This study focused on understanding the genetics behind developmental TDS, which is typically associated with male infertility and TGCC, sometimes combined with mild forms of genital malformations. Some studies suggest that a common genetic background for TDS is doubtful,1
and others have even implied that TDS does not exist.34
We believe that TDS is a real syndrome with predominantly environment-dependent pathogenesis and a relatively weak component of genetic predisposition. This means that the prevalence of TDS among Caucasian populations probably depends on the strength of the adverse environmental or lifestyle factors. However, as this study suggests, it is likely that a subset of cases of TDS do share predisposing genetic variants. Here, we identified, using a combination of GWAS and systems biology approaches, genomic variants in TGFBR3
between several TDS phenotypes, most notably cryptorchidism and testicular cancer. TGFBR3
has not previously been reported to be directly associated with male infertility or TDS, but participates in pathways that are highly relevant in terms of involvement in regulation of embryogenesis and oncogenesis. In the single-marker approach, we also found evidence in the discovery cohort for a possible association for rs17198432 and other markers in the HOXD
It should be mentioned that SNP markers found to be significant in GWASs are more often than not in intergenic regions and not in coding or known regulatory regions. Picking the closest gene as the affected one is an assumption; however, cis-eQTL studies35
have shown that it is a reasonable assumption. One of the strengths of our integrative approach is that it takes a gene focus that increases the probability of the gene being involved, as it requires the gene to have relevant expression, as well as protein–protein interactions of relevance to the phenotype.
The integrative systems biology analysis, which combined data from the GWAS with gene–phenotype associations found in complementary data types, including testis developmental expression data and protein–protein complexes associated with testicular developmental defects, identified several disparate loci that are functionally linked to the TGFβ superfamily signalling pathway, which were supported by the replication study. Using GWAS alone, these SNPs would fall into a ‘grey zone’,18
and would not have been selected for replication. Many GWASs suffer from this phenomenon because the number of samples required to attain genome-wide significance can be unfeasibly high.
Among the loci with the highest significance across all four TDS phenotypes was the TGFBR3
gene, which encodes the TGFβ receptor type III, a co-receptor for inhibins and TGFβ1–3, BMP2, BMP4 and BMP7. TGFBR3 and its co-receptors and ligands are expressed in most endocrine tissues, including the testis. TGFBR3 has been identified in human Sertoli cells and Leydig cells, both in the normal testis and in tubules with CIS.36
In this study, we found that TGFBR3 was expressed in both Leydig and peritubular cells in fetal as well as adult testis, but was absent from Sertoli cells. This expression pattern supports the idea that TGFBR3 is essential for embryonic development of the reproductive system, as silencing of the murine Tgfbr3
gene resulted in impaired function of fetal Leydig cells and testicular dysgenesis.37
In addition, the signalling partners of TGFBR3, the family of activin receptors, are present in fetal human testis38
and appear to be dysregulated in testes with TGCC.33
Of note, among the loci that scored high in both the pathway analysis and the integration analysis were BMP7 and other bone morphogenetic proteins, which can bind TGFBR3. The presence of several components of the same pathway, which have all previously been implicated in the development of the testis and early reproductive system, reinforces the validity of the identified SNPs as possible genetic factors predisposing to TDS.
The analysis of the TGCC subset of cases confirmed the association of the KITLG
locus (12q21) with increased risk of testicular cancer.13
However, it has been suggested that SNPs in the KIT
genes may also be involved in infertility,40
but we did not detect an association for other subtypes of TDS. Thus, the KITLG
region should be further studied to pinpoint causative variants within KITLG
or its regulatory regions. In the single-marker approach, we also found evidence for a possible association for rs17198432 and other markers in the HOXD
gene cluster (supplementary table 10 online). HOXD
genes encode a group of transcription factors well known for their importance in morphogenesis during early embryonic development of limbs and genitalia,41
and a direct association with cryptorchidism has been shown in one study.42
was not supported in the replication cohort, we have not discussed this in detail, but we mention it as a variant locus that would be worth screening for in other cohorts.
Here, we have successfully applied systems biology methodologies to prioritise important findings in a GWAS of a carefully collected discovery cohort of individuals of Danish ethnic background with detailed clinical records, and a subsequent replication cohort with similar phenotypes. The GWAS data were complemented by integration of multiple data types. This approach enabled us to identify potentially significant variations that would normally not be prioritised on the basis of genome-wide multiple testing corrected p values. Two of the three replicated loci were selected on the basis of systems biology, indicating that integrative analyses can be useful for candidate selection of markers that do not reach genome-wide significance in a single-marker association analysis. The candidates identified in this study, notably those involved in TGFβ superfamily signalling pathways, provide evidence that at least a subset of TDS cases may have a common genetic predisposition. The paucity of informative markers indicates the predominant role of environmental and lifestyle factors in the pathogenesis of this syndrome. However, association with the TGFβ signalling pathway has not been shown previously, and this should be examined further. The findings should be corroborated in further independent GWASs, fine-mapping and sequencing studies, or by mechanistic investigations of the implicated pathways. In any case, integration of prevailing information via systems biology approaches holds promise for enhancing future GWASs.