Membrane Filtration Test for Rapid Presumptive Differentiation of Four Candida Species

Abstract

A rapid enzymatic two-step test for the presumptive differentiation of four Candida species commonly occurring in various clinical samples is described. The technique involves membrane filtration of a liquid sample, followed by preincubation of the membrane filter on Sabouraud glucose agar supplemented with ticarcillin-clavulanic acid to yield microcolonies. In a separate assay step, parts of the filter are placed on absorbent pads impregnated with fluorogenic 4-methylumbelliferyl (4-MU) enzyme substrates (4-MU-N-acetyl-β-d-galactosaminide, 4-MU-phosphate, 4-MU-pyrophosphate, and 4-MU-β-d-galactoside) in combination with 0.1% digitonin acting as a membrane permeabilizer. The membrane filter in contact with the assay medium is incubated to allow cleavage of the enzyme substrate, resulting in fluorescent microcolonies under long-wavelength UV light. This approach, tested on 301 clinical samples, is able to presumptively differentiate C. albicans, C. glabrata, C. krusei, and C. tropicalis and to distinguish them from other Candida spp. in about 9 to 11 h. Overall agreement with the conventional methods of 94.4% (one Candida species present in the sample) to 83.8% (multiple Candida spp. present) was obtained. The false-negative rates with reference to identification by traditional methods were 1.3% (single species) and 3.8% (multiple species).

Yeast infections increasingly cause morbidity and mortality in immunosuppressed patients. Predisposing factors include underlying neutropenia and broad-spectrum and cytotoxic chemotherapy as well as the use of central venous catheters (2, 3, 5). The higher incidence and severity of yeast infections has led to an increase in the therapeutic and prophylactic use of antimycotics. Fluconazole is the standard treatment for candidiasis (5, 9). However, as Candida glabrata has become less susceptible to fluconazole, and as Candida krusei is intrinsically resistant to this drug, infections by these strains may necessitate alternative treatment with amphotericin B or itraconazole (5, 9).

Among the etiologic agents of yeast infections, C. albicans, C. glabrata, C. krusei and C. tropicalis predominate in many countries (5, 9), including Belgium (15). Hence, given their different susceptibilities to antimycotics, rapid differentiation between Candida spp. may be important, particularly in systemic candidiasis.

At present, the identification of Candida spp. is based on a variety of tests, including germ tube formation, production of chlamydospores, sugar and nitrate assimilation, biochemical reactions, and agglutination with specific antibodies (3, 6, 8). All these methods require pure cultures of the isolates and, consequently, take 24 to 48 h to complete.

So far, the most rapid biochemical tests rely on the detection of certain key enzymes. The differentiation between albicans and non-albicans spp., in particular, is frequently made according to the presence or absence of N-acetyl-β-d-galactosaminidase (4, 13, 18). This enzyme can be demonstrated starting from a pure culture by inoculating a test vial or strip containing a chromogenic or fluorogenic N-acetyl-β-d-galactosaminide substrate (6, 10). Alternatively, the primary isolation can be carried out on a selective agar already supplemented with this substrate, as for example in Albicans ID (bioMérieux Vitek, Hazelwood, Mo.) (2), CHROMagar Candida (CHROMagar Company, Paris, France), and Fluoroplate Candida agar (Merck, Darmstadt, Germany). These media contain chromogenic or fluorogenic substrates, for N-acetyl-β-d-galactosaminidase in C. albicans and for other unspecified key enzymes occurring in C. glabrata, C. krusei, and C. tropicalis, thus allowing a presumptive identification of these four species in 24 h (17, 19, 22).

These data suggest that the speed-limiting factor of the enzymatic tests is not the detection procedure itself but, rather, the growth phase preceding or accompanying it. Hence, our aim was to reduce the latter without compromising the detectability of the enzymes.

To this end, we adapted a previously reported procedure for the enzymatic detection of Escherichia coli and coliforms on membrane filters (16). An unusual, speed-enhancing feature of this method is the separation of the growth step from the actual enzyme assay step. In such a two-step approach, the membrane filter containing the bacteria is first incubated on a selective growth medium to yield microcolonies. The second step involves probing of a specific enzyme activity with a fluorogenic substrate in the presence of a membrane permeabilizer. Conversely, in a one-step enzymatic approach, the enzyme substrate is directly incorporated in the agar and cleaved in the course of growth. The gain in speed of a two-step test with reference to a one-step test for E. coli and coliforms is approximately 6 to 10 h, i.e., a reduction from 18 to 24 to 10 to 12 h.

This two-step concept has now been applied to the detection of enzymes in Candida spp. on a membrane filter. Bobey and Ederer reported the cleavage of 4-methylumbelliferone-derived substrates by a variety of enzymes in Candida spp. (4). We found that the combined use of three enzyme activities, i.e., N-acetyl-β-d-galactosaminidase, acid phosphatase, and pyrophosphatase, allowed differentiation of C. albicans, C. krusei, and C. tropicalis with a high degree of accuracy. Furthermore, C. glabrata was detected on the basis of its previously unrecognized orange fluorescence on membrane filters under long-wavelength UV light. Coupling this enzyme assay to microcolony formation permitted the presumptive identification of the four species in 9 to 11 h.

This study describes the development and the optimization of this test as well as the application to samples taken from a variety of hospitalized patients suffering from mucosal candidiasis. Preliminary results with spiked blood samples also indicate its potential applicability to candidemic subjects.

MATERIALS AND METHODS

Test strains for method development.

A total of 81 Candida spp. laboratory strains was used for the method development, i.e., C. albicans (n = 19), C. glabrata (n = 17), C. krusei (n = 18), C. tropicalis (n = 17), C. parapsilosis (n = 8), and C. lusitaniae (n = 2). The strains were obtained from Mycothèque de l’Universite Catholique de Louvain-la-Neuve (MUCL) (Louvain-la-Neuve, Belgium) (C. albicans 29800, 29903, 29919, 29981, 30112, and 30114; C. glabrata 27865, 29833, and 15664; C. krusei 29991, 29954, and 30068; and C. tropicalis 29817, 29870, 29893, 29952, and 30002), from IHEM (Brussels, Belgium) (C. krusei 1796, 3955, 4221, and 4562; C. tropicalis 4139, 4222, 4225, 4529, and 4550; C. parapsilosis 2305, 4224, 2052, 1716, 4223, 4024, 4606, and 6395; and C. lusitaniae 3974 and 3975), and from the American Type Culture Collection (Manassas, Va.) (C. albicans ATCC 749). Non-Candida yeast species tested included Trichosporon cutaneum (n = 4), Geotrichum candidum (n = 4), Cryptococcus neoformans (n = 6), and Saccharomyces cerevisiae (n = 3). They were obtained from the Institute of Hygiene and Epidemiology Mycology (IHEM) (T. cutaneum 2310, 2625, 3002, and 3988; G. candidum 1178, 1484, 3001, and 3136; C. neoformans 3552, 3553, 3554, 3555, 3409, and 4216; and S. cerevisiae 564, 2683, and 3179). Other strains were isolates from our own collection and were originally derived from patients from the Department of Otorhinolaryngology, University Hospital of Ghent, Ghent, Belgium.

Clinical specimens.

Clinical specimens were provided by the departments of internal medicine, division of infectious diseases; head and neck surgery; gastroenterology; and hematology of the University Hospital of Ghent. Samples included oral swabs (n = 224), bronchoalveolar liquid (n = 2), sputum (n = 7), esophageal biopsies (n = 7), esophageal brushings (n = 4), and tracheoesophageal voice prostheses (n = 57) (12, 23, 24).

Blood samples.

Blank serum and whole blood were supplemented with each of the four target Candida spp. in concentrations of 10² and 10³ cells per ml. Samples were diluted tenfold with physiological saline before filtration.

Filters, growth media, chemicals, and reagents.

Nylon membrane filters (47-mm diameter, 0.45-μm pore size) (Gelman Sciences, Ann Arbor, Mich.) were used for filtration of the samples. Other filters, including nitrocellulose, polyamide, polysulfone, mixed cellulose esters, and polytetrafluorethylene filters, were obtained from Millipore Corporation (Bedford, Mass.). Absorbent glass fiber pads were purchased from Gelman Sciences.

Sabouraud glucose agar (SGA) (Difco Laboratories, Detroit, Mich.) supplemented with 5,000 μg of a combination of ticarcillin (4,688 μg/ml)-clavulanic acid (312 μg/ml) (Timentin) (SmithKline Beecham Pharma, Genval, Belgium) (SGA-T)/ml was used as the standard growth medium. Other growth media included cornmeal agar with 0.5% Tween 80 (CMT) (Difco), yeast potato glucose agar (YPG) (Difco), SGA supplemented with chloramphenicol (SGA+) (bioMérieux Vitek), SGA (Difco) supplemented with gentamicin and vancomycin (Sigma Chemical Co., St. Louis, Mo.), Fluoroplate Candida agar (FCA), CHROMagar Candida (CHROMagar), Albicans ID, and SGA with a vitamin mixture (SGA-B), containing 0.1 ppm biotin, 1.0 ppm panthotenate, 1.0 ppm thiamine, 0.1 ppm nicotinic acid, and 1.0 ppm PABA (all extra ingredients from Sigma).

The following 4-methylumbelliferyl (4-MU) derivatives were tested as enzyme substrates: acetate, α-l-arabinofuranoside, β-d,d′-diacetylchitobioside, phosphate, β-d-galactoside, β-d-glucoside, α-l-iduronide, β-d-lactoside, oleate, α-l-rhamnopyranoside (all from Sigma), pyrophosphate, β-d-glucuronic acid, heptanoate, laurate, N-acetyl-β-d-galactosaminide, and palmitate (all from Melford Laboratories, Ipswich, United Kingdom).

A 0.1- to 1-mg quantity of each substrate was dissolved in 1 ml of dimethylsulfoxide (Fluka, Buchs, Switzerland), except 4-MU-palmitate and 4-MU-oleate, which were dissolved in dimethylformamide (Merck) (4). Stock solutions were diluted in 0.1 M citrate-phosphate buffer (pH 3.4 for 4-MU-phosphate and pH 4.5 for all other substrates) and sterilized by filtration over Nalgene disposable units (0.45-μm pore size, 250 ml) (Nalge Co., Rochester, N.Y.). These solutions were dispensed in test tubes and frozen at −20°C until use.

Procedure.

All clinical specimens were processed within 30 min of arrival. Swabs and other solid samples, e.g., biopsies and voice prostheses, were extracted by vortex mixing in 10 ml of physiological saline. The samples were filtered over a 47-mm-diameter nylon membrane filter with a pore size of 0.45 μm. After filtration, the membrane was placed on SGA-T medium and incubated for 9 to 11 h to yield microcolonies. The filter was subsequently removed and cut in four pieces. Each piece was placed on an absorbent fiberglass pad impregnated with 340 μl of a buffered solution of an enzyme substrate supplemented with 0.1% digitonin (Sigma), acting as a membrane permeabilizer, and 1 mM MgCl₂. The four substrates were 4-MU-N-acetyl-β-d-galactosaminide, 4-MU-phosphate, 4-MU-pyrophosphate, and 4-MU-β-d-galactoside. After incubation for 30 min at 30°C, the filter was sprayed with 1.2 M sodium hydroxide and inspected under a 366-nm UV lamp. Blue fluorescent microcolonies indicated a positive reaction. The whole assay phase, including incubation, took about 45 min.

Identification methods.

Specimens were inoculated by direct swabbing (oral swabs) or by streaking a loopful of cells from a liquid extract (other samples) on SGA+ and CHROMagar. Typical yeast colonies were identified on the basis of the following criteria: germ tube reaction, morphology on CMT, and patterns of sugar assimilation (3, 6, 14). Isolates that were chlamydospore and/or germ tube negative were further characterized with the API 20 C AUX yeast identification panel (bioMérieux Vitek) (21). Results were read after 24, 48, and 72 h. The confirmation of C. krusei was done by agglutination with the Krusei-Color test kit (Fumouze, Levallois Perret, France) (8). All cultures identified as C. albicans were grown on SGA at 45°C to distinguish them from C. dubliniensis which, unlike C. albicans, is unable to grow at 45°C (20).

RESULTS

When C. albicans was detected in a one-step streak plate procedure with FCA or SGA supplemented with a fluorogenic substrate for N-acetyl-β-d-galactosaminidase, intense fluorescent colonies were obtained after approximately 23 h. In a two-step membrane filtration approach involving preincubation on SGA followed by an enzyme assay with 4-MU-N-acetyl-β-d-galactosaminide as a substrate and digitonin as a membrane permeabilizer, fluorescent microcolonies of C. albicans (three strains) became visible after 9 h, their number approaching a plateau after 11 h. The contribution of the membrane filter to this increase in detectability with reference to plate streaking was demonstrated by placing a nylon membrane filter loaded with C. albicans (five different strains) on FCA and SGA supplemented with 4-MU-N-acetyl-β-d-galactosaminide. Fluorescent microcolonies showed up from about 12 (n = 3) to 14 (n = 2) h onwards. Numbers were constant after 14.5 h (n = 3) and 18 h (n = 2). The use of digitonin as the membrane permeabilizer in the assay step further reduced the detection time by 3 h.

The effects of the nature of the membrane filter and the membrane permeabilizer were assessed on the basis of the visually scored fluorescence intensity of the microcolonies obtained for three strains each of C. albicans, C. glabrata, C. krusei, and C. tropicalis. Among the membrane filters tested, nylon proved superior to cellulose nitrate, mixed cellulose esters, polyamide, polysulfone, and polytetrafluoroethylene.

Digitonin was chosen as the membrane permeabilizer from several candidates, including chloroform, sodium lauroyl sarcosinate, toluene, and amphotericin B (1, 7, 11).

For preincubation, four growth media were compared, i.e., SGA, CMT, YPG, and SGA-B. SGA gave the best results in terms of growth-promoting activity, as tested in liquid medium by turbidimetry (data not shown). However, when SGA was used in connection with clinical samples, it frequently permitted the growth of bacteria, which could potentially interfere with the enzymatic reactions. Chloramphenicol, gentamicin, and vancomycin were only partially effective in suppressing growth of five selected bacterial strains, i.e., Salmonella, Bacillus, Streptococcus, Pseudomonas, and Staphylococcus spp., and of the bacteria present in all clinical samples tested so far. In contrast, the combination of ticarcillin-clavulanic acid (Timentin) totally inhibited the growth of these five species and that of the bacterial contaminants in the clinical samples but did not affect the growth of the yeasts. SGA-T and SGA yielded comparable growth after 24 and 48 h for C. albicans, C. glabrata, C. krusei, and C. tropicalis, suggesting that the antibacterial agents had no negative effect.

The enzyme profiles of 219 laboratory strains and previously identified clinical strains belonging to the four target Candida species are listed in Table 1. The reactions with 4-MU-N-acetyl-β-d-galactosaminide, 4-MU-phosphate, and 4-MU-pyrophosphate afford excellent sensitivity (number of reactive strains) for C. albicans (100%), C. krusei (100%), and C. tropicalis (96%), respectively. No specific substrate was found for the tentative identification of C. glabrata. However, it was observed that in 87.5% (34 of 39) of the cases studied, colonies of this species exhibited native orange fluorescence on nylon membrane filters placed on SGA, unlike on other media. Although N-acetyl-β-d-galactosaminidase and pyrophosphatase show a relatively high degree of specificity for C. albicans and C. tropicalis, respectively, the differentiation of the four target Candida spp. requires the combined use of three substrates. Aberrant reactions were observed with a limited number of laboratory strains. Seven percent of C. albicans laboratory strains were pyrophosphatase positive, whereas 11% of C. tropicalis strains contained N-acetyl-β-d-galactosaminidase, and 4% were pyrophosphatase negative. C. parapsilosis and C. lusitaniae displayed the same profiles for the three key enzymes as did C. krusei and C. glabrata, respectively, but unlike the latter two also had β-galactosidase activity (Table 2). Also listed in Table 2 are the enzyme activities of other, clinically important yeasts not belonging to the genus Candida. Based on the four enzyme activities studied, T. cutaneum, G. candidum, and C. neoformans were indistinguishable from C. albicans, C. krusei, and C. krusei, respectively.

TABLE 1.

Enzyme activities in Candida species

Substrate (4-MU-derivative)	% of reactive strains
	C. albicans (n = 129)	C. glabrata (n = 39)	C. krusei (n = 25)	C. tropicalis (n = 26)
Acetate	100	92	68	69
Arabinofuranoside	5	0	0	0
Diacetylchitobioside	0	0	0	0
Galactoside	2	0	0	0
Glucoside	9	8	0	0
Glucuronic acid	0	3	0	0
Heptanoate	98	82	60	80
Iduronide	0	0	5	0
Lactoside	6	0	8	0
Laurate	92	100	44	54
N-acetylgalactosaminide	100	0	0	11
Oleate	7	59	28	0
Palmitate	5	3	2	0
Phosphate	57	15	100	65
Pyrophosphate	7	0	0	96
Rhamnopyranoside	0	0	0	0

TABLE 2.

Enzyme profiles of other yeast species tested

Reaction	% of positive strains
	C. parapsilosis (n = 14)	C. lusitaniae (n = 4)	T. cutaneum (n = 4)	G. candidum (n = 4)	C. neoformans (n = 7)	S. cerevisiae (n = 4)
N-Acetyl-β-d-galactosaminidase	0	0	100	0	0	0
Acid Phosphatase	100	0	100	100	100	0
Pyrophosphatase	0	0	0	0	0	0
Orange fluorescence	0	0	0	0	0	0
β-d-Galactosidase	93	100	0	0	0	100

The present method has been applied to 301 clinical specimens of mucosal origin containing either a single Candida species, multiple species, or no yeasts at all. The results are listed in Table 3. Overall agreement between the two-step and the conventional methods was 94.4 and 83.8% for single and multiple species, respectively. Calculated per individual species, the percentage agreement was 97.8 (C. albicans, n = 137), 78.6 (C. glabrata, n = 14), 87.5 (C. krusei, n = 8), and 100 (C. tropicalis, n = 3) for samples containing a single species. For mixed populations, the corresponding values were 97.1 (C. albicans, n = 35), 69.2 (C. glabrata, n = 13), 100 (C. krusei, n = 5), and 95.0% (C. tropicalis, n = 20). The false-negative rates for the two-step method with reference to identification by conventional methods were 1.3 (single species) and 3.8% (multiple species), respectively. In 2.6% of presumably negative specimens, the two-step method detected a single Candida sp. that was not detected by the streak or swab reference plate method. For multiple species, this apparent higher sensitivity of the two-step procedure increased to 10.0%. However, these results were not considered false positives, as they were confirmed by reincubating the membranes on fresh SGA and identifying the colonies formed by the panel of traditional biochemical tests. Among other Candida spp., C. parapsilosis and C. lusitaniae have not been isolated so far from clinical samples, unlike C. maris (n = 3), C. kefyr (n = 3), and C. guilliermondii (n = 2). These three species were phosphatase positive, and the latter two also yielded orange fluorescence. However, they all exhibited β-galactosidase activity, so that confusion with C. krusei and C. glabrata was excluded. None of the three potentially interfering non-Candida yeasts mentioned in Table 2 was found in clinical samples, except for Cryptococcus neoformans, which was isolated once. A general overview of the useful enzymatic reactions required for the differentiation of the four target Candida species and for their distinction from other Candida spp. is given in Table 4.

TABLE 3.

Numbers of species identified by the two-step and conventional methods

Identification by two-step method	No. of specimens (with single/multiple species) identified by conventional method as:
	C. albicans (n = 137/35)	C. glabrata (n = 14/13)	C. krusei (n = 8/5)	C. tropicalis (n = 3/20)	Negative (n = 69/7)
C. albicans	134/34	2/0	1/0	0/0	6/5
C. glabrata	0/1	11/9	0/0	0/0	0/1
C. krusei	1/0	0/1	7/5	0/1	0/1
C. tropicalis	0/0	0/0	0/0	3/19	0/1
Negative	2/0	1/3	0/0	0/0	63/NAa

^aNA, not applicable. 

TABLE 4.

Combined positive and negative reactions required for the differentiation of four Candida species from each other and from other Candida spp.

Reaction	Test result with:
	C. albicans	C. glabrata	C. krusei	C. tropicalis	Others
N-Acetyl-β-d-galactosaminidase	+	−	−
Acid phosphatase		−	+
Pyrophosphatase		−	−	+
Unknown mechanism		+a			+ab
β-d-Galactosidase		−	−		+c

^aOrange fluorescence. 

^bC. kefyr and C. guilliermondii. 

^cC. kefyr, C. guilliermondii, C. maris, C. parapsilosis, and C. lusitaniae. 

When the procedure was applied to blood or serum samples supplemented with any of the four target Candida spp., no quenching of fluorescence of the microcolonies was noted. In addition, good recoveries with reference to CHROMagar were obtained, i.e., 110.5% (C. albicans, n = 4), 111.3% (C. glabrata, n = 2), 92.0% (C. krusei, n = 2), and 103.0% (C. tropicalis, n = 2) from serum and 105.5% (C. albicans, n = 4), 98.5% (C. glabrata, n = 2), 98.5% (C. krusei, n = 2), and 110.0% (C. tropicalis, n = 2) from whole blood. Detection times in both tests were 12.5 h for the two-step approach and 24.5 h for the CHROMagar test.

DISCUSSION

The enzymatic test used for the rapid presumptive differentiation of Candida spp. in clinical specimens has several unusual aspects. First, it does not require a fully developed pure culture of the isolate or visible colonies on an agar medium but involves the formation of microcolonies on a nylon membrane filter. Second, although the procedure is agar based, the enzymatic activity is demonstrated not on the plate in the course of yeast propagation, as in the commercial CHROMagar, Albicans ID, and FCA, but in a separate assay step. Third, in this assay, a membrane permeabilizer is used to improve the cellular uptake of the substrate (1, 7, 11).

The combined effect of growth on a membrane filter and the use of fluorogenic substrates and a membrane permeabilizer, as well as the fluorescence enhancement by the microenvironment of the nylon filter, accounts for the enhanced detectability and, hence, the speed inherent in the present procedure. This increased detectability of enzyme activity minimizes the number of cells required, so that in turn the growth phase can be reduced to the level of microcolony formation. As a result of the interplay between all the above-mentioned phenomena, the present procedure requires less than half the time required by the existing enzymatic tests for Candida. For example, a one-step approach on FCA or SGA supplemented with a fluorogenic enzyme substrate yielded fluorescence for C. albicans after 23 h. CHROMagar containing chromogenic substrates also requires 24 h of incubation.

The contribution of the membrane filter to the gain in speed was estimated by determining the time required for fluorescence development on FCA or SGA inoculated by streaking and on a membrane filter placed on these media. A difference of approximately 9 h between the membrane filtration method and the direct inoculation method was obtained. Nylon filters displayed superior fluorescence characteristics, i.e., minimal background and maximum amplification of the signal compared to the other materials tested. The latter phenomenon is presumably rationalized by the hydrophobic microenvironment created by the nylon.

The use of digitonin as a membrane permeabilizer was useful to ensure the efficient cellular uptake of the substrate (1, 7). With digitonin, fluorescent microcolonies could already be visualized after 9 h, whereas without digitonin, the first microcolonies were detected after about 12 h. Because of its growth-inhibiting characteristics, digitonin could not be added to the preincubation medium. Consequently, the enzyme assay had to be performed in a separate step, combining the substrate and the membrane permeabilizer.

SGA was a useful selective medium for the preincubation step. However, to fully suppress bacterial growth, the addition of a broad-spectrum antibiotic was necessary. In theory, ticarcillin-clavulanic acid-resistant bacteria, exhibiting the key enzyme activities of the four target Candida spp., could interfere in the test. However, this hypothesis could not be verified experimentally because no such strains were available from our university hospital clinical microbiology laboratory. No bacterial growth has been observed in any of the 301 clinical samples tested so far.

The use of 4-MU-glycosides and esters to detect various hydrolases in microorganisms is widespread (4, 14, 18, 25). These substrates are cleaved to yield 4-MU, which is intensely fluorescent under long-wavelength UV light (366 nm) (4, 13). Bobey and Ederer had already investigated the cleavage of 17 4-MU substrates by Candida spp. (4). As a follow-up to their work, we tested 16 4-MU derivatives that differed partially from those used by Bobey and Ederer. However, both studies showed that most reactions were too nonspecific and/or too insensitive for practical application. They occurred either in different species or in only some of the strains within a given species. Still, three enzyme activities were useful for the differentiation of the four Candida spp. most commonly encountered in clinical practice, i.e., N-acetyl-β-d-galactosaminidase, acid phophatase, and pyrophosphatase. Unlike N-acetyl-β-d-galactosaminidase, the latter two were also recommended by Bobey and Ederer as useful identification tools in addition to β-d-glucosidase (4). The sensitivities (percentages of reactive strains) for N-acetyl-β-d-galactosaminidase, acid phosphatase, and pyrophosphatase in C. albicans (n = 123), C. krusei (n = 25), and C. tropicalis (n = 26) were 100, 100, and 96%, respectively. The specificity of the three enzyme activities is such that, in principle, only the combination of the three reactions can differentiate C. albicans, C. glabrata, C. krusei, and C. tropicalis. The addition of 4-MU-β-d-galactoside as a fourth substrate makes possible the distinction of the four above-mentioned Candida spp. from a group of other Candida spp., including C. kefyr, C. guilliermondii, C. maris, C. parapsilosis, and C. lusitaniae. However, no further differentiation on an enzymatic basis is possible within this group. No enzymatic reaction was sufficiently specific and sensitive to allow a presumptive characterization of C. glabrata. However, the previously unrecognized native orange fluorescence of C. glabrata on a membrane filter was a more useful criterion. Interestingly, this phenomenon was observed only with the combination of nylon filters and SGA, not with other membrane filters and growth media. Five of 39 isolates did not exhibit this orange fluorescence, but they were also acid phosphatase negative so they could not be confused with C. krusei. Possible confusion between C. albicans and C. tropicalis based on inconsistencies in the presence of N-acetyl-β-d-acetylgalactosaminidase and pyrophosphatase occurred only in laboratory strains and not in clinical samples. One isolate of C. tropicalis was misidentified as C. krusei because it was pyrophosphatase negative and acid phosphatase positive. The two-step method was also able to detect a mixture of two Candida species in the same sample. Fluorescent and nonfluorescent microcolonies were readily distinguishable from each other. For example, in a mixture of C. albicans and C. tropicalis, a portion of the microcolonies on the membrane filter showed a light blue fluorescence with 4-MU-N-acetyl-β-d-galactosaminide, while on a second piece of the filter the other portion of the microcolonies showed fluorescence with 4-MU-pyrophosphate.

The application of the two-step method to 301 clinical samples showed good agreement with the conventional identification methods. Given the lack of a positive reaction other than its orange fluorescence, the lower value for C. glabrata is not surprising. Particularly in mixed cultures, the orange fluorescence may be obscured by overgrowth by another species and/or the latter’s blue fluorescence. A salient feature of the present method is its low false-negative rate with reference to conventional methods. In addition, the two-step method has a substantially lower detectability, the theoretical limit being one cell per membrane filter and, hence, per sample, provided the yeasts are quantitatively eluted in physiological saline. This increased sensitivity rationalizes the higher number of positives found with reference to the traditional methods, by which the sample had not been concentrated but was directly applied on the primary plate by swabbing or streaking. This discrepancy has meanwhile been overcome by the use of a modified isolation procedure in the conventional test, i.e., membrane filtration and incubation on CHROMagar. False-positive reactions could be caused by other, non-Candida yeasts, including T. cutaneum, G. candidum, and C. neoformans.

Due to this margin of uncertainty, the present test should be considered presumptive rather than definitive and should, in any case, be complemented by confirmatory classical procedures. However, its lack of absolute specificity would be outweighed by its substantially enhanced speed, particularly in critical situations of candidemia or systemic candidiasis. Although this test has been evaluated with more easily accessible but clinically less relevant mucosal samples, preliminary experiments with spiked blood samples have shown the feasibility of the two-step approach for the rapid detection of candidemia, where sensitivity and speed are crucial. In contrast, the interpretation and diagnostic values of the low number of Candida spp. capable of being demonstrated by this highly sensitive test with mucosal samples, particularly containing multiple Candida spp. must be carefully considered.