CR1, also known as CD35, immune adherence receptor or complement component (3b/4b) receptor 1, is a large type-I transmembrane glycoprotein. It is an important member of the family of regulators of complement activation and a crucial, multifunctional mediator of innate immunity. Its wide range of functions can predominantly be perceived as the result of its capacity to bind complement components C3b and C4b. CR1 functions as a major immune adherence molecule on erythrocytes, transporting complement-opsonized particles to liver and spleen for clearance from the blood. In other cells, CR1 binding to complement components induces phagocytosis of the opsonized complexes. Lastly, CR1 functions as an inhibitor of the complement cascade by exerting decay accelerating activity toward the C3/C5 convertases and/or cofactor activity for factor I-mediated cleavage of C3b/C4b.^{3, 4}

Different CR1 isoforms have been identified, varying in size by units of ~30kDa.^{3, 5, 6} The most common isoform, designated CR1-F (also CR1-A or CR1^*1), has a frequency of 83%, and a molecular weight of 220kDa, whereas the second most common isoform, CR1-S (CR1-B or CR1^*2) has a frequency of 15% and a molecular weight of 250kDa (Figure 1). In addition, two rare isoforms have also been described of 190kDa (CR1-F′, CR1-C or CR1^*3; frequency 1%) and 280kDa (CR1-D or CR1^*4; frequency <1%).³ The CR1 extracellular domain is entirely composed of tandem repeated units of a conserved domain of 60–70 amino acids, known as short consensus repeat (or complement control repeat or sushi domain).³ All short consensus repeats, except the last two, are organized into higher order repeat units composed of seven short consensus repeats, termed long homologous repeats (LHRs)^{3, 7} (Figure 1). Of the four LHRs (LHR-A to -D; Figure 1), all but LHR-D bind C3b and/or C4b, resulting in inhibition of complement activation through decay accelerating activity or cofactor activity.^{8, 9} CR1 is located at chromosome 1q32 in a regulators of complement activation gene cluster,³ and encompasses a variable number of 18-kb long low-copy repeats (LCRs)—LCR1 and LCR2^{10, 11} (Figure 1). Both LCR1 and LCR2 are located in the CR1-coding region, and so is LCR1′, the extra LCR1-derived copy. The presence of LCR1′ explains the extra LHR, LHR-S in the long CR1-S isoform, creating additional C3b/C4b binding sites. The rare isoforms CR1-D and CR1-F′ result from either a triplication or deletion of one LCR1 copy. The third LCR, LCR3, is located in the neighboring CR1-like gene, CR1L.

Because of the myriad of functions ascribed to CR1, the protein and its potential role in disease has been intensively studied.⁴ Although biological evidence for a direct role of CR1 in AD etiology is lacking, CR1 might be involved at several levels. For instance, it has long been established that neuroinflammation and, more specifically, activation of the complement system, has an important role in pathological processes in AD brains.^{12, 13, 14} CR1 might be involved in amyloid β (Aβ) clearance from brain as fibrillar Aβ peptides were shown to activate the complement cascade and become covalently bound to C3b,^{15, 16, 17} and CR1 induces clearance and/or phagocytosis of C3b-opsonized particles. Interestingly, it has been demonstrated that in plasma, C3b mediates adherence of Aβ peptides to erythrocyte CR1, a process impaired early in the course of AD.¹⁸ Thus, CR1 may well contribute to AD by altering clearance of Aβ in the periphery (in line with the peripheral sink hypothesis stating that Aβ can be cleared from the brain by Aβ-binding molecules in the periphery). On the other hand, as CR1 is a potent regulator of the complement system, it could also participate in the disease process by inhibiting complement activation in brain, which in some studies of AD animal models had a detrimental rather than a beneficial effect.^{19, 20}

In an association analysis using 26 SNPs across the CR1 locus, two SNPs (rs4844610 and rs1408077), we showed significant association with AD in a Flanders–Belgian cohort. Both SNPs showed high pairwise LD (r²=0.97), and thus recognized the same association signal. They were also in strong LD (r²>0.87), with the two associated SNPs in the French GWA study, suggesting that they mimic the CR1 GWA cohort signal. Analysis of rs4844610 in the French cohort showed a similar, though nonsignificant trend. A combined cohort analysis provided an OR of 1.31 (95% CI: 1.11–1.55; P=0.0015). Haplotype analysis identified significant association (OR=1.24 (95% CI: 1.02–1.51; P_adj=0.027)) with one haplotype (HapB) within a 130kb LD-block containing CR1 except for its first and last exon. In agreement with the French GWA study,² single SNP and haplotype associations were strongest in APOE 4 carriers. Because CR1 is potentially involved in clearance of fibrillar Aβ peptides from the brain after complement activation and opsonization of Aβ peptides with C3b fragments,³⁴ we investigated whether genetic variability at CR1 could affect CSF Aβ₁₋₄₂ levels in patients. We calculated a significant association not with the two AD-associated SNPs (rs4844610, rs1408077), but with four other SNPs from within the associated CR1 region (rs646817/rs1746659 and rs11803956/rs12034383, showing strong pairwise LD). CSF Aβ₁₋₄₂ levels increased with ~20% in carriers homozygous for the rare allele compared with carriers of the common allele (Figure 3). Seemingly, the increased CSF Aβ₁₋₄₂ contradicts with the decreased CSF Aβ₁₋₄₂ levels that are considered indicative of AD in dementia patients and predictive in elderly people with or without mild cognitive impairment.³⁵ However, in our study, the increased Aβ₁₋₄₂ levels were not associated with the risk haplotype HapB but with haplotypes HapC and HapD. One possible explanation might be that functional variants are located on HapC and HapD that correspond to the so-called ‘intronic or exonic density polymorphisms', which have been correlated with number of CR1 molecules present on erythrocytes.^{4, 36} In plasma, C3b mediates adherence of Aβ peptides to erythrocyte CR1, a process, which is impaired early in the disease.¹⁸ However, it is doubtful that these variants can explain the observed association as genotype analysis of rs3811831, a SNP in high LD with the erythrocyte density polymorphisms, was not associated with AD or with CSF Aβ levels (data not shown). Additional studies of CR1 association and CSF Aβ₁₋₄₂ levels in AD will be needed to analyze a potential influence of CR1 genetic variability.

Using a MAQ dosage assay, we were able to show that the LCR variability in the CR1 locus was driven nearly exclusively by variability of the coding LCR1 in CR1. Association analysis of LCR1 copy number with AD indicated that individuals carrying an extra LCR1 copy had ~30% increased risk (P_adj=0.028, OR=1.32 (95% CI: 1.03–1.69)). Presence of an extra copy of LCR1 (i.e., LCR1′) defines CR1-S (allelic frequency 15%) that carries an extra set of C3b/C4b binding sites compared with the more frequent shorter CR1-F (allelic frequency 83%) (Figure 1). Further, we showed that the LCR1 CNV was in high LD with SNPs rs4844610 and rs1408077 (r²0.76), indicating that the association with CR1 can be explained by the intragenic CNV underlying CR1-S and CR1-F and, thus, number of C3b/C4b binding sites in CR1. Moreover, the observation that analysis of rs4844610 in the replication sample only showed a trend toward association, whereas the LCR1 CNV showed significant association in both study populations, independently, further strengthens the hypothesis that the CR1 CNV likely explains the association initially observed in the GWAS. It should be noted, however, that our MAQ dosage assay could not distinguish among carriers of four LCR1 copies those homozygous for CR1-S or heterozygous CR1-D/CR1-F. Also, our analysis did not include LCR2 variability (e.g., CR1-F′). However, in accordance with the reported frequencies, the respective isoforms are rare (1.7–1.8%) in our cohort, therefore unlikely to have wrongly affected the observed CR1 CNV association.

How the CR1 CNV genetic variations affect the biological processes that eventually culminate into AD remains unclear. One explanation could be that the longer CR1-S, with its additional C3b binding site (Figure 1), increases Aβ clearance from brain either through attacking Aβ plaques directly or by functioning as an Aβ-sink in the periphery, mediated by erythrocyte CR1. The observed association with AD with CR1-S, however, contradicts this hypothesis. Further, we did not observe an association between CR1-S and CSF Aβ₁₋₄₂, although we cannot exclude that the lack of association results from the use of levels of soluble Aβ₁₋₄₂,³⁵ whereas CR1-S perhaps only affects aggregated Aβ₁₋₄₂.^{16, 17} Nevertheless, we did obtain evidence of association between other genetic CR1 variants and CSF Aβ₁₋₄₂, indicating CR1 may have a role in AD pathogenesis through an effect on Aβ. Another explanation for the association with CR1-S might lie in the enhancement of the proteins' inhibitory potential (cofactor activity) towards complement activation, through its gained C3b site (Figure 1). One of the functions of CR1 is controlling amplification of the complement cascade. By binding to C3b-opsonized particles, CR1 will function as cofactor of Factor I, which converts C3b into its inactivated form C3b. The latter is unable to sustain or amplify the complement cascade, ultimately dampening the immune response and minimizing bystander tissue damage.³⁷ Therefore, CR1-S might represent a more potent inhibitor of complement cascade amplification than CR1-F, resulting in a shorter-lasting complement reaction and reducing the levels of C3b fragments that can opsonize Aβ aggregates. Thus, individuals carrying one or two CR1-S alleles might exhibit an elevated inhibition of a potentially neuroprotective process (e.g., clearance of Aβ aggregates from the brain), which is in agreement with several observations in AD mouse models, in which inhibition of complement activation was shown to correlate with enhanced Aβ plaque deposition and loss of neuronal integrity.^{34, 38} On the other hand, complement activation can elicit a potent immune reaction in the brain, which might kill healthy bystander neurons in the process. In this light, inhibition of complement activation could have a more beneficial outcome, as was likewise observed in AD mouse models in which the classical complement cascade was inhibited.^{19, 20} However, the fact that the longer CR1-S with more binding sites that can inhibit complement activation through enhanced cofactor activity, increases risk for developing AD argues against this idea.

The authors declare no conflict of interest.

Supplementary Methods

Supplementary Figures and Tables