The complexity of schizophrenia and how to understand it

An excellent, measured and thoughtful paper about the causes of schizophrenia appears in this week's Nature, a special issue on current knowledge about the disease.  Much recent research into this devastating disease has been gene-based, including of course genomewide association studies, but, as with all other complex traits, no simple genetic basis has been identified.  In this paper ("The environment and schizophrenia"), Jim van Os et al discuss reasons for this, and suggest ways to move the research forward.

GWAS have identified hundreds of genes for schizophrenia, or more, but currently only accounting for a small percent of the variation in disease presence or test-scores.  Depending on some assumptions and on what data one considers, estimates are that hundreds or even thousands of genes (including regulatory and other functional regions) contribute.  There are few that seem to make strong individual contributions, and one region that has been found to do that is in the HLA part of the immune system, a strange kind of finding.  Even with optimistic assumptions, predictive power will vary from sample to sample and population to population.  These will not, as currently designed, assess epigenetic changes, due to DNA modification.  And then there's the little trivial thing called the 'environment.'

Briefly, in this new paper van Os et al. argue that schizophrenia, and other 'psychotic syndromes', are the result of the interaction of the developing brain with environmental triggers during developmental sensitive periods in those with what is presumably a genetic susceptibility to disruptions in normal functioning of the brain.  We condense a long argument into one inadequate sentence here, but even so please note that it's a paean to complexity, flexibility as a response to the environment during development, and that genetic susceptibility is only one part of the picture.  

These researchers believe, based on a body of prior evidence, that the development of normal neuronal connections in the brain requires interaction with a variable environment, as they portray in the figure to the left.   Based on epidemiological data for associations of psychosis with environmental risk factors, they define risk as the stressors of urban environments, belonging to a minority group, developmental trauma, and/or cannabis use.  However, these factors are extremely common, while what they call the 'psychotic syndrome' is not.  This is where the genetic susceptibility comes in.

This suggests that beneath the relatively small marginal risks linking the environment to psychotic syndrome at the population level, vulnerable subgroups exist that are more sensitive to a particular environmental risk factor at a much larger effect size. Thus, the validity of observed associations with urban environment, developmental trauma, cannabis use and minority group position hinges on evidence of vulnerable subgroups. Genetically sensitive studies indicate that differential sensitivity to the psychosis-inducing effects of environmental factors may be mediated by genetic factors. For example, in siblings of patients with a psychotic disorder, who are at increased genetic risk to develop psychotic disorder, the psychotomimetic effect of cannabis is much greater than in controls, as is the risk to develop psychotic disorder when growing up in an urban environment

It's one thing to state this, and another thing to suggest ways to test this complex interaction of risk factors.  They do just this, however, describing a number of animal studies that can now be done to look at the effect of environmental factors on the developing brain of susceptible and non-susceptible animals.  The authors recognize that this won't be easy -- sensitive periods during development, risk factors, at-risk animals, and evidence of psychosis all must be well-defined and observed.

And then any potentially disastrous effects of methodological shortcomings, such as bias or confounding -- including genetic confounding -- must be ruled out before results can be considered credible.  E.g., are at-risk adolescents self-medicating with cannabis, making it appear that cannabis is the trigger when it's something intrinsic instead?  Does genetic susceptibility predispose to the use of cannabis, again making it look as though cannabis is the risk factor when it's genes?  They in fact include a box devoted to issues related to weighing the evidence.

Whether van Os et al. are correct in the details of which factors are most important to account for, or in the timing of the sensitive periods for specific aspects of normal growth we certainly can't say, but even so this is a beautifully nuanced and well-reasoned argument for accepting complexity, with suggestions for how to move forward from there, all based on the mountains of data that have come before.  And, the authors don't over claim, or say it will be easy. 

Indeed, the existence of many 'sensitive stages' during development means any number of pathways could be affected at any time to lead to disease.  Schizophrenia, like any trait, of course has genetic underpinnings, but given the 4-dimensional complexity of the developmental pathways in the brain,  this means that there are many ways that things could go awry, which only increases the difficulty with which they can be found, or, if found, be useful for prediction.

The authors conclude:

The human brain has evolved as a highly context-sensitive system, enabling behavioural flexibility in the face of constantly changing environmental challenges. There is evidence that genetic liability for psychotic syndrome is mediated in part by differential sensitivity to environments of victimization, experience of social exclusion and substances affecting brain functioning, having an impact during development. Given the complexity of the phenotype and evidence of dynamic developmental trajectories, with environmentally sensitive periods, longitudinal research on gene–environment interplay driving variation in behavioural expression of liability, that subsequently may give rise to more severe and more ‘co-morbid’ expressions of psychopathology and need for care, is required to identify the causes and trajectories of the psychotic syndrome. Examination of differential sensitivity to the environment requires technology to assess directly situated phenotypes indexing dynamic, within-person environmental reactivity as substrate for molecular genetic studies; parallel multidisciplinary translational research, using novel paradigms, may help identify underlying mechanisms and point the way to possible interventions.

So, this is an unusually sober and realistic treatment of a complex disease.  Is there anything surprising here?  Only that Nature is giving a number of pages to a nuanced treatment of a subject that is so often treated simply.

Genetic engineering

We often criticize the excess geneticization of diseases whose main cause is not genetic variation but lifestyle factors of various kinds. But some diseases seem clearly to be genetic, with little environmental input, one or only a few clearly known causal genes. In such cases, genetics is the right approach, and genetics of two kinds--to detect risk factors in genetic counseling for parents planning to have children who know a serious allele is in their family, and to treat the disease when it has arisen.

A good example described in last Sunday's New York Times is epidermolysis bullosa. The disease appears to be due to a defect in a collagen gene that produces structural strength and integrity to the skin. The victims have skin described as being as delicate as butterfly wings, and as a result have very compromised lives.

EB seems to be a perfect target for gene therapy--to replace the gene in the germline of parents, or of fertilized eggs in vitro before reimplantation in the mother, or therapeutically to replace the deficient skin cells with cells competent to produce the right type of collagen. Such efforts are described in the article.

Human beings are very good at engineering, and if science is good at anything, it's technology. EB is one of many problems that seem to be engineering rather than conceptual challenges. Science ought to work, in such cases--even if that doesn't mean it can happen overnight. This kind of challenge is where genetic investment should go, in our view.

For complex diseases that are mainly due to environmental or lifestyle factors, if those were ameliorated by social behavior (like better diet) and other measures (like removal of toxins), then for most diseases what would remain would be the truly genetic instances that would fortunately be rare, and fortunately be engineering challenges.

That doesn't mean it'll be easy. If we're good at engineering and have had thousands of years to practice it, if there's one thing that organisms have evolved over countless more thousands, it's to detect and prevent the outside world from attacking its cells. So these will be battles waged mano a mano at the molecular scale.

Many techniques already exist to replace genes, engineer vectors to put genes into cells, or make microorganisms (or culturable cells) produce a gene product. The best approach, perhaps, is to engineer stem cells from the affected person, redifferentiated to be of the needed tissue type, and then somehow introduce them into the affected tissue. There's a lot of progress along these lines, but only time will tell if this is the best approach. Whatever turns out to be the case, at least these are clear-cut problems for which technological solutions seem at least possible.

Genetics in clinical practice--the tail end of the spectrum?

One of the more interesting talks at the Bristol meeting, perhaps because it's an area we don't know much about, had to do with clinical applications of genetics. The question concerned when, where, and how genetic information can be useful in the clinic. But the issues go much deeper than that, and relate to main points in our 'Mermaid' book.

The idea presented was that genetic information can be useful, not so much for disease risk prediction, because genotypes have poor predictive power for 'complex' traits (that is, and somewhat circularly, those that are not due to a single, and hence predictive, gene!). Instead, the argument was that genotyping can help determine therapy in the diagnostic sense of determining which of many causes may apply to, and hence guide therapy for, individual cases of a given disease.

An example given was MODY, or 'maturity onset diabetes of the young. MODY runs in families and can mimic both type 1 and type 2 diabetes, though most often it is a mild form of T1D, but with patients continuing to make some insulin (T1D, or juvenile diabetes is due to a failure to make insulin; T2D, or adult diabetes, is failure of cells to respond to insulin). The speaker, Dr Andrew Hattersley, described one family in which each affected member was being treated differently, but once the causal gene was known, each member was treated more appropriately, and the disease then better controlled. This was very good to hear because all too often even knowing the causal gene can't inform treatment--Huntingdon's disease comes to mind.

Hattersley also discussed a gene that causes three forms of achondroplasia, or dwarfism, but the particular mutation in that gene determines whether an individual will be somewhat shorter than normal, or will be much more severely affected, and knowing the mutation during pregnancy helps to prepare the family and clinician.

He primarily discussed single gene disorders, although these are often disorders that can be caused by a number of different mutations or even different genes, but in the instances described, knowing the specific mutation can make the difference between useful medical intervention and none. The idea is essentially that the 'same' disease is not really the same in different individuals if enough specificity is known, and that when a gene's effects are understood, knowing the causal gene is clinically relevant.

By contrast, all too often in human genetics identifying genes whose variation is statistically associated in at least some studies with the occurrence of disease contributes nothing toward treatment, because the statistical connection is too weak to be useful. That is the problem that is now widely recognized in regard to GWAS (see our many earlier posts) for complex diseases, so we were heartened to learn that this is not always the case. In some cases, knowing the physiological pathway is very useful.

The much broader relevance of these points is that they relate to genetic determinism: the degree and specificity to which genes determine phenotypes (the same applies to environmental factors). Only to the extent that genes determine traits in organisms can knowing the genotype be used to predict the trait. That has everything to do with views, accurate or hyped, about the usefulness of huge genotyping studies to health. It has to do with the use of genetics in social behavioral and other societal contexts. And it has to do with the origins of traits and with evolution itself.

To the degree that natural selection molds and creates (produces) the traits we see in organisms--and that it does so is at the heart of Darwinian theory--genes must determine traits. That's because if genotypes do not determine the trait, selection for or against variation in the trait won't affect the frequency of genetic variation, and hence won't guide genetic evolution. Yet the modern Darwinian evolutionary theory is entirely based on genetic determinism.

With selection on phenotypes, not genotypes, if the connection between genotype and phenotype is weak, selection only weakly affects the relative success ('fitness') of genotypes, and their frequency changes over the generations will largely be the result of chance, not selection.

When a mutation strongly affects a trait, and the trait is important to organisms' success, then Darwinian theory works just fine as advertized. But this is not so if traits are so genotypically complex that selection hardly affects an individual gene.

There is a middle ground. If a trait, like a disease, is a strong reflection of a specific genotype, but many different genotypes can generate similar traits, then prediction can be weak but once the trait has arisen it can be useful to determine which of those causes is responsible. From an evolutionary point of view, if each instance of a trait is due only to variation in a single gene, then selection can affect the frequency of that variation.

The key difference is that for complex traits the phenotype is thought to be the simultaneous result of variation at many different genes. There are so many combinations of genotypes that can generate similar phenotypes, that one can't usefully predict the latter from the former--or, perhaps, not even go back the other way, unless variation at only one or a few genes is responsible in a given instance.

Thus, prediction in development, medicine, or evolution all depend centrally on the degree to which genes individually, or in aggregate, determine the traits you bear in life. In the 'tail' end of the causal spectrum, where specific variants exert strong effects, things work out nicely.