henhaobio
Sign in

bioRxivBiophysicsPreprint

Decoding Radiation-Induced Transcriptomic Signatures of Human Leukocytes using Long-Read RNA-Seq: Clinical and Biodosimetric Implications

Ahmed SalahiD, Heinz SchmidbergeriD, Fédérico MariniiD, Sebastian ZahnreichiD

PDFDOICodeData
In-depth full-text digest available · Jump to digest ↓

Abstract

Background: Gene expression profiling in radiation-exposed blood is a valuable tool for biodosimetry and clinical research. Evaluating the blood's transcriptomic radiation response provides insight into absorbed dose, hematotoxicities, and immune reactions. However, detailed analysis using long-read RNA sequencing is currently limited in its diffusion, despite the potential additional insights that could be extracted, including novel isoform discovery and on-the-field gene expression studies, owing to its portability. Results: In this study, we utilized Oxford Nanopore Technologies' long-read RNA sequencing on human peripheral blood leukocytes from three healthy donors 6 hours after exposure of whole peripheral blood to 4 Gy of X-rays. Compared to sham-irradiated (0 Gy) blood samples, gene-level differential expression analysis identified 117 upregulated and 66 downregulated genes, including canonical DNA damage repair and inflammatory responses. At the transcript level, 102 transcripts were significantly upregulated, and 17 were downregulated, revealing isoform-specific regulation that was not captured at the gene level. Notably, IL32, which was not detected as significantly changed at the gene level, exhibited significant upregulation of two transcript isoforms (log2 fold-changes of 1.19 and 1.99), while WDR74, ITM2B, AK2, and RPS19 displayed changes in transcript usage following irradiation. Leveraging long-read RNA sequencing, we further identified novel candidate isoforms, including protein-coding isoforms of PHPT1 and FBXO22 as well as non-coding candidates. Notably, the FBXO22 isoform showed evidence of altered protein domain architecture relative to the reference isoform. Conclusions: This study is the first to systematically investigate differential transcript usage in the leukocyte fraction of irradiated human whole blood using long-read RNA sequencing. We confirmed known radiation-response signatures at the gene level and identified isoform-level transcriptomic changes, including several isoforms that were undetectable in our previously generated matched short-read dataset. These findings demonstrate the potential of long-read RNA sequencing to provide a more detailed view of radiation-induced transcriptomic alterations, with potential applications in biodosimetry and clinical settings.

In-depth summary

Claude Opus 5.5 · xhigh effort · 8 Nov 2026 · from the full text of version 2 (29 Aug 2026) · CC BY 4.0

Nanopore long-read RNA-seq of leukocytes from 4 Gy X-irradiated blood (three donors, 6 h) recovers p53 signaling and reveals isoform-level regulation plus candidate unannotated FBXO22/PHPT1 isoforms.

Blood transcriptomic biodosimetry has relied mostly on microarrays, qPCR or short-read RNA-seq, which give limited isoform resolution, and the only prior nanopore study (isolated mononuclear cells) mainly validated established gene signatures. This work tests irradiated whole-blood leukocytes at gene, transcript and transcript-usage levels with long reads and benchmarks the calls against a matched short-read dataset from the same donors.

Key results

  1. ShownX-irradiation (4 Gy, 6 h) of whole blood induced a donor-adjusted leukocyte gene program dominated by p53-dependent DNA damage response and immune/inflammatory signaling, recovering canonical radiation-responsive genes.

    edgeR (~Donor + Treatment, FDR ≤ 0.05) on 13,020 filtered genes: 117 up, 66 down, incl. FDXR, CDKN1A, CD83; DESeq2 found 120 (117 shared), LFC R = 0.99. KEGG p53 signaling (hsa04115) GeneRatio = 15/89, padj = 1.8×10^−12; GO B cell proliferation padj = 0.0012. · Gene-level ONT RNA-Seq data reveal…; Differential expression and differential…

  2. SuggestedTranscript-level testing exposed isoform regulation missed at gene level: two IL32 isoforms rose with no gene-level change, and four genes (only ITM2B also a DEG) shifted isoform usage; matched short-read data could not test one IL32 isoform, found the other nonsignificant, and found no usage shifts.

    Transcript-level DE: 102 up, 17 down of 14,146 filtered transcripts. IL32 ENST00000440815 and ENST00000325568 up (LFC 1.19 and 1.99). DTU (diffSplice/stageR) hits RPS19, WDR74, AK2, ITM2B; WDR74 and ITM2B with two isoforms each. Short-read (SR): ENST00000440815 failed filtering, ENST00000325568 FDR = 0.90; no SR DTU gene at 5% FDR. · Transcript-level analysis reveals…; Discussion

  3. SuggestedIsoQuant discovery produced differentially expressed unannotated transcript models; combined coding-potential, homology, domain and junction-support screens prioritized FBXO22 and PHPT1 coding candidates, with FBXO22 candidate ORFs each keeping only one of the reference F-box and FIST C domains.

    26 unannotated DE transcripts (24 up, two down) from 22,512 models (3,579 after filtering). 8 coding by CPC2, CPAT and TransDecoder; seven of eight BLASTp ≥94% identity. SQANTI3 passed 19 of 26. PHPT1 novel junction: 36 SR reads, all six LR libraries; FBXO22 novel donor: no SR, 4 LR reads in two of six libraries (irradiated only). · Transcript-level analysis reveals…; Discussion

  4. ShownAgreement with the matched short-read dataset was partial: DEG sets overlapped modestly and a previously reported 34-gene radiosensitive signature was not fully replicated, though the full short-read DEG set showed a coordinated long-read response.

    0 vs 4 Gy, 6 h: 116 SR vs 183 LR DEGs, 58 shared. 34-gene SR signature: 29 testable, only G0S2 significant in LR; direction concordance 18/29 (62.1%, P = 0.265); LFC ρ = 0.41, P = 0.027; 116 SR DEGs as a set in LR, mroast P < 0.0005. Detected genes SR 12,329 vs LR 10,650; 1,470 LR-only, largely lncRNAs (504) and processed pseudogenes (574). · Discussion; Differential expression and differential…

  5. ShownGene-level effect sizes were stable when any single donor was removed, and depth subsampling showed detection near, but not fully at, saturation, with large-effect radiosensitive genes found even at shallow depth.

    Leave-one-donor-out: 76 to 124 of 183 DEGs (42–68%) stayed significant; LFC Spearman r = 0.80–0.86 vs full model. subSeq (20 replicates per depth): mean recovery of full-depth DEGs 7.8% at 5%, 70.1% at 50%, 93.3% at 90% depth; mean Spearman 0.364, 0.838, 0.976; 28 of 183 robust incl. CDKN1A, FDXR, AEN, MDM2. · Discussion; Differential expression and differential…

Methods, data and reproducibility

ONT PCR-cDNA long-read sequencing of total leukocyte RNA from sham or 4 Gy X-irradiated whole blood (three donors, 6 h ex vivo), quantified with minimap2 and IsoQuant (GENCODE v47 plus discovery), then donor-blocked edgeR DGE/DTE, diffSplice/stageR DTU, and multi-tool coding-potential and junction-support screening of unannotated transcripts.

  • Total leukocyte RNA (LeukoLOCK) from whole blood of healthy non-smokers (two male, one female; ages 29 to 46) after 0 or 4 Gy X-rays, 6 h ex vivo; ONT PromethION cDNA libraries · human · n = three donors; six libraries (0 and 4 Gy) · 10.5281/zenodo.22013472 (processed counts only)
  • Previously published short-read (150 nt) whole-blood RNA-seq of the same donors, re-analyzed for 0 vs 4 Gy at 6 h (original series 0.5-4 Gy, 2 and 6 h) · human · n = n = 3 donors per group
  • GRCh38 reference genome with GENCODE v47 (long-read) and v46 (short-read re-analysis) annotations · human
  • 34-gene radiosensitive signature from the authors' prior short-read study, tested for replication in long-read results · human · n = 34 genes (29 present in long-read results)
Methods & tools
Dorado v0.8.2 (SUP basecalling, trim, demux), NanoPlot 1.44.0, minimap2 2.28-r1209 (spliced mode), IsoQuant 3.6.2, edgeR 4.7.5 (filterByExpr, quasi-likelihood, diffSplice), DESeq2 1.49.4, stageR 1.31.0, subSeq 1.40.0, Salmon 1.10.3 / tximport 1.22.1 (short-read re-analysis), SQANTI3 v6.0.2 (psauron) with STAR 2.7.11b junctions, TransDecoder 5.7.1, CPC2, CPAT 3.0.5, GffCompare 0.12.10, BLASTp (NCBI nr), InterProScan v5.78-109.0, clusterProfiler (GO, KEGG, MSigDB C5 BP GSEA), GeneTonic 3.3.3, limma mroast 3.66.0
Reagents & environment
Repo provides Conda environment RNAseq_analysis.yaml and R Markdown workflow run in R v4.5.1; Dorado v0.8.2 SUP model dna_r10.4.1_e8.2_400bps_sup@v5.0.0; ONT SQK-PCB114.24 libraries on R10.4.1 PromethION flow cell (PromethION 2 Solo, 72 h run).
Reproducibility gaps
Raw individual-level RNA-seq reads are not publicly deposited owing to data-protection rules and are available only on reasonable request to the corresponding author; only processed gene- and transcript-level counts are public.
Limitations (6)
  • Only three donors per condition, one dose (4 Gy) and one time point (6 h), with ex vivo irradiation that does not fully mimic in vivo exposure; the authors call for orthogonal confirmation in a larger cohort balanced for sex and age. · stated by the authors
  • Without differential blood counts or computational deconvolution, changes in leukocyte subset proportions cannot be separated from true per-cell transcriptional responses. · stated by the authors
  • Unannotated isoform structures and proposed functions rest on computational evidence; targeted qPCR or Sanger sequencing across the novel junctions is still needed, particularly for lower-confidence candidates. · stated by the authors
  • The long- vs short-read comparison is confounded by different blood processing and RNA extraction, annotation versions (GENCODE v47 vs v46), quantification pipelines and DE frameworks (edgeR vs DESeq2), so platform effects cannot be isolated. · stated by the authors
  • FBXO22, ranked a top coding candidate, carries a novel splice donor with no short-read support and only 4 long reads in two irradiated libraries, leaving its full-length structure, and thus the predicted domain split, weakly supported. · noted in this summary
  • IsoQuant labeled 53.4% of 113,225,145 assigned reads as ambiguous (22.0% unique), so transcript-level and transcript-usage calls depend on ambiguous-read handling whose effect on results was not tested. · noted in this summary

Read this if you study radiation biodosimetry or blood transcriptomics and want a worked nanopore pipeline for isoform-level differential expression, transcript usage testing, and vetting of unannotated isoforms.

Citations
1

Reproducibility ledger

Nobody has reported trying to reproduce this paper yet. Did you replicate an experiment or run the code? Your report saves the next lab a week.

Sign in to file a reproducibility report.

Related papers

Similarity Map
SelectedSame fieldCross-field
THISSalah '25Rouchka '17Salah '26Chen '25Byrne '17Lane '26
66% semantic matchBiophysics
Ahmed Salah, Daniel Wollschläger, Maurizio Callari+3 more·Scientific Reports·Jun 5, 2025· 14 citations