The HLA-DRB1 Shared Epitope in Caucasians with Rheumatoid Arthritis: A Lesson Learned from Tic-Tac-Toe

The association of rheumatoid arthritis (RA) with the major histocompatibility complex (MHC) was first reported before our current appreciation of the complexity of the MHC genetic locus. Stastny (1) reported in 1978 that the B-cell alloantigen HLA-DRw4 was found in 70% of 54 RA white patients with erosive, rheumatoid-factor-positive rheumatoid arthritis, compared to 28% of 68 normal controls (P <0.00001). Gregersen et al. (2) recognized that despite the fact that DR4 haplotypes contain at least four closely linked genes, only HLA-DRB1 exhibited a difference between the various subtypes of DR4, which was restricted to the codons surrounding amino acid position 70 of the molecule. The stretch of amino acids from amino acid residues 70-74, which corresponds to the third hypervariable region, has since been referred to as the HLA-DRB1 shared epitope (SE) (Figure 1). The HLA-DRB1 alleles with this motif that are most common in Caucasian populations are: *0401; *0404; *0405; *0408; *0101; *0102; and *1001 (3). Weyand et al. (4) reported that among RA patients with chronic, rheumatoid factor positive erosive RA, the presence of two SE-containing alleles was associated with nodular disease and joint surgery.

Figure 1.

A. Schematic representation of an HLA-DRB1 molecule based on crystallographic structure data. The shared epitope at amino acid positions 70-74 is located in the α helix. Reprinted with permission from reference 20. B. Representation of peptide bound to crystalline HLA DR1 from human cells. The structure in orange represents that electron density of bound peptide, while that in blue represents the van der Walls surface of DR1. This shows a view from above, as probably recognized by T lymphocytes through the T-cell receptor. The α helix is on top and the β domain helix is on bottom, with the peptide N terminus on the left. Reprinted with permission from reference 21.

Because of the reproducibility of the association of the SE with RA in Caucasians, many analyses have focused on the amino acid region referred to as the SE and its role in the pathogenesis of RA, including susceptibility, severity, and response to treatment (5,6). Several modifications of the SE have been proposed, including those analyzed by Morgan et al. in this issue of Arthritis & Rheumatism(7). The Tezenas du Montcel classification system (8) divides HLA-DRB1 alleles into susceptibility (S) and non-susceptibility alleles (X) according to the presence or absence of the amino acid sequence arginine-alanine-alanine (RAA) at amino acid positions 72-74 (Table 1). The S alleles are then further subdivided according to the amino acid at position 71 (S₁: ARAA [S_1A] or ERAA [S_1E], S2: KRAA, S₃: RRAA). The S₃ alleles are subdivided further according to the amino acid at position 70 (S_3D: DRRAA, S_3P: QRRAA or RRRAA).

Table 1.

The Tezenas du Montcel classification system of the HLA-DRB1 shared epitope in RA.^*

	Amino Acid Residue
HLA-DRB1 Alleles	70	71	72	73	74	Classification
Susceptibility alleles
0401, 1303†		K	R	A	A	S2
0101, 0102, 0404, 0405, 0408, 10, 1402†, 1406†	Q/R	R	R	A	A	S3P
Low-risk alleles
1501, 1502, 1503		A	R	A	A	L (S1A)
0103, 0402, 1301, 1302		E	R	A	A	L (S1E)
1102†, 1103†, 1202†, 1305†, 16	D	R	R	A	A	L (S3D)
Nonsusceptibility alleles
03	Q	K	R	G	R	X
0403, 0406, 0407	Q	R	R	A	A	X
07	D	R	R	G	Q	X
08 (except 0803)	D	R	R	A	L	X
09	R	R	R	A	E	X
14 (except 1402 and 1406)	R	R	R	A	E	X

^*In this system described by Tezenas du Montcel et al, alleles are divided into susceptibility (S) and nonsusceptibility (X) alleles according to the presence or absence of the amino acid sequence RAA at amino acid positions 72–74 (see Table 1 in ref. 8). The S alleles are then further subdivided according to the amino acid at position 71 (for S1, ARAA [S_1A] or ERAA [S_1E]; for S2, KRAA; and for S3, RRAA). The S3 alleles are then subdivided further according to the amino acid at position 70 (for S3D, DRRAA, and for S3P, QRRAA or RRRAA). In the analysis by Morgan et al (7), the S_1A, S_1E, S_3D, and X alleles were pooled and classified as low-risk (L) alleles.

^†Allele is rare in Caucasians.

Letters represent the single-letter abbreviation for amino acids: K - lysine; R - arginine; A - alanine; Q - glutamic acid; G - glycine; D – aspartic acid; L – leucine; I – isoleucine.

The de Vries classification system (9) defines SE alleles as those encoding the amino acids LQKAA, LQRAA, or LRRRA at positions 67, 70, 71, 73, and 74. The SE-negative alleles are further subdivided into “protective” (P) and “neutral” (N) alleles. In their analysis of 167 Caucasian RA patients and 166 controls, de Vries and colleagues concluded that HLA-DRB1 *07, *1201, *1301, and *1501 alleles showed significant “protective” effects. The “protective” alleles often encoded isoleucine at position 67 or aspartic acid at position 70.

The classification system reported by Mattey et al. (10) focuses on the amino acid residue encoded at position 70 (Table 2). Those investigators reported that the presence of SE-containing alleles encoding a glutamic acid at position 70 conferred the greatest risk of RA in Caucasian populations from the United Kingdom and Spain. Alternatively, two non-SE alleles encoding an aspartic acid at position 70 were associated with the lowest risk. The severity of erosive damage did not appear to be associated with the amino acid substitution at position 70.

Table 2.

The Mattey classification system of the HLA-DRB1 shared epitope in RA.^*

	Amino Acid Residue
HLA-DRB1 Alleles	67	70	71	72	73	74	Classification
Susceptibility alleles
0401	L	Q	K	R	A	A	SE
0101, 0102, 0404, 0405, 0408	L	Q	K	R	A	A	SE
10	L	R	R	R	R	A	SE
Neutral (D70-) alleles
03	L	Q	K	R	G	R	N (D70-)
0403, 0406, 0407	L	Q	R	R	A	E	N (D70-)
09	F	R	R	R	A	E	N (D70-)
14	L	R	R	R	A	E	N (D70-)
15 (1504 is rare)	I	Q	A	R	A	A	N (D70-)
Protective (D70+) alleles
08 (0803 is rare)	F	D	R	R	A	L	P (D70+)
11	F	D	R	R	A	A	P (D70+)
16	F	D	R	R	A	A	P (D70+)
0103, 0402	I	D	E	R	A	A	P (D70+)
07	I	D	R	R	G	Q	P (D70+)
12 (1202 is rare)	I	D	R	R	A	A	P (D70+)
13	I	D	E	R	A	A	P (D70+)

^*In this system described by Mattey et al, shared epitope (SE)-negative alleles are further subdivided according to the amino acid at position 70 (10). Alleles without aspartic acid (D) at position 70 (D70-) are neutral (N), while those with an aspartic acid at position 70 (D70-) appear to confer protection (P).

Letters represent the single-letter abbreviation for amino acids: K - lysine; R - arginine; A - alanine; Q - glutamicacid; G - glycine; D - aspartic acid; L - leucine; I - isoleucine.

Several other groups of investigators have reported an association of RA with the SE, but with “protective” alleles as well. For example, Larsen et al. (11) found that HLA-DR1, DR5, DR2, and combinations of DR2/DR3 and DR3/DR7 were reduced in Canadian RA patients compared to controls. In an analysis of 852 Japanese RA patients, Wakitani et al. (12) found a positive association with HLA-DRB1 *0101 and *0405, and a negative association with HLA-DRB1 *0701, *0802, *1302 and *1405. Finally, Reviron et al. (13) examined the influence of SE-negative HLA-DRB1 alleles (HLA-DRB1*X) on the development of RA in Caucasians in France. In addition to validating the association between the SE and RA, they noted a set of “protective” alleles among those without the SE. They reported that alleles with a neutral or negative electric charge in their P4 pocket, such as HLA-DRB1 *0103, *0402, *07, *08, *11 (except *1107), *12, and *13 were “protective” against RA, while those with a positive electric charge in their P4 pocket, such as HLA-DRB1 *03, *0403, *0406, *0407, *0901, *1107, *14, *15, and *16, had no influence on the predisposition to RA. A recent study documented an association of HLA-DRB1 SE-containing alleles with RA in African-Americans, with the *1101 and *1302 alleles having significantly higher frequencies in controls compared to patients (14).

Morgan et al. (7) report the findings of their analysis of three different HLA-DRB1 allele classification systems in 1,325 Caucasian RA patients and 462 healthy Caucasian controls from the United Kingdom. They confirmed the association between the S₂ alleles (particularly *0401) which encode KRRAA at amino acid residues 70-74, and S_3P alleles (*0101, *0102, *0404, *0405, and *0408), which encode either RRRAA or QRRAA at those residues, as proposed by Tezenas du Montcel (8). A significant hierarchy of risk was observed; the highest risk was seen in individuals with two S alleles, particularly those individuals who were homozygous for S₂ alleles, with a moderate risk in individuals with one S₂ and one S_3P allele or two S_3P alleles. Thus, the study by Morgan et al. validated the hierarchy of risk alleles proposed by Tezenas du Montcel (8), which replicates a previously published validation by that group (15,16).

In the study by Morgan, there was no evidence of a significant association between *1001 and RA, but this may be due to a lack of statistical power. The frequency of HLA-DRB1*1001 in the RA control population was significantly higher than in the Welsh Bone Marrow Registry cohort, likely leading to a false negative association between *1001 and RA. This study highlights the usefulness of utilizing a large registry of normal subjects (in this case a bone marrow donor registry database) to assess the allele frequencies in a study's control population, especially if the control sample size is relatively small.

Results of the analysis by Morgan did not support the hypothesis that an isoleucine at position 67 confers protection for RA, other than in contrast to susceptibility (shared epitope) alleles. However, the data were compatible with the previous findings by Mattey et al. (10), in which all alleles that encoded an aspartic acid at amino acid 70 (D⁷⁰⁺) were pooled and shown to collectively confer a “protective” effect. In order to assess this, Morgan subdivided the SE-negative alleles into D⁷⁰⁺ and D⁷⁰⁻ alleles. When the 560 individuals with no SE alleles were analyzed, there was weak evidence of a lower risk associated with D⁷⁰⁺ than D⁷⁰⁻ SE-negative alleles. It is important to note, however, that when the D⁷⁰⁺ and D^70- SE-negative alleles were compared with all other alleles, both displayed a “protective” effect of a similar magnitude.

The common theme from all the papers cited above is that the RA shared epitope is consistently and reproducibly associated with RA in Caucasians. However, moving beyond that conclusion and definitively proving that there are HLA-DRB1 alleles that confer “protection” from RA is fraught with difficulty. Theories and models from non-medical fields can be used to glean insight into this situation. John von Neumann, a well-respected mathematician, launched the field of game theory in 1928 when he published his analyses of two-player “zero-sum” games, such as chess and tic-tac-toe (17). In a zero-sum situation, there is a fixed amount of measurable reward that must be divided between the parties. In such a situation, it is impossible for one party to advance its position without the other party suffering a corresponding loss (18). In the game tic-tac-toe (also called noughts and crosses, hugs and kisses, and other names), in order for one player to win, the other player must lose.

This situation is analogous to SE-positive versus SE-negative HLA-DRB1 alleles; if there is enrichment of SE-containing alleles among patients compared to controls, there must be a corresponding depletion of some alleles among controls compared to patients. If, in conjunction with the enrichment of SE-containing alleles, there are particular alleles whose frequencies are consistently diminished, then it is impossible to determine empirically whether the affected population has over-representation of a risk allele or under-representation of a “protective” allele. However, as can be seen from the data above, it appears that the alleles that are diminished are different among the various studies. This likely reflects differences in the background allele frequencies in the populations studied.

While alternative classifications of the SE appear useful in the populations in which they were initially tested, they are not completely applicable to all groups. Such a finding highlights both the complexity of appreciating MHC-encoded alleles in disease association studies and the lack of transferability of genetic associations between populations, especially among such polymorphic loci characterized by high levels of linkage disequilibrium. Thus, further evaluations of HLA-DRB1 classification systems in non-Caucasian RA subjects such as African-Americans and Asians are needed.

While alternative SE classification systems may not be universally applied, they may still be useful in some contexts and in specific populations once validated. For example, the presence of anti-cyclic citrullinated peptide (anti-CCP) antibodies has been associated with SE alleles in Caucasians (19), but this was not evaluated in the current study by Morgan. Perhaps, an alternative method could be useful in determining the genetic contribution of the SE in a clinical subgroup of RA patients such as those that are anti-CCP positive.

For now, the repeated validation of the shared epitope hypothesis and the subsequent modifications, including a hierarchy of risk alleles, by Tezenas du Montcel evokes thoughts of “Plus ça change, plus c'est la même chose.” Continued study of the SE in increasingly large groups of patients and those of diverse genetic backgrounds will hopefully bring us closer to understanding the role of HLA-DRB1 alleles in susceptibility to RA and its severity in other racial/ethnic groups, as well as furthering the goal of developing pharmacogenetic models for predicting treatment responses.