Lower germline mutation rates in young adults predict longer lives and longer reproductive lifespans

BACKGROUND:
Analysis of sequenced genomes from large three-generation families allows de novo mutations identified in Generation II individuals to be attributed to each of their parents9 germlines in Generation I. Because germline mutations increase with age, we hypothesized that they directly limit the duration of childbearing in women, and if correlated with mutation accumulation in somatic tissues, also reflect systemic aging in both sexes. Here we test whether the germline mutation rates of Generation I individuals when they were young adults predict their remaining survival, as well as the women9s reproductive lifespans.

METHODS:
Germline autosomal mutation counts in 122 Generation I individuals (61 women, 61 men) from 41 three-generation Utah CEPH families were converted to germline mutation rates by normalizing each subject9s number of mutations to the callable portion of their genome. Age at death, cause of death, all-site cancer incidence, and reproductive histories were provided by the Utah Population Database, Cancer Registry, and Utah Genetic Reference Project. Fertility analyses were restricted to the 53 women whose age at last birth (ALB) was at least 30 years, the approximate age when the decline in female fertility begins. Cox proportional hazard regression models were used to test the association of age-adjusted mutation rates (AAMRs) with aging-related outcomes. Linear regression analysis was used to estimate the age when adult germline mutation accumulation rates are established.

FINDINGS:
Quartiles of increasing AAMRs were associated with increasing all-cause mortality rates in both sexes combined (test for trend, p=0.009); subjects in the top quartile of AAMRs experienced more than twice the mortality of bottom quartile subjects (hazard ratio [HR], 2.07; 95% confidence interval [CI], 1.21-3.56; p=0.008; median survival difference = 4.7 years). Women with higher AAMRs had significantly fewer live births and a younger ALB. The analyses also indicate that adult germline mutation accumulation rates are established in adolescence, and that later menarche in women may delay mutation accumulation.

INTERPRETATION:
Parental-age-adjusted germline mutation rates in healthy young adults may provide a measure of both reproductive and systemic aging. Puberty may induce the establishment of adult mutation accumulation rates, just when DNA repair genes9 expression levels are known to begin their lifelong decline.


INTRODUCTION
The somatic mutation theory of aging 1 proposes that somatic mutations accumulate throughout life, resulting in apoptosis, cellular senescence, tumorigenesis, or other cellular pathologies, followed by tissue dysfunction, chronic disease, and death. DNA damage is continuous, 2 and while most of it is repaired, several classes of DNA damage are known to accumulate with age. [3][4][5] Serum levels of insulin-like growth factor I (IGF-I) peak during puberty, 6 suppressing the FOXO transcription factors and the DNA repair genes they upregulate. 7 DNA repair systems continue to decline throughout adult life. 8 Furthermore, developmental deficiency of the GH/IGF-1 axis in dwarf mice prevents the normal decline in DNA repair of adulthood and significantly extends lifespan. 9 These observations fit well with the evolutionary biology principle that the force of natural selection to maintain robust health begins to decline once the reproductive phase of life is attained. 10,11 Several monogenic segmental progeroid syndromes in humans are known to increase mutation accumulation rates and shorten the lifespan. 12 However, no studies have yet tested whether inter-individual differences in either somatic or germline mutation accumulation rates in healthy young adults predict differences in remaining lifespan, as would be expected if mutation accumulation contributes significantly to aging. Fertility in healthy women declines after age 30; 13 age at natural menopause varies from 40-60 years and is positively associated with longer lifespans; 14 and women with a high age at last birth and their siblings enjoy increased longevity. 15 Whether variation between healthy women in the decline and end of fertility may be attributable, at least in part, to differences in germline mutation accumulation rates is unknown. Similarly, while somatic mutations are known contributors to tumorigenesis, 16 a connection between mutation accumulation rates in healthy young adults and cancer risks has not yet been established.
De novo mutations (DNMs) in an individual's germline DNA can be found by genome sequencing of DNA extracted from somatic tissue samples (usually blood or saliva) collected from trios consisting of both parents and one of their offspring, identifying high-confidence sequence changes in the offspring not present in either parent, and finally attributing each mutation to the parental germline in which it arose [refs. 17,18 and Sasani et al. 2019, bioRxiv https://doi.org/10.1101/552117]. The number of parental germline DNMs increases with parental age, and increases more rapidly in the paternal germline than in the maternal germline. The higher mutation accumulation rate in men is expected, because there is a risk of mutation each time a cell divides, and cell divisions and spermatogenesis are ongoing in the male germline throughout adulthood, while oogenesis in the female is completed before she is born. For both sexes, at a given parental age the range of germline DNM counts can vary more than two-fold between individuals [ref. 18  Here, we used genome sequencing of 33 three-generation Utah families to identify mutations that were present in the germ cells of the grandparent generation when they were healthy young adults. We then tested whether the counts of autosomal DNMs are associated with two clinically important life history traits tracked in these same individuals over decades of follow-up: lifespan in both sexes and the duration of childbearing in women, as would be expected if germline DNM accumulation reflects rates of aging generally.

HUMAN SUBJECTS
The subjects were 122 grandparents (61 women and 61 men) from 33 of the 46 three-generation Utah families enrolled in the early 1980s for the purpose of building the first comprehensive human genetic linkage map. 19,20 All participants provided written informed consent, and studies were conducted with University of Utah Institutional Review Board approval. Each of these Utah CEPH (Centre d'Etude du Polymorphisme Humain) families consist of 5-15 siblings in the youngest generation, their two parents, and two to four grandparents. Differences between grandparents in their germline DNM rates are unlikely to be due to differences across the cohort in the presence or absence or degree of progression of various terminal illnesses, since all of the grandparents' offspring in whom the DNMs were identified are known to have reached maturity and subsequently conceived several children of their own.

OUTCOMES
Subjects were linked to the Utah Population Database (UPDB), a large and comprehensive resource of linked population-based information for demographic, genetic, and epidemiological studies (https://uofuhealth.utah.edu/huntsman/utah-population-database/acknowledging-updb.php). The UPDB is a dynamic genealogical and medical database that receives annual updates of Utah birth, death and cause of death, driver's license and health records. All research access to the UPDB requires the approval of the University of Utah's Resource for Genetic and Epidemiologic Research (RGE) and its Institutional Review Board (IRB). Mortality was ascertained based on Utah death certificates linked to UPDB. Causes of death were available in International Classification of Diseases (ICD) codes version 9-10 and aggregated into larger categories representing the leading causes of deaths. Cancer incidence records were from the Utah Cancer Registry (https://uofuhealth.utah.edu/utah-cancer-registry/). Fertility was assessed by parity (number of live births) and age at last birth, both derived from the UPDB. The University of Utah's RGE and IRB have approved this study.

STATISTICAL ANALYSIS
Demographic characteristics of male and female subjects were compared using t-tests for continuous variables and chi-square tests for categorical variables.
Sex-specific Cox proportional hazard models with adjustments for subject's birth year and subject's parental age, i.e. their age when the child in whom the DNMs were discovered (the index child) was born, were used to estimate the effects of germline DNM rates, treated as a continuous and categorical variable, on mortality and cancer risks, expressed as hazard ratios (HR). Results for both sexes combined were also adjusted for sex. Time was measured in years from parental age at the birth of the index child to time of death or last known living dates (2018). Cause-specific mortality was analyzed by fitting cause-specific hazard regression models with Cox regression, treating failures from the cause of death of interest as events and failure from other causes of deaths or those still living as right-censored observations. 21 For mortality and cancer incidence analyses, subjects were censored at age at death or age at last follow-up, whichever occurred first.
Poisson regression models were used to assess the effect of DNM rates on number of live births.
Logistic regression models were used to assess the association between DNM rates and age at last birth. All fertility analyses in women were adjusted for subject's birth year and subject's parental age and only included women with an age at last birth of at least 30 years, since cessation of childbearing prior to age 30 is unlikely due to reproductive aging. 13

CHARACTERISTICS OF THE SUBJECTS
All 122 subjects were from the grandparent generation of 33 three-generation Utah CEPH pedigrees. Each such pedigree consists of a large sibship with 5-15 children, their parents, and their living grandparents at the time of sample collection in the 1980s. For developing genetic linkage maps, the large sibship size allowed the segregation of genetic markers to be replicated in informative families, and the inclusion of grandparents helped to phase genetic loci being investigated. 19,20 The families were not selected for any disease, but for large sibship size. Selecting for large sibships may select somewhat for higher than average fertility, and selecting for living grandparents may select somewhat for higher than average lifespan; however, large sibships are common in Utah, and more than half of the grandparents were younger than age 72 at the time of the initial enrollment. Therefore, these families are unlikely to be strongly enriched for factors contributing to longer reproductive lifespans and longer life. Furthermore, since the same selection criteria were applied across all of the collected families, these criteria are not expected to introduce any biases that would affect the current study. Nearly all of the CEPH grandparents are of Northern European descent. Table 1 presents the demographic characteristics of the 61 women and 61 men who participated in this study.

SURVIVAL ANALYSES
The associations of DNM rates with all-cause mortality, cardiovascular disease (CVD) mortality and non-CVD mortality are presented in Table 2. All categorical comparisons used < 25 th percentile of DNM rates as the reference category. The analyses of both sexes combined, and of women only, found that germline DNM rates < 25 th percentile were associated with lower all-cause mortality than DNM rates ≥ 25th percentile; in men DNM rates < 25 th percentile were associated with lower all-cause mortality than DNM rates > 75th percentile. Tests for trend showed that for the full sample and for males, increasing DNM rates were significantly associated with greater mortality risk; for females this trend was not significant. Cardiovascular disease (CVD) mortality, which All rights reserved. No reuse allowed without permission.
author/funder, who has granted medRxiv a license to display the preprint in perpetuity.
The copyright holder for this preprint (which was not peer-reviewed) is the . https://doi.org/10.1101/19004184 doi: medRxiv preprint includes deaths from heart disease, stroke, and hypertension, was significantly increased only in men with DNM rates > 75 th percentile; this is supported by the test for trend (p=0.052). Non-CVD mortality increased significantly with increased DNM rates in both sexes combined (p=0.004), and in women analyzed independently (p=0.002). Non-CVD mortality is a category in which no single cause of death dominates. These results suggest a model whereby germline mutation rates correlate with somatic mutation rates across multiple tissues, with the accumulation of somatic mutations contributing to the development of a diverse set of agingrelated lethal diseases.
Adjusted survival curves by quartile of germline DNM rate showed that the median survival advantage for subjects with DNM rates < 25th percentile vs. those with DNM rates > 75th percentile was 8 years for women and 13 years for men ( Figure 1). Time was measured in years from parental age at the birth of the index child to time of death or last known living dates (2018).

CANCER INCIDENCE
In the set of 122 CEPH grandparents, there were 16 women and 18 men who received at least one cancer diagnosis in their lifetimes. We sought to test the hypothesis that in the full cohort of 122 subjects, lower sexand parental age-adjusted germline DNM rates would be associated with lower age-specific cancer risks (Supplementary Appendix, Table S1), as would be expected if germline mutation accumulation rates reflect somatic mutation accumulation rates. Associations were tested with germline DNM rates treated as a continuous variable and separately, as a categorical variable where the cancer risk in each higher quartile of DNM rates is compared to the cancer risk in the lowest quartile. No significant associations of DNM rates with cancer risk were found. Similar DNM and cancer incidence data from larger cohorts are needed to further investigate this hypothesis.

FERTILITY OF WOMEN
The fertility analysis was restricted to the 53 grandmothers whose age at last birth (ALB) was at least 30 years, since cessation of childbearing prior to age 30 is unlikely to be due to reproductive aging. 13 We hypothesized that germline DNM accumulation may contribute to oocyte atresia, lower rates of fertilization, higher rates of miscarriage, earlier menopause, and consequently shorter reproductive lifespans. Therefore, we tested whether women with higher germline DNM rates gave birth to fewer children, and also gave birth to their last child at younger ages, than women with lower germline DNM rates. The cut points for DNM rates (25 th All rights reserved. No reuse allowed without permission.
author/funder, who has granted medRxiv a license to display the preprint in perpetuity.
Among women with ALB ≥ 30 years, those in the top 50% of DNM rates had significantly fewer live births than those in the bottom 25% (p=0.026, Table 3), and higher DNM rates were significantly associated (p=0.036, Table 4) with a younger ALB (< 25th percentile).
We also tested the effect of germline DNM rates as a continuous or categorical variable on ALB as a continuous variable (Supplementary Appendix Table S2), and no significant association was found.

DISCUSSION
Here we show that lower sex-and parental age-adjusted germline DNM counts in young adults are associated with longer lifespans in both sexes, and higher parity and older age at last birth in women. Therefore, germline mutation accumulation rates in young adults may provide a measure, at least in part, of the rates of both reproductive and systemic aging. Corrections for multiple comparisons were not included in the analyses presented in Tables 2-4 for several reasons: 1) our comparisons are focused on a single prediction -that higher sex-and age-adjusted germline DNM counts are associated with the earlier occurrence of key milestones of aging (i.e. cessation of childbearing and death); 2) the comparisons are complementary, i.e. the same key prediction (more DNMs, earlier aging-related event) is being tested in several different ways -in both sexes, in each sex tested independently, in three categories of mortality, and in reproductive history -and all comparisons showing unadjusted significant differences support the prediction; and 3) the tests for trend that we performed across the categories of comparison also supported the key prediction, strongly suggesting that the low unadjusted p values we observed in several comparisons are unlikely due to chance alone. We encourage replications of our analyses in other populations, especially those with germline DNM counts and clinical follow-up data already in hand, to determine whether our findings and conclusions will be corroborated.
It is unlikely that germline mutation accumulation rates directly influence the rate of systemic aging, at least in men, since the lifespans of castrated and non-castrated men do not differ. 22 While mutation accumulation rates are lower in germline than in soma, 23 the effectors of DNA damage and the repair systems defending against it are likely similar across tissues. Therefore, ranking sex-and age-matched individuals by their germline mutation accumulation rates may effectively also rank them by their somatic mutation All rights reserved. No reuse allowed without permission.
author/funder, who has granted medRxiv a license to display the preprint in perpetuity.
The copyright holder for this preprint (which was not peer-reviewed) is the . https://doi.org/10.1101/19004184 doi: medRxiv preprint accumulation rates, in which case the associations we have identified may be interpreted as providing strong support for the somatic mutation theory of aging. Epidemiological studies of large cohorts with archived DNA samples and comprehensive clinical follow-up data, using sequencing technologies 4,5,24 that directly quantify somatic mutations in both nuclear and mitochondrial DNA, are needed to further investigate the role of somatically acquired mutations in the determination of the human healthspan and lifespan.
In considering possible causes for the variation in germline autosomal DNM counts between sex-and age-matched individuals, two likely contributors come to mind: differences in DNM accumulation rates and differences in the timing of puberty, if the establishment of adult mutation accumulation rates is triggered by puberty. Cross-sectional data for DNM accumulation in men (Sasani et al. Figure 2a This result is consistent with the mutation accumulation rates of adulthood becoming established around puberty, which is also known to have a five-year range for age of onset in both girls and boys. 25 These cross-sectional and longitudinal germline DNM accumulation data, taken together, support a model whereby there is a plateau of DNM counts prepubertally, followed by a resumption of mutation accumulation, at different rates in different people, that is triggered by puberty. Importantly, polygenic risk scores for later onset of puberty are associated with longer lifespans in both sexes; 26 and later puberty is All rights reserved. No reuse allowed without permission. author/funder, who has granted medRxiv a license to display the preprint in perpetuity. The copyright holder for this preprint (which was not peer-reviewed) is the . https://doi.org/10.1101/19004184 doi: medRxiv preprint associated with decreased all-cause mortality in women, 27,28 later menopause, 29 and reduced risk of cancer in both sexes; 30 as would be expected if puberty triggers the resumption of mutation accumulation, and consequently aging, in both germline and somatic tissues .
There are few studies to date of the dynamics of accumulation of somatic mutations in non-tumor tissues across the human lifespan. Surprisingly, at present most of the relevant available data come from analyzing DNA sequences from tumors, which include somatically acquired neutral (i.e. not tumor-promoting) mutations that were present in the cell of origin prior to tumorigenesis. By sequencing DNA from tumors that developed at various ages across the lifespan, it has been shown that somatic mutation accumulation begins in embryogenesis, pauses from approximately ages 5 to 14 years, and then doubles thereafter approximately every eight years. 31,32 Remarkably, intrinsic mortality rates are also plateaued from approximately age 5 -14 years, and double every eight years thereafter. 33 Thus somatic mutation accumulation after puberty may be a molecular correlate and major underlying cause of the increasing risk of dying with increasing age.
These observations fit well with the principle that the force of natural selection to evolve mechanisms that maintain robust health begins to decline once the reproductive phase of life is attained. 10,11 After puberty, both metabolic rate, which correlates with DNA damage rates, and DNA repair genes' expression levels decline with age 34,8 and their rates and relative levels of decline are likely to vary between individuals, due to both heritable genetic factors and differences in environmental exposures, including diet, exercise, and other lifestyle choices. This range of influences, many modifiable by personal choice, likely produces substantial inter-individual variation in somatic mutation accumulation rates and, therefore, rates of aging. However, significant variation in healthspan and lifespan may be expected even among individuals with identical puberty timing and mutation accumulation rates, if mutations are randomly distributed across the genome, only occasionally having pathogenic consequences.
While investigations of the causes of variation in the rate of aging in adult populations are likely to lead to novel therapies to postpone frailty and extend the human healthspan, further study of the effects of puberty on mutation accumulation rates may also lead to important medical breakthroughs. Rather than always rising with age, mortality rates are known to be plateaued in some contexts, e.g. in humans during prepubertal childhood 33 and at age 105 or older, 35 and in hydra 36,37 and asexual planaria 38 throughout life. These mortality plateaus may share gene expression profiles that robustly maintain the integrity of the genome (or at least All rights reserved. No reuse allowed without permission. author/funder, who has granted medRxiv a license to display the preprint in perpetuity. The copyright holder for this preprint (which was not peer-reviewed) is the . https://doi.org/10.1101/19004184 doi: medRxiv preprint prevent its further deterioration) and maintain other aspects of homeostasis as well, 3 effectively putting aging on hold.
The results presented here provide support for the hypothesis that mutation rates, at least in part, drive both reproductive and systemic aging. To our knowledge this is the youngest age range yet in which a molecular biomarker measured in healthy individuals has been found to predict remaining life expectancy.
Mutation accumulation with age may be the origin of or a contributor to one or more of the additional recognized hallmarks of aging. 3 Interventions in adults directed toward returning mutation accumulation rates to their very low prepubertal levels would be expected to have broad benefits, greatly lowering the risks for multiple aging-related diseases, and dramatically extending the human healthspan. Perhaps a relatively small number of genes that are master regulators of gene networks maintaining genome stability and homeostasis generally are down-regulated at puberty, but can be reprogrammed 39 or otherwise coaxed back to their prepubertal levels of activity by a combination of lifestyle, dietary, and/or pharmacological interventions.
All rights reserved. No reuse allowed without permission.
author/funder, who has granted medRxiv a license to display the preprint in perpetuity.
The copyright holder for this preprint (which was not peer-reviewed) is the . https://doi.org/10.1101/19004184 doi: medRxiv preprint author/funder, who has granted medRxiv a license to display the preprint in perpetuity.
The copyright holder for this preprint (which was not peer-reviewed) is the . https://doi.org/10.1101/19004184 doi: medRxiv preprint 0.025 Relative risks were estimated using Poisson regression to assess the effect of germline DNM rates as a continuous and categorical variable on number of live births of CEPH grandmothers, and additionally adjusted for the mother's own birth year and her parental age. All categorical comparisons used < 25 th percentile of DNM rates as the reference category. The number of live births decreased by 10.4% for each standard deviation increase in the maternal germline DNM rates (p = 0.104) with DNM rates treated as a continuous variable. Women with DNM rates >50 th percentile had significantly fewer live births than women with DNM rates <25 th percentile (p = 0.026).
The test for trend also shows a significant monotonic decrease in fertility as we move up the categories of DNM rates. 103 Logistic regression models were used to assess the effect of germline DNM rates as a continuous and categorical variable on the RR for ALB < 25 th percentile and additionally adjusted for the mother's birth year, her parental age, and whether she had any children with missing birth dates in the UPDB. All categorical comparisons used DNM rates < 25 th percentile as the reference category. For every standard deviation increase in the maternal germline DNM rate, the risk of having ALB in the < 25 th percentile category increased 2.45 times (95% CI 1.06-5.67, p = 0.036). With categorized DNM rates, the test for trend shows increasing DNM rate categories are associated with a greater likelihood of an ALB <25 th percentile.
All rights reserved. No reuse allowed without permission. author/funder, who has granted medRxiv a license to display the preprint in perpetuity.
The copyright holder for this preprint (which was not peer-reviewed) is the . https://doi.org/10.1101/19004184 doi: medRxiv preprint 213 Hazard ratios were estimated using Cox proportional hazard models to assess the effect of germline DNM rates as a continuous and categorical variable on cancer risk of CEPH grandparents and additionally adjusted for subject's birth year and subject's parental age. Time was measured in years from age when the index child was born (parental age) to time of death or last known living dates (2018) or first cancer diagnosis, whichever occurred first. For the categorical analyses, all comparisons used < 25 th percentile as the reference category. Cut points for DNM rates are the same as in Table 2. *HR (95% CI): the first row of each of the three sections (Both sexes, Females, and Males) presents the effect on cancer risk of a one standard deviation increase in germline DNM rates when DNM rates are treated as a continuous variable. For all other rows in the table DNM rates were treated as categorical variables. Multiple linear regression models were used to assess the effects of germline DNM rates as a continuous and categorical variable on ALB of CEPH grandmothers whose ALBs were ≥ 30 years. Effects were additionally adjusted for the mother's birth year, her age when the index child was born (parental age), and whether she had any children with missing birth dates in the UPDB. All categorical comparisons used DNM rates < 25 th percentile as the reference category. Cut points for DNM rates are the same as in Tables 3 and 4. The first row presents the effect on ALB of a one standard deviation increase in germline DNM rates when DNM rates are treated as a continuous variable. Age at last birth decreased by 12.3% (p = 0.167) for each standard deviation increase in the maternal germline DNM rate. For all other rows in the table DNM rates were treated as categorical variables. Figure S1. Germline Autosomal de novo Mutation Accumulation with Increasing Parental Age. The y intercepts of the linear regression lines, about 18 years for women, and 14 years for men, provide approximate lower bounds for the ages at which the observed mutation accumulation rates were established.

Supplementary Appendix
All rights reserved. No reuse allowed without permission.
author/funder, who has granted medRxiv a license to display the preprint in perpetuity.
The copyright holder for this preprint (which was not peer-reviewed) is the . https://doi.org/10.1101/19004184 doi: medRxiv preprint