A novel method to measure steroid hormone concentrations in walrus bone from archaeological, historical, and modern time periods using liquid chromatography/tandem mass spectrometry

Rationale A liquid chromatography/tandem mass spectrometry (LC/MS/MS) method was validated and utilized to measure and analyze four steroid hormones related to stress and reproduction in individual samples from a novel tissue, Pacific walrus (Odobenus rosmarus divergens, herein walrus) bone. This method determines steroid hormone concentrations in the remote walrus population over millennia from archaeological (>200 bp), historical (200–20 bp), and modern (2014–2016) time periods. Methods Lipids were extracted from walrus bone collected from these periods using methanol before LC/MS/MS analysis. Isotopically labeled internal standards for each target hormone were added to every sample. Analytical and physiological validations were performed. Additionally, a tissue comparison was done among paired walrus bone, serum, and blubber samples. A rapid resolution liquid chromatography system coupled to a QqQ mass spectrometer was used to analyze all samples after derivatization for progesterone, testosterone, cortisol, and estradiol concentrations. Multiple reaction monitoring was used for MS analysis and data were acquired in positive electrospray ionization mode. Results Progesterone, testosterone, cortisol, and estradiol were linear along their respective standard calibration curves based on their R2 values (all > 0.99). Accuracy ranged from 93–111% for all hormones. The recovery of extraction, recovery of hormones without matrix effect, was 92–101%. The overall process efficiency of our method for measuring hormones in walrus bone was 93–112%. Progesterone and testosterone concentrations were not affected by reproductive status among adult females and males, respectively. However, estradiol was different among pregnant and non‐pregnant adult females. Overall, steroid hormones reflect a long‐term reservoir in cortical bone. This method was also successfully applied to walrus bone as old as 3585 bp. Conclusions LC/MS/MS analysis of bone tissue (0.2–0.3 g) provides stress and reproductive data from elusive walruses that were alive thousands of years ago. Based on physiological validations, tissue comparison, and published literature, steroid hormone concentrations measured in walrus cortical bone could represent an accumulated average around a 10–20‐year time span. By investigating how stress and reproductive physiology may have changed over the past ~3000 years based on bone steroid hormone concentrations, this method will help answer how physiologically resilient walruses are to climate change in the Arctic.


| INTRODUCTION
Lipophilic steroid hormones, including cortisol, estradiol, progesterone, and testosterone, can provide important physiological information in marine mammals. Estradiol, progesterone, and testosterone are reproductive hormones and have been used to determine marine mammal reproductive status, including pregnancy in cetaceans, [1][2][3][4] pinnipeds, 5,6 and Pacific walruses (Odobenus rosmarus divergens, hereafter walrus). 7 Cortisol is produced in response to stress and naturally increases during times of high energy use including reproduction, molting, and mobilization of lipid stores in pinnipeds. 5,6,[8][9][10] However, when cortisol concentrations are consistently high, this is an indicator of a chronic stressor. 11 Pinnipeds are more susceptible to disease and may have poorer body condition when experiencing chronic stress compared to animals that have not been exposed to a chronic stressor and experience continuously elevated cortisol concentrations. 12,13 Clearly, steroid hormone studies of marine mammals provide relevant physiological data for current and future population health assessments. 14 Bone tissue contains lipids which are sequestered over the lifespan of an animal and do not significantly degrade after death, which means they can be detected in the bone after being buried for thousands of years. 15,16 Testosterone and estrogens, including estradiol, have been extracted and used to assign sex to human bones as old as 6961 calendar years before present (BP), 17 and testosterone and estradiol have been extracted and analyzed from rat (Rattus spp.) bone. 18 Bone has a slow turnover rate (3% cortical bone/year), 19 therefore hormone concentrations from bone are expected to represent a long-term accumulated average for an individual. This is beneficial when monitoring long-term physiological changes in a population, because bone hormone concentrations are not likely to be skewed by acute stressors or reproductive events like serum and blubber. 2,3,8,20,21 15 and steroid hormones have been extracted from rat bone and measured using enzyme immunoassays (EIAs). 18 Steroid hormones have been extracted from numerous matrices and measured with various immunoassay kits including enzyme linked immunoassays (ELISAs), radioimmunoassays (RIAs), and EIAs. Assays have been used to measure steroid hormones in marine mammal feces, 1,22 blubber, 6 serum, urine, 21 saliva, 7,23 baleen, 4,24 earwax, 25 and whale blow. 26 Immunoassay techniques are beneficial when sample mass is abundant, and they generally lower the cost of analysis. 27 However, immunoassays require relatively large sample masses and multiple assays for multi-hormone analyses, which increases required lab time. 28 In addition, cross-reactivity with target steroid hormone metabolites can lead to inflated hormone concentrations. 29,30 Furthermore, due to complicated logistics of collecting tissue samples from free-ranging marine mammals and animal care standards for managed populations, marine mammal biopsies, blow samples, fecal samples, etc., once obtained, are relatively small and are slated for multiple different analyses, (e.g., contaminants, fatty acids, disease). 31,32 Thus, researchers need to efficiently analyze tissue samples and have been transitioning from using immunoassays to more sophisticated analyses, like liquid chromatography/tandem mass spectrometry (LC/MS/MS). 28,30 LC/MS/MS analysis allows for greater utility of samples collected from rarely encountered species, such as marine mammals. For example, eight different hormones have been analyzed in a single 0.40 g blubber sample using LC/MS/MS 28 compared with a single hormone being measured in 0.15 g using ELISAs. 33

| Chemicals and reagents
Lipid extraction of powdered bone was performed by using 100% HPLC grade methanol from VWR BDH® Chemicals (Radnor, PA, USA).
Isotopically labeled internal standards, d 4 -cortisol, 13 C 3 -testosterone,  Samples were then stored in a − 80°C freezer until analysis.

| Blubber samples
The oxidized outer layer of walrus blubber from each full thickness slab was removed with sterilized individual razor blades exposing fresh blubber tissue. Two separate vertical strips of full thickness blubber weighing between 0.2-0.3 g were removed starting from below the skin and ending above the muscle and transferred to separate 2.8 mL ceramic bead homogenizer cryovials. Samples were homogenized, internal standards added, and lipids extracted with methanol as TABLE 1 Total sample sizes (n) of walrus bones collected for analyses during archaeological, historical, and modern time periods. Further, sample sizes are categorized into age class (adult, subadult, and unknown) and sex (female, male, and unknown) for each time period. "-" indicates no samples were collected for that category. See other tables for specific sample sizes for physiological validation (Table 3), time period (Table 6), and tissue comparison analyses (Tables 7 and 8

| Steroid hormone concentrations among different bone elements
To ensure that steroid hormone concentrations do not differ between walrus skeletal elements, we performed a pilot study on skull and mandible bone sampled from the same individual walruses (n = 7). All steroid hormone concentrations were similar between skulls and mandibles from the same individual (paired t-tests; cortisol P = 0.32, estradiol P = 0.08, progesterone P = 0.20, and testosterone P = 0.11, n = 7 pairs). These data agree with Yarrow et al, 18 where testosterone measured in tibias and femurs of rats were similar. This pilot study confirmed that different walrus skeletal elements used here result in comparable steroid hormone concentrations.

| Percent lipid correction factor
Walrus bones from different archaeological, historical, and modern time periods potentially have different lipid compositions, as lipid in cortical bone is already low, 19,42 and taphonomic processes could affect the lipid composition of archaeological bones buried for thousands of years. 15,43 In addition, there has been evidence of degradation of progesterone in cetacean blubber, 3  concentrations are also reported as the more traditional ng/g bone for reference and tissue comparison purposes.

| LC/MS/MS conditions and analysis of steroid hormones
Prior to analysis, each sample was reconstituted in 200 μL of methanol, split into two equal aliquots and dried again using an Samples with hormone concentrations below the detection limit for LC/MS/MS analysis (< 2.0 ng/g), were included in statistical analysis by assigning one-half the detection limit concentrations for each hormone with a non-detectable signal. 46,47 Extraction efficiencies were determined by comparing known volumes of added internal standards of each hormone that had been through the extraction process (i.e., blank samples that went through the steroid hormone extraction method with only added internal standards and methanol, n = 8, "Blank-Extraction"), with samples with internal standards and no extraction (i.e., added internal standard to vial and dried using nitrogen gas, n = 5, "Blank-Dried Internal Standards").

| Preparation of standards and stock solutions
Stock vials of isotopically labeled internal standards, d 4 -cortisol, 13 C 3testosterone, 2 H 9 -progesterone, and 2 H 5 -estradiol, were diluted to 10 ng/μL with methanol in separate 10 mL glass scintillation vials. Glass vials were then wrapped in aluminum foil and stored at −8°C. Non-labeled steroid hormone standards (hydrocortisone, β-estradiol, testosterone, and progesterone) used for creating calibration curves were diluted with methanol to both 10 ng/μL and 0.05 ng/μL per steroid hormone. Non-labeled steroid hormone standards were kept in amber 1 L bottles and stored at −8°C. All standards and stock solutions were brought to room temperature before analysis. progesterone, testosterone, cortisol, and estradiol, respectively). All samples went through the same extraction method as described above, including the addition of 100 ng of isotopically labeled internal standard for each hormone.

| Analytical validation of steroid hormones in bone
Linearity was determined by plotting the mean relative response ratios (n = 5) from bone powder spiked with concentrations of each hormone along their respective standard calibration curves ( Figure 1). 45 The mean relative response ratio is the peak area ratio of the analyte divided by the peak area of isotopically labeled internal standard. Accuracy was determined by using the equation: Where expected concentration (EC) is divided by mean actual concentration measured in spiked bone tissue (MAC) and then multiplied by 100 (

| Statistical analysis
Steroid hormone concentrations in walrus bone were not normally distributed; therefore, non-parametric Kruskal-Wallis ANOVAs were used to determine significant differences in hormone concentrations among known subadult and adult females of different reproductive statuses and between adult and subadult males to perform physiological validations (as described above). Kruskal-Wallis ANOVAs analyze differences in median values that are robust to outliers. All physiological validation data are reported as median ± 1 standard deviation (SD), with mean values reported for reference in ng/g lipid (Tables 3A and 3B). The samples used for the physiological validations are listed in Appendix 2.
The data used for tissue comparison analysis were log transformed to normalize distribution of steroid hormone concentrations in bone, blubber, and serum. The samples used for the tissue comparison are listed in Appendix 3. Three factorial ANOVA tests were used to test for differences in mean concentrations of steroid hormones among all tissues, between sexes, and the interaction of sex and tissue. If the differences among hormone concentrations were statistically significant among tissues, between sexes, and/or had a significant interaction of tissue and sex, a Tukey post hoc test was used to elucidate specific differences among the factors ("tissue", "sex", "sex*tissue") for each hormone. The majority of animals were classified as adult walruses (n = 25 adults, n = 3 subadults, and n = 5 unknown); therefore, sample size was too small to perform statistical  "*" includes one female that was lactating with a yearling. "-" represents no available data. P-values for comparison among subadult and adult females of different reproductive status are presented. Females were classified as subadult based on provenience data provided by museum records, hunter observations, and tooth age estimates 50 (sexually immature from approximately 1-9 years old). 51 Non-pregnant adult females were classified as adult based on tooth ages (i.e., > 9 years) and no fetus being present based on hunter observations. There were similar concentrations among adult females irrespective of pregnancy and reproductive status, except for estradiol    The recovery of extraction (%), or recovery of hormones without matrix effect, was 98%, 92%, 99%, and 101%, for progesterone, testosterone, cortisol, and estradiol, respectively. The overall process efficiency of our method for measuring hormones in walrus bone was 112%, 93%, 103%, and 110%, for progesterone, testosterone, cortisol, and estradiol, respectively. Equations used for validations are shown in Table 4.
The percent recovery of each internal standard was calculated by comparing the ratio of mean hormone concentration detected in "Blank-Extraction", divided by the mean hormone concentration

| Measurement of steroid hormones with LC/MS/MS
Multiple reaction monitoring was used for accurate detection of each steroid hormone. Thus, two product ions were checked for each steroid hormone using two different optimized collision energies (Table 5) Table 5.

| Bone
Progesterone, testosterone, cortisol, and estradiol were successfully Bone serves as a long-term reservoir for steroid hormones and lipids, but the reservoir is not metabolically inert nor do all steroid hormones biochemically behave similarly in cortical bone, 18 which could lead to ND concentrations in bone from different time periods.
Lipids marginally deteriorate in bone over time and hormones could potentially be degraded biologically and/or abiologically into more stable metabolites or leave bone completely with lipids through leaching, which could result in ND concentrations. 15

| Walrus tissue comparison
Tissues were compared to determine if bone steroid hormones were  (Table 8). This lends further support to the idea that bones are a long-term reservoir of steroid hormones with the possible exception of estradiol, and can be used to monitor long-term stress response that will not be skewed by acute stressors. 20 With the slow bone turnover rate of 3% cortical bone/year, 19 cortical bone might even be a longer-accumulated average of steroid hormone concentrations than the weekly to monthly average of blubber. 10

| Estradiol
Estradiol concentrations were significantly different between years (ANOVA, P < 0.001, Table 7). Thus, 2014 samples were tested for estradiol concentration differences among tissues, between sexes, TABLE 7 Mean estradiol concentrations (ng/for male and female walrus tissues (i.e., bone, blubber, and serum) ± 1 SD, median concentrations in ng/g lipid (for reference purposes only), concentration ranges (non-lipid corrected), and sample sizes (n). Walruses were harvested in 2014 and 2015 by Native subsistence hunters on St Lawrence Island, AK, USA. Significant differences among mean logtransformed estradiol concentrations are indicated with bolded P-values from three-way ANOVAs testing differences using two main factors (i.e., sex and tissue) and an interaction term (i.e., sex*tissue) for separate sampling years. If the tissue factor or the interaction term was significant, relevant  helping to increase bone mineral density, and is locally produced in bone. 16,18,64 It is still unknown how much this local production of estradiol contributes to overall estradiol concentrations compared with gonadal production, but bone is still an estradiol reservoir to some degree. 18 Therefore, estradiol may not have similar long-term reservoir times compared with other hormones measured in this study.

| Progesterone
Mean progesterone concentrations were significantly different between sexes (ANOVA, P < 0.001) and the interaction of sex and tissue (P = 0.009), but not tissue as a main effect (P = 0.27). Female blubber progesterone concentrations were driving the significant differences seen in the interaction term (i.e., sex*tissue). Female blubber progesterone concentrations were significantly higher compared with male blubber, bone, and serum progesterone concentrations (Tukey's post hoc test, P < 0.001, 0.007, 0.001, respectively). Bone progesterone concentrations for females were lower than blubber, but higher than serum progesterone concentrations (

| Testosterone
Mean testosterone concentrations were significantly different among tissues (ANOVA, P = 0.005), but not between sexes (P = 1.0), nor the interaction of sex and tissue (P = 0.75). Significant differences among walrus tissues were only found among bone and blubber testosterone concentrations (Tukey' post hoc test, P = 0.003), but not among serum and blubber (P = 0.33) or serum and bone (P = 0.26). Bone testosterone concentrations among walruses showed higher levels compared with serum and blubber ( . Testosterone is also known to be an important hormone for conversion into estradiol, which helps stimulate bone turnover in humans; 64 however, this was not clearly demonstrated in rats. 18 Further research into how testosterone is converted into estradiol in walrus bone is needed to determine the role bone turnover has on testosterone concentrations in walruses.  (Table 3A). We did see high variability in progesterone, testosterone, cortisol, and estradiol concentrations, which could indicate the reproductive success of individuals (Table 3A). That is, expected accumulation of progesterone should be higher in the bone of a female that has had three pregnancies   64 In addition, estradiol can be locally synthesized in bone by the aromatization of testosterone. 18 Thus, estradiol concentrations measured in walrus bone probably turn over on a shorter time scale, e.g., seasonally, unlike the other steroid hormone concentrations measured in this study (Table 3B). The apparent shorter reservoir time could explain why estradiol was significantly different between pregnant and non-pregnant females (Table 3B). While the non-pregnant females that were lactating and/or were accompanied by offspring did not contain a fetus, if they had a calf, there is a possibility they would have recently given birth and, hence, had similar estradiol concentrations compared with pregnant females (Table 3B).

| Physiological validations of steroid hormones in walrus bone
There were ample subadult female bones for the physiological validation analyses compared with the tissue comparison test (n = 18 and n = 0, physiological validations and tissue comparison, respectively).
Subadult females had significantly higher steroid hormone concentrations compared with adult females, with the exception of cortisol measured in known non-pregnant females (Tables 3A and 3B) (Table 3A). Combining our results of the tissue comparison and the physiological validation, steroid hormone concentrations measured in cortical bone represent a long-term reservoir of steroid hormones (Tables 3A, 3B, 5 and 6). This agrees with Yarrow et al, 18 who suggested there are reservoirs of estrogens and androgens in rat bones; however, they did not suggest a timeframe for that reservoir.
Steroid hormones are lipophilic, and lipids in bone are associated with cortical bone cells and its mineralized tissue. 16,80 There has been evidence of a strong positive linear relationship of bone cell turnover and lipid accumulation in rat bone. 81 While not directly transferrable to walruses, the relationship between bone cell turnover and lipid accumulation in bone supports our suggestion that steroid hormones, being lipidassociated molecules, have a slow turnover rate in walrus cortical bone (~3%/year). 19   Sample time period P (ng/g lipid) T (ng/g lipid) C (ng/g lipid) E (ng/g lipid) P (ng/g powder) T (ng/g powder) C (ng/g powder) E (ng/g powder)