Locomotion on a slope in leaf-cutter ants: metabolic energy use, behavioural adaptations and the implications for route selection on hilly terrain

The route an animal takes determines the energetic cost of, and the time taken to perform, an activity and, consequently, is an important aspect of animal behaviour. It has been suggested that it is advantageous for animals to use routes that minimise either the time taken or the metabolic energy consumed (Minetti, 1995; Alexander, 2000). The shortest route between two points is a straight line and, on level ground, this is the route that will also incur the minimum time and energetic costs. However, the presence of obstacles, such as hills or soft ground, will change the mechanical demands of a route and consequently incur costs such as increased metabolic energy expenditure and journey time. Such costs may result in the optimal route deviating from a straight line (Minetti, 1995; Alexander, 1998; Alexander, 2000). In a route selection experiment using foraging wood ants, it was shown that route was primarily determined by energetic cost (estimated indirectly from vertical height traversed), although journey time was also a factor (Denny et al., 2001). In addition, paths established by human travellers in mountainous regions over long periods of time have been shown, retrospectively, to correspond to those predicted to minimise metabolic energy expenditure (Minetti, 1995; Alexander, 2000). However, there is little direct, empirical evidence to show that animals select routes that minimise costs. It may therefore be valuable to develop a biological model system in which the relationship between obstacles, such as a slope, and the metabolic cost of locomotion can be measured. Such a system can be used to assess whether animals do indeed use optimal routes, and whether there are circumstances where factors other than energetic cost or journey time (e.g. predation risk, exploration for resources or environmental conditions) favour a deviation in route from that predicted to be optimal.

The cost of locomotion (C_path), defined as the mass-specific metabolic energy expended to move a unit distance on a slope, varies with gradient. C_path increases when travelling up an incline compared with level locomotion in a wide range of species, including both vertebrates and invertebrates (Margaria, 1968; Wunder and Morrison, 1974; Armstrong et al., 1983; Full and Tullis, 1990; Minetti et al., 1994; Kram and Dawson, 1998; Wickler et al., 2000; Minetti et al., 2002; Rubenson et al., 2006; Tullis and Andrus, 2011). The cost of locomotion on a decline has been less well studied. Where data exist, C_path appears to be lower than that of level locomotion (Margaria, 1968; Armstrong et al., 1983; Minetti et al., 2002; Hoyt et al., 2006) up to a certain gradient beyond which C_path increases again (Margaria, 1968; Minetti et al., 2002). For example, in humans, C_path is minimal on a slope of approximately –5 deg (–10%) but increases on steeper declines and during level and incline locomotion (Margaria, 1968; Minetti et al., 1994; Minetti et al., 2002). However, some studies suggest that in small species, such as leaf-cutter ants, that the mass-specific metabolic energy consumed per unit distance travelled does not change with incline (Lipp et al., 2005).

The primary aim of this study was to determine the effects of slope on the gross metabolic cost of locomotion in leaf-cutter ants. This will clarify whether there is indeed no additional metabolic cost associated with incline running in this genre. The second aim of this study was to use this information to calculate the effects of slope on the cost of undertaking journeys along routes involving vertical ascent/descent (Minetti et al., 2002), and therefore determine the optimal routes in terms of minimising either gross metabolic energy expenditure or vertical journey time. These predictions will provide the basis for further behavioural studies to investigate whether animals ever use optimal routes, or whether factors other than metabolic energy expenditure or journey time additionally influence route choice. Ants provide an excellent system in which to develop a testable model of foraging strategies as they are central place foragers (Holldobler and Wilson, 1990) and can travel relatively long distances, over complex terrain, between the nest and food resources. Their foraging routes can easily be manipulated experimentally to test predictions about the determinants of foraging behaviour (Denny et al., 2001; Ribeiro et al., 2009).

A single colony of leaf-cutter ants [Acromyrmex octospinosus (Riech 1793)] was collected in Trinidad. Workers from this colony were used for all experiments (body mass 11.1±1.7 mg, mean ± s.d., N=38). The colony was housed in a glass vivarium, 90×40×38 cm (length×width×height), maintained at 28–30°C and fed brown rice, fruit or privet leaves every 2–3 days. Worker ants on an outbound foraging route from the nest to a food source were randomly selected for the experiments.

A modification of the ‘running tube’ technique (Lighton and Feener, 1989) was used to allow measurement of CO₂ production in ants travelling on a range of slopes. The ant was sealed in a respirometry chamber constructed from a Perspex tube (30 cm long, 1.5 cm diameter), with an air inlet and outlet (Fig. 1). The respirometry chamber was attached to a baton marked with a calibration scale, which could be rotated to vary the gradient. The bottom half of the tube was lined with paper to provide a suitable surface to enable the ants to walk on slopes as steep as ±90 deg. A small magnetic stir bar was placed in the tube and could be moved using a magnet external to the respirometry chamber to encourage the ant to start walking. All experiments were carried out at room temperature (21–23°C).

The rate of CO₂ production in ants travelling on a range of slopes was determined using a constant-pressure variable-volume closed-loop respirometry system (Ellington et al., 1990). Air was continuously extracted from the chamber at a rate of 200 ml min^–1, passed through a drying column to remove water vapour and run through a CO₂ analyser (FoxBox field gas analysis system, Sable Systems, NV, USA) before being returned to the chamber. Therefore, increases in CO₂ were cumulative over the period of locomotion. The drying column contained both Drierite (Acros Organics, Geel, Belgium) and magnesium perchlorate (Alfa Aesar, Ward Hill, MA, USA). New Drierite interacts with CO₂ (White et al., 2006); therefore, predominately exhausted and recharged Drierite was used to remove water vapour from the air, with just a small amount of new Drierite, which had been exposed to room air, to act as an indicator of desiccator exhaustion (White et al., 2006). A semi-inflated balloon attenuated any pressure differences that could otherwise have arisen in the system. The system was calibrated by injecting a known volume of CO₂ into it after each test (Ellington et al., 1990). This procedure results in a slight change of volume of the system; however, as this change represented only 0.2% of the original volume, no correction was performed. Measurements of CO₂, pressure, temperature and mass flow were logged at 100 Hz from the FoxBox to a PC using a Powerlab data acquisition system (8/30; AD Instruments, Oxford, UK) and Chart software (v. 5.5.6 07; AD Instruments). Analysis of the data was carried out using the program IgorPro (Version 5.01; WaveMetrics, Tigard, OR, USA) and the gross rate of CO₂ production was determined for each trial. The rate of CO₂ production was converted into an estimate of the gross rate of energy consumption assuming a respiratory quotient of 0.71 (28 J ml^–1 CO₂) based on previous measurements in moving ants (Lighton and Wehner, 1993; Lipp et al., 2005). The gross metabolic energy expenditure per unit distance travelled on the slope (gross C_path) was calculated as the ratio between the gross rate of metabolic energy expenditure and average speed [J kg^–1 m^–1; following Minetti and Minetti et al. (Minetti, 1995; Minetti et al., 2002)].

An ant was introduced to the tube, and the tube resealed and oriented to give the desired slope. Seven slopes between –90 and +90 deg, at 30 deg increments, were used. The introduction of the ant into the tube allowed some mixing of room air to occur. Therefore, a brief period was allowed for the CO₂ partial pressure reading to stabilise as the air within the closed system was re-mixed. After the stabilisation period, the ant was encouraged to walk by moving the magnetic stir bar in the tube. When the ant reached the opposite end of the tube, the tube was rotated to the same angle so that, despite travelling from end to end in a sealed chamber, the ant travelled continuously at a specified slope (Fig. 1B; note that for level locomotion rotation of the tube was unnecessary). CO₂ was continually monitored and the ant's locomotion within the chamber filmed at 25 frames s^–1 using a video camera (GR-DVL 9800; JVC, London, UK) mounted approximately 1.5 m from the tube, perpendicular to the tube and level with its centre point. The video recordings were synchronised with the respirometry measurements and were used to identify a 2 min period where the ant was moving as desired. Because the ant was required to turn at the end of the tube, there was inevitably a short period of time where the ant was not moving on the desired slope. Only periods where the ants turned quickly, thereby moving as desired for at least 75% of the time, were used (actual time spent moving as desired was 78.7±0.47%). The mean rate of CO₂ production and the speed were calculated over this 2 min period. It should be noted that speed was calculated as the linear distance travelled divided by walking duration (excluding the time taken for the ant to turn in the tube). For each ant, measurements were attempted once on each slope; however, ants were not always co-operative and a complete set of data across the whole range of slopes was not always obtained.

Statistical analysis and modelling was carried out using the program R (R Development Core Team, 2009). General linear models (GLMs) or generalised additive models (GAMs), as appropriate, were used to model the relationships between slope and speed, rate of CO₂ production and gross C_path. They allowed identification of significant relationships between slope and these response variables. GAMs allow exploration of the extent to which the variation in a response variable is associated with variation in the explanatory variable without specifying the shape of the relationship between the variables (Crawley, 2005). This makes them particularly useful when, as here, the explanatory variable goes through zero, therefore making a hump-shaped relationship more likely than a linear one. The outputs of fitted models were used to calculate the predicted slope of the optimal paths to ascend or descend one unit distance in terms of minimising either the time taken or the metabolic energy expenditure.

The optimal slopes that minimised either vertical journey time or the gross vertical cost of locomotion were thus determined.

All results are presented as means ± s.e.m. (number of replicates) unless otherwise stated.

The gross rate of CO₂ production was not significantly affected by slope (GLM, adj. R²=0.0048, P=0.48; Fig. 2A) and was approximately 1.7 ml g^–1 h^–1 for all slopes, equivalent to a gross metabolic rate of approximately 47 J g^–1 h^–1. Speed was significantly affected by slope (GAM, adj. R²=0.66, P<0.001). The highest speed, 2.0±0.1 cm s^–1 (N=20), occurred during level walking, with speed decreasing at steeper gradients on both an incline and a decline, with a minimum value of 0.7±0.06 cm s^–1 (N=17) while ascending a 90 deg slope (Fig. 2B). Gross C_path was significantly affected by slope (GAM, adj. R²=0.488, P<0.001). There was an approximately U-shaped relationship between gross C_path and slope, with the minimum gross C_path (646.0±51.2 J kg^–1 m^–1, N=20) occurring during level locomotion and the maximum gross C_path (1961.3±167.3 J kg^–1 m^–1, N=17) occurring during ascent of a 90 deg slope (Fig. 2C).

The predicted gross vertical cost of locomotion and vertical journey time (Eqns 1, 2) for walking routes using different slopes are shown in Fig. 3. The path length to cover m_vert necessarily increases as slope angle decreases; path length tends to infinity as the slope approaches zero (i.e. level locomotion) (Fig. 3A,B). The gross metabolic cost of following a path of a given slope was calculated using Eqn 1 (Fig. 3A,C). Predicted minimum metabolic cost paths occur at –45 deg for a vertical descent of 1 m and 51 deg for a vertical ascent of 1 m (Fig. 3C). Time taken to cover a path on a given slope is given by Eqn 2 using the predicted speed for that slope (Fig. 3B). Predicted minimum time paths occur at –51 deg for a vertical descent of 1 m and 57 deg for a vertical ascent of 1 m (Fig. 3D).

The gross metabolic rate of leaf-cutter ants moving at their preferred speed is independent of slope (Fig. 2A). This finding is in contrast to studies in other species where metabolic rate has been found to increase with incline; however, in these previous studies, speed was kept constant for all gradients (Wunder and Morrison, 1974; Full and Tullis, 1990; Snyder and Carello, 2008). In our experiments, leaf-cutter ants were able to freely select their speed of locomotion and, rather than maintaining a constant speed, ants moderated speed, travelling fastest on the shallowest slopes and more slowly as slope increased (Fig. 2B). During level locomotion, gross metabolic rate increases with speed (Hoyt and Taylor, 1981; Taylor et al., 1982; Full and Tullis, 1990; Full et al., 1990); therefore, the modulation of speed with slope in ants (Fig. 2B) may be a strategy to maintain a constant metabolic rate.

To our knowledge, this study is the first to demonstrate an animal modulating speed to maintain a constant metabolic rate with changing slope. This behavioural adaptation suggests that the maintenance of a constant metabolic rate may confer some advantage to the animal as its preservation appears to be prioritised over the minimisation of the time taken to complete a journey. Maintenance of a constant metabolic rate represents a constant effort by the ants and it is possible that this allows them to sustain locomotion over long foraging periods. The moderation of speed with incline has also been noted in a variety of species under experimental (Wickler et al., 2000; Denny et al., 2001) and natural (Jayne and Irschick, 2000; Mogollon and Farji-Brener, 2009) conditions. In some of these species, the moderation of speed could – at least in theory – allow a constant metabolic rate to be maintained at different inclines [e.g. cockroaches (Full and Tullis, 1990); ghost crabs (Tullis and Andrus, 2011); red squirrels (Wunder and Morrison, 1974); guinea fowl (Rubenson et al., 2006); and horses (Wickler et al., 2000)]. However, in previous studies where the preferred speed has been measured [e.g. horses (Wickler et al., 2000)], animals were not found not to moderate speed to maintain a constant metabolic rate. This suggests that, in some species, factors other than endurance [e.g. minimisation of the time taken to complete a journey; reduction of the risk of navigational error (Denny et al., 2001)] are important, there is a difference depending on gait or there may be an effect of body mass (Taylor et al., 1972; Full and Tullis, 1990; Wickler et al., 2000).

The lack of difference in metabolic rate with varying slope could potentially have arisen because of a bias in the proportion of time spent running on each slope. However, there was no systematic variation in the proportion of running time across slopes suggesting that the constancy of gross metabolic rate was not due to ants spending more time resting on steeper slopes.

The result of the moderation of speed and the maintenance of gross metabolic rate is a U-shaped relationship between gross C_path and slope (Fig. 2C): gross C_path is lowest for ants travelling on the level and on shallow declines but increases on both inclines and steeper declines. This is broadly in line with previous findings in larger animals (Margaria, 1968; Full and Tullis, 1990; Wickler et al., 2000) but contradicts previous findings showing only small effects of slope on metabolic cost per unit distance travelled in another species of leaf-cutter ant (Lipp et al., 2005). The reasons for this difference are uncertain, but may be related to the much higher airflow rates used in the earlier study [resulting in air velocities of approximately 24 cm s^–1 (Lipp et al., 2005), compared with approximately 1.9 cm s^–1 in our study]. Gross C_path measured here for level locomotion lies within the range previously measured in Atta colombica (Lighton et al., 1987).

Locomotion on an incline as opposed to on the level is likely to incur a higher cost of locomotion as an increase in muscular work is required to raise the centre of mass; this is metabolically expensive (Fenn, 1923; Kushmerick and Davies, 1969; Bigland-Ritchie and Woods, 1976; Beltman et al., 2004). In addition, increasing muscular work must occur through changes in force generation and/or increased muscular shortening. Increased force could occur through the recruitment of a larger muscle volume; increased muscle shortening may also (though not necessarily, depending on whether muscle shortening velocity changes) require the recruitment of a greater muscle volume to compensate for a reduction in force because of the force–velocity relationship. Recruiting more muscle for either of these two reasons may further increase metabolic cost (Bigland-Ritchie and Woods, 1976; Roberts et al., 1997). However, the case for the increased cost of locomotion on a decline is less clear. During descent, muscles must absorb and dissipate as heat the potential energy lost by the centre of mass. This is achieved by active lengthening of the muscles, which is metabolically inexpensive (Curtin and Davies, 1973; Beltman et al., 2004). In addition, because actively lengthening muscles generate high forces (the negative part of the force–velocity relationship), a smaller volume of muscle may need to be recruited to absorb the potential energy, which is expected to further lower the metabolic costs during descent. However, the finding that gross C_path begins to increase beyond a certain gradient has been reported in other species (Margaria, 1968; Minetti et al., 2002). It has been suggested that, because of muscle fatigue and damage, energy consumption may increase during sustained downhill locomotion as more muscle fibres are recruited (Dick and Cavanagh, 1987). However, it is unlikely that either fatigue or muscle damage would be issues for ants exercising over the durations studied here. Hence, the underlying cause(s) for the decline locomotion being energetically expensive is currently unclear.

The predictions of gross vertical cost of locomotion and vertical journey time using paths of different gradients (Fig. 3) suggest that there are minima for both C_vert (Fig. 3C) and journey time (Fig. 3D). However, the effect of modifying speed means that the routes that minimise C_vert and t_vert are very similar (Fig. 3). The route that minimises vertical journey time incurs a 0.1% increase in the gross vertical cost of locomotion compared with the route that minimises C_vert, and the route that minimises C_vert incurs a 2% increase in vertical journey time compared with the route that minimises t_vert. Therefore, the optimal route for ants undertaking a journey involving vertical displacement is one following a path with a slope of between 51 and 57 deg on an incline and between –45 and –51 deg on a decline in order to keep both vertical journey time and vertical cost of locomotion close to their respective minima. The application of these findings in future behavioural studies will allow us to explore the strategies used by foraging ants to negotiate slopes and to determine whether routes that minimise either energetic or time costs of a journey are preferentially used or whether other factors result in a deviation from these predicted routes.

 
• C_path

  metabolic energy expenditure per unit distance travelled on the slope

 
• C_vert

  vertical metabolic cost of locomotion

 
• GAM

  generalised additive model

 
• GLM

  general linear model

 
• m_vert

  vertical metre

 
• t_vert

  vertical journey time

 
• v

  speed

We are most grateful to Dr William Hughes for the supply of the colony of Acromyrmex octospinosus, and to two anonymous referees for their comments.