J. Med. Microbiol. - Vol. 37 (l992), 56-61 0 1992 The Pathological Society of Great Britain and Ireland Classification of oral pigmented anaerobic bacilli by pyrolysis mass spectrometry and biochemical tests J. T. MAGEE", J. M. HINDMARCH, B. 1. DUERDEN" and L. GOODWIN Department of Experimental and Clinical Microbiology, University of Sheffield Medical School, Beech Hill Road, Sheffield SIO 2RX Summary. Clinical (66) and reference (5) strains of pigmented gram-negative anaerobic bacilli, identified as Prevotella intermedia (47), Pr. melaninogenica ( 1), Pr. corpora (8), Porphyromonas asaccharolyticus (12), P. endodontalis (1) and P. gingivalis (2), were examined by pyrolysis mass spectrometry (PMS) and in conventional tests. Numerical classification based on conventional test reaction patterns (CTRPs) resolved five clusters, four comprising strains identified as Pr. intermedia, Pr. corpora, Pr. melaninogenica, and P. gingivalis respectively, and one comprising strains identified as P. asaccharolyticus and P. endodontalis. Numerical classification based on PMS showed a similar division, with decreasing homogeneity of chemical composition in the order Pr. intermedia, Pr. corpora, P. asaccharolyticus, which agreed with the order of homogeneity in CTRPs. PMS clusters corresponding to the genus Porphyromonas were clearly distinct from those of the genus Prevotella. PMS and CTRP classification disagreed on cluster membership for six strains. PMS identification from blind challenge sets was in agreement with conventional identi- fication for 64 of 67 strains. Introduction infections at these In the mouth they are found in the gingival crevice in both health and Gram-negative anaerobic bacilli that produced di~ease.~Pr. melaninogenica is part of the normal black or brown colonies on blood-containing media commensal gingival flora' and appears to be of little were first described by Oliver and Wherry1 and called pathological significance,1° whereas Pr. intermedia is Bacterium (later Bac t ero ides) melaninogen icus . This more commonly associated with acute necrotising species encompassed all pigmented bacteroides but gingivitis and chronic periodontal disease.18-20 P. was subsequently divided into three subspecies.2 Sub- gingivalis is linked with severe, rapidly progressive sequently the asaccharolytic strains were removed, periodontitis and generalised juvenile periodontal first to a distinct species, B. asaccharolytic~s,~later as a disease.15'21 P. asaccharolyticus has not been associ- subgroup of species within the Bacteroides* and then ated with oral colonisation or disease, but in a study of to a separate genus, Porphyromonas, comprising three the gingival flora in patients with adult periodontal species, P. asaccharolyticus, P. endodontalis and P. disease22 we isolated large numbers of pigmented gingivalis. The saccharolytic pigmented strains re- strains, about half of which appeared asaccharolytic in mained members of the melaninogenicus-oralisgroup rapid produced small amounts of n-butyric of Bacteroides which included pigmented and non- acid, and were identified as P. asaccharolyticus. Most pigmented strains; but these too were later separated of the saccharolytic strains were identified as Pr. from the genus Bacteroides sensu stricto' (formerly the intermedia. Similar results were obtained in a subse- fragilis group of Bacteroides) as a new genus, quent longitudinal study of adult periodontal disease.26 Prevotella. The pigmented Prevotella spp. comprise The work described here was done partly because of Pr. intermedia, Pr. rnelaninogenica, Pr. corpora, Pr. doubts as to the validity of these identifications as loescheii and Pr. denticola. These changes were largely P. asaccharolyticus, based solely on rapid tests. the result of DNA homology studies. Pyrolysis mass spectrometry (PMS) is a rapid Pigmented bacteroides are members of the normal characterisation method that has been used in classi- flora of the mouth,8-10vaginal' and co10n,12 and are fication, identification and typing of micro- also implicated in a range of purulent and necrotising organism~.~~*28 The spectra produced reflect cell chemical composition as 'fingerprint ' or pattern data Received 21 May 1991; accepted 6 Sept. 1991. with a complex statistical structure. Analysis of these * The present address to which correspondence should be sent: data requires multivariate statistical strategies. How- Department of Medical Microbiology and Public Health Lab- ever, modern statistical program suites available for oratory, University of Wales College of Medicine, Heath Park, Cardiff CF4 4XN. micro- and mini-computers can be used to reduce raw 56 PYROLYSIS OF ORAL PIGMENTED ANAEROBES 57 spectral data to statistically meaningful strain similar- "strains". The replicate foils of each strain were ities, for classification, for typing, and to identify divided between four batches, each with a different spectra of unknown strains by comparison with a randomised strain sequence. The four batches of foils concurrently produced database of spectra for named were analysed over 2 days on a PYMS 200X strains. The approach is particularly relevant in studies automated pyrolysis mass spectrometer (Horizon of genera where characterisation is difficult because Instruments, Heathfield, Sussex). Immediately before strains are " biochemically unreactive " or require each batch run, six foils of three strains selected at prolonged incubation to yield readable results.28The random were pyrolysed and the spectra were disc- identification problem described above seemed a fit- arded. In our experience, this procedure limits excess ting application, which would also extend our investi- irreproducibility in spectra produced early in a batch gations into traditional and PMS numerical classi- run. fication of the anaerobic gram-negative bacilli. 29-31 The blind coding of the 73 " strains " examined was not broken until PMS classification was completed. In this study, discriminant analysis based on intensities of Materials and methods masses 50-120 yielded 11 canonical discriminant function axes (CDFAs) describing 846% of inter- Bacterial strains strain discrimination. Spectral co-ordinates on these The 71 strains of pigmented gram-negative an- CDFAs were processed in UPGMA hierarchical clust- aerobic bacilli tested comprised dental (53), vaginal (4) ering to produce a dendrogram (fig. 2). The squared and wound (9) isolates, together with the type strains Euclidian distance similarity coefficient (D,) used is an Pr. intermedia NCTC 9336, P. asaccharolyticus NCTC estimate of the x2 probability (10 do0 that spectra of 9337, P. gingivalis NCTC 11834, P. endodontalis HG the strains compared were statistically indistinguish- 370 and the collection strain P. gingivalis W50. These able.28 were preserved at - 70°C in glycerol 15 % broth, and PMS blind identification was performed as pre- cultured on BM agar3, at 37°C in an atmosphere of H, viou~ly~~"'*40 by inter-species discriminant analysis of 10 YO,N, 80 %, CO, 10 % for 48 h before testing. normalised intensities for masses 50-1 20. Unduly optimistic estimates of identification error rates are likely to be obtained if spectra of the strain to be Conventional tests identified are included in the characterisation data- The strains were examined in a range of conven- base.28 To avoid this, strains were sorted by con- tional tests, only 10 of which gave reproducible results ventional species identification, being put into the relevant in differentiation. These were : production of appropriate species set in the order of their initial acid from glucose, lactose and sucrose ;"* 25 production random numerical blind coding. The strains within of ind~le,~~~25 a-fucosidase, a-glucosidase and P-N- each species set were then numbered sequentially and acetyl glucosaminidase ;33 agglutination of sheep two identification analyses were performed. In the erythrocytes ;34 and production of n-butyric and phe- first, all spectra of strains assigned odd species set nyl acetic detected by gas-liquid chromato- numbers constituted the teaching set, describing the graph~.~~Tests which proved non-discriminatory or species characteristics in PMS, and all spectra of the irreproducible were : tolerance to sodium taurocholate, even numbered strains were the blind challenge set. In gentian violet and Victoria blue 4R; resistance to the second analysis, membership of these sets was penicillin (2 units), neomycin (1000 pg), kanamycin reversed. Spectra for the single strains representing Pr. (1000 pg) and rifampicin (1 5 pg) disks ;23-25 gas-liquid melaninogenica, P. endodontalis and P. gingivalis were chromatography for non-volatile fatty acids ;36 and omitted from these analyses, because no valid chal- fluorescence under UV ill~mination.~~37 lenge sets could be assembled. Spectra of the six strains Conventional identification was based on compari- showing disagreement between PMS and conventional son of conventional test reaction patterns (CTRPs) classification were included only in the challenge sets with those of type strains of the currently accepted in both analyses. PMS species identifications for each species. The UPGMA hierarchical clustering strat- spectrum in the challenge sets were noted, along with egy38and simple matching coefficient(S,,)39 were used the consensus of the results for the four replicate to produce a numerical classification based on the 10 spectra of each strain. If the identification results for a tests listed above, which gave the dendrogram shown