Clues From the Cloud
Schizophrenia is a uniquely human, enormously complex neurodevelopmental disorder (1, 2), characterized by a combination of genetic susceptibility factors and environmental insults that converge on a developmental timeline (3, 4). The recent emergence of high-throughput, unbiased genetic assessment methods such as genome-wide association studies (GWAS), combined with assessments of several thousand patients have provided us with a better understanding of the underlying genetics of schizophrenia (5). It appears that two often-interrelated genetic mechanisms are at work. The first mechanism encompasses rare, almost individually specific genome changes where small chromosomal regions are either deleted or duplicated (referred to as copy number variants [CNVs]) (6) and have strong disease-predisposing effects (7). The second mechanism is related to common variants in our genome that are found in both healthy individuals and those with disorders, but with slightly different frequencies. On their own, each of these common single-nucleotide polymorphisms (SNPs) are only small contributors to the emergence of the disease, but in large numbers and in the right combinations they can explain a significant proportion of the genetic underpinnings of the disease (5).
Many previously published study results suggest that genetic susceptibility, either from SNPs or CNVs, converge with environmental insults and jointly alter the gene expression (mRNA production) of the developing brain (3). This tips the balance of neurochemistry and alters the connectivity and information processing of the brain, ultimately resulting in complex behavioral changes, which we recognize as a diagnosis (8). This mechanism holds true not only for schizophrenia but for a host of other neurodevelopmental disabilities, including autism spectrum disorders (ASD).
Several years ago, a group of investigators led by Dr. Joel Kleinman created an interesting, and much underappreciated, data set, termed BrainCloud (9). This resource cataloged mRNA expression levels from 269 human postmortem dorsolateral prefrontal cortical samples ranging from fetal life up to 80 years of age, providing us with a glimpse into the molecular processes that shape the emergence of our cognitive abilities. The present study of Birnbaum et al. (10) in this issue of the Journal further builds on this data set, attempting to link brain development to neuropsychiatric diseases. The central premise of their study is simple, but often underappreciated: if we want to understand uniquely human neuropsychiatric and neurodevelopmental disorders, a significant set of clues will come from understanding human brain development itself (11).
Birnbaum et al. (10) first defined genes with “fetal effect,” and found that 25% of the more than 30,000 gene probes revealed transcript levels that were overexpressed or underexpressed in fetal life compared with the postnatal period. Then, they defined sets of risk genes that have been implicated as disease-predisposing for six neuropsychiatric disorders, including syndromic neurodevelopmental disorders, ASD, intellectual disability, schizophrenia, bipolar affective disorder, and neurodegenerative disorders. The authors then employed various statistical approaches as they explored the expression of these disease-predisposing gene sets in the BrainCloud data set. The outcome of this experiment was quite revealing: the authors observed increased fetal expression of disease-predisposing genes for syndromic neurodevelopmental disorders, intellectual disability, and ASD. In contrast, a statistically significant fetal underexpression of transcripts was reported for the gene set associated with neurodegenerative disorders. In a secondary analysis, the authors found that the ASD findings were a result of enrichment in transcripts originating from rare variants (CNVs) rather than the common variants (SNPs). Surprisingly, the group of schizophrenia or bipolar disease-predisposing genes (originating either from SNP, CNV, or exome sequencing studies) did not show enrichment in either direction. However, although the majority of schizophrenia susceptibility genes did not show fetal effects, these genes were still found to be important for neural development: gene ontology classification (12) revealed that they belonged to gene pathways related to “nervous system development” and “neuron differentiation.” Furthermore, within each CNV locus associated with schizophrenia, one or two genes were enriched in prenatal abundance, suggesting that they might be of particular relevance for disease-associated fetal pathogenesis.
Every excellent study raises a host of new questions and lays the foundation of new experiments, and the research endeavor of Birnbaum et al. is not different in this regard. First, this study can be considered a proof of principle endeavor, with limited resolution. The transcriptome of different brain cell types is unique, and we do not know if the observed correlations are limited to certain subpopulations of cells or if they are characteristic of multiple brain regions. Regional transcriptomes are complexly regulated, and combining the approach of the present study with high-quality spatio-temporal data sets and analyses (13–15) will further increase the resolution and specificity of these findings.
Second, it is surprising that present findings do not support a fetal effect associated with schizophrenia-predisposing genes derived from GWAS, CNV analysis, or exome sequencing of rare variants. However, the lack of this association should not be interpreted as proof that schizophrenia is not a neurodevelopmental disorder: the findings may suggest technical limitations of including inappropriately large or not well-established chromosomal regions in the study. Furthermore, we cannot exclude the possibility that the fetal effect is present in other regions (e.g., other neocortical, hippocampal, or subcortical areas), or that the schizophrenia-associated neurodevelopmental processes may be transcriptome-independent. Differentiating between these possibilities should be a focus of further studies.
Third, the authors found a surprising signature for genes involved in neurodegeneration. This is a very intriguing finding, and it suggests a dichotomy between neurodevelopmental and neurodegenerative mechanisms from the early onset: disruption of genes with enriched fetal expression leads to diseases that manifest with symptoms in early postnatal life, while the low expression of the neurodegenerative genes during prenatal development “postpones” the pathophysiology of neurodevelopmental disorders to later postnatal time periods. Simply, one can argue that disruption of genes with high expression during fetal life affects “brain construction,” while genes expressed at low levels prenatally are responsible for “brain maintenance” later in life.
The study also highlights the urgent need for better bioinformatics tools. We are generating GWAS, DNA and RNA sequence, and gene expression data at an amazing speed and volume. While tools for data mining of specific data types are developing at a reasonable pace, the tools and standards of integrating data sets from different sources significantly lag behind (16). Without them we are not able to observe the meaningful relationships between data sets, and the generated data remains only moderately informative.
Does the present study uncover the essence of schizophrenia, autism, or other neuropsychiatric disorders? Of course not. Over the last decades we have learned that nothing is ever conclusive in schizophrenia or ASD research. There is no smoking gun. There are no simple interpretations. But every day we are gathering more pieces of this enormous jigsaw puzzle, and the studies by Birnbaum et al. might just represent one of the important corner pieces. Recently, when visiting Cold Spring Harbor Laboratories I asked Dr. James Watson if he was aware of the enormity of the DNA discovery at the time when it was made. After a short pause and a walk back down memory lane, he responded with a smile: “We were not thinking about that.” Perhaps we should learn from this example and focus on accumulating high-quality data, noticing the important relationships rather than trying to solve the disease. And over time, when we have enough puzzle pieces from the many various research domains, the picture of schizophrenia and other disorders will emerge. But I admit: it is so hard to wait!
1 : Schizophrenia as a disorder of neurodevelopment. Annu Rev Neurosci 2002; 25:409–432Crossref, Medline, Google Scholar
2 : Implications of normal brain development for the pathogenesis of schizophrenia. Arch Gen Psychiatry 1987; 44:660–669Crossref, Medline, Google Scholar
3 : Schizophrenia as a disorder of molecular pathways. Biol Psychiatry (Epub ahead of print, Jan 10, 2014)Google Scholar
4 : Immune system disturbances in schizophrenia. Biol Psychiatry 2014; 75:316–323Crossref, Medline, Google Scholar
5 ;
6 : CNVs: harbingers of a rare variant revolution in psychiatric genetics. Cell 2012; 148:1223–1241Crossref, Medline, Google Scholar
7 : A polygenic burden of rare disruptive mutations in schizophrenia. Nature 2014; 506:185–190Crossref, Medline, Google Scholar
8 : Breaking the gene barrier in schizophrenia. Nat Med 2009; 15:488–490Crossref, Medline, Google Scholar
9 : Temporal dynamics and genetic control of transcription in the human prefrontal cortex. Nature 2011; 478:519–523Crossref, Medline, Google Scholar
10 : Prenatal expression patterns of genes associated with neuropsychiatric disorders. Am J Psychiatry 2014; 171:758–767Link, Google Scholar
11 : The developmental transcriptome of the human brain: implications for neurodevelopmental disorders. Curr Opin Neurol 2014; 27:149–156Crossref, Medline, Google Scholar
12 ;
13 : Spatio-temporal transcriptome of the human brain. Nature 2011; 478:483–489Crossref, Medline, Google Scholar
14 : Temporal specification and bilaterality of human neocortical topographic gene expression. Neuron 2014; 81:321–332Crossref, Medline, Google Scholar
15 : Functional and evolutionary insights into human brain development through global transcriptome analysis. Neuron 2009; 62:494–509Crossref, Medline, Google Scholar
16 : DAWN: a framework to identify autism genes and subnetworks using gene expression and genetics. Mol Autism 2014; 5:22Crossref, Medline, Google Scholar