High-Performance Passive Plasma Separation on OSTE Pillar Forest

Abstract

Plasma separation is of high interest for lateral flow tests using whole blood as sample liquids. Here, we built a passive microfluidic device for plasma separation with high performance. This device was made by blood filtration membrane and off-stoichiometry thiol–ene (OSTE) pillar forest. OSTE pillar forest was fabricated by double replica moldings of a laser-cut polymethylmethacrylate (PMMA) mold, which has a uniform microstructure. This device utilized a filtration membrane to separate plasma from whole blood samples and used hydrophilic OSTE pillar forest as the capillary pump to propel the plasma. The device can be used to separate blood plasma with high purity for later use in lateral flow tests. The device can process 45 $\mathsf{\mu }$L of whole blood in 72 s and achieves a plasma separation yield as high as 60.0%. The protein recovery rate of separated plasma is 85.5%, which is on par with state-of-the-art technologies. This device can be further developed into lateral flow tests for biomarker detection in whole blood.

1. Introduction

Blood contains many biomarkers that are related to the health conditions of human body, from indexes of nutrients (such as glucose and vitamin) to cancer markers (such as prostate-specific antigen) [1,2,3]. Therefore, many point-of-care diagnostic platforms use whole blood as sample liquids for testing. Plasma separation is needed for many tests to avoid the disturbance of hemoglobin from red blood cells. Active and passive methods are used to filter blood cells from whole blood using microfluidic systems [4]. An external power source is needed for active plasma separation, such as pressure source and electrical field. Tripathi et al. built a device that utilized biophysical and geometrical effects to filter plasma in a continuous flow fashion [5]. A standalone 3D-printed microfluidic device was developed to filter red blood cells using the microfiltration method [6]. Inertial microfluidics can be used to separate plasma from whole blood sample in a high-throughput way [7,8]. Blood separation can also be achieved on microfluidic devices using acoustophoresis [9] and dielectrophoretic force [10]. Household tools or toys including egg beaters [11], smartphones [12], and paper centrifuges [13] have been developed to separate plasma from blood. To fit the application circumstance of point-of-care diagnostics including lateral flow tests, passive methods are often chosen to separate plasma from whole blood sample. Maria et al. has developed a microfluidic device using the sedimentation effect of red blood cells to separate plasma [14]. Plasma filtration membrane is used together with capillary channels to achieve hemolysis-free plasma separation [15]. Shamsi et al. utilized the Zweifach–Fung effect and collected plasma in flow branches with a higher flow rate [16]. Guo et al. built a blood filter using OSTE synthetic paper, which induced the local agglutination of red blood cells by precoating agglutination antibodies on its surface [17]. Baillargeon et al. made a paper microfluidic device by integrating a prefilter, a plasma separation membrane, and absorbent material to yield a high separation efficiency [18]. According to the working principle, passive plasma separation methods including sedimentation, filtration, and hydrodynamic effects have been reviewed in detail in Reference [19]. For lateral flow tests using whole blood as sample liquids, usually, there are two options for plasma separation: pretreatment by centrifuge and microfiltration by filtration membrane. By integrating a filtration membrane on the test strip, lateral flow tests can separate plasma on site and use the plasma for later analyses [20]. For traditional lateral flow tests working on blood, the filtration membrane is laminated on the lateral flow test substrate: nitrocellulose paper, which acts as a capillary pump to propel the separated plasma. However, nitrocellulose paper has some disadvantages as a lateral flow test substrate, including nonuniform microstructures and strong autofluorescence. In this work, we developed a new substrate material for such applications: OSTE pillar forest. Pillar forest has been extensively used as a substrate for capillary flow [21,22,23,24]. Due to the uniform capillary flow rates provided by pillar forest, it is often used as a lateral flow test substrate [25,26,27], which can help to generate uniform signals and reduce the variation of testing results. OSTE is a photocurable polymer that was developed for biochip fabrication [28,29,30]. OSTE pillar forest was fabricated by double replica moldings of a laser-cut PMMA mold, which has a uniform microstructure. Moreover, the polymer OSTE has easily tuned surface properties and low autofluorescence [31]. With free thiol groups, it is easy to immobilize proteins on OSTE surface by click chemistry, such as the protocols thiol–yne and thiol–maleimide [32,33]. Synthetic paper fabricated by OSTE has been used as a substrate for protein microarray, which couples fluorescent immunoassays on its surface via a thiol–maleimide reaction [31]. Compared with nitrocellulose and glass, synthetic paper can improve the performance of microarray immunoassays [31]. The characteristics of OSTE pillar forest can make up for the shortcomings of nitrocellulose. We built the passive microfluidic device by gluing a filtration membrane onto the OSTE pillar forest substrate, and the other surface of OSTE pillar forest was covered by a hydrophilic tape to enhance the capillary action. High-performance plasma separation can be achieved on this passive device. We tested blood samples with different hematocrit values, calculated the plasma separation yield, and checked the purity of separated plasma. In addition, we analyzed the variation of sample flow rates among different devices and different blood samples. Moreover, we calculated the protein recovery rate of the separated plasma and purification efficiency of our devices.

2. Materials and Methods

The blood filtration membrane was purchased from Cobetter (PSM0180-B, Hangzhou, China). Polydimethylsiloxane (PDMS, Sylgard 184) was from Dow Corning Corporation (Midland, MI, USA) and prepared by mixing base and curing agent in a weight ratio as 10:1. The hydrophilic tape (ARflow^® 93049) was provided by Adhesives Research (Shanghai, China). The free nail glue was purchased from Feifanli (Shenzhen, China). The whole blood samples were from adult pigs provided by a local food provider. The blood samples were immediately stored in a blood collection tube (BD Vacutainer^®, K2 EDTA 10.8 mg, 6 mL, REF: 367863, Franklin Lakes, NJ, USA) after they were collected from pigs, and then stored at 4 ${}^{\circ }$C. All of the experiments were conducted within three days after the collection of blood samples. We measured the hematocrit of blood samples by centrifuging EDTA-treated whole blood in a capillary tube.

For the fabrication of OSTE pillar forest (as shown in Figure 1a), at first, we used a laser cutter (Model: Leidayu, Thunderlaser, Dongguan, China) to engrave the PMMA plate and form microgrooves on the PMMA surface. The parameters of PMMA engraving were power, 25%; speed, 100 mm/s; laser frequency, 20 KHz; and number of passes, 1. The cutting pattern on the PMMA plate was orthogonal lines, shown in Figure 1b. At first, we cut the lines along the X axis and then the lines along the Y axis. The long axis of the OSTE pillar forest strip in this study was along the Y axis. We tested two pitch distances (p in Figure 1a,b): 350 $\mathsf{\mu }$m and 500 $\mathsf{\mu }$m. After we prepared the PMMA mold, we poured PDMS on the PMMA and cured PDMS. After that, we removed the PDMS negative from the PMMA mold and used it as a mold to make an OSTE replica [34], of which the details can be found in Appendix A.1. The curing of the polymer was performed by flood ultraviolet (UV) irradiation (UV radiation lamp, Asiga, Australia). Then, we removed the OSTE from PDMS and obtained a piece of OSTE pillar forest substrate, of which the average thickness was 944 $\mathsf{\mu }$m. We measured and observed the microstructures of OSTE pillar forest by step profiler (Dektak XT, Bruker, Billerica, MA, USA) and SEM (Gemini 300, Zeiss, Germany). Then, we cut the OSTE pillar forest piece as well as the filtration membrane piece to the desired shapes using the laser cutter. The shape design was by the software LaserMaker (Thunderlaser, Dongguan, China). The shape of the filtration membrane was a circle with a diameter of 12 mm and a rectangle with the dimension 2 × 3 mm, and the shape of OSTE pillar forest was a circle with a diameter of 12 mm and a rectangle with the dimension 40 × 3 mm, as shown in Figure 1b. After cutting, we performed the hydrophilic treatment of OSTE pillar forest using oxygen plasma (at 1000 mTorr for 8 min, Plasma Cleaner, PDC-002, Harrick Plasma, Ithaca, NY, USA). Then, we assembled the microfluidic device using filtration membrane, OSTE pillar forest, and hydrophilic tape. In detail, at first, we used the hydrophilic tape to cover the surface of the rectangle plasma channel; then, we coated some glue on the edge of round part of OSTE pillar forest and glued the filtration membrane onto it, as shown in Figure 2a and Appendix A.2. The pore size on one side of the filtration membrane is bigger than that on the other side, as shown in Figure 2b. The side with smaller pores is in contact with OSTE pillar forest, while the other side is in the upside for sample loading. There was a wedge gap formed in the interface between hydrophilic tape, filtration membrane, and OSTE pillar forest, which can help to enhance the capillary action [4]. During the experiments, we attached the test strips on a glass slide using double-sided tape.

After the preparation of test strips, we dropped the whole blood sample onto the center of the filtration membrane by a pipette and used a digital camera (Canon EOS RP, Japan) to record the experimental video for data analyses. The data analyses were perforemd by a Video Analysis and Modeling Tool Tracker (https://physlets.org/tracker/, accessed on 18 January 2021). After the experiments, we cut the plasma channels off the test strips and used a mini centrifuge to spin out the plasma sample [17]. Then, we checked the purity of the plasma sample by microscope (Leica DM IL LED Fluo, Leica, Wetzlar, Germany) and measured the protein concentration in the plasma using a bicinchoninic acid (BCA) assay (BL521A, Biosharp Life Sciences, Hefei, China), of which the details can be found in Appendix A.3. We compared the protein concentration in separated plasma with that in the plasma got by centrifugation and calculated the protein recovery rate. For each blood sample, every experiment was repeated six times.

3. Results and Discussion

The head of the OSTE pillar is a rectangle, as shown in Figure 1c,d. For the dimensions of OSTE pillars, with p value as 350 $\mathsf{\mu }$m, the average side length is 76 $\mathsf{\mu }$m, the distance between pillars is 274 $\mathsf{\mu }$m, and the depth of groove is 54 $\mathsf{\mu }$m. With p value as 500 $\mathsf{\mu }$m, the average side length is 120 $\mathsf{\mu }$m, the distance between pillars is 380 $\mathsf{\mu }$m, and the depth of groove is 32 $\mathsf{\mu }$m. As the intensity distribution of the laser beam follows a Gaussian distribution, the profile of groove between OSTE pillars also resembles a Gaussian curve [35], as shown in Appendix A.4. We characterized the flow behavior and capillary flow rate of water on these two different substrates (more details can be found in Appendix A.5) and chose the OSTE pillar forest fabricated with p = 350 $\mathsf{\mu }$m as the substrate material, which can provide a bigger volume capacity at 6.32 $\mathsf{\mu }$L/cm${}^2$. The fabrication process of OSTE pillar forest is fast. All of the steps including pattern design, PDMS molding, OSTE molding, laser cutting, and hydrophilic treatments can be performed in less than three hours, which can benefit the iteration of device design.

After the preparation of the devices (as shown in Figure 2c), we loaded the whole blood sample on the center of filtration membrane. At first, the whole blood sample penetrated the filtration membrane; then, the separated plasma flowed out from the other side of the filtration membrane and formed a capillary bridge between the filtration membrane and OSTE pillar forest substrate. The OSTE pillar forest acted as a capillary pump to constantly inspire the plasma until the plasma channel was filled. During the experiments (as shown in Figure 2d), the plasma filled OSTE pillar forest below the filtration membrane at first and then the rectangle plasma channel. To find the suitable blood volume that the devices can process, we flowed blood samples with different volumes onto the filtration membrane. The amount of blood should be large enough to generate plasma to fill the plasma channel but cannot leave too much excess blood on the filtration membrane. We tested blood samples with different volumes from different pigs and chose to use 45 $\mathsf{\mu }$L as the optimal sample volume. For the process rate of our devices, 45 $\mathsf{\mu }$L of whole blood can be processed within 72 s, which is much higher than many passive devices for plasma separation [20,36,37,38]. We obtained the plasma separation yield by calculating the ratio of plasma separated to the total plasma in whole blood. We used Equation (1) to calculate the plasma separation yield:

$$Separationyield=\frac{Volum{e}_{separatedplasma}}{Volum{e}_{wholeblood}\times (1-hematocrit)}.$$

(1)

Volume of separated plasma can be calculated by Equation (2):

$$Volum{e}_{separatedplasma}=Volumecapacity\times Are{a}_{pillarforest}.$$

(2)

The average plasma separation yield of our devices is 60.0% ± 8.1%. We investigated the flow behavior of plasma on the plasma channel by studying the relation between pump distance and flow time. When the plasma reached the rectangle channel after filling the round region below the filtration membrane, the capillary pressure kept constant and the fluidic resistance increased with the capillary pumping, the distance–time curves should follow the Washburn Equation [39,40]: dx/dt ∼ t${}^{0.5}$, which is used to describe the behavior of capillary flow in capillary tubes or porous media (x indicates the penetrating distance, and t is the time). We used the Washburn Equation for fitting of the experimental data, and the details can be found in Reference [41]. The average R${}^2$ of the fitting for all of the tests is larger than 0.987.

We investigated the variation from device to device by comparing the flow behaviors of same blood sample (with hematocrit as 30.4%) on different devices. As shown in Figure 3, there are six curves of distance versus time, which are from one blood sample tested on six devices. As we can see from this figure, different curves are very similar. The average flow rate on different devices is 0.198 ± 0.023 $\mathsf{\mu }$L/s, with a 11.6% variation. Furthermore, we investigated the variation from sample to sample by comparing the flow behavior of different blood samples (with hematocrit from 30.4% to 56.1%) on the devices. We tested 14 blood samples in total and obtained their average curves of distance versus time, shown in Figure 4a. The average flow rate of different blood samples is 0.142 ± 0.028 $\mathsf{\mu }$L/s. Noticeably, there is a rough trend that the average flow rate of plasma decreased with the increase in hematocrit value, as shown in Figure 4b. A higher hematocrit could lead to a larger fluidic resistance and, therefore, a smaller flow rate. It is possible to further develop this device into a microfluidic device to measure hematocrit using the total flow time as the index [42].

We used Equation (3) to calculate the protein recovery rate of our devices:

$$Proteinrecoveryrate=\frac{Proteinconc{.}_{separatedplasma}-Proteinconc{.}_{control}}{Proteinconc{.}_{centrifugedplasma}}.$$

(3)

$Proteinconc{.}_{separatedplasma}$ refers to the protein concentration in the separated plasma by our devices, $Proteinconc{.}_{centrifugedplasma}$ refers to the protein concentration in the plasma obtained by centrifugation, and $Proteinconc{.}_{control}$ refers to the protein concentration in the separated sample when we load phosphate buffered saline (PBS) at the same volume onto our devices. We chose four samples to calculate the protein recovery rate and found that the protein recovery rate was 85.5%, which is similar to that in previous reports [15,17,43]. Furthermore, we checked the purity of separated plasma and found that there were few red blood cells in it, which indicated that the plasma separated by our devices had a high purity with a purification efficiency as 99.0% (see the details in Appendix A.3) [14].

We performed a comparison of our device with other passive separation devices in recent reports, as shown in

Table 1. Performance comparison between our device and other passive devices for plasma separation.

Reference	Methods	Hematocrit	Separation Yield	Protein Recovery Rate	Volume/Time (Process Rate)
Gao et al. [43]	Membrane filter	45%	71.70%	82.3%	60 $\mathsf{\mu }$L/360 s
Samy et al. [45]	Wetting and Sedimentation	-	22–49%	-	-
Shamsi et al. [16]	Zweifach–Fung effect	-	66.6 %	-	-
Kuo et al. [46]	Fishbone filtration	45%	15%	-	10 $\mathsf{\mu }$L/75 s
Baillargeon et al. [18]	Membrane filter	30%	53.8%	-	-
Liu et al. [47]	3D Parylene filter	-	42%	-	2000 $\mathsf{\mu }$L/300 s
Son et al. [15]	Membrane filter	38%	20%	89%	-
Maria et al. [14]	Wetting and Sedimentation	45%	-	-	10 $\mathsf{\mu }$L/900 s
Our device	Membrane filter	30.4–56.1%	60.0%	85.5%	45 $\mathsf{\mu }$L/72 s

‘-’ means not provided.

. From this table, we can see that the overall performance of our device is on par with state-of-the-art technologies. Our device can process whole blood samples with a wider hematocrit value range in comparison with other devices. For separation yield and process rate, the performance of our device is above the medium level among the nine devices listed. The separation yield of our device ranks third, and the process rate ranks second among the devices of which relevant data are available. In terms of shelf life, our device has a shelf life of at least two months (details can be found in Appendix A.6). The cost of materials for a single device in this work is ∼0.23 USD (details in Appendix A.7), which fits well for point-of-care diagnostics.

Depending on specific applications, it is easy to scale up or down our device to fit the sample volume processed. Mainly, the sample volume that can be processed is limited by the area of the filtration membrane and the volume capacity of the substrate (OSTE pillar forest). According to information provided by the manufacturer of filtration membrane, the recommended blood volume is 35–45 $\mathsf{\mu }$L/cm${}^2$. Therefore, the blood sample volume should be less than the capacity of the filtration membrane:

$$Volum{e}_{wholeblood}<=45\mathsf{\mu }{\mathrm{L}/\mathrm{cm}}^2\times Are{a}_{filtrationmembrane}.$$

(4)

Next, the area of pillar forest can be determined using Equations (1) and (2).

By improving the manufacturing method, it is possible to mass produce our devices. Injection molding can be used to fabricate OSTE pillar forest with a higher efficiency [25,27,44]. As different parts of our device are assembled by gluing or taping, commercial laminating technology is also applicable to the assembly of our device, which enables large-scale industrial manufacturing.

4. Conclusions

In summary, we developed a facile way to fabricate a substrate for capillary flow: OSTE pillar forest, and used it together with a filtration membrane to make a passive microfluidic device for plasma separation. Our device can process whole blood samples in a fast rate as 45 $\mathsf{\mu }$L/72 s. It works for different blood samples with a hematocrit from 30.4% to 56.1%. The performance of our device on separation yield and protein recovery rate are also comparable with that in recent reports. Since our device separates the plasma from whole blood in a lateral flow test format, it can be immediately developed into a lateral flow test for biomarker detection in whole blood samples. The unique surface chemistry and optical properties of polymer OSTE make it easy to couple immunoassays on the OSTE pillar forest substrate.