The main findings of the present study were upregulated TCF4 expression level in all psychotic diagnostic groups and that TCF4 risk variants associate with poorer verbal fluency in total sample. The exploratory analysis revealed that TCF4 variants associated with negative symptoms, cognitive dysfunction and cerebellar volume in the schizophrenia sample, age at onset in total patient sample and variants correlating with temporal cortical area and brain volume in total sample. This further implicates TCF4 in psychosis pathophysiology, and indicates a specific role in the development of clinical phenotypes related to aberrant neurodevelopment. To our knowledge, this is the first study on known TCF4 risk variants of schizophrenia and hitherto uncharacterized variants targeting psychosis phenotypes at different levels, including the peripheral TCF4 mRNA.
The analysis of known schizophrenia TCF4
risk variants revealed that the two linked rs12966547 and rs4309482 are associated with poorer executive function in the form of verbal fluency. Interestingly, reduced cognitive functioning has been associated with enlarged ventricular volume,41
corroborating our finding that the same risk variants, rs12966547 and rs4309482, were associated with larger ventricular volume even though this did not survive correction for multiple testing. Enlarged ventricular volume is one of the most consistent findings in schizophrenia,42
also found in bipolar disorder43
and Pitt–Hopkins syndrome.44
This indicates that these TCF4
risk SNPs might, at least partially, cause this common subphenotype in schizophrenia.41
The consistent pattern of effect across different phenotypes strengthens the findings.
The pattern of associations in the non-schizophrenia risk SNP analysis corroborates a relationship between TCF4
and changes in brain morphology. Several SNPs in one haploblock were associated with cerebellar volume, temporal cortical area or brain volume. The Allan Brain Atlas shows TCF4
expression in several parts of the brain, most pronounced in hippocampus and cerebellum (http://www.brain-map.org
). This is consistent with a role of TCF4
in early neurodevelopment45
previously shown in animal models as well as concordant with microcephaly as a core characteristic of Pitt–Hopkins syndrome,44
as common variants in CDK5RAP2
are associated with reduced brain volume and cortical area,47
while mutations cause primary microcephaly.48
Psychotic disorders are a heterogenous group of diseases and division into subgroups based on symptomatology might be desirable. The consistent effect of the TCF4
SNPs of several sequential haploblocks on negative symptoms in schizophrenia patients might be a path to define a subgroup of patients with more pronounced negative symptoms as previously suggested.49, 50
Interestingly, the variants that affect negative symptoms are situated in the exon-dense region of the gene in which Pitt–Hopkins syndrome–causing mutations (exons 9–19) are identified.51
Furthermore, we found that one SNP was associated with both an earlier age at onset and worse verbal learning. Cognitive decline and early age at onset are predictors of poor prognosis in schizophrenia.52
The pattern of clinical symptoms and cognitive dysfunction resembles what has been regarded as neurodevelopmental characteristics, and is consistent with an effect of TCF4
on abnormal neurodevelopment. The consistent pattern of effect across different SNPs on independent clinical, cognitive and brain imaging phenotypes suggests that the current findings are not due to chance.
mRNA levels were significantly higher in the patient groups compared with controls. Measurements in the blood allow for larger sample sizes and better standardization of technical procedures, but it might not reflect mRNA levels in the brain. It has previously been reported that the availability of an accessible tissue, such as blood, with similar gene expression to a more inaccessible tissue, such as brain, may advance research in neuropsychiatric disorders.53
Sullivan et al.53
reported that, on a transcriptome level, whole blood shares significant gene expression similarities with brain tissue. Thus, even though it was not perfectly correlated, gene expression in blood might be a useful surrogate marker for gene expression in the central nervous system.53
Our observation of increased level of TCF4
mRNA in patients with schizophrenia is consistent with increased cerebellar expression of TCF4
in schizophrenia patients shown in a genomic convergence analysis on post-mortem cerebellar cortices,54
indicating that in the case of TCF4
, blood and brain expression might correlate. Microarrays of gene expression in blood is a practical method to use in psychotic patients and the technique has become increasingly accepted for studying biomarkers for neuropsychiatric diseases.55, 56, 57, 58, 59, 60, 61, 62, 63
For example, gene expression in peripheral blood mononuclear cells in patients with multiple sclerosis was correlated to disease activity.64
This is in line with our finding of a correlation between mRNA level and severity of positive and negative symptoms in patients with schizophrenia.
The result of increased TCF4
mRNA in psychotic patients should be interpreted with caution as an effect on mRNA levels of prescribed psychotropic medications cannot be excluded. We did find an overall effect of medication on mRNA level in patients, with the highest level in patients on antipsychotics or treated with several types of psychotropics, even though it did not survive correction for multiple testing. Thus, the observed effect might be a mirror of medication or reflect a more severe form of psychosis in patients on regular medical treatment. The trend of higher TCF4
mRNA level in unmedicated patients compared with controls does corroborate an effect regardless of medication, though. Another small-size (41 patients) study observed less TCF4
expression in blood from patients with higher rates of delusions.39
mRNA level was positively correlated with positive and negative symptoms in our sample. This is in line with our finding of several SNPs associated with negative symptoms, which might reflect a difference in TCF4
expression in patients with high level of positive and negative symptoms and suggests that TCF4
variants may contribute to more pronounced symptoms. The reason for the apparent discrepancy between Kurian et al.
and our findings might reflect different assumptions. While they compared TCF4
mRNA level in patients with schizophrenia with high level of delusions and patients with no delusions, our sample is generally in a more chronic phase of disease. Further, our sample is much larger and more heterogeneous and includes bipolar disorder and other psychoses, while they include different ethnicities. Therefore, before TCF4
mRNA can be used as a biomarker for psychosis, any putative effect of the level of positive and negative symptoms and medication on TCF4
mRNA level needs further investigation.
Two recent studies have found allele-specific expression of TCF4
in post-mortem human brains.65, 66
No significant association between mRNA level and the common TCF4
polymorphisms was found in our data after correction for multiple testing, but there were nominal associations. The lack of significant support for cis
-regulatory effect could be due to Type II error, as there are many factors that affect expression levels and may mask the effect. It may also be due to differences in eQTLs in blood and brain tissues. Given the observed small differences in transcript levels and the deleteriousness of hemizygosity for TCF4
observed in Pitt–Hopkins syndrome, it seems reasonable that evolutionary constraint on variants that alter expression levels might exist. Another option is that the locus only exerts its regulatory effect in a developmental context, or is organ specific. The gene expression in uteri and postnatal life has recently been examined, and the brain TCF4
mRNA level is highest in the early brain development in uteri, decreases in childhood and stabilizes in early adulthood.67
Thus, epigenetic factors from early life may have a continued effect on expression68
and together with a decrease in variation explain the lack of association between GWAS SNPs and TCF4
mRNA level in adults.67
It is also possible that none of the current TCF
SNPs tagged the eQTLs.
Taken together, these results support earlier findings of TCF4 as an important gene in psychosis etiology, and suggest that TCF4 variants may lead to a subgroup of psychosis with clinical, cognitive and brain structural abnormalities related to neurodevelopmental mechanisms.