| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
Department of Medicine (J.M.H., O.J.B., J.D.C.-S., J.A.F., S.C.G.), Division of Medical Sciences, Institute of Biomedical Research, University of Birmingham, Birmingham B15 2TT, United Kingdom; and Wellcome Trust Centre for Human Genetics (J.C.B.), University of Oxford, Oxford OX3 7BN, United Kingdom
Address all correspondence and requests for reprints to: Dr. Joanne Heward, Department of Medicine, Division of Medical Sciences, Institute of Biomedical Research, University of Birmingham, Edgbaston, Birmingham B15 2TT, United Kingdom. E-mail: j.m.king{at}bham.ac.uk.
 |
Abstract |
|---|
 Top Abstract IntroductionSubjects and MethodsResultsDiscussionReferences |
|---|
Objective: Having previously reported significant association of the T allele of rs2476601 in a Graves’ disease (GD) cohort, we sought to determine whether novel rheumatoid arthritis-associated SNPs were also contributing to susceptibility to GD.
Design: Case control and family-based studies of five PTPN22 tag SNPs were performed.
Setting: An United Kingdom academic department of medicine was the setting for the study.
Patients or Other Participants: A total of 768 GD patients, 768 control subjects, and 313 families with autoimmune thyroid disease participated.
Main Outcome Measure: Tests for association with disease were the main outcome measure.
Results: No association with disease of any of the individual SNPs and no correlation between genotype and clinical phenotype were seen. However, haplotype analysis of the SNP markers with addition of rs2476601 did reveal a strong association of a haplotype containing the T allele, in both the case control (
2 = 29.13; P = 6.77 x 10–8) and family data sets (2 = 5.24; P = 0.02). Furthermore, a novel protective effect of a haplotype containing all six SNPs was observed (2 = 17.02; P = 3.7 x 10–5).
Conclusions: These data suggest that the association of SNPs within the PTPN22 region differs between autoimmune diseases, occurring individually and/or as part of a haplotype, indicating that the mechanisms by which PTPN22 confers susceptibility to GD may, in part, be disease specific.
|
Introduction |
|---|
|
TopAbstractIntroductionSubjects and MethodsResultsDiscussionReferences |
|---|
Recent studies have identified association of the T allele (W620) of a functional missense single nucleotide polymorphism (SNP), +1858 C > T encoding an arginine to tryptophan (R620W) substitution at codon 620 (dbSNP accession number rs2476601) of PTPN22 encoding lymphoid tyrosine phosphatase (LYP), with many different autoimmune diseases, including type 1 diabetes (5, 6, 7), systemic lupus erythematosus (8, 9), Graves’ disease (GD) (10, 11), rheumatoid arthritis (RA) (12, 13, 14), autoimmune hypothyroidism (AIH) (9), juvenile idiopathic arthritis (14), and vitiligo (15). LYP is a protein tyrosine phosphatase (PTP) that plays a negative role in T cell signaling by dephosphorylating the Src family of kinases, thus preventing initiation of T cell activation (16). It is encoded by PTPN22 on chromosome 1p13 (17) and exerts its effect on T cell activation via association with a variety of adaptor molecules, including Csk kinase (18), c-Cbl (17), and Grb2 (19). It is known that the presence of tryptophan at position 620 (W620) can disrupt the ability of LYP to bind to at least one of these adaptor molecules, namely Csk kinase, due to the inability of W620 to fit into the binding pocket on the Csk molecule (5). This has been shown experimentally in both Escherichia coli and COS cells, where only the construct containing R620 was precipitated by Csk (5). The combination of these genetic and functional data suggests that LYP plays an important role in regulating the autoimmune disease process.
A recent report genotyped 37 SNPs in and around PTPN22 in two large cohorts of patients with RA, resulting in the identification of several SNPs that exert both predisposing and protective effects on disease susceptibility (20). Strong linkage disequilibrium (LD) between these markers allowed the identification of 10 common haplotypes, two of which were associated with disease, with the strongest association being observed with the haplotype that contained the T allele of rs2476601, and two that appeared to be protective against disease development (20). Although the haplotype containing the risk allele of rs2476601 was strongly associated with disease, it was noted that this polymorphism alone could not fully explain the association between PTPN22 and RA, with two further SNPs being identified that may confer an independent risk to disease (20). Having previously reported significant association of the T allele of rs2476601 with GD (6), we sought to determine whether the novel SNPs most recently reported to be associated with RA were also contributing to susceptibility to GD. Therefore, we used a tag SNP approach to identify five common RA associated SNPs (dbSNP accession numbers rs2488458, rs12730735, rs1310182, rs1217413, and rs3811021) that would cover most of the common allelic variation within PTPN22, and genotyped them in a GD case control cohort and a family data set of 313 autoimmune thyroid disease families. Replication of the previously associated rs2476601 SNP was also sought in the family data set.
|
Subjects and Methods |
|---|
|
TopAbstractIntroductionSubjects and MethodsResultsDiscussionReferences |
|---|
All patients were unrelated white, United Kingdom Caucasians bled at thyroid clinics in Birmingham, Bournemouth, Exeter, and Walsall, as previously described (2). The presence of biochemical hyperthyroidism together with either the presence of dysthyroid eye disease [(NOSPECS) classification 2 or above] (21), or a diffuse goiter and a significant titer of microsomal, thyroglobulin, or TSH receptor (TSHR) autoantibodies defined GD. The presence of positive thyroid autoantibodies and biochemical evidence of hypothyroidism defined AIH. Gelatin particle agglutination (SERODIA_ATG; Fujirebio, Inc., Tokyo, Japan) was used to measure microsomal and thyroglobulin antibodies, and a titer of 1:100 was considered significant for both assays. A radioactive inhibition method (RSR Ltd., Cardiff, UK) determined TSHR autoantibody status. Serum free T4 concentrations were measured with the Bayer ACS 180 and Bayer Advia Centaur System (Newbury, Berkshire, UK; normal range, 9–20 pmol/liter). Control subjects, matched for age, ethnicity, and gender, with no personal history of autoimmune disease were bled at geographically matched sites, including the Blood Transfusion Services (Birmingham and Oxford, UK), Birmingham Heartlands Hospital (Birmingham, UK), and the Queen Elizabeth Hospital (Birmingham, UK). In total, 768 patients with GD and 768 control subjects were used in the case control study. All control subjects had tests of thyroid function and autoantibody status, and any showing evidence of subclinical autoimmune thyroid disease were removed from the study before genotyping. Simplex families were also recruited, with the majority consisting of both parents, an index case, and at least one unaffected sibling. In total, 313 families were recruited, of which 261 had an index case with GD, and 52 had an index case with AIH. All patients gave informed written consent, and the local ethics committee approved the project.
Genotyping
Genotyping was performed using TaqMan genotyping technology on an ABI 7900 HT (assay details available on request from the authors). SNPs were selected for genotyping because of their previous association with RA and their ability to tag other SNPs in the region with a minimum r2 of 0.8. This strategy enabled us to infer genotypes for 23 SNPs genotyped in the RA association study, with the remainder either being not associated or rare in the original study.
Phenotype analysis
Genotype data for all five SNPs typed in this study and the previously typed rs2476601 SNP were correlated with clinical characteristics (22), including the severity of ophthalmopathy indicated by the NOSPECS rating (0–1 vs. 2 or greater), age at biochemical diagnosis of thyroid dysfunction (<30 vs. greater than 30), presence or absence of goitre on physical examination (defined as palpable or visible thyroid enlargement), biochemical severity of thyroid dysfunction at diagnosis determined from serum concentrations of free T4 (<40 vs. greater than 40), and presence or absence of TSHR, thyroglobulin, and microsomal autoantibodies. This allowed us to establish whether any of the SNPs were contributing to the presence of the different clinical subphenotypes observed in GD.
Statistical and LD analyses
Statistical analysis of the case control data was performed using the 2 test within the Minitab package (MINITAB Release 14.1; Minitab, Inc., State College, PA), and P < 0.05 was considered significant. Power calculations and Hardy Weinberg Equilibrium were performed using Excel (Microsoft Office Excel; Microsoft Corp., Redmond, WA). LD was analyzed using the pairwise LD measure D' and haplotype blocks constructed with the use of the computer program Haploview version 3.2 (http://www.broad.mit.edu/mpg/haploview) (23), using the default algorithm for generating haplotype blocks based on methods established by Gabriel et al. (24). A D' value of one = complete LD, D' value greater than 0.8 = strong LD, D' value 0.2–0.8 = incomplete LD, and D' less than 0.2 = negligible LD. Haplotype analysis of the five SNPs genotyped in the case control study along with the previously associated rs2476601 was performed using the computer program Haploview version 3.2 (23). Transmission of both alleles and haplotypes in the family data set was analyzed using the transmission disequilibrium test (TDT) (25) within the computer program Haploview version 3.2 (23). Further conditional analyses and global haplotype tests were performed using WHAP software (http://pngu.mgh.harvard.edu/
purcell/whap/) developed by Shaun Purcell (Massachusetts General Hospital, Boston, MA) and Pak Sham (Hong Kong University, Hong Kong). This analysis was used to disentangle the correlation structure in the gene to exclude the possibility that multiple observed effects are caused by LD with a single true effect.
|
Results |
|---|
|
TopAbstractIntroductionSubjects and MethodsResultsDiscussionReferences |
|---|
No association of the five SNPs genotyped was seen in the GD case control data set (Table 1
) despite having greater than 99% power to detect an effect of all SNPs genotyped if present at an odds ratio (OR) of 1.5. An OR of 1.5 was selected because our previous study of rs2476601 in GD yielded an OR in excess of 1.4. However, adequate power was still attainable at ORs of 1.4 and 1.3 (98% and 85%, respectively). All cases and controls were in Hardy Weinberg Equilibrium.
|
Pairwise D' values of the five SNPs typed in the case control data set and the previously typed rs2476601 SNP were analyzed. The D' values ranged from 0.8 to 1, indicating strong or complete LD (Fig. 1). Haplotype blocks were constructed using the default algorithm based on methods established by Gabriel et al. (24). This algorithm requires that the lower confidence interval for D' values does not fall below 0.7. All markers within the haplotype block identified adhere to this criteria, with the exception of rs3811021 and rs2476601 (lower confidence interval = 0.62), indicating that these SNPs may only be in moderate LD with each other.
|
Because of the strong LD between these six markers, haplotype analysis was undertaken using the computer program Haploview version 3.2 (23). Six haplotypes were identified (Table 2), five of which correlated with haplotypes 1–5 identified in the report by Carlton et al. (20), with the sixth haplotype being a combination of haplotypes 6–10 identified in the same report. The lack of separation of these haplotypes was caused by rare or unassociated SNPs from the RA study not being taken forward for genotyping in the current study. One haplotype was strongly associated with GD (haplotype 2; 2 = 29.13; P = 6.77 x 10–8), and one was strongly protective (haplotype 3; 2 = 17.02; P = 3.7 x 10–5) (Table 2).
|
Analysis of the family data set revealed similar results. None of the five SNPs common to both data sets in this study showed significant overtransmission of any of the individual alleles (Table 3). However, the association of rs2476601 was replicated in the family data set (2 = 4.149; P = 0.04; Table 3) as was the association of the predisposing haplotype 2 (2 = 5.24; P = 0.02; Table 4).
|
|
Dissection of the haplotypes in the case control data set using WHAP software revealed that rs2476601 was solely responsible for the predisposing effect seen in haplotype 2. However, no particular SNP could be attributed with conveying the protective effect noted in haplotype 3 using this software. In addition to testing the association of individual haplotypes with GD, a global test of haplotype association was performed using WHAP, which showed a strong haplotypic effect with disease (2 with 6 df = 45.89; P = 9.6 x 10–5). The observation that the presence of risk haplotypes within data sets can create spurious protective effects for other haplotypes led us to perform the same global test conditioning, first for the predisposing haplotype 2 alone and then for both the predisposing haplotype 2 and protective haplotype 3. Conditioning solely for haplotype 2 revealed that this haplotype alone could not account for the full haplotypic association seen previously [2 (6 df) = 21.42; P = 0.00026], indicating that the protective effect seen with haplotype 3 is indeed a genuine effect. Conditioning for both haplotypes revealed that no significant haplotypic association remained in the data set after controlling for these two haplotypes [2 (6 df) = 4.22; P = 0.239].
Clinical phenotype correlations
No correlation was obtained between genotype at any SNP and clinical phenotype, including the severity of ophthalmopathy indicated by the NOSPECS rating (0–1 vs. 2 or greater), age at biochemical diagnosis of thyroid dysfunction (<30 vs. greater than 30), presence or absence of goitre on physical examination (defined as palpable or visible thyroid enlargement), biochemical severity of thyroid dysfunction at diagnosis determined from serum concentrations of free T4 (<40 vs. greater than 40), and presence or absence of TSHR, thyroglobulin, and microsomal autoantibodies (data not shown).
|
Discussion |
|---|
|
TopAbstractIntroductionSubjects and MethodsResultsDiscussionReferences |
|---|
LD analysis revealed all SNPs to be in strong to complete LD with one another and with the previously associated rs2476601 SNP, allowing the construction of haplotype blocks. All SNPs were found to form one block of LD, and subsequent haplotype analysis revealed the presence of one predisposing and one protective haplotype. Analysis of the family data set confirmed the association of the rs2476601 SNP and the predisposing haplotype. However, no excess transmission of the protective haplotype was observed, which is probably attributable to its rarity in this data set (Table 4
). Conditional analyses performed with the program WHAP indicate that the effect of the predisposing haplotype is caused solely by the presence of the rs2476601 variant on that haplotype. However, the protective haplotype is a novel finding not obviously caused by any genotyped SNP and, therefore, is likely to be in strong LD with an as yet unidentified variant responsible for the effect.
The haplotypes identified in this study correlated well with the 10 common haplotypes identified in the RA study. The same predisposing haplotype (haplotype 2) was identified in both studies and was the only haplotype to carry the T allele of the rs2476601 SNP. It differed from haplotype 1 only at this position, suggesting that this SNP was responsible for the predisposing effect of this haplotype, a result that the WHAP analysis confirmed. Haplotype 4 was also increased in the RA data set, but no effect of this haplotype was seen with GD in either the case control or family data sets. When examining the protective haplotypes between data sets, it must be noted that the haplotypes that appear to protect from disease differ between RA and GD. Haplotype 3 was strongly protective in the GD data set, whereas haplotypes 5 and 6 exerted a similar effect in RA. Although we can state that haplotype 5 has no protective effect in GD, it is more difficult to assess the role of haplotype 6 in this disease because the typing undertaken in this study did not allow for separation between haplotypes 6–10.
The lack of association of individual SNPs with GD could be attributed to a number of reasons. First, the highly significant associations seen in the original study could be a result of the first-time effect phenomenon, whereby the size of the genetic effect in the first positive report is biased upward, making replication in subsequent studies a more difficult task (27).
Second, the lack of replication and differing haplotype associations may be attributed to the different geographical origins of the data sets used. Although both were white Caucasian, those in the RA study were from North America, whereas our data set consisted of patients from the United Kingdom. It is well documented that LD patterns vary widely between different populations (28), and this in turn impacts on the transferability of tag SNPs between populations (29, 30). When comparing the minor allele frequencies obtained in the control populations between the two studies, it can be seen that there is a difference in both SNP and haplotype frequencies, ranging from 4–10%, which could be attributed to differing LD and tag SNP patterns between the populations, thus affecting the associations seen. However, a comparison of D' values showed a similar pattern to that seen in the RA study, indicating that LD patterns are not appreciably different between these two populations. It must also be noted that haplotype frequencies and associations differed between the two sample sets used in the RA study, which was attributed to a significant difference between the controls used in each study. This could also be contributing to the differing haplotype associations seen in our study when compared with the RA study. A comparison of both the RA and GD data sets with HapMap CEPH data for this gene shows that haplotype frequencies are noticeably different between the diseased populations studied and the CEPH data. However, the differences observed are not significantly different for the common haplotypes observed in each study, and may be attributed to geographical and/or data set size issues.
However, it seems more likely that although PTPN22 appears to be acting as a general autoimmunity locus, the lack of similarity between association of individual SNPs and haplotypes with RA and GD suggests the mechanisms by which PTPN22 confers susceptibility may, in part, be disease specific. A recent report in type 1 diabetes has identified infrequent variants at the PTPN22 locus, which appear to contribute to susceptibility to type 1 diabetes (31), thus providing support for our hypothesis. The SNPs typed in the type 1 diabetes study were different from those used in the RA and GD studies, and it would, therefore, be interesting to determine whether they are indeed disease specific or provide a more general effect on the autoimmune disease process. Although it is well documented that rs2476601 affects the ability of LYP to bind to Csk kinase (5), no work has been performed on how this, and other polymorphisms, may affect binding of LYP to its other adaptor molecules c-Cbl (17) and Grb2 (19). c-Cbl is a protooncogene that is constitutively expressed with LYP, and is known to be phosphorylated after stimulation of T cells and can negatively regulate Syk and ZAP-70, both of which are necessary to amplify signals from the T cell receptor (17). Phosphorylation of c-Cbl is reduced when LYP is overexpressed, suggesting that LYP might regulate the activity of c-Cbl by controlling its phosphorylation status (17). It has been shown that Grb2 binds to LYP via its SH3 domain, leading to formation of a functional complex resulting in negative regulation of T cells (19). It could be postulated that polymorphisms associated with RA may affect these pathways in a disease-specific manner, explaining the differing associations seen with GD.
Examining expression of the different isoforms of LYP (LYP1 and 2) in tissues from patients with both diseases may provide an alternative explanation for differential association of PTPN22 SNPs and haplotypes with RA and GD. Both isoforms share nucleotides 1 to 2097, but an alternative RNA splicing event of an LYP intron leads to different C-terminal sequences (17). Although it is unclear what impact these different sequences will have on LYP function, it can be postulated that the effect of any SNPs present within these regions will be affected by the relative expression of each isoform of LYP. It is documented that LYP2 only contains one SH3 binding domain site compared with four in LYP1 (17), which could reduce binding of the LYP molecule to its adaptor molecules, thus affecting regulation of T cell activation in different disease states. It must also be noted that the isoforms are differentially expressed in different cell types (17), with LYP1 being the predominant isoform in lymphoid tissue, and LYP2 being seen at higher levels in fetal liver tissue and resting T cells, indicating differences in regulation of expression, which could be altered between autoimmune diseases. To examine these hypotheses and determine whether the presence of certain SNPs is affecting the mechanisms by which the LYP molecule functions, it is first essential to obtain detailed information regarding the LD block structure within PTPN22. Identification of peaks of LD within this gene and, eventually the primary etiological variant(s), through fine mapping studies and logistic regression analysis will allow functional studies to focus on how this variant(s) may affect binding of LYP to its adaptor molecules and how it may affect the different isoforms of LYP.
In summary, our data suggest that genetic polymorphisms within PTPN22 are influencing susceptibility in a disease specific manner, a finding that warrants further functional analysis to elucidate their true role in GD and the autoimmune disease process in general.
|
Acknowledgments |
|---|
|
Footnotes |
|---|
Disclosure Statement: The authors have nothing to disclose.
First Published Online December 5, 2006
Abbreviations: AIH, Autoimmune hypothyroidism; GD, Graves’ disease; LD, linkage disequilibrium; LYP, lymphoid tyrosine phosphatase; OR, odds ratio; PTP, protein tyrosine phosphatase; RA, rheumatoid arthritis; SNP, single nucleotide polymorphism; TDT, transmission disequilibrium test; TSHR, TSH receptor.
Received September 20, 2006.
Accepted November 29, 2006.
|
References |
|---|
|
TopAbstractIntroductionSubjects and MethodsResultsDiscussionReferences |
|---|
This article has been cited by other articles:
 |
|
 |
|
  D. J. Smyth, J. D. Cooper, J. M.M. Howson, N. M. Walker, V. Plagnol, H. Stevens, D. G. Clayton, and J. A. Todd PTPN22 Trp620 Explains the Association of Chromosome 1p13 With Type 1 Diabetes and Shows a Statistical Interaction With HLA Class II Genotypes Diabetes, June 1, 2008; 57(6): 1730 - 1737. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
M. Zoledziewska, C. Perra, V. Orru, L. Moi, P. Frongia, M. Congia, N. Bottini, and F. Cucca Further Evidence of a Primary, Causal Association of the PTPN22 620W Variant With Type 1 Diabetes Diabetes, January 1, 2008; 57(1): 229 - 234. [Abstract] [Full Text] [PDF]
|
|
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
| Endocrinology | Endocrine Reviews | J. Clin. End. & Metab. |
| Molecular Endocrinology | Recent Prog. Horm. Res. | All Endocrine Journals |
