John A. McGrath and I (John J. McGrath—no relation) welcome any other like-named or like-minded researchers to contribute further to the quest to resolve the heterogeneity of the poorly understood group of brain disorders currently lumped under the label “schizophrenia.” This interim label has tenaciously guided our research efforts for decades, despite the fact that the research community is well aware of its deficiencies. Like intellectual fly-paper, the problems with this diagnostic label have probably shortened the life of many able researchers. Can we ever hope to fractionate the underlying disorders into more meaningful groupings?

John A. McGrath and colleagues from Baltimore have looked for heritability of a broad range of symptoms and measures of disability. All factors were found to be heritable, but some were more heritable than others. This is an interesting outcome. The challenge for the research community will be what to do next.

I would like to add a few friendly comments in order to stimulate debate. These comments reflect my personal biases. I am happy to acknowledge uncertainty around these comments and happy to revise my opinions in the face of data (hint—add your dissenting comments below so we can get some debate going).

1. Do we expect that the genetic architecture of schizophrenia will map obediently onto surface level phenotypes? With our current limited knowledge of developmental neurobiology and systems neuroscience, how optimistic should we be about linking factor-derived surface-level phenotypes with upstream gene variation? Would researchers interested in the genetic architecture underlying asthma gain much additional traction if they divided up cases into high-pitched versus low-pitched wheezing, polyphonic versus monotonic wheezing, wheezing with frequent versus infrequent coughing, etc. I am not arguing that this is not worth doing—it may provide important clues. We keep looking for the “low hanging fruit”—but is this realistic?

2. Can we enrich the phenotypes? Would the factors be more informative if they also included laboratory-based variables (e.g., brain structure, electrophysiology, eye movement, niacin flush, etc.) (Hallmayer et al., 2003; Price et al., 2006)? Would the “gain be worth the pain”? What do experts think?

3. How do we interpret the “scholastic” factor? This factor is particularly interesting in light of the convergent evidence from schizophrenia epidemiology. A range of population-based cohort studies have shown that individuals who go on to develop schizophrenia are more likely to have premorbid neurocognitive impairment (Maccabe, 2008; Welham et al., 2009). However, this intermediate phenotype is associated with a very wide range of adverse clinical and educational outcomes. On the one hand, this non-specificity is a cause for celebration—if we can find public health measures that can rescue these vulnerable brains, we may be able to avert much more disability than that only related to schizophrenia (Rose, 1992). On the other hand, it is a gloomy prospect if we acknowledge that there are an infinite number of ways to disrupt optimal brain development, and that reduced cognitive reserve is probably an intermediate phenotype of a wide range of clinical conditions (leaving aside educational and economic outcomes).

4. Where to now? Pulver and colleagues have already linked one of the factors with a candidate gene (Chen et al., 2009). Should we invest time in reclassifying existing or new patient groups based on surface level phenotypes and keep looking for new clues? This will be important, but I argue we know too little about normal brain function to be able to build realistic models linking genetic factors and surface level phenotypes. We should keep trying to build models based on the imperfect existing knowledge, but mostly these will be wrong. The best investment for our field would be to use the clues from genetics and risk factor epidemiology to help drive more neuroscience discovery (McGrath and Richards, 2009). The more knowledge we have about the brain, the less ridiculous our models will be. The faint clues that may emerge from genetic studies using taxons based on surface level phenotypes could be important catalysts for discovery in basic neuroscience. We have decades of work ahead of us, but the science is tractable.

References:

Chen, P. L., D. Avramopoulos, et al. (2009). Fine mapping on chromosome 10q22-q23 implicates Neuregulin 3 in schizophrenia. Am J Hum Genet 84(1): 21-34. Abstract

Hallmayer, J. F., A. Jablensky, et al. (2003). Linkage analysis of candidate regions using a composite neurocognitive phenotype correlated with schizophrenia. Mol Psychiatry 8(5): 511-23. Abstract

Maccabe, J. H. (2008). Population-based cohort studies on premorbid cognitive function in schizophrenia. Epidemiol Rev 30: 77-83. Abstract

McGrath, J. A., D. Avramopoulos, et al. (2009). Familiality of novel factorial dimensions of schizophrenia. Arch Gen Psychiatry 66(6): 591-600. Abstract

McGrath, J. J. and L. J. Richards. (2009). Why schizophrenia epidemiology needs neurobiology--and vice versa. Schizophr Bull 35(3): 577-81. Abstract

Price, G. W., P. T. Michie, et al. (2006). A multivariate electrophysiological endophenotype, from a unitary cohort, shows greater research utility than any single feature in the Western Australian family study of schizophrenia. Biol Psychiatry 60(1): 1-10. Abstract

Rose, G. (1992). The Strategy of Preventive Medicine. Oxford, Oxford University Press.

Welham, J., M. Isohanni, et al. (2009). The antecedents of schizophrenia: a review of birth cohort studies. Schizophr Bull 35(3): 603-23. Abstract

View all comments by John McGrath

While a number of genetic loci that appear to represent an increased susceptibility to schizophrenia have been identified, linkage, candidate gene and whole genome scan approaches have been largely unsuccessful in identifying these genes in any consistent way. This is partly because the clinical characteristics of schizophrenia are very far along the pathophysiological chain that extends from genes, through proteins, neurons, cognition, behavior, and symptoms, and finally to the DSM-IV construct of schizophrenia. Phenotypes closer to the gene effects in the pathophysiological pathway are better placed to help to untangle the genetic components of the illness.

McGrath and colleagues analyses of 73 indicators resulting in nine sign and symptom factors, or dimensions, could provide alternative phenotypes for molecular genetic studies in place of the DSM-IV construct of schizophrenia, assuming, of course, that twin and adoption studies show that they are heritable. McGrath and colleagues used a large number of core symptoms of schizophrenia and additional indicators of social, work, and educational dysfunction. It is the first study to demonstrate familial aggregation of child/adolescent sociability, scholastic performance, disability/impairment, and prodromal factors. These factors could be used to refine the schizophrenia phenotype; stratifying patients by their factor scores might help to identify the risk alleles that increase liability to this disorder.

View all comments by Timothea Toulopoulou