Paired ATAC- and RNA-seq offer insight into the impact of HIV on alveolar macrophages: a pilot study

Abstract

People with HIV remain at greater risk for both infectious and non-infectious pulmonary diseases even after antiretroviral therapy initiation and CD4 cell count recovery. These clinical risks reflect persistent HIV-mediated defects in innate and adaptive immunity, including in the alveolar macrophage, a key innate immune effector in the lungs. In this proof-of-concept pilot study, we leveraged paired RNA-seq and ATAC-seq analyses of human alveolar macrophages obtained with research bronchoscopy from people with and without HIV to highlight the potential for recent methodologic advances to generate novel hypotheses about biological pathways that may contribute to impaired pulmonary immune function in people with HIV. In addition to 35 genes that were differentially expressed in macrophages from people with HIV, gene set enrichment analysis identified six gene sets that were differentially regulated. ATAC-seq analysis revealed 115 genes that were differentially accessible for people with HIV. Data-driven integration of the findings from these complementary, high-throughput techniques using xMWAS identified distinct clusters involving lipoprotein lipase and inflammatory pathways. By bringing together transcriptional and epigenetic data, this analytic approach points to several mechanisms, including previously unreported pathways, that warrant further exploration as potential mediators of the increased risk of pulmonary disease in people with HIV.

Figure 1

xMWAS data integration based on correlations between the RNA-seq (genes shown in green squares) and ATAC-seq data (peaks shown in orange squares). xMWAS data-driven integration identified three distinct clusters, one centered on lipoprotein lipase (LPL) with corresponding accesible peak regions from ATAC-seq and two others involving multiple genes from inflammatory pathways with shared accessibility regions forming an interconnected subnetwork structure. The nodes are color-coded by communites detected by mulilevel community detection algorithm that is based on topology of the network, while the edges represent the high correlations (|ρ|> 0.85, red: positive, blue: negative) between the connected two nodes.