1 Introduction

Social bees are able to control the nest environment to allow optimal performance of the adults as well as optimal development of the brood (Wilson 1971; Jones and Oldroyd 2007). In particular, social bees are sensitive to, and invest considerable effort in controlling, the temperature within the nest. Honeybees (Apis mellifera) are very effective at controlling their internal nest temperature. The workers can perform quite well over a wide range of temperatures, but the brood is very sensitive to temperature fluctuations. The cognition ability of adult bees especially to foraging activities are affected even by 2 °C of difference between temperatures they were raised (Tautz et al. 2003). Moreover, mortality and malformation increase when brood is reared outside of a very narrow range of optimal temperatures (Himmer 1927; Heinrich 1993; Mardan and Kevan 2002)

Nest temperature can be increased through metabolic heating generated by the shivering of flight muscles in bumblebees (Hymenoptera, Apidae, Bombini) and honeybees (Hymenoptera, Apidae, Apini) (Jones and Oldroyd 2007). These groups also have the ability to cool their nests. The honeybee is able to maintain the internal nest temperature within a narrow range of 33 to 36 °C, even when the external temperatures are higher than 40 °C (Heinrich 1993). To accomplish this, workers fan their wings to create an air flow between the inside and the outside of the nest. This behaviour is combined with water collection, where small droplets of water are deposited on surfaces on the inside of the nest. This reduces the temperature of the nest through evaporative cooling (Lindauer 1954). In Bombus, only fanning behaviour has been reported and is correlated with an increase in nest temperature (Jones and Oldroyd 2007).

Stingless bees (Hymenoptera, Apidae, Meliponini) are the largest and most diverse group of highly eusocial bees (Michener 2000). They are restricted to tropical and subtropical habitats and are, therefore, often exposed to high temperatures. However, there is no evidence that they perform behaviours to actively cool the nest (Jones and Oldroyd 2007). Research shows that the internal nest temperature differs from ambient temperature, but most authors attribute this to the behaviour of nesting in insulated cavities and assume that this is the main way by which stingless bees avoid nest temperature rising to damaging levels during hot periods (Engels et al. 1995; Nogueira-Neto 1997; Jones and Oldroyd 2007). On the other hand, wing fanning has been described in several species in this group, leading some researchers to suggest that this might be a cooling behaviour (Nogueira-Neto 1948; Macías-Macías et al. 2011). However, this hypothesis has yet to be tested. Engels et al. (1995), working with Scaptotrigona postica, observed that colonies experiencing high temperatures had elevated numbers of bees performing wing fanning. Macías-Macías et al. (2011) demonstrated that workers of Melipona colimana kept individually in boxes inside incubators perform wing fanning behaviour when exposed to temperatures of 40 °C. However, Moritz and Crewe (1988) pointed out that wing fanning plays a role in air circulation inside the nest, regulating the CO2 and O2 levels inside the nest cavities of the ground nesting species Trigona denotii. Nogueira-Neto (1948) also reports the occurrence of fanning behaviour in stingless bee colonies during cold winter days, suggesting that this behaviour is related to gas exchange, rather than temperature regulation.

There are a few reports of water collection in stingless bee species (Nogueira-Neto 1997; Macías-Macías et al. 2011). However, the authors associate this behaviour with individual ingestion of water and minerals and reject the hypothesis that it plays a role in the cooling of the nest because the deposition of water droplets was never observed inside colonies and because cavity nesting is considered to provide a thermally stable environment (Engels et al. 1995). On the other hand, Macías-Macías et al. (2011), working with individual workers of Melipona colimana, found that high temperatures led to an increase in water consumption, and that workers regurgitated the water on the bottom of the boxes. The authors suggested that this individual behaviour could be used at the colony level to decrease the temperature of the nest.

In preliminary observations, we noted that Scaptotrigona depilis can maintain relatively stable temperature inside the nest during periods of extreme heat. This is in line with the observations of Engels et al. (1995), showing that colonies of Scaptotrigona postica, a closely-related species, are able to maintain the temperature of the brood area around 40 °C when exposed to high temperatures (close to 44 °C). Given the scattered, but suggestive, evidence for active cooling of the nest environment in stingless bees, we performed an experimental study to test the hypothesis that S. depilis has the ability to actively cool the nest environment. We addressed two main questions: First, to what extent is S. depilis able to cool the nest when exposed to high temperatures? Second, are there behavioural mechanisms for nest cooling? To answer the second question, we analysed two candidate behaviours, wing fanning and water collection. We also investigated the effects of temperature on the pupae survival and development time.

2 Material and methods

2.1 Study species and study site

The study was performed on the Ribeirão Preto campus of São Paulo University, SP, Brazil, during summer time (from December 2011 to March 2012—maximum temperature of 40 °C), where Scaptotrigona depilis is a common stingless bee species. This species was chosen because many aspects of its biology have been described in the literature, including some responses to high temperature, such as wing fanning behaviour (Engels et al. 1995). Colonies were kept outdoors, unshaded, in white wooden boxes covered by tiles under ambient conditions of light and temperature (Fig. S1A). Colonies usually contain several thousand workers and one, singly mated queen (Paxton et al. 2003).

2.2 Experimental setup

  1. a)

    Ability to cool the nest

    To investigate the bees’ ability to cool the nest during high temperature conditions, temperature probes (15 length × 5 diameter mm; SENSIRION, model SHT75; precision: 0.3 °C) were introduced into the hives through lateral holes (15-mm diameter). In each colony, the temperature of the brood area (T BROOD) was monitored (Fig. S1A). The data were collected and stored at intervals of 5 min and 24 h a day. Five colonies were monitored. Another empty wooden box with the same characteristics as the one used for keeping the colonies was monitored by a temperature probe (held by a piece of wood that prevented contact with the floor of the box) and placed in the same conditions as the experimental hives. This box, lacking bees but under the same environmental conditions, served as a control for the conditions the bee colonies were submitted to (T CONTROL). The ambient temperature was also recorded using a sensor placed in the shade, close to the hives (T AMB). All boxes were exposed to the sunlight, which produced high temperature conditions (>40 °C in the T CONTROL during the hottest time of the day), and conditions were similar in all hives. The data used in the analysis comprised 3 days in which the temperature conditions were hot enough.

  2. b)

    Water collection behaviour

    To test whether water collection behaviour is performed when the temperature increases, we performed further observations on two of the colonies mentioned above. The colonies had a transparent plastic tube attached to the entrance (15-cm length × 2.5-cm diameter), which had a small window (3 × 2 cm) that allowed the capture of foragers as they returned to the nest (Fig. S1C). Every hour, (from 08:00 to 17:00 hours) three foragers with neither pollen nor resin in their corbiculae were collected from each colony. These foragers were anaesthetized with CO2, and by pressing their abdomens, their crop content was collected with a graduated capillary tube (Fig. S1D). The sugar content of their crop content was then analysed using a digital refractometer (Krüss Optronic—Alemanha—DR201-95). Only the bees that contained at least 7 μL of liquid were analysed. This is the minimum sample volume required by our refractometer for accurate measurement of the solution’s concentration. Concentrations with less than 2 % of sugar were considered to be “water” (Leonhardt et al. 2007).

  3. c)

    Fanning behaviour

    To determine whether wing fanning behaviour occurs as temperature increases, we monitored three colonies as described above. A transparent plastic tube was attached to the entrance (12 cm length × 2.5 cm diameter) to extend the entrance tube (where the workers position themselves to perform fanning behaviours) and to allow counting of individual workers fanning their wings. This behaviour is very characteristic, with the workers standing on the substrate (in this case, the plastic tube), forming a queue, with the posterior legs stretched, and fanning their wings (Fig. S1B and S1E).

    Six bioassays were conducted in which the temperature of the boxes was gradually increased, using one 15 Watts incandescent lamp positioned under the hives (2 cm below the box). After 45 min, the lamps were turned off. After that, the number of workers fanning in the plastic tube was recorded every 5 min, until there were no more bees fanning. The bioassays were performed from 11:00 to 15:00 hours. Temperature data were sampled as described earlier.

  4. d)

    Effect of temperature on pupae mortality and development time

    The bioassay consisted of incubating brood combs containing pupae at five different temperatures: 22, 26, 30, 34 and 38 °C (BOD incubator). Towards the end of normal pupal developmental time, we observed the brood combs daily to check for emergence of adults. We noted individual mortality (pupae were considered dead if they failed to emerge or other characteristics of death, such as fungal growth, were observed) and, for the surviving pupae, the time until spontaneous emergence of the bees. We used 50 pupae from five different colonies in each treatment.

    To ascertain the age of the pupae, we uncapped 60 brood cells that contained larvae in pre-pupae stages, which were adjacent to the pupae. As in this species, the brood comb is built as a disc, with new cells being added from the centre to the margins, and the colonies produce more than 200 cells per day, almost all the pre-pupae which neighbour pupae pupated the next day. This day was considered day zero (Fig. S2). As we opened 60 cells, we removed the necessary amount of individuals to reach 50 female pupae in each comb. The removal was performed when the pupae reached the stage of black eyes, which allowed us to remove male pupae. The brood combs were placed inside a petri dish containing saturated solution of water and salt (NaCl) to maintain the humidity close to 75 %, which has previously been showed to be the ideal condition for in vitro queen rearing on this species (Menezes et al. 2013).

2.3 Data analysis

All analyses were performed in R 2.15.0 (R Development Core Team 2011). Cooling of the nest was considered to occur if the brood temperature (T BROOD) was lower than the control temperature (T CONTROL). The temperatures in this category (T BROOD < T CONTROL) were grouped into periods (cooling periods), during which the cooling happened. We described the temperature differences during this period and measured the duration of the cooling period. The probability that a forager bee was loaded with nectar and water (0 and 1, respectively) was modelled as a function of the nest temperature using a GLM with a binomial distribution, since the response variable followed a binomial distribution. We performed diagnostic analyses of the model, and found that the residual deviance did not indicate overdispersion. To test whether wing fanning was triggered and increased as the temperature inside the nest increased, we performed a linear regression with the number of fanning bees as a function of brood temperature and colony (Zuur 2009). All tests were two-tailed tests with a significance level set to α = 0.05. We used an ANOVA to test for differences in mortality of pupae reared at different temperatures.. The relationship between temperature and development time was analysed using a Kruskal-Wallis one-way ANOVA on Ranks test (Zuur 2009).

3 Results

  1. a)

    Nest cooling

    “Cooling periods” (when T BROOD was lower than T CONTROL) lasted about 5 h (306.6 ± 83.9 min), from approximately 10:00–14:00 hours. This was the time of highest environmental temperature (Figure 1).

    Figure 1.
    figure 1

    Temperature of the brood area (T BROOD—open dots showing mean and bars showing standard deviation) from the five colonies monitored, the control box (T CONTROL thick line), and ambient (TAMB thin line) during three non-consecutive days. The cooling periods (when T CONTROL is higher than average T BROOD) are highlighted (grey areas).

    Full size image

    These results suggest that there is a mechanism for cooling the nest during warm periods, while during the rest of the day, the brood was maintained at a higher temperature than the control (Figure 1; Table S1). The peaks of maximum differences between T CONTROL and T BROOD reached 1.6(±0.2)°C (averaged over all the colonies studied and all sampling days). The average maximum cooling achieved by the colonies was 2.5 (±0.28)°C for only one of the days sampled. The T AMB was considerably lower than the T CONTROL during the cooling periods (Figure 1).

  2. b)

    Water collection

    About 20 % of the foragers (47 of 231) in both colonies collected water (Table S2). Water collection started when the T BROOD was about 29 °C. The binomial GLM analyses were significant and showed that the probability of a forager collecting water increased significantly with the T BROOD (slope = 0.55 ± 0.09, P < 0.001, Figure 2).

    Figure 2.
    figure 2

    The relationship between water collection and brood temperature. The fitted values (solid line) represent the probability of a forager collecting water relative to the brood temperature (T BROOD) obtained by the binomial GLM analysis (slope = 0.55 ± 0.09, P < 0.001). The dots are the observed values (1 represent water collection).

    Full size image
  3. c)

    Fanning behaviour

    The number of fanning bees was positively correlated with the brood temperature (adjusted R 2 = 0.48, F = 158.8, P < 0.001): the higher the temperature, the more bees were observed fanning (Figure 3).

    Figure 3.
    figure 3

    Scatterplot showing the effect of brood temperature on the number of fanning bees. The line represents the best-fit line and was drawn based on the parameter estimates obtained from the linear regressions (adjusted R 2 = 0.48, F = 158.8, P < 0.001).

    Full size image
  4. d)

    Effect of temperature on pupae mortality and development time

    Mortality rates were lowest at 34 °C (1.2 %); however, there was no significant difference in mortality between 26, 30 and 34 °C. Both extreme temperatures, 22 and 38 °C, showed high levels of mortality, 100 % and 96.6 %, respectively (Figure 4; Table S4) (one-way ANOVA: F = 519.43; P < 0.001–Holm-Sidak: P < 0.05).

    Figure 4.
    figure 4

    Survival rate (bars) and development time (boxplots) of S. depilis pupae incubated at five different temperature. Bars, grey sections represent the percentage of survival, while the black sections represent percentage mortality. Boxplots, median (thick line), upper and lower quartiles (upper and lower limits of the boxes), 95 % of data distribution (stems) and the values out of the 95 % data distribution (ouliers) are represented. Different superscribed letters indicate groups statistically different (ANOVA and Kruskal-Wallis associate with Dunn’s test, P < 0.001).

    Full size image

    Development time showed significant variation among treatments (H = 675.1; P < 0.001). Non-significantly different values of development time were found only between 34 and 38 °C (Figure 4; Table S4). There was a strong negative correlation between incubation temperature and development time (spearman rank correlation = −0.95). The mean time of development halved from 26 to 34 °C.

4 Discussion

Our results show that S. depilis brood nest temperature is colder than the empty control box during the hottest part of the day. Although the difference between T BROOD and T CONTROL might seem to show only moderate cooling, it has been suggested that even small increases in temperature during very hot periods can be highly damaging to brood (Undurraga and Stephen 1980; Mardan and Kevan 2002; Tautz et al. 2003), as our data also showed for extreme temperatures (“Results” (d)). This damage is thought to mostly occur due to disruption of cellular physiological processes. Bees and most other animals seem to be much more robust to cold temperatures (Angilletta 2009).

Although the data we collected indicate that active mechanisms of fanning and water collection could be related to nest temperature decrease, we cannot rule out the possibility that nest structures and simple presence of the bees are not influencing the nest temperature oscillations, since they were not present in the empty control box. The brood, adult bees, wax structures, honey and pollen pots could result in thermal inertia, decreasing the rate of temperature rise in nests. However, both the adult workers and the brood produce metabolic heat (Roubik and Peralta 1983; Jones and Oldroyd 2007), which could have the opposite effect, increasing the nest temperature and make active cooling more difficult. Therefore, while further investigations are necessary to fully understand the mechanisms that decrease nest temperature in stingless bees, our data strongly suggest that active behavioural efforts play a role in nest thermoregulation. It is noteworthy that T AMB is lower than T CONTROL during the hottest period of the day, highlighting the necessity of a control box when studying the thermoregulation of social insects that nest in cavities. Such a control is absent in all other studies on nest cooling by social insects.

Our results suggest that S. depilis exhibit at least two active mechanisms which may play a role in the cooling of the nest: (i) fanning behaviour and (ii) water collection. Both behaviours are present in the behavioural repertoire of the response of honey bees to high temperatures described by Lindauer (1954).

Water collection for cooling purposes has up until now only been described in the Apini tribe and in several species of social wasps (Jones and Oldroyd 2007). Here, we provide the first evidence of water collection for the purpose of nest cooling in stingless bees. In Apini and wasps, the mechanism of cooling seems to be similar: the water is deposited in small droplets on the surfaces of the nest, or the surface is licked, and as the water evaporates, it lowers the temperature of the nest (Lindauer 1954; Simpson 1961; Nicolson 2009). Despite previous observations of water collection in Meliponini, this has not been linked to the cooling of the nest, but to nutritional needs (hydration and minerals). Such a link was probably not previously made because the deposition of water droplets or licking behaviour has not been observed before (Jones and Oldroyd 2007). As in honeybees, our results show that high ambient temperatures are linked to water collection (Lindauer 1954), but the mechanism underlying the process of cooling in stingless bees is still unknown. It is likely that the water is deposited on internal nest surfaces or evaporated through licking by the workers, but this awaits investigation. There is the possibility of the adults using this water to decrease their own body temperature, as has been described in honeybees (Heinrich 1979; Cooper et al. 1985). This would nonetheless be a way of decreasing the nest temperature. Our results suggest a causal link between nest temperature and water collection behaviour, rather than a coincidental occurrence of water collection at the time of highest ambient temperature, because stingless bee foraging activity for other resources (nectar, pollen or resin) normally decreases during extremely high temperatures (Silva et al. 2011; Hilário et al. 2012; Figueiredo-Mecca et al. 2013). The same phenomenon was observed in large bees, such as Bombus terrestris (Kwon and Saeed 2003). Water foragers may be able to tolerate foraging at higher temperatures by using evaporative cooling from their bodies (Prange 1996; Pereboom and Biesmeijer 2003).

We also confirm previous suggestions (Engels et al. 1995; Nogueira-Neto 1997; Jones and Oldroyd 2007) that the wing fanning behaviour is indeed a colony-level response to an increase in nest temperature and, therefore, is likely to be performed to cool the nest. The higher the temperature inside the nest, the more bees start to fan their wings in the entrance tube. We also observed that this behaviour occurs inside the nest, in regions not connected to the entrance. It is important to note that this behaviour is also a response to high levels of CO2 inside the nest in other social bees, and is used to improve gas exchange in the nest (Simpson 1961; Moritz and Crewe 1988; Weidenmuller et al. 2002). Indeed, high CO2 levels could also be the ultimate cause of the ventilation, since increased metabolic rates during high temperature periods could increase CO2 levels. This variable should be included in further studies to conclusively determine whether fanning is a response to overheating per se, or to an increase in CO2 concentrations caused by overheating. Lastly, another possibility is that the fanning behaviour could have a role of decreasing humidity inside the nest by facilitating the air exchange between the brood area and the outside. Such an integration between ventilation and evaporative cooling would be particularly important when the ambient air temperature exceeds the upper critical temperature. When this limit is reached, non-evaporative heat loss is progressively reduced because the air temperature is higher than the temperature inside the nest, causing simple ventilation itself to be ineffective. This hypothesis also remains to be tested.

The effects of temperature on brood development time varied greatly between intermediate temperature treatments (26, 30 and 34 °C; Figure 4). However, while we found a small positive correlation between an increase in temperature and pupae survival, this correlation was not significant. However, the extreme temperatures (22 and 38 °C) increased brood mortality (Figure 4). Our data showed that the T BROOD was on average 30.5 ± 2.9 °C (mean ± s.d.) over the whole day and 34 ± 1.4 °C (mean ± s.d.) during the cooling periods. These temperatures are most similar to the two temperature treatments which showed the fastest brood development and highest survival (30 and 34 °C). Extreme temperatures were found to be very damaging to the brood, highlighting the importance of temperature control. A change of even 1 °C could make the difference between survival or death. However, it is important to note that these temperatures were held constant in our development bioassay, but temperatures are not constant in nature. The impact of short episodes of extremes temperatures on brood development would benefit from investigation. Even 1 h of exposure to temperatures of 50 °C caused a mortality of 100 % in pupae and pre-pupae of the solitary bee Megachile rotundata (Hymenoptera, Megachilidae), while an exposure of several hours to 45 °C did not significantly increased mortality (Undurraga and Stephen 1980). The development time of Osmia bicornis (Hymenoptera, Megachilidae) brood was also decreased when the temperatures fluctuated, as compared to conditions of constant temperatures based on the average of the fluctuations (Radmacher and Strohm 2011).

In honeybees, a variation of 5 to 6 °C from the 33 °C optimal temperature for brood development greatly increases brood mortality and malformations (Himmer 1927; Mardan and Kevan 2002), while in S. depilis, mortality does not change over a range of at least 8 °C. However, the development of S. depilis is considerably slower than that of A. mellifera. The minimal development time from pupae to emergence of S. depilis is around 15.7 and 16.2 days, at 38 and 34 °C, respectively. In honey bees, the closed cell phase (prepupal and pupal) development time is around 9.5 to 12 days (Michener 1974). We suggest that the precise temperature control achieved by honeybees when compared to S. depilis has led to adaptation to a narrow range of temperatures. The stingless bees species which are poorer thermoregulators would be better adapted to a wider range of temperatures, leading to slower development, but one that is more robust to temperature fluctuations (Angilletta 2009). This still remains to be thoroughly tested, especially considering the difference in individual sizes and weight between S. depilis and A. mellifera. Stingless bee species are good candidates to test this hypothesis, since many different thermoregulatory strategies are present among closely related species.

Our results suggest that S. depilis have adaptations similar to honeybees for coping with high temperatures. This likely represents an incident of convergent evolution, since these adaptations depends on a social context, and the origin of eusociality in both groups is considered to have occurred independently (Michener 1974). Stingless bees do not thermoregulate their nest as precisely as honeybees do. However, our results suggest that S. depilis brood is more resistant to temperature variation during their development. We suggested that stingless bees’ strategy would be to increase the thermal resistance—coupled with proper nest site selection. In other words, increasing capacity to tolerate higher temperature variation, expanding the range of thermal tolerance by sacrificing higher performances at optimal temperatures, results in longer development time (Huey and Hertz 1984; Angilletta 2009).