A comparative analysis of thermogenic capacity and cold tolerance in small birds

Small birds wintering in temperate regions generally show markedly improved cold tolerance capabilities relative to summer-acclimatized birds(Marsh and Dawson, 1989;Swanson, in press). This winter acclimatization primarily results from an increased ability to sustain high levels of shivering thermogenesis over prolonged periods (Marsh and Dawson,1989). In birds showing marked winter improvement of cold tolerance, this improvement is also associated with expanded M_sum (summit metabolism or thermogenic capacity),typically measured by indirect calorimetry as the maximal rate of oxygen consumption under cold stress (Dawson and Smith, 1986; Marsh and Dawson,1989; Swanson,1990a; Cooper and Swanson,1994; Liknes and Swanson,1996; Liknes et al.,2002; Cooper,2002). Furthermore, birds that show relatively minor seasonal differences in cold resistance also show no, or only minor, seasonal differences in M_sum(Dawson et al., 1983a; Saarela et al., 1989, 1995; Swanson and Weinacht, 1997). Thus, during winter acclimatization the expanded M_sum is closely associated with increased shivering endurance at submaximal levels of cold challenge. Indeed, Swanson(2001) demonstrated that shivering endurance under a standardized cold exposure was positively correlated with M_sum in three species of small passerines. Moreover, expanded endurance is tied to enhanced maximal capacities for aerobic activity in vertebrate animals generally(Bennett, 1991).

Nevertheless, seasonal changes in shivering endurance and cold resistance in some species of small birds may occur without corresponding changes in M_sum, and geographic variation in cold resistance is not always associated with variation in M_sum(Dawson et al., 1983a; Swanson, 1993). Thus, cold tolerance and M_sum do not always change in lockstep and the extent of their phenotypic correlation is uncertain. Shivering endurance and M_sum are correlated intraspecifically in small birds(Swanson, 2001), but the interspecific relationship between cold tolerance and M_sumhas not been directly examined for birds. Intraspecific seasonal changes in cold tolerance in birds are concluded either when birds tolerate a static cold exposure longer in winter than in summer (e.g. Dawson and Carey, 1976; Dawson and Smith, 1986; Cooper and Swanson, 1994) or when colder temperatures are required to induce hypothermia in winter than in summer (Saarela et al., 1989, 1995; Liknes et al., 2002). Efforts to test the interspecific relationship between cold tolerance and M_sum have not yet been undertaken, and are potentially confounded by body size effects on metabolic rates and heat loss. Testing the relationship between M_sum and cold tolerance requires a standardized cold exposure among species and measurement of either shivering endurance or the temperature inducing hypothermia. Developing a standardized measure of shivering endurance requires a standardized cold challenge for all species measured, which is difficult, if not impossible, to attain because factors such as body size and thermal conductance vary among species and greatly impact heat loss to the environment(Aschoff, 1981). One way around this problem, however, is to hold shivering endurance essentially constant while measuring the temperature in helox (79% helium/21% oxygen) required to elicit hypothermia (or the temperature at the cold limit, T_CL; after Saarela et al., 1989).

The objective of this study was to examine the interspecific relationship between cold tolerance (measured as T_CL under a sliding helox cold exposure) and M_sum in both summer and winter in a phylogenetically diverse sample of small birds. We used both standard and phylogenetically corrected methods to analyze the interspecific M_sum/T_CL relationship to determine whether phylogeny influenced any correlation between M_sumand T_CL. To our knowledge, this is the first study to directly examine, using relevant comparative techniques, whether an interspecific phenotypic correlation between cold tolerance and M_sum exists for birds.

We measured M_sum and T_CL for 21 species in summer and 11 species in winter(Table 1); these species ranged in body mass M_b from 9.2 to 62.6 g. All birds used in the present study were captured by mist net in summer (late April-August) or winter (December-February) near Vermillion, Clay County, South Dakota, USA(42°47′N, 97°0′W). The species used in this study are all common summer, winter or permanent residents in this area. Following capture,birds were transported to the laboratory where they were held at room temperature (22°C), with food (bird seed, mealworm larvae and/or mulberries) and water ad libitum, until cold exposure tests later on the day of capture. Body mass was measured to the nearest 0.1 g immediately prior to cold exposure tests, which were conducted from 09:00 h-20:00 h CST in summer and from 10:00 h-19:00 h CST in winter. Following cold exposure tests,birds were banded with a standard US Fish and Wildlife Service leg band and released at the site of capture. M_sum and T_CL were measured only once on individual birds.

Standardized conditions for determining T_CL must be delineated to use cold tolerance data for comparative analyses. Swanson et al.(1996) suggested standard methods for eliciting M_sum in birds by sliding cold exposure in helox that we adapted for measurement of T_CL. Using this method, we exposed an individual bird to a declining series of temperatures in 79% helium/21% oxygen (helox), where temperature was decreased by 3°C at 25 min after the initiation of cold exposure, and every 20 min thereafter, until hypothermia was induced. We concluded that hypothermia had occurred when oxygen consumption decreased steadily, without rebounding, over several minutes, reaching levels lower than those recorded over the preceding portion of the cold exposure test. To verify hypothermia, we removed birds from the metabolic chamber and measured body temperature with a Cole-Parmer Model 8500-40 Thermocouple Thermometer (Chicago, IL, USA) by inserting a lubricated 20-gauge copper-constantan thermocouple into the cloaca to a depth(approximately 1 cm) where further insertion did not alter the temperature reading. We considered birds with body temperature T_b<37°C as hypothermic, and birds were invariably hypothermic when the conditions noted above had been met. We defined the helox temperature at the beginning of this steady decline in oxygen consumption as T_CL. One further matter in the standardization of T_CL measurement involves the temperature at which the sliding helox cold exposure is initiated. Because T_CL is affected by body mass, to keep thermogenic endurance roughly standardized among species, cold exposure tests must begin at higher temperatures for smaller birds.

where T_CL is in °C and M_b is in g. For subsequent T_CL experiments, sliding helox cold exposure was initiated at temperatures 6°C above the allometrically predicted T_CL. The initial temperature was then modified for each species, as needed, so that hypothermia did not occur too rapidly(<45 min) or too slowly (>2 h) for comparative purposes.

In the current study, we measured T_CL concurrently with M_sum determination on individual birds. We measured summit metabolic rate by open-circuit respirometry using a sliding cold exposure in helox (Swanson et al., 1996). Briefly, we placed birds into 1.9 l or 3.8 l paint cans (depending on body size), with the inner surface painted flat black to provide emissivities near 1.0, which served as metabolic chambers. Mean effective volumes of these chambers, calculated according to Bartholomew et al.(1981), were 1917 ml and 4688 ml for the 1.9 l and 3.8 l chambers, respectively. We achieved temperature control within metabolic chambers by immersing them into a bath of water and propylene glycol (Forma Scientific Model 2095; Marietta, OH, USA), which regulated chamber temperature to ±0.5°C. Prior to immersion, we flushed the chamber for at least 5 min with helox to replace air with helox. We maintained flow rates of dry, CO₂-free, helox at 1010-1030 ml min^-1 over the course of the experiments using a Cole-Parmer Precision Rotameter (Model FM082-03ST; Chicago, IL, USA), previously calibrated to ±1% accuracy. We measured fractional oxygen content in excurrent gas leaving the chamber using an Ametek S-3A oxygen analyzer(Pittsburgh, PA, USA). We recorded fractional oxygen content every 60 s over the test period and computed oxygen consumption according to the instantaneous equations of Bartholomew et al.(1981). We then calculated consecutive 10 min means for oxygen consumption rates over the test period(1-10, 2-11, etc.) and considered the highest 10 min mean, excluding the initial 10 min of measurements), as M_sum(Dawson and Smith, 1986). We corrected all values for oxygen consumption to STPD and converted oxygen consumption to metabolic rates (in W) by assuming an energy equivalent of 20.1 J ml^-1 O₂.

During cold exposure treatments, we exposed individual birds to a declining series of temperatures in helox (3°C every 20 min after 25 min at the initial test temperature) until a gradual decrease in oxygen consumption indicative of hypothermia occurred. These conditions have been shown to elicit M_sum in birds(Dutenhoffer and Swanson,1996; Swanson et al.,1996). For a few individuals in winter (three horned larks Eremophila alpestris, two northern cardinals Cardinalis cardinalis), birds did not become hypothermic after 3 h of cold exposure reaching the lowest temperatures the bath was capable of attaining(approximately -17.5°C), so it is not certain that these individuals attained M_sum and they did not reach T_CL. However, oxygen consumption in these individuals was essentially constant over at least the last hour of cold exposure, despite declining temperatures in helox, so it is likely that birds were very close to M_sum. In addition, since some individuals of these species did become hypothermic at similar temperatures, it is likely that these individuals had also approached T_CL. Although mean T_CL was undoubtedly slightly underestimated for these species, this should make the interspecific M_sum/T_CL relationship conservative, as these two species had higher M_sum than was allometrically predicted (Table 1).

We analyzed the relationship between M_sum and T_CL both by conventional statistical methods and by phylogenetically independent contrasts(Felsenstein, 1985; Garland et al., 1992). For conventional analyses, we performed least-squares regressions of logM_bvs logM_sum and logM_bvs logT_CL. We then calculated residuals from these allometric equations and performed least-squares regression of residuals of logT_CL against residuals of logM_sum. While this approach controls for the effects of mass on the M_sum/T_CLrelationship, it does not account for possible phylogenetic influence on the relationship.

Consequently, we calculated phylogenetically independent contrasts (PIC)for logM_b, logM_sum and logT_CL according to Garland et al.(1992, 1993). Calculation of phylogenetically independent contrasts requires knowledge of tree topology and branch lengths, which we garnered from Sibley and Ahlquist(1990)(Fig. 1). Most species for which we measured M_sum in this study either have branch length data provided directly in the study of Sibley and Ahlquist(1990) or are closely related to species that are listed, so that branch lengths can be determined. We used arbitrary branch lengths of 1.0 in the summer analysis for divergences of chipping (Spizella passerina) and field (S. pusilla)sparrows and for Baltimore (Icterus galbula) and orchard (I. spurious) orioles (based, respectively, on divergence distances within Melospiza sparrows of 1.3 or less and a divergence distance of 1.2 for orioles and New World blackbirds; Sibley and Ahlquist, 1990). In addition, we used a branch length of 2.8 for the Bell's-warbling vireo divergence, because that is the divergence distance between congeneric blue-headed and white-eyed vireos (Sibley and Ahlquist, 1990). In addition, analyses using PIC are robust to actual branch length variation (Garland et al., 1999), so the few arbitrary branch lengths used in this study are unlikely to influence PIC results. We initially standardized contrasts by dividing by branch lengths, but absolute values of contrasts were potentially correlated with their branch lengths, so branch lengths were log-transformed after first increasing the scale of the entire phylogenetic tree by a factor of 10. This reduced correlations to non-significant levels so that contrasts were weighted equally in subsequent analyses. Standardized contrasts were positivized on M_b according to Garland et al.(1992). We then performed least-squares regression through the origin on positivized contrasts of logM_bvs logM_sum and on logM_bvs logT_CL. We calculated residuals from logM_sum and logT_CL PIC allometric regressions and performed least-squares regression on residuals of logT_CL contrasts against residuals of logM_sum contrasts to test for phenotypic correlation independent of body mass and phylogeny.

To analyze phylogenetic diversity in the relationship between M_sum and T_CL, we calculated 95%confidence intervals around allometric regression lines for raw data and PIC regressions for M_sum and T_CL. We considered values for species (raw data) or for ancestral nodes (PIC) falling outside these confidence intervals as having high or low M_sum or T_CL (for allometric regressions).

We generated M_sum and T_CL data for 21 species in summer and 11 species in winter that ranged in body mass from 9.2-62.6 g (Table 1). Mean time to hypothermia (t_hypo) for different species ranged from 33 to 99 min in summer, with most values between 45 and 90 min, and from 40-170 min in winter, with most values between 65 and 105 min, so the goal of inducing hypothermia between 45 and 120 min was met for most species(Table 1).

Slopes of logT_CLvs logM_bregressions did not differ significantly between seasons(F_1,60=1.44, P>0.05), but the winter intercept was significantly lower than the summer intercept(F_1,29=87.99, P<0.001). Residuals of logM_sum/logM_b and logT_CL/logM_b regressions were significantly negatively related in both summer (R²=0.34, P=0.006) and winter (R²=0.40, P=0.037)(Fig. 3).

Least-squares regression through the origin of phylogenetically independent contrasts of logM_sum against logM_byielded significant positive relationships for both summer and winter birds. For summer birds, regression statistics were b=0.70, R²=0.66, P<0.001. Regression statistics for this relationship in winter were b=0.50, R²=0.73, P=0.001. Regressions through the origin for logM_band logT_CL (°K) contrasts were significantly negatively related in summer birds and showed a similar non-significant trend for winter birds. Regression statistics for the summer equation were b=-0.029, R²=0.29, P=0.012. Regression statistics for the winter equation were b=-0.012, R²=0.25, P=0.124. Residuals from allometric equations for logM_sum and logT_CLcontrasts were significantly negatively correlated in both summer(R²=0.47, P<0.001) and winter(R²=0.40, P=0.049)(Fig. 4).

Species exhibiting high M_sum in summer included downy woodpecker, house wren, black-capped chickadee, house sparrow, American goldfinch and field sparrow (Fig. 5A). Species with low M_sum in summer were eastern wood-pewee, Bell's vireo, gray catbird, white-breasted nuthatch,orchard oriole and rose-breasted grosbeak. Those species with high or low M_sum also generally showed low or high T_CL, respectively. Exceptions included gray catbird and rose-breasted grosbeak, which had low M_sum but typical T_CL, house wren and black-capped chickadee, which had high M_sum but typical T_CL, downy woodpecker, which had high M_sum but high T_CL, and house finch, which had typical M_sum but low T_CL.

Winter species with high M_sum included black-capped chickadee and house sparrow, whereas those with low M_sumwere downy woodpecker and white-breasted nuthatch; these latter two species also showed high T_CL(Fig. 5B). However, even though chickadees and house sparrows had high M_sum, their T_CL was typical for allometric predictions. American tree sparrows had low T_CL, despite exhibiting typical M_sum for their body size.

PIC analyses documented ancestral nodes showing high or low M_sum or T_CL(Fig. 5). For summer analyses,nodes with high M_sum included the root node for the entire tree, the vireo node, the house sparrow-sister taxon node, the Spizella node, the warbler-oriole/cardinalid node, and the oriole node (Fig. 5A). Nodes with low M_sum were the catbird node, the chickadee-nuthatch/wren node, the nuthatch-wren node, the Spizella-warbler/oriole/cardinalid node and the oriole-cardinalid node (Fig. 5B). Nodes showing high or low M_sum also generally showed low or high T_CL, respectively. Exceptions included the root node and the warbler-oriole/cardinalid node, which had high M_sum but typical T_CL, the oriole-cardinalid node, which had low M_sum but typical T_CL, and the woodpecker and nuthatch/wren/chickadee nodes,which had typical M_sum but low T_CL. The only winter node with high M_sum was the house sparrow-finch/sparrow/cardinalid node, but this node showed typical T_CL. The only winter node with low M_sum was the nuthatch-chickadee node, which also showed high T_CL. The nuthatch/chickadee-sister taxon and horned lark-sister taxon nodes both showed low T_CL but typical M_sum.

Cold tolerance and thermogenic capacity were positively correlated in both summer and winter on an interspecific basis for both standard and phylogenetically corrected analyses in this study. This indicates that species with higher thermogenic capacity also showed greater cold tolerance, as measured by lower T_CL, suggesting that cold tolerance and thermogenic capacity are functionally linked. These data are consistent with intraspecific data on thermogenic capacity and cold tolerance in small birds. Swanson (2001) measured cold tolerance as shivering endurance under cold stress in black-capped chickadees Poecile atricapillus, dark-eyed juncos Junco hyemalis and American tree sparrows Spizella arborea, all of which showed positive correlations between shivering endurance and thermogenic capacity. Thus, both within and among species comparisons demonstrate a positive correlation between cold tolerance and thermogenic capacity, strongly suggesting a functional link between them.

Such a correlation is also generally consistent with previous data on seasonal acclimatization in small birds. A few birds exhibit seasonal changes in cold tolerance without accompanying seasonal changes in thermogenic capacity, and geographic variation in cold tolerance is not always associated with corresponding variation in thermogenic capacity(Dawson et al., 1983a; Swanson, 1993; Saarela et al., 1995). Such data have cast doubt on the generality of the correlation between thermogenic capacity and cold tolerance. However, most species of small birds do show a significant winter increment of thermogenic capacity that is associated with substantial improvements in capacity to tolerate cold temperatures(Hart, 1962; Swanson, 1990a; Cooper and Swanson, 1994; O'Connor, 1995a; Liknes and Swanson, 1996; Liknes et al., 2002; Cooper, 2002; Arens and Cooper, 2005a). Winter increments of thermogenic capacity documented in these studies range from 16-55%. If thermogenic capacity and cold tolerance are generally elevated in winter relative to summer in small birds, then regression equations of logM_sum on logM_b should be elevated,and regression equations of logT_CL on logM_b should be lower, in winter compared to summer. Such was indeed the case in this study, as slopes of these regressions did not differ significantly between seasons, but intercepts were significantly higher for M_sum and significantly lower for T_CL in winter than in summer. In general, therefore,winter birds had higher thermogenic capacity and tolerated colder temperatures in helox than summer birds. For example, according to the regression equations in this study, a 20 g bird would have a 28.2% higher M_sumand would require a helox temperature 6.8°C lower to induce hypothermia in winter relative to summer. The seasonal temperature difference for hypothermia induction in helox substantially underestimates the actual seasonal temperature difference in air, as helox markedly increases thermal conductivity relative to air in small birds(Dawson and Smith, 1986; Swanson, 1993; Cooper, 2002), so seasonal differences in cold tolerance are quite marked for the species in this study.

Thus, winter increment of thermogenic capacity appears to be a common component of seasonal acclimatization in small birds. Taken together, data demonstrating concomitant seasonal variation in cold tolerance and thermogenic capacity and direct demonstration of correlations between cold tolerance and thermogenic capacity, both within and among species, strongly suggest that physiological adjustments promoting increased thermogenic capacity in small birds also promote elevated cold tolerance. This suggests that cold tolerance(i.e. thermogenic endurance) and thermogenic capacity are functionally linked,potentially through variation in muscle mass or by adjustments of mass-specific metabolic intensity or capacity to oxidize fuels, principally fat (Dawson et al., 1983b; Marsh and Dawson, 1989;Swanson, in press). Such a link is consistent with the general vertebrate pattern of coupled variation in endurance and aerobic capacity(Bennett, 1991).

which suggests that M_sum and T_CLshould be linked (e.g. Bozinovic and Rosenmann, 1989). However, two factors could influence this purported linkage. First, variation in M_sum is not the only factor that influences T_CL. Concurrent variation in C or T_b could offset any variation in M_sum, such that M_sum and T_CL might not be correlated. In essence, testing for a correlation between M_sum and T_CL is akin to testing for how much variation in T_CL is explained by variation in M_sum, rather than by other factors that affect C or T_b. Second, substituting M_sum and T_CL into the above equation assumes that T_CL always occurs concurrently with M_sum, but this is often not the case, as the highest metabolic rates (M_sum) during cold exposure treatments,such as those in this study, usually occur well before temperatures eliciting hypothermia (Swanson, 2001). Thus, substituting M_sum and T_CL into the equation describing metabolic rates in endotherms is probably not strictly appropriate.

R² values for regressions of residuals from allometric equations for M_sum and T_CL ranged from 34-47% in this study, indicating that interspecific variation in thermogenic capacity explained a substantial portion of the interspecific variation in cold tolerance. However, substantial variation in cold tolerance still remains unexplained, which suggests a role for other factors in affecting differences in cold tolerance among species and seasons. Such factors could include differences in insulation, control over thermal conductance, circulatory and ventilatory differences (Swanson,1990b; Breuer et al.,1995; Arens and Cooper, 2005a,b),or metabolic adjustments promoting shivering endurance without affecting thermogenic capacity (Marsh and Dawson,1982; Yacoe and Dawson,1983; Marsh et al.,1990).

Because seasonal acclimatization in birds is largely a metabolic process,with only a minor role played by seasonal changes in insulation(Dawson et al., 1983b; Marsh and Dawson, 1989; Swanson, 1991a), metabolic adjustments should play a prominent role in explaining both seasonal and interspecific variation in cold tolerance. Such metabolic adjustments could include those affecting fuel mobilization and supply to shivering muscles, as well as those promoting preferential use of lipid to fuel shivering(Marsh and Dawson, 1982; Yacoe and Dawson, 1983; Marsh et al., 1990; Swanson, 1991b; O'Connor, 1995b). These adjustments would not necessarily be reflected by increases in thermogenic capacity, but could increase cold tolerance by elevating the percentage of thermogenic capacity that could be sustained for prolonged periods. This model for seasonal variation in cold tolerance was posited by Marsh and Dawson(1989), largely from studies on American goldfinches and house finches. Liknes et al.(2002) termed this model the variable fraction model, because the model contends that it is the fraction of thermogenic capacity that is sustainable which varies seasonally, rather than the thermogenic capacity. In contrast to this model is the variable maximum model (Liknes et al., 2002),which posits that it is thermogenic capacity that varies seasonally. The winter increment of thermogenic capacity, in turn, increases thermogenic endurance in the cold, because as thermogenic capacity increases, the absolute rate of sustainable heat production also increases, even if the fraction of thermogenic capacity that is sustainable remains seasonally constant. Because the data in this study indicate a winter increment of thermogenic capacity and directly document a correlation between thermogenic capacity and cold tolerance in small birds, they are consistent with the variable maximum model. However, it is important to note that metabolic adjustments promoting maintenance of a higher sustained fraction of thermogenic capacity could further improve cold tolerance, and therefore might help account for some of the unexplained variation in cold tolerance in this study.

Some interesting general trends emerged from analyses of phylogenetic diversity in the relationship between M_sum and T_CL. For summer analyses, the root node had high M_sum, but typical T_CL based on allometric predictions, whereas in winter the root node was typical for both parameters. The summer data suggest that ancestral species had high thermogenic capacity, but were relatively poorly insulated, resulting in relatively poor cold tolerance for their metabolic abilities. However, in winter, where taxa not resident in cold climates were absent from the analyses, the root node was typical for both M_sum and T_CL, suggesting that it is taxa not resident in cold climates that were driving the uncoupling of M_sum and T_CL from summer analyses. Another factor likely influences this uncoupling, however, and that is the absence of a winter increase in M_sum in downy woodpeckers in this study. Because downy woodpeckers had high M_sum in summer and low M_sum in winter, and woodpeckers were one of the sister taxa at this node, the nodal values were likely influenced by the absence of a seasonal difference in M_sum in this species. The lack of a seasonal difference in M_sum in downy woodpeckers differs from that previously documented for this species by Liknes and Swanson(1996), where M_sum in winter was 52% greater than that in summer. The reason for the difference between these two studies is unknown, but may involve differences in winter weather among years, which can impact metabolic rates in birds (Swanson and Olmstead,1999).

Another noteworthy finding from summer analyses was that high M_sum and low T_CL, as well as low M_sum and high T_CL, occurred in taxa composed solely of migrants, as well as taxa with members wintering in cold climates. This suggests that physiological capacities for heat production or cold tolerance are not the sole determinant of wintering strategy within a taxon. Finally, although deviations from allometric predictions for M_sum and T_CL were usually coupled for species and for ancestral nodes, this was not always the case. This again suggests that while thermogenic capacity is a prominent factor influencing cold tolerance, there is still room for factors other than thermogenic capacity in establishing differences in cold tolerance among species and seasons.

Funding for this study was provided by NSF EPSCoR grant 0091948, the American Philosophical Society, and the USD Office of Research. We thank Karen Olmstead for advice on statistical analyses. Birds were captured under federal collecting permit 758442 and South Dakota collecting permits. All procedures in the study approved by the University of South Dakota IACUC and were consistent with guidelines provided by the American Ornithologists' Union.