates, around 95% of cells entering section (2) will at the end reside in the side streams.However, at high ow rates the so erythrocytes will, by deforming due to shear forces, avoid being laterally displaced and will end up as a contaminant among the parasites in outlet 3.At 3.8 ml min À1 around 20% of the erythrocytes exiting this section do so in the centre stream and are expected to end up as a contamination in outlet 3, leading to poor separation sensitivity.
Section (3) is identical to section (2) except for the direction of displacement and the depth, Fig. 6D.By the increase in depth to 9.0 mm, larger than the diameter of the erythrocytes, there is no steric hindrance to their orientation, leading to a decrease in their effective size as previously discussed.They are therefore not laterally displaced in this section and, if focused into the side streams in section (2), are sorted into outlet 2. Similar to the situation in section (1), it can be seen that if the ow rate is too low, the shear forces are not high enough to allow the cells to rotate and adopt the smaller effective size.Instead they have a maximum effective size equal to their diameter leading to a signicant lateral displacement.At low ow rates (0.2 ml min À1 ) 12% of the erythrocytes entering this section are sorted into outlet 3 even though they exit section (2) well-focused in the side streams.Consequently, the erythrocyte contamination in outlet 3 is here a result of unwanted lateral displacement in section ( 3).An increase of the ow rate leads to a decrease in the fraction of erythrocytes sorted out into outlet 3, however at too high pressure, the fraction once again increases.This is however believed to be an effect of erythrocytes not being well focused into the side streams when they exit section (2).This could potentially be avoided by increasing the width of section (2), leading to a decrease in the ow velocity (while maintaining the volumetric ow rate) and consequently the shear stress exerted on cells in this section.
For the T. cyclops the highest fraction of extracted cells is acquired at low pressures while the fraction decreases with increasing pressures.This shows how adjusting the depth of the channel can have a large impact on the behaviour of the cells, allowing for separation based on their differences in shape.
The measurements of T. cyclops and erythrocytes carried out separately at various ow rates are summarized in Fig. 7. Firstly, the fraction of the two cells types which ends up in outlet 3 is shown, here calculated based on the input sample.This gure is similar to Fig. 6D, but the cells removed in section (1) have also been taken into account.The most notable difference is the decreased capture rate of T. cyclops at low ow rates which is due to the increase in fraction of T. cyclops being sorted out into outlet 1.As a result, the separation efficiency is decreased at these ow rates.The largest enrichment (32 times) is reached at 1.9 ml min À1 .At this ow rate the fraction of erythrocytes entering outlet 3 is kept at a minimum (2.1%) while a large fraction of the trypanosomes are retained (66.5%).As a consequence, this was chosen as the ow rate for all the remaining measurements.However, it can be noted that even though the enrichment is the highest at this ow rate, the number of collected T. cyclops per minute increases with increasing ow rate.At a ow rate of 3.8 ml min À1 the number of collected T. cyclops is almost twice that of 1.9 ml min À1 .The optimum tradeoff between throughput and purity will be determined by the choice of detection scheme in the specic application.

Blood samples spiked with T. cyclops
As discussed, a ow rate of 1.9 ml min À1 was chosen for the continued studies of spiked blood samples in order to receive the maximum enrichment of T. cyclops.In these measurements, the sorting of leukocytes was also considered in order to show the versatility of the device, and its potential for applications other than erythrocyte and T. cyclops separation.
On the whole, the outcome, given in Fig. 8, agrees with the previous measurements.For the erythrocytes 21.7% are diverted to outlet 1 (leukocyte outlet).For the remaining erythrocytes which continue into section (2) the fraction successfully focused into the side stream amounts to 97.6%.From this number, a small fraction (0.6%) is not retained along the sides of the channel in section (3) but instead end up in outlet 3 (T.cyclops outlet) as a contamination.The remaining 97.0% which still remain in the side streams exit into the correct outlet which is close to the result of the separate measurements.
For the trypanosomes 18.3% is laterally displaced into outlet 1 while 92.8% of the remaining parasites are successfully focused into the side streams.This is, in agreement with previous measurements a smaller fraction than for the erythrocytes.But as the T. cyclops should exit into outlet 3 from the centre of section (3), this would lead to an increased sorting efficiency.In the end, 83.1% of the remaining T. cyclops exit into outlet 3, equivalent to a 30-fold enrichment in the number concentration.The leukocytes are, contrary to the other cells, laterally displaced to a great extent in section (1).The majority of them, 94.1%, are sorted out into outlet 1 while the remaining fraction continues to the shallower section (2).A large fraction of these leucocytes are not able to enter this shallow section, however over the time scale of our experiments the number of leucocytes ltered out was too small to adversely affect the device performance.This is believed to be an effect of the device's robustness to pressure changes, with section (2) being aimed at focusing all cells towards the wall.The result of this is that any smaller pressure variations will not have an effect on the device performance.
The total output from the device is summarized in Fig. 9, here we can see that 100% of the extracted leukocytes are found in outlet 1 owing to the large size and shape difference between leukocytes and the other cell types.There is a relatively large number of background cells present, around 20% of each cell type, as expected due to the design of section (1).The addition of a focusing section, similar to section (2), positioned before the sorting-out of leukocytes takes place, the number of background cells in outlet 1 could be greatly reduced.In outlet 2 the fraction of erythrocytes amounts to 76.0% (77.4%), while 67.9% (66.5%) of the T. cyclops are retained in outlet 3 (results of the earlier measurements on individual cell types are given in parenthesis).As can be seen, these fractions are in close agreement with the earlier measurements with a marginally smaller number of erythrocytes retained in outlet 2 while a slightly larger fraction of T. cyclops can be steered into outlet 3. It should however be noted that these numbers rely on the bifurcation of the ow into the outlets.The current device was designed to have, as described earlier, half the ow at the end of section (3) guided to outlet 2 and the other half to outlet 3. Changing these fractions would potentially lead to a better performance of the device.For example, decreasing the fraction of the ow in section (3) bifurcated into outlet 3 would remove more erythrocyte and would likely increase the specicity of the device.

Conclusions
We have demonstrated a proof-of-principle of a device that, while simple-to-use, can perform complex separations such as the fractionation of parasites, leukocytes and erythrocytes.Our method utilizes several steps along the device where differences in morphology and size between the particles are given different weights by varying the depth of the DLD array.The principle of using multiple successive arrays to focus and then separate particles could be applied to any of the particle parameters (or combinations thereof) that have been shown to affect behaviour in DLD devices such as size, shape, deformability and dielectric properties.What is more, the device functions without expensive pressure controllers and can be run using pressurized air from a single disposable syringe.
We believe the approach that we have presented here shows great promise as a point-of-care test for a multitude of diseases.In addition to our test system with African trypanosomiasis other protozoan parasites could be considered such as, but not limited to, leishmaniasis, Chagas disease and malaria, where particle size, shape and deformability can be leveraged for the enrichment of hard to nd objects.These diseases affect millions worldwide, not only via direct infection of human populations but also via the socioeconomic burden of livestock infection. 29hile we believe the current device offers several advantages over current methods there is still work to be done towards a fully functioning eld-ready diagnostic device.Most notably lacking is portable detection.One solution may be the devices developed by Ozcan et al. 30 that have shown great promise for eld-detection of pathogens combined with telemedicine.Another would be integrated electronic cell counting based on impedance spectroscopy. 31urther, the current experiments were performed in devices made using so lithography.This is a rapid prototyping technique that is suitable for research, but for eld applications large numbers of devices would be needed at lower cost.Foilbased uidics 32,33 and injection moulding approaches would provide the necessary basis for mass-produced devices and be fully compatible with our method.View Article Online

Fig. 1
Fig.1Complete device consisting of a syringe (I), a pressure valve (II), sample inlet (III) and outlets for sorted particles (IV).Here, the fluid is actuated by compressing air with a simple syringe.When the desired pressure is reached, the valve is closed allowing for a stable pressure throughout the measurement.The sample is injected into the single inlet reservoir prior to attaching the pressure system.Following an analysis, the sorted particles (erythrocytes, leucocytes and T. cyclops parasites) can be visualized in, or collected from the outlets for further processing.

Fig. 2 (
Fig. 2 (A) Schematic illustration of size-based sorting in DLD devices.Particles smaller than the critical size follow the direction of fluid flow while particles larger than the critical size follow rows of posts.(B-C) Non-spherical particles (such as erythrocytes) behave very differently when confined in shallow devices compared to in deep devices where they are able to rotate.In a shallow device (B) erythrocytes have an effective size corresponding to their diameter.In a deep device (C) they are instead sorted according to their much smaller thickness.
allowed to rest for at least 20 min before ushing with deionized water for another 20 min.Silicon inlets and outlets 12 mm and 5 mm outer diameter silicon tubing (228-0725 and 228-0707, VWR International LLC, Radnor, PA, USA) were glued (Elastosil A07, Wacker Chemie AG, Munich, Germany) onto the device.

Fig. 3 (
Fig. 3 (A) Basic principle of how orientation changes the effective size of particles of different shape.(B) Schematic overview with the three different sections colour coded.Arrows represent the direction of displacement in each section.(C) False-coloured mosaic image consisting of 34 time-integrated micrographs showing the trajectories of erythrocytes and the parasites through the device.
Fig.4 (A-B) Schematic overview of the device connected with a simple syringe that was used in all measurements to induce fluid flow.A manometer was initially used to relate the syringe compression with the resulting overpressure, ensuring that the correct pressure was achieved.Thereafter a valve is closed resulting in a consistent pressure over longer times.(C) Applied pressure measured over 90 minutes given as fraction of the initial pressure.(D) Throughput in terms of sample fed into the system measured as a function of the pressure difference across the device.

Fig. 5 (
Fig.5 (A-F) SEM and (G) confocal images of the device.(A-B) The channels connecting the different sections of the device are designed to be of equal fluidic resistance, and therefore their individual widths are chosen to compensate for the varying lengths.Due to these channels, the flow symmetry is maintained and the relative positions of the particles are conserved between subsequent sections allowing for longer and more complex devices to be created.By precise alignment of the complementary photolithographic masks in three separate UV-lithographic cycles, sections of different channel depth could be fabricated into a single continuous device, (E-F).After the final section in the device, section (3), the trypanosomes have been focused to the centre and thus exit to the trypanosome outlet where a dense array of pillars (C-D) is defined with the purpose of capturing the parasites.Note the inset (D) where a parasite is highlighted (coloured green), scale bar is 10 mm.(G) Confocal micrograph of a fluorescent solution (0.1% rhodamine B in H 2 O) passing through the step-change in depth between section (1) and (2), same as (E).The grey scale is mapped to the fluorescence signal.

Fig. 6
Fig. 6 Sorting efficiencies as a function of flowrate (i.e. the fraction of cells in section (1) that is sorted out into outlet 1, the fraction in section (2) which is focused along the channel walls and the fraction in section (3) which are sorted out into outlet 3).The numbers are given as the fraction of that cell type in that specific section.(A) Schematic overview of the device highlighting the origin of the data in the graphs.(B) After the first section the erythrocytes and parasites are sorted out in proportion to the fluid flow going into the outlet.(C) In section (2) most erythrocytes and parasites are focused to the side.(D) For the separation of T. cyclops and erythrocytes intermediate flow rates give the highest separation efficiency.Each data point represents a minimum of 1000 counts.For the highest flow rates the number of counts was around 7000 for both erythrocytes and T. cyclops.

Fig. 7
Fig. 7 Overall sorting performance versus applied pressure.(A) Fraction of erythrocytes and T. cyclops which are sorted into outlet 3. The numbers are given as the fraction of each cell type injected into the device.(B) Concentration enrichment of T. cyclops (ratio between the number of T. cyclops and the total number of particles) and the rate of trypanosomes arriving at the collection reservoir.

Fig. 8
Fig.8Distribution of the three different cell types at the end of each section.The numbers are given as the fraction of that cell type in that section.

Fig. 9
Fig.9The output of the device given as the fractional distribution of the three measured cell types.