Sample collection

Heparinized (sodium heparin) whole blood and plasma samples were collected from apparently healthy, hereafter referred to as control, loggerhead (whole blood n = 8), green (whole blood n = 12, plasma n = 10) and Kemp's ridley (whole blood n = 14, plasma n = 10) sea turtles captured in pound nets as part of population biology research by the National Marine Fisheries Service in North Carolina in October 2012 and between October and November of 2014. Control loggerhead plasma was not obtained, as this species was sampled directly on small boats, to streamline sampling and avoid potentially poor quality plasma samples. Heparinized whole blood and plasma samples from juvenile, in-water or beach-cast cold stun affected sea turtles of the same three species (loggerhead n = 9, green n = 11, Kemp's ridley n = 6) were collected for diagnosis and treatment purposes prior to rehabilitation of stranded turtles from 2012 to 2016. All samples were collected from the external jugular (dorsal cervical sinus) with a needle appropriate to the size of the individual turtle. Whole blood sample volumes were less than 6 ml or 5 ml kg^–1, whichever was smaller. Aliquots of approximately 100 µl of whole blood and plasma from each individual were reserved for this study.

For cold stun affected turtles, transport time to intake facility varied, but all were sampled immediately upon arrival, with body temperatures less than 15°C (minimum=4.8°C, median=9.7°C, maximum 13.8°C) and prior to receiving any treatment. All loggerheads recovered during rehabilitation and were released. One green turtle died in rehabilitation approximately four months post-stranding. There were no significant findings on necropsy. One Kemp's ridley was found to be near death one day post-stranding and died on the second day post-stranding. All other green and Kemp's ridley turtles recovered during rehabilitation and were released.

We placed all whole blood samples directly into predrilled wells in a block of dry ice (Nagase and Niwa 1964, Krause and Grove 1967). We immediately centrifuged additional whole blood for plasma samples (1000×g for approximately 5 min), separated the plasma, and froze it in the same manner. No hemolysis was noted. We transferred the frozen whole blood and plasma samples to cryogenic vials, which were stored at –80°C until final sample processing.

Sample preparation

Before use, we soaked centrifugal filters (Amicon Ultra 10K 0.5-ml) overnight (approximately 17 h) in ultrapure water and then rinsed each filter four times with 500 µl of fresh ultrapure water at 14 000 g for 20 min at 20°C. We thawed samples at room temperature (20–22°C), mixed 40 µl of sample with 400 µl of Drabkin's reagent (Ricca Chemical Co.) to induce complete and consistent hemolysis, and incubated the mixture at room temperature for 10 min (Niemuth and Stoskopf 2017). For consistency, plasma samples were treated in the same manner with the same solvent. We then filtered the samples at 14 000 g for 20 min and froze the filtrate at –80°C. We lyophilized the samples until dry (about 6 h), sealed each tube with laboratory wax film, and stored them at –80°C.

NMR data collection

We transported the frozen lyophilized samples to the NMR facility at –20°C. We thawed the sealed, lyophilized samples at room temperature. Once thawed, we rehydrated each sample with 70 µl of 100% D₂O containing 0.1 mM of the reference standard trimethylsilyl propanoic acid (TSP), 1 mM formate as a secondary reference standard (Kriat et al. 1992), and 20 mM phosphate buffer. After vortexing, we filtered (Fisherbrand SureOne 10 l, extended, filter, low-retention, universal-fit pipet tips) the rehydrated samples as well as aliquots of the rehydration solution at 3000 g for 2 min. These filtration steps were standard for operation of the microcoil NMR probe. We used the filtered aliquots of rehydration solution as ‘blanks’ (n = 16) run prior to and after each group of samples. We also removed 10 µl from each sample and combined them to create a control pool, cold stun pool and combined pool sample for each species and sample type for quality assurance. We capped all rehydrated samples, blanks and pooled samples and stored them at –20°C until NMR analysis.

We used a Varian Inova 600MHz multinuclear INOVA NMR spectrometer equipped with a Protasis microcoil NMR probe to obtain one-dimensions proton-NMR spectra at 25.08°C with a 1.138s acquisition time. The sweep width of 7193.60Hz acquired 8189 complex points and 4096 transients.

Data analysis

We used ACD labs 12.0 1D NMR processor (Advanced Chemistry Development, Toronto, Ontario, Canada) for spectral processing using our laboratory's standard protocol, as follows. We zero-filled each spectrum to 16 000 points and applied a Fourier transformation. We applied automatic phasing and baseline adjustments, making additional adjustments by hand when the automatic process was inadequate. We then referenced all spectra to the TSP peak at 0 ppm (Hz/MHz).

We identified metabolites using Chenomx NMR Suite 8.43 and the Human Metabolome Database (Wishart et al. 2013). We performed two-sided, two-sample, random sampling permutation tests (α=0.05, R=1,000) to test for significant differences in metabolite concentrations between sample types, species and cold stun status using R ver. 3.5.3 (< www.r-project.org>). One-sided, two-sample, random sampling permutation tests were used only if a significant difference was found with a two-sided test. Heat maps were constructed with the base heatmap function in R.

Overview

Wildlife studies are frequently limited by sample size and availability due to species rarity, body size, legal considerations and other factors, as was the case in this study. However, we demonstrated that high quality one-dimensional proton NMR spectra could be obtained from relatively small volume samples of whole blood or plasma. We were able to identify 18 metabolites involved in a variety of metabolic pathways (Table 1). The use of Drabkin's reagent was necessary to achieve complete and consistent hemolysis, but does limit the output to polar (i.e. aqueous) metabolites. Studies combining Drabkin's reagent with methods and solvents designed for nonpolar (i.e. lipophilic) metabolite extraction are necessary for a more complete metabolic profile, which would include metabolites such as triacylglycerides, cholesterol and fatty acids. While we were able to identify the majority of the peaks present in our spectra, there were sporadic peaks that did not fit with commonly identified metabolites. Additionally, across multiple whole blood spectra there was a very wide complex at approximately 3.1–3.2 ppm. We have not identified this metabolite, but believe it to be erythrocytic in origin, as it appears in relatively greater concentration in whole blood spectra compared to plasma spectra. In future studies, obtaining additional two-dimensional spectra, including those that incorporate other atoms such as carbon, could be useful for identification of unknown metabolites.

We elected to use a commercially available software suite for metabolite quantification and to assist in identification (Chenomx NMR Suite 8.43). This software is relatively user-friendly, has a large library of metabolites, and is widely used in the field (Viant et al. 2008, Wishart 2008, Grimes and O'Connell 2011, Mercier et al. 2011, Psychogios et al. 2011). One of the main benefits is the ability to visually examine the fit of proposed metabolites and make small adjustments if shifts have occurred, such as those due to individual variation in pH. This allowed us to fit metabolites in areas of the spectra where the peaks of multiple metabolites were overlapping, which is in contrast to methods that divide spectra into bins and quantify concentration based on integration, limiting use to areas without metabolite overlap. However, care must be taken to maintain consistency across spectra, such as always fitting metabolites in the same order and likely using measurements conducted by a single person. Quantification is semi-automated at best, and in our experience, the absolute concentrations obtained are often underestimated. While we expect relative differences in concentrations to be valid, the concentrations presented in Table 2 should not be used as reference values.

Table 2.

Median and interquartile range (IQR; quartile 1–3) concentrations (mM) for metabolites identified in the whole blood and plasma of loggerhead Caretta caretta, green Chelonia mydas and Kemp's ridley Lepidochelys kempii sea turtles in North Carolina, USA. Individuals with and without (control) cold stun syndrome were sampled between 2012–2016 and 2012–2014, respectively.

Figure 3.

Heat map of metabolite concentrations from one-dimensional proton-NMR spectra of whole blood and plasma samples from loggerhead Caretta caretta sea turtles in North Carolina, USA (data scaled by column; low=dark purple, high=yellow). Individuals with and without (control) cold stun syndrome were sampled from 2013 to 2016 and in 2014, respectively. Methanol, myo-inositol and trimethylamine N-oxide had significantly higher concentrations in all whole blood samples across species (p<0.001). Blood samples for loggerheads with cold stun syndrome had significantly greater concentrations of 3-hydroxybutyrate (p=0.005) and significantly lower concentrations of leucine (p=0.007) and propylene glycol (p=0.004). Adenosine, 2′inosine-5′monophosphate (IMP), and taurine were not identified in plasma samples.

Comparing whole blood to plasma

We expected, and found, a number of differences between whole blood and plasma metabolic profiles. Taurine was not identified in plasma samples and TMAO was found in significantly greater concentration in whole blood samples. Both metabolites are osmolytes important in maintaining cellular volume (Yancey 2005) and would be expected to be primarily intracellular. Concentrations of myo-inositol were also significantly greater in whole blood samples, which is likely due to protein linkages containing inositol that help protect erythrocytes from lysis (Hooper 1997). Differences in metabolites involved in cellular metabolism were also distinct in whole blood samples, such as the presence of adenosine and IMP, a purine nucleoside and nucleotide, respectively (Murray 2009). Glucose concentrations in green turtle whole blood were significantly lower than in plasma. This has been seen with green turtle samples analyzed by glucometer (Perrault et al. 2017) and may be due to in vitro glycolysis during the thawing and incubation steps prior to hemolysis in sample preparation (Stockham and Scott 2008), as glucose concentration in loggerhead samples have been shown to decrease over time (Eisenhawer et al. 2007). Lastly, succinate, which is generated in mitochondria via the carboxylic acid cycle (Murray 2009), was significantly greater in green turtle whole blood compared to plasma.

Figure 4.

Heat map of metabolite concentrations from one-dimensional proton-NMR spectra of whole blood and plasma samples from green Chelonia mydas sea turtles in North Carolina, USA (data scaled by column; low=dark purple, high=yellow). Individuals with and without (control) cold stun syndrome were sampled between 2012–2015 and 2012–2014, respectively. Methanol, myo-inositol and trimethylamine N-oxide had significantly higher concentrations in all whole blood samples across species (p<0.001). Glucose and isobutyrate concentrations were significantly less in green turtle whole blood samples versus plasma samples (p<0.001 and 0.018, respectively), whereas succinate concentrations were significantly greater in whole blood samples (p<0.001). In both whole blood and plasma samples, 3-hydroxybutyrate and acetate was significantly lower in individuals with cold stun syndrome (p=0.005 and 0.003 and p=0.001 and<0.001, respectively). Creatine and glycerol concentrations were significantly greater in plasma samples from individuals with cold stun syndrome (p=0.006 and 0.012, respectively). Isoleucine, leucine and trimethylamine N-oxide plasma concentrations were lower in individuals with cold stun (p=0.004, 0.002 and<0.001, respectively). Adenosine, 2′inosine-5′monophosphate (IMP) and taurine were not identified in plasma samples.

Differences between the whole blood and plasma concentrations of isobutyrate and methanol from control and cold stun affected groups were not expected. Isobutyrate can be formed by the oxidation of valine (Smith and Macfarlane 1997) and has been identified in the plasma of women and in human fecal extracts (Le Gall et al. 2011, Bahado-Singh et al. 2012, Shao et al. 2016). Gastrointestinal bacteria produce isobutyrate (Le Gall et al. 2011). Methanol was identified as a contaminant from our blank samples, but concentrations were greater than the background contamination. Methanol has also been identified as a human fecal metabolite (Ahmed et al. 2016, Shao et al. 2016) and gastrointestinal flora easily trap methanol (Garner et al. 2007). Methanol interacts with lipid bilayers (Sonmez et al. 2013), which may explain the significantly greater concentration in whole blood, but we do not have an explanation for why isobutyrate concentrations would be significantly less in whole blood versus plasma.

Figure 5.

Heat map of metabolite concentrations from one-dimensional proton-NMR spectra of whole blood and plasma samples from Kemp's ridley Lepidochelys kempii sea turtles in North Carolina, USA (data scaled by column; low = dark purple, high = yellow). Individuals with and without (control) cold stun syndrome were sampled between 2013–2015 and 2012–2014, respectively. Methanol, myo-inositol and trimethylamine N-oxide had significantly higher concentrations in all whole blood samples across species (p < 0.001). Isobutyrate concentrations were significantly less in Kemp's ridley whole blood samples versus plasma samples (p = 0.014). Acetate and taurine concentrations in whole blood and dimethyl sulfone concentrations in plasma were significantly lower in cold stun affected individuals (p = 0.002, 0.015 and 0.01, respectively). Plasma lactate, propylene glycol and succinate concentrations were significantly greater in individuals with cold stun syndrome (p = 0.027, 0.033 and 0.003, respectively). Adenosine, 2′inosine-5′monophosphate (IMP), and taurine were not identified in plasma samples.

Comparing cold stuns to controls and species differences

Sea turtles affected by cold stun syndrome become lethargic and cease feeding. We identified differences in several metabolites that are likely secondary to anorexia. Isoleucine and leucine are nutritionally essential in humans (Murray 2009) and both were significantly decreased in plasma samples from cold stunned green turtles. Leucine was also significantly decreased in whole blood from cold stunned loggerhead and green turtles. Dimethyl sulfone was significantly decreased in cold stunned Kemp's ridley plasma, which is likely a result of a combination of decreased dietary intake, endogenous metabolism and bacterial metabolism (Engelke et al. 2005). A significantly lower concentration of TMAO was found in cold stunned green turtle plasma samples, which may be related to decreased gastrointestinal microbial metabolism and decreased intake of choline and carnitine due to anorexia (Wang et al. 2011, Koeth et al. 2013). However, decreased TMAO concentrations may also reflect osmoregulatory disturbance (Yancey 2005). Significantly increased glycerol concentrations in cold stunned green turtle plasma are likely the result of triacylglycerol metabolism during anorexia (Rosenthal and Glew 2009). However, glycerol has also been noted as a cytoprotectant involved in redox balance and response to hypoxia (Yancey 2005). Additionally, increased glycerol concentrations may afford some cryoprotection, as seen in other species, including the painted turtle Chrysemys picta (Storey and Storey 1983, Kukal et al. 1988, Storey et al. 1988).

Table 3.

Summary of statistically significant diferences identified in the whole blood and plasma of loggerhead Caretta caretta, green Chelonia mydas and Kemp's ridley Lepidochelys kempii sea turtles in North Carolina, USA. (A) Statistically significant changes in metabolite concentrations in cold stunned versus control sea turtles. Individuals with and without cold stun syndrome were sampled between 2012–2016 and 2012–2014, respectively. (B) Statistically significant differences in metabolite concentrations in whole blood versus plasma samples of sea turtles.

We also identified differences in multiple metabolites related to exertion and anaerobic metabolism. Succinate was significantly increased in plasma of cold stunned Kemp's ridleys and is generated via ATP production in the carboxylic acid cycle (Murray 2009). Accumulation of succinate may also be an indicator of ischemia and has been associated with reperfusion injury (Chouchani et al. 2014, Pell et al. 2016). Similarly, increased consumption of acetate into the carboxylic acid cycle to attempt to generate energy aerobically (Rosenthal and Glew 2009) initially explains the significantly decreased acetate concentrations seen in cold stunned green turtle whole blood and plasma samples and Kemp's ridley whole blood samples. Lactate concentrations in cold stunned Kemp's ridley plasma were significantly increased, which has been found in cold stunned sea turtles and presumed secondary to hypoventilation, anaerobic metabolism and abnormal perfusion (Innis et al. 2007, Keller et al. 2012). Lactate is produced under hypoxic conditions and used by the liver for gluconeogenesis (Murray 2009). Creatine, a product of using creatine phosphate to generate ATP, is released from ischemically damaged tissue (Wyss and Kaddurah-Daouk 2000) and concentrations were significantly greater in the plasma of cold stunned green turtles. Additionally, creatine may help inhibit platelet aggregation and in humans with acute myocardial infarction, individuals with greater creatine concentrations had a lower incidence of death and complications up to two years later (Delanghe et al. 1991, Wyss and Kaddurah-Daouk 2000). Green and Kemp's ridley turtles with cold stun syndrome have lower clot strength and take longer to form clots compared to controls (Barratclough et al. 2019). Finally, while not a significant difference (p = 0.065 in Kemp's whole blood samples), adenosine trended toward greater concentrations in turtles with cold stun syndrome and has also been identified as a signal of hypoxia (Bickler and Buck 2007).

Changes in taurine and propylene glycol, in addition to glycerol, may potentially be a response to cold temperatures. Raheem (1980) found a marked increase in taurine in hibernating Varanus griseus and suggested that it may participate in central nervous system depression. However, we found taurine at significantly lower concentration in whole blood samples from Kemp's ridleys affected by cold stun. This may be due to changes in osmoregulation or consumption as an antioxidant (Yancey 2005) as efflux of taurine has been seen in avian erythrocytes before other more osmotically important amino acids presumably as a protective measure (Shihabi et al. 1989). Propylene glycol concentrations were significantly lower in whole blood samples from cold stunned loggerheads, but significantly greater in plasma samples from cold stunned Kemp's ridley turtles. This metabolite may be of bacterial or yeast origin (Enebo 1954, Suzuki and Onishi 1968), made biogenically by the turtles (Saxena et al. 2010), or an environmental contaminant. If a true metabolite, the turtles could be using it to provide energy and protect against ketosis through gluconeogenesis, the carboxylic acid cycle and glycogenesis (Shull and Miller 1960, Ruddick 1972, Morshed et al. 1988, Nielsen and Ingvartsen 2004) or potentially as a cryoprotectant (Niemuth et al. 2018). We do not have an explanation for the observed species difference. Species and seasonal changes in diet and foraging ground could contribute to the variance. Loggerheads are larger and strand later in the season than Kemp's ridleys (Still et al. 2005). The loggerheads could have depleted their propylene glycol as a result of being more chronically cold. Alternatively, the Kemp's ridleys early in the season could be stranding because they are physiologically disadvantaged in some way.

Loggerheads with cold stun syndrome were also found to have significantly greater concentrations of 3-hydroxybutyr-ate in whole blood samples versus green turtle whole blood and plasma samples, which had lower concentration in cold stun affected animals. Three-hydroxybutyrate is a ketone body produced secondary to anorexia, which allows the brain to function without glucose thereby decreasing the need for gluconeogenesis (Rosenthal and Glew 2009). As previously mentioned, loggerheads have greater body mass and strand later in the season, which may mean that they have been anorexic longer and have had greater time to convert to ketosis. Green turtles strand earlier and may have consumed their normal circulating ketones prior to increasing ketogenesis.

In the single Kemp's ridley that died, concentrations of both whole blood and plasma 3-hydroxybutyrate, acetate, dimethyl sulfone, glucose, myo-inositol, creatine and lactate were greater than the calculated third quartile value (Table 2) concentrations for its cohort with the creatine and lactate concentrations being much greater (greater than two-fold and 1.5-fold the median concentrations, respectively). Blood concentrations of adenosine and plasma concentrations of glycerol and methanol were also much greater than the third quartile value. The plasma concentration of succinate was below the calculated first quartile value (Table 2). These more severe perturbations suggest that this animal had been anorexic for some time, had hypoxia, was using anaerobic metabolism and had experienced acidosis, muscle damage and ischemia.

Sources of variation

We acknowledge that, as is the case in any study sampling wildlife, there are numerous sources of variation. The control cohort for this study was sampled in October and November, earlier in the season than most cold stun strandings occur in North Carolina. It was not possible to sample non-stranded, free-ranging turtles during cold stun season, so we do not know if those turtles that do strand are representative. Feeding status is also an unknown variable for free-ranging sea turtles and can result in some changes to biochemical parameters (Anderson et al. 2011a). The individuals sampled also varied somewhat in size and, except for the control green and Kemp's ridley turtles, were of unknown sex. Both body size and sex can affect blood parameters (Bolten and Bjorndal 1992). Additionally, variation in hydration status and red blood cell count could influence metabolite concentrations. No obvious relationship was apparent in our preliminary assessments, but it may be that a complex combination of parameters might provide some additional insights. Lastly, it took several years to accumulate the samples for this study and despite storage at ultralow temperature some metabolite degradation is likely. To our knowledge, there is no comprehensive review available of metabolite stability during storage in whole blood or plasma, but changes of varying degree have been documented in samples stored for clinical biochemistry panels for a variety of species (Thoresen et al. 1995, Cray et al. 2009, Brinc et al. 2012, Eshar et al. 2018).