How Much Can Enhanced DNA Repair Explain Longevity Differences Between Species?

Three measurements need separating. Damage is a DNA abnormality awaiting resolution; a mutation is a sequence change that remains; reporter restoration is recovery of a particular experimental output. After a double-strand break, homologous recombination usually uses a homologous template, whereas end joining connects ends. End joining can preserve sequence or leave insertions and deletions after end processing. Completing more repair events does not necessarily leave fewer errors; fixed-time reporter restoration also does not identify the true repair rate by itself. s9 s14

The 2025 study combines reporters, mismatch repair, an endogenous PTEN site, colony formation and chromosome-stability measurements. It does not show that every pathway is enhanced: nucleotide excision repair was broadly comparable and the higher base-excision trend was not significant. s1 A study across 18 rodents retained the double-strand-repair/lifespan association after phylogenetic correction and used SIRT6 residue swaps to alter repair and survival in a heterologous fly system. This strengthens a transferable mechanism, while leaving coevolution unresolved. A correlation model's explained variance is not a causal share of lifespan. s8

Whale fibroblasts did not require more perturbed pathways than human fibroblasts to form tumors experimentally. Additional transformation barriers therefore cannot by themselves explain the maintenance strategy. These are engineered-cell and transplantation models, not a survey of natural cancer incidence. Tumor variants also follow expansion and selection, rather than measuring natural accumulation across tissues of old whales. s1

Conditions matter. Whale cells were cultured at 33 degrees C, versus 37 degrees C for the main comparators; cooling human cells also accompanied higher CIRBP and reporter restoration. Physiological adaptation is relevant, but temperature and species contributions are not fully separated. This is not a cold-bath efficacy experiment. s1 s3 Cross-species assays generally used donor-derived lines, while some modified lines, clones and repeated experiments are technical repeats. Hundreds of nuclei or thousands of PCR reads cannot supply hundreds of independent donors. Significance from pooled individual-cell focus measurements needs interpretation at that hierarchy. s4 s5

The public PTEN family has three whale, three mouse, two cattle and four human treated lines, plus five public untreated controls, termed mocks below. Human HCA2 is neonatal, and four public lines are not unambiguously mapped to the three biological repeats in the figure. This audit processes all 17 experiments and 4,688,557 read pairs, with table means weighting the included treated lines equally. s3 s6 The table uses the prespecified physical R2, 30-bp cut window and local minimum Q20. Ranges describe variation across lines, not confidence intervals.

Restricting humans to the three adult-derived lines gives 31.2% mean unmodified; leaving out each human line in turn gives a mean range of 26.9%–31.5%. This is not an age-effect experiment: donor, culture and line differences remain mixed.

The five mocks also have only 89.6%–91.9% unmodified under the primary condition, with substitutions dominating nonreference calls; higher quality thresholds do not eliminate this background. Mocks are not matched to every treated line, so simple subtraction does not establish a true repair rate. Pre-existing sequence, amplification/sequencing and alignment contributions remain unresolved; every nonreference treated call cannot be attributed to error after cutting. Mock deletion >20-bp background is about 0.0016%–0.0146%, lower than treated fractions, but one locus still cannot identify cumulative repair histories. Whale and human unmodified line ranges overlap; species means do not rank every line. s6

An unmodified call only means an amplifiable, callable fragment matches the mock reference in the specified window. PCR does not count independently tagged original DNA molecules; unamplified large deletions and cells leaving the sampled population do not automatically enter the denominator. Read direction and equally optimal alignment representations can affect calls. Both orientations were evaluated; finite checks agreed with another implementation on principal categories but retained an equally optimal difference in deletion-length representation, so no complete exact deletion-length spectrum is claimed. s6

Figure: dots are treated lines, the diamond is neonatal HCA2, and bars are equal-line means, not confidence intervals. Left and middle denominators are called PCR reads; the right denominator is all input reads. Deletions >20 bp can include complex calls and are not added to the five exclusive principal categories. s6

Across read ends, windows and quality thresholds, species means retain the direction of more whale unmodified calls and fewer deletions >20 bp. Callability is uneven: cattle BT2 has only 11.6% callable under the primary condition, with about 56.8% of input excluded by this audit's whole-alignment substitution threshold. Genuine complex changes could also be excluded; these reads cannot all be diagnosed as sequencing-quality failures. The results are technical sensitivities of conditional spectra, not additional independent repeats. s6

Raising the minimum quality from Q10 to Q30 changes mean unmodified from 32.7% to 52.0% in whales and 21.7% to 25.9% in humans. Mean callability falls from 88.0% to 5.0% for whales and 88.3% to 6.3% for humans. The threshold changes the retained fragment family. Deleted reference bases have no quality value, so a minimum-local-quality filter can select intact and deleted fragments differently. A stricter threshold does not automatically yield a more reliable faithful-repair probability. All R1 and 20/40-bp conditions remain in the package as technical sensitivities of the same read family.

Figure: the same complete family, R2 and 30-bp window, with equal-weight means of treated lines. Lines are not independent experiments or confidence intervals. s6

Even perfectly measured unmodified fractions mix uncut DNA with accurately restored DNA. In a one-cut idealization without selection, cut probability c and faithful restoration probability r give 1-c+cr. Cutting 25% with 20% faithful restoration and cutting 100% with 80% restoration both give 80% unmodified. These are illustrative chosen parameters. Survival, proliferation and amplification can further change observed fractions. The original ddPCR time course, labeled-RNP transfection and pre-cut reporter controls matter: they weaken a simple failure-to-cut explanation. A transient broken state still does not identify every allele's cumulative cutting history. s1 s3

Figure: all parameters are chosen identification examples, without fitting to whales, flies or people. The one-cut model omits re-cutting; the competing-outcome model concerns one lesion.

CIRBP is more than an abundance correlate. Increasing it in human cells or depleting it in whale cells changes reporters. Damage-associated RNA and poly(ADP-ribose) signaling contribute to localization; cell-free experiments show improved repair-protein binding, joining and protection against excessive end degradation. These offer a mechanism for limiting harmful end processing. Cellular timing between protection, processing needed for homologous recombination and pathway coordination remains unresolved; protective condensates are a hypothesis, not an established complete mechanism. s1 An independent 2018 cell study supports its damage-response role. CIRBP depletion reduced repair and also reduced formation of gamma-H2AX, Rad51 and 53BP1 foci. Fewer foci alone therefore cannot establish less damage. s9

Stars do not establish effect size. The supplementary table reports an approximately 4.69-percentage-point increase in unmodified PTEN, with unadjusted P=0.0472; four stars in the figure are not reconciled with that single-category value. Without the original comparison output, this audit cannot confirm that they describe the same statistical layer. A seven-category multiplicity sensitivity does not supply comparably strong support. Other reporter, depletion and chromosome results remain relevant. s1 s3 Public Nanopore code is not the complete author matrix: some functions use insertion-plus-deletion event counts as denominators, which differ from all-allele fractions. s7

The flies provide actual animal survival outcomes. These female-fly ordinary-lifespan models account for cages and transfer day; medium induction gives hazard ratios of about 0.73 for human CIRBP and 0.75 for whale CIRBP. These are not 27% and 25% increases in lifespan. At high induction, the ordinary-lifespan ratios are about 1.21 for human protein and 0.95 for whale protein: an unadjusted opposite-direction signal and no clear extension, respectively. Survival after 650 Gy irradiation is a separate stress outcome. s1

An explicit BH sensitivity across all 12 protein, induction-level and endpoint comparisons gives about 0.054 for several whale effects and the high-human adverse signal; the low/medium human ordinary-lifespan positives remain. This diagnostic neither proves zero effects nor refits the trial. Whale medium post-irradiation survival has P=0.0043 inside the main plot but 0.043 in its caption and the complete table. Both are retained. The nonsignificant RU486 controls also have wide intervals, so they do not strictly exclude drug or genotype-dependent effects. s1

Figure: all 12 protein/induction/endpoint log hazard ratios and standard errors are transformed to hazard ratios with approximate 95% intervals, without refitting individual flies. Low, medium and high mean 50, 100 and 200 µM RU486 induction, not protein concentration. Ordinary lifespan and survival after 650 Gy are separate; HR does not translate into a lifespan-extension percentage. s1

Fly survival does not uniquely identify repair as the mediator. CIRBP binds RNA, and a recent mouse-muscle abstract reports ATP6/energy-metabolism and functional effects. Only that abstract was obtained here, so it cannot establish the complete mechanism or lifespan benefit. s15 Intracellular maintenance and extracellular signaling also differ: extracellular CIRP can activate inflammation experimentally, a different estimand from increasing intracellular expression. s10

Broader repair enhancement still needs organismal outcomes. Terminal brains in a DREAM loss-of-function model had fewer mutations, while earlier work on the same genetic model reported systemic amyloidosis and shorter survival. This does not establish that better repair itself is harmful; changing a regulator also changes other functions. Fewer brain mutations cannot replace net benefit across the organism. A September 2026 correction removed an external mutation-rate benchmark that did not account for culture expansion; it is not reused here. s11 s12 s13

Whale cells offer a testable maintenance strategy, and CIRBP/SIRT6 perturbations move evidence into selected cellular and animal causal settings. A chain from a specific repair change to whole-whale lifespan, or to long-term human function, disease and safety, remains missing. More discriminating work would separate temperature and protein effects, make donor/allocation units explicit, test repair versus other RNA-mediated pathways, and measure both benefit and cost in whole animals. A favorable unmodified fraction or a longevity label cannot replace those experiments.

Download code, public inputs and all conditional results. Original FASTQs must be obtained separately from the public repository. The archive documents external provenance, assumptions, source discrepancies and replay instructions.