Evidence for a Genetic Basis for Altitude Illness: 2010 Update

Individual patterns of susceptibility

Some individuals are predisposed to developing altitude illness, and previous affliction is one of the best predictors of subsequent affliction (Bartsch et al., 2001); however, care must be taken when categorizing an individual as altitude-illness susceptible versus resistant based on a single occurrence of the disease because, under extreme conditions (e.g., rapid ascent rate to very high altitude), most, or perhaps all, individuals will develop some form of altitude illness. Individuals of interest to many researchers in the field are those who repeatedly develop an altitude illness under mild conditions (e.g., slow ascent rate to moderate altitudes). The ability to divide populations into altitude tolerant versus altitude sensitive (or altitude illness− vs. altitude illness+) is an important tool in the search for a genetic component to acclimatization. Assignment into these categories is based on two basic criteria: (1) the presence or absence (or severity) of the condition and (2) the exposure. The former is generally standardized [e.g., use of the Lake Louise Score (LLS) for AMS], but the latter varies greatly among studies, in large part owing to the opportunistic nature of the studies (in most cases, the altitude and rate of ascent are not selected by the researcher). One potential strategy to address this problem would be to develop an exposure-severity metric that would quantify the environmental trigger. Again, to use AMS as an example, subjects could then be divided into susceptible (high LLS, low exposure score) versus tolerant (low LLS, high exposure). Such a binary scoring system would facilitate both intrastudy phenotyping and interstudy comparisons.

To demonstrate HAPE susceptibility, Dehnert and colleagues (2002) took a cohort of 76 mountaineers to 4559 m. Of those with a history of HAPE (approximately half of the group), 66% developed the condition on ascent, while there were no cases among those without previous history (Dehnert et al., 2002). A study of summiteers of Mt. Whitney (4420 m) reported that a history of AMS was a risk factor for developing this condition in their cohort of climbers (Wagner et al., 2006). These data are consistent with individual susceptibility; however, whether this is because the cohorts have some genetic factor in common or whether they share some environmental exposure or developmental event (or some combination of any of these factors) is unknown.

Identifying a common pathophysiology in a susceptible cohort is key to identifying potential candidate genes for study. For example, abnormally high pulmonary artery systolic pressures (PASP) in exercise and hypoxia have been demonstrated as potential risk factors for HAPE (Dehnert et al., 2005). Measuring PASP as a screening test to predict HAPE is not practical, and Dehnert's group was not able to show that it was a reliable predictor of HAPE susceptibility; however, genetic variations that affect pulmonary artery pressure response to stressors such as exercise have potential as candidates when looking for genetic contributors to a predisposition to HAPE. Similar strategies can be employed to identify other candidate genes based on other pathophysiological mechanisms implicated in HAPE and other altitude illnesses.

Familial patterns of susceptibility

Mendelian patterns of inheritance, familial clustering, and high concordance between individuals of similar genetic makeup are the classic markers of a genetically influenced trait; however, there is little evidence for or against this in the altitude-illness literature. In our previous review (Rupert and Koehle, 2006), we discussed the limited epidemiological studies that showed some familial clusters (see Table 1). The early data suggesting genetic transmission of susceptibility or resistance to altitude illness are tantalizing, but far from definitive.

In 2009, Lorenzo and colleagues described a small Han Chinese pedigree in which the proband (an 11-yr-old boy), his father, paternal grandfather, aunt, uncle, and cousin had a history of HAPE. The boy was from Chendu (500 m) and had experienced HAPE four times while visiting family in Lhasa (3648 m). Although no definitive conclusions can be drawn from the small pedigree, the pattern of inheritance is consistent with autosomal dominance with incomplete penetrance. Haplotype analysis for 34 genes of interest (see Table 1) was performed with the affected individuals, but all the genes investigated could be excluded as noncorrelative.

The overall lack of familial data in the field may simply be due to the paucity of family-exposure events (especially as there is likely a tendency to return to low altitude as soon as one family member gets ill). Hopefully, the development of registries and databases for altitude illnesses (e.g., the International HAPE Database; http://www.altitude.org/hape.php) will reveal patterns of affliction and contribute to further understanding of the genetic epidemiology of the conditions.

Population patterns of susceptibilities

Humans have occupied high altitude regions of the Andes, the Himalaya, and the East African highlands for thousands of years (reviewed in Rupert and Hochachka, 2001). Part of the adaptive response to hypoxia may have been selection against any genetic element that contributed to altitude- illness susceptibility, either among the initial migrants or in their offspring. A comparison between Tibetans and Han Chinese reported a lower frequency of AMS among the highlanders, which the authors postulate was owing to genetic adaptation (Wu et al., 2005). Resistance to AMS among Tibetans was also noted in a study of workers on the high altitude sections of the Qinghai–Tibet railroad (Wu et al., 2009). Wang and colleagues (2010a) reported that an allele in NOS3 (the gene encoding endothelial nitric oxide synthase) that they previously had reported to be associated with AMS in a Nepalese population (Wang et al., 2009) was less common in Andean Quechua compared with lowland Amerindians. The authors postulated that selection against AMS susceptibility may have occurred during the colonization of the Andean Altiplano. In Asia, an allele (5160A) in the CYP11B2 gene (which encodes aldosterone synthase) was associated with HAPE in lowland Indians (Qadar Pasha et al., 2005; see Table 2a); however, the same allele was not overrepresented in Himalayan natives compared to Indian lowlanders (Rajput et al., 2006), although an allele at a different polymorphic locus in the gene (−344T) was more common in the highlanders.

Separating adaptive changes in populations from developmental changes in individuals is challenging, but could be addressed by examining the susceptibility to AMS in people of highland ancestry born and living at low altitude. Such studies are challenging because the potential subject pool is limited, differences in environmental conditions (in addition to atmospheric pressure) are difficult to control, and the reason for the downward migration itself could be a confounder. Also to be considered is that the different adaptive strategies in the different highlander populations (e.g., see Beall, 2000) could result in differences in acclimatization among descendants born and raised at low altitudes.

CMS is believed to be less common in Tibetans than in similarly exposed Andeans (see review by Moore et al, 1998). Recent genomic studies in Tibetans have revealed a number of highly selected genes that play a role in maintaining hematocrit that may contribute to this apparent resistance, including EGLN1 (which encodes a prolyl hydroxylase in the HIF pathway; Simonson et al., 2010), EPAS1 (which encodes Hif 2a; Beall et al., 2010; Yi et al., 2010), FANCA, and PKLR (involved in red blood cell (RBC) production and maintenance (Yi et al., 2010). None of these genes has been tested for an association with CMS in Tibetans; however, their selection for in a CMS-resistant population ^* and their role in the regulation of erythropoiesis (variants in both EPAS1 and EGLN1 were associated with hematocrit and/or hemoglobin concentration) make them promising candidates for such a study. EPAS1, as well as a number of other genes in the HIF pathway, has been investigated in CMS in Andeans (Mejia et al., 2005; see Table 2c) and no association was observed; however, the sample size in the study was modest. Resistance to CMS alone seems unlikely to be a strong driver of selection, because the disease tends to occur later in life and the Andeans have successfully colonized the Altiplano despite being relatively susceptible to the condition. The low rates in CMS in Tibetans may be genetic, but as a secondary consequence of selection against higher hematocrits that is compensated for in another manner in the Andeans.

Polymorphisms, association studies, and candidate genes

One of the most common approaches used to investigate the role of specific genes in complex traits is the candidate-gene association study. The principles, as well as the strengths and limitations of this study design, were the subject of a series of didactic reviews (Attia et al., 2009a,b,c), but a brief description follows. Genes are selected as candidates for two reasons, either an a priori reason (often based on the known or predicted function of the gene product) to believe that the gene product may contribute to the phenotype (the trait or characteristic of interest) or because the gene is in a region of the genome shown to be associated with the phenotype [the latter approach is becoming more common as genome-wide association studies (GWAS) are being used to simultaneously interrogate regional associations across all chromosomes]. Once the gene is selected, individuals with and without the trait are genotyped for variants (alleles) of the candidate gene, and the correlations between alleles and the phenotype of interest are tested. These variants are typically single nucleotide polymorphisms (SNPs), insertion–deletion polymorphisms, or a variable number of tandem repeat polymorphisms (VNTRs). ^*

Of great importance is the correct classification of each subject's phenotype (in this instance, the diagnosis and possible assessment of the severity of the specific altitude illness). Standardized methods and procedures are available for the diagnosis of HAPE, AMS, high altitude pulmonary hypertension (HAPH), and CMS, but different studies may choose to assess patients differently, which can make comparisons between studies difficult. In addition, there may be factors commonly experienced at altitude other than the previously discussed altitude and ascent rate that might trigger, exacerbate, or complicate the diagnosis of altitude illness (e.g., cold, fatigue, dehydration, iron status, changes in diet, and time to testing venue). One advantage of using hypoxic chambers in altitude-illness research is the ability to control for these potentially confounding variables.

A number of potential limitations to candidate-gene association studies need to be considered when designing or evaluating such studies. First, and perhaps most importantly, gene associations are correlational rather than causal, and follow-up studies are required to establish the actual mechanisms involved (e.g., whether the allele of a hormone-encoding gene associated with a condition alters the level or function of the hormone). Errors in multiple testing are problematic for association studies, especially when a number of polymorphisms and/or a number of genes are tested or when post hoc stratification of the subjects is performed. To maintain an acceptable overall false positive rate, statistical adjustments such as the Bonferonni correction are often required. Subjects (cases and controls) should be biogeographically matched as closely as possible, because the frequency of genetic variants can differ greatly across populations, which can skew the results of a study. The advent of AIMs (ancestry-informative markers; see Tian and colleagues, 2008, for a review) will greatly facilitate quantification of sample stratification as well as reduce the reliance on self-reports or anecdotal data when defining the genetic backgrounds of study cohorts. Finally, an important test of a positive association study is reproducibility; however, because associations can be population specific owing to different patterns of linkage disequilibrium or differential representations of other interacting alleles (i.e., in a gene–gene interaction model) and because phenotypes can be influenced by a variety of environmental factors, failure to reproduce the association is not conclusive evidence of a false positive.

The following section reviews the genes that have been investigated for associations with altitude illness since 2006. We have divided the genes into two categories: (1) genes that have been previously investigated and were reviewed in Rupert and Koehle (2006), in which case we briefly discuss how the new data fit with the data from the earlier study and (2) genes investigated since 2006 as potential contributors to the etiology of altitude illness. Tables 2a–d are expansions of the table in Rupert and Koehle (2006) and summarize gene studies in the various categories of acute and chronic altitude illnesses. Table 3 presents a brief description of the candidate genes and the rationale behind choosing them for study.

Genes in the renin–angiotensin–aldosterone pathway: AGT, ACE, AGTR1, CYP11B2

The renin–angiotensin–aldosterone system (RAAS) plays a key role in maintaining fluid balance and regulating blood pressure. Because changes in blood flow and pressure in response to hypoxia likely play a role in altitude acclimatization, a number of genes encoding components of this pathway have been investigated in altitude-illness, gene-association studies. Since our 2006 review, AGT, ACE, AGTR1, and CYP11B2 have been revisited in several different populations.

Variants of AGT, ACE, AGTR1, and CYP11B2 were previously investigated for a role in HAPE, and the latter three genes were associated with HAPE. Among a cohort with repeated experiences in ascending to altitudes over 3500 m, Charu and colleagues (2006) found the ACE D allele (i.e., pooled I/D, D/D genotypes) to be overrepresented among HAPE-susceptible individuals. In the same study, the ACE I/I genotype was more common in HAPE-resistant individuals, albeit only in conjunction with specific EDN1 genotypes [longer (CT)n-(CA)n microsatellite repeats genotype and the G2288T G/G genotype), suggesting an interaction between the two genes. In a separate study, the association between HAPE and two polymorphisms from each of five RAAS genes (ACE, AGT, and CYP11B2, discussed in this section, and BDKRB2 and ACE2, discussed below) were tested in a Han Chinese population working on a railway at high altitude (Qi et al., 2008). The A allele and A/A genotype of the A240T ACE polymorphism (which had not been previously evaluated in HAPE) were associated with susceptibility, but variants of the ACE A2350G polymorphism were not (Qi et al., 2008). Of note, the A240T ACE polymorphism had a greater effect on serum ACE than the commonly assessed and previously associated ACE I/D polymorphism (Zhu et al., 2001). As was the case in a previous study (Qadar Pasha et al., 2005), polymorphisms of the AGT gene were also not associated with HAPE susceptibility in this study (Qi et al., 2008). Finally, alleles at two CYP11B2 gene polymorphisms that had not been previously investigated were strongly associated with HAPE. The T allele (C-344T) and the A (Lys) allele (LYS 173Arg), as well as the homozygote genotypes of both alleles, were overrepresented in HAPE-susceptible individuals; however, because alleles at these two polymorphisms are in strong linkage disequilibrium, ^** these may not be independent outcomes. Previously, variants from two different CYP11B2 gene polymorphisms (C5160A and intron 2 conversion) were associated with HAPE (Ahsan et al., 2004; Qadar Pasha et al., 2005). It is unclear which CYP11B2 gene variants are functionally linked to HAPE susceptibility and which variants may be nonfunctional markers linked to functional polymorphisms.

Associations between RAAS gene variants and AMS are inconsistent among populations and between studies. There was a weak association (associated on day 1 but not day 2 of a trek) between the ACE I/D genotype and AMS in a European population, but the authors stated that the observed heterozygote advantage was biologically implausible, and the genotype did not correlate with summiting success. The ACE D and AGT 235M alleles were each associated with AMS in a Han Chinese population, although the description of AMS diagnosis was unclear and seemed to include a variety of illnesses (Buroker et al., 2010). Conversely, two later studies failed to reproduce this association in both Nepalese (Koehle et al., 2006) and European populations (Kalson et al., 2009). These latter studies are in agreement with the earlier findings of Dehnert et al. (2002), who sampled a European population. There were also no associations between AMS and additional ACE gene polymorphisms in a Nepalese population (Koehle et al., 2006) or between AMS and angiotensin II receptor gene (AGTR1) polymorphisms in Nepalese (Koehle et al., 2006) and Han Chinese (Buroker et al., 2010) populations. Comparing studies is somewhat problematic because the altitudes achieved (3817 m, Tsianos et al., 2005; 4380 m, Koehle et al., 2006; 2700–5895 m, Kalson et al., 2009; 4559 m, Dehnert et al., 2002; 3670 m, Buroker et al., 2010) and the rates of ascent differed.

A previous association between HAPH and the ACE I allele was reproduced in two separate studies. The ACE I allele and I/I genotype were significantly overrepresented in Kyrgyz HAPH patients compared with unaffected Kyrgyz highlanders (Aldashev et al., 2002), as was the I-G-A haplotype (incorporating the I, G2215, A2350 alleles) (Aldashev, 2007). Tests of independent association for variants of the three polymorphisms were not presented for the second study, so the contribution of individual ACE gene polymorphisms is not entirely clear.

The AGT 235 M allele was associated with CMS in a Tibetan population living between 3600 and 4400 m, but variants of the ACE and AGTR1 genes were not associated (Buroker et al., 2010).

EDN1, endothelin-1

Vascular endothelial cells express the EDN1 gene, which encodes the vasoconstrictor endothelin. There was no correlation between repeat length at the EDN1 (CT)n-(CA)n microsatellite polymorphism and HAPE susceptibility (Charu et al., 2006), but, as previously mentioned, there may be an interaction between EDNI (CT)n-(CA)n and ACE I/D alleles. An association between EDN1 gene variants and HAPE was evident for the T2288G polymorphism: the T/T genotype (and pooled G/T + T/T genotypes) was overrepresented in HAPE-susceptible individuals. Further, the combination of the G/G genotype and ACE I/I genotype was associated with HAPE resistance (discussed later; the T2288G polymorphism was not examined in earlier studies). Variants of the two other EDN1 polymorphisms (−/A and G594T) were not associated with HAPE (Charu et al., 2006). The population was not reported in this study, but based on other studies from this group of authors, the subjects were likely Indian.

NOS3, endothelial nitric oxide synthase (eNOS)

The NOS3 gene encodes endothelial nitric oxide synthase (eNOS), which is responsible for the production of NO, a gaseous hormone involved in vasodilatation. Although variants of the NOS3 gene were previously associated with HAPE in two separate studies involving Indian (Ahsan et al., 2004) and Japanese populations (Droma et al., 2002), no attempts to reproduce these findings have been published since. In an attempt to look at functional implications of variants in the NOS3 gene, Smith and colleagues (2006) compared the rise in pulmonary artery systolic pressure (PASP) on ascent to 5200 m among individuals with different variants of the NOS3 gene. A rise in PASP has been demonstrated to be a potential contributor to the pathophysiology of HAPE, so one might expect a larger rise in PASP in those individuals with the genotypes associated with HAPE susceptibility. Unfortunately, in the Smith study, variants at the NOS3 gene were not associated with PASP; thus, a functional mechanism for this genetic association remains elusive.

With respect to AMS, the T allele (T/T and G/T genotypes combined) of the G894T polymorphism, which was associated with HAPE in two of three studies, was also associated with AMS in a study of Nepalese pilgrims at 4380 m (Wang et al., 2009). This allele was previously associated with lower concentrations of exhaled nitric oxide in asthmatics (Storm van's Gravesande et al., 2003), which could explain the physiological connection between the T allele and susceptibility to AMS. Six other NOS3 polymorphisms had no association with AMS (Wang et al., 2009).

Variants of the G849T and the 27 bp VNTR polymorphism were also assessed in a Kyrgyz population with and without HAPH; however, no associations were found (Aldashev, 2007).

Heat shock protein genes: HSPA1A and HSPA1B, heat shock 70 kDa proteins 1A and 1B

Heat shock proteins are a group of intracellular proteins that are upregulated during times of stress (e.g., heat, hypoxia, and oxidative stresses). Heat shock protein genes were not previously investigated in HAPE. Qi and colleagues (2009) found that the alleles and genotypes of the A-110C polymorphism of the HSPA1A gene and the alleles of the A1267G polymorphism of HSPA1B were unequally distributed between Chinese individuals with and without HAPE. Further, using haplotype analysis to compare the relative risk of HAPE among co-inherited alleles, individuals with the G-C-A [A1267G, G+190C (HSPA1A) and A-110C polymorphisms, respectively] and G-G-A haplotypes were more resistant to HAPE compared with those with the most common haplotype, A-G-A. When controlling for several other factors (e.g., age, BMI, etc.), these relationships remained significant, and the A-C-C haplotype showed a significantly elevated risk of HAPE when compared with the A-G-A haplotype. The risk associated with three other haplotypes was not significantly different from the A-G-A haplotype before or after correction. A diplotype analysis (comparing the relative risk of HAPE for combinations of haplotypes) showed an increased risk of HAPE for the A-G-A/A-C-C diplotype (a diplotype is two haplotypes) compared with the most common diplotype, A-G-A/A-G-A. After correction for other factors, this relationship was not statistically significant. The risk of HAPE for seven other diplotypes was not significantly different when compared to the most common diplotype, regardless of statistical adjustments. The functional role of two promoter polymorphisms (A-110C and A1267G) in gene expression was assessed to determine if either HSPA1A polymorphism could affect the phenotype. The HSPA1A A-110C polymorphism was a determinant of gene expression, with greater levels of expression from the −110A allele. This finding corroborates the positive genetic association, because the −110C allele was associated with increased risk of HAPE, and the HSPA1A gene product likely has a protective role.

Heat shock protein gene polymorphisms were previously associated with AMS (Li et al., 2004). Zhou and colleagues (2005) examined the same polymorphisms in the HSPA1A (i.e., hsp70-1) and HSPA1B (hsp70-2) genes that were previously investigated (Li et al., 2004), as well as polymorphisms in the HSPA1L (hsp70-hom) gene that were not previously studied. Similarly to Li and colleagues (2004), variants of the “b1/b2” (G+190C) polymorphism of the HSPA1A gene were not associated with AMS, and the HSPA1B “B/B” (A+1267G) genotype was associated with AMS. The HSPA1L “B/B” (G2437C ^* ) genotype was overrepresented in the AMS group. Subjects in both studies were ethnic Chinese.

Recently, a role for oxidative stress in the pathophysiology of AMS has gained support (reviewed by Bailey et al., 2009), and the association of heat shock protein gene polymorphisms with AMS supports this speculation, because their proteins allow cells to cope with the oxidative stress resulting from ischemia and reperfusion (see review by Misra et al., 2009). Reactive oxygen species (ROS) are highly reactive molecules containing an oxygen atom with unpaired valence electrons. ROS are important in cell signaling pathways, but can have harmful effects at elevated levels. Cells manage ROS with a series of enzymes, including superoxide dismutase (SOD), glutathione reductase, and catalase (reviewed in Scandalios, 2005), and these enzymes are likely involved in hypoxia management; for example, knocking down manganese–SOD gene expression in oral squamous cell carcinoma cells resulted in increased ROS, upregulation of HIF-1α mRNA, increased HIF-1α protein production, and inhibition of the VHL protein, indicating that ROS are important in the HIF-1 pathway (Sasabe et al., 2010). These data support the candidacy of genes in the antioxidant, or oxidative damage repair, pathways in future association studies.

BDKRB2; bradykinin receptor β2

The BDKRB2 gene encodes the bradykinin β2 receptor, which is the receptor for the vasodilator bradykinin peptide. Bradykinin and its receptor are part of the RAAS: upon binding to its receptor, bradykinin triggers a series of responses that ultimately decreases blood pressure. This gene has been investigated in HAPE and AMS. For both illnesses, the same polymorphisms (C-58T and ± 9 bp in exon 1) were investigated, and no associations were found (Qi et al., 2008; Wang et al., 2010b).

VEGFA; vascular endothelial growth factor A

Vascular endothelial growth factor A (VEGFA) is a mitogen that acts on endothelial cells, and the encoding gene (VEGFA) is upregulated by hypoxia. VEGFA is involved in vascular development and angiogenesis, and VEGFA also increases vascular permeability. Further, overexpression of VEGF in the lungs contributes to pulmonary edema in mice (Kaner et al., 2000). Although genetic variants affect the expression of VEGFA (e.g., Watson et al., 2000), Hanaoka and colleagues (2009) did not detect any association between five VEGFA variants and the development of HAPE in a Japanese population. VEGF has been hypothesized to play a role in HACE (Xu and Severinghaus, 1998), but no association studies with this gene and HACE have been published.

The effect, if any, of VEGF in AMS is unclear. Plasma VEGF was lower in mountaineers without AMS at 14,200 ft compared to sea-level controls (Maloney et al., 2000), and VEGF increased with ascent to high altitude in subjects with AMS, but not in subjects who remained well at altitude (Tissot van Patot et al., 2005). In contrast, other studies showed no difference in plasma VEGF between subjects with and without AMS (e.g., Dorward et al., 2007; Nilles et al., 2009). To our knowledge, VEGF gene polymorphisms have not been tested for an association with AMS.

ADRB2; β₂ adrenergic receptor

The β₂ adrenergic receptor is a G protein coupled receptor that acts as the primary catecholamine receptor in the lungs. Generally, plasma epinephrine increases at altitude and, by acting through β₂ receptors, facilitates oxygen uptake and delivery to tissues (Mazzeo and Reeves, 2003). Although Loeppky and colleagues (2003) showed that AMS patients (Americans, ethnicity unstated) had greater adrenergic tone (e.g., increased peripheral vasodilatation and heart rate) compared with controls, no association was observed between haploptyes encompassing the ADRB2 gene and AMS in a Nepalese population (Wang et al., 2007).

The A-654G ADRB2 promoter polymorphism was associated with HAPE in an Indian population, but the A46G and the G79C polymorphisms, along with five other polymorphisms, were not associated with HAPE when examined independently (Stobdan et al., 2010). Further analysis revealed that several haplotypes (some containing the A46G and G79C polymorphisms) were associated with HAPE susceptibility and resistance. The predictive capacity of the A46G and G79C polymorphisms was haplotype dependent, suggesting a model in which the phenotype results from the interaction of multiple alleles. These data demonstrate the importance of haplotypic analysis, because a study on only one of these SNPs would not have detected the effect.

Two functional missense polymorphisms in ADRB2 have been investigated as potential contributors to HAPH: the A46G and G79C (gln27glu) polymorphisms. These alleles have been associated with receptor sensitivity (e.g., the gly16 allele is associated with greater receptor downregulation; Green et al., 1994) and decreased vascular agonist response (gln27gln genotype; Cockcroft et al., 2000), both of which may contribute to the development of HAPH (reviewed in (Aldashev, 2000). Although neither allele at the ADRB2 A46G (arg16gly) polymorphism was associated with HAPH, G/G homozygotes for G79C were overrepresented in Kyrgyz highlanders with the condition compared with unaffected Kyrgyz controls (Aldashev, 2007).

GSTM1, glutathione S-transferase mu 1; GSTT1, glutathione S-transferase theta 1

Glutathione S-transferases (GSTs) are a group of intracellular enzymes that detoxify endogenous and exogenous substances by the addition of reduced glutathione. They are divided into several classes: alpha, mu, pi, theta, and zeta (Ates et al., 2005). Expression of glutathione-S-trasferase is upregulated in highland populations (Gelfi et al., 2004), possibility owing to the need to deal with ROS-induced tissue damage. Two of the genes in the GST family, the mu (GSTM1) and theta (GSTT1) genes, encode cytosolic enzymes, and the deletion of these genes is common (∼50% and 13% to 20%, respectively; Garte et al., 2001). The presence of one of these genes (GSTT1) and the absence of the other (GSTM1) were independently associated with a higher prevalence of AMS in a Chinese population (Jiang et al., 2005). The GSTM1 null genotype (−/−) and the GSTT1 positive genotype (+/−, +/+) were both significantly overrepresented in subjects with AMS compared with control subjects. Subjects who were both GSTM1 and GSTT1 null were at five times greater risk of developing AMS than subjects who were GSTM1 positive (+/+ or +/−) and GSTT1 null.

MTHFR, 5,10-methylenetetrahydrofolate reductase

The protein product of the MTHFR gene catalyzes the conversion of 5,10-methylenetetrahydrofolate to 5-methylenetetrahydrofolate, which is a substrate involved in the conversion of homocysteine (toxic) to methionine. MTHFR deficiency results in increased homocysteine and the condition known as hyperhomocysteinemia, which has been linked to certain cardiovascular diseases (Wald et al., 2002). Further, endothelium-dependent vasodilation by NO was lower in patients with hyperhomocysteinuria (Aldashev, 2007). The T allele of the C677T polymorphism is associated with markedly decreased enzyme function (35% and 70% lower in C/T heterozygote and T/T homozygotes, respectively; Frosst et al., 1995). The T allele and T/T genotype were overrepresented in a HAPH Kyrgyz cohort, perhaps owing to increased homocysteine in the blood causing hypoxia-induced endothelial dysfunction (Aldashev, 2007).

CDKN1B, cyclin-dependent kinase inhibitor 1B (p27, kip1)

CDKN1B encodes p27 kip1, a cyclin-dependent kinase inhibitor. p27 kip1 inhibits the cell cycle by binding to and preventing the activation of E-CDK2 and cyclin D-CDK4. Of importance to HAPH, this protein inhibits the growth of pulmonary artery smooth-muscle cells and is involved in hypoxia-induced pulmonary hypertension (Yu et al., 2005). The A allele of the C838A polymorphism, which correlates with decreased p27 expression, and the A/A genotype of this polymorphism were both associated with HAPH in a Kyrgyz population (Aldashev, 2007).

SLC6A4, solute carrier family 6 (neurotransmitter transporter, serotonin), member 4

The serotonin transporter (5-HTT) is an integral membrane protein that terminates the action of serotonin (5-HT) by transporting the neurotransmitter from the synapse into presynaptic neurons. Mice deficient in 5-HTT were less likely to develop pulmonary hypertension or show signs of vascular remodeling than control mice under hypoxic conditions (Eddahibi et al., 2000). The serotonin-transporter-linked promoter region is a commonly studied insertion–deletion polymorphism with a short (S) and long (L) allele. The L allele is expressed at a higher rate; accordingly, the L/L genotype is associated with increased 5-HTT expression and increased pulmonary artery smooth-muscle cell proliferation (Fumeron et al., 2002). No association was found between variants of the S/L polymorphism and HAPH, but the author notes that the LL genotype was very rare in the studied Kyrgyz population (Aldashev, 2007).

More RAAS genes: ACE2, angiotensin I converting enzyme 2

The ACE2 gene encodes angiotensin I converting enzyme 2 (ACE2), and it belongs to the RAAS pathway. ACE2 catalyzes two reactions: angiotensin I to angiotensin 1–9 and angiotensin II to angiotensin 1–7. Although angiotensin 1–9 does not have a known direct effect on blood vessels, the conversion of angiotensin I to angiotensin 1–9 would prevent angiotensin I-mediated vasoconstriction. In addition, ACE and other peptidases can convert angiotensin 1–9 to angiotensin 1–7, which is a vasodilator that antagonizes angiotensin II type 1 receptor stimulation-mediated vasoconstriction and lowers blood pressure (reviewed in Iwai and Horiuchi, 2009). Polymorphisms of the ACE2 gene have been tested in only one altitude-illness association study, in which variants of two ACE2 gene polymorphisms were not associated with HAPE susceptibility in a Chinese population (Qi et al., 2008).

Genes in the hypoxia response pathway: HIF1A, hypoxia inducible factor, and VHL, von Hippel–Lindau tumor suppressor protein

Hypoxia inducible factor-1α (HIF-1α) is a transcriptional factor that regulates the transcription of many genes involved in maintaining O₂ homeostasis. HIF-1 is a heterodimeric protein composed of HIF-1α and HIF-1β subunits (reviewed in Semenza, 2007). Both are constitutively expressed, but HIF-1α concentration varies with O₂ concentration. HIF-1α is degraded under normoxic conditions and upregulated under hypoxic conditions, during which it serves to upregulate the expression of genes that mediate the body's response to hypoxia (Smith et al., 2008). Genes involved in physiologic processes such as erythropoiesis, angiogenesis, and glycolysis have HIF-1 binding sites located within hypoxia response elements (HREs) and are regulated by HIF-1. One polymorphism of this gene was studied in Sherpas with and without a history of AMS (Droma et al., 2008), but no statistically significant difference in the distribution of variants between cohorts was detected.

Related to the HIF-1 gene, the von Hippel–Lindau tumor suppressor gene (VHL) encodes the VHL protein, which is part of a larger complex of proteins responsible for the ubiquination and degradation of HIF-1α under normoxic conditions (Smith et al., 2008). Based on an analysis of five SNPs in a cohort of Sherpas, Droma and colleagues (2008) concluded that there was no association between VHL and a history of AMS susceptibility.

ABO blood group gene

There are three main alleles at the ABO locus, and these determine a person's blood type (A, B, AB, or O). The A and B alleles encode glycosyltransferases that modify the H antigen (ABO precursor) on red blood cell surfaces to form the A or B antigen, respectively. The O allele has a deletion that causes a frameshift, resulting in the formation of a nonfunctional protein that does not modify the H antigen. The A and B alleles are codominant, and the O allele is recessive. The frequency of blood types varies ethnically, but the function of the A and B antigens is unknown. These antigens are important considerations for blood transfusions and hemolytic disease, but their association with most other diseases is controversial. Recently, a GWAS showed that alleles at the ABO locus were significant contributors to the mean ACE activity in a group of hypertensive Han Chinese patients, with two SNPs responsible for a combined 7.9% of the total variance (Chung et al., 2010). How these alleles affect ACE activity is unknown, but in doing so they may affect blood pressure. The potential significance of ACE and blood pressure in the pathophysiology of altitude illness implies that the ABO blood group gene is a potential candidate gene to be considered independently of or in combination with genes in the RAAS pathway. To our knowledge, only one study has examined blood type and altitude illness, and blood type was not a significant determinant of AMS development in a population of trekkers who ascended 3952 m on Yushan (Jade Mountain) in Taiwan (Wang et al., 2010c). This was a large epidemiological study based on self-reported blood types that did not control for factors that could confound a genetic association study of the ABO locus (e.g., subject ethnicity).

Apolipoprotein B (APOB) gene

The APOB gene encodes apolipoprotein B, the primary apolipoprotein in chylomicrons and low-density lipoproteins (LDL), which is essential for binding LDL particles to LDL receptors and consequently their uptake into hepatic cells. Plasma apolipoprotein B concentration is influenced by genetics (Lamon-Fava et al., 1991). Postulating that the effects of plasma LDL on blood pressure and pulmonary vascular tone could play a role in altitude illness, Buroker and colleagues (2010) typed this gene in Han Chinese with and without AMS and Tibetans with and without CMS; however, no associations were reported for either cohort.

Guanine nucleotide binding protein (G protein), beta polypeptide 3 (GNB3) gene

The GNB3 gene encodes the G protein β3 subunit, a component of the heterotrimeric G protein that integrates signals between receptor and effector proteins in transmembrane signaling systems. A polymorphism of the GNB3 gene (C825T) was previously associated with hypertension (Siffert, 2000). The GNB3 -350A allele, a variant of a polymorphism that is in strong linkage disequilibrium with the C825T polymorphism, was present only in individuals with AMS among a group of Han Chinese who ascended to approximately 3670 m; however, because blood pressure was not significantly different between the case and control groups, the potential functional significance of this association remains unclear. The −350A allele was not observed in Tibetan CMS cases or controls (Buroker et al., 2010).

Summary of polymorphisms, association studies, and candidate genes

Candidate-gene association studies have been the most common method for investigating the role of genetics in altitude illness. A total of 58 genes has been tested for an association with four of the major altitude illnesses (AMS, HAPE, CMS, HAPH). Seventeen of these genes had at least one variant associated with one or more altitude illnesses. The effect sizes (odds ratio; see Table 2) for these variants ranged from small (∼1) to large (∼9), but the majority of variants had relatively minor effects, suggesting that the altitude illnesses are likely polygenic conditions that result from the combination of multiple deleterious alleles and/or interactions of multiple genes. The genes with positive associations belong to a variety of biological pathways, providing further evidence for the involvement of multiple systems and the potential for the interaction of genes.

Caution is advised when evaluating these studies— many have limitations. The majority of reported studies used relatively small sample sizes, which limited their power to detect small effect sizes; however, small sample sizes can still rule out large effect sizes, which, from a clinical perspective, are more useful. Large-scale sampling is difficult for altitude illness because a large homogeneous population traveling to altitude is quite rare, resulting in reliance on opportunistic sampling and making a GWAS difficult. The description of subjects' biogeographical origins (i.e., ethnicity) was not clearly defined in some studies, and more specific descriptions of the population would be useful for comparing studies and ensuring that a homogeneous population was assessed. AIMs can be used to assess the ethnicity of subjects and controls and to test for population stratification, but this technique has not yet been implemented for studies of altitude illness. Many studies did not correct for multiple hypotheses testing, which could potentially lead to false positives. Ensuring that strict sampling criteria are enforced and that phenotyping (i.e., diagnosis) is accurate will also help strengthen studies.

Epigenetics

Unlike classical genetics, which focuses on the sequence of the DNA molecule and on the effects that changes in this sequence (e.g., mutations and polymorphisms) have on organisms, epigenetics is the study of modifications to the nucleotide bases of the nucleic acid–protein complex itself that, while not altering gene sequences, still affect the expression of genes and, consequently, the phenotype of the organism. Epigenetics is a rapidly expanding field, and a comprehensive review is beyond the scope of this article; however, recent applied reviews (among many) include Aguilera and colleagues (2010), environmental epigenetics; Brown and Rupert (2010), hypoxia epigenetics; Hirst and Marra (2009), clinical epigenetics; and Sharp (2008), sports epigenetics. In general, mammalian epigenetics involves modification of the proteins that associate with DNA to form chromatin or of the DNA base cytosine. Protein modification can include acetylation, methylation, or phosphorylation of the various histone proteins, whereas DNA modification is limited to methylation of cytosine bases (when followed by a guanine base, i.e., a CpG dinucleotide). By altering the compaction of the DNA, masking enzyme recognition sites, and/or inhibiting (or facilitating) the interaction of enzymes with the DNA molecule, these epigenetic "marks" regulate the extent to which genes are active. Of relevance to this review, the conserved HRE sequence contains a CpG dinucleotide (Wenger et al., 2005), so it is a potential target for DNA methylation. Because the marks are inherited by progeny cells, epigenetic programming is central to development and differentiation. All cells in an organism, regardless of structure and function, have an identical genetic complement, ^* and epigenetic marks contribute to the diversity necessary for organismal complexity, both anatomically and physiologically. There is also evidence for transgenerational transmission of epigenetic marks, which may be adaptive. Children of mothers exposed to famine during pregnancy have been shown to have increased insulin resistance, thereby conserving energy and restricting growth (Heijmans et al., 2008). Since maternal famine can alter DNA methylation, this may be part of a system to program offspring to deal with limited resources. The phenomenon can span multiple generations, with effects seen in the children of mothers exposed in utero.

There have been no investigations of epigenetic responses to environmental hypoxia in humans, but in mice exposed to 10.5% O₂ for 48 h, there were changes in placental gene expression, including genes involved in DNA methylation, and the authors speculate that embryonic hypoxia may cause long-term epigenetic changes (Gheorghe et al., 2007). Placental responses to altitude (Moore et al., 2004) raise the intriguing possibility that epigenetics plays a role in reproductive adaptation to hypoxia. Perhaps epigenetics could explain the historically documented delay in achieving reproductive success experienced by European women colonizing the Andes (cited in Rockwell et al., 2003). One challenge to epigenetic studies is that, unlike genetics, epigenetic information can be tissue specific and, in humans at least, this limits studies of epigenetic changes to a small number of accessible cell types (e.g., buccal cells, blood cells, and hair follicles).

Gene–gene interactions and pathway mapping

Altitude illnesses are complex traits for which many variants of many genes may contribute to susceptibility and resistance. The majority of studies to date have used a single-locus approach, with each variant studied in isolation. This approach is not congruent with the contemporary understanding of genetics, because gene products are usually only one factor in a larger physiological pathway. For example, the associations of the ACE D allele with myocardial infarct (Tiret et al., 1994; Fatini et al., 2000) and the risk for ischemic events (van Geel et al., 2001) were dependent on the AGTR1 C1166A C allele. Similarly, the ACE D/D genotype in combination with the AGT T704C C/C genotype predicted left ventricular mass in male cardiovascular disease patients, but individual genotypes of these alleles did not (Kim et al., 2000). If two genes interact, these interactions are known as gene– gene interactions or epistasis. Genetic interactions are complex, and the combination of different alleles may result in the augmentation or diminishment of a quantitative trait. Moore and Williams (2009) state that canalization, the development of redundant gene networks, may explain the robust nature of biological systems, for which many alterations that would not have an effect individually can collectively disturb the system. Hence, simultaneous investigations of variants of multiple genes may be more informative than independent investigations.

The requirement for large sample sizes has restricted studies of gene–gene interactions. Recently, new methods have been devised or applied to these interactions (reviewed in Cordell, 2009). One popular method, which can handle sparsely populated data cells, is multifactor dimensionality reduction (MDR) (Ritchie et al., 2001; Cordell, 2009). MDR is a nonparametric, genetic model-free data-mining approach that uses constructive induction to identify multilocus and environmental factors that can be used in predictive models of complex disease (reviewed in Cordell, 2009; Moore and Williams, 2009). The MDR approach was designed to complement single-locus approaches, because MDR can detect interactions that may not have significant independent associations.

To date, three studies have used the MDR approach to investigate the possibility of gene–gene interactions in HAPE. Using the tuned relief filter (TuRF), Qi and colleagues (2008) identified the five best polymorphisms in which to investigate gene–gene interactions: AGT M235T, ACE A-240T, A2350G, BDKRB2 C-58T, and CYP11B2 C-344T. The third-order model involving A-240T, A2350G, and C-344T was best; the interaction between A-240T and A2350G was strongly synergistic, whereas the interaction between these two SNPs and C-344T was moderately synergistic. Interestingly, A2350G was not associated with HAPE when examined independently, making it an example of epistasis. The authors identified the genotype A/A, G/G, T/T as carrying the highest risk for developing HAPE, which was corroborated by haplotype analysis. These interactions are biologically plausible, because both ACE and CYP11B2 are involved in blood-pressure regulation and salt–water balance (Qi et al., 2008). The same group performed a similar analysis of the same population (with eight additional cases and a much larger control population) using 5 SNPs belonging to three heat shock protein genes (discussed previously; Qi et al., 2009). There was a synergistic interaction between A-110C, A1267G, and G+190C, and these polymorphisms had an additive relationship with G2074C and T2437C. All possible models of 1–5 polymorphisms were analyzed. Using the three synergistically interacting polymorphisms as the best model, Qi and colleagues (2009) investigated the risk of developing HAPE for the seven haplotypes and nine diplotypes derived from these three polymorphisms, and associations with HAPE susceptibility and resistance were found. Finally, from eight ADRB2 SNPs, a model derived through MDR that included three SNPs (A46G, C79G, and C523A) was the best model of HAPE risk (Stobdan et al., 2010). The A/G, C/C, C/A genotype carried the highest risk (38 cases, 27 controls). These studies have not been replicated, and no other studies have employed this technique for altitude illness.