All procedures for animal care, handling, and experimentation were approved by the Animal Care and Use Committee, University of Alaska Fairbanks under protocol 06-049, which conforms to the American Society of Mammalogists guidelines for the use of mammals in research (Sikes, 2016).

We studied female reindeer (n = 6), caribou (n = 6), and muskoxen (n = 8) from a captive population maintained by the Institute of Arctic Biology in Fairbanks, Alaska (64.8795, −147.8640). Air temperature (°C) and solar radiation were measured on site (HOBO, Onset Computer, Bourne, MA, United States). All animals were bred in captivity and maintained on formulated rations with access to natural vegetation and grass hay in outdoor pens of one to two hectares. Water was available ad libitum as snow or from heated water troughs. This study was conducted during late winter (ordinal days 41–99; 11 February to 10 April 2009; Figure 1) when snow depths minimized foraging on natural vegetation. All females were held in harem with males during the breeding season in the autumn before the study and monitored for production of calves in the spring after the study.

We used three formulated rations that were provided as pellets: a high N “Control,” a high N “Spike” with high δN¹⁵, and low N ration with intermediate δN¹⁵ (Table 1; Alaska Pet and Garden, Anchorage, AK, United States). High N rations were similar to growing forbs that are available to wild reindeer, caribou and muskoxen in spring whereas the low N ration is similar to the low-quality forages available in winter (Parker et al., 2005; Barboza et al., 2018). The high N rations therefore foster maintenance or gain of body protein whereas the low N ration simulated conditions that are likely to limit protein gain and foster conservation of body N. The acute shift in isotopic value from the control to the spike ration was used to simulate an acute change in range from an area without marine derived N to an area with N subsidies from anadromous fish such as Pacific salmon (Oncorhynchus tshawytscha). Reindeer and caribou were provided with ad libitum ration in a shared feeding trough whereas muskoxen were provided each ration as an individual supplement each day (500 g⋅d^–1) with ad libitum grass hay (Bromus spp.). Muskoxen did not readily accept the Low N ration, which was as low in N content as the grass hay (Table 1). Consequently, muskoxen were only studied on the High N rations through ordinal day 75 (17 March).

We fed all animals each diet in three sequential periods. The control ration was used to maintain caribou and reindeer throughout the year whereas muskoxen were maintained on a mineral supplement (M ration Alaska Pet and Garden, Anchorage, AK, United States) with the same ingredients as the control ration but with a different mineral vitamin premix (Gustine et al., 2010). Animals were transitioned to each new ration for 3 days before feeding that ration for 7 days prior to sampling. We fasted animals for 2 days by removing rations from the reindeer and caribou pens and by removing hay and withholding supplemental ration from the muskoxen. Animals were allowed to feed again for 7 days before sampling in the “refed” state (Figure 1). Sampling days occurred before fasting, at the end of fasting, and 7 days after fasting ended for each period (Figure 1).

Individual animals were observed to collect fresh feces and urine on sampling days (Figure 1). We moved caribou and muskoxen into small handling pens (<0.4 ha) for up to 4 h to observe and collect excreta from snow as soon as it was voided by each individual in the pen. Reindeer were haltered and accustomed to standing while tethered in stalls, which allowed us to collect voided excreta directly into a cup without snow. All excreta samples were collected into 60 ml plastic bottles and frozen.

Animals were weighed on electronic load scales (±0.1 kg; Tru-Test Model 703, San Antonio, TX, United States) after collection of excreta. Animals were manually restrained in handling chutes or stalls to collect blood samples from the jugular vein. Blood was collected into glass tubes (Vacutainer, Becton Dickinson, Franklin Lakes, NJ, United States) without additives and allowed to coagulate. Clear sera were decanted from blood after centrifugation at 3,000 × g for 10 min and frozen for analysis. The clot of blood cells and clotting proteins was retained and frozen after draining residual sera.

We measured subcutaneous rump fat via ultrasound on reindeer and muskoxen standing in the handling chute without chemical immobilization (Rombach et al., 2002a; Barboza and Parker, 2006). Rump fat was not measured in caribou to minimize the stress of repeated handling.

Urine samples were freeze dried and rehydrated with 25 mL deionized water for analysis. We used steam distillation to collect urinary urea N from urine samples (Nolan and Leng, 1972; Barboza et al., 1997). Fecal samples were dried to constant mass at 50°C in a forced-air oven then ground in a Wiley mill (Arthur Thompson, Philadelphia, PA, United States) through a 1 mm screen. We isolated fibrous residues from feces by using polyester filter bags to individually boil ground fecal samples in separate beakers of deionized water (F57 filter bags, Ankom Technology, Macedon, NY, United States) for 20 min followed by 3 rinses with hot water (Gustine et al., 2011b, 2014b).

Clotted blood was freeze dried for analysis. Proteins were precipitated from blood sera with Na₂WO₄ (Barboza et al., 1997) and freeze-dried. We measured urea concentrations of deproteinized serum by reaction with diacetyl monoxime (Barboza et al., 2004). Serum amino acids were collected from deproteinized serum by ion-exchange chromatography (Gustine et al., 2010).

We assayed ¹⁵N by continuous flow isotope ratio mass spectrometry by using a Finnigan Delta V plus mass spectrometer (Thermo Scientific, Waltham, MA, United States) combined with a Costech Elemental Analyzer (Costech Analytical Technologies Inc., Valencia, CA, United States) at the Alaska Stable Isotope Facility at University of Alaska Fairbanks. Isotopic values for nitrogen were reported in delta notation and expressed in parts per thousand, relative to air {δ = [(isotope ratio sample/isotope ratio standard)−1] × 1000} (Gustine et al., 2011a). Peptone was used as a reference standard with an expected δ¹⁵N value of 7.00‰, which was assayed with an accuracy of ±0.28‰ (Vansomeren et al., 2017).

The distribution of N in blood was measured in triplicate samples of blood from four female reindeer and two female caribou during winter. We collected blood samples without additive for serum and with sodium heparin for plasma. Heparinized blood was used to measure the relative proportions of packed red cells and plasma as hematocrit. Whole blood, packed cells, serum, and plasma were subsampled and dried to constant mass in a freeze drier to measure dry matter content. We used plasma to measure total soluble protein (i.e., including clotting proteins) by reaction with Bradford reagent (Coomassie Blue dye at 595 nm; Sigma Chemicals, St. Louis, MO, United States). Urea and amino acid contents of deproteinized serum were measured as above. Total N of freeze-dried blood fractions were measured by elemental analyzer (LECO, St. Joseph, MO, United States). We allocated total N to red cells, plasma protein, free amino acids, and urea. Creatinine and other metabolites were estimated as the residual N from blood plasma.

We used two sets of mixed effects regression models with individual animal (constant) and time (ordinal days; covariate) as random effects to account for repeated measures. The first set of models examined body condition by using the following five dependent variables (Y_C): body mass, rump fat, serum urea concentration, δ¹⁵N of fecal fiber, and δ¹⁵N of blood clot. Each model included the fixed effect of taxon (TAXON; reindeer, caribou, or muskox). The first set of models on body condition included the fixed effects of reproductive status (REPRO; pregnant or not pregnant), diet treatment (DIET; Control, Spike, Low), and phase (PHASE; fed, fasted, refed) with interactions between taxon and reproductive status, taxon and diet treatment, and taxon and phase.

The full model for condition variables was:

The second set of models examined isotopic changes in three pools of body N as dependent variables (Y_N): δ¹⁵N value of serum amino acids, δ¹⁵N value of serum protein, and δ¹⁵N value of urinary urea. The second set of models included the fixed effects of taxon, phase, and their interaction. Reproduction was initially included but that fixed effect was not significant and thus removed from this second set of models. We also used δ¹⁵N of fecal fiber (FIBER15N) and δ¹⁵N of blood clot (CLOT15N) as covariates to represent isotopic endpoints of diet and body N, respectively, in the second set of models (Taillon et al., 2013; Gustine et al., 2014b).

The full model for variables of isotopic N was:

In each set of models, we used backward elimination of fixed effects by starting with the full model and progressively excluding interactions and independent variables with beta coefficients that were not significantly different from zero (P > 0.05). Margins (X¯ ± standard error) were calculated for the observed range of each fixed effect in the final model. We used Bonferroni’s adjustment for pairwise comparisons and contrasts of margins between diets and between phases (e.g., fasted vs. fed) within each taxon (STATA 16, StataCorp, College Station, TX, United States).

Finally, we derived simple linear relationships using a single fixed effect in the mixed model for two comparisons among taxa with individual animal (constant) and time (ordinal days; covariate) as random effects to account for repeated measures. Firstly, the relationship between the isotopic endpoint of the body (CLOT15N) and the isotopic endpoint of the diet (FIBER15N) on the High N control treatment (Figure 1). Secondly, relationships between body N pools (Y_N) and clotted blood cells among taxa.

Daily average air temperature ranged from −30°C to + 0°C as daily photoperiod increased from 420 to 900 min⋅day^–1 over the course of the study (Figure 1). All females were exposed to intact males before the study but one of six reindeer, two of six caribou, and five of eight muskoxen did not produce calves because they did not enter estrus or because they resorbed the fetus in utero. Births occurred on ordinal days 99–104 for reindeer, 128–136 for caribou, and 113–119 for muskoxen. Body mass was not significantly affected by reproductive state (Table 2) but did decline over the course of the study by −10.8 ± 2.8 kg in reindeer, −2.1 ± 1.0 kg in caribou, and −2.6 ± 0.8 kg in muskoxen (Table 2 and Supplementary Table S1). Reindeer lost 15.9 ± 1.9 mm of subcutaneous fat at the rump as body mass declined by 9.1% whereas muskoxen maintained rump fat depth over the course of the study even though body mass declined by 1.3%. Fasting did not significantly affect rump fat depth in reindeer or muskoxen (Table 2). However, body mass was lost in each 2-day fast by 2.3 ± 0.6 kg in reindeer, 2.1 ± 0.2 kg in caribou, and 5.8 ± 0.4 kg in muskoxen. Fasting increased serum urea concentrations by +5.1 ± 2.0 mg⋅dl^–1 in reindeer, +6.0 ± 1.9 mg⋅dl^–1 in caribou, and +16.2 ± 1.6 mg⋅dl^–1 in muskoxen. Serum urea was not significantly affected by the change from High N ration to High N Spike ration but the subsequent change to the Low N ration decreased serum urea by −5.7 ± 1.6 mg⋅dl^–1 in reindeer and by −11.8 ± 2.2 mg⋅dl^–1 in caribou.

Values for δ¹⁵N in fecal fiber were significantly affected by diet. An increase in δ¹⁵N value of the ration from High N Control to High N Spike (Table 1) significantly increased fecal fiber δ¹⁵N values by 2.2 ± 0.1‰ in reindeer, 1.7 ± 0.1‰ in caribou, and 0.8 ± 0.2‰ in muskoxen (Table 2 and Supplementary Table S1). The subsequent change to the Low N ration with a lower δ¹⁵N value (Table 1) significantly reduced fecal fiber δ¹⁵N values by 2.4 ± 0.2‰ in reindeer and by 1.5 ± 0.1‰ in caribou (Table 2 and Supplementary Table S1). Values for δ¹⁵N in fecal fiber were also significantly affected by fasting. Fasting increased fecal fiber δ¹⁵N values by 0.5 ± 0.1‰ in reindeer and by 0.3 ± 0.1‰ in caribou, which was reversed when the ration was fed ad libitum. In muskoxen, the 2-day fast did not significantly change values for δ¹⁵N in fecal fiber but resumption of ad libitum feeding decreased those values by −0.7 ± 0.1‰.

The random effects of individual and ordinal day accounted for less than 0.1% of the variance in the model for δ¹⁵N values in fecal fiber but more than 98% of the variance in δ¹⁵N values of the blood clots (Table 2). Most of the variance in δ¹⁵N values of the blood clots during the study was associated with changes over time within animals: values rose by +0.8 ± 0.3 and +0.5 ± 0.1‰ in reindeer and caribou, respectively, whereas values for muskoxen fell by −0.3 ± 0.1‰ over the course of the study (Supplementary Table S1). Fasting significantly affected blood clot δ¹⁵N values (Table 2) but pairwise differences between phases were only significant for reindeer, which increased by +0.2 ± 0.1‰ between the fed and fasted state. In reindeer and caribou, δ¹⁵N values of the blood clots increased by +0.2 ± 0.1‰ with each change in diet even though the δ¹⁵N values of those rations increased and then decreased (Table 1 and Supplementary Table S1). Conversely, δ¹⁵N values of the blood clots decreased by −0.2 ± 0.1‰ in muskoxen as the δ¹⁵N value of the ration increased (Table 1 and Supplementary Table S1).

The initial isotopic endpoints for the body were positively related to the long-term diet at the start of the study. Values for δ¹⁵N in blood clots increased by +0.21 ± 0.07‰ for each unit increase in δ¹⁵N values of fecal fiber during the first period of the study when animals were provided with the Control High N diet, which was similar to the diet provided through the previous year (Figure 2).

Values for δ¹⁵N in serum amino acids, serum protein, and urinary urea were not significantly affected by variation in the isotopic endpoint for the body indicated by blood clot δ¹⁵N values within taxon (Table 3). However, simple linear relationships across taxa indicated that each unit increase in δ¹⁵N value of the body endpoint was associated with an increase in +0.50 ± 0.20‰ of serum amino, and +1.61 ± 0.28‰ in urinary urea values (Figure 3).

Values for δ¹⁵N in serum amino acids, serum protein, and urinary urea were significantly affected by the dietary isotopic endpoint indicated by fecal fiber δ¹⁵N values (Table 3). Each unit increase in δ¹⁵N value of the dietary endpoint increased serum amino acid and protein values by only +0.28 ± 0.07 and +0.14 ± 0.04‰, respectively, whereas urinary urea values increased by +0.88 ± 0.1‰ (Figure 3). Fasting for 2 days did not affect δ¹⁵N values for either serum amino acids or protein (Table 3). However, δ¹⁵N values for urinary urea increased during fasting by +2.11 ± 0.40‰ in reindeer, +2.41 ± 0.30‰ in caribou, and +1.05 ± 0.34‰ in muskoxen. Values for δ¹⁵N of urinary urea declined again after ad libitum feeding resumed in each taxon.

Blood cells accounted for 48 ± 3% of the volume of heparinized whole blood. Cells accounted for more dry matter than plasma (15.67 ± 0.29 vs. 4.38 ± 1.28 g dry mass⋅dL^–1 in whole blood) but those dry components were similar in N concentration (15.93 ± 0.50 vs. 14.53 ± 2.09% N). Consequently, whole blood contained 20.05 ± 1.07 g dry mass⋅dL^–1, which was 15.78 ± 2.28% N. Plasma accounted for 21.6 ± 4.9% of dry mass and 19.8 ± 2.7% of total N in whole blood whereas blood cells were 78.4 ± 4.9% of dry mass and 80.2 ± 2.7% of N in whole blood (Figure 4). Serum contained 77.3 ± 5.9 mg of protein⋅mL^–1 that accounted for 12.82 ± 0.96 mgN⋅mL^–1, which was 19.2 ± 0.3% of total N in whole blood. Serum contained only 387 ± 201 μgN⋅mL^–1 as urea and only 12 ± 4 μgN⋅mL^–1 as amino acids, which was only 0.36 and 0.02% of the total N in whole blood, respectively. Residual N that was associated with creatinine and other metabolites was only 0.22% of N in whole blood (Figure 4).

The conceptual relationships between the metabolic pools of N and diet (Figure 5) characterize flows when body N is maintained or gained at positive N balance (i.e., when diet is sufficient to sustain the body store of protein for reproductive investments in the fetus during winter and in milk during spring). More importantly, the model characterizes flows of N when reproductive investments would be compromised by depletion of body N in a negative N balance. Values of δ¹⁵N in urinary urea can indicate short term (∼1 day) declines from adequate to negative N balance when diet is stable (Gustine et al., 2012). Amino acids may reflect medium term (∼7 days) switches in diet at positive balance but it may not indicate negative balance very quickly because it is buffered by intermediary metabolism. Similarly, clots can indicate changes in dietary values of δ¹⁵N over the long term (∼30–90 days) but isotopic changes in this large metabolic pool of N may be quite subtle. The projected effect of changing N balance on the average δ¹⁵N value of female reindeer was less than 1‰ when animals were held at positive or negative N balance observed on a high N diet during winter (Barboza and Parker, 2006).

Serum protein may provide the best indication of a declining body protein store in northern ungulates. Gustine et al. (2014a) hypothesized that serum protein could be compared with red blood cells to indicate relative increases in δ¹⁵N value of the circulating protein. However, the experimental data we present in this study indicates that the δ¹⁵N value of serum protein would probably decline not increase when N balance was negative. Removal of serum proteins from circulation by deposition in tissue or by oxidation to urea would remove N with δ¹⁵N values that exceed the values of dietary and circulating amino acids and muscle. Consequently, serum protein concentrations would be maintained by replacement from diet or tissue proteins with lower δ¹⁵N values that would likewise decrease the average value of the serum protein pool. Gustine et al. (2014a) reported values of δ¹⁵N in serum protein that were negatively related to those of red blood cells in female caribou during severe winters that resulted in low body mass of mothers and birth mass of their calves. Conditions that would predispose animals to depletion of body protein were associated with serum protein values that declined as isotopic values for red blood cells rose.

The investment of body stores in reproduction entails more than one currency because those investments are not just the loss of maternal body mass but a controlled mobilization of energy, protein, and minerals (Barboza and Parker, 2006). Body capital is used sparingly in relation to the supplies from diet and the costs of activity and weather. Female reindeer, caribou, and muskoxen are able to mobilize energy from body fat while conserving and reallocating protein and minerals to their offspring during pregnancy and lactation when food supplies are low in abundance and quality. Body protein is not an inert depot but a dynamic pool of nitrogen in cellular and circulating proteins and metabolites. Conservation of protein is therefore a process of controlling N flows rather than simply managing the loss from a static reserve (Barboza et al., 2009). Conservation of protein is achieved in two ways through the liver: control of flows to oxidation and routing protein in circulation. Firstly, dietary amino acids are burned before stored amino acids, that is stored amino acids are spared from oxidation to urea and that oxidized urea N is recycled. Secondly, the distribution of N from tissues such as muscle to the fetus is apparently managed with circulating proteins (Sand et al., 2015).

Capital breeding allows animals to uncouple primary production of food supplies from their reproductive investments to allow longer times for development and to facilitate flexibility in timing of births (Gustine et al., 2017). Capital breeding is most evident in northern ungulates, phocid seals, whales, and bears that schedule pregnancy or lactation when food intakes are low. However, most animals use some body stores to offset dietary supplies of lipid, protein, minerals, and vitamins when food supplies are low (McCue, 2010). Some degree of capital investment is involved in energy or nutrients for most animals. In muskoxen, copper is transferred from maternal stores in the liver to establish large hepatic stores of copper that the neonate uses until weaning (Rombach et al., 2002b; Swor, 2002). Calcium from bone is routed to developing offspring and milk in most mammals – the delay of restoring bone calcium from diet may be a few days in small animals but a season in large ungulates (Baksi and Newbrey, 1989; Schmidt and Hood, 2014). Behavioral and metabolic adaptations for intermittent feeding allow animals to reduce exposure to weather extremes or predation over diurnal to seasonal scales of time (Barboza and Hume, 2006). Capital breeding extends the duration and intensity of those metabolic adaptations to accomplish both survival and reproduction. In northern ungulates, the strategy is constrained by environmental variance in the quality and duration of forage available in summer and the capacity to store enough nutrients and energy for the variable demands of winter in both migratory and sedentary populations. Species that are widely distributed from north to south such as white-tailed deer (Odocoileus virginianus) and mule deer (Odocoileus hemionus) alter seasonal movements in relation to forage availability and body stores to balance maternal survival and maternal investments in offspring (van Beest et al., 2013; Monteith et al., 2014; Henderson et al., 2017). Summer foraging conditions and winter severity affect reproductive rates in northern populations of white-tailed deer (Simard et al., 2014; Ayotte et al., 2019), which is consistent with a capital breeding strategy that begins in a narrow seasonal window of mating before winter. Southern populations of white-tailed deer mate and birth over wider seasonal windows than their northern conspecifics (Bronson, 1985). Southern deer may be able to use the dietary income of energy and nutrients from forage available in winter and spring for pregnancy and lactation whereas growth and maturation may be constrained by summer forage production especially in stochastic environments with high temperatures and low precipitation (Ginnett and Young, 2000; Cain et al., 2006; DeYoung et al., 2019). Survival in these drought prone environments is favored by conservation of body water and body protein through the same metabolic routes that limit oxidation and facilitate recycling of body N in northern ungulates during winter (Barboza et al., 2009).