Organ transplantation as a therapy suffers from a critical lack of available donor organs. Anatomic and physiologic similarities between pigs and humans could permit the use of these animals as organs donors to eliminate this shortage.1,2 A key barrier to cross-species transplantation (xenotransplantation) is the destruction of donor tissue occurring when preformed human antibodies activate the complement system after binding to antigens expressed on pig tissue.3,4 Though multiple carbohydrate xenoantigens have been identified,5-10 early efforts at overcoming antibody-mediated rejection were limited by inefficient genetic engineering tools. Initial approaches eliminated the most abundant xenoantigen through inactivation of the pig glycoprotein α-galactosyl transferase 1 (GGTA1) gene, but this did not eliminate antibody-mediated rejection even when the GGTA1-deficient animals also expressed complement inhibitory receptor transgenes.11-13
To further reduce injury of pig tissue induced by human antibodies, we have used gene-editing systems to disrupt the swine GGTA1, cytidine monophospho-N-acetylneuraminic acid hydroxylase (CMAH) and β-1,4-N-acetylgalactosaminyl Transferase 2 gene (β4GalNT2) genes and eliminate the expression of the carbohydrate-based xenoantigens, α-gal, Neu5Gc, and DBA-reactive glycans, respectively.14 Comparing human antibody binding to peripheral blood mononuclear cells of wild type swine, and animals lacking either GGTA1, GGTA1/CMAH, or GGTA1/CMAH/β4GalNT2 showed that each successive gene deletion further reduced human antibody binding.
In this study, we have compared human antibody binding to RBC from various strains of pig using allogeneic and autologous human RBC as the standards of minimal interaction. We used RBC as a model system of low antigenicity because they are easy to study as a homogenous population of a single cell type, they can be transferred between mammals of the same species without needing immune suppression, and they do not express MHC molecules which frequently elicit humoral immunity in the setting of transplantation. Quantitative mass spectrometry suggested that triple knockout pig RBC and human RBC express similar levels of antigens. Flow cytometry confirmed that red blood cells (RBC) from a GGTA1/CMAH/β4GalNT2-deficient pig and human RBC bind similar levels of IgG and IgM. Reducing the xenoantigenicity of swine RBC, to levels approximating the antigenicity of human blood group O and auto RBC, suggests that it is possible to remove the humoral barrier to xenotransplantation through a gene knockout approach.
MATERIALS AND METHODS
Human and Pig Sample Preparation
Human serum and RBC samples were collected under an IRB approved protocol, or sera were purchased from an Food and Drug Administration-registered center using protocols approved by the American Association of Blood Banks (Valley Biomedical, Winchester, VA). Swine samples were collected under an IACUC-approved protocol. Erythrocytes were isolated from whole blood collected in acid-citrate-dextrose tubes (Becton Dickinson and Company, Franklin Lakes, NJ) using Ficoll-Paque Plus (GE Health, Uppsala, Sweden). Fresh serum was isolated by collecting whole blood in the absence of anticoagulant and centrifuging to remove clotted material. The production of the genetically modified pigs used in this study has been described in detail.14,15
RBC Stripping and Associated Flow Cytometry
After density isolation, RBCs were washed 3 times in PBS and suspended in 50% Alsever solution. Red blood cells were pelleted at 21 300g for 2 minutes Serum was added to the pelleted cells at a 1:1 of cell pellet volume to serum volume. Serum RBC mixture was mixed and incubated for 20 minutes at 4°C. Cells were pelleted at 21 300g for 2 minutes and washed once with Alsever solution. Cells were mixed with acid stripping buffer (pH 2.75 citric acid/phosphate at 300 mOs/kg) to remove bound antibodies and neutralized with 1 M Tris-base, pH 9.0. (Calbiochem, La Jolla, CA). Material that eluted during the low pH treatment was analyzed by sodium dodecyl sulfate-polyacrylamide gel electrophoresis (SDS-PAGE) and mass spectrometry (see below). Matching flow cytometric analysis for antibody elution samples was completed using 2 × 106 RBCs. Cells were suspended in EX-CELL 610-HSF Serum-Free Medium (Sigma, St. Louis, MO) with 0.1% sodium azide incubated for 30 minutes at 4°C with a mixture of 25% heat-inactivated serum. Red blood cells were washed 3 times and stained with goat antihuman IgG Alexa Fluor 488 or donkey antihuman IgM Alexa Fluor 488 (Jackson ImmunoResearch Laboratories Inc., West Grove, PA) antibodies. Secondary antibodies were incubated with RBCs for 30 minutes at 4 °C and washed. Flow cytometric analyses were completed on BD Accuri C6 flow cytometer (Accuri, Ann Arbor, MI, USA) and histograms were generated using FlowJo 7.6.5 (FlowJo LLC, Ashland, OR). Red blood cell gating was based on forward and side scatter (Figure 2 and Table 3). A representative gating is shown in Figure S4 (SDC, http://links.lww.com/TP/B301).
SDS-PAGE Gel Analysis of Proteins Eluting from RBC
After the stripping described above, neutralized eluates were precipitated by diluting 100-μL elution into 900-μL acetone, and then incubated in −20°C for 30 minutes. Precipitates were spun 20 000g at 4°C for 15 minutes. Supernatant was discarded and pellets were washed with 70% v/v ethanol in water. Precipitates were spun again at 20 000g at 4°C for 15 minutes. Supernatant was again discarded and pellets were dried in a Vacufuge Plus (Eppendorf, Hauppauge, NY) at 37°C for 30 minutes. Samples were then dissolved in 100 μL of 2× Laemmeli Buffer (Bio-Rad, Hercules, CA) containing 2-mercaptoethanol (Sigma-Aldrich) per manufacturer's instructions. Samples were then heated at 95°C for 5 minutes and then allowed to cool to room temperature. Fifteen microliters of each sample was then loaded onto a 26-well, 4% to 20% Stain-Free TGX gel (Bio-Rad) and electrophoresed under denaturing and reducing conditions at 200 V for 42 minutes using the Bio-Rad Criterion system. After electrophoresis, gel was removed and stained for 30 minutes with 100 mL G-250 Bio-Safe Coomassie Stain (Bio-Rad). Gel was then destained in water and imaged using a 700-nm laser on a Li-Cor Classic Imager (Li-Cor Biosciences, Lincoln, NE). Aliquots of proteins eluted from RBC were then shipped on dry ice for mass spectrometry preparation and analysis.
Mass Spectrometry Quantitation of Immunoglobulin Eluted from RBC (Performed by MSBioworks, LLC)
A 50-μL aliquot of each sample was treated with PNGase F (New England Biolabs) according to manufacturer's instructions. Each sample was acetone precipitated for 30 min followed by a wash with 70% ethanol at 4°C. Resultant precipitates were dried and resuspended in 40 μL 1.5× LDS loading buffer with DTT. Twenty microliters was run on a 4% to 12% bis tris SDS PAGE gel in the MOPS buffer system.
The region containing the heavy chain (50 kDa) was excised, and trypsin digestion was performed using a robot (ProGest, DigiLab) with the following protocol: (1) washed with 25 mM ammonium bicarbonate followed by acetonitrile, (2) reduced with 10 mM dithiothreitol at 60°C followed by alkylation with 50 mM iodoacetamide at room temperature, (3) digested with trypsin (Promega) at 37°C for 4 hours, (4) quenched with formic acid and the supernatant was analyzed directly without further processing. The gel digests were analyzed by nano LC/MS/MS with a Waters NanoAcquity HPLC system interfaced to a ThermoFisher Q Exactive. Peptides were loaded on a trapping column and eluted over a 75-μm analytical column at 350 nL/min; both columns were packed with Jupiter Proteo resin (Phenomenex). The mass spectrometer was operated in data-dependent mode, with MS and MS/MS performed in the Orbitrap at 70 000 FWHM resolution and 17 500 FWHM resolution, respectively. The fifteen most abundant ions were selected for MS/MS. The following peptides were used to determine relative quantities of each isotype: IgG1, EEQYDSTYR and TKPREEQYDSTYR; IgG2, EEQFDSTFR and TKPREEQFDSTFR; IgG3, EEQYDSTFR and TKPREEQYDSTFR; IgG4, EEQFDSTYR and TKPREEQFDSTYR; IgM, YKDNSDISSTR. The underlined aspartate residues (D) represent deamidated asparagine residues occurring as a consequence of PNGase F treatment.
Data were searched using a local copy of Mascot with the following parameters: Enzyme: Trypsin Database: Swissprot Human (forward and reverse appended with common contaminants) Fixed modification: Carbamidomethyl (C) Variable modifications: Oxidation (M), Acetyl (Protein N-term), Deamidation (NQ), Pyro-Glu (N-term Q) Mass values: Monoisotopic Peptide Mass Tolerance: 10 ppm Fragment Mass Tolerance: 0.02 Da Max Missed Cleavages: 2 Mascot DAT files were parsed into the Scaffold software for validation, filtering and to create a nonredundant list per sample. Data were filtered 1% protein and peptide level false discovery rate (FDR) and requiring at least 2 unique peptides per protein. Raw LC/MS data were inspected for the accurate m/z values of the target peptides the selected ion chromatograms were extracted in QualBrowser (Thermo); the peak area in each case was calculated.
After density isolation, RBC were washed 3 times with PBS and diluted 1:10 in PBS at room temperature (Figure 3). To determine the human antibody binding to RBC (2x105 cells/well) were incubated with diluted, heat-inactivated human serum for 30 min at 4 °C, final serum concentration 25%. The cells were washed 3 times in PBS with azide and stained with goat antihuman IgG Alexa Fluor 488 or donkey antihuman IgM Alexa Fluor 488 (Jackson ImmunoResearch Laboratories Inc., West Grove, PA, USA). Flow cytometry analysis was performed using an Accuri C6 flow cytometer and CFlow software (Accuri, Ann Arbor, MI, USA). RBC gating was based on forward and side scatter.
Cell surface analysis of porcine RBC cell surface glycans was performed using antibodies/lectin isolectin Griffonia simplicifolia GS-IB4 Alexa Fluor 647 (Invitrogen, Grand Island, NY) for αGal, a chicken anti-Neu5Gc antibody kit (BioLegend, San Diego, CA) for Neu5Gc, and Fluorescein Dolichos Biflorus Agglutinin (DBA) (Vector Laboratories, Inc, Burlingame, CA) for β4GalNT2-derived carbohydrates.
Figures refer to the various RBC as: autologous human (A), blood group O allogeneic human (H), wild type pig (W), GGTA1/CMAH knockout pig (D), and GGTA1/CMAH/β4GalNT2 knockout pig (T). Pigs were created using gRNA/Cas9 gene-editing tools, somatic cell nuclear transfer, and embryo implantation. This process has been described in detail previously.14,15 RBC were evaluated for expression of α-gal, Neu5Gc, and DBA-reactive glycans to verify the loss of antigenic structures with each successive gene disruption (Figure S1, SDC, http://links.lww.com/TP/B301). All pigs were generated using cells from the same founder animal in an effort to limit genetic variation between strains to the specifically targeted loci. The founder animal expressed the O blood type.
Quantitative Analysis of Human Antibody Binding to Various RBC Using Mass Spectrometry
Immunoglobulin binding was analyzed using a recently described quantitative mass spectrometry approach that measures the abundance of individual antibody isotypes.16 The Materials and Methods section and Figure 1A describe the protocol used in this approach. A single human serum was incubated with various RBC (pigs: W, D, T; and autologous human: A). Cells were washed once and incubated in low pH buffer to elute bound proteins. Eluted proteins were separated on SDS-PAGE and visualized by staining the gel with Coomassie Blue (Figure 1B). The presence of hemoglobin in the supernatant of serum-free samples indicated that the manipulations caused RBC lysis. Acid washing accelerated lysis and released increasing amounts of hemoglobin from the cells (triple arrowheads; compare lanes 1 and 2 in all RBC types). Other unidentified RBC polypeptides (*) were released at low pH even in the absence of serum and were not evaluated further. A protein migrating identically to human albumin was present in the acid wash supernatant of serum-incubated cells (single arrowhead; lanes 3 and 4 in all RBC types). Though autologous human RBC released less of this protein than the swine cells, this result was not reproduced in other experiments. After serum incubation, cells also released a faint band during the low pH wash that migrated identically to immunoglobulin G heavy chains (double arrowhead, lanes 3 and 4). The intensity of this band varied between the different RBC. Incubating cells at low pH for either 2 or 3 minutes released similar protein levels (compare lanes 3 and 4 in each sample). To precisely determine the quantity of antibody released from the cells, gel slices encompassing the approximate size of immunoglobulin heavy chain in the 2-minute elution samples, were collected and incubated with trypsin. Digested material eluting from these slices was fractionated by HPLC and peptides specific to each antibody isotype were identified using mass spectrometry. The area under the curve for each analyzed peptide enabled quantitative comparison of relative isotype abundance. Summing the area under the curve values for the peptides representing each isotype were used to calculate total IgG binding.
Table 1 provides quantitative mass spectrometry results on serum samples randomly collected from 4 unique individuals. Normalizing relative IgG and IgM binding values to the levels obtained on the triple knockout pig RBC ranked swine cells' antigenicity as W > D > T. Notably, mass spectrometry analyses also indicated that less human antibody bound to the triple knockout swine RBC than to autologous human RBC, where serum and cells were isolated from the same person. Though IgG1 and IgG2 accounted for the largest amount of IgG binding (percent of total, Table 1), large fluctuations in the binding occurred in each isotype (relative isotype levels, Table 1). Technical replicates of the mass spectrometry assay were performed to analyze the variability of the technique (Figure S2, SDC, http://links.lww.com/TP/B301). Per analysis, total relative IgG and IgM binding showed greater variability of relative isotype binding, indicating that the acid elution contributed more to the variation than did the mass spectrometric evaluation.
Comparing Mass Spectrometry and Flow Cytometry Techniques for Quantitating Antibody Binding to RBC
Though the mass spectrometry results suggested varying antigenicity across RBC isolated from different swine, the variability in the assay required confirmation of these results by a second method. Consequently, 3 of the sera used in mass spectrometry were also evaluated using flow cytometry (Figure 2). Each serum was incubated with its autologous RBC (A) and with pig RBC (W and T). Figures 2A and B show human IgG and IgM binding, respectively. Of all cell types tested, wild type pig RBC bound the most antibodies. The poor quality of histograms in the W samples likely resulted from cell destruction that occurred as a consequence of wild type RBC high antigenicity (Figure S4, SDC, http://links.lww.com/TP/B301). In agreement with the mass spectrometry results, histograms representing IgG and IgM binding to either autologous human RBC (A) or pig (T) RBC overlapped the secondary-only negative control suggesting minimal binding to each cell type. Because antibody binding to human autologous RBC was not detected, serum from a person with blood type O was incubated with blood type A RBC to verify that the flow cytometry assay could reveal increased antigenicity of human RBC (Figure 2C).
Flow Cytometric Analysis of Human Antibody Binding to Human Allogeneic RBC and Pig RBC
To expand the evaluation of antibody binding, human sera from 83 individuals were incubated with pig RBC (W, D, and T). Allogenic human RBC, expressing blood group O, were also used for comparison (H). Flow cytometry was used to detect bound antibodies. The median fluorescent intensities for all sera are shown (IgG and IgM in Figures 3A and B, respectively). Human antibody binding to the various RBC was maximal on wild type cells. Eliminating the GGTA1 and CMAH genes reduced binding of pig RBC to human IgG and IgM below the levels seen for antibody binding to allogeneic blood group O human RBC in 18 of 83 sera. Inactivating the β4GalNT2 gene in addition to GGTA1 and CMAH created pig RBC that bound fewer human antibodies than allogeneic blood group O human RBC in 25 of 83 tested sera (Figure 3C).
The pig has been the most studied animal when examining the use of animal tissue to eliminate the critical shortage of human organs. Several biological incompatibilities exist between pigs and humans that prevent clinical application of xenotransplantation. Major issues include destruction of pig tissue by the complement system, cellular rejection, and coagulation dysregulation.17 Though these incompatibilities are multifactorial, human immunoglobulins likely play a role in each. Antibody binding damages tissue by directly activating the complement cascade and by enabling antibody directed cellular cytotoxicity mediated by natural killer cells.3,4,18,19 Antibodies may also impair organ function through dysregulation of clotting factors and creation of a procoagulant state by affecting endothelial cells, which line the vasculature.20
Antibody-mediated damage of pig tissues in nonhuman primate models of xenotransplantation has been prevented by using organs that lack a single major xenoantigen (αGal) and express complement regulatory transgene(s).11-13 This approach has been key to protecting organs from preexisting xenoreactive antibodies that recognize antigens other than αGal. Continued pursuit of xenoantigen reduction in pigs is important because antibodies can injure transplanted tissue without activating complement.
We have been using recently improved gene editing tools to eliminate IgM and IgG targeting of glycan-based pig xenoantigens. Initially, we compared the binding of human antibodies to chimpanzee and pig cells.21,22 Chimpanzees provided a benchmark of low xenoantigenicity because their kidneys did not suffer immediate humoral rejection when implanted into humans.23 Disrupting the GGTA1 and CMAH genes eliminated αGal and Neu5Gc carbohydrate modifications from swine, making their cells less antigenic to humans than chimpanzee cells. Pigs lacking GGTA1 and CMAH gene function may express sufficiently few antigens to allow successful transplantation of their organs into some people.
To further lower antibody reactivity across a broader human population, we eliminated a third glycan modifying gene, β4GalNT2. Examining the human immunoglobulin binding among several porcine strains showed that peripheral blood mononuclear cells from swine lacking GGTA1, CMAH, and β4GalNT2 genes expressed the fewest xenoantigens.14 Given that the CMAH/GGTA1 knockout pig had already surpassed the nonhuman primate benchmark, here we compared human antibody binding to human cells versus cells isolated from a GGTA1/CMAH/β4GalNT2-deficient pig. We chose RBC as target cells because they could easily be isolated as a homogenous population. In addition, they do not express MHC-derived proteins which, though potentially important antigens, could confound the ability to analyze the impact of glycan modification on xenoantigen reduction. Using blood type O allogeneic and autologous human RBC as comparators should provide the most stringent criteria of low antigenicity. The analyses reported here indicated that RBC from the triple knockout swine bound as few or fewer human antibodies than did autologous human RBC (Table 1, Figures 2 and 3). These results demonstrate that gene knockouts dramatically reduce, and in some cases, eliminate the xenoantigenicity of a single pig cell type.
Although RBC likely display a less complex antigenic profile than an intact organ, their use was critical in the validation of a novel quantitative mass spectrometry assay that would describe the level of antibody binding. It should be possible to adapt this procedure to study the interaction of human IgM and IgG isotypes with other cell types and intact swine organs in ex vivo perfusion systems and in explanted organs that had been transplanted. Understanding the quantitative interaction of IgM and specific IgG isotypes with an organ may yield insight into aspects of humoral immunity that cannot be addressed thoroughly with the current technologies, such as flow cytometry and immunofluorescent/immunohistochemical analyses of tissue biopsies. For example, in vitro studies have shown that the IgG3 isotype can activate endothelial cells in a complement dependent manner, whereas IgG1, IgG2, and IgG4 cannot.24,25 Being able to examine the levels of each isotype binding to intact pig organs may help refine the understanding of how individual classes of antibody contribute to the rejection of xenotransplants.
The ability to eliminate the reactivity of preformed human antibodies with pig tissues has been in question for some time. Our results presented here suggest that this is an attainable goal. As with all gene editing, it will be important to produce healthy animals to ensure that their organs will be suitable as replacement tissues. We have now been able to successfully generate 4 of these triple knockout animals. Though our analyses are ongoing, the animals appear to be healthy (age range, 116-344 days). It is likely that some pig-specific antigens exist which may not pose a problem until after exposure to transplanted pig tissues. Further work will be necessary to determine the best combinations of gene editing (transgene insertion or gene knockouts) and pharmaceutical interventions to ensure long-term survival of xenotransplants. Several groups have successfully demonstrated prolonged survival of pig tissues in nonhuman primate models of xenotransplantation using these approaches.26-28
1. Cooper DKC. The case for xenotransplantation. Clin Transplant
2. Reichart B, Guethoff S, Mayr T, et al. Discordant Cellular and Organ Xenotransplantation—From Bench to Bedside. In: Organ Transplantation in Times of Donor Shortage
. Switzerland: Springer International Publishing; 2016:209–220.
3. Platt JL, Fischel RJ, Matas AJ, et al. Immunopathology of hyperacute xenograft rejection in a swine-to-primate model. Transplantation
4. Dalmasso AP, Vercellotti GM, Fischel RJ, et al. Mechanism of complement activation in the hyperacute rejection of porcine organs transplanted into primate recipients. Am J Pathol
5. Galili U, Shohet SB, Kobrin E, et al. Man, apes, and Old World monkeys differ from other mammals in the expression of alpha-galactosyl epitopes on nucleated cells. J Biol Chem
6. Joziasse DH, Oriol R. Xenotransplantation: the importance of the Galα1, 3Gal epitope in hyperacute vascular rejection. Biochim Biophys Acta
7. Galili U, Macher BA, Buehler J, et al. Human natural anti-alpha-galactosyl IgG. II. The specific recognition of alpha (1—3)-linked galactose residues. J Exp Med
8. Nishimaki T, Kano K, Milgrom F. Hanganutziu-Deicher antigen and antibody in pathologic sera and tissues. J Immunol
9. Nowak JA, Jain NK, Stinson MW, et al. Interaction of bovine erythrocyte N-glycolylneuraminic acid-containing gangliosides and glycoproteins with a human Hanganutziu-Deicher serum. Mol Immunol
10. Byrne GW, Stalboerger PG, Du Z, et al. Identification of new carbohydrate and membrane protein antigens in cardiac xenotransplantation. Transplantation
11. McGregor CGA, Ricci D, Miyagi N, et al. Human CD55 expression blocks hyperacute rejection and restricts complement activation in Gal knockout cardiac xenografts. Transplantation
12. Schuurman HJ, Pino-Chavez G, Phillips MJ, et al. Incidence of hyperacute rejection in pig-to-primate transplantation using organs from hDAF-transgenic donors. Transplantation
13. Mohiuddin MM, Corcoran PC, Singh AK, et al. B‐cell depletion extends the survival of GTKO. hCD46Tg pig heart xenografts in baboons for up to 8 months. Am J Transplant
14. Estrada JL, Martens G, Li P, et al. Evaluation of human and non‐human primate antibody binding to pig cells lacking GGTA1/CMAH/β4GalNT2 genes. Xenotransplantation
15. Li P, Estrada JL, Burlak C, et al. Efficient generation of genetically distinct pigs in a single pregnancy using multiplexed single‐guide RNA and carbohydrate selection. Xenotransplantation
16. Ladwig PM, Barnidge DR, Snyder MR, et al. Quantification of serum IgG subclasses by use of subclass-specific tryptic peptides and liquid chromatography–tandem mass spectrometry. Clin Chem
17. Griesemer A, Yamada K, Sykes M. Xenotransplantation: immunological hurdles and progress toward tolerance. Immunol Rev
18. Seebach JD, Yamada K, McMorrow IM, et al. Xenogeneic human anti-pig cytotoxicity mediated by activated natural killer cells. Xenotransplantation
19. Schaapherder AF, Daha MR, te Bulte MT, et al. Antibody-dependent cell-mediated cytotoxicity against porcine endothelium induced by a majority of human sera. Transplantation
20. Lin CC, Ezzelarab M, Hara H, et al. Atorvastatin or transgenic expression of TFPI inhibits coagulation initiated by anti‐nonGal IgG binding to porcine aortic endothelial cells. J Thromb Haemost
21. Burlak C, Paris LL, Lutz AJ, et al. Reduced binding of human antibodies to cells from GGTA1/CMAH KO pigs. Am J Transplant
22. Wang ZY, Burlak C, Estrada JL, et al. Erythrocytes from GGTA1/CMAH knockout pigs: implications for xenotransfusion and testing in non‐human primates. Xenotransplantation
23. Reemtsma K, McCracken BH, Schlegel JU, et al. Renal heterotransplantation in man. Ann Surg
24. Sæthre M, Lea T, Borgen MS, et al. Human complement‐activating immunoglobulin (Ig) G3 antibodies are essential for porcine endothelial cell activation. Xenotransplantation
25. Sæthre M, Sølvik UØ, Haraldsen G, et al. Human serum‐induced porcine endothelial cell E‐selectin expression is associated with IgG3 and IgM anti‐Gal antibodies. Xenotransplantation t
26. Iwase H, Liu H, Wijkstrom M, et al. Pig kidney graft survival in a baboon for 136 days: longest life‐supporting organ graft survival to date. Xenotransplantation
. 2015 ;22(4):302–309.
27. Higginbotham L, Mathews D, Breeden CA, et al. Pre‐transplant antibody screening and anti‐CD154 costimulation blockade promote long‐term xenograft survival in a pig‐to‐primate kidney transplant model. Xenotransplantation
28. Mohiuddin MM, Reichart B, Byrne GW, et al. Current status of pig heart xenotransplantation. Int J Surg