A Journal of Integrative Biology

Lung Ventilation During Treadmill Locomotion in aSemi-Aquatic Turtle, Trachemys scripta

TOBIAS LANDBERG1!, JEFFREY D. MAILHOT2, AND ELIZABETHL. BRAINERD3

1Ecology and Evolutionary Biology Department, University of Connecticut,Storrs, Connecticut2UMASS Memorial Medical Center, Worcester, MA3Ecology and Evolutionary Biology Department, Brown University, Providence,Rhode Island

ABSTRACT It is reasonable to presume that locomotion should have a mechanical effect onbreathing in turtles. The turtle shell is rigid, and when the limbs protract and retract, air in thelungs should be displaced. This expectation was met in a previous study of the green sea turtle,Chelonia mydas; breathing completely ceased during terrestrial locomotion (Jackson and Prange,1979. J Comp Physiol 134:315–319). In contrast, another study found no direct effect of locomotionon ventilation in the terrestrial box turtle, Terrapene carolina (Landberg et al., 2003. J Exp Biol206:3391–3404). In this study we measured lung ventilation during treadmill locomotion in a semi-aquatic turtle, the red-eared slider, Trachemys scripta. Sliders breathed almost continuously duringlocomotion and during brief pauses between locomotor bouts. Tidal volume was relatively small(!1mL) during locomotion and approximately doubled during pauses. Minute ventilation was,however, not significantly smaller during locomotion because breath frequency was higher than thatduring the pauses. We found no consistent evidence for phase coupling between breathing andlocomotion indicating that sliders do not use locomotor movements to drive breathing. We also foundno evidence for a buccal-pump mechanism. Sliders, like box turtles, appear to use abdominalmusculature to breathe during locomotion. Thus, locomotion affects lung ventilation differently inthe three turtle species studied to date: the terrestrial Te. carolina shows no measurable effect oflocomotion on ventilation; the semi-aquatic Tr. scripta breathes with smaller tidal volumes duringlocomotion; and the highly aquatic C. mydas stops breathing completely during terrestriallocomotion. J. Exp. Zool. 2008. r 2008 Wiley-Liss, Inc.

How to cite this article: Landberg T, Mailhot JD, Brainerd EL. 2008. Lung ventilationduring treadmill locomotion in a semi-aquatic turtle, Trachemys scripta. J. Exp. Zool.309A:[page range].

The body plan of turtles is unique in that theribs have fused with dermal bone into a carapaceand the gastralia and some pectoral girdle boneshave been incorporated into the plastron (re-viewed in Gilbert et al., 2008). The lungs, limbgirdles, abdominal muscles, and other viscera areall contained together within the shell formed bythese bony elements. Because the volume withinthe turtle shell is nearly constant, retraction of thepectoral or pelvic limb and girdle elements into theshell drives air out of the lungs, whereas protrac-tion of limb elements creates subatmosphericpressures that can produce inhalation (Gans andHughes, ’67; Gaunt and Gans, ’69). This constantvolume constraint suggests that during locomo-

tion, limb movements should affect the mechanicsof lung ventilation.Experimental evidence from adult female green

sea turtles (Chelonia mydas) suggests that loco-motion may interfere with breathing performance(Prange and Jackson, ’76; Jackson and Prange,’79). During terrestrial locomotion, C. mydas stops

Published online in Wiley InterScience (www.interscience.wiley.com). DOI: 10.1002/jez.478

Received 30 November 2007; Revised 28 May 2008; Accepted 12June 2008

Grant sponsor: US National Science Foundation; Grant numbers:9875245; and 0316174.

!Correspondence to: Tobias Landberg, Ecology and EvolutionaryBiology Department, University of Connecticut, 75 North EaglevilleRoad, Storrs, CT 06269. E-mail: tobias.landberg@uconn.edu

r 2008 WILEY-LISS, INC.

JOURNAL OF EXPERIMENTAL ZOOLOGY (2008)

breathing during bouts of locomotion and resumesbreathing during pauses in locomotion (Jacksonand Prange, ’79). Jackson and Prange (’79)suggested that the use of limb musculature forboth locomotion and breathing could prevent thetwo behaviors from being performed at the sametime.In contrast, a study of breathing and locomotion

in a terrestrial turtle, Terrapene carolina, foundthat box turtles breathe continuously duringlocomotion and there is no measurable effect ofthe stride cycle on lung ventilation (Landberget al., 2003). This remarkable result indicates thatbox turtles can decouple or minimize any effectsthe limbs have on body cavity pressure, most likelyby the action of their abdominal muscles.Two abdominal muscles, the transverse abdomi-

nis (TA) and the oblique abdominis (OA), areconsidered to be the primary ventilation musclesin turtles because they are present in all speciesstudied to date (George and Shaw, ’59; Shaw, ’62)and they are consistently found to be active duringlung ventilation (Gans and Hughes, ’67; Gauntand Gans, ’69). These antagonistic muscles alter-nate bilateral muscle activity to produce exhala-tion (TA) and inhalation (OA) in turtles at rest(McCutcheon, ’43; Gans and Hughes, ’67; Gauntand Gans, ’69). The OA is a thin, cup-shapedmuscle that lies just deep to the skin and spans theinguinal limb pockets anterior to each hindlimbbetween the carapace and the plastron. At rest,this muscle curves into the body cavity and has anaction similar to the mammalian diaphragm. As itcontracts, it flattens to move the flank postero-ventero-laterally, thereby reducing intrapulmon-ary pressure and driving inhalation when theglottis is open.The TA lies deep to the OA. It originates along

the inside of the carapace and is cupped aroundthe posterior half of each lung. The TA raisesintrapulmonary pressure as it contracts, produ-cing exhalation when the glottis is open. Theconvex sides of the TA and the OA face each otherso that when one muscle contracts and flattens,the other is stretched into a curved position fromwhich it can contract to reverse the motion. TheTA may act together with a diaphragmaticusmuscle that wraps around the front half of thelungs (Gaunt and Gans, ’69). The diaphragmati-cus is often largest in highly aquatic turtle speciessuch as softshelled turtles and absent in terres-trial turtles such as tortoises and box turtlessuggesting a primary role in buoyancy control(Shaw, ’62).

We chose Trachemys scripta for this studybecause the other two turtle species studied todate have a number of habitat specializations thatmay have influenced their breathing and/orlocomotor behavior. In contrast, Tr. scripta retainsa very typical semi-aquatic lifestyle that is com-mon to most members of the family Emydidae.The lack of obvious morphological specializationsof this species and its abundance in the wild andpet trade have made it the most common turtleused in laboratory studies.The goals of this study were to determine if

sliders breathe during locomotion and, if so, dothey: (1) use a buccal force pump; (2) userespiratory–locomotor coupling; (3) use a limbpump; (4) show any effects of the stride cycle onthe magnitude of breaths; (5) show any signs ofventilatory constraints imposed by locomotion.

MATERIALS AND METHODS

Experimental animals

Five subadult Tr. scripta elegans (Weid) werehoused together in a 170-L aquarium where theyhad room to swim and bask under light (14:10light cycle). Masses varied daily but rangedbetween about 225 and 300 g. These animals werekept at a water temperature of 26731C and fedcrickets, worms, fish, and Reptomin (Melle, Ger-many). Individuals were identified by a notch filedinto a different marginal scute. All experimentalprocedures and animal rearing protocols wereapproved and performed in accordance with theUniversity of Massachusetts Amherst Institu-tional Animal Care and Use Committee.

Airflow measurements

To measure ventilation in Tr. scripta duringlocomotion, a custom mask was constructed foreach turtle that covered both mouth and nares.These pneumotach masks were almost identicalto those developed for studying ventilation inTe. carolina (Landberg et al., 2003).The masks were made of high-viscosity rubber-

based dental impression material (Henry ScheinCompany, Melville, NY) and custom-fit to the headof each turtle. During each of the several stages inthe mask-building process, individual turtles worea padded restraint collar that prevented themfrom pulling their heads back into the shell.During the first stage, a small amount (!0.2 cm3)of modeling clay was applied over the nares of theturtle, which created an air-filled space in the

LANDBERG ET AL.2

J. Exp. Zool.

mask when it was removed at later stages. Themouth was uncovered during this stage allowinganimals to breathe through the mouth. Dentalimpression material was applied over the clay andthe head (except for the lower jaw and eyes). Onceset, the mask was removed from the turtle andallowed to cure overnight. The mask was thentrimmed, the clay was removed from inside, and abreathing port was glued into the front of themask (in front of the nares where the clay hadbeen). The breathing port (!0.5-cm section of aBic Round Stic pen, Shelton, CT; 0.8 cm outerdiameter and 0.6 cm inner diameter) was sealedand glued to the mask with cyanoacrylate. In stagetwo of mask construction, the mouth was coveredby the mask and turtle breathed via the naresthrough the open breathing port. A thin bead ofclay was placed over the area where the upper andlower beaks meet and another bead ran along themidline of the upper beak connecting to thebreathing port. Freshly mixed dental impressionmaterial was applied over the clay and the entirehead except the nares. The first, previously curedportion of the mask with the breathing port wasthen pressed onto the wet dental impressionmaterial. After the composite mask set, it wasremoved, allowed to cure overnight, and trimmed.When the clay was removed, the airspace in themask (!0.3 cm3) allowed the turtle to breathethrough either the nares or the mouth. Thefinished masks weighed less than 3 g. On the dayof the experiment, the mask was glued to theturtle’s head with surgical adhesive (cyanoacry-late) and the pneumotach was inserted into theport and sealed with a small amount of petroleumjelly. The entire mask was sealed airtight aroundthe nares and mouth channeling all ventilatoryairflow through the breathing port.The pneumotach was made from two 1-cm long

sections of a 1-cm3 syringe that were separated bya small piece of 53mm screen. Five millimeterpieces of a 18 gauge needle were inserted on eitherside of the screen. These needle segments con-nected 160pe tubing to the differential pressuretransducer (Validyne DP103-06, Northridge, CA).The pneumotach was calibrated with known air-flow rates and by injecting known volumes of airvia syringes. The voltage signal from the pressuretransducer was sent through a carrier demodula-tor (CD-15) to a computer, which displayed a tracein Superscope II 2.1. Exhalation and inhalationvolumes were measured as the area between thezero line and the trace above or below zero,respectively.

In order to allow subsequent frame-by-frameanalysis, a live feed of the breathing trace and astandard dorsal-view video of the animal(30 frames/sec) were synchronized and recordedonto video tape. The image was split so that thebottom half of the screen showed the turtlelocomoting on the treadmill, whereas the top halfof the screen showed the breathing trace.Experiments were run on a low-speed custom-

built treadmill (surrounded with clear acrylicwalls to keep the turtles on the treadmill). Cloacaltemperatures were monitored every hour duringthe experiment and maintained as close aspossible to 301C. A small space heater placed atone end of the treadmill helped control tempera-ture. On occasion, the animal rested near this heatsource resulting in a rapid rise in cloacal tem-perature. These increases (2–61C) could be rapidlyreversed and were most likely not representativeof core body temperature.The experiments were separated into four seg-

ments: acclimation, pre-exercise, locomotion, andrecovery. Before acclimation, the mask was glued onto the animal and the animal was placed into thetreadmill chamber. Acclimation lasted 1hr. Duringthe second hour, pre-exercise ventilation was mea-sured. An experimenter was present and visibleduring the entire experiment. This seemed to deterthe animals from wandering around the treadmillchamber during pre-exercise, pause, and recovery.Locomotion followed pre-exercise and was variablein duration. The goal was to record at least ten boutsof locomotion each containing at least ten breathsand ten strides per individual. For some experi-ments, this goal was never reached and theseexperiments were discarded. For the data reportedin this study it took 1–3hr to obtain ten satisfactorybouts. The animals most often initiated locomotorbouts themselves, but sometimes they were stimu-lated to locomote by starting the treadmill and/orrolling the treadmill back until the shell gentlytapped against the acrylic wall. Treadmill speed wasmanually controlled to match each turtle’s voluntarylocomotor speed. The final hour of the experimentrecorded post-exercise recovery ventilation, and thedata recorded in the first 20min immediately afterlocomotion were used for recovery analysis. Themasks were removed after the experiment and theanimals were returned to their aquarium.

Data analysis

Twenty-minute segments during pre-exercise,locomotion, and recovery were analyzed from four

TURTLE BREATHING DURING LOCOMOTION 3

J. Exp. Zool.

individuals (one experimental animal had anincomplete data set and was excluded from tidalvolume, breath frequency, and minute volumeanalyses). Pre-exercise segments were selected tominimize the animals’ movement in the chamber,which was usually the last 20min of the pre-exercise hour. The 20-minute segments of locomo-tion were chosen to maximize the amount of timespent actually locomoting (the remainder of thetime consisted of pausing between locomotorbouts). Each breath during the 20-min period oflocomotion was designated as occurring eitherduring a locomotor bout or during a pause betweenbouts and these breaths were analyzed separately.Lastly, the first 20-min segment (immediately afterlocomotion) of recovery was chosen to best repre-sent post-exercise recovery. This created fourexperimental ‘‘behaviors’’ to be analyzed (pre-exercise, locomotion, pause, and recovery).The tidal volume of every breath occurring

during the four behaviors was measured individu-ally. All tidal volume values from the fourbehaviors of four individuals were run in a two-way analysis of variance (ANOVA, StatView 4.5)to test for the effects of the behavior andindividual. Post hoc tests, Fisher’s protected leastsignificant difference (Fisher’s PLSD), were usedto test for significant differences in pairwisecomparisons between the four behaviors aftersignificant F-values were observed for behaviorin the ANOVA.Average minute volume (mL/min) was calcu-

lated by summing up the tidal volumes in eachexperimental segment and dividing this sum bythe duration of the segment (20min for pre-exercise and recovery and variable amounts oftime for locomotion and pause depending on theindividual animal’s behavior). This was done forall four behaviors in each of the four individuals,resulting in 16 average minute volumes. Averagebreath frequency (breaths/min) was calculatedsimilarly by counting the number of breathsduring each segment and dividing that numberby the segment duration. Average minute volumeand breath frequency were obtained to comparerelative differences in minute volume and breathfrequency between the four behaviors. As theabsolute value of minute volume and breathfrequency varied between individuals, pairedT-tests (StatView 4.5) were used for pairwisecomparisons between the four behaviors (valuespaired by individual).Phase analysis can reveal the relationship of

breaths to the stride cycle on a polar graph. The

polar graph is a circle that represents the stridecycle from 01 (start of stride) to 3601 (returnto original starting position). Peak airflow fromindividual breaths was plotted according tothe time of occurrence during the stride cycle ofwhich it occurred. This allowed us to test ifthe breaths were distributed randomly throughthe stride cycle or were clustered or absentat any point in the stride cycle. For each of thefive individuals, the first ten locomotor boutscontaining both ten strides and ten breaths wereused in this analysis (these all fell within thesame 20-min segments used for tidal volumeanalysis). A kinematically distinct point in thestride cycle, maximum right hindlimb extension(MHE), was chosen to anchor the polar plots (01).MHE was defined as the video frame where thegreatest knee and ankle extension occurred.Timing of MHE (0.0625 sec was the time intervalbetween frames) was measured from the videorecordings, whereas peak ventilatory airflow wasmeasured from the Superscope file (sampledat least at 200Hz). The timing of peak airflowcould be compared with the relative timingof the stride cycle through the equation: Relativetiming of breath peak5 ((actual time of breathpeak"actual time of MHE)/stride duration). Thetotal stride duration was calculated by subtractingthe MHE preceding the breath peak from theMHE succeeding the breath peak. The result ofthis equation was a number between 1 and 0 that,when multiplied by 360, could be plotted on apolar graph as a degree measure. All inhalationsand exhalations from each individual (range5225–259 breaths per individual) were pooled andanalyzed separately using Raleigh’s test of circularuniformity (Zar, ’96). Polar graphs were createdfor each of the ten bouts of locomotion perindividual showing not only the phase relation-ship of breaths to the stride cycle (around thecircle) but also the magnitude of each tidal volume(on the radius).The relationship between average breathing

frequency and average stride frequency per boutwas analyzed using analysis of covariance (ANCO-VA) with individual turtle as a factor and stridefrequency as a continuous covariate.

RESULTS

Average locomotor speed on the treadmill variedfrom 0.075m/sec to almost 0.18m/sec (Fig. 1).Both stride length and stride frequency contrib-uted to increases in speed, but stride length

LANDBERG ET AL.4

J. Exp. Zool.

increases were small relative to the increases instride frequency (Tables 1 and 2). The gait usedduring these trials was typical for turtles, adiagonal couplet with duty factors for each limbaround 75% of the stride cycle (Fig. 2).In agreement with previous studies (e.g. Vitalis

and Milsom, ’86a), we found that Tr. scriptabreathes intermittently when at rest, generallyin bouts of six to ten breaths interspersed withvariable periods of apnea. During treadmill loco-motion, however, pneumotach flow traces showthat Tr. scripta breathes continually with almostno apneic periods (Fig. 3).We observed some nonventilatory buccal oscilla-

tions at rest, but found no evidence that gularpumping contributes to lung ventilation at rest orduring locomotion in Tr. scripta. In pneumotachflow traces, gular pumping would create a

distinctive pattern of several small inspirationsfollowed by a large exhalation (Brainerd andOwerkowicz, 2006).Breath frequency did not increase or decrease

consistently with increasing stride frequency(ANCOVA, P5 0.99; Fig. 4). Average breathfrequency differed among individuals (ANCOVA,Po0.0001; Fig. 4); however, there was no indivi-dual by stride frequency interaction (ANCOVA,P5 0.23) indicating that all individuals had asimilar independence of breathing and stridefrequencies.

0.6

0.8

1.0

1.2

1.4

1.6

0.08

0.1

0.12

0.14

0.06 0.08 0.1 0.12 0.14 0.16 0.18

Ts 04

Ts 00Ts 01Ts 02Ts 03

Stride length (cm)

Stri

de fr

eque

ncy

(str

ides

per

sec

ond)

Speed (meters per second)

Fig. 1. Stride frequency and stride length vs. treadmillspeed during locomotion in Trachemys scripta. Each pointrepresents the average stride frequency or the average stridelength during a bout of treadmill locomotion (at least tenstrides per bout). All regressions are significant with theexception of stride length in Ts01 (see Tables 1 and 2 forregression statistics).

TABLE 1. Regression results for stride frequency vs. speed

Individual Slope Intercept R2 P-value

Ts00 7.68 0.23 0.94 o0.0001Ts01 8.52 0.22 0.79 o0.0001Ts02 5.04 0.42 0.48 0.0021Ts03 7.67 0.23 0.89 o0.0001Ts04 7.17 0.25 0.98 o0.0001

TABLE 2. Regression equations for stride length vs. speed

Individual Slope Intercept R2 P-value

Ts00 0.14 0.09 0.04 0.0014Ts01 0.16 0.08 0.17 0.0774Ts02 0.44 0.06 0.36 0.0111Ts03 0.13 0.09 0.25 0.0333Ts04 0.24 0.08 0.79 o0.0001

0˚

90˚

180˚

270˚

Right hindlimb

Right forelimb

Left forelimb

Left hindlimb

Fig. 2. Polar graph of footfall patterns during treadmilllocomotion. Each bar shows the average contact time (1standard deviation) of one limb. The end of right hindlimbcontact (01) shows no error because that point was defined as thebeginning of every stride cycle. This representative graph wasconstructed from the two locomotor bouts shown in Figure 3.

TURTLE BREATHING DURING LOCOMOTION 5

J. Exp. Zool.

Plotting the phase of peak inhalatory andexhalatory airflow relative to the 3601 stride cyclereveals no consistent relationship between footfalland breathing patterns in Tr. scripta (Fig. 5). Infour of the five individuals there was no statisti-cally significant phase relationship between thestride cycle and the timing of peak airflow(P40.05, Raleigh’s test), but individual Ts01showed a nonrandom distribution of breaths

relative to the stride cycle (both inhalations andexhalations were nonuniformly distributed;Po0.01, Raleigh’s test). Because the gait used bythese turtles is symmetrical, we also tested forbimodal clustering of breaths by axially trans-forming the angular data (Batschelet, ’81). Sub-sequent Raleigh’s tests were also nonsignificant,indicating uniform distributions and a lack ofbimodality for all individuals (P40.05).Breath magnitude might still be affected by the

stride cycle despite breaths being uniformly dis-tributed around the stride cycle. We examined themagnitude of breaths from all five individualsvisually by plotting both peak airflow rates andtidal volume as functions of the stride cycle. Wefound no observable, consistent effect of the stridecycle on either measure of breath magnitude (Fig.5, distance of the points from the bold zero lineindicates the magnitude of tidal volume).Tidal volumes measured in this study were

generally small, averaging just over 1mL duringlocomotion and doubling to over 2mL during thebrief pauses between locomotor bouts (Figs. 3 and6). Using two-way ANOVA with tidal volume asthe dependent variable and individual and beha-vior as factors, we found significant effects ofindividual and behavior. Individuals Ts00, Ts01,Ts03, and Ts04 were different from each other atthe Po0.0001 level (except Ts01 and Ts04 werenot different from each other P40.05; Fisher’sPLSD post hoc tests). Pre-exercise tidal volumeswere significantly different from locomotion,pause, and recovery (P5 0.0314, Po0.0001, andPo0.0013, respectively; Fisher’s PLSD post hoctests). Locomotion differed significantly frompauses between locomotor bouts but not fromrecovery (Po0.0001 and P40.05, respectively;

0 5 10 15 20 25 30

Airfl ow(ml min-1)

Inhalation

Exhalation

Left hindlimbLeft forelimb

Right forelimbRight hindlimb

Time (s)

LocomotionPauseLocomotion

0200400

-200-400

600

-600

Fig. 3. Representative airflow trace from individual 01 (above in red) and the footfall pattern (below in shades of gray)during two bouts of locomotion separated by a short pause. Contact times for each limb are shown as solid bars. Note thatbreathing is continuous throughout the sequence but the breaths are slightly larger during the pause than during the locomotorbouts.

0.40

0.60

0.80

1.0

1.2

1.4

0.6 0.8 1 1.2 1.4 1.6

Ts 04

Ts 00Ts 01Ts 02Ts 03

Bre

ath

freq

uenc

y (b

reat

hs p

er s

econ

d)

Stride frequency (strides per second)

Fig. 4. Breath frequency vs. stride frequency duringtreadmill locomotion in five Trachemys individuals. Eachpoint represents the average breath frequency of a treadmilllocomotor bout plotted against the average stride frequency ofthat same bout (at least ten strides per bout). Average breathfrequency is not affected by stride frequency (ANCOVA,P40.05; see Results for further details). ANCOVA, analysis ofcovariance.

LANDBERG ET AL.6

J. Exp. Zool.

Fisher’s PLSD post hoc tests). Recovery and pausealso differed significantly (Po0.0001; Fisher’sPLSD post hoc tests).Average breath frequency ranged from 6 breaths

per minute during pre-exercise to 46 breaths perminute during locomotion (Fig. 6). Breath fre-quency was higher during locomotion than duringpre-exercise, pause, and recovery (P5 0.0023,P5 0.0349, and P5 0.0204, respectively; T-testpaired by individual). Pre-exercise and pause alsodiffered significantly (P5 0.0022; T-test paired byindividual). Recovery values did not differ statis-tically from pre-exercise or pause (P40.05; T-testpaired by individual).Average minute volumes ranged from 8mL/min

during pre-exercise to 50mL/min during locomo-tion and 78mL/min during brief pauses inlocomotion (Fig. 6). Minute volumes during pre-exercise were significantly different from locomo-tion and pause values (P5 0.0086 and P5 0.0321,respectively; T-test paired by individual). Locomo-tion and pause values did not differ significantly(P40.05; T-test paired by individual). Recoveryvalues did not differ from any of the otherbehaviors (P40.05; T-test paired by individual).

DISCUSSION

In the words of Mitchell and Morehouse (1863),‘‘That the locomotive movements may, and per-haps do at times modify the respiratory process,may be taken for granted.’’ Locomotion is ex-pected to affect lung ventilation in turtles becauseboth limb girdles flank the lungs and cyclicalmovement of the limbs impinges on the spaceoccupied by the lungs. At rest, small bilaterallysymmetrical limb movements are sufficient todrive both inhalation (during limb protraction)and exhalation (during retraction) in many species(Mitchell and Morehouse, 1863; McCutcheon, ’43;Gans and Hughes, ’67). In green sea turtles,

-4-2024

Ts00

-4-2024

Ts04

-4-2024

Ts01

-4-2024

Ts02

-4-2024

Ts03

tidal

vol

ume

(ml)

tid

al v

olum

e (m

l)

tidal

vol

ume

(ml)

tid

al v

olum

e (m

l)

tidal

vol

ume

(ml) 0˚

90˚

180˚

270˚

Fig. 5. Polar graphs of tidal volume vs. locomotor phasefor five individual Trachemys scripta. Each graph shows themagnitude of inhalation (black circles, recorded as negativevolumes) and exhalation (white squares, recorded as positivevolumes) on the radius. The position of each breath aroundthe circle shows the time of peak airflow from that breathrelative to the stride cycle. Breaths were uniformly distributedwith respect to the stride cycle for all turtles except Ts01 (seeResults) indicating that there is not a fixed phase relationshipbetween locomotion and breathing. A small number ofexceptionally large breaths were cropped from these graphsto improve clarity.

TURTLE BREATHING DURING LOCOMOTION 7

J. Exp. Zool.

C. mydas, terrestrial locomotion interruptsbreathing entirely (Jackson and Prange, ’79).These large, almost entirely aquatic marineturtles crutch along the beach during the nestingperiod without ventilating the lungs and onlybreathe during the pauses between locomotorbouts. In contrast to C. mydas, box turtles (Te.carolina) breathe continually during treadmilllocomotion without any measurable effect of thestride cycle on ventilation (Landberg et al., 2003).We chose Tr. scripta for this study because it has asemi-aquatic lifestyle that is typical for turtles andTr. scripta is closely related to Te. carolina butlacks the morphological specializations associatedwith terrestriality (e.g. dome-shaped shell, verylarge lungs, and plastral hinge).Our results indicate that Tr. scripta breathes

almost continuously during treadmill locomotionusing relatively small, rapid breaths (Fig. 3), andminute ventilation increases substantially duringlocomotion (Fig. 6). This pattern is similar to thatof Te. carolina (Landberg et al., 2003) but differsmarkedly from the complete apnea exhibited byC. mydas during locomotion. This continuousbreathing pattern is also distinct from the typicalventilatory pattern seen in resting turtles that ischaracterized by relatively long periods of apneafollowed by short breathing bouts containingmultiple breaths (e.g. Jackson, ’71).The results of this study also indicate no phase

coupling between respiration and locomotion inTr. scripta. Only one of the five sliders showed anonrandom distribution of breaths relative to thestride cycle, and this individual (Ts01) exhibitedboth inhalations and exhalations around the entirestride cycle (but with some clumping) (Fig. 5). Wealso found no measurable effect of locomotion onbreathing patterns in our previous study of themore fully terrestrial box turtle, Te. carolina(Landberg et al., 2003). The independence ofbreathing and locomotor cycles seen in Te. carolinatherefore does not appear to be a specializationassociated with a terrestrial mode of life, butrather is a more general pattern shared with Tr.scripta, a member of the Emydidae, retaining themore typical semi-aquatic lifestyle of the family.Given the expected mechanical effects of limb

and girdle motions on lung volume in turtles, itseems almost unthinkable that locomotor move-ments could be completely decoupled from breath-ing, yet this is what we find now for bothTr. scripta and Te. carolina. How it is possiblethat the pattern of limb movements can have nomeasurable effect on lung ventilation?

0

10

20

30

40

50

60

Bre

ath

freq

uenc

y (b

reat

hs m

in-1

)

0

1

2

3

4T

idal

vol

ume

(ml)

0

20

40

60

80

100

120

Min

ute

volu

me

(ml m

in-1

)

pre-exercise

locomotion pause recovery

A

B

C

B

A

B

C

AC

A

B

B

AB

Fig. 6. Ventilation in Trachemys during four behaviors(n5 four individuals). (A) Tidal volume (mean1SD). Two-wayANOVA with tidal volume as dependent variable andindividual and behavior as factors with Fisher’s PLSD posthoc tests reveals that all pairwise comparisons are signifi-cantly different except locomotion and recovery (see Results;bars sharing a letter are not significantly different). (B)Breath frequency (mean1SD). Locomotion is significantlydifferent from all other behaviors and pause and pre-exerciseare also different (T-test paired by individual; see Results). (C)Minute volume (mean1SD). Pre-exercise differs significantlyfrom locomotion and pause (T-test paired by individual; seeResults). ANOVA, analysis of variance; PLSD, protected leastsignificant difference.

LANDBERG ET AL.8

J. Exp. Zool.

One possible explanation could be that turtlesbreathe with a buccal pump during locomotion,thereby decoupling lung ventilation from appen-dicular movements. However, no evidence of abuccal pump was found in this study. Buccalpumping behavior would have produced airflowtraces with multiple inhalations followed by alarge single exhalation (e.g. Fig. 8 of Brainerd andOwerkowicz, 2006). Buccal pumping would alsohave been evident in the lateral-view videorecordings as expansion–contraction cycles of thebuccal cavity, but again no such evidence wasfound. Many instances of buccal oscillation wereobserved in Tr. scripta at rest, but as has beenconcluded in every previous study of turtlebreathing (contra Agassiz, 1857), these oscillationswere ventilating the buccal cavity and not thelungs (Mitchell and Morehouse, 1863; McCutch-eon, ’43; Druzisky and Brainerd, 2001; Landberget al., 2003). We conclude that the ventilationmechanism used during locomotion in Tr. scriptadoes not involve the use of a buccal pump.Another possible explanation for the complete

decoupling of respiratory and locomotor cyclescould be that the respiratory muscles are activelycountering changes in pressure or volume inducedby limb movements. Ventilation during terrestriallocomotion in Tr. scripta is almost certainlyproduced by the action of the abdominal mus-

cles—the OA for inspiration and the TA forexhalation. Electromyography demonstrates activ-ity in these muscles during ventilation at rest andduring aquatic locomotion in Tr. scripta (Currie,2001, 2003), but electromyographic recordings ofthese muscles during terrestrial locomotion areyet to be made. Videofluoroscopy of breathingduring locomotion in Te. carolina shows move-ments consistent with OA and TA activity duringterrestrial locomotion (Landberg et al., 2003),providing some additional evidence that thesemuscles also are responsible for ventilation duringterrestrial locomotion in Tr. scripta.Evidence that the actions of the OA and the TA

could be coordinated with locomotor movementscomes from Currie’s (2003) findings that therespiratory and locomotor motor patterns blendtogether during simultaneous breathing andswimming. Unilateral respiratory muscle activa-tion coupled to hindlimb movement was super-imposed over, or blended with, a bilateral motorpattern that coincided with breathing. The in-dependence of stride cycle and breathing that wefind in emydid turtles could be explained by suchhybrid motor patterns if a bilateral breathingmotor pattern is driven by a central patterngenerator that is independent from a unilateralbreathing pattern synchronized with the stridecycle. As long as the central pattern generators for

hindlimbmovement

displacement

4 seconds

1 cm

exhalation

inhalation

airflow

inguinal flank

Fig. 7. X-ray kinematics of the hindlimb and inguinal flank with simultaneously recorded ventilatory airflow in Terrapenecarolina. Each spike in the hindlimb trace (top) corresponds to lifting of the right hindfoot during treadmill locomotion. A metalmarker was glued superficial to the oblique abdominis muscle on the inguinal flank and tracked from a digital X-ray video(middle trace). Airflow was measured from synchronized pneumotach mask (bottom trace). The inguinal flank reflects both thepatterns of the airflow trace as well as the limb kinematics consistent with the hypothesis that a hybrid (left–right alternatingand bilaterally symmetrical) motor pattern is driving the movements of the abdominal muscles. Methods are described in detailin Landberg et al. (2003).

TURTLE BREATHING DURING LOCOMOTION 9

J. Exp. Zool.

the unilateral and bilateral activities were inde-pendent and these independently controlled sti-muli did not interfere, the muscle could sum thesignals and could conceivably minimize effects ofthe limbs while driving ventilation. Evidence forsuch a motor pattern blending comes largely fromturtles (e.g. Stein et al., ’86; Earhart and Stein,2000), but a variety of vertebrates show the type ofmodularity necessary to hybridize or blend func-tions (reviewed in Tresch et al., 2002). We havesome kinematic evidence from Te. carolina that isconsistent with this explanation. We tracked theinguinal flank (just superficial to the OA) ofTe. carolina with a small metal marker in an X-ray videography experiment (Fig. 7). The motionsof the flank reflect aspects of both airflow and limbkinematics indicating that the flanks may bemoving under the influence of the hybrid alter-nating and bilateral motor pattern. Althoughthese results are suggestive, the substantial pitch,roll, and yaw that the animals undergo duringlocomotion could not be accounted for in thisanalysis.

Effects of locomotion on tidal volume andbreath frequency

Although we found no direct effect of limbmovements on the breathing cycle, we did findevidence that locomotion may constrain tidalvolume in a more global manner in Tr. scripta.Compared with pre-exercise, tidal volumes weresmall during locomotion, but increased to approxi-mately twice the average locomotion values duringbrief pauses in locomotion (Fig. 6). We also found asignificant decrease in breath frequency betweenlocomotion and pause. The increase in tidalvolumes and decrease in breath frequency can-celed each other out so that the minute volumesbetween locomotion and pause were not signifi-cantly different. Overall, breath frequencies dur-ing both locomotion and pause fell within thelowest part of breathing work curves forTr. scripta (Vitalis and Milsom, ’86a,b), indicatingthat the animals chose combinations of frequen-cies and tidal volumes that minimized work.Despite tidal volume being about only about

1mL in Tr. scripta during locomotion, this volumeis still sufficient to reach the gas exchangeportions of the lung. The dead space in Tr. scriptais 0.6mL/kg (Crawford et al., ’76), which for our0.2–0.3 kg animals would correspond to about0.15–0.2mL of dead space. Tidal volume goes upto almost 2mL during brief pauses in locomotion.

Minute ventilation remains the same as duringlocomotion, because frequency goes down, butalveolar ventilation will increase in these pausesowing to the increase in tidal volume relative tothe constant dead space.We do not have good evidence to suggest why

tidal volume is lower during locomotion thanduring pauses in Tr. scripta. It is possible thatlocomotion induces some general mechanical con-straint, such as an overall increase in body cavitypressure that would oppose inhalation. Theblended unilateral and bilateral motor patternhypothesized above could raise body cavity pres-sure over resting values. Increased abdominalpressure during locomotion has been recorded inlizards and a crocodilian (Farmer and Hicks, 2000;Kidd and Brainerd, 2000; Munns et al., 2005). It isalso possible that the neural control of breathing isaffected by locomotion in a manner that decreasestidal volume, either with or without some simul-taneous mechanical constraint.Unlike our current findings for Tr. scripta, our

previous results showed no difference in tidalvolume between periods of locomotion and briefpauses in Te. carolina (Landberg et al., 2003).Therefore, the relationship between ventilationand terrestrial locomotion for the three turtlespecies studied to date is different for all three: thefully terrestrial box turtle, Te. carolina, shows nomeasurable effect of locomotion on ventilation atall; the semi-aquatic red-eared slider, Tr. scripta,shows no phase effect of ventilation, but overalltidal volume is lower during locomotion; andbreathing stops completely during terrestriallocomotion in the green sea turtle C. mydas.

Evolution of lung ventilation mechanismsin turtles

We have documented effective lung ventilationduring terrestrial locomotion in Te. carolina andTr. scripta, and we have observed breathingmovements during terrestrial locomotion inPelomedusa subrufa, Chelydra serpentina, andGopherus polyphemus. On the basis of thesefindings, we hypothesize that the ability to breatheduring terrestrial locomotion is primitively sharedby all turtles and secondarily lost in adult C.mydas. We predict that juvenile cheloniid seaturtles that unlike the adults use a typical turtlegait are able to ventilate their lungs duringterrestrial locomotion. Additionally, the adultsand juveniles of leatherback sea turtles, Dermo-chelys coricacea, both use a bilateral crutching gait

LANDBERG ET AL.10

J. Exp. Zool.

on land. Comparison between leatherback andgreen sea turtles could provide useful data on theeffects of locomotor mode on breathing mechanics.One of the defining features of turtles is that

their ribs are fused into the carapace, therebymaking costal aspiration, which is the primitiveamniote breathing mechanism, impossible (Brai-nerd and Owerkowicz, 2006). Some researchershave implied that the carapace came first, ‘‘yonecan readily see that the immobility of the ribs andthe rigidity of the shell called for new ways ofaerating the lungs. This difficult problem hasgiven inquisitive mankind almost as much troubleas it must have once given the turtle.’’ (Pope, ’39).In fact, turtles must have had another mechanismfor breathing that was independent of the primi-tive costal ventilation mechanism before theyfused the ribs into the shell.We hypothesize that the respiratory functions

of the abdominal muscles of turtles evolved, asmany accessory ventilation mechanisms appearto have, in response to the primitive mechanicalconflict between respiration and locomotion hy-pothesized to have been present in early amniotes(Carrier, ’87, ’91). Duplication of ventilatoryfunction before evolving the shell would then haveset the stage for the ribs to abandon their role inlung ventilation.This hypothesis is supported by the conclusion

that the abdominal muscles seem to play a key rolein the remarkable and unexpected finding thatthere is no measurable phase coupling betweenthe respiratory and locomotor cycles in twoemydid turtles. The actions of the abdominalmuscles appear to dampen any effects of limbmovements, effectively decoupling the locomotorand breathing cycles. The complexity of inter-actions between ventilation and locomotion inturtles seems to grow with each study, as thethree most-carefully studied species have revealeddifferent modifications of breathing during loco-motion.

ACKNOWLEDGMENT

We thank Jim O’Reilly for lending us the turtlesfor this study. Scott Currie generously provided uswith figures of unpublished work. Mark Mandicaprepared the turtle drawings used in Figure 2.Emily Jerome edited the manuscript. This materi-al is based on work supported by the US NationalScience Foundation under grants 9875245 and0316174 to E. L. B.

LITERATURE CITED

Agassiz LD. 1857. Contributions to the natural history of theUnited States, Vol. 1. Boston: Little Brown.

Batschelet E. 1981. Circular statistics in biology. London:Academic Press.

Brainerd EL, Owerkowicz T. 2006. Functional morphologyand evolution of aspiration breathing in tetrapods. RespirPhysiol Neurobiol 154:73–78.

Carrier DR. 1987. The evolution of locomotor stamina intetrapods: circumventing a mechanical constraint. Paleo-biology 13:326–341.

Carrier DR. 1991. Conflict in the hypaxial musculo-skeletalsystem: documenting an evolutionary constraint. Amer Zool31:644–654.

Crawford EC, Gatz RN, Magnussen H, Perry SF, Piiper J. 1976.Lung volumes, pulmonary blood flow and carbon monoxidediffusing capacity of turtles. J Comp Physiol 107:169–178.

Currie SN. 2001. Turtle ‘‘respiratory muscles’’ switch from asynchronous to an alternating pattern of activity during thechangeover from breathing to swimming. Soc NeurosciAbstr 27:830.12.

Currie SN. 2003. Bilateral activity patterns of pelvic respiratorymuscles during breathing and swimming in turtles. J FASEB.17 Abstract ]793.1 (Suppl A1210).

Druzisky KA, Brainerd EL. 2001. Buccal oscillation and lungventilation in a semi-aquatic turtle, Platysternon megace-phalum. Zoology 104:143–152.

Earhart GM, Stein PSG. 2000. Scratch-swim hybrids in thespinal turtle: blending of rostral scratch and forward swim.J Neurophysiol 83:156–165.

Farmer CG, Hicks JW. 2000. Circulatory impairment induced byexercise in the lizard Iguana iguana. J Exp Biol 203:2691–2697.

Gans C, Hughes GM. 1967. The mechanism of lung ventilationin the tortoise Testudo graeca Linne. J Exp Biol 47:1–20.

Gaunt AS, Gans C. 1969. Mechanics of respiration in thesnapping turtle, Chelydra serpentina (Linne). J Morphol128:195–228.

George JC, Shaw RV. 1959. The structural basis of theevolution of the respiratory mechanism in Chelonia. J AnimMorphol Physiol 1:1–9.

Gilbert SF, Cebra-Thomas JA, Burke AC. 2008. How the turtlegot its shell. In: Wyneken J, Godfrey MH, Bels V, editors.Biology of turtles. Boca Raton, FL: CRC Press. p 1–16.

Jackson DC. 1971. The effect of temperature on ventilation in theturtle, Pseudemys scripta elegans. Respir Physiol 12:131–140.

Jackson DC, Prange HD. 1979. Ventilation and gas exchangeduring rest and exercise in adult green sea turtles. J CompPhysiol 134:315–319.

Kidd C, Brainerd EL. 2000. Abdominal pressure during highspeed locomotion in the Texas spiny lizard, Sceloporusolivaceus. Amer Zool 40:1085A.

Landberg T, Mailhot JD, Brainerd EL. 2003. Lung ventilationduring treadmill locomotion in a terrestrial turtle, Terra-pene carolina. J Exp Biol 206:3391–3404.

McCutcheon FH. 1943. The respiratory mechanism in turtles.Physiol Zool 3:255–269.

Mitchell SW, Morehouse GR. 1863. Researches upon theanatomy and physiology of respiration in the Chelonia.Smithsonian Contrib Knowl 159:1–42.

Munns SL, Hartzler LK, Bennett AF, Hicks JW. 2005.Terrestrial locomotion does not constrain venous return inthe American alligator, Alligator mississippiensis. J ExpBiol 208:3331–3339.

TURTLE BREATHING DURING LOCOMOTION 11

J. Exp. Zool.

Pope CH. 1939. Turtles of the United States and Canada. NewYork: Alfred A. Knopf.

Prange HD, Jackson DC. 1976. Ventilation, gas exchangeand metabolic scaling of a sea turtle. Respir Physiol 27:369–377.

Shaw RV. 1962. A comparative study of the respiratorymuscles in Chelonia. Brevoria 161:1–16.

Stein PS, Camp AW, Robertson GA, Mortin LI. 1986. Blends ofrostral and caudal scratch reflex motor patterns elicited bysimultaneous stimulation of two sites in the spinal turtle.J Neurosci 6:2259–2266.

Tresch MC, Saltiel P, d’Avella A, Bizzi E. 2002. Coordinationand localization in spinal motor systems. Brain Res Rev40:66–79.

Vitalis TZ, Milsom WK. 1986a. Pulmonary mechanics and thework of breathing in the semi-aquatic turtle, Pseudemysscripta. J Exp Biol 125:137–155.

Vitalis TZ, Milsom WK. 1986b Mechanical analysis ofspontaneous breathing in the semi-aquatic turtle, Pseud-emys scripta. J Exp Biol 125:157–171.

Zar JH. 1996. Biostatistical analysis. Upper Saddle River, NJ:Prentice-Hall.

LANDBERG ET AL.12

J. Exp. Zool.