Cardiorespiratory response to progressive hypoxia and hypercapnia in the turtle trachemys scripta

Semi-aquatic turtles are diving animals that exhibit an episodic breathing pattern similar to that of many other tetrapod lower vertebrates (see Shelton and Boutilier, 1982; Milsom, 1991; Glass, 1992). This breathing pattern appears to be an inherent component of the central respiratory rhythm generators, but its output may be dynamically altered by various afferent stimuli (Douse and Mitchell, 1990). The cues that modify this respiratory pattern have intrigued comparative respiratory physiologists for many years and, as such, the ventilatory responses to altered blood gas tensions are well documented (Frankel et al., 1969; Jackson, 1973; Hitzig and Jackson, 1978; Silver and Jackson, 1985; West et al., 1989). Lowered blood oxygen tensions have been demonstrated to induce an increase in minute ventilation primarily through increasing ventilatory frequency (decreasing the non-ventilatory period), with tidal volumes being progressively increased as hypoxia becomes more severe (Jackson, 1973; Milsom and Chan, 1986). These effects are thought to be mediated through O₂-sensitive chemoreceptors located in the carotid arteries and in the pulmonary and aortic arches (Frankel et al., 1969; Benchetrit et al., 1977; Ishii et al., 1985; Kusakabe et al., 1988). Hypercapnic acidosis, in contrast, increases ventilation through changes in both frequency and tidal volume, apparently via stimulation of both central and peripherally located chemoreceptors (Hitzig and Jackson, 1978; Benchetrit and Dejours, 1980; Milsom and Jones, 1980; Hitzig et al., 1985).

During breathing episodes, pulmonary blood flow (Q̇_pul) and heart rate (fH) increase to match lung perfusion with ventilation (Burggren, 1975; Shelton and Burggren, 1976; Burggren et al., 1977a; Wang and Hicks, 1996a). Because turtles breath episodically and use their lungs as a potential O₂ store during non-ventilatory periods, an independent control of pulmonary perfusion would enable a finer degree of control over the rate at which O₂ is extracted from the lungs during these non-ventilatory periods. Indeed, evidence exists that supports this idea, as shown by increases in arterial (and decreases in lung ⁠) during dives in the absence of ventilation (Shelton and Burggren, 1976; Burggren et al., 1989). Burggren et al. (1977a), using a similar protocol to that used in the present study, found that exposure to hypoxia induced much larger changes in pulmonary perfusion than in ventilation and therefore reduced the ventilation/perfusion ratio, further supporting an independent control of pulmonary perfusion. In addition, recent reports have indicated an apparent link between arterial oxygen content ⁠, when altered either chemically (Wang et al., 1996) or through reduced hematocrit (Wang et al., 1997), and net cardiac shunt, without concomitant changes in ventilation. However, with the apparent exception of hypoxemic hypoxia, both ventilation and pulmonary perfusion typically increase during hypoxic exposure, making the control of pulmonary perfusion difficult to resolve from the control of ventilation (Wang and Hicks, 1996a; Wang et al., 1997). Therefore, the degree to which the cardiovascular and respiratory systems are independent and capable of eliciting separate responses remains unclear. An independently controlled cardiovascular response may serve to maintain blood gas tensions during non-ventilatory periods or, alternatively, during periods of lowered O₂-carrying capacity (Wang and Hicks, 1996b; Wang et al., 1996, 1997). However, if the relationship between ventilation and pulmonary perfusion is strictly correlative, then hypercapnic stress should stimulate similar increases in both pulmonary perfusion and ventilation. The objectives of this study were therefore to examine the cardiorespiratory responses to altered steady-state arterial O₂ and CO₂ tensions.

Turtles [Trachemys scripta (Gray), body mass 0.6–1.5 kg] were purchased from Charles Sullivan Co. (Nashville, Tennessee, USA) and housed in large aquaria with free access to water and terrestrial environments. The turtles were kept on a 14 h:10 h L:D photoperiod for at least 2 months prior to experimentation to ensure health and ‘seasonal’ (long-day) status. Food was given twice weekly in the form of Trout Chow (Purina) or crickets, but was withheld for 5–7 days prior to experiments. During this time and throughout the experimental protocol, the turtles were maintained at 25 °C.

The turtles were anesthetized by intramuscular injections of Ketamine (140 mg kg⁻¹). Following induction of anesthesia, a 3 cm×5 cm rectangular hole was cut into the plastron above the heart using a Dremel tool. Doppler flow transducers (Triton Technology, San Diego, California, USA) were placed around the left pulmonary artery (2.0–2.8 mm i.d.), close to the bifurcation of the common pulmonary artery, and around the left aortic arch (2.4–3.0 mm i.d.). The leads from the flow transducers were passed through an incision cut into the skin at the right dorso-lateral margin of the neck and then through a hole drilled into the cranial border of the carapace. The right subclavian artery was non-occlusively cannulated (stretched PE-50 tubing) by cannulating the right ventral cervical artery and advancing the cannula into the subclavian artery. Once the cannula had been secured in place, it was flushed with heparinized Ringer’s solution to ensure patency and to prevent blood clotting. The arterial cannula was led out through the hole in the plastron and taped into place. The rectangular piece of the plastron was secured back into place using 5 min epoxy resin. To cannulate the lungs, the turtle was returned to a prone position, and two holes were drilled on each midpoint of the lateral border of the third dorsal scute. The lung tissue on each side was exposed, the underlying lung chamber was cannulated (PE-205) and the cannula was secured into place using a purse-string suture. Prior to placement, the cannulae were fitted through a hole in a rubber serum cork that was then seated into the hole in the carapace and sealed with 5 min epoxy resin to ensure an air-tight seal. The two lung cannulae were then connected via a T-piece with a sealed end so that normal lung ventilation could occur as the animals recovered.

The turtles were allowed 48–72 h to recover after induction of anesthesia to ensure full recovery from surgery and patency of flow probes. The turtles were placed into 5 l terraria 3 h prior to the start of recording. During this time, they were unidirectionally ventilated with the appropriate initial humidified gas (350 ml min⁻¹) using a gas-mixing flow meter (Cameron Instruments). The initial air mixture was either hyperoxic (30 % O₂, 3.5 % CO₂, remainder N₂) or hypocapnic (15 % O₂, 0 % CO₂, remainder N₂), depending on the protocol. A differential pressure transducer (Validyne, Northridge, California, USA) was attached via the side-arm of a T-piece to the air inflow cannulae to measure ventilation frequency and changes in pulmonary pressure. Lung ventilation frequency was measured via changes in inflow pulmonary pressure. While tidal volumes could not be assessed directly, the peak pressure of tidal movements within a bout was used as an index of tidal effort. The position of the differential pressure transducer on the inflow cannula allowed pulmonary pressure to be monitored and ensured that the unidirectional ventilation did not induce large increases in pulmonary pressure.

To alter the alveolar tension, the lungs were perfused for 2.5 h with 30 %, 15 % (normoxic ⁠), 5 %, 2.5 % or 0 % O₂ while holding CO₂ tension at 3.5 % (remainder N₂ in all cases). The O₂ concentrations were sequentially stepped down from 30 % to 0 % to induce progressive declines in and arterial during the experiments. To induce changes in CO₂ tension, the lungs were perfused for 2.5 h each at 0, 2, 4 and 8 % CO₂ while holding O₂ tension at 15 % (remainder N₂ in all cases). Measurements of responses to CO₂ were made by sequentially increasing CO₂ doses because of the long-lasting effects of acid–base change on blood ion concentrations (pH returns to the control value within 1 h of a return to normocapnia, but [HCO₃⁻] remains elevated for over 48 h; Silver and Jackson, 1985). The responses of turtles to varying levels were determined first, and the responses to changing levels were measured the following day. This ensured that CO₂-induced acidification did not confound the response curves.

Arterial blood was drawn (0.5 ml) via the ventral cervical cannulae during the last 5 min of each steady-state period in each condition to measure ⁠, arterial pH (pHa) and total CO₂ (during the CO₂ protocol). Total CO₂ was measured on whole-blood samples using infrared absorption (Capnicon V, Cameron Instruments) on 20 μl samples taken from the drawn blood. The Henderson–Hasselbalch equation was used to calculate estimated values of and [HCO₃⁻] using pK¹ and values derived in Nicol et al. (1983) and corrected to 25 °C. Arterial pH and were measured using pH and O₂ microelectrodes (Radiometer, Copenhagen, Denmark). After measurements had been made, the remainder of the blood was reinjected into the turtle to prevent anemia resulting from repeated blood sampling. The cannula was then flushed with heparinized Ringer’s solution to maintain patency. The electrode was calibrated using a zero-oxygen solution and water equilibrated with room air using a magnetic stir bar. The pH electrode was calibrated using standard buffers (pH 7.0 and 10.0). The electrodes were maintained at the same temperature (25 °C) as the animal during the experiment.

Blood-flow transducers were connected to a directional-pulsed Doppler flow meter (Bioengineering, University of Iowa, model 545C-4). Signals were converted to a digital signal using a DATAQ (model DI 220) A/D conversion board and recorded at 50 Hz using WINDAQ data-acquisition software (DATAQ Instruments, version 1.13 Akron, Ohio, USA). Heart rate was determined from peak-to-peak interval analysis of the left aortic blood flow. The flow probes were calibrated at five different flow rates by excising the left aorta from the animal following experiments and perfusing this vessel with blood that had been previously drawn from the animal. The pressure transducer was calibrated using a static water column prior to each experiment.

Blood flow was initially analyzed in 1 min bins for the last 30 min of each steady-state gas tension. Net cardiac shunt (Q̇_shunt, ml min⁻¹ kg⁻¹) was estimated as described by Wang and Hicks (1996a). In short, the total pulmonary blood flow (Q̇_pul) was estimated by doubling the blood flow in the left pulmonary artery (Q̇_LPA). Total systemic blood flow (Q̇_sys) was estimated by multiplying the blood flow in the left aorta (Q̇_LAo) by 2.8 (the combined blood flow in the right aortic arch, subclavian arteries and carotid arteries is 1.8 times the flow in the left aor tic arch; C omeau and Hicks, 1994). The difference between Q̇_pul and Q̇_sys yields an estimate of the net cardiac shunt: a negative value would indicate a net right-to-left shunt (R-L shunt), whereas a positive value would indicate a net left-to-right shunt (L-R shunt)To be statistically conservative, blood flows (Q̇_LPA and Q̇_LAo), fH, the frequency of tidal movements (f_vent; breaths min⁻¹), the frequency of breathing bouts (f_bout; bouts min⁻¹), the number of tidal movements per bout (breaths bout⁻¹) and peak tidal pressure (T_P) were averaged over the last 30 min of each steady-state gas tension. While analyzing the data in this manner masked the normal cardiorespiratory synchrony that took place during active lung ventilation, it gave a better estimate of the degree to which the cardiorespiratory variables had changed during the steady-state period. A repeated-measures analysis of variance (ANOVA) was used to analyze the effects of each steady-state gas tension on blood flow, fH and breathing variables. Post-hoc analysis using the Student–Newman–Keuls (SNK) method was used where significance was indicated by the ANOVA. A least-squares regression an alysis was used to determine the relationship between Q̇_LPA and f_bout. All statistical tests were performed using Statistica (StatSoft, version 4.5). The level of significance was set at P<0.05, and the data are presented as means ±1 S.E.M.

Ten turtles were exposed to progressive step decreases in while inflow was maintained constant. Decreasing significantly lowered (P<0.0001), while pHa did not differ among the different conditions (Table 1).

Heart rate (fH) appeared to increase slightly with decreasing ⁠, but this increase was not significant (P=0.07, Fig. 1A). Blood flow in the left aortic arch increased with decreasing (P<0.01, Fig. 1B). The prog ressive decline in also induced significant increases in Q̇_LPA (P<0.0001, Fig. 1C). As the turtles became progressively hypoxic, the calculated net L-R shunt flow increased significantly (P<0.05, Fig. 1D).

Reductions in did not alter the tidal frequency (f_vent) (Fig. 2A) but did increase the frequency of breathing bouts (f_bout) at the lowest level (P<0.0001, Fig. 2B). Since f_vent did not appear to change while f_bout increased, the number of breaths per bout was significantly reduced with reduced (P<0.001, Fig. 2C). Peak tidal pressure (T_P) increased significantly with decreasing (P<0.001, Fig. 2D).

Ten turtles were exposed to progressive step increases in while maintaining inflow constant. While there was no directional change in ⁠, there was a significant difference in between the 2 % and 8 % CO₂ treatments (P<0.05, Table 2). Increasing the of the unidirectional gas stream induced a significant decrease in pHa (P<0.001) and a significant elevation in (P<0.001) and in estimated (P<0.001) and [HCO₃⁻] (P<0.001, Table 2).

Altering the induced significant changes in fH (P<0.01) but not in a directional manner (Fig. 3A). Q̇_LAo did not appear to change significantly as the turtles became increasingly hypercapnic, although the ANOVA results were marginal (P=0.053, Fig. 3B). Q̇_LPA (Fig. 3C) and calculated net Q̇_shunt (Fig. 3D) also did not change with increasing ⁠.

Increasing had significant effects on all the respiratory variables (Fig. 4A–D). Tidal frequency (f_vent) increased in an almost linear fashion with progressive hypercapnia (P<0.001, Fig. 4A). Bout frequency (f_bout) also increased with each successive increase in CO₂ level (P<0.001, Fig. 4B). The number of breaths per bout increased initially (P<0.01) but then remained unchanged with further increases in (Fig. 4C). Peak tidal pressure also increased in an almost linear fashion with progressive hypercapnia (P<0.001, Fig. 4D).

Bout frequency and Q̇_LPA increased similarly during hypoxia and therefore their relationship was significantly correlated (r²=0.2285, P<0.001, Fig. 5A). This correlative relationship was not maintained during hypercapnic stress (r²=0.0439, P=0.1941, Fig. 5B).

The cardiorespiratory variables from the 15 % O₂ level on day 1 were compared with the 2 % and 4 % CO₂ values (normoxic/near-normocapnic) from day 2. Heart rate was slightly elevated on the second day of expe riments, b ut this did not result in significant elevations in Q̇_LPA or Q̇_LAo or in alterations in net Q̇_shunt. The respiratory variables were also not affected by the day of experiments. The only respiratory variable that differed between days was f_bout, which was slightly lower on day 1 than at the 4 % CO₂ level on day 2.

Progressive hypoxia and progressive hypercapnia were used separately in the present study to test the hypothesis that the control of pulmonary perfusion may be independent of the control of ventilation in the response to altered arterial blood gas tension. If the control of pulmonary perfusion is directed by output from respiratory control centers, then changes in pulmonary perfusion should match changes in respiratory output. Therefore, during both hypoxia and hypercapnia, ventilation and pulmonary perfusion should increase in a proportional manner. The hypoxic ventilatory response of turtles tends to be fairly moderate compared with their hypercapnic ventilatory response (Jackson, 1973; Burggren et al., 1977a; Hitzig and Jackson, 1978; Glass and Wood, 1983). Therefore, pulmonary perfusion should be higher during hypercapnia, reflecting the higher respiratory drive of hypercapnic stimulation. The major observation of the present study is that this is not the case and that central blood flow regulation may be uncoupled from ventilatory control during times of hypercapnic stress.

The gas concentrations used to achieve normoxic (15 % O₂) and normocapnic (3.5 % CO₂) conditions were chosen because they represent the and levels that have been reported in the lungs of turtles during moderate periods of apnea (Burggren and Shelton, 1979; Shelton and Boutilier, 1982). This percentage CO₂ resulted in and pHa levels that were comparable with the normal range reported in other studies (Frankel et al., 1969; Jackson et al., 1974). The values of 5 % and 2.5 % O₂ were chosen in an attempt to encompass the erythrocyte P₅₀ value of 3.2–3.5 kPa reported for this species (Burggren et al., 1977b). The inflow rate of the unidirectional gas stream was maintained at 350 ml min⁻¹ to try to ensure a constant alveolar gas tension. Because turtles have large multicameral lungs (Perry, 1978; Jones and Milsom, 1982), this flow rate could be used without generating large intrapulmonary pressures (<0.98 kPa for all turtles). However, with this type of lung, it is possible that unidirectional ventilation is not as efficient as normal ventilatory movements at equilibrating gases within the lung. Also, because the unidirectional flows were not corrected for body mass, lung washout during unidirectional ventilation in larger animals may not have been sufficient to overcome the increased ventilatory movements observed during the severely hypoxic periods. Therefore, despite the relatively high unidirectional flow rate used, ventilatory movements by some turtles were still sufficient to elevate the above the expected value. During the anoxic unidirectionally ventilated period, despite an inflow of less than 0.4 kPa, this protocol reduced below 1.33 kPa in only four of the 10 turtles.

Ventilatory frequency and amplitude increased with progressive hypoxia, as reported previously for other reptiles (Jackson, 1973; Benchetrit et al., 1977; Abe and Johansen, 1987). Breathing frequency increased primarily through a decrease in the non-ventilatory period, as evident from an increase in f_bout rather than an increased number of tidal movements (f_vent). In the present study, the increases in T_P preceded alterations in ventilation frequency, suggesting that tidal volume changes may be part of the initial ventilatory response to hypoxic exposure. Unlike the response reported in earlier studies (i.e. Burggren et al., 1977a; Wang et al., 1997), fH did not appear to increase as much with hypoxic exposure as did Q̇_LPA. The apparent lack of response in the present study is probably due to inter-animal variability because the ANOVA results were marginal (P=0.07). The results of the present study also differed from those of previous reports (i.e. Wang and Hicks, 1996a; Hicks and Wang, 1998) in that a net L-R shunt prevailed during normoxic/normocapnic conditions. This difference may have arisen, in part, because fH and Q̇_sys were lower in the present study during these conditions than values reported previously. However, even when fH was slightly elevated, as during the near-normocapnic conditions of the 2 and 4 % CO₂ protocols, the net L-R shunt flow prevailed in these turtles. Despite these differences, the relationship between Q̇_pul and is similar to that reported by Burggren et al. (1977a) during hypoxia and to that reported by Hicks and Wang (1998) during the transition to anoxia. Furthermore, progressive hypoxia led to an increase in the net L-R shunt flow and, therefore, to a larger proportional pulmonary stroke volume, a response similar to that reported in previous studies (Burggren et al., 1977a; West et al., 1992; Hicks and Wang, 1998). During normoxic ventilation, changes in Q̇_pul appear to be better correlated to f_bout than to the number of breaths taken within a bout (Wang and Hicks, 1996a). In the present study, as in the study of Wang et al. (1997), because both Q̇_pul and f_bout increased similarly with decreasing levels of ⁠, the correlative relationship between ventilation and pulmonary perfusion was qualitatively maintained.

During anoxic unidirectional ventilation, Q̇_LPA and fH were maximal. This differs from a recent report that pulmonary blood flow declines and fH decreases during chronic (>2 h) anoxic exposure (Hicks and Wang, 1998). There are two possible reasons for these differences. Hicks and Wang (1998) measured these variables in submerged turtles in which breathing was measured using an inverted funnel connected to a pneumotachograph. Many amphibians (Jones and Milsom, 1982), chelonian reptiles (Wang et al., 1997) and diving mammals and birds (see Jones and Milsom, 1982; Signore and Jones, 1996) show parasympathetic bradycardia during submergence. To test this possible influence on Q̇_LPA, we exposed four turtles to the step changes in while they were submerged in approximately 10 cm of water (data not shown). The blood flow responses to hypoxia in immersed and emersed turtles were not significantly different, suggesting that appears to be a more critical determinant of blood flow than does the immersion/emersion state of the animal. Recently, Crossley et al. (1998) reported that, in anesthetized turtles, pulmonary vascular resistance increases dramatically at a of approximately 1.5 kPa, a level that is lower than the mean achieved in the present study. In the study of Hicks and Wang (1998), in the anoxic turtles was reduced to below 0.3 kPa. In the four turtles from the present study in which was reduced below 1.3 kPa, Q̇_LPA was reduced below the level measured at 2.5 % ⁠, but not below that recorded at the 15 % O₂ level. At this low level, fH was also irregular, which may result from anoxic impairment of cardiac function or from a vagally mediated decrease in cardiac output (Jackson, 1987). Thus, the more severe anoxic exposure in the study of Hicks and Wang (1998) probably accounts for the observed differences. Taken together, this series of findings suggests that there may be a set point at which blood flow to the lungs is no longer a viable option, so that shunting blood to the systemic circulation is enhanced. This level (<1.5 kPa ⁠) appears to be below the reported P₅₀ for this species (Burggren et al., 1977b; West et al., 1989) and is nearer to the critical for the maintenance of standard metabolic rate (see Ultsch and Anderson, 1988). This set point may also be an important determinant in the uncoupling of the cardiorespiratory systems, as demonstrated in Fig. 2B from the study of Hicks and Wang (1998) in which the ventilatory movements of this turtle during anoxia were not accompanied by changes in Q̇_pul. We did not observe any turtles exhibiting this behavior in the present study, but this may once again reflect the differences during anoxic exposure between the present study and that of Hicks and Wang (1998).

Numerous studies have shown that hypercapnic exposure induces increases in chelonian ventilation (Frankel et al., 1969; Jackson et al., 1974; Burggren et al., 1977a; Hitzig and Jackson, 1978; Benchetrit and Dejours, 1980; Silver and Jackson, 1985). The increase in ventilation with hypercapnia is due to stimulation of both peripheral and central chemoreceptors in reptiles (Hitzig and Jackson, 1978; Hitzig et al., 1985). With the levels of CO₂ delivered to the turtles and the changes seen in the blood gas variables, there is little doubt that both sets of receptors participated in the hypercapnic responses. Assuming that pulmonary perfusion is correlated with, and possibly modified by, output from the respiratory centers in turtles, the increase in respiratory drive and respiratory output should also stimulate large increases in Q̇_LPA. This was not the case in the present study. Indeed, Q̇_LPA did not differ as the animals were stepped up from severe hypocapnia (0 %) to severe hypercapnia (8 %) despite an almost linear increase in the ventilatory variables. Thus, appears to be a potent respiratory stimulant, but is unlikely to be involved in cardiovascular control. Therefore, with this stimulus modality, the correlative nature between f_bout and Q̇_pul was not maintained.

These findings differ from that reported by Burggren et al. (1977a). These authors, using a similar protocol to the present study, reported that in an aquatic turtle, Pelomeduas subrufa, and in a tortoise, Testudo pardalis, both hypoxia and hypercapnia increased pulmonary blood flow. The difference in the hypercapnic responses may be due to a slight difference in experimental protocol between the two studies. Burggren et al. (1977a) exposed the turtles to each new gas concentration until an apparent steady state was reached in the ventilation and blood flow traces; typically, this occurred within minutes. With this protocol, it is conceivable that blood gas tensions and acid–base balance were not yet at steady state. Studies by Silver and Jackson (1985) have shown that acid–base adjustments take over an hour to equilibrate when the turtle is exposed to 5.7 % CO₂. Thus, the differences in the pulmonary blood flow response between these two studies may have arisen, in part, from acid–base equilibration differences. However, despite these differences, the results of the study of Burggren et al. (1977a) are similar to those of the present study in that both demonstrate that hypercapnia was a much more potent ventilatory than cardiovascular stimulus. Furthermore, pulmonary blood flow in the tortoise (Burggren et al., 1977a) and in T. scripta (present study) increased more vigorously during hypoxia than during hypercapnic exposure. These similarities further support the hypothesis that pulmonary perfusion and ventilation may be differentially modified in the response to altered arterial blood gas tension.

The question that still needs to be addressed is why blood flow is responsive to O₂ but not to CO₂. The ratio in turtles tends to be skewed towards ventilation when the turtles are breathing normoxic and normocapnic air (Burggren et al., 1977a; Hopkins et al., 1996). This ratio is brought closer to unity in response to hypoxia and increases in body temperature (Burggren et al., 1977a; Kinney et al., 1977). The labile nature of the ratio is suggestive of two independent yet tightly coupled systems. The present study further demonstrates this tight coupling in that, with respect to O₂ homeostasis, both ventilation and Q_pul are similarly modified. The main finding of the present study is that the correlative nature of the cardiorespiratory output is not obligatory, thereby demonstrating that the two systems are independent and may be uncoupled. Thus, the findings of the present study support the concept that oxygen homeostasis is more dependent on changes in pulmonary perfusion than is acid–base balance and provide a basis for further neurophysiological investigation of the sensory control of blood flow in turtles.

The authors would like to thank Dr K. D. O’Halloran for critically reviewing this manuscript. J.K.H. was in receipt of a University Scholar’s Fellowship from the University of Texas at Arlington.