Classification and Identification of Flavobacterium Species by Carbon Source Utilization

Transcription

1 JOURNAL OF CLINICAL MICROBIOLOGY, July 1987, p /87/ $02.00/0 Copyright 1987, American Society for Microbiology Vol. 25, No. 7 Classification and Identification of Flavobacterium Species by Carbon Source Utilization D. RASOAMANANJARA,' B. KOROSEC,2 AND H. MONTEILl* Institute de Bactériologie de la Faculté de Médecine, Université Louis Pasteur, Strasbourg,' France and Centre de Calcul du Centre National de la Recherche Scientifique, Strasbourg,7 Received 2 March 1987/Accepted 8 April 1987 Carbon substrates used as the sole source of carbon and energy were tested for the classification and identification of species of Flavobacterium: Flavobacterium meningosepticum, F. breve, F. odoratum, F. multivorum, F. thalpophilum, and Flavobacterium sp. group IIb. Hierarchical classification and stepwise discriminant analysis revealed three F. meningosepticum, two F. breve, two F. odoratum, and two Flavobacterium sp. group llb subgroups. Glucose, histidine, asparagine, tryptophan, maltose, citric acid, and glycine were selected as the most useful substrates to differentiate between the groups and subgroups. The various species of the genus Flavobacterium have been separated into distinct groups on the basis of DNA (3, 10), although few phenotypic features differentiate them absolutely (8). Our study was intended to determine the use of various carbon substrates as the sole source of carbon and energy by Flavobacterium meningosepticum, F. breve, Flavobacterium sp. group IIb, F. odoratum, F. multivorum, and F. thalpophilum. The advantages of such methods for the taxonomy and identification of members of the family Enterobacteriaceae and some members of the Vibrionaceae were confirmed by Véron (15) and Véron and Le Minor (16, 17), who found that the use profiles enabled the classification and, after restricting the number of substrates used, identification of the species. MATERIALS AND METHODS Bacterial strains. The following strains were tested: 13 F. breve (A40838, A54615, B33982, B38674, B44444, B49835, B54566, B58375, B64246, B79158, B97415, NCTC 200/75, and NCTC 666/76), 59 F. meningosepticum (A49822, A74007, A76407, B3332, B4167, B14430, B14441, B15859, B21632, B22418, B 23643, B26345, B26724, B26972, B29828, B30560, B31688, B32446, B32472, B34145, B35810, B36981, B37907, B38594, B38611, B38744, B37566, B39452, B40182, B40193, B40470, B40471, B40705, B41047, B41312, B41667, B41693, B41972, B43337, B43831, B44667, B44827, B46136, B46356, B47181, B47466, B47812, B58347, B60339, B63822, B64709, B68800, B69139, NCTC 10016, NCTC 10585, NCTC 10586, NCTC 10587, NCTC 10588, and NCTC 10589), 21 F. odoratum (A16987, A46456, A66955, A67866, A75632, B65910, B77131, B78400, B83305, B87050, B90332, B91164, K1031, K1518, K1689, K1713, K1778, K1807, K1830, K1899, and K1908), 27 Flavobacterium sp. group IIb (A28673, A70302, B1906, B9533, B10509, B12318, B15647, B20450, B23418, B24318, B32512, B37362, B46332, B46955, B48260, B49942, B54046, B56247, B57831, B69546, B77095, B77833, B82341, B91354, B92876, B94787, and F157), 10 F. multivorum (JKID, JKIJ, K1210, K1213, K1218, K1222, K1231, LK3A, NCTC 11033, and NCTC 11034), 1 F. * Corresponding author. thalpophilum (B86348), 9 Flavobacterium sp. (B38661, B40189, B50420, B50919, B52476, B59489, B90332, NK2C, and P3), and 6 strains that we called F. meningosepticumlike (A37574, A41986, A43935, A58537, B34262, and B47824) because they belonged to F. meningosepticum serological groups J and L, whereas other members of these serogroups were found to have affinities with F. breve or were not identified (8). Strain designations beginning with K were received from M. J. Pickett. Those designations beginning with NK, JK, and LK, from P. H. Hartman and others, were isolated at Strasbourg University Hospital from sputum, bronchial secretions, catheter tips, aspiration fluids, blood, urine, ulcers, abcesses, and vaginal and anal swabs and identified by cultural and biochemical tests as specified by Richard and Monteil (14). All strains were kept in brain heart-glycerol-horse serum (10:1:1, vol/vol/vol) at - 18 C. The strains were checked for purity on blood agar. Precautions were taken to avoid contamination of media by unwanted growth substances. After the usual cleaning, glassware was soaked for 2 days in sulfochromic acid and rinsed well in tap water and then in distilled water. Mineral medium. The mineral base of Owens and Keddie (12) with the following composition was used (in grams per liter of distilled water): K2HPO4 (0.52), KH2PO4 (0.375), (NH4)2SO4 (0.5), CaCI2 (0.05), MgSO4. 7H20 (0.2), NaCI (0.1), FeCI3.6H20 (0.01), MnSO4. 4H20 (0.002), and - Na2MoO4-2H20 (0.002), ph 6.8. Except for FeCI3 6H20, which was a solid, each component of the medium was from concentrated stock solutions. The mineral base was heated until boiling, then cooled to room temperature and filtered through filter paper. Agar (2.4%, wt/vol; Pastagar, Institut Pasteur Production) was then added to the mineral medium and sterilized for 30 min at 110 C. Growth factor requirements. The growth factor requirements were determined by using the mineral medium with the following carbon sources (wt/vol): 0.1% glucose, 0.1% acetate, and 0.1% ethanol. It was supplemented with various combinations of growth factors and with yeast extract ashes as used by Grant and Pramer (5). Yeast extract (Difco Laboratories) was dried to constant weight at 70 C, transferred to clean evaporating dishes, and then placed in a furnace, in which the temperature was gradually raised to 1285

2 1286 RASOAMANANJARA ET AL. J. CLIN. MICROBIOL. 450 C. Samples were ashed overnight. Different ash concen- of total rations were added to the mineral medium. Carbon substrates. Filter-sterilized carbon sources were added to the basal medium cooled to 50 C, and 25 ml was 100 placed in petri dishes and cooled. Among the substrates used by Véron (15) which are soluble at room temperature or after heating between 37 and 80 C were the following 35 çarbon 9 substrates which allowed growth for at least one species: at 2 g/iiter, L-alanine, L-arabinose, L-asparagine, L-arginine, L-cellobiose, L-cysteine, D-fructose, D-galactose, D-glucose, glycine, L-glutamine, D-lactose, L-histidine, L-leucine, L-iy- '0 sine, D-maltose, D-mannose, L-ornithine, L-proline, L- serine, starch, D-sucrose, D-trehalose, L-threonine, L-tryptOphan, D-xylose; at 1 g/liter, adipic acid, citric acid, ethanol, glycerol, isobutanol, methionine, meso-inositol, pyruvic acid; and at 0.25 g/liter, phenol. Cysteine did not allow growth of any strain but was considered a negative control test. Replicating method. Each strain was suspended in distilled sterile water (optical density, 1.3 at 600 nm), and all strains * 50 - U were placed on the different agar media with a Steers replicator. -O Interpretation of growth. Growth intensity was visually _ 40- coded from 0 to 5 to be as discriminating as possible. Growth a was interpreted after 2, 4, 6, 8, and 14 days of incubation at 300C.. 30 Statistical analysis. (i) Hierarchical classification. Hierarchical ascendant clustering by aggregation according to the. _._._. variance was performed on the whole sample, using the î 20 usual euclidian distance between two strains or two classes to build classes as homogeneous as possible. The withinclass variance defined the homogeneity of the classes. A new -_ s vaoai 100 El E2 E3 FIG. 2. Dendrogram of hierarchical aggregation clustering of 90 group E (58 strains). class was formed at each step by aggregation of the nearest 70 two classes, and the difference between the variance of the new class and the sum of the variances of the two former 60 classes was called the level of clustering (7). The hierarchical classification is represented by a dendrogram. (ài) Stepwise discriminant analysis. Stepwise discriminant analysis (1, 4) enables the selection of the variables for which the linear combination leads to the best separation of the 40 sclasses previously defined, using the minimum number of variables. The stepwise procedure for selecting the variables -,XOis based on the ratio of the variation within the class being considered. This ratio is called the F ratio. At each step, the variable is selected for which the F ratio is maximized. At Ï 20 step zero, the variable is selected for which the means of the classes (as determined by variance analysis) are then most 10 1 m r different. This is the first discriminating variable. At each ~-<- ~.. < ~ following step, the F ratio is computed, taking the previously selected variables into account. This procedure is repeated until the F ratio is too small. A set of new coordinates,,,..i,s, ",..., expressed as a linear combination of the selected variables A B C D E (canonical variable axes), is determined. Each strain of the FIG. 1. Dendrogram of hierarchical aggregation clustering of all studied classes was plotted on the first two canonical axes. 150 strains. CAHVOR (Association pour le développement des analyses 60

3 VOL. 25, 1987 IDENTIFYING FLAVOBACTERIUM SP. WITH CARBON SOURCES 1287 TABLE 1. Reaction of groups and subgroups to carbon substrates used Carbon- Reaction by subgroupa substrate A, A2 B1 B2 Cl C2 Di D2 E L-Leucine d (35) + (100) d (46) d (63) + (100) + (100) + (100) + (100) + (100) L-Lysine - (0) d (40) d (29) + (100) + (100) + (100) + (100) + (96) + (98) DL-Valine - (7) + (80) + (100) + (82) + (90) + (100) + (91) + (100) + (100) L-Alanine - (7) d (60) d (25) + (90) + (90) + (100) + (100) + (100) + (97) L-Histidine d (28) + (80) - (0) - (0) d (70) + (90) + (95) + (96) + (90) L-Tryptophan -(14) + (80) - (0) d (27) + (80) + (100) + (100) + (100) + (100) L-Methionine - (14) d (70) d (22) - (18) + (90) + (100) + (100) + (100) + (97) L-Cysteine - (0) - (0) - (0) - (0) - (O) - (0) - (0) - (0) - (0) Glycine -(14) d (40) d (29) d (27) d (30) - (20) + (92) + (96) + (96) L-Serine - (14) d (40) - (0) d (71) + (80) + (90) + (100) + (90) + (88) L-Asparagine - (0) + (100) + (81) + (100) + (100) + (100) + (100) + (100) + (100) L-Proline d (42) + (90) + (89) + (100) + (90) + (100) + (100) + (100) + + (100) L-Glutamine d (35) + (100) + (87) + (90) + (90) + (100) + (100) + (100) + (100) L-Threonine - (7) + (100) d (52) d (70) + (90) + (100) + (100) + (100) + (100) L-Arginine - (7) + (100) + (88) + (100) + (90) + (100) + (100) + (100) + (100) L-Ornithine - (14) + (80) + (100) d (70) + (90) + (100) + (100) + (100) + (100) D-Trehalose - (7) d (30) + (100) + (100) + (90) + (100) + + (100) + + (100) + (100) Starch - (7) +(100) +(100) + +(100) +(90) +(100) + +(100) + +(100) +(100) D-Fructose - (0) - (20) + (100) + (100) + (100) d (40) + + (100) + (100) + (100) D-Lactose - (0) - (10) + (89) + (100) d (40) d (50) + + (100) + (96) + (82) D-Galactose - (7) - (10) + (100) + (100) d (60) + (80) + + (100) + + (100) + (100) D-Sucrose - (14) - (0) + (82) d (27) d (50) + (100) + + (100) + (100) + (98) D-Mannose - (7) - (10) + (100) + (100) + (100) d (50) + (100) + + (100) + (100) D-Maltose -(7) - (10) + (94) + (100) + (80) + (100) + + (100) + (100) + (100) L-Arabinose - (0) - (10) + (88) + (100) d (40) + (100) + (100) + (100) + (100) D-Xylose - (0) - (10) + (84) + (100) d (40) + (90) + (100) + (100) + (100) D-Glucose -(0) - (10) + (94) + (100) d (40) + (100) + + (100) + + (100) + (100) D-Cellobiose - (14) - (0) + (82) + (100) d (30) + (90) + (100) + (100) + (100) Phenol - (0) - (10) d (52) + (100) d (20) d (70) + (100) + (100) + (98) Ethanol - (0) - (0) + (88) + (90) d (10) d (40) + + (100) + (100) + (100) meso-inositol - (7) - (10) d (52) d (45) - (0) d (60) + (96) + (100) + (100) Glycerol - (0) - (0) + (81) + (90) d (60) d (60) + + (100) + (100) + + (100) Isobutanol - (0) - (0) d (29) d (63) - (10) d (40) + + (100) + + (100) + (100) Pyruvic acid - (14) - (30) + (84) + (90) d (70) + (90) + + (100) + (100) + (100) Citric acid - (0) - (0) - (0) - (0) - (0) - (0) + (100) - (0) - (0) a The subgroups consisted of strains of the following Flavobacterium species: AI, 12 strains of F. odoratum; A2, 11 strains of F. odoratum; B1, 15 strains of Flavobacterium sp. group IIb; B2, 10 strains of Flavobacterium sp. group IIb; Cl, 8 strains of F. breve; C2, 13 F. meningosepticum-like strains; Dl, 15 strains off. multivorum; D2, 7 strains off. meningosepticum; and E, 59 strains of F. meningosepticum. Reactions were classified negative (-), positive (+), and strongly positive (+ +); d indicates that the strains gave different results. The percentage of positive strains (scores 2 to 5) is given in parentheses. The data in boldface type are from the substrates most useful in differentiating Flavobacterium species. de données, Institut Statistique des Universités, Paris) and BMDP 7M (University of California, Los Angeles) programs were used for hierarchical classification and stepwise discriminant analysis. They were run on an IBM 3081 MSV/XA computer at the Centre de Calcul du Centre National de la Recherche Scientifique, Strasbourg. RESULTS The addition of yeast extract ashes allowed growth of all the strains on mineral medium supplemented with glucose, acetate, and ethanol, whereas most of them did not grow on this medium with vitamins or growth factors. The addition of amino acids was not needed. Thus, the final medium was composed of the mineral medium, yeast extract ashes (1 g/liter of medium), agar, and carbon substrate. The hierarchical classification of the 150 strains is shown on the dendrogram (Fig. 1). By the naked eye, the sample was separated into five main clusters, A, B, C, D, and E. Clusters A, B, C, and D were each divided into two subgroups. Hierarchical classification was performed on cluster E, dividing it into three subgroups (Fig. 2). The subgroups were as follows: (i) the two subgroups of cluster A, consisting mainly of F. odoratum strains, (ii) the two subgroups of cluster B, one made up of Flavobacterium sp. group IIb strains and the other made up of Flavobacterium sp. group IIb strains and 1 F. meningosepticum strain, (iii) the two subgroups of cluster C, one consisting of 9 F. breve strains and the other made up of 4 F. breve, 1 Flavobacterium sp. group Il b, and the 5 F. meningosepticum-like strains, (iv) the one subgroup of cluster D mainly composed of F. multivorum strains and another subgroup composed of 5 F. meningosepticum strains, 1 Flavobacterium sp. group IIb, and 1 F. breve strain, and (v) the three subgroups of cluster E, one subgroup composed of six Flavobacterium sp. group IIb, 1 F. breve, and 1 F. meningosepticum strain, one composed of 16 F. meningosepticum and 4 Flavobacterium sp. group IIb strains, and another composed of 28 F. meningosepticum strains. Stepwise discriminant analysis was performed on the five main clusters, and eight variables (lysine, histidine, galactose, glucose, ethanol, glycerol, isobutanol, and citric acid) were selected as the most discriminating. The results of the canonical variable analysis are shown in Fig. 3. Other discriminant analyses were performed between the subgroups of clusters A and C, the subgroups of C and E, the

4 1288 RASOAMANANJARA ET AL. J. CLIN. MICROBIOL. 4e 6 3 strains belonging to GLC subgroups C and F (only one strain of the GLC subgroup A strain), while subgroup A2 consisted of strains belonging to GLC subgroup A (only one strain of the GLC subgroup F). These two subgroups are quite distinct (Fig. 4a). However, we do not know if the subgroups e a e o A. e, a.2 o c uil _6 -s -4 o Cononicul variable 1 4 s e' FIG. 3. Plot of all strains on first and second canonical axes. (The ellipses encompass the strains constituting the groups; the centers of the groups [0] are indicated.) two subgroups of D, and the three subgroups of group E, followed by canonical variable analyses. Between clusters A and C, the most discriminating variables were lysine, alanine, asparagine, sucrose, maltose, and glucose. The most discriminating variables between clusters C and E were glucose, glycine, tryptophan, ethanol, meso-inositol, and sucrose, and between subgroups D1 and D2, it was citric acid. The differentiation between the three subgroups of cluster E and between the two subgroups of cluster B required too many variables to be worthwhile. The results of the canonical variable analyses are shown in Fig. 4. The results of tests on the carbon source used are given in Table 1. These results were coded from 0 to 5 for the statistical analyses, which needed well-defined results to be finely discriminative. This was possible only when the tests were done on the same basal medium and were interpreted by the same operator, i.e., while the discrimination was very subtle. However, the identification does not need such an acute interpretation, and, for more practical purposes, the results are expressed here as negative (code 0 and 1 for negative to weakly positive), positive (code 2 and 3), and strongly positive (code 4 and 5). We selected glucose, histidine, asparagine, tryptophan, maltose, citric acid, and glycine as the most discriminating variables for all groups and subgroups, and an identification scheme is proposed in Fig. 5. s,~.3..,. 3~ el.3~.-s s. 9 -l -. -_ -i à i4li la.là * * 4 i 1 0 DISCUSSION Our study showed that no specific growth factor, except for some mineral components contained in yeast extract, was required for the growth of Flavobacterium species. Hierarchical ascendant clustering clearly separated the five species studied into five main clusters, which are quite distinct (Fig. 4). F. odoratum was split into two different subgroups; these findings agreed with base composition and DNA reassociation studies (8), electrophoretic protein patterns (R. J. Owen and P. J. H. Jackmann, Newsl. Flavobacterium-Cytophaga Group, 3:10-12, 1983), and gasliquid chromatographic (GLC) studies (13) that clearly showed that there were at least two subgroups of F. odoratum. Our F. odoratum subgroup A, consisted of Os.le -i -. i i i * i i* FIG. 4. Plot of some subgroups on first and second canonical axes. (a) Subgroups of A and C; (b) subgroups of C and E; (c) subgroups of D.

5 VOL. 25, 1987 IDENTIFYING FLAVOBACTERIUM SP. WITH CARBON SOURCES m STRAIN CLUCOSE HISTIDINE ASUINE NALT0U F.Odort 1 F.ultlvori CITRIC ACID +1~- GLYCINE Flavobcterlm F.mlotlcm Flavob lrm nlngoseptlcm-llken FIG. 5. Identification scheme for Flavobacterium species. sp. group IIb revealed by our studies correspond to those found by the other studies, because of the lack of reference strains. Group B was divided into two subgroups, but no nutritional feature differentiated them. Group C was divided into two subgroups, one with the eight F. breve strains and the other with the five F. meningosepticum-like strains, one Flavobacterium sp., and five F. breve strains, including the reference strain NCTC 666/76, although the two F. breve strains NCTC 200/75 and 666/76 were found to be closely related by DNA-DNA hybridization (9). While GLC studies of Flavobacterium had separated these two strains as did carbon substrate studies, our precedent hypothesis seems to be confirmed here (13); NCTC 666/76 might belong to another F. breve subgroup represented here by six strains. We first biochemically identified the F. meningosepticum-like strains as F. meningosepticum because they had the biochemical characteristics of F. meningosepticum, especially a rapidly positive O-nitrophenyl-B-D-galactopyranoside test (within 15 min) which was considered by Richard and Monteil (14) a distinguishing feature between F. breve and F. meningosepticum. The only test that distinguished them from the other F. meningosepticum strains was slow, weak esculin hydrolysis after 48 h, whereas all the others hydrolyzed esculin strongly within 24 h. In this study, these strains differed from F. breve strains by their growth in a minimal medium containing glucose (Fig. 5). Group D was divided into two subgroups, one containing all the F. multivorum and the F. thalpophilum strains and the other containing F. meningosepticum strains, including type strain NCTC 10016, which had been confirmed elsewhere as F.brl F.odoratm 2 atypical (11), and one F. breve strain which was probably misidentified. It was somewhat difficult to differentiate the strains of species F. meningosepticum and F. multivorum because all of them grew in most of the carbon substrate media, although F. meningosepticum generally grew less than F. multivorum, which is probably why some F. meningosepticum strains that grew well were agglomerated with the F. multivorum strains. However, growth with citric acid clearly distinguished F. multivorum from F. meningosepticum, and these subgroups are quite distinct (Fig. 4b). The F. breve and F. meningosepticum strains included in group E were probably misidentified. No nutritional or serological features differentiated the three F. meningosepticum group E subgroups, whereas canonical analysis clearly separated subgroup 1 from subgroups 2 and 3 (Fig. 4b). GLC analysis revealed three F. meningosepticum subgroups, but there was no concordance with the three subgroups found here. Some Flavobacterium sp. group IIb strains were included in this group E, probably because of the confusion existing between these two species. The many discrepancies between our results and those found in the nutritional studies of Bruun (2) are probably due to the addition of Casamino Acids and tryptophan in the media which gave results different from those obtained using ashes of yeast extract. We have been able to attribute most of the strains used in this study to their original biochemical species, although some of them were split into two or three subgroups. This study confirms the heterogeneity within F. odoratum, Flavobacterium sp. group IIb, and F. meningosepticum and reveals an intermediate group between F. meningosepticum and F. breve (the F. meningosepticum-like strains). The

6 1290 RASOAMANANJARA ET AL. concordance with the classification obtained by GLC was especially marked for F. odoratum and also for the F. meningosepticum-like strains which were assigned by GLC to F. breve, rather than to F. meningosepticum. ACKNOWLEDGMENT We are extremely grateful to Beatrice Lapeyre for her technical help. LITERATURE CITED 1. Bertier, P., and J. M. Bouroche Analyse des données multidimensionnelles. Presses Universitaires de France S.A., Paris. 2. Bruun, B Studies on a collection of strains of the genus Flavobacterium. 2. Nutritional studies. Acta Pathol. Microbiol. Immunol. Scand. Sect. B 91: Callies, E., and W. Mannheim Classification of the Flavobacterium-Cytophaga complex on the basis of respiratory quinones and fumarate respiration. Int. J. Syst. Bacteriol. 28: Dixon, W. D BMDP statistical software. University of California Press, Berkeley. 5. Grant, C. L., and D. Pramer Minor element composition of Yeast Extract. J. Bacteriol. 84: Holmes, B., R. J. Owen, and T. A. McMeekin Genus Flavobacterium Bergey, Harrison, Breed, Hammer and Huntoon 1923, 97AL, p In N. R. Kreig and J. G. Holt (ed.), Bergey's manual of systematic bacteriology, vol. 1. The Williams & Wilkins Co., Baltimore. 7. Jambu, M., and M. O. Lebeaux Classification automatique pour l'analyse des données, vol. 1, p Dunod, Paris. J. CLIN. MICROBIOL. 8. Owen, R. J., and B. Holmes Heterogeneity in the characteristics of desoxyribonucleic acid for Flavobacterium odoratum. FEMS Microbiol. Lett. 48: Owen, R. J., and B. Holmes Differentiation between strains of Flavobacterium breve and allied bacteria by comparison of deoxyribonucleic acids. Curr. Microbiol. 4: Owen, R. J., and B. Holmes Identification and classification of Flavobacterium species from clinical sources, p In H. Reichenbach and O. B. Weeks (ed.), Proceedings of the International Symposium on Yellow-Pigmented Gram-Negative Bacteria of the Flavobacterium-Cytophaga Group. Verlag Chemie, Weinheim, Federal Republic of Germany. 11. Owen, R. J., and J. J. S. Snell Desoxyribonucleic acid reassociation in the classification of Flavobacterium. J. Gen. Microbiol. 93: Owens, J. D., and R. M. Keddie A note on the vitamin requirements of some coryneform bacteria from soil and herbage. J. Appl. Bacteriol. 31: Rasoamananjara, D., F. Peladan, J. C. Turlot, H. Monteil, and C. Richard Characterization of Flavobacterium species by analysis of volatile fatty acid production. J. Gen. Microbiol. 132: Richard, C., and H. Monteil Isolement, identification et signification clinique des espèces du genre Flavobacterium. Ann. Biol. Clin. 41: Veron, M Nutrition and taxonomy of "Enterobacteriaceae" and related bacteria. I. Technical procedure for auxanograms. Ann. Microbiol. (Paris) 126A: Véron, M., and L. Le Minor Nutrition and taxonomy of "Enterobacteriaceae" and related bacteria. Il. General results and classification. Ann. Microbiol. (Paris) 126B: Véron, M., and L. Le Minor Nutrition and taxonomy of "Enterobacteriaceae" and related bacteria. III. Nutritional characters and differentiation of the taxonomic groups. Ann. Microbiol. (Paris) 126B: Downloaded from on May 11, 2018 by guest