A. Conventional protocol

The extraction of RNA from whole blood requires several biochemical steps, including lysis of cells, separation of DNA and RNA, isolation of RNA, and buffer exchanges. Since the protocols are lengthy and often complex to understand, we designed a visual way to describe the protocol used here in Fig. 1, where all steps and conditions are described in three parts, I, II, and III. In Part I, the red blood cells are lysed, and the white blood cells are isolated following the protocol for Millipore Sigma's Red Blood Cell Lysis Buffer (RLB). About 1.1 × 107 white blood cells (∼5500 per μl) were theoretically expected in this step.2121. M. S. Blumenreich, The White Blood Cell and Differential Count (Butterworths, 1990). Once the white blood cells were isolated, reagents from the RNA extraction kit are used to lyse the white blood cells, and β-mercaptoethanol is used to stabilize the RNA. At this point the white blood cell lysate can be stored at −20 °C or used immediately for extraction. In Part II, the RNA and DNA in the white blood cell lysate are bound to magnetic beads using a binding buffer, and the solution is constantly mixed for 5 min via pipetting up and down. The supernatant is removed, and similar mix steps are done for the wash steps. The solution is then degraded by a DNase I solution for 20 min at room temperature. This step removes any DNA contamination, which is a prerequisite for any RNA-based expression profiling. From there, in Part III, the RNA is re-bound to the magnetic beads using a binding buffer, and the RNA–magnetic bead complex is washed. The RNA is eluted using an elution buffer. About 3.2–16 μg of RNA was expected from 2 ml of blood,2222. See https://www.qiagen.com/us/resources/faq?id=06a192c2-e72d-42e8-9b40-3171e1eb4cb8&lang=en for “How much RNA does a typical mammalian cell contain?—QIAGEN.” which is a concentration of 16–80 ng/μl in 200 μl of elution buffer. This protocol takes approximately 3 h to complete for one sample.Following the method in Fig. 1, three human whole blood samples were tested, using disodium EDTA or sodium heparin as anticoagulants. These samples were stored for a variety of times after the blood was drawn: 2, 5, and 8 days, contributing to the standard deviation of the yield measurements. Disodium EDTA and sodium heparin anticoagulated blood yielded average RNA concentrations of 25 and 20 ng/μl, respectively (Table I). The average RNA integrity numbers (RINs) and average Cq values were 4.3 and 30.06, respectively, for disodium EDTA anticoagulated blood. The average RNA integrity numbers (RINs) and average Cq values were 6.7 and 29.01, respectively, for sodium heparin anticoagulated blood. There was no statistically significant difference between the values measured for disodium EDTA vs sodium heparin. In comparison to other methods, RNA integrity for the conventional protocol is similar; the concentration of the samples tested here is 2–3× fold higher than other methods.2323. D. Schwochow, L. E. K. Serieys, R. K. Wayne, and O. Thalmann, “Efficient recovery of whole blood RNA—A comparison of commercial RNA extraction protocols for high-throughput applications in wildlife species,” BMC Biotechnol. 12, 33 (2012). https://doi.org/10.1186/1472-6750-12-33

TABLE I. The conventional protocol was performed on three disodium EDTA and sodium heparin anticoagulated human whole blood. Using a t-test assuming unequal variance, the results of average concentration, average RIN, and average Cq value for three samples were compared between disodium EDTA and sodium heparin anticoagulants. There was no significant difference (p > 0.05) between any of the aforementioned values between the two anticoagulants. The volume of the reagents in the conventional protocol was scaled down by a factor of 22, and a separate experiment mixed the two wash buffers 1:1 to reduce the number of wash steps. Using the conventional protocol as a reference group, all measured values of the adjusted protocols were compared to that of the reference group and were not statistically significant. Note: For the “1/22 scale down” concentration values in the elution buffer—This is not the actual yield of the 1/22 scale down protocol, but for comparison purposes, the measured value was multiplied by 22.

AnticoagulantProtocolAverage RNA yield (ng/μl elution buffer) ± standard deviationAverage actual yield (ng/μl blood)Expected yield (ng/μl blood)Average RIN ± standard deviationAverage Cq value ± standard deviationDisodium EDTAConventional protocol25 ± 7.212.54.3 ± 3.030.06 ± 4.431/22 scale down54.05 ± 24.420.543.4 ± 0.529.60 ± 0.851:1 Wash buffer mix26.22 ± 4.002.61.6–84.7 ± 1.926.49 ± 1.16Sodium heparinConventional protocol20 ± 5.6726.7 ± 1.629.01 ± 2.971/22 scale down36.66 ± 15.020.373.6 ± 0.730.25 ± 1.541:1 Wash buffer mix31.83 ± 14.903.24.6 ± 0.626.78 ± 0.95

B. Scale down of Part I of the conventional protocol

Reducing the starting volume of the blood or the white blood cell lysate and scaling down the subsequent reagents linearly was essential for future microfluidic protocol development. Part I of the conventional protocol (Fig. 1) involves centrifuging 2 ml of blood and lysing the red blood cells to isolate the white blood cells. Thus, two methods of reducing the volume of the reagents in these steps were investigated. The first was reducing the starting volume of blood (from 2 ml to 500, 250, 100, or 50 μl) and the second was reducing the starting volume of the isolated white blood cells, lysis buffer, and β-mercaptoethanol (from 625 to 156, 78, 39, or 28.4 μl). The rest of the reagents were scaled down linearly with the starting sample. At the two highest volume reductions, reducing the starting volume of blood resulted in a lower concentration and RIN (measured using the Agilent Bioanalyzer 2100) than by reducing the volume of the white blood cell mixture. However, at the two lowest volume reductions, the concentration and RIN were similar between the two methods (data not shown).

We decided to move forward with the second method to sample from the white blood cell lysate from Part I of the protocolto reduce the sample preparation time for future experiments. Using this method, the starting volume of blood is always 2 ml and multiple small aliquots (28.4 μl) of the white blood cell lysate can be used for multiple experiments and analyses. When using the first method we investigated, many individual replicates of the white blood cell lysate would have needed to be prepared, while the second requires storing one large solution of white blood cell lysate that can be sampled from multiple times. Either method would be sufficient for reducing the volume of the conventional protocol. However, when isolating RNA from samples with lower white blood cell counts, the first method using lower volumes of blood may not have sufficient yield of less concentrated RNA.

C. Scale down of Parts II and III of the conventional protocol and combining wash buffers

The lowest volume tested used 1/22 of the 625 μl white blood cell lysate mixture, which is a starting volume of 28.4 μl after performing Part I of the conventional protocol. Therefore, the volumes of the reagents in Parts II and III of the conventional protocol were scaled down by a factor of 22. The scaled-down elution volume of 9 μl was multiplied by 2.2 for a final 20 μl elution volume to ensure sufficient volume for pipetting and downstream analyses. These experiments were done with five-day old blood. Additionally, in a separate experiment, to investigate the possibility of reducing the number of wash steps, wash buffers No. 3 and No. 4 were mixed 1:1. These experiments were done with two-day old blood, with two common anticoagulants: disodium EDTA and sodium heparin. We used two types of anticoagulated blood for investigational purposes, not for comparison. The results for average concentration (in elution buffer), average RIN, and average Cq are shown in Table I. The results for the altered protocols were compared to the conventional protocol, and no statistically significant difference was found between them. The actual yield per μl of blood for the scale downs is lower than the expected yield, which could be due to the age of the blood and lower volumes being more sensitive to measurements since the scaled-down concentration was found using the RNA Pico Kit. However, since the Cq and RIN values were comparable to the conventional protocol, we concluded that reducing the volume of reagents and combining the wash buffers into one wash buffer is acceptable to translate onto the microfluidic chip.

D. The microfluidic separator

Previous work in our group led to the development of a microfluidic device that integrates magnetic separation and electrokinetic purification (MSEP) of DNA.15–1715. K. Lee and A. Tripathi, “Parallel DNA extraction from whole blood for rapid sample generation in genetic epidemiological studies,” Front. Genet. 11, 374 (2020). https://doi.org/10.3389/fgene.2020.0037417. L. Schneider, F. Cui, and A. Tripathi, “Isolation of target DNA using synergistic magnetic bead transport and electrokinetic flow,” Biomicrofluidics 15(2), 024104 (2021). https://doi.org/10.1063/5.0045307 These microfluidic devices were developed to simplify the washing and purification steps of the DNA extraction process and are made of a PDMS mold bonded to a glass slide. The device operates by first placing a wash solution into the primary wash buffer loading well, which flows throughout the wash channel to the secondary wash buffer loading well [green, Figs. 2(a) and 2(b)]. Then, the sample, which contains nucleic acids, is pipetted into the sample well simultaneously with the elution well solution (a wash buffer or an elution buffer) [yellow, Fig. 2(c)]. Our prior work used a purified DNA sample1717. L. Schneider, F. Cui, and A. Tripathi, “Isolation of target DNA using synergistic magnetic bead transport and electrokinetic flow,” Biomicrofluidics 15(2), 024104 (2021). https://doi.org/10.1063/5.0045307 or diluted whole blood as the starting sample.1515. K. Lee and A. Tripathi, “Parallel DNA extraction from whole blood for rapid sample generation in genetic epidemiological studies,” Front. Genet. 11, 374 (2020). https://doi.org/10.3389/fgene.2020.00374 A magnet is then placed beneath the separator to transport the magnetic beads through the wash solution in the Wash Channel, washing the beads. This movement is repeated until all beads have been transported to the elution well, and the solution from the elution well is removed from the microfluidic separator simulatenous with the sample well's contents, which are discarded [Figs. 2(d) and 2(e)]. The elution well's contents are then used for downstream processing. The addition and removal of the sample well and elution wells' contents must be done simultaneously such that they do not flow into the elution well. Multiple separators are designed to be on one microfluidic chip, with six in parallel and in two rows (for a total of 12 separators), with each separator being 9 mm apart such that multiple purifications can be performed simultaneously with a multi-channel pipettor and a long magnet.

For RNA purification, which is investigated here, several changes were made to the microfluidic separator design used by Lee et al. First, the number of beads used in the RNA extraction protocol compared to the DNA extraction protocol used by Lee et al. is 4.25× higher, so the width of the narrowest section of the wash channel and the width of the entrance point of the sample and elution wells were increased to be 4.25× larger. This helped reduce the possibility of the magnetic beads getting stuck and unable to move through the separator and to keep the time to move the magnetic beads through the entire separator similar to that of the DNA extraction protocol. Next, the width of the entrance to the elution well was made to be the same as that of the sample well (instead of smaller) to ensure the magnetic beads could efficiently enter the elution well. The secondary wash buffer loading channel was made to be twice the width of the primary one because with a previous iteration of this RNA microfluidic separator, the wash solution would not flow from the primary wash buffer loading well to the secondary wash buffer loading well but would stop once the wash channel was filled. To reduce the hydrodynamic resistance in the secondary wash buffer loading well, the width of the channel was doubled, which did cause the wash channel and wells to be filled as desired. Finally, the height of the sample and elution wells (and subsequently the wash buffer loading wells) was increased via increasing the depth of the aluminum mold used to make the microfluidic chip, such that 100 μl of solution could fit into the Sample and Elution Wells.

E. Investigation of a reduced number of wash steps and hands-on time from the conventional protocol

Parts II and III of the conventional protocol have ∼25 min of hands-on washing steps each. The microfluidic separator was designed to simplify washing of magnetic beads; therefore, we investigated how the number of steps could be reduced for microfluidic purification. The blood samples here were from two different 2 ml aliquots from the same donor and were two days old. The controls (following the 1/22 scale down protocol) had an average Cq value of 26.88 for disodium EDTA anticoagulated blood and 28.08 for sodium heparin anticoagulated blood. There was no statistically significant difference between the two average Cq values of each anticoagulant.

To first test a reduction of the number of wash steps, all the wash steps were removed from the 1/22-scaled-down protocol (termed “No washes”), and the resulting products' Cq values were measured. Without the wash steps, the average Cq value for disodium EDTA anticoagulated blood was 34.05, and that of sodium heparin anticoagulated blood was 34.90 (Fig. 3), which were both higher than their respective control protocols. Additionally, out of six disodium EDTA replicates, only three amplified; for sodium heparin, two out of the six amplified. The inconsistency in amplification via RT-qPCR indicates that removing all the wash steps is not repeatable and that some of the unwashed species inhibit PCR.

We hypothesized that DNase I was essential to being washed off the magnetic beads for successful qPCR. To test this, the 1/22-scaled-down protocol was performed but without the wash steps after the DNase digestion (termed “No wash 2”). This trial was performed with the 1:1 mixture of wash buffers in place of two separate wash steps. The disodium EDTA samples had an average Cq value of 33.86, and the sodium heparin anticoagulated samples had an average Cq value of 37.61, which were statistically significantly higher than their respective controls. Four of the six disodium EDTA samples amplified, and one of the six sodium heparin samples amplified, leading to a similar conclusion to that of the “No washes” samples, where skipping the wash steps after the DNase I digestion are not repeatable.

Finally, to further reduce the number of hands-on steps for the microfluidic protocol, the need for the 5-min constant mixing after every reagent addition was investigated. For this experiment, the 1/22-scaled-down protocol was performed without any mixing where the conventional protocol indicates to do so. Instead, the sample was incubated for 5 min at room temperature during the indicated mixing times. For the disodium EDTA anticoagulated blood, the average Cq value was 33.14, and that of sodium heparin was 29.54. There was no statistically significant difference between no mixing with sodium heparin anticoagulated blood and the positive control; there was a statistically significant difference between no mixing with disodium EDTA anticoagulated blood and the positive control. Sodium heparin and disodium EDTA have both been shown to affect PCR.2424. J. F. Huggett et al., “Differential susceptibility of PCR reactions to inhibitors: An important and unrecognised phenomenon,” BMC Res. Notes 1, 70 (2008). https://doi.org/10.1186/1756-0500-1-70 Disodium EDTA chelates Mg2+ ions, while sodium heparin is thought to directly interact with nucleic acids.2525. C. Schrader, A. Schielke, L. Ellerbroek, and R. Johne, “PCR inhibitors—Occurrence, properties and removal,” J. Appl. Microbiol. 113(5), 1014–1026 (2012). https://doi.org/10.1111/j.1365-2672.2012.05384.x This could explain the more significant difference between the sodium heparin protocol variations and their corresponding control since RNA amplification may be more affected by the presence of heparin in the absence of washes.

To summarize, all of the conventional protocol variations tested showed that (1) the scaled-down volume of 1/22 and combining the wash steps into one 1:1 mixture had comparable Cq values and RINs to the conventional protocol, (2) washing of the RNA–magnetic bead complex after the DNase digestion was an essential step for repeatable results, and (3) the manual mixing steps could be replaced with room temperature incubations. DNase I is a PCR inhibitor since it degrades DNA, and this shows that DNase I must be sufficiently washed away for the resulting RNA sample to be RT-qPCR amplifiable. Taken together, these results allowed us to reduce the starting volume of the white blood cell lysate, reduce the volumes of other reagents used, combine the wash buffers, and reduce the amount of hands-on time needed by the researcher to perform RNA extraction. These are significant steps toward simplifying the RNA extraction process, and reducing the volume of reagents implies a reduction in cost to do RNA extraction. A detailed analysis into conventional protocol reduction has rarely been described in the literature, which is important for not only microfluidic translation but also in understanding essential steps in RNA extraction and purification.

F. The microfluidic separator—Solution loading procedure for RNA purification

We first tested the microfluidic separator to replace the wash steps in Part II. The starting sample at this stage contained white blood cell lysate, DNA–RNA-bound magnetic beads and binding buffer, and the 1:1 mix of the wash buffers was in the wash channel and elution well. We found that due to the high nucleic acid content of the sample, the magnetic beads were too difficult to transport from the sample well to the elution well. This is because both DNA and RNA are on the beads, and they have strong bonds between them, which cannot be broken up by the magnetic pull on the magnetic beads.1515. K. Lee and A. Tripathi, “Parallel DNA extraction from whole blood for rapid sample generation in genetic epidemiological studies,” Front. Genet. 11, 374 (2020). https://doi.org/10.3389/fgene.2020.00374 Thus, our focus shifted to using the microfluidic separator for the wash steps in Part III, after the DNAse I digestion and re-binding of the RNA to the magnetic beads.Using the insight from the conventional protocol's variations, the finalized microfluidic protocol is shown in Fig. 4. Part I remains the same between the conventional protocol and the microfluidic protocol. 28.4 μl of the solution from Part I was mixed with 3.4 μl of magnetic beads and 41 μl binding buffer No. 2 and incubated for 5 min at room temperature. The supernatant was removed and 27.4 μl of DNase I solution was added to the magnetic beads. The wash steps in Part II of the conventional protocol were not performed. Then, 41 μl of binding buffer No. 2 was added to the DNase I/beads solution and incubated for 5 min at room temperature. ∼50 μl of a 1:1 mixture of wash buffers 3 and 4 were added to the wash channel. Then, the DNase I/beads/binding buffer No. 2 solution (72 μl) was added to the sample well of a microfluidic separator simultaneously with a 1:1 mixture of the two wash buffers added to the elution well. The microfluidic separator was then used to clean the RNA–magnetic bead complex. The magnetic beads were moved from the sample well to the elution well until all magnetic beads are transported to the elution well, which was verified by visual inspection. Then, the solution from the elution well was placed into a tube, and the supernatant was removed. 20 μl of the elution buffer was added to the RNA–magnetic bead complex and incubated for 5 min. The supernatant contains the isolated total RNA. The 1:1 wash buffer mix was prepared fresh daily due to the uncertainty around the stability of that solution long-term. The volume of 85 μl was removed from the sample well (a larger volume than was input) to ensure all the contents of that well were removed from the separator such that none flows into the elution well. If the sample well contents flow into the elution well, the RNA did not amplify in RT-qPCR, due to the presence of DNase I (data not shown).

G. Can the microfluidic separator replace the post-DNase I treatment wash steps?

Following the steps in Fig. 4, we tested if the microfluidic separator was sufficient to perform the wash after the DNase I digestion and re-binding of the RNA to the beads (Fig. 5). The average Cq value for disodium EDTA anticoagulated samples was 31.82 and that of sodium heparin anticoagulated samples was 33.57. In this experiment, all six samples tested amplified, indicating repeatability and that the microfluidic separator was sufficient to wash away any PCR inhibitors, including DNase I. Additionally, there was no significant difference between using the microfluidic separator for washing compared to the conventional protocol or the 1/22-scaled-down protocol. “On-chip wash” does have more variability in Cq value than the control, which is likely due to the slight differences between separators and their formation. “On-chip wash” was performed on an earlier iteration of the microfluidic chip in which the elution well was a smaller diameter than the sample well (instead of being the same diameter), which made removing solutions from it via pipetting slightly challenging. This issue was solved by making the elution well's diameter larger and the variability between microfluidic separators was reduced as seen in subsequent results.

H. Electrophoretic applications to the microfluidic separator for increased purification

Building upon previous work in our group, we sought to investigate the use of electrokinetics to improve purification of the RNA samples. Here, when referring to electrokinetics, we are referring to both electroosmotic flow and electrophoretic transport. The total velocity of the solute would be the sum of the electroosmotic and the electrophoretic velocities,Electroosmotic flow is the bulk flow of fluid in the presence of an electric field. For this study, the glass slide to which the PDMS microfluidic chip is bonded is negatively charged, causing the positive charges of the fluid to be attracted to the glass slide creating an electric double layer.2626. A. A. S. Bhagat, S. Dasgupta, R. K. Banerjee, and I. Papautsky, “Effects of microchannel cross-section and applied electric field on electroosmotic mobility,” in TRANSDUCERS 2007—2007 International Solid-State Sensors, Actuators and Microsystems Conference (IEEE, 2007), pp. 1853–1856. Thus, along the surface of the glass slide, in the presence of an electric field, the bulk flow of the fluid is toward the negative electrode due to the positive charges of the fluid. Electrophoresis is the movement of a solute in response to an electric field, depending upon the charge of the solute. Cations will move toward a negative electrode, and anions will move toward a positive electrode.2727. D. Harvey, see https://chem.libretexts.org/Bookshelves/Analytical_Chemistry/Analytical_Chemistry_2.1_(Harvey)/12%3A_Chromatographic_and_Electrophoretic_Methods/12.07%3A_Electrophoresis for “12.7: Electrophoresis - Chemistry LibreTexts,” Chemistry LibreTexts, 2021. Schneider et al. and Deraney et al.17,1817. L. Schneider, F. Cui, and A. Tripathi, “Isolation of target DNA using synergistic magnetic bead transport and electrokinetic flow,” Biomicrofluidics 15(2), 024104 (2021). https://doi.org/10.1063/5.004530718. R. N. Deraney, L. Schneider, and A. Tripathi, “Synergistic use of electroosmotic flow and magnetic forces for nucleic acid extraction,” Analyst 145(6), 2412–2419 (2020). https://doi.org/10.1039/C9AN02191D found that electroosmotic flow improved the purity of the magnetic bead-bound DNA samples they were purifying because contaminants in the bulk solution would flow out of the wash channel into the wash buffer loading well that contained the negative electrode.Here, there are multiple contaminants in the solution that surrounds the beads, which are positively and negatively charged and will reduce RNA purity. The solutes are likely the DNase I enzyme, binding buffer salts, and fragmented genomic DNA. DNase I also requires positive ions such as Mg2+ and Ca2+ to degrade DNA.2828. M. Guéroult, D. Picot, J. Abi-Ghanem, B. Hartmann, and M. Baaden, “How cations can assist DNase I in DNA binding and hydrolysis,” PLoS Comput. Biol. 6(11), e1001000 (2010). https://doi.org/10.1371/journal.pcbi.1001000 The magnetic beads were dragged through the microchannel and the fluid drag on beads 6πηRvb separate any unbound DNase, salts, or fragmented genomic DNA. Fluid viscosity, bead aggregate radius, and bead velocity are denoted by η, R, and vb, respectively. However, the no-slip condition on the bead aggregate is less likely to separate the unbound species close to bead surfaces. Hence, electric field driven electrokinetic flows were investigated to effectively separate unbound species next to beads. More specifically, we were interested in exploiting electrophoretic transport as this would theoretically remove both positively and negatively charged contaminants from the wash channel.

I. Experimental investigation of electrophoresis for improving RNA purification on the microfluidic separator

Similar to our prior work, we applied an electric field onto the microfluidic separator via placing electrodes into the primary and secondary wash buffer loading wells to induce electrophoretic and electroosmotic velocities. To obtain uniform zeta potential between different microfluidic chips, it was essential for the microfluidic chip to not be used until it sat overnight following plasma bonding between the PDMS and glass slide. A negative electrode was placed into the primary wash buffer loading well, and a positive electrode was placed into the secondary one, causing the electroosmotic velocity to be in the opposite direction as magnetic bead motion and electrophoretic transport to occur according to the respective electrode (i.e., negatively charged contaminants flow toward the positive electrode and vice versa) [Fig. 6(a)]. Electroosmotic flow in the opposite direction of the bead motion would mean that the bulk flow of the fluid and contaminants would be away from the elution well to aid in purifying the sample. Electrophoretic transport toward both electrodes would mean that both positively and negatively charged contaminants are flowing out of the wash channel to the wash buffer loading wells.We compared the use of the electric field to an off-chip sample that followed the microfluidic protocol (Fig. 4), but instead of using the microfluidic chip in Part III, the wash was performed in a tube. We tested off-chip (average Cq = 27.28) and compared them to 0 V (average Cq = 27.12), 50 V (average Cq = 26.31), 150 V (average Cq = 27.35), and 300 V (average Cq = 25.62). There was no significant difference between the average Cq values of any of the tested conditions vs off-chip [Fig. 6(b)].Cq values are a function of both RNA concentration and RNA sample quality, i.e., the presence of inhibitors or degradation of RNA. Therefore, to further characterize the RNA extracted, capillary gel electrophoresis was performed using the Agilent Bioanalyzer [Fig. 6(c)]. In this analysis, each sample's replicate was measured three times with the Bioanalyzer, and those values were averaged. Then each replicate's measurements were averaged. The Bioanalyzer measures concentration and an RNA integrity number (RIN). The concentration values were highly variable with the Bioanalyzer measurements, but the yields of 0, 50, 150, and 300 V were not statistically significantly different than that of the off-chip. The lack of statistical difference is likely due to the high variability in the off-chip concentration measurements. The off-chip concentration measurements ranged from 871 pg/μl (RIN of 4) to 14 001 pg/μl (RIN of 2.1), and the median was 1,351 pg/μl (Table ST1 in the supplementary material). With RIN, there was a statistically significant difference between off-chip and 150 V, and off-chip and 300 V, with the RIN of 150 and 300 V being higher than that of off-chip.Since Cq values depend on a combination of purity and concentration, the combination of the RNA concentration and purity values in Fig. 6 explains why Cq does not show differences in purity. The average off-chip concentration was higher than that of any of the on-chip samples, and the average RIN for the on-chip samples were higher than those of the off-chip samples. Thus, this results in similar Cq values for all samples. A study by White et al.2929. K. White, P. Yang, L. Li, A. Farshori, A. E. Medina, and H. R. Zielke, “Effect of postmortem interval and years in storage on RNA quality of tissue at a repository of the NIH NeuroBioBank,” Biopreserv. Biobank. 16(2), 148–157 (2018). https://doi.org/10.1089/bio.2017.0099 showed some correlation between RIN and Cq value, with Cq values being lower with higher RIN. However, it has been shown that amplification of transcripts can happen despite RNA degradation.3030. M. Sidova, S. Tomankova, P. Abaffy, M. Kubista, and R. Sindelka, “Effects of post-mortem and physical degradation on RNA integrity and quality,” Biomol. Detect. Quantif. 5, 3–9 (2015). https://doi.org/10.1016/j.bdq.2015.08.002We tested a reversal of the electrodes, termed the negative configuration, to ensure electrophoretic transport was dominating over electroosmotic flow. If electroosmotic velocity was dominating, we would see a reduction of purity of the RNA samples because the contaminants would flow in the direction of the elution well,1818. R. N. Deraney, L. Schneider, and A. Tripathi, “Synergistic use of electroosmotic flow and magnetic forces for nucleic acid extraction,” Analyst 145(6), 2412–2419 (2020). https://doi.org/10.1039/C9AN02191D which we did not see (Fig. S1 in the supplementary material). We also tested disodium EDTA anticoagulated blood in the positive configuration and had similar results for the Cq values comparisons (Fig. S2 in the supplementary material). For the Bioanalyzer results, however, the RIN values could not be measured reliably, likely due to the concentration of EDTA being higher than the recommended limits for the RNA Pico 6000 Kit. These results are in Table ST2 in the supplementary material. The electropherogram traces for disodium EDTA samples were similar to that of the sodium heparin ones; thus, the threshold for the RIN measurements could be adjusted to estimate RIN values, but with less confidence.Overall, these results show that electrophoretic transport increased purity of the RNA extracted using the microfluidic separator. The higher the voltage the better the purity, i.e. the RIN of the sample increases. Even in the absence of electrokinetics, 0 V appeared to improve purity of the RNA (though not statistically significant) indicating that the microfluidic separator alone may clean the sample better than the off-chip protocol. This could be due to the differing geometries between an Eppendorf tube and the microfluidic wash channel, where the channel is more spread out allowing for the magnetic beads to experience advection more strongly. On the other hand, all samples tested on the microfluidic separator had a lower concentration than when using a tube off-chip. This demonstrates the importance of treating the microfluidic device with a solution to eliminate RNases (which was not done here), since the concentration of RNA purified on the microfluidic device was lower than that of the off-chip. The lower concentration could also be due to the adsorption of RNA to the walls of the PDMS or glass of the microfluidic chip.3131. J.-Y. Yoon and R. L. Garrell, “Biomolecular adsorption in microfluidics,” in Encyclopedia of Microfluidics and Nanofluidics, edited by D. Li (Springer, Boston, MA, 2008), pp. 68–76. Finally, since the concentration was not significantly reduced between the voltages tested, this shows that the RNA remained bound to the magnetic beads despite the effects of electrophoresis present in the microfluidic separator.

J. Mechanistic understanding of the flow of unbound species in the microfluidic channel

To validate our experimental results, we mathematically modeled the electric field on the microfluidic separator using COMSOL Multiphysics to observe the effect of electrokinetic velocities. For the wash channel, the solution chosen for modeling was ethanol, which is likely a primary component of the proprietary wash buffer used experimentally. The solute was chosen to be DNA fragments of 150 bp in size with a large negative charge, though experimentally the solutes are much more complex than that. Three modules were used in COMSOL: electric currents, laminar flow, and transport of diluted species, to solve for the electric field, electroosmotic flow, and the convection, diffusion, and electrophoresis of charged species, respectively. For further details regarding the model, see the supplementary material. Figure 7 shows the concentration of the solute at 3 min of the simulation time for 150 and 300 V with zi = −9000, where zi is the charge number of the ionic species (dimensionless). The electroosmotic velocity, electrophoretic flux, and electric field for 150 and 300 V are shown in Figs. S3 and S4 in the supplementary material. This shows that electrophoresis is what dominates the velocity of the solutes since the solutions (in red) are moving toward the positive electrode instead of the negative one. If the solution has a positive charge of the same magnitude (zi = 9000), the same profile is seen but toward the negative electrode (Fig. S5 in the supplementary material). Electroosmotic velocity is also occurring (Fig. S3 in the supplementary material), though it is not greater than electrophoretic velocity as evidenced by the concentration profile showing electrophretic flow despite increased electroosmotic velocity with increasing voltage.

Modeling in COMSOL aided in understanding the movement of the unbound species in the wash channel of the microfluidic separator. This showed that in the positive configuration, negatively charged ions move toward the bottom of the chip due to electrophoresis, while the positive ions travel upward due to electroosmosis and electrophoresis.