Publications

BACKGROUND: Anthracyclines such as doxorubicin are well recognized to induce dose-dependent cardiotoxicity. However, traditional imaging and blood biomarkers of myocardial injury are inadequate for early risk stratification. We aimed to discover plasma proteins that correlate with the subsequent development of anthracycline-induced systolic dysfunction in patients, with validation studies in preclinical models.

METHODS: We used an aptamer-based proteomics platform to measure over 7000 proteins in the plasma obtained at baseline (before anthracycline chemotherapy) in a prospectively enrolled cohort of patients with hematologic malignancies (n=34). The correlation between protein levels and change in left ventricular ejection fraction was assessed. Candidate biomarkers were validated in a second patient cohort treated with anthracyclines for hematologic malignancies (n=32), as well as in a mouse model of delayed doxorubicin cardiotoxicity.

RESULTS: In both discovery and validation cohorts, baseline levels of CPVL (carboxypeptidase vitellogenic-like protein) and PIGR (poly immunoglobulin receptor) measured in the plasma before anthracycline treatment correlated with a subsequent decline in left ventricular ejection fraction (PIGR: R=-0.74; P=0.0022; CPVL: R=-0.74; P=0.0022). Findings were confirmed in human plasma samples using ELISA or Western blot, and similar changes in proteins were observed in mice treated with doxorubicin. In further support of PIGR as a biomarker representing the systemic adaptive immune response, doxorubicin treatment was associated with increased circulating levels of IgA+ (immunoglobulin A) B cells in mice, which similarly correlated with the decline in left ventricular ejection fraction. The highest quartile of PIGR and CPVL levels was associated with decreased overall survival but not progression-free survival.

CONCLUSIONS: Preanthracycline levels of the plasma immunologic proteins CPVL and PIGR were associated with a subsequent decline in cardiac function, representing potentially new inflammatory biomarkers of anthracycline cardiotoxicity.

INTRODUCTION: Coronary artery disease (CAD) is a leading cause of death and disability worldwide. Although genome-wide association studies (GWAS) have identified over 300 loci associated with CAD risk, the molecular mechanisms linking these variants to disease and subclinical atherosclerosis are not fully understood.

METHODS: We performed integration of multi-ancestry CAD GWAS with transcriptomic data from the Multi-Ethnic Study of Atherosclerosis (MESA) obtained through the Trans-Omics for Precision Medicine (TOPMed) program. For integration, we applied Bayesian colocalization analysis with and without statistical fine-mapping to identify genes whose expression levels colocalize with CAD-associated loci. We further applied causal weighted gene co-expression network analysis (cWGCNA) to identify gene co-expression modules and key driver genes associated with subclinical atherosclerosis traits in MESA.

RESULTS: We identified 108 genes showing evidence of colocalization with CAD loci, including 24 shared between the two colocalization approaches and 48 novel genes not previously reported in CAD GWAS. Follow-up replication and validation analyses prioritized 5 novel ( CCDC30, ZEB1-AS1, ZPR1, PLEKHJ1 and AC018816.3 ) and 8 previously reported genes ( DHDDS, DDX59, LNPEP, DAGLA, ZKSCAN1, LIPA, OPRL1 and EIF2B2 ) with putative roles in both CAD and subclinical atherosclerosis. cWGCNA identified five gene modules significantly associated with subclinical atherosclerosis in MESA. Additionally, three key driver genes ( ATG9B, PRAM1 and ZBTB46 ) identified by cWGCNA were also identified as CAD-colocalized genes.

DISCUSSION: Our integrative analysis highlights key genetic drivers and regulatory networks underlying CAD and subclinical atherosclerosis. These findings underscore the value of incorporating statistical fine-mapping in colocalization studies and demonstrate the utility of combining colocalization with co-expression network analysis to prioritize functional genes and pathways.

Thrombosis remains a major cause of morbidity and mortality in patients with cancer. Existing risk models fail to reliably predict venous thromboembolism (VTE), underscoring the need for more accurate predictive models. In this study, we conducted a high-throughput proteomic analysis of 1105 plasma proteins in peripheral blood samples from patients with newly diagnosed lung or gastric cancer who were prospectively monitored for VTE development. Using a Bayesian probabilistic machine learning approach, we developed a predictive model incorporating 11 protein biomarkers and five clinical parameters (age, sex, history of VTE, body mass index, and hemoglobin), which outperformed the standardly used Khorana prediction score [c statistic 0.84 (0.79 to 0.90) as compared with 0.36 (0.27 to 0.45)]. Orthogonal validation in an external placebo cohort from a phase 3 trial confirmed the model's predictive power. Further investigation into the mechanistic role of CD200 receptor 1 (CD200R1), an immune checkpoint receptor known to limit leukocyte inflammatory response that contributed strongly to the model, showed that reduced concentrations in plasma correlated with higher D-dimer concentrations and thrombosis risk. CD200R1-deficient mice were characterized by features of a prothrombotic state, with elevated thrombin-antithrombin complexes, increased interleukin-17A (IL-17A), and endothelial inflammation. Administration of anti-IL-17A antibodies to CD200R1-deficient mice normalized thrombin-antithrombin complexes in vivo, and a meta-analysis of human COVID-19 studies showed reduced pulmonary thromboembolism in those on anti-IL-17A antibodies. These findings highlight the utility of plasma proteomics to improve prediction of thrombosis in patients with cancer and to identify unanticipated mechanistic insights and therapeutic targets in thrombo-inflammatory disease.

We aimed to improve understanding of the genetic architecture of type 2 diabetes and glycemic traits by leveraging whole-genome sequencing in diverse populations. Our goal was to identify novel variants, refine known loci, and link genetic signals to regulatory mechanisms through colocalization with expression quantitative trait loci. We discovered novel variants, significantly improved fine-mapping resolution, and identified 80 regulatory colocalization signals in diabetes-relevant tissues. These findings support precision medicine approaches by connecting genetic variation to functional biology in type 2 diabetes.

Bariatric surgery induces profound weight loss and improvement of obesity-associated metabolic dysfunction. Recent studies have shown that adipose tissue undergoes remodeling after weight loss, characterized by a reduction in proinflammatory immune cells, increased vascularization, and a shift in the adipocyte transcriptome, but these studies focused on time points long after surgery. We performed single nucleus RNA-seq (snRNA-seq) in subcutaneous white adipose tissue (SAT) samples from subjects with obesity undergoing bariatric surgery, collected at baseline and at one, six, and twelve months after surgery. We identify profound remodeling of SAT within the first month after surgery, characterized by a surge in lipid-associated macrophages and sharp reductions in specific populations of adipocytes, adipose stromal and progenitor cells (ASPCs), and endothelial cells. Transcriptional profiles strongly suggest that some adipocytes undergo apoptosis soon after surgery, while new adipocytes are generated by de novo differentiation. Mechanistically, the data are consistent with a model whereby coordinated early loss of a hedgehog signaling axis between endothelial cells and an anti-adipogenic population of ASPCs known as adipose regulatory cells (Aregs; ASPC PTCH2 ) enables a transient burst of adipogenesis to occur. Interestingly, very few of these early features seen in human subjects are represented in a mouse model of surgical weight loss.

BACKGROUND: Exercise unmasks limitations in multi-organ system reserve capacity characteristic of heart failure with preserved ejection fraction (HFpEF). However, the metabolic and genetic underpinnings of exercise deficits, and their cumulative contribution to HFpEF severity and prognosis, remain incompletely understood.

METHODS: We used invasive cardiopulmonary exercise testing (iCPET), metabolite profiling, and genomics to simultaneously characterize seven exercise physiologic deficits in HFpEF patients: reduced exercise stroke volume and heart rate, steep pulmonary capillary wedge pressure/cardiac output (PCWP/CO) slope, elevated pulmonary vascular resistance, pulmonary mechanical limitation to exercise, impaired peripheral oxygen extraction, and obesity-related exaggerated metabolic cost of initiating exercise. We first mapped the distribution, functional, and prognostic significance of these exercise deficits. We then applied LASSO regression to identify metabolite signatures of each exercise deficit, and measured the relation of these signatures with clinical-demographic features, cardiac magnetic resonance imaging, and incident HF in 6345 individuals in the Multi-Ethnic Study of Atherosclerosis (MESA) study with ≈20-year follow-up. Finally, we mapped deficit-implicated metabolites to tissue-specific genetic variation in ≈2M individuals with HF, and in the largest genome-wide association study (GWAS) studies of HFpEF comorbidities (obesity, renal disease, diabetes) to evaluate shared metabolic mechanisms of HFpEF pathophysiology.

RESULTS: Our iCPET HFpEF cohort (61.7±14.1 years, 54% female, BMI 30.6±6.7 kg/m2 ) exhibited a broad range of compound cardiac and extra-cardiac exercise deficits. Individuals with ≥5 exercise deficits had a nearly 4-fold higher hazard of incident cardiovascular event or mortality (HR 3.90, 95% CI 1.74-8.75, P<0.0001). The metabolite signature of exercise PCWP/CO slope conferred a HR of 1.43 per SD increment, 95% CI 1.20-1.71, P<0.001 for incident HF in MESA. Addition of all iCPET deficit metabolic signatures in a single model yielded ≈20% continuous net reclassification improvement over traditional HFpEF risk factors. Genes implicated by the exercise deficit metabolome were enriched in the HF GWAS (≈2M) and shared with obesity, renal dysfunction, and diabetes, highlighting a lifelong shared predisposition to HF (including HFpEF) and its comorbidities.

CONCLUSIONS: Organ-specific responses to exercise and their circulating metabolite signatures are strongly linked to HFpEF development and prognosis. These results offer a paradigm for parsing HFpEF subphenotypes and prioritizing metabolic mechanisms of HFpEF.

BACKGROUND: Knowledge of proteomic mechanisms explaining the link between psychosocial stress and cardiovascular disease is limited. This study aimed to (1) identify plasma proteins associated with psychosocial factors and (2) assess associational pathways between psychosocial factors, identified proteins, and incident cardiovascular disease events in a discovery cohort, JHS (Jackson Heart Study), and 2 replication cohorts, the CHS (Cardiovascular Health Study), and the MESA (Multi-Ethnic Study of Atherosclerosis).

METHODS: JHS participants from exam 1 (2000-2004) with SomaScan 1.3k platform proteomics data were included (n=2143, mean age=55.3). Depressive symptoms and perceived stress scores were measured via the 20-item Centers for Epidemiological Studies scale and an 8-item perceived stress scale adapted for the JHS, respectively. Multivariable linear regression models were used to test the association between psychosocial factors and plasma proteins, controlling for age, sex, proteomics batch, and estimated glomerular function. Meta-analyses were also performed across cohorts, using Bonferroni correction for multiple testing (P<3.782×10-5). Mediation analyses with Cox proportional hazards models were used to evaluate potential proteomic pathways in the association between psychosocial factors and coronary heart disease, heart failure, and stroke in JHS.

RESULTS: Angiopoietin-2 (𝛽=0.013, SE=0.002, P<0.001), contactin-5 (𝛽=-0.013, SE=0.002, P<0.001), growth/differentiation factor 15 or macrophage inhibitory cytokine 1 (𝛽=0.011, SE=0.002, P<0.001), neural cell adhesion molecule 120 (𝛽=-0.012, SE=0.002, P<0.001), and KYNU (kynureninase; 𝛽=0.014, SE=0.003, P<0.001) were each significantly associated with depressive symptoms, with angiopoietin-2, contactin-5, macrophage inhibitory cytokine 1, and neural cell adhesion molecule 120 replicating in CHS and MESA. Leukotriene A-4 hydrolase was associated with perceived stress (𝛽=-0.0235, SE=0.005, P<0.001). Macrophage inhibitory cytokine 1 partially accounted for the association between depressive symptoms and incident coronary heart disease in JHS (23%; P=0.0009).

CONCLUSIONS: Novel associations between psychosocial factors, plasma proteins, and cardiovascular disease were identified in JHS. Circulating proteomic profiles across 3 cardiovascular disease cohorts showed differences in protein concentrations by psychosocial measures. Future investigations should identify additional potentially targetable proteomic mechanisms by which psychosocial factors contribute to disease.

Exercise benefits numerous organ systems and tissues, however limited knowledge exists about its underlying molecular pathways. Identifying the exercise-induced biochemical changes that occur in the circulation may provide further insights into how exercise confers systemic health changes. Here, we perform large-scale plasma proteomic, metabolomic, and whole blood transcriptional profiling in sedentary human participants undergoing acute endurance exercise (EE), resistance exercise (RE), or a non-exercise control (CON) in up to 7 timepoints over a 24 hour period. We observe 7066 transcript, 189 protein, and 448 metabolite changes in response to EE or RE compared to CON. Our analyses reveal numerous shared biochemical responses between EE and RE modes, but also differences in immune cell responses, lipid metabolism, and pathways reflective of tissue repair and angiogenesis. Taken together, our findings highlight novel temporal and exercise mode-specific blood-based molecular responses to acute exercise, and provide a new resource for the scientific community.

BACKGROUND: Heart failure (HF) and its main subtypes, heart failure with preserved ejection fraction (HFpEF) and heart failure with reduced ejection fraction (HFrEF), impose an enormous health burden on elders. Assessment of the circulating proteome to illuminate pathogenesis could open new opportunities for treatment.

METHODS: We conducted a plasma proteomics screen of incident HF and its subtypes in 2 older population-based cohorts, the CHS (Cardiovascular Health Study) and the AGES-RS (Aging, Gene/Environment Susceptibility-Reykjavik Study). The 2 studies used SomaLogic platforms, with 4404 aptamers in common. Multivariable Cox models were fit to evaluate individual-protein associations with HF, HFpEF, and HFrEF separately in each cohort, and study-specific associations were combined by fixed-effects meta-analysis. Replication was performed in the ARIC (Atherosclerosis Risk in Communities) cohort. Two-sample Mendelian randomization of HF and its subtypes, along with colocalization analysis, was performed to support causal inference.

RESULTS: Among 8599 participants, 1590 experienced incident HF (536 HFpEF, 471 HFrEF). There were 119 proteins associated with HF, 15 proteins with HFpEF, and 11 proteins with HFrEF, at Bonferroni-corrected significance. Among these, 9 have never previously been identified for cardiovascular diseases, and another 61 represent new associations with incident HF or its subtypes. Of these 70 proteins, 55 of the 66 available replicated externally. Mendelian randomization analysis revealed 7 proteins genetically associated with HF at nominal significance; 2 were separately associated with HFpEF, and another 2 with HFrEF. Seven of these 9 proteins (NPDC1 [neural proliferation differentiation and control protein 1], APOF [apolipoprotein F], LMAN2 [lectin, mannose-binding 2], ADIPOQ [adiponectin], CD14 [cluster of differentiation 14], ARHGAP1 [Rho GTPase-activating protein 1], C9 [complement 9]) showed new, possibly causal associations, although we did not detect evidence for colocalization.

CONCLUSIONS: In this large-scale proteomic study involving 3 longitudinal cohorts of older adults, we identified and replicated 55 novel protein markers of HF or its subtypes, and 7 new, possibly causal proteins. These proteins may enhance risk prediction, improve understanding of pathobiology, and help prioritize targets for therapeutic development of these foremost disorders in elders.

Measures from affinity-proteomics platforms often correlate poorly, challenging interpretation of protein associations with genetic variants and phenotypes. Here, we examine 2157 proteins measured on both SomaScan 7k and Olink Explore 3072 across 1930 participants with genetic similarity to European, African, East Asian, and Admixed American ancestry references. Inter-platform correlation coefficients for these 2157 proteins follow a bimodal distribution (median r = 0.30). We evaluate protein measure associations with genetic variants, and find approximately 25-30% of the signals on each platform are likely driven by protein-altering variants. We highlight 80 proteins that correlate differently across ancestry groups likely in part due to differing protein-altering variant frequencies by ancestry. Furthermore, adjustment for protein-altering variants with opposite directions of effect by platform improves inter-platform protein measure correlation and results in more concordant genetic and phenotypic associations. Hence, protein-altering variants need to be accounted for across ancestries to facilitate platform-concordant and accurate protein measurement.