Human Leukocyte Antigen Profile Predicts Severity of Autoimmune Liver Disease in Children of European Ancestry

Genetic predisposition to autoimmune hepatitis (AIH) in adults is associated with possession of human leukocyte antigen (HLA) class I (A*01, B*08) and class II (DRB1*03, ‐04, ‐07, or ‐13) alleles, depending on geographic region. Juvenile autoimmune liver disease (AILD) comprises AIH‐1, AIH‐2, and autoimmune sclerosing cholangitis (ASC), which are phenotypically different from their adult counterparts. We aimed to define the relationship between HLA profile and disease course, severity, and outcome in juvenile AILD.

G enetic predisposition to autoimmune hepatitis (AIH), a progressive inflammatory liver disease with a female preponderance, has been associated with alleles of the major histocompatibility complex (MHC) class I and II genes in adult patients. Susceptibility to type 1 AIH (AIH-1), which is the most common form of the disease, characterized by antinuclear antibodies (ANA) and/ or anti-smooth muscle antibodies (SMA), has been linked to MHC class II human leukocyte antigen (HLA) DRB1 alleles encoding the similar amino acid sequences LLEQKR and LLEQRR at positions 67-72 of the DRβ polypeptide. These motifs are encoded by the DRB1*0301 and DRB1*0401 alleles, which predispose adults of European ancestry from northern Europe, North America, and Iran to AIH-1 (1)(2)(3) ; by DRB1*0405, the susceptibility allele in Japan and Argentina (4,5) ; and by DRB1*0404, the AIH-1 predisposing allele in Mexico, (6) although a recent investigation has questioned its role in susceptibility. (7) A recent paper suggests the impact of specific killer cell Ig-like receptor/HLA pairs in conferring susceptibility and influencing disease progression in Japanese patients with AIH-1. (8) DRB1*1501, which is associated with protection toward AIH-1, encodes alanine (A) at position 71, suggesting that the amino acid at this position is a primary determinant of disease susceptibility or resistance. (9)(10)(11)(12)(13) Childhood AIH differs from its adult counterpart in two key aspects: firstly, a third of patients are affected by AIH type 2 (AIH-2), characterized by positivity for anti-liver kidney microsomal type 1 (anti-LKM1) and/or anti-liver cytosol type 1 (anti-LC1) antibodies; secondly, some 50% of children with laboratory and histological features of AIH-1 (i.e. ANA and/or SMA positivity and interface hepatitis) have bile duct damage on cholangiography at disease onset and are diagnosed as having autoimmune sclerosing cholangitis (ASC). (14) Reports of MHC-encoded disease susceptibility in pediatric autoimmune liver disease (AILD) have been limited to either small numbers of patients or AIH subgroups (1,(15)(16)(17)(18)(19) and have not differentiated AIH-1 from ASC. Reported HLA associations in childhood AIH are summarized in Supporting Table S1. Of note, in contrast to adult patients, the DRB1*0401 allele is not a predisposing factor and can even exert a protective role, (20,21) and the DRB1*1301 allele reported to predispose to AIH-1 in Argentinian and Venezuelan children does not conform to the shared motif model mentioned above, harboring the sequence LIEDER at positions 67-72. (11,13,22) The aim of the present study was to investigate the HLA profile and its influence on disease predisposition, course, severity, and outcome in a large series of children with AILD, including AIH and ASC, followed up at a single center, King's College Hospital (KCH), London.

patIeNtS
A total of 236 unrelated children of European ancestry (152 female, 64%) fulfilling the criteria for the diagnosis of AILD according to the European Society for Paediatric Gastroenterology, Hepatology and Nutrition (ESPGHAN) scoring system for juvenile AILD, (23) were referred to our center between 1977 and 2016 and followed up to 2019. Retrospective data were collected from 1977 to December 1999 for 116 patients. From January 2000 until June 2016, 120 patients were recruited prospectively, and their blood was collected with full consent for research, including HLA typing. The study was approved by the Ethics Committee of KCH. The informed consent in writing was obtained from each patient or their guardians since January 2000 and the study protocol conformed to the ethical guidelines of the 1975 Declaration of Helsinki as reflected in a priori approval by the Ethics Committee of KCH. For retrospective data where informed consent was not possible, the requirement for informed consent was waived by the review committee of KCH.
To confirm the AILD diagnosis, all patients were scored retrospectively using the ESPGHAN diagnostic criteria. (23) Four patients from abroad were lost to follow-up soon after diagnosis, while 232 patients were followed up until June 2019 (median 14.5 years, range 20 days-42 years, 3,489.5 person-years). Cholangiograms by endoscopic retrograde cholangiopancreatography or MR cholangiopancreatography to assess possible bile duct damage were performed in all patients, at or soon after presentation. One hundred patients had AIH-1 (median age at diagnosis 12 years, range 2.4-17), 77 had ASC (median age at diagnosis 11.3 years, range 2.1-15.7), and 59 had AIH-2 (median age 7.4 years, range 0. [8][9][10][11][12][13][14][15][16]. Viral hepatitis, Wilson disease, NASH, and other causes of liver disease were excluded by appropriate investigations. Laboratory, clinical, and histological indices were recorded in a dedicated database, those at the time of diagnosis being shown in Table 1. Two hundred and nine local subjects of European ancestry (110 female, 53%, median age 31 years, range 23-62) served as healthy controls (HCs), after providing informed consent.

Hla DeteRMINatIoN
HLA class I genotyping for 14 HLA A and 17 HLA B antigens and class II genotyping for 37 HLA DRB, 8 DQA, and 15 DQB antigens was performed by PCR/sequence-specific primers using kits obtained from Biotest (Dreiech, Germany). (15,27) Alleles were assigned to broad antigens. HLA typing was carried out in the same laboratory, under similar conditions, and by the same investigator ( J.U. who died untimely before the writing up of the paper and to whom the paper is dedicated), who was unaware of the diagnosis and the details of the patients.

DeteCtIoN oF aUtoaNtIBoDIeS
ANA, SMA, anti-LKM1, and anti-LC1 were tested by indirect immunofluorescence on a composite substrate including rat liver, kidney, and stomach at the initial dilution of 1/10 in phosphate-buffered saline according to the recommendations of the International Autoimmune Hepatitis Group. (28,29) Positive sera were double-diluted to extinction. Titers of ≥1/20 were considered positive. (28,29) StatIStICal aNalySIS c 2 and one-tailed Fischer exact tests were used to compare HLA frequencies in different groups. In order to control family-wise error rate for multiple testing, the Holm-Bonferroni correction was applied. (30,31) Relative risk (RR) was calculated as ORs. (31) The normality of variable distributions was tested using the Kolmogorov-Smirnov goodnessof-fit test. Differences in aspartate aminotransferase (AST), IgG, and bilirubin levels and autoantibody titers among patients with different HLA alleles were analyzed either by t test for parametric data or the Wilcoxon rank sum test for nonparametric data. Transplant-free survival in years from diagnosis was calculated by a Kaplan-Meier survival plot. P < 0.05 was considered significant.

ClINICal Data
Eight of 100 patients with AIH-1, 14/77 patients with ASC, and 6/59 patients with AIH-2 were already on immunosuppression for 3 months to 3 years before referral to KCH. Liver biopsy at presentation was performed in 216 patients (92%), before starting immunosuppressive treatment in 194 and within 6 months from starting treatment in 22 with coagulopathy, after prothrombin time normalization. In the remaining 20 patients, liver biopsies were performed ≥8 months after referral while on immunosuppression or in their local hospitals before referral and reported as compatible with AILD, but these were not available for review. Four out of 20 patients had complications of portal hypertension (hematemesis/melena from esophageal/gastric varices, bleeding diathesis, splenomegaly), and 16 had evidence of severe fibrosis by noninvasive tests, including ultrasound, showing nodular liver parenchyma and splenomegaly, FibroScan, , severity of liver damage and degree of fibrosis were reviewed independently for the purpose of this study by two histopathologists (A.Q. and Y.Z.) and reported as histological activity index (HAI) and fibrosis score according to the Desmet and Ishak grading systems, (21,28) which record the degree of periportal or periseptal interface hepatitis, confluent necrosis, lytic necrosis, focal inflammation, portal inflammation, and fibrosis. These scoring systems were not applied to patients with ASC as they have not been validated for this condition. Ninety-eight patients (42%) had concurrent immune-mediated diseases, including type 1 diabetes, inflammatory bowel disease (IBD; ulcerative colitis, indeterminate colitis, or Crohn disease) or celiac disease ( Table 1). All patients were treated with prednisolone at a starting dose of 2 mg/kg/day (maximum 60 mg/ day), weaned within 6-8 weeks to a maintenance dose of 2.5-5 mg/day according to child's age and weight; azathioprine (1-2 mg/kg/day) was added in the absence of gradual improvement of transaminase levels on prednisolone alone. Patients with ASC were treated equally with addition of ursodeoxycholic acid (15 mg/kg/day). Response to treatment was assessed according to ESPGHAN criteria. (23) Mycophenolate mofetil or a calcineurin inhibitor was used as second-line treatment when standard immunosuppression failed. The 232 patients followed for up to 42 years were divided in three groups: good responders (GRs; 137, 59%) achieved remission (normal transaminase and IgG levels) and maintained it long term, with or without minor, easily treatable, transaminase flares (less than twice the upper limit of normal); suboptimal responders (SRs; 45, 19.4%), comprising 21 who did not achieve complete remission despite response to treatment and 24 frequent relapsers (up to 4 relapses per decade with transaminase levels more than twice the upper limit of normal), experienced liver disease progression not requiring liver transplant (LT); (29) and patients with end-stage liver disease (ESLD; 50, 21.5%), including 44 who required LT and 6 who died before LT. Transplant-free survival tended to be higher in AIH-1 than in AIH-2 (P = 0.06) (Fig. 1A).

Hla alleleS CoNFeRRINg SUSCeptIBIlIty to aIlD
The frequencies of 5 HLA class I alleles, 9 HLA class II alleles, and two haplotypes conferring susceptibility or resistance to AILD, chosen from 20 published reports since 1991 (Supporting Table S1), were compared to local controls ( Table 2). Frequencies significantly different from local controls are presented in Fig. 2. The authenticity of the local controls was confirmed by comparing their HLA profiles to those of the three national control groups (Supporting Table S2), though the frequencies of five individual HLA alleles and the two haplotypes were not available in the national control groups. The significance of the frequency of HLA alleles was confirmed after Bonferroni correction.
Children with AIH-1 possessed more frequently HLA A*01 and B*8 than HCs. DRB1*03 was more frequent in patients with AIH-1 than in those with ASC and AIH-2. The frequency of homozygous DRB1*03 was higher in patients with AIH-1 than in HCs.
Possession of HLA B*08 and homozygosity for DRB1*03 were more frequent in patients with ASC than in HCs. Frequencies of DRB1*13 and of DRB1*13 in DRB1*03-negative patients were higher in patients with ASC than in HCs and patients with AIH-2.
Patients with AIH-2 were more frequently positive for DRB1*03 or homozygous for DRB1*03 than HCs. There was an overrepresentation of HLA DRB1*07 in patients with AIH-2 compared to HCs and patients with AIH-1 or ASC. The frequency of DRB1*07 in DRB1*03-negative patients was significantly higher in patients with AIH-2 than in HCs and patients with AIH-1 or ASC.
DRB3*0101, which encodes DR52a, was more frequent in patients with AIH-1 and ASC than in HCs. DQB1*0201 was more frequent in patients with AIH-1, ASC, and AIH-2 than in HCs.
For HLA genes conferring resistance to AILD (Fig.  2), DRB1*04, an AIH-susceptibility allele in adult AIH-1, was less frequent in patients with AIH-1 and ASC than in HCs. Among HLA DRB1*03-negative patients, the frequency of DRB1*04 in AIH-1 was comparable to that in HCs, while it was lower in patients with ASC than in HCs.
Two genes, DRB1*15 and DQB1*0302 (DQ8), previously reported to confer resistance to AIH in adults, were also found less frequently in children with AILD than in HCs. Patients with AIH-2 had a DQ8 frequency similar to that of HCs. No significant differences in the HLA profile were detected in patients with AIH-1 or ASC divided according to presence or absence of IBD (Supporting Table S3).

UNDIVIDeD aIlD
Among undivided AILD, age at diagnosis was higher in patients possessing heterozygous or homozygous DRB1*03 or DRB1*13 than in patients with DRB1*07 or "other DRs" (Table 3). Female gender was more frequent in the DRB1*07-positive group than in patients possessing other HLA types. Acute presentation was more frequent in DRB1*07-positive than in DRB1*13-positive patients. Higher baseline AST levels were seen in patients possessing DRB1*03 (heterozygous or homozygous) or DRB1*07 compared to DRB1*13-positive patients and in patients heterozygous for DRB1*03 compared to those with "other DRs." Higher bilirubin levels were found in patients possessing heterozygous DRB1*03 compared to DRB1*13-positive patients; bilirubin levels were also higher in DRB1*07-positive patients than in those with homozygosity for DRB1*03 or with DRB1*13. The highest alkaline phosphatase (AP) levels were seen in patients heterozygous for DRB1*03 or with DRB1*13 compared to "other DRs." Gammaglutamyltransferase (GGT) levels were higher in DRB1*13-positive patients than in homozygous DRB1*03, in DRB1*07, or in "other DR" patients. IgG levels in patients heterozygous/homozygous for DRB1*03 were higher than in those with other HLA types. ANA titers in patients homozygous for DRB1*03 or positive for DRB1*07 were higher than in those with "other DRs." The highest titers of SMA were seen in patients possessing DRB1*07 and DRB1*13. In patients with AIH, the median HAI was similar among those possessing different HLA class II alleles. However, the median fibrosis score was higher in patients homozygous for DRB1*03 or    positive for DRB1*13 than in those possessing heterozygous DRB1*03, DRB1*07, or "other DRs." Time to remission was shorter in patients heterozygous or homozygous for DRB1*03 and in those with "other DRs" than in those with DRB1*07 or DRB1*13. The percentage of GRs was similar among those heterozygous for DRB1*03 and "other DRs," being higher than in patients homozygous for DRB1*03, DRB1*07, or DRB1*13. The percentage of SRs was lower among patients heterozygous for DRB1*03 or with "other DRs" than in those homozygous for DRB1*03, DRB1*07, or DRB1*13. The percentage of patients with ESLD was similar among the groups, but when SRs and patients with ESLD were considered together their percentage was lower among those heterozygous for DRB1*03 or "other DRs" than in those homozygous for DRB1*03, DRB1*07, or DRB1*13. Transplant-free survival was similar among patients possessing different HLA genotypes (Fig. 1B).

aIH-1
In patients with AIH-1, mode of presentation, frequency of female gender, baseline transaminase and bilirubin levels, and HAI were similar among those possessing different HLA DRB1 alleles (Supporting Table S4). Patients with heterozygous or homozygous DRB1*03 and DRB1*13 were older and had higher IgG levels and median SMA titers at presentation than those with "other DRs." In heterozygous DRB1*03-positive patients, levels of AP were higher than in those possessing "other DRs"; GGT levels were higher in heterozygous DRB1*03-positive patients than in those with DRB1*07 or with "other DRs" and in homozygous DRB1*03-positive patients than in those with "other DRs." The highest ANA titers were seen in patients with homozygous DRB1*03.
Time to remission was shorter in patients heterozygous for DRB1*03 and "other DRs" than in patients homozygous for DRB1*03, DRB1*07, or DRB1*13. The median fibrosis score was higher in patients with AIH homozygous for DRB1*03 and in those positive for DRB1*13 than in those possessing heterozygous DRB1*03, DRB1*07, or "other DRs." The percentage of GRs was similar in patients with heterozygous and homozygous DRB1*03 or "other DRs," being higher than in patients with DRB1*07 or DRB1*13. The percentage of SR+ESLD was lower in those heterozygous for DRB1*03 than in those with DRB1*07 or DRB1*13. The highest percentage of patients requiring transplant or dying was seen among those possessing DRB1*13. Transplant-free survival was similar among patients possessing different HLA genotypes (Fig. 1C).

aIH-2
In patients with AIH-2, the proportion of females, mode of presentation, levels of IgG, and LKM-1 titers were similar among those with different HLA allotypes (Supporting Table S5). Patients with DRB1*13 were older than patients possessing other allotypes. The highest AST levels were seen in patients heterozygous for DRB1*03 and in patients with DRB1*07 or "other DRs," reaching statistical significance when patients with DRB1*07 or "other DRs" were compared to homozygous DRB1*03. Baseline bilirubin levels were higher in patients with DRB1*07 and "other DRs" compared to homozygous DRB1*03. Baseline GGT levels were higher in DRB1*13 patients compared to those with all other HLA allotypes. The highest HAI was seen in patients with "other DRs" and the lowest with DRB1*13. DRB1*13-positive patients required longer time to achieve remission than patients heterozygous or homozygous for DRB1*03 with DRB1*07 or "other DRs." The median fibrosis score was higher in patients with AIH-2 homozygous for DRB1*03 or positive for DRB1*13 than in those possessing heterozygous DRB1*03, DRB1*07, or "other DRs." The lowest proportion of GRs was seen among patients with homozygous DRB1*03 or positive for DRB1*13. Most patients with heterozygous DRB1*03 or "other DRs" responded well to treatment. The percentage of SR+ESLD was lower in those heterozygous for DRB1*03 (30%) or with "other DRs" (33%) than in those with homozygous DRB1*03 (58%), DRB1*07 (52%), or DRB1*13 (58%); but the difference was not statistically significant due to the small case number. Transplant-free survival was similar among patients possessing different HLA genotypes (Fig. 1D).

aSC
In patients with ASC, mode of presentation and titers of SMA antibodies were similar among the different HLA allotypes (Supporting Table S6). Patients with "other DRs" or with DRB1*07 were younger than patients with heterozygous DRB1*03. There were more female patients among those possessing DRB1*07 than among those with all other HLA allotypes. Baseline AST levels were higher in patients homozygous for DRB1*03 or with DRB1*07 than in those with "other DRs." The highest levels of AP and GGT were seen in patients possessing DRB1*07. IgG levels in patients homozygous for DRB1*03 and in those with DRB1*07 were similar, being higher than in those possessing other HLA alleles. ANA titers in patients heterozygous for DRB1*03 or with DRB1*07 were higher than in those with DRB1*13 or "other DRs." Time to remission in patients heterozygous or homozygous for DRB1*03 or with "other DRs" was shorter than in those with DRB1*07 or DRB1*13. The percentage of GRs was similar among patients with heterozygous DRB1*03 or with "other DRs," being higher than in patients with homozygous DRB1*03 or with DRB1*07. All patients with heterozygous DRB1*03 or "other DRs" responded well to treatment. The percentage of SR+ESLD was similar among heterozygous DRB1*03 and "other DRs," being lower than in those homozygous for DRB1*03 or positive for DRB1*07. Transplant-free survival was similar among patients possessing different HLA genotypes (Fig. 1E).

Discussion
We report the largest study of HLA association in juvenile AILD, with the longest follow-up. We included only children of European ancestry, and owing to the single-center nature of the study, all were rigorously phenotyped and divided into the three main groups of juvenile autoimmune liver disease: AIH-1, AIH-2, and ASC.
We found that HLA B*08, -DRB1*03, and the A1-B8-DR3 haplotype, known to predispose to adult AIH-1, also predispose to all three types of childhood AILD. Because in populations of European ancestry the A1-B8-DR3 haplotype does almost invariably include DRB3*0101 (DR52a) and DQB1*0201 (DQ2), not surprisingly we found that DR52a and DQ2 also predispose to juvenile AILD. Genes that were underrepresented in children with AILD, indicating their protective role, were DRB1*04 (DR4), a susceptibility gene in adult AIH-1 (32) ; DRB1*1501 (DR2), which protects against AIH also in adults (1) ; and DQB1*0302 (DQ8), a gene reported to predispose to autoimmunity when associated to DR4. (33) A genome-wide association study in a large number of adult patients from northern Europe (33) reported an association between AIH-1 and possession of both DRB1*03 and DRB1*04. However, the latter was associated with later-onset disease, confirming previous reports (34) and in keeping with the lack of association, indeed protection, we observed between possession of DR4 and the juvenile form of the disease. Interestingly, DR4 has been reported to be associated with AIH susceptibility only in DR3-negative adult patients (1) and to predispose to a later age at onset and less severe disease. This observation has led to the suggestion that AIH in adults may consist of two distinct diseases with distinct HLA associations. In support of this hypothesis, DR3/DR4 heterozygosity in adults was not associated with susceptibility. Similarly, we found no evidence that possession of DR3/DR4 predisposes to AIH in children: only a small number of our patients possess DR3/DR4 (12% of AIH-1 and 10% of AIH-2), and clinical features and outcomes of these patients are similar to those who are DR3/DR4-negative. If DR4 does predispose to a distinct form of late-onset AIH, it is not surprising that it does not predispose to juvenile AILD. DR4 protection against juvenile AILD could derive either from an unrelated HLA association with a disease distinct from adult AIH or from DR4 binding an autoantigenic peptide and preventing it from binding to DR3 or another allotype containing the shared epitope, such as DR52a.
We have also defined individual disease risk alleles for each subgroup of AILD-HLA DRB1*03 for AIH-1, DRB1*13 for ASC, and DRB1*07 for AIH-2-indicating that the three conditions are nosologically distinct entities. These results are in partial agreement with previous studies in smaller pediatric cohorts.
In summary, we have defined both HLA class I and II profiles for each subgroup of childhood AILD: DRB1*03 for AIH-1, DRB1*03 plus DRB1*07 for AIH-2, and DRB1*13 for ASC. DRB1*03 and the A1-B8-DR3 haplotype are disease-predisposing genes for all three subgroups. The influence of HLA class II genes on disease severity is strong, DRB1*03 homozygosity and possession of DRB1*13 being associated to histologically more advanced disease from onset, while DRB1*07 is linked to the least optimal response to immunosuppression. The influence of the distinct HLA genes on effector and regulatory immune responses in the different AILD subgroups should be further explored.