Pressurized carbon dioxide as a potential tool for decellularization of pulmonary arteries for transplant purposes

Vascular bio-scaffolds produced from decellularized tissue offer a promising material for treatment of several types of cardiovascular diseases. These materials have the potential to maintain the functional properties of the extracellular matrix (ECM), and allow for growth and remodeling in vivo. The most commonly used methods for decellularization are based on chemicals and enzymes combinations, which often damage the ECM and cause cytotoxic effects in vivo. Mild methods involving pressurized CO2-ethanol (EtOH)-based fluids, in a supercritical or near supercritical state, have been studied for decellularization of cardiovascular tissue, but results are controversial. Moreover, data are lacking on the amount and type of lipids remaining in the tissue. Here we show that pressurized CO2-EtOH-H2O fluids (average molar composition, ΧCO2 0.91) yielded close to complete removal of lipids from porcine pulmonary arteries, including a notably decrease of pro-inflammatory fatty acids. Pressurized CO2-limonene fluids (ΧCO2 0.88) and neat supercritical CO2 (scCO2) achieved the removal of 90% of triacylglycerides. Moreover, treatment of tissue with pressurized CO2-limonene followed by enzyme treatment, resulted in efficient DNA removal. The structure of elastic fibers was preserved after pressurized treatment, regardless solvent composition. In conclusion, pressurized CO2-ethanol fluids offer an efficient tool for delipidation in bio-scaffold production, while pressurized CO2-limonene fluids facilitate subsequent enzymatic removal of DNA.

Two full-factorial design, one for each cosolvent, with three center points, were created in MODDE 10.1 (Sartorius Stedim Biotech, Malmö, Sweden) to investigate the impact of pressure (15.0-30.0 MPa), time (90-120 min) and temperature (35-40 °C) on tissue decellularization (Table 1). Higher pressures are obtained by adding more CO 2 to the mixture, thus modifying the composition of the mixture. To simplify the DoE, we have investigated pressure as a variable while we have given composition a fictitious constant value, reflected as the average molar fraction.
Total amount of remaining lipids per lipid class (μg of lipid class per mg of freeze-dried tissue) and remaining DNA were used as response variables to estimate the delipidation and DNA removal efficacy, respectively. PA pieces were subjected to neat scCO 2 treatment for 120 min at 30.0 MPa and 40 °C.
Enzymatic DNA removal. Tissue samples subjected to neat scCO 2 and pressurized CO 2 -cosolvent fluids were allowed to equilibrate in enzyme buffer (20 mM tris(hydroxymethyl)aminomethane, 20 mM NaCl, 2 mM MgCl 2 ) for 30 min, followed by treatment with 90 U/mL of benzonase endonuclease for 30 min at 37 °C, 1 mL per 5 mm 2 sample. Samples were subsequently washed three times in PBS during a total of 44 h, followed by fixation in formalin or quantification of residual DNA. evaluation of nuclei removal. The nuclei remaining in the specimen after treatment were estimated by hematoxylin and eosin (H&E) staining and quantification of residual DNA.
Hematoxylin and eosin staining of treated pulmonary arteries from pigs. Formalin fixed samples were dehydrated, embedded in paraffin and sectioned with a thickness of 4 µm. Sections corresponding to the central portion of the sample were selected for H&E staining. Untreated tissue as well as tissue submitted to neat scCO 2 were used as references.
DNA quantification. Residual double stranded DNA (dsDNA) were quantified by fluorescent nuclei acid staining using the Quant-iT TM PicoGreen TM dsDNA Assay Kit (Molecular Probes, Inc., Eugene, OR). Samples were lyophilized and homogenized using a Omni Tissue Homogenizer (Omni, Kennesaw, GA) followed by incubation with 200 U/mL of Proteinase K (Sigma-Aldrich) for 16 h at 37 °C. Samples were then centrifuged at 2000 × g for 10 min and dsDNA quantified in the supernatant (ng dsDNA/initial wet weight of the sample) according to manufacturer's instructions.  www.nature.com/scientificreports www.nature.com/scientificreports/ Fresh tissue submitted to detergent based decellularization 33 were used as positive controls. Briefly, pieces of PA were treated with a combination of 0.5% sodium deoxycholate (SDC) and 0.5% sodium dodecyl sulfate (SDS) for 24 h at room temperature with constant shaking. Samples were subsequently washed three times in phosphate buffered saline (PBS) for a total of 42 h, followed by fixation in formalin or quantification of residual DNA. extraction of lipid residues from treated tissues. Lipids were recovered by a dichloromethane/ methanol/water-based extraction method adapted for porcine pulmonary arteries 34 . Briefly, freeze-dried tissue pieces (see Supplementary Fig. S1) (in randomly assembled batches of 20 samples) were disrupted in a Qiagen TissueLyser (Qiagen GmbH, Hilden, Germany) for 10 min (1 min per cycle) at 25 Hz, followed by extraction of lipids as previously described in detail 34 . Extracts were dried under a stream of nitrogen gas, weighted for gravimetric analysis and stored at −80 °C until analysis.
Mass spectrometric detection was performed using a Xevo-2G quadrupole time-of-flight mass spectrometer (QTOF-MS; Waters, MA, USA). Make-up solvent (10 mM ammonium formate in methanol) was supplied at 0.25 mL/min and back-pressure regulated using two T-pieces, placed between the chromatographic system and the mass spectrometer (split ratio about 1:100) 34,35 . The capillary voltage was set at 3.0 kV and 2.5 kV for positive and negative electrospray ionization mode, respectively. The sampling cone voltage was set at 40 V, the cone gas flow rate at 100 L/h and the drying gas flow rate at 800 L/h, with a source and drying temperature of 120 °C and 200 °C respectively. The mass spectrometer was operated in MS E mode with a scanning range of m/z 150−1000, with a resolution of at least 20000 at m/z 500-900 using leucine-enkephalin (1500 ng/mL at 5 µL/min) for internal calibration. Data were processed in MassLynx v4.1(Waters, MA, USA) and Mzmine2 36 .
Lipids were identified by exact mass using LipidMaps ® Lipidomics gateway (San Diego, CA) and published data 34,37 retention times and by fragments.
Absolute quantification was performed using an external calibration curve. Calibrant mixtures were composed by tripalmitin, DPPC, sphingomyelin and stearic acid in Limit of detection (LOD) and limit of quantification (LOQ) were defined as three and ten times the signal-to-noise-ratio (S/N), respectively, for the analyte/internal standard-ratio and were calculated separately per lipid species (n = 3 per lipid species). Lipid species with levels below their respective LOD were considered absent and those with levels below LOQ were not quantified and hence not used for calculation of total amounts of lipid within the assessed lipid classes. A weighted calibration curve (1/Y) was used for TG and FAs, due to significant heteroscedasticity of the data 38 .
All analyses were conducted in a single randomized batch with alternation between positive and negative electrospray ionization. Calibration standards (n = 3) were analyzed prior to the first sample in the sequence and 8 blanks were evenly distributed in the batch. investigation of extracellular matrix integrity. The evaluation of the ECM integrity after treatment was approached by (1) the degree of tissue dehydration expressed as water retention (%; mg of retained water per 100 mg of water content in fresh tissue), (2) staining of elastin and collagen fibers of treated tissue, using the Modified Verhoeff Van Gieson Elastic Staining 39 , (3) transmission electron microscopy and (4) length-tension studies.
Transmission electron microscopy (TEM). Biopsies (2 mm sections) were taken out from four different paraffin blocks containing samples of native, detergent treated, pressurized CO 2 -EtOH-H 2 O treated or pressurized CO 2 -limonene treated. The samples had previously been sectioned for H&E staining and all had been treated with benzonase nuclease, except for the native control. The biopsy was dewaxed in xylene, washed in ethanol, stained with 0.05% methylene blue in ethanol, rinsed in ethanol, acetone, followed by 1:1 mixture of Polybed-acetone and finally embedded in pure Polybed 812. The polymerised block was sectioned with a Leica UC7 ultramicrotome (Leica Microsystems GmbH, Wetzlar, Germany) and sections were mounted on a pioloform coated copper Maxtaform H5 grid. The section was contrasted with 4% Uranyl acetate followed with 1% lead citrate. Images of the samples were analysed in a Tecnai BioTWIN transmission electron microscope (FEI Company, OR, USA) at two different magnifications.
Wire myography experiments for length-tension studies. Length-tension studies were carried out using wire myography. The samples studied were 2 mm long segments of porcine pulmonary artery either non-treated (native) or previously submitted to pressurized CO 2 -limonene, pressurized CO 2 -EtOH-H 2 O and detergent. The pressurized treatment corresponded to the central points in the experimental design. The samples were tied to pins in three Myograph Systems (610 M and 620 M from Danish Myotechnology a/s, Aarhus, Denmark) 40 using silk thread (6-0). The temperature inside the myograph chambers was maintained at 37 °C. Zero basal tension (2020) 10:4031 | https://doi.org/10.1038/s41598-020-60827-4 www.nature.com/scientificreports www.nature.com/scientificreports/ was ascertained in Hank's Balanced Salt Solution (HBSS, Sigma-Aldrich). Following equilibration, preparations were stretched in pre-specified steps and force was measured after 3 min. This procedure was repeated 10 times to obtain length-tension curves for each individual sample. Length and dry weight of each sample was used to obtain an approximation of the cross-sectional area that was used to normalize force.
Statistical and chemometric analysis. The full factorial designs were evaluated using multi-linear regression in MODDE 10.1 (Sartorius Stedim Biotech). Principal component analysis (PCA) was performed in SIMCA-P 12.0.1 (Sartorius Stedim Biotech). Differences between groups were evaluated by analysis of variance (ANOVA), with Tukeys test post hoc, and precision was estimated using an F-test. Significance was defined as p < 0.05. In this work, the applicability of scCO 2 and pressurized CO 2 -cosolvent fluids to decellularize porcine pulmonary arteries was studied.

Results and Discussion
Bio-scaffolds have been produced from animals and used for transplantations in humans 11 . A prerequisite for such an approach is that all genetic material and lipids can be removed, while maintaining the integrity of the ECM fibers 17,21,30 . In this work, the applicability of scCO 2 and pressurized CO 2 -organic solvent mixtures for decellularization of porcine pulmonary arteries was studied. Delipidation of pulmonary arteries by pressurized CO 2 . Next, we examined the effects of temperature, time and pressure on tissue delipidation, using a design of experiments (DoE) approach. However, we could not detect any significant effects of these parameters on any of the response variables with either of the solvent combinations. Hence, as TG, PC, SM, FA, and total extractable lipid residues did not differ between conditions, we pooled data for the different solvent combinations to gain power in our further analyses.
Subsequently, we compared total lipid residues in treated and untreated tissue, as determined by gravimetric analysis. Pressurized CO 2 -EtOH-H 2 O treatment efficiently reduced tissue lipid levels (p < 0.01) (see Supplementary Fig. S3). Unexpectedly, samples submitted to both scCO 2 and pressurized CO 2 -limonene showed higher levels of residual lipids than untreated samples (p < 0.01) (Supplementary Fig. S3). Since no limonene was detected during mass spectrometric analysis (the intensity of limonene adducts were below respective spectrometry base lines), these results are not due to residual limonene being extracted in the dichloromethane fraction. An interpretation is that the treatment with scCO 2 and pressurized CO 2 -limonene did not lead to a significant removal of lipids but it improved accessibility of lipids in subsequent extraction by dichloromethane.
To generate a more comprehensive depiction of the delipidation process, we studied in detail the lipid profiles generated by mass spectrometric analysis. TGs were partially reduced for all treatments and were found at detectable levels in all samples. All PCs and SMs were also found at detectable levels in all samples, except for PC 34:0 and PC 32:1 which were undetectable after pressurized CO 2 -EtOH-H 2 O treatment, and SM 41:1 which was absent in most samples treated with this solvent combination (Fig. 3a). Overall, pressurized CO 2 -EtOH-H 2 O treatment was found to be the most efficient solvent combination for delipidation, resulting in a majority of TG, PC and SM lipid species to be reduced to levels below their respective LOQs. Therefore, concentrations of no quantifiable lipids species were ranged from 0.027, 0.919 and 1. Pressurized CO 2 -limonene, on the other hand, mainly reduced levels of TG 48:2 (Fig. 3a). Pressurized CO 2 -EtOH-H 2 O was also found to be the most efficient solvent combination to reduce levels of FAs (Fig. 3b). FAs 18:1, 22:3, 22:5 and 22:6 were all absent. In addition, all long-and very long-chained FAs (C > 18), regardless their degree of unsaturation, showed levels below their respective LOQs (Fig. 3b, Supplementary Table S1).
Hence, the choice of cosolvent exerted a much more dramatic influence on PA delipidation, as compared to pressure, temperature and extraction time (Fig. 3). The extensive delipidation observed for pressurized CO 2 -EtOH-H 2 O, compared to scCO 2 and pressurized CO 2 -limonene, are in line with results obtained for porcine retina 32 . The relative permittivity of CO 2 is low, which makes it appropriate to dissolve mostly non-polar compounds of low molecular weight 41 . The static relative permittivity, and therefore the polarizability, of scCO 2 can be increased with the addition of cosolvents. As an example, supercritical CO 2 -cosolvent fluids have been effective in the extraction of edible lipids like triacylglycerols and fatty acids 42 . However, there is no evidence that the addition of such small amounts of cosolvent is enough to dissolve more polar lipids, like the ones known to be present in pulmonary artery 34 . As an alternative, higher amounts of cosolvents than what is soluble in supercritical CO 2 can be used. Under controlled conditions of pressure and temperature, this leads to a one-phase pressurized CO 2 -cosolvent fluid (see Supplementary Fig. S4), however not in the supercritical regime. Such fluids are rarely studied 31 , but they offer an even higher range of polarizability than supercritical mixtures, expanding the type of extractable lipids 32 . Mass transfer properties of such pressurized fluids are also modified with respect to the neat cosolvent, in favor of better extraction power. Furthermore, a one-phase pressurized CO 2 -cosolvent fluid prevents direct contact between the tissue and the liquid organic solvent, which would otherwise cause cytotoxic effects. By selecting ethanol as cosolvent in the pressurized fluid, we are increasing the amount of hydrogen-bonding www.nature.com/scientificreports www.nature.com/scientificreports/ interactions that can be created between solvent and solute. This results in higher solubilization of the polar lipids present in pulmonary arteries. In contrast to ethanol, limonene is not a polar molecule, but it is generally used to dissolve lipids in industrial applications. In the case of limonene as cosolvent, dispersion forces become stronger than for neat CO 2 , but these intermolecular interactions are still not strong enough to achieve full lipid removal of the less polar lipids (i.e. TGs) nor to dissolve the most polar lipid classes.
A PCA was calculated to visualize the impact of treatments on lipid profiles (described variation, R 2 = 0.83; predictive ability Q 2 = 0.76) (Fig. 4). The score scatter plot (Fig. 4a) revealed a clear separation between the untreated and treated PA along principal component (PC) 2, and a separation of CO 2 -EtOH-H 2 O treated PA from PA treated with CO 2 -limonene or neat scCO 2 along PC1. The loading plot (Fig. 4b), revealed that differences in TG levels did not affect clustering as much as the other species, possibly because all treatments were partially effective at removing TGs from the tissue. The biggest impact on the clustering was due to the more polar lipids, with the biggest difference most clearly observed for PCs and SMs. CO 2 -EtOH-H 2 O was more efficient than the other conditions in extracting polar lipids from the tissue, which corresponds well with its differentiated position along PC1 in Fig. 4a. These results are in agreement with the discussion above, based on the polarizability of the solvent mixtures and the type of intermolecular interactions present.
In line with the results from the PCA, univariate analysis of the lipid data revealed pressurized CO 2 -EtOH-H 2 O to be the most efficient of the tested methods for PA delipidation (Fig. 5a-c). Levels of TGs were reduced by all delipidation methods (p < 0.01), with CO 2 -EtOH-H 2 O treatment decreasing TG by 93.2% (p < 0.01) (Fig. 5a). This reduction was more significant than the reduction observed with CO 2 -limonene (81.0%). The correspondence between TG removal rates observed for scCO 2 and pressurized CO 2 -limonene is likely governed by the similar low polarity of these fluids. Levels of PCs and SMs differed more dramatically between treatment conditions and were mainly reduced after pressurized CO 2 -EtOH-H 2 O treatment (97.8% and 94.6%, respectively; p < 0.01). Unexpectedly, the total FA content was found to increase after treatment with neat scCO 2 (p < 0.01) and pressurized CO 2 -limonene (p < 0.05) (Fig. 5d), in line with the unexpected results from the gravimetric analysis. Similar trends have been observed for other tissue treated with pressurized CO 2 -based fluids 32 . The reason for this observation may be a result of the pressurized CO 2 (at both supercritical or pressurized state) impacting on FA availability and facilitating subsequent extraction with dichloromethane. This effect is also observed for saturated FAs in samples treated with pressurized CO 2 -EtOH-H 2 O. The fact that we can appreciate this effect is due to that saturated FAs were not removed by the pressurized treatment to the same extent as other FAs (Fig. 6).
Pressurized CO 2 -EtOH-H 2 O was found to more efficiently remove fatty acids, as compared to neat scCO 2 and pressurized CO 2 -limonene, presumably due to the possibility to form hydrogen bondings with the carboxylic acid moieties and OH … π−bonds with carbon-carbon double bonds 43 . The type of FA that remains in the tissue may affect the success of a subsequent recellularization. Notably, monounsaturated fatty acids (MUFAs) and polyunsaturated fatty acids (PUFAs), the most polar FA subclasses, were more efficiently removed by pressurized CO 2 -EtOH-H 2 O, as compared with the other treatments (Fig. 6). Levels of saturated FAs were more similar between conditions. Details on individual lipid species can be found in Supplementary Fig. S5. These are very promising results, considering that unsaturated FAs have been reported to impair proper cell adhesion, whereas the saturated FA stearic acid (FA 18:0) has been shown to induce cell adhesion 14 .
Notably, total levels of the pro-inflammatory fatty acid arachidonic acid (FA 20:4) 15 and arachidonic acid containing lipids such as TG 54:4, TG 54:5, PC 38:4 and PC 38:5 44,45 , were reduced by 48% after treatment of the tissue with pressurized CO 2 -EtOH-H 2 O, as compared to untreated samples. ScCO 2 treatment resulted in a less effective removal of pro-inflammatory lipids reduction (13.9%) followed by pressurized CO 2 -limonene (30.9%). Whether the reduction in arachidonic acid (FA 20:4) observed after pressurized CO 2 -EtOH-H 2 O treatment can lower proinflammatory responses after transplantation remain to be examined. www.nature.com/scientificreports www.nature.com/scientificreports/ Removal of DNA from pulmonary arteries by pressurized CO 2 . Following delipidation studies, we set out to examine whether any of the treatments also removed nuclei from the tissue, as has previously been suggested 30 , and questioned 17 for porcine aorta. We did not find any evidence of DNA removal, assessed by both dsDNA quantification and H&E staining, using any of the pressurized CO 2 conditions tested. In the case of scCO 2 , these results were expected due to the high polarity and molecular weight of DNA. In the case of CO 2 -cosolvent mixtures, these results are in agreement with the work by Casali et al. 17 but contradict that of Sawada et al. 30 . The former used a mixture of CO 2 -cosolvents in the supercritical state, which is not comparable to one-phase pressurized CO 2 -cosolvent fluids. The latter is more comparable to the fluid conditions used in this work, although their mixture contains higher amounts of ethanol. It is not clear at this point if this small difference in composition explains the discrepancies in results.
Considering However, after nuclease treatment, one third of the pressurized CO 2 -limonene treated tissues showed a notable nuclei removal in H&E staining (see example in Fig. 7g), suggesting the absence of dsDNA. A reduction up to 93.4% of the dsDNA content compared with fresh tissues (for samples submitted to 15 MPa and 40 °C for 90 min) was achieved, resulting in an improvement of a 12.2% compared to the results from detergent-enzyme treatment. The effect of treating the sample with pressurized CO 2 -EtOH-H 2 O and neat scCO 2 fluids prior to nuclei removal by detergent-enzyme treatment was moderate and null, respectively. There is no theory that can currently explain the promising effect of pressurized CO 2 -limonene fluid as facilitator of denaturalization of DNA by enzymes.

Assessment of extracellular matrix integrity.
Tissue dehydration is an important parameter determining the suitability of the tissue as a bio-scaffold, and a low water content has been suggested to impair the mechanical properties of the tissue 48 . However, the minimum hydration grade to preserve the functionality of a certain tissue remains unknown. All explored fluid combinations exerted a dehydration effect. Higher water retention was observed for pressurized CO 2 -limonene and pressurized CO 2 -EtOH-H 2 O treatments, i.e. 16.8% and 15.0% respectively, compared to neat scCO 2 treatment (11.4%). These results suggest that the presence of limonene and EtOH-H 2 O prevent the tissue from suffering extreme dehydration, which is in line with a previous study 17 .
We also conducted a histological examination of overall tissue structure and the integrity of elastic fibers, using a limited set of samples as indicated in Fig. 4a. This examination did reveal an overall conserved morphology with Data are presented as mean ± standard deviation for n = 5 (untreated and neat scCO 2 ) and n = 42 (CO 2 with cosolvent). Data were compared using ANOVA with Tukey's test post hoc. *p < 0.05 and **p < 0.01. Figure 6. Presence of fatty acids (%) in tissues after treatment. ARA, arachidonic acid. Untreated samples expressed as 100%. * And ** denote significant differences, p < 0.05 and p < 0.01 respectively, compared with their respective fatty acids group of the untreated samples. # And ## denote significant differences, p < 0.05 and p < 0.01 respectively, compared with their respective fatty acids group of the pressurized CO 2 -EtOH-H 2 O samples.
parallel aligned continuous elastic fibers without any differences between treated and untreated tissue. The spacing between elastic fibers tended to be decreased after treatment, consistent with extraction of lipids and other cellular material. The change in spacing between fibers was more pronounced for tissue samples treated with detergents ( Supplementary Fig. S6).
Tissue ultrastructure was evaluated in the same samples used in the histological evaluation by reprocessing paraffin embedded samples for TEM. Tissues were stained and imaged at up to 60000 times magnification. None of the pressurized treatments exerted a distinct effect on the ECM (Fig. 7i-p). Fibrils exhibited thickness and organization comparable to the native tissue regardless of treatment with pressurized CO 2 -EtOH-H 2 O, CO 2 -limonene or detergents. Collagen fiber organization appeared intact as far as can be judged by the density and parallel appearance of the collagen fibrils. These observations were supported by length-tension studies. Our results did not show any clear differences in length-tension relationships, and thus presumably stiffness, due to sample treatments or compared to the native samples ( Supplementary Fig. S7).
Decellularization and tissue integrity remains key for the production of bio-scaffolds. However, detergent and ezyme-based decellularization does not completely remove all immunogenic reactions. Previous studies have shown that decellularized xeno-transplanted tissue may still provoke immunogenic responses such as antibody formation in patients that have received decellularized porcine valves 49 . The immunogenic response appears to be induced by the ECM protein collagen VI, albumin and αGal epitopes 49,50 . These findings indicate that the ECM itself possess immunogenicity after decellularization by detergents and enzymes. Overall, the results in this work revealed that pressurized CO 2 fluids make an impact in the removal of lipids from pulmonary artery, and in the and not subjected to pressurized CO 2 mixtures, (e) artery subjected to neat scCO 2 (30.0 MPa, 40 °C, 120 min) followed by endonuclease treatment, (f-h) same treatments as in b, c, and d with the addition of subsequent endonuclease treatment. (j-l) and (n-p) correspond to TEM images at low and high magnification, respectively, of the respective samples above, i.e. f, g and h. Collagen fibers in cross section (black arrowheads), longitudinally sectioned collagen fibers (white arrowheads) and cell nucleus (asterisk). (2020) 10:4031 | https://doi.org/10.1038/s41598-020-60827-4 www.nature.com/scientificreports www.nature.com/scientificreports/ removal of DNA when added to the battery of decellularization strategies available. Further studies are needed to evaluate if treatment with pressurized CO 2 fluids has also a positive effect on reducing the immunogenic properties of ECM.
conclusion Pressurized CO 2 -EtOH-H 2 O fluid was found to be the most efficient pressurized solvent combination to achieve lipid removal from pulmonary artery tissue. It led to the efficient removal of TGs, PCs, SMs and most FAs including pro-inflammatory lipids, and a less efficient removal of saturated FAs. Pressurized CO 2 -limonene showed a low delipidation efficiency, similar to neat scCO 2 . Further studies are needed to evaluate whether the close to complete delipidation achieved in this work results in low immune responses, cellular invasion and functional recellularization. DNA removal was more efficient with pressurized CO 2 -limonene after endonuclease treatment. Importantly, this, as well as the other tested treatments, preserved ECM integrity.