To elucidate the actions of rs3865444 and CD33 in AD, we began by performing immunocytochemistry to localize CD33 expression in human brain. CD33 expression was observed in microglia as discerned by morphology (A
). Microglial localization was confirmed by double labeling brain sections for CD33 and IBA-1, a microglial protein, or GFAP, an astrocytic protein (C–F
). Overall, the predominant microglial localization is consistent with the hypothesized role of CD33 as a sialic acid receptor that inhibits monocyte-lineage cell activation (Lajaunias et al., 2005
Figure 1. CD33 immunohistochemistry in human brain. CD33-immunopositive cell profiles (arrows) show morphology consistent with microglia in both AD and non-AD samples (a, b). Immunofluorescence was used to help distinguish the CD33-immunopositive cell types. c/d (more ...) CD33
is encoded by seven exons, including the alternatively spliced exons 7A and 7B (Brinkman-Van der Linden et al., 2003
). To identify common CD33
isoforms in human brain, we performed PCR amplification on human brain cDNA using primers corresponding to exon 1 and exon 7A or 7B. Sequencing 50 random clones revealed frequent isoforms lacking the 381 bp exon 2 (D2-CD33
). Because the codon reading frame is maintained in the absence of exon 2, D2-CD33
encodes a protein that is identical to CD33 but lacks the IgV domain that mediates sialic acid binding in SIGLEC family members (Perez-Oliva et al., 2011
; for review, see Varki and Angata, 2006
). Both D2-CD33
are found on the cell surface of transfected cells (Perez-Oliva et al., 2011
We initially hypothesized that rs3865444 modulates CD33 expression because this SNP is 372 bp upstream of the CD33 transcription start site. To evaluate this possibility, CD33 expression was assessed by performing qPCR with primers corresponding to sequence within exons 4 and 5. Because immunostaining suggested that CD33 localized primarily to microglia, we analyzed CD33 expression relative to the geometric mean of two microglial mRNAs, CD11b and AIF-1 (IBA-1), as well as AD status and rs3865444 genotype. CD33 expression correlated strongly with microglial mRNA expression, was increased in AD, and was modestly decreased with the AD-protective rs3865444A allele (A,B); regression analysis showed a highly significant model (adjusted r2 = 0.76) wherein CD33 expression correlated significantly with microglial gene expression (p = 3.3 × 10−16, standardized β coefficient of 0.78), AD status (p = 5.9 × 10−6, standardized β coefficient of 0.29), and rs3865444 genotype (p = 0.012, standardized β coefficient of −0.17). These results are consistent with CD33 expression in microglia, increased CD33 expression in AD, and, more modestly, decreased CD33 expression with the protective rs3865444A allele.
Figure 2. CD33 isoform expression relative to AD status. CD33 expression correlated well with microglial gene expression (a, presented as geometric mean of CD11b and AIF-1, r2=0.64), as well as each microglial mRNA individually (data not shown). An association (more ...)
Because D2-CD33 was a common splice variant, we next quantified D2-CD33 by performing qPCR with primers corresponding to the exon 1–3 junction and exon 3. D2-CD33 expression was compared with total CD33 expression; isoform-specific standard curves were analyzed in parallel with samples, allowing absolute quantitation of each isoform. D2-CD33 expression corresponded well with CD33 expression and strikingly well with rs3865444 genotype (C). When D2-CD33 expression was considered as a percentage of total CD33 expression, the association with rs3865444 genotype was readily apparent (D); this impression was confirmed by linear regression analyses that found an overall highly significant model (adjusted r2 = 0.75) with D2-CD33 strongly associated with rs3865444 genotype (p = 1.01 × 10−13, standardized β coefficient of 0.81), as well as CD33 expression (p = 2.3 × 10−10, standardized β coefficient of 0.58), and modestly decreased with AD status (p = 0.013, standardized β coefficient of −0.19). Hence, the proportion of CD33 expressed as D2-CD33 showed a dose-dependent relationship with rs3865444 allele; the percentage of CD33 expressed as D2-CD33 increased by 10.7 ± 0.8% per copy of the AD-protective rs3865444A allele. Future work will include confirmation of these mRNA changes at the protein level.
Because rs3865444 resides in the CD33 promoter, this SNP is not likely to directly modulate exon 2 splicing. We hypothesized that rs3865444 is coinherited with an SNP near or within exon 2 that modulates exon 2 splicing efficiency. To investigate this hypothesis, we sequenced CD33 from 400 bp 5′ of the transcription start site through exon 4 in four rs3865444C/C and three rs3865444A/A individuals. We found four SNPs in these samples: (1) rs3865444; (2) rs2459141 (142 bp upstream of the transcription start site); (3) rs12459419 (the fourth base of exon 2); and (4) rs2455069 (the 168th base of exon 2). Among these variations, only rs12459419 was coinherited with rs3865444 in all seven individuals, i.e., rs3865444AA individuals were also rs12459419TT, whereas rs3865444CC individuals were rs12459419CC. Subsequent rs12459419 genotyping of the samples depicted in found that rs12459419 major and minor alleles were perfectly coinherited with rs3865444 major and minor alleles. Thus, the rs12459419C/T alleles substitute perfectly for the rs3865444C/A alleles in .
To evaluate whether rs12459419 is a functional polymorphism and directly modulates exon 2 splicing efficiency, we generated CD33
minigenes for each rs12459419 allele. These minigenes included exon 1 through exon 4, along with intervening introns. The only difference in minigene sequences was the rs12459419 alleles. These minigenes were transfected into BV2 microglial cells, and expression of minigene-derived D2-CD33
was quantified. This analysis found that D2-CD33
as a percentage of total CD33
increased approximately threefold between cells transfected with rs12459419C (3.4 ± 1.4, mean ± SE; n
= 3) versus rs12459419T minigenes (10.3 ± 2.3, mean ± SE; n
= 3, p
= 0.034). These findings confirm that rs12459419 is a functional polymorphism and are consistent with the human brain results. Although the mechanism underlying rs12459419 actions is beyond the scope of this study, in silico
analysis of RNA binding proteins (Smith et al., 2006
) for the RNA sequence containing rs12459419, GGG(C/U)CUG, predicts that the splicing factor SRSF2 binds when rs12459419C is present but not when rs124594149U is present (scores of 4.1 and 1.4, respectively, threshold score for binding of 2.4). SRSF2 is widely expressed, including immune cells (Visconte et al., 2012
). In summary, the most likely interpretation of these overall results is that rs12459419 mediates the rs3865444 association with D2-CD33
expression in brain because rs12459419 is highly coinherited with rs3865444, is the only SNP in exons 1–4 that is coinherited with rs3865444, resides within exon 2, and has a plausible mechanism to modulate splicing.