Search

Nephrotic syndrome (NS) is a rare kidney disease resulting from malfunction of the primary ultra-filtration unit in the kidney, the renal glomerulus, leading to excessive leak of protein into urine. NS has an annual incidence of 2 and 7 per 100,000 children and adults, and a prevalence of 1 to 15 per 100,000 depending on the ethnicity; NS is more common in African and South Asian populations. However, despite its rare disease status, it remains one of the most common kidney diseases to affect children and adults. It has a devastating impact on the health of affected individuals, with around 20% of cases developing end stage kidney failure and 60% of the severe group experiencing disease recurrence post kidney transplant. Mendelian inheritance appears only to explain around 30% of cases. Inheritance may be autosomal recessive or dominant with variable penetrance and more recently Xlinked has also been described. To date, causal genetic variants have been identified in 67 genes in NS patients. However, the molecular genetic mechanisms underlying the remaining 70% remain poorly understood and are likely to fall into a complex genetic category. The aim of this study was to identify causal genetic variation of NS, focusing on the 70% of cases currently unexplained by single mutations in previously established nephrotic syndrome genes. A cohort of 485 deeply phenotyped patients was available for analysis; all have undergone whole exome sequencing or whole genome sequencing. Data was analysed by applying computational approaches including linkage and association analyses. Based on this, a link with HLA was identified, confirming that despite a lack of typical inflammatory markers, NS in both children and adults falls into the category of an autoimmune disease.

With the sequencing of the genomes of individuals with rare Mendelian disease becoming routine, there is an emerging challenge in identifying and quantifying similarity between individual's phenotypes to assist in the identification of commonalities in the genetic variation contributing to disease. Whilst it is relatively easy to assess genetic similarities between individuals, it is less trivial to assess phenotypic similarity due to the complexity of phenotypic information. One route to systematically estimate similarity between phenotypes utilises computational approaches applied to standardised machine-readable phenotypic descriptors, such as those in the Human Phenotype Ontology (HPO) or structured patient questionnaires. This thesis describes advances in the representation of clinical phenotypes in machine-readable controlled vocabulary within the context of genetic studies of both the diagnosis of monogenic disease patients, and common variant association analysis of severe acne subtypes. When using genome sequencing for the genetic diagnosis of individuals with rare Mendelian diseases, a virtual gene panel approach is often taken wherein only a curated list of genes suspected to cause a phenotype are considered. With the number of known monogenic disease-gene pairs exceeding 5,000, manual curation of personalised gene panels based on the entire human phenotypic spectrum is challenging. Methods have previously been developed that formalise the approach using the patient phenotype to generate candidate genes, requiring both patients and known disorders to be defined in standardised machine-readable terms. Work in this project has investigated the ways by which established phenotypic descriptions (OMIM free-text) can be further leveraged using simple quantification of disease terms to gain a more nuanced description of known phenotypes with HPO terms, and how this helps to more efficiently generate candidate gene panels in real patient datasets. This project also examines the utility of extensive patient questionnaire records in patients with severe acne, enabling the identification of questionnaire response stratified subtypes of acne for use in downstream investigations seeking to identify new genetic determinants of the disease.

Frontal fibrosing alopecia (FFA), a clinical variant of follicular lichen planus, is a highly distressing inflammatory and scarring dermatosis of unknown pathobiology that affects almost exclusively post-menopausal women. Since FFA was first described in 1994, there has been rapid increase in reported incidence, culminating in speculation about likely environmental triggers. There is substantial evidence for an inherited component in the aetiology of FFA, as demonstrated by elevated incidence in first-degree relatives. To test the hypothesis that rare large effect size alleles underlie FFA, whole exome sequencing (WES) was utilised to identify variants shared by affected individuals within each of 12 FFA families of presumed Mendelian inheritance. To investigate the contribution of common genetic variation to disease susceptibility, a two-phase genome-wide association study (GWAS) was performed in a UK and a Spanish cohort in a total of 1,016 cases and 4,145 controls. Scalp skin RNA-seq was undertaken to define the FFA transcriptome. A case-control metabolomic analysis was performed to probe the relevance of xenobiotic processing in disease pathogenesis. Significant association with FFA was observed at genomic loci 2p22.2 (CYP1B1) and 6p21.1 (HLA-B), highlighting the role of xenobiotic and hormone metabolism and autoimmunity in the pathogenesis of FFA respectively. Transcriptomic analysis of affected skin implicated both innate and adaptive immune mechanisms. In short, this study provides insight into the genetic basis of FFA by characterising the condition as a complex immuno-inflammatory trait.

Ductal carcinoma in situ (DCIS) and lobular carcinoma in situ (LCIS) are clinically undetectable forms of non-invasive breast cancer. DCIS is considered a non-obligate precursor of invasive ductal carcinoma (IDC). LCIS shares many of the same genetic aberrations as invasive lobular breast cancer (ILC), which accounts for 10-15% of all invasive breast cancer. With the advent of screening mammography, the diagnosis of pure DCIS (with no invasive component) and LCIS has become more common, and approximately 20% of screen detected tumours are pure DCIS. The aim of this project is to test the hypothesis that breast cancer is a heterogeneous disease and that by focusing on specific histological subtypes we can increase the power to detect genetic variants that predispose to DCIS/LCIS/ILC. We also exploited the extreme phenotype hypothesis, having focused on cases with a severe phenotype such as early-onset or bilateral disease. During this PhD we assessed the role of rare coding variants using next generation sequencing approaches. We also interrogated data on 211,000 SNPs, genotyped on the iCOGS platform in 3,000 DCIS cases 2500 LCIS/ILC and 5000 controls, to evaluate common variants that predispose to these subtypes of breast cancer. Some of the key findings include the excess of CDH1 protein truncating variants in cases with bilateral lobular lesions (8%), and the identification of a novel lobular specific locus on 7q34. We were also able to estimate the prevalence of rare variants predisposing to breast cancer in the context of sporadic cases with DCIS/LCIS/ILC. Further analyses and validation is required in order to assess any of the novel putative genes can be linked with ILC development. Once such variants have been validated they can be used to predict which women are at high risk of developing DCIS/LCIS/ILC and such women can be offered intensive screening or chemoprevention.

Myoclonic astatic epilepsy (MAE) is a rare generalised childhood epilepsy with variable but poorly described neurodevelopmental outcome. Family studies suggest a major genetic influence as up to two thirds of relatives have seizures, or electroencephalographic (EEG) abnormalities. MAE is associated with 10 different genes, yet these genes account for less than 20% of the genetic aetiology of MAE leaving the majority unexplained. The aims of this thesis were (1) describe the epilepsy and neurodevelopmental phenotype of MAE cases, (2) perform EEG studies on first degree family members for familial EEG abnormalities and compare occurrence of epileptiform features to population prevalence and (3) to collect DNA and identify MAE causative genetic variants through exome sequencing. I assembled the largest MAE cohort (n=123) to date. The epilepsy phenotype is remarkably similar to previously published cohorts. I identified a severe neurodevelopmental phenotype: intellectual disability was reported in 64.9%, autism spectrum disorder in 21.3% and attention deficit hyperactivity symptoms in 41.0%. Additionally, extremely low adaptive behavioural scores were identified in 69.4% of cases. I performed EEG studies on 38 first-degree relatives of 13 MAE families, and found an excess of epileptiform EEG features in adults (>16 years), compared to controls (P=0.05, RR 6.82). I identified likely pathogenic or candidate variants in 11 of 109 cases. This comprised known genes associated with MAE: CHD2 n=1, SYNGAP1 n=2, SLC6A1 n=1, KIAA2022 n=1; epilepsy associated genes novel for MAE: KCNB1 n=1, MECP2 n=1, KCNH5 n=1, and three new candidate genes; SMARCA2 n=1, ASH1L n=1 and CHD4 n=1. Lastly, I highlight phenotypic features which help correlate with known and novel specific gene associations, discuss that MAE is a phenotypic and genetic nosological bridge between genetic generalised epilepsy and epileptic encephalopathy, and discussion applications and future directions leading on from this project.

Next generation sequencing generates a large quantity of sequence data which has the potential to be highly informative when evaluated using appropriate analytical methods. One of the key aims of human genetic disease studies is to use such methods to help identify sequence variants having some phenotypic effect. In the past few years, whole exome sequencing in particular has been used to identify single variants that cause many monogenic diseases. However, monogenic diseases in which genetic heterogeneity plays a role present a more difficult problem because different affected individuals in a study may not carry disease-causing mutations in the same gene. A major focus of my work is to develop and implement algorithms to identify disease-causing variants in such diseases. In particular I make use of functional information, such as that encoded by interaction networks, to prioritise genes for follow-up analysis. In this thesis I present two different analysis tools designed for this purpose. Simulated datasets are constructed to demonstrate the utility of these tools and test their performance under varying conditions. The tools are applied to a whole exome sequencing study for a genetically-heterogeneous monogenic disease (Adams-Oliver syndrome) with the aim of generating novel hypotheses regarding disease aetiology. This work also allows comparison and exploration of the challenges facing network-based methods in practice. The tools are also applied to a study of families exhibiting atypically strong recurrence of a complex disorder (Crohn’s disease), testing the hypothesis that one or a small number of rare highly-penetrant variants might be implicated in each family. In this way it is proposed that the application of network-based methods to next generation sequencing data can help to describe disease mechanisms that move beyond monogenic diseases and towards more complex genetic architectures.