Schizophrenia Research Forum - A Catalyst for Creative Thinking

Ambitious Genetic Integration Analysis of Schizophrenia Points to Early Brain Development

August 2, 2013. A new crop of rare genetic mutations in people with schizophrenia implicates genes involved in the early stages of brain development, reports a study published August 1 in Cell. Using exome sequencing, Mary-Claire King of the University of Washington in Seattle and colleagues identified spontaneously arising, or “de novo,” mutations likely to damage protein function in 54 genes in the schizophrenia group. Network analyses of the transcription patterns of these genes in healthy brains, and of the interactions between their protein products, suggest that these diverse genes work together in the prenatal stages of building a brain, particularly in dorsolateral and ventrolateral prefrontal cortex.

The findings support the long-standing notion that schizophrenia’s roots lie in early brain development and add to evidence of the vulnerability of prefrontal cortex, a region important for complex mental processes, in the disorder.

“These investigators are trying to take the next step by asking if there is anything that ties their genetic findings together,” said Daniel Geschwind of the University of California in Los Angeles, who was not involved in the study. “Given that they start with only 54 genes, it’s kind of remarkable there’s any coherence.”

Network analyses of all human genes have detected groups of genes that seem to work together in the brain. For example, Geschwind and colleagues have tracked transcription patterns in the brain to find genes that turn on and off together, with the idea that those that co-vary are likely engaged in the same function. With this approach, Geschwind’s team has identified groups of genes that work together, called modules, that are specific to humans (Oldham et al., 2006) and perturbed in autism (see SRF related news story). Last year, a similar approach by Vahram Haroutunian's group at New York’s Mount Sinai School of Medicine identified anomalous gene modules in schizophrenia (see SRF related news story). Another network analysis published last year by Dennis Vitkup and colleagues of Columbia University in New York grouped genes based on similarities in their likely phenotypic outcomes and found these groupings were enriched for schizophrenia suspects (see SRF related news story).

The new study combines transcriptional data with protein-protein interaction data to build a final network based only on genes fingered by the exome sequencing. As researchers use more and different kinds of data to decipher the functions of their genes, debate will ensue about how best to integrate the different kinds of information. “It would be nice to integrate the protein networks more formally with the transcriptional networks,” Geschwind said, noting that the protein database was relatively permissive in judging two proteins as interacting and not specific to conditions within neurons.

Round up the suspects
First authors Suleyman Gulsuner, Tom Walsh, and Amanda Watts and colleagues sequenced the exomes of 105 people with schizophrenia, 84 of their unaffected siblings, and 210 of their unaffected parents. The researchers were looking for de novo variants—those mutations that spontaneously occur in the sperm or egg cell prior to conception, rather than being inherited from parents. De novo mutations are prime suspects because they are more likely to be nasty, as natural selection hasn’t had a chance to weed them out. Though previous studies have found de novo mutations in schizophrenia (Xu et al., 2011, and see SRF related news story), they don’t necessarily constitute smoking guns because the background levels of de novo mutations are unclear. The unaffected siblings sequenced in the new study provide hard data on this.

The researchers report finding 103 de novo mutations in people with schizophrenia—meaning that these mutations were not found in a person’s parents or siblings. In comparison, 67 were found in unaffected siblings. The mutations consisted mainly of point mutations, in which a single base change is made in the DNA sequence, but included a few small insertions or deletions of DNA. Among the point mutations, slightly more were found in schizophrenia than in controls. Mutations likely to be damaging were also enriched in the schizophrenia group, which had 57 compared to 35 in sibling controls (that’s an average of 0.54 damaging mutations per person with schizophrenia compared to 0.42 per sibling controls). The researchers also noted that chances of carrying a de novo mutation increased with the father’s age at conception, consistent with previous studies (see SRF related news story and SRF Current Hypothesis). When all was counted and sorted, the researchers had 54 different genes that had been hit by potentially damaging mutations in schizophrenia.

Map them out
Many genetic studies would stop there, rounding out their discussion with a laundry list of speculations about how each gene might relate to a disorder. But the new study goes further, using network analyses to try to uncover what the diverse lot might have in common. To find evidence of genes working together, the researchers used transcription patterns available in the BrainSpan Atlas that were obtained by RNA sequencing of postmortem brain tissue from healthy subjects. The researchers asked whether transcription of their 54 genes co-varied more than those genes fingered by de novo mutations in unaffected siblings. The researchers calculated the correlation between the transcript levels of each pair of genes across different brain samples and judged those with correlation coefficients of 0.8 or above as “connected.” The resulting network of connections was compared to a control network, which was also built from pairwise comparisons of 54 different genes randomly selected from a pool of 264 genes hit by de novo mutations in the unaffected siblings from this and other studies (see SRF related news story).

The transcription data came from different brain regions at different stages of development, ranging from prenatal to adult tissue. Thus, the resulting networks were region and time specific. The researchers found that the schizophrenia network based on data from frontal cortex in fetal tissue had more connections than the control network, suggesting that these genes worked together on something there during early stages of development. Further analysis narrowed this to dorsolateral and ventrolateral prefrontal cortex—regions important for cognitive control, working memory, reward learning, and decision making. In contrast, genes hit by non-damaging variants in schizophrenia resulted in a network that was not different from the control networks.

Another way to gauge whether genes are working together is to see if their protein products physically interact. Using the GeneMANIA database of protein-protein interactions, the researchers found that the genes hit by de novo mutations in schizophrenia had a significantly greater degree of interaction than those found in healthy siblings. In contrast, for the genes hit by benign de novo mutations there was no difference between the schizophrenia gene and control networks.

Add them up
The researchers then combined this protein-based network with the transcription network for the fetal dorsolateral and ventrolateral prefrontal cortex, designating genes as connected by protein-protein interaction data and/or transcription data. This yielded a 50-gene network with 126 connections. These genes were associated with processes like neural proliferation, migration, cell signaling, and axon guidance—all crucial to laying down a brain’s foundations. The researchers noted that some of these genes influenced neurotransmitter signaling (ADCY9, SLC18A2, and GLS) and also highlighted CACNA1I, a gene encoding a subunit of a type of calcium channel, which was hit multiple times in their schizophrenia samples.

The findings support the neurodevelopmental hypothesis of schizophrenia and provide some causal evidence for these genes’ involvement in the disorder. Although opinions may differ on which is more persuasive, this study heralds the integrative nature of research to come.—Michele Solis.

Gulsuner S, Walsh T, Watts AC, Lee MK, Thornton AM, Casadei S, Rippey C, Shahin H, Nimgaonkar VL, Go RC, Savage RM, Swerdlow NR, Gur RE, Braff DL, King MC, McClellan JM. Spatial and temporal mapping of de novo mutations in schizophrenia to a fetal prefrontal cortical network. Cell . 2013 Aug 1 ; 154(3):518-29. Abstract

Comments on News and Primary Papers
Comment by:  Roger Boshes
Submitted 10 August 2013
Posted 20 August 2013

These data suggest a "stem" circuit that may be common to many patients with schizophrenia, but subsequent de novo mutations may explain the protean manifestations of the disorder. Alternatively, this prefrontal perturbation may be related to a heritable, i.e., not a somatic, mutation that explains 80 percent heritability but not the protean phenotypic expression of the condition. Finally, it may be the link between schizophrenia and some flavors of autism.


Boshes RA, Manschreck TC, Konigsberg W. Genetics of the schizophrenias: a model accounting for their persistence and myriad phenotypes. Harv Rev Psychiatry. 2012 May-Jun; 20(3):119-29. Abstract

View all comments by Roger Boshes

Comments on Related News

Related News: Autism Exome: Lessons for Schizophrenia?

Comment by:  Patrick Sullivan, SRF Advisor
Submitted 20 April 2012
Posted 23 April 2012
  I recommend the Primary Papers

Fascinating papers that likely presage work in the pipeline from multiple groups for schizophrenia. Truly groundbreaking work by some of the best groups in the business. Required reading for those interested in psychiatric genomics.

The identified loci provide important new windows into the neurobiology of ASD.

The results also pertain to the longstanding debate about the nature of ASD: does it result from many individually rare, Mendelian-like variants (potentially a different one in each person) and/or from the summation of the effects of many different common variants of subtle effects?

The multiple rare variant model now seems unlikely for ASD as, contrary to the expectations of some, ASD did not readily resolve into a handful of Mendelian-like diseases. (This comment is of course qualified by the limits of the technologies - which have, however, identified causal mutations for many monogenetic disorders.)

Readers might also want to read Ben Neale's comments on these papers at the Genomes Unzipped website.

View all comments by Patrick Sullivan

Related News: New Mutations Mount as Fathers Age

Comment by:  Dolores Malaspina
Submitted 27 August 2012
Posted 27 August 2012

The new report by Kong et al. (2012) demonstrates that paternal age is likely to be an important source of mutations that are relevant for schizophrenia, as we earlier hypothesized (Malaspina, 2001). Kong et al. demonstrated that the diversity in human mutation rates for offspring is dominated by the paternal age at conception. Following our initial observation that advancing paternal age was substantially associated with an increasing risk for schizophrenia, explaining a quarter of the population's attributable risk for schizophrenia (Malaspina et al., 2001), many scientists found it difficult to accept that the father’s age could be a risk pathway for schizophrenia. By contrast, the hypothesis that paternal age explained the risk for achondroplastic dwarfism achieved far greater immediate acceptance over 20 years ago (i.e., Thompson et al., 1986). While these new findings will surely advance our understanding of many de novo neuropsychiatric conditions, they also substantiate biological versus psychosocial causation theories for severe neuropsychiatric conditions.


Malaspina D. Paternal factors and schizophrenia risk: de novo mutations and imprinting. Schizophr Bull . 2001 ; 27(3):379-93. Abstract

Malaspina D, Harlap S, Fennig S, Heiman D, Nahon D, Feldman D, Susser ES. Advancing paternal age and the risk of schizophrenia. Arch Gen Psychiatry . 2001 Apr ; 58(4):361-7. Abstract

Thompson JN Jr, Schaefer GB, Conley MC, Mascie-Taylor CG. Achondroplasia and parental age. N Engl J Med. 1986 Feb 20;314(8):521-2. Abstract

View all comments by Dolores Malaspina

Related News: New Mutations Mount as Fathers Age

Comment by:  Patrick Sullivan, SRF Advisor
Submitted 27 August 2012
Posted 27 August 2012

Kong et al. sequenced 78 pedigree clusters (mostly parent-offspring trios) to around 30x coverage. After careful quality control, they identified an average of 63 new mutations per trio. These mutations were “de novo” in that they were absent in the parents but present in an offspring and assumed to have occurred during gametogenesis.

Intriguingly, more of these mutations occurred in older parents. The authors present several lines of evidence to implicate fathers rather than mothers, and estimated that there were about two extra de novo mutations per year of increase in paternal age. This conclusion is consistent with several of the exome sequencing papers published in Nature a few months ago.

Increased paternal age is an epidemiological risk factor for schizophrenia and autism, with relative risks on the order of two and five, respectively. This paper suggests a potential mechanism for the paternal age effect that might eventually prove to be relevant for some fraction of cases.

It is important to note that advanced paternal age is a risk factor, not a determining feature. Risk is increased, but not in a deterministic manner.

View all comments by Patrick Sullivan

Related News: A Bird’s Eye View of the Schizophrenia Transcriptome

Comment by:  Karoly Mirnics, SRF Advisor
Submitted 28 August 2012
Posted 28 August 2012

This is perhaps the best and most comprehensive transcriptome dataset of schizophrenia generated to date. It has multiple strengths, including the use of four different cortical regions, the correlation of genetics with transcriptomics, and the use of strong bioinformatics including WGCNA analysis. The findings are quite revealing. The results suggest that there is a strong, common signature across brain areas BA21, BA32, and BA38 that encompasses genes related to transcription/translation, signal transduction, the cell cycle, cell adhesion, the immune response, apoptosis, and the cytoskeleton.

Perhaps surprisingly, the expression signature was far less prominent in prefrontal cortical area BA 46, which is one of the most affected regions in schizophrenia. However, it was also clear that each brain region had a unique, region-specific schizophrenia signature. In addition, this study reproduced and validated a number of previously reported findings related to oligodendrocyte and mitochondrial transcript deficits. Nevertheless, the results of the current study disagree with the previously reported and replicated outcomes of similar assessments of other cohorts: in this study, GABA system genes were upregulated, and gene ontology categories related to immune response were downregulated in subjects with schizophrenia.

This is certainly noteworthy, and this apparent discrepancy in findings will have to be addressed by future experiments. I wish that the authors had addressed this issue in their discussion. The combined transcriptomics-genetics results suggest that the oligodendrocyte, GABA, and glutamate modules are (at least partially) driven by genetic vulnerability, while other gene expression changes might be secondary/adaptational in nature.

Finally, the study suggests that interregional coexpression is attenuated in schizophrenia. A very similar hypothesis, using WGCNA analysis of samples with autism, has been proposed in autism by the Geschwind laboratory (see Voineagu et al., 2011, and the commentary of Korade and Mirnics, 2011). Voineagu et al. reported that the differential patterns of gene expression that normally distinguish the frontal and temporal cortices are significantly attenuated in the autistic brain, potentially leading to loss of functional specifications across the affected cortical areas. This is certainly worth further exploration, and the schematic hypothesis presented in Figure 5 of the current paper is a logical blueprint that nicely maps out a possible sequence of pathophysiological events in schizophrenia. Still, the actual sequence of the proposed events can be debated at the current time: it is very likely that the cascades proposed by Figure 1 in the commentary by Korade et al. and Figure 5 in the current manuscript will have to be revised as our knowledge accumulates.


Voineagu I, Wang X, Johnston P, Lowe JK, Tian Y, Horvath S, Mill J, Cantor RM, Blencowe BJ, Geschwind DH. Transcriptomic analysis of autistic brain reveals convergent molecular pathology. Nature . 2011 May 25. Abstract

Korade Z, Mirnics K. Gene expression: the autism disconnect. Nature . 2011 June 15. Abstract

View all comments by Karoly Mirnics

Related News: New Mutations Mount as Fathers Age

Comment by:  John McGrath, SRF Advisor
Submitted 28 August 2012
Posted 28 August 2012
  I recommend the Primary Papers

In 2001, Dolores Malaspina alerted the research community to the link between advanced paternal age and increased risk of schizophrenia—she suggested that this may be due to de novo mutations in the male germ line (Malaspina et al., 2001). The study BY Kong et al. provides compelling evidence in support of this hypothesis (Kong et al., 2012). A related paper in Nature Genetics also demonstrates an association between paternal age and changes in microsatellite properties across generations (Sun et al., 2012).

While the hypothesis that de novo mutations accumulate due to copy error mutations in the production of germ cells in older males is compelling, it is still possible (albeit unlikely) that this association may be due to unmeasured confounding. For example, older men might be exposed to more environmental toxins that accumulate over time and subsequently cause mutations in the offspring of older dads as a byproduct of the greater exposure. There is also the evidence from Denmark indicating that, when adjusted for age of first child, the association between paternal age and risk of schizophrenia fades out (Petersen et al., 2011). This finding suggests that selective factors may also operate (e.g., perhaps related to personality of schizotypal men, etc.).

However, animal experiments can provide useful clues to this puzzle (Foldi et al., 2011). Mouse models of advanced paternal age indicate that the offspring of older sires differ from control animals on behavior and brain structure (Smith et al., 2009; Foldi et al., 2010). Of particular relevance for the study by Kong et al., a mouse experiment found that the offspring of older sires were significantly more likely to have de novo copy number variants (Flatscher-Bader et al., 2011).

We now have convergent evidence from risk factor epidemiology, animal experiments, and genetic studies. The evidence supports an increased risk of schizophrenia in the offspring of older fathers, and points to age-related mutagenesis in the male germ cell. It is still not clear why these age-related events seem to differentially impact on neurodevelopmental disorders (e.g., autism is also linked to paternal age). Perhaps neocortical development is less well "buffered" (compared to more phylogenetically ancient organs); thus, de novo mutations can more readily "decanalize" certain features of brain development (McGrath et al., 2011). From an evolutionary developmental biology perspective (evo-devo), the dictum goes “Last in, first to break.”

It is rare that different fields of research converge in such an obedient fashion. It is time that we pause and reflect on this important milestone—and also offer a rousing “three cheers for Dolores Malapsina!”


Flatscher-Bader T, Foldi CJ, Chong S, Whitelaw E, Moser RJ, Burne TH, Eyles DW, McGrath JJ. Increased de novo copy number variants in the offspring of older males. Transl Psychiatry. 2011 Aug 30;1:e34. Abstract

Foldi CJ, Eyles DW, McGrath JJ, Burne TH. Advanced paternal age is associated with alterations in discrete behavioural domains and cortical neuroanatomy of C57BL/6J mice. Eur J Neurosci. 2010 Feb;31(3):556-64. Abstract Foldi CJ, Eyles DW, Flatscher-Bader T, McGrath JJ, Burne TH. New perspectives on rodent models of advanced paternal age: relevance to autism. Front Behav Neurosci . 2011 ; 5():32. Abstract

Kong A, Frigge ML, Masson G, Besenbacher S, Sulem P, Magnusson G, Gudjonsson SA, Sigurdsson A, Jonasdottir A, Jonasdottir A, Wong WS, Sigurdsson G, Walters GB, Steinberg S, Helgason H, Thorleifsson G, Gudbjartsson DF, Helgason A, Magnusson OT, Thorsteinsdottir U, Stefansson K. Rate of de novo mutations and the importance of father's age to disease risk. Nature. 2012 Aug 23;488(7412):471-5. Abstract

Malaspina D, Harlap S, Fennig S, Heiman D, Nahon D, Feldman D, Susser ES. Advancing paternal age and the risk of schizophrenia. Arch Gen Psychiatry. 2001 Apr ; 58(4):361-7. Abstract

McGrath JJ, Hannan AJ, Gibson G. Decanalization, brain development and risk of schizophrenia. Transl Psychiatry. Abstract

Petersen L, Mortensen PB, Pedersen CB. Paternal age at birth of first child and risk of schizophrenia. Am J Psychiatry. 2011 Jan;168(1):82-8. Abstract

Smith RG, Kember RL, Mill J, Fernandes C, Schalkwyk LC, Buxbaum JD, Reichenberg A. Advancing paternal age is associated with deficits in social and exploratory behaviors in the offspring: a mouse model. PLoS One. 2009 Dec 30;4(12):e8456. Abstract

Sun JX, Helgason A, Masson G, Ebenesersdóttir SS, Li H, Mallick S, Gnerre S, Patterson N, Kong A, Reich D, Stefansson K. A direct characterization of human mutation based on microsatellites. Nat Genet. 2012 Aug 23. Abstract

View all comments by John McGrath

Related News: New Mutations Mount as Fathers Age

Comment by:  Georg Winterer (Disclosure)
Submitted 28 August 2012
Posted 28 August 2012
  I recommend the Primary Papers

Just a few thoughts:

One question is whether it is just age per se that produces de novo mutations or an accumulation of environmental effects like drug abuse, alcohol, or other potentially harmful toxic environments, etc. What I also would like to know is whether it is the number of sperm cycles; in that case, men who are sexually more active should have a greater risk to produce more de novo mutations.

View all comments by Georg Winterer

Related News: New Mutations Mount as Fathers Age

Comment by:  Michael O'Donovan, SRF AdvisorGeorge Kirov
Submitted 31 August 2012
Posted 31 August 2012

In a genomic sequencing study of 78 parent-proband trios (21 probands with schizophrenia, 44 with autism spectrum disorder [ASD]), Kong and colleagues (2012) identify almost 5,000 DNA single base changes that occurred as a result of new mutations. For five of the trios, the proband had a child who was also sequenced, and in this subset with three generations of data, Kong and colleagues were able to determine if the mutations had arisen on the paternal or maternal chromosomes. Although this subsample was small, paternal chromosomes showed much greater variance in the number of mutations than maternal chromosomes, suggesting that paternal variables are more relevant to variance in the overall de novo mutation rate than maternal variables. In the larger sample as a whole, although the parental origin of the mutations could not be determined, the number of new mutations carried by an individual could be almost completely explained by a combination of random variation and paternal age. Models of linear and of exponential increases in the number of mutations by paternal age both described the data well, the ability to distinguish between the two being constrained by a lack of fathers at the higher age. Children of fathers aged 40 had approximately twice the number of mutations as those aged 20. After accounting for random variation and paternal age, in this sample, there was very little residual variation to be explained by other factors, including maternal age and within-population environmental exposures. A possible impact of cross-population environmental exposures was not addressed, since all the subjects came from Iceland.

Overall, the findings from what is yet another impressive paper from the deCODE group support the proposition that paternal age is an important factor in determining the probability that a child might inherit a new mutation (see Goriely and Wilkie, 2012, for a wider discussion of earlier data on paternal age and mutation rates, particularly in sperm) and additionally quantify this effect in the context of other possible unexplained variables.

This is clearly an important paper for understanding factors dictating the rate by which new mutations occur, and is therefore a paper that will have wide relevance to diseases to which such mutations make a substantial contribution. But from the perspective of most readers of this Forum, it is more important to note what the study is not about.

There is good evidence that risk of schizophrenia increases with paternal age (Malaspina et al., 2001; Zammit et al., 2003; Frans et al., 2011). This is certainly compatible with the involvement of new mutations of the sort described in the paper by Kong and colleagues, but there are several alternative explanations. For example, fathers with high trait liability for schizophrenia might have subclinical characteristics making them less effective at reproduction (e.g., they may find it more difficult to find a partner) and, as a result, elderly fathers might be enriched for transmissible schizophrenia alleles. Consistent with this (and other explanations not dependent on new mutations), one large Danish study found that the paternal age effect was best explained by age at which fathers first reproduce, not the age (which is more relevant to new mutations) when the affected offspring was conceived (Petersen et al., 2011). Of general importance as it is, the study by Kong and colleagues makes no contribution to resolving to what extent the paternal age effect observed in schizophrenia (and autism) is explained by new mutations, or indeed to what extent new mutations are involved in these disorders at all. Indeed, as the authors point out, the fact that they have studied probands, the majority of whom are affected by schizophrenia or ASD, is an irrelevance; essentially identical findings would be expected if they had studied other types of families. This is because the average proband carries over 60 de novo mutations, of which, even under an extreme model in which all schizophrenia is caused by de novo mutations, at most, one or two (if any) might be schizophrenia or ASD relevant. Consequently, de novo mutations related to the phenotype of the proband cannot substantially contribute to the overall pattern of results.

Overall, this study provides empirical evidence for a mechanism by which some of the paternal age effects might be explained by de novo point mutations, but it is worth stressing that the fact that the authors have studied schizophrenia and ASD is incidental, and this study does not address the extent by which, if at all, mutations of this type make any contribution to schizophrenia (or autism). Finally, since the results of this paper have been widely reported (at least in the UK), we think it is important to note for the general reader that, while the paternal age effect of risk of schizophrenia (and autism) seems to be real, the vast majority of people with schizophrenia are not born to elderly fathers. More importantly, since the causal direction of the paternal age effect on schizophrenia risk is unknown, there is currently no strong reason to urge potential fathers to consider earlier reproduction as a strategy for reducing risk of this particular disorder.


Zammit S, Allebeck P, Dalman C, Lundberg I, Hemmingson T, Owen MJ, Lewis G. Paternal age and risk for schizophrenia. Br J Psychiatry. 2003 Nov;183:405-8. Abstract

Malaspina D, Harlap S, Fennig S, Heiman D, Nahon D, Feldman D, Susser ES. Advancing paternal age and the risk of schizophrenia. Arch Gen Psychiatry. 2001 Apr;58(4):361-7. Abstract

Goriely A, Wilkie AO. Paternal age effect mutations and selfish spermatogonial selection: causes and consequences for human disease. Am J Hum Genet. 2012 Feb 10;90(2):175-200. Review. Abstract

Frans EM, McGrath JJ, Sandin S, Lichtenstein P, Reichenberg A, Långström N, Hultman CM. Advanced paternal and grandpaternal age and schizophrenia: a three-generation perspective. Schizophr Res. 2011 Dec;133(1-3):120-4. Epub 2011 Oct 14. Abstract

Petersen L, Mortensen PB, Pedersen CB. Paternal age at birth of first child and risk of schizophrenia. Am J Psychiatry. 2011 Jan;168(1):82-8. Epub 2010 Oct 15. Abstract

View all comments by Michael O'Donovan
View all comments by George Kirov

Related News: New Mutations Mount as Fathers Age

Comment by:  Bernard Crespi
Submitted 3 September 2012
Posted 5 September 2012
  I recommend the Primary Papers

Kong et al. (2012) is an outstanding paper that provides the first detailed quantification of how human de novo mutations in sperm and eggs vary with parental age. The paper and its aftermath provide a number of important lessons for researchers studying neurodevelopmental disorders and parental age:

1. The work demonstrates directly that CpG dinucleotides contribute the lion's share of new mutations. CpG sites are of particular interest in understanding effects of de novo mutations because they differentially create new transcription factor binding sites (Zemojtel et al., 2011), as well as mediate the effects of methylation and genomic imprinting. Such findings might help to focus efforts at interpreting the functional importance of the myriad de novo variants that pepper each genome.

2. The work generates an apparent paradox: if, as the authors claim, paternal age so strongly predominates over maternal age in its de novo mutational effects, why do so many parental-age studies of autism and schizophrenia show clear effects of maternal age as well (e.g., Lopez-Castroman et al., 2010; Parner et al., 2012; Rahbar et al., 2012; Sandin et al., 2012)? Might maternal-age effects be mediated by different processes?

3. The X chromosome was not included in the analysis, despite its expected contribution to de novo mutational effects being much stronger than for autosomes, due to its hemizygosity (as found, e.g., in intellectual disability). A recent study also strikingly implicates the X chromosome in psychosis risk, perhaps involving epigenetic mechanisms (Goldstein et al., 2011).

4. It is important to avoid neurodevelopmental tunnel vision with regard to parental age effects. Advanced maternal age, for example, has been documented as a risk factor for a suite of other conditions, including hypertension, diabetes, cancer, and Alzheimer's (for a review, see Myrskylä and Fenelon, 2012), as expected if it exerts effects on all polygenic conditions.

5. As anyone following popular media accounts will have noticed, the paper has been fundamentally misinterpreted in translation from the scientific to popular literature. Contrary to almost all reports in the popular press (including, e.g., The New York Times), the paper clearly does not show that higher paternal age is associated with mutations that increase the risk of autism or schizophrenia. As noted by other commentators, to do so would require that the authors link paternal age with the number of new mutations that are actually known to contribute to autism or schizophrenia. This muddle should caution authors to be as clear in explaining what their findings do not show as they are in explaining what they actually demonstrate. If subsequent work shows that age-dependent point mutations themselves do not mediate increased autism or schizophrenia risk, scientific credibility will unjustifiably suffer.

6. Finally, the press has jumped on advanced parental age as an important possible factor in the increased diagnoses of autism over the past 30 or so years. But if increased mutation load has increased rates of autism, why haven't rates of schizophrenia increased in lockstep, albeit with a 20-year delay?

Parental age has been suspected as an important factor in genetically based, de novo conditions since Weinberg (of Hardy-Weinberg fame) noticed in 1912 that children with achondroplasia (a form of dwarfism) were later-born in sibships. One hundred years later, we are one large step closer to understanding why. Let us help to ensure that this step is free of de novo errors of interpretation and implication, and move forward with speed.


Goldstein JM, Cherkerzian S, Seidman LJ, Petryshen TL, Fitzmaurice G, Tsuang MT, Buka SL. Sex-specific rates of transmission of psychosis in the New England high-risk family study. Schizophr Res. 2011 May;128(1-3):150-5. Abstract

Kong A, Frigge ML, Masson G, Besenbacher S, Sulem P, Magnusson G, Gudjonsson SA, Sigurdsson A, Jonasdottir A, Jonasdottir A, Wong WS, Sigurdsson G, Walters GB, Steinberg S, Helgason H, Thorleifsson G, Gudbjartsson DF, Helgason A, Magnusson OT, Thorsteinsdottir U, Stefansson K. Rate of de novo mutations and the importance of father's age to disease risk. Nature. 2012 Aug 22; 488: 471-5. Abstract

Lopez-Castroman J, Gómez DD, Belloso JJ, Fernandez-Navarro P, Perez-Rodriguez MM, Villamor IB, Navarrete FF, Ginestar CM, Currier D, Torres MR, Navio-Acosta M, Saiz-Ruiz J, Jimenez-Arriero MA, Baca-Garcia E. Differences in maternal and paternal age between schizophrenia and other psychiatric disorders. Schizophr Res. 2010 Feb;116(2-3):184-90. Abstract

Myrskylä M, Fenelon A. Maternal Age and Offspring Adult Health: Evidence From the Health and Retirement Study. Demography . 2012 Aug 28. Abstract

Parner ET, Baron-Cohen S, Lauritsen MB, Jørgensen M, Schieve LA, Yeargin-Allsopp M, Obel C. Parental age and autism spectrum disorders. Ann Epidemiol. 2012 Mar;22(3):143-50. Abstract

Rahbar MH, Samms-Vaughan M, Loveland KA, Pearson DA, Bressler J, Chen Z, Ardjomand-Hessabi M, Shakespeare-Pellington S, Grove ML, Beecher C, Bloom K, Boerwinkle E. Maternal and Paternal Age are Jointly Associated with Childhood Autism in Jamaica. J Autism Dev Disord. 2012 Sep;42(9):1928-38. Abstract

Sandin S, Hultman CM, Kolevzon A, Gross R, MacCabe JH, Reichenberg A. Advancing maternal age is associated with increasing risk for autism: a review and meta-analysis. J Am Acad Child Adolesc Psychiatry. 2012 May;51(5):477-486.e1. Abstract

Zemojtel T, Kielbasa SM, Arndt PF, Behrens S, Bourque G, Vingron M. CpG deamination creates transcription factor-binding sites with high efficiency. Genome Biol Evol. 2011;3:1304-11. Abstract

View all comments by Bernard Crespi

Related News: Deciphering Themes for Schizophrenia’s Genetic Variation

Comment by:  Patrick Sullivan, SRF AdvisorDanielle Posthuma
Submitted 16 November 2012
Posted 16 November 2012

Gilman et al. pose exceptionally important and salient questions: given that increasingly detailed genomic data have established that many genes are now strongly implicated in the etiology of schizophrenia, how do we understand this? How can these different components of the “parts list” for schizophrenia be pieced together to derive a cogent etiological hypothesis for further testing?

The authors use a new computational approach to address these questions, and derive lists related to axon guidance, neuronal cell mobility, synaptic function, and chromosomal remodeling. Additional analyses suggest the coherence of their lists. These are good clues that deserve further evaluation.

It was intriguing that the authors included multiple types of genetic variation—rare but potent copy number variants (e.g., Kirov et al., 2012), rare exonic mutations (Xu et al., 2012), and common variations from genomewide association studies (Ripke et al., 2011)—as most authors have tended to conduct these analyses separately.

In sum, a nice contribution to the literature and initial steps towards tackling a tough problem in human genetics. But, there are four issues for readers to bear in mind in evaluating the results.

First, we hope that the authors make their program freely available. This is the standard in the field. Many of us are interested in evaluating the capacities of their program. To our knowledge, it is not now available, although it has been used in multiple published papers. We could find no link in the paper or on the senior author’s lab page.

Second, readers need to remember that this was an in-silico analysis. It produces hypotheses but does not (and cannot) provide proof. The methods are subject to multiple biases, and it was not clear how well these were controlled (see point 4 as well). We wondered whether known biases like gene size and LD patterns were well controlled.

Third, we would have liked to see greater scholarship. There is an unfortunate trend for computational biologists to produce tools without benchmarking them against existing tools or rigorously determining power and error rates. The lack of finding significant clusters in control sets is insufficient in showing the validity of their program. Are the authors’ claims that their new tool represents superiority truly justified?

Moreover, there are a lot of tools for performing analyses of these sorts (e.g., INRICH, FORGE, MAGENTA, Ingenuity, ALIGATOR, among many others). Indeed, these sorts of analyses are in the toolkits of most psychiatric genetics groups and are routinely applied. Given that there are many papers reporting results, a scholarly treatment of how their results compare to those of others and what the added value of their program is would have been useful.

Fourth, and most importantly, pathway analysis is completely dependent on the input—the genetic findings and the pathways. The findings that the authors used had issues. The CNV list is likely to change soon as the PGC CNV group completes its integrated analyses of tens of thousands of subjects. The exome list was based on a small and atypical sample, and much larger studies are in preparation (see SRF comment). The authors did not seem to confront the issue that all humans contain a lot of deleterious exonic variation. And (spoiler alert), the GWAS list is soon to increase markedly. More and more precise findings are sure to alter the results.

The pathways used were pretty standard—GO, KEGG, protein-protein interaction databases. Unfortunately, although widely used, these pathways have multiple issues. The content of many GO annotations and KEGG pathways have not been constructed by experts in the area. As one salient example, synaptic gene lists in standard pathway databases were quite imperfectly related to lists created by experts (Ruano et al., 2010). The authors also relied somewhat uncritically on the PPI databases. These have multiple issues, and some (unpublished) data suggest substantial error (i.e., large fractions of the predicted interactions are not, in fact, real or biologically meaningful). The fraction of the proteome screened adequately by these methods is small. Some interactions in these databases are non-specific, or occur between molecules that are never in the same place at the same time.

Indeed, the genes overrepresented in PPI databases were selected due to disease relevance or biological importance (e.g., there is a lot of work on P53). In general, the more a gene is investigated, the more interactions are found.

Still, this is a key paper, albeit a snapshot based on imperfect input data, and we look forward to seeing whether additional analyses confirm a role in schizophrenia of the networks identified currently with their program.


Kirov G, Pocklington AJ, Holmans P, Ivanov D, Ikeda M, Ruderfer D, Moran J, Chambert K, Toncheva D, Georgieva L, Grozeva D, Fjodorova M, Wollerton R, Rees E, Nikolov I, van de Lagemaat LN, Bayés A, Fernandez E, Olason PI, Böttcher Y, Komiyama NH, Collins MO, Choudhary J, Stefansson K, Stefansson H, Grant SG, Purcell S, Sklar P, O'Donovan MC, Owen MJ. De novo CNV analysis implicates specific abnormalities of postsynaptic signalling complexes in the pathogenesis of schizophrenia. Mol Psychiatry. 2012 Feb; 17(2):142-53. Abstract

Xu B, Ionita-Laza I, Roos JL, Boone B, Woodrick S, Sun Y, Levy S, Gogos JA, Karayiorgou M. De novo gene mutations highlight patterns of genetic and neural complexity in schizophrenia. Nat Genet. 2012 Oct 3. Abstract

Ripke S, Sanders AR, Kendler KS, Levinson DF, Sklar P, Holmans PA, Lin DY, Duan J, Ophoff RA, Andreassen OA, Scolnick E, Cichon S, St Clair D, Corvin A, Gurling H, Werge T, Rujescu D, Blackwood DH, Pato CN, Malhotra AK, Purcell S, Dudbridge F, Neale BM, Rossin L, Visscher PM, Posthuma D, Ruderfer DM, Fanous A, Stefansson H, Steinberg S, Mowry BJ, Golimbet V, de Hert M, Jönsson EG, Bitter I, Pietiläinen OP, Collier DA, Tosato S, Agartz I, Albus M, Alexander M, Amdur RL, Amin F, Bass N, Bergen SE, Black DW, Børglum AD, Brown MA, Bruggeman R, Buccola NG, Byerley WF, Cahn W, Cantor RM, Carr VJ, Catts SV, Choudhury K, Cloninger CR, Cormican P, Craddock N, Danoy PA, Datta S, de Haan L, Demontis D, Dikeos D, Djurovic S, Donnelly P, Donohoe G, Duong L, Dwyer S, Fink-Jensen A, Freedman R, Freimer NB, Friedl M, Georgieva L, Giegling I, Gill M, Glenthøj B, Godard S, Hamshere M, Hansen M, Hansen T, Hartmann AM, Henskens FA, Hougaard DM, Hultman CM, Ingason A, Jablensky AV, Jakobsen KD, Jay M, Jürgens G, Kahn RS, Keller MC, Kenis G, Kenny E, Kim Y, Kirov GK, Konnerth H, Konte B, Krabbendam L, Krasucki R, Lasseter VK, Laurent C, Lawrence J, Lencz T, Lerer FB, Liang KY, Lichtenstein P, Lieberman JA, Linszen DH, Lönnqvist J, Loughland CM, Maclean AW, Maher BS, Maier W, Mallet J, Malloy P, Mattheisen M, Mattingsdal M, McGhee KA, McGrath JJ, McIntosh A, McLean DE, McQuillin A, Melle I, Michie PT, Milanova V, Morris DW, Mors O, Mortensen PB, Moskvina V, Muglia P, Myin-Germeys I, Nertney DA, Nestadt G, Nielsen J, Nikolov I, Nordentoft M, Norton N, Nöthen MM, O'Dushlaine CT, Olincy A, Olsen L, O'Neill FA, Orntoft TF, Owen MJ, Pantelis C, Papadimitriou G, Pato MT, Peltonen L, Petursson H, Pickard B, Pimm J, Pulver AE, Puri V, Quested D, Quinn EM, Rasmussen HB, Réthelyi JM, Ribble R, Rietschel M, Riley BP, Ruggeri M, Schall U, Schulze TG, Schwab SG, Scott RJ, Shi J, Sigurdsson E, Silverman JM, Spencer CC, Stefansson K, Strange A, Strengman E, Stroup TS, Suvisaari J, Terenius L, Thirumalai S, Thygesen JH, Timm S, Toncheva D, van den Oord E, van Os J, van Winkel R, Veldink J, Walsh D, Wang AG, Wiersma D, Wildenauer DB, Williams HJ, Williams NM, Wormley B, Zammit S, Sullivan PF, O'Donovan MC, Daly MJ, Gejman PV. Genome-wide association study identifies five new schizophrenia loci. Nat Genet. 2011 Oct ; 43(10):969-76. Abstract

Ruano D, Abecasis GR, Glaser B, Lips ES, Cornelisse LN, de Jong AP, Evans DM, Davey Smith G, Timpson NJ, Smit AB, Heutink P, Verhage M, Posthuma D. Functional gene group analysis reveals a role of synaptic heterotrimeric G proteins in cognitive ability. Am J Hum Genet. 2010 Feb 12;86(2):113-25. Abstract

View all comments by Patrick Sullivan
View all comments by Danielle Posthuma

Related News: New Exome Evidence Points to Old Suspect in Schizophrenia

Comment by:  Francis McMahon, SRF Advisor
Submitted 23 January 2014
Posted 28 January 2014

I think these studies do represent real progress. Finding genetic support for particular pathways provides unique evidence for a causative role of these pathways in disease. Why didn't the case-control study point to individual genes? Disorders such as schizophrenia may be more like a plane crash than a typical inherited disease: Since many things can go wrong, each crash is different, but damage to key systems is very likely to lead to a bad outcome. The finding in Fromer et al. that there are 18 genes with recurrent deleterious de novo events should allow scientists to focus on these genes as especially important. The overlaps with autism and intellectual disability are interesting, though not entirely unexpected. Will we also see gene overlaps with illnesses such as bipolar disorder? It wouldn't surprise me if some of the same genes are involved, but with fewer, less deleterious hits.

View all comments by Francis McMahon