A simple and affordable calorespirometer for assessing the metabolic rates of fishes

The accurate measurement of metabolic rate has tremendous value across many disciplines in the life sciences. The rate at which an organism consumes and utilizes energy provides insight into its biology from the level of its cells to its ecology (Hochachka and Somero, 2002; Brown et al., 2004). The most widely used method for assessing metabolic rate is through the measurement of oxygen consumption rate (Ṁ_O2), which, under aerobic conditions, provides a reasonably good estimate of metabolic rate. However, under circumstances like hypoxia, where the metabolic rate of an organism cannot be solely supported by aerobic metabolism, and anaerobic processes are utilized to buffer ATP turnover, measurement of Ṁ_O2 could drastically underestimate metabolic rate. This is most evident in cases of anoxia-tolerant organisms like the painted turtle (Chrysemya picta), crucian carp (Carassius carassius) and goldfish (Carassius auratus auratus), where attempts to quantify metabolic rate via respirometry in anoxia are futile because of the organism's complete reliance on anaerobic processes to support energy turnover. Like aerobic pathways, however, these pathways yield heat as a by-product, and the total amount of heat lost by an animal to its environment is proportional to its total energy turnover (minus that conserved in carbon bonds) (Mendelsohn, 1964; Mclean and Tobin, 1987; Kaiyala and Ramsay, 2011). Measuring this heat via calorimetry is therefore an effective way of estimating an animal's metabolic rate in situations where aerobic metabolism may be compromised.

Despite direct animal calorimetry being the ‘gold standard for quantifying the fire of life’ (Kaiyala and Ramsay, 2011), it is a seldom-used technique owing to its reputed difficulty and expense when compared with respirometry. These difficulties are especially true when working with ectothermic animals like fish, whose lower metabolic rates produce less heat compared with similarly sized endotherms. Measuring these low levels of heat requires an especially sensitive calorimeter and, to date, these have been extremely expensive to purchase. Efforts have been made over the years to produce inexpensive systems to measure heat in fish (Davies, 1966; Stevens and Fry, 1970), but they have not been described in sufficient detail to facilitate their reconstruction. With high-density thermocouple Peltier units being widely available, it should be possible to construct a relatively simple and inexpensive calorimeter of high sensitivity. Here, we describe the construction and testing of such an apparatus, capable of converting a fish's metabolic heat to a voltage through the use of Peltier units and the thermoelectric effect (more specifically, the Seebeck effect). Furthermore, the fish chamber is designed to operate under flow-through conditions to enable environmental manipulations and the simultaneous measurement of inflowing and outflowing partial pressure of O₂ (P_O2), which can be used to calculate Ṁ_O2 (hence, calorespirometer). We tested the calorespirometer's function using goldfish, a species that is well known to undergo metabolic rate suppression in response to hypoxia/anoxia exposure (van Waversveld et al., 1988; Addink et al., 1991; Stangl and Wegener, 1996; Richards, 2009).

The Seebeck effect allows a heat flux to be converted to a voltage as it passes through a thermally conductive element such as a thermocouple. Peltier units are composed of a number of antimony telluride thermocouples connected in series that, with a Seebeck coefficient of 213 μV K⁻¹, are highly sensitive to temperature change. In building our calorespirometer, we placed a collection of Peltier units between a small fish/reference chamber and a large mass of aluminium so that the metabolic heat produced by the fish would flow through the Peltier units and into the mass of aluminium. The calorespirometer was assembled in a differential fashion with two identical fish/reference chambers attached on either side of the aluminium mass. To measure metabolic heat loss from the fish, the two chambers were treated identically apart from the presence of a fish (or resistor; see ‘Heat calibration and measurements’ section below) in one side, and we monitored the net voltage between the two chambers. This differential configuration accounted for any fluctuation in ambient temperature. Below, we detail the construction of the apparatus and its major components, and explain how it was assembled to optimize performance. A complete list of its essential and accessorizing components and their costs is shown in Table 1.

For this section, ‘calorimeter’ will refer exclusively to the component of the calorespirometer responsible for the detection of heat and its conversion to a voltage. This component, shown in Fig. 1A,B, was assembled symmetrically with two identical sides centred on a block of aluminium (98×48×48 mm). Two Peltier units (~40×40×4.7 mm, 127 couples; Custom Thermoelectric Peltier module 12711-5L31-03CQ, Bishopville, MD, USA) were affixed to each side of this block using an ultra-thin layer of silver conductive epoxy (MG Chemical no. 8331, Surrey, BC, Canada) and connected in series so as to maximize the voltage reading (Fig. 1A). A brass block (~78×26×26 mm) was affixed to the opposite side of each group of Peltier units using silver conductive epoxy. Brass was ideal for this component as its hardness and machinability allowed for especially thin walls and its high thermal conductivity optimized heat flow. The brass blocks had a 25 mm diameter bore into which a fish or reference chamber could be inserted. Together with the Peltier units and the brass blocks, the central block of aluminium was bolted to another aluminium block (~98×152×48 mm) into which two cylinders were bored (~25 mm diameter) and through which the fish chamber and reference chamber could be inserted into and removed from the calorimeter's brass blocks (Fig. 1B).

Identical 32 ml flow-through chambers were constructed to serve as the fish chamber and the reference chamber (Fig. 1B). These chambers were made of stainless steel tubing (~77 mm length, 25 mm o.d., 24 mm i.d., 0.5 mm wall thickness; McMaster-Carr no. 6622K152, Aurora, OH, USA), with a stainless steel cap of 0.5 mm thickness permanently welded to the upstream end of the chamber. Inserted through this cap were stainless steel inflow and outflow water lines (1 mm o.d., 0.8 mm i.d., 0.1 mm wall thickness), the inflow water line running along the chamber's bottom all the way to the downstream end, and the outflow water line situated at the chamber's top and mounted flush with the stainless steel cap at the upstream end of the chamber. The water lines were oriented this way to optimize mixing within the chamber and to allow an easy path of exit for any gas bubbles that may enter the chamber. Finally, a removable Plexiglas cap equipped with a rubber gasket was placed at the downstream end of the chamber through which the fish could be inserted and removed. This cap could also accommodate a P_O2 optode (Ocean Optics OR125, Dunedin, FL, USA) that was used to measure the P_O2 within the chamber. Apart from this optode, the fish chamber and reference chamber were identical.

For this section, ‘respirometer’ will refer exclusively to the component of the calorespirometer responsible for the measurement of P_O2 and determination of Ṁ_O2. This component was built in a flow-through fashion and incorporated exclusively on the fish chamber side of the calorimeter. Small stainless steel chambers of 1 ml (Fig. 1C) were built to accommodate P_O2 optodes (Ocean Optics OR125) on both the inflow and outflow water lines immediately adjacent to the bored-out aluminium block (i.e. as close to the fish chamber as possible), and the difference between the P_O2 values measured by these optodes enabled calculation of the fish's Ṁ_O2.

No heat or electrical signals from the activated P_O2 optodes could be detected by the calorimeter, thus their use did not affect measurements of metabolic heat loss.

To provide a thermally stable environment for the calorespirometer, it was placed within an insulated ice chest (Coleman 6-Day Xtreme, Golden, CO, USA) inside an additional enclosure (foam insulation, 5.08 cm thickness), located within a temperature-controlled (20±0.1°C) environmental chamber measuring 3×3×2.5 m. The insulated ice chest was lined with aluminium blocks totalling ~40 kg, and the calorespirometer was placed in the centre of the chest. The aluminium was used as a heat sink, drawing heat from the fish and reference chambers through the Peltier units. The aluminium's high thermal inertia, a function of its mass, thermal conductivity (237 W m⁻¹ K⁻¹) and (molar) heat capacity (24.2 J mol⁻¹ K⁻¹), made it an especially effective heat sink and ensured the Peltier units accounted for as much of the fish's metabolic heat as possible.

As heat from the chambers flowed through the Peltier units, the net voltage was measured using a Keithley Model 147 nanovoltmeter (Cleveland, OH, USA). The leads from the Peltier units were soldered to the pure copper lead from the voltmeter and this junction was affixed to the aluminium mass using electrical tape to minimize its possible (albeit small) effect on the measured voltage. The amplified signal was then digitally converted using a DATAQ DI-148 data acquisition system (Akron, OH, USA) and recorded on a Dell Precision M4300 laptop computer using DATAQ WinDaq software.

The water supplying the fish and reference chambers was sourced from a common 2 l recirculating volume. This volume was held in an insulated ice chest identical to the one housing the calorespirometer (minus the aluminium) and placed adjacently. Water was drawn out of the beaker by a peristaltic pump (Gilson Minipuls 3, Middleton, WI, USA), pushed into a gas equilibration chamber (see below), and then into the stainless steel tubing supplying the fish and reference chambers. Water flowed into and out of the chambers as described previously, and was emptied into the original 2 l beaker for recirculation.

To manipulate the gas tension in the fish and reference chambers, gas mixing was done using a precision gas mixer (Corning 192, Medfield, MA, USA) and the mixed gas was equilibrated with the water supply in two ways. First, the mixed gas was bubbled into the 2 l recirculating supply volume, and second, the mixed gas flowed into a 1.5 l glass gas equilibration chamber within which the supply water flowed through Silastic tubing before flowing into the stainless steel tubing supplying the fish and reference chambers (Fig. 2).

Goldfish, Carassius auratus auratus (Linnaeus 1758), of 0.756±0.087 g (mean ± s.e.m., N=5) wet body mass were acquired from a commercial fish dealer and held at the Department of Zoology's Aquatic Facility at The University of British Columbia (Vancouver, BC, Canada). Fish were held at a stocking density of <0.4 g l⁻¹ in a 76 l recirculating system and maintained in well-aerated, dechlorinated City of Vancouver tap water at 20°C under a 12 h:12 h light:dark cycle. Water in the recirculating system was replaced every 7–10 days. Fish were fed to satiation daily (Nutrafin Max Goldfish Flakes) except 24 h before being transferred to the calorespirometer, during which period feeding was suspended. The University of British Columbia Animal Care Committee approved all procedures involving fish.

Before each experiment, the P_O2 optodes were calibrated in air and 100% nitrogen. A single goldfish was then inserted into the fish chamber via the removable cap, and the chamber was sealed and slid into place within the calorespirometer's brass block. The peristaltic pump was turned on, supplying both fish and reference chambers with oxygenated water at a rate of 22 ml h⁻¹. Water temperature was maintained at 20°C throughout all experimental trials. The fish was allowed to habituate to the chamber for 16–18 h, which allowed sufficient time for both the thermal equilibration of the calorespirometer and the recovery of the fish from handling stress. After the habituation period, we conducted several experiments designed to test the calorimeter's function. To ensure it was capable of detecting variation in metabolic heat produced by the fish, we used carbonyl cyanide 4-(trifluoromethoxy) phenylhydrazone (FCCP; Sigma-Aldrich C2920, St Louis, MO, USA) in an attempt to increase metabolic rate via mitochondrial uncoupling, and benzocaine (Sigma-Aldrich E1501), an anaesthetic, to decrease metabolic rate. These experiments were repeated three times. The next set of experiments was carried out to ensure the apparatus was capable of detecting the previously observed O₂-dependent changes in metabolic heat produced by goldfish (van Waversveld et al., 1988; van Waversveld et al., 1989; Addink et al., 1991). Water P_O2 was decreased from ~40 kPa to 0–0.25 kPa, where it was held for 1.5 h before being returned to normoxia. This experiment was repeated five times.

Although the baseline heat signal remained stable over the duration of each run, it fluctuated between runs by ±0.03 mV. In order to accurately determine the baseline heat signal for each experiment, we introduced an overdose of anaesthetic in the fish chamber (final concentration ~300 μmol l⁻¹ benzocaine) to kill the fish in the chamber at the end of the experiment. Heat loss from the fish quickly subsided after the addition of the anaesthetic, stabilizing at a baseline value within ~25 min (preliminary experiments showed no further decrease in heat loss rate over 3 h). After ~1 h of stable baseline reading, the fish was removed from the calorespirometer and the experiment concluded.

Statistical analyses consisted of one-way analysis of variance, performed using SigmaStat version 4.0.

The calorespirometer was both stable and sensitive. Under flow-through conditions of 22 ml h⁻¹ and 20°C but without a fish present, heat flow measurements showed very small oscillations (±0.35 mW) around the baseline and there was no net drift in baseline heat detected over 72 h (data not shown). Changes in P_O2 of inflowing water (between 0 and 40 kPa) and turning the P_O2 probes on and off had no effect on heat flow (data not shown).

The heat calibration generated a linear relationship between applied wattage and measured voltage (Fig. 3; equation of the line mV=0.1371 mW) that could be used to convert the metabolic heat of a fish, measured in millivolts, to milliwatts. The heat pulses also revealed the calorespirometer's sensitivity to be 141.15 μV mW⁻¹ at a water flow rate of 22 ml h⁻¹, a sensitivity in close agreement with that of the only known commercially available calorespirometer that can accommodate a fish (see Addink et al., 1991).

The next step involved inserting a fish into the fish chamber to determine whether the apparatus was capable of measuring its metabolic heat under fully oxygenated conditions. The representative trace in Fig. 4 shows that ~15 h were required for the fish to habituate to its new surroundings and for the calorespirometer to thermally equilibrate after insertion of the fish (at time zero in Fig. 4). During these preliminary trials, water P_O2 was maintained at ~40 kPa to ensure adequate oxygen delivery and compensate for the low water flow rate (22 ml h⁻¹). The fish's rate of metabolic heat loss stabilized by 15 h and remained relatively constant at ~1.5 mW g⁻¹ (Fig. 4), with sporadic increases in heat probably due to episodes of activity.

To ensure we could detect variation in metabolic heat loss, pharmacological agents with known effects on metabolism were introduced into the fish and reference chambers by briefly transferring the inflow lines from the 2 l water supply beaker to a vessel holding the pharmacological agent. FCCP is an uncoupling agent that increases the permeability of the mitochondrial inner membrane, dissipating the proton gradient used to drive ATP production via oxidative phosphorylation. We predicted this would yield an increase in metabolic heat loss, and, in fact, sequential additions of 4 μmol l⁻¹ FCCP (up to 12 μmol l⁻¹ FCCP) resulted in incremental increases in heat loss, up to a 60.5±10.3% increase compared with controls (P<0.001; Table 2). Similarly, benzocaine, a widely used anaesthetic, was administered in the same way with the prediction that it would decrease metabolic heat. A single dose of ~100 μmol l⁻¹ benzocaine resulted in a 68.6±5.1% decrease in heat loss compared with controls (P<0.001).

Goldfish have been shown to reversibly suppress their metabolic rate by 60–70% when exposed to anoxia (van Waversveld et al., 1988; van Waversveld et al., 1989; Addink et al., 1991; Stangl and Wegener, 1996), and our results are consistent with these previous findings (Fig. 4, Fig. 5A). At the end of the 15 h habituation period, water P_O2 was decreased over 90 min to between 0 and 0.25 kPa. The fish was then held at this P_O2 for 1.5 h during which metabolic heat loss decreased and stabilized at an average value that was ~30% that of the average resting level (2.8 J g⁻¹ h⁻¹ at 0 kPa versus 9.3 J g⁻¹ h⁻¹ at 40 kPa; Fig. 5A). When P_O2 was returned to ~40 kPa, metabolic heat loss returned to levels that were equal to pre-hypoxia levels. Following a 2 h recovery period, the fish was killed with an overdose of anaesthetic to determine the baseline heat signal as described previously.

Measured Ṁ_O2 values from goldfish held in the calorespirometer showed similar responses to the measured heat fluxes discussed above. Specifically, high Ṁ_O2 values were measured over the initial 5 h after the fish was introduced to the calorespiromenter, gradually decreasing to stable levels after 12–15 h in the calorespirometer (Fig. 4). Our mass-specific routine Ṁ_O2 values are higher than those reported elsewhere for goldfish (van Waversveld et al., 1988), but this variance is probably accounted for by differences in size (our fish are 12 times smaller than those used by van Waversveld et al.), fasting regime, and habituation time and conditions between the studies. Upon exposure to anoxia/hypoxia, Ṁ_O2 fell to near-zero levels, returning to routine levels upon the reintroduction of O₂ (Fig. 5B).

In order to maximize the sensitivity of our calorespirometer for heat detection, we used a low rate of water flow through the fish and reference chambers, which affected the time domain over which Ṁ_O2 could be measured. After a change in inflowing P_O2, about 60 min were required for the P_O2 in the outflowing water to stabilize and, thus, during this equilibration period, calculations of Ṁ_O2 were inaccurate. Apart from this period, the fish's O₂ consumption could be accurately and constantly measured in real time in parallel with its rate of metabolic heat loss. It is important to note that this ~60 min equilibration period was not needed for the measurement of metabolic heat; the calorimeter responded instantly to changes in heat and stabilized within ~25 min (Fig. 3, inset).

Despite the calorespirometer's straightforward design and assembly, much attention was needed when preparing the apparatus for experimental use. Central to most of this was the extreme thermal sensitivity of the Peltier units. The differential design of our calorespirometer should theoretically account for fluctuations in ambient temperature, but effort was still required to ensure all heat produced by electrical equipment in the environmental chamber (e.g. computer, voltmeter, peristaltic pump, etc.) was evenly distributed across the enclosed, insulated ice chest. Fans and heat funnels were used for this, and any vulnerable parts on the insulated ice chest (especially drilled holes for the passage of water lines and electrical cables) were patched with form-fitting foam insulation. This was particularly important for holes in close proximity to the Peltier units. Although it was not a problem with our setup, care should also be taken to ensure the voltage reading is not being affected by electrical activity on the circuit into which the voltmeter is plugged.

The accurate and precise determination of a fish's metabolic rate demands a baseline heat signal that is known and stable. It is possible that with a highly controlled environment and a faithful duplication of experimental setup procedures and the orientation of all components, an identical inter-experiment baseline heat signal can be generated. However, despite our efforts, we noticed an inter-experiment fluctuation in baseline heat signal by ±0.03 mV (although mean intra-experimental baseline drift was negligible). This required the fish to be killed via an overdose of anaesthetic at the end of each experiment as described previously. Although this is not ideal, the accurate and precise determination of the fish's metabolic rate required it. It is possible that this approach may be needed in other calorespirometers built from this design.

Finally, a calorespirometer like the one described here will inevitably come with a few limitations that need addressing. Firstly, the flow-through design that allows for environmental manipulation and long experimental durations means some of the metabolic heat produced by the fish will be washed downstream, resulting in a possible underestimation of its metabolic rate. This effect will be minimized through the use of a relatively low flow rate, and all but eliminated by performing the heat calibration process at the experimental flow rate (see Materials and methods). Secondly, the use of a low water flow rate could result in the accumulation of metabolic end products (e.g. CO₂) in the fish chamber that could have their own effects on the organism. In our hands, measured P_CO2 values never exceeded 1.1 kPa in a typical 24 h experiment and thus were not likely to have a negative effect on the fish's metabolic rate (Fry et al., 1947). Should the outflowing water contain high P_CO2 or metabolic waste, a higher flow rate is recommended, though this will decrease the calorimeter's sensitivity. Thirdly, as discussed above, there are different time delays for the measurement of Ṁ_O2 and heat loss that must be taken into account when assessing metabolic rate. In general, if both measurements are required, the time resolution for measurements will be of the order of 1–2 h. And finally, because of the long habituation time required for accurate measurement of Ṁ_O2 and metabolic heat, the animals are in a fasted state. The duration of fasting has been shown to influence metabolic rate (Davies, 1966), so care must be taken to ensure that all animals are treated similarly.

We have constructed an inexpensive and sensitive calorespirometer for the simultaneous measurement of metabolic heat and Ṁ_O2 in fasted fish with a time resolution of 1–2 h, making it possible to measure metabolic rate in environments that compromise aerobic energy production. Combined with its low cost of construction and simple, modifiable design, this apparatus is obtainable to most researchers and has the potential to shed light on the metabolic responses of a broad range of species in any number of environments.

Special thanks to Bruce Gillespie in the University of British Columbia's Department of Zoology Machine Shop for his help with construction of the calorespirometer, and to the anonymous reviewers for their insightful comments.