Genetic correlations and genome-wide associations of cortical structure in general population samples of 22,824 adults

Cortical thickness, surface area and volumes vary with age and cognitive function, and in neurological and psychiatric diseases. Here we report heritability, genetic correlations and genome-wide associations of these cortical measures across the whole cortex, and in 34 anatomically predefined regions. Our discovery sample comprises 22,824 individuals from 20 cohorts within the Cohorts for Heart and Aging Research in Genomic Epidemiology (CHARGE) consortium and the UK Biobank. We identify genetic heterogeneity between cortical measures and brain regions, and 160 genome-wide significant associations pointing to wnt/β-catenin, TGF-β and sonic hedgehog pathways. There is enrichment for genes involved in anthropometric traits, hindbrain development, vascular and neurodegenerative disease and psychiatric conditions. These data are a rich resource for studies of the biological mechanisms behind cortical development and aging.

2) The authors provide an excellent analysis of co-heritability between the phenotypes they consider and brain-related traits. In particular, p-values in Table S16 for traits such as Parkinson's show more significance than in other studies such as Elliott et al. While still not significant at a BF level, this is a direction required in order to make progress in multiphenotype analysis. Also, the summarization of the genetic covariance in Figure S15 is much more clearly displayed and useful to researchers than the reports in Elliott et al. and other related work.
Minor point: 1) Page 25 line 11 "As recommended by the ldsc tool developers" -> "As recommended by the LDSC tool developers" Reviewer #3 (Remarks to the Author): This reviewer is satisfied with most responses from the authors. However, I still have a minor concern about the possible population structure and its potential consequences in this study. Specifically, this study integrates datasets from multiple different studies with different age range and MRI acquisition protocols. It seems that the authors ignored the heterogeneous population structure. This needs some discussions.
In addition, since the neuroimaging measures show a polygenic genetic architecture (i.e., many contributing loci are founded across the genome), it is of great interest to perform polygenic risk scores prediction (e.g., Vilhjálmsson et al., 2015 https://doi.org/10.1016/j.ajhg.2015.09.001) to check the out-of-sample genetics prediction power of these brain imaging phenotypes。

Reviewer #2 (Remarks to the Author):
The manuscript under review is "Genetic Determinants of Cortical Structure (Thickness, Surface Area and Volumes) in 2 General Population Samples of 22,824 Adults". The authors describe a meta analysis of many consortia, providing details for the mapping between genotype and brain phenotype.
This work is of great quality, and many improvements have been made since it was last reviewed by Nature Group, as follows.
1) More conservative thresholds for multiple test correction are now considered (in response to remark 1 of Reviewer #1). The authors note that BF correction reduces significant findings from 160 to 142. They now mention that BF destroys the CTh signal in the main text. 142 instead of 160 associations is still a valuable contribution to the effort to detail the mapping from genotype to brain phenotype. The authors may consider adding the 142 number to the main text or add a BF column to a table, to share this more standard analysis.

Response:
We agree with the reviewer and now write in Results / Genome-wide association analysis (page 9, lines 20-22): "If we had used a more stringent threshold of p discovery < 4.76x10 -10 = 5x10 -8 / 105, correcting for all the 105 GWAS analyses performed, we would have identified 142 significant associations ( Supplementary Tables 1-4)." In Supplementary Tables 1-4, discovery p-values < 4.76x10 -10 (= 5x10 -8 / 105) are now shown in bold. Table S16 for traits such as Parkinson's show more significance than in other studies such as Elliott et al. While still not significant at a BF level, this is a direction required in order to make progress in multiphenotype analysis. Also, the summarization of the genetic covariance in Figure S15 is much more clearly displayed and useful to researchers than the reports in Elliott et al. and other related work.

Response:
We thank the reviewer for this encouraging comment! Minor point: 1) Page 25 line 11 "As recommended by the ldsc tool developers" -> "As recommended by the LDSC tool developers"

Response:
We have changed "ldsc" to LDSC in the manuscript.

Reviewer #3 (Remarks to the Author):
This reviewer is satisfied with most responses from the authors. However, I still have a minor concern about the possible population structure and its potential consequences in this study. Specifically, this study integrates datasets from multiple different studies with different age range and MRI acquisition protocols. It seems that the authors ignored the heterogeneous population structure. This needs some discussions.

Response:
We now have now changed the "limitations" paragraph in the Discussion (page 19, lines 23-25 and page 20, lines 1-5): "A limitation of our study is the heterogeneity of the MR phenotypes between cohorts due to different scanners, field strengths, MR protocols and MRI analysis software. This heterogeneity as well as the different age ranges in the participating cohorts may have caused different effects over the cohorts. We nevertheless combined the data of the individual cohorts to maximize the sample size as it has been done in previous CHARGE GWAS analyses [31][32][33] . To account for the heterogeneity we used a sample-size weighted meta-analysis which does not provide overall effect estimates. This method has lower power to detect associations compared to inversevariance weighted meta-analysis and we therefore may have found less associations." In addition, since the neuroimaging measures show a polygenic genetic architecture (i.e., many contributing loci are founded across the genome), it is of great interest to perform polygenic risk scores prediction (e.g., Vilhjálmsson et al., 2015Vilhjálmsson et al., https://doi.org/10.1016Vilhjálmsson et al., /j.ajhg.2015 to check the out-of-sample genetics prediction power of these brain imaging phenotypes。