Identification of a pleiotropic effect of ADIPOQ on cardiac dysfunction and Alzheimer’s disease based on genetic evidence and health care records

Figures

• 

  Fig. 1. Identifying shared genetic architecture by repurposing scaled digital health data. Schematic workflow of a developing hypothesis in the observation phase based on digital health data (gray boxes) to study, using genomic data (white boxes). In the gray boxes, the database of HIRA and the electronic health records (EHR) of UCSF Medical Center were used to suggest the association of cardiac dysfunction (CD), such as hypertensive cardiac disease, with dementia and Alzheimer’s disease (AD/D); the public database of HIRA (Health Insurance Review and Assessment service) was used as a nationwide inpatient/outpatient diagnoses observation database in South Korea; the EHR of UCSF was used as the validation database. After presenting the propensity of CD amongst AD/D diagnoses, we examined the genetic associations between CD and AD/D in a discovery phase (i.e., white boxes).

• 

  Fig. 2. The preceding diagnosis of heart disease before dementia. A, B Traced models of disease diagnosis patterns using the directed acyclic graph (DAG) model. Our DAG model consisted of a node for the disease diagnosis and an edge for subsequent diagnoses presented that were not random (FDR < 0.1). A From the HIRA dataset, we found that 6.1% of the patients suffering from hypertensive heart and renal disease (n = 425) were diagnosed with unspecified dementia and Alzheimer’s disease (a major subtype of dementia). B We validated whether the identical pattern was repeated in the independent dataset obtained from the electronic health records (EHR) of the UCSF Medical Center. Of the 672 patients with hypertensive heart and kidney disease, 6.8% (n = 46) were diagnosed with dementia. C, D Due to the vague definition of the “hypertensive heart and kidney disease” diagnosis, we looked into a more detailed diagnosis code at the 5-digit level of ICD-10. C In the HIRA dataset, approximately 46% of dementia diagnoses made in patients who already had hypertensive heart and renal diseases were identified after heart failure. Therefore, the diagnosis of dementia after the diagnosis of hypertensive…

• 

  Fig. 3. Identification of the candidates of the pleiotropic variants for CD and AD/D. A Integration of the variant–disease associations metaset from our VARIMED and the whole-exome sequencing (WES) dataset. We selected 3 variants in MTHFD1L, DPP10, and ADIPOQ for further validation. B Reconfirmation of selected germline variants using Sanger sequencing.

• 

  Fig. 4. Validation of pleiotropic effects for CD and AD/D. A–D Validation of the functional impact of the 3 identified AD/D variants. We constructed site-directed mutant Neuro-2a cell lines (MTHFD1L-M, DPP10-M, and ADIPOQ-M). We assessed the abundance of each cell line of CP12, PHF1, and tau proteins via three biological replicates. A Loading abundance of CP13, PHF1, and tau proteins, as well as β-actin in neuronal cell lines by transfected genes. ‘M’ = mutated. ‘W’ = wild type. B–D Levels of aggregation of CP13, PHF1, and tau normalized by β-actin. ADIPOQ-M displayed abnormal aggregation of tau and CP13 (D) (p < 0.05, t-test). E Analysis of an enriched pathway of 473 selected differentially expressed genes (DEGs) (FDR p < 0.05, log2 fold change >5) in ADIPOQ knockout cells, H9c2 (rat cardiac cell). Cardiac dysfunction and cognition impairment pathways were enriched in 473 DEGs (FDR p < 0.05, hypergeometric test). F–H Functional impact of the ADIPOQ variant on a population scale using the UK Biobank. We selected 69 individuals from the minor allele group of ADIPOQ (c.268G>A) and 276 from the major allele group based on the propensity score matching analysis. F The average heart wall…