Assessing metabolic rates in zebrafish using a 3D-printed intermittent-flow respirometer and swim tunnel system

The metabolic rate of any animal reflects its cost of living. For endotherms the minimal metabolic rate (basal metabolic rate, BMR) is the resting metabolic rate of the fasting animal in its thermoneutral zone. For ectotherms, a resting and fasting animal is at its standard metabolic rate (SMR), which is temperature dependent. In either case, the BMR or SMR set the minimum food and energy requirement for survival and are useful physiological parameters for comparison of metabolic rates across species under identical conditions.

The maximal aerobic metabolic rate (MMR), also called aerobic capacity, is an important performance parameter used in the study of aerobic activity and reflects the maximal rate of ATP production by aerobic metabolism. A high MMR allows for a high sustained speed during persistent physical performance, which may allow to catch prey and/or avoid predators. In humans, MMR also serves as an important indicator of physical health. Because of the increased maintenance cost of a large metabolic engine, the evolution of a high MMR may have been the key driver for the evolution of a high SMR and BMR and ultimately for endothermy (Bennett and Ruben, 1979).

The difference between MMR and SMR is the absolute aerobic scope (AAS), which is the aerobic metabolic rate an ectotherm can spend on life processes like reproduction, growth and movement. It is widely considered to be the ultimate measure of aerobic performance (Fry and Hart, 1948) and shows large variation between individuals (Norin and Malte, 2011, 2012; Boldsen et al., 2013). While the difference between MMR and SMR describes the aerobic scope of the individual, the ratio between MMR and SMR, the so-called factorial aerobic scope, is often relevant in comparative studies and tells how many fold an organism can increase its aerobic metabolic rate, irrespective of the SMR.

The MMR is normally measured in fishes in one of two ways. One approach is measuring the O₂ consumption rate immediately after exhaustion of the fish in a post-exhaustion experiment. This is done by chasing or by swimming in a flume (Norin and Clark, 2016), followed by rapid transfer of the fish to a respirometer and immediate start of O₂ consumption measurements.

Another approach involves continuously monitoring O₂ consumption in a respirometer while progressively increasing swimming speed until exhaustion. This method traditionally used a step-by-step protocol and its goal was to determine the maximal sustainable swimming speed that can be kept for a prolonged (typically 30 min) period of time (Brett, 1964; 1967; Beamish, 1978). This speed is the so-called critical swimming speed and the O₂ consumption at this speed is thus the maximal sustainable aerobic metabolic rate. However, while this parameter provides a useful measure of an animal's performance, it does not qualify for a measure of MMR, which represents the maximal capacity of the O₂ transport and utilization system and is associated with a state that is clearly not sustainable.

Zebrafish (Danio rerio) is a powerful model organism for studying many aspects of physiology. Their small size, high fecundity, and genetic similarity to humans make them ideal for a wide range of experimental investigations (Seth et al., 2013). Obtaining reliable values of SMR as well as MMR for zebrafish is therefore key to characterize their metabolic performance and has become increasingly relevant in physiological and biomedical studies. However, divergent values of adult zebrafish metabolic rates are found in the literature. Studies have reported highly different SMR values (Lucas et al., 2014; Huang et al., 2020) and generally low factorial aerobic scopes (Lucas et al., 2014; Huang et al., 2020; Ripley et al., 2022; Morgan et al., 2022), suggesting that MMR might have been underestimated. Here, we describe how to accurately determine SMR and MMR of zebrafish under ramping exercise.

Currently there are no inexpensive, readily available solutions to measure MMR and SMR in zebrafish. In this study, we have developed a 3D-printed intermittent-flow respirometer and swim tunnel system specifically designed for accurate measure of SMR and MMR in adult zebrafish (Fig. 1). The internal volume of the swim tunnel is small enough to provide consistently accurate measurements of SMR at rest during low water flow rates, while capable of high flow rates that readily drive the fish to exhaustion, thus yielding reliable MMR measurements.

We here provide experimental protocols for measuring SMR and MMR in zebrafish in a swim tunnel and respirometer system. The 3D design is provided and can be printed and used by researchers to measure metabolic rates in zebrafish. The system can be customized further to accommodate other small fish species of various ages and sizes.

When measuring O₂ consumption overnight, it took between 3 and 8 h for the metabolic rate (MR) to stabilize (Fig. 2A). Furthermore, we observed occasional spikes of high O₂ consumption during the protocol, likely due to spontaneous activity (Fig. 2A-D). We measured a mean SMR of 119±15.9 µmol O₂·min⁻¹·kg⁻¹ using intermittent-flow respirometry overnight (Fig. 2E). This value is in good agreement with the value of ∼149 µmol O₂·min⁻¹·kg⁻¹ measured by Lucas et al. (2014) on similar-sized zebrafish, using a protocol with pre- and post-blank measurements, but at a slightly higher temperature of 28°C. The study by Huang et al. (2020), using a system similar to ours, reported a much higher SMR value of ∼450 µmol O₂·min⁻¹·kg⁻¹ in young adult zebrafish (∼130 mg). However, they did not correct for background respiration, which may have contributed to overestimation of the SMR.

The slow protocol for measuring MMR employs a 3 min flushing period followed by a 3 min closed period for each flowrate increment (Fig. 3A,C), and this is about as fast as intermittent-flow respirometry can measure O₂ consumption in this system. The advantage of this protocol is that pO₂ is held as high as possible to sustain aerobic metabolism and the measurement period is long enough to estimate O₂ consumption at each swimming speed. A potential drawback of this method is that the highest O₂ consumption rate could be missed during the extended flush period.

The fast protocol uses a single, extended close period starting before the expected yield of the fish (Fig. 3B). By eliminating the flush period in the fast protocol, it becomes possible to increase the frequency of flow rate increases to every 1 min instead of every 6 min. This makes the protocol faster and more akin to the continuous protocol used on humans to measure MMR, where workload is rapidly ramped until reaching a plateau in O₂ consumption (Duncan et al., 1997).

Data from the slow protocol was analyzed using a 90 s fixed window (Fig. 3A,C) or a 60 s sliding window yielding MMR of 767±411 and 877±444 µmol O₂·min⁻¹·kg⁻¹, respectively (Fig. 3D). Data from the fast protocol was measured with a 60 s sliding window (Fig. 3B), yielding a MMR of 982±330 µmol O₂·min⁻¹·kg⁻¹ (Fig. 3D). There was a tendency for the fast protocol to result in higher MMR values (Fig. 3D), albeit this was not significant (Kruskal–Wallis test, P=0.141). This is similar to the findings of Duncan et al. (1997) who compared the MMR obtained using stepwise versus continuous protocols in humans.

Our results indicate that the post-exhaustion method is not suitable for measuring MMR in zebrafish, where O₂ consumption drops immediately after exhaustion, as shown in the fast protocol experiments (Fig. 3B). In a typical post-exhaustion experiment, by the time the fish is transferred to the respirometer, it will already have recovered considerably from exhaustion. Additionally, our data shows that low pO₂ in the chamber did not limit MMR. We found a negative correlation between pO₂ during the measurement period and the calculated MMR in the fast protocol (Fig. S1), indicating that the decrease in pO₂ was a result of the high O₂ consumption of the fish and not a limiting factor.

Using the Shapiro–Wilk normality test, MMR values were normally distributed for the fast protocol (P=0.936), but not the slow protocol (P=0.0138 and P=0.0207 for the fixed and sliding windows, respectively), suggesting that more consistent data are obtained by using the fast protocol. Lastly, the speed of the fast protocol provides an obvious advantage in reducing the duration of the experiment.

The factorial aerobic scope (MMR·SMR⁻¹) was calculated using the three MMR methods tested here: the slow method with a fixed window (6.99±3.50), the slow method with a sliding window (8.05± 3.83), and the fast protocol using a sliding window (9.30±3.25) (Fig. 3E). Again, although the factorial aerobic scope tended to be higher when using the fast protocol, values were not significantly different (Kruskal–Wallis test, P=0.169). The factorial scopes in our study are higher than previously reported values in zebrafish of 2.5–4 (Lucas et al., 2014; Huang et al., 2020; Ripley et al., 2022; Morgan et al., 2022), where SMR and MMR might have been overestimated or underestimated, respectively. Conversely, our factorial scope values are similar to those of ∼7.5 found on ∼50 g brown trout (Norin and Malte, 2011, 2012) and also close to the range of 10–15 reported for mammals (Hinds et al., 1993). The advantage of our method is that it allows the researcher to completely exhaust the fish within the respirometer while simultaneously measuring O₂ consumption.

To conclude, we have developed a 3D-printed respirometer and swim tunnel system, designed for zebrafish but likely usable for other small fishes. We show that the respirometer can measure SMR and MMR reliably on the same fish in one continuous protocol with no interruption (Fig. S2). We provide the drawings for the system (Supplementary data file 1) and this will give researchers with access to a 3D-printer an inexpensive respirometry setup for accurately measuring SMR and MMR in zebrafish. Furthermore, we demonstrate that the fast protocol of measuring MMR in zebrafish is as accurate as the slower intermittent-flow respirometry alternative. Finally, our results show that zebrafish exhibit remarkably rapid post-exhaustion recovery, which may lead to incorrect estimates of their actual MMR when using post-exhaustion methods.

Zebrafish were bred in-house from AB founders from the European Zebrafish Resource Center of the Karlsruhe Institute of Technology (KIT). Fish were fed three times daily and maintained on a 14-h-light:10-h-dark cycle on recirculating housing systems at 28°C in reverse osmosis water conditioned to pH 7.2 and conductivity 700 µS using sodium bicarbonate and Instant Ocean Sea Salt. Embryos were obtained by natural crosses, reared in E3 buffer [5 mM NaCl, 0.17 mM KCl, 0.33 mM CaCl2, 0.33 mM MgSO4, 10⁻⁵% (w/w) methylene blue, 2 mM HEPES pH 7.2] at 28°C. All experiments were carried out according to Danish legislation and approved by the Danish Animal Experiments Inspectorate (Dyreforsøgstilsynet, permit number 2023-15-0201-01510). We used a total of 15 male and three female adult zebrafish with a standard length (SL) (Parichy et al., 2009) of 29±1.8 mm, a weight of 423±95.7 mg, and a condition factor of 1.67±0.225 (Bolger and Connolly, 1989) (Fig. S3).

The respirometer and swim tunnel system was 3D-printed in clear resin and equipped with an O₂ sensor and a propeller driven by a motor.

The system was designed in five sections (Fig. 1B; Fig. S4): three linear sections and two curved sections. Each rounded section has a port for inlet and outlet of water through connected tubes during flushing (Fig. 1B; Fig. S4). The system was drawn in Autodesk Inventor professional 2018 and printed using a stereolithography 3D printer (formlabs, #Form 3). The linear swim-section aimed to contain the fish was printed in clear resin (formlabs, #RS-F2-GPCL-04) while the remaining pieces were printed in durable resin (formlabs, #RS-F2-DUCL-02). Honeycombs were placed before and after the swim section to promote an even laminar flow. The O₂ sensing spot (Oxfoil, Pyroscience GmbH, Germany) was attached to the inside of the system just downstream of the swim-section (Fig. 1B,C; Fig. S4). All connections between sections were fitted with O-rings and non-toxic vacuum grease (Dow Corning, high vacuum grease) was applied to the O-rings to ensure easy assembly and disassembly, and to prevent leaks. Once assembled, the respirometer and swim tunnel system were fastened with two screws to an aluminium stand. An electric motor (either a small RS PRO 12 V dc, 431 rpm #834-7663 or a large 12 V dc, 2329 rpm #417-9655) controls the speed of the 3D-printed propeller using a large O-ring as drive belt. The motor was mounted onto a holder fixed to the aluminium stand keeping it out of the water and its speed was controlled by a digital, programmable power supply with a resolution of 0.01 V (ELC, #ALR3203). The correlation between water flow rate and motor voltage was determined by video recording using a standard cell phone camera (iPhone 13 mini) after injecting dye into the system at different motor voltages. Approximately 0.1 ml dye (green, non-toxic fruit dye) was injected with a 1 ml syringe fitted with a gauge 23 needle extended with 3–4 cm of PE-50 catheter through one of the flush ports while the other flush port was closed. The calibration procedure yielded a linear relation between motor voltage (y) and flow speed (x) of y=0.4057x+1.5113 (R²=0.9917) and y=0.0689x+1.2384 (R²=0.9971) for the small (Fig. S5A) and the large (Fig. S5B) motor, respectively. The flow profile across the respirometer was flat right after the propeller and was still close to uniform in the swim-section (Fig. S5C,D). We did not observe a preferred spot to swim in for any of the tested fish (e.g. Movie 1).

To measure O₂ consumption rates, pO₂ was measured over time by using a FireSting optical O₂ meter (#FSO2-C4, Pyroscience GmbH, Germany) and data collected using Pyro Oxygen Logger version 3.312 in 1 s intervals. The entire experiment for each fish consisted of a pre-blank, SMR, MMR slow, MMR fast, and post-blank measurements (Fig. S2). All experiments were performed at 26°C. Pre-experiment background respiration (pre-blank) was determined after sterilizing (by immersion in 70% EtOH for 5 min and then allowing the EtOH to evaporate) the swim tunnel and recording O₂ consumption for 20 min at a water flow rate of 1 cm·s⁻¹. The same pre-experimental background was used for all fish as it was found to be consistently very close to zero. Post-experiment background respiration (post-blank) was measured the same way after removal of the fish. All O₂ uptake rates were corrected for background respiration using pre- and post-experiment blank rates assuming a linear increase over time. The average background respiration contributed to a mean of 23.0±13.2% of the slopes used to determine O₂ consumption and SMR.

Fish weight was subtracted from the internal volume of the swim tunnel to correct for its replacement of water. SMR and MMR data were analyzed using the R package respR version 2.3.1 (Harianto et al., 2019). All zebrafish were fasted for 24 h before being transferred to the swim-section of the sterilized system.

To determine SMR, O₂ consumption was measured overnight using intermittent-flow respirometry in cycles of 5.5 min closed and 2.5 min of flushing at a flow rate of 0.5 SL·s⁻¹. The flow rate of 0.5 SL·s⁻¹ was low enough to allow the fish to rest at the bottom. At least 16 h of measurements (>120 periods) were recorded for each fish and the O₂ consumption rate in each close period was calculated using linear regression over 200 s, i.e. disregarding the first 90 and the last 40 s of the close period (using the respR function calc_rate.int with a wait period of 90 s and a measure period of 200 s). SMR was defined to be the lower 20th percentile of all O₂ consumptions rates (Chabot et al., 2016).

MMR was measured after completing the overnight SMR protocol. We tested a slow (stepwise) and a fast (continuous) protocol. Both protocols started with a water flow rate of 2 SL·s⁻¹, which was increased by 1 SL·s⁻¹ over a 6 min (slow) or 1 min (fast) period until the fish yielded (i.e. reached exhaustion). Yield was defined as the fish resting at the honeycomb for more than 3 s 3 times within 1 min. To test if the fish had effectively reached MMR, after 3 s of resting at the honeycomb, we applied a short interruption (∼0.5 s) of the water flow, which often resulted in the fish starting to swim again. Between the MMR protocols, the fish was given at least 1 h of rest at 0.5 SL·s⁻¹.

The slow protocol used intermittent-flow respirometry for 6 min (3 min closed and 3 min flushing) per water flow rate increment resulting in one O₂ consumption rate measurement per flow rate increment. The O₂ consumption was then calculated using linear regression on either a fixed or a sliding window on each period. The fixed approach measured for 90 s and disregarded the first 60 and the last 30 s of the closed period (using the respR function calc_rate.int with wait=60 and measure=90). The sliding approach employed a 60 s sliding window method (using the respR function auto_rate.int functions with window size of 60 s).

In the fast protocol, the system was flushed continuously until a swimming speed of 2–4 SL·s⁻¹ lower than that of yield in the slow protocol was reached. Here after, we closed the system and recorded pO₂ continuously over time while increasing speed stepwise, until the fish yielded. The O₂ consumption for the fast protocol was determined by applying a 60 s sliding window on the entire closed period. For all methods, MMR was defined as the highest O₂ consumption rate observed.

We measured leakage of the system by measuring O₂ consumption in the sterilized respirometer. The chamber was filled with partly deoxygenated water. No fish were present in the chamber during this measurement. A pre- and post-blank measurement was performed as described above and was subtracted from the O₂ consumption/leakage measured. We performed the leakage test at three different flow rates 0.5 SL·s⁻¹, 5 SL·s⁻¹, and 10 SL·s⁻¹ (corresponding to the mean SL of the fish in this study, 29 mm). Leakage was only detectable at 10 SL·s⁻¹. We found a negative correlation between chamber pO₂ and measured O₂ leakage with an intersection at 107 mmHg (Fig. S6). Thus, the leakage of the respirometer is negligible at normoxic condition and only worth considering if working at a low pO₂ and a high flow rate.

The Kruskal–Wallis tests were performed in R using stats (version 4.3.1). All graphs were produced in R using the tidyverse package (version 2.0.0) (Wickham et al., 2019).

Ricki Haubro Nielsen and John Svane (mechanical workshop at the Department of Biology, Aarhus University, Denmark) provided invaluable help in drawing and printing the respirometer and swim tunnel system.