The Smithsonian Conservation Biology Institute (SCBI) manages a captive herd of scimitar-horned oryx to facilitate breeding, scientific research and contribute to the in situ and ex situ conservation of the species. SCBI is a 1,440-ha facility for research and conservation of endangered species and their ecosystems and is located outside the town of Front Royal, VA, in the foothills of the Appalachian Mountains (Figures 1A, B). The scimitar-horned oryx herd at SCBI is maintained in several enclosures over an area of 7.47 ha (Figure 1B). Animals are moved between fenced pastures with free access to grazing and to shelters or barn facilities that protect animals from the weather and enable veterinary and animal care procedures. We selected eight adult scimitar-horned oryx (Figure 1C) for this study: one vasectomized male and seven non-pregnant females (Table 1). Initially, the studied animals were part of a social group of 18 adult females, five offspring of the year and the vasectomized male (“Sweeny”). After about 6 months, the male was permanently separated from the female herd and transferred to his own pasture and barn, in visual and olfactory proximity to other oryx. Animal care personnel interacted twice daily with scimitar-horned oryx for feeding, check-ups, care, and cleaning of the barn at approximately 08h00 and 14h00 local time. The animals had free access to pastures and hay and received pellet food once a day at about 15h00. Between August and November 2019, the female herd was temporarily kept in the largest grazing area (outlined by white lines in Figure 1B) due to barn maintenance.

We relied on the Reveal LINQ™ (Medtronic, Minneapolis, MN) biologgers equipped with the latest version of the custom software “B-Ware” (Laske et al., 2018) for heart rate monitoring. Several studies have successfully used identical or similar biologgers for heart rate monitoring in other mammalian species (Laske et al., 2021; Moraes et al., 2021). The device (4.0 mm × 7.2 mm x 44.8 mm; mass 2.4 g, volume 1.2 cc) is designed for subcutaneous positioning close to the heart (Laske et al., 2018). It has a long lifespan (up to 3 years), large storage capacity (three years for daily averages and up to 400 days for high density parameters added in B-Ware), and wireless capacity for remote web-based monitoring (Laske et al., 2018).

For the biologger placement and programming, we followed previously published procedures (Moraes et al., 2021), with some changes in anatomical location and detection parameters to account for unique cardiac anatomy and physiology in bovids that affect how electrocardiograms (ECGs) are generated (Santamarina et al., 2001; Tiusanen and Pastell, 2016). In bovids, the combination of a high voltage ventricular repolarization wave (T) and a long ST interval (time between the end of ventricular depolarization and beginning of ventricular repolarization) may cause dual heartbeat detection by automated algorithms due to T wave detection as ventricular depolarization (Santamarina et al., 2001; Tiusanen and Pastell, 2016).

Scimitar-horned oryx were anesthetized with a combination of 1.3–1.5 mg/kg of ketamine (ZooPharm, Laramie WY, 82070 United States) and 0.015–0.025 mL/kg of a commercial premix of butorphanol, azaperone, and medetomidine (BAM ™, ZooPharm, Laramie WY, 82070 United States) administered by intramuscular injection (IM). Naltrexone and atipamezole (ZooPharm, Laramie WY, 82070 United States) were administered IM at 2:1 butorphanol and 5:1 medetomidine, respectively, as reversal agents. Using aseptic techniques, we inserted the device subcutaneously in a left parasternal location over the heart area (∼45° on the sagittal plane) in six animals (Figure 1C) and tested a right parasternal placement in two females (named “Rose” and “Glenda”) (Table 1). We verified optimal positioning by monitoring the R-wave amplitude (>0.15 mV). To minimize double counting of each heartbeat, with R-waves being detected as ventricular senses, we adjusted the blank time (minimum time interval between ventricular senses) between 150 and 300 ms, as needed. The incision site was closed with simple interrupted sutures (2–0 absorbable monofilament poliglecaprone 25 or braided polyglactin 910 suture) in the subcutaneous and skin layers. To determine whether these subcutaneous biologgers can be used safely in scimitar-horned oryx, animal care staff visually inspected individuals twice daily to ensure we would detect and treat possible reactions (e.g., redness, swelling) at the location of the implant.

Animals were implanted between November 2018 and January 2020 and monitored on average for 592 days (Table 1). Data collection followed methods detailed for maned wolves (Chrysocyon brachyurus) in Moraes et al. (2021). We programmed the biologgers to continuously record 2-min average heart rate in beats per minute (bpm). In addition, the biologger recorded the ECGs from the last 10 episodes of tachycardia (heart rate ≥125 bpm) or asystole (at least 4.5 s between heartbeats). To measure activity, we relied on the biologger’s internal triaxial accelerometer to record the number of minutes that the animal was active (=movement exceeded the speed of ∼1,600 m h⁻¹) at 15 min intervals (Medtronic, Minneapolis, MN).

To download the different data types stored by the biologger, we used remote transmission and direct telemetry. For remote transmissions, we installed a monitor (MyCareLink^®, Medtronic) inside the barn, about 3 m from the animals’ usual resting place. The biologger was programmed to connect wirelessly to its respective monitor every 2 h and transmit a 10-s ECG in real time and the last 10 ECGs associated with any episode of tachycardia or asystole. ECG data were transmitted via the Global System for Mobile Communications (GSM) to the individual repository created for each animal on the Medtronic CareLink^® network. Because full data downloads can only be done through direct telemetry, we conducted opportunistic direct downloads during veterinary checks or management procedures that required physical restraint of the animals in a hydraulic Tamer^® (Fauna Research, Red Hook, New York). During the direct telemetry download, data were accessed by holding the telemetry head (CareLink^® model 2090 programmer) near the implant site (approximately 5 cm telemetry range). To improve the accuracy of ventricular detection, we also used the first data download opportunity (∼a month after biologger placement) to adjust the blank time interval of all eight biologgers to 300 ms.

To evaluate the accuracy of the biologger heartbeat detection, we analyzed 883 10-s ECGs automatically sent to the GSM network. We compared the heart rate calculated from the time intervals between R-waves (RR intervals) by the biologger with the RR intervals calculated using the findpeaks function (pracma package in R; Borchers, 2022); and ECG visual confirmation. Biologger detection was considered accurate when the absolute difference between the biologger heart rate (observed heart rate) and the validated heart rate (expected heart rate) was <10%.

Although the internal clocks of the biologgers were synchronized to their location at the time of placement, the time recordings exhibited slight offsets when verified during data download. To assure the accuracy of the recorded time, we used linear regressions on individual datasets according to Moraes et al. (2021). The combined data from the eight oryx were then filtered to exclude the 2-min averages obtained with blank intervals below 300 ms and heart rate values below 30 bpm (minimum value confirmed by ECG). To calculate individual and overall heart rate averages and ranges, we used the 2-min heart rate average dataset. For the analysis of factors influencing heart rate, we summarized 2-min heart rate averages into hourly intervals. For activity, the 15-min intervals data were summed to calculate the total number of minutes per hour (total hourly activity) and per day (total daily activity). In addition, as a proxy for resting heart rate, we summarized heart rate for the hourly intervals when animals were inactive (i.e., total activity = 0).

To assess the role of ambient temperature in influencing average heart rate we downloaded 30-min interval Single Aspirated Air Temperature data provided by the National Ecological Observatory Network (National Ecological Observatory Network (NEON), 2021) for SCBI, the core site for the Mid-Atlantic region of NEON. For our study, we averaged these data at hourly intervals.

To test the influence of day length on heart rate, we relied on the daylength function provided by the geosphere package in R (Hijmans et al., 2021). This function uses the latitude of the study location (SCBI; Lat: 38.89292) and date to compute the photoperiod. We also tested the influence of astronomical seasons on heart rate, with each season starting on the respective dates of the solstice and equinox in the Northern hemisphere.

Initial temporal plots of hourly heart rate and activity confirmed circadian and seasonal waveform patterns that are best approximated by polynomial functions. We used Generalized Additive Models for Location, Scale and Shape (GAMLSS; Rigby et al., 2005) to model the influence of animal activity, hour, ambient temperature, day length, and astronomical season on heart rate in scimitar-horned oryx. GAMLSS provides univariate models in which heart rate can be modeled as an additive function of multiple explanatory variables, allowing for the use of cubic splines for analyzing waveform data (Rigby et al., 2005). We converted the temperature from Celsius to Fahrenheit to avoid negative values and facilitate analysis with cubic splines. To account for repeated measurements and individual variation of heart rate, we included individuals as random effects. To determine which covariates were most important, we compared competing models with varying variable combinations and selected the best fit model based on its Akaike Information Criterion (AIC) value (Table 4).

We also checked whether autocorrelation resulted in an overestimation of partial fixed effects of our selected model. First, we used the acf function in R to determine the autocorrelation structure of the residuals of our null model (hrm0 in Table 4). We found strong autocorrelation coefficients for lag 1 (acf = 0.71) and lag 24 (acf = 0.55) (Supplementary Figure S1), confirming the daily rhythms observed in the hourly heart rate plots. Next, we fit a model with the same structure as our selected model (hrm8 in Table 4) but using a random subset of only 500 data points per animal and compared the results with the model fit with the full dataset (67,230 data points). Fitting the model with the randomly selected small dataset resulted in lower residual autocorrelation coefficients of lag 1 (acf = 0.34) and lag 24 (acf = 0.05) (Supplementary Figure S4). Yet, the outputs for the two models generated nearly identical values for the coefficient estimates, standard errors, t and P values (Supplementary Table S3). Thus, we confirmed that autocorrelation did not change the relative importance of covariates in influencing heart rate, as assessed by our selected model using the complete dataset.

All data processing and statistical analysis were performed in R (R Core Team, 2021) and RStudio (RStudio Team, 2020).

The devices used in our study required minimally invasive surgery for subcutaneous insertion and proved to be safe for use in scimitar-horned oryx. Not a single device was rejected or caused an infection. This is consistent with other research that utilized similar biologgers in brown bear (Ursus arctos; Støen et al., 2015; Evans et al., 2016), American black bear (Ursus americanus; Ditmer et al., 2015a; Laske et al., 2018), maned wolf (Moraes et al., 2021), moose (Alces alces; Græsli et al., 2020a), and that are currently being tested for as many as nine other species (Laske et al., 2021).

Accuracy assessments are rarely provided in current research on heart rate, though there is a high potential for biasing estimates upwards during high activity or stress periods because of oversensing. Depending on the cardiac electrophysiology of the study species (e.g., high or low average heart rate at rest relative to other species), muscle noise (i.e., electromyographic activity from skeletal muscle) and/or electric noise stemming from movement of the device or the surrounding subcutaneous tissue can reduce the performance of a device’s automatic detection algorithm which was optimized for human clinical use (Laske et al., 2021). Such issues are often addressed through data filtering. But without adequate reference data filtering may become a challenge.

Our results demonstrated that the biologgers were reliable in providing continuous data and the data was highly accurate, depending on the anatomical location of the implant and the detection settings used. We evaluated accuracy by calculating RR intervals from the ECG strip and comparing them with simultaneous RR intervals recorded by the device’s automatic detection algorithm. Similar accuracy assessment techniques have previously been used to investigate heart rate in Svalbard reindeer (Rangifer tarandus platyrhynchus) (Trondrud et al., 2021). Accuracies in our study were very similar to what they reported for their highest quality data, with 94% and 96% (our study) vs. 96% and 98.5% (Trondrud et al., 2021), for data deviating by less than 5% or 10% from ECG validated data, respectively. Similarly, we found minor differences in adjusted R-squared between recorded and validated heart rate in our study (0.92) and theirs (0.96).

We tested the placement of the biologgers in left and right parasternal positions and varying blank time intervals, trying to prevent dual counting of ventricular depolarizations by the device’s algorithm. Placement on the left parasternal location and setting the blanking period after a ventricular sense at 300 ms reduced oversensing and provided highly accurate data. Implanting the biologger on the right side did not improve heartbeat detection. Data from two animals had to be excluded from the analysis because of a higher rate of erroneous detection of T-waves and electric noise as heartbeats (Figure 2B). Due to a reduced number of individuals tested for right side placement, we cannot properly assess the source of the electrical noise, but the biologger could potentially be detecting electrical signals originating from the different compartments of the stomach (Wierzbicka et al., 2021). Erroneous automated detection of ventricular depolarization (QRS) may limit long-term monitoring of heart rate in other ungulate and cetacean species, which share similar anatomical and electrophysiological characteristics of the heart (Ono et al., 2009; De Almeida et al., 2015). In ungulates, Purkinje fibers, a major component of the mammalian ventricular electrical conduction and propagation system, form subendocardial and intramural networks that extend excitation across the ventricular walls. The morphology and function of the Purkinje cell network are reflected in the ECG electrical axis of the heart in ungulates (Detweiler, 2010) and can lead to larger and longer T waves, with relatively smaller QRS complexes as seen in sheep (Kviesulaitis et al., 2018) and pig (Hamlin, 2010). When using implantable biologgers (Græsli et al., 2020a) or sensors attached to the skin (Kviesulaitis et al., 2018), T waves might be detected by the algorithm as ventricular depolarization (QRS complex), leading to double counting of heartbeats.

Accurate reference values are essential for improving care (e.g., informing anesthesia protocols) and management of captive and wild scimitar-horned oryx. To our knowledge, our study is the first to report on physiological heart rate ranges in adult scimitar-horned oryx or other desert-adapted bovids. Previously reported values for anaesthetized desert bovids varied depending on the anesthetic protocol used, with values slightly above or below our overall average (60 bpm). Drug-induced stimulant or depressant effects in the heart have been observed in scimitar-horned oryx (Pearce and Kock, 1989), Arabian oryx (Ancrenaz et al., 1996), and addax (Portas et al., 2003). We also observed low heart rates (41 ± 11 bpm) during the procedure to implant the biologger, suggesting a cardiovascular depressant effect of the anesthetic protocol used in our study.

Despite differences in body size and experimental conditions, the range in heart rate observed for scimitar-horned oryx in our study is similar to data obtained by telemetry from unrestrained individuals of other bovid species, including the relatively small roe deer (Capreolus capreolus; Theil et al., 2004; Reimoser, 2012; Arnold, 2020) and the largest member of the deer family, the moose (Græsli et al., 2020b) (Supplementary Table S4). Our results were also similar to those reported for red deer (Cervus elaphus; Arnold et al., 2004; Turbill et al., 2011; Reimoser, 2012; Arnold, 2020), eland (Taurotragus oryx; Zizkova et al., 2013) and reindeer (Rangifer tarandus; Mesteig et al., 2000; Trondrud et al., 2021) (Supplementary Table S4). Heart rate monitors also have proved useful in detecting episodes of tachycardia (heart rate >125 bpm) in scimitar-horned oryx and relating these episodes to specific environmental challenges or events. For example, one animal (“Savannah”) exhibited a heart rate of 212 bpm while confined in a Tamer^® prior to a care procedure, which is similar to values recorded for roe deer (Theil et al., 2004) when trotting (184 bpm) or fleeing from humans (255 bpm) and moose when chased by dogs (195 bpm; Græsli et al., 2020a). Our findings are also consistent with those from a previous study on maned wolves, where tachycardia events were linked to confinement, social interactions, or other exogenous challenges (Moraes et al., 2021). Similarly, heart rate monitors were used to detect extremely low or high heart rates in wild black bears and brown bears (Laske et al., 2018).

Long-term heart rate patterns in mammals are closely tied to changes in metabolic rate as the result of seasonal variation in ambient temperature (Arnold et al., 2004; Arnold et al., 2006; Arnold, 2020), activity (Theil et al., 2004; Græsli et al., 2020b), food availability and nutrition (Arnold et al., 2004; Turbill et al., 2011; Arnold et al., 2018; Arnold, 2020), and reproduction (Arnold et al., 2004; Wascher et al., 2018; Ruf et al., 2021). These patterns are often overlaid, and sometimes masked, by physiological responses to other stressors, mainly intra- and inter-species interactions (e.g., dominance hierarchies, competition, predator avoidance; Chabot et al., 1996; Turbill et al., 2013), extreme environmental events (e.g., hard winters, heat; Das et al., 2016; Arnold et al., 2018), and human-caused disturbance (e.g., roads, hunting; Laske et al., 2011; Ditmer et al., 2015a; Ditmer et al., 2015b; Støen et al., 2015; Græsli et al., 2020a).

In our study, heart rate was most strongly influenced by activity, hour, season, and to a lesser extent by ambient temperature as can be seen in our results from GAMLSS (Figure 6). Daily rhythms of activity and heart rate were both characterized by increasing values during the day towards a peak near dusk, followed by decreasing values with low activity and heart rate at night, reaching the lowest values early in the morning (Figure 4). Because food availability was not a constraint, daily activity and heart rate patterns were mostly related to animal care, feeding routines, and weather conditions. Animals would lie inactive most of the night in the barn, moving out to the pastures around 08h00, during the barn cleaning routine. Outside foraging time was minimal in winter, with the herd spending most of the time inside the barn under heat sources. In summer, animals sought shelter only at midday due to the heat load and to avoid insects. In addition to locomotor activity, we believe food intake and ruminal activity, which affect the annual heart rate in deer (Mesteig et al., 2000; Turbill et al., 2011), may play a role in the daily heart rate fluctuations in the scimitar-horned oryx in our study, possibly due to combined changes in cardiac output, blood flow in the gastrointestinal tract, ruminal temperature and metabolic rate. These factors, however, remain to be tested in scimitar-horned oryx.

The seasonal heart rate in the captive scimitar-horned oryx was impacted by factors other than animal activity. Although scimitar-horned oryx were less active in winter, their heart rate was higher in winter than in summer (Figures 4, 5). This finding is in contradiction with what has been described for free-living wild northern ungulates, which tend to reduce their metabolism and heart rate over the winter in response to extreme cold (Arnold, 2020). Yet, this hypometabolic response to extreme cold can be prevented when the diet is unrestricted (Turbill et al., 2011), which could be the case in our study. As has already been observed in domesticated ruminants (Young, 1983; Brosh, 2007), the thermal stress caused by low ambient temperatures in winter may have activated thermoregulatory mechanisms to increase the metabolic rate and heat production of scimitar-horned oryx, resulting in increases in resting heart rate. Our results show that the overall average of hourly heart rate when scimitar-horned oryx were inactive was five bpm higher in winter (56.7 bpm) than in summer (51.9 bpm) and the partial cubic spline values for ambient temperature in our best fit model (Figure 6) suggest that scimitar-horned oryx elevate their heart rate at very low temperatures. In addition, we believe that psychological stress in winter, as a result of multiple animals sharing a heat source inside a confined shelter, should have impacted heart rate as well. Lack of space to avoid agonistic behaviors has long been recognized as a source for stress in captive animals (Morgan and Tromborg, 2006; Moraes et al., 2021) and research on fecal corticoid metabolites in captive scimitar-horned oryx herds also reported increased corticoid levels during winter months (Pauling et al., 2017), which were attributed to confined housing in the winter.

Biologging of heart rate may be especially valuable for use in reintroduction projects where animals are exposed to a wide range of stressors stemming from changes in care, management, and environment. Animals respond to these stressors through complex interactions between the brain and the heart that continuously modulate autonomous heart control and adrenal gland activity. These responses can be measured indirectly by quantifying glucocorticoids that accumulate over time in urine, feces, and hair (Heimbürge et al., 2019; Palme, 2019), or more directly by continuously recording heart rate. The latter provides high temporal resolution and facilitates the detection and timing of specific stressors. As animals adjust to new management conditions or environments, we expect peaks in heart rate and a reduction in heart rate variability due to a temporary increase in cardiac adrenergic tone and catecholamines blood levels. The magnitude and duration of the heart responses to stressors, would vary among individuals depending on the context and the animal’s personality and previous experiences.

Deploying heart rate monitors prior to the release should be relatively straightforward. During this stage, the animals are mostly under human care, which allows the implantation of monitors and regular downloads. Heart rate monitors would enable evaluation of different care and management actions and would be particularly effective in monitoring and reducing stress during transport (e.g., mode of transport, use of anesthetics, types of crates), acclimation (e.g., duration of acclimation period, social grouping), and release (e.g., timing, soft vs. hard release).

In combination with behavioral observations and experimentation prior to release, heart rate monitoring could also be used to categorize animal personalities, which have been shown to affect survival during reintroduction (Bremner-Harrison et al., 2004; Sinn et al., 2014; Germano et al., 2017). Several studies have linked heart rate and heart rate variability to individual’s behavioral traits in animals (Kovács et al., 2015; Finkemeier et al., 2019). Defining personality types for scimitar-horned oryx prior to release and then using satellite-GPS tracking could inform future selection of animals for release with a goal of maximizing reintroduction success. Existing research provides evidence that this may be feasible, because scimitar-horned oryx behavior is influenced by past experiences (i.e., management background) and social context (Mertes et al., 2019; Majaliwa et al., 2022; Mertes et al., 2022). In addition, we found considerable individual variation in heart rate patterns in our study (Supplementary Table S1).

The use of the biologgers in scimitar-horned oryx post-release is logistically challenging but possible. Ideally, individual animals would be implanted with a biologger and equipped with a satellite-GPS collar before to being shipped from their source population to the release site. Currently, released individuals are being recaptured after one or two years to replace the tracking collar (Mertes, pers. Comm), which could allow opportunistic data downloads from the heart rate monitors. In the near future, we hope to be able to send biologger data via Bluetooth to the satellite-GPS transmitter for upload to an internet server. When integrated with GPS tracking, post-release heart rate may be valuable for identifying the location, timing, and duration of stressors, and their short- and long-term impacts on the animal. Examples of assessing such stressors already exist for other species. Subcutaneous heart rate monitors have been used to evaluate stress in wildlife from exposure to unmanned aerial vehicles (Ditmer et al., 2015b), roads (Ditmer et al., 2018), and hunting (Støen et al., 2015). Continued biologging over longer time periods can be used to determine important thresholds for management, such as the minimum buffer distance from roads or human settlements. It can also provide an objective metric for how long it takes for an animal to adjust and habituate to new conditions.

Scimitar-horned oryx are among the most threatened ungulates globally (Newby, 1988; Devillers and Devillers-Terschuren, 2006; IUCN SSC Antelope Specialist Group, 2016; Woodfine and Gilbert, 2016). Although the restoration of a wild population appears to be successful, the species will require continued monitoring and conservation support. Heart rate biologging may prove to be a valuable tool in these restoration efforts to assess and improve conservation and reintroduction management, and to advance our knowledge about the ecological requirements of the species.