The NASA Twins Study: A multidimensional analysis of a year-long human spaceflight

Abstract

To understand the health impact of long-duration spaceflight, one identical twin astronaut was monitored before, during, and after a 1-year mission onboard the International Space Station; his twin served as a genetically matched ground control. Longitudinal assessments identified spaceflight-specific changes, including decreased body mass, telomere elongation, genome instability, carotid artery distension and increased intima-media thickness, altered ocular structure, transcriptional and metabolic changes, DNA methylation changes in immune and oxidative stress-related pathways, gastrointestinal microbiota alterations, and some cognitive decline postflight. Although average telomere length, global gene expression, and microbiome changes returned to near preflight levels within 6 months after return to Earth, increased numbers of short telomeres were observed and expression of some genes was still disrupted. These multiomic, molecular, physiological, and behavioral datasets provide a valuable roadmap of the putative health risks for future human spaceflight.

Copyright © 2019 The Authors, some rights reserved; exclusive licensee American Association for the Advancement of Science. No claim to original U.S. Government Works.

Figures

Fig. 1.. Study design and human spaceflight metrics.

The flight subject (TW, blue) and his identical twin (HR, green) were each studied over 25 months using a comprehensive set of health and biological metrics. (A) Earth-based and spaceflight collections include blood, sorted cells (CD4, CD8, CD19, and LD), stool, buccal, saliva, urine, and AR blood. TCR, T cell receptor sequencing; T- and U-Metabolomics, targeted and untargeted metabolomics; OSI, oxidative stress and inflammation; T- and U-Proteomics, targeted and untargeted proteomics; CVU, cardiac and vascular ultrasound; VSF, vascular structure and function. (B) Gene expression changes in TW inflight and postflight compared with the preflight period. All time points of HR were used to account for normal levels of variance and noise in gene expression. Genes that were significantly altered inflight after controlling for subject baselines and AR effect are reported. Gene expression changes were reported for any gene with q < 0.05 in a multivariate model that utilized expression values for polyadenylated [poly (A)+] and ribosomal RNA depleted transcripts. (C) Metabolites present at significantly different levels in HR and TW or between pre-, in-, and postflight periods. Heatmap represents median-normalized log2 intensity for each analyte, scaled across all samples. Red color indicates relative enrichment, whereas blue indicates relative depletion. (D) GO analysis of genes ranked on the basis of epigenetic discordance at their promoters. Comparisons of preflight samples to inflight (early and late, and combined early and late) and postflight samples are shown for both CD4 and CD8 cells. Heatmaps represent transformed enrichment values [square root (sqrt) of enrichment] for GO categories with a raw enrichment value >5. Reg., regulation; neg. reg., negative regulation. (E) C-means clustering of multiomics data reveals longitudinal patterns associated with spaceflight. Analyte abundance (scaled) plotted over time for the identified clusters from the integrated metabolome, proteome, cytokine, cognition, and microbiome datasets is shown. Median abundance (bold) per cluster and 5th and 95th percentiles of abundance (shaded) are indicated. The gray shaded region indicates the inflight period. n, number of analytes in cluster. (F) Individual plots of the different analyte types that compose the spaceflight-dependent cluster, cluster 3, from (E). Telomere levels are plotted adjacent to the cluster members for reference. Median abundance (bold) per cluster and 5th and 95th percentiles of abundance (shaded) are indicated. Thin lines show annotated examples of analytes from cluster 3. For Cognition, accuracy [MP(Acc)] and speed [MP(Spd)] on the motor praxis task as well as standardized speed across cognitive domains (speed) are shown. n, number of analytes in cluster; CAG, coabundance gene groups.

Fig. 2.. Telomere length dynamics and DNA damage responses.

(A) Relative average telomere length in PBMCs (DNA) pre-, in-, and postflight assessed by qRT-PCR for HR (green) and TW (blue). Significance was tested by one-way ANOVA, and error bars represent SEM. (B) Relative average telomere length for TW in sorted PBMC subpopulations, CD19 B cells, CD4 and CD8 T cells, and LD fractions, pre-, in-, and postflight. Boxplot whiskers show min and max. (C and D) Telo-FISH–generated histograms of individual telomere length distributions [shorter to longer, lower to higher relative fluorescent intensity (RFI)] for HR (C) and TW (D) preflight (blue), inflight (red), and postflight (green). (E) Cytogenetic analysis of DNA damage utilizing dGH paints (pink) for chromosomes 1, 2, and 3 facilitated simultaneous detection of translocations and inversions. Representative image of dGH on a metaphase chromosome spread illustrating an intrachromosomal inversion (yellow arrow) and an interchromosomal reciprocal translocation (white arrows). (F) Quantification of translocation (striped bars) and inversion (solid bars) frequencies for HR and TW pre-, in-, and postflight. Results were not statistically significant (one-way ANOVA). Error bars represent SEM.

Fig. 3.. Global changes in DNA methylation during spaceflight.

(A) PCA of distances derived from average CpG methylation levels in 1-kb intervals along the chromosomes in each sample (CD4 and CD8 lymphocytes collected from each subject). (B) Genome-wide distributions of MML and NME values in CD4 and CD8 lymphocytes collected from each subject at indicated time points. (C) Genome-wide distributions of JSDs within each subject in comparisons of preflight to the indicated inflight and postflight time points in CD4 and CD8 cells. (D) Heatmaps representing transformed enrichment values (square root of enrichment) for GO categories with a raw enrichment value >5 in TW for comparisons of preflight (L–162) samples to inflight (early, FD76; late, FD259) and postflight (R+104) samples in CD4 and CD8 cells. (E) University of California, Santa Cruz, Genome Browser images of NOTCH3 and SLC1A5 with peaks of JSD at their promoters in TW when comparing inflight (FD259, for NOTCH3, and FD76, for SLC1A5) to preflight (L–162) samples from CD4 (for NOTCH3) and CD8 (for SLC1A5) cells. Differential MML (dMML) and differential NME (dNME) values are also plotted, with negative values indicating reduced MML or NME at the inflight time points compared with preflight time points. For all boxplots, center lines indicate median, boxes indicate interquartile range (IQR), and whiskers indicate 1.5 × IQR. CGI, CpG island; chr. 19, chromosome 19.

Fig. 4.. Lipids and vaccine response.

(A) Cytokines present at significantly different levels in HR and TW or between pre-, in-, and postflight periods. Heatmap represents median-normalized log2 intensity for each analyte, scaled across all samples. Red color indicates relative enrichment, whereas blue indicates relative depletion. (B) Relative levels (median-normalized, scaled log2 intensity) of complex lipids containing ω-3 and ω-6 fatty acids in HR and TW, from untargeted plasma metabolomics. Red color indicates relative enrichment, whereas blue indicates relative depletion. LysoPC, lysophosphatidylcholine; LysoPE, lysophosphatidylethanolamine. (C) Immunological and postvaccination response to spaceflight. For each subject, the proportion of season 2 clones present in the database of influenza vaccination-responsive CDR3 clones derived from year 1 and 3 vaccinations is shown for each cell type and respective TCR chain (FDR < 0.05).

Fig. 5.. Effect of spaceflight on the fecal microbial community structure of the flight subject relative to the twin ground subject.

Samples are color coded by subject (TW, blue; HR, green), with open symbols representing flight samples from TW or flight-equivalent samples from HR. (A) Taxonomic alpha diversity parameters for HR and TW over nine sampling events, normalized to the value of the first sampling event for each subject. Shown are the normalized genus-level log2 SI values for both subjects and the F/B ratio. SI values at the first sampling event were 4.26 and 3.13 for HR and TW, respectively. F/B ratio values at the first sampling event were 2.18 and 0.87 for HR and TW, respectively. (B and C) Analysis of microbial community taxonomic structure at the genus level (B) and functional gene content (C) using mMDS of annotated shotgun metagenome sequence data. Data were log(x + 1) transformed, and a resemblance matrix, from Bray-Curtis dissimilarity, was generated. MDS axes 1 and 2 are plotted; two-dimensional stress values are 0.19 and 0.13 for taxonomic and functional gene analyses, respectively. Sample names represent dates relative to launch. (D) Number of microbial features at each taxonomic level that were differentially abundant between the inflight samples and the combined pre- and postflight samples of TW or between the equivalent sets of samples from HR. (E) Number of gene functional features at each level (ranged from SEED, that is, the most specific gene functions, to level 1, that is, the most general categories) that were differentially abundant between the inflight samples and the combined pre- and postflight samples of TW or between the equivalent sets of samples from HR. (F) Overall similarity of metagenomic sequence data from sample groups was assessed with Bray-Curtis similarity measurements for untransformed data and presented as boxplots. For each subject, all nine samples are compared and plotted (HR and TW; 36 comparisons), followed by pre- and postflight samples only (HR ground and TW ground; 10 comparisons) and flight samples only (HR flight and TW flight; six comparisons). (G) Metabolic products associated with microbial metabolism in humans. The heatmap shows relative levels of microbial metabolites detected in the metabolomics data (median-normalized, scaled, log2 intensity). Red color indicates relative enrichment, whereas blue indicates relative depletion. Row annotations mark different classes of microbial metabolites, including indoles (violet), phenyls (orange), and bile acids (green). Metabolites that were not significantly altered between HR and TW or between pre-, in-, and postflight periods are marked in black.

Fig. 6.. Mitochondrial functions and Seahorse assays.

(A) mtRNA content observed in RNA-seq data for the unsorted frozen PBMC (expressed relative to total number of aligned reads). Sequencing results after library preparation with both poly (A)+ selection and ribosomal depletion. (B to E) Mean OCR during pre-, in-, and postflight for basal mitochondrial respiration (B), spare reserve capacity (C), ATP-linked respiration (D), and nonmitochondrial respiration (E). L6 muscle cells were treated with plasma from TW and HR collected pre-, in-, and postflight with six to eight assay replicates per time point. OCR was measured before and after the addition of inhibitors (oligomycin, a complex V inhibitor; FCCP, a protonophore; and antimycin A and rotenone, complex III and I inhibitors, respectively) to derive parameters of mitochondrial respiration. Mean OCR was calculated for each parameter after normalizing for baseline OCR of the L6 cells before the addition of plasma or inhibitors. Error bars represent 95% confidence intervals of the mean OCR of all replicates at each time point for each flight event. (F) Boxplots comparing the concentration of lactic acid in urine normalized to urine creatinines (determined by GC-MS targeted metabolomics) pre-, in-, and postflight in TW and HR. (G) Boxplots representing the ratio of urine lactic acid to pyruvate pre-, in-, and postflight in TW and HR revealing an inflight shift to anaerobic metabolism in the flight twin.

Fig. 7.. Vascular adaptations.

(A and B) Carotid artery diameter during diastole (A) and carotid intima-media thickness (cIMT) (B) measured while supine on the ground and during spaceflight in TW and HR. “Pre” indicates the mean of the two preflight measurements. (C and D) IL1B (C) and PGF2α (D) as examples of biomarkers of inflammation and oxidative stress that were measured in these two subjects. (E and F) Boxplots of relative COL3A1 (E) and COL1A1 (F) levels pre-, in-, and postflight for both subjects. (G) Ratio of relative plasma levels of proteins APOB and APOA1 (APOB/APOA1) in both subjects pre-, in-, and postflight, measured using untargeted proteomics (LC-MS).

Fig. 8.. Body weight, bone formation, fluid regulation, and biochemical activity during spaceflight. Biochemical and biophysical measures during the mission. Data are mean ± SD.

(A) Body mass expressed as percent change (Δ) from preflight. (B) Relative urine AQP2 concentration in TW (blue circles) plotted along with serum sodium concentrations (yellow circles). (C) Calculated plasma osmolality for both subjects. (D) Relative urine AQP2 concentrations for both subjects. (E) Untargeted (LC-MS) urine proteomics reveals that relative angiotensinogen (AGT) levels inflight correspond with inflight weight loss. In (B) to (E), the shaded area represents inflight time points. (F and G) Bone breakdown (N-telopeptide) (F) and bone formation [bone-specific alkaline phosphatase (AP)] (G) markers. Data in (F) and (G) are expressed as percent change from preflight. (H) Exercise load using the advanced resistive exercise device (ARED) and bone-specific AP marker. 1 lb = 0.45 kg; U/L, units per liter.

Fig. 9.. Fluid shifts and ocular changes.

(A and B) Internal jugular vein cross-sectional area (A) and forehead skin thickness (B) measured by ultrasound as indicators of acute fluid shifting during the transition from seated to supine during ground-based testing and the chronic fluid shift in weightlessness in TW. (C and D) Concurrently, ocular measures of choroid thickness (C) and total retinal thickness (D) measured by optical coherence tomography under the same conditions. (E) Untargeted proteomics reveals decreased urine excretion of LRG1 in TW during flight relative to ground time points. (F) Serum folate is lower in astronauts who experienced ophthalmic issues and was similarly relatively low in both twins. TW’s serum folate increased during flight, mirroring the increase in telomere length.

Fig. 10.. Cognitive performance results.

(A) Representative images of the Cognition battery (10 tasks). MR motor praxis; VOLT, visual object learning test; F2B, fractal 2-back; AM, abstract matching; LOT, line orientation test; ERT, emotion recognition task; MRT, matrix reasoning test; DSST, digit symbol substitution task; BART, balloon analog risk test; RVT, psychomotor vigilance test. (B) Heatmap of cognitive performance scores of TW relative to HR during preflight (N = 3 tests), the first 6 months inflight (inflight 1 to 6, N = 6 tests), the second 6 months inflight (inflight 7 to 12, N = 5 tests), and postflight (N = 3 tests). Test scores were corrected for practice and stimulus set-difficulty effects. All data were then standardized relative to preflight ground data of 15 astronauts (including TW and HR).Test scores in the heatmap reflect TW scores minus HR scores expressed in SD units. The BART reflects risk-taking behavior and was thus not included in the accuracy score across cognitive domains (accuracy, ALL). Efficiency was calculated as the average of the speed score across cognitive domains (speed, ALL) and the accuracy score across cognitive domains (accuracy, ALL). (C) Standardized cognition performance scores for individual test bouts for the AM test (1) and speed (2), accuracy (3), and efficiency (4) across cognitive domains. The vertical lines indicate launch and landing dates for TW. The AM plot shows that HR (green) had a major insight mid-mission relative to the rules that govern the AM that TW (blue) did not have (50% is performance at chance level on the AM). The speed, accuracy, and efficiency plots demonstrate that the nature of the postflight decline in TW’s performance was protracted.