Extracorporeal membrane oxygenation (ECMO) is widely used in cases of respiratory failure that cannot be treated by conventional techniques . Membrane oxygenators are the commonly used type being less traumatic to the blood compared to the formerly used bubble or film-type oxygenators. In membrane oxygenators gas exchange takes place due to the oxygen concentration gradient along semi-permeable membranes that separate the gas phase from the blood phase. Because of this indirect oxygen blood contact membrane oxygenators imply a low risk of embolism . One of the greatest disadvantages of membrane oxygenation is the large foreign body area that is needed for sufficient gas transfer. The highest membrane surface area to priming volume ration can be accomplished through the use of fine hollow fibers in parallel array . More than 60% of all used membrane oxygenators are equipped with hollow fiber membranes for gas transfer .
Despite all efforts during the last decades, operating period of ECMO systems is still limited to a few weeks or, at best, months . Long-term application is impossible because of limited hemocompatibility, protein adsorption and bleeding induced by anticoagulation therapy. A tissue-engineering approach to extend the application period of ECMO is to apply a layer of endothelial cells onto the gas exchange membrane surfaces. This is to prevent inflammation, coagulation and a loss of gas transfer capacity caused by protein deposition over time . Hollow fiber membranes are not well suited for endothelialization because there is a lot of friction between the single fibers, whereby the cells are rubbed of the fiber surfaces. Consequently, flat membranes are used for implementation of an endothelialized oxygenator. One of the critical success factors for the future of this kind of biohybrid lung support systems is the increase of the gas transfer rates by developing suitable flat membrane assemblies, as it was found, that endothelialization of flat membranes decreases the gas transfer performance of flat membrane oxygenators .
Performance of flat membrane oxygenators is not efficiently, due to the thickness of the blood film as well as the laminar flow on the membranes : The wall friction leads to a flow profile with high flow rates in the middle of the channel and a slower blood flow close to the membrane surface. As a result these boundary layers work as a barrier for the oxygen to pass into the main flow as displayed in Figure 1A .
Higher gas transfer rates can be achieved by effective mixing of the blood phase. One approach to gain effective mixing was found by B.J. Bellhouse et al.  around 1960. It was described, that imprinting hollows onto the surfaces of flat membranes combined with a pulsatile blood flow causes vortices in the otherwise laminar flow. This effect is shown in Figure 1B.
The rates of gas transfer can be varied by the size of these hollows and the strength of the secondary flows induced . An appropriate membrane design was characterized to have a furrow wave length between 1.8 and 2.0 mm, furrow depth around 0.35–0.4 mm and a channel width of 0.4–0.6 mm [8–10]. Using these parameters an almost homogenous mixture of blood could be achieved. The application of the Bellhouse effect in the development of a biohybrid lung is the topic of this study.
A process to produce membrane assemblies for oxygenators suitable for endothelialization could be established. Therefore a sandwich structure (Figure 2) was developed comprising a textile spacer structure between two flat membrane sheets. To achieve higher gas transfer rates of the flat membrane oxygenator, the membrane surfaces were provided with hollows to profit from the so-called Bellhouse effect.
For imprinting these hollows onto the membranes two totally new procedures were generated.
Hollow imprinting for furrowed membrane assemblies
The aim of the study was the application of warp knitted spacer fabrics for the preparation of a new type of flat membranes in a biohybrid lung. Therefore a membrane, namely the lumox membrane, that is suitable for cell culturing was assembled to both sides of the spacer fabric. For assembly a vacuum-based and a pressure-based approach were applied. All parameters used for this hollow imprinting in this study are listed in Table 1.
Computed tomography (μ-CT) images (Figure 3) from the negative form of the membrane surfaces show that the two applied principles of hollow imprinting, by using vacuum as well as pressure, are feasible to produce furrowed membranes. The geometric form of the hollows produced by using pressure is less defined than the ones produced using vacuum. This is because the enclosed air between the membranes keeps the membranes from being as deeply pressed into the textile pores over the total textile structure. The dimension of the maximum hollow achieved with both procedures is on a comparable level.
It was demonstrated, that the depth of hollows imprinted is linked to the temperature and the time period. For imprinting hollows by vacuum a maximum hollow depth of 0.35 mm could be achieved at 75 °C and 5 min duration, which fits perfectly to the requirements defined by Bellhouse et al. Results for different parameters of the vacuum set-up are displayed in Figure 4.
The results for the pressure-imprinting approach are shown in Figure 5. The hollow depth ranges from approximately 0.19 mm achieved heat-setting at 70 °C to approx. 0.38 mm at 80 °C. Best results for the pressure set-up could be achieved using a heat-setting temperature of 80 °C for 5 min.
Evaluation of the endothelialization
HUVECs were seeded on lumox membrane and on gelatin as reference. The cells show the typical “cobblestone” pattern (Figure 6), with areas of confluence on the lumox membrane but also with areas, without any cell growth. The initial seeding density was 100,000 cells/mL.
Cell density of methanol-fixed HUVECs was determined after 3 days of cultivation in 24 well plates with Cell Profiler software after staining with DAPI (Figure 7).
The highest density and narrowest distribution of HUVECs was observed by cultivation on gelatin with the box located between 861 and 1048 cells/mm2. On lumox membrane the cell density is between 174 and 613 cells/mm2, so the distribution is much broader as on gelatin, which is probably caused by the inhomogeneity of cell growth.
As it was observed that some spots on the lumox membrane remained free and cells did not tend to grow there (Figure 6) contact angle measurements were performed in order to analyze the character of the surface and to evaluate the quality of the supplied material. The lumox film displays two different sides, one side is modified to enable cell culturing, which we call the cell contact side. The other side – back side – is not suitable for cell growth and probably reveals the bulk material. More details are not published by the manufacturer. Analysis of the two sides clearly shows the difference between the back and the cell contact side (see Figure 8).
The mean contact angle of the back side is 112°, with a homogeneous distribution. Instead, the contact angle of the cell contact side ranges between 48° and 90°. By comparing the results for the different spots on the cell contact side, it can be stated, that this side has two areas: One area shows a contact angle lower than 60°, the other area shows a contact angle around 87°. These findings are in accordance with the cell culture results.
Selection of spacer textiles
Flat membrane oxygenators often consist of alternating layers of sheet membranes and screen spacers. The screen spacers thereby provide a support for the thin membrane films, maintain a uniform distance between the membranes and create tortuous flow paths for gas and blood, which induce lateral mixing . The disadvantage of this way of mixing the blood phase is a high blood trauma which is why mixing of blood is in this study provided by imprinted hollows on the membrane surfaces. Nevertheless there still is a need for spacers in the gas phase. For efficiency reasons these spacers are also used to provide the mold for imprinting hollows onto the membranes. Characteristic aspects of suitable spacers are a high horizontal gas permeability to enable homogenous gas flow and a surface structure that provides molds for hollow imprinting. Textiles are recommended as spacers because their geometric properties are adjustable within wide limits . Different textile technologies can be used to get a textile that fulfills the defined requirements best. Nonwovens can be manufactured highly porous, which leads to a high horizontal permeability, but it is difficult to achieve a surface structure that can be used as mold for hollow imprinting. The surface structure of woven fabrics as well as warp knitted fabrics can be varied by the selection of weave constructions. As circular hollows are preferred, warp knitted fabrics are better suited than woven fabrics to be used as spacer. This textile is not expected to result in the required size of dimples referring to Bellhouse et al., but as a warp knitted spacer fabric it has a high horizontal permeability and can be used to associate the employed pore size to the resulting hollow size. This fabric can be used as a first part result on the way to find the most appropriate spacer structure.
Comparison of assembly procedures
Using pressure for hollow imprinting instead of vacuum on the other hand has the advantage that the membranes do not have to be punctured. On top of that there is the possibility to imprint the hollows after assembling the oxygenator device, so that the process itself causes less effort.
Table 2 lists the advantages and disadvantages of the two different imprinting procedures.
It was possible to show that the employed textile pore size is directly linked to the size of the hollows that emerge on the membrane surfaces. Using the warp knitted spacer fabric. The measured pore size was approximately the same as the textile pore size of 3.6 mm×5.8 mm.
Further research concerning the parameters and the process is needed to improve the shaping quality especially for imprinting hollows by pressure.
Assessment of cell seeding experiments
This study is part of a proof of principle on the applicability of furrowed membranes for the development of a biohybrid lung. The selected membrane for cell seeding, namely the lumox membrane, is to our knowledge the only available material with good cell adhesion properties and good gas-permeability. Nevertheless, our results show that a confluent layer of cells cannot be reached all over the surface due to the inhomogeneity of the surface modification. Beside these areas where no cells attach, endothelial cells are nicely grown and show the typical “cobblestone” morphology. Future experiments will show, if a confluent cell layer can be reached by an even more carefully treatment of the membrane. The next step will be to evaluate the cell cultivation under dynamic conditions and the determination of the gas transfer over the cell-membrane interface.
All in all it can be said, that it was so far possible to produce furrowed membrane assemblies that could be used for endothelialization. Still neither the selected textile and the developed imprinting techniques nor the cell-cultivation could result in perfect achievements, but can be used as first part results on the way to a biohybrid lung.
Materials and methods
Materials for furrowed membrane assemblies
A warp knitted spacer fabric (Pressless GmbH, Flöha, Germany), pictured in Figure 9, is selected for implementation into a flat membrane assembly: The employed fabric is made of polyester (PES), has a thickness of 2 mm and pore sizes about 3.6 mm×5.8 mm.
An ultra-thin, gas-permeable cell culture film (Lumox, foil, Sarstedt AG & Co., Nümbrecht, Germany) is used as cell contact membrane.
Hollow imprinting by vacuum
For the production of membrane assemblies suitable for endothelialization, one spacer structure is put between two flat cell contact membranes, which are welded together using an impulse welding machine (Gambro Instrumenta AB, Lund, Sweden). The approach is displayed in Figure 10: A cannula is inserted and a vacuum of approximately 0.35 bar is applied between the two membrane sheets, so that the membranes are sucked into the pores of the textile structure and dimples emerge on the membrane surfaces. Subsequently heat setting is processed in a water bath at 60 °C–75 °C.
Hollow imprinting by pressure
The set-up for hollow imprinting by pressure is shown in Figure 11. The membrane assembly is put into a uniquely designed frame that is itself put into a self-constructed, water filled pressure chamber. The chamber is tightly closed and the water inside subsequently heated up. Pressure of two bar is applied using a valve on top of the pressure chamber for about 5 min.
μ-CT (Ct alpha, Procon X Ray, Sarstedt, Germany) is performed of molds of the shaped membrane surfaces. A special feature of this μ-CT is the helical recording geometry, which results in a better resolution compared to conventional μ-CT images. To recieve a negative form of the imprinted structure the surfaces are molded with a two component silicone (Köraform 50, POLYchem Handelsges. m.b.H., Markt Allhau, Austria). Layer images from this negative form are measured using the software CorelDrawX5 (Corel GmbH, Munich, Germany).
Cell isolation and cultivation
Human umbilical cord vein endothelial cells (HUVECs) were used for this experiment. Cells were isolated from newborn umbilical cords following a protocol from Baudin et al. . Briefly, cells are isolated under sterile conditions by cannulating the vein with a blunt needle and fixation with a surgical clamping clip. The vein is flushed with a phosphate buffered saline (PBS) (Life Technologies, Carlsbad, CA, USA)-filled syringe until blood is removed. The second blunt end is introduced into of the other end of the vein and fixed again. After cannulation, collagenase (0.2% w/v, Life Technologies, Carlsbad, CA, USA) is inserted. Both sides are closed with stoppers and the umbilical cord is incubated (30 min at 37 °C). Cells are collected by flushing the vein with 15 mL PBS. The cells are centrifuged and resuspended in 10 mL endothelial basal medium (EBM-2) supplemented with endothelial cell growth medium (EGM-2, both Lonza, Basel, Switzerland) and seeded on 2% gelatin (from bovine skin, type B, Sigma-Aldrich, St. Louis, MO, USA) pre-coated tissue culture flasks in EBM-2/EGM-2. Cells maintain in a humidified incubator at 37 °C and 5% CO2. Upon 80% confluence, cells are serially passaged using a trypsin/EDTA solution in PBS (trypsin 0.05%, PAN Biotech GmbH, Aidenbach, Germany; Ethylendiamine tetraacetic acid disodium dihydrate/EDTA 0.02%, Sigma-Aldrich, St. Louis, MO, USA). Cells are cryopreserved by standard methods in passage 2, thawed and applied as cell pool in passage 4 for this study.
Cell seeding and pooling
A cell pool is obtained by pooling at least three different cell lines (n=5 for static conditions, n=3 for dynamic cultivation) in the same ratio, (e.g. 1 million living cells per cell line). Cells are suspended in EBM-2/EGM-2 and seeded on lumox membrane at a cell density of 100.000 cells/mL using CellCrowns (Scaffdex Oy, Tampere, Finland) or silicone rubber (ADDV M 4641, R&G Faserverbundwerkstoffe GmbH, Waldenbusch, Germany) for fixation of the membrane. Gelatin coated wells served as control samples. Cells were maintained in a humidified incubator at 37 °C and 5% CO2. Medium was changed on day 2.
Cells are fixed with methanol (–20 °C, 10 min; VWR International, Radnor, USA), three times rinsed with PBS and unspecific bonds blocked with 200 μL 3% bovine serum albumin (BSA, Sigma-Aldrich, St. Louis, MO, USA) for 30 min. After removal of blocking solution, 200 μL primary antibody is added, except for the negative control, which is filled with 1 mL washing liquid. Incubation time is 1 h at 37 °C. Secondary antibody is added after rising two times with washing liquid and incubated for 1 h at 37 °C in the dark. After washing twice with PBS or washing buffer, respectively, cells are counterstained with DAPI (4′, 6-Diamidin-2-phenylindol, Carl Roth GmbH & Co. KG, Karlsruhe, Germany) for 10 min and rinsed three times with PBS. Staining-specific applied antibodies and washing liquids are listed in Table 3.
Cell density and determination of focal adhesion
CellProfiler 2.2.1 software (Broad Institute, Cambridge, MA, USA) was used to identify and count DAPI-stained nuclei of whole well diagonals (13–25 images per well) . First, gray scaled images were cropped to a size of 1000 times 1000 pixels (“Crop”). Afterwards an enhancement step (“EnhanceOrSuppressFeatures”) of speckles (size 70) for higher contrast of nuclei on the rough surface (lummox) was added. Objects were identified with a global threshold strategy and the MoG thresholding method (“IdentifyPrimaryObjects”). Typical object parameter were set to minimal 14 pixels up to a maximum of 70 pixels. The software in return counted the objects, which were in this case the nuclei. Lastly data was exported (“ExportToSpreadsheet”). By additionally measuring the image area with AxioVision Rel 4.9, area-dependent cell densities were calculated.
Contact angle measurement
Measurements are performed with pocket Goniometer PG X (FIBRO System AB, Sweden) and Software PG, (Version 3.5) by the sessile drop method. The contact angle angle was determined on two different batches with 4 μL H2O on at least five spots.
Conventional light microscopy and fluorescence microscopy
Untreated samples were analyzed with routine bright field microscopy (Axio Imager; Carl Zeiss GmbH, Germany) and scaled images were acquired using a digital camera (AxioCam MRm; Carl Zeiss GmbH). Fluorescnece microscopy was performed using an Observer.Z1 Zeiss fluorescence microscope and the software AxioVision Rel 4.9 (both Carl Zeiss Microscopy GmbH, Oberkochen, Germany). Focusing was reached by DAPI-stained nuclei and exposure times were determined by the negative control (Table 3). Modification of microscopic images was done with AxioVision Rel. 4.9.
This study was funded by the Aachen Interdisciplinary Center for Clinical Research (IZKF) at University Hospital RWTH Aachen and by the Excellence Initiative of the German federal and state governments in the framework of the i3tm Seed Fund Program. The authors thank Jonas Rose and Mirco Katzenmeyer for assistance in cell isolation and culture and Sarah Menzel for performing contact angle measurements. The authors carry the responsibility for the content of this publication.
Bellhouse BJ, Bellhouse FH, Curl CM, Macmillan TI, Gunning AJ, Spratt EH, et al. A high efficiency in membrane oxygenator and pulsatile pumping system and its application to animal trials; Vol. XIX Tras. Am Soc Artif Int Organs 1973;19:72–9.CrossrefGoogle Scholar
Carpenter AE, Jones TR, Lamprecht MR, Clarke C, Kang IH, Friman O, et al. CellProfiler: image analysis software for identifying and quantifying cell phenotypes. Genome Biol 2006;7:R100.CrossrefGoogle Scholar
Cornelissen CG, Dietrich M, Gromann K, Frese J, Krueger S, Sachweh JS, et al. Fibronectin coating of oxygenator membranes enhances endothelial cell attachment. BioMed Eng OnLine 2013;12:7.Web of ScienceCrossrefGoogle Scholar
Cornelissen CG, Engwicht S, Finocchiaro N, Jockenhoevel S. ENDOXY-Gas transfer of endothelialized oxygenator membranes. Eur Res J 2014;44(Suppl 58):258.Google Scholar
Dorrington KL, Ralph ME, Bellhouse BJ, Sykes MK. Oxygen and CO2 transfer of a polypropylene dimpled membrane lung with variable secondary flows. J Biomed End 1985;7:89–99.Google Scholar
Gries T, Veit D, Wulfhorst B. Textile Fertigungsverfahren. München: Carl Hanser Verlag GmbH & Co. KG, 2014.Google Scholar
MacLaren G, Combes A, Bartlett RH. Contemporary extracorporeal membrane oxygenation for adult respiratory failure: life support in the new era. Intensive Care Med 2012;38:210–20.Google Scholar
Riccabona M. ECMO (Neonate, Pediatric) imaging. In: Baert AL, editor. Encyclopedia of diagnostic imaging, 2008:661–4.Google Scholar
Schmid C, Philipp A. Herz-Lungen-Maschine: Aufbau und Funktion. In: Schmid C, Philipp A, editors. Leitfaden extrakorporale Zirkulation. Heidelberg, Germany: Springer-Verlag, 2011:72–104.Google Scholar
About the article
Published Online: 2015-11-12
Published in Print: 2015-11-01