Deep RNA Sequencing Uncovers a Repertoire of Human Macrophage Long Intergenic Noncoding RNAs Modulated by Macrophage Activation and Associated With Cardiometabolic Diseases

Figures

• 

  Figure 1. Global discovery of human macrophage lincRNAs. Polyadenylated RNA sequencing data generated from M0 HMDM and M1‐ or M2‐activated HMDM were analyzed by the bioinformatics pipeline outlined for the annotation of known lincRNAs (the Cabili set and GENCODE V19) and newly annotated lincRNAs (from de novo assembly). FPKM indicates fragments per kilobase of transcript per million fragments mapped; HMDM, human peripheral blood mononuclear cell–derived macrophages; IFN‐γ, interferon γ; IL‐4, interleukin‐4; lincRNA, long intergenic noncoding RNA; LPS, lipopolysaccharide; PBMC, peripheral blood mononuclear cells.

• 

  Figure 2. Expression, synteny, and conservation of macrophage lincRNAs. A, Expression measured as FPKM for gene categories. The lincRNAs had overall lower expression than coding genes. B, Most lincRNAs were nonsyntenic or were syntenic but not annotated as expressed in mouse. Compared with known lincRNAs, newly annotated lincRNAs showed lower percentages of synteny and annotation in mouse. The lincRNAs that were syntenic and annotated in mouse showed higher expression than other lincRNAs. C, The lincRNA sequence conservation between human and mouse in syntenic regions was examined using BLASTN (see Methods). The lincRNAs demonstrated poor sequence conservation in the syntenic regions, and newly annotated lincRNAs were less conserved than known ones. FPKM indicates fragments per kilobase of transcript per million fragments mapped; lincRNA, long intergenic noncoding RNA.

• 

  Figure 3. Tissue enrichment of macrophage lincRNAs. A, Relative abundance of mRNA and lincRNA genes (rows) expressed in M0 HMDM across 16 human tissue types from the Human BodyMap demonstrate that macrophage lincRNAs were more tissue‐specific than mRNAs. Color intensity represents the fractional gene‐level expression across all tissues examined, as described.25 B, Certain lincRNAs were specifically enriched only in M1 or M2 HMDM. C, Expression of M0 HMDM‐enriched lincRNAs were higher than that of nonenriched lincRNAs. D, Most M0 HMDM‐enriched lincRNAs harbor PU.1 and C/EBPβ binding sites in ±2‐kb intervals centered on the lincRNA TSS and gene body, and this percentage was higher than that in nonenriched M0 HMDM lincRNAs. HMDM indicates human peripheral blood mononuclear cell–derived macrophages; lincRNA, long intergenic noncoding RNA; TSS, transcription start site.

• 

  Figure 4. Macrophage lincRNAs have distinct enhancer‐driven transcript signatures. A, Histone modification of H3K4me3 (promoter mark) and H3K4me1 (enhancer mark) across the ±1.5‐kb region around the transcription start site and the H3K4me1/H3K4me3 ratios of protein‐coding genes and lincRNAs show distinct enhancer‐driven transcript signatures in macrophage lincRNAs vs protein‐coding genes. Profiles were sorted based on mean H3K4me3 intensity. B, We used H3K4me1/H3K4me3 ratios of >1.2 and <0.8 to define enhancer and promoter states,39 respectively. The correlation of fold change in expression during M1 activation (M0 vs M1) of “elincRNAs” and their closest protein‐coding genes was significantly stronger than that for “plincRNAs.” elincRNA indicates enhancer‐associated lincRNA; H3K4me1, histone H3 lysine 4 monomethylation; H3K4me3, histone H3 lysine 4 trimethylation; lincRNA, long intergenic noncoding RNA; plincRNA, promoter‐associated lincRNA.

• 

  Figure 5. Macrophage activation induced widespread changes in lincRNA profile. Venn diagrams and volcano plot tabulating numbers of lincRNAs common or differentially expressed (fold change >2, FDR‐adjusted P<0.01) between M0 HMDM vs M1 HMDM (A) and M0 HMDM vs M2 HMDM (B). C and D, Ten lincRNAs were selected according to prioritization criteria described in text. The relative expression levels of these lincRNAs in M0, M1, and M2 HMDM were determined using qRT‐PCR and presented graphically, and the respective FPKM levels determined by RNA sequencing were listed showing consistent patterns with qRT‐PCR results. The ΔCt represents the mean cycle threshold for lincRNAs relative to human ACTB mRNA as the reference in each sample. Data were shown as mean±SD, # P<0.05 vs M0 HMDM. FDR indicates false discovery rate; FPKM, fragments per kilobase of transcript per million fragments mapped; lincRNA, long intergenic noncoding RNA; qRT‐PCR, quantitative reverse transcriptase–polymerase chain reaction.

• 

  Figure 6. MacORIS harbors variant associated with central obesity and acts as a repressor of IFN‐γ signaling. Genome browser view of MacORIS (annotated as RP11‐472N13.3) locus on chr10 (A and B) and linkage disequilibrium plot in European ancestry (C) shows that MacORIS overlaps SNP rs7081678, which is associated with WHR adjusted for BMI. MacORIS is abundant in M0 HMDM and monocytes but found at low levels in human adipose tissue and stromal vascular fraction–differentiated adipocytes (D, by RNA‐seq) and T cells (E, by qRT‐PCR). (F) Genome browser view of RNA‐seq tracks demonstrates lower expression level of MacORIS in M1 HMDM vs M0 HMDM, a consistent pattern also seen in M1 IPSDM vs M0 IPSDM. MacORIS has abundant PU.1 and C/EBPβ binding, both at transcription start site and gene body, in HMDM (F), but there was no annotation in mouse GRC38/mm10 and no expression in published RNA‐seq of murine bone marrow–derived macrophages (G). H, MacORIS is located primarily in cytoplasm (I), and qRT‐PCR validation confirmed that the expression of MacORIS was suppressed in M1 HMDM (n=8 subjects) and M1 THP1Φ (n=3 experiments). Because MacORIS was suppressed 6 h after M1 stimulation (J), the effects of…

• 

  Figure 7. IPSDM resembles the lincRNA expression and activation profiles of HMDM. A, Schematic figure of HMDM and IPSDM differentiation and activation. B, Multidimensional scaling plot confirmed distinct lincRNA transcriptome profile of iPSC from HMDM and IPSDM. M1 activation profoundly influences the lincRNA expression profile, whereas M2 activation resulted in more subtle expression changes in both HMDM and IPSDM. C, iPSC‐to‐IPSDM differentiation induced profound lincRNA expression changes, with 160 higher in iPSC and 153 higher in IPSDM (n=3 subjects). D, LincRNAs upregulated during iPSC‐to‐IPSDM differentiation were more abundantly expressed and demonstrated enriched PU.1 and C/EBPβ binding (E). F, Venn diagram shows remarkable overlap of lincRNA expression between M0 HMDM and M0 IPSDM (n=3 subjects). A small number (n=42) of lincRNAs were differentially expressed between HMDM and IPSDM with 25 higher in M0 HMDM and 17 higher in M0 IPSDM. G and H, IPSDM resembled lincRNA profiles of HMDM for both M1 and M2 activation. FDR, false discovery rate; FPKM, fragments per kilobase of transcript per million fragments mapped; HMDM, human peripheral blood mononuclear cell–derived macrophages;…