Salivary cortisol captures endocrine response to an acute stressor in captive female tufted capuchin monkeys (Sapajus apella)

Measuring glucocorticoids such as cortisol is a useful tool for exploring relationships among behavior, physiology, and well-being in primates. As cortisol circulates in blood, it moves into biological matrices such as hair, urine, feces, and saliva. Saliva sampling is a simple, noninvasive method to measure cortisol that can be easily implemented by training animals to voluntarily provide samples. The temporal lag between elevation of cortisol in the blood and elevation of cortisol in saliva likely varies by species and must be characterized to identify appropriate sampling regimens. In the present study we characterized the time course of cortisol changes in saliva following an acute psychological stressor in captive tufted capuchin monkeys (Sapajus apella). We trained eight free-moving female tufted capuchin monkeys to voluntarily produce clean saliva samples. We exposed them to the acute stressor of a veterinary catch net and observed behavior pre and post exposure. We collected salivary samples immediately pre exposure (0 min) and 30, 45, 60, 75, 90, and 120 min after exposure. Salivary cortisol was quantified using a Salimetrics kit. Behavioral and cortisol measures were compared within individuals to a control condition in which no stressor was presented. Capuchins showed a clear behavioral response to the stressor by demonstrating increased freezing and pacing, decreased feed foraging, nonsocial play, and scratching, and decreased willingness to provide saliva samples after stressor presentation. After stressor presentation, average salivary cortisol began to increase at 30 min and continued to increase through the 120 min sample period. There was individual variation in absolute cortisol levels, the timing of the cortisol increase, and the timing of the peak. Our results suggest that no single time-point can be reliably used to evaluate salivary cortisol response to an acute stressor across individuals, and instead we recommend the collection of a prolonged time series.