Physiological and behavioral responses of house sparrows to repeated stressors

Experimental design

To test the physiological and behavioral effects of repeated exposure to stressors, 22 wild free-living house sparrows (12 females, 10 males) were caught June 1–3 2016 in Medford, MA, USA. Individual birds were randomly assigned to one of two treatment groups and housed in cages in separate rooms according to this distinction. Randomization occurred such that each group contained both males and females and such that no group contained birds that were all captured on the same day. Birds were housed in pairs and/or in view of other birds in cages (45 cm × 37 cm × 33 cm) and kept on a natural 15L:9D light cycle. An acclimation period occurred for at least 2 weeks during which birds were disturbed only for routine husbandry. This period was determined based on previous research involving house sparrows, which showed hormonal acclimation occurring after 2 weeks in captivity (Fischer, Wright-Lichter & Romero, 2018).

This study was a repeated measures design in which the responses of individual birds during the experiment were compared to pre-stress samples. At the end of the initial 2-week acclimation period, a pre-experiment blood sample was taken (“Pre-Stress Control” in Fig. 1). This sample served as the control to which the responses to chronic stress were compared. We did not include a separate control group of birds that did not experience the chronic stress treatment because prior research has shown that no significant hormonal changes occur to house sparrows following acclimation (Dickens et al., 2009c; Lattin & Romero, 2014; Rich & Romero, 2001). The chronic stress protocol commenced 10 days after the end of the acclimation period. Birds in treatment group A received stressor A for four days followed by four days of stressor B, while birds in treatment group B received the stressors in the opposite order (Fig. 1). This was to ensure that the results were not stimulus or order-dependent. Stressor A (cage rolling) involved the cages being placed on lab carts and being gently rolled back and forth for 30 min twice per day. During stressor B (cage tapping), one person entered the experimental room every 2.5 min and tapped on the cages; this persisted for 30 min twice per day. The birds experienced these stressors at unpredictable times throughout the day, but always twice per day. Repeated variable application of acute stressors has been shown used to elicit physiological and behavioral changes consistent with chronic stress in many species (e.g., Cyr et al., 2007; Rich & Romero, 2005), including using both of these stimuli in previous studies involving house sparrows (DuRant et al., 2016a). Each stressor was presented for 4 days; a time period chosen based on previous studies that suggested physiological changes indicative of chronic stress occur over eight to ten days of stimulus presentation (Cyr et al., 2007; Rich & Romero, 2005).

Animals were collected under a Massachusetts state collection permit and all experiments were conducted with the approval of the Tufts Institutional Animal Care and Use Committee and in compliance with the Guidelines for Use of Wild Birds in Research (Fair, Paul & Jones, 2010).

Plasma sampling

Blood samples were taken at three time points to assess HPA axis functioning, immune capacity, and changes in uric acid concentrations (Fig. 1). The control sample was taken 10 days prior to exposure to the experimental stressors and at least two weeks after initiation of captivity. We decided on this separation between the control sample and the start of the stress protocol to remain under the ethical requirement for volume of blood for sampling. A second sample (“4 days” in Fig. 1) was taken after four complete days of the first stressor and a third and final sample (“8 days” in Fig. 1) was taken after the entire eight-day protocol. On each of these days, a series of blood samples was taken starting at 9:00am. Baseline samples were taken within 3 min of entering the room (Romero & Reed, 2005). Birds were then placed in an opaque cloth bag for 30 min, after which a stress-induced sample was taken. At this time, 1 mg/kg dexamethasone (12.5 µL; DEX; Phoenix Pharmaceuticals, Inc., St. Joseph, MO, USA), a synthetic glucocorticoid, was injected into the pectoralis muscle (30 G, BF PrecisionGlide™ Needle, Becton, Dickinson and Company, Franklin Lakes, NJ, USA) to assess efficacy of negative feedback in the HPA axis. Birds were sampled again 90 min after DEX injection, then 100 IU/kg ACTH (10 µL; Sigma Aldrich, St. Louis, MO, USA; Catalog No. A6303) was injected as before to assess the maximum capacity of the HPA axis. The final blood sample was taken after an additional 15 min of restraint. All blood samples were taken from the alar vein and collected in heparinized capillary tubes. Samples were stored on ice until processing, which occurred no more than 5 h after sampling. Centrifugation at ∼1,200 g for 10 min (Centrific Model 225, Fisher Scientific, Pittsburgh, PA, USA) was used to separate plasma, which was frozen at −20 °C until assay. Separate aliquots were created from the baseline samples in order to conduct CORT, bacterial killing, and uric acid assays.

Corticosterone assays

Radioimmunoassays (RIA) were used to quantify CORT concentrations in the baseline, stress-induced, negative feedback, and Max plasma samples (Wingfield, Vleck & Moore, 1992). Briefly, steroids were extracted from plasma samples (13.3 ± 5.9 µL) using dichloromethane, dried under N₂ gas, and then rehydrated with 550 µL phosphate-buffer saline with gelatin. A standard curve was created and the RIA was run using the antibody B3-163 (Esoterix, Calabasas Hills, CA, USA). Assay sensitivity was 1 ng/ml; baseline and negative feedback samples were often below the level of detection (20 samples) and therefore assigned this floor value. The inter and intra-assay CVs were 7.6% and 3.5% respectively.

Bacterial killing assays

Baseline plasma samples were also used for in vitro analyses of bacterial killing capacity. All samples were assayed within 10 days of sampling as longer periods of freezing have been shown to dramatically affect killing (Liebl & Martin, 2009). This assay assesses the proficiency of plasma antibodies and proteins to combat a pathogen and it has been validated for house sparrows (Liebl & Martin, 2009). Briefly, 1.5 µL of plasma was diluted with 34.5 µL PBS and was incubated with 12.5 µL of Escherichia coli (10⁴ bacteria/mL; ATCC^® 8739) at 37 °C for 30 min. Then, 250 µL of tryptic soy broth (TSB) was added to each tube, followed by an additional 12-hour incubation at 37 °C. Positive controls with only E. coli and TSB were also created. Finally, a NanoDrop spectrophotometer (NanoDrop 2000, Thermo Fisher Scientific, Waltham, MA, USA) was used to quantify absorbance of each sample at 300 nm. Killing capacity was calculated as 1 –(absorbance of sample / absorbance of positive control). All samples were run in triplicate when possible; only 4 samples out of all those that were assayed had to be run in duplicate. The inter-assay CV, based on the positive controls, was 11.5%.

Uric acid fluorometric assay

Baseline plasma samples were used to quantify uric acid using the Amplex^® Red Uric Acid/Uricase Assay Kit (Molecular Probes, Eugene, OR, USA). This assay relies on the conversion of uric acid to hydrogen peroxide, which subsequently reacts with Amplex^® Red reagent in the presence of horseradish peroxidase (HRP) to form a red-fluorescent product. Briefly, plasma samples were diluted with 200 µL of the provided reaction buffer and a uric acid standard curve was generated using the provided reagents. To have enough reagents to run all samples, the Mastermix was slightly altered to include 150 µL of Amplex^® Red reagent and 85 µL of the reaction buffer. To account for this change, spectrophotometric readings were taken at 30 and 45 min at 560 nm. No statistical differences were found between these samples and therefore all further analysis was completed using the 30-minute sample. A negative control, lacking uric acid, and a maximum-reading positive control were also included on the 96-well plates. Four separate assays were completed and samples were run in triplicate. The inter-assay CV from the positive controls was 7.3%.

Neophobia and perch hopping trials

In addition to the physiological data gathered from plasma sampling, changes in behavior were assessed by measuring the birds’ responses to novel objects. Three neophobia trials were conducted prior to any stimuli, after five days of the chronic stress protocol, and after seven days of the chronic stress protocol. Because blood and behavioral samples could not be taken the same day, we were limited which days the neophobia trials could occur. Beginning five days prior to the first neophobia trial, food dishes were removed each night (within 1 h before lights off) and replaced each morning (within 1 h after lights on) to acclimate the birds. At this time, the cage liners were also replaced to remove any excess food was removed from the bottom of each cage; this was to ensure that birds would reliably return to their food in the morning.

On the night directly before a trial, food dishes were removed and opaque blinders were placed between each cage to prevent the birds from seeing the neighboring trial, which could alter their own behavior (Fryday & Greig-Smith, 1994). The following morning, food was replaced with one of three possible novel objects—a colored plastic egg in the dish, a red ribbon across the dish, or a food dish painted red. All objects had been used previously in studies involving avian species and shown to elicit a neophobic response (Fischer & Romero, 2016). Behavior was recorded for 20 min after initial food presentation and then the novel objects and blinders were removed. Each individual’s latency to feed was recorded. Feed latency was defined as the time between replacing the food and the bird visibly feeding. Additionally, perch hopping was calculated during 15 min of the same trial. The first 5 min were discounted to avoid activity that occurred due to the presence of the experimenter replacing the food dish. A perch hop was defined as movement throughout the cage, typically between the perch, water dish, cage, and/or ground.

Three control trials were also conducted (one each day) prior to each neophobia trial to ensure that the stress treatments did not affect the birds’ normal feeding behavior. During these control trials, food dishes were returned in the morning without a novel object. Neophobia and control trials were conducted on days separate from blood sampling.

Data analysis

All statistical analyses were conducted in RStudio (RStudio Team, 2015). Before performing any linear mixed effects models or ANOVAs, the data were checked for heteroscedasticity using Bartlett’s test. In the event of a violation, data were log-transformed; this was the case with baseline and stress-induced CORT as well as with perch hopping.

Group and sex effects

There was no main effect of group in the datasets for stress-induced, negative feedback, and maximum capacity CORT, bacterial killing, uric acid, or neophobia datasets (p > 0.10 for all tests). There were also no significant interactions between group and sample (p > 0.54 for all tests). Therefore, stimulus order was removed from any further analyses of those datasets. We conducted the same check for sex effects and interactions and found none (p > 0.06 for all main effects tests and p > 0.28 for all interaction tests); sex was therefore removed from any further analyses in all datasets, except for the perch hopping data where we did detect a significant effect of sex.

Corticosterone responses

Group—the order that the stimuli were experienced in—significantly affected baseline CORT (F_1,60 = 5,73, p = 0.03; Fig. 2A). Birds in group A (experienced cage rolling first) had higher baseline CORT at each sample point. However, even when analyzed separately, we did not detect statistically significant differences in either group A (F_2,27 = 1.18, p = 0.33) or group B (F_2,29 = 1.57, p = 0.23) across the chronic stress protocol. We also did not detect any statistically significant differences in CORT levels at any of the other three bleed types (stress-induced, F_2,61 = 1.42, p = 0.25; negative feedback, F_2,60 = 0.07, p = 0.93; maximum capacity, F_2,63 = 1.24, p = 0.30) over the course of the chronic stress protocol (Figs. 2B–2D).

Immune capacity

The killing capacity of samples taken at the final time point was significantly lower than those for those at the control and second time point (F_2,62 = 19.7, p < 0.0005; Fig. 3).

Uric acid

Uric acid significantly decreased once the chronic stress protocol was initiated and remained low (F_2,61 = 12.8, p < 0.0005; Fig. 4).

Neophobia

We found no significant interactions between object type (red dish, red ribbon, plastic egg) and sample (F_4,61 = 0.58, p = 0.68). There was, however, a significant main effect of object (F_2,63 = 3.94, p = 0.03). Regardless of object type, feed latency decreased significantly after 4 days of the chronic stress protocol (F_2,63 = 4.03, p = 0.03; Fig. 5). Feed latencies also significantly decreased during control trials as the chronic stress protocol progressed (F_2,63 = 25.9, p < 0.00001; Fig. 5, white bars).

Perch hopping

Perch hopping was not significantly affected by neophobic object (F_2,56 = 1.02 p = 0.37) nor were there significant interactions between object and perch hops (object, F_4,54 = 0.37, p = 0.83). Perch hopping significantly decreased after the initiation of the stress protocol and remained reduced through the remainder of the protocol (F_2,56 = 11.16, p < 0.0005; Fig. 6A).

Perch hopping remained constant during the control trials. We did detect a significant main effect of sex; females were more active than males (F_1,57 = 6.23, p = 0.02), however neither females (F_2,30 = 0.27, p = 0.77) nor males (F_2,23 = 0.16, p = 0.86) significantly changed their activity during these control trials. We also found a significant interaction between group and perch hopping (F_6,52 = 7.52, p < 0.005) during the control trials when no novel objects were present. Further inspection of graphs revealed that this interaction was driven by birds in group A (experienced cage rolling first) increased their perch hopping at the final sample point, while those in group B (experienced cage tapping first) decreased their perch hopping (Fig. 6B).

Effects of chronic stress on HPA axis regulation

We found no significant changes in any of the four measurements of HPA axis regulation and responsiveness across the three sampling time points (Fig. 2). The lack of change in baseline CORT levels was not entirely surprising as a previous study using house sparrows and a rotating, repeated psychological stress protocol found no significant changes (Lattin & Romero, 2014). Interestingly, there was a main effect of group on baseline CORT. Despite randomizing the birds into groups A and B, group A birds tended to have higher baseline CORT, even prior to experiencing any stimuli (Fig. 2A). Regardless, when analyzed separately, we did not detect a statistically significant change in CORT over the course of the experiment; however, both groups appeared to exhibit a similar downward trend by day 8 (Fig. 2A). Additionally, stress-induced CORT levels were also trending down relative to the control levels at the start of the experiment; it is likely that had the experiment continued, we would have found statistically significant decreases, as shown in prior studies (Cyr et al., 2007; Cyr & Romero, 2007; Dickens, Earle & Romero, 2009b; Rich & Romero, 2005). We want to stress that our primary goal was not to induce significant changes in HPA axis function and regulation, but rather to push animals to the edge of when these changes might begin. Thus, this protocol was effective at pushing animals to the edge of experiencing homeostatic overload, at least as measured by metrics of the HPA axis. It is also unlikely that these results suggest acclimation or habituation to the chronic stress protocol since the stressors were encountered at random and unpredictable times throughout the day (see Rich & Romero, 2005).

Surprisingly, we found no changes in negative feedback strength or maximum CORT production throughout the course of the experiment (Fig. 2). We expected that negative feedback strength and production capacity would change as wear-and-tear increased throughout the protocol. Previous studies involving avian species have reported changes in both these measurements due to different types of chronic stress. For example, captivity combined with translocation of chukar (Alektoris chukar) significantly reduced feedback strength (Dickens, Delehanty & Romero, 2009a). In addition, HPA axis maximum capacity increased in house sparrows due to initial transference to captivity during two life history stages (late breeding and molt) (Lattin et al., 2012a). These differences suggest that experimental conditions play a role in the physiological changes that occur due to chronic stress. Though translocation, captivity, and a rotating, repeating, psychological stress protocol are each considered long-term “chronic” stressors, animals likely interpret these stressors differently; it is therefore plausible that the physiological changes that result from these stimuli are unique.

Immunological effects of chronic stress

Our results indicate that immune capacity was severely affected in chronically stressed birds. While birds maintained an in vitro immune response after the first bout of stressors, the building wear-and-tear throughout chronic stress protocol exposure induced immunosuppression following the second bout (Fig. 3). This finding was consistent with our predictions based on evidence that chronic stress leads to immunosuppression while acute stress can lead to immunoenhancement (Dhabhar, 2002; Glaser et al., 1987; Martin, 2009; McEwen, 1998). Furthermore, the data presented here are consistent with a previous study in which elevating allostatic load through topical CORT application slowed house sparrows’ healing of a superficial wound (DuRant et al., 2016a). What we find particularly intriguing about these results is that four days of the chronic stress protocol is not enough to induce changes in immune capacity, as it is in the case of uric acid and behavior. The initial maintenance of bacterial killing capacity is possibly indicative of an initial enhancement of immune function. It is possible that during the initial, “short-term” 4 days of the chronic stress protocol, immune function is actually enhanced. By the time the second sample is taken however, immune function is returning to original levels, before falling even further upon exposure to the second set of stressors. Prior studies have indicated the complex balance between stress and immunity with short-term exposure enhancing and long-term exposure suppressing (Dhabhar & McEwen, 1999; Martin, 2009).

We believe that this final decrease in in vitro immune function is representative of the birds transitioning into a “chronically stressed” state, or what the reactive scope model terms homeostatic overload (Romero, Dickens & Cyr, 2009). The fact that this is happening after approximately 8 days of the chronic stress protocol supports prior findings that have investigated the physiological changes during chronic stress (Cyr et al., 2007; Cyr & Romero, 2007; DuRant et al., 2016a; Rich & Romero, 2005). While we recognize that this is a single measure of immune function and relies on only one kind of pathogen, the BKA has been found to be correlated with other immune parameters and assays in avian species, including lipopolysaccharide challenge, phagocytosis assay, and acute phase proteins (Millet et al., 2007).

Metabolic changes during chronic stress

Interestingly, uric acid concentrations decreased immediately during the chronic stress protocol following the initial stimulus (Fig. 4). This was contrary to predictions based on previous studies that indicated that chronic corticosterone exposure or psychological or physical stress lead to elevations in uric acid in avian species (Azad et al., 2010; Lin, Decuypere & Buyse, 2004a; Lin, Decuypere & Buyse, 2004b). While a prior survey of nearly 100 bird species found that uric acid concentrations predominately fall following the acute stress response (Cohen, Klasing & Ricklefs, 2007), fewer studies have examined uric acid changes due to chronic stress, independent of CORT administration. We predicted that basal uric acid concentrations would increase over time because as birds experienced the chronic stress protocol, their metabolism would increase, thus leading to more uric acid as a byproduct of purine metabolism. However, these previous studies specifically relied on elevations of CORT to induce elevations in uric acid, whereas our stress protocol resulted in no changes in baseline CORT. Furthermore, previous studies report that uric acid is an important and potent antioxidant in avian species; one study specifically indicated that reduced levels of uric acid correlate with increased oxidative stress (Klandorf et al., 2001). In the present study, reduced uric acid concentrations could be an indication that the redox balance becomes upset within four days of the chronic stress protocol. In other words, the chronic stress protocol caused the house sparrows to use up their uric acid stores, which are not replenished, or at least not replenished fast enough to compensate for ongoing uric acid catabolism. This change suggests that birds may not have the same antioxidant buffer under chronic stress conditions as they do under “normal” conditions.

We find it interesting that this decrease in uric acid concentrations does not point to the same time frame for transitioning into chronic stress as the immune data. While the immune data do not show significant changes until after eight days of the chronic stress protocol, uric acid significantly decreases within the first four days. This finding emphasizes that different physiological metrics can suggest alternate time frames for when the shift into chronic stress occurs.

Behavioral effects of chronic stress

Finally, behavioral tests show that birds approached novel objects faster and reduced their activity as they progressed through the stress protocol. Note that a faster approach is unlikely to reflect habituation to the novel object because each bird was exposed to an unexperienced object for each trial, which in an earlier study elicited equivalent neophobic responses for each trial (Fischer & Romero, 2016). Interestingly in this study, object type did have a significant main effect on feed latency. Feed latency when the red dish and ribbon were present elicited similar declines over the course of the protocol whereas responses to the plastic egg did not vary. Importantly, object type did not have a main effect in the perch hopping data. Therefore, we feel that focusing on the overall pattern across all object types reflects the accurate biological response.

Faster approaches and lower activity were contrary to our original prediction, which was that neophobic and anxious behaviors would increase with chronic stress. This prediction was again, based on studies that suggested associations between CORT titers and fear/avoidance behavior (e.g., Korte, 2001; Skórzewska et al., 2006). The reduced neophobic behaviors may be a result of this experiment not inducing increased CORT concentrations, as in these other studies. Similarly, perch hopping frequency decreased through the chronic stress protocol, suggesting that overall activity decreased. Therefore, the faster feed times were not a result of random, flighty activity within the cage, but rather deliberate movements towards the novel object and food dish.

Interestingly, feed latencies during the no-object control trials significantly decreased after the initial trial (Fig. 5, white bars). This suggests that the chronic stress protocol caused the birds to approach their food dish faster in general. This does not lessen the significance of the pattern we see in the neophobia trials during which birds also approach the novel objects faster. Birds still took about seven times longer to feed during the neophobia times relative to the no-object controls. Therefore, the neophobic response did not disappear; birds were simply likely to more quickly approach the food dish regardless of there a novel object was present.

We found a significant interaction between group and perch hopping in the control samples (when no novel object was present) (Fig. 6B). Birds that experienced cage rolling first decreased their activity while those experiencing cage tapping first increased their activity by the end of the chronic stress protocol. It is not clear why order was related to perch hopping, but this kind of interaction effect was not apparent in any of the other physiological or behavioral data.

Although our original hypotheses for the behavioral measurements were based on prior studies connecting CORT and behavior, there is an alternative interpretation arising from the reactive scope model that could explain the decreases. The data suggest that chronic stress induced house sparrows to take additional risk when foraging. Rather than delaying their approach towards a novel, potentially dangerous, object, these birds more readily risked potential harm to forage. In certain circumstances this could be quite dangerous, particularly in the wild where novel objects might not be so innocuous as a colored food dish, ribbon, or plastic egg.

Many behavioral ecology studies have attempted to link CORT to broader indices of behavior, including explorative behavior and foraging activity (e.g., Atwell et al., 2012; Crino et al., 2017). The results of this study indicate that at least in some instances, chronic stress-induced changes to CORT and behavior happen at different time scales. This could be particularly important to consider in field-based studies when the stimuli might be on a much larger context, including urbanization and predator presence. Finally, we found it interesting that during the no-object control trials, females tended to be much more active than males. It is particularly intriguing that this sex difference is absent in the neophobia trials when novel objects are present. While the response to novel objects is not affected by sex, underlying sex differences appear to change baseline activity.