Genome-wide association analysis of cognitive domains in a sample

Genome-wide association analysis of cognitive domains in a sample of
schizophrenia patients – first results
Ozkan S1,2, Papiol S1,2, Rossner MJ1,3, Zill P1, Riedel M4, Spellmann I4, Musil R4
1
Molecular and Behavioral Neurobiology, Department of Psychiatry, Ludwig Maximillian University, Munich, Germany
2
Institute of Psychiatric Phenomics and Genomics (IPPG), Ludwig Maximilian University, Munich, Germany
3
Max-Planck-Institute of Experimental Medicine, Goettingen, Germany
4
Department of Psychiatry and Psychotherapy, Ludwig Maximilian University, Munich, Germany
* corresponding author: [email protected]
Background
Schizophrenia (SCZ) is a severe neuropsychiatric disorder with high heritability, estimated between 60% and
80% and affects up to 1% of the population worldwide.
According to genome-wide association studies (GWAS),
a) it is a highly polygenic disorder in which thousands of
genetic loci contribute to the disease risk, b) common
variation explains an important proportion of the genetic
risk (Ripke et al., 2014).
Neurocognitive deficits are increasingly recognized as a
core feature of SCZ. Genetic variation is one of the sources that can explain differences in cognitive abilities. Family and molecular genetics studies have reported the
remarkable heritability of cognitive traits (Davies et al.
2011, 2015). The polygenic architecture of cognitive traits
has been reported by GWAS based on these phenotypes.
The genetic overlap between SCZ and cognitive impairment has been already described in family studies (Fowler, 2012). However, a direct relationship between genetic variants and cognitive profile in SCZ has not yet been
ascertained.
Keywords
Objectives
The aims of this study are:
Methods
The sample under analysis consisted of 135 SCZ patients (age:32.16 ± 10.63; age of onset: 27.62 ± 9.46; female
39.7%) diagnosed according to DSM-IV-TR criteria. Patients were recruited in the context of different randomized
controlled, atypical antipsychotic monotherapy studies. Several neurocognitive variables were assessed in the sample
of patients: verbal memory, visual memory, working memory, executive function, reaction time and reaction quality.
The patient samples were genotyped using the Infinium PsychArray Bead¬Chip (Illumina®). After quality control,
~300,000 Single Nucleotide Polymorphisms (SNPs) covering the whole genome were ready for genome-wide association study (GWAS) using PLINK 1.07 (Purcell et al., 2007). These analyses were based on the 6 phenotypes aforementioned while controlling for age, sex, treatment and four principal components as regards population stratification. Genome-wide significance threshold is set as p = 5E-8.
For the calculation of the different SCZ polygenic risk scores, SNPs were selected using the latest SCZ GWAS (Ripke et al., 2014) as initial training sample. This information was applied, using PLINK 1.07 (Purcell et al., 2007), to
construct a score in our independent replication sample of SCZ patients by forming the weighted sum of associated
alleles within each subject across different P-value thresholds. Up to 106 different P-value thresholds ranging from
5E-8 until 1 were applied, with increasing numbers of SNPs as the P-values became less stringent.
Standardized values of cognitive variables were used as dependent variables in a linear regression model. Age, sex,
treatment and four population stratification principal components were used as covariates. R2 values derived from
a model including all of these covariates were subtracted from R2 values from a model including covariates plus the
respective polygenic score. The difference between the adjusted R2 values represents the increase in the variance
explained attributable to the polygenic score.
Figure 1: Genome-wide association analysis of cognitive domains
Executive Function
Reaction Quality
1.To identify, using GWAS, genetic loci involved in the
performance in 6 cognitive domains in a sample of 135
schizophrenia patients recruited in the context of different randomized controlled, atypical antipsychotic monotherapy studies.
2. To ascertain, in the same sample of patients, the influence of SCZ polygenic risk scores on these cognitive
domains.
Reaction Time
B
LN
TX
8
L9
CU orf10
C6
Verbal Memory
5
PN
PT
CH
X
MA
B
NT
1-F
C
UR
Visual Memory
Working Memory
E3
OX
AL
Results
P1
IN
BR C1
B
D
4A
MD
FR
EN
AV RM5
CH
Genome-wide analysis of the 6 cognitive domains did not
reveal any genome wide association (Figure 1). However,
some interesting candidate loci were identified in the analysis of reaction quality and visual memory.
Schizophrenia polygenic risk scores were not found to influence the cognitive performances of the schizophrenic
patients in our samples (Figure 2).
Figure 1: Genome-wide association analysis of executive function, reaction quality, reaction time, verbal memory, visual memory and working
memory. P value thresholds shown: 1E-5 (suggestive associations, blue line) and 5E-8 (genome-wide associations, red line). X-axis, chromosome
positions; Y-axis, -logarithm of the p values.
No genome-wide associations were found. Some suggestive associations were observed with p values between 1E-5 and 5E-8, particularly in
Discussion
A recent genome-wide study on cognition based on more
than 100,000 subjects found few associations between
common variants and cognition in the general population
despite the large statistical power (Davies et al. 2016).
Given these results, the lack of genome-wide associations
with cognition in our study can be explained to a larger
extent by a) a smaller sample size and b) the complex
and polygenic genetic architecture of cognitive traits.
reaction quality and visual memory. These association signals were further characterized in order to identify candidate genes in these loci. The most
promising candidates are shown in Manhattan plots.
Figure 2: Polygenic risk score analysis of cognitive domains
Figure 2: Influence of SCZ polygenic risk scores
on executive function, reaction quality, reaction
time, verbal memory, visual memory and
working memory. X-axis shows the 106 p value
These limitations may also account for the negative results in the polygenic risk score analysis using the same
cognitive domains as target variables.
thresholds used in this analysis; Y-axis shows
the amount of variance explained by polygenic
risk scores (R2 change). Color gradient in the
bars indicates statistical significance.
Perspectives
Although results of reaction quality and verbal
The characterization of the suggestive loci in terms of genes
(TXLNB, CUL9, C6orf108, CHURC1-FNTB, MAX, PTPN5,
BRINP1, DBC1, FRMD4A, AVEN, CHRM5, ALOXE3) and/or
regulatory elements is currently ongoing. Pathway analyses are to follow and interactions of genetic and clinical
variables are subject of further investigations.
risk scores, none of these effects remain
The effect of individual genetic variants or polygenic scores
on the changes of cognition over time have not been yet
studied. The sample analysed in this study is characterised by the availability of longitudinal information as regards cognition, psychopathology and other outcomes of
interest. While the present study is based on a cross-sectional design, ongoing efforts are directed towards the integration of genetic data and longitudinal information in
this sample.
memory suggest a certain influence of polygenic
significant after multiple testing correction.
References
1. Schizophrenia Working Group of the PGC, Nature 2014; 511(7510):421-7.
2. Davies et al., Molecular Psychiatry 2011; 16:996-1005.
3. Davies et al., Molecular Psychiatry 2015; 20:183-192.
4. Fowler et al., Archives of General Psychiatry 2012; 69:460–466.
5. Purcell et al., American Journal of Human Genetics 2007;81(3):559-75.
6. Davies et al., Molecular Psychiatry 2016; 21:758–767.