TIC-TOC — Cotton root growth & transcriptome in spaceflight

The finding

Against a shared, strongly tissue-specific spaceflight response (a canonical hypoxia/wounding induction with translational suppression in wild-type roots), the AVP-OX engineering acts as an attenuator, not an amplifier of the defence/stress programme — while the engineered lines sustain a larger primary-root spaceflight response than wild type. In short: AVP-OX cotton appears to experience microgravity as less stressful while growing more root — a favourable trait for space agriculture.

Supported by four independent analyses:

Morphometrics

AVP-OX larger primary-root flight response (P ≤ 0.04)

GO / KEGG

Jasmonate/wounding defence suppressed in AVP-OX

PhysioSpace

Lower osmotic/wounding stress scores vs WT

WGCNA

Defence module down in AVP-OX; growth-coupled module up

Figures

Experimental design and analysis workflow — **Fig 1.** Full factorial design (Treatment × Genotype × Tissue, 48 RNA-seq libraries), paired root imaging (days 3–6 → RSML), and the reproducible analysis workflow.

Root architecture over days 3-6 — **Fig 2.** Root growth days 3–6, Flight vs Ground by genotype. Total length diverges (lateral-driven); primary-root effect is the robust AVP-OX signal.

Differential expression counts per contrast — **Fig 3.** Spaceflight DEGs per genotype × tissue — root ≫ shoot; AVP-OX interaction contrasts down-biased.

WGCNA module-trait heatmap — **Fig 4.** Root co-expression modules (named by GO) vs Flight, AVP-OX and root architecture. Turquoise (signalling/isoprenoid) tracks growth; brown (defence) is down in AVP-OX.

GO enrichment dotplot — **Fig 5.** GO enrichment: WT roots induce hypoxia/wounding and suppress translation; A68 shifts to defence; the AVP-OX interaction down-regulates jasmonate biosynthesis.

PhysioSpace stress score heatmap — **Fig 6.** PhysioSpace decoding: spaceflight roots (left) activate a broad osmotic/drought/wounding stress signature; AVP-OX roots score lower than WT; shoots (right) are muted.

Autoencoder expression-program atlas — **Fig 8.** Deep-learning expression-program atlas — an autoencoder clusters genes into tissue-specific co-expression programs (bulk RNA-seq; programs, not single cells), annotated by GO function and Arabidopsis cell-type markers.

Program stress response WT vs AVP-OX — **Fig 9.** Program-level stress response to spaceflight, WT vs AVP-OX. The engineered lines attenuate the stress/defence programs (orange) while sustaining growth/metabolic programs (purple) — localising the whole-transcriptome effect to specific functional programs.

**Fig 7.** Integrative model: tissue-specific stress response, AVP-OX as an attenuator, and the growth-coupled module — less stress, more root.

Spaceflight DEGs (Flight vs Ground)

DESeq2, P.adj < 0.05, |log₂FC| ≥ 1 (ashr-shrunk). Root responds far more than shoot.

Contrast	Up	Down
WT · Root	2503	578
A68 (AVP-OX) · Root	2806	3610
D130 (AVP-OX) · Root	420	299
WT · Shoot	335	259
A68 · Shoot	79	419
D130 · Shoot	327	890

Reproducible pipeline

Every result regenerates from committed scripts (R 4.6; versions in ENVIRONMENT.txt).

deseq2/ — DESeq2 contrasts + factorial (V4) model
crosswalk/ — Gohir → Arabidopsis ortholog map (CottonGen BLASTP)
go_analysis/ — clusterProfiler GO / KEGG
physiospace/ — PhysioSpace stress-pattern decoding
wgcna/ — co-expression modules (GO-named)
morphometrics/ — RSML trait extraction + mixed models
integration/ — transcriptome ↔ root-architecture model

Honest limitations

The large total-length flight effect is lateral-root driven; absolute magnitude is pending Flight/Ground image-scale calibration. The primary-root AVP-OX result is robust to this.
RNA-seq libraries are not individually paired to imaged plants, so the expression ↔ architecture integration is group-level (n = 6) and correlational — pathways are consistent with driving growth, not proven causal.
Ortholog mapping is best-BLASTP-hit (approximate); single mission.

Data & code availability

Raw reads, count matrices, root images and RSML tracings: NASA OSDR/GeneLab (accession pending; released on publication). Analysis code and derived tables: this repository (data CC0-1.0; code MIT) and a forthcoming Zenodo archive. Grown in the ISS Veggie facility (custom Target Veggie Chambers); ISS National Laboratory (CASIS UA-2018-276) with support from the Target Corporation.