Convective oxygen transport during development in embryos of the snapping turtle Chelydra serpentina

For developing animals, the cardiovascular system is an essential component of the oxygen cascade that supports metabolic rate. A number of key variables allow an organism to regulate this system to meet tissue oxygen demands during development. These include changing cardiac output and blood oxygen content (Tazawa and Mochizuki, 1977; Burggren et al., 2011). The role of the embryonic chorioallantoic membrane (CAM), the primary gas exchange surface (Piiper et al., 1980; Corona and Warburton, 2000), has been studied. However, the ontogeny of key parameters such as stroke volume of the heart, blood oxygen content and selective organ perfusion of most egg-laying vertebrates is poorly understood. Quantification of these and other components of convective oxygen transport is necessary to understand how embryos of egg-laying species, such as reptiles, adjust oxygen delivery as they progress through development.

Egg oxygen consumption increases with development as the mass of embryonic tissues increases due to growth and differentiation (Thompson, 1989; Booth, 1998a,b; Booth et al., 2000; Crossley et al., 2017a) and the convective oxygen transport system must be sufficient to supply this demand. While the components of oxygen delivery have not been extensively investigated in developing animals, metabolic rate has been studied in numerous species representing all reptilian lineages (Lynn and Von Brand, 1945; Dmi'el, 1970; Ackerman, 1981; Thompson, 1989; Booth and Thompson, 1991; Whitehead and Seymour, 1990; Thompson, 1993; Aulie and Kanui, 1995; Birchard and Reiber, 1995; Thompson and Stewart, 1997; Booth, 1998a,b; Vleck and Hoyt, 1991; Thompson and Russell, 1998, 1999; Booth et al., 2000; Sartori et al., 2017; Crossley et al., 2017a). The patterns of whole-egg metabolism in many species are characterized as peaking at some point prior to hatching followed by either a plateau or a decrease (Gettinger et al., 1984; Thompson, 1989; Miller and Packard, 1992; Booth, 1998a,b; Booth et al., 2000; Crossley et al., 2017a). The constant or dropping rate of oxygen consumption during the final period of incubation has been speculated to represent a period of time allowing for catch-up growth and simultaneous hatching in many reptilian species (Spencer et al., 2001). However, this pattern may merely represent a transition in the functional properties of convective transport.

In embryonic reptiles, direct measurements of stroke volume and blood oxygen content have not been conducted; however, heart rate and heart mass have been used to estimate cardiac output (Birchard and Reiber, 1996; Crossley and Altimiras, 2005; Nechaeva et al., 2007; Crossley et al., 2017a). Heart mass has been used as a proxy for stroke volume but this approximation has never been validated. Further, when attempting to characterize potential oxygen delivery to tissues using estimates of cardiac output, it is assumed that oxygen content and hemoglobin affinity are constant during development, which may not be the case. In common snapping turtles (Chelydra serpentina), embryonic heart rate drops approximately 20% between 70% and 90% incubation, a change that could result in an overall reduction in cardiac output (Eme et al., 2012; Alvine et al., 2013). Using heart mass as a proxy for stroke volume, a previous study of embryonic snapping turtles reported that cardiac output does not increase in proportion to embryonic mass, which would constrain metabolic function (Birchard and Reiber, 1996). This developmental window corresponds to a relative reduction or plateau in metabolic rate in snapping turtles, which may be attributed to a reduced convective transport ability (Crossley and Burggren, 2009). However, changes in blood properties could offset the apparent limit imposed by the cardiovascular system. These could comprise increased hematocrit and hemoglobin content, while changes in oxygen affinity of the hemoglobin could also impact oxygen delivery (Tazawa and Mochizuki, 1977). This study is the first to investigate multiple aspects of convective oxygen transport during the embryonic development of an ectothermic vertebrate. To understand how convective oxygen transport in embryonic snapping turtles changes, we measured oxygen consumption (V̇_O2) and heart rate (f_H) and calculated oxygen pulse (O₂ pulse). We also analyzed blood collected from CAM arteries and veins to quantify blood oxygen content (Ca_O2 and Cv_O2), hematocrit (Hct), hemoglobin concentration ([Hb]) and hemoglobin oxygen affinity (P₅₀). Finally, we determined relative blood flow (%Q̇_sys) and estimated cardiac output (CO) of the CAM. We hypothesized that as development progressed, embryonic snapping turtles would fuel an increase in metabolic rate by coordinated changes in convective transport, control of blood flow distribution and increasing levels of oxygen carrying capacity of the blood supplying the developing tissues.

Eggs from common snapping turtles, Chelydra serpentina (Linnaeus 1758), were collected in Minnesota and transported to the Biology Department at the University of North Texas. Upon arrival, eggs were labeled, weighed and placed in plastic boxes (approximately 3 l volume), containing vermiculite mixed with water in a mass ratio of 1:1. Boxes were enclosed in plastic Ziploc^® bags ventilated with humidified atmospheric air at 4 l min⁻¹ to maintain the oxygen and water saturation at adequate levels. The bags were held in a walk-in environmental chamber set to 30°C. Water content of the vermiculite was maintained by weighing boxes three times a week and water was added as needed. A total of 104 eggs were used in this study. Eggs were taken at 50%, 70% and 90% incubation (total incubation 53 days), weighed and assigned for measurements of f_H and V̇_O2, collection of blood or injection of microspheres.

Six embryos from each incubation time were used to determine f_H. Two silver wire electrodes (26 gauge) were inserted just under the surface of the shell of each egg, and then eggs were returned to incubation conditions overnight. The electrode leads were connected to an impedance converter (UFI 2991, Morro Bay, CA, USA) with the output signal connected to a data acquisition system (Powerlab model 8/35). Data were collected at 100 Hz using LabChart Pro (V 7.2, ADInstruments, Colorado Springs, CO, USA). f_H was recorded for up to 1 h in an environmental chamber at 30±0.5°C. The average of three 3 min periods over the final 15 min of the recording was used to establish f_H of each embryo.

The same embryos were then used to determine embryonic oxygen consumption, using a modified version of the closed system respirometry as previously outlined (Crossley et al., 2017b). Briefly, individual eggs were enclosed in hermetically sealed glass chambers (67 ml) with two ports, each connected to a three-way stopcock valve. The valves were closed for a period of 240 min at 50% incubation and for 60 min at 70% and 90% incubation. A 60 ml syringe was connected to one stopcock valve, which was opened for withdrawal of the chamber gas sample (20 ml). Air samples were drawn into a 2 m long length of PE-200 tubing open to the room air via another 3-way stopcock valve (a ‘reservoir/sipper’ system; see fig. 1 in Warburton et al., 1995, for a schematic diagram of the system). The 3-way stopcock was then opened to the analyzers, and gas samples passed through a 10 cm drying tube (MLA0343, ADInstruments) in a modified desiccation chamber similar to a desiccant cartridge (MLA6024, ADInstruments). This was connected to the gas analyzers in series for measurement of O₂ (S-3A1, Thermox, Ametek, Berwyn, PA, USA) and CO₂ (CD-3A, Thermox, Ametek). The gas sample was drawn through the analyzers at a rate of 15 ml min⁻¹. The output from each analyzer was connected to a data acquisition system (Powerlab model 8/35, ADInstruments) and gas concentration data were recorded at 20 Hz using LabChart Pro acquisition software (V 7.2, ADInstruments). This procedure was repeated three times, with 10 min allowed between samples during which the chambers were opened to restore ambient gas concentrations. Mean V̇_O2 was calculated based on chamber volume and time. Mass-specific V̇_O2 (ml O₂ kg⁻¹ min⁻¹) was calculated based on embryo mass, which was measured at the completion of all studies (see below). For each egg, oxygen pulse (O₂ pulse) was calculated as the amount of whole-egg V̇_O2 per f_H and mass-specific O₂ pulse as mass-specific V̇_O2 per f_H.

A different group of embryos was used for blood analysis studies. Eggs were removed from incubation and candled to locate either a tertiary artery carrying mixed oxygenated and deoxygenated blood or a vein in the CAM carrying oxygenated blood. Eggs were placed in holders (4 cm diameter) and moved to a temperature-controlled surgical chamber at 30±0.5°C. A portion of the eggshell was removed (∼1 cm²) and the targeted chorioallantoic artery or vein was occlusively cannulated under a dissection microscope (Leica MZ6, Leica Microsystems, Waukegan, IL, USA) using a heat-pulled polyethylene tube (PE-50) filled with 0.9% saline with 50 IU ml⁻¹ heparin, as previously described (Crossley and Altimiras, 2000). A blood sample (approximately 50–80 µl) was immediately taken with a 100 µl Hamilton syringe for analysis.

A fraction of the whole blood was transferred to a 50 µl capillary tube and centrifuged (13,000 rpm) for Hct determination (N=13 for 50% and 70% incubation, N=12 for 90% incubation). The remaining fraction, ranging from 11 to 20 µl, was added to 4 ml of Drabkin's solution and analyzed spectrophotometrically (Shimadzu UV-1800, Cole Parmer, IL, USA) for determination of [Hb] (N=10 for 50% incubation, N=12 for 70% incubation and N=11 for 90% incubation). Fractions of the blood collected from CAM arteries of embryos at 70% and 90% incubation (N=5) were transferred to 60 µl capillaries, and analyzed in a previously calibrated automatic blood gas analysis system (StatProfile pHOx Plus L, Nova Biomedical, Waltham, MA, USA) to measure blood gases (arterial partial pressure of O₂ and CO₂: Pa_O2 and Pa_CO2).

For measurement of oxygen content (Tucker, 1967), a fraction of 8–20 µl of either an arterial or a venous blood sample was immediately inserted into a water-jacketed chamber (1.52 ml volume) containing an oxygen electrode (Radiometer, Copenhagen, Denmark) filled with degassed ferrocyanite solution, maintained at 30°C (±0.5°C) via a circulating water bath (VWR International, LLC, West Chester, PA, USA). The oxygen electrode was connected to a PHM 73 gas analyzer (Radiometer) and calibrated with degassed ferrocyanite solution (Tucker, 1967). Three oxygen partial pressure (P_O2) values were recorded, immediately before the injection of the sample into the chamber and after 1 and 2 min. The difference between the initial P_O2 value and the mean after injection was used for the calculation of oxygen content (ml O₂ 100 ml⁻¹ blood) according to the equation proposed by Tucker (1967). Blood oxygen content values from vessels afferent to and efferent from the CAM were expressed as arteriovenous difference (A–V diff) and used to calculate total blood flow (CO) through the CAM (see below).

From a different group of embryos (7 at 50% and 70% incubation, 8 at 90% incubation), a 50 µl blood sample was diluted with 5 ml of Hemox solution, a manufacturer-provided buffer, and drawn into a cuvette for analysis by an automatic blood oxygen dissociation analyzer (Hemox Analyzer, model B, TCS, New Hope, PA, USA) to determine oxygen binding curves. The system was maintained at 30°C with a circulating water bath (VWR International, LLC). After air calibration, the sample was equilibrated with a gas mixture of O₂ with either 2% or 6% CO₂ (G400, Qubit, Kingston, ON, Canada) to simulate, respectively, arterial and venous CO₂ conditions. After air calibration of the P_O2 value, the sample was deoxygenated with nitrogen, until P_O2 decreased below 0.3 kPa. The output signal was connected to a data acquisition system (Powerlab model 8/35, ADInstruments) at 20 Hz using LabChart Pro acquisition software (V 7.2, ADInstruments). Calculation of the Bohr shift was based on ΔlogP₅₀/ΔlogpH under the different conditions of CO₂. We considered that pH would not vary considerably under different Hb–O₂ saturations and used pH values measured with blood collected from 90% embryos diluted in the Hemox buffer, which were 7.17 under 2% CO₂ and 6.94 under 6% CO₂.

After all measurements and blood collection, embryos were killed with an overdose of isoflurane for measurement of mass of the embryo, heart, liver and yolk to the nearest 0.01 g. Embryonic stage was confirmed by comparison with the staging table of Yntema (1968) for C. serpentina embryos. At 50% of the incubation period, embryos corresponded to stages 19–20, at 70% to stages 23–24 and at 90% to stage 25. Relative cardiac index (relative COI) was calculated as the product of heart rate and heart mass corrected for embryonic body mass.

A separate group of embryos at each incubation time (5 at 50% incubation, 18 at 70% incubation and 15 at 90% incubation) were injected with colored polystyrene microspheres (Dye Track, Triton Technologies, San Diego, CA, USA) to measure regional blood flow. The microspheres were 15 µm in diameter, larger than the reported size of the snapping turtle red blood cells (Frair, 1977). Instrumentation was similar to that previously reported by our group. Briefly, a tertiary vein of the CAM was catheterized for injection of microspheres as described earlier. Microspheres were suspended in 0.9% saline containing 0.05% Tween 80 (to prevent agglomeration of the microspheres) at a final concentration of 1500 spheres µl^–1. Because of embryonic volume load limitation, a total of 52,500 yellow microspheres were injected at 50% incubation and a total of 75,000 at 70% and 90%. The injections were performed with a 100 µl glass Hamilton syringe into the catheterized vein, after a period of stabilization following the microsurgery (approximately 1 h). At the end of the protocol, a blood sample was withdrawn and stored in a 15 ml Falcon tube in order to detect whether any microspheres remained in the circulation. Embryos were killed with an intravenous dose of pentobarbital (50 µg kg⁻¹). Whole embryo, CAM and yolk membrane were immediately dissected, weighed and stored in either 15 or 50 ml Falcon tubes. The protocol for recovery of microspheres is described in detail in the guidelines provided by Triton Technologies Inc. (Mansfield, MA, USA) and summarized by Stecyk and colleagues (2004). We primarily used the sedimentation method based on a series of centrifugations. Briefly, the sedimentation recovery procedure consists of the following steps: alkaline digestion, a series of washes (Triton X-100, acidified ethanol and ethanol) and elution of the dye. We started tissue digestion with 1 mol l⁻¹ KOH (Sigma-Aldrich, St Louis, MO, USA) and added 10,000 blue microspheres to each tube to assess the effectiveness of the recovery process.

Tissues were sonicated to accelerate digestion, using an ultrasonic homogenizer (Fisher Scientific, Waltham, MA, USA), at 70% of full power for 30 s. The digestion step was repeated until a tissue pellet was no longer visible. Samples containing calcified tissue were subjected to additional treatment with a bone-digesting reagent (0.12% EDTA, 5.38% HCl, 94% H₂O). For tissues that were not digested after several repetitions of sonication, vortexing and alkaline digestion (particularly yolk and whole-embryo samples), a vacuum filtration method using a 25 mm diameter filter with a 10 µm pore size was applied (Triton Technologies Inc., part #31079).

After digestion, the samples were subjected to a series of washes, and the colored dye was eluted from the microspheres with acidified Cellosolve acetate (Sigma-Aldrich). Subsamples of each eluted sample (175 µl) were then transferred to a crystal 96-well plate for dye analysis using a spectrophotometer (Synergy H1, Biotek, VT, USA) at absorbance wavelengths of 650 nm for blue spheres and 440 nm for yellow spheres. To calculate total microspheres and recovery rates for each sample, we entered the absorbance values at each wavelength from each tube into a calculation Excel file, provided by the manufacturer. We calculated fractions of blood flow (%Q̇_sys) directed to the vascular bed of the embryo, CAM and yolk, considering the total microspheres recovered from the tissues of individual eggs and correcting for the microsphere recovery rate accessed by the blue control spheres.

By substituting our measured arterial oxygen content (Ca_O2), hemoglobin concentration ([Hb] in g dl⁻¹) and oxygen partial pressure (Pa_O2) at 70% and 90% into the oxygen content equation Ca_O2=[Hb]×1.34 ml O₂ g⁻¹ Hb×Sa_O2+Pa_O2×0.003 ml O₂ mmHg⁻¹ dl⁻¹, we estimated hemoglobin oxygen saturation (Sa_O2).

By combining our measured CAM A–V diff with the measured values for whole-egg V̇_O2 or mass-specific V̇_O2, we estimated CAM blood flow (Q̇_CAM) using the Fick equation (V̇_O2=CO×A–V diff). Then, Q̇_CAM was determined as the mean V̇_O2 calculated from one group of embryos divided by the A–V diff calculated with a different group of embryos.

Data from oxygen equilibrium curves were plotted as Hill allosteric curves using GraphPad Prism^® software. To measure Hb–O₂ affinity, data were fitted to linear regressions (r²≥0.95) in which P_O2 at 50% Hb saturation (P₅₀) and the corresponding cooperativity coefficient (n_H) corresponded to the zero intercept and slope, respectively, of Hill plots {log[Y/(1−Y)]} versus logP_O2 (where Y is the fractional saturation) fitted to seven saturation steps between 20% and 80%.

All data were tested for differences between the incubation periods using parametric one-way ANOVA but whenever data failed normality or homoscedasticity we used the non-parametric ANOVA on ranks. Post hoc Student–Newman–Keuls (SNK) or Dunn's tests were used for detection of differences across groups. For oxygen content, a t-test was used to account for differences between arterial and venous blood. The level of significance was established as P<0.05 and data are presented as means± s.e.m. All statistical analyses were conducted with Sigma Plot V. 11 software.

Egg mass did not differ between the incubation periods sampled (mean egg mass=12.9±0.2 g; Table 1), embryonic yolk-free body mass increased 4.4-fold from 1.62±0.08 g at 50% incubation to 7.16±0.16 g at 90% incubation (one-way ANOVA+SNK; P<0.001; Table 1). Liver also increased 4.5-fold from 0.047±0.002 g to 0.213±0.008 g, whereas heart mass increased 3-fold from 0.007±0.0002 g to 0.022±0.001 g during the same period (one-way ANOVA+SNK; P<0.001; Table 1). Relative heart mass decreased from 0.46±0.02% to 0.31±0.02% (one-way ANOVA+SNK; P<0.001; Table 1).

f_H decreased from 56±2 beats min⁻¹ at 50% incubation to 43±3 beats min⁻¹ at 90% incubation (one-way ANOVA+SNK; P<0.001; Fig. 1A) whereas V̇_O2 of whole eggs increased continuously from 8.5×10⁻³±5×10⁻⁴ ml O₂ egg⁻¹ min⁻¹ at 50% incubation to 19.9×10⁻³±7×10⁻⁴ ml O₂ egg⁻¹ min⁻¹ at 90% incubation (one-way ANOVA+SNK; P<0.001; Fig. 1B). Mass-specific V̇_O2 decreased progressively from 6.0±0.4 ml O₂ kg⁻¹ min⁻¹ at 50% incubation to 2.8±0.1 ml O₂ kg⁻¹ min⁻¹ at 90% incubation (one-way ANOVA+SNK; P<0.001; Fig. 1C) as embryonic mass increased. Oxygen pulse, the ratio of oxygen consumed per heart beat, increased from 1.5×10⁻⁴±1×10⁻⁵ ml O₂ beat⁻¹ at 50% incubation to 5.0×10⁻⁴±4×10⁻⁵ ml O₂ beat⁻¹ at 90% incubation (one-way ANOVA+SNK; P<0.001; Fig. 1D). When calculated with mass-specific V̇_O2, mass-specific O₂ pulse decreased from 0.11 ml O₂ beat⁻¹ kg⁻¹ to 0.07 ml O₂ beat⁻¹ kg⁻¹ (one-way ANOVA+SNK; P<0.001).

Hct and Hb did not change between incubation periods and were 22.4±1.3 and 9.5±0.4 ml O₂ 100 ml⁻¹ blood, respectively (Table 2). Partial pressure of blood gases, Pa_O2 and Pa_CO2, from CAM arteries containing mixed oxygenated and deoxygenated blood also did not differ during the periods measured (70% and 90%; Table 2).

Blood O₂ content from the CAM artery, Ca_O2, did not differ between the incubation periods studied (mean of 0.9±0.1 ml O₂ 100 ml⁻¹ blood; ANOVA on ranks; P=0.824) but CAM venous blood O₂ content, Cv_O2, increased from 2.7±0.4 ml O₂ 100 ml⁻¹ blood at 50% incubation to 5.0±0.5 ml O₂ 100 ml⁻¹ blood at 90% incubation (one-way ANOVA+SNK; P<0.05), resulting in a 2-fold increase in A–V diff (see Fig. 2).

Oxygen equilibrium curves did not differ between 2% and 6% CO₂ conditions in all incubation periods, indicating the lack of a marked Bohr shift (F=0.47; P=0.7; Table 2, Fig. 3). At 50% incubation, we found two patterns of dissociation curves, which reflected different embryonic stages, although they did not differ statistically in embryonic mass (Table 2). Therefore, the period was divided into two subgroups: ‘early’, which was equivalent to stage 19 of development (50%−E) and ‘late’, which was equivalent to stage 20 of development (50% – L). Late 50% presented a dissociation curve pattern similar to that at 70% and 90% incubation and also had similar P₅₀ values. Early 50% had a right-shifted curve and a higher P₅₀ (lower oxygen affinity) when compared with the late 50%, 70% and 90% incubation (one-way ANOVA+SNK; P<0.0001; Table 2, Fig. 3). Hill coefficients (n_H) for 2% and 6% CO₂ curves were all positive (>1) and did not differ between the incubation periods, indicating no significant change in the level of hemoglobin cooperativity (one-way ANOVA+SNK; P=0.41; Table 2).

The distribution of microspheres revealed that relative blood flow directed to the embryonic tissues (%Q̇_body), to the chorioallantoic membrane (%Q̇_CAM) and to the yolk sac (%Q̇_yolk) differed with incubation time (one-way ANOVA+SNK; P<0.0001; Fig. 4). The blue control spheres added to determine microsphere loss during tissue processing revealed a high efficiency of the recovery process, with rates of 89.76±0.02% at 50% incubation, 101.79±0.02 at 70% incubation and 103.30±0.02 at 90% incubation.

Throughout the last half of incubation the percentage of blood flow directed to embryonic tissues (%Q̇_body) was always greater than the flow to the CAM or to the yolk (Fig. 4). %Q̇_body was maximal at 50% incubation (84%) and minimal at 70% incubation (66%) (Fig. 4A,B). The percentage of blood flow directed to the CAM was initially low (9%) at 50% incubation then rose to 31% and fell to 25% between 70% and 90% incubation (Fig. 4). The relative blood flow directed to the yolk sac was 7% at 50% incubation and progressively decreased upon its depletion (Fig. 4), with the small volume of yolk remaining being incorporated into the abdominal cavity before hatching.

By substituting our measured parameters at 70% and 90% into the equation Ca_O2=[Hb]×1.34 ml O₂ g Hb⁻¹×Sa_O2+Pa_O2×0.003 ml O₂ mmHg⁻¹ dl⁻¹, we were able to estimate Hb oxygen saturation (Sa_O2), which increased from 34% to 49% (Table 2).

When we inserted our V̇_O2 and A–V diff data in the Fick equation (CO=V̇_O2/A−V diff) we found that blood flow through the CAM (Q̇_CAM) increased from 0.46 ml min⁻¹ at 50% incubation to 0.71 ml min⁻¹ at 70% incubation, but then decreased to 0.50 ml min⁻¹ at 90% (Fig. 4, Table 2). However, when using mass-specific V̇_O2, Q̇_CAM decreased from 0.33 ml min⁻¹ g⁻¹ at 50% incubation to 0.07 ml min⁻¹ g⁻¹ at 90% incubation (Fig. 4, Table 2). In agreement with this last finding, the relative COI also indicated a 50% decrease in total cardiac output from 50% to 90% incubation (Table 2).

The development of the components of the convective oxygen transport system in relation to changes in metabolic rate were largely unknown in reptilian embryos. In this study, we found that blood properties were relatively constant throughout snapping turtle incubation, while indexes of cardiac output, relative heart mass and mass-specific metabolic rate decreased over the course of incubation. If this pattern persists during the pre-hatching period (>90% of incubation), it could account for previously reported patterns of whole-egg V̇_O2 in snapping turtle embryos.

f_H of snapping turtle embryos decreased significantly at the end of incubation, a common pattern of members of the order Testudines (Alvine et al., 2013; Birchard and Reiber, 1996; Crossley, 1999; Birchard, 2000; McGlashan et al., 2015; Taylor et al., 2014) and in contrast to the relatively constant f_H during incubation of Squamata (Birchard and Reiber, 1996; Nechaeva et al., 2007; Du et al., 2009; Sartori et al., 2017). In snapping turtles, f_H reduction has been related to the onset of parasympathetic tone, which is present at 70% incubation (Alvine et al., 2013), in comparison to the late onset of vagal tone found in green iguana (Sartori et al., 2015) and the lack of vagal tone during incubation in the American alligator (Eme et al., 2011). f_H is a critical variable in the adjustment of cardiac output during periods of increased metabolic demand (Joyce et al., 2018) and its decrease, coupled to the appearance of parasympathetic control, may reflect an overall reduction in tissue demands. Mass-specific metabolic rate did fall at 50% and 90% incubation, supporting this speculation (Fig. 1C).

Snapping turtle embryos markedly increased in body mass (4.4-fold) during the second half of incubation concurrent with the depletion of the yolk (Table 1). To support this growth, egg V̇_O2 doubled during the last half of the incubation period but f_H decreased by 23%, which resulted in a 3-fold increase in O₂ pulse. Our previous study in a squamate reptile showed that V̇_O2 varied independently of f_H (Sartori et al., 2017), indicating that other physiological mechanisms were involved in matching O₂ supply to tissue demand during embryonic incubation. In the present study, heart mass tripled from 50% to 90% incubation (Table 1) and this was coupled to an increase in oxygen pulse (Fig. 1D). However, the size of the heart relative to the size of the embryo decreased from 0.46% at 50% incubation to 0.31% at 90% incubation (Table 1). This is also the case for the turtles Emys orbicularis (from approximately 0.56% at 50% incubation to 0.31% when close to hatching; Nechaeva et al., 2007) and Lepidochelys olivacea (from 0.46% at 70% incubation to 0.39% at 90%; Crossley et al., 2017a), and for the lizard Iguana iguana (from 0.35% at 30% incubation to 0.26% at 90% incubation; Sartori et al., 2017). This difference in heart growth, when compared with the body mass increase, could be offset to some extent by the reduction in tissue oxygen requirement of the embryo, reflected by the decrease in mass-specific V̇_O2 (Sartori et al., 2017) and a consequent decrease in mass-specific O₂ pulse, as described for the marine iguana by Butler and colleagues (2002).

Unlike prior studies, we found progressive increases in the V̇_O2 of snapping turtle eggs during the later stages of development (Fig. 1B). Previous reports of egg V̇_O2 for this species demonstrated that V̇_O2 reached a maximum value before rapidly declining immediately prior to hatching (Gettinger et al., 1984; Miller and Packard, 1992; Birchard and Reiber, 1995). While our findings differed from this, we did not measure V̇_O2 during the final 10% of incubation, which likely contributed to the apparent contradiction with prior work.

Venous blood leaving the CAM, the site of gas exchange, carried more O₂ in embryos at 90% incubation (Fig. 2), which would support the increase in total egg V̇_O2 (Fig. 1B). Factors that would enable the increase in oxygen carrying content of the venous blood could be: (i) an increase in the number of red blood cells or in Hb concentration; (ii) an increased saturation of the Hb, related to the modulation of Hb oxygen affinity; (iii) changes in the level of allantoic shunting, reducing the levels of deoxygenated blood mixed with the oxygenated blood; (iv) increases in CAM vascularization; and (v) decreases in diffusion resistance barrier. We were unable to monitor levels of shunting but found that Hct and Hb content remained unchanged throughout the last half of development. Further, with the exception of early 50% incubation embryos, P₅₀ was constant during the last half of the incubation (Table 2, Fig. 3). Calculated Hb saturation at 90% incubation was 15% higher than that at 70%, which likely contributed to the increase in blood oxygen content of blood returning from the CAM (Table 2). While physical measures of CAM development were not determined in our study, prior work reported that the CAM completely encloses the egg at approximately 60–70% incubation (Webb et al., 1987; Stewart and Florian, 2000; Nechaeva et al., 2007). Nevertheless, CAM mass can continue to increase thereafter, suggesting an expansion of the vascularization (Birchard and Reiber, 1993; Kam, 1993; Corona and Warburton, 2000; Blackburn et al., 2003). An increase in CAM vascularization late in incubation would then support the calculated increase in Hb saturation and allow greater surface area for oxygen loading, which would also impact the overall increase in egg V̇_O2 (Fig. 1B). The relative cardiac output measurements showed that during the second half of incubation the fraction of CO directed to the yolk was very small, further increasing the total amount flowing to embryonic tissues (Fig. 4). Over the final 30% of development, the CAM vasculature received a considerably higher fraction (25–32%) of CO, when compared with that at 50% incubation (Fig. 4). In comparison to other vertebrate groups, this value was lower than that reported for late-stage chicken embryos (44%; Mulder et al., 1997) and for the placenta fraction in near-term fetal sheep (41%; Jensen et al., 1991); however, this could be attributed to differences in developmental temperature and overall metabolic demand. The increase in CO directed to the CAM from 50% to 70% incubation in snapping turtles (Fig. 4A,B) may reflect a higher vascular density and an expanded surface for gas exchange to occur. The majority of CO was directed to the embryo at all stages measured, although it was reduced at 70% incubation in comparison to that at 50% and 90% incubation (Fig. 4). It is important to note that absolute values of blood flow to the organs were not determined because of limitations of the microsphere technique as reported by Mulder and colleagues (1997). The size of the snapping turtle embryos and their small blood volume restricted the collection of a reference blood sample simultaneously with microsphere injection and absolute measures of blood flow proved unsuccessful.

Our estimates of CAM blood flow suggest an overall decrease at 90% incubation (Table 2). Also, the calculated relative cardiac output index indicated that mass-specific cardiac output decreased throughout incubation, which was previously suggested to occur in developing turtles (Birchard and Reiber, 1996; Crossley and Burggren, 2009; Crossley et al., 2017a). This reduction could contribute to a slowing, a plateau or a decline in oxygen consumption previously reported for embryonic turtles (Ackerman, 1981; Booth and Astill, 2001; Reid et al., 2009; Thompson, 1993). In this study, we found that the estimated CO decreased with the progressive increase in egg V̇_O2; however, embryonic mass-specific V̇_O2 decreased as well (Fig. 1C). Nevertheless, the increase in oxygen carrying capacity of the blood and the lower mass-specific V̇_O2 at late stages of development compensate for any CO decline in snapping turtles and therefore it would not compromise metabolic function.

In summary, our findings indicate that snapping turtle embryos increase oxygen carrying capacity while COI and f­_H decrease late in development. While we lack sufficient data to definitively assign the change in mass-specific metabolic rate to changes in convective oxygen transport capacity, our data do suggest that the two are correlated. If this pattern persisted into the final 10% of snapping turtle incubation, it may contribute to the previously reported patterns of whole-egg V̇_O2 during the final stages of embryonic development.

We thank Prof. Turk Rehn for aid with egg collection and Dr Kevin Tate for help in egg and animal care.