henhaobio
Sign in

bioRxivGenomicsPreprint

Evaluating the reliability of tools for mRNA annotation and IRES studies

Gemma May, Christina AkirtavaiD, C. Joel McManusiD

PDFDOI

In-depth summary

Claude Opus 5.5 · xhigh effort · 8 Nov 2026 · from the full text of version 2 (25 Sept 2026) · CC BY 4.0

A back-splicing circRNA reporter yields linear trans-spliced artifacts; in orthogonal assays three reported cellular IRESes are inactive, and Iso-Seq 5' ends agree with nAnT-iCAGE start sites.

Plasmid reporters for cellular IRES activity are described as prone to promoter-driven false positives, and 5' UTR misannotation makes this worse. The paper rebuts a recently proposed toolbox (back-splicing plasmid, smFISH, qRT-PCR) and proposes an alternative platform.

Key results

  1. ShownThe ZKSCAN1 back-splicing reporter with Hoxa9 or Chrdl1 promoters as IRES candidates makes abundant polyadenylated, linear, trans-spliced GFP mRNA, so GFP output cannot be assigned to IRES activity.

    RNase R removed the spliced GFP RT-PCR band, including from poly-A selected RNA. Spliced GFP ranked in the top 1% of transcripts in polyA RNA-seq; 11-15% of spliced GFP was recovered by polyA selection versus rRNA-depleted data. · The toolbox study’s back-splicing plasmid…

  2. ShownThree reported cellular IRESes (Hoxa9, Chrdl1, Dlx1 promoters) show negligible IRES activity in the Tornado circRNA plasmid and in A-cap/stemloop mRNA transfection, whereas cap-dependent translation is far stronger.

    In Tornado (n = 6), HCV IRES gave at least 1,400-fold more luciferase than each candidate; CrPV gave at least 5-fold more. In mRNA transfection (n = 4), m7G-capped Hoxa9 5' UTR gave 1,000-fold more luciferase than the Hoxa9 A-cap/stemloop construct. · Orthogonal assays find no IRES activity from…; Figure legends

  3. ShownPacBio Iso-Seq 5' ends from mouse embryonic somites and neural tube agree closely with nAnT-iCAGE start sites, contradicting the claim that Iso-Seq 5' ends are unreliable.

    Genome-wide, per-nucleotide 5'-end frequencies correlated strongly, with median Pearson's R from 0.8 to 0.9, rising with expression. Cited prior work reports 80% to 96.5% of Iso-Seq TSSs matching CAGE. · PacBio Iso-Seq data are supported by…

  4. ShownThe toolbox study's own Iso-Seq data contain no reads for the long Hoxa9 'IRES' mRNA isoform; its assignment came from Isoquant being given incorrect annotations. Without them, de novo isoforms match the short 5' UTR.

    Isoquant rerun without Hoxa9/Hoxa10/Mir196b annotations gave five isoforms (a-e). Isoforms a (2,105 nt) and b (1,872 nt) held 93% of Hoxa9 reads; one read (0.96%) extended into the 'IRES' region; none matched the 3,226+ nt isoform. · Reanalysis of PacBio Iso-Seq data shows the…; Isoquant analysis without user-provided Hoxa9…

  5. ShownThe Hoxa9 'IRES' smFISH signal is overwhelmingly nuclear, consistent with pri-mir196b and intronic ncRNA. Separately, a non-translating spiked-in RNA sedimented in polysome fractions.

    In the published smFISH images, 96% of 'IRES' probe signal colocalized with DAPI nuclei. In HEK293T polysome lysate, in vitro transcribed 'Scramble 1' RNA migrated into polysomal fractions 4-6 by qRT-PCR (n = 4). · ncRNA from the Hoxa9 promoter (“IRES”) localizes…; Association of non-translating RNAs in polysome…

Methods, data and reproducibility

Recreated the back-splicing reporter plasmids and probed for trans-spliced artifacts by RT-PCR, RNase R and RNA-seq. Tested candidate IRESes in Tornado and mRNA-transfection luciferase assays, and reanalyzed published Iso-Seq, nAnT-iCAGE and smFISH data.

  • polyA+ RNA-seq of HEK293T cells transfected with back-splicing plasmids carrying Hoxa9, Chrdl1 or Dlx1 promoters (one replicate also rRNA-depleted) · human (HEK293T cell line) · n = two biological replicates per plasmid; one rRNA-depleted sample · PRJNA1526329
  • Luciferase assays of candidate IRESes in Tornado plasmid and A-cap/stemloop or m7G-cap mRNA, in HEK293T and C3H/10T1/2 cells · human; mouse (C3H/10T1/2) · n = n = 6 (Tornado); n = 4 (mRNA)
  • PacBio Iso-Seq reads from E11.5 mouse somites and neural tube, generated by Koch et al. and reanalyzed (mm39) · mouse
  • nAnT-iCAGE TSS data from matched E11.5 mouse embryonic tissue (Ivanov et al.) · mouse
  • Published smFISH images (Koch et al. Fig. 3D) reanalyzed for nuclear versus cytoplasmic puncta · mouse · n = six image files
  • Polysome sucrose gradient with spiked-in Scramble 1 RNA vs endogenous GAPDH mRNA · human (HEK293T lysate) · n = n = 4 technical replicates
Methods & tools
Tornado circRNA plasmid, PacBio Iso-Seq, nAnT-iCAGE, Isoquant 3.6.3, minimap2, cutadapt 4.5, hisat2, regtools, deepTools bamCoverage, Fiji with ComDet, IGV, RNase R
Reagents & environment
Versions given for Isoquant (3.6.3), minimap2, cutadapt (4.5), deepTools (3.51), bedtools (2.30.0), Fiji (2.16.0), ComDet (0.5.6) and R (4.2.0); genome mm39, GENCODE M36, RefSeq annotation dated March 15, 2025.
Reproducibility gaps
Raw RNA-seq will be released only upon publication. Custom Perl scripts are named but no repository link is given. Plasmid sequences, primers, luciferase and TSS tables are in the extended data files.
Limitations (4)
  • The Hoxa9 and Chrdl1 promoters gave slightly more luciferase than negative controls in Tornado, possibly from rare trans-ligation of linear products. · stated by the authors
  • PacBio Iso-Seq can still amplify non-coding RNAs and pre-mRNAs that contain internal genomic poly-A stretches, so it should be cross-checked against orthogonal TSS data. · stated by the authors
  • The authors say cellular IRESes may exist; only a few candidates (Hoxa9, Chrdl1, Dlx1, CDK11a) were tested, in two cell lines. · noted in this summary
  • Isoform read groups were sorted manually and rest on small read counts; the isoform and smFISH reanalysis centres on one locus (Hoxa9) using another group's published data. · noted in this summary

Read this if you design or interpret IRES reporter assays, or need to judge how reliable long-read 5' ends are for transcript annotation.

Citations
1

Abstract

Since the discovery of viral Internal Ribosome Entry Sites (IRESes), researchers have sought similar elements in mammalian genes, termed "cellular IRESes". However, the plasmids used to measure cellular IRES activity are vulnerable to false-positives due to promoter activity in candidate IRESes. Orthogonal methods are needed to validate putative IRESes while carefully avoiding known artifacts. Recently, Koch et al. proposed approaches for studying IRESes, primarily circular RNA-generating plasmids, and for validating mRNA transcripts using smFISH and qRT-PCR. Here, we demonstrate confounding variables and artifacts in each approach that can lead to unwarranted conclusions about potential cellular IRES activity. We show the back-splicing circRNA plasmid creates linear mRNA artifacts associated with false-positive IRES signals. Using orthogonal assays validated with viral IRESes, we find putative cellular IRESes reported using the back-splicing plasmid have no IRES activity. Furthermore, we demonstrate that smFISH and qRT-PCR can misidentify nuclear ncRNAs as mRNAs and we validate a single molecule sequencing assay for identifying mRNA 5' ends. Our work establishes reliable methods for robust transcript annotation and IRES studies that avoid documented artifacts.

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
THISMay '26Zhao '20Cortot '26Koch '26Akirtava '22Aoun '26
63% semantic matchGenetics
Gemma E May, C. Joel McManus·bioRxiv·May 9, 2026· 1 citations