Methods:
Experimental Design
To test these ideas, we constructed an experiment where we measured the behavioral responses of crayfish to alterations in total sensory stimuli emanating from predators. We altered these sensory stimuli by changing three different factors that would contribute to the type of chemical cues created in our mesocosms. The three different components were the number of predators present, the summed total length of the predators present, and the average gape ratio of the predators present. The number of predators serves as a proxy manipulation of the overall surface area of predator skin available to release predatory odors (Wood and Moore, 2020). Number of predators ranged from zero (controls) to a maximum of three. In a similar way, changing the summed total length of the predators in a trial also alters the overall surface area for the release of chemical cues. Summed total lengths of predators ranged from 44.9 cm to 80.6 cm. These two variables (number of predators and summed total length of predators) alter skin surface area in different ways. Finally, since these predators are gape limited, we measured the average gape ratio of predators present in a trial. Gape ratios ranged from 0.311 to 0.753 and were calculated by dividing the gape of the bass by the carapace width of the crayfish and taking the average of this measure if multiple bass predators were present (Wood and Moore, 2020). Historically, gape ratios of less than 0.9 for these predators indicated that the bass cannot consume or swallow the crayfish (Hill et al. 2004). The response to these altered threats was based upon changes in crayfish responses within a mesocosm. Behavioral measures included foraging, shelter use, and general movement variables. Within each mesocosm, a single crayfish was used in the behavioral assay regardless of the number of bass present in the predatory section of the mesocosm. Given the small numbers of bass available due to the COVID-19 pandemic effects on fish farms, bass were re-used but crayfish were not. Bass selection for trials was randomized, meaning that groupings of bass varied across trials. This also meant that bass were placed in randomized mesocosms each time. A total of eighty-three trials were run: ten with no bass (controls), thirty trials were run with a single bass, nineteen trials were run with two bass, and twenty-four trials were run with three bass.
Collection and Housing of Animals
Eighty-three form II (non-reproductive) male rusty crayfish (Faxonius rusticus (Girard, 1852)) (carapace width = 1.26 ± 0.026 cm [mean ± SEM]) were captured using minnow traps baited with sardines (Beach Cliff® sardines in soybean oil) from Carp Lake River in Emmet County, Michigan, USA (45.7497°N, 84.8292°W). All crayfish appendages were intact. Crayfish were stored in a flow-through steel cattle tank (200 x 60 x 60 cm: l x w x d) fed with unfiltered water from the East Branch of the Maple River (45.5280°N, -84.7738°W), a watershed containing rusty crayfish populations. Water entered the tank through a PVC delivery pipe (7.6 cm diameter) and exited the tank via a standpipe which kept the water depth at approximately 60 cm. Crayfish fed on naturally occurring detritus in the water column, and shelters made from clay pot halves were available in the storage tank. Crayfish were chosen randomly from the tank for trials and were held in the tank for less than 4 weeks before using. Each crayfish was used only once and then returned to the holding tank where a white triangular mark on their carapace distinguished them from animals that had not been used. Crayfish were not returned to the river upon completion of a trial due to the non-native status of F. rusticus in Michigan and were frozen as outlined in collection permit requirements.
Nineteen adult largemouth bass (total length = 25.6 ± 0.190 cm [mean ± SEM]) with no prior exposure to crayfish served as sources of predatory fish odors. The fish were purchased from Imlay City Fish Farm Inc., (Imlay City Michigan, USA). Fish were fed a diet a commercial fish food (Sportsman’s Choice® Trophy Fish Feed). Upon completion of the trials, bass were euthanized according following approved IACUC protocol. Fish were stored in two separate flow-through cattle troughs, utilizing water from the East Branch of the Maple River where largemouth bass naturally occur. The troughs (200 x 60 x 60 cm: l x w x d) utilized PVC pipes to deliver unfiltered river water to the systems as well as standpipes that kept the water depth at approximately 60 cm. Each trough also contained PVC pipes (7.6 cm diameter) that fish used as refuges. All fish and crayfish were kept outdoors under the natural temperature and daylight/darkness regime. Water temperatures fluctuated from 19-22°C throughout the experimental period. The natural daylight/darkness regime in northern Michigan consisted of 15.5 hours of light and 8.5 hours of darkness. Because treatments were randomized throughout the summer, any discrepancies due to water temperature or light were evenly spread across all treatments.
Ethical Approval
Largemouth Bass were maintained and handled following established animal care and use procedures. The use of vertebrate animals was approved by the Institutional Care and Use Committee at University of Michigan (Protocol: PRO00008892) and by the Institutional Care and Use Committee at Bowling Green State University (Protocol: 1411240-6).
Plant Collection and Storage
Samples of common muskgrass (Chara sp. (Linnaeus, 1753)) were collected from North Fishtail Bay of Douglas Lake, in Cheboygan County, Michigan, USA (45.5618°N, 84.6762°W). A macrophyte sampling rake, made by tying the heads of two rakes together so that the tines point outwards and attaching them to rope, was cast into mats of submerged vegetation to collect the aquatic plants. The collected macrophytes were held in one half head tank (59.1 x 59.1 x 43.5 cm: l x w x h) that acted as a storage stream and was filled with water from the East Branch of the Maple River. The half head tank had an overflow hole drilled into the side that allowed for water to constantly flow through. The macrophyte storage stream was placed in open sunlight to mimic a natural environment. A surplus of macrophyte samples was maintained, and macrophytes were cycled through the storage tank every two weeks from 5 June 2021 until trials ended on 22 July 2021.
Experimental Mesocosms
Cinderblocks were used to frame eight flow-through stream mesocosms (223.52 x 121.92 x 40.64 cm: l x w x d) which were lined with 6-mil polyethylene sheeting. A pair of 208 L plastic drums served as constant head tanks for the eight mesocosms and were filled with unfiltered water from the East Branch of the Maple River. Water entered into the drums via 7.6 cm PVC pipes that utilized nylon stockings to filter out macroinvertebrates. Each plastic drum fed four mesocosms with water using two garden hoses per mesocosm; hoses had diameters of 1.9 cm (flow rate = 0.1 ± 0.05 L/sec [mean ± SEM]). Each mesocosm (Figure 1) was comprised of a predator arena and a prey arena. Predator arenas were covered with egg crating to prevent bass escape and were always upstream of the prey arenas. The predator and prey arenas of each mesocosm were measured at 111.76 x 60.96 x 20.32 cm (l x w x d). The prey sections were lined with sand substrate (depth = 5.1 cm) which accumulated fine detrital material and provided a dark background against which the crayfish were easily observed in video recordings. This same construction technique has been used successfully in previous experiments (Beattie and Moore, 2018; Wood et al. 2018; Wood and Moore, 2019). Water flowed into the upstream predator section of each mesocosm before overflowing through a screened opening (28 x 12 cm opening with 1 x 1 mm screening) in a partial wall into the downstream prey section. The water overflowing through the screened opening did not exceed 5 mm in depth, which is inadequate for crayfish to see into the predator section of the arena (Wood and Moore, 2020). The water exited from the downstream end of the mesocosm through another screened opening. A single PVC half-pipe shelter (10 x 8.5 x 4 cm: l x w x h) with one opening was placed near the down current end of the prey section.
A wooden frame held an infrared DVR camera (Zosi ZR08ZN10) 1.3 m above the water surface of each mesocosm to record the crayfish’s nocturnal behaviors. Cameras have a frame rate of 30 fps which is high enough to capture crayfish movement (Moore et al. 2021). One low intensity red light bulb (Great Value brand: Model A19045 LED Lamp, 9 W, 145 mA, 120 V, 60 Hz, RED) was used to illuminate each mesocosm from above. Crayfish behavior is not impacted by the presence of red light due to crayfish insensitivity to red light wavelengths (Cronin and Goldsmith, 1982; Bruski and Dunham, 1987). An awning made from a black utility tarp (6.1 x 12.2 m) covered all mesocosms to prevent weather and water damage to the electrical equipment. The awning also eliminated glare from moonlight and starlight from the recordings. Sunlight was able to enter the system through 1.5 m openings located on the end of all downstream prey arenas, meaning sunlight exposure and water temperature in mesocosms remained similar throughout the experimental period.
Experimental Protocol
Each trial was run for 23 hours beginning on 7 June 2021 and concluding on 22 July 2021. Trial cycles began at 0830 with the selection and measurement of bass from the flow-through cattle streams. Bass were removed from the holding tanks and measured on a fish board to find total length to the nearest 0.1 millimeter. Bass were then placed into the predator arenas of the mesocosms. Next, a single crayfish was selected for each stream, meaning that all experimental mesocosms had a singular crayfish during the time of the trial. Crayfish carapace width (1.26 ± 0.026 cm [mean ± SEM]) was measured to the nearest 0.5 millimeter using calipers before crayfish were added to the prey arena. Crayfish were marked with a triangular white patch on their carapace before each trial using a non-toxic correction pen (BIC® Wite-Out® 2 in 1 Correction Fluid) to improve visibility for tracking in video recordings. The behavior of crayfish is not altered by the presence of Wite-Out application (Fero and Moore, 2008; Martin and Moore, 2008; Jurcak and Moore, 2018). Last, stems of Chara weighing approximately 5 g (5.00 ± 0.005 g [mean ± SEM]) in total were selected for each trial. Excess surface water was removed from each plant sample by spinning selected macrophytes in a salad spinner (Farberware Basics, Item No. 5158683) for 20 rotations before weighing to the nearest 0.001 g. The macrophyte stems were then attached to glass rods (255 x 6 mm: l x OD) with 26-gauge green painted floral wire. The loaded rods were placed into a hardware cloth bracket (24 x 19 cm: l x w) which held the samples in position during the feeding trial. This technique has been successful in the past (Wood and Moore, 2019; Wood and Moore, 2020).
Beginning at 2300 an automatic light timer activated the red lights illuminating the mesocosms. At 0000 the cameras above each mesocosm began recording the nocturnal behaviors of the crayfish. The cameras shut down at 0340 when behavioral recordings were complete. This time frame has been shown to be sufficient when observing the nocturnal behaviors of crayfish (Wood and Moore, 2019; Wood and Moore, 2020). Water flow through the mesocosms was slow enough that there was no visible surface distortion in the video recordings. All crayfish were removed from the mesocosms first on the following morning, followed by largemouth bass. Last, Chara samples were removed from each mesocosm and surface dried in the salad spinner again before being weighed a second time to find final weight. Upon completion of trials, the screened openings between the predator and crayfish section and the outflow from the crayfish section were brushed to remove any debris that might inhibit water/odor flow. Mesocosms were allowed to flush for at least 24 hours between trials, which is sufficient time for cue dissipation to occur (Wood and Moore, 2020).
Crayfish Behavior Analysis
Each 3 h 40 m video clip per trial was assessed by a viewer who was blind to treatment. The viewer scored the crayfish for its total time spent within the three zones (the foraging zone, sheltering zone, or the clear zone) located in the prey arena. The camera captured images at 30 fps and usually crayfish movement is digitized at 1 fps (Moore et al., 2021). Because crayfish feeding appendages are located on the underside of the animal, it was not possible to see when the crayfish was actually consuming the macrophytes. Still, crayfish were scored as actively foraging when the mark on their carapace was completely inside the foraging zone. Crayfish were also scored as sheltering when the mark was completely within the sheltering zone. Because we were only interested in how crayfish spent time between the zones, we did not observe any anti-predator behaviors in the form of raised chelae.
Foraging effort was calculated by dividing the total time (s) that crayfish spent in the foraging zone by 13,200 s (the total time of the video trial) and multiplying the quotient by 100. This resulting percentage is representative of the total time of the 3h 40m that the crayfish spent foraging. Time spent sheltering and time in the clear zone was also calculated this way. The number of transitions the crayfish exhibited between the three zones was also assessed. This helped to show the overall activity of the crayfish in the trial. Thus, the behavioral analysis produced four dependent variables: Percent of time spent in the clear zone, foraging zone, and shelter zone as well as number of transitions between zones.
Macrophyte Consumption Analysis
Consumption of the macrophyte Chara was assessed using the following formula:
(initial weight-final weight)/(initial weight) × 100
where the initial weight was determined before trials began, and the final weight was taken on the following day when trials ended. Macrophyte consumption was made into a percent to account for any possible minor discrepancies in weights throughout the experiment. This process was utilized across all trials.
Statistical Analysis
Dependent variables consisted of percent consumption of macrophyte, the percentage of time spent in each zone (neutral, foraging, sheltering), and the number of transitions between each zone. Data conditioning and treatment followed the steps outlined in Zuur et al. (2009) for mixed effects models. The first step in this process was creating dotcharts to examine potential outliers within the dataset. None of the 84 trails had outliers, so all of the raw data were included in further analysis. A collinearity analysis was performed between the independent variables of bass abundance, summed total length of predators, average gape ratio, and carapace width of the crayfish. Carapace width of the crayfish was used to calculate gape ratios. The two variables, bass abundance and summed total length of predators, were highly correlated and, because of this, were never run in the same statistical models. Next, histograms, q-q plots, and Shapiro-Wilk tests of normality were used to examine the underlying distribution of response variables. The percentage of time in the neutral zone and the number of transitions between the zones were normally distributed. For those variables that were not normal (percent consumption, time spent in the foraging zone and time spent in the shelter zone), “BestNormalized” was run to determine which data transformation was likely to produce the best normalized data set (CITE ). The consumption of macrophytes was not normally distributed, so a Yeo-Johnson transformation was performed on this data set. The percent of time that crayfish used the foraging zone and the time spent in the shelter zone were not normally distributed, so a square root transformation was applied to this data.
Once data conditioning was finished, dependent variables were analyzed using a linear mixed effects models by running the “lmer” function from the “lmerTest” package in R (Kuznetsova et al. 2017; R Core Team, 2019) for all behavioral responses as well as macrophyte consumption. Following appropriate model selection as outlined in Zuur et al. (2009) and recognizing that the bass abundance variable and summed total length variable were collinear, a top-down model selection process was used. This was a four-step process where 1) the initial two models contained the full interactions of the independent variables as well as the random effects of stream mesocosm (this step involved the construction of two separate models one with bass abundance and one with summed total length given their collinearity), 2) the judgement of the optimal model using the lowest AICs, 3) statistical output of the optimal model, and 4) validation of the optimal model by plotting the residuals against the fitted parameters (Zuur et al., 2009). Following model construction, the outputs were extracted using the “summary” and “anova” function from the “car” package in R (Fox and Weisberg, 2019). If a significant difference in an interaction term was found, a subsequent regression analysis was used to determine a significant linear relationship between the dependent variable and average gape ratio. This analysis was performed in OriginPro (2021b, Origin Lab Corporation).
