Enterobacteriaceae Biochemical Card

Transcription

1 JOURNAL OF CLINICAL MICROBIOLOGY, June 1980, p /80/ /09$0.00/0 Vol. 11, No. 6 Collaborative Investigation of the AutoMicrobic System Enterobacteriaceae Biochemical Card HENRY D. ISENBERG,`* THOMAS L. GAVAN, PETER B. SMITH,3 ALEX SONNENWIRTH,4 W. TAYLOR,5 W. J. MARTIN,6 DWANE RHODEN,3 AND ALBERT BALOWS3 Long Island Jewish-Hillside Medical Center, New Hyde Park, New York 11041; Cleveland Clinic and Foundation, Cleveland, Ohio ; Center for Disease Control, Atlanta, Georgia ; Jewish Hospital of St. Louis, St. Louis, Missouri ; St. Mary's Hospital, Chicago, Illinois 6065; and UCLA Clinical Center, Los Angeles, California The Enterobacteriaceae biochemical card was used in six separate laboratories to identify 170 representatives of Enterobacteriaceae. The AutoMicrobic System (Vitek Systems, Inc.) performed with an accuracy of 97.8% as compared with 98.1% by the standard method selected and 97.6% by a commercially prepared manual system approach. During this time, 5,450 clinical isolates belonging to Enterobacteriaceae were analyzed. Compared with the routine methods used in the various laboratories, the AutoMicrobic System identified 96.4% correctly. The commercially produced systems or kit approaches to the biochemical recognition of clinically significant bacteria and yeasts have provided standardized, easy-to-use reagents to the staffs of many clinical laboratories who heretofore could not offer the advantages of such service (4) to clinicians in their institutions or communities. Compared with other clinical laboratory disciplines, the clinical microbiology service has remained labor-consuming despite the proliferation of these kits or systems, all of which involve manual manipulation and individual evaluation and interpretation. The Auto- Microbic System (AMS; Vitek Systems, Inc., subsidiary of McDonnell-Douglas, Hazelwood, Mo.), a fully automated approach to the enumeration and recognition of selected bacteria in clinical urine specimens and certain antibiograms of machine-sequestered organisms (1, 5, 8, 9), has been expanded in its application to clinical microbiology by the introduction of a disposable card designed to identify members of Enterobacteriaceae. A feasibility study of this approach has been reported (M. C. Meyer, J. J. Underwood, R. Wilkinson, and L. V. Woods, Abstr. Annu. Meet. Am. Soc. Microbiol., C14, p. 333, 1979; H. D. Isenberg, J. Scherer, and S. Freedman, Abstr. Annu. Meet. Am. Soc. Microbiol., C 143, p. 334, 1979), and a carefully designed collaborative evaluation of this fully automated approach has been conducted simultaneously with the application of the Enterobacteriaceae biochemical card (EBC) to the identification of members of the family isolated from any clinical specimen. This report is a summary of these latter investigations. MATERIALS AND METHODS Bacteria used. For the collaborative challenge to the AMS, 170 members of the family Enterobacteriaceae, in numbers indicated in Table 1, were selected by the collaborating laboratory of the Center for Disease Control (CDC), coded, and shipped to each of the participants. At the same time, during and after this period of the collaborative investigation, any member of the family Enterobacteriaceae isolated in the laboratories of the participants was also examined. Materials used. All microbiological reagents used, other than the material required for the automated system, were supplied to all participants by a single commercial media manufacturer. These included the primary isolation agars, the various substrates for the standard manual method of determining the biochemical reactions of all bacteria, and those reagents required for detecting final reactions. All the materials were prepared for all laboratories at the same time and represenited the same lot of media and chemicals. Two of the laboratories (H.D.I. and T.L.G.) examined the bacteria used in the collaborative study with the API system (Analytab Products, Inc., Plainview, N.Y.), referred to subsequently as the routine method. Since this product is a frequently used systems approach to the identification of Enterobacteriaceae, no attempt was made to standardize the lot. Inoculum preparation. The challenge bacteria used in the collaborative study were subcultured on the blood agar and MacConkey agar provided for the study. After 18 to 4 h of incubation, several colonies were selected from blood agar and suspended in 1.8 ml of 0.5% NaCl until McFarland no. 1 standard density was obtained. Clinical isolates, grown on each laboratory's particular blood agar, MacConkey agar, or eosin methylene blue agar, were used. The inoculum was prepared as described. Before transfer into the single-barreled card 694

2 VOL. 11, 1980 AMS ENTEROBACTERIACEAE BIOCHEMICAL CARD 695 TABLE 1. Collaborative comparison ofams with standard and routine manual methods Bacterium' No. of strains No. identified correctly" submitted/ Total no. tested laboratory AMS Standard Routine' Arizona hinshawii (Salmonella arizonae) 1 11 (91.7) 1 (100) Citrobacter amalonaticus 1 10 (83.3) 3 (5) 0 (0) (intermedius) C. diversus (intermedius) 1 1 (100) 1 (100) C. freundii (90.5) 1 Edwardsiella tarda 1 1 (100) 1 (100) Enterobacter aerogenes (100) 60 (100) 0 (100) E. agglomerans (Erwinia (83.3) 34 (94.4) 1 (100) herbicola) E. cloacae (90.0) 59 (98.3) 0 (100) E. gergoviae (E. cloacae) (100) 4 (66.6) (100) Escherichia coli (100) 66 (100) (100) Hafnia alvei (100) 30 (100) 10 (100) Klebsiella ozaenae 1 1 (100) 1 (100) K. pneumoniae (100) 66 (100) (100) K. rhinoscleromatis (100) 11 (91.7) Morganella morganii (Proteus morganii) (100) 30 (100) 10 (100) Proteus mirabilis 10 P. vulgaris Providencia alcalifaciens (Proteus inconstans) P. rettgeri (Proteus rettgeri) P. stuartii (Proteus rettgeri) Salmonella sp. S. typhi Serratia liquefaciens S. marcescens S. rubidaea (Bacterium rubidaeum) Shigella dysenteriae S. sonnei Shigella sp. (S. boydii and S. flexneri) Yersinia enterocolitica Not members of Enterobacteriaceaed Total (100) 1 1 (100) 1 1 (100) ,00 30 (100) 11 (91.7) 16 (100) 1 (100) 30 (100) 10 (83.3) 1 (100) 66 (100) 36 (100) 1 (100) 60 (100) 998 (97.8) 1,001 (98.1) 33 (97.6) " Identified by reference laboratory at CDC; the clinical microbiology laboratory at CDC participated in the collaborative evaluation. Numbers in parentheses indicate percentage of the particular bacterium identified correctly. ' Applied consistently in the laboratories of two participants only. d Consisted of: Pseudomonas aeruginosa, two; P. maltophilia, one; P. cepacia, one; P. stutzeri, one; Acinetobacter calcoaceticus var. anitratus, one; A. calcoaceticus var. lwoffi, one; Flavobacterium meningosepticum, one; group CDC 11 K 1, one; group CDC VE 1, one. 60 (100) 1 (100) 1 (100) 30 (100) 10 (83.3) 16 (100) 1 (100) 30 (100) 1 (100) 1 (100) 65 (98.5) 35 (97.) 1 (100) 60 (100) 0 (100) 10 (100) 3 (75) 53 (98.1) 8 (100) 10 (100) 3 (75.0) (100) 11 (91.6) 0 (100) sample injector assembly, the suspension of bacteria was agitated on a Vortex mixer for 15 to 30 s. The EBC was affixed to the single-barreled card sample injector after it had been properly marked for computer recognition, and the contents of the injector were introduced into the EBC by the vacuum assembly furnished with the instrument. The operator inspected the card to verify that it was properly filled. The card was then placed into the reader/incubator for analysis. Members of Enterobacteriaceae are identified within 8 h, with all reactions recorded on a printout. Identification of each organism is accompanied by a probability number for the accuracy of this identification. The computer also provides the next closest possible identification. The differences in probabilities usually are extremely large. EBC. The EBC (Fig. 1) is a 30-compartment card containing 6 dried biochemical broths and a growth control broth. Three empty wells are for addition of other substrates in the future. The card contains 14 carbohydrate substrates, decarboxylases, 1 deaminase, and 1 dihydrolase, in addition to materials such as urea, citrate, malonate, and a system for detecting HS. The carbohydrate base, with the exception of the

3 696 ISENBERG ET AL. FIG. 1. EBC card. (A) Full-length guide ledge for proper orientation of EBC in incubator/reader; (B) bubble traps; (C) number entry blocks for identifying specimen, using up to seven numerals; (D) guide notches for internal instrument manipulation; (E) distribution channels connecting each compartment to inoculum injection port; (F) injection port for inoculum introduction. The substrates are distributed as designated: (1) DP300; () blank; (3) growth control medium; (4) blank; (5) p-coumaric acid; (6) indoxyl-,b-d-glucoside; (7) urea; (8) citrate; (9) malonate; (10) tryptophan; (11) raffinose; (1) rhamnose; (13) maltose; (14) sorbitol; (15) melibiose; (16) mannitol; (17) xylose; (18) sucrose; (19) inositol; (0) adonitol; (1) blank; () substrate for HS detection; (3) ONPG; (4) lactose; (5) arabinose; (6) glucose; (7) arginine; (8) lysine; (9) base control for decarboxylases and dihydrolase; (30) ornithine. compartment containing o-nitrophenyl-d-galactopyranoside (ONPG), consists of (per liter): g of proteose peptone no. 3 (Difco Laboratories, Detroit, Mich.); 0.5 g of KHPO1; 0.5 g of sodium deoxycholate; and 0. g of reduced aniline blue as indicator. Adonitol, arabinose, glucose, maltose, mannitol, sucrose, and xylose are present in concentrations of 10 g/liter; inositol, lactose, melibiose, raffinose, rhamnose, and sorbitol are present at 5 g/liter. J. CLIN. MICROBIOL. The ONPG base consists of (per liter): proteose peptone no. 3, 0. g; beef extract (Difco), g; ONPG, 0.1 g; i-propyl-d-thiogalactopyranoside, 0.1 g; and reduced aniline blue as indicator, 0. g. The base for the decarboxylases and arginine dihydrolase consists of the following (per liter): Thiotone (BBL Microbiology Systems, Cockeysville, Md.), g; glucose, 1 g; pyridoxyl 5-phosphate, g; and bromothymol blue as indicator, 0.1 g. The amino acids are added at a level of 1%. The tryptophan deaminase is dissolved in a base consisting of (per liter): NaCl, g; Trypticase (BBL), 1 g; tryptophan, 5 g; and ferric citrate as indicator, 10 g. Citrate medium has the following composition (per liter): Thiotone (BBL), 0.5 g; sodium citrate, g; MgSO4.7H0, 0. g; (NH4)HPO4, 1 g; bromothymol blue as indicator, 0. g. Malonate is prepared as follows (per liter): Thiotone, 0.5 g; (NH4)SO4, 0.5 g; malonic acid, 4 g; glucose, 0. g; bromothymol blue as indicator, 0. g. Urea medium contains (per liter): urea, 0 g; KHPO4, g; proteose peptone no. 3 (Difco), 1 g; sodium thioglycolate, 1 g; bromothymol blue as indicator, 0. g. To assess HS generation, the following substrate is provided, containing (per liter): Thiotone (BBL), g; beef extract (Difco), g; glucose, 0.5 g; sodium thiosulfate, 0.5 g; L-cystine, 0.1 g; ferric citrate as indicator, 0.3 g. Three novel substrates useful in the division of Enterobacteriaceae were incorporated into EBC. Indoxyl-D-glucoside (plant indican) in a concentration of 1 g/liter is suspended in proteose peptone no. 3 (Difco) (1 g/liter) and KHPO4 (1 g/liter). p-coumaric acid is suspended in a medium containing (per liter): coumaric acid, 1 g; KHPO4, 0.5 g; Trypticase (BBL), g; glucose, 10 g. Reduced aniline blue, 0.0 g/liter, is the indicator added. Irgasan (DP300), a broad-spectrum bacteriostat (CIBA-GEIGY, Nutley, N.J.), is suspended in medium containing (per liter): Thiotone (BBL), 3 g; sodium deoxycholate, 4 g; glucose, 10 g; DP300, 10 ml of 1% stock. Reduced aniline blue, 0. g/ liter, serves as indicator. Standard biochemical test method. All cultures submitted simultaneously to the collaborating laboratories for analysis were tested by the standard procedure advocated by CDC as outlined in Edwards and Ewing (3). As mentioned previously, all media were prepared in one commercial laboratory from the same lot and shipped to all participants at the same time. The following tests and media were included in this analysis: tests for oxidase, triple sugar iron agar, motility medium, methyl red-voges-proskauer broth, indole broth, citrate medium, Christensen urea, decarboxylase media for lysine and ornithine, arginine dihydrolase, phenylalanine deaminase, malonate, gelatin, deoxyribonucleic acid hydrolysis, and the prescribed media for carbohydrate fermentation containing arabinose, adonitol, inositol, sorbitol, raffinose, and rhamnose. This approach measured 3 characteristics of the microorganisms. Quality control. The reaction of the EBC was controlled with a series of six bacteria. These organisms were: Proteus mirabilis (ATCC 700), which reacts positively with glucose, ornithine, xylose, HS, urea, citrate, tryptophan (tryptophan deaminase), and p-coumaric acid; Citrobacter freundii (ATCC 6750),

4 VOL. 11, 1980 which is positive for glucose, ONPG, lactose, arabinose, mannitol, xylose, rhamnose, maltose, melibiose, HS, citrate, DP300, and coumaric acid; Shigella flexneri (ATCC 1661), which is positive for glucose and mannitol; Pseudomonas aeruginosa (ATCC 7315), which is positive for arginine, citrate, malonate, and p-coumaric acid; Klebsiella pneumoniae (ATCC 13883), which is positive for glucose, lysine, ONPG, lactose, arabinose, mannitol, xylose, sucrose, inositol, adonitol, raffinose, rhamnose, maltose, sorbitol, melibiose, plant indican, citrate, and malonate; and Enterobacter cloacae (ATCC 3355), which is positive for glucose, arginine, ornithine, ONPG, arabinose, mannitol, xylose, sucrose, raffinose, rhamnose, maltose, sorbitol, melibiose, plant indican, citrate, and malonate. RESULTS Table 1 summarizes the collaborative comparison of the AMS with standard and routine manual methods, using a total of 170 representatives of Enterobacteriaceae. Since there were six participants, the total numbers of each strain tested (column 3, Table 1) are the number of strains that should have been identified correctly by all participants. The results indicate that of 9 species analyzed by the AMS were identified correctly by all of the automated equipment in the six laboratories. The manual standard method identified all of 1 of these Enterobacteriaceae species, whereas the routine laboratory method (API) identified 4 of the species without error. Of course, it must be emphasized that only two of the participating laboratories used the so-called routine method consistently; therefore, the opportunities for correct and erroneous identification were reduced threefold. Correct identification of the species submitted as unknown samples to each laboratory depended on the number of opportunities each laboratory had to identify representatives of a specific species. These opportunities varied from 1 to as many as 7 in the case of the various salmonellae included. Percentages listed in the table reflect the total number of opportunities to identify the species submitted as unknowns. Thus, if two representatives of a species were submitted, there were 1 opportunities for the various methods to identify that particular organism. The total number of strains tested (column 3, Table 1) is the basis for determining the percentage identified correctly. The table does not reflect the performance of individual laboratories. The overall correct identification perfornance by the AMS in the six laboratories was as follows: A, 94.3%; B, 99%; C, 96.8%; D, 98%; E, 97.5%; and F, 96.%. Their performance with the otheir mn-thods was of the same order of magnitude. Discrepancies in AMS identifications of the collaborative study organisms are listed in Table AMS ENTEROBACTERIACEAE BIOCHEMICAL CARD 697. This table permits review of the reproducibility of the AMS in various laboratories in terms of correct as well as incorrect identification. It further permits an evaluation of the reasons for some of the discrepancies observed. The AMS in each of the laboratories participating in this collaborative evaluation of unknown representatives of Enterobacteriaceae was equipped with a tape-recording device which permitted a centralized computer to judge the results obtained in 6, 8, and 10 h. This device was helpful in establishing analyses of discrepant reactions. Thus, the one Arizona hinshawii was identified by one AMS as Hafnia alvei in 8 h. The reaction responsible for this mistake was negative citrate utilization. A 10-h reading of this culture by the same AMS identified the organism correctly. The additional h of incubation allowed the Arizona to express citrate utilization. The failure to react properly in 8 h may reflect one of two possibilities: (i) the reaction may have been slow in this particular EBC card, because substrate availability or enzyme expression was hindered, or (ii) since all other AMS instruments identified the organism correctly, the inoculum may have been prepared incorrectly. All reactions necessary to identify this organism as A. hinshawii were positive except for the citrate reaction. This result favored identification as H. alvei, with a probability of 49% as compared with 3% for A. hinshawii. When such close correspondence between identification probabilities is observed at low levels, one questions the adequacy of the reactions for identifying either organism. With a positive citrate reaction, the probability of identification for A. hinshawii becomes 96%, whereas that for H. alvei approaches 0%. Citrobacter amalonaticus was misidentified by two machines. Different strains of the organism were involved. In one instance the organism was identified as a C. freundii in 8 h and correctly in 10. Indoxyl-D-glucosidase was not expressed in the shorter incubation period, thus leading to the incorrect identification as C. freundii, with a probability of 6% as compared with 36% for C. amalonaticus. The indoxyl-dglucosidase is positive in 90% of C. amalonaticus and negative in 99% of the C. freundii. Once more, a slow reaction or an inadequately dense inoculum may explain the incorrect reaction in the 8-h period. The second AMS identified a different culture of C. amalonaticus as C. diversus. The identification was based on malonate utilization, which is 99% negative for C. amalonaticus. The reaction profile elicited identified this bacterium as a C. diversus with a probability of 94%. Comparing these reactions with the other five AMS results indicates that

5 698 ISENBERG ET AL. TABLE. AMS identification discrepancies in collaborative study bacteria v Standard Bacterium AMS identification No. AMS Reaction discrepancy' 10-h reading and/or Explanation routine Arizona hinshawii Hafnia alvei 1/6 CIT- A. hinshawii Correct Slow reaction Citrobacter ama- C. freundii 1/6 PLI- C. amalonaticus Incorrect Slow reaction lonaticus C. amalonaticus C. diversus 1/6 MAL+ C. diversus Incorrect See text C. freundii Escherichia coli 1/6 CIT- C. freundii Correct Slow reaction C. freundii Enterobacter ag- 1/6 HS- C. freundii Correct Slow reaction glomerans C. freundii Yersinia enter- /6 PLI+, CIT- C. freundii Correct See text ocolitica Enterobacter ag- Y. enterocolitica /6 ONP-, XYL-, Y. enterocolitica Incorrect See text glomerans DP3- ONP- E. agglomerans Correct Slow reaction E. agglomerans Y. enterocolitica 3/6 CIT- E. agglomerans Correct Slow reaction E. cloacae Y. enterocolitica 6/6 ARG-, LAC-, Y. enterocolitica Correct See text RAF-, MEL-, CIT- Providencia stuar- E. agglomerans 1/6 LAC+, ARA+, E. agglomerans Incorrect Clerical error tii MAN+, RHA+, SOR+ Serratia rubidaea Y. enterocolitica /6 LYS-, RHA-, CIT- S. rubidaea Correct Slow reaction 'Abbreviations: CIT, citrate; PLI, plant indican; MAL, malonate; ONP, ONPG; XYL, xylose; DP3, irgasan; ARG, arginine dihydrolase; LAC, lactose; RAF, raffinose; MEL, melibiose; ARA, arabinose; MAN, mannitol; RHA, rhamnose; SOR, sorbitol. all reactions were identical except for the malonate utilization. The only adequate explanation for this incorrect singular reaction is an improper threshold adjustment arising from the paucity of C. amalonaticus for inclusion in the data base. There were four misidentifications of C. freundii involving four AMS results and three different strains of the bacterium. In one machine C. freundii was called Escherichia coli in 8 h but correctly as C. freundii in 10 h (Table ), again suggesting a slow reaction. The substrates involved in this misidentification were negative 8- h utilizations of citrate and DP300. Citrate should be positive in more than 90% of reactions with C. freundii. The chance of encountering a citrate-negative representative is approximately 10%. Therefore, the reaction battery led to the designation of E. coli, characteristically negative for citrate at a 99% level. Citrate was utilized after 10 h of incubation, suggesting once again an inoculum or enzyme effect. In another instance, the identification in the 8-h period of C. freundii as Enterobacter agglomerans was based on a series of reactions requiring a positive HS for correct identification of this particular C. freundii. Although in the previous case C. freundii belonged to the 1% which do not generate HS, the failure to do so with this particular representative indicated the probability of a C. freundii at a level of only 9% and an E. agglomerans with a probability of 70%. The 10- h reading elicited the proper reaction, suggesting once again inoculum or slow enzyme effects. In two of the six AMS, one strain of C. freundii J. CLIN. MICROBIOL. was identified as Yersinia enterocolitica on the basis of two reaction discrepancies: false-positive plant indican and false-negative citrate. In 10 h the correct citrate was obtained. Indoxyl-D-glucoside is negative for C. freundii with a frequency of 99%, indicating the need for threshold adjustment. Two of the six representatives of E. agglomerans showed discrepancies in the AMS analysis. The first of these was identified in one of the laboratories as Y. enterocolitica even at the 10- h incubation period. The standard and routine methods also identified the orlg'nism incorrectly. Three incorrect reactions led to the misidentification. Therefore, the probability of identifying this organism as an E. agglomerans became very small, whereas Y. entetocolitica was correct at the level of 97%. FailurS- of the standard and routine methods to identify this particular bacterium correctly strongly suggests a clerical error in the numbering of this particular culture. In another laboratory, the organism was identified as a Y. enterocolitica on the basis of a negative ONPG in 8 h; it was identified as a Y. enterocolitica at 51% and as an E. agglomerans at 31% probability levels. The belatedly positive ONPG led to the correct identification with a probability of 90%. The second strain of E. agglomerans was identified in three laboratories as Y. enterocolitica on the basis of a negative citrate. In 10 h, this reaction was corrected in the three AMS, and the organism was identified properly. Again, either an inoculum, card, or threshold effect could explain the delayed diagnoses. Of course,

6 VOL. 11, 1980 we must also consider that the small differences in the rate of citrate transport and utilization as carbon source of a particular strain may play a significant role in the delays noted. One strain of Enterobacter cloacae was misidentified by all AMS instruments as a Y. enterocolitica through differences in five reactions. These were: negative arginine dihydrolase, failure to ferment lactose, raffinose, and melibiose, and inability to utilize citrate. Usually, E. cloacae in the AMS produces arginine dihydrolase at a frequency of 99%, ferments lactose at 76%, raffinose at 90%, and melibiose at 95%, and utilizes citrate at a 98% level. All of these reactions, with the possible exception of lactose fermentation, are significant for identifying E. cloacae. Arginine is probably the pivotal reaction, since the other reactions are shared by several species and genera. Thus, these reactions might have fit an exceptional S. liquefaciens. The negative reactions in the five substrates listed argue very strongly for a Y. enterocolitica despite the positive urease production usually encountered. This discrepancy in all of the instruments could have been due to an improperly labeled culture, but both the standard and routine methods identified the bacterium properly. In an attempt to understand these discrepancies, one participant tested the same culture at a different time with a different lot of EBC. Correct reactions were elicited, suggesting that the particular lot of EBC used in the collaborative study was inadequate for identifying this particular E. cloacae, requiring substrate or threshold adjustments. The error in identifying Providencia stuartii by one AMS suggests a clerical error. There were five reactions incompatible with the identification of P. stuartii, unchanged in 10 h and confirmed by routine and standard methods which also identified the organism as an E. agglomerans. One strain of Serratia rubidaea was identified as a Y. enterocolitica by two of the AMS in the 8-h reading. The misidentification was based on negative reactions for lysine, raffinose, and citrate. The 10-h reading in both laboratories reversed these negative reactions and identified the culture properly. The correct identification was also made by the manual methods applied. Once more, a slow reaction possibly resulting from an inadequate inoculum density may have been involved. While the collaborative analysis of the unknown cultures was in progress, 5,450 representatives of the family Enterobacteriaceae were isolated in the participating laboratories from clinical specimens. These were analyzed by the routine methods used in each of the laboratories and with the EBC. For the purposes of this AMS ENTEROBACTERIACEAE BIOCHEMICAL CARD 699 particular evaluation, we make the tacit assumption that the manual identification of these organisms was correct. Identifications by the AMS of these cultures were in agreement with the manual method in 5,55 instances (96.4%). At the same time, 93 such clinical isolates were tested in one laboratory by the standard method used in the collaborative study. Of these, 894 (96%) were correctly identified by the AMS. Among the Enterobacteriaceae encountered in the clinical specimens were organisms such as Klebsiella oxytoca and Providencia stuartii, urease positive. The computer memory of the AMS also has the capability of identifying Y. ruckeri, Y. pestis, and Y. pseudotuberculosis. However, identification of the first two of these yersiniae must await fine tuning of the instruments to accommodate the temperature limitations of Y. ruckeri and the incubation time requirement of Y. pestis. Because of the different methods used for identification, the unequal distribution of organisms tested in the various laboratories, and the distribution of representatives encountered in the clinical setting, presentation of a detailed analysis of these clinical samples would serve no purpose. DISCUSSION The capability of the AMS to identify Enterobacteriaceae at the same level as the classical and at least one of the manual systems approaches has been demonstrated in this study. This analysis pertains to clinical specimens as well as to the bacteria especially selected for the collaborative evaluation. The bacteria chosen challenge the limits of identification in all systems. Reproducibility from instrument to instrument was excellent. It is, however, the reaction file which is the critical part of the EBC processing by the AMS computer. The EBC reactions only characterize the unknown specimen; they cannot identify the bacterium. Experimental testing of many representatives of Enterobacteriaceae in the AMS was required to extract a statistical expression of the percentages (of each species) of positive and negative reactions for each substrate in the EBC for each identification. This is especially important because the environment of the AMS differs from the conditions that govern the manually prepared standard and the commercially manufactured systems. The passage of atmospheric gases, the small volume within each reaction chamber or well, the shorter incubation time, and the inoculum density contribute to these differences and require a special AMS identification file. Each species has the same initial probability. No biochemical substrate is

7 700 ISENBERG ET AL. given more weight regardless of its importance to delineate a species. No species receives additional advantages because of the frequency with which it occurs in clinical samples. The frequency with which each species acts on each substrate influences the conditional probability consideration of the computer. Before comparing the reactions obtained for a specific organism and attempting identification, the AMS performs several preliminary checks. Thus, the glucose reaction is screened to ascertain that the organism belongs to Enterobacteriaceae. If the glucose reaction is not positive, the control well is checked to establish that adequate proliferation has taken place. When the growth control well is positive but glucose is negative, the urease and ornithine decarboxylase compartments are checked. If these are positive, the computer assumes the glucose to be positive but reacting slowly. If glucose, urease, and ornithine are negative while the growth control well is positive, the computer will report that this particular organism is not a member of Enterobacteriaceae. The probability computation performed by the computer follows the usual Baysian calculation. The EBC card contains three novel substrates which aid in evaluating and determining the species that comprise the family Enterobacteriaceae. Two of these substrates, which act by interfering with glucose fermentation, are DP300 and p-coumaric acid. The organisms capable of fermenting glucose despite the presence of DP300 are Yersinia sp., Citrobacter sp., Serratia sp., Morganella (Proteus) morganii, Providencia (Proteus) rettgeri, and Providencia stuartii, including the urea-positive variant. The organisms that can ferment glucose in the presence of p-coumaric acid are Yersinia sp., Citrobacter sp., Serratia sp., Arizona hinshawii, Salmonella sp., Escherichia coli, and most species of Shigella. The indoxyl-f-d-glucoside is broken down by certain bacteria into indigo blue by a specific f8-glucosidase. Organisms which can produce this enzyme are the various species of the genus Serratia; the various Enterobacter species including E. gergoviae; the various Klebsiella species including K. oxytoca; Citrobacter diversus; and some Providencia rettgeri. The level of positive reaction for the last group is better than 90% for the organisms listed, whereas more than 80% of the genera listed for the DP300 and p- coumaric acid reactions are positive in fermenting glucose in the presence of these compounds. In identifying the various Enterobacteriaceae species, these positive reactions are helpful adjuncts for their automated recognition, especially in the absence of certain classical reactions J. CLIN. MICROBIOL. often performed, such as the test for indole. The EBC does contain a well with ordinary nutrient broth which can serve as a ready reservoir for this test and as a source of antigen for immunological corroboration. The discrepancies observed provide the opportunity to discuss the advantages inherent in this particular automated approach and propose means to improve the AMS further. In addition to its labor-saving aspects, the AMS reduces the chances for human error. One of the contributing factors is, of course, the computer, which can handle the numerous reactions and search its memory rapidly and accurately (7). Also, the computer's software can be updated continuously to accommodate changes in reaction sequences for newly established species or significant biotypes. It could also be instructed to critically scrutinize results which indicate a close probability of identification between organisms. Under those circumstances, the computer could instruct the instrument to prolong incubation for 1 or h or to resolve difficulties which might arise from inadequate or improper inoculum preparation or result from an inherent, slowly expressed enzyme sequence manifested in a particular organism. The computer also possesses the a bility to judge the purity of the culture.analyzed and could be instructed to evaluate a series of reactions which imply an impure culture. Also, it is possible to update computer instructions to request additional tests not only when immunological confirmation of identification is desirable, but also in those rare instances where additional biochemical tests should be performed to confirm the identification. To this end, the EBC cgntains a growth control compartment which can be entered, and the contents can be used for immunological and additional biochemical tests. The computer memory has the capability of dividing the genera Salmonella and the genus Shigella in various "species" or serotypes. This grouping of the salmonellae would seem unnecessary in identifying a clinical isolate, since reliance on immunological confirmation at least into groups is so well ingrained into clinical laboratory personnel that reliance solely on biochemical reactions might be deemed unacceptable. Only Salmonella typhi has therefore been sequestered from the remaining salmonellae. The frequency with which different shigellae are encountered in the clinical laboratory is recognized in the instructions to the AMS computer. Thus, Shigella dysenteriae and S. sonnei can be identified by their biochemical reactions, but the computer continues to advise immunological confirmation. S. boydii and S. flexneri are

8 VOL. 11, 1980 grouped together, for the same reasons as the salmonellae. In all instances, confirmation by immunological analysis is recommended on the computer printout. Some of the discrepancies observed which were ascribed to inoculum density can also be corrected by fine tuning the threshold setting in the computer for a particular well in the EBC. Each of these wells has a distinct point or threshold which discriminates between positive and negative reactions. This particular decision depends on the amount of light transmitted and recognized by the several detecting elements which scan the well at hourly intervals, as well as the ability of the detectors to recognize significant color changes. There are a number of reactions which may require a high threshold setting to accommodate equivocal areas. However, the computer linked to the detectors could be i istructed to direct the perusal of other diagnos-dc reaction compartments, to evaluate the equivocal threshold reading by interpolating the slope of the growth curve to a longer incubation time, and to decide on the identity of the particular isolate. Additionally, the contents in each EBC well can be fine tuned. Experience with many biotypes of the species may require adjustment in certain substrate concentrations or basal media to ease the recognition process. For example, slow-reacing citrate utilization and arginine dihydrolase production can be adjusted in this fashion. The 6-, 8-, and 10-h computer readings taken during the evaluation of the collaboratively analyzed cultures were intended to confirm the optimal incubation time commensurate with accuracy and the environment provided by the EBC. The shorter time period, compared with the standard and most commercially available systems, is an added advantage to the decision prowess of the AMS. Although it is possible to determine the presence of some bacterial enzymes very rapidly, the detection of others requires proliferation and synthesis, at times involving sequences of enzymes bearing on the final reaction. The excellent performance of the instrument indicates the suitability of the 8-h incubation. Adjustments for the majority of the inoculum-dependent discrepancies can be incorporated into the computer memory. Also, it may be possible to use the growth control well as control for adequate inoculum density. This specific compartment can be screened to determine the level of growth before identification is made. If a preset threshold is not reached in 8 h, as might be expected with slower-growing organisms or inadequately dense inocula, the AMS computer can be instructed to AMS ENTEROBACTERIACEAE BIOCHEMICAL CARD 701 prolong the incubation period before reaching a decision. Since 7 of the 11 discrepancies observed were the result of potential inoculum inadequacies, slowly elaborated enzyme sequences, or slowly synthesized enzyme combinations detected phenotypically, the advantages of this control procedure are evident. Although the reproducibility from instrument to instrument was excellent, fine tuning of individual instruments may be another means of overcoming any discrepant reactions. The family Enterobacteriaceae has received much attention by investigators attempting to bring some order to microbiological taxonomy. The initial, outstanding contributions of Edwards and Ewing (3) have been expanded by numerical taxonomy and deoxyribonucleic acid analysis () so that interspecies differences are now determined by phenotypic expressions of specific genomes. However, since numerical taxonomy and deoxyribonucleic acid analysis are beyond the capabilities of clinical microbiology laboratories, biochemical tests which reflect the species genome must be chosen and used for bacterial species identification. In the future, it may well be shown that a number of the reactions incorporated into the EBC and, for that matter, into the manual systems approaches or the standard method are superfluous. Adjustments certainly can be made in all of these systems to refine and possibly reduce the testing required to arrive at meaningful microbiological diagnoses. The EBC is an important addition to the utility of the AMS in the clinical microbiology laboratory, expanding the application of automation to another important time- and laborconsuming activity. The recent preliminary report on AMS general susceptibility and minimal inhibitory concentration capabilities (6) indicates even further progress in the application of automation and computer technologies by microbiology laboratories charged with the resp,ui. sibility of providing accurate and rapid service for the benefit of patients. ACKNOWLEDGMENTS We gratefully acknowledge the important contributions of the many individuals working in our laboratories. Without their efforts and devotion, this study could not have been conducted. LITERATURE CITED 1. Aldridge, C., P. W. Jones, S. Gibson, J. Lanham, M. Meyer, R. Vannest, and R. Charles Automated microbiological detection/identification system. J. Clin. Microbiol. 6: Brenner, D. J Characterization and clinical identification of Enterobacteriaceae by DNA hybridization. Prog. Clin. Pathol. 7:

9 70 ISENBERG ET AL. 3. Edwards, P. R., and W. H. Ewing Identification of Enterobacteriaceae, 3rd ed. Burgess Publishing Co., Minneapolis. 4. Isenberg, H. D Biochemical "rapid identification" of Enterobacteriaceae, p In J. E. Prier, J. Bartola, and H. Friedman (ed.), Modern methods in medical microbiology. University Park Press, Baltimore. 5. Isenberg, H. D., T. L. Gavan, A. Sonnenwirth, W. I. Taylor, and J. A. Washington II Clinical laboratory evaluation of automated microbial detection/ identification system in analysis of clinical urine specimens. J. Clin. Microbiol. 10: Isenberg, H. D., and J. Sampson-Scherer Clinical laboratory feasibility study of antibiotic susceptibility determined by the Auto Microbic System, p. 56- J. CLIN. MICROBIOL. 58. In J. D. Nelson and C. Grassi (ed.), Current chemotherapy and infectious disease. American Society for Microbiology, Washington, D.C. 7. MacLowry, J. D., E. A. Robertson, and R. J. Elin The place of the computer in diagnostic medical bacteriology. Proc. Clin. Pathol. 7: Smith, P. B., T. L. Gavan, H. D. Isenberg, A. Sonnenwirth, W. I. Taylor, J. A. Washington II, and A. Balows Multi-laboratory evaluation of an automated microbial detection/identification system. J. Clin. Microbiol. 8: Sonnenwirth, A. C Preprototype of an automated microbial detection and identification system: a developmental investigation. J. Clin. Microbiol. 6: Downloaded from on March 8, 019 by guest