Genome Wide Expression Analysis Suggests Perturbation of Vascular Homeostasis during High Altitude Pulmonary Edema

Background High altitude pulmonary edema (HAPE) is a life-threatening form of non-cardiogenic edema which occurs in unacclimatized but otherwise normal individuals within two to four days after rapid ascent to altitude beyond 3000 m. The precise pathoetiology and inciting mechanisms regulating HAPE remain unclear. Methodology/Principle findings We performed global gene expression profiling in individuals with established HAPE compared to acclimatized individuals. Our data suggests concurrent modulation of multiple pathways which regulate vascular homeostasis and consequently lung fluid dynamics. These pathways included those which regulate vasoconstriction through smooth muscle contraction, cellular actin cytoskeleton rearrangements and endothelial permeability/dysfunction. Some notable genes within these pathways included MYLK; rho family members ARGEF11, ARHGAP24; cell adhesion molecules such as CLDN6, CLDN23, PXN and VCAM1 besides other signaling intermediates. Further, several important regulators of systemic/pulmonary hypertension including ADRA1D, ECE1, and EDNRA were upregulated in HAPE. We also observed significant upregulation of genes involved in paracrine signaling through chemokines and lymphocyte activation pathways during HAPE represented by transcripts of TNF, JAK2, MAP2K2, MAP2K7, MAPK10, PLCB1, ARAF, SOS1, PAK3 and RELA amongst others. Perturbation of such pathways can potentially skew vascular homeostatic equilibrium towards altered vascular permeability. Additionally, differential regulation of hypoxia-sensing, hypoxia-response and OXPHOS pathway genes in individuals with HAPE were also observed. Conclusions/Significance Our data reveals specific components of the complex molecular circuitry underlying HAPE. We show concurrent perturbation of multiple pathways regulating vascular homeostasis and suggest multi-genic nature of regulation of HAPE.


Introduction
High altitude pulmonary edema (HAPE) is a life threatening clinical condition that occurs in otherwise healthy individuals who rapidly ascend to high altitude (above 3000 m). It is the major cause of death related to high altitude exposure. Though notable physiological and clinical studies in past two decades have underscored the importance of 'acclimatization schedules' in preventing HAPE, the individual susceptibility beyond acclimatization remains far from being deciphered. A prior history of HAPE [1], rapid ascent to high altitude [2], strenuous exercise at high altitude [3], preexisting respiratory infection [4] and genetic factors [5] are some of the proposed reasons for HAPE. Previous studies have shown exaggerated, unsustainable and non-uniform pulmonary vasoconstriction, leading to mechanically induced breaks at blood-gas barrier, to be concomittant with HAPE [6]. Elevated capillary pressure possibly due to inhomogeneous hypoxic pulmonary vasoconstriction (HPV) [7], [8], [9], transarteriolar leakage [10] and/or hypoxic venoconstriction [11] in pulmonary circulation are known to be causally linked to HAPE. Augmented hypoxic pulmonary vasoconstrictor (HPV) response to hypoxia [12], [13] and increased pulmonary artery systolic pressure (PASP) during exercise in normoxia [14] have also been reported in susceptible individuals. Reduced pulmonary synthesis of endothelium-derived vasodilator nitric oxide (NO) and impairment of NO mediated pulmonary vasodilation [15], [16] are some of the other factors of significance for HAPE susceptibility.
Amongst the aforementioned factors, it is difficult to identify primary and secondary factors which regulate individual HAPE susceptibility. Furthermore, the involvement of inflammation in primary etiology of HAPE remains unclear, suggesting complexity of possible underlying mechanisms. It is likely that increased hydrostatic pressure, inflammatory response or a combination of both could be important [6]. Notably, individuals who develop HAPE run a significant risk of recurrence suggesting involvement of genetic component in its etiology although little is known about the genetic basis of HAPE. An understanding of basic molecular mechanism underlying HAPE is thus necessary not only for designing of appropriate preventive or prophylactic strategies but more importantly for evaluating susceptibility. In the present study, we utilized a genome-wide approach to identify differentially expressed genes and pathways from peripheral blood of individuals who developed HAPE after acute induction to high altitude. Peripheral blood cells have been suggested to serve as a representative sample of the systemic state allowing for evaluation and profiling of multiple pathological and physiological pathways [17]. We implicate multiple genes within different pathways composing a complex network regulating vascular homeostasis and possibly, in consequence, lung fluid dynamics during HAPE.

Study Design
The study population consisted of male sea level residents of Indian origin (n = 17) who had developed HAPE after air induction (ascent by flying) to high altitude (Leh situated at 3250 m). Four volunteers were fresh inductees while the rest were re-inductees who travelled to high altitude after spending their leave period at sea level (Supplementary Table S1). After arrival at the Leh altitude, they were undergoing acclimatization as per prescribed acclimatization schedule but developed HAPE within 24-96 hrs. One individual developed HAPE after 120 hours. They were immediately admitted to Leh hospital and provided medical aid and oxygen supplementation. Diagnosis of HAPE was done by the specialist medical officers at the hospital who attended to the patients based on presence of pulmonary rales, cyanosis and chest radiography. The clinical symptoms included cough, dyspnoea at rest, breathlessness, chest tightening and congestion. The control group consisted of age, sex and duration matched lowlanders (n = 14) who were also inducted under similar conditions, underwent similar acclimatization schedule and did not develop HAPE (these volunteers were considered acclimatized). All the volunteers were healthy at sea level as per their annual medical record and did not suffer from any congenital anomaly or cardiovascular disease. The study was approved by 'Institute Ethics Committee of Defence Institute of Physiology and Allied Sciences (DIPAS), Delhi' and informed written consent was obtained from all volunteers. Clinical parameters of heart rate, systolic and diastolic blood pressure and arterial oxygen saturation were measured in HAPE individuals before oxygen supplementation and from acclimatized individuals before drawing blood.

RNA isolation and Aminoallyl labeling
Blood samples from HAPE individuals were collected immediately on admission prior to medication and oxygen supplementation except for three individuals from whom samples were collected 8 hrs after admission to hospital and stabilization of patients' condition (Supplementary Table S1). Blood samples were collected from acclimatized individuals within 45-72 hrs of arrival to high altitude except one (Supplementary Table S1). About 7 ml of venous whole blood was collected in PAXgene Blood RNA Tubes (PAXgene, PreAnalytiX, Hombrehtikon, Switzerland, distributed by Qiagen, catalog no. 762125 ex US) at room temperature using a blood collection accessory (Beckton Dickinson) and stored at 4uC. Within two days of collection, samples were transported on ice to the laboratory in Delhi situated at sea level. Isolation of total cellular RNA was done utilizing PAXgene Blood RNA kit (PreAnalytiX) as per manufacturer's protocol. An on-column DNase digestion was carried out for the removal of DNA using RNase-free DNase set (Qiagen) and purified RNA stored at -80uC. RNA concentration and quality was evaluated by measuring absorbance at 260 and 280 nm using a NanoDrop (NanoDrop, USA) and electrophoretic analysis on Agilent 2100 Bioanalyzer (Agilent Technologies Inc., Palo Alto, CA ) prior to labeling. Probes were generated by aminoallyl labeling using Amino AllylMessageAmp TM II aRNA Amplification Kit (Ambion, Life Technologies Corporation) and Cy3 TM post labeling reactive dye pack (GE Healthcare, UK).

Microarray hybridization and image analysis
Hybridization to Human Whole Genome oligochip 40K (Ocimum, Hyderabad, India) was carried out as per manufacturer's protocol (Ocimum, Hyderabad, India) and data submitted to GEO (Accession number: GSE52209). Experimental design, sampling, hybridization and data analysis were performed in compliance with Minimum Information About a Microarray Experiment (MIAME) guidelines (Supplementary Table S2). Each array representing a sample was scanned for three different PMT settings (40,50 and 60) on Affymetrix 428 TM Scanner. The images were loaded into the image analysis software (Imagene) and the transcript levels quantified. Initially, the data consisted of 40,320 probes, including empty spots and control probes which were removed leaving behind 39,400 probes. Out of these, 1620 probes corresponding to genes/ESTs that had at least one probe differentially expressed were chosen for further downstream filtering ( Figure 1). Probes were further filtered based on majority expression, +/strand and proximity to 3' end. The missing values were imputed by K-nearest neighbor (KNN) method. The intensity data for each sample was transformed on log 2 scale. Quantile normalization was used to overcome the systematic variation across the arrays. Three HAPE samples showed slightly different distribution profiles (bimodal) compared to other samples, however, for normalization, all samples were retained and distribution profiles obtained.

Differential expression analysis
Linear modeling approach was used for assessment of differential expression. The LIMMA [18] library from R-Bioconductor package was utilized to construct the linear models and gene wise linear models were set based on the experimental design. Analysis involved estimating consistent, closed form estimators for the hyper parameters using the marginal distributions of selected statistics. 'Robust linear method' (rlm), which is robust to outliers, measurement errors and other data irregularities, was used for estimating the model coefficients. These estimators were based on the two-step weighted least squares method, where weights were adaptively computed using the empirical distribution of residuals obtained from initial robust fit. To obtain differentially expressed genes, moderated t-statistic was used involving posterior residual standard deviations rather than the ordinary standard deviations. The design matrix consisted of two conditions/groups. The difference in the expression level of samples in the two conditions was studied by setting appropriate contrasts. The multiplicity of testing was performed using Benjamin & Hochberg (BH) correction adjusting for the false discovery rate (FDR) since each analysis involved large number of tests. The threshold adjusted pvalue was set to 0.05 and the fold change threshold was set to 1.5. These settings were retained throughout the analysis to select gene list across the comparisons. Expression levels of HAPE samples were compared with those of acclimatized control samples. The up/down regulation of a gene was decided only if the difference in the expression was statistically significant (p#0.05) and the absolute fold-change was greater than 1.5. If either of the criterions were not fulfilled, then the gene expression was treated as insignificant. Afold change of 1.5 was utilized in order to retain maximum biological information present in the data set. The data on log fold change and the p-value was further used to generate volcano plot (Supplementary Figure S1). A two-way hierarchical clustering was performed using the up and down regulated genes for the selected probes. Euclidean distance measure and average linkage method were used to obtain the relatedness of samples and genes and generation of heat maps.

Gene Enrichment analysis and Functional Annotation Clustering
The differentially expressed genes were studied for their overabundance in different GO terms as well as Pathways. The GO terms were categorized into biological process, molecular function and cellular component. The overabundance of a particular term was decided based on the number of significant genes in the analysis, the number of significant genes relevant to the term, the total number of genes for the organism and the number of genes that were relevant to the term for the organism. Fisher's exact test was used to determine the significance of each GO term; the significance of a term (p#0.05) was an indicator that it was enriched with genes. Functional Annotation Clustering was achieved using the Database for Annotation, Visualization and Integrated Discovery (DAVID) [19]. All genes which were identified to be both significantly (p#0.05) and numerically (61.5 fold change) expressed and also assigned with an official gene symbol and accession number, were used in the analysis by DAVID. Intergenic network (p#0.05) was extracted using Pathway Miner [20] for understanding the functional interactions between the differentially expressed genes.
Real-time quantitative validation of microarray data with TaqMan Low Density Array (TLDA) Confirmation of microarray results was performed on randomly selected differentially expressed genes (n = 20, FC.1.5, p#0.05) by Real Time Polymerase Chain Reaction on a microfluidic card assay system (TLDA) of Applied Biosystems (Foster, CA) using the same samples used in microarray experiments. The card included TaqMan probes and primer sets for amplification of target genes and an endogenous control of ribosomal RNA 18S. For each sample, cDNA was prepared with High Capacity Reverse transcription kit (Applied Biosystems). qRT-PCR was performed in duplicate on cDNA samples (200 ng) in each port of TLDA card through 50uC for 2 min, 94.5uC for 10 min, 40 cycles of 97uC for 30 s and 59.7uC for 1 min on ABI PRISM 7900 HT Sequence Detection System (Applied Biosystems, USA). The data was analyzed using comparative cycle threshold method (C T method) 2 2DDCt using the RQ software [21].

Results
The clinical parameters of heart rate (HR), systolic blood pressure (SBP), diastolic blood pressure (DBP) and arterial oxygen saturation (SaO 2 ) from HAPE and acclimatized individuals is presented in Supplementary Table S1 and were in the similar range as reported earlier by us [22]. Global analysis of differentially regulated genes and enriched functional clusters in individuals with HAPE We utilized RNA extracted from HAPE and acclimatized individuals to perform microarray experiments as detailed in previous section. From these experiments, we found significantly altered expression levels of 728 genes/ESTs (by more than/equal to 1.5 fold change) in peripheral blood cells of individuals with HAPE ( Figure 1). Amongst these, 420 genes/ESTs were upregulated and 308 genes/ESTs were downregulated. The complete list of genes along with information for respective biological process, molecular function, cellular localization is shown in Supplementary Table S3.
The hierarchical clustering of this data set distinctly separated the acclimatized lowlanders from individuals with HAPE ( Figure  2). We infer these results to be indicative of unique gene expression signatures of two groups of individuals (HAPE and acclimatized) and likely 'unifying' molecular responses within specific groups in response to high altitude exposure. Out of the 20 genes selected for validation, a subset of 17 genes yielded positive correlation and showed similarity of trends across the microarray and the RT-PCR platforms attesting overall authenticity of microarray data; only 3 transcripts showed a mismatch of expression between the two detection methods ( Figure 3).
Categorization of the differentially expressed genes based on their biological process, molecular functions and cellular component are shown in Supplementary Figure S2. Maximum number of differentially expressed transcripts in the molecular function category were for binding activity, followed by transcripts involved in catalytic activity. The other differentially expressed transcript categories of molecular function were representative of enzyme regulator activity, transcription regulator activity, molecular transducer activity and transporter activity. We further clustered the GO terms utilizing software 'BINGO' and as shown in Figure  4, observed enrichment of seven biological processes. These included amino acid metabolism and catecholamine synthesis, regulation of blood pressure, smooth muscle contraction, chronic inflammation, ion homeostasis, protein targeting and regulation of gene expression. The enrichment of inter-linked biological processes especially, the regulation of catecholamine biosynthesis, blood pressure, smooth muscle contraction and chronic inflam- Further analysis through DAVID revealed a number of biological processes in HAPE individuals which were significantly enriched (p#0.5, Fold Enrichment .1.1). Some of the critical processes were those involved with oxidative phosphorylation (OXPHOS) and ATP synthesis, ion homeostasis, apoptosis, voltage-gated channel activity, iron-sulfur cluster binding, intra cellular protein transport,TNF-associated molecules, regulation of systemic arterial blood pressure, Ras-guanyl exchange factor activity amongst others. Transcripts associated with smooth muscle contraction, lipid catabolic process, calcium ion transport, actin binding and inflammatory response were also differentially expressed in HAPE individuals. The high stringency functional annotaion cluster is shown in Supplementary Table S4.

HAPE-specific Inter Genic Networks
We next utilized Pathway Miner [20] to visualize the interactions between differentially regulated genes. Interacting nodes were extracted at p#0.05 and represented along with color scheme to indicate the respective fold changes ( Figure 5). Specific genes which feature within these pathways along with respective fold changes are represented as Supplementary Table S5. The integrated network contained a hub of common nodes which included TNF alpha, MAP2K2, PLA2G1B, PAK3, SOS1 ( Figure 5). These nodes were connected to smaller sub-networks with genes composing specific pathways such as regulation of actin cytoskeleton, MAPK signaling, VEGF signaling, Adipocytokine signaling, T-cell receptor signaling, gap junctions and focal adhesion, fatty acid metabolism amongst others. These sub-networks (specific pathways) as part of the bigger network are illustrated in Supplementary Figure S3A-R.

Discussion
There has been an ever increasing body of evidence in favour of 'stress failure' being one of the early phenomenon which initiates the cascade of events in HAPE. However, it has also been argued in literature that initial damage to blood-gas barrier because of 'stress failure' is self limiting owing to the propensity for quick closure of this barrier after pressure reduction [23]. This strongly suggests the existence of confounding parallel factors complementing initial stress failure for clinical presentation of HAPE which may arise from conjunction of multiple factors besides capillary stress failure [24]. Our data supports this notion and we suggest concurrent modulation of multiple factors which govern vascular homeostasis (balance between vascular injury and vascular repair) and perhaps perturbed lung fluid dynamics during HAPE. Vascular injury is usually induced by factors such as altered shear stress, oxidative stress, cytokine, and vascular tone modulators such as endothelin-1, angiotensin II amongst others. Notwithstanding the limitation of single time point study which is less likely to suggest cause/consequence relationship between various events, our global expression profiling for HAPE and acclimatized individuals evoked several interesting mechanistic possibilites besides generating molecular evidence for well known physiological processes involved in HAPE [23,25]. We observed multiple pathways to be modulated under these conditions which appear to be reminiscent of three broad physiological processes: 1) regulation of capillary pressure (stress failure), 2) modulation of barrier function (endothelial permeability/dysfunction) and 3) differential regulation of hypoxia sensing/response pathways.   Pathways which could directly regulate capillary pressure included smooth muscle contraction, cellular actin cytoskeleton rearrangements and regulation of systemic/pulmonary hypertension. Modulation of paracrine signaling through chemokines, lymphocyte activation and modulation of junctional proteins could potentially regulate endothelial permeability and dysfunction.
Notably, under conditions of perturbed endothelial permeability, HAPE can occur at much lower capillary pressure [25,26]. Lastly, differential regulation of metabolic pathways (OXPHOS, Fatty acid metablism), VEGF signaling and specific individual genes such as EGLN3 suggest significant differences in hypoxia sensing/ responses between acclimatized and HAPE individuals. Specific gene clusters, pathways and other individually curated genes from our dataset which support our proposition of perturbed vascular homeostasis through above mentioned pathways are discussed below and depicted schematically in Figure 6.

1) Pathways Regulating Capillary Stress Failure
Vasoconstriction and actin cytoskeletal organisation mechanical breakdown of basement membrane and pulmonary capillary endothelium resulting from excessive intravascular pressure [6], [27], [28]. This increase in intravascular pressure is a likely consequence of vasoconstriction at tissue level (blood vessels) regulated further by actin cytoskeletal reorganization at cellular level (vascular smooth muscle cells). Our data set revealed several transcripts which could be classified as potential regulators of vascular smooth muscle contraction (hsa:04270) (Table 1). Furthermore, gene transcripts associated with regulation of actin cytoskeleton namely PAK3, MAP2K2, SOS1, ARAF, NCKAP1L, PXN, (KEGG_pathway hsa:04810) were up regulated in HAPE (Table 1) though PTK2, MYLK3 and FGF7 were down regulated. Several other transcripts associated with actin cytoskeleton components like ACTR2, SVIL, SPTBN1, SLC9A3R1, TNNI1 and MYO5C were also up regulated in individuals with HAPE. Notably, we observed ARHGAP24 which encode negative regulators of Rho GTPases and are implicated in actin remodeling, cell polarity and cell migration [29] to be down regulated in our data set. Hypoxia per se is known to culminate in vasoconstriction. Further, we too have observed modulation of vascular smooth muscle contraction (has:04270) and actin cytoskeleton regulation (has:04810) in healthy volunteers on ascent to high altitude during process of acclimatization [30]. Taken together, HAPE-selective marked over-representation of genes regulating vasoconstriction and actin cytoskeletal rearrangement pathways, as seen in the present data set, is a possible evidence for 'exaggerated vasoconstriction' -an important factors which is likely to perturb pressure gradient across pulmonary capillaries.
Endothelin converting enzyme, endothelin receptor and hypoxic pulmonary vasoconstriction study, we observed several upstream regulators of vasoconstriction in HAPE compared to the acclimatized controls; one of them being upregulated transcript of endothelin converting enzyme (ECE 1). ECE 1 generates endothelin-1 (ET-1) from its inactive intermediate and its expression levels have been reported to increase in mouse carotid body under physiological hypoxia [31]. Increased lung endothelin content contributing to lung water accumulation and promoting development of pulmonary edema has been reported in young rats exposed to moderate environmental hypoxia [32], [33]. An augmented release of endothelin-1 and/or its reduced pulmonary clearance in HAPE-susceptible mountaineers was hypothesized to represent one of the mechanisms contributing to exaggerated pulmonary hypertension at high altitude [34]. Increased level of endothelin 1 [34], [35] and decreased availability of nitric oxide [36], [15] have emerged as strong mediators of HPV in HAPE. Our observation of upregulated expression of ECE1 in HAPE would logically imply higher production of ET-1 [37] consequently culminating in exaggerated HPV in such individuals.
Besides ECE 1, upregulated transcripts of endothelin receptor type A (EDNRA) were also observed in HAPE individuals. Endothelin receptor type A plays an important role in hypertension, vascular remodeling, cardiac hypertrophy and coronary artery disease [38] and is found on pulmonary vascular smooth muscle cells and adventitial fibroblasts. EDNRA antagonist in vivo has been shown to inhibit HPV [39], [40] and also ameliorates edema formation through decreased leakage of fluid from vasculature into the air spaces [32]. In view of functionally upstream regulation of key physiological pathways by EDNRA and ECE 1, we speculate these genes to be key nodes in pathways . Pulmonary capillary 'stress failure' is described as the . In the present regulating elevated pulmonary artery pressure in HAPE. These genes may act as potential genetic markers of high altitude acclimatization/maladaptation and these genomic loci should be scanned for identifying differences in HAPE and acclimatized individuals.

Regulation of systemic blood pressure in HAPE
observed up regulated expression of adrenergic alpha-1D receptor (ADRA1D), which is potentially involved in regulation of systemic blood pressure, to be upregulated in HAPE. Adrenergic receptors mediate sympathetic responses such as smooth muscle contraction causing vasoconstriction in blood vessels including arteries to heart [41], renal artery [42] as well as smooth muscle contraction of bronchioles. ADRA1D is one of the three subtypes of the a1 adrenergic receptor family that act as primary catecholamine receptor in the lungs. Enhanced a1 adrenergic trophic activity has been reported in hypoxic pulmonary hypertensive rats [43]. Expression studies from healthy human have shown all the three a1 receptor subtypes to be present in peripheral blood lymphocytes with expression of a1D being the least [44]. Considering that HAPE is characterized by distinctly increased pulmonary artery pressure preceding edema formation [1], upregulated ADRA1D along with other upregulated gene transcripts of ECE1, CYP11B2, PCSK5 and EDNRA, as observed in our data set, suggests cumulative role of multiple genes in elevating pulmonary arterial . We blood pressure thereby supporting multi-factorial dependence for clinical presentation of HAPE.
Heat shock protein (HSP) 90 and nitric oxide production in HAPE individuals was the downregulation of heat shock protein 90 kDa (HSP90AB3P) suggesting that many important signaling pathways may be compromised in these individuals. HSP90 is involved in HIF signaling pathway as well as many other important cellular events including protection of vital proteins of stress signal transduction pathway. Through altered binding of HSP90, activation state of nitric oxide synthase type 3 (NOS3) and nitric oxide (NO) formation was shown to be crucially regulated by polymerization state of beta actin in human platelets [45]. Studies have shown decreased NO concentration in exhaled air in HAPE susceptible individuals during hypoxic exposure at low altitude [16] and during development of HAPE at high altitude [15]. Further, lower concentration of nitrate and nitrite in bronchoalveolar lavage fluid was observed in mountaineers who developed HAPE compared to the controls [46]. It is plausible that decreased expression of HSP90 in HAPE, as seen in present study, may inhibit tertiary complex formation between NOS3, beta actin and HSP90 thereby resulting in decreased NO formation in HAPE.
Another intriguing observation was down regulation of histone deacetylase 9 (HDAC9) in HAPE. HDAC directly regulates HSP90 function for nuclear receptor activity [47]. Considering that both HSP90 and HDAC9 transcripts were downregulated in HAPE, the mechanism of action of these regulators and their significance in cellular adaptation to high altitude hypoxia should be probed further.

Leukocyte activation, inflammation and adhesion molecules
study also showed modulation of chemokine signalling pathway, T-cell receptor signalling pathway, leukocyte endothelial transmigration, adhesion and GAP junction molecules in circulating blood cells (Table 1). Notably, studies with lavage fluid from HAPE individuals have yielded conflicting evidence for the involvement of inflammation during early HAPE: while some studies have reported presence of pro-inflammatory cytokines in BAL fluid [48][49][50][51] others have not observed the same [46]. Interestingly, the present study showed up regulation of transcripts related to inflammation in the blood milieu of HAPE individuals. We raise a possibility that these gene products may engage in a positive feed forward loop pertaining to endothelial permeability/dysfunction and fluid leak ( Figure 6). Activation of such pathways will be capable of perturbing endothelial barrier function through multiple levels -directly through the altered expression of GAP junctions or indirectly through endothelial activation in response to lymphocyte-endothelial interactions and inflammatory molecules. It is also possible that such pathways can function cooperatively with initial vasoconstriction and stress failure to skew vascular homeostasis towards leakage phenotype observed in HAPE. In this context our observation of up regulated expression of tumor necrosis factor (TNF) in individuals with HAPE is also interesting. Our observation is in line with recent demonstration of increased TNF alpha along with IL-6, VEGF and NO in serum of HAPE patients [48]. Furthermore, studies have also shown that TNF alpha at high altitude increases lung permeability [49], [50], [51]. Our observation of increased expression of TNF could imply plausible role of this transcript in active modulation of lung permeability in HAPE.

Metabolic Pathways
phosphorylation (OXPHOS) (hsa:00190) and fatty acid metabolism (has:00071) were also observed to be differentially expressed in HAPE. At basic cellular level, hypoxia responses are known to culminate in dynamic metabolic adaptation mediated by HIF1a stabilization: up regulation of glycolic pathway genes and modulation of cytochrome oxidase subunit genes [52]. In view of this information, our observation of differential regulation of Early response to hypoxia in our study was differential expression of egl nine homolog 3 (EGLN3) which was upregulated 2.2 fold in HAPE compared to the acclimatized individuals (controls). We also observed near 1.5 fold change in EGLN1 transcript in HAPE although with insignificant p-value (0.1) (data not shown). EGLNs are a member of hypoxia inducible proline hydroxylases (PHD) family which function as cellular oxygen sensors and directly regulate HIF-a subunit stability [53] thereby regulating cellular response to oxygen availability. Under normoxic condition, EGLNs target HIF for proteosomal degradation through hydroxylation while under hypoxic condition, hydroxylation is significantly decreased leading to stabilization of HIF-a subunits [54]. EGLN isoforms show distinct patterns of expression in various cell types: while EGLN2 (PHD1) mRNA levels are unchanged or decreased by hypoxia, levels of EGLN1 (PHD2) and EGLN3 (PHD3) are increased by hypoxia [55]. PHD3 (EGLN3) has been suggested to contribute more substantially to the regulation of HIF-2a than HIF-1a [55]. Increased expression of EGLN3 transcript during HAPE in our study and in every likelihood its increased activity implies potential inhibition of HIF responsive element (HRE)containing genes which are selectively regulated by HIF-2a and thereby modulating basic adaptive response to hypoxia in these individuals. HAPE-specific differential expression of EGLNs observed in our data thus underscores marked differences in hypoxia sensing ability of susceptible individuals and perhaps differential activation of HIF system during high altitude responses. Notably, we did not observe statistically significant difference in the expression of HIF and von Hippel-Lindau tumor suppressor (VHL) transcripts between HAPE and acclimatized individuals.
We understand and note certain limitations of our study. Arguably, the tissue-and temporal-specificity of pathways is not clearly evident from the present data set. Pathogenesis of HAPE is likely to involve modulation of pathways at multiple tissue levels such as pulmonary vasculature and lung epithelium amongst others. An important challenge, therefore, would be to ascribe and validate tissue-specificity of these pathways. Studies with endothelium and/or pulmonary vascular smooth muscles from individuals with HAPE can be helpful in this regard. Furthermore, these pathways are likely to be involved in a temporal and ordered manner in specific cell types. Mechanistic understanding of such complex regulation will certainly benefit from future studies with samples collected pre-and post-induction to high altitude from same individuals.
The present study is one of the first global expression profiling in individuals who, after being air lifted to high altitude, succumbed to altitude stress and developed HAPE despite undertaking acclimatization schedule. Given a shear dearth of information on molecular aspects in this area, the current study takes our understanding a step ahead by revealing specific pathways in the complex molecular circuitry of HAPE. Despite certain limitations, our data presented here yields an unambiguous evidence of the multigenic nature of this condition and concurrent modulation of multiple pathways which may regulate vascular homeostasis and lung fluid dynamics. It is probable that individuals who succumb to HAPE on altitude exposure may have genomic sequence alterations, transcripts of which manifest when such individuals are exposed to the negative stress of high altitude. Approaches such as genome wide association and exome sequencing can therefore further enhance our understanding of this altitude illness  Figure 5. Individual pathways (sub-networks) interacting through common nodes generate the integrated network represented in Figure 5. Majority of these sub-networks directly link to the three broad physiological processes described in the manuscript. (PPTX)