Circular RNA Clock Ticks After Death: circRtn4 Emerges as Candidate Marker for Estimating Time of Death

One of the oldest questions in forensic science—how long ago did this person die—may soon be answered with a molecular stopwatch built from one of biology’s most resilient molecules. A team of researchers at Southwest Medical University in Luzhou, China, reports that a circular RNA called circRtn4, harvested from brain tissue, changes its abundance in a predictable, mathematically describable way after death. Their work, published in Frontiers in Veterinary Science, offers a new candidate biomarker for estimating the postmortem interval, or PMI: the often-frustrating gap between the moment an organism dies and the moment its body is found and examined.

Estimating the PMI has always been part science, part art. Pathologists traditionally rely on the classic postmortem signs taught in every forensic curriculum: the cooling of the body (algor mortis), the purplish pooling of blood (livor mortis), the stiffening of muscles (rigor mortis), and later, the progression of putrefaction or the developmental stage of insects colonizing a corpse. Each of these indicators is heavily influenced by ambient temperature, humidity, clothing, body mass, and a hundred other variables. The result is that conventional estimates are often empirical and imprecise, and most methods are only reliable within roughly the first 24 hours after death. Because a precise time of death can narrow a suspect pool, validate or destroy an alibi, and define the scope of an investigation, forensic scientists have long sought molecular markers whose degradation behaves like a dependable clock.

RNA has been an obvious candidate, but also a problematic one. Messenger RNA, the workhorse molecule that carries genetic instructions to the protein-making machinery, degrades rapidly after death, shredded by the ribonucleases that pervade both living organisms and the environment. That instability can itself be informative—several studies have used mRNA decay to estimate early PMI—but it also makes mRNA a poor internal reference standard for quantitative work. MicroRNAs, which are shorter and tougher, have shown promise, yet their degradation is still influenced by tissue origin, cause of death, and temperature. What forensic researchers ideally want is a molecule that is abundant, tissue-specific, and stable enough to survive hours or days of postmortem enzymatic attack while still changing on a measurable schedule.

Enter circular RNAs. First discovered in 1976 in viroids, circRNAs are a class of non-coding RNA molecules in which the strand has been spliced into a covalently closed loop. Unlike linear RNAs, they carry no 5-prime cap and no 3-prime poly(A) tail, which makes them largely invisible to the exonucleases that chew up ordinary RNA from the ends. This closed-loop architecture confers remarkable stability and long half-lives. CircRNAs are also highly abundant and show exquisite specificity to particular tissues and developmental stages—and nowhere are they more enriched than in the brain. Previous work has shown that circular RNA transcripts in mouse brain are abundant, conserved, and mirrored by similar transcripts in human brain tissue, making cross-species comparisons feasible. CircRNAs have already been explored in forensics for body fluid identification, bloodstain age estimation, and forensic age prediction, but a rigorous, temperature-controlled test of their use in PMI estimation has been lacking.

The Chinese team, led by Binghui Song, Jie Qian, and colleagues under the supervision of Jingliang Cheng and Junjiang Fu, began by mining public databases—circBase, circBank, the Tissue-Specific CircRNA Database, GTEx, and NCBI—to find circRNAs whose host genes are highly or specifically expressed in brain tissue. Four candidates emerged: circRtn4, circAtp5c1, circBcas1, and circBrsk2. The researchers designed divergent primers that only amplify the back-splice junction unique to circular transcripts, confirmed each molecule’s head-to-tail structure by Sanger sequencing, and then measured expression across 13 mouse tissues. CircRtn4 showed its highest expression in the brain, exactly as the database predictions had suggested. Of the four candidates, circRtn4 (technically mmu_circ_0000250, a 2,418-base-pair loop formed from exons 2 and 3 of the Rtn4 gene) stood out as the lead candidate, not least because it has a conserved human homolog, circRTN4 (hsa_circ_0054598), sharing 81 percent sequence similarity.

To turn degradation into a clock, the team needed stable reference molecules against which to normalize circRtn4’s signal. They screened eleven candidates, including common housekeeping mRNAs, ribosomal RNAs, a mitochondrial mRNA (mt-co1), and the well-known circular RNA circCDR1as. The results were telling. At 25 degrees Celsius, conventional mRNAs such as Tbp degraded fastest, while 28S rRNA, circCDR1as, and mt-co1 retained roughly 58 to 59 percent of their day-zero signal after eight days, with half-lives of about 8.6 to 9.5 days. The 18S rRNA, though initially stable, declined too quickly after day four to be useful. At the harsher 35 degrees Celsius condition, circCDR1as itself degraded too rapidly, leaving 28S rRNA and mt-co1 as the reference genes of choice. The lesson is one the authors emphasize repeatedly: reference gene selection cannot be assumed but must be empirically validated for every tissue, species, and temperature.

The modeling itself was meticulous. Sixty-two BALB/c mice were sacrificed under approved protocols, and brain tissue was collected at defined intervals—at 0, 1, 2, 4, and 8 days postmortem at 25 degrees Celsius, and at 0, 12, 24, 48, 72, and 96 hours at 35 degrees Celsius, each in humidity-controlled incubators set to 50 percent relative humidity. Using both semi-quantitative RT-PCR, where band intensity on agarose gels is quantified with ImageJ and standardized against a known DNA control loaded on every gel, and quantitative RT-qPCR, the team measured the circRtn4-to-reference ratio at each time point. They then fitted linear, quadratic, and cubic regression models, comparing them not just by R-squared values but with more rigorous statistics: the Akaike and Bayesian information criteria, leave-one-animal-out cross-validation error, nested partial F tests, and standard residual diagnostics. Quadratic models were statistically supported for the 25-degree data normalized to 28S rRNA and circCDR1as, while linear models were favored for mt-co1 at 25 degrees and for both references at 35 degrees; cubic terms added nothing.

The validation step revealed both the promise and the limitations of the approach. When models were tested on mouse samples with known times of death, prediction errors varied substantially across models, references, and time points. Some combinations performed impressively well—for example, the mt-co1-normalized linear model at 35 degrees estimated a 24-hour PMI with only a 3.8 percent error, and several 4-day predictions at 25 degrees were accurate within about 10 percent. But early PMI predictions, especially at the 2-day mark, sometimes erred by 30 to 85 percent. In the RT-qPCR analysis at 25 degrees, only the linear model showed a statistically significant relationship with PMI. A general pattern emerged: the longer the postmortem interval, the more accurate the estimation tended to be, echoing a known strength of molecular PMI methods for later intervals where traditional signs have faded.

Crucially, the team extended the work into human tissue, albeit in a strictly exploratory way. Cerebral cortex samples from a single healthy male donor, collected with informed consent roughly three hours after death, were incubated at 20 and 25 degrees Celsius and sampled across intervals extending to nearly five days. The human circRTN4 degraded sharply around two days postmortem and then remained low, and quadratic regression equations fitted to the data yielded R-squared values above 0.95 at both temperatures. The authors are careful to state that this single-donor analysis constitutes a proof of concept only—it cannot account for the variation in age, sex, cause of death, and environment that real forensic casework would encounter—but it demonstrates that the mouse findings have a plausible translational path.

What makes this study notable is its candor as much as its ambition. The authors openly acknowledge that circRtn4 performed poorly at 4 degrees Celsius, that mouse age (16-week versus 9-week animals across the two temperature experiments) may confound the comparison between temperature conditions, and that semi-quantitative RT-PCR, though cheap and visually intuitive, has a narrower dynamic range than the RT-qPCR or digital PCR platforms that future casework should adopt. Their conclusion is measured: circRtn4 may serve as a molecular PMI marker in mouse brain under the specific conditions tested, and postmortem circRNA models must be built species-by-species and temperature-by-temperature. Independent validation in larger human cohorts and veterinary-relevant species is required before any courtroom application. Still, the image is compelling—a loop-shaped RNA molecule, untouched by the enzymes that destroy its linear cousins, quietly recording the passage of time in the sealed vault of the skull. If larger studies bear it out, that silent loop could one day give investigators what they have always wanted: a molecular answer to the question of when death occurred.

Subject of Research: Use of the circular RNA circRtn4 as a candidate biomarker for postmortem interval estimation in mouse and human brain tissue

Article Title: CircRNA circRtn4 as a candidate molecular marker for postmortem interval estimation in mouse brain tissues

Article References: CircRNA circRtn4 as a candidate molecular marker for postmortem interval estimation in mouse brain tissues. (n.d.). https://doi.org/10.3389/fvets.2026.1941625

Image Credits: AI Generated

DOI: 10.3389/fvets.2026.1941625

Keywords: circular RNA, circRtn4, postmortem interval, forensic science, RNA degradation, PMI estimation, brain tissue, RT-qPCR, biomarkers, veterinary forensics, molecular forensics, reference genes