Association of primary open-angle glaucoma with mitochondrial variants and haplogroups common in African Americans

Abstract

Purpose: To estimate the population frequencies of all common mitochondrial variants and ancestral haplogroups among 1,999 subjects recruited for the Primary Open-Angle African American Glaucoma Genetics (POAAGG) Study, including 1,217 primary open-angle glaucoma (POAG) cases and 782 controls, and to identify ancestral subpopulations and mitochondrial mutations as potential risk factors for POAG susceptibility.

Methods: Subject classification by characteristic glaucomatous optic nerve findings and corresponding visual field defects, as defined by enrolling glaucoma specialists, stereo disc photography, phlebotomy, extraction of total DNA from peripheral blood or saliva, DNA quantification and normalization, PCR amplification of whole mitochondrial genomes, Ion Torrent deep semiconductor DNA sequencing on DNA pools ("Pool-seq"), Sanger sequencing of 3,479 individual mitochondrial DNAs, and bioinformatic analysis.

Results: The distribution of common African haplogroups within the POAAGG study population was broadly similar to prior surveys of African Americans. However, the POAG case population was found to be enriched in L1c2 haplogroups, which are defined in part by missense mutations m.6150G>A (Val83Ile, odds ratio [OR] 1.8, p=0.01), m.6253C>T (Met117Thr, rs200165736, OR 1.6, p=0.04), and m.6480G>A (Val193Ile, rs199476128, OR 4.6, p=0.04) in the cytochrome c oxidase subunit 1 (MT-CO1) gene and by a variant, m.2220A>G (OR 2.0, p=0.01), in MT-RNR2, which encodes the mitochondrial ribosomal 16s RNA gene. L2 haplogroups were predicted to be overrepresented in the POAG case population by Pool-seq, and the difference was confirmed to be significant with Sanger sequencing, that targeted the L2-associated variants m.2416T>C (rs28358580, OR 1.2, p=0.02) and m.2332C>T (OR 1.2, p=.02) in MT-RNR2. Another variant within MT-RNR2, m.3010G>A (rs3928306), previously implicated in sensitivity to the optic neuropathy-associated antibiotic linezolid, and arising on D4 and J1 lineages, associated with Leber hereditary optic neuropathy (LHON) severity, was confirmed to be common (>5%) but was not significantly enriched in the POAG cases. Two variants linked to the composition of the gut microbiome, m.15784T>C (rs527236194, haplogroup L2a1) and m.16390G>A (rs41378955, L2 haplogroups), were also enriched in the case DNA pools.

Conclusions: These results implicate African mtDNA haplogroups L1c2, L1c2b, and L2 as risk factors for POAG. Approximately one in four African Americans have these mitochondrial ancestries, which may contribute to their elevated glaucoma risk. These haplogroups are defined in part by ancestral variants in the MT-RNR2 and/or MT-CO1 genes, several of which have prior disease associations, such as MT-CO1 missense variants that have been implicated in prostate cancer.

Figure 1

Complete coverage of mtDNA was obtained by Pool-seq. A representative IGV coverage map from one of the 43 DNA pools is shown, with the positions of the major mitochondrial genes, and numbering corresponding to the rCRS mtDNA reference sequence. Positions with a variance frequency higher than 3% are indicated by colored bars to indicate the base composition of the variable reads (A=green, C=blue, G=brown, T=red), with coverage depth indicated by the height of each bar.

Figure 2

Pool-seq identified hundreds of variable positions on mtDNA, mostly associated with one or more African haplogroups. The mean variance frequency at each of the 16,549 positions on mtDNA is plotted as a percentage for all 43 pools, representing a total of 1,999 individuals (A). Variant positions that are associated the most common African-American mitochondrial macro haplogroups are indicated with colored markers. Uncolored circles represent variants associated with multiple haplogroups, or subgroups of the major groups listed in the legend. Panel B compares individual Sanger sequencing results on 1,999 individuals for a 605 bp region within the MT-CO1 gene to Pool-seq results on the same group. The location and population frequencies from Sanger sequencing (red solid diamonds) are plotted with those inferred by Pool-seq (gray open circles), and numeric labels indicate the position on the mtDNA reference sequence. “FP” denotes two of the positions, 6420 and 6442, that were categorized as noisy and excluded from analysis as false positives.

Figure 3

The distribution of major mtDNA haplogroups inferred for the POAAGG study population is consistent with African-American ancestry. For POAAGG subjects, the frequency of each haplogroup was estimated by the mean variance frequency from a set of associated positions on mtDNA by Pool-seq. The estimates for POAG cases (orange bars) and controls (green bars) are compared to those from an independent multicenter study of African-American haplogroups (gray bars). Error bars represent 95% confidence intervals.

Figure 4

Pool-seq on POAG cases versus controls implicated the L1c2 and L2 branches of the mitochondrial phylogenetic tree. The relationships among haplogroups detected in the POAAGG study population are represented in schematic form, with branch lengths representing approximate divergence times from the most recent common ancestor (MRCA), with branching order adapted from Behar et al. and Schlebush et al. [9,53] Colored shading indicates the degree of imbalance observed in the POAG versus case pools, expressed as odds ratios (OR) from estimated population frequencies for one or more variants associated with that lineage. Mitochondrial ancestry implicated in higher POAG risk (case: control OR >1.4) is indicated by red shading, moderate risk is indicated by blue, and mitochondrial ancestry that may be protective is indicated by green. All depicted haplogroups are African with the exception of branches R, M and N, which are ancestral to Asians and European populations. Haplogroup L1c2, is defined in part by the disease-associated missense variants m.6150G>A and 6253T>C in MT-CO1, with subgroups L1c2b1a'b also having the m.2220T>C variant in MT-RNR2. The L2 lineages are associated with variant m.2416T>C in MT-RNR2 and other variants. The m.3010G>A variant in MT-RNR2, linked to linezolid sensitivity, has arisen on branch L2a1c, and a recent European haplogroup, H65a (not shown).

Figure 5

Association of 381 variable positions on mtDNA with POAG in African-Americans. The number of individuals estimated to differ from the rCRS reference sequence in the POAG case pool (n=1,217) was compared to the number of estimated variant individuals in the control pool (n=782) for each variable position, the significance of the predicted difference was assessed using a Chi-square test, and the negative log₁₀ of the resulting p value is plotted. Variable positions annotated as having a disease association in MITOMAP or as “pathogenic” or “likely pathogenic” by the NCBI variant viewer are indicated by markers with solid red fill. Selected markers are identified by their coordinates on the rCRS reference sequence.