Human Erythrocyte Fragmentation During Ex Vivo Pig Organ Perfusion

Abstract

Platelet sequestration is a common process during organ reperfusion after transplantation. However, instead of lower platelet counts, when using traditional hemocytometers and light microscopy, we observed physiologically implausible platelet counts in the course of ex vivo lung and liver xenograft organ perfusion studies. We employed conventional flow cytometry (FC) and imaging flow cytometry (AMINS ImageStream X) to investigate the finding and found platelet-sized fragments in the circulation that are mainly derived from red blood cell membranes. We speculate that this erythrocyte fragmentation contributes to anemia during in vivo organ xenotransplant.

Introduction

Platelet counts usually drop quickly in the perfusate within 15 minutes after unmodified human blood is perfused through a pig lung or liver. ^1–3 Systematic studies revealed important roles for porcine von Willebrand factor (pvWF) and integrins, including GPIB and GPIIBIIIA.⁴ Surprisingly, during experiments designed to evaluate the role of various genetic modifications to protect the lung and liver from platelet sequestration and other features of organ xenograft injury, we uncovered that after the initial drop, platelet counts unexpectedly rose over time, occasionally significantly exceeding values measured in the blood perfusate before perfusion was initiated. To investigate this physiologically improbable finding, we evaluated multiple independent methods for quantifying platelet numbers.

Hematology analyzers are trusted to yield accurate whole blood platelet measurements in routine human and veterinary clinic practice.^5,6 However, hemocytometers depend on light-scattering properties to distinguish between formed blood components, and primarily use object size to identify and quantify platelets.⁷ Precision of hemocytometers is known to be reduced under some circumstances, such as in the setting of platelet aggregation to each other or to other formed blood cell elements^8,9, or in the context of severe thrombocytopenia.^10,11 In those instances, manual counting methods are generally used to more accurately quantify platelet numbers in whole blood, but manual counting is time-consuming, subjective and has previously been shown not to be accurate at low platelet numbers in some clinical contexts.^12,13 in contrast, flow cytometric platelet counting methods yield accurate, precise, and reliable platelet counts, whether enumerated relative to non-fluorescence RBCs¹⁴ or to precise numbers of fluorescent beads.¹⁵ Contemporary advances in platelet imaging tools like imaging flow cytometry allow detailed, mechanistically informative visualization of platelets biology and cell-cell interactions.¹⁶

Here we report that automated hemocytometry and classical microscopy platelet counting methods both systematically overestimated the number of platelets that remain in the perfusion circuit beyond the first 30 minutes after initiation of perfusion. We report for the first time that platelet-sized fragments are released during the interaction between human blood and pig lungs and livers. We show that these fragments are derived from red blood cells and can be distinguished from platelets by flow cytometry.

Methods and Materials

Blood preparation

Human blood was obtained from volunteers according to University of Maryland School of Medicine and Massachusetts General Hospital Institutional Review Board (IRB) approved protocols. Healthy human volunteers donated approximately 450 mL of whole blood into a blood collection bag containing citrate phosphate dextrose adenine (CPDA-1). Two units of blood were mixed with four units (~240 mL each) of ABO type-compatible thawed plasma. Total combined volume was approximately 2 liters. After the perfusate was heparinized (3 IU/mL blood, Heparin Sodium Injection; Sagent, Schaumburg, IL, USA), CPDA-1 was neutralized by addition of CaCl₂ (1.3 to 1.6 mg/ mL blood, American Regent, Shirley, NY, USA), and sodium bicarbonate (~0.84 mg/mL blood, Hospira, Lake Forest, IL, USA) was added to achieve physiologic pH. Blood, plasma and drugs were mixed thoroughly and decanted into the lung or liver perfusion circuit.

Animals

Lungs from pigs (6-15 kg) genetically engineered to lack the galactose-α(1,3)-galactose epitope (Galactosyl transferase knock-out, GalTKO) and expressing human membrane cofactor protein (hCD46), some of which expressed additional genetic modifications, were provided by Revivicor (Blacksburg, VA, USA). The generation of these pigs¹⁷ and their evaluation in the lung perfusion model have been previously reported.¹⁸ All procedures were approved by the IACUC Committees (The Institutional Animal Care and Use Committee) at the University of Maryland, School of Medicine and at the Massachusetts General Hospital, and were conducted in compliance with NIH (National Institutes of Health) guidelines for the care and use of laboratory animals.

Organ procurement

Pig anesthesia and surgical lung and/or liver dissection were performed as previously described.^19–21 First, 1-benzylimidazole (BIA, 5 mg/kg BW; Sigma-Aldrich, St. Louis, MO, USA, a thromboxane synthase inhibitor), and synthetic prostacyclin I₂ (Remodulin (Treprostinil, 0.06 mg/kg BW; United Therapeutics, Silver Springs, MD, USA) or Flolan (Epoprostenol, 0.03 mg/kg BW; GlaxoSmithKline, Research Triangle Park, NC, USA)) were administered intravenously and allowed to circulate for several minutes. Then, the lungs were flushed with pneumoplegia (0.5-1L; Perfadex; XVIVO, Gothenburg, Sweden) or the livers with Wisconsin solution (Organ Recovery Systems, Chicago, IL, USA). The lungs and livers were then explanted, cannulated as previously described, and stored in iced saline until ex vivo perfusion was initiated.

Ex vivo xenogeneic lung and liver perfusion

Lungs were perfused via the pulmonary artery using circuits fashioned from silicon tubing and polyurethane connectors as previously described.^20,21 Lung failure was defined as rise of pulmonary vascular resistance (>600 mmHg, min/l), loss of vascular barrier function, manifest by development of gross tracheal edema prohibiting lung ventilation, or loss of oxygen transport (step-up in blood oxygen content across the lung). Livers were perfused via the hepatic artery and portal vein as previously described¹⁹ with hepatic venous return draining by gravity into the reservoir. Liver failure was manifested between 4 and 24 hours by precipitous increase in hepatic arterial and portal venous vascular resistance.

Lung and livers from multiple lines of wild-type or genetically modified pigs with a GalTKO.hCD46 background, were perfused with fresh heparinized human blood perfusate to which the αGPIb 6B4 Fab, 1-benzylimidazole (BIA, 5 mg/kg BW, a thromboxane synthase inhibitor), and famotidine (0.5 mg/kg, antihistamine) were added; continuous insulin infusion (titrate based on glucometer readings; Midwest Veterinary Supply, Lakeville, MN, USA), heparin (titrated to ACT>400), and Remodulin (2 mg/h; United Therapeutics, Silver Spring, MD, USA) were also added. Some pigs also received DDAVP pre-treatment (Desmopressin Acetate, 3 mg/kg BW IV; Diamondback Drugs, Scottsdale, AZ, USA) on day −1 and prior to organ procurement to deplete intracellular endothelial von Willebrand’s factor.^4,19,22

Sampling

Blood perfusate samples for hematologic analyses were collected into tubes containing EDTA (ethylenediaminetetraacetic acid, BD catalogue # 367839) or CTAD (citrate, theophylline, adenosine, and dipyridamole, BD catalogue # 367599). A blood sample was collected after the blood mixture was circulated and warmed in the perfusion circuit for at least 3 min (“R0”). The R0 sample served as the reference (“baseline”) to which later samples are compared. Samples of pulmonary or hepatic vein effluent collected at 5, 15, 30, 60, 120, 240 and 480 min following the start of organ perfusion were handled as detail below.

Automated counting of platelets using hemocytometers

Blood platelet counts were measured by hemocytometry from perfusate collected in EDTA tubes. Samples were evaluated using standard automated techniques that are widely available in the clinical practice. In order to detect whether consistent ‘inter-machine’ differences exist, in 50 experiments identical samples from multiple points were evaluated in parallel with two or more hemocytometers. Samples were evaluated with ‘in-house’ hemocytometer (HC1) (Hemavet950FS, Drew Scientific, Miami Lakes, FL, USA), generally within 30 minutes after collection; if not processed immediately, samples were kept on rocker at room temperature until analyzed. ‘In-house’ measurements were typically performed in duplicate to assess whether ‘intra-machine’ variability contributes significantly to reduce assay result accuracy. After performing the in-house HC assay, the same samples were kept at 4ºC until shipment in batches at room temperature to a commercial laboratory (Advia120 machine (HC2), Antech Diagnostics, Rockville, MD, USA), where they were typically evaluated in less than 24 hours. Some fresh samples stored at room temperature were also analyzed on the day of collection using a third HC, a UniCel DxH 800 (Beckman Coulter, Inc., Miami, FL, USA), housed at the core facility of Baltimore VA Medical Center.

Manual counting of platelets with Thrombo-TIC®

Thrombo-TIC® (Bioanalytic GmbH, Germany) is a commercially available kit for manual counting of platelets by light microscopy. After transferring 10 μL of blood sample from the same EDTA tube used for the HC assays into a Thrombo-TIC® vial, and waiting 5 minutes for red blood cell lysis, samples were transferred into the Thrombo-TIC® counting chamber and platelets counted under a light microscope. Two independent measurements by two trained persons were performed, each counting 25 group squares, and using the sum to calculate the estimated platelet number using the formula: Sum*1000 = platelets/μL of blood.

Flow cytometry method:

Preparation of stock solution & diluent

A stock solution was prepared [22% BSA (bovine serum albumin, Sigma-Aldrich, St. Louis, MO, USA), 500 mM K₂-EDTA (Sigma-Aldrich, St. Louis, MO, USA) in 1X PBS (phosphate-buffered saline, Gaithersburg, MD, USA)] and stored at 4°C until further use as described previously by van der Meer et al.¹⁵ The stock solution stored at 4°C is stable up to a year (personal communication with Dr. van der Meer). A diluent was prepared by diluting the stock solution in a 1:50 ratio. This diluent was prepared fresh on the day of the experiment and used to dilute the blood samples and antibodies.

Preparation of antibodies

Anti-CD41a-FITC (Clone PM6/248, BIO-RAD, Hercules, CA, USA), anti-CD61-PerCP (Clone VI-PL2, Biolegend, San Diego, CA, USA) antibodies were diluted in small Eppendorf tubes using the freshly prepared diluent. IgG-FITC (1:30; BD) and IgG-PerCP (1:30; Biolegend) were used as proper isotype controls. The ratio of antibody dilutions was determined after serial titrations to reach the optimal staining (data not shown). Prepared antibody dilutions were stored in the dark at 4°C for up to a week until further use.

Handling and evaluation ‘fresh’ blood samples by flow cytometry

The blood in EDTA tubes was stored at room temperature, mixed by gentle inversions 8--10 times, and a sample stained within 60 minutes of collection.²³ Samples were first diluted in a 1:10 ratio with diluent (See ‘Preparation of stock solution & diluent’ section above). Twenty μL of the diluted blood sample and 20 μL of diluted antibodies were placed in a clean 5 mL polystyrene round-bottom FACS tube (Falcon, catalogue # 352052) and incubated for 15 minutes in the dark at room temperature. After incubation, 50 μL of counting beads [SPHERO™ AccuCount Particles (5-5.9um), Spherotech, Inc., Lake Forest, IL, USA] and 400 μL of diluent were added to each tube.

After completion of the staining steps, samples were promptly acquired on a BD FACSVERSE flow cytometry machine (BD Biosciences). Regular BD FACSVERSE calibration and maintenance were done according to the manufacturer’s recommendations. On the flow cytometer, the “modified” no lyse/no wash setting was used and a threshold on FSC (forward scatter) was set to 200 to maximize the acquisition of small particles (this threshold was set by the manufacturer and cannot be removed, creating a lower boundary constraint on data collection using this instrument.) There was no set threshold on SSC (side scatter), FITC or PerCP channels. A gate around ‘counting beads’ events was created, and the acquisition was stopped automatically after acquisition of 10,000 events in this gate. Data were analyzed using FlowJo single-cell flow cytometry analysis software (FlowJo, LLC., Ashland, OR, USA). On FSC-FITC or FSC-PerCP plot, counting beads, CD41-positive and CD61-positive events were gated and counted using FlowJo (Figure. 1A). Platelet /μL were calculated by multiplying the number of beads added to the tube with the number of platelets counted and the dilution factor, and divided by the number of beads counted and the sample volume.¹⁵

Figure 1.

Development and validation of flow cytometry-based assay in ‘fresh’ whole blood. (A) The platelet gating strategy is illustrated for CD41a (left) and CD61 (right). (B) Comparison of pre-perfusion platelet counts (‘PRE’) in fresh blood samples (n = 17), illustrating very good correlation of results for CD41a or CD46 by Pearson correlation (R = 0.9985, p<0.0001, n = 17). (C) Comparison of results using either CD41 or CD61 to detect platelets in fresh blood sampled over time during three representative ex-vivo perfusion experiments show close correlation between the 2 anti-platelet antibodies, suggesting that either antibody can be used to identify platelets (2way ANOVA not significant p = 0.1897).

Handling and evaluation of ‘fixed’ blood samples

To compare results between fresh and fixed platelets by the FC method, platelets counts in pre-perfusion whole blood samples were measured by the two hemocytometers, and an aliquot of each blood sample was fixed and stained for platelet markers. For the preparation of the fixative buffer, 4 mL of deionized water, 0.5 mL of 10x HBSS (Gibco, Gaithersburg, MD, USA) and 0.5 mL of 10% paraformaldehyde (Electron Microscopy Sciences, Hatfield, PA, USA) were combined. 100 μL of whole blood from CTAD or EDTA tubes were rapidly added to 1.4 mL of fixative buffer in Eppendorf tubes and kept at 4°C without mixing/shaking. Fixed samples were allowed to reach room temperature (15-30 minutes) before staining, as described for fresh samples. Fixed samples were evaluated by flow cytometry within 5 days after preparation, after vortexing at low speed to resuspend the sedimented cellular pellet; data acquisition and analysis were executed as described for fresh blood samples.

Imaging Flow Cytometry AMNIS:

Imaging flow cytometry using Image StreamX Mark II imaging flow cytometry (AMNIS Corporation) equipped with a 40× objective, six imaging channels, and 405, 488, and 642 nm lasers was employed to visualize and identify different blood population. 50 μL of fixed blood samples and 20 μL of diluted antibodies were incubated in the dark room temperature for 15 minutes. After incubation, 500 μL of diluent were added to each tube. The anti-CD41 antibody was used to identify platelets, whereas intact RBCs and RBC fragments were identified as CD235a positive events. The AMNIS analysis was performed using IDEAS Application Version 6.2.187.0 (AMNIS part of EMD Millipore). A Brightfield plot versus CD41a Intensity plot was used to identify and gate on platelets, intact RBCs, CD235a positive platelet-sized particles, and other platelet-sized events.

Hemolysis assay:

Hemolysis of human red blood cells was quantified over the course of ex-vivo perfusion of five pig lungs with GalTKO.hCD46 background and six GalKO.hCD46.hvWF livers. Free hemoglobin was determined by measuring the optical density in EDTA plasma (stored at −80°C until assayed) at 414nm wavelength using a spectrophotometer (SpectraMax, iD3 Molecular Devices) with reference to a standard curve of known human hemoglobin concentration, diluted with ddH₂O and expressed as milligrams per milliliter. DdH₂O was used as blank buffer.

Statistical analysis

All the statistical analysis (One-way and two-way ANOVA, Student’s t-test, Pearson correlation) were performed using GraphPad Prism Software (La Jolla California, USA). The Pearson’s correlation coefficient, R, was calculated to test strength of correlation. A P-value below 0.05 was considered statistically significant.

Results

Validation of platelet counts by blood flow cytometry in fresh and fixed samples

The gating strategy used to detect the platelet and bead events is illustrated in Figure 1A (CD41a left plot, CD61 right plot). Platelet counts by flowcytometry using CD41a and CD61 positive events as platelet markers in pre-perfusion fresh blood (Figure 1B) showed significant correlations (Pearson correlation. R = 0.9985, p < 0.0001, n = 17). Fixed whole blood samples were stained with CD41a and CD61 markers in three different ex-vivo perfusions (Figure 1C). The comparison between CD41a and CD61 in fresh and fixed samples resulted in a difference that was not statically significant (2way ANOVA, p = 0.1897, n = 3), showing that platelet activation did not significantly distort the relative expression of these 2 antigens. We conclude that antibodies against either CD41a or CD61 can be used interchangeably in this assay.

Comparison of platelet counts by flow cytometry and hemocytometery in fresh and fixed samples

Pre-perfusion platelet counts measured by FC in fresh whole blood (Figure 2A) correlated closely with both hemocytometers (FC vs. HC1 p = 0.1119 by paired t-test, n = 23; FC vs. HC2 p = 0.0720, n = 20). The comparison of the results between the two hemocytometers revealed a significant difference between HC1 and HC2 (paired t-test, HC1 vs. HC2 (p = 0.0014, n = 18). Comparing platelets counts in fixed blood (Figure 2B) by FC to measurements in fresh blood by the two hemocytometry devices showed non-significant variation in results between assay platforms (FC vs. HC1 p = 0.2690 by paired t-test; FC vs. HC2 p = 0.3533; HC1 vs. HC2 p = 0.1123, n = 45). FC platelet counts using fresh and fixed samples of pre-perfusion blood correlated significantly with each other (Pearson correlation: R = 0.7577, p < 0.0001, n = 22. Figure 2C). The comparison of platelet counts by FC using fresh and fixed blood samples during three ex-vivo perfusions showed no significant differences (2way ANOVA, p = 0.1387, n = 3. Figure 2D), validating use of the latter method when convenient.

Figure 2.

Comparison of absolute platelet count results measured in pre-perfusion blood (‘PRE’), using either fresh or fixed blood samples. Mean, 25--75 CI, and range of platelet counts was similar for both HCs and FC in ‘fresh’ blood (3A) and ‘fixed’ blood (3B). By paired t-test results with FC were similar to either HC method, whereas HC1 and HC2 differed significantly for fresh blood only: FC vs. HC1 (p = 0.1119, n = 23), FC vs. HC2 (p = 0.0720, n = 20), HC1 vs. HC2 ** (p = 0.0014, n = 18). For ‘fixed’ blood: FC vs. HC1 ns (p = 0.2690 n = 45), FC vs. HC2 ns (p = 0.3533, n = 45) HC1 vs. HC2 ns (p = 0.1123, n = 45). (C) Platelet counts measured by FC in fixed versus fresh ‘pre’ blood samples showed significant correlation (r = 0.7577, p < 0.0001, n = 22). (D). Platelet counts measured by FC in paired fixed and fresh blood samples from in three perfusion circuits showed no statistical differences, indicating that sample fixation does not consistently affect accuracy of the FC platelet enumeration method (2way ANOVA; P = 0.1387, n = 3).

Platelet counts during ex vivo pig lung and liver perfusion with human blood

Platelet counts and % remaining platelets were measured by two HCs and FC during five GalKO.hCD46 pig lung (Figure 3A) and six GalKO.hCD46.hvWF pig liver (Figure 3B) ex-vivo perfusion experiments using human blood. While the platelet counts measured by FC showed a decreasing trend over time, the platelet counts by HC showed an increasing trend and large variations within the first 4 hrs of lung and liver perfusion. Differences in platelet counts between FC and each of the two HCs are statistically significant in both lung and liver ex-vivo perfusions (for ex-vivo lung perfusions: paired t-test HC1 vs. HC2 p = 0.011; FC vs. HC1 p = 0.0289; FC vs. HC2 p = 0.0038; for ex-vivo liver perfusions: paired t-test HC1 vs. HC2 p = 0.0181; FC vs. HC1 p = 0.0006, FC vs. HC2 p = 0.0015). FC method showed similar patterns across different individual experiments, whereas large variations were observed between two HC methods (Supplementary Figure 1).

Figure 3.

Comparison of platelet counts measured by flowcytometry (FC) relative to results from two hemocytometers (HC1, HC2). In five GalTKO.hCD46 ex-vivo lung perfusion experiments (A) and in six GalKO.hCD46.hvWF ex-vivo liver perfusion experiments (B), platelet counts by two HCs (HC1 in green and HC2 in blue) showed an increasing trend beyond 15 minutes during lung perfusions, with large variations between experiments. The steady, consistent decline in platelet counts measured by FC (red lines) was significantly attenuated when measured by either HC. As reflected by the differences in SEM magnitude for each platelet enumeration method, relatively large variations between experiments in platelet counts by HC were not observed in FC results. Results are expressed as mean +/− SEM relative to pre-perfusion platelet counts in the fresh human blood perfusate.

Detection and characterization of platelet-like events

FC analysis demonstrates that the higher measured platelet numbers in HC and manual count data after 15 minutes of perfusion likely reflect reporting of platelet-sized CD41- and CD61-negative events that are detected by flow cytometry in steadily increasing numbers (Figure 4A). These CD41/CD61-negative events appear to largely account for the discrepancy between the various counting methods, since platelet counts obtained by hemocytometry correlated closely with the sum of FC-measured platelets and CD41/CD61-negative platelet-sized objects, as illustrated in GalTKO.hCD46 pig lungs experiments (Figure 4C). Manual counting of platelets in one GalTKO.hCD46.hCD55.hEPCR.TFPI.hCD47 pig lung by two independent technicians did not distinguish platelets from platelet-sized cell fragments (Supplementary Figure 2).

Figure 4.

Change in number of platelet-sized events detected by FC through the course of a representative lung perfusion experiment. CD61 positive platelet events (middle gates, 4A panels) decline significantly over time (T60 and T480, relative to initiation of perfusion at T0), while the population of platelet- sized events lacking human platelet markers (leftward CD61 low/negative gate in 4A, 4B) and that express CD235a (gate, 4B) increase throughout the perfusion. (4C) Absolute counts of platelets by HC (black bars) are compared to the sum of platelets (red bars) and platelet-sized events (hatched bars) recorded by FC in paired samples; aggregated results from five GalTKO.hCD46 ex-vivo lung perfusion experiments are shown. The sum of platelets and platelet- sized events by FC approximates the platelet counts reported using the HC instrument for the first 15 minutes, after which FC consistently detects larger numbers of events.

Detection of human red blood cells fragmentation

The majority of CD41/CD61-negative platelet-like events detected in FC are positive for CD235a, a red blood cell marker (62 ± 13% after 4 hours of perfusion, Figures. 4B, 5B). Staining for pig- (pCD45) or human-specific (hCD45) white blood cell markers, pig endothelial cells (CD31), myelo-monocyte lineage antigens (CD14) and for the prevalently-expressed swine-specific cell surface marker SW3 were interpreted as negative (less than 2% positivity among platelet-sized particles -- data not shown). The AMNIS imaging flow cytometry method confirmed that platelet-sized events are positive for CD235a (Figure 5A and 5B), and negative for CD41 (Figure 5B). The population of CD235a+/CD41− platelet-sized events increased over the four hours of the perfusion (Supplementary Figure 3).

Figure 5.

Visualizing platelets and characterize platelet-sized events using the AMNIS imaging flow cytometry system. A representative scatter plot (4A) illustrates the gating strategy used to identify and image (4B) various cell populations by fluorescent microscopy. Representative images of individual RBCs (4A light blue gate; 4B first panel), RBC clusters (4A green gate; 4B second panel), platelets (4A light orange gate; 4B 3^rd and 7^th panels), RBC-platelet aggregates (4A purple gate; 4B 4^th panel) and damaged RBC’s (4A pink gate; 4B 5^th panel), platelets-sized RBC fragments (4A dark orange gate; 4B last two panels), and flow beads (4A grey gate; 4B 6^th panel) are show as visualized using this technique. AMNIS imaging confirms that platelet-sized events that are negative for platelet markers are typically positive for CD235a.

Human plasma hemoglobin levels were increased towards the end of lung and liver pig perfusions. The mean increases from initial values were statistically significant after 60 minutes of lung perfusions (Paired t-test, 0’ vs. 60’ p = 0.0298, 0’ vs. 120’ p = 0.0260, 0’ vs. Final p = 0.0452, n = 5, Figure 6A) and after 4 hours of liver perfusions (Paired t-test, 0’ vs. 240’ p = 0.0298, 0’ vs. Final p = 0.0070, n = 6, Figure 6B). These measurements are consistent with the increasing number of RBC fragments. Thrombocytopenia and hemolysis were not observed when human blood was perfused through the circuit without a pig lung (data not shown), demonstrating that these phenomena are not an artifact of the perfusion circuit, but rather a consequence of human blood interacting with porcine organs.

Figure 6:

Plasma free hemoglobin during ex vivo perfusion of pig organs with human blood. During five GalTKO.hCD46 lung ex-vivo perfusion experiments (6A) and six GalKO.hCD46.hvWF liver ex-vivo perfusion experiments (6B), plasma free hemoglobin increased significantly, reflecting progressive . RBC lysis. The mean differences from initial were statistically significant after 60 minutes, 120 minutes and final in lung perfusions (Paired t-test, 0’ vs. 60’ p = 0.0298, 0’ vs. 120’ p = 0.0260, 0’ vs. Final p = 0.0452, n = 5) and after 4 hours of liver perfusions (Paired t-test, 0’ vs. 240’ p = 0.0298, 0’ vs. Final p = 0.0070, n = 6). Data are expressed as mean +/− SEM relative to pre-perfusion free hemoglobin in the human blood perfusate.

Discussion

Thrombocytopenia is commonly observed in organ xenograft ex vivo perfusion models^2,4, and in in vivo transplant models.^24–30 In order to study the mechanisms contributing to thrombocytopenia, it is essential to accurately enumerate platelets and to distinguish them from cell fragments that may appear in whole blood in the context of our cross-species organ perfusion studies. Here we show that, in ex vivo perfusion models, conventional platelet counting methods fail to discriminate platelets from RBC fragments. The platelet counts by HC are accurate for the first 15 -- 20 minutes, as confirmed here by FC. After this interval, we show that traditional techniques used to enumerate platelets in whole blood overestimate platelet numbers, Specifically, we utilize an adaptation of 2 previously described flow-cytometry-based platelet enumeration methods^14,15 and demonstrate that red blood cell fragments appear in increasing numbers over time during ex vivo liver or lung perfusion and could interfere with platelet count results from conventional methods.

RBC loss and fragmentation during porcine organ perfusion^31–33 or by neutrophil extracellular traps^34,35 has been studied by different groups. Sialoadhesin-mediated immunologic reactions and ‘bystander’ complement activation were identified as contributors to free hemoglobin release in those xeno organ perfusion models.^31,32 Erythrocyte lysis, suspected clinically in the later timepoints of our lung and liver xenogeneic perfusion models since the plasma became darker compared to initial pre plasma samples, was confirmed by increased plasma free hemoglobin elaboration and discovery of platelet-sized erythrocyte fragments by FC. Importantly, identification of this phenomenon of RBC fragmentation and development of the FC method will facilitate simultaneous investigation of both platelet sequestration and erythrocyte injury mechanisms in ex vivo xenogenic organ perfusion models.

There are multiple prior reports that not only RBC fragments but also bacteria, fungi, and fragments of nucleated cells may also cause seemingly normal/high platelet counts in various clinical situations, especially in the context of severe thrombocytopenia.^36,37 For example, in the presence of RBCs with extremely low volume, hemocytometers may report spuriously elevated platelet counts.^37–39 In addition to RBC fragments or schistocytes, fragmented cytoplasm of nucleated cells (such as leukemic blasts, monoblasts, lymphoblasts etc.) may also cause pseudo-normal/high platelet counts.^36,40,41. As reported in a neonate on veno-arterial ECMO support, severe RBC fragmentation may result in erroneously high platelet counts by automated counting methods, masking the severity of thrombocytopenia,⁴² Our findings corroborate the conclusion reached by other researchers: that immunoplatelet counting provides more reliable platelet counts than do hemocytometry-based assays. Indeed, the FC approach has been reported to allow avoidance of unnecessary platelet transfusions.^43,44 Our ex vivo studies support this technical guidance and extend it to studies of ex vivo perfusion of pig lungs and livers. By distinguishing a platelet-sized population that increased over time simultaneously with worsening thrombocytopenia, FC enables dissection of the various mechanisms driving these two phenomena. AMNIS imaging flow cytometry combines the tools of light microscopy with conventional flow cytometry, adding complementary additional information regarding cellular morphology along with the ability to simultaneously detect multiple fluorescents markers and thus reveal spatial distribution of several different cell surface molecules.^35,45

Whether the phenomenon of erythrocyte fragmentation and artificial overestimation of platelet counts occurs in vivo in association with organ xenotransplantation is unknown. Importantly, in our preliminary work to date, we find that HC platelet counts appear to be accurate when correlated with FC assays in the peripheral blood of lung and heart xenograft recipients. Specifically, RBC fragments are not observed in the peripheral blood when we have looked for them. If baboon or cynomolgus RBCs are damaged in vivo consequent to interaction with porcine endothelium, we presume that damaged RBCs are efficiently scavenged by the reticuloendothelial system of the recipient, likely accounting for our inability to detect them. Whether more subtle perturbations in actual platelet counts or elaboration of either donor or recipient cell fragments can be detected in in vivo xenotransplantation models should be revealed by systematic application of the flow-based immunostaining methods in relevant models.

Variations in platelet results between different HCs, and sometimes even in the re-runs of the same sample in the same machine, were more frequent and greater in magnitude than we expected. Superimposed upon prominent and mechanistically implausible fluctuations over the course of some individual experiment, in our estimation this variability renders HC data insecure and its interpretation challenging in the ex vivo model. Therefore, we conclude that FC should be considered the ‘reference method’ for ex vivo organ perfusion studies. FC appears to avoid artefactual detection of platelet-like particles when that phenomenon occurs and appears to consistently yield accurate platelet counts relative to conventional methods (HC, microscopy) under circumstances where RBC fragments are not present. We conclude that our technical modifications of the 2 previously reported FC reference methods allows accurate enumeration of platelets in blood sampled over time from ex vivo lung and liver perfusion circuits when used to model xenotransplantation of these organs. Moreover, our discovery of RBC fragmentation using FC to accurately enumerate platelets, as independently confirmed by AMNIS imaging, will allow mechanistic study of this phenomenon. We hypothesize that red blood cell damage is associated with transit through an organ xenograft and may be related to the anemia that is frequently observed in organ xenograft models, which in our anecdotal experience is usually out of proportion to surgical blood loss.