Behaviour and genesBMJ 1999; 319 doi: https://doi.org/10.1136/bmj.319.7201.37 (Published 03 July 1999) Cite this as: BMJ 1999;319:37
- a Social Genetic and Developmental Psychiatry Research Centre, Institute of Psychiatry, Kings College London, London SE5 8AF
- b Division of Psychological Medicine, University of Wales College of Medicine, Cardiff CF4 4XN
- Correspondence to: P McGuffin
New discoveries in genetics seem to hit the headlines almost daily, and probably the most eyecatching and controversial are those dealing with human behaviour. Thus, there has been popular media interest in reports of genes conferring susceptibility to psychiatric diseases and a whole range of “genes for” traits such as aggression, intelligence, and neuroticism. The scope for sensationalism and oversimplification is great. Here, we outline some of the basic concepts and lines of evidence from quantitative genetics indicating that genes do have important influences in determining human behaviour but that this nearly always involves an interplay with the environment. We then look at ways in which molecular methods are being used to locate and identify genes and how such approaches may impact on clinical practice.
Behaviour runs in families
We tend to resemble our parents, siblings, and other close relatives not just in the way we look but in the way we behave. However, the types, patterns, and causes of familial behavioural traits are extremely varied. They range from rare single gene disorders, such as Huntington's disease, that present with dramatic changes in behaviour in adult life, to commoner but more genetically complex disorders, such as schizophrenia and manic depression, through to normal variation in traits that are usually measured quantitatively, such as personality or intelligence.1 They can also include characteristics such as political or religious persuasion and career choice.
One of the basic pitfalls in studying the genetics of behaviour (or indeed any other type of trait) is to assume that familiality necessarily implies genetic transmission or that strong evidence of familial clustering implies single gene effects. Geneticists have long been aware that mendelian inheritance can be simulated by other mechanisms.2 To illustrate this, McGuffin and Huckle studied the educational histories of the families of a cohort of medical students: they found that attending medical school was roughly 80 times more common in the students' first degree relatives than in the general population and that when complex segregation analysis (a computerised method of exploring the mode of transmission) was performed the “trait” closely resembled an autosomal recessive one.3
A more satisfactory basic model for complex traits is that the phenotype results from a combination of the genotype and the environment to which it is exposed. Environment canbe split into two broad types—shared environment, which acts on all family membersto make them similar, and non-shared experiences that are specific to individuals, whichwould be expected to cause differences in how family members behave.
Identification of relevant genes will improve understanding of the molecular neurobiology of psychiatric disorders
This will lead to the development of more efficacious and more specific drugs
DNA testing may be used in predicting response to treatment and susceptibility to side effects
DNA testing will be useful in counselling patients' relatives at high risk of heritable disease but not for population screening
Public perception of psychiatric disorder will change: improved understanding of the causes and mechanisms of disease is likely to reduce stigma
Experiments of nature
Two types of “natural experiment” enable us to estimate the extent to which complex traits are familial because of shared genes, shared environment, or a combination of both. The first is afforded by twinning and the second by adoption.
The basis of twin studies is that monozygotic or identical twins are naturally occurring clones, having all of their genes in common. On the other hand, dizygotic or fraternaltwins share on average half of their genes. If we assume that the environment shared by twins is roughly the same for monozygotic and dizygotic pairs, then any greater observed similarities in monozygotic than dizygotic pairs should reflect the action of genes. This “equal environments assumption” is open to criticism since there is some evidence that monozygotic pairs report greater environmental similarities (such as dressing the same and sharing friends) than do dizygotic pairs, especially during childhood.4 Nevertheless, the assumption can be tested by incorporating environmental measures in twin studies or by looking at the effects of mistaken zygosity (the twins and their parents believe that they are monozygotic when they are actually dizygotic or vice versa). Sometimes it is also possible to compare the resemblance of monozygotic twins reared together with those reared apart. In practice, these checks suggest that the equal environments assumption is generally valid.5
Figure 1 summarises results from twin studies on various disorders and traits including psychological “symptoms” in normal twins. There are clear differences between monozygotic and dizygotic twins for some phenotypes—schizophrenia,6 manic depressive disorder,7 unipolar depression,4 childhood autism,8 attention deficit hyperactivity disorder,1 and cognitive ability as measured by IQ test1—suggesting genetic effects. For bulimic behaviour 10 and disabling fatigue in childhood,11 there is evidence of familiality (there are positive correlations) but less clear genetic effects (modest differences in the monozygotic and dizygotic correlations).
Adoption studies have been used less extensively than twin studies but have been crucial in studying some conditions, such as schizophrenia. Heston studied 47 index adoptees separated from their schizophrenic mothers within 72 hours of birth and compared them in their mid-30s with 50 control adoptees who did not have schizophrenic parents.12 Five of the index adoptees (11%) became schizophrenic (roughly the rate expected in non-adopted offspring of schizophrenics) compared with none of the controls. Subsequently, a series of studies was carried out in Denmark; the most recent study showed that the frequency of schizophrenia was 16% in the biological relatives of schizophrenics adopted early in life, compared with 1.8% in adoptive relatives and the relatives of control adoptees.13
Quantifying nature and nurture
Major advances in quantitative psychiatric genetics have occurred in the past two decades with the increased availability of high speed computing. This has enabled researchers to go beyond merely estimating heritability— that is, the proportion of total phenotypic variance explained by additive genetic affects. It is now possible to accurately quantify both genetic and environmental effects and assess the extent to which a reduced model (such as one with no additive genetic or no common environmental effects) explains the data compared with a full model. A recent application to schizophrenia showed that the heritability may be as high as 80%, and, although this leaves 20% of variance to be explained by the environment, this seems to be entirely of the non-shared type.6 The table shows examples of behavioural disorders or traits that are thought to have a genetic component.
It is also possible to carry out multivariate model fitting that examines two or more disorders or traits at a time.14 For example, life events are generally thought to have a causal relation with depression, but a recent study suggested that at least part of the co-occurrence of life events and depressive symptoms results from genetic and shared environmental influences on the extent to which both are reported.15 A similar approach can be applied to psychiatric syndromes that often occur together. It had been shown that the same genes influence both anxiety and depressive disorder but that the environmental influences on the two types of disorder are apparently distinct.16
Such findings are not an end in themselves. Demonstrating that a disorder is substantially heritable or that a pair of disorders overlap genetically but have different environmental influences leads to two other types of study. The first is to identify the specific environmental factors that co-act or interact with genes and the second is to locate and identify the genes themselves. Here we consider only the latter.
Mapping and positional cloning
Figure 2 shows the general process of positional cloning. A chromosomal region containing a gene that confers susceptibility to the disorder of interest is identified by linkage mapping. The region is then narrowed down by means of various methods until the gene itself is identified. Subsequently, the mutations (or variations) that confer susceptibility to disease are identified. The distribution, level of expression and functions of the gene products can then be studied. The process of positional cloning carries the possibility of incremental benefits for clinical practice in the forms of predictive testing, refined diagnosis and eventually the development of targeted specific treatments.
However, conventional linkage analysis requires several assumptions: that there are major gene effects to be detected, that the sample is genetically homogeneous (that is, allof the affected individuals have the same cause of disorder), and that the mode of transmission of the disorder is known. A study of large, multiply affected pedigrees with early onset of Alzheimer's disease has identified mendelian subforms of the disease: three distinct forms have been mapped, and the mutations identified.17 This has not been the case with other psychiatric disorders such as schizophrenia or manic depression. Here a bewildering and apparently contradictory array of positive linkage study results has been reported,18 19 suggesting that single genes of large effect are rare or non-existent and that these disorders are oligogenic (resulting from the combined effects of several genes) or perhaps polygenic (resulting from many genes).
Sib pair analysis
An alternative approach that is useful in oligogenic or polygenic diseases is sib pairanalysis. It is possible to take several hundred DNA markers roughly evenly spaced alongthe 23 pairs of human chromosomes and carry out genotyping in a series of sibling pairs ofwhich both have the same disorder. The probability that siblings share 0, 1, or 2 alleles at any marker locus are, respectively, 0.25, 0.5 and 0.25. However, if a marker locus is close to (and therefore linked with) a locus conferring susceptibility to the disease this will be detectable as increased allele sharing at the marker. This approach has been successful in identifying susceptibility loci for disorders such as type 1 diabetes, and a variant of the method has been used to map genes involved in reading disability (or dyslexia) on chromosomes 6 and 15.20
The main drawback is that susceptibility loci of very small effect (such as conferring a relative risk of less than 2) may require large numbers of sib pairs—in the region of 600 to 800—to be detected.19 In a disorder such as schizophrenia the relative risk in a sibling of an affected individual is about 10; thus, if several additive genes are involved, none may individually have a relative risk of more than 2.
Just as it is possible to search through the entire genome for disease susceptibility genes using classic linkage or sib pair methods, it is now becoming feasible to do the same thing using allelic association to detect linkage disequilibrium. Linkage disequilibrium occurs when a marker allele and the locus for susceptibility to a disease are so closely linked that their association is preserved over many generations of recombination. The advantage of mapping genes using linkage disequilibrium is that it can detect very small effects1 and might the only way of finding the genes involved in polygenic disorders. The disadvantage is that, because linkage disequilibrium takes place only over very short distances, several thousand markers are required to complete a genome search. New methods of high throughput genotyping are now being developed that can accomplish this based on DNA pooling 21 or studying single nucleotide polymorphisms on microarrays.22 The first “top to bottom” search of a whole chromosome by means of DNA pooling has now been published,23 and such methods should allow whole genome scans for linkage disequilibrium to be completed in the foreseeable future.
Predicting the clinical impact of mapping genes
The most immediate benefit of identifying genes contributing to common familial psychiatric diseases will be in our understanding of the basic neurobiology of disease, but this should lead on to the discovery of new and more specific drug treatments. It is as much commercial farsightedness as scientific altruism that has led the major pharmaceutical companies into serious investment in genotyping technology and the development of detailed maps of single nucleotide polymorphisms.24 Safer, more effective treatments will obviously be to the benefit of patients, but it is also possible to envisage DNA testing being used to predict which patients will respond to different types of antipsychotic or antidepressant treatment or who will be susceptible to particular side effects. Preliminary evidence suggests that these approaches will work. For example, response to antipsychotic drugs seems to be influenced by genotype at the gene for serotonin receptor 5-HT2A.25
What about predicting whether someone will develop a psychiatric disorder and attempting to prevent it? Prediction is already possible for rare dementias of early onset, such as Huntington's disease, or single gene forms of Alzheimer's disease. Understanding the molecular basis of these conditions should lead on to the development of effective treatments and preventive methods, but none exists yet. Somewhat paradoxically, genetic predictive testing is much more difficult for commoner disorders such as depression or schizophrenia, where effective treatments exist but were the genetic basis is complex. Psychiatric genetic counselling is already available in some specialist centres, but at present it is possible only to offer empirical figures for those at high genetic risk. For example, the child of a schizophrenic parent has about 10 times the risk of developing the disorder compared with a member of the general population, where the lifetime risk is 1%. The child's risk increases to 16 times if a sibling is already affected and is over 40 times the population risk when both parents have schizophrenia.26
It is likely that individual risk prediction will become more precise once the molecular genetic basis of schizophrenia is better understood, but it is unlikely that risk prediction will ever be better than about 50% accurate since monozygotic twins are discordant for schizophrenia 50% of the time. In our view, this means that DNA based population screening for complex psychiatric disorders (including Alzheimer's disease of late onset) will never become a reality but that screening for high risk relatives probably will. This could have obvious benefits in advising relatives on avoiding risk factors (such as the recreational use of cannabis in the case of schizophrenia) or, more controversially, in attempting prevention with a low dose antipsychotic agents.
There will also be less immediately tangible benefits to sufferers from psychiatric disorders. It has sometimes been feared that “geneticisation” could contribute to the stigma of mental disorder. So far the experience has been just the opposite. Alzheimer's disease is now widely perceived as a “real” disorder with a rapidly unfolding molecular aetiology. We predict that this is the start of a trend and that identifying the genes involved and understanding causation will do much to improve public perception and acceptance of other psychiatric disorders.
Funding PMcG is director of an MRC funded research centre. NM holds an MRC studentship.
Competing interests None declared.