Genetic background of systemic lupus erythematosus

Systemic lupus erythematosus (SLE) is a prototype of systemic autoimmune diseases. Epidemiological data indicate that genetic factors play a substantial role in the disease, and intensive searches for susceptibility genes are underway by genome-wide as well as candidate gene approaches (Tsao 2004).

Genome-wide linkage analyses have been reported from five groups from North America and Europe. Collectively, more than 50 regions have been identified as meeting the suggestive level of linkage (LOD≥1.9 or P<0.01), including seven regions showing a significant level of linkage (LOD≥3.3), 1q22-23, 1q41-42, 2q37, 4p16, 6p21-11, 16q13 and 17p13 (Kelly et al. 2002).

Studies on biological and positional candidate genes have reported a number of potential susceptibility genes. Among these, the HLA-DRB1 (DRB1*1501 in Asians and Caucasians, DRB1*1503 in African-Americans and DRB1*0301 in Caucasians) and C4A*Q0 alleles, both located within the MHC complex at 6p21.3, are the most well-established. As we will see later, genes within the low affinity FCGR cluster at 1q23 (FCGR2A, 3A, 3B) have frequently been reported to be associated with SLE, although the results are not very consistent; as is the case with mannose-binding lectin (10q11.2-12). PDCD1, coding for an inhibitory receptor PD-1 on the T cells, has been identified as being associated with SLE through a positional candidate approach (Prokunina et al. 2002). Recently, an association of IRF5 polymorphism was reported in northern European populations (Sigurdsson et al. 2005), and this has been convincingly replicated in a large-scale study of four cohorts (Sweden, Spain, Argentina and USA) (Graham et al. 2006), suggesting that IRF5 might be a susceptibility gene common to various populations.

We have been trying to identify susceptibility genes for rheumatic diseases through polymorphism screening and association studies of the functional genes in the immune system (Tsuchiya et al. 2002). The role of B lymphocytes in the pathogenesis of SLE has been established from a number of observations, such as the production of an array of autoantibodies and abnormalities in the peripheral blood B cell subsets, as well as a therapeutic effect of the B cell-depleting anti-CD20 monoclonal antibody (mAb) rituximab (Anolik and Sanz 2004); thus, genes coding for B cell regulatory molecules are considered to be strong candidates for susceptibility genes. Here, we will review a series of studies performed by us on the association of polymorphisms of two B cell inhibitory receptors with SLE.

Detection of FCGR2B polymorphism in the transmembrane region and its association with SLE in Asian populations

FCGR2B codes for a receptor for the Fc portion of IgG expressed on B cells and monocytes, Fcγ receptor IIb (FcγRIIb). FcγRIIb contains immunoreceptor tyrosine-based inhibitory motif (ITIM) in its cytoplasmic region, which becomes phosphorylated when FcγRIIb is coligated with B cell receptor (BCR) by antigen–IgG complexes and recruits SH2-containing 5′-inositolphosphate phosphatase (SHIP). SHIP hydrolyzes phosphatidylinositol-3,4,5-triphosphate (PIP3) and attenuates activation and survival signals in B cells (Takai 2002). Thus, FcγRIIb represents one of the molecules that prevents excessive B cell activation.

Indeed, Fcgr2b-deficient mice become susceptible to autoimmune diseases such as collagen-induced arthritis (Yuasa et al. 1999) and Goodpasture’s syndrome (Nakamura et al. 2000). Moreover, depending on their genetic backgrounds, they spontaneously develop lupus-like disease (Bolland and Ravetch 2000; Yajima et al. 2003) with the production of rheumatoid factor and anti-DNA antibody (Moll et al. 2004). Such mice exhibit an increased frequency of anti-DNA antibody-producing plasma cells, suggesting a role of FcγRIIb in the maintenance of peripheral tolerance (Fukuyama et al. 2005). Furthermore, mice strains which spontaneously develop lupus or autoimmune disease, such as NZB, MRL and NOD, all share Fcgr2b polymorphisms (Jiang et al. 1999, 2000), and the development of autoimmunity is inhibited in such mice by the transfer of the wild-type Fcgr2b gene (McGaha et al. 2005).

Human FCGR2B gene and other classical FCGR genes such as FCGR2A, 2C, 3A and 3B form a gene family clustered within ~200-kb of 1q23 (Su et al. 2002). These genes, which are presumed to have developed through unequal crossover events, are extremely homologous to one another, not only in exons but also in introns and intergenic regions.

Association with FCGR2A, 3A and 3B polymorphisms as well as linkage to this region has been reported on multiple occasions (Hatta et al. 1999; Tsuchiya and Kyogoku 2005). Unlike FCGR2B, FCGR2A, 3A and 3B mediate activation signals and phagocytosis in cell lineages such as neutrophils, monocytes and NK cells. All of these genes have functional polymorphisms, and, interestingly, alleles that are reported to be associated with SLE at each locus (FCGR2A-131Arg, FCGR3A-176Phe [also referred to as 158Phe], FCGR3B-NA2) uniformly exhibit lower affinity to IgG or decreased phagocytotic potential compared with the nonassociated alleles. The mechanism of association has been explained by defective clearance of the immune complex (Dijstelbloem et al. 2001). However, prior to our studies, linkage disequilibrium (LD) among these genes and FCGR2B had not been investigated; therefore, the possibility that FCGR2B might be the causative susceptibility gene that explains the linkage and association signals of this region had not been formally excluded.

Based on such a background, FCGR2B was considered a strong candidate susceptibility gene to SLE. However, polymorphism screening of FCGR2B proved to be extremely difficult because the nucleotide sequence of FCGR2B was almost identical to FCGR2C from the 5′ flanking region to exon 6, in terms of both exons and in introns, and the specific amplification of FCGR2B was not successful. Furthermore, some of the single nucleotide differences between FCGR2B and 2C found in the database turned out to be polymorphisms.

We employed nested PCR to specifically amplify the FCGR2B gene, and identified a single nucleotide polymorphism (SNP) (c.695T>C, rs1050501) which led to an amino acid substitution within the transmembrane domain at position 232, Ile232Thr (Kyogoku et al. 2002). (The same SNP is also referred to as Ile187Thr by excluding the signal sequence; Li et al. 2003.) A case-control association study on Japanese indicated that there was a significant association of the 232Thr/Thr genotype with SLE (Kyogoku et al. 2002). We subsequently replicated the association of this polymorphism with SLE in Thai (Siriboonrit et al. 2003) and Chinese populations (Chu et al. 2004). When the data from these three Asian populations were combined using meta-analysis, the odds ratio of 232Thr/Thr for developing SLE was 2.45 (95% confidence interval 1.49–4.02, P=0.0004) against 232Ile/Ile genotype (Table 1a). This association was predominantly observed in the patients with lupus nephritis (Chu et al. 2004).

Table 1 Population differences in the FCGR2B alleles associated with SLE
Full size table

FCGR2B was found to be in LD with FCGR3B, 3A and 2A. In Asian subjects, association with SLE was also found for FCGR3B-NA2 and 3A-176Phe. The association of each of the three FCGRs could not be explained solely by LD with other FCGR; thus, each of the FCGRs appeared to have its own contribution (Kyogoku et al. 2002; Chu et al. 2004).

Population difference

It is becoming increasingly clear that susceptibility genes are often different among populations. For example, the association of PTPN22-Arg620Trp with rheumatoid arthritis (RA) has now been replicated in multiple Caucasian populations (Gregersen and Batliwalla 2005), but this allele is not found in Japanese (Mori et al. 2005).

Although the results are not consistent, many studies have demonstrated associations of FCGR2A, 3A and 3B with SLE. Recent meta-analyses have demonstrated that FCGR2A is associated with the development of SLE (Karassa et al. 2002), while FCGR3A is associated with the presence of nephritis (Karassa et al. 2003). These meta-analyses are mostly based on data from Caucasian populations. As described above, we observed associations of FCGR3A and 3B (in addition to FCGR2B) in three Asian populations, but not an association of 2A (Chu et al. 2004). This difference was not caused by differences in the genotyping methods, because we observed an association of FCGR2A, but not 3A, with SLE in Caucasians (Kyogoku et al. 2004). Thus, this difference seems to be caused by differences by the populations.

As for FCGR2B, association with FCGR2B-232Thr was found in three Asian populations. However, this allele was rare in Caucasians, and significant association with SLE was not detected (Li et al. 2003; Kyogoku et al. 2004; Magnusson et al. 2004) (Table 1a).

On the other hand, two groups demonstrated an association of FCGR2B promoter polymorphisms in Caucasians (Su et al. 2004a; Blank et al. 2005). However, this promoter allele was not found in Japanese (unpublished results). This might be another example of a difference between populations, but it also may represent an allelic heterogeneity between populations (Table 1b). The finding that two different polymorphic sites of the same gene are associated with the same disease in different populations may provide strong evidence to support the role of the gene in the disease. Differences between the alleles associated with SLE among populations have also been reported in PDCD1. While the A allele at an intron 4 SNP called PD1.3 has been shown to be significantly increased among Swedish, European American and Mexican patients (Prokunina et al. 2002), the same allele was significantly decreased in Spanish SLE (Ferreiros-Vidal et al. 2004).

The two groups reported conflicting results regarding the activity of the SLE-associated promoter allele. While Su et al. (2004a, 2004b) reported a SLE-associated promoter haplotype containing the −386C allele (rs3219018), Blank et al. (2005) reported a lower promoter activity for the −386C allele (they referred to the same SNP as −343G>C by counting from the transcription start site). This discrepancy may be caused by differences between the experimental systems used, although lower promoter activity is in better agreement with other lines of evidence, including our own, as described later.

A copy number polymorphism (CNP) is known to exist for FCGR3B. In addition to the common haplotype carrying one FCGR3B locus, haplotypes carrying no or two FCGR3B loci are present. Thus, the number of FCGR3B loci in each individual varies from 0 to 4. FCGR3B CNP appears to be common in Caucasians, and it was recently reported that low copy number of FCGR3B is associated with the presence of nephritis in SLE in the northern European population (Aitman et al. 2006). Detection of FCGR3B CNP is technically challenging, and population data are not available for other populations. However, we found that homozygosity of the reduction-of-function allele NA2 is associated with nephritis in SLE in Asians (Chu et al. 2004), supporting the idea that a reduction of function for FcγRIIIb is somehow associated with development of nephritis.

FCGR2B-232Thr is a reduction-of-function inhibitory receptor

Since homozygous deletion of FCGR2B in mice has been repeatedly shown to be susceptible to autoimmune diseases (Takai 2002; Nimmerjahn and Ravetch 2006), functional difference between the polymorphic FCGR2B allele products was the next focus of interest. Human FcγRIIb exists in two isoforms, FcγRIIb-1 and b-2, and in the latter, the first cytoplasmic exon that inhibits receptor internalization is spliced out. It is generally accepted that the b-1 isoform is expressed in B cells, and the b-2 isoform in myeloid cell lineage including dendritic cells (Daeron 1997). Recent works highlighted the critical role of FcγRIIb in B cells in preventing tolerance breakdown, based on the observations that retroviral FcγRIIb gene transfer to B cells effectively suppresses autoimmunity associated with reduced FcγRIIb expression in NZB, MRL and NOD mice (McGaha et al. 2005).

In order to characterize the polymorphic human FCGR2B allele products in the context of B cells, we took advantage of the ST486 human B cell line, which is inherently devoid of FcγRIIb expression (Koncz et al. 1998). The IIA1.6 subline of mouse A20 cells, which also lacks FcγRIIb, has so far been utilized as a competent cell line. However, IIA1.6 is obviously inappropriate in this setting for studying Ile232Thr single amino acid substitution, in light of the modest amino acid identity (∼60%) of human and mouse FcγRIIb. Indeed, a recent functional analysis of the polymorphic FcγRIIb in IIA1.6 (Li et al. 2003) has led to different conclusions from ours using ST486 human B cells (Kono et al. 2005), and also from those using human macrophages (Floto et al. 2005).

Each allele of the human FcγRIIb-1 isoform was introduced into ST486 cells, and two groups of cells with equivalent FcγRIIb-1 expression were established. First we determined ligand-binding parameters. The results showed that the dissociation constants of human FcγRIIb for IgG are in the order of 10−7 M, values comparable with those in the previous studies (Nimmerjahn and Ravetch 2006), and that Ile232Thr substitution does not significantly affect the binding affinity.

Recent analysis of receptor-proximal events in a variety of systems has featured a novel view of initial signaling that involves specialized plasma membrane domains, variously referred to as lipid rafts, detergent-insoluble membrane domains (DRMs) or detergent-insensitive membranes (DIMs) (Simons and Ikonen 1997; Ikonen 2001; London 2002). Lipid rafts are small membrane patches (diameter <100 nm) consisting of a plasma membrane bilayer and constructed with specific lipids including cholesterol and glycosphingolipid. Lipid rafts accumulate receptor-proximal signaling molecules harboring saturated acyl-chain anchors, including Src family kinases, heterotrimeric G proteins and adaptor proteins such as linker for activation of T cells (LAT) and C-terminal Src kinase (Csk) binding protein (Cbp) (Pierce 2002). Receptors such as T and B cell receptors, Fc receptors and TNF family receptors become associated with, or induce the aggregation of, lipid rafts shortly after agonist addition (Fig. 1). It is convincingly argued that this process precedes tyrosine phosphorylation signaling, and, moreover, is required for the de novo generation of signaling (Honda et al. 2000; Suzuki et al. 2000; Kono et al. 2002; Pierce 2002). Since we have unveiled the critical roles of their transmembrane domains in the association of Fc receptor with lipid rafts (Kono et al. 2002), we hypothesized that Ile232Thr substitution influences FcγRIIb redistribution to lipid rafts.

Fig. 1
figure 1

Schematic hypothetical representation of the FcγRIIb-232Ile redistribution to lipid rafts (upper panel) and 232Thr exclusion from lipid rafts (lower panel). Note that, even if 232Thr is (partially) excluded from lipid rafts, this does not necessarily mean that 232Thr is concomitantly excluded from B cell receptor (BCR) signaling machinery. IC immune complex

Full size image

Distribution of receptor to lipid rafts is best measured by the surface probing of the receptor with specific antibody Fab labeled with 125I, followed by its solubilization with nonionic detergents and fractionation under density gradient ultracentrifugation (Field et al. 1997; Holowka et al. 2000). However, we found that it was impossible to prepare anti-human FcγRIIb Fab with sufficient affinity for the analysis from the available FLI8.26 and AT-10 mAbs. We thus employed two different strategies to perform the assay (Kono et al. 2005). First, we created chimeric FcγRIIbs in which extracellular domain of mouse FcγRIIb was fused to transmembrane (containing 232Ile or Thr) and to intracellular domain of human FcγRIIb, and we probed this with high affinity 125I-2.4G2 mAb Fab. Second, human FcγRIIb was probed with 125I-AT10 whole IgG. Both systems yielded consistent results; 232Thr was relatively excluded from lipid rafts as compared to 232Ile under resting conditions and after coligation with BCR, and the differences were statistically significant (Kono et al. 2005).

It is widely accepted that FcγRIIb is able to attenuate BCR signaling only when it is physically coligated with BCR (Muta et al. 1994; Ravetch and Lanier 2000). These features are different from the other B cell inhibitory coreceptors CD22 and CD72, which also possess ITIM: CD22 and CD72 are able to suppress BCR signaling without coligation (Adachi et al. 2000; Nitschke and Tsubata 2004). Although this concept is supported by several series of reports, it may be worth bearing in mind that this idea derives from the observations in mouse B cells. In the current study, it was reproducibly observed that human FcγRIIb attenuates BCR signaling in ST486 immortalized human B cells even when BCR was ligated separately from FcγRIIb (Kono et al. 2005).

With respect to the allelic differences, we found that the extent of either constitutive or coligation-induced inhibition was significantly greater in 232Ile than that in 232Thr by monitoring a variety of signals including phosphatidylinositol 3,4,5-trisphosphate (PIP3) levels, phospholipase Cγ2 (PLCγ2) and Akt tyrosine phosphorylation and calcium mobilization (Kono et al. 2005). From the functional aspects, therefore, it was concluded that human FcγRIIb 232Thr represents a reduction-of-function receptor isoform when taking 232Ile as a standard. This conclusion was supported by Floto et al. (2005), who showed that FcγRIIb 232Ile exerts greater inhibitory effects than 232Thr in human macrophages on FcγRI- and FcγRIII-mediated phagocytosis, superoxide generation, cytokine production and MHC class I and II expression. In addition, it should be noted that, under the experimental conditions of Floto et al., FcγRIIb was not coligated with FcγRI. These findings indicate common mechanisms underlying constitutive inhibition by FcγRIIb in human B cells and macrophages.

Does the altered subcellular distribution of 232Thr, the partial exclusion from lipid rafts, provide a mechanistic basis for its phenotype? Floto et al. (2005) coincidentally showed defective 232Thr distribution to lipid rafts, and concluded it to be responsible for the loss-of-function characteristics of 232Thr. It is certainly attractive to speculate that 232Thr is excluded from lipid rafts, and concurrently from BCR signaling complex (Fig. 1). However, it should be emphasized that no firm evidence exists in support of the idea that association with lipid rafts is a prerequisite for human FcγRIIb to inhibit signaling from BCR or activating FcγRs. Indeed, Fas-mediated apoptosis has been shown to depend upon its association with lipid rafts in certain cell types (type I cells), but not in others (type II cells) (Muppidi and Siegel 2004). The determinant in the transmembrane structure required for FcγRIIB to associate with lipid rafts, and to efficiently attenuate BCR signaling needs to be elucidated in future studies.

Polymorphisms affecting the alternative splicing of CD72

CD72 is a 45 kDa type II transmembrane protein containing a C-type lectin domain expressed in B cells, dendritic cells and macrophages as a homodimer. CD72 contains two ITIMs in its cytoplasmic tail, one of which recruits SH2-containing tyrosine phosphatase (SHP)-1, and functions as a negative regulator of BCR signaling (Adachi et al. 2000; Hokazono et al. 2003; Nitschke and Tsubata 2004).

Human CD72 gene is encoded on human chromosome 9p13.2. In a murine model of SLE, MRL/lpr, CD72c allele has been shown to be associated with systemic vasculitis phenotype (Qu et al. 2000), indicating that CD72 accounts for one of the background genes of MRL.

Based on such information, we considered CD72 as a candidate susceptibility gene for human SLE. Polymorphism screening revealed four polymorphic sites with minor allele frequencies >0.01, two of which were 13-nucleotide repeat polymorphism (once or twice) and four-nucleotide deletion in intron 8. These four polymorphic sites were in LD, forming two major haplotypes. The haplotypes were represented by the number of the 13-nucleotide repeat used as the tag polymorphism (*1 and *2) (Hitomi et al. 2004).

A case-control association study did not show significant association with the development of SLE; however, genotype *1/*1 was increased and *2/*2 was decreased in the SLE patients with nephritis. Furthermore, when the combination of the FCGR2B and CD72 genotypes was examined in the Japanese and Thai populations, it was revealed that the association of FCGR2B-232Thr/Thr with SLE was more striking in individuals with CD72-*1/*1 genotype (P=0.009), while the risk of this FCGR2B genotype was substantially decreased and did not reach statistical significance in individuals with CD72-*1/*2 and *2/*2 genotypes (Hitomi et al. 2004) (Fig. 2). Thus, CD72 polymorphisms were shown to affect the risk of FCGR2B polymorphisms for SLE, although CD72 per se had only a slight effect, if any, on the development of SLE. It is interesting to note that FCG2B-232Thr/Thr also showed a preferential association in the subgroup of SLE with nephritis (Chu et al. 2004), suggesting that genetic interaction with CD72 may play a role.

Fig. 2
figure 2

Genetic interaction between FCGR2B and CD72 polymorphisms against the development of SLE. Risk of SLE conferred by FCGR2B-232Thr/Thr genotype only reached statistical significance in individuals with CD72-*1/*1 (P=0.009), not in individuals with CD72-*1/*2 and *2/*2 genotypes. The odds ratios for the FCGR2B genotype alone were 2.38 (232Thr/Thr) and 1.10 (232Ile/Thr) versus 232Ile/Ile, and those for CD72 alone were 1.07 (*1/*1), 1.18 (*1/*2) versus *2/*2. The data were derived from Japanese and Thai association studies combined via meta-analysis (Hitomi et al. 2004)

Full size image

In view of the two polymorphic sites within intron 8, we then examined whether the polymorphisms influence splicing. RT-PCR of RNA derived from peripheral blood mononuclear cells using primers placed within exons 6 and 9 revealed two amplification products, one of which was found to encode a novel alternative splicing isoform caused by exon 8 skipping (AS isoform). While the termination codon appears within exon 8 in the common isoform, exon 8 is skipped in the AS isoform. This would result in the translation of exon 9 sequences until the new termination codon appears, leading to the replacement of 42 C-terminal amino acids in the membrane-distal portion of the extracellular region with 49 new amino acids in the AS form. The predicted amino acid sequence maintains the C-type lectin domain, suggesting that the alternative isoform may be functional at the protein level. Indeed, we confirmed that the protein product of this isoform is expressed when the full-length cDNA is introduced into the COS-7 cells (Hitomi et al. 2004).

We then examined whether the polymorphisms affect the efficiency of alternative splicing, and this was indeed found to be the case. Thus, although both common and AS isoforms were found in each individual, the proportion of AS isoform was clearly increased in individuals with *2/*2 genotype, followed by *1/*2. We confirmed that the 13-nucleotide repeat and four-nucleotide deletion directly regulate the splicing efficiency by minigene assay (Hitomi et al. 2004).

These results imply that the CD72-*2 allele increases the relative proportion of the CD72 AS isoform, and protects its carriers from the risk of SLE conferred by the FCGR2B-232Thr/Thr genotype. The molecular mechanisms of this protection are currently being investigated.

Concluding remarks

It is surprising that our studies that focused on two genes provided potential keys to so many new fields of biological significance, such as the relationships between transmembrane polymorphism, lipid raft and signaling, as well as genomic polymorphism, alternative splicing and epistatic interaction.

Human systems are much more complicated than we could possibly imagine, and immune systems are already known to be highly polymorphic, with most genes being subject to alternative splicing (Lynch 2004). Most of our current understanding of the immune system is based on the analysis of the interaction among the major isoform of the major allele of each gene. Although such analysis provides an indispensable scheme for the molecular interaction, in order to fully understand the diversity of immune responses among human populations, and the etiology and pathogenesis of human disease, it will be necessary to study multiple isoforms and alleles of each gene and their interactions as well.