1. Genetics Journal Club
Sumeet A. Khetarpal
10 December 2015

2. A common overall problem…

3. A common overall problem…
Phenotype/Disease
Unbiased genomic discovery
Causal variant?
Mechanism?
Variants
Gene
Pathway
Actionable deliverable

4. Published Genome-Wide Associations through 2013
~12,000 SNP-trait associations
NHGRI GWA Catalog

5. Maris lab at CHOP: Heritability of Neuroblastoma
Bosse & Maris. Cancer. 2015

6. Neuroblastoma is a Pediatric Cancer
Tumor of sympathetic nervous system
most frequently presents in adrenal gland
650 cases/year in USA
10% of all pediatric cancers and 15% of all pediatric cancer deaths
<50% of high-risk cancers contain recurrent somatic mutations
(Pugh et al. 2013)
Germline genetic variation may substantially impact heritability
Germline mutations in the ALK/PHOX2B genes are implicated in rare familial neuroblastoma cases
J Nucl Med, 2004

7. Neuroblastoma susceptibility loci from GWAS
DUSP12
PLoS Genet 2011
HSD17B12
DDX4/IL31RA
Low-Risk
BARD1
Nat Genet 2009
LINC00340
NEJM 2008
LMO1
Nature 2011
High-Risk
HACE1/LIN28B
Nat Genet, 2012
TP53
JNCI, 2014

8. LIM Domain Only 1 (LMO1) Transcriptional co-regulator
Part of a family of related proteins (LMO1-4) with oncogenic activity in T-ALL, breast cancer, neuroblastoma
Matthews et al. Nat Rev Cancer. 2013

9. (Protective Genotype) (Protective Genotype)
LMO1 is a neuroblastoma oncogene implicated in both tumor initiation and maintenance
LMO1 Knockdown
(Risk Genotype)
LMO1 Knockdown
(Protective Genotype)
LMO1 Overexpression
(Protective Genotype)
Wang et al. Nature, 2011.

10. Hypothesis
GWAS associated noncoding SNPs at LMO1 perturb LMO1 gene expression
Goal
Determine the causal variant and mechanism underlying association of LMO1 noncoding SNPs with neuroblastoma risk

11. Approach Fine mapping of the LMO1 locus
Characterization of SNPs using publicly available datasets
Impact of lead SNP on gene expression
Impact of lead SNP on gene transcription

12. Fine Mapping of the LMO1 Locus
Neuroblastoma GWAS ~2800 cases vs controls
Germline SNP Imputation from 1000 Genomes
2100
4200

13. Fine Mapping of the LMO1 Locus 27 SNPs with MAF>0.01 and P<10-5
Fig. 1a

14. Direct genotyping and meta-analysis of rs2168101 association
3200 Cases vs Controls
Common in Europeans + East Asians, rare/absent in Africans
Direct genotyping of 146 / 2101 cases from GWAS  86% imputation accuracy
Table 1

15. Conditional analysis of other LMO1 SNPs after adjusting for rs2168101
Ext. Fig. 1a

16. Conditional analysis of rs2168101 after adjusting for other LMO1 SNPs
rs is the lead SNP of the lone signal at the LMO1 locus
Ext. Fig. 1b

17. ENCODE ChIP Seq overlap of LMO1 SNPs in SKNSH neuroblastoma cell line
Fig. 1b

18. Does the LMO1 lead SNP affect transcription factor binding?
JASPAR Database
Consensus TF binding motifs based on experimental ChIP-Seq data
202 nonredundant profiles in vertebrates

19. Does the LMO1 lead SNP affect transcription factor binding?
Lead LMO1 SNP falls in conserved enhancer and GATA binding motif
Fig. 1b

20. Impact of lead SNP on LMO1 expression
LMO1 RNA reads in 127 high-risk tumors separated by genotype
No T/T genotypes found in high-risk neuroblastoma tumors 
risk alleles ‘predispose’ to the high-risk subset
Fig. 2a

21. Impact of lead SNP on LMO1 expression
Allelic imbalance of LMO1 expression in tumors from hets
Fig. 2b – 12 tumors were heterozygous.
Fig. 2a-b

22. Impact of lead SNP on LMO1 expression
Allelic imbalance of nascent LMO1 expression in heterozygous cell lines
Ancestral GATA allele of lead SNP  increased LMO1 gene expression
Cis regulation?
Fig. 2a-c

23. G(O)TA find the GATA
GATA2 and GATA3 are highly expressed GATA factors in neuroblastoma tumors
Neuroblastoma cell lines overexpressing GATA3 show increased LMO1 expression if G allele present
Ext. Fig. 5a

24. GATA3 binding at LMO1 by ChIP
GATA3 ChIP-Seq in neuroblastoma cell lines with differing rs genotype
Fig. 4a

25. GATA3 binding at LMO1 by ChIP
Allele-specific GATA3 ChIP-Seq at LMO1 lead SNP in cell lines
~Zero GATA3 binding to rs T allele!
Fig. 4b

26. LMO1 lead SNP lies in active enhancer
Fig. 4c, e

27. Summary
Fine mapping validates LMO1 locus-neuroblastoma association and identifies single signal
Lead SNP lies in conserved GATA TF binding motif
Protective allele disrupts GATA3 binding, enhancer activity, gene expression
Protective allele associated with reduced cell growth & increased survival in high-risk disease

28. Lingering Questions Busted enhancer, broken TF binding, both,
or one-in-the-same?
Molecular mechanism of disruption?
Why is this particular GATA binding site so important for this region?
Other long-range interactions disrupted?
(Is it really all just LMO1?)
Is locus really a ‘super-enhancer’?

29. Can this locus be defined as a super-enhancer?
Proposed model for impact of lead SNP on GATA3-mediated LMO1 transcription
Risk “G” allele of rs :
Exon 1
GATA3
AGATAA
Exon 1
Exon 2
Exon 3
Exon 4
LMO1 Gene
Can this locus be defined as a super-enhancer?
Protective “T” allele of rs :
GATA3
Exon 2
Exon 3
Exon 4
Exon 1
ATATAA
LMO1 Gene

30. Enhancers and Super-enhancers
DNA element that activates transcription from a distance
Enrichment of H3K4me1, H3K27ac, DNase I hypersensitivity
Few 100 bp
Super-enhancers
Unusually strong enrichment for coactivators (e.g. Med1)
Enrichment of Pol II, eRNA, p300, CBP, H3K27ac, H3K4me1/2, DNase I hypersensitivity
Few 1000 bp
High controversy over distinction from enhancers
Poor agreement over definition
Pott & Lieb. Nat Genet. 2015

31. ‘Super-enhancers’ are controversial

32. Back to our common problem…

33. Back to our common problem…
Phenotype/Disease
How can we find causal variants and their mechanisms?
Unbiased genomic discovery
Variants
Gene
Pathway
Actionable deliverable

34. There are some tools to help…
Spain & Barrett. Hum Mol Genet. 2015

35. …but it’s still a hot mess
Spain & Barrett. Hum Mol Genet. 2015

36. ...and only a few have ‘succeeded’ so far
Spain & Barrett. Hum Mol Genet. 2015

37. Germline genetic variation and pediatric cancer risk
Whole-genome or whole-exome sequencing in ~1100 pediatric participants with cancer vs. ~960 cancer-free participants from 1000 genomes
8.5% of cancer patients had deleterious germline mutation
Underestimate of total germline heritability

38. Thanks for coming!