This study was conducted in the Hatay Mountain Gazelle Wildlife Development Area, located near the Syrian border in Hatay Province, Türkiye (36°32′N, 36°32′E; elevation 200–450 m, Figure 1). The region in which the study was conducted encompasses an area of 13,228 hectares, consisting of grassland vegetation and shrublands, extensive agricultural land, and rocky hills. The study area is characterized by a warm climate, with mild winters and hot, dry summers with no recorded precipitation. Spring and autumn are transitional periods with moderate temperatures and low rainfall, reflecting semi-arid climatic conditions (3, 25). As of 2025, the population of mountain gazelles in this area is estimated to be 1,504 individuals (data from Nature Conservation and National Parks, Hatay Branch), whereas at the time this study was conducted in 2023 in this region, the estimated population stood at 1,387. This habitat, which supports a diverse range of mammalian fauna, is actively utilized by the gazelle throughout the year and represents the northernmost distribution of the species and is the only known habitat of mountain gazelles in Türkiye (26). Notably, the protected area contains both free-ranging and captive populations of mountain gazelles that accordingly experience similar macroenvironmental conditions, including photoperiods and climatic conditions. Moreover, the areas inhabited by both free-ranging and captive populations are characterized by a similar vegetation (Figure 2). However, whereas animals in the captive population have constant access to barley, hay, and water supplied by the staff of the Direction of Nature Conservation and Natural Parks of Hatay (Figure 2), in the case of the free-ranging population, access to water tends to be limited. Captive population is housed within a 12-hectare fenced area at the Hatay Mountain Gazelle Production Center, located inside the study region (3, Figure 1). At the time of sampling, the population consisted of 39 females and 11 males. The Production Center was originally established to support the small local population and to facilitate reintroduction efforts, though this goal has not yet been realized. It also provides care for injured individuals and orphaned newborns. Due to habituation to humans, these animals are not released back into the wild. The population breeds within the enclosure, and offspring remain with their parents. Veterinary interventions are limited to essential cases such as injury or illness. As such, the management conditions differ considerably between the free-ranging and captive populations. For the purpose of this study, we use the term “population status” to indicate whether a population is captive or free-ranging.

Fecal samples were collected from the ground in the natural habitat of mountain gazelles from both free-ranging and captive unknown individuals. Prior to fecal sample collection, mountain gazelles were observed at a minimum distance of 500 m for free-ranging individuals and 50 m captive individuals. Gazelles tend to form groups during certain periods, while at other times some individuals may remain solitary (3). During our observations, when an individual or group was seen, the area was promptly approached either on foot or by vehicle, and the freshest fecal samples were collected as quickly as possible. Approximately 7 g of feces were gathered from each sample, placed into polypropylene tubes, and kept at 4 °C in a cooler during transport to the laboratory. Although animals were monitored closely before sample collection, it was not possible to assign fecal samples to specific individuals. As a result, the sex and age of the animal that produced each sample remain unknown. Furthermore, mountain gazelles, particularly those in free-ranging populations, are highly vigilant and are likely to withdraw when encountering unfamiliar stimuli or anthropogenic activity. This behavior, observed in both wild and captive populations, limited our ability to achieve individual-level identification at the time of non-invasive sampling. Nevertheless, based on direct observation of defecation events and spatiotemporal separation during sampling, we assumed that each fecal sample originated from a different individual. Sample collection months were selected to represent biologically relevant reproductive periods (i.e., mating in December, gestation in April, lactation in July, and non-reproductive in September), which also correspond to calendar seasons (winter, spring, summer, and autumn, respectively). This alignment allowed for evaluation of both reproductive status and potential seasonal variation in hormone levels.

The fecal samples were stored in separate plastic bags, labeled with the time, date, and location of collection and maintained at −20 °C until used for analysis. The analysis of hormone metabolite concentrations began with drying the samples at 50 °C for 24 h. Thereafter, each dried sample was ground using a mortar and pestle, and then 0.5 g of dried fecal powder was transferred to tube to which 5 mL of 80% methanol was added. Following vortexing for 15–20 s, the samples were extracted for 30 min with shaking, after which, they were centrifuged for 15 min at 3,500×g. After centrifugation, 100 μL of each extract was pipetted into two Eppendorf tubes (one for progesterone metabolites determination and other for testosterone metabolites determination) and 900 μL of phosphate-buffered saline (PBS) was added to each tube (first dilution step). The diluted extracts were vortexed and stored at −20 °C until progesterone and testosterone metabolite concentrations were determined by ELISA (within one to two days). After ELISA, diluted extracts were kept at −20 °C.

Two hours before ELISA analysis, the diluted extracts were taken from the freezer and thawed at room temperature. ELISA was performed using a Progesterone ELISA kit (DE 1561; Demeditec, Kiel, Germany) and a Testosterone ELISA kit (DE 1559; Demeditec, Kiel, Germany), with concentrations in the fecal extracts being measured in duplicate, following the manufacturer’s instructions, and absorbance measured at 450 nm using a Multiskan FC microtiter plate reader (Thermo Fisher Scientific, Waltham, United States). The results obtained for progesterone and testosterone metabolite concentrations in terms of nanograms per milliliter of extract were converted to nanograms per gram of dry feces. The diluted extracts (first dilution step) containing concentrations of progesterone metabolites that exceeded the detection limit of the ELISA (4,000 ng/g) were further diluted (100 μL in 900 μL PBS—second dilution step) and determined again.

As partial validation of the measurements, we determined intra- and inter-assay coefficients of variation (CV), sensitivity, recovery, linearity and stated cross-reactivity provided by manufacturer. To determine CVs for the progesterone test, two fecal extracts were measured 10 times in a single assay and four times in each additional assay, whereas for testosterone, the two fecal extracts were measured four times in one assay and four times in each additional assay. For progesterone metabolites, we accordingly obtained intra-assay CV values of 8.27 and 12.11% for samples containing 1980 and 2,155 ng/g metabolites, respectively, whereas the corresponding inter-assay CV values for the same samples were 8.18 and 16.05%. Similarly, for testosterone metabolites, intra-assay CV values of 7.45 and 5.78% were obtained for samples containing 195 and 710 ng/g testosterone metabolites, respectively, with corresponding inter-assay CVs of 4.66 and 10.06%.

The sensitivity of the assay for detecting fecal progesterone and testosterone metabolites is reflected in the detection ranges of the respective kits. For progesterone metabolites, the detection range was 40–4,000 ng/g when using fecal extracts with the initial (first) dilution step. Samples with concentrations exceeding 4,000 ng/g, were subjected to an additional (second) dilution, extending the measurable range to 400–40,000 ng/g. Overall, the combined detection range for fecal progesterone metabolites was 40–40,000 ng/g. For testosterone fecal metabolites, the detection range of the assay measurements was 20–1,600 ng/g.

Recovery rates were assessed using dried fecal samples with previously determined concentrations of progesterone and testosterone metabolites. Aliquots of 0.5 g of dry fecal samples were spiked with 1,500 ng of progesterone standard and either 100 ng and 500 ng of testosterone standard (both from Sigma-Aldrich, St. Louis, USA), previously diluted in methanol. After spiking, the samples were subjected to the same methanol extraction protocol as described above. Recovery was calculated comparing the measured hormone concentrations to the expected values, yielding recovery rates of 113% for progesterone and 96 and 115% for testosterone, respectively.

Linearity was assessed using two fecal samples: one with previously determined high concentration of progesterone and the other with a high concentration of testosterone metabolites. Fecal extracts from both samples were serially diluted with PBS in a stepwise 1:1 ratio. The results are presented in Table 1.

The coefficients of determination (R²) were 0.986 for progesterone metabolite and 0.992 for testosterone metabolite measurements.

Details of the cross-reactivity of the ELISAs with other steroids were provided by the manufacturer. The cross-reactivities of the progesterone ELISA test are as follows: pregnenolone 0.35%; 17α OH progesterone 0.3%; corticosterone 0.2%; 11-deoxycorticosterone 1.1%; and individual natural estrogens, testosterone, and other glucocorticoids ≤0.1%. The cross-reactivities of the testosterone ELISA are as follows: dihydrotestosterone 12.9%; 11β-hydroxytestosterone and 19-nortestosterone 3.3%; androstenedion 0.9%; 5α-dihidrotestosterone 0.8%; and epitestosterone, progesterone, cortisol and individual natural estrogens <0.1.

Biological evaluation was only partially performed, as the exact age, reproductive status and health condition of individual animals were unknown. However, it is likely that most females were pregnant in April and lactating in July, based on the species’ reproductive biology. The observed high fecal progesterone metabolite concentrations in April and low concentrations in July are consistent with this expected physiological pattern and support biological relevance of the assay. Furthermore, we found a significant positive correlation between the concentrations of testosterone and progesterone fecal metabolites as detailed in the Results section. A positive correlation between testosterone and progesterone has been described previously (27, 28), so our results showing this phenomenon can be considered as evidence of a correct method of analysis. Collectively, these results provide partial biological validation of the applied method.

The sex of the animals was determined by detecting the SRY gene in the DNA extracted from fecal samples, as described previously (3). Of the 246 fecal samples analyzed, 125 were assessed to be derived from females and 121 were from males. These results are in line with the expected proportions and enabled us to determine the composition of the studied animal groups, as shown in Table 2.

Data analysis was performed with R¹ using the stats package for model fitting, emmeans for post hoc comparisons, and ggplot2 for data visualization. Fecal testosterone and progesterone metabolite concentrations were analyzed using generalized linear models (GLMs) with a gamma distribution and a log-link function suitable for modeling continuous positive data with a skewed distribution. Each hormone was modeled using a three-way interaction between sex, population status, and season as follows:

and

Model selection was based on the lowest Akaike’s information criterion (AIC) and the lowest residual deviance compared with those of simpler models based on two-way interactions or only main effects. Model adequacy was assessed by examining the quantile residuals, and influential observations were identified using Cook’s distance and leverage values. Observations with unusually high values, particularly those with leverage exceeding twice the average were flagged for further inspection. Data points deemed physiologically implausible or overly influential were excluded prior to refitting the model. Model fit was assessed using McFadden’s and Nagelkerke’s pseudo-R², along with the AIC values.

Pairwise comparisons were performed to assess the differences between captive and free-ranging individuals within each sex and season using the estimated marginal means. Sidak’s correction, which maintains statistical power while adjusting for the family-wise error rate, was applied to control for multiple comparisons. Owing to the log-link function, all comparisons were performed using a logarithmic scale. The estimates were back-transformed by exponentiation, and the results are reported as multiplicative effects (ratios) rather than absolute differences in logarithms. The correlation between fecal progesterone metabolite and testosterone metabolites concentrations was assessed using Pearson’s correlation coefficients. To account for seasonal and sex-specific variations, analyses were conducted separately for each combination of season and sex. For tests, the statistical significance was set at p < 0.05.

The descriptive statistics pertaining to data obtained for the fecal testosterone metabolites are shown in Table 5. In the analysis, we identified three measurements as outliers, which were accordingly removed prior to model fitting (Table 4).

The GLM applied for analysis of fecal testosterone metabolite concentrations utilizing a gamma distribution was found to have moderate explanatory power, as indicated by Nagelkerke’s pseudo-R² value of 0.39, whereas the value obtained for McFadden’s pseudo-R² was lower at 0.038. The fit of the model, as indicated by the AIC, was 2,926, with null and residual deviance values of 62.3 and 39.5, respectively. A summary of the model estimates and their confidence intervals is provided in Supplementary Table A, with the model-predicted values being shown in Figure 3.

Post hoc comparisons of fecal testosterone metabolite concentrations in samples obtained from captive and free-ranging individuals of G. gazella among seasons and in both sexes are shown in Table 6. Overall, we established that the feces derived from free-ranging individuals contained consistently higher and more variable concentrations of testosterone metabolites than the samples obtained from captive individuals.

With respect to females, we detected significant differences among the samples collected in December, April, and July, when free-ranging individuals had approximately 2.0- (p = 0.039), 2.3- (p < 0.001), and 1.6-fold (p = 0.001) higher testosterone metabolite levels, respectively. Moreover, seasonal variation was established to be more pronounced in free-ranging females, peaking in April, whereas in captive females, concentrations tended to relatively stable across the seasons.

For males, we detected similarities between the free-ranging and captive populations with respect to seasonal patterns in testosterone metabolite concentration, free-ranging males were found to have consistently higher concentrations throughout the year, which were 2.0-, 1.67-, and 1.61-fold higher in December (p = 0.001), July (p = 0.001), and September (p = 0.007), respectively.

Our analysis of fecal progesterone metabolite concentrations revealed seven measurements deemed to be outliers, which were accordingly removed prior to model fitting. Descriptive statistics for the remaining fecal progesterone metabolites data are presented in Table 7.

The GLM developed for analysis of fecal progesterone metabolite concentrations using a gamma distribution was found to have moderate explanatory power, as indicated by a Nagelkerke’s pseudo-R² value of 0.50, whereas McFadden’s pseudo-R² was lower at 0.034. Model fit as assessed using the AIC was 3,683, and the null and residual deviances of the model were 259.9 and 162.8, respectively. A summary of the model estimates and confidence intervals can be found in Supplementary Table A, with the model-predicted values being shown in Figure 4.

Post hoc comparisons of progesterone metabolite concentrations in samples obtained from captive and free-ranging populations across seasons and for both sexes are presented in Table 8. Overall, we detected comparable concentrations of progesterone metabolites in samples derived from males and females, although we noted a distinct seasonal pattern. In females, concentrations were elevated from December through April, followed by a decline in July and September. Contrastingly, for males we detected a single peak in April, whereas levels were relatively stable and lower in the remaining months.

Furthermore, a comparison with respect to population status revealed differences in certain seasons. Although progesterone metabolite levels were generally low in July, the concentrations in samples obtained for free-ranging females were 3.4-fold higher than those in the samples obtained for captive females (p = 0.002). Similarly, in September, concentrations in samples derived from free-ranging males were 3.5-fold higher than those in samples from captive males (p = 0.006). Moreover, throughout the year, free-ranging individuals of both sexes showed a wider dynamic range of progesterone metabolite levels than the captive individuals.

It is important to note that only two samples were obtained from captive females in December, thereby contributing to highly uncertain model estimates. The wide confidence intervals observed for this group reflect this limitation and should accordingly be interpreted with caution.

The dataset used for analyses of the correlations between fecal progesterone and testosterone metabolite concentrations was the same as that used for modeling, with the previously identified outliers excluded. The results obtained for separate evaluations performed for each season and sex are summarized in Table 9 and illustrated in Figure 5 for females and Figure 6 for males.

Among females, we detected moderate positive correlations in July (r = 0.50, p < 0.001) and September (r = 0.35, p = 0.025), whereas weaker associations were observed in December (r = 0.27, p = 0.269) and April (r = 0.51, p = 0.051). Contrastingly, for males, there were positive correlations in all seasons, with stronger associations being detected in April (r = 0.50, p = 0.007), July (r = 0.38, p = 0.023), and September (r = 0.51, p < 0.001). Thus, the strongest associations in males and females were apparent in September and July, respectively.