An Innovative Setup for High-Throughput Respirometry of Small Aquatic Animals

Metabolic rate is often measured as a phenotype in evolutionary genetics studies because it impacts organismal fitness, is repeatable and heritable, and is responsive to numerous environmental variables. Despite a wide body of literature about metabolic rates, key questions remain unanswered: 1) why do individuals from the same population exhibit up to three fold differences in metabolic rate, 2) how does metabolic rate change during an individual’s lifetime, and 3) what metabolic rate is advantageous in a specific environment? Current low throughput approaches to measure metabolic rate make it difficult to answer these and other relevant ecological and evolutionary questions that require a much larger sample size. Here we describe a scalable high-throughput intermittent flow respirometer (HIFR) design and use it to measure the metabolic rates of 20 aquatic animals simultaneously while reducing equipment costs and time by more than 50%.


INTRODUCTION
22 Metabolic rate is often measured as a phenotype in evolutionary genetics studies because it is 23 known to impact organismal fitness, is repeatable and heritable, and is affected by a variety of 24 environmental variables (1-5). The relationship between metabolic rate and a variable of interest, 25 such as temperature, oxygen availability, or toxicant exposure, has been investigated frequently, 26 which has led to a rich literature on metabolic rates in many species (7-11). Despite this wide 27 body of literature, key questions about metabolic rates remain unanswered including 1) why do 28 individuals from the same population exhibit up to three fold differences in metabolic rate under 29 similar acclimation conditions and activity levels, 2) how does metabolic rate change during an 30 individual's lifetime, and 3) what metabolic rate is advantageous in a specific environment (7)? 31 32 33 Flow through respirometry, intermittent-flow respirometry (IFR), and closed respirometry are 34 techniques used to measure metabolic rates in terrestrial and aquatic organisms. Flow through 35 respirometry is achieved by measuring the amount of oxygen entering and leaving a chamber 36 relative to the flow rate of air or water through the chamber (12). In IFR the respirometer cycles 37 between open and closed periods. During open periods the chamber is flushed to remove waste 38 and oxygen is replenished and during closed periods the animal is using oxygen sealed in the 39 chamber ( Fig. 1) (12, 13). Closed respirometry places an organism in a sealed chamber of known 40 volume and measures oxygen or carbon dioxide partial pressures at multiple time points 41 throughout the trial. The sealed chamber during closed respirometry may result in the 42 accumulation of nitrogenous waste and carbon dioxide, which can increase stress, and may cause 43 loss of equilibrium (LOE) in aquatic organisms (14). Measurement periods (gray shading) occur when the chamber is sealed, and the decrease in oxygen concentration reflects oxygen consumption by the organism. The slopes of the lines (oxygen vs. time) during the measurement periods are used to calculate metabolic rate. Flush periods are when the rapid increase in oxygen occurs as fully oxygenated water is pump into the chamber. Data displayed are from the setup described here.

MATERIALS AND METHODS
84 The custom HIFR system is a large water bath with a PVC rack that holds 20 glass chambers. 85 Each chamber has tubing and pumps to flush the chamber and re-circulate water that passes by 86 an oxygen sensor (Fig. 2).

88
The large water bath (1.2 m long, 1.1 m wide, 0.3 m deep) was constructed out of 0.635 89 cm thick plexiglass and sealed with plastic weld and silicone glue to prevent leaking. A PVC 90 rack with twenty slots separated by small PVC pieces was placed in the water bath, and a 0.300 91 L glass chamber was placed in each slot and sealed with two rubber stoppers. Each rubber 92 stopper had two 0.635 cm stainless steel tubes to attach flexible tubing connected to pumps. 93 Glass chambers were then paired and attached to one peristaltic pump (60 mL/minute) and two 94 separate flush pumps with flow directed to 10 chambers each (300 mL/minute per chamber). A 95 flow-through-cell with an oxygen sensor with a fiber optic cable (FTC) was placed in line with 96 the peristaltic pump. The fiber optic oxygen sensor was attached to a 10-channel oxygen meter 97 (PreSens Precision Sensing, Regensburg Germany). A separate PT-100 temperature probe was 98 placed in the large water bath (PreSens Precision Sensing, Regensburg Germany). PreSens 99 Measurement Studio 2.0 software was used to record oxygen over time as the peristaltic pump 100 recirculated water through the FTC and past the oxygen sensor then back to the chamber. 101 An Arduino Uno with a 5V relay and a set of double pole double throw relays was used to 102 control the direction the peristaltic pump turned and the power to the flush pumps. One-way 103 valves were used to control the flow path to and from the peristaltic pump such that when the Circuit 1 (red), circuit 2 (blue). One-way values (black arrows) control flow direction. By changing the polarity of the peristaltic pump motor, the peristaltic pump direction changes. B) Overall schematic of HIFR. The basic design is a PVC rack that holds and secures glass chambers with their rubber stoppers, which is placed in a large water bath. Each chamber is connected to flush pumps and re-circulating pumps with oxygen sensors. Throughput is limited by the number of channels on the oxygen meter (N) with this design able to measure 2N individuals simultaneously. 146 starting. An R-Markdown script detailing the processing of raw data files is available on github 147 (https://github.com/mxd1288/FunHe_Genomics/blob/master/Raw_Metabolic_Rate_Pipeline.Rm 148 d). 149 The slope of oxygen levels over time was extracted using a linear model for each replicate 150 measurement period, and MO 2 in mol O 2 l -1 was calculated using the equation y=KV then 151 converted to mg O 2 l -1 for comparability, where y=MO 2 (mol O 2 min -1 ), K=slope (mol min -1 ), 152 V= volume of the respirometer (including tubing) minus volume of the organism (liters) (13). 153 Any data collected while the lights were on in the room (before 23:00h or after 06:30h) or a 154 slope with an R 2 value less than 0.9 were excluded from the analysis. Between midnight and 155 06:30h at least 25 measurement periods were completed for each individual, of those at least 20 156 were used for analysis after exclusion based on R 2 value. The lower 10 th percentile values from 157 the cumulative frequency distribution of all replicates from that individual were used to estimate 158 standard metabolic rate (SMR). Using the lower 10 th percentile value from the cumulative 159 frequency distribution did not average the lowest two metabolic rate measures. One value for 160 each individual that lay on the continuous cumulative frequency distribution at the 10 th percentile 161 was selected to represent each individual. This lower 10 th percentile value captures the time 162 period when the fish were most at rest during measurement and excludes the lowest tail of the 163 data distribution, which may be sensitive to outliers (16, 21, 22). (Fig. 3).  165 To compare metabolic rates among individuals that vary in size, metabolic rate must be corrected 166 for body mass. Fish were weighed to the nearest 0.1g the day of metabolic rate measurement. 167 After calculating SMR the residuals of the model metabolic rate (log transformed) vs. body mass 168 (log transformed) were used as the body mass corrected SMR (23).
169 Background respiration: 170 In order to correct for oxygen used by bacteria and other microorganisms in the HIFR, blank 171 runs were completed in between each use of the HIFR and average background respiration 172 subtracted from the MO 2 of each fish (Eq. 2). 173 Eq 2: MO 2 _corrected = MO 2 -background respiration 174 Where MO 2 is the minimum metabolic rate of each fish as previously described and background 175 respiration was a chamber specific value calculated by averaging the oxygen consumption over 176 time in each empty chamber across three replicate blank runs. An empty chamber was 177 additionally run in parallel each night and the background respiration did not change over the 178 course of the night validating the decision to not use a time corrected value of background 179 respiration.

RESULTS AND DISCUSSION
181 System design and testing: 182 Water at 28°C (1°C) and 15ppts was used to fill the custom water bath and recirculated 183 with an aquarium system to maintain temperature and reduce ammonium load. To validate that 184 the flush period was long enough to fully replenish oxygen empty chambers were filled with 185 water at a low oxygen concentration (~60% a.s.), achieved by bubbling in nitrogen, and flushed 186 for over 8 minutes. Between 4 and 5 minutes after turning on the flush pump the oxygen level in 206 207 Repeatability 208 A random set of 19 fish was measured in the HIFR, each in three different chambers over the 209 course of one week (Monday, Wednesday, and Friday night). Log SMR was regressed against 210 log body mass (y=2.66 + 1.08x, R 2 =0.59, N=57, Fig. 4A), and a body mass correction was 211 calculated as described above. The mean coefficient of variation (CV) within an individual was 212 18.03% (Fig. 4B). SMR is repeatable (Fig. 4C), and the variance for each individual for three 213 SMR measured in three different chambers is much smaller than the variance among individuals 214 (ratio of variance in group means/mean of within individual variance = 74.54:1). To measure 215 repeatability (R) directly: where s 2 a equals the difference in the mean sum of 219 measures per individual yields a repeatability of the tenth percentile value of metabolic rate, used 220 here to represent SMR, of 0.96. 221 Metabolic rates measured in HIFR are comparable with values from previously reported 222 metabolic rate values for F. heteroclitus (5%) and other teleost fish (40%) further validating 223 the methods described here (16, 27, 28). To determine this, the metabolic rate reported for F. 224 heteroclitus or other species acclimated to various temperatures was used to interpolate 225 metabolic rate for an 8-gram individual at 28°C. Differences between values reported from our 226 HIFR and other studies using F. heteroclitus may also be due to the type of metabolic rate (i.e. 227 SMR, RMR, MMR) being measured and the environmental parameters (acclimation vs. acute 228 temperature or hypoxia exposure, maximum vs. standard metabolic rate, etc.). The 18.03% CV 229 within individuals for SMR is similar to previous studies that reported 12-14% CV in SMR, 230 MMR, and aerobic scope of brown trout (29). 231 Depending on the oxygen sensing technology and software, a single respirometer (including the 232 cost of oxygen sensing technology and software) may cost between $2,000 and $4,000, a 233 significant investment especially considering that they are designed to measure one individual at 234 a time, which may take hours or days. Some companies additionally offer high-throughput 235 versions (up to 8 chambers) for small animals; however, the cost remains high (>$2000 per 236 chamber). The HIFR presented here was assembled using basic materials and a moderately 237 priced oxygen meter and oxygen sensors. Including the cost of the meter, sensors, and materials, 238 HIFR costs $855.50 per chamber to assemble, a 57% reduction in cost per respirometer 239 compared to purchasing a Loligo high-throughput system. Additionally, the HIFR can 240 simultaneously run up to 20 respirometers at once, greatly reducing the total time needed to 241 achieve a large sample size, which holds value far beyond monetary savings. For example, 242 within a one-week period at least 100 individuals could be run under the same experimental 243 conditions introducing little variation due to time and requiring only 5 nights of respirometry set 244 up with daily background respiration measures. The flexibility of the HIFR offers the additional 245 advantage of allowing organisms of various sizes to be measured. By changing the size of the 246 glass chambers and altering the flow rate of peristaltic pumps and flush pumps by changing the 247 tubing size the system can easily be adapted to fit the desired organism. This further decreases 248 costs for groups who may wish to measure a single species at various ages and stages of life or 249 different species that may vastly differ in size (30). 250 Costs could be cut further by using less expensive peristaltic pumps or a different water bath than 251 described here. However, the lifespan of a given pump varies greatly depending on the quality of 252 the motor and the tubing. Several peristaltic pumps ranging from $3 to $50 were tested to 253 determine the appropriate tubing material and motor design that could withstand frequent long-254 term use and alternation of motor polarity without rapidly burning out. Generally, it is 255 recommended to use a peristaltic pump that has a brushed motor and tygon tubing and to 256 determine the tubing size based on the desired flow rate. While there are large peristaltic pumps 257 available it should be noted that depending on chamber size this may not provide enough mixing 258 to prevent the stratification of water in the chamber (31). The addition of a closed loop mixing 259 pump could mitigate this problem and provide adequate mixing, although this has not been tested 260 here. Including the mixing pump would increase the total setup cost and without it the size of the 261 chambers (and organisms) that can be measured with this system will be limited to those that can 262 be adequately mixed with only a peristaltic pump. 263 The plexiglass tank served as a water bath for the chambers and could be replaced with a cheaper 264 alternative as long as it could hold the appropriate volume of water needed to maintain a stable 265 temperature and prevent the buildup of nitrogenous waste over the course of the run. It would 266 also need to be large enough to hold the desired number of chambers of a specific size. In 267 general, the respirometer volume should be 20 to 50 times larger than the organism to achieve a 268 measurable decrease in oxygen over a reasonable period of time (several minutes) (6, 12). If the 269 respirometer volume does not fit within this ratio for a given organism the measurement period 270 length can be adjusted to allow for the appropriate drop in oxygen (above 80% O 2 saturation) as 271 long as routine movements are not inhibited by chamber size (6, 12). If variation in body mass of 272 individuals is large, adjusting the measurement period to prevent low O 2 levels for larger 273 individuals may mean that smaller individuals do not have a large enough O 2 decrease to get a 274 reliable slope measurement. Using the HIFR design it is possible to use chambers of various 275 sizes within a single run as long as pump flow rates were adjusted (tubing sizes) to accommodate 276 this change. This allows for added flexibility of running different sized respirometry chambers 277 simultaneously. 278 The throughput of this design is limited by the number of channels available on the oxygen 279 meter. Any flow through oxygen sensing cells can be interchanged for the ones used here; 280 however, the oxygen meter needs N channels (one per sensor) to allow measurement of 2N 281 individuals at once. If an oxygen meter were available with 20 channels, for example, it is 282 feasible that this design could be scaled to measure 40 individuals over the course of one night. 283 An oxygen meter with fewer channels could be used to design a similar HIFR with fewer 284 individual respirometers. Additionally, a HIFR could be built to measure ten individuals with ten 285 channels by eliminating the double pole double throw relays and using an Arduino to turn the 286 flush pump on and off. 287 Due to the size of the water bath and the available equipment, the most practical solution to 288 maintaining a constant temperature in the water bath was to recirculate the water through a 289 temperature-controlled aquaria system. This made it possible to pump the HIFR water bath into 290 the same system the fish had been housed in prior to measuring metabolic rate so the temperature 291 along with pH and salinity were not variable between the HIFR and the acclimation conditions 292 (32). 293 It should be noted that the HIFR was built by an early career biology graduate student with little 294 prior knowledge of electrical engineering or plumbing. The easy to learn techniques used make 295 this methodology highly accessible. 296 The ability to precisely measure metabolic rates in a high-throughput manner without 297 significantly increasing the necessary effort has application for physiologists, ecologist, 298 geneticists, and comparative biologists alike. This method reduced the total system cost from 299~$2,000 per respirometer to ~$900 per respirometer including the cost of the FTC, oxygen 300 sensor, and oxygen meter. The HIFR also greatly reduced the effort needed to measure metabolic 301 rate in a large sample size making it possible to answer questions relevant to ecological and 302 evolutionary biology.