Interactions of Cardiopulmonary Bypass and Erythrocyte Transfusion in the Pathogenesis of Pulmonary Dysfunction in Swine

Background:Allogeneic erythrocyte transfusion in cardiac surgical patients is associated with a fourfold increase in pulmonary complications. Our understanding of the processes underlying these observations is poor and there is no experimental model of transfusion-related acute lung injury that shows homology to cardiac surgical patients. Our objective was to develop a novel swine recovery model to determine how two clinical risk factors, allogenic erythrocyte transfusion and cardiopulmonary bypass, interact in the genesis of postcardiac surgery acute lung injury. Methods:Thirty-six pigs were infused with allogeneic 14- or 42-day-old erythrocytes or they underwent cardiopulmonary bypass with or without transfusion of 42-day erythrocyte. Controls received saline. All pigs were recovered and assessed for pulmonary dysfunction, inflammation, and endothelial activation at 24 h. Results:Transfusion of stored allogeneic erythrocytes in pigs compared with sham caused pulmonary dysfunction characterized by reduced lung compliance (mean difference −3.36 [95% CI, −5.31 to −1.42] ml/cm H2O), an increase in protein levels in bronchoalveolar lavage fluid, histological lung injury inflammation, and endothelial activation. Transfusion of blood stored for up to 42 days resulted in greater protein levels in bronchoalveolar lavage fluid, macrophage infiltration, platelet activation, and depletion of T-lymphocytes in recipient lungs versus 14-day-old blood. Transfusion interacted with cardiopulmonary bypass to increase lung injury in the absence of platelet activation. Conclusions:In this novel large animal model of allogeneic erythrocyte transfusion, pulmonary dysfunction occurs in the absence of any priming event, is increased when combined with other inflammatory stimuli, and is mediated by monocyte activation and T-lymphocyte depletion.

Patel et al.

Erythrocyte Transfusion and Pulmonary Dysfunction
P ULMONARY morbidity is an important contributor to mortality and resource use after cardiac surgery with cardiopulmonary bypass (CPB). 1,2 Transfusion of allogeneic erythrocytes increases the risk of pulmonary morbidity postcardiac surgery by as much as fourfold. 3,4 This effect may be influenced by the duration of erythrocyte storage before transfusion. Koch et al. 4 demonstrated an increased risk of respiratory insufficiency and the need for prolonged ventilation in patients receiving erythrocytes stored for greater than 14 days versus those receiving blood stored for less than 14 days. Transfusion-related acute lung injury (TRALI), defined by a consensus definition 5 of hypoxaemia (PaO 2 /FIO 2 <300 mmHg), bilateral infiltrates on chest radiograph, pulmonary artery occlusion pressure of less than 18 mmHg, and the acute onset of features within 6 h after a transfusion is rare, occurring in 2.4% of cardiac surgical patients. 6 However, transfusion-associated circulatory overload and transfusionassociated dyspnoea are other syndromes of transfusionrelated morbidity, which overlap with the features of TRALI. These syndromes are poorly defined and consistently underreported. Consequently, the true incidence of transfusionassociated pulmonary morbidity is unclear. 7 Moreover, the associations between pulmonary complications and erythrocyte transfusion are derived from observational studies, so causality has therefore, not been established, and the underlying pathophysiological mechanisms are poorly understood. The latter can be attributed to the lack of experimental models with homology to TRALI, as observed in a clinical setting. Current TRALI models are also limited by cross-species or ex-vivo design, the use of priming events such as lipopolysaccharide that have limited homology to clinical events, and methods of erythrocyte storage dissimilar to those used for human erythrocytes. [8][9][10][11][12] At present, there is no large animal model where transfusion of an allogeneic cellular blood component causes acute lung injury. The objectives of this study were therefore: (1) to determine whether transfusion of allogeneic erythrocyte causes pulmonary dysfunction in swine, (2) to assess whether erythrocyte storage duration affects severity of pulmonary dysfunction, and (3) to evaluate the interaction of erythrocyte transfusion with CPB in the genesis of postcardiac surgery pulmonary dysfunction. We hypothesized that the transfusion of older erythrocytes would cause greater pulmonary dysfunction compared with younger erythrocytes and that erythrocyte transfusion would interact with CPB to increase the severity of pulmonary dysfunction.

Materials and Methods
Animals received care in accordance with and under license of the Animals (Scientific Procedures) Act 1986 (London, United Kingdom) and conform to the Guide for the Care and Use of Laboratory Animals published by the U.S. National Institutes of Health (Publication No. 85-23, revised 1996). The study received local institutional review board (University of Bristol, Bristol, United Kingdom) approval.

Intervention
Thirty-six pigs were quasirandomized to the following groups: After 30 min of CPB or sham procedure, 500 ml of crystalloid or allogeneic erythrocytes were transfused over 2 h during the remainder of the intervention period.

Outcomes
Storage-related Changes in Stored Porcine Erythrocyte Units. Five milliliter aliquots were removed from representative bags for evaluation of storage-related changes. Biochemical changes in the supernatant, plasma hemoglobin using spectromorphometry, and erythrocyte adenosine triphosphate concentrations using high-performance liquid chromatography, were assessed at weekly intervals, as previously described. 13,15 The hemolysis index was defined as plasma (hemoglobin × hematocrit)/donor unit hemoglobin. Scanning electron microscopy was performed of stored porcine erythrocytes at days 0, 14, and 42 of storage. Assessment of Pulmonary Dysfunction. Pulmonary dysfunction was determined according to functional, histological, and biochemical parameters of acute lung injury, as recommended by the American Thoracic Society. 16 Lung compliance, PaO 2 /FIO 2 ratio, airways resistance, and work of breathing were measured in-vivo at baseline, 1.5 and 24 h postintervention, using the SERVO-i Universal Ventilator (Maquet GmbH, Rastatt, Germany),which used volumecontrolled ventilation with a tidal volume of 10 ml/kg, FIO 2 of 0.5, respiratory rate of 12 breaths/min, and peak endexpiratory pressure of 5 cm H 2 O. Total protein in bronchoalveolar lavage samples was measured using a Bradford Protein Assay (Quick Start Bradford Protein Assay, Hercules, CA) at 24 h postintervention by an investigator blinded to intervention allocation. The lower lobe of the left lung was harvested 24 h postintervention, immediately fixed in 10% formalin, and six lung sections taken sequentially across the resected lung were stained with hematoxylin and eosin. A histology scoring system for lung injury was used by investigators blinded to intervention allocation, as previously described. 16 Briefly, the following parameters were scored on a scale of 0-2: (1) neutrophils in the alveolar space, (2) neutrophils in the interstitial space, (3) hyaline membranes, (4) proteinaceous debris filling the airspaces, and (5) alveolar septal thickening. The sum of each of the five variables (each with a different weighting) was normalized to the number of fields evaluated. The resulting score is a continuous value between 0 and 1 (inclusive). The mean lung injury score was then obtained for each group. Pulmonary Inflammatory and Platelet Cell Infiltration and Activation, and Endothelial Activation. Tumor necrosis factor-α (R&D Systems, Abingdon, United Kingdom) was measured in homogenized lungs obtained 24 h postintervention using solid phase enzymelinked immunosorbent assay according to manufacturer's protocol, and normalized to lung protein concentrations.

Statistical Analysis
The current study represents an analysis of pulmonary dysfunction from a series of experiments powered to detect differences in creatinine clearance as a primary endpoint. 13,14,17 The study outcomes to assess pulmonary dysfunction were prespecified secondary outcomes in these experiments, however, no power calculation was performed a priori, and our analysis should be considered exploratory.
Comparisons between groups were performed using oneway and repeated measures ANOVA with the Bonferroni correction. General linear model ANOVA, evaluating time, group, and the interaction of time and group, was used for repeated measures with adjustment for baseline values. Data were reported throughout as mean (±SEM) for normally distributed or as geometric mean (±95% CIs) for nonnormally distributed data. Treatment differences were reported as mean difference (95% CIs) or as the ratio of geometric means (95% CIs). Statistical significance was defined as a P value less than 0.05, using two-tailed tests. All analyses were conducted using SPSS 18.0 (SPSS Inc., Chicago, IL).

Results
Thirty-two of 36 animals completed the experimental protocol to recovery and reassessment. Baseline characteristics, including lung function, lung tidal volumes, gas exchange, and inflammatory markers were similar between groups (see appendices 1 and 2). CPB pigs required a greater volume of intravenous fluids due to the pump priming volume. Four experiments (2 pigs in D14 Tx group and 2 pigs in CPB + Tx group) were terminated prematurely due to cardiovascular instability or refractory hypoxemia. These pigs were excluded from our analyses. To assess whether this may represent a source of bias, our analysis was repeated incorporating these animals up until the time of death. However, this did not alter our findings (see Supplemental Digital Content 1, tables 1-4, http:// Patel et al.

Erythrocyte Transfusion and Pulmonary Dysfunction
links.lww.com/ALN/A929, which are tables demonstrating the sensitivity analyses for experiment 1 and 2).

Stored Porcine Erythrocytes Develop a "Storage Lesion"
To determine homology between porcine and human erythrocyte units, biochemical changes of porcine erythrocytes were analyzed and compared with human data ( fig. 1A). Hematocrit concentrations remained constant during storage, although the mean hematocrit was lower in porcine units compared with human units, a reflection of the lower hematocrit in porcine venous blood. The day-14 porcine erythrocyte units developed a storage lesion, which showed considerable homology to that observed in 42-dayold human erythrocyte units, as shown by an increase in potassium and oxygen levels and a decrease in sodium and erythrocyte adenosine triphosphate concentrations. However, acidic pH in sucrose-adenosine-glucose-mannitol stored porcine units appeared to inhibit glycolysis; glucose was not utilized and adenosine triphosphate levels became depleted more rapidly than those observed in human units. Concentrations of 2,3-diphosphoglycerate in porcine erythrocytes decreased over storage time in a similar manner to human erythrocytes. The hemolysis index was higher in porcine units at day 42, but the mean hemolysis index (mean 0.87 [±0.59]) remained less than 1, a quality control standard for human erythrocyte storage. Porcine erythrocytes underwent morphological change over storage time, characterized by the loss of biconcavity and development of echinocytosis, as observed on scanning electron microscopy ( fig. 1B).

Transfusion of 14-day and 42-day Porcine Erythrocytes Causes Pulmonary Dysfunction in Swine
Pulmonary Dysfunction. To determine whether stored erythrocytes caused pulmonary dysfunction, adult pigs received an allogeneic transfusion with 14-or 42-day-old erythrocytes and lung function was assessed in vivo at 1.5 and 24 h posttransfusion. Despite a sustained rise in hematocrit, there was no difference between the groups in PaO 2 / FIO 2 ratio after erythrocyte transfusion ( fig. 2, A and B). There was no difference in other measures of pulmonary function including tidal volumes, PaCO 2, or central venous pressures between groups or over time ( fig. 2, C-E). D14 and D42 erythrocyte transfusion caused significant reductions in lung compliance at both 1.5 and 24 h compared with sham ( fig. 3A). Both D14 and D42 Tx pigs demonstrated histological evidence of pulmonary injury characterized by neutrophils in the interstitium and alveolar space, hyaline membrane formation, alveolar wall thickening, proteinaceous debris in the alveolar space, and significantly elevated lung injury scores ( fig. 3, B and C). Erythrocyte transfusion increased protein concentrations in bronchoalveolar lavage fluid compared with sham, although, this was significantly less in pigs receiving D14 versus D42 transfusions ( fig. 3D).
To determine the effect of erythrocyte transfusion on pulmonary endothelial activation, lung homogenates from sham and transfused pigs underwent Western blot analysis and were probed with antibodies to markers of endothelial activation. Both D14 and D42 erythrocytes caused pulmo-  6D). CPB also caused a pulmonary inflammatory infiltrate characterized by an increase in neutrophils, macrophages, and T-lymphocytes ( fig. 6E and appendix 3A), which was associated with an increase in tumor necrosis factor-α concentrations in porcine lung tissue (appendix 3B). CPB resulted in a reduction in total platelets, as indicated by CD41 staining but caused an increase in the number of activated platelets in lung tissue ( fig. 6E). CPB also increased P-selectin and endothelin-1, and attenuated VE-cadherin expression, but had no significant effect on E-selectin expression ( fig. 6F and appendix 3C).
Transfusion of D42 erythrocytes to pigs undergoing CPB reversed hemodilutional anemia ( fig. 5A), but increased severity of pulmonary dysfunction by reducing lung compliance at 24 h, despite augmenting arterial oxygen tensions and increasing lung injury scores, bronchoalveolar lavage protein concentration, pulmonary neutrophil and macrophage infiltration, and pulmonary endothelial activation (E-selectin expression) compared with CPB alone ( fig. 6 and  appendix 3). In contrast, erythrocyte transfusion + CPB had no effect on pulmonary T-lymphocytes, tumor necrosis factor-α concentrations, platelet activation, and P-selectin and endothelin-1 expression ( fig. 6 and appendix 3). Patel et al.

Main Findings
We describe a novel in-vivo porcine model of pulmonary dysfunction, using a protocol for the preparation of allogeneic porcine erythrocyte units identical to that used by the United Kingdom National Health Service Blood and Transplant. Our main findings are: (1) allogeneic erythrocyte transfusion causes pulmonary dysfunction in the absence of any clear priming event; (2) pulmonary dysfunction is characterized by pulmonary neutrophil infiltration and endothelial cell activation; (3) older blood is associated with increased macrophage infiltration, T-lymphocyte depletion, and platelet activation compared with younger blood; (4) stored erythrocyte transfusion in the presence of CPB exacerbates pulmonary dysfunction characterized by marked neutrophil and macrophage infiltration, endothelial activation but not platelet activation.

Strengths and Limitations
The current porcine model has significant advantages over other in vivo models because: (1) transfusion of cross-matched allogeneic erythrocytes in swine mimics clinical transfusion as compared with models that have used either syngeneic blood or xenotransfusion 8,10,11 ; (2) stored, leukodepleted erythrocytes were used to induce pulmonary dysfunction, which again mimics clinical transfusion as compared with the use of plasma or membrane-derived proinflammatory factors,

PERIOPERATIVE MEDICINE
nonphysiological doses of proinflammatory components of the storage supernatant, or antibodies 10,11,18 ; (3) the combination of erythrocyte transfusion with a clinical priming event, CPB, models a common clinical scenario associated with pulmonary dysfunction unlike other models 10,11,18 ; (4) the model allows us to link changes in biochemical and histological parameters to clinically relevant outcomes, such as changes in lung compliance, in a model homologous to human physiology. Swine are excellent models of respiratory disease as anatomy, biochemistry, physiology, size, and genetics resemble those of humans. 19 As a result, porcine lungs have been used to study many respiratory diseases and therapeutics, including surfactant function and therapy, 20 reperfusion injury, 21 pulmonary artery hypertension, 22 and the effects of mechanical ventilation. 17 Importantly, our results are comparable with a recent clinical study in which healthy volunteers when transfused with autologous blood developed subclinical pulmonary dysfunction associated with increases in markers of pulmonary inflammation. 23 Our results must also be interpreted with appropriate consideration of the limitations of this preclinical model. Importantly, we did not detect hypoxia in this model, the principal feature of TRALI, and the changes in lung compliance,   (4)(5)(6) T-lymphocytes using the CD-3 antibody, (7)(8)(9) constitutive platelets using the CD41 antibody, and (10-12) activated platelets using the platelet-activating complex (PAC; antigpiib/iiia, α iib β 3 epitope) antibody is shown in green and 4′,6-diamidino-2-phenylindole (DAPi)-stained nuclei in blue in porcine lung tissue (×630, MAC-387, CD41, PAC; ×400, CD3) for sham, 14-day, and 42-day erythrocyte transfused pigs. (B) Quantification of inflammatory cell and platelet staining demonstrates that 42-day erythrocyte transfusion elicits a significant macrophage infiltrate, T-lymphocyte depletion, and platelet activation, which is attenuated by 14-day erythrocyte transfusion. Quantification of lung neutrophil infiltration based on H&E sections demonstrates significantly increased lung neutrophil counts in both 14-and 42-day transfused pigs. (C) Tumor necrosis factor-α concentration is increased in whole lung lysates of 42-day transfused pigs, which is determined using solid phase enzyme-linked immunosorbent assay. (D and E) Western blot analysis demonstrates that markers of endothelial activation (E-selectin, P-selectin, endothelin-1) are increased in whole lung lysates of pigs that have received erythrocyte transfusion. Erythrocyte transfusion also reduces expression of VE-Cadherin. images from the same gel have been grouped, as indicated by black dividing lines. Data represents mean (±SEM) for normally distributed data or geometric mean (±95% Ci) for nonnormally distributed data. * P < 0.05 versus sham. † P < 0.05 versus D14 Tx. ET-1 = endothelin-1; H&E = hematoxylin and eosin; kDA = kilodaltons; prot = protein; TNFα = tumor necrosis factor-α; Tx = transfusion.

PERIOPERATIVE MEDICINE
although statistically significant, are unlikely to equate to clinical TRALI. There are several possible reasons for this. First, we evaluated posttransfusion lung injury only at two time points, 1.5 and 24 h. By evaluating these relatively early time points, we may not have detected important late pathological changes, for example postcardiac surgery TRALI typically manifests later in the clinical setting, often 48-72 h postsurgery. 3 Second, stored erythrocytes are known to preserve gas exchange resulting in systemic normoxia, despite the increased oxygen affinity secondary to 2,3-diphosphoglycerate depletion. 24,25 However, transfusion of stored erythrocytes causes microvascular defects in the recipients, which leads to tissue hypoxia which is not evident systemically. This has been shown in experimental studies in baboons and hamsters 26,27 and is in agreement with the findings of the study by Weiskopf et al., 23 where autologous erythrocyte transfusion resulted in subtle deficits in gas exchange (defined by a change in the alveolar to arterial difference in oxygen partial pressure) but no significant change in arterial oxygen tensions. Third, it is possible that the functional effects we observed might be accentuated by preexisting disease states where they become clinically significant, as has been reported in a clinical study. 28 These limitations notwithstanding, it must also be remembered that severe respiratory distress and hypoxia are not appropriate in a recovery model for reasons of refinement and three experiments in the current study were terminated prematurely for this reason. Severe hypoxia in these animals was most evident at approximately 6-8 h postintervention and was therefore, not reflected in our results.
Another limitation is that the porcine erythrocyte storage lesion in our study had important differences to the human storage lesion. Stored porcine erythrocytes do not utilize glucose during storage unlike stored human erythrocytes for two reasons; first porcine erythrocytes, unlike other mammalian erythrocytes, contain a hexokinase III, which is responsible for 98% of the total glucose phosphorylating activity. 29 The presence of this hexokinase reduces the erythrocyte cell membranes ability to transport glucose and therefore, porcine erythrocytes have a reduced capacity to metabolize glucose. 30 Second, the starting acidic pH inhibits relevant enzyme systems. As a result glycolysis does not occur and therefore, a reduction in pH and a rise in lactate are not observed. Erythrocyte adenosine triphosphate concentrations rapidly decline, as they are utilized as the only energy source, but not replenished via glycolysis. Despite this limitation, we have established a model of allogenic erythrocyte transfusion, although with an advanced storage lesion, and demonstrated that transfusion of these cells results in a sustained increase in hematocrit for up to 24 h.

Erythrocyte Transfusion and Pulmonary Dysfunction
Finally, although the sham group presented with the same fluid load as the experimental groups, 500 ml of crystalloid would not have the same effect on intravascular volume. Although difference in the 14-and 42-day-old blood groups argue against a pure volume effect, this does represent a limitation in the study.

Translational Relevance
Our observations are consistent with previous studies showing that neutrophils and endothelial activation play an important role in the development of TRALI. 31 Our findings are consistent with a two-event hypothesis of TRALI, Erythrocyte transfusion in the presence of CPB prevented T-lymphocyte infiltration. CPB + erythrocyte transfusion augmented pulmonary platelet infiltration but did not affect platelet activation. (F) Erythrocyte transfusion in the presence of CPB augmented E-selectin expression, but did not alter P-selectin and VE-cadherin expression compared with CPB alone. Erythrocyte transfusion also prevented CPB-induced endothelin-1 (ET-1) expression. Data represents mean (±SEM) for normally distributed data or geometric mean (±95% Ci) for nonnormally distributed data. * P < 0.05 versus sham. † P < 0.05 versus CPB. BAL = bronchoalveolar lavage; FiO 2 = fraction of inspired oxygen; PAC = platelet-activating complex; PaO 2 = arterial oxygen tension.

PERIOPERATIVE MEDICINE
in which CPB and transfusion interact to cause more severe TRALI than either risk factor in isolation. However, for the first time, we demonstrate that erythrocyte transfusion in the absence of any identified priming event can also cause pulmonary dysfunction in an in-vivo large animal model using an allogeneic cellular blood component. This finding is unlikely to have been confounded by: (1) a preexisting inflammatory state, as leukocyte counts, C-reactive protein levels, PaO 2 /FIO 2 ratios, and core temperatures were within normal limits for swine and similar between groups at baseline; (2) circulatory overload, as central venous pressures and total volumes of intravenous fluids infused were comparable between groups; (3) immune cross-reactivity, as all blood was cross-matched. One consideration is that anesthesia and ventilation served as a priming event, however, we do not believe this to have confounded our results because ventilation was standardized across the groups and did not elicit any evidence of pulmonary injury or inflammation by the measures used.
Our study also suggests an important role for macrophage infiltration and T-lymphocyte depletion in the pathogenesis of erythrocyte transfusion-mediated pulmonary dysfunction. These findings are supported by recent observations in rodents. [32][33][34] Our results also suggest that postcardiac surgery pulmonary dysfunction associated with erythrocyte transfusion may occur despite significant attenuation of platelet activation, in this case as a result of CPB. This is in contrast to recent studies in mice, which demonstrates that depletion of platelets or antiplatelet therapy prevents antibody-mediated TRALI. 18 The pathogenesis of CPB-induced acute lung injury in this model differs from lipopolysaccharide-induced acute lung injury and highlights the importance of developing animal models with more clinically relevant insults to aide data interpretation.
In conclusion, we have developed a novel in-vivo porcine model of pulmonary dysfunction, using allogeneic porcine erythrocytes that has homology to cardiac surgical patients. These findings are consistent with observational studies in cardiac surgical patients showing strong associations between erythrocyte transfusion and organ injury. 3,4,6 Transfusionmediated pulmonary dysfunction in this model occurs both in the absence, and the presence of CPB. This model represents an ideal platform for the evaluation of therapies that prevent pulmonary dysfunction before translation into the clinical setting. Patel  Data expressed as mean (SD). * Nonnormally distributed data expressed as geometric mean (± 95% Ci). FiO 2 = fraction of inspired oxygen; iV = intravenous; PaO 2 = arterial oxygen tension; Tx = transfusion.