1. Introducción
1.5 Patología
1.5.1 Etiopatogenia de las rupturas agudas del tendón de Aquiles
To elucidate potential mechanisms underlying these genetic associations we investigated the functional role of variants at the four significant loci. For each locus, we examined all genes lying within a 2 Mb window of the lead variant and prioritized genes based on their spatial expression patterns in six human brains (Figure 2A, Supplementary Figure S3; see Online Methods). Of particular interest were genes expressed in the white matter and in the grey matter of frontal, temporal, and central structures (Figure 2B). Next, we investigated whether the lead variants and those in linkage disequilibrium (LD) were quantitative trait loci (QTL) or predicted to be damaging by searching publicly available databases (Supplementary Table S2, see Online Methods). Finally, we screened the putative causal genes for (rare) functional variants using exome sequencing (N = 1,479) and a dedicated exome chip (N = 8,087) in subsamples of our study populations (Supplementary Table S3, see Online Methods). 3p11.1 and EPHA3
The 3p11.1 signal surrounds the EPHA3 gene and its downstream region (Figure 1B), with no other genes in close proximity of the lead variant. EPHA3 belongs to the Eph family of receptor tyrosine kinases, which is one of the five canonical families of axon guidance proteins. The
mouse homolog EphA3 mediates segregation and pathfinding of callosal axons.24 In humans,
it is expressed in grey matter structures of the brain that are connected by the anterior com- missure, including the piriform cortex (Figures 2A-B). The lead variant rs7650184 lies in the 3’-UTR and influences expression of EPHA3 in brain tissue obtained from the temporal cortex
(p = 3.3 x 10-7). Also, the missense variant rs35124509 is in high linkage disequilibrium (r2
= 0.93; D’ = 0.99). Exome analyses of the EPHA3 region supported the association of an-
other variant in the 3’-UTR (rs73139148; p = 3.3 x 10-7), but did not reveal any additional
functional variants (Supplementary Table S3). 5P15.2 AND CTNND2
For 5p15.2, the signal is located in a narrow region inside the CTNND2 gene (Figure 1C). The lead variant rs11748929 and those in LD are all intronic and have enhancer histone marks (H3K4me1, H3K27ac, and H3K9ac) almost exclusively in brain tissues (Supplemen- tary Table S2). The CTNND2 gene is also expressed in the brain (Figure 2A-B), but largely
undetected in other parts of the body.25 CTNND2 encodes a neuron-specific member arma-
dillo protein, also known as δ-Catenin or NPRAP, that is a member of the β-catenin super-
family. It was initially discovered as an interaction partner of Presenilin-1,25,26 mutations of
which cause familial Alzheimer’s Disease. Subsequently, CTNND2 has been related to various
cognitive disorders, including Cri-du-Chat syndrome,27 reading problems and mild intellectu-
al disability,28 and a variety of psychiatric disorders.29 Zebrafish knockdowns show migration
cognitive function in vivo.30 Exome sequencing of CTNND2 identified a rare, but synonymous
exonic variant (chr5:10973781; minor allele frequency (MAF) = 0.4%; p = 9.2 x 10-4).
5q23.1 and SEMA6A
Also on chromosome 5, we find the strongest signal of our genome-wide association study at 5q23.1 (Figure 1D). The lead variant rs11948331 lies close to LINC00992, a long non-cod- ing RNA that is not expressed in the brain and on which little is known. Additionally, there was an apparent second signal on 5q23.1 from a less common variant (rs148925592; minor
allele frequency = 4%; r2 = 0.002; D’ = 0.21). Therefore, we performed conditional analy-
ses in a subset of studies, and found this to be an independent signal (Supplementary Table S4). This variant has an almost two-fold larger effect and lies within another long non-coding RNA, LOC102467224, which remains uncharacterized.
The nearest coding gene to the primary signal is SEMA6A, which is 762 kb further in the prox- imal direction. SEMA6A is highly expressed in the human brain, particularly in the white matter (Figure 2A-B). Interestingly, murine Sema6a is crucial for neurons to cross the midline, and
misrouted axons of the anterior commissure are still present in adult mice mutants.31 However,
we found no evidence for QTLs nor any enhancer marks that might implicate the identified variants in regulating the expression of SEMA6A or other genes. One rare exonic nonsynony- mous variant with a predicted damaging effect was nominally significant (rs200578077, MAF
= 1.0%, p=0.046), but it was not in LD with rs11948331 (r2 = 0.002; D’ = 0.01) and thus
unlikely to explain its effect. However, given the expression pattern of SEMA6A and the phe- notype of Sema6a mutants, but lack of a direct link with our association signal, we hypothe- sized there might exist long-distance interaction between the region harboring the associated variants and the SEMA6A promotor. We first searched a curated database of genome-wide chromosome interaction and found that two long-distance interactions were described using Hi-C for both regions, but these did not include interaction with the promotor (Supplemen-
tary Table S5).32 However, since none of the available cell lines for Hi-C were derived from
the brain, we performed Targeted Chromatin Capture (T2C)33 in human neural progenitor
cells to selectively interrogate the 5q23.1 region at a high resolution (see Online Methods). We identified three topological associated domains (TADs) in a 5 Mb region surrounding the variants of interest (Figure 2D). The larger TAD contains both the SEMA6A promotor and the associated variants, indeed showing that, three-dimensionally, these genomic regions are in close proximity (Figure 2E).
18q12.3 and RIT2
The final locus spans an intergenic region on chromosome 18 that is flanked by two recom- bination hotspots (Figure 1E). Conditional analyses revealed that, independent of the lead
variant rs346205, there was another association at 18q12.3 with rs144695388 (Supple- mentary Table S4). The lead variant lies inside the long non-coding RNA LINC00907 and between the coding genes PIK3C3 and RIT2. Brain expression data shows PIK3C3 is highly expressed in the white matter (Figure 2A-B), whereas RIT2 as well as the more distal gene
SYT4 have a particularly low expression (Figure 2A). Putative functional variants in LD have
enhancer marks in the brain and some acting as expression QTLs (Supplementary Table S2), including rs346212 that influences RIT2 expression in the anterior cingulate cortex (p
= 8.6 x 10-7). RIT2 belongs to the Ras superfamily of small guanosine triphosphate binding
proteins and expressed in neurons, particularly in the substantia nigra.34 A large GWAS of
Parkinson’s disease (PD) discovered rs12456492,35 which is 660 kb distal of our lead variant,
but it was not associated with the anterior commissure (p = 0.30). Through exome analyses of all nearby genes a rare missense mutation was found in RIT2 (rs142911081; MAF = 1.5%;
p = 1.0 x 10-3) that was evolutionarily constrained and had an active transcription start site
chromatin state in all samples taken from the brain, but in none of the other tissues (Supple- mentary Table S5).