RESEARCH ARTICLE
House Finch Responses to Mycoplasma gallisepticum Infection do not Vary With Experimentally Increased Aggression JAMES STEPHEN ADELMAN*, IGNACIO TOMÁS MOORE, AND DANA MICHELLE HAWLEY Department of Biological Sciences, Virginia Tech, Blacksburg, Virginia
ABSTRACT
Aggression can alter infectious disease dynamics through two, non‐exclusive mechanisms: 1) increasing direct contact among hosts and 2) altering hosts' physiological response to pathogens. Here we examined the latter mechanism in a social songbird by manipulating intraspecific aggression in the absence of direct physical contact. We asked whether the extent of aggression an individual experiences alters glucocorticoid levels, androgen levels, and individual responses to infection in an ecologically relevant disease model: house finches (Haemorhous mexicanus) infected with Mycoplasma gallisepticum (MG). Wild‐caught male finches were housed in one of three settings, designed to produce increasing levels of aggression: 1) alone, with no neighbor (“no neighbor”), 2) next to a sham‐ implanted stimulus male (“sham neighbor”), or 3) next to a testosterone‐implanted stimulus male (“testosterone neighbor”). Following one week of social treatment, focal males were experimentally infected with MG, which causes severe conjunctivitis and induces sickness behaviors such as lethargy and anorexia. While social treatment increased aggression as predicted, there were no differences among groups in baseline corticosterone levels, total circulating androgens, or responses to infection. Across all focal individuals regardless of social treatment, pre‐infection baseline corticosterone levels were negatively associated with the severity of conjunctivitis and sickness behaviors, suggesting that corticosterone may dampen inflammatory responses in this host‐pathogen system. However, because corticosterone levels differed based upon population of origin, caution must be taken in interpreting this result. Taken together, these results suggest that in captivity, although aggression does not alter individual responses to MG, corticosterone may play a role in this disease. J. Exp. Zool. 323A:39–51, 2015. © 2014 Wiley Periodicals, Inc.
How to cite this article: Adelman JS, Moore IT, Hawley DM. 2015. House finch responses to Mycoplasma gallisepticum infection do not vary with experimentally increased aggression. J. Exp. Zool. 323A:39–51, 2015 J. Exp. Zool. 323A:39–51.
Aggressive behaviors, specifically fighting or threatening to fight (Dugatkin, 2009), can alter infectious disease dynamics by simultaneously augmenting physical exposure to and physiological susceptibility to parasites (Hawley et al., 2011a; Fairbanks and Hawley, 2012). For example, because aggression can involve direct physical contact among individuals, it has been included in a suite of traits that covary with contact rates and, thus, risk of infection in free‐living mammals, including feral cats and spotted hyenas (Courchamp et al., 1998; East et al., 2001). Furthermore, in mammals and birds, stressful social settings, including those associated with increased aggression (e.g., competition for food
Grant sponsor: National Science Foundation; grant numbers: IOS-1054675, IOS-1145625; grant sponsor: Franklin Life Science Institute at Virginia Tech. Conflict of interest: The authors declare that they have no conflicts of interest. Correspondence to: James Stephen Adelman, Department of Biological Sciences, Virginia Tech, Derring Hall, Room 2119 (MC 0406), 1405 Perry Street Blacksburg, Virginia 24061. E‐mail: [email protected] Received 10 July 2014; Revised 15 August 2014; Accepted 25 August 2014 DOI: 10.1002/jez.1894 Published online 11 November 2014 in Wiley Online Library (wileyonlinelibrary.com).
© 2014 WILEY PERIODICALS, INC.
40 or mates), have been shown to reduce immune responses such as the production of antibodies and inflammatory cytokines (Svensson et al., 2001a,b; Hawley et al., 2006; Hawley et al., 2007; Willette et al., 2007). These changes in immunity suggest that in addition to altering contact rates, aggression can hinder physiological means of pathogen clearance, increasing pathogen loads in infected individuals. Increased pathogen loads can subsequently increase the probability of pathogen transmission (Alizon et al., 2009), as exemplified by monarch butterflies infected with the protozoan parasite Ophryocystis elektroscirrha (de Roode et al., 2009). Moreover, because many instances of aggression in the wild do not escalate to intense physical contact (Thompson, 1960; Maynard Smith and Price, 1973), the physiological consequences of aggression may be particularly relevant to understanding disease ecology for many wildlife systems. However, to our knowledge, no studies of wildlife disease have tested the effects of aggression on physiological and behavioral responses to infection in the absence of direct, physical contact between hosts. Here we experimentally remove the effects of aggression on contact rates, while testing how aggression affects individual responses to infection in an ecologically relevant wildlife disease system: house finches (Haemorhous mexicanus) infected with Mycoplasma gallisepticum (MG), a bacterium that causes severe conjunctivitis and significantly reduces host survival in the wild (Faustino et al., 2004; Kollias et al., 2004). The house finch‐MG interaction is a particularly relevant host‐ pathogen system for examining the effect of aggression on host physiology. MG prevalence peaks during the fall and winter months, when house finches form large flocks at bird feeders and engage in frequent aggressive interactions (Thompson, 1960; Altizer et al., 2004). Additionally, aggressive encounters in this species most frequently involve behavioral threats and displacement from perches without direct contact (Thompson, 1960). As such, examining the effects of aggression without direct physical contact is particularly important in this system. Moreover, because bird feeders can facilitate MG transmission (Dhondt et al., 2007), and both host pathogen load and conjunctivitis increase MG deposition onto feeders (Adelman et al., 2013a), the influence of social interactions on within‐host processes are likely to have important consequences for transmission in this system. Past work has shown that competition for food can suppress immune responses in captive house finch flocks (Hawley et al., 2006), and dominance status within males predicts the severity of conjunctivitis following experimental infection with MG (Hawley et al., 2007). However, conjunctival pathogen loads in response to social stress or dominance status were not measured in prior work, so it remains unclear how aggression may influence the relevant currency for transmission potential in this system. Furthermore, understanding the physiological mediators, such as hormones, that may link aggression to pathogen responses is particularly important because these J. Exp. Zool.
ADELMAN ET AL. mechanisms may help drive covariation in behavioral exposure and susceptibility among individuals (Hawley et al., 2011a). Hormones provide one potential mechanism linking aggression and within‐host disease processes. Notably, in diverse vertebrates, glucocorticoid and androgen levels increase in response to aggression and thus are prime candidates for mediating the effects of aggression on disease dynamics (Creel et al., 2013; but see Goymann, 2009; Rosvall et al., 2014). Glucocorticoids (principally corticosterone in birds) are steroid hormones released during periods of stress, which can include aggressive interactions, predation attempts, and severe weather (Sapolsky et al., 2000; Romero, 2004; Creel et al., 2013; Deviche et al., 2014). These hormones have wide‐ranging effects on physiology, altering processes from glucose metabolism to neuronal activity and behavior (Sapolsky et al., 2000). While short‐term stressful events can help redistribute or augment certain immune defenses, when glucocorticoids remain elevated for extended periods (e.g., days to weeks), these hormones can inhibit a number of immune defenses, notably decreasing inflammatory immune responses (Sapolsky et al., 2000; Martin, 2009; Dhabhar, 2014). In addition to glucocorticoids, androgens could play an important role in linking aggression and disease processes. Circulating levels of androgens, including testosterone, can rise during aggressive encounters in numerous species (Wingfield et al., 1990; Hirschenhauser and Oliveira, 2006; but see Goymann, 2009; Rosvall et al., 2014) and experimentally increased levels can reduce immune responses and increase disease susceptibility in some reptilian, mammalian, and avian species, including house finches, although other species within these taxa do not show testosterone‐mediated immunosuppression (Folstad and Karter, 1992; Duckworth et al., 2001; Roberts et al., 2004; Deviche and Cortez, 2005; Roberts et al., 2007). In addition, because hormones, notably androgens (Nelson, 2000; Adkins‐Regan, 2005), can facilitate aggressive behaviors, positive feedback loops among aggression and hormones could exacerbate the effects of aggression on immune responses. The potential role for hormones in suppressing inflammatory immune responses for days to weeks is particularly relevant to MG infection, as host responses to this pathogen typically include several inflammatory processes, such as the influx of white blood cells to the site of infection, localized swelling, systemic production of pro‐inflammatory cytokines, fever, and the sickness behaviors of lethargy and anorexia (Luttrell et al., '98; Mohammed et al., 2007; Hawley et al., 2012; Adelman et al., 2013b). Recent evidence suggests that the degree of pro‐ inflammatory signaling during MG infection in house finches predicts resulting disease pathology (Adelman et al., 2013b). Therefore, while hormonal responses to aggression may reduce immune defenses during MG infection, such reductions could not only increase pathogen levels, but somewhat paradoxically, decrease levels of pathology. Specifically, because animals
AGGRESSION AND DISEASE IN HOUSE FINCHES showing lower levels of inflammatory immune responses exhibited less‐pronounced conjunctivitis (Adelman et al., 2013b), hormonally mediated reductions in inflammation could lead to reduced pathology. Additionally, (Belthoff et al., '94) have shown that immediately following periods of aggression in captive house finches, corticosterone levels closely match those induced by restraint‐stress (Lindström et al., 2005). If heightened levels of aggression persist, such increases in corticosterone may lead to down‐regulation of inflammatory responses (Martin, 2009). While (Belthoff et al., '94) found no alteration to testosterone with increased aggression, trials were performed during the non‐breeding season when testosterone levels are extremely low. Given that house finches show seasonal alterations of testosterone typical of many socially monogamous passerines, we predicted that experimentally increased aggression should further increase levels of testosterone when finches are in breeding condition (Wingfield et al., '90). To experimentally isolate the physiological effects of aggression and to test the hypothesis that hormone levels link aggression to within‐host responses to MG infection, we housed male house finches in one of three social settings: 1) no aggression (“no neighbor”) —focal birds were housed alone in one side of a cage divided into two compartments, 2) low aggression (“sham neighbor”)—focal birds were housed in one side of a divided cage with a single, sham‐implanted neighbor in the other compartment, or 3) high aggression (“testosterone neighbor”)— birds were housed in one side of a divided cage with a neighbor that had been implanted with testosterone in the other compartment (Searcy and Wingfield, '80). We predicted that increasing aggression from the neighbor would elevate both corticosterone and total androgens in the focal bird, which, because of their anti‐inflammatory and immunomodulatory properties, would lead to higher pathogen loads, but reduced
41 clinical signs of infection, measured here as conjunctivitis and sickness behaviors (anorexia and lethargy).
MATERIALS AND METHODS Permits Experiments were conducted in accordance with NIH guidelines and all required permits: Virginia Tech IACUC (10‐059‐BIOL), Alabama Department of Conservation and Natural Resources (5436), Virginia Department of Game and Inland Fisheries (038781 and 044569), and United States Fish and Wildlife Service (MB158404‐1). At the termination of the experiments, animals were euthanized using an overdose of the anesthetic, isoflurane, as approved by the Virginia Tech IACUC. Study Subjects We only used males in this experiment to remove any confounding effects of sex on analyses of aggressive interactions. Male house finches were captured using feeder traps on the campus of Auburn University in Auburn, Alabama, USA (32° 250 29.7800 N, 85° 292013.5900 W, n ¼ 14 birds) on September 9–10, 2011 and on the campus of Virginia Tech in Blacksburg, VA (37° 130 23.0300 N, 80° 250 39.6700 W, n ¼ 10 birds) between November 30, 2011 and January 15, 2012. Upon capture, birds were housed indoors in 46 46 76 cm cages under a constant light cycle (12L:12D) and temperature (22 °C), either alone or paired with a female. On February 24, 2012, all birds were moved to individual cages. Birds were placed into social treatments on March 21, 2012, with experimental inoculation following on March 27, 2012 (see timeline in Fig. 1). Based upon prior work, a 12L:12D cycle successfully terminates the photorefactory period in house finches (Hamner, '66); thus all birds should have been photosensitive by the beginning of the experiment. While no
Figure 1. Experimental timeline for stimulus birds and focal birds.
J. Exp. Zool.
42
ADELMAN ET AL.
cloacal protuberances were visible, birds could be heard singing throughout the experiment, suggesting that males were in the early stages of breeding condition. Housing and Social Treatments Ten weeks before the start of the experiment, birds were assigned to one of two groups: 1) focal birds (n ¼ 24): MG‐naïve males that would be experimentally infected and 2) stimulus birds (n ¼ 17): MG‐vaccinated (see below), but healthy, males used to modulate social stress of the focal birds. We then randomly assigned each focal male to one of three social treatments: 1) “no neighbor” (n ¼ 7): housed alone, 2) “sham neighbor” (n ¼ 9): housed next to a stimulus male implanted with empty silastic tubing (see below), or 3) “testosterone neighbor” (n ¼ 8): housed next to a stimulus male implanted with testosterone (see below). While the “no neighbor” treatment could induce stress through isolation in some species, during the time of year that experiments were performed, male house finches typically interact with few individuals in the wild (Badyaev et al., 2012), suggesting that minimal interactions would be unlikely to induce stress in this species. Moreover, animals could hear and see other individuals in the same experimental room. Birds from each population of origin were distributed across treatment groups (Table 1). Within each population of origin, individuals were assigned to treatment groups randomly. For implants, silastic tubing (Catalog Number 508‐006, Dow Corning, Midland, MI, USA inner diameter 1.47 mm, outer diameter 1.96 mm) was cut into 10 mm sections and sealed at one end with 1 mm of silicone sealant. Implants were then packed with crystalline testosterone (Catalog Number A6950‐000, Steraloids, Inc., New Port, RI, USA) or left empty (sham) and sealed at the other end with 1 mm of silicone sealant. Implants were inserted subcutaneously on each bird's left flank and incisions were sealed with Vetbond surgical adhesive (3M, St. Paul, MN, USA). Similar procedures in this species have yielded circulating androgen levels at the high end of the normal breeding season range (Duckworth et al., 2001; Stoehr and Hill, 2001; Duckworth et al., 2004). The efficacy of implants was validated in this study by radioimmunoassay (see Results).
Six days prior to experimental infections, all birds were placed into 46 46 76 cm cages, which were divided into two compartments using a plane of 1 cm wire mesh affixed along a central diagonal. One compartment housed a focal bird, while the other compartment either housed a stimulus bird or remained empty (“no neighbor” treatment). The orientation of diagonal dividers (back‐left to front‐right or vice versa) was divided equally within social treatments. Focal birds were always housed in the front compartment to facilitate the recording of video for behavioral analyses. Birds could both see and hear other birds (in the same and separate cages), though the fine mesh eliminated physical contact. Each bird had ad libitum access to water and food, with feeding stations placed next to one another along the central barrier to maximize aggressive interactions between the focal and stimulus cagemates. Water dishes were placed away from the barrier. Social treatments were divided equally among five rooms maintained at identical light and temperature conditions. Vaccination of Stimulus Birds To minimize the chance that stimulus birds would contract MG during the experiment, which would likely affect their behavior toward focal birds (Bouwman and Hawley, 2010), we experimentally infected stimulus birds with MG prior to the start of the experiment and subsequently treated them with tylosin (Tylan, Elanco, Greenville, IN, USA), a broad spectrum antibiotic. Birds were inoculated in the palpebral conjunctiva of both eyes using 25 mL of Mycoplasma gallisepticum in Frey's media. We used an expansion of the 7th in vitro passage of an isolate of MG collected from a house finch with conjunctivitis in Virginia in 1994 (7994‐ 1 7P 2/12/09; D. H. Ley, North Carolina State University, College of Veterinary Medicine, Raleigh, NC, USA)(Ley et al., 1996). The MG strain and volume used were the same as that used for focal birds (see below). However, the vial used to inoculate stimulus birds had been thawed and refrozen prior to use, which reduces MG viability. Before freeze/thaw, the viable count of the inoculum was 2.24 107 color changing units (CCU) per milliliter, determined by the most probable number method (Meynell and Meynell, 1970).
Table 1. Sample sizes of focal birds in each social treatment, partitioned by population of origin. Social treatment
Birds from Alabama Birds from Virginia Total Focal Birds
No neighbor
Sham‐implanted neighbor
Testosterone‐implanted neighbor
3 (1 excluded) 4 7
7 2 9
5 3 (1 excluded) 8
Two birds were excluded from analyses because they were found to have low levels of Mycoplasma gallisepticum before the start of the experiment (see Methods).
J. Exp. Zool.
AGGRESSION AND DISEASE IN HOUSE FINCHES Beginning three weeks after inoculation and continuing for seven weeks, stimulus birds were given drinking water with 2 mg/ mL tylosin mixed in each day. This protocol ensured that stimulus birds would have antibodies against the same strain of MG used in the experiment, so would be unlikely to contract the pathogen during interactions with focal birds. Following treatment, none of the stimulus birds showed clinical signs of infection. Experimental Infection of Focal Birds Six days after being housed in experimental social groups, focal birds were inoculated with the same volume and strain of MG as above (Fig. 1). The vial used for this inoculation had not been repeatedly frozen, so its viable count was 2.24 107 CCU/mL. Behavior Three days prior to the inoculation of focal birds, and on days 4, 16, and 30 post‐inoculation, we recorded video of all birds from 08:00 (lights‐on) through 11:00. Cameras were placed in each room before lights‐off the day prior to recording (approximately 16:00). On the day of recording, cameras were turned on manually while rooms were still dark (07:30–07:50). From these videos, behaviors of focal birds were observed during two 15 min sampling periods, the first beginning immediately after lights‐on (08:00), the second beginning approximately 1.5 hr later (09:20– 09:30). We assessed the following behaviors relevant to the prototypical sickness behaviors of lethargy and anorexia: the number of hops or flights, the number of bill swipes, the amount of time spent immobile (no detectable movement at all), the time spent within one body‐width of the feeder, and the number of pecks made at food. In addition, we quantified the instances of aggression initiated by and received by the focal bird. Aggression was defined as a rapid movement (by either focal or stimulus bird) made directly at the other bird. An aggressive encounter was considered “escalated” if the focal bird made physical contact with the barrier separating cage compartments. Because involvement in an aggressive encounter can alter an animal's physiology, regardless of whether that animal initiated aggression or not (Wingfield et al., 1990; Creel et al., 2013), we considered the total observed instances of aggression as the most appropriate metric for this study. For analyses, we therefore summed aggression received and aggression initiated by the focal bird to yield a single value of total aggression in each cage. Because other authors have reported asymmetrical immune and physiological changes in response to aggression received or reductions in social status (Tuchscherer et al., 1998; Devoino et al., 2003; Hawley, 2006), we note that our results are qualitatively identical whether analyzing aggression received, aggression initiated, or total aggression. We defined the outcome of an aggressive interaction as a “win” if the focal bird maintained his position and the stimulus bird moved away after the interaction, and as a “loss” if the stimulus
43 bird maintained his position and the focal bird moved away after the interaction. The outcome was defined as “neutral” if both birds maintained their positions for 3 sec or longer after the interaction. We classified focal birds as dominant, subordinate, or equal to their neighbors using pre‐infection data on aggression. Specifically, for each focal bird, we subtracted the total number of losses during pre‐infection sampling from the total number of wins during pre‐infection sampling. Positive numbers indicated that focal birds were dominant, negative numbers indicated that focal birds were subordinate, 0 indicated birds were equal. Of all aggressive interactions recorded, 66% resulted in a clear winner and loser. Considering only pre‐infection sampling, this percentage was the same. Conjunctival Pathogen Load and Pathology Pathogen load was assessed using previously published methods (Hawley et al., 2011b). Briefly, for each eye, a sterile cotton swab was rotated for 5 sec on the inside of the conjunctiva after being dipped in tryptose phosphate broth (TPB). The swab was then swirled in a microcentrifuge tube containing 300 mL of TPB and wrung out on the side of the tube. These samples were then frozen at 20 °C for up to four months. DNA was extracted from these samples using Qiagen DNeasy 96 Blood and Tissue kits (Qiagen, Valencia, CA, USA). We then performed quantitative polymerase chain reaction, probing for the Mycoplasma mgc2 gene on a MyiQ Single Color Real‐Time PCR Detection System (Bio‐Rad Laboratories, Hercules, CA, USA), using previously published parameters and primers (Grodio et al., 2008; Hawley et al., 2011b). Conjunctival pathology was assessed using eye scores, based upon a 4‐point scale, as defined in previously published studies (e.g., Hawley et al., 2011b). Briefly, this scale is defined as: 0) no pathology, 1) slight redness and swelling of the conjunctiva, 2) moderate swelling and some eversion of the conjunctiva, 3) severe swelling of the conjunctiva and noticeable exudate. Hormone Levels Blood samples were taken by puncturing the wing vein using a 26 gauge needle and collecting blood into heparin‐coated capillary tubes. All samples were collected between the hours of 11:00 and 12:00, local time. All samples used to assess corticosterone were obtained within 3 min of entering the experimental rooms; samples for testosterone were taken within 10 min. Samples for both androgens and corticosterone were taken from focal birds one day before infection (five days after co‐housing) (Fig. 1). To ensure the efficacy of hormone implants in stimulus birds, samples for androgens were taken from these animals six days after implantation (six days before co‐housing) and 32 days after implantation (20 days after co‐ housing; Fig. 1). All samples were kept on ice and centrifuged to separate plasma within 2 hr of collection. Plasma was then frozen at 20 °C for up to two months. J. Exp. Zool.
44
ADELMAN ET AL.
Table 2. Summary of the principal components analysis on sickness behaviours (using metrics of locomotion and foraging as proxies for lethargy and anorexia, respectively).
Variable Number of hops and flights Number of bill‐swipes Proportion of time immobile Proportion of time within one body‐width of feeder Number of pecks at food Cumulative proportion of variance explained
PC1
PC2
Rotation PC3
PC4
PC5
0.39 0.33 0.51 0.44 0.54 0.45
0.62 0.11 0.34 0.61 0.34 0.67
0.34 0.90 0.11 0.24 0.01 0.85
0.44 0.02 0.71 0.22 0.51 0.95
0.39 0.25 0.34 0.57 0.58 1.00
Concentrations of total androgens and corticosterone were measured in plasma samples using a direct radioimmunoassay after extraction in dicholoromethane, using published methods (Wingfield et al., '91; Moore et al., 2002). All samples were run in duplicate. Androgens were measured in a single assay (intra‐assay % CV ¼ 16.8, detection limit 0.13 ng/mL); corticosterone was measured in a separate, single assay (intra‐assay % CV ¼ 10.7, detection limit 1.0 ng/mL).
Data Analysis Data were analyzed using mixed effects models in R version 11.1 (Pinheiro and Bates, 2000; R Development Core Team, 2011). Data on corticosterone, total androgens, and total aggression were transformed to better meet the assumptions of homoscedascity and normally distributed errors (corticosterone and total androgens were log transformed; total aggression was log þ 1 transformed). All other variables were untransformed. Because birds were captured from two populations several months apart, and either population origin or time in captivity could affect physiological responses to infection (Bonneaud et al., 2011; Bonneaud et al., 2012; Adelman et al., 2013a; Adelman et al., 2013b), we included population as a random effect in all analyses. Initially, random effects allowed the effect of social treatment to vary with population of origin. However, in no case did this improve the fit of the models (likelihood ratio tests, in all cases P > 0.3), so final random effects structures controlled only for different intercepts between populations. In all initial models, we included prior housing condition (with or without female) as a fixed effect. This term was removed from all but one model, as it did not improve model fits (see results). In addition to the above analyses testing for differences among treatment groups, to determine if corticosterone could affect disease responses independent of treatment, we also performed linear regressions of pre‐infection corticosterone levels against J. Exp. Zool.
responses to infection (pathogen load, eye pathology, and sickness behaviors). Finally, we asked whether initial condition (quantified as the residuals of a regression of mass vs. tarsus before co‐housing occurred) either differed across treatment groups or predicted individual variation in corticosterone levels. For analysis of sickness behaviors, the five behavioral metrics described above were collapsed into a single variable using principal components analysis. The first principal component (PC1) explained 46% of the variation in all behaviors, with active behaviors (bill swipes, hops and flights) loading with the same sign as feeding behaviors (proportion of time spent within one body‐width of the feeder, number of pecks at food) and with the opposite sign of proportion of time spent immobile (Table 2). PC1 was used to provide a single metric for all subsequent analyses of sickness behavior. Measurement of total aggression and circulating hormones pre‐infection were taken from single time points. However, post‐ infection responses (pathogen load, eye score, and PC1 above) were measured at several time points. To generate a single data point for each bird that accounted for both the magnitude and duration of post‐infection responses, we estimated the integral of each variable across time as in Adelman et al. (2013b). Briefly, we calculated the area under a curve generated for each response, with the response variable on the y‐axis and time on the x‐axis (including the pre‐infection metrics at x ¼ 0, and post infection metrics on days 4, 7, 14, 21, 28, and 35 after inoculation for pathogen load and pathology; and post infection metrics on days 4, 16, and 30 after inoculation for behavioral responses, see Figure 1). We assumed a linear change in y‐variables between each time point. Two focal birds, one originally assigned to the “no neighbor” group (from AL) and one originally assigned to the “testosterone neighbor” group (from VA), were omitted from all analyses, as each was found to have detectable levels of MG (over 300 copies of the mgc2 gene detected) one day before experimental
AGGRESSION AND DISEASE IN HOUSE FINCHES inoculation. Sample sizes given above and in Table 1 reflect these omissions.
RESULTS Testosterone Treatment of Stimulus Birds Both before and after being housed in social treatments, stimulus birds treated with testosterone implants had significantly higher levels of circulating total androgens than those treated with empty implants (mean 1SE ng/mL, Before Co‐Housing: testosterone: 13.16 1.04, empty: 0.16 0.04; After Co‐Housing: testosterone: 9.00 0.90, empty: 0.34 0.20; implant type: F1,14 ¼ 149.09, P < 0.001, implant type x date: F1,16 ¼ 20.51, P 0.18). Initial body condition, which could influence subsequent levels of aggression, hormones, or responses to infection, did not differ across social treatments (F2,20 ¼ 0.51, P ¼ 0.61, data not shown) or with population of origin (Likelihood ratio (model with and without random effect of population) ¼ 1.26, P ¼ 0.26). Contrary to our predictions, pre‐infection corticosterone levels, pre‐infection androgen levels, and post‐infection responses to MG did not differ with social treatment (Fig. 2C–G, all F2,20 0.26). Circulating androgen concentrations
45 were low among focal birds, with 9 of 24 pre‐infection samples falling at or below the detection limit of the assay. Additionally, including population of origin as a random effect improved the model of corticosterone levels, with pre‐ infection corticosterone higher in birds from Virginia than Alabama (Virginia: 7.16 0.94, Alabama: 3.12 0.29; Likelihood ratio (model with and without random effect of population) ¼ 10.79, P < 0.001). Similar analyses showed a trend for sickness behaviors to be lower among birds from Virginia (Likelihood ratio (model with and without random effect of population) ¼ 2.55, P ¼ 0.11), but no other patterns with population emerged for aggression, androgen levels, or responses to infection (all Likelihood ratios (models with and without random effect of population) 0.49). Results were similar when analyzed by dominance rather than social treatment. As with social treatment, dominance status within a cage did not influence pre‐infection corticosterone levels or post‐infection responses to MG (eye pathology, sickness behaviors, or pathogen load) (data not shown, all F3,19 0.66). “Testosterone neighbor” males tended to have fewer aggressive interactions with a clear winner (median ¼ 60%) than “sham neighbor” males (median ¼ 80%, Wilcoxon test P ¼ 0.06), suggesting that dominance may have been slightly less stable in the “testosterone neighbor” treatment. However, the percentage of aggressive interactions in which the focal bird was the clear winner did not influence pre‐infection corticosterone levels or post‐infection responses to MG (eye pathology, sickness behaviors, or pathogen load) (data not shown, all F1,12 < 0.93, all P > 0.36). Corticosterone and Responses to Infection: Individual‐Level Analyses Regressions independent of treatment group revealed that pre‐ infection corticosterone levels correlated negatively with eye pathology and sickness behaviors, but not pathogen load during infection (eye score, Fig. 3A: F1,22 ¼ 4.93, P ¼ 0.04, r2 ¼ 0.18; sickness behaviors, Fig. 3B: F1,22 ¼ 7.59, P ¼ 0.01, r2 ¼ 0.26; pathogen load, Fig. 3C: F1,22 ¼ 2.17, P ¼ 0.16, r2 ¼ 0.09). Because population of origin correlated with pre‐infection corticosterone levels (see above), population of origin was not included as a covariate in this analysis. When these regressions were performed separately for each population, estimated slopes were similarly negative, though no relationships were statistically significant (all F1,13(AL) or 1,7(VA) < 1.59, P > 0.23, r2
House Finch Responses to Mycoplasma gallisepticum Infection do not Vary With Experimentally Increased Aggression JAMES STEPHEN ADELMAN*, IGNACIO TOMÁS MOORE, AND DANA MICHELLE HAWLEY Department of Biological Sciences, Virginia Tech, Blacksburg, Virginia
ABSTRACT
Aggression can alter infectious disease dynamics through two, non‐exclusive mechanisms: 1) increasing direct contact among hosts and 2) altering hosts' physiological response to pathogens. Here we examined the latter mechanism in a social songbird by manipulating intraspecific aggression in the absence of direct physical contact. We asked whether the extent of aggression an individual experiences alters glucocorticoid levels, androgen levels, and individual responses to infection in an ecologically relevant disease model: house finches (Haemorhous mexicanus) infected with Mycoplasma gallisepticum (MG). Wild‐caught male finches were housed in one of three settings, designed to produce increasing levels of aggression: 1) alone, with no neighbor (“no neighbor”), 2) next to a sham‐ implanted stimulus male (“sham neighbor”), or 3) next to a testosterone‐implanted stimulus male (“testosterone neighbor”). Following one week of social treatment, focal males were experimentally infected with MG, which causes severe conjunctivitis and induces sickness behaviors such as lethargy and anorexia. While social treatment increased aggression as predicted, there were no differences among groups in baseline corticosterone levels, total circulating androgens, or responses to infection. Across all focal individuals regardless of social treatment, pre‐infection baseline corticosterone levels were negatively associated with the severity of conjunctivitis and sickness behaviors, suggesting that corticosterone may dampen inflammatory responses in this host‐pathogen system. However, because corticosterone levels differed based upon population of origin, caution must be taken in interpreting this result. Taken together, these results suggest that in captivity, although aggression does not alter individual responses to MG, corticosterone may play a role in this disease. J. Exp. Zool. 323A:39–51, 2015. © 2014 Wiley Periodicals, Inc.
How to cite this article: Adelman JS, Moore IT, Hawley DM. 2015. House finch responses to Mycoplasma gallisepticum infection do not vary with experimentally increased aggression. J. Exp. Zool. 323A:39–51, 2015 J. Exp. Zool. 323A:39–51.
Aggressive behaviors, specifically fighting or threatening to fight (Dugatkin, 2009), can alter infectious disease dynamics by simultaneously augmenting physical exposure to and physiological susceptibility to parasites (Hawley et al., 2011a; Fairbanks and Hawley, 2012). For example, because aggression can involve direct physical contact among individuals, it has been included in a suite of traits that covary with contact rates and, thus, risk of infection in free‐living mammals, including feral cats and spotted hyenas (Courchamp et al., 1998; East et al., 2001). Furthermore, in mammals and birds, stressful social settings, including those associated with increased aggression (e.g., competition for food
Grant sponsor: National Science Foundation; grant numbers: IOS-1054675, IOS-1145625; grant sponsor: Franklin Life Science Institute at Virginia Tech. Conflict of interest: The authors declare that they have no conflicts of interest. Correspondence to: James Stephen Adelman, Department of Biological Sciences, Virginia Tech, Derring Hall, Room 2119 (MC 0406), 1405 Perry Street Blacksburg, Virginia 24061. E‐mail: [email protected] Received 10 July 2014; Revised 15 August 2014; Accepted 25 August 2014 DOI: 10.1002/jez.1894 Published online 11 November 2014 in Wiley Online Library (wileyonlinelibrary.com).
© 2014 WILEY PERIODICALS, INC.
40 or mates), have been shown to reduce immune responses such as the production of antibodies and inflammatory cytokines (Svensson et al., 2001a,b; Hawley et al., 2006; Hawley et al., 2007; Willette et al., 2007). These changes in immunity suggest that in addition to altering contact rates, aggression can hinder physiological means of pathogen clearance, increasing pathogen loads in infected individuals. Increased pathogen loads can subsequently increase the probability of pathogen transmission (Alizon et al., 2009), as exemplified by monarch butterflies infected with the protozoan parasite Ophryocystis elektroscirrha (de Roode et al., 2009). Moreover, because many instances of aggression in the wild do not escalate to intense physical contact (Thompson, 1960; Maynard Smith and Price, 1973), the physiological consequences of aggression may be particularly relevant to understanding disease ecology for many wildlife systems. However, to our knowledge, no studies of wildlife disease have tested the effects of aggression on physiological and behavioral responses to infection in the absence of direct, physical contact between hosts. Here we experimentally remove the effects of aggression on contact rates, while testing how aggression affects individual responses to infection in an ecologically relevant wildlife disease system: house finches (Haemorhous mexicanus) infected with Mycoplasma gallisepticum (MG), a bacterium that causes severe conjunctivitis and significantly reduces host survival in the wild (Faustino et al., 2004; Kollias et al., 2004). The house finch‐MG interaction is a particularly relevant host‐ pathogen system for examining the effect of aggression on host physiology. MG prevalence peaks during the fall and winter months, when house finches form large flocks at bird feeders and engage in frequent aggressive interactions (Thompson, 1960; Altizer et al., 2004). Additionally, aggressive encounters in this species most frequently involve behavioral threats and displacement from perches without direct contact (Thompson, 1960). As such, examining the effects of aggression without direct physical contact is particularly important in this system. Moreover, because bird feeders can facilitate MG transmission (Dhondt et al., 2007), and both host pathogen load and conjunctivitis increase MG deposition onto feeders (Adelman et al., 2013a), the influence of social interactions on within‐host processes are likely to have important consequences for transmission in this system. Past work has shown that competition for food can suppress immune responses in captive house finch flocks (Hawley et al., 2006), and dominance status within males predicts the severity of conjunctivitis following experimental infection with MG (Hawley et al., 2007). However, conjunctival pathogen loads in response to social stress or dominance status were not measured in prior work, so it remains unclear how aggression may influence the relevant currency for transmission potential in this system. Furthermore, understanding the physiological mediators, such as hormones, that may link aggression to pathogen responses is particularly important because these J. Exp. Zool.
ADELMAN ET AL. mechanisms may help drive covariation in behavioral exposure and susceptibility among individuals (Hawley et al., 2011a). Hormones provide one potential mechanism linking aggression and within‐host disease processes. Notably, in diverse vertebrates, glucocorticoid and androgen levels increase in response to aggression and thus are prime candidates for mediating the effects of aggression on disease dynamics (Creel et al., 2013; but see Goymann, 2009; Rosvall et al., 2014). Glucocorticoids (principally corticosterone in birds) are steroid hormones released during periods of stress, which can include aggressive interactions, predation attempts, and severe weather (Sapolsky et al., 2000; Romero, 2004; Creel et al., 2013; Deviche et al., 2014). These hormones have wide‐ranging effects on physiology, altering processes from glucose metabolism to neuronal activity and behavior (Sapolsky et al., 2000). While short‐term stressful events can help redistribute or augment certain immune defenses, when glucocorticoids remain elevated for extended periods (e.g., days to weeks), these hormones can inhibit a number of immune defenses, notably decreasing inflammatory immune responses (Sapolsky et al., 2000; Martin, 2009; Dhabhar, 2014). In addition to glucocorticoids, androgens could play an important role in linking aggression and disease processes. Circulating levels of androgens, including testosterone, can rise during aggressive encounters in numerous species (Wingfield et al., 1990; Hirschenhauser and Oliveira, 2006; but see Goymann, 2009; Rosvall et al., 2014) and experimentally increased levels can reduce immune responses and increase disease susceptibility in some reptilian, mammalian, and avian species, including house finches, although other species within these taxa do not show testosterone‐mediated immunosuppression (Folstad and Karter, 1992; Duckworth et al., 2001; Roberts et al., 2004; Deviche and Cortez, 2005; Roberts et al., 2007). In addition, because hormones, notably androgens (Nelson, 2000; Adkins‐Regan, 2005), can facilitate aggressive behaviors, positive feedback loops among aggression and hormones could exacerbate the effects of aggression on immune responses. The potential role for hormones in suppressing inflammatory immune responses for days to weeks is particularly relevant to MG infection, as host responses to this pathogen typically include several inflammatory processes, such as the influx of white blood cells to the site of infection, localized swelling, systemic production of pro‐inflammatory cytokines, fever, and the sickness behaviors of lethargy and anorexia (Luttrell et al., '98; Mohammed et al., 2007; Hawley et al., 2012; Adelman et al., 2013b). Recent evidence suggests that the degree of pro‐ inflammatory signaling during MG infection in house finches predicts resulting disease pathology (Adelman et al., 2013b). Therefore, while hormonal responses to aggression may reduce immune defenses during MG infection, such reductions could not only increase pathogen levels, but somewhat paradoxically, decrease levels of pathology. Specifically, because animals
AGGRESSION AND DISEASE IN HOUSE FINCHES showing lower levels of inflammatory immune responses exhibited less‐pronounced conjunctivitis (Adelman et al., 2013b), hormonally mediated reductions in inflammation could lead to reduced pathology. Additionally, (Belthoff et al., '94) have shown that immediately following periods of aggression in captive house finches, corticosterone levels closely match those induced by restraint‐stress (Lindström et al., 2005). If heightened levels of aggression persist, such increases in corticosterone may lead to down‐regulation of inflammatory responses (Martin, 2009). While (Belthoff et al., '94) found no alteration to testosterone with increased aggression, trials were performed during the non‐breeding season when testosterone levels are extremely low. Given that house finches show seasonal alterations of testosterone typical of many socially monogamous passerines, we predicted that experimentally increased aggression should further increase levels of testosterone when finches are in breeding condition (Wingfield et al., '90). To experimentally isolate the physiological effects of aggression and to test the hypothesis that hormone levels link aggression to within‐host responses to MG infection, we housed male house finches in one of three social settings: 1) no aggression (“no neighbor”) —focal birds were housed alone in one side of a cage divided into two compartments, 2) low aggression (“sham neighbor”)—focal birds were housed in one side of a divided cage with a single, sham‐implanted neighbor in the other compartment, or 3) high aggression (“testosterone neighbor”)— birds were housed in one side of a divided cage with a neighbor that had been implanted with testosterone in the other compartment (Searcy and Wingfield, '80). We predicted that increasing aggression from the neighbor would elevate both corticosterone and total androgens in the focal bird, which, because of their anti‐inflammatory and immunomodulatory properties, would lead to higher pathogen loads, but reduced
41 clinical signs of infection, measured here as conjunctivitis and sickness behaviors (anorexia and lethargy).
MATERIALS AND METHODS Permits Experiments were conducted in accordance with NIH guidelines and all required permits: Virginia Tech IACUC (10‐059‐BIOL), Alabama Department of Conservation and Natural Resources (5436), Virginia Department of Game and Inland Fisheries (038781 and 044569), and United States Fish and Wildlife Service (MB158404‐1). At the termination of the experiments, animals were euthanized using an overdose of the anesthetic, isoflurane, as approved by the Virginia Tech IACUC. Study Subjects We only used males in this experiment to remove any confounding effects of sex on analyses of aggressive interactions. Male house finches were captured using feeder traps on the campus of Auburn University in Auburn, Alabama, USA (32° 250 29.7800 N, 85° 292013.5900 W, n ¼ 14 birds) on September 9–10, 2011 and on the campus of Virginia Tech in Blacksburg, VA (37° 130 23.0300 N, 80° 250 39.6700 W, n ¼ 10 birds) between November 30, 2011 and January 15, 2012. Upon capture, birds were housed indoors in 46 46 76 cm cages under a constant light cycle (12L:12D) and temperature (22 °C), either alone or paired with a female. On February 24, 2012, all birds were moved to individual cages. Birds were placed into social treatments on March 21, 2012, with experimental inoculation following on March 27, 2012 (see timeline in Fig. 1). Based upon prior work, a 12L:12D cycle successfully terminates the photorefactory period in house finches (Hamner, '66); thus all birds should have been photosensitive by the beginning of the experiment. While no
Figure 1. Experimental timeline for stimulus birds and focal birds.
J. Exp. Zool.
42
ADELMAN ET AL.
cloacal protuberances were visible, birds could be heard singing throughout the experiment, suggesting that males were in the early stages of breeding condition. Housing and Social Treatments Ten weeks before the start of the experiment, birds were assigned to one of two groups: 1) focal birds (n ¼ 24): MG‐naïve males that would be experimentally infected and 2) stimulus birds (n ¼ 17): MG‐vaccinated (see below), but healthy, males used to modulate social stress of the focal birds. We then randomly assigned each focal male to one of three social treatments: 1) “no neighbor” (n ¼ 7): housed alone, 2) “sham neighbor” (n ¼ 9): housed next to a stimulus male implanted with empty silastic tubing (see below), or 3) “testosterone neighbor” (n ¼ 8): housed next to a stimulus male implanted with testosterone (see below). While the “no neighbor” treatment could induce stress through isolation in some species, during the time of year that experiments were performed, male house finches typically interact with few individuals in the wild (Badyaev et al., 2012), suggesting that minimal interactions would be unlikely to induce stress in this species. Moreover, animals could hear and see other individuals in the same experimental room. Birds from each population of origin were distributed across treatment groups (Table 1). Within each population of origin, individuals were assigned to treatment groups randomly. For implants, silastic tubing (Catalog Number 508‐006, Dow Corning, Midland, MI, USA inner diameter 1.47 mm, outer diameter 1.96 mm) was cut into 10 mm sections and sealed at one end with 1 mm of silicone sealant. Implants were then packed with crystalline testosterone (Catalog Number A6950‐000, Steraloids, Inc., New Port, RI, USA) or left empty (sham) and sealed at the other end with 1 mm of silicone sealant. Implants were inserted subcutaneously on each bird's left flank and incisions were sealed with Vetbond surgical adhesive (3M, St. Paul, MN, USA). Similar procedures in this species have yielded circulating androgen levels at the high end of the normal breeding season range (Duckworth et al., 2001; Stoehr and Hill, 2001; Duckworth et al., 2004). The efficacy of implants was validated in this study by radioimmunoassay (see Results).
Six days prior to experimental infections, all birds were placed into 46 46 76 cm cages, which were divided into two compartments using a plane of 1 cm wire mesh affixed along a central diagonal. One compartment housed a focal bird, while the other compartment either housed a stimulus bird or remained empty (“no neighbor” treatment). The orientation of diagonal dividers (back‐left to front‐right or vice versa) was divided equally within social treatments. Focal birds were always housed in the front compartment to facilitate the recording of video for behavioral analyses. Birds could both see and hear other birds (in the same and separate cages), though the fine mesh eliminated physical contact. Each bird had ad libitum access to water and food, with feeding stations placed next to one another along the central barrier to maximize aggressive interactions between the focal and stimulus cagemates. Water dishes were placed away from the barrier. Social treatments were divided equally among five rooms maintained at identical light and temperature conditions. Vaccination of Stimulus Birds To minimize the chance that stimulus birds would contract MG during the experiment, which would likely affect their behavior toward focal birds (Bouwman and Hawley, 2010), we experimentally infected stimulus birds with MG prior to the start of the experiment and subsequently treated them with tylosin (Tylan, Elanco, Greenville, IN, USA), a broad spectrum antibiotic. Birds were inoculated in the palpebral conjunctiva of both eyes using 25 mL of Mycoplasma gallisepticum in Frey's media. We used an expansion of the 7th in vitro passage of an isolate of MG collected from a house finch with conjunctivitis in Virginia in 1994 (7994‐ 1 7P 2/12/09; D. H. Ley, North Carolina State University, College of Veterinary Medicine, Raleigh, NC, USA)(Ley et al., 1996). The MG strain and volume used were the same as that used for focal birds (see below). However, the vial used to inoculate stimulus birds had been thawed and refrozen prior to use, which reduces MG viability. Before freeze/thaw, the viable count of the inoculum was 2.24 107 color changing units (CCU) per milliliter, determined by the most probable number method (Meynell and Meynell, 1970).
Table 1. Sample sizes of focal birds in each social treatment, partitioned by population of origin. Social treatment
Birds from Alabama Birds from Virginia Total Focal Birds
No neighbor
Sham‐implanted neighbor
Testosterone‐implanted neighbor
3 (1 excluded) 4 7
7 2 9
5 3 (1 excluded) 8
Two birds were excluded from analyses because they were found to have low levels of Mycoplasma gallisepticum before the start of the experiment (see Methods).
J. Exp. Zool.
AGGRESSION AND DISEASE IN HOUSE FINCHES Beginning three weeks after inoculation and continuing for seven weeks, stimulus birds were given drinking water with 2 mg/ mL tylosin mixed in each day. This protocol ensured that stimulus birds would have antibodies against the same strain of MG used in the experiment, so would be unlikely to contract the pathogen during interactions with focal birds. Following treatment, none of the stimulus birds showed clinical signs of infection. Experimental Infection of Focal Birds Six days after being housed in experimental social groups, focal birds were inoculated with the same volume and strain of MG as above (Fig. 1). The vial used for this inoculation had not been repeatedly frozen, so its viable count was 2.24 107 CCU/mL. Behavior Three days prior to the inoculation of focal birds, and on days 4, 16, and 30 post‐inoculation, we recorded video of all birds from 08:00 (lights‐on) through 11:00. Cameras were placed in each room before lights‐off the day prior to recording (approximately 16:00). On the day of recording, cameras were turned on manually while rooms were still dark (07:30–07:50). From these videos, behaviors of focal birds were observed during two 15 min sampling periods, the first beginning immediately after lights‐on (08:00), the second beginning approximately 1.5 hr later (09:20– 09:30). We assessed the following behaviors relevant to the prototypical sickness behaviors of lethargy and anorexia: the number of hops or flights, the number of bill swipes, the amount of time spent immobile (no detectable movement at all), the time spent within one body‐width of the feeder, and the number of pecks made at food. In addition, we quantified the instances of aggression initiated by and received by the focal bird. Aggression was defined as a rapid movement (by either focal or stimulus bird) made directly at the other bird. An aggressive encounter was considered “escalated” if the focal bird made physical contact with the barrier separating cage compartments. Because involvement in an aggressive encounter can alter an animal's physiology, regardless of whether that animal initiated aggression or not (Wingfield et al., 1990; Creel et al., 2013), we considered the total observed instances of aggression as the most appropriate metric for this study. For analyses, we therefore summed aggression received and aggression initiated by the focal bird to yield a single value of total aggression in each cage. Because other authors have reported asymmetrical immune and physiological changes in response to aggression received or reductions in social status (Tuchscherer et al., 1998; Devoino et al., 2003; Hawley, 2006), we note that our results are qualitatively identical whether analyzing aggression received, aggression initiated, or total aggression. We defined the outcome of an aggressive interaction as a “win” if the focal bird maintained his position and the stimulus bird moved away after the interaction, and as a “loss” if the stimulus
43 bird maintained his position and the focal bird moved away after the interaction. The outcome was defined as “neutral” if both birds maintained their positions for 3 sec or longer after the interaction. We classified focal birds as dominant, subordinate, or equal to their neighbors using pre‐infection data on aggression. Specifically, for each focal bird, we subtracted the total number of losses during pre‐infection sampling from the total number of wins during pre‐infection sampling. Positive numbers indicated that focal birds were dominant, negative numbers indicated that focal birds were subordinate, 0 indicated birds were equal. Of all aggressive interactions recorded, 66% resulted in a clear winner and loser. Considering only pre‐infection sampling, this percentage was the same. Conjunctival Pathogen Load and Pathology Pathogen load was assessed using previously published methods (Hawley et al., 2011b). Briefly, for each eye, a sterile cotton swab was rotated for 5 sec on the inside of the conjunctiva after being dipped in tryptose phosphate broth (TPB). The swab was then swirled in a microcentrifuge tube containing 300 mL of TPB and wrung out on the side of the tube. These samples were then frozen at 20 °C for up to four months. DNA was extracted from these samples using Qiagen DNeasy 96 Blood and Tissue kits (Qiagen, Valencia, CA, USA). We then performed quantitative polymerase chain reaction, probing for the Mycoplasma mgc2 gene on a MyiQ Single Color Real‐Time PCR Detection System (Bio‐Rad Laboratories, Hercules, CA, USA), using previously published parameters and primers (Grodio et al., 2008; Hawley et al., 2011b). Conjunctival pathology was assessed using eye scores, based upon a 4‐point scale, as defined in previously published studies (e.g., Hawley et al., 2011b). Briefly, this scale is defined as: 0) no pathology, 1) slight redness and swelling of the conjunctiva, 2) moderate swelling and some eversion of the conjunctiva, 3) severe swelling of the conjunctiva and noticeable exudate. Hormone Levels Blood samples were taken by puncturing the wing vein using a 26 gauge needle and collecting blood into heparin‐coated capillary tubes. All samples were collected between the hours of 11:00 and 12:00, local time. All samples used to assess corticosterone were obtained within 3 min of entering the experimental rooms; samples for testosterone were taken within 10 min. Samples for both androgens and corticosterone were taken from focal birds one day before infection (five days after co‐housing) (Fig. 1). To ensure the efficacy of hormone implants in stimulus birds, samples for androgens were taken from these animals six days after implantation (six days before co‐housing) and 32 days after implantation (20 days after co‐ housing; Fig. 1). All samples were kept on ice and centrifuged to separate plasma within 2 hr of collection. Plasma was then frozen at 20 °C for up to two months. J. Exp. Zool.
44
ADELMAN ET AL.
Table 2. Summary of the principal components analysis on sickness behaviours (using metrics of locomotion and foraging as proxies for lethargy and anorexia, respectively).
Variable Number of hops and flights Number of bill‐swipes Proportion of time immobile Proportion of time within one body‐width of feeder Number of pecks at food Cumulative proportion of variance explained
PC1
PC2
Rotation PC3
PC4
PC5
0.39 0.33 0.51 0.44 0.54 0.45
0.62 0.11 0.34 0.61 0.34 0.67
0.34 0.90 0.11 0.24 0.01 0.85
0.44 0.02 0.71 0.22 0.51 0.95
0.39 0.25 0.34 0.57 0.58 1.00
Concentrations of total androgens and corticosterone were measured in plasma samples using a direct radioimmunoassay after extraction in dicholoromethane, using published methods (Wingfield et al., '91; Moore et al., 2002). All samples were run in duplicate. Androgens were measured in a single assay (intra‐assay % CV ¼ 16.8, detection limit 0.13 ng/mL); corticosterone was measured in a separate, single assay (intra‐assay % CV ¼ 10.7, detection limit 1.0 ng/mL).
Data Analysis Data were analyzed using mixed effects models in R version 11.1 (Pinheiro and Bates, 2000; R Development Core Team, 2011). Data on corticosterone, total androgens, and total aggression were transformed to better meet the assumptions of homoscedascity and normally distributed errors (corticosterone and total androgens were log transformed; total aggression was log þ 1 transformed). All other variables were untransformed. Because birds were captured from two populations several months apart, and either population origin or time in captivity could affect physiological responses to infection (Bonneaud et al., 2011; Bonneaud et al., 2012; Adelman et al., 2013a; Adelman et al., 2013b), we included population as a random effect in all analyses. Initially, random effects allowed the effect of social treatment to vary with population of origin. However, in no case did this improve the fit of the models (likelihood ratio tests, in all cases P > 0.3), so final random effects structures controlled only for different intercepts between populations. In all initial models, we included prior housing condition (with or without female) as a fixed effect. This term was removed from all but one model, as it did not improve model fits (see results). In addition to the above analyses testing for differences among treatment groups, to determine if corticosterone could affect disease responses independent of treatment, we also performed linear regressions of pre‐infection corticosterone levels against J. Exp. Zool.
responses to infection (pathogen load, eye pathology, and sickness behaviors). Finally, we asked whether initial condition (quantified as the residuals of a regression of mass vs. tarsus before co‐housing occurred) either differed across treatment groups or predicted individual variation in corticosterone levels. For analysis of sickness behaviors, the five behavioral metrics described above were collapsed into a single variable using principal components analysis. The first principal component (PC1) explained 46% of the variation in all behaviors, with active behaviors (bill swipes, hops and flights) loading with the same sign as feeding behaviors (proportion of time spent within one body‐width of the feeder, number of pecks at food) and with the opposite sign of proportion of time spent immobile (Table 2). PC1 was used to provide a single metric for all subsequent analyses of sickness behavior. Measurement of total aggression and circulating hormones pre‐infection were taken from single time points. However, post‐ infection responses (pathogen load, eye score, and PC1 above) were measured at several time points. To generate a single data point for each bird that accounted for both the magnitude and duration of post‐infection responses, we estimated the integral of each variable across time as in Adelman et al. (2013b). Briefly, we calculated the area under a curve generated for each response, with the response variable on the y‐axis and time on the x‐axis (including the pre‐infection metrics at x ¼ 0, and post infection metrics on days 4, 7, 14, 21, 28, and 35 after inoculation for pathogen load and pathology; and post infection metrics on days 4, 16, and 30 after inoculation for behavioral responses, see Figure 1). We assumed a linear change in y‐variables between each time point. Two focal birds, one originally assigned to the “no neighbor” group (from AL) and one originally assigned to the “testosterone neighbor” group (from VA), were omitted from all analyses, as each was found to have detectable levels of MG (over 300 copies of the mgc2 gene detected) one day before experimental
AGGRESSION AND DISEASE IN HOUSE FINCHES inoculation. Sample sizes given above and in Table 1 reflect these omissions.
RESULTS Testosterone Treatment of Stimulus Birds Both before and after being housed in social treatments, stimulus birds treated with testosterone implants had significantly higher levels of circulating total androgens than those treated with empty implants (mean 1SE ng/mL, Before Co‐Housing: testosterone: 13.16 1.04, empty: 0.16 0.04; After Co‐Housing: testosterone: 9.00 0.90, empty: 0.34 0.20; implant type: F1,14 ¼ 149.09, P < 0.001, implant type x date: F1,16 ¼ 20.51, P 0.18). Initial body condition, which could influence subsequent levels of aggression, hormones, or responses to infection, did not differ across social treatments (F2,20 ¼ 0.51, P ¼ 0.61, data not shown) or with population of origin (Likelihood ratio (model with and without random effect of population) ¼ 1.26, P ¼ 0.26). Contrary to our predictions, pre‐infection corticosterone levels, pre‐infection androgen levels, and post‐infection responses to MG did not differ with social treatment (Fig. 2C–G, all F2,20 0.26). Circulating androgen concentrations
45 were low among focal birds, with 9 of 24 pre‐infection samples falling at or below the detection limit of the assay. Additionally, including population of origin as a random effect improved the model of corticosterone levels, with pre‐ infection corticosterone higher in birds from Virginia than Alabama (Virginia: 7.16 0.94, Alabama: 3.12 0.29; Likelihood ratio (model with and without random effect of population) ¼ 10.79, P < 0.001). Similar analyses showed a trend for sickness behaviors to be lower among birds from Virginia (Likelihood ratio (model with and without random effect of population) ¼ 2.55, P ¼ 0.11), but no other patterns with population emerged for aggression, androgen levels, or responses to infection (all Likelihood ratios (models with and without random effect of population) 0.49). Results were similar when analyzed by dominance rather than social treatment. As with social treatment, dominance status within a cage did not influence pre‐infection corticosterone levels or post‐infection responses to MG (eye pathology, sickness behaviors, or pathogen load) (data not shown, all F3,19 0.66). “Testosterone neighbor” males tended to have fewer aggressive interactions with a clear winner (median ¼ 60%) than “sham neighbor” males (median ¼ 80%, Wilcoxon test P ¼ 0.06), suggesting that dominance may have been slightly less stable in the “testosterone neighbor” treatment. However, the percentage of aggressive interactions in which the focal bird was the clear winner did not influence pre‐infection corticosterone levels or post‐infection responses to MG (eye pathology, sickness behaviors, or pathogen load) (data not shown, all F1,12 < 0.93, all P > 0.36). Corticosterone and Responses to Infection: Individual‐Level Analyses Regressions independent of treatment group revealed that pre‐ infection corticosterone levels correlated negatively with eye pathology and sickness behaviors, but not pathogen load during infection (eye score, Fig. 3A: F1,22 ¼ 4.93, P ¼ 0.04, r2 ¼ 0.18; sickness behaviors, Fig. 3B: F1,22 ¼ 7.59, P ¼ 0.01, r2 ¼ 0.26; pathogen load, Fig. 3C: F1,22 ¼ 2.17, P ¼ 0.16, r2 ¼ 0.09). Because population of origin correlated with pre‐infection corticosterone levels (see above), population of origin was not included as a covariate in this analysis. When these regressions were performed separately for each population, estimated slopes were similarly negative, though no relationships were statistically significant (all F1,13(AL) or 1,7(VA) < 1.59, P > 0.23, r2
