Using genetics to test the disease consequences of higher adiposity uncoupled from its adverse metabolic effects

Obesity is associated with many diseases in addition to type 2 diabetes and heart disease. People who are carrying too much weight are also at greater risk of diseases of the joints (e.g. osteoarthritis) , gut and intestinal tract (e.g. gastro-oesophageal reflux disease, diverticular disease), diseases of mental health (e.g. depression) and some cancers (e.g. post-menopausal breast cancer). For many of these conditions, we do not know if being overweight is a cause or is simply correlated due to shared factors such as ageing and poor general health. Even if we do know that being overweight is a direct cause, we do not know which aspect of the excess fat mass is important. For example, for a disease of the joints, such as osteoarthritis, the excess weight itself could be more important or the inflammation that normally accompanies excess weight could be more important, in causing the disease.

In this proposal, we aim to understand the causal effects of two separate aspects of being overweight on common diseases - the excess fat, and the adverse consequences that normally accompany excess fat. These adverse consequences include reduced sensitivity to the beneficial effects of insulin, higher levels of fat and cholesterol in the blood, and inflammation. We will use a novel genetic approach where we characterise two sets of genetic variants - those associated with "normal fat" and those associated with "favourable fat". The "normal fat" genetic variants are associated with more fat and higher risk of metabolic diseases such as type 2 diabetes. The "favourable fat" variants are associated with more fat but lower risk of metabolic diseases. We will then compare the effects of these two sets of genetic variants on non-metabolic diseases using very large datasets such as the UK Biobank and other large collections of 10,000s of people with and without diseases. If both the normal and favourable fat genetic variants are associated with risk of a disease such as osteoarthritis, it will indicate the excess weight is important. In contrast, if only the "normal fat" genetic variants are associated, it will indicate the adverse metabolic consequences are more important. Diseases we will study include osteoarthritis, gastro-oesophageal reflux, depression, osteoporosis, breast cancer and psoriasis.

Understanding the mechanism by which excess weight causes conditions such as osteoarthritis and reflux disease is of fundamental importance to developing effective interventions and treatments. For example if excess weight is shown to cause a condition purely through the adverse metabolic effects, this raises the possibility of treatments that modify the metabolic effects, such as inflammation or insulin sensitivity. If effects are purely driven by the excess weight itself, these approaches will be futile.

In this proposal we will improve understanding of the causal effects of two separate aspects of higher adiposity on common non-metabolic diseases - the excess weight, and the adverse metabolic consequences that normally accompany excess weight. We will use a novel genetic approach to separate the two components. In stage 1, we will use multivariate statistical approaches to identify additional "favourable adiposity" alleles using genotype data and adiposity related phenotype data from 100,000s of people. These alleles are those associated with higher adiposity but lower metabolic-disease risk. We will build a "favourable adiposity" genetic score from these alleles. We will also update the metabolically unfavourable "normal adiposity" genetic score. In stage 2, we will use abdominal MRI imaging data from four studies to understand the detailed fat distribution phenotypes of the "favourable adiposity" genetic score. These phenotypes will include subcutaneous, visceral, liver and pancreatic adipose and we will also assess lean mass. We will confirm associations with measures of insulin sensitivity. We will also use data from children to test whether or not the "favourable adiposity" genetic score is a proxy for favourable adiposity from early in life. In stage 3 we will characterise the non-metabolic diseases associated with higher adiposity in the UK Biobank. We anticipate taking forward approximately 20 diseases that are not clearly known consequences of the adverse metabolic effects of obesity. We will prioritise those diseases that have a sufficiently strong association with adiposity and with a large enough numbers of cases to have good statistical power for stage 4. In stage 4 we will use the metabolically "favourable adiposity" and "normal adiposity" genetic scores in a Mendelian Randomisation approach to test separately the effects of excess weight and the adverse metabolic effects on non-metabolic diseases.

1. Who will benefit from this research?
In the short to medium term researchers in academia, clinical trials and the pharmaceutical industry will benefit from this research. In the long term, we anticipate any improvement in knowledge of the biology of obesity related diseases could have a large economic and social impact because obesity is so common. Sadly, there appears no realistic prospect of a solution, as seen by the difficulty people have in losing weight and maintaining weight loss.

2. How will they benefit from this research?
Academics will benefit from our research through better knowledge of the causes of obesity-related diseases. Academics working in clinical trials will benefit because they will have a stronger knowledge base with which to test and follow up interventions. Especially relevant will be those trialling weight loss, anti-inflammatory and insulin sensitizing interventions. One immediate pathway to impact in clinical trials is through our colleague Prof Rob Andrews, who sits on the NHS diabetes prevention programme strategy committee and is a leader in weight loss interventions such as the ACTID trial.

The pharmaceutical industry could benefit from our research because drug targets supported by human genetic evidence are more likely to reach market. One of the known "favourable adiposity" variants provides an excellent proof of principle - a common allele in the PPARG gene encodes the target protein for the anti-diabetes class of agents, thiazolodinediones. Our long term contacts at GSK, Rob Scott (Stevenage) and Dawn Waterworth (Pennsylvania) will be our first points of contact. Rob and Dawn lead human genetic research teams at GSK with the main aim of improving drug discovery pipelines. Whilst GSK has no diabetes therapy pipeline they invest heavily in cardiovascular disease and our work will be very relevant. Dawn is a co-supervisor of an MRC CASE studentship with Prof. Frayling.

To ensure maximum impact we will disseminate our research findings, data and software through multiple avenues. We have a strong track record of early release of new findings through abstracts and biorxiv. The applicants are all regular speakers at conferences and forums related to obesity and diabetes. Prof. Bell and Dr Tyrrell and Dr Yaghootkar have all been invited to talk at the annual UK Biobank meeting. Dr Yaghootkar has a Diabetes UK fellowship that means she regularly attends events for early career researchers. For a wider audience all applicants regularly speak at café scientifique, Pint-of-Science, local patient and public involvement groups, and 6th form colleges.

The dissemination of our research goes beyond normal publication and speaking engagements. We recognise the importance of making data and software publically available. We make all summary stats from genome wide association studies available at publication via the online EBI GWAS catalogue (e.g. https://www.ebi.ac.uk/gwas/studies/GCST006675) and consortia websites (e.g. https://portals.broadinstitute.org/collaboration/giant/index.php/GIANT_consortium_data_files). We were early contributors to the Accelerating Medicines Partnership effort to create a Type 2 diabetes knowledge portal http://www.type2diabetesgenetics.org/informational/data. This portal provides a publically available forum for querying genotype and phenotype data by subset (e.g. by sex or BMI cut off). We also post software to our github page (https://github.com/t2diabetesgenes/).