Variability of Agar Dilution-Replicator Method of Yeast Susceptibility Testing

The agar dilution method of in vitro susceptibility testing of fungi was analyzed with a Steers-type inoculum replicator, ten strains, and three drugs. The replicator reproducibly delivered the same inoculum to each series of plates. The minimal inhibitory concentrations of ketoconazole (an imidazole) and 5-fluorocytosine, but not that of the polyene nystatin, were dependent on the initial inoculum size. With the former two drugs, but not with the latter, minimal inhibitory concentrations were highly variable depending on the time of reading. Results with agar and broth dilution methods were divergent, and the differences in minimal inhibitory concentrations were variable in serial comparisons by the two methods. If the agar dilution minimal inhibitory concentrations were determined at first appearance of control growth, a commonly used end point, small variations in the time of reading (as could occur by variation in observer perception of when initial growth appears) induced large variations in the minimal inhibitory concentrations of 5-fluorocytosine and ketoconazole, particularly with rapidly growing strains. Results at 35 and 30°C were similar. The differences in results with the three drugs suggest different mechanisms of action. The variability quantitated with the agar dilution method could result in variability in results between laboratories and even observers in the same laboratory.

A variety of in vitro susceptibility testing methods are available for use with fungi to determine the minimal inhibitory concentration (MIC) of a drug. These include broth dilution, agar dilution, and disk diffusion techniques. Of these, agar dilution is one of the more common methods used in clinical laboratories (3,4), but the variables inherent in this system have not been clearly defined.
Since the latter method enables MICs for multiple strains to be determined easily and thus offers certain advantages, the variables in this method were studied utilizing a Steers-type inoculum replicator. Ten fungal species were tested for susceptibility against nystatin, 5-fluorocytosine (5-FC), and ketoconazole. Ketocona Agents Chemother., 18th, Atlanta, Ga., Abstr. No. 51, 1978).
(This paper was presented in part at the 18th Interscience Conference on Antimicrobial Agents and Chemotherapy, Atlanta, Ga., October 1978.) MATERIALS AND METHODS Ten yeast strains were used: Candida albicans (seven strains), Torulopsis glabrata (one strain), and Candida parapsilosis (two strains). Nine strains were clinical isolates; one was a laboratory-passagedl strain of uncertain origin.
These strains were grown in yeast nitrogen base broth overnight at 35°C. The growth was then adjusted to 60% ransmittance at 540 nm, using a Coleman Junior spectrophotometer. This reading approximates a concentration of 2 x 10r cells per ml (1). It was necessary to first determine that the replicator apparatus could reproducibly deliver the same inoculum after the repeated dipping in the wells needed to inoculate several plates. Inocula of one Candida strain were prepared at 2 x 104 and 2 x 103 cells per ml as described, the seed block was filled, and 25 yeast nitrogen base plates were serially inoculated. To assay the delivery quantitatively, the contact point was removed from each plate by undercutting with a sterile surgical scalpel. The fament of agar removed was then placed in 10 ml of yeast nitrogen base agar in a tube at 560C. It was necessary to undercut as small a volume of solid agar as possible to ensure complete liquifaction in the diluent agar. The tube contents were then blended in a Vortex mixer and poured into petri dishes, which were then incubated at 350C. The resultant colonies were counted at 24 h.
In all experiments, control plates containing only yeast nitrogen base were inoculated at the beginning and end of each series of drug concentration. This provided a control in each run for changes in inoculation by the replicator, uneven mixing of the inoculum in the wells during inoculation, or loss of viability of portions of the inoculum during an experiment.
Broth dilution end points were determined at 48 h and then daily for 7 days, with the tubes being shaken before each reading. Agar plates were also examined daily; the MIC was defined as the lowest drug concentration at which no growth occurred.
In one experiment, the agar plates were inspected frequently, beginning during the first 24 h after inoculation, in an effort to determine the MIC at the time when growth was first apparent on the control agar plate. This point for MIC determination is often used in clinical laboratories. Further readings were made at frequent intervals after the initial reading to ascertain changes in MIC over the first 56 h (3). All readings in an experiment were performed by the same observer.

RESULTS
Reproducibility ofreplicator delivery. An experiment was performed with two inoculum sizes, by the method described, to study variability in replicator delivery from plate to plate and also possible differences induced by different depths of fluid in the wells after repeated sampling. Without refilling the wells, there were minute differences in inoculum delivery between plates 1 and 25 in a series. For example, from an inoculum of 2 x 10i organisms per ml, 16 colonyforming units were delivered to plate 1 and 15 colony-forming units to plate 25; with an inoculun of 2 x 103 organisms per ml, the counts were 5 and 4 colony-forning units, respectively.
Inoculum dependence of MIC by agar dilution method for some drugs. The effect of inoculum size on the MIC when determined by the agar dilution method was examined with six strains of yeasts. For these studies, the MIC was determined, by the criteria described, at 48 h. The inoculum sizes used represent a range of those in common use.
In studies with ketoconazole (Table 1), the effect of inoculum size was large with three strains. With these, the strains were interpreted as highly susceptible when 103 colony-forming units/ml was used as the inoculum but highly resistant when 10' colony-forming units/ml was used. Some effect of inoculum size was apparent with all strains tested, although the effect was modest with one strain which had a long generation time.
Similar studies were perforned with nystatin (Table 1). No effect of inoculum size was noted.
Results with 5-FC were similar to those with ketoconazole. However, the MICs for five of six strains were only 0.25 ,ug/ml even with the highest inoculum tested (105/ml). With three ofthese five strains the variation from 103 to 105/ml was only four to eight-fold. The MIC for the sixth With ketoconazole, and with several inocula, the MIC increased as the incubation time was prolonged ( Table 2). This occurred with both agar dilution and broth dilution methods.
In contrast, the MIC of nystatin for these strains was relatively constant over time ( Table  2). The modest increase observed at 48 h is consistent with the observed relative instability of nystatin and other polyenes at 35°C for prolonged periods.
Results with 5-FC again resembled those with ketoconazole. With five of six strains, however, the absolute variation in MICs was less marked since the MICs were c2 gg/ml with all inocula tested and with readings made during 48 to 168 h. With these, the MIC varied sfourfold between 48 and 168 h (five organisins, three inocula each), with three exceptions. With two, the variation was indeterminate because the initial MIC at 103/ml was below the range ofdilutions tested, and in the other the variation was eightfold. The MIC for one strain was 264 ,tg/ml with all inocula at 168 h, when readings were made and the variation between 48 and 168 h for this "resistant" strain was 1,000-fold with two of the inocula tested.
The data for the strains tested can be shown by comparing the percentage of strains showing a 2fourfold increase in the MICs of nystatin and ketoconazole at different time periods during incubation compared to the 48-h readings ( Table  3). The effect oftime on the MIC ofketoconazole in contrast to that of nystatin is readily apparent from the data presented.
Comparison of broth and agar dilution MIC8. A comparison of two commonly used methods, agar dilution and broth dilution, was performed with several organis, several inoculum sizes, and several times of reading, using ketoconazole and nystatin as previously shown to be inoculumand time-dependent and inoculumand time-independent drugs, respectively. The results can be displayed as the ratio of MIC in broth to that in agar (Table 4). There was no concordance between the two methods; the broth dilution method gave a consistently higher Ketoconazole 5/6 NT 5/6 6/6 Nystatin 0/6 0/6 1/6 1/6 a Six yeast strains were tested, with an initial inoculum of 103 organisms per ml. NT, Not tested.  Table 4 shows.
Effect of determining results at firs control growth and effect of small variations in time of initial reading on agar dilution MIC. As stated previously, agar dilution susceptibility testing results are often read at the time of first appearance of control growth. Studies were made at this point in time, serial readings were made at later periods, and comparisons were made with drug concentrations resulting in 50% inhibition (IC1/2).
Results with two representative strains and ketoconazole are shown in Effects of temperature on MIC determinations. These studies were repeated at 30°C with ketoconazole and nystatin, with several strains and several inoculum sizes, and the results were compared with those at 350C. Some laboratories use the lower temperature for susceptibility testing.
With six strains and ketoconazole, variation in the MIC, as measured at 30°C by the agar dilution method and dependent on inoculum size and time of reading, showed only minor differences from the data obtained from studies done at 350C. Inoculum dependence of the MIC result was less prominent at the inital reading (17.5 h). For example, inocula of 10i and 106 cells per ml gave identical results, whereas only the MIC with an inoculum of 103/ml showed the marked  In the course of these studies we also showed that the variables described in the broth dilution method at 370C with miconazole and amphotericin (1) were aLso present in the broth dilution method with ketoconazole and nystatin at 350C. DISCUSSION These results demonstrate several key variables in the agar dilution method of susceptibility testing of yeast not previously quantitated with these organisms.
MICs of ketoconazole and 5-FC, but not that of nystatin, were dependent on inoculum size. Variability of MIC results with time of incubation was seen with the former two drugs but not with nystatin. With ketoconazole and nystatin, there was consistently poor concordance in repeated serial determinations between agar dilution susceptibility test results and broth dilution susceptibility test results. If the agar dilution MICs were determined at first appearance of control growth, as is commonly done, small variations in the time of reading (as could occur by variation in observer perception of when initial growth appears) induced large MIC variations, particularly with rapidly growing strains.
Variability of results was generally greater with ketoconazole than with 5-FC. Against most of the strains used, MICs of 5-FC were consistently well below achievable plasma concentrations, whereas the variability of MICs seen with ketoconazole resulted in a range sufficiently large to encompass concentrations well below those achievable in plasma and concentrations well above those achievable in plasma (Brass and Stevens, unpublishe4.data). The8e mod1est differences in results between these two drugs may be due to the exquisite sensitivity of five of the six species used in tests with 5-FC. Growth could occur (the total mass dependent on inoculum size), but the rate could be so slow that at some concentrations it would not reach the visual threshold over the period of observation. Alternatively, at some drug concentrations growth in agar may plateau at concentrations of organism which are below the visual threshold.
The differences in results between 5-FC and ketoconazole may disappear if organisms with a broader distribution of susceptibility to 5-FC are used.
Thus, with some drugs, MIC results by the agar dilution method are dependent on several variables, and one's choices in controlling such variables will arbitrarily determine the final MIC result. This variability would result in considerable discrepancy in results between different laboratories and, with some variables, even between observers in the same laboratory. These variables are also present in the usual method of determining MIC results by dilution in liquid medium and by use of a visual end point, as previously described (1). A spectrophotometric method, which determines end point as a function of control growth, was shown to be free of dependence on inoculum size, and if read during log-phase growth in the controls, the end point did not vary during this interval (1).
Major differences were seen between 5-FC and ketoconazole on the one hand, which were dependent on all the variables indicated, and nystatin on the other hand, which was not dependent on these variables. Similar results were reported previously with the broth dilution method for 5-FC and miconazole (like ketoconazole, an imidazole drug), on the one hand, and amphotercin B (like nystatin, a polyene drug) on the other (2). It was shown in that system that the onset of growth inhibition by the polyene was uniform and independent of generation time ofthe fungus, in contrast to the other drugs. We hypothesize that these differences reflect mechanism of action, with 5-FC and imidazoles principally acting by metabolic inhibition and requiring actively growing organisms and polyenes acting by membrane binding, presumably with an irreversible effect at any phase of growth.
If correlations are to emerge between in vitro susceptibility and clinical results with antifungal chemotherapy, in vitro methods with uncontrolled variables are undesirable. We have found the spectrophotometric method (1) easy to perforn, and due to its independence of growth characteristics and inoculum size of the organisn, we suggest that this method should be used widely for in vitro susceptibility testing. If other methods are used, such as the agar dilution-replicator method, attempts should be made to control all variables (e.g., inoculum size) possible.