Tissue-dependent variation of hydrogen sulfide homeostasis in anoxic freshwater turtles

Hydrogen sulfide (H₂S) has emerged as an important signaling molecule belonging to the family of gasotransmitters together with nitric oxide (NO) and carbon monoxide (Kolluru et al., 2017; Olson and Straub, 2016; Wang, 2002). First discovered as a modulator in the brain (Abe and Kimura, 1996), H₂S was later found to control other key physiological functions, such as vascular tone and cytoprotection (Elrod et al., 2007; Yang et al., 2008; Zhao et al., 2012). H₂S conveys its biological activity by binding reversibly to ferric heme proteins (Jensen and Fago, 2018; Pietri et al., 2011) and by S-sulfhydration of thiol groups (Mustafa et al., 2009; Paul and Snyder, 2012), but also via other sulfur-containing compounds, such as persulfides (R-S-SH), polysulfides (R-S_n-R, n>2) and thiosulfate (S₂O₃²⁻) (Kimura, 2017; Olson et al., 2013; Shen et al., 2017). Endogenous H₂S is produced in the cytosol enzymatically by cystathionine β-synthase (CBS) and cystathionine γ-lyase (CSE), using (homo)-cysteine and cystathionine as main substrates. In the mitochondria, 3-mercaptopyruvate sulfurtransferase is the primary enzyme producing H₂S from mercaptopyruvate assisted by cysteine aminotransferase, though this enzyme machinery is also present in the cytosol (Nagahara et al., 1998). In addition to enzymatically produced H₂S, free H₂S can also be regenerated by reduction of bound stores of persulfides, polysulfides and thiosulfate, collectively representing the bound sulfane sulfur (BSS) fraction (Koike and Ogasawara, 2016; Shen et al., 2012, 2015). During low O₂ conditions, H₂S can be regenerated from thiosulfate (Olson et al., 2013), as part of a potential oxygen-sensing mechanism (Olson, 2013), as well as from persulfides and polysulfides (Ishigami et al., 2009; Kimura, 2014a).

The discovery that exposure to exogenous H₂S gas triggers reversible suspended animation in mice (Blackstone et al., 2005) has fostered the idea that endogenous H₂S may control the biological metabolic suppression of hibernating animals. To fit with this proposed in vivo role, H₂S is an in vitro reversible inhibitor of cytochrome c oxidase (Collman et al., 2009; Cooper and Brown, 2008; Pietri et al., 2011), thus being able to modulate mitochondrial oxygen consumption. In a previous analysis of blood samples from free-ranging summer active and winter-hibernating brown bears (Revsbech et al., 2014), we detected changes in free H₂S and BSS during hibernation, consistent with the generation of free H₂S from circulating storage pools that may contribute to the suppression of mitochondrial respiration in tissues. Moreover, the substrate cysteine shifted towards the synthesis of RBC glutathione (GSH), which increased dramatically (Revsbech et al., 2014).

In hibernating bears, O₂ consumption rate falls to ∼25% of basal levels (Tøien et al., 2011); in comparison, some turtle species are capable of overwintering in complete anoxia, sustaining even more pronounced metabolic suppression, down to 5–10% of basal levels (Bickler and Buck, 2007; Ultsch, 2006). Among these turtle species, the red-eared slider, Trachemys scripta elegans (Wied-Neuwied 1839), is among the most extreme vertebrates as it can survive for weeks in anoxia at low temperatures, relying only on glycolysis for energy production (Bundgaard et al., 2019; Ultsch, 1989, 2006; Warren et al., 2006). Thus, we hypothesized that T. scripta elegans endogenous levels of free H₂S and BSS in various tissues should change more markedly upon cold acclimation and anoxia than found in the brown bears (Revsbech et al., 2014). Furthermore, NO metabolites are known to increase dramatically in the blood (Jacobsen et al., 2012) and tissues (Jensen et al., 2014) of anoxia-acclimated T. scripta elegans and may contribute further to the strong metabolic suppression of this species (Fago and Jensen, 2015), although not necessarily to cytoprotection against oxidative stress at reoxygenation, at least in the heart (Bundgaard et al., 2018).

All chemicals were obtained from Sigma-Aldrich unless otherwise stated.

Adult red-eared sliders of both sex (Trachemys scripta elegans) kept in aquaria at the animal care facility at Zoophysiology, Aarhus University, were used for experiments. While fasted, the turtles were progressively acclimated to low temperature (5°C) followed by randomized exposure to either normoxic (N=5, 0.66±0.13 kg) (termed ‘cold normoxia’) or anoxic conditions (N=4, 0.64±0.07 kg) (termed ‘cold anoxia’) for 9 days, according to a previously described protocol (Bundgaard et al., 2018; Bundgaard et al., 2019; Jensen et al., 2014). A control group (N=5, 0.61±0.04 kg) of non-fasting turtles taken directly from the aquaria at 25°C was also included in the study (termed ‘warm normoxia’). All procedures were performed in accordance with regulations of animal care and experimentation in Denmark under permit 2015-15-0201-00544.

Turtles were killed as previously described (Bundgaard et al., 2018). In brief, an overdose of pentobarbital (50 mg kg⁻¹) was given via the supravertebral venous sinus. When reflexes were lost, the head was cut off and an opening was made in the plastron. After removing the heart (used for another study), a blood sample was quickly collected in heparin-filled syringes from the thoracic cavity and centrifuged (2000 g, 3 min) to separate RBCs from plasma; 250 µl RBCs were diluted in 800 µl 100 mmol l⁻¹ Tris-HCl buffer (pH 9.5, 0.1 mmol l⁻¹ DTPA) and snap-frozen in liquid N₂. Tissue samples (∼150 mg), including liver, brain, lung and kidney) were harvested and conserved in 800 µl of 100 mmol l⁻¹ Tris-HCl buffer (pH 9.5, 0.1 mmol l⁻¹ DTPA) followed by snap-freezing in liquid N₂. All samples were weighed to calculate the exact dilution factor. The whole procedure was completed within 20 min. Samples were shipped on dry ice to Louisiana State University Health Sciences Center, Shreveport, LA, USA, where H₂S metabolite analysis took place. Additional tissue samples were collected and stored at −80°C for enzyme activity measurements at Aarhus University.

Bioavailable biochemical forms of H₂S including free H₂S and BSS (including persulfides, polysulfides and thiosulfate) were measured by a fluorescent monobromobime (MBB) assay coupled with RP-HPLC as previously described (Shen et al., 2011, 2012, 2015). This method has been extensively refined to determine specific incubation pH conditions and reaction times (<30 min) of MBB to derivatize biological sulfide and its various forms with minimal interference or artifact (Shen et al., 2011). The MBB assay reported here was performed under exacting specifications to ensure minimal chemical interference and maintain the chemical identity and abundance of various sulfide species, as previously confirmed in a separate publication using LC/MS non-radioisotopic standardization (Shen et al., 2014). In this method, the tissue homogenates were prepared in alkaline 100 mmol l⁻¹ Tris-HCl buffer (to convert all free H₂S gas to the HS⁻ anionic form). HS⁻ was then derivatized by incubation with excess MBB (10 mmol l⁻¹ in CH₃CN) for 30 min; 1 mmol l⁻¹ TCEP [tris(2-carboxyethyl)phosphine hydrochloride] was used to liberate total stored sulfide (Shen et al., 2012, 2015). The whole procedure was done at room temperature at 1% O₂ in a hypoxic chamber in the dark. All measurements were obtained in technical triplicate. MBB was separated from the product, sulfide dibimane (SDB), by a RP-HPLC equipped with a fluorescence detector (λ_ex: 390 nm and λ_em: 475 nm). Retention time was 16.5 and 17.6 min, respectively. The amount of SDB in each sample was quantified from a standard curve. H₂S is a weak acid in solution (pK_a1 ∼6.8, 37°C) (Li and Lancaster, 2013), and under physiological conditions, ∼80% of total sulfide is present as hydrosulfide anion. When using the MBB method, the equilibrium H₂S⇌HS⁻+H⁺ is right-shifted as a result of alkaline conditions, and thus the method measures both H₂S and HS⁻. For simplicity, the two species are here collectively referred to as H₂S.

The activity of H₂S-producing enzymes was measured in warm normoxic turtles. Tissue homogenates were prepared in 100 mmol l⁻¹ potassium phosphate buffer+1 mmol l⁻¹ EDTA, pH 7.4, mixed with 90 mmol l⁻¹ pyridoxal 5′-phosphate and incubated in a 1 ml glass chamber with continuous stirring (500 rpm) at room temperature. A final concentration of 250 mmol l⁻¹ cysteine was added to initiate the reaction. Production of H₂S over time was measured with an amperometric H₂S-specific microsensor (Unisense A/S, Aarhus, Denmark), calibrated with a freshly made Na₂S stock solution of known concentration (r²=0.99). Activity was expressed as nanomoles of H₂S produced per milligram total protein per hour. Protein content was measured by a colorimetric assay using bovine serum albumin as a standard. Protein concentration in RBCs was measured as Hb concentration determined from the oxyHb extinction coefficients at 541 nm of 13.8 l mmol⁻¹ cm⁻¹, 577 nm of 14.6 l mmol⁻¹ cm⁻¹ and 415 nm of 125 l mmol⁻¹ cm⁻¹.

Plasma lactate concentrations were measured in a subsample of normoxic and anoxic turtles by a colorimetric l-lactate assay kit (ab65331, Abcam, Denmark).

Statistical analyses were done in Prism (GraphPad Software Inc., La Jolla, CA, USA). Shapiro–Wilk normality test was applied to determine whether data were normally distributed. Significant differences between warm normoxic turtles, cold normoxic turtles and cold anoxic turtles were assessed by one-way ANOVA with Dunnett's multiple comparisons test (normally distributed) and the few non-normally distributed datasets were assessed by non-parametric Kruskal–Wallis test with Dunn's multiple comparisons test (Table S1). The significance level was P<0.05. All data are reported as means±s.e.m.

To examine for changes in H₂S metabolism during the extreme metabolic suppression of cold-acclimated and anoxic turtles, we kept turtles at 25°C under normoxia and at 5°C under either normoxia or anoxia for 9 days, as described previously (Bundgaard et al., 2019). We then measured in vivo pools of free H₂S and BSS in kidney, brain, liver, lung and RBC samples using the MBB method (Shen et al., 2012, 2015). An increase in plasma lactate from 0.21±0.04 mmol l⁻¹ (cold normoxic turtles) to 19.4±9.7 mmol l⁻¹ (cold anoxic turtles) confirmed that glycolysis was upregulated and that turtles were anoxic. Physiological basal levels (expressed in µmol l⁻¹) of free H₂S and BSS in kidney, brain, liver and lung homogenates of freshwater turtle are shown in Fig. 1. An overview of H₂S levels normalized to total protein content is given in Fig. S1. To our knowledge, this is the first time that H₂S metabolites have been measured in a reptile.

Our analysis showed that the levels of free H₂S in tissues of warm normoxic turtles were highest in the kidneys (9.3±0.4 µmol l⁻¹) and lowest in the brain, liver and lung (0.9–2.5 µmol l⁻¹) (Fig. 1A). Upon cold acclimation of turtles to 5°C, these values did not change, except in the liver, where free H₂S decreased significantly (Fig. 1A). No increase was observed in free H₂S after acclimation of turtles to anoxia for 9 days in any of these tissues (Fig. 1A). This finding suggests that free H₂S is not involved in overall metabolic suppression, although we cannot exclude that the lack of O₂ would enhance the inhibitory effect of H₂S on cytochrome c oxidase activity. O₂ and H₂S levels are inversely linked (Olson and Straub, 2016), as a result of the oxidative degradation of H₂S (Hildebrandt and Grieshaber, 2008), and H₂S lifetime increases when O₂ is limiting.

Although not in our experimental setup, the natural environment of overwintering anoxic turtles is presumably sulfide rich, which could add to the in vivo H₂S levels. A recent study on amphibian fish inhabiting sulfide-rich mangroves found no change in aquatic H₂S sensitivity between wild-caught and laboratory-reared fish in terms of behavior (Cochrane et al., 2019); however, further studies are needed on the potential adaptations to environmental H₂S in vertebrates.

That free H₂S is high in the kidneys (Fig. 1A) is not unique to turtles, as mammals also exhibit abundant H₂S levels (Shen et al., 2013). The enzyme d-amino acid oxidase constitutes an additional enzymatic pathway for H₂S production and is highly expressed in mammalian kidneys (Shibuya et al., 2013). In mammals, endogenous H₂S is proposed to be important for normal kidney function (Lobb et al., 2015) and H₂S has been shown to increase glomerular filtration rate (Xia et al., 2009) and salt excretion by inhibition of Na⁺/K⁺-ATPase (Ge et al., 2014), and to suppress renin release (Cao and Bian, 2016). Given that H₂S is an ancient vasoregulatory molecule across multiple vertebrate clades (Dombkowski et al., 2004), our data indicate that H₂S could also have a similar regulatory role in the renal function of turtles. Protein content in the kidney did not change during acclimation, but free H₂S normalized to total protein decreased significantly in cold anoxic turtles compared with warm normoxic ones (Fig. S1). Compared with other tissues such as brain and liver, the kidneys of anoxic turtles exhibit the highest reduction in blood flow (Stecyk et al., 2004), suggesting H₂S-mediated regulation of vascular tone in this organ.

BSS levels were overall similar in tissues of all three groups of turtles, except for the brain, where BSS decreased significantly when cold­-acclimated turtles were exposed to anoxia (Fig. 1B) without increasing free H₂S (Fig. 1A). In mammals, H₂S and H₂S-derived polysulfides in the brain are implicated in neuroprotection and have been shown to regulate the activity of essential receptors such as N-methyl-d-aspartate receptor (NMADR) and transient receptor potential ankyrin (TRPA) by persulfhydration (Abe and Kimura, 1996; Kimura, 2014b; Kimura et al., 2013; Li et al., 2017). In the turtle brain, NMDAR is strongly inhibited during anoxia by phosphorylation (Bickler and Buck, 2007) and cysteine S-nitrosation (Takahashi et al., 2007) to limit neuronal activity and protect the turtle brain from damage and cell death (Bickler and Buck, 2007). One can speculate that an increased persulfhydration may enhance the activity of NMDAR (Abe and Kimura, 1996) and other brain receptors at low temperatures in normoxia (Fig. 1B) and promote subsequent inhibition by increased S-nitrosation in hypoxia and anoxia (Fago and Jensen, 2015; Jensen et al., 2014). There is emerging evidence of complex H₂S/NO cross-talk of protein cysteine redox modifications that remains to be investigated in detail (Cortese-Krott et al., 2015, 2017; Hosoki et al., 1997; Miyamoto et al., 2017). Besides possible interactions including production of polysulfides (Miyamoto et al., 2017), the two gasotransmitters have been suggested to have several overlapping functions (Kolluru et al., 2013).

The level of free H₂S in RBCs was 10.5±0.8 µmol l⁻¹ in warm normoxic turtles (Fig. 2A), which is not far from the value reported for human RBCs measured using a MBB-based method (3.8±1.0 µmol l⁻¹) (Tan et al., 2017). RBC free H₂S increased 2-fold in cold-acclimated turtles in normoxia but did not change further in anoxia (Fig. 2A), as for the other tissues (Fig. 1A). BSS concentrations in RBCs were remarkably high in warm normoxic turtles (68±3 µmol l⁻¹) compared with those of other tissues and increased even more in cold-acclimated normoxic and anoxic turtles (121±8 and 91±9 µmol l⁻¹, respectively) (Fig. 2B). These high values of BSS in RBCs appear to be consistent with the high content (∼24 mmol l⁻¹) of total reactive thiols in T. scripta elegans RBCs, including those of hemoglobin (Hb) and GSH (Damsgaard et al., 2013; Jacobsen et al., 2012; Petersen et al., 2018). Thiols can potentially be partly sulfhydrated (i.e. R-S-SH) and contribute to the BSS pool, and in T. scripta elegans Hb, we have previously identified bound persulfide and polysulfides by mass spectrometry (Petersen et al., 2018). Considering that protein sulfhydration is a pervasive modification (Mishanina et al., 2015) estimated to be up to 10–25% for certain proteins (Mustafa et al., 2009), the values of ∼70–120 µmol l⁻¹ detected here indicate a maximum of 0.25–0.5% of total protein thiol sulfhydration in turtle RBCs, which is not unrealistic. Sulfur extraction from GSH is expected to be only 0.01% under the conditions used in the MBB methods (Montoya et al., 2015), and would therefore not contribute much. In addition to protein sulfhydration, some of the BSS pool may originate from H₂S coordinated to the ferric heme in metHb (Jensen and Fago, 2018). MetHb in freshwater turtles is ∼1.0% of total Hb (Maginniss et al., 1983), which corresponds to ∼20 µmol l⁻¹ ferric heme capacity available for H₂S binding and release (Jensen and Fago, 2018). Therefore, these data suggest that T. scripta RBCs may act as a circulating storage pool of bound sulfide, that can either take up or regenerate free H₂S by varying cellular redox conditions, pH and body temperature (Fig. 2).

We added cysteine to tissue homogenates to detect the generation of H₂S by the combined enzymatic activity of CBS and CSE (Vicente et al., 2016) by using a H₂S-specific microsensor (Fig. 3A). The enzymatic activity of H₂S production was highest in the kidneys, which aligns with the high concentration of free H₂S in this tissue (Fig. 3B). RBCs, brain, liver and kidney all showed comparable enzymatic activity (Fig. 3B). This finding implies that the high levels of free H₂S and BSS detected in RBCs (Fig. 2) are not due to a particularly high capacity for enzymatic production, and further supports a role for RBCs as a circulating H₂S reservoir.

H₂S and NO are involved in numerous physiological functions. NO upregulation may contribute to metabolic suppression in T. scripta (Fago and Jensen, 2015; Jensen et al., 2014), but apparently not in hibernating brown bears (Revsbech et al., 2014). In contrast, free H₂S and BSS decrease in the blood of hibernating bears (Revsbech et al., 2014), whereas in turtle blood, free H₂S and BSS increase during cold acclimation, but show no further change during anoxia (Fig. 2). This suggests that NO may be the major gasotransmitter regulating physiological responses in turtles during anoxia. But, H₂S could play a role in cytoprotection, as suggested by the enlarged BSS storage pool of cold-acclimated turtles (Fig. 2B).

In conclusion, we observed tissue-specific changes in endogenous levels of free H₂S and BSS upon cold acclimation and 9 days of anoxia in T. scripta elegans liver and brain and especially RBCs. In warm normoxic turtles, the levels of free H₂S were highest in RBCs and kidney and lowest in brain, liver and lung, which matched the observed enzymatic production rates of H₂S. BSS levels were 100-fold higher in RBCs than in other tissues, indicating that turtle RBCs may function as a circulating reservoir for bioactive H₂S bound to Hb's ferric heme and thiols as persulfide and polysulfides. Besides changes in brain BSS, free H₂S and BSS levels increased during cold acclimation only in the RBCs, where they remained high during anoxia. Interestingly, levels of NO metabolites have been found to increase dramatically in the blood (Jacobsen et al., 2012) of anoxia-acclimated T. scripta. In contrast, in plasma of hibernating bears, free H₂S and BSS decrease, while nitrite did not change significantly (Revsbech et al., 2014). These differences suggest distinct effects on aerobic or anaerobic metabolic suppression mediated by NO and H₂S, respectively, that are yet to be fully unraveled.

We thank Amanda Bundgaard for assisting with turtle acclimation, Inge Grønvall Revsbech for support during tissue sampling and Elin E. Pedersen for technical assistance with lactate measurements.