Relative purity of thrombin-based hemostatic agents used in surgery

SURGICAL FORUM The opening act of the drama of academic surgery often has been set at a session of the Surgical Forum. The brainchild of Owen H Wangensteen, the Surgical Forum has been a proving ground for surgical research and an environment where dialogue is established with other investigators with similar interests. Each session is a concentrated distillate of current scientific solutions as they pertain to Surgery. Focused subject matter invites suggestions, amplifications, and criticism by peers. As such, the Surgical Forum is a uniquely enriching event. As manuscripts based on Surgical Forum abstracts presentations are reviewed and accepted, they will appear in this section of JACS throughout the year. Seymour I Schwartz, MD, FACS, Editor in Chief

Relative Purity of Thrombin-Based Hemostatic Agents Used in Surgery Jonathan G Schoenecker, MD, PhD, Rachel K Johnson, BA, Ryan C Fields, MD, Aaron P Lesher, BA, Taymon Domzalski, BA, Kamran Baig, MBBS, MRCS, Jeffrey H Lawson, MD, PhD, William Parker, PhD Hemostatic agents used in surgery contain thrombin isolated from either a bovine or human source. The use of thrombin derived from a bovine source has been associated with the development of an abnormal immune response, but a study of the immunoreactivity of the various commercially available thrombin preparations has not been conducted. This study determined the relative purity of commercially available thrombin preparations, if humans have natural antibodies that recognize these preparations, and if elicited antibodies against bovine thrombin cross-react with other bovine or human hemostatic agents. STUDY DESIGN: The purity of hemostatic agents was determined by protein and substrate assays, electrophoresis, and immunoblotting. The natural antigenicity and cross-reactivity of elicited antibodies were measured by ELISA using serum samples from 82 donors from the Red Cross and serum collected from patients exposed to bovine thrombin, respectively. RESULTS: All of the bovine thrombin preparations were found to contain the xenogeneic carbohydrate galactose␣1-3galactose. The natural antigenicity of the bovine thrombin preparations was greater than that of a human thrombin preparation and similar to that of porcine aortic endothelial cells. Antibodies elicited against bovine thrombin were found to cross-react with other bovine preparations and other xenoantigens but not with human hemostatic preparations. CONCLUSIONS: All patients have antibovine thrombin antibodies, even before exposure to bovine thrombin– containing hemostatic agents. The cross-reactivity of elicited antibovine thrombin antibodies indicates that if a patient has been sensitized to a bovine product, it is likely safer to use a human-derived product in lieu of a bovine product. ( J Am Coll Surg 2003;197:580-590. © 2003 by the American College of Surgeons) BACKGROUND:

Numerous surgical hemostatic agents are available for use in the United States. Although these products are

used interchangeably, their components and mechanisms of action differ greatly. Despite these differences, nearly all hemostatic products contain or are combined with the coagulation protein thrombin. Bovine thrombin was chosen over human thrombin as a therapeutic agent for reasons that may no longer be valid. Concerns over the transmission of viruses from pooled human plasma date back more than 30 years,1 and because of these concerns, hemostatic agents de-

No competing interests declared.

This work was supported by the American Heart Association (JHL), the Department of Surgery at Duke University Medical Center, and the Fannie E Rippel Foundation. Dr Lawson is a Clinician Scientist Awardee from the American Heart Association and Genentech, Inc (95004380). Abstract presented at the American College of Surgeons 88th Annual Clinical Congress, Surgical Forum, San Francisco, CA, October 2002. Received January 30, 2003; Revised April 25, 2003; Accepted April 25, 2003. From the Departments of Surgery (Johnson, Lesher, Domzalski, Baig, Parker, Lawson) and Pathology (Schoenecker, Fields, Lawson), Duke University Medical Center, Durham, NC.

© 2003 by the American College of Surgeons Published by Elsevier Inc.

Correspondence address: William Parker, PhD, Department of Surgery, Duke University Medical Center, Box 2605, Durham NC 27710.

580

ISSN 1072-7515/03/$21.00 doi:10.1016/S1072-7515(03)00670-7

Vol. 197, No. 4, October 2003

rived from human plasma were not available for many years in the United States. Even though the possibility of disease transmission from human-derived products has been greatly reduced and human-derived hemostatic agents have recently become available, most hemostatic agents still contain bovine thrombin. Although bovine thrombin continues to be used clinically, reports have suggested that its use can result in the development of an abnormal immune response, resulting in either a hemorrhagic phenotype associated with the development of autoreactive antibodies2-5 or an immunoglobulin E (IgE)–mediated anaphylaxis.6-8 In almost all cases, patients were found to mount an immune response against bovine thrombin, with complications occurring on a second exposure. Tadokoro and colleagues,6 using a simple skin test to characterize the immune response to bovine thrombin and to identify patients at risk of bovine thrombin–induced anaphylaxis, found that at least 11% of patients exposed to bovine thrombin preparations develop antibovine thrombin IgE antibodies. Further, we recently demonstrated that exposure of non–autoimmune-prone mice to a single dose of bovine thrombin can lead to the development of autoantibodies associated with systemic autoimmunity.9 These reports provide mounting evidence indicating that exposure to bovine thrombin can stimulate an abnormal immune response. When humans are exposed to bovine thrombin, they are exposed to xenogeneic antigens. For that reason, at least in terms of exposure to antigen, exposure to bovine thrombin is similar in many regards to a xenotransplantation. The primary complication preventing the successful xenotransplantation of organs and cells is the host immune response against xenogeneic, nonself determinants.10 These determinants include discordant posttranslational modifications, such as galactose␣13galactose (Gal␣1-3Gal), and differences in the primary structure of proteins. We propose that, like other xenogeneic tissue, the cause of the immune response against bovine thrombin results, at least in part, from xenogeneic, nonself determinants. We recently demonstrated that bovine thrombin contains the xenogeneic carbohydrate Gal␣1-3Gal and that when humans are exposed to bovine thrombin they mount an immune response against this epitope.11 Despite evidence that the xenogeneic nature of bovine thrombin may contribute to the abnormal immune response associated with these products, a prospective,

Schoenecker et al

Purity of Bovine Thrombin

581

randomized clinical trial comparing bovine and human thrombin preparations has not been conducted. Currently, such a study is not feasible because of limitations in the types of FDA-approved hemostatic agents. Approved hemostatic agents containing thrombin can be divided into two categories: thrombin-based products and fibrinogen-based products. Thrombin-based products consist primarily of thrombin, and fibrinogenbased products contain mostly fibrinogen, with smaller quantities of thrombin. Importantly, the hemostatic agents derived from human plasma are fibrinogen based, and the agents derived from bovine plasma are thrombin based. A direct comparison of the antigenic properties of human versus bovine thrombin may not necessarily provide conclusive evidence about the safety of a given thrombin preparation. The immunostimulatory properties of bovine thrombin may not be limited to its xenogeneic origin. Both the proteolytic activity and exosite domains of thrombin have been shown to be immunostimulatory.12 Further, the fact that currently available human products are fibrinogen based and bovine products are thrombin based complicates any potential comparison between human and bovine thrombin preparations. Until a human-derived thrombin-based hemostatic agent becomes available and the appropriate trials are performed, a surgeon must decide which products are safe to use, but there are few studies that critically compare the contents and antigenicity of currently available hemostatic agents that contain thrombin. The principal goals of this study were to determine: 1. The relative purity of commercially available thrombin preparations 2. If the proteins in bovine thrombin preparations contain the same or similar nonself determinants as do other xenoantigens 3. To what extent humans have natural antibodies that recognize bovine thrombin preparations 4. If elicited antibodies against bovine thrombin preparations cross-react with other hemostatic agents

METHODS Materials

Thrombogen was obtained from Johnson and Johnson, Thrombin-JMI from Jones Pharma Incorporated, and Tisseel from Baxter Healthcare Corporation, Hyland Division Galactose␣1-3galactose (galactobiose; Gal␣1-

582

Schoenecker et al

Purity of Bovine Thrombin

3Gal) and human serum albumin conjugated with galactose␣1-3galactose(␤1-4)N-acetylglucosamine (HSA␣Gal) were obtained from V-Laboratories. Human serum albumin (HSA), bovine serum albumin, alkaline phosphatase–conjugated goat antihuman IgM and IgG, Tween-20, 2-mercaptoethanol, p-nitrophenyl phosphate, porcine thyroglobulin, collagenase, endothelial cell growth supplement, and smooth muscle ␣-actin were purchased from Sigma. Spectrozyme-TH was obtained from American Diagnostica Inc. Purified bovine ␣-thrombin was obtained from Hematelogic Technologies Inc. Sodium dodecyl sulfate-polyacrylamide gel electrophoresis (SDS-PAGE) 5% to 15% and DC Protein Assay kit were purchased from Biorad. Dulbecco’s modified eagle medium (DMEM), penicillinstreptomycin, porcine-derived heparin, and L-glutamine were from Gibco (Grand Island, NY). Fetal bovine serum was from Hyclone (Logan, UT). Fluorescentlabeled acetylated-LDL was obtained from Molecular Probes. 5-Bromo-4-chloro-3-indolyl phosphate/nitroblue tetrazolium was purchased from Promega. Nitrocellulose was from Scleicher & Schuell.

J Am Coll Surg

practice was requested. The decision to administer topical bovine thrombin was made by the surgeon during the operative procedure, and the topical bovine thrombin preparation used in all cases was Thrombogen. The amount of topical bovine thrombin administered during these procedures was typically between 5,000 and 20,000 U. During the procedures, the topical bovine thrombin was used alone as a topical spray or in conjunction with Gelfoam powder, Gelfoam sheets, or both (Pharmacia and Upjohn) depending on the surgeon’s preference. Out of 52 patients exposed to Thrombogen who provided samples during the initial study,2 five were selected at random for this study. Pre- and postexposure samples from these five patients were pooled separately and used as a “preexposed pool” and a “postexposed pool.” The five patients selected at random were all Caucasian men and had a mean age of 56.2 years (range 46 to 66 years) at the time of operation. Postoperative samples from these five patients were obtained between 7 and 12 weeks (mean 8.2 weeks) after surgical exposure to Thrombogen. Preparation of porcine aortic endothelial cells

Human serum samples

Eighty-two plasma samples were obtained from the American Red Cross. These samples were used either individually or collectively and are designated “normal donor” and “normal donor pool,” respectively. When less than 82 samples were required for a particular analysis, samples were selected at random to perform analysis. In addition, two plasma samples obtained from the American Red Cross that contained a high concentration of anti-Gal␣1-3Gal antibodies13 were pooled and used as “standard serum.” Patient serum was collected at Duke University Medical Center after informed consent was obtained from all patients as approved by Duke’s Institutional Review Board and in accordance with the ethical standards of the Helsinki Declaration of 1975 (revised 1983). Serum was obtained before operation and between 2 and 36 weeks (mean 11.6 weeks) after surgical exposure to Thrombogen. All patients undergoing primary coronary artery bypass grafting at Duke University Medical Center were eligible for the study. Exclusionary criteria for the study included abnormal baseline coagulation studies not attributable to anticoagulant therapy and enrollment in a separate study that precluded inclusion in this study. No change in the attending physician’s surgical

Porcine aortic endothelial cells (PAECs) were harvested from the aortas of pigs at the Duke University Medical Center Vivarium. Briefly, the aortas of sacrificed pigs were dissected free of surrounding tissues and placed in DMEM under sterile conditions. The vessel was then opened lengthwise and bathed in 1% collagenase for 5 minutes. PAECs were harvested by gently rolling a wooden swab over the luminal surface, and they were cultured in DMEM supplemented with 20% fetal bovine serum, 4 mmol/L L-glutamine, 10 U/mL penicillinstreptomycin, 1 U/mL porcine-derived heparin, and 15 mg/mL endothelial cell growth supplement. Endothelial cell morphology was visualized by light microscopy and positive uptake of fluorescent-labeled acetylated-LDL and negative staining of smooth muscle ␣-actin. Cells were used between passages 4 and 10. Homology between human and bovine prothrombin

The amino acid sequences for human prothrombin (GenBank accession number NM_000506) and bovine prothrombin (GenBank accession number J00041) were compared in a National Center for Biotechnology Information standard protein-protein Basic Local Alignment Search Tool (http://www.ncbi.nlm.nih.gov/ BLAST) using the default settings. These findings were

Vol. 197, No. 4, October 2003

confirmed using the ClustalW 1.8 global progressive alignment algorithm, which is available through the European Bioinformatics Institute (http://www.ebi.ac.uk). Specific activity of thrombin preparations

Thrombogen, Thrombin-JMI, and the human thrombin component of Tisseel were diluted and loaded in duplicate onto a Falcon Pro-Bind (Becton Dickinson) flat-bottom assay plate. Spectrozyme-TH in PBS was added to each well, and the rate of change in absorbance at 405 nm was then measured on an EL 340 Microplate Reader from Bio Kinetics (Winooski, VT). Bovine thrombin was used as a standard to calculate units of thrombin activity in each preparation. The protein concentration of each preparation was determined by a Lowry assay. Determination of relative purity of Thrombogen, Thrombin-JMI, and Tisseel

The preparations were reduced in 5% (vol/vol) 2-mercaptoethanol in SDS-PAGE sample buffer, separated by SDS-PAGE14 on a 5% to 15% gradient, and stained with Coomassie blue. To compare the preparations at therapeutically equivalent concentrations, 25 mU of each preparation was applied to each lane. To compare equal amounts of protein in each preparation, 25 ␮g of preparation was applied to each lane. Determination of Gal␣1-3Gal by Western blot

Equal amounts of the preparations were separated as described earlier. After transfer to nitrocellulose, the membranes were cut into strips and blocked with blocking buffer (0.1% HSA and 0.05% Tween-20 in PBS) for 12 hours at 4°C. The strips were probed for 1 hour at 25°C with 3% “standard serum” or 3% “standard serum” incubated with 10 mmol/L galactobiose to block binding of anti-Gal␣1-3Gal IgM. Incubation of strips using buffer alone was performed as a negative control (data not shown). The strips were then incubated at 4°C for 1 hour with goat antihuman IgM (␮-chain specific) conjugated with alkaline phosphatase, which was diluted 1:700 in blocking solution. The strips were then developed using a solution containing 5-bromo-4chloro-3-indolyl phosphate/nitroblue tetrazolium as the chromogenic substrate. Determination of natural IgM and antigen content by ELISA

The amount of “natural” (preimmune) IgM in normal human serum specific for various hemostatic prepara-

Schoenecker et al

Purity of Bovine Thrombin

583

tions was determined by ELISA under conditions in which the antibody concentration was limiting. The relative amount of antigen in a given thrombin preparation was approximated by ELISA under conditions in which the antibody concentration was high. For these experiments, Nunc Maxisorb 96-well plates were incubated for 3 hours at 24°C with Thrombogen, Thrombin-JMI, or each component of the Tisseel kit (human fibrinogen, human thrombin, and bovine aprotinin). Equal amounts of protein (1.5 ␮g per well) from each preparation were used. A portion of each plate was incubated with 1.5 ␮g per well HSA as a control. The wells were blocked with blocking buffer (0.1% HSA and 0.05% Tween-20 in PBS) and then incubated with each normal donor’s serum for 1 hour at 24°C in duplicate. For approximation of antigen content, 100% serum was used, and 2% serum was used to determine the relative concentration of antibody. After incubation with human serum, the plates were washed and then incubated at 23°C for 1 hour with antihuman IgM (1:700) conjugated with alkaline phosphatase. The plates were developed using p-nitrophenyl phosphate in diethanolamine buffer, and the change in absorbance at 405 nm was measured. The rate of change in absorbance in wells coated with HSA only was taken to be “background” and was subtracted from the rate in wells coated with antigen. Each assay was standardized using the “standard serum.” For determining the antigen content of PAECs, 104 cells were placed in each well of a Falcon 96-well flatbottom tissue culture plate (Becton Dickson) and allowed to grow to confluency (3 days). Plates were then washed with PBS three times and fixed in 1% formaldehyde at room temperature for 5 minutes. The “normal donor pool” in the ELISA protocol (described earlier) was used as a source of antibody. The binding of antibody to Gal␣1-3Gal was taken to be the binding that was inhibited by 10 mmol/L galactobiose, and binding to epitopes other than Gal␣1-3Gal was taken to be that binding not inhibited by the presence of 10 mmol/L galactobiose. Determination of cross-reactivity by ELISA

Detection of IgG binding to Thrombogen, ThrombinJMI, each individual component of the Tisseel kit, porcine thyroglobulin, and PAECs were determined by ELISA as described earlier. The primary antibody used in these assays was that of the preexposed pool or the

584

Schoenecker et al

Purity of Bovine Thrombin

J Am Coll Surg

Table 1. Currently Used Hemostatic Agents

Product Manufacturer

Thrombin-JMI, Jones Pharma Inc. Thrombogen, Johnson & Johnson Hemaseel APR, Haemacure Tisseel, Baxter Costasis US Surgical FloSeal Fusion Cryo Gel Tessenderlo Platelet Gel, Harvest Technologies AGF, Interpore Cross

Xenogeneic protein*

Yes Yes Yes Yes Yes Yes Yes Yes Yes

Source of thrombin

Bovine Bovine Human Human Bovine Bovine Bovine Bovine Bovine

Other ingredients

Mode of action (thrombin or fibrinogen based)

None None Fibrinogen, aprotinin, calcium chloride Fibrinogen, aprotinin, calcium chloride Collagen, patient’s plasma Gelatin matrix Cryoprecipitate unit from blood bank Platelet-rich plasma, calcium chloride Patient’s blood

Thrombin Thrombin Fibrinogen Fibrinogen Fibrinogen Other (matrix formation) Fibrinogen Fibrinogen Fibrinogen

*Some preparations contain xenogeneic components even though the primary component is human.

postexposed pool (2%) with and without 10 mmol/L galactobiose for 1 hour at 24°C. Antihuman IgG conjugated to alkaline phosphatase was used as a secondary antibody. The plates were developed and measured as described earlier. RESULTS Hemostatic agents containing thrombin

Three sources of thrombin are available in the United States (Table 1). Thrombogen and Thrombin-JMI are bovine thrombin preparations, and they are obtainable as stand-alone products or they are included in other products. Tisseel and Hemaseel (Haemacure Corp) are fibrinogen-based products containing human fibrinogen, human thrombin, and bovine aprotinin. In this study, only the human thrombin component of Tisseel was characterized, because Tisseel and Hemaseel are identical products. Homology between human and bovine prothrombin

Prothrombin is the proenzyme that, when cleaved appropriately, yields proteolytically active thrombin. A proteolytically inactive fragment is also released during cleavage of prothrombin. This fragment is a common contaminant in thrombin preparations and contains the gla and kringle domains of prothrombin. When compared with the bovine sequence, human prothrombin showed 81% identity (503 of 617 amino acids) and 89% similarity (554 of 617 amino acids) with three amino acid gaps. All functional domains, including the gla, kringle, antithrombin III binding, and active site of the protease, were highly conserved. The longest continuous nonhomologous stretch was four amino acids occurring in a nonfunctional domain. Further, in areas of nonhomology, 45% (51 of 114 amino acids) of amino

acid substitutions were conservative in nature. These results were confirmed using the Clustal W Medicine alignment tool. Relative purity of hemostatic agents containing thrombin

Thrombogen and Thrombin-JMI were found to contain a large number of contaminants. When these products were separated by SDS-PAGE (Fig. 1), many bands that did not correspond to either chain A or chain B of ␣-thrombin were revealed. When compared at equivalent units of thrombin activity (Fig. 1), Thrombin-JMI was evidently a relatively more pure preparation than Thrombogen. Nevertheless, impurities of ThrombinJMI were observed when more preparation was loaded (Fig. 1). The relative purity of these thrombin preparations was further evaluated by comparing their respective specific activities (units of enzyme activity per unit protein; Table 2). Based on this evaluation, ThrombinJMI is approximately 10 times more pure than Thrombogen. Because of the apparent method of preparation, it proved difficult to determine the purity of the thrombin component of Tisseel. After electrophoretic separation of the proteins in the preparation and staining, one broad band was apparent in the gel. The broad band (39 to 81 kd; Fig. 1) in the thrombin component of Tisseel was determined to be HSA based on reactivity with an antihuman serum albumin antibody (evaluated by immunoblotting). It is unlikely that the HSA in this preparation is a contaminant, because albumin is not a common contaminant in thrombin preparations. Rather, the albumin is apparently used as a stabilizer for the enzyme preparation.

Vol. 197, No. 4, October 2003

Schoenecker et al

Purity of Bovine Thrombin

585

Presence of Gal␣1-3Gal on xenogeneic hemostatic agents

Some of the glycoproteins in Thrombogen and Thrombin-JMI contain the Gal␣1-3Gal epitope, but bovine aprotinin (a component of Tisseel) does not (Fig. 2). Normal human serum contained natural IgM, which bound to multiple bands on Thrombogen, ThrombinJMI, and bovine aprotinin (Fig. 2). Binding of natural IgM to Thrombogen and Thrombin-JMI was decreased in the presence of galactobiose (Fig. 2), which inhibits the binding of anti-Gal␣1-3Gal antibodies to Gal␣13Gal epitopes. Using this approach, Thrombogen was found to have multiple glycoproteins that contain Gal␣1-3Gal. In contrast, Thrombin-JMI was found to have a single protein (apparent subunit weight 91 kd) that contains Gal␣1-3Gal, and bovine aprotinin was not found to contain Gal␣1-3Gal. Quantification of the amount of Gal␣1-3Gal epitopes on the bovine thrombin preparations (Table 2) revealed that Thrombogen has about 6.5-fold more Gal␣1-3Gal per unit of protein than does ThrombinJMI. Comparing units of thrombin activity, this corresponds to roughly 60-fold more Gal␣1-3Gal on Thrombogen than on Thrombin-JMI. Based on the analysis just described, most but not all of the IgM that bound to the bovine thrombin preparations was specific for Gal␣1-3Gal. The remaining antigens on the bovine thrombin preparations have not been identified, but some studies would suggest that xenogeneic proteins possess a variety of epitopes that are recognized by polyreactive natural antibodies present in the serum of all individuals.18 The binding of polyreactive antibodies to bovine thrombin preparations has not been verified, and additional studies will be required to address this issue. Reactivity of natural IgM with hemostatic agents

The antigen content of the bovine thrombin preparations and of the human thrombin preparation were determined by ELISA (Fig. 3). The bovine preparations contained substantial amounts of antigen, but the human thrombin preparation did not (Fig. 3A). In fact, the human thrombin preparation, on average, showed less reactivity with human antibodies than did HSA. This might be because some natural antibodies bound to the HSA preparation or because the human thrombin prevented nonspecific binding to the plate better than did HSA.

Figure 1. Proteins present in Thrombogen (TGEN), Thrombin-JMI (TJMI), and the thrombin component of Tisseel (HT). The preparations were separated by SDS-PAGE and stained with Coomassie blue. In the first three lanes, 25 mU were loaded per lane. In the second three lanes, 25 ␮g were loaded per lane. The position of the band corresponding to chain A (6 kd) of bovine thrombin is indicated by “-A-.” Chain B (31 kd) of bovine ␣-thrombin (factor IIa, the biologically active form of thrombin) is labeled “-B␣.-” Chain B of ␤-thrombin, a partially active proteolytic degradation product of ␣-thrombin, is labeled “-B␤.-” Chains A, B␣, and B␤ are all readily visible in the lane loaded with 25 ␮g of TJMI. The major band between chain A and chain B␤ in the lane loaded with 25 ␮g of TJMI is likely a fragment of chain B, although this has not been confirmed. Proteolytic cleavage of ␣-thrombin to yield ␤-thrombin and other degradation products is expected15 in the absence of protease inhibitors, which are not added to the therapeutics.

There were higher levels of natural antibodies in normal human serum that recognized the bovine preparations compared with the human preparation (Fig. 3B). Under the conditions used, about one-third of the individuals tested did not show reactivity with the bovine thrombin

586

Schoenecker et al

Purity of Bovine Thrombin

J Am Coll Surg

Table 2. Purity and Immunoreactivity of Thrombin Preparations Measurement

Thrombin units/mg prep mg protein/mg prep Thrombin units/mg protein Relative amount of antigen-specific antibody in preimmune serum† Relative antigen content‡ Gal␣1–3Gal/protein§

Thrombogen

Thrombin-JMI

Tisseel IIa

Human IIa*

62.50 0.653 95.7 1.0 ⫾ 0.25

153.85 0.171 898.0 0.86 ⫾ 0.14

8.33 1.000 8.33 No antibody found

2,530 1.0 2,530

1.0 ⫾ 0.10 0.55 ⫾ 0.04

0.53 ⫾ 0.07 0.084 ⫾ 0.021

No antigen found ⫺0.001 ⫾ 0.01

*The specific activity (units/mg protein) of human thrombin was defined by Friberger.16 A comparison of bovine thrombin and human thrombin15 has shown that purified bovine thrombin may contain more ␤-thrombin and ␥-thrombin, which are proteolytic breakdown products of ␣-thrombin (refer to Fig. 1 for example). Both ␤-thrombin and ␥-thrombin effectively cleave chromogenic substrates used in thrombin assays, but unlike ␣-thrombin, they are biologically inactive. † The level of antibody specific for a given thrombin preparation was determined by ELISA under conditions in which the amount of antibody was limiting (2% serum). Serum from 62 normal individuals was tested separately, and the standard error of the mean for all 62 samples is shown. ‡ Reactivity of antibody with a given amount (by weight) of protein in a particular preparation in the presence of high antibody concentrations (100% serum) was taken as a measure of the amount of antigen in that preparation. Serum from 64 normal individuals was tested separately, and the standard error of the mean for all 64 samples is shown. § The units of Gal␣1–3Gal are expressed as the number of residues per 100,000 kd of protein mass. The number of Gal␣1–3Gal residues was approximated by comparing the binding of anti-Gal␣1–3Gal antibodies to the thrombin preparation with the binding of the same antibodies to porcine thyroglobulin, which contains 0.91 Gal␣1–3Gal residues per 100,000 kd of protein mass.17 A direct ELISA was used for the comparison, as described in the Methods section.

products, but under more physiologic conditions (100% serum, Fig. 3A), reactivity of the bovine thrombin preparations was observed with almost all human sera. This is consistent with the observation, described earlier, that the bovine preparations contain Gal␣1-3Gal, an antigen that is recognized by natural antibodies present in all immunocompetent individuals. Thrombogen and Thrombin-JMI were found to have approximately the same natural immunoreactivity as porcine thyroglobulin (Fig. 3), a glycoprotein known to react with anti-Gal␣1-3Gal antibodies.17 Further, Thrombogen and Thrombin-JMI were found to have approximately the same natural immunoreactivity as PAEC (Fig. 3), a commonly used model for the immune nature of porcine xenografts. The bovine aprotinin preparation (component of Tisseel) was found to have considerably less natural immunoreactivity than Thrombin-JMI and Thrombogen, but there was some immunoreactivity of the bovine aprotinin with normal human serum, because the mean immunoreactivity with the aprotinin preparation was substantially greater than reactivity with either HSA, human thrombin, or human fibrinogen. Figure 2. Binding of anti-Gal␣1-3Gal antibodies to proteins present in Thrombogen (TGEN), Thrombin-JMI (TJMI), and the bovine aprotinin (BA) component of Tisseel. The preparations were separated by SDS-PAGE, transferred to a membrane, and stained with human antibody in the presence (indicated by a plus sign) or absence of 10 mmol/L galactobiose as described in the Methods section. The serum used as a source of antibody was a mixture of two normal sera containing high concentrations of anti-Gal␣1-3Gal immunoglobulin M (“standard serum” described in Methods). Twenty-five ␮g of protein were loaded per lane.

Cross-reactivity of antibodies

Antibodies elicited by exposure to Thrombogen were found to cross-react with Thrombin-JMI, PAEC, and porcine thyroglobulin, but not with bovine aprotinin, human thrombin, or human fibrinogen (Fig. 4). When individuals are exposed to porcine organs, one of the primary antigens recognized by the IgG elicited by the exposure is Gal␣1-3Gal. In contrast, only 1% and 2% of

Vol. 197, No. 4, October 2003

Schoenecker et al

Purity of Bovine Thrombin

587

Figure 3. Immunoreactivity of normal human serum with Thrombogen (TGEN); Thrombin-JMI (TJMI); and the bovine aprotinin (BA), human thrombin (HT), and human fibrinogen (HF) components of Tisseel. Immunoreactivity was determined by ELISA as described in the Methods section. (A) The amount of antigen in the preparations was determined using a constant amount of antigen and high concentrations of antibody (100% serum). (B) The amount of antibody in normal human serum that reacted with the various preparations was determined using a constant amount of antigen and low concentrations of antibody (2% serum). The means for each group of data are shown as a solid line. The reactivity of pooled normal human serum with porcine thyroglobulin (dashed line) and porcine aortic endothelial cells (dotted line) is shown.

the binding of elicited IgG to Thrombogen and Thrombin-JMI, respectively, were inhibited by the presence of 10 mmol/L galactobiose, indicating that the majority of the elicited antibodies are not specific for Gal␣1-3Gal. This is despite the observation that increases in anti-Gal␣1-3Gal IgG are observed in the serum of most individuals after exposure to bovine thrombin preparations.11 These findings suggest that bovine thrombin preparations contain numerous epitopes or contain predominant epitopes other than Gal␣1-3Gal. DISCUSSION The present findings demonstrate that the two currently available bovine thrombin preparations used in the United States are recognized by IgM from the sera of normal, unsensitized individuals, but a human thrombin preparation is not. Importantly, Thrombogen is 10 times less pure than Thrombin-JMI and, on a per-unit basis, contains as much as 60-fold more Gal␣1-3Gal than does Thrombin-JMI. This finding is important,

because all immunocompetent individuals have antiGal␣1-3Gal antibodies in their serum. Despite the fact that Thrombin-JMI is more pure than Thrombogen and less reactive with natural antibodies than Thrombogen, the preparation still has considerable immunoreactivity—similar to that of porcine endothelial cells. This finding is of particular interest because the reaction between porcine endothelial cells and human natural antibodies is strong enough to initiate the hyperacute rejection of xenografts.19 The xenogeneic nature of the bovine thrombin preparations likely contributes to their observed immunostimulatory properties. We previously reported that human and bovine coagulation factors share approximately 70% homology.11 Here, we demonstrate that although most of the functional domains are conserved between species, 10% to 20% of the primary amino acid structure differs between human and bovine prothrombin. These differences likely represent antigenic determinants that can lead to the development of autoantibodies

588

Schoenecker et al

Purity of Bovine Thrombin

Figure 4. Cross-reactivity of antibodies elicited against Thrombogen. The binding of immunoglobulin G against Thrombogen (TGEN); Thrombin-JMI (TJMI); the bovine aprotinin (BA), human thrombin (HT), and human fibrinogen (HF) components of Tisseel; porcine thyroglobulin (PT); and porcine aortic endothelial cells (PAEC) was determined by ELISA.

against conserved domains through molecular mimicry (see Schoenecker and colleagues11 for review). Further, the binding of natural antibodies, present in the serum of all immunocompetent individuals, to Gal␣1-3Gal acts effectively as an adjuvant,20 leading to the production of antibodies against other antigens. It is likely that the combination of protein and carbohydrate differences between bovine and human thrombin plays a key role in the elicited immune response against bovine thrombin.

J Am Coll Surg

Some studies strongly suggest that the xenogeneic nature of an antigen can dramatically increase the likelihood that, on exposure to that antigen, an adverse immune reaction will occur. For example, exposure to xenogeneic but not allogeneic ␤-2-glycoprotein I (a 50-kd serum cofactor also known as apolipoprotein H) results in development of autoimmune disease in a mouse model of antiphospholipid syndrome.21 Further, exposure to bovine collagen elicits a strong immune response in humans22 and is associated with the development of an autoimmune response in some patients.23 These observations indicate that bovine thrombin is not the only bovine reagent associated with a pathogenic autoimmune response and suggest that xenogeneic material should be used therapeutically only when autologous materials are not available or are deemed unsafe. The presence of impurities in the bovine thrombin preparations, especially Thrombogen, may contribute to their immunostimulatory properties. Fastenau and McIntyre24 demonstrated that most of the impurities in Thrombogen are plasma proteins. These impurities may contribute to the immunostimulatory properties of these products in several ways. First, because of the differences in purity, patients may be exposed to greater doses of antigen than necessary. Thrombin-JMI has a specific activity 10 times higher than Thrombogen. Because these products are dosed by units, patients who receive Thrombogen will receive 10 times more xenogeneic protein than those receiving Thrombin-JMI. Second, many of the coagulation factor impurities are less homologous to their corresponding human factor than thrombin is. Further, the main source of Gal␣1-3Gal in these preparations is expressed not on thrombin but rather on the contaminating proteins. For example, almost all of the Gal␣1-3Gal in Thrombin-JMI is expressed on contaminating proteins, so it is possible that these contaminating factors are more immunogenic than thrombin itself. Most of the reported adverse reactions against thrombin support this idea, associating the use of bovine thrombin with the development of antibodies against factor Va.2 The fact that bovine factor Va and human factor Va are only 73% homologous may contribute to the antigenic nature of bovine factor Va. Further, bovine factor Va, unlike bovine thrombin, is heavily glycosylated and so may contain xenogeneic carbohydrate recognized by natural antibodies. It is estimated that more than 500,000 patients per year are exposed to bovine thrombin in the United

Vol. 197, No. 4, October 2003

States.11 Many of these patients require a second operation. The warning label on these products cautions a surgeon not to use these products if a patient has antibodies against bovine proteins. This warning is generally ignored, because all normal, unsensitized individuals have natural antibodies against the preparation. Further, we have previously shown that 90% of humans exposed to a bovine thrombin preparation develop IgG antibodies against one or more of the proteins in the product,2 so according to the warning label, 90% of patients who have already been exposed should not be reexposed. In further support of this point, we previously found an increase in postoperative adverse events (adjusted odds ratio of 5.40 with a 95% confidence interval of 1.54 to 18.8) in patients with high levels of preoperative antibodies against two or more bovine coagulation proteins.2 With this in mind, it is important to know whether antibodies elicited against one bovine thrombin product will cross-react with other hemostatic products. Samples for this study were collected at a time when Thrombin JMI was not available and when Thrombogen was in common use. Because Thrombin JMI is a relatively new product, we felt it was important to address questions regarding the reactivity of antibodies to Thrombin JMI in patients exposed to Thrombogen. We determined that the antibodies elicited against Thrombogen cross-reacted with Thrombin-JMI, but these antibodies did not bind to any of the components in the Tisseel kit, including bovine aprotinin. If a patient has been exposed to bovine thrombin, the use of a humanderived preparation would likely be safer than the repeated use of another bovine thrombin product. Although incisive clinical trials are lacking, we recommend that surgeons reconsider the use of bovine thrombin because of its antigenicity and because of the case reports associated with its use. Unfortunately, consideration of the alternatives is not encouraging at the present time. There is currently no commercially available stand-alone source of human thrombin. As explained earlier, the bovine thrombin products are thrombin based and the human-derived products are fibrinogen based. If a surgeon wishes to use a human-derived hemostatic agent, a fibrinogen-based product must be used. Because thrombin-based products facilitate formation of fibrin clots primarily at the site of vascular injury, and the fibrinogen-based products form a fibrin clot at the application site, regardless of any fibrinogen released from the vessel, some surgeons prefer thrombin-based

Schoenecker et al

Purity of Bovine Thrombin

589

products to fibrinogen-based products. Consequently, until a stand-alone human thrombin product is available, if a surgeon has a preference for a human-derived thrombin-based product, the best option may be to use the human thrombin component of the human fibrinogen kits. This practice is not FDA approved, and the human fibrinogen kits contain 250 to 2,500 U of thrombin, compared with bovine thrombin kits, which contain 5,000 to 20,000 U of thrombin. If a surgeon feels that a bovine thrombin product must be used, we strongly recommend using Thrombin-JMI in lieu of Thrombogen. This recommendation, given the absence of clinical trials, is based on the supposition that the less immunogenic of the two preparations is preferable. More emphatically, we urge the rapid development of a human-derived thrombin-based therapeutic. Author Contributions

Study Conception and Design: Schoenecker, Lawson, Parker Acquisition of Data: Schoenecker, Johnson, Fields, Lesher, Domzalski, Baig Analysis and Interpretation of Data: Schoenecker, Lawson, Parker Drafting of Manuscript: Schoenecker, Johnson, Fields, Domzalski, Baig, Lawson, Parker Critical Revision: Schoenecker, Lesher, Parker, Lawson Statistical Expertise: Parker, Lawson Obtaining Funding: Lawson, Parker Supervision: Parker, Lawson Acknowledgment: We would like to thank Meg Mercer and Betty Thames for collection of patient samples. We are also grateful to Susan H Schoenecker and Susanne Meza-Keuthen for invaluable assistance.

REFERENCES 1. Committee on Plasma and Plasma Substitutes, Division of Medical Sciences, National Research Council. Statement on normal (whole, pooled) human plasma. Transfusion 1968;8: 57–59. 2. Ortel TL, Mercer MC, Thames EH, et al. The immunologic impact and clinical outcomes following surgical exposure to bovine thrombin. Ann Surg 2001;233:88–96. 3. Banninger H, Hardegger T, Tobler A, et al. Fibrin glue in surgery: frequent development of inhibitors of bovine thrombin and human factor V. Br J Hematol 1993;85:528–532. 4. Knobl P, Lechner K. Acquired factor V inhibitors. Baillieres Clin Haematol 1998;11:305–318. 5. Dorion RP, Hamati HF, Landis B, et al. Risk and clinical sig-

590

6. 7. 8. 9. 10. 11.

12. 13. 14. 15.

Schoenecker et al

Purity of Bovine Thrombin

nificance of developing antibodies induced by topical thrombin preparations. Arch Pathol Lab Med 1998;122:887–894. Tadokoro K, Ohtoshi T, Takafuji S, et al. Topical thrombin– induced IgE-mediated anaphylaxis. RAST analysis and skin test studies. J Allergy Clin Immunol 1991;88:620–629. Rothenberg DM, Moy JN. Anaphylactic reaction to topical bovine thrombin. Anesthesiology 1993;78:779–782. Pope M, Johnston KW. Anaphylaxis after thrombin injection of a femoral pseudoaneurysm: recommendations for prevention. J Vasc Surg 2000;32:190–191. Schoenecker JG, Johnson RK, Lesher AP, et al. Exposure of mice to topical bovine thrombin induces systemic autoimmunity. [comment] Am J Pathol. 2001;159:1957–1969. Lai L, Kolber-Simonds D, Park K-W, et al. Production of ⫺1,3galactosyltransferase knockout pigs by nuclear transfer cloning. Science 2002;295:1089–1092. Schoenecker JG, Hauck RK, Mercer MC, et al. Exposure to topical bovine thrombin during surgery elicits a response against the xenogeneic carbohydrate galactose(alpha1–3)galactose. J Clin Immunol 2000;20:434–444. Strukova SM. Thrombin as a regulator of inflammation and reparative processes in tissues. Biochem (Moscow) 2001;66:8– 18. Parker W, Bruno D, Holzknecht ZE, Platt JL. Characterization and affinity isolation of xenoreactive human natural antibodies. J Immunol 1994;153:3791–3803. Laemmli UK. Cleavage of structural proteins during the assembly of the head of bacteriophage T4. Nature 1970;227:680– 685. Witt I. Test systems with synthetic peptide substrates in haemostaseology. Eur J Clin Chem Clin Biochem 1991;29:355–374.

J Am Coll Surg

16. Friberger P. Synthetic peptide substrate assays in coagulation and fibrinolysis and their application on automates. Semin Thromb Hemost 1983;9:281–300. 17. Thall A, Galili U. Distribution of Gal alpha 1-3Gal beta 1-4GlcNAc residues on secreted mammalian glycoproteins (thyroglobulin, fibrinogen, and immunoglobulin G) as measured by a sensitive solid-phase radioimmunoassay. Biochemistry 1990; 29:3959–3965. 18. Gaca JG, Lee W, Aksoy O, et al. Evidence for polyreactive xenoreactive antibodies in the repertoire of human anti-swine antibodies: the “next” humoral barrier to xenotransplantation? Transplant Immunol 2001;9:19–27. 19. Cooper DK, Koren E, Oriol R. Oligosaccharides and discordant xenotransplantation. Immunol Rev 1994;141:31–58. 20. Galili U, LaTemple DC. Natural anti-Gal antibody as a universal augmenter of autologous tumor vaccine immunogenicity. Immunol Today 1997;18:281–285. 21. Gharavi AE, Sammaritano LR, Wen J, Elkon KB. Induction of antiphospholipid autoantibodies by immunization with beta 2 glycoprotein I (apolipoprotein H). J Clin Invest 1992;90:1105– 1109. 22. McClelland M, Delustro F. Evaluation of antibody class in response to bovine collagen treatment in patients with urinary incontinence [comment]. J Urol 1996;155:2068–2073. 23. Cukier J, Beauchamp RA, Spindler JS, et al. Association between bovine collagen dermal implants and a dermatomyositis or a polymyositis-like syndrome [comment]. Ann Intern Med 1993;118:920–928. 24. Fastenau DR, McIntyre JA. Immunochemical analysis of polyspecific antibodies in patients exposed to bovine fibrin sealant. Ann Thorac Surg 2000;69:1867–1872.