Molecular Psychiatry (2007) 12, 886–890

& 2007 Nature Publishing Group All rights reserved 1359-4184/07 $30.00
www.nature.com/mp

GUEST EDITORIAL

Endophenotypes in psychiatric genetics

Molecular Psychiatry (2007) 12, 886–890; doi:10.1038/ Whilst there is widespread acceptance of the
sj.mp.4002068 importance of these criteria, it is disconcerting how
few endophenotypes used in psychiatric research
There has been a recent surge in interest in the use of actually fulfil them. For example, there is limited
endophenotypes in psychiatric research, although the evidence for co-segregation with illness, let alone
concept was introduced to psychiatry by Gottesman heritability for many neurophysiological and neuro-
and Shields as long as 35 years ago.1 This has been cognitive measures that have been suggested as
driven by concerns about the limited success and schizophrenia endophenotypes.9,10 The extent to
poor reproducibility of existing approaches as well as which this matters depends upon the use to which
the fact that current diagnostic systems in psychiatry an endophenotype is being put and it is helpful to
by and large lack aetiological justification. In addi- consider the possibilities in some detail.
tion, the identification of disease endophenotypes Broadly speaking, endophenotypes are used under
offers the prospect of creating animal models relevant two circumstances. First, as originally proposed,1,2
to human psychopathology, which will be suitable for they are used to aid in the discovery of novel genes.
experimental approaches and greatly facilitate the The critical assumption here as we have seen is that
development and screening of novel therapeutics. the genetic architecture of the endophenotype is
Gottesman and Shields,1 who adapted the term simpler than that of the disease phenotype and this,
from insect biology, described endophenotypes as together with the opportunity to study clinically
internal phenotypes that lie on the pathway between unaffected relatives, who display the endophenotype
genes and disease. Fundamental to the concept is the by virtue of their increased risk of disorder, should
assumption that variation in an endophenotype will increase power. However there are a number of
depend upon variation in fewer genes than the more concerns that need to be addressed before the whole-
complex disease phenotype and therefore be more sale adoption of endophenotype approaches in gene-
tractable to genetic analysis.2 There has been a finding studies. First, it is often not clear how state-
number of attempts to devise criteria to define the independent many of the measures proposed really
optimal characteristics of an endophenotype.2–8 There are, with the potential for contamination not just by
is a general agreement that an endophenotype should fluctuations in course of illness and drug treatment,
occur at a higher frequency in individuals with the but also by factors such as smoking and menstrual
disease than the general population; moreover, this cycle phase.9,11 Second, there are uncertainties about
association should derive from shared genes. It reliability and particularly inter-laboratory variation
therefore follows that an endophenotype should be for many of the methods espoused. For example, there
heritable, tend to co-segregate with the illness in are no generally agreed protocols for reliably eliciting
multiply affected families, be found in unaffected electrophysiological deficits in pre-pulse inhibition
relatives of cases at a higher rate than in the general (PPI) and P50 sensory gating, which have been
population and ideally show evidence for common proposed as endophenotypes in schizophrenia, as
genetic risk factors from twin studies. In addition, if well as concerns regarding test–retest reliability.9 The
an endophenotype truly lies on the causal pathway neuroimaging community is still struggling with
between genes and disease, it should be state- issues of reliability as well as in developing methods
independent, in that it manifests in an individual to allow data obtained from different scanners to be
whether or not illness is active (although it may meaningfully combined.12 The neurocognition litera-
require challenge or provocation) and to an extent ture continues to be overwhelmed by the use of
that is not critically dependent upon the degree of multiple variations of tests for the same cognitive
activity of the illness. However, it is important to note domains and poor examination of reliability issues,
that even if a putative endophenotype satisfies all although consensus approaches and computer-deliv-
these criteria, this does not exclude the possibility ered batteries should help.10,13 Fortunately, many of
that it is epiphenomenal with respect to the disease. these issues are being addressed by the development
In other words, it might occur as a pleiotropic of multisite initiatives such as the Consortium on the
consequence of the risk gene or genes and not lie on Genetics of Endophenotypes in Schizophrenia.14,15
the disease pathway (see Figure 1 and Table 1). A Such projects represent an advance in this field in
further widely supported criterion is that an endo- aiming to address reliability concerns by standardiz-
phenotype should have good psychometric proper- ing methodologies for electrophysiological and neuro-
ties, especially reliability and validity, and be cognitive measures with regular monitoring of
sufficiently sensitive to detect individual differences. procedures, training and reliability.
 Editorial

887
1 G Endo P

Endophenotype on Disease Pathway

(ii) State dependent

2 eg medication

G P (i) Endo

Endophenotype is consequence of (i) disease symptoms or (ii) state

3
Endo

P
Endophenotype is epiphenomenom

4
G1 P1

G2 Endo

Different genes associated with endophenotype and disease

5
Enviro Endo P

Endophenotype arises from non-genetic factors

Figure 1 This shows in simplified, schematic terms some of the possible relationships between putative endophenotypes,
gene and disease. In reality, different combinations of these simplified scenarios are likely. G, genes; Enviro, environmental
factors; Endo, putative endophenotype; P, disease phenotype.

The main justification for the use of endopheno- lity of phenotypes that are more reliable and objective
types in gene-finding studies is their assumed genetic than those based on patient report as well as the
simplicity in comparison to complex disease pheno- potential to harness the increased power that accrues
types. Originally, it was hoped that this would from the use of quantitative phenotypes. However, it is
allow disorders such as schizophrenia to be decom- also important to note that the cost of measuring some
posed into a set of single-gene deficits. Most recent endophenotypes, particularly those based on neuro-
commentators accept that this is unlikely although imaging, currently prohibits their application to the
they still believe that the genetic architecture of large samples required for gene-finding studies.
endophenotypes will be significantly simpler than It is generally assumed that to be useful as an
that of disease. Recently, even this assumption has endophenotype, a trait should lie on the causal
been challenged on both empirical and theoretical pathway between genes and the disorder.2–7 Indeed,
grounds.16 The authors acknowledge that their this would seem to be fundamental to the concept.
work has some limitations, but it seems likely that However, this is one of the most difficult criterion to
endophenotypes, like complex diseases, will reflect demonstrate. It is often indirectly inferred if a trait
the operation of many genes of small effect. This work that is associated with a disorder is found in
cautions against unfettered enthusiasm, but it would unaffected relatives. However, as we have seen,
seem to be premature to exclude the possibility that this does not exclude the possibility that the trait
endophenotypes will be useful in defining more bears an epiphenomenal relationship with respect
aetiologically homogeneous groups. Moreover, the to genes and the disorder (Figure 1 and Table 1). To
use of carefully designed measures offers the possibi- prove causality, longitudinal studies are required,

Molecular Psychiatry
Molecular Psychiatry

888
Table 1 The relationship between genes and a putative endophenotype impacts on the various criteria proposed for endophenotypes

From Genotype/Endophenotype Heritable Associated State m in Cosegregates Associated Useful in On gene-

Figure 1 relationship in independent unaffected with illness with disease gene-finding disease
population relatives allele experiments pathway

1 | | | | | | | |

2  /| |   | |  

Editorial
Endo
G
3 | | | | | | | 
P

4 |  |     

5  | | |/  |/    

See text for further explanation.

 Editorial

889
preferably in a genetically informative design,8 However, a number of dangers present themselves
although even here intervention studies are really in the absence of robust evidence for disease associa-
required to prove that the putative endophenotype tion with a specific allele or alleles, as is currently the
lies on the causal path to disorder. It is worth pointing case for most, if not all, associations in psychiatric
out the proof that a trait lies on the disease pathway genetics. In particular, there is the potential for ill-
is not strictly speaking required for a trait to be a substantiated disease associations to gain spurious
useful aid to finding disease genes; a trait that is support from associations with endophenotypes; the
epiphenomenal will do just as well here as long as it combination of uncertain genetic associations leading
simplifies the genetic architecture by defining a to multiple testing of alleles or haplotypes together
more genetically homogeneous disease subgroup with the analysis of multiple endophenotypes leads
or identifies carriers of the risk genotype among to the potential generation of false positives that gain
unaffected relatives. However, some might argue that spurious credibility from appeals to the endopheno-
the more general term ‘biomarker’ is more appropriate type concept. The solutions to this problem are
in this instance. clear, if not easily attained.23 Such studies should be
Given these difficulties, it is perhaps not surprising based upon robust evidence for association with
that to date there are few examples of the use of the primary disease phenotype, which will usually
endophenotypes leading to the identification of novel require large samples, low P-values and evidence of
risk genes or even robust linkages for psychiatric replication. Fine mapping studies are then required to
phenotypes. An elegant and instructive exception is a allow exclusion of many polymorphisms in the region
series of studies by the Collaborative Study of the as causal variants. This will then leave a subset of
Genetics of Alcoholism (COGA) in which electro- putative candidate variants that will usually show
physiological endophenotypes were studied in linkage disequilibrium with each other. These will
addition to clinical diagnoses.17 Here the use of then be suitable for further genetic studies of
endophenotypes substantially improved both the endophenotypes, where the per-gene effect size might
strength and localization of linkage findings and be larger than those involving the clinical phenotype,
allowed the identification of GABRA2 and CHRM2 although it is here that researchers will need to
as genes associated with predisposition to alcohol remain vigilant that the endophenotypes studied are
dependence.17,18 In this case, the utility of analysing true mediating variables.
electrophysiological data seems to have been that it It has not been our intention to try and dissuade
allowed broad linkage signals reflecting linkage to researchers from using the endophenotype approach.
more than one locus to be decomposed into consti- Indeed, we see great potential for its use in
tuent signals reflecting variation in individual genes; psychiatric genetics. However, we do believe that
gains in power resulted from a combination of greater the use of endophenotypes needs to be more care-
genetic homogeneity and the use of quantitative fully considered and more attention is given to the
phenotypes. Further studies have used individual choice of possible measures with specific use and
and combined endophenotypes in genetic linkage setting in mind. In our view, the use of endopheno-
studies of schizophrenia, although as yet these types in gene finding, by both linkage and association
studies have not resulted in the unambiguous approaches, is most likely to add value when multiple
identification of susceptibility genes.19–21 measures are employed in combination with
The second use of endophenotypes is to study the clinical data.17,20,24 Putative endophenotypes should
functional consequences of risk alleles rather than be chosen on the basis of robust evidence that they
as a means of identifying novel risk genes.22 This are not only associated with the disease but also
approach is becoming increasingly popular as evi- that association reflects shared genes (Table 1).
dence for susceptibility genes accumulates since it is Fortunately, increasing attention is now being given
perceived as holding much promise for establishing to this latter issue.9,10,25–28 Whether or not the genetic
disease mechanisms, for example, by seeking associa- architecture of endophenotypes will be substantially
tions of risk alleles with structural or neurocognitive simpler overall than that of the diseases to which
variables.22 However, there are dangers arising they are related remains to be seen,16 but there are
from the fact that, as we have seen, it is difficult to encouraging signs that, from the work from COGA
establish that traits are actually on the disease cited above and work on non-psychiatric disorders,29–
31
pathway even if all the generally agreed criteria for their use might make complex traits more tractable
an endophenotype have been met (Figure 1 and to genetic analysis.
Table 1). Most would agree that if a trait is associated There also seems to be great potential to use
with a disease, satisfies the other criteria for an objectively measurable endophenotypes to illuminate
endophenotype and is associated with the presence of the brain mechanisms linking specific gene actions
a robustly associated risk allele, then this is prima and products to the subjective experience of psycho-
facie, if not conclusive, evidence that it lies on the pathological symptoms. However, this work must be
disease pathway. Indeed once these criteria are met, based upon robust genetic associations that have been
there are statistical approaches available to determine simplified by fine mapping, and researchers should
the extent to which a putative endophenotype might remain vigilant for evidence of pleiotropy. Given the
mediate a gene–disease association.4 complexities of psychiatric phenotypes and our lack

Molecular Psychiatry
 Editorial

890
of understanding of disease pathogenesis, the penal- 13 Green MF, Nuechterlein KH. The MATRICS initiative: developing
ties for not adhering to this could be severe. a consensus cognitive battery for clinical trials. Schizophr Res
Finally, we believe that the above considerations are 2004; 72: 1–3.
14 Calkins ME, Dobie DJ, Cadenhead KS, Olincy A, Freedman R,
relevant to the question of nomenclature. There are Green MF et al. The consortium on the genetics of endopheno-
some who prefer the term ‘intermediate phenotype’ to types in schizophrenia: model recruitment, assessment, and
‘endophenotype’ because it implies a position on the endophenotyping methods for a multisite collaboration.
pathway between genes and disease. It is for precisely Schizophr Bull 2007; 33: 33–48.
15 Braff DL, Freedman R, Schork NJ, Gottesman II. Deconstructing
this reason that we prefer the more mechanistically
schizophrenia: an overview of the use of endophenotypes in order
neutral term ‘endophenotype’ at least until convincing to understand a complex disorder. Schizophr Bull 2007; 33: 21–32.
evidence for mediation has been obtained. 16 Flint JI, Munafo MR. The endophenotype concept in psychiatric
genetics. Psychol Med 2007; 37: 163–180.
17 Dick DM, Jones K, Saccone N, Hinrichs A, Wang JC, Goate A et al.
Endophenotypes successfully lead to gene identification: results
Acknowledgments from the collaborative study on the genetics of alcoholism. Behav
Genet 2006; 36: 112–126.
We thank Irv Gottesman, Mike O’Donovan and Nick 18 Hinrichs AL, Wang JC, Bufe B, Kwon JM, Budde J, Allen R et al.
Craddock for their comments on this paper. Functional variant in a bitter-taste receptor (hTAS2R16) influences
risk of alcohol dependence. Am J Hum Genet 2006; 78: 103–111.
19 Freedman R, Coon H, Myles-Worsley M, Orr-Urtreger A, Olincy A,
JTR Walters and MJ Owen Davis A et al. Linkage of a neurophysiological deficit in
1
Department of Psychological Medicine, School schizophrenia to a chromosome 15 locus. Proc Natl Acad Sci
of Medicine, Cardiff University, Cardiff, UK USA 1997; 94: 587–592.
E-mail: owenmj@cardiff.ac.uk 20 Hallmayer JF, Kalaydjieva L, Badcock J, Dragovic M, Howell S,
Michie PT et al. Genetic evidence for a distinct subtype of
schizophrenia characterized by pervasive cognitive deficit. Am J
Hum Genet 2005; 77: 468–476.
References 21 Paunio T, Tuulio-Henriksson A, Hiekkalinna T, Perola M, Varilo T,
Partonen T et al. Search for cognitive trait components of
1 Gottesman II, Shields J. Schizophrenia and Genetics; a Twin Study schizophrenia reveals a locus for verbal learning and memory on
Vantage Point. Academic Press: New York, 1972. 4q and for visual working memory on 2q. Hum Mol Genet 2004; 13:
2 Gottesman II, Gould TD. The endophenotype concept in psychia- 1693–1702.
try: etymology and strategic intentions. Am J Psychiatry 2003; 160: 22 Meyer-Lindenberg A, Weinberger DR. Intermediate phenotypes
636–645. and genetic mechanisms of psychiatric disorders. Nat Rev
3 Gould TD, Gottesman II. Psychiatric endophenotypes and the Neurosci 2006; 7: 818–827.
development of valid animal models. Genes Brain Behav 2006; 5: 23 Todd JA. Statistical false positive or true disease pathway? Nat
113–119. Genet 2006; 38: 731–733.
4 Waldman ID. Statistical approaches to complex pheno- 24 Price GW, Michie PT, Johnston J, Innes-Brown H, Kent A, Clissa P
types: evaluating neuropsychological endophenotypes for atten- et al. A multivariate electrophysiological endophenotype, from a
tion-deficit/hyperactivity disorder. Biol Psychiatry 2005; 57: unitary cohort, shows greater research utility than any single
1347–1356. feature in the Western Australian family study of schizophrenia.
5 Bearden CE, Freimer NB. Endophenotypes for psychiatric dis- Biol Psychiatry 2006; 60: 1–10.
orders: ready for primetime? Trends Genet 2006; 22: 306–313. 25 Cannon TD, Huttunen MO, Lonnqvist J, Tuulio-Henriksson A,
6 Skuse DH. Endophenotypes and child psychiatry. Br J Psychiatry Pirkola T, Glahn D et al. The inheritance of neuropsychological
2001; 178: 395–396. dysfunction in twins discordant for schizophrenia. Am J Hum
7 Munafo MR. Candidate gene studies in the 21st century: meta- Genet 2000; 67: 369–382.
analysis, mediation, moderation. Genes Brain Behav 2006; 26 Cannon TD, Thompson PM, van Erp TGM, Toga AW, Poutanen
5(Suppl 1): 3–8. V-P, Huttunen M et al. Cortex mapping reveals regionally specific
8 Viding E, Blakemore S-J. Endophenotype approach to develop- patterns of genetic and disease-specific gray-matter deficits in
mental psychopathology: implications for autism research. Behav twins discordant for schizophrenia. Proc Natl Acad Sci USA 2002;
Genet 2006; 37: 51–60. 99: 3228–3233.
9 Turetsky BI, Calkins ME, Light GA, Olincy A, Radant AD, 27 van Erp TGM, Saleh PA, Huttunen M, Lonnqvist J, Kaprio J,
Swerdlow NR. Neurophysiological endophenotypes of schizo- Salonen O et al. Hippocampal volumes in schizophrenic twins.
phrenia: the viability of selected candidate measures. Schizophr Arch Gen Psychiatry 2004; 61: 346–353.
Bull 2007; 33: 69–94. 28 Hall MH, Schulze K, Rijsdijk F, Picchioni M, Ettinger U, Bramon E
10 Gur RE, Calkins ME, Gur RC, Horan WP, Nuechterlein KH, et al. Heritability and reliability of P300, P50 and duration
Seidman LJ et al. The consortium on the genetics of schizophrenia: mismatch negativity. Behav Genet 2006; 36: 845–857.
neurocognitive endophenotypes. Schizophr Bull 2007; 33: 49–68. 29 Anderson GG, Leaves NI, Bhattacharyya S, Zhang Y, Walshe V,
11 Swerdlow NR, Light GA, Cadenhead KS, Sprock J, Hsieh MH, Broxholme J et al. Positive association to IgE levels and a physical
Braff DL. Startle gating deficits in a large cohort of patients with map of the 13q14 atopy locus. Eur J Hum Genet 2002; 10: 266–270.
schizophrenia: relationship to medications, symptoms, neuro- 30 Keating M, Atkinson D, Dunn C, Timothy K, Vincent GM, Leppert
cognition, and level of function. Arch Gen Psychiatry 2006; 63: M. Linkage of a cardiac arrhythmia, the long QT syndrome, and the
1325–1335. Harvey ras-1 gene. Science 1991; 252: 704–706.
12 Cho RY, Ford JM, Krystal JH, Laruelle M, Cuthbert B, Carter CS. 31 Zhang Y, Leaves NI, Anderson GG, Ponting CP, Broxholme J,
Functional neuroimaging and electrophysiology biomarkers for Holt R et al. Positional cloning of a quantitative trait locus on
clinical trials for cognition in schizophrenia. Schizophr Bull 2005; chromosome 13q14 that influences immunoglobulin E levels and
31: 865–869. asthma. Nat Genet 2003; 34: 181–186.

Molecular Psychiatry