Engineered membranes for residual cell trapping on microfluidic blood plasma separation systems. A comparison between porous and nanofibrous membranes

Blood-based clinical diagnostics require challenging limit-of-detection for low abundance, circulating molecules in plasma. Micro-scale blood plasma separation (BPS) has achieved remarkable results in terms of plasma yield or purity, but rarely achieving both at the same time. Here, we proposed the first use of electrospun polylactic-acid (PLA) membranes as filters to remove residual cell population from continuous hydrodynamic-BPS devices. The membranes hydrophilicity was improved by adopting a wet chemistry approach via surface aminolysis as demonstrated through Fourier Transform Infrared Spectroscopy and Water Contact Angle analysis. The usability of PLA-membranes was assessed through degradation measurements at extreme pH values. Plasma purity and hemolysis were evaluated on plasma samples with residual red blood cell content (1, 3, 5% hematocrit) corresponding to output from existing hydrodynamic BPS systems. Commercially available membranes for BPS were used as benchmark. Results highlighted that the electrospun membranes are suitable for downstream residual cell removal from blood, permitting the collection of up to 2 mL of pure and low-hemolyzed plasma. Fluorometric DNA quantification revealed that electrospun membranes did not significantly affect the concentration of circulating DNA. PLA-based electrospun membranes can be combined with hydrodynamic BPS in order to achieve high volume plasma separation at over 99% plasma purity.


Introduction
Blood plasma separation is usually required for several blood-based clinical diagnostics [1]. In this context, plasma purity superior to 99% is a crucial requirement since presence of blood cells and hemoglobin, may interfere with immunoassays and other assays [1][2][3]. On the other hand, cell lysis, such as hemolysis, but also leukolysis, should be avoided as DNA or RNA released from lysed cells may interfere with circulating cell-free DNA and RNA levels, resulting in erroneous readings, particularly in the case of liquid biopsy assays. In recent years, there has been a growing interest in micro-scale blood plasma separation (BPS) both as a replacement for traditional centrifugation and as preliminary modules to complete Lab-On-Chip devices. BPS studies have demonstrated remarkable results in terms of yield or purity, but have rarely achieved both at the same time [2].
Micro-scale BPS strategies adopted until now depend on several factors such as the volume of plasma required for subsequent clinical tests as well as the number and type of tests to be performed on the same sample. For instance, blood gas, electrolytes and metabolites (BGEM) tests can be carried out on plasma from a drop of blood, typically in the order of 10-200 µL. In this volume range, paper-based technologies or filtration are the most common approaches [4][5][6][7].
If larger blood volumes are needed (milliliter range), such in the case of amplification of rare circulating DNA from plasma [8], hydrodynamic blood plasma separation (HBPS) is preferred because it avoids the saturation usually quickly reached by membrane-based separation. These approaches permit achieving plasma purity close to 100%, although plasma yields are usually low (1-15%) and often the plasma purity decreases upon increasing plasma yields [9][10][11][12][13].
The combination of hydrodynamic and membrane separation in a BPS system is proposed to achieve high yield and high purity simultaneously, when processing large volumes of blood, by allowing large amount of plasma to be removed with suboptimal purity, followed by removal of residual RBCs by filtration without loss of plasma volume.
Membrane filtration at microscale is often used as a sieving process to retain large molecules or particulate matter [14][15][16] However, in order to accomplish high performance separation, membranes should meet challenging requirements such as low pressure drop and high plasma purity. The development of engineered membranes could allow the control of chemical, physical and mechanical properties of biopolymeric highly porous structure susceptible to meet these requirements. Among the several approaches developed for fabricating biopolymeric porous structures, thermally or diffusion induced phase separation (TIPS or DIPS) as well as electrospinning are gaining more attention [17,18].
TIPS and DIPS techniques are based on thermodynamic de-mixing of a homogeneous polymersolvent solution into a polymer-rich and a polymer-lean phase. De-mixing can be achieved by a change in temperature (TIPS) or by changing the exposure of the solution to another immiscible solvent (DIPS). [17,18] Both these approaches lead to the production of well interconnected porous structures exhibiting tunable pore size.
On the other hand, electrospinning produces ultrafine fibers ranging from micro-to nano-meter in diameters from polymeric solutions or melts [19]. Electrospun (ES) membranes show high porosity and, tunable mechanical properties as well as high surface to volume ratio due to their small fiber dimension and they were successfully integrated in microfluidic chips for cell culture or for immunoassay optimization [20][21][22][23].
One of the most common biopolymers used for these applications is polylactic acid (PLA). Due to its commercial availability at an affordable cost, PLA has been extensively studied and used in many biomedical purposes including wound healing [24,25] and tissue engineering scaffolds [26,27]. Recently, we have shown that PLA can be used as a novel substrates for microfluidic devices [28]. During the last decades, researchers have contributed to widening the application range of PLA through chemical and physical alterations to the bulk and/or surface by introducing hydrophilic and biocompatible moieties [29,30]. Several modifications strategies such as plasma [31] or chemical [29] treatments have been employed to enable specific surface properties on PLA.
For instance, surface modification such as the aminolysis technique has been widely used for the addition of amine groups in order to covalently bind bioactive molecule on PLA [32].
In an effort to develop a rapid and effective residual cell trapping, the efficacy of PLA-based membranes prepared by TIPS, DIPS, and electrospinning for blood plasma separation was assessed.
To the best of our knowledge, it is the first time that PLA-based porous membranes have been tested as a microfiltration unit (MFU) for microfluidic blood plasma separation purposes. Initially, a comparative study on PLA membranes degradation at extreme pH values was carried out in order to ascertain their usability in microfluidic protocols requiring the use of non-neutral buffers or after acid or basic treatments. Secondly, plasma purity, free hemoglobin concentration and pressure drop across the DIPS, TIPS and ES membranes have been evaluated employing plasma-diluted blood corresponding to output-plasma tested from already existing HBPS systems [33] i.e. with an initial hematocrit up to 5% flowing at 1, 2 or 3 mL h -1 . In addition, we have compared the performance of pristine with surface modified membranes, using commercially available BPS membranes as a benchmark. Finally, non-specific binding of circulating DNA was evaluated for each of the membranes via fluorometric quantification with Qubit TM analysis in order to assess their suitability in use in clinical applications. Vivid TM GX (Vivid B) from Pall. All the reactants were ACS grade (purity > 99%).

Thermally Induced Phase Separation membrane preparation
TIPS membranes were prepared by dissolving PLLA (4 wt%) in an 87/13 w/w dioxane-water mixture at 120 °C. The solution, kept above 60 °C to ensure homogeneity, was hot poured in a cylindrical high-density polyethylene (HDPE) mold. The mold filled with the polymer solution was first immersed for 10 minutes in a 0 °C bath and then moved to a -20 °C bath for 20 minutes in order to allow the complete freezing of the as-obtained structure. Successively, the foams were extracted from the mold and subjected to a washing step in distilled water. Finally, the foams were dried under vacuum for 24 h to ensure the complete removal of any solvent trace. The so obtained membranes were then cut to obtain slices 1 mm thick.

Diffusion Induced Phase Separation membrane preparation
DIPS membranes were produced by dissolving PLLA (8 wt%) in neat dioxane at 120°C. A stainless steel slide was immersed into the solution, cooled at 30°C and then pulled out at a constant speed of 3 cm min -1 (dip coating) to ensure a proper final average membrane thickness of 120 µm, based on previous work [17]. The coated slide was immerged for 5 minutes into an 87:13 w:w dioxane-water coagulation bath (in order to avoid this so called "skin effect"), and then it was immerged into a pure water bath for other 5 minutes to ensure a complete phase separation. The temperature of both coagulation baths temperature was fixed at 30 °C.

Electrospun membrane preparation
The polymeric solution for electrospinning was obtained by adding PLA (10 wt%) to 20 mL of TCM:Ac (2:1 v:v), and completely dissolved by stirring overnight at room temperature to obtain homogeneous PLA/TCM:Ac solution [26]. A semi-industrial electrospinning equipment (NF-103, MECC CO., LTD., Japan) was used to prepare PLA nanofibers membranes. The polymeric solution was filled in a 10-mL syringe fitted with a 19 gauge stainless steel needle. The electrospinning was then performed using the following constant parameters: flow rate, 1 mL h -1 ; distance between the needle tip and the collector, 13 cm; supplied high voltage, 15 kV; temperature, 25 °C and relative humidity, 40%. The nanofibers obtained were collected on a grounded rotary drum (diameter = 10 cm, speed = 10 rpm) wrapped in an aluminum foil for 180 min in order to obtain membranes approximately 100 μ m thick. The collected PLA ES membranes were subsequently dried for at least 2 days under fume hood in order to remove any residual solvents.

Membrane surface modification
All the membranes were surface modified through a solution of methanol:water (3:7 v:v) and HMDA (1% w:v). The solution was directly pipetted onto each side of the PLA membranes. A 12hour drying period at room temperature in a fume cupboard followed.

Morphological analysis
The morphology of the nanofiber mats was evaluated by scanning electron microscopy, (Phenom ProX, Phenom-World, The Netherlands). The samples were attached on an aluminum stub using an adhesive carbon tape.

Fiber diameter and pore size distribution
Fiber diameter distribution of ES membranes and pore size distribution of TIPS and DIPS membranes were determined using ImageJ as an image processing software. In particular, the plugin DiameterJ is able to analyze an image obtained by SEM and to find the diameter of nanofibers at every pixel along a fiber axis. The software produces a histogram of these diameters and summarizes statistics such as mean fiber diameter [26].

pH stability test
The hydrolysis of the ES mats (10 mm × 40 mm × 100 µm) was performed in 10 mL of solutions (pH 1.0, pH 7.0 and pH 13.0) at 25 °C up to 4 hours. At fixed times, i.e. 0.5, 1, 2 and 4 hours, the membranes were washed thoroughly with distilled water at room temperature, followed by drying under chemical hood for 1 day. After drying, the samples were weighed (m dry ).

FTIR-ATR analysis
The chemical surface properties of the samples were assessed by spectroscopic analysis. FTIR-ATR analysis was carried out by using a PerkinElmer FTIR/NIR Spectrum 400 spectrophotometer; the spectra were recorded in the range 4000-400 cm -1 .

Water Contact Angle measurements
Static contact angles were measured on all the samples by using distilled water as fluids with an FTA 1000 (First Ten Ångstroms, UK) instrument. More in detail, 4 μ L of distilled water were dropped on the membranes. Images of the water on the surface were taken at a time of 10 s. At least 7 spots of each composite nanofiber mat were tested and the average value was taken. Layers were aligned in specific custom-made plates endowed with metallic pins; therefore, individual layers were featured with corresponding alignment holes. The membranes were positioned in the proper layer and then ethanol was spread between PMMA layers with a pipette.

PMMA device preparation and membranes integration
The bonding was carried out at 70 °C and 100 bar for 5 minutes in a Carver laboratory press [34].

Microfluidic Device testing
A WPI Al300-220 syringe pump loaded with a 10 mL syringe was used to impose a constant flow through the membranes. Plasma-diluted human blood with initial Hct (Hct in ) equal to 1%, 3% and 5% were used for this test. All kinds of functionalized and unfunctionalized membranes were tested. The blood flow rates were 1, 2 and 3 mL h -1 which corresponds to tested output flow rates from existing HBPS system [33]. Measurements were carried out at three volume points of extracted plasma i.e. 0.5 mL, 1 mL and 2 mL. LabSmith Pressure Sensor was used to measure the blood pressure upstream to the membranes.
Separated plasma was collected and analyzed in aliquots at different volume points (ܸ ) i.e. 0.5 mL, 1 mL and 2 mL. The cumulative Hct (%) was calculated according to the following Equation (1): is the hematrocit evaluated for the i-th sample; ‫ܪ‬ ܿ ‫ݐ‬ is the Hct of the feeding blood.
The plasma purity [%] was evaluated according to Equation (2): The extent of hemolysis in plasma samples was determined through UV-Vis spectroscopic measurement according to Cripps method. In particular, Cripps number was calculated according to where A 560 , A 576.5 and A 593 are the absorbance values at 560 nm, 576.5 nm and 593 nm, respectively. Before UV-Vis spectroscopy, each plasma sample was centrifuged for 10 min at 1,600 g and only the supernatant absorbance was measured in order to ensure to count only free hemoglobin present in the sample. Absorbance was measured using Jenway 7315 Spectrophotometer in 500-630 nm wavelength range at 1 nm interval.
In order to report hemolysis values only due to the cell damage resulting from the microfluidic membranes (H corrected ), the H value of the centrifuged feed (H control ) was subtracted to H value of the extracted volume (H out ) according to Equation (4): In order to correlate the Cripps number with a concentration of Hgb in plasma, a series of centrifuged plasma containing 0.1 g dL -1 , to 0.3 g dL -1 of Hgb (measured with the hemoanalyzer) were used to obtain a calibration curve, correlating the H value and the hemoglobin concentration.
In the concentration range herein investigated, the calibration curve was found to be a line with an angular coefficient of 2.17 up to a H value of 0.33. For higher H value, the calibration curve still remained a line but with a slope of 3.17. In the case of Hgb concentrations higher than 0.3 g dL -1 , the results from the hemoanalyzer were considered.
Finally, the cumulative Free Hgb (g dL -1 ) concentration was calculated according to the following Equation (5): is the hemoglobin concentration evaluated for the i-th sample; The score index calculation was evaluated on the basis of the two following arbitrary parameters: The parameter related with the Hgb concentration (K Hgb ) was determined according to Equation (6): Where, Hgb th is the Hgb threshold value chosen i.e. 0.15 g dL -1 . Note that if Hgb i > Hgb th then K Hgb < 0 and these values appears in the heat map as a black square.
The parameter related with the plasma purity (K RBC ) was determined according to Equation (7): Where, PP i is the plasma purity evaluated for the i-th sample and PP th is the plasma purity threshold value chosen i.e. 85 %. Note that if PP i < PP th then K RBC < 0 and these values appears in the heat map as a black square.
The positive values of K Hgb a n d K RBC were normalized in the 0-1 range and called nK Hgb and nK RBC.
The score index calculation (SIC) was then evaluated according to the following Equation (8):

DNA binding
In order to assess the binding capacity of the membranes the mixture of fragmented T47-D cell line

Statistical analysis
Statistical analyses of the data were performed through ANOVA by using the GraphPad Prism version 8 software (GraphPad Software Inc., La Jolla, CA, USA). Significant differences among mean values, where applicable, were determined by the means of ANOVA. In all cases, a value of p<0.05 was considered statistically significant.
In an attempt to achieve rapid and effective blood plasma separation, compatible with Lab-on-Chip or in-line processing, we have developed an hybrid strategy, combining HBPS published previously [33], and a MFU ( Figure 1A).

and (ii) front side of Vivid ® membrane; (iii) ES membrane; (iv) TIPS membrane; (v) Back side and (vi) Front side of DIPS membrane.
This two-step strategy should allow the simultaneous achievement of high yield (via the HBPS device) and high purity (via the collection of residual blood cells from the HBPS device on the MFU) in the milliliter range blood plasma separation. In fact, HBPS performance decreases upon increasing plasma yields [9][10][11][12][13]. Combining in the same BPS system hydrodynamic and membrane separation may permit removing the residual RBCs due to an increase of the plasma yields.
Accomplishing high performance separation is not straightforward. Membranes should meet challenging requirements such as low pressure drop and high plasma purity.
In this respect, we have evaluated the use of PLA-based membranes prepared by TIPS, DIPS and ES for the removal of residual red blood cells following HBPS. The performance of these PLA engineered membranes were compared to two commercially available membranes for RBC removal i.e. Vivid TM GF (Vivid A) and Vivid TM GX (Vivid B). As shown in Figure 1B Figure 1Ciii, it was determined that all the fibers in the ES membranes were in the nanoscale range (mean fiber diameter 0.83 µm ± 0.11 µm), randomly oriented and smooth, without any presence of beads. Image processing analysis revealed that the mean pore diameter of this system was 3.53 µm ± 1.67 µm. Figure 1Civ shows the TIPS membrane morphology which was characterized by pores with a mean pore diameter of 3.31 µm ± 0.81 µm and a pore wall of approximately 1 µm. Figures 1Cv and vi show, the morphologies of the back (or bath side) and front sides (slide side) of the DIPS membrane, respectively. DIPS membranes presented considerable differences in pore morphology between the two sides with the mean pore diameter of the front side (in contact with the stainless steel slide) being 1.02 µm ± 0.19 µm while that for the back side (the side immersed in the water/dioxane solution) was 3.05 µm ± 0.36 µm. This asymmetry in pore morphology can be ascribed to the water diffusion kinetic into the polymeric solution. In particular, during desiccation, the back side (directly exposed to air) is characterized by faster water removal compared to the front side (in contact with stainless steel).
For this reason, the external membrane surface is more likely to be closed, as faster water removal could re-establish a single-phase condition [17].

Comparative study on PLA membranes degradation at extreme pH values
Chemical based surface modification protocols may be used to provide various functionalities to the membranes. In order to compare the effects of surface modification chemistries, the membranes were subjected to extreme pHs (acidic and basic treatment) for up to 4 hours, as shown in Figure 2. ES membranes immersed in acidic medium and distilled water for up to 4 hours did not show any significant change in their morphology. In fact, only a slight increase in the fiber diameter ( Figure   2B) in acidic medium after 4 hours was observed. On the other hand, ES membranes immersed in the medium at pH 13 for 4 hours exhibited a considerable number of broken fibers (see the inset) and a slight reduction in the mean fiber diameter that did not affect the average pore size.
Similarly, TIPS and DIPS membranes showed a more pronounced morphology change in the basic medium. Although the mean pore diameter of the TIPS remained almost constant ( Figure 2C), it can be clearly seen that, at pH 13, the edge of the pores had become partially eroded, particularly after 4 hours.
DIPS membranes immersed in all kinds of medium show a more significant change in mean pore diameter as a function of time. In Figure 2D it is possible to observe that the pore size of the front side gradually increased while that of the back side gradually decreased, in particular for membranes immersed in pH 1 and pH 13 solutions. As can be seen in Figure 2A, when the polymer matrix was immersed in a medium at pH 1, the back side seemed to swell, thus inducing a reduction in the pore size, while the edges of the pores on the front side partially eroded with time, thereby slightly increasing the pore diameter. A similar phenomenon was observed for the membranes immersed in distilled water, although less pronounced. In contrast, DIPS membranes immersed in a medium at pH 13 seem to be affected by deeper erosion. In fact, after 4 hours both side of the membranes show a similar morphology likely due to the degradation of the membrane external layers (Figure 2A).
The above observations can be explained by considering the mechanism of hydrolysis of PLA, which is primarily determined by the amount, presence and location of water molecules. Recent studies correlated the differences in the degradation behavior of PLA to the hydrophilicity of this polymer matrix exposed in media at different pH [37]. PLA exposed to basic media (pH 13) maintains its non-polar (hydrophobic) character, thus reducing its absorption capacity of water. Therefore, water cannot permeate the polymer matrix thus causing superficial degradation.
Differently, in acid medium (pH 1) PLA molecular chains change their character from hydrophobic to hydrophilic thus allowing PLA swelling and the bulk degradation of the samples [19]. In both cases, during PLA degradation, lactic acid oligomers, lactide and lactic acid are formed [38]. In conclusion, the PLA membrane durability in the acid medium is very similar to that in the neutral medium and higher than that in the alkaline environment. Probably the PLA bulk degradation due to acidic medium is slower than surface degradation due to basic environment thus affecting the membranes morphology less than the basic environment during the 4 hours of the test time.

Wettability and surface functionalization of PLA-based porous or fibrous membranes
In order to accomplish high performance separation, membranes should meet challenging requirements such as low pressure drop that can be reduced by modifying their wettability.
Several approaches have been investigated in order to enhance the wettability of PLA-based porous or fibrous membranes in both dry and wet chemistry [29][30][31]. In this work, an aminolysis surface treatment via hexamethylenediamine (HDMA) was carried out. SEM images of PLA-based membranes after aminolysis are shown in Figure 3.  Morphological inspection clearly reveals that the surface treatment eroded the porous structure of all membranes. In particular, from Figure 3A, for the m-ES, it is possible to observe several fractured fibers, as already reported by others [39]. Furthermore, it can be seen that some fibers are attached to each other in bundles of 2 to 6. The reason for this phenomenon may be due to the increased interaction now possible following surface modification. Similar to the treatment in basic medium, the TIPS membrane ( Figure 3B) showed erosion of the pore edges while the DIPS membrane showed partial erosion of the superficial layer of the membrane.

SEM images of (A) ES membrane; (B) TIPS membrane; (C) Front side and (D) back side of DIPS membrane after aminolysis. Effect of surface modification on WCA of (A') ES membrane; (B') TIPS membrane; (C') DIPS membranes; (E)
In Figure 3A'C' the effect of surface modification on the water contact angles are reported.
Unmodified ES, TIPS and DIPS membranes showed WCA equal to 118°, 116° and 89°, respectively. It is well known that the water contact angle is dependent on surface chemical functional groups, porosity and surface roughness. Since the polymer matrix is the same in all cases, this result can be reasonably ascribed to the higher porosity of both the TIPS and ES membranes with respect to the DIPS membranes. In fact, porosity allows the material to entrap air when the surface is immersed in water, thus enhancing the hydrophobicity of the PLA membranes [19]. At the same time, m-ES and m-TIPS membranes showed the highest WCA reduction, i.e.
100% and 71% respectively, while WCA decrease of m-DIPS membranes was only 29%. This result can be likely ascribed to the higher specific surface of ES and TIPS membranes that permits exposing more amine moieties to water drops than DIPS membranes. deformation, which give rise to peaks at 3353, 3262, 3175 and 1585 cm -1 , respectively [40].
Although absorption in the range 3000 -3500 cm -1 in amino-modified membranes is barely visible due to the vibration of amide groups that lie in the same regions and to the presumable presence of water due to the enhanced wettability, the peak at 1585 cm -1 is clearly visible in particular for m-TIPS and m-ES membranes. Moreover, after aminolysis of ES, TIPS and, to a lesser extent, DIPS membranes showed another new peak at 1650 cm −1 which corresponded to amide I stretch [41].
The intensities of the amine-related peaks, including the peak related to the primary amide, decrease progressively for ES, TIPS and DIPS membranes and this is consistent with the WCA results of amino-modified membranes showing that the reactive amine density is highest at the highest specific surface exposed by the membrane [40].

Membrane performance: Plasma purity
In order to evaluate the membranes performance, they were included in a microfluidic device denominated MFU. The MFU design is shown in Figure 1B  Similar experiments performed using both the surface modified and unmodified TIPS and DIPS membranes resulted in clogging (reaching a pressure higher than 300 kPa) before 0.5 mL of plasma had been extracted. The results have been compiled into a three-color heat map and shown in Figure   4A.   a  n  i  z  e  d  i  n  q  u  a  d  r  a  n  t  s  f  o  r  c  l  a  r  i  t  y  .  T  h  e  b  a  s  e  l  i  n  e  v  a  l  u  e  i  s  s  e  t  a  t  9  5  %  a  n  d  r  e  p  r  e  s  e  n  t  e  d  i  n  y  e  l  l  o  w  ,  t  o  e  n  a  b  l  e  r  a  p  i  d  r  e  a  d  i  n  g  a  n  d  s  e  l  e  c  t  i  o  n  (  g  r  e  e  n  c  o  r  r  e  s  p  o  n  d  i  n  g  t  o  p  l  a  s  m  a  p  u  r  i  t  y  >  9  5  %  )  .  C  r  o  s  s  e  s  i  n  d  i  c  a  t  e  s  e  t  s  o  f  p  a  r  a  m  e  t  e  r  s  w  e  r  e  d  a  t  a  w  a  s  n  o  t  a  c  q  u  i  r  e  d  ,  b  e  c  a  u  s  e  t  h  e  d  e  v  i  c  e  s  c  o  u  l  d  n  o  t  h  a  n  d  l  e  h  i  g  h  p  r  e  s  s  u  r  e  The general trend observed is that there is a decrease in plasma purity with increasing flow rate, residual hematocrit and extracted volume. The highest plasma purity was achieved using ES 10L.
Employing this device, it was possible to extract up to 2 mL of pure plasma (100% of plasma purity) from plasma with 1% residual Hct in with all the flow rates investigated. For plasma samples containing 3% residual Hct in , the plasma purity decreased to 96%, 93%, and 90% using flow rates of 1 mL h -1 , 2 mL h -1 and 3 mL h -1 , respectively. It is worth noting that 90% of plasma purity due to the MFU fed with 3% of Hct in represents a global plasma purity (i.e. due to both HBPS chip and MFU) higher than 99%. In fact, assuming a whole blood hematocrit of 40%, 3% Hct in can be achieved by an HBPS chip plasma purification of around 92.5%. After a further microfiltration step able to achieve 90% plasma purity, the global plasma purity of the hybrid system would be 99.25%.
The worst plasma purity was observed in devices containing m-ES 1L membranes. The commercial membranes compared relatively well at the lowest flow rate, and the lowest residual Hct, but rapidly clogged, and did not yield results beyond this (no data collection represented by squares on heat map), showing limited use at higher volumes.

Membrane performance: Hemolysis
Membrane performance was also evaluated by measuring the free hemoglobin concentration in the extracted plasma due to the RBCs rupture during their passage through the membrane.
The two performance indicators, plasma purity and hemolysis, are needed to reflect on the true performance of the device as a high RBCs removal on its own would not necessarily reflect an overall good performance if the filtrates have high levels of hemoglobin.
In order to assess the amount of hemolysis due to membrane separation, the free hemoglobin concentration was determined according to Equation 5, as described in the 'materials and methods' section. The results have been reported in a heat map format ( Figure 4B). As expected, the general trend observed is that there is an increase in the free Hgb concentration upon increasing the flow rate, Hct in and the extracted volume. In this case, the best performances were shown by the surface modified (aminolyzed) membranes. Specifically, if 1% Hct in is used as input for the devices incorporating either m-ES 1L and m-ES 10L membranes, the concentration of free hemoglobin in the output samples was lower than 0.04 g dL -1 . On the contrary, RBCs lysis was found to be particularly pronounced for devices including the commercial porous membranes, Vivid A and Vivid B, at all the flow rates here investigated. For instance, Vivid B membranes showed a concentration of free hemoglobin equal to 0.21 g dL -1 after 1 mL of plasma extracted, with a flow rate equal to 3 mL h -1 .
Upon increasing the Hct in up to 3% and 5% the free Hgb concentration increased for all the system investigated but the trend remained similar ( Figure 4B).

Membrane performance: Combined score
It is worth noting that high separation efficiencies can result in high hemolysis value. Both performance indicators need to be balanced, to achieve and overall good separation with maximum plasma purity and minimal hemolysis. For this reason, in Figure 4C we report a combined score calculated using Equation 8 which takes both plasma purity and hemolysis into account. The higher the score, the higher the performance of the membrane. A double threshold can be set, so that systems displaying one or both indicators beyond certain thresholds can be excluded from the selection matrix. In our example in Figure 4C, we have arbitrarily set a threshold at 85% plasma purity and 0.15 g dL -1 hemoglobin level. Membranes displaying indicators respectively below or above these levels are represented in black squares. Threshold can be set up to facilitate selection when testing a large number of membranes. Here, the 0.15 g dL -1 hemoglobin level corresponds to a threshold for moderate hemolysis according to Snyder et al. [42]. The heat map clearly reveals that the best combined performance was achieved by the ES 10L membranes since they offered the best compromise between the plasma purity and a relatively low hemolysis degree. Combined score separation index heat maps obtained with other thresholds are available in Supporting Information (SI 2).

Membrane pressure drop
It is worth noting that hemolysis is strongly affected by the pressure drop across the membranes because it can stress the RBC membrane. In Figure 5 it is reported the pressure drop across all ES (amino-modified or not) and commercial Vivid membranes as a function of the hemolysis (free hemoglobin concentration). As expected, Figure 5 shows that free Hgb concentration increases The pressure drop differences can partially explain the differences in hemolysis levels between Vivid and ES membranes. It is envisaged that the fibrous morphology of the latter's PLA-based mats and their superior interconnection degree reduced RBCs lysis during plasma separation. In fact, during contact with the red blood cell, the ES fibers may bend thus reducing the stress on the RBC extracellular membrane and, as a consequence, their lysis.

Non-fouling performance
The cell separation membranes have been designed to be integrated into a complete workflow for circulating DNA extraction. Avoiding DNA adsorption onto the surface of the membrane is therefore crucial for the functioning of the membrane. To investigate the non-fouling properties of the membrane, DNA extraction after permeation of a plasma sample through the membrane placed in a column has been performed.
In Figure 6 the DNA binding test procedure draws and results are reported. In brief, in order to assess this test, silica membranes were removed from commercial spin columns and replaced with the ES membranes ( Figure 7A) [36]. The columns were loaded with the DNA solution (1 ng µL -1 ), incubated, and then centrifuged to pull the solution through the membranes ( Figure 6B, Materials and Methods).  The histogram in Figure 6C represents the concentration of DNA before (positive control) and after the permeation trough ES layers. Results showed that ES 10L and m-ES 1L and m-ES 10L did not significantly (p < 0.01) affect the DNA concentration. Only ES 1L lead to a slight decrease of DNA concentration, within the instrument deviation range. These results highlight that both the unmodified and modified ES PLA-based membranes are low binding and suitable for use in assays where maximum DNA recovery from raw samples is of importance.

Conclusion
PLA-based membranes prepared by DIPS, TIPS and electrospinning were modified trough aminolysis and characterized. The membranes resistance was tested at extreme pH values for up to 4 hours. Basic media was the most aggressive environment affecting pore morphology. The successful amine functionalization was demonstrated by FTIR-ATR while the subsequent change in wettability of the samples was assessed though WCA analysis. Both modified and unmodified TIPS and DIPS membranes clogged before extracting 0.5 mL of plasma and were excluded from further examination. ES membranes were selected for blood plasma separation tests at low hematocrit levels (1, 3 and 5%), corresponding to output-plasma tested from already existing HBPS systems.
Single layers (ES 1L) or stacks of 10 ES membranes were tested in a microfluidic configuration devise. The stack of 10 layers (ES 10L) allowed the collection of up to 2 mL of RBCs-free and lowhemolyzed (free Hgb = 0.078 g dL -1 ) plasma at 1% Hct in and flow rate up to 2 mL h -1 . With the same membrane and flow rate, upon increasing the Hct in the plasma purity decreased but it was also possible to collect 1 mL of RBCs-free and low-hemolyzed (free Hgb = 0.062 g dL -1 ) with 3% Hct in .
There is evidence to show that ES membranes decreased the degree of hemolysis when compared to the commercially available Vivid ® membranes. The ES membranes also exhibited lower pressure drops while maintaining comparable plasma purity, thereby showing superior separation performance at low hematocrit levels. We have shown that the PLA-based ES membranes may be suitable for liquid biopsy workflows, where maximum recovery of circulating DNA is crucial.
PLA-based ES membranes can be readily integrated into microfluidic workflows and show enhanced separation performance compared to commercial membranes, for the recovery of high purity plasma in the hundreds of microliters to milliliter range.