The brains of six African mole-rat species show divergent responses to hypoxia

Metabolism in the adult brain is almost completely aerobic and the brain receives a large percentage of bodily oxygen from the blood supply (∼12%), so changes in oxygen availability can have a substantial effect on brain function (Buck and Pamenter, 2018; Nortje and Gupta, 2006; Rolett et al., 2000). The critical oxygen tension is the partial pressure of oxygen (P_O2) where O₂ consumption matches its delivery to the tissue, and it varies for each species, but it can also vary with angiogenesis, stress, exercise and hypoxia (Dunn et al., 2000). Hypoxia occurs when tissues do not receive an adequate oxygen supply and is a stress associated with some diseases such as cancer, stroke and cardiac arrest (Buck and Pamenter, 2018). Below the critical oxygen tension, a decrease in ATP is accompanied by a rise in ADP and AMP (Ashcroft and Ashcroft, 1990; Folbergrová et al., 1990). Hypoxia/ischemia cause reductions in ATP levels within a few minutes which can have an impact on cellular health (Ashcroft and Ashcroft, 1990; Buck and Pamenter, 2018). Hypoxia also creates a favorable environment for the generation of free radicals and their leakage from the mitochondria into the cytosol and extracellular space, which has the potential of causing significant cell damage to membranes, proteins, lipids and other cellular components (Smith et al., 2017; Terraneo et al., 2017). Free radicals are also produced upon reoxygenation. However, several mole-rat species such as Spalax galili and Heterocephalus glaber are resistant to decreases in oxygen tension (Larson and Park, 2009; Schülke et al., 2012; Shams et al., 2004), and other species of African mole-rats (Bathyergidae) may also show hypoxia tolerance (Ivy et al., 2020).

African mole-rats spend the majority of their life in subterranean environments where oxygen concentration may be as low as 15% (McNab, 1966). They have developed mechanisms to adapt to their hypoxic, hypercapnic and high ammonia tunnel systems where they are able to tolerate extremely low oxygen tensions for several hours without any significant cellular damage (Johansen et al., 1976; Larson and Park, 2009). Some species of African mole-rats, like the naked mole-rat (H. glaber), have been shown to tolerate P_O2 of 3% or lower, for several hours (Pamenter et al., 2018; Park et al., 2017), and respond with a decrease in oxygen consumption of 85% and a reduction in neuronal electron transport chain by 50% relative to normoxic controls (Pamenter et al., 2015, 2018).

This study focused on the hypoxia tolerance of six species of African mole-rats: the Cape dune mole-rat [Bathyergus suillus (Schreber 1782)], three subspecies of the common mole-rat [Cryptomys hottentotus hottentotus (Lesson 1826), Cryptomys hottentotus mahali (Roberts 1913) and Cryptomys hottentotus pretoriae (Roberts 1913)], the Cape mole-rat [Georychus capensis (Pallas 1778)] and the naked mole-rat (H. glaber Rüppell 1842). Phylogenetic analyses have concluded that H. glaber is the ancestral species of the Bathyergidae family, whereas mole-rats of the genus Cryptomys evolved more recently (∼6–7 mya) (Fig. 1) (Faulkes et al., 2004). Naked mole-rats were the first mammalian species identified as being eusocial (having a social structure with a queen and workers) (Jarvis, 1981). They burrow in very dry, hard soil, and because of their characteristic lack of fur, are poor endotherms. Living in large eusocial colonies would help the naked mole-rat thermoregulate, but it would also contribute to increased O₂ consumption and higher CO₂ production in closed-off burrows, which theoretically could trigger a higher hypoxia tolerance in the naked mole-rat (Tregub et al., 2013). By contrast, the closely related B. suillus and G. capensis species are strictly solitary and must dig their less-complex tunnel systems alone. The hypoxia tolerance of these solitary species may differ as a result of differences in the soils they reside in; B. suillus live in soft, sandy soils, which are porous to air and water, but are easier for this large species to dig in; G. capensis occur in harder soils that are less porous to air and water and would require more effort to move through, and therefore heightened respiration (Thomas et al., 2009, 2013). In addition, these two species are known to occasionally emerge from their tunnel systems and forage on the surface; this is most frequently seen in B. suillus. Cryptomys, in contrast, is a social genus of mole-rat that has a large range, partially because it is one of the most specious genera (Ingram et al., 2004). Although C. h. hottentotus and G. capensis have different levels of sociability, they have similar lifestyles and spend most of their time in their nests, estimated as 81% and 90% of their time, respectively (Bennett, 1992). However, Cryptomys species live in much harsher environments and burrow in packed, rocky ground. Cryptomys hottentotus species would therefore be expected to be the most hypoxia-tolerant species, as a result of living in social colonies, rarely coming to the surface, and digging in harder soils than G. capensis: factors that would result in increased levels of CO₂ in their sealed-off tunnel systems.

In order to assess the mole-rats' ability to tolerate hypoxic stress at the molecular level, several molecular markers of cell stress and microRNA (miRNA) profiles were measured in the brains of these six species, comparing normoxic and hypoxic conditions. Specifically, miRNAs that are upregulated in response to hypoxia were investigated in the six species of African mole-rat, with most having roles in regulating cell death, inflammation and angiogenesis (Kulshreshtha et al., 2007; Truettner et al., 2013). Hypoxia damage to the brain was estimated by analyzing relative protein carbonylation and 8-hydroxy-2′-deoxyguanosine (8-OH-dG) levels. Protein carbonylation, an irreversible modification that most often occurs on cysteine and methionine residues, is a biomarker of oxidative stress that occurs in disease (Alzheimer's, ischemia–reperfusion) and natural aging (Dalle-Donne et al., 2006). 8-OH-dG has been established as a strong indicator of hypoxia-induced tissue damage and can be detected in both the brain tissue and urine of mammals exposed to acute hypoxia and ischemic stroke (Nair et al., 2012; Nakajima, 2012). Based on reports of altered antioxidant enzyme levels and activity in selected normoxic mole-rat species relative to other rodent species (Andziak et al., 2006; Caballero et al., 2006; Fang et al., 2014; Schmidt et al., 2016), overall antioxidant capacity was also assessed to determine whether mole-rat species increase the use of antioxidants to prevent damage in response to oxidative stress. Protein levels of the hypoxia-inducible transcription factor (HIF-1α) and its binding capacity to DNA were also evaluated to determine whether hypoxia signaling differed among the six species. Finally, the relative expression of 27 miRNAs that have been linked with low oxygen adaptation in other mammals was assessed. The results suggested that the mole-rat species examined respond to hypoxia in a species-specific manner, where the species with the lowest tolerance to low oxygen would also be the least likely to experience hypoxia in the wild.

Overall, this study investigated whether different mole-rat species experience damage induced by hypoxia (oxidized protein and DNA markers), and explored the mechanisms they may use to tolerate hypoxia (HIF-1α DNA binding and protein levels, miRNA, antioxidant capacity), which may provide insight into the treatment and prevention of human diseases involving hypoxic insults, such as stroke and cardiac arrest. Furthermore, understanding how naturally occurring hypoxia-tolerant animals deal with hypoxia at the molecular level will help in the development of treatments for solid cancer tumors, as hypoxia is a common characteristic of locally advanced solid tumors that has been associated with diminished therapeutic response and malignant progression.

Except for naked mole-rats, which came from a laboratory-raised colony, all other mole-rat species were wild-captured in South Africa with permits and the experimental procedures that were approved by the animal ethics committee of the University of Pretoria (EC069-17). Mean body mass of the animals was: C. h. mahali 109.1±6.6 g, C. h. pretoriae 114.8±12.3 g, C. h. hottentotus 78.7±5.5 g, B. suillus 587.0±90.8 g, G. capensis 133.2±19.0 g and H. glaber 51.3±3.7 g. Wild-caught mole-rats (randomly separated into n=4 groups of control and hypoxic animals) were acclimated to laboratory conditions 3 weeks prior to the hypoxic exposures, using a simulated 12 h light:12 h dark cycle and with food supplied ad libitum. For hypoxia exposures, animals were placed individually and unrestrained inside 4.7 l (B. suillus) or 1 l (all other species) Plexiglas chambers held at ∼28°C. Animals were provided with a thin layer of bedding on the floor of the experimental chamber. The chamber was sealed and constantly supplied with dried incurrent air and nitrogen, mixed to the desired fractional gas composition by pre-calibrated rotameters (Matheson Model 7400 Gas Mixer, E700 and E500 flowtubes, Oakville, ON, Canada) at a flow rate of 600 ml min⁻¹ (small chamber) or 2 l min⁻¹ (large chamber), which was assessed by a calibrated mass flow meter (Alicat Scientific, Tucson, AZ, USA). Excurrent air flow leaving the animal chamber was continuously subsampled at 200 ml min⁻¹, with water vapor measured with a water vapor analyzer (RH-300, Sable Systems, Las Vegas, NV, USA), dried with pre-baked drierite (Drierite, W.A. Hammond Drierite Co. Ltd, Xenia, OH, USA), and analyzed for O₂ and CO₂ fraction using a galvanic fuel cell O₂ analyzer and infrared CO₂ analyzer (FOXBOX, Sable Systems).

Naked mole-rats were group-housed in interconnected multi-cage systems at 30°C and 21% O₂ in 70% humidity with a 12 h light:12 h dark cycle. Animals were fed fresh tubers, vegetables, fruit and Pronutro cereal supplement ad libitum. Animals were not fasted prior to experimental trials. All experimental procedures were approved by the University of Ottawa Animal Care Committee in accordance with the Animals for Research Act and by the Canadian Council on Animal Care. Non-breeding (subordinate) naked mole-rats do not undergo sexual development or the expression of sexual hormones and thus we did not take sex into consideration when evaluating our results (Holmes et al., 2009).

Animals were randomly assigned to either (i) 3 h of normoxia (21 kPa O₂, balance N₂; control) or (ii) 3 h of hypoxia at 5 kPa O₂ for most species or 7 kPa O₂ for B. suillus; exposures were preceded by a 30 min step at 12 kPa O₂. Hypoxia exposures were chosen to be just above the low oxygen minimum that the animals withstood in preliminary experiments (Ivy et al., 2019). Following experimentation, animals were euthanized immediately by cervical dislocation followed by decapitation and brain samples were extracted on ice and stored at −80°C until analyzed.

Whole frozen brain from n=4 control and n=4 hypoxic animals of each species was weighed (∼60–80 mg) and crushed into small pieces under liquid nitrogen. Tissue samples were homogenized 1:4 w:v in Cell Signaling lysis buffer (Millipore-Sigma, 43-040) supplemented with 1 mmol l⁻¹ sodium orthovanadate, 10 mmol l⁻¹ sodium fluoride, 10 mmol l⁻¹ β-glycerophosphate and 10 µl ml⁻¹ Protease Inhibitor cocktail (BioShop, PIC001.1) using a Polytron PT10 homogenizer. Each sample was placed on ice, sonicated for 10 s, and then gently vortexed every 10 min for 30 min before centrifugation at 13,500 g for 20 min at 4°C. The supernatant containing soluble proteins was collected and total protein concentration was determined for ELISA samples using the Bio-Rad method (Bio-Rad, 500-0005). Samples for the protein carbonylation assay were standardized to 3 µg µl⁻¹, samples for the HIF-1α transcription factor DNA binding activity assay were standardized to 1 µg µl⁻¹, and samples for the total HIF-1α assay were standardized to 20 µg µl⁻¹ for all species except for H. glaber, which were standardized to 16.5 µg µl⁻¹ by adding appropriate amounts of Cell Signaling lysis buffer. Protein samples for the antioxidant capacity assay were diluted to 2.2 µg µl⁻¹ in antioxidant assay buffer provided with the kit.

Protein carbonylation (Cayman Chemical Company, 10005020) was assessed as per the kit manufacturer's instructions. The amount of protein-hydrazone produced from the reaction of 2,4-dinitrophenylhydrazine (DNPH) with protein carbonyls in tissue homogenates was quantified spectrophotometrically as a measure of protein carbonyl content. For each sample from normoxic and hypoxic individuals, 100 µl of total protein extract was standardized to 3 µg µl⁻¹ and then treated with either 400 µl of DNPH (experimental) or 400 µl 2.5 mol l⁻¹ HCl (control). Both experimental and control solutions were incubated in the dark at room temperature (RT) for 1 h with intermittent vortexing every 15 min. Then, 0.5 ml of 20% TCA solution was added to each sample and vortexed, followed by a 5 min incubation on ice and then centrifugation at 10,000 g for 10 min at 4°C. The supernatant was discarded and the pellet resuspended in 0.5 ml of 10% TCA and placed on ice. Samples were centrifuged again, and the pellet was resuspended in 0.5 ml 1:1 v:v ethanol/ethyl acetate followed by vortexing and centrifugation at 10,000 g for 10 min at 4°C. This step was repeated two more times. Next, protein pellets were resuspended in 250 µl guanidine hydrochloride, centrifuged again under the same conditions and 220 µl of the supernatant from each sample was transferred to a separate well of a 96-well microplate. Absorbance was measured at 370 nm and the signal from each control well was subtracted from the signal of the corresponding experimental well. Protein carbonyl concentration (nmol ml⁻¹) was then calculated as per the manufacturer's instructions.

Antioxidant capacity was assayed using a commercially available kit (Cayman Chemical Company, 709001), which utilizes the ability of antioxidants in the sample to inhibit the oxidation of ABTS [2,2′-azino-di-(3-ethylbenzthiazoline sulfonate)] to ABTS⁺. A standard curve was prepared by combining increasing amounts of Trolox (6-hydroxy-2,5,7,8-tetramethylchoman-2-carboxylic acid) diluted in antioxidant assay buffer (5 mmol l⁻¹ potassium phosphate, pH 7.4, with 0.9% sodium chloride and 0.1% glucose) to obtain concentrations between 0 and 0.330 mmol l⁻¹. Briefly, 10 µl of Trolox solution or 2.2 µg µl⁻¹ protein samples was combined with 10 µl of metmyoglobin and 150 µl of chromogen. Reactions were initiated by addition of 40 µl of the hydrogen peroxide working solution and were incubated on a shaker in the dark for 5 min at RT. Absorbance was read at 750 nm and quantified as Trolox equivalents (mmol l⁻¹ mg⁻¹ wet mass).

An assay for HIF-1α transcription factor DNA binding activity was performed as directed by the manufacturer (Abcam, ab133104), using total protein extracts prepared as above. To ensure the signal detected was indeed due to HIF-1α binding to DNA, one well containing NiCl₂-stimulated HeLa cell nuclear extract was used as a positive control and another contained competitive dsDNA. The blank contained the 1× antibody binding buffer (ABB) but no antibodies, and non-specific binding was assessed via the use of wells that contained the 1× ABB and both primary and secondary antibodies. To each well, 10 µl of each control or hypoxia protein extract was added along with 90 µl of complete transcription factor binding assay buffer. The plate was sealed and incubated overnight at 4°C. The wells were washed as directed before adding 100 µl of transcription factor (TF) HIF-1α primary antibody (1 in 100 using 1× ABB) to each well, except the blank wells. The plate was sealed and incubated for 1 h at RT, then washed and 100 µl of TF goat anti-rabbit HRP conjugate secondary antibody (1 in 100 using 1× ABB) was added to all wells except the blank, followed by incubation for 1 h at RT. The plate was washed as directed and then 100 µl of TF developing solution was added to each well followed by incubation for 45 min at RT with gentle agitation while protected from light. Once wells were visibly blue, 100 µl of TF stop solution was added and absorbance readings were obtained at 450 nm using a plate reader.

HIF-1α protein abundance was determined with a commercially available HIF-1α ELISA kit (Abcam, ab171577). All buffers, reagents and standards were prepared according to the manufacturer's instructions. First, a HIF protein standard with a stock concentration of 200 ng ml⁻¹ was diluted in cell extraction buffer to concentrations between 15 and 0.23 ng ml⁻¹. Protein lysates standardized to 20 µg µl⁻¹ (or 16.5 µg µl⁻¹ for H. glaber samples) and standards were added to their appropriate wells, with an equal volume of 1× antibody cocktail (a mixture of capture and detection antibodies). The plate was sealed and incubated for 1 h at RT on a plate shaker. Next, each well was rinsed 3 times with 350 µl of 1× prepared wash buffer ensuring complete removal of liquid after each wash. Then, 100 µl of TMB substrate was added to each well and incubated for 10 min in the dark on a plate shaker, followed by addition of 100 µl of stop solution to each well and shaking for 1 min to mix. Optical density (OD) was then measured at 450 nm using a plate reader.

DNA was extracted from both normoxic and hypoxic mole-rats (n=4 each) using a Quick-DNA Plus Kit (Zymo Research, D4068) according to the manufacturer's instructions. Briefly, samples weighing 25 mg were lightly homogenized in proteinase K solution (47.5% solid tissue buffer, 5% proteinase K) using a glass pestle and then digested overnight. The next day, 400 µl of genomic binding buffer was added to each sample, followed by 10–15 s vortexing. Samples were then transferred to a Zymo-Spin™ column and centrifuged at 12,000 g for 1 min. The flow through was discarded, and 400 µl of DNA pre-wash buffer was added. After another centrifugation, a total of 900 µl of g-DNA wash buffer was used to wash the collected DNA. Finally, the DNA was eluted from the column after a 5 min incubation in 50 µl of DNA elution buffer. The DNA was digested by incubation at 55°C for 1 h with 4 U of nuclease P1 (Millipore-Sigma, N8630). To digest nucleotides to nucleosides, 1 U of alkaline phosphatase (NEB, M0290S) was added and each sample was incubated at 37°C for 30 min, then boiled for 10 min and placed on ice until use.

DNA/RNA oxidative damage ELISA was performed according to the kit manufacturer's instructions (Cayman Chemical, 589320). Standards were prepared from stock reagents supplied in the kit and reconstituted to concentrations from 10.3 to 3000 pg ml⁻¹. DNA samples from normoxic and hypoxic mole-rat brains were diluted in 1× ELISA buffer (Cayman Chemical, 400060) to 15 ng µl⁻¹. A non-specific binding blank (100 µl ELISA buffer, 50 µl tracer dye), a maximum binding blank (50 µl ELISA buffer, 50 µl tracer and 50 µl of monoclonal antibody) and a background absorbance blank (Ellman's reagent) were used. For both conditions and all six mole-rat species, 50 µl each of DNA, AChE tracer and monoclonal antibody specific to 8-OH-dG were combined into each well of the supplied microplate, covered with an adhesive strip and incubated overnight in the dark on a shaker at 4°C. Subsequently, the wells were washed 5× with wash buffer, 200 µl of Ellman's reagent was added and the plate was incubated for 90 min on a shaker at RT in the dark. OD was measured spectrophotometrically at 412 nm with blank subtraction.

RNA isolation and miRNA amplification were performed as described previously (Biggar et al., 2018). Total RNA from mole-rat brains was isolated using Trizol™ (Invitrogen, 15596018). Briefly, 50 mg samples of tissue were each homogenized in 1 ml Trizol using a Polytron PT10 homogenizer. Next, 200 µl of chloroform was added followed by a 10 min incubation on ice and brief vortexing at low speed. Samples were then centrifuged at 10,000 g for 15 min at 4°C after which the upper aqueous layer was collected. RNA was precipitated with 500 µl of 2-propanol, followed by a 10 min incubation at RT, and centrifugation as above. The supernatant was removed and the pellet was washed with ethanol and centrifuged again. Excess ethanol was allowed to evaporate for 10 min. RNA pellets were then resuspended in 40 µl of sterile RNAse-free water. RNA purity was assessed by measuring the A₂₆₀/A₂₈₀ ratio. RNA integrity was determined by visualizing 18S and 26S ribosomal bands on a 1% agarose gel with SybrGreen staining. All samples were diluted to 1.5 µg µl⁻¹ using RNAse-free water.

Polyadenylation was performed using a polymerase A tailing kit (Epi-Bio, PAP5104H). Reaction mixes were prepared with 1 µl of 10 mmol l⁻¹ ATP, 0.5 µl E. coli polyadenylate polymerase (2 U) and 2 µg of total RNA, diluted with RNase-free water to a final volume of 10 µl. Reactions were incubated at 37°C for 30 min, then 95°C for 5 min, and transferred to ice. To ligate the stem-loop adapters to the RNA, 10 µl aliquots of polyadenylated RNA were incubated with 5 µl of 250 nmol l⁻¹ universal stem-loop RT primer at 95°C for 5 min and then at 60°C for 5 min. After cooling on ice, reverse transcription was performed by adding 4 µl 5× first strand buffer (ThermoFisher Scientific, 18080044), 2 µl 0.1 mol l⁻¹ DTT, 1 µl 25 mmol l⁻¹ dNTP (ThermoFisher Scientific, R1121) and 1 µl M-MLV reverse transcriptase (2 U; ThermoFisher Scientific, 18080044) to each RNA sample. Reactions proceeded at 16°C for 30 min, 42°C for 30 min, followed by 85°C for 5 min. Finally, cDNA products were serially diluted and stored at −20°C until further use. The miRNAs to be evaluated were selected from the literature by doing searches on NCBI PubMed using keywords including miRNA, hypoxia and brain to identify numerous miRNAs that have been shown to be differentially expressed under hypoxia stress in rat brain (Truettner et al., 2013). The miRNAs of interest were evaluated for sequence conservation where the available mature miRNA stemloop sequences from Mus musculus obtained from miRBase database (release 22) were aligned with those of H. glaber using EMBL-EBI Clustal Omega. The universal reverse primer (5′-CTCACAGTACGTTGGTATCCTTGTG-3′) and the forward primers listed in Table S1 were synthesized by Integrated DNA Technologies and were used in all qPCR reactions performed on a CFX96 real-time PCR detection system (BioRad Laboratories Inc.). Denaturation for 3 min at 95°C preceded 40 cycles of 95°C for 10 s and 60°C for 30 s. Melt curve analysis was used to confirm the amplification of a single amplicon.

For all analyses, data were analyzed by Student's t-test (n=4 independent replicates, P<0.05) to determine statistical significance between conditions (normoxia and hypoxia) for each species, independently of one another. All statistical analyses and figures were generated using RStudio or R-BioPlot (Zhang and Storey, 2016).

For each ELISA, the means (±s.e.m.) of the normoxic controls and hypoxic datasets for each species were calculated, and control values were normalized to 1 and the hypoxia mean was expressed relative to the normoxic control mean. In addition, a one-way ANOVA with a Tukey post hoc test (n=4 independent replicates, P<0.05) was used to compare the parameters assessed via ELISA between species at normoxic conditions. The analysis of HIF-1α protein levels excluded H. glaber because less total protein was used in the assay.

For all miRNAs examined, following amplification, expression data for the miRNAs of interest and the loading control were linearized using 2^−Cq as per the comparative ΔΔCq method. Sno-66 was used as the reference gene for all six species because its expression did not significantly change with hypoxia stress in any of the species (Student's t-test P>0.05). ΔΔCq was obtained by standardizing the miRNA expression data to the loading control. Then, the means of both normoxic and hypoxic miRNA data were generated. For each species, the normoxic (control) mean was normalized to 1 for graphing purposes and the hypoxia mean was expressed relative to the normoxic mean.

Principal component analysis (PCA) was computed in R using ‘FactominR’ version 1.41. Missing values were imputed using ‘missMDA’ version 1.14 (Josse and Husson, 2016). Relative protein, damage marker and gene expression data were used in the comparison of control and stress groups. Then, the top five contributing factors for each principal component were identified by the function fviz_contrib using choice=‘var’. These are reported in Table S3.

The relative protein carbonylation level did not change in any of the six mole-rat species in response to hypoxia exposure (Fig. 2). Furthermore, there were no changes in the level of 8-OH-dG after hypoxia in any of the species (Fig. 3).

There was no significant change in antioxidant capacity comparing normoxia controls and hypoxia conditions for any of the species (Fig. 4A). However, when comparing all species in normoxia, antioxidant capacity was the only parameter assessed via ELISA to show statistical significance between species, with C. h. mahali having 2.11(±0.28)-fold increased antioxidant capacity relative to B. suillus.

The protein levels and transcription factor DNA-binding activity of HIF-1α were examined (Figs 5 and 6, respectively) and revealed species-specific results. Total HIF-1α levels increased in response to hypoxia exposure in both B. suillus and C. h. hottentotus to 3.26(±0.25)-fold and 2.59(±0.18)-fold, respectively, over normoxic controls, but were unchanged in the other species (Fig. 5). However, HIF-1α transcription factor binding activity did not change in any of the mole-rat species in response to hypoxia (Fig. 6).

The levels of 27 miRNAs that have been implicated in hypoxia-related processes were evaluated in the brain of all six mole-rat species. Fig. 7 is a heat map comparing the species whereas Fig. S1 shows the individual data for each species and Table S3 summarizes the roles of miRNAs chosen for this study. The greatest number of significant changes in miRNAs occurred in B. suillus, with strong upregulation of 19 miRs ranging from 1.72(±0.18)-fold to 14.6(±4.04)-fold, as compared with controls. In C. h. hottentotus, only two significant changes were noted, a 2.07(±0.27)-fold increase in miR-125b-5p and a decrease in miR-24-3p levels to 69±7.1% of control values (Fig. 7; Fig. S1B). The brain of C. h. mahali showed decreased expression of three miRNAs: miR-210-3p, miR-150-5p and miR-592-3p decreased to 53±13%, 48±14% and 58±12% of control values, respectively (Fig. 7; Fig. S1C). However, the brain of the related C. h. pretoriae did not differentially express any miRNAs in response to hypoxia (Fig. 7; Fig. S1D). The Cape mole-rat, G. capensis, significantly changed the levels of 9 of the 27 miRNAs assessed. Levels of four miRNAs increased (miR-147-3p, miR-150-5p, miR-199a-3p and miR-325-5p) with changes ranging from 1.54- to 3.31-fold compared with the control value. Five miRNAs decreased in expression (miR-24-5p, miR-24-3p, miR-27a-3p, miR-140-5p and miR-146a-3p), with levels reduced to 22–43% of the normoxic mean (Fig. 7; Fig. S1E). Finally, the brain of H. glaber significantly increased five miRNAs in response to hypoxia with values ranging from 1.27- to 2.17-fold the normoxic value for miR-24-5p, miR-24-3p, miR-207, miR-27a/b-3p and miR-592-3p (Fig. 7; Fig. S1F). In contrast, three miRNAs (miR-210-5p, miR-26b-5p and miR-126a-5p) were suppressed by 31–52%.

PCA plots were used to uncover the underlying trends in the molecular data. There were no statistical differences between control and hypoxic groups of any of the three Cryptomys hottentotus subspecies when considering all molecular data (Fig. S2) or when only mapping the trends from the top five contributing factors for PC1 and PC2 (Table S2; Fig. 8). The molecular markers that we used were sufficient to identify major differences between control and hypoxic mole-rats for the three other species, B. suillus, G. capensis and H. glaber. However, when considering only the top five contributing variables for these three species, the significant difference between control and hypoxia was only seen for B. suillus and G. capensis.

When comparing the molecular data of hypoxic animals from all species, it was observed that B. suillus was different from all other species (Fig. 9); the top five variables for each principal component are shown for clarity.

In this study, we evaluated the relative expression of molecular markers linked to hypoxia in the brains of six species of mole-rats to determine how various African mole-rat species respond to hypoxia. Firstly, it was interesting that none of the mole-rat species showed increased oxidative damage during hypoxia (Figs 2 and 3). The unchanging level of protein carbonylation and 8-OH-dG DNA during hypoxia suggests that mole-rats may use mechanisms to limit oxidative damage and/or remove the by-products of reactive oxygen species (ROS) reactions. Analysis of liver and urinary metabolites suggests that mole-rats accumulate more oxidized protein, lipid and DNA than hypoxia-intolerant mice, possible as a result of having 70-fold lower liver glutathione peroxidase (GPx) activity levels than mice in normoxic conditions (Andziak et al., 2005). Even though mice have highly active GPx and high antioxidant enzyme gene expression at normal oxygen levels, mice exposed to 9% oxygen still showed increased protein carbonylation and other markers of oxidative injury (Veasey et al., 2004). Recently, research has shown H. glaber muscle and heart can detoxify mitochondrial ROS better than those of mice in normoxic conditions, suggesting enhanced tissue-specific antioxidant capabilities (Munro et al., 2019). Brain from anoxia-tolerant turtles and mole-rats increases or maintains high gene/protein expression and enzyme activity levels of antioxidant enzymes other than GPx in low oxygen conditions, which may help them handle episodic oxidative stress (Andziak et al., 2006; Schülke et al., 2012; Willmore and Storey, 1997). Herein, all six species also showed no change in global antioxidant capacity with hypoxia (Fig. 4). Sustained antioxidant capacity in conjunction with no change in oxidative damage of DNA or protein during hypoxia is consistent with these previous reports of constitutively high and increased expression and activity patterns of certain, but not all, antioxidant enzymes in animals treated with hypoxia. Further, the relatively short time frame of the hypoxia exposure (3 h) in the current study may not have necessitated an enhancement of antioxidant defenses. Increased antioxidant capacity may be more important upon reoxygenation, which was not studied herein. Interestingly, antioxidant capacity was the only parameter to significantly differ between different species in normoxic conditions. Specifically, C. h. mahali had over 2-fold higher antioxidant capacity than B. suillus, suggesting different mole-rat species may have different baseline antioxidant capacities to help them adapt to different levels of natural hypoxic exposures. Also, considering B. suillus had a reduced capacity to withstand experimental hypoxic exposures as severe as those tolerated by C. h. mahali, B. suillus may have a lower antioxidant capacity in normoxia because this solitary species does not encounter low oxygen environments as regularly as the more social C. h. mahali might, in the wild.

Instead of using increased antioxidant capacity, mole-rat tissues (including brain) may minimize accumulation of damaged proteins by removing them as they are formed, as evidenced by greater protein ubiquitination and proteasome activity compared with that in mice (Garbarino et al., 2015; Perez et al., 2009). Mole-rats may also use comparable mechanisms to minimize or prevent the accumulation of DNA damage. As previously mentioned, mole-rats may increase the activity of certain antioxidant enzymes to prevent DNA oxidation. Similar to the Spalacidae family of mole-rats, Bathyergidae mole-rats may have genomic differences compared with hypoxia-intolerant and terrestrial rodents that allow them to express DNA repair proteins under hypoxia stress (Shams et al., 2013). Compared with hypoxia-intolerant species, H. glaber have a higher copy number of the two genes (TINF2 and CEBPG) that are involved in protecting DNA integrity and repairing damaged DNA, and they show more efficient nucleotide and base excision repair and PARylation activities than mouse cells in response to both H₂O₂ and UV irradiation damage to DNA (Evdokimov et al., 2018; Macrae et al., 2015). Overall, the lack of change in DNA and protein oxidation in response to hypoxia in the six species analyzed in the present study supports the concept that mole-rats have evolved unique molecular mechanisms that suit them for subterranean life under low oxygen conditions.

HIF-1α levels increased strongly in the brains of B. suillus and C. h. hottentotus under hypoxia, providing an indication that O₂ levels in the brains of these two species were at or below a crucial low O₂ tension that necessitated a metabolic response to alleviate hypoxia, which may indicate variability in hypoxia resistance among the six species. Bathyergus suillus and C. h. hottentotus happen to have overlapping ecological niches (Thomas et al., 2013) and C. h. pretoriae and C. h. mahali do not, even though they are more closely related. This suggests that the environment may influence the molecular mechanisms that are utilized for hypoxia tolerance. Bathyergus suillus and C. h. hottentotus may upregulate HIF-1 to enhance glycolysis by directly inducing transcription of the glucose transporter (GLUT1) and several glycolytic enzymes as well as upregulating pyruvate dehydrogenase kinase 1 (PDK1) to inhibit pyruvate catabolism by the tricarboxylic acid cycle, and thereby promote lactate production as the end product of glycolytic ATP production (Mirtschink and Krek, 2016; Papandreou et al., 2006). Under normoxic conditions, HIF-1α levels are very low as a result of rapid degradation by the ubiquitin-proteasome system whereas HIF-1β is constitutively expressed. During hypoxia, however, HIF-1α levels are stabilized, leading to the formation of the HIF-1 dimer and then HIF-1 binding to the hypoxia response element (HRE) of target genes to upregulate their transcription (Dengler et al., 2014). HIF-1α from H. glaber may have amino acid substitutions that prevent HIF-1α ubiquitination and degradation such that it may be constitutively expressed (Kim et al., 2011; Xiao et al., 2017), which could help explain why HIF-1α protein levels were simply maintained in the other four African mole-rat species. Xiao et al. (2017) also demonstrated that HIF-1α protein levels only increased following 4 h of hypoxia, suggesting future time-course studies are warranted in the exploration of furry African mole-rat responses to hypoxia.

Indeed, the natural history of these species can help explain why certain mole-rats show marked changes in miRNA expression or HIF-1α protein abundance, but not others. Bathyergus suillus and G. capensis are solitary species, whereas the Cryptomys hottentotus subspecies are social, and H. glaber live in large, eusocial colonies (Bennett and Faulkes, 2000; Robb et al., 2012). Their sociality undoubtedly impacts their foraging and burrowing behaviors and, as a result, their capacity to tolerate hypoxia. Bathyergus suillus live in soft, sandy soils and alone; consequently, they do not dig complex tunnel systems like C. h. hottentotus (Thomas et al., 2009, 2013). In contrast to other mole-rats, B. suillus sometimes forage on the surface for grasses and forbes where O₂ is not limited, so it has less of a need for hypoxia tolerance compared with other species of mole-rat, such as C. h. hottentotus, which spends 100% of its foraging time underground, searching for the underground storage organs of geophytes, in hypercapnic burrows with many other individuals. Indeed, wild-captured B. suillus could not withstand the same degree of hypoxia at the beginning of this experiment, nor could it reduce its metabolic rate, or change its tidal volume to compensate for changes in O₂ availability, to the same extent as the other African mole-rat species during hypoxia (Ivy et al., 2019). As a result, B. suillus is less hypoxia tolerant than the other species and may require more numerous and/or larger fold-changes in certain miRNAs to cope with decreases in oxygen availability. When comparing the molecular responses to hypoxia of all species (Fig. 9), B. suillus is statistically separate from the other species, supporting the premise that it has a unique response to hypoxia.

Specifically, to cope with low oxygen stress, B. suillus strongly upregulated several miRNA, including miR-26b-5p, miR-124-3p, let-7f-5p, miR-378-3p and miR-210-3p (Fig. 7). These five miRNA are noteworthy because they are the top five variables that contribute to PC1, which helps explain 92% of the differences seen between control and hypoxic brains (Fig. 8A). Each of these miRNAs has protective roles during hypoxia, and could be important for the survival of B. suillus, the least hypoxia-tolerant African mole-rat (Ivy et al., 2019). For instance, miR-210-3p is a hypoxia-inducible miRNA with many roles including preventing cell death and promoting tissue repair in response to ischemic and hypoxic brain damage (Meng et al., 2018; Zeng et al., 2011). Additionally, miR-210 can inhibit glycerol 3-phosphate dehydrogenase 1-like (GPD1L) expression, a protein that tags HIF for degradation (Bavelloni et al., 2017), which is consistent with elevated HIF-1α levels in brain tissue from B. suillus. Another miRNA that may prevent apoptosis during hypoxia in B. suillus is miR-378, which is reported to inhibit caspase-3 in mouse brain cells injured by ischemia (Fang et al., 2012; Zhang et al., 2016). Interestingly, hypoxia-intolerant rodents show decreased let-7f levels in their brains with hypoxia (Jeyaseelan et al., 2008), but this miRNA is strongly upregulated in hypoxic B. suillus brain. let-7f targets N-myc downstream-regulated gene (NDRG3) to inhibit angiogenesis and cell growth through the Raf-ERK pathway (Yao et al., 2017), and it may be important to upregulate it in B. suillus brain to maintain tissue homeostasis during hypoxia. The brain of B. suillus also upregulated miR-124-3p, a marker of stroke with hypoxia-inducible roles in the inhibition of autophagy, neuronal inflammation and the expression of neurodegenerative proteins such as amyloid beta and phospho-Tau (Gu et al., 2016; Leung et al., 2014; Sørensen et al., 2017). miR-124 is associated with improved neurologic outcomes and inhibition of neuroinflammation in mice after traumatic brain injury (Huang et al., 2018). Overall, increased levels of these miRNAs under hypoxic stress are likely protective against neurological damage in B. suillus.

The PCA plot indicates that G. capensis significantly increases its molecular response to hypoxia, similar to B. suillus (Fig. 8E). This is interesting because it is also a solitary species and occasionally forages above ground (Robb et al., 2012), suggesting that this lifestyle may be correlated with poorer hypoxia tolerance. Further, B. suillus and G. capensis have some similarities in their physiological response to hypoxia. In moderate (7–9 kPa) and severe hypoxia (3–5 kPa), both solitary species increase breathing frequency and do not suppress their metabolic rate as strongly as the more social species like C. h. mahali and C. h. pretoriae (Ivy et al., 2019). Although many molecular adaptations that G. capensis use are similar to those of B. suillus, as determined by overlap in the most important variables contributing to the principal components (Table S2), there are still differences in the response to hypoxia between B. suillus and the other five mole-rat species (Fig. 9). Importantly, G. capensis may be able to survive lower P_O2 than B. suillus and adapt to these conditions using fewer and lower fold-changes in cell signaling proteins and miRNA expression. Many of the molecular targets identified by the PCA, which explain over 75% of the variance between control and hypoxic G. capensis groups for PC1, were the same targets that were differentially regulated in B. suillus (Table S2). One of these, miR-147, was upregulated in both G. capensis and B. suillus during hypoxia. miR-147 decreases in hypoxia-treated rat adrenal medulla cells used to study cerebral ischemia, which is associated with increased apoptosis and decreased cell migration, resulting in increased Sox2 and activation of NFκB and Wnt signaling pathways (Han et al., 2017). Thus, an increase in miR-147 during hypoxia could be important to limit apoptosis in B. suillus and G. capensis brains.

If only the top five contributing factors for PC1 and PC2 are considered, there is no difference between control and hypoxic H. glaber groups, but if all data are considered, a difference between control and hypoxia can be detected (Fig. 8F; Fig. S2F). This suggests that H. glaber uses more numerous, smaller fold-changes to respond to hypoxia, rather than a few key molecular adaptations. Interestingly, the brains of H. glaber and B. suillus upregulate one or both miR-24 isomiRs while the opposite is true for G. capensis and C. h. hottentotus. Some studies implicate miR-24 in apoptosis-promoting pathways via the inhibition of anti-apoptotic XIAP (Zhou and Zhang, 2014), but decreases in miR-24 can also be protective in ischemic rat cardiomyocytes (Li et al., 2012). Most studies identify pro-apoptotic BIM as a major miR-24 target, which means that an upregulation of miR-24 prevents apoptosis (Kulshreshtha et al., 2007; Roscigno et al., 2017; Sun et al., 2016). Therefore, an increase in miR-24 in hypoxic H. glaber and B. suillus brains would limit apoptosis. By contrast, the decrease or maintenance of this miRNA in other species could be attributed to the inherent redundancy of the miRNAome, where many different miRNAs can regulate the same process (e.g. miR-147 regulates apoptosis in G. capensis brain). As a result, it is important to stress that we may not have identified the main miRNAs that each species uses in response to hypoxia using the RT-qPCR method, and future studies could contribute a more thorough analysis of the miRNAome of these hypoxia-tolerant species, to identify the most important hypoxamiRs used by the brains of each species.

Lastly, all three Cryptomys hottentotus subspecies did not show any statistical differences between control and hypoxic groups (Fig. 8B–D), suggesting that the markers we hypothesized as being important in the hypoxic response are not the ones that these subspecies use to tolerate low oxygen or, rather, a lifestyle of foraging underground and sharing sealed burrows with other members has benefited these species by increasing their tolerance to hypoxia. Indeed, their high hypoxia tolerance may even facilitate species like C. h. mahali living at a range of altitudes (1600–3200 m) (Broekman et al., 2006).

Herein, we present species-specific hypoxia-mediated changes in HIF-1α and miRNA profiles of brains from six hypoxia-tolerant rodents. None of the mole-rat species showed evidence of protein or DNA damage during hypoxia. Furthermore, brain antioxidant capacity did not change, suggesting that all six species have sufficient antioxidant capacity to handle changes in oxygen availability. PCA analysis of all molecular data suggest that the solitary species, B. suillus and G. capensis, are the least hypoxia tolerant, requiring the highest number of miRNAs and degree of molecular regulation during hypoxic stress, whereas the least responsive species were all C. hottentotus subspecies. Indeed, B. suillus upregulated the most miRNAs during hypoxia, and most were shown to be anti-apoptotic, anti-inflammatory and anti-angiogenic. The few miRNAs that were differentially expressed in C. h. hottentotus, C. h. mahali and H. glaber brains were also implicated in anti-apoptotic and anti-inflammatory processes. Finally, most of the differentially expressed miRNAs in brain from G. capensis were implicated in suppressing apoptosis but were largely downregulated. However, the few miRNAs that did increase in response to hypoxia are postulated to have protective roles for the hypoxic brain, including inhibiting genes involved in apoptosis and neuroinflammation. These results suggest that the brain of each mole-rat species uses unique molecular mechanisms to increase hypoxia tolerance, and that C. hottentotus subspecies, as a result of their social and fully subterranean lifestyle, may be the most equipped to tolerate low oxygen.

We would like to thank Dr Bill Milsom and Ryan Sprenger, University of British Columbia, as well as Catherine Ivy, McMaster University, for expert help with the animal hypoxia exposures.