Communal nesting is the optimal strategy for heat conservation in a social marsupial: lessons from biophysical models

Organisms exist in thermodynamic equilibria with their environment (Porter and Gates, 1969), and cold conditions create especially tough challenges for small, endothermic animals with high rates of heat dissipation, owing to their high surface area-to-volume ratio (Canals et al., 1989). Although fur and feathers provide good insulation, at very small sizes these are not enough to maintain sustained euthermia. Thus, additional strategies were promoted by natural selection, among which some are physiological, such as increase in thermogenic capacity and heterothermic responses (Bozinovic et al., 1987; Nespolo et al., 1999; Sharbaugh, 2001), while some others are behavioral, such as the formation of clusters or huddling, the use of nests or a combination of both (i.e. communal nesting) (Bozinovic et al., 1988; Canals et al., 1989; Canals et al., 1997; Gilbert et al., 2010; Scholander, 1955).

The fact that clustered individuals conserve heat more efficiently than isolated ones is a generalized biophysical consequence of the exponential reduction of surface-to-volume ratio (Canals et al., 1997). Thus, it is not surprising that huddling is widespread in small birds and mammals of cold environments (e.g. marsupials: Fisher et al., 2011; rodents: Antinuchi and Busch, 2001; Kauffman et al., 2003; birds: Lubbe et al., 2018; see reviews in Gilbert et al., 2010; Wojciechowski et al., 2011). However, communal nesting is also attributed to other factors, such as social behavior, protection and structural support (reviewed by Ebensperger and Labra, 2020). For instance, in the Siberian flying squirrel, clustering behavior is explained by subsequent mating rather than kinship or thermoregulating benefits (Selonen et al., 2014). The avian cup-shaped nest design is primarily explained by structural support, and not insulation (Heenan and Seymour, 2011). Similarly, in the social degu (Octodon degus), communal nesting is explained by kinship, rather than thermoregulation (Ebensperger and Bozinovic, 2000; Ebensperger et al., 2004). However, it has been challenging to test these competing hypotheses on live animals, either in the laboratory or in the field, owing to the obvious limitations of obtaining the animals, the precision of the technique for measuring energy consumption, or even obtaining the nests.

Here, we performed a laboratory study with biophysical models focused on the social marsupial monito del monte (Dromiciops gliroides Thomas 1894), which, according to several observations in both captive and free-ranging individuals, exhibits advanced levels of sociality (reviewed in Nespolo et al., 2022b). They normally cluster together, at any season, in groups that can have 10 individuals (typical cluster size is 3–4) (Franco et al., 2011; Nespolo et al., 2022a) and build sophisticated nests (Vazquez et al., 2020). Detailed descriptions of Dromiciops biology, physiology and life cycle were recently published (Fonturbel et al., 2022; Nespolo et al., 2022b). Briefly, D. gliroides is an arboreal, nocturnal small mammal (∼30 g) endemic to the temperate rainforests of southern South America, which includes high Andean locations where winter temperatures reach freezing values, but also mild locations near the coast with mean winter temperatures of approximately 5°C (Gurovich et al., 2015; Mejías et al., 2021). Two species of monito are now recognized (D. gliroides and D. bozinovici), which are intimately associated with the temperate rainforests of southern South America, an ecosystem that is patchily distributed over a latitudinal range of approximately 1000 km at both sides of the Andes range (from 35°S to 43°S). The activity cycle of monitos is highly seasonal, characterized by a hibernation period of approximately 6 months and a breeding period of 4 months, followed by a fattening stage (see details in Nespolo et al., 2022b). However, monitos are opportunistic heterotherms with variable hibernation periods depending on latitude and food availability. For instance, active animals have been observed during winter (Nespolo et al., 2022a; Oda et al., 2019), and torpid animals during breeding periods (females with pups, all torpid, see Nespolo et al., 2021).

Monitos elaborate complex spherical nests, built by interlaced quila leaves (Chusquea quila, an endemic bamboo), covered with moss inside, and the whole structure is usually located within tree cavities (Celis-Diez et al., 2012; Franco et al., 2011; Honorato et al., 2016; Vasquez et al., 2018). It is likely that monitos build communal nests as a heat conservation and energy saving strategy, which could explain the origin of their sociality. However, two facts complicate this conclusion. First, measurements of energy expenditure in clustered and isolated individuals did not produce significant differences (Franco et al., 2012). Second, as it is known that marsupials have a tropical origin (Mitchell et al., 2014), it could be the case that the sociality of monitos preceded the colonization of cold environments (Fonturbel et al., 2022; Nespolo et al., 2022b; Russell, 1984). Therefore, showing a net energetic benefit of communal nesting over solitary strategies would support the idea that sociability arose in part from having to face a cold environment.

Here, we performed laboratory experiments for simulating clustering, combined with nest use, and tested whether these strategies are synergistic in conferring energetic savings. If communal nesting represents an adaptive strategy to cope with the cold, then clustered individuals within nests should significantly reduce the endothermic cost of thermoregulation compared with isolated individuals without nests.

We used real monito nests obtained from a bird conservation program run by one of the authors (T.A.). We had access to 240 nesting boxes located at Pucón (39°18′51″S, 71°52′50″W, 1100 m.a.s.l.) that are normally colonized by birds in spring and monitos in autumn. During the 2019 winter, we obtained 17 monito nests (dry mass 55.1±2.4 g, mean±s.e.m.) (Fig. 1) found within nest boxes at the end of the winter. Each nest was individually placed in a paper bag and stored until the start of our experiment, 2 months later.

The taxidermic models (‘mannequins’, hereafter) were manufactured with skins of dead monitos, provided by the National Services for Wildlife (Servicio Agrícola y Ganadero and Corporación Nacional Forestal, Chile). Nine carcasses were used, which were freshly dissected and left for 2–3 days stretched, covered by a layer of sodium chloride and boron sulphate to dry out. Two skins were needed to manufacture one individual, thus we built five mannequins in total (Fig. 1). Each mannequin was fabricated using a 40 g agar bead, because it has similar thermal properties with living tissue. In the range of 5–40°C, agar has on average, a conductance of 0.56 W m⁻¹ K⁻¹ (Zhang et al., 2011), whereas the thermal conductance of living soft tissue (e.g. brain and abdomen) ranges from 0.53 to 0.55 W m⁻¹ K⁻¹ (Rough and Bharathan, 2005). The agar was covered by the skin and sutured with surgical sutures. In the middle of the agar sphere, a 4×9 mm (diameter×length) cylindrical space was left to introduce the temperature sensor (see below), and a Velcro closure was included for opening. The agar was prepared with 3 g of agar–agar and 100 ml of boiling water. The sphere was wrapped in a rubber balloon so that the agar did not dehydrate at high temperatures. The mannequin's mass was 55.1±2.4 g, with a volume of 39.8±2.7 ml and a density of 1.4±0.09 g ml⁻¹. This is within the range of well-fed monitos (typical of autumn; range: 30–60 g; see Bozinovic et al., 2004; Nespolo et al., 2021).

For the cooling curves, we used two ‘Pelt’ chambers (60×40×30 cm, Sable Systems, North Las Vegas, NV, USA) to maintain constant temperatures, one for warming up the mannequins to 40±1°C and the other for cooling it at 5±1°C (the actual cooling curve). This temperature is representative of the coastal habitat of D. gliroides, which exhibits a winter mean temperature of 5.5°C (max.=8.1°C; min.=3.3°C) (Nespolo et al., 2022a).

We placed a wire frame in the floor of the chamber so that neither mannequins nor nests were in direct contact with the chamber material. For each cooling experiment, the procedure consisted of removing the mannequin from the warm chamber, waiting for it to cool down to 35°C (the euthermic body temperature of a monito), and placing it within the cold chamber. We performed two cooling curves per day. To continuously record the core temperature of the mannequins, we used temperature data loggers (model DST-micro, Star Oddi, Iceland; cylindrical 25 mm long and 5 mm in diameter, resolution: 0.003°C) with a data-gathering frequency of one per minute. According to the manufacturer, the devices are calibrated at the factory over a temperature range of 5 to 45°C. We also tested the devices in a beaker with water at 40°C that was allowed to cool to room temperature (10°C), with temperature recorded every 2 min, using a copper-constantan thermocouple (Cole Parmer). The linear regression between water and logger temperature (20 points) was highly significant (R²=0.99, P=0.001). The environmental temperature within the chambers was recorded, in addition to the chamber's thermocouple, with a HOBO environmental temperature data logger (model pro v2, Onset Computer Corporation, Bourne, MA, USA) attached to the internal walls of the chamber, with the probe hanging in the air. This probe was configured to record one datapoint per minute. Every cooling trial had a duration of 593 min, and we performed a total of 50 curves combining solitary, clustered and individuals within nests. This was an arbitrary duration decided after inspecting a few preliminary trials, which indicated that temperatures reach an asymptotic value after approximately 8 h (500 min; see Fig. 4C). We ran five curves with a bare datalogger (the data logger without agar nor fur), five curves with the agar sphere (no fur), 10 curves with a solitary mannequin (data logger with agar with fur), 10 curves with a solitary mannequin within a nest, 10 curves with a group of three mannequins and 10 curves with a group of three mannequins within a nest. For visualization (and for solely illustrative purposes), we took infrared thermographic images at different stages of the warming using a thermograph (FLIR i40, Sweden). The thermograph was located 25 cm away from the mannequins and set to an emissivity of 0.98. Images were taken at 0, 4 and 8 h in solitary mannequins, and also repeated for grouped mannequins. These were different trials than those used for the presented analyses (for which we did not open the chamber). Each cooling curve was adjusted to a one-phase exponential decay given by the equation: y=(y₀–plateau)e^−k×time+plateau, where y is the temperature at each time interval, y₀ is y at time=zero, the plateau is the asymptotic temperature of the taxidermic model, and k is the rate constant, expressed as the reciprocal of time (the higher the k, the higher the heat loss rate). We also calculated the time constant (τ), as the reciprocal of k, and the half-time, as ln(2)/k.

In addition to estimating curve parameters, we predicted the euthermic cost of maintenance by calculating the energy that a monito would spend to maintain its body temperature constant (E_cost), for every minute. Thus, we used Newton's cooling equation for passive cooling, which is normally used for estimating metabolic rates in steady-state conditions among endotherms (Mejías et al., 2022; Nicol and Andersen, 2007; Rezende and Bacigalupe, 2015), and calculated E_cost as: C(T_B−T_A), where C is the minimum thermal conductance of euthermic D. gliroides (C=3.4848 J g⁻¹ h⁻¹ °C⁻¹) (Bozinovic et al., 2004), T_B is body temperature (°C) and T_A is ambient temperature (°C). We also calculated the overall E_cost (=net E_cost) for the entire cooling curve. Curve parameters (i.e. rate constant, half-time, time constant) and net E_cost were compared among the six treatments: bare data logger, data logger with agar, data logger with fur (=mannequin), grouped mannequins (three individuals), individual mannequin within a nest and grouped mannequins within a nest. These were compared with a one-way ANOVA, using Statistica software (StatSoft, Tulsa, OK, USA). Parametric assumptions were checked using Levene’s test for homogeneity of variance and the Shapiro–Wilk test for normality.

We found significant differences among treatments in rate constant (one-way ANOVA, F_5,43=495.5, P=0.0001), half-time (F_5,43=65.7, P=0.0001) and time constant (F_5,43=65.8, P=0.0001; see Table 1). The agar (i.e. the difference in cooling rate of the data logger with and without agar) reduced the rate of heat loss by a factor of 12 (from 0.37±0.02 to 0.031±0.002 min⁻¹, Fisher post hoc test, P=0.001; Table 1). In contrast, the fur (i.e. the difference in cooling rate of the mannequin with and without skin) reduced the rate of heat loss by a factor of 2.3 (from 0.031±0.002 to 0.014±0.0006 min⁻¹, Fisher post hoc test, P=0.001; Table 1). However, the rate constant was not significantly different between solitary, grouped, solitary in nests and grouped in nests mannequins (Fisher post hoc tests, P>0.6; Table 1).

The infrared thermal images showed how the core of the body maintained the heat (33.8°C, in red) until the fourth hour of the cooling, while maintaining a cooler envelope (skin with fur, in yellow, Fig. 2A–D). These images also show, qualitatively, the better heat conservation of clustered individuals compared with solitary individuals. The adjustment of every one of the 50 cooling curves had an R² greater than 0.98 (all curves are plotted in Fig. 3).

In terms of heat conservation, grouped mannequins within nests conserved heat 31% longer (half-time: 91.7±4.71 min; Table 1) than solitary mannequins in nests (half-time: 63.2±2.8 min; Fisher post hoc test, P=0.008; Table 1). However, there were non-significant differences between half-times of solitary individuals in nest and grouped individuals without a nest (63.2±2.7 versus 64.1±3.6; Fisher post hoc test, P=0.45; Table 1). Solitary individuals without a nest (half-time: 52.6±3.0 min) conserved heat 18% lower than grouped individuals without a nest (half-time: 64.1±3.6 min; Fisher post hoc test, P=0.001; Table 1).

The amount of metabolic heat that a monito should produce every minute to maintain a constant body temperature represents the euthermic cost of maintenance (E_cost). This was calculated using Newton's passive cooling equation (see Materials and Methods) and estimated for clustering mannequins (mimicking a group of monitos, as shown in Fig. 4A) and mannequins within nests (Fig. 4B), and is presented in Fig. 4C. This comparison shows that the configuration that minimizes E_cost is to be grouped within a nest (red line in Fig. 4C). Summing E_cost values across the cooling period confirms that the net euthermic cost of maintenance is minimized by the combined use of nest in groups (Fig. 4D). Indeed, significant differences were found after a one-way ANOVA (F_5,44=49.4, P=0.0001; Fig. 4D). Specifically, being grouped in the nest has an E_cost (44.5±0.83 kJ) that is significantly lower than that of being grouped outside a nest (Fisher post hoc test, 48.1±1.38 kJ, P=0.008). However, being solitary in the nest (50.8±0.62 kJ) is equivalent (not significantly different) to being grouped without a nest (48.1±1.38 kJ; Fisher post hoc test, P=0.07). There were significant differences in E_cost between the mannequin with and without fur (51.4±0.55 versus 62.1±1.85 kJ; Fisher post hoc test, P=0.0001). Finally, the calculated E_cost of the bare data logger and the mannequin without fur was not significant (64.2±1.70 versus 62.1±1.85 kJ; Fisher post hoc test, P=0.26).

There is a historic discussion about whether nest sharing in clusters has a bioenergetic significance (i.e. a heat-conservation strategy) or represents a by-product of the benefits of group living (Dausmann and Glos, 2015; Ebensperger, 2001; Franco et al., 2011; Gilbert et al., 2010; Heenan and Seymour, 2011; Lubbe et al., 2018; Madikiza and San, 2020; Olson et al., 2018; Selonen et al., 2014; Vickery and Millar, 1984; Vogt and Lynch, 1982; Willis and Brigham, 2007; Withers and Jarvis, 1980). The problem arises because often, species with advanced levels of sociality also benefit from clustering during cold periods. For instance, in the woodland dormouse (Graphiurus murinus), social clustering seems to be explained by mating behavior rather than by thermoregulatory benefits (Madikiza and San, 2020), and in North American flying squirrels (genus Glaucomys), interspecific social nesting appears to be driven by sociality (Olson et al., 2018; Selonen et al., 2014). However, in birds, it has been documented that social thermoregulation combined with the use of insulative roosts and nests provides energy savings of over 50% compared with isolated birds outside nests (Lubbe et al., 2018). This is also the case of Siberian hamsters, which compensate for the reduction in insulation owing to experimental haircuts by using nest materials that were provided experimentally (Kauffman et al., 2003). Thus, in general, grouping and nest use reduce the total conductance of the group, reducing the lower limit of thermoneutrality of the whole cluster (Bozinovic et al., 1988). Therefore, our results, indicating that the optimal strategy is clustering and nest use, are in line with thermodynamic predictions.

In the field, monitos exhibit high phenotypic flexibility in nest-building behaviour and may build nests within tree cavities or in more exposed locations on tree branches (Vazquez et al., 2020). Also, both the mass and the volume of the nests increase with elevation, thus suggesting that animals build more insulated nests in colder environments (Altamirano et al., 2019). This strategy is combined with huddling behavior, as larger groups are often found in comparatively colder locations (Altamirano et al., 2019; Celis-Diez et al., 2012). Indeed, according to Canals et al. (1997), the average relative surface area reduction during huddling in small mammals ranges from 28.7 to 39.1%, where the maximum reduction in surface area is attained with three individuals. Indeed, according to Bozinovic et al. (1988), the reduction in minimum thermal conductance owing to huddling and nest use is approximately 42% (compared with animals housed individually), and represents an important fraction of the energy budget of a small marsupial or rodent, especially in winter. However, these authors used n=5 individuals and experimental temperatures of −10°C, which is outside the natural range for D. gliroides. In our experiments (with minimum T_A of 5°C), the nests provided additional energy savings of approximately 10% (a reduction in E_cost from 49 to 44 kJ; see Fig. 4D). Considering that the basal metabolic rate of D. gliroides is approximately 13 kJ day⁻¹ (Nespolo et al., 2022a), these 5 kJ represent almost half of energy consumption per day in a resting, thermoneutral condition.

In heterothermic species [i.e. animals that express daily or multiday torpor (hibernation), sensuRuf and Geiser, 2015], of which D. gliroides is an example (Nespolo et al., 2021), huddling has been associated with increased survival (Boratyński et al., 2015; Boyles et al., 2008; Patil et al., 2013). However, when torpid, animals reduce the set point of temperature regulation, thus promoting heat loss instead of heat conservation, until this set point is attained (Geiser, 2011; Humphries et al., 2002; Mejías et al., 2022; Nespolo et al., 2021). Then, if communal nesting is a strategy for conserving heat, it should be beneficial at relatively high temperatures, where animals promote euthermic thermoregulation (Nespolo et al., 2022a; Nowack and Geiser, 2016). Although several results support the idea that heterothermic species cluster together when euthermic (Bustamante et al., 2002; Gilbert et al., 2010; Jefimow et al., 2011; Nespolo et al., 2022a; Willis and Brigham, 2007; Wojciechowski et al., 2011), extreme hibernators (e.g. animals that hibernate near 0°C, such as arctic ground squirrels, hoary marmots or monitos at high Andean locations) tend to hibernate solitarily (Barnes, 1989; Mejías et al., 2021; Patil et al., 2013).

Recently, an experimental field study conducted on D. gliroides conclusively showed that overfed individuals that aborted hibernation formed larger groups and used experimental hibernacula more frequently, compared with those that received less food and hibernated normally (which were found more often outside the hibernacula; see details in Nespolo et al., 2022a). Thus, the ‘torpor-prone’ individuals (i.e. animals with a fall in blood glucose, the main trigger of torpor, see Lo Martire et al., 2018; Westman and Geiser, 2004) search for cold places and avoid huddling to minimize energy expenditure. This has been observed both in D. gliroides (Nespolo et al., 2022a) and in hibernating bats, which even experience internal migrations during short arousals, within roosting caves, to select cooler sites (Ryan et al., 2019). On the contrary, individuals reluctant to use torpor (i.e. those that are well fed) promote euthermic thermoregulation by selecting refugia and huddling. This explains the non-significant differences found by Franco et al. (2012) when comparing energy expenditure of clustered and isolated animals in the laboratory: torpid and active animals were mixed in the chamber. These authors concluded that the main response of D. gliroides to low ambient temperature was reduced body temperature and torpor, irrespective of huddling. Now we know, from the cited field studies, that this is incorrect.

In summary, the results presented in the present study, together with the cited field experiments performed in D. gliroides, in our opinion clarifies the fact that communal nesting (including the crucial isolating effect of the Dromiciops nest) is essential to this species during cold periods, and when euthermic.

The use of biophysical models in physiological ecology dates back to Porter and Gates's seminar paper (Porter and Gates, 1969; Tracy, 1972; see an update in Kearney and Porter, 2009), which defined the energy budget of animals as a balance between energy losses and gains in a variety of environments, organism types and thermal conditions. Later, a number of authors applied them in different contexts (e.g. climatic impact on species range: Huey et al., 2012; Kearney et al., 2009; heat loss and behavioral thermoregulation in endotherms: Kenagy et al., 2002; McCafferty et al., 2011; reconstructing dinosaur physiology: Seebacher, 2003), and recently, for predicting habitat suitability for mammalian conservation (McComb et al., 2021). Here, we show that simple simulations could provide quantitative data for testing a specific hypothesis about the origin of sociality in a marsupial. Today, Dromiciops seems to be the only American marsupial with advanced levels of sociality, in which social groups seem to be maintained in time, above the level of the family (Nespolo et al., 2022b). Our results supporting a net energetic benefit of communal nesting warrant further field studies for analyzing how this social behavior extrapolates in nature.