Department of Integrative Biology, University of Wisconsin-Madison, Madison, USA.
Sci Rep. 2022 Jan 7;12(1):108. doi: 10.1038/s41598-021-04020-1.
Depression is a complex mental health disorder that is difficult to study. A wide range of animal models exist and for many of these data on large-scale gene expression patterns in the CNS are available. The goal of this study was to evaluate how well animal models match human depression by evaluating congruence and discordance of large-scale gene expression patterns in the CNS between almost 300 animal models and a portrait of human depression created from male and female datasets. Multiple approaches were used, including a hypergeometric based scoring system that rewards common gene expression patterns (e.g., up-up or down-down in both model and human depression), but penalizes opposing gene expression patterns. RRHO heat maps, Uniform Manifold Approximation Plot (UMAP), and machine learning were used to evaluate matching of models to depression. The top ranked model was a histone deacetylase (HDAC2) conditional knockout in forebrain neurons. Also highly ranked were various models for Alzheimer's, including APPsa knock-in (2nd overall), APP knockout, and an APP/PS1 humanized double mutant. Other top models were the mitochondrial gene HTRA2 knockout (that is lethal in adulthood), a modified acetylcholinesterase, a Huntington's disease model, and the CRTC1 knockout. Over 30 stress related models were evaluated and while some matched highly with depression, others did not. In most of the top models, a consistent dysregulation of MAP kinase pathway was identified and the genes NR4A1, BDNF, ARC, EGR2, and PDE7B were consistently downregulated as in humans with depression. Separate male and female portraits of depression were also evaluated to identify potential sex specific depression matches with models. Individual human depression datasets were also evaluated to allow for comparisons across the same brain regions. Heatmap, UMAP, and machine learning results supported the hypergeometric ranking findings. Together, this study provides new insights into how large-scale gene expression patterns may be similarly dysregulated in some animals models and humans with depression that may provide new avenues for understanding and treating depression.
抑郁症是一种复杂的心理健康障碍，难以研究。目前存在广泛的动物模型，其中许多动物模型具有中枢神经系统（CNS）的大规模基因表达模式数据。本研究的目的是通过评估近 300 种动物模型与男性和女性数据集创建的人类抑郁症画像之间 CNS 中大规模基因表达模式的一致性和差异性，来评估动物模型与人类抑郁症的吻合程度。本研究使用了多种方法，包括基于超几何分布的评分系统，该系统奖励常见的基因表达模式（例如，在模型和人类抑郁症中均为上调-上调或下调-下调），但惩罚相反的基因表达模式。RRHO 热图、一致流形逼近图（UMAP）和机器学习被用于评估模型与抑郁症的匹配程度。排名最高的模型是一种组蛋白去乙酰化酶（HDAC2）条件性敲除的前脑神经元模型。阿尔茨海默病的多种模型（包括 APPsa 敲入模型（总体排名第二）、APP 敲除模型和 APP/PS1 人源化双突变模型）也具有较高的排名。其他排名较高的模型包括线粒体基因 HTRA2 敲除（在成年期致命）、一种改良的乙酰胆碱酯酶、亨廷顿病模型和 CRTC1 敲除。评估了 30 多种与应激相关的模型，虽然其中一些与抑郁症高度匹配，但其他模型则不匹配。在大多数顶级模型中，均鉴定到 MAP 激酶途径的一致失调，并且 NR4A1、BDNF、ARC、EGR2 和 PDE7B 等基因与人类抑郁症一样持续下调。还评估了单独的男性和女性抑郁症画像，以识别与模型的潜在性别特异性抑郁症匹配。还评估了个体的人类抑郁症数据集，以允许在相同的大脑区域进行比较。热图、UMAP 和机器学习结果支持超几何评分结果。总的来说，本研究为一些动物模型和人类抑郁症中大规模基因表达模式可能相似失调提供了新的见解，这可能为理解和治疗抑郁症提供新的途径。
