Genetic Analysis of Hypoxia Tolerance and Susceptibility in Drosophila and Humans

Abstract

Oxygen is essential for metazoans’ life on earth. Oxygen deprivation, or hypoxia, contributes significantly to the pathophysiology of many human diseases. A better understanding of the fundamental molecular and genetic basis for adaptation to low-oxygen environments will help us develop therapeutic strategies to prevent or treat diseases that have hypoxia as a major part of their pathogenesis. Different cells and organisms have evolved different ways to cope with this life-threatening challenge, and the molecular and genetic mechanisms remain largely unknown. The current revolution of genomic technology has advanced our understanding of the genetic basis of many diseases and conditions, including hypoxia tolerance and susceptibility. In this review, we highlight the progress made in understanding the molecular responses to hypoxia in an animal model organism (Drosophila melanogaster) and genetic adaptation to high-altitude hypoxia in humans.

INTRODUCTION

Modern humans have conquered many challenging environments over the past 50,000 years, including those with extreme temperatures and (markedly) oxygen concentrations. Although it is generally difficult for humans to adapt over shorter periods (days and weeks), and short durations in extreme environments can have debilitating consequences for some humans, over prolonged periods (hundreds and thousands of years) adaptation that allows humans to tolerate such environments has been evident.

High altitude is associated with reduced barometric pressure and hence with decreased oxygen levels in inhaled air. Clearly, the higher the altitude is, the more rarified oxygen is and the more difficult the environment is to tolerate. For example, at an altitude of 3,000 m, the oxygen partial pressure is 537 mm Hg (~70% of the oxygen available at sea level), and at an altitude of 5,000 m, the oxygen partial pressure is 420 mm Hg (~55% of the oxygen available at sea level). At the top of Mount Everest, the oxygen level is ~33% of that at sea level. Although some humans have reached the top of Mount Everest without oxygen tanks, the majority of climbers have used them, for obvious reasons (60). In an effort to study adaptation in an animal model, over the past 10 years we have raised Drosophila melanogaster flies in chambers and have decreased the oxygen concentrations every few generations, with flies now reaching a formidable adaptation to 4% O₂—an oxygen level less than 20% of that at sea level and equivalent to that of ~4,000 m above Mount Everest (63).

Besides humans, other mammals can also tolerate low oxygen, some of which were discovered at high altitude. For example, some amphibians, such as red-eared turtles (Trachemys scripta elegans), are extremely resistant to anoxia, tolerating a lack of oxygen in their tissues for weeks. Although to our knowledge there have been no attempts to sequence the whole genome of this particular turtle species, the genome of the yak—which lives at very high altitude in Tibet—was recently sequenced (50). This sequencing provided evidence for a role of rapidly evolving genes in energy metabolism as well as for proteins involved in sensing the extracellular environment, hypoxic stress, and nutrition.

The ability to rapidly sequence a human genome (in days or, more recently, even in hours) has changed the landscape for investigations of the basis of diseases and disease processes, especially in relation to high-altitude hypoxia tolerance. To a large extent, our goal of understanding the pathobiology of stress tolerance has become attainable. The genomic revolution has already made an impact on our understanding of many diseases and conditions, and we have already started to make major progress in understanding the basis of high-altitude tolerance and susceptibility. In this article, we first review what we and others have accomplished in an animal model system (Drosophila melanogaster) regarding the molecular and genetic responses to hypoxia. We then review what we and others have done experimentally regarding genetic adaptation to high-altitude hypoxia in humans.

DROSOPHILA MELANOGASTER: GENETIC AND GENOMIC STUDIES

Genetic Dissection of Mechanisms Underlying Anoxia Tolerance in Adult Drosophila Flies

An earlier study in our laboratory revealed a remarkable phenotype in adult Drosophila flies: They tolerate low-oxygen environments extremely well (24). After hours of anoxia, flies can recover from anoxia-induced stupor without apparent tissue injury. We took advantage of this characteristic and performed a genome-wide mutagenesis screen to investigate the genetic factors that regulate the hypoxia-tolerance phenotype. In these studies, we measured the time it took for the flies to recover from anoxic stupor after the start of reoxygenation and used this as an index to screen for mutants with prolonged recovery times (23). After screening >10,000 flies carrying mutations on the X chromosome, we identified four X-linked mutants in three separate loci, and named them hypnos-1^N, hypnos-2^L, hypnos-2^P, and hypnos-3^I. The longer the delay in recovery was, the more sensitive these mutants were to lack of oxygen. Additional studies showed that the hypnos-2^L and hypnos-2^P mutants induced a deletion and a functional disruption in the gene that encodes Drosophila pre-mRNA adenosine deaminase (dADAR), which is expressed almost exclusively in the adult central nervous system. Disruption of dADAR resulted in completely unedited and functionally different proteins for sodium (Para), calcium (Dmca1A), and chloride (DrosGluCl-α) channels. The consequence was prolonged recovery from anoxic stupor, vulnerability to heat shock, and increased oxygen demands as well as neuronal degeneration in adult flies (40). Although this X-ray mutagenesis screen was not saturated and was restricted to the X chromosome, the fact that screening >10,000 mutagenized flies resulted in only three positive loci suggested that the anoxia-tolerance phenotype is regulated by a limited number of genes or pathways.

Our laboratory also performed a misexpression screen using Drosophila enhancer/promoter (EP) insertion lines (52) to identify genes that regulate anoxia responsiveness on all fly chromosomes (31). In this study, we screened 1,600 EP lines with a daughterless (da)–Gal4 driver and identified the CG14709/dMRP4 gene [a homolog of human multidrug resistance protein 4 (MRP4/ABCC4)] as regulating anoxia response. Ubiquitous expression of CG14709/dMRP4 in adult flies increased their sensitivity to anoxia and led to dramatically prolonged recovery times from anoxic stupor. Constitutive overexpression of CG14709/dMRP4 caused larval lethality under hypoxic conditions owing to growth arrest. Interestingly, selective overexpression of this gene in neurons, but not in glia or muscles, was sufficient to induce an anoxia-sensitive phenotype, demonstrating again that (similar to the hypnos-2 mutants) disruption of genes specifically in neurons plays a critical role in reducing anoxia tolerance in flies.

Genetic Dissection of Mechanisms Underlying Tolerance to Constant or Intermittent Hypoxia in Drosophila

The impact of hypoxia depends on the strength of the hypoxic level, the duration of exposure, and the paradigm of insult used—i.e., constant hypoxia (CH) or intermittent hypoxia (IH). Organisms respond differently to CH and IH (15, 17, 18, 64). In a recent study, we screened the critical genes that regulate tolerance to either CH or IH in adult flies on a genome-wide scale (4). We used microarrays to measure the gene expression changes associated with severe short-term CH (2.5 h, 1% O₂) or IH (cycles of 1–21% O₂) in adult flies. Gene expression under CH and IH varied both in the number of responding genes and in the categories of gene families: Many fewer significantly altered genes were identified following acute IH as compared with acute CH (approximately a 1:10 gene ratio). Several gene families were overrepresented in the CH-treated flies, including those involved in the response to unfolded proteins, lipids, carboxylic acid, amino acid metabolic processes, and immunity, whereas the overrepresented gene families in IH were related to drug resistance.

Using P elements and EP insertion lines, we then further studied the role of candidate genes that were up- or downregulated under CH or IH. We observed that the P-element lines overexpressing the Hsp70 and Hsp23 genes significantly increased adult survival during CHand that overexpression of Mdr49 improved adult survival during IH, demonstrating that the increases in transcript levels of these genes play an important role in adult survival under severe hypoxic conditions. However, the role of the genes induced under CH or IH was not generalizable but very much specific to the paradigm. For example, Hsp70, which regulates CH tolerance, had no effect on IH tolerance; similarly, overexpression of Mdr49 enhanced adult survival under IH but not under CH (Figure 1). In addition, Hsp70 overexpression in specific tissues using the upstream activation sequence (UAS)–Gal4 system revealed that overexpression of Hsp70 in the heart (in cardiac or pericardial cells as well as in hemocytes) and to a certain degree in the brain (including the mushroom body and antennal lobes) increased adult survival under CH. These data provided further evidence that CH or IH leads to changes in gene expression that play an important role in the survival of the organism in such extreme conditions.

Figure 1.

Distinct mechanisms underlying tolerance to intermittent and constant hypoxia in Drosophila melanogaster. Alterations on l(2)08717 (i.e., P{ry[+t7.2] = PZ}l(2)08717) and Mdr49 (i.e., Mi{ET1}Mdr49) specifically enhanced survival of adult flies under intermittent hypoxia but not under constant hypoxia, whereas alterations on Hsp70Aa (i.e., P{wHy}Hsp70Aa) and Hsp70Bbb (i.e., P{Mae-UAS}Hsp70Bbb) improved survival of adult flies under constant hypoxia but not under intermittent hypoxia.

Genome-Wide Dissection of Hypoxia Adaptation Over Generations in Drosophila

Investigations in the past few decades have provided insight into the fundamental basis of hypoxia and the mechanisms by which it leads to injury. Furthermore, the mechanisms that underlie adaptation to and survival in hypoxic conditions have also begun to be identified. In the past decade, we have employed an approach that combines population genetics, transcriptome profiling, and genetic manipulation strategies to dissect the mechanisms underlying hypoxia tolerance to chronic CH conditions (65). In this study, we have taken a population genetic approach (laboratory selection), decreasing the level of inhaled oxygen every few generations to create a Drosophila strain that can develop and live perpetually in severe hypoxic conditions (containing only 4% O₂). The hypoxia selection was initiated at 8% O₂ because at 6% O₂ only a small proportion of the embryos reached adulthood, and 4% O₂ was embryonic lethal to fertilized eggs. By the 13th generation, we had obtained populations that could develop and live perpetually at 5% O₂; by the 32nd generation, we had obtained populations that could develop and live perpetually at 4% O₂, a lethal condition for naive flies.

We next asked whether this hypoxia tolerance is heritable. To answer this question, we cultured a subset of embryos from the hypoxia-selected flies under normoxia for several consecutive generations; after eight generations under normoxia, we then reintroduced embryos into the 4% O₂ environment. The majority (>80%) of embryos completed their development and were able to live perpetually in this extreme environment. This result strongly suggested that the hypoxia-tolerance phenotype in the hypoxia-selected flies was indeed heritable. In these flies, several remarkable phenotypic changes were also observed, including enhanced motility/activity performance under severe hypoxia, significant reductions in body weight and size, significant reductions in cell number and cell size in the wings, reduced rates of mitochondrial oxygen consumption and reactive oxygen species production, and rearrangements in the mitochondrial ultrastructure (2, 49, 65, 66) (Figure 2). This experiment demonstrated that the hypoxia-selected flies can serve as a powerful model to genetically dissect mechanisms that contribute to hypoxia adaptation, including cell signaling and metabolic regulation.

Figure 2.

Ultrastructural modifications in the mitochondria of hypoxia-selected Drosophila melanogaster flies. (a) Longitudinal slice through a tomographic volume of a mitochondrion found in the thoracic muscle of a control (naive) fly. This slice exhibits a normal morphology, including stacks of ordered cristae with a light intracristal space and darker matrix. (b) Segmented volume of the mitochondrial membrane structure of the control fly. The cristae (colored regions) are all lamellar in form, as expected. This mitochondrial volume contains 42 cristae. (c) Oblique view of the mitochondrion from the control fly, seen partially through the outer membrane (blue), which has been made translucent to show the arrangement of the cristae. (d) Slice through a tomographic volume of the thoracic muscle of a hypoxia-selected fly. This large mitochondrion displays a few regions devoid of cristae (blue arrowheads) and, more notably, small regions where the cristae have been fractured (e.g., the boxed area, which is expanded in panel g). (e) Top view of the membrane segmentation of the volume from the hypoxia-selected fly. This mitochondrial volume contains 641 cristae. (f) Oblique view of the mitochondrion from the hypoxia-selected fly, seen partially through the outer membrane (blue), which has been made translucent to show the arrangement of the cristae. (g–i) Expanded views of different areas of panel d, showing dense, rodlike cores between fractured cristae; panel g is specifically an expansion of the boxed area in panel d. The edge of the rod is the fracture curve of the cristae (blue arrowheads in panel g) where the cristae fractured almost uniformly in a localized region (black lines in panel h) and the ends annealed, i.e., the crista membrane closed around the break. The dense core (red arrowhead in panel g) appears to be the degenerated cristae (black outline in panel i) that are separated by a white (translucent) band from the annealed crista membrane curve. (j) Top view of the segmented region of panel i, showing the dense core (blue) surrounded by the crista fragments (red). (k) Side view perpendicular to the view in panel j, looking down the axis of the dense core, which resembles an avenue with the cristae as spaced sentinels. (l,m) Examples of lamellar cristae in the mitochondrion in panel d. The cristae in panel l are not as extensive as those in the control fly and often exhibit fenestrations. A greater percentage of these cristae are similar to the highly branched and severely fenestrated crista shown in panel m. (n,o) Comparisons of crista surface area and number of cristae in the thoracic muscles of control and hypoxia-selected flies. The crista surface area, normalized to the mitochondrial outer membrane surface area, was significantly greater in the hypoxia-selected fly (panel n). The number of crista, normalized to the mitochondrial cross-sectional area, was also significantly greater in the hypoxia-selected fly (panel o). The asterisks denote p < 0.5; the error bars show the standard error of the mean. Adapted from Reference 49.

We next performed genome-wide transcriptome profiling and whole-genome resequencing to further investigate the transcriptional signature in this hypoxia-selected Drosophila strain. To measure expression at different developmental stages, we identified the differentially expressed genes in both third-instar larvae and adult flies. This experiment detected 2,749 significant alterations (1,534 upregulations and 1,215 downregulations) at the larval stage, which is approximately 20 times greater than those found in the adults (138 significant alterations, comprising 95 upregulations and 43 downregulations). This suggested that hypoxia had a significantly stronger effect on the transcriptional activity in the developing flies (i.e., larval-stage fly, in which cells are rapidly differentiating and proliferating) than it did in the adult flies (in which these processes have largely ceased). Among the differentially expressed genes, the majority of the commonly downregulated genes encode proteins related to metabolism (135 genes were related to carbohydrate metabolism). In contrast, the upregulated genes encode multiple components of signal transduction pathways involved in EGF, insulin, Notch, and Toll/Imd signaling transduction.

We also sequenced the whole genomes of the hypoxia-adapted flies using high-throughput sequencing technology. The genomic regions under hypoxia selection contained 188 protein-coding genes (63), and a total of 68 genes remained high priority after applying three complementary tests: (a) the McDonald-Kreitman test between hypoxia-selected and control flies for evidence of adaptive evolution based on correlation between fixed (i.e., >90% frequency) and nonsynonymous mutations in a population (p ≤ 0.05); (b) ≥1.5-fold transcriptomic change under hypoxia; and (c) sorting intolerant from tolerant (SIFT) evaluation of the impact of fixed, nonsynonymous mutations on the functions of proteins encoded by the genes in these regions (p ≤ 0.05). These included 12 genes that interact with, activate, or are targets of the Notch pathway. For example, there were two members of the Notch repressor complex (Hairless and HDAC4) that are located in the hypoxia-selected regions in the hypoxia-selected flies. As previous transcriptome profiling studies have demonstrated that the Notch pathway is upregulated in hypoxia-selected flies, the results obtained from both whole-genome resequencing and transcriptome profiling strongly suggested that this pathway plays an important role in hypoxia tolerance. Experimentally, the role of Notch activation in hypoxia tolerance was further determined in mutants carrying homozygous-viable loss-of-function (N[Ax-tsl] and N[fa-1]) or gain-of-function (N[spl-1]) Notch alleles (21, 33, 56). We found that the loss-of-function Notch mutants were hypersensitive to hypoxia and had a lower survival rate, even in much milder hypoxic conditions (6% O₂). In contrast, the gain-of-function Notch mutants exhibited remarkable hypoxia tolerance and survived even in the most severe hypoxic conditions (4% O₂), much like the hypoxia-selected flies but without adaptation (Figure 3).

Figure 3.

Role of Notch activation in hypoxia tolerance in Drosophila melanogaster. (a) Enrichment of fixed single-nucleotide polymorphisms (SNPs) and insertions or deletions (indels) in an extended Notch signaling pathway. Upregulated and downregulated genes in larvae are in cyan and yellow, respectively; genes showing no change in expression are in gray; and untested genes are in white. Genes for which at least one SNP and/or indel became fixed are indicated with stars; red stars denote genes located within a hypoxia-selected region, and blue stars denote all others. (b) Contribution of Notch signaling to hypoxia tolerance, as determined by culturing the homozygous-viable Notch mutants in either relatively mild (6% O₂) or severe (4% O₂) hypoxic conditions. Canton-S was used as a control. Flies with N[spl-1], a gain-of-function Notch allele, had a dramatically increased survival rate in hypoxic conditions. In contrast, flies with N[fa-1], a loss-of-function Notch allele, had little or no eclosion in the severe hypoxic condition and had a reduced survival rate even in the mild hypoxic condition. Adapted from Reference 63.

The notion that Notch signaling regulates survival under hypoxia was further strengthened by a current unbiased genome-wide screen of hypoxia-tolerant P-element insertion lines in our laboratory. In this study, we screened 2,187 lines covering 1,870 genes located on all chromosomes; of these, 44 lines representing 44 individual genes exhibited significantly higher eclosion rates (>70%) compared with those of the controls (~7–8%) under hypoxic conditions. These 44 genes can be linked to multiple signaling pathways, such as Notch, Wnt, Jnk, and Hedgehog. In particular, 20 of the 44 genes are linked to the Notch signaling pathway (5). These results, again, clearly demonstrate that Notch signaling is one of the major factors regulating development and survival under chronic hypoxic conditions. Given that multiple genes/pathways seem to be involved in different cell signaling systems, the contribution of crosstalk between these genes and pathways to hypoxia tolerance or susceptibility deserves further investigation.

Role of HIF Signaling in Drosophila

Hypoxia-inducible factor (HIF), a heterodimeric complex of two basic helix-loop-helix proteins of the PAS family, was first identified in mammalian cells (55). It regulates oxygen-dependent transcriptional activation of many mammalian genes, including those that control glucose metabolism, erythropoiesis, and vascular growth (41, 59). This oxygen-dependent transcription regulation mechanism was later found to be evolutionarily conserved in metazoans (53) (Figure 4).

Figure 4.

Overview of PHD and HIFα gene products in selected metazoan lineages. The left column of blocks shows the PHD proteins, and the right column of blocks shows the HIFα proteins. In the PHD blocks, “M” indicates the presence of a MYND-type zinc-finger domain. In the HIFα blocks, the number indicates the length of the protein in amino acids. For both proteins, the most ubiquitously expressed human paralogs (hsPHD2 and hsHIF-1α) were chosen as a reference for amino acid identity comparisons, which are indicated as percentage values (the catalytic domain for the PHD proteins, and the whole coding sequence for the HIFα proteins). In HIFα, “P” denotes proline hydroxylation motifs; when there is only one symbol, it aligns with the C-terminal ODD domain motif, and when there are two symbols, both N- and C-terminal ODD domain motifs are present. The arrows below these symbols indicate a change in the relative importance of the ODD domain motif in the regulation. The “partial sequence” boxes indicate sequence fragments from smooth dogfish (HIF-3α) and elephant shark (PHD3). Adapted from Reference 53.

The HIF transcription factor complex is regulated by oxygen concentration in the cell and mediated by the α subunits (HIF-1α, HIF-2α, and HIF-3α). In addition, oxygen-dependent regulation of the α subunits is mediated by enzymatic hydroxylation of specific amino acid residues in these proteins. Two distinct processes have been identified: A prolyl hydroxylation is catalyzed by a group of prolyl hydroxylases [prolyl hydroxylase domain–containing proteins 1, 2, and 3 (PHD1, PHD2, and PHD3, respectively)] (12, 16), and an asparaginyl hydroxylation is catalyzed by factor inhibiting HIF (FIH), an asparaginyl hydroxylase (27, 35). Under oxygen-limited conditions, the activity of these enzymes is reduced, which allows nonhydroxylated HIFα subunits to escape destruction, recruit the p300/CBP coactivator, and form a transcriptionally active complex. Previous studies have shown that HIF and the oxygen-dependent regulation system are conserved between Drosophila and mammals, and these studies also identified Drosophila homologs of HIFα [i.e., similar (sima)], HIFβ [i.e., tango (tgo)], and PHD [i.e., fatiga (fga)] (6, 37, 44). Unlike their vertebrate counterparts, flies carrying a null mutation in sima are fully viable in normoxic conditions (13).

The availability of a wide array of mutants and the simplicity of the methods used for gene silencing, transgenesis, and overexpression studies have provided an ideal framework for investigating the cellular basis of poorly understood aspects of HIF biology at a genomic scale. For example, an unbiased genome-wide RNA-interference (RNAi) screen in Drosophila cells identified 30 genes as critical HIF regulators in hypoxia, most of which had not been previously associated with HIF (14). The majority of hits belonged to just a few multiprotein complexes or signaling pathways, including the PI3K/TOR pathway, the eIF3 and eEF2 complexes, and the Brahma/SWI/SNF complex (chromatin remodeling). One of the remarkable hits from this screen was the argonaute 1 (ago1) gene, a central element of the microRNA translational silencing machinery. Using a double-stranded RNA–mediated knockdown strategy, the authors further confirmed that functional microRNA machinery is required for HIF-dependent transcription and cell viability under hypoxic conditions.

Mechanisms Regulating Growth and Metabolic Activities Under Hypoxia in Drosophila

Hypoxia-induced suppression of somatic growth is observed in many animal species, including rodents (42), humans (20, 22, 32), and flies (25, 65). For instance, we observed gradual reductions in body size and mass in the hypoxia-selected flies as we progressively lowered the oxygen levels in their environment from 8% to 4% O₂. The body masses of hypoxia-selected flies reared in the 4% O₂ environment decreased by 40%, and they exhibited changes in both cell number and size as measured in the wing. Of interest is that the reductions in body size and mass can be reversed to control levels when the hypoxia-selected flies are reared in normoxia, demonstrating that these reductions are controlled mainly by developmental mechanisms (65, 66). The convenience of manipulating genes involved in hypoxic signaling provides an opportunity and powerful tools to investigate the mechanisms regulating this developmental plasticity.

The evolutionarily conserved PI3K and TOR pathways play a central role in growth control and cell size determination (1, 36, 45). An unbiased genome-wide gain-of-function approach using EP-element screening of more than 4,000 EP lines found that activation of scylla, the Drosophila ortholog of human REDD1, suppresses PKB/PDK1-dependent growth under hypoxic conditions (51). As scylla expression can be induced by hypoxia (51, 66), this finding suggested potential crosstalk between growth regulation and oxygen-sensing mechanisms. It has been observed that the pupa size and growth rate are significantly reduced in flies carrying mutant fatiga, the Drosophila HIFα/sima negative regulator (13). The cells in fatiga loss-of-function mitotic clones in the larval fat body are clearly smaller than wild-type cells in the same organ. Conversely, overexpression of fatiga in wing imaginal discs leads to increased cell size (19). Further studies demonstrated that fatiga regulates cell growth and size through its downstream target, sima, by showing that the pupa size and growth rate can be restored in sima-fatiga double-mutant flies.

Although the sima-dependent role of fatiga in growth control is intriguing, a sima-independent mechanism cannot be excluded. For example, a genetic screen for modifiers of CycD/Cdk4-induced overgrowth in eye imaginal discs showed that fatiga mutation–induced suppression of cell growth does not involve upregulation of the Drosophila HIF complex (i.e., the sima/tango dimer) (19), suggesting that fatiga may play a sima/tango-independent role in cell growth control.

Reduction in energy demand is a common response of organisms to balance cellular energy homeostasis under hypoxic conditions. This energy-saving strategy includes an inhibition of macro-molecular synthesis (26), which may cause developmental delay and decrease in body size following prolonged hypoxia exposure. In mammalian cells, hypoxia suppression of protein synthesis is regulated at least in part by direct modifications of the mammalian TOR (mTOR) pathway and is independent of Akt/protein kinase B phosphorylation, AMP-activated protein kinase phosphorylation, ATP levels, ATP:ADP ratios, and HIF-1 activation (3, 34, 39, 61). These findings suggest that mTOR may directly link decreased oxygen levels with inhibition of protein synthesis. However, the signaling pathways modulating translational suppression during hypoxia are largely unknown. An unbiased genome-wide RNAi screen in Drosophila was recently performed to identify relevant genes that contribute to this translational suppression phenotype under hypoxia (38). This screen identified mitochondrial genes and gene regulators of translation such as two negative regulators of TOR, Tsc1 and Tsc2. Furthermore, this study also revealed the contribution of a tyrosine phosphatase, Ptp61F, to hypoxia responses, and identified it as a novel regulator important for the downregulation of TOR activity and translation during hypoxia (38).

One important compensatory response in hypoxia is metabolic suppression. For example, Hochachka et al. (28–30) have argued that anoxia-tolerant organisms depress their metabolism to minimize the mismatch between supply and demand. Although this idea is intuitively appealing, there was no information about how various metabolic enzymes could be coordinated to survive severe, long-lasting hypoxia. Taking advantage of the hypoxia-selected flies, we detected significant changes in a family of genes regulating cellular respiration and metabolism, with the overwhelming majority of these genes being downregulated (66). For example, besides one gene encoding a pyruvate kinase isoform (CG12229), most of the genes encoding glycolytic enzymes were dramatically downregulated. Similarly, the expression of genes encoding TCA cycle enzymes, lipid β-oxidation enzymes, and respiratory chain complexes was also significantly downregulated (Figure 5). Because the majority of the genes encoding TCA cycle enzymes were downregulated, we hypothesized that such downregulation was coordinated, possibly at a transcriptional level.

Figure 5.

Suppression of genes encoding glycolytic and TCA cycle enzymes in hypoxia-selected Drosophila melanogaster flies. (a) Schematic illustration of changes in genes encoding glycolytic and TCA cycle enzymes. Genes in green are downregulated, genes in red are upregulated, and genes in black show no significant change in expression. (b) Cluster map of changes in genes encoding glycolytic enzymes in control (naive) and hypoxia-selected larvae. (c) Cluster map of changes in genes encoding TCA cycle enzymes in control and hypoxia-selected larvae. Each cluster contains nine hybridizations of control fly samples and eight hybridizations of hypoxia-selected fly samples. Adapted from Reference 66.

Using bioinformatics, we therefore performed a study to identify which transcriptional regulators control the coordinated suppression by analyzing the cis-regulatory regions of these downregulated genes. In this analysis, the genes related to the TCA cycle were divided into two groups: one containing the significantly downregulated genes (the experimental group, comprising 16 genes) and one containing the insignificantly altered or upregulated genes (the reference group, comprising 7 genes). Of major interest was that the binding elements of the Drosophila transcriptional suppressor hairy were overrepresented in the regulatory regions of the downregulated genes (15 of 16) but not in the reference group (1 of 7). Indeed, the expression level of hairy was significantly upregulated in the hypoxia-selected flies.

These results suggested that hairy reduces the expression of the TCA cycle genes under hypoxic conditions. Furthermore, the physical binding of hairy to the cis-regulatory regions of the candidate TCA cycle genes was confirmed by a chromatin immunoprecipitation–polymerase chain reaction (ChIP-PCR) assay in Drosophila Kc cells treated with hypoxia. This hypoxia-induced suppression of TCA cycle genes was abolished in the hairy loss-of-function mutants (i.e., h¹ and h^1j3). Following these studies, we used these two hairy loss-of-function mutants to determine their survival rates in relatively mild hypoxic conditions (6% O₂) (66). We found that both mutants had a much lower survival rate compared with the controls, which proved the role of hairy in metabolic suppression and hypoxia tolerance in flies.

HUMAN HIGH-ALTITUDE DWELLERS: GENETIC AND GENOMIC STUDIES

Studies on the Genetic Basis of High-Altitude Human Traits

It is estimated that more than 140 million people live at high altitudes (>2,500 m) (47). Three large high-altitude populations (Andeans, Tibetans, and Ethiopians) have been studied and have adapted in diverse ways to these harsh hypoxic environments. Physiological studies have found distinct physiological traits in each population, suggesting that the adaptive phenotypes might have evolved independently and use different mechanisms. For example, hemoglobin concentrations are elevated in high-altitude Andean populations but not in Tibetans or Ethiopians, and oxygen saturation is reduced in Andeans and Tibetans but not in Ethiopians (7). The heritability estimates for these traits also vary among high-altitude populations. For example, hemoglobin concentration has strong heritability in Tibetans, but oxygen saturation does not show the same strong heritability (7).

Although it has long been proposed that there are genetic factors for adaptation to high altitude, these factors have remained largely unknown. Some studies have recently started to address this. Moore and colleagues (57) analyzed ~11,000 genome-wide single-nucleotide polymorphisms (SNPs) and identified a subset of variants located in genes involved in the HIF pathway as well as other targets of positive selection in the high-altitude Andean population, including ADRA1b (alpha-1B-adrenergic receptor), EDN1 (endothelin 1), NOS2A (nitric oxide synthase 2), and PHD3. Other studies using a much larger set of SNPs (~500,000) also identified a subset of genes involved in the HIF pathway, namely CDH1 (cadherin 1), EDNRA (endothelin receptor A), ELF2 (E74-like factor 2), NOS2A, PIK3CA (phosphoinositide-3-kinase, catalytic, alpha polypeptide), PRKAA1 (protein kinase, AMP-activated, alpha 1 catalytic subunit), TNC (tenascin C), and VEGF (vascular endothelial growth factor) (9, 11, 48, 58). EGLN1 (egl nine homolog 1)—encoding the enzyme EGLN1, also known as HIF prolyl hydroxylase 2, which catalyzes the posttranscriptional formation of 4-hydroxyproline—was also identified as one of the candidate genes.

Additional studies have identified candidate genes and regions that correlate to high-altitude adaptation in Tibetan populations (58, 62). Simonson et al. (58) identified several loci, including EGLN1, HMOX2 (heme oxygenase 2), CYP17A1 (encoding a cytochrome P450 enzyme), PPARA (peroxisome proliferator-activated receptor alpha), PTEN (encoding a phosphatase), EPAS1 (endothelial PAS domain protein 1), CYP2E1 (encoding a cytochrome P450 enzyme), EDNRA, ANGPTL4 (angiopoietin-like 4), and CAMK2D (calcium/calmodulin-dependent protein kinase II delta). In addition, the authors used haplotypes (sets of associated SNPs) with extended haplotype homozygosity (a target of selection) to test for association between genotypic and phenotypic variations.

At the same time, Yi et al. (62) reported a study of sequence variation in exomes and flanking regions in high-altitude Tibetans and identified a candidate intronic SNP in EPAS1 that is significantly associated with hemoglobin levels and erythrocyte count. Similarly, Beall et al. (9) identified a signal of positive selection in a sample of high-altitude Tibetans at the EPAS1 locus. Additional analyses were carried out that demonstrated a significant association between the EPAS1 SNPs and hemoglobin concentration in two separate samples of high-altitude Tibetans (54, 62), and similar studies have been performed in high- and low-altitude Ethiopian populations (54). A set of candidate genes for positive selection was actually identified in the high-altitude population, including genes such as CBARA1, VAV3, ARNT2, and THRB. Although most of these genes had not been identified in previous studies in Andean or Tibetan populations, two of them (ARNT2 and THRB) play a role in the HIF-1 pathway, which had been implicated in the previous Andean and Tibetan studies.

Although the population genomic studies described above have furthered our understanding of the genetic basis of adaptation to high altitude, the increased accessibility of next-generation sequencing, advances made using bioinformatics tools and functional assays, future work characterizing sequence variation in the coding regions, and analysis of variants in the transcription regulatory regions (combined with gene expression profiling) will play an important role in discovering the basis for high-altitude adaptation in humans. One of the important issues at high altitude is that a certain percentage of people adapt in a way that is not beneficial to long-term survival at such high altitude, as illustrated by the high incidence of myocardial infarction and stroke (43). Indeed, they adapt very much as humans do who travel to acute high altitudes, increasing O₂ carrying capacity by increasing the hematocrit and hemoglobin at the risk of vessel occlusion and reduction of blood flow to vital organs (maladaptation). Patients who have such deleterious adaptation also have a variety of nervous system problems (such as headaches, mental confusion, and paresthesia) and may end up with the whole syndrome of chronic mountain sickness (CMS), which is debilitating. Although most of the world’s high-altitude populations have adapted rather well, a certain percentage of individuals living at high altitude have CMS, with more individuals suffering from it in the Andes than in either Tibet or Ethiopia. These populations—both those who adapt well and those who do not—provide an opportunity to study the nature of adaptation, including the genetics involved and whether we can elicit potential drug targets for humans at high altitude and even at sea level.

A number of studies have focused either on high-altitude dwellers in comparison with sea-level populations or on differences between CMS and non-CMS populations in comparison with sea-level populations. However, these studies unfortunately fell short in three areas. First, to our knowledge, there have been no studies other than our own that have carried out whole-genome sequencing of patients from CMS and non-CMS populations. There is a distinct possibility that structural variants could exist in coding as opposed to noncoding regions; it may be more difficult to functionally characterize these variants and interpret their significance, but without whole-genome sequencing, they will be missed entirely. Two sequencing-based partial genomic analyses have been performed in high-altitude populations (46, 62), but as indicated above, such studies surveyed only a small portion of the genome (i.e., exons and flanking regions) or a focused group of candidate genes, and the interpretation of these results is limited. Second, the studies that have been performed often choose the gene(s) of interest, and such studies therefore may not be entirely unbiased. Third, there has been no experimental proof that genes that have a differential frequency distribution among the various populations at high altitude and at sea level are the basis of the adaptation. This has been a major issue because drug targets would have to be proven essential for the adaptation before they could be used as drug targets.

Whole-Genome Sequencing in High-Altitude Human Dwellers

We previously performed the first whole-genome-based, unbiased analysis of genes that contributed to hypoxia adaptation in our hypoxia-selected Drosophila strain, which was generated through laboratory evolution (63). This study allowed us to examine the complete genome at single-nucleotide resolution and to detect fine changes in the allele frequency spectrum consistent with natural selection. We have recently extended the analytical strategy developed in this previous study and used it for a whole-genome resequencing-based analysis to identify genes that modulate high-altitude adaptation in humans. We focused on 17 high-altitude (~3,500 m) native Ethiopian residents, as Ethiopian highlander populations have been found to be well adapted to high altitudes (8, 10). Specifically, we sequenced 10 individuals of Oromo heritage living in the Bale Mountains (labeled “Oromos”) and 7 individuals residing on the Chennek Plateau in the Simen Mountains (labeled “Simen”). We also tested whether particular genes highlighted by the study play a role in adaptation to low-oxygen conditions, by using RNAi to target their respective orthologs in Drosophila.

In the Ethiopian population, we searched for regions with evidence of a selective sweep: a loss of genetic diversity and computed test statistics that measure this loss in diversity. Of the 10 regions that proved to be statistically significant, only 1 showed up as significant in both the Oromos and Simen populations. This gene-rich 208-kb region on chromosome 19 contains a block of 135 differential SNPs with a significant change in frequency relative to that of the control populations. Eight genes in this region point to many intriguing candidates. For example, the differential SNPs include two missense mutations in an enzyme that is essential in lipid metabolism (lipase). Other genes include a transcriptional suppressor involved in early organ development, a gene associated with angiogenesis, and a gene associated with coronary artery disease and organ development. Thus, these results point to a cluster of putative hypoxia response genes. As these genes are associated with phenotypes such as lipid metabolism, transcription regulation, and angiogenesis, they illustrate the potential for a variety of human adaptive mechanisms to high altitude (8).

To provide further evidence of the role of these genes in hypoxia, we used Drosophila as a model organism to test the hypothesis that the differential regulation of expression of the orthologs in flies has an effect on hypoxia tolerance or susceptibility. The fixation (or near fixation) of SNP variations in the candidate genes may cause either gain- or loss-of-function changes. For conceptual and practical reasons, we first used the UAS-RNAi/Gal4 system in Drosophila to analyze whether the downregulation/knockdown of the Drosophila orthologs representing the human candidate genes located in the selected region on chromosome 19 affects hypoxia tolerance. Upon testing the fly orthologs, we found that three of the human genes improved hypoxia tolerance in flies, increasing the survival rate by approximately 40–80%, a two- to fourfold increase over that of control flies. Thus, this study demonstrated that genes identified from whole-genome sequencing in humans can indeed affect hypoxia tolerance in a model system, and provided evidence for the importance of these genes obtained from human studies.

SUMMARY

The accelerated pace of discovery in the present scientific era is well illustrated in investigations related to hypoxia tolerance and susceptibility, whether at sea level or at high altitude. Next-generation sequencing and advances in bioinformatics have provided the tools to deepen our understanding of disease. Clearly, the impetus for these investigations comes from the need to mitigate, eliminate, or better treat various diseases or conditions. The morbidity and mortality from respiratory and cardiovascular diseases such as stroke, myocardial infarction, diabetes and vascular insufficiency, renal ischemic disease, and diseases in childhood and adulthood pertaining to hypoxic and ischemic attacks or episodes take a global toll and have immense consequences second to no other condition. Research that could address these medical problems through better treatments or cures could have a major impact on societal health.