1. Detailed genetic characterization of the psoriasis-associated gene IL23R
S. J. Schrodi1, M. Chang1, V. E. Garcia1, N. Matsunami2, M. F. Leppert1, G. G. Krueger2, A. B. Begovich1
1. Celera, Alameda, CA, USA University of Utah, Salt Lake City, UT, USA
INTRODUCTION
SINGLE SNP ANALYSES
HAPLOTYPE ANALYSES
Using a multi-tiered, case-control association design, scanning over 25,000 gene-centric SNPs, we recently identified two psoriasis susceptibility genes: IL12B and IL23R (Cargill et al, 2007). One of the IL23R missense SNPs implicated in psoriasis has also been strongly associated with predisposition to inflammatory bowel disease (Duerr et al, 2006; Van Limbergen et al, 2007; Rioux et al, 2007; Libioulle et al, 2007; Dubinsky et al, 2007). To refine our initial association results and further investigate the region involved in psoriasis risk, we extended our genetic analysis by resequencing this gene in 96 psoriatic individuals. Twelve novel IL23R SNPs were discovered including one missense, two synonymous and three 3’ UTR SNPs. Using a series of selection criteria, we identified 58 additional IL23R–linked SNPs which were genotyped in our three independent, white North American sample sets (>2800 individuals in toto). Single marker and preliminary haplotype analyses have been performed on the full set of interrogated SNPs. A sliding window of haplotype association demonstrates co-localization of psoriasis susceptibility within the boundaries of IL23R across all sample sets, thereby decreasing the likelihood that neighboring genes, particularly IL12RB2 which lies directly adjacent to IL23R, are driving the association of this region with psoriasis. The refinement of disease association patterns within IL23R to specific predisposing and protective genetic variants will play an important role in the elucidation of the causes of psoriasis and possibly additional autoimmune phenotypes as well as the role of IL23R genetic variants in response to therapy and dosage.
Figure 2 shows a plot of the allelic association P-values as a function of position for the 61 SNPs selected to interrogate the IL23R region. Results for the three sample sets are plotted separately. The peak of psoriasis association occurs over the IL23R coding region.
Figure 6
Using a sliding window of haplotype association on 3 adjacent SNPs, we identified two neighboring windows composed of rs , rs , rs and rs yielding the peak global association with psoriasis when analyses are combined across the three independent sample sets (window 1 combined P-value = 1.28E-04 and window 2 combined P-value = 6.42E-05). Both windows were significant (P<0.05) in each of the three studies. These four SNPs reside within a 12kbp region within IL23R.
Figure 2. IL23R Single SNP Analysis: 61 SNPs Genotyped
IL12RB2
C1ORF141
SERBP1
IL23R
Global P-Value
Figure 6. IL23R region sliding window Haplotype association
3-SNP Windows
EXPERIMENTAL DESIGN
25,215 SNPs
Discovery UPI Sample Set
40 Orthogonal Pools (465 cases / 460 controls)
Replication GCI Sample Set
42 Orthogonal Pools (494 cases / 495 controls)
P<0.05
Comb P
< 0.05
Confirmation: Individual Genotype
Replication #2 GCI/BCW sample set
(481 cases / 424 controls)
Fine-scale mapping
Haplotypes/Diplotypes
C1orf141
IL23R
IL12RB2
SERBP1
GENOTYPE ANALYSES
Six individual SNPs with combined P-values below 0.05 are presented in detail showing genotype frequencies in cases and controls across the three sample sets. Notably, the Q381R SNP (rs ) yielded the most significant combined association results.
Table 2
SNPs significantly associated using a Fisher’s combined P-value across all three studies (6 SNPs from Table 1) were examined and partitioned into three groups on the basis of LD structure and significance. Group1: rs , rs , rs ; Group2: rs and rs ; Group3: rs One representative SNP from each group was used in a haplotype analysis where phase was estimated through a pseudo-Gibbs sampling algorithm (SNPAnalyzer program). Haplotypes segregating at rs rs rs were then evaluated for statistical association. One psoriasis-predisposing haplotype (C-G-G), one neutral haplotype (C-G-T) and two protective haplotypes (T-G-T and C-A-T) were found.
Table 1. Genotype frequencies for significantly associated SNPs across sample sets
SNP SELECTION
We undertook a multifaceted approach to identify SNPs to genotype individually in a fine-scale mapping effort in the IL23R region. A 336 kbp region was selected across a portion of C1orf141 through SERBP1. SNPs in moderate to high LD (r2>0.20) with the two originally-identified missense SNPs from Cargill et al (2007), rs and rs , were initially selected and reduced so that those pairs within this group in extremely high LD (r2>0.97) were represented by one or two SNPs. Next, we ran the tagging SNP program Redigo (Hu et al, 2004) to select a set of tagging SNPs among those in weak LD (r2<0.20) with the original two SNPs. Further, any SNP with putative function annotation were selected to be genotyped. In all, 61 SNPs were identified to cover the IL23R region.
0.723
0.262
0.015
0.889
0.107
0.004
0.728
0.259
0.013
0.885
0.111
0.370
0.475
0.155
0.736
0.248
IL23R
dbSNP ID
Gene
Type
Position (kbp)
Cases
(n=465)
Controls
(n=460)
Frequency in
(n=494)
(n=495)
Genotype
(n=481)
(n=424)
Discovery
Sample Set
Replication
Sample Set 1
Sample Set 2 rs rs rs rs rs rs
L310P
Intron
Q381R
3' of IL23R
97.357
0.801
0.190
0.009
0.924
0.074
0.002
0.805
0.186
0.918
0.078
0.316
0.492
0.192
0.795
0.196
0.810
0.180
0.010
0.887
0.109
0.808
0.012
0.888
0.105
0.006
0.340
0.454
0.206
0.800
0.788
0.016
0.846
0.144
0.198
0.014
0.832
0.158
0.362
0.467
0.171
0.762
0.228 CC CT TT GT GG AG AA AT
0.793
0.199
0.008
0.929
0.068
0.794
0.195
0.932
0.066
0.318
0.476
0.183
0.214
0.024
0.871
0.124
0.005
0.760
0.026
0.873
0.122
0.359
0.472
0.169
0.768
0.204
0.028
Combined
Analysis
Pcomb
Table 2. Effect sizes for haplotypes at rs rs rs
Haplotype
Cases
(n=465)
Controls
(n=460)
Counts in
(n=494)
(n=495)
Discovery
Sample Set
Replication
Sample Set 1 OR
(n=424)
Sample Set 2
(n=481)
CGG
CGT
TGT
CAT
Others
404
385 95 40
360
376
133 55 2
1.22
1.04
0.68
0.72
N/A
425
407 93 56 5
403
390
113 88
1.11
1.09
0.81
0.62
424
398
104 34
344
338
110
1.16
1.07
0.82
0.52
Figure 1. Physical map of 338kb around IL23R on human chromosome 1
Chr1 (p31.3)
B36 Position
Genes
LD Regions
Genotyped Markers rs rs
IL23R
C1orf141
IL12RB2
SERBP1
CONCLUSIONS
Fine-scale mapping of the IL23R region across three independent psoriasis case-control sample sets shows variants segregating at IL23R that are significantly associated with disease. This observation, in conjunction with many immunology and clinical results in the literature (e.g. Park et al, 2005; Ivanov et al, 2006; Krueger et al, 2007), substantially aids our understanding of the role of IL-23 pathobiology underlying chronic inflammatory conditions.
As with several other inflammation disease susceptibility genes exhibiting pleiotropy, variants at IL23R now appear to confer risk/protection for psoriasis and IBD.
A novel SNP-selection procedure to fine-scale map regions with initial association data is discussed.
Single marker analysis produces little evidence that the IL23R-neighboring gene IL12RB2 (also a functional candidate) is generating the patterns of psoriasis-association in the region.
Three-SNP haplotype sliding window analysis yields a peak association of 6.42E-05 located at IL23R with the adjacent SNPs rs rs rs
Additional haplotype analyses with representative SNPs derived from LD groups, demonstrate that haplotypes segregating at rs rs rs (L310P – Q381R – putative transcription factor binding site) generate susceptible, neutral and protective effects with regard to psoriasis.
LINKAGE DISEQUILIBRIUM
Figure 3. Pairwise LD between the 61 IL23R-region SNPs as measured by r2 in the cases & controls of all three sample sets combined
Figure 4. LD with rs : 61 follow-up IL23R SNPs genotyped
Figure 5. Psoriasis association decay with rs LD
Pairwise linkage disequilibrium was calculated across cases and controls combined as an exploratory analysis to better understand the detailed structure in the IL23R region (Abecasis et al, 2000). The correlation statistic r2 was employed. Results are displayed in the Figure 3 heat-map. As much of the association signal was being driven by rs , we specifically sought to elucidate the decay of rs LD in a positional fashion as shown in Figure 4. Only a single SNP attained a r2 value greater than 0.20 in our sample sets: rs , located in an IL23R intron. A genotypic-based squared correlation coefficient was calculated between the two sites: Discovery cases g2 = 0.902, Discovery controls g2 = 0.888, Repl 1 cases g2 = 0.705, Repl 1 controls g2 = 0.870, Repl 2 cases g2 = 0.914, and Repl 2 controls g2 = Two-SNP analyses of rs and rs appear to indicate that recombinants between the two SNPs are reasonably rare in our samples. Figure 5 shows the decay of statistical association with psoriasis as a function of LD with Q381R (rs ). 1 r2
Cases and Controls 2 61 60 rs rs
C1orf141
IL23R
IL12RB2
SERBP1
REFERENCES
Cargill et al (2007) AJHG 80:
Duerr et al (2006) Science 314:
Van Limbergen (2007) Gut Epub Mar 2.
Rioux et al (2007) Nat Genet 39:
Libioulle et al (2007) PLoS Genet 3:e58 Epub Apr 20.
Dubinsky et al (2007) Inflamm Bowel Dis 13:
Hu et al (2004) Hum Hered 57:
Abecasis et al (2000) Bioinformatics 16:
Yoo et al (2005) Nucleic Acids Res 33:W483-W488.
Park et al (2005) Nat Immunol 6:
Ivanov et al (2006) Cell 126:
Krueger et al (2007) N Engl J Med 356: