Using Large Datasets to Understand CKD

Genome-wide association studies (GWAS) have been instrumental in identifying genetic variants that can influence a particular phenotype or disease. In 2009, using a large European population, this powerful technique was applied to CKD, and it was successful in identifying several polymorphisms that had significant correlation with CKD.¹ When a single-nucleotide polymorphism (SNP) is identified by GWAS, it is important to note that the relationship with the phenotype is often correlative but not causative. Additional information is needed to understand the basis for the relationship. In that initial study, uromodulin variants had the strongest correlation with CKD, which was not unexpected, because uromodulin was already known to be important in kidney function.

The identification of SHROOM3 as a candidate was much more surprising, because only scant expression data supported its involvement in the kidney. Shroom3, originally simply named Shroom, was originally discovered in a gene trap experiment in mice that identified it as a gene essential for neural tube closure.² This phenotype is the source of the unusual name, because the brain “mushrooms” out of the skull. The severe exencephaly in those mice also makes it difficult to study, because homozygous mutant mice die at birth. In the neural tube, Shroom3 is needed to drive apical constriction in epithelial cells of the neural plate, helping the epithelial sheet curl into a tube. Shroom3 drives apical constriction of cells by binding both apical actin filaments and Rho-associated protein kinase (ROCK) at the apical end of the cell. When brought into close proximity to the actin, ROCK then phosphorylates nonmuscle myosin II, causing contraction of the actin filaments. Two Apx/Shrm-specific domains, termed ASD1 and ASD2, define the small family of Shroom proteins and mediate the binding to actin and ROCK, respectively.³ Shroom3 can also drive elongation of epithelial cells through a much less understood microtubule-based mechanism, and its presence in many adult tissues suggests that additional roles in regulating cytoskeletal dynamics remain to be discovered.

However, the known role for Shroom3 in apical constriction provides little aid in understanding its role in the kidney. In addition, the SNP initially correlated with CKD (rs17319721) is located in an intron, typical for most SNP genotyping technology, complicating our understanding of the GWAS result. However, since that original finding, the validity of SHROOM3’s role in CKD has been supported in two ways. Multiple subsequent GWAS consistently identify SHROOM3 variants correlating with different aspects of kidney function and disease.⁴ Most importantly, studies on animal models showed that Shroom3 plays important roles in the development of the kidney and kidney function, including a role in the podocyte that can potentially explain the link to CKD.^5–7

The studies of Shroom3 in animal models provide important impetus for a more detailed examination of SHROOM3 variation in humans. In this issue of the Journal of the American Society of Nephrology, Prokop et al.⁸ have started to clarify many of the questions concerning human SHROOM3 and how it relates to CKD. The study makes extensive use of a large database of human whole-genome and exon sequences that represent over 100,000 individuals. In addition, there is extensive RNA sequencing data from multiple tissues and a detailed analysis of epigenetic marks. One straightforward but important achievement is the characterization of many novel SHROOM3 variants that have the potential to impair function as well as the assessment of the allele frequency of those variants in different populations. The group has previously shown that one of the variants (P1244L) has impaired function in their zebrafish functional assay.⁵ In this study, sequence comparison with over 100 species helped identify a 14-3-3 binding site and a predicted substrate site for phosphorylation by LATS 1/2 kinase in close proximity to P1244L. An in vitro kinase assay showed that both the wild type and the variant could indeed be phosphorylated by LATS 1/2, but crystal structures of the variant and the wild type followed by molecular simulations suggest that the 14-3-3 binding would be impaired in the variant.

When Shroom3 was initially characterized in the mouse, two isoforms were identified: a full-length one and a shorter isoform that lacks a PDZ domain found near the amino terminus.² This study shows that the same isoforms are found in humans and provides RNA sequencing data that indicate that the full-length isoform is predominant in neural tissue, whereas other tissues, including the kidney, predominantly express the short isoform. Another short isoform is also identified, although this encodes the same protein, and the start methionine of the two short isoforms is at amino acid 177 of the long isoform. In the zebrafish functional assay, the short isoform has the same activity as the full-length protein. In other studies, assays for the ability to drive apical constriction in cells established that the PDZ domain is also not required for driving apical constriction.³ An additional shorter isoform was identified only in podocytes and is predicted to be missing the PDZ and ASD1 domain.

The short isoforms also provide a probable explanation for why the SNP rs17319721 is associated with CKD. The associated allele of this SNP results in increased transcription of the short isoform and is found between binding sites for the transcription factors FOXO1 and TCF7L2. Interestingly, TCF7L2 has also been linked to CKD.⁹ Analysis of publicly available data on chromatin looping contacts predicts that rs17319721 directly interacts with the transcriptional start site of the short isoforms. Taken together, these data suggest that rs17319721 is a functional SNP that is important for transcriptional regulation of the short isoform in the kidney.

This study is a remarkable example of how large datasets can be harvested to generate insights into a candidate gene for a given disease (Figure 1). In applying this bioinformatics pipeline to SHROOM3, the analysis revealed an explanation for the function of this intronic SNP that is correlated with CKD. It identifies potential upstream regulators and potential interactions with two well established signaling systems. Finally, it also provides an extensive list of potential human variants with impaired function. Much remains to be done to now experimentally verify the many insights that this study revealed. Furthermore, there are a number of steps necessary to determine if the sequence information has clinical benefit. Even SNPs with a significant correlation with CKD can have little predictive value at the level of the individual.¹⁰ The main value for studies such as the one in this issue will be to develop insight into the molecular mechanisms that are linked to kidney health and hopefully, generate novel targets for intervention. Finally, the bioinformatics pipeline shown here can serve as a model for any candidate gene associated with a particular phenotype can be put through this bioinformatics pipeline in order to generate novel hypotheses about SNPs within the candidate gene, gene function, and how the gene is regulated.

Figure 1.

Large datasets can be fed into a bioinformatics pipeline to generate insights into the biology of a candidate gene. This study shows an integrated analysis of very large datasets of genomic data. In this case, the candidate gene was SHROOM3, but any gene could be analyzed in this manner. The data generated uncovered novel genetic variants, important insights into known variants, and potential novel interactions at the genomic and protein level for SHROOM3. ChIP-Seq, Chromatin Immunoprecipitation followed by next generation sequencing; HiChIP, Protein-Centered Chromosome Conformation Capture; RNA-seq, RNA sequencing; SNP, single-nucleotide polymorphism.

Disclosures

None.