Human leucocyte antigen determinants of susceptibility to Barrett's oesophagus in Asians – a preliminary study


Dr S. Rajendra, Division of Gastroenterology, Department of Medicine, Royal Perak College of Medicine, Ipoh, Malaysia.


Background : Characteristic immune profiles have been demonstrated in gastro-oesophageal reflux disease. However, the genetic basis of gastro-oesophageal reflux disease remains unclear.

Aim : To investigate whether certain human leucocyte antigen genes are associated with Barrett's oesophagus.

Methods : Asian patients of Malay, Chinese and Indian descent with Barrett's oesophagus (n = 59) and those without reflux symptoms and a normal oesophagus (n =60) were recruited prospectively using endoscopic and histopathological criteria. Human leucocyte antigen class I and II typing was performed using a polymerase chain reaction sequence-specific primers method.

Results : The HLA-B7 allele was present in 17% (10 of 59) of patients with Barrett's oesophagus when compared with 0% (zero of 60) of controls [P = 0.0006, corrected P = 0.0171, OR = 25.67]. Subgroup analysis revealed that the HLA-B7 allele was confined almost exclusively to Indians with Barrett's oesophagus, 43% (nine of 21) vs. 0% (zero of 19) Indian controls (P = 0.0014, corrected P = 0.0406, OR = 29.64). No class II associations, protective human leucocyte antigens or extended haplotypes for disease susceptibility were identified.

Conclusions : Barrett's oesophagus in Asians, particularly Indians, is strongly positively associated with HLA-B7; reinforcing a genetic component to gastro-oesophageal reflux disease. A larger sample size and different ethnic populations should be genotyped to further confirm this association and identify possible additional risk factors in the human leucocyte antigen locus.


Gastro-oesophageal reflux disease (GERD) is characterized by a diverse phenotypic response. It includes symptomatic disease with a normal mucosa to oesophagitis, stricture formation, Barrett's oesophagus and oesophageal adenocarcinoma. Endoscopically visible GERD-associated lesions, i.e. erosive oesophagitis and Barrett's oesophagus better define a population with true GERD than symptoms, pH-metry or response to antisecretory therapy.1–7 Barrett's oesophagus is found in 10% of patients with persistent reflux symptoms referred for endoscopy.8–11 It is a premalignant condition in which the transformation from squamous to specialized columnar lining of the lower third of the oesophagus predisposes to oesophageal adenocarcinoma.

There is mounting evidence for a genetic component to GERD as evidenced by familial aggregation of GERD symptoms,12, 13 Barrett's oesophagus and oesophageal adenocarcinoma.14, 15 Two recent studies of GERD symptoms in Swedish and British monozygotic (MZ) and dizygotic (DZ) twins revealed a substantial genetic contribution to the aetiology of GERD.16, 17 Furthermore, ethnic differences in the prevalence of heartburn, endoscopic oesophagitis and Barrett's oesophagus have been reported in Western18 and Asian populations.19–22 All these lines of evidence suggest a genetic aetiology and or a common exposure to environmental factors.

At present, no GERD genotype has been identified,23 and the genetic basis of the disease remains obscure. In this report, we examine if the human leucocyte antigen (HLA) gene complex, important for immune responsiveness, may be a susceptibility locus for this disease.

Indeed, characteristic immune profiles have been demonstrated in the varied phenotypic responses to GERD.24 This suggests that the immune environment may be important in determining whether an individual will develop cancer. HLA molecules perform a crucial function in the regulation of the immune response. As there is now clear evidence of immunosurveillance in cancer, it is reasonable to hypothesize that HLA types might predispose some individuals to particular malignancies.25 Furthermore, many studies have found associations between HLA alleles and susceptibility or resistance to certain diseases. As such, it would be pertinent to investigate whether certain HLA subtypes are predisposed to Barrett's oesophagus.

We therefore studied the frequencies of HLA class I and II antigens in a cohort of Asian patients to identify susceptible and preventive HLAs.

Materials and methods


Patients and controls were prospectively recruited from Hospital Ipoh between May 2003 and September 2004. These patients had been referred from the general medical and specialist clinics or their general practitioners to the open access endoscopy unit of Hospital Ipoh. This hospital is located in the city of lpoh, within the ethnically heterogeneous Kinta District in the State of Perak. This State has a population of approximately 2 million with a median age of 25.1 years of whom 32.6% are below the age of 15 years.26 The racial composition of Kinta District is as follows: 35.6% Malays, 46.6% Chinese, 14.3% Indians, 1.2% other indigenous people and 2.3% others.26

Procedures were performed by a single experienced endoscopist (SR) using Olympus GIF-130 videoendoscopes (Olympus Optical, Tokyo, Japan). About 10 mL of peripheral blood was collected in ethylenediaminetetraacetic acid (EDTA)-coated tubes for HLA typing. Demographic details of all patients including ethnicity were recorded. Written informed consent was obtained from all patients for this study. The study protocol was reviewed and approved by the Medical Research and Ethics Committee, Ministry of Health, Malaysia.

Fifty-nine Asian patients (18 Malays, 20 Chinese and 21 Indians) with an endoscopic and histological diagnosis of Barrett's oesophagus were recruited. These patients were a mix of newly diagnosed Barrett's oesophagus and those on long-term proton-pump inhibitor therapy undergoing oesophageal cancer surveillance. Assessment for Barrett's oesophagus was done if the squamo-columnar junction was located above the gastro-oesophageal junction. The presence of Barrett's oesophagus defined as columnar epithelium with specialized intestinal metaplasia obtained from any length segment of the tubular oesophagus was recorded and proven by histological examination of the involved mucosa. In addition to Barrett's oesophagus surveillance, paired biopsies were obtained from the midpoint of the Barrett's segment for research. The length of columnar lined segments containing specialized intestinal metaplasia was documented as was the presence and grade of dysplasia. Exclusion criteria included those patients with oesophageal disease other than Barrett's oesophagus. Patients with secondary causes of GERD, oesophageal varices, severe comorbidity, on anticoagulation treatment or with any contraindication to the taking of endoscopic biopsies were also excluded.

Sixty unrelated Asian control patients (20 Malays, 21 Chinese and 19 Indians) consisted of individuals undergoing upper gastrointestinal endoscopy for reasons other than reflux symptoms, Barrett's oesophagus surveillance or any form of dyspepsia. This group of patients included those who were being investigated for anaemia/faecal occult blood-positive stools, chronic diarrhoea requiring small bowel biopsy, irritable bowel syndrome and familial adenomatous polyposis. Exclusion criteria included those who had consumed antacids or antisecretory therapy in the past 2 years, prior gastric/oesophageal surgery and had a current or past history of peptic ulcer disease/gastric cancer. In control patients, two biopsies of macroscopically normal oesophageal tissue, 2 cm above the gastro-oesophageal junction were obtained. Squamous oesophageal samples were subjected to microscopic examination to exclude oesophagitis characterized by the presence of basal layer hyperplasia and papillary elongation27 and eosinophilic oesophagitis. By definition, the control group had no endoscopic or histopathological evidence of inflammation.

HLA typing

White cells were separated using Ficoll-Hypaque and the DNA extracted using the QIAamp DNA Blood Mini Kit (Qiagen, Hilden, Germany), following the manufacturer's instructions. The purity and concentration of the DNA was estimated using an Eppendorf Bio Photometer (Hamburg, Germany). The optimal DNA concentration for polymerase chain reaction (PCR) work in our laboratory was determined to be 25–250 ng/μL (70 ng/μL was optimal) and DNA purity of 1.6–1.9 (A260/A280).

The HLA-A, -B, -C, -DR and -DQ alleles for a total of 59 patients and 60 controls were typed using the PCR-sequence-specific primer (SSP) Micro SSP HLA DNA kits (One Lambda, Canoga Park, CA, USA). DNA samples, recombinant Taq polymerase (One Lambda) and a buffer mix (Micro SSP D-mix; One Lambda) were added to preoptimized primers presented dried in different wells of a 96-well tray for PCR. Negative genotype control assays which amplify a conserved region of the human β-globin gene were used to control for reaction specificity. Briefly, PCR amplification was carried out in a GeneAmp PCR system (Perkin-Elmer 9600, Cetus Instruments, Norwalk, CT, USA). After the PCR process, the amplified DNA fragments were separated by agarose gel electrophoresis (Micro SSP Gel System; One Lambda) and visualized by staining with ethidium bromide before being read in the UV transilluminator (Bio-Rad Gel Doc 1000, Hercules, CA, USA). The results were interpreted using the worksheet provided by the manufacturer. Loci with only one allele detected were considered to be homozygous.

Statistical analysis

The frequencies of HLA alleles of patients and controls were compared using the chi-square analysis of 2 × 2 tables and Yates correction. Fisher's exact test was used when any value in the cell was <5. P-values were subjected to correction according to the Bonferroni inequality method (corrected P, Pc) by multiplying the P-values with the number of alleles tested for each locus. P-values were considered statistically significant if its value was <0.05. The strength of the association was estimated by calculating the odds ratio (OR). If an allele was present or absent in all patients, Halden's modification of Woolf's formula was applied.28 EPI INFO 2000 (Centers for Disease Control and Prevention, Atlanta, GA, USA) was used for statistical analysis.


A total of 119 patients were evaluated. The mean age (s.d.) of patients with Barrett's oesophagus was 55.8 (10.3) when compared with 50.6 (12.4) for the control group (P < 0.05). The male to female sex ratio was 1.8 in the Barrett's oesophagus group vs. 0.9 in controls (P = 0.05). Both these demographic features are consistent with previously published data.29 Body mass index was not statistically different between patient groups.

The frequencies of HLA class I alleles in patients and controls are given in Table 1. The only significant association was HLA-B7 which was present in 16.9% (10 of 59) of Barrett's oesophagus patients and was absent in the control group (P = 0.0006, Pc = 0.0171, OR = 25.67). HLA-Cw3 was found in 10.2% (six of 59) of patients with Barrett's oesophagus vs. 0% (zero of 60) controls (P = 0.0130, Pc = 0.195).

Table 1.  Frequencies for HLA class I alleles for Barrett's oesophagus patients and controls
HLA allelesPatients (n = 59), n (%)Controls (n = 60), n (%)P-valuePcOR
  1. HLA, human leucocyte antigen; N.S., not significant; Pc, corrected P; OR, odds ratio.

A110 (16.9)5 (8.3)N.S.  
A223 (39.0)22 (36.7)N.S.  
A36 (10.2)4 (6.7)N.S.  
A1125 (42.4)23 (38.3)N.S.  
A231 (1.7)0 (0.0)N.S.  
A2423 (39.0)16 (26.7)N.S.  
A261 (1.7)3 (5.0)N.S.  
A300 (0.0)1 (1.7)N.S.  
A314 (6.8)2 (3.3)N.S.  
A321 (1.7)0 (0.0)N.S.  
A3311 (18.6)14 (23.3)N.S.  
A342 (3.4)3 (5.0)N.S.  
A680 (0.0)4 (6.7)N.S.  
B710 (16.9)0 (0.0)0.00060.017125.67
B85 (8.5)4 (6.7)N.S.  
B137 (11.9)4 (6.7)N.S.  
B186 (10.2)6 (10.0)N.S.  
B271 (1.7)3 (5.0)N.S.  
B356 (10.2)10 (16.7)N.S.  
B371 (1.7)0 (0.0)N.S.  
B381 (1.7)4 (6.7)N.S.  
B391 (1.7)1 (1.7)N.S.  
B403 (5.1)0 (0.0)N.S.  
B443 (5.1)7 (11.7)N.S.  
B464 (6.8)7 (11.7)N.S.  
B481 (1.7)1 (1.7)N.S.  
B490 (0.0)2 (3.3)N.S.  
B5110 (16.9)9 (15.0)N.S.  
B522 (3.4)7 (11.7)N.S.  
B541 (1.7)0 (0.0)N.S.  
B554 (6.8)5 (8.3)N.S.  
B561 (1.7)0 (0.0)N.S.  
B575 (8.5)3 (5.0)N.S.  
B581 (1.7)1 (1.7)N.S.  
B608 (13.6)8 (13.3)N.S.  
B615 (8.5)3 (5.0)N.S.  
B627 (11.9)5 (8.3)N.S.  
B631 (1.7)0 (0.0)N.S.  
B701 (1.7)0 (0.0)N.S.  
B711 (1.7)1 (1.7)N.S.  
B759 (15.3)13 (21.7)N.S.  
B771 (1.7)2 (3.3)N.S.  
Cw111 (18.6)9 (15.0)N.S.  
Cw36 (10.2)0 (0.0)0.01300.195 
Cw411 (18.6)16 (26.7)N.S.  
Cw51 (1.7)0 (0.0)N.S.  
Cw66 (10.2)4 (6.7)N.S.  
Cw727 (45.8)32 (53.3)N.S.  
Cw88 (13.6)12 (20.0)N.S.  
Cw92 (3.4)2 (3.3)N.S.  
Cw1014 (23.7)13 (21.7)N.S.  
Cw111 (1.7)0 (0.0)N.S.  
Cw125 (8.5)13 (21.7)N.S.  
Cw147 (11.9)6 (10.0)N.S.  
Cw155 (8.5)4 (6.7)N.S.  
Cw162 (3.4)1 (1.7)N.S.  
Cw171 (1.7)0 (0.0)N.S.  

No class II HLA-DR and -DQ associations, protective HLAs or extended haplotypes for disease susceptibility were identified (data not shown).

As there are ethnic differences in the prevalence of GERD in Malaysia,20–22 we reanalysed all the data in relation to frequencies of HLA class I and II alleles in all patients and controls, segregating them into Malay, Chinese and Indian groups. In general, the results of subgroup analysis were similar to the results of the total patient and control population (data not shown). Nevertheless, subgroup analysis revealed that the HLA-B7 allele was confined almost exclusively to Indian patients with Barrett's oesophagus, i.e. 42.8% (nine of 21) vs. 0% (zero of 19) Indian controls (P = 0.0014, Pc = 0.0406, OR = 29.64). As subgroup analysis reduces the sample size, we reanalysed these results with an additional 232 Indian controls drawn from the Malaysian Bone Marrow Donor Registry (unpublished data). A substantial number of the Bone Marrow Donor Registry group may well have GERD, as a third of Malaysians experience heartburn annually, 9.7% monthly and 6% weekly.20 Moreover, nearly 40% of Malaysian patients with Barrett's oesophagus do not report reflux symptoms.21 Nevertheless, the association of HLA-B7 in Indians with Barrett's oesophagus remained significant when compared with this group of controls (P = 0.0183, OR = 3.30, data not shown). Amongst the Chinese, HLA-A24 was present in 30% (six of 20) of patients compared with 4.8% (one of 21) in the ethnically matched control group (data not shown). However, the difference lost significance when the P-value was corrected (P = 0.0448, Pc = 0.5824). There was no significant association of class I in the Malay population with Barrett's oesophagus when compared with the control group.

Table 2 documents clinical and histopathological characteristics of HLA-B7-positive and -negative patients with Barrett's oesophagus. HLA-B7-positive patients with Barrett's oesophagus had a significantly higher family history of heartburn when compared with their HLA-B7-negative counterparts.

Table 2.  Comparison of clinical and histopathological characteristics of HLA-B7-positive and -negative patients with Barrett's oesophagus
 HLA-B7- positive (n = 10)HLA-B7- negative (n = 49)P-value
  1. * Fisher's exact test.

Mean age at diagnosis51.4 (8.3)56.7 (10.5)N.S.
Male sex632N.S.*
Mean body mass index27.3 (3.6)24.9 (5.2)N.S.
Family history of heartburn6110.026*
Extraoesophageal symptoms419N.S.*
On proton-pump inhibitors428N.S.*
Long segment Barrett's oesophagus325N.S.*


The HLA system with its extensive polymorphism is an excellent marker for population genetic analyses and disease association studies. We believe this is the first published study on HLA determinants of susceptibility to Barrett's oesophagus. Our results reveal that the B7 allele is associated with Barrett's oesophagus in Asians, especially Indians who have been shown to have the highest prevalence of GERD in Malaysia.20–22 A recent, prospective cross-sectional survey using a race-stratified disproportionate random sample of the Malaysian population (n = 949) revealed that the Indians suffered significantly more heartburn at least once a year (42%) when compared with the Malays (29%) or Chinese (29.3%).20 In Singapore, a country with a similar multiracial mix, Ho et al. have reported that reflux symptoms were twice as common in Indians when compared with Chinese.19 In Malaysia, a study of 1985 consecutive patients undergoing gastroscopy in Hospital Ipoh between 1997 and 2000 revealed that the prevalence of endoscopic oesophagitis was significantly higher in Indians (8.5%) when compared with the Chinese (4.5%) or Malays (5.6%).21 Barrett's oesophagus (both long and short) was similarly found to be significantly higher in Indians (8.2%) when compared with the Chinese (5.7%) and Malays (4.4%). Furthermore, another hospital-based study in Kuala Lumpur, Malaysia revealed that reflux oesophagitis was significantly more common amongst Indians when compared with the other races.22 Non-erosive reflux disease was also more frequently seen in Indians and Malays when compared with the Chinese.22

A genetic component to GERD has been demonstrated in severe paediatric reflux disease with an autosomal dominant inheritance in five families. However, this association between a locus on chromosome 13q14 and severe paediatric GERD,30 could not be demonstrated in another linkage study involving six families with a similar autosomal dominant inheritance GERD phenotype. This raises the possibility of genetic heterogeneity in the inheritance of paediatric GERD.31

It is well-known that familial clustering of reflux symptoms is seen in relatives of patients with reflux symptoms and increased oesophageal acid exposure and in relatives of patients with Barrett's oesophagus.12 Romero et al. also reported that reflux symptoms are more common among relatives of patients with Barrett's oesophagus or oesophageal adenocarcinoma but not among relatives of patients with endoscopic oesophagitis.13 Interestingly, the HLA-B7-positive patients with Barrett's oesophagus in our study had a significantly higher family history of heartburn, compared with their HLA-B7-negative counterparts indicating that this genetic component may be associated with particular disease manifestations of GERD.

A study of GERD symptoms in 8411 Swedish twin pairs over the age of 55 years found a casewise concordance for GERD of 31% among female MZ twins when compared with 21% in female DZ twins.16 Heritability was estimated to account for 30% of the liability to GERD. A British study of 1960 twin pairs to determine the relative contribution of genetic and environmental influences revealed that casewise concordance were significantly higher for MZ than DZ twins (42% vs. 26%; P < 0.001).17 Multifactorial liability threshold modelling suggested that 43% of the variation in liability to GERD was due to multiple small genetic effects.

Although many other diseases have been associated with HLA genes, exactly which loci predispose to disease is uncertain. Neighbouring alleles can ‘hitchhike’ if association between them is maintained artificially high by linkage disequilibrium.32 It is possible that HLA-B7 is a disease marker for Barrett's oesophagus in Asians; an allele ‘hitchhiking’ with the true disease predisposing gene. Moreover, the highly polymorphic nature of the HLA complex and linkage to the region suggests that additional disease susceptibility genes may be identified. Association with major histocompatibility complex (MHC) class I genes may reflect linkage disequilibrium to other MHC or non-MHC genes or may reflect a role for MHC class I molecules in disease pathogenesis. Malignant transformation of cells is frequently associated with altered HLA class I expression and/or function.33 These abnormalities provide tumour cells with escape routes from immune recognition.34 Class I molecules are also important to fight some viral and bacterial infections, although no infective aetiology has yet been implicated in the pathogenesis of Barrett's oesophagus.

In our study, racial subgroup analysis revealed no significant HLA association(s) in the Malays or Chinese. Subgroup analysis further reduces the sample size, which may have been insufficient to detect possible associations reliably. On the basis of 59 cases, an α-value of 0.05 and a prevalence of a given allele of up to 5% in the control group, we had about 80% power to detect a difference between patients and controls with an OR of 5 or more. In our study therefore, we were able to detect only strong associations. When the Bonferroni correction is made for the number of comparisons, the power is much lower, suggesting that the associations actually noted are likely to be seen as even more significant in a larger study. It is conventional to multiply the observed significance level by the number of alleles tested in genetic comparison to avoid rejecting the null hypothesis too readily.35, 36 Unfortunately, reducing the type I error for null associations increases the type II error for the associations that are not null.37 Furthermore, HLA genes might only account for a small proportion of the heritability for GERD in some racial groups and non-HLA genes and or environmental factors may have a greater effect. In GERD, environmental influences acting in concert with genetic factors contribute to disease susceptibility and the clinical phenotype. Ethnic variations in relative contributions of the environment and genetics may account for the insignificant HLA associations noted in the non-Indians. Given the ethnic variation in frequencies of MHC alleles, it is not surprising that different alleles may account for susceptibility and resistance to Barrett's oesophagus in various racial groups. This is exemplified in different MHC class II alleles being linked to susceptibility to Crohn's disease in Japanese (DQB1*0401, DQB1*0402)38 and Caucasian patients (DRB3*0301/DRB1*1302).39

In conclusion, Barrett's oesophagus was found to be significantly positively associated with HLA-B7 in Asians, particularly Indians. It is tempting to speculate that the disease gene responsible for the development of Barrett's oesophagus maybe the HLA-B7 allele or other gene(s) tightly linked to this allele. These findings further reinforce the concept of a genetic basis for GERD. A larger sample size and different ethnic populations should be genotyped to further confirm this association and identify possible additional risk factors in the HLA. Furthermore, higher resolution analysis of the HLA-B7 locus should be performed to determine which particular allelic subtype or linked gene is associated with GERD. Finally, genome wide scans to identify additional non-HLA genes for disease susceptibility are equally important.


Authors thank Prof. Kannan Kutty (Universiti Teknologi MARA, Shah Alam, Selangor) and Dr Mohan Yadav (Hospital Pantai, Kuala Lumpur) for helpful suggestions and critical reading of the manuscript. Drs Fauziah Jaya, Ker Hong Bee, Letchumanan Ramanathan and Arivindan Thurairaj (Hospital Ipoh) kindly helped with patient recruitment. Authors gratefully acknowledge Mohamad Azmi Hamdi, Mohamed Rasyeed, Mohamed Zakaria, Prem Singh and Tan Mai Chin of the Endoscopy unit, Hospital Ipoh for their technical assistance.

This work was supported by grants from AstraZeneca Malaysia, Malaysian Society of Gastroenterology and Hepatology and Pfizer Malaysia. The study sponsors had no involvement in the study design, collection, analysis and interpretation of the data or in the writing of the paper. The decision to submit this paper for publication was solely that of the authors.

Conflict of interest

None declared.