Lipidomic predictors of heart failure in chronic kidney disease

摘要 慢性肾脏病（CKD）是一个重大的公共卫生问题。心血管疾病（CVD）是主要的 CKD发病率和死亡率的原因，心力衰竭（HF）是主要的心血管（CV）事件。HF 以显著的心肌和全身脂质代谢紊乱为特征。然而时过境迁 关于功能、形态学和临床表现的代谢改变过程不是 建立，也没有脂质代谢紊乱在HF发病机制中的作用进行了研究。因此有 不清楚脂质代谢改变是原因还是结果。我们假设脂质代谢的改变 在CKD患者中，代谢途径可预测代谢事件HF，并且代谢事件HF是连续的， 这一过程始于脂质代谢途径的改变，导致左心室（LV） 配置，并最终表现为临床HF。为了验证这一假设，我们将进行以下研究： 具体目的：1）确定伴有和不伴有糖尿病的CKD患者中发生HF的脂质组学预测因子， 2)确定CKD伴和不伴糖尿病患者左室构型的脂质组学决定因素，以及3） 探索区分CKD伴和不伴糖尿病患者CVD表型的不同脂质网络。 方法：这是一项具有CV事件前瞻性数据的观察性纵向研究。研究将 在慢性肾功能不全队列（CRIC）的参与者中进行，作为训练子集， 密歇根大学的患者作为验证队列。入选标准：基线eGFR>30 mL/min， 入组时200 µL血浆样本的可用性、第1年超声心动图（CRIC）和CVD结局 后续数据。采样：实施入选和排除标准后，来自CRIC的2931例患者 包括341名HF患者，1148名验证队列参与者，包括610名HF患者， 选择538名无HF的受试者。预测因子：基于质谱的游离脂肪酸定量值 脂肪酸（FA）、酰基肉毒碱（AC）和由甘油脂、磷脂和脂肪酸组成的复合脂质。 鞘磷脂结局：目标1和3的主要结局是HF事件。目标2的主要结局是 LV构型的超声心动图评估。分析：对于目标1，分析将包括利用 在基线和结局前采用序贯脂质组学测量的校正考克斯回归模型， 识别HF事件随时间变化的独立脂质预测因子。对于目标2，分析将包括利用率 偏最小二乘判别分析（PLS-DA），随机森林，混合线性模型，惩罚 多项式回归模型，以区分4类LV配置。对于目标3，分析将是 基于差分网络富集分析（DNEA）来识别脂质的差分网络 区分研究结果。预期成果：预计将有一个由FA、AC和 基线复合脂质预测HF事件，并与LV形态差异相关。也是 预期DNEA将揭示受研究结果差异调节的脂质代谢途径。

ABSTRACT Chronic kidney disease (CKD) is a major public health issue. Cardiovascular diseases (CVD)s are the leading causes of morbidity and mortality in CKD, with heart failure (HF) being a major cardiovascular (CV) event. HF is characterized by significant myocardial and systemic lipid metabolic derangements. However, the time course of metabolic alterations in regard to functional, morphological and clinical presentations is not established, nor the role of lipid metabolic derangements in pathogenesis of HF is investigated. Therefore, it is unclear if the lipid metabolic alterations are cause or effect. We hypothesize that alteration in lipid metabolic pathways predicts metabolic incident HF among CKD patients, and that metabolic incident HF is continuum of a process that starts with alterations in lipid metabolic pathways that leads to changes of left ventricular (LV) configuration, and eventually manifests as clinical HF. To test this hypothesis, we will pursue the following specific aims: 1) To identify the lipidomic predictors of incident HF in CKD patients with and without diabetes, 2) To determine the lipidomic determinants of LV configuration in CKD with and without diabetes, and 3) To explore differential lipid networks discriminating CVD phenotypes in CKD patients with and without diabetes. Methods: This is an observational longitudinal study with prospective data on CV events. Study will be performed on participants of the Chronic Renal Insufficiency Cohort (CRIC) as the training subset, and the HF patients at the University of Michigan as the validation cohort. Inclusion criteria: baseline eGFR>30 mL/min, availability of 200 µL of plasma sample at enrollment, echocardiography at year 1 (CRIC), and CVD outcome data at follow up. Sampling: After implementation of inclusion and exclusion criteria 2931 patients from CRIC including 341 with HF, and 1148 participants for validation cohort including 610 patients with HF matched by 538 participants without HF are selected. Predictors: mass spectrometry based quantified values of free fatty acids (FA)s, acylcarnitines (AC)s, and complex lipids consisting of glycerolipids, phospholipids, and sphingomyelins. Outcomes: Primary outcome for aims 1 and 3 is incident HF. Primary outcome for aim 2 is echocardiographic assessment of LV configurations. Analysis: For aim 1, the analysis will include utilization of adjusted Cox regression models with sequential lipidomic measurements at baseline and prior to outcome to identify time-varying independent lipid predictors of incident HF. For aim 2, the analysis will include utilization of partial least square-discriminant analysis (PLS-DA), random-forest, mixed linear models, and penalized multinomial regression models to discriminate 4 categories of LV configurations. For aim 3, the analysis will be based on Differential Network Enrichment Analysis (DNEA) to identify differential network of lipids discriminating the study outcomes. Expected outcomes: It is expected that a differential panel of FAs, ACs, and complex lipids at baseline predict incident HF and differentially correlate with LV morphology. It is also expected that DNEA will disclose lipid metabolic pathways that are differentially regulated by study outcomes.