Expression-to-chromatin decoding of the Arabidopsis spaceflight transcriptome

No spaceflight chromatin assay exists for any plant. ChromDecode-At sidesteps that gap. It maps every Arabidopsis gene to its ground-state chromatin context and tests whether the genes spaceflight changes are concentrated in particular chromatin classes. Applied to the Arabidopsis RNA-seq datasets in NASA's Open Science Data Repository (OSDR), it shows that spaceflight-repressed genes sit in Polycomb (H3K27me3) chromatin.

NASA OSDRPCSD 36-stateAraENCODE 12-state PlantCADBJASPAR 2024Python 3.11 MIT code · CC-BY 4.0 data

Manuscript draft (PDF) → Supplement (PDF) Tables S1–S68 Code

Odds 8.0

OSD-37 flight-repressed genes in the AraENCODE Repression state (FDR 3×10⁻¹⁰⁵); PCSD S15 odds 18.0.

2 ecotypes

The bias replicates in ISS-grown roots: Col-0 odds 3.06 (OSD-193) and 3.12 (OSD-218), and WS 3.74.

0 / 265

SOG1 ChIP-seq peaks at Polycomb-down promoters after DNA damage (OSD-496): not a SOG1 damage response.

~6–8

Replicates per group of a single genotype needed for a stable root-level test (9–12 if genotypes are mixed).

Approach

The decoder works from coordinates, not from a learned model. Each gene's TSS-anchored region (gene body + 2 kb upstream) is assigned a chromatin state from two independent annotations, and differentially expressed genes are tested for state enrichment by Fisher's exact test with BH-FDR. A learned expression→state model was tried and failed honestly (block-CV accuracy 0.533 against a 0.549 baseline), which is why the deterministic backbone carries the analysis.

Results

1 · Flight-repressed genes concentrate in Polycomb chromatin

replicatedIn OSD-37 seedlings (four ecotypes; 529 up / 526 down), down-regulated genes are enriched for Polycomb_repressed chromatin (5-group odds 7.01, FDR 1.7×10⁻⁹²) and for the AraENCODE Repression state (odds 8.0). At 36-state resolution they converge on the H3K27me3 + accessible-promoter block: S15 18.0, S16 6.4, S14 3.9, S13 3.3. Of the 526 down genes, 265 sit in Polycomb states S11–S15. Constitutive heterochromatin is inert in both directions.

PCSD 36-state enrichment heatmap of down-regulated genes across OSDR datasets — PCSD 36-state enrichment of down-regulated genes per OSDR dataset (Table S12). The OSD-314 row uses the sweep's original contrast; see the audit below.

2 · The repressed class is a redox/peroxidase module

specificAgainst the genome, 18 GO terms are significant, led by peroxidase activity (11.1×), hydrogen peroxide catabolism (13.6×) and heme binding (7.2×). Against a state-matched Polycomb control, 11 remain, so the flight-repressed subset is functionally distinct within the Polycomb class. Its promoters are rich in AT-hook (AHL) and ZHD motifs, but no motif survives the state-matched comparison: the motif layer tracks the state, not the response.

GO dot plot comparing the Polycomb-down set to the genome and to a state-matched control — GO over-representation of Polycomb-down genes against the genome and against the state-matched Polycomb control (Table S19).

The module also replicates in an independent ISS centrifuge experiment. In OSD-314 (0.3g vs 1g), all 10 top OSD-37 GO terms are enriched at BH < 0.05 among 268 Polycomb-down genes (Table S63).

Four-panel OSD-314 diagnostics comparing published and corrected contrasts — OSD-314 diagnostics on the published pooled contrast (grey) and the corrected contrasts (blue): null calibration, dose-response, threshold-free enrichment and GO replication (Tables S58–S63; `code/qc_osd314_figure.py`).

3 · No methylation change, and no SOG1 binding

nullOSD-217 WGBS (Wassilewskija, Ws; root and leaf) shows no flight-induced methylation change at the module that survives correction (minimum padj 0.07 gene body, 0.099 promoter). In OSD-496 (Col-0 SOG1-3xFLAG seedlings after 75 Gy gamma irradiation), no SOG1 peak falls in a redox-module promoter (0/23) or a Polycomb-down promoter (0/265) at either timepoint. The spaceflight repression is therefore not a recapitulation of SOG1's direct DNA-damage targets.

SOG1 peak overlap with gene-set promoters — Fraction of promoters with a SOG1 ChIP-seq peak, by gene set and timepoint (Table S26).

4 · The bias replicates in ISS-grown roots

replicatedSample conditions for OSD-193 and OSD-218 were recovered from NCBI BioSample. With age as a covariate, Polycomb-state odds for down-regulated root genes are 3.06 in OSD-193 Col-0 (padj 2.2×10⁻¹⁰), 3.12 in OSD-218 Col-0 (padj 7.6×10⁻¹¹) and 3.74 in OSD-218 WS (padj 6.5×10⁻⁷), about half the seedling effect. OSD-624, a suborbital flight, is null.

Bar charts comparing published and corrected Polycomb odds ratios — Polycomb odds ratios under the published contrast (grey) and the corrected contrast (blue) (Tables S48, S54; `code/qc_figure.py`).

5 · Power and design

design limitSubsampling the Col-0 roots of OSD-218 gives 70% power at 6 replicates per group, and the full 8 v 8 set is significant. A single-genotype design therefore needs about 6–8 replicates per group. The original estimate of 9–12 came from a pool that mixed Col-0 and WS. A stratified OSD-120 flight × light analysis bounds the interaction but cannot resolve it: no module gene shows a significant interaction with either the pooled or the Col-0 module.

Power curves for pooled and Col-0-only subsampling, and OSD-120 module medians per light stratum — Empirical power for pooled (grey) and Col-0-only (blue) OSD-218 subsampling, and the OSD-120 module bound (Tables S43, S65–S67; `code/qc_col0_figure.py`).

Pre-submission audit

Before submission, every dataset's design was re-read from its OSDR study record. The analysis was then re-run from the public counts with the original pinned environment, and it reproduced the archived numbers exactly. The audit found four sample-coding or labelling errors, all corrected in the manuscript:

Dataset	Problem	After correction
OSD-314	Mars-gravity (0.3g) arm pooled into "0G"	Polycomb enrichment of up genes holds on every applicable diagnostic, but moderate (odds 1.5–3.6, not S11 22.4); the redox GO module replicates 10/10
OSD-218	16 Wassilewskija (WS) samples labelled Col-0	Odds unchanged, now attributed to two ecotypes
OSD-406	Three genotypes and two rockets pooled	Odds 3.17 with covariates (discovery level)
OSD-624	Described as orbital	Virgin Galactic suborbital flight

A stale 5-group sweep table was replaced, a column-order bug that misplaced significance stars in one heatmap was fixed, and a wrong GEO accession for the SOG1 data was corrected (GSE112529). All references were checked against Crossref.

Conclusions

The Polycomb-state bias of spaceflight repression is robust across datasets, tissues, ecotypes and state resolutions.
The repressed genes form a redox/peroxidase module that is distinguishable by function, but not by motif, within the Polycomb class.
The mechanism is not the canonical SOG1 damage pathway, and DNA methylation does not detectably change, so the upstream regulator remains unidentified.
The decisive next experiment is a spaceflight H3K27me3 ChIP-seq or CUT&RUN assay with ≥8 replicates per group of a single genotype.

Limitations. All chromatin states are ground-state annotations, so the decoder tests enrichment in pre-existing states, not dynamic mark changes. OSD-193 and OSD-218 share a payload. Motif and GO results are correlative. All results are discovery-only.

Code & data

Folder	Contents
`code/`	Analysis scripts (`chromdecode.py` is the core module), QC re-runs, HPC provenance
`supplementary_tables/`	Tables S1–S68 (CSV) with an index of content, producing script and section
`figures/`	All figures, SVG + PNG
`manuscript/`	LaTeX manuscript and supplement, Crossref-verified bibliography, figure audit
`DATA_SOURCES.md`	Every public input (OSDR, PCSD, AraENCODE, PlantCADB, JASPAR, GO, SRA)

Source data are public and not redistributed. Zenodo DOI: to be assigned on first release.