Brain, Vol. 122, No. 11, 2015-2032,
November 1999
© 1999 Oxford University Press
Invited review |
The genetic basis of cognition
Institute of Molecular Medicine, John Radcliffe Hospital, Oxford, UK
Correspondence to:
Institute of Molecular Medicine, John Radcliffe Hospital, Oxford OX3 9DS, UK E-mail: jf{at}molbiol.ox.ac.uk
 |
Abstract |
|---|
 Top Abstract IntroductionGenetic influences on...Specific cognitive effects of...Complex genetic disorders with...Which genes are implicated...Non-specific mental retardationSyndromic mental retardation:...Syndromic mental retardation:...What effect do genes...Animal models of learning...Specific effects of mutations...Systems neuroscience and...ConclusionReferences |
|---|
The molecular characterization of single-gene disorders or chromosomal abnormalities that result in a cognitive abnormality (predominantly mental retardation) and of the genetic variants responsible for variation in intellectual abilities (such as IQ, language impairment and dyslexia) is expected to provide new insights into the biology of human cognitive processes. To date this hope has not been realized. Success in finding mutations that give rise to mental retardation has not been matched by advances in our understanding of how genes influence cognition. In contrast, the use of engineered mutations in mice to study models of learning and memory has cast new light on the molecular basis of memory. A comparison of studies of human and mouse mutations indicates the limitations of current genetic approaches to the understanding of human cognition. It is essential to interpret a mutation's effect within a well-characterized neural system; mutations can be used to define gene function only when the mutation has an effect on a system whose constituents form a serial causal chain, such as the molecular components of a signal transduction pathway. Typically, however, genetic mutations with a cognitive and behavioural phenotype are characterized by specific effects on different systems whose inter-relationships are unknown. Genetic approaches are currently limited to exploring neuronal function; it is not yet clear whether they will throw light on how neuronal connections give rise to cognitive processes. We need a much greater integration of different levels of understanding of cognition in order to exploit the genetic discoveries. In short, a rapprochement between molecular and systems neuroscience is required.
cognition; genetics; mental retardation; memory; learning
|
Introduction |
|---|
|
TopAbstractIntroductionGenetic influences on...Specific cognitive effects of...Complex genetic disorders with...Which genes are implicated...Non-specific mental retardationSyndromic mental retardation:...Syndromic mental retardation:...What effect do genes...Animal models of learning...Specific effects of mutations...Systems neuroscience and...ConclusionReferences |
|---|
Genetic influences on cognitive processes have long been considered to be non-specific, a view based on two considerations: first, there is a lack of convincing evidence from individuals with genetic or chromosomal anomalies that genetic damage can result in specific neuropsychological deficits; secondly, there is an insufficient number of genes (~100 000 in the entire human genome) to allow the determination of either the number or the interconnections of neurons that give rise to human cognitive processes. Genes, it was thought, could set up the scaffold of the brain, but would have a limited influence on the mind. It seemed hardly plausible that there could be genes for memory or learning, let alone genes determining reading abilities. Nevertheless, there are now claims that such genes are close to being found.
In this review I ask what the study of genetic variants has told us about cognitive processes and, perhaps more importantly, what we can expect it to tell us. I start by showing that there are some genetic disorders that have associated cognitive phenotypes, although they are not as specific as is sometimes claimed. Recognizing `pure' cognitive phenotypes in genetic disorders would considerably strengthen the idea that genetic mutations can directly influence cognition, so neuropsychological investigation of people with genetic and chromosomal disorders is potentially very useful.
It might be expected that the molecular characterization of genetic defects that result in a relatively pure cognitive phenotype would open a new window into the biology of cognitive processes. I review what is known about the genetic mutations and variants that give rise to cognitive dysfunction, work that has not, I argue, significantly advanced our understanding of the biology of cognitive processes.
By contrast, the analysis of mutations in mice has been spectacularly successful in dissecting the molecular basis of memory and learning. The disparity with the analysis of human mutations is at first sight curious. Why should the two not be equally enlightening? Both involve the characterization of mutants, a classic way of defining the function of genes (or more accurately gene products). Perhaps if we could answer this question we might know how best to move forward the genetic dissection of human cognitive phenotypes.
I argue that the most promising approach is to bridge the divide between systems neuroscience and molecular neuroscience. One of the impediments to scientific advance here might be described as cultural: there is relatively little communication between those involved in cloning the genetic basis of complex human traits and those busy trying to understand brain function from the perspectives of cognitive neuroscience and neuropsychology. Neither approach on its own can give a complete description. To demonstrate this, I show that cognitive processes are not determined in a direct fashion by gene action, so that the study of mutants cannot, on its own, tell us much about the biological basis of cognitive processes.
|
Genetic influences on intelligence and specific cognitive processes |
|---|
|
TopAbstractIntroductionGenetic influences on...Specific cognitive effects of...Complex genetic disorders with...Which genes are implicated...Non-specific mental retardationSyndromic mental retardation:...Syndromic mental retardation:...What effect do genes...Animal models of learning...Specific effects of mutations...Systems neuroscience and...ConclusionReferences |
|---|
Throughout this review the terms intelligence, cognitive process, intellectual function and ability are largely interchangeable, reflecting uncertainty about how many neural systems are involved and how they interrelate. However, this should not obscure an important distinction between IQ measures of cognition, which attempt to capture a generalized measure of mental ability, cognition understood as composed of specific cognitive processes such as memory and language.
I am primarily concerned here with the analysis of mutations and their effects on specific cognitive processes, because these currently hold out most hope for revealing something about the underlying biology. As I discuss in more detail later, genetic dissection is most appropriate when there is a direct, unmediated relationship between genetic mutation and phenotype, of the sort that may exist in certain mental retardation syndromes. The relationship between IQ and a genetic variant is likely to be much more complex; the available genetic data suggest that the genetic influences on IQ are relatively non-specific.
Psychometricians have studied individual differences in intelligence, primarily using IQ measures, and have accumulated evidence for a genetic influence; 50% of the variation in IQ test scores can be attributed to genetic variation (Devlin et al., 1997
; McClearn et al., 1997; Bouchard, 1998). Correlations between different measures of mental abilities are always positive, indicating the validity of IQ as a measure of general intelligence, but some correlations are larger than others, indicating also that there are separable broad types of mental ability (such as verbal and spatial ability). The relatively good correlations between intelligence test component scores suggest that the genes that determine variation in, for example, visuospatial abilities will also influence variation in verbal abilities. In other words, the genes that influence IQ scores have a relatively non-specific effect. Furthermore, multivariate genetic analyses implicate common genetic factors as the major contributors to phenotypic correlations between intelligence and some personality dimensions (genetic correlations are reported to be as high as 0.4) (Harris et al., 1998). In turn, genetic effects on personality determine other behaviours; e.g. the genes that predispose to the personality trait neuroticism overlap with those that predispose to some mood disorders (Kendler et al., 1993). With genetic correlations this high, the chances of finding discrete genetic effects are slim.
This is not to say that the study of individual differences in cognitive phenotypes is unrewarding. As I show below, genetic analysis of dyslexia has been successful, but success here is due primarily to a much more detailed understanding of the phenotype than is currently possible with general measures of intelligence such as IQ.
|
Specific cognitive effects of genetic lesions in humans |
|---|
|
TopAbstractIntroductionGenetic influences on...Specific cognitive effects of...Complex genetic disorders with...Which genes are implicated...Non-specific mental retardationSyndromic mental retardation:...Syndromic mental retardation:...What effect do genes...Animal models of learning...Specific effects of mutations...Systems neuroscience and...ConclusionReferences |
|---|
Examples of single-gene or chromosome disorders with a purely cognitive phenotype are important as they provide evidence that genes can have a relatively direct effect on cognition, and strengthen the case for using molecular genetics to understand the biology of cognition. However, it is extremely difficult to find the necessary examples. While some genetically determined syndromes are associated with specific and characteristic behaviours (Flint, 1996
) [such as the unusual self-mutilation in Lesch–Nyhan syndrome (Nyhan, 1976)], the case is much harder to make for the existence of specific cognitive phenotypes. Indeed, if we consider only single-gene or chromosomal disorders it is probably true to say that there are no such examples (but it should also be borne in mind that there are insufficient data to come to a definitive conclusion on this point).
The most promising examples are families in which a cognitive disability that is not associated with physical abnormalities segregates as a single Mendelian trait. Such conditions are termed non-specific (or non-syndromic) mental retardation (to differentiate them from the >800 forms of syndromic mental retardation that have a complex physical phenotype) (Wahlstrom, 1990; Simonoff et al., 1996). Table 1 lays out a classification of the genetic and chromosomal conditions that are associated with cognitive disabilities; these are discussed in more detail below.
|
Non-specific cognitive dysfunction can be found segregating in families (Hurst et al., 1990
; Gedeon et al., 1996) but there are so few affected individuals that it is very difficult to decide whether, in addition to the cognitive defects, there are personality and behavioural characteristics. With such small numbers, the confounding effects of genetic background and environmental differences cannot be controlled and we cannot be certain that there is a specific immediate genetic effect on cognition. However, it has to be admitted that the neuropsychological investigation of families with non-specific mental retardation is very limited. There is, in fact, just one good example in the literature that illustrates the difficulties.
About half of the males and females of a four-generation family were found to have a speech and language disorder, leading to claims that affected members had a specific impairment in grammar (Gopnik, 1990; Hurst et al., 1990; Gopnik and Crago, 1991). For example, family members could produce plurals of familiar but not unfamiliar words and failed to generate inflected forms of regular but not irregular verbs. On the basis of this family, there have been claims that `there is suggestive evidence for grammar genes, in the sense of genes whose effects seem most specific to the development of the circuits underlying parts of grammar' (Pinker, 1994). Further assessment of the family has shown that speech and language difficulties are not as specific as first thought and are part of a broader syndrome that includes lower than average IQ scores (Vargha-Khadem et al., 1995). Furthermore, the phenotype is not purely cognitive; affected members have a pronounced impairment in articulation, a praxic deficit, that also involves non-linguistic oral and facial movements. Nevertheless, the phenotype in this family is intriguing and deserves further study.
Compared with families in which there is non-specific mental retardation, cognitive phenotypes associated with syndromic mental retardation are less easy to use as evidence for an immediate genetic determination of cognition because of the suspicion that mental retardation could be secondary (or even more distantly related) to the other abnormalities that constitute the syndrome. On the other hand, the larger numbers of affected individuals that can be examined means that the investigation of syndromic mental retardation with characteristic cognitive profiles does go some way towards showing that cognitive functions can be dissected genetically. In other words, just as components of cognition can be defined at the anatomical or even the cellular level, they can also be recognized at the genetic level.
The two most thoroughly investigated examples are the Williams and Turner syndromes (Table 1
). Both conditions are due to abnormalities of the chromosomal content: Williams syndrome is due to a deletion of part of chromosome 7q (technically it is therefore a monosomy) (Ewart et al., 1993), and Turner syndrome is due to monosomy of all, or part, of the X chromosome. Both conditions are the consequence of monosomy for more than one gene, which might make it hard to detect the effects of single genes; however, as described later, a combination of molecular and neuropsychological investigations has gone some way towards that goal.
Williams syndrome is a rare congenital condition diagnosed on the basis of developmental delay, a characteristic face and a heart defect (supravalvular aortic stenosis and peripheral pulmonary artery stenosis) (Martin et al., 1984; Burns, 1986; Morris et al., 1998). Neuropsychological testing of Williams syndrome children shows a consistent discrepancy between verbal and performance IQ. Compared with age- and IQ-matched controls (with Down syndrome), Williams syndrome children have preserved syntax and lexical abilities but diminished visuoperceptual skills (Udwin et al., 1987; Crisco et al., 1988; Bellugi et al., 1990; Wang et al., 1992; Wang and Bellugi, 1994). While investigators agree that there are significant discrepancies in verbal and performance IQ, there is less consensus about how to interpret this finding. Bellugi and colleagues have argued strongly that the language abilities of Williams syndrome children are intact despite the presence of serious cognitive defects. But the evidence is not definitive, as on some tests of comprehension Williams syndrome children perform worse than expected. Using a test of grammatical gender assignment in French-speaking Williams syndrome children, Karmiloff-Smith and colleagues show that the subjects had deficits in syntax production compared with normal children (Karmiloff-Smith et al., 1997).
There is also evidence that the superior verbal abilities of Williams syndrome children may not be the result of a relatively immediate genetic determination of a cognitive function. Verbal ability develops at a faster rate than non-verbal ability in Williams syndrome children, implying that the genetic abnormality affects a developmental trajectory rather than directly determining visuospatial abilities (Jarrold et al., 1998). Furthermore, Williams syndrome children do not just have a cognitive phenotype; there appear to be other behavioural and personality components. Inappropriate social interaction, hyperactivity, anxiety, sleep disturbance and hyperacusis occur at significantly higher rates than in controls matched for age, gender and level of mental retardation (Einfeld et al., 1997; Greer et al., 1997). It may be that interaction between the other features of the syndrome gives rise to the cognitive phenotype. However, as we shall see later, combined molecular and neuropsychological investigation does argue in favour of a direct genetic effect on some parts of the cognitive phenotype.
The second chromosomal disorder that may have a specific cognitive profile is Turner syndrome. This syndrome is not a mental retardation syndrome, as although IQ scores are lower than normal, mental retardation is unusual in children and adults with the syndrome. Claims that women with Turner syndrome have a specific deficit in non-verbal ability or spatial functioning, sometimes attributed to non-dominant hemisphere function, go back over 30 years (Alexander et al., 1964; Money and Mittenthal, 1970; Rovet and Netley, 1980, 1982; Netley and Rovet, 1982; Money, 1993). For example, a twin study of Turner syndrome showed that the affected twin had a performance IQ 18 points below her sister while her verbal IQ was only 7 points lower (Weiss et al., 1982; Reiss et al., 1993). Further investigation has identified difficulties in spatial and mathematical abilities (Temple and Carney, 1993; Mazzocco, 1998; Siegel et al., 1998).
A study of `social cognition' in Turner syndrome has strengthened the case for a cognitive profile (Skuse et al., 1997). Using molecular techniques to categorize the Turner syndrome sample into those whose X chromosome is maternally derived (Xm) and those with a paternal X chromosome (Xp), Skuse and colleagues gave the two groups IQ and behavioural inhibition tests. In the latter, the subject first learned an association (say `one' when you see `1') and then had to inhibit the response (say `two' when you see `1'). They found that scores were higher in Xp than in Xm females and that Xp females did as well as normal females on the behavioural inhibition task. The group concluded that there is a gene influencing `social cognition' on the X chromosome that carries an imprint of the parent of origin.
Does Turner syndrome provide an example of a relatively direct genetic effect on cognition? Again, the evidence could be interpreted to mean that the genetic effects are distant, and the psychological profile of Turner syndrome girls could represent an exaggeration of normal sex differences (Temple and Carney, 1993), involving many different genes and genetic interactions.
Thus, although there is some evidence that distinctive cognitive phenotypes are associated with different genetic disorders, as we would expect if the genes had a part in determining cognitive processes, there is still room to doubt that the genetic effect is immediate.
|
Complex genetic disorders with cognitive phenotypes |
|---|
|
TopAbstractIntroductionGenetic influences on...Specific cognitive effects of...Complex genetic disorders with...Which genes are implicated...Non-specific mental retardationSyndromic mental retardation:...Syndromic mental retardation:...What effect do genes...Animal models of learning...Specific effects of mutations...Systems neuroscience and...ConclusionReferences |
|---|
Single-gene disorders might not result in specific cognitive effects, but this is not to say that specific cognitive defects do not have a genetic basis. The clearest examples are to be found in the literature on specific language impairment and dyslexia. The distinction between specific language impairment and global cognitive impairment classically rests on the presence of a large verbal–non-verbal discrepancy in IQ tests that is not explained by any other cause, just as specific reading retardation (developmental dyslexia) depends on a poor reading ability incommensurate with the child's age and IQ. Twin studies have made clear that in many respects this is an artificial distinction; when discrepancies alone are used for diagnostic classification then neither condition is genetically distinct from less specific disorders (Pennington et al., 1992
; Bishop et al., 1995). On the other hand, twin studies also show that some patterns of disability are specific. The patterns for specific language impairment remain to be discovered, but a lot is known about dyslexia.
Assessments of rhyme judgements (for instance comparing word groups like hat, cat and dog) (Bradley and Bryant, 1983) of verbal working memory (Hulme and Roodenrys, 1995), phonetic elisions (such as `cat' without `c' is `at') and rapid naming of letters, numbers and colours (Bowers and Wolf, 1993) have shown that dyslexia is characterized by abnormal phonological processing. A combination of functional neuroimaging and psychophysical experiments has begun to define the neuroanatomical basis of the disorder. For instance, in men with dyslexia the strong functional connection between the left angular gyrus and other left hemisphere regions involved in normal reading is absent (Horwitz et al., 1998). Thus, dyslexia is an example of a partly genetically determined syndrome with a pure cognitive phenotype that has well-defined psychological abnormalities. Indeed, we may even know the neuroanatomical basis of the neural systems that are disordered in dyslexia. Genetic approaches have begun to build on these data to show how individual genetic loci can contribute to specific elements of the disorder.
One way in which it has been possible to detect the specific effect of genes on a cognitive process is by using the methods of genetic linkage and association. Note that genetic mapping is first of all a statistical process. By itself it does not find genes but associations between part of a chromosome and a trait. The first successful example of genetic mapping applied to a complex cognitive process was the identification of a locus on chromosome 6p that influences reading disability or dyslexia (Cardon et al., 1994). The method used exploited the fact that the presence of genetic influences on a trait will partly determine the degree of regression to the mean for a trait. The same method also gives estimates of genetic correlations between traits and can therefore be used to determine the specificity of the effect, but it is necessary to know what to look for.
What other phenotypes might this locus influence? As already mentioned, the cognitive phenotype of dyslexia is phonological; it depends substantially on variation in the accuracy and speed of single-word recognition (Perfetti, 1985). Dyslexics are less accurate in reading pseudo-words than readers with similar accuracy in reading single real words (Rack et al., 1992). They are deficient in the ability to recognize phonemes (Pennington et al., 1990), the speech sounds that correspond roughly to the sounds of the alphabet. Are there common genetic influences on these processes? Quantitative genetic analyses, using regression, show that phonological coding, phoneme awareness and single-word reading have common genetic determinants.
Genetic mapping goes even further. Using the neuropsychological dissection of dyslexia, it is possible to look for genetic loci that influence the subcomponents of dyslexia. There is one report that the genetic effects are separable, though they are not completely independent. A single-word reading phenotype was found to be linked to a marker on chromosome 15q and phoneme awareness to 6p (Grigorenko et al., 1997). More recent genetic mapping studies, while confirming the existence at 6p21.3 of a reading-related locus, have not substantiated a specific effect on phoneme awareness but instead implicate the effect of the locus on several components of the phenotype (Fisher et al., 1999; Gayan et al., 1999). Nevertheless, all studies agree that the locus does not affect more general measures of intelligence and is indeed relatively specific for reading disability.
The conjunction of neuropsychological and genetic research in dyslexia is exceptional, and shows that genetic dissection of a cognitive phenotype is possible. It is perhaps surprising that complex genetics is associated with a more specific cognitive effect than a single genetic mutation, and one explanation for this is considered later. First, however, I discuss what is known about genetic mutations that result in cognitive dysfunction.
|
Which genes are implicated in human cognitive processes? |
|---|
|
TopAbstractIntroductionGenetic influences on...Specific cognitive effects of...Complex genetic disorders with...Which genes are implicated...Non-specific mental retardationSyndromic mental retardation:...Syndromic mental retardation:...What effect do genes...Animal models of learning...Specific effects of mutations...Systems neuroscience and...ConclusionReferences |
|---|
I have presented the evidence that specific cognitive phenotypes can be found in individuals with genetic disorders. I have argued that cognitive deficits are not completely specific, and associated behavioural, personality and other psychological characteristics are also found. Nevertheless, an important distinction between non-specific and syndromic mental retardation indicates that single-gene disorders can present with effects limited to psychological manifestations, although it must be admitted that we have only limited knowledge of the phenotype in non-specific mental retardation. Inadequate neuropsychological investigation of non-specific mental retardation due to single-gene mutations is a major obstacle to further progress.
I now review what is known about the genes that are abnormal (either by mutation or dosage) in disorders with cognitive disabilities. It is worth first clearing the ground of a number of autosomal Mendelian disorders that include cognitive disabilities in their phenotype. Do these disorders tell us how genes affect cognitive processes? Disappointingly, the short answer is that they do not.
The commonest examples are neurodegenerative disorders (cortical and subcortical dementias). The relatively specific deterioration in memory in Alzheimer's disease might suggest that genetic analysis would lead to new advances in the biology of memory. This has not happened. The dominant model explaining the aetiology of Alzheimer's disease is the amyloid cascade hypothesis, which states that the overproduction of amyloid-beta peptide or failure to clear this peptide results in amyloid deposition, neurofibrillary tangles and cell death (Hardy, 1997; Hardy et al., 1998).
Molecular cloning of rare inherited mutations that result in Alzheimer's disease supports the amyloid cascade hypothesis. Mutations have been found in the amyloid precursor protein and in the presenilin genes, which are involved in amyloid precursor protein trafficking (Goate et al., 1991; Levy-Lahad et al., 1995a, b; Sherrington et al., 1995). None of this tells us much about the biology of memory. Frontotemporal dementia with parkinsonism is clinically distinct from Alzheimer's disease, presenting with behavioural and personality changes before cognitive decline, but the molecular pathology, which involves mutations in the tau gene leading to neuronal death (Hutton et al., 1998; Spillantini et al., 1998), does not account for the difference.
Huntington's disease, an autosomal dominant progressive neurodegenerative disorder due to a gain of function mutation in the huntingtin gene (Huntington Disease Collaborative Research Group, 1993), is also characterized by cognitive deterioration, although the specificity of the impairment is open to debate (Foroud et al., 1995; Giordani et al., 1995; Jason et al., 1997). The relatively specific neuropathology (affecting predominantly the caudate, putamen and globus pallidus) appeared to be due to neuronal intranuclear inclusions of huntingtin (Davies et al., 1997) until more recent studies showed that nuclear aggregates do not initiate the disease (Klement et al., 1998; Saudou et al., 1998). Analysis of transgenic mice containing an abnormal Huntington's disease gene suggest that there is altered expression of multiple neurotransmitters that cannot be explained by generalized dysfunction of a particular cell type (Cha et al., 1998). Possibly more specific neuropathological processes are operating to produce cognitive dysfunction, but to date there is no genetic explanation for the cognitive decline.
Duchenne muscular dystrophy is an X-linked myopathy in which one-third of patients have non-progressive cognitive impairment compared with age-matched normal boys and boys with other neuromuscular disorders (Yoshioka et al., 1980; Leibowitz and Dubowitz, 1981; Whelan, 1987; Anderson et al., 1988; Dorman et al., 1988; Billard et al., 1992). The disorder is due to mutations in the dystrophin gene (Hoffman et al., 1987). Despite considerable heterogeneity in the cognitive profile of affected children (Dorman et al., 1988), there is some evidence of a specific cognitive defect: verbal skills are more severely impaired than non-verbal skills (Billard et al., 1992; Bresolin et al., 1994), and there is one report that children with Duchenne muscular dystrophy may have a phonological deficit akin to dyslexia (Billard et al., 1998). A lot is known about the protein product of the dystrophin gene and the finding that deletion of exon 52 of the dystrophin gene is associated with mental retardation suggested that cognitive dysfunction was a direct effect of the mutation (Rapaport et al., 1991). However, subsequent work has not supported this hypothesis (Bushby et al., 1995). Thus, although there is some evidence of a specific cognitive deficit in Duchenne muscular dystrophy, molecular characterization of the defect has not explained its pathogenesis.
|
Non-specific mental retardation |
|---|
|
TopAbstractIntroductionGenetic influences on...Specific cognitive effects of...Complex genetic disorders with...Which genes are implicated...Non-specific mental retardationSyndromic mental retardation:...Syndromic mental retardation:...What effect do genes...Animal models of learning...Specific effects of mutations...Systems neuroscience and...ConclusionReferences |
|---|
Given the assumption that a very large number of genes must be involved in cognitive processes, a view supported by the fact that there a large number of known genetic conditions, it comes as a surprise that mutations characterized in three non-specific mental retardation families are in biochemically related gene products. The three genes identified are GDI1 (guanine nucleotide dissociation inhibitor 1), oligophrenin and PAK3 (P21-activated kinase), and all are involved in Rho GTPase signalling networks (Allen et al., 1998
; Billuart et al., 1998; D'Adamo et al., 1998). GDI1 inhibits GDP dissociation from Rab3a by binding to GDP-bound Rab proteins. Rab3a is a small GTP-binding protein that plays a role in the recruitment of synaptic vesicles for exocytosis (Geppert et al., 1994; Sudhof, 1995). Oligophrenin encodes a rhoGAP protein that stimulates the intrinsic GTPase activity of the small G proteins Rho, Rac and Cdc42. PAK3 links Rac and Cdc42 to the actin cytoskeleton and to transcriptional activation via JNK and p38 proteins. How might any of this relate to human cognitive function?
There are, at present, two possibilities. The first is that normal development of axonal connections is disrupted in patients with mutations in these genes. As yet this hypothesis is unproved, but it fits with what is known about the cell biology of the Rho GTPases (Van Aelst and D'Souza-Schorey, 1997). Growth cones, the specialized structures of developing axons, find their way through the brain by sampling molecular signals, helped by GTPases. Cdc42 and Rac1 are involved in the formation of lamellipodia and filopodia (Nobes and Hall, 1995), both of which are linked to PAK proteins; for instance, the formation of dendrites is reduced by inhibiting Rho, Rac and Cdc42 (Threadgill et al., 1997). Perhaps cognitive dysfunction in the three mental retardation syndromes is due to a failure of cortical development.
The second possibility is that synapse function is compromised. The main evidence here comes from our understanding of the function of Rab3a, a protein that is expressed only in neurons and neuroendocrine cells and is localized in secretory vesicles (Sudhof, 1995). Synaptic vesicles contain Rab3a, which is the most abundant Rab protein in the brain. In one model, exocytosis of synaptic vesicles leads to the dissociation of Rab3a from the vesicle (Sudhof, 1995). Since Rab3a-deficient mice have no fundamental deficits in synaptic vesicle exocytosis (Geppert et al., 1997), the protein is not essential to the process, but it is required to maintain a normal reserve of synaptic vesicles. The GDI1 mutation, by disrupting Rab3a traffic, is expected to alter neurotransmitter release, which might, in turn, account for the intellectual impairment.
Both explanations raise an issue that has been repeated many times in studies of genetic effects on cognition. Why is the effect of the mutation specific? Both the developmental and the synaptic transmission account of Rho GTPase involvement must explain why only neurons involved in cognitive systems are disrupted. One likely explanation is that the mutations only partly disrupt the brain system on which they operate. I return to this question below, when considering the effects of mutations on learning and memory in other species.
The fourth gene that has been implicated in non-specific mental retardation is FMR2 (Gecz et al., 1996; Gu et al., 1996). We know next to nothing about what the protein product does, other than that the gene encodes a nuclear protein that may regulate transcription (Gecz et al., 1997a, b). The dearth of information about its function is more typical of our knowledge of genes involved in mental retardation than is suggested by the success with the Rho GTPase family of genes.
|
Syndromic mental retardation: single genes |
|---|
|
TopAbstractIntroductionGenetic influences on...Specific cognitive effects of...Complex genetic disorders with...Which genes are implicated...Non-specific mental retardationSyndromic mental retardation:...Syndromic mental retardation:...What effect do genes...Animal models of learning...Specific effects of mutations...Systems neuroscience and...ConclusionReferences |
|---|
Despite an immense amount of work, molecular analysis of syndromal mental retardation has not been particularly enlightening about the neurobiology of cognition. The reason for this is largely because the genetic effect is not immediate. Either the genes are transcription factors, acting on many other interacting systems (e.g. the gene for
-thalassaemia mental retardation X-linked syndrome, ATRX) or the phenotype requires the action of more than one gene (e.g. Prader–Willi syndrome). A division into conditions due to mutations in a single gene and those due to chromosomal rearrangements is helpful for descriptive purposes, but this probably does not reflect a major distinction in pathogenesis.
The mutation in the FMR1 gene that is responsible for the fragile X syndrome, a common heritable cause of mental retardation, was characterized in 1991 and found to be an unusual trinucleotide repeat expansion, the first example of a mutational mechanism common to a series of predominantly neurological and neuropsychiatric conditions (Pieretti et al., 1991; Verkerk et al., 1991). The repeat expansion has been studied exhaustively but it tells us nothing about how the mutation results in mental retardation. Furthermore, 8 years after the FMR1 gene was cloned its function is still unknown. In the normal brain the FMR protein is found in nearly all neurons. It can bind RNA, including its own transcript, and it has been postulated that the protein has a role in the machinery of translation (Eberhart and Warren, 1996). FMR1 knockout mice have macro-orchidism and impaired spatial learning abilities, but nothing specific that tells us what the gene does (Dutch–Belgian Fragile X Consortium, 1994). Biochemical and immunofluorescence studies reveal a tight co-localization of FMR protein with cytoplasmic ribosomes, similar to that observed for translation factors (Khandjian et al., 1996; Feng et al., 1997a, b). Possibly the protein is involved in the control of gene products, but this still does not explain why absence of the protein should manifest as a cognitive defect.
More is known about the genetic basis of the ATRX syndrome (Gibbons et al., 1995). The disorder is X-linked, and patients have an anaemia (-thalassaemia), a characteristic facial appearance, profound developmental delay, neonatal hypotonia and genital abnormalities. The gene product (ATRX) contains a PHD finger (a putative zinc-binding domain) and a motif that relates ATRX to a group of proteins called helicases. Other members of this group are known to bind to chromatin, and ATRX may be involved in chromatin remodelling, which is considered to be a crucial step in the control of gene expression. ATRX is widely expressed in the brain, heart and skeletal muscles, so pleiotropic effects are expected; perhaps it operates by regulating the expression of a restricted class of genes, hence accounting for the phenotype.
The final example is Optiz G/BBB syndrome, an X-linked multiple organ disorder that includes developmental delay. Mutations have been found in a gene called MID1, the features of which suggest that it is involved in developmental regulation by protein–protein interactions (Quaderi et al., 1997). Again, this has not advanced our understanding of the biology of cognition.
|
Syndromic mental retardation: segmental aneusomy syndromes |
|---|
|
TopAbstractIntroductionGenetic influences on...Specific cognitive effects of...Complex genetic disorders with...Which genes are implicated...Non-specific mental retardationSyndromic mental retardation:...Syndromic mental retardation:...What effect do genes...Animal models of learning...Specific effects of mutations...Systems neuroscience and...ConclusionReferences |
|---|
A number of syndromes associated with cognitive dysfunction have been found to be due to chromosomal rearrangements, among which Down syndrome (trisomy 21) is by far the commonest (accounting for about one-third of all cases with moderate to severe retardation) (Hou et al., 1998
; Murphy et al., 1998). The complex phenotypes of chromosomal rearrangements are thought to arise because of the loss (in the case of monosomy) or addition (in the case of trisomy) of dosage-sensitive genes, of unrelated function, that happen to lie next to each other on the chromosome. Most syndromes occur because of relatively small regions of aneusomy (Down syndrome is an exception here) and are consequently known as segmental aneusomy syndromes. For example, in the WAGR (Wilms tumour, aniridia, genitourinary abnormalities, mental retardation) syndrome, monosomy for the PAX6 gene results in aniridia type 2 and monosomy for the WT1 gene results in Wilms tumour (Van Heyningen and Hastie, 1992). The two genes lie close together on chromosome 11. Since many segmental aneusomy syndromes include mental retardation in their phenotype, their molecular characterization could lead to the identification of dosage-sensitive genes that affect cognitive function.
In two cases this approach has been successful. In Williams syndrome it has been possible to identify genes that contribute to different components of the syndrome; families have been found with a mutation affecting the elastin gene and presenting with supravalvular aortic stenosis and facial features typical of Williams syndrome, but not with the cognitive profile (Frangiskakis et al., 1996). Other, very rare, families have some facial features, supravalvular aortic stenosis, verbal ability and short-term memory similar to those of unaffected members, but marked impairment of visuoconstructive skills. Molecular characterization of these individuals showed that the chromosomal deletion was small (only 84 kb compared with >500 kb in the majority of Williams syndrome patients), which permitted the researchers to isolate a candidate gene, LIMK1 kinase (Frangiskakis et al., 1996). The LIM domain, first identified in three homeodomain (developmental) proteins, is a zinc finger motif believed to function as a protein–protein binding module in neural development (Wanaka et al., 1997). LIMK1 binds to several isoforms of protein kinase C and to neuregulin (Wang et al., 1998). Transmembrane neuregulins interact with LIMK1 and co-localize at the neuromuscular synapse, suggesting that the two proteins have a role in synapse formation and maintenance, though how this explains the defect in visuospatial cognition in Williams syndrome patients is a mystery.
Rubinstein–Taybi syndrome is characterized by abnormal craniofacial features, broad thumbs and mental retardation. It arises from monosomy of a small region on chromosome 16p13.3, where mutations have been documented in the Cbp gene (Petrij et al., 1995). The protein product of the Cbp gene binds to the cyclic AMP (cAMP) response element binding protein and to several elements of the basal transcriptional machinery, suggesting that mutations will disturb the transcription of numerous genes. Thus, as is the case with the ATRX syndrome, the molecular basis of Rubinstein–Taybi syndrome lies in a gene with effects on many different systems.
From the point of view of advancing our knowledge of genetic influences on cognitive processes, dissection of other segmental aneusomy syndromes has been less successful. Two clinically distinct disorders, the Prader–Willi and Angelman syndromes, arise from deletions of a small region of 15q11-q13 (Nicholls et al., 1998). The two syndromes have characteristic and distinct neurobehavioural profiles: in Angelman syndrome the retardation is severe (very few affected individuals can talk) and there is ataxia, seizures, hyperactivity and paroxysmal laughter. By contrast, in Prader–Willi syndrome the mental retardation may be only mild and there is the specific behavioural abnormality of hyperphagia, resulting in severe obesity. An immense amount of work has gone into the characterization of the small deleted region. The basic defect is not simply a dosage effect; it turns out that about one-quarter of cases of Prader–Willi syndrome are due not to a deletion but to the inheritance of two maternal copies of chromosome 15 (rather than the usual situation of one maternal and one paternal chromosome 15). Conversely, two paternal copies of chromosome 15 result in Angelman syndrome. The chromosomal region at 15q11–13 is said to bear a parent-of-origin imprint, of which the molecular signature is a difference in DNA methylation. Further work suggests that, although the molecular deficit is the same in the two disorders (a failure of parent-of-origin-specific expression), it looks increasingly unlikely that the genes act directly or specifically on the CNS.
Mutations in a ubiquitin protein ligase gene (UBE3A) have been found in a few rare families with Angelman syndrome (Kishino et al., 1997), and it has been proposed that the UBE3A gene is maternally expressed. If mutations in UBE3A are the cause of Angelman syndrome, they are unlikely to tell us much about the origin of the cognitive phenotype. The gene product is part of a widely used ubiquitin-mediated protein degradation pathway. The deletion almost certainly has pleiotropic effects that will be hard, if not impossible, to disentangle.
Prader–Willi syndrome is probably not the result of a defect in a single gene. Seven genes (and candidate genes) have been identified in the Prader-Willi syndrome region at 15q11–13, all of which appear to be brain-specific (Budarf and Emanuel, 1997). The function of these genes is as yet unknown: one gene that has been identified, IPW, does not even code for a protein and is therefore similar to two other imprinted non-translated RNAs, H19 and Xist. Potentially, therefore, the phenotype arises from deficits in all these genes. It is not known how the genetic defect causes intellectual impairment.
Similar problems beset attempts to understand how deletions of 22q11 give rise to cognitive disabilities. The DiGeorge, velocardio-facial and conotruncal anomaly face syndromes are different manifestations of deletions of 22q11 (Scambler, 1993). The DiGeorge and velocardio-facial syndromes are both associated with mental retardation; additionally, psychosis is found in some patients with 22q11 deletions. The region most consistently deleted is large (>1.5 Mb), containing at least 14 genes. Cloning and sequencing of the entire region has not identified any obvious candidates for the cognitive defect and it now seems likely that the syndromes arise from combined monosomy of more than one gene (Budarf and Emanuel, 1997).
Attempts have been made in Down syndrome to correlate regions of trisomy with different phenotypic abnormalities and hence to infer the location of specific genes. While no genes have been identified solely on this basis, the information has been crucial in driving attempts to make a mouse model of Down syndrome. First, mice with three copies of chromosome 16 (the mouse homologue of human chromosome 21) show many of the features of Down syndrome (Reeves et al., 1995). The astrocytosis, craniofacial abnormalities and seizures seen in trisomy 16 mice mimic the phenotype of Down syndrome. Secondly, attention has been focused on 21q22.2 as a potential site for dosage-sensitive genes that affect learning and behaviour. By putting pieces of human DNA from 21q22.2 into mice and testing them for deficits in memory, one gene from this region has been identified. It is a human homologue of the Drosophila minibrain gene, which is a tyrosine/serine kinase expressed in developing neuroblasts. The use of transgenic mice to isolate minibrain is one possible way of dissecting complex phenotypes (including cognition) down to their molecular basis, although it is suitable only for dissecting trisomies, where there is an extra copy of the gene. As discussed later, changes in dosage are a potent way of altering a phenotype and we cannot expect this system to be widely applicable. Furthermore, there is as yet no proof that the minibrain gene in humans is either dosage-sensitive or a critical determinant of mental retardation in Down syndrome.
|
What effect do genes have on cognitive processes? |
|---|
|
TopAbstractIntroductionGenetic influences on...Specific cognitive effects of...Complex genetic disorders with...Which genes are implicated...Non-specific mental retardationSyndromic mental retardation:...Syndromic mental retardation:...What effect do genes...Animal models of learning...Specific effects of mutations...Systems neuroscience and...ConclusionReferences |
|---|
It is clear that success in finding mutations that give rise to mental retardation has not been matched by success in understanding how these mutations give rise to cognitive deficits. In fact the difficulties in interpreting the results of the gene cloning experiments prompt us to ask whether a genetic approach is really likely to tell us much about the biology of cognition. It goes without saying that knowing which gene is mutated implicates that gene in the phenotype, but that is not the same as using a mutation to infer function, as is done in classical genetics.
How we relate a phenotype to a mutation depends on how immediately the mutation affects the system under study. In systems where there is a direct causal pathway connecting a series of proteins, knocking out one protein (by mutating the cognate gene) defines the position of the protein in that pathway and therefore defines its function. Mutations in animals do not necessarily fit such a model. First, a genetic mutation operates throughout development, so the phenotype may not be directly due to the mutated gene; the mutation could disrupt the expression of a series of developmental genes, which in turn determine the tissue-specific regulation, in the adult animal, of proteins directly controlling the phenotype of interest. Secondly, mutations are influenced by interactions with other genes. This can happen in a number of ways. The phenotype of a mutation depends on the strain into which it is introduced (de Belle and Heisenberg, 1996; Crawley et al., 1997), or it may arise as a consequence of changes far downstream of the mutation. There may be a large number of different pathways affected, each with its own specific outcome. If these interact, attributing the final cause to the mutation, while true, does not tell us much about the immediate processes that give rise to the phenotype we are studying.
Do these considerations invalidate the use of genetics to investigate cognitive processes? Success in studying learning and memory in animals shows that they do not, as illustrated in the following review of genetic mutations that are used to dissect animal models of cognitive processes.
|
Animal models of learning and memory |
|---|
|
TopAbstractIntroductionGenetic influences on...Specific cognitive effects of...Complex genetic disorders with...Which genes are implicated...Non-specific mental retardationSyndromic mental retardation:...Syndromic mental retardation:...What effect do genes...Animal models of learning...Specific effects of mutations...Systems neuroscience and...ConclusionReferences |
|---|
One intensively studied animal model of learning is the gill and siphon withdrawal reflex of the sea-slug Aplysia (Bailey and Kandel, 1993
). While a single stimulus gives rise to sensitization lasting minutes to hours, repetitive stimulation produces sensitization that can last from days to weeks. The short- and long-term changes have different properties: long-term, but not short-term, sensitization is blocked by inhibitors of RNA and protein synthesis; long-term changes involve the growth of new synaptic connections. The key point here is that long-term sensitization involves gene induction, thus pushing research in the direction of looking for genes whose expression increases following neural activity.
At least part of the molecular pathway involved has now been identified: cell surface receptors activate protein kinase A via an increase in cAMP. Protein kinase A acts in the nucleus to activate CREB1 (cAMP response element binding protein 1) and to relieve CREB2-mediated repression (Kandel and Schwarz, 1982; Bailey et al., 1996; Byrne and Kandel, 1996). One consequence is the downregulation of cell adhesion molecules (in Aplysia these are termed apCAMs), decreasing the interaction of neurites and resulting in the generation of new synaptic connections.
A similar mechanism operates in the fruitfly Drosophila. Drosophila can be taught to associate an odour with an electric shock, and this association can be divided into long- and short-term components on the same basis as in Aplysia in that long-term association requires gene expression (Tully, 1996). One of the great advantages of working with Drosophila is that it is relatively easy to set up a screen for mutations that alter a phenotype, and then clone the mutated gene. The same genetic approach has been applied to behaviour; in this work flies are mutated and those that cannot learn are examined in detail. Ten mutations have been isolated from four separate screens (dunce, rutabaga, radish, turnip, cabbage amnesiac, latheo, linotte, nalyot and golovan) (Tully, 1996). Cloning of the first few of these mutations pointed to the importance of the cAMP second messenger system. The dunce gene encodes a camp-specific phosphodiesterase and rutabaga an adenylyl cyclase. Mutations in protein kinase A or its subunits disrupt olfactory learning.
The role of CREB was further investigated in flies by transgenic analysis. Expression of one form of CREB (dCREB2b) selectively blocks long-term memory (Yin et al., 1994) and overexpression of another CREB gene (dCREB2a) facilitates memory; a task normally requiring multiple spaced trials was acquired in one training trial (Yin et al., 1995). Just as in Aplysia, different forms of CREB have contrasting roles: dCREB2b suppresses and dCREBa enhances the long-term acquisition of olfactory learning.
The same mechanism is responsible for some forms of memory in mammals. Activity-dependent changes in synaptic strength are the most likely correlates of associative learning in any brain. One good example of such changes is long-term potentiation, which was first observed in the major synaptic pathways of the hippocampus (Bliss and Collingridge, 1993). It is conventionally elicited by giving a 25–100 Hz electrical stimulus for ~1 s to the hippocampal pathway, resulting in a lasting increase in synaptic strength. Of the various forms that have been found in the CNS, the most studied is NMDA (N-methyl-D-aspartate) receptor-dependent long-term potentiation, which is the form discussed in the remainder of this article.
Demonstrating the relationship between long-term potentiation and learning has not been easy (Eichenbaum, 1995). For instance, measuring hippocampal extracellular field potentials while an animal is learning spatial relationships would seem to be a good way of showing that neural activity in the hippocampus is due to this learning. Yet these changes have been shown to be due, at least in part, to variation in brain temperature (Moser et al., 1993).
Evidence that long-term potentiation is memory at work comes from studies of fear conditioning, in which a tone is associatively paired with an electric shock to the foot of a rodent. Studying fear conditioning in humans is not impossible but employs less fearful stimuli. The advantage of studying fear conditioning over studying neural activity in the hippocampus is that the neural circuitry from auditory stimulus to motor output is well defined. This makes it much easier to argue that any changes in receptor activity are correlates of the cognitive state rather than just bystander activity secondary to some more decisive process elsewhere in the brain. It has been shown recently that once a tone has been associated with shock, its enhanced response in the amygdala is specific (i.e. it does not reflect a general increase in responsiveness) and is identical to that elicited by electrically stimulated long-term potentiation (McKernan and Shinnick-Gallagher, 1997; Rogan et al., 1997).
The requirement that long-term potentiation must last longer than ~4 h for mRNA and protein synthesis to take place divides long-term potentiation into early and late forms and suggests that the mechanisms for associative learning are similar in Aplysia and Drosophila. In fact, the same molecular machinery is employed. Mice with a targeted mutation in the CREB gene were tested for fear conditioning, long-term potentiation and spatial memory deficits (Bourtchuladze et al., 1994). The animals could be conditioned to a tone and could learn to find a hidden platform in a water bath, but both skills were lost after 30 min. In other words, the deficit specifically affected long-term memory. Successful short-term training ruled out motivational sensory or motor deficits as an explanation. Electrophysiological studies showed that long-term potentiation was disrupted.
Eric Kandel and colleagues have argued that the system bears some similarities to the checks and balances on cell division (Abel et al., 1998). They argue that there are memory suppressor genes (equivalent to tumour suppressor genes) such as ApCREB2, which encodes a transcription factor that represses CREB1-mediated transcription. They also point to the importance of phosphatases, e.g. calcineurin, that inhibit kinases known to be important in the production of long-term memory. Memory suppressors may act as a control point for the laying down of memories, as they provide a mechanism by which a decision based on the salience of the stimulus can be translated into a decision to form an association; conceivably they may also control forgetting.
|
Specific effects of mutations are due to partial loss of function |
|---|
|
TopAbstractIntroductionGenetic influences on...Specific cognitive effects of...Complex genetic disorders with...Which genes are implicated...Non-specific mental retardationSyndromic mental retardation:...Syndromic mental retardation:...What effect do genes...Animal models of learning...Specific effects of mutations...Systems neuroscience and...ConclusionReferences |
|---|
While these results are major advances in our understanding of how associative learning is established at a molecular level, they are also problematic. Why are the effects apparently so specific? CREB proteins are not specific to the hippocampus. Indeed, they are widely distributed throughout the tissues of the body and are involved in many fundamental cellular processes. For instance, the cAMP response element mediates activation of gluconeogenic enzymes, which are crucial for neonatal survival (Schmid et al., 1993
), yet CREB knockout mice were not abnormal in this respect. CREB-deficient