Evaluation of the Biolog system for the identification of food and beverage yeasts

Letters in Applied Microbiology 1997, 24, 455 459 Evaluation of the Biolog system for the identification of food and beverage yeasts W. Praphailong 1, M. Van Gestel, G.H. Fleet 1 and G.M. Heard Cooperative Research Centre for Food Industry Innovation and 1 Department of Food Science and Technology, University of New South Wales, Australia 1208/96: received 25 September 1996 and accepted 26 October 1996 W. PRAPHAILONG, M. VAN GESTEL, G.H. FLEET AND G.M. HEARD. 1997. The inconvenience of conventional yeast identification methods has resulted in the development of rapid, commercial systems, mainly for clinical yeast species. The Biolog system (Biolog Inc., Hayward, CA, USA) is a new semi-automated, computer-linked technology for rapid identification of clinical and non-clinical yeasts. The system is based around a microtitre tray and includes assimilation and oxidation tests. This paper evaluates the Biolog system for the identification of 21 species (72 strains) of yeasts of food and wine origin. Species correctly identified included Saccharomyces cerevisiae, Debaryomyces hansenii, Yarrowia lipolytica, Kluyveromyces marxianus, Kloeckera apiculata, Dekkera bruxellensis and Schizosaccharomyces pombe. Zygosaccharomyces bailii and Zygosaccharomyces rouxii were identified correctly 50% of the time and Pichia membranaefaciens 20% of the time. INTRODUCTION Yeasts are significant in foods because they cause spoilage and conduct desirable fermentations (Deak 1991; Fleet 1992; Tudor and Board 1993). Routine methods for their identi- fication are needed. Conventional procedures for identi- fication are time consuming, inconvenient and require expertise for interpretation of results (Kregar van Rij 1984; Barnett et al. 1990 ; Deak 1993). Systems have been developed for rapid yeast identification, but mainly for species of clinical significance. These systems are based on a few highly dis- criminative tests (Deak and Beuchat 1987) that are packaged into diagnostic kits such as the API 20C and ATB 32C kits (biomérieux, France) and enzyme-based kits including the API Yeast Ident and the Baxter Microscan Yeast panel (Deak and Beuchat 1988 ; Török and King 1991 ; Deak and Beuchat 1993 ; Deak and Beuchat 1995). These kits have been applied to identification of food yeasts with various degrees of success (Rohm et al. 1990 ; Deak and Beuchat 1993). Currently avail- able kits include only a small range of tests and their databases do not include some important food spoilage yeasts. Heard and Fleet (1990) overcame some of these limitations by incorporating a large range of tests for yeast identification into a microtitre tray. While this approach was convenient and gave reliable identification, it was not interfaced with a computer database and not automated. Correspondence to: Dr G. M. Heard, Department of Food Science and Technology, University of New South Wales, Sydney, NSW 2052, Australia. The Biolog system (Biolog Inc., Hayward, CA, USA) is a new semi-automated, computer-linked technology for yeast identification. It is based around a 96-well microtitre tray containing a range of dehydrated carbon sources for assimilation and oxidation tests (Bochner 1989 ; Biolog 1993). The profile of growth responses provides a metabolic fingerprint for each isolate (Bochner 1989) and is compared to profiles of the 267 yeast species in the Biolog database, to provide an identification. Biolog systems have been developed for bac- teria (Stager and Davis 1992) and evaluations have been given mixed endorsements of reliability (Miller and Rhoden 1991 ; Klinger et al. 1992 ; Jones et al. 1993 ; Miller et al. 1993). The Biolog deviates from conventional methods for yeast identification (Kreger van Rij 1984 ; Barnett et al. 1990) by introducing a number of co-metabolism tests and many assimilation tests and oxidation tests not common to the traditional systems. This paper evaluates the Biolog for the identification of food-associated yeasts. Some performance limits of the system, including reproducibility and the effects of culture preparation, are reported. MATERIALS AND METHODS Yeast species Seventy-two strains of yeasts, representing 21 species, were examined (Table 1). These cultures were obtained from CSIRO Division of Food Science and Technology, North 1997 The Society for Applied Bacteriology

456 W. PRAPHAILONG ET AL. Number Strains (%) correctly Number of times identified to of strains Number correctly Yeast species tested of tests* identified Genus Species Candida krusei 3 5 5 100 100 Candida parapsilosis 3 9 9 100 100 Debaryomyces hansenii 5 8 8 100 100 Dekkera bruxellensis 7 16 16 100 100 Dekkera naardenensis 1 2 2 100 100 Kloeckera apiculata 6 17 17 100 100 Kluyveromyces marxianus 4 6 6 100 100 Pichia anomala 1 2 2 100 100 Saccharomyces cerevisiae 7 23 23 100 100 Schizosaccharomyces pombe 2 2 2 100 100 Yarrowia lipolytica 6 7 7 100 100 Candida multisgemmis 1 3 0 100 0 Candida sake 1 4 0 100 0 Rhodotorula glutinis 1 2 0 100 0 Octosporomyces octosporus 1 2 0 100 0 Zygosaccharomyces bailii 7 14 7 50 50 Zygosaccharomyces rouxii 2 6 3 50 50 Pichia membranaefaciens 4 10 2 30 20 Saccharomyces bayanus 2 4 0 0 0 Saccharomycodes ludwigii 2 2 0 0 0 Torulaspora delbrueckii 7 10 0 0 0 * Identification was completed using conditions recommended by Biolog. Some strains were tested more than once. Table 1 Yeast species identified by the Biolog system Ryde, Australia; Australian Wine Research Institute, Glen substrates for assimilation reactions, and the lower section is Osmond; South Australia Centraalbureau voor Schimmeltetrazolium for oxidation reactions. Oxidation wells contain a redox dye, cultures, Yeast Division, Delft, the Netherlands ; and Departif violet, which changes from colourless to purple ment of Food Science and Technology, University of New the substrate is oxidized. The inoculated microplate was South Wales, Sydney, Australia. The identities of these spec- incubated at 25 C for 24 h, results were recorded by the ies were confirmed by conventional methods (Kreger van Rij Microplate reader, and processed for identification by the 1984 ; Barnett et al. 1990) and by the microtitre tray method Microlog software. Microplates were also read after 48 to 72 of Heard and Fleet (1990). Cultures were maintained on h. A similarity index may be calculated, based on the reaction slants of malt extract agar (MEA) (Oxoid, Sydney, Australia). profiles for the 10 species most closely related to the test strain. At 24 h, an acceptable species identification must have Biolog Microstation a similarity (SIM) value of 0 75 or above, and at subsequent readings (48 72 h) a SIM value of 0 5 or above is needed Standard procedure. The Biolog Microstation consists of (Biolog 1993). For comparison, the system gives the names, Biolog microplates, each containing 94 test wells and two expected profiles and SIM values of the 10 closest species. control wells, a multichannel pipetter, a turbidimeter, a com- Reproducibility. Four yeasts, Debaryomyces hansenii, Dekkera puter-linked microplate reader and Biolog MicroLog bruxellensis, Kloeckera apiculata and Saccharomyces cerevisiae, software. Before use, yeasts were subcultured onto plates were each tested five times in the Biolog system using the of Biolog Universal Yeast Agar (BUY) (Biolog Inc.) and standard procedure. Each replication used a freshly grown incubated at 25 C for 1 2 d. The inoculum was prepared by culture. suspending cells from the plates in sterile distilled water to Effect of culture medium, inoculum density and give 44 51% transmittance (%T) with the Biolog turbiculture age on identification dimeter. Inoculum (100 ml) was dispensed into each well of a Biolog yeast (YT) microplate. The YT plate is divided into Seven yeasts were used for these experiments : Yarrowia two sections : the top section consists of wells containing lipolytica, Deb. hansenii, Dekk. bruxellensis, Kloeck. apiculata,

BIOLOG IDENTIFICATION OF FOOD YEASTS 457 The effects of varying culture medium and inoculum prep- aration on identification of seven yeast species are shown in Table 2. All yeast species were correctly identified (SIM 0 75) after growth on BUY, MEA and YNB. With the exception of Deb. hansenii, cultures could be grown for up to 7 d prior to inoculation of YT plates without affecting identification. Zygosaccharomyces bailii did not give sufficient growth on BUY after 1 2 d for reliable identification and needed incubation for 3 4 d to produce an inoculum sus- pension with 50%T. At an inoculum density of 20 50%T, correct identifications were made for all yeasts (SIM 0 70). All yeasts except Dekk. bruxellensis were correctly identified at 80%T, but SIM values for Y. lipolytica and Kluy. marx- ianus were less than 0 5. Kluyveromyces marxianus, S. cerevisiae and Zygosaccharomyces bailii. To determine the effect of culture media on identification, yeasts were grown at 25 C on yeast nitrogen base 5% glucose agar (YNB) (Difco, MI, USA), MEA and BUY (standard medium) for 1 2 d prior to inoculation into the Biolog YT plates. All other steps followed the standard procedure. To determine the effect of inoculum density, inocula from BUY plates were prepared at 20%T (10 7 10 8 cfu ml 1 ), 50%T (10 6 10 7 cfu ml 1 ) and 80%T (10 5 10 6 cfu ml 1 ). Standard procedures were followed for inoculation of the YT plates. The populations (cfu ml 1 ) were determined by viable counts on MEA. To determine the effect of culture age, yeasts were grown at 25 C on BUY for 1 2 d (recommended by Biolog) and for up to 7 d. After incubation, cultures were prepared for identification by the standard procedure. Effects of growth medium, culture age and inoculum density Effect of suspension medium on identification Effect of suspension medium on identification Candida krusei, Deb. hansenii, Dekk. bruxellensis and Sac- Sterile saline (0.85%) was compared with sterile distilled charomyces cerevisiae were correctly identified using either water as a medium for suspending yeast cells prior to inocu- saline or water for preparation of the cell suspension. Using lation into YT plates. Nine yeasts, Candida krusei, Deb. water as suspension medium, M. pulcherrima was not corhansenii, Dekk. bruxellensis, Kloeck. apiculata, Metschnikowia rectly identified and P. membranaefaciens was correctly identipulcherrima, Pichia membranaefaciens, S. cerevisiae, Z. bailii fied only 20% of the time, but both cultures were correctly and Z. rouxii were used in the evaluation. Apart from varying identified (SIM 0 8) using saline. For Z. bailii and Z. the suspension medium, standard preparation and inoculation rouxii, no differences were found between identification procedures were followed for all cultures. results using the two suspension media. RESULTS Comparison of assimilation and oxidation test results Identification of yeasts using the Biolog system Twenty-nine of the substrates used in the Biolog system Of the 21 species examined, 15 were identified correctly to for yeast identification are included as both assimilation and genus level (Table 1). Metschnikowia pulcherrima, S. bayanus, oxidation tests. For all species examined (Table 1), a com- Saccharomycodes ludwigii and Torulaspora delbrueckii were not parison was made of results from the assimilation and oxiidentified to genus or species by the Biolog. All of the Candida dation tests for these 29 substrates. There were no differences species were correctly identified to genus but species identiany between results for assimilation and oxidation reactions for fications were incorrect for C. multisgemmis and C. sake. Only of the species. For each strain tested, where a substrate 11 of the yeasts were consistently identified to correct species was oxidized, the substrate was also assimilated. level. These included S. cerevisiae, food spoilage yeasts such as Deb. hansenii, Y. lipolytica and Kluy. marxianus, C. parapsilosus, Dekk. naardenensis and P. anomala and wine-related DISCUSSION species including C. krusei, Dekk. bruxellensis, Kloeck. apic- The Biolog aims to provide a rapid, convenient approach to ulata and Schizosaccharomyces pombe. Pichia membranaefaciens yeast identification. With a database of 267 species, including was identified correctly to species level only 20% of the time many spoilage yeasts and a range of 94 tests, it should overcome and Z. bailii and Z. rouxii could be identified only 50% of some of the limitations of other rapid kits developed the time. for yeast identification. However, of these 94 tests, only 26 The Biolog gave reproducible identification of Deb. are common with the biochemical and physiological tests hansenii, Dekk. bruxellensis, Kloeck. apiculata and S. cerevisiae. listed by Kreger van Rij (1984) and Barnett et al. (1990). Each of these species was identified correctly in five replications. Moreover, Biolog introduces oxidation tests, new metabolic reactions such as assimilation of palatinose, turanose, mal-

458 W. PRAPHAILONG ET AL. Table 2 Effects of growth medium, culture age and inoculum density on identification of yeasts by the Biolog system Similarity index value Yarrowia Debaryomyces Dekkera Kloeckera Kluyveromyces Saccharomyces Zygosaccharomyces Condition lipolytica hansenii bruxellensis apiculata marxianus cerevisiae bailii Culture growth medium MEA 0 814 0 752 0 994 0 756 0 958 0 987 0 756 BUY* 0 849 0 752 0 995 0 966 0 958 0 935 0 764 YNB 0 798 0 765 0 820 0 861 0 946 Culture age 1 2 d* 0 989 0 868 0 844 0 959 1 000 0 945 No ID 3 4 d 0 901 0 753 0 994 0 794 0 932 0 822 0 765 5 d 0 957 0 353 0 769 0 924 0 908 7 d 0 904 0 127 0 767 0 914 0 894 Inoculum density 20%T 0 899 0 869 0 991 0 966 1 00 0 851 0 93 (10 7 10 8 cfu ml 1 ) 50%T 0 72 0 849 0 935 0 828 0 982 0 878 0 917 (10 6 10 7 cfu ml 1 )* 80%T 0 682 0 875 No ID 0 753 0 5 0 878 0 836 * Conditions recommended by Biolog. Not tested. Insufficient growth to prepare inoculum suspension. totriose and stachyose, and a range of compounds combined with D-xylose in co-metabolism tests. In this study, 49 of 72 yeast strains (68%) were correctly identified by the Biolog system. However 27 strains, including several important food and beverage yeasts, were not correctly identified to species level. Results obtained here for all common tests in the Biolog plates agreed with reaction responses given by Kreger van Rij (1984) and Barnett et al. (1990). Responses to the remaining tests in the YT plates have not been described in the literature but agreed with profiles from the Biolog database. Thus, no particular reactions in the Biolog system were incorrect. The fault may lie with the generally poor reactivity of these species with many substrates (Kreger van Rij 1984 ; Barnett et al. 1990). However, for both conventional and Biolog systems, variable reactions are given for many assimilation tests. Consequently, it is important that reaction results of a number of strains of each species be included in the database to enable discrimination between species. Improvement of the database for some important spoilage yeasts is recommended. In the light of the present results, the relevance of co-metabolism tests and the role of oxidation tests should perhaps be reconsidered. Preparation of yeast cultures prior to inoculation into identification kits can influence the reliability of the results (Biolog 1993). Consequently, manufacturers specify conditions for inoculum preparation, including growth media and cell suspensions. However, it is not always possible to conform totally to these specifications. The data in Table 2 show that considerable variation to inoculum preparation did not affect correct identification of the yeasts examined, although there was some variation in response of very young and old cultures. Choice of diluent for preparing cell suspensions can affect the viability of yeast cells (Mian et al. 1997). Distilled water, the suspension medium recommended by Biolog, may not be suitable for the suspension of all yeast species. This study has revealed the potential for using the Biolog system for routine identification of some food and beverage yeasts. The Biolog has overcome several problems of existing diagnostic kits, by providing a large number of tests and a large database including many food-related species. Nevertheless, further refinements to the database and tests used may permit the system to be used with greater confidence. REFERENCES Barnett, J.A., Payne, R.W. and Yarrow, D. (1990) Yeasts : Characteristics and Identification, 2nd edn. Cambridge : Cambridge University Press.

BIOLOG IDENTIFICATION OF FOOD YEASTS 459 Biolog (1993) YT Microplate : Instructions for Use. Hayward, Cal- Klinger, J.M., Stowe, R.P., Obenhuber, D.C., Groves, T.O., ifornia : Biolog Inc. Mishra, S.K. and Pierson, D.L. (1992) Evaluation of the Biolog Bochner, B.R. (1989) Breathprints at the microbial level. ASM News automated microbial identification system. Applied and Environmental 55, 536 539. Microbiology 58, 2089 2092. Deak, T. (1991) Foodborne yeasts. Advances in Applied Microbiology Kreger van Rij, N.J.W. (1984) The Yeasts. A Taxonomic Study. 36, 179 278. Amsterdam : Elsevier. Deak, T. (1993) Simplified techniques for identifying foodborne Mian, M., Fleet, G.H. and Hocking, A.D. (1997) Effect of diluent yeasts. International Journal of Food Microbiology 19, 15 26. type on viability of yeasts enumerated from foods or in pure Deak, T. and Beuchat, L.R. (1987) Identification of foodborne culture. International Journal of Food Microbiology 53, 103 107. yeasts. Journal of Food Protection 50, 243 264. Miller, J.M. and Rhoden, D.L. (1991) Preliminary evaluation of Deak, T. and Beuchat, L.R. (1988) Evaluation of simplified and Biolog, a carbon source utilisation method for bacterial identification. commercial systems for identification of foodborne yeasts. International Journal of Clinical Microbiology 29, 1143 1147. Journal of Food Microbiology 7, 135 145. Miller, J.M., Biddle, J.W., Quenzer, V.K. and McLaughlin, J.C. Deak, T. and Beuchat, L.R. (1993) Comparison of the SIM, API (1993) Evaluation of Biolog for identification of members of the 20C, and ID 32C systems for identification of yeast isolated from family Micrococcaceae. Journal of Clinical Microbiology 31, 3170 fruit juice concentrates and beverages. Journal of Food Protection 3173. 56, 585 592. Rohm, H., Lechner, F. and Lehner, M. (1990) Evaluation of the Deak, T. and Beuchat, L.R. (1995) Evaluation of the MicroScan API ATB 32C system for the rapid identification of foodborne enzyme-based system for the identification of foodborne yeasts. yeasts. International Journal of Food Microbiology 11, 215 224. Journal of Applied Bacteriology 79, 439 446. Stager, C.E. and Davis, J.R. (1992) Automated systems for identi- Fleet, G.H. (1992) Spoilage yeasts. Critical Reviews in Biotechnology fication of microorganisms. Clinical Microbiology Reviews 5, 302 12, 1 44. 327. Heard, G.M. and Fleet, G.H. (1990) A convenient microtitre tray Török, T. and King, A.D. (1991) Comparative study on the identification procedure for yeast identification. Journal of Applied Bacteriology of food borne yeasts. Applied and Environmental Microprocedure 68, 447 451. biology 57, 1207 1212. Jones, J.B., Chase, A.R. and Harris, G.K. (1993) Evaluation of the Tudor, A. and Board, R.G. (1993) Food-spoilage yeasts. In The Biolog GN microplate system for identification of some plant- Yeasts Vol. 5 ed. Rose, A.H. and Harrison J.S. pp. 435 515. pathogenic bacteria. Plant Disease 77, 553 558. London : Academic Press.