Correlating Behavior with Non-Invasive Physiological Measures to Evaluate Mating Strategies in Belugas

With conservation and ethical concerns facing cetaceans, minimally invasive research on reproduction is important for population management. Belugas (Delphinapterus leucas) are endangered in parts of their range, yet little is known about their breeding behavior and much of the existing research depends on postmortem sampling. To date, there have been no directed studies of social interactions between belugas during the breeding season, and few studies have attempted to correlate reproductive physiology with behavior in any species of cetacean. Improved understanding of reproductive strategies in this species would facilitate management. This work describes the development of minimally invasive methods for determining sex, maturity, and reproductive cycle stage in belugas, and the utilization of these methods to assess relationships between reproductive physiology and behavior in a social group of belugas in an aquarium. The results of this work are interpreted in the context of the current understanding of beluga reproductive physiology and ecology. Chapter 1 describes the physiological validation of blow (exhale) sampling for measuring testosterone and progesterone in belugas. Concentrations of both progesterone and testosterone in blow are correlated with circulating concentrations and reflect variation in the reproductive status of individuals. Chapter 2 presents the development of DNA isolation from beluga blow samples and determines the relationship between various sample characteristics and DNA yield and performance in polymerase chain reactions (PCR). Although yield and quality varied greatly among samples, single exhale samples from wild and aquarium belugas enabled PCR amplification of genes used in sex determination or population genetics. Chapter 3 describes the seasonal variation in testicular volume and testosterone in male belugas studied longitudinally. This work revealed a seasonal increase in testes size in belugas of approximately 50%, filling a significant gap in knowledge for wild belugas. Chapter 4 utilizes methods developed and validated in Chapters 1 and 3 to evaluate social behavior in a group of aquarium belugas in the context of reproductive physiology. Reproductive seasonality and the occurrence of reproductive events were detected using non-invasive techniques and used to contextualize patterns of association and the frequency of social behaviors of interest, including courtship. Appendix 1 presents a general review of the literature regarding beluga mating strategies and reproductive biology; specific literature reviews are included with each chapter. Appendix 2 provides a synthesis of the dissertation and discusses the implications of these findings on the current understanding of beluga mating strategies, and by extension, the management and conservation of this species. Appendix 3 presents laboratory protocols utilized in this dissertation, and provides supplementary data that was not included in Chapter 1.


LIST OF TABLES
. "Receptiveness score" scheme used to assign receptivity to genital presents.
Each occurrence received one score from each of the three categories (swim speed, orientation, and contact), and the total score is the sum of these three component  Progesterone concentrations in 64 matching blood and blow samples from 11 females were also significantly positively correlated (F 1, 62 = 94.0, p < 0.0001). Testosterone concentrations in blow varied seasonally in males monitored longitudinally, with the peak occurring during the breeding season (February -April). Testosterone concentrations (mean ± SD) in blow samples collected from adult males during the breeding season (February -April, 136.95 ± 33.8 pg/ml) were significantly higher than in those collected outside of the breeding season (May -January, 99.4 ± 39.5 pg/ml). Both adult male groups had blow testosterone concentrations that were higher than that of a juvenile male (<8 years) (59.4 ± 6.5 pg/ml) or female belugas (54.1 ± 25.7 pg/ml). Although there is a high degree of overlap between out of season adult males, juvenile males, and females at low testosterone concentrations, high testosterone concentrations in blow can be used to identify adult males. Matching blood and blow samples collected from wild belugas in Alaska (8 males and 2 females) were not positively correlated, but only males had blow testosterone

Introduction
Given the fundamental relationship between reproduction and endocrinology, reproductive steroid hormone determinations are frequently used to assess the reproductive status of individual animals, yielding information such as the sex, maturity status, and reproductive cycle stage. At the population level, this information is necessary to assess demographics and viability and is crucial to the development and long term monitoring of conservation plans (Labrada-Martagón et al. 2014;Kersey and Dehnhard 2014). Cetaceans (whales, dolphins, and porpoises) particularly benefit from hormone monitoring due to the management concerns facing many populations.
The utility of reproductive steroid determination in cetaceans has been established through the use of blood samples. The measurement of progesterone in female blood samples can be used to diagnose pregnancy (Stewart 1994;Kellar et al. 2013a) and detect the luteal phase of the estrous cycle (Sawyer-Steffan et al. 1983).
Testosterone determination in male blood samples has been used to assess sexual maturity (Daoquan et al. 2006;Desportes et al. 1994; and identify the breeding season in seasonally breeding species Hao et al. 2007). While collecting samples for this purpose is complicated by the aquatic environment, advances in minimally invasive sampling methodologies and alternative matrices are replacing the need for blood sampling and enabling hormone determination research in wild, free-swimming cetaceans (reviewed in Amaral 2010).
The application of hormone determination in alternative matrices to cetacean research has allowed the assessment of sex, maturity status and reproductive seasonality, as well as population-level assessments of pregnancy rate that can inform management and conservation efforts Vu et al. 2015;Kellar et al. 2013b). To date, the most commonly used matrices for hormone determination in wild cetaceans are blubber samples obtained via remote biopsy sampling (Trego et al. 2013;Perez et al. 2010) and fecal samples collected from the water's surface (Rolland et al. 2005). Despite the value of these methodologies, they may not be appropriate for all species or all populations. Biopsy sampling may not be ideal if repeated sampling is required, or if conservation concerns make even small risks to the animal's health undesirable. These risks may limit the sampling population available if investigators are required to avoid sampling mothers with young calves or the calves themselves (Kellar et al. 2014). Defecation may be infrequently observed, and collecting sufficient feces from smaller species of cetaceans is more difficult, especially if the feces rapidly disperse upon excretion, precluding collection (Green et al. 2007). Additionally, assigning fecal samples to particular individuals can be difficult in highly social species. These and other limitations in existing methodologies have led to a recent effort to develop "blow" (exhale) sampling as a source of hormones for analysis in cetaceans.
Blow, which is also referred to as respiratory vapor or exhaled breath condensate, can be collected without contacting the animal when the whale surfaces to breathe. Blow samples contain steroid and thyroid hormones at detectable levels, as demonstrated in a variety of cetaceans (Hogg et al. 2005;Hogg et al. 2009;Thompson et al. 2014;Hunt et al. 2014). Hunt et al. (2014) and Thompson et al. (2014) have shown that relatively inexpensive enzyme immunoassays can be validated to measure steroid hormones and thyroid hormones in cetacean blow, improving the accessibility of this technique. Additionally, blow sampling can be used to gather genetic information on the individual cetacean (Frère et al. 2010) or the microorganisms associated with the respiratory tract (Acevedo-Whitehouse et al. 2010). Due to the minimally invasive nature of this sampling, coupled with its ability to provide a wide variety of information, blow sampling has the potential to be a very useful tool in monitoring cetacean populations and assessing the health and reproductive status of individuals (Hunt et al. 2013).
In order for blow sampling to reach its potential as a diagnostic tool for use in population assessments, the measurement of the various hormones in this matrix must be shown to be physiologically relevant. Testosterone (3 species) and progesterone (5 species) have been detected in blow samples from both mysticete and odontocete cetaceans (Hogg et al. 2005;Hogg et al. 2009;Hunt et al. 2014;Dunstan et al. 2012;Tizzi et al. 2010). However, without knowledge of how blow hormone concentrations relate to blood concentrations or the reproductive status of an individual, the ability to interpret blow hormone concentrations has been limited to assessing presence or absence. The use of blow as a diagnostic tool would be greatly enhanced if the relationship between hormone concentrations in blow and sex, maturity status, pregnancy status, or breeding season was known.
Belugas are an ideal study species for advancing the use of blow sampling as a diagnostic tool. Their reproductive biology is well understood, due to extensive postmortem studies from subsistence harvests (e.g.  as well as in depth longitudinal studies of live animals conducted in aquaria (e.g. ). Progesterone in blood or progesterone metabolites in urine can be used to detect pregnancy and luteal activity Stewart 1994).
Testosterone levels in blood can be used to detect sexual maturity and reproductive seasonality in males, with peak secretion occurring between January and April . Belugas are widely held under professional managed care, making it possible to collect sufficient blow samples from known individuals to develop hormone assays and to sample the same individuals longitudinally to assess changes in reproductive status over time. Although reproductive steroids have not been measured in beluga blow samples, blow sampling has already been shown to be an effective way to assess cortisol secretion in this species (Thompson et al. 2014). There is also potential to apply the methodology to wild belugas (Thompson et al. 2014), as well as a need for additional information to improve management, especially in endangered populations (National Marine Fisheries Service 2015).
The aim of this study was to determine if testosterone and progesterone concentrations in beluga blow samples are biologically relevant, laying the foundation for the development of blow sampling as a diagnostic tool in this species.
Commercially available enzyme immunoassays for testosterone and progesterone were validated, and a physiological validation was performed using samples collected from aquarium belugas. Specifically, testosterone and progesterone measurements in matched blood and blow samples were compared, and the relationship between the concentrations of these hormones in blow and the sex, age, and reproductive status of these belugas was explored. Further validation was performed by measuring testosterone in the blow samples of wild belugas that are temporarily restrained for a health assessment project in Bristol Bay, Alaska to determine the applicability of this methodology in the field.

Zoological facilities
All blow and blood samples were collected from belugas with the voluntary cooperation of the animals via trained behaviors. Blow samples were collected from a total of 20 belugas (8 male, 12 female) from four different zoological facilities, ranging in age from 3-33 years ( Table 1). Males less than 8 years old and females less than 6 years old were considered juveniles . All belugas were housed in mixed sex groups with at least 2 males. Three of the four adult males sampled were proven sires. To assess seasonality of testosterone secretion, two males were sampled in all 12 calendar months, and all males were sampled between February and April, representing much of the period of peak testosterone secretion in belugas (January -April, . This project was approved by the Institutional Animal Care and Use Committees of Mystic Aquarium (Project #12001) and the University of Rhode Island (Project #AN12-02-016).

Ultrasonography and Characterization of Reproductive Status in Adult Females
To monitor for pregnancy, ultrasound exams were conducted on female belugas approximately twice per month at irregular intervals that varied by individual.
Ultrasound exams were performed with the voluntary cooperation of the animal while the animal lied unrestrained in lateral recumbency at the water's surface. For the purposes of this study, a female later observed with a viable fetus was considered pregnant starting on the date that fluid was first visible in the uterus. The presence of a corpus luteum (CL) was detected via ovarian ultrasound or inferred by high progesterone levels in blood (>3 ng/ml) . In the absence of ultrasound or blood progesterone data, a CL was presumed to be present 30 days prior to a pregnancy diagnosis and to persist for 14 days from detection in non-conceptive cycles .

Blow sample collection
Blow samples were collected onto a nylon mesh (110 µm, Elko Filtering Co., Miami, FL) stretched over a petri dish (100 mm diameter) and secured with a rubber band (Thompson et al. 2014). The nylon mesh and rubber bands were cleaned prior to use by soaking in 70% ethanol for 15 minutes, rinsed with Nanopure water, and air dried. For sample collection, the whales were trained to lift their head so that the blowhole was above the water's surface. The whale would then exhale once to clear any excess water from the blowhole surface. Then, 2-8 successive exhales were collected onto the mesh. The nylon was used to soak up any fluid that may have passed through onto the petri dish, and was immediately put into a 15 ml conical tube pre-loaded with half of a plastic syringe stopper in the bottom. The syringe stopper would allow for the later centrifugation of the nylon to retrieve the fluid blow sample.
This tube was then frozen within 10 minutes at -20˚ C for up to 2 weeks until processing, or at -80˚ C until shipment on dry ice to the laboratory.
To retrieve the respiratory vapor, the tubes were thawed for 15 minutes and then centrifuged at 10˚ C for 30 minutes at 2600 x g. The volume of each sample was recorded and the fluid was frozen in 1.5 ml cryovials at -80˚ C until analysis. When possible, blood samples were collected into serum separator or sodium heparinized vacutainer tubes within one hour of blow sampling from a ventral fluke vein with the voluntary cooperation of the animal as a part of routine veterinary monitoring. Matching blood and blow samples were collected in the morning hours, typically between 0900 and 1000 hours. One ml of serum or sodium heparin plasma was obtained through centrifugation (2000 x g for 10 minutes at 10˚ C) and stored at -80˚ C.

Bristol Bay, Alaska
Blow samples were collected from 10 wild belugas (8 males and 2 females) in Bristol Bay, Alaska between August 25 and September 2, 2014 while they were being temporarily restrained for health assessment and satellite tagging (described in . Samples were collected under National Marine Fisheries Service Marine Mammal Research Permit #14245. Animals were sexed and all were judged to be mature adults based on growth curves for this stock ). Four exhales were collected per sample as described above, but due to field conditions it was not always possible to ensure that the blowhole was free of all water prior to collecting an exhale as it was with trained belugas. Once sealed inside a 15 ml conical tube, the samples were stored in a cooler on ice packs for 2-6 hours prior to centrifugation. The volume of each sample was measured and the fluid was transferred to 1.5 ml cryovials tubes for storage. The samples were frozen in liquid nitrogen within 8 hours of collection. Blood samples were also collected from these animals from a dorsal or ventral fluke vein, centrifuged to obtain sodium heparin plasma, and stored in liquid nitrogen for transport.

Testosterone Enzyme Immunoassay Validation
A commercially available testosterone enzyme immunoassay (Cayman Chemical, Ann Arbor, MI, Item #582701) was validated for use with beluga blow, serum, and plasma. This kit has 100% reactivity with testosterone. Cross-reactivities reported by the manufacturer were 140% for 19-nortestosterone, 27% for 5αdihydrotestosterone, 18.9% for 5β-dihydrotestosterone, 4.7% for methyl testosterone, 3.7% for androstenedione, and 2.2% for 11-keto testosterone; all other crossreactivities were below 1%. All blood samples were extracted with diethyl ether (Sigma-Aldrich, St. Louis, MO, Catalog #346136) according to the EIA kit manufacturer's instructions; blow samples were assayed without a sample preparation step. Parallelism to the standard curve was tested for blow, serum, and plasma using pooled male samples serially diluted from neat to 1:16 for blow and 1:10 to 1:100 for serum and plasma. Accuracy was tested for blow, serum, and plasma using pooled male samples spiked with an equal volume of known amounts of testosterone standard (125, 62.5, 31.3, 15.6, and 7.8 pg/ml). Accuracy was performed at 1:2 dilution for blow and 1:20 dilution for serum and plasma. Extraction efficiency in blood samples was tested by adding a known amount of testosterone (0, 250, 500, 1000 and 2000 pg) to subsamples of unextracted serum or plasma sample pools and performing the extraction protocol. Extraction efficiency was calculated as the observed amount of hormone quantified in the assay divided by the total amount of hormone expected (native testosterone present in unspiked sample plus amount of spiked testosterone) multiplied by 100. To test for assay interference and recovery from the collection material, nylon mesh was spiked with assay buffer (negative control, n = 10) or known concentrations of testosterone (250, 125, 62.5, 31.3, 15.6, and 7.8 pg/ml) in three replicates. The spiked nylon was treated identically to a biological sample (placed in 15 ml conical tube, frozen >24 hours, thawed, centrifuged to recover spiked fluid, and refrozen until assay). Testosterone recovery (observed divided by expected multiplied by 100) was measured.

Progesterone Enzyme Immunoassay Validations
A commercially available progesterone enzyme immunoassay (Cayman Chemical, Ann Arbor, MI, Item #582601) was validated for use with beluga blow, serum and plasma. This kit has 100% reactivity with progesterone. Cross-reactivities reported by the manufacturer were 14.0% for pregnenolone, 7.2% for 17β-estradiol, 6.7% for 5β-pregnan-3α-ol-20-one, and 3.6% for 17α-hydroxyprogesterone; all other cross-reactivities were below 1%. All blood samples were extracted with dichloromethane (Sigma-Aldrich, St. Louis, MO, Catalog #270997) according to the manufacturer's instructions. Parallelism to the standard curve was tested for serum and plasma using pooled female samples diluted 1:10, 1:20, 1:40, 1:60, and 1:80. Accuracy was performed for serum and plasma at 1:40 dilution using pooled female samples spiked with an equal volume of known amounts of progesterone standard (500, 250, 125, 62.5, and 31.25 pg/ml). Extraction efficiency in blood samples was tested by adding a known amount of progesterone (4000, 2000, 1000, and 500 pg) to subsamples of unextracted serum or plasma sample pools and performing the extraction protocol; extraction efficiency was calculated as previously described for testosterone.
Untreated blow samples displayed parallelism to the standard curve, but failed the accuracy test, indicating matrix interference. Four extraction protocols for the extraction of progesterone from various matrices were tested (dichloromethane liquidliquid microextraction and 3 variations of a solid phase extraction protocol) and all failed either parallelism or accuracy tests (See Appendix 3 for detailed methods). A diethyl ether liquid-liquid extraction was considered optimal and used for all samples.
Samples (55 or 60 µl) were placed in glass test tubes and 0.5 ml of diethyl ether was added. The samples were vortexed for 2 minutes, and then the aqueous layer was frozen in an ultralow freezer. The ether layer was poured off and the extraction procedure was repeated. The two ether layers were combined, dried under compressed air, and reconstituted in 110 or 120 µl of assay buffer (for a final dilution of 1:2). Extraction efficiency in blow samples was tested by adding a known amount of progesterone standard (4000, 2000, 1000, and 500 pg) to subsamples of unextracted blow sample pools and performing the extraction protocol. Paralellism to the standard curve was tested for blow samples using pooled female samples serially diluted from neat to 1:8. Accuracy was performed for blow samples at 1:2 dilution using pooled female samples spiked with an equal volume of known amounts of progesterone standard (250 ,125, 62.5, 31.25, and 15.625 pg/ml). To test for recovery from the collection material, nylon mesh was spiked with assay buffer (negative control, n = 4) or known amounts of progesterone (1000, 500, 250, and 125 pg) in two replicates.
The spiked nylon was treated identically to a biological sample (placed in 15 ml conical tube, frozen >24 hours, thawed, centrifuged to recover spiked fluid, and refrozen until assay). Progesterone recovery (observed divided by expected multiplied by 100) was measured.

Assay of biological samples
Blow samples were centrifuged at 8,000 x g for 10 minutes to remove particulates and assayed at a 1:2 dilution for testosterone. Samples assayed for progesterone were centrifuged prior to extraction; extracted blow samples were assayed at 1:2 dilution. Extracted blood samples were assayed primarily at 1:20, but ranged between 1:2 and 1:80 depending on the expected concentration of testosterone or progesterone . All samples were assayed in duplicate and the means were used in calculations. Individual samples with a %B/B o between 20 and 80% and a coefficient of variation (CV) < 20% were accepted. Samples with CV >20% were re-assayed, and blood samples outside of the range of the kit were reassayed at a different dilution. Two female blow samples assayed for testosterone had CV above the 20% cutoff (24.2 and 23.1), but were kept in the analysis due to lack of volume to re-run the assay for these samples. Blood samples with low progesterone (<300 pg/ml) were prone to higher CV; if re-assaying these samples did not result in a CV below the 20% threshold, the concentration measurement with the lowest CV was used for analysis (n=8). Two standard controls were run in each assay (testosterone: 100 and 25 pg/ml, n = 12; progesterone: 200 and 50 pg/ml, n = 14). Inter-assay variation was calculated by determining the CV for the two standard controls on each plate. Intra-assay variation was calculated by averaging the CV for all of the samples with 20-80% binding on each plate.

Sample handling and storage
To test the effect of long term storage on testosterone concentration in blow, a pool of male blow samples was constructed and aliquoted into separate cryovial tubes and stored at -80˚ C. The pool was assayed 2 days after construction, then again 3, 17, 20, and 21 months later.
To determine if temporary chilled storage used during field work influenced testosterone concentration in blow, a separate pool of male blow samples was constructed and aliquoted in duplicate into separate cryovial tubes. Two samples were frozen immediately and replicate tubes were stored in a cooler on ice packs for 2, 4, 6, 8, and 10 hours before freezing at -80 ˚C to simulate field conditions. Both replicates from the 0 hour time point from this experiment had an unacceptable CV (>30%), so another experiment was conducted to examine the time between 0 and 2 hours. For this experiment, a third pool of male blow samples was aliquoted in duplicate into separate cryovial tubes; 2 samples were frozen at -80 ˚C immediately after construction, the others were stored in a cooler on ice packs for 30, 60, 90, and 120 minutes. The cooler remained at 4-7˚C for the duration of these experiments.
To test the effect of the freeze-thaw-freeze cycle that blow samples in this study were subjected to on testosterone concentration in blow, a pool of male blow samples was constructed and aliquoted in cryovial tubes. One sample was frozen at -80 ˚C immediately (no subsequent thaw-freeze), while the other tubes were stored in a -20˚C freezer for 1, 2, or 4 weeks. After the appropriate duration, they were thawed and refrozen at -80 ˚C until analysis.
To assess within sample variation, samples were collected on two separate collection devices; the first contained the blow from the first, third, and fifth exhale, while the second contained the blow from the second, fourth, and sixth exhale. A total of five samples collected in this manner from two different males were assayed for testosterone. To explore the influence that centrifugation might have on testosterone measurements, 4 samples from a male beluga were divided into two subsamples. One subsample was centrifuged prior to being assayed as described above, while the other was mixed thoroughly prior to being assayed.

Statistics
All statistical analyses were performed in R (R Core Team, 2015). Linear regression models (ANCOVA) were used to test for parallelism between serially diluted sample pools and the standard curve. Accuracy was tested by linear regression and considered acceptable if the slope of the line was not significantly different from 1. Linear regression was used to test for the effect of sample volume or sample handling regime on testosterone concentration. Assessing the correlation between hormone concentrations measured in blow and matching blood samples in aquarium belugas using linear regression required hormone concentrations to be log transformed to meet normality assumptions; Bristol Bay beluga blood and blow testosterone concentrations were correlated separately and were not transformed. For aquarium belugas with more than 5 matched observations, each individual's correlation was examined separately to account for the uneven number of samples available from each beluga.
To describe seasonality of testosterone concentrations in blow collected from adult males, an additive modeling approach was used to identify the polynomial regression model that best described the data using a centered time variable (month).
Random intercept and random slope terms were tested to account for observations being clustered by individual using the {lme4} package in R (Bates et al. 2015).
Model fits were compared using ANOVA, AIC, and Log-Likelihood.
To test for differences in log transformed blow testosterone concentrations in belugas with varying sex and reproductive status, mixed effects regression models were constructed using the predictors of sex, status (adult or juvenile), and season (Feb -Apr or May -Jan). To account for clustered observations within individual, a random intercept term was incorporated into the model. Significant interaction terms were evaluated graphically by constructing effects plots using the {effects} package in R (Fox 2003). The same method was used for log transformed progesterone concentrations in blow, using a predictor combining females with a corpus luteum or a pregnancy as well as a random intercept term. Males were left out of the model to avoid multicollinearity between sex and luteal activity and pregnancy, and thus sex differences will not be interpreted. Means are presented ± 1 SD and significance levels were set at p < 0.05.

Results
Males had larger blow sample volumes per exhale than females; blow sample characteristics are presented in Table 2. to 109%, with a mean of 96.5%. The extraction efficiency for plasma ranged from 69.5% to 97.9%, with a mean of 79.2%.
The average lower limit of detection (80% B/B o ) was 10.7 pg/ml. Intra-assay variation was 8.37%; inter-assay variation was 6.3% for the 100 pg/ml control and 10.3% for the 25 pg/ml control.

Progesterone Assay Validation
The binding of a serially diluted pool of ether extracted female blow samples was parallel to the standard curve (F 1, 6 = 0.05, p = 0.83) (Fig. 4). Progesterone was detectable in the sample pool for dilutions from neat to 1:8. The recovery of progesterone from spiked blow sample pools was 102 ± 17% (y = 1.06x -5.62, R 2 = .97). The slope of the regression line was not significantly different from 1 (95% CI [0.74, 1.37]), demonstrating good accuracy (Fig. 5). The extraction efficiency for blow ranged from 100% to 117%, with a mean of 110% (Fig 6). For the results of extraction methods that were not effective, see Appendix 3. Figure 4. Parallelism between a serially diluted female blow sample pool (triangles) and the progesterone standard curve (diamonds). Blow sample pool dilutions range from neat (1:1) to 1:8.   the study were serum samples.
The average lower limit of detection (80% B/B o ) was 32.6 pg/ml. Intra-assay variation was 10.9%; inter-assay variation was 6.6% for the 200 pg/ml control and 11.3% for the 50 pg/ml control.

Sample Handling Experiments
Storage of a pooled blow sample at -80˚C for up to 21 months did not affect testosterone concentration beyond the inter-assay variation (CV = 7.6, F 1,3 = 1.82, p = 0.27). The initial experiment testing the effects of storage in a cooler on testosterone concentration showed a significant effect of storage time on concentration (F 1, 10 = 7.87, p = 0.02). The CV for the 12 measurements of the same pool was 17%.
However, both replicates of the 0 hours time point had high CV (34 and 36%). After dropping this time point, storage in a cooler for 2-10 hours before freezing did not significantly affect testosterone concentration (F 1, 8 = 0.51, p = 0.50); the CV for the 10 measurements of the same pool was 5.3%. A second experiment evaluated the effect of storage time on testosterone concentration from 0-2 hours more closely.
Storage on ice packs for up to 2 hours did not affect testosterone concentration (F 1, 8 = 2.14, p = 0.18); the CV for the 10 measurements was 6.5%. Performing a thaw-freeze cycle after 1, 2, or 4 weeks at -20˚C did not affect testosterone concentration (F1, 2 = 1.37, p = 0.36), with the 8 measurements of the same pool having a CV of 5.8%.
For samples that were collected onto two separate collection devices, the average CV for the two samples for each divided sample was 9.0%. For samples that were divided and either centrifuged or mixed prior to assay, the average CV for the two samples for each divided sample was 6.6%. There was an insignificant negative relationship between sample volume and testosterone concentration in male samples (y = -0.04x + 121, F 1, 104 = 2.69, p = 0.1).

Assay of individual samples: Testosterone
Testosterone assay results are summarized in Table 3. Blow testosterone concentration was greater than the mean negative control concentration (45.0 pg/ml) for 90% (122/136) of the biological samples. Of those that were less than the mean negative control concentration, 7 were from females, two were from juvenile males, and 5 were from adult males. Two biological samples (both female) had testosterone concentrations lower than the lowest negative control sample (27.7 pg/ml).

Relationship between blood and blow
There was a significant positive correlation between matching log transformed blow and blood testosterone concentrations from aquarium belugas (F 1, 38 = 41.7, p < 0.0001) (Fig. 8). Individual correlations for the 3 belugas sampled more than 5 times are shown in Table 4. After correcting for the mean negative control value and removing two observations with negative concentrations, blow testosterone concentration was 4.9 ± 4.7% of the matching blood concentration.

Biological Relevance of Testosterone Concentrations in Blow
Seasonality of testosterone secretion was found in adult male blow samples.
Variation in log transformed testosterone concentrations by month was best described by a quartic polynomial function with a random intercept term (y = -.0843x + 0.0096x 2 + 0.0033x 3 -0.0004x 4 + 1.9615) (Fig 9). A model selection summary is shown in Table 5. A comparison between samples from belugas with varying sex and reproductive status showed a significant effect of individual (intercept, p < 0.001), as well as significant interactions between sex and status (adult or juvenile) (p < 0.01) and between sex and season (Feb -Apr or May -Jan) (p < 0.001). Effects plots demonstrated that adult males had higher testosterone in blow than juvenile males or females, and that adult males sampled during breeding season had higher testosterone concentrations in blow than all other groups (Figs. 10 and 11). Variation in blow testosterone concentrations by sex and reproductive status is shown in Fig. 12.

Application to wild belugas
There was no significant relationship between blood and blow testosterone for belugas from Bristol Bay (F 1, 8 = 0.79, p = 0.40) (Fig. 13). Blow testosterone concentrations were 3.3 ± 3.1% of those in the matching blood sample. Three male samples exceeded 100 pg/ml; all six animals with blow testosterone concentrations >65 pg/ml were male.

Assay of individual samples: Progesterone
Progesterone results are summarized in Table 6. All blow samples had concentrations higher than the mean negative control samples; two samples had concentrations lower than the highest negative control sample (151.7 pg/ml). Three belugas became pregnant during the study period. For one pregnant female (DL12), samples prior to pregnancy, during pregnancy, and after parturition were available.
Samples from DL11 were available before and during pregnancy, while samples collected during pregnancy and after parturition were available from DL16. Luteal activity was detected via ultrasound in 2 females (DL14 and DL16). For DL14, the CL persisted approximately 21 days. In DL16, the CL was visualized using ultrasound prior to pregnancy diagnosis. Two luteal phases were detected in DL12; a non-conceptive luteal phase inferred from a blood sample with a high progesterone concentration, as well as an inferred conceptive luteal phase approximately 10 weeks later, prior to the confirmation of pregnancy via ultrasound.

Relationship between Blood and Blow
There was a significant positive correlation between matching log transformed blow and blood progesterone concentrations from aquarium belugas (F 1, 62 = 94.0, p < 0.0001) (Fig. 14). Individual correlations for the 4 belugas sampled are shown in Table 7. After correcting for the mean negative control value, blow progesterone concentration was 3.3 ± 1.4% of the matching blood concentration for pregnant females and 44.3 ± 31.3% of the matching blood concentration for non-pregnant females.

Biological Relevance of Progesterone Concentrations in Blow
Pregnant belugas and belugas with luteal activity had higher progesterone concentrations in blow than females without luteal activity or ongoing pregnancies (p < 0.0001) (Fig. 15). A blow concentration >330 pg/ml was a valuable diagnostic threshold: 6% of the blow samples collected from non-pregnant females and 80% of the blow samples collected from pregnant females or females in the luteal phase had progesterone concentrations that exceeded 330 pg/ml. Only pregnant belugas had   progesterone concentrations in blow >420 pg/ml. All three belugas that were pregnant during the study demonstrated temporal variation that was associated with changes in pregnancy status (Fig. 16). Although samples collected while a CL was active were rare (4 from DL12, 2 from DL16, and 6 from DL14), blow progesterone concentrations were higher than in females without luteal activity, and longitudinal variation in blow progesterone was observed with changes in luteal activity in DL14 ( Fig 17).

Discussion
This study has validated commercially available enzyme immunoassays for testosterone and progesterone in beluga blow samples and has demonstrated that the concentrations of these hormones in beluga blow samples are biologically relevant.
Although both progesterone and testosterone have been detected in blow samples collected from other cetacean species, the ability to interpret hormone concentrations in blow samples has been limited due to a lack of this physiological validation. In this study, collecting relatively undiluted blow samples from belugas of known reproductive states allowed for comparisons between samples, establishing the value of progesterone and testosterone determination in beluga blow.

Comparison with Blood Samples
Both testosterone and progesterone concentrations in blow positively correlated with those in blood, demonstrating that blow sample analysis can serve as an indicator of the relative activity of these hormones in circulation. Blow testosterone concentrations reflected the expected variation in testosterone concentrations in blood due to reproductive status and seasonality that were observed both in this study and in previous studies on belugas . Similarly, high progesterone levels in blood associated with pregnancy or luteal activity in belugas in this study and others Stewart 1994) were also detected in blow samples. Most individuals sampled repetitively demonstrated positive correlations between the two matrices; those that did not either lacked high (DL20) or low (DL3) hormone concentration observations to anchor the regression lines. Thus, the value of blow sampling is in the ability to make important distinctions between reproductive states (e.g. pregnant vs. non-pregnant) that are associated with relatively large differences in hormone concentrations, as opposed to fine-scale changes in hormone secretion.
The relative amount of testosterone and progesterone in blow compared to matching blood samples (approximately 3-5%, except for progesterone in nonpregnant females) is similar to the relative amount of steroid hormones in human saliva when compared to matching blood measurements (10%) (Gröschl 2008). As steroids likely enter saliva via passive diffusion, this similarity in relative concentration supports the hypothesis that steroids also enter the fluid lining the respiratory tract in cetaceans via passive diffusion (Hogg et al. 2009). The relatively high amount of progesterone in blow samples collected from non-pregnant females is unlikely to be due to a change in the mechanics of how the hormones enter blow samples in belugas of different reproductive states. Instead, the relatively high concentrations are more likely due to a matrix effect that artificially inflates progesterone concentrations in all biological samples and is amplified in samples collected from belugas with low blood progesterone concentrations.
Testosterone and progesterone levels in blow are much lower than those reported in the blubber or feces of other cetaceans. Unlike other matrices, there is little opportunity for the hormone to accumulate in the fluid lining the respiratory tract over time given the frequency with which belugas breathe (4-6 times per minute at rest). Fluid, and the hormones contained within it, is continually ejected with each exhale, while several hours may pass between subsequent defecations. All of the belugas sampled in this study were breathing at baseline rates; it is possible that a long breath hold (as during a dive) could result in higher hormone concentrations in blow as more time is allowed for diffusion to occur. The comparison of samples consisting of the 1 st , 3 rd , and 5 th or the 2 nd , 4 th , and 6 th exhales suggest that the hormone concentrations vary little from breath to breath when there is not a long breath hold in between exhales. Any variation would depend on the relative rates of respiratory fluid production and steroid diffusion into this fluid. Further experiments on belugas in zoological facilities can be conducted to determine the effect of breath hold duration on hormone concentration in blow for improved application to free-swimming wild belugas. Given the low testosterone concentrations found in both immature males and females in this study and others , as well as the low blow testosterone to blood testosterone ratio, this method is currently insufficiently sensitive to reliably differentiate juvenile males from females in this species. Many of the female and juvenile male blow samples contained very little testosterone, as the concentrations were within the range of the negative control samples. However, the three highest testosterone concentrations found in female samples were the only samples collected from pregnant females. In right whales, pregnant females have higher testosterone in fecal samples than non-pregnant females (Rolland et al. 2005). Testosterone has also been detected in presumed pregnant humpback whale blow samples (Hogg et al. 2009). However, the two female belugas sampled in Bristol Bay for this study were presumed pregnant based on blood progesterone values (data not shown), yet their blow testosterone concentrations were in the range of the negative controls. Because approximately 1/3 of female belugas are pregnant in wild populations , it will be important to sample additional pregnant females to improve the diagnostic capability of this technique.

Biological Relevance of Progesterone Determination in Blow
Progesterone measurements in blow reflected reproductive cycle stage in female belugas, with detectable increases in progesterone with the onset of pregnancy or the presence of a corpus luteum in individual females monitored longitudinally.
Although an increase in progesterone would be expected with these reproductive events, the relative increase in progesterone in blow with pregnancy state did not match those found in blood samples from this study or others (Stewart 1994;. While pregnancy resulted in more than a 30-fold increase in blood progesterone concentration, progesterone in blow only increases by a factor of 1.7, perhaps due to the inability of the steroid to accumulate in the fluid lining the respiratory tract. The lack of proportionality reduces the value of any one sample, especially when compared to other matrices, where distinctions between states are clearer (Rolland et al. 2005;Kellar et al. 2013a). Despite some ambiguity, most samples would be informative, as there was a diagnostically useful threshold (>330 pg/ml) that indicated pregnancy with reasonable certainty. Blow sampling may also be used to detect ovulation, although repeated sampling would be necessary to discriminate between a non-conceptive cycle and a pregnancy. The time of year that the samples are collected would further resolve this uncertainty, as estrous cycles are rare from July through December . Additional sampling is required to fully develop the diagnostic value of this method.
Progesterone concentrations in blow were not valuable in determining the sex of the beluga in the absence of pregnancy or luteal activity. This was to be expected based on the available data on progesterone in male cetaceans; progesterone concentrations in beluga serum , bowhead blubber (Kellar et al. 2013a) and right whale feces (Rolland et al. 2005) have been similar for males and non-pregnant females. Given the low concentrations of progesterone in blood for non-pregnant adults and juvenile females, this method is also not sensitive enough to detect maturity in females unless there is luteal activity or an ongoing pregnancy. This was also the case for progesterone measurements in the blubber of several other odontocete species (Trego et al. 2013).

Other factors influencing hormone concentrations in blow
In addition to reproductive status, other physiological factors may influence testosterone concentration in blow. In other species, including bottlenose dolphins in aquaria, testosterone secretion is affected by diurnal rhythm, with highest concentrations occurring in the morning (Funasaka et al. 2011). For this study, a majority of the samples were collected in the morning between 8:00 AM and 12:00 PM, including all of the matched blood and blow samples. However, the potential for a diurnal rhythm to affect these results cannot be dismissed. If the variation is significant, blow sampling would be an ideal method to use to study diurnal rhythm of testosterone secretion in belugas because blow samples can be collected with greater frequency than blood samples. Testosterone secretion may also be influenced by stress (Lynn et al. 2015) or contaminant load, which may be a particular problem for marine mammals that bioaccummulate endocrine disrupting toxins (Oskam et al. 2003;Subramanian et al. 1987). In future work it will be possible to measure both cortisol and testosterone in the same blow sample to investigate the influence of stress on testosterone in blow.
Hormone concentrations in blow samples are also affected by the sampling procedure. The belugas in this study exhaled once to clear away pooled water from their blowhole, but the volume of water remaining on or around the blowhole likely varied from sample to sample, leading to some of the variation in blow hormone concentrations within groups found in this study. Increased risk for water contamination during sample collection in Bristol Bay under restraint conditions may have led to low testosterone concentrations in male blow samples when compared to matching blood samples and obscured the correlation between blood and blow that was found to occur in aquarium belugas. The probable water contamination is reflected in the higher sample volumes collected from wild belugas when compared to aquarium belugas. However, all of the wild male samples fell within the range of testosterone concentrations found in aquarium beluga samples collected in the summer or fall. Increasing sample size will aid in determining if testosterone in blow is correlated with testosterone in blood in wild belugas as it is in aquarium belugas.
Although it did not interfere with the assays for testosterone or progesterone, there is non-specific binding that likely results from the collection material used in this study. Collection material is known to affect salivary testosterone measurements in humans (Celec and Ostatníková 2012). This material was selected based on its performance with a Cayman Chemical cortisol EIA kit with beluga blow samples (Thompson et al. 2014). The collection material was cleaned following Thompson et al. (2014), who did not report negative controls, but spiked nylon with cortisol standard and observed recoveries similar to the recoveries in this study after correcting for the negative control. Perhaps the method used to clean the nylon for cortisol measurement is inappropriate for testosterone or progesterone measurements. Hunt et al. (2014) also tested for interference from negative control collection material spiked with testosterone and cortisol. They did not find detectable testosterone from negative control samples, but their assay had a sensitivity of 50 pg/ml, which is higher than the mean negative control testosterone concentration found in this study (45 pg/ml) and thus a similar amount of interference would not have been detected using their assay.
This non-specific binding is undesirable and future work should identify methods to eliminate it, either by different cleaning methods or through the use of a different collection material. However, all of the samples used in this study were collected in the same way, and validation experiments demonstrate that this effect was consistent across samples, allowing comparisons between samples for the purposes of this study.
Eliminating this non-specific binding may also improve the diagnostic value of the method.

Application to unrestrained wild belugas
In order for hormone determination in blow to be applied to wild, freeswimming belugas, a single exhale should contain enough sample to perform the assay, as collecting multiple exhales from an unrestrained whale is unlikely. The hormone concentration should also not vary significantly from exhale to exhale, so that conclusions can be drawn from the analysis of the blow sample that is collected.
In a majority of the cases in this study, a single exhale would have yielded a large enough fluid volume to perform the assay at a 1:2 dilution. With additional validation experiments, the sensitivity of this assay would allow for samples to be diluted even further prior to analysis, requiring as little as 10 µl of sample. Collecting samples on two different sample collection devices demonstrated that testosterone content is relatively consistent from one breath to the next during a sampling event. Hunt et al. (2014) found large variation in hormone concentrations in successive breaths sampled in right whales. The findings in this study support their conclusion that this variability was most likely due to varying environmental water contamination and not physiological changes between exhales, although there is potential for increased accumulation of hormone via diffusion over time during long duration breath holds.
Sampling in field conditions also requires that samples be stored appropriately until analysis. Several sample handling experiments demonstrated that testosterone in blow is stable both during temporary storage while chilled and long term storage while frozen. Thus, we can be confident that the storage protocols in this study do not affect the interpretation of testosterone concentrations. These findings are consistent with the stability of testosterone in serum (Stroud et al. 2007) and saliva (Durdiaková et. al 2013) in humans. They are also consistent with the stability of cortisol in right whale blow (Hunt et al. 2014). Hogg et al. (2005) found that testosterone was not stable in frozen bottlenose dolphin blow samples, which they attributed to the activity of bacteria that may metabolize steroids, as occurs in feces (Khan et al. 2002). Therefore an antibiotic was added to the dolphin blow samples to improve stability. The observed differences with this study may be due to species differences in the bacterial communities within the respiratory tract, the variation in spiked testosterone levels for stability tests, or the different analytical methods used. Under the field conditions in this study, there was no evidence to suggest the use of an antibiotic was necessary to preserve beluga blow samples. This allows for an accurate measurement of the volume of the blow sample and eliminates the need for an extraction step prior to immunoassay for testosterone. It was also unnecessary to perform the alcohol extraction from the collection material described for right whale blow (Hunt et al. 2014) to acquire measurable testosterone or progesterone, although collection from free-swimming animals with high levels of environmental water contamination (as found in the right whale sampling) may ultimately necessitate an additional extraction step in belugas as well.
As recognized by others studying cetacean blow samples (Hogg et al. 2009;Hunt et al. 2014;Thompson et al. 2014), the most significant limitation to the application of this methodology to free-swimming cetaceans is dilution from environmental water. The use of exhaled breath condensate in human medicine faces similar challenges, and dilution markers may vary by analyte (Ahmadzai et al. 2013).
The use of hormone ratios or other potential markers of dilution may partially alleviate this limitation and should be explored. Although insignificant, the negative relationship between blow sample volume and testosterone concentration in this study suggests that in addition to variation in environmental water contamination, variable fluid production within the respiratory tract may also have an effect on the interpretation of hormone concentrations. Therefore, it may be necessary to identify a dilution factor for fluid production within the animal as well as dilution by environmental water to fully standardize results and allow strict comparisons between samples.
In addition to the ability to accurately measure hormones in blow, the reproductive state of an unknown beluga should ideally be distinguished from a single sample. Sampling a larger number of belugas in various reproductive states would allow for the development of statistical models that could be used to determine the probability that an unknown beluga is in a particular reproductive state based on hormone determinations in blow. Although the number of belugas in this study was small, this sample represents >60% of the belugas in US zoological facilities, including all of the living sires (Myers 2014). There is also the possibility that some samples were incorrectly classified by reproductive status due to the irregularity of blood sampling and ultrasound exams. The most likely misclassification was associated with the failure to detect luteal phases in non-pregnant females. Also, without an assessment of sperm production in the males studied, the maturity status may have been incorrectly assigned. The closely managed beluga breeding program in US zoological facilities will allow for increased sampling with greater frequency in the future.

Samples
An important advantage of blow sampling in zoological facilities is the relative ease with which samples can be collected, enabling frequent longitudinal monitoring of individuals. Blow sampling can be used to monitor the attainment of maturity in individuals, track reproductive cycles, and identify pregnancy in belugas that are not trained for blood or urine sampling or ultrasound examinations. The possibility of incorrectly identifying reproductive status would be reduced due to the ease of collecting multiple samples from the same animal. The ability to detect luteal activity, and thus ovulation, is important for belugas because they are facultative induced ovulators . The detection of an active luteal phase in a female with access to a male implies that the female likely copulated. Copulation is rarely observed in belugas, even when directed behavioral observations are conducted Chapter 4, this dissertation). Therefore, hormone determination via blow sampling would improve reproductive management and also creates new research opportunities by allowing frequent, repetitive sampling of individuals.
For wild belugas, this method is immediately applicable to mass stranding events such as those that occur in Cook Inlet, Alaska, where groups of belugas may temporarily strand between high tides. This scenario improves the ability to interpret blow hormone concentrations by removing the variability that would be associated with environmental water contamination. One such event involved the live, temporary stranding of at least 76 belugas, representing nearly a quarter of the 315 belugas estimated to comprise that stock (National Marine Fisheries Service 2015).
Significant portions of this endangered population could be sampled during a single event, providing important demographic information. Blow sampling would strike a balance between minimizing the stress of handling imposed on the animals while they are already under stress and maximizing the information that could be obtained during these events.
The ability to identify pregnant females non-invasively in this population would be a great benefit to population management. A significant effort is made in this population to identify individuals and to count calves during aerial censuses; having the ability to identify pregnant females and follow up on her success at birthing and rearing the calf would help identify potential causes for the lack of recovery in this population (National Marine Fisheries Service 2015). The ability to identify males in breeding condition is also important, as reduced availability of breeding males can negatively impact population viability ). Additionally, detecting maturity in belugas is difficult to perform through visual inspection alone; an improved understanding of the proportion of belugas that are mature through the use of blow sampling would aid in developing more accurate population models (Mosnier et al. 2015).
Taken together, the variation in blow testosterone and progesterone concentrations with reproductive status may also help to identify the breeding season in wild beluga populations. While there is a great deal of overlap, the peak of the breeding season appears to vary among stocks ; Heide-Jørgensen and Teilmann 1994), or is unknown or unreliably documented for some stocks due to the difficulties associated with determining the length of gestation in wild belugas . Determining when ovulations are occurring through progesterone determination and when males are in breeding condition through testosterone determination will help clarify when conceptions are actually occurring and thus identify periods when wild beluga populations may be more vulnerable to disturbance. In conjunction with photo identification of individuals and aerial surveys, blow sampling temporarily stranded individuals would allow for better management of endangered populations, such as the Cook Inlet stock.
Additional research is required for blow sampling to provide the same diagnostic capabilities as biopsy sampling in free-swimming cetaceans. However, the perceived risks associated with biopsy sampling may preclude research from being conducted or limit sampling capabilities. With further development, blow sampling will provide a less invasive, yet informative alternative for monitoring population demographics and viability.

Conclusion
This study advances the use of cetacean blow samples for hormone determination by demonstrating that testosterone and progesterone concentrations in beluga blow samples are biologically relevant, varying by sex, maturity status, season, and reproductive cycle stage. The positive correlation between testosterone and progesterone in blow and blood found in this study is an important step in the development of cetacean blow sampling as a diagnostic tool for studying population demographics, and further justifies the continued study of other analytes that can be detected in blow samples. Continued validation experiments and method development using samples collected from cetaceans in zoological facilities will improve the application of blow sampling to wild cetaceans. This study provides a framework for interpreting testosterone or progesterone concentrations found in blow samples, which can be used to monitor breeding in zoological facilities, as well as to identify pregnant females or males in breeding condition in groups of wild belugas that are temporarily stranded.

Literature Cited
Acevedo-Whitehouse K, Rocha-Gosselin A, and Gendron D. 2010. A novel noninvasive tool for disease surveillance of free-ranging whales and its relevance to conservation programs.  Gladden et al. 1997;Colbeck et al. 2012). Post-mortem genetic sampling has provided information that is critically important for population management in this species . Recently, blow (exhale) sampling has been identified as a source of DNA for analysis in cetaceans (Frère et al. 2010). Due to the forcefulness with which cetaceans exhale, cellular debris is commonly ejected along with the respiratory vapor that is also a source of steroid and thyroid hormones for analysis (Frère et al. 2010, Hunt et al. 2013  Although similar to biopsy sampling in that a boat must approach the animal in order to collect the sample, blow sampling does not impose physical harm upon the animal and thus may provide a less invasive alternative for acquiring molecular data. To date, cetacean DNA sampling in blow samples has been reported for bottlenose dolphins (Frère et al. 2010) and harbor porpoises (Borowska et al. 2014) in aquariums, and one wild bottlenose dolphin (Frère et al. 2010). Both studies found that DNA sequences isolated from blow samples matched those obtained from DNA isolated from blood from the same individual, validating the technique for use in these species. However, both of the studies utilized blow samples that consisted of more than one exhale (harbor porpoises: 5-6 breaths; dolphins: 4 breaths). While bow riding species may allow the collection of more than one exhale from the same animal, for many species, including the beluga, a single exhale is the most realistic sampling outcome. Therefore, while blow sampling for molecular analyses has been successful in these species, further investigation is required to determine if a single exhale would yield enough DNA to perform common analyses utilized in population management.
For blow sampling to be a reasonable alternative to biopsy sampling in cetaceans, DNA yield from a single exhale should be sufficient to allow for multiple experiments; for example, investigators may wish to identify the sex, mtDNA haplotype, and microsatellite genotype of an individual from the same sampling event.
Therefore Additionally, for blow sampling to be a realistic alternative to biopsy sampling, the DNA that is recovered from blow samples would not be excessively fragmented and allow for the amplification of a variety of target sequences, regardless of size.
Currently, the largest nuclear DNA fragment that has been amplified from cetacean DNA extracted from blow samples is approximately 160 base pairs (Borowska et al. 2014;Frère et al. 2010). Amplifying longer sequences from blow would be useful for molecular sex determination, which is often necessary because sex is difficult to determine in a free swimming beluga (Petersen et al. 2012). In belugas, molecular sex determination is commonly accomplished using a fragment of the ZF gene, which is approximately 1000 bp long (Shaw et al. 2003). The ability to amplify larger gene targets would also allow for research investigating evolutionary trends in immune function or the impact of anthropogenic effects on populations (O'Corry-Crowe et al.

2008
). Therefore, validating the ability to amplify longer nuclear sequences from host DNA isolated from blow samples would be valuable.
Belugas may be a good candidate for single-exhale blow sampling for DNA analysis, given their large size relative to previously studied species. Perhaps with a larger exhale volume, more cells will be carried up with each exhalation, leading to higher DNA yields per exhale relative to smaller cetaceans. The availability of belugas in zoological facilities enables method development, and blow samples from wild animals can also readily be collected from wild belugas that are temporarily restrained for tagging purposes (Thompson et al. 2014).
Using blow samples collected from wild and aquarium belugas, this project aims to determine the relationship between number of exhales collected per sample and DNA yield, as well as the downstream performance of the extracted DNA in polymerase chain reactions (PCR). PCR performance with both mtDNA and large nuclear target sequences (>900 bp) will be assessed. To maximize the potential utility with wild belugas, additional factors that may influence DNA yield from blow samples will also be explored.

Study Animals
Blow sampling in aquarium animals (

Blow Sample Collection and Handling
Sample collection methods were similar to Frère et al. (2010). Blow samples consisting of one, two, or four successive exhales were collected into a polypropylene Samples were shipped to the laboratory on dry ice or in liquid nitrogen.
A total of 11 aquarium belugas and 29 wild belugas were sampled (

DNA Extraction
After thawing, the 50 ml conical tubes were again rolled by hand to coat the inner surface of the tubes with buffer, and were then centrifuged for 10 minutes at 2060 x g. After pipetting up and down several times to dislodge material from the bottom of the tube, the fluid was pipetted from the conical tube into a 1.5 ml microcentrifuge tube. This tube was then centrifuged in a microcentrifuge for 10 minutes at 13400 x g.
The presence or absence of a cell pellet was then recorded. DNA was isolated using the Qiagen DNEasy Blood and Tissue kit (Valencia, CA), following the manufacturer's tissue protocol with the following modifications.
The addition of buffer ATL was reduced to account for any leftover volume of TE in the sample tube. The duration of the lysis step lasted one hour or in rare cases, until the pellet (if visible) was completely lysed, up to 3 hours. Samples were vortexed every 15-20 minutes during lysis. DNA was eluted in 50 µl of the provided buffer AE. This elution step was repeated into a separate tube. Two 50 µl elutions were preferred to a single 100 µl elution because the DNA was more likely to be concentrated enough in the first elution to then be used in PCR reactions without concentration.

Yield and Purity
DNA concentration (ng/µl) and purity (ratio of absorbance of 2 µl of sample at 260 and 280 nm) was then assessed using a NanoDrop 8000 Spectrophotometer (Thermo Scientific, Waltham, MA). The total yield from each elution step was calculated (assuming 50 µl sample volume), and the yield from these two separate elutions were added together to calculate the total yield. Reported A 260 /A 280 ratios are from the first elution, when the DNA was most concentrated.

Blood Sampling
For comparison, blood samples were collected from trained belugas at Mystic Aquarium (n = 4) from the ventral fluke vein with the voluntary cooperation of the whale. DNA was isolated from 100 µl of whole blood using the Qiagen DNeasy Blood and Tissue kit (blood protocol) using the manufacturer's instructions. Blood DNA was eluted once into 200 µl of the provided buffer AE. DNA concentration and purity was assessed via NanoDrop. For each beluga's blood sample, 3 or 4 separate extractions were performed for a total of 14 extractions.

Molecular sex determination via Polymerase Chain Reaction
PCR performance was tested for Mystic Aquarium samples through a molecular sex determination test. The zinc finger gene (ZF), which has a sex-linked polymorphism, was amplified using primers LGL331 (    blow and blood from three belugas was used in a mtDNA PCR; resulting bands were sequenced from one blood and one blow sample to ensure that results were replicated from the two DNA sources for each individual.

Sample Collection and Handling Effects on Yield
To determine if the strength of the breath influences DNA yield or PCR performance, 4 separate "calm breaths" were collected from 2 aquarium belugas (2 samples per beluga). Typically, the exhale collected from trained aquarium belugas is of similar force to the exhale used when the whale surfaces to breathe. The "calm breaths" were collected while the belugas were resting at the surface. The force of these breaths is much lower than the typically sampled breaths, and is similar to the force of the Bristol Bay beluga exhales collected under restraint conditions. DNA was isolated using the protocol described above, and an mtDNA PCR was attempted using 30 ng of template DNA.
Wild beluga samples were kept on ice for up to 6 hours while in the field until they could be frozen. To test the effect of temporary chilled storage on DNA yield and PCR performance, 4 single-exhale blow samples were collected from 2 aquarium belugas (2 samples per beluga). Buffer TE was added to the samples, and they were stored in a 4˚C refrigerator for 6 hours before being placed in a -20˚ freezer. DNA was isolated using the protocol described above, and an mtDNA PCR was attempted using 30 ng of template DNA.
To test the effect of long term storage on the ability to isolate DNA from blow samples, two single-exhale blow samples were stored at -80˚ C in 1.5 ml microcentrifuge tubes. One sample was stored for 22 months and the other was stored for 23 months before the DNA extraction protocol was performed. Yield and purity were assessed via NanoDrop. A ZF PCR was attempted using 30 ng of template DNA.
To test the influence of lysis time on DNA yield, two single exhale blow samples were collected from an aquarium beluga. After transferring the sample to a microcentrifuge tube, each sample was vortexed thoroughly and divided equally into two tubes. DNA was isolated from both tubes following the protocol described above, with one tube from each sample undergoing a 1 hour lysis time, and the matching tube undergoing a 6 hour lysis time. Yield and purity were assessed via NanoDrop.
To estimate the contribution of microorganism DNA from environmental water in the blow sample, DNA was isolated from 1 ml of water from various locations in the beluga whale exhibit at Mystic Aquarium (n = 3) or within Bristol Bay (n = 3 proportional with the number of exhales for either group (Table 3). Among aquarium samples, total yield was influenced by individual ( Fig. 2) as well as the presence or absence of a cell pellet following centrifugation prior to the extraction protocol (Fig. 3).   Table 3. The fold increase in mean DNA yield relative to the mean DNA yield for single exhale samples from the same population (aquarium or Bristol Bay).     Mitochondrial DNA sequences obtained from DNA isolated from blood and blow were identical for each beluga tested at Mystic Aquarium (gel electrophoresis results shown in Fig. 7, sequencing alignment for one beluga shown in Fig. 8

Discussion
This study has shown that DNA can be reliably extracted from beluga blow samples, and that single exhale blow samples can yield sufficient DNA to perform common molecular analyses, as well as those that require larger fragments of DNA.
The results of mtDNA haplotype sequencing from DNA extracted from single-exhale blow samples were identical to the results obtained from blood, validating the use of this method in belugas. This study also demonstrated that blow sampling for molecular analyses from temporarily restrained wild belugas can easily be performed while other tests or sampling are being conducted. The large number of blow samples studied revealed a wide variation in DNA yield, even within the same individual. Despite this variability, most blow samples ultimately allowed for PCR amplification of a gene target that is commonly used in beluga research, and many would have allowed for the study of more than one genetic marker. Blow sampling is associated with greater risk for the investigator than biopsy sampling given the range of possible DNA yield outcomes; the amount of DNA that can be acquired from a biopsy sample would undoubtedly be higher, even when compared to the highest quality blow sample.
However, the potential for blow sampling to serve as a less invasive alternative for acquiring DNA in belugas is clear, perhaps enabling research that would otherwise not be conducted.

DNA Yield from Blow Samples
The number of exhales collected per sampling event had a large effect on DNA yield, especially in Bristol Bay samples. Unexpectedly, yield was not proportional to the number of exhales collected, suggesting that the expulsion of cellular debris varied from exhale to exhale. The wide variation in yield from single exhale samples in aquarium belugas, even from the same individual, further supports this observation.
Instead, the forcefulness of the breath and chance collection of large pieces of cellular debris likely shaped the relationship between the number of exhales and the DNA yield. Often, a forceful breath would result in a large piece of mucous-rich debris to be expelled; collecting multiple exhales would increase the odds of this occurring during the sampling event. This could explain the increase in DNA yield per exhale seen in Bristol Bay beluga samples, and is also reflected in the higher yields observed from samples that had a cell pellet following centrifugation compared to those that did not.
While flow rates were not measured, the lower DNA yields from Shedd Aquarium samples may be related to the force of the breath, as these samples were less likely than Mystic Aquarium samples to have pellets following centrifugation.
In aquarium samples, the declining median DNA yield per exhale with increasing number of exhales per sample could also be explained by an initial expulsion of cellular debris and a small volume of water pooled on top of the blowhole in the first breath leading to relatively high yields, followed by a declining amount of debris and environmental water in successive exhales that leads to lower yields.
Submerging the blowhole and increasing the amount of time between exhales collected would more closely simulate the breathing pattern of a swimming beluga, and may have resulted in higher yields in successive exhales.
The forcefulness of the exhale that is collected likely resulted in the lower yields from Bristol Bay beluga samples relative to aquarium belugas. Aquarium belugas were trained to exhale forcefully to simulate the forceful exhale of a swimming beluga. By exhaling forcefully, a free-swimming whale ensures the clearing of water from the blowhole prior to inspiration and reduces the time spent at the surface. In contrast, the breaths of the Bristol Bay belugas were deep, yet calm and much less forceful than the breaths of trained or free-swimming belugas. The breaths observed under restraint conditions were similar to those seen in belugas under human care while sleeping or calmly lying at the surface. The breathing pattern of Bristol Bay belugas may be due to restraint conditions during temporary capture. In a study of pulmonary function with dolphins under human care, Brodsky et al. (2012) observed that dolphins that were voluntarily "beached" during testing had a 2-5 fold decrease in flow rates when compared to dolphins that were fully supported in the water. While the belugas were not completely beached during restraint in Bristol Bay, the thorax was usually touching the bottom, which may have reduced their pulmonary flow rates as a result.
Alternatively, the Bristol Bay belugas were exhibiting calm breathing patterns as an energy saving mechanism during capture. Either way, the resulting reduction in force or flow rates may have resulted in lower sample volumes, which would be expected to be related to the amount of cellular material present. In a study of dolphin exhaled breath condensate, less forceful breaths from one individual resulted in reduced sample volumes relative to other dolphins (Aksenov et al. 2014). In addition to reducing sample volume, perhaps reduced force led to the ejection of fewer cells with each exhale, and the chance ejection of larger pieces of cellular debris became less likely, ultimately leading to reduced yields. The 65% reduction in yield from aquarium samples consisting of "calm" breaths further supports this observation.
In the absence of large pieces of cellular debris, the volume of the blow sample (which would be influenced by both the number of exhales and the forcefulness of the breath) likely affects DNA yield. Blow sample volumes were not recorded in this study because the amount of buffer that was added to the sample could not be determined with sufficient accuracy in the field. In a separate study of more than 100 beluga blow samples, fluid volumes per exhale were routinely between 50 and 70 µl, ranging from less than 10 to several hundred µl (Ch. 1, this dissertation). While highly variable, this is considerably higher than blow sample volumes observed in harbor porpoises, which ranged from 15-50 µl from 5-6 breaths, or approximately 3-10 µl per exhale (Borowska et al. 2014). The larger size of the beluga likely results in larger sample volumes relative to the much smaller harbor porpoise. Borowska et al. (2014) observed that DNA yield increased with sample volume, so a wide variation in blow sample volume could therefore account for some of the variation in per exhale yield observed in this study.
The mean DNA yield per exhale determined from single-exhale aquarium samples in this study compares favorably to those found in other odontocetes, being higher than either of the smaller species previously studied (Frère et al. 2010;Borowska et al. 2014 Water from Bristol Bay also had higher DNA yields than aquarium exhibit water.
While no effort was made to reduce contamination by environmental water, the amount of water contamination was probably higher in restraint conditions in Bristol Bay than in the more controlled setting found in aquariums, further reducing the ratio of target DNA to total DNA for Bristol Bay belugas. However, when the quality of the sample is high, very little template is required for mtDNA sequencing, leaving sufficient template to perform multiple experiments with the same sample.
The purity of DNA samples also influences PCR success (Boesenberg-Smith et al. 2012). Spectrophotometry was chosen to assess sample purity in this study due to its relative ease and the small sample size with which the assessment can be made. The different handling conditions the Bristol Bay samples were exposed to could also have led to a reduction in DNA quality and thus PCR success rate.
However, there was no evidence that short term storage of the samples at cool temperatures prior to freezing had an effect the ability to extract or amplify DNA from aquarium blow samples, with yields similar to those achieved from samples that had been frozen immediately. More experiments are needed to confirm that this is the case, but the reduced DNA yield and PCR success in wild beluga samples relative to aquarium samples is more likely a result of the differences in breathing pattern than sample collection or handling methods that varied in the two conditions.

Future Directions and Current Utility
Further method development should focus on collecting as much of the blow sample as possible from a given exhale. The collection method used in this study facilitates the collection of the sample through centrifugation, but does not collect the entire sample. Due to the diameter of the tube compared to the diameter of a beluga's blowhole, it is likely that less than half of the actual blow sample is collected in this method. For a free swimming animal, a collection device large enough to capture as much of the blow as possible should be devised, with the ability to efficiently transfer the sample from the collection device into a tube that can then be centrifuged.
Acevedo-Whitehouse (2010) utilized sterilized plastic sheets to create a large surface area for sample collection from wild whales and used swabs to retrieve the samples.
We are currently investigating the use of a sheet of parafilm for this purpose; the hydrophobic nature of the parafilm facilitates sample transfer via pipette into a tube that can then be centrifuged. Collecting a larger portion of the exhaled plume will likely lead to increased yields and therefore improve the utility of this technique. The forceful nature of the exhale of a swimming beluga will also increase the likelihood that sufficient DNA can be collected from single breath.
PCR success can be improved by optimizing PCR conditions for low template inputs, and DNA yield may be improved by using different extraction protocols. The relatively high level of mucous in blow samples may impact yield. For saliva samples, an overnight proteinase K digestion can be utilized to improve yields (Zakharkina et al. 2011). In the pilot experiment reported here, increasing the proteinase K lysis time by 5 hours had no effect on yield, although the relative mucous content may vary from sample to sample. DNA extraction kits that are more efficient for smaller amounts of starting tissue may also improve yield.
Experiments should also be performed to determine the relationship between DNA yield and the distance between the sampling device and the blowhole, as the proximity of the sampling device to the blowhole in this study is unlikely to be replicated with a free-swimming beluga. Studying belugas in zoological facilities will allow for further method development before wild whales are approached for sample collection, minimizing the impact to wild belugas while maximizing the likelihood that this method can be utilized effectively.
Although further development is required for the application of these techniques to free-swimming belugas, blow sampling is immediately applicable to suitable for pixel intensity measurements) and 119 blood samples were available for analysis. Significant seasonal variation in testes volume, blood testosterone concentration, and testicular pixel intensity were observed, with peak activity occurring between January and April. Seasonality of testicular volume was best described by a cubic function, while seasonal variations in testosterone and pixel intensity were best described by quadratic functions. Individuals differed significantly in both testes size and rate of change. On average, testes size increased by 60% from minimum to maximum values. These results are consistent with observations of reproductive seasonality both in the wild and in zoological facilities, and suggest a relatively low demand for sperm in this species that is consistent with their classification as induced ovulators.

Introduction
In seasonally breeding mammals, males often demonstrate seasonal variation in the energetic investment for sperm production, conserving energy by reducing testes size or function when conceptions are unlikely to occur (Kenagy and Trombulak 1986). Among seasonally breeding odontocetes, seasonal variation has been detected in testosterone levels, testes size, sperm production, or seminiferous tubule diameter Belugas (Delphinapterus leucas), an Arctic and subarctic species of odontocete, breed in the late winter or early spring so that births occur in the summer, 15.5 months later . With this seasonal reproductive pattern, male belugas would be expected to undergo changes in testes size or function. Male belugas in zoological facilities have a seasonal pattern of testosterone production, with peak concentrations occurring from January through April . Conceptions also occur seasonally in zoological facilities, with 80% occurring in March -May, a range that agrees with estimates of breeding season in wild belugas Brodie 1971).
Postmortem evaluations of wild beluga testes demonstrate that most males reduce spermatogenesis outside of the breeding season, evidenced by testicular histology or epididymides devoid of sperm in mature adults Seaman 1988, Heide-Jørgensen and. Seasonal variation in testes size also appears to occur in wild belugas .
However, wild belugas are primarily sampled during the summer, several months after the presumed peak in breeding. Therefore, insufficient data is available to definitively describe the extent of seasonal variation in testes size. Collectively, postmortem studies have reported testes size measurements from more than 300 adult male belugas, yet only 1 observation is available for the months of December through March, and relatively few observations are available for April, May, and November relative to June through October (Brodie 1971;Finley et al. 1982;Kleinenberg et al. 1969;Sergeant 1973). Based on the current understanding of beluga breeding seasons, this gap in data occurs at a crucial time when testicular recrudescence is predicted to occur.
Longitudinal studies of live males with known reproductive histories would fill the existing temporal sampling gap and help describe the seasonal variation in testes size and function in belugas. The testes of odontocetes are located within the abdominal cavity, necessitating the use of ultrasonography to monitor live animals.
Ultrasonography is commonly used to assess testicular function in domestic species and has been applied to the study of reproductive function in male odontocetes in zoological facilities and in the wild (Kastelic and Brito 2012;Robeck et al. 2009;Alves et al. 2012). Estimates of testicular volume from ultrasound images correlate well with actual testicular volume measurements (Gouletsou et al. 2008). In addition to size, ultrasonography can also be used to determine the density of the testis tissue Given the seasonal variation known to exist in male beluga testosterone concentrations and the common pattern of seasonal variation in testes size in seasonally breeding odontocetes, it is likely that belugas also undergo seasonal changes in testes size. However, the degree of this change is unknown. The aim of this project was to determine if the suspected seasonal variation in testes size in belugas can be detected via the longitudinal monitoring of males with known reproductive histories. Blood testosterone concentrations will also be monitored for comparison, and the effectiveness of testicular pixel intensity measurements in evaluating seasonality will also be assessed.

Methods
The testicular volumes and blood testosterone concentrations of 5  a "Age" refers to age at the beginning of the study period. b Length and weight measurements were taken once during the study period.

Calculating testicular volume via ultrasonography
Ultrasound exams were performed with the voluntary cooperation of the animal while the animal lay unrestrained in lateral recumbency at the water's surface.
Exams were attempted once or twice per month in July through December and twice per month in the months of January through June by a single operator at each zoological facility. Although the specific ultrasound equipment varied by facility, all exams were performed with a convex 3.5 MHz probe.
To ensure that measurements were being made at the appropriate angle, the observer first visualized the testicular mediastinum, the thin hyperechoic band passing through the center of the testis (Brook et al 2000). Two still digital images of the longitudinal view of the testis were saved for analysis. If the length of the testis exceeded the footprint of the probe, then as much of the testis as possible was visualized with the caudal border contained within the image. Two still images of the transverse view at the midpoint of the organ were taken for each testis, for a total of 4 images per testis per exam.
Using these still images, measurements to the nearest hundredth of a cm were performed by the ultrasound operator using analysis software available on the ultrasound machine. Dorsoventral diameter (depth) and lateral diameter (width) were measured on the transverse images. Length was measured in longitudinal view. If the testis did not fit on the screen, then the length to the midpoint was calculated by measuring from the caudal border of the testis to the widest point of the testis. This measurement was then doubled to calculate the total length. The fat pads present on the ventrolateral surface of belugas made an indirect measurement of length using a ruler placed on the abdomen (as described Brook et al. 2000) unreliable in pilot observations, necessitating this direct approach. Both testes were measured within the same day. Each measurement was taken on both images of the same view, and the average of these two measures was calculated and used for analyses. While operators varied by facility, the same operator performed all of the exams and measurements for an individual animal throughout the study.
Total testicular volume (TTV), or the sum of the volumes of the right and left testes, was then calculated using Lambert's formula for the volume of an ellipsoid applied to each testis: V= (LWD)(0.71) (Brook et al. 2000).

Calculating Testicular Pixel Intensity
The pixel intensities (PI) of the testicular parenchyma relative to the pixel intensity of the blubber layer from testicular ultrasounds from DL1 and DL2 were determined using Image J (http://imagej.nih.gov/ij/) as an indicator of tissue density.
Because many factors can influence the pixel intensity of an ultrasound image, images were utilized for this analysis from a single ultrasound machine (at Mystic Aquarium) when the gain was equal to 50 and the scan depth was the same for both testes examined on the same day (DL1 = 34 observations, DL2 = 37 observations). Three points of analysis per image were averaged to determine blubber PI while six points of analysis were averaged to determine PI of the testicular parenchyma. Points were selected from homogenous regions of the image, avoiding areas that would artificially increase or decrease the PI, including the relatively brighter mediastinum and relatively darker shadowed areas. Blubber PI was subtracted from the PI of the testicular parenchyma to normalize for differences in the pressure applied to the probe by the operator.
Various scanning depths may have been used to adjust for seasonal changes in blubber thickness in an individual (range: 17 -25 cm). Altering the scanning depth will affect the PI measurements, which inhibits its use as an indicator of tissue density.
To correct for variation in depth between images, a correction factor was developed. Extracted blood samples were assayed primarily at 1:40, but ranged between 1:10 and 1:80 depending on the expected concentration of testosterone ). All samples were assayed in duplicate and the means were used in calculations. Individual samples with a %B/Bo between 20 and 80% and a coefficient of variation (CV) < 15% were accepted. Samples with CV >15% were re-assayed, and blood samples outside of the range of the kit were re-assayed at a different dilution. Significance was set at p < 0.05.

Seasonality of Total Testicular Volume
The appearance of beluga testes on ultrasound was as described in studies of other odontocetes (Brook et al. 2000) (Fig. 1). Due to variation in the animals' behavior, weather, or faulty equipment, some scheduled ultrasound examinations were not conducted. As a result, DL5 was not measured in the months of July or September, and DL2 was not measured in September or November of 2012 or March of 2013. TTV by month for each individual is shown in Table 2.
The length, width, and depth of the testes varied within and among individuals (Table 3). This variation occurred seasonally, with a seasonal pattern apparent for four of the five individuals (Fig. 2). TTV was significantly higher from Jan -Apr (p < 0.0001) and from Mar -May (p < 0.0001) compared to all other months, but the effect was stronger using the predictor season of Jan -Apr. There were significant differences between individuals in both tests of seasonality (random intercept term, p < 0.001 for Jan -Apr and p < 0.01 for Mar -May). Seasonality of TTV was best described by a cubic fixed effect model with a random intercept term and random linear and quadratic slope effects (Table 4). TTV was generally highest in winter/spring, and lowest in late summer/fall, with individual TTV increasing by 60% from the minimum measurement to the maximum measurement on average ( Table 5).

Seasonality of Blood Testosterone Concentration
Testosterone concentrations in blood varied significantly between Jan -Apr and all other months (p < 0.0001), but did not vary between Mar -May and all other months (p > 0.05). The effect of individual (random intercept term) was not significant for the comparison between Jan -Apr and all other months (p > 0.05), but was significant for the comparison between Mar -May and all other months (p < 0.01). Seasonal variation in testosterone occurred in four of the five whales (Fig 4).     Seasonality of testosterone was best described by a quadratic fixed effects model with a random intercept and random linear slope term. Intra-assay variation for groups A and B were 17.6 and 10.3%, respectively. Interassay variation for group B was 10.0% for the 100 pg control and 17.7% for the 25 pg control. The relationship between seasonal variation in testosterone and TTV is shown in Fig 5.

Seasonality of Pixel Intensity
A seasonal variation in the PI of testicular ultrasound images was apparent in both DL1 and DL2 (Fig. 6). Pixel intensity was significantly higher from Jan -Apr compared to all other months (p < 0.05), but was not different from Mar -May compared to all other months (p > 0.05). Seasonality of PI was best described by a quadratic fixed model with a random intercept and random linear slope term. An increase in echodensity of testicular tissue preceded the increase in testes size in DL1, while it was coincident with the increase in testes size in DL2. In both animals, echodensity decreased prior to the decrease in testes size (Fig. 6).

Fig. 5. Seasonal variation in TTV (points separated by individual on primary y axis)
and testosterone (bars on secondary y axis; mean of individual means ± SD). Filled circles represent the mean of the individual means of TTV by month, with the cubic fixed effects regression model plotted (gray line). TTV is plotted in raw form (not normalized to body length, as in statistical analyses). Fig. 6. Seasonal variation in TTV, blood testosterone, and the pixel intensity of testicular ultrasound images (from right to left) for DL1 (top row) and DL2 (bottom row). Open symbols represent individual observations, while closed circles represent the monthly mean. Lines represent the fitted curves determined from statistical analyses (TTV: cubic; testosterone and pixel intensity: quadratic).

Year to Year Variation within Individuals
Both DL1 and DL2 showed similar patterns of seasonal change in TTV, testosterone, and testicular PI (Fig. 6). While DL2 reached similar peak TTV measurements from year to year, DL1 reached a higher TTV in the second year relative to the first (Fig. 7). This was also reflected in a different rate of increase in   The testes sizes measured in this study are similar to those found in postmortem studies of belugas (linear measurements: Heide-Jørgensen and Teilmann 1994; volume measurements: Kleinenberg et al. 1969, Brodie 1971. As testes mass is the most commonly reported measure of testes size, determining the relationship between testes volume measured via ultrasound and testes mass would be helpful in expanding the utility of this method for assessing reproductive function in wild belugas. To our knowledge, only one report is available where mass and volume of the same testes were reported (Kleinenberg et al. 1969). Using the small data set reported by Kleinenberg et al. (1969) (n = 5 adults), a relationship between mass and volume can tentatively be obtained (M (g)= 1.13V(cm 3 ) -27.5, r 2 = 0.997). Using this equation and the minimum and maximum volumes observed in this study (Table 4), the belugas studied had combined testes masses of 605-1941 g. These values are very similar to published values, although with a higher peak in this study, likely due to the lack of data from peak season in post-mortem studies.
Significant individual variation was found in testes size and in the rate of change in testes size with season. In post-mortem studies of belugas, wide variation in testes size has been found for belugas of similar body length (e.g. Sergeant 1973)  in hormone assays between studies could explain some of the variation, it is also possible that some males do not maintain sufficiently high testosterone throughout the year, resulting in the individual variation in spermatogenic activity found in . However, without additional information on spermatogenesis to complement testes size or testosterone data, the relationship between these measures of reproductive function are currently unknown.
Comparisons of testes size between individuals in this study should be made cautiously, as different observers were necessarily used at each zoological facility.
Although variation between measures can be expected when using ultrasound on an unrestrained live animal, the degree of intra-observer variation was relatively small and did not obscure the seasonal pattern of testicular growth and regression. Intraobserver variation for this study was similar to that found in a study of dolphin testes via ultrasound, where replicate linear measures generally varied by less than 0.5 cm (Yuen et al. 2009  . The difference in testicular growth that season relative to the others may have been a response to this change in social grouping. One possible mechanism for this difference could be the "challenge hypothesis," where a new social challenge causes an increase in reproductive activity in a male that previously was not challenged for breeding opportunities ). Alternatively or in addition, the "Coolidge effect" may have stimulated higher reproductive activity via the introduction of novel females to the social group (Dewsbury 1981). In contrast, DL2 did not experience changes to social grouping during the study and the degree of seasonal variation was consistent from year to year.
Social influences on sperm production have been suggested for managed groups of bottlenose dolphins (Tursiops truncatus), but the effect was hypothesized to be inhibitory (Robeck and O'Brien 2004). In the wild, adult male belugas travel together , and thus maintaining multi-male social groupings has been a goal of the cooperative managing belugas in zoological facilities in the US and Canada.
Reproductive rate is also presumed to be higher in multi-male groups, and the enhanced reproductive activity in DL1 coincident with a social change may provide a mechanistic explanation for this observation.
DL5 apparently did not undergo seasonal variation in either testes size or testosterone concentration during the study period, although both testes size and testosterone concentration were within the range of values found in the other belugas studied. DL4 shared the same environment and displayed seasonal variation in both testes size and testosterone, indicating that environmental cues were not a significant factor. DL5 has not sired a calf in his lifetime, suggesting this pattern may be abnormal, but other factors such as access to females in breeding condition can contribute to breeding history. Disease can cause senescence in adult male bottlenose dolphins (Kemper et al. 2014), but this animal had no signs of illness and continues to be healthy 7 years later. As the oldest animal in the study, it is also possible that the lack of seasonal variation may be due to age-related senescence. Some degree of senescence has been observed in other odontocetes, including pilot whales (Desportes et al. 1993) and finless porpoises (Wu et al. 2010a). Detecting senescence in wild male belugas would be difficult because sampling typically occurs out of breeding season, when most adult males are in the regression phase .
With an understanding of typical seasonal changes, further longitudinal study of belugas in zoological facilities could be performed to assess age-related changes to this pattern.
These data suggest that the spring and autumn equinoxes may carry important photoperiod information for belugas. Testicular regression appeared to begin following the spring equinox, and recrudescence appeared to begin following the autumn equinox. Many physiological changes associated with the change in season are driven by photoperiod, including changes in circulating testosterone levels and spermatogenesis in some mammals (Goldman 1999 The apparent time lag between peak testosterone level and peak testes size is consistent with the time lag observed between peak testosterone and conceptions in . This time lag would be expected to occur, as high testosterone is required to support spermatogenesis and thus recrudescence of testicular tissue. Due to the cycle of the seminiferous epithelium, there is also a lag between the initiation of testicular growth and peak spermatogenesis, or even between the initiation of testicular growth and the initiation of spermatogenesis (Martinet 1984 and loss of echodensity of testicular tissue at this time suggests that spermatogenesis is slowing as well. It is possible that spermatogenesis is suspended by this time, as has been found in wild belugas sampled as early as April that lacked sperm in the epididymis . This contrasts with harbor porpoises, which maintain elevated testes size for 1 month beyond when females are typically receptive . This apparent decline in demand for sperm prior to the termination (or perhaps even the peak) of breeding season suggests that belugas are able to establish sufficient sperm stores by April to allow for conceptions later in the season. This further supports a spermatogenic cycle of about 60 days, as it would be energetically beneficial to stop or slow sperm production 2 months in advance of the end of the breeding season.
The degree of seasonal variation found in these belugas was less than the degrees of change found in some other species of seasonally breeding odontocetes, which may experience 4-5x increases in testes size. However, the degree of change was similar to the seasonal increase found in seasonally breeding sheep (~67% increase) and long-finned pilot whales (Globicephala melas) (~50% increase) (Ortavant et al. 1988;Desportes et al. 1993). Even at maximum size, belugas have small testes relative to other cetaceans. The linear measurements of maximum testes size in this study were similar to those made on a harbor porpoise that weighed 37.5 kg, only 4.2% of the weight of the smallest beluga in this study (Desportes et al. 2003).

Testes size is often used to infer mating systems in mammals (Kenagy and
Trombulak 1986). In other odontocetes, relatively large testes are thought to be in response to high copulation rates during very short breeding seasons (approximately 2 weeks in the harbor porpoise, Read 1990) or protracted breeding seasons with a high level of sperm competition (common dolphin, Murphy et al. 2005). Either situation creates high demand for sperm to increase reproductive success. In addition to small relative testes size, the small seasonal change in testes size and regression of the testes prior to the end of the breeding season observed in this study all imply relatively low demand for sperm in the beluga. However, next to nothing is known about the mating system of belugas, with several disparate strategies proposed with varying degrees of pre-and post-copulatory competition between males based on morphological characteristics such as sexual dimorphism and relative testes size . However, this apparent low demand for sperm is consistent with the recent discovery that belugas are facultative induced ovulators . In induced ovulators, the first male to mate with a female has a much higher chance to sire offspring than successive males, and these species tend to have smaller testes than males in spontaneously ovulating species . The low testosterone levels during peak breeding season in this study suggests that male-male agonistic competition is also low in this species, as high levels are associated with aggressive behavior in other mammals (Trainor et al. 2009  Studies of beluga breeding behavior and paternity are needed to determine the mating rate and perhaps degree of promiscuity to improve our understanding of beluga mating strategies.
The improved understanding of the seasonality of reproduction will aid in the management of individual belugas in zoological facilities, allowing managers to identify the best time of year to train voluntary semen collection for use with artificial insemination, establish maturity, assess reproductive capabilities, and diagnose reproductive abnormalities. The methods employed here could also be used for nonlethal assessments of reproductive function in wild belugas that are temporarily restrained for satellite tagging (e.g. . Longitudinal sampling throughout the year, made possible by the study of belugas in zoological facilities, will continue to help fill gaps in our understanding of beluga reproduction. Given the logistical impediments to making direct observations of this Arctic-dwelling odontocete and the resulting difficulty in identifying the timing of key reproductive events, these studies will have an impact on wild beluga conservation and management.

Acknowledgements
Considerable assistance leading to the initiation of this project was provided by Tracey Several measures indicated that the adult female in this study preferentially engaged in courtship with one of the males. As paired physiological and behavioral observations of individual wild belugas throughout the year are not feasible, these observations provide an important complement to the study of wild belugas.

Introduction
The inherent link between reproductive behavior and population dynamics creates a need for an understanding of mating strategies to facilitate population management. As some strategies may be more resilient to perturbation than others, understanding the strategy employed by a species is important information when faced with rapid environmental change . Belugas (Delphinapterus leucas) are an Arctic and sub-Arctic species of cetacean that are well adapted for life among sea ice, using the ice edge for foraging (Asselin et al. 2012), a barrier for predation by killer whales (Sergeant and Brodie 1969), and as a driver of seasonal migration patterns and habitat use (Hornby et al. 2015). Sea ice loss or increasing water temperatures may thus be expected to alter beluga behavior. Variation in sea surface temperature can affect beluga migration patterns (Bailleul et al. 2012) and sea ice loss can affect their distribution . These changes could also affect social behavior in belugas, as males and females are known to segregate outside of the breeding season and exploit differing habitats and ice conditions . A reduction in habitat variability through the loss of sea ice may alter malefemale association patterns, which could impact reproductive behavior.
Belugas breed in the late winter or early spring, when most populations are inaccessible to human observers . Thus, next to nothing is known about the breeding behavior of wild belugas. Observations of male-female associations are possible during the summer months in some areas (e.g. Alekseeva et al. 2013), but these interactions are unlikely to be associated with true breeding behavior given the timing of births and the gestation period of belugas . Instead, these observations may reflect socio-sexual behavior, which is known to be a component of the social relationships of other cetaceans (Mann 2006). The availability of belugas in aquaria has allowed descriptive studies of reproductive behavior, but these observations have not yet been evaluated in the context of reproductive physiology and thus possible mating strategies in this species ).
Due to the absence of direct observations, beluga mating strategies have been inferred primarily using morphological characteristics ). However, the conclusions drawn from these inferences do not agree across studies. Belugas have alternatively been thought to have a polygynous mating system with competition between males for access to mates, or a more promiscuous system in which sperm competition may play an important role in belugas relative to the narwhal, a close relative.  point out the lack of knowledge of the beluga mating system, but suggest that both pre-and post-copulatory selection are weak in this species. In their analysis,  noted that belugas lacked obvious male weaponry or behavioral displays for demonstrating male quality, and demonstrated a relatively small degree of sexual size dimorphism and investment in testes size.
Recent discoveries about beluga reproductive physiology will improve the ability to infer beluga mating strategies in the absence of direct observations.
Seasonal variation in testes size and tissue density supports the finding that some males may suspend sperm production outside of the breeding season (Chapter 3, this dissertation; . The relatively small maximum size of beluga testes as well as the small seasonal change relative to other seasonally breeding odontocetes indicates that demand for sperm is low in this species Chapter 3, this dissertation  . With the observation that most conceptions occur after testosterone concentrations begin to fall and testes begin to regress ; Chapter 3, this dissertation), the frequency of courtship behavior may be unrelated to testosterone in this species, in contrast to convention in vertebrates . Although very few individuals have been studied, belugas also appear to have small ejaculate volumes relative to other cetaceans Alexa McDermott, personal communication). Taken together, these observations suggest a reduced role for sperm competition or male contest competition relative to other species of cetaceans. These findings could all be related to the recent discovery that belugas are facultative induced ovulators .

Induced (vs. spontaneous) ovulation is associated with a different set of
predictions of reproductive behavior, as the ability of an individual male to monopolize paternity is generally increased and male-male postcopulatory competition is reduced . Copulatory behavior and association patterns may depend on the amount of time between copulation and ovulation . In belugas induced to ovulate with a GnRH analog, this time period is approximately 36 hours ). If subsequent mates have a chance to fertilize a female's ova, then the first male to copulate with the female might be expected to engage in longer courtship or mate guarding to thwart breeding attempts by other males to ensure paternity . Given the difficulty of monopolizing mates in a three dimensional marine environment, females may select to engage in lengthy courtship with a superior male as a way to ensure suboptimal mates do not copulate with her. In addition, if post-copulatory selection mechanisms are lessened with induced ovulation, then pre-copulatory selection by females would be of greater importance in ensuring high quality mates . Therefore, female behavioral mechanisms for employing pre-copulatory selection of mates would be expected to occur.
To most accurately describe the mating strategies of a species, it is necessary to determine the reproductive condition of an individual at the time that reproductive behavior is observed, as individuals would be expected to behave differently depending on their reproductive condition (e.g. Muraco and Kucjaz 2015). Gonad function, measured through testes size or follicular development, is a clear indicator of an individual's reproductive condition and is often used to assess maturity and reproductive condition in adults Muraco and Kucjaz 2015).
Reproductive steroid measurements can be used to detect changes in gonad function, and thus reproductive condition. Therefore, studies that correlate steroid hormone concentrations and behavior in wild animals are common across taxa, including several species of marine mammals . Studies that incorporate measures of gonad function, hormone measures, and behavior are also feasible in some wild terrestrial mammals. For example, a study of free-ranging Soay sheep correlated scrotal circumference, testosterone concentrations in blood, and breeding behavior (Preston et al. 2012). However, longitudinal sampling of behavior and physiology in free-ranging cetaceans poses prohibitive logistical challenges.
Hormone sampling in wild cetaceans is limited by logistics, with blubber (via biopsy sampling) or feces presenting the most viable methods for obtaining reproductive steroids for analysis Rolland et al. 2005). However, fecal sampling can only be performed opportunistically, and biopsy sampling is inappropriate for repeated sampling of individuals. In zoological facilities, cetaceans can be trained for blood sampling or urine sampling as less invasive ways to monitor reproductive steroids . However, repeated blood sampling is often undesirable in an effort to limit inflammation, and urine sampling is trained less frequently than blood sampling in zoological facilities. As a result, paired studies of reproductive behavior and endocrinology in cetaceans are rare, but yield important insights on the reproductive biology of a species Muraco and Kucjaz 2015; Dudzinski et al. 2012). Additionally, research on beluga reproductive physiology in aquaria has yielded results that are consistent with and often enhance the knowledge of wild beluga reproductive physiology, which is obtained primarily in the summer, outside of the breeding season .
While a study of a small population of belugas in an aquarium cannot be used to infer the mating system or specific reproductive strategies of all wild belugas, the more intensive data collection that is possible in this setting can provide valuable information that can be used to inform studies of wild belugas.
In this study, non-invasive methods will be utilized to assess the relationship between reproductive physiology and behavior of a group of aquarium belugas. These results will be interpreted in the context of the current state of knowledge of their reproductive biology as well as their behavior in the wild. Intersexual interactions are expected to show seasonality, as would intersexual socio-sexual behavior. Thus, courtship behavior is expected to be seasonal in nature and occur most frequently when females are most receptive (the follicular phase of the estrous cycle). Based on the understanding of conception timing relative to reproductive physiology, courtship is expected to be unrelated to testosterone concentration or testes size in males. The relative importance of male contest competition will be evaluated by quantifying male-male aggression and affiliative behavior. Following predictions based on ovulation mode and testes size in this species, a low copulation rate is expected, and opportunities for females to employ precopulatory mate choice are expected to occur. -a Length and mass were measured once during the study period. b Age refers to the age of the animal at the beginning of the study period.

Reproductive Hormone Analysis in Blood and Blow
Blow (exhale) samples were collected twice per week from the males and once per week from the females, and analyzed as described in Richard et al. (Chapter 1,this dissertation). Briefly, blow samples were collected onto a pre-cleaned nylon mesh (110 µm, Elko Filtering Co., Miami, FL) stretched over a petri dish and secured with a rubber band. For sample collection, the whales were trained to lift their head so that the blowhole was above the water's surface. The whale would then exhale once to clear any excess water from the blowhole. Then, 2-8 successive exhales were collected onto the same mesh. Fluid blow samples were retrieved from the nylon via centrifugation, and samples were stored at -80˚ C until analysis.
Blood sampling was attempted once per month with F1 with the voluntary cooperation of the animal as a part of routine veterinary monitoring. Blood samples were collected into sodium heparinized vacutainer tubes from the ventral fluke vein.
One ml of serum or sodium heparin plasma was obtained through centrifugation (2000 x g for 10 minutes at 10˚ C) and stored at -80˚ C. Blood and blow samples were collected in the morning hours between 0900 and 1130, typically during the first training session of the day.
Male samples were assayed for testosterone using a commercially available enzyme immunoassay (Cayman Chemical, Ann Arbor, MI, Item #582701) validated for use with beluga blow samples by Richard et al. (Chapter 1,this dissertation). A commercially available enzyme immunoassay (Cayman Chemical, Ann Arbor, MI, Item #582601) validated for use with beluga blood and blow samples (Chapter 1, this dissertation) was used to assay female samples for progesterone. For both testosterone and progesterone measurements, 55 µl of sample was required; blow samples with volumes <55 µl were unavailable for assay. A diethyl ether extraction step was performed on female blow samples assayed for progesterone (Chapter 1, this dissertation). Blood samples were extracted according to the manufacturer's instructions. All samples were assayed in duplicate and the means were used in calculations. Individual samples with a %B/B o between 20 and 80% and a coefficient of variation (CV) < 20% were accepted. Samples with CV >20% were re-assayed.
Two standard controls were run in each assay (testosterone: 100 and 25 pg/ml, n = 12 assays; progesterone: 200 and 50 pg/ml, n = 14 assays). Inter-assay variation was calculated by determining the CV for the two standard controls on each plate. Intraassay variation was calculated by averaging the CV for all of the samples with 20-80% binding on each plate. When available, multiple samples collected within the same week were averaged to obtain weekly testosterone values. Some of the hormone measurements in blow samples have been presented previously (Chapter 1, this dissertation), but are repeated here to contextualize behavioral observations.

Ultrasonographic assessments of reproductive condition
Total testicular volume (TTV) was determined for the males in this study as described previously (Chapter 3, this dissertation). Briefly, ultrasound exams were performed with the voluntary cooperation of the animal while the animal lied unrestrained in lateral recumbency at the water's surface. Exams were attempted twice per month by a single operator using a convex 3.5 MHz probe. Two still images of the longitudinal view and 2 still images of the transverse view at the midpoint of the testis were taken for each testis, for a total of 4 images per testis per exam. Using these still images, measurements to the nearest hundredth of a cm were performed by the ultrasound operator using analysis software available on the ultrasound machine to the nearest hundredth of a cm. Dorsoventral diameter (depth) and lateral diameter (width) were measured on the transverse images. Length was measured in longitudinal view. Each measurement was taken on both images of the same view, and the average of these two measures was calculated and used for analyses. Total testicular volume, or the sum of the volumes of the right and left testes, was then calculated using Lambert's formula for the volume of an ellipsoid applied to each testis: V= (LWD)(0.71) (Brook et al. 2000). Data on M1 has been reported previously (Chapter 3, this dissertation), but is repeated here for comparison to behavioral observations.

Inferring ovulation
The occurrence of ovulation in F1 was inferred through progesterone measurements in blow, using known estrous cycle stage lengths in belugas reported by . Previous work has shown that ovulation leads to an increase in progesterone concentrations in blow (Chapter 1, this dissertation). In F1, the presence of a corpus luteum (CL) was associated with an increase in blow progesterone that lasted approximately 20 days, from a baseline of 248.5 ± 62.5 pg/ml to 326.5 ± 33 pg/ml (mean ± SD). However, progesterone measurements in blow made when a CL was absent occasionally occurred within this elevated range, while two samples collected when a CL was present had progesterone concentrations <300 pg/ml (Chapter 1, this dissertation). To minimize the chance of a false positive, ovulation was inferred to have occurred if two consecutive weekly samples exceed 326.5 pg/ml, indicative of the luteal phase lasting 29-32 days in non-conceptive cycles. The timing of ovulation was inferred by considering the lengths of the estrous cycle stages and all progesterone measurements available because the first elevated progesterone sample may not correspond to the start of the luteal phase given the time between samples.
The follicular phase, which lasts 14-27 days, was conservatively presumed to occur in the week that ovulation was inferred to have occurred, as well as the two weeks preceding the week of ovulation. The follicular phase was allowed to overlap with the luteal phase of a previous cycle. Inferred ovulations must be separated by >30 days, given the inter-estrus interval of 33-34 days in two animals studied by Steinman et al. (2012).
If available, progesterone concentrations in blood samples were used to confirm or refute inferences. Progesterone data from blood were considered more clearly interpretable than those from blow, given the wider variation within a reproductive condition that can be found in blow relative to blood (Ch. 1, this dissertation). Progesterone concentrations in blood <1000 pg/ml was considered baseline (no pregnancy or luteal activity); pregnant females or females with an active CL have progesterone concentrations in blood >4000 pg/ml (Ch 1, this dissertation).

Ethogram
An ethogram was developed using published descriptions of beluga behavior DiPaola et al. 2007), as well as pilot observations of the study group performed in the breeding season (Feb -Apr) of the previous year. A social interaction was defined as occurring when two or more whales are performing any of the social behaviors listed in the ethogram. The behaviors of interest for this study are listed in Table 2. Of particular interest was the "genital present" (Fig. 1). One whale length (~4 m) was used as a distance frame of reference for several behaviors.  (2007), , and pilot observations of the study group.
Behavior Definition

States Group Swim
Two or more whales swim in the same direction at approximately the same velocity; all whales are within 2 m of at least one other whale in the group; bodies can be aligned or staggered (one whale swims ahead of the other), but one whale may not be completely behind another; body orientation of individuals vary

Social Milling
Two or more whales actively swim, drift passively, or lie still with no discernible pattern or in variable directions within 4 m of each other; may be associated with social displays

Events Approach
A whale alters swim direction or speed to initiate interaction with another whale(s) while the other whale(s) does not alter swim speed or direction; resulting position is less than 4 m of recipient whale; interaction is initiated

Separate
A whale alters swim direction or speed to terminate interaction with another whale(s) while the other whale(s) does not alter swim direction or speed; resulting position is > 4 m of previously interacting whale(s), terminating the interaction

Open Mouth
With rostrum pointed in the direction of another whale, whale opens mouth wide enough so that the tongue is (or would be) visible

Bite
A whale makes contact with another whale with an open mouth and partially closes jaws upon contact

Bite Threat
With an open mouth, whale moves toward another whale to a distance of 1 m or less, but does not make physical contact

Rake
A whale drags open mouth along the body of another whale so that either or both jaws make contact with the body of the other whale

Jaw Clap
With rostrum pointed in the direction of another whale, whale rapidly and forcefully claps jaws together, once or several times in rapid succession; creates a percussive sound that may or may not be audible to the observer Chase A whale swims rapidly at another, who flees in response

Head Thrust
A whale sharply directs rostrum in the direction of another whale and rapidly returns head to its original orientation; depending on the location of the other whale, the direction of the thrust may vary; may be associated with corresponding body movement anterior to the dorsal ridge

Melon Shake
A whale vigorously shakes head in dorsal/ventral plane, causing the melon to shake; behavior has a recipient if rostrum is directed at another whale

Melon Flat
Anterior portion of the melon is compressed, flattening the melon along the maxilla, reducing or eliminating the normal rounded shape of the melon

Melon Push
The anterior portion of the melon is pushed outward, moving the normal rounded shape of the melon toward the rostrum of the whale leaving a depression behind the melon

Erection
Any part of the penis is visible during any social interaction; may occur simultaneously with any other social behavior in this ethogram

Ventral Present
Whale rotates along long axis so that ventral surface is oriented toward another whale

Genital Present
Male whale stops active forward progress by terminating fluke beating and drifts in the direction of another whale while arching their caudal peduncle so that the genital region is pushed closer to the recipient whale; caudal end of the caudal peduncle is correspondingly angled dorsally; rostrum is often directed toward the recipient whale for some portion of the presentation causing the body to assume an "S" shape; flukes and flippers may be held at various angles to control the drift towards the recipient whale; may result in contact of the genital region with the recipient or resumption of locomotion Genital Present Posture Same as genital present except the acting whale does not drift in the direction of another whale; recipient is the whale closest to the presenting whale at the time of the present Fig. 1. A sequential view of a genital present from M1 (bottom of frame) toward F1 (center of frame). Note the ventral orientation of F1 relative to M1 in frame 1 and the lateral orientation of F1 relative to M1 by frame 3. The "S" shape is clear in frame 4.

Observation Protocol
Continuous observations were conducted on the social group using a tripod- The days of the week where observations occurred were not consistent throughout the study, but observation effort each week was consistent; therefore behavioral data will be grouped by week (#1-52) and will be compared to weekly assessments of reproductive physiology. The distribution of morning to afternoon filming was not equal in weeks 1 (3 PM, 1 AM), 29 (3 PM, 1 AM), and 52 (1 PM, 3 AM). In weeks 21, 32, and 40, 5 observation sessions were required to reach 4 h of observations in that week. A total of 211 observation sessions were conducted for a total of 208 h of observation.

Quantifying Behavior
Video was first screened for the presence of social interactions by one of two observers. The animals engaging in the interaction were identified, as was the duration of the interaction to the nearest 5 s (minimum 5 s), and the animal that initiated ("Approach") and terminated ("Separate") the interaction. Approaches and separations may not have been identified for interactions that either started or ended prior to or after the observation session, if the interaction was initiated or terminated out of view of the camera, if the whales interacted while swimming past each other, or if the whales appeared to initiate or terminate the interaction mutually. At this stage, the behavioral state was also identified. One behavioral state was assigned to each interaction. Some interactions were relatively long and included both milling and group swim behavioral states; these interactions were assigned group swim as the behavioral state. It was possible for the social grouping to change during a continuous interaction if a third whale joined 2 interacting whales. This observation was considered two separate interactions (the 2 whale interaction and the 3 whale interaction). Using these data, a weekly record of social interactions was created. The total number of interactions, the average duration of interactions, and the total amount of time interacting was calculated for each social grouping and each behavioral state.
In the next stage of analysis, behavioral events performed during social interactions were coded and quantified using CowLog software (Hänninen and Pastell 2009) and continuous recording (Martin and Bateson 2007) by one of three observers.
For each event, the actor, behavior, and recipient were recorded. In social groups of more than 2 whales, if the acting whale performed a behavior that was not directed toward a clear recipient, the recipient whale was considered to be the whale closest to the acting whale. Behavioral event coding was not possible for interactions that occurred in the satellite pools that lacked underwater viewing, although the occurrence, duration, and participants of the interactions in these pools were recorded.
Weekly coefficients of association (COA) were calculated using the simple COA described in Cairns and Schwager (1987), with the amount of time spent interacting divided by the total amount of time observed per week, in minutes. Unless otherwise noted, the COAs between M1 and M2 were calculated using only interactions that occurred in the absence of a female. Weekly behavioral frequencies (events per minute of observation) were calculated separately for each individual, unless otherwise noted.

Genital presents
Every occurrence of a genital present between a male and female was reviewed by a single observer to describe the behavior in greater detail and to determine the recipient's response to the genital present display. The behavioral state and social grouping at the time of the genital present was recorded. The general area of the recipient's body the actor directed the display toward (right lateral, left lateral, ventral, or dorsal surfaces), the relative body position of the actor during the display (horizontal, head angled toward bottom, head angled toward surface), and the position of the whales in the water column (completely submerged or any part of the body at the surface) were also recorded. A "receptiveness score" was developed for the recipient and assigned to each occurrence of a male toward female genital present (Table 3), with higher scores indicating increased receptivity. Table 3. "Receptiveness score" scheme used to assign receptivity to genital presents. Each occurrence received one score from each of the three categories (swim speed, orientation, and contact), and the total score is the sum of these three component scores.

Swim Speed
Recipient alters swim speed to increase distance from genital present 0 Recipient continues to make forward progress during genital present (swim speed unchanged) 1 Recipient stops active swimming and drifts during genital present 2

Orientation
Recipient rolls ventral surface away from the genital present 0 Recipient rolls ventral surface toward the genital present 2 Recipient remains stable along long axis so that genital present ends closer to dorsal surface than ventral surface 1 Recipient remains stable along long axis so that genital present ends closer to ventral surface than dorsal surface 2

Contact
Contact does not occur 0 Contact occurs on lateral or dorsal surfaces 1

Inter-observer reliability
Inter-observer reliability for the coding of behavioral events was assessed by having all three observers code the same 10 observation sessions selected from 10 different months. This video (10 hours total) included 133 minutes of social interaction (90 min of male/female interactions and 43 min of male/male interaction).
The behavioral event coding for each observer was then aligned by the time each behavior occurred, allowing side by side comparison of each event. For each behavior, a pairwise kappa statistic was determined, using observer #1 as the reference observer. Behaviors coded by both observers #2 and #3 with "good" agreement or better (κ > 0.6) with observer #1 were considered in further analyses (Kaufman and Rosenthal 2009).

Statistical Analysis
Small sample sizes and dependent measures limited the ability to perform rigorous statistical analysis. Descriptive statistics (mean ± standard deviation) were used to assess seasonality of behaviors (January -June vs. July -December) and variation with the occurrence of follicular phases in F1. Box plots displayed show the interquartile range (box), the median (bold line), and the whiskers show the maximum (Q3) or minimum (Q1) value ≤1.5 times the interquartile range. Individual linear regressions were performed for each male to describe the relationship between weekly testosterone and the weekly frequency of genital presents or aggression in R (R Core Team 2015).

Testosterone Assays
Two blow samples per week were available for all weeks for M1 except weeks 3, 25, 36, 39, 42, 42, 50, 51, where only a single sample was available for assay. Two samples per week were available for all weeks for M2 except weeks 35, 39, and 42.
Three samples were available from M2 in weeks 9, 24, and 48. The average lower limit of detection (80% B/B o ) was 10.9 pg/ml. All biological samples assayed exceeded the lower limit of detection. Testosterone intra-assay variation was 7.7%; inter-assay variation was 5.7% for the 100 pg/ml control and 10.8% for the 25 pg/ml control.
Both males demonstrated seasonal variation in testosterone concentrations in blow (Table 4). M1 displayed a more gradual increase to approximately week 26, when testosterone began to gradually decline (Fig. 2). M1's testosterone remained elevated (greater than the mean value of all weekly observations, 107.0) from week 15 to week 36, with only two relatively low testosterone weeks during that time. M1's blow testosterone doubled between weeks 14 and 26. M2's blow testosterone concentration had a sharper peak that occurred between weeks 26 and 39 (greater than the mean value of all weekly observations, 97.0 pg/ml), with only one relatively low testosterone week occurring during that period (Fig. 3). M2's blow testosterone nearly tripled from week 22 to week 26. Testes Ultrasounds

M1 demonstrated seasonal variation in TTV, with peak TTV occurring in
February and an apparent plateau in TTV between weeks 22-33 (January 21 -April 7) ( Fig. 2). Total testicular volume remained above average (993.3 cm 3 ) from week 18 -38, with only one measurement below the average value during that period (April 21).
M2's testes remained below 375 cm 3 throughout the study period, and while the smaller TTV made a clear seasonal pattern less evident, in general the largest volumes were observed in March through May (Fig. 3). Measurements were unavailable for M2 prior to week 23. Total testicular volume data for each male is presented in Table   5.

Progesterone Assays
Once weekly blow samples were available for F1. Blood progesterone measurements were available from F1 for all months except for August 2013 and April 2014. Blow samples were commonly less than 55 µl for F2, or had volumes that were too small to re-assay the samples if the CV was above 20%. Thus, only 13/21 weeks were sampled for F2 ( Table 6). The average lower limit of detection (80% B/B o ) was 30.5 pg/ml. All biological samples exceeded the lower limit of detection.
Progesterone intra-assay variation was 11.7%; inter-assay variation was 5.8% for the 200 pg/ml control and 15.5% for the 50 pg/ml control.

Inferring ovulation in F1
Elevated progesterone was not observed in any blow samples collected from The third period of elevated progesterone in blow was longer in duration than the previous two, occurring between weeks 44 and 48. If the second non-conceptive cycle was immediately followed by a third cycle, ovulation would have occurred between June 12 -14 at the earliest, with regression of the CL occurring by approximately July 11-16. Blow progesterone did indeed remain elevated during this time until July 22, with the first low progesterone sample collected on August 1.
However, a low progesterone concentration was observed in the blood sample collected on July 3, close to when peak progesterone secretion should have been occurring from an ovulation occurring in mid June, decreasing the likelihood of an ovulation at that time. The low progesterone concentration measured in the blood sample on August 6 suggests that any subsequent luteal activity must have ceased by then. Using this date as an approximate cycle end date, ovulation could have occurred 29-32 days prior, between July 5 and 8, which is consistent with the low progesterone observed in blood on July 3, as well as the sustained elevated progesterone in blow through July 22. Thus, a third ovulation was inferred to have occurred between July 5 and 8 (weeks 45/46), with the associated follicular phase spanning weeks 43-45. This inferred ovulation also did not result in a pregnancy.

Behavioral Observations
Social interactions were observed in 197 of the 211 observation sessions and comprised 12% of the total time observed. A summary of the available data for analysis is presented in Table 7. A total of 45.2 minutes of interactions occurred in pools that lacked underwater viewing and thus lacked behavioral event data. Every possible social grouping was seen at least once. F2 was involved in just 7% of the recorded interactions, and there were no genital present behaviors presented toward her during the study. Therefore, any descriptions of courtship behavior will be in reference to F1. Interactions with F1 were comprised mostly of group swim interactions (by duration), while M1-M2 interactions were primarily milling interactions ( Table 8). The approaching whale was identified for 2124 of the interactions (89%). The separating whale was identified for 2086 of the interactions (87%). Of all male-female interactions, males initiated 95% and terminated 30%. Of the male-male interactions, M2 initiated 80% and terminated 42%.

Interobserver reliability
Observers #1-3 coded 57, 37, and 6% of the total interactions (by duration) that occurred in this study, respectively. The video used for determining interobserver reliability contained 9% of the total duration of interactions. The identity of the participants in an interaction had 100% agreement between all three observers.
Pairwise kappas calculated for each behavior are presented in Table 9. Melon push and melon flat had pairwise kappas less than 0.55, with a high rate of disagreement on whether the behavior occurred. These were also the most frequently coded behaviors, so pairwise kappas were calculated after these behaviors were removed from the matrix. Based on these results, ventral present, genital present posture, chase, head threat, and jaw clap were omitted from analysis. The extremely low frequency of chase and jaw clap contributed to the lack of agreement between observers. The aggressive behaviors bite, rake, and bite threat were condensed into a single aggressive category to improve agreement and allow further analysis. The following behaviors had "excellent" agreement (κ ≥ 0.80) for both observer pairings: open mouth, melon shake, genital present, and erection (although there was only one observation of an erection in the video used for interobserver analysis).

Seasonal Variation in COA
Interactions involving males and females occurred in every week of the year, and in 74% of the observation sessions. However, COA varied widely throughout the year, with 85% of the time of male-female interaction (811 minutes) and 96% of the total duration of group swims (682 minutes) occurring between January and June.
There was less seasonal variability in interactions that only involved males; 59% of the time of male-male interaction (285 min) occurred between January and June (Fig.   5). There were not large differences in COA for the groupings of interest in relation to F1's inferred follicular phases (Fig. 6).  Male interest in F1 generally occurred in short intervals, and appeared to be unrelated to testosterone concentrations, as much of the social interaction occurred after testosterone concentrations began to decline for both M1 and M2 (Fig. 7).
Coefficients of association >0.05 were observed for M1 in weeks 21, 26-27, and 33-35, and for M2 in weeks 21, 26-27, 31, 33-36, 38-41, and 46. The COA of M1 and M2 was influenced by the occurrence of the triad interaction between M1, M2, and F1, where M1 was generally observed group swimming with F1 while M2 milled nearby without engaging either M1 or F1. To remove this potential bias, the COA between M1 and M2 was calculated after excluding interactions that included a female. The resulting M1-M2 COA remained relatively consistent throughout the year, with two periods of increased socialization (COA > 0.10) in weeks 10 and 31-32 (Fig. 8).
The proportion of interactions by week initiated by a male was generally >0.9.
There were 4 weeks with >10 male-female interactions where the proportion of male approaches was less than 0.9 (F1 more likely to approach males): weeks 37, 39, 42, 45. The proportion of interactions terminated by a female was generally >0.5. There were 4 weeks with > 10 male-female interactions where the proportion of interactions terminated by a female was <0.5 (females less likely to end an interaction with a male): weeks 2, 22, 38, and 39.

Seasonal Variation in Genital Presents
A total of 541 genital presents were observed in the study, for an overall frequency of 0.04 per minute of observation, or .38 per minute of social interaction (once every 158 seconds of social interaction) (Table 10). Relatively few sessions contained a genital present, and 97% (270/278) of the male toward female genital presents occurred between weeks 26 -41 (Fig. 9).

Characterizing courtship
Observation sessions that contained a male toward female or female toward male genital present were investigated in greater detail in comparison to observation sessions that did not in an attempt to identify behavioral patterns indicative of courtship (Table 11). Genital presents tended to occur in sessions with high COAs between males and females, and 98% occurred during group swims. Of the genital presents from M1 to F1, 73% occurred within a social grouping of M1, M2, and F1, while only 3% of M2's genital presents occurred in this social grouping. No male toward male genital presents were observed in sessions when a male toward female genital present was also observed. All of F1's genital presents towards M2 occurred in sessions where M2 also displayed a genital present toward F1. Copulation was not observed in this study. Erections were rarely seen during male-female interactions (2 from M1, 1 from M2). Both of the erections observed from M1 occurred with a genital present, once on April 7, and once on April 25 (weeks 33 and 35).
Despite their small number relative to the total number of observation sessions, sessions that contained male toward female genital presents contained large proportions of the total occurrence of some behaviors, most notably group swim, open mouth, and melon shake (Fig. 11)

F1's Receptivity to Genital Present Displays
Genital presents were performed with approximately equal frequency toward F1's left or right side, and were rarely presented toward her ventral or dorsal surfaces, although M1 was more likely to present toward the ventral surface than M2 (Table   12). Males performed this behavior with their long axis parallel to the bottom or with their head angled down toward the bottom; there were no observations of this behavior with the male's head angled toward the surface. This behavior was also never performed by either male at the water's surface.
F1's receptivity toward genital present displays varied by male (Fig. 12). She was more likely to suspend active forward swimming during M1's displays, and was more likely to allow M1 to make contact during displays (Table 13). She was not observed to actively distance herself from any of M1's displays, but actively distanced herself from 21% of M2's displays. Receptivity during an inferred follicular phase was higher for genital presents by M1 (2.57 ± 1.00) than for M2 (1.45 ± 1.11), but receptivity with respect to inferred follicular phases did not vary within individual males (Fig 13). Mean receptivity for all genital present displays was higher for M1 when he performed a melon shake behavior in the same session (2.58, 4 sessions) than when he did not (2.08, 7 sessions). F1's receptivity scores for the genital presents from M1 with an erection were both 2. In both observation, F1 rolled her ventral surface away from M1 during the display, and contact did not occur with either of these presents.

Male Aggressive Behavior toward F1
Aggressive behavior (bite, bite threat or rake) was rare between M1 and F1, with only 27 events from M1 to F1 and 0 events from F1 to M1. All 27 events directed toward F1 occurred after week 28, by which point testosterone had already been in decline. More than half of the events directed toward F1 (14) occurred in one session (April 25) which was also the session in which genital presents from M1 to F1 were most frequent (Fig. 9). Aggression was more common between M2 and F1, with 51 events directed by M2 toward F1, and 8 events directed by F1 toward M2.
Aggressive events between M2 and F1 were approximately evenly distributed throughout the year (Fig. 14). Aggression frequency toward F1 was not correlated with blow testosterone concentration for either male (M1: F 1, 50 = 0.32, p > 0 .05; M2: F 1, 50 = 0.03, p > 0.05) (Fig. 15).   (Fig. 16). However, the rate of aggression from M1 to M2 per minute of observation was higher between January and June (0.048) than it was in July to December (0.026). M1's rate of aggression toward M2 was also higher when F1 was in an inferred follicular phase (0.052) than when she was not (0.033). M2's rate of aggression was higher in January through June (0.11) than it was in July through December (0.052), but there was not a clear difference when F1 was in an inferred follicular phase (0.084) or not (0.080).
Male toward male genital presents were more common in the fall and spring, but occurred in every month of the year except January (Fig. 16).

Physiological Assessments
The relationship between testes size and blow testosterone concentration in M1 in this study further validates the use of blow sampling as an indicator of reproductive function in belugas. Testes size remained elevated after testosterone began to decline in M1, consistent with previous observations in this species (Chapter 3, this dissertation). Although there also appeared to be a seasonal peak in M2's testicular volume, the sensitivity of ultrasound measurements may be insufficient to detect actual changes of those magnitudes, reducing the confidence with which seasonality can be assessed (Chapter 3, this dissertation). The linear measurements of M2's testes were outside of the 95% confidence interval for measurements from both immature and mature belugas examined by , and were approximately half the volume of the smallest adult male testes previously measured in aquaria (Chapter 3, this dissertation). Brodie (1971) found one animal with a testis volume of 130 cm 3 with evidence of spermatogenesis, but sexual maturity was not evident for all belugas until testis volume exceeded 360 cm 3 (TTV of approximately 720 cm 3 ). In the same study, the youngest sexually mature male had 14 growth layer groups in the cross section a tooth, which previously corresponded to an age of 7 years, but is now thought to correspond to an age of 14 years (M2 was 11-12 years old during the study) (Stewart et al. 2006). In a study of belugas in aquaria,  found that the youngest male to sire a calf was 9 years old, but that the mean age of first reproduction in male belugas was approximately 13 years. The seasonal pattern of testosterone, while present, also differed from M1 and from other males studied previously, with a period of elevation shorter in duration (Chapter 3, this dissertation). Therefore, in the absence of a conception or a semen sample for evaluation, M2 was likely immature during the study period. This suggests that testosterone concentration measurements alone are insufficient for identifying maturity status in male belugas.
The identification of estrous cycle stages in F1 was critical to interpreting the behavior of all individuals in the group. Unfortunately, ultrasonographic or urinary hormone conjugate data, which can be effectively used to characterize the estrous cycle in belugas , was unavailable for F1. Instead, progesterone measurements in blow samples were used to identify luteal phases and to infer the occurrence and stages of the estrous cycle. While the utility of progesterone sampling in blow for this purpose has been demonstrated for belugas, there is a risk of falsely identifying a reproductive condition based on a single sample due to the variable nature of dilution in each sample (Chapter 1, this dissertation). This necessitated the conservative standard for identifying luteal phases in order to reduce the likelihood of falsely identifying an estrous cycle. The first two inferred estrous cycles identified fit with the estrous cycle stage durations described by , and occurred when estrous cycles frequently occur . There also appeared to be behavioral correlates for these estrous cycles. The third estrous cycle identified was less clearly interpreted, and fell during a time when estrous cycles are less likely to occur . Although aquarium belugas most often have two estrous cycles in a given breeding season, there are observations of up to seven estrous cycles in a single year for one female (Katsumata et al. 2006). Increased frequency of blow sampling, or ultrasound examinations at key times to confirm or refute findings from blow sampling would aid in identifying reproductive events with greater certainty.
The lack of a confirmed conception from three estrous cycles in the presence of a male has been documented in an aquarium beluga previously (Katsumata et al. 2006). Two possible explanations for this observation are linked to ovulation mode.
While the mechanism for ovulation induction is unknown in belugas, in many species it is triggered by the physical act of copulation . If this is the case in belugas, then M2 could have induced ovulation without siring an offspring.
Alternatively, it is possible that F1 spontaneously ovulated in the absence of copulation. Belugas have been observed to ovulate spontaneously in the absence of a male, which  speculated may be related to self-stimulation by the female. F1 would also periodically engage in this behavior, perhaps triggering ovulation without copulation.

Behavioral Measures
The use of pair-wise behavior specific kappas revealed that several behaviors were subject to higher levels of disagreement. This emphasizes the importance of the use of these targeted evaluations of interobserver reliability in studies of behavior, so that high rates of agreement for more obvious behaviors do not mask poor agreement for others (Kaufman and Rosenthal 2009). Eliminating several behaviors in the intended ethogram did not affect the ability to assess seasonal variation in association, aggression, or courtship in this group.
Behavioral definitions should be altered in future studies. The open mouth display, which occurs in both aggressive and courtship contexts, should be accurately timed, as longer displays seemed to be associated with courtship. Bite threat is another behavior that might vary in different contexts, as evidenced by the reaction of the recipient. In aggressive contexts, this behavior appeared to be a threat of physical contact, or failed physical contact due to evasive action of the recipient. The movement toward the recipient in courtship contexts subjectively appeared slower and was associated with a less evasive response from the recipient relative to bite threats in aggressive contexts. Evasive behaviors performed by the recipient in response to bite threats may aid in interpreting this behavior.
The nature and contexts of voluntary movements of the melon require additional attention. The belugas in this study frequently altered the shape of their melons, a behavior that has been documented previously for this species DiPaola et al. 2007). Melon movements could possibly function as a behavioral modifier, altering the signal conveyed to the recipient by similar behaviors. For example, the belugas in this study frequently performed open mouth displays while altering the shape of their melon. Perhaps one shape is associated with an aggressive signal, while a different shape allows the open mouth display to convey an affiliative signal. The lack of agreement between observers in this study prevented such an analysis, but refined behavioral definitions could allow such a study to be performed with the existing video data.

Patterns of Intersexual Association
One pattern of behavior in this study that was similar to wild beluga behavior was sexual segregation, with intersexual association occurring almost exclusively between January and June. Several studies have documented sexual segregation of belugas in the summer and fall . Observations of male-female interactions in July through Dec were short in duration and rarely appeared to be affiliative, with a very low occurrence of group swims and genital presents. In estuarine habitats, female belugas appear to actively avoid groups of male belugas . The observations in this study suggest a functional sexual segregation despite continuous physical proximity. The period of high intersexual association observed in this study (January through June) corresponds with a time period that belugas are difficult to observe in the wild, although improving remote telemetry technology should enable further study of movement patterns during this time. Belugas typically do not enter more readily observable nearshore estuarine habitats until late June or July in most regions (Hornby et al. 2016;Richard et al. 2001), emphasizing the importance of behavioral observations in aquaria earlier in the year.
While relatively high levels of intersexual association occurred when estrous cycles are most likely to occur , there did not appear to be differences in association patterns during F1's inferred follicular phases, when she would be expected to be most receptive. The uncertainty associated with identifying estrous cycle stages in this study could account for this observation. However, if female receptivity was the only factor influencing patterns of intersexual association, even shorter periods of relatively high intersexual association might be expected given the rarity of intersexual associations outside of the breeding season. Instead, the observed pattern of association might indicate a relatively long courtship period that lasts longer than the time of peak receptivity from the female. This may have occurred due to the lack of additional receptive females for the males to interact with, or in the case of M1, serious adult male competitors. However, if this is a consistent pattern of behavior in belugas, longer courtship could allow ovulation-inducing males to thwart breeding attempts by other males, while females prevent suboptimal males from breeding with her while she is most fertile by choosing to associate with a higher quality mate. This may also provide the female with opportunity to evaluate potential mates, as induced ovulation likely reduces the opportunity for postcopulatory selection . Relatively long periods of consortships have also been observed in humpback whales and bottlenose dolphins .
Intersexual interactions were typically initiated by males and terminated by females in this group. This pattern was strong and consistent throughout the year.
Two of the weeks where F1 was more likely than normal to approach the male occurred close to her inferred second and third ovulations. This suggests that periods with a reversal or reduced strength of the typical pattern may indicate important changes in behavior, and thus this behavior should be a component of future research.

Courtship Behavior
The strong seasonal nature of intersexual genital presents in this study that corresponds with the breeding season suggest a reproductive function for this behavior, as opposed to a more social function, although socio-sexual behavior for non-reproductive purposes is common in odontocetes . Therefore, observation sessions containing male toward female genital presents were considered to be observations of courtship in this social group. Copulation was not observed to confirm this assumption, but copulation is rarely observed in belugas in aquaria . This contrasts with other species of odontocetes in both aquaria and the wild, where copulation rates can be high . This could be a reflection of beluga mating strategy, as lower copulation rates might be expected for induced ovulators depending on a male's ability to both induce ovulation and ensure paternity.
As proposed by , group swimming served an important function in beluga social interactions observed in this study, particularly with intersexual associations. Group swimming was relatively rare, occurred almost exclusively during the breeding season, and was concentrated primarily in the observation sessions that also contained male toward female genital presents. Genital presents also almost exclusively occurred during group swim interactions. Group swimming was not necessarily synchronous, as described in other species (Connor et al. 2006), and it was clear that either the male or the female could lead the direction and pace of swimming during a group swim. Swimming in this manner requires cooperation from all participants, and could thus be used to assess mate choice in female belugas. In this study, both males were frequently simultaneously observed in a group swim with F1. However, courtship behavior during these triad interactions were almost exclusively between M1 and F1, with M2 apparently observing these interactions and rarely displaying toward F1 during them. With both males available for interaction at the same time, this is perhaps indicative of F1's choice to interact with M1 as opposed to M2 at those times. The close association between group swimming and genital presents further suggests that the female can select which males might be able to display toward her by choosing who to swim in close association with. This behavior is obviously constrained in aquaria, but its potential function in mate choice merits further study.
The genital present described in this study is similar to the pelvic thrust behavior described by  and a combination of the horizontal "S" posture and pelvic thrust behaviors described by . The behavior was described differently in this study in order to: more clearly define the body posture of the actor, remove any orientation restrictions on the occurrence of the behavior, identify the necessary presence of a recipient, clearly separate this behavior from copulation, account for the cessation of active swimming, and reduce the implication that the behavior is forcible by eliminating the use of the term "thrust," because the genital present was generally slow and deliberate in the study group. The rarity of erections during this behavior suggests it has more of a display function in courtship as opposed to a direct relationship with copulation.
The overall frequency of genital present displays was lower than in , which may be due to the differences in social group size or composition in these two studies.  also found that the occurrence of this behavior had a strongly seasonal pattern, with a peak occurring in March. In this study, the frequency of genital presents peaked later, perhaps as a reflection of the estrous cycle timing in F1. Male belugas would likely adjust their courtship behavior to match the periods of receptivity of females. M2 performed more genital presents, but M1 performed a much larger proportion of the observed genital presents within F1's inferred follicular phases. This could be the result of F1 preferentially choosing to swim with M1 during these times, enabling increased displays. As an adult and proven sire, M1 may also have been more able to detect F1's receptiveness during these periods, and thus concentrated displays during this time. It is interesting that M1's genital present behavior toward F1 varied in frequency with the occurrence of the follicular phase, but association did not; this further supports importance of prolonged courtship in this pair of belugas.
In addition to selecting which male to swim with and thus allow genital presents from, F1 could further control breeding opportunities by behaving differently in response to genital presents, and presumably copulation attempts as well. The slow, deliberate pace of the genital present, as well as the associated termination of active forward propulsion by the displaying male leaves clear opportunity for the female recipient to respond in a way that either allows contact to occur or not, as well as where on the body that contact can occur. In this study, F1 was observed to variably alter her pace or body orientation in response to genital presents, allowing the quantification of receptivity. The body position of belugas during copulation has not been formally described. Here, we assume that copulation must be performed in a ventral to ventral position, as in other species of odontocetes , and thus locomotion and orientation patterns that would lead to contact on the ventral surface were considered to be associated with a higher level of receptivity. F1 displayed higher receptivity toward M1 than M2, allowing a far higher rate of contact and only displaying active avoidance in response to M2's displays. Some of F1's responses to genital presents can be considered analogous to the evasive behavior displayed by wild dusky dolphins (Lagenorhynchus obscurus) that variably roll their body away from males attempting to copulate . On one occasion, F1 was observed to drift during a genital present from M1 with the flukes curled dorsally. Fluke posture could be explored as a further indication of receptivity.
The prevalence of open mouth and melon shake displays during observation sessions with male toward female genital presents suggests a role in courtship for these behaviors. Open mouth displays and mouthing have previously been suggested to have socio-sexual functions in addition to agonistic functions in belugas . The melon shake is especially interesting, as it almost exclusively occurred in courtship contexts and is clearly distinct from behaviors that occur in other contexts.
There is also some evidence that melon shake displays from M1 were associated with higher levels of receptivity from F1, although this was not the case for M2. The melon of a beluga is remarkable in comparison to all other odontocetes due to its mobility, discussed above in the context of the difficulty of categorizing the range of movements possible. The melon is described as sexually dimorphic in belugas, with males having broader melons than females (Martin 1996). The primary function of the melon is to focus echolocation signals (Cranford et al. 1996), and focusing within the melon could aid in directing echolocation signals in this species (Penner et al. 1986). Altering the shape of the melon may also be associated with creating "facial expressions" . With the prominent role that melon movement behaviors appear to have in social interactions and especially courtship, perhaps this sexual dimorphism is a result of sexual selection by females for males with larger, broader, and thus more expressive melons. The differently shaped melons in males could also be related to sexually selected acoustic displays, as may be the case in sperm whales (Physeter macrocephalus) (Cranford 1999  . It is possible that M2 observed courtship between M1 and F1 as part of a social learning process, and then continued to associate with F1 after M1's interest had waned. This is reflected in the shared peaks of association when courtship behavior from M1 was frequent and a lack of shared peaks of association when M2's courtship behavior was high. F1's choice to interact with M2 during these times is less clearly explained, although socio-sexual interactions have been observed between adults and juveniles in previous studies of belugas . Perhaps the most striking observation through the paired study of physiological and behavioral measures is that the majority of presumed courtship behavior in this social group occurred after peak testosterone and for M1, at a time when testes size was about to decline. This finding agrees with the observation that most conceptions occur after peak testosterone has been reached, and testicular volume is declining in size Chapter 3, this dissertation). This is in contrast with many species of mammals, including spinner dolphins (Stenella longirostris), where testosterone is associated with breeding behavior in males . This also contrasts with a study of finless porpoises (Neophocaena phocaenoides asiaeorientalis), where breeding behavior was correlated with testes size  Aggression was not an important component of courtship in these belugas, and in general, aggression was mild in this social group. In several species of odontocetes, including the beluga's closest living relative, the narwhal (Monodon monoceros), aggression that results in physical injury from teeth is common . In this group, aggression did not result in a single rake mark throughout the study. There was no evidence that F1 was coerced into interacting with either male, as occurs in bottlenose dolphins (Scott et al. 2004). The relatively higher rates of bites, bite threats, and rakes that occurred during observation session with genital presents seemed to be associated with gentle mouthing that has previously been described as socio-sexual in belugas.
This study is not considered to be a The study of additional males will allow for more comparisons, and perhaps even an exploration into the effect of social learning on courtship behavior, as males that are housed together may show similar behavior to each other that may be different from other social groupings.

Male-Male Social Interactions
Direct observations of wild belugas, as well as data from telemetry studies, suggest that male-male social relationships may be important in this species, at least for some parts of the year. In this social group, association between the males did not vary by season, in contrast to associations between males and females. Despite the extreme seasonality seen in genital presents toward F1, male toward male genital presents occurred throughout the year, indicating a potential social function of this behavior. Even though they were of different reproductive conditions, both males displayed aggressive and presumably affiliative behavior (genital presents) toward each other, often in succession. There was not a clear dominance/submission relationship between the two; in fact aggression was more commonly displayed from M2 toward M1. Some of the seemingly aggressive behavior may have been related to play or socio-sexual behavior, which is common between males in other odontocetes (Mann 2006). M2 was more likely to initiate interactions and M1 was more likely to terminate them, which may be related to play behavior in M2 and the reduced willingness of M1 to engage in such behavior given his age (Hill and Ramirez 2014 (Brodie 1971), implying that at least young calves would have an opportunity to observe courtship and mating.
Interestingly, male toward male genital presents were more common when F1 was in an inferred follicular phase. If male contest competition was an important feature of reproductive behavior, the opposite might be expected. In a highly competitive environment, a male engaged in courtship behavior would also be expected to be intolerant of the proximity of other males, yet M1 was very tolerant of M2's presence during courtship bouts with F1. Although rates of aggression between the males were higher during the breeding season, the relatively low level of aggression in both frequency and intensity between M1 and M2 suggests that competition for mates was minimal in this group. M1 may not have identified M2 as a threat due to M2's immaturity, and thus did not need to expend energy in competitive behavior. Further study of groups with multiple adult males is needed for comparison.

Conclusion
Despite the limitations of studying behavior in a small group of aquarium housed animals, several observed patterns were consistent with the limited knowledge of social behavior in wild belugas, including seasonal association patterns between males and females resulting in functional sexual segregation, and the relatively greater importance of male-male social relationships. The near absence of interaction between males and females outside of the breeding season suggests that the observed pattern had a reproductive function, as opposed to socialization in general.
Periods of higher intersexual association were not necessarily confined to inferred follicular phases, suggesting relatively long courtship periods. These associations contained a variety of behaviors, including some that may have multiple functions depending on context, and required a high degree of female cooperation.
These relatively long, complex interactions provided multiple opportunities for males to display toward the female, and for the female to actively respond to such displays by choosing whether or not to swim with the displaying male as well as choosing how to respond to behavioral displays. This observation is also consistent with predictions of reproductive behavior based on ovulation mode. The relative importance of these behavioral displays in pre-copulatory selection could account for the apparent low investment in pre-or post-copulatory traits in belugas found in an analysis of sexual selection in cetaceans .
The relatively low copulation rate is also consistent with the relatively small testes found in this species, as well as the relatively small seasonal variation in testes size (Chapter 3, this dissertation). Further study of groupings with multiple adults of both sexes will aid in assessing the relative importance of postcopulatory selection (sperm competition) in belugas. The apparent reduced association between measures of male reproductive physiology and courtship behavior is consistent with previous observations of male reproductive seasonality and the timing of conceptions ; Chapter 3, this dissertation). This suggests that signaling from the female indicating receptivity plays a more important role in eliciting courtship behavior from the males than internal physiological cues.
Despite its limitations, blow sampling enabled a greater understanding of the reproductive physiology of the belugas in this study than would have otherwise been possible. Blow sampling is a comparatively easy, non-invasive way to assess reproductive physiology and it involves less intensive training than urine collection, blood sampling, or ultrasound examinations. This methodology will greatly facilitate paired studies of behavior and physiology of belugas in aquaria. With further development, blow sampling also has greater potential for application in wild belugas than more invasive forms of physiological assessments. However, personal experience with multi-male social groups in an aquarium setting suggests that male-male aggression may occur less frequently than expected. Adult males in the wild associate closely with one another and often form their own samesex pods Barber et al. 2001). Adult males in aquaria may form social bonds that are readily apparent (pers. obs.). These associations could have a reproductive function, as in wild bottlenose dolphins (Tursiops truncatus), which may form male alliances to sequester and guard mates (Randic et al. 2012). Alternatively, the sexual dimorphism observed in belugas could be unrelated to mating strategies, stemming instead from phylogenetic relationships or ecological drivers (González-Suárez and Cassini 2014). Larger males are able to exploit different habitats than females, perhaps reducing intraspecific competition in a habitat that may be nutrient poor for part of the year . Unraveling the selection mechanisms that drive sexual size dimorphism requires an understanding of the relative importance of body size in beluga mating strategies.
Sperm competition may also be a mechanism for males to compete or for females to employ mate choice. Relative testes size can be used to infer the importance of sperm competition in the mating strategy of mammalian species . In species such as the harbor porpoise, relatively large testes (4% of the body mass) suggest the importance of sperm competition . In belugas, sperm competition is less likely to play a dominant role in male reproductive success; the testes account for ~0.1% of body mass (calculated from data in Kleinenberg et al. 1969), which places them among the smallest ratios in odontocetes studied thus far Mesnick and Ralls 2002). However, testes size has primarily been assessed in the summer, when carcasses are readily available but breeding does not occur  This may be important for species that are often segregated sexually in highly seasonal environments, such as the beluga (Barber et al. 2001;. Induced ovulation is also expected to lead to some degree of mate guarding during the short period between copulation and fertilization

Monitoring Reproductive Function
In order to sufficiently test these predictions, it is necessary to accurately describe the reproductive condition of an individual at the time that behavior is observed. Gonad function, measured through testes size or follicular development, is a clear indicator of an individual's reproductive condition and is often used to assess maturity and reproductive condition in adults  Therefore, studies that correlate steroid hormone concentrations and behavior in wild animals are common across taxa, including several species of marine mammals . Studies that incorporate measures of gonad function, hormone concentration, and behavior are also feasible in some wild terrestrial mammals. For example, a study of free-ranging Soay sheep correlated scrotal circumference, testosterone concentrations in blood, and breeding behavior (Preston et al. 2012). However, research on hormonal correlates of behavior among cetaceans has been particularly sparse. Although methods are available for determining hormone concentrations in the feces and blubber of wild cetaceans (Hunt et al. 2013), opportunities have been limited because individuals are exceedingly difficult to monitor for long periods in their natural marine environment and behavioral observations are generally limited to surface observations.
Studying cetaceans in zoological facilities would alleviate many of these challenges. Belugas are already common research subjects; most of the available information about beluga reproductive physiology, parturition and calf rearing has been elucidated by studying trained belugas in zoological facilities , Russell et al. 1997 testosterone concentration could be correlated with an increase in the frequency of certain breeding behaviors .  were able to correlate the frequency of reproductive behavior (contacting erect penis to the genital region of another animal) to both testicular volume and blood testosterone concentration in a captive group of finless porpoises (Neophocaena phocaenoides asiaeorientalis).  were able to assess the estrous cycle stage of bottlenose dolphins (Tursiops truncatus) and identify behaviors associated with estrus in this species. However, the physiological measures used for these studies were collected infrequently and/or omitted samples from one of the sexes. Future studies in zoological facilities would benefit from sampling both sexes more frequently. Studies of wild cetaceans would become more feasible if a method of non-invasively monitoring steroid hormones became available.

Blow as a method of non-lethal sample collection from cetaceans
Due to the legal challenges facing cetacean researchers, many have worked to develop non-lethal methods for acquiring tissue from free-ranging cetaceans for genetic and hormone analyses, most of which require firing a biopsy dart into the animal (Noren and Mocklin 2012; ). Recently, blow sampling has emerged as a less invasive alternative for studying wild cetaceans. Blow samples were first used to determine testosterone concentrations in trained dolphins by Hogg et al. (2005). Since then, testosterone, progesterone, or cortisol has been assayed in the blow of six different species of cetaceans, including cortisol in belugas (Hunt et al. 2014; Thompson et al. 2014;Hogg et al. 2009;Miller and Hall 2012). The presence of epithelial cells in a blow sample also allows for genetic analysis of the individual, a technique that has been developed for trained bottlenose dolphins and harbor porpoises (Phocoena phocoena) (Frère et al. 2010;Borowska et al. 2014). While these studies highlight the potential of blow sampling as a research tool, there are factors that currently limit the application of the technique. Genetic studies required 4-8 exhales from the same individual to accumulate sufficient cellular material, an unlikely scenario under field conditions (Frère et al. 2010;Borowska et al. 2014).
Many individuals sampled for hormone analysis are of unknown gender or reproductive condition or belong to species with unknown endocrine cycles (Hogg et al. 2009). More importantly, reproductive hormone concentrations in blow have not been compared to those circulating in the blood for any species, preventing an understanding of the biological significance of hormone concentrations in blow.
The hormone cycles of belugas are well understood , and the availability of trained individuals of known gender and reproductive condition in zoological facilities offers an opportunity for the advancement of this methodology. Hormone concentrations measured in blow could be compared to those in matching blood samples to establish a correlation. Due to the variable amount of water vapor present in each blow sample, it is unlikely that the hormone concentration in blow will be an exact measure of the hormone concentration in circulation. However, given the 5-10 fold difference in circulating testosterone concentration in adults of both sexes in belugas , it is reasonable to hypothesize that a biologically relevant threshold concentration of testosterone can be established, above which all animals sampled are male. The dramatic difference in progesterone in pregnant and non-pregnant female belugas may allow for a similar technique to be used to separate pregnant from non-pregnant females (Calle et al. 1993;Stewart 1994). A threshold may also be identified for mature adults of both sexes, as explored by  with reproductive hormones in beluga blood. These authors proposed a testosterone threshold of 3.30 nmol/l for identifying mature males; a similar predictive value may be identified for hormone concentrations in blow.
Previously, liquid chromatography-tandem mass spectrometry (LC-MS/MS) has been utilized to detect the presence and concentration of testosterone and progesterone in blow from other species, due to the small sample volume of a single exhale (Hogg et al. 2005). However, this technique requires prohibitively specialized equipment and may require advanced separation methods to provide the necessary sensitivity for measuring concentrations (Dunstan et al. 2012;Miller and Hall 2012).
Validating enzyme immunoassays (EIAs) for the measurement of reproductive hormones in blow would greatly improve the applicability of the technique in aquarium managed, live-stranded, or temporarily restrained wild cetaceans, where multiple exhales can be collected. Species with more voluminous exhales may produce enough fluid for EIA analysis of a single exhale; as one of the largest odontocetes, belugas may be one such species. EIAs are already the standard for measuring steroid hormones in beluga blood, and have been validated for use with blow samples collected from right whales (Eubalaena glacialis) and for the measurement of cortisol in beluga blow samples (Hunt et al. 2014;; Thompson et al. 2014).
Our understanding of wild beluga population dynamics would benefit from the development of this research tool. Hormone and genetic sampling would be simplified in endangered populations such as the Cook Inlet stock, where live strandings are common (Balsiger 2003). The care of belugas in zoological facilities would also be improved by reducing reliance on blood samples and enabling increased sampling frequency for research or medical purposes. If the utility of blow sampling is validated in belugas, method development would be fostered in other species whose conservation status necessitates non-invasive sampling.
The principal aims of this work are to develop methodologies that can be used to assess beluga reproductive condition in known individuals, and reproductive condition and gender in unknown individuals, and then utilize these methodologies to simultaneously assess physiology and behavior in a social group of belugas in an aquarium setting. The results of this study will be interpreted in the context of the current understanding of beluga reproductive biology in an effort to complement and inform studies of wild beluga behavioral ecology. Additionally, establishing the value of blow sampling as a minimally invasive tool to assess reproductive condition will allow this type of study to become more feasible for both wild and aquarium-managed belugas.
The aims of this dissertation are presented below.

SYNTHESIS AND SPECULATIVE DISCUSSION
Improved understanding of the mating system or sex-specific mating strategies can improve species management and allow for more robust predictions of how a population might respond to perturbation. The mating strategies of belugas have not been determined with confidence because direct observations during the breeding season are not logistically feasible. Paternity studies, which can provide information on reproductive success in the absence of direct observation, are also lacking in this species. Despite the lack of data, several authors have attempted to infer beluga mating strategies from morphological or demographic information because this information provides context for the numerous studies of beluga population genetics and is important for the management of this species.
Among the features used to infer mating strategies in belugas are sexual size dimorphism, testes size, and the operational sex ratio. In this speculative discussion, I aim to evaluate these features in the context of the results of this dissertation as well as other recent observations relating to the reproductive biology of belugas, in order to develop new hypotheses that can be explored through studies in the wild and in zoological facilities to improve the understanding of beluga mating strategies.

Sexual size dimorphism and male contest competition
Sexual size dimorphism is considered to most often be associated with sexual selection (Shine 1989). In species in which males are the larger sex, this typically indicates a polygynous mating system with contest competition between males, where larger body size confers a competitive advantage for males. As such, body size has been proposed to indicate contest competition in belugas . However, the degree of sexual dimorphism varies among populations in belugas (Doidge 1990).
Given the latitudinal cline in feeding ecology found in belugas (Yurkowski et al. 2016), it is possible that the observed pattern of sexual dimorphism may be driven by feeding ecology (Shine 1989;Gonzalez-Suarez and Cassini 2014;).
In marine mammals, large body size confers a thermoregulatory advantage as well as increased diving capacity due to increased oxygen storage ability, primarily in muscle (Berta and Sumich 2003). Larger male belugas are known to travel through areas that have greater ice coverage than smaller female or juvenile belugas . Increased diving capacity and thermoregulatory capacity due to larger body size may allow male belugas to reach feeding areas that are not attainable by other conspecifics, perhaps reducing intraspecific competition. These food sources would be expected to be high in quality given the degree of risk and energy invested in these movements ). If so, this high quality prey would support the observed dimorphism. In addition to limitations due to body size, female belugas are also limited in their ability to penetrate areas with heavy ice cover because they are accompanied by calves with limited dive capacity, forcing them to take fewer risks in their movements among heavy sea ice . Females are also unable to invest as much energy to growth as males, as they mature at an earlier age and bear the high costs of lactation . These factors could lead to the development of sexual size dimorphism in the absence of strong sexual selection (Shine 1989).
Variation in diet between the sexes, extent of sexual segregation, and sexspecific diving behavior could be compared to the degree of sexual size dimorphism observed within a group of belugas to determine the relative contribution of feeding ecology to sexual size dimorphism in belugas. Populations with a lesser degree of sexual size dimorphism might be expected to have less sex-specific variation in diet, less exaggerated sexual segregation in space and time, and fewer differences in sexspecific diving (i.e. foraging) behavior. Some observations are available that seem to lend support to this hypothesis; in lower latitudes, where food may be more readily available, females and males showing similar dietary patterns and apparently reduced levels of sexual segregation in space, but not time (Yurkowski et al. 2016;. If sexual size dimorphism is primarily driven by natural selection, then it would be inappropriate to infer high levels of male contest competition in belugas because of the presence of sexual size dimorphism.
If contest competition is important in belugas, then males might be expected to travel separately from each other during the breeding season in search of receptive females, yet at least in the summer and fall, male belugas are typically found in groups Cobeck et al. 2012). In sperm whales, roving males sexually segregate from females outside of the breeding season to feed in more productive areas, and then rove between pods of females, forming temporary consortships . In the absence of coercion or herding behavior , there would be no fitness benefit to travelling in a group of adult males, as doing so would put an individual male in direct competition with its social partner(s). Malemale associations outside of the summer and fall are largely unknown for belugas.
Male movement patterns in relation to females might be explored using improved telemetry technology to determine if males travel more widely (presumably in search of females) during the breeding season. Interbreeding between stocks is known to occur in belugas while sharing wintering grounds, supporting the concept of males that rove between groups of females in search of mates . If possible, males found associating closely in summer would be fitted with telemetry devices simultaneously to further explore the persistence of these associations.  Harwood et al. 2002) may provide another way to assess male-male contest competition. If sex is not associated with varying longevity, then male contest competition is likely less important in this species. If males live longer than females, it may support the alternative hypothesis that sexual size dimorphism is driven by natural selection. Larger males may be able to store more blubber or have more plastic foraging behavior due to increased dive capacity, and thus be more likely to survive during times of nutritional unpredictability, improving longevity. Although age structures were not different between the sexes, the data presented by Harwood et al. (2002) from one beluga stock seems to suggest that males may reach older ages more frequently than females despite selective hunting that preferentially removes males from the population.

Relative testes size and sperm competition
Measuring relative testes size offers great insight to the mating system of mammals, with larger testes conferring an advantage to males in multi-male systems (polyandry or polygynandry) where sperm competition is relatively high . Two studies have attempted to use testes size in belugas to infer mating strategies. In , belugas were inferred to have a more promiscuous mating system in comparison to narwhals, given their larger relative testes size and rapid testicular growth at sexual maturity. Although sperm competition may be more prevalent in belugas than it is in narwhals,  considered belugas to have low investment in postcopulatory traits, indicating relatively low levels of sperm competition compared to other cetaceans. However, as the work of  used the data from , there was a lack of sampling during beluga breeding season, with only 2 individuals assessed between December and April.
Considering testes size is approximately 50% larger during the breeding season (Chapter 3), both studies may have underestimated the relative importance of postcopulatory selection in belugas.
Comparing relative testes sizes between cetaceans is not always clear because testes size does not increase allometrically in this group (MacLeod 2010). Therefore, in the absence of additional statistical modeling, comparisons between species of similar body size, as conducted by , would be most informative.
The closest species included in the analysis conducted by MacLeod (2010) was the pilot whale, which has testes that account for 0.31% of the body mass, approximately 3 times the percentage in belugas (mean = 0.07% reported by  allowing for a 50% increase during the breeding season would increase the mean to approximately 0.11%). Using the regression developed from a study of a wide variety of mammals in Kenagy and Trombulak (1986), a 1000 kg beluga would be expected to have a testes mass that accounts for approximately 0.5% of their body mass (acknowledging that this is only a rough estimate given the lack of allometry in this trait among cetaceans). Therefore, even after accounting for a 50% increase in testes mass during the breeding season (Chapter 3), the observed testes size in belugas is still relatively small, perhaps for mammals in general. The degree of seasonal change is also small relative to other species of odontocetes (Chapter 3). These data suggest a lower demand for sperm in belugas when compared to other cetaceans, indicating that sperm competition is less important for belugas than other cetaceans, in concordance with the view presented by . While  found that sperm competition was likely more important for belugas than it is for narwhals, it is possible that narwhals are an extreme case among cetaceans, with the characteristic tusk of male narwhals allowing for greater levels of polygyny than is generally possible among cetaceans, resulting in smaller testes sizes. The finding that the narwhal tusk may serve as an honest indicator of male quality in narwhals ) supports this suggestion and merits further study.

Other indicators of the relative importance of sperm competition in belugas
Sperm competition theory makes several predictions about male reproductive physiology in addition to selection for larger testes . Therefore, the current knowledge of beluga anatomy, physiology and behavior, revealed through this study and others, can be evaluated to help determine the relative importance of sperm competition in belugas in the absence of direct observations. These suggestions are tentative, as other factors can influence these traits, and different groups of mammals can respond differently to similar pressures (Gomendio et al. 2011). Due to lack of data, sperm morphology and female reproductive anatomy (genital tract length and complexity, for example) will not be considered here, but are promising areas of future research in relation to mating strategies (Miller et al. 2002;Plön and Bernard 2006;).
Higher levels of sperm competition are associated with higher testosterone levels among primates (Dixson and Anderson 2004). Adult male belugas have relatively low circulating testosterone concentrations. In this study, belugas in aquaria had testosterone that ranged from 0.2 -6.9 ng/ml, which was similar to the range reported by  for aquarium belugas and Høier and Heide-Jørgensen (1994) for wild belugas, although measurements are unavailable for wild belugas in the breeding season. This is considerably lower than maximum testosterone concentrations measured in species where sperm competition is thought to be more important: harbor porpoises (30 ng/ml, Desportes et al. 2003), Dall's porpoises (20 ng/ml, Temte 1991), and bottlenose dolphins (54 ng/ml, . The observed range is similar to the killer whale , a species where sperm competition is thought to be less important ). In addition to suggesting a reduced importance of sperm competition, relatively low testosterone concentrations also support the suggestion that male contest competition is not an important mating strategy in this species ).
Ejaculatory frequencies are also higher in primates with more promiscuous mating systems . This study and others  suggest that copulatory frequency, and presumably ejaculatory frequency, is low for belugas. Rates of copulation can be much higher among other odontocete species with larger relative testes sizes . The low observed copulation rate in belugas also corresponds with the apparent low sperm demand for this species.
Sperm competition is also associated with greater numbers of motile sperm per ejaculate  representative, this would be consistent with the observation that there is a low demand for sperm in this species, and would imply a reduced role of sperm competition.
Taken together, there are several lines of evidence that suggest that sperm competition is not an important reproductive strategy in belugas. Certainly these observations do not rule out the potential for sperm competition, they only suggest a reduced role of this strategy relative to other species. Additional insight would be gained through more detailed investigations of the female reproductive tract, particularly the vagina, where sperm competition would primarily be expected to occur . Molecular investigations into paternity, such as those accomplished in the killer whale (Ford et al. 2011), would be particularly informative. Kinship studies are already common in belugas (e.g. Colbeck et al. 2012). This type of study may be facilitated in wild belugas, especially within small populations, through the use of minimally invasive blow sampling as a source of DNA for analysis (Chapter 2).

Operational sex ratio and competition for mates
One demographic feature that is predicted to have a strong influence on mating system is the operational sex ratio (OSR). This measure is different from the actual sex ratio (ASR) in that it quantifies the ratio of sexually available females to males (Reynolds 1996). Female belugas breed in two to three year intervals ).
Therefore, there would be two to three available females for each breeding age male beluga in a given population. Male belugas also have no role in parental care , so remaining with any one female long term does not carry a fitness advantage. This lack of male parental care and highly male skewed OSR would be expected to be associated with intense competition between males for mates . However, the presumed competition need not be confined to contest competition. Sperm competition via multiple mating, scramble competition between males to find receptive females, and female mate choice are other mechanisms of male-male competition . If contest competition and sperm competition are relatively unimportant male strategies, then perhaps scramble competition or female mate choice are of greater importance.
Male belugas have been observed to travel together outside of the breeding season in groups of 8-10, while groups of females with dependent young travel in much larger groups . If belugas maintain these group sizes during the breeding season, and belugas are widely dispersed during the breeding season (low population density), then conceivably these small groups of male belugas may rove between groups of females during the breeding season . The ability to find and consort with groups containing receptive females (scramble competition) would have a fitness advantage. Then, in any given encounter between such groups, there would be expected to be a greater number of receptive females relative to adult males, reducing the need for contest competition between these associating males to gain access to mates.
Once an encounter occurs, females may further ensure that a given male is a high quality mate by mating with multiple males and encouraging sperm competition and/or employing precopulatory mate choice. The apparently low copulation rate in the groups of belugas studied thus far, as well as the prolonged and complex courtship behavior observed in this study (Chapter 4), suggest that precopulatory mate choice may be an important strategy in this species, perhaps in addition to some degree of scramble competition. These strategies would still allow for a polygamous mating system, with both females and males breeding with multiple mates over a given breeding season, given the relatively long duration of the breeding season and the fact that female belugas are seasonally polyestrous . Observations of social behavior in wild belugas during the breeding season, perhaps inferred through the use of telemetry, are needed to evaluate these suggestions.

Theoretical effects of induced ovulation on mating strategies
The relative unimportance of contest competition or sperm competition as mating strategies in belugas are in concordance with theoretical predictions for the influence of ovulation mode on mating strategies. For female induced ovulators, copulation is required to trigger ovulation . Induced ovulation is thought to reduce post-copulatory competition in mammals because the first male to copulate is the most likely to induce ovulation and sire the resulting offspring . Thus, demand for sperm would be reduced, in agreement with the findings presented in Chapter 3. Males might be expected to require fewer sperm per ejaculate, and to copulate less frequently, assuming one copulation is sufficient to induce ovulation .
These predictions are also consistent with the current understanding of beluga reproductive biology.
Induced ovulation is thought to be selected for when males and females are less likely to come into contact, as mature follicles persist for long periods in induced ovulators prior to copulation. This reduces the need of both sexes to be in the same place within a narrow fertile period. If this is the selection mechanism that resulted in induced ovulation in belugas, it might indirectly indicate the rarity with which groups of females encounter groups of males, perhaps further emphasizing the importance of scramble competition relative to other strategies in this species.
The lengthy estrus would also create sufficient time for females to employ precopulatory mate choice. The relatively long periods of association between males and the female, as well as frequent behavioral displays by males toward females in this study provided ample opportunity for the occurrence of mate choice. The variable response of the female to these association attempts and displays further supports the occurrence of mate choice (Chapter 4). Females may also employ postcopulatory mate choice in the absence of sperm competition through ovulation induction mechanisms that require threshold levels of stimulation for ovulation to occur (Lariviere and Ferguson 2003).
The ability of the first male to copulate with a female to monopolize paternity in induced ovulators also suggests that there would be a fitness advantage for males to be associating with a female when she first becomes receptive . In species with a relatively diffuse breeding season, like the beluga, males might be expected associate with females prior to estrus. The observations in this study provide some support for this suggestion (Chapter 4). A short period of postcopulatory association may also be expected, both for males to thwart breeding attempts that might result in sperm competition, as well as for females to ensure only the selected male has an opportunity to copulate. This study also provided some evidence of this, with periods of exclusivity of particular male-female associations close to ovulation (Chapter 4).
Therefore, the combined data on reproductive physiology and behavior from this species are in accord with predictions that follow from the presence of induced ovulation in belugas, a condition that is thus far unique among odontocetes. These mating strategies would also coincide with the wealth of genetic information available for this species (e.g. O'Corry-Crowe et al. 2010). They also do not fundamentally challenge the inferred mating system of the species, which is generally considered to be some form of polygamous system . Given the limited ability for direct observations, these suggestions are tentative but may provide further guidance in interpreting studies of wild belugas.

Future studies of belugas in aquaria
This discussion highlights the difficulty in trying to infer the mating system of the beluga in the absence of direct observations. Although the application of studies of behavior in zoological facilities to wild populations can be limited, there are some cases where behavioral observations of managed groups can inform wild studies . Evaluating mating strategies in belugas may provide one such example, especially because direct observations of belugas during the breeding season are not logistically feasible. Studies in aquaria could therefore fill knowledge gaps that are unlikely to be filled with direct observations of wild belugas. Indeed, several observations in this study agree with the current understanding of beluga social behavior in the wild, namely sexual segregation outside of the breeding season, the strong seasonality of courtship behavior, and male-male association patterns (Chapter 4).
Studying existing groups with multi-male/multi-female social composition would allow for further exploration of the topics addressed here. Key issues to study in such groups would be: the timing of follicular phases within a group of females to determine the degree of synchrony, and thus degree of competition that would be expected for a given female; male-male aggression during receptive periods to further assess male-male contest competition; female responses to behavioral displays to males of various size and age as a measure of precopulatory mate choice; and ultimately the reproductive success of males of various age, size, and frequency or type of courtship behavior. Especially critical information would be to determine if females copulate with more than one male in a given period of receptivity. This is essential for determining the relative importance of sperm selection in belugas. A greater quantity of data, perhaps through remotely operated cameras, would be more likely to capture the act of copulation and allow for these analyses.
In addition to the knowledge that can be gained by studying behavior in aquarium belugas, the study of beluga reproductive physiology in aquaria has complemented or augmented the understanding gained from the study of wild belugas (Chapters 1 and 3, . Additionally, and perhaps critically, the access to abundant sample material from aquarium belugas allows the development and validation of research tools that will improve the ability to study, and thus manage, wild belugas (Chapters 1 and 2).
Because of the ability to collect data throughout the year, longitudinally sample individuals, and observe behavior that is rare or difficult to see from the surface, studies in aquaria serve as important complements to studies of wild belugas.
For example, the work by  provides an excellent complement to this work, as those authors had access to considerable amounts of material from wild belugas, but were restricted to morphological data collected post-mortem, primarily from the summer and fall. Although the current study had a vastly smaller sample size and includes few samples from wild belugas, it directly addresses the sampling gaps in the work of , leading to a more complete picture of beluga biology when considered in tandem than if either study were to be evaluated separately.

Management Implications of beluga mating strategies
Species with more promiscuous mating systems are thought to have lower extinction risks at small population sizes than other mating systems (Lee et al. 2011).
The degree of polygyny can also influence extinction risk; if a small proportion of males are responsible for large proportion of offspring, any perturbation to a male's ability to sire offspring can have a large effect on female productivity. While few males are required to support a polygynous mating system, a reduction in quality males (adults, larger animals, animals with fully developed courtship behavior) could reduce fecundity in females that employ precopulatory mate choice (Lee et al. 2011;. In the absence of quality mates, females may delay breeding until a quality mate is encountered, and as a result could miss the opportunity to breed in a given season. The apparent low level of postcopulatory selection, suggesting a lesser degree of promiscuity, as well as the potential for female mate choice in this species, make belugas theoretically more susceptible to extinction at small population sizes.
The potential for sea ice loss in the Arctic to expand beluga distribution , and thus decrease population density, could feasibly reduce encounter rates between the segregated sexes and enhance these extinction risks through reduced fecundity.
The sex ratio of a population can also affect how mating system is related to extinction risk. Female skewed sex ratios create greater degrees of stochasticity in polygynous populations, leading to higher extinction risk (Lee et al. 2011). This is an important consideration for small populations of belugas, as subsistence harvests generally target larger belugas. This can result in harvests composed primarily of adult males, due to their larger size Harwood et al. 2002). The current evidence suggests that male belugas have relatively low sperm production capability, limiting the number of females a given male could inseminate over a short period of time. A relatively low capacity for increasing mating rate, in concert with prolonged periods of courtship that enable female precopulatory mate choice, means that if males in a beluga population were relatively rare, female fecundity could be greatly reduced.
In saiga antelope, where males are preferentially hunted for their antlers, male limitation has been implicated in the collapse of wild populations (Milner-Gulland et al. 2003). A variety of effects of selective hunting have been documented in other mammals (reviewed by ).
While most subsistence hunts are sustainable (e.g. Harwood et al. 2002) and no ill effects would be expected from this selection, an unsustainable harvest occurred in Cook Inlet, Alaska in the last two decades of the 20 th century. The population declined from approximately 1,300 belugas to fewer than 400 in a twenty year span, with as much as 20% of the population removed annually through hunting (NMFS 2015). This population is currently listed as Endangered because it has not increased in size, despite the absence of subsistence harvests in the last 10 years (NMFS 2015).
The current factors that limit the recovery of this population are unknown. A number of possible threats have been identified for this population that could be impeding recovery, including nutritional limitation, noise pollution, and predation, all of which have the ability to limit fecundity and thus population growth (NMFS 2015). Most likely due to a lack of information on population demographics, or perhaps operating under the assumption that relatively few males are required for maximum productivity in belugas, mating strategies, OSR, and ASR were not identified as potential limitations to recovery. Additionally, directed efforts to evaluate these factors have not been specifically proposed in the recovery plan for this population (NMFS 2015).
The sex is unknown for most of the belugas that were harvested or struck and lost in Cook Inlet during the period of heavy exploitation (NMFS 2015). If the unsustainable harvest preferentially targeted adult males, as it does in other areas, it is possible that a very large proportion of the adult males in the population were removed during the period of heavy hunting pressure. If males are limited in their potential fecundity due to the combined effects of physiology and mating strategies, as suggested by the results of this study, then perhaps this population is male limited.
Given the long time to maturity in this species , it could take an extended period of time for the sex ratio to balance sufficiently. Further, if learning is associated with courtship behavior, as suggested by this study, then perhaps removal of adult males through hunting also reduced behavioral diversity in this population.
Inappropriate courtship behavior could then interfere with female precopulatory mate choice. If the threshold for accepting mates is not plastic, and fewer males meet threshold requirements for selection by females, either by size, behavior, or ability to induce ovulation, then female fecundity could be greatly reduced.
The methodologies developed in this work have their greatest potential application in the management of the endangered Cook Inlet belugas. The study of temporarily stranded belugas through minimally invasive blow sampling could yield important information on the sex ratio of the population, pregnancy rate, the structure of social groupings, and the relatedness (and perhaps paternity) of members of a social group. These temporarily stranded groups can consist of large proportions of the entire population at a given time (NMFS 2015) and therefore offer a tremendous opportunity to better understand the dynamics of this population. Blow sampling of free-swimming belugas could be employed in this population more feasibly than in others, given that a long-running photo ID project already exists, and a more extensive biopsy program is proposed (NMFS 2015). Both of these activities require close approaches to belugas, where blow sampling might be accomplished. While biopsy sampling might not be performed on very young belugas or females with attendant calves, blow sampling could be performed on these individuals with reduced welfare Pass: Pregnant 2X Nonpregnant *Diethyl ether liquid-liquid extraction was selected for use in assaying blow samples for progesterone due to lower levels of matrix interference in control samples. Turn off the power and unplug the electrophoresis box from the power supply.