Schizophrenia risk gene ZNF804A does not influence macroscopic brain structure: an MRI study in 892 volunteers

The single nucleotide polymorphism (SNP) rs1344706 in the ZNF804A gene was the first SNP to reach unequivocal genome-wide significance for schizophrenia, but little is known about how it confers susceptibility. Here, we looked for neuroanatomical correlates of rs1344706, and of other SNPs across the gene, in a large structural MRI data set of healthy young adults, using brain volumetry and voxel-based morphometry (VBM). Neither rs1344706 nor the other SNPs, individually or in combination, affected any volumetric or VBM parameter. As such, the association of ZNF804A with psychosis seems unlikely to be mediated through an influence of rs1344706 or any other SNP on macroscopic brain structure.

Schizophrenia and bipolar disorder are highly heritable disorders. Data from genome-wide association studies (GWAS) provide compelling evidence for association to the ZNF804A (zinc finger protein 804A) gene, at chromosome 2q32.1. Association to schizophrenia was originally reported in 2008,¹ and a recent meta-analysis of >21 000 cases and 38 000 controls found an odds ratio (OR) of 1.10, P=2.5 × 10⁻¹¹ for schizophrenia, and OR 1.11, P=4 × 10⁻¹³ for schizophrenia and bipolar disorder combined.² The genetic signal arises predominantly from a single, intragenic SNP, rs1344706³ or possibly a haplotype that includes it.² The roles of ZNF804A and the functional correlates of rs1344706 are unclear, although the latter is predicted to create a transcription factor-binding site and to affect gene expression of ZNF804A.⁴

Several studies have used neuroimaging to examine whether rs1344706 impacts upon brain structure. Lencz et al.⁵ showed in 39 healthy volunteers that carriers of the risk allele (A) of rs1344706 had significantly larger total white matter but reduced gray matter volumes in several regions comprising the ‘default mode network'. Other studies reported that the risk allele affected volumes of the amygdala and hippocampus in 70 patients but not in 38 healthy controls,⁶ and cortical thickness was reduced in the anterior cingulate, posterior cingulate and superior temporal gyrus in 62 healthy volunteers, with no changes in white matter tracts.⁷ A Chinese study found decreased white matter density in the left prefrontal lobe of 69 healthy risk allele carriers.⁸ However, all these studies were limited in terms of their sample sizes and no ZNF804A SNPs other than rs1344706 were examined.

Here, we studied the effects of ZNF804A on brain structure in a sample of 892 healthy young adults participating in the Brain Imaging Genetics (BIG) study (see Supplementary Table 1).⁹ Of these, 422 volunteers were scanned at 1.5 Tesla (T) and 470 at 3T. We implemented two analysis methods—volumetry and VBM—to look at the effects of rs1344706, and the individual and combined effects of 266 other SNPs in ZNF804A. The genetic data were available from an Affymetrix Genome-Wide Human SNP Array 6.0 and a subsequent imputation step using HapMap2 as reference sample (CEU data, NCBI build 36, UCSC hg18, http://www.sph.umich.edu/csg/abecasis/MaCH/download/). The gene-wide analysis method used a statistical approach as described by Hoh et al.¹⁰ where the analysis consisted of a SNP-by-SNP linear regression and the estimation of the effect of the complete gene on the volumes. The single SNP and gene-wide effects were assessed for gray matter, white matter and total brain volume, as well as eight circumscribed brain structures: nucleus accumbens, amygdala, brainstem, caudate nucleus, globus pallidus, hippocampus, putamen and thalamus. With VBM, we looked for genetic effects on regional differences across the whole brain in both the 1.5T and the 3T groups using a full-factorial ANCOVA with age, sex, scan protocol and total brain volume as covariates. To increase statistical sensitivity, we applied small volume corrections for regions of interest (dorsolateral prefrontal cortex, hippocampus, amygdala, anterior cingulate cortex and posterior cingulate cortex). See Supplementary Methods for details of imaging methods and genetic analysis.

The volumetry results for rs1344706 are presented in Table 1. They reveal no effects on either total brain volume, gray or white matter, or the regions of interest measured (all P>0.05).

Table 1. Brain volumes according to rs1344706 genotype.

Genotype	C/C	C/A	A/A	P-value
N	160	434	297
Frequency	0.18	0.49	0.33
Accumbens	0.67 (0.12)	0.69 (0.12)	0.68 (0.11)	0.86
Amygdala	2.15 (0.39)	2.14 (0.36)	2.14 (0.37)	0.52
Brainstem	19.23 (2.32)	19.39 (2.32)	19.22 (2.33)	0.37
Caudate nucleus	5.45 (0.66)	5.52 (0.61)	5.52 (0.58)	0.45
Gray matter	829.9 (79.57)	838.0 (71.49)	838.0 (77.80)	0.27
Globus pallidus	1.88 (0.23)	1.89 (0.21)	1.89 (0.21)	0.84
Hippocampus	5.33 (0.68)	5.37 (0.64)	5.31 (0.63)	0.35
Putamen	6.21 (0.70)	6.24 (0.69)	6.23 (0.72)	0.82
Total brain volume	1304 (129.1)	1317 (113.4)	1314 (120.1)	0.70
Thalamus	11.43 (1.20)	11.55 (1.08)	11.44 (1.12)	0.32
White matter	474.1 (56.74)	479.5 (51.22)	476.2 (51.81)	0.56

Values are mm³, mean (s.d.). Genotype data missing for one subject.

Similarly, VBM revealed no effects of rs1344706, either in the whole-brain analysis or in any of the regions of interest (Supplementary Table 2). Results for the gene-wide approach were also negative (Supplementary Tables 3–5). We also compared rs1344706 AA homozygotes versus CC homozygotes; again, there was no influence on any brain measure analysed (data not shown).

In summary, we have shown that genetic variation in ZNF804A, including the genome-wide significant psychosis risk allele rs1344706, does not affect total or regional brain volumes in healthy young adults. Compared with previous reports, we studied a far larger sample, used both volumetry and VBM and also conducted a gene-wide approach. The method has been shown previously to be effective for the detection of genetic effects on MRI volumes in the BIG database.⁹ Together, our results strongly suggest that rs1344706 in particular, and sequence variation in ZNF804A in general, do not impact upon macroscopic brain structure. While we cannot rule out such effects occurring in childhood or in clinical populations, it is more likely that any pathophysiological correlates of genetic variation in ZNF804A occur via modulation of brain function or connectivity.

The authors declare no conflict of interest.