Development of novel membranes for blood purification therapies based on copolymers of N-vinylpyrrolidone and n-butylmethacrylate

Institute for biomedical technology and Science and Technology, University of Twe P.O. Box 217, 7500 AE Enschede, The Neth nl; Fax: +31 53 4892155; Tel: +31 53 48946 Interface BIOmaterials BV, Burg. Lemm Netherlands Department of Internal Medicine, Immun Medical Center, PO Box 5800, 6202 AZ Maa Department of Internal Medicine, Divisio Medical Center, PO Box 5800, 6202 AZ Maa Department of Biomedical Engineering/Bio Faculty of Health, Medicine & Life Sciences, University of Malaya, Department of B Malaysia Department of Internal Medicine, Clin. Maastricht University Medical Center, PO Netherlands Department of Internal Medicine, Cent University Medical Center, PO Box 5800, 62 Cite this: J. Mater. Chem. B, 2013, 1, 6066


Introduction
In blood purication treatments, membranes with optimal separation properties as well as optimal biocompatibility are required. For example, apheresis membranes, plasma separation membranes and hemodialysis membranes have optimized molecular weight cut offs for their applications and show good biocompatibility for a few hours of application. New developments in blood purication therapies oen require longer durations. For example, nocturnal dialysis (where 7.85 hour treatment instead of 3.75 h for conventional hemodialysis is applied) leads to improvements in patient survival. 1,2 However, longer ltration times mean more contact with the membrane material, requiring long-term membrane biocompatibility. This holds for wearable and portable articial kidneys too. Davenport et al. reported about a pilot study with a wearable articial hemodialysis device, however, 2 out of 8 patients had clotting problems and the treatment was discontinued, highlighting the need for improved biocompatible membranes. 3 Currently, blood purication membranes made of synthetic polymers dominate the market. Oen polymer blends are used in which a hydrophobic membrane material (e.g. polysulfone or polyethersulfone) is combined with a hydrophilic additive (e.g. polyvinylpyrrolidone (PVP) or polyethylene glycol (PEG)). Also hydrophilic or hydrophilized copolymers are used (poly(ethylene vinyl alcohol) or polyacrylonitrile methallylsulfonate). 4 Membranes with hydrophilic patches in a hydrophobic membrane matrix are generally thought to be biocompatible. 5 However, the hydrophilic modier can elute from the dialysis membrane surface into the patients' circulation. 6 Another approach is the use of surface modications such as heparin immobilization onto polysulfone membranes 7 or albumin graing onto polyethersulfone based membranes. 8 However, in most of these studies the biocompatibility of blood purication membranes is not fully investigated. Although the effect of biocompatibility of different membrane materials is hard to show in clinical trials, 9 both nurses and physicians prioritize biocompatibility for future innovations on blood purication membranes above efficiency and removal capacity. 10 In this study, we develop via phase inversion membranes suitable for blood purication therapies using the SlipSkinÔ biomaterial which is a copolymer of N-vinylpyrrolidone (NVP) (hydrophilic building block) and n-butylmethacrylate (BMA) (hydrophobic building block). This is actually the most common method for preparation of membranes for blood purication. 11 In general, due to the combination of hydrophilic and hydrophobic properties in one material, copolymers may offer favorable and oen tunable biocompatibility and cytocompatibility properties. However, at the same time this combination oen imposes challenges in the fabrication of porous membranes with such copolymers. As a result very few studies discuss in detail the science and technology for their optimization for specic applications. In this work, we pay specic attention to this and aim to understand the mechanism of pore formation and the tailoring of the membrane's mechanical and transport properties to obtain the optimal membrane for blood purication therapies. We use various polymer compositions in which the molar ratio of the hydrophilic and hydrophobic building blocks is varied and prepare under various process conditions (tailoring solvent type and solvent evaporation time) porous at sheet membranes suitable for plasma fractionation. Although SlipSkin has been used before as a surface coating for several blood-contacting medical devices (catheters and guide wires) and its cytocompatibility and hemocompatibility have been documented, 12-14 the biocompatibility of SlipSkinÔ porous membranes has not been studied yet. In our opinion, this study is absolutely necessary since it is well known that the membrane biocompatibility can be seriously inuenced by the membrane processing conditions, surface roughness, pore and surface morphology, even if the membrane forming material has good biocompatibility. Therefore here, we also thoroughly investigate the biocompatibility of the new membranes based on the ISO guidelines 10993: "Biological evaluation of medical devices, part 4, selection of tests for interactions with blood". Following the advice of these guidelines, for external communication devices like hemodialysis equipment, we perform tests in all 5-test categories of thrombosis, coagulation, platelets, hematology and complement system. In all cases, we compare the performance of the newly developed membranes with several benchmark at membranes.
For biocompatibility tests, we selected several benchmark membranes for comparison with our SlipSkin membranes. PES/ PVP blends are oen used by the industry for the preparation of blood purication membranes. 4,15 Commercial hemodialysis or plasma fractionation membranes are always in hollow ber conguration, but for our hemocompatibility testing, at sheet membranes were necessary. Therefore we prepared at sheet (home-made) membranes, based on a PES/PVP polymer blend. Besides, commercial at sheet PES based membranes (14650-47-D, molecular weight cut off ¼ 50 000 Da, Sartorius, Nieuwegein, The Netherlands) were selected. A at sheet polyester membrane, which is in clinical use for blood transfusion applications, was selected as well and was taken out of blood transfusion lters SQ40SE (PALL, Portsmouth, UK). For complement activation experiments, two more benchmark materials, regenerated cellulose (RC58, Whatman, Sigma-Aldrich) and DEAE cellulose (DE81 DEAE cellulose, Whatman, Sigma-Aldrich), were used as positive control as described before. 16,17 Glass coverslips were obtained from VWR (Amsterdam, The Netherlands).
The following reagents for a thrombin generation assay were purchased from Thrombinoscope BV (Maastricht, The Netherlands). FluCa reagent contains 2.5 mM uorogenic substrate for thrombin (Z-Gly-Gly-Arg-AMC) and 87 mM calcium chloride. The MP reagent contains 24 mM phospholipids at 20 : 20 : 60 mol% PS : PE : PC. The thrombin calibrator contains a xed amount of thrombin-a2-macroglobulin complex. For the second thrombin generation assay, a Z-Gly-Gly-Arg-AMC uorogenic substrate was purchased from Bachem (Bubendorf, Switzerland), dimethyl sulfoxide was purchased from Acros Organics (Geel, Belgium), and calcium chloride was purchased from Merck (Darmstadt, Germany). For the platelet adhesion assay triton X100 was purchased from Acros Organics and CytoTox 96Ò non-radioactive cytotoxicity kit was purchased from Promega Madison (USA). Platelets were xed with glutaraldehyde from Sigma-Aldrich (Zwijndrecht, The Netherlands).

Membrane preparation
SlipSkinÔ membranes were prepared from 10 wt% SlipSkinÔ in acetone or NMP polymer dopes. Different types of SlipSkinÔ co-polymers were used with NVP : BMA molar ratios of 30 : 70, 50 : 50 and 70 : 30. Home-made PES membranes were prepared from 15 wt% PES and 7 wt% PVP in NMP. 18 The solutions were mixed on a roller bank, and aerwards degassed overnight at room temperature. All membranes were prepared by phase inversion. Solutions were cast onto a glass plate using a casting knife with a slit of 300 mm for the Slip-SkinÔ membranes and a 150 mm slit for the homemade PES membranes. Casting was immediately followed by immersion in the coagulation bath unless otherwise stated. Coagulation baths of ultrapure water and 60 wt% NMP and 40% ultrapure water were employed for the SlipSkinÔ and homemade PES membranes, respectively. In some cases, a solvent evaporation step of 30 seconds, 5 or 10 minutes was performed aer which the cast lms were immersed into the coagulation bath. Aer the membrane formation process, the membranes were rinsed and stored in ultrapure water for further use.
All membranes with accompanying codes and details are described in Table 1. SlipSkinÔ membranes are indicated with M, and the rst number in this code indicates the molar ratio of NVP : BMA building blocks in the polymer composition. When NMP was used as solvent for SlipSkinÔ, it was also indicated in the code while for the other SlipSkinÔ membranes, acetone was used as solvent. Furthermore, the evaporation step was indicated as 'ev' in the code for a solvent evaporation time of 30 seconds, while the solvent evaporation times of 5 and 10 minutes were indicated in the code as ev5 and ev10, respectively.

Membrane characterization
2.3.1. Scanning electron microscopy. For scanning electron microscopy (SEM), membranes were dried overnight in air. To obtain cross-sections, the dried membranes were broken cryogenically using liquid nitrogen. These, as well as dried membrane surfaces, were put on a SEM holder. Then all samples were dried overnight under vacuum and gold coated (BalzersUnion SCD 040 sputter coater, OerlikonBalzers, Belgium). Membrane cross-sections and surfaces were examined using a JEOL JSM-5600LV scanning electron microscope (JEOL, Japan).
2.3.2. Membrane transport properties. Clean water permeances were determined using an air-pressurized dead end 'Amicon type' ultraltration cell and ultrapure water. Membranes were pre-pressurized at 0.5 or 1 bar and then at least 4 pressure steps were applied for 30 minutes. The clean water ux (J) at each pressure was determined by taking the time and weight of permeated water and membrane surface area into account. The clean water permeance (L p ) of the membrane was calculated from the slope of the linear part of the clean water ux vs. transmembrane pressure relationship and was dened as where J represents the clean water ux expressed in L m À2 h À1 , DP is the transmembrane pressure difference in bar and L p is the clean water permeance in L m À2 h À1 bar À1 . The membrane sieving coefficient (SC) was determined using an air pressurized dead end Amicon ultraltration cell. The following compounds were dissolved in ultrapure water with initial concentrations of 0.1 mg mL À1 creatinine, 0.25 mg mL À1 myoglobin, 1 mg mL À1 BSA, 0.5 mg mL À1 g-globulin and 1 mg mL À1 thyroglobulin and pressurized through the membrane (active surface area: 12.6 cm 2 ) at 0.5 bar for 30 minutes. Then the concentrations of the feed and permeate compartments were analyzed using a photospectrometer (Varian, Cary 300 Scan UV-visible spectrophotometer) at 230, 409, 278, 278, and 280 nm for creatinine, myoglobin, BSA, g-globulin and thyroglobulin, respectively. The SC was calculated as follows: where C p is the concentration of the compound in the permeate and C f is the concentration in the feed. SC ¼ 1 means that the compound passes freely through the membrane, while SC ¼ 0 means that the membrane rejects the compound completely.

Biocompatibility tests
2.4.1. Thrombin generation. Fresh blood was obtained from 3 healthy human volunteers (donors 1, 2, and, 3; all volunteers gave informed consent; the study was conducted in agreement with the Declaration of Helsinki) and collected in Venosafe terumo citrated tubes (nal citrate concentration 3.2 w/v%) (Terumo Europe N.V., Leuven, Belgium).
Platelet poor plasma (PPP) was prepared by a centrifugation step at 2000 Â g for 5 minutes and a second centrifugation step at 10 000 Â g for 10 minutes. PPP was frozen at À80 C until further use. To investigate contact activation, circular membrane pieces (diameter ¼ 6 mm) and glass coverslips for 5 repeats were put on the bottom of a at bottom 96 well plate in ultrapure water. Also empty wells without membranes were used to incubate PPP as blank. Before addition of 120 mL PPP, the ultrapure water was aspirated. The PPP was incubated at room temperature under gentle shaking conditions for 15 minutes while no calcium ions or phospholipids were present. Then 80 mL of the incubated PPP was aspirated and transferred to round bottom 96 well plates. Also a control of plasma, which was not incubated, and a pooled normal plasma internal control were taken into account. 20 mL MP reagent was added to all wells and the plate was pre-warmed at 37 C for 10 minutes. The calibrated automated thrombogram (CAT) assay was performed as previously described. 19 Thrombin generation was started by addition of 20 mL FluCa reagent. The calibration was performed with a mixture of 80 mL plasma and 20 mL of the thrombin calibrator and 20 mL FluCa reagent. Fluorescence of the thrombin split uorogenic substrate was read in a Fluoroskan Ascent reader (Thermo Labsystems OY, Helsinki, Finland) equipped with a 390/460 lter set. Thrombin generation curves were generated with the Thrombinoscope soware (Thrombinoscope BV, Maastricht, The Netherlands). Also the data analysis was performed using this soware. The lag time and peak height were calculated from each separate curve.
For the second thrombin generation test, the bottom of the wells of a 96 well plate was covered with the test membranes and glass coverslips. Fresh blood was taken in Vacuette precitrated tubes from 3 healthy donors (donors 4, 5, and 6). The whole blood was anticoagulated with citrate (11 mM) and platelet rich plasma (PRP) was obtained by a centrifugation step at 200g for 15 minutes. The obtained PRP was aspirated and used within 2 h. The uorogenic substrate was added to the PRP and just before the start of the experiment, PRP was recalcied. 200 mL of this mixture was added to the wells and also empty wells were used as blank. This thrombin generation test was performed as described before. 14 Fluorescence was measured at 365/460 nm lter set with a SpectraMax M2 microplate reader (Molecular Devices, Sunnyvale, USA). For quantication, the time between the start of the experiment and the moment at which the thrombin concentration rst exceeds the 5 nM level was used (thrombin generation time).
2.4.2. Platelet adhesion and aggregation and lactate dehydrogenase assay. For these tests PRP was obtained from donors 4, 5 and 6 (all volunteers gave informed consent) as described in Section 2.4.1.
For qualitative platelet adhesion, PRP of two donors was used and membranes and glass coverslips were put on the bottom of a 24 well plate in ultrapure water. Before addition of 100 mL PRP (of donors 5 and 6) the ultrapure water was aspirated. The PRP was incubated for 45 minutes at 37 C. Then the PRP was removed. The test membranes and glass coverslips were washed in phosphate buffered saline (PBS) 3 times and then transferred to a clean vial and 900 mL of 2.5% glutaraldehyde was added and incubated for >1 h at 4 C. Aerwards, the test membranes and glass coverslips were washed in a 0.1 M phosphate buffer (pH ¼ 7.2) and with ultrapure water. Aer drying, the membranes and glass coverslips were put on a SEM holder, gold sputtered and examined using SEM, as described earlier.
Platelet adhesion was quantied using the lactate dehydrogenase (LDH) assay. 20 Membrane circles (diameter ¼ 6 mm) were immersed in ultrapure water (in duplo) in a 96 well plate, and 250 mL PRP (of donors 4, 5, and 6) was added and incubated at 37 C for 60 minutes aer aspiration of ultrapure water. Then, the PRP was removed, and the membrane samples were washed three times with PBS to remove non-adherent platelets. The membrane samples were transferred into clean vials and incubated for 60 minutes at room temperature in 200 mL lysis buffer. 50 mL of the supernatant was transferred in duplo to a 96 well plate and 50 mL substrate mix was added and this mixture was incubated for 30 minutes in the dark. Then, 50 mL stop solution was added and the optical density was measured at 490 nm on the SpectraMax M2 microplate reader (Molecular Devices, Sunnyvale, USA). Glass was not used as a control here because of non-accurate estimation of adhered platelets on glass. 20 Platelets were quantied by a standard curve plotted from samples containing known amounts of platelets measured with a Beckman Coulter Ac-T diff analyzer (California, USA).
2.4.3. Hematology tests. These tests were adapted, but based on previous literature. 21,22 Calcium ions are important for leukocyte adhesion, therefore heparin was selected as an anticoagulant rather than citrate. 23 Fresh blood was taken from 3 healthy donors (donors 1, 2, and 3) in 4 mL sodium heparin tubes (nal concentration: 68 IU) (Becton & Dickinson, Franklin Lakes, USA) and per donor pooled together and mixed by decanting. The bottoms of 24 well plates were covered with the test membranes (in triplicate) and 2Â washed in sterile physiological saline solution. 600 mL blood was added to each well aer aspiration of the saline, and incubated for 1 h at room temperature on a gentle shaker.
For the leukocyte adhesion test, empty wells were also taken into account for the incubation of the donor blood. Aer the incubation, blood was aspirated and analyzed using a Sysmex XE-5000 (Sysmex Europe GmbH, Germany) for leukocyte, basophil, eosinophil, neutrophil, monocyte and lymphocyte counts and hematocrit.
For the hemolysis test, additional wells were taken into account containing 600 mL saline or 600 mL sterile ultrapure water; these wells were used as negative and positive controls, respectively. 24 Aer the incubated step, blood was aspirated and centrifuged at 4000 rpm for 6 minutes. Light absorbance was measured at 542 nm on the supernatant to analyze the free hemoglobin concentration. The percentage of hemolysis (H) was obtained as follows: where C s is the concentration of free hemoglobin in the sample (mmol L À1 ), C neg is the concentration of free hemoglobin in the negative control (mmol L À1 ) while C pos is the concentration of free hemoglobin in the positive control (mmol L À1 ).

Complement tests.
The ISO guideline proposes among others a CH50 test for the complement activation category. To our knowledge, in all Dutch hospital laboratories, the CH50 test has been replaced by the classical pathway ELISA test. Therefore we used this ELISA test to investigate classical pathway complement activation. Since articial surfaces are known to activate complement rather via the alternative pathway than the classical pathway, we also performed an alternative pathway complement ELISA test.
The bottom of 24 well plates was covered with 15 mm circles of the test membranes in duplo and washed in sterile ultrapure water. Blood was taken from 3 healthy donors (donors 1, 2, and 3), collected in BD Vacutainer SSTII Advance tubes (BD Plymouth, UK) and le to clot for 30 minutes. Then serum was obtained by a centrifugation step at 2000g for 10 minutes and the collected serum was pooled together per donor. The incubation step procedure was based on ref. 21, our own experience and the amount of available serum. Aliquots were immediately frozen for the not incubated serum control. Furthermore, empty wells were taken into account for the blank incubation of serum. 800 mL serum was added per well and incubated for 1 h at 37 C under gentle shaking. Aer incubation, the serum was aspirated and directly analyzed further. The classical and alternative pathways were measured by ELISA using the Wieslab Complement system classical pathway and Wieslab complement system alternative pathway kits (Euro Diagnostica, Malmö, Sweden) according to the instructions of the manufacturer. 25 The internal controls with a normal and pathological pooled serum show results within the valid range conrming the validity of both ELISA tests. Although the mechanical stability of these membranes was not systematically studied, during handling we observed that the M70NMP membranes were very fragile, whereas the M50NMP and M30NMP membranes were less fragile. All these membranes are porous, and below a thin skin layer large macro voids throughout the membrane cross-section are present. The mutual affinity (or miscibility) between the solvent-non-solvent pair NMP-water is very high, leading to instantaneous demixing and macro void formation during membrane fabrication. [26][27][28][29] Fig. 2b presents SEM images of membranes formed using polymer dopes of 10 wt% SlipSkinÔ dissolved in acetone. The M30 (A) and M50 (B) membranes have a porous sponge like structure. M70 membranes are brittle and a representative SEM picture could not be obtained. The solvent-non solvent pair acetone-water has lower mutual affinity than NMP-water and therefore delayed onset of de-mixing takes place, leading to macro void free membranes. 27 Similar effects of solvents on membrane structures made from different polymers have been described before. For example, using water as non-solvent, a dope of PMMA dissolved in NMP gave nger like macro void structures while a dope of PMMA dissolved in acetone resulted in a symmetric porous sponge like structure. 28 Besides, it seems that M50 shrinks more during the phase inversion process than the M30 membrane. This may be due to the lower amount of BMA present in the M50 membrane, which is the hydrophobic building block providing strength to the membrane during solidication. Fig. 2c shows SEM images of membranes formed from a 10 wt% 50 : 50 SlipSkinÔ in acetone with no solvent evaporation steps (A, B, and C), or solvent evaporation for 30 seconds (D, E and F), 5 minutes (G, H, and I) and 10 minutes (J, K, and L). All membranes seem to have a porous structure. The M50ev membranes with 30 second solvent evaporation seem to have bigger pores in the cross-sectional image compared to the M50 membranes. However, the surface images of the M50ev membranes show somewhat smaller pores. Due to the solvent evaporation the membranes with 5 and 10 minute evaporation steps seem to have pores through the membrane thickness but it looks like they are closed cellular pores without good pore interconnectivity.

Membrane fabrication
Macroscopically, shrinkage was observed while drying the membranes in air. Also, the thicknesses in SEM pictures are smaller than the thicknesses of the membranes measured in a wet state using the micrometer, see Table 2. As expected, thinner membranes are formed aer longer solvent evaporation. However, there was no difference in thickness between the membranes obtained aer the 5 and 10 minute evaporation step.

Membrane transport properties
3.2.1. Clean water ux measurements. Fig. 3a presents the clean water ux-pressure relationship of the membranes formed using the solvent NMP. The M30NMP membranes have a clean water permeance of 2226 AE 210 L À1 m À2 h À1 bar À1 and the clean water ux-pressure relationship is linear. The M70NMP membranes have a clean water permeance in a similar range, 2485 AE 671 L À1 m À2 h À1 bar À1 . This clean water permeance is based on the ux-pressure relationship up to 0.2 bar since at higher pressures this relationship is not linear. Probably the membrane compacts at higher pressures because of poor mechanical stability. The M50NMP membranes have a lower clean water permeance of 570 AE 121 L À1 m À2 h À1 bar À1 with a linear ux-pressure relationship and without any indications of membrane compaction. Fig. 3b presents the clean water ux versus pressure relationship for the M50 and M50ev membranes. The M50 membranes have a clean water permeance of 186 AE 14 L À1 m À2 h À1 bar À1 and the ux-pressure relationship is linear. Although the SEM pictures show a porous structure, the M30 membranes show no water permeation aer applying a pressure of 1 bar for several hours. The M70 membranes have poor mechanical stability and could not be tested for the clean water permeance. The M50ev membranes have a lower clean water permeance of 112 AE 20 L À1 m À2 h À1 bar À1 ( p < 0.05) compared to the M50 membranes fabricated without a solvent evaporation step probably due to membrane densication. The ux-pressure relationship there is linear. Membranes M50ev5 and M50ev10 fabricated with 5 minute and 10 minute evaporation have no water permeability when they were tested for several hours. This is probably due to the lack of pore interconnectivity, consistent with the SEM analysis of these membranes, which mainly shows a closed cell pore structure. Since our target here is to develop ultraltration membranes with good mechanical properties we continued further investigation of sieving coefficients and biocompatibility only for the M50 and M50ev membranes.
3.2.2. Sieving curve. Fig. 3c presents the sieving curves of the M50 and M50ev membranes. Because of the different clean water permeances of these membranes one would also expect a difference in their sieving curves, since the pore size can have a large impact on water permeability. 30,31 However, their sieving curves appear to be similar. For the molecules creatinine (113 Da) and myoglobin (17.8 kDa) the SC is approximately 1, while the BSA (66 kDa) is around 0.83. For larger molecules like g-globulin (150 kDa) and thyroglobulin (670 kDa) the SC is around 0.52 and 0.10, respectively. We speculate that this maybe due to a decreased number of pores (at the surface or throughout the membrane structure) due to the evaporation step for the M50ev membranes, leading to lower clean water permeance while the pore size stays the same, leading to a similar sieving curve compared to M50 membranes. It is also important to note that our M50 and M50ev membranes have a quite sharp separation curve and separation characteristics  similar to some commercial lters in plasma fractionation treatments. 32,33 For example Evaux 4A secondary lters show sieving coefficients for albumin of $0.8 and for IgM (950 kDa) of $0.1, close to our sieving coefficients ($0.8 for BSA and $0.1 for thyroglobulin). 34 Evaux 4A can be successfully applied for treatment of for example hyperglobulinemia where large globulins such as IgM are removed. 35 For most of the commercial membranes however, there is hardly any information available in the scientic literature about biocompatibility. In the next section we investigate extensively the biocompatibility properties of our SlipSkinÔ membranes M50 and M50ev and compare this to several benchmark materials.

Biocompatibility results
We performed several biocompatibility tests. The tests are selected based on the ISO guidelines. For the category of coagulation, two different thrombin generation tests were performed. In the category of thrombosis, platelet adhesion and aggregation were analyzed by SEM, while a LDH assay was performed for a quantitative analysis of platelet adhesion in the category of platelets. Leukocyte adhesion experiments and hemolysis evaluation were performed for the hematology category. For the complement category, we performed a classical and alternative pathway complement test.
3.3.1. Thrombin generation. Thrombin plays a central role in the coagulation cascade, and converts brinogen into brin so that an actual clot can be formed. 36 Furthermore it inuences platelet aggregation and endothelial cell function. In this assay, the intrinsic coagulation cascade is responsible for the generation of thrombin, since there is no tissue factor present in the test system. The membrane can lead to contact activation aer which the intrinsic coagulation cascade starts. During the incubation period with the membranes there is no calcium or phospholipids present and the coagulation cascade will hold. Aer addition of phospholipids and calcium ions, the intrinsic pathway will be continued and eventually thrombin will be generated. The peak height of generated thrombin is an indicator for the contact activation caused by the responsible material. Fig. 4a shows typical thrombin generation curves of the positive control glass, the M50 membrane and the negative control blank (incubated PPP). The time until the rapid increase in thrombin concentration (thrombin generation time, TGT) as well as the peak height is analyzed and the peak height gives the most distinguishable and consistent results. The SlipSkinÔ M50 membrane peak height is between the peak height values of the positive control glass and the negative control blank. Fig. 4b presents the average peak height of the thrombin concentration (nM) of donors 1, 2, and 3. The not incubated PPP shows similar results to the incubated plasma, the blank. The thrombin peak height of our SlipSkinÔ membranes M50 and M50ev is similar to the benchmark materials and much lower than the glass (positive control) showing that the SlipSkinÔ membranes have excellent properties concerning contact activation. Besides the above, we also measured thrombin generation real time, where PRP was incubated on the membrane in the presence of calcium. In this assay, the thrombin generation time was used for further analysis and was an indicator for the thrombogenicity of the material. 14,37 Fig. 5a shows typical results of thrombin generation curves over time for the M50 membrane and negative (blank) and positive (glass) control. Fig. 5b presents the results of the thrombin generation times and again, the SlipSkinÔ membranes M50 and M50ev show very good properties, comparable to currently clinical in use PE membranes and other benchmark membranes in terms of thrombogenicity.
3.3.2. Platelet adhesion and aggregation. Blood platelets are important during coagulation. They can form a platelet plug to stop bleeding, their presence is necessary for the coagulation cascade and activated platelets can release biologically active compounds. 36 Platelet adhesion and activation on hemodialysis membranes can lead to reactive oxygen species generation by neutrophils. 38 Here, we have performed two tests: SEM image analysis for a qualitative measure for platelet adhesion and aggregation, while quantitative results about platelet adhesion were obtained using the LDH platelet adhesion assay. 20 Fig. 6a shows SEM pictures of platelets, which adhered onto the tested membranes and glass. Both SlipSkinÔ membranes (M50 and M50ev) (C, D, E, and F) show little platelet adhesion, no severe cluster formation and the adhered platelets are mostly round with little or no pseudopodia formation. Besides there are hardly any deposits. On glass, single platelets and small clusters are adhered with mostly long pseudopodia and small amounts of deposits (A and B). The PESh membranes (G and H) show deposits on the surface, where the platelets are mostly round with small pseudopodia but no clusters are observed. Fig. 6a (I and J) show clusters of platelets with pseudopodia on the PESc surface, and a lot of deposits on the membrane surface. On the PE blood transfusion lter, round and at platelets are observed, mostly without pseudopodia, although some platelets have long pseudopodia (K and L). Fig. 6b presents the average adhered platelets per cm 2 material. In every donor a similar trend can be observed. Only a few platelets adhere to the M50 and M50ev membranes, whereas many platelets seem to  adhere to the other benchmark materials, especially to the PESc. The low platelet adhesion to the M50 and M50ev membranes could be due to the formation of a mosaic of hydrophilic and hydrophobic regions on the membrane surface leading to relatively low protein adhesion and thereby low platelet adhesion. The search for this optimal mosaic structure is one of the major reasons for investigating membrane fabrication using copolymers.
3.3.3. Leukocyte adhesion. Leukocytes are one of the major players in the inammatory response, and leukocyte activation can result in release of inammatory mediators. White blood cells can adhere to hemodialysis membranes 39 and hemodialysis is associated with production of cytokines by white blood cells. 40 Fig. 7 shows the average number of leukocytes present in blood aer incubation with various membranes. The rst bar represents the number of leukocytes in not incubated blood. All donors are in the normal range of leukocyte numbers although donor 1 is on the upper range of the spectrum. The number of leukocytes in blood is similar to the number of leukocytes in blank, which indicates that there is hardly undesired adhesion of leukocytes on the well plate material. Generally, there is hardly any leukocyte adhesion and the M50 and M50ev membranes perform equally well to clinical in use PE membranes and the other benchmark membranes PESh and PESc. Based on the hematocrit levels we conclude that there is no evaporation (data not shown), which could lead to elevated leukocyte concentrations in the blood and misinterpretation of the data. Besides, the following categories of leukocytes were separately analyzed as well: monocytes, basophils, eosinophils, neutrophils and lymphocytes. No major adhesion of any of these cell types was observed for all the membranes (data not shown).
3.3.4. Hemolysis. Hemolysis is oen used as a screening test to assess biocompatibility. 41 Hemolysis less than 5% is regarded as non-toxic according to the ASTM F-756-08 standard. 24 Table 3 presents the average percentage of hemolysis of the tested membranes. The M50 and M50ev SlipSkinÔ membranes did not induce toxic levels of hemolysis in this in vitro hemolysis-screening test and show comparable good results to the benchmark materials PESh and PESc and the clinical in use PE membranes.
3.3.5. Complement. The complement system is an important part of the body's innate immune system. 42 Three pathways are known for the activation of this system, the classical pathway, alternative pathway and mannose binding lectin pathway. All these pathways eventually cause effect via the formation of opsonisation factors, which facilitate phagocytosis, production of leukocyte attracting factors and the production of the membrane attack complex which can lyse e.g. bacteria. 42 Fig . 8 shows the average % complement activity of the classical (Fig. 8a) and alternative (Fig. 8b) pathways of the complement system. The not incubated serum and the blank (incubated serum) show similar results in both cases. This indicates that the incubation step itself did not lead to complement activation; therefore a decrease in complement activity in these assays would indicate a relative deciency of any of the complement components, due to complement activation via the responsible pathway by the membrane. Overall, there was no activation of the classical pathway of the complement system by any of them, see Fig. 8a. In donors 1 and 3 all the materials show values close to 100%   complement activity. In donor 2, the complement activity for DEAE cellulose seems to be lower than the rest, but this was not observed in the other donors. Activation of the classical pathway usually occurs via immune complexes rather than articial surfaces. The alternative pathway of the complement system, on the other hand, can be activated by articial surfaces such as membranes.

Conclusions
This work presents the preparation and characterization of at sheet SlipSkinÔ membranes with excellent biocompatibility. In our tests these membranes have lower platelet adhesion and have equivalently good properties in relation to contact activation, thrombogenicity based on thrombin generation, leukocyte adhesion, hemolysis and complement activation compared to benchmark materials PESh, PESc and clinical in use PE membranes.
The new membranes have a sieving curve similar to cascade lters in plasma fractionation applications. In the future we plan to develop SlipSkinÔ hollow ber membranes suitable for hemodialysis with a lower molecular weight cut-off so that albumin is retained. Besides membranes purely made from SlipSkinÔ, this material may be used for preparing porous coatings of existing membrane supports, so that it can offer its biocompatibility characteristics while the mechanical stability can be provided by a support membrane material. Furthermore, the very promising biocompatibility results of our membranes make it interesting to further investigate their biocompatibility under ow conditions.

Conflict-of-interest notification
The authors Y. Aldenhoff, L. H. Koole and J. Janssen declare that they are shareholders of the company INterface BIOmaterials BV (Geleen, the Netherlands) that manufactures and commercializes the SlipSkinÔ biomaterial.