Genetic Modifications Designed for Xenotransplantation Attenuate Sialoadhesin-dependent Binding of Human Erythrocytes to Porcine Macrophages

Abstract

The phenomenon of diminishing hematocrit after in vivo liver and lung xenotransplantation and during ex vivo liver xenoperfusion has largely been attributed to action by resident liver porcine macrophages which bind and destroy human erythrocytes. Porcine sialoadhesin (siglec-1) was implicated previously in this interaction. This study examines the effect of porcine genetic modifications, including knockout of the CMAH gene responsible for expression of Neu5Gc sialic acid, on the adhesion of human red blood cells (RBCs) to porcine macrophages. Wild type (WT) porcine macrophages and macrophages from several strains of genetically engineered pigs, including CMAH gene knockout and several human transgenes (TKO+hTg), were incubated with human RBCs and “rosettes” (≥3 erythrocytes bound to one macrophage) were quantified by microscopy. Our results show that TKO+hTg genetic modifications significantly reduced rosette formation. The monoclonal antibody 1F1, which blocks porcine sialoadhesin, significantly reduced rosette formation by WT and TKO+hTg macrophages compared to an isotype control antibody. Further, desialation of human RBCs with neuraminidase before addition to WT or TKO+hTg macrophages resulted in near complete abrogation of rosette formation, to a level not significantly different from porcine RBC rosette formation on porcine macrophages. These observations are consistent with rosette formation being mediated by binding of sialic acid on human RBCs to sialoadhesin on porcine macrophages. In conclusion, the data predict that TKO+hTg genetic modifications, coupled with targeting of porcine sialoadhesin by the 1F1 mAb, will attenuate erythrocyte sequestration and anemia during ex vivo xenoperfusion and following in vivo liver, lung, and potentially other organ xenotransplantation.

Introduction

Utilization of swine organs promises to address multiple critical and currently unmet clinical needs. For example, circulating a patient’s blood through a pig liver -- extracorporeal porcine liver perfusion (ECLP) -- could provide a patient in fulminant liver failure with supportive liver function while their own liver regenerates or until a human liver transplant becomes available^1–4. However, despite significant progress, major barriers still need to be overcome. Prior work in porcine liver xenoperfusion experiments by Rees et al. demonstrated that the pigs’ activated Kupffer cells (resident macrophages of the liver) bind and destroy more than 85% of perfused human erythrocytes in 72 hours⁵. Anemia out of proportion to surgical blood loss has also been observed after in vivo liver and lung xenotransplantation^6–9, and in association with heart and kidney xenografts^10,11. The observed erythrocyte sequestration, destruction, and resulting anemia pose a serious issue for liver, lung and other xenografts and remains one significant barrier to safe clinical translation of pig-to-human xenotransplantation or extracorporeal liver perfusion. Thus, understanding the mechanism by which erythrocyte damage or sequestration occurs may prove critical to making liver xenoperfusion and transplantation of organ xenografts safer.

Diminishing hematocrit levels in ex vivo liver perfusions have been attributed to graft versus host activity⁵. Specifically, porcine Kupffer cells have been implicated for their ability to bind human red blood cells (RBCs) in the absence of prior opsonization^5,12. This interaction is reportedly mediated by porcine sialoadhesin, as demonstrated by a loss of human RBC binding to a porcine cell line expressing a mutant sialoadhesin. Similarly, human erythrocyte binding was inhibited by nearly 90% following the addition of a monoclonal antibody (41D3) directed against porcine sialoadhesin¹³.

Whether genetic modifications of pig donors to alter their carbohydrate expression profile or improve their physiologic compatibility affects the binding of human erythrocytes to pig macrophages has not been evaluated. In this study, we examined the ability of macrophages from wild type (WT) and genetically engineered pigs lacking three carbohydrate antigens (Gal α1,3 Gal or αGal, Neu5Gc, Sd(a)) and carrying several human transgenes (TKO + hTg) to bind human erythrocytes. In addition, the role of sialic acid expression and binding to porcine sialoadhesin was evaluated.

Methods

Animals:

Three lines of genetically modified pigs were provided by eGenesis (Cambridge, MA). Each line lacked the expression of the three carbohydrate antigens, α−Gal, Neu5Gc, and Sd(a) (TKO), and expressed different human transgenes (hTgs) to regulate complement, coagulation, and innate immune responses (TKO + hTgs); Supplemental Figure 1 specifies the genotypes of the TKO + hTg cells used. Naive wild type pigs (Yorkshire) were received through Tufts University. Lungs were harvested as described previously¹⁴. All procedures were approved by the Institutional Animal Care and Use Committee at the Massachusetts General Hospital and were conducted in compliance with the National Institutes of Health guidelines for the care and use of laboratory animals.

Pulmonary Alveolar Macrophage Collection and Storage:

Alveolar macrophages were utilized as a substitute for Kupffer cells due to the relative convenience and low cost of isolation at high yield, and their high sialoadhesin expression¹³. Macrophages were isolated from pig lungs harvested via median sternotomy during a multi-organ procurement surgery. After exposure and division of adjacent vascular structures, the trachea was clamped with the lungs inflated, and the lungs removed en bloc from the chest cavity and kept on ice until use. To collect lung macrophages, bronchoalveolar lavage was performed under sterile conditions by pouring 500cc of cold RPMI medium (Lonza, Walkersville, MD) into the trachea and then re-occluding the trachea. The lungs were gently massaged for 30 seconds, the tracheal clamp was removed, and fluid was drained by gravity into a sterile bowl. The lavage was repeated a second time with fresh RPMI. In a sterile culture hood, the lavage fluid was filtered through a 100-micron strainer to remove fat or other contaminants. The remaining fluid was centrifuged for 10 minutes at 400xg at 4^oC in 50cc centrifuge tubes and the supernatant media removed. If pig red blood cell contamination was visually prominent, ACK lysing Buffer (Gibco, Grand Island, NY) was added at 5cc for 5 minutes, followed by 5cc of RPMI. After centrifugation at 400xg for 10 minutes, cells were resuspended in RPMI and quantified by trypan blue exclusion using a hemocytometer. Cells were frozen at densities ranging from 2 to 20×10⁶ cells/mL in 90% fetal bovine serum (FBS, Sigma-Aldrich, St. Louis, MO) supplemented with 10% dimethyl sulfoxide (DMSO, Sigma-Aldrich, St. Louis, MO). Tubes were placed in isopropanol-filled cryovial containers at −80^oC for one day prior to transfer to liquid nitrogen for long-term storage.

Flow cytometry verification of 1F1 anti-sialoadhesin mAb binding to primary porcine alveolar macrophages:

Isolated primary porcine alveolar macrophages were assessed for surface expression of CD45 (pan-leukocyte marker; Serotec, cloneK252-1E4, cat# MCA1222A647, Raleigh, NC, USA), CD172a (macrophage/monocyte marker; Becton Dickinson, clone 74–22-15A, cat#561498, Franklin Lakes, NJ, USA ), SWC9 (pig monocyte-macrophage differentiation marker; Serotec, clone PM 18–7, cat#MCA 1973F, Raleigh, NC, USA ) and a mouse anti-pig macrophage antibody (clone BA4D5; Serotec, cat#MCA2317F, Raleigh, NC, USA ). The primary porcine alveolar macrophages were also analyzed for Mouse Anti-Porcine CD169 mAb clone 1F1 or isotype control Ab binding¹⁵. The Mouse Anti-Porcine CD169 [1F1] monoclonal antibody was produced by the Nonhuman Primate Reagent Resource (NIH Nonhuman Primate Reagent Resource Cat# PR-0066, RRID:AB_2916076). In this assay, porcine macrophages were suspended in blocking buffer (phosphate-buffered saline (PBS) containing 10% heat inactivated porcine serum) and incubated at 4°C for 30 minutes then stained with 2.98 mg/ml 1F1 mAb or a low endotoxin, azide-free (LEAF) purified mouse IgG2a (isotype control Ab; Biolegend, cat#400224, San Diego, CA, USA) at 10 µg/ml concentration for 30 minutes at 4°C. Then, cells were washed and stained with 1 ug/ml PerCP conjugated goat anti-mouse IgG secondary antibody (Jackson Inc. cat#115-126-072, West Grove, PA, USA) at 4°C for 30 minutes. Samples were acquired on a BD FACSVERSE (Becton Dickinson, Franklin Lakes, NJ, USA) and the data analyzed using FlowJo software (FlowJo, LLC).

Rosetting Assay:

Sterile, tissue-culture treated flat bottom 96-well plates (Corning Inc., Corning, NY) were coated with human fibronectin (Advanced BioMatrix, Carlsbad,CA) at 100µg/ml in PBS for 1–2 hours at room temperature. Macrophages were thawed and resuspended in RPMI media (RPMI+ 10% FBS + Gentamicin (Life Technologies Corporation, Grand Island, NY)). Macrophages were then centrifuged at 700xg for 10 minutes at 4^oC and resuspended in 1ml of RPMI media for counting. Fibronectin coating solution was then removed from all wells and macrophages were plated at 30,000 cells/100µl/well and incubated overnight at 37^oC in 5% CO₂. The following morning, macrophages were observed by brightfield microscopy for adherence to the plate. The media was removed from all wells, and red blood cells were added according to one of four treatment groups: pig RBC, human RBC, human RBC with isotype control IgG, and human RBC with 1F1 mAb. RBC alone groups were plated with 100ul of RPMI media while antibody-treated human RBC wells received RPMI media supplemented with 10ug/mL of isotype control or 1F1 antibody (Non-Human Primate Research Resource, Cambridge, MA), respectively. For the antibody treated wells, the antibodies were added to wells with the macrophages and then incubated for one hour before adding erythrocytes.

For RBC preparation, 1mL of pig or human red blood cells (Innovative Research, Novi, MI) were centrifuged at 900xg for 10 min at 4^oC. Supernatant fluid was removed, and the RBC pellet diluted 1:200 in Hanks’ Balanced Salt Solution without calcium or magnesium (HBSS, Lonza, Walkersville, MD) to yield an RBC concentration of 0.5%. In one experiment, RBCs were treated with neuraminidase from Clostridium perfringes (Sigma-Aldrich, St. Louis, MO) to remove sialic acid residues. In these experiments, 50ul of pelleted RBCs were added to two tubes each containing 20cc of RPMI to which 40ul of neuraminidase solution (25 units/mL in PBS; treated) or 40 µl of PBS (untreated) was then added. After incubation at 37^oC for 1 hour and centrifugation at 900xg for 10 min at 4^oC, the supernatant fluid was removed and the remaining RBC pellet was diluted 1:200 as in other experiments.

The rosetting assay was performed by removing supernatant media from each well and adding 100µl/well of 0.5% RBC, either with or without 10µg/mL IgG or 1F1 mAb according to treatment group, and incubating for 2 hours at 37^oC. Following incubation, wells were washed with HBSS to remove unbound RBC and were imaged by brightfield microscopy at 10x magnification on a Zeiss Observer.Z1 microscope (Zeiss, Oberkochen, Germany). Five to ten images per well were collected using the image acquisition camera for Bioflux 1000 (Fluxion, South San Francisco, CA). Each image was analyzed using the multipoint feature in ImageJ software (National Institutes of Health) to determine the percentage of macrophages which form rosettes (defined as three or more RBCs bound to a single macrophage). GraphPad Prism was used for statistical analysis by unpaired t-test.

Results

Characterization of porcine pulmonary alveolar macrophages

By light microscopy, 65–85% of the cells recovered from the lung lavage exhibited the expected morphology for porcine alveolar macrophages (not illustrated). By flow cytometry, virtually all the CD45+ leukocytes recovered from the lavage fluid of representative WT (95–99%; Fig 1A) and TKO + hTg lungs (95–99%; Fig 1B) expressed macrophage-specific a-mac antigen and SWC9, as well as CD172a and SLA class II (not illustrated), all characteristic markers of mature porcine macrophages¹⁶. Expression of these markers was consistent across all cell lines evaluated. Expression of sialoadhesin, as detected by antibody 1F1 binding, was similar on WT and TKO + hTg cells (Fig 1C; Supplemental Figure 2), demonstrating that the TKO phenotype or other genetic modifications present in these three TKO + hTg lines does not significantly alter expression of porcine sialoadhesin in cell lines evaluated here.

Figure 1. Flow Cytometric Characterization of WT and TKO + hTg pulmonary alveolar macrophages.

Cells recovered from 3 WT and 4 TKO + hTg lung lavage preparations were characterized by flow cytometry. More than 95% of CD45 positive WT (Fig 1A) or TKO + hTg cells (Fig 1B) express SLA Class II (not shown) and SWC9, CD172 and a-mac, three monocyte-macrophage-specific cell lineage markers. WT (115–13) and TKO + hTg Payload 5 (animal ID 1942) cells shown here are representative of the four TKO + hTg lines and three WT preparations studied. TKO + hTg macrophages are phenotypically similar to WT lung macrophages as defined by these markers, and for sialoadhesin expression (Fig 1C), expressed as mean fluorescence intensity revealed by binding of 1F1 mAb.

Pig and human erythrocytes binding to WT and TKO + hTg porcine macrophages

Very minimal allogeneic binding of pig RBCs to pig macrophages was observed: 1.7±0.7% of pig macrophages formed rosettes with pig erythrocytes (Figure 2A, left panel and B). In contrast, 49.4±2.4% of WT porcine macrophages formed rosettes with human erythrocytes (Figure 2A right panel, and 2B). This binding data demonstrates an inter-species incompatibility between pigs and humans. Compared to WT macrophages, 27.2±1.4% TKO + hTg macrophages exhibited rosetting under the same conditions (Figure 3A, top right panel and 3B), demonstrating a significant reduction in rosetting relative to WT macrophages for all three TKO + hTg cell lines evaluated (TKO + hTg 5, 9 and 10; Supplemental Figure 3A). Thus, one or more genetic modifications present in the TKO + hTg macrophages significantly reduces xenogeneic erythrocyte binding.

Figure 2: Human RBC form rosettes to WT pig macrophages.

Representative images of observed rosetting following incubation of WT porcine macrophages with pig (Fig 2A left) or human (Fig 2A right) red blood cells. Rosettes were defined as macrophages bound to three or more RBCs, examples of which are highlighted in black squares. Wells were imaged by brightfield microscopy at 10x magnification. Images are displayed at 30% of original size. Figure 2B illustrates summary data showing that 1.7±0.7% of WT porcine macrophages formed rosettes with pig erythrocytes, whereas 49.4±2.4% of WT porcine macrophages formed rosettes with human erythrocytes. Summary data was derived from images from four individual experiments utilizing macrophages from three different WT pigs (WT 1, WT 25031, and WT 115–13) incubated with either pig (n=11 wells) or human RBCs (n=20 wells). Compiled results are shown as the mean ± SEM. ****p<0.0001 by Unpaired t test.

Figure 3: TKO + hTg genetic modifications and sialoadhesin blockade reduce xenogeneic RBC rosetting.

A. As illustrated in representative examples (Fig 3A), blockade of sialoadhesin with 1F1 mAb, but not isotype control IgG, inhibits RBC rosetting to WT (left panels) and TKO + hTg macrophages (right panels). This observation was confirmed for all three lines of WT (WT 1, WT 25031, and WT 115–13) and four lines of TKO + hTg (1940, 1942, 1977, A9956) macrophages (Supplemental Figure 1). Images are displayed at 30% of original size, and examples of “rosettes” are highlighted in black. A significantly higher percentage of WT porcine macrophages (49.4±2.4% n=20 wells) than TKO + hTg macrophages (27.2±1.4% n=20 wells; p<0.0001) formed rosettes with human erythrocytes (Fig 3B, Supplemental Figure 1A). Results for untreated WT vs. TKO + hTg macrophages are plotted as mean % rosetting ± SEM. Sialoadhesin binding mediates human erythrocyte rosetting by both WT and TKO + hTg macrophages, since 1F1 significantly reduces human RBC rosette formation to WT (3.4±0.5% vs. 42.3±3.2%) and TKO + hTg macrophages (4.2±0.6% vs. 18.3±1.9%) (Figure 3C; ****p<0.0001 by Unpaired t test). The effect of 1F1 treatment was consistently observed for both WT (n=23) and TKO + hTg macrophages (n=24). Rosette data is displayed as mean % rosetting ± SEM. Results with 1F1 and isotype control for human erythrocyte rosetting with each macrophage phenotype, as well as pig erythrocyte controls, are illustrated in Supplemental Figure 1B.

Porcine sialoadhesin mediates adhesion of human erythrocytes to porcine macrophages

Treatment with 1F1 mAb significantly attenuated binding of human RBC to pig macrophages when compared to isotype control (IgG) in both WT (3.4±0.5% vs. 42.3±3.2% p<0.0001) and TKO + hTg (4.2±0.6% vs. 18.3±1.9%, p<0.0001) groups (Figure 3A and C). A reduction in rosette formation in the presence of 1F1 was observed with all four individual TKO + hTg lines (Supplemental Figure 3B), although minimal residual binding (3–5%) was detectable above the pig allogenic control level (1.7%±0.7%). These data suggest that rosetting of human erythrocytes by WT and TKO + hTg macrophages is mediated primarily by porcine sialoadhesin.

Sialic acid mediates adhesion of human erythrocytes to porcine macrophages

Treatment with neuraminidase efficiently removed surface sialic acid residues from human RBCs, resulting in exposure of peanut agglutinin lectin (PNA) binding sites as confirmed by flow cytometry. The geometric mean (gMFI) of PNA binding by neuraminidase treated human RBCs (2779 gMFI) was significantly higher compared to untreated human RBCs (430 gMFI) (Figure 4A). While 47.5±2.7% of WT macrophages (Figure 4B, top left panel and 4C) and 20.8±0.7% of TKO + hTg macrophages (Figure 4B, top right panel and 4C) formed rosettes with untreated human RBCs, only 0.7±0.1% of WT (Figure 4B, bottom left panel and 4C) and 1.2±0.2% of TKO + hTg macrophages (Figure 4B, bottom right panel and 4C) formed rosettes with human RBCs after treatment with neuraminidase. We conclude that sialic acid mediates RBC rosetting by both WT and TKO + hTg macrophages. Inhibition of rosette formation by the porcine genetic modifications and desialation was consistent across all individual TKO + hTg and WT cell sample lines (Supplemental Figure 4).

Figure 4: Adhesion of human erythrocytes to WT and TKO + hTg porcine macrophages is sialic acid-dependent.

A. Flow cytometry demonstrated PNA binding on human RBCs treated with neuraminidase (NA) as compared with untreated human RBCs (Fig 4A). Geometric mean of PNA binding on NA-treated human RBCs was higher than untreated human RBCs (gMFI, treated human RBCs 2779 vs. untreated human RBCs 430), demonstrating PNA binding to desialated glycoproteins on the NA-treated RBC surface membrane. NA treatment abrogated rosetting to both WT (Fig 3B, 3C, left panels) and TKO + hTg macrophages (Fig 3B, 3C, right panels), confirming that sialic acid decoration of human RBCs accounts for virtually all binding of human RBCs to pig macrophages. After treatment with neuraminidase, 0.7±0.1% of WT (down from 47.5±2.7%) and 1.2±0.2% of TKO + hTg macrophages (down from 20.8±0.7%) formed rosettes. Results from 1 experiment are reported as mean % rosetting ± SEM for all cell lines (n=5 for each pig phenotype and condition). **** p<0.0001 by Unpaired t test. Because NA treatment results in lower binding than 1F1 blockade, and the latter is not increased by higher concentrations of 1F1 (not illustrated), we conclude that the residual rosetting after 1F1 treatment is mediated by sialic acid interacting with some other ligand on pig macrophages.

Discussion

Our results reinforce the understanding that the majority of macrophage-erythrocyte binding reflects a fundamental xenogeneic incompatibility between swine and humans, one that is mediated by sialic acid residues expressed on the surface of human but not porcine erythrocytes. The observed rosetting is consistently reduced by either the addition of the 1F1 monoclonal antibody or by desialation of human RBC with neuraminidase, confirming the central role of porcine sialoadhesin in this phenomenon, and predicting that 1F1 will be effective to prevent analogous phenomena during ex vivo organ perfusion or following organ xenotransplantation (discussed further below). In addition, we show for the first time that one or more of the modifications present in the genetically engineered TKO + hTg pigs reduce erythrocyte binding, and potentially sequestration, relative to wild type pigs.

Based on prior work by others¹⁷, we hypothesize that this reduction in erythrocyte rosetting is attributable mainly to the carbohydrate gene knockout in TKO + hTg pigs. Specifically, we speculate that preventing the production of Neu5Gc in CMAH KO animals partially “humanizes” the sialic acid decoration of porcine cells, accounting for the partial reduction in sialic-acid-mediated TKO + hTg macrophage rosetting. Sialoadhesin (also known as Siglec-1, sialic acid-dependent immunoglobulin-like family member lectin 1) is a lectin that was originally described as the molecule on the surface of sheep erythrocytes that binds to sialic acid. Pigs, as well as all mammals except humans, predominantly express sialic acid in the context of N-Glycolylneuraminic acid (Neu5Gc), produced by the enzyme CMP-Neu5Ac hydroxylase. Encoded by the cytidine monophosphate-N-acetylneuraminic acid hydroxylase (CMAH) gene, CMP-Neu5Ac hydroxylase converts CMP-Neu5Ac to CMP-Neu5Gc, resulting in high levels of Neu5Gc expression. Evolutionary changes have rendered the human CMAH gene non-functional, and humans therefore express Neu5Ac (N-Acetylneuraminic acid) and have lost the ability to express Neu5Gc. Importantly, ligation of Neu5Ac by murine macrophages primes them toward a more inflammatory and phagocytic state^18–20. Porcine sialoadhesin preferentially binds terminal α2,3-linked sialic acids, more specifically binding to Neu5Acα2-3Galβ1 residues^21,22, and exhibits a strong binding preference for Neu5Ac over Neu5Gc^23–25. Based on this extensive prior work, we presume that this particular difference in sialic acid expression is the primary basis for the reduced TKO+hTG macrophage recognition of human erythrocytes. This is further supported by our observation of less rosetting following desialation of human RBC’s than the residual binding and rosetting seen between porcine macrophages with porcine RBC’s. Of note, experiments perfusing livers from GalTKO.hCD46.Neu5GcKO pigs exhibited significantly higher mean hematocrit levels than with GalTKO.hCD46 livers in an ex vivo xenoperfusion model, supporting the hypothesis that the additional Neu5Gc knockout contributes to reduced erythrocyte sequestration²⁶. We infer that sialoadhesin on porcine macrophages formed in an environment lacking the CMAH gene exhibit reduced ‘recognition’ of the Neu5Ac present on human erythrocytes as foreign. However, our data clearly demonstrate that sialoadhesin on TKO + hTg macrophages mediates rosetting of human RBCs via sialic acid, presumably expressed on either Neu5Ac, or decorating other human glycoproteins. However, some TKO + hTg macrophages do not bind 1F1 and thus appear to be negative for sialoadhesin, perhaps also contributing to the observed reduction in RBC rosetting.

In addition to the triple carbohydrate gene knockout phenotype, each of the TKO + hTg constructs evaluated here also express human CD47 and human leukocyte antigen (HLA-E), albeit at different levels. For example, TKO + hTg lines 5 and 9 express low levels of both of these transgenes, whereas TKO + hTg line 10 expresses both of these transgenes at higher levels (eGenesis unpublished observations; Supplemental Fig 3)²⁷. HLA-E and human CD47 are human self-recognition receptors that suppress cytotoxicity by human leukocytes, as first shown by transfection into swine endothelial cells that were then incubated with human macrophages²⁸. Expression of human CD47 could affect RBC rosetting to porcine macrophages since human CD47 might interact with human SIRP1a on human RBCs to inhibit macrophage activation. However, absence of a consistent difference between TKO + hTg line 10 and the other two pig lines does not support this hypothesis, given no difference in erythrocyte rosetting was observed between the three TKO + hTg lines despite differences in human CD47 expression. HLA-E expression is intended to inhibit activation of human NK cells but should not be relevant to macrophage interactions with human erythrocytes, which lack the relevant NK cell surface ligands for HLA-E.

Overall, the TKO + hTg genetic modifications reduce RBC rosetting by about 50%. The 1F1 mAb is highly effective to block residual rosetting of TKO + hTg macrophages, demonstrating that the majority of remaining binding is mediated by porcine sialoadhesin, and implicating sialic acid residues expressed by human RBC Neu5Gc. The effect of 1F1 is extremely consistent, reducing rosetting to around 4% regardless of pig macrophage phenotype, very near that of pig macrophages with pig RBCs (~2%) and pig macrophages with desialated human RBCs (~1%). Thus, targeting porcine sialoadhesin with an antibody Fab or Fc-silenced antibody could be useful as a therapeutic agent or potential solution to the problem of human erythrocyte sequestration and destruction by pig livers^29,30 and other organs^{7,31, 32}. Treatment with the 1F1 antibody decreased the rate of loss of human RBCs during extracorporeal perfusion of WT pig livers over a 72 hour period, but did not prevent loss completely¹⁷. Further research will be required to determine what dose of 1F1 is needed to ensure complete saturation and blockade of sialoadhesin in ex vivo organ perfusion conditions, and to identify whether additional sialoadhesin-independent mechanisms exist to drive RBC sequestration.

A limitation of our analytic method is that we only enumerate macrophages which are bound to three or more erythrocytes. Whether macrophages that bind one or two RBCs represent a physiologically important phenomenon is unknown, but seems unlikely, since pig macrophages and pig RBCs also exhibit this behavior in the in vitro rosette assay. There was no significant difference in rosetting across the different TKO + hTg lines despite varying levels of expression of additional complement, thromboregulatory, and anti-inflammatory transgenes²⁷, suggesting that these transgenes do not play a major role in erythrocyte binding under the conditions studied here. However, this important question will require further investigation as pig lines and macrophages with consistently higher expression of each of these individual transgenes become available, enabling comprehensive blocking studies. CD18, a component of the MAC-1 complex, has been implicated in sialic acid-dependent platelet sequestration by porcine Kupffer cells³⁰. Whether that receptor participates in erythrocyte sequestration, independent of sialoadhesin, merits investigation. Finally, whether porcine pulmonary alveolar macrophages accurately reflect the biology of liver Kupffer cells and the detailed carbohydrate expression phenotype of both cell types remain to be investigated.

Conclusion

In summary, we conclude that sequestration of human RBCs by WT or TKO + hTg pig xenografts is mediated by sialic acid on RBCs binding to porcine sialoadhesin, and is efficiently prevented by blocking their interaction using the 1F1 antibody against sialoadhesin. This sequestration may significantly contribute to the observed phenomenon of anemia in xenoperfusion and xenotransplantation experiments. While treatment with 1F1 may provide benefits for the use of swine livers as a bridge to allotransplantation and reduce injury to the recipient’s erythrocytes, it is not practical as a long-term treatment. However, a long-term solution may not be needed if the sialoadhesin system becomes saturated and ceases to contribute significantly to erythrocyte sequestration or damage and anemia. This possibility seems likely, since baboon recipients of pig livers and lungs exhibit recovery from perioperative anemia within a few days without an ongoing transfusion requirement. We predict that transient inhibition of sialoadhesin will be sufficient to reduce early recipient erythrocyte loss and reduce perioperative transfusion requirements. Although it is not known whether accumulation of RBCs within the graft has possible adverse consequences for the organ xenograft, survival of non-human primate recipients of life-supporting heart and kidney xenograft for over six months is reassuring, at least for these organs. If anemia or ongoing sequestration of RBCs in organ xenografts are encountered, particularly for liver and lung xenografts, our work suggests that humanizing porcine sialoadhesin or silencing its expression in the pig organ prior to xenotransplantation will likely address the problem. Genetic modification has the potential to attenuate erythrocyte sequestration to the point where it is manageable, especially in combination with other genetic modifications already shown to be beneficial to block other well-described mechanisms of injury, such as in lung xenograft injury.

Supplementary Material

fS1

Supplemental Figure 1: Gal a1-3Gal and Neu5Gc expression on WT and TKO macrophages. Porcine macrophages were incubated with fluorescently labeled IB4 lectin or antibody against Neu5Gc and evaluated by flow cytometry. Relative to expression on wild-type macrophages, expression of both carbohydrates was significantly lower on the four cell lines examined, confirming the expected phenotype of TKO pigs. Expression of b4Gal was not assessed in this assay. Expression of various human transgenes that were included in each cell line’s gene construct are show as present (+), absent (−), or not included (N/A).

fS3

Supplemental Figure 3: Rosette formation by individual pig lines Rosette formation results for macrophages from each of the 7 pig genotypes studied here are shown (Suppl Fig 3A). Porcine macrophages were incubated with untreated human RBCs, imaged, and quantified. Results from 2 individual experiments are reported as mean % rosetting ± SEM for all cell lines (total number of observations with each macrophage phenotype, from left to right: n=5, n=5, n=3, n=5, n=5, n=5, n=5). Analysis by Unpaired t test, p<0.05 vs WT 1 (*), WT TBRC 25031 (#), or WT 115–13 (+). To compare erythrocyte rosetting and relative dependence on sialoadhesin across the 7 pig genotypes evaluated, porcine macrophages of these genotypes are incubated with porcine RBCs, human RBCs, and human RBCs with isotype control or 1F1 mAb (Suppl Fig 3B). The consistent trend toward reduced rosette formation associated with isotype antibody may be due to reduced macrophage activation in the setting of FcR occupancy without crosslinking, but this phenomenon was not further investigated here. Similarly, differences between WT lines or between TKO + hTg lines in rosette formation with pig or human erythrocytes under reference conditions were not investigated since they were relatively small compared to effects associated with TKO + hTg, sialoadhesin blockade and desialation (shown in Figures 2–4), and within the range of day-to-day variability for this assay in our experience. Results from 2 individual experiments are reported as mean % rosetting ± SEM for all cell lines (n≥3 wells for all groups). Analysis by Unpaired t test, **** p<0.0001, ** p<0.01 vs isotype control.

fS2

Supplemental Figure 2: 1F1 mAb binding to porcine macrophages by flow cytometry. Porcine macrophages were incubated with FC-PBS alone, or with isotype control antibody, the 1F1 mAb, or antibody specific to MHC II (negative control) and assessed by flow cytometry and the data displayed as MFI. (Supplemental 2A). For comparison, 1F1 mAb or isotype LEAF binding to porcine macrophages was shown for each cell line (Supplemental 2B).

fS4

Supplemental Figure 4: Effect of Neuraminidase treatment on rosette formation by individual pig macrophage cell lines. Rosette formation was consistently reduced to 1–4% by neuraminidase treatment for each of 7 cell lines studied. Porcine macrophages were incubated with untreated or neuraminidase treated human RBCs, imaged, and quantified. Results from 1 experiment are reported as mean % rosetting ± SEM for all cell lines (n=5 wells for all treatment groups). Analysis by Unpaired t test, **** p<0.0001 vs untreated human RBC.

Abbreviations:

ACK

Ammonium-Chloride-Potassium buffer

CMAH

Cytidine monophosphate-N-acetylneuraminic acid hydroxylase gene

CMAH KO

Cytidine monophosphate-N-acetylneuraminic acid hydroxylase gene knockout

DMSO

Dimethyl sulfoxide

ECLP

Extracorporeal liver perfusion

FBS

Fetal bovine serum

HBSS

Hanks’ Balanced Salt Solution

HLA-E

Human Leukocyte Antigen E

IgG

Isotype Control Antibody (immunoglobulin G)

mAb

Monoclonal antibody

MFI

Mean Fluorescence Intensity

Neu5Ac

N-Acetylneuraminic acid

Neu5Gc

N-Glycolylneuraminic acid

PBS

Phosphate buffer saline

RBC

Red blood cel

RPMI

Roswell Park Memorial Institute Medium

WT

Wild type