HLA‐DQ2 homozygosis increases tTGA levels at diagnosis but does not influence the clinical phenotype of coeliac disease: A multicentre study

Magnitude of gluten‐specific T‐cell responses in coeliac disease (CD) might be dependent on HLA‐DQ2 gene dose. We aimed to investigate the effects of HLA‐DQB1*02 allele dose on clinical outcomes.


| INTRODUC TI ON
Coeliac disease (CD) is a highly heritable, immune-mediated systemic disorder. The consumption of gluten-containing cereals triggers a T-cell-mediated immune cascade resulting in impairment of small intestinal villous architecture and systemic manifestations.
DQ molecules are MHC class II proteins serving critical functions in the immune response to foreign antigens. Gliadin peptides are presented (with rare known exceptions) on DQ2 or DQ8 molecules.
Tissue transglutaminase generates negatively charged peptide residues in the deamidation of gliadin. Acting as high-affinity ligands, the peptide residues bind to HLA-DQ2 and DQ8 molecules of antigen-presenting cells' (APCs). Then, CD4+ T cells recognize this complex, initiate cellular activation and trigger T-cell immune responses.
The diagnosis of CD can be excluded by HLA typing because of its outstanding negative predictive value approaching 100% (Diaz-Redondo, Miranda-Bautista, Garcia-Lledo, Gisbert, & Menchen, 2015;Rubio-Tapia et al., 2013). The current paediatric guideline recommends HLA typing if patients are diagnosed without intestinal biopsy (Husby et al., 2012). Although the diagnostic applicability of HLA typing is clear, its role in risk stratification is still under debate. Homozygotes have the potential to synthesize the highest possible number of identical DQ2 molecules with effective antigenpresentation properties. Unlike heterozygotes, who have multiple potential allele combinations, only some of them can present gliadin peptides (van Belzen et al., 2004). Accordingly, DQ2.5 homozygotes have a fivefold risk of CD, as compared to heterozygotes (Abraham & Inouye, 2015;Koning, 2012;Pietzak, Schofield, McGinniss, & Nakamura, 2009;Vader et al., 2003;van Belzen et al., 2004).
Early identification of high-risk patients would be of utmost importance. A closer follow-up and a stricter gluten-free diet might help them to avoid the development of life-threatening complications (e.g., malignancies) (Megiorni & Pizzuti, 2012;Romanos et al., 2009).
We aimed to investigate the association between HLA-DQB1*02 allele dose with clinical parameters thereby attempting to define the role of HLA status in clinical risk stratification of CD patients.

| Patients
We included patients who were (a) diagnosed with CD by the current guidelines in one of three Hungarian university clinics (1st Clinical data were retrieved from medical files retrospectively by independent investigators, blinded to the HLA status of the patients.
Since HLA genotyping is not mandatory in CD patients, it is not incorporated in our routine diagnostic management.

Scientific and Research Ethics Committee of the Medical
Research Council has granted ethical approval of this research project (45098-2/2016/EKU).

| HLA genotyping
Genomic DNA was isolated from peripheral venous blood (QIAamp DNA Blood Mini Kit). We used polymerase chain reaction with sequence-specific primers (PCR-SSP) and sequence-specific oligonucleo-

| Clinical features
Age at diagnosis corresponded to the date of definite diagnosis when gluten-free diet was introduced. We defined the clinical presentation as per Oslo criteria as classical CD (with malabsorptive syndrome, e.g., diarrhoea and weight loss, irrespective of extraintestinal manifestations) and non-classical CD (without malabsorptive syndrome e.g., atypical gastrointestinal symptoms, extraintestinal manifestations) (Ludvigsson et al., 2013). We classified diagnostic small intestinal histology according to Corazza-Villanacci (Corazza & Villanacci, 2005). Levels of tissue transglutaminase antibody (tTGA) were measured at diagnosis with ELISA. Positive tTG serology was further divided into two groups: patients with high and low titre levels were defined as >10 times or ˂10 times of the upper limit of normal (ULN), respectively. Haemoglobin levels <130 and <120 g/L indicated anaemia in men and women, respectively. Metabolic bone disease (including osteopenia and osteoporosis) was defined as measuring a T-score <−1.0 standard deviation by dual-energy X-ray absorptiometry (DEXA). Concurrent autoimmune diseases, malignancies and dermatitis herpetiformis were assessed, as well.

| Statistical analysis
We performed Pearson's chi-squared test to analyse the association between HLA risk and categorical variables and one-way ANOVA to compare age at diagnosis across HLA risk groups. p < 0.05 indicated the rejection of the null hypothesis. Statistical analysis was carried out by IBM SPSS Statistics v 20.0 (IBM's Corporate, New York, USA).

| Malignant tumours
Malignant tumours were diagnosed in three patients: a female patient (DQ2.2/DQ7 heterozygote) developed malignant melanoma at the age of 38 years (CD was diagnosed at the age of 36 years), another female patient (DQ2.5 homozygote) died of pancreatic adenocarcinoma at the age of 75 years (CD was diagnosed at the age of 59 years), and a male patient (DQ2.5 homozygote) had lung adenocarcinoma at the age of 55 years (CD was diagnosed at the age of 46 years). Here, the low case number did not allow us to perform statistical analysis.
Refractory CD did not occur in the study population.

| D ISCUSS I ON
Although the theoretical background suggests a significant gene dose effect in CD, we proved only association between HLA-DQ2 gene dose and tTGA level but not in connection with the clinical features.
Gene dose effect might determine magnitude of T-cell responses. APCs extracted from DQ2.5 homozygotes induced more prominent T-cell proliferation and interferon γ production, as compared to HLA-DQ2.5/DQX heterozygotes (X corresponded to any haplotypes except for DQ2.5). Taken together, the number of HLA-DQ2.5 molecules on APCs determines the magnitude of T-cell responses (Vader et al., 2003), in other words "quantity matters" (Koning, 2012). Enhanced antibody response, observed in our study in homozygotes as well, seems to support this theory.
Here, the chance of synthesizing (high affinity) DQ2.5 molecules is only 25%; therefore, this haplotype is deemed to be an intermediate-risk one (van Belzen et al., 2004;Vader et al., 2003). Other HLA haplotypes, not producing functioning DQ2 molecules, are accompanied by low risk of CD.
On the contrary, a few findings oppose gene dose effect. Equal magnitude of specific T-cell responses characterizes homo-and heterozygotes. In addition to heterozygotes, the amount of DQA1*05 and DQB1*02 mRNS exceeded the expected 50% of those measured in homozygotes (Pisapia et al., 2016).
Coeliac disease is strongly HLA-linked but the effect of non-HLA loci (not taken into account in this study) can outweigh that of HLA. In a study, HLA risk stratification was complemented with 10 non-HLA loci which resulted in the allocation of 10% of study population from the moderate-to the high-risk group (Romanos et al., 2009). Besides, nongenetic (environmental) factors contribute to CD phenotype (Greco et al., 1998;Gudjonsdottir et al., 2009;Piccini et al., 2012;Vermeulen et al., 2009