The substrates used were 50 mM CDNB, 20 mM PNPA and 8 mM EA, giving final assay concentrations of 2.5 mM, 1 mM and 0.4 mM respectively. Measurements were conducted at wavelengths used previously.

The activities recorded are presented in three formats. Firstly, the activity was calculated as dABS per minute, with the value of the average uncatalysed reaction rate for the appropriate day subtracted. The full set of these values is given in Appendix B. The original results for obtained for each sample are also given. Due to the somewhat different assay conditions in the original study, the original readings were linearly extrapolated to account for the concentration of CDNB (2.703 mM) and enzyme (1:38 dilution) used.

Secondly, the activities thus calculated were logarithmically transformed. This was done because it had been previously observed that the distribution of GST activity is skewed towards lower values, with a median less than the mean. It was hoped that the logarithmic values would have a more normalised distribution.

Thirdly, the activities were scaled allometrically to give dABS/second/gram of tissue. This was done because it forms part of the current standard operating procedure (SOP) for GST analysis (Garrood, A.C., Beverley, M., Johnson, I., 1990). Enzyme activities and tissue wet weights (recorded in the original data) were logarithmically transformed and subjected to regression analysis to give an equation of the form:

log Activity = a + b log Weight

From this the activity of each sample was back extrapolated to give the activity per gram:

log Activity(corrected) = log Activity - (b log Weight)

Given that earlier studies had revealed a significant loss of activity, the original results recorded for the samples tested were also included in the analysis to provide an estimation of the loss of activity.

The results of the regression analysis of log weight vs log activity are given below.

The results for the upstream and downstream sites are presented below. They were assessed by a one way ANOVA and are represented graphically in Figures 5.1.

Means + S.E.M.

Means + S.E.M.
* Significant difference at 5% level.

The activities determined for the various substrates were linearly regressed in order to ascertain the relationship between them.

1. Comparison of different substrates

Some activity towards EA was detected in the digestive gland. The level of activity is in concordance with the results obtained previously in Sphaerium, where EA activity was found to be approximately 10% that of CDNB (Johnson et al., 1992). However in the downstream gills the activity is not significantly different from zero, and that in the upstream gills is only 2.5% of CDNB. Similarly PNPA activity in the digestive gland is around 20-25% of CDNB activity, and activity in the gills is around 10% of that with CDNB.

It is likely that both EA and PNPA are substrates of an isoenzyme which is either inactive towards CDNB or is only partly responsible for CDNB activity. Analysis of covariance shows no significant correlation of the activity towards the substrates in individual mussels. This could indicate that the isoenzymes responsible are differentially regulated from the isoenzyme(s) responsible for CDNB activity.

At first this seems to hold some promise of the use of an assay based on differential induction to provide additional information on pollutant exposure. Unfortunately it seems that only gill GST is enhanced in response to waterbourne toxicants. The activity towards EA is too low to be of any use (though this study does suggest the possibility of an extremely simple binomial test: Activity = clean, no activity = polluted). The activity with PNPA shows a 50% reduction in downstream samples, but the difficulties of detecting such a low level of activity obscures any significance.

2. Effects of different transformations

It is apparent that there is very little correlation between tissue weights and activity. This mirrors previous studies (eg Crane et al., 1993) where tissue wet weight is uncorrelated with activity. This result is puzzling, but it is evident that whatever the sample weights represent they are not an accurate measure of the size of the glutathione conjugating organ. Either GST is localised in a small portion of the gill, or it may well be the case that, during the blotting procedure, a variable amount of fluid is lost across the thin gill membrane. It may also be due to the fact that the dilution factor takes no account of tissue water. All extracts are assumed to be in a volume of 5 ml, whereas the actual volume is 5 ml + tissue water. The amount of tissue water is larger for large samples, thereby leading to a greater dilution and lower estimated activity. The final possibility is that the fluid content of the tissues may naturally show a marked variance.

The effect of this low correlation is that the regression line is placed almost at random, and so the effect of allometric scaling is to introduce a randomising factor and so reduce the significance of the results.

On the other hand, a logarithmic transformation effects a substantial reduction in the standard error of the mean (by normalising the distribution) and enhances the significance. This approach is not appropriate at very low levels of activity, since some samples will give an apparent negative dABS. These cannot be transformed, and so a bias is introduced into the data set.

3. Comparison of Original Data with New Data

The correlation between the two sets of data is very good (r=0.85 to 0.95, p<0.001). The regression equation gives indicates the new estimation for activity in the digestive gland is around 63% of the original estimation. The gill data is around 125% of the original, suggesting that activity has increased over the intervening period. This peculiar result is probably due to the difficulty in converting previous results to the equivalent in the new assay conditions. However it is noticeable that in the original study the activity in the digestive gland was around twice that of the gills, whereas in this study the activity was found to be approximately equivalent. This is probably due to the presence of proteolytic enzymes in the digestive gland, which would degrade the enzyme possibly while frozen, but more likely when the samples were defrosted during both the current and previous analyses.

An additional point worth noting is that significance of the new results for the gills is around double that of the original results. Although this may well be a stochastic effect, various modifications of the assay procedure have been identified and used in this study to enhance the accuracy of the results:

1. Measurement of the background rate. Although this will not affect the relationship of the untransformed data, the logarithmic and allometric transformations will produce different results when the background rate is subtracted.

2. Semi-automation of the procedure. The automated timing and reading of the results reduces the potential for operator error. This also increases the rapidity of the assay since whilst one sample is read, the next can be prepared.

3. Restriction of measurements to 90 seconds. To be an accurate measurement of activity, it is essential that the increase in absorbance be approximately linear over the duration of the readings. Highly active samples are often non-linear over three minutes, making the assessment of activity lower than that actually present. The standard operating procedure calls for three readings to be taken over the three minutes, and the measurements to be repeated at a lower extract concentration if they are non-linear. In practice this estimation of linearity is very difficult, and so errors are incurred. The simple expedient of measurement over a shorter time period virtually eliminates this problem, and also means that the assay is more rapid.

4. Monitoring of temperature. The room in which the spectrophotometer is housed is not temperature controlled, and was found to warm substantially through the day due to operation of the equipment. Although the experiment was designed to minimise any temperature effects it was found that the ambient temperature could be kept within a narrow band by simply leaving the door ajar as appropriate.

5. Repetition of measurements where replicates disagree by more than 10%. There are two reasons for the large intra-sample variations encountered, both of which are largely specific to this experiment. Firstly, low readings are susceptible to random fluctuations. Secondly, storage of frozen extracts appears to denature the protein in the sample, leading to large conglomerates of suspended particulate matter. This made accurate pippetting difficult. A critical disagreement level of 10% was chosen because it is easy to assess, and it resulted in a substantial decrease in random variability.

o Two alternative substrates has been identified, but the low level of activity limits their utility as a test substrate.

o Allometric scaling reduces test significance, but logarithmic transformation appears to improve the test.