Read-through (cis-SAGe) chimeras: most 'fusions' called from RNA-seq in childhood ALL are normal conjoined genes

Prostate models; the CTCF-regulation component of this paper was subsequently qualified by the same laboratory (see E8).

• The founding report of this chimera assigned it to trans-splicing, and said so explicitly. Characterising SLC45A3-ELK4 in benign prostate and prostate cancer, and showing by quantitative PCR of DNA that its expression is NOT exclusive to cases carrying a chromosomal rearrangement, the original authors concluded it 'most commonly occurs through a mechanism other than chromosomal rearrangement (e.g., trans-splicing)'. The cis-splicing assignment is a later reinterpretation of the same molecule by a different group, not an uncontested finding, and the two accounts are still both in the literature. Rickman DS et al. 2009

EST/mRNA-driven catalogue from the pre-deep-RNA-seq era; the true number is certainly larger and the validation subset was selected rather than random.

• RT-PCR validation does not exclude the dominant artefact class, because the artefact is generated by reverse transcription itself. Yan et al. estimated from a cross-species spike-in that approximately 1% of RNA-seq reads are artifactually chimeric through random RT template switching. A conjoined-gene call validated by RT-PCR has passed a test that shares its failure mode with the original assay; the controls that do discriminate are reproducibility across independent biological isolates and enrichment experiments that raise true signal without raising chimeras. 291 of 353 validated is therefore a weaker number than it looks, though the existence of real conjoined genes is not in doubt. Read: abstract only. Yan B et al. 2021

Three cell systems (HEK-293T, HUVEC, LO2) plus donor blood; the validated set is small relative to the called set, so it bounds existence rather than prevalence.

• Some chimeras that look like a transcription-level phenomenon in 'normal' cells are germline DNA structural variants. A bottom-up search for population-specific chimeric RNAs in the GTEx cohort found 58 of them with distinct patterns across ancestry groups, provided direct evidence for 29 polymorphic chimeric RNAs with associated structural variants, and uncovered 13 novel rare structural variants. A validated chimera in a handful of cell lines and one donor blood panel therefore cannot be assumed to be readthrough transcription: the alternative is that the donor carries the rearrangement. Elfman J et al. 2024

10 cases; the exact proportions depend on the caller and on the normal reference used for subtraction, so 6.1% is illustrative of the magnitude rather than a fixed constant.

• There is a third category underneath both 'rearrangement' and 'conjoined gene': the chimeric read that corresponds to no molecule at all. Yan et al. measured the background directly by building RNA-seq libraries from human RNA spiked with RNA from an unrelated species (fruit fly) and counting the resulting cross-species chimeras, and estimated that approximately 1% of RNA-seq reads are artifactually chimeric, arising from random template switching during reverse transcription and from alignment error. Their discriminator is reproducibility: genuine splicing and genome-rearrangement events generate the same chimeric fragment across independent biological isolates, whereas artefactual chimeras do not. Severgnini's 245 candidate calls from 10 cases were not, on the abstract, filtered on cross-isolate reproducibility, so the 6.1% leukaemia-exclusive figure is an upper bound on real conjoined-gene biology and part of the remaining 93.9% may be neither leukaemic nor conjoined. This makes the entry's conclusion stronger and its mechanism attribution weaker. Read: abstract only. Yan B et al. 2021

TCGA-era pipelines; normal-tissue coverage was uneven across lineages, and haematopoietic normals in particular were sparse.

• Normal-tissue subtraction does not rescue a targeted assay, because the false positives appear in targeted assays too. Song et al. applied sensitive nested RT-PCR to a large series of healthy peripheral blood samples and detected every leukaemia-associated marker tested, including MLL-AF4, BCR-ABL p190 and p210, AML1-ETO, PML-RARA and CBFB-MYH11, at various prevalences in healthy individuals, plus aberrations in cord blood from healthy neonates. Filtering an untargeted caller against normal RNA-seq (E5) does nothing for a targeted nested PCR run on a single patient sample, which is the assay format most diagnostic and screening proposals in this beat actually use. Read: abstract only (PMC full text returned empty). Song J et al. 2011

A corrigendum was published (Int J Biochem Cell Biol 2019;111:11, PMID 30965217) and should be cited alongside; the numbers quoted are from the original abstract and were not re-checked against the corrigendum.

• A corrigendum was published to this paper and is cited here so the numbers are not quoted from the superseded abstract. The entry's own caveat flags that the 19,547 / 271 / 13 figures were taken from the original abstract and not re-checked; the corrigendum is the primary record of what changed and must be read before any of those counts is reused. Zhu D et al. 2019

56 non-malignant samples is a modest baseline; 'cancer-restricted' is relative to that panel.

• A 56-sample non-malignant baseline cannot exclude an ancestry-structured germline fusion. Population-specific chimeric RNAs exist at appreciable frequency - 58 identified in GTEx with distinct patterns between ancestry groups, 29 of them traced to underlying structural variants - so a chimera absent from a modest, likely ancestry-skewed control panel can still be a polymorphism rather than a cancer-restricted event. 'Cancer-restricted' is a statement about the control panel's size and composition. Elfman J et al. 2024

A negative/qualifying result from the same laboratory as the original claim, which is credible but leaves the regulatory mechanism unexplained.

• NO PUBLISHED OPPOSITION FOUND, and the falsifier is specific. This negative result has not been re-tested by an independent group: PubMed 'SLC45A3-ELK4[tiab] OR (SLC45A3[tiab] AND ELK4[tiab])' returns 12 records (checked 2026-09-19), and title-checking every one leaves only two that address CTCF at this locus - the original cis-SAGe report and this qualifying paper, both from the same laboratory. What would falsify the qualification is a study that measures CTCF occupancy directly at the SLC45A3/ELK4 intergenic boundary in the SAME cells in which the chimera is quantified, rather than comparing across cell lines whose genotypes and CTCF landscapes differ; no such experiment is published. The nearest adjacent work, a later study of the same chimera by the same group, characterises it as a long non-coding chimeric RNA and does not revisit CTCF. Qin F et al. 2017