Our laboratory works in human genetics, molecular epidemiology and functional genomics, with a specific focus on the role of obesity and insulin resistance in development of cardiovascular disease (CVD). We combine big data analytics in large population-based studies with functional work in model systems. Specifically, we try disentangle the complex relationship between obesity, insulin resistance and CVD using modern -omics technologies combined with functional characterization of candidate genes using CRISPR-Cas9 gene editing. Our research is translational; bridging epidemiology, molecular biology and clinical medicine to gain new insights into the pathophysiology of CVD, identify biomarkers that improve risk prediction, and discover novel drug targets.
We are planning many population-based projects in the UK Biobank, which is an excellent example of science in the new era of open science initiatives and big data analytics. In 2006-2010, the UK Biobank recruited 502,650 participants aged 37-73 years to undergo physical measurements, detailed assessments about risk factors and future disease events, and sampling of blood, urine and saliva. Genome-wide genotyping on the UK Biobank Axiom Array and imputation to ~80 million variants has been performed in all participants. They have also been extensively examined, including anthropometry, whole-body bioimpedance measures, blood pressure, heart rate, hand grip strength, spirometry, cognitive testing, bone mineral density, arterial stiffness, hearing test, eye exam, cardiorespiratory fitness test, a biomarker array, and imaging (MRI brain, heart, abdomen and a DEXA scan). Outcome events are captured via registries, and there is an expert adjudication group who reviews all register-reported cases of specified conditions, including CVD.
We are working on a wide range of projects using this excellent cohort, which has served as model for the NIH Precision Medicine Initiative. These projects include traditional epidemiological studies and GWAS addressing important, but understudied conditions, such as peripheral vascular disease, heart failure and infectious diseases – including risk prediction studies to improve patient stratification, as well as studies of environmental risk factors, genetic determinants and their interactions; but also more novel approaches which aims at addressing causality of risk factors and biomarkers and importantly, at finding druggable targets using genomic approaches. The statistical power, as well as the opportunities to study new research questions, are unprecedented given the very large sample size (ten- to hundred-fold larger than all previous studies) and the extreme richness of the data.
For studies of causality, we use Mendelian randomization (MR) - a method using genetic variants as robust proxies for an environmentally modifiable exposure to study its causal relationship with health outcomes. This method has become hugely popular recently, and we are amongst the leaders in this field. We are working with a whole range of projects addressing clinically important questions, such as salt intake, periodontitis, sleep duration, migraine and coffee intake in relation to CVD and T2D-related outcomes. As an illustration of these methods, we analyzed the associations of BMI and coffee intake with systolic blood pressure (SBP). While traditional observational analyses indicated (as previous epidemiological studies) that higher BMI and coffee intake was associated with higher and lower SBP, respectively; the MR analyses showed that it was only the association of BMI with SBP that was causal, while there was no hint of a causal association of coffee intake with SBP. These kinds of analyses are hugely important for public health, and we have a range of projects with important clinical, biological and/or public health questions lined up within this framework. In fact, my second R01 is focusing on the potential causal role of 36 circulating biomarkers representing different biological systems for development of CVD and T2D. These biomarkers include amongst others vitamin D, IGF-1, Apo-AI, urate, SHBG, estrogen and testosterone – all of which are being debated as to whether they are causally related to disease, and hence if their perturbation should be part of a preventive strategy or not.
In another exciting project, we are investigating the phenome-wide characteristics of individuals carrying gene-disrupting alleles, and hence characterize phenotypic effects as a function of number of functional copies of specific genes. Using this "human knockout model" in the 502,650 participants of the UK Biobank, we can simulate and predict what the result would be if blocking the corresponding protein. Combining this information with data from drug target databases, we can predict effects of perturbing drug target genes. Such increased knowledge about downstream effects of naturally occurring null mutations in druggable genes will provide important insights into disease mechanisms, predict potential for repurposing drugs, and unknown side effects. Our investigations will further characterize the functional impact of gene variation while bypassing the inherent translational uncertainty of model systems.
In addition to UK Biobank, I am also working with several other datasets. I am still the PI for a range of –omics projects in several Swedish cohorts – ULSAM, PIVUS, TwinGene and EpiHealth. These include genomics, transcriptomics, epigenomics, proteomics and metabolomics, often used in combination - aiming at increasing the biological knowledge of obesity, insulin resistance and CVD, and to identify new biomarkers for risk prediction and novel drug targets.
To further characterize gene function after various –omics studies and use of in silico data on gene regulation and transcription from public resources, we proceed to studies of gene function in model systems. We use CRISPR-Cas9 techniques for gene editing in human SGBS adipocytes, HepG2 hepatocytes, HMCL-7304 skeletal myocytes, and murine 3T3-L1 adipocytes to study phenotypes related to obesity and insulin resistance. We transfect cells using our custom-built lentivirus CRISPR-Cas9 constructs, and assess the effect of knockdown or overexpression of candidate genes on basal and insulin-stimulated glucose uptake (using 14C-labeled deoxyglucose) and lipolysis (measuring glycerol after insulin and isoprenaline exposure), as well as insulin signaling proteins and adipogenesis. We address downstream effects of gene knockdown using transcriptomic and metabolomic profiling on cell lysates. Although this is a new line of research for our group since the past 2-3 years, and we have relocated from Uppsala to Stanford in this period, I have already got a R01 to do this line of work, and we have several striking results that we are currently writing up as manuscripts. For example, by combining our GWAS on waist-hip ratio (Shungin, Nature 2015) with data from CAGE-seq in adipocytes during adipogenesis, we identified a regulatory region in 7p15 that we have characterized in detail. We have shown that the GWAS variant has an allele-specific effect on the enhancer, that it is very strong enhancer that responds to insulin and in a range of cell types using a luciferase reporter assay. We have proceeded to knock out the enhancer followed by RNA sequencing and qPCR to pinpoint the culprit gene by which the GWAS variant is linked to adipogenesis.
Another example of how we take findings from our large GWAS forward to functional studies is a locus in 4q22 which is associated with fasting insulin (Scott, Nat Genet 2012). Using expression data from subcutaneous fat, we found a strong eQTL, and proceeded with CRISPR-Cas9 knockout of the likely culprit gene. Our characterization of adipocytes and hepatocytes shows that knocking this gene increase glucose uptake in fat and liver cells, as well as increases glycogen content in liver cells. Our results indicates involvement of the insulin signaling pathway, and we are currently breeding knockout mice to further investigate the function of this gene that seems central in development of insulin resistance.