Ovulation induction in anovulatory southern white rhinoceros (Ceratotherium simum simum) without altrenogest

Lack of ovulation is common in captive southern white rhino females and contributes to poor reproductive success. We show that ovulation can be induced efficiently with exogenous hormones using a single treatment protocol. This information can lead to improved genetic management of captive populations acting as insurance against extinction.


Introduction
The Sumatran (Dicerorhinus sumatrensis), Javan (Rhinoceros sondaicus) and black rhinoceros (Diceros bicornis) are des-ignated as Critically Endangered. The greater one-horned (Rhinoceros unicornis) is listed as Vulnerable and the southern white (Ceratotherium simum simum) as Near Threatened by the IUCN (iucn.org, downloaded 07 June 2018). Survival in the wild is threatened by habitat loss and illegal poaching for their horns (Talukdar et al., 2008;van Strien et al., 2008a, b;Emslie, 2012a, b), rendering captive populations important genetic reservoirs for continued survival of these species. However, population growth is slow due to the lengthy 16month gestation in all rhinoceros species. Additionally, each species has distinct reproductive characteristics (e.g. cycle length and pre-ovulatory follicle size; Roth, 2006). The southern white rhinoceros (SWR) experiences additional challenges in ex situ as a substantial number of captive females are diagnosed as 'acyclic' or 'irregularly cyclic'  and females born in captivity often experience infertility associated with high phytoestrogen diets (Tubbs et al., 2012(Tubbs et al., , 2016. While the exact mechanism for acyclicity is unknown, it presents a specific barrier to overcome before assisted reproductive technique (ART) development can be pursued in earnest. ARTs such as ovarian control [including follicle stimulation, ovulation induction, corpus luteum (CL) disruption, etc.], artificial insemination (AI), ovum pickup and in vitro techniques can help overcome the challenges of rhinoceros reproduction in captivity. Development of these techniques relies heavily on a thorough understanding of female reproductive physiology (Roth, 2006;Hildebrandt et al., 2007;Hermes et al., 2009a, b;Roth et al., 2018). Despite substantial published hormone data describing SWR reproduction (Hindle et al., 1992;Schwarzenberger et al., 1998;Patton et al., 1999;Brown et al., 2001;Roth, 2006;Hermes et al., 2009bHermes et al., , 2012van der Groot et al., 2013), relatively little longitudinal ultrasound data exist on estrous cycle characterization in this species (Radcliffe et al., 1997;Patton et al., 1999).
Estrous cycle parameters have been determined for the SWR and include two distinct cycle lengths of ∼30 or 70 days (Radcliffe et al., 1997;Patton et al., 1999;Roth, 2006;Brown et al., 2001;Ververs et al., 2015). Since most studies on white rhinos have utilized only progesterone analysis to characterize estrous cycles (Schwarzenberger et al., 1998;Brown et al., 2001;Carlstead and Brown, 2005;Hermes et al., 2007), the acyclic designation may be incomplete. Although progesterone metabolite profiles indeed confirm that ovulation does not occur, the inability to accurately measure estrogen metabolites in white rhinos Roth et al., 2018) leaves the question of cyclic follicle growth unanswered. Without concurrent ultrasound examination the possibility of ovarian activity cannot be excluded. For the purposes of this paper, we refer to females that exhibit follicular growth, but fail to ovulate, as anovulatory. The next steps in understanding anovulation (i.e. follicular growth without ovulation) include investigating the efficiency of ovulation induction with exogenous hormone treatment.
However, published reports provide few details describing follicle growth, time from GnRH treatment to ovulation, follicle or reproductive tract characteristics or resulting luteal parameters. The omission of valuable information on ovarian events leading up to GnRH administration as well as those following induced ovulation Hermes et al., 2009aHermes et al., , 2012 identified the need for additional data collection as well as investigation of a different treatment protocol.
Here we describe the utilization of an injectable GnRH analog (deslorelin acetate, SucroMate TM ), commonly used in domestic horse reproductive management, to induce ovulation in anovulatory SWR. We employed specific criteria for GnRH treatment and report resulting luteal phase parameters. Our aims were to (i) achieve ovulation from a novel protocol utilizing follicle growth and size as criteria for treatment, (ii) observe specific time of ovulation following treatment and (iii) describe differences between resulting cycle lengths.
As wild rhinoceros populations continue to decrease, captive populations will become critical resources for potential reintroduction efforts. Genetic diversity is a key element in a healthy population and can be a challenge to maintain in captivity. ARTs will enable more efficient use of genetic material, but they require a sound understanding of reproductive physiology as well as reproductive obstacles.

Animals
This research was conducted as a part of a larger project and in accordance with animal use protocols approved by San Diego Zoo Global Institutional Animal Care and Use Committee (IACUC #15-009). Six SWR females, #2194, #2195, #2196, #2197, #2198 and #2199 (SB#s), were brought to the San Diego Safari Park Rhino Rescue Center (San Diego, CA, USA) in November 2015 from South Africa where they were semi-free ranging on three different conservancies. All females (n = 6) were wild-born and estimated ages ranged between 4 and 7 years old at the beginning of the study (Supplementary Material). Females were housed as a bachelorette herd at the time of data collection. One female #2197 came into captivity pregnant and delivered a stillborn calf in November, 2016. All females had access to a barn, cable yards, concretewalled yards or exhibit (4 acres total) dependent upon daily management requirements. Each animal received 1.8-2.3 kg of commercial pellet individually, as well as supplemental timothy hay cubes, orchard grass hay and produce for training purposes. The group received up to two bales of Bermuda hay and free access to mineral blocks. Data were collected from September 2016 to August 2017 for all individuals except #2197 whose data were collected post-partum from December 2016 to August 2017.

Serial ultrasound exams
Reproductive tract ultrasound exams were performed in a custom-designed chute allowing personnel protected access through a rear gate. The chute was equipped with stationary sides, such that rhino movement was limited but not restricted. Rhinos were trained with positive reinforcement and operant conditioning to stand voluntarily in the chute without sedation for transrectal ultrasound exams one to three times per week. Prior to entering the chute, individuals were behaviorally assessed. Keepers provided food items and verbal reinforcement during ultrasound sessions and all animal participation was voluntary. If the animal became agitated or restless the exam was terminated and she was released from the chute. Sessions typically lasted 10-30 min, including manual fecal voiding. A Sound LogiqE ultrasound unit with a 3.5-5 MHz convex probe was used for all exams. Each exam intended to image the cervix, uterine bifurcation, uterine horns and both ovaries. A probe extension allowed visibility of entire uterine horns and ovaries. Ovarian structures were classified, counted, measured and recorded.

GnRH treatment trials
Treatment with a GnRH analog as a part of a criteria-specific protocol was evaluated. Deslorelin acetate (SucroMate TM , 4.5 mg, 2.5 ml) was administered via single injection delivered to the neck by remote dart or by hand. Darts were immediately removed and all injections delivered the full dose. Injections were given when the following criteria were met: (i) follicle growth was confirmed by serial ultrasound exams and (ii) follicle diameter was ∼35 mm. Females were examined via ultrasound 24, 36 and 48 hours post-injection. Following the 48-hour time point, animals were not examined again until their next twice weekly exam to confirm CL formation. Animal participation was voluntary and varied through the study period; therefore trial replications occurred more frequently in some animals than others.

Fecal steroid extraction
Fecal samples were collected at least three times per week and stored at −20 • C until processing prior to hormone analysis. Samples were lyophilized, pulverized and sifted through a 0.045-inch mesh screen to remove vegetation and debris. For all samples up to 5 May 2017, 0.1 g of sifted fecal material in a 16 × 100 mm glass tube was combined with 5 ml of 90% aqueous ethanol (EMD Millipore, Billerica, MA) and boiled at 80 • C for 20 min. Samples were centrifuged (Thermo Scientific Sorvall Lengend XTR, Waltham, MA) for 10 min at 1000 x g at room temperature and the supernatant was recovered. Remaining fecal material was mixed with 5 ml 90% aqueous ethanol, pulse vortexed and centrifugation was repeated. Supernatants were combined, dried under air and resuspended in 1 ml ethanol.
For samples after 5 May 2017 0.2 g sifted fecal material in a 50 ml polypropylene tube was combined with 20 ml of 80% methanol (Fisher, Waltham, MA) in sterile water and vortexed at room temperature for 30 min (Fisher Scientific Multi-Tube Vortexer, Waltham, MA). Samples were centrifuged for 10 min at 4000 x g at room temperature (Thermo Scientific Sorvall Lengend XTR, Waltham, MA) and the supernatant was recovered.
Transition to a new fecal extraction method was initiated to decrease sample processing time and expedite analysis. The 80:20 methanol method was chosen based on reports by Palme et al. (1997Palme et al. ( , 2013 demonstrating this as the most effective method of hormone extraction. To ensure values were comparable between methods, extractions and radioimmunoassays (RIAs) for progestogens were performed with each method simultaneously (n = 196, distributed throughout 12 extractions and assays). Progestagen values did not differ between the two extraction methods (Pearson correlation r = 0.948, Paired T-test for significant difference P = 0.735). Methanol extraction efficiency measuring recovery of added tritiated progesterone (Perkin Elmer, Waltham, MA) was 82% ± 2.8%.
Fecal extract dilutions were determined by parallelism and adjusted as necessary to ensure optimal assay antibody binding (20-85%). Two replicates of each sample extract were brought to 0.5 ml in 0.1 M phosphate buffered saline (PBS) pH 7.0 with 1% gelatin before assay.  counted in a Beckman liquid scintillation counter (LS6500) for 1 min.
The assay was validated with a parallelism of serially diluted fecal neat and a progesterone standard curve (r = 0.986) and an accuracy test determined by recovery of known quantities of progesterone standards added to a pool of fecal extract (R 2 = 0.986). Assay sensitivity was 7.55 pg/tube. Average intra-assay coefficient of variation was 8.59% and inter-assay coefficients of variation were 4.67% for low controls (average 45.5 pg/tube) and 3.52% for high controls (average 206.7 pg/tube). The RIA was also verified with reverse phase high performance liquid chromatography (Waters Nova-Pak Reverse Phase C18 4 mm 3.9 × 150 mm column, Milford, MA) to demonstrate progestagen detection.

Statistical analysis
When ovulation was not induced, dominant follicle growth and subsequent regression coincident with no substantial progestagen rise above baseline was recorded as an anovulatory cycle. Baseline progestagen values were determined for each individual by an iterative process in which values higher than the mean plus 2 SD were excluded. The mean was then recalculated and the value exclusion process was repeated until no values were higher than the mean plus 2 SD . Cycle length was determined to be either short (<50 days) or long (≥50 days) by calculating the length of time progestagens were elevated above baseline following ovulation induction (female rhinos #2194, #2196, #2197 and #2198) or observed spontaneous ovulation  between previously determined cycle lengths of 30 or 70 days (Radcliffe et al., 1997;Patton et al., 1999;Brown et al., 2001;Roth, 2006;Ververs et al., 2015) and has been used previously to determine long cycle lengths in SWR . Cycle parameters were as follows: number of days from observed ovulation to progestagen rise above baseline, number of days progestagens were elevated above baseline, maximum progestagen concentration (ng/g), number of days the CL was visible and maximum observed luteal structure size (mm). Cycle length and parameters are inclusive of days that were not sampled in order to calculate longest/largest possible time frames. All statistical analyses were performed using R studio (version 1.1.383). Welch's t-test was used to determine significant differences between cycle parameters and cycle designations ('short' or 'long') as well as effect of treatment on follicle outcome ('ovulation' or 'regression'; stats package, R Development Core Team, 2008). All data are presented as mean ± SD and considered significantly different at P < 0.05.

Results
Ovarian follicular activity was apparent in all females examined during the study, although hormone analysis indicated anovulation in five of six females prior to treatment (Fig. 1).
Notably, a single female (#2195) ovulated consistently, evidenced by regular fluctuations in fecal progestagen and single corpora lutea that were not concurrently visible (i.e. visible corpora lutea did not overlap temporally; Fig. 1A). This female exhibited both short (n = 5) and long (n = 3) cycles. Progestagens remained above baseline 22.0 ± 3.2 days (range, 19-26) in short cycles and 59.0 ± 9.5 days (range, 52-70) in long cycles (Table 1). Additional ultrasonographic information is not available as this female was only examined once per week. No ovulation inductions were performed on this individual.
Two females (#2194 and #2198; Fig. 1C and D) showed no signs of ovulation prior to induction trials. Two females (#2196 and #2199; Fig. 1F and B) experienced brief elevations in progestagen without treatment, but no CL was visualized concurrent with progestagen rise and were therefore not considered to be ovulatory. Finally, #2197 came to the Safari Park pregnant and calved on 13 November 2016. She displayed a single ovulation ∼45 days post-partum as evidenced by visualization of a CL and progestagen rise above base-line, but failed to ovulate spontaneously thereafter (Fig. 1E). During anovulatory cycles in all females except #2195, follicles grew to pre-ovulatory sizes, but then regressed. Follicles typically developed on alternating ovaries cyclically and growth often occurred during regression of a previously large follicle. Follicles appeared clear, without particulate infiltration or webbing ( Fig. 2A, B and E). When follicles reached maximal size, visible indicators of regression were apparent such as separation of the granulosa cell layer from the follicle wall and the follicle wall becoming irregular or thickened (Fig. 2E). Anovulation was confirmed when fecal progestagens remained below baseline and there was no visible CL. Four females were induced to ovulate with GnRH treatment (#2194, #2196, #2197 and #2198), while one female (#2199) was not treated and did not form any luteal structures during the study period (Fig. 1B). Following ovulation induction and subsequent CL regression, the next dominant follicle was allowed to proceed without treatment. These unstimulated dominant follicles (n = 6) reached a maximal diameter (43.8 ± 6.1 mm; range, 37-54 mm), followed by regression and growth of another follicle that was subsequently induced to ovulate.
GnRH treatment resulted in formation of a luteal structure, whereas no action resulted in regression of the follicle (P < 0.001). GnRH treatment resulted in ovulation 9 of 11 times (81.8%) and hemorrhagic anovulatory follicle (HAF) formation twice (18.2%; Table 2, Fig. 4). Both HAFs formed in the same female (#2198) and all long cycles after treatment occurred in the same female (#2194). Follicle size at treatment (0 hours) ranged from 34 to 36 mm in diameter. Following treatment, the follicle was observed at 24 and 36 hours but had disappeared by 48 hours. Ovulation was confirmed by CL formation (Fig. 3). HAF formation was determined by the persistence of the follicle at 48 hours followed by a substantial increase in size, up to 76 mm diameter, and development of fibrous echogenic bands that quivered upon ballotment (Fig. 2D). Mean follicle sizes were 35 mm at 24 and 36 hours for ovulations as well as at 48 hours post-treatment in follicles that formed HAFs, with luteal tissue averaging 36 mm 4-9 days post-treatment (36.6 ± 5.9; Fig. 4). These measurements are inclusive of HAFs as they fell within confirmed CL size ranges up to 4 days post-ovulation. However, maximal sizes were observed 7 and 15 days post-injection for each HAF (Fig. 1D). Progestogen levels were attenuated in HAF cycles (maximum, 1622 ng/g).  Ovulation was confirmed following treatment by elevation of progestagen above baseline for each individual (Fig. 1A-F) and visualization of a CL (Fig. 2C). Progestagens were elevated following HAF formation, and these data are included as cycle parameters (Fig. 1D). Both short (n = 8) and long (n = 3) cycles resulted from GnRH treatment ( Table 2). Duration of progestagen elevation (days) was significantly different (P < 0.001) between short and long cycles, while visibility of luteal structures was not significant (P = 0.11). In treated females, progestagens were elevated for 27.4 ± 7.6 and 62.7 ± 5.5 days (range, 19-40 and 57-68 days) and luteal structures were visible for 37.1 ± 13.6 and 55.7 ± 13.1 days (range, 24-67 and 41-66 days) for short and long cycles, respectively. Average time from confirmed ovulation to progestagen rise above baseline, maximum progestagen and time to reach maximum progestogen levels were not significantly different between short and long cycles (Table 2). Interestingly, time from treatment to progestagen rise above baseline (days) was very similar (∼7 days) for both short and long cycles. Substantial differences were seen in maximal luteal size between CL (37.4 ± 5.7 mm) and HAF formation (72.5 ± 7.8 mm), although only two HAFs contribute to this measurement.

Discussion
We investigated the efficacy of ovulation induction in anovulatory SWR. We show that a novel ovulation induction protocol, based on natural follicle growth in anovulatory females, results in ovulation efficiently and reliably. Brown et al. (2001) found numerous acyclic, or irregularly cyclic females at multiple institutions by fecal progesterone analysis. It is possible, however, that the females determined to be acyclic in this study experienced follicular growth but were anovulatory. Follicular waves without ovulation have been observed in SWR (Hermes et al., 2006;Roth et al., 2018) and were demonstrated in this study. The terms used to describe animals that fail to ovulate may need clarification: we suggest that anovulation is a form of acyclicity, from the perspective that ovulation is the defining aspect of the cycle and therefore a necessary component. Acyclic animals, determined by progesterone analysis, may grow follicles in a cyclical pattern without ovulation and can be identified by serial ultrasonography. More complete information about an individual designated as acyclic may expand the designation to anovulatory indicating that management options like ARTs might salvage reproductive potential and avoid the presumption of infertility commonly associated with acyclicity.
Exogenous GnRH resulted in luteal formation following all treatments in this study, although two instances resulted in HAF formation rather than ovulation. The injectable form of deslorelin acetate in oil was chosen for its straightforward application (single injection). This deslorelin preparation efficiently induces ovulation in horses, the model species for rhinos (Squires and Simon, 2011;Ferris et al., 2012). GnRH treatment in rhinos is intended to facilitate AI Hermes et al., 2009b;Stoops et al., 2016 Each shape represents an individual and color indicates treatment number. Circle (•) represents #2194, square ( ) represents #2196, triangle ( ) represents #2197 and diamond ( ) represents #2198. Open shapes indicate first treatment, solid the second, gray the third and divided the fourth. Gray and divided diamonds ( , #2198) identify treatments that resulted in HAF formation and were therefore present at 48 hours post-treatment. therefore ovulation efficiency is critical. We observed increased efficiency (81.8%) compared to a previous study (60.5%) with a similar sample size utilizing injectable GnRH in SWR (Hermes et al., 2012). In that study, a timed protocol used a 45-day treatment of altrenogest and induced ovulation 9 days after synthetic progestin withdrawal. Our dose of 4.5 mg GnRH was higher than the 3.0 mg dose utilized by Hermes et al. (2012) and we generated a narrowed window of ovulation between 36 and 48 hours postinjection. Furthermore, our time of observed response was longer than previously reported Hermes et al., 2009b) in which ovulation was reported within 24 hours of GnRH administration. We tailored treatment to the individual animal and minimum follicle size, similar to the equine model (Squires and Simon, 2011;Ferris et al., 2012), rather than treat on an expected timed response, like the cattle model (Lamb et al., 2010). By customizing the treatment to each individual through serial ultrasonography and applying specific follicle growth and size criteria for treatment, rather than a timed approach, we observed rates of ovulation similar to those achieved in commercial equine practices (Finan et al., 2016;Squires and Simon, 2011).
Precise ovulation timing and predictability expand the potential of AI and utility of cryopreserved sperm. We have also demonstrated that, similar to domestic horses, cycle lengths can be quite variable and a timed approach may not be suitable. Serial ultrasonography is beneficial to both domestic and exotic species reproductive management (Adams et al., 1991) and was critical to the efficiency achieved here. We also observed a highly predictable time from injection to rise in fecal progestagen above baseline (∼7 days) regardless of resulting luteal phase designation. With such efficiency and predictability achieved in ovulation induction, next steps include pairing with AI or developing other techniques such as embryo transfer.
Relying on hormone analysis alone to determine ovulation, as opposed to HAF formation, is cautioned against (Roth et al., 2018) and we observed a progestagen rise above baseline from HAF structures that mimicked luteal levels and duration. HAFs are commonly hormonally active and have been documented in all studied rhinoceros species (Roth et al., , 2018Stoops et al., 2004) and in horses (Ginther et al., 2006(Ginther et al., , 2008Cuervo-Arango and Newcombe, 2012). Our observations of HAFs in only one female suggest certain individuals maybe pre-disposed to HAF formation. In horses, similar individual variability has also been observed (Ginther et al., 2006(Ginther et al., , 2008. Interestingly, although the HAFs observed in this study reached extremely large diameters, the duration of the HAF was designated as 'short' on both occasions, indicating the ability to resolve the structures quickly. Confirmation of ovulation, or lack thereof, was important as aged HAFs appeared similar to CL and highlights the need for frequent longitudinal ultrasonography when employing ARTs like AI. We corroborate previous findings that white rhinos display two distinct estrous cycle lengths (Hindle et al., 1992;Radcliffe et al., 1997;Schwarzenberger et al., 1998;Patton et al., 1999;Brown et al., 2001). Length of time progestagens were elevated above baseline was significantly different between cycle designations, as expected. However, visibility of the resulting luteal structure was not significant between cycle designations and did not temporally correlate with progestagen elevation (data not shown). We suggest this may be due to the lower frequency of ultrasound data compared to that of fecal sampling. Fecal samples were collected at least three times per week while ultrasound data were collected one to three times per week with the exception of post-injection exams. The observation of luteal structures beyond progestagen elevation in short cycles, but not long cycles (Table 1), even when HAFs were excluded from analysis (data not shown) is interesting and warrants further investigation.
Long cycles have been purportedly caused by pathologies, such as pyometria, failed pregnancy or endometritis, and have been suggested to be abnormal (Radcliffe et al., 1997;Patton et al., 1999), while other studies indicate that the frequency of occurrence suggests that long cycles are normal (Schwarzenberger et al., 1998;Brown et al., 2001). Here we found that both cycle types can be exhibited within an individual (#2194 and #2195). Others (#2196, #2197 and #2198) displayed a single cycle type following ovulation induction, although the number of trials is too few per individual to conclude that these females do not exhibit both types. Longitudinal ultrasonography identified no uterine pathologies in any females throughout the study, and all cycles were non-conceptive with no potential for insemination, indicating that long cycles are not necessarily caused by pathology or failed pregnancy. Observation of both cycle types following ovulation induction in anovulatory females indicates that exogenous GnRH treatment produces luteal phases similar to spontaneous cycles. Although it is not known what dictates different cycle lengths, it appears both types are normal features of white rhino reproduction. The duration of the CL, and associated progestagen secretion that result in different luteal phase lengths, may be driven by the CL itself or the uterus. The conventional understanding for most mammals is that uterine prostaglandin secretion is down regulated by progestagen secretion from a CL and delays the onset of luteolysis. The functional capacity of the CL may be associated with the vascularity of the structure after ovulation and throughout the luteal phase as angiogenesis is a vital aspect of luteal formation and achievement of function (Fraser and Wulff, 2003;Tamanini and de Ambrogi, 2004). Increased vascularity, established at or after ovulation, may prolong the CL and enable sustained progestagen secretion, resulting in long cycles. Alternatively, because increased luteal blood flow has been shown to immediately precede prostaglandin secretion and luteolysis in cows (Miyamoto et al., 2005), perhaps highly vascularized rhino CLs may be more responsive to uterine prostaglandin secretion and result in short cycles. Future ultrasonographic studies utilizing color flow Doppler should focus on the extent of CL vascularity throughout the luteal phase to identify any relationships between blood flow and cycle length. Additionally, measurement of prostaglandin levels may provide more insight into the luteolytic process.
The criteria of follicle size (∼35 mm) utilized in this study is larger than the previously documented pre-ovulatory follicle size (30 mm) for SWRs (Roth, 2006). Despite this, we observed a reliable response to GnRH. It is possible, though, that a range of effective sizes exists, as in horses (Ferris et al., 2012), and these limits have not been explored for rhinoceros species. We also propose that response to GnRH may vary in different housing situations, particularly if a male is present. Additionally, females that are determined to be ovulatory may have an altered response to GnRH as well, and the criteria for treatment may need to be amended. Nonetheless, since a large proportion of captive SWR females have been determined to be anovulatory via progesterone evaluation Hermes et al., 2007), the next steps may be to ultrasonographically assess ovaries, monitor follicle growth and consider GnRH for induction of ovulation to facilitate breeding or AI. Efficient use of ARTs would not only help overcome reproductive issues, like those common in captive SWR females, but also could improve animal welfare by obviating the need to transport animals between locations for breeding based on genetic matches. They may also circumvent behavioral incompatibilities experienced by an intended breeding pair. Furthermore, semen cryopreservation combined with AI or in vitro fertilization will maintain or improve the genetic diversity in a captive population by enabling breeding of genetic matches simultaneously at multiple locations. Information such as the data presented here will provide a route for more efficient use of other ARTs like AI and ET, which could ultimately modify the way captive populations are managed.
to the Northern White Rhino Initiative and made this research possible. The authors also thank the tireless efforts of the rhino training team (Jonnie Capiro, Jill Van Kempen, Marco Zeno, Weston Popichak and Todd Schwenk) that enabled completion of this work. We appreciate the San Diego Zoo Safari Park veterinary team's support of our work and the maintenance of the health of the study rhinos. We also thank Dr Chris Tubbs and Dr Tom Jensen for insightful discussion and manuscript preparation.

Funding
This work was supported by San Diego Zoo, project number 55352.