Estimating energetics in cetaceans from respiratory frequency: why we need to understand physiology

ABSTRACT The accurate estimation of field metabolic rates (FMR) in wild animals is a key component of bioenergetic models, and is important for understanding the routine limitations for survival as well as individual responses to disturbances or environmental changes. Several methods have been used to estimate FMR, including accelerometer-derived activity budgets, isotope dilution techniques, and proxies from heart rate. Counting the number of breaths is another method used to assess FMR in cetaceans, which is attractive in its simplicity and the ability to measure respiration frequency from visual cues or data loggers. This method hinges on the assumption that over time a constant tidal volume (VT) and O2 exchange fraction (ΔO2) can be used to predict FMR. To test whether this method of estimating FMR is valid, we measured breath-by-breath tidal volumes and expired O2 levels of bottlenose dolphins, and computed the O2 consumption rate (V̇O2) before and after a pre-determined duration of exercise. The measured V̇O2 was compared with three methods to estimate FMR. Each method to estimate V̇O2 included variable VT and/or ΔO2. Two assumption-based methods overestimated V̇O2 by 216-501%. Once the temporal changes in cardio-respiratory physiology, such as variation in VT and ΔO2, were taken into account, pre-exercise resting V̇O2 was predicted to within 2%, and post-exercise V̇O2 was overestimated by 12%. Our data show that a better understanding of cardiorespiratory physiology significantly improves the ability to estimate metabolic rate from respiratory frequency, and further emphasizes the importance of eco-physiology for conservation management efforts.


INTRODUCTION
Marine mammals live a life of dual constraints, with food located underwater, and the oxygen (O 2 ) required to fuel aerobic metabolism available only at the surface. Empirical data and optimal foraging theory suggests that aerobic dives should maximize time underwater, thereby increasing foraging efficiency (Carbone and Houston, 1996;Kooyman et al., 1983). For an animal that mainly utilizes aerobic metabolism, the time available to forage underwater depends on the amount of O 2 available in the lung, blood and tissue stores, and the rate at which these stores are utilized (the metabolic rate).
Marine mammals have several traits that enhance time underwater such as: increased muscle and blood O 2 stores, a shape that minimizes hydrodynamic drag, and a physiological response during diving that helps to manage gases and maximize the aerobic dive time (Butler and Jones, 1997;Ponganis et al., 2011;Scholander, 1940). While many of these traits have been described for a range of pinnipeds and smaller cetaceans (Fahlman et al., 2008a;Kooyman et al., 1971;Reed et al., 1994Reed et al., , 2000Sparling and Fedak, 2004;Sparling et al., 2007;Williams et al., 1993;Yazdi et al., 1999), we still know very little how larger, free ranging cetaceans manage their energy budgets (Kooyman et al., 1975;Sumich, 2001;Wahrenbrock et al., 1974).
Energy budgets of free-ranging animals are critical variables in bioenergetics models used to assess important issues such as the impact of climate change, or variation in food distribution, on marine mammal populations (Bowen, 1997;Winship et al., 2002). Current bioenergetics models are more sensitive to uncertainty in energy budgets and metabolic rates than to diet composition or population size (Winship et al., 2002). Better resolution of energy budgets requires an understanding of how animals manage their time underwater, their physiological limitations, and the energy expenditures associated with different activities.
Estimating the field metabolic rate (FMR) in breath-hold diving animals is challenging, as it is logistically difficult to investigate free-ranging animals, especially when they spend considerable time underwater and migrate over large areas. Several indirect methods have been used to assess FMR in diving species (Dolphin, 1987a;Fahlman et al., 2013;Folkow and Blix, 1992;Hindell and Lea, 1998). These include the use of stable isotopes (Boyd et al., 1995;Costa, 1988), and relating activity levels or heart rate to metabolic rates in calibrated studies (Boyd et al., 1995;Butler et al., 2004;Fahlman et al., 2013Fahlman et al., , 2008bHalsey et al., 2008;McPhee et al., 2003;Williams et al., 2004;Young et al., 2011a,b).
While it is known that both vary with activity level and following diving Reed et al., 1994Reed et al., , 2000, it may be valid to assume that over time, average values of these parameters can be used to estimate FMR from breathing frequency. This method to estimate FMR is attractive in its simplicity, but the assumptions about VT and ΔO 2 have not been validated. While validation experiments are logistically difficult in larger cetaceans, the current study sought to determine how accurate we could estimate metabolic cost during rest and following a bout of aerobic exercise. The vital capacity (VC) of cetaceans is high and estimated to be approximately 80-90% of their total lung capacity (TLC) Kooyman and Cornell, 1981;Kooyman et al., 1975;Olsen et al., 1969). Some studies have assumed that VT is independent of activity level and is close to VC (Dolphin, 1987a;Williams and Noren, 2009). If all breaths are close to VC, increased O 2 demand, as occurs during periods of increased activity, would be achieved through an increase in breathing frequency (Kooyman et al., 1971). Based on these relationships, it has been expected that breathing frequency is tightly linked to energy requirements and may be used to assess or compare the metabolic cost of different activities, e.g. resting or swimming at the surface versus diving.
However, there appears to be considerable variability in VT during spontaneous breaths in both small Olsen et al., 1969;Sumich, 2001) and large cetaceans (Wahrenbrock et al., 1974). Tidal volumes of resting pilot whales (Globicephala scammoni/macrorhynchus) and bottlenose dolphins (Tursiops truncatus) range from 20-88% and 24-40% of TLC, respectively Olsen et al., 1969). In two publications on gray whales VT was reported to vary by up to 50% between breaths (Wahrenbrock et al., 1974) or to range by two orders of magnitude (1 litre to approximately 180 litre, Sumich, 2001). Like VT, ΔO 2 is highly variable between breaths, increasing with the preceding breath-hold duration (Reed et al., 2000;Sumich, 2001) and decreasing with breath number following diving (Ridgway et al., 1969). Thus, VT and ΔO 2 appear to vary both within and between species, which raises concern over the validity of estimating energy expenditure based on static estimates of these parameters.
In order to refine computation of free-swimming marine mammal energetics, we determined how accurately we could estimate FMR in a small cetacean, the bottlenose dolphin, before and after exercise, from respiratory frequency alone. We compared 3 different methods to estimate metabolic rate from breathing frequency and compared the results to experimentally derived measurements. Our results indicate that with an understanding of the underlying physiology, the metabolic rate before and after exercise can be estimated to within 2% and 12%, respectively.
During the pre-exercise resting trials, neither the number of breaths, the end-expiratory O 2 , the volume of O 2 exchanged per breath, average O 2 content per breath, nor VT changed (mixed model ANOVA, P>0.1). Following exercise, VT (Fig. 1), and the volume of O 2 exchanged per breath decreased, while end-expiratory and average O 2 content increased (Fig. 2, P<0.05).

Estimated V̇O 2 from breaths
We used three methods (A-C) to estimate metabolic rate from breathing frequency. Methods A and B used Eqns 2A-C and assumptions used to estimate FMR in free-ranging mysticetes. For method C, we used the measured VT and the integrated volume of O 2 taken up during each breath to estimate the average O 2 content for that breath, Eqn 1D. The estimated average O 2exp was used to determine average ΔO 2 , which in turn was used to estimate V O2 and  (Fahlman et al., 2011;Kooyman, 1973). VȮ 2 for method C. Thus a mass-independent estimate of respiratory effort (Eqn 1B) was used estimate VT.

DISCUSSION
We used breath-by-breath respirometry to determine if counting the number of breaths can be used to estimate metabolic rate in a small cetacean at rest, or during recovery from exercise. Our data clearly indicate that both VT and ΔO 2 vary following a bout of exercise in bottlenose dolphins. We compared different assumptions to estimate VȮ 2 , most which have been used to estimate FMR in large mysticetes following diving or during migration (Armstrong and Siegfried, 1991;Blix and Folkow, 1995;Christiansen et al., 2014;Dolphin, 1987a,b;Folkow and Blix, 1992). These methods overestimated measured metabolic rates by 216-501%. Accounting for temporal changes following exercise, we created a model with which we were able to estimate the resting and post-exercise recovery metabolic rate to within 2% and 12%, respectively (Method C). While our data suggest that improved knowledge of the physiology of these animals significantly enhances the ability to more accurately predict energy needs, there are some important caveats to consider. First, we studied a small odontocete and both the differences in size (allometry) and physiology warrant consideration. Second, these additional data were performed on dolphins during rest and following a standardized bout of aerobic exercise at the surface. Thus, direct comparisons with diving animals may not be valid. Similar to our findings, previous data in both pinnipeds (Kerem et al., 1975;Kooyman et al., 1971;Reed et al., 1994) and cetaceans (Reed et al., 2000;Ridgway et al., 1969) indicate that both ΔO 2 (Fig. 1) and VT (Fig. 2) change significantly as animals recover from exercise. Method A and B, which have been used to estimate FMR in humpback whales (Dolphin, 1987a) and minke whales (Armstrong and Siegfried, 1991;Blix and Folkow, 1995;Christiansen et al., 2014;Dolphin, 1987a,b;Folkow and Blix, 1992), overestimated VȮ 2 by between 216-501%. These methods assume that all breaths during a surface interval have an average VT between 60-80% of VC and that ΔO 2 is 10-11.6% (Armstrong and Siegfried, 1991;Blix and Folkow, 1995;Christiansen et al., 2014;Dolphin, 1987a,b;Folkow and Blix, 1992). There are a number of possible differences that may explain this large error such as species or allometric differences, and the difference in behaviors investigated.
Our data from bottlenose dolphins in this and our previous study  show that in this small odontocete most voluntary breaths and maximal VTs following a bout of aerobic exercise range from 25-45% and 63-95% of TLC, respectively, or between 27-102% of VC assuming a minimum air volume of 7% of TLC . Only seven breaths following exercise exceeded 80% of TLC est , and most breaths were only about 37-40% of TLC est (Fig. 1B). Cetaceans have the capacity to exchange much of their TLC, and probably do so at times for a few breaths following a long dive or intense exercise, allowing them to rapidly replenish the O 2 stores and reduce the surface interval and recovery period (Boutilier et al., 2001;Fahlman et al., 2008a). However, we observed considerable variation in VT and average values far below 60-80% of VC (Fig. 1B). Thus, the assumed average VT in Method A and B is a major contributor to its large error and overestimation of metabolic rate. As our experimental design only involved experiments on a small cetacean during rest and recovery from exercise, this conclusion may have to be viewed with caution as there may be considerable differences during versus after exercise and between species, especially large cetaceans.
The VT has been reported to vary considerably in large cetaceans both at rest and following apnea (Epple et al., 2015;Kasting et al., 1989;Kooyman et al., 1975;Sumich and May, 2009;Wahrenbrock et al., 1974; Kriete, PhD thesis, The University of British Columbia, 1995). The average VT for a gray whale in human care (Gigi II) was 220 litres when she weighed 6200 kg (Wahrenbrock et al., 1974). At this size, TLC est would be approximately 416 litres, and with a minimum air volume of 7% of TLC, VC would be 387 litres (Fahlman et al., 2011;Kooyman, 1973). Thus, the average VT was about 57% of VC. Similar calculations during growth (2000-6200 kg) showed that average VT increased steadily from 15% to 55% of VC (Wahrenbrock et al., 1974). Similar calculations for the killer whale and beluga whale (Delphinapterus leucas) suggest that VT ranges from 35-95% of VC in the former (Kasting et al., 1989; Kriete, PhD thesis, The University of British Columbia, 1995) and from 23-80% in the latter (Epple et al., 2015;Kasting et al., 1989). Consequently, this large variation between studies may at least in part explain the large difference between observed VȮ 2 and that estimated from Methods A and B.
Another reason for the large error for Method B stems from the method to estimate VC (Table 2). For example, the range of estimated VC for Method A ranged from 10.3-15.5 litres for animals ranging from 168-250 kg, while for Method B the range was 19.2-26.6 litres. Thus, the equation for VC in Method B may only be valid for minke whales and inappropriate for smaller cetaceans.
Additional error may for Methods A and B may stem from the value assumed average O 2 content of the lung of 10 and 11.6%, respectively (10 and 11.6 kPa). In previous studies the end-tidal P O2 in the gray whale ranged from 7.1 to 11.2 kPa (Wahrenbrock et al., 1974), and from 7.5 to 16.0 kPa in the harbour porpoise (Reed et al., 2000). In the current study the end-expired O 2 ranged from 8.1 to 17.8 kPa and from 5.8 to 18.1 kPa before and after exercise, respectively. Thus, our end-expiratory values are within the range of those previously reported for both a small and a large cetacean. In another study, average O 2 content of 5.4 kPa was reported from the first breath from a bottlenose dolphin after a 200 m dive, increasing to 11.9 kPa for the second breath (Ridgway et al., 1969). In the killer whale, the average O 2 content was around 12-16 kPa for short apneas at all activity levels and decreased both with activity level and apnea duration, and the lowest measured levels were around 9.5 kPa (Figs 2-5 in B. Kriete, PhD thesis, The University of British Columbia, 1995). However, these values are not directly comparable to those collected in the current study or from Wahrenbrock et al. (1974) or Reed et al. (2000). In the deep diving dolphin and killer whale the air was collected in a bag and subsequently analysed for mixed average O 2 for the entire breath. In the current study, on the other hand, the average expired O 2 was the average of exhaled air during the exhalation. During the expiratory phase in bottlenose dolphins, respiratory flow increases rapidly to a more or less constant rate and O 2exp decreases until alveolar gas is sampled Kooyman and Cornell, 1981;Mortola and Seguin, 2009). If we assume that the cumulative O 2 taken up per breath is the integrated product of the instantaneous flow-rate and O 2exp , we can estimate the expected average O 2 content for each breath. In other words, the integrated volume of O 2 taken up per breath and the measured VT can be used to more accurately to predict the expected average lung O 2 content (Fig. 2B). This resulted in an average O 2 content of all breaths of 15.2±1.8 kPa and a range from 9.2-19.8 kPa, which is similar to those of the killer whale (B. Kriete, PhD thesis, The University of British Columbia, 1995) and harbour porpoise (Reed et al., 2000), but higher than the values from the dolphin following a dive to 200 m (Ridgway et al., 1969). Thus, ΔO 2 (10-11.6%) used in the previous studies (Armstrong and Siegfried, 1991;Blix and Folkow, 1995;Dolphin, 1987a;Folkow and Blix, 1992) is much higher than those observed in the dolphins. While Methods A and B may provide reasonable methods to estimate FMR, careful consideration of selected parameters are required and may explain the large error when compared with the dolphin during rest and recovering from exercise.
Using Eqn 1D allowed us to predict O 2exp during the recovery from exercise to account for the O 2 debt that develops as the animal swims actively (Method C). This method allowed us to estimate the metabolic rate from breath numbers to within 12% following exercise. Diving mammals develop an O 2 debt while foraging underwater as the internal O 2 stores are used to fuel aerobic metabolism (Fahlman et al., 2008a;Kooyman and Ponganis, 1998;Scholander, 1940). Respiration rates often show a positive correlation with increasing activity (Blix and Folkow, 1995;Sumich, 1983) and dive depth and/or duration (Dolphin, 1987b;Würsig et al., 1986), and changes in exhaled gas content suggest that lung O 2exp decreases during the recovery (Boutilier et al., 2001;Fahlman et al., 2008a;Reed et al., 1994Reed et al., , 2000Ridgway et al., 1969). Whether this relationship is similar following periods of prolonged apnea is likely but needs to be verified.
In summary, we used breath-by-breath respirometry to test if breathing frequency can be used to estimate metabolic rate in a smaller cetacean during rest and recovery from exercise. Our results suggest that the average O 2 extraction and VT are lower than assumptions used in previous studies estimating FMR in mysticetes during migration and following diving (Armstrong and Siegfried, 1991;Blix and Folkow, 1995;Christiansen et al., 2014;Dolphin, 1987a,b;Folkow and Blix, 1992), contributing to overestimation by these particular methods. We found that accurate estimates of the average VT (Eqn 1B, Fig. 1B) and estimated O 2exp (Eqn 1D, Fig. 2B) content provided reasonable estimates of measured VȮ 2 (method C), to within 2% for pre-exercise rest and within 12% for post-exercise recovery data.
The two methods that used the upper limits for VT and O 2 extraction overestimated measured oxygen consumption by 216-501%. One explanation may be the different behaviours investigated, where the previous studies focused on FMR during foraging, while in the current study we looked at the effect of rest and recovery from exercise. Individual breaths can be detected on acoustic bio-logging tags (e.g. Génin et al., 2015), and have been used to calculate breathing rates in free-ranging animals; however, the use of acoustic information (e.g. duration, frequency, or amplitude) from each breath has not been explored in its potential to provide measures of the variability critical for accurate estimates of energetics from breathing patterns (Sumich and May, 2009;van der Hoop et al., 2014b; our personal observations). Current work is underway to do so, and would greatly improve our ability to incorporate respiratory dynamics to properly estimate bioenergetics in free-ranging marine mammals. In addition, attempts are underway to determine if these relationships are valid following surface apneas and may shed additional information that may improve this potentially simple and powerful technique. Such physiological studies are crucial to determine the physiological dynamics, variability, and limitations of marine mammals and will enhance our ability to predict how they may respond to changes in the environment.

Animals
All experiments were conducted under IACUC approval from the Animal Care Committee of Texas A&M University -Corpus Christi (AUP04-14), and the research committee of Dolphin Quest. Four adult male Atlantic bottlenose dolphins in a closed lagoon were used for all experiments; the name, body mass, straight length, and approximate age of the animals are summarized in Table 1. All experiments were performed using operant conditioning and participation by the dolphins was voluntary; animals were not restrained and could refuse to participate or withdraw at any point during experimental trials. Prior to initiating the study, animals were desensitized to the equipment and trained for novel research-associated behaviours. Thus, data on respiratory variables and metabolism were collected in dolphins that were in a relaxed, normal physiological state, before and after exercise.
Apparatus, Holliston, MA), connected to the pneumotachometer with two, 310 cm lengths of 2.0 mm I.D., firm walled, flexible tubing (see Figs 1 and 2 in Fahlman et al., 2015). The pneumotachometer was calibrated using a 7.0 litre calibration syringe (Series 4900, Hans-Rudolph Inc, Shawnee, KS). The signal was integrated and the flow determined by calibrating the pneumotachometer using the syringe immediately before and after each trial, through a series of pump cycles at various flow speeds  to calibrate the differential pressure and flows for the expiratory and inspiratory phases to be determined.

Resting and post exercise metabolic rates
Each trial consisted of resting, exercise and recovery period. For the resting period, the dolphins rested (remained stationary) for at least 5 min, during which time the respiratory flows and expired gas composition were measured. The average O 2 consumption rate (VȮ 2 ) was measured for the last 2 min of this resting period, during which time the animal was acclimated to the task and the VȮ 2 was stable and provided an accurate estimate of resting VȮ 2 . Next, the dolphin was instructed to either swim in a predetermined path for 10 min, or to swim behind a remote control model boat at a set pace of 3 m s −1 for 5 min. The dolphin was free to breathe at will while exercising. At the completion of the swimming task, the dolphin returned immediately to the measurement station, where postexercise respiratory flow and expired gas composition were continuously measured for up to 5 min (see Fig. 3B in van der Hoop et al., 2014a).

Respiratory gas composition
Respiratory gases were subsampled via a port in the pneumotachometer and passed through a 310 cm length of 2 mm I.D., firm walled, flexible tubing and a 30 cm length of 1.5 mm I.D. Nafion tubing, to fast-response O 2 and CO 2 analysers (ML206, Harvard Apparatus, Holliston, MA, USA) at a flow rate of 200 ml min −1 , with a response time for a 90% change to equilibrium for O 2 and CO 2 of 67 ms and 94 ms, respectively . The gas analysers were connected to the data acquisition system and sampled at 200 Hz. The respiratory gas signals were phase-corrected to match the respirations, to account for the lag resulting from gas flow through the tubing. The expiratory flow-rate and expired O 2 and CO 2 content were multiplied to calculate the instantaneous VȮ 2 and carbon dioxide production rate (VĊ O2 , litre CO 2 min −1 ). The instantaneous VȮ 2 and VĊ O2 were integrated to yield the total volume of O 2 and CO 2 exchanged during each breath. The volumes were summed for each breath during the trial period and divided by the duration of the trial to provide an estimate of the VȮ 2 and VĊ O2 for that time period. The gas analyser was calibrated before and after the experiment using a commercial mixture of 5% O 2 , 5% CO 2 , and 90% N 2 , certified accurate to at least 0.01%. Ambient air was used to check the calibration before and after each experimental trial. Mean daily air temperature and humidity were 27.8±1.2°C (range 25.0-30.0°C) and 58.6±7.5% (46-77%). The average water temperature in the lagoon was 24.6±0.5°C. All gas volumes were converted to standard temperature pressure dry (STPD, Quanjer et al., 1993). Exhaled air was assumed saturated at 37°C, inhaled air volume was corrected for ambient temperature and relative humidity.

Estimated V̇O 2 from breaths
Three different methods (A-C) were used to estimate VȮ 2 from the dolphins. Each model incorporated measured or assumed values to determine which variables have the greatest impact on the VȮ 2 estimate ( Table 2). All methods (A through C) used the same basic equations to estimate VȮ 2 , where the volume of O 2 per breath (V O2 , litre O 2 ) was estimated as: where VT is the tidal volume, and ΔO 2 (% O 2 breath −1 ) was determined as the difference in O 2 concentration between inspired (O 2insp , 20.94% O 2 ) and expired (O 2exp , %) air.
The total volume of O 2 taken up per breath was calculated separately for the pre-and post-exercise periods, and divided by the total duration (t, min) of all breaths in that trial period to estimate the VȮ 2 : For method A, we used the assumption that VT is on average 80% of the vital capacity (VC), as estimated from Stahl's allometric equation (Stahl, 1967), and ΔO 2 to be 10% (Armstrong and Siegfried, 1991;Dolphin, 1987a). For method B, we used the assumptions from those used to estimate FMR in minke whales (Blix and Folkow, 1995;Christiansen et al., 2014;Folkow and Blix, 1992), where VT was assumed to be 60% of the vital capacity, Eqns 4 and 7 in Folkow and Blix (1992), and ΔO 2 was assumed to be 11.6%. For method C, we used the measured VT and integrated volume of O 2 taken up during each breath to estimate the average O 2 content for that breath, i.e. average O 2exp The estimated average O 2exp was used to determine average ΔO 2 , which, in turn was used to estimate V O2 and VȮ 2 in method C.

Data processing and statistical analysis
Metabolic data are reported as the average VȮ 2 for a trial. The relationship between a dependent variable and experimental covariates was analysed using mixed effects models (lme, R: A Language and Environment for Statistical Computing, R Foundation for Statistical Computing, version 3.1.0, 2014). Individual animals were treated as random effects to account for correlation between repeated measurements on the same individual (Littell et al., 1998). Best models were chosen by the Akaike information criterion (AIC) against the null model (AIC null ) and significant parameters assessed by the t-value between the estimate and its standard error. P-values ≤0.05 were considered as significant. Data are presented as the mean± standard deviation (s.d.), unless otherwise stated.