High Frequency of Haplotype HLA-DQ7 in Celiac Disease Patients from South Italy: Retrospective Evaluation of 5,535 Subjects at Risk of Celiac Disease

Background Celiac disease (CD) has a strong genetic component mainly due to HLA DQ2/DQ8 encoding genes. However, a minority of CD patients are DQ2/DQ8-negative. To address this issue, we retrospectively characterized HLA haplotypes in 5,535 subjects at risk of CD (either relatives of CD patients or subjects with CD-like symptoms) referred to our center during a 10-year period. Methods We identified loci DQA1/DQB1/DRB1 by sequence-specific oligonucleotide-PCR and sequence-specific primer-PCR; anti-transglutaminase IgA/IgG and anti-endomysium IgA by ELISA and indirect immunofluorescence, respectively. Results We diagnosed CD in 666/5,535 individuals, 4.2% of whom were DQ2/DQ8-negative. Interestingly, DQ7 was one of the most abundant haplotypes in all CD patients and significantly more frequent in DQ2/DQ8-negative (38%) than in DQ2/DQ8-positive CD patients (24%) (p<0.05). Conclusion Our data lend support to the concept that DQ7 represents an additive or independent CD risk haplotype with respect to DQ2/DQ8 haplotypes but this finding should be verified in other large CD populations.


Conclusion
Our data lend support to the concept that DQ7 represents an additive or independent CD risk haplotype with respect to DQ2/DQ8 haplotypes but this finding should be verified in other large CD populations.

Introduction
Celiac disease (CD) is an immune-mediated enteropathy triggered by gluten ingestion that may occur at any age in genetically predisposed individuals, and affects approximately 1% of the general population in Europe [1]. The strong genetic component of CD [2] is mainly due to HLA class II genes that encode the DQ2 and DQ8 molecules. Most (95-99%) CD patients, irrespective of age, carry these molecules that account for about 40% of the disease heritability [3,4], whereas other CD-associated genetic factors contribute little to the disease risk [5][6][7]. DQ2 exists in two highly homologous variants, DQ2.5 encoded by alleles DQA1 Ã 05/DQB1 Ã 02 and DQ2.2 encoded by alleles DQA1 Ã 02/DQB1 Ã 02; DQ2.2 is usually considered to entail a lower CD risk than DQ2.5 [8,9]. The presence of HLA-DQ2/DQ8 is necessary but not sufficient for the disease development. In fact, these molecules are also present in 30-40% of unaffected Caucasian subjects, but their absence is very rarely associated with a diagnosis of CD [1].
Given its high negative predictive value, HLA molecular typing is widely used to predict CD risk, particularly among relatives of CD patients [10,11], rather than to diagnose CD.
Recently, the European Society of Gastroenterology, Hepatology and Pediatric Nutrition (ESPGHAN) recommended HLA-typing to reinforce CD diagnosis and so avoid small intestinal biopsy in children and adolescents with gastrointestinal symptoms, IgA-tTG levels greater than 10 times the upper reference limit and EMA positivity [12]. However, a minority of DQ2/ DQ8-negative CD patients, in a variable percentage depending on geographical area, develops the disease [13].
The above considerations prompted us to investigate the CD-associated HLA haplotypes in a cohort of 5,535 individuals at risk of CD (either relatives of CD patients or subjects with CDlike symptoms) from south Italy, enrolled over a period of 10 years (2003-2013), to identify common and uncommon HLA haplotypes associated with the disease in our geographical area.

Materials and Methods
The retrospective analysis of HLA molecular typing was performed in 5,535 subjects at risk of CD referred to our Department of Laboratory Medicine of the University of Naples/Center of Advanced Biotechnology (CEINGE) of Naples, Italy, in a time window of ten years (2003)(2004)(2005)(2006)(2007)(2008)(2009)(2010)(2011)(2012)(2013). A total of 1,785 subjects presented CD-like symptoms (1,254 18 years and 531 >18 years) and 3,750 subjects were relatives of patients affected by CD (1,805 18 years and 1,945 >18 years). Written informed consent to the study was obtained both from the adult enrolled subjects and from a parent or legal guardian of the enrolled children. The study was approved by the Ethics Committee of University "Federico II" of Naples and was conducted according to the Helsinki II declaration.
Anti-transglutaminase (tTG) IgA, or IgG in subjects with IgA deficiency, were tested by ELISA using human recombinant tTG as antigen (DIA Medix Corp., Miami, FL, USA). Total serum IgA was evaluated by a nephelometric assay (BN ProSpec System; Behring, Marburg Germany). Anti-endomysium IgA levels were measured by indirect immunofluorescence on rhesus monkey esophagus substrate (Eurospital, Trieste, Italy) in tTG IgA-positive subjects to exclude false positive results.
In the presence of CD-associated antibodies and depending on the patient's age, CD was diagnosed in subjects with CD-like intestinal lesions at biopsy according to the Marsh classification [14] and ESPGHAN criteria [12].
Genotype and haplotype frequencies were reported as absolute value and in percentages. The statistical significance of differences between groups were evaluated by χ2 test and by binomial logistic regression analysis, p values <0.05 were considered significant. Statistical analysis was conducted with the PASW package for Windows (v18; SPSS Inc Headquarters Chicago IL, USA).

Results and Discussion
Among the 5,535 subjects at risk of CD enrolled in this study (3,750 relatives of CD patients and 1,785 with CD-like symptoms subjects), 3,059 were aged < 18 years and 2,476 were between 19 and 82 years of age (Fig 1). Celiac disease was diagnosed in 666/5,535 (12%) individuals; 75% of them (497 CD cases) were <18 years old (Fig 1). The prevalence of CD in the overall population was 9.6% in relatives of CD patients and 17% among subjects with CD-like symptoms, which is in agreement with previous data, namely, from 4% to 17% in CD relatives [11,15,16] and from 12% to 50%, in symptomatic subjects [15]. The prevalence of CD was higher (2.38:1) in children [497/3,059 (16.24%)] than in adults [169/2476 (6.8%)] (p<0.001), which tallies with previous Italian data (1.57:1) [17]. It was also higher in females than in males (p<0.001), the female/male ratio being 1.8-2.7 depending on the patients' age (data not shown), which also coincides with previous reports (1.5-2.0) [15,18]. The HLA genotype distribution, after normalization for age and family history differed between males and females (Table 1A).
Notably, our group of 4,869 unaffected subjects showed higher frequencies of the DQ2/ DQ8 and DQ7 haplotypes (Fig 1 and Table 2), as expected being constituted by relatives of CD patients (consequently they had a similar genetic background as their relatives affected by CD) and by subjects with gastrointestinal symptoms (in whom high DQ2/DQ8 frequencies were previously described) [20].
In our cohort, 4.2% of CD patients were DQ2/DQ8-negative (DQX/DQX), which is lower than the percent of DQ2/DQ8-negative CD patients (about 6%) previously reported in a large European CD population [13]. This discordance reflects the fact that the latter study was performed before DQ2.2 was identified as a CD-predisposing molecule [9], and hence was not considered a CD risk molecule. Furthermore, our results are in line with a higher prevalence of DQ2/DQ8-negative CD patients in south Europe than in north Europe [13].
The weight of the DQ7 (DQA05-DQB1 Ã 0301) haplotype in CD risk has been previously reported only in the presence of the DQ2.2 (DQA1 Ã 02-DQB1 Ã 0202) haplotype and was implicated in the production of the DQ2.5 molecule in trans [8].
Very recently a support to our descriptive data was given by Bergseng et al [21]. In fact, these authors demonstrated in lymphoblastoid cell lines and by relative quantitative proteomics that HLA-DQ2.5,-DQ2.2 and-DQ7.5 molecules have different specificity requirements for peptide binding and consequently distinct risks for celiac disease [21].
In conclusion, our results obtained in a large Italian cohort of children and adult CD patients lend support to the concept that DQ7 represents an additive or independent CD risk haplotype with respect to DQ2/DQ8 haplotypes. Moreover, our data questions the negative predictive value generally attributed to the absence of HLA-DQ2/DQ8 molecules in subjects at risk of CD. In fact, based on our results a diagnosis of CD should not be ruled out a priori in HLA-DQ2/DQ8-negative individuals carrying the HLA-DQ7 molecule, but this finding should be verified in other large CD populations.