Transcriptomic profiling of mock-infected primary CD4+ T cells and a model of HIV latency treated with suberoylanilide hydroxamic acid (SAHA) and Romidepsin (RMD). Transcriptomic profiling of mock-infected primary CD4+ T cells and a model of HIV latency treated with suberoylanilide hydroxamic acid (SAHA) and Romidepsin (RMD)

Purpose: The goal of this study is to identify host genes whose expression is perturbed in primary CD4+ T cells by histone deacetylase (HDAC) inhibitors (HDACi) SAHA and RMD, which have different potencies and specificities for various HDACs. The study aims to evaluate the effects of SAHA and RMD that may promote or inhibit reactivation of HIV provirus out of latency. Methods: Peripheral blood mononuclear cells were collected from 4 HIV-seronegative donors. CD4+ T cells were isolated and utilized to generate an in vitro model of latent HIV infection (model developed in the Spina laboratory and previously described in Spina et al., 2013). Mock-infected cells were cultured in parallel to evaluate effects of SAHA and RMD that may be dependent on the exposure of cells to virus. Following generation of the model, cells were treated with SAHA, RMD or their solvent dimethyl sulfoxide (DMSO) for 24 hours. Mock-infected cells were treated in parallel. The experiment had 4 biological replicates, 6 conditions for each, for a total of 24 samples. ERCC spikes (Thermo Fisher Scientific, Inc.) were added to cell lysates based on cell number in each sample (10 ul of 1:800 dilution per million cells). Mix 1 was used for DMSO- and mix 2 for SAHA- and RMD-treated cells. After all samples were collected, RNA was extracted and subjected to deep sequencing by Expression Analysis, Inc. Sequence reads that passed quality filters were mapped using Tophat (human genome) or Bowtie (ERCC spikes and HIV) and counted using HTSeq. ERCC spikes with the same concentration in mixes 1 and 2 were utilized to remove unwanted technical variation. Any human gene which did not achieve at least 1 count per million reads in at least 4 samples or any ERCC that did not achieve at least 5 reads in at least 4 samples was discarded. Differential gene expression analysis was performed using library EdgeR in Bioconductor R. National Center for Biotechnology Information (NCBI) HIV-1 Human Interaction Database was then searched for genes that have been implicated in controlling HIV latency. EdgeR output was used to extract expression information of the genes of interest from the NCBI database to identify genes implicated in HIV latency that were modulated by SAHA and RMD. The resulting lists were manually curated to verify relevance to HIV latency, using the Description column of the NCBI database, as well as available PubMed references. Results: Using a custom built data analysis pipeline, ~100 million reads per sample were mapped to the human genome (build hg38). After applying filtering criteria, 16058 human transcripts, 19 ERCC spikes transcripts, and HIV NL4-3 transcripts were identified with the Tophat/Bowtie and HTSeq workflow. Differential expression analysis was performed between SAHA or RMD-treated and DMSO-treated cells. In addition, differential modulation of gene expression by SAHA and RMD in the model of HIV latency and mock-infected cells was assessed using EdgeR. In mock-infected cells, SAHA upregulated 3,971 genes and downregulated 2,940 genes; RMD upregulated 5,068 genes and downregulated 4,050 genes. In the model of HIV latency, SAHA upregulated 3,498 genes and downregulated 2,904 genes; RMD upregulated 5,116 genes and downregulated 4,053 genes (FDR < 0.05). SAHA modulated 6, and RMD 11 genes differentially between mock-infected cells and the model of HIV latency. Following search of the NCBI HIV-1 Human Interaction Database, 27 genes upregulated and 29 downregulated in common between SAHA and RMD were found to be relevant to regulation of HIV latency; 31 were up- and 32 downregulated by RMD only; and 6 were up- and 2 were downregulated by SAHA only. Conclusions: This study demonstrates that SAHA and RMD, which have different potencies and specificities for HDACs, modulate a set of overlapping genes implicated in regulation of HIV latency. Some of these genes may be explored as additional host targets for improving the outcomes of “shock and kill” strategies. Overall design: Transcriptomic profiling of the in vitro model of HIV latency and mock-infected cells treated with SAHA, RMD or the solvent DMSO (N=4 donors) by deep sequencing at Expression Analysis, Inc.

本研究旨在探究经组蛋白去乙酰化酶（histone deacetylase, HDAC）抑制剂SAHA与RMD处理后，原代CD4+T细胞中表达发生扰动的宿主基因——这两种抑制剂对不同HDAC亚型具有不同的效力与特异性。同时，本研究旨在评估SAHA与RMD可能发挥的作用，即促进或抑制潜伏状态HIV前病毒的激活。

方法：本研究从4名HIV血清阴性供者体内采集外周血单个核细胞。分离得到CD4+T细胞后，用于构建体外HIV潜伏感染模型（该模型由Spina实验室开发，此前已在Spina等人2013年的研究中发表）。同时平行培养模拟感染细胞，以评估SAHA与RMD的效应是否依赖于细胞接触病毒。模型构建完成后，分别用SAHA、RMD及其溶剂二甲基亚砜（dimethyl sulfoxide, DMSO）处理细胞24小时，同步对模拟感染细胞进行相同处理。本实验设置4次生物学重复，每个供者对应6种处理条件，总计24个样本。根据每个样本中的细胞数量，向细胞裂解液中加入ERCC外参（Thermo Fisher Scientific公司）：每百万细胞添加10μL 1:800稀释液。其中，DMSO处理组使用Mix 1，SAHA与RMD处理组使用Mix 2。所有样本收集完成后，提取RNA并由Expression Analysis公司进行深度测序。
对通过质量过滤的序列读段，分别采用Tophat（比对人类基因组）或Bowtie（比对ERCC外参与HIV基因组）进行比对，并通过HTSeq进行计数。利用两组混合液中浓度一致的ERCC外参，去除不必要的技术变异。对于人类基因，若在至少4个样本中均未达到每百万读段1个计数，或ERCC外参若在至少4个样本中均未达到5个读段，则将其剔除。使用Bioconductor R中的EdgeR库进行差异基因表达分析。随后，检索美国国家生物技术信息中心（National Center for Biotechnology Information, NCBI）的HIV-1人类相互作用数据库，筛选出已被证实与HIV潜伏调控相关的基因。利用EdgeR的输出结果，从NCBI数据库中提取目标基因的表达信息，以鉴定经SAHA或RMD调控的、与HIV潜伏相关的基因。最终通过NCBI数据库的描述栏以及已发表的PubMed文献，对得到的基因列表进行人工手动注释，以验证其与HIV潜伏的相关性。

结果：通过自建的数据分析流程，每个样本约得到1亿条可比对至人类基因组（hg38版本）的读段。应用过滤标准后，通过Tophat/Bowtie与HTSeq流程共鉴定到16058个人类转录本、19条ERCC外转录本以及HIV NL4-3转录本。分别对SAHA或RMD处理组与DMSO处理组进行差异表达分析。此外，通过EdgeR评估了SAHA与RMD在HIV潜伏感染模型与模拟感染细胞中对基因表达的差异调控效应。在模拟感染细胞中，SAHA上调3971个基因、下调2940个基因；RMD上调5068个基因、下调4050个基因。在HIV潜伏感染模型中，SAHA上调3498个基因、下调2904个基因；RMD上调5116个基因、下调4053个基因（错误发现率false discovery rate, FDR < 0.05）。SAHA与RMD在模拟感染细胞与HIV潜伏感染模型之间存在6个差异调控基因，RMD则存在11个。检索NCBI HIV-1人类相互作用数据库后发现，SAHA与RMD共同上调的基因有27个、共同下调的有29个，均与HIV潜伏调控相关；仅RMD调控的基因中，上调31个、下调32个；仅SAHA调控的基因中，上调6个、下调2个。

结论：本研究证实，对不同HDAC亚型具有不同效力与特异性的SAHA与RMD，可调控一系列与HIV潜伏调控相关的重叠基因。其中部分基因可作为潜在宿主靶点，用于优化“激活并清除（shock and kill）”HIV治疗策略的效果。

整体实验设计：由Expression Analysis公司通过深度测序，对经SAHA、RMD或溶剂DMSO处理的体外HIV潜伏感染模型与模拟感染细胞进行转录组分析，共纳入4名供者的样本（N=4）。