Plasma proteomic changes during hypothermic and normothermic cardiopulmonary bypass in aortic surgeries

Deep hypothermic circulatory arrest (DHCA) is a protective method against brain ischemia in aortic surgery. However, the possible effects of DHCA on the plasma proteins remain to be determined. In the present study, we used novel high-throughput technology to compare the plasma proteomes during DHCA (22°C) with selective cerebral perfusion (SCP, n=7) to those during normothermic cardiopulmonary bypass (CPB, n=7). Three plasma samples per patient were obtained during CPB: T1, prior to cooling; T2, during hypothermia; T3, after rewarming for the DHCA group and three corresponding points for the normothermic group. A proteomic analysis was performed using isobaric tag for relative and absolute quantification (iTRAQ) labeling tandem mass spectrometry to assess quantitative protein changes. In total, the analysis identified 262 proteins. The bioinformatics analysis revealed a significant upregulation of complement activation at T2 in normothermic CPB, which was suppressed in DHCA. These findings were confirmed by the changes of the terminal complement complex (SC5b-9) levels. At T3, however, the level of SC5b-9 showed a greater increase in DHCA compared to normothermic CPB, while 48 proteins were significantly downregulated in DHCA. The results demonstrated that DHCA and rewarming potentially exert a significant effect on the plasma proteome in patients undergoing aortic surgery.


Introduction
Hypothermia <32-34˚C has been clinically utilized only in cardiovascular surgery with cardiopulmonary bypass (CPB) as cardiac function becomes unstable with the decrease in temperature. Profound (<20˚C) to deep (<25˚C) hypothermic circulatory arrest (HCA) combined with ante grade selective cerebral perfusion (SCP) is a protective strategy against brain ischemia in complex aortic surgery (1). Although there is relatively little evidence of adverse effects of profound/deep hypothermia, there is some concern that HCA may have some drawbacks, such as coagulation disorders (2,3). As a consequence, moderate HCA (25-28˚C) combined with increased SCP flow rates for aortic surgery has been utilized as the mainstay of treatment (4).
The development of isobaric tag-based quantitative mass spectrometry has provided a reliable method for high-throughput analysis of proteomes (5). Proteomic changes in human plasma samples have been successfully analyzed through the use of the isobaric tag for relative and absolute quantification (iTRAQ) labeling strategy (6,7). In the present study, a comparison was made of the plasma proteomes of patients undergoing aortic surgery using deep hypothermic circulatory arrest (DHCA) and SCP (DHCA + SCP) to those undergoing aortic surgery using normothermic CPB.
Tokyo, Japan) and polymer-coating cardiotomy reservoir (only used in hypothermic CPB; BO-HC2821; Maquet), centrifugal pump (ROTAFLOW, BO-RF32; Maquet), arterial filter (40 µm Pall arterial line filter, AL6; Pall, Port Washington, NY, USA) and heparin-coated CPB circuit (Mera or JMS, Tokyo, Japan). The pump was primed with bicarbonated Ringer's solution, mannitol and methylprednisolone (500 mg). After systemic anticoagulation with 3 mg/kg heparin, CPB was instituted at a flow rate of 2.0-2.8 l/min/M 2 . Blood gases, electrolytes, glucose and the hemoglobin (Hb) concentration were continuously monitored by CDI 500 (Terumo, Tokyo, Japan) and blood gas analysis was performed every 30 min (ABL800 FLEX Blood Gas Analyzer; Radiometer, Copenhagen, Denmark). Each parameter was maintained as follows: PaO 2 : 300-500 mmHg, PaCO 2 : 35-45 mmHg (α-stat management), blood glucose: 100-300 mg/dl, Hb: 7-9 g/dl in the two groups. Rectal and urinary bladder temperatures were monitored. Dilutional ultrafiltration was continuously performed during CPB for the patients in the two groups to correct hyperkalemia and fluid imbalance (Blood concentrator PLUS, BC-140 plus; Maquet). In DHCA + SCP, the body was cooled to 22˚C with topical head cooling. After cardioplegic cardiac arrest and initiation of HCA, SCP was performed following cannulation into the three arch vessels and maintained by monitoring pressures (both bilateral radial artery and SCP cannula tip) and using an INVOS Oximeter (INVOS 5100C; Somanetics, Troy, MI, USA). Rewarming was initiated following completion of open distal anastomosis and initiation of distal perfusion.
Blood sampling. Three blood samples were collected from an arterial line into ethylenediaminetetraacetic acid (EDTA) tubes as follows: T1 (before cooling, 5 min after pump initiation), T2 (before aortic declamping, 30-160 min after pump initiation in normothermic CPB or during deep hypothermia, 30-124 min after pump initiation in DHCA + SCP), T3 (just before termination of CPB, 90-193 min after pump initiation in normothermic CPB or 121-314 min after pump initiation in DHCA + SCP). Blood samples were centrifuged at 1,400 x g for 5 min, and the plasma layers were stored at -80˚C.
Immunodepletion of abundant proteins. The two most abundant plasma proteins, albumin and immunoglobulin (Ig) G, were removed using an immunodepletion column (Albumin & IgG Depletion SpinTrap; GE Healthcare, Buckinghamshire, UK) according to the manufacturer's instructions and as previously described (7).
iTRAQ labeling and strong cation exchange (SCX) chromatography. Samples were prepared according to the manual published by AB Sciex (Foster city, CA, USA) and as described previously (7). In brief, equal amounts of immunodepleted T1, T2, and T3 samples from each patient were denaturated and reduced, the cysteines were alkylated, and then digested with trypsin (AB Sciex). Each digest was labeled with a different iTRAQ tag using the iTRAQ reagent 4-plex kit (AB Sciex). iTRAQ label 114 was used for the T1 sample, and iTRAQ labels 115, 116 or 117 were randomly selected for the T2 and T3 samples, after which the three samples of each patient were combined. The combined samples were then fractionated into six fractions by SCX chromatography according to the manufacturer's instructions (AB Sciex) and each fraction was desalted according to the manufacturer's instructions (Waters, Milford, MA, USA).
NanoLC and MALDI-TOF/TOF MS/MS analysis. One fraction from the SCX chromatography (see above) was fractionated to 171 spots using a DiNa nanoLC system (KYA Technologies, Tokyo, Japan) and collected onto an Opti-TOF LC/MALDI 384 target plate (AB Sciex) according to the manufacturer's instructions and as previously described (7). Spotted peptide samples were analyzed by a 5800 MALDI-TOF/TOF MS/MS Analyzer with TOF/TOF Series software (version 4.0; AB Sciex). MS/MS data were analyzed using ProteinPilot™ software (version 3.0) and the Paragon™ protein database (AB Sciex). Quantitative changes of proteins at T2 or T3 were calculated using the iTRAQ ratios T2:T1 or T3:T1, respectively.
iTRAQ data analysis and bioinformatic analysis. Proteins identified as showing expression changes were tested for conformity to the following conditions: i) a false discovery rate (FDR) <5% (FDR was estimated by 'decoy database searching' using the ProteinPilot Software); and ⅱ) protein confidence >99% ('unused ProtScore' >2). Unused ProtScore is defined as -log (1-% confidence/100). Proteins fulfilling these criteria were considered to have 'statistical significance' (7,8). PANTHER software (version 8.1, http://www.pantherdb. org) was used to test for statistical overrepresentation of Gene Ontology (GO, http://www.geneontology.org) terms as described in detail elsewhere (9,10). If the number of identified genes in a GO term was significantly larger than that in the whole genome classified by the same GO term, i.e., the number of observed genes in a GO term is significantly larger than the number of expected genes in the same GO term by the binomial test, the GO term was described here as 'overrepresented' with statistical significance after a Bonferroni correction for multiple testing. The annotations of identified proteins were obtained from the Uniprot database (http://www.uniprot.org/).
Western blot analysis and enzyme-linked immunosorbent assay (ELISA). Western blot analyses were performed as described previously (7). In brief, plasma samples were separated by sodium dodecyl sulfate-polyacrylamide gel electrophoresis (SDS-PAGE), and immunoblotted using rabbit monoclonal carbonic anhydrase 1 (CA1) antibody (Abcam, Tokyo, Japan) and anti-rabbit IR dye 680-conjugated IgG (LI-COR, Lincoln, NE, USA). Protein bands were visualized using an Odyssey (LI-COR) infrared imaging system and their intensities measured for densitometric analyses of CA1. Plasma levels of the complement proteins C5a and SC5b-9 were determined using a commercially available solid-phase sandwich ELISA kit (Quidel Corp., San Diego, CA, USA) following the manufacturer's recommendations.
Statistical analysis. Continuous variables, such as blood gas analysis and hemodynamic variables were expressed as means ± standard deviation (SD). The variables were tested for statistically significant differences between patient groups using the Student's t-test. Categorical variables were compared using Fisher's exact test. For analysis of iTRAQ ratios, i.e. T2:T1 or T3:T1 ratios for each significantly identified protein, P-values were calculated by the one sample t-test of averaged protein ratio against 1 to assess the validity of the protein expression change (11). To assess ELISA and western blot analysis data, statistical comparisons were performed by two-way analysis of variance (ANOVA) (general linear model) with repeated measures followed by a post-hoc Bonferroni test to detect individual differences. The relationships between complement activation (SC5b-9) and the iTRAQ ratios of each protein including CA1 values (western blot analysis) were calculated by single and stepwise multiple regression analyses (forward selection method) (StatFlex version 6; Artech, Osaka, Japan). P<0.05 was considered to indicate statistical significance.

Results
Patient characteristics and CPB data. Baseline characteristics were similar in both groups except for minimal core temperature and CPB time (Table Ⅰ). There were several small but significant differences in blood gas analysis, Hb and lactate levels in both groups (Table Ⅱ). Pump flow rate, perfusion pressure, and core temperatures were similar between the two groups at T1 and T3. At T2, there was a significant difference in core temperature between the two groups: 23.0±1.2˚C vs. 36.0±1.0˚C (P<0.0001). The timing of T1 and T2 sampling from the initiation of CPB did not differ between the two groups (P=1.000 for T1, P=0.250 for T2); however, sampling at T3 was performed significantly later in the DHCA + SCP patient group than in the normothermic CPB group (Table Ⅰ). The HCA time of the lower body averaged 59±41 min in DHCA + SCP patients.
iTRAQ data analysis and bioinformatics. Mass spectro metry identified 322 proteins, of which 262 were found to be statistically significant, i.e., the proteins fulfilled both the FDR <5% and protein confidence >99% criteria. Of the 262 proteins, those that increased or decreased (>1.2-fold or <0.833-fold) in at least four of the seven samples in each treatment group were identified to obtain an adequate sample size and statistical power (5,12), subsequently analyzed by PANTHER software. The statistical overrepresentation test identified six GO categories (biological processes) with five proteins significantly increased in the normothermic CPB group at T2 (Tables Ⅲ and Ⅴ). By contrast, only one GO category showed significant overrepresentation in the DHCA + SCP group at T2 (Table Ⅲ). At T3 in the latter treatment group, 11 GO categories were significantly overrepresented with 48 proteins significantly decreased, whereas only one GO category was significantly overrepresented in the normothermic CPB group (Tables Ⅳ and Ⅵ). In the normothermic CPB group, the iTRAQ ratio of CA1 increased significantly to 1.50±0.23 at T2 and to 1.58±0.30 at T3 in comparison to T1 (P=0.0241 at T2, and P=0.0299 at T3, Table Ⅴ). The ratio also increased in the DHCA + SCP group to 1.52±0.45 at T2 and 2.65±1.65 at T3, although without statistical significance due to missing data. This increase in the CA1 ratio at T3 is a noteworthy finding as all other proteins decreased significantly in the DHCA + SCP patients, with the exception of HBA1 and HBB (Table Ⅵ).
Western blot analysis and ELISA. To validate the iTRAQ ratios and complement activation identified by the PANTHER analysis, the protein CA1 was analyzed by western blot analysis and the complement proteins C5a and SC5b-9 were measured by ELISA. In both patient groups, the levels of CA1 continuously and significantly increased during CPB (P<0.0001), although the differences between the groups were not significant (Fig. 1). The level of complement C5a increased, albeit non-significantly, at T3 in the DHCA + SCP group compared to the normothermic group (Fig 2A). The level of SC5b-9 increased significantly at T2 in the normothermic group, and the differences between T1 and T2 and between T1 and T3 were both highly significant (P<0.0001, Fig 2B). In the DHCA + SCP group, the level of SC5b-9 increased slightly from T1 to T2, and then increased Figure 1. Results of the western blot analysis of carbonic anhydrase 1 (CA1) levels during cardiopulmonary bypass (CPB). CA1 levels increased significantly with time during CPB ( * P<0.0001) in the deep hypothermic circulatory arrest (DHCA) and selective cerebral perfusion (SCP) (DHCA + SCP) and normothermic CPB patients with no significant differences between the two groups. * Significant differences (P<0.0001) between sampling times in the two groups. T1, 5 min after initiation of CPB; T2, during deep hypothermia or before declamping; T3, just before termination of CPB. rapidly and significantly from T2 to T3. There were significant differences in SC5b-9 values between T3 and T1 and between T3 and T2 in the DHCA + SCP group (P<0.0001), and between both groups at T2, respectively (P= 0.008). Since the innate immunity response is regulated by complex protein-protein interactions (13), it is difficult to clarify the causative relationship between the SC5b-9 level and iTRAQ ratios of identified proteins. Therefore, single regression analyses were performed to search for the relationships between the SC5-9 values and the iTRAQ ratios of the 42 proteins identified in all samples including the CA1 value (western blot analysis). Seven proteins (CP, HPX, VTN, F2, A1BG, GC, and CA1) showed significant correlations with SC5b-9 levels. A stepwise multiple regression analysis identified a significant coefficient (R= 0.486) and ceruloplasmin as an independent variable (standardized β=-0.468, P=0.012).

Discussion
Exposure of blood components to a CPB circuit activates various blood cell types, endothelial cells and proteins. This activation can improperly stimulate biological processes that may Table Ⅲ . GO categories showing significant overrepresentation for plasma proteins at T2. contribute to the development of postoperative complications resulting from the dysfunction of blood cells, coagulation, fibrinolysis, and the kinin-kallikrein and immune systems (14)(15)(16). Various modifications to the inner surface of the CPB circuit, including heparin-and polymer-coating, reduce adverse effects on complement activation, blood coagulation, proinflammatory cytokines and improve clinical outcomes (17)(18)(19)(20). The C5 complement inhibitor pexelizumab and a complement factor 1 esterase inhibitor have been tested clinically (21) and in an animal study (22) for their abilities to reduce complement activation during CPB. Although a recent proteomic study demonstrated the possibility of identifying predictive plasma biomarkers in patients undergoing CPB (23), this approach has not been used to investigate the effect of hypothermia/rewarming on the biocompatibility of a CPB circuit.

Homo sapiens
The proteins that were identified as showing significant protein confidence in this experiment were all included in the list of proteins reported previously for the human plasma proteome (24). Our analysis using iTRAQ labeling mass spectrometry demonstrated that several GO categories related to the complement system and proteolysis were significantly overrepresented at T2 in normothermic CPB in spite of using a heparin-coated CPB circuit and polymer-coated oxygenator. Of note, these changes were not observed at T2 in DHCA + SCP patients. To confirm these findings, the levels of plasma C5a and SC5b-9 were measured by ELISA, which showed a significantly larger increase in SC5b-9 at T2 in patients undergoing normothermic CPB compared to DHCA + SCP. This reduction in the complement activation level in the DHCA + SCP group is similar to the findings from investigations of a randomized  The test for statistical overrepresentation of each Gene Ontology (GO) term (biological process) was perfomed using PANTHER software (version 8.1). Upregulated genes are shown in bold. a The total number of genes in the whole genomes (Homo sapiens, n=20,000) classified by the GO term; b the number of genes from the input list [n=11 (mapped genes, 11) in the normothermic cardiopulmonary bypass (CPB) group, n=57 (mapped genes, 52; unmapped gene, 5) in the deep hypothermic circulatory arrest (DHCA) with selective cerebral perfusion (SCP) group (DHCA + SCP)] classified by the GO term; c the number of genes that would be detected in the input list for a particular GO category on the basis of the reference list (Homo sapiens). T3, just before termination of CPB.

At T2 At T3 ---------------------------------------------------------------
by chelatable iron via Fenton chemistry and, thereby, to accelerate complement activation (30). On the other hand, both reperfusion injury and the impact of cooling/rewarming on liver metabolism may underlie the downregulation of some of the identified proteins at T3. A number of proteins were found to be downregulated at T3 in the DHCA + SCP patients (Tables Ⅳ and Ⅵ). This downregulation seems inconsistent with the complement activation identified by the ELISA analysis. Possibly, this inconsistency might be explained by the tight control of complement activation by its regulators, suggesting that the downregulation of regulators such as vitronectin (VTN) or C1-inhibitor (SERPING1) contribute to complement activation (Table Ⅵ) (29). CA1 had a significant positive correlation with SC5b-9 levels by a single regression analysis. It has been demonstrated that CA1 causes kallikrein-kinin system activation (31). There is evidence of an interaction among the complement, coagulation and kinin systems (29,32) in which kallikrein is an initiator of the complement cascade by digestion of factor XIIa into factor XIIf (33) and possibly through the cleavage of C3 or C5 into active forms (32). By contrast, ceruloplasmin, which had a significant negative correlation with the SC5b-9 level by stepwise multiple regression analysis, was downregulated at T3 in the DHCA + SCP. This protein has ferroxidase activity to oxidize Fe 2+ to Fe 3+ , contributing to the antioxidant capacity because of the reduction of oxidation capacity via the Fenton reaction (34). Ceruloplasmin is also reported to be an endogenous inhibitor of myeloperoxidase (35), but is downregulated by reactive oxygen species (34). Taken together, CA1, an upregulated protein at T3, may be a positive regulator of the complement system via kallikrein-kinin system activation. By contrast, ceruloplasmin, a downregulated protein at T3, may be a major negative regulator of the complement system via its antioxidant capacity. Thus, both the increased complement activator and several decreased complement regulators could result in the rapid rise of the SC5b-9 level at T3 in the DHCA + SCP patients. During normothermic CPB, the levels of C5a and SC5b-9 did not change significantly from T2 to T3 after an initial rise at T2, a finding consistent with the relatively unchanged iTRAQ ratio and with reports from previous studies (14,16). Although we used state-of-the-art iTRAQ labeling technology mass spectroscopy, a considerable number of proteins expressed at low levels may not have been detected because of the large dynamic range in plasma proteins (24). Nevertheless, this technology may become a useful analytical tool in the field of research of biocompatibility of CPB in the future. The comparisons here involved two groups in which the two sets of patients underwent similar thoracic aortic surgery, i.e., DHCA + SCP and normothermic CPB. However, significant differences between the two groups occurred in the CPB time, which in turn affected T3 sampling time, implementation of circulatory arrest and SCP (hypothermic ischemia of lower body), and levels of PaCO 2 and Hb at T2. These differences may affect plasma proteomic profiles. To elucidate the effects of hypothermia (and rewarming) per se on the biocompatibility of CPB, it may be necessary to perform a study using an animal model in the future.
In conclusion, results of the comparative iTRAQ proteomic analysis demonstrated that CPB-induced complement activation (and proteolysis) was initially suppressed during DHCA and later upregulated at the end of CPB after rewarming. Single and multiple regression analyses showed that this complement activation/suppression was associated with changes in expression of ceruloplasmin and six other proteins including CA1.