Paternal age, mental illness, and the perils of statistics



Carpeaux: Ugolino and his sons. 1860-2

So I’ve been thinking about the to-ing and fro-ing concerning the role paternal mutations play in the development of mental disorders. It’s an interesting subject because firstly, it has relieved some of the vaguely sexist genetic blame on older women being responsible for contributing to the chromosomal disorder that results in Down Syndrome (and who knows what else, those selfish older women who insist on trying to have it all), and secondly because studying how mutations move through generations has really become easier with the truly exponential increase in whole genome sequencing facility. So there are  science-hype-publicity crap implications and genuine biological interest that relate to this subject. And I am always looking for new ways to torture my students with interesting genetic conundrums.

Biologically, understanding the role of age in parental mutation contribution depends on understanding how the germ cells, the gametes, eggs and sperm, are formed. In mammals like us, there is a big difference between how this process works in men vs. women. The cells that are destined to become the gametes are set aside in embryogenesis, and for men, spermatogenesis begins at puberty. The cell division that forms gametes is called meiosis, which consists of two stages – in Meiosis I the number of chromosomes is halved, and in Meiosis II, the chromosomes themselves divide to produce four cells, each with half the regular number of chromosomes. These meiotic products then undergo gametogenesis, or germ cell formation. For men, this process is continuous virtually throughout life.

oogen & sperm 02

For women, a limited number of eggs is produced during embryogenesis – a million or so. Those pre-egg cells then proceed to the first stage of Meiosis I, and then arrest. When a girl reaches puberty, Meiosis I is completed, resulting in two daughter cells of which only one survives (the other degrades).  Meiosis II is initiated, and of the two daughter cells from this second part of meiosis, only one cell survives again, and also arrests again unless the egg is fertilized. In men, for every meiotic pre-sperm cell, 4 mature sperm are formed, but for women, for every meiotic pre-egg cell, only one egg ever makes it.  And of course, after menopause, the ovaries the pack up. As for men, who knows when they begin to shoot blanks.

So think about what aging entails. It entails, basically, falling apart. A living organism is a complex biochemical machine, driven by thousands and thousands of interconnected chemical reactions, that slowly slide towards equilibrium, as all things must (in this universe, anyway). There are elaborate molecular machineries associated with the formation of germ cells. For women, errors in the machinery of cell division are most likely to have profound impacts on older conceptions (pregnancies arising in older women). Most commonly of course is Down Syndrome – which can arise by other means, but a failure in the machinery that properly segregates the chromosomes in completing those arrested meioses described earlier leads to the formation of eggs with sometimes missing, sometimes extra chromosomes (this is called aneuploidy). Most gametes that are aneuploid are targeted for destruction, or if fertilization happens, the embryo does not survive. Having extra or missing chromosomes results in an imbalance in the expression of genes that reside on those chromosomes, upsetting the delicate balance of biochemistry that makes a functioning organism. But sometimes the aneuploidy is tolerated, though not without consequence – Down Syndrome arises largely from the offspring receiving an extra chromosome 21.

For a long time this story of Down Syndrome was a warning for women to reproduce early, OR ELSE. Rarely, when I was growing up, did I hear about the problems associated with aging fathers. I’ve always found that rather annoying. But Haldane alluded to this problem in the 1930’s, long before the ideas of DNA and mutagenesis were realized. And yes, men have to worry about entropy in their machinery also. For men, the formation of sperm is continuous, meaning many, many, many millions if not billions of rounds of DNA replication to produce sperm throughout their lives. DNA replication is a remarkable and odd process that is probably one of the most difficult things I have to teach in introductory cell and molecular biology classes, with a fascinating evolutionary history. Our 46 chromosomes add up to over 6 billion base pairs of DNA, which have to be faithfully replicated every time a cell divides. A profound selective pressure for accuracy in this vast process has resulted in the evolution of a magnificent capacity for proof-reading and correcting errors during and after replication. This is the machinery that men have to worry about as they age, contributing to an increasing likelihood of errors accumulating the DNA of their continually produced sperm cells. (For a comprehensive review, see PMID: 11262873)

There have been a lot of studies about this recently. Look here and here for some nice summaries. These reviews refer to an Icelandic study from 2012 (PMID:22914163) and a Swedish study from 2014 (PMID:24577047). A very recent study (PMID: 27213288)  was published by an Australian group (summary here) which appears to contradict the claims made by the earlier studies, or at least, offer an alternative (and to my mind, rather unconvincing) hypothesis.

Full disclaimer: I am not a statistician, not an epidemiologist, not a population geneticist. I am a fruit fly geneticist. If it doesn’t happen in a fly, then I tend not to get it right away. And all these papers are seriously heavy on the statistical front. I found that my eyes ran and my nose bled for some of the more impenetrable parts. Like most statistical analyses and epidemiological surveys, the data tend to be heavy on correlation, and the impenetrable part comes from an attempt to use statistics to control for things that are virtually uncontrollable (socioeconomic stuff, I.Q., psychological propensities, meaning of life, folly of human existence etc). Of the lot, I really only felt I gained some understanding from the Icelandic study, which had a biological dimension that made some sense.

So in the 2012 Icelandic study, 78 parent-offspring groups had their genomes fully sequenced. What’s a genome? Very messy definitions abound: in my class, I define a genome as the information represented in a haploid set of chromosomes. As we are diploid, resulting from the fusion of haploid eggs and sperm, each of our (non egg or sperm) cells possesses two genomes (excepting some weird polyploid tissues – don’t worry about that). As humans, we divide our genomes into 23 chunks called chromosomes and as we get one set of 23 chromosomes from mum and the other from dad, each of those 23 chromosomes are arranged in so-called homologous (maternal and paternal) pairs. Algebraically, humans are 2n=46 (n=23 haploid state in the egg or sperm). Sequencing a full genome, (messy definition, see?) means presumably the diploid content. So the 2012 Icelandic study sequenced 219 complete diploid genomes, representing 78 families consisting of two parents and one or more offspring. 65 of the offspring were diagnosed with Autism Spectrum Disorder (ASD) or Schizophrenia. (More than 2000 additional Icelandic people were sequenced for statistical controls and artefact reduction purposes – all part of a large genome sequencing project in Iceland where the population has remained somewhat isolated and therefore very interesting to geneticists).

The researchers were looking for de novo or new mutations, changes in the DNA sequence that appeared in the offspring but were not present in either parent. So how do the researchers in this Icelandic study know if a mutation originated in the sperm, the egg, or the embryo? For most of these families, technically they don’t know. Previous sequencing studies likewise didn’t know either (example PMID: 22495306) but did notice a strong correlation between paternal age and the appearance of these de novo mutations (many of which occur in brain-expressed genes, considered candidate genetic contributors to ASD). The Icelandic paper however did include data from 5 families in which a third generation was sequenced, and this provided a stronger link to paternal age. The rationale is complex, and highly statistical, as it should be, and requires an understanding of something called a haplotype – which is essentially the DNA sequence context of a mutation. So, we are a very recently evolved species, and our genomes across the planet are very uniform, differing in perhaps 1 for every 1000 nucleotides in sequence. This means, that across a given stretch of DNA sequence, there is virtually no difference in DNA sequence between mum and dad, but the few differences there are provide a way of identifying the sequence context of a given mutation. So if a mutation occurs in the sperm, then it will appear in the fathers haplotype or sequence context, in the progeny. Likewise, if it occurred in the egg, then it will appear in the maternal context in the progeny. But that isn’t sufficient – how would you then distinguish a mutation that occurred in the embryo, that happened to occur on the paternal chromosome (and therefore in the paternal sequence context or haplotype)? Likewise for an embryonic mutation that occurred in the maternal context?

That’s were the third generation comes in. If mutation can occur randomly  in the maternal haplotype or paternal haplotype, then you would expect the offspring to transmit that mutation in the maternal or paternal haplotype with approximately equal frequency to the next generation. But if the mutation occurred more frequently in the sperm from the first generation, then the mutation would segregate (be passed down together) with that paternal context, or haplotype, more frequently than expected. And that is indeed what was observed in the five families in which three generations were sequenced. Here are the data – the five families in which three generations were sequenced so that the paternal vs maternal assignment of de novo mutations could be made:

denovo icelandic

Notice two things. Firstly, even in younger parents, the segregation of de novo mutations is predominantly from the paternal line. Secondly, notice how the ratio of paternal to maternal de novo mutations is dramatically larger in the single family with the oldest parents. This is a very small data set, but it provides some biological support for the correlation linking older fathers with the appearance of ASD/Schizophrenia in their offspring observed in the fully sequenced two generation family groups from this study.  This correlation between paternal age and mental illness in progeny is supported by both epidemiological studies (like PMID:24577047 – essentially analyses of medical records rather than anything molecular or biological), and other sequencing studies (like PMID: 22495306 – studies that did not perform the three generation analysis that permits parental assignment of de novo mutation).

So what are we to conclude? Do these findings support these kinds of headlines:

Father Time: Children with Older Dads at Greater Risk for Mental Illness

More Bad News for Older Dads: Higher Risk of Kids With Mental Illness

Too Old to Be a Dad?

Father’s Age is Linked to Autism and Schizophrenia

Devastating’ health risks older fathers may pass to children

Children with older fathers have lower IQs

….and so on. There probably isn’t any overt social engineering going on here (breed young, have nice nuclear families with 2.3 children and buy a dog called Spot etc) but do these studies, any of them, warrant sound-byte headlines designed to scare the crap out of new parents? All these studies show  (or at least the ones I have attempted to digest) is that the chances of a de novo mutation appearing in offspring increase with paternal age. Just like the chances of chromosomal segregation defects occuring increase with maternal age. But NONE of these studies, nor any others I have found, have identified the genetic basis for Autism Spectrum Disorder or Schizophrenia. (One possible exception, and not initiated by interest in paternal age and disease onset in offspring – read here: (PMID: 26814963)).  Indeed, one study I alluded to earlier from Australia (PMID: 27213288) makes the peculiar argument that de novo mutations in the male germline might not play a significant role at all – rather that men with mental illness of necessity tend to delay fatherhood and then pass on their mental illness genes. (But most of the older fathers I know are simply older because they married a second, or third time, and were wealthy enough to have a second round of kids – and I imagine this is supremely common whereas mental illness is not).

From what I can tell, the more interesting questions relate to  how the machinery of cell division and DNA replication age, and what the genetic basis for complex mental disorders might be. Like as not, the genetic basis will be highly complex – mutations in multiple genes yielding an inherited propensity at best. And don’t forget about the epigenetic dimension that inevitably plays a role in the phenotypic interpretation of genotypes – none of which is amenable to raw DNA sequence analysis. Dissecting the polygenic basis of complex traits is both frustrating and fascinating, and not reducible to socially provocative headlines. Dads, try not to worry. Welcome to Mums’ world.