This document summarizes genetic predisposition to papillary thyroid cancer. It discusses heritability estimates for various cancers including thyroid cancer. Attempts to identify predisposing genes through linkage analysis and next generation sequencing have had limited success due to genetic heterogeneity and overdiagnosis of thyroid cancer. Genome-wide association studies have identified several loci associated with small increased risks. Whole exome sequencing of families identified a potentially pathogenic variant in the SRRM2 gene segregating with disease in one family. Overall, most heritability is likely due to many common low-penetrance variants, though rare high-penetrance mutations also exist. Gene discoveries have had modest clinical impact to date.
3. Heritability of selected cancers
Thyroid 8.48 12.42
Testis 8.57 8.5
Multiple myeloma 4.29 5.62
Prostate 2.21 9.41 0.42
Colorectum 2.54 4.41 0.35
Breast 1.83 2.01 0.27
Lung 2.55 3.16 0.26
All cancers 2.15 3.53
Site Family risk ratio
Utah1 Sweden2
Twin study3
(proportion of
variance)
1. Goldgar et al. 1994 2. Dong and Hemminki 2001 3. Lichtenstein et al. 2000
Adapted from Risch 2001
4. Odds ratio (OR)
Odds ratio = likelihood of acquiring the
phenotype (e.g. PTC) if marker
present relative to when it is absent
OR >1 is “predisposing”
OR <1 is “protective”
In the last few years the term odds ratio
has begun to be called “effect size”
OR and effect size are equivalent to
“penetrance”
5. Searching for predisposing
genes
• Loss of heterozygosity
• Association studies
• Next generation sequencing
• Linkage analysis in families
6. Linkage in PTC
Literature
• Between 1997 and 2006 at least 5
loci were proposed: 1p21, 2q21,
8p23, 14q32, 19p13
• Despite vigorous efforts no genes
have been found
• These data suggest genetic
heterogeneity, multigenic and
multifactorial inheritance, probably
low penetrance.
7. Main reason for failure of linkage
analysis in PTC is
Overdiagnosis (?)
Unfortunately, the Next Generation
Sequencing suffers from the same problem.
8. • Definition of overdiagnosis
“Diagnosis of those that would not, if left alone, result in
symptoms or death”
• “Overdiagnosis accounting for thyroid cancer in women
South Korea 90%
USA, Italy, France 70-80%
Japan, Nordic countries, UK 50%”
• “There is no evidence of new risk factors or increased
exposure”
NEJM 2016
13. Summary of candidate
genes found by linkage
OSU
Chromosome Gene description No. families Ref
8q24 A lncRNA inside the TG gene 7 1
12q14 SRGAP1 4 2
4q32 An enhancer of unknown function 1 3
1. He et al. Cancer Res 2009
2. He et al. JCEM 2013
3. He et al. PLoS One 2013
14. He et al. JCEM 2013
Whole genome linkage analysis of 38 PTC families
15. Plot of the genome-wide linkage scan with posterior PPL from 38 families.
He et al. JCEM 2013 (V. Vieland)
16. SRGAP1, Slit-Robo Rho GTPase
activating protein 1 gene
• Located in 12q14; found by linkage
• Different missense mutations segregate with
PTC in 3 families
• Missense mutations occur in sporadic cases
and controls (OR 1.21, p=0.0008)
• Missense variants impair the inactivation of
CDC42, a key function
• SRGAP1 is a candidate gene for PTC
susceptibility and has low-medium penetrance
17. A>C enhancer mutation in gene desert found by targeted deep sequencing
He et al. PLoS One 2013
18. Long-range enhancer
mutation 4q32 A>C
• Enhancer element is highly conserved
• ChIP assay confirms enrichment of
enhancer marks, e.g. H3K4me1
• Enhancer binds TFs POU2F and YY1
• Risk allele (C) greatly impairs TF binding
• Enhancer RNA is greatly downregulated
in thyroid tumors
19. Enhancer A>C mutation in 4q32
is ultra rare
• Found in 11 affected individuals of one
large non-medullary thyroid carcinoma
family
• Not found in 38 other families
• Not found in 2676 sporadic cases
• Not found in 2470 controls
• Target genes not yet found
This suggests an ultra-rare, high-
penetrance mutation
20. Enhancer mutation in 4q32
Counseling
• Initial pedigree had 11 affected individuals
• Extensive counseling has resulted in much
larger pedigree
• Testing for mutation: positive 34/68
negative 34/68
• Sex ratio in mut. positive individuals:
males n=17
females n=17
21. Hypothesis
Most of the heritability in PTC is due to
(common?) low-penetrance genes
The way of approaching these is….
23. Genome-wide association studies, GWAS
• Principle: search for marker that is more common in cases
than in controls
• Discovery test: type many (e.g. 1 million) SNPs in e.g. 1000
cases, 1000 controls
• Validation test: type top SNPs (e.g. 50) in e.g. a further 2000
cases, 2000 controls
• Replication test: type top SNPs (e.g. 5) in further cases and
controls from different populations
• Because of multiple testing, apply rigorous significance
standards (e.g. p 10-8)
24. GWAS-generated loci for PTC
Summary, first two GWAS
• 9q22 (OR 1.8) intergenic, apparently related to one or two
lincRNA genes (data shown). FOXE1 nearby
• 14q13 (OR 1.37) intergenic, risk allele affects thyroid
specific lincRNA (data shown). NKX2-1 nearby
• 2q35 (OR 1.34) in DIRC3 gene
• 8p12 (OR 1.36) in first intron of neuroregulin (NRG1) gene.
Risk allele lowers gene expression
• 14q13 (OR 2.09) intergenic, close to but independent of
first 14q13 locus
25. Putative lincRNA transcripts in 9q22
PTCSC = papillary thyroid cancer susceptibility candidate
unspliced transcript of PTCSC2 (>60 Kb) spans the genomic region
containing SNP rs965513.
rs965513
Shared haplotype
rs1877431
rs10983700
rs1561960
rs7871887
rs1867277
FOXE1
PTCSC2-
unspliced
PTCSC2-
spliced
26. Three enhancers and 4 functional
variants in a ~33 kb block in 9q22
He et al. PNAS 2015
29. 0.20.40.60.8
2^-(DeltaCt)
Normal, NKX2.1 Vs rs944289
CC
n=11
CT
n=33
TT
n=28
0.20.40.60.8
2^-(DeltaCt)
Tumor, NKX2.1 Vs rs944289
CC
n=11
CT
n=31
TT
n=29
rs944289[T]
Adjacent unaffected Tumor
CC
n=11
CT
n=31
TT
n=29
CC
n=11
CT
n=31
TT
n=29
2^-deltaCt
2^-deltaCt
The risk allele [T] of the SNP increases the
expression of NKX2-1 (TTF1) in thyroid tissue
Kruskal test p-value =0.0899
Pairwise Wilcoxon test,
TT vs CC, p value= 0.14
TT vs CT, p value= 0.046
Kruskal test p-value =0.0200
Pairwise Wilcoxon test,
TT vs CC, p value= 0.074
TT vs CT, p value= 0.0074
30. PTC: Examples of clinical association
at the 14q13 locus
Data from genotyping 1216 cases and 1416 controls
• rs965513 associates with larger tumor size (p=0.025) and
extrathyroidal expansion (OR=1.29, p=0.045)
• Rs2439302 associates with lymph node metastasis (OR 1.24,
p=0.016) and multifocality of the tumor (OR 1.24, p=0.012)
Much more to come…
Jendrzejewski et al. Thyroid 2016
31. Predictive power of GWAS loci
Towards the development of a risk
panel
• 5 loci described so far
• ORs range from ~1.4 to ~2.1
• Are these ratios additive?
2 large cohorts of cases and controls
genotyped for the 5 loci
Ohio 747 cases 1047controls;
Warsaw 1795 cases 2090 controls
Additive risks sought
Liyanarachchi et al. Thyroid 2013
32. Cumulative odds ratios relative to
number of risk alleles for 5 GWAS SNPs
Liyanarachchi et al. Thyroid 2013
33. Third GWAS deCODE + OSU
Paper submitted, Gudmundsson et al. 2016
•Previous 5 loci confirmed
•5 new loci detected
•Involvement of coding genes observed
34. Next generation sequencing (NGS)
* Whole exome sequencing WES
* Whole genome sequencing WGS
• In principle predisposing genes can be found in
individual patients by whole genome sequencing and
perfect bioinformatics
• In practice this does not work
• Power of resolution can be improved by studying
families searching for variants shared by affecteds
• When families are reasonably large linkage can help
focus search for relevant variants
• Discrimination power can be enhanced by
haplotyping
• NB overdiagnosis of PTC
35. WES of PTC
• Our first NGS experiment
• Study 7 PTC families with > 4
affected
• Do WES on 2 affected/family
• Present results:
Positive finding in 2
37. Filtering principles and results
Conditions and filtering No. of variants
• Variants detected ~1 million/individual
• Quality filtering ~200,000/individual
• Elimination of common variants (>0.01) ~10,000/individual
• Variant shared by the 2 affecteds ~2000/family
• Not found in other than one family ~400/family
• Deleterious by nature of variant & conservation ~100/family
• Expressed in thyroid ~40/family
Akagi, Symer et al.
38. How to filter the remaining
candidates (n=40)
• Validate mutation by Sanger
• Literature (cancer involvement?)
• Databases (same mutation seen?)
• Cosegregation in the family
• Genotyping results from deCODE
• Linkage (peak or valley)
• Haplotype sharing
• Population occurrence
39. Segregation of SRRM2 c.1037 (S346F) variant in Family 7
Haplotype sharing
Typing of PTC cases and controls:
7/1170 sporadic cases
0/1404 controls
OR = 8.14; p-value = 0.049
0/138 familial cases
40. Haplotype sharing:
a good filter to eliminate candidate variants
(Final 21 candidates from Family 7)
43. Heat map of 1642 alternative splicing
events
RNA-Seq data: alternatively spliced transcripts in the cases
were differentially expressed when compared to controls
PSI: the ratio of the “included”
expression level vs. the sum of both
spliced isoforms.
Yellow: higher PSI.
Blue: lower PSI.
44. Main problems
• Only coding DNA typed
• Unexpectedly common mutation (>1%; >3%
etc.) would be filtered out
• Only mutations classified as “pathogenic”
are considered
47. Whole genome sequencing in PTC families
Summary of results in 8 families
• Filtering excluded everything except coding variants
of genes expressed in thyroid tissue and with
predicted pathogenicity
• The median number of remaining candidate genes
per families was 27 (range 14-83)
• Efforts to identify the correct gene(s) are underway
48. So What?
Consequences of gene discoveries
• Improved molecular insight; pathways?
Yes but slow
• Diagnosis?
Yes but mainly in families
• Clinical stratification?
Promising but so far modest impact
Editor's Notes
Now to our story. Our study was designed based on the availability of PTC affected families.