Genetics and Neuropsychology: A Merger Whose Time Has Come

Abstract

Genetics and neuropsychology have historically been two rather distant and unrelated fields. With the very rapid advances that have been taking place in genetics, research and treatment of disorders of cognition in the 21^st century are likely to be increasingly informed by individual differences in genetics and epigenetics. Although neuropsychologists are not expected to become geneticists, it is our view that increased training in genetics should become more central to training in neuropsychology. This relationship should not be unidirectional. Here we note ways in which an understanding of genetics and epigenetics can inform neuropsychology. On the other hand, given the complexity of cognitive phenotypes, neuropsychology can also play a valuable role in informing and refining genetic studies. Greater integration of the two should advance both fields.

Genetics and Neuropsychology: A Merger Whose Time Has Come

For most of the 20^th century there was no real need for an integration of the fields of genetics and neuropsychology. Human genetics was, for the most part, focused on the determinants of disease and utilized methods that were not conducive to elucidating brain-behavior relationships. In the latter part of the century, however, rapid advances in the sequencing of the human genome and in high-throughput techniques for identifying single nucleotide polymorphisms (SNPs) and quantifying gene expression have made it possible for researchers to aggressively search for the genes associated with a variety of normal and pathological phenotypes. It is now possible to interrogate millions of SNPs as well as perform genome-wide expression analyses, and with these technologies come the opportunity to gain a better understanding of the fundamental biological mechanisms that drive cognition and disorders thereof. Although substantial progress is being made, it is also true that the task of gene discovery has proven to be rather complex.

To date, twin studies, extended pedigree studies, and gene association studies have provided overwhelming evidence that genes play a major role in brain structure and function. As such, it is our view that in the 21^st century there needs to be a stronger integration of genomics and neuropsychology. In all likelihood treatment and rehabilitation recommendations will become increasingly individualized as individual differences based on genomic differences are elucidated. We do not expect neuropsychologists to become geneticists, but a greater familiarity with the genetics of brain and cognition is becoming increasingly relevant. In part, the purpose of this special section is to highlight the importance of this transition in neuropsychology. Although there are numerous topics relevant to the integration of neuropsychology and genetics, here we briefly discuss some specific topics that we feel are of particular importance. Rather than presenting a comprehensive literature review, we simply use examples of recently published studies, as well as examples from our own research for purposes of illustrating the points to be made. These topics highlight the numerous ways in which genetics can contribute to neuropsychology and vice versa. We believe that a better appreciation of them can only serve to expedite the integration of these two fields of study.

Polygenicity

Nearly a century ago, Fisher (Fisher, 1918) proposed the theory of polygenic inheritance, positing that continuous variation in a trait is caused by many genes, each of which contribute in a small degree to the phenotype in question. Polygenic inheritance quickly became one of the fundamental principles of behavior genetics, laying the foundation for concepts such as heritability, research methods such as the classical twin design, and the more recent methodological advancements such as genome-wide complex trait analysis (GCTA) (Yang et al., 2010). With the advent of genome-wide association studies (GWAS), polygenic inheritance has moved from the realm of theory to that of established reality. Study after study has shown that traits such as stature, personality, brain size, and intelligence, as well as disorders such as schizophrenia, multiple sclerosis, Parkinson’s disease, and late-onset Alzheimer’s disease are influenced by many common genetic variants each of relatively small effect (Davies et al., 2011; International Multiple Sclerosis Genetics et al., 2010; International Schizophrenia et al., 2009; Lambert et al., 2013; Nalls et al., 2014; Stein et al., 2012; Yang et al., 2010)

Because virtually all behavioral, cognitive, and brain phenotypes are likely to be highly polygenic, it will be important for neuropsychological studies to shift toward the examination of multiple genetic effects. Many of the genetically-informative studies that have been published in this and other journals have been candidate-gene studies, i.e., studies focused on a single gene with putative theoretical relevance to some cognitive function. These studies have been informative, but it is time to begin utilizing genetic data in new and more biologically plausible ways. The use of polygenic risk scores is one way in which this goal can be accomplished (International Schizophrenia et al., 2009).

Derived from large, consortia-based GWAS results, where the large sample sizes allow for the detection of multiple minor SNP effects, polygenic risk scores have emerged as a means of quantifying the polygenic nature of a phenotype (in most cases a disease or other discrete classification). These scores can then be used in predictive analyses in the same way that candidate genes have traditionally been utilized. For example, Chauhan and colleagues (2015) derived a polygenic risk score for late-onset Alzheimer’s disease based on 24 risk loci and found a significant association with hippocampal volume in non-demented older adults (Chauhan et al., 2015). This effect remained significant even after accounting for APOE genotype. Accounting for APOE genotype is important because the APOE-ɛ4 is a major risk allele for Alzheimer’s disease (Saunders et al., 1993). Similarly designed studies have found associations between the polygenic risk for Alzheimer’s disease and cortical thickness in clinically normal older adults (Sabuncu et al., 2012), mental status and verbal fluency (Vivot et al., 2015), as well as amnestic and non-amnestic mild cognitive impairment (Adams et al., 2015). Although polygenic risk scores are not without their limitations—the method does not account for interactions between genes, proper methods for weighting individual SNP effects remain unclear, and their predictive power can be weak—they represent a significant step toward integrating polygenicity with current methods of evaluating the effects of disease risk on brain structure and function.

Complexity of Cognitive Phenotypes

Why are neuropsychological test batteries notoriously lengthy? Readers of this journal know that the reason is that all neuropsychological tests are multi-determined. Put another way, they are phenotypically complex. It is necessary to compare and contrast performance on a large number of different tests in order to make sound inferences about specific cognitive processes and abilities that are spared or impaired. Given that cognitive phenotypes are highly polygenic, the same principle applies when trying to understand the underlying genetic determinants of cognition.

Episodic memory serves as a good example. Tests of episodic memory may involve list learning, story recall, recall of designs, paired associates, or other formats, and can be used to examine encoding, recall at different delay intervals, and recognition. In a multivariate twin analysis of the California Verbal Learning Test (CVLT; Delis, Kramer, Kaplan, & Ober, 2000), we showed that the same genes influenced short- and long-delay free recall (Panizzon et al., 2011). However, there were some genetic influences that were specific to the learning trials and were independent of the genetic influences on free recall. The learning trials involve both acquisition and recall, but the recall trials involve no acquisition. Therefore, we surmised that genetic influences specific to the learning trials must represent genes that influence acquisition in episodic memory.

In another twin study of episodic memory, we examined short/immediate and long/delayed recall on the CVLT, and the Wechsler Memory Scale-III (Wechsler, 1997) Logical Memory and Visual Reproductions (W. S. Kremen et al., 2014). There was a highly heritable general episodic memory factor. The general factor indicates that there are significant genetic correlations among the phenotypes, i.e., there are shared genetic influences. Although there was a general factor, it is equally important to point out that approximately 30% of the genetic variance in Logical Memory and in Visual Reproductions was specific to each of those tests and independent of the general factor. Even CVLT and Logical Memory, which are both verbal recall measures, are not entirely influenced by the same genes.

In the twin studies we have described, the genes are “anonymous”, by which we mean that we do not know which genetic variants or how many of them are involved. Nevertheless, these results are highly informative for studies that target genes directly because they allow one to make a priori predictions about the degree of genetic overlap (pleiotropy) among the phenotypes; in other words, integrating epidemiological (e.g., twin) and molecular genetic (e.g., GWAS) approaches can be advantageous (W.S. Kremen et al., 2011; Panizzon et al., 2011; Papassotiropoulos & de Quervain, 2011). One of the biggest problems facing genome-wide association studies (GWAS) is the failure to replicate results across different samples. Rather than representing false positive findings, these failures to replicate might reflect real genetic differences underlying phenotypes that are assumed to assess the same latent cognitive construct. If in the case of episodic memory different phenotypes are simply lumped together across studies, then, in effect, those different phenotypes will be treated as if “memory is memory” and any memory measure will do. Our results and those of other researchers clearly demonstrate that episodic memory tests are not simply interchangeable at the genetic level. We must assume that the same is true for other cognitive constructs. A degree of sophistication with respect to cognitive phenotypes is therefore important for good genetic studies of cognition.

Cognitive Architecture

Phenotypic Versus Genetic

Phenotypic architecture refers to the (factor) structure and organization of cognition: are there factors and subfactors?; do certain measures/functions represent similar or different underlying abilities? Neuropsychology can contribute to genetic studies, in part, because it is a field in which much attention is paid to cognitive architecture. Factor analytic methods are a common approach to elucidating the phenotypic architecture of cognition. This approach, of course, goes back to testing Spearman’s g and variants of that theoretical model (reviewed by Panizzon et al., 2014). It might follow that factor analytic studies of cognition will be informative for GWAS, but we note two caveats regarding this notion. First, phenotypic factor analysis may be insufficient. Applying phenotypic factor analytic results to GWAS means that there is an unwritten assumption that the phenotypic factor structure is the same as the underlying genetic architecture, but that is not necessarily the case. As an example, we factor analyzed components for the Delis-Kaplan Executive Function System (Delis, Kaplan, & Kramer, 2001) Trail-Making Test. There was only a single factor at the phenotypic level, but the genetic factor analysis revealed that there were significant genetic influences on the switching condition that were independent of the general factor. Thus, the genetic factor analysis suggests a different approach to phenotype definition for genetic association studies. In another example, we found that there was only one phenotypic factor accounting for Tower of London scores; however, there were two genetic factors underlying the same data (W.S. Kremen et al., 2009).

In contrast to factor analytic models, neurocognitive structural models organize cognitive functions and infer neurocognitive structure on the basis of established findings from lesion studies and functional neuroimaging studies of intact brain networks. As such, these models should map more closely onto brain structure and function than factor analytic models. With respect to genetic models, the same principle applies. That is, it is simply an empirical question as to whether the underlying genetic structure is the same as is found in a phenotypic neurocognitive structural model. We have examined something similar regarding cortical, rather than cognitive, structure. Cortical atlases have traditionally been based on what is known about structure and function (e.g., Brodmann areas), but we found that genetically defined cortical regions did differ from more traditional regions based on sulcal-gyral boundaries (Chen et al., 2012).

Where To Begin?

There is the question of what population to start with. Paralleling the issue of phenotypic versus genetic architecture, we cannot assume that the genetic architecture of cognition is the same in healthy individuals and those with psychiatric and neurological disorders. Neuropsychology is, of course, primarily focused on psychiatric and neurological disorders. If, for example, we want to understand the genetics of cognitive dysfunction in schizophrenia, one may be tempted to start with a factor analysis of cognitive test scores in people with schizophrenia. However, we have to acknowledge that we are only in the very early stages of understanding the genetic underpinnings of cognition. We think it is probably best to first explore the genetic underpinnings of cognition in general population samples and then focus on particular disorders.

Cognitive Endophenotypes

Gottesman introduced the endophenotype concept to psychiatry and psychology (Gottesman & Gould, 2003). In part, endophenotypes were thought to represent clues to the underlying genetics of a disorder because they are less genetically complex and would involve fewer genes that the disorder itself. Gottesman and Gould noted that the rationale was that endophenotypes would be “very specialized and represent relatively straightforward and putatively more elementary phenomena (as opposed to behavioral macros)” (p. 637). Unfortunately, cognitive endophenotypes are not very elementary. Two prime examples are working and episodic memory as endophenotypes for schizophrenia. Both represent complex, multidimensional and multi-determined phenotypes with their own complex genetic architectures (Flint & Munafò, 2007; W.S. Kremen et al., 2007; W. S. Kremen et al., 2014; Papassotiropoulos & de Quervain, 2011). Nevertheless, both show a pattern consistent with that of an endophenotype for schizophrenia, in that the unaffected relatives of patients with schizophrenia perform more poorly on tests of these domains compared with healthy controls, and the degree of deficit is worse with increasing genetic relationship to an affected individual (i.e., unaffected MZ co-twins of patients more deficient than unaffected DZ co-twins of patients) (Cannon et al., 2000; Glahn et al., 2003; van Erp et al., 2008). While the genetic complexity of these cognitive phenotypes is probably on a par with that of schizophrenia, these cognitive phenotypes are still likely to be closer to the mechanisms of abnormal gene action than the schizophrenia phenotype and any genetic associations with them can be tested for confirmation in experimental studies with transgenic models (Cannon & Keller, 2006).

Gene-Environment Interplay

Genes do not act in isolation, but rather through constant interactions with one another and the organism’s environment. While a discussion of gene-gene interactions goes beyond the scope of this commentary, the dynamic interplay between genes and the environment is of particular relevance. This dynamic can be described in terms of two phenomena: gene-environment correlation and gene-environment interaction. Gene-environment correlation refers to the process by which an individual’s genotype influences the environment that they inhabit (Eaves, Chen, Neale, Maes, & Silberg, 2005). These processes can be passive (e.g., parents with higher IQs raising their children in more intellectually enriched environments), evocative (e.g., a child demonstrating superior intelligence is given more scholastic opportunities), or active (e.g., an individual with superior intelligence seeking out more intellectually challenging employment). In each of these examples, the environmental factors are likely to enhance the expression of genes predisposing to higher intellectual function.

Gene-environment interaction, on the other hand, represents the phenomenon whereby the response to an environmental factor varies as function of one’s genotype. Gene-environment interaction is perhaps of greater significance to neuropsychology as it demonstrates the mutability of genetic effects and identifies opportunities for psychosocial or behavioral intervention. The impact of physical activity on genetic risk factors for age-related cognitive decline is an ideal example of such an interaction. Ferencz and colleagues (2014) created a genetic risk score based on the genes PICALM, BIN1, and CLU, which have been associated with episodic memory. They found that the effect of that risk score on episodic memory differed as a function of physical activity level. Specifically, being physically active was protective; a high genetic risk score was associated with memory impairment only in physically inactive individuals. Findings such as this adhere to what McClearn (2006) referred to as “contextual genetics,” i.e., placing genetic effects into a broader biological context and trying to understand the dynamic systems at work. This way of thinking is not inconsistent with how neuropsychologists already approach cognition, either at the level of the individual or clinical populations.

It is also worth noting that while changes in the apparent effects of genes may be due to changes in strength of genetic effects, these changes may also result from the reduction of environmental influences in the presence of stable genetic effects. In an application of the classical twin design, we demonstrated this principle with analysis of reading ability, as assessed by the Wide Range Achievement Test (W.S. Kremen et al., 2005). The heritability of reading recognition differed as a function of parental education, ranging from .21 when parental education was lowest to .69 when parental education was highest. There were, however, no changes in the magnitudes of the genetic effects. Instead this change was due to substantial change in the effects of the common or shared environment. Common environmental factors accounted for approximately 50% of the variance at the lowest level of parental education, but accounted for no variance at the highest end. These results would support the value of environmental interventions such as Head Start for children at the lower end of the socioeconomic spectrum, who would, on average, have parents with lower education.

Beyond Genotyping

Genotyping consists of assessing the structure of DNA. Differences in the DNA sequence are static. However, epigenetics refers to dynamic processes by which the actions of genes are modified. In other words, it is about functionally relevant changes to the genome that do not involve changes in the DNA sequence. DNA methylation is one example. It refers to a chemical modification of the DNA by the addition of a methyl group, which effectively shuts off the gene. If the gene is shut off, it means that it is not expressed. There is evidence that early life events can cause lasting DNA methylation changes that can affect behavior many years later (e.g., McGowan et al., 2009). Gene expression (transcriptomics) refers to transcribing of DNA into RNA, which may then be translated into proteins that affect functioning (Feinberg, 2008). DNA methylation or gene expression may be dysregulated in particular disorders. However, whereas one’s genotype is the same in all tissues, DNA methylation and gene expression not only change over time but they are also in different tissues. An obvious limitation for the field of neuropsychology is that it is not possible to examine these phenomena in the living human brain. However, other tissues, such as blood may still have some utility (Breen et al., 2015). These processes reflect some of the mechanisms through which gene-environment interactions play out.

Concluding Remarks

There seems to be little doubt that in the 21^st century, elucidating the genetics of cognition and brain-behavior relationships in healthy and pathological conditions will be a central component of research and treatment. Recognition of that reality suggests a two-way street. A stronger focus on genetics should, thus, strengthen training in neuropsychology by helping neuropsychologists to be better informed about genetics. On the other hand, there is no shortage of expertise in neuropsychology regarding the complexity of cognitive phenotypes. As a major source of that expertise, neuropsychology can also make valuable contributions toward refining genetic studies of cognitive phenotypes. This special section is one small step toward the goal of a greater integration of the fields of genetics and neuropsychology.