Rapid Serodiagnosis of Gram-positive Bacterial Endocarditis

Journal of Infection (2001) 42, 140–144 doi:10.1053/jinf.2001.0809, available online at http://www.idealibrary.com on

Rapid Serodiagnosis of Gram-positive Bacterial Endocarditis M. Connaughton1, S. Lang3, S. E. Tebbs2, W. A. Littler1, P. A. Lambert3 and T. S. J. Elliott*2 Departments of 1Cardiology and 2Microbiology, University Hospital, Birmingham B15 2TH, UK, 3Aston Pharmacy School, Aston University, Aston Triangle, Birmingham B4 7ET, UK Objectives: To characterize a serological test for diagnosing endocarditis caused by Gram-positive cocci. Methods: We have developed an indirect enzyme-linked immunosorbent assay (ELISA) for the serological detection of Gram-positive infections. The test measures serum IgG directed towards lipid S, a recently identified exocellular glycolipid antigen which is related to lipoteichoic acid. We have previously shown the test to be of value in serodiagnosis of central venous catheter-associated sepsis and infection of orthopaedic prostheses caused by coagulase-negative staphylococci. We now describe the application of this test in endocarditis. Results: Serum IgG levels to lipid S were significantly elevated in 34 patients with Gram-positive bacterial endocarditis confirmed as ‘definite’ by the Duke criteria as compared to 50 control patients. The test had a sensitivity of 88% and a specificity of 88%. Conclusions: The assay is independent of culture results or endocardial imaging, making it complementary to currently used investigations. It may therefore be possible to refine the current Duke criteria for diagnosing endocarditis. We describe an algorithm which incorporates lipid S serology into a positive diagnostic strategy, © 2001 The British Infection Society

Introduction Bacterial endocarditis (BE) is a serious illness with an overall mortality of 12–25%.1,2 The Duke criteria for diagnosis of BE uses a probabilistic approach based on evidence of infection and endocardial involvement.3 It is superior to previous diagnostic criteria,4,5 but its utility is limited in cases where cultures are negative, or if endocardial imaging is technically difficult or non-diagnostic. Although blood cultures are generally positive prior to antimicrobial therapy, recovery of the microorganism is reduced to 64% of samples from patients pretreated with antimicrobial agents.6 The best available imaging with transoesophageal echocardiography is diagnostic in only 50–80% of cases of prosthetic valve BE.7,8 Methods of diagnosing BE therefore need to be improved further in situations where conventional tests such as echocardiography are not discriminatory. Early efforts to develop a routine serodiagnostic test for staphylococcal infection focused primarily on cellular antigens and bacterial cell wall teichoic acid. Crowder and White9 attempted to distinguish Staphylococcus aureus BE

* Please address all correspondence to: T. S. J. Elliott, Department of Clinical Microbiology, Queen Elizabeth Hospital, Edgbaston, Birmingham, B15 2TH, U.K. Accepted for publication 27 February 2001. 0163-4453/01/020140;05 $35.00/0

from other S. aureus infections, such as bacteraemia, without clinical evidence of cardiac involvement, by the measurement of anti-wall teichoic acid antibody levels in serum. Cell wall teichoic acid precipitins were detected in patients with S. aureus or coagulase-negative staphylococcirelated BE. Patients with other S. aureus infections also had detectable levels of precipitins. Precipitins, however, were not detected in patients with BE caused by other organisms.9 Background levels of anti-wall teichoic acid antibody in the general population, inter-laboratory variability (sensitivity of the assay in patients with S. aureus-associated BE ranged from 61% to 100%) and low reproducibility have limited subsequent development of this technology.10–12 Following the limited success of wall teichoic acid as a serological marker of infection, other cellular antigens have therefore been investigated, including whole cell preparations,13 peptidoglycan and lipoteichoic acid.14,15 Background levels of antibody, low sensitivity and crossreactivity have also limited the success of these tests. Assays based on exocellular rather than cellular antigens have subsequently been investigated in an attempt to distinguish serious staphylococcal infection from uncomplicated septicaemia.16,17 Serum antibody levels to a number of exocellular S. aureus antigens distinguished osteomyelitis caused by S. aureus from those due to other bacteria.18,20 As a progression of this approach, an enzyme-linked immunosorbent assay (ELISA) based on the detection of © 2001 The British Infection Society

Serodiagnosis of Endocarditis anti-lipid S antibody has been developed.21,22 The value of the test has been demonstrated in the diagnosis of central venous catheter-associated sepsis22 and orthopaedic infections due to Gram-positive bacteria.23 In this report we have investigated the value of the test as an adjunct to the Duke criteria for use in the diagnosis of BE caused by Gram-positive cocci. To achieve ‘proof of concept’ for the utility of this assay in diagnosing BE, we aimed to establish its capacity to distinguish patients with ‘definite’ or ‘possible’ BE from a comparative group with no evidence of infection and to characterize the sensitivity and specificity profile of the test.

Materials and Methods Patient Groups Thirty-four patients referred to the tertiary cardiology service at University Hospital Birmingham between 1996–1999 were included in the ‘endocarditis’ group. These were classified as ‘definite’ BE (n:29) or ‘possible’ BE (n:5) following a combination of clinical, echocardiographic, bacteriological and, where relevant, intraoperative assessment, using the standard Duke criteria for diagnosis.3 Fifty patients who were undergoing routine coronary artery bypass surgery (CABG) during the same period, and had no clinical or bacteriological evidence of BE, were selected as controls. A 10 ml sample of clotted blood was obtained from each of the 84 patients (34 ‘definite’ or ‘possible’ BE; 50 CABG controls) and assessed for anti-lipid S serum lgG titres. Ethical approval and patient consent were obtained for the study. Preparation of the Antigen The serological test was carried out as described previously.21 Essentially, exocellular material was prepared by growing seven Gram-positive coccal strains (five S. epidermidis NCIMB 40896, 40945, 40946, 40947, 40948, one S. haemolyticus NCIMB 40949 and one Micrococcus kristinae NCIMB 40950 isolated from patients with various infections at the Queen Elizabeth Hospital, Birmingham) in brain-heart infusion broth (Oxoid Ltd., UK). The culture supernatant was recovered by centrifugation and concentrated 10-fold by freeze-drying. Combined material from the seven strains (1 ml) was applied to a gel permeation chromatography column (Superose 12 HR 10/30; Amersham Pharmacia Biotech, St. Albans, Hertfordshire, UK) and eluted with water (0.8 ml fractions). Bacterial antigen was eluted from the column in fractions 10–15 (elution volume 8–12 ml) before the components of the brain-heart infusion. The lipid S antigen contained in these fractions is

141

a short chain length form of the lipoteichoic acid found in the cytoplasmic membrane and is common to a range of Gram-positive cocci of medical importance.22,24 Coating and Blocking ELISA Plates Purified antigen contained in fractions 10–15 was diluted with 100 volumes of sodium carbonate/bicarbonate buffer (0.05 M, pH 9.6). Each well of a 96-well polystyrene microtitre plate (Immulon 2HB, Dynex Technologies, Chantilly, VA, USA) was coated with 0.1 ml of diluted antigen for 18 h at 4 ºC. Excess antigen was discarded, the wells washed three times in TBS-Tween solution (0.0 l M Tris-HCl pH 7.4, 0.9% (w/v) NaCl, 0.3% (v/v) Tween-20) and the unbound sites on the plate blocked with TBSTween solution for 3 h at 4 ºC. The Assay Serial dilutions of patient serum were reacted with bound antigen at 4 ⬚C for 18 h. Bound anti-lipid S lgG antibody was detected by incubation with protein A-peroxidase (0.25 ␮g/ml in TBS- Tween; Sigma, Poole, Dorset, UK) followed by 0.1 ml of a chromogenic peroxidase substrate. This was prepared by dissolving 10 mg of 3,3⬘,5,5⬘tetramethylbenzidine (Sigma) in 1 ml of dimethyl sulphoxide and diluting with 100 ml of sodium acetate/citrate buffer (0.1 M, pH 6.0) containing 10 ␮l of H2O2 (6% v/v) at room temperature (20 ºC). The blue colour was developed for 5 min and then the reaction stopped by addition of 0.1 ml of sulphuric acid (1 M) to each well. The absorbance of the yellow coloured product in each well at 450 nm was read on a Anthos 2001 plate reader (Labtech, Ringmer, East Sussex, UK). Results were plotted as absorbance at 450 nm versus serum dilution. Standard positive and negative sera were included on all plates (the positive control was set to give a titre of 1 : 100 000 and the negative control 1 titre of 1 : 10 000) and the same batch of antigen was used throughout the study. From the graphs the serum dilution (titre) of each test sample which reduced the absorbance to 0.1 was estimated.

Results Thirty-four patients with Gram-positive coccal BE (staphylococcal BE, n:18; streptococcal BE, n:7; enterococcal BE, n:3; polymicrobial infection (including a Grampositive strain), BE n:6) were tested for anti-lipid S lgG (Fig. 1). The organisms recovered from the patients are shown in Table I, together with the Duke diagnosis of ‘definite’ or ‘possible’ BE. Major and minor Duke criteria were applied in the diagnosis of each patient3 and the causative

M. Connaughton et al.

142

between those patients with infection and the control groups. Applying this cut-off value to the endocarditis patients gave a sensitivity of 88% (30 true positives out of 34 patients with BE) and a specificity of 88% (44 true negatives out of 50 controls). Positive and negative predictive values were 83% and 92%, respectively. The likelihood ratio of a positive test was 7.33, for a negative test it was 0.14 and the accuracy was 88%. The five BE patients with polymicrobial infection, each of which included a Gram-positive bacterium, gave positive titres.

Discussion

Figure 1. Scattergram of anti-lipid S serum IgG enzyme-linked immunosorbent assay (ELISA) titres. The control sera were collected from patients undergoing coronary artery bypass surgery (n:50). The patient groups included staphylococcal (n:18), streptoccocal/ enterococcal (n:10), or polymicrobial (associated with a Gram-positive microorganism) (n:6) infections defined as ‘definite’ or ‘possible’ endocarditis using the Duke criteria. Staph, staphylococcal BE; Strep, streptococcal BE; Ent, enterococcal BE; polymicrobial, polymicrobial BE. Table I. The microorganisms recovered from blood cultures of patients diagnosed as ‘definite’ or ‘possible’ BE by the Duke criteria. Organism

Number of patients ‘Definite’ BE ‘Possible’ BE

Staphylococcus aureus Coagulase-negative staphylococci ‘Viridans’ streptococci Beta – haemolytic streptococci Other streptococci Enterococci Polymicrobial infection including Gram-positive coccal species Total

Total

1 14

0 3

1 17

5 1

0 0

5 1

1 2 5

0 1 1

1 3 6

29

5

34

organism was confirmed in every case by multiple positive blood cultures. Serum lgG titres to lipid S were elevated in BE patients (30 out of 34 patients had elevated titres) compared to those of the CABG control sera (six out of 50 had elevated titres) (means
The lipid S assay is not specific for BE, but it is an excellent indicator of persisting Gram-positive infection. Our results indicate that the assay can be used reliably to distinguish patients with BE caused by Gram-positive cocci from control patients. Thus, if BE is suspected, and other infections associated with Gram-positive cocci can be excluded, this test has potential use in confirming or excluding BE. In practice, excluding other infections associated with Gram-positive cocci should be straightforward and lipid S serology may then prove a valuable adjunct to current diagnostic methods for BE. Furthermore, application of the test to the ‘possible’ Gram-positive BE group indicated that the test might allow conversion of such patients into the ‘definite’ or ‘rejected’ Duke categories. Three of the patients in the ‘possible’ group had high lipid S titres (40 000, 50 000 and 100 000) associated with the culture of two coagulase-negative staphylococci and a mixed culture of S. epidermidis and Escherichia coli, respectively. The titres suggest that these patients belong to the ‘definite’ group. Confirming the clinical utility of lipid S serology as a diagnostic test for BE will require further prospective assessments of patients with this infection. Indeed, the use of lipid S serology may have distinct advantages over current diagnostic methods for BE. Since the host antibody response is independent of factors affecting successful in vitro culture or technical imaging problems, its validity is not affected if standard bacteriological, and imaging investigations are inconclusive. Such scenarios include situations where blood cultures have not been obtained and possible contaminant organisms have been cultured, ‘culture-negative’ cases of BE, and those where echocardiography is technically inadequate or where prosthetic valves do not allow sufficiently clear imaging. Serological testing for antibodies to lipid S may therefore complement and refine the standard Duke criteria for diagnosing BE, particularly in cases classified as ‘possible’ BE. We have prepared an algorithm for the investigation of presumptive BE which incorporates the use of lipid S serological

Serodiagnosis of Endocarditis

143

with staphylococci BE. In comparison, the titres for patients with streptococcal and enterococcal BE did not reach the levels attained in some of the staphylococcal cases. Other marker antigens specific to staphylococci, the viridans streptococci and enterococci, would complement the lipid S serological test. Some candidate antigens have been described, including EfaA, an exocellular protein from Enterococcus faecalis,25–27 a 112 kDa antigen of E. faecalis,28 Esa, an endocarditis specific antigen of Streptococcus sobrinus,29 and SsaA, an antigen of S. epidermidis.30 Further work is, however, needed to determine whether any of these antigens would be complementary to lipid S in serodiagnosis of BE. Acknowledgements

Figure 2. Duke scheme for management of patients with suspected bacterial endocarditis (BE).

This study was partly funded by the British Heart Foundation, PG96192. The contributions of Dr. T. Worthington in processing some of the serum samples and that of Prof. S. J. Eykyn in providing additional serum samples and making valuable suggestions are acknowledged.

References

Figure 3. Proposed scheme for management of patients with suspected bacterial endocarditis (BE) incorporating lipid S serological testing.

testing, and compared this to our current strategy based on the Duke diagnostic criteria (Figs 2 and 3). The clinical utility of this approach is being investigated by a prospective trial in our unit. Although the lipid S test is valuable in detecting Grampositive BE, it does not distinguish between the causative organisms including staphylococci, streptococci and enterococci. The highest lgG levels to lipid S were found in patients

1 Dyson C, Barnes R, Harrison G. Infective endocarditis: an epidemiological review of 128 episodes. J Infect 1999; 38: 87–93. 2 Sandre RM, Shafran SD. Infective endocarditis: review of 135 cases over 9 years. Clin Infect Dis 1996; 22: 276–86. 3 Durack DT, Lukes AS, Bright DK. New criteria for diagnosis of infective endocarditis: utilization of specific echocardiographic findings. Am J Med 1994; 96: 200–209. 4 Pelletier LL, Petersdorf RG. Infective endocarditis: a review of 125 cases from the University of Washington hospitals, 1963–1972. Med 1977; 56. 5 von Reyn CF, Levy B, Arbeit R, Friedland G, Crumpacker C. Infectious endocarditis: an analysis on strict case definitions. Ann Intern Med 1981; 94: 505–581. 6 Pazin GJ, Saul S, Thompson ME. Blood culture positivity, suppression by outpatient antibiotic therapy in patients with bacterial endocarditis. Arch Intern Med 1982; 142: 263–268. 7 Schulz R, Fuchs WG. Clinical outcome and echocardiographic findings of native and prosthetic valve endocarditis in the 1990’s. Eur Heart J 1996; l72: 281–288. 8 Lengyel M. The impact of transesophageal echocardiography on the management of prosthetic valve endocarditis: experience of 31 cases and review of the literature. J Heart Valve Dis 1997; 6: 204–211. 9 Crowder JG, White A. Teichoic acid antibodies in staphylococcal and nonstaphylococcal endocarditis. Ann Intern Med 1972; 77: 87–90. 10 Julander I, Granström M, Hedström S, Möllby R. The role of antibodies against alpha-toxin and teichoic acid in the diagnosis of staphylococcal infections. Infect 1983; 11: 77–83. 11 West TE, Cantey J, Apicella MA, Burdash N. Detection of antiteichoic acid immunoglobulin G antibodies in experimental Staphylococcus epidermidis endocarditis. Infect Immun 1983; 42: 1020–1026. 12 Jackson L, Sottile M, Aguilar-Torres F. Correlation of antistaphylococcal antibody titres with severity of staphylococcal disease. Am J Med 1978; 64: 629–633. 13 Kjerulf A, Espersen F, Tvede M. IgG antibody response in bacterial endocarditis using ELISA with multiple antigens. APMIS 1994; 120: 736–742. 14 Wergeland H, Endresen C, Natas O, Aasjord P, Oeding P. Antibodies to Staphylococcus aureus peptidoglycan and lipoteichoic acid in sera from blood donors and patients with staphylococcal infections. Acta Path Microbiol lmmunol Scand Sect B 1984; 92: 265–269.

144

M. Connaughton et al.

15 Verbrugh HA, Nelson RD, Peterson PK, Wilkinson BJ, Thompson RL. Serology of Staphylococcus aureus infections using multiple antigens and serum samples. J Infect Dis 1983; 148: 608. 16 Christensson B, Fehrenbach F, Hedström S. A new serological assay for Staphylococcus aureus infections: detection of lgG antibodies to S. aureus lipase with an enzyme-linked immunosorbent assay. J Infect Dis 1985; 152: 286–292. 17 Espersen F, Wheat LJ, Bemis AT, White A. Enzyme-linked immunosorbent assay for detection of Staphylococcus epidermidis antibody in experimental S. epidermidis endocarditis. J Clin Microbiol 1986; 23: 339–342. 18 Lambert P, van Maurik A, Parvatham S, Akhtar Z, Fraise A, Krikler S. Potential of exocellular carbohydrate antigens of Staphylococcus epidermidis in the serodiagnosis of orthopaedic prosthetic infection. J Med Microbiol 1996; 44: 355–361. 19 Lambert PA, Krikler SJ, Patel R, Parvathan S. Enzyme-linked immunosorbent assay for the detection of antibodies to exocellular proteins of Staphylococcus aureus in bone infection. FEMS Microbiol Lett 1992; 100: 67–70. 20 Krikler S, Lambert PA. Extracellular proteins as a potential marker of active Staphylococcus aureus infection in bone. J Med Microbiol 1992; 37: 227–23l. 21 Elliott TSJ, Tebbs SE, Moss HA, Worthington T, Spare M, Faroqui M. A novel serological test for the diagnosis of central venous catheterassociated sepsis. J Infect 2000; 40: 262–266. 22 Lambert PA, Worthington T, Tebbs SE, Elliott TSJ. Lipid S, a novel exocellular glycolipid antigen from Staphylococcus epidermidis for the

23 24 25 26

27 28 29 30

serodiagnosis of Gram-positive coccal infections. FEMS Microbiol lmmunol 2000; 29: 195–202. Rafiq M, Dias R, Lambert PA, Tebbs S, Stirling A, Treacy R, Elliott, TSJ. Serological detection of Gram-positive bacterial infection of prosthetic joints. J Bone Joint Surgery (Brit) 2000; 82-B: 1156–1161. Fischer W. Lipoteichoic acid and lipids in the membrane of Staphylococcus aureus. Med Microbiol lmmunol 1994; 183: 61–76. Aitchison EJ, Lambert PA, Smith EG, Farrell ID. Serodiagnosis of Streptococcus faecalis endocarditis by immunoblotting of surface protein antigens. J Clin Microbiol 1987; 25: 211–215. Shorrock PJ, Lambert PA, Aitchison EJ, Smith EG, Farrell ID, Gutschik E. Serological response in Enterococcus faecalis endocarditis determined by enzyme-linked immunosorbent assay. J Clin Microbiol 1990; 28: 195–200. Lowe A, Lambert PA, Smith A. Cloning of an Enterococcus faecalis endocarditis antigen: homology with adhesins from some oral streptococci. Infect Immun 1995; 63: 703–706. Burnie JP, Clark I. Diagnosing endocarditis with the cloned 112 kDa antigen of Enterococcus faecalis. J Immunol Meth 1989; 123: 217–225. Brooks W, Burnie JP. Cloning and sequencing the endocarditis immunodominant antigen of Streptococcus sobrinus strain MUCOB 263. J Med Microbiol 1994; 40: 330–337. Lang S, Livesley MA, Lambert PA, Littler WA, Elliott TSJ. Identification of a novel antigen from Staphylococcus epidermidis. FEMS Immunol Med Microbiol 2000; 29: 213–220.