Paradoxically frequent? Psychiatric disorders in the evolutionary perspective

Given the high heritability, significant impairment, morbidity and increased rates of mortality of psychiatric disorders, several hypotheses have been proposed to explain why natural selection does not remove the causal alleles of these phenotypes from the population.

This question is universal for any frequent phenotype that reduces the fitness of the carriers of the risk variants, and literature is full of examples of phenotypes were natural selection seems to paradoxically behave in an opposite way than expected. Nevertheless, once one looks deeper into the biology of the analyzed system, realizes that selective forces are still acting according to the laws of Darwinian selection. For example, thalassemias are a type of anemia that are caused by mutations in the genes coding for the hemoglobin chains; despite of the impairing conditions associated to the disease, thalassemias are present at a relatively high frequency in particular geographic regions. A simplistic interpretation of the principles of Darwinian selection would suggest that this situation is impossible. However, from a genomic point of view, it has been shown that most causal alleles show genomic signatures of selective sweeps. Why mutations that produce anemia are naturally maintained in the population? This paradox can be solved by realizing that thalassemias occur in regions that show (or have traditionally shown) a high prevalence of malaria, a disease produced by a protozoa that needs to infect blood cells during its reproductive cycle. Digging a bit more into the epidemiology of thalassemia and malaria, it can be seen that individuals that do not carry thalassemia mutations are more prone to suffer from malaria, whereas individuals that carry two thalassemia mutations tend to have anemia. However, carriers of only one thalassemia-causing allele (heterozygotes) are more resistant to malaria and do not suffer from anemia.Thus, from a selective point of view, heterozygotes have a fitness advantage over both types of homozygotes and more chances of producing descendants.

This classical example of population genetics provides some main lines that any study that attempts to explain the prevalence of a disorder in a population must take into account:

  • The hypothesis of selection must be validated in the causal alleles of the phenotype. This can be cumbersome if the causal variant of the phenotype is unknown.
  • Identification of the factor triggering the selective pressure requires deep knowledge of the etiology of the phenotype, which is usually something not really available.
  • The factor triggering the selective pressure must be set in a historical context.

In the case of complex phenotypes, such as psychiatric disorders, all these lines are compromised, making everything more difficult. First of all, complex phenotypes involve many causal loci (i.e. they are polygenic) with a very small effect on the phenotype; moreover, alleles associated to a particular complex phenotype are usually identified by means of genome-wide association studies (GWAS), but the causal loci are not known. Second, statistical tests for detecting polygenic adaptation are still in their infancy and classical tests for detecting selective sweeps are usually underpowered for detecting selection at complex phenotypes. Finally, each locus can affect more than one phenotype (pleiotropy), each under different selective pressures.

Nevertheless, this does not imply that detecting signatures of polygenic adaptation in psychiatric disorders is not affordable. Recently, Polimanti and Gelernter published a study in Plos Genetics ( where they conducted analyses to detect polygenic adaptation in five psychiatric disorders: attention deficit hyperactivity disorder (ADHD), autism spectrum disorder (ASD), bipolar disorder (BP), major depressive disorder (MDD), and schizophrenia (SCZ). The authors take advantage that there are GWAS (however, some of them are underpowered) for all these phenotypes from the Psychiatric Genomics Consortium (, and maps of signals of different types of positive selection in the human genome for different human populations ( However, since the GWAS have been conducted in individuals of recent European ancestry, the authors focus on European populations. Out of the five phenotypes analyzed, the authors identified signals of polygenic adaptation in ASD and SCZ-associated alleles with incomplete (i.e. it has not reached fixation) selection in European populations. Next, they analyzed genetic correlation of ASD with some other phenotypes ( and concluded that ASD is related to several advantageous traits: years of schooling, college completion, childhood intelligence and opennessto experience. Based on these results, the authors hypothesize that ASD alleles present beneficial effects with respect to cognitive ability, a feature under selective pressure.

It is unclear whether the absence of statistically significant results for the other phenotypes is due to the lack of power of the GWAS for detecting associations due to low sample size of the studies, lack of power of the applied boosting statistics for detecting polygenic selection or to true lack of selective advantage of these phenotypes. In the case of ADHD, several hypotheses have been proposed to explain the high prevalence of the phenotype in current populations (see GWAS with larger sample sizes and the use of more powerful selective tests will be required in order to properly test these hypotheses.

3D genomic maps. Better than Google maps for understanding the functional role of GWAS hits

Genome wide association scan (GWAS) is a popular method among genetic epidemiologists for identifying genetic variants associated to a trait of interest. GWAS power relies on analyzing the statistical association between a phenotype of interest (i.e. “autism” versus “non-autism” individuals) in a panel of genetic markers (usually single nucleotide polymorphisms (SNPs)) that densely covers the genome. The description of millions of genetic variants in the human genome, advances in genotyping technologies, the generation of large databases of individuals for the traits of interest and the invaluable international cooperation among different groups of investigators identified during the last decade thousands of genetic markers associated to hundreds of phenotypic traits ( However, understanding the biological meaning of these hits is a challenge for the scientific community, and translating such knowledge to patient treatment is at its infancy for most of the analyzed traits.

For scientists, one of the many intriguing issues of a GWAS outcome is that most of the reported genetic variants tend to be non-coding, falling far away from any described gene or any other known functional element of the genome. This empirical observation is theoretically explained by the fact that statistical hits do not necessarily represent a causal variant, but more likely are “spatially close” to the (not genotyped) causal genetic variant. Here, “spatially close” is interpreted at one dimension. As a consequence, genetic epidemiologists tend to report as putative functional candidates all the genes that are at an ad hoc genomic distance from a hit; further in-deep analyses for identifying the causal gene and the causal genetic variant are performed in silico by applying powerful bioinformatics tools that combine the expertise from different fields of knowledge. Nevertheless, even after performing all these analyses, in most of the cases the conclusion of the study is just a general overview of the physiology of the phenotype summarized in a sentence like “among all the genes close to the hits we observe gene enrichment for a pathway X” and a list of associated SNPs of unknown function. This (quite frustrating) view is rapidly changing with the incorporation of new technological advances that do not consider the one-dimensional nature of DNA, which do not capture many regulatory interactions, but the tridimensional conformation of DNA within the cellular nucleus.

In a recent paper published in Nature (, Won et al described the tridimensional map of chromosome conformation during human corticogenesis. The role of human brain development in regulating gene networks has been claimed as one of the factors explaining neurodevelopmental disorders, such as autism or schizophrenia. After describing the topology of the 3D map of chromatin conformation during human corticogenesis, the authors conducted several analyses for understanding the evolutionary impact of this 3D conformation. More interestingly for understanding the etiology of schizophrenia, Won et al analyzed how this topological conformation relates to previously reported 108 SNPs associated to that disease. Similar to other GWAS studies, nearly all of these loci reside in relatively uncharacterized non-coding regions of the genome. The authors show that most of these SNPs, which were previously described to be associated to the disease only from a statistical point of view, are close in the 3D space to genes with a likely role in postsynaptic density, acetylcholine receptors, neuronal differentiation, and chromatin remodellers. Amazingly, most of these genes are not in linkage disequilibrium with the GWAS hits, but proximity –and likely interaction- is only at tridimensional level.

Overall, this study shows that attacking the “functional hit problem” requires a multidimensional perspective from a multidisciplinary and genomic point of view. Exciting times are coming for understanding the etiology of complex phenotypes and -we hope- for developing new treatment strategies.