NEWS & VIEWS G E NETICS O F D ISEASE
Associations with depression Two genetic regions associated with major depressive disorder have been revealed for the first time, through whole-genome sequencing of a population of Han Chinese women. PAT R I C K F. S U L L I VA N
O
f all complex human illnesses, major depressive disorder (MDD) has arguably proved the trickiest to understand. Despite decades of research, there is little certainty about its biological basis, in part because genetic clues to its aetiology have been hard to find1. The combination of relatively high prevalence and relatively low heritability seems to indicate that MDD does not lend itself to genetic analysis, although the genetic dissection of type 2 diabetes mellitus, which has a similar prevalence and heritability, has been much more productive. In a paper published on Nature’s website today, the CONVERGE consortium2 identifies the first two longawaited genetic associations for MDD. Our ignorance about MDD is in marked contrast to its impact on people and public health3. The disease is common, costly and associated with high rates of morbidity and mortality. As such, it stands to reason that this research is exciting for those who study MDD. But it also exemplifies a sometimes neglected issue — how an informed approach to improving the definition of a complex illness can lead to success where other approaches have failed. Why is defining MDD so complicated? Sadness is normal and integral to the human condition. However, much too frequently, sadness becomes pervasive, persistent, unshakable and associated with signs and symptoms characteristic of MDD, such as changes in sleeping habits, appetite and cognition, and the onset of suicidal tendencies. But where should we draw the line between normality and pathology? This question is echoed throughout medicine, for example when using normal fasting blood glucose levels to delineate normal physiology from that of type 2 diabetes, or when separa ting normal blood pressure from hypertension. The difference is that the measures for these latter two conditions are more objective than those for assessing ‘sadness’, and the consequences of each disease more readily assessed. There is no laboratory test that will help us to
know when sadness becomes MDD. The CONVERGE consortium authors reasoned that the core issue hampering the discovery of MDD-associated genes is heterogeneity — in a group of people who all have the same MDD symptoms, the aetiology of the MDD may in fact be different. Some people might have a highly genetic form of the disease, whereas in others, MDD may be brought on by environmental factors such as poverty, physical or sexual abuse, or an unhealthy lifestyle. Still others may have a primary problem such as alcoholism, of which MDD is a secondary consequence. This long-held concept of heterogeneity has made defining ‘true’ An informed MDD something of approach to a holy grail. improving the The researchers definition of a made a set of intelcomplex illness ligent decisions can lead to when defining who success where to study. They reaother approaches soned that morehave failed. severe cases would have a clearer and less-complex genetic signal — an approach widely used in human genetics to minimize heterogeneity in complex illnesses. Unlike more-inclusive approaches4, the consortium authors implemented several measures that they thought would maximize their chances of success. They worked in China, where the prevalence of MDD is lower than that in the United States or Europe, studied only women and selected relatively severe cases, in which the women had experienced two or more episodes of MDD, using psychiatric inpatient and outpatient facilities. Unusually, they genotyped their samples using low-coverage whole-genome sequencing (genotyping determines the identity of genetic variations across the genome). They identified two regions in which genetic changes, or variants, are associated with MDD — one near the SIRT1 gene and the other in an intron (a non-protein-coding region) of the gene LHPP.
doi:10.1038/nature14635
The typical approach to genotyping in human genetics is to use an array containing a fixed set of between 500,000 and 1 million genetic markers — DNA variants at known chromosomal locations. To my know ledge, this is the only published study in which genotyping involved low-coverage sequencing of the whole genome. Because of decreases in the costs of genotyping arrays, it may be one of the last. The authors’ low-coverage sequencing had relatively high error rates — around 2% of the genetic variants that they identified could not be replicated with a different method, compared with less than 0.5% for an inexpensive array. Moreover, despite their wish to gain traction on MDD-causing genetic associations that have yet to be described in China, the two variants that they found have been in standard databases for a decade. This is because genetic variants such as these, which are common in China, are evolutionarily old, and so likely to be found across the globe. Wisely, the authors confirmed the variants with a second method and replicated the results in an independent sample, ensuring that their associations meet typical standards for significance and replication. The consortium suggests that the proximity of one of the variants to SIRT1 implicates abnormalities in mitochondria (the cell’s energy-producing centres) in MDD, because one role of the SIRT1 protein is to regulate mitochondrial function. If this assertion holds up, it is likely to imply that many other genetic variants are involved in altered mitochondrial function and have associations with MDD that near the threshold for significance. A standard way to evaluate such a hypothesis is to investigate whether a specific genetic pathway, such as that involving the genes that affect mitochondrial function, is statistically more likely to have smaller P-values for association with MDD than expected by chance. This analysis was not reported in the current study. A previous systematic pathway analysis of MDD and other major psychiatric disorders did not implicate mitochondrial biology5. As such, although the researchers’ hypothesis is intriguing, it requires replication, extension, integrated analysis and more biological evidence. I wish the authors had formally tested their fundamental premise, namely that their sample was more homogeneous than those studied previously. If that is the case, then the heritability of the common variants (the proportion of variance contributing to MDD that can be accounted for by the genetic variation they measured) should be notably high. But although several methods exist to check this6,7, the authors did not report such an analysis. More unsettling, the two variants identified | NAT U R E | 1
© 2015 Macmillan Publishers Limited. All rights reserved
RESEARCH NEWS & VIEWS have almost no association signal in samples taken from European populations8. The reasons for this discrepancy are unclear. Perhaps these variants are truly causative for MDD only in severe cases from China. However, many other common associations for complex illnesses hold across the world. The authors tested the comparability of their findings with European samples and found some evidence of overlap, but more-refined analyses would be of keen interest. This first identification of replicable, significant genome-wide associations for MDD is exceptional. Although further work is required, it is to be hoped that these results will provide therapeutic targets for MDD. The
drug-discovery pipeline for MDD has never been based on solid biological foundations, but the work begun here could improve the focus of the field. The authors’ study marks the beginning of the beginning for the genetic dissection of MDD. The CONVERGE consortium has provided an excellent starting point for what should be an intriguing voyage of discovery. ■ Patrick F. Sullivan is in the Departments of Genetics and Psychiatry, University of North Carolina, Chapel Hill, North Carolina 275997264, USA, and in the Department of Medical Epidemiology and Biostatistics, Karolinska
2 | NAT U R E |
© 2015 Macmillan Publishers Limited. All rights reserved
Institutet, Stockholm, Sweden. e-mail: [email protected] 1. Levinson, D. F. et al. Biol. Psychiatry 76, 510–512 (2014). 2. CONVERGE consortium. Nature http://dx.doi. org/10.1038/nature14659 (2015). 3. Whiteford, H. A. et al. Lancet 382, 1575–1586 (2013). 4. Kendler, K. S. Arch. Gen. Psychiatry 54, 299–304 (1997). 5. The Network and Pathway Analysis Subgroup of the Psychiatric Genomics Consortium. Nature Neurosci. 18, 199–209 (2015). 6. Cross-Disorder Group of the Psychiatric Genomics Consortium. Nature Genet. 45, 984–994 (2013). 7. Bulik-Sullivan, B. K. et al. Nature Genet. 47, 291–295 (2015). 8. Major Depressive Disorder Working Group of the Psychiatric GWAS Consortium. Mol. Psychiatry 18, 497–511 (2013).
Associations with depression Two genetic regions associated with major depressive disorder have been revealed for the first time, through whole-genome sequencing of a population of Han Chinese women. PAT R I C K F. S U L L I VA N
O
f all complex human illnesses, major depressive disorder (MDD) has arguably proved the trickiest to understand. Despite decades of research, there is little certainty about its biological basis, in part because genetic clues to its aetiology have been hard to find1. The combination of relatively high prevalence and relatively low heritability seems to indicate that MDD does not lend itself to genetic analysis, although the genetic dissection of type 2 diabetes mellitus, which has a similar prevalence and heritability, has been much more productive. In a paper published on Nature’s website today, the CONVERGE consortium2 identifies the first two longawaited genetic associations for MDD. Our ignorance about MDD is in marked contrast to its impact on people and public health3. The disease is common, costly and associated with high rates of morbidity and mortality. As such, it stands to reason that this research is exciting for those who study MDD. But it also exemplifies a sometimes neglected issue — how an informed approach to improving the definition of a complex illness can lead to success where other approaches have failed. Why is defining MDD so complicated? Sadness is normal and integral to the human condition. However, much too frequently, sadness becomes pervasive, persistent, unshakable and associated with signs and symptoms characteristic of MDD, such as changes in sleeping habits, appetite and cognition, and the onset of suicidal tendencies. But where should we draw the line between normality and pathology? This question is echoed throughout medicine, for example when using normal fasting blood glucose levels to delineate normal physiology from that of type 2 diabetes, or when separa ting normal blood pressure from hypertension. The difference is that the measures for these latter two conditions are more objective than those for assessing ‘sadness’, and the consequences of each disease more readily assessed. There is no laboratory test that will help us to
know when sadness becomes MDD. The CONVERGE consortium authors reasoned that the core issue hampering the discovery of MDD-associated genes is heterogeneity — in a group of people who all have the same MDD symptoms, the aetiology of the MDD may in fact be different. Some people might have a highly genetic form of the disease, whereas in others, MDD may be brought on by environmental factors such as poverty, physical or sexual abuse, or an unhealthy lifestyle. Still others may have a primary problem such as alcoholism, of which MDD is a secondary consequence. This long-held concept of heterogeneity has made defining ‘true’ An informed MDD something of approach to a holy grail. improving the The researchers definition of a made a set of intelcomplex illness ligent decisions can lead to when defining who success where to study. They reaother approaches soned that morehave failed. severe cases would have a clearer and less-complex genetic signal — an approach widely used in human genetics to minimize heterogeneity in complex illnesses. Unlike more-inclusive approaches4, the consortium authors implemented several measures that they thought would maximize their chances of success. They worked in China, where the prevalence of MDD is lower than that in the United States or Europe, studied only women and selected relatively severe cases, in which the women had experienced two or more episodes of MDD, using psychiatric inpatient and outpatient facilities. Unusually, they genotyped their samples using low-coverage whole-genome sequencing (genotyping determines the identity of genetic variations across the genome). They identified two regions in which genetic changes, or variants, are associated with MDD — one near the SIRT1 gene and the other in an intron (a non-protein-coding region) of the gene LHPP.
doi:10.1038/nature14635
The typical approach to genotyping in human genetics is to use an array containing a fixed set of between 500,000 and 1 million genetic markers — DNA variants at known chromosomal locations. To my know ledge, this is the only published study in which genotyping involved low-coverage sequencing of the whole genome. Because of decreases in the costs of genotyping arrays, it may be one of the last. The authors’ low-coverage sequencing had relatively high error rates — around 2% of the genetic variants that they identified could not be replicated with a different method, compared with less than 0.5% for an inexpensive array. Moreover, despite their wish to gain traction on MDD-causing genetic associations that have yet to be described in China, the two variants that they found have been in standard databases for a decade. This is because genetic variants such as these, which are common in China, are evolutionarily old, and so likely to be found across the globe. Wisely, the authors confirmed the variants with a second method and replicated the results in an independent sample, ensuring that their associations meet typical standards for significance and replication. The consortium suggests that the proximity of one of the variants to SIRT1 implicates abnormalities in mitochondria (the cell’s energy-producing centres) in MDD, because one role of the SIRT1 protein is to regulate mitochondrial function. If this assertion holds up, it is likely to imply that many other genetic variants are involved in altered mitochondrial function and have associations with MDD that near the threshold for significance. A standard way to evaluate such a hypothesis is to investigate whether a specific genetic pathway, such as that involving the genes that affect mitochondrial function, is statistically more likely to have smaller P-values for association with MDD than expected by chance. This analysis was not reported in the current study. A previous systematic pathway analysis of MDD and other major psychiatric disorders did not implicate mitochondrial biology5. As such, although the researchers’ hypothesis is intriguing, it requires replication, extension, integrated analysis and more biological evidence. I wish the authors had formally tested their fundamental premise, namely that their sample was more homogeneous than those studied previously. If that is the case, then the heritability of the common variants (the proportion of variance contributing to MDD that can be accounted for by the genetic variation they measured) should be notably high. But although several methods exist to check this6,7, the authors did not report such an analysis. More unsettling, the two variants identified | NAT U R E | 1
© 2015 Macmillan Publishers Limited. All rights reserved
RESEARCH NEWS & VIEWS have almost no association signal in samples taken from European populations8. The reasons for this discrepancy are unclear. Perhaps these variants are truly causative for MDD only in severe cases from China. However, many other common associations for complex illnesses hold across the world. The authors tested the comparability of their findings with European samples and found some evidence of overlap, but more-refined analyses would be of keen interest. This first identification of replicable, significant genome-wide associations for MDD is exceptional. Although further work is required, it is to be hoped that these results will provide therapeutic targets for MDD. The
drug-discovery pipeline for MDD has never been based on solid biological foundations, but the work begun here could improve the focus of the field. The authors’ study marks the beginning of the beginning for the genetic dissection of MDD. The CONVERGE consortium has provided an excellent starting point for what should be an intriguing voyage of discovery. ■ Patrick F. Sullivan is in the Departments of Genetics and Psychiatry, University of North Carolina, Chapel Hill, North Carolina 275997264, USA, and in the Department of Medical Epidemiology and Biostatistics, Karolinska
2 | NAT U R E |
© 2015 Macmillan Publishers Limited. All rights reserved
Institutet, Stockholm, Sweden. e-mail: [email protected] 1. Levinson, D. F. et al. Biol. Psychiatry 76, 510–512 (2014). 2. CONVERGE consortium. Nature http://dx.doi. org/10.1038/nature14659 (2015). 3. Whiteford, H. A. et al. Lancet 382, 1575–1586 (2013). 4. Kendler, K. S. Arch. Gen. Psychiatry 54, 299–304 (1997). 5. The Network and Pathway Analysis Subgroup of the Psychiatric Genomics Consortium. Nature Neurosci. 18, 199–209 (2015). 6. Cross-Disorder Group of the Psychiatric Genomics Consortium. Nature Genet. 45, 984–994 (2013). 7. Bulik-Sullivan, B. K. et al. Nature Genet. 47, 291–295 (2015). 8. Major Depressive Disorder Working Group of the Psychiatric GWAS Consortium. Mol. Psychiatry 18, 497–511 (2013).
