simplebooklet thumbnail

of 0
OPEN
ORIGINAL ARTICLE
Association between polygenic risk for schizophrenia,
neurocognition and social cognition across development
L Germine
1,2,3,4
, EB Robinson
4,5
, JW Smoller
1,2,4,6
, ME Calkins
7
, TM Moore
7
, H Hakonarson
8
, MJ Daly
4,5
,PHLee
1,4
, AJ Holmes
9
,
RL Buckner
3,10,11
, RC Gur
7
and RE Gur
7
Breakthroughs in genomics have begun to unravel the genetic architecture of schizophrenia risk, providing methods for
quantifying schizophrenia polygenic risk based on common genetic variants. Our objective in the current study was to understand
the relationship between schizophrenia genetic risk variants and neurocognitive development in healthy individuals. We rst used
combined genomic and neurocognitive data from the Philadelphia Neurodevelopmental Cohort (4303 participants ages 821
years) to screen 26 neurocognitive phenotypes for their association with schizophrenia polygenic risk. Schizophrenia polygenic risk
was estimated for each participant based on summary statistics from the most recent schizophrenia genome-wide association
analysis (Psychiatric Genomics Consortium 2014). After correction for multiple comparisons, greater schizophrenia polygenic risk
was signicantly associated with reduced speed of emotion identication and verbal reasoning. These associations were signicant
by age 9 years and there was no evidence of interaction between schizophrenia polygenic risk and age on neurocognitive
performance. We then looked at the association between schizophrenia polygenic risk and emotion identication speed in the
Harvard/MGH Brain Genomics Superstruct Project sample (695 participants ages 1835 years), where we replicated the association
between schizophrenia polygenic risk and emotion identication speed. These analyses provide evidence for a replicable
association between polygenic risk for schizophrenia and a specic aspect of social cognition. Our ndings indicate that individual
differences in genetic risk for schizophrenia are linked with the development of aspects of social cognition and potentially verbal
reasoning, and that these associations emerge relatively early in development.
Translational Psychiatry (2016) 6, e924; doi:10.1038/tp.2016.147; published online 18 October 2016
INTRODUCTION
Schizophrenia is among the most debilitating and highly heritable
of mental disorders. Recent shifts in our conceptualization of
neuropsychiatric illnesses suggest that such disorders might be
better understood in terms of underlying behavioral or neurobio-
logical dimensions rather than as categories.
1
Evidence from
neuroscience,
25
behavioral genetics
6
and prospective clinical
studies
7,8
suggest that schizophrenia is associated with quantita-
tive variations in neurobiological and neurocognitive systems.
These observations have led to several hypothesized relationships
between schizophrenia and neurocognitive abilities,
911
though
genetic studies examining the association between schizophrenia
and general cognitive ability or educational attainment have been
inconsistent.
12,13
Recent genome-wide associations studies (GWAS) provide a
window into the genetic architecture of schizophrenia, and
support a complex model of psychosis liability. First, these studies
demonstrate that, individually, common single-nucleotide poly-
morphisms explain very little of the variation in schizophrenia
liability.
14
The effects of common genetic variants must be taken
in aggregate to explain a meaningful proportion of schizophrenia
risk, indicating that the genetic liability for schizophrenia is the
result of differences across many hundreds to thousands of genes
and regulatory regions. GWAS data now provide the means to
quantify aggregate genetic risk for schizophrenia (hereafter
referred to as schizophrenia polygenic risk) for any individual,
regardless of their familial or phenotypic risk,
14,15
paving the way
for a renewed interrogation of intermediate phenotypes (also
known as endophenotypes).
1618
We use the term intermediate
phenotype to refer to measurable variations in biological or
information processing systems that are thought to lie along a
causal pathway between genetic risk and mental disorder and
might be used to understand the downstream effects of validated
genetic-risk factors.
19
Understanding downstream effects of
schizophrenia genetic-risk variants in terms of intermediate
phenotypes can highlight potential pathways that lead from
genetic variation to schizophrenia, as well as potential biologically
informative phenotypes to study outside of case populations.
18,19
Here we take advantage of a data set of ~ 4300 individuals ages
821 years collected through the Philadelphia Neurodevelop-
mental Cohort (PNC)
2022
and a replication sample of ~ 700
individuals tested as part of the Harvard/Massachusetts General
Hospital (MGH) Brain Genomics Superstruct Project (GSP).
23
The
PNC data set includes genome-wide data on all participants and a
1
Psychiatric and Neurodevelopmental Genetics Unit, Massachusetts General Hospital, Boston, MA, USA;
2
Psychiatry Department, Harvard Medical School, Boston, MA, USA;
3
Psychology Department, Harvard University, Cambridge, MA, USA;
4
Broad Institute of MIT and Harvard, Boston, MA, USA;
5
Analytic and Translational Genetics Unit,
Massachusetts General Hospital, Boston, MA, USA;
6
Department of Epidemiology, Harvard T.H. Chan School of Public Health, Boston, MA, USA;
7
Neuropsychiatry Section,
Department of Psychiatry, Perelman School of Medicine, University of Pennsylvania, Philadelphia, PA, USA;
8
Center for Applied Genomics, The Childrens Hospital of Philadelphia,
Philadelphia, PA, USA;
9
Psychology Department, Yale University, New Heaven, CT, USA;
10
Center for Brain Sciences, Harvard University, Cambridge, MA, USA and
11
MGH/HST
Athinoula A. Martinos Center for Biomedical Imaging, Massachusetts General Hospital, Boston, MA, USA. Correspondence: Dr L Germine, Division of Depression and Anxiety
Disorders, McLean Hospital, Belmont, MA 02478, USA.
E-mail: lgermine@mclean.harvard.edu
Received 27 June 2016; accepted 4 July 2016
Citation: Transl Psychiatry (2016) 6, e924; doi:10.1038/tp.2016.147
www.nature.com/tp
comprehensive assessment of neurocognitive function across
domains of general and social cognition, where measures were
selected and designed to map onto specic neural circuitry.
2426
We took an unbiased approach towards understanding the
relationship between schizophrenia polygenic risk and multiple
domains of neurocognitive performance, by rst exploring the
relationship between polygenic risk and performance across all
neurocognitive measures. Genetic effects on complex traits are
known to be broadly pleiotropica phenotypic screening
approach allows us to identify the prole of associations between
schizophrenia genetic risk and neurocognition, accounting for such
pleiotropy within the neurocognitive phenotypes assessed. We
then examined whether there was any evidence of developmen-
tally specic effects of polygenic risk on neurocognition. We
operationalize polygenic risk for schizophrenia as the weighted
sum of the effects of many thousands of risk alleles across the
genome, identied through large-scale schizophrenia GWAS
analyses.
14,15
Previous research has suggested that schizo-
phrenia vulnerability might impact neurocognition at specic
developmental transitions that happen during puberty and
adolescence,
27,28
and thus associations between polygenic risk
and neurocognition might be restricted to a particular age range or
not begin until a critical developmental period begins. Using the
resources of the PNC and GSP, we explore and replicate tests of the
hypothesis that individual differences in schizophrenia genetic risk
are related to quantitative dimensions of neurocognition.
MATERIALS AND METHODS
Participants
Our primary analytic sample was drawn from the PNC, a Childrens Hospital
of Philadelphia health network-based sample of ~ 9500 individuals ages 8
21 years from the greater Philadelphia area (details in Calkins et al.
20
and
Gur et al.
24
). Participants who provided assent/consent gave genetic
samples and written permission to be recontacted for further research. The
University of Pennsylvania and Childrens Hospital of Philadelphia
Institutional Review Boards approved the study. After stratication based
on sex, age and ethnicity, PNC participants were recruited through random
selection. Inclusion criteria were (1) ability to provide informed consent
(parental consent where age o 18 years), (2) English prociency and (3)
physical and cognitive functioning sufcient to complete clinical assess-
ment interviews and cognitive testing on a computer. Participants were
not excluded for any other medical concerns, including psychiatric illness.
No recruitment was done at psychiatric clinics, however, so the sample is
not enriched for individuals seeking psychiatric care.
Our replication sample was drawn from the Harvard/MGH Brain GSP. The
GSP is a study cohort that includes neuroimaging, genomic and cognitive
data on over 4000 healthy participants.
23
The present sample included an
age-restricted subsample that performed behavioral tasks compatible with
the discovered effects in the PNC. Because of the differences in protocols,
only one PNC assessment could be tested for replication. All GSP
participants provided written informed consent for biomedical research
approved by the Partners Healthcare Institutional Review Board or the
Harvard University Committee on the Use of Human Subjects. Inclusion
criteria were (1) English prociency, (2) age 1835 years, (3) no history of
psychiatric illness or major health problems and (4) physical and cognitive
functioning sufcient to complete magnetic resonance imaging scanning
and cognitive testing.
Genetic analysis
As polygenic risk scores are sensitive to ancestry, we restricted our analysis
to genotyped individuals with self-described white non-Hispanic ancestry.
Within this subsample, we further excluded population outliers (prior to
imputation) based on directly genotyped single-nucleotide polymorphism
data. Data cleaning and imputation were performed using standard
procedures (see Robinson et al.
21
for PNC data; Hibar et al.
29
for GSP data;
see also Ripke et al.
30
). In the PNC data set, the ancestry threshold was
relaxed to a pi_hat of 0.1 (0.125 for the GSP data set), which is a level of
relatedness equal to or less than that of rst cousins. Imputed data were
used to generate individual schizophrenia polygenic risk scores using the
procedures described in Purcell et al.
15
and Ripke et al.
14
In brief, polygenic
risk scores estimate genome-wide common variant liability for a trait
through a weighted sum of many thousands of risk alleles. The polygenic
risk score used here was generated using summary statistics from the
Psychiatric Genomics Consortium recent meta-analysis of schizophrenia,
14
and includes only single-nucleotide polymorphisms with Po 0.05. This
version of the score was selected because it most commonly maximized
the schizophrenia risk explained in independent casecontrol samples
(~20% of casecontrol variation).
14
In addition to limiting the analysis to
participants of European descent, the rst 10 principal components of
ancestry were controlled for in all analyses. There was no relationship
between schizophrenia polygenic risk and age or sex in either sample.
Phenotypic neurocognitive assessment and analysis
PNC participants completed the Computerized Neurocognitive Battery
(CNB).
24,25
The CNB was developed from tasks that map onto specic brain
systems, as identied through functional neuroimaging.
26
Psychometric
properties
(adult and pediatric samples) and task descriptions for the CNB
measures are included elsewhere.
21,2426,31
The CNB provides accuracy and
speed measures of: executive function (abstraction and mental exibility,
attention and working memory), memory (verbal, spatial and facial),
complex cognition (verbal reasoning, nonverbal reasoning and spatial
processing), social cognition (emotion identication, emotion differentia-
tion and age differentiation) as well as speed measures for sensorimotor
and pure motor function. All PNC neurocognitive tests were completed on
a computer in the laboratory or at home, according to family or participant
preference. For specic test names, see Table 1.
GSP participants completed a battery of personality, cognitive and
behavioral measures assessing a broad range of domains. A full list of
measures and details of neurocognitive assessment for the GSP are
described in Holmes et al.
23
A subset of measures that are available in the
PNC were also available in the GSP data set (in comparable or identical
form). All GSP neurocognitive tests were completed using online versions
of the tests, on a participant's own computer.
Differences in performance attributable to age and sex were regressed
out of all neurocognitive variables (speed and accuracy) prior to analysis.
Any individual scores more than four s.d. from the mean for a particular
test were excluded. Outliers were determined based on age group means.
Linear regression was then used to examine the relationship between
schizophrenia polygenic risk and neurocognitive performance. In our
primary analytic sample (PNC), Bonferroni correction was applied to correct
for the number of comparisons across all neurocognitive variables (26
comparisons: 12 accuracy variables and 14 speed variables) with an alpha
threshold of 0.05. Bonferroni correction is appropriate for family-wise error
adjustment even in the case where all phenotypes are independent of one
another. As neurocognitive performance across difference measures tends
to be somewhat correlated, this correction ensures that the probability of a
false positive is (at most) 0.05, and provides a strict standard of statistical
Table 1. Specic test names are given corresponding to each
neurocognitive domain assessed
Domain Test name
Abstraction/cognitive exibility Penn conditional exclusion test
Attention Penn continuous performance test
Working memory Letter N-back task
Verbal memory Penn word memory task
Face memory Penn face memory task
Spatial memory Visual object learning test
Verbal reasoning Penn verbal reasoning test
Nonverbal reasoning Penn matrix reasoning test
Spatial reasoning Penn line orientation test
Emotion identication Penn emotion identication test
Emotion discrimination Penn emotion differentiation test
Age discrimination Penn age differentiation test
Motor speed Computerized nger tapping test
Sensorimotor speed Mouse practice task
Abbreviation: GSP, Genomics Superstruct Project. The emotion identica-
tion test from our replication sample (GSP) was also the Penn emotion
identication test.
Schizophrenia polygenic risk and cognition
L Germine et al
2
Translational Psychiatry (2016), 1 7
evidence in these analyses. Thus, this correction is conservative given the
correlation structure of the neurocognitive phenotypes.
21
RESULTS
The nal analytic sample from the PNC included 4303 participants
(50% female) ranging in age from 8 to 21 years (near uniform
distribution with mean age of 13.8 years). As described in
Robinson et al.,
21
most of the 838 genotyped white non-
Hispanic PNC individuals excluded from this analysis were
removed for outlying ethnicity or relatedness to another person
in the data set. After Bonferroni correction, two neurocognitive
measures were signicantly associated with schizophrenia poly-
genic risk at Po 0.05 (Figure 1a). These were verbal reasoning
speed (analogies) (β = 0.058 l, Po 5E 4 uncorrected, Po 5E 3
corrected) and emotion identication speed (β = 0.066,
Po 5E 5 uncorrected, P o 1E 3 corrected). For both variables,
increases in polygenic risk were related to linear decreases in
speed of responses, with the highest quartile showing the slowest
speeds (longest reaction times) and the lowest quartile showing
the highest speeds (shortest reaction time; Figure 1b). This was
not the case for other variables related to response speed, for
example sensorimotor speed, where no differences were observed
between the lowest and highest quartiles of schizophrenia genetic
risk (Figure 1b). Schizophrenia polygenic risk was unrelated to
matrix (nonverbal) reasoning ability, a common proxy for general
intelligence (no relationship after correction for multiple compar-
isons; P=0.035 uncorrected, P=0.91 corrected) or general
cognitive ability (that is, general intelligence or g, based on
factor analysis; P= 0.64). We estimated g using standard
methods,
32
specically by taking scores from the rst principal
component (unrotated) based on a principal component analysis
using all accuracy variables (tests 112; Table 1) in the PNC
sample.
To understand the specicity of the relationship between
schizophrenia polygenic risk and speed of verbal reasoning and
emotion identication, we performed the same analysis, control-
ling for overall motor speed, sensorimotor speed, as well as
matrix/nonverbal reasoning ability. Both associations were similar
after controlling for motor speed and sensorimotor speed
(emotion identication speed: β = 0.079, P o 5E 7 uncorrected,
Po 1E 5 corrected; verbal reasoning speed: β = 0.065,
Po 5E 5 uncorrected, Po 1E 3 corrected) and matrix reasoning
ability (emotion identication speed: β = 0.065, Po 5E 5
uncorrected, Po 1E 3 corrected; verbal reasoning speed:
β = 0.057, Po 5E 4 uncorrected, Po 1E 2 corrected).
To understand the impact of speed accuracy trade-offs on the
strength of each association, we also looked at both emotion
Figure 1. Schizophrenia polygene scores and neurocognitive performance. (a) Linear regression was used to estimate associations between
schizophrenia polygenic risk, estimated from genome-wide data, and performance for each neurocognitive variable (labels shown on the
right). To best illustrate the strength of the evidence for each association, relationships are plotted in terms of the negative base-10 logarithm
of the P-value, when regressing neurocognitive performance on schizophrenia polygenic risk estimates for each participant. The gray line
shows the threshold for statistical signicance based on P o 0.05, uncorrected. The black line shows the threshold for statistical signicance
after Bonferroni correction for all 26 comparisons (P o 0.05 corrected). Red bars show variables where an association exceeded the threshold
for statistical signicance (verbal reasoning speed and emotion identication speed). For nonsignicant associations, black bars indicate a
negative relationship between schizophrenia polygenic risk and neurocognitive performance (that is, greater polygenic risk associated with
poorer performance) and gray bars indicate a positive relationship. (b) The relationship between schizophrenia polygenic risk and
neurocognitive performance is shown for the two speed variables (emotion identication and verbal reasoning) where associations were
statistically signicant after correction for multiple statistical tests and (for comparison purposes) associations with a general measure of
response speed. Participants from the primary analytic sample (PNC data set) were divided into four groups of approximately equal size based
on level of schizophrenia polygenic risk. Quartile 1 (Q1) includes individuals with the lowest schizophrenia polygenic risk. Quartile 4 (Q4)
includes individuals with the highest schizophrenia polygenic risk. Mean z-score is plotted on the y axis, with higher values reecting better
performance. For both emotion identication and verbal reasoning speed, increasing polygenic risk was linearly associated with decrease in
neurocognitive performance. There was no consistent relationshipsignicant or otherwisebetween schizophrenia polygenic risk and
sensorimotor speed.
Schizophrenia polygenic risk and cognition
L Germine et al
3
Translational Psychiatry (2016), 1 7
identication speed and verbal reasoning speed, controlling for
emotion identication accuracy and verbal reasoning accuracy,
respectively. Emotion identication speed was associated with
schizophrenia polygenic risk, even after controlling for emotion
identication accuracy (β = 0.065, Po 5E 5 uncorrected,
Po 1E 3 corrected). Verbal reasoning speed was also associated
with schizophrenia polygenic risk after controlling for verbal
reasoning accuracy (β = 0.056, Po 5E 4 uncorrected, P o 1E 2
corrected).
For both variables, schizophrenia polygenic risk accounted for
approximately 0.30.5% of the variation in neurocognitive
performance.
Developmental specicity
We evaluated the relationship between polygenic risk and speed
over development, by looking at the interaction between
schizophrenia polygenic risk and age for both emotion identica-
tion speed and verbal reasoning speed (Figure 2). There was no
interaction between schizophrenia polygenic risk and age for
either emotion identication speed (P=0.38) or verbal reasoning
speed (P=0.11). The relationship between schizophrenia poly-
genic risk and neurocognitive performance was statistically signi-
cant by 8 years of age for verbal reasoning speed (β = 0.094,
Po 5E 2) and by 9 years of age for emotion identication speed
(β = 0.11, Po 5E 2). Thus, there was no evidence that the
associations between schizophrenia polygenic risk and emotion
identication, or between schizophrenia polygenic risk and verbal
reasoning, were emerging later in development (for example, after
puberty). There was also no evidence for the opposite pattern
early associations that disappeared later in development due to
potential compensatory changes. Overall, our analyses did
not provide any indication of modulation in the relationship
between schizophrenia polygenic risk and emotion identication
or schizophrenia polygenic risk and verbal reasoning across
development.
Replication in an independent sample
In the GSP data set, there were 695 participants completed the
emotion identication measures used in the CNB and met our
inclusion criteria. These participants were 53% female and ranged
in age from 18 to 35 years (mean: 21.5 years; s.d.: 3.2 years). This
independent sample allowed us to test replication of the
relationship between schizophrenia polygenic risk and emotion
identication speed. The stimuli for the emotion identication task
were shared from the CNB at the initiation of the GSP data
collection effort, allowing for a true, independent replication using
the same stimuli and test in an independent cohort. The
association between schizophrenia polygenic risk and emotion
identication speed was replicated, despite GSP participants being
sampled from a very different population (β = 0.09, P o 0.05; see
Figure 3). This association again survived controlling for general
intelligence (as estimated by matrix reasoning performance) and a
measure of psychomotor response speed. The CNB emotion
identication test was the only neurocognitive test that over-
lapped between the two samples. Verbal reasoning speed was not
one of the neurocognitive measures included in the GSP, so we
were unable to assess replication for this association.
DISCUSSION
In two large samples, spanning middle childhood to adulthood,
we identify a small, but statistically replicable relationship
between schizophrenia polygenic risk and speed of emotion
identication. We also found an equally signicant association
between schizophrenia polygenic risk and speed of verbal
reasoning in our developmental sample (ages 821 years). Both
associations survived correction for multiple comparisons in our
developmental (discovery) sample and the inclusion of covariates
related to general intelligence and sensorimotor speed. Critically,
the association between schizophrenia risk and emotion identi-
cation replicated in an independent, demographically distinct
sample of adults.
We found no evidence that either association depended on
developmental phase: effects were signicant by 9 years of age
and effect sizes were consistent across middle childhood,
adolescence and early adulthood. These results suggest a
relatively early perturbation affecting neurocognitive develop-
ment that is present before the typical onset of psychotic illness in
young adulthood. Finally, we found that even intermediate levels
of schizophrenia polygenic risk were associated with reductions in
emotion identication and verbal reasoning speed. In other words,
the relationship between schizophrenia polygenic risk and
performance was evident across the spectrum of polygenic risk,
Figure 2. Schizophrenia polygene scores as predictors of emotion
identication speed and verbal reasoning speed, across age. Shown
are associations between schizophrenia polygenic risk and speed of
emotion identi cation and verbal reasoning, at each year of age.
Bars give ± 1 s.e. of the effect size estimate. Although both measures
were signicantly associated with schizophrenia polygenic risk,
there was no signicant interaction of polygenic risk and age on
neurocognitive performance for either variable.
Figure 3. Schizophrenia polygenic risk and emotion identication
speed. Replication in an independent sample of adults. Shown are
effect size relationships between emotion identication speed
(controlling for age and sex) and schizophrenia polygenic risk in
the original discovery sample (PNC ages 821 years) and the
replication sample (GSP ages 1835 years). Black bars show s.e. of
the β values in both samples. GSP, Genomics Superstruct Project;
PNC, Philadelphia Neurodevelopmental Cohort.
Schizophrenia polygenic risk and cognition
L Germine et al
4
Translational Psychiatry (2016), 1 7
as opposed to being observed only in those at the highest
polygenic risk.
33
Our results indicate that common variants that
increase risk for schizophrenia impact the development of specic
aspects of social cognition and possibly verbal reasoning.
The nding of a relationship between schizophrenia polygenic
risk and emotion identication performance is consistent with a
large body of literature emphasizing the importance of social
cognition in schizophrenia, individuals who go on to develop
schizophrenia and individuals at risk of schizophrenia. Schizo-
phrenia is associated with profound and consistent decits in
social cognition.
3437
These decits appear early, sometimes
decades before the onset of illness
3840
and are strongly related
to core aspects of symptomatology and everyday social
functioning.
4145
Differences in social cognition are observable in
healthy individuals with a potential genetic predisposition
towards developing schizophrenia.
4648
Early environmental
factors linked with the development of schizophrenia
49,50
are also
associated with adult differences in social cognition.
51
Even in
healthy populations, quantitative differences in psychosis-like
characteristics are linearly related to differences in emotion
identication.
33
There is extensive literature with functional
neuroimaging indicating abnormalities in regional brain activation
to emotion identication tasks in patients with schizophrenia
4,5,52
and those with psychosis spectrum features,
53
including indivi-
duals from the PNC who underwent neuroimaging.
54
Our specic
ndings provide further support, based on polygenic risk
estimates, that schizophrenia is a disorder that is fundamentally
related to the development of social cognition, specically the
efciency of emotion identication.
Abnormalities in verbal reasoning have also been identied
in schizophrenia and rst-degree relatives of schizophrenia
patients.
55
Although this association was not available for
replication due to the absence of a comparable phenotype in
our replication sample, the effect size difference was comparable
to our nding relating polygenic risk and emotion identication
speed. Perhaps decits in processing emotion and verbal
communication combine in creating difculties in social commu-
nication, contributing to social functioning decits related to
schizophrenia risk. Abnormalities in the development of social
understanding and communicationin the presence of other life
events or cognitive vulnerabilitiescould set the stage for
psychosis later in life. This possibility is consistent with the
ndings in the entire PNC sample that individuals with psychosis
spectrum features are delayed in their neurocognitive develop-
ment already by 8 years of age.
56
Notably, the most pronounced
delays were in complex reasoning and social cognition.
These results have implications for our understanding of
schizophrenia genetic risk in relation to neurocognition, particu-
larly social cognition. First, our ndings validate the notion that
risk genes might perturb cognition in a dimensional fashionwith
a linear (as opposed to threshold) relationship between genetic
load and cognitive impairment. We found such a relationship for
both emotion identication speed and verbal reasoning speed,
indicating that quantitative dimensions of genetic risk and
cognitive vulnerability can provide critical information for under-
standing psychopathology. Further, that these relationships did
emerge so early may make it unlikely that social decits arise as a
consequence of the expression of psychotic or psychosis-like
characteristics (for example, stigma or social rejection could drive
social isolation and subsequent deterioration of social cognitive
skills in patients with schizophrenia). Indeed, the differences at
even low levels of polygenic risk point to a more direct and
fundamental relationship.
It is noteworthy that our ndings were mainly for speed of
processing rather than accuracy. This effect is consistent with
earlier studies that showed reduced speed in individuals
genetically related to probands with schizophrenia and may
indicate compensatory strategies to mitigate vulnerability.
57
Furthermore, processing speed has been implicated in meta-
analyses as a major decit domain in schizophrenia.
11
Long-
itudinal studies are needed to examine whether reduced speed of
processing during development portends decits that extend to
accuracy and eventually to schizophrenia.
The present study has several limitations. The wide age range
spanning from childhood to early adulthood is an epoch
associated with rapid improvement in cognitive performance.
While enabling the identication of developmental effects, this
characteristic of the PNC may mask effects that could be detected
in equally sized samples of adults with a narrower age range. The
effects we report, although signicant, are smallaccounting for
only between 0.5 and 1% of variance in neurocognitive
performance. Given the small effect size, the association between
schizophrenia polygenic risk and neurocognitive performance
reported is primarily of theoretical interest by illustrating how
polygenic risk is linked to a specic dimensional cognitive domain.
The small effect sizes reported here suggest that although current
schizophrenia polygenic risk models may be useful for identifying
mechanisms, they should not be used for making predictions
about neurocognitive performance or making conclusions about
individuals. The size of the identied associations may also reect
the overall normative nature of the sample; estimated risk could
still confer clinically signicant effects in vulnerable individuals.
Another limitation of these analyses is that we only looked at
schizophrenia polygenic risk, and not polygenic risk for other
neuropsychiatric diseases. Emotion identication speed might also
be related to polygenic risk for other neuropsychiatric disorders
beyond schizophrenia. Finally, only one of the effects could be
tested in an independent sample because the other measure was
not available, underscoring the need for harmonizing measures
across genomic projects.
Genome-wide association analyses now provide robust and
replicable methods for quantifying genetic risk for schizophrenia.
Although such developments are an important milestone toward
unraveling the biological architecture of mental disorders, they
have been broadly acknowledged as only rst steps toward
understanding mechanisms. Here we show that variations in
emotion identication speed are one replicable downstream
effect of schizophrenia polygenic risk. This nding suggests
specic hypotheses about the relationship between schizophrenia
polygenic risk and neurobiology that could be addressed by
future work. Given the large body of research on the neural basis
of emotion identication, future studies might focus on the link
between schizophrenia polygenic risk and neural responses in
subcortical and superior temporal regions that have been
consistently linked with emotion perception. Future work might
also look at the developmental trajectories of these regions in
relation to schizophrenia polygenic risk, to understand how
differences in early development might lead to neurocognitive
variations in middle childhood and beyond, and ultimately confer
psychiatric vulnerability.
CONCLUSION
Notwithstanding these limitations, our results highlight a new and
important role for intermediate phenotypes in the GWAS-era: as a
means of understanding mechanism from validated genetic
predictors of disease.
19
In this role, intermediate phenotypes are
not only useful but also fundamental, as they allow us to
understand how genes contribute to the development of disease.
Finally, our results point to one potential mechanism linking
genetic risk for schizophrenia to psychosis through alterations in
the efciency of emotion identication. These alterations arise
early in development before typical onset of psychosis. This
nding is consistent with decades of research on the fundamental
nature of social decits in schizophrenia, further indicating that
these social decits may arise from schizophrenia risk-related
Schizophrenia polygenic risk and cognition
L Germine et al
5
Translational Psychiatry (2016), 1 7
common genetic variants. The nding that schizophrenia has a
genetic basis in individual differences opens up many pathways
for further study, including the translation of knowledge from
social neuroscience and social cognitive psychology toward the
elucidation of the roots of major psychiatric illness.
CONFLICT OF INTEREST
The authors declare no conict of interest.
ACKNOWLEDGMENTS
This study was supported by National Institutes of Health grants F32MH102971 (Dr
Germine), MH089983 and MH096891 (Dr Gur) and MH089924 (Dr Hakonarson),
K08MH079364 (Dr Calkins), K99MH101367 (Dr Lee), U01MH094432 (Dr Daly),
K01MH099286 (Dr Robinson) and K24MH094614 (Dr Smoller).
REFERENCES
1 Insel T, Cuthbert B, Garvey M, Heinssen R, Pine DS, Quinn K et al. Research domain
criteria (RDoC): toward a new classication framework for research on mental
disorders. Am J Psychiatry 2010; 167:748751.
2 Phillips ML, Drevets WC, Rauch SL, Lane R. Neurobiology of emotion perception II:
implications for major psychiatric disorders. Biol Psychiatry 2003; 54:515528.
3 Brunet E, Sarfati Y, Hardy-Baylé M-C, Decety J. Abnormalities of brain function
during a nonverbal theory of mind task in schizophrenia. Neuropsychologia 2003;
41: 15741582.
4 Gur RE, Loughead J, Kohler CG, Elliott MA, Lesko K, Ruparel K et al. Limbic acti-
vation associated with misidentication of fearful faces and at affect in schizo-
phrenia. Arch Gen Psychiatry 2007; 64: 13561366.
5 Gur RE, McGrath C, Chan RM, Schroeder L, Turner T, Turetsky BI et al. An fMRI
study of facial emotion processing in patients with schizophrenia. Am J Psychiatry
2002; 159: 19921999.
6 Seidman LJ, Hellemann G, Nuechterlein KH, Greenwood TA, Braff DL, Cadenhead
KS et al. Factor structure and heritability of endophenotypes in schizophrenia:
Findings from the Consortium on the Genetics of Schizoph renia (COGS-1). Schi-
zophr Res 2015; 163:7379.
7 de Paula ALD, Hallak JEC, Maia-de-Oliveira JP, Bressan RA, Machado-de-Sousa JP.
Cognition in at-risk mental states for psychosis. Neurosci Biobehav Rev 2015; 57:
199208.
8 Fusar-Poli P, Deste G, Smieskova R, Barlati S, Yung AR, Howes O et al. Cognitive
functioning in prodromal psychosis: a meta-analysis. Arch Gen Psychiatry 2012; 69:
562571.
9 Heinrichs RW, Zakzanis KK. Neurocognitive decit in schizophrenia: a quantitative
review of the evidence. Neuropsychology 1998; 12: 426.
10 Green MF, Kern RS, Braff DL, Mintz J. Neurocognitive decits and functional
outcome in schizophrenia: are we measuring the right stuff
? Schizophr Bull 2000;
26: 119.
11 Snitz BE, MacDonald AW, Carter CS. Cognitive decits in unaffected rst-degree
relatives of schizophrenia patients: a meta-analytic review of putative endophe-
notypes. Schizophr Bull 2006; 32:179194.
12 Bulik-Sullivan B, Finucane HK, Anttila V, Gusev A, Day FR, Perry JR et al. An atlas of
genetic correlations across human diseases and traits. bioRxiv 2015; 47:
12361241.
13 McIntosh AM, Gow A, Luciano M, Davies G, Liewald DC, Harris SE et al. Polygenic
risk for schizophrenia is associated with cognitive change between childhood and
old age. Biol Psychiatry 2013; 73:938943.
14 Consortium SWGotPG. Biological insights from 108 schizophrenia-associated
genetic loci. Nature 2014; 511: 421427.
15 International Schizophrenia C, Purcell SM, Wray NR, Stone JL, Visscher PM,
O'Donovan MC et al. Common polygenic variation contributes to risk of schizo-
phrenia and bipolar disorder. Nature 2009; 460:748752.
16 Gottesman I, Gould T. The endophenotype concept in psychiatry: etymology and
strategic intentions. Am J Psychiatry 2003; 160:636645.
17 Meyer-Lindenberg A, Weinberger DR. Intermediate phenotypes and genetic
mechanisms of psychiatric disorders. Na t Rev Neurosci 2006; 7:818827.
18 Braff D, Freedman R, Schork N, Gottesman I. Deco nstructing schizophrenia: an
overview of the use of endophenotypes in order to understand a complex dis-
order. Schizophr Bull 2007; 33:2132.
19 Hall M-H, Smoller JW. A new role for endophenotypes in the GWAS era: functional
characterization of risk variants. Harv Rev Psychiatry 2010; 18:6774.
20 Calkins ME, Merikangas KR, Moore TM, Burstein M, Behr MA, Satterthwaite TD
et al. The Philadelphia Neurodevelopmental Cohort: constructing a deep phe-
notyping collaborative. J Child Psychol Psychiatry 2015; 56: 13561369.
21 Robinson EB, Kirby A, Ruparel K, Yang J, McGrath L, Anttila V et al.
The genetic
architecture of pediatric cognitive abilities in the Philadelphia Neurodevelop-
mental Cohort. Mol Psychiatry 2014; 20:454458.
22 Satterthwaite TD, Elliott MA, Ruparel K, Loughead J, Prabhakaran K, Calkins ME
et al. Neuroimaging of the Philadelphia neurodevelopmental cohort. Neuroimage
2014; 86: 544553.
23 Holmes AJ, Hollinshead MO, OKeefe TM, Petrov VI, Fariello GR, Wald LL et al. Brain
Genomics Superstruct Project initial data release with structural, functional, and
behavioral measures. Sci Data 2015; 2: 150037.
24 Gur RC, Richard J, Calkins ME, Chiavacci R, Hansen JA, Bilker WB et al. Age group
and sex differences in performance on a computerized neurocognitive battery in
children age 8 21. Neuropsychology 2012; 26:251.
25 Gur RC, Richard J, Hughett P, Calkins ME, Macy L, Bilker WB et al. A cognitive
neuroscience-based computerized battery for efcient measurement of individual
differences: standardization and initial construct validation. J Neurosci Methods
2010; 187:254262.
26 Roalf DR, Ruparel K, Gur RE, Bilker W, Gerraty R, Elliott MA et al. Neuroimaging
predictors of cognitive performance across a standardized neurocognitive bat-
tery. Neuropsychology 2014; 28:161.
27 Meier MH, Caspi A, Reichenberg A, Keefe RS, Fisher HL, Harrington H et al. Neu-
ropsychological decline in schizophrenia from the premorbid to the postonset
period: evidence from a population-representative longitudinal study. Am J Psy-
chiatry 2014; 171:91101.
28 O'Donnell P. Adolescent onset of cortical disinhibition in schizophrenia: insights
from animal models. Schizophr Bull 2011; 37:484492.
29 Hibar DP, Stein JL, Renteria ME, Arias-Vasquez A, Desrivières S, Jahanshad N et al.
Common genetic variants inuence human subcortical brain structures. Nature
2015; 520:224229.
30 Ripke S, Sanders AR, Kendler KS, Levinson DF, Sklar P. Schizophrenia Psychiatric
Genome-Wide Associations Study (GWAS) Consortium. Genome-wide association
study idenes ve new schizophrenia loci. Nat Genet 2011; 43
: 969976.
31 Moore TM, Reise SP, Gur RE, Hakonarson H, Gur RC. Psychometric properties of
the Penn computerized neurocogn itive battery. Neuropsychology 2015; 29: 235.
32 Spearman C. General Intelligence, Objectively determined and measured. Amer J
Psychol 1904; 15:201293.
33 Germine LT, Hooker CI. Face emotion recognition is related to individual differ-
ences in psychosis-proneness. Psychol Med 2011; 41:937948.
34 Couture SM, Penn DL, Roberts DL. The functional signicance of social cognition
in schizophrenia: a review. Schizophr Bull 2006; 32(suppl 1): S44S63.
35 Penn DL, Corrigan PW, Bentall RP, Racenstein J, Newman L. Social cognition in
schizophrenia. Psychol Bull 1997; 121: 114.
36 Kohler CG, Walker JB, Martin EA, Healey KM, Moberg PJ. Facial emotion perception
in schizophrenia: a meta-analytic review. Schizophr Bull 2009; 36: 10091019.
37 Heimberg C, Gur RE, Erwin RJ, Shta sel DL, Gur RC. Facial emotion discrimination:
III. Behavioral ndings in schizophrenia. Psychiatry Res 1992; 42: 253265.
38 Walker E, Lewine J. Prediction of adult-onset schizophrenia. Am J Psychiatry 1990;
1: 47.
39 Walker EF, Grimes KE, D avis DM, Smith AJ. Childhood precursors of schizophrenia:
Facial expressions of emotion. Am J Psychiatry 1993; 150: 16541660.
40 Schiffman J, Lam CW, Jiwatram T, Ekstrom M, Sorensen H, Mednick S. Perspective-
taking decits in people with schizophrenia spectrum disorders: a prospective
investigation. Psychol Med 2004; 34: 15811586.
41 Hooker C, Park S. Emotion processing and its relationship to social functioning in
schizophrenia patients. Psychiatry Res 2002; 112:4150.
42 Fett A-KJ, Viechtbauer W, Penn DL, van Os J, Krabbendam L. The relationship
between neurocognition and social cognition with functional outcomes in schi-
zophrenia: a meta-analysis. Neurosci Biobehav Rev 2011; 35
:573588.
43 Brüne M, Abdel-Hamid M, Lehmkämper C, Sonntag C. Mental state attribution,
neurocognitive functioning, and psychopathology: what predicts poor social
competence in schizophrenia best? Schizophr Res 2007; 92:151159.
44 Farzin I, Platek SM, Panyavin IS, Calkins ME, Kohler C, Siegel SJ et al. Self-face
recognition and theory of mind in patients with schizophrenia and rst-degree
relatives. Schizophr Res 2006; 88: 151160.
45 Kohler CG, Bilker W, Hagendoorn M, Gur RE, Gur RC. Emotion recognition decit in
schizophrenia: association with symptomatology and cognition. Biol Psychiatry
2000; 48: 127136.
46 Bediou B, Asri F, Brunelin J, Krolak-Salmon P, D'Amato T, Saoud M et al. Emotion
recognition and genetic vulnerability to schizophrenia. Br J Psychiatry 2007; 191:
126130.
47 Eack SM, Mermon DE, Montrose DM, Miewald J, Gur RE, Gur RC et al. Social
cognition decits among individuals at familial high risk for schizophrenia. Schi-
zophr Bull 2010; 36: 10811088.
48 Glahn DC, Williams JT, McKay DR, Knowles EE, Sprooten E, Mathias SR et al. Dis-
covering schizophrenia endophenotypes in randomly ascertained pedigrees. Biol
Psychiatry 2015; 77:7583.
Schizophrenia polygenic risk and cognition
L Germine et al
6
Translational Psychiatry (2016), 1 7
49 Read J, Jv Os, Morrison A, Ross CA. Childhood trauma, psychosis and schizo-
phrenia: a literature review with theoretical and clinical implications. Acta Psy-
chiatr Scand 2005; 112: 330350.
50 Morgan C, Fisher H. Environment and schizophrenia: environmental factors in
schizophrenia: childhood traumaa critical review. Schizophr Bull 2007; 33:310.
51 Germine L, Dunn EC, McLaughlin KA, Smoller JW. Childhood adversity is asso-
ciated with adult theory of mind and social afliation, but not face processing.
PLoS ONE 2015; 10: e0129612.
52 Pinkham AE, Loughead J, Ruparel K, Overton E, Gur RE, Gur RC. Abnormal
modulation of amygdala activity in schizophrenia in response to direct-and
averted-gaze threat-related facial expressions. Am J Psychiatry 2011; 168:
293301.
53 Germine LT, Garrido L, Bruce L, Hooker C. Social anhedonia is associated with
neural abnormalities during face emotion processing. Neuroimage 2011; 58:
935945.
54 Wolf DH, Satterthwaite TD, Calkins ME, Ruparel K, Elliott MA, Hopson RD et al.
Functional Neuroimaging abnormalities in youth with psychosis spectrum
symptoms. JAMA Psychiatry 2015; 72:456465.
55 Calkins ME, Tepper P, Gur RC, Ragland JD, Klei L, Wiener HW et al. Project among
African-Americans to explore risks for schizophrenia (PAARTNERS): evidence for
impairment and heritability of neurocognitive functioning in families of schizo-
phrenia patients. Am J Psychiatry 2010; 167:459472.
56 Gur RC, Calkins ME, Satterthwaite TD, Ruparel K, Bilker WB, Moore TM et al.
Neurocognitive growth charting in psychosis spectrum youths. JAMA Psychiatry
2014; 71:366374.
57 Gur RC, Braff DL, Calkins ME, Dobie DJ, Freedman R, Green MF et al. Neurocog-
nitive performance in family-based and case-control studies of schizophrenia.
Schizophr Res 2015; 163:1723.
This work is licensed under a Creative Commons Attribution 4.0
International License. The images or other third party material in this
article are included in the articles Creative Commons license, unless indicated
otherwise in the credit line; if the material is not included under the Creative Commons
license, users will need to obtain permission from the license holder to reproduce the
material. To view a copy of this license, visit http://creativecommons.org/licenses/
by/4.0/
© The Author(s) 2016
Schizophrenia polygenic risk and cognition
L Germine et al
7
Translational Psychiatry (2016), 1 7