Developmental programming of sarcoplasmic reticulum function improves cardiac anoxia tolerance in turtles

The environment an organism experiences during embryonic or fetal development has a powerful, and often permanent, influence over its phenotype (Moczek, 2015). This phenomenon, termed developmental plasticity, can result in individuals that share a single genotype displaying substantially different phenotypes, depending on the developmental environment (Bateson et al., 2014). Developmental processes are susceptible to oxygen availability, and a lack of oxygen (hypoxia) during development is a particularly potent stressor that can affect multiple organ systems and alter the developmental trajectory of vertebrates (Burggren and Reyna, 2011). In the cardiovascular system, developmental hypoxia has a profound, life-long impact, which, for fetal mammals and birds, leads to remodelling that produces maladaptive phenotypes and predisposes individuals to disease susceptibility in adulthood (Giussani, 2021). Interestingly, recent evidence suggests developmental hypoxia might improve cardiovascular outcomes in some ectothermic vertebrates (Galli et al., 2023). Accordingly, understanding the mechanisms that programme adaptive cardiovascular phenotypes in ectothermic vertebrates has broad ecological and clinical implications.

As with many North American freshwater turtles, the developmental environment of common snapping turtles (Chelydra serpentina) is characterised by routine fluctuations in oxygen (Ackerman and Lott, 2004; Ultsch, 2006). Turtles are oviparous ectotherms, and lay their eggs in subterranean nests, which can become chronically hypoxic, depending on factors that include nest depth, egg metabolic activity, microbial respiration and flooding (Ackerman and Lott, 2004). Consequently, eggs laid in the deepest part of a nest might experience O₂ tensions as low as 10%, whereas eggs at the surface have access to atmospheric tensions of up to 21% O₂ (Ackerman and Lott, 2004; Booth, 2000). Based on mammalian and avian outcomes (Galli et al., 2023), such dramatic differences in developmental oxygen are expected to severely alter the turtle cardiovascular system. Indeed, numerous studies show that embryonic turtles respond to hypoxia very similarly to mammals and birds, with long-lasting alterations to cardiovascular form and function (Eme et al., 2014, 2013, 2021; Sartori et al., 2019; Tate et al., 2015). However, unlike mammals and birds, these changes appear to enhance aspects of cardiovascular performance after hatching in juvenile snapping turtles (Ruhr et al., 2021; Smith et al., 2022; Wearing et al., 2017, 2016). The most distinguishing feature of the hypoxia-programmed turtle heart is its extreme tolerance to anoxia. Snapping turtles from hypoxic incubations maintain cardiac power output and heart rate roughly twofold higher than individuals from normoxic incubations, even during 2 h in anoxic conditions (Ruhr et al., 2021). Hypoxia presumably acts as an external cue that primes an embryonic snapping turtle's physiology to a future life in low-oxygen habitats or activities (e.g. diving underwater for extended periods to forage, thermoregulate or escape a threat). When underwater, freshwater turtles routinely engage in breath-hold dives, during which they are subjected to bouts of hypoxia (Gatten, 1980; Jackson and Silverblatt, 1974). Improved cardiac function, by previous exposure to developmental hypoxia, could conceivably provide an advantage by more rapidly delivering O₂/nutrients to the muscles and removing of CO₂/wastes from the tissues, thereby enhancing underwater performance. Considering that wild freshwater turtles actively exploit oxygen-deprived environments and overwinter in anoxic, ice-covered lakes throughout their lives (Jackson, 2000; Ultsch, 2006), the programming of cardiac anoxia tolerance could be beneficial for their future survival in hypoxic environments.

Developmental programming typically occurs by epigenetic modifications, whereby DNA is differentially methylated, and these changes can persist throughout the life of an individual (Hanson et al., 2014). We have recently begun to unravel the mechanisms underlying the programming of cardiac anoxia tolerance in snapping turtles, and found evidence of adaptive plasticity at the cellular (Ruhr et al., 2019), metabolic (Galli et al., 2021) and molecular (Ruhr et al., 2021) levels. We showed that the remarkable anoxia tolerance of the juvenile turtle heart is apparent in isolated cardiac cells (cardiomyocytes), which is supported by enhanced myofilament calcium (Ca²⁺) sensitivity and low basal production of reactive oxygen species (ROS) (Ruhr et al., 2019). These adaptive phenotypes are associated with changes in gene expression and the differential methylation of genes linked to cardiomyocyte excitation and contraction (Ruhr et al., 2021). Curiously, we revealed that ventricular cardiomyocytes isolated from snapping turtles previously exposed to developmental hypoxia had much smaller Ca²⁺ transients than their normoxic counterparts (Ruhr et al., 2019). This finding suggests that Ca²⁺-cycling pathways and/or Ca²⁺-storage sites in ventricular cardiomyocytes have been permanently modified by developmental hypoxia. Thus, it is possible that modifications to intracellular Ca²⁺ handling could be partly responsible for the programming of turtle cardiac anoxia tolerance.

Maintenance of excitation–contraction (EC) coupling within cardiomyocytes is critically important to surviving hypoxia. Briefly, an action potential initiates the opening of sarcolemmal L-type Ca²⁺ channels that triggers an influx of extracellular Ca²⁺ (I_Ca), which causes a larger release of Ca²⁺ stored in the sarcoplasmic reticulum (SR) through ryanodine receptors (RyRs) (Bers, 2002). This process, termed Ca²⁺-induced Ca²⁺ release, culminates in a rise of cytosolic free Ca²⁺ concentration ([Ca²⁺]_i). The Ca²⁺ then acts as a secondary messenger that binds to troponin-C on the myofilaments, leading to cross-bridge cycling and cardiomyocyte contraction (Gillis, 2011). During relaxation, Ca²⁺ is either pumped back into the SR, through the SR Ca²⁺-ATPase (SERCA) or transported across the sarcolemma, through the Na⁺/Ca²⁺-exchanger (NCX) (Bers, 2002). The magnitude of the rise in [Ca²⁺]_i – the Ca²⁺ transient (Δ[Ca²⁺]_i) – is the chief regulator of contractile force (Bers, 2002).

While EC coupling is a ubiquitous and conserved process among vertebrate classes, there are substantial differences in the ways species cycle Ca²⁺ during contraction and relaxation. Mammals rely heavily on the SR for Ca²⁺ cycling, but ectothermic vertebrates typically possess poorly developed cardiac SR (Bossen and Sommer, 1984; Lillywhite et al., 1999) and rely almost exclusively on trans-sarcolemmal Ca²⁺ flux for contraction and relaxation (Cros et al., 2014; Shiels and Galli, 2014; Shiels et al., 2002b; Vornanen et al., 2002), rather than SR Ca²⁺ release (Tibbits et al., 1990). However, exceptions exist in nature, and it is thought that SR Ca²⁺ cycling is more prevalent in physically active species of ectotherms (Galli et al., 2009a,b; Keen et al., 1992; Shiels et al., 2011, 1999; Shiels and Farrell, 2000) and also maintains cardiac contractility during environmental stress (Kubly and Stecyk, 2019; MacCormack et al., 2003; Shiels and Farrell, 1997). This can occur through many mechanisms, including changes in SR structural organisation, SR density and the expression of SR proteins (e.g. RyR and SERCA). The sensitivity of the myofilaments to Ca²⁺ is also increased by acute and chronic environmental stressors, leading to enhanced contractility (Churcott et al., 1994; Fanter et al., 2017; Klaiman et al., 2014; Ruhr et al., 2019). Mechanisms that increase myofilament Ca²⁺ sensitivity include an increase in pH, reduced adrenergic tone and changes in the abundance or activity of myofilament proteins, such as troponin and myosin (Chung et al., 2016). These earlier studies suggest that modulation of pathways involved in SR Ca²⁺ cycling and myofilament Ca²⁺ sensitivity could play a role in the programming of cardiac anoxia tolerance in turtles.

Although North American freshwater turtles are relatively sedentary and are usually reliant on cardiomyocyte sarcolemmal Ca²⁺ cycling, we propose that developmental hypoxia could permanently alter cardiac SR function in turtles, leading to improved anoxia tolerance. Therefore, in this study, we determined the functional contribution of SR Ca²⁺ cycling during anoxia and reoxygenation in ventricular cardiomyocytes from juvenile common snapping turtles that were previously incubated in either normoxia (21% O₂) or chronic hypoxia (10% O₂) during embryonic development. To our knowledge, we provide the first evidence of developmental plasticity of cardiac SR function. We also demonstrate that freshwater turtles depend on the SR when the heart is challenged by anoxia, highlighting the importance of SR Ca²⁺ cycling (rather than transarcolemmal Ca²⁺ transport) that is observed in the stressed hearts of other vertebrate species (Shiels and Galli, 2014).

Eggs from multiple clutches of common snapping turtles, Chelydra serpentina (Linnaeus 1758), were collected from the wild, in Minnesota, USA, and transported for incubation to the University of North Texas (Denton, TX, USA). Permission to collect the eggs was granted to D.A.C. by the Minnesota Department of Natural Resources (permit no. 21232). Two eggs from each clutch were staged to determine embryonic age. All eggs were artificially incubated by embedding them to their midpoint in vermiculite, inside plastic incubators (2.5-L Ziploc Container, SC Johnson, Racine, WI, USA) that were stored in a walk-in Environmental Control Room (model IR-912L5, Percival Scientific, Perry, IA, USA). The vermiculite was mixed in a 1:1 ratio with water, as previously described (Crossley and Altimiras, 2005).

At approximately 20% development (9–12 days after laying; determined by embryonic staging), eggs were randomly assigned to two incubation groups: either normoxia (atmospheric 21% O₂) or hypoxia (10% O₂). Given that atmospheric gas composition is nominally 21% O₂ and to distinguish between the experimental oxygen treatment of normoxia, individuals that were incubated in 21% O₂ are denoted as ‘N21’, while those incubated in 10% O₂ are denoted as ‘H10’. These oxygen levels were chosen because they occur naturally in wild turtle nests (Ackerman and Lott, 2004) and produce distinct whole-organ, cellular, metabolic, transcriptomic and methylomic cardiac phenotypes in juvenile and adult snapping turtles (Galli et al., 2021; Ruhr et al., 2021, 2019; Wearing et al., 2017; Wearing et al., 2016).

Because sex determination in turtles is dependent on embryonic temperature, all eggs were incubated at a female-determining 30°C (Alvine et al., 2013; Yntema, 1968). To achieve the desired oxygen levels, parallel gas inflow and outflow tubes were attached to the incubators, and gas mixtures were set to a flow rate of 2–3 l min⁻¹ using rotameters (Sho-Rate Brooks Instruments Division, Hatfield, PA, USA); room air was used for the N21 incubations and a compressed-N₂/room-air mixture was used for the H10 incubations. The gas mixtures passed through an H₂O bubbler to ensure 80–95% relative humidity, and their compositions were monitored continuously with an oxygen analyser (S-3AI, AEI Technologies, Pittsburgh, PA, USA). The eggs began to hatch approximately 55 days after they were laid; thus, the duration of the incubations was between 43 and 46 days.

Upon hatching, all turtles were housed in the common-garden conditions of 21% O₂ and 26°C. Turtles were fed dry crocodilian food (Mazuri, PMI Nutrition International, Brentwood, MO, USA) 2–4 times weekly ad libitum, kept on a daily 12 h:12 h light:dark cycle, and provided with both distilled water and basking areas. At 7 months post-hatch, the hatchlings were transported by air-freight to the University of Manchester (Manchester, UK), where they were individually housed in 21% O₂ and 26°C, kept on a daily 12 h:12 h light:dark cycle, and fed fresh krill and Tetra ReptoMin (Spectrum Brands, Blacksburg, VA, USA), ad libitum, 3 times weekly. Turtle husbandry and experimental procedures were carried out in accordance with University of Manchester handling protocols, which adhere to the UK Home Office legislation. Experiments were carried out when the turtles reached a juvenile age, between 15 and 24 months; N21 turtles weighed (mean±s.e.m.) 314.8±27.4 g (range: 113–477 g) and H10 turtles weighed 717.4±46.7 g (range: 375–984 g), with no significant differences in body mass between the two groups.

Ventricular cardiomyocytes were isolated by enzymatic dissociation, as previously described (Galli et al., 2006b; Ruhr et al., 2019). Briefly, turtles were killed by rapid decapitation; the brain and spinal cord were destroyed by pithing and the head was flash frozen in liquid nitrogen. The heart was quickly excised and cannulated by retrograde perfusion, through the aortic arch into the ventricle. The ventricle was initially perfused with isolation solution for 8–10 min (to remove blood, debris, Ca²⁺ and other blood-plasma ions) and then with enzymatic dissociation solution for 26–30 min (see Table 1 for solution compositions). Solutions were heated to between 28 and 30°C, with an In-line Heater/Cooler Peltier (model SC-20, Warner Instruments, Hamden, CT, USA), connected to a Single-Channel Temperature Controller (model TC-324C, Warner Instruments). After perfusion, the ventricle was removed and minced into small pieces (≤3 mm), and individual ventricular cardiomyocytes were released by gentle agitation. Finally, cardiomyocytes were suspended in isolation solution and stored at 4°C, for up to 8 h. A total of 25 juvenile turtles were used in this study (N=14 and 11 individuals from normoxic and hypoxic incubations, respectively), and their body and heart masses are recorded in Table S1.

The composition of all saline solutions used in this study is provided in Table 1. To our knowledge, the concentration of plasma ions of snapping turtles in normoxia or anoxia has not been determined experimentally, at room temperature (20–22°C). Therefore, our saline solutions were designed according to ion concentrations measured in the plasma of red-eared slider turtles (Trachemys scripta) breathing 21% O₂, at 25°C (Warren and Jackson, 2007), which we used in a previous study (Ruhr et al., 2019). To reveal differences in anoxia tolerance between the N21 and H10 cardiac phenotypes, we subjected the isolated cardiomyocytes to an anoxic saline that combined three main elements of in vivo anoxia: zero oxygen, CO₂ retention and lactate build-up. The precise values of dissolved gases and ion concentrations in plasma were also based on those in T. scripta, after 4 h of anoxia, at 25°C (Warren and Jackson, 2007). While the concentrations of other plasma ions change with in vivo anoxia (e.g. Ca²⁺, Mg²⁺ and K⁺), we limited changing some variables for our simulated anoxic challenge to facilitate the interpretation of phenotypic differences between N21 and H10 cardiomyocytes. All chemicals were purchased from Sigma-Aldrich, unless otherwise stated.

We used epifluorescent microscopy to simultaneously measure cardiomyocyte shortening, intracellular [Ca²⁺] ([Ca²⁺]_i), and intracellular pH (pH_i), during three distinct experimental oxygen periods: normoxia, an anoxic challenge and reoxygenation (Fig. S1A). The isolated cardiomyocytes were placed in a perfusion bath (model Series RC-21BRFS, Warner Instruments; Fig. S1B) with built-in field-stimulation electrodes, which were connected to a stimulator (model SD-9, Grass Instruments, Astro-Med, Inc., West Warwick, RI, USA). The perfusion bath was inserted into a custom-built environmental chamber (Bold-Line Top Stage Incubator, model H301-NIKON-TI-SR, Okolab, Ottaviano, NA, Italy; Fig. S1C), into which an oxygen probe was inserted (mini-sensor model TROXF1100 and FireStingO₂ optical oxygen meter, PyroScience GmbH, Aachen, Germany; Fig. S1C). Gas-tight perfusion lines were also inserted into the chamber and were connected to external gas cylinders. Epifluorescence images of the cardiomyocytes were recorded by mounting the environmental chamber onto an inverted microscope (model Eclipse TE-2000U, Nikon, Surbiton, UK) that was coupled to an Optoscan photomultiplier tube, a monochromator and a high-intensity xenon arc lamp (Cairn Research Instruments, Faversham, UK). Signals were digitised with a Digidata 1440A and analysed with pClamp software (v10.7, Axon Instruments, Sunnyvale, CA, USA).

To measure [Ca²⁺]_i and pH_i simultaneously, resuspended cardiomyocytes were co-loaded with the acetoxymethyl (AM)-ester, cell-permeant fluorescent indicators Fura-2 AM and BCECF AM [2′,7′-bis(2-carboxyethyl)-5(6)-carboxyfluorescein AM; Invitrogen, Loughborough, UK), respectively, at room temperature. Cells were loaded with the indicators by co-incubating them with Fura-2 (0.075 μmol l⁻¹), for 10 min, and BCECF (0.8 μmol l⁻¹), for 30 min. To prevent leakage across the plasma membrane, we de-esterified the indicators, by resuspending the cells in fresh isolation solution, for 15–20 min. To inhibit SR Ca²⁺ cycling, a subset of resuspended cells was co-incubated with ryanodine (10 μmol l⁻¹; a RyR blocker) and thapsigargin (2 μmol l⁻¹; a SERCA blocker), for 30 min, at room temperature.

Following the loading protocol, cardiomyocytes were placed in the recording chamber (at room temperature), perfused with normoxic saline, and stimulated to contract at the physiologically relevant frequency of 0.2 Hz (Frische et al., 2000; Ruhr et al., 2021; Wearing et al., 2017). After a 10 min stabilisation period, the perfusate was switched to the anoxic saline, for 20 min, and cells were subsequently reoxygenated by switching back to the normoxic saline, for a further 10 min (Fig. S1A). Oxygen was measured continuously with the FireStingO₂ probe throughout the protocol and was undetectable during the anoxic period. We note that measurements were taken at a common frequency of 0.2 Hz throughout the experimental protocol (i.e. normoxia, anoxia and reoxygenation). This is a limitation to our study, given that turtles experience a profound bradycardia during anoxia (Jackson, 2000). Therefore, the effects of anoxia on cellular shortening and SR Ca²⁺ cycling might have been different at lower frequencies.

All excitation light waves were filtered with a Nikon T510lpxru dichroic (Chroma Technology Corp, Olching, Germany) long-pass filter and emitted light waves were collected using a HQ535/50 m filter (Chroma) to capture Fura-2 and BCECF emissions. Fura-2 and BCECF were excited sequentially with 340 nm/380 nm and 490 nm/440 nm wavelengths, respectively, and light was collected at 515 nm. Fura-2 input and exit slit widths were each set to 10 nm and BCECF slit widths were set to 7 nm (input) and 4 nm (exit). Fura-2 and BCECF were calibrated using previously published protocols (James-Kracke, 1992; Lattanzio, 1990; Shiels et al., 2002a).

For the BCECF calibration, cells were perfused with a series of solutions of increasing pH (6.5 to 8.0) that contained nigericin (7 µmol l⁻¹), a K⁺ ionophore (Table 1). A four-step pH calibration was used to generate a standard curve, in which the fluorescence output measurements were converted into pH_i.

For the Fura-2 calibration, cells were perfused with an extracellular solution (Table 1) that contained the metabolic inhibitors rotenone (2 µmol l⁻¹), sodium iodoacetate (5 µmol l⁻¹), carbonyl cyanide 3-chlorophenylhydrazone (5 μmol l⁻¹; prevents Ca²⁺ sequestration in the mitochondria), and ionomycin (10 µmol l⁻¹; a Ca²⁺ ionophore). To measure maximum (R_max) and minimum (R_min) Ca²⁺ fluorescence (see below), EGTA (10 mmol l⁻¹; a Ca²⁺-chelating agent) and CaCl₂ (2 mmol l⁻¹) were added to the extracellular solution, respectively.

A high-resolution, infra-red camera (model 902B EIA, Watec, Newburgh, NY, USA) was connected to the Nikon microscope and recorded cardiomyocyte contractions. Cardiomyocyte contractions were collected concurrently with Fura-2 fluorescence, recorded onto DVDs and analysed using the open-source software CellX (https://github.com/jake-rigby/CellX/releases/tag/v1.0.3). CellX allows relaxed and shortened cell length to be accurately measured, but does not have the resolution to faithfully record the kinetics of cellular contraction and relaxation.

GraphPad Prism (v10, GraphPad Software, Boston, MA, USA) and SPSS (v29, IBM, Armonk, NY, USA) were used for figure production and statistical analyses, respectively. Data were tested for equal variances and normality.

Generalised linear models (GLMs), followed by Šidák post hoc tests (for multiple comparisons), were used to determine differences in body mass, heart mass and heart-to-body-mass ratios. For these GLMs, the fixed factor was embryonic oxygen. The data for body and heart mass were normally distributed, but those for the heart-to-body-mass ratios were not. Thus, we fitted the mass data to normal distributions, while we log-transformed the ratio data and fitted them to a gamma distribution.

Repeated-measures, generalised linear mixed-effects models (GLMMs), followed by sequential Šidák post hoc tests, were used to determine differences in cardiomyocyte shortening, Ca²⁺ transients (Δ[Ca²⁺]_i), normalised Δ[Ca²⁺]_i, diastolic [Ca²⁺]_i, systolic [Ca²⁺]_i, pH_i, and kinetics (times to rise and half-decay). When analysing the anoxia/reoxygenation protocol, the GLMM fixed factors were embryonic oxygen incubation [21% O₂ (normoxia) and 10% O₂ (hypoxia)], inhibitor treatment (ryanodine and thapsigargin, to inhibit SR Ca²⁺ cycling), experimental oxygen/time. Because the data for all variables were not normally distributed, we transformed the data using either log or square-root functions and fitted them to normal or gamma distributions, depending on which model provided the best fit (Bolker, 2015), as indicated by Akaike's information criterion (AIC). Δ[Ca²⁺]_i was calculated by subtracting the value of the diastolic [Ca²⁺]_i from the systolic [Ca²⁺]_i.

Finally, to determine the effects of developmental hypoxia, SR inhibition and experimental oxygen on myofilament Ca²⁺ sensitivity, regression analyses and GLMMs were conducted on log- or square-root-transformed data. There were two sets of initial analyses: first, to determine experimental oxygen-dependent effects (i.e. N21 normoxia versus anoxia versus reoxygenation and H10 normoxia versus anoxia versus reoxygenation) and, second, both developmental and experimental oxygen-dependent effects (i.e. normoxia N21 versus H10, anoxia N21 versus H10, and reoxygenation N21 versus H10). The dependent variable was shortening, the fixed factors were experimental oxygen, developmental oxygen and SR inhibition, and the covariate was either Δ[Ca²⁺]_i or pH_i. These initial analyses revealed that the slopes of the regressions did not differ between experimental oxygen treatments alone or when interacting with developmental oxygen. Thus, data were pooled to determine the effect of developmental hypoxia and SR inhibition, with shortening as the dependent variable, developmental oxygen and SR inhibition as the fixed factors, and Δ[Ca²⁺]_i as the covariate.

All data are presented as means±s.e.m. and considered significantly different when P≤0.05. Note that the number of surviving cells is shown Table S2; survival was defined as cells not displaying alternans or not shrinking in size (due to cell death) during the experimental period.

Although body and heart masses of turtles from normoxic and hypoxic incubations did not differ, developmental hypoxia resulted in a significantly higher heart-to-body-mass ratio and significantly shorter cardiomyocytes (Table 2).

Representative traces of the effects of developmental hypoxia and inhibited SR Ca²⁺ cycling on the size, shape and duration of the Δ[Ca²⁺]_i during simulated normoxia, anoxia and reoxygenation are shown in Fig. 1. The GLMs and GLMMs revealed significant effects of developmental oxygen, SR inhibition and experimental oxygen (i.e. anoxia and reoxygenation) on multiple aspects of juvenile turtle cardiomyocyte function. The significant effects of these experimental parameters are apparent in the size, shape and duration of Δ[Ca²⁺]_i between the N21 and H10 groups (Fig. 1). F- and P-values for the fixed factors and their interactions are presented in Table 3, and full statistical results for the findings are given in Table S3 (i.e. test statistics, degrees of freedom and exact P-values).

Prior to anoxia exposure, there were no differences in shortening and pH_i between the control cardiomyocytes (i.e. non-SR-inhibited cells) of N21 and H10 juvenile turtles (Fig. 2). However, similar to our previous study (Ruhr et al., 2019), absolute Δ[Ca²⁺]_i in N21 cells was more than twice the size of that in H10 cells (Fig. 2C,D). The large difference between Δ[Ca²⁺]_i values was due to significantly lower H10 systolic [Ca²⁺]_i levels, whereas diastolic [Ca²⁺]_i levels were similar (Fig. 3). Although hypoxic development significantly reduced Δ[Ca²⁺]_i in H10 cells, it had no impact on shortening.

SR Ca²⁺ cycling was inhibited with ryanodine and thapsigargin, which blocked the ryanodine Ca²⁺ channel and SERCA, respectively. Under normoxic conditions, SR inhibition significantly reduced cardiomyocyte shortening and Δ[Ca²⁺]_i in N21 cells but not in H10 cells (Fig. 2A–D). Therefore, SR inhibition abolished the large difference in Δ[Ca²⁺]_i between H10 and N21 control cells. The reduction in Δ[Ca²⁺]_i for SR-inhibited N21 cells was due to changes in [Ca²⁺]_i, in which diastolic [Ca²⁺]_i tended to rise and systolic [Ca²⁺]_i tended to fall, although these changes were not statistically significant (Fig. 3). The same pattern was observed in H10 cells; SR inhibition significantly decreased diastolic [Ca²⁺]_i at 10 min, but had no effect on systolic [Ca²⁺]_i (Fig. 3). SR inhibition also significantly hastened the time to half-decay in N21 and H10 cardiomyocytes, but did not alter the time to rise in either group (Tables 4 and 5). The overall effect of SR inhibition was a sharp reduction in shortening and Δ[Ca²⁺]_i only in N21 cardiomyocytes, thereby demonstrating their reliance on SR Ca²⁺ cycling, under normoxic conditions, unlike H10 cells.

At the onset of the anoxic perfusion, oxygen rapidly dropped inside the environmental chamber to undetectable levels (<1 min), which had a significant negative inotropic effect on shortening in N21 and H10 cells (Fig. 2A,B), in association with significant reductions in Δ[Ca²⁺]_i (Fig. 2C,D) and pH_i (Fig. 2G,H). However, while shortening in N21 cells remained depressed throughout the anoxic period, it recovered to pre-anoxic levels in H10 cells after 20 min of anoxia (Fig. 2A,B), which was linked to a parallel recovery in Δ[Ca²⁺]_i after 25 min of anoxia (Fig. 2D), despite a persistent and significant acidosis (Fig. 2H). Anoxia also significantly slowed the time to rise in both N21 and H10 cells and the time to half-decay only in H10 cells (Tables 4 and 5). Moreover, the time to half-decay was significantly slower in H10 cells, relative to N21 cells (Table 5). In line with our previous study (Ruhr et al., 2019), anoxic perfusion initially reduced shortening and Δ[Ca²⁺]_i in N21 and H10 cells, yet these variables recovered for H10 cardiomyocytes, despite the absence of oxygen.

Reoxygenation restored shortening to pre-anoxic levels in N21 cells, in association with a recovery in Δ[Ca²⁺]_i and a significant alkalosis (Fig. 2). Nevertheless, diastolic [Ca²⁺]_i remained significantly, although modestly, elevated in N21 and H10 cells relative to pre-anoxic levels, whereas systolic [Ca²⁺]_i was modestly, yet significantly, higher in H10 cardiomyocytes, but not in N21 cells. (Fig. 3). While the times to rise and half-decay were restored to pre-anoxic levels in H10 cells, they were significantly faster in N21 cells (Tables 4 and 5). In contrast to the normoxic period, cardiomyocyte shortening was significantly greater in H10 cells than in N21 cells after reoxygenation, but this was not linked to increased Δ[Ca²⁺]_i, which was significantly higher in N21 cells versus H10 cells (Fig. 2). At the end of the reoxygenation period, diastolic [Ca²⁺]_i was significantly higher in the N21 versus H10 cells (Fig. 3). Even though reoxygenation restored shortening and Δ[Ca²⁺]_i in N21 and H10 cells, H10 shortening rebounded to a significantly higher level than in the N21 group.

The effects of simulated anoxia and reoxygenation on N21 and H10 cardiomyocyte shortening, Δ[Ca²⁺]_i and pH_i were similar between SR-inhibited cells and control cells, but the effects were significantly more pronounced in N21 cells (Fig. 2). Shortening of N21 and H10 cardiomyocytes was significantly smaller in SR-inhibited cells during anoxia and reoxygenation across all time points, compared with that of control cells (Fig. 2A,B). The recovery in shortening of H10 cells in the latter stages of anoxia was abolished by SR inhibition, which was mirrored by significant decreases in Δ[Ca²⁺]_i and pH_i (Fig. 2B,D,H), as well as significantly higher diastolic [Ca²⁺]_i (Fig. 3B). In N21 cells, the reduction in shortening was also associated with significant decreases in Δ[Ca²⁺]_i and pH_i (Fig. 2A,C,G), which was driven by higher diastolic and systolic Ca²⁺ (Fig. 3A,C). SR inhibition during anoxia also significantly hastened the time to rise in N21 cells and the time to half-decay in N21 and H10 cells (Tables 4 and 5, Fig. 1). When Δ[Ca²⁺]_i was normalised to pre-anoxic levels, it was clear that SR inhibition in H10 and N21 cells did not change the shape of the Δ[Ca²⁺]_i response to anoxia and reoxygenation (Fig. 2C,D), it just reduced the amount of Ca²⁺ available for contraction (Fig. 2C,D). SR inhibition had the expected effect, in that it significantly reduced shortening and Δ[Ca²⁺]_i in N21 and H10 cells to the same levels, thus highlighting the reliance of N21 and H10 cells on SR Ca²⁺ mobilisation under anoxic conditions.

During reoxygenation, shortening, Δ[Ca²⁺]_i and pH_i in SR-inhibited N21 and H10 cells recovered to pre-anoxic levels, but shortening and pH_i remained significantly lower than in control cells (Fig. 2). The depression in shortening was not associated with any significant differences in Δ[Ca²⁺]_i, pH_i, diastolic Ca²⁺ or systolic Ca²⁺ (Figs 2 and 3), but was linked to significant decreases in the time to half-decay in both the N21 and H10 cells, relative to control cells (Table 5). The effect of reoxygenation on SR-inhibited cells contrasted the effect on control cells, in that shortening of SR-inhibited N21 and H10 cells rebounded to the same level during reoxygenation (i.e. no significant difference between SR-inhibited N21 and H10 cells).

There were significant positive regressions between shortening and Δ[Ca²⁺]_i for control and SR-inhibited N21 cells (Fig. 4A) and for SR-inhibited H10 cells, but not control H10 cells (Fig. 4B). When controlling for Δ[Ca²⁺]_i (i.e. using Δ[Ca²⁺]_i as the covariate), the GLMM revealed a significant effect of developmental oxygen. That is, for a given Δ[Ca²⁺]_i value, control or SR-inhibited H10 cells shortened significantly more than N21 cells (i.e. control H10 cells shortened more than control N21 cells and SR-inhibited H10 cells shorten more than SR-inhibited N21 cells). Additionally, there was a significant effect of SR inhibition; for a given Δ[Ca²⁺]_i value, control N21 or H10 cells shortened significantly more than their SR-inhibited counterparts (i.e. control N21 cells shorten more than SR-inhibited N21 cells and control H10 cells shorten more than SR-inhibited H10 cells). See Table S3 for full details of these statistical interactions.

An effect of pH_i on shortening was also observed. There were significant positive regressions between shortening and pH_i in control and SR-inhibited N21 cells, but not in H10 cells (Fig. S2). When controlling for pH_i, the GLMM revealed no significant effect of developmental oxygen (i.e. pH_i affected N21 and H10 control cells and N21 and H10 SR-inhibited cells similarly), but demonstrated a significant effect of SR inhibition. Namely, for a given pH_i value, control N21 or H10 cells shortened significantly more than their SR-inhibited counterparts (i.e. control N21 cells shorten more than SR-inhibited N21 cells and control H10 cells shorten more than SR-inhibited H10 cells). See Table S3 for full details of these statistical interactions.

These analyses demonstrate that shortening in H10 cardiomyocytes is more sensitive to [Ca²⁺]_i, leading to greater shortening by H10 versus N21 cells for the same Δ[Ca²⁺]_i value, which is abolished by SR inhibition. The effect of pH_i on cardiomyocyte shortening is also clear; control H10 shortening is unaffected by changes in pH_i, when SR Ca²⁺ stores are available, unlike SR-inhibited H10 shortening.

The developing vertebrate heart is capable of extraordinary phenotypic plasticity in response to oxygen deprivation. Although the extent of cardiac remodelling depends on embryonic stage (Burggren and Mueller, 2015), chronic developmental hypoxia consistently changes cardiovascular function in all vertebrate classes [e.g. mammals (Peyronnet et al., 2002), birds (Villamor et al., 2004), reptiles (Crossley et al., 2022), amphibians (Pan and Burggren, 2013) and fish (Robertson et al., 2014)], as well as invertebrates [e.g. crustaceans (Gopel and Burggren, 2024)]. Nevertheless, only a handful of studies have characterised the cellular and molecular mechanisms that underpin the phenotypes of hypoxia-programmed hearts in ectotherms. The SR of the ectothermic vertebrate heart is a structure that can be modified, providing a source of phenotypic flexibility to elevate heart performance, when faced with an environmental stressor (Shiels and Galli, 2014).

In the present study, we found that chronic developmental hypoxia reduced turtle cardiomyocyte reliance on SR Ca²⁺ cycling under normoxic conditions and increased the sensitivity of the myofilaments to Ca²⁺. We also show that, regardless of developmental oxygen levels, cardiomyocyte shortening during anoxia and reoxygenation is dependent on SR Ca²⁺ cycling. However, hypoxic development increased the reliance on the SR, given that anoxic recovery of shortening and intracellular calcium transients (Δ[Ca²⁺]_i) in the H10 cells is abolished by SR inhibition. These results collectively suggest that chronic developmental hypoxia programmes SR function; specifically, it alters intracellular Ca²⁺ cycling in the ventricular cardiomyocytes of juvenile snapping turtles, which has implications for contractility in normoxia and during anoxia/reoxygenation. Accordingly, we demonstrate that freshwater turtles are dependent on the SR when the heart is challenged by anoxia.

Developmental hypoxia has a potent effect on the snapping-turtle heart. It causes relative cardiac enlargement in embryos (Tate et al., 2015) and, in agreement with prior studies (Galli et al., 2021; Ruhr et al., 2021), we show that this phenotype persists into juvenile life (Table 2). Despite the cardiac enlargement, developmental hypoxia decreased the length of snapping turtle ventricular cardiomyocytes (Table 2), which is coupled to alterations in cellular function, a result we have previously shown (Ruhr et al., 2019). The most notable observation from the present and previous studies (Ruhr et al., 2019) is that H10 cardiomyocytes produce remarkably smaller Δ[Ca²⁺]_i than N21 cells in normoxia, without affecting shortening (Fig. 2A–D). Here, we show that SR inhibition reduced N21 Δ[Ca²⁺]_i to H10 levels, and this was associated with a reduction in systolic Ca²⁺. These findings suggest that developmental hypoxia reduces ventricular cardiomyocyte reliance on SR Ca²⁺ cycling, leading to smaller Δ[Ca²⁺]_i. Despite the lower Δ[Ca²⁺]_i, H10 shortening was similar to that of N21 cells in normoxia, and we previously showed this is driven by enhanced myofilament Ca²⁺ sensitivity (Ruhr et al., 2019) that is likely epigenetically programmed by hypoxia during embryogenesis, which we discuss below (Ruhr et al., 2021).

A hallmark of ventricular myofibrils in ectothermic vertebrates is high Ca²⁺ sensitivity, relative to endotherms, and modulation of Ca²⁺ sensitivity commonly occurs to maintain contractility in response to acute or chronic changes in oxygen, pH, temperature and phosphate (Churcott et al., 1994; Fanter et al., 2017; Klaiman et al., 2014; Ruhr et al., 2019). These studies suggest that an environmental stressor, such as developmental hypoxia, might reorganise EC-coupling pathways in turtle cardiomyocytes, which acts to reduce SR Ca²⁺ reliance and increase myofilament Ca²⁺ sensitivity. This notion is supported by studies on fetal sheep subjected to chronic, high-altitude hypoxia, which alters the abundance of RyRs, lowers activator Ca²⁺ availability and increases myofilament Ca²⁺ sensitivity (Browne et al., 1997; Onishi et al., 2004). Developmental hypoxia could likewise modify EC coupling in snapping turtles, leading to similar shortening of N21 and H10 cardiomyocytes under normoxic conditions, but achieved through different mechanisms.

Plasticity of SR morphology and function has been demonstrated in numerous adult ectothermic vertebrates that are exposed to chronic stress. For instance, mobilisation of SR Ca²⁺ increases contractility at cold temperature in the highly active bluefin tuna (Thunnus thynnus), bigeye tuna (Thunnus obesus), mahi-mahi (Coryphaena hippurus) and swordfish (Xiphias gladius) (Galli et al., 2009a; Shiels et al., 2011). Furthermore, SR Ca²⁺ becomes important for maintaining force production in rainbow trout and Alaskan blackfish, when they are acutely or chronically warmed (Kubly and Stecyk, 2019; Shiels and Farrell, 1997). More pertinent to our results is the reliance on SR Ca²⁺ cycling that is required to tolerate anoxia in the heart of the Amazonian armoured catfish (MacCormack et al., 2003). As with the snapping turtle, inhibition of the SR, during anoxia, substantially reduced heart function in the catfish. Taken together, these earlier studies demonstrate the cardioprotective role the SR plays in the hearts of stress-challenged ectotherms.

The cellular and molecular mechanisms that drive reduced SR Ca²⁺ cycling and enhanced myofilament Ca²⁺ sensitivity in H10 cells are largely unknown, but might involve epigenetic regulation. Myofilament Ca²⁺ sensitivity is susceptible to perturbations in the developmental environment, and developmental programming can modulate the expression of proteins associated with myofilament–Ca²⁺ binding affinity. For example, gene expression patterns of cardiac troponin are modified by DNA methylation in embryonic mice (Xu et al., 2015). Likewise, we have shown developmental hypoxia differentially methylates and enhances the expression of genes encoding cardiac-type troponin T2 (tnnt2), tropomyosin 3 (tpm3) and myosin-light chain 1 (myl1) in juvenile snapping turtles (Ruhr et al., 2021). Mutations in these genes change the sensitivity of the myofilaments to Ca²⁺ in mammalian systems, but through distinct mechanisms (Marston, 2018). Because these proteins modulate Ca²⁺-dependent myofilament contractility, their elevated synthesis could enhance myofibril Ca²⁺ sensitisation in turtles (Fig. 4), as it does in mammals. Moreover, developmental hypoxia upregulates the genes bmp10 (ligand) and bmpr2 (BMP10 receptor) in snapping turtle hearts (Ruhr et al., 2021), which are linked to healthy contractile function in mice (Umbarkar et al., 2019). When rat BMPR2 is not synthesised, it impairs cardiac myofilament–Ca²⁺ binding and decreases contractility (Hautefort et al., 2019).

We can only speculate on the mechanism driving the reduced reliance on SR Ca²⁺ cycling in H10 cardiomyocytes. It is possible that developmental hypoxia alters the gene expression of SR proteins via epigenetic regulation, such as DNA methylation. A reduction in RyR protein abundance would explain the lower levels of systolic Ca²⁺ in H10 versus N21 cardiomyocytes, whereas a reduction in SERCA protein abundance could reduce SR Ca²⁺ content and slow the Δ[Ca²⁺]_i decay. While we did not find differences in the methylation patterns of genes encoding SR proteins, we did find differences in genes that are known to regulate SR function. For instance, the promoter regions of ryr3 (RyR) and atp2a3 (SERCA) are differentially methylated by developmental hypoxia in the snapping turtle heart (Ruhr et al., 2021). Methylation of these genes' promoter regions reduces their RNA expression and protein abundance in the mammalian heart and brain (Kao et al., 2010; Mackrill, 2010; Rabkin and Wong, 2023; Torres and Kerr, 2021). Furthermore, we found that cardiac genes encoding the transcription factors hypoxia-inducible factor 1 (hif1A) and nuclear-factor kappa beta (nfkβ) were upregulated in H10 turtles (Ruhr et al., 2021). In mammals, higher protein abundance of HIF1A and NFKβ suppresses SERCA protein synthesis (Ronkainen et al., 2011; Tsai et al., 2015; Williams et al., 2019). Alternatively, reliance on SR Ca²⁺ cycling could be reduced by an overall reduction in cardiac SR density or a reorganisation of SR–couplon activity, which occurs in response to either chronic oxygen deprivation or metabolic inhibition in mammals (Chantawansri et al., 2008; Guski et al., 1981; Schaper et al., 1977). Lastly, developmental hypoxia might alter SR Ca²⁺ cycling via post-translational modifications, such as oxidation of RyR and calmodulin (Oda et al., 2015).

In contrast to our findings in snapping turtles, our previous work on red-eared sliders suggests cardiomyocyte contractility is entirely dependent on trans-sarcolemmal Ca²⁺ cycling in normoxia, without any contribution from the SR (Galli et al., 2006a,b). It is thought that SR Ca²⁺ cycling is more common in active species of ectotherms (Galli et al., 2009a,b, 2006a; Keen et al., 1992; Shiels et al., 2011, 1999; Shiels and Farrell, 2000), since they require higher levels of cardiac output. In this scenario, the SR provides an additional source of Ca²⁺ that can be released during systole, leading to higher force generation. Considering that North American freshwater turtles are relatively sedentary reptiles and that common snapping turtles and slider turtles share similar ecological niches and life histories (Ultsch, 2006), it is surprising to discover that snapping turtle cardiomyocytes use SR Ca²⁺ stores for contraction and relaxation, unlike slider turtles. However, this result is consistent with earlier findings that ventricular force production in isolated muscle strips is roughly twofold higher in snapping turtles than in slider turtles (Galli et al., 2006a; Smith et al., 2022). The reason for the difference in SR reliance between snapping and slider turtles remains to be investigated, but it might be related to several factors, including laboratory conditions (e.g. age, sex, developmental conditions, husbandry, etc.) or wild behaviours (e.g. foraging, diet, daily movements/migrations, etc.). Nevertheless, previous measures of cardiovascular function are comparable between the two species (e.g., heart rate, stroke volume, cardiac output, ventricular pressure) (Galli et al., 2004; Gatten, 1988; Overgaard et al., 2002; Ruhr et al., 2021; Smith et al., 2022; Stecyk and Farrell, 2007). Therefore, it would be interesting to see whether this is also the case during some form of exercise, such as foraging.

Acute anoxia severely impaired shortening in N21 and H10 cardiomyocytes, which is consistent with in vivo and in vitro studies on snapping turtles (Ruhr et al., 2021; Ruhr et al., 2019), painted turtles [Chrysemys picta (Jackson, 1987)] and slider turtles (Stecyk et al., 2009). As in our previous study (Ruhr et al., 2019), the anoxia-induced reduction in shortening was due to a combination of reduced Δ[Ca²⁺]_i and a profound intracellular acidosis (Fig. 2). Here, we show that the effects of anoxia were comparable between SR-inhibited N21 and H10 cells, but significantly more pronounced than in control N21 and H10 cells. The lower levels of shortening in SR-inhibited cells during anoxia was driven by a reduction in Δ[Ca²⁺]_i that was caused by increased diastolic [Ca²⁺]_i. SR inhibition also exacerbated the intracellular acidosis during anoxia (Fig. 2G,H), which is known to decrease Δ[Ca²⁺]_i by reducing myofilament–Ca²⁺ binding affinity (Orchard and Kentish, 1990). While we did not investigate the mechanism, the lower pH_i of SR-inhibited cells could be driven by the blockade of SERCA-mediated Ca²⁺ uptake, which would potentiate Ca²⁺ efflux and Na⁺ influx via NCX, leading to more Na⁺ being exchanged for protons via the Na⁺/H⁺-exchanger (compared with control cells). Collectively, these results suggest N21 and H10 cardiomyocytes recruit SR Ca²⁺ during anoxia to maintain contractile function.

In line with our previous study (Ruhr et al., 2019), H10 shortening and Δ[Ca²⁺]_i recovered to pre-anoxic levels at the latter stages of anoxia, while N21 shortening remained depressed throughout the anoxic challenge. We previously showed that the recovery of H10 force is partly due to a Ca²⁺-dependent mechanism (increased myofilament Ca²⁺ sensitivity) and a Ca²⁺-independent mechanism (a superior ability to supress ROS) during anoxia (Ruhr et al., 2019). Here, we show that the recovery is also dependent on the SR, because SR inhibition of anoxic H10 cells prevents a rebound in shortening.

Our results are consistent with other studies that show embryonic and juvenile American alligators (Alligator mississippiensis) and snapping turtles exposed to developmental hypoxia have attenuated responses to an acute hypoxic challenge, with blunted responses to heart rate and blood pressure (Crossley and Altimiras, 2005; Crossley et al., 2017, 2022, 2023; Eme et al., 2011; Kam, 1993; Smith et al., 2019). Developmental programming of stress-tolerant phenotypes has also been observed in other ectothermic vertebrates and can occur by either direct (embryonic) or indirect (parental) exposure to environmental stressors. For example, exposing embryonic zebrafish (Danio rerio) to hypoxia lowers their critical oxygen tension (Robertson et al., 2014) and boosts metabolic plasticity (Barrionuevo et al., 2010) at the larval stage, and parental exposure to hypoxia or thermal stress results in offspring with superior hypoxia or thermal tolerance, respectively (Ho and Burggren, 2012; Massey and Dalziel, 2023). Similarly, common killifish (Fundulus heteroclitus) and white suckerfish (Catostomus commersonii) living in chronically polluted waters produce offspring with improved resistance to the pollutants (Bozinovic et al., 2022; Meyer et al., 2003; Munkittrick and Dixon, 1988). Of course, it should be noted that environmental stressors can also lead to maladaptive programming in ectothermic and endothermic vertebrates (Galli et al., 2023; Itani et al., 2018; Wu, 2009), but our results support the growing body of evidence that early exposure to hypoxia can prime the physiology of some ectotherms to better tolerate oxygen-deprived environments in later life.

Although anoxia can initiate catastrophic cell-death cascades in mammals (Bundgaard et al., 2020), reoxygenation is arguably one of the gravest threats to cellular survival, as this causes a surge in ROS production that can trigger the activation of apoptotic pathways (Murphy and Steenbergen, 2008). The mammalian response to reoxygenation stands in stark contrast to that of turtles, which do not suffer any overt cardiac dysfunction (Bundgaard et al., 2018; Galli et al., 2021; Ruhr et al., 2021, 2019; Wasser et al., 1997, 1992). Indeed, upon reoxygenation, all variables in this study returned to pre-anoxic levels in control N21 and H10 cardiomyocytes, apart from pH_i, which rebounded beyond control levels (i.e. a relative alkalosis). SR inhibition did not affect the bimodal responses to reoxygenation in N21 and H10 cardiomyocytes, but did reduce the amount of shortening relative to control cells. This suggests that SR Ca²⁺ cycling also plays a role in maintaining contractility during reoxygenation. Similar to the results during the pre-anoxic period, H10 cardiomyocytes were able to maintain shortening during reoxygenation at higher levels than N21 cardiomyocytes, despite similar levels of Δ[Ca²⁺]_i and pH_i. While increased myofilament Ca²⁺ sensitivity undoubtedly plays a role, other factors might contribute to greater shortening of H10 cardiomyocytes during reoxygenation, such as ROS management. Excessive ROS production can impede cardiac force production, by damaging sarcomere proteins and lowering myofilament Ca²⁺ sensitivity (Sumandea and Steinberg, 2011). Compared with N21 turtles, snapping turtles from hypoxic incubations possess a superior ability to suppress cardiac ROS production (Galli et al., 2021), including the mitigation of oxidative damage to the heart. Indeed, exposure to developmental hypoxia leads to upregulation of the gene encoding peptidylprolyl isomerase A (Ruhr et al., 2021), which is an enzyme that cells produce in response to oxidative stress to suppress mitochondrial ROS production (Satoh et al., 2011). Consequently, turtles from hypoxic incubations might better protect EC-coupling components during reoxygenation, such as the SR and myofibrils. Moreover, better management of mitochondrial Ca²⁺ could help maintain contractile function and avoid triggering apoptotic pathways, such as the mitochondrial permeability transition pore (Morciano et al., 2017). Nevertheless, our results suggest that the turtle heart, irrespective of developmental oxygen levels, can withstand reoxygenation without damaging cellular function and ion homeostasis, if it exploits SR Ca²⁺.

North American freshwater turtles possess the exceptional ability to survive in the complete absence of oxygen for periods lasting several months (Jackson, 2000). We have provided evidence that this anoxia-tolerant phenotype is improved in the heart by early exposure to chronic embryonic hypoxia. On a mechanistic level, our results build on our previous work to demonstrate that the hearts of snapping turtles from hypoxic incubations are less reliant on SR Ca²⁺ stores to contract under normoxic conditions, but recruit them during acute anoxia. Together, SR Ca²⁺ cycling and higher myofilament Ca²⁺ sensitivity work in concert to sustain routine levels of contractile function during anoxia and reoxygenation. These findings represent the first examples of developmental plasticity of SR function and are noteworthy, because the SR is a vital source of activator Ca²⁺ for contraction and is an essential Ca²⁺ sink during relaxation. Therefore, developmental programming of this structure has important implications for EC coupling in the vertebrate heart. From a broader ecological perspective, snapping turtles likely rely on the SR to protect cardiac function during anoxia, which allows them to thrive in their natural habitats.

We thank the staff of the Biological Services Facility, at the University of Manchester (UoM), for their help with turtle husbandry. We thank Prof Andrew Trafford (UoM) for experimental advice and Peter Ceuppens (UoM) with our statistical analyses.