Viscoelastic Separation and Concentration of Fungi from Blood for Highly Sensitive Molecular Diagnostics

Isolation and concentration of fungi in the blood improves sensitivity of the polymerase chain reaction (PCR) method to detect fungi in blood. This study demonstrates a sheathless, continuous separation and concentration method of candida cells using a viscoelastic fluid that enables rapid detection of rare candida cells by PCR analysis. To validate device performance using a viscoelastic fluid, flow characteristics of 2 μm particles were estimated at different flow rates. Additionally, a mixture of 2 μm and 13 μm particles was successfully separated based on size difference at 100 μl/min. Candida cells were successfully separated from the white blood cells (WBCs) with a separation efficiency of 99.1% and concentrated approximately 9.9-fold at the center outlet compared to the initial concentration (~2.5 × 107 cells/ml). Sequential 1st and 2nd concentration processes were used to increase the final number of candida cells to ~2.3 × 109 cells/ml, which was concentrated ~92-fold. Finally, despite the undetectable initial concentration of 101 CFU/ml, removal of WBCs and the additional buffer solution enabled the quantitative reverse transcription (RT)-PCR detection of candida cells after the 1st concentration (Ct = 31.43) and the 2nd concentration process (Ct = 29.30).


Results
Working principle. The characteristics of flow and particle migration in the viscoelastic flow are defined by non-dimensional numbers. The Reynolds number (Re) describes the ratio of the inertial force to the viscous force while the Weissenberg number (Wi) describes the ratio of the elastic force to the viscous force as follows: (2) c here, ρ, V m , D h , η c , λ, and γ c denote the solution density, mean flow velocity, hydraulic diameter of the particle, characteristic viscosity of the solution, fluid relaxation time, and characteristic shear rate, respectively. The particles suspended in the viscoelastic fluid experience the simultaneous effect of elastic and inertial lift forces. Therefore, the relative effect of fluid elasticity to inertia is estimated by using the elasticity number (El = Wi/Re).
www.nature.com/scientificreports www.nature.com/scientificreports/ The elastic lift force (F e ) is induced by non-uniform differences in the first normal stress (N 1 ) which create additional tension along the flow streamlines as follows 38 : Here, a, x, W, and Q denote the particle diameter, lateral distance, width of the microchannel, and volumetric flow rate, respectively. The elastic force drives the suspended particles to low shear rate regions at the centerline and the corners of the microchannel. Conversely, the inertial lift force also involves the lateral migration of particles in a viscoelastic fluid. The inertial lift force has two counteracting components, the shear-gradient lift force toward the channel walls (F i,s ), and the wall repulsion force toward the channel center (F i,w ). The shear-gradient lift force diminishes to zero near the corners of the microchannel and the wall repulsion force becomes dominant. The expression is as follows: i is iw , , 4 2 As shown in Eq. (3) and (4), the elastic lift force and the inertial lift force are substantially affected by the particle diameter.
A schematic of the device for sheathless viscoelastic particle focusing and separation is shown in Fig. 1. A sample mixture containing small candida cells and relatively larger WBCs was injected at the inlet and the cells were randomly distributed ( Fig. 1(a)). With respect to small cells (candida cells) relative to the channel size in the low-AR channel, inertial lift force drove cells away from the channel walls and the center while the elastic lift force drove the particles to the centerline of the channel. However, with respect to the cells with a relatively high blockage ratio (β = a/H, H denotes the channel height), elastic normal stresses drove the particles toward the channel walls, which was different from the cells with a low blockage ratio. Therefore, candida cells were tightly focused at the center of the microchannel while WBCs migrated toward the two equilibrium positions between the channel center and the side walls ( Fig. 1(b)). Finally, highly concentrated candida cells were collected at the center outlet (Outlet A) while WBCs were eliminated from the initial sample mixture toward the rear outlet (Outlet B).
Effect of viscoelasticity on flow characteristics. In order to examine the effect of viscoelasticity on flow characteristics of 2 μm fluorescent particles (β = 0.08), flow rate-dependent distributions of particles suspended in deionized (DI) water and 0.1% HA solution were observed over a wide range of flow rates from 20 μl/min (Re = 9.87, Wi = 2.77, El = 0.28) to 100 μl/min (Re = 49.38, Wi = 13.86, El = 0.28). Figure 2 shows the normalized fluorescence intensity in the expansion region during the flow of 2 μm particles. In DI water, 2 μm particles were randomly distributed in the microchannel over a wide range of flow rates from 20 μl/min to 100 μl/min. However, 2 μm particles suspended in 0.1% HA solution were focused into a single band along the channel center due to the elastic lift force toward the center. Candida cells were separated and concentrated at outlet A due to size-dependent viscoelastic separation while WBCs were removed at outlet B in a low aspect ratio microchannel.
www.nature.com/scientificreports www.nature.com/scientificreports/ Effect of the particle size on flow characteristics. Flow rate-dependent particle distributions in the expansion region were monitored with varying flow rates between 20 μl/min and 100 μl/min by using particles with diameters of 5 μm (β = 0.2) and 10 μm (β = 0.4) in 0.1% HA solution. Figure 3 shows that 5 μm particles (β = 0.2) were focused along the centerline while 10 μm particles (β = 0.4) were focused into two fluorescent streams at both sides of the centerline in the entire range of flow rates (20-100 μl/min). An increase in the flow rate increased the Wi, i.e., elastic force increased, and suspended particles were more tightly focused at the equilibrium positions. These results indicated significant consistency with previous studies in which off-center focusing into two streams was reported in numerical studies (β ≥ 0.25) and microfluidic experiments involving the use of viscoelastic fluid (β = 0.3 with El = 0.028 and 0.11). Flow rate-dependent distributions of smaller particles (1 μm diameter, β = 0.04) were also examined (Fig. S1).
The equilibrium position of particles was determined by the simultaneous effect of flow inertia and flow elasticity, shear thinning effect during the flow, and blockage ratio of each particle in the microchannel. The off-center focusing of large particles (β = 0.4) was attributed to the elastic lift force that acted on the side walls (F el,w ). The elastic lift toward the wall was due to fluid elasticity and shear thinning effects, and was strongly dependent on particle size. Therefore, the equilibrium positions of large particles were pushed toward side walls and off-centered. Our device can be applied to sheathless and label-free particle/cell separation based on the size-dependent equilibrium positions during the flow. Figure 4 shows the viscoelastic separation of binary mixtures of 2 μm (β = 0.08) and 13 μm particles (β = 0.52) in 0.1% HA solution at a fixed flow rate of 100 μl/min (Re = 49.38, Wi = 13.86). A stacked microscopic image at the channel inlet ( Fig. 4(a), left) shows a random distribution of both 2 μm and 13 μm particles. At the outlet expansion, three distinct streams were formed with the outer two streams of 13 μm particles and a single stream of 2 μm particles at the channel center. The outer two streams of 13 μm particles approximately corresponded to 1/4 of the channel width from the channel sidewalls and flowed to the side outlet channels. Furthermore, 2 μm particles were separated from larger particles and collected at outlet A. The microscopic images in Fig. 4(b) show the sample collected from each outlet. In outlet A, approximately 97.8% of 2 μm particles were collected with high purity (approximately 99%) while approximately 99.7% of 13 μm particles were collected from outlet B.

Separation of binary mixtures of particles.
Effect of viscosity of a lysed blood sample on separation performance. The separation of candida cells from WBCs was performed to examine the potential of our device for clinical application. The viscosity of a lysed blood sample with a hematocrit of 50 was known to be higher than that of blood plasma (~2.4 cP) 56 , which could affect the size-dependent lateral migration to the equilibrium position by the flow inertia and flow elasticity. The effect of the increased viscosity of the suspension was examined using viscous aqueous glycerol solution www.nature.com/scientificreports www.nature.com/scientificreports/ containing 2 and 13 μm particles (Fig. S2). In glycerol solution with the viscosity between 1 and 2.5 cP, the streamline of 2 and 13 μm particles was briefly almost similar to those in Figs 2 and 4. Separation of 2 and 13 μm particles was successfully achieved, even from a high viscosity suspension at ~2.5 cP. Therefore, we found that the viscosity of the lysed blood sample would not affect the device performance of separation and concentration of candida cells from WBCs in a lysed blood sample. However, for further use of lysed blood samples, possible biological effects of plasma proteins on the device performance should be considered. Figure 5(a) shows the microscopic images at the inlet and outlet of the microchannel during the separation process superimposed from bright-field images for candida and fluorescent images for WBCs at the fixed flow rate of 100 μl/min. At the inlet, a binary mixture containing both cells was randomly distributed in the microchannel. At the outlet, candida and WBCs exhibited separate streamlines flowing to different outlets. Candida was tightly focused at a single equilibrium position in the center of the microchannel due to relatively small sizes (approximate diameter of 3 μm, β = 0.12), while the equilibrium positions of WBCs were off-center shifted toward the side walls. According to the flow characteristics of particles in Figs 2 and 3, particles larger than 10 μm (β ≥ 0.4) were clearly removed at outlet B. The diameter of WBCs was 9-15 μm (0.36 ≤ β ≤ 0.6). Fluorescent streamlines of WBCs were approximately 1/3 of the channel width from the channel sidewalls, and this was similar to the equilibrium positions of 13 μm particles.

Separation of candida cells from WBCs.
Flow cytometric analysis was conducted to evaluate the separation performance of the device. Additionally, 10,000 events were counted with preset gates determined by each sample and cell distribution was achieved, as shown in Fig. 5(b). Prior to the separation, the binary mixture sample contained WBCs and candida. After the separation, candida was successfully extracted at outlet A and most WBCs were removed at outlet B. Separation efficiency is the ratio of the number of target particles at the target outlet to the total number of particles found at both outlets. In outlet A, 99.1% of the desired candida was collected from the total number flowing through both outlets while 96.1% of unwanted WBCs were removed at outlet B. The purity of candida collected at outlet A was approximately 97%, and this is defined as the ratio of the number of target particles to the total number of particles collected at outlet A. The results indicate that WBCs at outlet B exhibited slightly lower separation efficiency compared to that of 13 μm particles. This might be due to the heterogeneous size distribution and deformability of WBCs. However, a slight reduction in separation performance did not significantly affect the separation of candida cells, and this was verified by the PCR analysis. enhancement of concentration factor. During the separation process, candida could also be concentrated by removing additional suspending medium at outlet B. To enhance the concentration of candida cells www.nature.com/scientificreports www.nature.com/scientificreports/ suspended in a lysed blood sample, the suction flow rate at outlet A could be controlled to remove additional buffer solution at outlet B. The flow rate factor (FF) is the ratio of the inlet flow rate to the outlet flow rate at outlet A of the trifurcation outlet region 57 . With respect to the separation process, the inlet flow rate was fixed at 100 μl/min while the outlet flow rate at outlet A was determined by the ratio of the channel width at the trifurcation region. In this study, the widths of the outlet channels were designed as 300 μm, 100 μm, and 300 μm, respectively, such that the initial flow rate factor was determined to be 7. In order to enable further manipulation of the flow rate factor to optimize the concentration performance of our device, the suction flow rate at outlet A was controlled by a syringe pump (KDS210, KD Scientific). To show the effect of the suction flow rate on the flow characteristics at the outlet trifurcation, the numerical simulation was conducted (Fig. S3). Figure 6 shows the concentration performance based on different flow rate factors. The suction flow rates at the center outlet were determined based on the simulation results in Fig. S3. As shown in Fig. 6(a), center-focused candida flowed to outlet A at a suction flow rate of 14 μl/min (FF = 7), which was equal to the designed ratio. When the FF was increased to 10 (suction flow rate = 10 μl/min), all the candida was still collected at outlet A. However, when FF was further increased to 17.5 (suction flow rate = 5.7 μl/min), a few of the center-focused candida could not be recovered at outlet A and deflected into the side channels. Figure 6(b) shows the flow rate factor-dependent concentration factor, which is defined as the ratio of cell concentration of the sample collected at outlet A to the initial cell concentration at the inlet. The concentration of candida depending on different flow rate factors is determined by the suction flow rate from outlet A. An increase in the flow rate factor from 7 (suction flow rate = 14 μl/min) to 10 (suction flow rate = 10 μl/min) increased the concentration factor to approximately 9.9. The concentration factor continued to exceed 9.5 when the flow rate factor increased to 11.7. However, as the flow rate factor increased further (FF = 14.2 and 17.5), the concentration factor decreased to 7.5 and 6.1, because a certain amount of candida flowed to outlet B. Therefore, the optimal flow rate factor was 10 in this study with respect to the concentration of candida. The concentration of the fluorescent-dyed candida was 2.5 × 10 7 cells/ml in the inlet sample prior to the separation process. After separation, the amount of fluorescent candida was 2.4 × 10 8 cells/ml in the sample collected from outlet A. The sample collected after the 1st concentration was used in our device for the 2nd concentration process. The final concentration of candida was increased to 2.3 × 10 9 cells/ml, and this corresponded to a 92-fold increase in the concentration. For PCR analysis, the initial sample volume of 500 μl was reduced to the final volume ~5 μl at the optimal flow rate of 100 μl/min and the flow rate factor of 10, which was processed in ~6 min. In our previous www.nature.com/scientificreports www.nature.com/scientificreports/ work on malaria parasite separation using the two-stage device 14 , malaria parasites were separated from WBCs with a 94% separation efficiency, 99% purity, and 7-fold concentration increase at 100 μl/min. In comparison, candida cells were separated with 99.1% separation efficiency and 97% purity, and concentrated by 9.9-fold in one-time device processing at the same flow rate (100 μl/min).

Validation of the device performance.
In order to evaluate the capability of the device to detect C. albicans, SYBR green quantitative RT-PCR and conventional PCR analyses were conducted using an undetectable concentration (10 1 CFU/ml, Table S1) of candida cells in blood. As shown in Fig. 7(a,b), C. albicans was barely detected due to the low candida cell numbers and the high number of contaminants, such as WBCs (Ct = 35), prior to the separation. Outlet A1 and outlet B1 indicated the center and side outlets from the 1st process in our device while outlet A2 and outlet B2 indicated outlets in the 2nd concentration process. After the separation and the concentration process, Ct values for the samples collected at outlet A1 and outlet A2 were measured as 31.43 and 29.30, respectively. The samples collected at outlets B1 and B2 and a negative control (Human serum gDNA) were not detected in the PCR condition (Ct = 35). Figure 7(B) shows the results from agarose gel electrophoresis of C. albicans samples collected from the device. The amplicon band (approximately 273 bp) of outlet A2 exhibited a stronger band relative to that of outlet A1. This was due to the concentration of candida cells by depleting WBCs and removing the additional buffer solution at outlet B. Furthermore, when the candida cells in the blood were  www.nature.com/scientificreports www.nature.com/scientificreports/ concentrated without leukocyte removal, 10 X blood samples showed lower amplification than 1 X blood samples in PCR, despite the high concentrations of candida cells (Fig. S4).

Discussion
As shown in Figs 2 and 3, particles with different sizes migrated toward each equilibrium position over a wide range of flow rates from 20 μl/min to 100 μl/min. These results implied the flow rate-insensitive nature of particle patterning in the low-AR channel. Therefore, size-dependent particle migration to the equilibrium positions and particle separation was achieved by non-powered manual operation and negated the external flow injection system with accurate flow rate control.
Compared to our previous works 14,49,50 , simultaneous separation and concentration of candida cells were successfully achieved in the device used in this study. Previously developed devices enabled sensitive size-dependent separation. However, it was difficult to concentrate cells by removing additional buffer solution based on the slight difference between lateral displacements. In addition, in the previous devices, platelets in blood samples could be activated by wall shear stress during initialization of all cells at the 1st bifurcation. However, in the device used in this study, it is assumed that shear-induced platelet activation will not be detected because there is no cell initialization process along the channel walls.
In our device, HA solution was chosen as the viscoelastic fluid with low viscosity (0.9 mPa·s) to achieve the high-throughput processing, compare to PEO solutions used in previous reports (2.3 mPa·s 51,52,54 and 4 mPa·s 53 ). Due to its cost effectiveness, it can be helpful to use PEO solution for the high-throughput processing in our device. However, the effect of molecular structure of PEO and HA solution on the viscoelastic flow has not been examined yet. Therefore, as further study, it is worth examining the flow characteristics using PEO solution by modulating the polymer concentration and the flow conditions.
The capability of high-throughput processing was examined by increasing the flow rate further to 500 μl/min by 100 μl/min interval for candida separation. The throughput of our device can be increased up to 400 μl/min with high separation efficiency (η = 98.8 ± 0.6%), which is due to flow-rate insensitivity. This was constrained by the burst of unstable tubing interconnection at the inlet at 500 μl/min 58 . However, according to the previous report, it can be confirmed that the viscoelastic focusing could be achieved at higher flow rate (~20 ml/min) using a rigid chip made of epoxy resin 59 . The microchannel used in previous research was with the width and the height of 80 μm, and the length of 35 mm, which had ~37-fold smaller flow resistance compared to our device. Therefore, our low-AR channel device can be fabricated in rigid thermoplastic to address the present limitation for device mass production during commercialization process. Moreover, the device throughput can be further enhanced by stacking or multiplexing the devices in parallel [60][61][62] , since our device is comprised of a straight channel with a single inlet without introducing any sheath flows.
As with platelets, there are also cells and residues of various sizes in biological samples such as blood, urine, and the cerebrospinal fluid. Based on the results shown in Figs 2, 3 and 4, particles with various sizes laterally migrated toward the size-dependent equilibrium positions. Particles/cells with the blockage ratio of 0.08 ≤ β ≤ 0.2 were tightly focused at the center of the microchannel with the aspect ratio of 0.5, while particles/cells with the blockage ratio of β ≥ 0.4 were patterned into two separate streams. Therefore, the capability of our device can be further expanded to manipulate cells and residues of different sizes, for example, separation and simultaneous concentration of malaria parasites (1.5-2 μm), bacteria (0.5-2 μm), circulating tumor cells (15-25 μm), etc., from WBCs.

Materials and Methods
Device design and fabrication. The microfluidic channel consisted of an inlet and two outlets. The width and the height of the main channel were 50 μm and 25 μm, respectively, such that the aspect ratio (AR) of the channel was defined as 1/2 (AR = height/width). The length of the main channel was 4 cm. The width of the expansion outlet region was 700 μm to visualize the flow streams of cells.
A polydimethylsiloxane (PDMS) microfluidic channel was fabricated using a soft lithography technique. A replica mold was fabricated using an SU-8 negative photoresist (MicroChem, Newton, MA) on a silicon wafer. The PDMS base and curing agent were mixed at a ratio of 10:1 (Sylgard 184, Dow Corning, USA), degassed in a vacuum chamber, and thermally cured in an oven for 1 h at 80 °C. Subsequently, the cured PDMS channels were peeled off from the replica mold, cut, and bonded on a glass slide with oxygen plasma (CUTE, Femto Science, South Korea). To minimize non-specific protein binding, surface modification was applied to the PDMS channel walls using Triton X-100 surfactant 63 . sample preparation. As a viscoelastic non-Newtonian fluid, hyaluronic acid (HA) sodium salt (357 kDa, Lifecore Biomedical) in phosphate-buffered saline (PBS) at 0.1 (w/v)% was prepared. The zero-shear viscosity and relaxation time of 0.1% HA solution were reported as 0. 9 mPa s and 0.26 ms, respectively 59 .
Fluorescent polystyrene particles with 1 μm (green, G0100, ThermoFisher), 2 μm (blue, B0200, ThermoFisher), 5 μm (green, G0500, ThermoFisher), 10 μm (green, G1000, ThermoFisher), and 13 μm (red, 36-4B, ThermoFisher) diameters were used to examine flow characteristics and verify device performance prior to their application to the candida samples. The particle diameters were selected by considering the sizes of target cells. The particles were suspended in 0.1% HA solution at a final concentration of approximately 2.1 × 10 7 particles/ml. Fresh human whole blood with an anticoagulant, ethylenediaminetetraacetic acid (EDTA), was obtained from healthy volunteers at Korea University Guro Hospital (Seoul, Korea) with informed consent from all subjects. Human whole blood samples were obtained from Korea University Guro Hospital complied with safety regulations through an Institutional Review Board (IRB) approved collection method (2017GR0769, approved by Korea University Guro Hospital). The hematocrit of the whole blood sample was ~48%, which was measured using a Hematospin (Hanil Scientific Inc. Korea). C. albicans SC5413 was provided by Dr. Jeong-Yoon Kim at the (2019) 9:3067 | https://doi.org/10.1038/s41598-019-39175-5 www.nature.com/scientificreports www.nature.com/scientificreports/ Department of Microbiology & Molecular Biology, Chungnam University, Korea. Yeast were cultured overnight at 30 °C in 10 ml of yeast extract-peptone-dextrose (YPD) broth (Qbiogene), and the cultured cells were then quantified with phase-contrast microscopy (40x power) by using a counting grid. For the application of candida separation, 100 μl of whole blood was mixed with 700 μl 1x BD FACS lysing solution (BD Biosciences), 100 μl of 1x SYBR Green and 100 μl of 1% HA solution. Therefore, the final concentration of HA solution was 0.1%. To reduce the effect of cell-cell interaction which can degrade the separation efficiency, RBCs in whole blood sample were required to be lysed. The concentration of WBCs was approximately 5 × 10 6 cells/ml. C. albicans was spiked at approximately 2.5 × 10 7 cells/ml, which is a relatively high concentration compared to those in clinical cases, to examine the possibility of clinical applications.
Experimental procedure. The sample solution was infused into a microchannel by using a syringe pump (LSP01-1A, Longer Precision Pump). During the experiment, flow of particles and cells in the microchannel were observed by an inverted microscope (CKX41, Olympus) with a high-speed camera (V611, Phantom) and a fluorescent camera (CS230B, Olympus). In order to evaluate the separation performance of the device, the inlet sample and each sample from two outlets were quantitatively analyzed using a flow cytometer (Accuri C6, BD Bioscience, CA). SYBR Green quantitative RT-PCR and conventional PCR. DNA was extracted from whole blood as described in previous reports with minor modifications 64 . Erythrocytes were briefly lysed in Red Cell Lysis Buffer (Invitrogen, USA) for 10 min at 37 °C. After centrifugation at 3,000 rpm for 10 min, the pellets were treated with 200 μl of 1 M Sorbitol with 5 U/μl of Zymolyase (Invitrogen) at 37 °C for 30 min. DNA from each sample was extracted using the QIAmp DNA Mini Kit (Qiagen) following manufacturer instructions. Real-time (RT) PCR was performed using a SYBR Green Kit (Bio-Rad). For C. albicans quantification, primers specific to the ITS1-ITS2 region of C. albicans were used: forward: TTTATCAACTTGTCACACCAGA and reverse: ATCCCGCCTTACCACTACCG 65 . The qPCR reactive mixture contained 2 μl of gDNA, 10 μl of iQ SYBR Green Supermix (2x), 1 μl of forward primer (5 pmol/μl), 1 μl of reverse primer (5 pmol/μl), and RNase-free water to a total volume of 20 μl. RT-PCR was performed using the CFX-96 instrument (BioRad). Cycling conditions were as follows: initial denaturation at 94 °C for 3 min followed by 35 cycles of denaturation at 95 °C for 20 s, and annealing and elongation at 60 °C for 40 s. For conventional PCR, the PCR reactive mixture contained 2 μl of gDNA, 10 μl of iQ Multiplex Powermix (2x, Bio-Rad), 1 μl of forward primer (5 pmol/μl), 1 μl of reverse primer (5 pmol/μl), and RNase-free water to a total volume of 20 μl. Thermal cycling parameters were the same as the above RT-PCR condition.

Conclusion
In summary, we described a novel continuous cell separation and concentration device by using a viscoelastic fluid to detect extremely rare candida cells. Polystyrene particles with different blockage ratios were used to estimate the flow-rate and size-dependent flow characteristics using the viscoelastic fluid in a slit microchannel. Particles with a blockage ratio lower than 0.08 (β < 0.08) were distributed in the microchannel due to the low blockage ratio. Particles 2 μm and 5 μm in diameter were tightly focused at the center of the microchannel, while particles larger than 10 μm were patterned into two fluorescent streams. Therefore, 2 μm and 13 μm particles were successfully separated at 100 μl/min with high efficiency (approximately 97.8%). The optimized condition was used to separate candida cells with a 99.1% separation efficiency and 97% purity at a flow rate of 100 μl/min. The suction flow rate at outlet A was controlled to achieve the maximum concentration factor of 9.9 with a flow rate factor of 10. Finally, undetectable candida cells at an extremely low concentration (10 1 CFU/ml) became detectable using the RT-PCR analysis by separation and sequential concentration process. Therefore, the device enables continuous separation and concentration processes for the pre-treatment of extremely rare disease-related cells to improve detection sensitivity.

Data Availability
All data generated or analyzed during this study are included in this published article (and its Supplementary  Information files).