Genetic variation that determines TAPBP expression levels associates with the course of malaria in an HLA allotype-dependent manner

Abstract

HLA class I (HLA-I) allotypes vary widely in their dependence on tapasin (TAPBP), an integral component of the peptide-loading complex, to present peptides on the cell surface. We identified two single-nucleotide polymorphisms that regulate TAPBP messenger RNA (mRNA) expression in Africans, rs111686073 (G/C) and rs59097151 (A/G), located in an AP-2α transcription factor binding site and a microRNA (miR)-4486 binding site, respectively. rs111686073G and rs59097151A induced significantly higher TAPBP mRNA expression relative to the alternative alleles due to higher affinity for AP-2α and abrogation of miR-4486 binding, respectively. These variants associated with lower Plasmodium falciparum parasite prevalence and lower incidence of clinical malaria specifically among individuals carrying tapasin-dependent HLA-I allotypes, presumably by augmenting peptide loading, whereas tapasin-independent allotypes associated with relative protection, regardless of imputed TAPBP mRNA expression levels. Thus, an attenuated course of malaria may occur through enhanced breadth and/or magnitude of antigen presentation, an important consideration when evaluating vaccine efficacy.

Keywords: HLA; malaria; tapasin.

Fig. 1.

rs111686073 and rs59097151 genotypes associate with TAPBP mRNA expression level in PBMCs. (A) Schematic representation of TAPBP and adjacent genes roughly to scale, with relevant SNPs and the direction of transcription indicated. (B and C) TAPBP mRNA expression level was measured by using qPCR (FRESH and SK) and microarray data (CAPRISA) in each of three (for rs111686073) or two (for rs59097151) Black South African cohorts, and for the Ugandan cohorts, mRNA expression was measured by using qPCR in PROMOTE and RNA-Seq in AFRICOS: rs111686073 in the FRESH (n = 221), Sinikithemba (n = 134), CAPRISA (n = 186), PROMOTE (n = 156), and AFRICOS (n = 30) cohorts (B); and rs59097151 in the FRESH (n = 243), Sinikithemba (n = 132), PROMOTE (n = 156), and AFRICOS (n = 30) cohorts (C). Black symbols in the combined groups indicate homozygotes. (D) Expression across the combined genotypes of rs111686073 and rs59097151 in the pooled FRESH and SK cohorts (n = 384) and in the PROMOTE cohort (n = 156). mRNA levels were converted to Z-scores to allow for the combination of datasets in the case of the FRESH and SK cohorts. The numbers of individuals with each genotype are as follows: for the FRESH and SK cohort: n_CC/GG = 18, n_CC/AG = 90, n_CG/AG = 15, n_CC/AA = 186, n_CG/AA = 68, and n_GG/AA = 7; and for the PROMOTE cohort: n_CC/GG = 5, n_CC/AG = 35, n_CG/AG = 13, n_CC/AA = 69, n_CG/AA = 29, and n_GG/AA = 5. (E) TAPBP mRNA expression within tumors from African-American individuals in the TCGA breast cancer cohort (BRCA; n = 137). Expression is presented as Z-score normalization to the grand mean mRNA expression of all TCGA tumors. The numbers of individuals with each genotype for the TCGA cohort are as follows: n_CC/GG = 4, n_CC/AG = 27, n_CG/AG = 2, n_CC/AA = 88, n_CG/AA = 15, and n_GG/AA = 1. Mean ± SEM are shown, and significance was determined by using unpaired, two-sided Mann–Whitney tests. *P < 0.05; **P < 0.01; ***P < 0.001; ****P < 0.0001. B2M, β2M.

Fig. 2.

rs111686073G induces higher luciferase expression than the C variant. The pGL3-based luciferase constructs with a 623-nt promoter sequence containing the core promoter and SNP of interest was constructed by using Gibson assembly and transfected into HeLa cells for the initial analysis of the effect of rs111686073 on TAPBP mRNA expression. To examine the effects of rs111686073 more accurately, a shorter, 157-nt promoter sequence containing the proximal promoter and either the SNP of interest or a 10-nt deletion was constructed by using Gibson assembly and transfected into MCF7 cells. Luciferase activity was measured 48 h posttransfection and normalized to the expression of Renilla and the empty pGL3-basic vector (Promega). (A) Map of the area around rs111686073 with the promoter sequences for the luciferase assay indicated (not to scale). (B) Schematics of the pGL3-based luciferase constructs, with “Pro” indicating the location of the promoter insert (i.e., TAPBP 5′ UTR) containing rs111686073 and “Pro_del” indicating the promoter with the 10-nt deletion. (C) The 623-nt construct containing the G variant shows significantly higher luciferase expression, as compared to that with the C variant. (D) The 157-nt construct containing the G variant shows significantly higher luciferase expression, as compared to the C variant, or a 10-nt deletion that deletes the AP-2α binding site. The mean ± SEM of at least three independent experiments is shown, and significance was determined by using unpaired, two-sided Student’s t tests.

Fig. 3.

AP-2α binds rs111686073G more strongly than rs111686073C. EMSAs were performed with oligonucleotides containing the two respective variants of the SNP and consensus sequences for Sp1 and AP-2α. (A) Web-tool Alibaba2 predicted Sp1 and AP-2α binding sites in the 5′ UTR of TAPBP overlapping rs111686073. The SNP variants are indicated in red, and the lines indicate the predicted binding sites. (B and C) Nuclear extract from 293T cells (B) and HeLa cells (C) bound to both the C- and G-variant oligomers (lower arrows), and shifts were observed in the presence of antibody to AP-2α (higher arrows), but not in the presence of the control IgG or antibody to Sp1. Binding of AP-2α to the TAPBP-G oligomer appears stronger than binding to the TAPBP-C oligomer. Positive control Sp1 and AP-2α consensus binding-site oligomers (Sp1 Con and AP-2α Con, respectively) show protein shifts with antibodies to Sp1 and AP-2α, respectively. (D) A cold competition assay was performed with HeLa nuclear extract using 1:0, 1:5, and 1:10 ratios of hot to cold oligomer run on a single gel. Cold G-variant oligomer was better able to compete with hot G-variant oligomer and resulted in fainter bands relative to the cold C-variant oligomer. (E) Binding of nuclear extracts from 293T and HeLa cells is abrogated when the overlapping AP-2α and AP-2γ sites are mutated (TAPBP-mut oligomer). Arrows indicate location for the band associated with AP-2α binding. (F) Nuclear extract from MCF7 cells also shows abrogation of binding to the TAPBP-mut oligomer. AP-2α binds both the TAPBP-C and TAPBP-G oligomers (lower arrow), as indicated by the shifts in the presence of antibody to AP-2α (higher arrow).

Fig. 4.

rs59097151A constructs express more luciferase than do rs59097151G constructs. pGL3-based luciferase constructs with abbreviated 3′vUTR sequences of TAPBP (1,360 bp and 448 bp) were constructed by using Gibson assembly and site-directed mutagenesis. The constructs were transfected into 293T cells, and luciferase activity was measured 48 h posttransfection. Luciferase activity was normalized to the expression of Renilla and the empty pGL3-basic vector (Promega). (A) Schematics for the various luciferase constructs and maps of the area around the SNPs (not to scale). Only variant combinations of rs59097151 and rs73410010 genotypes found in our African samples were examined. Haplotypes for the two long constructs are indicated by the connected nucleotides shown beneath the two SNPs, and the variants for the shortest construct (448 bp) by the nucleotides below the included SNP (i.e., rs59097151). (B) Nested PCR of 3′ RACE confirmed the length of the full 3′vUTR of TAPBP with RNA isolated from PBMCs of three donors. Bands indicated were positively identified as the TAPBP 3′vUTR via Sanger sequencing. (C) The 293T luciferase experiments with constructs containing the two 3′vUTR haplotypes described in A or the two variants of rs59097151 (448 bp). The haplotypes of the 1,360-bp constructs are indicated by A/A and G/G and the alleles at rs59097151 of the 448-bp constructs by A and G. The mean ± SEM of four independent experiments is shown, and significance was determined by using unpaired, two-sided Student’s t tests. *P < 0.05. n.s., not significant.

Fig. 5.

miR-4486 binds rs59097151G and results in decreased luciferase expression. (A) RNAhybrid 2.2 predictions for miRNA binding and the minimum free energy (MFE) of the duplexes are shown. The nucleotide variant of rs59097151 is indicated in bold and highlighted. (B) Expression of mature miR-4486 as a percentage of RNU48 (a small nucleolar RNA) expression in 293T cells, HeLa cells, and PBMCs from three individuals and B-cell lines derived from eight individuals. miR-4486 expression was quantitated by using the 2^-ΔΔCt method with miR-423 as the endogenous control and RNU48 as the reference for expression. In the B-cell-line category, the turquoise symbols indicate individuals homozygous for rs59097151G (n = 3), and the remaining in dark blue (n = 5) are composed of two heterozygotes and three rs59097151A homozygotes. Samples were measured by using TaqMan miR qPCR assays in triplicate. (C and D) The 448-bp TAPBP 3′ UTR luciferase constructs containing only the rs59097151 variants were transfected into 293T cells with Renilla and either an miR-4486 mimic and control miRNA mimic (C) or an miR-4486 inhibitor and control miRNA inhibitor (D), as indicated. After 2 d, cells were lysed, and luciferase activity was read and normalized to the expression of Renilla and the empty pGL3-basic vector. The pGL3-Promoter construct expression is included as a positive control. The mean ± SEM of four independent experiments is shown, and significance was determined by using unpaired, two-sided Student’s t tests. Only the G variant is susceptible to down-regulation by miR-4486 mimic (C), while there is significantly increased expression for the G variant in the presence of the miRNA inhibitor (D).

Fig. 6.

Higher TAPBP mRNA i-expression associates with lower parasite prevalence. (A) The two malaria outcomes were examined for an association with TAPBP mRNA i-expression based on rs111686073 genotype (GG/CG vs. CC) or rs59097151 genotype (AA vs. AG/GG), each as a dichotomous variable and as a combined continuous variable (continuous TAPBP mRNA i-expression) with three levels of mRNA i-expression, based on data in Fig. 1D (denoting rs111686073/rs59097151 genotypes, respectively): CC/GG and CC/AG (lowest); CG/AG and CC/AA (intermediate); and CG/AA and GG/AA (highest). (B) Associations between malaria outcomes and tapasin dependence as a dichotomous variable are shown for the combined HLA-A, -B, and -C allotypes and each locus individually. Empirically defined cutoffs (HLA-ABC: 1.7, HLA-A: 1.1, HLA-B: 1.4, and HLA-C: 0.8) were used to distinguish tapasin dependence vs. independence (Materials and Methods). Malaria incidence = number of symptomatic malaria episodes per person-year (IRR ± bounds); parasite prevalence = having at least one symptomatic or asymptomatic parasitemic visit per quarter (OR ± bounds). FDR q-values are displayed. *q < 0.05.

Fig. 7.

Higher TAPBP mRNA i-expression associates with protection against malaria outcomes among subjects with tapasin-dependent HLA-I allotypes specifically. Expression was examined as either a dichotomous variable based on rs111686073 (GG/CG vs. CC) or rs59097151 genotypes (AA vs. AG/GG) or as a continuous variable with three groups of expression levels (as described in the Fig. 6 legend). Empirically defined cutoffs (HLA-ABC: 1.7, HLA-A: 1.1, HLA-B: 1.4, and HLA-C: 0.8) were used to distinguish tapasin-dependent vs. -independent levels (Materials and Methods). Among subjects with tapasin-dependent HLA-I allotypes, high TAPBP mRNA i-expression levels associate with lower malaria incidence (A; number of symptomatic malaria episodes per person-year) and lower parasite prevalence (B; defined as having at least one symptomatic or asymptomatic parasitemic visit per quarter), compared to those with low TAPBP mRNA i-expression levels. Weaker to no effect of TAPBP mRNA i-expression levels were observed among those with tapasin-independent allotypes. See SI Appendix, Table S5 for numbers in each subgroup (total n = 833 for malaria incidence, total n = 835 for parasite prevalence). IRR: upper and lower bounds are displayed. FDR q-values are displayed. *q < 0.05; **q < 0.01; ***q < 0.001.

Fig. 8.

High TAPBP mRNA i-expression levels in individuals with tapasin-dependent HLA-I allotypes confer equal protection to that observed among all individuals with tapasin-independent allotypes. Individuals with tapasin-dependent allotypes who have either high or low TAPBP mRNA i-expression levels, as measured by rs59097151 as a dichotomous variable (AA vs. AG/GG), were compared to all individuals with tapasin-independent HLA allotypes to determine their relative effects on malaria outcomes. Empirically defined cutoffs (HLA-ABC: 1.7, HLA-A: 1.1, HLA-B: 1.4, and HLA-C: 0.8) were used to distinguish tapasin-dependent vs. independent levels (Materials and Methods). Individuals with tapasin-dependent allotypes who had high i-expression levels of TAPBP mRNA showed equivalent protection to those with tapasin-independent allotypes against malaria incidence (A) and against parasite prevalence (B). H/D, high TAPBP mRNA i-expression and tapasin-dependent allotypes; L/D, low TAPBP mRNA i-expression and tapasin-dependent allotypes; I, tapasin-independent allotypes. Malaria incidence, number of symptomatic malaria episodes per person-year; parasite prevalence, having at least one symptomatic or asymptomatic parasitemic visit per quarter. IRR: upper and lower bounds are displayed. FDR q-values are displayed. *q < 0.05; ***q < 0.001.