FACTORS AFFECTING PHENOTYPIC HOMOGENEITY

1
FACTORS AFFECTING PHENOTYPIC HOMOGENEITY IN
AUTISM GENETIC STUDIES
Vanessa Hus1, Catherine Lord1, Andrew
Pickles2 1University of Michigan Autism and
Communication Disorders Center 2University of
Manchester
Abstract
Language Acquisition
RRSBs
Savant Skills
Many chromosomal regions for susceptibility to
Autism Spectrum Disorders (ASDs) have been
identified but few have reached genome-wide
significance. In response, researchers have
attempted to increase the power of their analyses
by stratifying samples to increase phenotypic
homogeneity. Although homogeneity is often
defined by similar scores on a single variable,
groups identified in this way often differ in
other ways that may be directly pertinent to the
phenotype. The present study examines group
differences in chronological age, gender, Verbal
and Nonverbal IQ, and measures of autism
severity, as related to ADI-R items and domains
previously used for subsetting or QTL analysis.
Findings suggest that, out of a number of
potential grouping variables, only restricted and
repetitive behaviors associated with Insistence
on Sameness was independent of age, IQ, and
autism severity.
BACKGROUND Increased LOD scores for markers on
Chromosome 2q (Buxbaum, et al., 2001 Shao, et
al., 2002), 7q and 13q (Bradford, et al., 2001)
were found using subsamples of families
characterized by delayed phrase speech.
Chromosomal regions on 3q, 7q, and 17q that may
be associated with QTLs for age of first words
have been identified (Alarcón, et al., 2002
Alarcón, et al., 2005). Another language QTL
associated with the region on chromosome 7q was
attributable to a subset of families with the
earliest language development (Shao, et al.,
2003).
BACKGROUND Increased linkage on chromosome 15
was found in a subset of families with high
scores on the Insistence on Sameness (IS) factor
(Shao, 2003). Using Tadevosyans Compulsions
factor, four coding substitutions and 15 variants
were identified (Sutcliffe, et al., 2005), and
increased evidence for linkage was found on 17q
(McCauley, et al., 2004) and a serotonin
transporter intron 2 polymorphism (Mulder, et
al., 2005).
BACKGROUND Using a subset of families
demonstrating at least one savant skill according
to items loading on Tadevosyans Savant Skills
Factor, Nurmi, et al. (2003) found that LOD
scores increased for the chromosome region 15q.
Ma, et al. (2005) and Sutcliffe, et al. (2005)
also used savant skills scores, but did not
replicate Nurmis finding.
Groups were based on the Tadevosyan, et al.,
2003s Savant Skills Factor, with current and
ever scores of four ADI-R items visuospatial
ability, memory skill, musical ability, and
computation ability. Item scores were summed and
divided by total number of items to generate a
score between 0 and 1. Missing scores were coded
0. Participants were divided into two groups
Savant-positive ? at least one savant skill
(score 0.16) Savant-negative ? no savant
skills (score Groups were based on Cuccaro, et al. 2003s
factors Repetitive Sensory Motor Actions (RSMA)
and Resistance to Change. For clarity, we refer
to the latter as Insistence on Sameness (IS), a
term which others have called their similar
factor (Shao, et al., 2003 Bishop, et al., in
press). Eight ADI-R items were summed to yield
an RSMA (5 items) and IS (3 items) score.
Participants missing any items were excluded,
slightly reducing sample sizes. Confirmatory
factor analyses using MPlus3.0 replicated
Cuccaros factors in the present sample (see
Bishop, et al., in press). LRSMA ? (Low RSMA,
score 6) HRSMA ? (High RSMA, score 6)   LIS
? (Low IS, score 0) MIS ? (Medium IS,
score 1 or 2) HIS ? (High IS, score 2)
Groups were based on the ADI-R NDW ? (Not
Delayed Words, acquired words ? (Delayed Words, acquired words 24
months) NW ? (No Words, no words at time
of ADI-R)   NDP ? (Not Delayed Phrases,
acquired phrases (Delayed Phrases, acquired phrases 33
months) NP ? (No Phrases, no phrases at
time of ADI-R)
Introduction
CURRENT ANALYSES Chi-Square analyses showed no
significant differences in diagnostic category,
gender, or race between Savant-Negative (SN) and
Savant-Positive (SP) see Table 3. As shown in
Figures 5 and 6, T-tests indicated that
participants in the SN group were significantly
younger and had significantly lower Verbal and
Nonverbal IQs in comparison to SP. The SN also
had significantly higher ADI-R RRSB and Nonverbal
Comm. scores than SP, with no other significant
differences.
Approaches Used to Increase Phenotypic Homogeneity

• Using familiality to identify homogeneous
  phenotypes
• Assuming that, if a variable shows high
  familiality/broad heritability (through
  sibling/relative correlations), using it to
  subset families will increase genetic homogeneity.

Stratifying samples Research groups have
also used scores on the Autism Diagnostic
Interview Revised (ADI-R LeCouteur, Lord,
Rutter, 2003) to divide their samples into more
homogenous groups. For example
Language Acquisition Commonly used to stratify
samples is age of language acquisition, based on
age of first words or phrases. Delayed language
is defined on the ADI-R by age of first words
24 and age of first phrases 33-36 months.
Use of Empirically Derived Factors Several
research groups have conducted factor analyses
and found two distinct factors (though what they
call them has varied) within the RRSB domain
of the ADI-R that they have used to stratify
their samples Insistence on Sameness (IS)
and Restricted, Repetitive Sensory-motor
Behaviors (RSMA Shao, et al., 2003 Cuccaro, et
al., 2003 Bishop, et al., in press, Szatmari,
et al., in press). Another factor used for
stratification has been Tadevosyan-Lefer, et
al.s (2003) Savant Skills Factor.
Figure 5 Age and IQ by savant skills group
Figure 6 ADI-R scores by savant skills group
CURRENT ANALYSES Results for grouping by age at
first words indicated an association with
diagnosis, ?2(2, 983)34.95 and race, ?2 (2,
968)37.91 (both pTable 1.) One-Way ANOVAs indicated significant
associations with Verbal and Nonverbal IQs, but
no association with chronological age see Figure
1. As shown in Figure 2, there were significant
differences between word acquisition groups for
all ADI-R and ADOS domain scores.

• Identifying Quantitative Trait Loci (QTL)
• Some researchers have searched for loci that
  affect endophenotypes of specific behaviors
  related to autism.

Similarly, age at first phrases was significantly
associated with diagnosis, ?2 (2, 983)128.78 and
race, ?2 (2, 968)44.36 (both pgender see Table 1. As shown in Figures 1 2,
similar to age of first words, there were
significant differences between phrase
acquisition groups in mean Verbal and Nonverbal
IQs, as well as ADI-R and ADOS scores. Unlike
age of first words, there was a significant
effect of age on phrase acquisition groups.
CURRENT ANALYSES Analyses of RSMA groups
revealed a relationship between RSMA and
diagnosis, ?2 (2, 625)59.84, prace or gender see Table 2. As shown in Figures
3 and 4, ANOVAs and post hoc tests also indicated
that Verbal and Nonverbal IQs, as well as ADI-R
and ADOS domain scores, differed significantly
for all groups.
Research suggests that many aspects of ASD,
including some of the stratifications/approaches
to language or behavioral phenotypes that are
being used in genetic studies, are correlated
with variables such as age, IQ, and general
severity of autistic symptoms. The impact of
stratification on a single measure, such as
speech delay on other, more general variables,
such as IQ, has rarely been reported in genetic
papers. Knowledge about the impact of subsetting
based on major demographic and background
cognitive child variables could improve
understanding of phenotypes (e.g. Alárcon, et
al., 2005) and should increase replicability
across investigations. A better understanding of
the consequences of grouping according to a
single measure or factor would not negate
previous genetic findings, but could provide
important insights into potential interactions
that might decrease measurement error and allow
better interpretation of both extremes of a
subset. Some of these variables may even more
directly account for results attributed to
increased phenotypic homogeneity for a specific
ADI-R item or factor.
Discussion
Analyses conducted to examine the differences
between IS groups found that groups were only
significantly different in their ADI-R RRSB
scores, which would be expected. As shown in
Table 2 and Figures 3 4, there were no
significant differences between IS groups in
gender, race, diagnosis, IQ, ADI-R Social or
Communication Scores, or ADOS scores.

• Stratifications by language delay result in
  samples that differ in many more characteristics
  than onset of first words or phrases.
  Participants with delayed words or phrases
  consistently had lower Verbal and Nonverbal IQs
  and higher symptom scores on all defining domains
  of autism, as measured on both the ADI-R and
  ADOS.
• Results for Repetitive-Sensory Motor Actions
  (RSMA) followed similar patterns as language
  delay, with groups defined by RSMA scores
  differing on IQ and all symptom areas on the
  ADI-R and ADOS.
• The Savant skills domain followed another
  pattern with differences between its two groups
  strongly associated with age and Nonverbal and
  Verbal IQs. However, it was not associated with
  severity of autism symptoms. What this means is
  not clear, but suggests that stratifying on IQ
  before stratifying on savant skills might provide
  a more specific measure of the presence of
  specific strengths.
• Overall, findings suggest that stratifying
  samples by age and IQ first may result in more
  representative phenotypic groupings than using
  age of language acquisition, RSMA, or savant
  skills on their own. However, in contrast to
  results using those variables for stratification,
  Insistence on Sameness (IS) was relatively
  independent of gender, race, diagnosis,
  chronological age, Verbal and Nonverbal IQ and
  autism symptom domains on the ADI-R and ADOS.
  This suggests that the IS factor may be a useful
  strategy to stratify within an ASD sample because
  it offers a relatively independent dimension that
  varies considerably within ASD and other
  populations.

Figure 1 Age and IQ by language acquisition
group
Figure 3 Age and IQ by RRSB group
Methods
Participants
Figure 2 ADI-R scores by language acquisition
group
983 research participants and clinic referrals
for assessment of possible autism to clinics in
Illinois, Michigan, and North Carolina (812
male), 4 to 52 years old (M age7.75 years,
SD4.58), mostly Caucasian (N 760). Each
participant was evaluated using the Autism
Diagnostic Interview Revised (ADI-R), Autism
Diagnostic Observation Schedule (ADOS), a
cognitive assessment, and received a best
estimate diagnosis from a psychologist-child
psychiatrist team. Based on standardized
assessments, 663 were diagnosed with Autism, and
320 with an ASD (including 310 with PDD-NOS 10
with Aspergers Syndrome). Verbal and Nonverbal
IQ scores ranged from 2 to 153. Written informed
consent was obtained from all individuals and/or
their parents, in accordance with IRB-approved
protocols.
Figure 4 ADI-R scores by restricted
repetitive behavior group
References
(See Handout)
This research was funded by grants NIMH
R25MH067723 R01MH066496. Paper submitted to
Biological Psychiatry.