Chapter 2. Genetics and Epigenetics of Addiction

Abstract

The last half-decade or so has seen massive progress in psychiatric genetics. We have exited the age of candidate gene studies. With the benefit of hindsight, we realize that this is a good thing; most candidate gene findings have not survived into the current age of unbiased genomewide investigation (Border et al. 2019; Duncan et al. 2019). Most of this progress is attributable to the use of large samples for genomewide association studies (GWASs) and subsequent analyses based on the results of these GWASs. Psychiatric traits are genetically complex—that is, they involve many genes, each of small effect—and for complex traits, large samples are required if we are to make credible progress. Our ideas of “large” also have shifted. We still occasionally see published articles describing samples of a few hundred patients as “large.” This may be the case for some areas of investigation, but it is not so for genetics. In the field of genetics, hundreds or thousands of participants may be adequate to identify the risk loci of largest effect, but for most traits, this is not the case, and we may need more than 10,000 subjects; to be called “large,” certainly more than 100,000 would be needed. Such studies are possible because of the advent of large meta-analysis consortia, such as the Psychiatric Genomics Consortium (PGC; Sullivan et al. 2018), large biobank samples such as the UK Biobank (UKB; Collins 2012) and the Million Veteran Program (MVP; Gaziano et al. 2016), and the direct-to-consumer company 23andMe (Check Hayden 2017). When traits can be identified that span multiple large-participant collections, sample sizes of more than 1,000,000 subjects are possible (Liu et al. 2019). These large samples are often crude tools when phenotype is concerned. Investigators have to work with what is available. In the case of meta-analyses, this frequently means use of a lowest-common-denominator phenotype to which every contributing investigator has access. For biobanks, the traits studied must be of medical interest or fall into the broad category of things the originators of the sample thought were useful or interesting. These limitations have been surmountable for psychiatric traits such as schizophrenia and, to a lesser extent, major depressive disorder. These are instructive examples of what can be obtained with adequate data. For large genetics studies, substance use disorders (SUDs) fall into two categories. The legal substance dependencies—alcohol dependence and nicotine dependence—have benefited from the availability of large biobank samples (Liu et al. 2019). Alcohol and tobacco have such widespread and well-recognized effects on health that information about these two substances is generally collected in any study focused on health outcomes. Alcohol dependence and tobacco dependence (i.e., alcohol use disorder and tobacco us disorder) are also fairly common, and although they are stigmatized, the stigma is not huge. Accordingly, we review a great deal of data on these substances later in this chapter. The situation is very different for the illegal substance dependencies, such as opioid and cocaine use disorder. These are less common than the legal substance dependencies; data are often not collected systematically in biobank samples, and when they are collected, we presume that because of stigma and other factors, the false-negative rate is higher. Thus, genetics studies of the illegal substance dependencies have had to rely mostly on purpose-collected samples (Gelernter et al. 2014b, 2014c) or on samples that used very ad hoc diagnostic criteria. Because funding institutes (public and private) have given only limited support to the recruitment of informative cohorts for these illicit drug dependencies, studies to date tend to be underpowered, replications are difficult to come by, and results are therefore more difficult to interpret than those for alcohol- and nicotine-related traits. SUDs are inherently different from other psychiatric traits such as schizophrenia, because they include key pharmacogenomic components. This architecture predicts that there may be variants of relatively large effect for these traits. Indeed, a few variants of comparatively large effect have been identified for SUDs and related traits, as is discussed later in this chapter. With these issues in mind, we now discuss the current state of the art.

Topics & Concepts