Excluding individuals with a history of smoking, 90 young male volunteers (age, 23 ± 3.7 years) acutely ascended to Lhasa (3648 m), China, by train within 48 h, were included in this study. All volunteers had the same diet during the investigation. Among them, 22 individuals were randomly selected for microRNA array screening, while the remaining 68 constituted the validation set (Table 1). All blood samples were collected in EDTA tubes using standard operating procedures. The study protocol was approved by the ethics committees of third Military Medical University, China, in accordance with the Declaration of Helsinki; all participants provided written informed consent.

Blood samples from volunteers were collected in the morning at sea level, and 2 days after exposure to high altitude. Self-assessment questionnaires of the Lake Louise Scoring System and main parameters, including blood pressure, heart rate, and blood oxygen saturation, were recorded prior to departure, and at each day during high altitude exposure (7 days).

Thirty-three phenotypes were recorded (Supplementary Table 1), including stress hormones, lipid profiles, hematological indices, myocardial enzyme spectrum, and liver and kidney function related enzymes. High performance liquid chromatography was carried out to detect blood catecholamine amounts, and chemiluminescent immunometeic assay was used to measure adrenocorticotropic hormone and cortisol levels. Serum concentrations of C-reactive protein and biochemical indexes were assessed by immunonephelometry and on a Roche Chemistry Analyzer, respectively. Hematological indices, including red blood cell, hematocrit, hemoglobin, white blood cell and platelet levels, were evaluated on an XE-2100 analyzer (Sysmex, Kobe, Japan) with commercial reagents, immediately after blood collection.

Acclimatization of volunteers at high altitude was evaluated by using the Lake Louise Scoring System (LLS), which comprises a questionnaire and a scorecard that determine severity. During the 7 day study period, individuals with LLS ≥ 3 (including a headache score ≥ 1), at least once, were considered to be un-acclimatized, while those without headache or showing LLS <3 were considered to have a better acclimatization at high altitude.

Total RNA was isolated using TRIzol (Invitrogen) and purified with RNeasy mini kit (QIAGEN) according to manufacturers' instructions. RNA quality and quantity were measured on a NanoDrop spectrophotometer (ND-1000, Nanodrop Technologies). Detailed RNA quality and quantity data are listed in Supplementary Table 2. After quality control, the miRCURY™ Hy3™/Hy5™ Power labeling kit (Exiqon, Vedbaek, Denmark) was used according to the manufacturer's guideline for microRNA labeling. Then, Hy3™-labeled samples were hybridized on a miRCURYTM LNA Array (v.18.0) (Exiqon), according to the manufacturer's protocols. Following hybridization, the slides were washed several times with a Wash buffer kit (Exiqon), and scanned on an Axon GenePix 4000B microarray scanner (Axon Instruments, Foster City, CA). Scanned images were then imported into the GenePix Pro 6.0 software (Axon) for grid alignment and data extraction. Replicated microRNAs were averaged, and probes with intensities ≥30 in all samples were selected for normalization. After normalization, significantly differentially expressed microRNAs between the two groups were identified using Fold change and P-value cutoffs of 2 and 0.05, respectively.

For qRT-PCR, a synthetic Caenorhabditis elegans microRNA, cel-miR-39 (Qiagen, Valencia, CA, USA), was added to plasma samples as a control prior to RNA extraction. Total RNA was extracted from 200 μL individual plasma samples with a miRNeasy extraction kit (Qiagen, Valencia, CA, USA) according to the manufacturer's instructions. RNA (6 μl) was reverse transcribed into cDNA (in a final volume of 10 μl) using a reverse transcription kit (GenePharma, Shanghai, China). Subsequently, quantitative real-time PCR was performed on an iQ™5 Real-Time PCR Detection System (Bio-Rid, USA) using SYBR Green. The primers used for qRT-PCR are listed in Table 2. Relative microRNA levels were determined by the ΔΔCT method.

After microRNA profiling, we searched three well-known target prediction databases, including TargetScan (http://www.targetscan.org) (Lewis et al., 2003), miRanda (http://www.microrna.org) (Enright et al., 2003) and miRBase (http://www.mirbase.org/) (Griffiths-Jones, 2004), to identify mRNA targets of the altered microRNAs. To reduce the chance of false positives, only the targets obtained in all three programs were considered putative candidates for Gene Ontology (GO) (Ashburner et al., 2000) and KEGG pathway analysis (Kanehisa et al., 2008).

Normality was assessed for all datasets by the Shapiro-Wilk's test. Then, statistically significant differences in phenotypes were identified by paired-samples t-test or Wilcoxon signed-rank test. The Pearson coefficient correlation method was used for correlation analyses between phenotypes and microRNA expression. Independent t-test or Mann-Whitney U test was carried out to assess phenotypic differences between acclimatized and un-acclimatized individuals. Areas under the receiver-operating characteristic curves (AUCs) were evaluated using Swets classification (Gasparini et al., 2015): AUC = 0.5, no diagnostic value; 0.5 < AUC <0.7, accuracy to only a small degree; 0.7 < AUC <0.9, fair accuracy; 0.9 < AUC ≤ 1, high accuracy. P < 0.05 was considered statistically significant. All statistical analyses were performed with the R software (version 3.2.3, R Foundation for Statistical Computing, Vienna, Austria) using the “gmodels” package for principle component analysis (PCA). The suitability of PCA was assessed prior to analysis using the Kaiser-Meyer-Olkin and Bartlett's test.

Exposure of healthy lowlanders to high altitude was accompanied by progressively increased heart rate and blood pressure during the first week at high altitude (Supplementary Figure 1), which are expected cardiovascular adjustments to the unavoidable reduction in oxygen availability. Oxygen saturation of arterial hemoglobin decreased with the ascent due to a reduction in inspired oxygen partial pressure (Supplementary Figure 1). To identify the most dramatic phenotypic changes at high altitude, 33 parameters were examined, and 19 of them were significantly altered at high altitude compared with pre-exposure values (Figure 1; Supplementary Figure 2; Supplementary Table 3). Among these, hematological indices, including red blood cell, hematocrit and hemoglobin levels, were significantly increased after exposure to high altitude as expected. In addition, cortisol (relevant to stress) and triglyceride, total cholesterol, lactate dehydrogenase, high-density lipoprotein-cholesterol and low-density lipoprotein-cholesterol (involved in metabolism significantly) levels were significantly altered by high altitude hypoxia.

A microRNA expression array was utilized to examine circulating microRNA expression changes at high altitude. Of the 2383 probes assessed, 86 microRNAs (79 up-regulated and 7 down-regulated) showed differential expression with statistical significance, and were selected for subsequent analysis (Figure 2A).

Meanwhile, computational predictions of the target genes were performed for the top 10 upregulated and all downregulated miRNAs, and relevant Gene Ontology categories and signaling pathways for all the target genes were identified. A large number of targets were significantly enriched in several key signaling pathways implicated in macromolecule metabolism (Figure 2B).

To further identify the cmiRNAs potentially responsible for the observed phenotypic changes, the associations of 19 significantly altered phenotypes with 86 robustly differentially expressed cmiRNAs were assessed. Of the 86 differentially expressed microRNAs, 67 were associated with the 19 significantly changed phenotypes at high altitude (Figure 3; Supplementary Table 4). Interestingly, the vast majority of microRNAs were substantially related to metabolic parameters and stress hormones. Importantly, 32 and 25 microRNAs were strongly associated with total cholesterol (TCH) and low-density lipoprotein-cholesterol (LDLC) elevations, respectively; meanwhile, 22 microRNAs were strongly correlated with increased cortisol (F).

Based on the LLS of all individuals in the microRNAs array screening set, 13 participants were considered un-acclimatized individuals with acute mountain sickness, whilst the remaining 9 had better acclimatization at high altitude. The phenotypic changes in acclimatization individuals presented no significant differences compared with un-acclimatized subjects (Supplementary Table 5). To ensure whether there are differences in circulating microRNA expression profiles between acclimatized and un-acclimatized individuals, we analyzed circulating microRNA expression profiles in both groups. Interestingly, 91 (55 up-regulated and 36 down-regulated) cmiRNAs were differentially expressed under the set criteria (p < 0.05 and fold change ≥2) (Figure 4A). Among the 91 differentially expressed cmiRNAs, there was an overlap of 11 cmiRNAs with the 86 cmiRNAs differentially expressed after exposure to high altitude. The suitability of PCA was assessed prior to analysis. Inspection of the correlation matrix showed that all variables had at least one correlation coefficient greater than 0.3. The overall Kaiser-Meyer-Olkin (KMO) score was 0.78, with individual KMO measures all above 0.7, classified as “middling” to “meritorious” according to Kaiser (1974). The Bartlett's test of sphericity was statistically significant (p < 0.0001), indicating that the data are likely factorizable. PCA further confirmed these results, showing that un-acclimatized individuals form a separate group from acclimatized individuals (Figure 5A).

To further explore the sensitivity and specificity of cmiRNAs in discriminating between acclimatized and un-acclimatized individuals, we measured the expression levels of has-miR-676-3p (MIMAT0018204), has-miR-3591-3p (MIMAT0019877), has-miR-181b-5p (MIMAT0000257) and has-miR-193b-5p (MIMAT0004767), the most differentially up-regulated cmiRNAs, by qRT-PCR. As shown in Figure 4B, qRT-PCR data were consistent with microarray findings. Indeed, has-miR-676-3p, has-miR-3591-3p, has-miR-181b-5p, and has-miR-193b-5p were differentially expressed between acclimatized and un-acclimatized individuals. Moreover, among the four cmiRNAs, has-miR-3591-3p, has-miR-676-3p, and has-miR-181b-5p yielded ROC AUCs of 0.805 (P < 0.0001), 0.713 (P < 0.01), and 0.734 (P < 0.01), respectively, indicating that these cmiRNAs were fairly accurate in determining un-acclimatization diagnosis (Figure 5B).

To further assess the pathogenesis of un-acclimatization at high altitude hypoxia, GO and KEGG pathway analyses were performed to disclose functional enrichment of genes predicted to be regulated by the top 10 differentially up- and down-regulated cmiRNAs between acclimatized and un-acclimatized individuals. Interestingly, several signaling pathways were regulated by these cmiRNAs. Specifically, the HIF-1 signaling pathway was suppressed by cmiRNAs up-regulated in un-acclimatized individuals (Table 3).

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.