|Home | About | Journals | Submit | Contact Us | Français|
Extracorporeal membrane oxygenation (ECMO) is a life-saving support system used in neonates and young children with severe cardiorespiratory failure. Although ECMO has reduced mortality in these critically-ill patients, almost all patients treated with ECMO develop a systemic inflammatory response syndrome (SIRS) characterized by a ‘cytokine storm’, leukocyte activation, and multisystem organ dysfunction. We used a neonatal porcine model of ECMO to investigate whether rising plasma concentrations of inflammatory cytokines during ECMO reflect de novo synthesis of these mediators in inflamed tissues, and therefore, can be used to assess the severity of ECMO-related SIRS.
Three-week-old previously-healthy piglets were subjected to venoarterial ECMO for up to 8 hours. SIRS was assessed by histopathological analysis, measurement of neutrophil activation (flow cytometry), plasma cytokine concentrations (enzyme immunoassays), and tissue expression of inflammatory genes (polymerase chain reaction/western blots). Mast cell degranulation was investigated by measurement of plasma tryptase activity.
Porcine neonatal ECMO was associated with systemic inflammatory changes similar to those seen in human neonates. TNF-α and interleukin-8 (IL-8) concentrations rose rapidly during the first 2 hours of ECMO, faster than the tissue expression of these cytokines. ECMO was associated with increased plasma mast cell tryptase activity, indicating that increased plasma concentrations of inflammatory cytokines during ECMO may result from mast cell degranulation and associated release of preformed cytokines stored in mast cells.
TNF-α and IL-8 concentrations rose faster in plasma than in the peripheral tissues during ECMO, indicating that rising plasma levels of these cytokines immediately following the initiation of ECMO may not reflect increasing tissue synthesis of these cytokines. Mobilization of preformed cellular stores of inflammatory cytokines such as in mucosal mast cells may play an important pathophysiological role in ECMO-related SIRS.
Extracorporeal membrane oxygenation (ECMO) is used for providing cardiopulmonary support in critically-ill neonates and young children with respiratory failure, congenital heart disease, and overwhelming sepsis. While the availability of ECMO has reduced mortality in these patients by nearly 80%, concerns remain about the near-universal occurrence of a systemic inflammatory response syndrome (SIRS) during ECMO that is associated with multi-system organ dysfunction and considerable morbidity. This inflammatory response usually manifests within the first few hours of ECMO with hypotension, decreased urine output, decreased lung compliance, anasarca, and liver dysfunction, changes that frequently persist for several days and delay recovery from the underlying disease that led to ECMO in the first place (1–3). Elucidation of the mechanisms of ECMO-related SIRS is a critical step in the development of effective anti-inflammatory strategies, which may prevent adverse effects and thereby reduce the average duration of ECMO, encourage its institution earlier in the clinical course of cardiorespiratory failure rather than as a ‘rescue’ therapy of last resort, and possibly allow its application in a wider range of indications (1, 3).
ECMO-related inflammatory complications are more frequent in younger patients. Neonates and young infants are at greater risk because the enormous disparity between the patient’s circulating blood volume and that of the circuit (circuit volumes are often 200–300% greater than patient’s circulating blood volume) increases the activation of various blood components upon exposure to the foreign surface of the ECMO circuit (4, 5). The high incidence of ECMO-related complications in neonates is not altered by the configuration (veno-arterial vs. veno-venous) of ECMO or by the underlying cause of cardiorespiratory failure (6); developmental factors such as high metabolic demands, reactivity of the pulmonary vasculature, and immaturity of vascular autoregulation are believed to be more important (5). Our understanding of the inflammatory effects of extracorporeal circulation is based primarily on studies on adult patients treated with cardiopulmonary bypass (CPB) during cardiac surgery (7, 8). However, the extrapolation of data from CPB to ECMO is difficult because of major pathophysiological differences between the two modalities, such as the presence of defined ischemia-reperfusion sequences in CPB (related to the placement and release of aortic cross-clamps during cardiac surgery) but not in ECMO (9, 10).
In this study, we used a porcine neonatal model of ECMO to investigate the hypothesis that rising plasma concentrations of inflammatory cytokines during ECMO reflect de novo synthesis of these mediators in inflamed tissues, and therefore, can be used to assess the severity of ECMO-related SIRS. We subjected previously-healthy 3–6 kg piglets to venoarterial ECMO in the laboratory and detected inflammatory changes that were similar to those seen in human neonates. However, plasma concentrations of tumor necrosis factor-alpha (TNF-α) and interleukin-8 [IL-8, CXC-ligand 8 (CXCL8)] increased within the first two hours of ECMO, faster than the increase in tissue expression of these cytokines. These findings indicated that the increased plasma concentrations of inflammatory cytokines could not be completely explained on the basis of increased de novo synthesis in inflamed tissues and may have resulted, at least partially, from redistribution of preformed stores of these cytokines. Plasma TNF-α and IL-8 levels correlated with plasma mast cell tryptase activity, indicating that mast cell degranulation may contribute to the rapid rise of plasma cytokine concentrations during ECMO.
Mixed-breed neonatal piglets of either gender weighing 3–6 kg were subjected to venoarterial ECMO after approval by the UAB Institutional Animal Care and Use Committee. Sham animals received anesthesia, assisted ventilation, vascular cannulation, and heparinization similar to ECMO animals, but were not connected to the circulatory pump device. Data in this study represent 8 animals each in sham and ECMO groups; 3 animals in each group were euthanized after 2 hours of treatment and the remaining 5 after 8 hours.
The piglets were anesthetized with intramuscular atropine (0.04 mg/kg), tiletamine HCl/zolazepam HCl (4.4 mg/kg), and xylazine (4.4 mg/kg). All animals were intubated with a size 3.5 cuffed endotracheal tube and mechanically ventilated (volume controlled, tidal volume = 15 mL/kg, Hallowell EMC 2000 veterinary anesthesia ventilator) at a rate of 10–15 cycles/minute with the pig in a dorsally-recumbent position. Anesthesia was maintained by inhalation of isoflurane (1.5 to 2.5%) administered in 100% oxygen. Surface lead II EKG recordings, arterial blood pressure (ABP), and core (rectal) temperature was monitored. Water-heated pads were used to maintain normothermia (37 ± 1°C). Serum electrolyte levels and arterial blood gases were monitored throughout the study. Animals were given intravenous 0.9% saline at 2–5 mL/kg/hour during the experiment and also provided boluses as required to maintain normal ABP.
A multiple side-hole 8–10F Biomedicus venous cannula (Medtronic Perfusion Systems, Minneapolis, MN) was inserted into the external jugular vein and advanced into the right atrium. The common carotid artery was cannulated using a 6–8F Biomedicus arterial cannula. Placement of these cannulae was confirmed by fluoroscopy.
After circuit priming and before starting ECMO, animals were bolused with porcine heparin (Sigma, 150 U/kg), followed by additional intravenous heparin injections to maintain activated clotting times (ACT) of 180–220 seconds after initiation of ECMO.
ECMO circuit was primed with 1 liter 0.9% saline followed by 300 mL porcine whole blood collected from a similarly-anesthetized pig, 10 mL sodium bicarbonate, 1500 units of porcine heparin, and 50 mg calcium chloride. The venoarterial ECMO system consisted of a Biomedicus BP-50 centrifugal pump (Medtronic, Shoreview, MN), and a Minimax hollow fiber oxygenator (Medtronic) with an integral heat exchanger. Water bath temperatures were maintained at 37°C using a Sarns Dual Heater/Cooler (Soma Technology, Cheshire, CT). Gas flow rates to the membrane oxygenator were maintained at 0.5 L/min of 100% oxygen. After starting ECMO, the flow rates in the circuit were advanced to 350 mL/min or 1.5 L/min/m2. Additional volume requirements to maintain circuit flow or desired hemoglobin level were met by the addition of a porcine whole blood or saline boluses. Animals were euthanized after 2 or 8 hours of ECMO by administration of potassium chloride.
Neutrophils were isolated using an established density-centrifugation method (11). Briefly, porcine blood was mixed with 4% gelatin (Sigma, St. Louis, MO) in 0.9% NaCl and allowed to settle at 37°C for 30 min. The supernatant was then layered over Histopaque-1083 separation media (Sigma) and centrifuged at 400 g for 25 min at room temperature. Neutrophils were isolated from the resultant pellet after hypotonic lysis of the red blood cells and suspended in Hanks’ balanced salt solution (HBSS; Sigma) (12). Cells were suspended in RPMI-1640 supplemented with 10% fetal bovine serum, and antibiotics.
Harvested tissues were fixed in formalin and embedded in paraffin. Histopathological examinations were performed on hematoxylin-eosin stained sections. Paraffin-embedded tissue sections were immunostained for c-kit (CD117, a mast cell marker), TNF-α, and IL-8/CXCL8 as described previously (12, 13). Briefly, tissue sections were deparaffinized with xylene and graded ethanols and antigen retrieval was achieved by heating the sections in 10 mM sodium citrate buffer, pH 6.0 at 95°C × 10 min followed by treatment with proteinase K (20 μg/mL; Sigma, St. Louis, MO) at 37°C × 20 minutes. After blocking overnight (Superblock, Pierce, Rockford, IL), the sections were stained overnight (4°C) with primary antibodies (mouse antibodies against c-kit and goat antibodies against TNF-α, and IL-8 from Santa Cruz Biotechnology, Santa Cruz, CA). Secondary staining was performed at room temperature for 30 min with Alexa 488- and/or Alexa 568-conjugated donkey anti-goat IgG and rabbit anti-mouse IgG (Invitrogen, San Diego, CA). Controls included slides with no primary antibody and appropriate isotype control. Cell nuclei were stained with 4′,6-diamidino-2-phenylindole (DAPI; Calbiochem, San Diego, CA), diluted 1:1000 in PBS, applied for 3 min. Imaging was performed using a fluorescence microscope (Carl Zeiss Microimaging, Inc., Thornwood, NY).
Neutrophils were stained for neutrophil activation markers CD18, CD35, CD62L, and CD11b using our previously-reported protocol (14). Controls included cells with no antibody and appropriate isotype control.
Real-time PCR was performed using our previously-described protocol (14). Total RNA was isolated using the acid guanidinium thiocyanate-phenol-chloroform extraction protocol (Trizol reagent, Invitrogen). First-strand cDNA was synthesized using oligo-dT primers and Moloney murine leukemia virus reverse transcriptase (Invitrogen). Real-time PCR primers were designed using the Beacon Design software (Bio-Rad, Hercules, CA). Two-step real-time PCR was performed using a SYBR Green protocol described elsewhere (14). Data were normalized against glyceraldehyde-3-phosphate dehydrogenase and gene expression was compared between samples by using the 2−ΔΔCT method.
Plasma concentrations of porcine cytokines TNF-α, IL-8/CXCL8, IL-6, and IL-1β were measured by ELISA (R&D systems, Minneapolis, MA). ELISA kits for porcine CRP and C5a were purchased from Alpco Diagnostics, Salem, NH and Immunobiological Laboratories, Minneapolis, MN, respectively.
Western blots were performed for TNF-α using our previously described protocol (14) and were analyzed using the NIH scion software.
We used a commercially-available assay (Millipore, Bedford, MA) based on the spectophotometric detection (405 nm) of a chromophore p-nitroaniline (pNA) released by tryptase action on a labeled substrate tosyl-gly-pro-lys-pNA.
Autopsy samples from full-term human neonates who died during treatment with ECMO for pulmonary hypertension (n = 3) were obtained with approval of the local Institutional Review Board.
Parametric and non-parametric tests were applied using the SigmaStat 3.1.1 software (Systat, Point Richmond, CA). The number of samples and the statistical analyses are described in figure legends. A p value of <0.05 was considered significant.
The ECMO animals developed tachycardia and hypotension within 1–2 hours of initiation of ECMO. We treated hypotension with intravenous boluses of normal saline or porcine whole blood transfusions (given if hematocrit was <20%). In the ECMO group, capillary leakiness became clinically evident by about 4 hours of ECMO with abdominal distension due to ascites. Animals subjected to ECMO needed more frequent fluid boluses and blood transfusions (total administered volume 18 ± 9 mL/kg in sham animals vs. 178 ± 16 mL/kg in animals treated with ECMO, p < 0.05). The vital parameters of animals in sham and ECMO groups did not differ from each other.
Leukocyte infiltration, focal hemorrhages, and/or edema were evident in most tissues examined including the lung, intestine, liver, skin, and kidney after 2 hours of ECMO. These inflammatory changes increased in severity after 8 hours. These changes were most prominent in the lung and the intestine (Fig. 1A). These histopathological changes of inflammation in our porcine model were generally similar to changes noted in autopsy tissue samples from human neonates who died during ECMO (Fig. 1B). While the contribution of the underlying illness (that led to cardiopulmonary failure) to the inflammatory changes in these neonates cannot be clearly ascertained, these findings provide indirect evidence for our observations in the porcine model.
Because neutrophil activation is a key event during human ECMO (15, 16), we measured the expression of CD18, CD35, CD62L, and CD11b on circulating neutrophils in hourly samples from sham and ECMO animals by flow cytometry. Compared to sham animals, the expression of all four activation markers was significantly upregulated in ECMO animals. The expression of these markers at 2 hours (time point chosen to compare with histopathological changes of inflammation) is depicted in Fig. 1C.
Using a real-time PCR microarray, we next measured mRNA expression of 48 well-characterized porcine inflammatory genes in lung and intestinal tissue. As shown in Fig. 2A, we detected increased expression of neutrophil chemokines IL-8/CXCL8 and growth-related oncoprotein-β/CXCL2 and mononuclear cell chemokines interferon-inducible protein-10/CXCL10, CXCL12, macrophage inflammatory protein-1β/CC-ligand 4, CCL11, and CCL21. Inflammatory cytokines such as TNF-α, interferon-γ, IL-1β, IL-6, IL-12, IL-15, and IL-18 were also upregulated. We also found increased tissue expression of the cognate receptors for TNF-α (TNF-receptor II), IL8 and GRO-β/CXCL2 (CXC receptors CXCR1 and CXCR2), CXCL10 (CXCR3), and MIP-1β/CCL4 (CCR3). The expression of nuclear factor-κB and activator protein 1 (AP-1) transcription factors was also increased. These changes showed that ECMO activated a wide range of pro-inflammatory genes encoding for important chemokines, cytokines, cytokine/chemokine receptors, and critical transcriptional factors that regulate the expression of these cytokines and chemokines.
In support of data obtained from the PCR array shown in Fig. 4A, we compared the plasma concentrations of TNF-α, IL-8, IL-6, and IL-1β in sham and ECMO animals. We detected a rapid rise in plasma levels of TNF-α and IL-8 within the first two hours of ECMO, which was not observed in sham-treated animals (Fig. 2B). Plasma IL-1β and IL-6 levels increased more slowly, rising significantly after 4–6 hours of ECMO.
We also investigated whether the onset of the systemic inflammatory response after initiation of ECMO was associated with changes in blood leukocyte counts or plasma CRP concentrations, two acute phase reactants that are frequently used in the clinical setting. Total leukocyte counts were 13.1 ± 2 × 109/L and 12.4 ± 3.2 × 109/L at 4 hours of treatment in sham vs. ECMO animals, respectively (differences not statistically significant). The absolute neutrophil counts were also not affected by ECMO (8.9 ± 1.2 × 109/L vs. 7.85 ± 2.1 × 109/L in sham and ECMO animals, respectively). Similarly, plasma CRP concentrations were also not altered by ECMO-related SIRS (Fig. 3). Other acute phase reactants such as erythrocyte sedimentation rate and fibrinogen levels were considered but not evaluated during ECMO because of the presence of multiple confounding factors such as transfusions and heparinization.
To investigate whether increased plasma concentrations of TNF-α and IL-8 during ECMO reflect increased de novo synthesis of these cytokines in inflamed tissues, we next measured mRNA and protein expression for TNF-α and IL-8 in tissues harvested from sham and ECMO animals after 2 hours of treatment. Although the mRNA expression of TNF-α increased in various tissues at 2 hours of ECMO (Fig. 4A), we did not detect important differences in TNF-α protein that could explain the rise in plasma TNF-α concentrations during ECMO at this time point. In contrast to the 8-fold increase in plasma TNF-α concentration at 2 hours of ECMO, TNF-α expression was decreased in the intestine and increased 1.5–2-fold in tissue samples from the liver and spleen (Fig. 4B). Similarly, we measured a 0.9–5-fold change in IL-8 mRNA in the tissues but the amount of IL-8 protein did not increase significantly in the tissues after 2 hours of ECMO; the expression of IL-8 was also decreased in the intestine at the 2 hour time-point (Supplemental Fig. 1).
Since the rapid rise in plasma TNF-α and IL-8 levels during ECMO could not be explained on the basis of increased de novo synthesis of these cytokines, we next investigated whether ECMO mobilized preformed tissue stores of TNF-α and IL-8 into the plasma. As seen in the western blots in Fig. 4B, the intestine contained the largest amounts of TNF-α in control animals. These data were consistent with evidence that mast cells present in mucosal organs such as the intestine comprise the largest store of pre-formed TNF-α in the body (17, 18) and play an important role in experimental models of SIRS (19). In Fig. 4B, the reduction in TNF-α in the ECMO intestine was also consistent with the hypothesized release of cellular stores during ECMO. Therefore, we first used immunofluorescence microscopy to confirm whether porcine mucosal mast cells contained TNF-α. As seen in Fig. 5A, strong immunoreactivity for TNF-α was detected in c-kit+ mast cells in the intestine in both sham and ECMO animals. We also detected TNF-α in mast cells in sham/ECMO porcine lung, although the lung contained fewer mast cells than the intestine (not depicted). Mast cells also showed strong immunoreactivity for IL-8 in both sham and ECMO animals (not depicted). To confirm the pathophysiological relevance of mast cells in ECMO-related SIRS in human neonates, we also stained intestinal tissues from human neonates treated with ECMO (autopsy samples); a similar co-localization of c-kit and TNF-α was detected (Fig. 5B).
To investigate whether the initiation of ECMO in our porcine model was associated with mast cell degranulation, we next compared plasma samples from sham and ECMO animals for tryptase (mast cell protease 7) activity, a serine protease present almost exclusively and in large amounts in mast cell granules (20, 21). As shown in Fig. 5C, plasma tryptase activity after 1 hour of ECMO was nearly 4-fold higher than in sham animals and correlated with plasma concentrations of TNF-α and IL-8/CXCL8 (p < 0.05). Mast cell degranulation is a plausible event during ECMO because activation of the complement pathway in the membrane oxygenator releases several mast cell secretagogues such as C5a, C3a, and C3adesArg (22). Consistent with this hypothesis, we detected increased plasma C5a levels in our porcine model after 1 hour of ECMO (inset, Fig. 5C).
The histopathological changes of inflammation, activation of neutrophils, and increased plasma concentrations of various pro-inflammatory cytokines in our piglet model were generally similar to those reported in human ECMO (16, 23–25). Existing clinical and experimental evidence on the mechanisms of ECMO-related SIRS emphasizes the activation of neutrophils and other leukocytes as a pivotal event during ECMO, which occurs due to contact with the foreign surface of the circuit, shear stress, expression of inflammatory cytokines, activation of complement, coagulation and fibrinolytic pathways, and increased concentrations of bioactive lipids (16, 22–28). These activated neutrophils adhere to the capillary/venular endothelium and undergo degranulation to produce cytokines, arachidonic acid metabolites, and reactive oxygen species, causing widespread microvascular injury (15, 27) and multi-organ dysfunction (16, 22–28).
The rapid rise in plasma concentrations of TNF-α and IL-8 almost immediately after the initiation of ECMO suggests that these cytokines may be important mediators in the development of ECMO-related inflammation. TNF-α can induce the expression of a wide range of inflammatory genes, including its own (29), which suggests that the early rise in circulating TNF-α levels during ECMO may form a ‘feed-forward’ loop to augment the inflammatory responses. In human neonates, higher plasma TNF-α concentrations have been noted in patients who died during ECMO than in ‘responders’ who recovered completely (16, 26, 30). In other studies, plasma concentrations of IL-8 correlated with the severity of SIRS and cardiopulmonary impairment during ECMO (16, 23, 24, 31).
The marked elevation in plasma cytokine levels in our model contrasted with the low sensitivity of acute phase reactants such as leukocyte counts and plasma CRP concentrations. Elevated CRP levels have been noted in one previous report on human ECMO (31) where CRP concentrations were higher in non-responders as compared to patients who improved on ECMO, although CRP was found to be less sensitive as a prognostic indicator than IL-6. The differential expression of IL-6 and CRP in our model was surprising because IL-6 is a potent inducer of CRP production in the liver and the two mediators are usually identified as bound variables in various inflammatory states (32). A possible explanation may be related to the 8-hour duration of our experiments, where we might have missed a late rise in plasma CRP concentrations (following the delayed rise in IL-6 levels). Alternatively, the CRP response may have been dampened by the relative resistance of porcine hepatocytes to IL-6, a species-specific feature. IL-6 induces CRP expression in hepatocytes via a CCAAT-enhancer-binding protein-beta (C/EBP-β)-mediated pathway (33), which carries unique posttranslational modifications and/or compartmentalization of C/EBP-β in pigs (34).
The increase in plasma TNF-α and IL-8 concentrations within 2 hours of initiation of ECMO in our model was not associated with a proportionate increase in the tissue synthesis of these cytokines at the same time. Considering the rapid change in plasma cytokine levels and also because the expression of TNF-α protein was actually decreased in the intestine at 2 hours, we hypothesized that ECMO mobilized these cytokines from preformed tissue stores of TNF-α into plasma. Of all the major cellular sources of TNF-α, mast cells (35) and eosinophils (36) are the only cell types considered capable of storing preformed TNF-α. We focused on mast cells and not eosinophils in this study because the gut mucosa normally contains a large reservoir of mast cells (17, 18). The plausibility of mast cell degranulation, as shown in our model, is supported by existing data on activation of complement and consequent release of mast cell secretagogues such as C5a, C3a, and C3adesArg during human ECMO (37). C3a levels have also been shown to correlate with plasma TNF-α levels and the severity of inflammation during ECMO (30).
Mast cells are now increasingly recognized as a major source of pre-formed TNF-α in SIRS in diverse settings such as in Gram-negative and Gram-positive bacterial sepsis, peritonitis, severe alveolar hypoxia, and portal hypertension (38–42). Unlike other major cellular sources of TNF-α such as macrophages and T-/B-lymphocytes that contain little or no detectable TNF-α unless appropriately stimulated (43, 44), mast cells present in mucosal organs contain large amounts of pre-formed TNF-α (17, 18) and play an important role in experimental models of SIRS (19). Histamine liberation has been previously reported during cardiopulmonary bypass (CPB) in older children and adults, but circulating basophils, and not mast cells, were incriminated as the source (45, 46). As we mentioned earlier, mechanistic differences in the inflammatory responses associated with ECMO vs. CPB are not unexpected and illustrate the need for specifically-directed studies for these two forms of ECC. Even though ECMO and CPB are both performed using similar circulatory pumps and other major equipment, CPB is associated with several additional pathophysiological variables such as ischemia-reperfusion injury from aortic cross-clamping and due to the vasoconstrictive effects of hypothermia and hemodilution, pro-inflammatory effects of surgical trauma and tissue débris, and more frequent use of protamine (7, 8, 47–49).
In conclusion, the institution of ECMO in our porcine model induced systemic inflammatory changes that were remarkably similar to those reported during human ECMO. Although our original hypothesis that increased plasma TNF-α/IL-8 levels reflect tissue synthesis of these cytokines during ECMO was disproven, our data reaffirmed the higher sensitivity of inflammatory cytokines as indicators of inflammation during ECMO as compared to blood leukocyte counts and C-reactive protein. Further studies are needed to develop a panel of cytokines that are synthesized de novo during ECMO-related SIRS and not released from preformed stores. We have also identified mast cell degranulation and the release of TNF-α as early events in ECMO-related SIRS. The availability of agents that can stabilize enteric mast cells (50) as well as effective neutralizing anti-TNF-α antibodies raises important therapeutic possibilities (51) and indicates a need to aggressively investigate these targets in preclinical and clinical settings.
Sources of Funding: Supported in part by the American Heart Association grant 0665155B, American Gastroenterological Association 2006 Research Scholar Award, and NIH awards K12HD043397, RHD059142 to A.M. The work was made possible in part by Research Facilities Improvement Grant C06RR15490 from the National Center for Research Resources.
Disclosures: No conflicts of interest to disclose.