Evaluation of the Enteric-Tek System for Identifying Enterobacteriaceae

Transcription

1 JOURNAL OF CLINICAL MICROBIOLOGY, Mar. 1982, p Vol. 15, No /82/ $02.00/0 Evaluation of the Enteric-Tek System for Identifying Enterobacteriaceae A. 0. ESAIAS,* D. L. RHODEN, AND P. B. SMITH Center for Infectious Diseases, Centers for Disease Control, Atlanta, Georgia Received 26 June 1981/Accepted 5 October 1981 The Enteric-Tek wheel (Flow Laboratories), consisting of 14 different biochemical parameters for rapidly identifying Enterobacteriaceae, was evaluated and compared with the conventional method for completely identifying 301 enteric cultures, representing 36 species. The Enteric-Tek system correctly identified 264 (97.8%) of the 270 common or typical strains and 26 (83.9%) of the 31 unusual or atypical strains tested, demonstrating an overall identification accuracy rate of 96.3%. There were 80 (26.6%) correctly identified strains requiring additional tests. Of the 11 (3.6%) misidentifications, 5 (3 Klebsiella and 2 Salmonella strains) were correctly identified at the genus level. When 4,228 individual tests in the Enteric-Tek wheel were compared with the conventional tubed media, 96.4% of the tests agreed; urease, citrate, adonitol, and lactose agreed less than 97%. The Enteric-Tek system was found to be reliable and accurate in producing identifications at the genus and species level within 18 to 24 h. Since a substantial proportion of the workload generated in the clinical microbiology laboratory involves isolating and identifying Enterobacteriaceae, the development of rapid identification systems for enteric bacteria has increased, as evidenced by the numerous commercially available kits (2, 14). Many of these systems have been studied and evaluated by various investigators. A new commercial identification system, the Enteric-Tek wheel (Flow Laboratories, Roslyn, N.Y.), is designed to identify the Enterobacteriaceae at the genus and species level within 18 to 24 h of isolation on primary plating media. By the use of 14 different biochemical parameters, the system generates a five-digit octal profile number derived from the biochemical reactions. In our study, interpretation of the reactions was made according to the manufacturer's recommendations so an identification could be derived from the computer code book. The purpose of this evaluation was to determine the ability of the Enteric-Tek system to identify both typical and atypical members of Enterobacteriaceae. This study presents data which describe the performance and accuracy of the Enteric-Tek wheel as compared with a conventional identification system. MATERIALS AND METHODS Bacterial cultures tested. We tested 301 of our stock cultures from the General Bacteriology Laboratory, Centers for Disease Control, culture collection. These cultures had been assigned code numbers by a third party. The selected cultures, well distributed among 36 species, included 270 common or typical strains and 31 unusual or atypical strains. All cultures were maintained in sealed Trypticase (BBL Microbiology Systems, Cockeysville, Md.) soy agar stabs and stored at rodm temperature in the dark. Each culture had previously been identified by conventional procedures (5, 8-11). The identity of each culture remained unknown until all results were compiled and the testing was completed. Conventional method for identification. Each culture was streaked onto a MacConkey agar plate and a blood agar plate. Cultures were then identified by conventional biochemical tests (5, 8-10). The conventional tests routinely performed included reactions on triple sugar iron agar (H2S production); Christensen urea agar; indole; methyl red; Voges-Proskauer (VP) medium; citrate; lysine and omithine decarboxylase; arginine dihydrolase; motility; phenylalanine; malonate; o-nitrophenyl-o-d-galactopyranoside; and production of acid from glucose, lactose, sorbitol, dulcitol, inositol, adonitol, mannitol, sucrose, salicin, arabinose, raffinose, and rhamnose. These media were inoculated from culture suspensions in tryptone broth and were incubated at 35C. After 18 to 24 h of incubation, reagents were added for indole, methyl red, VP (acetoin), and phenylalanine tests. Tests which were not positive within 18 to 24 h were observed for a maximum of 7 days. When required, additional tests were performed to complete the identification, e.g., growth in KCN and serological confirmation of Salmonella and Shigella. All cultures were identified by the nomenclature and taxonomy described by Edwards and Ewing (8, 9) and Brenner et al. (5). When necessary, the Enterobacteriology Section, CDC, was consulted as a reference laboratory. Enteric-Tek system. The Enteric-Tek system is a round, multicompartment wheel, consisting of a central well and 11 individual peripheral wells, all of which contain solid media. The system provides for 419

2 420 ESAIAS, RHODEN, AND SMITH determining 14 different biochemical parameters which include: indole production; tryptophan deaminase; H2S production; citrate; malonate; lysine and ornithine decarboxylase; urease; and acid production from glucose, lactose, rhamnose, adonitol, sorbitol, and arabinose. All organisms tested were inoculated onto MacConkey agar and blood agar plates and incubated for 18 to 24 h at 35 C. A suspension from several well-isolated colonies of each organism was made in 2 ml of sterile distilled water. Before inoculation, the wheels were allowed to warm to room temperature and were labeled appropriately. Inoculation was done as described in the detailed instructions provided by the manufacturer. Each wheel was placed in an upright position, incubated at 35 C for 18 to 24 h, then removed from the incubator, and observed for color changes in the solid media. The only manipulation required to observe a biochemical reaction involved rolling a cotton swab saturated with a special indole reagent (Flow Laboratories) over growth in the center well. Any redness developing within 30 s indicated a positive indole reaction. The remainder of the wells were visually observed, and their color was read, as suggested by the manufacturer. All reactions were recorded on a printed form and tabulated into a fivedigit profile number for identifying organisms listed in the manufacturer's code book. The data incorporated into the computer code book were generated from the percentage charts of Edwards and Ewing (8) and from in-house studies. Identification was based on the probability that a set of biochemical reactions would occur with more likelihood for one particular organism than for another and allowed for the possibility of atypical reactions. The code book lists additional biochemical tests required to complete an identification. The additional tests include: arginine, cellobiose, DNase, esculin, gas from glucose, inositol, Jordan tartrate, potassium cyanide, maltose, mannitol, methyl red, motility, raffinose, VP, xylose, and serology for Salmonella and Shigella. RESULTS Agreement of tests. Biochemical reactions for the various tests in the Enteric-Tek system were compared with those obtained by the conventional method. Of the 14 common tests, 10 showed more than 97% agreement with their conventional counterparts, whereas urease, citrate, adonitol, and lactose agreed 83.4, 91.4, 94.0, and 94.4%, respectively. These four tests comprised 75% of all of the test discrepancies. Certain groups of organisms were responsible for lowering the agreement for some of the tests. The 50 differences in the urease test, all false-positive in the Enteric-Tek system except for 1, occurred with 6 Citrobacter freundii, 10 Citrobacter diversus, 14 Citrobacter amalonaticus, 4 Klebsiella pneumoniae, 4 Klebsiella oxytoca, 8 Enterobacter cloacae, 1 Hafnia alvei, 1 Morganella morganii, and 1 Yersinia enterocolitica strain. One Proteus rettgeri strain gave a false-negative reaction despite a 4+ reaction in the conventional Christensen urea agar slant. The 26 citrate test discrepancies included J. CLIN. MICROBIOL. 22 false-negatives and 4 false-positives but were not responsible for any misidentifications. Of the 18 differences in the adonitol test, all of the 11 strains of Serratia marcescens, 1 Serratia liquefaciens, 1 K. pneumoniae, 1 Shigella flexneri, 2 Yersinia pseudotuberculosis, and 1 Y. enterocolitica strain were false-positive, but only 2 of the latter 3 cultures were misidentified. One false-negative adonitol reaction occurred with a Providencia alcalifaciens strain. The 17 discrepancies in lactose were divided almost equally between the false-negative (9) and falsepositive (8) reactions. Differences in the lactose results did not change the identifications. Identification. Overall, the Enteric-Tek system correctly identified, at the genus and species level, 96.3% of the 301 strains tested. As shown in Table 1, the system correctly identified 264 (97.8%) of the typical strains and 26 (83.9%) of the atypical strains. Of the 290 cultures correctly identified, 62 of 270 typical strains (23.0%) and 18 of 31 atypical strains (58.1%) required additional tests. The number of additional tests ranged from one test for 7 strains, two tests for 42 strains, and three tests for 30 strains, to five tests for 1 atypical Escherichia coli strain. We performed the suggested additional tests when an identification was given with less than 95% probability. Although correct identifications for one Arizona hinshawii, one S. flexneri, two Shigella boydii, and one Y. enterocolitica were listed as first-choice identifications, serological confirmation was required because the given probabilities were less than 95%. Twenty-one strains required serological confirmation because the assigned identifications were listed as second, third, and fourth choices in the code book. The identification accuracy of individual species by the Enteric-Tek system is shown in Table 2. The system provided 252 (86.9%) of the correct identifications as the first choice. There were 38 identifications listed as second, third, or fourth choices (29 as second, 6 as third, and 3 as fourth). The Enteric-Tek system correctly identified 100% of the strains for 29 of 36 species (80.6%) tested. The number of strains tested for each species ranged from 2 to 14, except for 25 E. coli strains, which included H2S-positive and atypical strains. Two atypical strains of both E. coli and C. freundii were misidentified, decreasing their identification rates from 100% to 92 and 82%, respectively. Four strains each of Salmonella paratyphi A, Salmonella typhi, and Y. enterocolitica were tested; one strain of each was misidentified, resulting in a 75% identification rate for each species. The misidentified strain of S. typhi was atypical. Both Salmonella strains were identified as S. enteritidis, listed as the second choice in the manufacturer's comput-

3 VOL. 15, 1982 EVALUATION OF THE ENTERIC-TEK SYSTEM 421 TABLE 1. Summary of identifications of unknown cultures with the Enteric-Tek system ~~~~~~~~~~~~~~additional Organism categoryno. No. tested No. correct No. correct requiring tests' Common or typical (97.8)b 62 (23.0)c Unusual or atypical (83.9) 18 (58.1) Total (96.3) 80 (26.6) a Including Salmonella and Shigella serology when required. b Number in parentheses gives the percentage for the category. c Number in parentheses indicates the percentage of the number correct in the category. er code book. Klebsiella ozaenae, with nine strains tested, showed the lowest identification rate (66.6%), because three strains did not meet the expected species level of identification, as claimed by the manufacturer. DISCUSSION In comparison with the conventional methods, the Enteric-Tek system demonstrated a highly acceptable level of identification accuracy for the 301 stock cultures tested. The overall identification rate of the Enteric-Tek system (96.3%) compares favorably with the Micromedia' System (97%) and Entero-Set 20 (96%) and is better than the Micro-ID (94%), API 20E (92%), and Enterotube (84%) systems when each system is compared with conventional methods (2, 14). Another study reported a higher identification rate (97%) for both the API 20E and Micro-ID systems (3). Thirty-six enteric species were tested with the Enteric-Tek wheel, as compared with 21 to 28 enteric species tested with the other rapid identification kits. For identifying a large variety of enteric bacterial strains to the species level, the Enteric-Tek system shows potential advantages over some of the rapid systems previously evaluated (2-4, 6, 14). The Enteric-Tek system can correctly identify Serratia strains as to species, whereas the API 20E system has frequently been reported to identify them at the genus level only (4, 14). Additionally, a test for adonitol fermentation is included in this system, as in the Micro-ID and Entero-Set 20 but not in API 20E, allowing for the differentiation of H2S-negative C. freundii from C. diversus without the need for additional tests (4). Unlike the Micro-ID data base, the Enteric-Tek system includes identification for Enterobacter gergoviae and C. amalonaticus strains (6). Additional species of varied levels of difficulty were used in this study, and the Enteric-Tek system correctly identified all strains for 29 of 36 species (81%) tested. By comparison, 20 enteric species were tested against the following rapid systems, with the indicated number of species having all of their strains correctly identified: the Micro-Media system with 17 (85%), Micro-ID with 13 (65%), Enterotube with 12 (60%), and API 20E with 11 (55%) (14). The scope of the Enteric-Tek system data base was demonstrated by the 185 different profile numbers generated in deriving the identifications, with a range of 1 to 9 different profile codes for each species tested. The Enteric-Tek system showed the ability to identify two new Serratia species, S. fonticola and S. odorifera. This ability was clearly shown when the one S. odorifera strain tested was correctly identified as a first-choice identification, having a 87.71% probability. Also, with 8 of 10 correctly identified Enterobacter aerogenes strains, the Enteric-Tek system listed S. fonticola, E. aerogenes, and S. odorifera, in that order, as first-, second-, and third-choice identifications. E. aerogenes was given as a secondchoice identification, showing a 47.07% probability, as compared with S. fonticola (52.91%) and S. odorifera (0.02%). In both of these situations, two additional tests, VP and DNase, were suggested by the manufacturer to easily differentiate these species. All S. fonticola strains are listed as VP negative, whereas all E. aerogenes strains are VP positive. All strains for both S. fonticola and E. aerogenes have been shown to be DNase negative, whereas all S. odorifera strains are DNase positive. Strains of S. fonticola have been isolated from water samples and reported as Citrobacter-like bacteria with lysine decarboxylase production (12). Researchers have reported that with commercial identification systems, S. odorifera may look like atypical (gelatin-positive, anaerogenic) E. aerogenes. S. odorifera strains are nonpigmented and have a characteristic odor that resembles the smell of vegetable matter, which is helpful in making a correct identification. Strains of S. odorifera have been recovered from clinical specimens and may be clinically significant because of the strains studied (13). The reasons for the 11 (3.6%) erroneous identifications (Table 3) were as follows: 3 (27.3%) resulted from reaction discrepancies, 5 (45.4%) displayed aberrant biochemical patterns, and 3 (27.3%) required insufficient additional tests. As shown in Table 4, the three reaction discrepancies in the common tests consisted of one false-

4 422 ESAIAS, RHODEN, AND SMITH TABLE 2. Accuracy of identification by the Enteric-Tek system with unknown enteric cultures Organism No. correct/no. tested % Correct Proteus vulgaris Proteus mirabilis (3)' Morganella morganii (1) Providencia alcalifaciens Providencia stuartii (3) Providencia rettgeri Edwardsiella tarda (2) Salmonella enteritidis Salmonella, atypical (5) Salmonella paratyphi A Salmonella typhi (1) Salmonella cholerae-suis Arizona hinshawii (1) Citrobacter freundii Citrobacterfreundii, atypical (2) Citrobacter diversus Citrobacter amalonaticus (1) Shigella dysenteriae Shigella flexneri Shigella boydii Shigella sonnei Escherichia coli Escherichia coli, H2S positive Escherichia coli, atypical (10) Yersinia enterocolitica Yersinia pseudotuberculosis Klebsiella pneumoniae (1) Klebsiella oxytoca Klebsiella ozaenae Klebsiella rhinoschleromatis Enterobacter cloacae Enterobacter sakazakii Enterobacter aerogenes (1) Enterobacter gergoviae Enterobacter agglomerans Hafnia alvei Serratia liquefaciens Serratia marcescens Serratia rubideae Serratia odorifera 12/12 8/8 12/12 3/4 3/4 4/4 11/11 0/2 14/14 3/3 2/2 8/10 4/5 3/4 6/9 6/6 2/2 8/8 11/11 1/1 a Number in parentheses indicates the number of atypical strains tested. positive adonitol; one false-positive adonitol, lactose, and sorbitol; and one false-positive lysine and false-negative sorbitol strain. For three cultures of K. ozaenae, differentiation required additional conventional biochemical tests beyond the two tests, gas and motility, shown in the code book. These cultures were incorrectly identified by the Enteric-Tek system as Klebsiella rhinoschleromatis, the first-choice identification. K. ozaenae was given as the second choice on the basis of a negative reaction for gas production. The problem might be eliminated by including the additional tests, esculin and gas from cellobiose, in their data base. A positive esculin reaction occurs in 75% of K. ozaenae strains, as compared with 15.4% of K. rhinoschleromatis strains. A positive reaction with gas from cellobiose has been shown to occur in 70% of K. ozaenae strains, whereas all K. rhinosch- J. CLIN. MICROBIOL leromatis strains show a negative gas reaction (11). Five atypical strains with aberrant reaction patterns were misidentified. The organisms involved were two atypical E. coli strains, which were indole and lactose negative; two atypical C. freundii strains (one H2S-negative strain and one H2S-negative indole-positive strain); and one atypical ornithine-positive and H2S-negative S. typhi strain. The atypical S. typhi strain was incorrectly identified as S. enteritidis, as is suggested by percentage charts supplied by the manufacturer (Flow Laboratories) showing a 100% negative ornithine reaction. These charts also show a 4% probability for E. coli strains to have negative indole and lactose reactions and a 10% probability for H2S-negative and indolepositive C. freundii strains. Misidentification of these atypical E. coli and C. freundii strains by

5 VOL. 15, 1982 EVALUATION OF THE ENTERIC-TEK SYSTEM 423 TABLE 3. Misidentifications of unknown cultures by the Enteric-Tek system Reason No. % Misidentifications cu Total cultures Biochemical reaction discrepancy Atypical pattern Insufficient biochemical tests indicated Total the Enteric-Tek system might be eliminated if these reactions were included in the data base. Of the 11 misidentifications, 6 were of typical strains and 5 were of atypical strains (Table 4). Of the 11 misidentifications, 10 required supplemental tests, including 5 serological confirmations for 1 S. paratyphi A and 1 atypical S. typhi strain and Shigella serology for 1 Y. enterocolitica and 2 atypical E. coli strains, because Shigella was listed as first, second, and third choices. The two Salmonella strains, identified as S. enteritidis, did not meet expected species level of identification as claimed by the manufacturer. However, had these two Salmonella been identified only to genus, with subsequent serology required, this would have increased the overall identification rate to 97.0%. Identifications for the two atypical E. coli strains were listed as Shigella for all three choices for one strain and for the first three choices of the other strain, with Y. enterocolitica given as a fourth choice. Shigella serology was required for the one misidentified Y. enterocolitica strain because Shigella was given as the third-choice identification. The Y. enterocolitica strain, misidentified by the Enteric-Tek system as K. rhinoschleromatis, presented an interesting situation. The four identification choices and their probability percentages were listed in this order: K. ozaenae (75.56%), K. rhinoschleromatis (21.19%), S. flexneri (3.06%), and Y. enterocolitica (0.05%). S. flexneri could be eliminated as a possible identification, on the basis of serology. We did the three additional tests-gas, KCN, and raffinose-as suggested by the manufacturer, and all three gave a negative reaction. Since K. ozaenae showed a greater probability for positive reactions in all three tests, we eliminated this choice as a possible identification. On the basis of a greater probability percentage given for K. rhinoschleromatis (25.9%), we decided on this choice instead of Y. enterocolitica (0.05%) as the identification. The three negative test reactions favor the identification as Y. enterocolitica rather than K. rhinoschleromatis. This situation illustrates the potential for misidentifying an organism if the probability percentage given for each choice identification is the sole criterion. This example emphasizes the need to consider equally all given information and use colony characteristics before making an identification. Had the manufacturer suggested a motility test at 25 C, a correct identification would have been made. At the time our study was done, we were unaware that the manufacturer could provide identification probability percentages based on Enteric-Tek, plus additional tests. K. ozaenae presented the greatest challenge to the system, with three misidentifications causing 27.3% of all misidentifications. The remaining erroneous identifications involved two species, with two misidentifications each, and four species with only one misidentification each. The analysis of the 4,228 individual test comparisons revealed the Enteric-Tek System to be very sensitive and specific, showing high test TABLE 4. Identification errors made by the Enteric-Tek system Organism Enteric-Tek identification Reason for misidentification S. paratyphi A (1)' S. enteritidis Lysine positive and sorbitol positiveb S. typhi (1)C S. enteritidis Ornithine positive and H2S negative C. freundii (2)c E. agglomerans H2S negative E. coli H2S negative and indole positive E. coli (2)c Y. enterocolitica Indole negative and lactose negative Unidentified Indole negative and lactose negative Y. enterocolitica (1) K. rhinoschleromatis Adonitol positive b Y. pseudotuberculosis (1) K. rhinoschleromatis Adonitol, lactose, and sorbitol positiveb K. ozaenae (3) K. rhinoschleromatis (3) Gas and motilityd a Number in parentheses indicates the number of cultures. b Biochemical reaction discrepancy. c Atypical strain. d Insufficient biochemical test indicated.

6 424 ESAIAS, RHODEN, AND SMITH correlations when compared with other rapid identification systems. For comparison, an analysis of the common test reactions for several rapid systems showed the following results for the number of tests with a 97% or better agreement, when each was compared with the conventional counterpart: Enteric-Tek with 10 of 14 (71.4%), Micro-ID with 6 of 11 (54.5%), API 20E with 6 of 15 (40%), and the Entero-Set 20 system with 6 of 16 tests (37.5%) (1, 2). The comparison showed that urease, citrate, arabinose, adonitol, and malonate differed substantially in their agreement percentages. The urease test agreement was poorer in the Enteric-Tek (83.4%) and Entero-Set 20 (95%) systems (1-4). The highly sensitive urease test in the Enteric-Tek system did not cause any misidentifications, whereas the urease test in the Micro-ID system has caused misidentifications because of its lack of sensitivity and specificity (3). The Enteric-Tek system performed better in the citrate test with 91.4% agreement, as compared with the Entero- Set 20 (86%) and API 20E (83 to 86%) systems (1, 2). The Entero-Set 20 showed poorer agreement for two additional tests, adonitol (86%) and malonate (89%), whereas the Enteric-Tek and Micro-ID each gave similar agreement percentages for adonitol (94%) and malonate (99%) (1, 2, 4). A low agreement percentage for the arabinose test (39 to 82%) in API 20E has been reported, as compared with 99 to 100% in Enteric-Tek, Micro-ID, and Entero-Set 20 (1, 2). It must be noted that such comparisons do have inherent weaknesses, as pointed out by Edberg et al. (7). A rapid and accurate identification system for the Enterobacteriaceae is highly desirable in the clinical laboratory. Such a system would prove beneficial in directing antimicrobic therapy, as well as increasing identification capabilities without additional demand on trained personnel. We found the Enteric-Tek system to be accurate and easy to use. The advantages included easy inoculation, minimal manipulation, minimal growth for sufficient inoculum, useful listings of supplemental tests and probability percentages, and easy reading of color changes in the media. The major limitations involved distinguishing between weakly positive and negative H2S reactions and reading the indole test with weakly positive strains. Although the Enteric-Tek wheels are stackable, their size and shape may require more storage room, as compared with some of the other rapid kits. Overall, we found that the Enteric-Tek system provided a highly acceptable level of identification for the Enterobacteriaceae within 18 to 24 h and served as an alternative procedure to the conventional method. Because this study was designed only to determine the ability of the J. CLIN. MICROBIOL. Enteric-Tek system to identify the Enterobacteriaceae, additional testing will be necessary in evaluating its performance in a clinical setting. Since atypical strains present the greatest challenge to the clinical laboratory worker, the need for subjective evaluation, particularly of colony morphology, pigment, and odor, remains essential when an unusual organism is encountered. LITERATURE CITED 1. Aldridge, K. E., B. B. Gardner, S. J. Clark, and J. M. Matsen Comparison of Micro-ID, API 20E, and conventional media systems in identification of Enterobacteriaceae. J. Clin. Microbiol. 7: Aldridge, K. E., and R. L. Hodges Correlation studies of Entero-Set 20, API 20E, and conventional media systems for Enterobacteriaceae identification. J. Clin. Microbiol. 13: Barry, A. L., and R. E. Badal Rapid identification of Enterobacteriaceae with the Micro-ID system versus API 20E and conventional media. J. Clin. Microbiol. 10: Blazevic, D. J., D. L. MacKay, and N. M. Warwood Comparison of Micro-ID and API 20E systems for identification of Enterobacteriaceae. J. Clin. Microbiol. 9: Brenner, D. J., J. J. Farmer III, F. W. Hickman, M. A. Asbury, and A. G. Steigerwait Taxonomic and nomenclature changes in Enterobacteriaceae. Center for Disease Control, Atlanta, Ga. 6. Buesching, W. J., D. L. Rhoden, A. 0. Esaias, P. B. Smith, and J. A. Washington H Evaluation of the modified Micro-ID system for identification of Enterobacteriaceae. J. Clin. Microbiol. 10: Edberg, S. C., B. Atkinson, C. Chambers, M. H. Moore, L. Palumbo, C. F. Zorzon, and J. M. Singer Clinical evaluation of the Micro-ID, API 20E, and conventional media systems for identification of Enterobacteriaceae. J. Clin. Microbiol. 10: Edwards, P. R., and W. H. Ewing (ed.) Identification of Enterobacteriaceae, 3rd ed. Burgess Publishing Co., Minneapolis, Minn. 9. Ewing, W. H Differentiation of Enterobacteriaceae by biochemical reactions-reprinted. Center for Disease Control, Atlanta, Ga. 10. Ewing, W. H., and B. R. Davis Media and tests for differentiation of Enterobacteriaceae. U.S. Department of Health, Education and Welfare, National Communicable Disease Center, Atlanta, Ga. 11. Flfe, M. A., W. H. Ewing, and B. R. Davis The biochemical reactions of the tribe Klebsielleae-reprinted. Center for Disease Control, Atlanta, Ga. 12. Gavini, F., C. Ferragut, D. hard, P. A. Trinel, H. Le- Clerc, B. LeFebvre, and D. A. A. Mossel Serratia fonticola, a new species from water. Int. J. Syst. Bacteriol. 29: Grimont, P. A., F. Grimont, C. Richard, B. R. Davis, A. G. Stelgerwalt, and D. J. Brenner Deoxyribonucleic acid relatedness between Serratia plymuthica and other Serratia species, with a description of Serratia odorifera sp. nov. (type strain: ICPB 3995). Int. J. Syst. Bacteriol. 28: Kelly, M. T., and J. M. Latimer Comparison of the AutoMicrobic System with API 20E, Enterotube, Micro- ID, Micro-Media Systems, and conventional methods for identification of Enterobacteriaceae. J. Clin. Microbiol. 12: