The research received approval from the Human Research Ethics Committee (HS1/2564) and the Institutional Animal Care and Use Committee at the Faculty of Veterinary Medicine (FVM), Chiang Mai University (CMU), located in Chiang Mai, Thailand (FVM-ACUC; S4/2564).

A total of 58 adult elephants (14 males and 44 females) from six elephant camps in the Mae Taeng district of Chiang Mai province was studied and initially described by Supanta et al. (15, 16) (Table 1). Preceding the nationwide lockdown, elephants participated in five tourist-related activities: riding with a saddle, riding bareback, elephant shows, feeding, bathing or observation only. As a result of the lockdown, all tourist-related activities ceased. The inclusion criteria for elephants in this study were: adult age (18–50 years); deemed healthy by camp veterinarians or veterinary consultants; females were not pregnant. Samples were not collected from males during musth for safety reasons. These are the same elephants included in the two companion papers (15, 16).

Data collection for this prospective study began in April 2020 (T01), a month after the implementation of a travel ban due to the COVID-19 pandemic, and continued through April 2022 (T07) as camps began to open again to tourists. The study consisted of in-person interviews with camp owners, veterinarians and/or mahouts (elephant keepers) to gather information about the numbers of tourists, elephants, and mahouts, tourist activities (riding, walking, shows, feeding), chaining duration and length, and provisioning of roughage and supplements (14). The questionnaire is shown in Table 1. Interviews were conducted, and samples were collected monthly for the first year [April 2020 to April 2021 (T01-T04)] and then every 4 months during the second year [May 2021 to April 2022 (T05-T07)]. This methodology led to a total of 16 on-site visits that were subsequently segmented into seven time periods for analysis [April 2020 (T01), May–August 2020 (T02), September–December 2020 (T03), January–April 2021 (T04), May–August 2021 (T05), September–December 2021 (T06), January–April 2022 (T07)] as described previously (15, 16). Queries about camp management before COVID-19 were included in the first interview to capture data on operations in 2019 (16). A total of seven camp managers, 50 mahouts, and two veterinarians were interviewed across the 16 data collection periods. Supplementary information on camp management and elephant activities pre-COVID-19 was sourced from Bansiddhi et al. (4, 11, 12) and Norkaew et al. (5–7).

Fecal samples were collected monthly in the morning between 08.00 to 12.00 h to avoid variation in the circadian rhythm of cortisol secretion (23, 48, 49). Defecation was either observed or indicated to have occurred within the hour by the mahout. Fecal boluses were broken open and approximately 50 grams of material taken from the middle and placed into zip lock plastic bags labeled with the elephant name, time of collection and date. Samples were transported at 2–8°C in a cool box to the Faculty of Veterinary Medicine, Chiang Mai University and then stored at −20°C until fecal extraction and hormonal analysis.

Blood (~8 mL) was collected from a caudal auricular vein and placed into two tubes: serum separator tubes (SST) (~7 mL) for analysis of oxidative stress (MDA, 8-OHdG), and ethylenediaminetetraacetic acid tubes (EDTA) (~1 mL) for the stress leukogram (heterophils, lymphocytes, monocytes and eosinophils). Blood in SST tubes was centrifuged at 1,200×g for 15 min within 6 h of collection (50), while whole EDTA blood was analyzed within 24 h of blood smear preparation. Serum was stored at −20°C until processing and analysis.

Fecal samples were extracted as described by Norkaew et al. (5). Frozen feces were thawed at room temperature (RT) and dried in a conventional oven (60°C) for 24–48 h. Mixed powdered feces (~0.1 g) were placed into glass tubes, 4.5 mL of EtOH and 0.5 mL of distilled water was added, and tubes were vortexed briefly. Samples were extracted by boiling in a water bath (90°C) for 20 min, with 95% EtOH added to keep the volume at 5 mL, and then centrifuged at 1,200×g for 20 min. The fecal extracts were combined, dried in a 90°C water bath, re-suspended in 3 mL of 95% EtOH, dried down again, and finally re-suspended in 1 mL of 50% methanol. Samples were stored at −20°C until analysis.

Concentrations of fGCM were measured in extracts diluted 1:3 in assay buffer (0.0137 M Trizma base, 0.2 M Tris–HCl, 0.2 M NaCl, 0.2 M EDTA, 0.001% BSA, and 0.001% Tween 20; pH 7.5) using a double-antibody EIA with a polyclonal rabbit anti-corticosterone antibody (CJM006, Coralie Munro, UC Davis, CA) previously used in Thailand (5). Samples and corticosterone standards (50 μL) were added to wells in duplicate followed by corticosterone-HRP (25 μL; 1:30,000) and anti-corticosterone antibody (25 μL, 1:100,000). Plates were incubated in the dark at RT for 2 h, followed by 100 μL of TMB solution and incubation for 20–35 min. Stop solution was added (50 μL, 1 N HCl) and absorbance measured at 450 nm (TECAN, Männedorf, Switzerland). Assay sensitivity was 0.12 ng/mL. Intra- (at 90% binding) and inter-assay coefficients of variation (CV, based on dose values) were < 10% (all duplicates over 10% were reanalyzed) and 12.28%, respectively.

MDA was quantified by a thiobarbituric acid (TBA)-reacting substances (TBARS) assay described previously (51, 52). Briefly, 50 μL of serum samples and standards were mixed with 750 μL of 0.44 M phosphoric acid, 250 μL of 42 mM TBA and 450 μL of distilled water. Tubes were placed into a boiling water bath for 15 min, cooled on ice for 5 min, and then centrifuged at 900×g for 5 min. Absorbance of the supernatant was measured at 532 nm using a UV–VIS spectrophotometer (Shimadzu, Japan). Serum MDA concentrations were calculated from standard curves of MDA equivalents generated by the acid-catalyzed hydrolysis of 1,1,3,3-tetramethoxypropane (TMP) within the range of 1.25–80.00 μM. Assay sensitivity was 0.15 μM. Intra- and inter-assay coefficients of variation (CV) were 1.1 and 2.3%, respectively.

Serum 8-OHdG was measured using a DetectX^® DNA Damage Enzyme Immunoassay Kit (Cat #K059-H5, Arbor Assays) validated for Asian elephants (52, 53). Assay sensitivity was 0.072 ng/mL and intra-assay and inter-assay CVs were < 10 and 6.3%, respectively.

White blood cell count differentials were done manually on blood smears stained with Wright–Giemsa (RVL Supply, Pathumthani, Thailand) (45). A ratio of heterophils to lymphocytes was calculated for each sample.

Elephant data from management surveys are presented as mean ± standard error of the mean (SEM) and a range depending on the type of data. Statistical analyses were conducted using R version 4.3.1 (54) package geepack, function geeglm, under the optimal autoregressive order 1 correlation structure (55). Repeated measures data were analyzed using Generalized Estimating Equations (GEE) to determine the effect of camp management practices on stress parameters. A univariate analysis was performed, and any variable with a significance level of p < 0.15 was included in the subsequent multivariate analysis. Post hoc tests were conducted using Tukey’s multiple comparisons of least-squares mean (LS-mean) to identify differences between variable categories. For the multivariate model, Sex and Age were incorporated as covariate factors. The selection of the final model in the GEE analysis was determined based on the smallest quasi-likelihood under the independence model criterion (QIC) values (56) using package MuMIn, function dredge (57) with the significance set at α = 0.05. Reference categories were chosen based on specific questions as described previously (11, 12), including Sex = male; Time = T01 (April 2020, the beginning of the travel ban); and Camp = F (the smallest camp). Differences in mean fGCM, oxidative stress markers, stress leukograms and H/L ratio among three seasons (summer: February 16–May 15; rainy: May 16–October 15; winter: October 16–February 15) (53, 58, 59) were analyzed by Tukey’s post-hoc tests following GEE analyses.

Camp differences in management practices before the COVID-19 travel ban (T00) compared to the start of the lockdown (T01) and each subsequent year (T04, T07) are described in Supanta et al. (15) and Table 1. A summary of Time and Camp differences in biomarkers of adrenal activity, oxidative stress, DNA damage, stress leukogram, and the heterophil/lymphocyte ratio are shown in Supplementary Table S1. Univariate and multivariate GEE results are presented in Supplementary Tables S1–S9. Table 2 summarizes the results of multivariate GEE analyses of category and continuous camp management variables associated with stress biomarkers, with LS-mean data from the final GEE models presented in Table 3. Stress biomarkers were compared to published reference ranges before COVID-19 in Table 4. Moreover, seasonal effects are presented in Supplementary Table S10.

Overall average fGCM concentration was 52.9 ± 1.1 ng/g (21.2–147.4 ng/g), with lower overall concentrations observed in T01 (p < 0.001) (Figure 1 and Table 4). Variables significantly affecting fGCM concentrations in the multivariate model were Time and Camp (Tables 2, 4 and Supplementary Table S2), with a rising trend in three of six camps (Camps A, C, E) when data were compared across T01, T04 and T07 time periods (Supplementary Table S1). Elephants at Camps B and F exhibited the highest fGCM concentrations, while those at Camp C had the lowest overall. There was a seasonal effect on fGCM, with the lowest concentrations in the summer and highest values in the rainy season (Supplementary Table S10).

Serum MDA concentrations averaged 2.2 ± 0.5 μM (1.3–3.8 μM), and were affected by Time, Chain length, Chain hour and Roughage in the multivariate analysis (Tables 2, 4 and Supplementary Table S3). Chain hour and Roughage were positive relationships, while Chain hour was negative (Table 1 and Supplementary Table S3). There was a declining trend over time (Figure 2A and Table 4) in all six camps (Supplementary Table S1), but no seasonal effect (Supplementary Table S10).

Overall average serum 8-OHdG concentration was 7.5 ± 1.4 mg/mL (3.3–18.2 mg/mL) and related to Time, Camp, Chain length and Chain hour in the final multivariable model (Tables 2, 4 and Supplementary Table S4). Chain length was positively associated with serum 8-OHdG concentration, while a negative relationship with Chain hour approached significance (Table 2 and Supplementary Table S4). There was an increasing trend over the 2-year period (Figure 2B and Table 4), albeit significant in only two camps (Camp C and E) (Supplementary Table S1). Overall, Camp F had consistently higher 8-OHdG values, while Camp B had among the lowest concentrations. There was no seasonal effect on serum 8-OHdG concentrations (Supplementary Table S10).

Overall average heterophil counts were 2.9 ± 0.1 × 10³ cells/μL (1.1–8.9 × 10³ cells/μL) and affected only by Roughage (positive) in the multivariate analysis (Tables 2, 4 and Supplementary Table S5). There were no differences among Time periods (Figure 3A and Table 4), and counts were not affected by season (Supplementary Table S10).

Monocyte counts averaged 3.7 ± 0.1 × 10³ cells/μL (0.0–8.7 × 10³ cells/μL). From the multivariate analysis, factors that had the greatest influence were Age, Time, and Camp (Tables 2, 4 and Supplementary Table S6). Monocyte counts were higher in younger elephants (Table 1 and Supplementary Table S6) and exhibited an increasing trend over the 2-year period (Figure 3B and Table 4). At each time period, Camp A elephants consistently had lower monocyte counts (Supplementary Table S1). There was no seasonal effect (Supplementary Table S10).

Mean lymphocyte counts overall were 4.2 ± 0.1 × 10³ cells/μL (1.1–9.9 × 10³ cells/μL). From the multivariate analysis, Age, Time, and Chain hour had the greatest influence (Tables 2, 4 and Supplementary Table S7), with a negative relationship to Age and positive relationship to Chain hour (Table 2 and Supplementary Table S7). An overall declining trend over time was observed (Figure 3C and Table 2) that was significant in four camps (Camps A, C, E, F) (Supplementary Table S1). There was no effect of season (Supplementary Table S10).

Eosinophil counts averaged 0.3 ± 0.0 × 10³ cells/μL (0.0–1.7 × 10³ cells/μL), and from the multivariate analysis, Age, Time, Camp and Roughage had significant effects. Eosinophil counts were lower in older elephants, and there was a positive relationship with Roughage (Table 2 and Supplementary Table S8). However, changes over time revealed no discernable pattern (Figure 3D and Table 4), and neither were their seasonal effects (Supplementary Table S10). Elephants at Camp C had the lowest eosinophil counts, while those at Camp F were generally the highest (Table 4).

For the overall H/L ratio (0.8 ± 0.01; range, 0.2–3.4) Age, Time, Camp and Walking distance had significant effects. H/L ratio was higher in older elephants, with a positive relationship to Walking distance (Table 2 and Supplementary Table S9) and increasing trend over the 2-year period (Figure 3E and Table 4) for four camps (Camps A, C, E, F) (Supplementary Table S1). Camp E had the highest H/L ratio, while Camp C was the lowest. There were no seasonal effects (Supplementary Table S10).

Concentrations of fGCM increased over time and were highest at T05 (May–August 2021), a year into the lockdown. The range in this study (16.9–239.7 ng/g) aligns with previous values in Asian elephants in northern Thailand [37.4–98.3 ng/g (5–7); 39.2–69.0 ng/g/feces (12); 25.8–39.3 ng/g (46); 22.2–106.1 ng/g (53)], albeit at the higher end as the travel ban progressed. While there are limited studies on how tourist or logging activities impact stress hormone responses, there are indications that yearly patterns of GCs may be related more to work activities than environmental influences (52, 53). In India, fGCM concentrations were higher in elephants involved in an intensive and crowded public festival, but not other tourist activities (61). In Myanmar, elephants have the best body condition, but also experience the highest fGCM concentrations as they enter the more stressful logging and monsoon seasons (62). By contrast, a different study of the Myanmar population found the highest fGCM concentrations in the winter and lowest in the monsoon season, with intermediate values in the summer (63). That study was conducted on non-working elephants only; thus, it appears to demonstrate a true seasonal pattern.

In Thailand, a series of correlational studies aimed at identifying captive management factors that exert significant influences on welfare outcomes have been conducted by our group over the past 6 years (4–7, 11, 12, 14–16, 45–47, 49, 52, 53, 64). Lower fGCM concentrations were found to be associated with riding or walking activities, being kept in open areas during the day, and in the forest or under a tree at night (12). High BCSs were linked to limited exercise and feeding of high-calorie treats like banana and sugarcane, and also higher fGCM concentrations (5–7); all these were similar to findings observed in U.S. zoo elephants (8, 9). However, the relationship between fGCM and BCS during COVID-19 differed, as there was a slight, but significant negative relationship between fGCM and BCS (Spearman’s rank correlation; rho = −0.14; p = 0.0172). So, as body condition decreased, presumably due to limited roughage and supplements, fGCM increased. The reason for this is not clear, as there were no clear relationships between fGCM and management factors found to be important in earlier studies, like Walking distance, Chain hours, Chain length, Roughage or Supplements. Rather it could simply be due to the fact that elephants during the COVID-19 pandemic were chained most, if not all, of the time and had nothing to do, and body condition was reduced to levels not seen in earlier studies. In a study of wild elephants in India, fGCM were highest in individuals with the lowest BCSs, lower than those observed in this study (65). So, it might not be surprising there were little differences among camps in elephant adrenal responses to management, as it was all inadequate. Follow up studies are now planned as tourism begins to return normal in Thailand, although as of the end of 2023 it had not.

A significant new finding was related to season, where the pattern in this study was opposite to that found earlier. Previously, higher fGCM concentrations during the winter months suggested a possible link between colder temperatures and increased adrenal activity (5, 6, 46). However, those months (November–February) also correspond to the high tourist season (5, 6), so until now, it has been impossible to tease apart tourist activity versus climate conditions on adrenal activity. But this and another study (53) conducted during the COVID-19 lockdown strongly suggest that tourist activities are the most likely cause of increased fGCM excretion during the winter, high tourist season months. Thus, further studies are needed to better understand not only what tourist activities affect elephant health and welfare, but how. That said, it is critical to consider context, as increases in adrenal GC activity can occur in response to both positive and negative stimuli, so higher concentrations do not necessarily equate to poor welfare, nor do lower levels indicate good welfare (66).

Serum MDA concentrations (1.3–3.8 μM) were similar to previous ranges [0.93–3.0 μM (44, 45)], but showed a declining trend as the study progressed, with T01 (April 2020) being the highest and T07 (January–April 2020) the lowest. There was no obvious seasonal effect, which agrees with another study conducted during the same time period in Thai elephants (53), although a follow-up study did find somewhat lower concentrations during the rainy season (52). Polyunsaturated fatty acids are major substrates for lipid peroxidation (59, 61, 62), and MDA is an end-product produced by the decomposition of the lipid peroxidation process (61). We recently found that elephants during COVID-19 received reduced high-energy diets from a lack of tourists (16), resulting in lower lipid concentrations and BCSs (16). Thus, one possible explanation is a reduction in fat consumption due to dietary changes, as the factor, Roughage was positively associated with MDA concentrations and generally contains 4.1–5.5% crude fat (64). During COVID-19, amounts of roughage decreased, leading to possible reductions in lipid intake. Moreover, roughage, especially young Napier grass, has antioxidant properties (67) known to inhibit the activity of free radicals (68, 69), which might also explain the observed lower MDA concentrations during the lockdown.

Several studies have suggested monitoring serum 8-OHdG as an indicator of DNA damage might be a useful biomarker for both acute (70, 71) and chronic stress (72). In this study, serum 8-OHdG concentrations (3.7–15.2 ng/mL) were similar to a previous study in elephants (3.3–18.2 ng/mL, (52, 53), but with an increasing trend over time, being lowest at the beginning and highest at the end of the study. A previous study found inactive mice had higher oxidative stress markers (lipid hydroperoxides, vascular superoxide release and nicotinamide adenine dinucleotide phosphate oxidase) than active mice (42). In humans, positive correlations have been observed between urinary 8-OHdG and average working hours and serum cortisol concentrations (69), suggesting that increased GC activity may lead to higher levels of oxidative DNA damage (37). In zoo-housed grizzly bears (Ursus arctos horribilis), fecal 8-OHdG concentrations were positively related to fGCM concentrations (71). Thus, it is possible that the increased GCs noted in this study may have contributed to increased DNA damage (73), although other factors may certainly have been involved.

In human and animal medicine, a stress leukogram is characterized by neutrophilia, lymphopenia, eosinopenia, and monocytosis, and specifically refers to the effects of corticosteroids on white blood cell populations (36). While the hematology values during COVID-19 were generally within the range of healthy elephants (45, 74, 75), comparatively, data point to the presence of monocytosis and lymphopenia, two of four criteria in a stress leukogram. Monocytosis was most likely the result of a shift from a marginalized to a circulating population of monocytes. Changes in the expression of adhesion molecules and chemotactic cytokines that interfere with the trafficking of leukocytes into tissues (76) could play a role in changes in WBC populations. Lymphopenia is the result of an immediate shift of lymphocytes from the circulating blood to other tissues, as well as decreased lymphopoeisis (77). In addition, GCs can induce apoptosis of lymphoid cells in mammals (78) and thus interfere with immune functionality. Although no elephants became visibly sick during the study, we cannot rule out subclinical changes as a result of long-term stress associated with restrictive management changes.

The H/L values in this study ranged from 0.1 to 2.5, which agrees with earlier reports of 0.30 to 2.8 (34) and 0.2 to 3.5 (63) in captive Asian timber elephants. The H/L ratio reflects the ability of an animal to cope with infection through injury (via heterophils) or disease (via lymphocytes) (79). While heterophils form the first line of immune defense against bacterial pathogens in inflammatory lesions, lymphocytes are involved in host defense through cytokine production and immune responses (80). As early as 1989, the H/L ratio has been used as an indicator of welfare and how stress affects immune function in response to adrenal-corticoid hyperactivity (81, 82), with lower ratios suggesting better immune function or as a proxy of lower GC responses. During the travel ban, H/L ratios increased over time, which agrees with concomitant increases in GCs that are known to affect the H/L. In elephants, Ukonaho et al. (63) found elephants in Myanmar had higher fGCM concentrations and H/L ratios during the cold season, indicating increased stress. Heterophilia occurs only under stressful conditions, similar to the situation with monocytes, but in this study was not enough to be statistically different. In relation to management factors, there was a positive correlation between lymphocyte numbers and Chain Time, which is consistent with a report showing captive rodents presented higher lymphocyte counts than free-ranging ones (83). Thus, results further suggest elephants in Thailand were exposed to chronic stress during the COVID-19 lockdown.