Summary of spatial arrangement analyses

arrangement analyses conducted on images were unexpected. It was thought that Slurry FISH images would exhibit a more random distribution than Aggregate FISH images, as measured by the pair cross-correlation values and nearest neighbor values. This was not the case. In all instances, the Slurry FISH images had g(r) values greater than 1,

indicating clustering, whereas the Aggregate FISH images had g(r) values approaching 1, indicating a random distribution. The most likely reason for this finding is the

association of cells and pyrene with soil particles in the Slurry FISH images.

The g(r) values were in contrast to the calculated indices of dispersion which were, on the whole, greater for Aggregate FISH. There are many possible explanations for this finding. First, the analysis did not take into account the association between pyrene, cells, and soil particles. In Aggregate FISH images, this association could not be comfortably made without confocal microscopy. It was not possible to distinguish between cells and/or pyrene floating in a plane above the soil. In Slurry FISH images, there was a distinct association of pyrene and soil particles (see section 4.4.9.2 above). This phenomenon would explain the large g(r) values calculated at the shortest distances in the slurried samples and may also explain the large confidence intervals obtained for these samples. Whether this association was due to the increased contact between pyrene, cells, and soil particles during the fixation step of Slurry FISH or to the

possibility that cells and pyrene were originally adherent to soil in the static incubation is unknown. Secondly, it is possible that the distribution of pyrene itself in the incubating

microcosms was random. In this case, cells responding to pyrene presence would also assume a random distribution. This phenomenon would not be observed in the Slurry FISH images, since the original spatial relationships in those samples would have been destroyed. Thirdly, the daime program has been utilized on images at a lesser degree of magnification than those of this study. It is possible that, over longer distances, the pair cross-correlation values of Aggregate FISH samples would show a greater tendency toward non-randomness. Initially, images were taken at a lower level of magnification in addition to 1000X. However, at this level, it was not possible to distinguish between cells, pyrene, or soil particles, such that these images could not be used for analysis. Fourthly, it is also possible that the greater areas occupied by cells and pyrene in the Aggregate FISH images led to lower cross-correlation values, even though this value is supposedly normalized for density (see equation in section 4.4.8). Not only do these ratios decrease the g(r) value, but the numbers affect the totals of hits versus misses, which can also lower the cross-correlation value. Finally, Reed and Howard (1999) considered g(r) values of ~1.8 to be highly positively correlated. Therefore, values such as those found in Aggregate FISH samples could very well indicate clustering, but not as significantly as with Slurry FISH.

4.5 Method Limitations The chief limitation to the Aggregate FISH method is its cost.

Aggregate FISH requires a greater amount of probe than does Slurry FISH. Additionally, success of the embedding reagent is dependent upon soil type and moisture content. The higher the sand and the lower the clay contents of the soil, the easier the acrylamide can penetrate into a core. This was true for the CMN soil compared with the KKY soil and

especially with the CNC soil. Unfortunately, a textural analysis was not available for the latter. The presence of moisture also causes the acrylamide embedding to be less

effective. The samples for this study contained 70% of each soil’s water holding capacity. The higher a soil’s clay content, the higher that holding capacity will be, thus exacerbating the problem of penetration. However, the embedding was shown to be effective on a separate clay soil (21%) that was dry. It appears that clay’s expandability in the presence of moisture precludes acrylamide penetration and makes the embedded core less stable. Most FISH protocols have dehydration steps involving ethanol.

However, these were not feasible for the contaminated soils, as ethanol tended to extract PAHs from the samples. Ideally, cores could be re-embedded in resin, then sectioned and examined with confocal laser scanning microscopy.

Because the CNC was not characterized, it is also difficult to say whether the differences in probe hybridization were due to the soil’s physico-chemistry. However, core integrity was a problem when processing this soil, and it is likely that this is a key cause of differences between this soil and the others.

An additional cause of bias was the lack of random image generation. In order to compensate for the lack of an automated stage, the stage was moved manually while visualizing the sample with the bright field filter. The direction was usually in a

“stairstep” pattern (up/down and over), until the edge of the sample was reached; then the stage was shifted either left or right, and the stairstep repeated. It is possible that samples were overcounted as a result of this process, but this error should still have been