For all experiments, the pH ranged between 7.4 and 7.9; the oxygen content was always higher than 75% saturation; and the temperature ranged between 12.2°C and 14.0°C. The measured pH was close to the optimum (7.2 to 7.8) for gammarids given by Schellenberg [[25]]. Oxygen content and temperature of the test medium fulfilled the conditions preferred by G. pulex [[26]].
General findings
The overall feeding rate of the first experiment (impact of food source and parasite infection) was 0.21 ± 0.14 mg (food)/(mg (gammarid) × d) without distinguishing the influence of food source and parasite infection (Figure 1, column 2). This overall feeding rate ranged between 0.17 ± 0.13 and 0.37 ± 0.20 mg (food)/(mg (gammarid) × d) when calculated on a daily basis. A reduction of the variability of the test results by 1.6% was observed on discarding the first feeding period from the data analysis (Figure 1, columns 1 and 2). Data for the first feeding period (t0h to t24h) were excluded from further analysis because of a significant difference in feeding rate relative to subsequent periods.
The overall feeding rate in the next experiment (impact of body mass) without distinguishing on body mass was 0.37 ± 0.28 mg (food)/(mg (gammarid) × d) (Figure 1, column 8) and ranged between 0.23 ± 0.25 and 0.51 ± 0.54 mg (food)/(mg (gammarid) × d) when calculated on a daily basis.
Overall, the intra-specific variability in feeding rate was very high. The standard deviations of the measured feeding rates were 66% and 75% of the average values for the two experiments described above when no differentiation in the three tested factors (food source, parasite infection and body mass) was made (Figure 1, columns 1 and 8). The large variability indicated a low statistical power in any tests of individual feeding rate of G. pulex at a daily resolution and thus the need for further work to understand and reduce intra-specific variability.
Food source
The feeding rate of non-infected gammarids was significantly influenced (p < 0.001, power = 1, overall difference amongst treatment groups; ANOVA, Holm-Sidak test) by leaf type (Figure 1, columns 3 to 5) for experiments where gammarids were fed with different sources of horse chestnut. The food source nD (not decomposed) gave the lowest feeding rate (0.07 ± 0.05 mg (food)/(mg (gammarid) × d)), the feeding rate of organisms fed with source DC (decomposed with Cladosporium for 2 weeks) was intermediate (0.22 ± 0.09 mg (food)/(mg (gammarid) × d))), and food source D (decomposed with Cladosporium for 3 months) resulted in the highest feeding rate (0.36 ± 0.10 mg (food)/(mg (gammarid) × d))). Dangles and Guerold [[27]] found the same relationship for the freshwater amphipod Gammarus fossarum, and Graça et al. [[28]] observed that G. pulex ate twice as much when leaf material was conditioned. Other references show that food preferences for freshwater detritivores are related to the time of inoculation with microorganisms [[29],[30]].
Results demonstrated that standardised food preparation and storage can reduce the variability of the feeding rate of G. pulex in laboratory studies (Figure 1, food quality) which would increase the potential for detecting stressor-related effects ex situ. Within this experiment, a maximal reduction of the variability in feeding rate by 38% was found when data were distinguished by food quality. The higher the food quality and thus the feeding rate, the lower the variability (Figure 1, food quality).
Parasite infection
Organisms infected with acanthocephalan parasites showed a lower feeding rate for both food types tested (p = 0.064 and p = 0.099 for food sources DC and D, respectively; power = 0.999; ANOVA, Holm-Sidak test) (Figure 1). The feeding rate decreased with parasite infection from 0.36 ± 0.10 to 0.28 ± 0.09 mg (food)/(mg (gammarid) × d) when fed with food of source D, and from 0.22 ± 0.09 to 0.11 ± 0.08 (mg (food)/(mg (gammarid) × d)) by feeding with leaves of source DC (Figure 1, column 4 vs. 7). The results also suggest that the intensity of the influence may have been related to the food source. A reduction in feeding rate caused by parasite infection of 22% and 50% was found when fed with the leaf types D and DC, respectively. Excluding infected organisms from laboratory studies reduced the variability of the test results by up to 33% (Figure 1, column 2 vs. 6).
The results of the present study combined with those of Brown and Pascoe [[21]] show that a separation of the organisms according to whether or not they are infected with acanthocephalan parasites will reduce the variability of the test results and thus increase the power in a toxicity study to detect any effects caused by a stressor. Standardisation of either parasite infection or food quality might be suitable to reduce the intra-specific variability in the individual feeding rate for successful toxicity studies at a daily resolution. However, it might be advisable to standardise both because significant differences in feeding rate were observed for the tested food sources and infection status.
Body size
A strong relationship (R2 = 0.79) between feeding rate and body mass (given in dry weight (dw)) was observed for all observation periods. Figure 2 shows the average of the feeding rate as a function of body mass for the whole experimental duration. The feeding rate was consistently higher for smaller organisms.
Figure 3 shows the individual feeding rate over time for three sizes classes of the organism. The feeding rate of each group was compared over time, and it was found that the variability was the greatest and significant for small organisms (<5 mg), whereas for larger organisms, no significant differences in feeding rate over time were observed. Smaller organisms (body mass <5 mg) had two to three times higher feeding rates than organisms with a body mass of >5 mg when calculated over the whole experimental duration (Figure 1). Specification in terms of body mass reduced the variability in feeding rate compared to the mixed groups by 35%, 57% and 49% for the groups <5 mg, 5 to 10 mg and >10 mg, respectively.
When making feeding activity assays with G. pulex, it is advisable to use organisms of a very specific body mass (for example, 2.0 to 2.5 mg) to reduce the variability of the test results and thus increase the possibility of a successful toxicity study. However, restricting the size range of organisms reduces the relevance of test results for the mixed populations found in the environment. A further option would be to use organisms of a higher body mass because of the decreasing strength of the relationship between feeding rate and body mass with increasing body weight. It was observed that organisms >5 mg showed a less distinct and not significant fluctuation in feeding rate over time (p = 0.304 and p = 0.554 for organisms between 5 and 10 mg and organisms >10 mg, respectively; Kruskal-Wallis test) when compared to smaller organisms (Figure 3); however, this was offset by the disadvantage that the feeding rate was lower than that for smaller organisms (Figure 1). It was observed that the feeding rate of organisms <5 mg fluctuated greatly over time, yielding significant differences between different observation periods. Such fluctuations must be excluded for toxicity studies. Thus, the results would suggest the use of organisms >5 mg for toxicity tests. Use of larger organisms reduces uncertainties from weighing, and such organisms can be collected throughout the year [[31]]. We suggest the use of body length as a measure of an organism's dry weight for practicality in future feeding assays. A correlation between body length and dry weight suitable for organisms between 2 and 16 mm is given by Graça et al. [[28]].
Food quality (C-N ratio)
Large variability in the C-N ratio of the food type DC was observed, and this was related to the colour of the leaf discs (Figure 4). The relatively large variability in the C-N ratio for the food type DC correlates with a large variability in feeding rate for gammarids fed with this food type. A further separation of the food source within one preparation procedure by nutrient content (here the C-N ratio) might reduce the variability of the test results even more than the 38.4% observed in this study, because the whole set of type DC leaf discs was used. Furthermore, it was observed that G. pulex had the smallest C-N ratio tested (5.55 ± 0.02) followed by Cladosporium sp. (10.32 ± 0.04). These C-N ratios are clearly smaller than those of all horse chestnut leaf discs tested (Figure 4). Benthic consumers often contain higher amounts of nitrogen and thus have a lower C-N ratio than their food sources [[32]]. There seem to be exceptions for this observation as the C-N ratio for the food source DS was close to that of the gammarids themselves (Figure 4).
Figure 4 also shows a relationship (R2 = 0.99) between the C-N ratio of food eaten (leaf discs excluding the veins) and feeding rate. The feeding rate decreased with increasing C-N ratio. The decline in the C-N ratio with decomposition time of the leaf discs is caused by microbial activity and was also observed for alder and beech leaves in the field [[33]]. This microbial activity, called conditioning, is an important part of leaf litter processing in aquatic ecosystems which increases the palatability of detritus for shredding organisms [[8]]. The literature suggests that aquatic shredders prefer food of lower C-N ratios because the quality is higher resulting in better nutritional status (original reference in [[34]]). However, whether food preference really depends on C-N ratio is unclear, as information in the literature range from no relationship between those two factors at all [[34],[35]] to a strong relationship [[36]]. Nutritional composition is a determining factor of food quality [[37]], and adaptation in feeding activity provides a compensation for sub-optimal composition of available food [[32],[38]].
Under the assumption that G. pulex only eats the amount of food needed to sustain the energy budget and nitrogen is the limiting factor, food consumption would increase as nitrogen content of the food decreased. However, the opposite relationship was observed (Figure 4). An explanation could be that the content of other important nutrients (e.g. phosphorus) might have been decreased during the decomposition by microbial activity and was then limiting. The compensation for this limitation then forced the gammarids to increase their feeding rate. For G. fossarum, another aquatic shredder, it has recently been shown that their growth, which is influenced by feeding, is negatively correlated to the C-P ratio of the food source [[39]]. Our results show that a comparison of feeding activity data generated in experiments using differing leaf species decomposed using different methods might be possible when the C-N ratio of the provided food is measured. However, it is advisable to include phosphorus into the testing of nutritional status of food sources; phosphorus is an essential component of food quality alongside carbon and nitrogen [[37]], and which nutrient is limiting depends on the actual content of the nutrient rather than the ratio.
General discussion
Decreased intra-specific variability in feeding rate resulted when measurement of feeding rate focused on a sub-group of gammarids. Selecting organisms of a sub-group in terms of parasite infection and body mass resulted in a reduction in intra-specific variability of up to 50% and 57%, respectively. Using a food source of particular quality reduced the variability by up to 38%. Acclimation to test conditions only reduced the variability in test results by 1.6%. Certainly, taking into account each of these options to reduce the intra-specific variability for an ex situ feeding assay will maximise the reduction in intra-specific variability, but the result will not be additive.
There are contrasting strategies which can be followed for a feeding assay depending on the objective of the conducted study. The results suggest letting the organisms acclimate to the test conditions for at least 1 day and using organisms that are either all infected or all uninfected. The optimum should be the use of non-infected individuals as the infection tends to reduce feeding rate (present study and [[21]]), thus reducing the chance of measuring negative impacts due to the tested stressor. Furthermore, it is not known whether both parasites of G. pulex influence feeding in the same manner as here these parasites were not distinguished. The proportion of infection and the intensity of the infection with the parasites will depend on geographical location and season, and may in turn cause differences in impacts on feeding. Therefore, the use of uninfected organisms is recommended. Further research could include the investigation of xenobiotic impacts on feeding of infected organisms as those were shown to be more sensitive than uninfected organisms [[21]].
The results show that when conducting feeding assays with gammarids, attention has to be given to the selection of the test organisms in terms of their body size. In order to increase the chance of measuring influences of the treatment, the results suggest conducting experiments with organisms of a specific size class. Which size class to use might depend on the length of the planned study. Short-term experiments may be conducted with juvenile organisms of a very particular size because their feeding rate is in general higher than that for adults. Juveniles have been shown to have a higher sensitivity to toxicants [[40],[41]], making this size class a good candidate for toxicity studies. A further reason to select juvenile individuals may be the increased representativeness of the test for the field situation. The density of organisms within the larger size class (adults) in the field is lower than that of smaller organisms [[42]]. However, as the feeding rate of juveniles fluctuates over time, they may only be suitable for short-term experiments. One may want to increase the number of replicates in such an experiment as the total amount of food consumed by juveniles within a day is rather low which increases the measuring uncertainty. Long-term experiments are particularly important to observe recovery potential following a treatment and for investigation of effects from pulsed exposure. Such experiments should be conducted with adult gammarids to stabilise the control feeding rate over time. A further reason to select adults is their importance for sustainability of the population as these individuals reproduce.
Some attention should be drawn to the food source to be used in a feeding assay. The results suggest using conditioned food prepared in a single batch and the C, N and P content of the food should be measured. Furthermore, more than one leaf disc should be provided per organism in order to reduce the variability of the feeding rate caused by the variability in the food quality. Generally, the longer the leaf material is inoculated with microorganisms, the higher is the feeding rate which, again, increases the chance of measuring negative impacts of the treatment. However, one might want to consider that there is likely a maximum feeding rate which is determined by the food handling time of gammarids. Conducting a feeding assay at such a shredding rate might eliminate the possibility to measure treatment-related increases (i.e. hermetic effects).