INTERNATIONALJOURNAL OF SYSTEMATICBACTERIOLOGY, Jan. 1992, p. 12-18 Vol. 42, No. 1 0020-7713/921010012-07$02.OO/O Copyright 0 1992, International Union of Microbiological Societies Application of Multivariate Analyses of Enzymic Data to Classification of Members of the Actinobacillus- Haemophilus-Pasteurella Group VESLEMJbY MYHRVOLD,' ILIA BRONDZ,2t AND INGAR OLSEN'" Department of Microbiology, Dental Faculty, University of Oslo, Oslo, Norway,' and Research Department, National Institute of Occupational Health, UmeB, Sweden2 Outer membrane vesicles and fragments from Actinobacillus actinomycetemcomitans, Actinobacillus lignier- esii, Actinobacillus ureae, Haemophilus aphrophilus, Haemophilus paraphrophilus, Haemophilus influenzae, Haemophilus parainfluenzae, Pasteurella haemolytica, and Pasteurella multocida were isolated and examined semiquantitatively for 19 enzyme activities by using the API ZYM micromethod. The enzyme contents of vesicles and fragments were compared with the enzyme contents of whole cells of the same organisms. Enzymic data were analyzed by using principal-componentanalysis and soft independent modeling of class analogy. This technique allowed us to distinguish among the closely related organisms A. actinomycetemcomitans, H. aphro- philus, and H. paraphrophiius. A. actinomycetemcomitans was divided into two groups of strains. A. lignieresii fell outside or on the border of the A. actinobacillus class. A. ureae, H. injluenzae, H. parainjluenzae, P. haemolytica, and P. multocida fell outside the A. actinomycetemcomitans, H. aphrophilus, and H. paraphro- philus classes. Organisms belonging to the Actinobacillus-Haemophilus- The enzymic characterization data were treated statisti- Pasteurella group, which constitute the family Pasteurel- cally by using multivariate analyses. We have previously laceae, have assumed increasing clinical importance in med- used such methods as auxiliary techniques to define the icine and dentistry over the last few years (see references 30, following bacterial and yeast genera: Actinobacillus, Hae- 34, 38, 39). Unfortunately, generic separation in this family mophilus, and Pasteurella (6, 11); Porphyromonas (3, 8, 9); has been questioned, and with conventional biochemical Prevotella, Bacteroides, Campylobacter, and Wolinella (2, tests it can be difficult to distinguish among the sometimes 8, 9); Treponema (4); and Candida, Torulopsis, and Saccha- confusingly similar species of these genera. This particularly romyces (7, 10). applies to Actinobacillus (Haemophilus) actinomycetem- comitans, Haemophilus aphrophilus, and Haemophilus para- MATERIALS AND METHODS phrophilus. Therefore, additional criteria should be sought to assist in the taxonomic separation of organisms belonging Bacteria. The bacterial species and strains which we used to the Actinobacillus-Haemophilus-Pasteurellagroup. and their sources are shown in Table 1. The organisms were Like a number of other gram-negative rod-shaped bacte- cultured in an atmosphere containing 80% N,, 10% H,, and ria, organisms belonging to the genera Actinobacillus, Hae- 10% CO, at 37°C for 3 days. Each cell culture was divided mophilus, and Pasteurella form budding extrusions (exten- into two parts; one part was used for isolation of vesicles and sions of the outer membrane) which can be either attached to fragments that were assessed enzymically, and the other part the bacterial cell surface or released to the environment was used for enzymic assessment of whole cells. In order to during growth (16, 19, 23, 25, 35). These structures are remove vesicles and fragments attached to whole-cell sur- referred to below as outer membrane vesicles or outer faces, a light ultrasonic treatment was performed before membrane fragments. They serve as vehicles for toxins and enzymic registration. enzymes (for a review, see reference 24), and their small Collection of outer membrane vesicles and fragments. The sizes (21 to 500 nm) permit them to cross epithelial barriers outer membrane vesicles and fragments were obtained after that are impermeable to whole cells. ammonium sulfate precipitation, differential centrifugation, The aim of this study was to examine whether the enzymic and dialysis (14). Briefly, cells from 0.5 liter of a 3-day contents of vesicles and fragments and the enzymic contents culture of bacteria were collected by centrifugation at 10,000 of whole cells, as determined by using the API ZYM X g for 15 min. Then, over a period of 2 h, 120 g of micromethod, can be used for taxonomic distinction among ammonium sulfate (May & Baker, Ltd., Dagenham, En- members of the Actinobacillus-Haemophilus-Pasteurella gland) was added to the culture supernatant (40% satura- group. Previously, the API ZYM method has been used to tion). After a second centrifugation at 20,000 x g for 40 min, distinguish whole cells of A. actinomycetemcomitans, Acti- the resulting pellet was suspended in 15.0 ml of 50 mM Tris nobacillus lignieresii, and H. aphrophilus (29). Another buffer (pH 9.5) containing 0.5 mM dithiothreitol (Sigma purpose of this study was to determine whether the enzyme Chemical Co., St. Louis, Mo.). This suspension was dia- contents of vesicles and fragments differed from the enzyme lyzed for 16 h at 4°C against 3 liters of the same buffer. The contents of whole cells within the same species. vesicles and fragments obtained were then collected by centrifugation at 27,000 x g for 40 min and resuspended in 5 ml of 50 mM Tris buffer (pH 7.2) containing 0.5 mM * Corresponding author. dithiothreitol. After a final centrifugation at 27,000 x g for 40 ? Present address: Department of Microbiology, Dental Faculty, min, the vesicles and fragments were resuspended in 1.0 ml University of Oslo, Oslo, Norway. of Tris buffer (pH 7.2) and kept at -20°C. The method 12 VOL.42, 1992 API ZYM PROFILES OF PASTEURELLACEAE 13 TABLE 1. Origins and sources of the strains of the incubated, and read according to the manufacturer’s direc- Actinobacillus-Haemophilus-Pasteurellagroup investigated tions. The enzymic tests were carried out aerobically after Strain Sample Site of isolation Sourcea inoculation with 65 r~.lof vesicles and fragments or whole bacterial cells emulsified in 2 ml of sterile, distilled water to A. actinomycetemcomitans a turbidity between McFarland no. 5 and 6 standards. All FDC Y4 1 Periodontitis FDC tests were performed three times by using material collected SUNY 75 2 Dental plaque SUNY from cultures obtained on different days. BK435 3 Pus Kilian Statistical analyses. The statistical analyses were based on FDC 511 4 Periodontitis FDC FDC 2097 5 Periodontitis FDC mean scores from three experiments. In this study the FDC 2112 6 Periodontitis FDC following two multivariate approaches were used: principal- ATCC 29523 7 Blood ATCC component analysis (18, 37) and soft independent modeling ATCC 29524 8 Chest aspirate ATCC of class analogy (SIMCA) (36). In the principal-component ATCC 33384T 9 Lung abscess ATCC method the original space for variable measurements is A. lignieresii projected down onto two low-dimension subspaces. One of ATCC 19393 17 Bovine lesion ATCC these is sample related, and the other is variable related. NCTC 4189T 25 Glands, cattle NCTC This projection also decides which of the variables contrib- A. ureae KWF 3520/59= 24 Ozaena KWF utes most to the sample-related projection. The complexity H. aphrophilus FDC 621 10 Periodontitis FDC of both models was determined by cross-validation (36, 37). FDC 626 11 Periodontitis FDC The approximate class borders for the SIMCA analysis were FDC 654 12 Periodontitis FDC constructed by using an F test. FDC 655 13 Periodontitis FDC ATCC 13252 14 ATCC ATCC 19415 15 Endocarditis ATCC ATCC 33389= 16 Endocarditis ATCC RESULTS H. paraphrophilus API ZYM data. The vesicles and fragments and the whole KWF IMoAI 27 Cerebellar abscess KWF bacterial cells obtained from 33 bacterial strains and as- KWF IIMoAI 28 Cerebellar abscess KWF KWFIIIRA43 29 Sputum KWF sessed in three independent experiments by using cultures H76 30 Oral cavity Sims produced on different days were negative for the following HK477 31 Dental plaque Kilian enzyme activities: lipase, trypsin, chymotrypsin, a-galac- ATCC 29240 32 Parietal abscess ATCC tosidase, P-galactosidase, P-glucuronidase, P-glucosidase, ATCC 29241T 33 Paronychia ATCC N-acetyl-P-glucosaminidase, a-mannosidase, and a-fucosi- H. infiuenzae dase. The results of the enzymic reactions for the bacterial ATCC 31441 21 Clinical isolate ATCC strains in the tests for the nine remaining enzymes are shown ATCC 33533 22 Blood ATCC in Table 2 (vesicles and fragments) and in Table 3 (whole ATCC 8143T 26 NCTC cells). H. aphrophilus and H. paraphrophilus produced H. parainfiuenzae NCTC 4101 19 Human tongue NCTC a-glucosidase, while A. actinomycetemcomitans, A. lignier- ATCC 7901 20 ATCC esii, Actinobacillus ureae, Haemophilus influenzae, Pas- P. haemolytica NCTC 9380T 23 NCTC teurella haemolytica, and Pasteurella multocida did not. P. multocida NCTC 10322T 18 NCTC Minor differences were occasionally found between the enzymic profiles of vesicles and fragments and the enzymic a ATCC, American Type Culture Collection, Rockville, Md.; NCTC, profiles of whole cells for specific organisms. In such cases National Collection of Type Cultures, London, England; FDC, Forsyth Dental Center, Boston, Mass.; SUNY, State University of New York at the activity tended to be highest in the whole-cell prepara- Buffalo, Buffalo, New York; Kilian, M. Kilian, Royal Dental College, Aarhus, tions. This was particularly