Food restriction delays seasonal sexual maturation but does not increase torpor use in male bats

Living in seasonal, often unpredictable environments is challenging for many animals. The optimization of assimilated energy allocation among competing life-history traits is central to maximize fitness (Arlettaz et al., 2001; Hou et al., 2008). Energy allocation strategies may thus involve trade-offs affected by numerous factors such as environmental conditions (food predictability and availability), sex-specific life history, reproductive status and energy-saving mechanisms (Perrin and Sibly, 1993). Small endothermic animals have especially high energy requirements (Speakman and Thomas, 2003). Indeed, they have to generate significant amounts of endogenous heat to compensate for large heat losses, due to their large surface area to volume ratio (Aschoff, 1981). Moreover, reproduction increases their energy demands (Gittleman and Thomson, 1988). This increase in energy expenditure has to be balanced by physiological adjustments and an increase in foraging effort (Speakman, 2008; Rödel et al., 2016). Thus, trade-offs between costly reproductive processes and body maintenance or thermoregulation are particularly challenging for small animals (Grinevitch et al., 1995; Willis et al., 2006). Resource limitation may further exacerbate competition between life-history traits (Boggs and Ross, 1993; Cotter et al., 2011; Chinn et al., 2018).

Even among small mammals, energy demands of bats are exceptionally high (Speakman and Thomas, 2003). Beside their small body size, their wings represent large naked surfaces prone to heat and water loss, which increases thermoregulatory costs (Speakman and Thomas, 2003; Dzal and Brigham, 2013). Moreover, flight as their only mode of locomotion is extremely energetically expensive (Thomas and Suthers, 1972). For insectivorous species, adverse weather conditions such as cold ambient temperature or precipitation, strongly decrease prey activity and can affect bats' ability to fly and detect prey (Griffin, 1971; Voigt et al., 2011; Ruczyński et al., 2017). Even excess fat which increases weight is disadvantageous during flight (Norberg and Rayner, 1987).

In many small endotherms, torpor (i.e. the controlled decrease of metabolic rate) is a crucial physiological adaptation to save energy (Geiser, 2004; Bozinovic et al., 2007) and, under some conditions, to enable fat deposition (Geiser and Brigham, 2012). Food availability, body condition, season and/or weather conditions are thought to be the main determinants of torpor use in bats (Wojciechowski et al., 2007; Matheson et al., 2010; Dzal and Brigham, 2013; Becker et al., 2013). However, torpor use also incurs costs, such as reduced immunocompetence (Humphries et al., 2003). Furthermore, it has been suggested that torpor can slow down fetal development and milk production (Racey and Swift, 1981; Wilde et al., 1999; McLean and Speakman, 1999). As in rodents (Fietz et al., 2004), torpor may also reduce rates of testicular recrudescence and sperm production in bats (Entwistle et al., 1998). Consequently, torpor expression can be influenced by a trade-off between reproduction and saving energy and should be optimized depending on current body conditions and the reproductive state of individuals.

Many temperate zone bat species are promiscuous, and females only bear one or two offspring per year (Campbell, 2008). Furthermore, among the family Vespertilionidae, females can store sperm for several months (Racey, 1979). Sperm competition is strong, and male reproductive traits are under selection (Orr and Zuk, 2013; Fasel et al., 2018), putatively favouring increased sperm production. Sperm production in temperate zone bats is seasonal, extending from spring to late summer. Spermatozoa are transferred from testicles to the cauda epididymides and stored until the mating period, from the end of summer until female ovulation in the following spring (e.g. Thomas et al., 1979, Gebhard, 1995, Furmankiewicz et al., 2013, Pfeiffer and Mayer, 2013). In the context of our study, we define male sexual maturity as the seasonal readiness to mate (Racey, 1974a; Entwistle et al., 1998). Both spermatogenesis and the resulting intensive growth in testicular mass by up to 8% of body mass are costly (Wilkinson and McCracken, 2003). During spermatogenesis, males remain normothermic in order to maintain a high rate of sperm production (Entwistle et al., 1998; Boyles et al., 2007) and increase the amount of sperm cells produced (Jolly and Blackshaw, 1987; McWilliam, 1988). Once spermatogenesis is terminated, torpor is probably beneficial for sperm storage (Racey, 1972; Geiser and Brigham, 2012; Sharifi and Javanbakht, 2016). However, not only body temperature but also food and thus body condition (Beasley and Zucker, 1986; Abedi-Lartey, 2016) may be important in the development of sexual maturity. Indeed, only individuals in a certain ‘target’ body condition initiate testes recrudescence and sperm production (Speakman and Racey, 1986; Young et al., 2000). Thus, sperm production is synchronized with periods of highest insect availability in several insectivorous bats (Taylor, 1963; Racey, 1982), enabling males to complete spermatogenesis quickly. Then, when insect availability declines, bats resume torpor use to accumulate fat in preparation for the mating period and hibernation (Encarnação, 2012; Geiser and Brigham, 2012).

While it is known that energy shortage affects fetus and/or pup survival (e.g. Chinn et al., 2016), the consequences of environmental limitation on energy resource allocation in male bats have not yet been studied experimentally. We manipulated food availability to determine how male parti-coloured bats, Vespertilio murinus, balance the costs of thermoregulation against sperm production. Vespertilio murinus is highly specialized on ephemeral and thus patchy insect swarms. Interestingly, it is one of the few temperate zone bat species that form male colonies in summer (Safi and Kerth, 2007). As these male aggregations overlap with the period of spermatogenesis, it has been postulated that males may forage in groups and use passive information transfer to find insect swarms more efficiently (Safi and Kerth, 2007; Dechmann et al., 2009; Dechmann et al., 2010) and/or benefit from social thermoregulation to limit torpor use (Safi, 2008).

Here, we investigated the influence of food availability on torpor use and the resulting effect on sexual maturation (i.e. testicular growth and subsequent epididymal sperm storage). We tested the following predictions: (i) torpor use, as an energy-saving strategy, should be more frequent in food-restricted individuals; (ii) torpor use should be lower during spermatogenesis and should increase when spermatogenesis is completed and sperm are stored; and (iii) frequent torpor use during spermatogenesis should slow down sexual maturation.

The experiment was carried out at the Mammal Research Institute of the Polish Academy of Sciences in Białowieża (NE Poland). We captured 36 experimental adult males from a colony of more than 100 male Vespertilio murinus (Linnaeus 1758) in mid-June 2016 with a harp trap (Austbat, Bat Conservation and Management) and a custom-built funnel trap as they emerged from the roost in the Bieszczady region (SE Poland) in the evening. Experimental individuals were chosen opportunistically. Body mass at capture ranged from 9.9 to 13.3 g (mean 12.0 g) and forearm length ranged from 41.53 to 46.27 mm (mean 44.32 mm). During an initial 7 day acclimation period, we trained animals to feed independently, which they all learned within 3–4 days. The diet (maintained during the experiment) consisted of mealworms (larvae of Tenebrio molitor). Vitamins (solBiosupervit, Biofaktor, Skierniewice, Poland) were added to the water every fourth day (Boratyński et al., 2015). During the acclimation period, we allowed bats to freely choose group size and composition during the day in six wooden roosts that mimic natural tree cavities (Ruczyński et al., 2007). The same roosts were used for housing bats during the experiment. Throughout the captive period, we allowed all bats to fly freely in an isolated flight room (5.3 m×6.9 m×3.4 m) every evening. Light and temperature in the holding room followed outside conditions. We recorded temperatures inside each roost every 10 min using iButtons (DS1922L-F5, Dallas Semiconductor). We obtained local outside temperatures measured at 2 m above ground from the local weather station in Białowieża. After the experiment was terminated, we released bats at the site of capture.

Our experiment started in mid-June when V. murinus males form colonies and lasted until the beginning of September. The captivity period coincided with sperm production and maturation. As males in eastern Poland stop being social in July (Hałat et al., 2018), we covered periods of both social and solitary lifestyles. In the period of June–July there is a noted high insect abundance in eastern Poland (Ruczyński et al., 2019). At the beginning of the experiment, we assigned bats to two treatment groups of equal size each containing the same distribution of body conditions (ratio of forearm length to body mass on the first day). We exposed bats to two feeding regimes (hereafter: ‘high food’ versus ‘low food’). In the high food treatment, each bat received 5 g of mealworms daily and in the low food treatment, each received 1.5 g. For feeding, we placed bats in individual boxes. To determine the amount of food given in each regime, we carried out a pilot study to determine the amount of food bats required to maintain mass. The pilot study was carried out 1 year before the experiment, from the end of June till the beginning of August. The resulting amount was provided to the low food bats. The high food treatment was determined based on the amount of mealworms spontaneously eaten by unrestricted bats daily during the first 2 weeks of captivity. We recorded intensive body mass gain in the high food bats over time, probably because our bats were less active than in nature. The average difference between initial and final body mass of high food bats was 9.7 g, but some individuals even doubled their initial body mass. Therefore, we gradually decreased the amount of food given to them after the second half of August until 1 September, when the experiment was completed.

From the beginning of the experiment until their release, bats were kept in six groups of six individuals (three bats from each treatment in each group). We kept individuals from the two treatments together, to expose them to the exact same ambient conditions, resulting in food availability being the only different environmental parameter. During the experiment, bats were restricted to a roosting cavity except during the night time when each group was allowed to fly in the flight room.

Because of injury, four individuals (one from the high food treatment and three from the low food treatment) were excluded from the final analysis.

We monitored body mass (M_b) every evening with an electronic balance (±0.1 g; Pesola PTS3000 General Electronic Scale) after bats had flown and before feeding. We used body mass gain (i.e. the day to day difference in M_b of each individual) as the measure for mass. We are aware that fat is not the only component of body composition that affects mass. We could not measure real fat stores with non-invasive methods [e.g. total body electrical conductivity (Reynolds et al., 2009) or quantitative magnetic resonance imaging (Guglielmo et al., 2011)], so we made the assumption that M_b increase reflects adipose tissue accumulation, even if in an unknown ratio.

We recorded skin temperature as a proxy for body temperature every 10 min with iButtons miniaturized according to Lovegrove (2009). Although skin temperature is probably lower than core body temperature, they are highly correlated in small endotherms particularly during torpor (e.g. Willis and Brigham, 2003). We trimmed the fur on the back of the bats and then glued iButtons directly onto their skin with tissue adhesive (Sauer Hautkleber, Manfred Sauer, Germany). At the beginning of the experiment, we fitted six individuals from each treatment (i.e. 12 bats) with iButtons. The next set of 12 bats received iButtons 2–3 weeks after the first fitting, even if some animals from the previous batch were still equipped with loggers. The original iButtons were not removed, but left to detach spontaneously (after 2 weeks to 2 months), resulting in longer measurements from a subset of bats. With this approach at least one-third of all experimental animals were always equipped with iButtons. A bat only received a new iButton after the previous logger had been detached for at least 2 weeks or longer. The iButtons weighed approximately 1.6 g (5–13% of individuals' M_b) and did not appear to disturb the animals or impair their ability to fly.

We excluded time periods when bats were flying and being fed from the HI analyses. We also excluded 14 days when other manipulations were done (e.g. metabolic rate measurements, data not shown here). We calculated the HI for every 24 h period (HI₁) and for 5 day time slots (HI₅). We used HI₁ for analyses concerning factors affecting the thermoregulatory patterns of the bats. We used HI₅ when analysing sperm production rate to account for day-to-day variability in torpor usage.

We described sexual maturation stages based on testicular growth and epididymal filling, assessed through external visual examination of the testicles and caudae epididymides. Development stages were categorized into five classes (Safi, 2006), from 0 (no to small testicles, situated at the base of the penis, with no epididymis visible), through 2 (testicles large, situated below the anus; epididymes clearly visible as dark patches, but flat) to 4 (with fully filled epididymes; testicles regressed and not visible); classes 1 and 3 were intermediate between these (Fig. S1). A single observer made all assessments (E.K.) every 2–3 weeks. Partial filling of the epididymes indicates that spermatogenesis is highly progressed, and spermatozoa are being moved from the testicles to the epididymes for maturation (Racey, 1974a; Krutzsch, 2000). We assumed that the time it takes a bat to move to the next class indicates sperm production rate (sexual maturation) and modelled it for each individual (see below).

We used general additive mixed models (GAMM) to determine factors affecting body mass gain and HI_1. The best models for both body mass gain and HI₁ were determined based on Akaike's information criterion (AIC; model selections in Tables S1 and S3). For M_b gain analysis, we included four explanatory variables: forearm length, time (expressed as consecutive days of the experiment) and mean daily roost temperature recorded the previous day (T_roost−1) – all three are continuous variables – and treatment (factor with two levels: low food and high food). We suspected that M_b would change non-linearly over time and ambient temperature according to the bats' reproductive status, environmental conditions and treatment. Thus, we fitted time and T_roost−1 with thin plate regression splines separately for the two treatments. We used T_roost−1 for this analysis, as we assumed that conditions from the previous day should have a greater influence on current M_b than current conditions.

The GAMM investigating HI₁ considered forearm length, mean daily roost temperature recorded on a given day (T_roost), treatment, and time grouped by treatment, as explanatory variables. Moreover, we also considered M_b as a fifth explanatory variable (scaled within treatment to disentangle obvious dependence between treatment and M_b; using the ‘scale’ function in R). The time, temperature and scaled M_b were fitted using splines. We included the random effects of bat individual (ID) and group in both models. We also set the upper limit of degrees of freedom for T_roost−1, T_roost and M_b to 5, to keep the fit relatively simple, while the upper limit of degrees of freedom for time was 10, to account for potentially larger day-to-day variability of the response variables. All models were performed in the ‘mgcv’ package (Wood, 2006) in R (version 3.2.5, http://www.R-project.org/). We assumed that the differences in the effect of each factor between treatments on M_b gain and HI were significant when confidence intervals (CI) of treatments did not overlap. We did not use body condition index (BCI, i.e. ratio of body mass and forearm length) as it does not accurately reflect body condition in bats according to McGuire et al. (2018). However, we decided to include forearm length in models together with M_b as a constant measure of body size for each individual. This approach allowed us to control for changes in M_b.

We estimated the effects of targeted explanatory variables (i.e. food treatment, HI₅ and M_b corrected by forearm length) on sexual maturation separately with three hierarchical ordinal logistic models using Bayesian inference (Jackman, 2009). In detail, the slope coefficients for a latent sexual maturation variable were estimated using a random intercept model. The time since the beginning of the experiment was considered as a fixed effect and bat identity as a random factor. We also estimated the transition time points, when animals switched from one sexual maturation class to the next. In addition, we considered effects of the explanatory variables on transition time points. We made the a priori assumption that there was no difference in the time of transition from class 0 to 1 between bats from different treatments, and that this first transition point was not affected by any of the explanatory variables. The first transition time point was thus fixed at 15 days for all individuals. This value is biologically sound and led to more adequate chain convergences than larger or smaller values (data not presented). In order to ease the convergence of chains, time in days since the beginning of the experiment was divided by 365. Three different Markov chains, starting at random initial values in the range of parameter space, were run during 100,000 iterations and the initial convergence phase was excluded by dropping the first 70,000 iterations. We thinned Markov chains with a factor of three and used the Brooks–Gelman–Rubin criterion (Brooks and Gelman, 1998) to assess the convergence of chains. We specified vague prior distributions for target variables; namely, a normal distribution with a mean of 0 and precision of 0.001 for the slope and the explanatory variables effects, and a uniform distribution [0,5] for the standard deviation of the normal distribution centred on 0 for the individual random effects. The prior distributions for the transition time points were uniform [0.041,100]. The Bayesian analyses were run using the function ‘jags’ (package: ‘jagsUI’; https://CRAN.R-project.org/package=jagsUI). Effects were considered significant when their 95% CI did not cover 0.

All experimental procedures were authorized by the General Director for Environmental Protection (authorization no. DZP-Wg.6401.09.2.2014, DZP-Wg.6401.09.1.2015, DZPWg.6401.09.5.2016) and by the Local Ethical Commissions in Białystok and Olsztyn (authorization no. 11/2014, 14/2015, 120/2015, 150/2015, 15/2015, 45/2015).

Temperature in the holding room and T_roost were highly correlated with outside temperature in Białowieża (r=0.81, P<0.001). Mean daytime T_roost decreased over the duration of the experiment. From mid-June until the end of July, T_roost was high (mean 24°C), but varied from 17 to 30°C along with outside temperature. From August until the beginning of September, mean T_roost was lower but more stable (mean 20°C, range 19–26°C; Fig. 1A).

We recorded M_b gain over time in both treatments. In the high food group, the mean difference between initial and final M_b was greater than that in the low food group. In the low food group, the mean M_b of bats increased from 11.8 g (range 9.9–13.3 g) to 12.7 g (range 10.1–14.9 g), while in the high food group it increased from 12.1 g (range 9.6–13.3 g) to 21.9 g (range 16.6–24.8 g) (Fig. 1B). M_b gain in both treatments was significantly positively affected by time (low food, P=0.02, F=3.31; high food, P<0.001, F=9.96; Fig. 2A; Table S2). Specifically, from the beginning of the experiment (mid-June) until day 15 (end of June), M_b gain in both treatments was systematically negative (Fig. 2A). From the beginning of July, the mean M_b gain in both treatments stayed around zero. High food bats started gaining mass after 2 weeks (first half of July), while low food bats did not gain mass until the end of July. From the end of July until the end of August, M_b gain in low food bats was constantly increasing. After mid-July, M_b gain was significantly higher in the high food than in the low food bats most of the time. However, we observed fluctuations in M_b gain in high food bats throughout the study. At the beginning of August (ca. day 40), M_b gain decreased to 0. There was also a large M_b loss at the end of August when we decreased the amount of food given (Fig. 2A).

The effect of T_roost−1 was significant only for high food bats (low food, P=0.092, F=2.11; high food, P=0.002, F=4.71; Table 1). The mass gain increased together with the increase of T_roost−1, but only up to 25°C, when a drop was recorded (Fig. 2B).

There was no significant effect of forearm length on M_b gain (P=0.99, t=−0.007; Table S2).

Torpor use, expressed as HI₁, was highly correlated with both daily mean and daily minimum T_sk (respectively: r=−0.97, P<0.001 and r=−0.91, P<0.001). Except for a short period in the second half of July, there was no significant difference in HI₁ between treatment groups. In both treatments, HI₁ increased with time (low food, P<0.001, F=16.368; high food, P<0.001, F=20.724; Table S4). The average heterothermy patterns in the two treatments were similar (Figs 1C and 3A). Torpor use was variable but low during the first part of the experiment, which coincided with low sexual maturation classes (0–2) in both treatments. From the beginning of the experiment (mid-June) until mid-July, the mean HI₁ in high food bats was 4.4°C (range 0.9–16.9°C) and in low food bats it was 4.3°C (range 0.7–15.3°C). In the second half of July, HI₁ remained the same in high food bats (mean 4.8°C, range 0.9–15.9°C), while in low food bats the mean increased to 6.8°C and the difference between treatments was significant. However, the range did not change (0.6–15.6°C). HI₁, and therefore torpor use, increased at the end of July/early August coinciding with an increase in M_b of the high food bats. After the beginning of August, coinciding with reaching high sexual maturation (classes 3 and 4) in high food individuals, HI₁ increased to 11.9°C in the high food bats (range 7.8–15.3°C) and to 10.0°C in the low food bats (range 6.5–17.4°C). We also found a significant increase of HI₁ along with increasing scaled M_b as the experiment progressed (P=0.002, F=5.514; Fig. 3C; Table S4). In addition, HI₁ was higher at low T_roost (P<0.001, F=32.942; Fig. 3B). In both treatments, HI₁ was strongly positively affected by time (i.e. torpor use increased with time; low food, P<0.001, F=16.368; high food, P<0.001, F=20.724; Table S4).

We did not observe a significant time difference in switching from sexual maturation class 1 to 2 between treatments (Fig. 4, Table 1). Bats reached class 2 on average on day 36 (second half of July). High food bats shifted from class 2 to 3 around day 50 (beginning of August) and from class 3 to 4 on day 71 (end of August). High food bats switched to classes 3 and 4 significantly faster than low food bats, with low food bats showing a delay of 11 and 9 days, respectively, for the two maturation classes (Fig. 4, Table 1). M_b was negatively correlated with the transition time points to the two latest classes but did not affect the transition time point from class 1 to 2 (Table 1). Forearm length, as well as HI₅, did not significantly impact sexual maturation (Table 1).

In nature, unfavourable environmental conditions can reduce food availability. We investigated how such food restriction affects thermoregulation, sexual maturation and M_b in males of an insectivorous bat species from the temperate zone.

We found that changes in M_b differed dramatically between treatments. High food bats had much higher M_b increase over the course of the study (Fig. 1B). However, in both treatments, we initially observed no mass gain. This M_b maintenance may be due to nutrient requirements and the necessity to defend normothermia during testicular growth and spermatogenesis (Encarnação and Dietz, 2006; Takahashi and Parris, 2009). Similarly, there were no M_b changes in male Daubentons' bats during intensive spermatogenesis despite a more than 2-fold increase in food intake (Encarnação and Dietz, 2006). Indeed, small mammals can use up reserves and subsequently lose mass to balance the costs of thermoregulation and spermatogenesis if they cannot increase food intake (Entwistle et al., 1998; Dietz and Hörig, 2011; Becker et al., 2013; de Bruin et al., 2018). In our study, we did not observe M_b loss in the low food bats. However, our food restriction was moderate for ethical reasons. The decrease in M_b gain observed in high food bats in mid-August and at the end of August was due to the reduction in the amount of food we gave them to reduce obesity.

Contrary to our expectations, we did not find substantial differences in torpor patterns between feeding regimes, despite the growing asymmetry in M_b and difference in food received. Our results indicate that torpor use in V. murinus males is driven by their life history stage (spermatogenesis) more than by current ambient conditions such as food availability or temperature. Low torpor use in both treatments in the period from mid-June until the beginning of August might thus be dependent on sex hormone levels. According to observations of Hałat et al. (2018) in wild-ranging V. murinus males, spermatogenesis was ongoing at this time of the year. Plasma testosterone, which peaks during this period, may act to maintain metabolism and body temperature at normothermic levels during testicle growth (e.g. Racey, 1974b; Entwistle et al., 1998; Martin and Bernard, 2000; Dietz and Hörig, 2011; Becker et al., 2013). However, we did find low levels of torpor use in both treatments during testicle growth. The rare cases of torpor use during this time period in both treatments did occur at low ambient temperatures. This observation is consistent with previous work showing that male bats and rodents may employ short torpor bouts during periods of cold weather in order to save energy, even during spermatogenesis (e.g. Grinevitch et al., 1995; Lovegrove and Raman, 1998). However, in agreement with our predictions, low food bats expressed significantly more torpor than high food bats during these cold periods. This was the only period of significant difference in torpor use between treatments. Wojciechowski et al. (2007) and Encarnação (2012) also showed that at low ambient temperature, bats with higher food availability use less torpor than those with limited food supplies. Low torpor use between the end of June and the beginning of August, independent of food regimes, supports the idea that bats prioritize sperm production over the potential costs associated with thermoregulation at low ambient temperature or reduced food availability.

We showed the combined effect of M_b and torpor use on sexual maturation rate. The completion of spermatogenesis (i.e. reaching class 4) was delayed by about 9 days in low food bats. The lag in transition from class to class was driven by food availability and M_b but not torpor use. The impact this lack of torpor use had on sexual maturation rate is surprising in view of other studies. It has been shown that torpor use by male bats may lead to the reduction of testicular growth and maintenance, suppression of accessory gland activity and a reduction of spermatogenesis (Jolly and Blackshaw, 1987; Entwistle et al., 1998; Dietz and Kalko, 2006). Based on our observations, we suggest an indirect effect of thermoregulatory patterns on sexual maturation. We showed that food abundance affected M_b gain, sexual maturation and, partially, torpor use, while torpor had no effect on sexual maturation. Thus, we propose that in male bats, food resources are primarily allocated to thermoregulation, which seems to be driven by the circannual rhythm. The remaining energy resources are invested in maintaining M_b, with only surplus energy resources allocated to the production of sperm itself. Thus, sexual maturation rate depends on M_b, which is the result of the combined effect of food abundance and costs of thermoregulation. Bats may not be able to balance the high costs of thermoregulation and M_b maintenance while faced with insufficient food resources, which results in delayed sexual maturation. In conclusion, our results support the idea that males in better condition are prepared for the mating season earlier. Other potential costs of food restriction, such as lower sperm density, were not recorded but would be interesting to assess in future studies.

The effect of food limitation, such as the one we artificially induced here, may not apply to all bat species equally. The time it takes to complete sperm production differs among temperate zone bat species (e.g. Entwistle et al., 1998; Dietz and Hörig, 2011; Pfeiffer and Mayer, 2013). This difference might be related to seasonal behaviour of females, but could also be due to interspecific differences in foraging strategy. Aerial hawking species, such as V. murinus, are probably more vulnerable to the unpredictability of food resources than gleaning species. Thus, they may employ other mechanisms, such as alterations in behaviour, to save energy. Vespertilio murinus is one of a few temperate zone species forming male colonies. In nature, these bats stay together for just a few weeks and the colonial lifestyle overlaps with spermatogenesis (Safi, 2008). According to Safi (2008), males with higher progress in reproductive development leave the colony earlier. Thus, males may benefit from social thermoregulation and/or social foraging to buffer against variation in energy expenditure and resource availability during times of energy-costly spermatogenesis.

In summary, our results indicate that the effect of food availability on resource allocation in male bats may be more complex than expected, at least during sexual maturation. Males allocated energy resources primarily to maintaining normothermia. This strategy allowed them to maintain testicle development. However, the combined pressure of food availability, M_b and torpor use slowed down sexual maturation, causing a severe potential cost in the reproductive arms race.

We thank our students and volunteers for help: M. Kroeze, M. Walesiak and E. Farragiana. We are grateful to Z. Hałat for her support in the field, Michał Żmihorski for consultations on statistical analyses, the two reviewers for their valuable comments and Jenna Kohles, Sue Anne Zollinger and Eleanor Stockwell for correcting the English in the manuscript.