The Energetic Cost of Limbless Locomotion

Reprint Series 3 August 1990, Volume 249, pp. 524-527

MICHAELWALTON, BRUCE C. JAYNE, AND ALBERTF. BENNETT

The net energetic cost of terrestrial locomotion by the snake Coluber constrictor, moving by lateral undulation, is equivalent to the net energetic cost of running by limbed animals (arthropods, lizards, birds, and mammals) of similar size. In contrast to lateral undulation and limbed locomotion, concertina locomotion by Coluber is more energetically expensive. The findings donot support the widely held notion that the energetic cost of terrestrial locomotion by limbless animals is less than that of limbed animals.

PECIES WITH REDUCED LIMBS OR NO contact with the ground (4, 12). In narrow limbs and elongate bodies have passageways such as tunnels, snakes often Sevolved independently from limbed perform concertina locomotion exclusively antecedents in several groups of vertebrates: (4). Snakes performing concertina locomo- salamanders, caecilians, amphisbaenians, liz- tion stop periodically, and certain parts of ards, and snakes (1). An important factor the body are moved forward while others proposed to explain the evolution of limb- maintain static contact with the ground. In lessness is its presumptively low energetic passageways, snakes alternately press them- cost, such that energetic expenditure during selves against the sides by forming a series of locomotion by limbless animals is expected bends and then extend themselves forward to be less than that of limbed animals of from the region of static contact (4, 13). In similar size (1,2). Biomechanical arguments comparison to lateral undulation, concertina advanced in support of the low energetic locomotion involves higher momentum cost of limbless locomotion include no costs changes (4), resistance due to static (as well associated with vertical displacement of the as sliding) friction (4), and usually slower center of gravity (1, 3, 4), no costs to forward speed (11) and, therefore, probably accelerate or decelerate limbs (3), and low entails higher energetic costs. cost for support of the body (I). A prelimi- Although these considerations logically nary study, published only as an abstract, suggest differential costs of the two locorno- reported that the energetic cost of locomotor modes, only if we directly determine the tion of the garter snake (Thamnophis sirtalis) metabolic rates of moving animals can these was only 30% of that predicted for a qua- be verified and compared to anticipated drupedal lizard of similar size (5). Although values for limbed animals. In the current that study was preliminary, it has been wide- study, we measured energy expenditure as ly cited in review articles (1-3, 6, 7) and the rate of oxygen consumption of snakes at textbooks (4, 8) in support of mechanical rest (~02rest), in the moments just before arguments for the low cost of limbless loco- locomotory exercise ( i/Oz ,e.,,), and dur- motion. We sought to test the generality of ing locomotion at several speeds (0.2 to 1.0 these conclusions by examining the energet- km hour-' for lateral undulation, 0.06 to ic cost of locomotion in a snake, the black 0.14 km hour-' for concertina locomotion) racer (Coluber constrictor) . on motorized treadmills (14, 15). Endur- A snake may utilize a variety of locomo- ance, measured as time sustained on tread, tor modes, depending on both speed and as a function of speed and locomotor mode surface encountered (4, 9-1 1). Lateral undu- was also determined (16). Videotapes were lation and concertina locomotion are two used to verify locomotor mode and to corre- common modes that use lateral vertebral late frequency of movement with energy movements to generate propulsive forces. expenditure. By dividing oxygen consump- During lateral undulation on the ground, tion by frequency of movement, we estimat- snakes move along an approximately sinu- ed energetic costs of single cycles of lateral soidal trajectory. Bends in the body contact undulatory and concertina movement. with the substrate and push posteriorly on The metabolic response of VO~to speed projections from the ground, propelling the in Coluber constrictor [mass = 102.8 ) 6.1 body forward. All parts of the body move (SE) g, n = 71 locomoting by lateral undu- simultaneously with the same overall speed, lation is similar to that observed in many while forward and lateral components of terrestrial vertebrates with limbs (17): ~02 velocity change as a result of the sinusoidal increases as a linear function of speed (18) trajectory (10, 11). Snakes moving with throughout the range of sustainable speeds lateral undulation experience only sliding (0.2 to 0.5 km hour-'), above which Vo2is constant and endurance decreases (speeds Deparnnent of Ecology and Evolutionary Biology, Uni- greater than 0.5 km hour-') (Fig. 1, A and versity of California, Irvine, CA 92717. B). Oxygen consumption also increased as a

SCIENCE, VOL. 249 linear function of speed during concertina Fig. 2. (A) Frequency of movement as a locomotion by C. constrictor (19) but with a function of speed. For lateral undulation (0) the rcgression equation relating frequcncy of greater slope than that of lateral undulation movement (J)(in cycles per minute) to mean (t = 3.67, df = 28, P < 0.001, Fig. 1A). forward speed (spced) (in kdometcrs pcr The point of intersection between the in- hour) is: J= 54.9 x speed + 1.6, n = 41, creasing and constant phases of the metabol- P < 0.001. Sample sizes for lateral undulation ic resDonse lot defines the maximal rate of are as follows: 0.2 km hour-', n = 6; 0.3 km hour-', n = 6; 0.4 km hour-', n = 6; 0.5 Ian oxyg;n corknption ( VO~ ,,,) and the hour-', n = 5; 0.6 km hour-', n = 7; 0.8 km lowest speed at which i/02 ,, is achieved, hour-', n = 7; 1.0 km hour-', 11 = 3. Thc 0- often termed the maximum aerobic speed regression equation for concertina locomotion 0.0 0.2 0.4 0.6 0.8 1.0 (MAS) (20). Coluber constrictor performing (A) is J= 132 x speed + 1.2, n= 10, Mean speed (km hour-') P < 0.05. Sample slzes for concertina locomo- lateral undulation achieves h2,, = 0.83 B tion arc 0.15 km hour-', n = 5 and 0.2 km ml of O2per gram per hour + 0.06 (n = 5) hour-', n = 5. (B) Gross energetic cost of a at MAS = 0.5 km hour-' (Fig.- 1A). This single cycle of movement as a function of ~0~ is nine times the resting rate and is speed. We calculated these values by dividing similar to the maximum rate previously re- V02 at a particular specd by the frequency of movcmcnt over the same time interval. Sam- ported for this species (21). The MAS is ple sizes are as in Fig. 1A for corresponding one-tenth of the maximum burst specd (22) speeds. of these animals (mean maximum burst speed = 5.5 2 0.4 km hour-'). Endurance decreased too precipitously during concertina locomotion (Fig. 1B) to permit us to Mean speed (km hour -') determine i/02,ax and MAS for this locomotor mode. na locomotion is indistinguishable from by both aerobic metabolism and increasing Rates of oxygen consumption extrapolat- zero (t = 0.89, df = 7, P > 0.40), i/02 contributions of anaerobic metabolism. The ed to zero speed (the y intercept) are often (I= 1.5O,df=7,P>O.lO),and bQ ,,,, energetic cost of a single cycle of lateral elevated above resting rates, and earlier in- (t = 2.36, df = 7, P > 0.50). The y inter- undulation does not change within the vestigators have interpreted this increment cepts for the metabolic responses during range of aerobically sustainable speeds [F(3, as the energetic cost of postural support concertina and lateral undulation are also 9) = 2.57, P = 0.121 (Fig. 2B) but does (20). For C. constrictor performing lateral indistinguishable from each other (t = 1.84, vary significantly among individual snakes undulation, the y intercept is elevated above df = 28, P > 0.05). [F(3, 26) = 33.9, P < O.OOl] (23). zero (t = 2.61, df = 21, P < 0.02) but is Snakes increase speed by increasing the Concertina locomotion has an increased indistinguishable from the resting (t = 1.39, number of cycles of movement per unit time energetic cost compared to that of lateral df=21, P>0.10) or pre-exercise (Fig. 2A). The frequency of lateral undula- undulation. The oxygen consumption of (t = 0.33, df = 21, P > 0.50) rates of oxy- tion continues to increase beyond the range animals performing concertina locomotion gen consumption (Fig. 1A). The y intercept of sustainable speed. Thus, undulation at exceeds ~02predicted for snakes perform- for the metabolic response during concerti- nonsustainable speeds is probably supported ing lateral undulation at similar speeds (Fig. 1A). Furthermore, endurance during concertina locomotion is much less than for Fig. l..(A) Stcady-statc rate of oxygen consumption (PO2) as a hction of speed for seven lateral undulation at similar speeds (Fig. individuals performing lateral undulation (0)and 1B). The elevated energetic cost of concerti- three individuals performing concertina locomo- na locomotion is attributable to two factors: tion (A). Mean speed is reported, because animals (i) snakes performing concertina locomo- moving by concertina locomotion periodically stop. The mean resting rate of oxygen consumption require more cycles of movement to tion is indicated by the x at zero speed; prc- sustain the same speed than they do during exercise ratcs of oxygen consumption are indicat- lateral undulation [two-tailed paired t test ed by circles at zero speed. The net cost of comparing the rate (in cycles per minute) of transport for lateral undulation is represented by animals locomoting at 0.2 km hour-', the slope of the line in the increasing region of the n=5, t=8.37, P=0.001, Fig. 2A], and plot (0.2 km hour-' 5 spccd 5 0.5 km hour-'). Mean speed (km hour-') Thc Icasf-squaresestimate of the equation for this (ii) the energetic cost of a single cycle of line is VOz = 1.153 (20.205 SE) x spccd + concertina locomotion is greater than that of 0.222 (20.085 SE), n = 23, P = 0.0001 (curved a single cycle of lateral undulation [Mann- dashcd lincs rcpment 595% confidence limits of predicted values of VOz).An alternative method Whitney U test, U(9, 23) = 179, P < is to calculate separately a regression for each 0.001, Fig. 2B]. snake and then determine thc mean slope and The energetic cost of locomotion by C. intercept of these individual regressions. This constrictor is expressed by the slope of the line method indicated a +ation similar to that of the i/02 model I regression: V02 = 0.916 (20.199 SE) relating to speed within the range of X speed + 0.285 (e0.069 SE) (n = 6). Oxygen aerobically sustainable speeds (17) (Fig. consumption is not related to speed at 20.5 km 1A). This slope, ofien termed the net cost of hour-'. Thc rcgression cquation for this rcgion is Mean speed (km hour-') transport (No, indicates the amount of V02 = -0.069 x.speed + 0.845, P = 0.84. The energy required to move a unit mass of cquation rclating V02 to speed during concertina animal a given distance (20) and is frequent- lo~omotionis h2= 8.494 (51.676 SE) - 0.151 (A0.170 SE), n = 9, P = 0.002. (B) Endurance (time to exhaustion) as a function of speed. All snakes performing latcral undulation at 0.4 km hour-' ly used for comparisons among taxa and sustained locomotion for 120 min; thc trials wcrc stopped at that point. locomotor modes (24, 25). The NCT of

3 AUGUST I990 REPORTS 525 lateral undulatory locomotion for C. constric- tion may neutralize proposed energetic withdrawn, and the chamber was ventilated with a constant flow of fresh air. We determined the frac- tor (1.15 * 0.21) is virtually equivalent to benefits associated with limblcssness. For tional oxygcn content of the samples by injecting the that predicted for a limbed lizard (26) of example, limbless animals probably encoun- sample at a constant flow ratc into an Amctek S3A similar mass (predicted NCT = 1.14, Fig. ter grcater external frictional resistive forces oxygen analyzer. Gas samples were injected through columns of water (Drierite) and C02 (Ascarite) 3). In fact, the NCT of C. constrictor is similar than limbed animals (3). Lateral accelera- absorbents before entering the analyxr sensor. Gas to that predicted for terrestrial locomotion tions of the body during limbless movement volumcs were corrected to standard temperature, of birds, mammals, and arthropods of simi- should also add to the energetic cost. Final- ptessutc, and density. The rate of oxygcn consumption was detennined according to the method of D. lar mass (25).Furthermore. the NCT of C. ly, limbless locomotion is not necessarily \, Vlcck [ J. Appl. Phyriol. 62, 2103 (1987)J. constrictor performing concertina locomotion without energetic cost for body support, 15. We measured rates of oxygcn consumption during locomotor exercise at 30°C over a rangc of speeds on (8.49 I+- 1.68) is seven times the NCT for either from muscular activity to maintain motori7rd madmills designed to elicit either latcral C. constrictor lateral undulation and substan- rigidity of the ribs or to elevate the head and undulation or concertina locomotion. A snake at a tially greater than that predicted for limbed anterior regions of the body above the body temperature of 30°C was fitted with a light- we&, dear plastic mask, which was secured to the animals of similar mass (Fig. 3). In contrast ground as the animal moves. head with a small rubber spacer inssrtcd between the to previous observations (5) and widely held In addition to multiple independent ori- dorsal surface of the head and the mask. Room air opinions concerning the cnergetic cbst of gins of limblessness in amphibians and rep- was drawn through the mask at a constant flow ratc of 200 to 300 ml min-I by a length offlexible plastic limbless locomotion (1-4, 649,lateral un- tiles, many fossorial lizards and salamanders tubing attached to the end of the mask. The expired dulatory and concertina locomotion by C. are characterized by elongate bodies and air stream passed from the mask through absorbent columns, through the Amctek S3A analyzer, and constrictor are not morc economical than small but fully functional limbs (1). One through a metered air pump. We determined the walking or running by limbed animals. Be- explanation for the presence of small limbs oxygcn consumption by comparing the fractional cause the data for garter snakes (Thamnophis in such taxa is that they represent a transi- oxygcn content of room air to that ofthe expired air. The rate of oxygcn consumption was calculated sirtalis) were published only as an abstract tional stage toward the evolution of limb- according to the method of P. C. Withers [ J. Appl. (S), it is diffi< to assess this discrepancy. lessness (27). In this view, limbs are seen as Physiol. 42, 120 (1977)l. In the open-flow mask For our study, we verified locomotor mode encumbrances during locomotion through technique it is assumed that oxygen consumption is dependent on pulmonary and buccopharyngcal, but using videotape; and we used a snake, C. narrow tunnels or crevices. Our results con- not cutaneous, respiration. Cutaneous oxygcn con- constrictor, that is known to have a relatively cerning the high cost of concertina locomo- sumption, however, represents only a small fraction high capacity for aerobic metabolism (21). tion suggest an alternative hypothesis. Many (4%)of total gas exchange among tcrrcsuial rcp- tiles [rM. E. Fcder and W. W. Burggren, Biol. Rev. Thus, we are certain that the data we used to limbless lower vertebrates must switch to (Cambridge) 60, 1 (1985)l. Pre-exercise rates of calculate NCT involved only aerobically sus- energetically costly concertina locomotion oxygcn consumption wcrc measurcd during the last 2 to 3 min of a rest period (30 min to 1 hour) More tainable speeds. In the earlier study (5), the within tunnels. In contrast, small but fully the exercise nial. The madmill was thcn stad, and locomotor mode was not identified, al- functional limbs enable animals to perform oxygcn consumption during movement was record- though lateral undulation seems probable, limbed locomotion in narrow tunnels, cd. Snakes usually crawled voluntarily, but many required light mps on the uil with ow fingers or a and the snakes in that study were-tested at which may also convey an energetic benefit soft brush to initiate and encourage continual move- speeds (up to 0.9 krn hour-') that may have that favors their evolutionary retention. ment. Data for individuals that did not mainrain elicited extensive anaerobiosis and, there- continual movement wcrc not uscd in subsequent REFERENCES AND NOTES analyses. To elicit lateral undulation, we uscd a fore, underestimated NCT. straight meadmill (21 cm wide by 80 cm long), the Why is terrestrial limbless locomotion not 1. C. Gans, Am. Zool. 15, 455 (1975). surface of which was covered with dcial grass. 2. J. L. Edwards, in Functional VMebrafe~Uo~phology, IM. To provide surface projections, small artificial grass energetically less expensive than limbed lo- Hildebrand, D. M. Bramble, K. F. Liem, D. B. squares (2.5 cm by 2.5 cm) wcre afIixed to the comotion despite plausiblc biomechanical Wake, Eds. (Behap, Cambridge, MA, 1985). pp. surface 4 un from the treadmill walls and at 17-em arguments for its low cost? Equdy plausible 159-172. intervals in two parallel rows 6 cm apart. The 3. G. ~&s~ink,in Mechanics a~ldEnergetics ojAnimal concertina treadmill consisted of a circular parallel- arguments suggest that certain energetically Loromofiotr, R. McN. Alexander and G. Goldspink, sidcd tunnel (2 m in outside diametcr) mounted on costly features used during limbless locomo- Eds. (Wiley, New York, 1977), pp. 153-167. wheels, which rotated about the center in the hori- 4. C. Gans, Biomechanics: Atr Apprwch lo Vmebrate zontal plane. The floor of the tunnel was 7 cm wide Biology (Univ. of Michigan Press, Ann Arbor, and consisted ofsmooth plywood. Friction tape was 1974). &cd to the vertical sides to provide that surface 5. R. E. Chodrow and C. R Taylor, Fed. Proc. Fed. o - Lizards with a high coefficient of friction. On both tread- Coluber lateral undulation Am. Soc. Exp. Biol. 32,422 (1973). mills, oxygcn consumption was recorded when the - 6. C. Gans, Herpelologica 42, 33 (1986). A - Coluber concertina snakes had achicvcd several minutcs of steady-state 7. H. B. Lillywhite, in &rakes: Ecology and Evolutionary oxygcn consumption. During the lateral undulation Biology, R. A. Seigel, J. T. Collins, S. S. Novak, Eds. trials, spccd was thcn increased by 0.1 or 0.2 km (~Macmillan,New York, 1987), pp. 422-477. hour-' and oxygen consumption was measured 8. W. N. McFarland ei a/., Vertebrate Lije (Maunillan, again. We procccdcd in this fashion until the snake New York, cd. 2, 1985). could no longer maintain position on the tread. We 9. W. Mosauer, Science 76, 583 (1932). attempted to exercisc cach animal at each spccd. 10. J. Gray, j. Exp. Biol. 23, 101 (19%). 16. Endurance was measured as the time to exhaustion 11. B. C. Jaync, Copeia 1986,915 (1986). during locomotion at a particular speed and locomo- 12. , J. ~Morphol. 197, 159 (1988). tor mode. Exhaustion was determined as the pint 13. , J. Exp. Biol. 140, 1 (1988). at which the snake could no loneer kcm Dace with L. 14. The resting rate of oxygcn consumption was mea- the ueadmill. surcd for snakrs maintained in darkened isolated, 17. V. A. Tuckcr, Comp. Biochem. Physiol. 34, 841 cylindrical plastic mmbolic chambers. Snakes were (1970); K. Schmidt-Niclscn, Science 177, 222 placed in the chambers and left undisturbed for 48 (i972j. hours at 30°C. [Allmeasurements were conducted at 18. A total of seven snakes were uscd in the lateral Fig. 3. Net cost of transport (NCT) as a function 30°C a typical field active body temperature for C. undulation experiments. However, only one animal of body mass plotted on a log-log scale. Data for constrictor; H. S. Fitch, Univ. Kans. Publ. Mur. Nar. perfonncd satisfactorily at all spccds. Therefore, limbed lizards wcrc compiled by John-Aldcr et al. Hist. 15, 351 (1963); H. F. Hirth and A. C. King, samplc si7rs at each spud are 56 individuals. Ac- (26). The solid line indicates the rcgression line J. Herpetol. 16, 101 (1969); L. J. Vitr, Copeia 1971, cording to the procedures of J. H. Zar [Biosfatisfical calculated for the Lizard data (curved dashed lines 255 (1971).] Fresh air was ventilated through the A~ialysis (Prentice-Hall, Englcwood CEh, NJ, indicate *95% confidence limits for the predicted chambers at a constant flow rate. At 1900 hours on 1984)], a model I rcgrcssion (multiple values of y the day of the measurements, an initial 20-mI gas for cach value of x) was calculated to examine the values of NCT). Vemcal bars on the lateral sample was drawn for oxygen analysis with a gas- relation betwan i/ozand speed of locomotion by undulation and concertina NCTs for Coluber in&- tight glass syringe. The chamber was sealcd and left lateral undulation. A model 1 rcgression partitions cate 2 1 SE. The garter snake (Thamnophis) datum undisnubcd in the darkened tempencure cabinet for variability into among-pup, regression, deviations is from Chodrow and Taylor (5). 3 hours. A second 20-ml gas sample was thcn from repsion, and within-group components.

SCIENCE, VOL. 249 The test for the significance of the regression was conducted for each snake, and the fastest 50-cm 1982), vol. 13, pp. 155-199; C. F. Herreid, in based on a comparison of the regression and devi- interval of all three trials was selected as the snake's Locomofim and Energetics of Arthropods, C. F. Herreid ations from rcgression mean squares. Confidence maximum burst spccd. and C. R. Fourtncr, Eds. (Plcnum, New York, limits wcrc based on a residual mean square that Using the four individuals with b'Q data at each of 1981), pp 491-526; R. J. Full, B. D. Anderson, C. pools deviations from rcgrmion and within-groups the four sustainable speeds of lateral undulation M. Finncrry, M. E. Feder, J. Exp. Biol. 138, 471 components. (0.2,0.3,0.4, and 0.5 km hour-'), we calculatcd the (1988); M. Walton and B. D. Anderson, ibid. 136, 19. Only three snakes pcrfonned satisfacrory concertina cost per cycle for each 1-min interval within the 2- to 273 (1988). locomotion during the oxygen consumption cxpcri- 5-min records of bQ uscd to calculate the average 25. C. R. Taylor et a/., J. Exp. Biol. 97, 1 (1982); R. J. ments [mass = 107.1 2 6.2 (SE) g]. Two of these values shown in Fig. 1A. These values were com- Full, in Energy Transjomofions in Cells and Organisms, were used during lateral undulation experiments as pared by mans of a two-way analysis of variance W. Wciscr and E. Gnaigcr, Eds. (Thicmc Vcrlag, well. (ANOVA) with spccd as a fued cffect and individ- New York, 1989), pp. 175-182. 20. A. F. Bennett, in Funnioml Vetfebrate Morphology, M. ual as a random cffect [M. J. Norusis, Advanced 26. H. B. John-Alder, T. Garland, A. F. Bennett, Physi- Hidebrand, D. M. Bramble, K. F. Licm, D. B. S/atisticr, SPSS/PC+ (Statistical Package for the 01. Zool. 59, 523 (1986). Wake, Eds. (Behap, Cambridge, MA, 1985), pp. Social Scicnces (SPSS), Inc., Chicago, 1986)l. Fol- 27. R. Landc, Evolurio~~32, 73 (1978). 173-184. lowing the guidclincs of J. H. Zar [Bios/a/isficol 28. Financial support was provided by NSF grants BSR 21. J. A. Ruben, J. Comp. Physiol. 109, 147 (1976). Analysis (Prenticc-Hall, Englcwood Cli&, w, 8600066 and DCB 8812028 to A.F.B. and NSF 22. Burst sped at 3OT was measured as snakes crawled 19841, we calculated F statistics using comparisons grant BNS 8919497 to B.C.J. We are grateful to J. over a 4-m track, 21 cm wide and covered with among mean squares (MS) as follows: FrFed = Davis for technical assistance and to B. Full for aruficial grass and surface projections (as above). Wc MSqpeedMSrpeed x indiwdual and Findividual = though& comments on an early draft of this vigorously tapped rhc tail with a soft brush to elicit MSmdividudMScrmr report. rapid movement and used the timer fcaturc of the A. F. Bennett, in Biology ojthe Reptilia, C. Gans and video system to determine velocity. Three trials were F. H. Pough, Eds. (Academic Press, New York, 16 January 1990; accepted 27 April 1990