Very low protein diets lead to reduced food intake and weight loss, linked to inhibition of hypothalamic mTOR signaling, in mice. Very low protein diets lead to reduced food intake and weight loss, linked to inhibition of hypothalamic mTOR signaling, in mice

Purpose: To investigate graded levels of low protien dites effects on hypothalamic gene expression pattern, RNA-seq was performed on hypothalamus of mice treated with different levels of low protein and normal protein under both 60% fat and 20% fat conditions. Methods: There were 8 diets in total, the fat content of first 4 diets were fixed at 60% fat level by energy and protein was varied from 1% to 20% (1%, 2.5%, 5% and 20% respectively) by energy (D14121903, D14121904, D14071601,D14071604), another 4 kinds of diets contained the same graded levels of protein content as the first 4 diets but fat contents were fixed at 20% by energy (D14121905, D14121906, D14071607, D14071610). The remaining energy was compensated of carbohydrate which included corn starch and maltodextrose. Casein was used as the protein source in all diets. To mimic the typical western diet a mix of cocoa butter, menhaden oil, sunflower oil, palm oil and coconut oil was used as the fat source and was designed to generate a 47.5: 36.8: 15.8 proportion of saturated, mono-unsaturated and polyunsaturated fats and 14.7: 1 proportion of n-6 and n-3 fatty acids, the proportions of different fat compositions didn’t change under two different 60% and 20% fat conditions. Sucrose and cellulose were fixed at 5% level by energy and standard vitamin and mineral mix were also added to all diets as well (Hu et al., 2018). The two series were isocaloric within each series. Diets can be ordered direct from research diets (https://researchdiets.com) using the diet codes provided. The hypothalamus of 6 individuals were collected for each diet group, 3 of which were pooled together as one sample, resulting in each group having 2 pooled samples of 3 hypothalamus. Hypothalamic RNA was extracted using Trizol method described in star methods in paper and then sent to Beijing Genomic Institute (BGI) for RNA sequencing. All RNA sequencing was measured by using the BGI-seq 500 sequencer. To determine the sequencing quality of each sample, FASTQ raw data files were measured by using fastQC (www.bioinformatics.bbsrc.ac.uk/projects/fastqc/) quality control tool, after confirmed sequencing quality of each sample, raw reads were aligned to the Mus musculus reference genome (GRCm38) using HISAT2-2.1.0 (Kim et al., 2015; Pertea et al., 2016) and Samtools-1.2 modules (Li et al., 2009), and then featureCounts (Liao et al., 2014) tool in Subread-5.0 (Liao et al., 2013) was used to get counts for each sample from BAM files which was obtained from alignment step. To exclude the low counts, only genes with counts per million (CPM) value > 1 were included in the further analysis. Counts were normalized by the trimmed means of M values (TMM normalization) (Robinson and Oshlack, 2010) method in the edgeR (Anders et al., 2013; Lund et al., 2012; McCarthy et al., 2012; Robinson et al., 2010; Robinson and Oshlack, 2010; Robinson and Smyth, 2007) package. To investigate different dietary protein contents under fixed two levels fat content effects on gene expression levels Generalized Linear Modelling (GLM) function in R-3.5.3 edgeR package was used, The GLM model used in this study were: ~P+F+P:F, which means regression against protein (P) and fat contents (F) of diets plus their interaction (P:F). If the interaction effect was not significant (p > 0.05), the interaction was not included in the analysis and a revised model (~P+F) was utilized (Hu et al., 2018). To explore significantly correlated genes with dietary protein content respectively under 60% fat and 20% fat conditions, Pearson correlation analysis was performed for normalized log counts of all genes by using Pearson correlation method in R-3.5.3. Then selected the significantly correlated genes with dietary protein content (GLM: p 0.05), the interaction was not included in the analysis and a revised model (~P+F) was utilized (Hu et al., 2018). To explore significantly correlated genes with dietary protein content respectively under 60% fat and 20% fat conditions, Pearson correlation analysis was did for normalized log counts of all genes by using Pearson correlation method

研究目的：为探究不同梯度低蛋白饮食对小鼠下丘脑基因表达模式的影响，本研究对分别在60%和20%脂肪供能条件下、经不同水平低蛋白及正常蛋白处理的小鼠下丘脑开展了RNA测序（RNA-seq）。
实验方法：本研究共设置8种饲料，其中前4种饲料的脂肪供能比固定为60%，蛋白质供能比分别为1%、2.5%、5%和20%（对应饲料编码依次为D14121903、D14121904、D14071601、D14071604）；后4种饲料与前4种的蛋白质梯度一致，但脂肪供能比固定为20%（对应编码依次为D14121905、D14121906、D14071607、D14071610）。剩余能量由包含玉米淀粉与麦芽糊精的碳水化合物补充。所有饲料均以酪蛋白作为蛋白质来源。为模拟典型西式饮食，脂肪来源由可可脂、鲱鱼油、葵花籽油、棕榈油及椰子油复配而成，其饱和脂肪酸、单不饱和脂肪酸、多不饱和脂肪酸的比例为47.5:36.8:15.8，n-6与n-3脂肪酸的比例为14.7:1，且60%与20%脂肪供能条件下的脂肪组成比例保持一致。蔗糖与纤维素的能量占比均固定为5%，所有饲料均添加标准维生素与矿物质混合物（Hu等，2018）。两个饲料系列内部均保持等热量。饲料可通过研究饲料官网（https://researchdiets.com）使用对应编码直接订购。
每个饲料组收集6只小鼠的下丘脑组织，每3只组织混合为一个样本，因此每个饲料组包含2个混合样本，每个样本由3个下丘脑组织构成。采用Trizol法提取下丘脑总RNA，具体操作参照论文中STAR方法部分所述步骤，随后将RNA送至北京基因组研究所（BGI）进行RNA测序，测序平台为BGI-seq 500测序仪。
使用fastQC质控工具（www.bioinformatics.bbsrc.ac.uk/projects/fastqc/）对FASTQ原始数据文件进行测序质量评估。确认样本测序质量合格后，使用HISAT2-2.1.0（Kim等，2015；Pertea等，2016）与Samtools-1.2模块（Li等，2009）将原始reads比对至小鼠（Mus musculus）参考基因组GRCm38。随后使用Subread-5.0中的featureCounts工具（Liao等，2014），从比对得到的BAM文件中获取每个样本的基因计数。为排除低表达基因，仅保留每百万reads计数（CPM）>1的基因用于后续分析。使用edgeR包中的截尾均值标准化法（TMM标准化，trimmed means of M values）对计数数据进行标准化。
为探究固定脂肪水平下不同膳食蛋白质含量对基因表达水平的影响，本研究使用R-3.5.3的edgeR包中的广义线性模型（Generalized Linear Modelling, GLM）函数，采用的GLM模型为：~P+F+P:F，即针对饲料的蛋白质含量（P）、脂肪含量（F）及其交互作用（P:F）进行回归分析。若交互作用不显著（p>0.05），则移除交互项，采用修正后的模型~P+F进行分析（Hu等，2018）。
为分别探究60%和20%脂肪供能条件下与膳食蛋白质含量显著相关的基因，本研究使用R-3.5.3中的Pearson相关分析法，对所有基因的标准化log计数进行Pearson相关分析。筛选出与膳食蛋白质含量显著相关的基因（GLM分析p 0.05），若交互作用不显著则采用修正后的模型~P+F进行分析（Hu等，2018）。为分别探究60%和20%脂肪供能条件下与膳食蛋白质含量显著相关的基因，本研究使用R-3.5.3中的Pearson相关分析法，对所有基因的标准化log计数开展Pearson相关分析。