FOXP2 expression and gray matter density in the male brains of patients with schizophrenia

Common genetic variants of FOXP2 may contribute to schizophrenia vulnerability, but controversial results have been reported for this proposal. Here we evaluated the potential impact of the common FOXP2 rs2396753 polymorphism in schizophrenia. It was previously reported to be part of a risk haplotype for this disease and to have significant effects on gray matter concentration in the patients. We undertook the first examination into whether rs2396753 affects the brain expression of FOXP2 and a replication study of earlier neuroimaging findings of the influence of this genetic variant on brain structure. FOXP2 expression levels were measured in postmortem prefrontal cortex samples of 84 male subjects (48 patients and 36 controls) from the CIBERSAM Brain and the Stanley Foundation Array Collections. High-resolution anatomical magnetic resonance imaging was performed on 79 male subjects (61 patients, 18 controls) using optimized voxel-based morphometry. We found differences in FOXP2 expression and brain morphometry depending on the rs2396753, relating low FOXP2 mRNA levels with reduction of gray matter density. We detected an interaction between rs2396753 and the clinical groups, showing that heterozygous patients for this polymorphism have gray matter density decrease and low FOXP2 expression comparing with the heterozygous controls. This study shows the importance of independent replication of neuroimaging genetic studies of FOXP2 as a candidate gene in schizophrenia. Furthermore, our results suggest that the FOXP2 rs2396753 affects mRNA levels, thus providing new knowledge about its significance as a potential susceptibility polymorphism in schizophrenia. Electronic supplementary material The online version of this article (10.1007/s11682-020-00339-x) contains supplementary material, which is available to authorized users.


Statistical power of the study
The statistical power obtained for the postmortem samples was 0.65 using the ClinCalc calculator (Rosner, 2011) at https://clincalc.com/Stats/Power.aspx, and 0.58 using the G*Power version 3.0.10 (Faul et al., 2007). When the sample was divided in genotypes, the statistical power was reduced to 0.34 (G*Power) due to the small number of individuals in the "CC" genotype.
For neuroimaging data, we applied the G*Power version 3.0.10 (Faul et al., 2007).Using a 10 mm sphere on each cluster of the resulting parametric maps, we have obtained an averaged statistical power of 0.73 for the comparison between patients and controls.
Regarding the imaging analysis between genotype groups, the averaged statistical power falls to 0.54 mainly due to the small number of subjects in each group.

Brain tissue samples
All samples belonged to Caucasian male donors with the exception of one Hispanic case in the schizophrenia patients group.

Subjects
None of the participants had previous electroconvulsive therapy or severe head trauma, present or past criteria for drug abuse (except for tobacco and cannabis), or any standard contraindications to the MRI examination.

RT-qPCR
RNA extraction from the CIBERSAM samples was carried out in RNAse-free environment. The purified total RNA was eluted in RNAse-free water and stored at -80ºC.
Nucleic acid concentration and purity were measured at 260 nm and 260nm/280nm respectively by spectrophotometry (Eppendorf BioPhotometer plus), obtaining purity values of 2. The integrity of RNA was further assessed calculating the RNA Integrity Number (RIN) using an Agilent 2100 Bioanalyzer. These samples showed RIN values between 4.0 and 8.0. Regarding the RNA from the Stanley array collection, poor RNA quality is an exclusion criteria for all specimens included in the collection (http://www.stanleyresearch.org/brain-research/). Primers for the transcript amplification of FOXP2 were designed to amplify a 116 bp product between exons 9 and 11 to avoid potential gDNA contamination (Tolosa et al. 2010).
Beta-actin (ACTB) was used to normalize the expression levels of FOXP2 since ACTB was the most stable housekeeping gene analyzed in our samples using the free Norm Finder software (http://www.mdl.dk/publicationsnormfinder.htm), previously developed and validated (Andersen et al. 2004). Each sample was run in triplicate. Following a 95 ºC denaturation for 3 min, the reactions were cycled 50 times with a 95 ºC denaturation for 30 s, a 57 ºC annealing step for 30 s and a 72 ºC extension for 30 s. A melt curve was obtained to calculate primers efficiency for the target and the reference gene. We obtained efficiency values of 94.961% for

Quantitative neuroimaging
After acquisition, the MR images were qualitatively reviewed by a radiologist and a computer engineer, who were both blind to clinical and genetic data, to ensure data quality. The images were then anonymized for analysis and post-processing.
To reduce the bias derived from using predefined templates (Shen et al. 2007), a custom set of tissue-templates was created that included all subjects. First, the raw images were normalized to the standard Montreal Neurological Institute of McGill University Health Center (MNI152) template using affine transformations. Second, the normalized MR images were segmented, averaged and smoothed using a 3D Gaussian smoothing kernel with an 8 mm FWHM to create whole brain, GM and WM tissue templates. Next, the original MR volumes of