Erratum to: CD28/CTLA-4/ICOS haplotypes confers susceptibility to Graves’ disease and modulates clinical phenotype of disease

In the original publication, the copyright line in the article PDF and print is incorrectly published with Springer copyright statement instead of “The Authors”. In addition, the license statement is missing in online and print version. The license statement should read as: Open Access This article is distributed under the terms of the Creative Commons Attribution 4.0 International License (http:// creativecommons.org/licenses/by/4.0/), which permits use, duplication, adaptation, distribution, and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license, and indicate if changes were made.


Introduction
The pathogenesis of Graves' disease (GD) requires a specific sequence of events, involving both genetic factors and the ensuing environmental factors, for the final clinical phenotype to manifest [1][2][3]. The orbital inflammatory process in Graves' orbitopathy (GO) is believed to be driven by T cells in response to thyroid-orbit shared antigens accessing and infiltrating the orbital space [4]. As T cells play a key role in the pathogenesis of GO, and CD28/CTLA-4/ICOS co-stimulatory molecules are critical in T-cell activation, they may be considered to be candidate risk factors.
Disruption to the precise balance of expression of costimulatory and inhibitory molecules is known to play a role in the pathogenesis of autoimmune diseases [5]. This is supported by studies demonstrating that adolescents and children newly diagnosed with GD had an abnormal percentage of lymphocytes expressing co-stimulatory molecules [6,7]. Similarly, differences in the expression pattern of CD28 in T cells of the thyroid gland and peripheral blood T lymphocytes was observed [6]. Differential regulation of these molecules could easily affect T-cell function and hence the regulation of the immune response. Abnormal expression and/or dysfunction of these molecules may be caused by polymorphisms in the corresponding gene [8][9][10][11][12][13][14][15][16][17][18][19][20][21][22][23][24][25][26]. In previous work, we linked the presence of guanine (G) in CT60 and Jo31 single-nucleotide polymorphism (SNPs), located within the 3′UTR of CTLA-4, with an elevated level of the soluble CTLA-4 isoform in GD, especially in cases of severe outcome [26]. Competitive binding of sCTLA-4 with CTLA-4 common ligands inhibited T-cell proliferation via increased activation of CD28 [27]. The prevalence of sCTLA-4-bearing genetics in GD may indicate abnormal expression of CTLA-4 on T-regulatory cells (Treg). In turn, this may affect the precise balance of suppressive and effector activity of Treg, which is strongly linked to pathogenesis of autoimmune diseases [27].

Patients
One hundred and seventy-two unrelated Polish Caucasian GD patients consecutively admitted to the tertiary care academic medical center were enrolled into the study. The initial diagnosis of GD was confirmed by routine clinical and laboratory tests. Detailed patient characteristics are presented in Table 1. Ninety-six patients responded to firstline medical treatment and required no further therapy for relapse within next 18 months.

Assessment of ocular changes
Ophthalmological examination was performed at the time of blood collection. The severity and activity of orbitopathy was assessed using criteria proposed by Bartalena [39] (detailed in Table 1).

Control group
The control group consisted of 389 apparently healthy volunteers from the same geographical area (225 (57.8 %) females and 164 (42.2 %) males).

Ethical approvals
The local ethics committee approved the study protocol.

Statistical analyses
The Hardy-Weinberg equilibrium (HWE) was tested for the studied group. Categorical data between GD patients and controls, and within the GD group according to clinical outcome, were compared using the χ 2 test with appropriate correction. In case of multiple comparisons, Bonferroni multiple adjustments were employed. Haplotype analysis and LD coefficients were estimated using the SHEsis software (http://202.120.7.14/analysis/myAnalysis.php).
The combined effects of polymorphic markers and environmental factors (gender and smoking habits) on the development of GD and GD outcome were analyzed by unconditional multivariate logistic regression analysis with a logit regression model using a quasi-Newton estimation method (software EPIINFO Ver. 7.1.1.14).
Differences were considered statistically significant if the p-value < 0.05.

Results
Genotype frequencies in both groups were in HWE ( Table 2).
Smokers were statistically more prone to severe eye symptoms (p = 0.04), and patients with non-severe GO were the best responders to anti-thyroid treatment (p = 0.001) (Supplementary Table 1).

CD28/CTLA-4/ICOS gene polymorphisms and GD risk
In univariate analysis a significant association with GD was observed only with two of the studied polymorphisms: g.*642AT(8_33) and CT60 (rs3087243) ( Table 2).
Other studied polymorphisms were not associated with GD risk (Supplementary Table 2).
In the controls robust LD was observed between the pair of CTLA-4 polymorphisms, and weaker association was seen between CTLA-4 polymorphisms and CD28c.17 +3T>C (rs3116496), while in patients such relations were observed only between CTLA-4 gene polymorphisms (Supplementary Table 3).
Although the association of CTLA-4g.319C>T (rs5742909) with GO development was observed only at the allele level (p = 0.04) ( Table 5), this SNP was associated with GO outcome (Supplementary Table 11). The rare [T]      Albeit the CTLA-4c.49A>G (rs231775) and Jo31 (rs11571302) SNPs were not related to GO development, they were associated with the GO outcome (Supplementary Table 12 Table 12).

The course of GO-haplotype associations
A statistically significant difference in haplotype distribution between GD patients regarding GO occurrence and healthy controls was observed (Supplemental Table 13). When severity and activity status were taken into account, different global haplotypes distribution was seen in case of absence of GO symptoms compared to severity and activity status, but they did not reach statistical significance (in both p corrected = 0.06) (Supplemental Table 14). Haplotype TCG [AT [16][17][18][19][20][21] ]GG(m) favors not only direct GO development, but also the severe and active GO course (p = 0.015201; p = 0.015448; p = 0.015266) (Supplementary Tables 13  and 14).

The course of GO-multivariate associations
The unconditional multivariate logistic regression analysis of all studied polymorphisms, including four prone genetic

Discussion
Since a number of studies have established a central role for T cells in GD [2,[5][6][7], the region containing genes encoding the immune regulatory molecules, CTLA-4, CD28, and ICOS, is good candidate for disease risk factor. Selected for our study polymorphic sites located within genes encoded, these co-stimulatory molecules had to have been reported to be associated with altered immune response and/or susceptibility to autoimmune disease [9,13,[18][19][20][21][22][24][25][26]. Based on the fact that all three genes are close to one another within the chromosomal region 2q33, and that LD was observed between polymorphisms located within these genes, we postulated that the association between GD (and/or its clinical phenotype) and this region might represent the effects of a combination of polymorphic variations of these three genes. We therefore looked for a possible association of studied polymorphisms as one susceptible genetic region defined as a haplotype block with the GD phenotype.
Our study indicated specific haplotypes significantly linked with sporadic GD and haplotypes over-represented in patients with a familial history of thyroid disease. These results led us to the conclusion that patients with sporadic incidence and those with a familial background of GD are characterized by a combination of various markers, and it is therefore possible that they are genetically different. These markers may be predictive factors for the development of GD in families in whom thyroid disease is observed. Although the 2q33 chromosomal region is linked to GD, the association between the two was found to be different in female and male subjects. Our group of patients was too small for stratification by gender at the haplotype level in the context of GO outcome, therefore the possibility that differences in our overall population at the univariate level will also be true in GO subgroups cannot be excluded. To date, no literature concerning such a relationship has been published. In terms of predicting the clinical outcome, two specific haplotypes are possible: one reducing (TTA[AT [16][17][18][19][20][21] ]GT(l)) and another (TCA[AT <16 ]AT(m)) increasing the successful outcome of anti-thyroid treatment.
The roles of genetic factors in the GO pathomechanism are not well characterized. Although a genetic predisposition to GD has been confirmed, to increase one's risk of developing the disease, the significance of genotype-phenotype correlations in GD remains speculative, and the results of family-based studies are conflicting. So far-and only at a univariate levelpolymorphisms located in CTLA-4 have been linked to GO risk and outcome predictors. Therefore, we tested the combined impact of the 2q33 region and GO course, and observed statistically significantly different haplotype associations in GO status. In our search for deeper insight, specifically GO course (severity and activity), the differences were more evident. Particularly, one specific haplotype TCG[AT [16][17][18][19][20][21] ]GG(m) favored not only the development of GO but also a severe outcome. The haplotypes TCA[AT >21 ]GG(m) and TCG[AT <16 ]GG(m), on the other hand, were directly linked to GD without any eye symptoms. As this study is the first association analysis based on the aforementioned concept, our observations require confirmation by future studies. It must be noted that we were able to collect a relatively large cohort of patients with severe eye symptoms and compare them with subjects with the milder form of GO.
When we used the classical approach-a univariate association model-the association with GD in our southern Polish population was seen only in case of two polymorphisms located within the CTLA-4 3′UTR region: CT60 (rs3087243) and g.*642AT(8_33). When we looked deeper, taking ancestral background into account, three CTLA-4 polymorphic variants (g.319C>T (rs5742909) [CC], c.49A>G (rs231775) [G], g.*642AT(8_33)[AT >21 ]/ [AT >21 ]) were associated with the family burden of GD. This finding may provide some evidence in support of our hypothesis; specifically that the superiority of the association defined as one susceptible locus expressed as the haplotype over the individual polymorphism association. As compared to other studies, which describe only univariate analyses [28,29,30,31], our research takes a deeper and more detailed approach. The most GD-essential genetic marker in our study, CTLA-4g.*642AT(8_33), was also relevant in other populations [2,18,25]. Importantly, this marker has functional implications; specifically the long AT-repeat allele is associated with reduced control of T-cell proliferation and thus contributes to the pathogenesis of GD [25].
Other Polish studies [29,30,31] focused mostly on CTLA-4c.49A>G (rs231775) and revealed a significant association of this marker with disease; however, we did not. On the other hand, the association of the marker CT60 (rs3087243) observed in our group was not as significant as in another study of Polish patients with GD [29]. Our observation of a significant association of the CT60 (rs3087243) [G] allele aligns with the results of other reports [9,18,28], including a large meta-analysis where this polymorphic site strongly increased GD risk [28].
Since the goal of anti-thyroid drug therapy (to restore an euthyroid state) may be achieved in up to 50 % of GD patients, defining the genetic predictors of remission could be helpful in clinical practice; therefore, we tested studied markers to forecast the clinical outcome. At a univariate level, CT60 (rs3087243) [AA] and Jo31 (rs11571302) [TT] homozygosity significantly increased the success of antithyroid treatment, but multivariate logistic regression analysis only pointed CT60 (rs3087243) as an independent risk factor for disease progression.
Previous studies focusing on the CTLA-4c.49A>G (rs231775) polymorphism [32,33,34,35] showed its association with a lower chance of remission after discontinuation of propylthiouracil treatment [32,33,34], indicating that GD patients with the CTLA-4c.49A>G (rs231775) [GG] genotype may not be the best candidates for anti-thyroid drug therapy. Moreover, GD patients with the [G] allele at this SNP needed to continue anti-thyroid drug treatment for longer periods to achieve remission, indicating that this polymorphic variant may be an indicator of poor prognosis [33,34]. In contrast, but in line with our results, Kim et al. [35] did not observe such a relationship when it came to CTLA-4g.319C>T (rs5742909) and CTLA-4c.49A>G (rs231775). So far, no research has assessed the impact of CD28 and ICOS variation on the success of medical treatment, and only variation within CTLA-4 has been documented to be a prognostic factor for remission of disease after treatment. The observed association is probably due to the functional consequence of studied polymorphic markers.
We found an association between CTLA-4 polymorphic variations and the development and course of GO on a univariate level. If severity of GO was considered an outcome, associations were observed in the case of markers for CTLA-4c.49A>G (rs231775), CTLA-4g.*642AT(8_33), CT60 (rs3087243), and Jo31 (rs11571302). When looking for a combined effect (genetics and environment), CTLA-4g.*642AT(8_33) and male gender were found to be independent risk factors for GO. CTLA-4c.49A>G (rs231775) was linked to GO course but not development. A large meta-analysis [37] of inconsistent results of studies, focusing on the association between CTLA-4 gene polymorphisms and GO, revealed no evidence in support of an association between this SNP and eye changes.
Moreover, in the one-marker association analysis no relationship between studied polymorphisms within CD28 and ICOS and GD or the course of the disease was observed in the overall group association analysis. Our results confirmed previous data showing a lack of association between these polymorphic sites and GD [9,27]. The absence of a relationship between the CD28c.17+3T>C (rs3116496) SNP and the disease was evidenced by the lack of impact on thyroid antibody production [35]; however, stratification by gender showed that females with long alleles at the ICOSc.1554+4GT(8_15) polymorphic site were more prone to GD. This is the first gender-stratified analysis; therefore, corroboration by future studies is necessary.
We, and others, have provided evidence to support the idea that, in a single gene polymorphic variation, changes within CTLA-4 may be considered to be GD risk factors and factors capable of modifying the clinical course of the disease. We also postulated that because of the location-in one common chromosomal region-and LD, the analyzed CD28/CTLA-4/ICOS gene polymorphic variability should be considered at the haplotype level, rather than by the single-marker model. Moreover, since GD and GO are associated with gender and familial autoimmune thyroid history, these factors should also be taken into account when a patient's genetic background is analyzed.