COSMETIC PRESERVATION 207 cells in clumps then disperse to result in single cells that give a higher CFU at 14 days but now are more susceptible to the biocide, and so a reduction follows at 21 days. Although this model needs further testing, it is satisfying that it logically describes what has heretofore been passed off as experimental error or biological variability or the development of resistance. In doing plate counts, one can use either pour plates or spread plates to determine how many CFU/gm survive. In pour plates, the diluted product (about 1 ml) is vortexed into a test tube of about 15 ml of melted agar at 46 -+ 2øC. The agar is then poured into a Petri dish. Alternatively, the dilution may be placed into the Petri dish and agar poured on top of it while the experimenter swirls the plate in a "figure 8" motion. With spread plates, the diluted product (about 0.1 ml) is spread onto prepoured solidified agar plates using a bent glass rod. Spread plating allows easy processing of samples. The main advantage is that it avoids exposing microorganisms to heated media. However, pour plating allows for more exposure to neutralizing agents in the agar. Some published information finds that the two methods give similar results (81,82). One can also perform enrichments of the product to detect low levels of potentially T=Od 18 CFU/ml T=14d • 21 CFU/rnl T=7d 10 CFU/ml T=21d 6 CFU/rnl Figure 1. Bacteria exist as clumps in Poisson distribution. This model helps explain anomalous results in PET testing. Each clump gives rise to a colony rather than each individual doing so. After 7 days, the CFU/ml is reduced due to the death of organisms existing outside of the protection of the clump. A few of the organisms at the periphery of the clump are killed, but the clump still forms an individual colony. The fact that some of the organisms in the clump died is not detected upon plating. After 14 days, the clumps break up to provide more CFU. Now the individuals are no longer within the protection of the clump and are more susceptible to exposure to the biocide. Therefore, at day 21 the total CFU is decreased.

208 JOURNAL OF THE SOCIETY OF COSMETIC CHEMISTS recoverable organisms (83). The sensitivity or detection limit of typical dilution and plate count methods is usually from 10 to 20 CFU/gm of product. In enrichment, at least 10 g of product is put into 1 liter of broth and incubated. Any turbidity (or color change if one uses either a pH or redox indicator) indicates at least one organism was present in the 10-g sample. This approach makes the detection limit 1 CFU/10 gm of product (theoretically 0.1 CFU/gm). Thus the sensitivity is increased by 100X com- pared to traditional plate count methods. It is most useful in determining if, after the 28-day period, low levels of inoculum still exist that may still be capable of growing later on. Adaptation and resistance. All of the recognized tests require long incubation periods (28-56 days). These long periods are supposed to account for the phenomenon of adaptation. After some lag phase the microbes "grow back" to high enough levels to be detected again. The mechanism(s) for this regrowth are not well understood. Perhaps it is due to survival and adaptation. Perhaps it is due to in situ recovery of injured organisms (84). It may be due to container-associated organisms that slough off into the product (85,86). It may even be due to inadequate mixing and inconsistent plating methods, since bacteria display Poisson distribution in the sample. The paradigm of organisms existing as clumps is also a possible explanation to help explain "grow-back," without needing to claim microbial adaptation, or recovery of injured cells or the "Phoenix Phenomenon" (87,88). These latter two explanations need not be the sole or even primary explanations the clumping paradigm also explains what appears to be anomalous results when cells die off but then "recover" during a PET. Whether or not the clumping paradigm is a more valid explanation for these anomalous results than adaptation or the "Phoenix Phenomenon" remains to be shown empirically. Certainly there are cases where adaptation occurs. However, where adaptation is claimed for preservatives that have multiple modes of action, resistance is rarely via an individual occurrence of plasmid acquisition, mutation, or lifting of repression (89), as is often found with antibiotics but rather is a result of enhancement of the expression of a characteristic within a population due to genetic drift. This may occur as a shift in the amount of capsule production, clumping, stimulation of production of glutathione, or even physical community developments within biofilms where certain organisms exist as protector guilds for other organisms (90-92). Such resistance is typified by whole-cell poisons such as chlorine (93,94). Few cases of true chlorine resistance occur (e.g., point mutations by a single mutant cell that survives). Instead, any "resistance" seen is really a population or community effect of cells existing within the protection of a biofilm or surrounded with a capsule composed of extracellular polymeric substances that excludes the chlorine or use of cellular energy to produce higher levels of glutathione (90,95). Perhaps in these examples a more proper term to use would be biocide "tolerance" rather than resistance. The establishment of a biofilm or clumps of organisms provides an adaptation mechanism for tolerance to biocides using extracellular polymers in the form of a capsule. This biofilm then leads to an inoculum source that is constantly being exposed to sublethal or subinhibitory levels of biocide. Once established, adaptation via increased production of glutathione or a slowdown of metabolism (or even perhaps mutation) can result in a resistant phenotype (or even genotype), and the problem becomes compounded (personal communication, J. S. Chapman, Rohm, and Haas). Genetic adaptation to biocides at the individual rather than population level is a pos- sibility in some cases (96). However, several papers claiming to have demonstrated this