Liquid Chromatography - Europe PMC

0 downloads 0 Views 859KB Size Report
S. sanguis, were examined by curie-point pyrolysis gas-liquid chromatography. A linear discriminant function based upon three items from the output data was.
APPLIED AND ENVIRONMENTAL MICROBIOLOGY, Jan. 1978, p. 45-50 Copyright © 1978 American Society for Microbiology

Vol. 35, No. I Printed in U.S.A.

Discrimination Between Oral Streptococci by Pyrolysis GasLiquid Chromatography M. V. STACK,* HELEN D. DONOGHUE,t AND J. E. TYLER Medical Research Council Dental Unit, University of Bristol Dental School, Bristol, BSI 2LY, England

Received for publication 3 August 1977

Washed organisms, including strains of Streptococcus mitior, S. mutans, and S. sanguis, were examined by curie-point pyrolysis gas-liquid chromatography. A linear discriminant function based upon three items from the output data was adequate for segregating the strains according to species. Strains with intermediate properties were also encountered. Sources of variability in cultures were evaluated, chromatographic performance was maintained throughout the investigation, and matching performance from a duplicate pair of chromatographic columns was demonstrated.

For many years the oral streptococci have remained a poorly characterized group. Particular human species may now be distinguished fairly readily (7), but there is a residual category of organisms which cannot yet be assigned to any particular group. Hardie and Bowden, for example, found that 15% of oral streptococci did not fit into their identification scheme (7). The placement of such strains might be facilitated by access to continuous numerical data, rather than by using key characters. The normal physiological tests used in identification of microorganisms have the disadvantage of requiring an additional period of growth by the organism being identified, during which selection or adaptation may occur. Pyrolysis gas-liquid chromatography (PGLC) is a technique that has been used with success in the identification of high-molecular-weight biological macromolecules (19). Reiner pioneered the application of PGLC to the identification of microorganisms (16), and its role in this and other biomedical studies has been established (1, 6, 9-11, 17, 20, 21). The underlying assumptions of the PGLC technique are that two organisms having the same chemical composition will give pyrolysis products of the same types and in the same proportions, and also that slight differences between two types of organism will be reflected by similar differences in pyrolysis products. Because the data are generated by a standard procedure which is recorded automatically, the acquisition of a proportion of redundant information, similar to that found using physiological tests to differentiate organisms (8), is of minor importance. However, the parameters involved in this type of analysis re-

quire careful evaluation before standardization and reproducibility can be made acceptable (13-15). Adequate treatment of the data is also necesary to obtain maximal information from the output (4, 10). In the present study, the main sources of variability in pyrograms from whole washed cells of oral streptococci were examined. With adequate control of parameters, it was then possible to use the technique to distinguish between oral streptococci of current interest: Streptococcus mitior, S. mutans, and S. sanguis. There was less requirement for such distinction in the case of S. mutans strains, since these may be identified readily by traditional methods as well as by PGLC. The use of a simple discriminant function applied to a small proportion of the data made it possible to segregate groups of these strains into their species and to suggest positions for some isolated strains with properties intermediate between those of the discrete species. MATERIALS AND METHODS Bacterial strains. More than 50 bacterial strains, many of them streptococci, were freshly isolated from human dental plaque. A collection of named, reference strains was made from cultures received from workers in other laboratories; those referred to in this study are listed in Table 1. Organisms were maintained by weekly subculture on plates of brain heart infusion (BHI) agar (Difco Laboratories, Detroit, Mich.). Sample preparation. Bottles (6 ounce [ca. 180 ml]) containing 100 ml of warm BHI broth (Difco) were inoculated with 10 ml of an 18- to 40-h culture. Depending on the growth rate of the strain, incubation was continued for a further 18 to 40 h at 36°C. Aeration was not routinely controlled. Cultures were spun down at 50C and washed three times with sterile deionized water. The resulting paste was sampled for PGLC

t Present address: School of Medical Sciences, University of Bradford, Bradford, England.

analysis. 45

Appi,. ENVIRON. MICROBIOL.

STACK, DONOGHUE, AND TYLER

46

TABLE: 1. Description and source of bacterial strains from other laboratories Species

S. mutans S. mutans S. mutans S. mutans S. mutans S. mutans S. mutans S. mutans S. mutans S. mutans S. mutans

S. S. S. S. S.

sanguis sanguis sanguis sanguis sanguis

Strain no.

AHT BHT BHT FA-1 NCTC 10449

Ingbritt C67-1 C67-25 C180-2 Kl LM-7 JENA 2697 T175-1

L/SANG GPB1 GSG2

LPA1 S. mitiora S5723 S. mitior' FW75 S. mitior' Received as a strain of S. sanguis. 'Previously termed "S. sanguis, type II." c NCTC 10712.

Source

Bleiweis Bleiweis Drucker Drucker Colman Bowden and Hardie de Stoppelaar de Stoppelaar de Stoppelaar Colman Colman

Originator Zinner Zinner Zinner

Fitzgerald Sims Krasse de Stoppelaar de Stoppelaar de Stoppelaar

Fitzgerald Bratthall

Colman de Stoppelaar Bowden and Hardie Luoma Luoma

de Stoppelaar Bowden and Hardie Luoma Luoma

Luoma Bowden and Hardie Colman

Luoma Seelemann Colman

Pyrolysis conditions. A Pye curie-point pyrolyzer pyrolysis, column heating, and cooling was completed (Pye-Unicam, Cambridge, U.K.) was used with sol- within 30 min. Column rejuvenation. To maintain column pervent-cleaned iron pyrolysis wires (curie point, 770'C) acting as carriers for submilligram quantities of formance the porous Teflon seals, and the first 10-cm washed bacterial cells. The bacterial paste was dried length of the column substrate, located within the onto the wires, which were then placed in quartz injection zone of the instrument, were replaced every pyrolysis tubes. The control unit was set to give 15-s 4 weeks, using fresh portions of the coated support heating times at the curie point, which was attained material. The rejuvenated columns were held at 230°C in less than 1 s with the customary sample loading of for 6 h before re-use. The flame ionization detectors less than 0.5 mg and the gas chromatograph injection and their associated inlet connectors were cleaned at port temperature maintained at 200°C. Pyrolysis wires the same time to remove accumulated high-boilingwere stored over molecular sieve granules (type 3A, point residues. Performance monitoring. The PGLC system was B.D.H. Ltd., Poole, U.K.) and used only once. Quartz tubes were replaced after every tenth pyrolysis. Base- monitored at intervals by analyzing samples from a line "noise" was sometimes seen if re-use of pyrolysis single batch of washed, dried, 100-mesh-powder bakwires was attempted or if quartz tubes were replaced ers' yeast (Distillers Co., Ltd., Bristol, U.K.), which less frequently. The quality of the base line was was used throughout the investigation. The quantitachecked from time to time by blank firings; "peak tive response of the separating system and its replicability were evaluated in terms of pyrolyzed standard ghosts" were not observed. Gas chromatography. The pyrolyzer delivered yeast samples of known weight. Numerical evaluation. Preliminary observations products from bacterial cells to a Pye series 104, model 64, gas chromatograph. All analyses were run on dual showed no advantage in determining peak areas in3-m glass columns (ID, 3 mm) containing acid-washed stead of peak heights. The typical chromatograph siliconized Chromosorb W support phase carrying 15% output (pyrogram) comprised 35 to 40 distinguishable Carbowax 20M (a polyethylene glycol); the columns peaks, some of which were complex. Heights of the were aged overnight at 230°C, and the operating col- selected 16 peaks, i.e., those shown to express differumn was brought to optimum performance within 1 ences between the streptococci of interest most effecor 2 weeks of commissioning. Pure nitrogen was used tively, were normalized. Values for these peaks, desas the carrier gas, at a flow rate of 20 cm:'/min. Both ignated A through P, provided the basis for taxonomic flame ionization detectors were supplied with 40 cm:' interpretations as well as identifications. At least three of hydrogen and 500 cm 3of air per min. The attenuator pyrograms were evaluated for each culture tested. If range was set at 1 x 10-9 or 2 x 10-" A for full-scale more than three were considered necessary, the set deflection; that for the chart recorder was 1 mV, with of three showing an acceptable level of reproducibility a chart speed of 25 mm/min. The separations were was taken to represent the data generated from the temperature-programmed, for an initial temperature culture. In the first instance, a set of three pyrograms, of 70°C up to 230°C at 12°C/min. Each cycle of represented by the 16 normalized peak values, was

VOL. 35, 1978

compared with another set by determining the mean ratio of heights of corresponding peaks. The lower value for each pair was taken as the numerator. Such ratios are termed similarity coefficients (S values) or coefficients of correspondence (10). When pyrogram data from strains of S. mitior, S. sanguis, and S. mutans were compared, it was noted that data for peak D alone were adequate to segregate nearly all strains examined. Less marked species differences were seen if other peaks were compared. A discriminant function was used for more efficient segregation of pyrogram data for ten strains of S. mitior, including three reference strains, and ten strains of S. sanguis, including five reference strains. The function was based upon three peaks (D, F, and K), with peak heights expressed as percentages of the total for the 16 selected peaks. Multiplying factors (d, f, and k) were determined, such that the resultant expression (dD + fF + kK) provided the maximum discrimination between the two species. The values for the factors (d, 4.92; f, 1.94; k, -1.64) were calculated from a set of simultaneous equations. Ranges of the two groups of function values did not overlap and, with a few exceptions, further isolates thought to be one or other of the above species could also be identified by this discriminant function. Moreover, it was possible to use the same function to segregate the ten strains of S. mutans received from other laboratories.

RESULTS Standardization of conditions. Weighed yeast samples (0.5 to 1.7 mg) were pyrolyzed, and the heights of the group of 16 selected peaks (denoted A through P in Table 1 and Fig. 24) were summed. These totals were proportional to the sample weights (Fig. 1). The peak heights were normalized for each of the five programs, and the individual means and standard deviations were calculated (Table 2). Individual peak height differences were 8% when they were normalized after repeated analyses of a single culture. At least 60 cultures of the strains listed in Table 1 were analyzed on the fourth chromatographic column. A somewhat greater peak height difference of 10% was typical when several parameters were investigated, using BHI broth and single strains for each parameter: (i) comparison of cultures of the same streptococcal strain on different occasions, (ii) comparison of a culture early and late in its growth cycle, (iii) comparison of cultures of an organism grown on consecutive batches of BHI broth, (iv) comparison of cultures grown in the presence and absence of oxygen (v) comparison of cultures grown in BHI broth and on BHI agar, and (vi) comparison of analyses of a single culture examined by using a mature chromatographic column immediately after testing its designated successor. Discrimination between oral streptococci. The major requirement of this study was to facilitate rapid discrimination of streptococci

47

PGLC OF ORAL STREPTOCOCCI

mg of yeast pyrolysed .K0

00

Total response, summation of peak heights mm. 1000

0

300

FIG. 1. Summations of heights of 16 selected pyrogram peaks in relation to the weight of the reference yeast sample before pyrolysis.

TABLE 2. Means and standard deviations (SD) of the selected peak heights (normalized) in pyrograms of five reference yeast samples (0.5 to 1. 7 mg) Peak

Percent of total

SD

Peak

Percent of total

A B C D E F G H

15.5 17.0

0.3 0.6 0.3 0.4 0.6 0.3 0.5 0.4

I J K L M N 0 P

1.6 2.5 8.2 2.5 9.2 4.7 1.7 5.8

e

4.4

2.9 4.5 6.9 6.5 5.8

SD

S 0.1 0.1 0.5 0.5 0.4 0.4 0.2 0.4

such as S. sanguis and S. mitior and to assist with the classification of strains having properties intermediate between those characteristic of existing species. This became possible after attention had been given to the controftof various parameters as described in the previous section. Figure 2 shows typical pyrograms for the three species mentioned above. Peak D differed most obviously in these examples, but other peaks (F and K) were of considerable value in distinguishing between strains of these species. The linear discriminant function described above also yielded data required for applying Hotelling's T2 test; this was found to indicate statistical significance for differences between the strains representing these species and also for their comparison with strains of S. mutans.

The mean function values, and their standard deviations, for 35 strains of the three species studied are listed in Table 3. The much greater range of these values for the S. mutans strains as compared with S. sanguis strains (Table 4) was consistent with the greater heterogeneity of this species. It was thought likely that several locally isolated strains were similar to S. mitior or S. sanguis, according to bacteriological tests. Discriminant function values for three of them were between 2 and 3, suggesting a closer relationship to S. mitior strains already examined. A value of 12.5 was found for each of the two other

48

Appi,. ENVIRON. MICROBIOL.

STACK, DONOGHUE, AND TYLER

ducibility of the Carbowax 20M columns was not expected to be high since, at the peak of every thermal cycle, the polymer was exposed to the recommended upper limit of temperature. However, the discriminant function calculated to be the most effective for the S. mitior and S. sanguis strains was also found to demonstrate good column reproducibility. The function values for the S. mutans neotype strain (NCTC 10449), for example, were 46.5, 54.5, and 46.5 for young, normal, and old cultures grown early in the year during which the third pair of columns was operated. At the end of this year, a value of 49.5 was noted, and values of 53 and 52.5 were calculated for later cultures analyzed using the fourth and fifth column pairs; the fourth column pair was in use during a 2-year period. Table 5 shows values for the discriminant function applied to some of the strains analyzed on columns IV and V. The mean values matched adequately with one exception, the variation being one-tenth of the mean.

mV

DISCUSSION The continuous numerical data generated by PGLC can be interpreted by the methods of

mV

TABLE 4. Values for a linear discriminant function applied to pyrogram data (peaks D, F, and K) for S. sanguis and S. mutans strains (see Table 1) Function value

FIG. 2. Pyrograms of strains of S. mitior, S. mutans, and S. sanguis; the selected peaks are denoted A through P, peak D being the most distinctive.

TABLE 3. Values for a linear discriminant function applied to pyrogram data (peaks D, F, and K) for S. mitior, S. sanguis, and S. mutans strains: means and standard deviations (SD) Species

No. of strains"

Mean (SD)

15 -2 (2) S. mitior 10 18 (3) S. sanguis 10 47 (11) S. mutans a Three, five, and ten strains, respectively, received from other laboratories.

strains, indicating a closer relationship to S. sanguis. More effective differentiation of these and other intermediate strains was demonstrated by using canonical variates analysis. Comparison of data from two columns. When organisms were to be compared over a period of several years, it was considered inevitable that even a rejuvenated chromatographic column would require replacement. The repro-

.

S. mutans 16 AHT 25 BHTa 22 FA-1 13 10449 17 Ingbritt 21 C67-1 17 C67-25 17 C180-2 Kl 19 16 LM7 a Received from D. B. Drucker.

S. sanguis JENA 2697 T175-1 L/SANG GPB1 GSG2 E M6 M9 84 119

Function value

36.5 60 60 52.5 32.5 54 52 55 30.5 39

TABLE 5. Comparison ofpyrogram data from two Carbowax 20M columns: values for a linear discriminant function (peaks D, F, and K) Discriminant function value

Organism

S. mutans BHTa S. mutans AHT

S. sanguis T175-1 S. sanguis 84 S. mitior S5723

Column IV 60 36.5

Column V

25

25.5

19 0 -2 S. mitior 91 a Received from D. B. Drucker.

62.5 34 27 2 -4

VOL. 35, 1978

numerical taxonomy. A disadvantage may arise if zero values are encountered. The study of oral streptococci by PGLC, however, has led to the conclusion that, for the strains investigated, values for peak heights are always finite. Nevertheless, the peak height values sometimes represent more than one product, and the identities of the components are unknown. The pyrograms are thus considered as fingerprints. Other studies of streptococci, in which fatty acid profiles have been explored, have shown that these organisms are particularly stable in composition (2, 3). In the present study, however, it was considered necessary to monitor performance when a new batch of medium was used, and this led to the rejection of one batch. It was possible to attain the required level of reproducibility when organisms were washed off BHI agar slopes as an alternative to harvesting organisms grown in BHI broth, but the risk of contamination by agar did not appear to be warranted. Reproducible pyrograms have, however, been obtained from cells that were harvested from membrane filters without washing (12). The demonstration that intercolumn comparisons are feasible is of great importance, since lack of comparability has been considered to be one of the main objections to the use of PGLC (1, 18). Successful rejuvenation of a column of established performance by periodic replacement of the initial portion of the column packing material, which accumulates high-boiling tarry residues, has also been of great value. It has thus been possible to complete a long-term investigation by using the same column throughout, as well as compare results from several columns. In our view, other investigators have sometimes placed too much reliance upon visual comparisons of pyrograms, and data have not always been examined rigorously enough. It is therefore not possible to decide whether parameters were adequately controlled. In some instances (e.g., 5) this may have led to the conclusion that the technique could not be recommended for successful differentiation of strains. The importance of adequate numerical treatment of PGLC data has been emphasized in recent years (4, 10). The discriminant function utilized is not considered to provide adequate numerical treatment for more than the limited purpose outlined in this study. This function, designed for discriminating between S. mitior and S. sanguis strains, although also found to be effective in discriminating S. mutans strains, was not effective in discriminating between S. sanguis strains and a few strains of two other streptococcal species

PGLC OF ORAL STREPTOCOCCI

49

residing in the mouth. Other, empirical functions were designed such that the resultant values for the three S. salivarius and two S. milleri strains differed by at least 4 standard deviations from the sets of values for S. sanguis strains. By using similarity coefficients (20) as well as discriminants, the degree of relatedness between previously encountered oral streptococci and unknown strains can be achieved rapidly, and our results confirm the value of PGLC for this purpose. ACKNOWLEDGMENTIS Thanks are due to Margaret Marshall and Gillian Rapson for skilled and consistent technical assistance, to colleagues mentioned in Table 1 who provided reference strains, and to Colin Gutteridge, of the Agricultural Research Council Meat Research Institute, Langford, Somerset, for canonical variates analysis of pyrogram data.

LITERATURE CITED 1. Carmichael, J. W., A. S. Sekhon, and L. Sigler. 1973. Classification of some dermatophytes by pyrolysis-gasliquid chromatography. Can. J. Microbiol. 19:403-407. 2. Drucker, D. B., C. J. Griffith, and T. H. Melville. 1972. Effect of changes in growth on the fatty acid profiles of Streptococcus mutans. J. Dent. Res. 51:1276. 3. Drucker, D. B., C. J. Griffith, and T. H. Melville. 1974. Fatty acid fingerprints of streptococci: variability due to carbohydrate source. Microbios 9:187-189. 4. Eshuis, W., P. G. Kistemaker, and H. L. C. Meuzelaar. 1977. Some numerical aspects of reproducibility and specificity, p. 151-166. In C. E. R. Jones and C. A. Cramers (ed.), Proceedings of the Third International Symposium on Analytical Pyrolysis, Amsterdam, 1976. Elsevier, Amsterdam. 5. Fontanges, R., G. Blandenet, and R. Queignec. 1967. Difficultes d'application de la chromatographie en phase gazeuse a l'identification des bacteries. Ann. Inst. Pasteur Paris 112:10-23. 6. Haddadin, J. M., R. M. Stirland, N. W. Preston, and P. Collard. 1973. Identification of Vibrio cholerae by pyrolysis gas-liquid chromatography. Appl. Microbiol.

25:40-43. 7. Hardie, J. M., and G. H. Bowden. 1976. Physiological

classification of oral viridans streptococci. J. Dent. Res. 55 (Spec. Issue A):166-176. 8. Holmberg, K., and C.-E. Nord. 1975. Numerical taxonomy and laboratory identification of Actinomyces and Arachnia and some related bacteria. J. Gen. Microbiol.

91:17-44. 9. Huis in't Veld, J. H. J., H. L. C. Meuzelaar, and A. Tom. 1973. Analysis of streptococcal cell wall fractions

by curie-point pyrolysis gas-liquid chromatography. Appl. Microbiol. 26:92-97. 10. Meuzelaar, H. L. C., P. G. Kistemaker, and A. Tom. 1975. Rapid and automated identification of bacterial strains by fully automated curie point pyrolysis gasliquid chromatography, p. 165-178. In C.-G. Heden and T. Illeni (ed.), New approaches to the identification of microorganisms. John Wiley, New York. 11. Needleman, M., and R. Stuchberry. 1977. The identification of microorganisms by pyrolysis-gas-liquid chromatography, p. 77488. In Jones and Cramers (ed.), Proceedings of the Third International Symposium on Analytical Pyrolysis, Amsterdam, 1976. Elsevier, Amsterdam. 12. Oxborrow, G. S., N. D. Fields, and J. R. Puleo. 1976.

50

13.

14.

15. 16. 17.

APPT. ENVIRON. MICROBIIOL.

STACK, DONOGHUE, AND TYLER Preparation of pure microbiological samples for pyrolysis-gas-liquid chromatography studies. Appl. Environ. Microbiol. 32:306-309. Oxborrow, G. S., N. D. Fields, and J. R. Puleo. 1977. Pyrolysis gas-liquid chromatography studies of the genus Bacillus. Effect of growth time on pyrochromatogram reproducibility, p. 69-76. In C. E. R. Jones and C. A. Cramers (ed.), Proceedings of the Third International Symposium on Analytical Pyrolysis, Amsterdam, 1976. Elsevier, Amsterdam. Oxborrow, G. S., N. D. Fields, and J. R. Puleo. 1977. Pyrolysis gas-liquid chromatography studies of the genus Bacillus. Effect of growth media on pyrochromatogram reproducibility. Appl. Environ. Microbiol. 33:865-870. Quinn, P. 1974. Development of high resolution pyrolysisgas chromatography for the identification of microorganisms. J. Chromatogr. Sci. 12:796-806. Reiner, E. 1965. Identification of bacterial strains by pyrolysis-gas-liquid chromatography. Nature (London) 206:1272-1274. Reiner, E. 1977. The role of pyrolysis-gas liquid chromatography in biomedical studies, p. 49-56. In C. E. R.

18.

19.

20.

21.

Jones and C. A. Cramers (ed.), Proceedings of the Third International Symposium on Analytical Pyrolysis, Amsterdam, 1976. Elsevier, Amsterdam. Sekhon, A. S., and J. W. Carmichael. 1973. Column variation affecting a pyrolysis-gas-liquid chromatographic study of strain variation in two species of Nannizzia. Can. J. Microbiol. 19:409-411. Stack, M. V. 1968. A review of pyrolysis gas chromatography of biological macromolecules, p. 109-118. In Proceedings of the Seventh International Symposium on Gas Chromatography and Its Exploitation, Copenhagen, 1968. Institute of Petroleum, London. Stack, M. V., H. D. Donoghue, J. E. Tyler, and M. Marshall. 1977. Comparison of oral streptococci by pyrolysis GLC, p. 57-68. In C. E. R. Jones and C. A. Cramers (ed.), Proceedings of the Third International Symposium on Analytical Pyrolysis, Amsterdam, 1976. Elsevier, Amsterdam. Vincent, P. G., and M. M. Kulik. 1973. Pyrolysis-gasliquid chromatography of fungi: numerical characterization of species variation among members of the Aspergillus glaucus group. Mycopathol. Mycol. Appl. 51:251-265.