Analyzing Atypical Development: Causes and Comorbidities

1
Analyzing Atypical DevelopmentCauses and
Comorbidities

• Bruce F. Pennington
• University of Denver

2
Outline

• Overall Framework
• Phenotype Analyses of RD
• A. Cognitive components and subtypes of RD
• B. Comorbidities of RD
• III. The interaction of genotypic and phenotypic
  analyses of RD
• A. Do different risk loci for RD act on
  different
• cognitive components?
• B. How do we explain comorbidities?
• IV. Phenotypic and genotypic analyses of the
  relations
• among RD and speech and language disorders
• A. Children at Risk for RD have SSD
• (Lefly Pennington, 2001)
• B. Coheritability of SSD and RD (Tunick, M.A.)
• C. Implicit Phonological Representations in RD
• (Boada, Ph.D.)
• D. SSD children have preliteracy deficits
• (Raitano, M.A.)

3
Figure 1. Complex Disease Model
Non-Independence at each Level Gx E
interaction G-E Correlation Interactive
Development Comorbidity
Level of Analysis Etiologic Risk and
Protective Factors Cognitive
Causes Complex Behavioral Disorders
G1
E1
G2
E2
G3

Phon
Sem
OSM
RD
SSD
SLI
KEY G genetic risk or protective factor, E
environmental risk or protective factor, Phon
Implicit phonological development, Sem Semantic
development, OSM Orosensory and oromotor
development, RD Reading disability, SSD
Speech sound disorders, SLI Specific language
impairment

4
(No Transcript)
5
(No Transcript)
6
Common Complex Diseases Multiple G E risk
factors non-CNS Cardiovascular
Disease Hypertension Cancer Diabetes Obesi
ty CNS Stroke Alzheimer Disease All
Psychiatric Disorders Some apparently
environmental conditions, like PTSD motor
vehicle accidents, have a G background
7
Behavioral Genetics Key Ideas Caveats

• Given that 40 of genes uniquely expressed in
  brain and genes are polymorphic, then genetic
  influences on individual differences in behavior
  are virtually Inevitable.
• But environmental influences important too, both
  
• for differences and universals.
• Genetic influences are probabilistic, not
  deterministic recipe not blueprint.
• Epigenesis is important Complex developmental
• pathways run from genes to behavior.

8
Three Components of Typical Development
Shared Genotype (99.9)
Epigenesis or Chance
Species Typical Environment
Atypical development results when genetic or
environmental risk factors alter these
interactions.
9
Are there Multiple Subtypes of Developmental
Dyslexia?

• Theoretically, there could be a subtype
  associated with each cognitive component of
  reading text.
• If so, genetic studies would be difficult
  because we would need to distinguish all these
  subtypes.

10
Evidence for Subtypes of Developmental Dyslexia
(Castles and Coltheart, 1993 Manis et al., 1996)

• 20 Surface -- selectively bad on exception word
  reading (orthographic coding)
• 20 Phonological -- selectively bad on nonword
  reading (phonological coding)
• 60 Mixed -- bad on both
• Errors in the surface subtype (but not the
  phonological subtype) are like those of younger
  normal readers at a similar word recognition
  level (Bryant Impey, 1986 Manis et al., 1996
  Stanovich et al., 1998). Substantial genetic
  covariance between orthographic and phonological
  coding deficits (Olson et al., in press)
• Therefore, little evidence for a distinct surface
  subtype. Most developmental dyslexics have a
  phonological coding deficit.

11
Comorbidities of RD
1.) Externalizing disorders ADHD, CD, ODD a)
ADHD accounts for relation to CD, ODD b) Inatt.
Subtype of ADHD, rather than HI c) Greater in
males with RD 2.) Internalizing symptoms
Depression and Anxiety a) Greater in females
with RD b) Are secondary to RD and not
genetically correlated 3.)
Speech and language disorders - Why?
12
Steps in a Genetic Analysis
Question Method 1. Is it familial? Familial
Risk Studies 2. Is familiality genetic? Twin
Studies, Adoption Studies 3. What is the
mode of Segregation Analysis
transmission? 4. Where are the gene(s)? Linkage
and Association Analyses
13
Family and Twin Studies of Normaland Abnormal
Reading
1. Normal and abnormal reading runs in
families A. In general population, first
degree relatives correlated at .40 B.
First degree relatives of RD proband 4-14 times
more likely to have RD
(Hallgren, 1950 Finucci et al, 1976
Vogler et al, 1985 Gilger et al, 1991) 2.
Normal (h2 .56 - .73) and abnormal (h2g .50
.11) variations in reading are moderately
heritable (e.g. DeFries Gillis, 1993)
14
Recurrence Risk

• Rate in Family Members
• Base Rate in Population
• Rate in Base Recurrence
• Disorder Sibs Rate Risk
• Huntingtons Disease .50 .0001 5000
• Cystic Fibrosis .25 .0004 500
• Alzheimers Disease .40 .10 4
• Autism .05 .0004 125
• Reading Disability .40 .05 8
• ADHD .35 .05 7

15
The DeFries-Fulker multiple regression model
(DeFries Fulker, 1985 1988)
The distribution of reading achievement scores in
the population
µ
16
Probands are selected due to reading difficulties
RD Cutoff
µ
MZ and DZ Proband Means
17
Cotwin means if reading difficulties are due to
nonshared environmental influences
MZ and DZ Cotwin Means
18
Cotwin means if reading difficulties are due to
shared and nonshared environmental influences
µ
MZ and DZ Cotwin Means
19
Cotwin means if reading difficulties are due to
genetic influences
µ
MZ Cotwin Mean
DZ Cotwin Mean
20
The Basic Regression Model

• C B1 P B2 R
• C is the cotwin score
• P is the proband score
• R is the coefficient of relationship
• (1 for MZ pairs, 0.5 for DZ pairs)
• B1 twin resemblance independent of zygosity.
• B2 A direct estimate of h2g
• (estimates twice the difference between the
  MZ and DZ cotwin means after
• covariance adjustment for differences in
  scores of MZ and DZ probands)

21
Etiology of reading deficits(Wadsworth et al.,
2000)
h2g 0.57 .08
MZ and DZ Proband Reading Means -2.80 SD
DZ Cotwin Reading Mean -1.55 SD
MZ Cotwin Reading Mean -2.41 SD
22
Is heritability worthless?
23
Interpreting heritability estimates

• Heritability is population and time specific
• Heritability within groups tells us nothing about
  the causes of differences between groups
• Heritability does not apply to individuals
• Human universals are typically not heritable
  (but are usually genetic!)

24
Assumptions of the twin design

• No nonadditive genetic influences
• Increases h2, decreases c2
• No assortative mating
• Decreases h2, increases c2
• No epistasis
• Could increase or decrease h2 or c2
• No rater bias or sibling contrast effects
• Typically increase h2, decrease c2

25
Linkage Analysis
Two genes are said to be linked when they are
positioned close together on the same chromosome,
such that recombination between them is
significantly decreased. If a trait is found to
be linked to a marker with known position, the
position of the gene for the trait is then known.
26
Y Chromosome Map
27
Recombination and Linkage (Thomas Hunt Morgan)
Q1 Why is Mendels second law of
independent assortment of genes mostly
correct? A1 Because of recombination, the
dealer that shuffles the genetic
deck. Q2 Why not completely correct? A2 Recomb
ination does not shuffle perfectly adjacent
genes can be linked.
28
(No Transcript)
29
(No Transcript)
30
(No Transcript)
31
(No Transcript)
32
(No Transcript)
33
(No Transcript)
34
(No Transcript)
35
(No Transcript)
36
Recombination the basis of linkage analysis

• Recombination occurs about 1.5 times per
  chromosome.
• Recombination occurs more in some areas (hot
  spots) than others
• The presence of one crossover inhibits the
  development of others nearby.
• If a marker locus and a disease locus are far
  apart, recombination will typically occur between
  the loci.
• If a marker locus and a disease locus are very
  close together, recombination will rarely occur
  between the two loci.

37
Recombination Frequency
Initial parental chromosomes with varying
distance between marker locus and disease locus
Percentage of the time marker and disease
co-occur after meiosis
50
75
87
95
99
Recombination Frequency
.50
.25
.13
.05
.01
38
Genetic Distance

• Centimorgan the distance at which two loci
  recombine only 1 of the time

Recombination Frequency
.50
.25
.13
.05
.01
Centimorgans apart
50
25
13
5
1
39
Classical Linkage Analysis

• Works well for rare single-gene disorders
• the mode of genetic transmission must be
  specified.
• Large family pedigrees are necessary to provide
  sufficient statistical power.

40
Weaknesses of CLA

• Most disorders are polygenic.
• Most disorders are heterogeneous.
• Developmental / cohort effects.
• The dreaded categorical phenotype.

41
Quantitative Trait Loci

• QTL are genes of relatively small effect that
  influence a multifactorial trait.
• The selected sibling-pair design provides
  excellent power to detect the effect of a QTL.
• This is a simple extension of the DF regression
  method for twins.

42
Summary of the selected sib-pair design to test
for linkage

• A sample of sibling pairs is selected in which at
  least one sib scores above the cutoff on the
  phenotype of interest.
• Genotypes are collected from parents and sibs.
• Phenotypes are measured for the sibs, but are not
  necessary for the parents.
• Complex algorithms are utilized to estimate the
  proportion of alleles shared identical by descent
  (IBD) at each marker locus.
• An extension of the DF method is used to test if
  each marker locus is linked to a QTL associated
  with the phenotype.

43
Applying the DF regression method to search for
QTLs
C B1 P B2 ? K C is the
cosib score P is the proband score ? is
the estimated proportion of alleles shared IBD at
the marker K is the regression
constant B2 tests for significant linkage at
that chromosomal location
44
(No Transcript)
45
(No Transcript)
46
(No Transcript)
47
(No Transcript)
48
RD and ADHD are Comorbid Why?

• Rejected Hypotheses
• Not a selection artifact Comorbidity found in
  population samples (eg Willcutt
• Pennington, 2000)
• Not a secondary phenocopy Comorbid subjects
  have both EF and PA deficits
• (Willcutt et al, 2001), contrary to Pennington
  et al (1993)
• Not cross-assortment (Friedman et al,
  submitted)
• Supported Hypothesis Shared Etiological
  Influences
• Bivariate h2g for RD and ADHD (Stevenson et al,
  1993 Light et al, 1995)
• Bivariate h2g for RD and Inatt is about .40,
  whereas NS for RD and HI
• (Willcutt et al, 2000)
• QTL for RD on 6p21.3 is also linked to ADHD and
  shows bivariate linkage
• with RD phenotypes (Willcutt et al, in press)

49
An example with some real data(Willcutt et al.,
2002)

• Sample 48 sib-pairs in which at least one sib
  has ADHD.
• Markers 8 polymorphic markers spanning 14.7
  centiMorgans of the short arm of chromosome 6p.
• Markers in this area have been shown to be linked
  to a QTL for RD in four independent samples
  (Cardon et al., 1994, 1995 Fisher et al., 1999
  Gayan et al., 1999 Grigorenko et al., 1997)

50
Description of Markers
51
Symptoms of ADHD exhibited by the cosibs of
probands with ADHD as a function of the
proportion of alleles shared IBD at marker D6S105
52
Linkage of ADHD to markers on chromosome 6
P .001
D6S291
D6S461
D6S276
D6S105
D6S306
D6S439
D6S258
D6S1019
6pter
6cen
5 cM
53
Bivariate linkage of RD and ADHD to markers on
chromosome 6p
P .01
D6S291
D6S461
D6S276
D6S105
D6S306
D6S439
D6S258
D6S1019
6pter
6cen
5 cM
54
Summary of Linkage Analysis

• Family-based technique.
• Designed to screen large sections of the genome
  for possible genes.
• Strength can be detected over relatively long
  distances if the gene has a large effect.
• Weakness Due to relative infrequency of
  recombination across the genome, regions of
  significant linkage are often quite large, and it
  is hard to detect small effects.

55
Genetic Association and Candidate Gene Studies

• compare large samples of individuals with and
  without the disorder.
• test if an allele at a specific genetic locus is
  associated with the disorder in the population as
  a whole.
• a simple ?2 tests significance of association
• Example
• ADHD 40 have 7-repeat Dopamine D4 receptor
  allele
• Control 20 have 7-repeat allele

56
Possible implications of significant association

• True functional association
• identifies a polymorphism in a gene that plays a
  direct functional role in the disorder.
• Population Stratification

57
Population Stratification

• potential fatal flaw of the case-control design
• Different ethnic groups often have very different
  gene frequencies

58
Population frequency of the 7-repeat risk
allele of the dopamine D4 receptor
gene(summarized from data described by Ken Kidd,
M.D. http//www.med.yale.edu/genetics/kkidd/)
59
An alternative to the case-control designThe
sibling-pair association study

• A proband is selected with the disorder.
• Test whether a sibling who shares the risk
  allele is more likely to also have the disorder
  than a sibling without the risk allele (or has a
  higher mean score in dimensional analyses)
• The use of siblings controls for population
  stratification because full siblings are from the
  same ethnic group
• What if the proband has no sibling?

60
The Transmission Disequilibrium Test
 

• A proband with the disorder is identified.
• The proband and both parents are genotyped.
• We test whether the affected proband receives the
  risk allele from a heterozygous parent
  significantly more often than the other allele.

 
 
 
61
A Subset of the Genes Associated with ADHD

• Dopamine Transporter Gene (e.g., Cook et al.,
  1995)
• Dopamine D4 receptor (e.g., LaHoste et al., 1996)
• Dopamine D2 receptor (e.g., Comings et al., 1999)
• Dopamine D5 receptor (e.g., Daly et al., 1999)
• Dopamine-beta-hydroxylase gene (e.g., Daly et
  al., 1999)
• Serotonin Transporter gene (Manor et al., 2001)
• Serotonin receptor 2B (Quist et al., 2001)
• Monoamine oxidase A (Jiang et al., 2000)
• DxS7 region (Jiang et al., 2000)
• A gene on chromosome 6p (Willcutt et al., 2002)

62
IV. Developmental Dyslexia is a Language
Disorder

• Phonological development is difficult,
  protracted and
• exhibits wide individual differences.
• Studies of infants and toddlers born to
  dyslexic parents
• find problems in speech perception (Lyytinen,
  1999)
• and expressive syntax (Scarborough, 1990).
• Children with speech production problems have
  higher
• rates of later dyslexia and dyslexics have
  higher rates
• of earlier speech production problems. The
  two disorders
• are co-familial and co-heritable.
• We have begun a study to examine the genetic
  and
• cognitive relations between dyslexia and
  speech and
• language disorders.

63
Phenotypic Overlap Among SSD, LI, RD
SSD
RD
LI
64
Main Hypotheses Genetic

• 1.) The comorbidity among these three disorders
  is caused
• by a partly shared genetic etiology.
  Therefore, some of
• the risk loci already
  identified for RD will also be risk loci
• for SSD and LI.
• 2.) This genetic overlap will vary by subtypes
  of SSD LI.
• For SSD, the overlap will be found for SD but
  not
• for Dyspraxia.
• For LI, the overlap will be found for
  deficits in
• structural (lexical syntactic) but not
  functional
• language.

65
Main Hypotheses Neuropsychological
1.) The shared neuropsychological phenotype
among SSD, LI, RD is a deficit in
implicit phonological representations.
2.) This shared neuropsychological phenotype
will vary by subtypes of SSD LI present
in SD and structural LI, but absent in
Dyspraxia and PLI.
66
Recent Relevant Evidence From My Lab
A. Children at Risk for RD have SSD (Lefly
Pennington, 2001) B. Coheritability of SSD and
RD (Tunick, M.A.) C. Implicit Phonological
Representations in RD (Boada,
Ph.D.) D. SSD children have preliteracy
deficits (Raitano, M.A.)
67
A. Children at Risk for RD(Pennington Lefly,
2001)
Who Became Dyslexic? 34 (22/64) of High-Risk
Group 6 (3/49) of Low-Risk Group Relative Risk
5.7
68
Speech Production Problems(Pennington Lefly,
2001)
Parent Report Age 5 28 (18/64) of High-Risk
Group plt.05 13 (6/48) of Low-Risk Group GFW
Ratings Age 5 58 (11/19) of High-Risk
RD plt.001 0 (0/11 of High-Risk NRD Late 8
Ratings Age 7 58 (7/12) of High-Risk
RD plt.05 0 (0/11) of High-Risk NRD
69
Shared Etiologies Among SSD, LI, RD

• SSD LI are coheritable (Bishop et al., 1995)
• SSD RD are co-familial (Lewis et al, 1989
• 1990 1992)
• LI RD are coheritable (Bishop et al., 1999)
• So far, no molecular tests of shared etiologies

70
B. Coheritability of SSD and RD (Tunick, MA
Thesis)

• In CLDRC Twin sample, retrospective SSD and RD
  are comorbid
• SSD
• -
• Retrospective SSD is heritable
• MZ concordance 76, DZ33
  
• X2 83.77, p, lt.001 RD
• RD and SSD are co-heritable -
• MZ cross-concordance 28, DZ 19
• X2 9.35, p, lt.05
• X271.4, plt.001

153 655 19 81
77 1096 7 93
71
Shared Neuropsychology Among SSD, LI, RD

• All three disorders share deficits in
  phonological
• memory, as measured by NW repetition (Bishop
  et al.,1996 Brady, 1997 Gathercole Baddeley,
  1990
• Snowling, 1981)
• All three disorders share deficits in phoneme
• awareness (Bishop et al, 1995 Lewis
  Freebairn,
• 1992 Wagner Torgesen, 1987)
• Deficit in NW rep. is coheritable with LI
  (Bishop, et al,
• 1996) and PA deficit is coheritable with RD
  (Olson et
• al, 1994)
• These results could be explained by a problem
  in the
• development of phonological representations

72
C. Implicit Phonological Representations in
RD(Boada, Ph.D. Thesis)

• Compared RD only, RD SSD/LI, CA, and RA
  controls. N20 each.
• All four groups similar in SES, gender NVIQ.
  First 3 groups similar
• in age (11-13y).
• Used three measures of implicit phonological
  representations
• - Lexical gating
• - Priming
• - Syllable similarity (Treiman Breaux, 1982)
• No differences between RD only and RD SSD/LI
  groups

73
Lexical Gating(sum of gates required)
RD RDSSD/LI CA RA 93.6 95.3 80.8
87.5 (11.0) (10.4) (6.8)
(10.9) F(2,77) 13.12, plt.001 RD
combined gt CA, RA
74
Priming Task
75
Syllable Similarity (c-s)
76
D. SSD and RD Project (Raitano, et al, in
revision)

• SSD (N101), Controls (N 41)
• Similar in age (5-6y), parental education,
  gender
• ethnicity. NVIQ lower in SSD
• Examined 2 subtype dimensions of SSD, persistence
  and LI, in a 2x2 design.
• Compared on pre-literacy measures letter
• knowledge, PA, RSN

77
SSD and RD Project Initial Results(Raitano, et
al, in revision)
1.) Entire SSD group worse than controls on PA
and letter knowledge, but not RSN. 2.) Within
SSD group, main effects of persistence and LI
status on PA, with NVIQ covaried. 3.) Even the
normalized SSD, no LI subgroup worse than
controls on PA.
78
SSD and RD Project Future

• Follow-up groups at age 7 to test literacy
  outcomes
• Test for Shared Neuropsychology
• Auditory (Backward masking)
• Phonological memory (NW rep)
• Implicit Phon. Representations (Boada measures)
• PA
• Test for Shared Etiologies Sib Pair Linkage
  Study
• with Shelley Smith