A test of altitude-related variation in aerobic metabolism of Andean birds

Endotherms that are native to high-altitude environments must contend with physiological challenges posed by the reduced partial pressure of O₂ (P_O2) and low ambient temperature (T_a). Reduced P_O2 may compromise the maximum capacities for aerobic exercise (MMR; maximum metabolic rate) and shivering thermogenesis because of the reduced availability of O₂ to fuel ATP synthesis (Chappell et al., 2007; Hayes, 1989a; McClelland and Scott, 2019; Storz and Scott, 2019; Storz et al., 2010). BMR may be elevated in highland species as a result of increased thermoregulatory demands or as a correlated response to changes in MMR that entail higher maintenance costs (Hayes and Garland, 1995; Portugal et al., 2016; Rezende et al., 2004). Non-proportional changes in BMR and MMR entail changes in absolute aerobic scope, defined as the difference between the two rates (MMR–BMR), which reflects an animal's capacity to increase its rate of aerobic metabolism above maintenance levels (Bennett, 1991; Hochachka, 1985).

In comparison with lowland relatives, mammals native to high altitude often have higher mean MMR in hypoxia and suffer a smaller decrement in MMR with increasing hypoxia (Chappell and Dlugosz, 2009; Cheviron et al., 2012, 2014; Lau et al., 2017; Lui et al., 2015; Schippers et al., 2012; Storz et al., 2019; Tate et al., 2017, 2020). In addition, when measured at their native elevations, BMR is consistently higher in high-altitude deer mice (Peromyscus maniculatus) relative to lowland conspecifics (Hayes, 1989a,b), although it is not known to what extent the elevated BMR reflects an evolved change or a reversible acclimatization response. Available evidence for birds is not conclusive on whether BMR varies with elevation, and it is unknown whether MMR exhibits a consistent pattern of altitudinal variation among species.

Studies evaluating changes in BMR with elevation show mixed results. Highland and lowland populations of rufous-collared sparrows (Zonotrichia capensis), measured at their native altitudes, had no differences in BMR (250–4540 m; Castro et al., 1985), thermogenic capacity (600–3000 m; Novoa et al., 1990) or field metabolic rates (600–3000 m; Novoa et al., 1991). High elevation amethyst sunbirds (Chalcomitra amethystina) had higher BMR even after acclimatization to low elevation, but the effect was statistically significant only in winter (540–1550 m; Lindsay et al., 2009a,b). In contrast, high elevation fiscal shrikes (Lanius collaris) showed lower BMR after acclimatization (130–1800 m; Soobramoney et al., 2003). Interspecific comparisons of birds of paradise found that high elevation (>1000 m) species had higher BMR but only after accounting for diet (McNab, 2003). However, a recent study involving a phylogenetically diverse set of more than 250 Neotropical bird species found no significant association between BMR and native elevation (400–3000 m; Londoño et al., 2015). The effects of altitude may be more readily detectable in fine-grained comparisons between pairs of closely related species that are native to different elevations but that are otherwise ecologically similar. Moreover, it remains unclear whether exercise-induced MMR and aerobic scope exhibit consistent patterns of altitude variation.

We measured BMR, MMR and aerobic scope in 10 Andean passerines, representing five pairs of closely related species with contrasting elevational ranges (Fig. 1A,B). We used a paired-lineage design (Felsenstein, 2004) such that the five pairwise comparisons were phylogenetically independent (Fig. 1A).

Birds were captured between June and August from 2017 to 2019 at several field sites in the western Andes of Colombia (Fig. 1A; Table S1). We compared closely related species that had contrasting elevation ranges (Fig. 1B) but which are otherwise similar ecologically (Table S2). High elevation species were captured between 2300 and 2500 m; low elevation species were captured between 500 and 1400 m. Annual mean temperature and ambient P_O2 differed, respectively, by approximately 3.6°C and 2.3 kPa for the spinetails; 6.1°C and 3 kPa for the thrushes; 6.7°C and 3.7 kPa for the wood wrens and the warblers; and 5.9°C and 2.9 kPa for the tanagers.

We mist-netted birds (aided by playback of vocalizations) during the day (09:00 to 18:00 h). We released juveniles (identified on the basis of plumage or color of bill gape) and adults with brood patches. We measured MMR immediately after capture and subsequently kept birds inside cloth bags (∼20×30 cm) in a dark and isolated room to minimize stress. Keeping birds inside cloth bags will cause stress; however, most individuals remained calm for the time they were restrained and seemed less agitated than when we tried retrieving them from cages immediately before BMR experiments. We took the birds out of the bags every 2 h to provide water and food (banana for fruit-eating birds, mealworm larvae and adults for insectivorous birds) offered ad libitum while they were hand-held for a few minutes. To ensure that the birds were post-absorptive (Karasov and del Rio, 2007), we stopped providing food and water 4 h before the onset of BMR measurements. All BMR measurements were conducted after sunset, between 19:00 h and 01:00 h. Body mass was measured immediately after capture and also before and after BMR measurements. Birds were euthanized for tissue collection after all measurements were taken, and deposited at the ICESI zoological collection. All procedures were approved by the University of Nebraska IACUC (project ID 1499) and Colombian research permits granted to G.A.L. (Permit 536, May 20/2016).

We used a positive-pressure flow-through respirometry system to measure metabolic rates. Incurrent air was dried with silica gel and the flow was then divided into four metered channels using a FlowBar (Sable Systems, Las Vegas, NV, USA). One channel was used for reference air. The other three supplied air continuously to three acrylic metabolic chambers (2.7 liters), each equipped with a thermocouple that measured excurrent air temperature. Another thermocouple measured T_a in the incubator. We measured one to three birds per night and matched incurrent airflow into chambers (200–1000 ml min⁻¹ at standard temperature and pressure (STP)) with the body mass of each tested bird. Excurrent air flows were sampled sequentially by a multiplexer (Sable Systems RM-8). Subsamples (50–200 ml min⁻¹) of excurrent airflow were scrubbed of CO₂ and H₂O (using soda lime and silica gel, respectively) and routed through a Sable Systems FoxBox to measure O₂ content, which was calibrated against atmospheric air (20.95% O₂). O₂ content deflections ranged between 0.5% and 1.5%. Flow rates were adjusted to STP conditions, as the mass flow meters in the FlowBar compensated for variations in atmospheric pressure and temperature. Birds were allowed at least 30 min of acclimation to experimental conditions before the measurements. Each bird was monitored for 15 min and reference air was measured for 2.5 min before switching between individuals; this pattern was repeated until the measurements were finished. T_a was kept constant (±0.5°C) using a PELT-5 controller and a PTC-1 cabinet (Sable Systems). Birds were first measured at T_a=34°C and then at T_a=30°C, remaining at each stable temperature for at least 1 h. We recorded T_a, flow rate and O₂ content every second using Warthog LabHelper (www.warthog.ucr.edu) interfaced to a Sable UI-2 A-D converter.

We elicited MMR using forced exercise in a hop-flutter wheel (Chappell et al., 1999). This method reliably elicits behavioral exhaustion, with repeatable MMR (Chappell et al., 1996), but may not elicit maximum power output of the flight muscles. Accordingly, rates of oxygen consumption (V̇O₂) values measured with this technique are usually lower than those measured in wind tunnels (Chappell et al., 2011; McKechnie and Swanson, 2010). We ran these experiments at ambient temperature, between 15 and 18°C at high elevation and at 20–26°C at low elevation. During exercise trials, incurrent air flow was 750–1000 ml min⁻¹ and no multiplexer was used. The chamber (14.5 liters) was manually rotated for a maximum of 10 min or until birds showed signs of exhaustion (coordination loss and heavy breathing) and we verified that they reached their maximum rates of oxygen consumption – typically shortly after the beginning of the exercise bout. All tested birds attained similar levels of behavioral exhaustion during the exercise trials and no birds were injured during experiments. Birds were then fed and allowed to rest for at least 4 h before the onset of the BMR experiments (see field protocol section).

BMR was computed as the lowest continuous average V̇O₂ over 5 min during periods of low and stable V̇O₂, and the lowest value of the two temperature measurements per bird (30 and 34°C) was chosen for subsequent analyses. Before obtaining MMR V̇_O2,max, we applied the instantaneous correction (Bartholomew et al., 1981) to compensate for the mixing characteristics of the system (i.e. the blunted response to rapid changes in O₂ concentration). We calculated MMR as the highest continuous averaged V̇O₂ over 1 min during periods of high and stable V̇O₂ values. Finally, we calculated the absolute aerobic scope for each bird as the difference between MMR and BMR.

We used log-transformed metabolic rates in interspecific comparisons and we included log-transformed body mass as a covariate in the analyses. For each species and metabolic measurement, we discarded data points that fell outside ±2 standard deviations from the mean, resulting in the removal of two data points for BMR, six for MMR, and five for aerobic scope.

Second, in order to better understand the trends observed in the linear models, we mass-corrected each of our observed V̇O₂ values (BMR, MMR, aerobic scope) by dividing them by M_b^S (Gillooly et al., 2001), where M_b is body mass for each individual and S is the allometric scaling coefficient obtained by McKechnie and Wolf (2004) for passerine birds (S=0.667). We then used t-tests to evaluate the statistical significance of the difference in mass-corrected V̇O₂ between high- and low-elevation species within each pair. Considering that scaling coefficients for BMR, MMR, and aerobic scope may vary, we ran additional analyses using the coefficients reported by McKechnie and Swanson (2010), but the overall patterns remained unchanged (results not shown). We also ran a post-hoc power analysis (package pwr v1.3; https://cran.r-project.org/web/packages/pwr/) using Hedges’ G (Hedges, 1983) to estimate effect sizes with a significance level of 0.05.

Finally, to assess the influence of other variables on metabolic rates, we built linear mixed models that included various combinations of mass and the interaction between species group (categorical; 5 levels) and elevation (categorical; high, low) as fixed factors, and age (categorical; immature, adult), sex (categorical; male, female), molt state (categorical; absent if bird had no developing feathers, moderate if few, abundant if several), and year of capture (2017, 2018, 2019) as random factors. In all mixed effect models the variance explained by any variable other than mass was negligible (usually<<0.1), and the model that included mass as the only predictor had the lowest AIC (ΔAIC=46; results not shown).

We measured 96 wild-caught birds, with sample sizes of 8–12 individuals per species (Fig. 2; Table S3), except for Turdus serranus (n=4). Both the linear mixed model and the ANCOVA explained a high percentage of variation in the data (average R² for mixed model: 0.98; average adjusted R² for ANCOVA: 0.69; Tables S4 and S5). Elevation was correlated with BMR, MMR, and aerobic scope, but only after accounting for the effects of species pair and body mass, as evidenced by the difference between the mean marginal R² (0.006; variance explained by the fixed effect) and the mean conditional R² (0.98; variance explained by the full model) (Table S4). Likewise, after accounting for mass, elevation also had a significant effect on variation in BMR, MMR and aerobic scope in each species pair (Table S5), with the exception of BMR in thrushes. In all species pairs other than thrushes, high-elevation species had higher metabolic rates than their lowland counterparts, as indicated by the negative slopes estimated for the effect of elevation on each species group (Table S5). These results are noteworthy considering the relatively small elevational difference (around 1500 m) between our sampling localities.

Our pairwise comparisons of mass-corrected metabolic rates revealed significant differences between high- and low-elevation species in most but not all cases. BMR was significantly higher in high-elevation wrens, tanagers and warblers (Fig. 3A; Table S3), MMR was significantly higher for high-elevation wrens, tanagers and warblers (Fig. 3B) and aerobic scope was significantly higher for high-elevation thrushes, tanagers, and warblers (Fig. 3C). Spinetails did not exhibit significant elevational differences in BMR, MMR or aerobic scope. Our results are similar to previous studies in birds that found either no differences in BMR with elevation (Castro et al., 1985; Londoño et al., 2015) or higher BMR in high elevation populations (Lindsay et al., 2009a,b; McNab, 2003).

The available evidence suggests that the relationship between elevation and avian metabolic rates likely depends on each species’ ecology. It has been hypothesized that BMR may depend on food quality, availability and predictability (food habits hypothesis; McNab, 1988, 2019). Previous studies have shown that changes in BMR with elevation are more apparent in species with more frugivorous diets (McNab, 2003), similar to the case of the tanagers in our study, although a direct test of this hypothesis did not support it (Sabat et al., 2010). Similarly, McNab (2019) found that BMR variation in birds is correlated with the frequency and intensity of flight, with higher rates in more active birds. In our set of sampled taxa, comparisons between species with greater dispersal propensity (i.e. tanagers and warblers) revealed larger elevation-related differences in metabolic rates than in comparisons between pairs of more sedentary species (i.e. spinetails and wrens) (Sheard et al., 2020). Our results and those of other comparative studies suggest that there is no simple, unitary explanation regarding causes of elevational variation in aerobic metabolism, and it seems equally clear that relevant interactions between ecological and physiological factors vary among taxa.

In our small sample of Andean passerines, we found a significant effect of elevation on BMR, MMR and aerobic scope after accounting for effects of body mass (Tables S3–S5). In most cases, high-elevation species had higher BMR and MMR than closely related and ecologically similar low-elevation species. Overall, we observed significant differences in rates of aerobic metabolism between individual pairs of species, but we did not document a uniformly consistent elevational trend. Our study and most others to date have investigated elevational variation in aerobic metabolism by measuring wild-caught birds in their native habitat (e.g. Jones et al., 2020 preprint; Londoño et al., 2015, 2017). In the future, common-garden or reciprocal-transplant experiments that control for acclimatization effects should help reveal whether bird species native to high elevations have generally evolved increased aerobic performance capacities in cold, hypoxic conditions. Such experiments would complement results of comparative studies by revealing the genetic and environmental components of variation in aerobic metabolism within and among species.

We thank all the students whose invaluable help in the field made this project possible, especially to Jorge Lizarazo, Isabel Cifuentes, Valentina Echeverry, Jose Riascos, and Maria Laura Mahecha. For facilitating logistics at fieldwork sites in Colombia we thank the staff at ICESI University, staff at CELSIA, PNN Farallones de Cali, Alirio Bolivar, Ana Tulia Montes, and Gustavo Giraldo. Additional help with fieldwork preparation was provided by Andrew Crawford and Santiago Herrera. Gwen Bachman offered vital guidance while processing metabolic rate data. We are thankful to Kate Lyons, John DeLong, Graham Scott, Grant McClelland, and Elizabeth G. King for discussions and assistance with the statistical analyses. This work benefited from discussions with Gwen Bachman, Chris Witt, Kristi Montooth, Colin Meiklejohn, and Jamilynn Poletto. Finally, we thank two anonymous reviewers for their comments, which greatly improved a previous version of this manuscript.