Available online at www.sciencedirect.com
General and Comparative Endocrinology 156 (2008) 27–33 www.elsevier.com/locate/ygcen
Repeatability of baseline corticosterone concentrations L. Michael Romero *, J. Michael Reed Department of Biology, Tufts University, Medford, MA 02155, USA Received 2 August 2007; revised 24 September 2007; accepted 1 October 2007 Available online 5 October 2007
Abstract One major assumption for endocrine studies is that hormone titers are consistent within an individual so that if hormone titers are low relative to the cohort on one day, they are relatively low compared to the cohort on other days. This is an especially important assumption for most field studies where researchers may have access to an individual animal only once. We used a laboratory study with captive house sparrows (Passer domesticus) to test this assumption using glucocorticoid titers. Baseline corticosterone titers were measured five different times for each bird under six different experimental conditions: during both day and night while birds were held on a short day photoperiod (11L, 13D), a long day photoperiod (19L, 5D), and while birds were undergoing a prebasic molt. Although the variation within an individual was often larger than the variation between individuals, the relative ranks of birds compared to their cohort were consistent during the night in all three conditions. In contrast, during the day the relative ranks of birds compared to their cohorts were only consistent on short days; on long days and during molt there was no significantly consistent ranking among individuals. Furthermore, the overall rank of an individual in its cohort was often different during the day and night. These data indicate that it is not always a good assumption that birds can be categorized as individuals with higher and lower titers, which will complicate analyses of the causes of interindividual variation. Ó 2007 Elsevier Inc. All rights reserved. Keywords: Individual variation; Baseline; Basal; Corticosterone; Birds
1. Introduction Studies of Hypothalamic–Pituitary–Adrenal (HPA) axis function in free-living or captive wild animals often focus on glucocorticoid titers, and for studies in birds, the major glucocorticoid of interest is corticosterone (Holmes and Phillips, 1976). Standard practice for these studies is to determine baseline, or near baseline, corticosterone concentrations by taking blood samples within minutes of capture (Romero and Reed, 2005; Romero and Romero, 2002), then determining stress-induced corticosterone concentrations using handling or restraint (Wingfield and Romero, 2001). Alternatively, researchers can collect a general measure of corticosterone by analyzing fecal steroid metabolites (Millspaugh and Washburn, 2004).
*
Corresponding author. Fax: +1 617 627 3805. E-mail address:
[email protected] (L.M. Romero).
0016-6480/$ - see front matter Ó 2007 Elsevier Inc. All rights reserved. doi:10.1016/j.ygcen.2007.10.001
Different individuals under the same conditions (i.e., captured and bled at the same time) differ in baseline and stress-induced corticosterone concentrations. This individual variation is typically ignored during analysis except for looking for patterns associated with life-history differences, such as sex (e.g. Romero et al., 2006), age (e.g. Angelier et al., 2007; Heidinger et al., 2006), body mass or condition (e.g. Silverin et al., 1997), or season (reviewed by Romero (2002)). In making these comparisons, it is assumed, usually implicitly, that the corticosterone titer measured for an individual represents its true state, so patterns driving individual differences could be discerned if they are biologically significant. In other words, it is assumed that patterns in individual differences found on one sampling occasion would be found consistently across sampling occasions. If this assumption were false, low individual consistency in repeated samples of corticosterone titers could mask potential patterns among individuals. For example, research specifically on differences among
28
L.M. Romero, J.M. Reed / General and Comparative Endocrinology 156 (2008) 27–33
individuals, or where breeding high-corticosterone and low-corticosterone lineages is needed as a tool, presumes that single samples are representative of the individuals (and for the latter type of study, that they are heritable) (Evans et al., 2006; Jones et al., 1994a). For a recent review of why understanding individual differences is important, see Williams (in press). It is uncommon to compare replicate samples from the same individual under the same conditions, especially for wild animals and in field studies (Williams, in press). Consequently, this assumption is largely untested. In fact, we could find only three studies from the published literature that compared corticosterone titers from multiple samples of the same individual for a wild bird species (Cockrem and Silverin, 2002; Kralj-Fiser et al., 2007; Love et al., 2003). In general, these studies reported high repeatability of corticosterone titers within an individual. That this consistency has a genetic basis is supported by successful artificial selection studies for high and low corticosterone titers in domesticated Japanese quail, chickens, and turkeys (Brown and Nestor, 1973; Gross and Siegel, 1985; Jones et al., 1994b; Satterlee and Johnson, 1988; Satterlee and Jones, 1997). Our goal was to determine the consistency of baseline corticosterone values across sampling periods for captive house sparrows (Passer domesticus). In particular, we wanted to know if there was consistency among individuals in their relative baseline corticosterone values; i.e., does an individual with a low corticosterone titer relative to other individuals during one sampling period consistently have relatively low corticosterone titers at other sampling periods? This is a different approach to investigating the question of individual consistency of corticosterone titers than that posed by earlier studies because we are looking at relative corticosterone titers across individuals rather than evaluating variance in absolute measures.
19L:5D cycle with lights on at 02:00), and a long-day photoperiod during which the house sparrows were undergoing prebasic molt. House sparrows usually undergo a prebasic molt between August and October (Lowther and Cink, 1992), beginning to replace feathers three to four months after the onset of long days. Birds remained on long days until they began to molt. Birds were given a minimum of 2 weeks to acclimate to the new photoperiod before sampling commenced. We took five blood samples at the same time of day from each bird, and repeated this for each experimental condition. Under each experimental condition, blood samples were obtained within 2–3 min of entering the room by puncturing the alar vein. Based on field studies, samples collected within 2 min are equivalent to baseline concentrations, and those collected between 2–3 min are near baseline (Romero and Reed, 2005). We collected approximately 40 ll of the upwelling blood in heparinized microhematocrit tubes and used cotton to stanch the bleeding. Separate samples were collected during both day (10:30–11:00) and night (22:00–22:30) for all three experimental conditions. All samples were spaced over at least a 6-week period and no bird was bled more than twice per week to ensure that the birds had sufficiently replenished their blood volume between bleeds and HPA axis function had recovered from the previous sampling. We also verified that the birds were unaffected by the repeated sampling by ensuring that they lacked any symptoms of long-term, or chronic, increases in corticosterone titers (such as loss of body weight, lethargy, etc. Sapolsky et al., 2000). Samples collected during the night were done while using a white light bulb filtered to allow only blue light into the room. This provided sufficient light for sample collection, but does not penetrate the avian skull to stimulate photoreceptors on the pineal as do other wavelengths (Oishi and Lauber, 1973). Lighting malfunctions and a few mortalities over the long time in captivity prevented all birds from being sampled during all six periods (day and night during three experimental conditions). Birds were entered into the experiment at different points so that each experimental condition had some birds that were included in another experimental condition and some birds that were not. This meant that samples were independent within an experimental condition, but not between experimental conditions. Blood samples were immediately centrifuged at 400g for 5 min and plasma was extracted and frozen. Corticosterone concentrations were measured using a radioimmunoassay (RIA) described by Wingfield et al. (1992) and used previously for house sparrows (Rich and Romero, 2001). Intra- and inter-assay variability was less than 7 and 12%, respectively.
2. Materials and methods 2.2. Statistical procedures 2.1. Birds and assays The data reported here were gathered in conjunction with another published study (Romero and Rich, 2007) using wild house sparrows. Here, we report corticosterone concentrations from the baseline bleeds, which were not reported in the previous study. The experimental design and housing conditions are repeated here. Nineteen wild house sparrows were captured in Eastern Massachusetts and brought into captivity. Birds were a mixture of males and females, but no distinction was made for sex since earlier work indicated that male and female captive house sparrows do not differ in their HPA axis responses (Rich and Romero, 2001). Birds were housed in large indoor flight aviaries for at least two weeks to acclimate to captivity and subsequently transferred to an experimental room. They were then housed in pairs in separate cages with all cages placed next to each other so that no bird was isolated. The temperature in all rooms was set at 25 °C and birds were provided food and water ad libitum. All procedures were performed according to AALAC procedures and approved by the Tufts University Institutional Animal Care and Use Committee. The experiments were performed under three conditions: when the experimental room was held on a short-day photoperiod (an 11L:13D light:dark cycle with lights on at 07:00), a long-day photoperiod (a
We were interested in determining whether relative baseline corticosterone levels for an individual are consistent across time. That is, does a bird with relatively low corticosterone values compared to other birds in a sample on one day, have relatively low values on other days? The contrast, or null hypothesis, would be that a bird with a relatively low baseline corticosterone value on one day compared to other birds would have a randomly ranked value on other days relative to new samples from the same birds. To test this hypothesis, we took blood samples for measuring baseline corticosterone from a suite of birds on 5 different days. For the first set of samples (those designated Day 1), we ranked individuals from lowest to highest corticosterone values. This was repeated for each of the next four sets of bleeds, resulting in five sets of ranks, each from 1 to n (where n was the number of individuals in a sample). We then did a 1way analysis of variance (ANOVA) looking for consistency in ranks among individuals across sampling times. In the model, individual identity was the independent variable, and we were looking to explain variance in ranked corticosterone values. If the null hypothesis is true, an individual’s rank should be inconsistent (random) across samples, resulting in a lack of significant effect. If ranks are consistent within individuals, we would reject the null hypothesis, concluding that individual identity explains variability in ranks.
L.M. Romero, J.M. Reed / General and Comparative Endocrinology 156 (2008) 27–33 Analyses were done separately by season and by time of day (day vs. night) because corticosterone concentrations are known to vary daily and seasonally in both captive (Rich and Romero, 2001) and free-living (Romero et al., 2006) house sparrows. Of the 111 potential data points in our combined analyses, we were missing three values. Rather than remove the remaining 12 records of these individuals from our analyses, these missing values were assigned a rank that was the nearest integer to the mean of the other four ranks; this created a small bias toward increasing consistency of ranks. Even though all of the corticosterone measures were baseline or nearbaseline, there is still some increase in corticosterone levels over time after first disturbance (Romero and Reed, 2005). This variability would tend to obscure the patterns we are investigating. Consequently, bleed time was included as a covariate in each of the analyses and Type III statistical values are reported. That is, we evaluated the importance of individual identity on consistency of rank hormone values after the effects of bleed time had been removed. All analyses were done with SAS 9.1 (GLM procedure). For each analysis of ranked corticosterone values, we also calculated the repeatability statistic (r) described by Lessells and Boag (1987), where r = s2A/(s2 + s2A). Variance components are calculated from the ANOVA table, where s2 = error mean squares and
29
s2A = (model mean squares s2)/n0; for all of our analyses; n0 = 5, number of repeated values for each individual. (Lessells and Boag refer to ‘model mean squares’ as ‘mean squares among individuals’, and ‘error mean squares’ as ‘mean squares within individuals’; we use the former terms because they are commonly used in the output of many statistical packages.).
3. Results There was considerable individual variation in corticosterone concentrations between different baseline bleeds (Figs. 1–3, left hand panels). In many cases, the within subject variation seems as large as the between subject variation. However, variability of ranked hormone levels do show some consistent relationship between individual identification and ranked corticosterone concentrations (Figs. 1–3, right hand panels). Note that because the right hand panels are graphing mean and variability (standard error) of ranked hormone val-
Fig. 1. Corticosterone concentrations in captive house sparrows held on a short day photoperiod. Left hand panels show each individual corticosterone value for each individual bird sampled during the day (upper panels) and night (lower panels). Individual birds are placed on the X-axis in order of increasing mean corticosterone concentrations from the five bleeds. Note that rankings are different during the day and night, as are the ranges on the Yaxis. Right hand panels depict means ± SE of ranked hormone levels from the five separate rankings. Labels for individual animals are consistent across panels and across figures. Dotted lines indicate a perfect ranking, where the bird with the lowest corticosterone mean is ranked #1 all five times, the bird with the second lowest corticosterone mean is ranked #2 all five times, etc.
30
L.M. Romero, J.M. Reed / General and Comparative Endocrinology 156 (2008) 27–33
Fig. 2. Corticosterone concentrations in captive house sparrows held on a long day photoperiod. See legend of Fig. 1 for explanation of panels.
ues, the variation for each individual is not equivalent to the variation in actual corticosterone titers shown in the left-hand panels. For birds held on short days (Fig. 1), corticosterone concentration rankings were consistent across individuals both during the day (F12,64 = 2.50, p < 0.012; r = 0.23) and night (F12,64 = 2.98, p < 0.004; r = 0.19). Likewise, corticosterone concentration rankings were consistent across individuals at night held on long days both while not (Fig. 2; F13,69 = 4.98, p < 0.0001; r = 0.33) and while (Fig. 3; F11,59 = 2.62, p < 0.01; r = 0.26) undergoing a prebasic molt. However, the null hypothesis (rankings were not consistent among individuals) could not be rejected during the day for birds held on long days (Fig. 2; F13,69 = 0.89, p = 0.57; r = 0.05) or during the day for molting birds (Fig. 3; F9,49 = 1.72, p = 0.12; r = 0.09). Note also that the ranking of an individual during the day was often different from the ranking of that same individual at night. A graphical representation comparing mean corticosterone titers during day and night in each treatment group illustrates that the means are not consistent across individuals (Fig. 4).
4. Discussion Of the six experimental treatments, four showed significant consistency of corticosterone titers between repeated bleeds. Three other studies on individual variation in corticosterone titers from wild birds also reported consistent individual responses. The first study to look for individual repeatability in corticosterone responses in wild birds was by Cockrem and Silverin (2002), who bled the same 13 captive great tits (Parus major) on each of three different occasions and measured baseline and stress-induced corticosterone titers at
