Parental Age and Reproductive Procedures Leave Distinct Signatures on De Novo Mutations

A child’s genome is not a perfect copy of the genetic information inherited from the parents. Around the time of fertilization, “de novo mutations” not found in either parent occasionally arise. Most mutations have no apparent effect, but a small number may be involved in congenital disorders. As people increasingly delay parenthood and assisted reproduction becomes more common, a large family-sequencing study asked a further question: Do age and different reproductive procedures leave identifiable traces in the next generation’s genome?

The research team analyzed whole-genome data from 24,030 people in the Jiangsu Birth Cohort Study, including 7,851 parent–child families that passed quality control and 8,328 infants. They identified 390,924 high-quality de novo single-nucleotide variants and 28,032 small insertions or deletions. Each additional year of paternal age was associated with an average increase of about 1.06 paternally derived de novo mutations; each additional year of maternal age was associated with an increase of about 0.38 maternally derived mutations. The maternal change did not follow a simple linear increase. The model showed that the rate of increase accelerated after approximately the 29-to-32-year age range, but this was a statistical pattern in the study population and should not be treated as a clear clinical age threshold.

Assisted reproduction is also not a single exposure. After adjustment for parental age and infant sex, singleton infants conceived through assisted reproduction had an average of 2.38 more de novo single-nucleotide variants than naturally conceived infants. When the procedures were examined separately, intracytoplasmic sperm injection was associated with an increase in paternally derived mutations, while ovarian stimulation using a gonadotropin-releasing hormone antagonist was associated with an increase in maternally derived mutations and showed a dose trend in the study. The number of paternally derived mutations with conventional in vitro fertilization did not differ statistically from that in the natural-conception group, indicating that different technologies should not be broadly grouped as posing the same risk.

Procedures performed after embryo formation presented a different picture. Frozen embryo transfer and blastocyst-stage transfer were associated with increases in early postzygotic mosaic mutations; compared with natural conception, frozen blastocyst transfer was associated with an average of 1.23 more mutations of this type. Among them, the pattern of C-to-A variants resembled the signature of oxidative DNA damage, leading the researchers to speculate that longer in vitro culture or related procedures might contribute to their formation. However, similarity between mutation signatures can provide only mechanistic clues and cannot prove that the culture environment is the direct cause.

The study also linked genetic data with outcomes at birth and at the one-year follow-up. A higher number of paternally derived de novo single-nucleotide variants was associated with a slightly shorter gestational period, lower birth weight, and a higher probability of preterm birth, but explained only a small part of the differences in gestational duration associated with assisted reproduction or intracytoplasmic sperm injection. For every additional early mosaic C-to-A mutation, the relative likelihood of delayed cognitive development at age one increased by 14%; the total number of mosaic mutations itself did not show the same association, and a single short-term assessment is insufficient to infer long-term development.

These findings particularly should not be interpreted as showing that assisted reproduction inevitably harms children. The study was an observational analysis. Although it adjusted for multiple parental, clinical, and procedural factors, residual confounding cannot be excluded. The sample included only families with a live birth who completed the one-year follow-up, so it could not include miscarriages, stillbirths, or pregnancy terminations, and it might also have missed mosaic mutations present only in specific tissues. An average difference in mutation counts does not mean an increase in disease-causing mutations, and the findings cannot be used directly to compare reproductive options for an individual family.

Database records also provide additional details on the study’s scale and reproducibility: the data initially submitted by Nanjing Medical University covered 7,857 families and 8,334 infants, slightly more than the final figures after quality control in the paper. The variant files use the GRCh37 reference genome and are deposited under GVM001081, linked to PRJCA041582. Statistical source data and analysis code are available for review, but individual-level VCF files contain sensitive human genetic data and can be accessed only through a reviewed application. Further studies in other populations and reproductive centers, together with longer-term follow-up, are still needed to determine whether these subtle genomic differences have reproducible, clinically meaningful consequences.