A broad swath of neuropsychiatric conditions, including autism, depression and schizophrenia, share genetic underpinnings. But individual variants can have strikingly different effects on clinical diagnoses: A variant that increases the likelihood of one condition may reduce that of another, according to two new preprints.
“When you delete and duplicate genes, you get very different, often diametrically opposed, outcomes,” says Jonathan Sebat, professor of psychiatry and cellular biology at the University of California, San Diego and co-principal investigator of the studies. He and his colleagues analyzed whole-genome sequencing data from 574,965 people, collected at dozens of institutions as part of the Psychiatric Genomics Consortium, a massive international project aimed at finding common variants in psychiatric conditions.
The work, posted on medRxiv in July, revealed nearly three dozen links between copy number variants (CNVs)—deletions or duplications of small segments of DNA—and any of six different neuropsychiatric conditions: autism, attention-deficit/hyperactivity disorder (ADHD), bipolar disorder, major depressive disorder, post-traumatic stress disorder (PTSD) and schizophrenia.
Nearly all of these rare variants are tied to more than one diagnosis, one of the studies shows. But duplications and deletions can have different and sometimes opposite effects: If a deletion raises a condition’s likelihood, the corresponding duplication may lower it, and vice versa.
Autism was unique in that no deletions or duplications decreased autism probability, whereas some variants lowered the chances of having schizophrenia or one of the other conditions, Sebat says.
“What was surprising is that sometimes a gene duplication can be protective,” says Hyejung Won, associate professor of genetics at the University of North Carolina School of Medicine, who did not contribute to the studies. “That provides a lot of food for thought. How does excessive expression of a gene protect you from another disorder?”
The six neuropsychiatric conditions analyzed in the first study converge on a common set of molecular pathways but are associated with distinct cell types, brain regions and gene dosages, according to the other study.
“The overlapping biology of these disorders might not be quite as overlapping as we would have thought,” says Jason Lerch, professor of neuroscience at the University of Oxford, who did not contribute to the research. “This is important stuff.”
B
ecause the list of thousands of genes linked to neuropsychiatric conditions doesn’t say much about how those genes affect psychiatric traits, Sebat and his colleagues turned to deletions and duplications, which predictably alter gene dosage—an approach akin to studying gain- or loss-of-function variants, he says.Genome-wide association studies involving participants grouped according to six psychiatric diagnoses revealed a total of 35 associations between rare CNVs and least one of the conditions. In the case of schizophrenia, for example, deletions in the 22q11.21 A-D chromosomal region significantly increase likelihood, whereas duplications decrease it. And the effect sizes of schizophrenia-linked reciprocal deletions and duplications are inversely correlated, indicating a linear dose-dependent relationship between gene copy number and a person’s chances of having schizophrenia.
A similar pattern held for other chromosomal regions and diagnostic categories, but not all. Autism was a notable exception because no CNV decreased the likelihood of having the condition—perhaps because autism encompasses a particularly broad spectrum of traits, Sebat says.
The second study identified gene sets based on the genes’ roles in molecular pathways, along with cell types, brain regions and gene dosage (duplication or deletion), and then determined which genes sets correlated with which psychiatric conditions by way of a gene-set burden analysis.
All of the 35 CNVs flagged in the first study act on a small set of molecular pathways, such as MAPK signaling, synaptic transmission or chromatin regulation, the second study showed. “All of the same 19 canonical pathways are involved in all disorders,” Sebat says. And separating these pathways by cell type and brain region in addition to gene dosage showed further distinctions among conditions, he says.
This multidimensional analysis provided three unique factors, or signatures, that captured the neurobiological processes involved in each condition. Schizophrenia, for example, is strongly associated with factor 1, a set of genes involved in chromatin regulation and MAPK signaling, specifically in postnatal excitatory neurons, microglia and vascular cells. And it is linked primarily to duplications in cell signaling and metabolic pathways in postnatal excitatory neurons and neurovascular cells, as well as to deletions in inhibitory interneurons.
This pattern supports the hypothesis that schizophrenia is associated with an imbalance in excitation and inhibition, Sebat says.
Autism is also strongly correlated with factor 1—a finding that aligns with past studies that point to microglia and postnatal excitatory neurons, as well as chromatin remodeling. But both duplications and deletions linked to autism act on the same pathways and cell types in most cases. In addition, deletions that increased autism likelihood were linked to a decrease in schizophrenia risk, the study showed.
Factor 2, which is associated with mood disorders, isn’t strongly tied to specific pathways and involves changes in non-neuronal cells. Factor 3 separates conditions by brain region: In bipolar disorder, for instance, deletions are concentrated in frontotemporal regions, whereas in ADHD, depression and schizophrenia, they are more common in sensory regions.
T
he directional effects of duplications and deletions are interesting, says Mark Gerstein, professor of biomedical informatics at Yale University, who was not involved in the research. But duplications and deletions do not lead only to increases or decreases in protein expression, he says. In practice, both can also lead to novel protein functions or disruptions that weren’t captured in the study’s analysis.Sebat contends that the general equivalence between gene copy number and protein expression holds, but he adds that deletions and duplications can both disrupt gene functions in unpredictable ways. “There’s still a lot we don’t know,” he says.
Ultimately, the findings could one day help clinicians distinguish different clinical presentations of heterogeneous diagnoses such as autism, Sebat says.
Won agrees. If this line of study can identify “some disease specificity,” she says, “we can diagnose individuals with better genetic etiology and find tailored therapeutic options.”